PLoS Genetics

Schons-Fonseca, Luciane;Lazova, Milena D.;Smith, Janet L.;Anderson, Mary E.;Grossman, Alan D.

doi: 10.1371/journal.pgen.1010564pmid: 36574412

Introduction DNA replication is essential for all living organisms, and a variety of events can perturb replication, including DNA damage and replication arrest due to “roadblocks” such as DNA-binding proteins or transcription. In any of these cases, some or all the components of the replication machinery (the replisome) may disassemble from the complex, leading to the collapse of the replication fork [1–3]. If a subsequent replisome reaches unrepaired lesions, nicks, or collapsed forks, it can turn these defects into double-strand breaks, which, if not repaired, are lethal. Proper cell growth and division following replication arrest depend on the restart of replication. A wide range of bacteria use a similar mechanism for replication restart: the DNA is processed to enable the formation of a structure resembling a replication fork, followed by reassembly of the replisome (reviewed in [4]). This processing typically involves homologous recombination mediated by RecA [5, 6]. To initiate homologous recombination, helicases and nucleases process DNA at the arrested fork to generate single-stranded DNA (ssDNA), in a process called end-resection [7, 8]. Bacillus subtilis has two nucleases capable of performing end-resection, RecJ and AddAB. In response to DNA damage, the exonuclease activity of RecJ can expand ssDNA gaps by several kilobases [7, 9]. In contrast, the helicase-nuclease complex AddAB, a functional homolog of RecBC in Escherichia coli, binds blunt or near blunt DNA ends. AddAB degrades both strands until reaching a Chi site [10] where it switches to degrade only a single DNA strand in the 5’ to 3’ direction [8]. Both the RecJ and AddAB end-resection pathways in B. subtilis lead to loading of RecA onto the DNA by the loader protein RecO [11]. RecA forms nucleofilaments that can search and invade complementary double-stranded DNA, creating cross-shaped DNA structures known as Holliday junctions. Finally, Holliday junctions are processed into a substrate for the helicase and replication restart protein PriA. PriA functions downstream of AddAB and RecJ, binds to stalled replication forks and promotes assembly of the replicative helicase and primosome enabling replication restart (reviewed in [12, 13]). Most of what is known about the bacterial response to replication arrest comes from experiments using DNA damage to disrupt DNA replication. Repair of damaged DNA and replication restart are both required for optimal survival, making it difficult to differentiate between processes related to the damage itself or the arrest of replication forks. We were interested in identifying genes that influence the ability of cells to survive replication arrest, without the confounding effects of chemical damage (e.g., pyrimidine dimers, cross-links) to DNA. To cause replication fork arrest independent of DNA damage, we used HPUra (6-(p-hydroxyphenylazo)-uracil), an azopyrimidine that reversibly binds to and inhibits the catalytic subunit of DNA polymerase III (PolC) in B. subtilis [14]. We used Tn-seq (transposon insertion mutagenesis with massively parallel sequencing) to identify candidate genes that affect cell survival following the arrest of replication elongation caused by treatment with HPUra. We found that many processes appeared to be important for survival, including: regulation of oxidative homeostasis, cell wall stability, and DNA repair pathways. Interestingly, we found that one of the DNA end-resection pathways that is important for surviving DNA damage was detrimental to surviving replication arrest. Loss of recJ, recO, and recF was beneficial to cell survival, indicating that the functional genes were detrimental. In contrast, functional addA, recN, and recU were beneficial for cell survival following replication fork arrest. Our findings indicate that although AddAB and RecJ both carry out end-resection, they commit replication fork repair to distinct pathways. We found that these pathways led to different amounts of RecA that formed filaments on the ssDNA. Following replication fork arrest, RecJ or the recombinase loader and regulator (recO and recF, respectively) led to excessive RecA loading, accumulation of unrepaired forks, and ultimately an increase in cell death. Our data indicate that B. subtilis activates at least two pathways to repair arrested replication forks due to replisome malfunction and that one pathway (AddAB) is beneficial and the other (RecJ) is detrimental to cell survival. Results Identification of genes involved in cell survival following replication arrest Using Tn-seq, we identified non-essential genes that significantly affected the ability of cells to survive and recover from replication arrest caused by treatment with HPUra. We used a previously described library of ~1.7 x 105 unique transposon insertions of a modified version of the magellan6 transposon, magellan6x, distributed throughout the B. subtilis genome [15]. This library of cells was split in two and treated or not with HPUra for one hour. The cells were then removed from HPUra, allowing replication and cell growth to resume. Aliquots of cells were harvested 2, 3, and 4 hours later, and the location and relative number of transposon insertions in a given chromosomal site were determined by high-throughput sequencing. In a population of random transposon insertions, mutations that lead to decreased survival would be underrepresented, and the ratio between the frequency of insertions in treated and untreated samples would be <1. Conversely, insertion mutations that improved survival would be overrepresented, resulting in frequency ratios >1. The vast majority of genes had a similar number of insertions (frequency ratios of ~1) with or without HPUra, indicating that these are neutral (S1 Table). Thirty-five genes were classified as affecting survival and recovery from HPUra (see Material and Methods for analysis and criteria), and 13 of these were validated using targeted gene disruptions (Table 1). Several cellular processes appeared to affect survival, including: (i) iron-sulfur and oxidative homeostasis (rex, nadABC, defB, and nifS), likely related to the induction of oxidative stress and genes regulating NAD/NADH following treatment with HPUra [16]; (ii) cell wall stability (ltaS, rseP, oppAB, ydiL, cwlO and ponA); (iii) induction of lysogenic phage (yoyI, yonH from the SPß prophage and xpf, xkdBC from PBSX); and (iv) DNA repair and recombination. Below, we focus on several of the DNA recombination and repair genes (Fig 1A). We also validated seven of the other genes identified in the screen using targeted gene disruptions (Table 1), but did not further evaluate them as part of this study. The number of insertion sites and read frequencies for all open reading frames are presented (S1 Table). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Pathways and genes involved in cell survival after arrest of DNA replication. A. General view of the roles of the DNA repair and recombination genes identified in the Tn-seq screen. The dashed arrows represent multistep pathways. B. Percent survival of wild type (strain JMA222) and recU (GP891), polA (LSF41), recN (LSF298), addAB (LSF253) and recJ (LSF200) null mutants after 3 h of replication arrest caused by HPUra treatment. Percent survival is relative to cell viability at the time of arrest. All mutants were statistically different from wild type, as assessed by two-sample t-tests (P < 0.025). C. Effects of recJ and addAB null mutations on cell survival following replication arrest due to treatment with the DNA damaging agent mitomycin-C (MMC) or inactivation of the replicative helicase (DnaC). Left panel: survival of wild-type (JMA222), ΔaddAB (LSF253) and ΔrecJ (LSF200) cells after 3 h of treatment with 0.33 μg/ml of MMC. Right panel: wild-type, ΔrecJ and ΔaddAB cells carrying a dnaCts allele (LSF176, LSF274, and LSF326) were grown at permissive temperature (30°C) to early exponential phase and shifted to non-permissive temperature (49°C) for 4 h. Percent survival is relative to the population before treatment (left) or the temperature shift (right). The deletion mutants in both conditions were statistically different from wild type, as assessed by two-sample t-tests (P < 0.025). Throughout this work, error bars correspond to one standard deviation of the mean and each point is from an independent culture (biological replicates). https://doi.org/10.1371/journal.pgen.1010564.g001 Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. Genes identified in Tn-seq analysis of cells surviving replication arrest in the absence of DNA damage. https://doi.org/10.1371/journal.pgen.1010564.t001 Effects of null mutations in genes involved in DNA repair and recombination on survival after replication arrest In order to confirm the importance of the DNA repair and recombination genes identified in the Tn-seq screen, we made null mutations in six of the loci identified (polA, recU, recN, recJ, walJ, and addAB; note that addB and addA comprise an operon and for all experiments except where indicated, both genes were deleted) and tested their effects on survival after HPUra-mediated replication fork arrest. We treated wild-type and mutant cells with HPUra for 3 h and then removed HPUra and measured the number of viable cells remaining. Approximately 4.5% of wild-type cells survived this prolonged replication arrest (Fig 1B). In contrast, only 0.5–1% of polA, recU, recN, and addAB mutant cells survived. In contrast to the detrimental effects of the loss of polA, recU, recN, and addAB on surviving replication fork arrest, we found that loss of recJ promoted survival. In the recJ null mutant, approximately 45% of cells remained viable after 3 hr of replication arrest (Fig 1B). These results indicate that normally RecJ is detrimental to survival of replication arrest in the absence of DNA lesions. Together, these results confirm the initial Tn-seq data that indicated that these genes are important for promoting survival following replication fork arrest in the absence of DNA damage. Comparison of recJ and addAB mutants in response to replication arrest Both RecJ and AddAB are involved in end-resection (Fig 1A), and both improve survival after treatment with DNA-damaging agents [17–20]. Therefore, it was somewhat surprising that a recJ null mutant had increased survival in response to replication fork arrest in the absence of externally induced chemical damage to DNA. We confirmed that ΔrecJ and ΔaddAB mutants were more sensitive than wild-type cells to treatment with the DNA damaging agent mitomycin C (MMC). Approximately 1.5% of wild-type cells survived treatment with MMC (0.33 μg/ml) for 3 h (Fig 1C). In contrast, addAB and recJ mutants had 7 and 4-fold lower survival, respectively. To verify that the effects of recJ and addAB on survival and recovery from HPUra were due to replication arrest, rather than some other possible effect, we tested their survival under conditions in which replication was arrested with a temperature-sensitive allele of the replicative DNA helicase (dnaCts) [21]. Inactivation of the replicative helicase mirrored the phenotypes observed during HPUra-mediated replication arrest (Fig 1C). We grew dnaCts (LSF176), dnaCts ΔaddAB (LSF326), and dnaCts ΔrecJ (LSF274) strains at a permissive temperature (30°C). Cells were then shifted to non-permissive temperature (49°C) to arrest replication elongation. After 4 h at 49°C, we measured the number of viable cells from each of the three strains. In the otherwise wild-type cells, prolonged replication arrest at the non-permissive temperature caused a decrease in cell viability to about 20% of the original population (Fig 1C). The loss of addAB exacerbated this effect, and only 4% of cells were viable. In contrast, the loss of recJ largely protected the cells from the outcome of replication arrest due to inactivation of the replicative helicase, and 65% of cells survived (Fig 1C). Together, our results indicate that recJ is detrimental to surviving replication fork arrest in the absence of DNA lesions, and confirm previous work [17, 19] demonstrating that recJ contributes to survival following chemical damage to DNA. Loss of RecA loader RecO or regulator RecF protects ΔaddAB cells from arrest-induced death End-resection by AddAB or RecJ produces the substrate for interaction with RecA. RecO is the main protein responsible for loading RecA onto the ssDNA substrate (Fig 1A), and RecF stabilizes the resulting RecA nucleofilaments [11]. Insertions in recO and recF were not recovered in our initial Tn-seq screen because they were depleted from the initial library prior to treatment with HPUra (S1 Table). We suspect that this depletion was because cells were not kept in the dark during library preparation and these mutants are probably sensitive to ambient UV light. To test if recO and recF contribute to survival following replication arrest, we constructed recO and recF null mutants. We found that deletion of either recO or recF significantly increased cell survival following replication fork arrest (HPUra treatment) in comparison to wild type (Fig 2A). Both deletions suppressed the sensitivity of ΔaddAB mutants to HPUra (Fig 2A, see addAB recO and addAB recF), indicating that loading RecA was at least partially responsible for decreased survival of the ΔaddAB mutant. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. RecA loading is detrimental to survival during HPUra-induced replication arrest. A. Survival after HPUra treatment is greater in the absence of RecA loader and regulator (RecO, RecF). Fold change in survival after 3 h of arrest for wild type (JMA222), ΔrecO (LSF814), ΔrecF (LSF270), ΔaddAB (LSF253), ΔaddAB ΔrecO (LSF817), ΔaddAB ΔrecF (LSF709), ΔrecJ (LSF200), ΔrecJ ΔrecO (LSF815), ΔrecJ ΔrecF (LSF282). Significant difference in means (two-sample T-test): * P < 0.05; ** P < 0.01. B. RecA activity is increased in the absence of addAB and decreased in the absence of recJ. Transcription from the promoter of the SOS-inducible gene yneA is derepressed when RecA is active and serves as an indicator of RecA activity. Cultures of wild-type, ΔrecJ, and ΔaddAB strains containing amyE::PyneA-lacZ (LSF633, LSF634, and LSF635, respectively) were treated or not with HPUra. After 30 min, aliquots were collected for measuring ß-galactosidase activity. Results from three independent experiments are presented. https://doi.org/10.1371/journal.pgen.1010564.g002 In contrast to effects on the addAB mutant, deletion of recF had little if any effect on the sensitivity of the ΔrecJ mutant to HPUra, indicating that RecF and RecJ likely affect the same process–formation of stable RecA filaments (Fig 2A). The SOS response is reduced in the absence of recJ and increased in the absence of addAB We postulated that if the increased survival of ΔrecJ is a consequence of lower stability of RecA filaments, then ΔrecJ might have lower RecA activity. Once assembled onto ssDNA, RecA induces autocleavage of the SOS repressor LexA, allowing expression of many genes involved in the response to genotoxic stress [22–24]. Consequently, expression of genes repressed by LexA can be used as a proxy for the levels of activated RecA. yneA encodes an inhibitor of cell division [25], and is repressed by LexA and de-repressed in response to DNA damage and replication fork arrest [26]. We used a fusion of the promoter region of yneA to lacZ (PyneA-lacZ) and measured expression in the mutants in the presence and absence of HPUra. For wild-type cells in defined minimal medium, ß-galactosidase activity from the PyneA-lacZ fusion increased approximately 20-fold after 30 min of replication arrest (Fig 2B). In the ΔrecJ mutant, there was a basal level of activity similar to that in wild type, but only an approximately 6-fold increase after addition of HPUra, indicating that recJ plays an important role in activation of RecA. The absence of addAB led to very high PyneA-lacZ activity during exponential growth (Fig 2B), indicating that there is an abundance of active RecA in these cells. Lysogenic phages are not responsible for the effects of recJ and addAB on survival after replication arrest It seemed plausible that the impact of recJ and addAB on survival following replication stress could be due to different levels of phage induction, as both mutations affect the activity of RecA. The B. subtilis strains used in the experiments described above have two resident lysogenic phages, both of which undergo RecA-dependent activation after genotoxic stress: SPß and the defective phage PBSX [16, 26–28]. Both encode toxins and autolysins which, when expressed, can cause cell death. We found that insertions in two genes from SPß (yoyI, yonH) and three genes involved in the early steps of PBSX induction (xpf, xkdBC) were overrepresented in the screen, indicating that loss of these genes helped cells survive arrest of replication elongation (Table 1). We measured survival of otherwise wild-type cells missing either PBSX, SPß, or both following replication arrest with HPUra. In these experiments, the survival of wild-type cells (containing both lysogenic phages) was 3.5% (Fig 3A). Loss of either phage increased survival to approximately 7%, and loss of both phages increased survival to 15–20% (Fig 3A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Effect of the resident lysogenic phages SPß and PBSX on survival after replication arrest. A. The lysogenic phages SPß and PBSX (defective) negatively affect cell survival after replication arrest (3 h HPUra treatment). Percent survival of strains JMA222 (which has both phages), LSF204 (only SPß; ΔPBSX), LSF203 (only PBSX; ΔSPß), and LSF225 (no phage; ΔPBSX, ΔSPß). B. recJ and addAB affect survival following replication arrest, even in cells without SPß and PBSX. Survival after 3 h of replication arrest for strains lacking both phage is plotted as dark gray triangles with black error bars; wt (LSF225), ΔaddAB (LSF254), and ΔrecJ (LSF231). Data from corresponding strains with both phages (JMA222, LSF253, and LSF200, respectively), plotted as light gray circles with gray error bars, are shown for comparison, and are the same as presented in Fig 1B. Statistically different means, by two-tailed t-test: * P < 0.05; ** P < 0.001; *** P < 0.0001). https://doi.org/10.1371/journal.pgen.1010564.g003 Although the presence of SPß and PBSX contributed to cell death following replication arrest, we found that addAB and recJ affected survival following replication arrest even in cells missing both PBSX and SPß. In strains cured of phages, loss of addAB caused a 10-fold drop in survival, indicating that this mutant was still more sensitive to replication arrest than wild-type cells (Fig 3B). This effect was comparable to that caused by loss of addAB in strains with both phages (~8-fold). The absence of recJ significantly increased the survival of cells after treatment with HPUra to ~50%, regardless of the presence or absence of PBSX and SPß. The increase in survival of the phage null mutant caused by loss of recJ was approximately 3-fold (Fig 3B). Together, these results indicate that the effects of addAB and recJ on survival are largely independent of killing by SPß or PBSX. End-resection is needed for survival following replication fork arrest, but RecA is not We sought to determine how strains lacking recJ and addAB-encoded nucleases involved in end resection or lacking recA behaved following arrest of replication forks with HPUra. RecA is essential for survival following treatment with DNA damaging agents. Due to the inability to efficiently load RecA in the absence of the nucleases AddAB and RecJ, a double ΔaddAB ΔrecJ mutant has a similar extreme sensitivity to DNA damaging agents as ΔrecA [17, 19, 29]. The survival following replication arrest of a strain missing both addAB and recJ was 0.3% (Fig 4). This is significantly worse than the effect caused by ΔaddAB and indicates that the loss of both end-resection pathways is synergistic and that end-resection is needed for survival in response to replication arrest in the absence of chemical damage to the DNA. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Effect of RecA and end-resection on survival of HPUra-induced replication arrest, in the absence of phages. Percent survival after 3 h of replication arrest for wild-type (LSF225), and recA (LSF658), addAB (LSF254), recJ addAB (LSF648), and recJ (LSF231) null mutants. ** Statistically different means, by two-tailed t-test (P < 0.005). https://doi.org/10.1371/journal.pgen.1010564.g004 In contrast, recA did not detectably affect survival following replication arrest. Survival of the recA null mutant was similar to that of otherwise wild-type cells, but lower than the recJ mutant (Fig 4). Based on these results, we conclude that end-resection mediated by either AddAB or RecJ is essential for recovery from replication fork arrest, but that RecA and RecA-mediated homologous recombination is not required. It is formally possible that RecA is required in the absence of RecJ. For example, survival in the absence of RecJ depends on AddAB and this could be due to end processing by AddAB and subsequent assembly of RecA onto the ssDNA. If true, then survival of a recJ recA double mutant should be less than that of a recJ single mutant. We found that the increased survival conferred by the loss of recJ was not dependent on the presence of recA (Table 2), indicating that the AddAB pathway that is required for surviving replication arrest in the absence of RecJ was not dependent on RecA. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 2. recA is not needed for increased survival to replication stress in the absence of recJ. https://doi.org/10.1371/journal.pgen.1010564.t002 RecA activity correlates with the accumulation of repair centers formed during HPUra-induced replication arrest Since RecA itself was not required to survive HPUra-induced replication arrest, but mutants with lower RecA activity like ΔrecJ and ΔrecF had increased survival, we considered the possibility that the activity of RecJ and the formation of RecA filaments were interfering with the proper processing of the arrested replication forks. If true, then there should be greater accumulation of recombination intermediates and exposed DNA ends in strains with more loading of RecA. Upon formation of a double-strand break, the DNA ends are processed by end-resection allowing RecA to be loaded onto the ssDNA. It was previously believed that following chemical damage to DNA, RecN was recruited directly to dsDNA breaks before RecA loading [19, 30, 31]. However, recent evidence indicates that RecN is recruited to the already-processed DNA in a manner dependent on RecA [17]. Nonetheless, the formation of foci of RecN can be used as a proxy for DNA repair centers, likely as a result of a dsDNA break or gaps that have been processed to allow RecA to assemble into a filament [17, 19, 30, 31]. RecN with a C-terminal tag has been shown to be functional in DNA repair and was used previously to monitor repair centers formed following DNA damage [17, 19, 30]. We constructed strains in which the native recN was replaced by recN-mNeongreen, and monitored formation of foci to determine the frequency of repair centers arising from DNA damage-independent replication arrest. Different mutants containing the fusion were treated with HPUra and scored by the presence of foci after 30 min of replication arrest (Fig 5A). In wild-type cells growing exponentially, cells with RecN-mNeongreen foci were uncommon (< 2% cells). Thirty minutes after replication arrest, approximately 20% of wild-type cells had at least one focus of RecN-mNeongreen (Fig 5A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Accumulation of exposed DNA ends during replication arrest by HPUra. A. Repair centers in response to exposed DNA ends were visualized by fluorescence microscopy of cells expressing the RecN-mNeongreen fusion. Cell membranes were stained with FM 4–64 and DNA was stained with DAPI. Wild-type (LSF708), ΔrecU (LSF740), ΔaddA (LSF650), ΔrecO (LSF818), ΔrecJ (LSF649), ΔrecF (LSF816), ΔrecJ ΔaddA (LSF706), and ΔrecA (LSF756) cells were analyzed without (gray) or 30 min after treatment with HPUra (black). At least 800 cells from two biological replicates were analyzed from each strain for the presence of RecN-mNeongreen foci. The means and standard errors are presented. B. Loss of recN causes a phenotype in cells with or without addAB or recJ. Percent survival of wild type (LSF225), ΔrecN (LSF536), ΔaddAB (LSF254), ΔrecN ΔaddA (LSF652), ΔrecJ (LSF231), and ΔrecN ΔrecJ (LSF651) are plotted. *Statistically different means according to two-tailed t-test: * P < 0.05. https://doi.org/10.1371/journal.pgen.1010564.g005 Effects of mutations on the formation of RecN-mNeongreen foci were significant. In the absence of recU or addA, 56% and 37% of cells, respectively, had at least one focus of RecN-mNeongreen (Fig 5A). In contrast, only ~12% of ΔrecJ cells contained foci and a double ΔrecJ ΔaddA mutant behaved similarly to ΔrecJ (Fig 5A). Finally, deleting recA or recO or recF, the genes encoding the RecA-loader and stabilizer, greatly reduced the number of foci of RecN-mNeongreen (Fig 5A). This indicates that association of RecN to the DNA damage centers is dependent on RecA during damage-independent replication arrest and is consistent with similar findings after treatment with agents that induce DNA damage [17]. These results support the model that RecJ and the recombinase loader and regulator impede the proper processing of the replication fork via excessive formation of RecA filaments in the absence of chemical damage to DNA, and indicate that the RecJ pathway is the major source of RecA assembly following HPUra-induced replication arrest. The neutral effect of recA in survival (Fig 4) may be a combination of increased survival due to lack of loading on DNA processed by RecJ and a decrease in RecA-mediated processes for restart. In addition, there are likely to be RecA-independent processes for replication restart (see Discussion). RecN is important for survival if RecA is assembled onto DNA Our results indicate that there are breaks or gaps in dsDNA that lead to DNA processing and RecA loading in wild-type cells following replication fork arrest. Furthermore, we found that the loss of recN caused a decrease in survival in otherwise wild-type cells following replication fork arrest (Table 1 and Fig 5B). In the absence of addAB, there was a further decrease in survival upon loss of recN (Fig 5B), consistent with the results indicating that RecA is more active and there are more dsDNA breaks in the absence of addAB than in otherwise wild-type cells. In addition, loss of recN also decreased survival of the recJ mutant (Fig 5B), indicating that RecN is beneficial regardless of which pathway (AddAB or RecJ) is used to process DNA ends. Identification of genes involved in cell survival following replication arrest Using Tn-seq, we identified non-essential genes that significantly affected the ability of cells to survive and recover from replication arrest caused by treatment with HPUra. We used a previously described library of ~1.7 x 105 unique transposon insertions of a modified version of the magellan6 transposon, magellan6x, distributed throughout the B. subtilis genome [15]. This library of cells was split in two and treated or not with HPUra for one hour. The cells were then removed from HPUra, allowing replication and cell growth to resume. Aliquots of cells were harvested 2, 3, and 4 hours later, and the location and relative number of transposon insertions in a given chromosomal site were determined by high-throughput sequencing. In a population of random transposon insertions, mutations that lead to decreased survival would be underrepresented, and the ratio between the frequency of insertions in treated and untreated samples would be <1. Conversely, insertion mutations that improved survival would be overrepresented, resulting in frequency ratios >1. The vast majority of genes had a similar number of insertions (frequency ratios of ~1) with or without HPUra, indicating that these are neutral (S1 Table). Thirty-five genes were classified as affecting survival and recovery from HPUra (see Material and Methods for analysis and criteria), and 13 of these were validated using targeted gene disruptions (Table 1). Several cellular processes appeared to affect survival, including: (i) iron-sulfur and oxidative homeostasis (rex, nadABC, defB, and nifS), likely related to the induction of oxidative stress and genes regulating NAD/NADH following treatment with HPUra [16]; (ii) cell wall stability (ltaS, rseP, oppAB, ydiL, cwlO and ponA); (iii) induction of lysogenic phage (yoyI, yonH from the SPß prophage and xpf, xkdBC from PBSX); and (iv) DNA repair and recombination. Below, we focus on several of the DNA recombination and repair genes (Fig 1A). We also validated seven of the other genes identified in the screen using targeted gene disruptions (Table 1), but did not further evaluate them as part of this study. The number of insertion sites and read frequencies for all open reading frames are presented (S1 Table). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Pathways and genes involved in cell survival after arrest of DNA replication. A. General view of the roles of the DNA repair and recombination genes identified in the Tn-seq screen. The dashed arrows represent multistep pathways. B. Percent survival of wild type (strain JMA222) and recU (GP891), polA (LSF41), recN (LSF298), addAB (LSF253) and recJ (LSF200) null mutants after 3 h of replication arrest caused by HPUra treatment. Percent survival is relative to cell viability at the time of arrest. All mutants were statistically different from wild type, as assessed by two-sample t-tests (P < 0.025). C. Effects of recJ and addAB null mutations on cell survival following replication arrest due to treatment with the DNA damaging agent mitomycin-C (MMC) or inactivation of the replicative helicase (DnaC). Left panel: survival of wild-type (JMA222), ΔaddAB (LSF253) and ΔrecJ (LSF200) cells after 3 h of treatment with 0.33 μg/ml of MMC. Right panel: wild-type, ΔrecJ and ΔaddAB cells carrying a dnaCts allele (LSF176, LSF274, and LSF326) were grown at permissive temperature (30°C) to early exponential phase and shifted to non-permissive temperature (49°C) for 4 h. Percent survival is relative to the population before treatment (left) or the temperature shift (right). The deletion mutants in both conditions were statistically different from wild type, as assessed by two-sample t-tests (P < 0.025). Throughout this work, error bars correspond to one standard deviation of the mean and each point is from an independent culture (biological replicates). https://doi.org/10.1371/journal.pgen.1010564.g001 Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. Genes identified in Tn-seq analysis of cells surviving replication arrest in the absence of DNA damage. https://doi.org/10.1371/journal.pgen.1010564.t001 Effects of null mutations in genes involved in DNA repair and recombination on survival after replication arrest In order to confirm the importance of the DNA repair and recombination genes identified in the Tn-seq screen, we made null mutations in six of the loci identified (polA, recU, recN, recJ, walJ, and addAB; note that addB and addA comprise an operon and for all experiments except where indicated, both genes were deleted) and tested their effects on survival after HPUra-mediated replication fork arrest. We treated wild-type and mutant cells with HPUra for 3 h and then removed HPUra and measured the number of viable cells remaining. Approximately 4.5% of wild-type cells survived this prolonged replication arrest (Fig 1B). In contrast, only 0.5–1% of polA, recU, recN, and addAB mutant cells survived. In contrast to the detrimental effects of the loss of polA, recU, recN, and addAB on surviving replication fork arrest, we found that loss of recJ promoted survival. In the recJ null mutant, approximately 45% of cells remained viable after 3 hr of replication arrest (Fig 1B). These results indicate that normally RecJ is detrimental to survival of replication arrest in the absence of DNA lesions. Together, these results confirm the initial Tn-seq data that indicated that these genes are important for promoting survival following replication fork arrest in the absence of DNA damage. Comparison of recJ and addAB mutants in response to replication arrest Both RecJ and AddAB are involved in end-resection (Fig 1A), and both improve survival after treatment with DNA-damaging agents [17–20]. Therefore, it was somewhat surprising that a recJ null mutant had increased survival in response to replication fork arrest in the absence of externally induced chemical damage to DNA. We confirmed that ΔrecJ and ΔaddAB mutants were more sensitive than wild-type cells to treatment with the DNA damaging agent mitomycin C (MMC). Approximately 1.5% of wild-type cells survived treatment with MMC (0.33 μg/ml) for 3 h (Fig 1C). In contrast, addAB and recJ mutants had 7 and 4-fold lower survival, respectively. To verify that the effects of recJ and addAB on survival and recovery from HPUra were due to replication arrest, rather than some other possible effect, we tested their survival under conditions in which replication was arrested with a temperature-sensitive allele of the replicative DNA helicase (dnaCts) [21]. Inactivation of the replicative helicase mirrored the phenotypes observed during HPUra-mediated replication arrest (Fig 1C). We grew dnaCts (LSF176), dnaCts ΔaddAB (LSF326), and dnaCts ΔrecJ (LSF274) strains at a permissive temperature (30°C). Cells were then shifted to non-permissive temperature (49°C) to arrest replication elongation. After 4 h at 49°C, we measured the number of viable cells from each of the three strains. In the otherwise wild-type cells, prolonged replication arrest at the non-permissive temperature caused a decrease in cell viability to about 20% of the original population (Fig 1C). The loss of addAB exacerbated this effect, and only 4% of cells were viable. In contrast, the loss of recJ largely protected the cells from the outcome of replication arrest due to inactivation of the replicative helicase, and 65% of cells survived (Fig 1C). Together, our results indicate that recJ is detrimental to surviving replication fork arrest in the absence of DNA lesions, and confirm previous work [17, 19] demonstrating that recJ contributes to survival following chemical damage to DNA. Loss of RecA loader RecO or regulator RecF protects ΔaddAB cells from arrest-induced death End-resection by AddAB or RecJ produces the substrate for interaction with RecA. RecO is the main protein responsible for loading RecA onto the ssDNA substrate (Fig 1A), and RecF stabilizes the resulting RecA nucleofilaments [11]. Insertions in recO and recF were not recovered in our initial Tn-seq screen because they were depleted from the initial library prior to treatment with HPUra (S1 Table). We suspect that this depletion was because cells were not kept in the dark during library preparation and these mutants are probably sensitive to ambient UV light. To test if recO and recF contribute to survival following replication arrest, we constructed recO and recF null mutants. We found that deletion of either recO or recF significantly increased cell survival following replication fork arrest (HPUra treatment) in comparison to wild type (Fig 2A). Both deletions suppressed the sensitivity of ΔaddAB mutants to HPUra (Fig 2A, see addAB recO and addAB recF), indicating that loading RecA was at least partially responsible for decreased survival of the ΔaddAB mutant. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. RecA loading is detrimental to survival during HPUra-induced replication arrest. A. Survival after HPUra treatment is greater in the absence of RecA loader and regulator (RecO, RecF). Fold change in survival after 3 h of arrest for wild type (JMA222), ΔrecO (LSF814), ΔrecF (LSF270), ΔaddAB (LSF253), ΔaddAB ΔrecO (LSF817), ΔaddAB ΔrecF (LSF709), ΔrecJ (LSF200), ΔrecJ ΔrecO (LSF815), ΔrecJ ΔrecF (LSF282). Significant difference in means (two-sample T-test): * P < 0.05; ** P < 0.01. B. RecA activity is increased in the absence of addAB and decreased in the absence of recJ. Transcription from the promoter of the SOS-inducible gene yneA is derepressed when RecA is active and serves as an indicator of RecA activity. Cultures of wild-type, ΔrecJ, and ΔaddAB strains containing amyE::PyneA-lacZ (LSF633, LSF634, and LSF635, respectively) were treated or not with HPUra. After 30 min, aliquots were collected for measuring ß-galactosidase activity. Results from three independent experiments are presented. https://doi.org/10.1371/journal.pgen.1010564.g002 In contrast to effects on the addAB mutant, deletion of recF had little if any effect on the sensitivity of the ΔrecJ mutant to HPUra, indicating that RecF and RecJ likely affect the same process–formation of stable RecA filaments (Fig 2A). The SOS response is reduced in the absence of recJ and increased in the absence of addAB We postulated that if the increased survival of ΔrecJ is a consequence of lower stability of RecA filaments, then ΔrecJ might have lower RecA activity. Once assembled onto ssDNA, RecA induces autocleavage of the SOS repressor LexA, allowing expression of many genes involved in the response to genotoxic stress [22–24]. Consequently, expression of genes repressed by LexA can be used as a proxy for the levels of activated RecA. yneA encodes an inhibitor of cell division [25], and is repressed by LexA and de-repressed in response to DNA damage and replication fork arrest [26]. We used a fusion of the promoter region of yneA to lacZ (PyneA-lacZ) and measured expression in the mutants in the presence and absence of HPUra. For wild-type cells in defined minimal medium, ß-galactosidase activity from the PyneA-lacZ fusion increased approximately 20-fold after 30 min of replication arrest (Fig 2B). In the ΔrecJ mutant, there was a basal level of activity similar to that in wild type, but only an approximately 6-fold increase after addition of HPUra, indicating that recJ plays an important role in activation of RecA. The absence of addAB led to very high PyneA-lacZ activity during exponential growth (Fig 2B), indicating that there is an abundance of active RecA in these cells. Lysogenic phages are not responsible for the effects of recJ and addAB on survival after replication arrest It seemed plausible that the impact of recJ and addAB on survival following replication stress could be due to different levels of phage induction, as both mutations affect the activity of RecA. The B. subtilis strains used in the experiments described above have two resident lysogenic phages, both of which undergo RecA-dependent activation after genotoxic stress: SPß and the defective phage PBSX [16, 26–28]. Both encode toxins and autolysins which, when expressed, can cause cell death. We found that insertions in two genes from SPß (yoyI, yonH) and three genes involved in the early steps of PBSX induction (xpf, xkdBC) were overrepresented in the screen, indicating that loss of these genes helped cells survive arrest of replication elongation (Table 1). We measured survival of otherwise wild-type cells missing either PBSX, SPß, or both following replication arrest with HPUra. In these experiments, the survival of wild-type cells (containing both lysogenic phages) was 3.5% (Fig 3A). Loss of either phage increased survival to approximately 7%, and loss of both phages increased survival to 15–20% (Fig 3A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Effect of the resident lysogenic phages SPß and PBSX on survival after replication arrest. A. The lysogenic phages SPß and PBSX (defective) negatively affect cell survival after replication arrest (3 h HPUra treatment). Percent survival of strains JMA222 (which has both phages), LSF204 (only SPß; ΔPBSX), LSF203 (only PBSX; ΔSPß), and LSF225 (no phage; ΔPBSX, ΔSPß). B. recJ and addAB affect survival following replication arrest, even in cells without SPß and PBSX. Survival after 3 h of replication arrest for strains lacking both phage is plotted as dark gray triangles with black error bars; wt (LSF225), ΔaddAB (LSF254), and ΔrecJ (LSF231). Data from corresponding strains with both phages (JMA222, LSF253, and LSF200, respectively), plotted as light gray circles with gray error bars, are shown for comparison, and are the same as presented in Fig 1B. Statistically different means, by two-tailed t-test: * P < 0.05; ** P < 0.001; *** P < 0.0001). https://doi.org/10.1371/journal.pgen.1010564.g003 Although the presence of SPß and PBSX contributed to cell death following replication arrest, we found that addAB and recJ affected survival following replication arrest even in cells missing both PBSX and SPß. In strains cured of phages, loss of addAB caused a 10-fold drop in survival, indicating that this mutant was still more sensitive to replication arrest than wild-type cells (Fig 3B). This effect was comparable to that caused by loss of addAB in strains with both phages (~8-fold). The absence of recJ significantly increased the survival of cells after treatment with HPUra to ~50%, regardless of the presence or absence of PBSX and SPß. The increase in survival of the phage null mutant caused by loss of recJ was approximately 3-fold (Fig 3B). Together, these results indicate that the effects of addAB and recJ on survival are largely independent of killing by SPß or PBSX. End-resection is needed for survival following replication fork arrest, but RecA is not We sought to determine how strains lacking recJ and addAB-encoded nucleases involved in end resection or lacking recA behaved following arrest of replication forks with HPUra. RecA is essential for survival following treatment with DNA damaging agents. Due to the inability to efficiently load RecA in the absence of the nucleases AddAB and RecJ, a double ΔaddAB ΔrecJ mutant has a similar extreme sensitivity to DNA damaging agents as ΔrecA [17, 19, 29]. The survival following replication arrest of a strain missing both addAB and recJ was 0.3% (Fig 4). This is significantly worse than the effect caused by ΔaddAB and indicates that the loss of both end-resection pathways is synergistic and that end-resection is needed for survival in response to replication arrest in the absence of chemical damage to the DNA. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Effect of RecA and end-resection on survival of HPUra-induced replication arrest, in the absence of phages. Percent survival after 3 h of replication arrest for wild-type (LSF225), and recA (LSF658), addAB (LSF254), recJ addAB (LSF648), and recJ (LSF231) null mutants. ** Statistically different means, by two-tailed t-test (P < 0.005). https://doi.org/10.1371/journal.pgen.1010564.g004 In contrast, recA did not detectably affect survival following replication arrest. Survival of the recA null mutant was similar to that of otherwise wild-type cells, but lower than the recJ mutant (Fig 4). Based on these results, we conclude that end-resection mediated by either AddAB or RecJ is essential for recovery from replication fork arrest, but that RecA and RecA-mediated homologous recombination is not required. It is formally possible that RecA is required in the absence of RecJ. For example, survival in the absence of RecJ depends on AddAB and this could be due to end processing by AddAB and subsequent assembly of RecA onto the ssDNA. If true, then survival of a recJ recA double mutant should be less than that of a recJ single mutant. We found that the increased survival conferred by the loss of recJ was not dependent on the presence of recA (Table 2), indicating that the AddAB pathway that is required for surviving replication arrest in the absence of RecJ was not dependent on RecA. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 2. recA is not needed for increased survival to replication stress in the absence of recJ. https://doi.org/10.1371/journal.pgen.1010564.t002 RecA activity correlates with the accumulation of repair centers formed during HPUra-induced replication arrest Since RecA itself was not required to survive HPUra-induced replication arrest, but mutants with lower RecA activity like ΔrecJ and ΔrecF had increased survival, we considered the possibility that the activity of RecJ and the formation of RecA filaments were interfering with the proper processing of the arrested replication forks. If true, then there should be greater accumulation of recombination intermediates and exposed DNA ends in strains with more loading of RecA. Upon formation of a double-strand break, the DNA ends are processed by end-resection allowing RecA to be loaded onto the ssDNA. It was previously believed that following chemical damage to DNA, RecN was recruited directly to dsDNA breaks before RecA loading [19, 30, 31]. However, recent evidence indicates that RecN is recruited to the already-processed DNA in a manner dependent on RecA [17]. Nonetheless, the formation of foci of RecN can be used as a proxy for DNA repair centers, likely as a result of a dsDNA break or gaps that have been processed to allow RecA to assemble into a filament [17, 19, 30, 31]. RecN with a C-terminal tag has been shown to be functional in DNA repair and was used previously to monitor repair centers formed following DNA damage [17, 19, 30]. We constructed strains in which the native recN was replaced by recN-mNeongreen, and monitored formation of foci to determine the frequency of repair centers arising from DNA damage-independent replication arrest. Different mutants containing the fusion were treated with HPUra and scored by the presence of foci after 30 min of replication arrest (Fig 5A). In wild-type cells growing exponentially, cells with RecN-mNeongreen foci were uncommon (< 2% cells). Thirty minutes after replication arrest, approximately 20% of wild-type cells had at least one focus of RecN-mNeongreen (Fig 5A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Accumulation of exposed DNA ends during replication arrest by HPUra. A. Repair centers in response to exposed DNA ends were visualized by fluorescence microscopy of cells expressing the RecN-mNeongreen fusion. Cell membranes were stained with FM 4–64 and DNA was stained with DAPI. Wild-type (LSF708), ΔrecU (LSF740), ΔaddA (LSF650), ΔrecO (LSF818), ΔrecJ (LSF649), ΔrecF (LSF816), ΔrecJ ΔaddA (LSF706), and ΔrecA (LSF756) cells were analyzed without (gray) or 30 min after treatment with HPUra (black). At least 800 cells from two biological replicates were analyzed from each strain for the presence of RecN-mNeongreen foci. The means and standard errors are presented. B. Loss of recN causes a phenotype in cells with or without addAB or recJ. Percent survival of wild type (LSF225), ΔrecN (LSF536), ΔaddAB (LSF254), ΔrecN ΔaddA (LSF652), ΔrecJ (LSF231), and ΔrecN ΔrecJ (LSF651) are plotted. *Statistically different means according to two-tailed t-test: * P < 0.05. https://doi.org/10.1371/journal.pgen.1010564.g005 Effects of mutations on the formation of RecN-mNeongreen foci were significant. In the absence of recU or addA, 56% and 37% of cells, respectively, had at least one focus of RecN-mNeongreen (Fig 5A). In contrast, only ~12% of ΔrecJ cells contained foci and a double ΔrecJ ΔaddA mutant behaved similarly to ΔrecJ (Fig 5A). Finally, deleting recA or recO or recF, the genes encoding the RecA-loader and stabilizer, greatly reduced the number of foci of RecN-mNeongreen (Fig 5A). This indicates that association of RecN to the DNA damage centers is dependent on RecA during damage-independent replication arrest and is consistent with similar findings after treatment with agents that induce DNA damage [17]. These results support the model that RecJ and the recombinase loader and regulator impede the proper processing of the replication fork via excessive formation of RecA filaments in the absence of chemical damage to DNA, and indicate that the RecJ pathway is the major source of RecA assembly following HPUra-induced replication arrest. The neutral effect of recA in survival (Fig 4) may be a combination of increased survival due to lack of loading on DNA processed by RecJ and a decrease in RecA-mediated processes for restart. In addition, there are likely to be RecA-independent processes for replication restart (see Discussion). RecN is important for survival if RecA is assembled onto DNA Our results indicate that there are breaks or gaps in dsDNA that lead to DNA processing and RecA loading in wild-type cells following replication fork arrest. Furthermore, we found that the loss of recN caused a decrease in survival in otherwise wild-type cells following replication fork arrest (Table 1 and Fig 5B). In the absence of addAB, there was a further decrease in survival upon loss of recN (Fig 5B), consistent with the results indicating that RecA is more active and there are more dsDNA breaks in the absence of addAB than in otherwise wild-type cells. In addition, loss of recN also decreased survival of the recJ mutant (Fig 5B), indicating that RecN is beneficial regardless of which pathway (AddAB or RecJ) is used to process DNA ends. Discussion The replisome pauses or collapses not only when it encounters damaged DNA, but also when roadblocks, such as DNA binding proteins, are encountered, or when the replisome itself is inhibited or damaged. We specifically focused on how cells recover from replication arrest by replisome inhibition, in the absence of chemical damage to DNA. It is well known that RecA activity and both the RecJ and AddAB end-resection pathways are critically important for surviving exposure to DNA damage-inducing agents [32–34]. In contrast, we found a different scenario when there is no DNA damage and a replisome component is inhibited (chemically or using a thermosensitive allele). In this situation, end-resection was necessary, but the RecJ pathway and high RecA activity were detrimental, and the AddAB pathway was beneficial. Our results show that different types of genotoxic stresses require different levels of RecA activity, which are obtained by the AddAB vs. RecJ pathways, and that the choice of pathway affects survival. The consequences of using each end-resection pathway depend on the kind of replication arrest As depicted in Fig 6A, when replication is blocked, both end-resection pathways work together to process the collapsed replication fork. After fork regression, which involves reannealing of the template DNA and annealing of the two daughter strands, AddAB can unwind and degrade the daughter strands until it reaches a Chi site, after which it begins generating a 3’ ssDNA tail [10]. RecJ can target the collapsed fork directly, or an already resected fork [35], probably extending the ssDNA substrate generated by AddAB and leading to longer RecA nucleofilaments. RecFO(R) then ensures RecA nucleofilament loading and stability [11]. In this case, the role of RecA is not to promote recombination, but rather to stabilize the DNA at the fork until the DNA damage has been repaired [32, 33]. Our results indicate that when replication stops because a replisome component is inhibited, AddAB activity supports optimal recovery (black arrows in the left side of Fig 6B) and after fork regression, AddAB generates less substrate for RecA loading than does RecJ. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. The role of end-resection pathways AddAB and RecJ on survival outcomes after replication arrest. A. When replication arrest is caused by lesions on the DNA template or other genotoxic stresses requiring RecA activity, both end-resection pathways work together, generating stable RecA filaments that delay restart and support DNA repair. B. If replication arrest is caused by the inhibition of replisome components or other stresses that do not require RecA, AddAB favors a pathway (black arrows) with limited RecA filaments that promotes survival. In this situation, RecJ and the RecA loader RecO and regulator RecF impair survival by promoting a pathway (gray arrows) with high RecA activity. This could prevent proper processing of the fork, exposing DNA ends and creating a futile circle. https://doi.org/10.1371/journal.pgen.1010564.g006 Our results indicate that RecJ and the loader RecO (and regulator RecF) negatively affect the process of fork repair by enhancing RecA activity and preventing fork regression (gray arrows by the right side of Fig 6B). RecJ can target a range of DNA substrates [9, 36]. In addition to competing with AddAB for access to an already regressed fork, it can degrade a daughter strand in the collapsed fork, which would prevent fork regression and the generation of the blunt ends AddAB requires. Also, RecJ and RecO localize to the replication fork via interaction with SSB [37], and therefore have immediate access to the replication fork, whereas AddAB does not. In the absence of DNA damage, the resulting stable RecA filaments that help promote survival in other conditions may expose DNA ends that unnecessarily delay fork processing. Any DNA breaks or exposed ends arising during this step would need to go through an additional round of end-resection and RecA loading, feeding the detrimental pathway. Different types of genotoxic stresses require different levels and duration of RecA activity The different effects that RecA has on survival, depending on the kind of replication arrest, could be due to the complexity of the resulting collapsed fork and duration of RecA activity necessary for restart. Treatment with MMC, for example, impedes replication due to template lesions and “dirty” DNA breaks, i.e. ones that cannot be ligated because they lack a 3’ hydroxyl or 5’ phosphate. The recovery from this kind of arrest can take up to 3 h in B. subtilis and relies on high levels of RecA activity [29, 30]. The repair of a “clean”, ligatable single break in the chromosome, however, requires very transient RecA filaments and takes less than 20 min, at least in E. coli [38]. When a DNA-binding protein acting as a roadblock arrests replication, E. coli cells are able to process these forks and recover normally, regardless of the presence of recA [39]. In this case, there are no chemical lesions to be repaired or complex structures to be disassembled at the fork before processing. Likewise, our data indicate that the collapsed forks resulting from the inhibition of a replisome component (PolC by HPUra treatment or arrest of the replicative helicase DnaC using at temperature sensitive mutant) readily undergo fork regression and processing. In this scenario, we found that the excess RecA activity that results from RecJ and the recombinase loader and regulator is not only unnecessary but also detrimental to survival. Replication-transcription conflicts also cause replication arrest (reviewed in [3, 40]). The importance of RecA in replication restart at these sites is unclear. One report found that RecA was not required in E. coli [41], but another found that RecA was needed in B. subtilis [42]. The fact that RecA is required in B. subtilis for resolving head-on replication-transcription conflicts, but not for recovery from replication arrest caused by HPUra may be due to the different structures that are formed after each of these treatments. HPUra treatment has been shown to leave many of the replisome components largely intact, at least initially [43–46], whereas replication-transcription conflicts likely cause replisome disassembly [47], and this may predispose arrested forks to different restart pathways. RecJ leads to increased RecA loading and DNA repair centers and increased cell death when fork repair does not require RecA activity We hypothesize that during HPUra-induced arrest in a wild-type cell, the stable RecA nucleofilaments favored by RecJ prevent optimal repair of the collapsed fork by unnecessarily delaying restart, perhaps by inhibiting the ability of PriA to function or increasing the frequency of double-strand breaks, or both. We noted a positive correlation between the predicted levels of RecA activity in different mutants following HPUra treatment and the accumulation of DNA repair centers (increased RecN-mNeongreen foci; Fig 5A). These are likely a result of increased DNA breaks or gaps that have undergone end-resection and RecA loading, followed by RecN recruitment [17, 48]. Stable RecJ-dependent RecA filaments may factor into the propensity to cause breaks in these long stretches of coated ssDNA, leading to more DNA repair centers. Similar to what is observed for repair in the context of DNA damage, we found that deletion of recN decreased survival of HPUra treatment in wild type, ΔaddAB, and ΔrecJ mutants, indicating that under all of these conditions, recruitment of RecN helps to repair the DNA, regardless of the level of activity of RecA. These two mechanisms–increased DNA repair centers (likely from increased dsDNA breaks or gaps) and unnecessarily stalled restart–presumably synergistically contribute to cell death during HPUra treatment. Few studies have focused on the survival outcome of recombination mutants to replisome inhibition in the absence of external DNA damage. Nevertheless, there are some hints of a disadvantageous role of RecJ in situations where RecA is not required in E. coli, which could be explained by excessive loading of the recombinase. A strain lacking recB (analogous to ΔaddAB in B. subtilis) can still process a fork after collapse due to a roadblock, but survival was low, while a strain lacking recA was indistinguishable from wild type [39]. Although a mechanism for these findings was not proposed, the results indicate that a pathway parallel to RecBCD (presumably RecJ/gap repair) may be processing the collapsed forks in a way that decreases viability. Toxicity caused by RecJ, RecFOR, and RecA also occurs in the absence of the helicases (DNA translocases) encoded by rep and uvrD in E. coli and pcrA in B. subtilis. This toxicity is apparently due to excess or improper loading of RecA at blocked replication forks and led to the understanding that the helicases normally help remove or clear RecA to ensure proper replication restart [49–51]. These examples highlight the importance of controlling RecA loading and how improper or excessive loading can be detrimental to the cell. RecJ homologs are widespread from bacteria to eukaryotes [52]. It is not known if the detrimental nature of RecJ and excessive loading of RecA in response to replication arrest is widespread, but the results from E. coli discussed above, and the conserved nature of RecJ and RecA indicate that this might be the case. Possible mechanisms of surviving replication fork arrest in the absence of RecA Our data indicate that although RecA is dispensable, and at times detrimental, to the repair of replication arrest in absence of external DNA damage, end-resection is still required for survival. This indicates that formation of RecA-dependent Holliday junctions is not required for survival, and that the end-resection has a purpose other than loading of RecA. In E. coli, replication fork reversal can be accomplished independently of RecA, using RecBCD (analogous to AddAB in B. subitilis) and RuvAB, and has been observed in situations of stalled forks in the absence of chemical damage to DNA, including fork stalling caused by mutations in the DNA polymerase holoenzyme (reviewed in [2]), conflicts caused by DNA binding proteins [39], or head-on collisions with RNA polymerase [41]. A similar mechanism may occur in the absence of RecA after HPUra treatment, however it is also possible there are other RecA-independent mechanisms in B. subtilis, perhaps including non-homologous end-joining [53] or a role for DNA polymerase I [54, 55]. Additionally, breaks that occur at the replication fork resulting in damage to only one copy of the chromosome would yield one viable cell, rather than two, but would not be lethal nor require recombination. Indeed, in E. coli, in the absence of RecA, damaged chromosomes are degraded by RecBCD and can result in cells containing an odd number of chromosomes due to selective degradation of the damaged chromosomes [56]. Bacteria use all pathways for repair, even when it hinders proper recovery Each end-resection mechanism seems adapted to a kind of genotoxic stress in B. subtilis: AddAB for double-strand breaks from damage or fork collapse, and RecJ for gaps from DNA repair. However, our data highlighted that bacteria may use an unfavorable (“wrong”) pathway for repair. RecJ commits repair of some DNA damage-independent collapsed forks to a pathway that decreases viability, and it probably competes with AddAB for blunt DNA ends. RecJ and stable RecA loading only promote survival during DNA repair, but our data indicate that they are also used during replication stresses that do not require RecA. Ultimately, this indicates that cells do not have an effective mechanism to discriminate between these different types of broken forks, potentially leading to increased cell death in some, but enabling robust repair and restart in other circumstances. The consequences of using each end-resection pathway depend on the kind of replication arrest As depicted in Fig 6A, when replication is blocked, both end-resection pathways work together to process the collapsed replication fork. After fork regression, which involves reannealing of the template DNA and annealing of the two daughter strands, AddAB can unwind and degrade the daughter strands until it reaches a Chi site, after which it begins generating a 3’ ssDNA tail [10]. RecJ can target the collapsed fork directly, or an already resected fork [35], probably extending the ssDNA substrate generated by AddAB and leading to longer RecA nucleofilaments. RecFO(R) then ensures RecA nucleofilament loading and stability [11]. In this case, the role of RecA is not to promote recombination, but rather to stabilize the DNA at the fork until the DNA damage has been repaired [32, 33]. Our results indicate that when replication stops because a replisome component is inhibited, AddAB activity supports optimal recovery (black arrows in the left side of Fig 6B) and after fork regression, AddAB generates less substrate for RecA loading than does RecJ. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. The role of end-resection pathways AddAB and RecJ on survival outcomes after replication arrest. A. When replication arrest is caused by lesions on the DNA template or other genotoxic stresses requiring RecA activity, both end-resection pathways work together, generating stable RecA filaments that delay restart and support DNA repair. B. If replication arrest is caused by the inhibition of replisome components or other stresses that do not require RecA, AddAB favors a pathway (black arrows) with limited RecA filaments that promotes survival. In this situation, RecJ and the RecA loader RecO and regulator RecF impair survival by promoting a pathway (gray arrows) with high RecA activity. This could prevent proper processing of the fork, exposing DNA ends and creating a futile circle. https://doi.org/10.1371/journal.pgen.1010564.g006 Our results indicate that RecJ and the loader RecO (and regulator RecF) negatively affect the process of fork repair by enhancing RecA activity and preventing fork regression (gray arrows by the right side of Fig 6B). RecJ can target a range of DNA substrates [9, 36]. In addition to competing with AddAB for access to an already regressed fork, it can degrade a daughter strand in the collapsed fork, which would prevent fork regression and the generation of the blunt ends AddAB requires. Also, RecJ and RecO localize to the replication fork via interaction with SSB [37], and therefore have immediate access to the replication fork, whereas AddAB does not. In the absence of DNA damage, the resulting stable RecA filaments that help promote survival in other conditions may expose DNA ends that unnecessarily delay fork processing. Any DNA breaks or exposed ends arising during this step would need to go through an additional round of end-resection and RecA loading, feeding the detrimental pathway. Different types of genotoxic stresses require different levels and duration of RecA activity The different effects that RecA has on survival, depending on the kind of replication arrest, could be due to the complexity of the resulting collapsed fork and duration of RecA activity necessary for restart. Treatment with MMC, for example, impedes replication due to template lesions and “dirty” DNA breaks, i.e. ones that cannot be ligated because they lack a 3’ hydroxyl or 5’ phosphate. The recovery from this kind of arrest can take up to 3 h in B. subtilis and relies on high levels of RecA activity [29, 30]. The repair of a “clean”, ligatable single break in the chromosome, however, requires very transient RecA filaments and takes less than 20 min, at least in E. coli [38]. When a DNA-binding protein acting as a roadblock arrests replication, E. coli cells are able to process these forks and recover normally, regardless of the presence of recA [39]. In this case, there are no chemical lesions to be repaired or complex structures to be disassembled at the fork before processing. Likewise, our data indicate that the collapsed forks resulting from the inhibition of a replisome component (PolC by HPUra treatment or arrest of the replicative helicase DnaC using at temperature sensitive mutant) readily undergo fork regression and processing. In this scenario, we found that the excess RecA activity that results from RecJ and the recombinase loader and regulator is not only unnecessary but also detrimental to survival. Replication-transcription conflicts also cause replication arrest (reviewed in [3, 40]). The importance of RecA in replication restart at these sites is unclear. One report found that RecA was not required in E. coli [41], but another found that RecA was needed in B. subtilis [42]. The fact that RecA is required in B. subtilis for resolving head-on replication-transcription conflicts, but not for recovery from replication arrest caused by HPUra may be due to the different structures that are formed after each of these treatments. HPUra treatment has been shown to leave many of the replisome components largely intact, at least initially [43–46], whereas replication-transcription conflicts likely cause replisome disassembly [47], and this may predispose arrested forks to different restart pathways. RecJ leads to increased RecA loading and DNA repair centers and increased cell death when fork repair does not require RecA activity We hypothesize that during HPUra-induced arrest in a wild-type cell, the stable RecA nucleofilaments favored by RecJ prevent optimal repair of the collapsed fork by unnecessarily delaying restart, perhaps by inhibiting the ability of PriA to function or increasing the frequency of double-strand breaks, or both. We noted a positive correlation between the predicted levels of RecA activity in different mutants following HPUra treatment and the accumulation of DNA repair centers (increased RecN-mNeongreen foci; Fig 5A). These are likely a result of increased DNA breaks or gaps that have undergone end-resection and RecA loading, followed by RecN recruitment [17, 48]. Stable RecJ-dependent RecA filaments may factor into the propensity to cause breaks in these long stretches of coated ssDNA, leading to more DNA repair centers. Similar to what is observed for repair in the context of DNA damage, we found that deletion of recN decreased survival of HPUra treatment in wild type, ΔaddAB, and ΔrecJ mutants, indicating that under all of these conditions, recruitment of RecN helps to repair the DNA, regardless of the level of activity of RecA. These two mechanisms–increased DNA repair centers (likely from increased dsDNA breaks or gaps) and unnecessarily stalled restart–presumably synergistically contribute to cell death during HPUra treatment. Few studies have focused on the survival outcome of recombination mutants to replisome inhibition in the absence of external DNA damage. Nevertheless, there are some hints of a disadvantageous role of RecJ in situations where RecA is not required in E. coli, which could be explained by excessive loading of the recombinase. A strain lacking recB (analogous to ΔaddAB in B. subtilis) can still process a fork after collapse due to a roadblock, but survival was low, while a strain lacking recA was indistinguishable from wild type [39]. Although a mechanism for these findings was not proposed, the results indicate that a pathway parallel to RecBCD (presumably RecJ/gap repair) may be processing the collapsed forks in a way that decreases viability. Toxicity caused by RecJ, RecFOR, and RecA also occurs in the absence of the helicases (DNA translocases) encoded by rep and uvrD in E. coli and pcrA in B. subtilis. This toxicity is apparently due to excess or improper loading of RecA at blocked replication forks and led to the understanding that the helicases normally help remove or clear RecA to ensure proper replication restart [49–51]. These examples highlight the importance of controlling RecA loading and how improper or excessive loading can be detrimental to the cell. RecJ homologs are widespread from bacteria to eukaryotes [52]. It is not known if the detrimental nature of RecJ and excessive loading of RecA in response to replication arrest is widespread, but the results from E. coli discussed above, and the conserved nature of RecJ and RecA indicate that this might be the case. Possible mechanisms of surviving replication fork arrest in the absence of RecA Our data indicate that although RecA is dispensable, and at times detrimental, to the repair of replication arrest in absence of external DNA damage, end-resection is still required for survival. This indicates that formation of RecA-dependent Holliday junctions is not required for survival, and that the end-resection has a purpose other than loading of RecA. In E. coli, replication fork reversal can be accomplished independently of RecA, using RecBCD (analogous to AddAB in B. subitilis) and RuvAB, and has been observed in situations of stalled forks in the absence of chemical damage to DNA, including fork stalling caused by mutations in the DNA polymerase holoenzyme (reviewed in [2]), conflicts caused by DNA binding proteins [39], or head-on collisions with RNA polymerase [41]. A similar mechanism may occur in the absence of RecA after HPUra treatment, however it is also possible there are other RecA-independent mechanisms in B. subtilis, perhaps including non-homologous end-joining [53] or a role for DNA polymerase I [54, 55]. Additionally, breaks that occur at the replication fork resulting in damage to only one copy of the chromosome would yield one viable cell, rather than two, but would not be lethal nor require recombination. Indeed, in E. coli, in the absence of RecA, damaged chromosomes are degraded by RecBCD and can result in cells containing an odd number of chromosomes due to selective degradation of the damaged chromosomes [56]. Bacteria use all pathways for repair, even when it hinders proper recovery Each end-resection mechanism seems adapted to a kind of genotoxic stress in B. subtilis: AddAB for double-strand breaks from damage or fork collapse, and RecJ for gaps from DNA repair. However, our data highlighted that bacteria may use an unfavorable (“wrong”) pathway for repair. RecJ commits repair of some DNA damage-independent collapsed forks to a pathway that decreases viability, and it probably competes with AddAB for blunt DNA ends. RecJ and stable RecA loading only promote survival during DNA repair, but our data indicate that they are also used during replication stresses that do not require RecA. Ultimately, this indicates that cells do not have an effective mechanism to discriminate between these different types of broken forks, potentially leading to increased cell death in some, but enabling robust repair and restart in other circumstances. Materials and methods Bacterial strains Unless otherwise indicated, all B. subtilis strains used in this work derive from JMA222 [57], a version of JH642 [58] cured of the integrative and conjugative element ICEBs1. Strain genotypes and references for previously described strains are listed in Table 3. Previously described alleles were introduced into JMA222 by transforming naturally competent cells with genomic DNA from the relevant strain. These alleles include: dnaC30(ts)-mls [21], recU::cat [59], recA260 [60], walJ::erm [61], recF::spc [62], amyE::Pyne-lacZ cat [63], ponA::spc [64], and addAB::kan [65]. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 3. B. subtilis strains used. https://doi.org/10.1371/journal.pgen.1010564.t003 Introduction of new alleles was done by transforming cells with linear DNA products generated either by traditional cloning or by isothermal assembly [69]. Insertion-deletion constructs contained antibiotic resistance cassettes typically flanked by 800–1000 bp of genomic sequences upstream and downstream of region to be deleted. These alleles include addA::spc, recJ::spc, recN::kan, recO::cat, recF::spc, ycgE::spc, hrcA::cat, polA::cat and cwlO::spc. To generate deletions of SPß (LSF203) and PBSX (LSF204), each was substituted by a kanamycin resistance gene flanked by loxP sites, generating strains LSF197 (ΔSPß::lox-kan-lox) and LSF198 (ΔPBSX::lox-kan-lox). The cassettes were removed by the Cre recombinase (from bacteriophage P1) using the cre expression vector pDR244 as previously described [70], leaving a 75 bp insertion (’scar’ containing a single lox site). A strain devoid of both phages (LSF225) resulted from transforming LSF203 (ΔSPß::lox) with genomic DNA from LSF201 (a ΔPBSX::lox-kan-lox strain), and recombining the antibiotic cassette out. The ΔPBSX allele also removes spoIISABC (toxin-antitoxin-antitoxin) which is just downstream of PBSX but is not normally thought to be part of it [71]. To visualize DNA repair centers, we constructed a strain with a recN-mNeongreen (recN-mng) fusion, inserted in the native locus such that the fusion is the only recN allele. The full linear DNA product contained the 800 bp at the 3’ end of recN, an in-frame linker (5’- CTCGAGGGATCTGGCCAAGGAAGCGGC-3’; encoding 9 amino acids), the mNeongreen coding sequence (a gift from Ethan Garner), a kanamycin resistance cassette, and 800 bp genomic sequence downstream of the recN stop codon. Transformation of this construct into JMA222 generated strain LSF284 that contains recN::[recN-9aa-mNeongreen kan]; referred to as recN-mng kan. The recN-mng kan allele from LSF284 was then introduced into various strains, including LSF225 (generating LSF708). Media and antibiotics All experiments were performed in defined minimal medium containing 50 mM MOPS (S750) and supplemented with 1% glucose, 0.1% glutamate, 40 μg/ml phenylalanine and 40 μg/ml tryptophan [72]. When required, the following concentration of antibiotics were used: 5 mg/ml chloramphenicol, 100 μg/ml spectinomycin, 5 μg/ml kanamycin, and 0.5μg/ml erthyromycin plus 12.5 μg/ml lincomycin to select for macrolide-lincosamide-streptogramin B (MLS) resistance. Serial dilutions of cultures for spot-plating were made in Spizizen minimal salts medium [73]. Viability assays B. subtilis strains were streaked from -80°C freezer stocks on LB plates and grown overnight at 37°C. Single colonies were transferred to S750, and dilutions of those were grown overnight with vigorous shaking at 37°C. Starter cultures between OD600 0.05–0.5 were diluted to OD600 0.025, grown for at least three generations, and adjusted to OD600 0.1. recA mutants were grown in the dark (ΔrecA mutants, and many other mutants altered in the DNA damage response, are sensitive to ambient UV light). HPUra (6-(p-hydroxyphenylazo)-uracil; [11]), 38 μg/ml, or mitomycin C (Sigma-Aldrich), 0.33 μg/ml, was added as indicated, and cultures were kept at 37°C with shaking for 3 h. Viability was assessed before and at various times after addition of HPUra or MMC by making 10-fold serial dilutions from 100 μl of cultures in a 96-well plate and spotting 10 μl on LB plates, in triplicate. Results are in percentage of colony forming units present after 3 h treatment in relation to immediately before addition of HPUra or MMC. Temperature sensitive mutants with the dnaCts (replicative helicase) allele [21] were grown at 30°C. At OD600 0.1, the cultures were shifted to 49°C, the non-permissive temperature, to arrest DNA replication. Dilutions of the cultures were plated before and every hour after the shift to assess survival. Results are percentages of colony forming units (at 30°C) present after 4 h in relation to immediately before the temperature shift. Statistical comparison between different groups was performed in R, using the varequal and t.test functions. Two-tailed, paired t-tests were computed using a pooled estimate of variances for similarly distributed samples or were estimated separately for both groups and the Welch modification to the degrees of freedom was used. Tn-seq screen The transposon insertion library used in this work has been described in detail [15]. The library contains ~1–2 x 105 unique transposon insertions in strain JMA222. An aliquot of the transposon insertion library was diluted in minimal medium, and grown from an OD600 of 0.02 to 0.3 in defined minimal medium at 37°C. The culture was then split and HPUra (38 μg/ml) was added to half to arrest replication, and the other half was untreated. After 1 h at 37°C, cells from both cultures were harvested by centrifugation, washed, and resuspended in fresh minimal medium, diluting the cultures back to OD600 0.15. They were allowed to recover for 4 h, and aliquots were harvested every hour during this time. DNA was prepared for sequencing as described previously [15], and sequencing was done using an Illumina HiSeq by the MIT BioMicro Center. Tn-seq data analysis Tn-seq data were initially processed as described [15]. Sequences adjacent to the transposon were mapped to JMA222 genome using Bowtie2 [74]. The resulting files contained the number of reads per genomic coordinate in each sample. Our resulting mapped libraries had approximately 105 independent insertions each, with an average of one insertion per 37 base pairs, and 19 insertions per non-essential gene. Subsequent analyses and file manipulations were performed using custom-made R scripts. Any genomic position with less than 3 reads was discarded, to avoid potential noise due to rare insertions or misaligned reads. Then, we used inter-sample quantile normalization from the preprocessCore package (available at www.bioconductor.org) to ensure that different samples were comparable. Finally, we calculated the number of reads interrupting each gene in every sample–only insertions mapping to 5–95% internal sequence were considered since insertions in the extremities of essential genes are sometimes tolerated. To assess enrichment or depletion of insertion mutants, we calculated the ratio between the number of reads in the treated and control samples. For proper analysis of Tn-seq, it is important that libraries being compared were expanded roughly the same number of generations before sequencing. Since replication and cell division in the treated library was arrested during 1 h (leaving them one generation “behind” the control libraries), we compared the HPUra-treated libraries with the controls harvested an hour earlier: the HPUra sample harvested after 4 h of recovery was compared to the control harvested after 3 h, and HPUra 2 h with control cells harvested after 1 h of recovery. Finally, for a gene to be considered to affect survival in HPUra in relation to the control, it had to satisfy the following requisites: (i) be longer than 200 bp; (ii) have more than 5 insertions in the corresponding control library; (iii) have log2 fold change > 1 at 4 h; (iv) have an amplified change in read frequency over time. These criteria aimed to restrict the number of candidate genes and decrease false-positives. The importance of additional candidate genes of interest (e.g., recO, recF, and recA) that did not meet all of these criteria were evaluated using targeted gene disruptions. ß-Galactosidase activity assay Cultures in mid-exponential phase growing in minimal medium were diluted to OD600 0.1. One-milliliter aliquots were permeabilized with 15 μl toluene and stored at -20°C. For cells undergoing replication arrest, HPUra (38 μg/ml) was added for 30 min, 1.5 ml aliquots were centrifuged at 3000 x g for 2 min and cells were resuspended in 1.5 ml of fresh medium. One milliliter aliquots were permeabilized and frozen, and the remainder was used to measure OD600 to correct for cell recovery. ß-galactosidase specific activity was determined as previously described [75, 76]. Fluorescence microscopy We used RecN-mNeongreen as a marker to measure DNA repair centers. Cells containing this construct were grown in minimal media until OD600 ~ 0.1. Cells were either untreated or treated with HPUra (38 μg/ml) for 20 min. Aliquots were then added to a tube containing 4’,6’-diamidino-2-phenylindole (DAPI, 1 μg/ml final concentration) and FM4-64 (2.5 μg/ml), to stain nucleoids and membranes, respectively, incubated for 10 min at 37°C, and then prepared for imaging. For imaging, cultures were concentrated roughly four-fold by centrifugation (2 min at 2000x g) and 2 μl cells were placed on a slice of 1.5% UltraPure agarose (Invitrogen) made with minimal medium. The agarose slice was placed, cells down, on standard coverslips and imaged on a Nikon Ti-E inverted microscope. Fluorescence was generated using excitation/emission of 500/535 nm for RecN-mNeongreen (1 ms exposure), 350/460 nm for DAPI-stained nucleoids (100 ms) and 510/630 for FM4-64-stained membranes (100 ms). Image processing was performed using Fiji [77], where foci were detected with the Laplacian of Gaussians (LoG) detector in the TrackMate plugin [78], with a 0.7 nm blob diameter and threshold between 10 and 20 (determined by the control samples). Bacterial strains Unless otherwise indicated, all B. subtilis strains used in this work derive from JMA222 [57], a version of JH642 [58] cured of the integrative and conjugative element ICEBs1. Strain genotypes and references for previously described strains are listed in Table 3. Previously described alleles were introduced into JMA222 by transforming naturally competent cells with genomic DNA from the relevant strain. These alleles include: dnaC30(ts)-mls [21], recU::cat [59], recA260 [60], walJ::erm [61], recF::spc [62], amyE::Pyne-lacZ cat [63], ponA::spc [64], and addAB::kan [65]. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 3. B. subtilis strains used. https://doi.org/10.1371/journal.pgen.1010564.t003 Introduction of new alleles was done by transforming cells with linear DNA products generated either by traditional cloning or by isothermal assembly [69]. Insertion-deletion constructs contained antibiotic resistance cassettes typically flanked by 800–1000 bp of genomic sequences upstream and downstream of region to be deleted. These alleles include addA::spc, recJ::spc, recN::kan, recO::cat, recF::spc, ycgE::spc, hrcA::cat, polA::cat and cwlO::spc. To generate deletions of SPß (LSF203) and PBSX (LSF204), each was substituted by a kanamycin resistance gene flanked by loxP sites, generating strains LSF197 (ΔSPß::lox-kan-lox) and LSF198 (ΔPBSX::lox-kan-lox). The cassettes were removed by the Cre recombinase (from bacteriophage P1) using the cre expression vector pDR244 as previously described [70], leaving a 75 bp insertion (’scar’ containing a single lox site). A strain devoid of both phages (LSF225) resulted from transforming LSF203 (ΔSPß::lox) with genomic DNA from LSF201 (a ΔPBSX::lox-kan-lox strain), and recombining the antibiotic cassette out. The ΔPBSX allele also removes spoIISABC (toxin-antitoxin-antitoxin) which is just downstream of PBSX but is not normally thought to be part of it [71]. To visualize DNA repair centers, we constructed a strain with a recN-mNeongreen (recN-mng) fusion, inserted in the native locus such that the fusion is the only recN allele. The full linear DNA product contained the 800 bp at the 3’ end of recN, an in-frame linker (5’- CTCGAGGGATCTGGCCAAGGAAGCGGC-3’; encoding 9 amino acids), the mNeongreen coding sequence (a gift from Ethan Garner), a kanamycin resistance cassette, and 800 bp genomic sequence downstream of the recN stop codon. Transformation of this construct into JMA222 generated strain LSF284 that contains recN::[recN-9aa-mNeongreen kan]; referred to as recN-mng kan. The recN-mng kan allele from LSF284 was then introduced into various strains, including LSF225 (generating LSF708). Media and antibiotics All experiments were performed in defined minimal medium containing 50 mM MOPS (S750) and supplemented with 1% glucose, 0.1% glutamate, 40 μg/ml phenylalanine and 40 μg/ml tryptophan [72]. When required, the following concentration of antibiotics were used: 5 mg/ml chloramphenicol, 100 μg/ml spectinomycin, 5 μg/ml kanamycin, and 0.5μg/ml erthyromycin plus 12.5 μg/ml lincomycin to select for macrolide-lincosamide-streptogramin B (MLS) resistance. Serial dilutions of cultures for spot-plating were made in Spizizen minimal salts medium [73]. Viability assays B. subtilis strains were streaked from -80°C freezer stocks on LB plates and grown overnight at 37°C. Single colonies were transferred to S750, and dilutions of those were grown overnight with vigorous shaking at 37°C. Starter cultures between OD600 0.05–0.5 were diluted to OD600 0.025, grown for at least three generations, and adjusted to OD600 0.1. recA mutants were grown in the dark (ΔrecA mutants, and many other mutants altered in the DNA damage response, are sensitive to ambient UV light). HPUra (6-(p-hydroxyphenylazo)-uracil; [11]), 38 μg/ml, or mitomycin C (Sigma-Aldrich), 0.33 μg/ml, was added as indicated, and cultures were kept at 37°C with shaking for 3 h. Viability was assessed before and at various times after addition of HPUra or MMC by making 10-fold serial dilutions from 100 μl of cultures in a 96-well plate and spotting 10 μl on LB plates, in triplicate. Results are in percentage of colony forming units present after 3 h treatment in relation to immediately before addition of HPUra or MMC. Temperature sensitive mutants with the dnaCts (replicative helicase) allele [21] were grown at 30°C. At OD600 0.1, the cultures were shifted to 49°C, the non-permissive temperature, to arrest DNA replication. Dilutions of the cultures were plated before and every hour after the shift to assess survival. Results are percentages of colony forming units (at 30°C) present after 4 h in relation to immediately before the temperature shift. Statistical comparison between different groups was performed in R, using the varequal and t.test functions. Two-tailed, paired t-tests were computed using a pooled estimate of variances for similarly distributed samples or were estimated separately for both groups and the Welch modification to the degrees of freedom was used. Tn-seq screen The transposon insertion library used in this work has been described in detail [15]. The library contains ~1–2 x 105 unique transposon insertions in strain JMA222. An aliquot of the transposon insertion library was diluted in minimal medium, and grown from an OD600 of 0.02 to 0.3 in defined minimal medium at 37°C. The culture was then split and HPUra (38 μg/ml) was added to half to arrest replication, and the other half was untreated. After 1 h at 37°C, cells from both cultures were harvested by centrifugation, washed, and resuspended in fresh minimal medium, diluting the cultures back to OD600 0.15. They were allowed to recover for 4 h, and aliquots were harvested every hour during this time. DNA was prepared for sequencing as described previously [15], and sequencing was done using an Illumina HiSeq by the MIT BioMicro Center. Tn-seq data analysis Tn-seq data were initially processed as described [15]. Sequences adjacent to the transposon were mapped to JMA222 genome using Bowtie2 [74]. The resulting files contained the number of reads per genomic coordinate in each sample. Our resulting mapped libraries had approximately 105 independent insertions each, with an average of one insertion per 37 base pairs, and 19 insertions per non-essential gene. Subsequent analyses and file manipulations were performed using custom-made R scripts. Any genomic position with less than 3 reads was discarded, to avoid potential noise due to rare insertions or misaligned reads. Then, we used inter-sample quantile normalization from the preprocessCore package (available at www.bioconductor.org) to ensure that different samples were comparable. Finally, we calculated the number of reads interrupting each gene in every sample–only insertions mapping to 5–95% internal sequence were considered since insertions in the extremities of essential genes are sometimes tolerated. To assess enrichment or depletion of insertion mutants, we calculated the ratio between the number of reads in the treated and control samples. For proper analysis of Tn-seq, it is important that libraries being compared were expanded roughly the same number of generations before sequencing. Since replication and cell division in the treated library was arrested during 1 h (leaving them one generation “behind” the control libraries), we compared the HPUra-treated libraries with the controls harvested an hour earlier: the HPUra sample harvested after 4 h of recovery was compared to the control harvested after 3 h, and HPUra 2 h with control cells harvested after 1 h of recovery. Finally, for a gene to be considered to affect survival in HPUra in relation to the control, it had to satisfy the following requisites: (i) be longer than 200 bp; (ii) have more than 5 insertions in the corresponding control library; (iii) have log2 fold change > 1 at 4 h; (iv) have an amplified change in read frequency over time. These criteria aimed to restrict the number of candidate genes and decrease false-positives. The importance of additional candidate genes of interest (e.g., recO, recF, and recA) that did not meet all of these criteria were evaluated using targeted gene disruptions. ß-Galactosidase activity assay Cultures in mid-exponential phase growing in minimal medium were diluted to OD600 0.1. One-milliliter aliquots were permeabilized with 15 μl toluene and stored at -20°C. For cells undergoing replication arrest, HPUra (38 μg/ml) was added for 30 min, 1.5 ml aliquots were centrifuged at 3000 x g for 2 min and cells were resuspended in 1.5 ml of fresh medium. One milliliter aliquots were permeabilized and frozen, and the remainder was used to measure OD600 to correct for cell recovery. ß-galactosidase specific activity was determined as previously described [75, 76]. Fluorescence microscopy We used RecN-mNeongreen as a marker to measure DNA repair centers. Cells containing this construct were grown in minimal media until OD600 ~ 0.1. Cells were either untreated or treated with HPUra (38 μg/ml) for 20 min. Aliquots were then added to a tube containing 4’,6’-diamidino-2-phenylindole (DAPI, 1 μg/ml final concentration) and FM4-64 (2.5 μg/ml), to stain nucleoids and membranes, respectively, incubated for 10 min at 37°C, and then prepared for imaging. For imaging, cultures were concentrated roughly four-fold by centrifugation (2 min at 2000x g) and 2 μl cells were placed on a slice of 1.5% UltraPure agarose (Invitrogen) made with minimal medium. The agarose slice was placed, cells down, on standard coverslips and imaged on a Nikon Ti-E inverted microscope. Fluorescence was generated using excitation/emission of 500/535 nm for RecN-mNeongreen (1 ms exposure), 350/460 nm for DAPI-stained nucleoids (100 ms) and 510/630 for FM4-64-stained membranes (100 ms). Image processing was performed using Fiji [77], where foci were detected with the Laplacian of Gaussians (LoG) detector in the TrackMate plugin [78], with a 0.7 nm blob diameter and threshold between 10 and 20 (determined by the control samples). Supporting information S1 Table. Data from Tn-seq analysis with the number of insertions and read frequency of transposon insertions in each gene with (indicated as ’treated’) and without (indicated as ’control’) treatment with HPUra. Samples were collected and analyzed 2, 3, and 4 hours after treatment, or from parallel untreated cultures. The sequencing data used for these analyses have been deposited in the NCBI Gene Expression Omnibus [79] and are accessible through GEO Series accession number GSE221151. https://doi.org/10.1371/journal.pgen.1010564.s001 (XLSX) S1 Data. Underlying raw data for experiments presented. The excel spreadsheet contains the underlying data for the experiments presented in each of the figures and Table 2. https://doi.org/10.1371/journal.pgen.1010564.s002 (XLSX) Acknowledgments We thank Ethan Garner and Alexandre Bisson-Filho for providing the mNeongreen allele and initial help with microscopy, Juan Alonso for helpful comments on the manuscript, and Lyle Simmons for providing HPUra.

