Nucleic Acids Research

doi: N/Apmid: N/A

Transcription coupled nucleotide excision repair by isolated Escherichia coli membrane-associated nucleoidsC.G.Lin, O.Kovalsky and L.GrossmanNucleic Acids Res. (1998), 26, 1466–1472The authors wish to note that there is an error in Figure 2; the numerical values within the lower graph of (A) describing the concentration of Rif are incorrectly in reverse order. The correct figure is shown below.

doi: 10.1093/nar/26.12.ipmid: N/A

Abstract Transcription coupled nucleotide excision repair by isolated Escherichia coli membrane-associated nucleoids C.G.Lin, O.Kovalsky and L.Grossman Nucleic Acids Res. (1998), 26 , 1466–1472 The authors wish to note that there is an error in Figure 2; the numerical values within the lower graph of (A) describing the concentration of Rif are incorrectly in reverse order. The correct figure is shown below. View largeDownload slide View largeDownload slide © 1998 Oxford University Press

Jordan, Peter;Carmo-Fonseca, Maria

doi: 10.1093/nar/26.12.2831pmid: 9611224

Abstract Cis -diammininedichloroplatinum(II) (cisplatin or cis -DDP) is a DNA-damaging agent that is widely used in cancer chemotherapy. Cisplatin crosslinks DNA and the resulting adducts interact with proteins that contain high-mobility-group (HMG) domains, such as UBF (upstream binding factor). UBF is a transcription factor that binds to the promoter of ribosomal RNA (rRNA) genes thereby supporting initiation of transcription by RNA polymerase I. Here we report that cisplatin causes a redistribution of UBF in the nucleolus of human cells, similar to that observed after inhibition of rRNA synthesis. A similar redistribution was observed for the major components of the rRNA transcription machinery, namely TBP, TAF I s and RNA polymerase I. Furthermore, we provide for the first time direct in vivo evidence that cisplatin blocks synthesis of rRNA, while activity of RNA polymerase II continues to be detected throughout the nucleus. The clinically ineffective trans isomer ( trans -DDP) does not alter the localization of either UBF or other components of the RNA polymerase I transcription machinery. These results suggest that disruption of rRNA synthesis, which is stimulated in proliferating cells, plays an important role in the clinical success of cisplatin. Introduction The inorganic compound cis -diammininedichloroplatinum(II), commonly referred to as cisplatin or cis -DDP, is one of the most widely used anticancer drugs with well established effectiveness against a number of cancers, particularly metastatic testicular tumors (for recents reviews see 1, 2 ). Although the mode of action of cisplatin has been under intensive study since its discovery more than 30 years ago ( 3 ), the molecular basis for the biological effects of this drug are not entirely clear. Cisplatin forms covalent adducts with many biological molecules, but its principal target is DNA. The most prevalent DNA lesion caused by cisplatin is a 1,2-intrastrand crosslink, with the platinum covalently bound to the N 7 positions of adjacent purine bases. Other platinum-DNA adducts include 1,3-intrastrand, interstrand and protein-DNA crosslinks. The formation of the major 1,2-intrastrand cross-links is most probably responsible for the biological activity of cisplatin, since the stereoisomer trans -DDP cannot form this type of adduct and is clinically inactive ( 1 , 2 , 4 ). Recently, much attention has been focused on high-mobility-group (HMG)-domain proteins, which bind specifically to cisplatin-damaged DNA and may therefore participate in the cytotoxic effects of the drug ( 4 ). The HMG domain is a DNA-binding motif of ∼80 amino acid residues and HMG-domain proteins are architectural proteins that function to bend DNA ( 5 ). This class of proteins recognizes DNA structural elements present in linear, cruciform or cisplatin-modified DNA ( 6 ). Importantly, HMG-domain proteins bind to DNA lesions induced by cisplatin but not by trans -DDP, and the binding is specific for the major 1,2-intrastrand adducts. This strongly suggests that HMG-domain proteins are involved in cisplatin cytotoxicity, and several models have been proposed to explain their role. One possibility is that recruitment of HMG-domain proteins to cisplatin-lesioned DNA masks the DNA damage from being repaired. Another possibility