Crucianelli, Carlo;Jaiswal, Janhvi;Maya, Ananthakrishnan Vijayakumar;Nogay, Liyne;Cosolo, Andrea;Grass, Isabelle;Classen, Anne-Kathrin

doi: 10.1371/journal.pgen.1010516pmid: 36520882

Introduction Tissue regeneration in many systems relies on the induction of cell proliferation in stem cells [1] or other tissue-resident cell types [2] to restore the damaged tissue. The signaling pathways that regulated regenerative proliferation have been extensively explored [3, 4]. Yet how different types of tissue damage may activate distinct signaling pathways and how these different signals converge on regenerative proliferation is less well defined. Epithelia are tissues with high regenerative capacity. Epithelia cover the surfaces of many organs and execute one core function: to act as barrier between internal and external environments [5]. This essential function may be challenged by different types of damage, such as toxins, wounding and infection. Mild insults may just induce elevated cell death. Cell death in epithelia is highly regulated to maintain junctional integrity and barrier function, even at decreasing cell density [6–8]. Thus, regenerative proliferation needs to respond to a geometrically altered environment (reduced cell density) which still performs normal epithelial functions. However, if the rate of cell death exceeds regenerative proliferation, barrier function breaks down [9]. Similarly, physical wounding or pathological processes strongly disrupt epithelial barrier integrity. In response, inflammation and associated cellular responses, which strongly alter cell behavior, are activated. These include upregulation of cytokine signaling, ROS production, migratory behaviors and even senescence, all geared towards preventing infection and efficiently closing the wound to restore the barrier [10–12]. Whether these different inflammatory and non-inflammatory scenarios activate distinct signaling pathways to drive regenerative proliferation and whether both types of tissue disruption target the same proliferative program is less well understood. Many tissues that undergo regenerative proliferation increase the number of proliferating cells. For example, quiescent stem cells which reenter the cell cycle can support tissue repair [13, 14]. Yet other models increase cell numbers by stimulating cell cycle acceleration [15, 16]. Importantly, both strategies may co-exist. However, each must be controlled by distinct mechanisms. Cell cycle re-entry necessitates control at the G1/S transition, whereas cell cycle acceleration must control progression through G1, S, G2 and M-phases individually [17–20]. Cell cycle acceleration has been attributed to mitogenic signals driving gap phase dynamics thereby also allowing more frequent S-phase entry [21, 22]. Acceleration of S-phases or M-phases themselves have rarely been described in regeneration [23, 24]. While S-phase length is emerging as a novel regulator of cell fate decisions [25–28], acceleration of DNA replication must be tightly controlled to prevent replication stress. Not surprisingly, replication stress can drive diseases, such as cancer [29–32]. To better understand how the type of tissue damage, proliferative signals and cell cycle controls are integrated during regeneration, we chose to investigate compensatory proliferation in Drosophila imaginal discs [33–35]. In imaginal discs, compensatory proliferation is mediated by resident cells near the site of damage [15, 33, 34, 36] and is regulated by the conserved TNFα/JNK/AP-1, Cytokine/JAK/STAT, EGF/ERK, Myc and Hippo/Yki signaling pathways [11, 37–39]. However, like in other regeneration models, little is known about how these signals may be adapted to different types of epithelial damage and if these very different signaling pathways converge on the same cell cycle program to drive compensatory proliferation. Previous studies explored cell cycle alteration during regenerative proliferation in imaginal discs [40], but the unexpected complexity of spatial organization of signaling and cell cycle patterns in damaged discs [35, 41] invited a renewed analysis of this question. Results Compensatory proliferation in hid-expressing discs is associated with short G1, G2 and S-phases To understand how the cell cycle may be adapted during regeneration, we examined two very distinct models of wing disc damage [33–35]. Briefly, we induced apoptosis in the wing pouch by expression of the pro-apoptotic transgenes head involution defective (hid) or eiger (egr) under the control of rn(rotund)-GAL4 and a temperature-sensitive GAL80ts for 24 h during third instar stages (see Fig 1H). Both cell ablation systems have been previously demonstrated to undergo compensatory proliferation to regenerate the damaged disc [33–35]. Importantly, regenerative responses can be detected as early as 7–8 h after hid or egr-expression is initiated (S1A–S1D Fig) and continue well into subsequent recovery periods [41, 42]. An analysis of the imaginal discs directly after 24 h of hid or egr-expression therefore captures both the characteristics of the tissue damage, as well as the immediate regenerative responses. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Compensatory proliferation in hid-expressing discs is associated with short G1, G2 and S-phases. (A,B) Control wing disc (A) and wing disc after 24 h of hid-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei (A,B) and cleaved Dcp1 (A’,B’), a marker of apoptosis. (C,D) Control wing disc (C) and wing disc after 24 h of hid-expression in the pouch domain (D). Discs were stained for E-Cadherin to label adherens junctions. (E,F) Control wing disc (E,E’) and wing disc after 18 h of hid-expression in the pouch domain and 6 h into the recovery period (F,F’). Discs express two ‘flip-out’ construct to generate labelled clones, either controlling expression of GFP (green) or of Lac-Z (red). As both constructs are induced independently, clones either express GFP (green), LacZ (red) or both (yellow). Magnified views of pouch domain shown in (E’,F’). (G) Quantification of number of cells per clone expressing either GFP, LacZ or both in the pouch domain. Mean and 95% confidence interval (CI) are shown, Welch’s test was performed to test for statistical significance. (G) GFP clones (green), WT, n = 84 clones and Hid, n = 70 clones, **** <0.0001. (G’) Lac-Z clones (red), WT, n = 83 clones and Hid, n = 50 clones, **p = 0.0038. (G’) GFP and Lac-Z clones (yellow), WT, n = 75 clones and Hid, n = 43 clones, **** <0.0001. (H) Schematic representation of the imaginal wing disc. The rotund-GAL4 expressing domain is indicated in grey and a blue dotted line. Characteristic folds in the pouch, hinge and notum are represented by red dotted lines. (I,J) Control wing disc (I) and wing disc after 24 h of hid-expression in the pouch domain (J). Discs were stained with DAPI to visualize nuclei (I, J) and express the FUCCI reporter system of ubi-GFP-E2f11-230 (green in overlay) and ubi-mRFP-NLS-CycB1-266 (red in overlay). Cells in G1 express GFP, cells in early S-phase lack expression of both FUCCI constructs, cells in late S-phase express RFP, cells in G2 express both reporters. See also S1H–S1J Fig for characterization. (K-N) Quantification of FUCCI profiles to determine cell cycle phase distribution for each genotype. G1 (K), Early S (L), Late S (M), G2 (N). n = 14 discs for each genotype, Welch’s test was performed to test for statistical significance. Quantifications were performed in lateral sections, thereby omitting apical mitotic cells (M-phase) from the analysis. (O) Flow cytometry analysis of DNA content in the pouch of undamaged control wing discs (grey) and wing disc after 24 h of hid-expression (green). The pouch of the wing disc was labeled by rnGAL4-driven expression of UAS-GFP and only GFP-positive flow cytometry events were plotted as counts scaled to mode against fluorescence intensity of the DNA stain Hoechst. GFP-negative events outside the pouch domain are plotted in S1K Fig. (P) Schematic representation of relative cell cycle length and cell cycle phase distribution in undamaged control tissues and in tissues undergoing compensatory proliferation after 24 h of hid-expression. (Q,R) Control wing disc (Q) and wing disc after 24 h of hid-expression in the pouch domain (R) were assessed for DNA replication activity by EdU incorporation. (S) Schematic representation of wing disc tissue; white nuclei do not incorporate EdU, red nuclei incorporate EdU, different shades of red visualize intensity of EdU incorporation. (T) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that are positive for EdU incorporation. This serves as a proxy for the number of nuclei undergoing DNA replication. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Hid, n = 10 discs, ****p = <0.0001). (T’) Quantification of incorporated EdU intensity in the pouch of the wing disc, measured as the mean EdU intensity within the EdU area of the pouch. This serves as a proxy for the speed of nucleotide incorporation during S-phase. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Hid, n = 10 discs, ****p = <0.0001). (U-X) Control wing disc (U,W) and wing disc after 24 h of hid-expression in the pouch domain (V,X) were assessed for DNA replication activity by allowing larvae to feed on EdU for 2 h (U,V) and 18 h (W,X) into the recovery period. Imaging conditions were adjusted for each timepoint individually. (Y) Schematic representation of wing disc tissue visualizing localization of compensatory proliferation. Graphs display mean and 95% confidence interval (CI). Maximum projections of multiple confocal sections are shown in (A,B,I,J,Q,R,U,V,W,X); single sections are shown in (E,F); Local Z Projector was used to generate (C,D). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g001 In hid-expressing disc, the IAP-inhibitor Hid induces cell death by directly activating caspases [43] and results in controlled delamination of apoptotic cells (Fig 1A and 1B). While this causes a reduction in cell density and tissue size, this process maintains junctional integrity, epithelial barrier and planar disc morphology intact (Fig 1C and 1D). Consequently, only low activity of the central epithelial stress response pathway JNK are observed in hid-expressing disc, which is normally robustly activated by loss of epithelial polarity and integrity (Fig 3A and 3B) [6]. Thus, hid-driven cell ablation transiently challenges epithelial homeostasis by elevating the rate of cell death, and therefore models an environment of non-inflammatory, accelerated tissue turn-over where epithelial barrier function is maintained. Previous studies demonstrated that a local increase in mitotic divisions drives compensatory proliferation in hid-expressing discs. Specifically, clones within the pouch and proximal hinge grow to larger sizes than in the disc periphery [34]. We confirmed these observations by comparing the size of TIE-DYE clones marked independently by expression of LacZ or GFP in control and hid-expressing discs (Fig 1E–1G) [44]. Control clones in the pouch usually contained 2.3 cells indicating that they had divided 1.2 times since clone induction 24 h prior, placing the length of one cell cycle at 20.2 h. Clones in the hid-expressing pouch contained on average 4.9 cells, indicating that they had divided 2.3 times in the same span of time, placing the length of one cell cycle at 10.6 h. The real cell cycle length in hid-expressing discs is likely shorter than this estimate, as cell survival and clone growth is limited by hid-induced apoptosis, which occur at the same time as clone growth. Of note, clone sizes in more peripheral tissues, such as the notum, are comparable between control and ablated discs, demonstrating that compensatory proliferation in hid-expressing disc is regulated by local signals at the site of hid-induced cell death (S1E–S1G Fig). To better characterize the cellular program of regenerative proliferation, we analyzed the cell cycle in hid-expressing discs. We utilized the FUCCI cell cycle reporter, EdU incorporation assays and flow cytometry to specifically describe G1, G2 and S-phase dynamics. The FUCCI cell cycle reporter expresses GFP- and RFP-tagged peptides of the cell cycle genes E2f1 and Cyclin B, which are degraded in a cell cycle-dependent manner [45]. Correlating GFP and RFP levels with EdU incorporation patterns allowed us to establish a precise FUCCI read-out for G1, early S-phase, late S-phase and G2 cells in our hands (S1H–S1J Fig). We then asked if the cell cycle changed during compensatory proliferation and analyzed basolateral tissue sections, which omitted apically localize M-phase cells from the assay (see also S1I Fig). Of note, M-phase is a relatively low frequency event, which reflects the short time cells spend in mitosis, and thus not central to our analysis (see [42] and also S2A Fig). Strikingly, cells in the pouch domain of hid-expressing discs displayed a gap-phase profile different from control discs (Fig 1I and 1J). The proportion of cells in G1 and G2 was significantly reduced. In contrast, the proportion of cells in early and late S-phase was strongly increased (Fig 1K–1N). Indeed, flow cytometry of UAS-GFP-labelled cells from the rn-GAL4, hid-expressing domain confirmed a dramatic shift towards a S-phase dominated cell cycle profile, which was absent in cells outside the rn-GAL4 domain (Figs 1O and S1K). These observations are further supported by high nuclear area fractions of EdU incorporation in the pouch domain, indicative of more DNA-replicating cells (Fig 1Q–1T). Combining the clone growth and cell cycle analysis in hid-expressing discs reveals that a very short cell cycle drives compensatory proliferation when the wing disc is challenged by massive cell death. While our conclusion on reduced length of the cell cycle is in agreement with previous reports [15, 33, 34, 46], we specifically demonstrate that the cell cycle is characterized by dramatic gap phase shortening. Moreover, even though the proportion of cells in early and late S-phase is high in regenerating discs, the overall length of S-phase must also be shortened to match the extent of cell cycle acceleration (Fig 1P). Our finding that S-phase is shortened in compensatory proliferation was unexpected. A shortened S-phase would require accelerated DNA replication to replicate the genome. Indeed, compensatory proliferation in hid-expressing discs was not just characterized by a high nuclear area fraction of EdU incorporation, reflecting more DNA-replicating cells. In replicating cells, EdU intensities were strongly elevated, indicating that EdU was incorporated at higher-than-normal rates into the DNA of these cells (Fig 1T’). Importantly, high rates of EdU incorporation were not caused by endoreplication, an alternative regenerative strategy reported for other tissues in vivo (S1L Fig) [47–50]. Elevated EdU incorporation was also not an artifact of locally altered EdU uptake due to disturbed epithelial barrier function. Neither disruption of the epithelial barrier by knock-down of the septate junction protein Cora, nor disruption of the basement membrane by targeted expression of MMP1 or MMP2, altered nucleotide incorporation in control discs (S1M–S1Q Fig). To provide further evidence that DNA replication and thus S-phase is accelerated and that endoreplication does not occur, we tested if EdU incorporation in control and hid-expressing tissues is ultimately reaching the same saturation levels after one round of DNA replication. We thus fed EdU to larvae for 2 h or 18 h during regeneration. After 2 h of EdU incorporation in vivo, we observed the expected differences in both area and intensity of EdU incorporation between control and hid-expressing discs, confirming that cells enter S-phase frequently and undergo accelerated DNA replication in hid-expressing discs (Fig 1U and 1V). However, after 18 h of EdU incorporation, EdU was incorporated equally across control and hid-expressing wing discs suggesting that cells had gone through S-phase at least once and that EdU incorporation was saturated at comparable levels (Fig 1W and 1X). Combined, our data demonstrate that surviving cells inside the hid-expressing domain undergo compensatory proliferation, and that the short compensatory cell cycle is characterized by short gap phases, and importantly, by a short S-phase facilitated by accelerated DNA replication (Fig 1P and 1Y). Non-autonomous proliferation in egr-expressing discs is also associated with S-phase acceleration To understand if accelerated DNA replication was generally associated with compensatory proliferation, we also analyzed egr-expressing discs. In contrast to hid-expressing discs, expression of the TNFα-homologue Eiger strongly activates the epithelial stress response pathway JNK via receptor-mediated signaling [51]. This drives apoptosis, but also disrupts overall tissue architecture, junctional integrity and epithelial polarity (Fig 2A–2G, also compare Fig 3A–3C) [33, 35]. Due to the high activation of a JNK-dependent stress response program, egr-expression reproduces many hallmarks of highly inflammatory wounds [35, 41, 52]. Nevertheless, egr-expressing discs undergo compensatory proliferation to regenerate the damage [33–35]. Compensatory proliferation, however, straddles the high JNK-signaling domain, which is created by rn-GAL4-driven egr-expression in the pouch, and which cell-autonomously represses proliferation by inducing G2-arrested cells [35]. Thus, JNK-signaling cells lack EdU incorporation and the mitotic marker phospho-histone 3 (Figs 2G and S2A). Yet, in the pouch periphery, larger clone sizes can be detected [33]. In EdU incorporation assay, cells with high EdU area fraction and, importantly, increased EdU intensities form a ring around the JNK-signaling domain (Fig 2I–2M). Analysis of these cells by flow cytometry is hampered by the lack of genetic labeling opportunities. However, we analyzed the FUCCI-profile in the band of cells just outside the JNK-signaling domain. This analysis confirmed the presence of highly G2-arrested cells in the JNK-signaling domain (Fig 2N and 2P). Yet, just outside the G2-shifted JNK-signaling domain, a band of cells with reduced G1 and late G2-phase markers, but elevated markers for early and late S-phase could be observed (Fig 2O and 2Q). Combined, this data suggests that the compensatory cell cycle in hid- and egr-expressing discs is characterized by a short gap phases and accelerated S-phases. This is a surprising conclusion, as DNA replication speed during S-phase may need to be restrained to prevent replicative stress, whereas gap phases may be more safely exploited to accelerate cellular growth and cell cycle progression. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Non-autonomous proliferation in egr-expressing discs is also associated with S-phase acceleration. (A,B) Control wing disc (A) and wing disc after 24 h of egr-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei (A,B) and cleaved Dcp1 (A’,B’), a marker of apoptosis. Control wing disc also shown in Fig 1A and 1A’. (C-F) Control wing disc (C,E) and wing disc after 24 h of egr-expression in the pouch domain (D,F). Discs were stained for E-Cadherin (Ecad) to label adherens junctions (C,D) and for Discs-large (Dlg) to asses apical-basal polarity (E,F). (G,H) Wing disc after 24 h of egr-expression in the pouch domain stained with DAPI to visualize nuclei (G). Discs also express the JNK activity reporter TRE-RFP (G’, red in G”‘ and H) and were assessed for DNA replication activity by EdU incorporation (G”, cyan in G”‘,H). Magnified section shown in (H). (I,J) Control wing disc (I) and wing disc after 24 h of egr-expression in the pouch domain (J). Discs were assessed for DNA replication activity by EdU incorporation (I’,J’). (K) Schematic representation of nuclei in wing disc tissue. White nuclei do not incorporate EdU, red-shaded nuclei incorporate EdU. Different shades represent intensity of detected EdU. (K’) Schematic representation of localization of compensatory proliferation in egr-expressing wing discs. (L) Quantification of the percentage of cells in the pouch domain of the wing disc that are positive for EdU incorporation, mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, egr, n = 10 discs, **p = <0.0083) (M) Quantification of incorporated EdU intensity in the pouch of the wing disc, mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Eig, n = 10 discs, ****p = <0.0001). (N-Q) Control wing disc (N,O) and wing disc after 24 h of egr-expression in the pouch domain (P,Q). Discs were stained with DAPI to visualize nuclei (N,P) for MMP-1 (a JNK target gene) to visualize JNK activity, (N‘,P‘) and express the FUCCI reporter system of ubi-GFP-E2f11-230 (green in overlay N”-O‘) and ubi-mRFP-NLS-CycB1-266 (red in overlay, P”-Q‘). Cells in G1 express GFP, cells in early S-phase lack expression of both FUCCI constructs, cells in late S-phase express RFP, cells in G2 express both reporters. See also S1H–S1J Fig for characterization. Maximum projections of multiple confocal sections are shown in (A,B,C,G); Local Z Projector was used to generate (D,E,F); single sections are shown in (K,L,N,O,P,Q). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g002 Lastly, levels of EdU incorporation in hid- and egr-expressing wing discs remained low in the notum, supporting the notion that cell cycle and S-phase acceleration are controlled by the local signaling environment of tissue damage (S2B–S2E Fig). However, in egr-expressing discs, the accelerated cell cycle is not directly localized in the domain of cell death as in hid-expressing discs, indicating that different local and non-autonomous cues may be involved in the regenerative process (compare Fig 1Y and Fig 2K’). JNK signaling cannot cell-autonomously promote cell cycle acceleration To begin to understand how the accelerated cell cycle was regulated, we asked if our observations may reflect a reversion to a more primordial cell cycle, i.e. one used during rapid growth in earlier development. We thus analyzed EdU incorporation and FUCCI profiles in wing discs throughout larval development (D5-D8 AEL). Importantly, EdU incorporation rates in early stages were comparable to those in late developmental stages, and lower than those observed in hid-expressing discs (S3A–S3E Fig). Similarly, the analysis of the FUCCI profile confirmed a developmentally regulated increase of cells in G2 which was matched by a relative decrease of G1 cells (S3F–S3J Fig). Thus, the compensatory cell cycle does not reflect early developmental features, a conclusion supported by previous studies [40]. To understand which signaling pathways may then be required to produce a compensatory cell cycle profile, we closely analyzed the signaling environment in hid- and egr-expressing domain. As accelerated DNA replication clearly defines the compensatory cell cycle, we used EdU incorporation to faithfully track proliferative domains in both systems. We first mapped activity of the most central stress coordinator JNK, which was previously shown to modulate the cell cycle cell-autonomously [35]. Based on the JNK-reporter TRE>RFP, hid-expressing discs displayed mildly elevated TRE>RFP activity in the pouch where cells undergo compensatory proliferation (Figs 3A, 3B and S3K–S3M), whereas proliferating cell in egr-expressing discs localized just outside the very high JNK-signaling domain (Fig 2G). Thus, low levels of JNK can be detected in proliferating cells of both models. We thus asked if very low levels of JNK may somehow cell-autonomously support progression through a compensatory cell cycle. We therefore tested if independently activating JNK at mild levels was sufficient to promote EdU incorporation. A brief knock-down of the negative JNK regulator puckered [53] in wing discs caused low levels of JNK-associated cell death (S3N and S3O Fig). Yet, these discs did not exhibit elevated proliferation nor EdU incorporation (Fig 3D–3F). Conversely, we found that a hid-expressing wing disc hemizygous for the hepR75 JNKK-allele [35] did not display any changes to EdU incorporation patterns in the pouch (Fig 3G–3I). Similarly, expression of a dominant-negative JNK (bskDN) [35] in wild type wing discs or in hid-expressing discs did not alter EdU incorporation dynamics (S3P–S3R Fig). Combined, these observations suggest that low JNK activity cannot cell-autonomously account for an accelerated cell cycle and S-phase profiles. This conclusion is consistent with the reported opposite role of JNK in promoting cell cycle stalling and even arrest in the G2-phase [35]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. JNK signaling cannot cell-autonomously promote cell cycle acceleration. (A-C) Control wing disc (A) and wing disc after 24 h of hid-expression (B) and after 24 h of egr-expression (C) in the pouch domain. Discs were stained with DAPI to visualize nuclei (A-C). JNK activity is detected by activation of the TRE-RFP reporter (A’-C’). (D,E) Control wing disc (D) and wing disc after 12 h of puc-RNAi expression in the pouch domain (E). Discs were stained with DAPI to visualize nuclei (D,E). Discs were assessed for DNA replication activity by EdU incorporation (D’,E’). (F) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that are positive for EdU incorporation. This serves as a proxy for the number of nuclei undergoing DNA replication. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 6 discs, puc-RNAi 12h, n = 6 discs, p = 0.7581). (F’) Quantification of incorporated EdU intensity in the pouch of the wing disc, measured as the mean EdU intensity within the EdU area of the pouch. This serves as a proxy for the speed of nucleotide incorporation during S-phase. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 6 discs, puc-RNAi 12h, n = 6 discs, p = 0.9774). (G-I) Control wing disc (G) and wing disc after 24 h of hid-expression in the pouch domain (H), or a hid-expressing disc hemizygous for the hypomorphic hepR75 allele (I). Discs were stained with DAPI to visualize nuclei (G-I). Discs were assessed for DNA replication activity by EdU incorporation (G’-I’). Single sections are shown in all figure panels. Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g003 Yorkie activity and ERK signaling are elevated in proliferating cells of hid-expressing discs To understand which signaling pathways may then be required to produce a compensatory cell cycle profile, we closely analyzed the signaling environment in hid-expressing disc. We focused on pathways known to promote proliferation during tissue regeneration, specifically the growth-promoting and pro-survival pathways Hippo/Yki, Ras/ERK, JAK/STAT and Myc, predicting that the regulation of these pathways would positively correlate with high EdU intensity in hid- or egr-expressing discs. We first monitored signaling through the Hippo/Yki pathway by nuclear localization of the effector Yorkie (Yki) [54]. Strikingly, Yki distinctly localized to nuclei in proliferating cells in the hid-expressing pouch, but not in normally cycling cells in the disc periphery (Figs 4A, 4C, 4D, S4A and S4B). Similarly, when we monitored signaling through the ERK pathway using the miniCic reporter system [55], we found that ERK signaling was specifically elevated in proliferating cells of hid-expressing discs (Figs 4B, 4E, 4F, S4C and S4D). Utilizing a reporter for activated STAT [56], we found that proliferating cells in hid-expressing discs did not activate JAK/STAT signaling (Figs 4G, 4H, S4E and S4F). Similarly, only cells of the anterior compartment maintained an ancestral Myc expression pattern also observed in undamaged control discs (Fig 4I and 4J). We conclude that Myc is not upregulated de novo or expressed in all proliferating cells of hid-expressing discs. Combined, this systematic analysis revealed that compensatory proliferation in hid-expressing disc highly correlates with nuclear localization of Yki and elevated ERK activity. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Yorkie activity and ERK signaling are elevated in proliferating cells of hid-expressing discs. (A) Schematic representation of nuclear shuttling of Yki-GFP. High levels of nuclear Yki-GFP represent Yki-activation. (B) Schematic representation of nuclear shuttling of the miniCic-mCherry reporter. Low levels of nuclear miniCic represent high ERK activity and vice versa. (C-F) Control wing disc (C,E) and wing disc after 24 h of hid-expression in the pouch domain (D,F). Discs either express Yorkie-GFP (C, D) or the ERK reporter miniCic-mCherry (E,F). Magnified view of the pouch domain (C”-F”). Discs were stained with DAPI to visualize nuclei. (G-J) Control wing disc (G,I) and wing disc after 24 h of hid-expression in the pouch domain (H,J). Discs either express the JAK/STAT reporter 10xStat92E>dGFP (G,H) or an endogenously tagged Myc-GFP construct (I,J). Discs were stained with DAPI to visualize nuclei. Maximum projections of multiple confocal sections are shown in (G,H,I,J); single sections are shown in (C,D,E,F). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g004 JAK/STAT signaling and Myc-expression are elevated in proliferating cells of egr-expressing discs To understand if a similar signaling signature was associated with non-autonomous compensatory proliferation in egr-expressing discs, we analyzed the same reporter panel for changes in the domain of proliferating cells. Strikingly, in contrast to hid-expressing discs, Yki was not enriched in nuclei of proliferating cells but instead localized to the nuclei of cell cycle arrested, high JNK-signaling cells in egr-expressing discs, as reported before (Figs 5A, 5B, S5A and S5B) [57]. Similarly, no consistent correlation could be detected for ERK-activation in proliferating cells of egr-expressing discs (Figs 5C, 5D, S5C and S5D). However, in contrast to hid-expressing discs, domains of compensatory proliferation correlated well with activation of the JAK/STAT reporter (Figs 5E, 5F, S5E and S5F) and de novo expression of Myc in the peripheral pouch and hinge domains, a region where it is normally not expressed (Fig 5G and 5H). Importantly, the pattern of this signaling signature did not change during regeneration and could still be detected 24 h after egr-expression well into the recovery period (S5G–S5N Fig). Of note, hid-expressing discs maintained their signaling signature as well (S5O and S5P Fig). Combined, we find that JAK/STAT activity and Myc expression are specifically detected in cells undergoing compensatory proliferation that must be driven by non-autonomous signaling from egr-expressing domains. Indeed, JAK/STAT-activating, secreted ligands of the Unpaired family are expressed in JNK-signaling cells [41, 42, 52, 58–60]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. JAK/STAT signaling and Myc-expression are elevated in proliferating cells of egr-expressing discs. (A-D) Control wing disc (A,C) and wing disc after 24 h of egr-expression in the pouch domain (B,D). Discs either express Yorkie-GFP (A,B) or the ERK reporter miniCic-mCherry (C,D). Magnified view of the pouch domain. Discs were stained with DAPI to visualize nuclei (A-D). (E-H) Control wing disc (E,G) and wing disc after 24 h of egr-expression in the pouch domain (F,H). Discs either express the JAK/STAT reporter 10xStat92E>dGFP (E,F) or an endogenously tagged Myc-GFP construct (G,H). Discs were stained with DAPI to visualize nuclei (E-H). Magnified view of the pouch domain (E”-H”). Images with increased brightness show the presence of Myc-GFP in the regenerative domain (G”‘,H”‘). We suggest that the Myc-expressing cells in the anterior pouch domain of control disc are killed by egr-expression and a new expression pattern of Myc is set up de novo by tissue damage signals. Maximum projections of multiple confocal sections are shown in (E,F,G,H); single sections are shown in (A-D). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g005 As a result, this analysis left us with the surprising conclusion, that completely different signaling signatures can be associated with compensatory proliferation and specifically, with accelerated nucleotide incorporation and thus DNA replication speed. These results suggest that at least two distinct regulatory circuits may converged on compensatory proliferation and the same cell cycle adaptation upon distinct damaging challenges. Yki and EGF cooperate to drive compensatory proliferation in response to non-inflammatory damage To investigate which of these signaling pathways may truly be required for compensatory proliferation, we systematically analyzed sufficiency and necessity of Hippo/Yki and Ras/ERK signaling in hid-expressing disc. We first asked if Hippo/Yki or Ras/ERK activation alone were sufficient to induce accelerated EdU incorporation. However, neither expression of a phospho-ablative YkiS168A construct nor RNAi-mediated knock-down of Warts altered the rate of EdU incorporation in mosaic clones, or upon expression in the pouch (Figs 6A–6C, S6A and S6B). Similarly, expression of oncogenic RasV12 alone failed to phenocopy an accelerated S-phase profile (Fig 6D–6F). However, to understand if Hippo/Yki and Ras/ERK are necessary for S-phase acceleration, we created hid-expressing discs heterozygous mutant for a null allele of ykiB5. Indeed, in the very rare discs that we were able to recover due to high lethality, we observed a reduction in EdU incorporation, if compared to control discs (Figs 6G, 6H, S6C and S6D). This suggests, that Hippo/Yki is necessary to drive nucleotide incorporation during S-phase in hid-expressing discs. We performed experiments to test the necessity of Ras/ERK signaling in S-phase acceleration. We analyzed discs that either co-expressed a dominant-negative Egfr (EgfrDN) in hid-expressing cells (S6E–S6H Fig) or that were heterozygous for Ras1 (S6I–S6N Fig). Both strategies failed to reveal changes to EdU incorporation dynamics. However, it is possible that EgfrDN-expressing cells die too quickly in the context of hid-coexpression, and that Ras1 heterozygosity may not sufficiently interfere with ERK function. Thus, other genetic strategies may be needed to perform these experiments. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Yki and ERK cooperate to drive compensatory proliferation in response to non-inflammatory damage. (A) A wing disc expressing the act-GAL4 ‘flip-out’ system controlling the mosaic expression of GFP and UAS-yki.S168A (green in A”‘). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (magenta). (B, C) Control wing disc (B), wing disc after 24 h of UAS-yki expression in the pouch domain (C). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (D) A wing disc expressing the act-GAL4 ‘flip-out’ system controlling the mosaic expression of GFP and UAS-RasV12 (green in D”‘). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (magenta). (E,F) Control wing disc (E) and wing disc after 24 h of UAS-RasV12 expression in the pouch domain (F). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (G,H) Wing disc after 24 h of hid-expression (G) and a wing disc heterozygous for yki B5 after 24 h of hid-expression (H). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. White frame marks the magnified view of the pouch domain shown in (G”-H”‘). (I,J) Control wing disc (I) and a wing disc after 24 h of UAS-yki-GFP and UAS-RasV12 expression in the pouch domain (J). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (K) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that was positive for incorporated EdU in control wing discs, or Yki and RasV12 expressing wing discs. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 8 discs, UAS-yki-GFP, UAS-RasV12, n = 7 discs, **p<0.01). (L) Quantification of incorporated EdU intensity, measured as the mean EdU intensity within the EdU area of the pouch in control wing discs and wing disc after 24 h of Yki and RasV12 expression. A Welch’s test was performed to test for statistical significance. (WT, n = 8 discs, UAS-yki-GFP, UAS-RasV12, n = 7 discs, ****p = <0.0001). (M,N) Control wing disc (M) and a wing disc after 24 h of UAS-yki-GFP and UAS-RasV12 expression in the pouch domain (N). Discs were stained for cleaved Dcp1, a marker of apoptosis. Graphs display mean and 95% CI. Maximum projections of multiple confocal sections are shown in (G,H,I,J). Single sections are shown in (A-F). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g006 Importantly, though, as neither pathway was sufficient to accelerate nucleotide incorporation individually, we tested if Hippo/Yki and Ras/ERK pathways cooperate. Indeed, the combined expression of RasV12 and Yki was sufficient to drive elevated nucleotide incorporation in the pouch (Fig 6I–6L). Not surprisingly, the pouch overgrew, demonstrating that both pathways cooperate in promoting proliferation. This was not associated with elevated levels of apoptosis, indicating that accelerate nucleotide incorporation was directly caused by RasV12 and Yki co-expression (Fig 6M and 6N). Combined these observations demonstrate, that the hid-expressing model of local, non-inflammatory regeneration uses Hippo/Yki and Ras/ERK activation to promote cell cycle adaptations for compensatory proliferation. JAK/STAT and Myc are sufficient to drive S-phase acceleration in response to inflammatory damage Since Hippo/Yki and Ras/ERK signaling did not robustly correlate with domains of compensatory proliferation in egr-expressing disc, we asked if JAK/STAT activation and Myc expression may directly control cell cycle acceleration. We first tested if Myc and JAK/STAT alone were sufficient to induce accelerated EdU incorporation. Indeed, overexpression of Myc alone was sufficient to drive S-phase acceleration, aligning with mammalian reports that Myc can accelerate S-phase progression (Fig 7A–7C) [31]. Similarly, expression of the transcription factor Stat92E was sufficient to cell-autonomously drive high levels of EdU incorporation, confirming that JAK/STAT is a mitogenic pathway strongly implicated in compensatory proliferation (Fig 7D–7F) [39, 61, 62]. We wanted to understand, if Myc or Stat92E activity are rate-limiting for EdU incorporation. We thus generated egr-expressing discs heterozygous for the null allele Stat92E85C3. However, we failed to detect any changes in cells undergoing compensatory proliferation, suggesting that heterozygosity for Stat92E is not rate-limiting for EdU incorporation, or alternatively, that Myc upregulation can compensate for reduced Stat92E function (S7A and S7B Fig). Due to lethality of egr- and hid-expressing larvae heterozygous for dMyc alleles, we were unable to specifically test the necessity of Myc in mediating cell cycle acceleration. However, our observations suggest that activation of JAK/STAT signaling or elevated expression of Myc alone are sufficient to accelerate DNA replication during compensatory proliferation. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. JAK/STAT and Myc are sufficient to drive S-phase acceleration in response to inflammatory damage. (A,B) Control wing disc (A), wing disc after 24 h of UAS-Myc expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (C) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in control wing discs or UAS-Myc expressing wing discs. This serves as a proxy for the number of nuclei undergoing DNA replication. (C’) Quantification of incorporated EdU, measured as mean EdU intensity in the DAPI area within the pouch. A Welch’s test was performed to test for statistical significance. (WT, n = 9 discs, UAS-Myc, n = 9 discs, ****p = <0.0001). (D,E) Control wing disc (D), and a wing disc after 24 h of UAS-Stat92E-expression (E). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (F) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in control wing discs or UAS- Stat92E expressing wing discs. This serves as a proxy for the number of nuclei undergoing DNA replication. (F’) Quantification of incorporated EdU, measured as mean EdU intensity in the DAPI area within the pouch. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, UAS-Stat, n = 10 discs, ***p<0.001, ****p<0.0001). Graphs display mean and 95% CI. Single sections are shown in (A,B,D,E). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g007 Compensatory proliferation is not associated with replication stress Many reports highlight the emergence of replicative stress upon pathological acceleration of DNA replication, for example in tumors [29–32]. To understand if accelerated DNA replication during compensatory proliferation was associated with elevated replication stress, we assessed levels of DNA double-strand breaks in hid- and egr-expressing discs [63, 64]. While very occasionally apoptotic cells displayed high levels of phosphorylated H2Av staining, we failed to detect a general increase in this DNA damage marker in areas of compensatory proliferation (Fig 8A–8F). This suggests that mechanisms exist which ensure that accelerated DNA replication does not generally cause replication stress and DNA damage. It suggests that DNA-replication can be safely accelerated to increase cell cycle progression. Even though we report here that S-phases are accelerated, we also observe that gap phases nearly disappear, suggesting that, ultimately, safe DNA replication speed is still rate-limiting for cell cycle length. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 8. Compensatory proliferation is not associated with replication stress. (A-C) Control wing disc (A), and a wing disc after 24 h of hid-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA damage by staining for phosphorylated γH2A. White frame marks position of views shown in (“, “‘) panels. (C) Quantification of H2Aγ staining intensity within the DAPI area of the pouch domain normalized to DAPI intensity to correct for fluctuations in DNA density. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, Hid, n = 10 discs, ns, p = 0.2533). (D-F) Control wing disc (D), and a wing disc after 24 h of egr-expression in the pouch domain (E). Discs were stained with DAPI to visualize nuclei and were assessed for DNA damage by staining for phosphorylated γH2A. Yellow dashed line in (E) demarcates the area of high JNK reporter activity (cyan star) as assessed by TRE-RFP expression (not shown). Compensatory proliferation occurs in a band outside of the JNK-signaling domain (cyan bracket). White frame marks position of views shown in (“, “‘) panels. (F) Quantification of H2Aγ staining intensity within the DAPI area of domain outside the JNK-signaling domain normalized to DAPI intensity to correct for fluctuations in DNA density. Welch’s test was performed to test for statistical significance (WT, n = 9 discs, Egr, n = 9 discs, **p = 0.0019). (G) Model of signaling environment driving compensatory proliferation and accelerated DNA replication in response to two distinct challenges to tissue health. Maximum projections of multiple confocal sections are shown in (A,B,D,E). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g008 Compensatory proliferation in hid-expressing discs is associated with short G1, G2 and S-phases To understand how the cell cycle may be adapted during regeneration, we examined two very distinct models of wing disc damage [33–35]. Briefly, we induced apoptosis in the wing pouch by expression of the pro-apoptotic transgenes head involution defective (hid) or eiger (egr) under the control of rn(rotund)-GAL4 and a temperature-sensitive GAL80ts for 24 h during third instar stages (see Fig 1H). Both cell ablation systems have been previously demonstrated to undergo compensatory proliferation to regenerate the damaged disc [33–35]. Importantly, regenerative responses can be detected as early as 7–8 h after hid or egr-expression is initiated (S1A–S1D Fig) and continue well into subsequent recovery periods [41, 42]. An analysis of the imaginal discs directly after 24 h of hid or egr-expression therefore captures both the characteristics of the tissue damage, as well as the immediate regenerative responses. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Compensatory proliferation in hid-expressing discs is associated with short G1, G2 and S-phases. (A,B) Control wing disc (A) and wing disc after 24 h of hid-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei (A,B) and cleaved Dcp1 (A’,B’), a marker of apoptosis. (C,D) Control wing disc (C) and wing disc after 24 h of hid-expression in the pouch domain (D). Discs were stained for E-Cadherin to label adherens junctions. (E,F) Control wing disc (E,E’) and wing disc after 18 h of hid-expression in the pouch domain and 6 h into the recovery period (F,F’). Discs express two ‘flip-out’ construct to generate labelled clones, either controlling expression of GFP (green) or of Lac-Z (red). As both constructs are induced independently, clones either express GFP (green), LacZ (red) or both (yellow). Magnified views of pouch domain shown in (E’,F’). (G) Quantification of number of cells per clone expressing either GFP, LacZ or both in the pouch domain. Mean and 95% confidence interval (CI) are shown, Welch’s test was performed to test for statistical significance. (G) GFP clones (green), WT, n = 84 clones and Hid, n = 70 clones, **** <0.0001. (G’) Lac-Z clones (red), WT, n = 83 clones and Hid, n = 50 clones, **p = 0.0038. (G’) GFP and Lac-Z clones (yellow), WT, n = 75 clones and Hid, n = 43 clones, **** <0.0001. (H) Schematic representation of the imaginal wing disc. The rotund-GAL4 expressing domain is indicated in grey and a blue dotted line. Characteristic folds in the pouch, hinge and notum are represented by red dotted lines. (I,J) Control wing disc (I) and wing disc after 24 h of hid-expression in the pouch domain (J). Discs were stained with DAPI to visualize nuclei (I, J) and express the FUCCI reporter system of ubi-GFP-E2f11-230 (green in overlay) and ubi-mRFP-NLS-CycB1-266 (red in overlay). Cells in G1 express GFP, cells in early S-phase lack expression of both FUCCI constructs, cells in late S-phase express RFP, cells in G2 express both reporters. See also S1H–S1J Fig for characterization. (K-N) Quantification of FUCCI profiles to determine cell cycle phase distribution for each genotype. G1 (K), Early S (L), Late S (M), G2 (N). n = 14 discs for each genotype, Welch’s test was performed to test for statistical significance. Quantifications were performed in lateral sections, thereby omitting apical mitotic cells (M-phase) from the analysis. (O) Flow cytometry analysis of DNA content in the pouch of undamaged control wing discs (grey) and wing disc after 24 h of hid-expression (green). The pouch of the wing disc was labeled by rnGAL4-driven expression of UAS-GFP and only GFP-positive flow cytometry events were plotted as counts scaled to mode against fluorescence intensity of the DNA stain Hoechst. GFP-negative events outside the pouch domain are plotted in S1K Fig. (P) Schematic representation of relative cell cycle length and cell cycle phase distribution in undamaged control tissues and in tissues undergoing compensatory proliferation after 24 h of hid-expression. (Q,R) Control wing disc (Q) and wing disc after 24 h of hid-expression in the pouch domain (R) were assessed for DNA replication activity by EdU incorporation. (S) Schematic representation of wing disc tissue; white nuclei do not incorporate EdU, red nuclei incorporate EdU, different shades of red visualize intensity of EdU incorporation. (T) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that are positive for EdU incorporation. This serves as a proxy for the number of nuclei undergoing DNA replication. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Hid, n = 10 discs, ****p = <0.0001). (T’) Quantification of incorporated EdU intensity in the pouch of the wing disc, measured as the mean EdU intensity within the EdU area of the pouch. This serves as a proxy for the speed of nucleotide incorporation during S-phase. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Hid, n = 10 discs, ****p = <0.0001). (U-X) Control wing disc (U,W) and wing disc after 24 h of hid-expression in the pouch domain (V,X) were assessed for DNA replication activity by allowing larvae to feed on EdU for 2 h (U,V) and 18 h (W,X) into the recovery period. Imaging conditions were adjusted for each timepoint individually. (Y) Schematic representation of wing disc tissue visualizing localization of compensatory proliferation. Graphs display mean and 95% confidence interval (CI). Maximum projections of multiple confocal sections are shown in (A,B,I,J,Q,R,U,V,W,X); single sections are shown in (E,F); Local Z Projector was used to generate (C,D). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g001 In hid-expressing disc, the IAP-inhibitor Hid induces cell death by directly activating caspases [43] and results in controlled delamination of apoptotic cells (Fig 1A and 1B). While this causes a reduction in cell density and tissue size, this process maintains junctional integrity, epithelial barrier and planar disc morphology intact (Fig 1C and 1D). Consequently, only low activity of the central epithelial stress response pathway JNK are observed in hid-expressing disc, which is normally robustly activated by loss of epithelial polarity and integrity (Fig 3A and 3B) [6]. Thus, hid-driven cell ablation transiently challenges epithelial homeostasis by elevating the rate of cell death, and therefore models an environment of non-inflammatory, accelerated tissue turn-over where epithelial barrier function is maintained. Previous studies demonstrated that a local increase in mitotic divisions drives compensatory proliferation in hid-expressing discs. Specifically, clones within the pouch and proximal hinge grow to larger sizes than in the disc periphery [34]. We confirmed these observations by comparing the size of TIE-DYE clones marked independently by expression of LacZ or GFP in control and hid-expressing discs (Fig 1E–1G) [44]. Control clones in the pouch usually contained 2.3 cells indicating that they had divided 1.2 times since clone induction 24 h prior, placing the length of one cell cycle at 20.2 h. Clones in the hid-expressing pouch contained on average 4.9 cells, indicating that they had divided 2.3 times in the same span of time, placing the length of one cell cycle at 10.6 h. The real cell cycle length in hid-expressing discs is likely shorter than this estimate, as cell survival and clone growth is limited by hid-induced apoptosis, which occur at the same time as clone growth. Of note, clone sizes in more peripheral tissues, such as the notum, are comparable between control and ablated discs, demonstrating that compensatory proliferation in hid-expressing disc is regulated by local signals at the site of hid-induced cell death (S1E–S1G Fig). To better characterize the cellular program of regenerative proliferation, we analyzed the cell cycle in hid-expressing discs. We utilized the FUCCI cell cycle reporter, EdU incorporation assays and flow cytometry to specifically describe G1, G2 and S-phase dynamics. The FUCCI cell cycle reporter expresses GFP- and RFP-tagged peptides of the cell cycle genes E2f1 and Cyclin B, which are degraded in a cell cycle-dependent manner [45]. Correlating GFP and RFP levels with EdU incorporation patterns allowed us to establish a precise FUCCI read-out for G1, early S-phase, late S-phase and G2 cells in our hands (S1H–S1J Fig). We then asked if the cell cycle changed during compensatory proliferation and analyzed basolateral tissue sections, which omitted apically localize M-phase cells from the assay (see also S1I Fig). Of note, M-phase is a relatively low frequency event, which reflects the short time cells spend in mitosis, and thus not central to our analysis (see [42] and also S2A Fig). Strikingly, cells in the pouch domain of hid-expressing discs displayed a gap-phase profile different from control discs (Fig 1I and 1J). The proportion of cells in G1 and G2 was significantly reduced. In contrast, the proportion of cells in early and late S-phase was strongly increased (Fig 1K–1N). Indeed, flow cytometry of UAS-GFP-labelled cells from the rn-GAL4, hid-expressing domain confirmed a dramatic shift towards a S-phase dominated cell cycle profile, which was absent in cells outside the rn-GAL4 domain (Figs 1O and S1K). These observations are further supported by high nuclear area fractions of EdU incorporation in the pouch domain, indicative of more DNA-replicating cells (Fig 1Q–1T). Combining the clone growth and cell cycle analysis in hid-expressing discs reveals that a very short cell cycle drives compensatory proliferation when the wing disc is challenged by massive cell death. While our conclusion on reduced length of the cell cycle is in agreement with previous reports [15, 33, 34, 46], we specifically demonstrate that the cell cycle is characterized by dramatic gap phase shortening. Moreover, even though the proportion of cells in early and late S-phase is high in regenerating discs, the overall length of S-phase must also be shortened to match the extent of cell cycle acceleration (Fig 1P). Our finding that S-phase is shortened in compensatory proliferation was unexpected. A shortened S-phase would require accelerated DNA replication to replicate the genome. Indeed, compensatory proliferation in hid-expressing discs was not just characterized by a high nuclear area fraction of EdU incorporation, reflecting more DNA-replicating cells. In replicating cells, EdU intensities were strongly elevated, indicating that EdU was incorporated at higher-than-normal rates into the DNA of these cells (Fig 1T’). Importantly, high rates of EdU incorporation were not caused by endoreplication, an alternative regenerative strategy reported for other tissues in vivo (S1L Fig) [47–50]. Elevated EdU incorporation was also not an artifact of locally altered EdU uptake due to disturbed epithelial barrier function. Neither disruption of the epithelial barrier by knock-down of the septate junction protein Cora, nor disruption of the basement membrane by targeted expression of MMP1 or MMP2, altered nucleotide incorporation in control discs (S1M–S1Q Fig). To provide further evidence that DNA replication and thus S-phase is accelerated and that endoreplication does not occur, we tested if EdU incorporation in control and hid-expressing tissues is ultimately reaching the same saturation levels after one round of DNA replication. We thus fed EdU to larvae for 2 h or 18 h during regeneration. After 2 h of EdU incorporation in vivo, we observed the expected differences in both area and intensity of EdU incorporation between control and hid-expressing discs, confirming that cells enter S-phase frequently and undergo accelerated DNA replication in hid-expressing discs (Fig 1U and 1V). However, after 18 h of EdU incorporation, EdU was incorporated equally across control and hid-expressing wing discs suggesting that cells had gone through S-phase at least once and that EdU incorporation was saturated at comparable levels (Fig 1W and 1X). Combined, our data demonstrate that surviving cells inside the hid-expressing domain undergo compensatory proliferation, and that the short compensatory cell cycle is characterized by short gap phases, and importantly, by a short S-phase facilitated by accelerated DNA replication (Fig 1P and 1Y). Non-autonomous proliferation in egr-expressing discs is also associated with S-phase acceleration To understand if accelerated DNA replication was generally associated with compensatory proliferation, we also analyzed egr-expressing discs. In contrast to hid-expressing discs, expression of the TNFα-homologue Eiger strongly activates the epithelial stress response pathway JNK via receptor-mediated signaling [51]. This drives apoptosis, but also disrupts overall tissue architecture, junctional integrity and epithelial polarity (Fig 2A–2G, also compare Fig 3A–3C) [33, 35]. Due to the high activation of a JNK-dependent stress response program, egr-expression reproduces many hallmarks of highly inflammatory wounds [35, 41, 52]. Nevertheless, egr-expressing discs undergo compensatory proliferation to regenerate the damage [33–35]. Compensatory proliferation, however, straddles the high JNK-signaling domain, which is created by rn-GAL4-driven egr-expression in the pouch, and which cell-autonomously represses proliferation by inducing G2-arrested cells [35]. Thus, JNK-signaling cells lack EdU incorporation and the mitotic marker phospho-histone 3 (Figs 2G and S2A). Yet, in the pouch periphery, larger clone sizes can be detected [33]. In EdU incorporation assay, cells with high EdU area fraction and, importantly, increased EdU intensities form a ring around the JNK-signaling domain (Fig 2I–2M). Analysis of these cells by flow cytometry is hampered by the lack of genetic labeling opportunities. However, we analyzed the FUCCI-profile in the band of cells just outside the JNK-signaling domain. This analysis confirmed the presence of highly G2-arrested cells in the JNK-signaling domain (Fig 2N and 2P). Yet, just outside the G2-shifted JNK-signaling domain, a band of cells with reduced G1 and late G2-phase markers, but elevated markers for early and late S-phase could be observed (Fig 2O and 2Q). Combined, this data suggests that the compensatory cell cycle in hid- and egr-expressing discs is characterized by a short gap phases and accelerated S-phases. This is a surprising conclusion, as DNA replication speed during S-phase may need to be restrained to prevent replicative stress, whereas gap phases may be more safely exploited to accelerate cellular growth and cell cycle progression. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Non-autonomous proliferation in egr-expressing discs is also associated with S-phase acceleration. (A,B) Control wing disc (A) and wing disc after 24 h of egr-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei (A,B) and cleaved Dcp1 (A’,B’), a marker of apoptosis. Control wing disc also shown in Fig 1A and 1A’. (C-F) Control wing disc (C,E) and wing disc after 24 h of egr-expression in the pouch domain (D,F). Discs were stained for E-Cadherin (Ecad) to label adherens junctions (C,D) and for Discs-large (Dlg) to asses apical-basal polarity (E,F). (G,H) Wing disc after 24 h of egr-expression in the pouch domain stained with DAPI to visualize nuclei (G). Discs also express the JNK activity reporter TRE-RFP (G’, red in G”‘ and H) and were assessed for DNA replication activity by EdU incorporation (G”, cyan in G”‘,H). Magnified section shown in (H). (I,J) Control wing disc (I) and wing disc after 24 h of egr-expression in the pouch domain (J). Discs were assessed for DNA replication activity by EdU incorporation (I’,J’). (K) Schematic representation of nuclei in wing disc tissue. White nuclei do not incorporate EdU, red-shaded nuclei incorporate EdU. Different shades represent intensity of detected EdU. (K’) Schematic representation of localization of compensatory proliferation in egr-expressing wing discs. (L) Quantification of the percentage of cells in the pouch domain of the wing disc that are positive for EdU incorporation, mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, egr, n = 10 discs, **p = <0.0083) (M) Quantification of incorporated EdU intensity in the pouch of the wing disc, mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 10 discs, Eig, n = 10 discs, ****p = <0.0001). (N-Q) Control wing disc (N,O) and wing disc after 24 h of egr-expression in the pouch domain (P,Q). Discs were stained with DAPI to visualize nuclei (N,P) for MMP-1 (a JNK target gene) to visualize JNK activity, (N‘,P‘) and express the FUCCI reporter system of ubi-GFP-E2f11-230 (green in overlay N”-O‘) and ubi-mRFP-NLS-CycB1-266 (red in overlay, P”-Q‘). Cells in G1 express GFP, cells in early S-phase lack expression of both FUCCI constructs, cells in late S-phase express RFP, cells in G2 express both reporters. See also S1H–S1J Fig for characterization. Maximum projections of multiple confocal sections are shown in (A,B,C,G); Local Z Projector was used to generate (D,E,F); single sections are shown in (K,L,N,O,P,Q). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g002 Lastly, levels of EdU incorporation in hid- and egr-expressing wing discs remained low in the notum, supporting the notion that cell cycle and S-phase acceleration are controlled by the local signaling environment of tissue damage (S2B–S2E Fig). However, in egr-expressing discs, the accelerated cell cycle is not directly localized in the domain of cell death as in hid-expressing discs, indicating that different local and non-autonomous cues may be involved in the regenerative process (compare Fig 1Y and Fig 2K’). JNK signaling cannot cell-autonomously promote cell cycle acceleration To begin to understand how the accelerated cell cycle was regulated, we asked if our observations may reflect a reversion to a more primordial cell cycle, i.e. one used during rapid growth in earlier development. We thus analyzed EdU incorporation and FUCCI profiles in wing discs throughout larval development (D5-D8 AEL). Importantly, EdU incorporation rates in early stages were comparable to those in late developmental stages, and lower than those observed in hid-expressing discs (S3A–S3E Fig). Similarly, the analysis of the FUCCI profile confirmed a developmentally regulated increase of cells in G2 which was matched by a relative decrease of G1 cells (S3F–S3J Fig). Thus, the compensatory cell cycle does not reflect early developmental features, a conclusion supported by previous studies [40]. To understand which signaling pathways may then be required to produce a compensatory cell cycle profile, we closely analyzed the signaling environment in hid- and egr-expressing domain. As accelerated DNA replication clearly defines the compensatory cell cycle, we used EdU incorporation to faithfully track proliferative domains in both systems. We first mapped activity of the most central stress coordinator JNK, which was previously shown to modulate the cell cycle cell-autonomously [35]. Based on the JNK-reporter TRE>RFP, hid-expressing discs displayed mildly elevated TRE>RFP activity in the pouch where cells undergo compensatory proliferation (Figs 3A, 3B and S3K–S3M), whereas proliferating cell in egr-expressing discs localized just outside the very high JNK-signaling domain (Fig 2G). Thus, low levels of JNK can be detected in proliferating cells of both models. We thus asked if very low levels of JNK may somehow cell-autonomously support progression through a compensatory cell cycle. We therefore tested if independently activating JNK at mild levels was sufficient to promote EdU incorporation. A brief knock-down of the negative JNK regulator puckered [53] in wing discs caused low levels of JNK-associated cell death (S3N and S3O Fig). Yet, these discs did not exhibit elevated proliferation nor EdU incorporation (Fig 3D–3F). Conversely, we found that a hid-expressing wing disc hemizygous for the hepR75 JNKK-allele [35] did not display any changes to EdU incorporation patterns in the pouch (Fig 3G–3I). Similarly, expression of a dominant-negative JNK (bskDN) [35] in wild type wing discs or in hid-expressing discs did not alter EdU incorporation dynamics (S3P–S3R Fig). Combined, these observations suggest that low JNK activity cannot cell-autonomously account for an accelerated cell cycle and S-phase profiles. This conclusion is consistent with the reported opposite role of JNK in promoting cell cycle stalling and even arrest in the G2-phase [35]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. JNK signaling cannot cell-autonomously promote cell cycle acceleration. (A-C) Control wing disc (A) and wing disc after 24 h of hid-expression (B) and after 24 h of egr-expression (C) in the pouch domain. Discs were stained with DAPI to visualize nuclei (A-C). JNK activity is detected by activation of the TRE-RFP reporter (A’-C’). (D,E) Control wing disc (D) and wing disc after 12 h of puc-RNAi expression in the pouch domain (E). Discs were stained with DAPI to visualize nuclei (D,E). Discs were assessed for DNA replication activity by EdU incorporation (D’,E’). (F) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that are positive for EdU incorporation. This serves as a proxy for the number of nuclei undergoing DNA replication. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 6 discs, puc-RNAi 12h, n = 6 discs, p = 0.7581). (F’) Quantification of incorporated EdU intensity in the pouch of the wing disc, measured as the mean EdU intensity within the EdU area of the pouch. This serves as a proxy for the speed of nucleotide incorporation during S-phase. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance (WT, n = 6 discs, puc-RNAi 12h, n = 6 discs, p = 0.9774). (G-I) Control wing disc (G) and wing disc after 24 h of hid-expression in the pouch domain (H), or a hid-expressing disc hemizygous for the hypomorphic hepR75 allele (I). Discs were stained with DAPI to visualize nuclei (G-I). Discs were assessed for DNA replication activity by EdU incorporation (G’-I’). Single sections are shown in all figure panels. Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g003 Yorkie activity and ERK signaling are elevated in proliferating cells of hid-expressing discs To understand which signaling pathways may then be required to produce a compensatory cell cycle profile, we closely analyzed the signaling environment in hid-expressing disc. We focused on pathways known to promote proliferation during tissue regeneration, specifically the growth-promoting and pro-survival pathways Hippo/Yki, Ras/ERK, JAK/STAT and Myc, predicting that the regulation of these pathways would positively correlate with high EdU intensity in hid- or egr-expressing discs. We first monitored signaling through the Hippo/Yki pathway by nuclear localization of the effector Yorkie (Yki) [54]. Strikingly, Yki distinctly localized to nuclei in proliferating cells in the hid-expressing pouch, but not in normally cycling cells in the disc periphery (Figs 4A, 4C, 4D, S4A and S4B). Similarly, when we monitored signaling through the ERK pathway using the miniCic reporter system [55], we found that ERK signaling was specifically elevated in proliferating cells of hid-expressing discs (Figs 4B, 4E, 4F, S4C and S4D). Utilizing a reporter for activated STAT [56], we found that proliferating cells in hid-expressing discs did not activate JAK/STAT signaling (Figs 4G, 4H, S4E and S4F). Similarly, only cells of the anterior compartment maintained an ancestral Myc expression pattern also observed in undamaged control discs (Fig 4I and 4J). We conclude that Myc is not upregulated de novo or expressed in all proliferating cells of hid-expressing discs. Combined, this systematic analysis revealed that compensatory proliferation in hid-expressing disc highly correlates with nuclear localization of Yki and elevated ERK activity. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Yorkie activity and ERK signaling are elevated in proliferating cells of hid-expressing discs. (A) Schematic representation of nuclear shuttling of Yki-GFP. High levels of nuclear Yki-GFP represent Yki-activation. (B) Schematic representation of nuclear shuttling of the miniCic-mCherry reporter. Low levels of nuclear miniCic represent high ERK activity and vice versa. (C-F) Control wing disc (C,E) and wing disc after 24 h of hid-expression in the pouch domain (D,F). Discs either express Yorkie-GFP (C, D) or the ERK reporter miniCic-mCherry (E,F). Magnified view of the pouch domain (C”-F”). Discs were stained with DAPI to visualize nuclei. (G-J) Control wing disc (G,I) and wing disc after 24 h of hid-expression in the pouch domain (H,J). Discs either express the JAK/STAT reporter 10xStat92E>dGFP (G,H) or an endogenously tagged Myc-GFP construct (I,J). Discs were stained with DAPI to visualize nuclei. Maximum projections of multiple confocal sections are shown in (G,H,I,J); single sections are shown in (C,D,E,F). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g004 JAK/STAT signaling and Myc-expression are elevated in proliferating cells of egr-expressing discs To understand if a similar signaling signature was associated with non-autonomous compensatory proliferation in egr-expressing discs, we analyzed the same reporter panel for changes in the domain of proliferating cells. Strikingly, in contrast to hid-expressing discs, Yki was not enriched in nuclei of proliferating cells but instead localized to the nuclei of cell cycle arrested, high JNK-signaling cells in egr-expressing discs, as reported before (Figs 5A, 5B, S5A and S5B) [57]. Similarly, no consistent correlation could be detected for ERK-activation in proliferating cells of egr-expressing discs (Figs 5C, 5D, S5C and S5D). However, in contrast to hid-expressing discs, domains of compensatory proliferation correlated well with activation of the JAK/STAT reporter (Figs 5E, 5F, S5E and S5F) and de novo expression of Myc in the peripheral pouch and hinge domains, a region where it is normally not expressed (Fig 5G and 5H). Importantly, the pattern of this signaling signature did not change during regeneration and could still be detected 24 h after egr-expression well into the recovery period (S5G–S5N Fig). Of note, hid-expressing discs maintained their signaling signature as well (S5O and S5P Fig). Combined, we find that JAK/STAT activity and Myc expression are specifically detected in cells undergoing compensatory proliferation that must be driven by non-autonomous signaling from egr-expressing domains. Indeed, JAK/STAT-activating, secreted ligands of the Unpaired family are expressed in JNK-signaling cells [41, 42, 52, 58–60]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. JAK/STAT signaling and Myc-expression are elevated in proliferating cells of egr-expressing discs. (A-D) Control wing disc (A,C) and wing disc after 24 h of egr-expression in the pouch domain (B,D). Discs either express Yorkie-GFP (A,B) or the ERK reporter miniCic-mCherry (C,D). Magnified view of the pouch domain. Discs were stained with DAPI to visualize nuclei (A-D). (E-H) Control wing disc (E,G) and wing disc after 24 h of egr-expression in the pouch domain (F,H). Discs either express the JAK/STAT reporter 10xStat92E>dGFP (E,F) or an endogenously tagged Myc-GFP construct (G,H). Discs were stained with DAPI to visualize nuclei (E-H). Magnified view of the pouch domain (E”-H”). Images with increased brightness show the presence of Myc-GFP in the regenerative domain (G”‘,H”‘). We suggest that the Myc-expressing cells in the anterior pouch domain of control disc are killed by egr-expression and a new expression pattern of Myc is set up de novo by tissue damage signals. Maximum projections of multiple confocal sections are shown in (E,F,G,H); single sections are shown in (A-D). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g005 As a result, this analysis left us with the surprising conclusion, that completely different signaling signatures can be associated with compensatory proliferation and specifically, with accelerated nucleotide incorporation and thus DNA replication speed. These results suggest that at least two distinct regulatory circuits may converged on compensatory proliferation and the same cell cycle adaptation upon distinct damaging challenges. Yki and EGF cooperate to drive compensatory proliferation in response to non-inflammatory damage To investigate which of these signaling pathways may truly be required for compensatory proliferation, we systematically analyzed sufficiency and necessity of Hippo/Yki and Ras/ERK signaling in hid-expressing disc. We first asked if Hippo/Yki or Ras/ERK activation alone were sufficient to induce accelerated EdU incorporation. However, neither expression of a phospho-ablative YkiS168A construct nor RNAi-mediated knock-down of Warts altered the rate of EdU incorporation in mosaic clones, or upon expression in the pouch (Figs 6A–6C, S6A and S6B). Similarly, expression of oncogenic RasV12 alone failed to phenocopy an accelerated S-phase profile (Fig 6D–6F). However, to understand if Hippo/Yki and Ras/ERK are necessary for S-phase acceleration, we created hid-expressing discs heterozygous mutant for a null allele of ykiB5. Indeed, in the very rare discs that we were able to recover due to high lethality, we observed a reduction in EdU incorporation, if compared to control discs (Figs 6G, 6H, S6C and S6D). This suggests, that Hippo/Yki is necessary to drive nucleotide incorporation during S-phase in hid-expressing discs. We performed experiments to test the necessity of Ras/ERK signaling in S-phase acceleration. We analyzed discs that either co-expressed a dominant-negative Egfr (EgfrDN) in hid-expressing cells (S6E–S6H Fig) or that were heterozygous for Ras1 (S6I–S6N Fig). Both strategies failed to reveal changes to EdU incorporation dynamics. However, it is possible that EgfrDN-expressing cells die too quickly in the context of hid-coexpression, and that Ras1 heterozygosity may not sufficiently interfere with ERK function. Thus, other genetic strategies may be needed to perform these experiments. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Yki and ERK cooperate to drive compensatory proliferation in response to non-inflammatory damage. (A) A wing disc expressing the act-GAL4 ‘flip-out’ system controlling the mosaic expression of GFP and UAS-yki.S168A (green in A”‘). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (magenta). (B, C) Control wing disc (B), wing disc after 24 h of UAS-yki expression in the pouch domain (C). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (D) A wing disc expressing the act-GAL4 ‘flip-out’ system controlling the mosaic expression of GFP and UAS-RasV12 (green in D”‘). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (magenta). (E,F) Control wing disc (E) and wing disc after 24 h of UAS-RasV12 expression in the pouch domain (F). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (G,H) Wing disc after 24 h of hid-expression (G) and a wing disc heterozygous for yki B5 after 24 h of hid-expression (H). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. White frame marks the magnified view of the pouch domain shown in (G”-H”‘). (I,J) Control wing disc (I) and a wing disc after 24 h of UAS-yki-GFP and UAS-RasV12 expression in the pouch domain (J). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (K) Quantification of the percentage of DAPI area in the pouch domain of the wing disc that was positive for incorporated EdU in control wing discs, or Yki and RasV12 expressing wing discs. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 8 discs, UAS-yki-GFP, UAS-RasV12, n = 7 discs, **p<0.01). (L) Quantification of incorporated EdU intensity, measured as the mean EdU intensity within the EdU area of the pouch in control wing discs and wing disc after 24 h of Yki and RasV12 expression. A Welch’s test was performed to test for statistical significance. (WT, n = 8 discs, UAS-yki-GFP, UAS-RasV12, n = 7 discs, ****p = <0.0001). (M,N) Control wing disc (M) and a wing disc after 24 h of UAS-yki-GFP and UAS-RasV12 expression in the pouch domain (N). Discs were stained for cleaved Dcp1, a marker of apoptosis. Graphs display mean and 95% CI. Maximum projections of multiple confocal sections are shown in (G,H,I,J). Single sections are shown in (A-F). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g006 Importantly, though, as neither pathway was sufficient to accelerate nucleotide incorporation individually, we tested if Hippo/Yki and Ras/ERK pathways cooperate. Indeed, the combined expression of RasV12 and Yki was sufficient to drive elevated nucleotide incorporation in the pouch (Fig 6I–6L). Not surprisingly, the pouch overgrew, demonstrating that both pathways cooperate in promoting proliferation. This was not associated with elevated levels of apoptosis, indicating that accelerate nucleotide incorporation was directly caused by RasV12 and Yki co-expression (Fig 6M and 6N). Combined these observations demonstrate, that the hid-expressing model of local, non-inflammatory regeneration uses Hippo/Yki and Ras/ERK activation to promote cell cycle adaptations for compensatory proliferation. JAK/STAT and Myc are sufficient to drive S-phase acceleration in response to inflammatory damage Since Hippo/Yki and Ras/ERK signaling did not robustly correlate with domains of compensatory proliferation in egr-expressing disc, we asked if JAK/STAT activation and Myc expression may directly control cell cycle acceleration. We first tested if Myc and JAK/STAT alone were sufficient to induce accelerated EdU incorporation. Indeed, overexpression of Myc alone was sufficient to drive S-phase acceleration, aligning with mammalian reports that Myc can accelerate S-phase progression (Fig 7A–7C) [31]. Similarly, expression of the transcription factor Stat92E was sufficient to cell-autonomously drive high levels of EdU incorporation, confirming that JAK/STAT is a mitogenic pathway strongly implicated in compensatory proliferation (Fig 7D–7F) [39, 61, 62]. We wanted to understand, if Myc or Stat92E activity are rate-limiting for EdU incorporation. We thus generated egr-expressing discs heterozygous for the null allele Stat92E85C3. However, we failed to detect any changes in cells undergoing compensatory proliferation, suggesting that heterozygosity for Stat92E is not rate-limiting for EdU incorporation, or alternatively, that Myc upregulation can compensate for reduced Stat92E function (S7A and S7B Fig). Due to lethality of egr- and hid-expressing larvae heterozygous for dMyc alleles, we were unable to specifically test the necessity of Myc in mediating cell cycle acceleration. However, our observations suggest that activation of JAK/STAT signaling or elevated expression of Myc alone are sufficient to accelerate DNA replication during compensatory proliferation. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. JAK/STAT and Myc are sufficient to drive S-phase acceleration in response to inflammatory damage. (A,B) Control wing disc (A), wing disc after 24 h of UAS-Myc expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (C) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in control wing discs or UAS-Myc expressing wing discs. This serves as a proxy for the number of nuclei undergoing DNA replication. (C’) Quantification of incorporated EdU, measured as mean EdU intensity in the DAPI area within the pouch. A Welch’s test was performed to test for statistical significance. (WT, n = 9 discs, UAS-Myc, n = 9 discs, ****p = <0.0001). (D,E) Control wing disc (D), and a wing disc after 24 h of UAS-Stat92E-expression (E). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (F) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in control wing discs or UAS- Stat92E expressing wing discs. This serves as a proxy for the number of nuclei undergoing DNA replication. (F’) Quantification of incorporated EdU, measured as mean EdU intensity in the DAPI area within the pouch. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, UAS-Stat, n = 10 discs, ***p<0.001, ****p<0.0001). Graphs display mean and 95% CI. Single sections are shown in (A,B,D,E). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g007 Compensatory proliferation is not associated with replication stress Many reports highlight the emergence of replicative stress upon pathological acceleration of DNA replication, for example in tumors [29–32]. To understand if accelerated DNA replication during compensatory proliferation was associated with elevated replication stress, we assessed levels of DNA double-strand breaks in hid- and egr-expressing discs [63, 64]. While very occasionally apoptotic cells displayed high levels of phosphorylated H2Av staining, we failed to detect a general increase in this DNA damage marker in areas of compensatory proliferation (Fig 8A–8F). This suggests that mechanisms exist which ensure that accelerated DNA replication does not generally cause replication stress and DNA damage. It suggests that DNA-replication can be safely accelerated to increase cell cycle progression. Even though we report here that S-phases are accelerated, we also observe that gap phases nearly disappear, suggesting that, ultimately, safe DNA replication speed is still rate-limiting for cell cycle length. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 8. Compensatory proliferation is not associated with replication stress. (A-C) Control wing disc (A), and a wing disc after 24 h of hid-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA damage by staining for phosphorylated γH2A. White frame marks position of views shown in (“, “‘) panels. (C) Quantification of H2Aγ staining intensity within the DAPI area of the pouch domain normalized to DAPI intensity to correct for fluctuations in DNA density. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (WT, n = 10 discs, Hid, n = 10 discs, ns, p = 0.2533). (D-F) Control wing disc (D), and a wing disc after 24 h of egr-expression in the pouch domain (E). Discs were stained with DAPI to visualize nuclei and were assessed for DNA damage by staining for phosphorylated γH2A. Yellow dashed line in (E) demarcates the area of high JNK reporter activity (cyan star) as assessed by TRE-RFP expression (not shown). Compensatory proliferation occurs in a band outside of the JNK-signaling domain (cyan bracket). White frame marks position of views shown in (“, “‘) panels. (F) Quantification of H2Aγ staining intensity within the DAPI area of domain outside the JNK-signaling domain normalized to DAPI intensity to correct for fluctuations in DNA density. Welch’s test was performed to test for statistical significance (WT, n = 9 discs, Egr, n = 9 discs, **p = 0.0019). (G) Model of signaling environment driving compensatory proliferation and accelerated DNA replication in response to two distinct challenges to tissue health. Maximum projections of multiple confocal sections are shown in (A,B,D,E). Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.g008 Discussion In this study we find that two distinct signaling signatures can drive compensatory proliferation during tissue regeneration by shortening gap phases, and surprisingly, accelerating DNA replication during S-phase (Fig 8G). We find that, in the absence of epithelial barrier damage and inflammation, hid-expressing discs accelerate the cell cycle by cooperativity of Hippo/Yki and Ras/ERK signaling. Hippo/Yki and Ras/ERK signaling have been reported to respond to local changes in cell density and may specifically respond to changes in the force balance of cell-cell junctions [6, 65–69]. As the cell density in hid-expressing discs continuously decreases but junctions remain intact, Hippo/Yki and Ras/ERK may be ideally suited to control proliferation in this non-inflammatory environment. Inflammatory tissue damage that disrupts the epithelial barrier is characterized by high JNK activation. High JNK signaling induces a senescent cell cycle arrest which does not support proliferation [35]. Thus, compensatory proliferation must occur distally to the wound and is guided by paracrine factors secreted from the site of damage. In fact, the JAK/STAT activators of the Unpaired cytokine-like family and the Myc activator Wg of the Wnt family are expressed at high levels in egr-expressing and JNK-signaling cells [33, 42]. We observe strong activation of JAK/STAT and Myc outside JNK-signaling domains and either is sufficient to drive accelerated S-phase profiles in the disc. JAK/STAT signaling has been implicated in driving S-phase entry via Cyclin E expression [70]. Myc may facilitate the metabolic drive needed for cell growth in gap phases via targeting of protein synthesis and Cyclin E [71, 72]. Strikingly, metabolic drive may be sufficient for S-phase acceleration as expression of a constitutively active Insulin receptor is sufficient to promote nucleotide incorporation in EdU assays (S8A–S8C Fig). While Myc and JAK/STAT have known access points into the cell cycle, it is not clear how they would specifically promote nucleotide incorporation and thus S-phase acceleration. However, Myc expression has also been found to increase replication speed in mammalian cells [31], suggesting that conserved mechanisms may confer S-phase acceleration. Previous studies have analysed cell cycle changes in regenerating and transdetermining imaginal discs [15, 40]. Specifically, the dissection of the ‘blastema’ cell population and subsequent analysis by flow cytometry revealed both an increase in S-phase and G2 cells, as well as an increase in cell size [40]. Based on more detailed spatial studies from our lab, we now suggest that the G2-component and large cell sizes arise from cells undergoing a JNK-controlled G2-arrest directly at the wound sites [35]. The S-phase component arises from cell undergoing compensatory proliferation next to JNK-controlled wound site (this study). Thus, our work sheds a new light on the spatial organisation of cell cycle adaptations in regenerating tissues. Little is known about molecular details that modulate of S-phase length. S-phase length is controlled by the number of active replication forks and their velocity. The number of active replication forks could be increased via recruitment of dormant ORCs. ORC usage is lineage specific and correlates with transcriptional and epigenetic states [25, 27, 73–77]. The signaling environment of compensatory proliferation alters transcriptional activity, and as a result may recruit dormant ORCs for S-phase acceleration. In contrast, processivity and velocity of the replication complex can be regulated by components of the cell cycle machinery [78–85]. In addition, enhancing access to DNA can promote replication complex processivity. Indeed, the DNA helicase Top3a was previously identified to be genetically required for compensatory proliferation [86]. An additional function for accelerated DNA replication, that goes beyond its role in proliferation, is not known [see also 16]. However, S-phase length has been implicated in cell fate decisions [25–28]. If S-phase acceleration may thus support reprogramming of imaginal disc cells during tissue repair would form the basis of an interesting line of future research. Methods Fly genetics All fly stocks and experimental crosses were maintained on standard media and raised at 18 °C unless otherwise specified. For detailed genotypes, please refer to S1 Table. Flip-out clones GAL4/UAS-driven ‘flip-out’ experiments utilized heat-shock-driven expression of a flipase. The respective crosses were allowed to lay eggs for 72 h at 25°C followed by a heat-shock at 37°C for 5–25 min. Larvae were dissected at wandering 3rd instar stage or as indicated (30 h or 54 h after heat-shock). To analyze the growth of clones in hid-expressing discs, GFP was expressed under the control of a ubiquitin promoter upon ’flip-out’ of an FRT cassette. After a 7 minutes heat-shock at 37°C, the cross was shifted to 30°C for 18 h to activate expression of hid in the rn-GAL4 domain. Larvae were dissected and fixed 6 h into the recovery period. Nuclei were counted for each GFP-positive clone in the pouch or notum. Cell ablation using GAL4/UAS/GAL80ts system to express UAS-hid or UAS-egr To induce expression of egr or hid, experiments were carried out as described in [33, 35, 42] with few modifications. Briefly, larvae of genotype rn-GAL4, tub-GAL80ts (rnts) and carrying the desired UAS-transgenes were staged by a 6 h egg collection and raised at 18°C at the density of 50 larvae/vial. Overexpression of transgenes was induced by shifting the temperature to 30°C for 24 h at day 6 or 7 after egg deposition (AED), as indicated. Larvae were subsequently dissected for analysis (recovery time point R0) or allowed to recover at 18°C for the indicated time. All images represent R0, unless noted otherwise. Control genotypes were either rnts control crosses, or the sibling larvae (+/TM6B, tubGAL80) [33]. At least 20 discs were dissected for each genotype. Immunofluorescence microscopy Wing discs from third instar larvae were dissected and fixed for 15 min at room temperature (RT) in 4% paraformaldehyde in PBS. Washing steps were performed in PBS containing 0.1% TritonX-100 (PBT). Discs were then incubated with primary antibodies in PBT, gently mixing overnight at 4°C. The following antibodies were used: rabbit anti-cleaved Dcp-1 (Cell Signaling, 9578, 1:200), mouse anti-β-Galactosidase (Promega, Z3783, 1:1000), chicken anti-GFP (Abcam, ab13970, 1:1000), rabbit anti-GFP (Invitrogen, G10362, 1:200), rabbit anti-H2Av-pS137 (Rockland, 600-401-914, 1:500), mouse anti-H3-pS10 (Abcam, ab14955, 1:2000), rat anti-HA (MAB facility at the Helmholtz Zentrum München, 3F10, 1:20), mouse anti-MMP1 (DSHB, a mix of 3A6B4, 3B8D12 and 5H7B11, each 1:30), mouse anti-RFP (Abcam, ab65856, 1:100), rat anti-RFP (MAB facility at the Helmholtz Zentrum München, 5F8, 1:20). Tissues were counterstained with DAPI (0.25 ng/μl, Sigma, D9542) or Phalloidin-Alexa Fluor 488/647 (1:100, Life Technologies) or Phalloidin-conjugated TRITC (1:400, Sigma) during incubation with cross-absorbed secondary antibodies coupled to Alexa Fluorophores (Invitrogen or Abcam) at room temperature for 2 h. Tissues were mounted using SlowFade Gold Antifade (Invitrogen, S36936). Whenever possible, experimental and control discs were processed in the same vial and mounted on the same slides to ensure absolute comparability in staining conditions between different genotypes. Genotypes were distinguished on the slide by deliberately co-expressed fluorescence markers (GFP, RFP, HA, LacZ). Of note, the signals of the following fluorescent reporters were further amplified by anti-GFP or anti-mCherry antibody staining: miniCiC-mCherry, Yki-GFP. Images were acquired using the Leica TCS SP8 Microscope (DFG Project 414136422), using sample-matched confocal settings. EdU Labelling EdU incorporation was performed after crude dissection and detected using the Click-iT Plus EdU Alexa Fluor 647 Imaging Kit (Invitrogen, C10640) prior to primary antibody incubation. Briefly, larval cuticles were inverted in Schneider’s medium and incubated with EdU (10μM final concentration) at RT for 15 minutes. Cuticles were then fixed in 4% PFA/PBS for 15 minutes, washed for 30 minutes in PBT 0.5%. EdU-Click-iT labeling was performed according to manufacturer’s guidelines. Tissues were washed in PBT 0.1%, after which additional immunostainings, sample processing and imaging were carried out as described above. BrdU Labelling Larval cuticles were inverted in Schneider’s medium and incubated with BrdU (10μM final concentration) at RT for 15 minutes. Cuticles were then fixed in 4% PFA/PBS for 20 minutes and washed in 0.5% PBT. Samples were then incubated in HCl at 2N concentration for 45 minutes and subsequently washed twice for 2 min in 0.1M Na3BO3 pH 8.5. After three washes in PBT 0.5%, discs were incubated with mouse anti-BrdU (BD, 555627, 1:100) in PBTN, gently mixing overnight at 4°C. Tissues were washed in PBT 0.5%, then counterstained with DAPI (0.25 ng/μl, Sigma, D9542) and secondary antibody coupled to Alexa Fluorophores (Invitrogen) at room temperature for 2 h. Tissues were washed again in PBT 0.5% and PBS before mounting. Mounting and imaging were performed as described above. EdU feeding experiment After hid-expression using the GAL4/UAS/GAL80ts system, larvae were transferred to fly food containing EdU (100 μM final concentration). Larvae were left feeding for 2 h or 18 h. Only larvae still roaming in the food were chosen to dissect after the feeding. After dissection, larvae were fixed in 4% PFA/PBS for 15 minutes, and EdU was detected using the Click-iT Plus EdU Alexa Fluor 647 Imaging Kit (Invitrogen, C10640) as described above. Flow cytometry Cell cycle analysis of wing imaginal discs by flow cytometry was performed as described [87]. Wing imaginal discs from at least 10 larvae were dissected in PBS and incubated for 2 h in PBS containing 9X Trypsin-EDTA (Sigma, T4174) and 0.5 μg/ml Hoechst 33342 (Invitrogen, H3570). Cells were analyzed with an LSRFortessa cell analyzer (BD Biosciences) or FACS Aria II cell sorter (BD Biosciences). Univariate cell cycle analysis was performed using the Watson Pragmatic algorithm in FlowJo v10 (FlowJo). Image analysis and quantification General comments. Where possible, control and experimental samples were fixed, processed and mounted together to ensure comparable staining and imaging conditions. Positive results were verified with a minimum of n = 3 replicates. Images were processed, analyzed and quantified using tools in Fiji (ImageJ v2.0.0) [88] (see below). Great care was taken to apply consistent methods (i.e. number of projected sections, thresholding methods, processing) within experimental settings. Statistical analyses were performed in Graphpad Prism (see below). Briefly, every data set was checked for normality of distribution and homogeneity of variances by applying Shapiro’s and Bartlett’s test, respectively. The α value for each analysis was set to 0.01 (α = 0.01). A Welch’s Test was then performed on one of the replicates as indicated in the figure legends. Confidence interval (95%) is shown for each dataset to show the range of the true population’s mean, for a fairer display of the variability of the sample compared to a point estimate. Statistical tests are indicated in the figure legends. Statistical significance is shown as: * = p val<0.05, ** = p val<0.01, *** = p val<0.001, **** = p val<0.0001. Figure panels were assembled using Affinity Design. EdU area proportion and EdU intensity analysis Depending on the genetic background, a single confocal section or a substack of 3 sections was chosen within each disc, capturing the highest nuclear area of the disc proper. Intensity-based thresholding was used to generate a binary mask of nuclei (DAPI) and EdU areas (EdU). Masks were then expanded with the functions "fill holes" and "dilate". Pixel fluorescence intensities for all channels in these masks were obtained using the ‘Save XY Coordinates’ function, either on the whole tissue or in a ROI selected with the ‘Freehand Selection’ tool in Fiji. The percentage area of replicating DNA in the image was calculated as the ratio of the number of EdU-positive pixels in the DAPI mask over the number of pixels in the DAPI mask. This measure approximates the percentage of cells undergoing DNA replication, circumventing technical difficulties of segmenting individual nuclei in the tightly packed pseudostratified wing epithelium. The average intensity of EdU was calculated by averaging pixel intensities within the EdU mask only. If a stack of 3 section was used, only the EdU percentage was calculated. FUCCI-based cell cycle analysis The proportion of cell-cycle phases (G1, S, LateS, G2) was calculated in Fiji/ImageJ using FUCCI-expressing imaginal discs. A DAPI mask was generated to analyze nuclear levels of FUCCI.NLS markers, by following the same steps described above. After intensity-based thresholding based on subtracting disc-specific non-nuclear background, the fluorescence intensities for each nuclear pixel coordinate was exported using the ‘save XY coordinates’ function. The pixel population was then divided into 4 cell-cycle phases: Ubi-GFP.E2f.1-230 positive area (G1), Ubi-mRFP1.NLS.CycB.1-266/CyO positive area (Late S), Ubi-GFP.E2f.1-230 and Ubi-mRFP1.NLS.CycB.1-266/CyO double positive area (G2), Ubi-GFP.E2f.1-230 and Ubi-mRFP1.NLS.CycB.1-266/CyO double negative area (S). γH2Av intensity analysis A single confocal section was chosen which maximally captured the nuclear DAPI area of the disc proper. For hid-expressing discs, the proliferative ROI was created by free-hand selection of the rn-GAL4 defined pouch area. For egr-expressing discs, the proliferative ROI was created by (1) free-hand selection of the JNK-signaling area in the TRE-RFP channel and then (2) subtracting it from a free-hand selection of the pouch area using characteristic folding patterns as guides. This approach created a ROI of the proliferative domain and excluded the JNK-signaling area with G2-arrested cells and inflammatory signaling. Average intensity was measured for the DAPI channel and the γH2A channel in the ROI. All γH2A intensities were normalized by the DAPI intensity. Fly genetics All fly stocks and experimental crosses were maintained on standard media and raised at 18 °C unless otherwise specified. For detailed genotypes, please refer to S1 Table. Flip-out clones GAL4/UAS-driven ‘flip-out’ experiments utilized heat-shock-driven expression of a flipase. The respective crosses were allowed to lay eggs for 72 h at 25°C followed by a heat-shock at 37°C for 5–25 min. Larvae were dissected at wandering 3rd instar stage or as indicated (30 h or 54 h after heat-shock). To analyze the growth of clones in hid-expressing discs, GFP was expressed under the control of a ubiquitin promoter upon ’flip-out’ of an FRT cassette. After a 7 minutes heat-shock at 37°C, the cross was shifted to 30°C for 18 h to activate expression of hid in the rn-GAL4 domain. Larvae were dissected and fixed 6 h into the recovery period. Nuclei were counted for each GFP-positive clone in the pouch or notum. Cell ablation using GAL4/UAS/GAL80ts system to express UAS-hid or UAS-egr To induce expression of egr or hid, experiments were carried out as described in [33, 35, 42] with few modifications. Briefly, larvae of genotype rn-GAL4, tub-GAL80ts (rnts) and carrying the desired UAS-transgenes were staged by a 6 h egg collection and raised at 18°C at the density of 50 larvae/vial. Overexpression of transgenes was induced by shifting the temperature to 30°C for 24 h at day 6 or 7 after egg deposition (AED), as indicated. Larvae were subsequently dissected for analysis (recovery time point R0) or allowed to recover at 18°C for the indicated time. All images represent R0, unless noted otherwise. Control genotypes were either rnts control crosses, or the sibling larvae (+/TM6B, tubGAL80) [33]. At least 20 discs were dissected for each genotype. Immunofluorescence microscopy Wing discs from third instar larvae were dissected and fixed for 15 min at room temperature (RT) in 4% paraformaldehyde in PBS. Washing steps were performed in PBS containing 0.1% TritonX-100 (PBT). Discs were then incubated with primary antibodies in PBT, gently mixing overnight at 4°C. The following antibodies were used: rabbit anti-cleaved Dcp-1 (Cell Signaling, 9578, 1:200), mouse anti-β-Galactosidase (Promega, Z3783, 1:1000), chicken anti-GFP (Abcam, ab13970, 1:1000), rabbit anti-GFP (Invitrogen, G10362, 1:200), rabbit anti-H2Av-pS137 (Rockland, 600-401-914, 1:500), mouse anti-H3-pS10 (Abcam, ab14955, 1:2000), rat anti-HA (MAB facility at the Helmholtz Zentrum München, 3F10, 1:20), mouse anti-MMP1 (DSHB, a mix of 3A6B4, 3B8D12 and 5H7B11, each 1:30), mouse anti-RFP (Abcam, ab65856, 1:100), rat anti-RFP (MAB facility at the Helmholtz Zentrum München, 5F8, 1:20). Tissues were counterstained with DAPI (0.25 ng/μl, Sigma, D9542) or Phalloidin-Alexa Fluor 488/647 (1:100, Life Technologies) or Phalloidin-conjugated TRITC (1:400, Sigma) during incubation with cross-absorbed secondary antibodies coupled to Alexa Fluorophores (Invitrogen or Abcam) at room temperature for 2 h. Tissues were mounted using SlowFade Gold Antifade (Invitrogen, S36936). Whenever possible, experimental and control discs were processed in the same vial and mounted on the same slides to ensure absolute comparability in staining conditions between different genotypes. Genotypes were distinguished on the slide by deliberately co-expressed fluorescence markers (GFP, RFP, HA, LacZ). Of note, the signals of the following fluorescent reporters were further amplified by anti-GFP or anti-mCherry antibody staining: miniCiC-mCherry, Yki-GFP. Images were acquired using the Leica TCS SP8 Microscope (DFG Project 414136422), using sample-matched confocal settings. EdU Labelling EdU incorporation was performed after crude dissection and detected using the Click-iT Plus EdU Alexa Fluor 647 Imaging Kit (Invitrogen, C10640) prior to primary antibody incubation. Briefly, larval cuticles were inverted in Schneider’s medium and incubated with EdU (10μM final concentration) at RT for 15 minutes. Cuticles were then fixed in 4% PFA/PBS for 15 minutes, washed for 30 minutes in PBT 0.5%. EdU-Click-iT labeling was performed according to manufacturer’s guidelines. Tissues were washed in PBT 0.1%, after which additional immunostainings, sample processing and imaging were carried out as described above. BrdU Labelling Larval cuticles were inverted in Schneider’s medium and incubated with BrdU (10μM final concentration) at RT for 15 minutes. Cuticles were then fixed in 4% PFA/PBS for 20 minutes and washed in 0.5% PBT. Samples were then incubated in HCl at 2N concentration for 45 minutes and subsequently washed twice for 2 min in 0.1M Na3BO3 pH 8.5. After three washes in PBT 0.5%, discs were incubated with mouse anti-BrdU (BD, 555627, 1:100) in PBTN, gently mixing overnight at 4°C. Tissues were washed in PBT 0.5%, then counterstained with DAPI (0.25 ng/μl, Sigma, D9542) and secondary antibody coupled to Alexa Fluorophores (Invitrogen) at room temperature for 2 h. Tissues were washed again in PBT 0.5% and PBS before mounting. Mounting and imaging were performed as described above. EdU feeding experiment After hid-expression using the GAL4/UAS/GAL80ts system, larvae were transferred to fly food containing EdU (100 μM final concentration). Larvae were left feeding for 2 h or 18 h. Only larvae still roaming in the food were chosen to dissect after the feeding. After dissection, larvae were fixed in 4% PFA/PBS for 15 minutes, and EdU was detected using the Click-iT Plus EdU Alexa Fluor 647 Imaging Kit (Invitrogen, C10640) as described above. Flow cytometry Cell cycle analysis of wing imaginal discs by flow cytometry was performed as described [87]. Wing imaginal discs from at least 10 larvae were dissected in PBS and incubated for 2 h in PBS containing 9X Trypsin-EDTA (Sigma, T4174) and 0.5 μg/ml Hoechst 33342 (Invitrogen, H3570). Cells were analyzed with an LSRFortessa cell analyzer (BD Biosciences) or FACS Aria II cell sorter (BD Biosciences). Univariate cell cycle analysis was performed using the Watson Pragmatic algorithm in FlowJo v10 (FlowJo). Image analysis and quantification General comments. Where possible, control and experimental samples were fixed, processed and mounted together to ensure comparable staining and imaging conditions. Positive results were verified with a minimum of n = 3 replicates. Images were processed, analyzed and quantified using tools in Fiji (ImageJ v2.0.0) [88] (see below). Great care was taken to apply consistent methods (i.e. number of projected sections, thresholding methods, processing) within experimental settings. Statistical analyses were performed in Graphpad Prism (see below). Briefly, every data set was checked for normality of distribution and homogeneity of variances by applying Shapiro’s and Bartlett’s test, respectively. The α value for each analysis was set to 0.01 (α = 0.01). A Welch’s Test was then performed on one of the replicates as indicated in the figure legends. Confidence interval (95%) is shown for each dataset to show the range of the true population’s mean, for a fairer display of the variability of the sample compared to a point estimate. Statistical tests are indicated in the figure legends. Statistical significance is shown as: * = p val<0.05, ** = p val<0.01, *** = p val<0.001, **** = p val<0.0001. Figure panels were assembled using Affinity Design. General comments. Where possible, control and experimental samples were fixed, processed and mounted together to ensure comparable staining and imaging conditions. Positive results were verified with a minimum of n = 3 replicates. Images were processed, analyzed and quantified using tools in Fiji (ImageJ v2.0.0) [88] (see below). Great care was taken to apply consistent methods (i.e. number of projected sections, thresholding methods, processing) within experimental settings. Statistical analyses were performed in Graphpad Prism (see below). Briefly, every data set was checked for normality of distribution and homogeneity of variances by applying Shapiro’s and Bartlett’s test, respectively. The α value for each analysis was set to 0.01 (α = 0.01). A Welch’s Test was then performed on one of the replicates as indicated in the figure legends. Confidence interval (95%) is shown for each dataset to show the range of the true population’s mean, for a fairer display of the variability of the sample compared to a point estimate. Statistical tests are indicated in the figure legends. Statistical significance is shown as: * = p val<0.05, ** = p val<0.01, *** = p val<0.001, **** = p val<0.0001. Figure panels were assembled using Affinity Design. EdU area proportion and EdU intensity analysis Depending on the genetic background, a single confocal section or a substack of 3 sections was chosen within each disc, capturing the highest nuclear area of the disc proper. Intensity-based thresholding was used to generate a binary mask of nuclei (DAPI) and EdU areas (EdU). Masks were then expanded with the functions "fill holes" and "dilate". Pixel fluorescence intensities for all channels in these masks were obtained using the ‘Save XY Coordinates’ function, either on the whole tissue or in a ROI selected with the ‘Freehand Selection’ tool in Fiji. The percentage area of replicating DNA in the image was calculated as the ratio of the number of EdU-positive pixels in the DAPI mask over the number of pixels in the DAPI mask. This measure approximates the percentage of cells undergoing DNA replication, circumventing technical difficulties of segmenting individual nuclei in the tightly packed pseudostratified wing epithelium. The average intensity of EdU was calculated by averaging pixel intensities within the EdU mask only. If a stack of 3 section was used, only the EdU percentage was calculated. FUCCI-based cell cycle analysis The proportion of cell-cycle phases (G1, S, LateS, G2) was calculated in Fiji/ImageJ using FUCCI-expressing imaginal discs. A DAPI mask was generated to analyze nuclear levels of FUCCI.NLS markers, by following the same steps described above. After intensity-based thresholding based on subtracting disc-specific non-nuclear background, the fluorescence intensities for each nuclear pixel coordinate was exported using the ‘save XY coordinates’ function. The pixel population was then divided into 4 cell-cycle phases: Ubi-GFP.E2f.1-230 positive area (G1), Ubi-mRFP1.NLS.CycB.1-266/CyO positive area (Late S), Ubi-GFP.E2f.1-230 and Ubi-mRFP1.NLS.CycB.1-266/CyO double positive area (G2), Ubi-GFP.E2f.1-230 and Ubi-mRFP1.NLS.CycB.1-266/CyO double negative area (S). γH2Av intensity analysis A single confocal section was chosen which maximally captured the nuclear DAPI area of the disc proper. For hid-expressing discs, the proliferative ROI was created by free-hand selection of the rn-GAL4 defined pouch area. For egr-expressing discs, the proliferative ROI was created by (1) free-hand selection of the JNK-signaling area in the TRE-RFP channel and then (2) subtracting it from a free-hand selection of the pouch area using characteristic folding patterns as guides. This approach created a ROI of the proliferative domain and excluded the JNK-signaling area with G2-arrested cells and inflammatory signaling. Average intensity was measured for the DAPI channel and the γH2A channel in the ROI. All γH2A intensities were normalized by the DAPI intensity. Supporting information S1 Fig. Compensatory proliferation is associated with short G1, G2 and S-phases and EdU incorporation is not sensitive to tissue architecture defects. (A-D) Control wing disc (A,C) and wing disc after 8 h of hid-expression in the pouch domain (D,E) were assessed for DNA replication activity by EdU incorporation. Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. Pyknotic nuclei in (C) confirm onset of hid-induced cell death with the wing disc pouch domain where hid is expressed under the control of rn-GAL4 (cyan dotted line). S-phase-specific incorporation of the nucleotide analogue EdU into replicating DNA is already elevated in the pouch after 8 h of hid-expression (D). (E,F) Nota of control wing disc and (E) wing disc after 18 h of hid-expression in the pouch domain and at 6 h into the recovery period (F). Discs express two ‘flip-out’ construct to generate labelled clones, either controlling expression of GFP (green) or of Lac-Z (red). As both constructs are induced independently, clones either express GFP (green), LacZ (red) or both (yellow). (G) Quantification of number of cells per clone in the notum domain, from control or hid-expressing discs. Mean and 95% confidence interval (CI) are shown. Welch’s test was performed to test for statistical significance. (WT, n = 42 clones and Hid, n = 52 clones, ns p = 0.0747). (H) Peripodium of wild type wing was stained with DAPI to visualize nuclei (H), expresses the FUCCI reporter system, ubi-GFP-E2f11-230 (green in overlay) and ubi-mRFP-NLS-CycB1-266 (red in overlay) (H’,H”,H”“,H”“‘). Discs were assessed for DNA replication activity by EdU incorporation (H”‘,H”“‘). (H”“) Composite view of H,H’,H”. (H”“‘) Composite view of H’,H”,H”‘. Euchromatin correlates with lower DAPI staining and is replicated early (magenta arrow). Satellite repeats (heterochromatin) correlate with bright DAPI staining and replicate late (purple arrow). (I) Wild type wing disc expressing the FUCCI reporter system, ubi-GFP-E2f11-230 (green in overlay I,I”) and ubi-mRFP-NLS-CycB1-266 (red in overlay I,I”) and stained with DAPI to visualize nuclei. An extremely apical section through the wing pouch visualizes mitotic cells. Mitosis occurs exclusively on the apical surface in imaginal discs. Magnified view in I’ and I” is indicated by blue frame in I. DAPI staining visualizes progression through M-phase: metaphase plates (white arrows) and two separate nuclei (blue arrows). (J) Schematic representation of cell cycle phase identification using the FUCCI reporters validated by EdU incorporation assays and mitotic markers. (K) Flow cytometry analysis of DNA content in undamaged control wing discs (grey) and in wing disc after 24 h of hid-expression (red). The pouch of the wing disc was labeled by rn-GAL4-driven expression of UAS-GFP and thus cells outside the rn-GAL4 domain can be distinguished by the lack of GFP expression. GFP-negative events were plotted. The cell cycle of cells outside of the control and hid-expressing domains in the pouch is not different. (L) Flow cytometry analysis of DNA content in undamaged control wing discs (grey) and in wing disc after 24 h of hid-expression (green in I, red in I’), as shown in Fig 1O. The gating of Hoechst-channel (DNA) was opened (if compared to Fig 1O) to also visualize flow cytometry events with higher Hoechst intensity. No difference in the proportion of these higher intensity events between control and hid-expressing discs can be detected. This suggests that hid-expressing discs do not experience a specific increase in events that may represent endoreplicating nuclei. (M-Q) Control wing disc (M,P) and wing disc after 24 h of ectopic MMP1 (N), MMP2 (M) and cora-RNAi (Q) expression in the pouch domain. Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by BrDU incorporation or EdU incorporation. Cora-RNAi expressing discs were stained for Cora to assess knock-down efficiency. Maximum projection of the apical domain is shown (P’,Q’). Maximum projections of multiple confocal sections are shown in (B,D); Single confocal sections are shown in (A,C,E,F,H,I). Maximum projection of the apical domain is shown (P’,Q’). Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s001 (PDF) S2 Fig. EdU incorporation is not sensitive to tissue architecture defects. (A) Wing disc after 24 h of egr-expression (E) in the pouch domain. Discs also express the JNK-reporter TRE-RFP (A‘, red in A”“). The disc was stained with DAPI to visualize nuclei (A, red in A”“‘), for the mitotic marker phospho-His3 to visualize M-phase cells (A”‘, green) and was assessed for DNA replication activity by EdU incorporation (grey). Compare number of phospho-His3 positive events to the number of EdU labelled nuclei to estimate relatively low frequency of M-phase cells in discs. (B-E) Nota of control wing disc (B,D) and and wing disc after 24 h of hid-expression (C) and after 24 h of egr-expression (E) in the pouch domain. Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (B’-E’). Cyan star in (E) marks small domain of frequent transdetermination as described in M. I. Worley, L. A. Alexander and I. K. Hariharan, CtBP impedes JNK- and Upd/STAT-driven cell fate misspecifications in regenerating Drosophila imaginal discs, Elife 2018 Vol. 7. Cells in this patch undergo compensatory-like proliferation as part of the transdetermination program and therefore incorporate more EdU. Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s002 (PDF) S3 Fig. Compensatory proliferation does not revert to a developmentally younger cell cycle and JNK signaling cannot cell-autonomously promote cell cycle acceleration. (A-E) Wing discs at different developmental stages (day 5, 6, 7 and 8 after egg lay)(A-D), and a wing disc after 24 h of hid-expression (E). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity to visualize S-phase cells by EdU incorporation. Please compare (A-D) to (E). (F-I) Wing disc expressing the FUCCI reporter system, ubi-GFP-E2f11-230 (green in overlay) and ubi-mRFP-NLS-CycB1-266 (red in overlay) at different developmental stages (day 5, 6, 7 and 8 after egg lay). Discs were stained with DAPI to visualize nuclei. (J) Quantification of cell cycle phase distribution using the FUCCI profile at day 5 and day 8 after egg lay. Phases were defined as described in experimental procedures. A Welch’s test was performed to test for statistical significance between day 5 and day 8 wing discs: G1 (p = 0.0269 *), S (p = 0.3828 ns), Late S (p = 0.9363 ns), G2 (p = 0.0092 **). n = 5 disc for each day. (K-M) Control wing disc (K) and wing disc after 24 h of hid-expression in the pouch domain (L, M). Discs were stained with DAPI to visualize nuclei. Discs were assessed for JNK activity by TRE-RFP reporter activity (K,L). Discs were assessed for DNA replication activity by EdU incorporation (M). (N,O) Control wing disc (N) and wing disc after 12 h of puc-RNAi-expression in the pouch domain (O). Discs were stained with DAPI to visualize nuclei. Basal section of the disc from Fig 3D–3E are shown. Pyknotic nuclei visualize cell death patterns and indicate that, as expected, JNK-activity is elevated upon knock-down of puc. (P) Wing disc after 24 h of bskDN-expression in the engrailed domain using en-GAL4 (P’, red in P”‘). Discs were stained with DAPI to visualize nuclei (P). Discs were assessed for DNA replication activity by EdU incorporation (P”, cyan in P”‘). (Q, R) Control wing disc (Q) and wing disc after 24 h of co-expressing hid and bskDN in the pouch domain (R). Discs were stained with DAPI to visualize nuclei (Q,R). Discs were assessed for DNA replication activity by EdU incorporation (Q’,R’). Graphs display mean and 95% confidence interval (CI). Single confocal sections are shown. Scale bars: 50 μm. Dotted lines (red) outline stereotypic folds in the wing discs. https://doi.org/10.1371/journal.pgen.1010516.s003 (PDF) S4 Fig. Yorkie activity and ERK signaling are elevated in proliferating cells of hid-expressing discs. (A-D) Control wing disc (A,C), wing disc after 24 h of hid-expression in the pouch domain (B,D). Discs either express Yorkie-GFP (A,B) (green) or the ERK reporter miniCic-mCherry (C,D) (green). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation (A’-D’, or magenta). Magnified view of the pouch domain shown. (E, F) Control wing disc (E) and wing disc after 24 h of hid-expression in the pouch domain (F). Discs express the JAK/STAT reporter 10xStat92E>dGFP (green) and DNA replication activity was assessed by EdU incorporation (magenta). Magnified view of the pouch domain shown (E”,F”). Same disc as in Fig 4G and 4H are shown. Maximum projections of multiple confocal sections are shown in (E,F); single sections are shown in (A,B,C,D). Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s004 (PDF) S5 Fig. JAK/STAT signaling and Myc-expression are elevated in proliferating cells of egr-expressing discs and signaling signature is still the same 24 h into the recovery period. (A-D) Control wing disc (A,C), wing disc after 24 h of egr-expression in the pouch domain (B,D). Discs either express Yorkie-GFP (A,B) (green) or the ERK reporter miniCic-mCherry (C,D) (green). Discs were stained with DAPI to visualize nuclei (A”‘-D”‘) and were assessed for DNA replication activity by EdU incorporation (A’-D’ or magenta). Magnified view of the pouch domain shown. (E,F) Control wing disc (E) and wing disc after 24 h of egr-expression in the pouch domain (F). Discs express the JAK/STAT reporter 10xStat92E>dGFP (green) and DNA replication activity was assessed by EdU incorporation (magenta). Magnified view of the pouch domain (E”,F”). Please note that these discs are the same as shown in Figs 5E and 4G. (G-P) Control wing disc (G,I,K,M,O), or wing disc after 24 h of expression in the pouch domain and then analyzed 24 h into the recovery period, after egr-expression (H,J,L,N) or hid-expression (P) was stopped. Discs either express Yorkie-GFP (G,H), the ERK reporter miniCic-mCherry (I,J), the JAK/STAT reporter 10xStat92E>dGFP (K,L) or an endogenously tagged Myc-GFP construct (M-P). Images with increased brightness show the presence of Myc-GFP in the regenerative domain (M”,N”). We suggest that the Myc-expressing cells in the anterior pouch domain of control disc are killed by egr-expression and a new expression pattern of Myc is set up de novo by tissue damage signals, which is maintained throughout the regenerative period. The interspersed apoptosis and the lack of a JNK-driven wound response program in hid-expressing disc maintains the original myc-expression pattern in the anterior pouch. Discs were stained with DAPI to visualize nuclei. Maximum projections of multiple confocal sections are shown in (E,F); single sections are shown in (A-D). Scale bars: 50 μm. Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s005 (PDF) S6 Fig. Yki and ERK cooperate to drive compensatory proliferation in response to non-inflammatory damage. (A,B) A wing disc expressing the act-GAL4 ‘flip-out’ system controlling the mosaic expression of GFP (A’,B’, or green) and UAS-Warts-RNAi (A), or UAS-Hippo-RNAi (B). Discs were stained with DAPI to visualize nuclei (A,B) and were assessed for DNA replication activity by EdU incorporation (A”,B”, or magenta). (C,D) Control wing disc (C), and a wing disc heterozygous for ykiB5 (D). Discs were stained with DAPI to visualize nuclei (C,D) and were assessed for DNA replication activity by EdU incorporation (C’,D’). (E-H) Control wing disc (E), and a control wing disc after 24 h of Egfr.DN-expression in the pouch domain (F). A control wing disc after 24 h of hid-expression (G) and a wing disc after 24 h of hid- and Egfr.DN-co-expression in the pouch domain (H). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. Egfr.DN does not change EdU incorporation dynamics in wild type and hid-expressing discs. (I-L) Control wing disc (I) and a control wing disc heterozygous for Ras1 (J). A control wing disc after 24 h of hid-expression (K) and a wing disc after 24 h of hid-expression and heterozygous for Ras1 (L). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. Heterozygosity for Ras1 does not change EdU incorporation dynamics in wild type and hid-expressing discs. (M) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in hid-expressing discs and hid-expressing discs heterozygous for Ras1. This serves as a proxy for the number of nuclei undergoing DNA replication. Mean and 95% CI are shown. Welch’s test was performed to test for statistical significance. (Hid, n = 8 discs; Hid, Ras1/+, n = 7 discs, ns, p = 0.266). (N) Quantification of incorporated EdU, measured as the mean EdU intensity in the EdU area within the pouch of hid-expressing discs and hid-expressing discs heterozygous for Ras1. This serves as a proxy for the dynamics of nucleotide incorporation. A Welch’s test was performed to test for statistical significance. (Hid, n = 8 discs, Hid, Ras1/+, n = 7 discs, ns, p = 0.255). Single sections are shown in (A-F,I-L). Maximum projections of multiple confocal sections are shown in (G,H). Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s006 (PDF) S7 Fig. Analysis of necessity of Stat92E for accelerated nucleotide incorporation. (A,B) Control wing disc after 24 h of egr-expression (A) and a wing disc heterozygous for the Stat92E85C3 null allele after 24 h of egr-expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. https://doi.org/10.1371/journal.pgen.1010516.s007 (PDF) S8 Fig. Insulin signaling is sufficient to drive accelerated DNA replication. (A,B) Control wing disc (A), wing disc after 24 h of UAS-InR-DA expression in the pouch domain (B). Discs were stained with DAPI to visualize nuclei and were assessed for DNA replication activity by EdU incorporation. (C) Quantification of the percentage of DAPI areas that were positive for incorporated EdU in control wing discs or UAS-InR-DA expressing wing discs. This serves as a proxy for the number of nuclei undergoing DNA replication (WT, n = 9 discs, UAS-InR-DA, n = 9 discs, p = 0.1711). (C’) Quantification of incorporated EdU, measured as mean EdU intensity in the EdU area within the pouch. A Welch’s test was performed to test for statistical significance. (WT, n = 9 discs, UAS-InR-DA, n = 9 discs, *p = 0.0187). Single sections are shown in (A,B). Scale bars: 50 μm. https://doi.org/10.1371/journal.pgen.1010516.s008 (PDF) S1 Table. Fly strains used in this study. https://doi.org/10.1371/journal.pgen.1010516.s009 (DOCX) Acknowledgments We thank the reviewers for critical comments on the manuscript. We thank the LIC facility at the University of Freiburg for technical help with imaging. We thank the Bloomington Drosophila Stock Center (BDSC), the Vienna Drosophila Stock Collection (VDRC) and the Developmental Studies Hybridoma Bank (DSHB) for providing fly stocks and antibodies. We thank David Bilder, Erica Bach, Barry Thompson, Gines Morata, Iswar Hariharan, Dirk Bohmann and Romain Levayer for sharing reagents.

Poll, Sarah R.;Martin, Renan;Wohler, Elizabeth;Partan, Elizabeth S.;Walek, Elizabeth;Salman, Shaima;Groepper, Daniel;Kratz, Lisa;Cernach, Mirlene;Jesus-Garcia, Reynaldo;Haldeman-Englert, Chad;Choi, Yoon Jae;Morris, Carol D.;Cohen, Bernard;Hoover-Fong, Julie;Valle, David;Semenza, Gregg L.;Sobreira, Nara L. M.

Riva, Sabrina;Ispizua, Juan Ignacio;Breide, María Trinidad;Polcowñuk, Sofía;Lobera, José Ricardo;Ceriani, María Fernanda;Risau-Gusman, Sebastian;Franco, Diana Lorena

doi: 10.1371/journal.pgen.1010258pmid: 36548223

Introduction In most animals, endogenous circadian clocks coordinate physiological and behavioral processes to keep the entire organism in synchrony with the 24 hours day-night cycling environment. In Drosophila melanogaster the circadian clock is composed by approximately 150 neurons, where the levels of clock-related proteins oscillate with periods close to 24 hours. These neurons are organized in different clusters [1,2] whose interaction is necessary for a coherent and plastic control of behavior [3]. The ventrolateral neurons (LNvs) are important because they drive rhythmicity under free-running conditions [4–7], mostly through the release of pigment dispersing factor (PDF), a neuropeptide expressed in the small (sLNvs) and large (lLNvs) which is key for the communication between them and other group of the clock neurons [5–8]. In many species, mating induces changes in the behavior and physiology of females; these changes are known as postmating responses (PMR). In Drosophila females, mating induces a reduction in partner receptivity [9,10], an increase in egg production and oviposition [11], stimulation of the immune response [12] and changes in sleep and activity patterns [13,14], as well as in nutritional preferences [15]. These PMRs are mediated by the sex peptide (SP), transferred from males to females during mating [16,17]. SP acts mostly via the sex peptide receptor (SPR), on a small number of uterine sex peptide sensory neurons (SPSN) that co-express the sex-determination genes doublesex (dsx), fruitless (fru), and pickpocket (ppk), triggering various PMRs [17–19]. It has also been suggested that SP could act via the hemolymph [20], which is consistent with SPR being expressed broadly in the CNS [21]. Fruit flies display two activity bouts, one at dawn (“morning peak”) and the other at dusk (“evening peak”). The early start of the morning activity peak shows that flies can anticipate the transition between night and day. Mated females usually have greater activity than virgin females and males, and less daytime sleep [13,14], a change modulated mainly by SP through SPR within SPSN and Sex Peptide Abdominal Ganglion (SAG) neurons [22]. This PMR has also been observed in females of other Drosophila species such as D. suzukii [23]. In general, sleep is not only regulated by the circadian clock, but also by a homeostatic process [24], even in Drosophila [25]. Thus, the reduction of daytime sleep displayed by mated females could be due to an interference with any of these two processes. An important question is whether the mating status can influence the circadian clock to modify the temporal behavior of mated females. We show here that mated females lose their ability to anticipate the night-day transition, a clear output of the circadian clock. This PMR is probably mediated by SP acting on ppk+ neurons, since decreased expression of SPR in these neurons restores morning anticipation in mated females. We searched for projections of ppk+ expressing neurons near circadian clusters and found that the pdf+ sLNvs are postsynaptic targets of ppk+ neurons. This connection, along with the relationship between PDF and morning anticipation [4], suggested that ppk+ neurons are involved in the inhibition of either expression or transport of PDF. Accordingly, we found that PDF levels are reduced in the dorsal termini of mated females, which can be partially rescued through downregulation of SPR in ppk+ neurons. Thus, our results are consistent with a model whereby mating-triggered signals are delivered to the clock network in order to modulate the temporal organization of the behavior. Results Mating induces loss of morning anticipation in mated females Locomotor activity has been extensively studied in males, but much less in female (mated or virgin) flies, and most published studies were performed using the Drosophila Activity Monitor tracking system (DAM, Trikinetics). On this system, flies are housed in small tubes (¬320 mm3), and their movement is recorded only when they cross an infrared beam. It has been reported [26] though, that activity records differ between DAM and the more accurate video tracking systems. Moreover, activity of mated flies depends on the size of the arena where they are placed [26]. Thus, in order to provide mated females with a more adequate environment where movement can be accurately recorded, we developed a system where flies are placed in a set of relatively large transparent chambers (80 x 8 x 8 mm), and are tracked using a video system (for a detailed description please refer to the Methods section). In order to validate our setup, we compared the activity recorded by our video tracking method, which directly measures activity as distance travelled per second, with recordings obtained using the DAM system, which estimates activity as the number of interruptions of a single infrared beam, firstly in male flies. Both methods show that flies display bimodal locomotor activity patterns over several 12 h: 12 h LD cycles, and no significant differences were found in the morning anticipation activity between both systems (S1A–S1C Fig). However, a close examination of their sleep pattern revealed that flies seem to sleep significantly more when locomotor activity is monitored using the DAM system than when they are registered using our video tracking method (S1D and S1E Fig). This is probably because the DAM system fails to detect a portion of the activity that flies display along the day, for instance when they stand on one side of the tube, without crossing the infrared beam, likely resulting in an overestimation of sleep duration. These results confirm that even though single-fly recordings generated either by the DAM system and our system produce highly similar locomotor activity profiles, validating our experimental setup, video tracking acquisition provides more accurate register of the activity patterns. Next, we compared the locomotor activity of males, virgin and mated females of different control laboratory lines, white (w1118), Oregon R (OreR) and Canton-S (CS) under 12 h: 12 h LD cycles. Fig 1A shows the average activity profile (AAP) for males, virgin and mated females of w1118 flies. Male flies have two pronounced peaks of daily activity that start in anticipation to the lights-on and lights-off transitions, separated by a relatively long interval with very low activity (siesta). In contrast, mated females displayed a sustained and robust activity during daytime while at night the activity remains as low as in males. Virgin females, on the other hand, are more active than males during daytime but less so than mated females [13,14]. A closer visual inspection of AAPs showed that males and virgin females have a gradual increase in activity before lights-on (termed morning anticipation), whereas mated females did not display this morning anticipation, but they had a high amplitude morning startle response coincident with lights-on (Fig 1B). A more quantitative assessment of the degree of anticipation is given by the morning anticipation index (MAI, defined in Materials and Methods) [27]. Fig 1C shows a significant reduction in MAI in mated females compared with males supporting a strongly dimorphic trait. Additionally, the figure shows that mated females display a significant reduction in MAI when compared to virgin females, to the point that morning anticipation is virtually suppressed (on average), suggesting a postmating effect. Interestingly, all three control strains showed a reduced morning anticipation in mated females, supporting the idea that this effect occurs across different genetic backgrounds. Thus, our data shows that one of the postmating responses characteristic of mated females is the suppression of their ability to anticipate the night-day transition. In this article we focus on characterizing this postmating behavior. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Mating suppresses morning anticipation in females of Drosophila melanogaster. (A) Average locomotor activity profile of white (w1118) flies in LD. The number in the horizontal axes refers to days after males were removed. (B) Zoom of the region enclosed by dashed lines in panel A to highlight the increase of the locomotor activity at the end of the night. (C) Anticipation indexes for males (blue), virgin (orange) and mated (red) females of white, CantonS and Oregon flies. For w1118 flies: males (n = 62), virgin females (n = 64) and mated females (n = 66). For CS flies: males (n = 39), virgin females (n = 37) and mated females (n = 43) and for Oregon flies: males (n = 34), virgin females (n = 35) and mated females (n = 37). Each dot corresponds to the average index of three complete days calculated for a single fly. The mean and SEM are shown as bars. Statistical analysis: Kruskal–Wallis test; for the MAI of white flies, X2 = 78.02, p< 0.0001; for the MAI of CS flies, X2 = 49.48, p< 0.0001; for the MAI of OreR flies, X2 = 44.43, p< 0.0001. Pairwise comparisons were performed using Dunn’s multiple comparisons test. Statistically significant differences are represented by *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. https://doi.org/10.1371/journal.pgen.1010258.g001 Loss of morning anticipation is mediated by sex peptide in ppk+ neurons Although SPR is detected broadly on the female reproductive tract, the ventral nerve cord and the brain [21], only a restricted subset of fru+/ppk+/dsx+ sex peptide sensory neurons (SPSN) expressing SPR in the reproductive system are necessary and sufficient to induce some SP-mediated postmating responses [17–19]. To examine whether SPSN neurons are involved in the modulation of morning anticipation through SP signaling, we evaluated the impact of SPR downregulation by RNA interference (RNAi) (21) on activity patterns of mated females. As expected, SPR downregulation in SPSN neurons induced a significant increase in morning anticipation compared with mated controls (Fig 2A). Some PMRs are controlled by a larger group of neurons than the SPSN. For example, oviposition recruits more dsx+ neurons that the ones present in the SPSN [19,20]. Likewise, fru+ neurons control many sexual behaviors in males and females [28]. Additionally, whereas SPSN neurons communicate with the brain using an indirect pathway (that passes through SAG neurons) it has been reported that some ppk+ neurons project directly to the brain [29], the main center of circadian control. For these reasons we decided to study separately the effect of the downregulation of SPR in each of these three neuronal groups. SPR knock down in fru+ and dsx+ neurons resulted in a significant increase in the MAI compared to mated controls (Fig 2B and 2C). This increase in MAI was similar to the value obtained when SPR expression was reduced in SPSN neurons. However, SPR knock down in ppk+ neurons resulted in a dramatic increase in this index (Fig 2D) suggesting a distinctive role of ppk+ neurons in the control of this postmating response. Remarkably, we did not observe any significant difference in morning anticipation when we downregulated SPR in ppk+ neurons in virgin females. Since all virgin genotypes have morning anticipation, the regulation by SPR is indeed mating-state-dependent (Fig 2E). These findings demonstrate the need for SPR function in SPSN neurons to respond to and transmit SP-generated signals to control morning anticipation in mated females and also support the hypothesis that additional spr+ ppk+ neurons beyond the SPSN domain play a role in the modulation of morning anticipation (Fig 2F). Interestingly these additional neurons do not seem to play any role in the control of the loss of daytime sleep (S2 Fig) which is a different PMR that has been reported to be regulated by the action of SP in the SPSN neurons [22]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Downregulation of SPR in ppk+ neurons restores morning anticipation. Morning anticipation index in mated females upon chronic SPR knock down in different groups of neurons and their controls. (A) SPSN-Gal4>UAS-Dicer2, SPR-IR1 (n = 45); SPSN-Gal4>+ (n = 44); UAS-Dicer2, SPR-IR1>+ (n = 44). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test: X2 = 36.50 p < 0.0001. (B) fru-Gal4>UAS-Dicer2, SPR-IR1: (n = 46), fru-Gal4>+ (n = 46); UAS-Dicer2, SPR-IR1>+ (n = 42). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test X2 = 15.38 p < 0.001. (C) dsxGal4>UAS-Dicer2, SPR-IR1 (n = 23); dsxGal4>+ (n = 23); UAS-Dicer2, SPR-IR1>+ (n = 22). One-way ANOVA with Tukey’s post hoc tests, F = 19.58 p < 0.0001. (D) Mated females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 72); ppk-Gal4>+ (n = 71); UAS-Dicer2, SPR-IR1>+ (n = 62). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test X2 = 107.5 p < 0.0001. (E) Virgin females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 41); ppk-Gal4>+ (n = 23); UAS-Dicer2, SPR-IR1>+ (n = 45). One-way ANOVA with Tukey’s post hoc tests F = 4 p < 0.0210. (F) Mated females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 25); SPSN-Gal4>UAS-Dicer2, SPR-IR1 (n = 20); UAS-Dicer2, SPR-IR1>+ (n = 24). One-way ANOVA with Tukey´s post hoc test F = 75.04 p < 0.00018. Dots represent independent flies, the mean and SEM are shown. Statistically significant differences are represented by *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns = not significant. https://doi.org/10.1371/journal.pgen.1010258.g002 Since the SPSN neurons contact the central brain indirectly, as it passes through neurons in the abdominal ganglion [17–19,30], and that some ppk+ neurons have been shown to project to the brain [29], we wondered if ppk+ neurons that control morning anticipation could provide an additional, more direct route for mating signals to reach the central brain. LNvs are postsynaptic targets of ppk+ neurons Our observations suggest that SPR activation mostly in ppk+ neurons alters the temporal organization of rest-activity cycles in mated females inducing the loss of morning anticipation, a feature regulated by the circadian clock. We next wondered how the signal provided by ppk+ neurons would reach the circadian network. Thus, we first analyzed the expression pattern of the ppk-Gal4 driver in adult females by expressing a membrane-bound green fluorescent protein (UAS-mCD8-GFP) under the control of ppk-Gal4. We observed GFP expression in the reproductive tract, in the abdominal ganglia as well as in a few somas in the dorsomedial region of the central brain (Fig 3). To visualize putative dendritic and axonal terminals of ppk+ neurons, both DenMark, a specific somatodendritic marker [31] and synaptotagmin (syt)-GFP, a presynaptic marker localized to synaptic vesicles [32], were expressed in mated females. ppk+ neurons displayed strong labelling of presynaptic syt-GFP, in the subesophageal ganglion zone (SEZ), and in the dorsal protocerebrum where they are in a close apposition to pdf+ terminals. Also, the somatodendritic compartment of ppk+ neurons displayed specific labelling in the dorsal protocerebrum and the SEZ (Fig 4A–4C and S1 Video). Then, to explore whether ppk+ neurons could provide direct input to circadian neurons, we employed the anterograde trans-synaptic tracing tool trans-Tango [33] to define specific downstream synaptic partners of ppk+ neurons. Expression of the tethered trans-Tango ligand under a Gal4 neuronal pattern of choice triggers mtd-Tomato expression in postsynaptic targets. We co-expressed the trans-Tango ligand and GFP to label pre-synaptic ppk+ neurons, and mtd-Tomato in the whole brain, in males, virgin and mated females in the event there were sexual and/or postmating differences in postsynaptic connectivity. Trans-synaptic labeling revealed that pdf+ ventral lateral neurons, both the l-LNvs and s-LNvs, are postsynaptic targets to ppk+ cells in the three experimental groups (Figs 4D–4F and S3). We also observed a clear postsynaptic mtdTomato labeling in several unidentified neurons within the suboesophageal ganglia in male and female brains (Figs 4D–4F and S3). The subesophageal ganglion is a region within the CNS that is likely to contain circuits that mediate behavioral responses to mating [17,18]. In addition, it has been reported that ppk+ projections are in close proximity to the somas of DN1 clock neurons, which are required to control the timing (onset) of the siesta in response to temperature [29]. Although trans-Tango provided evidence of direct synaptic contact between ppk+ and pdf+ neurons, it failed to detect postsynaptic labeling within DN1 somas as previously reported, [29] despite additional–yet unidentified- postsynaptic targets became apparent at the dorsal protocerebrum. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Expression pattern of the ppk-Gal4 line used in this work. (A, B, C) Confocal immunostaining of a representative brain and ventral nerve cord (A), ovaries (B) and uterus (C) of ppk-Gal4>UASmCD8GFP females stained with anti-GFP (green). Asterisks indicate the positions of stained ppk+ somas. Scale bar 100 μm for ovaries and 10 μm for the uterus. (D) Schematic diagram showing the location of ppk+ neurons. https://doi.org/10.1371/journal.pgen.1010258.g003 Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. pdf+ LNv neurons are postsynaptic targets of ppk+ neurons. (A) The dendritic arbors and presynaptic terminals of ppk+ neurons were visualized by expression of the postsynaptic marker DenMark (orange) and the presynaptic marker syt-GFP (green), in a mated female brain. pdf+ neurons were visualized with anti-PDF (red). The bar indicates 10 μM. (B) A higher magnification of the dashed region shown in (A). (B1,B2,B3) Brains of flies co-expressing UAS-DenMark and UAS-syt-GFP reporters in ppk+ neurons, stained with anti-PDF (B1), with anti-DsRed (DenMark, orange, postsynaptic) (B2) and with anti-GFP (green, presynaptic) (B3). (C1,C2,C3) Single plane confocal images of the region (C) highlighted in panel (B) displaying the area in the protocerebrum where the presynaptic termini of ppk+ neurons are in close apposition to pdf+ neurons. Scale bars, 10 μm. (D) Trans-synaptic labeling using trans-Tango. Representative confocal image of a ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango mated female brain. Brains were co-stained with anti-PDF (cyan) to identify pdf+ neurons, anti GFP (green) to highlight ppk+ presynaptic neurons and anti-DsRed (red) to identify putative ppk postsynaptic partners. The bar indicates 60 μm. (E, F) Higher magnification of the dorsal (E) and accessory medulla (F) regions of a mated female brain. The rightmost image of B and C panel shows single focal planes of an overlap between presynaptic and postsynaptic ppk+ contacts at PDF-expressing neurons in the accessory medulla (C) and in the dorsal termini of pdf+ axons (B). Bars indicate 20 μm. https://doi.org/10.1371/journal.pgen.1010258.g004 Overall, these results confirm a contact between ppk+ and pdf+ neurons suggesting a direct interaction between circuits involved in reproduction and the circadian clock. We next inquired what would be the impact of this communication on the circadian clock. Mating reduces PDF levels in females The sLNvs and ILNvs are required for normal circadian behavior under free running conditions and are the only neurons within the circadian network that express the PDF neuropeptide [4–6]. PDF levels at the dorsal sLNv terminals as well as in the somas oscillates in a circadian fashion; at dawn, levels are high while at dusk they are low [34]. PDF released from the sLNvs appears to be responsible for the maintenance of free-running activity rhythms [35]. Ablation of the LNvs and/or PDF results in the loss of morning anticipation, an advance in the evening activity peak under LD cycles, and a decrease in the free-running period [4–6]. Thus, we reasoned that the phenotype associated to the loss of morning anticipation in mated females could possibly involve alterations in PDF levels, cycling, or both. To test this idea, we evaluated PDF levels by performing immunofluorescence in whole brains of mated and virgin females as well as in males. Fig 5A–5F shows that PDF immunoreactivity in males and virgin females displays a normal cycling pattern; however, PDF levels at the axonal termini are particularly reduced in mated females, especially in the morning (albeit some residual cycling is still observed), suggesting that the mating state alters the circadian modulation of PDF levels (Fig 5G). To confirm whether reduced PDF levels in mated females could be mediated by SP signaling we knocked-down SPR in ppk+ neurons. Fig 5H show that this restores PDF cycling in mated females and causes a significant increase of PDF levels in the morning when compared to mated controls. In fact, Fig 5H also shows that PDF levels are like those observed in matched genetic controls as well as in ppk-Gal4>SPR-IR1 males. Surprisingly, we also noticed that virgin females devoid of SPR in ppk+ neurons displayed a significant decrease in PDF levels when compared to control virgin females. Overall, the results for mated females indicate that PDF levels are responsive to postmating regulation in females. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. PDF levels are reduced in mated females (A, C, E) Confocal images of representative sLNv dorsal projections of individual male, virgin and mated female flies showing their PDF content during the day (top) and night (bottom). Left: control flies (UAS-Dicer2, SPR-IR1>+). Right: flies with SPR downregulated in ppk+ neurons (ppk-Gal4>UAS-Dicer2, SPR-IR1). Brains were dissected at ZT02 and ZT14 and standard anti-PDF immunofluorescence detection was performed. The bar indicates 10 μm. (B, D, F) PDF quantitation of the sLNv dorsal projections for the twelve conditions mentioned above. Dots represent individual brains. B, Males (n = 31–37); D, Mated females (n = 26–33), F, Virgin females (n = 27–33). Two-way ANOVA with Tukey’s post hoc tests. (G and H) Quantification of PDF signal intensity at day (ZT02) of control (UAS-Dicer2, SPR-IR1>+) and SPR downregulation in ppk+ neurons (ppk-Gal4>UAS-Dicer2, SPR-IR1) respectively. Kruskal-Wallis test with Dunn’s multiple comparisons test, p < 0.0001. Statistically significant differences were represented as *p< 0.05, **p< 0.01, ***p< 0.001, ****p < 0.0001. ns, not significant. https://doi.org/10.1371/journal.pgen.1010258.g005 Mating induces loss of morning anticipation in mated females Locomotor activity has been extensively studied in males, but much less in female (mated or virgin) flies, and most published studies were performed using the Drosophila Activity Monitor tracking system (DAM, Trikinetics). On this system, flies are housed in small tubes (¬320 mm3), and their movement is recorded only when they cross an infrared beam. It has been reported [26] though, that activity records differ between DAM and the more accurate video tracking systems. Moreover, activity of mated flies depends on the size of the arena where they are placed [26]. Thus, in order to provide mated females with a more adequate environment where movement can be accurately recorded, we developed a system where flies are placed in a set of relatively large transparent chambers (80 x 8 x 8 mm), and are tracked using a video system (for a detailed description please refer to the Methods section). In order to validate our setup, we compared the activity recorded by our video tracking method, which directly measures activity as distance travelled per second, with recordings obtained using the DAM system, which estimates activity as the number of interruptions of a single infrared beam, firstly in male flies. Both methods show that flies display bimodal locomotor activity patterns over several 12 h: 12 h LD cycles, and no significant differences were found in the morning anticipation activity between both systems (S1A–S1C Fig). However, a close examination of their sleep pattern revealed that flies seem to sleep significantly more when locomotor activity is monitored using the DAM system than when they are registered using our video tracking method (S1D and S1E Fig). This is probably because the DAM system fails to detect a portion of the activity that flies display along the day, for instance when they stand on one side of the tube, without crossing the infrared beam, likely resulting in an overestimation of sleep duration. These results confirm that even though single-fly recordings generated either by the DAM system and our system produce highly similar locomotor activity profiles, validating our experimental setup, video tracking acquisition provides more accurate register of the activity patterns. Next, we compared the locomotor activity of males, virgin and mated females of different control laboratory lines, white (w1118), Oregon R (OreR) and Canton-S (CS) under 12 h: 12 h LD cycles. Fig 1A shows the average activity profile (AAP) for males, virgin and mated females of w1118 flies. Male flies have two pronounced peaks of daily activity that start in anticipation to the lights-on and lights-off transitions, separated by a relatively long interval with very low activity (siesta). In contrast, mated females displayed a sustained and robust activity during daytime while at night the activity remains as low as in males. Virgin females, on the other hand, are more active than males during daytime but less so than mated females [13,14]. A closer visual inspection of AAPs showed that males and virgin females have a gradual increase in activity before lights-on (termed morning anticipation), whereas mated females did not display this morning anticipation, but they had a high amplitude morning startle response coincident with lights-on (Fig 1B). A more quantitative assessment of the degree of anticipation is given by the morning anticipation index (MAI, defined in Materials and Methods) [27]. Fig 1C shows a significant reduction in MAI in mated females compared with males supporting a strongly dimorphic trait. Additionally, the figure shows that mated females display a significant reduction in MAI when compared to virgin females, to the point that morning anticipation is virtually suppressed (on average), suggesting a postmating effect. Interestingly, all three control strains showed a reduced morning anticipation in mated females, supporting the idea that this effect occurs across different genetic backgrounds. Thus, our data shows that one of the postmating responses characteristic of mated females is the suppression of their ability to anticipate the night-day transition. In this article we focus on characterizing this postmating behavior. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. Mating suppresses morning anticipation in females of Drosophila melanogaster. (A) Average locomotor activity profile of white (w1118) flies in LD. The number in the horizontal axes refers to days after males were removed. (B) Zoom of the region enclosed by dashed lines in panel A to highlight the increase of the locomotor activity at the end of the night. (C) Anticipation indexes for males (blue), virgin (orange) and mated (red) females of white, CantonS and Oregon flies. For w1118 flies: males (n = 62), virgin females (n = 64) and mated females (n = 66). For CS flies: males (n = 39), virgin females (n = 37) and mated females (n = 43) and for Oregon flies: males (n = 34), virgin females (n = 35) and mated females (n = 37). Each dot corresponds to the average index of three complete days calculated for a single fly. The mean and SEM are shown as bars. Statistical analysis: Kruskal–Wallis test; for the MAI of white flies, X2 = 78.02, p< 0.0001; for the MAI of CS flies, X2 = 49.48, p< 0.0001; for the MAI of OreR flies, X2 = 44.43, p< 0.0001. Pairwise comparisons were performed using Dunn’s multiple comparisons test. Statistically significant differences are represented by *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. https://doi.org/10.1371/journal.pgen.1010258.g001 Loss of morning anticipation is mediated by sex peptide in ppk+ neurons Although SPR is detected broadly on the female reproductive tract, the ventral nerve cord and the brain [21], only a restricted subset of fru+/ppk+/dsx+ sex peptide sensory neurons (SPSN) expressing SPR in the reproductive system are necessary and sufficient to induce some SP-mediated postmating responses [17–19]. To examine whether SPSN neurons are involved in the modulation of morning anticipation through SP signaling, we evaluated the impact of SPR downregulation by RNA interference (RNAi) (21) on activity patterns of mated females. As expected, SPR downregulation in SPSN neurons induced a significant increase in morning anticipation compared with mated controls (Fig 2A). Some PMRs are controlled by a larger group of neurons than the SPSN. For example, oviposition recruits more dsx+ neurons that the ones present in the SPSN [19,20]. Likewise, fru+ neurons control many sexual behaviors in males and females [28]. Additionally, whereas SPSN neurons communicate with the brain using an indirect pathway (that passes through SAG neurons) it has been reported that some ppk+ neurons project directly to the brain [29], the main center of circadian control. For these reasons we decided to study separately the effect of the downregulation of SPR in each of these three neuronal groups. SPR knock down in fru+ and dsx+ neurons resulted in a significant increase in the MAI compared to mated controls (Fig 2B and 2C). This increase in MAI was similar to the value obtained when SPR expression was reduced in SPSN neurons. However, SPR knock down in ppk+ neurons resulted in a dramatic increase in this index (Fig 2D) suggesting a distinctive role of ppk+ neurons in the control of this postmating response. Remarkably, we did not observe any significant difference in morning anticipation when we downregulated SPR in ppk+ neurons in virgin females. Since all virgin genotypes have morning anticipation, the regulation by SPR is indeed mating-state-dependent (Fig 2E). These findings demonstrate the need for SPR function in SPSN neurons to respond to and transmit SP-generated signals to control morning anticipation in mated females and also support the hypothesis that additional spr+ ppk+ neurons beyond the SPSN domain play a role in the modulation of morning anticipation (Fig 2F). Interestingly these additional neurons do not seem to play any role in the control of the loss of daytime sleep (S2 Fig) which is a different PMR that has been reported to be regulated by the action of SP in the SPSN neurons [22]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Downregulation of SPR in ppk+ neurons restores morning anticipation. Morning anticipation index in mated females upon chronic SPR knock down in different groups of neurons and their controls. (A) SPSN-Gal4>UAS-Dicer2, SPR-IR1 (n = 45); SPSN-Gal4>+ (n = 44); UAS-Dicer2, SPR-IR1>+ (n = 44). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test: X2 = 36.50 p < 0.0001. (B) fru-Gal4>UAS-Dicer2, SPR-IR1: (n = 46), fru-Gal4>+ (n = 46); UAS-Dicer2, SPR-IR1>+ (n = 42). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test X2 = 15.38 p < 0.001. (C) dsxGal4>UAS-Dicer2, SPR-IR1 (n = 23); dsxGal4>+ (n = 23); UAS-Dicer2, SPR-IR1>+ (n = 22). One-way ANOVA with Tukey’s post hoc tests, F = 19.58 p < 0.0001. (D) Mated females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 72); ppk-Gal4>+ (n = 71); UAS-Dicer2, SPR-IR1>+ (n = 62). Statistical analysis: Kruskal-Wallis test with Dunn’s multiple comparisons test X2 = 107.5 p < 0.0001. (E) Virgin females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 41); ppk-Gal4>+ (n = 23); UAS-Dicer2, SPR-IR1>+ (n = 45). One-way ANOVA with Tukey’s post hoc tests F = 4 p < 0.0210. (F) Mated females: ppk-Gal4>UAS-Dicer2, SPR-IR1 (n = 25); SPSN-Gal4>UAS-Dicer2, SPR-IR1 (n = 20); UAS-Dicer2, SPR-IR1>+ (n = 24). One-way ANOVA with Tukey´s post hoc test F = 75.04 p < 0.00018. Dots represent independent flies, the mean and SEM are shown. Statistically significant differences are represented by *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns = not significant. https://doi.org/10.1371/journal.pgen.1010258.g002 Since the SPSN neurons contact the central brain indirectly, as it passes through neurons in the abdominal ganglion [17–19,30], and that some ppk+ neurons have been shown to project to the brain [29], we wondered if ppk+ neurons that control morning anticipation could provide an additional, more direct route for mating signals to reach the central brain. LNvs are postsynaptic targets of ppk+ neurons Our observations suggest that SPR activation mostly in ppk+ neurons alters the temporal organization of rest-activity cycles in mated females inducing the loss of morning anticipation, a feature regulated by the circadian clock. We next wondered how the signal provided by ppk+ neurons would reach the circadian network. Thus, we first analyzed the expression pattern of the ppk-Gal4 driver in adult females by expressing a membrane-bound green fluorescent protein (UAS-mCD8-GFP) under the control of ppk-Gal4. We observed GFP expression in the reproductive tract, in the abdominal ganglia as well as in a few somas in the dorsomedial region of the central brain (Fig 3). To visualize putative dendritic and axonal terminals of ppk+ neurons, both DenMark, a specific somatodendritic marker [31] and synaptotagmin (syt)-GFP, a presynaptic marker localized to synaptic vesicles [32], were expressed in mated females. ppk+ neurons displayed strong labelling of presynaptic syt-GFP, in the subesophageal ganglion zone (SEZ), and in the dorsal protocerebrum where they are in a close apposition to pdf+ terminals. Also, the somatodendritic compartment of ppk+ neurons displayed specific labelling in the dorsal protocerebrum and the SEZ (Fig 4A–4C and S1 Video). Then, to explore whether ppk+ neurons could provide direct input to circadian neurons, we employed the anterograde trans-synaptic tracing tool trans-Tango [33] to define specific downstream synaptic partners of ppk+ neurons. Expression of the tethered trans-Tango ligand under a Gal4 neuronal pattern of choice triggers mtd-Tomato expression in postsynaptic targets. We co-expressed the trans-Tango ligand and GFP to label pre-synaptic ppk+ neurons, and mtd-Tomato in the whole brain, in males, virgin and mated females in the event there were sexual and/or postmating differences in postsynaptic connectivity. Trans-synaptic labeling revealed that pdf+ ventral lateral neurons, both the l-LNvs and s-LNvs, are postsynaptic targets to ppk+ cells in the three experimental groups (Figs 4D–4F and S3). We also observed a clear postsynaptic mtdTomato labeling in several unidentified neurons within the suboesophageal ganglia in male and female brains (Figs 4D–4F and S3). The subesophageal ganglion is a region within the CNS that is likely to contain circuits that mediate behavioral responses to mating [17,18]. In addition, it has been reported that ppk+ projections are in close proximity to the somas of DN1 clock neurons, which are required to control the timing (onset) of the siesta in response to temperature [29]. Although trans-Tango provided evidence of direct synaptic contact between ppk+ and pdf+ neurons, it failed to detect postsynaptic labeling within DN1 somas as previously reported, [29] despite additional–yet unidentified- postsynaptic targets became apparent at the dorsal protocerebrum. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Expression pattern of the ppk-Gal4 line used in this work. (A, B, C) Confocal immunostaining of a representative brain and ventral nerve cord (A), ovaries (B) and uterus (C) of ppk-Gal4>UASmCD8GFP females stained with anti-GFP (green). Asterisks indicate the positions of stained ppk+ somas. Scale bar 100 μm for ovaries and 10 μm for the uterus. (D) Schematic diagram showing the location of ppk+ neurons. https://doi.org/10.1371/journal.pgen.1010258.g003 Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. pdf+ LNv neurons are postsynaptic targets of ppk+ neurons. (A) The dendritic arbors and presynaptic terminals of ppk+ neurons were visualized by expression of the postsynaptic marker DenMark (orange) and the presynaptic marker syt-GFP (green), in a mated female brain. pdf+ neurons were visualized with anti-PDF (red). The bar indicates 10 μM. (B) A higher magnification of the dashed region shown in (A). (B1,B2,B3) Brains of flies co-expressing UAS-DenMark and UAS-syt-GFP reporters in ppk+ neurons, stained with anti-PDF (B1), with anti-DsRed (DenMark, orange, postsynaptic) (B2) and with anti-GFP (green, presynaptic) (B3). (C1,C2,C3) Single plane confocal images of the region (C) highlighted in panel (B) displaying the area in the protocerebrum where the presynaptic termini of ppk+ neurons are in close apposition to pdf+ neurons. Scale bars, 10 μm. (D) Trans-synaptic labeling using trans-Tango. Representative confocal image of a ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango mated female brain. Brains were co-stained with anti-PDF (cyan) to identify pdf+ neurons, anti GFP (green) to highlight ppk+ presynaptic neurons and anti-DsRed (red) to identify putative ppk postsynaptic partners. The bar indicates 60 μm. (E, F) Higher magnification of the dorsal (E) and accessory medulla (F) regions of a mated female brain. The rightmost image of B and C panel shows single focal planes of an overlap between presynaptic and postsynaptic ppk+ contacts at PDF-expressing neurons in the accessory medulla (C) and in the dorsal termini of pdf+ axons (B). Bars indicate 20 μm. https://doi.org/10.1371/journal.pgen.1010258.g004 Overall, these results confirm a contact between ppk+ and pdf+ neurons suggesting a direct interaction between circuits involved in reproduction and the circadian clock. We next inquired what would be the impact of this communication on the circadian clock. Mating reduces PDF levels in females The sLNvs and ILNvs are required for normal circadian behavior under free running conditions and are the only neurons within the circadian network that express the PDF neuropeptide [4–6]. PDF levels at the dorsal sLNv terminals as well as in the somas oscillates in a circadian fashion; at dawn, levels are high while at dusk they are low [34]. PDF released from the sLNvs appears to be responsible for the maintenance of free-running activity rhythms [35]. Ablation of the LNvs and/or PDF results in the loss of morning anticipation, an advance in the evening activity peak under LD cycles, and a decrease in the free-running period [4–6]. Thus, we reasoned that the phenotype associated to the loss of morning anticipation in mated females could possibly involve alterations in PDF levels, cycling, or both. To test this idea, we evaluated PDF levels by performing immunofluorescence in whole brains of mated and virgin females as well as in males. Fig 5A–5F shows that PDF immunoreactivity in males and virgin females displays a normal cycling pattern; however, PDF levels at the axonal termini are particularly reduced in mated females, especially in the morning (albeit some residual cycling is still observed), suggesting that the mating state alters the circadian modulation of PDF levels (Fig 5G). To confirm whether reduced PDF levels in mated females could be mediated by SP signaling we knocked-down SPR in ppk+ neurons. Fig 5H show that this restores PDF cycling in mated females and causes a significant increase of PDF levels in the morning when compared to mated controls. In fact, Fig 5H also shows that PDF levels are like those observed in matched genetic controls as well as in ppk-Gal4>SPR-IR1 males. Surprisingly, we also noticed that virgin females devoid of SPR in ppk+ neurons displayed a significant decrease in PDF levels when compared to control virgin females. Overall, the results for mated females indicate that PDF levels are responsive to postmating regulation in females. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. PDF levels are reduced in mated females (A, C, E) Confocal images of representative sLNv dorsal projections of individual male, virgin and mated female flies showing their PDF content during the day (top) and night (bottom). Left: control flies (UAS-Dicer2, SPR-IR1>+). Right: flies with SPR downregulated in ppk+ neurons (ppk-Gal4>UAS-Dicer2, SPR-IR1). Brains were dissected at ZT02 and ZT14 and standard anti-PDF immunofluorescence detection was performed. The bar indicates 10 μm. (B, D, F) PDF quantitation of the sLNv dorsal projections for the twelve conditions mentioned above. Dots represent individual brains. B, Males (n = 31–37); D, Mated females (n = 26–33), F, Virgin females (n = 27–33). Two-way ANOVA with Tukey’s post hoc tests. (G and H) Quantification of PDF signal intensity at day (ZT02) of control (UAS-Dicer2, SPR-IR1>+) and SPR downregulation in ppk+ neurons (ppk-Gal4>UAS-Dicer2, SPR-IR1) respectively. Kruskal-Wallis test with Dunn’s multiple comparisons test, p < 0.0001. Statistically significant differences were represented as *p< 0.05, **p< 0.01, ***p< 0.001, ****p < 0.0001. ns, not significant. https://doi.org/10.1371/journal.pgen.1010258.g005 Discussion In most animals, many behaviors display a high degree of temporal organization, even in the absence of any temporal cues from the environment, which is a clear indication of the presence of an endogenous circadian clock. Fruit flies have two daily bouts of consolidated activity, the first of which starts some time before lights on, under laboratory conditions, a property that is known as morning anticipation. Morning anticipation is a trait that has been extensively characterized in Drosophila males [36]. However, it has not been explored in much depth in females. Particularly, we noticed that virgin females display less consolidated morning anticipation compared to males (Fig 1), which contributes to a certain degree of variability when assessing this trait: any single day does not robustly describe the ability to anticipate the coming transition. However, when this trait is scored over the course of several days, morning anticipation is indeed present in virgin females (Fig 1C). After mating females display important changes in their behavioral repertoire, such as an increase in the oviposition rate and a reduction of daytime sleep [11–14]. We show here that the temporal organization of locomotor behavior is also significantly altered after mating, since females lose their ability to anticipate the incoming morning (Fig 1). As it has been described with most postmating responses [17–19], such loss of anticipation can probably be related to the transfer of the sex peptide during copulation. It has been shown that many neurons have sex peptide receptors, which gives them the ability of initiating various postmating responses. Here we describe that the spr+ ppk+ neurons are responsible for the suppression of morning anticipation. Furthermore, we show that the SPR activation in these neurons correlates with a decrease in the concentration of PDF, a key circadian neuropeptide, which is known to be actively involved in clock control of the morning peak of activity in Drosophila. Suppression of morning anticipation as a postmating response Even though during copulation, males transfer more than 100 proteins to receptive females [37], SP is largely responsible for most postmating changes described in Drosophila. Most SP-dependent responses, including the loss of morning anticipation described here, are mediated by a small group of sensory neurons (SPSN) located within the female genital tract that co-express ppk, dsx and fru genes [17–19]. However, the diverse properties of neurons that separately express ppk, fru or dsx make it unlikely that the SP response is only regulated by neurons that co-express the three genes. In fact, egg laying and receptivity are differently affected by manipulations of subsets of these neurons [17–20,30,38]. This appears to be the case with the loss of morning anticipation after mating. Indeed, our results show that this PMR is also mediated by the action of SPR in ppk+ neurons, since silencing SPR expression in them restores to a greater extent the ability to anticipate light/dark transitions in mated females. In SPSN, fru+ or dsx+ neurons, SPR downregulation achieved a much smaller effect, pointing to a more relevant SPR function in ppk+ neurons for this behavior (Fig 2). On the other hand, when we analyzed another PMR such as the loss of daytime sleep [22], SPR knock down in ppk+ and SPSN neurons produces the same daytime sleep recovery (S2 Fig). These results reinforce the idea that distinct SPR neurons regulate different postmating responses [20,39]. An intriguing question is how the signals triggered by mating reach higher-order neurons in the brain to coordinate the sensory integration of the postmating response. Mating signals can act by two major pathways [20,30]. One is through SPSN neurons that can detect SP and alter its activation rate accordingly. This propagates the mating signal to the abdominal ganglia where, through direct synaptic contact, they silence the SAG neurons, which in turn relay mating information to the brain [17–19,30]; alternatively, the sex peptide can enter the hemolymph and act in higher centers of the brain [20]. Very recently, it was shown that this last pathway is being used to transmit mating information to the brain to regulate female receptivity [39]. In this work we suggest an additional PMR that could be using, at least partially, this second pathway through which SP could induce a postmating response: SPR activation in ppk+ neurons, probably in the brain, provides synaptic inputs to pdf+ neurons that would then drive morning anticipation in a manner dependent both on the clock and on mating status. Interestingly, we also observed this connection between ppk+ and pdf+ neurons in male brains, underscoring that this connection is not a sexually dimorphic feature (Figs 4 and S3). The exact identity of the ppk+ neurons that are presynaptic to pdf+ neurons could not be determined since the ppk-Gal4 driver employed is expressed in the reproductive tract, in the abdominal ganglia as well as in some brain somas (Fig 3). However, our data (Fig 4 and S1 Video) strongly suggests that some of the presynaptic contacts observed might correspond to projections from ppk+ neurons in the brain. Interestingly, very recently it was shown that ppk+ neurons in the brain control receptivity in mated females [39]. Thus, it is tempting to hypothesize that ppk+ neurons in the brain can also control, at least partially, the suppression of morning anticipation. Further work is necessary to pinpoint which specific ppk+ neurons contact pdf+ neurons directly. SP receptor plays a dual role in the temporal organization of locomotor behavior The optimal time of day to engage in a particular behavior can vary depending upon environmental factors, and according to the internal physiology of the animal. In Drosophila, PDF differentially coordinates the activity of circadian neuronal groups to optimize behavioral output [5,7,40–42]. PDF immunoreactivity at the axonal terminals of the sLNv oscillates in a circadian fashion [34]. We show here that mated females have reduced levels of PDF compared with virgin females and males, and these levels can be restored when SPR expression is reduced in ppk+ neurons (Fig 5). This is expected because SP activates the inhibitory G-protein coupled receptor (GPCR) SPR in the SPSN, which upon activation induces PMR by silencing the SPSN directly and the SAG neurons indirectly [30]. Additionally, we hypothesize that some ppk+ feed excitatory synaptic inputs to pdf+ neurons, and that the presence of SP silences ppk+ neurons, which in turn weakens or suppresses excitatory inputs to pdf+ neurons, inducing a decrease in PDF levels in the dorsal terminals [43]. Further work will be necessary to confirm this hypothesis. In virgin females PDF cycling/levels resembles the cycling shown in males; however, SPR downregulation in ppk+ neurons results in decreased PDF levels, in contrast to the effect observed in mated females (Fig 5). Although understanding this regulation is beyond the scope of the present manuscript, it is tempting to speculate that reduced SPR levels impairs the binding of additional neuropeptide/s that might also contribute to the neuronal communication among ppk+ and pdf+ neurons. This putative neuropeptide would likely be a GPCR ligand, just as SP is, but it may recruit different signaling pathways downstream of SPR. One attractive candidate is the myoinhibitory peptide (Mip), a potent agonist for SPR [44,45]. Like other neuromodulators, Mip is highly pleiotropic, and it has been shown to modulate behaviors as diverse as sleep, feeding and mating receptivity [46–48]. SPR mediates the sleep function of Mip but is dispensable for its feeding function and normal mating in virgin females [47,48]. These findings should be thoroughly characterized in the future, as they suggest a different regulation of the excitability of sLNv through ppk+ neurons in virgin and mated females. To the best of our knowledge, the loss of morning anticipation is the first postmating response that can be clearly ascribed to the circadian clock. The obvious question that arises is what could be the benefit of losing the ability to anticipate the end of the night. Fujii and colleagues (2007) showed that male-female couples are highly active throughout the night and early morning, and this locomotor activity rhythm is associated with courtship [49]. Thus, it is possible that, once mated, females lose the ability to anticipate dawn in order to reduce the chance of encounters with other males thus saving their energy for the daytime hours, which are probably better suited for feeding and oviposition activities, including searching for egg-laying sites. More generally, it is tempting to hypothesize that both the loss of morning anticipation and the suppression of the siesta could be part of a larger strategy of downregulating the circadian clock after mating, in order to restrict activity to daylight hours- whenever they come. Further confirmation of this hypothesis would require experiments to ascertain the preference for daytime over nighttime for the typical activities displayed by mated females (such as oviposition, feeding, searching for suitable egg-laying sites, etc.). Establishing the ecological relevance of the suppression of morning anticipation is particularly difficult in the case of natural conditions because the transition to lights-on is continuous, complicating the definition of morning anticipation. A similar observation can be made regarding the relevance of the suppression of the siesta, since under seminatural conditions flies show an afternoon peak at approximately the same times when the siesta is observed under laboratory conditions [50,51]. In fact, it will be interesting to study how these PMRs appear in the case of seminatural conditions. In our opinion, the main value of this new PMR (suppression of morning anticipation) lies in its unambiguous relationship to the endogenous clock, which makes it a useful tool to probe the influence of mating on the female circadian clock. Other postmating responses, although stronger, do not have such a clear relationship with the circadian clock. For example, the suppression of the siesta evidently depends on the system that regulates sleep, which in turn depends not only on circadian but also on homeostatic processes [24]. As further extensions of this work it will be interesting to study the dependence of this PMR with temperature, and with time. After the initial suppression of the siesta, mated females begin to recover this ability after about 7 days [14]. We monitored the loss of morning anticipation for 4 days after male removal and did not observe any decrease in this PMR, but it would be interesting to study whether it disappears after a longer period of time. To summarize, we show here that mating impairs the ability to anticipate dawn in females. This is probably also achieved by SPR-mediated silencing a subset of ppk+ neurons, which in turn acts on the circadian release of PDF, thus obliterating a major signal that times locomotor activity to specific windows along the day. In nature, signaling of a wide range of sensory modalities along with internal cues must concomitantly be considered to match the onset of activity with environmental conditions. Our work provides a framework to unravel how mating triggered signals impinge upon clock neurons in the Drosophila nervous system and modify the dynamic control of activity. Suppression of morning anticipation as a postmating response Even though during copulation, males transfer more than 100 proteins to receptive females [37], SP is largely responsible for most postmating changes described in Drosophila. Most SP-dependent responses, including the loss of morning anticipation described here, are mediated by a small group of sensory neurons (SPSN) located within the female genital tract that co-express ppk, dsx and fru genes [17–19]. However, the diverse properties of neurons that separately express ppk, fru or dsx make it unlikely that the SP response is only regulated by neurons that co-express the three genes. In fact, egg laying and receptivity are differently affected by manipulations of subsets of these neurons [17–20,30,38]. This appears to be the case with the loss of morning anticipation after mating. Indeed, our results show that this PMR is also mediated by the action of SPR in ppk+ neurons, since silencing SPR expression in them restores to a greater extent the ability to anticipate light/dark transitions in mated females. In SPSN, fru+ or dsx+ neurons, SPR downregulation achieved a much smaller effect, pointing to a more relevant SPR function in ppk+ neurons for this behavior (Fig 2). On the other hand, when we analyzed another PMR such as the loss of daytime sleep [22], SPR knock down in ppk+ and SPSN neurons produces the same daytime sleep recovery (S2 Fig). These results reinforce the idea that distinct SPR neurons regulate different postmating responses [20,39]. An intriguing question is how the signals triggered by mating reach higher-order neurons in the brain to coordinate the sensory integration of the postmating response. Mating signals can act by two major pathways [20,30]. One is through SPSN neurons that can detect SP and alter its activation rate accordingly. This propagates the mating signal to the abdominal ganglia where, through direct synaptic contact, they silence the SAG neurons, which in turn relay mating information to the brain [17–19,30]; alternatively, the sex peptide can enter the hemolymph and act in higher centers of the brain [20]. Very recently, it was shown that this last pathway is being used to transmit mating information to the brain to regulate female receptivity [39]. In this work we suggest an additional PMR that could be using, at least partially, this second pathway through which SP could induce a postmating response: SPR activation in ppk+ neurons, probably in the brain, provides synaptic inputs to pdf+ neurons that would then drive morning anticipation in a manner dependent both on the clock and on mating status. Interestingly, we also observed this connection between ppk+ and pdf+ neurons in male brains, underscoring that this connection is not a sexually dimorphic feature (Figs 4 and S3). The exact identity of the ppk+ neurons that are presynaptic to pdf+ neurons could not be determined since the ppk-Gal4 driver employed is expressed in the reproductive tract, in the abdominal ganglia as well as in some brain somas (Fig 3). However, our data (Fig 4 and S1 Video) strongly suggests that some of the presynaptic contacts observed might correspond to projections from ppk+ neurons in the brain. Interestingly, very recently it was shown that ppk+ neurons in the brain control receptivity in mated females [39]. Thus, it is tempting to hypothesize that ppk+ neurons in the brain can also control, at least partially, the suppression of morning anticipation. Further work is necessary to pinpoint which specific ppk+ neurons contact pdf+ neurons directly. SP receptor plays a dual role in the temporal organization of locomotor behavior The optimal time of day to engage in a particular behavior can vary depending upon environmental factors, and according to the internal physiology of the animal. In Drosophila, PDF differentially coordinates the activity of circadian neuronal groups to optimize behavioral output [5,7,40–42]. PDF immunoreactivity at the axonal terminals of the sLNv oscillates in a circadian fashion [34]. We show here that mated females have reduced levels of PDF compared with virgin females and males, and these levels can be restored when SPR expression is reduced in ppk+ neurons (Fig 5). This is expected because SP activates the inhibitory G-protein coupled receptor (GPCR) SPR in the SPSN, which upon activation induces PMR by silencing the SPSN directly and the SAG neurons indirectly [30]. Additionally, we hypothesize that some ppk+ feed excitatory synaptic inputs to pdf+ neurons, and that the presence of SP silences ppk+ neurons, which in turn weakens or suppresses excitatory inputs to pdf+ neurons, inducing a decrease in PDF levels in the dorsal terminals [43]. Further work will be necessary to confirm this hypothesis. In virgin females PDF cycling/levels resembles the cycling shown in males; however, SPR downregulation in ppk+ neurons results in decreased PDF levels, in contrast to the effect observed in mated females (Fig 5). Although understanding this regulation is beyond the scope of the present manuscript, it is tempting to speculate that reduced SPR levels impairs the binding of additional neuropeptide/s that might also contribute to the neuronal communication among ppk+ and pdf+ neurons. This putative neuropeptide would likely be a GPCR ligand, just as SP is, but it may recruit different signaling pathways downstream of SPR. One attractive candidate is the myoinhibitory peptide (Mip), a potent agonist for SPR [44,45]. Like other neuromodulators, Mip is highly pleiotropic, and it has been shown to modulate behaviors as diverse as sleep, feeding and mating receptivity [46–48]. SPR mediates the sleep function of Mip but is dispensable for its feeding function and normal mating in virgin females [47,48]. These findings should be thoroughly characterized in the future, as they suggest a different regulation of the excitability of sLNv through ppk+ neurons in virgin and mated females. To the best of our knowledge, the loss of morning anticipation is the first postmating response that can be clearly ascribed to the circadian clock. The obvious question that arises is what could be the benefit of losing the ability to anticipate the end of the night. Fujii and colleagues (2007) showed that male-female couples are highly active throughout the night and early morning, and this locomotor activity rhythm is associated with courtship [49]. Thus, it is possible that, once mated, females lose the ability to anticipate dawn in order to reduce the chance of encounters with other males thus saving their energy for the daytime hours, which are probably better suited for feeding and oviposition activities, including searching for egg-laying sites. More generally, it is tempting to hypothesize that both the loss of morning anticipation and the suppression of the siesta could be part of a larger strategy of downregulating the circadian clock after mating, in order to restrict activity to daylight hours- whenever they come. Further confirmation of this hypothesis would require experiments to ascertain the preference for daytime over nighttime for the typical activities displayed by mated females (such as oviposition, feeding, searching for suitable egg-laying sites, etc.). Establishing the ecological relevance of the suppression of morning anticipation is particularly difficult in the case of natural conditions because the transition to lights-on is continuous, complicating the definition of morning anticipation. A similar observation can be made regarding the relevance of the suppression of the siesta, since under seminatural conditions flies show an afternoon peak at approximately the same times when the siesta is observed under laboratory conditions [50,51]. In fact, it will be interesting to study how these PMRs appear in the case of seminatural conditions. In our opinion, the main value of this new PMR (suppression of morning anticipation) lies in its unambiguous relationship to the endogenous clock, which makes it a useful tool to probe the influence of mating on the female circadian clock. Other postmating responses, although stronger, do not have such a clear relationship with the circadian clock. For example, the suppression of the siesta evidently depends on the system that regulates sleep, which in turn depends not only on circadian but also on homeostatic processes [24]. As further extensions of this work it will be interesting to study the dependence of this PMR with temperature, and with time. After the initial suppression of the siesta, mated females begin to recover this ability after about 7 days [14]. We monitored the loss of morning anticipation for 4 days after male removal and did not observe any decrease in this PMR, but it would be interesting to study whether it disappears after a longer period of time. To summarize, we show here that mating impairs the ability to anticipate dawn in females. This is probably also achieved by SPR-mediated silencing a subset of ppk+ neurons, which in turn acts on the circadian release of PDF, thus obliterating a major signal that times locomotor activity to specific windows along the day. In nature, signaling of a wide range of sensory modalities along with internal cues must concomitantly be considered to match the onset of activity with environmental conditions. Our work provides a framework to unravel how mating triggered signals impinge upon clock neurons in the Drosophila nervous system and modify the dynamic control of activity. Material and methods Fly strains All fly strains used in this study are detailed in Table 1. Flies were reared and maintained on standard cornmeal/agar medium at 25°C and 60% humidity in a 12 hr:12 hr LD cycle unless stated otherwise. Per vial, six 0–5 day-old females were mated with 3 males during 72 h. After that period males were discarded and mated females were used for locomotor experiments. 5–8 day-old adult males, virgin or mated females were used for all locomotor activity experiments. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. List of fly stocks, chemicals and antibody used throughout this study. https://doi.org/10.1371/journal.pgen.1010258.t001 Video-based acquisition systems and locomotor behavioral analysis Flies were entrained to 12:12h LD cycles during their entire development, and males, virgin and mated females were individually housed in open ended transparent chambers measuring 80 x 8 x 8 mm, containing approx. 1ml of banana medium [54] in one end and a cotton plug in the other end. A collective enclosure containing 26 of such chambers was built with two transparent acrylic boards (forming the floor and roof of all chambers) and white-opaque acrylic separations between chambers (to avoid visual contact between flies). The enclosures were placed inside an incubator (set at 25°C) over a translucid (not transparent) white plaque, and were illuminated from below with an array of white and infrared LEDs (850nm), in order to record activity both in LD and in constant darkness. Each enclosure was recorded from a distance of 18 cm, with two webcams whose infrared filters were manually removed. Locomotor activity was monitored for 3–4 days in LD conditions. Data from the webcams was analyzed in a laptop computer (running Ubuntu, a Linux distribution) using custom made software (written in Python3 and using OpenCV libraries). The software implements a tracking paradigm: from the video signal the position of each fly is extracted in real time. The output is a text file containing a pair of coordinates (x, y) for each fly, sampled at one second intervals. These files are then processed with an analysis software we developed (in Bash) which provides statistics for activity (distance traveled, morning anticipation indexes, etc.) and sleep (daytime and nighttime sleep duration, number and duration of sleep bouts, etc.). Sleep is defined as the absence of "significant activity" (defined as a displacement of at least one body length per second) for at least 5 consecutive minutes. The Morning Anticipation Index (MAI) is defined as the ratio between the total activity in ZT21-24 and the total activity in ZT 18–24 [26]. Given that there is a large variability in the MAI across different days in individual flies, we averaged the MAI of the first three complete days for every experiment for a more robust quantitation. All the software used is open-source and freely available upon request. Immunofluorescence detection Adult flies were fixed with 4% p-formaldehyde (pH 7.5) for 60 min at room temperature. Brains were dissected and rinsed six times in PT buffer (PBS with 0.1% Triton X-100) for 30 min. Samples were incubated with rat anti-PDF (1:500; (45)) in 7% normal goat serum at 4°C for two days. Next, samples were washed in PT 6x10 min, and incubated with Cy5-conjugated anti-rat (1:500; Jackson ImmunoResearch, USA) for 2h at room temperature. Samples were washed 4x15 min in PT and mounted in Vectashield antifade mounting medium (Vector Laboratories, USA). For Trans-Tango staining ppk-Gal4 males were crossed with trans-Tango females and kept at 25°C. Immediately after eclosion, adult males, virgin and mated females were separated from the progeny and aged for 15 days at 18°C. The immunofluorescence procedure was the same as described above, except for the length of the incubation with primary antibody, which was extended to 5 days at 4°C. The following primary antibodies were used: rabbit anti-DsRed (1:1000, Rockland), chicken anti-GFP (1:1000, Aves Labs) and rat anti-PDF (1:1000; (45)). The following secondary antibodies were used: Cy2-conjugated anti-chicken, Cy5-conjugated anti-rat, and Cy3-conjugated anti-rabbit (1:500, Jackson ImmunoResearch Laboratories, Inc). Images were acquired with a ZEISS LSM 880 Confocal Laser Scanning Microscope. Images from Fig 4A–4C were acquired with a ZEISS LSM 980 Confocal Laser Scanning Microscope. PDF levels quantification For the quantification of PDF intensity at the sLNv projections, we assembled a maximum intensity z-stack that contains the whole projection (approximate 10 images) and constructed a threshold image to create a ROI to measure immunoreactivity intensity using ImageJ (NIH) [43]. Data was analyzed with GraphPad Prism. Statistical analysis The following statistical analyses were used in this study: one-way ANOVA and two-way ANOVA with post hoc Tukey’s or Holm-Sidak’s test for multiple comparisons of parametric data, and non-parametric Kruskal-Wallis statistical analysis with multiple comparisons as specified in figure legends. Parametric tests were used when data were normally distributed and showed homogeneity of variance, tested by D’Agostino & Pearson test. When data was not normally distributed we applied a log transformation, and checked again for normality. Sidak’s and Dunn’s multiple comparisons tests were performed after parametric and non-parametric ANOVA when GraphPad software was used. A p value < 0.05 was considered statistically significant. Throughout the manuscript n represents the total number of measurements compared in each experimental group (behaviour of an individual or brain), and N represents the number of independent times an experiment was repeated. In dot plots for MAI, daytime sleep and immunofluorescence quantification lines represent the mean value; error bars depict the standard error of the mean. No statistical methods were used to determine sample size. Sample sizes are similar to those generally used in this field of research. Samples were not randomized and analyzers were not blind to the experimental conditions. In all the figures we show results of two or three independent experiments. Fly strains All fly strains used in this study are detailed in Table 1. Flies were reared and maintained on standard cornmeal/agar medium at 25°C and 60% humidity in a 12 hr:12 hr LD cycle unless stated otherwise. Per vial, six 0–5 day-old females were mated with 3 males during 72 h. After that period males were discarded and mated females were used for locomotor experiments. 5–8 day-old adult males, virgin or mated females were used for all locomotor activity experiments. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. List of fly stocks, chemicals and antibody used throughout this study. https://doi.org/10.1371/journal.pgen.1010258.t001 Video-based acquisition systems and locomotor behavioral analysis Flies were entrained to 12:12h LD cycles during their entire development, and males, virgin and mated females were individually housed in open ended transparent chambers measuring 80 x 8 x 8 mm, containing approx. 1ml of banana medium [54] in one end and a cotton plug in the other end. A collective enclosure containing 26 of such chambers was built with two transparent acrylic boards (forming the floor and roof of all chambers) and white-opaque acrylic separations between chambers (to avoid visual contact between flies). The enclosures were placed inside an incubator (set at 25°C) over a translucid (not transparent) white plaque, and were illuminated from below with an array of white and infrared LEDs (850nm), in order to record activity both in LD and in constant darkness. Each enclosure was recorded from a distance of 18 cm, with two webcams whose infrared filters were manually removed. Locomotor activity was monitored for 3–4 days in LD conditions. Data from the webcams was analyzed in a laptop computer (running Ubuntu, a Linux distribution) using custom made software (written in Python3 and using OpenCV libraries). The software implements a tracking paradigm: from the video signal the position of each fly is extracted in real time. The output is a text file containing a pair of coordinates (x, y) for each fly, sampled at one second intervals. These files are then processed with an analysis software we developed (in Bash) which provides statistics for activity (distance traveled, morning anticipation indexes, etc.) and sleep (daytime and nighttime sleep duration, number and duration of sleep bouts, etc.). Sleep is defined as the absence of "significant activity" (defined as a displacement of at least one body length per second) for at least 5 consecutive minutes. The Morning Anticipation Index (MAI) is defined as the ratio between the total activity in ZT21-24 and the total activity in ZT 18–24 [26]. Given that there is a large variability in the MAI across different days in individual flies, we averaged the MAI of the first three complete days for every experiment for a more robust quantitation. All the software used is open-source and freely available upon request. Immunofluorescence detection Adult flies were fixed with 4% p-formaldehyde (pH 7.5) for 60 min at room temperature. Brains were dissected and rinsed six times in PT buffer (PBS with 0.1% Triton X-100) for 30 min. Samples were incubated with rat anti-PDF (1:500; (45)) in 7% normal goat serum at 4°C for two days. Next, samples were washed in PT 6x10 min, and incubated with Cy5-conjugated anti-rat (1:500; Jackson ImmunoResearch, USA) for 2h at room temperature. Samples were washed 4x15 min in PT and mounted in Vectashield antifade mounting medium (Vector Laboratories, USA). For Trans-Tango staining ppk-Gal4 males were crossed with trans-Tango females and kept at 25°C. Immediately after eclosion, adult males, virgin and mated females were separated from the progeny and aged for 15 days at 18°C. The immunofluorescence procedure was the same as described above, except for the length of the incubation with primary antibody, which was extended to 5 days at 4°C. The following primary antibodies were used: rabbit anti-DsRed (1:1000, Rockland), chicken anti-GFP (1:1000, Aves Labs) and rat anti-PDF (1:1000; (45)). The following secondary antibodies were used: Cy2-conjugated anti-chicken, Cy5-conjugated anti-rat, and Cy3-conjugated anti-rabbit (1:500, Jackson ImmunoResearch Laboratories, Inc). Images were acquired with a ZEISS LSM 880 Confocal Laser Scanning Microscope. Images from Fig 4A–4C were acquired with a ZEISS LSM 980 Confocal Laser Scanning Microscope. PDF levels quantification For the quantification of PDF intensity at the sLNv projections, we assembled a maximum intensity z-stack that contains the whole projection (approximate 10 images) and constructed a threshold image to create a ROI to measure immunoreactivity intensity using ImageJ (NIH) [43]. Data was analyzed with GraphPad Prism. Statistical analysis The following statistical analyses were used in this study: one-way ANOVA and two-way ANOVA with post hoc Tukey’s or Holm-Sidak’s test for multiple comparisons of parametric data, and non-parametric Kruskal-Wallis statistical analysis with multiple comparisons as specified in figure legends. Parametric tests were used when data were normally distributed and showed homogeneity of variance, tested by D’Agostino & Pearson test. When data was not normally distributed we applied a log transformation, and checked again for normality. Sidak’s and Dunn’s multiple comparisons tests were performed after parametric and non-parametric ANOVA when GraphPad software was used. A p value < 0.05 was considered statistically significant. Throughout the manuscript n represents the total number of measurements compared in each experimental group (behaviour of an individual or brain), and N represents the number of independent times an experiment was repeated. In dot plots for MAI, daytime sleep and immunofluorescence quantification lines represent the mean value; error bars depict the standard error of the mean. No statistical methods were used to determine sample size. Sample sizes are similar to those generally used in this field of research. Samples were not randomized and analyzers were not blind to the experimental conditions. In all the figures we show results of two or three independent experiments. Supporting information S1 Fig. Comparisons between DAM and video acquisition systems. (A) Left: Tube where a fly is housed in the DAM system. Right: Set of chambers where flies are housed in our video tracking system. (B) Comparison between group average locomotor activity of male CantonS flies in LD, obtained using both systems. (C) Comparison between Morning Anticipation Index of male CantonS flies, obtained using DAM (red) and Video sytems (orange). Each dot corresponds to the index calculated for a single fly. Statistical analysis: Scheirer–Ray–Hare test. Post hoc tests Wilcoxon rank tests for every day, corrected for multiple comparisons (Benjamini-Hochberg). (D) Comparison between the times spent sleeping (in 30 minutes bins) of male CantonS flies obtained using both systems under LD conditions. (E) Comparison between total sleep of male CantonS flies obtained using DAM (red) and video (orange) systems under LD conditions. Each dot corresponds to the percentage of sleep per day calculated for a single fly. Statistical analysis: Scheirer–Ray–Hare test, as Post hoc tests we have applied Wilcoxon rank tests for every day, corrected for multiple comparisons (Benjamini-Hochberg). Dots represent independent flies, the mean and SEM are shown. *p< 0.05, **p< 0.01 ***p < 0.001. ns, not significant. https://doi.org/10.1371/journal.pgen.1010258.s001 (TIF) S2 Fig. SPR knock down in ppk+ or SPSN neurons increases daytime sleep. Total daytime sleep (ZT0-12) of the first and second day after male removal of mated females of the indicated genotypes. ****p < 0.0001; ***p ≤ 0.001; ns, not significant. https://doi.org/10.1371/journal.pgen.1010258.s002 (TIF) S3 Fig. pdf+ LNv neurons are postsynaptic targets of ppk+ neurons in males as in virgin females. Trans-synaptic labeling using trans-Tango. (A) Schematic diagram of a Drosophila brain. (B) Higher magnification of the dorsal region of ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango male brain. (C, D) Single focal plane of the image shown in B displaying the overlap of the PDF staining (cyan) with the postsynaptic ppk+ partners (red). E, Higher magnification of the accessory medulla region of ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango male brain. (F, G) Single focal plane of the image shown in E to exhibit the overlap of the PDF stain with the postsynaptic ppk partners (red). (H) Schematic diagram of a fly brain. (I) Higher magnification of the dorsal region of ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango virgin female brain. (J, K) Single focal plane of the image shown in I displaying the overlap of the PDF stain (cyan) with the postsynaptic ppk partners (red). (L) Higher magnification of the accessory medulla region of ppk-Gal4>UAS-myr-GFP, QUAS-mtdTom; trans-Tango virgin brain. (M, N) Single focal plane of the image shown in L to show the overlap of the PDF stain with the postsynaptic ppk partners (red). https://doi.org/10.1371/journal.pgen.1010258.s003 (TIF) S1 Video. Close Apposition of ppk+ and pdf+ Neurons. Confocal stacks showing overlap of arborizations between presynaptic terminals of ppk+ and pdf+ projection neurons in dorsal protocerebrum. The dendritic arbors and presynaptic terminals of ppk+ neurons were visualized by expression of the postsynaptic marker DenMark (orange) and the presynaptic marker syt-GFP (green), in a mated female brain. pdf neurons were visualized with anti-PDF (red) https://doi.org/10.1371/journal.pgen.1010258.s004 (MP4) S1 Table. All datasets and statistical analysis on which the conclusion are based. https://doi.org/10.1371/journal.pgen.1010258.s005 (XLT) Acknowledgments We thank Carolina Rezával and Lucas Mongiat for helpful discussions and critical reading of the manuscript. We are also grateful to Carolina Rezával and José Duhart for sharing fly stocks with us. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537), were used in this study.