is that the cisplatin-DNA adducts may titrate or ‘hijack’ HMG-domain proteins from their normal sites of action thereby disrupting normal gene expression. Alternatively, binding of HMG-domain proteins to cisplatin-lesioned DNA may displace certain tumor-specific regulatory DNA-binding proteins, resulting in tumor cell death. It is also possible that binding of HMG-domain proteins to unplatinated DNA may create a favorable DNA conformation for cisplatin to act upon (reviewed in 2 ). The model that cisplatin-damaged DNA may serve as a molecular decoy was based on the observation that in vitro , human upstream binding factor (hUBF, a member of the HMG-domain protein family) binds with high avidity to the intrastrand crosslinks produced by the drug. As UBF is an essential transcription factor for RNA polymerase I, it was suggested that rRNA synthesis may be disrupted if the adducts hijack hUBF in vivo ( 7 ). Here, we have analysed the effects of cisplatin on the subnuclear distribution of UBF and other components of the machinery for RNA polymerase I transcription. The results show that cisplatin blocks synthesis of rRNA and causes a redistribution of UBF, RNA polymerase I, TATA-binding protein (TBP) and TBP-associated factors for RNA polymerase I (TAF I s) to the periphery of the nucleolus, where they co-localize with inactive rDNA genes. In contrast, activity of RNA polymerase II continues to be detected throughout the nucleus. This suggests that disruption of rRNA synthesis, which is stimulated in proliferating cells, plays an important role in the clinical success of cisplatin. Materials and Methods Cell culture and drug treatment HeLa monolayer cultures were maintained mycoplasm-free in Dulbecco's modified minimum essential medium supplemented with 10% fetal calf serum. Stock solutions of 10 mg/ml cis -DDP and trans -DDP (Sigma) were prepared in dimethylformamide and stored at −20°C. Alternatively, cisplatin was dissolved in sterile phosphate buffered saline (PBS) (1 mg/ml; 3.3 mM) and each stock was used for no longer than 3 days (cf. 8). Immunofluorescence For indirect immunofluorescence the cells were grown on 10 × 10 mm glass coverslips and harvested at 50–70% confluency. Coverslips with attached cells were washed twice in PBS, fixed with 3.7% formaldehyde in PBS for 10 min at room temperature, and subsequently permeabilized with 0.5% Triton X-100 in PBS for 15 min at room temperature. Alternatively, the cells were first permeabilized with 0.5% Triton X-100 in CSK buffer ( 9 ) containing 0.1 mM PMSF for 1 min on ice, and then fixed with 3.7% formaldehyde in CSK for 10 min at room temperature. Formal-dehyde (7.4% stock solution) was prepared from paraformaldehyde and stored frozen until use. After fixation and permeabilization, the cells were rinsed in PBS containing 0.05% Tween 20 (PBS-T), incubated for 30–60 min with primary antibodies diluted in PBS-T, washed, and incubated for 30 min with the appropriate secondary antibodies conjugated to either fluorescein or Texas Red (Vector Laboratories, UK). Finally, the coverslips were mounted in VectaShield (Vector Laboratories, UK) and sealed with nail polish. The following antibodies were used: rabbit polyclonal antibody E29 raised against human UBF ( 10 ), auto-immune human anti-RNA polymerase I serum S18 (kindly provided by Dr U. Scheer), rabbit anti-hTBP KD 55/1 ( 11 ) and rabbit anti-hTAF I 63 sera ( 12 ). Visualization of transcription sites Visualization of transcription sites was performed essentially according to Jackson et al . ( 13 ). Briefly, cells grown as monolayers on coverslips were washed twice in PBS and incubated with 0.05% Triton X-100 in PB buffer for 2 min on ice, in order to permeabilize selectively the plasma membrane. Then, the cells were incubated with a transcription-mix containing bromo-UTP (Sigma) for 20 min at 33°C. Subsequently, the cells were further incubated in 0.2% Triton X-100 for 10 min, fixed with 3.7% formaldehyde for 10 min, and the incorporated bromo-uridine detected using a monoclonal antibody directed against bromodeoxyuridine (Boehringer Mannheim, Germany). In order to visualize the sites of transcription by distinct RNA polymerases, the transcription assay was performed in the presence of selective inhibitors. The drugs were added for 10 min prior to, as well as during incubation with the modified nucleotide. In the presence of α-amanitin at a concentration of 100 µg/ml (to inhibit RNA polymerases II and III), the labeling was exclusively observed in the nucleolus. In the presence of actinomycin D at a concentration of 0.08 µg/ml (to inhibit transcription by polymerase I), the labeling was detected in the nucleoplasm but not in the nucleolus, and in the presence of 5 µg/ml actinomycin D (which inhibits transcription by all RNA polymerases), all labeling was abolished. Visualization of DNA replication For in vivo labeling of replication sites, cells were pulse-labeled in culture medium with 1 µM bromodeoxyuridine (BrdU; Boehringer Mannheim) for 30 min. For detection of incorporated BrdU the cells were washed with PBS, fixed in pre-cooled 70% ethanol, 50 mM glycine for 20 min at −20°C, rinsed in PBS, incubated with 4 N HCl for 8 min and rinsed again in PBS. The cells were then blocked with 5% fetal calf serum in PBS for 15 min and incubated with mouse monoclonal anti-bromodeoxyuridine antibody (Boehringer Mannheim) and an appropriate secondary antibody coupled to FITC. After washing the samples were mounted in VectaShield. Microscopy Samples were analysed using the laser scanning microscope Zeiss LSM410 equiped with an Argon Ion laser (488 nm) to excite FITC fluorescence and a Helium-Neon laser (543 nm) to excite Texas Red fluorescence. For double-labeling experiments, images from the same focal plane were sequentially recorded in different channels and superimposed. In order to obtain a precise alignment of superimposed images the equipment was calibrated using multicolor fluorescent beads (Molecular Probes, Eugene, USA), and a dual-band filter that allows simultaneous visualization of red and green fluorescence. Run-on transcription assay Cells were grown in 10 cm plates and scraped-off in PBS. To isolate nuclei, the cells were incubated on ice for 5 min in lysis buffer (10 mM Tris-HCl, pH 7.4, 3 mM MgCl 2 , 10 mM NaCl, 0.5% NP-40) and centrifuged at 500 g . Nuclei were resuspended in 120 µl glycerol buffer (50 mM Tris-HCl, pH 8.3, 5 mM MgCl 2 , 40% glycerol) and mixed with 1 vol 2× reaction buffer (10 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 300 mM KCl, 1 mM of each GTP, ATP, CTP and 60 µCi [α- 32 P]UTP) and incubated for 30 min at 30°C. Subsequently, DNA was digested for 30 min at 30°C by adding 360 µl DNase-buffer (10 mM Tris-HCl, pH 7.4, 50 mM MgCl 2 , 2 mM CaCl 2 , 500 mM NaCl) containing 15 U of RNase-free DNase (Promega). Protein was then digested for 30 min at 42°C by adding 120 µl PK-buffer (500 mM Tris-HCl, pH 7.4, 125 mM EDTA, 5% SDS) containing 150 µg Proteinase K. RNA was isolated by phenol/chloroform extraction, precipitated with ethanol, dissolved in 200 µl TE and separated from unincorporated nucleotides by gel-filtration through a Sephadex G50 spin-column. The labeled, total RNA was then denatured for 10 min on ice by adding 15 µl 2 N NaOH, neutralized with 30 µl 1 N HCl and 12 µl 1 M Tris-HCl, pH 7.5, before hybridization. As target DNA, a plasmid containing 28S rDNA (pBS28S; 14 ) was linearised by digestion with Eco RI, then denatured for 30 min at room temperature with 0.1 vol of 0.1 N NaOH, neutralized with 10 vol of 6× SSC and dot blotted onto a nitrocellulose membrane. After prehybridisation of the membrane for 3 h at 42°C in hybridisation buffer (50% formamide, 6× SSC, 1× Denhardt's reagent, 100 µg/ml denatured tRNA), the labeled RNA was added to the mixture and hybridized for 3 days at 42°C. The membrane was then washed once in 1× SSC, 0.1% SDS for 20 min at 42°C, three times in 0.2× SSC, 0.1% SDS for 20 min at 68°C, and finally exposed to a Kodak X-Omat film. Figure 1 View largeDownload slide In vivo effects of cis - and trans -DDP on DNA replication. HeLa cells were treated for 20 h with either cis - ( E and F ) or trans -DDP ( C and D ) (final concentration 20 µg/ml, from 10 mg/ml stock solutions dissolved in dimethylformamide). Mock treated cells were incubated for the same time in the presence of 0.2% dimethylformamide ( A and B ). Following drug treatment, the cells were incubated with 1 µM BrdU for 30 min and the bromyl residues incorporated into newly synthesized DNA strands were visualized by indirect immunofluorescence. (A), (C) and (E) depict confocal fluorescent images recorded and printed using precisely the same settings