Hampton, Brea K.;Plante, Kenneth S.;Whitmore, Alan C.;Linnertz, Colton L.;Madden, Emily A.;Noll, Kelsey E.;Boyson, Samuel P.;Parotti, Breantie;Xenakis, James G.;Bell, Timothy A.;Hock, Pablo;Shaw, Ginger D.;de Villena, Fernando Pardo-Manuel;Ferris, Martin T.;Heise, Mark T.

doi: 10.1371/journal.pgen.1010548

Rizzo, Marzia;Soisangwan, Natthapon;Vega-Estevez, Samuel;Price, Robert Jordan;Uyl, Chloe;Iracane, Elise;Shaw, Matt;Soetaert, Jan;Selmecki, Anna;Buscaino, Alessia

doi: 10.1371/journal.pgen.1010576pmid: 36574460

Introduction Understanding how organisms survive and thrive in changing environments is a fundamental question in biology. Genetic variation is central to environmental adaptation because it facilitates the selection of fitter genotypes better adapted to a new environment. Different types of genetic changes contribute to genetic variability, including (i) whole-chromosome or segmental-chromosome aneuploidy, (ii) translocations and (iii) mutations [1]. Furthermore, diploid cells can undergo loss of heterozygosity (LOH) driven by mitotic events, such as cross-over, gene conversion or meiotic reversion [1,2]. Whole-chromosome or segmental chromosome aneuploidies have the greatest effect on adaptation as they generate copy number variations (CNVs) of multiple genes. These result in divergent phenotypes which may be selectively advantageous [3]. Genome plasticity–the ability to generate large-scale genomic variation–is emerging as a critical adaptive mechanism in human microbial pathogens that need to adapt quickly to extreme environmental shifts because it provides genetic diversity upon which selection can act [4–8]. One such organism is Candida albicans, a common human fungal pathogen and a prevalent cause of death due to systemic fungal infections [9]. C. albicans is part of the normal microbiota of most healthy individuals but, in immunocompromised individuals, it is a dangerous pathogen causing a wide range of infections, including life-threatening disseminated diseases [10]. Azole antifungal agents, such as fluconazole (FLC), are the most commonly prescribed drugs for treating C. albicans infections [9,11,12]. Several lines of evidence suggest that C. albicans genome plasticity provides a competitive advantage under host-relevant stress environments. C. albicans is a diploid organism with a heterozygous genome organised into 2 × 8 (2n = 16) chromosomes (Chr) [13,14]. Seven chromosomes are designated Chr1 to Chr7 according to size, while one is termed ChrR because it contains the rDNA locus [15]. Genomic analysis of clinical isolates reveals that many C. albicans strains have large-scale genomic changes including segmental and whole chromosome aneuploidies [16–19]. Furthermore, specific chromosomal variants are selected during host-niche colonisation [17,20–25]. Accordingly, many drug-resistant isolates exhibit karyotypic diversity that can confer resistance due to increased copies of specific genes. For example, CNV for the gene ERG11 encoding for the target of FLC, lanosterol 14-alpha-demethylase is often observed in drug-resistant isolates [4,16,26–28]. Several studies suggest that C. albicans genome instability is not random as it occurs more frequently at specific hotspots which are often repetitive [17,19,23,29,30]. Subtelomeric regions and the rDNA locus are among the most unstable genomic sites [17,31]. C. albicans subtelomeric regions are enriched in repetitive sequences derived from transposons and protein-coding genes [29,32]. Most notable are the telomere-associated TLO genes, a family of 14 closely related paralogues encoding proteins similar to the Mediator 2 subunit of the Mediator transcriptional regulator [33–35]. Most TLO genes are located at subtelomeric regions except TLO34, located at an internal locus on the left arm of Chr1 [33]. The rDNA locus consists of a tandem array of a ~12 kb unit repeated 50 to 200 times; rDNA length polymorphisms frequently occur [14,17]. Despite the clear correlation between genomic variation and environmental adaptation, pinpointing the environmental pressure(s) selecting specific genotypes and understanding how complex karyotypes are formed is often difficult. In this study, we performed a genetic screening to identify modulators of C. albicans genome stability. The screen led to the identification of the ULP2 gene, encoding for a SUMO protease. ULP2 deletion causes increased genome instability and enhanced genome variation leading to fitness defects and hypersensitivity to genotoxic agents. We show that loss of ULP2 combined with exposure to an additional stress (FLC) leads to the selection of multichromosome segmental aneuploidies with adaptive power. Long-read genomic sequencing demonstrates that these novel segmental aneuploidies are selected by a two-step process producing chromosomal fragments with breakpoints at microhomology regions and DNA repeats. Thus, exposure to stress can increase tolerance to unrelated stress by selecting novel complex genotypes. Results A systematic genetic screen identifies Ulp2 as a regulator of C. albicans genotoxic stress response To identify factors regulating C. albicans genome integrity, we utilised a deletion library comprising a subset (674/3000) of C. albicans genes that are not conserved in other organisms or have a functional motif potentially related to virulence [36]. As defects in genome integrity lead to hypersensitivity to genotoxic agents [37], the deletion library was screened for hypersensitivity to two DNA damaging agents: Ultraviolet (UV) irradiation which induces formation of pyrimidine dimers [38], and methyl methanesulfonate (MMS), which leads to replication blocks and base mispairing [39]. Genotoxic stress hypersensitivity was semi-quantitatively scored by comparing the growth of treated versus untreated on a scale of 0 to 4, where 0 indicates no sensitivity, and 4 specifies strong hypersensitivity (Fig 1A). The screen identified 28 gene deletions linked to DNA damage hypersensitivity (UV or MMS score ≥2). Of those deletion mutants, 9/28 hits show sensitivity to both UV and MMS, 6/28 hits are sensitive only to UV, and 13/28 hits are sensitive only to MMS (S1 Table). Functional prediction analysis demonstrated that ~43% of the hits are genes predicted to encode components of the DNA damage response pathway (5/28) or for proteins necessary for cell division (7/28) (S1 Table). For example, the top 4 hits of the screen were GRR1, KIP3, MEC3 and RAD18 genes (S1 Table). C. albicans GRR1 and KIP3 are required for cell cycle progression [40] and mitotic spindle organisation, respectively [41]. Although C. albicans MEC3 and RAD18 are uncharacterised, they encode for proteins conserved in other organisms that are universally involved in sensing DNA damage (Mec3) [42] and in DNA post-replication repair (Rad18) [43]. Of the remaining hits, 3/28 genes encode proteins with no apparent ortholog in the two well-studied yeast model systems (Saccharomyces cerevisiae and Schizosaccharomyces pombe). The last 13 genes encodes for proteins with diverse functions, including stress response (HOG1) [44], transcriptional and chromatin regulation (SPT8, SIN3) [45–47], transport and trafficking (DUR35, NPR2, FCY2, PEP7, VAC14) [48–52], protein folding (CNE1) [53], MAP kinase pathway (STT4) [54], phosphatase (PTC2) [47], immune evasion (GPD2) [55] and cell wall biosynthesis (KRE5) [56]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. ULP2 is a regulator of the C. albicans genotoxic stress response. (A) Schematic representation of the screening strategy. 674 C. albicans deletion strains were screened for hypersensitivity to UV and MMS. Hypersensitivity was scored by comparing the growth of treated vs untreated on a scale of 0 (white) to 4 (magenta). Black *: genes encoding for DNA damage and sensing repair pathway components, Blue *: genes encoding for cell division and chromosome segregation machinery, Green arrow: ulp2Δ/Δ (B) Data for a plate containing ulp2Δ/Δ strain (magenta circle). Growth on Non-selective (N/S) media or following UV and MMS treatment is shown (C) Colony-forming unit assay (% survival) of UV treated WT and ulp2Δ/Δ strain. Statistical analysis was performed using the Kruskal-Wallis test with Mann-Whitney U test for post hoc analysis (D) Growth curve of WT and ulp2Δ/Δ strains grown in non-selective (N/S) and MMS-containing liquid media. Error bars: standard deviation (SD) of three biological replicates (E) Growth curve of WT and ulp2Δ/Δ strains grown in non-selective (N/S) and HU-containing liquid media. Error bars: standard deviation (SD) of three biological replicates. https://doi.org/10.1371/journal.pgen.1010576.g001 One of the highest-ranked genes on our screen is ULP2 (CR_03820C/ orf19.4353: MMS score:3, UV score:3) encoding for a SUMO protease (Fig 1A and 1B and S1 Table). Colony-forming unit (CFU) assays of UV-treated cells confirmed the importance of C. albicans ULP2 in DNA damage resistance as UV treatment reduced the number of CFU in a ulp2Δ/Δ strain (~14.5% survival) compared to a wild-type (WT) strain (~33.7% survival) (Fig 1C). Furthermore, the ulp2Δ/Δ strain also displayed a reduced growth rate in liquid media containing MMS or Hydroxyurea (HU), a chemotherapeutic agent that challenges genome integrity by stalling replication forks [57] (Fig 1D and 1E). Thus, ULP2 has a role in responding to a wide range of genotoxic agents. ULP2 but not ULP1 is required for survival under stress C. albicans contains three putative SUMO-deconjugating enzymes: Ulp1, Ulp2 and Ulp3 (Fig 2A). Sequence comparison between the three C. albicans Ulp proteins and the two well-characterised S. cerevisiae Ulps (Ulp1 and Ulp2) reveals that although the C. albicans proteins are poorly conserved, the amino acid residues essential for catalytic activity are conserved (Fig 2A and 2B). Accordingly, recombinantly expressed C. albicans Ulp1, Ulp2 and Ulp3 have SUMO-processing activity in vitro [58]. Similarly to S. cerevisiae ULP1, C. albicans ULP3 is an essential gene and was not investigated further in this study [59,60]. Previous studies failed to detect a poly-histidine tagged Ulp2 protein by Western blot analyses of C. albicans protein extracts [58]. These results suggested that Ulp2 is unstable or expressed at undetectable low levels. We reassessed Ulp2 protein levels by generating strains expressing, at the endogenous locus, an epitope-tagged Ulp2 protein (Ulp2-HA). Western blot analyses show that Ulp2-HA expression is readily detected in extracts from four independent integrant strains (Fig 2C). Thus, a stable Ulp2 protein is expressed in cells grown under standard laboratory growth conditions (YPD, 30°C). To assess whether ULP1, similarly to ULP2, is involved in genotoxic stress response, we engineered homozygous deletion strains for ULP1 (ulp1Δ/Δ) and ULP2 (ulp2Δ/Δ). Growth analysis demonstrated that deletion of ULP2 reduces fitness as the newly generated ulp2Δ/Δ strain is viable, but cells are slow-growing (Fig 2D and 2E). In contrast, the ulp1Δ/Δ strain grows similarly to the WT control in solid and liquid media (Fig 2D and 2E). Spot dilution assay confirmed that ULP2 is an important regulator of C. albicans stress response as, similarly to the deletion library mutant, the newly generated ulp2Δ/Δ strain was sensitive to different stress conditions including treatment with DNA damaging agents (UV and MMS), DNA replication inhibitor (HU), oxidative stress (H2O2) and high temperature (39°C) (Fig 2E). In contrast, deleting ULP1 did not cause any sensitivity to the tested stress conditions (Fig 2E). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. ULP2 is necessary for survival under stress. (A) Schematic representations of C. albicans Ulp1, Ulp2 and Ulp3 proteins. The systematic name and the amino acid (aa) number is indicated for each protein. Blue box: putative catalytic UD SUMO protease domain (B) Protein alignment of C. albicans Ulp proteins (Ulp1, Ulp2 and Ulp3) and S. cerevisiae Ulp2 proteins (Ulp1 and Ulp2). Magenta arrows: amino acids essential for SUMO protease activity (C) Western blot analysis of 4 ULP2-HA integrants and the progenitor untagged control (No Tag). Top: anti-HA Western blot, Bottom: anti-Tubulin Western blot serving as a loading control (D) Growth curves of WT, ulp1Δ/Δ and ulp2Δ/Δ strains grown in non-selective (N/S) liquid media. Error bars: standard deviation (SD) of three biological replicates (E) Serial dilution assay of WT, ulp1Δ/Δ and ulp2Δ/Δ strains grown in unstressed (N/S) or stress (UV, MMS, HU, H2O2 and 39°C) growth conditions. https://doi.org/10.1371/journal.pgen.1010576.g002 Although ULP-1, ULP-2 and ULP-3 may have some partially redundant functions, our results suggest that ULP-1 does not play a major role in genotoxic stress response. In summary, loss of ULP2 leads to poor growth in standard laboratory growth conditions and hypersensitivity to multiple stresses. Genome instability is exacerbated in the absence of ULP2 To assess whether the hypersensitivity to DNA damage agents observed in the ulp2Δ/Δ strain was indeed due to enhanced genome instability, we deleted ULP2 from a set of strains containing a heterozygous URA3+ marker gene inserted in three different chromosomes (Chr 1, 3 and 7) [61]. We quantified the frequency of URA3+ marker loss by plating on media containing the counter-selective drug 5-Fluoroorotic Acid (FOA) and scoring the number of colonies able to grow on FOA-containing media compared to non-selective (N/S) media. Deletion of ULP2 caused a dramatic increase in LOH rate at all three chromosomes (Chr1: 378X, Chr3: 18X, Chr7: 96X), indicating that ULP2 is required for maintaining genome stability across the C. albicans genome (Fig 3A). In C. albicans, hypersensitivity to genotoxic stress leads to filamentous growth [37,62–65]. Accordingly, the ulp2Δ/Δ strain formed wrinkled colonies on solid medium and displayed a higher frequency of abnormal morphologies than the WT strain, including filamentous cells (Fig 3B and 3C). To assess whether the exacerbated ulp2Δ/Δ genome instability is linked to defective chromosome segregation, we deleted the ULP2 gene in a reporter strain in which TetO sequences are integrated adjacent to the centromere on Chr7 (CEN7) and a TetR-GFP fusion protein is expressed from the gene-free NEUT5L locus [66,67]. The binding of TetR-GFP to TetO sequences allowed the visualisation of Chr7 duplication and segregation during the cell cycle. We found that deletion of ULP2 leads to abnormal Chr7 segregation. This included cells with no TetR-GFP signals or multiple TetR-GFP-foci, that were ~5 fold higher in the ulp2Δ/Δ strain compared to the WT control strain (Fig 3D). Thus, deletion of C. albicans ULP2 leads to increased genome instability. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Loss of ULP2 leads to increased genome instability. (A) Quantification (%) of loss of a heterozygous URA3+ marker gene inserted in Chr1, Chr3 and Chr7 in WT and ulp2Δ/Δ strain. The fold difference of URA3+ marker loss between ulp2Δ/Δ and WT strains is indicated. Statistical differences were calculated using the Kruskal-Wallis test and the Mann-Whitney U test for post hoc analysis **: Chr1 (4.11 E-07) and Chr7 (6.74 E-05) p-value, *: Chr3 (2.87 E-02) p-value (B) Left: Representative images displaying colony morphologies of WT and ulp2Δ/Δ strains. Right: Quantification (%) of smooth and wrinkled colonies in WT and ulp2Δ/Δ strains (C) Left: Representative images displaying the morphologies of WT and ulp2Δ/Δ strains. Right: Quantification (%) of yeast and filamentous morphologies in WT and ulp2Δ/Δ strains. Error bar: Standard deviation of 3 biological replicates (D) Top: schematics of the CEN7 TetO and TetR-GFP system. Bottom: nuclear morphology and segregation pattern of centromere 7 (CEN7) in WT and ulp2Δ/Δ strain. Quantification (%) of abnormal GFP-CEN7 patterns is indicated. Error bar: Standard deviation of 3 biological replicates. https://doi.org/10.1371/journal.pgen.1010576.g003 ULP2 loss coupled with stress triggers selection of segmental aneuploidies Previous studies performed in the model system S. cerevisiae demonstrated that loss of ULP2 leads to the accumulation of a specific multichromosome aneuploidy (amplification of both ChrI and ChrXII). This aneuploidy rescues the lethal defects of ulp2 deletion by amplification of specific genes on both chromosomes [68,69]. To assess whether loss of C. albicans ULP2 triggers the selection of gross karyotypic abnormalities, we analysed the genome of WT and ulp2Δ/Δ strains at the beginning (Day 0) and the end (Day 30) of an in vitro evolution experiment where strains were passaged daily for 30 days in rich media (YPD 30°C) (Fig 4A). Clamped homogeneous electrical field (CHEF) electrophoresis analysis did not detect any major chromosome rearrangements in both sets of evolved strains (Fig 4B). To further investigate the impact of ULP2 loss on genome organisation, we sequenced the genome of 3 randomly selected ulp2Δ/Δ colonies by whole genome Illumina sequencing (WGS) and compared their genome to the C. albicans reference genome. This analysis revealed that loss of ULP2 leads to very few (<10 across the 3 isolates) de novo mutations (S2 Table). Although we did not detect CNVs, we identified novel LOH tracts on different chromosomes in two of the three sequenced colonies (Fig 4C). For example, chromosome mis-segregation followed by reduplication of the remaining homologue is detected in isolate U1 (U1: ChrR) and the genome of U2 contains a long-track LOH (U2:Chr3L) that occurred within 4.6 kb of a repeat locus on Chr3L (PGA18, [19]) (Fig 4C). Therefore, loss of ULP2 can trigger selection of large chromosomal variations. We hypothesised that exposure of ulp2Δ/Δ cells to stress could facilitate the selection of novel adaptive karyotypes. To test this hypothesis, we challenged the ulp2Δ/Δ strain with high concentrations of FLC (128 μg/ml; ~1000 fold above susceptibility breakpoint [70]) and isolated a FLC-adapted isolate (FLC-1) that was still able to grow at high drug concentration following two passages (T1 and T2) in non-selective (N/S) media (Fig 5A and 5B). The phenotypes associated with the loss of ULP2 were partially rescued in FLC-1. as this isolate was less sensitive than the ulp2Δ/Δ progenitor to UV treatment and high temperature (39°C) (Fig 5B). Furthermore, fewer wrinkled colonies are present in FLC-1 than ulp2Δ/Δ and the number of elongated cells was reduced in FLC-1 compared to ulp2Δ/Δ (Fig 5C and 5D). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Karyotypic changes are detected in the absence of ULP2. (A) Schematics of laboratory evolution strategy (B) Karyotype organisation of C. albicans WT and ulp2Δ/Δ strains at the start (Day 0) and the end (Day 30) of the evolution experiment (C) Whole genome sequencing analysis of the progenitor (WT:SN152) and three single ulp2Δ/Δ colonies (U1, U2, and U3). Data were plotted as the log2 ratio and converted to chromosome copy number (y-axis, 1–4 copies) as a function of chromosome position (x-axis, Chr1-ChrR) using the Yeast Mapping Analysis Pipeline (YMAP) [109]. Heterozygous (AB) regions are indicated with grey shading, and homozygous regions (loss of heterozygosity) are indicated by shading of the remaining haplotype, either AA (cyan) or BB (magenta). https://doi.org/10.1371/journal.pgen.1010576.g004 Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Genomic variants are selected in ulp2Δ/Δ cells challenged with additional stress. (A) Schematic of experimental design. The FLC-1 ulp2Δ/Δ isolate was selected from a casitone agar plate containing 128 μg/ml fluconazole (FLC) passaged two times (2X) non-selective (N/S) agar plates and its genome sequenced by Illumina technology. FLC-2, FLC-3 and FLC-4 ulp2Δ/Δ isolates were selected from a casitone agar plate containing 128 μg/ml fluconazole (FLC) and the genome was sequenced by Illumina technology. (B) Serial dilution assay of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1) in unstressed (N/S) or stress (UV, MMS, HU, H2O2 and 39°C) growth conditions. (C) Left: Representative images displaying colony morphologies of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1). Right: Quantification (%) of smooth and wrinkled colonies in WT and ulp2Δ/Δ strains (D) Left: Representative images displaying the morphologies of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1). Right: Quantification (%) of yeast and filamentous morphologies in WT and ulp2Δ/Δ strains. Error bar: Standard deviation of 3 biological replicates (E) Whole genome sequencing data for four single colonies isolated from 128 μg/ml fluconazole plates (FLC1-FLC4). The chromosome copy number is plotted along the y-axis (1–4 copies). https://doi.org/10.1371/journal.pgen.1010576.g005 To identify mutations that underlie the above phenotypes, we sequenced the genome of three FLC-1 single colonies (FLC-1a, b and c) by Illumina technology (Figs 5A and S1 and S3–S5 Tables). We also sequenced the genome of 3 additional ulp2Δ/Δ FLC-adapted isolates (FLC-2, FLC-3 and FLC-4) randomly selected from FLC plates and unable to grow at high FLC doses following passaging in N/S medium. The whole-genome sequencing revealed that the FLC-adapted colonies have a genotype distinct from the ulp2Δ/Δ progenitor (Fig 5E and S3 Table). FLC-1, but not FLC-2, FLC-3, or FLC-4 isolates, is marked by a segmental aneuploidy: a partial Chr1 amplification (~1.3 Mbp) containing 535 protein-coding genes (Fig 5E and S4 Table). Furthermore, all sequenced isolates have a partial deletion (~ 388 Kb) of the right arm of ChrR (ChrRR-Deletion). ChrRR-deletion occurs at the ribosomal DNA and extends to the right telomere of ChrR (ChrR:1,897,750 bp—2,286,380 bp), reducing the copy number of 204 genes from two to one (Fig 5E and S5 Table). In contrast, we detected very few (<10) de novo point mutations, and none of these are common among all the sequenced FLC isolates (S3 Table). Thus, exposure to an antifungal drug triggers the selection of adaptive chromosomal variations in the absence of ULP2. Segmental aneuploidy is selected via a two-step process Illumina sequencing is an inadequate technology for resolving complex chromosomal abnormalities because of the generated short reads. For example, we could not establish whether the increased Chr1 CNV was due to the formation of an extrachromosomal fragment or to a chromosomal fusion. Therefore, to understand further the genomic structure of FLC-1 we sequenced the genome of this isolate using long-read Oxford Nanopore Technologies (ONT) sequencing. To establish the temporal trajectory of FLC-1 aneuploidies, timepoints T1 (1X passage in non-selective media following FLC treatment) and T2 (2X passages in non-selective medium following FLC treatment) were sequenced (Fig 5A). Using this method, we could resolve the structure of FLC-1 aneuploidies completely (Fig 6A). We discovered that FLC-1 contains, in addition to the two endogenous Chr1 homologous chromosomes, an extra linear Chr1 (linChr1) copy that is selected in a two-step process. At time T1, we detected a linChr1 (~ 1.9 Mbp) containing an intact right arm and a truncated left arm (Fig 6B). At the 5’ breakpoint, Chr7 subtelomeric and telomeric regions are fused to Chr1. This chromosomal fusion occurs within the internal TLOα34 gene deleting ~1.3 Mbp. Sequence homology between TLOα34 and the subtelomeric TLOγ16 gene is likely to have guided the fusion between Chr1 and Chr7. At T2, linChr1 is further processed at its right arm by deletion of ~ 0.6 Mbp and the addition of telomeric repeats (Fig 6B). The 3’ breakpoint contains a microhomology tract (6 bp 5’-TTCTTG-3’) between internal sequences of Chr1 and telomeric repeats. The resulting linChr1 spans the centromere and is flanked by terminal telomeric repeats. To assess whether linChr1 was necessary for the ulp2Δ/Δ phenotypic rescue, we passaged FLC-1 in N/S media and selected three independent phenotypic revertants (R-1, R-2 and R-3) that, similarly to the ulp2Δ/Δ strain, are less able to withstand FLC and form wrinkled colonies on solid media (Fig 6C and 6D). Diagnostic PCR analysis with primers specific for linChr1 indicates that linChr1 was lost in the R-1, R-2 and R-3 revertants, while an amplification product was detected in DNA isolated from FLC-1 (Fig 6E). This result suggests that linChr1 has an adaptive value. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Microhomology tracts and DNA repeats drive the formation of segmental aneuploidies. (A) Circular plots of the long-read coverage across the FLC-1 genome at timepoint T1 (left) and T2 (right). Magenta and Blue arrows: CNV breakpoints on ChrR and Chr1 respectively (B) Schematics of the events leading to linChr1 formation. Top: Schematics of the endogenous Chr1 and Chr7. Middle: Schematics of linChr1 structure at timepoint T1. Bottom: Schematics of linChr1 structure at timepoint T2. Tel: telomeric repeats. Microhom: microhomology tract found on Chr1 and telomeric repeats. (C) Serial dilution assay of ulp2Δ/Δ parental (P) fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates in non-selective (N/S) agar plate or plates containing 128 μg/ml fluconazole (FLC) (D) Quantification (%) of smooth and wrinkled colonies in ulp2Δ/Δ parental (P) fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates (E) Top: Schematics of linChr1 highlighting the position (magenta line) of linChr1 specific primer. Bottom: LinChr1 diagnostic PCR in the ulp2Δ/Δ parental (P), fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates. Loading Control: Chr1 internal primers (F) Schematics of ChrRR-deletion. Tel: telomeric repeats Microhom: microhomology tract found on ChrR and telomeric repeats. https://doi.org/10.1371/journal.pgen.1010576.g006 Finally, we discovered that the ChrRR-deletion is selected early following FLC treatment as this deletion is present at timepoint T1 and T2 (Fig 6A). A microhomology tract between the RDN18 (encoding for the 18S rRNA) genes and telomeric repeats facilitated the addition of telomeric repeats to one copy of the RDN18 gene stabilising the truncated chromosome (Fig 6E). Thus, rearrangements at microhomology regions and DNA repeats guide the formation of adaptive segmental aneuploidies via two temporally separated events. A systematic genetic screen identifies Ulp2 as a regulator of C. albicans genotoxic stress response To identify factors regulating C. albicans genome integrity, we utilised a deletion library comprising a subset (674/3000) of C. albicans genes that are not conserved in other organisms or have a functional motif potentially related to virulence [36]. As defects in genome integrity lead to hypersensitivity to genotoxic agents [37], the deletion library was screened for hypersensitivity to two DNA damaging agents: Ultraviolet (UV) irradiation which induces formation of pyrimidine dimers [38], and methyl methanesulfonate (MMS), which leads to replication blocks and base mispairing [39]. Genotoxic stress hypersensitivity was semi-quantitatively scored by comparing the growth of treated versus untreated on a scale of 0 to 4, where 0 indicates no sensitivity, and 4 specifies strong hypersensitivity (Fig 1A). The screen identified 28 gene deletions linked to DNA damage hypersensitivity (UV or MMS score ≥2). Of those deletion mutants, 9/28 hits show sensitivity to both UV and MMS, 6/28 hits are sensitive only to UV, and 13/28 hits are sensitive only to MMS (S1 Table). Functional prediction analysis demonstrated that ~43% of the hits are genes predicted to encode components of the DNA damage response pathway (5/28) or for proteins necessary for cell division (7/28) (S1 Table). For example, the top 4 hits of the screen were GRR1, KIP3, MEC3 and RAD18 genes (S1 Table). C. albicans GRR1 and KIP3 are required for cell cycle progression [40] and mitotic spindle organisation, respectively [41]. Although C. albicans MEC3 and RAD18 are uncharacterised, they encode for proteins conserved in other organisms that are universally involved in sensing DNA damage (Mec3) [42] and in DNA post-replication repair (Rad18) [43]. Of the remaining hits, 3/28 genes encode proteins with no apparent ortholog in the two well-studied yeast model systems (Saccharomyces cerevisiae and Schizosaccharomyces pombe). The last 13 genes encodes for proteins with diverse functions, including stress response (HOG1) [44], transcriptional and chromatin regulation (SPT8, SIN3) [45–47], transport and trafficking (DUR35, NPR2, FCY2, PEP7, VAC14) [48–52], protein folding (CNE1) [53], MAP kinase pathway (STT4) [54], phosphatase (PTC2) [47], immune evasion (GPD2) [55] and cell wall biosynthesis (KRE5) [56]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. ULP2 is a regulator of the C. albicans genotoxic stress response. (A) Schematic representation of the screening strategy. 674 C. albicans deletion strains were screened for hypersensitivity to UV and MMS. Hypersensitivity was scored by comparing the growth of treated vs untreated on a scale of 0 (white) to 4 (magenta). Black *: genes encoding for DNA damage and sensing repair pathway components, Blue *: genes encoding for cell division and chromosome segregation machinery, Green arrow: ulp2Δ/Δ (B) Data for a plate containing ulp2Δ/Δ strain (magenta circle). Growth on Non-selective (N/S) media or following UV and MMS treatment is shown (C) Colony-forming unit assay (% survival) of UV treated WT and ulp2Δ/Δ strain. Statistical analysis was performed using the Kruskal-Wallis test with Mann-Whitney U test for post hoc analysis (D) Growth curve of WT and ulp2Δ/Δ strains grown in non-selective (N/S) and MMS-containing liquid media. Error bars: standard deviation (SD) of three biological replicates (E) Growth curve of WT and ulp2Δ/Δ strains grown in non-selective (N/S) and HU-containing liquid media. Error bars: standard deviation (SD) of three biological replicates. https://doi.org/10.1371/journal.pgen.1010576.g001 One of the highest-ranked genes on our screen is ULP2 (CR_03820C/ orf19.4353: MMS score:3, UV score:3) encoding for a SUMO protease (Fig 1A and 1B and S1 Table). Colony-forming unit (CFU) assays of UV-treated cells confirmed the importance of C. albicans ULP2 in DNA damage resistance as UV treatment reduced the number of CFU in a ulp2Δ/Δ strain (~14.5% survival) compared to a wild-type (WT) strain (~33.7% survival) (Fig 1C). Furthermore, the ulp2Δ/Δ strain also displayed a reduced growth rate in liquid media containing MMS or Hydroxyurea (HU), a chemotherapeutic agent that challenges genome integrity by stalling replication forks [57] (Fig 1D and 1E). Thus, ULP2 has a role in responding to a wide range of genotoxic agents. ULP2 but not ULP1 is required for survival under stress C. albicans contains three putative SUMO-deconjugating enzymes: Ulp1, Ulp2 and Ulp3 (Fig 2A). Sequence comparison between the three C. albicans Ulp proteins and the two well-characterised S. cerevisiae Ulps (Ulp1 and Ulp2) reveals that although the C. albicans proteins are poorly conserved, the amino acid residues essential for catalytic activity are conserved (Fig 2A and 2B). Accordingly, recombinantly expressed C. albicans Ulp1, Ulp2 and Ulp3 have SUMO-processing activity in vitro [58]. Similarly to S. cerevisiae ULP1, C. albicans ULP3 is an essential gene and was not investigated further in this study [59,60]. Previous studies failed to detect a poly-histidine tagged Ulp2 protein by Western blot analyses of C. albicans protein extracts [58]. These results suggested that Ulp2 is unstable or expressed at undetectable low levels. We reassessed Ulp2 protein levels by generating strains expressing, at the endogenous locus, an epitope-tagged Ulp2 protein (Ulp2-HA). Western blot analyses show that Ulp2-HA expression is readily detected in extracts from four independent integrant strains (Fig 2C). Thus, a stable Ulp2 protein is expressed in cells grown under standard laboratory growth conditions (YPD, 30°C). To assess whether ULP1, similarly to ULP2, is involved in genotoxic stress response, we engineered homozygous deletion strains for ULP1 (ulp1Δ/Δ) and ULP2 (ulp2Δ/Δ). Growth analysis demonstrated that deletion of ULP2 reduces fitness as the newly generated ulp2Δ/Δ strain is viable, but cells are slow-growing (Fig 2D and 2E). In contrast, the ulp1Δ/Δ strain grows similarly to the WT control in solid and liquid media (Fig 2D and 2E). Spot dilution assay confirmed that ULP2 is an important regulator of C. albicans stress response as, similarly to the deletion library mutant, the newly generated ulp2Δ/Δ strain was sensitive to different stress conditions including treatment with DNA damaging agents (UV and MMS), DNA replication inhibitor (HU), oxidative stress (H2O2) and high temperature (39°C) (Fig 2E). In contrast, deleting ULP1 did not cause any sensitivity to the tested stress conditions (Fig 2E). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. ULP2 is necessary for survival under stress. (A) Schematic representations of C. albicans Ulp1, Ulp2 and Ulp3 proteins. The systematic name and the amino acid (aa) number is indicated for each protein. Blue box: putative catalytic UD SUMO protease domain (B) Protein alignment of C. albicans Ulp proteins (Ulp1, Ulp2 and Ulp3) and S. cerevisiae Ulp2 proteins (Ulp1 and Ulp2). Magenta arrows: amino acids essential for SUMO protease activity (C) Western blot analysis of 4 ULP2-HA integrants and the progenitor untagged control (No Tag). Top: anti-HA Western blot, Bottom: anti-Tubulin Western blot serving as a loading control (D) Growth curves of WT, ulp1Δ/Δ and ulp2Δ/Δ strains grown in non-selective (N/S) liquid media. Error bars: standard deviation (SD) of three biological replicates (E) Serial dilution assay of WT, ulp1Δ/Δ and ulp2Δ/Δ strains grown in unstressed (N/S) or stress (UV, MMS, HU, H2O2 and 39°C) growth conditions. https://doi.org/10.1371/journal.pgen.1010576.g002 Although ULP-1, ULP-2 and ULP-3 may have some partially redundant functions, our results suggest that ULP-1 does not play a major role in genotoxic stress response. In summary, loss of ULP2 leads to poor growth in standard laboratory growth conditions and hypersensitivity to multiple stresses. Genome instability is exacerbated in the absence of ULP2 To assess whether the hypersensitivity to DNA damage agents observed in the ulp2Δ/Δ strain was indeed due to enhanced genome instability, we deleted ULP2 from a set of strains containing a heterozygous URA3+ marker gene inserted in three different chromosomes (Chr 1, 3 and 7) [61]. We quantified the frequency of URA3+ marker loss by plating on media containing the counter-selective drug 5-Fluoroorotic Acid (FOA) and scoring the number of colonies able to grow on FOA-containing media compared to non-selective (N/S) media. Deletion of ULP2 caused a dramatic increase in LOH rate at all three chromosomes (Chr1: 378X, Chr3: 18X, Chr7: 96X), indicating that ULP2 is required for maintaining genome stability across the C. albicans genome (Fig 3A). In C. albicans, hypersensitivity to genotoxic stress leads to filamentous growth [37,62–65]. Accordingly, the ulp2Δ/Δ strain formed wrinkled colonies on solid medium and displayed a higher frequency of abnormal morphologies than the WT strain, including filamentous cells (Fig 3B and 3C). To assess whether the exacerbated ulp2Δ/Δ genome instability is linked to defective chromosome segregation, we deleted the ULP2 gene in a reporter strain in which TetO sequences are integrated adjacent to the centromere on Chr7 (CEN7) and a TetR-GFP fusion protein is expressed from the gene-free NEUT5L locus [66,67]. The binding of TetR-GFP to TetO sequences allowed the visualisation of Chr7 duplication and segregation during the cell cycle. We found that deletion of ULP2 leads to abnormal Chr7 segregation. This included cells with no TetR-GFP signals or multiple TetR-GFP-foci, that were ~5 fold higher in the ulp2Δ/Δ strain compared to the WT control strain (Fig 3D). Thus, deletion of C. albicans ULP2 leads to increased genome instability. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Loss of ULP2 leads to increased genome instability. (A) Quantification (%) of loss of a heterozygous URA3+ marker gene inserted in Chr1, Chr3 and Chr7 in WT and ulp2Δ/Δ strain. The fold difference of URA3+ marker loss between ulp2Δ/Δ and WT strains is indicated. Statistical differences were calculated using the Kruskal-Wallis test and the Mann-Whitney U test for post hoc analysis **: Chr1 (4.11 E-07) and Chr7 (6.74 E-05) p-value, *: Chr3 (2.87 E-02) p-value (B) Left: Representative images displaying colony morphologies of WT and ulp2Δ/Δ strains. Right: Quantification (%) of smooth and wrinkled colonies in WT and ulp2Δ/Δ strains (C) Left: Representative images displaying the morphologies of WT and ulp2Δ/Δ strains. Right: Quantification (%) of yeast and filamentous morphologies in WT and ulp2Δ/Δ strains. Error bar: Standard deviation of 3 biological replicates (D) Top: schematics of the CEN7 TetO and TetR-GFP system. Bottom: nuclear morphology and segregation pattern of centromere 7 (CEN7) in WT and ulp2Δ/Δ strain. Quantification (%) of abnormal GFP-CEN7 patterns is indicated. Error bar: Standard deviation of 3 biological replicates. https://doi.org/10.1371/journal.pgen.1010576.g003 ULP2 loss coupled with stress triggers selection of segmental aneuploidies Previous studies performed in the model system S. cerevisiae demonstrated that loss of ULP2 leads to the accumulation of a specific multichromosome aneuploidy (amplification of both ChrI and ChrXII). This aneuploidy rescues the lethal defects of ulp2 deletion by amplification of specific genes on both chromosomes [68,69]. To assess whether loss of C. albicans ULP2 triggers the selection of gross karyotypic abnormalities, we analysed the genome of WT and ulp2Δ/Δ strains at the beginning (Day 0) and the end (Day 30) of an in vitro evolution experiment where strains were passaged daily for 30 days in rich media (YPD 30°C) (Fig 4A). Clamped homogeneous electrical field (CHEF) electrophoresis analysis did not detect any major chromosome rearrangements in both sets of evolved strains (Fig 4B). To further investigate the impact of ULP2 loss on genome organisation, we sequenced the genome of 3 randomly selected ulp2Δ/Δ colonies by whole genome Illumina sequencing (WGS) and compared their genome to the C. albicans reference genome. This analysis revealed that loss of ULP2 leads to very few (<10 across the 3 isolates) de novo mutations (S2 Table). Although we did not detect CNVs, we identified novel LOH tracts on different chromosomes in two of the three sequenced colonies (Fig 4C). For example, chromosome mis-segregation followed by reduplication of the remaining homologue is detected in isolate U1 (U1: ChrR) and the genome of U2 contains a long-track LOH (U2:Chr3L) that occurred within 4.6 kb of a repeat locus on Chr3L (PGA18, [19]) (Fig 4C). Therefore, loss of ULP2 can trigger selection of large chromosomal variations. We hypothesised that exposure of ulp2Δ/Δ cells to stress could facilitate the selection of novel adaptive karyotypes. To test this hypothesis, we challenged the ulp2Δ/Δ strain with high concentrations of FLC (128 μg/ml; ~1000 fold above susceptibility breakpoint [70]) and isolated a FLC-adapted isolate (FLC-1) that was still able to grow at high drug concentration following two passages (T1 and T2) in non-selective (N/S) media (Fig 5A and 5B). The phenotypes associated with the loss of ULP2 were partially rescued in FLC-1. as this isolate was less sensitive than the ulp2Δ/Δ progenitor to UV treatment and high temperature (39°C) (Fig 5B). Furthermore, fewer wrinkled colonies are present in FLC-1 than ulp2Δ/Δ and the number of elongated cells was reduced in FLC-1 compared to ulp2Δ/Δ (Fig 5C and 5D). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Karyotypic changes are detected in the absence of ULP2. (A) Schematics of laboratory evolution strategy (B) Karyotype organisation of C. albicans WT and ulp2Δ/Δ strains at the start (Day 0) and the end (Day 30) of the evolution experiment (C) Whole genome sequencing analysis of the progenitor (WT:SN152) and three single ulp2Δ/Δ colonies (U1, U2, and U3). Data were plotted as the log2 ratio and converted to chromosome copy number (y-axis, 1–4 copies) as a function of chromosome position (x-axis, Chr1-ChrR) using the Yeast Mapping Analysis Pipeline (YMAP) [109]. Heterozygous (AB) regions are indicated with grey shading, and homozygous regions (loss of heterozygosity) are indicated by shading of the remaining haplotype, either AA (cyan) or BB (magenta). https://doi.org/10.1371/journal.pgen.1010576.g004 Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Genomic variants are selected in ulp2Δ/Δ cells challenged with additional stress. (A) Schematic of experimental design. The FLC-1 ulp2Δ/Δ isolate was selected from a casitone agar plate containing 128 μg/ml fluconazole (FLC) passaged two times (2X) non-selective (N/S) agar plates and its genome sequenced by Illumina technology. FLC-2, FLC-3 and FLC-4 ulp2Δ/Δ isolates were selected from a casitone agar plate containing 128 μg/ml fluconazole (FLC) and the genome was sequenced by Illumina technology. (B) Serial dilution assay of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1) in unstressed (N/S) or stress (UV, MMS, HU, H2O2 and 39°C) growth conditions. (C) Left: Representative images displaying colony morphologies of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1). Right: Quantification (%) of smooth and wrinkled colonies in WT and ulp2Δ/Δ strains (D) Left: Representative images displaying the morphologies of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-1). Right: Quantification (%) of yeast and filamentous morphologies in WT and ulp2Δ/Δ strains. Error bar: Standard deviation of 3 biological replicates (E) Whole genome sequencing data for four single colonies isolated from 128 μg/ml fluconazole plates (FLC1-FLC4). The chromosome copy number is plotted along the y-axis (1–4 copies). https://doi.org/10.1371/journal.pgen.1010576.g005 To identify mutations that underlie the above phenotypes, we sequenced the genome of three FLC-1 single colonies (FLC-1a, b and c) by Illumina technology (Figs 5A and S1 and S3–S5 Tables). We also sequenced the genome of 3 additional ulp2Δ/Δ FLC-adapted isolates (FLC-2, FLC-3 and FLC-4) randomly selected from FLC plates and unable to grow at high FLC doses following passaging in N/S medium. The whole-genome sequencing revealed that the FLC-adapted colonies have a genotype distinct from the ulp2Δ/Δ progenitor (Fig 5E and S3 Table). FLC-1, but not FLC-2, FLC-3, or FLC-4 isolates, is marked by a segmental aneuploidy: a partial Chr1 amplification (~1.3 Mbp) containing 535 protein-coding genes (Fig 5E and S4 Table). Furthermore, all sequenced isolates have a partial deletion (~ 388 Kb) of the right arm of ChrR (ChrRR-Deletion). ChrRR-deletion occurs at the ribosomal DNA and extends to the right telomere of ChrR (ChrR:1,897,750 bp—2,286,380 bp), reducing the copy number of 204 genes from two to one (Fig 5E and S5 Table). In contrast, we detected very few (<10) de novo point mutations, and none of these are common among all the sequenced FLC isolates (S3 Table). Thus, exposure to an antifungal drug triggers the selection of adaptive chromosomal variations in the absence of ULP2. Segmental aneuploidy is selected via a two-step process Illumina sequencing is an inadequate technology for resolving complex chromosomal abnormalities because of the generated short reads. For example, we could not establish whether the increased Chr1 CNV was due to the formation of an extrachromosomal fragment or to a chromosomal fusion. Therefore, to understand further the genomic structure of FLC-1 we sequenced the genome of this isolate using long-read Oxford Nanopore Technologies (ONT) sequencing. To establish the temporal trajectory of FLC-1 aneuploidies, timepoints T1 (1X passage in non-selective media following FLC treatment) and T2 (2X passages in non-selective medium following FLC treatment) were sequenced (Fig 5A). Using this method, we could resolve the structure of FLC-1 aneuploidies completely (Fig 6A). We discovered that FLC-1 contains, in addition to the two endogenous Chr1 homologous chromosomes, an extra linear Chr1 (linChr1) copy that is selected in a two-step process. At time T1, we detected a linChr1 (~ 1.9 Mbp) containing an intact right arm and a truncated left arm (Fig 6B). At the 5’ breakpoint, Chr7 subtelomeric and telomeric regions are fused to Chr1. This chromosomal fusion occurs within the internal TLOα34 gene deleting ~1.3 Mbp. Sequence homology between TLOα34 and the subtelomeric TLOγ16 gene is likely to have guided the fusion between Chr1 and Chr7. At T2, linChr1 is further processed at its right arm by deletion of ~ 0.6 Mbp and the addition of telomeric repeats (Fig 6B). The 3’ breakpoint contains a microhomology tract (6 bp 5’-TTCTTG-3’) between internal sequences of Chr1 and telomeric repeats. The resulting linChr1 spans the centromere and is flanked by terminal telomeric repeats. To assess whether linChr1 was necessary for the ulp2Δ/Δ phenotypic rescue, we passaged FLC-1 in N/S media and selected three independent phenotypic revertants (R-1, R-2 and R-3) that, similarly to the ulp2Δ/Δ strain, are less able to withstand FLC and form wrinkled colonies on solid media (Fig 6C and 6D). Diagnostic PCR analysis with primers specific for linChr1 indicates that linChr1 was lost in the R-1, R-2 and R-3 revertants, while an amplification product was detected in DNA isolated from FLC-1 (Fig 6E). This result suggests that linChr1 has an adaptive value. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Microhomology tracts and DNA repeats drive the formation of segmental aneuploidies. (A) Circular plots of the long-read coverage across the FLC-1 genome at timepoint T1 (left) and T2 (right). Magenta and Blue arrows: CNV breakpoints on ChrR and Chr1 respectively (B) Schematics of the events leading to linChr1 formation. Top: Schematics of the endogenous Chr1 and Chr7. Middle: Schematics of linChr1 structure at timepoint T1. Bottom: Schematics of linChr1 structure at timepoint T2. Tel: telomeric repeats. Microhom: microhomology tract found on Chr1 and telomeric repeats. (C) Serial dilution assay of ulp2Δ/Δ parental (P) fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates in non-selective (N/S) agar plate or plates containing 128 μg/ml fluconazole (FLC) (D) Quantification (%) of smooth and wrinkled colonies in ulp2Δ/Δ parental (P) fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates (E) Top: Schematics of linChr1 highlighting the position (magenta line) of linChr1 specific primer. Bottom: LinChr1 diagnostic PCR in the ulp2Δ/Δ parental (P), fluconazole-recovered (FLC-1) and FLC-1 revertants (R-1, R-2 and R-3) isolates. Loading Control: Chr1 internal primers (F) Schematics of ChrRR-deletion. Tel: telomeric repeats Microhom: microhomology tract found on ChrR and telomeric repeats. https://doi.org/10.1371/journal.pgen.1010576.g006 Finally, we discovered that the ChrRR-deletion is selected early following FLC treatment as this deletion is present at timepoint T1 and T2 (Fig 6A). A microhomology tract between the RDN18 (encoding for the 18S rRNA) genes and telomeric repeats facilitated the addition of telomeric repeats to one copy of the RDN18 gene stabilising the truncated chromosome (Fig 6E). Thus, rearrangements at microhomology regions and DNA repeats guide the formation of adaptive segmental aneuploidies via two temporally separated events. Discussion There is a significant gap in our understanding of how, in microbial organisms, genome instability leads to increased fitness in stress and non-stress environments. Here, we identified the SUMO protease Ulp2 as a key protein ensuring genome stability in C. albicans. We demonstrated that ULP2 loss leads to enhanced genome variation and that dysregulation of the SUMO system combined with drug treatment facilitates the selection of adaptive segmental aneuploidies via a two-step process. Ulp2 is a critical regulator of C. albicans genome stability SUMOylation is a post-translational protein modification signalling in a large number of cellular processes by targeting nuclear eukaryotic proteins [71–74]. SUMO peptides are covalently attached to target proteins by the concerted action of E1, E2 and E3 enzymes while SUMOylation is reversed by SUMO-specific proteases [75–79]. We discovered that the SUMO protease Ulp2 promotes genome stability in C. albicans. These findings are consistent with the emerging role of SUMO proteases as a guardian of genome integrity across eukaryotes. Indeed, SUMO proteases ensure genome stability throughout eukaryotes [68,80]. We hypothesise that C. albicans ULP2 promotes genome stability by modulating SUMO levels of several substrates including: (i) kinetochore and centromere-associated proteins, (ii) the DNA replication machinery and (iii) factors involved in DNA repair. Indeed, it is well established that SUMO homeostasis modulates kinetochore function, DNA replication and DNA repair and that defects in these pathways lead to exacerbated genome instability [77,78,81,82]. For example, several S. cerevisiae kinetochore and centromere-associated proteins, including the centromeric-specific histone H3 variant Cse4CENP-A, are SUMOylated and the knockdown of SENP6, the human ortholog of Ulp2, leads to mis-localisation of the inner kinetochore complex CENP-H/I/K causing chromosome segregation defects [74,83–85]. Furthermore, SUMOylation of replication and repair proteins increases upon DNA damage [74,83,86–89]. We still know very little about SUMOylation effect on C. albicans biology and its adaptation to hostile host environments. However, the observation that C. albicans protein SUMOylation patterns are different in normal and stress growth conditions agrees with our data and suggests that this post-translation modification has a critical role in adaption [90]. The adaptive power of segmental aneuploidy to overcome multiple stresses It is well established that exposure to moderate stress can increase tolerance to unrelated stresses [91,92]. This increased tolerance is usually the result of coordinated gene expression changes known as the core stress response [91]. In contrast, we show that exposure to a stress (antifungal drug) can increase tolerance to unrelated stress (loss of ULP2) by selecting segmental aneuploidies. We demonstrate that two different segmental aneuploidies can co-exist when ulp2Δ/Δ cells are challenged with FLC: an amplification of a Chr1 fragment via the formation of an extra chromosome (linChr1) and a partial deletion of ChrR. We posit that these changes in karyotype provide a synergistic fitness advantage in response to the two independent stressors (the presence of FLC and lack of ULP2) by simultaneously changing the copy number of multiple genes. Indeed, linChr1 amplifies 535 protein-coding genes (S4 Table) and GO analyses demonstrated that 41 of those genes are associated with a “response to drug” phenotype (S6 Table). Among these, amplification of UPC2 encoding for the Upc2 transcription factor is likely to be critical. Indeed, it is well established that UPC2 overexpression leads to FLC resistance by ERG11 transcriptional upregulation [93,94]. Similarly, amplification of CCR4 and NOT5 (part of linChr1) might play a key role in rescuing the fitness defects of the ulp2Δ/Δ strain. Ccr4 and Not5 are subunits of the evolutionarily conserved Ccr4-Not complex that modulate gene expression at multiple levels, including transcription initiation, elongation, de-adenylation and mRNA degradation [95]. It has been shown that S. cerevisiae CCR4 and NOT5 overexpression rescues the lethal defects associated with a ulp2 deletion strain [68]. Similarly, GO analysis revealed that ChrRR-Deletion leads to a reduced copy number of 34/204 genes associated with "response to stress" and 18/204 genes are linked to "response to drug" (S7 Table). Selection of segmental aneuploidies via a two-step process involving microhomology tracts One key question is to understand how complex novel genotypes are selected in C. albicans. We provide evidence that two temporally separated events led to the formation of linChr1 via regions with (micro)homology to telomeric and subtelomeric regions. Our findings support a model in which non-allelic homologous recombination (NAHR) between the TLOα34 (Chr1) gene and Chr7 subtelomeric sequences resulted in a first deletion on Chr1 that is stabilised by the addition of Chr7 telomeric regions. We propose that, in a second step, microhomology-mediated break-induced replication (MMBIR) involving a microhomology region (6 bp 5’-TTCTTG-3’) between internal sequences of Chr1 (position Chr1: 2594815–2594820 bp) and the 23 bp telomeric repeats caused the second Chr1 deletion and addition of telomeric repeats. We hypothesise that MMBIR is also responsible for the stabilisation of ChrRR-deletion. Indeed, a microhomology tract (4 bp AxGG) between the RDN18 regions and telomeric repeats is found at ChrRR-deletion breakpoint resulting in the addition of telomeric repeats at the rDNA locus and subsequent stabilisation of the broken chromosome. As MMBIR is caused by stalled replication forks [96], it is likely that defects in DNA replication trigger MMBIR and aneuploidy selection. We suspect that these mechanistic pathways are common in C. albicans and other fungal pathogens. Indeed, complex aneuploidies have been observed in other stress conditions [30,97,98]. Complex novel karyotypes are most likely the result of independent events that accumulate over time. Furthermore, CNVs downstream of the rDNA locus have been described in C. albicans clinical isolates as well as in cells treated with the antifungal posaconazole [16,30,97]. Analysis of these genotypes by long-read sequencing will be instrumental in fully resolving these karyotypes and unveiling their origins. Ulp2 is a critical regulator of C. albicans genome stability SUMOylation is a post-translational protein modification signalling in a large number of cellular processes by targeting nuclear eukaryotic proteins [71–74]. SUMO peptides are covalently attached to target proteins by the concerted action of E1, E2 and E3 enzymes while SUMOylation is reversed by SUMO-specific proteases [75–79]. We discovered that the SUMO protease Ulp2 promotes genome stability in C. albicans. These findings are consistent with the emerging role of SUMO proteases as a guardian of genome integrity across eukaryotes. Indeed, SUMO proteases ensure genome stability throughout eukaryotes [68,80]. We hypothesise that C. albicans ULP2 promotes genome stability by modulating SUMO levels of several substrates including: (i) kinetochore and centromere-associated proteins, (ii) the DNA replication machinery and (iii) factors involved in DNA repair. Indeed, it is well established that SUMO homeostasis modulates kinetochore function, DNA replication and DNA repair and that defects in these pathways lead to exacerbated genome instability [77,78,81,82]. For example, several S. cerevisiae kinetochore and centromere-associated proteins, including the centromeric-specific histone H3 variant Cse4CENP-A, are SUMOylated and the knockdown of SENP6, the human ortholog of Ulp2, leads to mis-localisation of the inner kinetochore complex CENP-H/I/K causing chromosome segregation defects [74,83–85]. Furthermore, SUMOylation of replication and repair proteins increases upon DNA damage [74,83,86–89]. We still know very little about SUMOylation effect on C. albicans biology and its adaptation to hostile host environments. However, the observation that C. albicans protein SUMOylation patterns are different in normal and stress growth conditions agrees with our data and suggests that this post-translation modification has a critical role in adaption [90]. The adaptive power of segmental aneuploidy to overcome multiple stresses It is well established that exposure to moderate stress can increase tolerance to unrelated stresses [91,92]. This increased tolerance is usually the result of coordinated gene expression changes known as the core stress response [91]. In contrast, we show that exposure to a stress (antifungal drug) can increase tolerance to unrelated stress (loss of ULP2) by selecting segmental aneuploidies. We demonstrate that two different segmental aneuploidies can co-exist when ulp2Δ/Δ cells are challenged with FLC: an amplification of a Chr1 fragment via the formation of an extra chromosome (linChr1) and a partial deletion of ChrR. We posit that these changes in karyotype provide a synergistic fitness advantage in response to the two independent stressors (the presence of FLC and lack of ULP2) by simultaneously changing the copy number of multiple genes. Indeed, linChr1 amplifies 535 protein-coding genes (S4 Table) and GO analyses demonstrated that 41 of those genes are associated with a “response to drug” phenotype (S6 Table). Among these, amplification of UPC2 encoding for the Upc2 transcription factor is likely to be critical. Indeed, it is well established that UPC2 overexpression leads to FLC resistance by ERG11 transcriptional upregulation [93,94]. Similarly, amplification of CCR4 and NOT5 (part of linChr1) might play a key role in rescuing the fitness defects of the ulp2Δ/Δ strain. Ccr4 and Not5 are subunits of the evolutionarily conserved Ccr4-Not complex that modulate gene expression at multiple levels, including transcription initiation, elongation, de-adenylation and mRNA degradation [95]. It has been shown that S. cerevisiae CCR4 and NOT5 overexpression rescues the lethal defects associated with a ulp2 deletion strain [68]. Similarly, GO analysis revealed that ChrRR-Deletion leads to a reduced copy number of 34/204 genes associated with "response to stress" and 18/204 genes are linked to "response to drug" (S7 Table). Selection of segmental aneuploidies via a two-step process involving microhomology tracts One key question is to understand how complex novel genotypes are selected in C. albicans. We provide evidence that two temporally separated events led to the formation of linChr1 via regions with (micro)homology to telomeric and subtelomeric regions. Our findings support a model in which non-allelic homologous recombination (NAHR) between the TLOα34 (Chr1) gene and Chr7 subtelomeric sequences resulted in a first deletion on Chr1 that is stabilised by the addition of Chr7 telomeric regions. We propose that, in a second step, microhomology-mediated break-induced replication (MMBIR) involving a microhomology region (6 bp 5’-TTCTTG-3’) between internal sequences of Chr1 (position Chr1: 2594815–2594820 bp) and the 23 bp telomeric repeats caused the second Chr1 deletion and addition of telomeric repeats. We hypothesise that MMBIR is also responsible for the stabilisation of ChrRR-deletion. Indeed, a microhomology tract (4 bp AxGG) between the RDN18 regions and telomeric repeats is found at ChrRR-deletion breakpoint resulting in the addition of telomeric repeats at the rDNA locus and subsequent stabilisation of the broken chromosome. As MMBIR is caused by stalled replication forks [96], it is likely that defects in DNA replication trigger MMBIR and aneuploidy selection. We suspect that these mechanistic pathways are common in C. albicans and other fungal pathogens. Indeed, complex aneuploidies have been observed in other stress conditions [30,97,98]. Complex novel karyotypes are most likely the result of independent events that accumulate over time. Furthermore, CNVs downstream of the rDNA locus have been described in C. albicans clinical isolates as well as in cells treated with the antifungal posaconazole [16,30,97]. Analysis of these genotypes by long-read sequencing will be instrumental in fully resolving these karyotypes and unveiling their origins. Material and methods Yeast strains and growth conditions Strains used in this study are listed in S8 Table. Routine culturing was performed at 30°C in Yeast Extract-Peptone-D-Glucose (YPD) liquid and solid media containing 1% yeast extract, 2% peptone, 2% dextrose, 0.1 mg/ml adenine and 0.08 mg/ml uridine, Synthetic Complete (SC-Formedium) or Casitone (5 g/L Yeast extract, 9 g/L BactoTryptone, 20 g/L Glucose, 11.5 g/L Sodium Citrate dehydrate, 15 g/L Agar) media. When indicated, media were supplemented with 1 mg/ml 5-Fluorotic acid (5-FOA, Melford), 200 μg/ml Nourseothricin (clonNAT, Melford), 15 μg/ml and 128 μg/ml fluconazole (Sigma #F8929), 6m H2O2 (Sigma #H1009), 12 mM and 22 mM Hydroxyurea (Sigma #H8627), 0.005% MMS (Sigma #129925). Genetic screening The genetic screening was performed using a C. albicans homozygous deletion library [36] arrayed in 96 colony format on YPD plates (145x20 mm) using a replica plater (Sigma #R2508). Control non-selective (N/S) plates were grown at 30°C for 48 hours. UV treatment was performed using UVitec (Cambridge) with power density of 7.5 μW/cm2 (0.030 J for 4 seconds). Following UV treatment, plates were incubated in the dark at 30°C for 48 hours. For MMS treatment, the library was spotted onto YPD plates (145x20 mm) containing 0.005% MMS and incubated at 30°C for 48 hours. UV and/or MMS sensitivity of selected strains was confirmed by serial dilution assays in control (YPD) and stress (UV: power density of 7.5 μW/cm2, MMS: 0.005%) plates. Correct gene deletions were confirmed by PCR using gene-specific primers (S9 Table). Yeast strain construction Integration and deletion of genes were performed by transforming PCR products containing a marker gene and the appropriate target-gene sequence integration site [99]. Oligonucleotides and plasmids used for strain construction are listed in S9 and S10 Tables, respectively. For Lithium Acetate transformation, overnight liquid yeast cultures were diluted in fresh YPD and grown to an OD600 of 1.3. Cells were harvested by centrifugation and washed once with dH2O and once with SORB solution (100 mM Lithium acetate, 10 mM Tris-HCl pH 7.5, 1 mM EDTA pH 7.5/8, 1M sorbitol; pH 8). The pellet was resuspended in SORB solution containing single-stranded carrier DNA (Sigma-Aldrich) and stored -80°C in 50 μl aliquots. Frozen competent cells were defrosted on ice, mixed with 5 μL of PCR product and 300 μL PEG solution (100 mM Lithium acetate, 10 mM Tris-HCl pH 7.5, 1 mM EDTA pH 8, 40% PEG4000). Following incubation for 21–24 hours at 30°C, cells were heat-shocked at 44°C for 15 minutes and grown in 5 mL YPD liquid for 6 hours before plating on selective media at 30°C. UV survival quantification Following dilution of overnight liquid cultures, 500 cells were plated in YPD control plates and 1500 cells were plated in YPD stress plates and UV irradiated with a power density of 7.5 μW/cm2 (0.030 J for 4 seconds). Plates were incubated at 30°C for 48 hours in the dark. Colonies were counted using a colony counter (Stuart Scientific). Experiments were performed in 5 biological replicates, and violin plots were generated using R and R Studio IDE (http://www.r-project.org/). Growth curve Overnight liquid cultures were diluted to 60 cells/μL in 100 μL YPD and incubated at 30°C in a 96 well plate (Cellstar, #655180) with double orbital agitation of 400 rpm using a BMG Labtech SPECTROstar nanoplate reader for 48 hours. When indicated, YPD media was supplemented with MMS (0.005%) and HU (22 mM). Graphs show the mean of 3 biological replicates, error bars show the standard deviation. Serial dilution assay Overnight liquid cultures were diluted to an OD600 of 4, serially diluted 1:5 and spotted into agar plates with and without indicated additives using a replica plater (Sigma Aldrich, #R2383). Images of the plates were taken using Syngene GBox Chemi XX6 Gel imaging system. Experiments were performed in 3 biological replicates. Protein extraction and Western blotting Yeast extracts were prepared as described [100] using 1 × 108 cells from overnight cultures grown to a final OD600 of 1.5–2. Protein extraction was performed in the presence of 2% SDS (Sigma) and 4 M acetic acid (Fisher) at 90°C. Proteins were separated in 2% SDS (Sigma), 40% acrylamide/bis (Biorad, 161–0148) gels and transferred into PVDF membrane (Biorad) by semi-dry transfer (Biorad, Trans Blot SD, semi-dry transfer cell). Western-blot antibody detection was performed using anti-HA mouse monoclonal primary antibody (12CA5 Roche, 5 mg/ml) at a dilution of 1:1000 in PBS containing 0.2% Tween and 5% w/v non-fat dry milk, recombinant anti-alpha Tubulin (Abcam #ab184970) at a dilution of 1:10000 in PBS containing 0.2% Tween and 5% w/v non-fat dry milk, anti-mouse IgG-peroxidase (A4416 Sigma) at dilution of 1:30000, anti-rabbit IgG-peroxidase (A0545 Sigma) at a dilution of 1:30000, and Clarity ECL substrate (Bio-Rad). URA3+ marker loss quantification Strains were first streaked onto synthetic solid media lacking uracil and uridine (SC–Uri) to ensure the selection of cells carrying the URA3+ marker gene. Parallel liquid cultures were grown for 16 hours at 30°C in YPD and plated on SC plates containing 1 mg/ml 5-FOA (5-fluorotic acid; Sigma) and on N/S SC plates. Colonies were counted after 2 days of growth at 30°C. The frequency of the URA3+ marker loss was calculated using the formula F = m/M, where m represents the median number of colonies obtained on 5-FOA medium (corrected by the dilution factor used and the fraction of culture plated) and M the average number of colonies obtained on YPD (corrected by the dilution factor used and the fraction of culture plated) [63]. Statistical differences between results from samples were calculated using the Kruskal-Wallis test and the Mann-Whitney U test for post hoc analysis. Statistical analysis was performed and violin plots were generated using R Studio (http://www.r-project.org/). Microscopy 30 ml of yeast cultures (OD600 = 1) grown in SC were centrifuged at 1550 x g for 5 minute and washed once with dH2O. Cells were fixed in 10 ml of 3.7% paraformaldehyde (Sigma #F8775) for 15 minutes, washed twice with 10 ml of KPO4/Sorbitol (100 mM KPO4, 1.2 M Sorbitol) and resuspended in 250 μl PBS containing 10 μg of DAPI. Cells were then sonicated and resuspended in a 1% low melting point agarose (Sigma Aldrich) before mounting under a 22 mm coverslip of 0,17 μm thickness. Samples were imaged on a Zeiss LSM 880 Airyscan with a 63x/1.4NA oil objective. Airyscan images were taken with a relative pinhole diameter of 0.2 AU (airy unit) for maximal resolution and reduced noise. GFP was imaged with a 488 nm Argon laser and 495–550 nm bandpass excitation filter. The DAPI channel was imaged on a PMT with standard pinhole of 1AU and brightfield images were captured on the trans-PMT with the same excitation laser of 405 nm. DAPI and brightfield images were taken with the same pixel size and bit depth (16bit) as the airyscan images. Images were of a 42.7x42.7μm field of view with a 33 nm pixel size resolution. z-stacks were taken containing cells of z interval of 500 nm. Airyscan Veena filtering was performed with the inbuilt algorithms of Zeiss Zen Black 2.3. Experiments were performed in 3 biological replicates and >100 cells/replicate were counted. Drug selection For fluconazole selection, strains were incubated overnight in Casitone liquid media at 30°C with shaking. 104 cells were plated in a small plate (10 cm) containing Casitone medium plus 256μl DMSO or 128 μg/mL fluconazole. Plates were incubated at 30°C for 7 days. Colonies able to grow on fluconazole- were streaked (2X) on non-selective (N/S) plates and tested by spotting assay in Casitone +DMSO, or Casitone+FLC. For selection of FLC-1 revertants, 100 cells were plated in YPD agar plates and single colonies were assessed for their ability to grow on casitone medium plus 256μl DMSO or 128 μg/mL fluconazole by serial dilution assays. Contour-clamped homogeneous electric field (CHEF) electrophoresis Intact yeast chromosomal DNA was prepared as previously described [101]. Briefly, cells were grown overnight, and a volume equivalent to an OD600 of 6 was washed in 50 mM EDTA and resuspended in 20 μl of 10 mg/ml Zymolyase 100T (Amsbio #120493–1) and 300 μl of 1% Low Melt agarose (Biorad # 1613112) in 100 mM EDTA. Chromosomes were separated on a 1% Megabase agarose gel (Bio-Rad) in 0.5X TBE using a CHEF DRII apparatus. Run conditions were as follows: 60-120s switch at 6 V/cm for 12 hours followed by a 120-300s switch at 4.5 V/cm for 26 hours at 14°C. The gel was stained in 0.5x TBE with ethidium bromide (0.5 μg/ml) for 60 minutes and destained in water for 30 minutes. Chromosomes were visualised using a Syngene GBox Chemi XX6 gel imaging system. Whole-genome Illumina sequence analysis Illumina genome sequencing data have been deposited in the Sequence Read Archive under BioProject PRJNA781758. Genomic DNA was isolated using a phenol-chloroform extraction as previously described [26]. Paired-end (2 x 151 bp) sequencing was carried out by the Microbial Genome Sequencing Center (MiGS) on the Illumina NextSeq 2000 platform. Read trimming was conducted using Trimmomatic (v0.33 LEADING:3 Trailing:3 SLIDINGWINDOW:4:15 MINLEN:36 TOPHRED33) [102]. Trimmed reads were mapped to the C. albicans reference genome (SC5314 A21-s02-m09-r08) from the Candida Genome Database (http://www.candidagenome.org/download/sequence/C_albicans_SC5314/Assembly21/archive/C_albicans_SC5314_version_A21-s02-m09-r08_chromosomes.fasta.gz). Reads were aligned to the reference using BWA-MEM (v0.7.17) with default parameters [103]. The BAM files, containing aligned reads, were sorted and PCR duplicates removed using Samtools (v1.10 samtools sort, samtools rmdup) [104]. Qualimap (v2.2.1) analysed the BAM files for mean coverage as aligned to the SC5314 A21 reference genome; coverages ranged from 73.7x to 89.3x coverage [105]. Variant detection was conducted using the Genome Analysis Toolkit (Mutect, v2.2–25), with the SC5314 A21 reference genome as the reference fasta input, and SN152 as the normal bam input [106]. Variants were annotated using SnpEff (V4.3) [107] using the SC5314 A21 reference genome fasta and gene feature file above, and filtered using SnpSift for missense, nonsense, and synonymous mutations. Variants were verified visually using the Integrative Genomic Viewer, and variants present in SN152 were removed. (IGV, v2.8.2) [108]. Read depth and breakpoint analysis of short-reads sequencing Whole-genome sequencing data were analysed for copy number and allele ratio changes as previously described [19,30]. Aneuploidies were visualised using the Yeast Mapping Analysis Pipeline (YMAP, v1.0) [109]. BAM files aligned to the SC5314 reference genome as described above were uploaded to YMAP and read depth was determined and plotted as a function of chromosome position, using the inbuilt SC5314 A21 reference genome and haplotype map. Read depth was corrected for both chromosome-end bias and GC-content. The GBrowse CNV track and allele ratio track identified regions of interest for CNV and LOH breakpoints, and more precise breakpoints were determined visually using IGV. LOH breakpoints are reported as the first and last informative homozygous position in a region that is heterozygous in the parental genome. CNV breakpoints were identified as described previously [19,30]. Long-read DNA sequencing Oxford Nanopore Technologies (ONT) sequencing data have been deposited in the Sequence Read Archive under BioProject PRJNA879282. DNA was extracted from an overnight culture in YPD using the QIAGEN genomic tip 100/G kit (Qiagen, #10243) according to manufacturing protocol. Long read sequencing libraries were prepared using the SQK-LSK109 Ligation Sequencing Kit with the EXP-NBD104 Native Barcoding Kit (Oxford Nanopore Technologies) from approximately 1μg of high molecular weight genomic DNA, following the manufacturer’s protocol. Long read libraries were sequenced on R9.4.1 Spot-On Flow cells (FLO-MIN106) using the GridION X5 platform set to super accurate base calling. Long-read genome assembly Long reads were quality controlled using NanoPlot v1.30.1 [110], and adapters and barcodes were trimmed using Porechop v0.2.4 (https://github.com/rrwick/Porechop) with default parameters. Reads shorter than 1kb or with a quality score less than Q9 were removed using Filtlong v0.2.1 https://github.com/rrwick/Filtlong. Long reads were assembled using NECAT v0.0.1_update20200803 [111] using a genome size of 16Mb, all other parameters were left as default. Error correction was performed by aligning the long reads to the assemblies with Minimap2 v2.17-r941 [112] to inform one iteration of Racon v1.4.20 [113], followed by one iteration of Medaka v1.5.0 (https://github.com/nanoporetech/medaka) using the r941_min_high_g360 model. Assembly statistics were generated using a custom Python script, and single copy ortholog analysis was performed using BUSCO v5.2.2 [114], using the saccharomycetes_odb10 database. Identification of segmental aneuploidies Raw long-reads were first aligned to the reference SC5314 genome using Minimap2 (v 2.17-r941) [112] and coverage plotted using Circos (v0.69–8) [115]. Raw reads sorting and indexing was performed with Samtools (v1.11) [116], bam to fasta conversion was conducted with bedtools (v2.30.0) [117] and visualised using IGV (v2.11.9). Presence of telomeres was confirmed by extracting raw reads at the target regions using samtools (v1.11). Individual long reads spanning the breakpoints were investigated and annotated in SnapGene Viewer. Yeast strains and growth conditions Strains used in this study are listed in S8 Table. Routine culturing was performed at 30°C in Yeast Extract-Peptone-D-Glucose (YPD) liquid and solid media containing 1% yeast extract, 2% peptone, 2% dextrose, 0.1 mg/ml adenine and 0.08 mg/ml uridine, Synthetic Complete (SC-Formedium) or Casitone (5 g/L Yeast extract, 9 g/L BactoTryptone, 20 g/L Glucose, 11.5 g/L Sodium Citrate dehydrate, 15 g/L Agar) media. When indicated, media were supplemented with 1 mg/ml 5-Fluorotic acid (5-FOA, Melford), 200 μg/ml Nourseothricin (clonNAT, Melford), 15 μg/ml and 128 μg/ml fluconazole (Sigma #F8929), 6m H2O2 (Sigma #H1009), 12 mM and 22 mM Hydroxyurea (Sigma #H8627), 0.005% MMS (Sigma #129925). Genetic screening The genetic screening was performed using a C. albicans homozygous deletion library [36] arrayed in 96 colony format on YPD plates (145x20 mm) using a replica plater (Sigma #R2508). Control non-selective (N/S) plates were grown at 30°C for 48 hours. UV treatment was performed using UVitec (Cambridge) with power density of 7.5 μW/cm2 (0.030 J for 4 seconds). Following UV treatment, plates were incubated in the dark at 30°C for 48 hours. For MMS treatment, the library was spotted onto YPD plates (145x20 mm) containing 0.005% MMS and incubated at 30°C for 48 hours. UV and/or MMS sensitivity of selected strains was confirmed by serial dilution assays in control (YPD) and stress (UV: power density of 7.5 μW/cm2, MMS: 0.005%) plates. Correct gene deletions were confirmed by PCR using gene-specific primers (S9 Table). Yeast strain construction Integration and deletion of genes were performed by transforming PCR products containing a marker gene and the appropriate target-gene sequence integration site [99]. Oligonucleotides and plasmids used for strain construction are listed in S9 and S10 Tables, respectively. For Lithium Acetate transformation, overnight liquid yeast cultures were diluted in fresh YPD and grown to an OD600 of 1.3. Cells were harvested by centrifugation and washed once with dH2O and once with SORB solution (100 mM Lithium acetate, 10 mM Tris-HCl pH 7.5, 1 mM EDTA pH 7.5/8, 1M sorbitol; pH 8). The pellet was resuspended in SORB solution containing single-stranded carrier DNA (Sigma-Aldrich) and stored -80°C in 50 μl aliquots. Frozen competent cells were defrosted on ice, mixed with 5 μL of PCR product and 300 μL PEG solution (100 mM Lithium acetate, 10 mM Tris-HCl pH 7.5, 1 mM EDTA pH 8, 40% PEG4000). Following incubation for 21–24 hours at 30°C, cells were heat-shocked at 44°C for 15 minutes and grown in 5 mL YPD liquid for 6 hours before plating on selective media at 30°C. UV survival quantification Following dilution of overnight liquid cultures, 500 cells were plated in YPD control plates and 1500 cells were plated in YPD stress plates and UV irradiated with a power density of 7.5 μW/cm2 (0.030 J for 4 seconds). Plates were incubated at 30°C for 48 hours in the dark. Colonies were counted using a colony counter (Stuart Scientific). Experiments were performed in 5 biological replicates, and violin plots were generated using R and R Studio IDE (http://www.r-project.org/). Growth curve Overnight liquid cultures were diluted to 60 cells/μL in 100 μL YPD and incubated at 30°C in a 96 well plate (Cellstar, #655180) with double orbital agitation of 400 rpm using a BMG Labtech SPECTROstar nanoplate reader for 48 hours. When indicated, YPD media was supplemented with MMS (0.005%) and HU (22 mM). Graphs show the mean of 3 biological replicates, error bars show the standard deviation. Serial dilution assay Overnight liquid cultures were diluted to an OD600 of 4, serially diluted 1:5 and spotted into agar plates with and without indicated additives using a replica plater (Sigma Aldrich, #R2383). Images of the plates were taken using Syngene GBox Chemi XX6 Gel imaging system. Experiments were performed in 3 biological replicates. Protein extraction and Western blotting Yeast extracts were prepared as described [100] using 1 × 108 cells from overnight cultures grown to a final OD600 of 1.5–2. Protein extraction was performed in the presence of 2% SDS (Sigma) and 4 M acetic acid (Fisher) at 90°C. Proteins were separated in 2% SDS (Sigma), 40% acrylamide/bis (Biorad, 161–0148) gels and transferred into PVDF membrane (Biorad) by semi-dry transfer (Biorad, Trans Blot SD, semi-dry transfer cell). Western-blot antibody detection was performed using anti-HA mouse monoclonal primary antibody (12CA5 Roche, 5 mg/ml) at a dilution of 1:1000 in PBS containing 0.2% Tween and 5% w/v non-fat dry milk, recombinant anti-alpha Tubulin (Abcam #ab184970) at a dilution of 1:10000 in PBS containing 0.2% Tween and 5% w/v non-fat dry milk, anti-mouse IgG-peroxidase (A4416 Sigma) at dilution of 1:30000, anti-rabbit IgG-peroxidase (A0545 Sigma) at a dilution of 1:30000, and Clarity ECL substrate (Bio-Rad). URA3+ marker loss quantification Strains were first streaked onto synthetic solid media lacking uracil and uridine (SC–Uri) to ensure the selection of cells carrying the URA3+ marker gene. Parallel liquid cultures were grown for 16 hours at 30°C in YPD and plated on SC plates containing 1 mg/ml 5-FOA (5-fluorotic acid; Sigma) and on N/S SC plates. Colonies were counted after 2 days of growth at 30°C. The frequency of the URA3+ marker loss was calculated using the formula F = m/M, where m represents the median number of colonies obtained on 5-FOA medium (corrected by the dilution factor used and the fraction of culture plated) and M the average number of colonies obtained on YPD (corrected by the dilution factor used and the fraction of culture plated) [63]. Statistical differences between results from samples were calculated using the Kruskal-Wallis test and the Mann-Whitney U test for post hoc analysis. Statistical analysis was performed and violin plots were generated using R Studio (http://www.r-project.org/). Microscopy 30 ml of yeast cultures (OD600 = 1) grown in SC were centrifuged at 1550 x g for 5 minute and washed once with dH2O. Cells were fixed in 10 ml of 3.7% paraformaldehyde (Sigma #F8775) for 15 minutes, washed twice with 10 ml of KPO4/Sorbitol (100 mM KPO4, 1.2 M Sorbitol) and resuspended in 250 μl PBS containing 10 μg of DAPI. Cells were then sonicated and resuspended in a 1% low melting point agarose (Sigma Aldrich) before mounting under a 22 mm coverslip of 0,17 μm thickness. Samples were imaged on a Zeiss LSM 880 Airyscan with a 63x/1.4NA oil objective. Airyscan images were taken with a relative pinhole diameter of 0.2 AU (airy unit) for maximal resolution and reduced noise. GFP was imaged with a 488 nm Argon laser and 495–550 nm bandpass excitation filter. The DAPI channel was imaged on a PMT with standard pinhole of 1AU and brightfield images were captured on the trans-PMT with the same excitation laser of 405 nm. DAPI and brightfield images were taken with the same pixel size and bit depth (16bit) as the airyscan images. Images were of a 42.7x42.7μm field of view with a 33 nm pixel size resolution. z-stacks were taken containing cells of z interval of 500 nm. Airyscan Veena filtering was performed with the inbuilt algorithms of Zeiss Zen Black 2.3. Experiments were performed in 3 biological replicates and >100 cells/replicate were counted. Drug selection For fluconazole selection, strains were incubated overnight in Casitone liquid media at 30°C with shaking. 104 cells were plated in a small plate (10 cm) containing Casitone medium plus 256μl DMSO or 128 μg/mL fluconazole. Plates were incubated at 30°C for 7 days. Colonies able to grow on fluconazole- were streaked (2X) on non-selective (N/S) plates and tested by spotting assay in Casitone +DMSO, or Casitone+FLC. For selection of FLC-1 revertants, 100 cells were plated in YPD agar plates and single colonies were assessed for their ability to grow on casitone medium plus 256μl DMSO or 128 μg/mL fluconazole by serial dilution assays. Contour-clamped homogeneous electric field (CHEF) electrophoresis Intact yeast chromosomal DNA was prepared as previously described [101]. Briefly, cells were grown overnight, and a volume equivalent to an OD600 of 6 was washed in 50 mM EDTA and resuspended in 20 μl of 10 mg/ml Zymolyase 100T (Amsbio #120493–1) and 300 μl of 1% Low Melt agarose (Biorad # 1613112) in 100 mM EDTA. Chromosomes were separated on a 1% Megabase agarose gel (Bio-Rad) in 0.5X TBE using a CHEF DRII apparatus. Run conditions were as follows: 60-120s switch at 6 V/cm for 12 hours followed by a 120-300s switch at 4.5 V/cm for 26 hours at 14°C. The gel was stained in 0.5x TBE with ethidium bromide (0.5 μg/ml) for 60 minutes and destained in water for 30 minutes. Chromosomes were visualised using a Syngene GBox Chemi XX6 gel imaging system. Whole-genome Illumina sequence analysis Illumina genome sequencing data have been deposited in the Sequence Read Archive under BioProject PRJNA781758. Genomic DNA was isolated using a phenol-chloroform extraction as previously described [26]. Paired-end (2 x 151 bp) sequencing was carried out by the Microbial Genome Sequencing Center (MiGS) on the Illumina NextSeq 2000 platform. Read trimming was conducted using Trimmomatic (v0.33 LEADING:3 Trailing:3 SLIDINGWINDOW:4:15 MINLEN:36 TOPHRED33) [102]. Trimmed reads were mapped to the C. albicans reference genome (SC5314 A21-s02-m09-r08) from the Candida Genome Database (http://www.candidagenome.org/download/sequence/C_albicans_SC5314/Assembly21/archive/C_albicans_SC5314_version_A21-s02-m09-r08_chromosomes.fasta.gz). Reads were aligned to the reference using BWA-MEM (v0.7.17) with default parameters [103]. The BAM files, containing aligned reads, were sorted and PCR duplicates removed using Samtools (v1.10 samtools sort, samtools rmdup) [104]. Qualimap (v2.2.1) analysed the BAM files for mean coverage as aligned to the SC5314 A21 reference genome; coverages ranged from 73.7x to 89.3x coverage [105]. Variant detection was conducted using the Genome Analysis Toolkit (Mutect, v2.2–25), with the SC5314 A21 reference genome as the reference fasta input, and SN152 as the normal bam input [106]. Variants were annotated using SnpEff (V4.3) [107] using the SC5314 A21 reference genome fasta and gene feature file above, and filtered using SnpSift for missense, nonsense, and synonymous mutations. Variants were verified visually using the Integrative Genomic Viewer, and variants present in SN152 were removed. (IGV, v2.8.2) [108]. Read depth and breakpoint analysis of short-reads sequencing Whole-genome sequencing data were analysed for copy number and allele ratio changes as previously described [19,30]. Aneuploidies were visualised using the Yeast Mapping Analysis Pipeline (YMAP, v1.0) [109]. BAM files aligned to the SC5314 reference genome as described above were uploaded to YMAP and read depth was determined and plotted as a function of chromosome position, using the inbuilt SC5314 A21 reference genome and haplotype map. Read depth was corrected for both chromosome-end bias and GC-content. The GBrowse CNV track and allele ratio track identified regions of interest for CNV and LOH breakpoints, and more precise breakpoints were determined visually using IGV. LOH breakpoints are reported as the first and last informative homozygous position in a region that is heterozygous in the parental genome. CNV breakpoints were identified as described previously [19,30]. Long-read DNA sequencing Oxford Nanopore Technologies (ONT) sequencing data have been deposited in the Sequence Read Archive under BioProject PRJNA879282. DNA was extracted from an overnight culture in YPD using the QIAGEN genomic tip 100/G kit (Qiagen, #10243) according to manufacturing protocol. Long read sequencing libraries were prepared using the SQK-LSK109 Ligation Sequencing Kit with the EXP-NBD104 Native Barcoding Kit (Oxford Nanopore Technologies) from approximately 1μg of high molecular weight genomic DNA, following the manufacturer’s protocol. Long read libraries were sequenced on R9.4.1 Spot-On Flow cells (FLO-MIN106) using the GridION X5 platform set to super accurate base calling. Long-read genome assembly Long reads were quality controlled using NanoPlot v1.30.1 [110], and adapters and barcodes were trimmed using Porechop v0.2.4 (https://github.com/rrwick/Porechop) with default parameters. Reads shorter than 1kb or with a quality score less than Q9 were removed using Filtlong v0.2.1 https://github.com/rrwick/Filtlong. Long reads were assembled using NECAT v0.0.1_update20200803 [111] using a genome size of 16Mb, all other parameters were left as default. Error correction was performed by aligning the long reads to the assemblies with Minimap2 v2.17-r941 [112] to inform one iteration of Racon v1.4.20 [113], followed by one iteration of Medaka v1.5.0 (https://github.com/nanoporetech/medaka) using the r941_min_high_g360 model. Assembly statistics were generated using a custom Python script, and single copy ortholog analysis was performed using BUSCO v5.2.2 [114], using the saccharomycetes_odb10 database. Identification of segmental aneuploidies Raw long-reads were first aligned to the reference SC5314 genome using Minimap2 (v 2.17-r941) [112] and coverage plotted using Circos (v0.69–8) [115]. Raw reads sorting and indexing was performed with Samtools (v1.11) [116], bam to fasta conversion was conducted with bedtools (v2.30.0) [117] and visualised using IGV (v2.11.9). Presence of telomeres was confirmed by extracting raw reads at the target regions using samtools (v1.11). Individual long reads spanning the breakpoints were investigated and annotated in SnapGene Viewer. Supporting information S1 Fig. Genomic variants are selected in ulp2Δ/Δ cells challenged with Fluconazole. (A) Whole genome sequence data were plotted as the log2 ratio and converted to chromosome copy number (y-axis, 1–4 copies) as a function of chromosome position (x-axis, Chr1-ChrR) using YMAP. Heterozygous (AB) regions are indicated with gray shading and homozygous regions are indicated by haplotype AA (cyan) or BB (magenta). Allele ratio changes that occur within a CNV are indicated as dark blue (AAB) or purple (ABB). Colony B and C had allele ratio colouring that was corrected using IGV and allele frequency information. (B) Serial dilution assay of ulp2Δ/Δ parental (P) and fluconazole-recovered isolates (FLC-2, FLC-3, FLC-4, FLC-1a, FLC-1b and FLC-1c) in non-selective (N/S) or media containing 128 μg/ml fluconazole (FLC). https://doi.org/10.1371/journal.pgen.1010576.s001 (PDF) S1 Table. Genetic screen top hits (score ≥2). https://doi.org/10.1371/journal.pgen.1010576.s002 (DOCX) S2 Table. List of SNPs detected in the ulp2Δ/Δ colonies (U1, U2 and U3) sequenced in this study. https://doi.org/10.1371/journal.pgen.1010576.s003 (XLSX) S3 Table. List of SNPs detected in ulp2Δ/Δ isolates selected from Fluconazole (128 μg/ml) plates. https://doi.org/10.1371/journal.pgen.1010576.s004 (XLSX) S4 Table. Coordinates of the Chromosome 1 amplification (lin-Chr1) detected in the FLC-1 ulp2Δ/Δ isolate. https://doi.org/10.1371/journal.pgen.1010576.s005 (XLSX) S5 Table. Coordinates of the Chromosome R deletion (ChrRR-Deletion) detected in FLC-1, FLC-2, FLC-3 and FLC-4 ulp2Δ/Δ isolates. https://doi.org/10.1371/journal.pgen.1010576.s006 (XLSX) S6 Table. List of genes associated with "response to drugs" GO terms for the Chromosome 1 amplification (lin-Chr1). https://doi.org/10.1371/journal.pgen.1010576.s007 (XLSX) S7 Table. List of genes associated with "response to stress" and "response to drugs" GO terms for the Chromosome R deletion (ChrRR-deletion). https://doi.org/10.1371/journal.pgen.1010576.s008 (XLSX) S8 Table. Strains used in this study. https://doi.org/10.1371/journal.pgen.1010576.s009 (DOCX) S9 Table. Oligonucleotides used in this study. https://doi.org/10.1371/journal.pgen.1010576.s010 (DOCX) S10 Table. Plasmids used in this study. https://doi.org/10.1371/journal.pgen.1010576.s011 (DOCX) Acknowledgments We thank J. Berman for reagents, strains and materials and A. Pidoux for critically reading the manuscript.

Lama, Jyoti;Srivastav, Satyam;Tasnim, Sadia;Hubbard, Donald;Hadjipanteli, Savana;Smith, Brittny R.;Macdonald, Stuart J.;Green, Llewellyn;Kelleher, Erin S.

doi: 10.1371/journal.pgen.1010080pmid: 36477699

Introduction Transposable elements (TE) are mobile DNA sequences that spread through host genomes by replicating in germline cells. Although individual TE insertions are sometimes beneficial, genomic TEs are foremost genetic parasites reviewed in [1]. Unrestricted transposition not only produces deleterious mutations, but also double-stranded breaks (DSBs) that lead to genotoxic stress in developing gametes. The mechanisms by which hosts enact silencing of resident TEs through the heritable production of regulatory small RNAs is extensively studied and broadly conserved [2,3]. However, host genomes are frequently invaded by new TE families, against which they lack small RNA-mediated “immunity” [4–7]. In the context of such novel TEs, genetic variation in the host’s ability to produce gametes could be a critical determinant of fitness. The presence and mechanisms of such host variation remain largely unstudied. P-element DNA transposons, which invaded natural populations of Drosophila melanogaster around 1950, provide a unique opportunity to uncover host genetic variation in transposition-dependent sterility [8–10]. Strains isolated from natural populations prior to this invasion, referred to as M strains, do not contain genomic P-elements, and do not produce maternally-transmitted piRNAs that control their expression and transposition. When females from M strains are mated to males bearing genomic P-elements (P-strains), they produce dysgenic offspring that do not negatively regulate P-elements in germline cells [11]. A range of fertility effects result from unregulated P-element transposition, including the complete loss of germline cells and sterility [12]. However, naive M genotypes differ in their propensity to produce dysgenic progeny suggesting genetic variation in dysgenic sterility [8,10,13,14]. One potential source of variation in P-element dysgenic sterility arises from the response of germline cells to DSBs arising from transposition. In dysgenic females, primordial germ cells (PGCs) are lost beginning in the second instar larval stage, most likely due to unrepaired DSBs [15–17]. Furthermore, mutations in DNA damage response and repair proteins are known to enhance dysgenic germ cell loss [18,19]. Therefore, it is predicted that genotypes that enact more efficient DSB repair should be more tolerant of P-element transposition, and maintain germline cells. A related response is the production of de novo piRNAs, which is triggered by the activation of the DNA damage response protein checkpoint kinase 2 (CHK2) in the adult female germline [20,21]. These de novo piRNAs transcriptionally silence P-elements in a process analogous to maternally transmitted silencing, and restore fertility as dysgenic females age [20,21]. If de novo piRNA production activates more readily in some genotypes than others, it could lead to fertility differences in dysgenic crosses. Another potential source of variation in dysgenic sterility lies with host co-factors of transposition, including host proteins that regulate the transcription and splicing of transposase-encoding RNA, or the activity of P-element transposase enzyme. In particular, differences in splicing cofactors between germline and somatic cells ensure P-element transposase-encoding transcripts are only produced in germline cells [22]. However, individual germlines could differ in the production of these cofactors. Beyond transposase production, P-elements insert preferentially near origins of replication, a strategy that may facilitate their spread through the genome by ensuring they are replicated multiple times in a given S-phase [23]. Differences in the timing or composition of this machinery could therefore drive differences in transposition rates, and ultimately downstream germline loss. We recently isolated natural variation in dysgenic sterility through QTL mapping, using a panel of highly recombinant inbred lines derived from M strains (Drosophila Synthetic Population Resource, DSPR, Population A RILs, [24]). We mapped a major effect QTL surrounding the gene bruno, a female germline differentiation factor [14]. Here we present results from a second QTL mapping study in an independent panel of DSPR RILs (Population B, [24]). We describe two QTL that determine differences in dysgenic sterility, one in young females only, and one in aged females only. Focusing on young dysgenic females, we further interrogated mechanisms underlying fertility differences by contrasting RNA and small RNA expression, radiation sensitivity, and P-element expression and splicing between fertile and sterile genotypes. Our results suggest that natural variation in dysgenic sterility arises through differences in both germline DNA repair and P-transposase mRNA splicing, revealing considerable complexity in host factors that modulate the fitness costs associated with transposition. Results QTL mapping The DSPR RILs are all P-element free M-strains, which were derived from founders isolated from natural populations before the P-element invasion [24]. We therefore screened for alleles that influence dysgenic sterility among the panel B RIL genomes by crossing RIL females to males from the reference P-strain Harwich, and examining the morphology of F1 ovaries (Fig 1A). Atrophied ovaries are indicative of germline loss resulting from P-element activity [14,25]. Since dysgenic sterility changes across development [15], and some females exhibit age-dependent recovery from P-element hybrid dysgenesis through the production of de novo piRNAs [20], we phenotyped F1 females at two developmental time points: 3 days and 21 days post-eclosion. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. QTL mapping of variation in P-element induced ovarian atrophy. a) Crossing scheme to phenotype the variation in F1 ovarian atrophy among RIL offspring. Representative images of atrophied and non-atrophied ovaries are from Kelleher et al. [14] b) The log of odds (LOD) plot for QTL mapping of ovarian atrophy using 3 day-old (gold) and 21 day-old (blue) F1 females. The dotted line is the LOD threshold and x-axis represents the chromosomal positions. c) Zoomed-in figure of QTL mapping from 3 days (gold) and 21 days (blue). The colored boxes show the genomic interval that likely contains the causative genetic variant of each QTL, based on a Δ2LOD drop from the peak position [26]. The pairs of dotted lines indicate the peak Δ2LOD scores that determines the interval. The solid horizontal line is the LOD significance threshold based on 1,000 permutations of the phenotype data. d) Cytological map depicting the interval of the two QTL peaks [27,28]. e) Residual F1 atrophy (y-axis) associated with each of the eight founder alleles (x-axis) at the QTL peaks after accounting for random effects. All the QTL peaks show 2 phenotypic classes: sterile (light green) and fertile (dark green). (f-g) Percentage of (f) ovarian atrophy and (g) sterility among dysgenic female offspring from crosses between Harwich males and isogenic females carrying sterile (B6) and fertile (B8) alleles. Proportions were compared between samples using χ2 tests of independence. h) Number of F2 offspring produced by individual dysgenic F1 females from crosses between Harwich males and isogenic fertile and sterile females. The horizontal line indicates the mean, which was compared between samples using permutation tests. Superscripts1,2 and 3 in (f-h) denote isogenic lines that were independently generated, these sometimes differ between experiments because sterile_B61 became contaminated and was replaced with sterile_B63. Error bars in e, f and g represent the standard error. The data used to generate plot in panels b,c, and e are provided in S3 and S4 Tables and that used for plot in panels f, g and h are provided in S17 and S18 Tables respectively. * denotes P < 0.05, ** denotes P < 0.01, *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g001 Similar to our observations with the Population A RILs [14], we found continuous variation in the frequency of ovarian atrophy among dysgenic offspring of different RIL mothers, indicating genetic variation in dysgenic sterility that is unrelated to maternally deposited piRNAs (S1 and S2 Tables). Based on a combined linear model of F1 atrophy among 3 and 21 day-old females, we estimated the broad-sense heritability of dysgenic ovarian atrophy in our experiment to be ~42.5%. The effect of age on the proportion of F1 atrophy was significant but minimal (χ2 = 7.03, df = 1, p-value = 0.008) with 21 day-old females showing only 0.7% increase in atrophy as compared to 3 day-old females. This suggests that age-dependent recovery from dysgenic ovarian atrophy through the production of de novo piRNAs is not common among the genotypes we sampled. To identify the genomic regions associated with genetic variation in dysgenic ovarian atrophy, we performed QTL mapping using the published RIL genotypes [24]. We found a large QTL peak near the 2nd chromosome centromere in both 3 and 21 day-old F1 females (Fig 1B and Tables 1, S3, and S4). However, the genomic intervals of the two QTL are non-overlapping (Fig 1C and Table 1). The major QTL in 21 day-old females (hereafter, QTL-21d) resides in the euchromatic region and is quite small (990 kb) compared to the major QTL in 3 day-old females (hereafter QTL-3d), which spans the centromere and pericentromeric regions (9.6 Mb, Fig 1D). Therefore, there are likely at least two polymorphisms that influence tolerance near the 2nd chromosome centromere, one of which has a larger effect in young 3-day old females, with the other having a larger effect in 21 day-old females. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. QTL positions in 3 and 21-day old females. The peak position, Δ2LOD drop confidence interval (2LOD CI), and the Bayesian Credible Interval (BCI) in dm6 [29] are provided for each analysis. The data used to identify the LOD peaks and intervals for 3 and 21-day old females can be found in S3 and S4 Tables, respectively. https://doi.org/10.1371/journal.pgen.1010080.t001 We further evaluated the effect of the two linked QTL through haplotype analysis. We modeled residual F1 ovarian atrophy as a function of QTL haplotype for the 3 day and 21 day peaks, thereby disentangling synergistic (e.g. sterile 3d, sterile 21d) from opposing (e.g. sterile 3d, fertile 21d) allelic combinations (S4 Fig). We observed that the 3 day-old QTL allele is solely-determinant of ovarian atrophy in the 3 day-old offspring. However, in 21 day-old offspring only the genotypes containing fertile alleles at both QTL show decreased atrophy. This suggests that QTL-3d may determine germ cell maintenance in the larval, pupal and early adult stages, but QTL-21d may be additionally required to maintain germline cells in aging females. The presence of two QTL is further supported by the phenotypic classes we detected among founder alleles (B1-B8) for each of the QTL peaks (Fig 1E). For QTL-21d, both B2 and B6 founder alleles are associated with greatly increased dysgenic ovarian atrophy. By contrast for QTL-3d, only the B6 founder allele is associated with increased ovarian atrophy. We next sought to determine whether reduced ovarian atrophy corresponds to restored fertility, or merely allows for the production of inviable gametes. To this end, we generated isogenic lines that carry either high-atrophy (B6) or low-atrophy (B8) alleles at both QTL loci in an otherwise identical genetic background through 6 generations of backcrossing to a marker stock (S5 Fig). Consistent with our QTL mapping, B8 alleles display less F1 ovarian atrophy (24–31%) than B6 strains when crossed with Harwich males (Fig 1F and S17 Table). Furthermore, while B6 dysgenic females produced no offspring, 13–29% of dysgenic B8 females were fertile and produced offspring (Fig 1G and S18 Table). For one B8 stock, offspring counts were significantly higher when compared to B6 (Fig 1H). In light of these observations, we refer to the low-atrophy and high-atrophy alleles hereafter as “fertile” and “sterile”. The fertility rescue conferred by these alleles would be highly beneficial in populations where dysgenic crosses are common. Sterile and fertile alleles differ in chromatin regulation Both the QTL regions contain large numbers of protein coding and non-coding RNA genes, piRNA clusters, and repeats, which could influence dysgenic sterility (Fig 1D). To better understand the differences between fertile and sterile genotypes, we compared their gene expression profiles in the ovaries of young 3–5 day-old females by stranded total RNA-seq. To avoid the confounding effects of germline loss under dysgenic conditions, we focused on RIL females rather than their dysgenic offspring. To account for potential background effects, we examined three pairs of RILs that carried either a sterile (B6) or fertile (B4) QTL haplotype across the QTL region (dm6 2L:19,010,000-2R:7,272,495) in otherwise similar genetic backgrounds (shared 44–47% of founder alleles outside the QTL). Please note that these lines differ from the B6 and B8 isogenic stocks we utilize in Fig 1 and later in the manuscript, which are more closely matched for genetic background. Principal component analysis (PCA) of read counts reveals two independent axes that resolve sterile and fertile gene expression profiles, which together account for 40% and 16% of variation (Fig 2A and S14 Table). One biological replicate of RIL 21188 (fertile) was an outlier, which we excluded from our downstream analysis of differentially expressed genes. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Fertility is associated with increased chromatin modification, whereas sterility is associated with increased expression of replication-dependent histones. a) PCA analysis of gene expression data for pairs of sterile (B6) and fertile (B4) RILs, that carry founder B6 and B4 haplotypes across the QTL window. Members of the same RIL pair with otherwise similar genetic backgrounds are represented by the same shape. b) GO terms enriched among genes upregulated in fertile and sterile genotypes. c) Log2 fold differences in expression for chromatin modifiers between sterile and fertile genotypes. d) Log2 fold increase in RD histone expression in sterile genotypes. e). Probability density plot of log2 fold change values for all euchromatic (blue), pericentromeric (red), telomeric (green) genes and 4th chromosome (gray) between strains carrying sterile and fertile. The mean of each distribution is represented by a dotted line, and is compared between distributions with a two-sample t-test. The x-axis boundaries were confined from (-1.5 to 2) for a better visualization. The pericentromere-euchromatin boundaries were drawn from [29,45] and subtelomeric-euchromatin boundary coordinates from [46–48]. The data represented in panel a is provided in S14 Table and plot in panels c, d, and e in S5 Table). *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g002 We found a total of 530 genes differentially expressed between sterile and fertile genotypes (Benjamini-Hochberg adjusted p-value < = 0.05, fold-change > 1.5; S5 Table). The most significantly enriched gene ontology (GO) term among genes upregulated in fertile ovaries is chorion assembly (Bonferroni corrected P value <0.01, Fig 2B and S7 Table). This suggests a larger number of late-stage oocytes in fertile ovaries, as other genes that are upregulated in late oogenesis (stages 12–14) show a similar increase in expression (S6 Fig, [30]). Because atrophy results from the loss of larval PGCs and pre-meiotic adult cysts (GSCs), larger numbers of late-stage oocytes are likely unrelated to dysgenic sterility [15–17,19]. The second most significant GO term upregulated in the ovaries of fertile genotypes is covalent chromatin modification (Fig 2B). Strikingly, we discovered fertile ovaries exhibit a systematic upregulation of multiple chromatin modification complexes with key roles in oogenesis, including polycomb group, trithorax group, and the TIP60 complex, as well as many individual SET domain lysine methyltransferases (Fig 2C) [31–33]. The TIP60 complex in particular is involved in cell cycle progression and differentiation in pre-cystoblasts [32]: daughter cells of germline stem cells in which P-elements transpose [19,21]. Interestingly, the TIP60 complex is also involved in DSB repair [34], which could promote dysgenic germ cell survival. Similarly, polycomb-dependent gene silencing initiates in nurse cells concurrently with meiosis I: a window in which germ cell cysts with large numbers of DSBs undergo apoptosis [35]. Genes upregulated in the sterile genotypes are enriched for functions in chromatin assembly and transcription, cell division, and translation. However, a careful inspection of genes underlying these enriched terms reveals that, with the exception of translation, they are primarily explained by the increased expression of replication-dependent (RD) histone gene copies (Fig 2D). While these expression increases are modest (<2 fold), they are likely an underestimate of the true degree of histone upregulation. Histone gene expression increases dramatically in late-stage oocytes (beyond stage 10 [36]), which appear to be reduced in the ovaries of sterile genotypes (S6 Fig). Overexpression of RD histones is associated with increased sensitivity to DNA damage [37–41], and excess histones are reported to compete with DNA repair proteins for binding to damage sites [38]. Thus, while TIP60 activity might increase DSB repair in fertile ovaries promoting germ cell survival, histones might decrease repair in sterile ovaries increasing germ cell death. Differences in ovarian chromatin modification between fertile and sterile genotypes may further be connected to histone regulation. Replication-dependent histones occur in a tandemly duplicated gene cluster that exhibits coordinated and dosage-compensated regulation in a specialized nuclear compartment known as the histone locus body (HLB, [42]). In particular, negative regulation of histone expression relies on multiple heterochromatin factors [41,43]. Consistent with reduced heterochromatin formation in the ovaries of sterile genotypes, they show higher expression of pericentromeric genes (two-sample t-test, t141 = -9.32, p-value = 2.3x 10–16), as well as genes on the heterochromatic 4th chromosome (two-sample t-test, t53 = -4.56, p-value = 3.0x10-5, Fig 2E). The sterile B6 haplotype also exhibits increased expression of pericentromeric genes in a previously published microarray dataset from head tissue ([44] S1 Fig). Sterile genotypes exhibit silenced piRNA loci, but no systematic TE dysregulation In addition to gene expression, differences in the regulation of resident TEs could modulate the degree of dysgenic sterility. The D. melanogaster genome harbors >100 resident TE families [49,50], many of which are transpositionally active and show variable transposition rates in wild-type strains [51–53]. If sterile alleles establish weaker regulation of some resident TEs, their transposition could add to DNA damage resulting from P-element transposition, thereby promoting germ cell loss. Resident TEs are regulated by piRNAs, and two features of our data suggest differences in piRNA biogenesis between sterile and fertile alleles. First, QTL-3d contains numerous piRNA clusters, including the major ovarian piRNA cluster 42AB, which could differ in the regulation or resident TEs between sterile and fertile alleles (Fig 1D). Second, differences in chromatin regulation between sterile and fertile alleles could impact piRNA cluster expression (Fig 2E and 2F), which is dependent upon the heterochromatic histone modification, histone 3 lysine 9 trimethylation (H3K9me3) [54,55]. To look for differences in resident TE regulation, we performed small RNA-seq on the same ovarian samples from RIL females (mothers) that we used for total RNA-seq, and quantified the expression of piRNAs from clusters throughout the genome. A PCA of piRNA cluster expression reveals that sterile and fertile genotypes differ in the ovarian expression of some piRNA clusters, and are resolved by the second principal component, accounting for 22% variation in expression (Fig 3A and S15 Table). In particular, we discovered two small pericentromeric piRNA clusters located within QTL-3d that were active in fertile genotype ovaries but largely quiescent in sterile genotype ovaries (Figs 3B, 3C, 3D, S2, and S3 and S16 Table). However, the major piRNA clusters—including 42AB—do not differ in ovarian expression in between sterile and fertile alleles, suggesting that the proposed reduction in heterochromatin formation in sterile genotypes does not majorly reduce piRNA transcription (Fig 3B and S8 Table). Furthermore, the differentially active clusters in QTL-3d seem unlikely to regulate transpositionally active resident TE families, since they are largely composed of TE fragments that are relatively divergent from the consensus (65 to 95% sequence similarity; Figs 3C, 3D, S2, and S3 and S9 Table), or are most similar to a consensus TE from other (non-melanogaster) Drosophila species. Transpositionally active TEs are generally highly similar to the consensus sequence [56], and piRNA silencing is disrupted by mismatches between the piRNA and its target [57]. Nevertheless, we cannot rule out the possibility that sterile or fertile genotypes could harbor TE insertions in this locus that are not represented in the dm6 reference genome. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Dysgenic sterility is not related to differential activity of piRNA clusters or TE deregulation. a) PCA analysis for piRNA cluster expression data of sterile (S) and fertile (F) genotypes. Members of the same RIL pair are represented by the same shapes. b) Heat map showing the expression of seven major piRNA clusters [61] and the two differentially expressed QTL clusters in QTL-3d. RIL pairs are plotted adjacent to each other. c and d) Uniquely mapping piRNAs within two differentially active QTL-3d piRNA clusters are compared between sterile (21183) and fertile (21213) genotypes. Positive value indicates piRNAs mapped to the sense strand of the reference genome and negative value indicates those from the antisense strand. TE insertions in each cluster are presented according to family by different colors; TE-others indicate the insertion was most similar to a consensus TE from a sibling species of D. melanogaster. See S2 and S3 Figs for cluster expression in the remaining RIL pairs. For b, c and d, piRNA cluster expression levels are estimated by log2 scale transformed of reads per million mapped reads [log2(RPM+1)]. e) Genome-wide differences in TE family expression between sterile and fertile genotypes (fold change = 1.5, base mean > = 100, adjusted p-value < = 0.05), based on alignment to consensus sequences. The data used to plot panel a is provided in S15 Table, for panel b in S8 Table, for panels c and d in S16 and S9 Tables, and for panel e in S10 Table). https://doi.org/10.1371/journal.pgen.1010080.g003 To directly address if fertile and sterile genotypes differ in resident TE regulation, we compared their genome-wide resident TE expression in our RNA-seq data. None of the TE families represented in the QTL-3d piRNA clusters were upregulated in sterile genotypes (Fig 3E and S10 Table). Furthermore, while some TE families are differentially expressed, there is no systematic increase in TE activity in the sterile genotypes. Rather, more TE families are upregulated in the ovaries of fertile genotypes (13 TEs) when compared to sterile (4 TEs) genotypes. Upregulation of certain TEs may be consistent with more late-stage egg-chambers in fertile genotype ovaries, since many retrotransposons accumulate transcripts in the oocyte over the course of oogenesis [58–60]. Therefore, despite the conspicuous position of QTL-3d surrounding piRNA producing-regions, as well as evidence for differential chromatin regulation that could impact piRNA biogenesis (Fig 3B and 3E), we find no evidence that fertility in dysgenic crosses is determined by resident TE silencing. Sterile alleles increase P-element mRNA expression and transposase mRNA splicing Increased dysgenic sterility associated with sterile alleles could also reflect increased P-element transposition, resulting from increased P-element mRNA expression or splicing. In particular only transcripts in which the third intron (intervening sequence, IVS3) is spliced will produce P-transposase, and regulation of IVS3 splicing is a key determinant of P-element transposition [22]. We therefore examined the abundance of different P-element transcripts in the ovaries of F1 dysgenic offspring of sterile (B6) and fertile (B8) isogenic females. Dysgenic offspring in these experiments were reared at 22°C to avoid germline loss [20]. Consistent with differential production of P-transposase, we observed differences in overall abundance of P-element transcripts between the F1 dysgenic offspring of sterile and fertile females (Fig 4A, t10 = 13.09, p = 2.31x10-15). On average, sterile females showed a 34% increase in P-element transcripts (95% CI 27–42%). Transposase-encoding (IVS3 spliced) transcripts show an even more pronounced 59% increase in expression in sterile genotypes (Fig 4A, t10 = 10.27, p = 2.91x10-10, 95% CI: 42–81%). By contrast, unspliced (IVS3 retaining) transcripts were not significantly differentially expressed between ovaries of F1 dysgenic offspring of sterile (B6) and fertile (B8) isogenic females (Fig 4A, t10 = 1.68, p = 0.11), although certain individual comparisons between strains were significant. To directly address whether splicing is more efficient in the ovaries of sterile dysgenic offspring, we compared the ratio of spliced to unspliced P-element transcripts (Fig 4B). The ratio of spliced to unspliced transcripts differed significantly between the ovaries sterile and fertile dysgenic offspring (t10 = 7.45, p = 7.30x10-5), suggesting that splicing itself is more efficient in sterile genotypes. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Decreased expression of P-transposase in fertile genotypes. A) Differential expression of P-element transcripts between fertile and sterile genotypes. Three separate qPCRs were performed, which detect all transcript isoforms, as well as IVS3 spliced and unspliced isoforms. qPCRs are normalized to rpl32. B) Ratios of IVS3 spliced to unspliced isoforms. C) Differential expression of splicing factors between sterile and fertile genotypes based on RNA-seq data. Dark green bars indicate factors that are significantly upregulated in fertile genetic backgrounds. Significant differences in qPCR data are based on linear models to detect differences between sterile and fertile genotypes, or Tukey-HSD comparisons to detect differences between genotypes containing the same allele. *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g004 We also observed differences in splicing and expression between isogenic stocks carrying the same allele. In particular, the Fertile_B81 and Fertile_B82 differed in both spliced and unspliced transcripts, with a particularly pronounced 2.89-fold increase in spliced transcript expression in Fertile_B82 as compared with Fertile_B81 (Fig 4A, t10 = 12.14, p = 9.78x10-12, 95% CI: 2.61–3.17 fold). While the sample size is too small to draw any conclusions, it is notable that Fertile_B82 does not exhibit the same degree of fertility rescue as Fertile_B81, further pointing to a connection between spliced transcript production and dysgenic sterility (Fig 1G and 1H). In germline cells, the splicing of IVS3 is known to be repressed by piRNA mediated transcriptional silencing, which is initiated by maternally transmitted piRNAs or through the production of de novo piRNAs in aged females [16,21]. In addition to the absence of maternally transmitted piRNAs, the splicing differences we observe here are in young (3–4 day old) dysgenic females, as opposed to aged dysgenic females, suggesting they are likely independent of the piRNA pathway. However, the splicing of IVS3 is also repressed by several host splicing factors in somatic cells, and it is proposed that some of these factors may also partially repress splicing in germline cells [62–64]. Consistent with piRNA-independent differences in splicing, we discovered that three splicing factors known to promote IVS3 retention in somatic cells, hrp36, hrp38 and P-element somatic inhibitor (Psi) show increased expression in fertile genotypes in our ovarian RNA seq data (Fig 4C). This suggests that decreased splicing in the dysgenic offspring of fertile isogenic lines may result from increased abundance of host splicing factors. Sterile alleles increase radiation sensitivity and accumulated mutations Our gene expression data suggest that sterile and fertile alleles may differ in their capacity to repair germline DSBs in young (3 day) dysgenic females. Fertile alleles exhibit upregulation of the TIP60 complex (which is involved in DSB repair [34]), while sterile alleles exhibit upregulation of replication dependent histones (which may complete with DNA repair machinery [38]). Mutations in DSB repair genes are widely known to cause radiation sensitivity, which is easily quantified by measuring lethality following larval radiation exposure [65–69]. While this assay occurs in whole larvae as compared to female germlines, larvae are composed of numerous classes of mitotically dividing cells, similar to the primordial and premeiotic stages of gamete production in which P-element transposition occurs [16,21]. Furthermore, numerous key factors for germline DNA damage response, as well as germline P-element excision repair, exhibit larval radiation sensitivity phenotypes [70–75]. We therefore compared the X-ray radiation sensitivity of larvae from isogenic lines containing sterile (B6) and fertile (B8) alleles. Note these lines are the same as those in Fig 1F–1H). After exploring a range of radiation doses, we found that doses above 10 Gy showed high lethality, making it difficult to detect differences in radiation sensitivity between the genotypes (S19 Table). Therefore, we compared the response of sterile and fertile larvae to radiation doses of 0 Gy, 5 Gy and 10 Gy. We observed that fertile genotypes had significantly higher survival (53–58%) than the sterile genotypes (25–30%) at 10 Gy (Fig 5A). Given that X-ray radiation produces predominantly DSBs, these results are consistent with differences between fertile and sterile alleles in DSB repair. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Sterile alleles exhibit reduced DNA repair. A) The percentage of mock treated and irradiated (5 Gy and 10 Gy) larvae that survived to adulthood for the fertile (B8), sterile (B6) and the control genotypes. CS refers to Canton-S and marker refers to the multiply marked stock b cn (#44229), which was used to generate isogenic lines. The numbers in the brackets refer to the sample size. The significance of comparisons between genotypes was determined by the χ2 test-of-independence. B and C) New mutations that accumulate in RIL genomes as detected by MSG. B) New SNPs and C) new indels. An excess of new mutations was detected by a t-test comparing Sterile B6 RILs to all others. The data represented in the Fig is provided in S19 and S20 Tables) ** denotes P< 0.01, *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g005 We further looked specifically for differences in germline DSB repair by examining whether RILs carrying B6 alleles at the QTL-3d peak have accumulated more de novo base substitutions and small insertions or deletions. The DSPR RILs underwent 50 generations of inbreeding, and have since been maintained as isogenic lab stocks for ~175 generations, allowing ample time for new mutations to accrue as a consequence of deficient repair. To detect these new mutations, we generated multiplexed shotgun genotyping (MSG) libraries for 792 population B RILs [76]. This low coverage method (mean 2.9x) will uncover only a random subset of new mutations in each RIL, thereby underestimating the true amount of mutation accumulation. Nevertheless, we were able to detect 102,476 novel base substitutions and 5,026 novel insertions or deletions among the RIL MSG libraries. After accounting for differences in sequencing depth and plate effects, the founder allele at QTL-3d was associated with differences in the number of new base substitutions (-2ΔlnL = 15.62, df = 6, p = 0.016). Furthermore, RILs carrying B6 alleles at the QTL-3d peak exhibit an average increase of 14.58 new base substitutions (95% CI 5.23–24.12), when compared to those carrying another founder allele (t782 = 3.043, P = 0.0024, Fig 5B). In contrast, there was no significant association between founder allele at QTL-3d and new indels (-2ΔlnL = 1.37, df = 6, p = 0.97). Given the low coverage data as well the limited potential of short-read sequencing data to identify larger structural variation [77], we cannot be conclusive about a relationship between QTL allele and indel accumulation rate. Nevertheless, the increase in base substitution supports a deficiency in germline DNA repair in association with B6 alleles for QTL-3d. Identifying candidate genes The QTL we map here are quite large and contain numerous candidate genes whose differential function could influence dysgenic sterility. Nevertheless, we next sought to identify candidate genes that influence dysgenic sterility for future study. We combined our own expression and mapping data with previously published polymorphism and single cell expression data to narrow candidates based on four criteria: 1) location within a QTL, 2) expression in primordial germ cells or early, pre-meiotic cysts [78,79], 3) differential expression between sterile and fertile adult ovaries, and 4) the presence of “in-phase” single nucleotide polymorphisms (SNPs) (S11, S12 and S13 Tables). In-phase SNPs are those where the genotypic differences between the founder alleles are consistent with their phenotype class (Figs 1E and 6A [80]). Of 530 differentially expressed genes, 43 are within the QTL region, representing an approximately five-fold enrichment in the QTL regions compared to the rest of the genome (χ2 = 255.54, df = 1, p-value < 2.2e-16, Fig 6B). Ultimately, we identified 12 and 5 differentially expressed genes and early germ cell expressed genes that also carry in-phase SNPs within the QTL-3d and 21d, respectively (Fig 6C and 6D and S12 Table). Furthermore, we identified 32 genes in QTL-3d and 3 genes in QTL-21d that exhibit early germ cell expression and also contain in-phase non-synonymous SNPs, which may affect the function of the encoded protein (S13 Table). Collectively these genes represent the strongest candidates to contain causative variants. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Differential expression and in-phase SNPs identify candidate genes. a) The proportion of genes differentially expressed (DEG) is compared inside and outside the QTL. The dotted line is the genome wide average. b) Hypothetical in-phase and out of phase SNPs are shown. Sequences of each of the B founder strains are colored based on their phenotypic classification, either fertile or sterile (Fig 1E). Bold letters indicate SNPs. c and d) Venn diagrams showing the overlap of differentially expressed genes (DEG), genes carrying in-phase synonymous and non-synonymous SNPs, and genes expressed in primordial or pre-meiotic germ cells for QTL-21d (c) and QTL-3d (d).The data for differential expression of genes for fertile and sterile genotypes is provided in S5 Table. The data on in-phase polymorphisms for each QTL peak are provided in S11 Table. List of candidate genes that have both in-phase polymorphisms and are differentially expressed, and those having non-synonymous in-phase polymorphisms are provided in S12 and S13 Tables, respectively. https://doi.org/10.1371/journal.pgen.1010080.g006 We next scoured our list of candidate genes for those with known functions in chromatin regulation, DSB repair, or alternative splicing, whose differential function or regulation are plausibly related to the phenotypic differences associated with sterile and fertile alleles. None of the three splicing factors we discovered are differentially expressed reside within the QTL (Fig 4C), suggesting their expression differences arise as a consequence of regulatory differences in trans. While we did not discover any transcription factors located in the QTL that are differentially expressed in fertile and sterile ovaries, we did discover three C2H2 zinc finger transcription factors, tio and CG30431 (QTL-3d) and CG17568 (QTL-21d), that are located within QTL and contain in-phase non-synonymous SNPs. Unfortunately, the genomic binding sites of these transcription factors are undetermined, so it remains unknown if they are regulators of hrp36, hrp38, or psi transcription. With respect to differences in chromatin state and/or DNA repair, two genes within QTL-3d, stand out as particularly attractive candidates; Nipped-A and jing. Nipped-A contains a non-synonymous in-phase SNP and is expressed in both PGCs and in germline cells throughout the earliest stages of oogenesis (S13 Table). Nipped-A is a member of the TIP60 complex, which has functions in DSB repair, chromatin modification and chromatin remodeling. Additionally, we identified multiple TIP60 components upregulated in fertile ovaries (Fig 2C). The non-synonymous SNP that separates sterile and fertile alleles of this gene are located in the HEAT2 domain, which is predicted to be essential for protein-protein interaction, and could have important implications for the function of this multi-protein complex [81–83]. Jing contains in-phase synonymous and non-synonymous SNPs, is upregulated in fertile ovaries, and exhibits a similar expression pattern in germline cells to that of Nipped-A (Fig 2E and S12 and S13 Tables). Based on a yeast-two hybrid screen Jing physically interacts with inverted repeat binding protein 18 (IRBP18): a DNA binding protein that comprises part of a heterodimer that binds directly to P-element’s transcribed inverted repeats, and facilitates repair of donor DNA after excision [84,85]. Furthermore, irbp18 mutants exhibit larval radiation sensitivity, similar to our sterile genotypes (Fig 5A). If Jing determines differential activity of IRBP18 it could have a strong impact on dysgenic sterility. Beyond this function, Jing acts as an important cofactor of polycomb repressor complex 2, many of which showed increased expression in fertile ovaries Fig 2C, [86,87]. QTL mapping The DSPR RILs are all P-element free M-strains, which were derived from founders isolated from natural populations before the P-element invasion [24]. We therefore screened for alleles that influence dysgenic sterility among the panel B RIL genomes by crossing RIL females to males from the reference P-strain Harwich, and examining the morphology of F1 ovaries (Fig 1A). Atrophied ovaries are indicative of germline loss resulting from P-element activity [14,25]. Since dysgenic sterility changes across development [15], and some females exhibit age-dependent recovery from P-element hybrid dysgenesis through the production of de novo piRNAs [20], we phenotyped F1 females at two developmental time points: 3 days and 21 days post-eclosion. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. QTL mapping of variation in P-element induced ovarian atrophy. a) Crossing scheme to phenotype the variation in F1 ovarian atrophy among RIL offspring. Representative images of atrophied and non-atrophied ovaries are from Kelleher et al. [14] b) The log of odds (LOD) plot for QTL mapping of ovarian atrophy using 3 day-old (gold) and 21 day-old (blue) F1 females. The dotted line is the LOD threshold and x-axis represents the chromosomal positions. c) Zoomed-in figure of QTL mapping from 3 days (gold) and 21 days (blue). The colored boxes show the genomic interval that likely contains the causative genetic variant of each QTL, based on a Δ2LOD drop from the peak position [26]. The pairs of dotted lines indicate the peak Δ2LOD scores that determines the interval. The solid horizontal line is the LOD significance threshold based on 1,000 permutations of the phenotype data. d) Cytological map depicting the interval of the two QTL peaks [27,28]. e) Residual F1 atrophy (y-axis) associated with each of the eight founder alleles (x-axis) at the QTL peaks after accounting for random effects. All the QTL peaks show 2 phenotypic classes: sterile (light green) and fertile (dark green). (f-g) Percentage of (f) ovarian atrophy and (g) sterility among dysgenic female offspring from crosses between Harwich males and isogenic females carrying sterile (B6) and fertile (B8) alleles. Proportions were compared between samples using χ2 tests of independence. h) Number of F2 offspring produced by individual dysgenic F1 females from crosses between Harwich males and isogenic fertile and sterile females. The horizontal line indicates the mean, which was compared between samples using permutation tests. Superscripts1,2 and 3 in (f-h) denote isogenic lines that were independently generated, these sometimes differ between experiments because sterile_B61 became contaminated and was replaced with sterile_B63. Error bars in e, f and g represent the standard error. The data used to generate plot in panels b,c, and e are provided in S3 and S4 Tables and that used for plot in panels f, g and h are provided in S17 and S18 Tables respectively. * denotes P < 0.05, ** denotes P < 0.01, *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g001 Similar to our observations with the Population A RILs [14], we found continuous variation in the frequency of ovarian atrophy among dysgenic offspring of different RIL mothers, indicating genetic variation in dysgenic sterility that is unrelated to maternally deposited piRNAs (S1 and S2 Tables). Based on a combined linear model of F1 atrophy among 3 and 21 day-old females, we estimated the broad-sense heritability of dysgenic ovarian atrophy in our experiment to be ~42.5%. The effect of age on the proportion of F1 atrophy was significant but minimal (χ2 = 7.03, df = 1, p-value = 0.008) with 21 day-old females showing only 0.7% increase in atrophy as compared to 3 day-old females. This suggests that age-dependent recovery from dysgenic ovarian atrophy through the production of de novo piRNAs is not common among the genotypes we sampled. To identify the genomic regions associated with genetic variation in dysgenic ovarian atrophy, we performed QTL mapping using the published RIL genotypes [24]. We found a large QTL peak near the 2nd chromosome centromere in both 3 and 21 day-old F1 females (Fig 1B and Tables 1, S3, and S4). However, the genomic intervals of the two QTL are non-overlapping (Fig 1C and Table 1). The major QTL in 21 day-old females (hereafter, QTL-21d) resides in the euchromatic region and is quite small (990 kb) compared to the major QTL in 3 day-old females (hereafter QTL-3d), which spans the centromere and pericentromeric regions (9.6 Mb, Fig 1D). Therefore, there are likely at least two polymorphisms that influence tolerance near the 2nd chromosome centromere, one of which has a larger effect in young 3-day old females, with the other having a larger effect in 21 day-old females. Download: PPT PowerPoint slide PNG larger image TIFF original image Table 1. QTL positions in 3 and 21-day old females. The peak position, Δ2LOD drop confidence interval (2LOD CI), and the Bayesian Credible Interval (BCI) in dm6 [29] are provided for each analysis. The data used to identify the LOD peaks and intervals for 3 and 21-day old females can be found in S3 and S4 Tables, respectively. https://doi.org/10.1371/journal.pgen.1010080.t001 We further evaluated the effect of the two linked QTL through haplotype analysis. We modeled residual F1 ovarian atrophy as a function of QTL haplotype for the 3 day and 21 day peaks, thereby disentangling synergistic (e.g. sterile 3d, sterile 21d) from opposing (e.g. sterile 3d, fertile 21d) allelic combinations (S4 Fig). We observed that the 3 day-old QTL allele is solely-determinant of ovarian atrophy in the 3 day-old offspring. However, in 21 day-old offspring only the genotypes containing fertile alleles at both QTL show decreased atrophy. This suggests that QTL-3d may determine germ cell maintenance in the larval, pupal and early adult stages, but QTL-21d may be additionally required to maintain germline cells in aging females. The presence of two QTL is further supported by the phenotypic classes we detected among founder alleles (B1-B8) for each of the QTL peaks (Fig 1E). For QTL-21d, both B2 and B6 founder alleles are associated with greatly increased dysgenic ovarian atrophy. By contrast for QTL-3d, only the B6 founder allele is associated with increased ovarian atrophy. We next sought to determine whether reduced ovarian atrophy corresponds to restored fertility, or merely allows for the production of inviable gametes. To this end, we generated isogenic lines that carry either high-atrophy (B6) or low-atrophy (B8) alleles at both QTL loci in an otherwise identical genetic background through 6 generations of backcrossing to a marker stock (S5 Fig). Consistent with our QTL mapping, B8 alleles display less F1 ovarian atrophy (24–31%) than B6 strains when crossed with Harwich males (Fig 1F and S17 Table). Furthermore, while B6 dysgenic females produced no offspring, 13–29% of dysgenic B8 females were fertile and produced offspring (Fig 1G and S18 Table). For one B8 stock, offspring counts were significantly higher when compared to B6 (Fig 1H). In light of these observations, we refer to the low-atrophy and high-atrophy alleles hereafter as “fertile” and “sterile”. The fertility rescue conferred by these alleles would be highly beneficial in populations where dysgenic crosses are common. Sterile and fertile alleles differ in chromatin regulation Both the QTL regions contain large numbers of protein coding and non-coding RNA genes, piRNA clusters, and repeats, which could influence dysgenic sterility (Fig 1D). To better understand the differences between fertile and sterile genotypes, we compared their gene expression profiles in the ovaries of young 3–5 day-old females by stranded total RNA-seq. To avoid the confounding effects of germline loss under dysgenic conditions, we focused on RIL females rather than their dysgenic offspring. To account for potential background effects, we examined three pairs of RILs that carried either a sterile (B6) or fertile (B4) QTL haplotype across the QTL region (dm6 2L:19,010,000-2R:7,272,495) in otherwise similar genetic backgrounds (shared 44–47% of founder alleles outside the QTL). Please note that these lines differ from the B6 and B8 isogenic stocks we utilize in Fig 1 and later in the manuscript, which are more closely matched for genetic background. Principal component analysis (PCA) of read counts reveals two independent axes that resolve sterile and fertile gene expression profiles, which together account for 40% and 16% of variation (Fig 2A and S14 Table). One biological replicate of RIL 21188 (fertile) was an outlier, which we excluded from our downstream analysis of differentially expressed genes. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Fertility is associated with increased chromatin modification, whereas sterility is associated with increased expression of replication-dependent histones. a) PCA analysis of gene expression data for pairs of sterile (B6) and fertile (B4) RILs, that carry founder B6 and B4 haplotypes across the QTL window. Members of the same RIL pair with otherwise similar genetic backgrounds are represented by the same shape. b) GO terms enriched among genes upregulated in fertile and sterile genotypes. c) Log2 fold differences in expression for chromatin modifiers between sterile and fertile genotypes. d) Log2 fold increase in RD histone expression in sterile genotypes. e). Probability density plot of log2 fold change values for all euchromatic (blue), pericentromeric (red), telomeric (green) genes and 4th chromosome (gray) between strains carrying sterile and fertile. The mean of each distribution is represented by a dotted line, and is compared between distributions with a two-sample t-test. The x-axis boundaries were confined from (-1.5 to 2) for a better visualization. The pericentromere-euchromatin boundaries were drawn from [29,45] and subtelomeric-euchromatin boundary coordinates from [46–48]. The data represented in panel a is provided in S14 Table and plot in panels c, d, and e in S5 Table). *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g002 We found a total of 530 genes differentially expressed between sterile and fertile genotypes (Benjamini-Hochberg adjusted p-value < = 0.05, fold-change > 1.5; S5 Table). The most significantly enriched gene ontology (GO) term among genes upregulated in fertile ovaries is chorion assembly (Bonferroni corrected P value <0.01, Fig 2B and S7 Table). This suggests a larger number of late-stage oocytes in fertile ovaries, as other genes that are upregulated in late oogenesis (stages 12–14) show a similar increase in expression (S6 Fig, [30]). Because atrophy results from the loss of larval PGCs and pre-meiotic adult cysts (GSCs), larger numbers of late-stage oocytes are likely unrelated to dysgenic sterility [15–17,19]. The second most significant GO term upregulated in the ovaries of fertile genotypes is covalent chromatin modification (Fig 2B). Strikingly, we discovered fertile ovaries exhibit a systematic upregulation of multiple chromatin modification complexes with key roles in oogenesis, including polycomb group, trithorax group, and the TIP60 complex, as well as many individual SET domain lysine methyltransferases (Fig 2C) [31–33]. The TIP60 complex in particular is involved in cell cycle progression and differentiation in pre-cystoblasts [32]: daughter cells of germline stem cells in which P-elements transpose [19,21]. Interestingly, the TIP60 complex is also involved in DSB repair [34], which could promote dysgenic germ cell survival. Similarly, polycomb-dependent gene silencing initiates in nurse cells concurrently with meiosis I: a window in which germ cell cysts with large numbers of DSBs undergo apoptosis [35]. Genes upregulated in the sterile genotypes are enriched for functions in chromatin assembly and transcription, cell division, and translation. However, a careful inspection of genes underlying these enriched terms reveals that, with the exception of translation, they are primarily explained by the increased expression of replication-dependent (RD) histone gene copies (Fig 2D). While these expression increases are modest (<2 fold), they are likely an underestimate of the true degree of histone upregulation. Histone gene expression increases dramatically in late-stage oocytes (beyond stage 10 [36]), which appear to be reduced in the ovaries of sterile genotypes (S6 Fig). Overexpression of RD histones is associated with increased sensitivity to DNA damage [37–41], and excess histones are reported to compete with DNA repair proteins for binding to damage sites [38]. Thus, while TIP60 activity might increase DSB repair in fertile ovaries promoting germ cell survival, histones might decrease repair in sterile ovaries increasing germ cell death. Differences in ovarian chromatin modification between fertile and sterile genotypes may further be connected to histone regulation. Replication-dependent histones occur in a tandemly duplicated gene cluster that exhibits coordinated and dosage-compensated regulation in a specialized nuclear compartment known as the histone locus body (HLB, [42]). In particular, negative regulation of histone expression relies on multiple heterochromatin factors [41,43]. Consistent with reduced heterochromatin formation in the ovaries of sterile genotypes, they show higher expression of pericentromeric genes (two-sample t-test, t141 = -9.32, p-value = 2.3x 10–16), as well as genes on the heterochromatic 4th chromosome (two-sample t-test, t53 = -4.56, p-value = 3.0x10-5, Fig 2E). The sterile B6 haplotype also exhibits increased expression of pericentromeric genes in a previously published microarray dataset from head tissue ([44] S1 Fig). Sterile genotypes exhibit silenced piRNA loci, but no systematic TE dysregulation In addition to gene expression, differences in the regulation of resident TEs could modulate the degree of dysgenic sterility. The D. melanogaster genome harbors >100 resident TE families [49,50], many of which are transpositionally active and show variable transposition rates in wild-type strains [51–53]. If sterile alleles establish weaker regulation of some resident TEs, their transposition could add to DNA damage resulting from P-element transposition, thereby promoting germ cell loss. Resident TEs are regulated by piRNAs, and two features of our data suggest differences in piRNA biogenesis between sterile and fertile alleles. First, QTL-3d contains numerous piRNA clusters, including the major ovarian piRNA cluster 42AB, which could differ in the regulation or resident TEs between sterile and fertile alleles (Fig 1D). Second, differences in chromatin regulation between sterile and fertile alleles could impact piRNA cluster expression (Fig 2E and 2F), which is dependent upon the heterochromatic histone modification, histone 3 lysine 9 trimethylation (H3K9me3) [54,55]. To look for differences in resident TE regulation, we performed small RNA-seq on the same ovarian samples from RIL females (mothers) that we used for total RNA-seq, and quantified the expression of piRNAs from clusters throughout the genome. A PCA of piRNA cluster expression reveals that sterile and fertile genotypes differ in the ovarian expression of some piRNA clusters, and are resolved by the second principal component, accounting for 22% variation in expression (Fig 3A and S15 Table). In particular, we discovered two small pericentromeric piRNA clusters located within QTL-3d that were active in fertile genotype ovaries but largely quiescent in sterile genotype ovaries (Figs 3B, 3C, 3D, S2, and S3 and S16 Table). However, the major piRNA clusters—including 42AB—do not differ in ovarian expression in between sterile and fertile alleles, suggesting that the proposed reduction in heterochromatin formation in sterile genotypes does not majorly reduce piRNA transcription (Fig 3B and S8 Table). Furthermore, the differentially active clusters in QTL-3d seem unlikely to regulate transpositionally active resident TE families, since they are largely composed of TE fragments that are relatively divergent from the consensus (65 to 95% sequence similarity; Figs 3C, 3D, S2, and S3 and S9 Table), or are most similar to a consensus TE from other (non-melanogaster) Drosophila species. Transpositionally active TEs are generally highly similar to the consensus sequence [56], and piRNA silencing is disrupted by mismatches between the piRNA and its target [57]. Nevertheless, we cannot rule out the possibility that sterile or fertile genotypes could harbor TE insertions in this locus that are not represented in the dm6 reference genome. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. Dysgenic sterility is not related to differential activity of piRNA clusters or TE deregulation. a) PCA analysis for piRNA cluster expression data of sterile (S) and fertile (F) genotypes. Members of the same RIL pair are represented by the same shapes. b) Heat map showing the expression of seven major piRNA clusters [61] and the two differentially expressed QTL clusters in QTL-3d. RIL pairs are plotted adjacent to each other. c and d) Uniquely mapping piRNAs within two differentially active QTL-3d piRNA clusters are compared between sterile (21183) and fertile (21213) genotypes. Positive value indicates piRNAs mapped to the sense strand of the reference genome and negative value indicates those from the antisense strand. TE insertions in each cluster are presented according to family by different colors; TE-others indicate the insertion was most similar to a consensus TE from a sibling species of D. melanogaster. See S2 and S3 Figs for cluster expression in the remaining RIL pairs. For b, c and d, piRNA cluster expression levels are estimated by log2 scale transformed of reads per million mapped reads [log2(RPM+1)]. e) Genome-wide differences in TE family expression between sterile and fertile genotypes (fold change = 1.5, base mean > = 100, adjusted p-value < = 0.05), based on alignment to consensus sequences. The data used to plot panel a is provided in S15 Table, for panel b in S8 Table, for panels c and d in S16 and S9 Tables, and for panel e in S10 Table). https://doi.org/10.1371/journal.pgen.1010080.g003 To directly address if fertile and sterile genotypes differ in resident TE regulation, we compared their genome-wide resident TE expression in our RNA-seq data. None of the TE families represented in the QTL-3d piRNA clusters were upregulated in sterile genotypes (Fig 3E and S10 Table). Furthermore, while some TE families are differentially expressed, there is no systematic increase in TE activity in the sterile genotypes. Rather, more TE families are upregulated in the ovaries of fertile genotypes (13 TEs) when compared to sterile (4 TEs) genotypes. Upregulation of certain TEs may be consistent with more late-stage egg-chambers in fertile genotype ovaries, since many retrotransposons accumulate transcripts in the oocyte over the course of oogenesis [58–60]. Therefore, despite the conspicuous position of QTL-3d surrounding piRNA producing-regions, as well as evidence for differential chromatin regulation that could impact piRNA biogenesis (Fig 3B and 3E), we find no evidence that fertility in dysgenic crosses is determined by resident TE silencing. Sterile alleles increase P-element mRNA expression and transposase mRNA splicing Increased dysgenic sterility associated with sterile alleles could also reflect increased P-element transposition, resulting from increased P-element mRNA expression or splicing. In particular only transcripts in which the third intron (intervening sequence, IVS3) is spliced will produce P-transposase, and regulation of IVS3 splicing is a key determinant of P-element transposition [22]. We therefore examined the abundance of different P-element transcripts in the ovaries of F1 dysgenic offspring of sterile (B6) and fertile (B8) isogenic females. Dysgenic offspring in these experiments were reared at 22°C to avoid germline loss [20]. Consistent with differential production of P-transposase, we observed differences in overall abundance of P-element transcripts between the F1 dysgenic offspring of sterile and fertile females (Fig 4A, t10 = 13.09, p = 2.31x10-15). On average, sterile females showed a 34% increase in P-element transcripts (95% CI 27–42%). Transposase-encoding (IVS3 spliced) transcripts show an even more pronounced 59% increase in expression in sterile genotypes (Fig 4A, t10 = 10.27, p = 2.91x10-10, 95% CI: 42–81%). By contrast, unspliced (IVS3 retaining) transcripts were not significantly differentially expressed between ovaries of F1 dysgenic offspring of sterile (B6) and fertile (B8) isogenic females (Fig 4A, t10 = 1.68, p = 0.11), although certain individual comparisons between strains were significant. To directly address whether splicing is more efficient in the ovaries of sterile dysgenic offspring, we compared the ratio of spliced to unspliced P-element transcripts (Fig 4B). The ratio of spliced to unspliced transcripts differed significantly between the ovaries sterile and fertile dysgenic offspring (t10 = 7.45, p = 7.30x10-5), suggesting that splicing itself is more efficient in sterile genotypes. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Decreased expression of P-transposase in fertile genotypes. A) Differential expression of P-element transcripts between fertile and sterile genotypes. Three separate qPCRs were performed, which detect all transcript isoforms, as well as IVS3 spliced and unspliced isoforms. qPCRs are normalized to rpl32. B) Ratios of IVS3 spliced to unspliced isoforms. C) Differential expression of splicing factors between sterile and fertile genotypes based on RNA-seq data. Dark green bars indicate factors that are significantly upregulated in fertile genetic backgrounds. Significant differences in qPCR data are based on linear models to detect differences between sterile and fertile genotypes, or Tukey-HSD comparisons to detect differences between genotypes containing the same allele. *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g004 We also observed differences in splicing and expression between isogenic stocks carrying the same allele. In particular, the Fertile_B81 and Fertile_B82 differed in both spliced and unspliced transcripts, with a particularly pronounced 2.89-fold increase in spliced transcript expression in Fertile_B82 as compared with Fertile_B81 (Fig 4A, t10 = 12.14, p = 9.78x10-12, 95% CI: 2.61–3.17 fold). While the sample size is too small to draw any conclusions, it is notable that Fertile_B82 does not exhibit the same degree of fertility rescue as Fertile_B81, further pointing to a connection between spliced transcript production and dysgenic sterility (Fig 1G and 1H). In germline cells, the splicing of IVS3 is known to be repressed by piRNA mediated transcriptional silencing, which is initiated by maternally transmitted piRNAs or through the production of de novo piRNAs in aged females [16,21]. In addition to the absence of maternally transmitted piRNAs, the splicing differences we observe here are in young (3–4 day old) dysgenic females, as opposed to aged dysgenic females, suggesting they are likely independent of the piRNA pathway. However, the splicing of IVS3 is also repressed by several host splicing factors in somatic cells, and it is proposed that some of these factors may also partially repress splicing in germline cells [62–64]. Consistent with piRNA-independent differences in splicing, we discovered that three splicing factors known to promote IVS3 retention in somatic cells, hrp36, hrp38 and P-element somatic inhibitor (Psi) show increased expression in fertile genotypes in our ovarian RNA seq data (Fig 4C). This suggests that decreased splicing in the dysgenic offspring of fertile isogenic lines may result from increased abundance of host splicing factors. Sterile alleles increase radiation sensitivity and accumulated mutations Our gene expression data suggest that sterile and fertile alleles may differ in their capacity to repair germline DSBs in young (3 day) dysgenic females. Fertile alleles exhibit upregulation of the TIP60 complex (which is involved in DSB repair [34]), while sterile alleles exhibit upregulation of replication dependent histones (which may complete with DNA repair machinery [38]). Mutations in DSB repair genes are widely known to cause radiation sensitivity, which is easily quantified by measuring lethality following larval radiation exposure [65–69]. While this assay occurs in whole larvae as compared to female germlines, larvae are composed of numerous classes of mitotically dividing cells, similar to the primordial and premeiotic stages of gamete production in which P-element transposition occurs [16,21]. Furthermore, numerous key factors for germline DNA damage response, as well as germline P-element excision repair, exhibit larval radiation sensitivity phenotypes [70–75]. We therefore compared the X-ray radiation sensitivity of larvae from isogenic lines containing sterile (B6) and fertile (B8) alleles. Note these lines are the same as those in Fig 1F–1H). After exploring a range of radiation doses, we found that doses above 10 Gy showed high lethality, making it difficult to detect differences in radiation sensitivity between the genotypes (S19 Table). Therefore, we compared the response of sterile and fertile larvae to radiation doses of 0 Gy, 5 Gy and 10 Gy. We observed that fertile genotypes had significantly higher survival (53–58%) than the sterile genotypes (25–30%) at 10 Gy (Fig 5A). Given that X-ray radiation produces predominantly DSBs, these results are consistent with differences between fertile and sterile alleles in DSB repair. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Sterile alleles exhibit reduced DNA repair. A) The percentage of mock treated and irradiated (5 Gy and 10 Gy) larvae that survived to adulthood for the fertile (B8), sterile (B6) and the control genotypes. CS refers to Canton-S and marker refers to the multiply marked stock b cn (#44229), which was used to generate isogenic lines. The numbers in the brackets refer to the sample size. The significance of comparisons between genotypes was determined by the χ2 test-of-independence. B and C) New mutations that accumulate in RIL genomes as detected by MSG. B) New SNPs and C) new indels. An excess of new mutations was detected by a t-test comparing Sterile B6 RILs to all others. The data represented in the Fig is provided in S19 and S20 Tables) ** denotes P< 0.01, *** denotes P < 0.001. https://doi.org/10.1371/journal.pgen.1010080.g005 We further looked specifically for differences in germline DSB repair by examining whether RILs carrying B6 alleles at the QTL-3d peak have accumulated more de novo base substitutions and small insertions or deletions. The DSPR RILs underwent 50 generations of inbreeding, and have since been maintained as isogenic lab stocks for ~175 generations, allowing ample time for new mutations to accrue as a consequence of deficient repair. To detect these new mutations, we generated multiplexed shotgun genotyping (MSG) libraries for 792 population B RILs [76]. This low coverage method (mean 2.9x) will uncover only a random subset of new mutations in each RIL, thereby underestimating the true amount of mutation accumulation. Nevertheless, we were able to detect 102,476 novel base substitutions and 5,026 novel insertions or deletions among the RIL MSG libraries. After accounting for differences in sequencing depth and plate effects, the founder allele at QTL-3d was associated with differences in the number of new base substitutions (-2ΔlnL = 15.62, df = 6, p = 0.016). Furthermore, RILs carrying B6 alleles at the QTL-3d peak exhibit an average increase of 14.58 new base substitutions (95% CI 5.23–24.12), when compared to those carrying another founder allele (t782 = 3.043, P = 0.0024, Fig 5B). In contrast, there was no significant association between founder allele at QTL-3d and new indels (-2ΔlnL = 1.37, df = 6, p = 0.97). Given the low coverage data as well the limited potential of short-read sequencing data to identify larger structural variation [77], we cannot be conclusive about a relationship between QTL allele and indel accumulation rate. Nevertheless, the increase in base substitution supports a deficiency in germline DNA repair in association with B6 alleles for QTL-3d. Identifying candidate genes The QTL we map here are quite large and contain numerous candidate genes whose differential function could influence dysgenic sterility. Nevertheless, we next sought to identify candidate genes that influence dysgenic sterility for future study. We combined our own expression and mapping data with previously published polymorphism and single cell expression data to narrow candidates based on four criteria: 1) location within a QTL, 2) expression in primordial germ cells or early, pre-meiotic cysts [78,79], 3) differential expression between sterile and fertile adult ovaries, and 4) the presence of “in-phase” single nucleotide polymorphisms (SNPs) (S11, S12 and S13 Tables). In-phase SNPs are those where the genotypic differences between the founder alleles are consistent with their phenotype class (Figs 1E and 6A [80]). Of 530 differentially expressed genes, 43 are within the QTL region, representing an approximately five-fold enrichment in the QTL regions compared to the rest of the genome (χ2 = 255.54, df = 1, p-value < 2.2e-16, Fig 6B). Ultimately, we identified 12 and 5 differentially expressed genes and early germ cell expressed genes that also carry in-phase SNPs within the QTL-3d and 21d, respectively (Fig 6C and 6D and S12 Table). Furthermore, we identified 32 genes in QTL-3d and 3 genes in QTL-21d that exhibit early germ cell expression and also contain in-phase non-synonymous SNPs, which may affect the function of the encoded protein (S13 Table). Collectively these genes represent the strongest candidates to contain causative variants. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Differential expression and in-phase SNPs identify candidate genes. a) The proportion of genes differentially expressed (DEG) is compared inside and outside the QTL. The dotted line is the genome wide average. b) Hypothetical in-phase and out of phase SNPs are shown. Sequences of each of the B founder strains are colored based on their phenotypic classification, either fertile or sterile (Fig 1E). Bold letters indicate SNPs. c and d) Venn diagrams showing the overlap of differentially expressed genes (DEG), genes carrying in-phase synonymous and non-synonymous SNPs, and genes expressed in primordial or pre-meiotic germ cells for QTL-21d (c) and QTL-3d (d).The data for differential expression of genes for fertile and sterile genotypes is provided in S5 Table. The data on in-phase polymorphisms for each QTL peak are provided in S11 Table. List of candidate genes that have both in-phase polymorphisms and are differentially expressed, and those having non-synonymous in-phase polymorphisms are provided in S12 and S13 Tables, respectively. https://doi.org/10.1371/journal.pgen.1010080.g006 We next scoured our list of candidate genes for those with known functions in chromatin regulation, DSB repair, or alternative splicing, whose differential function or regulation are plausibly related to the phenotypic differences associated with sterile and fertile alleles. None of the three splicing factors we discovered are differentially expressed reside within the QTL (Fig 4C), suggesting their expression differences arise as a consequence of regulatory differences in trans. While we did not discover any transcription factors located in the QTL that are differentially expressed in fertile and sterile ovaries, we did discover three C2H2 zinc finger transcription factors, tio and CG30431 (QTL-3d) and CG17568 (QTL-21d), that are located within QTL and contain in-phase non-synonymous SNPs. Unfortunately, the genomic binding sites of these transcription factors are undetermined, so it remains unknown if they are regulators of hrp36, hrp38, or psi transcription. With respect to differences in chromatin state and/or DNA repair, two genes within QTL-3d, stand out as particularly attractive candidates; Nipped-A and jing. Nipped-A contains a non-synonymous in-phase SNP and is expressed in both PGCs and in germline cells throughout the earliest stages of oogenesis (S13 Table). Nipped-A is a member of the TIP60 complex, which has functions in DSB repair, chromatin modification and chromatin remodeling. Additionally, we identified multiple TIP60 components upregulated in fertile ovaries (Fig 2C). The non-synonymous SNP that separates sterile and fertile alleles of this gene are located in the HEAT2 domain, which is predicted to be essential for protein-protein interaction, and could have important implications for the function of this multi-protein complex [81–83]. Jing contains in-phase synonymous and non-synonymous SNPs, is upregulated in fertile ovaries, and exhibits a similar expression pattern in germline cells to that of Nipped-A (Fig 2E and S12 and S13 Tables). Based on a yeast-two hybrid screen Jing physically interacts with inverted repeat binding protein 18 (IRBP18): a DNA binding protein that comprises part of a heterodimer that binds directly to P-element’s transcribed inverted repeats, and facilitates repair of donor DNA after excision [84,85]. Furthermore, irbp18 mutants exhibit larval radiation sensitivity, similar to our sterile genotypes (Fig 5A). If Jing determines differential activity of IRBP18 it could have a strong impact on dysgenic sterility. Beyond this function, Jing acts as an important cofactor of polycomb repressor complex 2, many of which showed increased expression in fertile ovaries Fig 2C, [86,87]. Discussion Although small RNA mediated TE regulation is widely studied, little is known about genetic variation in host factors that modulate the germline transposition of invading TEs and their associated fertility effects. Here we uncovered natural variation in dysgenic sterility imposed by P-element DNA transposons. Our work points to two major differences between sterile and fertile genotypes, which likely explains the differential occurrence of dysgenic sterility between them (Fig 7). First, fertile alleles suppress the splicing of transposase-encoding mRNA, which likely reduces the occurrence of germline DSBs that drive germ cell loss. Second, fertile alleles are more tolerant of DSBs, perhaps due to enhanced repair, which may allow them to retain germ cells despite the genotoxic effects of transposition. We propose that these differences highlight two axes of host-TE interaction: permissivity and tolerance. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. Schematic of phenotypic differences between sterile and fertile alleles. Differences in sterile and fertile alleles between IVS3 splicing and DNA repair are represented. In fertile alleles, host splicing suppressor expression is increased, leading to reduced production of spliced P-transposase encoding transcripts. As a consequence of reduced transposase production, it is predicted that fewer DSBs are produced in dysgenic females. However, it is also predicted that those breaks that are produced are repaired more efficiently. https://doi.org/10.1371/journal.pgen.1010080.g007 Host splicing factors determine differences in permissivity As intracellular parasites, TEs rely on host machinery for transcription, translation and replication. Variation in host co-factors that modulate these processes could drive differences in permissivity to TE proliferation. The concept of permissivity is prevalent in virology, and refers to the degree to which an individual cell type allows a virus to replicate [88–90]. With respect to TEs, permissivity is distinguishable from repression in that host-cofactors modulating permissivity precede the introduction of a TE into the genome, and the primary function of permissivity factors is not in TE regulation. For example, it has long been known that P-elements do not transpose in somatic cells due to the presence of host splicing factors that prevent the correct splicing of P-transposase encoding mRNA [22,63,91]. While the primary function of these splicing factors is not to regulate P-elements, their expression renders somatic cells non-permissive. In germline cells, maternally deposited piRNAs also regulate P-element transposition by promoting IVS3 retention [16]. However, even in the absence of maternally deposited piRNAs, IVS3-retaining transcripts are common, suggesting other host factors may also modulate P-element splicing in the germline [16,64]. Our work here reveals that there is host genetic variation in IVS3 splicing that is independent of maternally deposited piRNAs, which has potentially dramatic impacts on host fitness in dysgenic crosses (Figs 1B, 1G, 1H and 4A). While we cannot completely rule out a potential role for de novo piRNA production in driving these splicing differences, their occurrence in young dysgenic females and the absence of strong age effects on dysgenic sterility in our experiment points away from this explanation. Rather, the upregulation of multiple host splicing regulators in fertile genotypes suggest that the same factors that regulate IVS3 splicing in somatic cells may also modulate splicing in germline cells (Fig 4C). While we did not directly address whether these differences lead to differential transposition of P-transposons, these results suggest that fertile genotypes, like somatic cells, reduce permissivity through splicing regulation. Tolerance and DSB repair. In our previous work on natural variation in dysgenic sterility, we proposed that host genotypes may differ in tolerance: the ability of germline cells to persist and divide despite the damaging effects of transposition [14]. Because hybrid dysgenesis occurs through the loss of larval primordial germ cells and adult germline stem cells [15,16], the variable expression of factors that determine stem cells maintenance and differentiation could be an important source of tolerance variation. In particular, we found that the function of bruno, a differentiation factor in early pre-meiotic cysts, increases dysgenic sterility [14]. Conversely, the overexpression of the stem cell factor myc in dysgenic larval PGCs suppresses their loss, and by association, decreases dysgenic sterility [17]. Our work here suggests that genetic variation in DNA damage response or repair provides another potential mechanism of tolerance. Since transposition results in DSBs at sites of insertion and excision, enhanced ability to detect and repair these breaks would help reduce dysgenic germ cell loss. We observed that fertile genotypes are significantly more resilient to X-ray radiation (Fig 5A), a phenotype that is widely associated with increased activity of DNA repair genes [65–69]. Indeed, the magnitude of the differences in radiation sensitivity is large, and mirrors previous comparisons between wild-type and DNA damage response mutants such as p53 or checkpoint kinase 1 [74,75]. We further observed that fertile genotypes exhibit fewer accumulated base substitutions (Fig 5B), suggesting heritable differences in DNA repair. While DNA damage signaling is a clear determinant of dysgenic germ cell loss [15,18,19,21], to our knowledge this is the first evidence that natural variation in DNA repair may modulate the sterility effects of transposition. However, we cannot rule out the possibility reduced DSB repair also increases permissivity, by prolonging S-phase thereby allowing more time for P-elements to transpose. Something that remains puzzling about our observations regarding DNA damage and repair is that it is not intuitively obvious how deficiencies in DSB repair would lead to an accumulation of base substitutions in RILs carrying sterile alleles. However, DNA repair pathways are interdependent, with many components impacting multiple repair processes. For example, mei-9 is required for both nucleotide excision repair and meiotic recombination [92], and mutant alleles are known to enhance germline loss in P-element dysgenic males [18,93]. Beyond this, homologous repair of DSBs leads to an increased rate of base substitution, potentially due to the sensitivity of single stranded repair intermediates to other forms of DNA damage reviewed in [94]. Therefore, if the DSB repair deficiency associated with sterile alleles results in a larger number of DSBs repaired by homologous pathways, or a delay in those repairs, an increase in base substitution is predicted. Host splicing factors determine differences in permissivity As intracellular parasites, TEs rely on host machinery for transcription, translation and replication. Variation in host co-factors that modulate these processes could drive differences in permissivity to TE proliferation. The concept of permissivity is prevalent in virology, and refers to the degree to which an individual cell type allows a virus to replicate [88–90]. With respect to TEs, permissivity is distinguishable from repression in that host-cofactors modulating permissivity precede the introduction of a TE into the genome, and the primary function of permissivity factors is not in TE regulation. For example, it has long been known that P-elements do not transpose in somatic cells due to the presence of host splicing factors that prevent the correct splicing of P-transposase encoding mRNA [22,63,91]. While the primary function of these splicing factors is not to regulate P-elements, their expression renders somatic cells non-permissive. In germline cells, maternally deposited piRNAs also regulate P-element transposition by promoting IVS3 retention [16]. However, even in the absence of maternally deposited piRNAs, IVS3-retaining transcripts are common, suggesting other host factors may also modulate P-element splicing in the germline [16,64]. Our work here reveals that there is host genetic variation in IVS3 splicing that is independent of maternally deposited piRNAs, which has potentially dramatic impacts on host fitness in dysgenic crosses (Figs 1B, 1G, 1H and 4A). While we cannot completely rule out a potential role for de novo piRNA production in driving these splicing differences, their occurrence in young dysgenic females and the absence of strong age effects on dysgenic sterility in our experiment points away from this explanation. Rather, the upregulation of multiple host splicing regulators in fertile genotypes suggest that the same factors that regulate IVS3 splicing in somatic cells may also modulate splicing in germline cells (Fig 4C). While we did not directly address whether these differences lead to differential transposition of P-transposons, these results suggest that fertile genotypes, like somatic cells, reduce permissivity through splicing regulation. Tolerance and DSB repair. In our previous work on natural variation in dysgenic sterility, we proposed that host genotypes may differ in tolerance: the ability of germline cells to persist and divide despite the damaging effects of transposition [14]. Because hybrid dysgenesis occurs through the loss of larval primordial germ cells and adult germline stem cells [15,16], the variable expression of factors that determine stem cells maintenance and differentiation could be an important source of tolerance variation. In particular, we found that the function of bruno, a differentiation factor in early pre-meiotic cysts, increases dysgenic sterility [14]. Conversely, the overexpression of the stem cell factor myc in dysgenic larval PGCs suppresses their loss, and by association, decreases dysgenic sterility [17]. Our work here suggests that genetic variation in DNA damage response or repair provides another potential mechanism of tolerance. Since transposition results in DSBs at sites of insertion and excision, enhanced ability to detect and repair these breaks would help reduce dysgenic germ cell loss. We observed that fertile genotypes are significantly more resilient to X-ray radiation (Fig 5A), a phenotype that is widely associated with increased activity of DNA repair genes [65–69]. Indeed, the magnitude of the differences in radiation sensitivity is large, and mirrors previous comparisons between wild-type and DNA damage response mutants such as p53 or checkpoint kinase 1 [74,75]. We further observed that fertile genotypes exhibit fewer accumulated base substitutions (Fig 5B), suggesting heritable differences in DNA repair. While DNA damage signaling is a clear determinant of dysgenic germ cell loss [15,18,19,21], to our knowledge this is the first evidence that natural variation in DNA repair may modulate the sterility effects of transposition. However, we cannot rule out the possibility reduced DSB repair also increases permissivity, by prolonging S-phase thereby allowing more time for P-elements to transpose. Something that remains puzzling about our observations regarding DNA damage and repair is that it is not intuitively obvious how deficiencies in DSB repair would lead to an accumulation of base substitutions in RILs carrying sterile alleles. However, DNA repair pathways are interdependent, with many components impacting multiple repair processes. For example, mei-9 is required for both nucleotide excision repair and meiotic recombination [92], and mutant alleles are known to enhance germline loss in P-element dysgenic males [18,93]. Beyond this, homologous repair of DSBs leads to an increased rate of base substitution, potentially due to the sensitivity of single stranded repair intermediates to other forms of DNA damage reviewed in [94]. Therefore, if the DSB repair deficiency associated with sterile alleles results in a larger number of DSBs repaired by homologous pathways, or a delay in those repairs, an increase in base substitution is predicted. Tolerance and DSB repair. In our previous work on natural variation in dysgenic sterility, we proposed that host genotypes may differ in tolerance: the ability of germline cells to persist and divide despite the damaging effects of transposition [14]. Because hybrid dysgenesis occurs through the loss of larval primordial germ cells and adult germline stem cells [15,16], the variable expression of factors that determine stem cells maintenance and differentiation could be an important source of tolerance variation. In particular, we found that the function of bruno, a differentiation factor in early pre-meiotic cysts, increases dysgenic sterility [14]. Conversely, the overexpression of the stem cell factor myc in dysgenic larval PGCs suppresses their loss, and by association, decreases dysgenic sterility [17]. Our work here suggests that genetic variation in DNA damage response or repair provides another potential mechanism of tolerance. Since transposition results in DSBs at sites of insertion and excision, enhanced ability to detect and repair these breaks would help reduce dysgenic germ cell loss. We observed that fertile genotypes are significantly more resilient to X-ray radiation (Fig 5A), a phenotype that is widely associated with increased activity of DNA repair genes [65–69]. Indeed, the magnitude of the differences in radiation sensitivity is large, and mirrors previous comparisons between wild-type and DNA damage response mutants such as p53 or checkpoint kinase 1 [74,75]. We further observed that fertile genotypes exhibit fewer accumulated base substitutions (Fig 5B), suggesting heritable differences in DNA repair. While DNA damage signaling is a clear determinant of dysgenic germ cell loss [15,18,19,21], to our knowledge this is the first evidence that natural variation in DNA repair may modulate the sterility effects of transposition. However, we cannot rule out the possibility reduced DSB repair also increases permissivity, by prolonging S-phase thereby allowing more time for P-elements to transpose. Something that remains puzzling about our observations regarding DNA damage and repair is that it is not intuitively obvious how deficiencies in DSB repair would lead to an accumulation of base substitutions in RILs carrying sterile alleles. However, DNA repair pathways are interdependent, with many components impacting multiple repair processes. For example, mei-9 is required for both nucleotide excision repair and meiotic recombination [92], and mutant alleles are known to enhance germline loss in P-element dysgenic males [18,93]. Beyond this, homologous repair of DSBs leads to an increased rate of base substitution, potentially due to the sensitivity of single stranded repair intermediates to other forms of DNA damage reviewed in [94]. Therefore, if the DSB repair deficiency associated with sterile alleles results in a larger number of DSBs repaired by homologous pathways, or a delay in those repairs, an increase in base substitution is predicted. Conclusion The degree to which innate differences among hosts govern the propagation of an invading TE, as well as its fitness effects during invasion, is an understudied aspect of TE invasion biology. Here we have uncovered two different forms of host genetic variation in dysgenic sterility, which alter the permissivity of host cells to transposition, as well as their tolerance to transposition’s effects. These observations add complexity to our current understanding of how host genetic variation can modulate the fitness effects of an invading TE. The precise pathways and genetic factors whose differential function underlie these tolerance and permissivity phenotypes remain to be resolved. Similarly, the degree to which these processes reflect the pleiotropic effects of a single gene, or the combined action of multiple factors remains an important question to be addressed by future work. Methods Drosophila strains and husbandry The recombinant inbred lines are described in King et al. [24]. Harwich (#4264) and b cn (#44229), were obtained from the Bloomington Drosophila stock center. Canton-S was obtained from Brigitte Dauwalder. All flies were maintained on standard cornmeal media. Alleles of the second chromosome centromeric region, containing both QTL, were extracted from three recombinant inbred lines carrying B6 QTL allele (#21076, #21218, #21156) and two RILs carrying B8 QTL allele (#21077, #21154) into a common background by crossing them to multiply marked stocks b cn (#44229). After 7 rounds of backcrossing followed by inbreeding, the final isogenic lines (Sterile_B61, Sterile_B62, Sterile_B63 and Fertile_B81, Fertile_B82) were generated. The lines were made homozygous for the 2nd chromosome by inbreeding and selecting for wild type phenotype. The genotype of the isogenic lines were verified through PCR using five different primers within the two QTL. chr2L:19383155–19383970: AACCCTTTTTCGCTGACAATAACA, ATTATCAGCAGGAGCCGGAAACTT; chr2L:21333500–21334300: AAGTGAAGCTAACAACGTGACAAC, CGTTTGACCATCGCTTACAACTAA; chr2R:2392800–2393600: AACAGGAGGTCGAAAGCCAAATA, ATGCAGAGTCATATTCTGGGTTGG; chr2R:6203290–6204284: AATGGAGACCGTTGATTTTGGTAA, CTTTTCTGCGGCATCAGGTG; chr2R:6058000–6059000: TGGCAATTGCAATCCTTTTGGTAT, ATAACACGAACTACGACCTTTCCA. Phenotyping. Phenotyping of ovarian atrophy was performed as described previously in Kelleher et al. [14]. Briefly, crosses between virgin RIL females and Harwich males were transferred to fresh food every 3–5 days. Since crosses reared at a restrictive temperature (29oC) result in complete gonadal atrophy in F1 offspring, we reared our crosses at a lower permissive temperature (25oC), which produces an intermediate phenotype that better reveals the variation in severity of dysgenesis [12,14,15,95]. F1 offspring were maintained for 3 days or 21 days, at which point their ovaries were examined using a squash prep [95]. 21 day- old females were transferred onto new food every 5 days as they aged to avoid bacterial growth. Females who produced 1 or more chorionated egg chambers were scored as having non-atrophied ovaries, and females producing 0 egg chambers were scored as having atrophied ovaries. Crosses and phenotyping were performed for 673 RILs across 22 experimental blocks for 3 day-old F1 females, and 552 RILs across 18 experimental blocks for 21 day-old F1 females. If fewer than 21 F1 offspring were phenotyped for the same cross, it was discarded and repeated if possible. In total, we phenotyped >20 3-day old and 21 day-old F1 female offspring for 595 RILs and 456 RILs, respectively. QTL mapping. QTL mapping was performed as described in Kelleher et al. [14]. Briefly, for each developmental time point, we modeled the arcsine transformed proportion of F1 ovarian atrophy as a function of two random effects: experimental block and undergraduate experimenter. Regression models were fit using the lmer function from the lme4 package [96]. We then used the residuals as a response for QTL mapping with the DSPRqtl package [24] in R 3.02 [97]. The LOD significance threshold was determined from 1,000 permutations of the observed data, and the confidence interval around each LOD peak was identified by a difference of -2 from the LOD peak position (Δ2-LOD) [26], or from the Bayes Confidence Interval [98]. For Δ2-LOD intervals, we took the conservative approach of determining the longest contiguous interval where the LOD score was within 2 of the peak value. We further calculated the broad sense heritability of ovarian atrophy as in Kelleher et al. [14]. Estimation of founder phenotypes and QTL phasing. To estimate the phenotypic effect associated with each founder allele at the QTL peak, we considered the distribution of phenotypes from all RILs carrying the founder haplotype at the LOD peak position (genotype probability >0.95%) [24]. QTL were then phased into allelic classes by identifying the minimal number of partitions of founder haplotypes that describe phenotypic variation associated with the QTL peak, as described previously [14,24]. Fertility assays. Virgin female offspring from dysgenic crosses between isogenic lines carrying fertile_B81/B82 (21077, 21154) and sterile_B62/B63 (21218, 21156) alleles and Harwich males were collected daily and individually placed in a vial containing two Canton-S males. Females were allowed to mate for 5 days and were transferred to a new vial for another 5 days after which the parents were discarded. The presence and total number of F2 individuals were counted from the two vials. Selection of paired RILs with alternate QTL alleles. We identified background matched RILs containing either the B6 (sterile) or B4 (fertile) haplotypes from the start position of the QTL-21d confidence interval (2L: 19,010,000) to the end position of QTL-3d confidence interval (2R: 6,942,495) (P > 0.9), based on their published HMM genotypes [24]. For all possible RIL pairs (B6 and B4), we then calculated the number of 10 Kb genomic windows in which they carried the same RIL haplotype (P > 0.9). We selected three pairs of RILs, which carry the same founder genotype for 47% (21213 & 21183), 46% (21147 & 21346) and 44% (21291 & 21188) of genomic windows outside of the QTL. Small RNA-seq and total RNA-seq. RILs were maintained at 25°C, and three biological replicates of 20 ovaries were dissected from 3–5 day-old females. Ovaries were homogenized in TRIzol (invitrogen) and stored at -80°C until RNA extraction. 50 μg of total RNA from each of 18 biological samples (3 biological replicates x 3 pairs) was size fractionated in a 15% denaturing polyacrylamide gel and the 18–30 nt band was excised. 2S-depleted small RNA libraries for Illumina sequencing were then constructed according to the method of Wickersheim and Blumenstiel [99]. Ovarian small RNA libraries were published previously SRP160954, [100]. Ribodepleted and stranded total RNA libraries were generated from the same ovarian samples using NuGen total RNA kit (TECAN). All 18 small RNA and total RNA libraries were sequenced on an Illumina Nextseq 500 at the University of Houston Seq-N-Edit Core. Small-RNA analysis. Sequenced small RNAs were separated based on size into miRNAs/siRNAs (18-22nt) and piRNAs (23-30nt) [11]. Reads corresponding to contaminating rRNAs, including 2S-rRNA, were removed from each library by aligning to annotated transcripts from flybase [101]. To determine the piRNA cluster activity we first uniquely aligned the piRNAs to reference genome (dm6 [29]) using Bowtie1 (-v 1 -m 1) [102]. We then used a customized perl script to count reads that mapped to a set of previously annotated piRNA clusters from the same genotypes (497 piRNA clusters, [103]). Read counts normalized to total mapped microRNAs for each library were used to infer differential expression using DESeq2 [104]. Sliding window estimates of piRNA abundance (Fig 2C and 2D) were calculated using bedtools genomecov [105], normalizing the read counts to total mapped miRNA reads. Total RNA analysis. Residual ribosomal RNAs (rRNAs) were identified in ribo depleted libraries based on alignment to annotated rRNAs from flybase [101], and excluded from further analysis. Retained reads aligned to the library of consensus satellite and TE sequences from repbase [106], plus additional satellite consensus sequences from Larracuente [107]. For TE expression, the total reads mapped to TE sequences were counted using unix commands (uniq -c). Remaining reads that failed to map were pseudoaligned to D. melanogaster transcriptome (dm6/BDGP6) using Kallisto with default parameters [108]. Differentially expressed TEs and genes were identified from a combined analysis in DESeq2 [104]. Genes and TEs with base mean > = 100, Adjusted P-value < = 0.05 and whose expression pattern differed (fold change > = 1.5) were considered differentially expressed between the B6 and B4 QTL haplotype. Radiation sensitivity. Third instar larvae were either mock treated or irradiated in a Rad Source RS 1800 X-ray machine set at 12.5 mA and 160 kV. To obtain 3rd instar larvae, embryos were collected for 24 hr and aged for 5 days at 25°C. The food vials containing larvae were then X-ray irradiated at doses from 5–80 Gray after which an optimal dose that clearly depicts the phenotypic difference was selected. Survival to adulthood was determined by scoring the number of empty and full pupal cases at 10 days after radiation. Identification of novel mutations in RIL genomes. 10 females from each of 795 RILs were deposited into a well of a 96 well plate (Axygen, P-96-450R-C) on ice. DNA isolation was then executed in plates using the Gentra Puregene Cell Kit (Qiagen, 158788) using extensions of the manufacturer’s protocol. Subsequently, DNA was further purified (Qiagen, QIAquick 96 PCR Purification kit, 28183), quantified via a fluorometer (ThermoFisher, Qubit dsDNA HS kit, Q32854), and diluted to 2-ng/μl. Each RIL sample was then subjected to the MSG (Multiplexed Shotgun Genotyping) approach developed by Andolfatto et al. [76], which is a form of RADseq (restriction-site associated DNA sequencing, Baird et al. [109]). Starting with 10-ng DNA, samples were first digested using MseI (New England Biolabs, R0525L). This enzyme has the restriction site T/TAA, cuts frequently along the genome, and unlike traditional rare-cutter RADseq strategies, yields sequencing reads spread somewhat evenly along the genome. Next, plate well-specific barcoded adapters are independently ligated onto the cut ends of each DNA sample. Fragmented, barcoded samples from a given plate are then mixed, and the 96-plex pool precipitated, purified, and size-selected to 250-300-bp via a Pippin Prep (Sage Science). Each multiplexed sample is then PCR amplified, during which a DNA plate-specific Illumina-compatible index is incorporated, and then purified. Finally, each of the independently-indexed 96-plex pools are quantified, mixed at equal concentration, and sequenced over multiple lanes of an Illumina HiSeq 2500 on “high output” mode, yielding single-end 100-bp reads (KU Genome Sequencing Core). With this MSG approach, reads from each RIL are computationally distinguished by both an Illumina index sequence (which marks the plates), and an “in line barcode” (the first 6 Read1 bases, which marks samples on any given plate). SNP and indel variants were identified from MSG short-read data following GATK best practices for sample groups [110–112]. Briefly, Cutadapt (version 3.5; [113]) was used to de-multiplex samples and trim adaptors, while alignment to the D. melanogaster reference genome (dm6 [29]) were performed using BWA (version 0.7.17-GCC-10.2.0;[114]). The resulting BAM files were sorted and indexed using Samtools (Li et al., 2009 [114]). Individual GVCF files were generated using HaplotypeCaller and then joint-genotyped using Genotype_GVCFs. Both SNPs and indels were extracted and filtered out following the GATK Best Practices hard filters using VariantFiltration (SNP: "QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < -12.5”; indel “"QD < 2.0 || FS > 200.0”, and converted to TSV in R for further analysis using the vcfR package [115]. To identify novel base substitutions that arose in RILs and not present in the founders, we filtered out all alleles with spanning deletions as well as annotated SNPs from the founder lines (https://wfitch.bio.uci.edu/~dspr/Data/index.html). Since founder SNPs were called in dm5, we converted their coordinates to dm6 using the NCBI Genome Remapping Service (https://www.ncbi.nlm.nih.gov/genome/tools/remap) before filtering. For indels, short indels were not called from the original founder sequence data. We therefore considered an indel to be novel if they were unique to a RIL and sequenced among at least 50 RILs. Differences in the number of SNPs and indels rate was modeled using linear mixed model fitted with lmer function in the R package lme4 (Bates et al., 2013 [96]). Three models were compared, a null model (mutations ~ plate + depth), a founder allele model (mutations ~ plate + depth + founder allele), and a dysgenic sterility allele model (mutations ~ plate + depth + sterile/fertile). Founder or dysgenic sterility allele referred to the QTL-3d peak: (B1-B5,B6-B8) or sterile/fertile (B6/all other founders). Note that B5 alleles are not present among the RILs at the QTL-3d peak. The models were compared using a likelihood ratio test to determine whether founder allele or allelic class explained variation in the number of novel SNPs or indels between RILs. The effect of sterile alleles on SNP and indel number was evaluated by t-test. qRT-PCR of P-element transcripts 3 biological replicates including 20 pairs of 3–5 day F1 dysgenic ovaries from crosses between fertile (B81, B82) or sterile (B62, B63) females and Harwich males were dissected and homogenized in TRIzol. Crosses were maintained at 22°C. RNA was treated with DNA-free (ThermoFisher) and reverse transcribed using oligo-dT primers and superscript IV (ThermoFisher) according to manufacturer instructions. Three different primer sets were used to amplify the 3’ end of all P-element transcripts, IVS3 spliced transcripts and IVS3 unspliced transcripts, as well as rpl32. Transcripts were amplified and quantified in three technical replicates using power-SYBR green (ThermoFisher) and normalized to rpl32 for the same sample. Primer sequences were as follows: rpl32-F: 5’-CCGCTTCAAGGGACAGTATC, rpl32-R: 5’-GACAATCTCCTTGCGCTTCT, P-element-all F: 5’-CACCGAAATGGATGAGTTGACG, P-element-all R: 5’-TAATAAGTCCGCCGTGAGACAC, P-element IVS3 spliced F: 5’-AATAGCCAGGAATACAGAAATG, P-element IVS3 spliced R: 5’-AACATTTCTGTATTCCTGGCTA, P-element IVS3-unspliced F: 5’-GACAAAACACAATAGACAGCACA, P-element IVS3-unspliced R: 5’-TGTGCTGTCTATTGTGTTTTGTC. Identification of in-phase polymorphisms. The SNP data of B founders that used to infer in-phase SNPs is based on dm3 [24]. To identify in-phase SNPs we looked for alternate SNP alleles that match the predicted phenotypic class for each of the QTL peaks. For QTL-21d we used the criteria: sterile class (B2, B6) and the fertile class (B1, B3, B4, B7, B8), whereas for QTL-3d: sterile class (B6) and the fertile class (B1, B2, B3, B4, B7, B8). Drosophila strains and husbandry The recombinant inbred lines are described in King et al. [24]. Harwich (#4264) and b cn (#44229), were obtained from the Bloomington Drosophila stock center. Canton-S was obtained from Brigitte Dauwalder. All flies were maintained on standard cornmeal media. Alleles of the second chromosome centromeric region, containing both QTL, were extracted from three recombinant inbred lines carrying B6 QTL allele (#21076, #21218, #21156) and two RILs carrying B8 QTL allele (#21077, #21154) into a common background by crossing them to multiply marked stocks b cn (#44229). After 7 rounds of backcrossing followed by inbreeding, the final isogenic lines (Sterile_B61, Sterile_B62, Sterile_B63 and Fertile_B81, Fertile_B82) were generated. The lines were made homozygous for the 2nd chromosome by inbreeding and selecting for wild type phenotype. The genotype of the isogenic lines were verified through PCR using five different primers within the two QTL. chr2L:19383155–19383970: AACCCTTTTTCGCTGACAATAACA, ATTATCAGCAGGAGCCGGAAACTT; chr2L:21333500–21334300: AAGTGAAGCTAACAACGTGACAAC, CGTTTGACCATCGCTTACAACTAA; chr2R:2392800–2393600: AACAGGAGGTCGAAAGCCAAATA, ATGCAGAGTCATATTCTGGGTTGG; chr2R:6203290–6204284: AATGGAGACCGTTGATTTTGGTAA, CTTTTCTGCGGCATCAGGTG; chr2R:6058000–6059000: TGGCAATTGCAATCCTTTTGGTAT, ATAACACGAACTACGACCTTTCCA. Phenotyping. Phenotyping of ovarian atrophy was performed as described previously in Kelleher et al. [14]. Briefly, crosses between virgin RIL females and Harwich males were transferred to fresh food every 3–5 days. Since crosses reared at a restrictive temperature (29oC) result in complete gonadal atrophy in F1 offspring, we reared our crosses at a lower permissive temperature (25oC), which produces an intermediate phenotype that better reveals the variation in severity of dysgenesis [12,14,15,95]. F1 offspring were maintained for 3 days or 21 days, at which point their ovaries were examined using a squash prep [95]. 21 day- old females were transferred onto new food every 5 days as they aged to avoid bacterial growth. Females who produced 1 or more chorionated egg chambers were scored as having non-atrophied ovaries, and females producing 0 egg chambers were scored as having atrophied ovaries. Crosses and phenotyping were performed for 673 RILs across 22 experimental blocks for 3 day-old F1 females, and 552 RILs across 18 experimental blocks for 21 day-old F1 females. If fewer than 21 F1 offspring were phenotyped for the same cross, it was discarded and repeated if possible. In total, we phenotyped >20 3-day old and 21 day-old F1 female offspring for 595 RILs and 456 RILs, respectively. QTL mapping. QTL mapping was performed as described in Kelleher et al. [14]. Briefly, for each developmental time point, we modeled the arcsine transformed proportion of F1 ovarian atrophy as a function of two random effects: experimental block and undergraduate experimenter. Regression models were fit using the lmer function from the lme4 package [96]. We then used the residuals as a response for QTL mapping with the DSPRqtl package [24] in R 3.02 [97]. The LOD significance threshold was determined from 1,000 permutations of the observed data, and the confidence interval around each LOD peak was identified by a difference of -2 from the LOD peak position (Δ2-LOD) [26], or from the Bayes Confidence Interval [98]. For Δ2-LOD intervals, we took the conservative approach of determining the longest contiguous interval where the LOD score was within 2 of the peak value. We further calculated the broad sense heritability of ovarian atrophy as in Kelleher et al. [14]. Estimation of founder phenotypes and QTL phasing. To estimate the phenotypic effect associated with each founder allele at the QTL peak, we considered the distribution of phenotypes from all RILs carrying the founder haplotype at the LOD peak position (genotype probability >0.95%) [24]. QTL were then phased into allelic classes by identifying the minimal number of partitions of founder haplotypes that describe phenotypic variation associated with the QTL peak, as described previously [14,24]. Fertility assays. Virgin female offspring from dysgenic crosses between isogenic lines carrying fertile_B81/B82 (21077, 21154) and sterile_B62/B63 (21218, 21156) alleles and Harwich males were collected daily and individually placed in a vial containing two Canton-S males. Females were allowed to mate for 5 days and were transferred to a new vial for another 5 days after which the parents were discarded. The presence and total number of F2 individuals were counted from the two vials. Selection of paired RILs with alternate QTL alleles. We identified background matched RILs containing either the B6 (sterile) or B4 (fertile) haplotypes from the start position of the QTL-21d confidence interval (2L: 19,010,000) to the end position of QTL-3d confidence interval (2R: 6,942,495) (P > 0.9), based on their published HMM genotypes [24]. For all possible RIL pairs (B6 and B4), we then calculated the number of 10 Kb genomic windows in which they carried the same RIL haplotype (P > 0.9). We selected three pairs of RILs, which carry the same founder genotype for 47% (21213 & 21183), 46% (21147 & 21346) and 44% (21291 & 21188) of genomic windows outside of the QTL. Small RNA-seq and total RNA-seq. RILs were maintained at 25°C, and three biological replicates of 20 ovaries were dissected from 3–5 day-old females. Ovaries were homogenized in TRIzol (invitrogen) and stored at -80°C until RNA extraction. 50 μg of total RNA from each of 18 biological samples (3 biological replicates x 3 pairs) was size fractionated in a 15% denaturing polyacrylamide gel and the 18–30 nt band was excised. 2S-depleted small RNA libraries for Illumina sequencing were then constructed according to the method of Wickersheim and Blumenstiel [99]. Ovarian small RNA libraries were published previously SRP160954, [100]. Ribodepleted and stranded total RNA libraries were generated from the same ovarian samples using NuGen total RNA kit (TECAN). All 18 small RNA and total RNA libraries were sequenced on an Illumina Nextseq 500 at the University of Houston Seq-N-Edit Core. Small-RNA analysis. Sequenced small RNAs were separated based on size into miRNAs/siRNAs (18-22nt) and piRNAs (23-30nt) [11]. Reads corresponding to contaminating rRNAs, including 2S-rRNA, were removed from each library by aligning to annotated transcripts from flybase [101]. To determine the piRNA cluster activity we first uniquely aligned the piRNAs to reference genome (dm6 [29]) using Bowtie1 (-v 1 -m 1) [102]. We then used a customized perl script to count reads that mapped to a set of previously annotated piRNA clusters from the same genotypes (497 piRNA clusters, [103]). Read counts normalized to total mapped microRNAs for each library were used to infer differential expression using DESeq2 [104]. Sliding window estimates of piRNA abundance (Fig 2C and 2D) were calculated using bedtools genomecov [105], normalizing the read counts to total mapped miRNA reads. Total RNA analysis. Residual ribosomal RNAs (rRNAs) were identified in ribo depleted libraries based on alignment to annotated rRNAs from flybase [101], and excluded from further analysis. Retained reads aligned to the library of consensus satellite and TE sequences from repbase [106], plus additional satellite consensus sequences from Larracuente [107]. For TE expression, the total reads mapped to TE sequences were counted using unix commands (uniq -c). Remaining reads that failed to map were pseudoaligned to D. melanogaster transcriptome (dm6/BDGP6) using Kallisto with default parameters [108]. Differentially expressed TEs and genes were identified from a combined analysis in DESeq2 [104]. Genes and TEs with base mean > = 100, Adjusted P-value < = 0.05 and whose expression pattern differed (fold change > = 1.5) were considered differentially expressed between the B6 and B4 QTL haplotype. Radiation sensitivity. Third instar larvae were either mock treated or irradiated in a Rad Source RS 1800 X-ray machine set at 12.5 mA and 160 kV. To obtain 3rd instar larvae, embryos were collected for 24 hr and aged for 5 days at 25°C. The food vials containing larvae were then X-ray irradiated at doses from 5–80 Gray after which an optimal dose that clearly depicts the phenotypic difference was selected. Survival to adulthood was determined by scoring the number of empty and full pupal cases at 10 days after radiation. Identification of novel mutations in RIL genomes. 10 females from each of 795 RILs were deposited into a well of a 96 well plate (Axygen, P-96-450R-C) on ice. DNA isolation was then executed in plates using the Gentra Puregene Cell Kit (Qiagen, 158788) using extensions of the manufacturer’s protocol. Subsequently, DNA was further purified (Qiagen, QIAquick 96 PCR Purification kit, 28183), quantified via a fluorometer (ThermoFisher, Qubit dsDNA HS kit, Q32854), and diluted to 2-ng/μl. Each RIL sample was then subjected to the MSG (Multiplexed Shotgun Genotyping) approach developed by Andolfatto et al. [76], which is a form of RADseq (restriction-site associated DNA sequencing, Baird et al. [109]). Starting with 10-ng DNA, samples were first digested using MseI (New England Biolabs, R0525L). This enzyme has the restriction site T/TAA, cuts frequently along the genome, and unlike traditional rare-cutter RADseq strategies, yields sequencing reads spread somewhat evenly along the genome. Next, plate well-specific barcoded adapters are independently ligated onto the cut ends of each DNA sample. Fragmented, barcoded samples from a given plate are then mixed, and the 96-plex pool precipitated, purified, and size-selected to 250-300-bp via a Pippin Prep (Sage Science). Each multiplexed sample is then PCR amplified, during which a DNA plate-specific Illumina-compatible index is incorporated, and then purified. Finally, each of the independently-indexed 96-plex pools are quantified, mixed at equal concentration, and sequenced over multiple lanes of an Illumina HiSeq 2500 on “high output” mode, yielding single-end 100-bp reads (KU Genome Sequencing Core). With this MSG approach, reads from each RIL are computationally distinguished by both an Illumina index sequence (which marks the plates), and an “in line barcode” (the first 6 Read1 bases, which marks samples on any given plate). SNP and indel variants were identified from MSG short-read data following GATK best practices for sample groups [110–112]. Briefly, Cutadapt (version 3.5; [113]) was used to de-multiplex samples and trim adaptors, while alignment to the D. melanogaster reference genome (dm6 [29]) were performed using BWA (version 0.7.17-GCC-10.2.0;[114]). The resulting BAM files were sorted and indexed using Samtools (Li et al., 2009 [114]). Individual GVCF files were generated using HaplotypeCaller and then joint-genotyped using Genotype_GVCFs. Both SNPs and indels were extracted and filtered out following the GATK Best Practices hard filters using VariantFiltration (SNP: "QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < -12.5”; indel “"QD < 2.0 || FS > 200.0”, and converted to TSV in R for further analysis using the vcfR package [115]. To identify novel base substitutions that arose in RILs and not present in the founders, we filtered out all alleles with spanning deletions as well as annotated SNPs from the founder lines (https://wfitch.bio.uci.edu/~dspr/Data/index.html). Since founder SNPs were called in dm5, we converted their coordinates to dm6 using the NCBI Genome Remapping Service (https://www.ncbi.nlm.nih.gov/genome/tools/remap) before filtering. For indels, short indels were not called from the original founder sequence data. We therefore considered an indel to be novel if they were unique to a RIL and sequenced among at least 50 RILs. Differences in the number of SNPs and indels rate was modeled using linear mixed model fitted with lmer function in the R package lme4 (Bates et al., 2013 [96]). Three models were compared, a null model (mutations ~ plate + depth), a founder allele model (mutations ~ plate + depth + founder allele), and a dysgenic sterility allele model (mutations ~ plate + depth + sterile/fertile). Founder or dysgenic sterility allele referred to the QTL-3d peak: (B1-B5,B6-B8) or sterile/fertile (B6/all other founders). Note that B5 alleles are not present among the RILs at the QTL-3d peak. The models were compared using a likelihood ratio test to determine whether founder allele or allelic class explained variation in the number of novel SNPs or indels between RILs. The effect of sterile alleles on SNP and indel number was evaluated by t-test. Phenotyping. Phenotyping of ovarian atrophy was performed as described previously in Kelleher et al. [14]. Briefly, crosses between virgin RIL females and Harwich males were transferred to fresh food every 3–5 days. Since crosses reared at a restrictive temperature (29oC) result in complete gonadal atrophy in F1 offspring, we reared our crosses at a lower permissive temperature (25oC), which produces an intermediate phenotype that better reveals the variation in severity of dysgenesis [12,14,15,95]. F1 offspring were maintained for 3 days or 21 days, at which point their ovaries were examined using a squash prep [95]. 21 day- old females were transferred onto new food every 5 days as they aged to avoid bacterial growth. Females who produced 1 or more chorionated egg chambers were scored as having non-atrophied ovaries, and females producing 0 egg chambers were scored as having atrophied ovaries. Crosses and phenotyping were performed for 673 RILs across 22 experimental blocks for 3 day-old F1 females, and 552 RILs across 18 experimental blocks for 21 day-old F1 females. If fewer than 21 F1 offspring were phenotyped for the same cross, it was discarded and repeated if possible. In total, we phenotyped >20 3-day old and 21 day-old F1 female offspring for 595 RILs and 456 RILs, respectively. QTL mapping. QTL mapping was performed as described in Kelleher et al. [14]. Briefly, for each developmental time point, we modeled the arcsine transformed proportion of F1 ovarian atrophy as a function of two random effects: experimental block and undergraduate experimenter. Regression models were fit using the lmer function from the lme4 package [96]. We then used the residuals as a response for QTL mapping with the DSPRqtl package [24] in R 3.02 [97]. The LOD significance threshold was determined from 1,000 permutations of the observed data, and the confidence interval around each LOD peak was identified by a difference of -2 from the LOD peak position (Δ2-LOD) [26], or from the Bayes Confidence Interval [98]. For Δ2-LOD intervals, we took the conservative approach of determining the longest contiguous interval where the LOD score was within 2 of the peak value. We further calculated the broad sense heritability of ovarian atrophy as in Kelleher et al. [14]. Estimation of founder phenotypes and QTL phasing. To estimate the phenotypic effect associated with each founder allele at the QTL peak, we considered the distribution of phenotypes from all RILs carrying the founder haplotype at the LOD peak position (genotype probability >0.95%) [24]. QTL were then phased into allelic classes by identifying the minimal number of partitions of founder haplotypes that describe phenotypic variation associated with the QTL peak, as described previously [14,24]. Fertility assays. Virgin female offspring from dysgenic crosses between isogenic lines carrying fertile_B81/B82 (21077, 21154) and sterile_B62/B63 (21218, 21156) alleles and Harwich males were collected daily and individually placed in a vial containing two Canton-S males. Females were allowed to mate for 5 days and were transferred to a new vial for another 5 days after which the parents were discarded. The presence and total number of F2 individuals were counted from the two vials. Selection of paired RILs with alternate QTL alleles. We identified background matched RILs containing either the B6 (sterile) or B4 (fertile) haplotypes from the start position of the QTL-21d confidence interval (2L: 19,010,000) to the end position of QTL-3d confidence interval (2R: 6,942,495) (P > 0.9), based on their published HMM genotypes [24]. For all possible RIL pairs (B6 and B4), we then calculated the number of 10 Kb genomic windows in which they carried the same RIL haplotype (P > 0.9). We selected three pairs of RILs, which carry the same founder genotype for 47% (21213 & 21183), 46% (21147 & 21346) and 44% (21291 & 21188) of genomic windows outside of the QTL. Small RNA-seq and total RNA-seq. RILs were maintained at 25°C, and three biological replicates of 20 ovaries were dissected from 3–5 day-old females. Ovaries were homogenized in TRIzol (invitrogen) and stored at -80°C until RNA extraction. 50 μg of total RNA from each of 18 biological samples (3 biological replicates x 3 pairs) was size fractionated in a 15% denaturing polyacrylamide gel and the 18–30 nt band was excised. 2S-depleted small RNA libraries for Illumina sequencing were then constructed according to the method of Wickersheim and Blumenstiel [99]. Ovarian small RNA libraries were published previously SRP160954, [100]. Ribodepleted and stranded total RNA libraries were generated from the same ovarian samples using NuGen total RNA kit (TECAN). All 18 small RNA and total RNA libraries were sequenced on an Illumina Nextseq 500 at the University of Houston Seq-N-Edit Core. Small-RNA analysis. Sequenced small RNAs were separated based on size into miRNAs/siRNAs (18-22nt) and piRNAs (23-30nt) [11]. Reads corresponding to contaminating rRNAs, including 2S-rRNA, were removed from each library by aligning to annotated transcripts from flybase [101]. To determine the piRNA cluster activity we first uniquely aligned the piRNAs to reference genome (dm6 [29]) using Bowtie1 (-v 1 -m 1) [102]. We then used a customized perl script to count reads that mapped to a set of previously annotated piRNA clusters from the same genotypes (497 piRNA clusters, [103]). Read counts normalized to total mapped microRNAs for each library were used to infer differential expression using DESeq2 [104]. Sliding window estimates of piRNA abundance (Fig 2C and 2D) were calculated using bedtools genomecov [105], normalizing the read counts to total mapped miRNA reads. Total RNA analysis. Residual ribosomal RNAs (rRNAs) were identified in ribo depleted libraries based on alignment to annotated rRNAs from flybase [101], and excluded from further analysis. Retained reads aligned to the library of consensus satellite and TE sequences from repbase [106], plus additional satellite consensus sequences from Larracuente [107]. For TE expression, the total reads mapped to TE sequences were counted using unix commands (uniq -c). Remaining reads that failed to map were pseudoaligned to D. melanogaster transcriptome (dm6/BDGP6) using Kallisto with default parameters [108]. Differentially expressed TEs and genes were identified from a combined analysis in DESeq2 [104]. Genes and TEs with base mean > = 100, Adjusted P-value < = 0.05 and whose expression pattern differed (fold change > = 1.5) were considered differentially expressed between the B6 and B4 QTL haplotype. Radiation sensitivity. Third instar larvae were either mock treated or irradiated in a Rad Source RS 1800 X-ray machine set at 12.5 mA and 160 kV. To obtain 3rd instar larvae, embryos were collected for 24 hr and aged for 5 days at 25°C. The food vials containing larvae were then X-ray irradiated at doses from 5–80 Gray after which an optimal dose that clearly depicts the phenotypic difference was selected. Survival to adulthood was determined by scoring the number of empty and full pupal cases at 10 days after radiation. Identification of novel mutations in RIL genomes. 10 females from each of 795 RILs were deposited into a well of a 96 well plate (Axygen, P-96-450R-C) on ice. DNA isolation was then executed in plates using the Gentra Puregene Cell Kit (Qiagen, 158788) using extensions of the manufacturer’s protocol. Subsequently, DNA was further purified (Qiagen, QIAquick 96 PCR Purification kit, 28183), quantified via a fluorometer (ThermoFisher, Qubit dsDNA HS kit, Q32854), and diluted to 2-ng/μl. Each RIL sample was then subjected to the MSG (Multiplexed Shotgun Genotyping) approach developed by Andolfatto et al. [76], which is a form of RADseq (restriction-site associated DNA sequencing, Baird et al. [109]). Starting with 10-ng DNA, samples were first digested using MseI (New England Biolabs, R0525L). This enzyme has the restriction site T/TAA, cuts frequently along the genome, and unlike traditional rare-cutter RADseq strategies, yields sequencing reads spread somewhat evenly along the genome. Next, plate well-specific barcoded adapters are independently ligated onto the cut ends of each DNA sample. Fragmented, barcoded samples from a given plate are then mixed, and the 96-plex pool precipitated, purified, and size-selected to 250-300-bp via a Pippin Prep (Sage Science). Each multiplexed sample is then PCR amplified, during which a DNA plate-specific Illumina-compatible index is incorporated, and then purified. Finally, each of the independently-indexed 96-plex pools are quantified, mixed at equal concentration, and sequenced over multiple lanes of an Illumina HiSeq 2500 on “high output” mode, yielding single-end 100-bp reads (KU Genome Sequencing Core). With this MSG approach, reads from each RIL are computationally distinguished by both an Illumina index sequence (which marks the plates), and an “in line barcode” (the first 6 Read1 bases, which marks samples on any given plate). SNP and indel variants were identified from MSG short-read data following GATK best practices for sample groups [110–112]. Briefly, Cutadapt (version 3.5; [113]) was used to de-multiplex samples and trim adaptors, while alignment to the D. melanogaster reference genome (dm6 [29]) were performed using BWA (version 0.7.17-GCC-10.2.0;[114]). The resulting BAM files were sorted and indexed using Samtools (Li et al., 2009 [114]). Individual GVCF files were generated using HaplotypeCaller and then joint-genotyped using Genotype_GVCFs. Both SNPs and indels were extracted and filtered out following the GATK Best Practices hard filters using VariantFiltration (SNP: "QD < 2.0 || FS > 60.0 || MQ < 40.0 || MQRankSum < -12.5”; indel “"QD < 2.0 || FS > 200.0”, and converted to TSV in R for further analysis using the vcfR package [115]. To identify novel base substitutions that arose in RILs and not present in the founders, we filtered out all alleles with spanning deletions as well as annotated SNPs from the founder lines (https://wfitch.bio.uci.edu/~dspr/Data/index.html). Since founder SNPs were called in dm5, we converted their coordinates to dm6 using the NCBI Genome Remapping Service (https://www.ncbi.nlm.nih.gov/genome/tools/remap) before filtering. For indels, short indels were not called from the original founder sequence data. We therefore considered an indel to be novel if they were unique to a RIL and sequenced among at least 50 RILs. Differences in the number of SNPs and indels rate was modeled using linear mixed model fitted with lmer function in the R package lme4 (Bates et al., 2013 [96]). Three models were compared, a null model (mutations ~ plate + depth), a founder allele model (mutations ~ plate + depth + founder allele), and a dysgenic sterility allele model (mutations ~ plate + depth + sterile/fertile). Founder or dysgenic sterility allele referred to the QTL-3d peak: (B1-B5,B6-B8) or sterile/fertile (B6/all other founders). Note that B5 alleles are not present among the RILs at the QTL-3d peak. The models were compared using a likelihood ratio test to determine whether founder allele or allelic class explained variation in the number of novel SNPs or indels between RILs. The effect of sterile alleles on SNP and indel number was evaluated by t-test. qRT-PCR of P-element transcripts 3 biological replicates including 20 pairs of 3–5 day F1 dysgenic ovaries from crosses between fertile (B81, B82) or sterile (B62, B63) females and Harwich males were dissected and homogenized in TRIzol. Crosses were maintained at 22°C. RNA was treated with DNA-free (ThermoFisher) and reverse transcribed using oligo-dT primers and superscript IV (ThermoFisher) according to manufacturer instructions. Three different primer sets were used to amplify the 3’ end of all P-element transcripts, IVS3 spliced transcripts and IVS3 unspliced transcripts, as well as rpl32. Transcripts were amplified and quantified in three technical replicates using power-SYBR green (ThermoFisher) and normalized to rpl32 for the same sample. Primer sequences were as follows: rpl32-F: 5’-CCGCTTCAAGGGACAGTATC, rpl32-R: 5’-GACAATCTCCTTGCGCTTCT, P-element-all F: 5’-CACCGAAATGGATGAGTTGACG, P-element-all R: 5’-TAATAAGTCCGCCGTGAGACAC, P-element IVS3 spliced F: 5’-AATAGCCAGGAATACAGAAATG, P-element IVS3 spliced R: 5’-AACATTTCTGTATTCCTGGCTA, P-element IVS3-unspliced F: 5’-GACAAAACACAATAGACAGCACA, P-element IVS3-unspliced R: 5’-TGTGCTGTCTATTGTGTTTTGTC. Identification of in-phase polymorphisms. The SNP data of B founders that used to infer in-phase SNPs is based on dm3 [24]. To identify in-phase SNPs we looked for alternate SNP alleles that match the predicted phenotypic class for each of the QTL peaks. For QTL-21d we used the criteria: sterile class (B2, B6) and the fertile class (B1, B3, B4, B7, B8), whereas for QTL-3d: sterile class (B6) and the fertile class (B1, B2, B3, B4, B7, B8). Identification of in-phase polymorphisms. The SNP data of B founders that used to infer in-phase SNPs is based on dm3 [24]. To identify in-phase SNPs we looked for alternate SNP alleles that match the predicted phenotypic class for each of the QTL peaks. For QTL-21d we used the criteria: sterile class (B2, B6) and the fertile class (B1, B3, B4, B7, B8), whereas for QTL-3d: sterile class (B6) and the fertile class (B1, B2, B3, B4, B7, B8). Supporting information S1 Fig. Sterility is associated with increased expression of pericentromeric genes in the head. a) Mean expression of genes located in the pericentromere, euchromatin, telomere and the fourth chromosome from RILs carrying each of the eight B founder genotypes at the QTL-3d region. Error bars represent the standard deviation among mean expression levels of different genes. The sterile/B6 (light green) shows high pericentromeric gene expression compared to the fertile strains (dark green) (Anova; F6,494 = 7.775, P < 5.24e-08). The letters indicate significantly different expression levels based on Tukey-HSD comparisons between RILs with different founder alleles. https://doi.org/10.1371/journal.pgen.1010080.s001 (PDF) S2 Fig. Expression profile of QTL piRNA clusters in sterile and fertile RIL pair 2. The piRNA expression between sterile and fertile genotypes from the 21188–21291 RIL pair along the two QTL piRNA clusters: 2L:23,328,000–23,337,026 and 2L:23,222,004–23,246,024, respectively. Only uniquely mapping piRNAs are considered. The TE families at the top of each panel are represented by different colors. TE-others represent the repeat families coming from sibling species of D. melanogaster. Positive value indicates piRNAs mapped to the sense strand of the reference genome and negative value indicates those from the antisense strand. The piRNA cluster expression levels are estimated by log2 scale transformed of reads per million mapped reads [log2(RPM+1)]. https://doi.org/10.1371/journal.pgen.1010080.s002 (PDF) S3 Fig. Expression profile of QTL piRNA clusters in sterile and fertile RIL pair 3. The piRNA expression between sterile and fertile genotypes from the 21346–21147 RIL pair along the two QTL piRNA clusters: 2L:23,328,000–23,337,026 and 2L:23,222,004–23,246,024, respectively. Only uniquely mapping piRNAs are considered. The TE families at the top of each Fig are represented by different colors. TE-others represent the repeat families coming from sibling species of D. melanogaster. Positive value indicates piRNAs mapped to the sense strand of the reference genome and negative value indicates those from the antisense strand. The piRNA cluster expression levels are estimated by log2 scale transformed of reads per million mapped reads [log2(RPM+1)]. https://doi.org/10.1371/journal.pgen.1010080.s003 (PDF) S4 Fig. Sterility among 3 and 21 day females based on QTL haplotype. Four haplotypes are compared, which comprise all possible combinations of sterility alleles at 2 QTL. The allele at the 3 day QTL is indicated first and is represented by the color of the violin plot (light green = sterile, dark green = fertile). The allele at the 21 day QTL is indicated second and represented by the color of the points on the scatter plot. Y-axis is residual variation in F1 atrophy after accounting for student experimenter and block. Among 3 day old females, haplotypes containing different alleles for the 3 day old QTL are significantly different from each other (Tukey HSD P = 0.016–0). However, haplotypes containing alternative QTL for the 21d only do not differ from each other (Tukey HSD P>0.74). This suggests phenotypic variation in 3 day old females is not influenced by their genotype at the 21 day QTL. In contrast, among 21 day old females tolerant alleles in both QTL loci are required to significantly decrease sterility below the sterile allele containing haplotypes (Tukey HSD P = 0.01–0). https://doi.org/10.1371/journal.pgen.1010080.s004 (PDF) S5 Fig. Crossing scheme to generate sterile and fertile alleles. https://doi.org/10.1371/journal.pgen.1010080.s005 (PDF) S6 Fig. Increased expression of genes upregulated in late-stage egg chambers in fertile ovaries. Upregulation in stage 9–10 and stage 12–14 egg chambers is from Tootle et al. [30]. Genes are separated into eggshell components (top) and non-eggshell components (bottom). Dark green bars indicate genes significantly upregulated in fertile genotypes whereas light green indicates genes upregulated in sterile genotypes. https://doi.org/10.1371/journal.pgen.1010080.s006 (PDF) S1 Table. Provided are the proportion of atrophy for 3-day old F1 females when recombinant inbred lines were crossed to Harwich males. https://doi.org/10.1371/journal.pgen.1010080.s007 (XLSX) S2 Table. Provided are the proportion of atrophy for 21-day old F1 females when recombinant inbred lines were crossed to Harwich males. https://doi.org/10.1371/journal.pgen.1010080.s008 (XLSX) S3 Table. Residuals from 3-day-old F1 females used for QTL mapping. https://doi.org/10.1371/journal.pgen.1010080.s009 (XLSX) S4 Table. Residuals from 21-day-old F1 females used for QTL mapping. https://doi.org/10.1371/journal.pgen.1010080.s010 (XLSX) S5 Table. Results of DESeq2 analysis of differential gene expression between sterile and fertile alleles. https://doi.org/10.1371/journal.pgen.1010080.s011 (XLS) S6 Table. List of differential expressed of Tip60 members and one of its interactors. https://doi.org/10.1371/journal.pgen.1010080.s012 (XLS) S7 Table. List of genes upregulated in sterile and fertile ovaries, as well as associated enriched GO terms. https://doi.org/10.1371/journal.pgen.1010080.s013 (XLS) S8 Table. Analysis of piRNA cluster expression and abundance of P and I element derived piRNAs in sterile and fertile ovaries. https://doi.org/10.1371/journal.pgen.1010080.s014 (XLS) S9 Table. TE composition of differentially active piRNA clusters in QTL-3d. https://doi.org/10.1371/journal.pgen.1010080.s015 (XLSX) S10 Table. Results of DESeq2 analysis of differential TE expression between sterile and fertile alleles. https://doi.org/10.1371/journal.pgen.1010080.s016 (XLS) S11 Table. In phase polymorphisms in QTL-3d and QTL-21d. https://doi.org/10.1371/journal.pgen.1010080.s017 (XLSX) S12 Table. List of candidate genes that are differentially expressed between sterile and fertile alleles and contain non-coding in-phase SNPs. https://doi.org/10.1371/journal.pgen.1010080.s018 (XLSX) S13 Table. List of candidate genes that are contain non-synonymous in-phase SNPs. https://doi.org/10.1371/journal.pgen.1010080.s019 (XLSX) S14 Table. PCA analysis of gene expression data of background-matched recombinant inbred lines. https://doi.org/10.1371/journal.pgen.1010080.s020 (XLS) S15 Table. PCA analysis of piRNA cluster expression data of background-matched recombinant inbred lines. https://doi.org/10.1371/journal.pgen.1010080.s021 (XLSX) S16 Table. Uniquely mapping read coverage of piRNAs from sterile and fertile ovaries for both plus and minus strand of differentially active piRNA clusters in QTL-3d. https://doi.org/10.1371/journal.pgen.1010080.s022 (XLSX) S17 Table. Incidence of ovarian atrophy among in F1 females from crosses between isogenic Fertile/Sterile lines and Harwich males. https://doi.org/10.1371/journal.pgen.1010080.s023 (XLSX) S18 Table. Fertility of F1 females from crosses between isogenic tolerant/sensitive lines and Harwich males. https://doi.org/10.1371/journal.pgen.1010080.s024 (XLS) S19 Table. Radiation sensitive of sterile and fertile alleles and controls. https://doi.org/10.1371/journal.pgen.1010080.s025 (XLSX) S20 Table. Number of novel SNPs and indels in population B RIL genomes. https://doi.org/10.1371/journal.pgen.1010080.s026 (XLSX) S21 Table. qPCR estimates of P-element derived mRNA abundance in dysgenic ovarian RNA. https://doi.org/10.1371/journal.pgen.1010080.s027 (XLSX)

Kabakci, Zeynep;Reichle, Heidi E.;Lemke, Bianca;Rousova, Dorota;Gupta, Samir;Weber, Joe;Schleiffer, Alexander;Weir, John R.;Lehner, Christian F.

doi: 10.1371/journal.pgen.1010547pmid: 36480577

Introduction Regular chromosome transmission during mitotic and meiotic divisions depends on sister chromatid cohesion. The corresponding ties that keep sister chromatids paired are formed already during S phase, concomitant with chromosome replication. They consist primarily of cohesin, a protein complex based on three core subunits, SMC1, SMC3 and an α-kleisin [1]. The latter provides binding sites for additional cohesin subunits that are known as HAWK proteins (HEAT repeat proteins Associated With Kleisins) [1,2]. One of the HAWKs, a member of the stromalin protein family, which includes Scc3 of budding yeast, Drosophila SA-1 and the human proteins STAG1-3 (also designated as SA1-3), is permanently bound. Moreover, Pds5 or Scc2/Nipped-B/NIPBL proteins are recruited interchangeably by another binding site on α-kleisin. Cohesin might mediate sister chromatid cohesion in a topological manner, by forming a proteinaceous ring around the sister chromatid pair [1]. Thereby, this topological embrace also occasions an effective way of releasing sister chromatid cohesion by proteolytic opening of the ring. Evidently, after biorientation of all chromosomes within the spindle, cohesion between sister chromatids needs to be severed during mitosis, as well as during the second meiotic division (M II), to permit the segregation of sister centromeres to opposite spindle poles. The final release of sister chromatid cohesion is known to be promoted by the endoprotease separase, which cleaves specifically the α-kleisin subunit of cohesin after separase activation at the metaphase-to-anaphase transition [3–6]. Sister chromatid cohesion is also essential for the regular segregation of homologous chromosomes during the first meiotic division (M I). In combination with crossovers (COs), cohesion of sister chromatids in chromosome arm regions on the telomeric side of COs precludes CO terminalization. Therefore, the combination of COs and distal sister chromatid cohesion maintains paired homologs as bivalent chromosomes during canonical meiosis [7–11]. After biorientation of all the bivalents in the M I spindle, homolog separation at the metaphase-to-anaphase transition is also induced by separase [9–11], similar as in mitosis and M II. However, α-kleisin cleavage by separase during M I is spatially controlled [11,12]. While α-kleisin of cohesin on chromosome arms is cleaved during M I, it is protected from cleavage by separase within pericentromeric regions. The spatial control of cohesin elimination during M I depends on the expression of Rec8-type, meiosis-specific α-kleisins and their spatially differentiated phosphorylation, while the Scc1/Rad21 proteins function as α-kleisins during mitosis. Strikingly, male meiosis in higher dipteran species including Drosophila melanogaster is achiasmate [13,14]. Neither the formation of a synaptonemal complex, which mediates synapsis of homologous chromosomes all along their length during canonical meiosis, nor meiotic recombination and CO formation occur. Nevertheless, homologous chromosomes are also paired into bivalents before separation to opposite spindle poles during anaphase of M I. The initial pairing of homologous chromosomes in spermatocytes might be driven by the same mechanisms that mediate the wide-spread pairing of homologous chromosomes in somatic cell types of D. melanogaster [15,16]. However, somatic pairing appears to be disrupted by chromosome condensation at the onset of mitotic divisions [17,18]. In contrast, in spermatocytes, homolog pairing is maintained until the onset of anaphase I by a special conjunction system that functions as a replacement of COs [13]. Four genes have been identified as specifically required for this alternative homolog conjunction (AHC) [19–22]. Null mutations in these AHC genes result in premature separation of bivalents into univalents and random segregation of the univalents during M I in males, but not in females where meiosis is canonical. The AHC gene teflon (tef) is required predominantly for normal formation and segregation of autosomal bivalents [19]. In contrast, the three additional AHC genes, stromalin in meiosis (snm) (or also SA-2), modifier of mdg in meiosis (mnm) and univalents only (uno), are equally important for autosomal bivalents and the sex chromosome bivalent [20,22]. The molecular details of how the AHC proteins contribute to maintenance of chromosome conjunction in bivalents are still poorly understood. TEF appears to function in early spermatocytes, during establishment of AHC [21,23]. In late spermatocytes, TEF is no longer detectable [23]. Thus, it does not appear to be a component of the chromosomal glue that mediates AHC until anaphase I. TEF contains three zinc fingers [21]. It can bind to chromosomes and recruit MNM, which binds directly to TEF [23]. In contrast to TEF, MNM as well as the additional AHC proteins SNM and UNO are presumably part of the chromosome conjunction glue. SNM, UNO and MNM, abbreviated as SUM in the following, are co-localized on bivalents [20,22]. Their presence is particularly prominent on the sex chromosome bivalent. While the sex chromosomes of D. melanogaster do not share any extended euchromatic homology, they both contain rDNA repeats embedded in centromere-proximal heterochromatin, and these rDNA loci serve as pairing centers during male meiosis [24]. In particular, a 240 bp repeat normally located within the intergenic spacers of the rDNA repeats was demonstrated to be sufficient for chromosome conjunction [20,24–26]. The SUM proteins are associated with the 240 bp repeats in sub-nucleolar foci in mid to late spermatocytes [20]. During chromosome condensation at the onset of the first meiotic division (M I), the sub-nucleolar SUM foci coalesce into a single large dot at the XY chromosome (chr) pairing site [20,22]. On autosomal bivalents, the SUM proteins are present in much smaller dots that are not necessarily always detectable above background [20,22,23,27]. The number of SUM dots on autosomal bivalents appears to be restricted to around one or two per bivalent at the start of M I [23,27]. Their chromosomal localization is presumably distinct in each spermatocyte, as genetic analyses have revealed a wide-spread distribution of meiotic conjunction potential throughout the euchromatic regions rather than dedicated invariant autosomal conjunction sites [28,29]. Possibly, therefore, AHC is established during spermatocyte maturation in a rather randomly chosen restricted chromosomal region in case of autosomal bivalents, analogous to the placement of COs during canonical meiosis [27]. Strong support for the notion that the SUM protein dots on the sex chromosome and autosomal bivalents represent the physical conjunction between chromosomes is their precipitous, separase-dependent disappearance at the onset of anaphase I [20,22,30]. Mutational elimination of a separase cleavage site motif in UNO was shown to block the disappearance of the chromosomal SUM protein dots as well as homolog separation during anaphase I [22]. For further clarification of the molecular mechanisms whereby the SUM proteins, which do not include well-known bona fide DNA-binding motifs, might achieve physical conjunction of chromosomes, we have initiated biochemical analyses of their interaction domains and their DNA-binding. Our findings reveal that co-option of the cohesin-related proteins SNM and UNO for AHC has conferred the ability to regulate chromosome conjunction by separase activity, as well as an ability to bind DNA. However, our observations indicate that SUM proteins do not conjoin chromosomes in a topological manner akin to cohesin. Rather, effective chromosome conjunction appears to be achieved by tight multivalent DNA-binding resulting from multimerization via both UNO and MNM. Results The C-terminal region of UNO that binds to SNM is derived from α-kleisins UNO homologs can be detected exclusively within Diptera (suborder Brachycera). They have conserved N- and C-terminal domains separated by a putative disordered linker region (Fig 1A). UNO of D. melanogaster was identified originally because of co-purification with both SNM-EGFP and MNM-EGFP from testis extracts [22]. To assess whether SNM or MNM might bind directly to UNO, we performed co-immunoprecipitation experiments after transient expression of tagged proteins in Drosophila S2R+ cells. In this cell line, endogenous UNO expression occurs at most at very low levels according to RNA-seq data, while SNM mRNA is not detectable [31,32]. Similarly, by immunoblotting, we failed to detect endogenously expressed UNO and SNM in S2R+ cells (S1 Fig). In case of MNM, which is but one of many distinct isoforms expressed from the complex mod(mdg4) locus [20,33], RNA-seq data is consistent with endogenous expression in S2R+ cells at low levels [31,32]. As antibodies specific for MNM are not available, the potential presence of endogenous MNM protein in S2R+ cells remains to be confirmed. However, immunoblotting with anti-M_CP, an antibody that reacts with the N-terminal common part (CP) present in all the different Mod(mdg4) isoforms [34], demonstrated that endogenous MNM, if present, is at a level far lower than that generated from transfected expression plasmids (S1 Fig). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. SNM and UNO form a complex that binds MNM. (A) Structure of UNO and interactions with AHC proteins. UNO includes two regions with stronger sequence conservation close to the N- and C-termini, as well as a highly conserved match to the separase cleavage-site consensus. Regional extent of amino acid identity (%) among drosophilid UNO orthologs is indicated with grey shading. Protein interactions of SNM, MNM and full-length UNO, as well as the UNO fragments UNO_N, UNO_M and UNO_C were analyzed by co-immunoprecipitation experiments after transient expression in S2R+ cells. The main findings are summarized schematically (with reference to relevant figure panels in brackets). (B-F) Co-immunoprecipitation after transient expression of the indicated proteins was analyzed by immunoblotting. Protein tags were EGFP (E), mCherry (C) or a myc epitope (myc). Tetracycline repressor fused to a nuclear localization signal and EGFP (TetR-E) was used for control experiments. Presence or absence of proteins in the extracts used for immunoprecipitation (input) or in the samples immunoprecipitated with the indicated antibodies (IP) were analyzed with anti-EGFP (anti-E), anti-mCherry (anti-C), anti-myc (anti-myc) and anti-SNM. A non-specific band (*) recognized by anti-E and the positions of molecular weight markers are indicated on the right side, as well as the bands representing proteins of interest. https://doi.org/10.1371/journal.pgen.1010547.g001 After transfection of S2R+ cells with plasmids for co-expression of SNM-mCherry and UNO-EGFP, we observed efficient co-immunoprecipitation of these two proteins (Fig 1B). In contrast, co-immunoprecipitation of MNM-mCherry and UNO-EGFP could not be observed in an analogous experiment (Fig 1C). According to our initial bioinformatic analyses with standard BLAST searches [22], the predicted amino acid (aa) sequence of UNO failed to display significant similarities to proteins with known functions. However, because of the apparent direct binding of UNO to SNM, a member of the stromalin family of α-kleisin-binding proteins, and because UNO was previously shown to include a functionally essential cleavage site cut by separase (Fig 1A) [22], which cleaves primarily α-kleisins, we focused specifically on detecting α-kleisin similarities within UNO. Interestingly, a region close to the C-terminus of UNO (aa 289–364) was found to be similar to an internal α-kleisins region (aa 320–394 of human Rad21) (Figs 1A and S2). This conserved internal α-kleisin region is known to bind stromalin/SA/STAG proteins [35,36]. We conclude that UNO includes a region that was co-opted from an α-kleisin. This region of UNO comprises a separase cleavage site and the C-terminal stromalin-binding region. Co-immunoprecipitation experiments clearly confirmed that a C-terminal fragment of UNO (UNO_C, aa 241–417) mediates binding of UNO to SNM-mCherry (Fig 1A and 1D). The N-terminal region of UNO (UNO_N, aa 1–78) and the middle part (UNO_M, aa 79–240) did not bind to SNM (Fig 1A and 1D). We conclude that SNM and UNO appear to form a complex, which will be designated as SU complex in the following. The co-immunoprecipitation experiments also indicated that SU complex formation was accompanied by mutual stabilization of the interacting proteins. After transient co-expression, the resulting levels of UNO and SNM were around sixfold higher compared to those detected after individual expression (S3 Fig). Analyses with UNO subregions indicated that UNO_C stability was highly dependent on the presence of SNM, while UNO_N and UNO_M appeared to be more stable and not dependent on SNM (S3 Fig). The N-terminal region of UNO mediates self-association Other characteristic hallmarks of α-kleisins, the N- and C-terminal domains, which bind to the cohesin core subunits SMC3 and SMC1, respectively, are absent in UNO. UNO does not extend C-terminally beyond the SNM-binding region. At the N-terminus, UNO has a domain that is distinct from those present in α-kleisins. The N-terminal domain of UNO (UNO_N, aa 1–78) is predicted to form β-strands, while the N-terminal SMC3-binding region of α-kleisins is α-helical. Moreover, a region with high sequence similarity to UNO_N is present in a poorly characterized Drosophila protein (CG32117) that has no similarity to α-kleisins [22]. Interestingly, our co-immunoprecipitation experiments demonstrated that (UNO_N) mediates self-association. After co-expression of UNO_N-EGFP and UNO_N-mCherry, we observed their efficient co-immunoprecipitation (Fig 1A and 1E). UNO_N-mCherry was also co-immunoprecipitated by full-length UNO-EGFP along with SNM (Fig 1E). In contrast, only SNM but not UNO_N-mCherry was co-immunoprecipitated by UNO_C-EGFP (Fig 1E). These results suggest that the distinctive N-terminal domain of UNO (UNO_N) mediates self-association, even when SNM is bound to UNO. SNM does not appear to have the ability to self-associate, in contrast to UNO. We did not detect co-immunoprecipitation of SNM-EGFP and SNM-mCherry (Fig 1A and 1B). A complex of SNM and UNO binds MNM The lack of co-immunoprecipitation of MNM-mCherry and UNO-EGFP (Fig 1C) suggested that the co-purification of MNM-EGFP and UNO from Drosophila testis extracts [22] was not the result of a direct interaction between these two proteins. Thus, we considered the possibility that SNM might function as a bridging factor, by binding not only to UNO, but perhaps also to MNM. However, SNM-mCherry binding to MNM-EGFP could not be detected in co-immunoprecipitation experiments (Fig 1A and 1B). Similarly, untagged SNM was not co-immunoprecipitated by MNM-mCherry (Fig 1C). However, when the three proteins SNM, UNO-EGFP and MNM-mCherry were co-expressed, we clearly detected their co-immunoprecipitation (Fig 1C). Moreover, we also observed efficient co-immunoprecipitation of the C-terminal domain of UNO (UNO_C-EGFP) with SNM-mCherry and MNM-myc (Fig 1F). In conclusion, the SU complex formed by SNM and UNO can bind MNM, while individually SNM and UNO do not appear to have this ability (Fig 1A). The complex formed by the three proteins SNM, UNO and MNM will be designated as SUM complex in the following. MNM self-association permits binding of TEF to the SUM complex MNM is only one of more than thirty distinct protein isoforms expressed by mod(mdg4) [20,34]. Mod(mdg4) products, including MNM, share an N-terminal common part (CP) followed by an isoform-specific C-terminal region (Fig 2A). To address whether SU binds only MNM or also other Mod(mdg4) proteins, we performed additional co-immunoprecipitation experiments. The Mod(mdg4) isoforms T, P and C were analyzed in these experiments. The T isoform (also designated as 67.2) represents the most extensively characterized Mod(mdg4) product, which functions along with Su(Hw) and CP190 in the gypsy insulator for example [37]. The C and P isoforms might be expressed in testis according to RNA-seq data [31]. In addition, we used a construct for expression of the shared N-terminal common part (CP) of the Mod(mdg4) proteins. Our experiments revealed that none of the analyzed Mod(mdg4) variants (C, P, T and CP, all with C-terminal mCherry) were able to co-immunoprecipitate SNM-EGFP and UNO-myc, in contrast to MNM-mCherry (Fig 2B). We conclude that the isoform-specific C-terminal region of MNM (MNM_C) is required for binding to the SU complex. Additional experiments confirmed that MNM_C fused to mCherry (MNM_C-mCherry) is co-immunoprecipitated by UNO-EGFP after co-expression with SNM (Fig 2G). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. MNM self-association is required for formation of SUMT complexes containing SNM, UNO, MNM and TEF. (A) Structure of MNM and interactions with AHC proteins. MNM has an N-terminal (Mod(mdg4)_CP) that is also present in other Mod(mdg4) isoforms (including _T, _P and _C). This part includes a BTB domain. In addition, MNM has a C-terminal MNM-specific region. Protein interactions were analyzed by co-immunoprecipitation experiments after transient expression in S2R+ cells. The main findings are summarized schematically (with reference to relevant figure panels in brackets). (B-G) Co-immunoprecipitation after transient expression of the indicated proteins was analyzed by immunoblotting. Protein tags were EGFP (E), mCherry (C) or a myc epitope (myc). Tetracycline repressor fused to a nuclear localization signal and EGFP (TetR-E) or a C-terminal fragment of Mad1 tagged with a myc-epitope (myc-Mad1_C) was used for control experiments. Presence or absence of proteins in the extracts used for immunoprecipitation (input) or in the samples (IP) immunoprecipitated with the indicated antibodies were analyzed with anti-EGFP (anti-E), anti-mCherry (anti-C), anti-myc (anti-myc), and anti-Mod(mdg4)_CP (anti-M_CP). The positions of molecular weight markers are indicated on the right side, as well as the bands representing the indicated proteins of interest. (B) While MNM-mCherry (MNM-C) co-immunoprecipitated SNM-EGFP (SNM-E) and UNO-myc (UNO-myc), other Mod(mdg4) isoforms (T, P and C) as well as the common part (CP) tagged with mCherry did not co-immunoprecipitate SNM-EGFP and UNO-myc. (C) The mCherry fusions of Mod(mdg4) variants (CP, T, P and C) all co-immunoprecipitated MNM-EGFP (MNM-E) specifically, as demonstrated by first two lanes, which present a positive and negative control experiment, respectively. (D) MNM-E co-immunoprecipitated MNM-myc specifically, as demonstrated by the negative control experiment with TetR-E. (E) The C-terminal MNM-specific region as an mCherry fusion (MNM_C-C) was not co-immunoprecipitated by MNM-E in contrast to full length MNM-mCherry (MNM-C). (F) UNO-E co-immunoprecipitated SNM-mCherry, MNM and TEF-myc specifically [23], as demonstrated by the negative control experiment with TetR-E. (G) UNO_C-E co-immunoprecipitated only SNM-mCherry, MNM_C-C but not TEF-myc, even though TEF-myc was observed to co-immunoprecipitate MNM_C-C, indicating that MNM_C can bind in a mutually exclusive manner to either the SNM-UNO complex or to TEF. https://doi.org/10.1371/journal.pgen.1010547.g002 The N-terminal common part of the Mod(mdg4) proteins includes a BTB/POZ domain, which is present in a wide range of proteins with distinct functions (Fig 2A). BTB/POZ domains appear to mediate protein-protein interactions, including dimerization, tetramerization and multimerization [38,39]. Therefore, we evaluated whether MNM self-associates via the N-terminal common part (CP) and hence also with Mod(mdg4) proteins other than MNM. Indeed, MNM-EGFP was readily co-immunoprecipitated with all the analyzed Mod(mdg4) variants C, P, T and CP fused to mCherry (Fig 2C). Similarly, we detected co-immunoprecipitation of MNM-EGFP and MNM-myc (Fig 2D). We conclude that MNM has the potential to associate with itself and with other Mod(mdg4) proteins via the N-terminal common part. The isoform-specific C-terminal region of MNM fused to mCherry (MNM_C-mCherry) was not observed to co-immunoprecipitate with MNM-EGFP (Fig 2E). The isoform-specific C-terminal region of MNM (MNM_C) does not only bind to the SU complex (Fig 2A and 2B) but also to the N-terminal region of TEF [23]. Moreover, consistent with MNM’s ability to bind to both the SU complex and to TEF, all four AHC proteins (SNM, UNO, MNM and TEF) are co-immunoprecipitated after co-expression in S2R+ cells (Fig 2F) [23]. This apparent SUMT complex could arise if the binding of SU and TEF to MNM was not mutually exclusive. However, even if mutually exclusive, SUMT complex formation might still succeed, because of the self-association of MNM. For example, one of two associated MNM proteins might bind to SU and the other to TEF. To address the mode of TEF binding in the SUMT complex, we co-expressed TEF-myc with MNM_C-mCherry, SNM and UNO_C-EGFP (Fig 2G). As shown above, MNM_C cannot self-associate, but it can bind to TEF and also to the SU complex. UNO_C cannot self-associate, but it can bind to SNM and recruit MNM_C. As expected, UNO_C-EGFP was observed to co-immunoprecipitate MNM_C-mCherry after co-expression of the four proteins (TEF-myc, MNM_C-mCherry, SNM and UNO_C-EGFP), but TEF-myc was not co-immunoprecipitated (Fig 2G). Thus, after binding to the SU_C-EGFP complex, MNM_C-mCherry was no longer able to bind TEF-myc. This result suggests that the binding of TEF and SU to MNM is mutually exclusive. Accordingly, formation of SUMT assemblies, which contain all four known AHC proteins, depends on self-association of either MNM or possibly also of UNO. Purification and structural analysis of a complex of SNM and UNO_C For further characterization of the AHC protein-protein interactions, we succeeded in purifying recombinant versions of some of these proteins for analysis in vitro. To characterize the SU complex, we expressed full-length SNM and UNO_C (aa 281–417) in baculovirus-infected insect cells. To promote solubility, SNM was fused N-terminally with a maltose-binding protein (MBP). UNO_C was fused to an N-terminal twin Strep-II affinity tag. Using streptactin affinity chromatography followed by size exclusion chromatography, we purified an MBP-SNM/Strep-II-UNO_C complex to homogeneity (from here on referred to as SU_C) (Fig 3A). We next determined the stoichiometry of SU_C using both mass photometry (MP) and multi-angle light scattering coupled to size exclusion chromatography (SEC-MALS). MP and SEC-MALS measured a molecular mass of 178 and 211.9 kDa, respectively (Fig 3B and 3C). As the theoretical molecular mass of a 1:1 dimer was 173.4 kDa, these results provided strong evidence of a 1:1 stoichiometry. Based on this stoichiometry, we generated a de novo structure prediction using AlphaFold 2.2.0 Multimer (AF2) [40,41] (Fig 3D). In order to test the AF2 model, we made use of bifunctional chemical cross-linking coupled to mass spectrometry (XL-MS). The observed cross-links were mapped onto the 2D representation of the complex and onto the 3D model (Fig 3E). Given that Disuccinimidyl Dibutyric Urea (DSBU), which was used as cross-linker, has a fixed length (12.5 Å), we would expect the observed cross-links to occur between residues that are no further apart in the 3D model than the maximum possible cross-link Cα-Cα distance. The maximum distance is usually considered to be linker length plus 2 x lysine length (12.8 Å), and an additional tolerance of between 2-5Å [42]. The majority of the detected cross-links mapped onto the AF2 model within an appropriate distance limit (i.e. <30Å) (Fig 3E), thus suggesting that the SU_C structure model is likely a good fit. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. SNM and UNO_C form a stable heterodimeric complex. (A) SNM and UNO_C, with N-terminally fused maltose-binding protein (MBP) and twin Strep-II affinity tag (Stag), respectively, form a stable SU_C complex. Size exclusion chromatography profile from the final purification step and Coomassie-stained gel for analysis of the indicated peak are shown. (B) Mass photometry analysis of SU_C. The indicated molecular mass was determined by a Gaussian fit of the distribution of counts calibrated against a molecular mass standard. (C) SEC-MALS analysis of SU_C, revealing the indicated molecular mass. (D) AF2 model of SU_C with predicted alignment plot on the right. Regions associated with high (30 Å) and low (0 Å) error as predicted by the algorithm are shown in red and blue, respectively. (E) Analysis of SU_C by cross-linking mass spectrometry (XL-MS). The observed cross-links (with false discovery rate <1%) were plotted onto 2D representations of SNM and UNO_C. Moreover, cross-links were also modelled onto the AF2 model using XMAS [74], with separation distance of cross-linked positions color-coded. A plot with the distance distribution of the observed cross-links mapped to the AF2 model is presented on the right. The majority of cross-links fall within the distance range expected for the DSBU linker (~27 Å, see text and [42]. The longer distance outliers may be caused by errors in the model prediction, or flexibility within the structure. https://doi.org/10.1371/journal.pgen.1010547.g003 DNA-binding of the SNM-UNO_C complex The predicted SU_C structure displayed striking similarity to the structures reported for complexes of kleisins bound to HAWK-type subunits of both cohesin and condensin complexes [2]. The SU_C structure was particularly similar to that of the human cohesin subunits STAG1 and RAD21 [36] (Fig 4A). Like SNM, STAG1 belongs to the stromalin family, and RAD21 is an α-kleisin. The similarity between SU_C and STAG1/RAD21 included the apparent conservation of three surface patches (P1-P3) with positively charged residues that are involved in the binding of cohesin to DNA [43] (Fig 4B). The regions corresponding to P1 and P3 in particular were found to be positively charged. Interestingly, compared to RAD21, UNO has an extra α-helix at the C-terminus (Fig 4B). This C-terminal helix (ch) is highly positively charged and its position within the AF2 model was also suggestive that it might cooperate with the basic patches for DNA-binding (Fig 4B). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. The SU_C complex binds to DNA. (A) Structural comparison of SU_C with STAG1/RAD21 from the human cohesin complex. The structure of STAG1, RAD21 and dsDNA (PDB 6wg3, [36]) was superposed on the SU_C model using Chimera MatchMaker, giving a local C-alpha RMSD of 1.31 Å over 272 residues, and a global C-alpha RMSD of 8.1 Å over all 863 pairs. (B) Surface representation of SNM colored by electrostatic potential to highlight three potential DNA-binding patches that are common with STAG1. UNO_C is shown as a cartoon representation. The inset (right) shows the positively charged residues of the C-terminal helix of UNO that were mutated (see below). The position of dsDNA is taken from the STAG1/RAD21 model shown in (A). (C) DNA-binding of SU_C. For analysis by EMSA, SU_C was incubated with a mix of two distinct DNA fragments, fluorescently labeled with a green and a red dye, respectively. The sequence of the first DNA fragment corresponded to that of the 240 bp repeat previously implicated in sex chromosome conjunction before male M I [26]. A scrambled version of the same sequence was present in the second fragment. Band shifts revealed DNA-binding by SU_C without preference for the 240 bp repeat. (D) SU_C and mutant versions of this complex were purified and analyzed by SDS-PAGE and Coomassie staining. SP1m: SNM with mutant basic patch 1, SP2m: SNM with mutant basic patch 2, Cchm: UNO_C with mutations in C-terminal helix. SP2m resulted in a reduced presence of UNO_C (asterisk) and Cchm in an altered electrophoretic mobility of UNO-C. (E) Representative EMSAs for comparison of DNA-binding activity of SU_C complex variants (wild-type SU_C, SP1mU_C, and SU_Cchm). The range of the analyzed protein concentrations was from 36 nM to 1.7 μM in each experiment. Quantification of the EMSA experiments is shown on the right. Three independent experiments were carried out and the fraction of bound DNA at each data point was calculated. Error bars show s.d. A non-linear regression fit was made to determine apparent KD. A good fit (R-squared 0.955) was obtained for the wild-type SU_C dataset, resulting in a KD of 555 nM (+/- 37), with a hill-factor of 6. For the mutant SP1mU_C, the fit was less good, with an approximate apparent KD of 1.3 μM. For the mutant SU_Cchm, no curve could be fitted. https://doi.org/10.1371/journal.pgen.1010547.g004 To evaluate whether SU_C might indeed bind to DNA, we performed electrophoretic mobility shift assays (EMSAs). Thereby, SU_C was observed to have DNA-binding activity (Fig 4C) that appeared comparable or even higher than that reported for human cohesin [43]. In spermatocytes, SNM, UNO and MNM are all co-localized on the sex chromosome pairing regions, i.e., on the rDNA loci [20,24–26]. Specifically, the 240 bp repeats located within the intergenic spacers of the rDNA repeats were shown to be sufficient for chromosome conjunction. Therefore, we tested whether SU_C might bind preferentially to the 240 bp repeat sequence with an EMSA competition experiment. We mixed two DNA sequences, one corresponding to the 240 bp repeat sequence, and a second with the same sequence composition but scrambled. The former was labelled with Cy5 and the latter with fluorescein. Both sequences were bound equally well by SU_C (Fig 4C). Thus, SU_C does not appear to have a binding preference for the 240bp repeat sequence. In order to test the potential involvement of the basic patches (P1, P2 and P3) of SNM and the C-terminal helix (ch) of UNO in DNA-binding, we generated mutants. The basic patches were altered with a series of point mutations analogous to the work on STAG1/RAD21 [43], and were designated as SNMP1m, SNMP2m and SNMP3m. In case of the C-terminal helix of UNO, we generated the mutant UNOchm with six basic residues changed into either alanine of glutamic acid (R395E, K397A, R398E, R401A, R405E, and R406A). During purification of these four mutant SU_C complexes, SNMP3m proved to be unstable and was therefore not pursued further. The remaining three mutant complexes could be purified to homogeneity like wild-type SU_C (Fig 4D). Compared to U_C, the mutant U_Cchm had a lower mobility during SDS-PAGE (Fig 4D). As SNMP2m did not appear to bind stoichiometric amounts of UNO_C (Fig 4D, asterisk), we also did not analyze it further. However, we compared the efficiency of DNA-binding of the three complexes wild-type SU_C, SP1mU_C and SU_Cchm with EMSAs. Wild-type SU_C bound DNA with a high affinity (KD of 555 nM +/- 37) (Fig 4E). SP1mU_C bound with an approximate three-fold lower affinity. SU_Cchm bound with an even lower affinity (Fig 4E). These results demonstrate that SU_C binds efficiently to DNA. Moreover, they implicate the characteristic highly basic C-terminal helix of UNO in DNA-binding. The positively charged C-terminal helix of UNO promotes chromatin binding and chromosome conjunction To further evaluate the role of the strongly positively charged C-terminal helix of UNO, we generated an UASt-unochm-EGFP transgene for analyses in vivo. Wild-type UNO was previously shown to bind to polytene chromosomes in larval salivary gland after co-expression with SNM [23]. To determine whether UNOchm can still bind to polytene chromosomes, we co-expressed UASt-snm-mCherry and UASt-unochm-EGFP with the salivary gland-specific driver Sgs3-GAL4. In spread polytene chromosome preparations, the yellow signals in chromosomal bands resulting from strictly co-localized SNM-mCherry and UNOchm-EGFP were strongly reduced in intensity compared to controls co-expressing SNM-mCherry and wild-type UNO-EGFP (Fig 5A). The reduced polytene chromosome-binding resulting from co-expression of SNM-mCherry with UNOchm-EGFP did not arise because of lower expression levels, as demonstrated by quantification of fluorescent signals in whole mount preparations (Fig 5C). On the contrary, co-expression of SNM-mCherry with UNOchm-EGFP resulted in higher expression levels compared to controls with SNM-mCherry and wild-type UNO-EGFP (Fig 5C). However, the combination of SNM-mCherry with UNOchm-EGFP was localized predominantly in between polytene chromosomes (Fig 5B), in contrast to the combination of SNM-mCherry with wild-type UNO-EGFP that was preferentially on polytene chromosomes (Fig 5B). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Mutations eliminating positive charge from the C-terminal helix of UNO decrease chromatin binding of SU and interfere with normal chromosome conjunction during meiosis. (A) After co-expression with SNM-mCherry in larval salivary glands, UNOchm-EGFP with a mutant C-terminal helix is still strictly co-localized with SNM-mCherry, as indicated by yellow signals, but its association with polytene chromosome spreads is much weaker in comparison to wild-type UNO-EGFP. UASt transgenes and Sgs3-GAL4 were used for salivary gland-specific expression. (B,C) Expression of SNM-mCherry and UNOchm-EGFP resulting with the UASt transgenes and Sgs3-GAL4 was stronger than the analogous expression of SNM-mCherry and wild-type UNO-EGFP, as revealed by microscopic analysis of whole mount preparations of larval salivary glands. However, the former combination (with UNOchm-EGFP) resulted in signals predominantly in between polytene chromosomes in contrast to the latter combination (with wild-type UNO-EGFP), which was primarily chromosome-associated, as revealed by the optical slices of nuclei. The bar diagram (C) displays average intensities of the mCherry and EGFP signals in the salivary gland nuclei expressing the indicated transgenes, as well as s.d., n = 6 glands (UNO-EGFP) and 7 glands (UNOchm-EGFP). (D) UNOchm-EGFP and wild-type UNO-EGFP are co-immunoprecipitated to a comparable degree along with MNM by SNM-mCherry. For further explanations, see legend of Fig 1. (E-H) Function of UNOchm-EGFP in spermatocytes. The driver bamP-GAL4-VP16 was used for expression of either UASt-unochm-EGFP or UASt-uno-EGFP for comparison. (E) Extent of chromosome missegregation during meiosis in the indicated genotypes was estimated by analysis of the variability of the DNA content detected microscopically in nuclei of early round spermatids in testis squash preparations. Bars indicate the average of the standard deviation of the DNA content distribution observed in distinct spermatid cysts, and whiskers indicate s.d. of these averages. The first two bars on the left represent data reported previously [22]. (F) Squash preparations of uno null mutant testes with either bam> uno-EGFP or bam>unochm-EGFP were labeled with a DNA stain. Representative spermatocytes at the indicated stages reveal an abnormal UNOchm-EGFP localization in the nuclei (dashed circumference). (G) Spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and either bam>uno-EGFP or bam>unochm-EGFP were analyzed by time-lapse imaging. UNOchm-EGFP dot signals were far weaker than those formed by UNOchm-EGFP on the sex chromosome bivalent in spermatocytes from cysts at the S6 stage just before NEBD I. Moreover, the number of major chromosome territories (arrowheads) was increased above the normal number of three in the spermatocytes expressing UNOchm-EGFP instead of endogenous UNO. (H) Progression through M I in spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and bam>unochm-EGFP was analyzed by time-lapse imaging. An UNOchm-EGFP dot moving very rapidly without an associated chromosome mass is marked by an arrowhead until its abrupt disappearance during anaphase I. Premature separation of bivalents before the onset of anaphase I is apparent at the time points 30:45 and 39:45 (min:sec after onset of NEBD I). Scale bars = 5, 10, 5, 10 and 2 μm in A, B, F, G and H, respectively. https://doi.org/10.1371/journal.pgen.1010547.g005 To confirm that UNOchm still binds to SNM and MNM, we performed co-immunoprecipitation experiments after transient co-expression in S2R+ cells. The results clearly confirmed that UNOchm-EGFP still binds to SNM-mCherry and MNM (Fig 5D). Moreover, the purification of recombinant SU_C complexes (Fig 4D) had also demonstrated normal SNM binding by UNOchm_C. According to the results of our analyses in larval salivary glands, in S2R+ cells and in vitro, the mutations in UNOchm, which eliminate positive charge from the C-terminal helix, did not interfere with SNM and MNM binding but they strongly reduced chromosomal recruitment. To analyze UNOchm function during male meiosis, we expressed UASt-unochm-EGFP in uno null mutants using bamP-GAL4-VP16. Analogous expression of wild-type UASt-uno-EGFP in uno null mutants was shown to preclude premature disjunction and random segregation of homologs during M I [22]. In contrast, meiotic chromosome segregation was clearly defective in case of bam> unochm-EGFP in uno null mutants. This was revealed by measuring the nuclear DNA content in individual post-meiotic nuclei of early spermatids. Regular chromosome segregation during wild-type meiosis generates a population of haploid nuclei with highly comparable DNA content, while random chromosome segregation during M I in uno null mutants generates post-meiotic nuclei with a highly variable DNA content [22]. In case of bam>unochm-EGFP in uno null mutants, early spermatids displayed an extent of variability of the nuclear DNA content that was intermediate between that in wild-type controls and uno null mutants (Fig 5E) [22]. Therefore, UNOchm-EGFP does not restore normal meiotic chromosome segregation. Meiotic chromosome segregation was also not entirely normal when UASt-unochm-EGFP was expressed with bamP-GAL4-VP16 in heterozygous unocc1 spermatocytes (Fig 5E), presumably reflecting a dominant-negative effect of UNOchm-EGFP. For further functional characterization of UNOchm-EGFP, we analyzed its localization during spermatogenesis. UNOchm-EGFP expression was readily detectable in squash preparations of testes from uno null mutants with bam> unochm-EGFP. Beginning at around the S3 stage, the subcellular localization of UNOchm-EGFP diverged from that of wild-type UNO-EGFP. In contrast to wild-type UNO-EGFP, which relocated from a diffuse nucleolar distribution into sub-nucleolar foci, UNOchm-EGFP signals maintained a homogenous distribution in the nucleolus (Fig 5F). Eventually, in S5 spermatocytes, UNOchm-EGFP was enriched around rather than within the nucleolus (Fig 5F). Beyond these diffuse perinucleolar UNOchm-EGFP signals, some late spermatocytes displayed also one or two strong EGFP dots in the nucleus. In contrast, wild-type UNO-EGFP was still predominantly in strong intranucleolar foci during the S5 stage (Fig 5F). To characterize chromosome segregation during M I in uno null mutants with bam> unochm-EGFP, we applied time-lapse imaging. As reported earlier for controls (uno null mutants with bam> uno-EGFP), a coalescence of sub-nucleolar UNO-EGFP foci into a strong dot on the pairing center of the chrXY bivalent occurs at the end of the S6 stage (Fig 5G) [22]. In comparison, the UNOchm-EGFP dot signals in late S6 spermatocytes were weaker and more variable in uno null mutants with bam> unochm-EGFP (Fig 5G). Regarding the number of major chromosome territories, most spermatocytes in uno null mutants with bam> unochm-EGFP appeared to be still normal at the onset of M I. Ninety two percent of the spermatocytes displayed three major territories, as in controls (n = 66 from seven distinct cysts). The remaining 8% had four completely separated territories (Fig 5G). However, during prometaphase I, 94% of the spermatocytes displayed premature separation of major bivalents, followed by random segregation of univalents during anaphase I (Fig 5H and S1 Movie). Only four spermatocytes (6%) progressed through M I normally. Unexpectedly, 17% of the spermatocytes analyzed, displayed an EGFP dot with aberrant characteristics. These EGFP dots were not linked to chromosomal His2Av-mRFP masses, and their movements were far more rapidly than those of chromosomes (Fig 5H and S1 Movie). During anaphase I, these EGFP dot signals, as well as those associated with a chromosome disappeared rapidly (Fig 5H and S1 Movie), like wild-type UNO-EGFP [22]. In conclusion, our characterization of UNOchm-EGFP function demonstrated the importance of the C-terminal helix of UNO for normal male meiosis. This conserved helix is characteristic for UNO orthologs and absent from α-kleisins. Consistent with the observations concerning DNA-binding in vitro, mutational elimination of the positive charge in this helix impaired binding to chromosomes and homolog conjunction in spermatocytes. Multimerization by N-terminal domains of UNO and MNM The N-terminal part of UNO, which is most conserved but unrelated to α-kleisins, permits self-association according to our co-immunoprecipitation experiments. For further structural characterization, we purified recombinant UNO_N (aa 1–73). As for SNM, UNO_N was made more soluble with an N-terminal MBP fusion. MBP-UNO_N was purified using amylose affinity chromatography, ion-exchange chromatography and ultimately size exclusion chromatography (SEC). Although MBP-UNO_N was largely free of contaminants, it was present in two peaks during the final SEC (S4A Fig). We kept these peaks separate for measurement of molecular masses using SEC-MALS. Peak 1 and 2 had a mass of 200.2 and of 101.2 kDa, respectively (S4B Fig). These data suggested that peak 1 is a tetramer (theoretical molecular mass 204.56 kDa) and peak 2 a dimer (theoretical molecular mass 102.28 kDa) of UNO_N. We again made use of AF2 modelling using the knowledge of the two UNO_N stoichiometries in our input parameters. Both UNO_N dimer and tetramer generated high confidence models (S4 Fig). These results confirm that UNO can self-associate and indicate that it can form dimers and/or tetramers. Our co-immunoprecipitation experiments revealed self-association not only for UNO_N but also in case of Mod(mdg4)_CP, the common N-terminal part shared between MNM and the other Mod(mdg4) protein isoforms. For further structural characterization, we expressed Mod(mdg4)_CP in E. coli with an N-terminal MBP moiety to promote solubility. After affinity purification using amylose beads, MBP was removed with 3C protease. Ultimately, the protein was purified to homogeneity using SEC (S5A Fig). Strikingly, the elution volume of Mod(mdg4)_CP was considerably larger than expected (S5A Fig). Thus, we measured the molecular mass using SEC-MALS (S5B Fig). The resulting value of 80.92 kDa was a nearly perfect match for a hexamer of Mod(mdg4)_CP (theoretical molecular mass of 81.76 kDa). Mod(mdg4)_CP contains a BTB/POZ motif, for which self-association has been extensively reported [38,39], although never as hexamer. We therefore used AlphaFold 2.2.0 multimer to generate structural models of the Mod(mdg4)_CP hexamer, which suggested a ring-like arrangement of the subunits (S5 Fig). For confirmation of the hexameric ring architecture, we carried out negative stain electron microscopy (NS-EM), which revealed small ring-like assemblies of Mod(mdg4)_CP with a diameter consistent with the AlphaFold model (S5 Fig). SUM complexes on the sex chromosome pairing region are stable To evaluate whether the nucleolar SUM foci in spermatocytes are temporally stable or dynamic, we used fluorescence recovery after photobleaching (FRAP) for analysis. Stable foci are expected, if the physical linkage of chromosomes into bivalents is achieved by SUM protein complexes functioning like a conventional glue. However, in principle, homologs might also be conjoined by temporally dynamic protein assemblies, if some of many remain in place at any given time point, for instance as in liquid-liquid phase separated droplets. The FRAP analysis was done with S5 spermatocytes expressing either SNM-EGFP, UNO-EGFP or MNM-EGFP (using bamP-GAL4-VP16 and UASt transgenes). Because autosomal signals were too weak for analysis, we focused on the intense nucleolar signals. A region comprising about half of the nucleolus was bleached rapidly (Figs 6A and S6). We obtained comparable results from eight cells per genotype and each cell was from a distinct cyst. For analysis of FRAP, we quantified EGFP signal intensities and used the EGFP signals in nucleoli of neighboring spermatocytes, which were not bleached as reference (S6 Fig). Recovery of EGFP signals in the bleached region was observed to be slow and partial (Figs 6A, 6B, and S6). After 90 minutes, recovery of the EGFP signal intensities was maximal in case of SNM-EGFP (50%), intermediate for UNO-EGFP (29%) and minimal for MNM-EGFP (14%) (Fig 6B). However, recovery was restricted to the weaker diffuse nucleolar signals around the intense sub-nucleolar foci. Signals in the sub-nucleolar foci, which remained in a stable spatial pattern in neighboring non-bleached control spermatocytes, did not recover in the bleached nucleolar regions. For further confirmation, we analyzed MNM-EGFP recovery 240 minutes after photobleaching, and again did not detect any recovery of the signals in the sub-nucleolar foci in three independent experiments (Fig 6A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. SUM complexes on the sex chromosome pairing region are stable. (A) FRAP analyses were completed with S5 cysts released from pupal testes of males expressing the indicated UASt transgenes driven by bam-GAL4-VP16. A subregion of the nucleolus (dashed oval) was bleached in one of the spermatocytes at t = 0. Still frames from representative control and bleached spermatocytes are displayed immediately before bleaching (pre) and at the indicated time points (min). (B) The extent of fluorescent signal recovery at 90 minutes after photobleaching was quantified. Bars indicate average and whiskers s.d.; n = 8 spermatocytes per genotype. (C) Intensities of the EGFP signals in sub-nucleolar foci and of the surrounding diffuse nucleolar signals were quantified during stage S5 (late). Dots indicate diffuse signal intensities relative to the intensities in sub-nucleolar foci, which were set as 100%. Each dot represents a distinct spermatocyte. Averages +/- s.d. are indicated as well; n = 16, 22 and 21 spermatocytes (from left to right). (D) Transition of the subcellular localization of UNO-EGFP in nucleoli during spermatocyte maturation (S1 to S4 stage) from diffuse to sub-nucleolar foci as observed in squash preparations of bam>uno-EGFP testes labeled with a DNA stain. Scale bars = 5 μm. https://doi.org/10.1371/journal.pgen.1010547.g006 The extent of recovery of the weak diffuse nucleolar signals that was observed for the different AHC proteins was correlated with the distinct levels of these diffuse signals before photobleaching (Fig 6C). Compared to the average pixel intensity in the sub-nucleolar foci, the diffuse nucleolar signals were highest in case of SNM-EGFP (60%), intermediate for UNO-EGFP (48%) and lowest for MNM-EGFP (18%), when analyzed at the S5 stage, which was also studied in the FRAP experiments (Fig 6C). Interestingly, the diffuse nucleolar signals were dominant in early spermatocytes, and sub-nucleolar foci developed during spermatocyte maturation, as illustrated for UNO-EGFP (Fig 6D) and reported earlier for MNM- and SNM-EGFP [44]. Overall, the results of our FRAP experiments indicated that chromosome conjunction in the sex chromosome bivalents is provided by stable SUM protein complexes. Moreover, the decrease of the diffuse nucleolar signals during normal spermatocyte maturation is consistent with the notion that focal chromosomal SUM complexes are protected against the degradation that apparently occurs with unassembled SUM proteins. Proteolytic cleavage of UNO is sufficient to abolish chromosome conjunction Separase is required for homolog separation during M I in Drosophila males [30], and UNO includes a motif matching the consensus of separase cleavage sites (Fig 7A) [6,22]. We have previously shown that expression of UNOnc-EGFP, a mutant with a non-cleavable (nc) variant of the separase cleavage motif (E130A and R113A), instead of endogenous UNO (bam> unonc-EGFP in uno null mutants) prevented the rapid disappearance and homolog separation during anaphase I [22]. Thus, UNO cleavage by separase was proposed to abolish homolog conjunction during wild-type M I, so that homologs can be pulled apart to opposite spindle poles. To assess whether UNO cleavage is sufficient for elimination of homolog conjunction, we generated a mutant transgene (UASt-unoTEV-EGFP), in which the separase cleavage site was replaced by three repeats of a target sequence for tobacco etch virus (TEV) protease (Fig 7A). After expression of UNOTEV-EGFP instead of endogenous UNO (bam> unoTEV-EGFP in uno null mutants), UNOTEV-EGFP did not disappear rapidly during anaphase I (Fig 7C), in contrast to UNO-EGFP [22]. Moreover, homolog separation was inhibited and massive chromosome bridging was observed during telophase I (Fig 7C). Thus, behavior and effects of UNOTEV-EGFP were identical to those previously observed with UNOnc-EGFP in analogous experiments [22]. Since UNOTEV-EGFP no longer contained the separase cleavage site, this corresponded precisely to expectations. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. UNO cleavage is sufficient for elimination of homolog conjunction. (A) UNO contains a separase cleavage site preceded by a threonine (T128 in D. melanogaster) that is highly conserved among drosophilids [22], as illustrated with the sequence alignment. Positions essential for UNO cleavage are indicated by arrows. In the consensus sequence of separase cleavage sites [6,49], Φ indicates hydrophobic, X any amino acid and ζ hydrophilic residues. Separase cleaves C-terminally after R. The UNOTEV-EGFP mutant contains three repeats of the recognition sequence cleaved by TEV protease (grey shading), replacing the separase cleavage site. (B) Squash preparations uno null mutant testes with bam>unoTEV-EGFP and either no TEV transgene, exumP-TEV or betaTub85DP-TEV (bTub85DP-TEV) as indicated were labeled with a DNA stain and analyzed microscopically to reveal presence and localization of UNOTEV-EGFP during spermatocytes maturation. As indicated (dashed line) in the images with the apical testes regions on the left, a premature UNOTEV-EGFP disappearance was induced by the TEV transgenes. This disappearance occurred after the S3 stage, which still displayed a normal chromosome territory organization and UNOTEV-EGFP signals of normal level and localization in sub-nucleolar foci even in the presence of betaTub85DP-TEV (right panel). (C-E) Progression through M I in spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and bam>unoTEV-EGFP was analyzed by time-lapse imaging. The spermatocytes expressed either no TEV transgene (C) or betaTub85DP-TEV (bTub85DP-TEV) (D). Still frames of representative spermatocytes are displayed with time indicated (min:sec) with t = 0 at the onset of NEBD I. Chromosome organization at NEBD I (number of major chromosome territories), during prometaphase I (presence of univalents) and telophase I (presence of chromosome bridges) was scored. The results are displayed in the bar diagram (E) with numbers of analyzed spermatocytes (n) indicated. (F) Summary of the protein-protein and protein-DNA interactions of the AHC proteins SNM, UNO, MNM and TEF. See discussion for the implications for the mechanism of alternative homolog conjunction. Scale bars = 20 (B, left), 5 (B, right), and 3 (C,D) μm. https://doi.org/10.1371/journal.pgen.1010547.g007 When a transgene driving expression of TEV protease specifically in late spermatocytes was added into the UNOTEV-EGFP expressing background, the resulting phenotype was very distinct. Two TEV transgenes, exumP-TEV and betaTub85DP-TEV, were generated and used. The former resulted in an onset of TEV protease effects at a slightly earlier stage compared to the latter (Fig 7B). The phenotypic effects were highly similar with both TEV transgenes and will be documented primarily for betaTub85DP-TEV. In uno null mutant spermatocytes with bam> unoTEV-EGFP and betaTub85DP-TEV, UNOTEV-EGFP disappearance occurred well after the S3 stage, and S3 spermatocytes still displayed UNOTEV-EGFP signals of normal intensity and normal localization in sub-nucleolar foci (Fig 7B). Moreover, these S3 spermatocytes also displayed normal chromosome territories (Fig 7B). Thus, betaTub85DP-TEV did not affect the initial establishment and maintenance of AHC. However, after the betaTub85DP-TEV-induced disappearance of UNOTEV-EGFP from late spermatocytes, bivalents were prematurely separated into univalents. Cytological analyses with testis squash preparations (S7 Fig) and time-lapse imaging (Fig 7D) revealed an increase in the number of chromosome territories in S6 spermatocytes, as well as the presence of univalents during prometaphase I. After a delayed onset of anaphase I, the univalents were segregated randomly onto the two spindle poles, without formation of chromosome bridges during telophase I (Figs 7D and S7). For a quantitative confirmation of the phenotypic differences resulting from absence or presence of a TEV transgene, we scored the number of major chromosome territories around NEBD I, the presence of univalents during prometaphase I and of chromosome bridges during telophase I after time-lapse imaging (Fig 7E). Overall, these results of our experiments with UNOTEV-EGFP suggested that proteolytic cleavage of UNO is sufficient for elimination of alternative homolog conjunction in spermatocytes. An additional control experiment ruled out that the premature separation of bivalents after co-expression of UNOTEV-EGFP and TEV protease in uno null mutants resulted from some unexpected dominant off-target effect of TEV protease unrelated to UNOTEV-EGFP cleavage. When UNOTEV-EGFP and TEV protease were co-expressed in spermatocytes that also produced UNO from the endogenous locus (bam> unoTEV-EGFP and betaTub85DP-TEV in uno-/+), bivalents were not separated prematurely (S7 Fig). Interestingly, in this uno-/+ background, UNOTEV-EGFP remained weakly detectable as a dot on the chrXY pairing region until anaphase I, while in the uno null mutant background, UNOTEV-EGFP disappearance was complete already before the S6 stage (S7 Fig). As discussed below, the longer perdurance of chromosomal UNOTEV-EGFP in the TEV expressing uno+ background appears to be consistent with a mechanism for chromosome linking involving multimerized SUM protein assemblies. Also consistent with this suggestion, we observed a perdurance of weak sub-nucleolar EGFP foci until M I after expression of UNO_N- or UNO_C-EGFP in uno+, but not in uno null mutant spermatocytes, with UASt transgenes and bamP-GAL4-VP16 (S8 Fig). Expression of these UNO fragments (UNO_N- and UNO_C-EGFP) did not rescue meiotic chromosome segregation in uno null mutants (S8 Fig), and in the uno+ background, they did not have an evident dominant-negative effect (S8 Fig), presumably as their level of expression was low according to EGFP signal intensities. Analogous expression of UNO_M-EGFP resulted in higher expression, but this fragment was present only transiently in early spermatocytes without any enrichment in the nucleolus (S8 Fig). In an attempt to address the control of the separase-mediated UNO cleavage during M I, we generated UASt transgenes expressing mutant UNO-EGFP versions with alterations in a conserved potential phosphorylation site preceding the separase cleavage site (Fig 7A). Experimental evidence from fungal and mammalian organisms [12,45–49] have emphasized that α-kleisin phosphorylation promotes cleavage by separase, in particular in case of the meiotic Rec8 isoforms. The UNO orthologs of Drosophila species all contain a threonine residue (T128 in D. melanogaster) followed by a proline residue immediately upstream of the separase cleavage site (Fig 7A). Phosphorylation of a serine at an equivalent position in a fungal α-kleisin or also its phosphomimetic mutation to glutamic acid was demonstrated to stimulate separase-mediated cleavage in vitro [49]. To evaluate the potential role of T128 phosphorylation in UNO, we mutated the corresponding codon to encode either alanine (T128A), which cannot be phosphorylated, or the phosphomimetic aspartic acid (T128D). Expression of UNOT128A-EGFP in uno null mutants (bam>unoT128A-EGFP in uno null) resulted in homolog conjunction that resisted elimination during M I (S9 Fig). While these results supported a potential significance of T128 phosphorylation, we observed an identical phenotype unexpectedly also with the phosphomimetic UNOT128D-EGFP (bam>unoT128D-EGFP in uno null) (S9 Fig). While our observations demonstrate that T128 is crucial for UNO cleavage by separase during M I, further work will be required to clarify the role of phosphorylation at this position. The C-terminal region of UNO that binds to SNM is derived from α-kleisins UNO homologs can be detected exclusively within Diptera (suborder Brachycera). They have conserved N- and C-terminal domains separated by a putative disordered linker region (Fig 1A). UNO of D. melanogaster was identified originally because of co-purification with both SNM-EGFP and MNM-EGFP from testis extracts [22]. To assess whether SNM or MNM might bind directly to UNO, we performed co-immunoprecipitation experiments after transient expression of tagged proteins in Drosophila S2R+ cells. In this cell line, endogenous UNO expression occurs at most at very low levels according to RNA-seq data, while SNM mRNA is not detectable [31,32]. Similarly, by immunoblotting, we failed to detect endogenously expressed UNO and SNM in S2R+ cells (S1 Fig). In case of MNM, which is but one of many distinct isoforms expressed from the complex mod(mdg4) locus [20,33], RNA-seq data is consistent with endogenous expression in S2R+ cells at low levels [31,32]. As antibodies specific for MNM are not available, the potential presence of endogenous MNM protein in S2R+ cells remains to be confirmed. However, immunoblotting with anti-M_CP, an antibody that reacts with the N-terminal common part (CP) present in all the different Mod(mdg4) isoforms [34], demonstrated that endogenous MNM, if present, is at a level far lower than that generated from transfected expression plasmids (S1 Fig). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. SNM and UNO form a complex that binds MNM. (A) Structure of UNO and interactions with AHC proteins. UNO includes two regions with stronger sequence conservation close to the N- and C-termini, as well as a highly conserved match to the separase cleavage-site consensus. Regional extent of amino acid identity (%) among drosophilid UNO orthologs is indicated with grey shading. Protein interactions of SNM, MNM and full-length UNO, as well as the UNO fragments UNO_N, UNO_M and UNO_C were analyzed by co-immunoprecipitation experiments after transient expression in S2R+ cells. The main findings are summarized schematically (with reference to relevant figure panels in brackets). (B-F) Co-immunoprecipitation after transient expression of the indicated proteins was analyzed by immunoblotting. Protein tags were EGFP (E), mCherry (C) or a myc epitope (myc). Tetracycline repressor fused to a nuclear localization signal and EGFP (TetR-E) was used for control experiments. Presence or absence of proteins in the extracts used for immunoprecipitation (input) or in the samples immunoprecipitated with the indicated antibodies (IP) were analyzed with anti-EGFP (anti-E), anti-mCherry (anti-C), anti-myc (anti-myc) and anti-SNM. A non-specific band (*) recognized by anti-E and the positions of molecular weight markers are indicated on the right side, as well as the bands representing proteins of interest. https://doi.org/10.1371/journal.pgen.1010547.g001 After transfection of S2R+ cells with plasmids for co-expression of SNM-mCherry and UNO-EGFP, we observed efficient co-immunoprecipitation of these two proteins (Fig 1B). In contrast, co-immunoprecipitation of MNM-mCherry and UNO-EGFP could not be observed in an analogous experiment (Fig 1C). According to our initial bioinformatic analyses with standard BLAST searches [22], the predicted amino acid (aa) sequence of UNO failed to display significant similarities to proteins with known functions. However, because of the apparent direct binding of UNO to SNM, a member of the stromalin family of α-kleisin-binding proteins, and because UNO was previously shown to include a functionally essential cleavage site cut by separase (Fig 1A) [22], which cleaves primarily α-kleisins, we focused specifically on detecting α-kleisin similarities within UNO. Interestingly, a region close to the C-terminus of UNO (aa 289–364) was found to be similar to an internal α-kleisins region (aa 320–394 of human Rad21) (Figs 1A and S2). This conserved internal α-kleisin region is known to bind stromalin/SA/STAG proteins [35,36]. We conclude that UNO includes a region that was co-opted from an α-kleisin. This region of UNO comprises a separase cleavage site and the C-terminal stromalin-binding region. Co-immunoprecipitation experiments clearly confirmed that a C-terminal fragment of UNO (UNO_C, aa 241–417) mediates binding of UNO to SNM-mCherry (Fig 1A and 1D). The N-terminal region of UNO (UNO_N, aa 1–78) and the middle part (UNO_M, aa 79–240) did not bind to SNM (Fig 1A and 1D). We conclude that SNM and UNO appear to form a complex, which will be designated as SU complex in the following. The co-immunoprecipitation experiments also indicated that SU complex formation was accompanied by mutual stabilization of the interacting proteins. After transient co-expression, the resulting levels of UNO and SNM were around sixfold higher compared to those detected after individual expression (S3 Fig). Analyses with UNO subregions indicated that UNO_C stability was highly dependent on the presence of SNM, while UNO_N and UNO_M appeared to be more stable and not dependent on SNM (S3 Fig). The N-terminal region of UNO mediates self-association Other characteristic hallmarks of α-kleisins, the N- and C-terminal domains, which bind to the cohesin core subunits SMC3 and SMC1, respectively, are absent in UNO. UNO does not extend C-terminally beyond the SNM-binding region. At the N-terminus, UNO has a domain that is distinct from those present in α-kleisins. The N-terminal domain of UNO (UNO_N, aa 1–78) is predicted to form β-strands, while the N-terminal SMC3-binding region of α-kleisins is α-helical. Moreover, a region with high sequence similarity to UNO_N is present in a poorly characterized Drosophila protein (CG32117) that has no similarity to α-kleisins [22]. Interestingly, our co-immunoprecipitation experiments demonstrated that (UNO_N) mediates self-association. After co-expression of UNO_N-EGFP and UNO_N-mCherry, we observed their efficient co-immunoprecipitation (Fig 1A and 1E). UNO_N-mCherry was also co-immunoprecipitated by full-length UNO-EGFP along with SNM (Fig 1E). In contrast, only SNM but not UNO_N-mCherry was co-immunoprecipitated by UNO_C-EGFP (Fig 1E). These results suggest that the distinctive N-terminal domain of UNO (UNO_N) mediates self-association, even when SNM is bound to UNO. SNM does not appear to have the ability to self-associate, in contrast to UNO. We did not detect co-immunoprecipitation of SNM-EGFP and SNM-mCherry (Fig 1A and 1B). A complex of SNM and UNO binds MNM The lack of co-immunoprecipitation of MNM-mCherry and UNO-EGFP (Fig 1C) suggested that the co-purification of MNM-EGFP and UNO from Drosophila testis extracts [22] was not the result of a direct interaction between these two proteins. Thus, we considered the possibility that SNM might function as a bridging factor, by binding not only to UNO, but perhaps also to MNM. However, SNM-mCherry binding to MNM-EGFP could not be detected in co-immunoprecipitation experiments (Fig 1A and 1B). Similarly, untagged SNM was not co-immunoprecipitated by MNM-mCherry (Fig 1C). However, when the three proteins SNM, UNO-EGFP and MNM-mCherry were co-expressed, we clearly detected their co-immunoprecipitation (Fig 1C). Moreover, we also observed efficient co-immunoprecipitation of the C-terminal domain of UNO (UNO_C-EGFP) with SNM-mCherry and MNM-myc (Fig 1F). In conclusion, the SU complex formed by SNM and UNO can bind MNM, while individually SNM and UNO do not appear to have this ability (Fig 1A). The complex formed by the three proteins SNM, UNO and MNM will be designated as SUM complex in the following. MNM self-association permits binding of TEF to the SUM complex MNM is only one of more than thirty distinct protein isoforms expressed by mod(mdg4) [20,34]. Mod(mdg4) products, including MNM, share an N-terminal common part (CP) followed by an isoform-specific C-terminal region (Fig 2A). To address whether SU binds only MNM or also other Mod(mdg4) proteins, we performed additional co-immunoprecipitation experiments. The Mod(mdg4) isoforms T, P and C were analyzed in these experiments. The T isoform (also designated as 67.2) represents the most extensively characterized Mod(mdg4) product, which functions along with Su(Hw) and CP190 in the gypsy insulator for example [37]. The C and P isoforms might be expressed in testis according to RNA-seq data [31]. In addition, we used a construct for expression of the shared N-terminal common part (CP) of the Mod(mdg4) proteins. Our experiments revealed that none of the analyzed Mod(mdg4) variants (C, P, T and CP, all with C-terminal mCherry) were able to co-immunoprecipitate SNM-EGFP and UNO-myc, in contrast to MNM-mCherry (Fig 2B). We conclude that the isoform-specific C-terminal region of MNM (MNM_C) is required for binding to the SU complex. Additional experiments confirmed that MNM_C fused to mCherry (MNM_C-mCherry) is co-immunoprecipitated by UNO-EGFP after co-expression with SNM (Fig 2G). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. MNM self-association is required for formation of SUMT complexes containing SNM, UNO, MNM and TEF. (A) Structure of MNM and interactions with AHC proteins. MNM has an N-terminal (Mod(mdg4)_CP) that is also present in other Mod(mdg4) isoforms (including _T, _P and _C). This part includes a BTB domain. In addition, MNM has a C-terminal MNM-specific region. Protein interactions were analyzed by co-immunoprecipitation experiments after transient expression in S2R+ cells. The main findings are summarized schematically (with reference to relevant figure panels in brackets). (B-G) Co-immunoprecipitation after transient expression of the indicated proteins was analyzed by immunoblotting. Protein tags were EGFP (E), mCherry (C) or a myc epitope (myc). Tetracycline repressor fused to a nuclear localization signal and EGFP (TetR-E) or a C-terminal fragment of Mad1 tagged with a myc-epitope (myc-Mad1_C) was used for control experiments. Presence or absence of proteins in the extracts used for immunoprecipitation (input) or in the samples (IP) immunoprecipitated with the indicated antibodies were analyzed with anti-EGFP (anti-E), anti-mCherry (anti-C), anti-myc (anti-myc), and anti-Mod(mdg4)_CP (anti-M_CP). The positions of molecular weight markers are indicated on the right side, as well as the bands representing the indicated proteins of interest. (B) While MNM-mCherry (MNM-C) co-immunoprecipitated SNM-EGFP (SNM-E) and UNO-myc (UNO-myc), other Mod(mdg4) isoforms (T, P and C) as well as the common part (CP) tagged with mCherry did not co-immunoprecipitate SNM-EGFP and UNO-myc. (C) The mCherry fusions of Mod(mdg4) variants (CP, T, P and C) all co-immunoprecipitated MNM-EGFP (MNM-E) specifically, as demonstrated by first two lanes, which present a positive and negative control experiment, respectively. (D) MNM-E co-immunoprecipitated MNM-myc specifically, as demonstrated by the negative control experiment with TetR-E. (E) The C-terminal MNM-specific region as an mCherry fusion (MNM_C-C) was not co-immunoprecipitated by MNM-E in contrast to full length MNM-mCherry (MNM-C). (F) UNO-E co-immunoprecipitated SNM-mCherry, MNM and TEF-myc specifically [23], as demonstrated by the negative control experiment with TetR-E. (G) UNO_C-E co-immunoprecipitated only SNM-mCherry, MNM_C-C but not TEF-myc, even though TEF-myc was observed to co-immunoprecipitate MNM_C-C, indicating that MNM_C can bind in a mutually exclusive manner to either the SNM-UNO complex or to TEF. https://doi.org/10.1371/journal.pgen.1010547.g002 The N-terminal common part of the Mod(mdg4) proteins includes a BTB/POZ domain, which is present in a wide range of proteins with distinct functions (Fig 2A). BTB/POZ domains appear to mediate protein-protein interactions, including dimerization, tetramerization and multimerization [38,39]. Therefore, we evaluated whether MNM self-associates via the N-terminal common part (CP) and hence also with Mod(mdg4) proteins other than MNM. Indeed, MNM-EGFP was readily co-immunoprecipitated with all the analyzed Mod(mdg4) variants C, P, T and CP fused to mCherry (Fig 2C). Similarly, we detected co-immunoprecipitation of MNM-EGFP and MNM-myc (Fig 2D). We conclude that MNM has the potential to associate with itself and with other Mod(mdg4) proteins via the N-terminal common part. The isoform-specific C-terminal region of MNM fused to mCherry (MNM_C-mCherry) was not observed to co-immunoprecipitate with MNM-EGFP (Fig 2E). The isoform-specific C-terminal region of MNM (MNM_C) does not only bind to the SU complex (Fig 2A and 2B) but also to the N-terminal region of TEF [23]. Moreover, consistent with MNM’s ability to bind to both the SU complex and to TEF, all four AHC proteins (SNM, UNO, MNM and TEF) are co-immunoprecipitated after co-expression in S2R+ cells (Fig 2F) [23]. This apparent SUMT complex could arise if the binding of SU and TEF to MNM was not mutually exclusive. However, even if mutually exclusive, SUMT complex formation might still succeed, because of the self-association of MNM. For example, one of two associated MNM proteins might bind to SU and the other to TEF. To address the mode of TEF binding in the SUMT complex, we co-expressed TEF-myc with MNM_C-mCherry, SNM and UNO_C-EGFP (Fig 2G). As shown above, MNM_C cannot self-associate, but it can bind to TEF and also to the SU complex. UNO_C cannot self-associate, but it can bind to SNM and recruit MNM_C. As expected, UNO_C-EGFP was observed to co-immunoprecipitate MNM_C-mCherry after co-expression of the four proteins (TEF-myc, MNM_C-mCherry, SNM and UNO_C-EGFP), but TEF-myc was not co-immunoprecipitated (Fig 2G). Thus, after binding to the SU_C-EGFP complex, MNM_C-mCherry was no longer able to bind TEF-myc. This result suggests that the binding of TEF and SU to MNM is mutually exclusive. Accordingly, formation of SUMT assemblies, which contain all four known AHC proteins, depends on self-association of either MNM or possibly also of UNO. Purification and structural analysis of a complex of SNM and UNO_C For further characterization of the AHC protein-protein interactions, we succeeded in purifying recombinant versions of some of these proteins for analysis in vitro. To characterize the SU complex, we expressed full-length SNM and UNO_C (aa 281–417) in baculovirus-infected insect cells. To promote solubility, SNM was fused N-terminally with a maltose-binding protein (MBP). UNO_C was fused to an N-terminal twin Strep-II affinity tag. Using streptactin affinity chromatography followed by size exclusion chromatography, we purified an MBP-SNM/Strep-II-UNO_C complex to homogeneity (from here on referred to as SU_C) (Fig 3A). We next determined the stoichiometry of SU_C using both mass photometry (MP) and multi-angle light scattering coupled to size exclusion chromatography (SEC-MALS). MP and SEC-MALS measured a molecular mass of 178 and 211.9 kDa, respectively (Fig 3B and 3C). As the theoretical molecular mass of a 1:1 dimer was 173.4 kDa, these results provided strong evidence of a 1:1 stoichiometry. Based on this stoichiometry, we generated a de novo structure prediction using AlphaFold 2.2.0 Multimer (AF2) [40,41] (Fig 3D). In order to test the AF2 model, we made use of bifunctional chemical cross-linking coupled to mass spectrometry (XL-MS). The observed cross-links were mapped onto the 2D representation of the complex and onto the 3D model (Fig 3E). Given that Disuccinimidyl Dibutyric Urea (DSBU), which was used as cross-linker, has a fixed length (12.5 Å), we would expect the observed cross-links to occur between residues that are no further apart in the 3D model than the maximum possible cross-link Cα-Cα distance. The maximum distance is usually considered to be linker length plus 2 x lysine length (12.8 Å), and an additional tolerance of between 2-5Å [42]. The majority of the detected cross-links mapped onto the AF2 model within an appropriate distance limit (i.e. <30Å) (Fig 3E), thus suggesting that the SU_C structure model is likely a good fit. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. SNM and UNO_C form a stable heterodimeric complex. (A) SNM and UNO_C, with N-terminally fused maltose-binding protein (MBP) and twin Strep-II affinity tag (Stag), respectively, form a stable SU_C complex. Size exclusion chromatography profile from the final purification step and Coomassie-stained gel for analysis of the indicated peak are shown. (B) Mass photometry analysis of SU_C. The indicated molecular mass was determined by a Gaussian fit of the distribution of counts calibrated against a molecular mass standard. (C) SEC-MALS analysis of SU_C, revealing the indicated molecular mass. (D) AF2 model of SU_C with predicted alignment plot on the right. Regions associated with high (30 Å) and low (0 Å) error as predicted by the algorithm are shown in red and blue, respectively. (E) Analysis of SU_C by cross-linking mass spectrometry (XL-MS). The observed cross-links (with false discovery rate <1%) were plotted onto 2D representations of SNM and UNO_C. Moreover, cross-links were also modelled onto the AF2 model using XMAS [74], with separation distance of cross-linked positions color-coded. A plot with the distance distribution of the observed cross-links mapped to the AF2 model is presented on the right. The majority of cross-links fall within the distance range expected for the DSBU linker (~27 Å, see text and [42]. The longer distance outliers may be caused by errors in the model prediction, or flexibility within the structure. https://doi.org/10.1371/journal.pgen.1010547.g003 DNA-binding of the SNM-UNO_C complex The predicted SU_C structure displayed striking similarity to the structures reported for complexes of kleisins bound to HAWK-type subunits of both cohesin and condensin complexes [2]. The SU_C structure was particularly similar to that of the human cohesin subunits STAG1 and RAD21 [36] (Fig 4A). Like SNM, STAG1 belongs to the stromalin family, and RAD21 is an α-kleisin. The similarity between SU_C and STAG1/RAD21 included the apparent conservation of three surface patches (P1-P3) with positively charged residues that are involved in the binding of cohesin to DNA [43] (Fig 4B). The regions corresponding to P1 and P3 in particular were found to be positively charged. Interestingly, compared to RAD21, UNO has an extra α-helix at the C-terminus (Fig 4B). This C-terminal helix (ch) is highly positively charged and its position within the AF2 model was also suggestive that it might cooperate with the basic patches for DNA-binding (Fig 4B). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. The SU_C complex binds to DNA. (A) Structural comparison of SU_C with STAG1/RAD21 from the human cohesin complex. The structure of STAG1, RAD21 and dsDNA (PDB 6wg3, [36]) was superposed on the SU_C model using Chimera MatchMaker, giving a local C-alpha RMSD of 1.31 Å over 272 residues, and a global C-alpha RMSD of 8.1 Å over all 863 pairs. (B) Surface representation of SNM colored by electrostatic potential to highlight three potential DNA-binding patches that are common with STAG1. UNO_C is shown as a cartoon representation. The inset (right) shows the positively charged residues of the C-terminal helix of UNO that were mutated (see below). The position of dsDNA is taken from the STAG1/RAD21 model shown in (A). (C) DNA-binding of SU_C. For analysis by EMSA, SU_C was incubated with a mix of two distinct DNA fragments, fluorescently labeled with a green and a red dye, respectively. The sequence of the first DNA fragment corresponded to that of the 240 bp repeat previously implicated in sex chromosome conjunction before male M I [26]. A scrambled version of the same sequence was present in the second fragment. Band shifts revealed DNA-binding by SU_C without preference for the 240 bp repeat. (D) SU_C and mutant versions of this complex were purified and analyzed by SDS-PAGE and Coomassie staining. SP1m: SNM with mutant basic patch 1, SP2m: SNM with mutant basic patch 2, Cchm: UNO_C with mutations in C-terminal helix. SP2m resulted in a reduced presence of UNO_C (asterisk) and Cchm in an altered electrophoretic mobility of UNO-C. (E) Representative EMSAs for comparison of DNA-binding activity of SU_C complex variants (wild-type SU_C, SP1mU_C, and SU_Cchm). The range of the analyzed protein concentrations was from 36 nM to 1.7 μM in each experiment. Quantification of the EMSA experiments is shown on the right. Three independent experiments were carried out and the fraction of bound DNA at each data point was calculated. Error bars show s.d. A non-linear regression fit was made to determine apparent KD. A good fit (R-squared 0.955) was obtained for the wild-type SU_C dataset, resulting in a KD of 555 nM (+/- 37), with a hill-factor of 6. For the mutant SP1mU_C, the fit was less good, with an approximate apparent KD of 1.3 μM. For the mutant SU_Cchm, no curve could be fitted. https://doi.org/10.1371/journal.pgen.1010547.g004 To evaluate whether SU_C might indeed bind to DNA, we performed electrophoretic mobility shift assays (EMSAs). Thereby, SU_C was observed to have DNA-binding activity (Fig 4C) that appeared comparable or even higher than that reported for human cohesin [43]. In spermatocytes, SNM, UNO and MNM are all co-localized on the sex chromosome pairing regions, i.e., on the rDNA loci [20,24–26]. Specifically, the 240 bp repeats located within the intergenic spacers of the rDNA repeats were shown to be sufficient for chromosome conjunction. Therefore, we tested whether SU_C might bind preferentially to the 240 bp repeat sequence with an EMSA competition experiment. We mixed two DNA sequences, one corresponding to the 240 bp repeat sequence, and a second with the same sequence composition but scrambled. The former was labelled with Cy5 and the latter with fluorescein. Both sequences were bound equally well by SU_C (Fig 4C). Thus, SU_C does not appear to have a binding preference for the 240bp repeat sequence. In order to test the potential involvement of the basic patches (P1, P2 and P3) of SNM and the C-terminal helix (ch) of UNO in DNA-binding, we generated mutants. The basic patches were altered with a series of point mutations analogous to the work on STAG1/RAD21 [43], and were designated as SNMP1m, SNMP2m and SNMP3m. In case of the C-terminal helix of UNO, we generated the mutant UNOchm with six basic residues changed into either alanine of glutamic acid (R395E, K397A, R398E, R401A, R405E, and R406A). During purification of these four mutant SU_C complexes, SNMP3m proved to be unstable and was therefore not pursued further. The remaining three mutant complexes could be purified to homogeneity like wild-type SU_C (Fig 4D). Compared to U_C, the mutant U_Cchm had a lower mobility during SDS-PAGE (Fig 4D). As SNMP2m did not appear to bind stoichiometric amounts of UNO_C (Fig 4D, asterisk), we also did not analyze it further. However, we compared the efficiency of DNA-binding of the three complexes wild-type SU_C, SP1mU_C and SU_Cchm with EMSAs. Wild-type SU_C bound DNA with a high affinity (KD of 555 nM +/- 37) (Fig 4E). SP1mU_C bound with an approximate three-fold lower affinity. SU_Cchm bound with an even lower affinity (Fig 4E). These results demonstrate that SU_C binds efficiently to DNA. Moreover, they implicate the characteristic highly basic C-terminal helix of UNO in DNA-binding. The positively charged C-terminal helix of UNO promotes chromatin binding and chromosome conjunction To further evaluate the role of the strongly positively charged C-terminal helix of UNO, we generated an UASt-unochm-EGFP transgene for analyses in vivo. Wild-type UNO was previously shown to bind to polytene chromosomes in larval salivary gland after co-expression with SNM [23]. To determine whether UNOchm can still bind to polytene chromosomes, we co-expressed UASt-snm-mCherry and UASt-unochm-EGFP with the salivary gland-specific driver Sgs3-GAL4. In spread polytene chromosome preparations, the yellow signals in chromosomal bands resulting from strictly co-localized SNM-mCherry and UNOchm-EGFP were strongly reduced in intensity compared to controls co-expressing SNM-mCherry and wild-type UNO-EGFP (Fig 5A). The reduced polytene chromosome-binding resulting from co-expression of SNM-mCherry with UNOchm-EGFP did not arise because of lower expression levels, as demonstrated by quantification of fluorescent signals in whole mount preparations (Fig 5C). On the contrary, co-expression of SNM-mCherry with UNOchm-EGFP resulted in higher expression levels compared to controls with SNM-mCherry and wild-type UNO-EGFP (Fig 5C). However, the combination of SNM-mCherry with UNOchm-EGFP was localized predominantly in between polytene chromosomes (Fig 5B), in contrast to the combination of SNM-mCherry with wild-type UNO-EGFP that was preferentially on polytene chromosomes (Fig 5B). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. Mutations eliminating positive charge from the C-terminal helix of UNO decrease chromatin binding of SU and interfere with normal chromosome conjunction during meiosis. (A) After co-expression with SNM-mCherry in larval salivary glands, UNOchm-EGFP with a mutant C-terminal helix is still strictly co-localized with SNM-mCherry, as indicated by yellow signals, but its association with polytene chromosome spreads is much weaker in comparison to wild-type UNO-EGFP. UASt transgenes and Sgs3-GAL4 were used for salivary gland-specific expression. (B,C) Expression of SNM-mCherry and UNOchm-EGFP resulting with the UASt transgenes and Sgs3-GAL4 was stronger than the analogous expression of SNM-mCherry and wild-type UNO-EGFP, as revealed by microscopic analysis of whole mount preparations of larval salivary glands. However, the former combination (with UNOchm-EGFP) resulted in signals predominantly in between polytene chromosomes in contrast to the latter combination (with wild-type UNO-EGFP), which was primarily chromosome-associated, as revealed by the optical slices of nuclei. The bar diagram (C) displays average intensities of the mCherry and EGFP signals in the salivary gland nuclei expressing the indicated transgenes, as well as s.d., n = 6 glands (UNO-EGFP) and 7 glands (UNOchm-EGFP). (D) UNOchm-EGFP and wild-type UNO-EGFP are co-immunoprecipitated to a comparable degree along with MNM by SNM-mCherry. For further explanations, see legend of Fig 1. (E-H) Function of UNOchm-EGFP in spermatocytes. The driver bamP-GAL4-VP16 was used for expression of either UASt-unochm-EGFP or UASt-uno-EGFP for comparison. (E) Extent of chromosome missegregation during meiosis in the indicated genotypes was estimated by analysis of the variability of the DNA content detected microscopically in nuclei of early round spermatids in testis squash preparations. Bars indicate the average of the standard deviation of the DNA content distribution observed in distinct spermatid cysts, and whiskers indicate s.d. of these averages. The first two bars on the left represent data reported previously [22]. (F) Squash preparations of uno null mutant testes with either bam> uno-EGFP or bam>unochm-EGFP were labeled with a DNA stain. Representative spermatocytes at the indicated stages reveal an abnormal UNOchm-EGFP localization in the nuclei (dashed circumference). (G) Spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and either bam>uno-EGFP or bam>unochm-EGFP were analyzed by time-lapse imaging. UNOchm-EGFP dot signals were far weaker than those formed by UNOchm-EGFP on the sex chromosome bivalent in spermatocytes from cysts at the S6 stage just before NEBD I. Moreover, the number of major chromosome territories (arrowheads) was increased above the normal number of three in the spermatocytes expressing UNOchm-EGFP instead of endogenous UNO. (H) Progression through M I in spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and bam>unochm-EGFP was analyzed by time-lapse imaging. An UNOchm-EGFP dot moving very rapidly without an associated chromosome mass is marked by an arrowhead until its abrupt disappearance during anaphase I. Premature separation of bivalents before the onset of anaphase I is apparent at the time points 30:45 and 39:45 (min:sec after onset of NEBD I). Scale bars = 5, 10, 5, 10 and 2 μm in A, B, F, G and H, respectively. https://doi.org/10.1371/journal.pgen.1010547.g005 To confirm that UNOchm still binds to SNM and MNM, we performed co-immunoprecipitation experiments after transient co-expression in S2R+ cells. The results clearly confirmed that UNOchm-EGFP still binds to SNM-mCherry and MNM (Fig 5D). Moreover, the purification of recombinant SU_C complexes (Fig 4D) had also demonstrated normal SNM binding by UNOchm_C. According to the results of our analyses in larval salivary glands, in S2R+ cells and in vitro, the mutations in UNOchm, which eliminate positive charge from the C-terminal helix, did not interfere with SNM and MNM binding but they strongly reduced chromosomal recruitment. To analyze UNOchm function during male meiosis, we expressed UASt-unochm-EGFP in uno null mutants using bamP-GAL4-VP16. Analogous expression of wild-type UASt-uno-EGFP in uno null mutants was shown to preclude premature disjunction and random segregation of homologs during M I [22]. In contrast, meiotic chromosome segregation was clearly defective in case of bam> unochm-EGFP in uno null mutants. This was revealed by measuring the nuclear DNA content in individual post-meiotic nuclei of early spermatids. Regular chromosome segregation during wild-type meiosis generates a population of haploid nuclei with highly comparable DNA content, while random chromosome segregation during M I in uno null mutants generates post-meiotic nuclei with a highly variable DNA content [22]. In case of bam>unochm-EGFP in uno null mutants, early spermatids displayed an extent of variability of the nuclear DNA content that was intermediate between that in wild-type controls and uno null mutants (Fig 5E) [22]. Therefore, UNOchm-EGFP does not restore normal meiotic chromosome segregation. Meiotic chromosome segregation was also not entirely normal when UASt-unochm-EGFP was expressed with bamP-GAL4-VP16 in heterozygous unocc1 spermatocytes (Fig 5E), presumably reflecting a dominant-negative effect of UNOchm-EGFP. For further functional characterization of UNOchm-EGFP, we analyzed its localization during spermatogenesis. UNOchm-EGFP expression was readily detectable in squash preparations of testes from uno null mutants with bam> unochm-EGFP. Beginning at around the S3 stage, the subcellular localization of UNOchm-EGFP diverged from that of wild-type UNO-EGFP. In contrast to wild-type UNO-EGFP, which relocated from a diffuse nucleolar distribution into sub-nucleolar foci, UNOchm-EGFP signals maintained a homogenous distribution in the nucleolus (Fig 5F). Eventually, in S5 spermatocytes, UNOchm-EGFP was enriched around rather than within the nucleolus (Fig 5F). Beyond these diffuse perinucleolar UNOchm-EGFP signals, some late spermatocytes displayed also one or two strong EGFP dots in the nucleus. In contrast, wild-type UNO-EGFP was still predominantly in strong intranucleolar foci during the S5 stage (Fig 5F). To characterize chromosome segregation during M I in uno null mutants with bam> unochm-EGFP, we applied time-lapse imaging. As reported earlier for controls (uno null mutants with bam> uno-EGFP), a coalescence of sub-nucleolar UNO-EGFP foci into a strong dot on the pairing center of the chrXY bivalent occurs at the end of the S6 stage (Fig 5G) [22]. In comparison, the UNOchm-EGFP dot signals in late S6 spermatocytes were weaker and more variable in uno null mutants with bam> unochm-EGFP (Fig 5G). Regarding the number of major chromosome territories, most spermatocytes in uno null mutants with bam> unochm-EGFP appeared to be still normal at the onset of M I. Ninety two percent of the spermatocytes displayed three major territories, as in controls (n = 66 from seven distinct cysts). The remaining 8% had four completely separated territories (Fig 5G). However, during prometaphase I, 94% of the spermatocytes displayed premature separation of major bivalents, followed by random segregation of univalents during anaphase I (Fig 5H and S1 Movie). Only four spermatocytes (6%) progressed through M I normally. Unexpectedly, 17% of the spermatocytes analyzed, displayed an EGFP dot with aberrant characteristics. These EGFP dots were not linked to chromosomal His2Av-mRFP masses, and their movements were far more rapidly than those of chromosomes (Fig 5H and S1 Movie). During anaphase I, these EGFP dot signals, as well as those associated with a chromosome disappeared rapidly (Fig 5H and S1 Movie), like wild-type UNO-EGFP [22]. In conclusion, our characterization of UNOchm-EGFP function demonstrated the importance of the C-terminal helix of UNO for normal male meiosis. This conserved helix is characteristic for UNO orthologs and absent from α-kleisins. Consistent with the observations concerning DNA-binding in vitro, mutational elimination of the positive charge in this helix impaired binding to chromosomes and homolog conjunction in spermatocytes. Multimerization by N-terminal domains of UNO and MNM The N-terminal part of UNO, which is most conserved but unrelated to α-kleisins, permits self-association according to our co-immunoprecipitation experiments. For further structural characterization, we purified recombinant UNO_N (aa 1–73). As for SNM, UNO_N was made more soluble with an N-terminal MBP fusion. MBP-UNO_N was purified using amylose affinity chromatography, ion-exchange chromatography and ultimately size exclusion chromatography (SEC). Although MBP-UNO_N was largely free of contaminants, it was present in two peaks during the final SEC (S4A Fig). We kept these peaks separate for measurement of molecular masses using SEC-MALS. Peak 1 and 2 had a mass of 200.2 and of 101.2 kDa, respectively (S4B Fig). These data suggested that peak 1 is a tetramer (theoretical molecular mass 204.56 kDa) and peak 2 a dimer (theoretical molecular mass 102.28 kDa) of UNO_N. We again made use of AF2 modelling using the knowledge of the two UNO_N stoichiometries in our input parameters. Both UNO_N dimer and tetramer generated high confidence models (S4 Fig). These results confirm that UNO can self-associate and indicate that it can form dimers and/or tetramers. Our co-immunoprecipitation experiments revealed self-association not only for UNO_N but also in case of Mod(mdg4)_CP, the common N-terminal part shared between MNM and the other Mod(mdg4) protein isoforms. For further structural characterization, we expressed Mod(mdg4)_CP in E. coli with an N-terminal MBP moiety to promote solubility. After affinity purification using amylose beads, MBP was removed with 3C protease. Ultimately, the protein was purified to homogeneity using SEC (S5A Fig). Strikingly, the elution volume of Mod(mdg4)_CP was considerably larger than expected (S5A Fig). Thus, we measured the molecular mass using SEC-MALS (S5B Fig). The resulting value of 80.92 kDa was a nearly perfect match for a hexamer of Mod(mdg4)_CP (theoretical molecular mass of 81.76 kDa). Mod(mdg4)_CP contains a BTB/POZ motif, for which self-association has been extensively reported [38,39], although never as hexamer. We therefore used AlphaFold 2.2.0 multimer to generate structural models of the Mod(mdg4)_CP hexamer, which suggested a ring-like arrangement of the subunits (S5 Fig). For confirmation of the hexameric ring architecture, we carried out negative stain electron microscopy (NS-EM), which revealed small ring-like assemblies of Mod(mdg4)_CP with a diameter consistent with the AlphaFold model (S5 Fig). SUM complexes on the sex chromosome pairing region are stable To evaluate whether the nucleolar SUM foci in spermatocytes are temporally stable or dynamic, we used fluorescence recovery after photobleaching (FRAP) for analysis. Stable foci are expected, if the physical linkage of chromosomes into bivalents is achieved by SUM protein complexes functioning like a conventional glue. However, in principle, homologs might also be conjoined by temporally dynamic protein assemblies, if some of many remain in place at any given time point, for instance as in liquid-liquid phase separated droplets. The FRAP analysis was done with S5 spermatocytes expressing either SNM-EGFP, UNO-EGFP or MNM-EGFP (using bamP-GAL4-VP16 and UASt transgenes). Because autosomal signals were too weak for analysis, we focused on the intense nucleolar signals. A region comprising about half of the nucleolus was bleached rapidly (Figs 6A and S6). We obtained comparable results from eight cells per genotype and each cell was from a distinct cyst. For analysis of FRAP, we quantified EGFP signal intensities and used the EGFP signals in nucleoli of neighboring spermatocytes, which were not bleached as reference (S6 Fig). Recovery of EGFP signals in the bleached region was observed to be slow and partial (Figs 6A, 6B, and S6). After 90 minutes, recovery of the EGFP signal intensities was maximal in case of SNM-EGFP (50%), intermediate for UNO-EGFP (29%) and minimal for MNM-EGFP (14%) (Fig 6B). However, recovery was restricted to the weaker diffuse nucleolar signals around the intense sub-nucleolar foci. Signals in the sub-nucleolar foci, which remained in a stable spatial pattern in neighboring non-bleached control spermatocytes, did not recover in the bleached nucleolar regions. For further confirmation, we analyzed MNM-EGFP recovery 240 minutes after photobleaching, and again did not detect any recovery of the signals in the sub-nucleolar foci in three independent experiments (Fig 6A). Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. SUM complexes on the sex chromosome pairing region are stable. (A) FRAP analyses were completed with S5 cysts released from pupal testes of males expressing the indicated UASt transgenes driven by bam-GAL4-VP16. A subregion of the nucleolus (dashed oval) was bleached in one of the spermatocytes at t = 0. Still frames from representative control and bleached spermatocytes are displayed immediately before bleaching (pre) and at the indicated time points (min). (B) The extent of fluorescent signal recovery at 90 minutes after photobleaching was quantified. Bars indicate average and whiskers s.d.; n = 8 spermatocytes per genotype. (C) Intensities of the EGFP signals in sub-nucleolar foci and of the surrounding diffuse nucleolar signals were quantified during stage S5 (late). Dots indicate diffuse signal intensities relative to the intensities in sub-nucleolar foci, which were set as 100%. Each dot represents a distinct spermatocyte. Averages +/- s.d. are indicated as well; n = 16, 22 and 21 spermatocytes (from left to right). (D) Transition of the subcellular localization of UNO-EGFP in nucleoli during spermatocyte maturation (S1 to S4 stage) from diffuse to sub-nucleolar foci as observed in squash preparations of bam>uno-EGFP testes labeled with a DNA stain. Scale bars = 5 μm. https://doi.org/10.1371/journal.pgen.1010547.g006 The extent of recovery of the weak diffuse nucleolar signals that was observed for the different AHC proteins was correlated with the distinct levels of these diffuse signals before photobleaching (Fig 6C). Compared to the average pixel intensity in the sub-nucleolar foci, the diffuse nucleolar signals were highest in case of SNM-EGFP (60%), intermediate for UNO-EGFP (48%) and lowest for MNM-EGFP (18%), when analyzed at the S5 stage, which was also studied in the FRAP experiments (Fig 6C). Interestingly, the diffuse nucleolar signals were dominant in early spermatocytes, and sub-nucleolar foci developed during spermatocyte maturation, as illustrated for UNO-EGFP (Fig 6D) and reported earlier for MNM- and SNM-EGFP [44]. Overall, the results of our FRAP experiments indicated that chromosome conjunction in the sex chromosome bivalents is provided by stable SUM protein complexes. Moreover, the decrease of the diffuse nucleolar signals during normal spermatocyte maturation is consistent with the notion that focal chromosomal SUM complexes are protected against the degradation that apparently occurs with unassembled SUM proteins. Proteolytic cleavage of UNO is sufficient to abolish chromosome conjunction Separase is required for homolog separation during M I in Drosophila males [30], and UNO includes a motif matching the consensus of separase cleavage sites (Fig 7A) [6,22]. We have previously shown that expression of UNOnc-EGFP, a mutant with a non-cleavable (nc) variant of the separase cleavage motif (E130A and R113A), instead of endogenous UNO (bam> unonc-EGFP in uno null mutants) prevented the rapid disappearance and homolog separation during anaphase I [22]. Thus, UNO cleavage by separase was proposed to abolish homolog conjunction during wild-type M I, so that homologs can be pulled apart to opposite spindle poles. To assess whether UNO cleavage is sufficient for elimination of homolog conjunction, we generated a mutant transgene (UASt-unoTEV-EGFP), in which the separase cleavage site was replaced by three repeats of a target sequence for tobacco etch virus (TEV) protease (Fig 7A). After expression of UNOTEV-EGFP instead of endogenous UNO (bam> unoTEV-EGFP in uno null mutants), UNOTEV-EGFP did not disappear rapidly during anaphase I (Fig 7C), in contrast to UNO-EGFP [22]. Moreover, homolog separation was inhibited and massive chromosome bridging was observed during telophase I (Fig 7C). Thus, behavior and effects of UNOTEV-EGFP were identical to those previously observed with UNOnc-EGFP in analogous experiments [22]. Since UNOTEV-EGFP no longer contained the separase cleavage site, this corresponded precisely to expectations. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. UNO cleavage is sufficient for elimination of homolog conjunction. (A) UNO contains a separase cleavage site preceded by a threonine (T128 in D. melanogaster) that is highly conserved among drosophilids [22], as illustrated with the sequence alignment. Positions essential for UNO cleavage are indicated by arrows. In the consensus sequence of separase cleavage sites [6,49], Φ indicates hydrophobic, X any amino acid and ζ hydrophilic residues. Separase cleaves C-terminally after R. The UNOTEV-EGFP mutant contains three repeats of the recognition sequence cleaved by TEV protease (grey shading), replacing the separase cleavage site. (B) Squash preparations uno null mutant testes with bam>unoTEV-EGFP and either no TEV transgene, exumP-TEV or betaTub85DP-TEV (bTub85DP-TEV) as indicated were labeled with a DNA stain and analyzed microscopically to reveal presence and localization of UNOTEV-EGFP during spermatocytes maturation. As indicated (dashed line) in the images with the apical testes regions on the left, a premature UNOTEV-EGFP disappearance was induced by the TEV transgenes. This disappearance occurred after the S3 stage, which still displayed a normal chromosome territory organization and UNOTEV-EGFP signals of normal level and localization in sub-nucleolar foci even in the presence of betaTub85DP-TEV (right panel). (C-E) Progression through M I in spermatocytes homozygous for the unocc1 null mutation with His2Av-mRFP and bam>unoTEV-EGFP was analyzed by time-lapse imaging. The spermatocytes expressed either no TEV transgene (C) or betaTub85DP-TEV (bTub85DP-TEV) (D). Still frames of representative spermatocytes are displayed with time indicated (min:sec) with t = 0 at the onset of NEBD I. Chromosome organization at NEBD I (number of major chromosome territories), during prometaphase I (presence of univalents) and telophase I (presence of chromosome bridges) was scored. The results are displayed in the bar diagram (E) with numbers of analyzed spermatocytes (n) indicated. (F) Summary of the protein-protein and protein-DNA interactions of the AHC proteins SNM, UNO, MNM and TEF. See discussion for the implications for the mechanism of alternative homolog conjunction. Scale bars = 20 (B, left), 5 (B, right), and 3 (C,D) μm. https://doi.org/10.1371/journal.pgen.1010547.g007 When a transgene driving expression of TEV protease specifically in late spermatocytes was added into the UNOTEV-EGFP expressing background, the resulting phenotype was very distinct. Two TEV transgenes, exumP-TEV and betaTub85DP-TEV, were generated and used. The former resulted in an onset of TEV protease effects at a slightly earlier stage compared to the latter (Fig 7B). The phenotypic effects were highly similar with both TEV transgenes and will be documented primarily for betaTub85DP-TEV. In uno null mutant spermatocytes with bam> unoTEV-EGFP and betaTub85DP-TEV, UNOTEV-EGFP disappearance occurred well after the S3 stage, and S3 spermatocytes still displayed UNOTEV-EGFP signals of normal intensity and normal localization in sub-nucleolar foci (Fig 7B). Moreover, these S3 spermatocytes also displayed normal chromosome territories (Fig 7B). Thus, betaTub85DP-TEV did not affect the initial establishment and maintenance of AHC. However, after the betaTub85DP-TEV-induced disappearance of UNOTEV-EGFP from late spermatocytes, bivalents were prematurely separated into univalents. Cytological analyses with testis squash preparations (S7 Fig) and time-lapse imaging (Fig 7D) revealed an increase in the number of chromosome territories in S6 spermatocytes, as well as the presence of univalents during prometaphase I. After a delayed onset of anaphase I, the univalents were segregated randomly onto the two spindle poles, without formation of chromosome bridges during telophase I (Figs 7D and S7). For a quantitative confirmation of the phenotypic differences resulting from absence or presence of a TEV transgene, we scored the number of major chromosome territories around NEBD I, the presence of univalents during prometaphase I and of chromosome bridges during telophase I after time-lapse imaging (Fig 7E). Overall, these results of our experiments with UNOTEV-EGFP suggested that proteolytic cleavage of UNO is sufficient for elimination of alternative homolog conjunction in spermatocytes. An additional control experiment ruled out that the premature separation of bivalents after co-expression of UNOTEV-EGFP and TEV protease in uno null mutants resulted from some unexpected dominant off-target effect of TEV protease unrelated to UNOTEV-EGFP cleavage. When UNOTEV-EGFP and TEV protease were co-expressed in spermatocytes that also produced UNO from the endogenous locus (bam> unoTEV-EGFP and betaTub85DP-TEV in uno-/+), bivalents were not separated prematurely (S7 Fig). Interestingly, in this uno-/+ background, UNOTEV-EGFP remained weakly detectable as a dot on the chrXY pairing region until anaphase I, while in the uno null mutant background, UNOTEV-EGFP disappearance was complete already before the S6 stage (S7 Fig). As discussed below, the longer perdurance of chromosomal UNOTEV-EGFP in the TEV expressing uno+ background appears to be consistent with a mechanism for chromosome linking involving multimerized SUM protein assemblies. Also consistent with this suggestion, we observed a perdurance of weak sub-nucleolar EGFP foci until M I after expression of UNO_N- or UNO_C-EGFP in uno+, but not in uno null mutant spermatocytes, with UASt transgenes and bamP-GAL4-VP16 (S8 Fig). Expression of these UNO fragments (UNO_N- and UNO_C-EGFP) did not rescue meiotic chromosome segregation in uno null mutants (S8 Fig), and in the uno+ background, they did not have an evident dominant-negative effect (S8 Fig), presumably as their level of expression was low according to EGFP signal intensities. Analogous expression of UNO_M-EGFP resulted in higher expression, but this fragment was present only transiently in early spermatocytes without any enrichment in the nucleolus (S8 Fig). In an attempt to address the control of the separase-mediated UNO cleavage during M I, we generated UASt transgenes expressing mutant UNO-EGFP versions with alterations in a conserved potential phosphorylation site preceding the separase cleavage site (Fig 7A). Experimental evidence from fungal and mammalian organisms [12,45–49] have emphasized that α-kleisin phosphorylation promotes cleavage by separase, in particular in case of the meiotic Rec8 isoforms. The UNO orthologs of Drosophila species all contain a threonine residue (T128 in D. melanogaster) followed by a proline residue immediately upstream of the separase cleavage site (Fig 7A). Phosphorylation of a serine at an equivalent position in a fungal α-kleisin or also its phosphomimetic mutation to glutamic acid was demonstrated to stimulate separase-mediated cleavage in vitro [49]. To evaluate the potential role of T128 phosphorylation in UNO, we mutated the corresponding codon to encode either alanine (T128A), which cannot be phosphorylated, or the phosphomimetic aspartic acid (T128D). Expression of UNOT128A-EGFP in uno null mutants (bam>unoT128A-EGFP in uno null) resulted in homolog conjunction that resisted elimination during M I (S9 Fig). While these results supported a potential significance of T128 phosphorylation, we observed an identical phenotype unexpectedly also with the phosphomimetic UNOT128D-EGFP (bam>unoT128D-EGFP in uno null) (S9 Fig). While our observations demonstrate that T128 is crucial for UNO cleavage by separase during M I, further work will be required to clarify the role of phosphorylation at this position. Discussion Drosophila male meiosis is achiasmate and therefore dependent on dedicated proteins (SNM, UNO and MNM, or SUM for abbreviation) that maintain conjunction between homologous chromosomes in replacement for the missing crossovers. Our main findings (summarized in Fig 7F) provide insight into the biochemical basis of (1) how the SUM proteins achieve this alternative homolog conjunction (AHC), and (2) how AHC is eliminated in time at the transition from metaphase to anaphase of M I to permit separation of the homologs to opposite spindle poles. In addition, our results are informative concerning the evolution of the AHC system. We demonstrate that SNM and the C-terminal domain of UNO form a stable heterodimeric complex (SU_C). Based on sequence comparisons, AlphaFold structural predictions and XL-MS with recombinantly expressed and purified proteins, the SU_C complex is homologous to that formed by stromalin and the stromalin-binding region of α-kleisin. Stromalins and α-kleisins are components of cohesin complexes [1]. While SNM was recognized as highly similar to stromalins early on [20], the very limited similarity of UNO to α-kleisins has escaped detection until now. The important and conserved N- and C-terminal domains of α-kleisins, which mediate its binding to the SMC heterodimer in cohesin, are not present in UNO. From an α-kleisin precursor, UNO has thus retained only the stromalin-binding region and the previously identified separase cleavage site [22]. Stromalin, via positively charged surface patches, has recently been shown to promote DNA-binding of cohesin in vitro [43]. We find that purified SU_C also binds DNA (Fig 7F). At least one of stromalin’s positively charged surface patches [43] is clearly also present in SNM and contributes to the DNA-binding of SU_C, according to our in vitro analysis with mutant versions of SU_C. In addition, a conspicuous, positively charged α-helix at the very C-terminus of UNO, which is absent from α-kleisins, makes a contribution to the DNA-binding of SU_C that is even more important than the basic SNM patch. Apart from DNA-binding, the interactions with the other AHC proteins were still normal in case of UNOchm-EGFP, a mutant with acidic or neutral residues in place of the six basic residues in the C-terminal α-helix. In vivo, UNOchm-EGFP displayed strongly reduced chromosome-binding and failed to provide normal AHC during male meiosis. These results strongly argue for the physiological importance of the DNA-binding activity of SU_C. We speculate that the C-terminal α-helix of UNO might clamp down on a DNA double helix bound to the basic surface patches of SNM and thereby strongly increase the strength of DNA-binding. The binding of SU_C to DNA does not appear to be sequence specific. Clearly, in competition with the scrambled DNA sequence, we have not detected increased binding to the 240 bp repeat sequence from the rDNA intergenic spacers, which appears to mediate sex chromosome conjunction [24–26]. Beyond DNA, SU_C binds to MNM (Fig 7F). Neither SNM nor UNO interact with MNM individually, indicating that prior association of SNM and UNO is required for MNM binding. These conclusions are based on our co-immunoprecipitation experiments after transient expression in S2R+ cells. Of note, we have not accomplished SUM complex formation with purified proteins in vitro so far. Our attempts at expressing and purifying full length MNM were not successful. Moreover, the successfully purified C-terminal region of MNM (MNM_C), which mediates the binding to SU_C (Fig 7F) according to our co-immunoprecipitation experiments, did not bind to SU_C in vitro. It is conceivable, therefore, that binding of MNM to SU depends on prior post-translational processing steps. At present, our inability to generate SUM complexes in vitro precludes a straightforward clarification of the issue whether SU can bind simultaneously to both MNM and DNA. However, the extended contacts between the C-terminal domain of UNO and SNM over long stretches (Fig 7F) provide ample space with interface potential, thereby increasing the likelihood of simultaneous binding of MNM and DNA to SU. MNM_C mediates binding not only to SU_C but also to TEF (Fig 7F), as revealed by the co-immunoprecipitation experiments reported here and previously [23]. The MNM-TEF interaction also remains to be re-constituted with purified proteins in vitro. However, in case of MNM_C, simultaneous binding of both TEF and SU_C is not feasible according to our co-immunoprecipitation experiments. Beyond the interaction domains discussed above, our analyses demonstrated the presence of multimerization domains in both UNO and MNM. We suggest that these domains are likely of crucial importance for the molecular mechanism whereby the SUM proteins generate AHC. In case of UNO, the N-terminal domain (UNO_N), which is highly conserved in UNO homologs, self-associates (Fig 7F), forming dimers and tetramers when expressed and purified from bacteria. This UNO_N region has a predicted structure that is very distinct from that of the conserved N-terminal region of α-kleisins, indicating that the evolution of uno involved substitution of N-terminal in addition to deletion of C-terminal coding sequences in an ancestral α-kleisin gene. Multimerization in case of MNM is also mediated by the N-terminal region MNM_N (Fig 7F). The primary sequence of MNM_N is identical to that of the N-terminal region present in the considerable number of alternative isoforms expressed from the complex mod(mdg4) locus. This common part of the Mod(mdg4) protein isoforms (thus also designated as Mod(mdg4)_CP) contains a BTB/POZ domain. This protein interaction domain present in many eukaryotic proteins with diverse functions has been shown to mediate homomeric dimerization [38,39]. In addition, in case of the particular type of BTB domain that is also present in the Mod(mdg4) protein products, heteromeric and higher order multimerization has been reported based on SEC, native gel electrophoresis and crosslinking studies [50]. Our results confirm and extend these findings. Purified MNM_N/Mod(mdg4)_CP was found to form stable hexamers according to SEC-MALS. The hexamers were readily modeled by AlphaFold2 as a ring-like complex with three dimers, and negative-stain electron microscopy revealed ring-like complexes of an appropriate dimension. A recent preprint describing similar structural analyses of the same type of BTB domains (i.e. the TTK-type) derived from other Drosophila proteins (Lola and CG6765) and also from Mod(mdg4) provides further confirmation of the ring-shaped hexameric structure [51]. Because of the multimerization domains (UNO_N and MNM_N/Mod(mdg4)_CP), the SUM proteins have presumably the potential to form extended protein assemblies that include many copies of the SU_C DNA-binding site (Fig 7F). Thereby, they might be empowered to effectively and stably conjoin distinct double-stranded DNA molecules (Fig 7F) and function as a chromosome glue. Accordingly, AHC would not rely on a topological ring-like embrace as proposed to be provided by cohesin in case of sister chromatid cohesion [1]. Since the tetramers and hexamers formed in vitro by purified UNO_N and MNM_N/Mod(mdg4)_CP are stable, SUM protein assemblies formed on bivalents in spermatocytes are expected to adopt a more solid rather than a liquid state. Indeed, our FRAP analyses confirmed that the SUM proteins in the dots associated with the sex chromosome pairing regions do not undergo dynamic exchange. The proposed extended SUM protein assemblies with their multitude of DNA-binding sites are unlikely to conjoin exclusively homologous DNA strands. Presumably, sister DNA strands are connected as well (and perhaps even neighboring regions on the same strand). Previous characterizations of meiotic mutant phenotypes are consistent with this view. Absence of AHC function results in premature separation of bivalents into univalents in late spermatocytes and early in M I [20,22]. The SOLO and SUNN proteins, which appear to function similar to the Rec8 cohesin complexes of other eukaryotes [52–54], still assure in these univalents a functional unification of sister centromeres for organization of a single kinetochore unit, as well as well as pericentromeric sister chromatid cohesion. In solo and sunn mutants, sister centromeres and pericentromeric regions lack cohesion, but bivalents are still present until the onset of anaphase I [52,53]. As sister chromatid cohesion within the regions of chromosome arms is normally lost after territory formation already during spermatocyte maturation, the presence of bivalents in solo and sunn mutants during early M I suggests that the SUM proteins conjoin not just homologous chromatids but also sister chromatids. In support of this interpretation, snm solo double mutants display univalents during early M I [52]. We emphasize that a chromosomal glue that conjoins DNA strands indiscriminately, as proposed for the SUM protein assemblies (i.e., sister strands and homologous strands in trans and perhaps also neighboring regions in cis) should be perfectly adequate if it is applied at the right time during spermatocyte maturation, i.e., after disruption of non-homologous chromosomal associations by territory formation but before complete disruption of homolog associations. Clearly, our proposal that AHC relies on extended assemblies of SUM proteins providing a high number of DNA-binding sites remains speculative and requires further investigation. For example, understanding how the formation of SUM protein assemblies is controlled and restricted to limited chromosomal regions will be crucial. The mechanism whereby SUM protein assemblies are targeted to the sex chromosome pairing rDNA loci on chromosome X and Y remains unexplained. In case of autosomal bivalents, TEF is likely involved in the initial establishment of SUM protein assemblies [23]. However, after ectopic expression in larval salivary glands, TEF as well as SUM bind to a large number of polytene chromosome bands [23]. In contrast, in mature S6 spermatocytes, the autosomal SUM protein assemblies are spatially restricted to one or two dots per chromosome arm [27]. Targeting of SUM protein assemblies to the sex chromosome pairing site and into autosomal dots might involve interactions with additional chromosomal proteins. Mod(mdg4)_T (also designated as 67.2 or 2.2), the most extensively characterized isoform expressed from the complex mod(mdg4) locus, interacts and co-operates with several chromatin architectural proteins (including CP190, HIPP1 and SuHw) at the gypsy insulator [37]. Moreover, Mod(mdg4)_T in combinations with chromatin architectural proteins in various combinations is generally enriched at boundaries between topologically associated chromatin domains and also at button loci that promote the somatic pairing of homologous chromosomes [55,56]. Recently, Mod(mdg4) function has been implicated in a striking example of chromosome pairing-dependent regulation of physiological gene expression [57]. Thus, multimerization by Mod(mdg4)_CP is likely crucial for chromosomal associations other than AHC during male meiosis. Accordingly, by recruitment of the Mod(mdg4)_H isoform MNM for AHC, evolution might have co-opted a pre-adaption that achieves chromosomal associations by Mod(Mdg4)_CP multimerization. While we found Mod(mdg4) isoforms other than MNM to be unable of binding to SU, these other isoforms clearly have the potential to form heteromeric associations with MNM according to our co-immunoprecipitation experiments. Moreover, based on yeast two-hybrid (Y2H) analyses, various other proteins with TTK-type BTB domains might also form heteromeric associations with MNM [51]. Whether such heteromeric interactions are relevant of AHC remains to be clarified. However, phenotypic analyses with various mod(mdg4) alleles have argued against contributions to AHC by Mod(mdg4) isoforms other than MNM [33]. Moreover, while heteromeric associations of Mod(mdg4)_T with other Mod(mdg4) isoforms can readily be detected by Y2H and co-immunoprecipitation after overexpression in S2 cells, their occurrence on chromosomes without overexpression is questionable according to chromatin-immunoprecipitation [58]. Importantly, AHC must provide conjunction between homologs in bivalents that is very robust and yet also amenable to rapid and complete elimination after biorientation of all the bivalents in the M I spindle, so that homologs can be separated to opposite poles during anaphase I. Efficient destructibility of AHC was achieved by the evolutionary co-option of the α-kleisin-derived protein UNO. Like α-kleisin, UNO includes a separase cleavage site that is highly conserved among UNO orthologs (Fig 7F) [22]. This cleavage site was shown to be required for AHC elimination and homolog separation during anaphase I [22]. Here, by exchanging the separase cleavage site in UNO with that cleaved by the bio-orthogonal TEV protease, we provide evidence that UNO cleavage is indeed sufficient to eliminate AHC. In our experiments, TEV was expressed under control of cis-regulatory sequences from exu or betaTub85D in mid spermatocytes. The presence of normal chromosome territories and of a normal subcellular localization of UNOTEV-EGFP at the onset of TEV expression indicated that this TEV expression occurred after successful AHC establishment, which occurs early during spermatocyte maturation [44]. However, as a consequence of TEV expression, bivalents were prematurely converted into univalents, as clearly indicated by cytological analyses and by time lapse imaging of progression into and through M I. UNO cleavage separates the multimerization domain UNO_N from UNO_C, which mediates DNA-binding in conjunction with SNM. Therefore, we propose that UNO cleavage dissociates the chromosomal SUM protein assemblies to an extent where the number of associated DNA-binding sites is no longer sufficient for tight linkage of distinct double-stranded DNA molecules. Clearly, alternative mechanisms of AHC elimination by UNO cleavage remain conceivable, and further work will be required to clarify the mechanistic details of AHC and its elimination. Materials and methods Plasmids For transient expression of proteins in S2R+ cells, we co-transfected pCaSper4-Act5C-GAL4 (kindly provided by Christian Klämbt, Westfälische Wilhelms-Universität, Germany) with pUASt constructs. Several of these pUASt constructs have been described previously: pUASt-snm, pUASt-mnm, pUASt-EGFP-snm, pUASt-snm-EGFP, pUASt-EGFP-mnm, pUASt-mnm-EGFP [44]; pUASt-uno-EGFP [22]; pUASt-snm-mCherry, pUASt-mnm-mCherry, pUASt-Mod(mdg4)_CP-mCherry, pUASt-Mod(mdg4)_C-mCherry, pUASt-Mod(mdg4)_P-mCherry, pUASt-Mod(mdg4)_T-mCherry, pUASt-teflon-10xmyc, pUASt-teflon-mCherry [23]. The plasmid pUASt-nls-tetR-EGFP was kindly provided by Stefan Heidmann (Universität Bayreuth, Bayreuth, Germany). The plasmid pUASt-uno-myc was generated by amplifying the uno coding region from pUASt-uno-EGFP with the primer pair OL005/ZK017 (see S1 Table for all oligonucleotide sequences). After digestion of the resulting fragment with EcoRI, it was inserted into the corresponding site of pUASt-mcs-10xmyc [23]. To generate pUASt-mnm-myc, pUASt-mcs-10xmyc was first modified. By digestion with EcoRI, followed by insertion of a linker obtained by annealing LV026/LV027, the EcoRI site was converted into a NotI site. By mutagenic plasmid amplification [59] with MS018 as primer the reading frame between the NotI site and the region coding for the myc epitopes was adjusted. Thereafter the mnm coding region was isolated from pUASt-mnm-EGFP as a NotI fragment and inserted into the corresponding site of the modified vector. To generate pUASt-myc-mad1_C, we enzymatically amplified the C-terminal mad1 coding region (aa 500–730) from the cDNA clone GM14169 [60] using AF65/AF66. The fragment was digested with NotI and Acc65I, followed by ligation into the corresponding sites of pUASt, yielding pUASt-Mad1_C. For insertion of the sequences coding for 10 copies of the myc epitope, we isolated a NotI fragment from pUASt-10myc-Mps1 [61] and inserted it into the corresponding site of pUASt-Mad1_C. For construction of pUASt-mnm_C-mCherry, we enzymatically amplified the C-terminal mnm region from pUASt-mnm-EGFP with ZK055/AF049. After digestion with NotI, we inserted the fragment in the corresponding site of pUASt-mcs-mCherry [23]. Additional constructs were made starting from pUASt-attB [62]. In a first step, the derivative pUASt-attB-mcs-EGFP was generated by enzymatic amplification of the EGFP coding sequence with the primer pair RAS079/RAS080, followed by digestion with KpnI/XbaI and insertion into the corresponding restriction sites of pUASt-attB. In a second step, insert fragments coding for UNO_N, UNO_M and UNO-C were amplified from pUASt-uno-EGFP with the primer pairs JW082/ZK013, ZK014/ZK015 and ZK016/JW083, respectively. After digestion with BglII/XhoI, the fragments were inserted into the corresponding sites of pUASt-attB-mcs-EGFP, yielding pUASt-attB-uno_N-EGFP, pUASt-attB-uno_M-EGFP and pUASt-attB-uno_C-EGFP. For the construction of pUASt-attb-uno_N-mCherry, we exchanged the region coding for EGFP with that encoding mCherry. The pUASt-attB derivatives for the expression of either UNO, UNOTEV, UNOT128A, UNOT128D and UNOchm were made starting from pUASt-attB-mCherry-uno-EGFP [23]. In a first step, the mCherry coding region was excised with EcoRI/BglII and replaced with a double-stranded DNA oligonucleotide generating by annealing of the oligos CL338/CL339, yielding pUASt-attB-uno-EGFP. For the generation of pUASt-attB-unoTEV-EGFP, we replaced its EcoRI—AgeI fragment with the synthetic DNA fragment CL342 after digestion with the same restriction enzymes. In case of pUASt-attB-unoT128A-EGFP and pUASt-attB-unoT128D-EGFP, we also replaced the EcoRI—AgeI fragment of pUASt-attB-unoTEV-EGFP with the EcoRI/AgeI-digested synthetic DNA fragments CL340 and CL341, respectively. In case of pUASt-attb-unochm-EGFP, the BsiWI—XhoI fragment of pUASt-attB-uno-EGFP was replaced with the BsiWI/XhoI-digested synthetic DNA fragment CL419. To generate exumP-TEV transgenic flies, we modified pattB-exumP-EGFP [22] by replacing the EGFP coding region with that encoding TEV. The region coding for TEV with an N-terminal SV40 nuclear localization signal and a V5 epitope tag as well as two C-terminal SV40 nuclear localization signals was excised with EcoRI/NotI from pUASp1-TEV, a plasmid identical to that described earlier [63] except that it did not contain the S219V mutation. Thereafter, the EcoRI—NotI fragment was used to replace the EcoRI—NotI fragment of pattB-exumP-EGFP. The plasmid pattB-betaTub85DP-TEV contained cis-regulatory 5’ and 3’ sequences of the spermatocyte-specific betaTub85D gene, as also present in pattB-Nslmb-vhh4-GFP4 [44]. The region coding for TEV and the N- and C-terminal extension described above was enzymatically amplified with SCH035/NT022 and inserted between the 5’ and 3’ betaTub85D sequences after digestion with KpnI/NotI. All constructs for expression of recombinant proteins were cloned using the InteBac system [64] for either insect cells or bacterial expression. Drosophila lines Several lines with mutations or transgenes that we have used for our analyses have been described earlier: UASt-snm-EGFP and UASt-mnm-EGFP [44], UASt-uno-EGFP and unocc1 [22], UASt-snm-mCherry [23], g-His2Av-mRFP [65], bamP-GAL4-VP16 [66], Sgs3-GAL4 (Bloomington Drosophila Stock Center (BDSC) #6870), Df(2R)Exel7094 (BDSC #7859). Fly lines carrying the transgenes UASt-unochm-EGFP, UASt-unoTEV-EGFP, UASt-unoT128A-EGFP and UASt-unoT128D-EGFP were generated (BestGene Inc., Chino Hills, CA, USA) by integration of the corresponding pUASt-attB constructs described above into the landing site P{CaryP}attP2. For production of fly lines with the transgenes exumP-TEV, betaTub85DP-TEV, UASt-uno_N-EGFP, UASt-uno_M-EGFP and UASt-uno_C-EGFP with the attB plasmids described above, we used the integration site P{CaryP}attP40. Standard crossing and generation of recombinant chromosomes were used to produce the various strains used for experimental analyses. The genotypes of the flies analyzed are described in detail in the supporting information (S2 Table). Bioinformatic analyses of the predicted UNO amino acid sequence UNO orthologs were collected with NCBI blast within the NCBI non-redundant protein database or the UniProt reference proteomes applying significant E-value thresholds below 0.001 [67]. We used HHPRED with alignments of the N- and C-terminal conserved regions of UNO for remote homology detection in the PDB and PFAM database [68]. The search with the N-terminal conserved domain was not informative, but the C-terminal region of UNO (aa 289–364) hit with a probability of 83.2 to the middle region of human Rad21 (aa 320–394) from the structure 4PJW [35]. The C-terminal hit region of UNO was aligned with the conserved middle region of Rad21, Rad21L and Rec8 orthologs using mafft (-linsi, v7.427) [69] and visualized with Jalview [70]. Cell culture and transfection S2R+ cells were cultured in Schneider’s medium (Gibco, cat# 21720, Thermo Fisher Scientific, Waltham, MA), 10% fetal bovine serum (Gibco, cat# 10500–064) and 1% Penicillin-Streptomycin (Gibco, cat# 15140) at 25°C. Transfections were performed using FuGENE HD (Promega, cat# E2311) in 6-well plates, T25 or T75 flasks. In case of 6-well plates, 1.2x106 cells were plated into one well in 2 ml complete medium. In T25 flasks, 5.2×106 cells were plated in 4 ml complete medium, and in T75 flasks, 15.6×106 cells in 8 ml complete medium. One hour after plating transfection mix was added. In case of 6-well plates, 100 μl transfection mix containing 1 μg plasmid DNA and 4 μl FuGENE HD in Schneider’s medium were added. For T25 flasks, 200 μl of transfection mix was used containing 2 μg plasmid DNA and 8 μl FuGENE HD in Schneider’s medium, and for T75 flasks, 400 μl of transfection mix containing 4 μg plasmid DNA and 16 μl FuGENE HD in Schneider’s medium. Cells were incubated for 2 days before analysis in co-immunoprecipitation experiments. Immunoprecipitation and immunoblotting For the analysis of whole cell lysates by immunoblotting (S1 and S3 Figs), transfected cells in T25 flasks or 6-well plates were transferred on ice and carefully washed twice with cold phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 1.47 mM KH2PO4, 6.46 mM Na2HPO4, pH 7.4). Using a scraper, cells were harvested with a total of 150 μl resp. 60 μl of 3x Lämmli buffer (62.5 mM Tris-HCl, 10% glycerol, 5% β-mercaptoethanol, 3% SDS, 0.01% Bromophenol Blue) and the mixture was transferred to Eppendorf tubes. All samples were boiled for 8 minutes at 96°C, aliquoted, snap frozen in liquid nitrogen and stored at -80°C until analysis by immunoblotting. For immunoprecipitation, the transfected S2R+ cells were detached with a cell scraper and the resulting cell suspension was centrifuged at 580 x g for 5 minutes in a 15 ml Falcon tube. The cell pellet was washed with 1 ml cold phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 1.47 mM KH2PO4, 6.46 mM Na2HPO4, pH 7.4) and transferred to a 1.5 ml Protein LoBind tube (Eppendorf, cat# 022431081), which were also used for all subsequent steps. Cells were sedimented at 600 x g for 5 minutes at 4°C. All subsequent steps were performed on ice or at 4°C with ice cold solutions. Cells were lysed in lysis buffer (LB): 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 2 mM MgCl2, 0.1% Nonidet P-40 Substitute (Sigma Aldrich, cat# 74385), 5% glycerol, 0.5 mM EGTA, 1 mM DTT, 50 U/ml Benzonase Nuclease ultrapure (Sigma Aldrich, cat# E8263) and 1 tablet Roche protease inhibitor c0mplete per 10 ml lysis buffer (Mini EDTA-free, EASYpack, Roche, cat# 04693159001). The cells harvested from a T25 flask were lysed in 500 μl LB and in 1000 μl LB in case of T75 flasks by pipetting up and down twice, each time followed by a 15-minute incubation. Cell lysates were cleared by centrifugation for 15 minutes at 16100 x g. A small aliquot of the supernatant was removed for analysis by immunoblotting (input samples). The rest of the supernatant was added to pre-washed 25 μl nano-trap agarose beads (GFP-Trap agarose, ChromoTek, cat# gta-20, or RFP-Trap agarose, ChromoTek, cat# rta-20, or MYC-Trap agarose beads ChromoTek, cat# yta-20). Beads were incubated for 1 hour on a rotating wheel at 15 rpm. Thereafter, beads were washed 3 times with LB and centrifugation at 2500 x g for 2 minutes. For the final wash samples were transferred into a fresh tube. For elution, beads were resuspended in 80 μl 3x Lämmli Buffer (62.5 mM Tris-HCl, 10% glycerol, 5% β-mercaptoethanol, 3% SDS, 0.01% Bromophenol Blue) and boiled for 8 minutes at 96°C. After rapid cooling on ice, beads were sedimented and the supernatant was distributed in three aliquots of 25 μl (IP samples). All samples were snap frozen in liquid nitrogen and stored at -80°C until analysis by immunoblotting. Samples were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using a mini-gel system (BioRAD) and gels with between 7% and 12% polyacrylamide for 3 hours at 90 V. Molecular weight markers were PageRuler Plus Prestained Protein Ladder (Thermo Scientific, cat# 815-968-0747). Protein transfer to nitrocellulose membranes (Amersham Protran 0.45 μm, cat# 10600002) was achieved by tank blotting at room temperature for 1 hour with 100 V. After transfer, Ponceau S staining was performed, followed by a blocking step using 5% dry milk (w/v) in PBS with 0.02% NaN3. This solution was also used for incubation with the primary antibody for 2 hours at room temperature or overnight at 4°C. Three washes with 5% dry milk (w/v) in PBS were done before incubation with the secondary antibody, which was applied in 5% dry milk (w/v) in PBS at room temperature for 1 hour protected from light. After 2 washes with 5% dry milk (w/v) in PBS and 2 washes with PBS, 0.1% Tween-20, signals were detected with ECL reagents (WesternBright ECL, Advansta, cat# K-12045-D50) in an Amersham Imager 600. The following antibodies were used for immunoblotting: rabbit polyclonal antibodies anti-GFP diluted 1:800 (ChromoTek, cat# pabg1) or 1:2000 (Torrey Pines Biolabs, cat# TP401); mouse monoclonal antibody anti-RFP (ChromoTek, cat# 6g6), 1:1000; rat monoclonal antibody anti-c-MYC (ChromoTek, cat# 9e1) at 1:1000; rabbit polyclonal antibody anti-ModC [13] (kindly provided by Rainer Dorn, Universität Halle, Halle, Germany) at 1:4000; rabbit polyclonal antibody anti-SNM C-terminal peptide (DIAHLKEYRNALRPRKTKSYPQAT) [20] (kindly provided by Bruce McKee, University of Tennessee, Knoxville, TE, USA); rabbit polyclonal antibody anti-UNO [22]; HRP-conjugated AffiniPure goat anti-rabbit IgG polyclonal antibody (Jackson ImmunoResearch, cat# 111-035-003) at 1:1000; HRP-conjugated AffiniPure goat anti-mouse IgG polyclonal antibody (Jackson ImmunoResearch, cat# 115-035-003) at 1:1000; HRP-conjugated goat anti-rat IgG antibody (Thermo Scientific, cat# 62–9520) at 1:5000. We note that after transient expression of TetR-E and immunoblotting with anti-EGFP, we always detected two bands for unknown reasons. Protein expression and purification For expression of SNM and UNO_C, baculovirus was generated as previously described [71]. Briefly, EMBacY cells were transformed with constructs for either UNO_C-Strep or MBP-SNM. Bacmids were extracted from positive transformants and used to transfect Sf9 cells. Three rounds of viral amplification were performed to produce a “V2” virus suspension. For each SNM/UNO_C prep, 2 x 400 ml flasks of Hi5 cells were infected with V2 at a 1:100 dilution (i.e. 4 ml of SNM V2 virus and 4 ml of UNO_C V2 virus), and incubated at 27°C for 72 hours. Cells were harvested, washed with PBS, and flash frozen in liquid nitrogen. For protein purification, cells were resuspended in lysis buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 0.1% Triton X-100, 1 mM MgCl2 + 5 mM BME, DNAse, SERVA protease inhibitor) and lysed by sonication. Cleared lysate was loaded onto a 5 ml Strep Tactin Superflow Plus column (Qiagen), washed with lysis buffer, and eluted with elution buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 1 mM MgCl2 + 5 mM BME, 2.5 mM Desthiobiotin). Fractions corresponding to the SNM-UNO_C complex (SU_C) were concentrated with a 10 kDa MWCO concentrator (Pierce) and loaded onto a Superose 6 10/300 column equilibrated with storage buffer (25 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 1 mM MgCl2 + 5 mM TCEP). Eluted fractions were concentrated and snap frozen. N-terminally 6xHis-MBP tagged UNO_N was expressed in BL21 cells at 18°C for 16 hours grown in TB. For each purification, cells from a 1L culture were used. Harvested cells were resuspended in lysis buffer (50mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 0.1% Triton X-100, 1 mM MgCl2, 5 mM β-mercaptoethanol, DNAse, SERVA protease) using 5 ml/g pellet. Resuspended bacteria were lysed by two passages through an Emulsiflex (Avestin). Cleared lysate was passed over a 5 ml MBP Trap column (Cytiva) equilibrated in MBP-Trap A buffer (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 5 mM β-mercaptoethanol) at 2.5 ml /minute. MBP-Trap was washed with 15 CV of MBP-Trap A buffer followed by elution with 10 CV of MBP-Trap buffer B (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 1 mM Maltose 5 mM β-mercaptoethanol) collecting 5 ml fractions. Fractions corresponding to MBP-UNO_N were pooled and the NaCl concentration adjusted to 100 mM. Sample was then loaded onto a 6 ml ResourceQ column (Cytiva) pre-equilibrated with ResQ Buffer A (50 mM Na-HEPES pH7.5, 100 mM NaCl, 5% glycerol, 5 mM β-mercaptoethanol). Unbound sample was washed away with 10 CV of ResQ Buffer A. MBP-UNO_N was eluted by running a gradient of 0–60% ResQ Buffer B (50 mM Na-HEPES pH7.5, 1 M NaCl, 5% glycerol, 5 mM β-mercaptoethanol) over 30 CVs. Fractions corresponding to MBP-UNO_N were concentrated in a Amicon 30 kDa MWCO concentrator to a volume of ~1.5 ml. Concentrated protein was loaded onto a Superdex 200 16/600 column, pre-equilibtated with SEC Buffer (50mM Na-HEPES pH7.5, 300mM NaCl, 5% Glycerol, 1mM TCEP) and run at 1 ml/minute. Fractions corresponding to the two resulting UNO_N peaks were pooled, concentrated, flash frozen, and stored at -80°C. N-terminally 6xHis-MBP tagged MNM_N was expressed in BL21 cells at 18°C for 16 hours grown in TB. For each purification, cells from a 1L culture were used. Harvested cells were resuspended in lysis buffer (50mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 0.1% Triton X-100, 1 mM MgCl2, 5 mM β-mercaptoethanol, DNAse, SERVA protease) using 5 ml/g pellet. Resuspended bacteria were lysed by two passages through an Emulsiflex (Avestin). Cleared lysate was passed over a 5 ml MBP Trap column (Cytiva) equilibrated in MBP-Trap A buffer (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 5 mM β-mercaptoethanol) at 2.5 ml /minute. MBP-Trap was washed with 15 CV of MBP-Trap A buffer followed by and additional wash step with buffer containing 1 mM ATP, followed by elution with 10 CV of MBP-Trap buffer B (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 1 mM Maltose 5 mM β-mercaptoethanol) collecting 5 ml fractions. Fractions corresponding to MBP-MNM_N were pooled and the NaCl concentration adjusted to 100 mM. Sample was then loaded onto a 6 ml ResourceQ column (Cytiva) pre-equilibrated with ResQ Buffer A (50 mM Na-HEPES pH7.5, 100 mM NaCl, 5% glycerol, 5 mM β-mercaptoethanol). Unbound sample was washed away with 10 CV of ResQ Buffer A. MBP-MNM_N was eluted by running a gradient of 0–100% ResQ Buffer B (50 mM Na-HEPES pH7.5, 1 M NaCl, 5% glycerol, 5 mM β-mercaptoethanol) over 10 CVs. Fractions corresponding to MBP-MNM_N were concentrated in a Amicon 30 kDa MWCO concentrator to a volume of ~1.5 ml. The MBP moiety was removed by the addition of GST-3C protease, followed by incubation on ice for 3 hours. Concentrated protein was loaded onto a Superdex 200 16/600 column, pre-equilibtated with SEC Buffer (50mM Na-HEPES pH7.5, 300mM NaCl, 5% Glycerol, 1mM TCEP) and run at 1 ml/minute. To ensure removal of the MBP and GST-3C protease from the sample a 5 ml MBP-Trap and a 5 ml GST-Trap column were connected in-line and downstream of the Superdex 200 16/600 column Fractions corresponding to the two resulting MNM-N peaks were pooled, concentrated in an Amicon 10 kDa MWCO concentrator, flash frozen, and stored at -80°C. Size exclusion chromatography SEC experiments were run on an Akta Pure 25 system (Cytiva). Absorbances were measured at both 280 nm (blue traces) or 254 nm (red traces). Molecular weight markers (Bio-Rad) were used as a reference (grey dotted traces). Samples from SEC experiments were collected and run on SDS-PAGE gels stained with Coomassie brilliant blue. SEC-MALS 50 μL samples at 5–10 μM concentration were loaded onto a Superose 6 5/200 (run at 0.3 ml/min) or Superdex 75/150 (run at 0.5 ml/min) analytical size exclusion column (Cytiva) equilibrated in buffer containing 50 mM HEPES pH 7.5, 1 mM TCEP, 300 mM NaCl (for samples without nucleosomes) or 150 mM NaCl (for samples with nucleosomes) attached to an 1260 Infinity II LC System (Agilent). MALS was carried out using a Wyatt DAWN detector attached in line with the size exclusion column. Cross-linking mass spectrometry (XL-MS) For XL-MS analysis proteins were diluted in 200 μL of XL-MS buffer (30 mM HEPES 6.8, 150 mM NaCl, 5% glycerol, 1 mM MgCl2, 1 mM TCEP) to the final concentration of 3 μM, mixed with 3 μL of DSBU (200 mM) and incubated for 1 hour at 25°C. The reaction was stopped by adding 20 μL of 1 M Tris-HCl pH 8.0 and incubated for another 30 min at 25°C. The cross-linked sample was precipitated by addition of 4X volumes of 100% cold acetone followed by overnight incubation at -20°C. Samples were analyzed as previously described [72]. For interaction network visualization XVis software was used and for visualization of the crosslinks on the PDB model PyXlinkViewer [73] and XMAS [74] was used. Each time a different cutoff for the cross-linking credibility was selected depending on the quality of the cross-linking data. Alphafold2 predictions Predicted structures were calculated using AlphaFold Multimer (2.2.0) [40] run on GPU nodes of the Raven HPC of the Max Planck Computing and Data Facility (MPCDF), Garching. Each job was run on a single node consisting of 4 x Nvidia A100 NVlink 40 GB GPUs. Multiple predictions were generated for each run, and the best model (determined by pTM score) was then used. PAE plots were generated using a custom script (Vikram Alva, MPI Biology Tübingen). Mass photometry Mass Photometry was performed in the mass photometry buffer (MP) containing 30 mM HEPES pH 7.8, 150 mM NaCl, 5% glycerol, 1 mM MgCl2, and 1 mM TCEP. Protein samples (3 μM) were pre-equilibrated for 1 hour in the MP buffer. Measurements were performed using Refeyn One (Refyn Ltd., Oxford, UK) mass photometer. Directly before the measurement, the sample was diluted 1:100 with the MP buffer. Molecular mass was determined in Analysis software provided by the manufacturer using a NativeMark (Invitrogen) based standard curve created under the identical buffer composition. Negative-stain electron microscopy 4 μl of MNM at 27 μg/ml were adsorbed at 25°C for 2 minutes onto glow-discharged carbon-coated grids. The grids were washed three times with water and negatively stained with three washes of 1% uranyl acetate, followed by a 5-minute incubation at 25°C. Samples were imaged with a Tecnai G2 Spirit BioTWIN microscope equipped with a LaB6 cathode operated at 120 kV. Images were recorded at low-dose conditions (19 electrons/Å2) at a corrected magnification of 82553x on a 4k × 4k CMOS camera F416 (TVIPS, Oslo, Norway). EMSAs The binding reactions (10 μL volume) were carried out in EMSA buffer (25 mM HEPES pH 7.5, 0.1 μg/μL BSA, 60 mM NaCl) containing indicated fluorescently labelled DNA substrate (10 nM). The reactions were started by addition of increasing amounts of SU_C protein complexes (36.25, 72.5, 108.75, 145, 217.5, 290, 435, 580, 870, 1160 and 1740 nM) and incubated for 20 min at 30°C. After the addition of 2 μL of the gel loading buffer (60% glycerol, 10 mM Tris–HCl, pH 7.4, 60 mM EDTA, 0.15% Orange G), the reaction mixtures were resolved in 0.8% agarose gel in 1x TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA). The gels were scanned using Amersham Typhoon scanner (Cytiva) and quantified in ImageJ. Microscopic analyses with larval salivary glands By standard crossing, we combined the Sgs3-GAL4 driver with UASt transgenes. Wandering third instar larvae were used for dissection of salivary glands after development at 25°C. Salivary gland preparations were made as described [75,76] with the modifications reported previously [23]. Imaging and signal quantification was also performed as described [23]. Microscopic analyses with testis preparations For whole-mount testis preparations, dissections from young adult males (0–1 day after eclosion) were performed in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl, pH 6.8). Testes were fixed in PBST containing 4% formaldehyde in 0.2 ml Eppendorf tubes for 20 minutes on a rotating wheel. Testis squash preparations were made essentially as described previously [77]. For DNA staining, testes were incubated for 10 minutes in PBS, 0.1% Triton X-100 (PBTx) containing Hoechst 33258 (1 μg/ml). After three washes with PBS, preparations were mounted under a coverslip in a drop of mounting medium. Microscopic quantification of the DNA content of nuclei of early round spermatids was done as described [44]. Preparations of ovaries and testes were analyzed with a wide-field fluorescence microscope (Zeiss Axio Observer HS) using 40×/1.3, 63×/1.4 and 100×/1.4 oil immersion objectives. Maximum intensity projections of image stacks are presented. Time-lapse imaging of progression through meiosis was performed as described [78]. Testes were dissected from pupal or young adult males in Schneider’s Drosophila Medium (Invitrogen, #21720) supplemented with 10% fetal bovine serum (Invitrogen) and 1% penicillin/streptomycin (Invitrogen, #15140). The dissected pupal testes were transferred into 45 μl of medium in a 35 mm glass bottom dish (MatTek Corporation, #P35G-1.5-14-C) and opened with fine tungsten needles to release the cysts. In case of adult testes, 150 μl of medium were used. To reduce sample movements, 15 μl of 1% w/v methylcellulose (Sigma, #M0387) was added to pupal testes preparations and 50 μl to adult testes preparations. A wet filter paper was placed inside along the dish wall before sealing the lid with parafilm. Imaging was performed at 25°C in a room with temperature control using a spinning disk confocal microscope (VisiScope with a Yokogawa CSU-X1 unit combined with an Olympus IX83 inverted stand and a Photometrics evolve EM 512 EMCCD camera, equipped for red/green dual channel fluorescence observation; Visitron systems, Puchheim, Germany). A 60×/1.42 oil immersion objective was used. We acquired z-stacks with 30–40 focal planes spaced by 500 nm at 45-second intervals. Maximum intensity projections were generated using ImageJ or ZEN software for wide-field images and IMARIS (Bitplane) for spinning disk confocal images. Figures display maximum intensity projections unless stated otherwise. Export of projections from IMARIS as movies or still frames after live imaging was made with interpolated image display. Moreover, display parameters for the His2Av-mRFP were adjusted manually over time to reveal chromosomes clearly throughout the movies, thereby correcting photobleaching and partially also the changes in the extent of chromosome condensation during M I. Graphs were generated with Microsoft Excel or GraphPad Prism. P values were calculated using a two-tailed student t-test (* = p < 0.05; ** = p < 0.01; *** = p < 0.001). Adobe Photoshop and Adobe Illustrator were used for production of figures. FRAP analyses Testes were isolated from early pupae and cysts were released for imaging in dishes with a glass coverslip at the bottom as described [78]. Before bleaching, we acquired five z-stacks with 40 optical sections spaced by 500 nm at one-minute intervals using a FV1000 Olympus laser scanning confocal microscope with a PLAPON 60XO/1.42 Objective with a zoom factor of 5.4. Thereafter, we photobleached a part of the nucleolus in one of the spermatocytes by using an circular region of interest (diameter = 30 pixels) and 100 iterations of tornado scanning with maximal 488 nm laser intensity within one z section. After photobleaching the EGFP signals in part of a nucleolus, imaging of z-stacks was continued, initially as before photobleaching. However, after five z-stacks, the time interval between z-stack acquisition was increased from one to 15 minutes. In case of long-term FRAP analyses with bam>mnm-EGFP spermatocytes, we acquired only three z-stacks at one-minute intervals immediately after photobleaching, followed by acquisition of three additional z-stacks at one-minute intervals four hours later. For the quantitative analysis of EGFP signal recovery after photobleaching over time (S6 Fig), we used spot detection by IMARIS software to identify spheres containing the bleached nucleolus or unbleached nucleoli in neighboring spermatocytes of the imaged cyst. After creating a first set of spheres with a diameter of five μm, a second set of spheres with a diameter of seven μm was generated for background correction of signal intensities in the smaller spheres. For further processing with Microsoft Excel, the signal intensities detected in the spheres were exported from IMARIS. The difference in signal intensities observed in the large and small sphere, respectively, was used for estimation of background signal intensity, which was subtracted from the total intensity value within the small sphere. Moreover, fluorescence intensities were normalized to the average detected during the five z-stacks acquired before the bleaching of a nucleolus. For the quantification of the overall recovery of EGFP signals 90 minutes after photobleaching (Fig 6A and 6B), we generated intensity sum projections using Image J. Representative projection images are shown in Fig 6A. A circular ROI with a diameter of 30 pixels, as used before for photobleaching, was placed over the bleached region. Moreover, a bean-shaped ROI covering the unbleached part of the targeted nucleolus was selected manually. Average pixel intensities in these ROIs were quantified for the five pre-bleaching time points, as well as for the first and last post-bleaching time points. Pixel intensities were normalized to the average of the five pre-bleaching time points. The intensity difference between the first and last post-bleaching time points observed in the non-bleached and bleached parts of the nucleolus were compared to estimate the extent of FRAP corrected for photobleaching during image acquisition after pulse-bleaching of a part of a nucleolus. To express the extent of FRAP after 90 minutes in percent of the total signal intensity loss induced by the pulse-bleaching within the bleached part of the nucleolus (Fig 6B), this total signal intensity loss was calculated as the difference between the intensities at the first post-bleaching time point in the non-bleached and bleached region, respectively, of the targeted nucleolus. To compare the intensity of EGFP signals that were either diffusely distributed throughout the nucleolus or within sub-nucleolar foci, we analyzed intensity sum projections of the pre-bleaching z-stacks with Image J. The projection images were segmented by using the threshold tool of Image J and selecting first the top 2% intensity pixels, which represented the sub-nucleolar foci well. Thus, the resulting intensity values were used as a measure of signal intensity in sub-nucleolar foci. Thereafter, a second segmentation was applied for selection of the top 10% intensity pixels, which covered the complete nucleoli. To estimate the diffuse nucleolar signals, we subtracted the intensities detected within the sub-nucleolar foci from those of the complete nucleoli. Plasmids For transient expression of proteins in S2R+ cells, we co-transfected pCaSper4-Act5C-GAL4 (kindly provided by Christian Klämbt, Westfälische Wilhelms-Universität, Germany) with pUASt constructs. Several of these pUASt constructs have been described previously: pUASt-snm, pUASt-mnm, pUASt-EGFP-snm, pUASt-snm-EGFP, pUASt-EGFP-mnm, pUASt-mnm-EGFP [44]; pUASt-uno-EGFP [22]; pUASt-snm-mCherry, pUASt-mnm-mCherry, pUASt-Mod(mdg4)_CP-mCherry, pUASt-Mod(mdg4)_C-mCherry, pUASt-Mod(mdg4)_P-mCherry, pUASt-Mod(mdg4)_T-mCherry, pUASt-teflon-10xmyc, pUASt-teflon-mCherry [23]. The plasmid pUASt-nls-tetR-EGFP was kindly provided by Stefan Heidmann (Universität Bayreuth, Bayreuth, Germany). The plasmid pUASt-uno-myc was generated by amplifying the uno coding region from pUASt-uno-EGFP with the primer pair OL005/ZK017 (see S1 Table for all oligonucleotide sequences). After digestion of the resulting fragment with EcoRI, it was inserted into the corresponding site of pUASt-mcs-10xmyc [23]. To generate pUASt-mnm-myc, pUASt-mcs-10xmyc was first modified. By digestion with EcoRI, followed by insertion of a linker obtained by annealing LV026/LV027, the EcoRI site was converted into a NotI site. By mutagenic plasmid amplification [59] with MS018 as primer the reading frame between the NotI site and the region coding for the myc epitopes was adjusted. Thereafter the mnm coding region was isolated from pUASt-mnm-EGFP as a NotI fragment and inserted into the corresponding site of the modified vector. To generate pUASt-myc-mad1_C, we enzymatically amplified the C-terminal mad1 coding region (aa 500–730) from the cDNA clone GM14169 [60] using AF65/AF66. The fragment was digested with NotI and Acc65I, followed by ligation into the corresponding sites of pUASt, yielding pUASt-Mad1_C. For insertion of the sequences coding for 10 copies of the myc epitope, we isolated a NotI fragment from pUASt-10myc-Mps1 [61] and inserted it into the corresponding site of pUASt-Mad1_C. For construction of pUASt-mnm_C-mCherry, we enzymatically amplified the C-terminal mnm region from pUASt-mnm-EGFP with ZK055/AF049. After digestion with NotI, we inserted the fragment in the corresponding site of pUASt-mcs-mCherry [23]. Additional constructs were made starting from pUASt-attB [62]. In a first step, the derivative pUASt-attB-mcs-EGFP was generated by enzymatic amplification of the EGFP coding sequence with the primer pair RAS079/RAS080, followed by digestion with KpnI/XbaI and insertion into the corresponding restriction sites of pUASt-attB. In a second step, insert fragments coding for UNO_N, UNO_M and UNO-C were amplified from pUASt-uno-EGFP with the primer pairs JW082/ZK013, ZK014/ZK015 and ZK016/JW083, respectively. After digestion with BglII/XhoI, the fragments were inserted into the corresponding sites of pUASt-attB-mcs-EGFP, yielding pUASt-attB-uno_N-EGFP, pUASt-attB-uno_M-EGFP and pUASt-attB-uno_C-EGFP. For the construction of pUASt-attb-uno_N-mCherry, we exchanged the region coding for EGFP with that encoding mCherry. The pUASt-attB derivatives for the expression of either UNO, UNOTEV, UNOT128A, UNOT128D and UNOchm were made starting from pUASt-attB-mCherry-uno-EGFP [23]. In a first step, the mCherry coding region was excised with EcoRI/BglII and replaced with a double-stranded DNA oligonucleotide generating by annealing of the oligos CL338/CL339, yielding pUASt-attB-uno-EGFP. For the generation of pUASt-attB-unoTEV-EGFP, we replaced its EcoRI—AgeI fragment with the synthetic DNA fragment CL342 after digestion with the same restriction enzymes. In case of pUASt-attB-unoT128A-EGFP and pUASt-attB-unoT128D-EGFP, we also replaced the EcoRI—AgeI fragment of pUASt-attB-unoTEV-EGFP with the EcoRI/AgeI-digested synthetic DNA fragments CL340 and CL341, respectively. In case of pUASt-attb-unochm-EGFP, the BsiWI—XhoI fragment of pUASt-attB-uno-EGFP was replaced with the BsiWI/XhoI-digested synthetic DNA fragment CL419. To generate exumP-TEV transgenic flies, we modified pattB-exumP-EGFP [22] by replacing the EGFP coding region with that encoding TEV. The region coding for TEV with an N-terminal SV40 nuclear localization signal and a V5 epitope tag as well as two C-terminal SV40 nuclear localization signals was excised with EcoRI/NotI from pUASp1-TEV, a plasmid identical to that described earlier [63] except that it did not contain the S219V mutation. Thereafter, the EcoRI—NotI fragment was used to replace the EcoRI—NotI fragment of pattB-exumP-EGFP. The plasmid pattB-betaTub85DP-TEV contained cis-regulatory 5’ and 3’ sequences of the spermatocyte-specific betaTub85D gene, as also present in pattB-Nslmb-vhh4-GFP4 [44]. The region coding for TEV and the N- and C-terminal extension described above was enzymatically amplified with SCH035/NT022 and inserted between the 5’ and 3’ betaTub85D sequences after digestion with KpnI/NotI. All constructs for expression of recombinant proteins were cloned using the InteBac system [64] for either insect cells or bacterial expression. Drosophila lines Several lines with mutations or transgenes that we have used for our analyses have been described earlier: UASt-snm-EGFP and UASt-mnm-EGFP [44], UASt-uno-EGFP and unocc1 [22], UASt-snm-mCherry [23], g-His2Av-mRFP [65], bamP-GAL4-VP16 [66], Sgs3-GAL4 (Bloomington Drosophila Stock Center (BDSC) #6870), Df(2R)Exel7094 (BDSC #7859). Fly lines carrying the transgenes UASt-unochm-EGFP, UASt-unoTEV-EGFP, UASt-unoT128A-EGFP and UASt-unoT128D-EGFP were generated (BestGene Inc., Chino Hills, CA, USA) by integration of the corresponding pUASt-attB constructs described above into the landing site P{CaryP}attP2. For production of fly lines with the transgenes exumP-TEV, betaTub85DP-TEV, UASt-uno_N-EGFP, UASt-uno_M-EGFP and UASt-uno_C-EGFP with the attB plasmids described above, we used the integration site P{CaryP}attP40. Standard crossing and generation of recombinant chromosomes were used to produce the various strains used for experimental analyses. The genotypes of the flies analyzed are described in detail in the supporting information (S2 Table). Bioinformatic analyses of the predicted UNO amino acid sequence UNO orthologs were collected with NCBI blast within the NCBI non-redundant protein database or the UniProt reference proteomes applying significant E-value thresholds below 0.001 [67]. We used HHPRED with alignments of the N- and C-terminal conserved regions of UNO for remote homology detection in the PDB and PFAM database [68]. The search with the N-terminal conserved domain was not informative, but the C-terminal region of UNO (aa 289–364) hit with a probability of 83.2 to the middle region of human Rad21 (aa 320–394) from the structure 4PJW [35]. The C-terminal hit region of UNO was aligned with the conserved middle region of Rad21, Rad21L and Rec8 orthologs using mafft (-linsi, v7.427) [69] and visualized with Jalview [70]. Cell culture and transfection S2R+ cells were cultured in Schneider’s medium (Gibco, cat# 21720, Thermo Fisher Scientific, Waltham, MA), 10% fetal bovine serum (Gibco, cat# 10500–064) and 1% Penicillin-Streptomycin (Gibco, cat# 15140) at 25°C. Transfections were performed using FuGENE HD (Promega, cat# E2311) in 6-well plates, T25 or T75 flasks. In case of 6-well plates, 1.2x106 cells were plated into one well in 2 ml complete medium. In T25 flasks, 5.2×106 cells were plated in 4 ml complete medium, and in T75 flasks, 15.6×106 cells in 8 ml complete medium. One hour after plating transfection mix was added. In case of 6-well plates, 100 μl transfection mix containing 1 μg plasmid DNA and 4 μl FuGENE HD in Schneider’s medium were added. For T25 flasks, 200 μl of transfection mix was used containing 2 μg plasmid DNA and 8 μl FuGENE HD in Schneider’s medium, and for T75 flasks, 400 μl of transfection mix containing 4 μg plasmid DNA and 16 μl FuGENE HD in Schneider’s medium. Cells were incubated for 2 days before analysis in co-immunoprecipitation experiments. Immunoprecipitation and immunoblotting For the analysis of whole cell lysates by immunoblotting (S1 and S3 Figs), transfected cells in T25 flasks or 6-well plates were transferred on ice and carefully washed twice with cold phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 1.47 mM KH2PO4, 6.46 mM Na2HPO4, pH 7.4). Using a scraper, cells were harvested with a total of 150 μl resp. 60 μl of 3x Lämmli buffer (62.5 mM Tris-HCl, 10% glycerol, 5% β-mercaptoethanol, 3% SDS, 0.01% Bromophenol Blue) and the mixture was transferred to Eppendorf tubes. All samples were boiled for 8 minutes at 96°C, aliquoted, snap frozen in liquid nitrogen and stored at -80°C until analysis by immunoblotting. For immunoprecipitation, the transfected S2R+ cells were detached with a cell scraper and the resulting cell suspension was centrifuged at 580 x g for 5 minutes in a 15 ml Falcon tube. The cell pellet was washed with 1 ml cold phosphate-buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 1.47 mM KH2PO4, 6.46 mM Na2HPO4, pH 7.4) and transferred to a 1.5 ml Protein LoBind tube (Eppendorf, cat# 022431081), which were also used for all subsequent steps. Cells were sedimented at 600 x g for 5 minutes at 4°C. All subsequent steps were performed on ice or at 4°C with ice cold solutions. Cells were lysed in lysis buffer (LB): 20 mM Tris-HCl pH 7.5, 300 mM NaCl, 2 mM MgCl2, 0.1% Nonidet P-40 Substitute (Sigma Aldrich, cat# 74385), 5% glycerol, 0.5 mM EGTA, 1 mM DTT, 50 U/ml Benzonase Nuclease ultrapure (Sigma Aldrich, cat# E8263) and 1 tablet Roche protease inhibitor c0mplete per 10 ml lysis buffer (Mini EDTA-free, EASYpack, Roche, cat# 04693159001). The cells harvested from a T25 flask were lysed in 500 μl LB and in 1000 μl LB in case of T75 flasks by pipetting up and down twice, each time followed by a 15-minute incubation. Cell lysates were cleared by centrifugation for 15 minutes at 16100 x g. A small aliquot of the supernatant was removed for analysis by immunoblotting (input samples). The rest of the supernatant was added to pre-washed 25 μl nano-trap agarose beads (GFP-Trap agarose, ChromoTek, cat# gta-20, or RFP-Trap agarose, ChromoTek, cat# rta-20, or MYC-Trap agarose beads ChromoTek, cat# yta-20). Beads were incubated for 1 hour on a rotating wheel at 15 rpm. Thereafter, beads were washed 3 times with LB and centrifugation at 2500 x g for 2 minutes. For the final wash samples were transferred into a fresh tube. For elution, beads were resuspended in 80 μl 3x Lämmli Buffer (62.5 mM Tris-HCl, 10% glycerol, 5% β-mercaptoethanol, 3% SDS, 0.01% Bromophenol Blue) and boiled for 8 minutes at 96°C. After rapid cooling on ice, beads were sedimented and the supernatant was distributed in three aliquots of 25 μl (IP samples). All samples were snap frozen in liquid nitrogen and stored at -80°C until analysis by immunoblotting. Samples were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using a mini-gel system (BioRAD) and gels with between 7% and 12% polyacrylamide for 3 hours at 90 V. Molecular weight markers were PageRuler Plus Prestained Protein Ladder (Thermo Scientific, cat# 815-968-0747). Protein transfer to nitrocellulose membranes (Amersham Protran 0.45 μm, cat# 10600002) was achieved by tank blotting at room temperature for 1 hour with 100 V. After transfer, Ponceau S staining was performed, followed by a blocking step using 5% dry milk (w/v) in PBS with 0.02% NaN3. This solution was also used for incubation with the primary antibody for 2 hours at room temperature or overnight at 4°C. Three washes with 5% dry milk (w/v) in PBS were done before incubation with the secondary antibody, which was applied in 5% dry milk (w/v) in PBS at room temperature for 1 hour protected from light. After 2 washes with 5% dry milk (w/v) in PBS and 2 washes with PBS, 0.1% Tween-20, signals were detected with ECL reagents (WesternBright ECL, Advansta, cat# K-12045-D50) in an Amersham Imager 600. The following antibodies were used for immunoblotting: rabbit polyclonal antibodies anti-GFP diluted 1:800 (ChromoTek, cat# pabg1) or 1:2000 (Torrey Pines Biolabs, cat# TP401); mouse monoclonal antibody anti-RFP (ChromoTek, cat# 6g6), 1:1000; rat monoclonal antibody anti-c-MYC (ChromoTek, cat# 9e1) at 1:1000; rabbit polyclonal antibody anti-ModC [13] (kindly provided by Rainer Dorn, Universität Halle, Halle, Germany) at 1:4000; rabbit polyclonal antibody anti-SNM C-terminal peptide (DIAHLKEYRNALRPRKTKSYPQAT) [20] (kindly provided by Bruce McKee, University of Tennessee, Knoxville, TE, USA); rabbit polyclonal antibody anti-UNO [22]; HRP-conjugated AffiniPure goat anti-rabbit IgG polyclonal antibody (Jackson ImmunoResearch, cat# 111-035-003) at 1:1000; HRP-conjugated AffiniPure goat anti-mouse IgG polyclonal antibody (Jackson ImmunoResearch, cat# 115-035-003) at 1:1000; HRP-conjugated goat anti-rat IgG antibody (Thermo Scientific, cat# 62–9520) at 1:5000. We note that after transient expression of TetR-E and immunoblotting with anti-EGFP, we always detected two bands for unknown reasons. Protein expression and purification For expression of SNM and UNO_C, baculovirus was generated as previously described [71]. Briefly, EMBacY cells were transformed with constructs for either UNO_C-Strep or MBP-SNM. Bacmids were extracted from positive transformants and used to transfect Sf9 cells. Three rounds of viral amplification were performed to produce a “V2” virus suspension. For each SNM/UNO_C prep, 2 x 400 ml flasks of Hi5 cells were infected with V2 at a 1:100 dilution (i.e. 4 ml of SNM V2 virus and 4 ml of UNO_C V2 virus), and incubated at 27°C for 72 hours. Cells were harvested, washed with PBS, and flash frozen in liquid nitrogen. For protein purification, cells were resuspended in lysis buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 0.1% Triton X-100, 1 mM MgCl2 + 5 mM BME, DNAse, SERVA protease inhibitor) and lysed by sonication. Cleared lysate was loaded onto a 5 ml Strep Tactin Superflow Plus column (Qiagen), washed with lysis buffer, and eluted with elution buffer (50 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 1 mM MgCl2 + 5 mM BME, 2.5 mM Desthiobiotin). Fractions corresponding to the SNM-UNO_C complex (SU_C) were concentrated with a 10 kDa MWCO concentrator (Pierce) and loaded onto a Superose 6 10/300 column equilibrated with storage buffer (25 mM HEPES pH 7.5, 300 mM NaCl, 5% glycerol, 1 mM MgCl2 + 5 mM TCEP). Eluted fractions were concentrated and snap frozen. N-terminally 6xHis-MBP tagged UNO_N was expressed in BL21 cells at 18°C for 16 hours grown in TB. For each purification, cells from a 1L culture were used. Harvested cells were resuspended in lysis buffer (50mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 0.1% Triton X-100, 1 mM MgCl2, 5 mM β-mercaptoethanol, DNAse, SERVA protease) using 5 ml/g pellet. Resuspended bacteria were lysed by two passages through an Emulsiflex (Avestin). Cleared lysate was passed over a 5 ml MBP Trap column (Cytiva) equilibrated in MBP-Trap A buffer (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 5 mM β-mercaptoethanol) at 2.5 ml /minute. MBP-Trap was washed with 15 CV of MBP-Trap A buffer followed by elution with 10 CV of MBP-Trap buffer B (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 1 mM Maltose 5 mM β-mercaptoethanol) collecting 5 ml fractions. Fractions corresponding to MBP-UNO_N were pooled and the NaCl concentration adjusted to 100 mM. Sample was then loaded onto a 6 ml ResourceQ column (Cytiva) pre-equilibrated with ResQ Buffer A (50 mM Na-HEPES pH7.5, 100 mM NaCl, 5% glycerol, 5 mM β-mercaptoethanol). Unbound sample was washed away with 10 CV of ResQ Buffer A. MBP-UNO_N was eluted by running a gradient of 0–60% ResQ Buffer B (50 mM Na-HEPES pH7.5, 1 M NaCl, 5% glycerol, 5 mM β-mercaptoethanol) over 30 CVs. Fractions corresponding to MBP-UNO_N were concentrated in a Amicon 30 kDa MWCO concentrator to a volume of ~1.5 ml. Concentrated protein was loaded onto a Superdex 200 16/600 column, pre-equilibtated with SEC Buffer (50mM Na-HEPES pH7.5, 300mM NaCl, 5% Glycerol, 1mM TCEP) and run at 1 ml/minute. Fractions corresponding to the two resulting UNO_N peaks were pooled, concentrated, flash frozen, and stored at -80°C. N-terminally 6xHis-MBP tagged MNM_N was expressed in BL21 cells at 18°C for 16 hours grown in TB. For each purification, cells from a 1L culture were used. Harvested cells were resuspended in lysis buffer (50mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 0.1% Triton X-100, 1 mM MgCl2, 5 mM β-mercaptoethanol, DNAse, SERVA protease) using 5 ml/g pellet. Resuspended bacteria were lysed by two passages through an Emulsiflex (Avestin). Cleared lysate was passed over a 5 ml MBP Trap column (Cytiva) equilibrated in MBP-Trap A buffer (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 5 mM β-mercaptoethanol) at 2.5 ml /minute. MBP-Trap was washed with 15 CV of MBP-Trap A buffer followed by and additional wash step with buffer containing 1 mM ATP, followed by elution with 10 CV of MBP-Trap buffer B (50 mM Na-HEPES, pH7.5, 300 mM NaCl, 5% Glycerol, 1 mM Maltose 5 mM β-mercaptoethanol) collecting 5 ml fractions. Fractions corresponding to MBP-MNM_N were pooled and the NaCl concentration adjusted to 100 mM. Sample was then loaded onto a 6 ml ResourceQ column (Cytiva) pre-equilibrated with ResQ Buffer A (50 mM Na-HEPES pH7.5, 100 mM NaCl, 5% glycerol, 5 mM β-mercaptoethanol). Unbound sample was washed away with 10 CV of ResQ Buffer A. MBP-MNM_N was eluted by running a gradient of 0–100% ResQ Buffer B (50 mM Na-HEPES pH7.5, 1 M NaCl, 5% glycerol, 5 mM β-mercaptoethanol) over 10 CVs. Fractions corresponding to MBP-MNM_N were concentrated in a Amicon 30 kDa MWCO concentrator to a volume of ~1.5 ml. The MBP moiety was removed by the addition of GST-3C protease, followed by incubation on ice for 3 hours. Concentrated protein was loaded onto a Superdex 200 16/600 column, pre-equilibtated with SEC Buffer (50mM Na-HEPES pH7.5, 300mM NaCl, 5% Glycerol, 1mM TCEP) and run at 1 ml/minute. To ensure removal of the MBP and GST-3C protease from the sample a 5 ml MBP-Trap and a 5 ml GST-Trap column were connected in-line and downstream of the Superdex 200 16/600 column Fractions corresponding to the two resulting MNM-N peaks were pooled, concentrated in an Amicon 10 kDa MWCO concentrator, flash frozen, and stored at -80°C. Size exclusion chromatography SEC experiments were run on an Akta Pure 25 system (Cytiva). Absorbances were measured at both 280 nm (blue traces) or 254 nm (red traces). Molecular weight markers (Bio-Rad) were used as a reference (grey dotted traces). Samples from SEC experiments were collected and run on SDS-PAGE gels stained with Coomassie brilliant blue. SEC-MALS 50 μL samples at 5–10 μM concentration were loaded onto a Superose 6 5/200 (run at 0.3 ml/min) or Superdex 75/150 (run at 0.5 ml/min) analytical size exclusion column (Cytiva) equilibrated in buffer containing 50 mM HEPES pH 7.5, 1 mM TCEP, 300 mM NaCl (for samples without nucleosomes) or 150 mM NaCl (for samples with nucleosomes) attached to an 1260 Infinity II LC System (Agilent). MALS was carried out using a Wyatt DAWN detector attached in line with the size exclusion column. Cross-linking mass spectrometry (XL-MS) For XL-MS analysis proteins were diluted in 200 μL of XL-MS buffer (30 mM HEPES 6.8, 150 mM NaCl, 5% glycerol, 1 mM MgCl2, 1 mM TCEP) to the final concentration of 3 μM, mixed with 3 μL of DSBU (200 mM) and incubated for 1 hour at 25°C. The reaction was stopped by adding 20 μL of 1 M Tris-HCl pH 8.0 and incubated for another 30 min at 25°C. The cross-linked sample was precipitated by addition of 4X volumes of 100% cold acetone followed by overnight incubation at -20°C. Samples were analyzed as previously described [72]. For interaction network visualization XVis software was used and for visualization of the crosslinks on the PDB model PyXlinkViewer [73] and XMAS [74] was used. Each time a different cutoff for the cross-linking credibility was selected depending on the quality of the cross-linking data. Alphafold2 predictions Predicted structures were calculated using AlphaFold Multimer (2.2.0) [40] run on GPU nodes of the Raven HPC of the Max Planck Computing and Data Facility (MPCDF), Garching. Each job was run on a single node consisting of 4 x Nvidia A100 NVlink 40 GB GPUs. Multiple predictions were generated for each run, and the best model (determined by pTM score) was then used. PAE plots were generated using a custom script (Vikram Alva, MPI Biology Tübingen). Mass photometry Mass Photometry was performed in the mass photometry buffer (MP) containing 30 mM HEPES pH 7.8, 150 mM NaCl, 5% glycerol, 1 mM MgCl2, and 1 mM TCEP. Protein samples (3 μM) were pre-equilibrated for 1 hour in the MP buffer. Measurements were performed using Refeyn One (Refyn Ltd., Oxford, UK) mass photometer. Directly before the measurement, the sample was diluted 1:100 with the MP buffer. Molecular mass was determined in Analysis software provided by the manufacturer using a NativeMark (Invitrogen) based standard curve created under the identical buffer composition. Negative-stain electron microscopy 4 μl of MNM at 27 μg/ml were adsorbed at 25°C for 2 minutes onto glow-discharged carbon-coated grids. The grids were washed three times with water and negatively stained with three washes of 1% uranyl acetate, followed by a 5-minute incubation at 25°C. Samples were imaged with a Tecnai G2 Spirit BioTWIN microscope equipped with a LaB6 cathode operated at 120 kV. Images were recorded at low-dose conditions (19 electrons/Å2) at a corrected magnification of 82553x on a 4k × 4k CMOS camera F416 (TVIPS, Oslo, Norway). EMSAs The binding reactions (10 μL volume) were carried out in EMSA buffer (25 mM HEPES pH 7.5, 0.1 μg/μL BSA, 60 mM NaCl) containing indicated fluorescently labelled DNA substrate (10 nM). The reactions were started by addition of increasing amounts of SU_C protein complexes (36.25, 72.5, 108.75, 145, 217.5, 290, 435, 580, 870, 1160 and 1740 nM) and incubated for 20 min at 30°C. After the addition of 2 μL of the gel loading buffer (60% glycerol, 10 mM Tris–HCl, pH 7.4, 60 mM EDTA, 0.15% Orange G), the reaction mixtures were resolved in 0.8% agarose gel in 1x TAE buffer (40 mM Tris, 20 mM acetic acid, 1 mM EDTA). The gels were scanned using Amersham Typhoon scanner (Cytiva) and quantified in ImageJ. Microscopic analyses with larval salivary glands By standard crossing, we combined the Sgs3-GAL4 driver with UASt transgenes. Wandering third instar larvae were used for dissection of salivary glands after development at 25°C. Salivary gland preparations were made as described [75,76] with the modifications reported previously [23]. Imaging and signal quantification was also performed as described [23]. Microscopic analyses with testis preparations For whole-mount testis preparations, dissections from young adult males (0–1 day after eclosion) were performed in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl, pH 6.8). Testes were fixed in PBST containing 4% formaldehyde in 0.2 ml Eppendorf tubes for 20 minutes on a rotating wheel. Testis squash preparations were made essentially as described previously [77]. For DNA staining, testes were incubated for 10 minutes in PBS, 0.1% Triton X-100 (PBTx) containing Hoechst 33258 (1 μg/ml). After three washes with PBS, preparations were mounted under a coverslip in a drop of mounting medium. Microscopic quantification of the DNA content of nuclei of early round spermatids was done as described [44]. Preparations of ovaries and testes were analyzed with a wide-field fluorescence microscope (Zeiss Axio Observer HS) using 40×/1.3, 63×/1.4 and 100×/1.4 oil immersion objectives. Maximum intensity projections of image stacks are presented. Time-lapse imaging of progression through meiosis was performed as described [78]. Testes were dissected from pupal or young adult males in Schneider’s Drosophila Medium (Invitrogen, #21720) supplemented with 10% fetal bovine serum (Invitrogen) and 1% penicillin/streptomycin (Invitrogen, #15140). The dissected pupal testes were transferred into 45 μl of medium in a 35 mm glass bottom dish (MatTek Corporation, #P35G-1.5-14-C) and opened with fine tungsten needles to release the cysts. In case of adult testes, 150 μl of medium were used. To reduce sample movements, 15 μl of 1% w/v methylcellulose (Sigma, #M0387) was added to pupal testes preparations and 50 μl to adult testes preparations. A wet filter paper was placed inside along the dish wall before sealing the lid with parafilm. Imaging was performed at 25°C in a room with temperature control using a spinning disk confocal microscope (VisiScope with a Yokogawa CSU-X1 unit combined with an Olympus IX83 inverted stand and a Photometrics evolve EM 512 EMCCD camera, equipped for red/green dual channel fluorescence observation; Visitron systems, Puchheim, Germany). A 60×/1.42 oil immersion objective was used. We acquired z-stacks with 30–40 focal planes spaced by 500 nm at 45-second intervals. Maximum intensity projections were generated using ImageJ or ZEN software for wide-field images and IMARIS (Bitplane) for spinning disk confocal images. Figures display maximum intensity projections unless stated otherwise. Export of projections from IMARIS as movies or still frames after live imaging was made with interpolated image display. Moreover, display parameters for the His2Av-mRFP were adjusted manually over time to reveal chromosomes clearly throughout the movies, thereby correcting photobleaching and partially also the changes in the extent of chromosome condensation during M I. Graphs were generated with Microsoft Excel or GraphPad Prism. P values were calculated using a two-tailed student t-test (* = p < 0.05; ** = p < 0.01; *** = p < 0.001). Adobe Photoshop and Adobe Illustrator were used for production of figures. FRAP analyses Testes were isolated from early pupae and cysts were released for imaging in dishes with a glass coverslip at the bottom as described [78]. Before bleaching, we acquired five z-stacks with 40 optical sections spaced by 500 nm at one-minute intervals using a FV1000 Olympus laser scanning confocal microscope with a PLAPON 60XO/1.42 Objective with a zoom factor of 5.4. Thereafter, we photobleached a part of the nucleolus in one of the spermatocytes by using an circular region of interest (diameter = 30 pixels) and 100 iterations of tornado scanning with maximal 488 nm laser intensity within one z section. After photobleaching the EGFP signals in part of a nucleolus, imaging of z-stacks was continued, initially as before photobleaching. However, after five z-stacks, the time interval between z-stack acquisition was increased from one to 15 minutes. In case of long-term FRAP analyses with bam>mnm-EGFP spermatocytes, we acquired only three z-stacks at one-minute intervals immediately after photobleaching, followed by acquisition of three additional z-stacks at one-minute intervals four hours later. For the quantitative analysis of EGFP signal recovery after photobleaching over time (S6 Fig), we used spot detection by IMARIS software to identify spheres containing the bleached nucleolus or unbleached nucleoli in neighboring spermatocytes of the imaged cyst. After creating a first set of spheres with a diameter of five μm, a second set of spheres with a diameter of seven μm was generated for background correction of signal intensities in the smaller spheres. For further processing with Microsoft Excel, the signal intensities detected in the spheres were exported from IMARIS. The difference in signal intensities observed in the large and small sphere, respectively, was used for estimation of background signal intensity, which was subtracted from the total intensity value within the small sphere. Moreover, fluorescence intensities were normalized to the average detected during the five z-stacks acquired before the bleaching of a nucleolus. For the quantification of the overall recovery of EGFP signals 90 minutes after photobleaching (Fig 6A and 6B), we generated intensity sum projections using Image J. Representative projection images are shown in Fig 6A. A circular ROI with a diameter of 30 pixels, as used before for photobleaching, was placed over the bleached region. Moreover, a bean-shaped ROI covering the unbleached part of the targeted nucleolus was selected manually. Average pixel intensities in these ROIs were quantified for the five pre-bleaching time points, as well as for the first and last post-bleaching time points. Pixel intensities were normalized to the average of the five pre-bleaching time points. The intensity difference between the first and last post-bleaching time points observed in the non-bleached and bleached parts of the nucleolus were compared to estimate the extent of FRAP corrected for photobleaching during image acquisition after pulse-bleaching of a part of a nucleolus. To express the extent of FRAP after 90 minutes in percent of the total signal intensity loss induced by the pulse-bleaching within the bleached part of the nucleolus (Fig 6B), this total signal intensity loss was calculated as the difference between the intensities at the first post-bleaching time point in the non-bleached and bleached region, respectively, of the targeted nucleolus. To compare the intensity of EGFP signals that were either diffusely distributed throughout the nucleolus or within sub-nucleolar foci, we analyzed intensity sum projections of the pre-bleaching z-stacks with Image J. The projection images were segmented by using the threshold tool of Image J and selecting first the top 2% intensity pixels, which represented the sub-nucleolar foci well. Thus, the resulting intensity values were used as a measure of signal intensity in sub-nucleolar foci. Thereafter, a second segmentation was applied for selection of the top 10% intensity pixels, which covered the complete nucleoli. To estimate the diffuse nucleolar signals, we subtracted the intensities detected within the sub-nucleolar foci from those of the complete nucleoli. Supporting information S1 Fig. Characterization of endogenous expression of SNM, UNO and MNM in S2R+ cells by immunoblotting. (A-C) Total extracts of S2R+ cells, either untransfected or transfected for transient expression of EGFP-SNM (E-SNM), SNM-EGFP (SNM-E), UNO-EGFP (UNO-E), EGFP-MNM or MNM-EGFP (MNM-E) were analyzed by immunoblotting with the indicated antibodies. Prestained marker proteins with the indicated molecular weights (kDa) are displayed on the left. Relative amounts of extract loaded are indicated on top. Ponceau S staining of the membrane after protein transfer for control of loading is presented below the immunoblots. (A) A band indicating expression of endogenous SNM is not detected by anti-SNM, in contrast to transiently expressed SNM-EGFP, confirming antibody functionality. (B) A band indicating expression of endogenous UNO is not detected by anti-UNO, in contrast to transiently expressed UNO-EGFP, confirming antibody functionality. (C) An antibody (anti-M_CP) against the common part present in MNM and the other isoforms expressed from the mod(mdg4) locus detects bands in untransfected S2R+ cells that are much weaker than those reflecting transiently expressed MNM-EGFP. The anti-M_CP immunoblot is shown after short (top) and long exposure (bottom) to reveal these weak bands. https://doi.org/10.1371/journal.pgen.1010547.s001 (PDF) S2 Fig. Sequence similarity of UNO and α-kleisins. An alignment of the predicted amino acid sequences of the indicated proteins reveals similarities between a region close to the C-terminus of UNO and an internal α-kleisin region, which is known to bind to stromalin/SA/STAG proteins. The family of α-kleisins includes members that function in meiosis (Rec8 proteins) beyond those functioning also during mitosis (Rad21 and Rad21L proteins). While the genome of D. melanogaster does not contain a canonical rec8 ortholog, it includes c(2)M, a more distant meiotic α-kleisin [79]. Accession numbers from the UniProt database (Dmel_uno, Dere_uno, Dpse_uno, Mdom_uno) or the NCBI database (all others) are provided next to the gene names with species abbreviated: Drosophila melanogaster (Dmel), Drosophila erecta (Dere), Drosophila pseudoobscura (Dpse), Musca domestica (Mdom), Ceratitis capitata (Ccap), Bactrocera dorsalis (Bdor), Homo sapiens (Hsap), Mus musculus (Mmus). Residues (triangles) of the human α-kleisin Rad21 that directly interact with human SA1 [36], as well as the α-helices (magenta rods) in Rad21 are indicated below the alignment. https://doi.org/10.1371/journal.pgen.1010547.s002 (PDF) S3 Fig. Comparison of protein levels after individual or combined expression of SNM and UNO. Co-expression of SNM with UNO or with the UNO_C fragment results in higher expression levels in comparison to the individual expression of these proteins. Levels of tagged proteins resulting after transient expression of UNO (A) or UNO fragments (B) together with the indicated proteins in S2R+ cells were analyzed by immunoblotting. SNM was tagged with mCherry (SNM-C). UNO (A) and the fragments UNO_N, UNO_M and UNO_C (B) were tagged with EGFP (UNO-E, UNO_N-E, UNO_M-E and UNO_C-E). A plasmid for expression of a C-terminal fragment of Mad1 tagged with a myc-epitope (myc-Mad1_C) was co-transfected for comparison of transfection efficiencies. Total cell extracts were resolved and analyzed with anti-mCherry (anti-C), anti-EGFP (anti-E), and anti-myc (anti-myc). Relative amounts of extract loaded are indicated on top. Ponceau S staining of the membrane after protein transfer for control of loading is presented below the immunoblots. The positions of prestained marker proteins with the indicated molecular weights (kDa) and of the bands representing the indicated proteins of interest are displayed on the right. https://doi.org/10.1371/journal.pgen.1010547.s003 (PDF) S4 Fig. Purification and structural analysis of UNO_N. (A) Final purification step of UNO_N. The SEC profile revealed two peaks (Peak 1 and Peak 2), both containing only UNO_N according to SDS-PAGE and Coomassie staining. (B) SEC-MALS of the UNO_N complexes in Peak 1 and Peak 2 resulted in the indicated molecular masses. (C) Topology map of a UNO_N monomer. (D) AF2 model of the UNO_N dimer with predicted alignment error shown below. (E) AF2 model of the UNO_N tetramer with predicted alignment error shown below. https://doi.org/10.1371/journal.pgen.1010547.s004 (PDF) S5 Fig. Purification and structural analysis of N-terminal part of MNM. (A) Final purification step of Mod(mdg4)_CP, the N-terminal part of MNM, which is also present in other Mod(mdg4) protein isoforms. Analysis by SDS-PAGE and Coomassie staining indicates that the peak obtained by SEC contains pure Mod(mdg4)_CP in a multimerized form. (B) Molecular mass determination of purified Mod(mdg4)_CP by SEC-MALS resulted in a value of 80.92 kDa, consistent with a hexamer formation. (C) AF2 model of Mod(mdg4)_CP hexamer with predicted alignment error shown below. (D) Ribbon diagram of one of the three Mod(mdg4)_CP dimers that form the hexamer according to the AF2 model. (E) Surface electrostatics on the Mod(mdg4)_CP hexamer showing a lack of obvious DNA-binding regions. (F) NS-EM image of the Mod(mdg4)_CP hexameric rings. https://doi.org/10.1371/journal.pgen.1010547.s005 (PDF) S6 Fig. FRAP analysis with spermatocytes expressing SNM-, UNO- or MNM-EGFP. FRAP analyses were completed with S5 cysts released from pupal testes of males expressing the indicated UASt transgenes driven by bam-GAL4-VP16. A subregion of the nucleolus was bleached in one of the spermatocytes of the cyst, while neighboring spermatocytes were used as controls, as illustrated with the still frames acquired during an experiment with SNM-EGFP just before and after the bleaching. The timeline on top illustrates the image acquisition sequence. EGFP signal intensities in the nucleoli were quantified and are plotted after normalization to the average of the intensities observed before bleaching. https://doi.org/10.1371/journal.pgen.1010547.s006 (PDF) S7 Fig. Localization and effects of UNOTEV-EGFP in different genotypes. (A) Phenotypic characterization with squash preparations of testes from uno null mutants with bam>unoTEV-EGFP and either no TEV transgene or betaTub85DP-TEV (bTub85DP-TEV) provided further confirmation of the findings revealed by time-lapse imaging. In the absence of a TEV transgene UNOTEV-EGFP was readily detectable at the S6 stage with a sub-cellular localization identical to that of wild-type UNO-EGFP [22]. The S6 spermatocytes also displayed a normal number of major chromosome territories. However, in the presence of bTub85DP-TEV, UNOTEV-EGFP was not detectable in S6 spermatocytes, which displayed an increased number of chromosome territories. Each micrograph is shown twice, on the left with enhanced EGFP signal intensities, which reveal weak autosomal UNOTEV-EGFP dots in the absence of a TEV transgene and still no signals in the presence of bTub85DP-TEV. Still frames from cysts during telophase I analyzed by time-lapse imaging (bottom), illustrate the presence of massive chromosome bridges with persisting non-degradable UNOTEV-EGFP, but only in the absence of a TEV transgene. (B) Still frames after time-lapse imaging of His2A-mRFP expressing uno null mutants with bam>unoTEV-EGFP and exumP-TEV during M I reveal premature bivalent separation and absence of chromosome bridges during telophase I. Time (min:sec) with t = 0 at the onset of NEBD I is indicated. (C) Comparison of UNOTEV-EGFP signal intensities at the onset of NEBD I in indicated genotypes after time-lapse imaging. While UNOTEV-EGFP is completely eliminated by bTub85DP-TEV in unocc1 homozygous null mutant spermatocytes (uno-/-), residual UNOTEV-EGFP is detectable in unocc1 heterozygous spermatocytes (uno-/+) despite the presence of bTub85DP-TEV. (D) Disappearance of residual UNOTEV-EGFP (arrows) during anaphase I and absence of chromosome bridges during telophase I in uno-/+ spermatocytes with bam>unoTEV-EGFP and bTub85DP-TEV, as revealed by time-lapse imaging of spermatocytes expressing His2Av-mRFP. Time (min:sec) with t = 0 at the onset of NEBD I is indicated. Scale bars = 10 (A) and 3 (B-D) μm. https://doi.org/10.1371/journal.pgen.1010547.s007 (PDF) S8 Fig. Expression and localization of the UNO fragments UNO_N-, UNO_M- and UNO_C-EGFP in spermatocytes. (A) The driver bamP-GAL4-VP16 was used for expression of the indicated UASt transgenes in an uno+ background. Testis squash preparations for comparison of expression level and pattern were labeled with a DNA stain. Apical testes regions are displayed. The images presenting merged DNA and EGFP channels (top row) are shown with identical settings during acquisition and display, indicating that expression levels were maximal in case of UNO_M-EGFP, weaker for UNO_N-EGFP and considerably weaker for UNO_C-EGFP. In case of the images displaying only the EGFP signals in grey values (bottom row), display settings were enhanced in case of UNO_C-EGFP to reveal its weak expression. (B) After peak expression of UNO_N-, UNO_M- and UNO_C-EGFP in early spermatocytes (see panel A), maintenance at low levels in late spermatocytes was detectable in case of UNO_N-EGFP and UNO_C-EGFP, as documented by high resolution of images from testes squash preparations. While at the S5 stage, UNO_N-EGFP and UNO_C-EGFP were in sub-nucleolar foci in spermatocytes heterozygous for the unocc1 null allele (uno-/+) (comparable to full-length UNO-EGFP, see Fig 5F), they displayed an abnormal diffuse nucleolar localization in unocc1 homozygous spermatocytes (uno-/-). At the S6 stage, UNO_N-EGFP and UNO_C-EGFP were in weak dots on the sex chromosome bivalent in uno-/+ spermatocytes and not detectable in uno-/- spermatocytes. Scale bars = 20 (A) and 5 (B) μm. https://doi.org/10.1371/journal.pgen.1010547.s008 (PDF) S9 Fig. UNOT128A-EGFP and UNOT128D-EGFP promote homolog conjunction that cannot be eliminated during anaphase I. (A,B) Time-lapse imaging of progression through M I with His2Av-mRFP expressing spermatocytes was performed for the characterization of phenotypic consequences of the T128A (A) and T128 (B) mutations that alter a conserved potential phosphorylation site immediately upstream of the separase cleavage site in UNO (see Fig 7A). The mutants were expressed in uno null mutant spermatocytes with UASt transgenes and the driver bamP-GAL4-VP16. Time (min:sec) with t = 0 at the onset of NEBD I is indicated. Scale bars = 2 μm. https://doi.org/10.1371/journal.pgen.1010547.s009 (PDF) S1 Movie. Progression through M I in uno null mutant spermatocyte with bam>unochm-EGFP and His2Av-mRFP. Time-lapse analysis was used for analysis of progression through M I. Maximum intensity projections of z-stacks acquired at 45 sec intervals of the spermatocyte, which is also displayed in Fig 5H, are shown. https://doi.org/10.1371/journal.pgen.1010547.s010 (MP4) S2 Movie. Progression through M I in uno null mutant spermatocyte with bam>unoTEV-EGFP and His2Av-mRFP. Time-lapse analysis was used for analysis of progression through M I. Maximum intensity projections of z-stacks acquired at 45 sec intervals of the spermatocyte, which is also displayed in Fig 7C, are shown. The images sequence is presented three times. During the first period, display settings do not saturate the strong UNOTEV-EGFP dot on the chrXY bivalent. During the second period, green signal intensities are enhanced to reveal the weak UNOTEV-EGFP dots on autosomal bivalents. During the final repetition, only the enhanced green signals are displayed as grey values. https://doi.org/10.1371/journal.pgen.1010547.s011 (MP4) S3 Movie. Progression through M I in uno null mutant spermatocyte with bam>unoTEV-EGFP, betaTub85DP-TEV and His2Av-mRFP. Time-lapse analysis was used for analysis of progression through M I. Maximum intensity projections of z-stacks acquired at 45 sec intervals of the spermatocyte, which is also displayed in Fig 7D, are shown. For comparison with S2 Movie, the images sequence is also presented three times with distinct display settings, as described for S2 Movie, even though UNO-TEV-EGFP is not detectable during M I in this genotype. https://doi.org/10.1371/journal.pgen.1010547.s012 (MP4) S4 Movie. Progression through M I in uno-/+ mutant spermatocyte with bam>unoTEV-EGFP, betaTub85DP-TEV and His2Av-mRFP. Time-lapse analysis was used for analysis of progression through M I. Maximum intensity projections of z-stacks acquired at 45 sec intervals of the spermatocyte, which is also displayed in S7D Fig, are shown. For comparison with S2 Movie, the images sequence is also presented three times with distinct display settings, as described for S2 Movie. A weak UNOTEV-EGFP dot on the chrXY bivalent is barely detectable with the display settings during the first period, but it is readily apparent with the enhanced settings during the second and third period. UNOTEV-EGFP dot on autosomal bivalents cannot be detected even with the enhanced display settings. https://doi.org/10.1371/journal.pgen.1010547.s013 (MP4) S1 Table. Synthetic DNA fragments. https://doi.org/10.1371/journal.pgen.1010547.s014 (XLSX) S2 Table. Description of the analyzed genotypes. https://doi.org/10.1371/journal.pgen.1010547.s015 (XLSX) S3 Table. Source data. https://doi.org/10.1371/journal.pgen.1010547.s016 (XLSX) Acknowledgments We thank Rainer Dorn and Bruce McKee for providing antibodies, Katharina Hipp (MPI for Biology, Tübingen, Germany) for help with NS-EM experiments, Franziska Müller and Petra Janning (MPI of Molecular Physiology, Dortmund, Germany) for the XL-MS data sets, Veronika Altmannova (Weir Lab) for help with EMSA experiments, as well as Sina Moser and Hiro Yamada for technical help.

Kalbfuß, Nils;Strohmayr, Alexander;Kegel, Marcel;Le, Lien;Grosse-Holz, Friederike;Brunschweiger, Barbara;Stöckl, Katharina;Wiese, Christian;Franke, Carina;Schiestl, Caroline;Prem, Sophia;Sha, Shuyao;Franz-Oberdorf, Katrin;Hafermann, Juliane;Thiemé, Marc;Facher, Eva;Palubicki, Wojciech;Bolle, Cordelia;Assaad, Farhah F.

Hughes, Hannah Q.;Christman, Nicholas D.;Dalia, Triana N.;Ellison, Courtney K.;Dalia, Ankur B.

doi: 10.1371/journal.pgen.1010561pmid: 36542674

Introduction Type IVa pili (T4aP) are used by a wide range of bacteria to interact with their environment [1]. These fibers extend and retract from cell surfaces to accomplish a wide range of behaviors, including twitching motility [2,3], the uptake of DNA during natural transformation [4,5], and initial attachment of cells to a surface [6–8]. The frequency of pilus dynamic activity and the number of surface pili maintained at any given time directly impacts the ability of the cell to perform pilus-related functions. For example, the Neisseria meningitidis pilus system has been characterized to perform different functions based on the number of surface pili, ranging from DNA uptake, where a single pilus is sufficient, to interaction with host cells, which optimally requires five pili [9]. Elucidating the regulation of pilus dynamic activity and pilus number is therefore crucial to understanding pilus function. T4aP are primarily composed of repeating subunits of the major pilin, which are extruded from the inner membrane and assembled into a filament through the activity of a dedicated extension ATPase. The process is reversed through the activity of a retraction ATPase, commonly called PilT, which deposits major pilins back into the inner membrane [10]. In these dynamic pilus systems, the activity of the extension and retraction motors plays an important role in regulating surface piliation. Current models propose that these motors compete to interact with the T4aP machine, such that the motor that interacts with the platform dictates the direction of dynamic activity [11,12]. If the extension motor is inhibited, pilus retraction dominates and pilus number decreases [13–15]. Conversely, if the retraction motor is disrupted, extension occurs undeterred, resulting in hyperpiliation [5,16–19]. However, these trends are not universally observed, indicating that alternate mechanisms of pilus regulation exist and contribute to surface piliation. Deletion of pilT does not result in hyperpiliation of the T4P in Francisella tularensis [20], Clostridium perfringens [21], or V. cholerae mannose-sensitive hemagglutinin (MSHA) pili [22–25]. Indeed, for each of these systems, deletion of pilT results in a decrease in piliation. This observation is counterintuitive based on our current understanding of T4aP and suggests that T4aP dynamic activity may be regulated by more than a simple competition between extension and retraction motors. In this study, we use the MSHA T4aP of V. cholerae as a model system to investigate PilT-dependent surface piliation because of the extensive genetic and cell biological tools that have recently been developed to study the dynamic activity of these pili [7,22,23,25,26]. Results Vibrio cholerae expresses two distinct T4aP systems: MSHA pili and competence pili (also known as ChiRP pili [27]). MSHA pili facilitate biofilm formation by promoting the initial attachment of cells to a surface [8,27–29], while competence pili promote DNA uptake for horizontal gene transfer by natural transformation [4,5]. The dynamic activity of both T4aP systems can be studied in live cells via the use of a major pilin cysteine knock-in mutation (mshAT70C for MSHA pili; pilAS67C for competence pili) and subsequent labeling with thiol-reactive fluorescent maleimide dyes [4,7,26]. Competence pili are highly dynamic and pilus retraction generally occurs immediately following extension. In contrast, MSHA pili are maintained in an extended state and retract in response to distinct stimuli [23]. One stimulus that induces MSHA pilus retraction is the imaging condition used to visualize fluorescently labeled pili. Specifically, repeated exposures during timelapse imaging (3 sec intervals) results in the uniform retraction of MSHA pili across a field of view within 3 minutes–a phenomenon that we term ‘imaging-induced retraction’. Although the molecular mechanism underlying this response remains unclear, it nevertheless provides a reliable approach by which we can assess MSHA pilus retraction. While these T4aP have distinct extension motor ATPases (MshE for MSHA pili; PilB for competence pili), they both rely on the same ATPase motor, PilT, for efficient retraction (Fig 1A and 1B), which is consistent with a number of previous studies [4,22,23]. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 1. PilT has a distinct impact on V. cholerae MSHA and competence T4aP surface piliation. (A) Representative montage of piliated AF488-mal labeled cells for the pilus system indicated. Strains contain either mshAT70C (to label MSHA pili) or pilAS67C (to label competence pili). Phase images (left) are included from the first frame of the montage and show the cell boundary. Fluorescent images show AF488-mal labeled pili and there are 6-s intervals between frames. Scale bar = 1 μm. (B) Quantification of MSHA or competence T4aP retraction. Graph displays the percentage of piliated cells in each replicate that exhibited any pilus retraction during a three-minute timelapse. n ≥ 25 piliated cells analyzed for each of the three biological replicates. (C) Quantification of MSHA and competence pili in parent and ΔpilT cells as indicated. n = 300 cells analyzed from three independent biological replicates for all samples. All data are displayed as the mean ± SD. https://doi.org/10.1371/journal.pgen.1010561.g001 Deletion of pilT, however, exhibits distinct effects on these two T4aP systems. Loss of pilT results in a marked increase in the number of surface competence pili, which is characteristic of the unregulated extension observed in many other T4aP systems [14,30–33]. Conversely, the loss of PilT dramatically reduces the number of MSHA pili (Fig 1A and 1C). In fact, in the ΔpilT mutant, most cells lack MSHA pili, and the rare cells that do make pili generally display a single pilus that is longer than those observed in the parent (Fig 1A and 1C). This change in pilus number and length has been previously quantified and shown to differ significantly from the parent strain [23]. These phenotypes have also been qualitatively observed in multiple prior studies [4,15,22–25]. While the change in pilus length is visually striking, the major population-level change is the marked reduction in the frequency of piliated cells in the ΔpilT background (Fig 1C). One model that has been proposed for the observed reduction in MSHA surface piliation is that PilT is necessary to disrupt pilus extension after it is initiated [25]. Thus, in ΔpilT cells, a single MSHA T4aP machine undergoes uncontrolled “runaway” extension that exhausts the MshA major pilin pool of the cell. This would account for the markedly longer pili observed in the piliated ΔpilT cells (Fig 1A). This model predicts that a similar proportion of cells should be piliated in both the parent and ΔpilT strains, since initiation of extension would not be affected. However, as mentioned above, the ΔpilT strain shows a striking reduction in the percentage of piliated cells (Fig 1C). One possible explanation for this discrepancy could be that the long MSHA pili in the ΔpilT cells are sheared off the surface. If runaway extension was followed by shearing and loss of surface pili, we would expect ΔpilT cells to exhibit a distinct reduction in the level of cell-associated MshA major pilin. However, Western blot analysis indicates that cell-associated MshA major pilin is similar in the parent and ΔpilT strains (Fig 2A). In the T4aP systems of Pseudomonas aeruginosa and Myxococcus xanthus, major pilin gene transcription is increased in ΔpilT cells due to feedback regulation by the PilS-PilR two-component system [34–37]. V. cholerae lacks homologs of the PilS-PilR system, but it remains possible that ΔpilT cells also compensate for the loss of sheared pili by increasing expression of MshA major pilin. If so, the total amount of major pilins present in the cells + supernatant (i.e., unassembled + assembled, shed pili) should be markedly increased in the ΔpilT strain compared to the parent. However, we found that MshA protein levels were similar in the parent and ΔpilT samples even when the supernatants were included (Fig 2A). These results are not consistent with runaway extension of MSHA pili in ΔpilT cells. We next assessed whether ΔpilT cells were instead experiencing a biogenesis defect. If so, we would expect them to retain parental levels of intracellular major pilin. The Alexa Fluor 488-maleimide dye (AF488-mal) used to fluorescently label cysteine-modified MshA major pilins (MshAT70C) can pass through the outer membrane of V. cholerae and label pilins retained in the inner membrane [26]. Thus, the cell body fluorescence of labeled cells can serve as a proxy for the concentration of unassembled major pilins. To ensure that comparisons in this assay were not obscured by the fact that some strains have extended pili while others do not, we first labeled cells with AF488-mal, induced pilus retraction (see Methods for details), and then analyzed the cell-associated fluorescence of unpiliated cells as an indication of the concentration of MshA major pilin they possess in their inner membrane. We compared the cell-associated fluorescence of the parent strain, a ΔpilT strain, a ΔmshE strain, and a strain lacking the mshAT70C mutation. The ΔmshE mutant lacks the canonical extension ATPase and correspondingly cannot assemble pili (no surface pili are ever observed in this mutant background). It was therefore included as a control to demonstrate the cell body fluorescence of a strain that retains all its MshA major pilins in the inner membrane. The strain lacking the mshAT70C mutation (i.e., unlabelable) was included to indicate the background fluorescence observed in the absence of any labelable major pilins. We found that the cell body fluorescence of unpiliated AF488-mal labeled ΔpilT cells was more similar to the parent and ΔmshE strains than to the unlabelable strain (Fig 2B). This suggests that unpiliated ΔpilT cells retain a significant concentration of MshA in their inner membrane, a finding inconsistent with the runaway extension model. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 2. Reduced MSHA piliation of ΔpilT cells is likely due to a lack of pilus biogenesis rather than “runaway” extension. (A) Western blot analysis of MshA protein levels in parent and ΔpilT strains. Samples represent either cell-associated major pilin (cells only) or cell associated + pili shed in the supernatant (cells + sup). Band intensities are normalized to the RpoA loading control for quantification. Data are from three independent biological replicates. Data are displayed as the mean ± SD. (B) Quantification of cell body fluorescence in AF488-maleimide labeled cells as indicated. The parent, ΔpilT, and ΔmshE strains contain the mshAT70C mutation required for labeling the major pilin, while the unlabelable strain lacks this mutation. The latter is included here to denote the background cell body fluorescence observed in the absence of major pilins that can be labeled. The fluorescence of unpiliated cells was analyzed (as a measure of major pilin load in the inner membrane), and n ≥ 600 cells from three independent biological replicates for each strain. Data are presented as violin plots that demarcate the median (solid line) and quartile range (dotted lines). Representative images of cells used for quantification are included below. Phase images show cell boundaries and fluorescence images show AF488-mal labeled cells. Lookup tables are equivalent between fluorescence images. Scale bar = 1 μm. (C) Quantification of piliation in the indicated strains before and after induction with 0.2% arabinose. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples and data are displayed as the mean ± SD. (D) Representative montage of timelapse imaging of cells quantified in C. Scale bar = 2 μm. https://doi.org/10.1371/journal.pgen.1010561.g002 Another way to test the runaway extension hypothesis is to directly observe pilus biogenesis and extension dynamics in parent and ΔpilT cells. Because the ΔmshE cells retained fluorescently labeled MshA in their inner membrane (Fig 2B), we hypothesized that these AF488-mal labeled cells could be induced to extend their fluorescently labeled pilins into pilus filaments via ectopic induction of mshE. Furthermore, if mshE is induced in these cells under the microscope, this approach would circumvent any issues associated with pilus shearing, since pilus extension would be directly observed in single cells trapped under a gelzan pad. Therefore, if ΔpilT cells were undergoing runaway extension, we would expect most cells in the population to extend a single long pilus when mshE is induced. However, if the ΔpilT cells have an MSHA biogenesis defect, we would instead expect most cells in the population to remain unpiliated following mshE induction. To test this, we deleted the native copy of mshE and added an ectopic copy under the control of an arabinose-inducible promoter (PBAD-mshE). While parent cells (where pilT was intact) were able to extend pili after mshE was induced, ΔpilT cells showed a stark defect in pilus biogenesis (Fig 2C and 2D), providing further evidence against the runaway extension model. A lack of pilus biogenesis in the ΔpilT background could be due to an indirect downregulation of MshE extension motor activity. If so, we hypothesized that increased expression of mshE should recover surface piliation in the ΔpilT mutant. Overexpression of mshE, however, did not recover piliation in the ΔpilT background (S1 Fig). MshE activity is also allosterically regulated by cyclic-di-GMP (c-di-GMP), such that high levels of c-di-GMP stimulate MshE-dependent pilus extension [22,38]. Thus, the lack of MSHA biogenesis in ΔpilT cells could be due to a decrease in cellular levels of c-di-GMP, diminishing MshE activity. To test this, we increased the intracellular concentration of c-di-GMP in ΔpilT cells by overexpressing a previously characterized diguanylate cyclase, DcpA [39]. Elevated c-di-GMP also induces expression of the vps and rbm loci [40], which encode Vibrio polysaccharide and biofilm matrix proteins, respectively. To avoid cell clumping due to the expression of these biofilm factors and for ease of pilus quantification in these experiments, we deleted both the vps and rbm loci (ΔVC0917-VC0939; here called Δvps) in the background where dcpA was overexpressed. We found that overexpression of dcpA was insufficient to restore piliation in the ΔpilT background (S1 Fig). To further test whether c-di-GMP was involved in regulating MSHA piliation in the ΔpilT background, we took advantage of a previously-characterized allele of mshE–mshEL10A/L54AL/58A (denoted here as mshE*)–that genetically mimics the c-di-GMP bound state of MshE and promotes MSHA extension independently of c-di-GMP concentration [23,38]. Native expression and overexpression of mshE*, however, both failed to overcome the MSHA surface piliation defect of the ΔpilT background (S1 Fig). These results suggest that PilT does not promote MSHA biogenesis by indirectly affecting MshE abundance or interaction with c-di-GMP. Because PilT did not indirectly alter pilus number through known mechanisms of extension regulation, we hypothesized that PilT actively participates in MSHA biogenesis. If this were true, we would expect ΔpilT cells to extend MSHA pili very rapidly following PilT induction. To test this, we generated strains where pilT expression could be tightly controlled via an arabinose- and theophylline-inducible ectopic expression construct [41] (PBAD-ribo-pilT), and we measured piliation in cells before and after induction of the ectopic pilT construct. While only minor changes in piliation were observed in a strain still expressing native pilT, induction of ectopic pilT in the ΔpilT background rapidly recovered piliation to parental levels (Fig 3A and 3B). This increase was not observed in strains lacking the PBAD-ribo-pilT construct, indicating that this response was not due to a nonspecific effect of the arabinose and theophylline inducers (Fig 3A and 3B). Importantly, this effect was extremely rapid (within minutes) and occurred in single cells prior to cell division, using existing, prelabeled pilins (Figs 3 and S2). It is therefore unlikely that PilT is indirectly promoting biogenesis through regulation of T4aP machine assembly, because the structural elements of the T4aP machine are thought to be inserted into the cell envelope during cell growth and division [42]. Indeed, a fluorescent fusion to MshJ, a component of the MSHA T4aP machine, demonstrated that the number of MSHA machines was similar in the parent and ΔpilT backgrounds (Fig 3C). Because ectopic expression of pilT in these experiments stimulated MSHA biogenesis, these data also further confirm that ΔpilT cells are not simply unpiliated due to runaway extension. Together, these data suggest that in addition to its defined role in promoting MSHA retraction, PilT also participates in the biogenesis of MSHA pili. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 3. PilT promotes MSHA pilus extension. (A) Quantification of piliation in strains imaged before and after induction with 0.2% arabinose + 1.5 mM theophylline. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples and data are displayed as the mean ± SD. (B) Representative images of cells from strains in A. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Each sample is shown before (“-”) and after (“+”) induction. Scale bar = 2 μm. (C) Representative images of parent and ΔpilT strains containing a functional MshJ-mCherry fusion to label MSHA T4P machines. Phase images (top) show cell boundaries, green images (middle) show AF488-mal labeled pili, and red images (bottom) show MshJ-mCherry localization. Scale bar = 2 μm. https://doi.org/10.1371/journal.pgen.1010561.g003 Some T4aP pili can undergo motor-independent retraction via the spontaneous depolymerization of the pilus filament [43]. Taking this into account, there are at least two distinct mechanisms by which PilT could support MSHA biogenesis. PilT can either (1) help promote MSHA pilus assembly, or (2) it can help maintain extended MSHA pili on the surface. For the latter model to be true, ΔpilT cells must be continually undergoing extension to generate short pili that cannot be resolved by our fluorescent imaging approach, and the rapid motor-independent retraction of these short filaments in the absence of PilT prevents the processive extension necessary to generate visible surface pili in the parent strain. Under this model, once these pili are extended, PilT acts to prevent motor-independent retraction of these pili. If PilT is needed to maintain extended MSHA pili in this manner, we hypothesized that depleting PilT protein from piliated cells should result in the loss of surface pili due to the subsequent motor-independent retraction of those pili. The depletion of PilT was accomplished by adapting a previously described orthogonal degron system from Mesoplasma florum [44,45]. The translational fusion of a degron tag (called pdt2) to the C-terminus of native PilT (i.e., PilT-pdt2) allows for the specific and temporally regulated degradation of PilT by induction of the cognate M. florum Lon protease (mf-Lon). Strains with pilT-pdt2 exhibited parental levels of piliation without mf-Lon expression, while still exhibiting imaging-induced retraction, indicating that this fusion protein is functional (S3A–S3C Fig). Importantly, depletion of PilT in these experiments is performed in conditions where cells are not actively dividing, which allows us to assess the effect of PilT depletion on cells with pre-extended pili, after initial biogenesis has already taken place. Piliation in strains with untagged PilT did not change upon mf-Lon induction, indicating that this protease does not pleiotropically affect MSHA piliation. The PilT-pdt2 strain also showed no change in piliation upon induction of the mf-Lon protease (S3A and S3B Fig), indicating that the loss of PilT does not result in the loss of surface pili via motor-independent retraction. To assess whether PilT-pdt2 was indeed degraded in these experiments, we measured PilT-dependent retraction following depletion. Using imaging-induced retraction, we found that all strains exhibited normal retraction except for the PilT-pdt2 strain with induced mf-Lon (S3C Fig), which is consistent with the depletion of PilT in this sample. Together, these data indicate that MSHA pili do not exhibit spontaneous retraction in the absence of PilT. Thus, PilT is not required to maintain extended pili, but instead helps to promote MSHA pilus biogenesis. Because PilT is necessary for both MSHA extension and retraction, we next wanted to investigate whether both activities relied on its ATPase activity. In addition to PilT, V. cholerae also expresses PilU, a PilT-dependent accessory retraction motor ATPase. Loss of PilU alone (ΔpilU) did not impact MSHA pilus biogenesis, and the loss of PilU did not further exacerbate the piliation defect of the ΔpilT mutant (ΔpilTU) (Figs 4A, 4B and S4). PilU is incapable of mediating retraction in the absence of PilT, but it can promote retraction in the presence of an ATPase defective allele of PilT [15,25]. To test if PilTU-dependent MSHA retraction and extension were linked, we also assessed piliation in the ATPase defective Walker A mutants of PilT (pilTK136A) and PilU (pilUK134A). Although the pilTK136A allele is ATPase-defective, retraction can still occur in this background through the ATPase-activity of PilU (Fig 4C). While the retraction-capable pilTK136A and pilUK134A mutants exhibited parental levels of MSHA piliation, the retraction-deficient pilTK136A ΔpilU and pilTK136A pilUK134A mutants exhibited a loss of MSHA piliation (Fig 4A–4C). These data suggest that an ATPase active retraction motor–whether that is (1) PilT alone, or (2) a PilTU complex with functional PilU when PilT ATPase activity is compromised–is sufficient to promote MSHA pilus assembly. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 4. Retraction motor ATPase activity is required for MSHA surface piliation. (A) Representative images of piliated cells containing the indicated mutations to their retraction motor ATPases. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Scale bar = 1 μm. (B) Quantification of piliation in strains from A. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples. Data for the parent and ΔpilT strains are identical to that presented in Fig 1C and are included again here for ease of comparison. (C) Quantification of retraction in strains from A. Graphs display the percentage of cells in each replicate that exhibited any pilus retraction during a three-minute timelapse. n ≥ 25 piliated cells analyzed for each of the three biological replicates. All data are displayed as the mean ± SD. Parent and ΔpilT data from Fig 1C are included for comparison. https://doi.org/10.1371/journal.pgen.1010561.g004 Our data thus far are consistent with a model wherein the PilT(U) retraction motor also promotes MSHA extension. We hypothesized that if PilT contributes to MSHA extension, it may directly interact with the MshE extension motor in order to carry out this function. If this interaction is specific to the role of PilT in extension of MSHA pili, we would predict no such interaction between PilT and the competence pilus extension motor, PilB, since PilT is not required for biogenesis of competence T4aP (Fig 1). Because our data suggest that PilU can promote MSHA extension in the pilTK136A background, we also wanted to assess whether this accessory retraction ATPase exhibited a specific interaction with the MSHA extension motor. To assess interactions between these motors, we carried out bacterial adenylate cyclase two-hybrid (BACTH) analysis of the retraction motors (PilT and PilU) with the MSHA extension motor MshE and the competence pilus extension motor PilB. Interactions of the retraction motors with the MSHA platform protein (MshG) were also included as a positive control. The BACTH analysis showed that both PilT and PilU interacted with the MshG platform protein as expected (S5 Fig). Additionally, we found that the retraction motors (PilT and PilU) do interact with MshE (S5 Fig). However, these assays also showed interactions between the retraction motors and PilB (S5 Fig). Because the interaction between extension and retraction motors is not specific to the MSHA system, the physiological relevance for these interactions is unclear. Thus far, our data suggest that the PilT(U) retraction motor is required for MSHA pilus assembly. However, its precise role in the process is not yet established. One possibility is that PilT actually drives processive pilus extension via its ATPase activity. It has been shown in a number of T4P systems that the ATPase activity of the motor that drives dynamic activity correlates with the speed of extension/retraction [15,43,45,46]. Thus, if PilT’s ATPase activity drives processive extension, we hypothesized that a mutation that slows its ATPase activity should correspondingly slow down extension speed. Previous work has defined mutant alleles of T4P motor ATPases that reduce ATPase activity ~1.5-2-fold [15,43,45,47]. Therefore, we generated mutants of both pilT (pilTL201C; aka pilTslow) and mshE (mshEL390C; aka mshEslow) to slow their ATPase activity, and we tested the impact of these mutations on MSHA extension speed. MSHA pilus extension, however, is rarely observed in parent cells because the imaging conditions used to visualize pili induce MSHA retraction, as previously discussed. To counteract this limitation, we serendipitously found that some strains of V. cholerae are less susceptible to imaging-induced MSHA retraction. Namely, while the E7946 strain largely employed in this study is highly susceptible to imaging-induced retraction, the A1552 strain is much less sensitive under the same conditions. The A1552 strain exhibits the same loss of surface piliation when pilT is deleted, suggesting that the activity of PilT that promotes MSHA assembly is conserved across these V. cholerae strains [22]. Therefore, we employed the A1552 background to study pilus extension in these experiments. To facilitate direct observation of pilus extension, we generated strains where mshE could be induced while cells were imaged via timelapse microscopy (ΔmshE native + ectopic mshE expression construct). These strains also lacked native pilU, thereby eliminating any confounding effects this accessory motor ATPase might have on extension speed. When we assessed pilus extension speed in these strains, we found that while mshEslow did reduce extension speed, pilTslow had no impact on extension speed (Fig 5A). Furthermore, cells that harbored both mshEslow pilTslow exhibited an extension speed that was indistinguishable from mshEslow (Fig 5A). These data suggest that PilT does not drive processive extension of MSHA pili. Instead, the canonical extension ATPase in this system, MshE, is responsible for this activity. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 5. PilT and MshE promote distinct activities during MSHA extension. (A) The speed of MSHA T4P extension was determined in strains containing native motor ATPases (parent), or mutants that slow the ATPase activity of PilT (pilT slow) and/or MshE (mshE slow) as indicated. Because PilU is not required for MSHA piliation, all strains contained a ΔpilU mutation to prevent any confounding effects this motor ATPase might have on extension speed. For all strains, n = 75 from three independent experiments. Each data point indicates the extension speed for a dynamic T4P event. Box plots represent the median and the upper and lower quartile, while the whiskers demarcate the range. Statistical comparisons were made by One way ANOVA with Tukey’s post-test. NS = not significant; *** = p < 0.0001. (B) Schematic of the experimental setup to test whether PilT and MshE activity can be temporally separated in cells. In the absence of arabinose (Ara) induction (top), PilT-pd2 is left intact during mshE induction via anhydrotetracycline (ATc) (“PilT intact”). When cells are incubated with arabinose (bottom), PilT-pdt2 is degraded by Mf-Lon prior to the expression of mshE (“PilT depleted”). (C) Representative montage of cells subjected to the experiment described in B. Cells were imaged at 0, 20, and 40 minutes after induction of mshE expression. Scale bar = 3 μm. (D) Quantification of extension in the samples shown in C. Graph displays the percentage of cells in each replicate that exhibited pilus extension during a 40-minute window. n ≥ 100 cells analyzed for each of the three biological replicates and data are shown as the mean ± SD. (E) Quantification of imaging-induced retraction in samples shown in C. Graph displays the percentage of cells in each replicate that exhibited any pilus retraction during a three-minute timelapse. n ≥ 25 piliated cells analyzed for each of the three biological replicates and data are shown as the mean ± SD. https://doi.org/10.1371/journal.pgen.1010561.g005 Our data above suggest that PilT does not actively drive processive extension, a finding consistent with the observation that ΔpilT cells occasionally generate a single long pilus. This indicates that processive extension can still occur, albeit infrequently, even when PilT is absent. Based on these observations, we hypothesized that PilT may instead help initiate pilus extension. In canonical T4aP, pilus extension is initiated (or primed) by a complex of minor pilins—proteins that structurally resemble the major pilin but are expressed at significantly lower levels. One possibility is that the MSHA T4aP system minor pilins are not naturally capable of initiating pilus extension without the aid of PilT, which may help minor pilins assemble and/or mature into a complex that is capable of initiating pilus assembly. If this were the case, PilT and MshE would be acting at distinct steps during MSHA pilus assembly, and we would hypothesize that they would not need to be in the cell at the same time to promote extension. To test this, we used a strain where PilT-pdt2 could be degraded via ectopic induction of mf-Lon (PBAD-mf-Lon) and where mshE expression was under the control of a tightly inducible promoter (pLTetO-mshE). In these experiments, we grew cells without any inducers (so mshE was not expressed and pili were not extended) and then incubated them with AF488-mal to fluorescently label the MshA major pilin in their inner membrane. Next, these labeled cells were incubated either with or without arabinose to induce mf-Lon, which, if expressed, would deplete PilT-pdt2 from cells (Fig 5B). Importantly, the depletion of PilT in these experiments is performed in conditions where cells are not actively dividing. Thus, the effect that PilT exerts on cells prior to depletion could theoretically be maintained (i.e., if PilT facilitated maturation and/or assembly of the minor pilins into a complex) (Fig 5B). Following this incubation, cells were placed under gelzan pads containing anhydrotetracycline (ATc), which induces mshE expression, and pilus extension was assessed via microscopy. If PilT and MshE activity can be temporally separated, we would expect cells to still be able to extend pili even when PilT was depleted. When we performed this experiment, we found that many cells could still extend pili, even when PilT was depleted (Fig 5C and 5D). The percent of cells that extend pili is slightly lower when PilT is depleted (Fig 5C). However, the amount of extension observed is still much greater than when cells lacked PilT altogether (Fig 2C). The slightly reduced extension observed when PilT is depleted could be attributed to a short half-life for mature initiation complexes and/or due to the production of additional minor pilins while PilT was being depleted (and therefore could not be “matured” / assembled into initiation complexes). Importantly, we confirmed that piliated cells could not undergo imaging-induced retraction under the conditions where mf-Lon was induced, which phenotypically confirms that PilT-pdt2 is depleted in these experiments (Fig 5E). All together, these results suggest that PilT and MshE activity can be temporally separated in cells. This is consistent with a model in which PilT facilitates initiation of pilus assembly, while MshE promotes processive extension. Furthermore, because the activity of PilT and MshE can be temporally separated, these data also reaffirm that PilT is not indirectly affecting MshE expression and/or activity. To better understand factors regulating MSHA pilus extension, we also took an unbiased genetic approach. The presence of multiple surface pili is required for optimal MSHA-dependent attachment to surfaces [48]. Therefore, we took advantage of the poor attachment of ΔpilT cells to the walls of plastic culture tubes to select for suppressor mutants with improved binding. We hypothesized that this would occur through the restoration of MSHA surface piliation. Through this genetic selection, we isolated 14 spontaneous suppressor mutants that exhibited parental levels of binding and restored MSHA piliation. Whole genome resequencing revealed that all of the suppressor mutants had point mutations in the mshA major pilin, with 6 distinct mshA suppressor alleles identified (S6A Fig). These suppressor mutations predominantly clustered around the predicted kink within the N-terminal α-helix (Fig 6A). We reconstructed each of these mutations in clean genetic backgrounds, avoiding the use of rare codons, and saw that all of these mshA alleles increased MSHA piliation in the ΔpilT background (Fig 6B and 6C). Correspondingly, these mutations also restored surface attachment to parental levels (S6B Fig). Western blot analysis indicated that these alleles did not increase piliation by increasing the expression of the major pilin (Fig 6B). To confirm that these mshA suppressor alleles relied on the canonical MSHA extension machinery, we also assessed piliation in a ΔmshE background. We found that none of the mshA suppressor alleles displayed surface pili in the ΔmshE background (n >1,000 cells analyzed per strain), which confirmed that extension of these mshA suppressor alleles was still dependent on the canonical MSHA extension machinery. Furthermore, all of these suppressor mshA mutants still exhibited PilT-dependent imaging-induced retraction, just like the mshAparent (Fig 6D). Thus, these suppressor mutations specifically enhanced MSHA surface piliation in the ΔpilT background without markedly altering MSHA piliation or retraction in the presence of pilT. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 6. Point mutations in the major pilin mshA recover the piliation defect of a ΔpilT mutant. (A) Predicted structure of MshA, generated using AlphaFold [64,65]. Residues mutated in the suppressor selection are colored magenta. (B) Representative images of piliated cells and Western blotting of MshA in suppressor mutants. RpoA is included as a loading control. Phase images (top) show cell boundaries and fluorescence images (below) show AF488-mal labeled pili. Scale bar = 1 μm. In all panels, strains that retain native pilT are denoted “+” and strains with ΔpilT mutations are denoted “-”. (C) Quantification of piliation in strains from B. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples. (D) Quantification of retraction in strains from B. Data display the percentage of cells in each replicate that exhibited any pilus retraction during a three-minute timelapse. n ≥ 25 piliated cells analyzed for each of the three biological replicates. Data in C and D are shown as the mean ± SD. Parent and ΔpilT data from Fig 1C are included for comparison. https://doi.org/10.1371/journal.pgen.1010561.g006 Since our previous data suggest that PilT helps initiate MSHA pilus extension, we hypothesized that the mshA suppressor mutants bypass the need for PilT by independently promoting pilus extension. As mentioned above, in canonical T4aP, minor pilins form an initiation complex that primes pilus assembly [49–51]. However, the MSHA minor pilins may not naturally be capable of initiating pilus extension without the aid of PilT. If so, the suppressor mutations in mshA could help enhance MSHA pilus extension in the absence of PilT by functioning similarly to a canonical T4aP minor pilin. As a result, we hypothesized that only low concentrations of the mutated MshA suppressor allele would be required to initiate pilus extension, since minor pilins are generally expressed at low levels. We chose mshAP31S as a representative suppressor because it was centrally located among the other suppressors, it was hit most frequently in the selection, and because we predicted that the disruption of the proline might generate the strongest impact on the structure of the protein. To test induction of low levels of mshAP31S, we generated constructs that ectopically expressed mshAP31S under the control of a tightly regulated Ptac-riboswitch promoter, in a background that still expressed native mshA but lacked pilT (i.e., Ptac-ribo-mshAP31S ΔpilT). If MshAP31S helps initiate pilus extension like a minor pilin, we hypothesized that it should be able to recover piliation even at low levels of induction. However, we found that full induction of the mshAP31S allele was required to increase pilus number (S7A and S7B Fig). An equivalent construct inducing the wild-type mshA allele showed no change in pilus number, indicating that the observed effect on piliation requires ectopic expression of the mshAP31S suppressor allele. MshA expression from the Ptac-riboswitch-mshA construct was assessed by Western blot analysis in a background that lacked native mshA (S7C Fig), demonstrating that full induction resulted in native levels of MshA expression (S7C Fig). Furthermore, we found that ectopic mshA induction in the presence of native mshA (i.e., the scenario exhibited by the strains used in S7B Fig) resulted in additive levels of MshA (S7C Fig). These data suggest that the MshA suppressor alleles do not initiate pilus extension at low levels (akin to a canonical T4aP minor pilin). However, these alleles may still promote initiation of pilus assembly, albeit inefficiently. If MshAP31S does help initiate pilus extension in the absence of PilT, we hypothesized that the expression of this allele should promote assembly of wildtype MshA major pilins co-expressed in the same background. This cannot be distinguished in the experiments described above (S7A–S7C Fig) because both the native mshA allele and the ectopically expressed mshA allele contain the T70C mutation that allows for fluorescent labeling. We therefore generated strains where only one of the two mshA alleles contained the T70C mutation, so that the assembly of each allele could be distinguished. If MshAP31S can help initiate pilus assembly in the ΔpilT background, then we would expect ectopic expression of unlabeled MshAP31S to promote the assembly of natively expressed, labelable MshAT70C into fluorescent pilus filaments. Conversely, if MshAP31S does not help initiate pilus assembly and instead assembles into distinct filaments, then no change in surface piliation should be observed when unlabeled MshAP31S is induced. When we performed this experiment, we found that surface piliation was recovered following induction of MshAP31S (native mshAT70C + ectopic mshAP31S) to a level that resembled the scenarios using labelable MshAP31S (native mshAWT + ectopic mshAP31S,T70C; native mshAT70C + ectopic mshAP31S,T70C) (Fig 7). Importantly, ectopic overexpression of the mshAparent alleles that lacked the P31S mutation (native mshAWT + ectopic mshAT70C; native mshAT70C + ectopic mshAWT) did not restore surface piliation (Fig 7). Similar experiments using the mshAV27F and mshAR32L alleles also promoted assembly of labelable MshAT70C into pilus filaments, suggesting this activity was not unique to the mshAP31S allele (S8 Fig). Together, these data indicate that these mshA suppressor alleles promote the assembly of mshAWT into pilus filaments, which is consistent with a model wherein the mshA suppressor alleles increase PilT-independent piliation by promoting initiation of pilus extension. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 7. The MshAP31S suppressor allele promotes assembly of wild-type MshA. Representative images of cells from ΔpilT strains expressing the indicated mshA allele at the native locus (native mshA allele) and ectopic locus (Ptac-riboswitch-mshA allele). Cells were either grown with (”+”) or without (”-”) 100 μM IPTG + 1.5 mM theophylline to induce the ectopic Ptac-riboswitch-mshA allele in the strain as indicated. Only the alleles containing the T70C mutation can be labeled with AF488-mal and are denoted in green text in the table below the images. Lookup tables are equivalent between fluorescent images to reflect the expression level of the labelable mshA allele. Scale bar = 2 μm. https://doi.org/10.1371/journal.pgen.1010561.g007 Discussion Here, we have explored the mechanistic basis for the regulation of MSHA piliation in V. cholerae and found that piliation is dependent on both PilT and the major pilin (Fig 8). We demonstrate that PilT is not needed to disrupt runaway extension or maintain pili in their extended state. Instead, our data are consistent with a model in which PilT counterintuitively promotes MSHA pilus extension in addition to its established role in pilus retraction. The strongest evidence for this model comes from the observation that the tightly controlled ectopic expression of pilT in ΔpilT cells results in the rapid extension of MSHA pili and the observation that tightly controlled ectopic expression of mshE only recovers piliation when pilT is intact. Furthermore, we show that both PilT-dependent extension and retraction rely on the presence of a functional ATPase retraction motor, whether that is accomplished by PilT alone, or through the ATPase activity of PilU when PilT ATPase activity is inactivated. Download: PPT PowerPoint slide PNG larger image TIFF original image Fig 8. Model for the proposed mechanisms that promote MSHA extension and surface piliation. In wildtype cells (left), PilT promotes initiation of pilus extension, potentially by maturing the minor pilins into an initiation complex. Then, MshE can promote processive extension of the pilus in a c-di-GMP-dependent manner. The PilT-dependent accessory retraction ATPase motor PilU is dispensable for promoting pilus extension, but it can facilitate extension if PilT’s ATPase activity is inactivated. If PilT is absent (middle), initiation is inhibited, so processive extension is greatly diminished. Finally, mutations to the MshA major pilin (right), like MshAP31S, can promote initiation of MSHA extension even in the absence of PilT, suggesting that the structure of the major pilin is an important determinant of pilus extension. https://doi.org/10.1371/journal.pgen.1010561.g008 We found that while PilT(U) retraction motor ATPase activity is required for MSHA pilus assembly, that ATPase activity likely does not drive processive extension, and may not even need to be present during processive extension. These data are most consistent with a model wherein PilT is required for initiating pilus assembly (Fig 8). In canonical T4aP, the minor pilins spontaneously assemble within the pilus machine prior to processive extension [12,51]. Therefore, one possibility is that PilT promotes retraction early on when only the minor pilins are sitting in the pilus machine to reorient the initiation complex (i.e., the molecular equivalent of rotating a screw counterclockwise to ensure sure that the threads are seated properly). Alternatively, PilT could promote the interaction between minor pilins and the first major pilin to generate a “mature” initiation complex. Studying the impact of minor pilins on MSHA pilus assembly and the potential role for PilT in facilitating their assembly will be the focus of future work. Our results suggest that MSHA pilus biogenesis and dynamics may not simply be the result of direct competition between the extension and retraction motors in V. cholerae. Instead, we find that the combined activity of these motors is required for pilus extension. Recent work highlights that distinct motor ATPases can coordinate to promote pilus extension and/or retraction. As discussed above, PilU is an accessory retraction motor that works in conjunction with PilT to promote retraction of T4aP [15,25]. In Acinetobacter baylyi, optimal T4aP extension depends on the combined activity of two distinct extension motors, TfpB and PilB, where TfpB initiates pilus extension and PilB promotes processive extension [14]. MSHA extension may be similarly regulated, through the cooperative activities of an extension and retraction motor (i.e., MshE and PilT). Specifically, our data show that PilT is likely required for initiation of pilus extension, while MshE drives processive extension. PilT, however, is also clearly required for MSHA retraction. This bidirectional activity resembles Caulobacter crescentus CpaF, which is the sole motor ATPase expressed in this T4cP tad pilus system, and which drives both pilus extension and retraction [46]. Indeed, recent structural analysis of the transitions adopted by PilT suggest that this family of ATPase motors may be able to adopt distinct configurations that can switch the directionality of their activity (from retraction to extension) [52]. Because our data demonstrates that PilT ATPase activity does not directly drive extension speed, this distinguishes it from the bifunctional CpaF motor. Our suppressor screen also implicated the structure of the major pilin in regulating MSHA biogenesis. Most of the suppressor mutations obtained were localized near a proline in the widely conserved N-terminal α-helix of T4aP major pilins (i.e., the α1 subdomain) [53]. This proline introduces a kink into α1 [54], and is hypothesized to affect the packing of major pilin subunits within the filament. Indeed, mutations near the conserved proline in the major pilin of the Myxococcus xanthus T4aP have been shown to impact pilus assembly and/or retraction [55]. Because our data indicate that PilT promotes initiation of pilus extension, we hypothesize that these MshA suppressor alleles also enhance initiation of pilus extension to allow for pilus biogenesis in the absence of PilT. One possibility is that these suppressor alleles interact more efficiently with the immature minor pilins so that pilus assembly does not depend on the “maturation” of the initiation complex carried out by PilT(U) retraction motor ATPase activity (Fig 8). This is supported by the observation that the MshA suppressor alleles promote assembly of the MshAparent major pilin into a pilus filament. Additionally, we have recently shown that the packing of major pilin subunits within the pilus filament (as inferred by altered chemical properties of sheared pilus filaments) plays an important role in regulating motor-independent retraction of T4aP [43]. This suggests that the major pilin and the structure of the pilus filament are factors that likely contribute to the regulation of both T4aP extension and retraction. This study describes two unappreciated factors that influence pilus extension of the V. cholerae MSHA pili: the PilT(U) retraction motor ATPase and the structure of the MSHA major pilin. Because PilT is also required for piliation in other T4aP systems [20,21], it is likely that the mechanisms described here are not limited to the MSHA pili of V. cholerae. Also, recent phylogenetic analysis indicates that MSHA pilus systems form a discrete cluster from canonical T4aP [1]. In particular, MSHA systems contain pilus components of unknown function that are distinct from canonical T4aP (i.e., MshM, MshQ, and MshF) [1]. It is possible that these pilus components contribute to PilT-dependent extension. Thus, characterizing the function of the pilus components that are unique to MSHA T4aP moving forward may yield further insight into the broad diversity of mechanisms that regulate pilus dynamic activity. Ultimately, our findings describe a previously uncharacterized role for PilT. In addition to its canonical function as a retraction ATPase motor, we demonstrate that it is also required for the initiation of MSHA pilus assembly. These results expand our understanding of the factors that regulate the biogenesis and dynamic activity of these widely conserved bacterial surface appendages. Methods Bacterial strains and culture conditions Vibrio cholerae E7946 was used as the background for all strains employed in this study. Cultures were grown in LB broth (Miller) and plated on LB agar, and incubations were carried out at 30°C. Descriptions of all strains used in this study are found in S1 Table. Where appropriate, media were supplemented with kanamycin 50 μg/mL, trimethoprim 10 μg/mL, spectinomycin 200 μg/mL, chloramphenicol 1 μg/mL, carbenicillin 100 μg/mL, or zeocin 100 μg/mL. Construction of mutant strains Splicing-by-overlap extension PCR was used to generate mutant constructs, as described previously [56]. Natural transformation and cotransformation were used to introduce mutant constructs into chitin-induced competent V. cholerae as described previously [57,58]. Mutations were confirmed by PCR and/or sequencing. All ectopic expression constructs were derived from previously published strains [41]. The primers used to generate mutant constructs can be found in S2 Table. Microscopy and image analysis All phase contrast and fluorescence images were obtained with a Nikon Ti-2 microscope with a Plan Apo 60X objective, GFP filter cube, Hamamatsu ORCAFlash 4.0 camera, and Nikon NIS Elements imaging software. Images were prepared with Fiji [59,60]. Pili were visualized by Alexa Fluor 488-maleimide (AF488-mal) labelling. Cells were grown in LB to an OD600 of ~0.5 at 30°C rolling, with the exception of samples for imaging competence pili (which were grown in LB + 100 μM IPTG + 20 mM MgCl2 + 10 mM CaCl2 to an OD600 of ~2.5) or samples imaging mshJ-mCherry (which were grown in LB overnight). Cells were then washed and resuspended in instant ocean medium (7 g/L; Aquarium Systems) and incubated with 25 μg/mL AF488-mal for 10–15 min at room temperature. Cells were washed twice to remove excess dye and resuspended in instant ocean medium. After labelling, cells were loaded on a glass coverslip, covered with a 0.2% gelzan pad, and imaged. Lookup tables were chosen to best represent piliation for each image unless otherwise noted. Pilus number was assessed by randomly selecting 100 cells in a field of view in the phase channel for each replicate and then manually scoring piliation in each cell. The percent of piliated cells that exhibited retraction was assessed via imaging-induced retraction for each replicate by randomly selecting ~25 piliated cells and then observing retraction during a 3-minute timelapse (frames taken with 100 ms exposure every 3 sec). The retraction of any pili within a cell was used as a positive indicator of retraction. Three biological replicates were included for each dataset. Extension speeds were obtained by first labelling cells as above. Cells were then loaded onto a glass coverslip and covered with an 0.2% gelzan pad containing 0.2% arabinose inducer to induce mshE expression. Timelapses (3 second intervals over 3 minutes) were taken ~15–30 minutes after addition of the inducer. Pilus length was measured in the final frame of the extension event, and extension speed was calculated by dividing that length by the total extension time. To quantify cell body fluorescence, cells were labelled and imaged as described above, with the exception that cells were incubated with AF488-mal for 45 min. Pilus retraction (to ensure cells were unpiliated) was induced via imaging-induced retraction (i.e., a 4-minute timelapse was performed where images were captured every 3 seconds). The final image of the timelapse was used for quantification of cell body fluorescence. Cell body detection and fluorescence quantification were carried out using MicrobeJ [61]. Data sets were manually curated to remove piliated cells, overlapping cells, and dead cells. Temporally regulated induction of pilT Cells were first labelled, added to a coverslip, and covered with an 0.2% gelzan pad as described above. For S2 Fig, to observe how quickly cells could respond to inducer, cells were imaged under pads that either contained or lacked inducers (1.5 mM theophylline + 0.15% arabinose). An image was taken immediately upon exposure to the gelzan pad, and again after 15 min incubation at room temperature. For a more quantitative comparison of piliation following induction as shown in Fig 3, cells were first loaded under a gelzan pad containing no inducer and allowed to recover for 5 min. A second 0.2% gelzan pad containing 3 mM theophylline + 0.4% arabinose was then layered on top of the first gelzan pad. The use of stacked gelzan pads in this experiment allowed for the slower diffusion of inducer molecules to cells (i.e., via the diffusion of inducers through the gelzan pad in immediate contact with the cells). This ensured that any change in piliation was due to the effect of the inducer rather than recovery from conditions experienced during labeling and setup. Cells were imaged immediately (“- inducer”) and then again after 45 min, which we empirically determined was sufficient for the inducer to diffuse into the lower pad (“+ inducer”). Piliation states were quantified as above for each of the three biological replicates. Degradation of pilT Strains were labelled as described above and then allowed to incubate for 2.5 hrs with either inducer (0.2% arabinose) or vector control (an equivalent volume of water) to induce mf-Lon. Cells were then loaded onto a coverslip, covered with an 0.2% gelzan pad, and imaged as described above. Piliation state was scored as described above. Retraction was scored over a 3 min timelapse as described above, but categorized into either “no pili retract”, “some pili retract”, or “all pili retract”. This was to reflect the more subtle effects seen during PilT depletion compared to when pilT is deleted. For the experiment testing temporally distinct expression of MshE and PilT, cells were first labelled and PilT was degraded with 0.2% arabinose as described above. Cells were then loaded onto a coverslip and covered with a gelzan pad containing 50 ng/mL ATc to induce expression of mshE. Cells were imaged as above, after 0 min, 20 min, or 40 min in the presence of the inducer and scored for pilus extension. Pili were induced to retract through timelapse imaging and retraction was scored as described above. Western blotting Cells were grown to an OD600 of ~0.5 at 30°C rolling. To assess cell associated MshA, cells were spun down (~1.5 x 109 CFUs) and resuspended in an equivalent volume of instant ocean medium (i.e., cells only). To assess the amount of MshA that was cell associated and shed, an aliquot of culture was used directly for Western blot analysis (i.e., cells + sup). Samples were mixed 1:1 with 2 x SDS sample buffer [200 mM Tris pH 6.8, 25% glycerol, 1.8% SDS, 0.02% Bromophenol Blue, 5% β-mercaptoethanol] and then incubated at 95°C for 10 min, separated by SDS PAGE (15% gel), and then electrophoretically transferred to a PVDF membrane. This membrane was blocked with 5% milk and then incubated simultaneously with rabbit polyclonal α-MshA (1:1000, gift of J. Zhu) and mouse monoclonal α-RpoA (1:1000, Biolegend) (as a loading control) primary antibodies. Blots were washed and then incubated simultaneously with α-mouse and α-rabbit secondary antibodies conjugated to horseradish peroxidase enzyme. Blots were developed with Pierce ECL Western Blotting Substrate and then imaged with a ProteinSimple FluorChem R system. Bacterial adenylate cyclase two-hybrid assay Genes of interest were amplified via PCR and cloned into the BACTH vectors pKT25 and pUT18C to produce N-terminal fusions between the target protein and the T25 and T18 fragments of adenylate cyclase, respectively. Miniprepped plasmids (Qiagen) were then co-transformed into E. coli BTH101 cells. As a positive control, T25-zip and T18-zip vectors were cotransformed into BTH101. As a negative control, the pKT25 and pUT18C empty vectors were cotransformed into BTH101. Transformations were plated onto LB + kanamycin (50 μg/mL) + carbenicillin (100 μg/mL) plates to select for transformants that received both plasmids. These transformants were then picked and grown overnight at 30°C in LB + kanamycin (50 μg/mL) + carbenicillin (100 μg/mL). Finally, 3 μL of the overnight cultures were spotted onto LB agar plates supplemented with 500 μM IPTG, kanamycin (50 μg/mL), carbenicillin (100 μg/mL), and X-gal (40 μg/mL). These plates were allowed to incubate statically at 30°C for ~ 24 hours prior to imaging on an Epson Perfection V600 photo flatbed scanner. Genetic suppressor selection 20 independent cultures of ΔpilT Δvps were started from ~100 CFUs each to generate independent lineages for this suppressor screen. Cells were grown in polystyrene culture tubes to an OD600 of ~1.5 at 30°C rolling. A small aliquot was removed from each culture tube to be used for quantitative plating on LB agar (total cells). Tubes were then dumped and rinsed 5 times with fresh LB to remove unbound cells. Bound cells were then removed from the walls of the culture tube by adding 3 mL of fresh LB medium and vortexing vigorously for 30 sec. The cell slurry was then centrifuged, concentrated, and subjected to quantitative plating on LB agar (bound cells). The binding frequency was calculated for each sample by dividing the CFU after rinsing (bound cells) by the CFU before rinsing (total cells). Cells retrieved after binding were used to inoculate the next round of selection. Iterative rounds of selection were continued for each lineage until the binding frequency was at least 100x higher than the parental ΔpilT Δvps control (4–8 rounds selection). Genomic DNA from select clones was used to generate sequencing libraries by homopolymer tail mediated ligation PCR (HTML-PCR) exactly as previously described [62] and then sequenced on an Illumina Next-Seq instrument. The resulting sequencing data was mapped to the N16961 genome [63] using CLC Genomics WorkBench and the Basic Variant Analysis tool was used to identify SNVs. Binding assays Cells were grown to an OD600 of ~2.5 at 30°C shaking and 4 mL of culture was loaded into a new polystyrene culture tube. Tubes were incubated on a roller drum at room temperature for 20 min, during which time cells were allowed to bind to the walls of the tube. After incubation, a small aliquot was removed from each culture tube to be used for quantitative plating on LB agar (total cells). Then, tubes were dumped and rinsed 5 times with fresh LB to remove unbound cells. Bound cells were then removed from the walls of the culture tube by adding 1 mL of fresh LB medium and vortexing vigorously for 30 sec. The cell slurry was then centrifuged, concentrated, and subjected to quantitative plating on LB agar (bound cells). The binding frequency was calculated for each sample by dividing the CFU after rinsing (bound cells) by the CFU before rinsing (total cells). Three biological replicates were carried out for each strain. Bacterial strains and culture conditions Vibrio cholerae E7946 was used as the background for all strains employed in this study. Cultures were grown in LB broth (Miller) and plated on LB agar, and incubations were carried out at 30°C. Descriptions of all strains used in this study are found in S1 Table. Where appropriate, media were supplemented with kanamycin 50 μg/mL, trimethoprim 10 μg/mL, spectinomycin 200 μg/mL, chloramphenicol 1 μg/mL, carbenicillin 100 μg/mL, or zeocin 100 μg/mL. Construction of mutant strains Splicing-by-overlap extension PCR was used to generate mutant constructs, as described previously [56]. Natural transformation and cotransformation were used to introduce mutant constructs into chitin-induced competent V. cholerae as described previously [57,58]. Mutations were confirmed by PCR and/or sequencing. All ectopic expression constructs were derived from previously published strains [41]. The primers used to generate mutant constructs can be found in S2 Table. Microscopy and image analysis All phase contrast and fluorescence images were obtained with a Nikon Ti-2 microscope with a Plan Apo 60X objective, GFP filter cube, Hamamatsu ORCAFlash 4.0 camera, and Nikon NIS Elements imaging software. Images were prepared with Fiji [59,60]. Pili were visualized by Alexa Fluor 488-maleimide (AF488-mal) labelling. Cells were grown in LB to an OD600 of ~0.5 at 30°C rolling, with the exception of samples for imaging competence pili (which were grown in LB + 100 μM IPTG + 20 mM MgCl2 + 10 mM CaCl2 to an OD600 of ~2.5) or samples imaging mshJ-mCherry (which were grown in LB overnight). Cells were then washed and resuspended in instant ocean medium (7 g/L; Aquarium Systems) and incubated with 25 μg/mL AF488-mal for 10–15 min at room temperature. Cells were washed twice to remove excess dye and resuspended in instant ocean medium. After labelling, cells were loaded on a glass coverslip, covered with a 0.2% gelzan pad, and imaged. Lookup tables were chosen to best represent piliation for each image unless otherwise noted. Pilus number was assessed by randomly selecting 100 cells in a field of view in the phase channel for each replicate and then manually scoring piliation in each cell. The percent of piliated cells that exhibited retraction was assessed via imaging-induced retraction for each replicate by randomly selecting ~25 piliated cells and then observing retraction during a 3-minute timelapse (frames taken with 100 ms exposure every 3 sec). The retraction of any pili within a cell was used as a positive indicator of retraction. Three biological replicates were included for each dataset. Extension speeds were obtained by first labelling cells as above. Cells were then loaded onto a glass coverslip and covered with an 0.2% gelzan pad containing 0.2% arabinose inducer to induce mshE expression. Timelapses (3 second intervals over 3 minutes) were taken ~15–30 minutes after addition of the inducer. Pilus length was measured in the final frame of the extension event, and extension speed was calculated by dividing that length by the total extension time. To quantify cell body fluorescence, cells were labelled and imaged as described above, with the exception that cells were incubated with AF488-mal for 45 min. Pilus retraction (to ensure cells were unpiliated) was induced via imaging-induced retraction (i.e., a 4-minute timelapse was performed where images were captured every 3 seconds). The final image of the timelapse was used for quantification of cell body fluorescence. Cell body detection and fluorescence quantification were carried out using MicrobeJ [61]. Data sets were manually curated to remove piliated cells, overlapping cells, and dead cells. Temporally regulated induction of pilT Cells were first labelled, added to a coverslip, and covered with an 0.2% gelzan pad as described above. For S2 Fig, to observe how quickly cells could respond to inducer, cells were imaged under pads that either contained or lacked inducers (1.5 mM theophylline + 0.15% arabinose). An image was taken immediately upon exposure to the gelzan pad, and again after 15 min incubation at room temperature. For a more quantitative comparison of piliation following induction as shown in Fig 3, cells were first loaded under a gelzan pad containing no inducer and allowed to recover for 5 min. A second 0.2% gelzan pad containing 3 mM theophylline + 0.4% arabinose was then layered on top of the first gelzan pad. The use of stacked gelzan pads in this experiment allowed for the slower diffusion of inducer molecules to cells (i.e., via the diffusion of inducers through the gelzan pad in immediate contact with the cells). This ensured that any change in piliation was due to the effect of the inducer rather than recovery from conditions experienced during labeling and setup. Cells were imaged immediately (“- inducer”) and then again after 45 min, which we empirically determined was sufficient for the inducer to diffuse into the lower pad (“+ inducer”). Piliation states were quantified as above for each of the three biological replicates. Degradation of pilT Strains were labelled as described above and then allowed to incubate for 2.5 hrs with either inducer (0.2% arabinose) or vector control (an equivalent volume of water) to induce mf-Lon. Cells were then loaded onto a coverslip, covered with an 0.2% gelzan pad, and imaged as described above. Piliation state was scored as described above. Retraction was scored over a 3 min timelapse as described above, but categorized into either “no pili retract”, “some pili retract”, or “all pili retract”. This was to reflect the more subtle effects seen during PilT depletion compared to when pilT is deleted. For the experiment testing temporally distinct expression of MshE and PilT, cells were first labelled and PilT was degraded with 0.2% arabinose as described above. Cells were then loaded onto a coverslip and covered with a gelzan pad containing 50 ng/mL ATc to induce expression of mshE. Cells were imaged as above, after 0 min, 20 min, or 40 min in the presence of the inducer and scored for pilus extension. Pili were induced to retract through timelapse imaging and retraction was scored as described above. Western blotting Cells were grown to an OD600 of ~0.5 at 30°C rolling. To assess cell associated MshA, cells were spun down (~1.5 x 109 CFUs) and resuspended in an equivalent volume of instant ocean medium (i.e., cells only). To assess the amount of MshA that was cell associated and shed, an aliquot of culture was used directly for Western blot analysis (i.e., cells + sup). Samples were mixed 1:1 with 2 x SDS sample buffer [200 mM Tris pH 6.8, 25% glycerol, 1.8% SDS, 0.02% Bromophenol Blue, 5% β-mercaptoethanol] and then incubated at 95°C for 10 min, separated by SDS PAGE (15% gel), and then electrophoretically transferred to a PVDF membrane. This membrane was blocked with 5% milk and then incubated simultaneously with rabbit polyclonal α-MshA (1:1000, gift of J. Zhu) and mouse monoclonal α-RpoA (1:1000, Biolegend) (as a loading control) primary antibodies. Blots were washed and then incubated simultaneously with α-mouse and α-rabbit secondary antibodies conjugated to horseradish peroxidase enzyme. Blots were developed with Pierce ECL Western Blotting Substrate and then imaged with a ProteinSimple FluorChem R system. Bacterial adenylate cyclase two-hybrid assay Genes of interest were amplified via PCR and cloned into the BACTH vectors pKT25 and pUT18C to produce N-terminal fusions between the target protein and the T25 and T18 fragments of adenylate cyclase, respectively. Miniprepped plasmids (Qiagen) were then co-transformed into E. coli BTH101 cells. As a positive control, T25-zip and T18-zip vectors were cotransformed into BTH101. As a negative control, the pKT25 and pUT18C empty vectors were cotransformed into BTH101. Transformations were plated onto LB + kanamycin (50 μg/mL) + carbenicillin (100 μg/mL) plates to select for transformants that received both plasmids. These transformants were then picked and grown overnight at 30°C in LB + kanamycin (50 μg/mL) + carbenicillin (100 μg/mL). Finally, 3 μL of the overnight cultures were spotted onto LB agar plates supplemented with 500 μM IPTG, kanamycin (50 μg/mL), carbenicillin (100 μg/mL), and X-gal (40 μg/mL). These plates were allowed to incubate statically at 30°C for ~ 24 hours prior to imaging on an Epson Perfection V600 photo flatbed scanner. Genetic suppressor selection 20 independent cultures of ΔpilT Δvps were started from ~100 CFUs each to generate independent lineages for this suppressor screen. Cells were grown in polystyrene culture tubes to an OD600 of ~1.5 at 30°C rolling. A small aliquot was removed from each culture tube to be used for quantitative plating on LB agar (total cells). Tubes were then dumped and rinsed 5 times with fresh LB to remove unbound cells. Bound cells were then removed from the walls of the culture tube by adding 3 mL of fresh LB medium and vortexing vigorously for 30 sec. The cell slurry was then centrifuged, concentrated, and subjected to quantitative plating on LB agar (bound cells). The binding frequency was calculated for each sample by dividing the CFU after rinsing (bound cells) by the CFU before rinsing (total cells). Cells retrieved after binding were used to inoculate the next round of selection. Iterative rounds of selection were continued for each lineage until the binding frequency was at least 100x higher than the parental ΔpilT Δvps control (4–8 rounds selection). Genomic DNA from select clones was used to generate sequencing libraries by homopolymer tail mediated ligation PCR (HTML-PCR) exactly as previously described [62] and then sequenced on an Illumina Next-Seq instrument. The resulting sequencing data was mapped to the N16961 genome [63] using CLC Genomics WorkBench and the Basic Variant Analysis tool was used to identify SNVs. Binding assays Cells were grown to an OD600 of ~2.5 at 30°C shaking and 4 mL of culture was loaded into a new polystyrene culture tube. Tubes were incubated on a roller drum at room temperature for 20 min, during which time cells were allowed to bind to the walls of the tube. After incubation, a small aliquot was removed from each culture tube to be used for quantitative plating on LB agar (total cells). Then, tubes were dumped and rinsed 5 times with fresh LB to remove unbound cells. Bound cells were then removed from the walls of the culture tube by adding 1 mL of fresh LB medium and vortexing vigorously for 30 sec. The cell slurry was then centrifuged, concentrated, and subjected to quantitative plating on LB agar (bound cells). The binding frequency was calculated for each sample by dividing the CFU after rinsing (bound cells) by the CFU before rinsing (total cells). Three biological replicates were carried out for each strain. Supporting information S1 Fig. The piliation defect in the ΔpilT mutant cannot be recovered by increasing MshE expression or activity. (A) Quantification of piliation in strains with altered extension motor regulation. Strains that retain native pilT are denoted “+” and strains with ΔpilT mutations are denoted “-”. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples and data are displayed as the mean ± SD. All strains with ectopic PBAD constructs were induced with 0.2% arabinose. Parent and ΔpilT data from Fig 1C are included for comparison. (B) Representative images of piliated cells from strains in A. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Strains that retain native pilT are denoted “+” and strains with ΔpilT mutations are denoted “-”. Scale bar = 1 μm. https://doi.org/10.1371/journal.pgen.1010561.s001 (TIFF) S2 Fig. Induction of pilT rapidly recovers MSHA surface piliation. Representative images of a PBAD-ribo-pilT ΔpilT strain when pilT expression is rapidly induced under the microscope. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Samples were applied to a slide with a gelzan pad either lacking inducer (left; “- inducer”) or including 0.2% arabinose and 1.5mM theophylline as an inducer (“+ inducer”). Each sample was imaged immediately after application of the gelzan pad (“0 min”) and again after 15 min (“+15 min”). Scale bar = 2 μm. https://doi.org/10.1371/journal.pgen.1010561.s002 (TIFF) S3 Fig. PilT is not required to maintain MSHA surface piliation. (A) Representative images of cells where mf-Lon was (“+”; 0.2% arabinose) or was not (“-”) induced. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Scale bar = 1 μm. (B) Quantification of piliation in samples from A. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples. (C) Quantification of imaging-induced retraction in samples from A. Graph displays the percentage of cells in each replicate that retracted all, some, or no pili during a three-minute timelapse. n ≥ 25 piliated cells analyzed for each of the three biological replicates. All data are displayed as the mean ± SD. https://doi.org/10.1371/journal.pgen.1010561.s003 (TIFF) S4 Fig. PilT does not require PilU to promote MSHA surface piliation as long as its ATPase activity is intact. Fluorescence microscopy of the indicated strains. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Each sample is shown before (-) and after (+) induction. Scale bar = 4 μm. Data are representative of three independent experiments. https://doi.org/10.1371/journal.pgen.1010561.s004 (TIFF) S5 Fig. PilT and PilU directly interact with the MshE and PilB extension motors. Representative image of a BACTH assay between T25- and T18- fusions of the indicated proteins. Both PilT and PilU displayed strong interactions with themselves (“self”), the extension motors (MshE and PilB), as well as the MSHA platform protein (MshG). “E.V.” denotes a pairing with an empty vector, and “+” indicates the BACTH positive control (T25-Zip + T18-Zip). This image is representative of three independent experiments. https://doi.org/10.1371/journal.pgen.1010561.s005 (TIFF) S6 Fig. MshA suppressor mutants enhance adherence. (A) Table of the MshA mutations isolated in the ΔpilT suppressor screen. The relative position of the mutated residues is indicated relative to the MshA start codon (location in unprocessed pilin) and relative to the first amino acid in the pilin after being processed by the pre-pilin peptidase (location in processed pilin).”Number of independent hits” denotes the number of distinct genetic lines in which the indicated suppressor mutation was isolated. (B) Binding frequency of the indicated strains to the wall of culture tubes. Strains that retain native pilT are denoted “+” and strains with ΔpilT mutations are denoted “-”. Data are from three independent biological replicates and shown as the mean ± SD. https://doi.org/10.1371/journal.pgen.1010561.s006 (TIFF) S7 Fig. Full induction of MshAP31S is required to recover piliation in the absence of pilT. (A) Representative images of piliated cells from strains under increasing induction of either Ptac-riboswitch-mshAparent or Ptac-riboswitch-mshAP31S. All mshA alleles in these strains (native and ectopic) contain the T70C mutation needed for AF488-mal labeling. Concentrations of inducer used from left to right: (1) 0 μM IPTG + 0 μM theophylline, (2) 5 μM IPTG + 75 μM theophylline, (3) 20 μM IPTG + 300 μM theophylline, and (4) 100 μM IPTG + 1.5 mM theophylline. Phase images (top) show cell boundaries and fluorescence images (bottom) show AF488-mal labeled pili. Scale bar = 2 μm. (B) Quantification of piliation in samples from A. Cells were categorized as either having no pili (white bars), 1–2 pili (light gray bars), or at least 3 pili (dark gray bars). n = 300 cells analyzed from three independent biological replicates for all samples. (C) Western blot quantification of cell-associated MshA in the indicated strains in the induction conditions used in A and B. Band intensities are normalized to the RpoA loading control. Data are from three independent biological replicates. All data are displayed as the mean ± SD. https://doi.org/10.1371/journal.pgen.1010561.s007 (TIFF) S8 Fig. The MshAV27F and MshAR32L suppressor alleles promotes assembly of wild-type MshA. Representative images of cells from ΔpilT strains expressing the indicated mshA allele at the native locus (native mshA allele) and ectopic locus (Ptac-riboswitch-mshA allele). Cells were either grown with (”+”) or without (”-”) 100 μM IPTG + 1.5 mM theophylline to induce the ectopic Ptac-riboswitch-mshA allele in the strain as indicated. Only the alleles containing the T70C mutation can be labeled with AF488-mal and are denoted in green text in the table below the images. Scale bar = 2 μm. https://doi.org/10.1371/journal.pgen.1010561.s008 (TIFF) S1 Table. Strains used in this study. https://doi.org/10.1371/journal.pgen.1010561.s009 (PDF) S2 Table. Primers used in this study. https://doi.org/10.1371/journal.pgen.1010561.s010 (PDF) Acknowledgments We would like to thank Clay Fuqua, Julia van Kessel, and Tuli Mukhophadyay for helpful discussions.