in order to allow comparison of the signal intensities. (B), (D) and (F) depict the corresponding phase contrast images; note that cells were initially plated at similar densities. Bar, 10 µm. Figure 1 View largeDownload slide In vivo effects of cis - and trans -DDP on DNA replication. HeLa cells were treated for 20 h with either cis - ( E and F ) or trans -DDP ( C and D ) (final concentration 20 µg/ml, from 10 mg/ml stock solutions dissolved in dimethylformamide). Mock treated cells were incubated for the same time in the presence of 0.2% dimethylformamide ( A and B ). Following drug treatment, the cells were incubated with 1 µM BrdU for 30 min and the bromyl residues incorporated into newly synthesized DNA strands were visualized by indirect immunofluorescence. (A), (C) and (E) depict confocal fluorescent images recorded and printed using precisely the same settings in order to allow comparison of the signal intensities. (B), (D) and (F) depict the corresponding phase contrast images; note that cells were initially plated at similar densities. Bar, 10 µm. Results Cis - and trans -DDP affect DNA replication in vivo Since it has been previously demonstrated that both cis - and trans -DDP block DNA replication in vitro ( 4 ), we analysed whether a similar effect occurs in vivo . Stock solutions of cis - and trans -DDP were prepared in dimethylformamide and added to the tissue culture medium at a final concentration of 20 µg/ml for 20 h. This corresponds roughly to the estimated blood concentration of cisplatin in a patient administered intravenously with the clinically used dosage of 50–120 mg/m 2 body surface area ( 8 ). Mock-treated cells were incubated for the same time in dimethylformamide (final concentration 0.2%). After drug treatment, bromo-deoxyuridine (BrdU, 1 µM) was added to the culture medium and incubated for 30 min. Subsequently the cells were fixed and the incorporated bromyl residues were detected by indirect immunofluorescence. In mock-treated samples, ∼28 % of the cells were brightly labeled ( Fig. 1A and B ), indicating that DNA synthesis was actively taking place. In samples treated with trans -DDP, the proportion of labeled cells decreased to ∼15%; moreover, the fluorescence intensity in these cells was significantly lower than in controls ( Fig. 1 C and D), indicating a general reduction in DNA synthesis. Treatment with cisplatin completely abolished incorporation of BrdU indicating a complete block of DNA replication ( Fig. 1E and F ). As cisplatin is commonly administered to patients diluted with physiologic saline ( 8 ), we next prepared fresh stocks of the drug in PBS and performed a time course analysis of its effect on DNA synthesis. A decrease in both the number and the intensity of BrdU-labeled cells was first noticed at 3 h, whereas by 5 h after drug treatment no labeling was observed (data not shown). In conclusion, these results show that both cis - and trans -DDP affect DNA replication in vivo . However, the inhibitory effect of cisplatin is significantly higher than that of trans -DDP. Cisplatin induces a redistribution of UBF and other major components of the RNA polymerase I transcription machinery According to a currently proposed model, when HMG-domain proteins such as UBF bind cisplatin-damaged DNA, they can be displaced from their natural binding sites on the genome. To further address this view, we analysed the effect of cisplatin on the localization of UBF in the nucleus of human cells. Initially, cells were either mock-treated with 0.2% dimethylformamide or treated with 20 µg/ml cis - or trans -DDP for 20 h. In both mock-treated cells and cells exposed to trans -DDP, UBF is localised in intranucleolar foci ( Fig. 2A and B ). A similar distribution pattern of UBF was previously described in non-treated cells and shown to correspond to sites of transcription by RNA polymerase I ( 15 , 16 ). Following exposure to cisplatin, UBF is detected in large aggregates which form ‘caps’ at the periphery of the nucleolus ( Fig. 2 C). This redistribution of UBF is detected as early as 5 h after treatment with a concentration of 20 µg/ml cisplatin. Decreasing the concentration of cisplatin to 10 µg/ml produces a similar effect within 8 h of treatment, whereas at 1 µg/ml the redistribution of UBF occurs after 20 h of exposure to the drug ( Fig. 3A and B ). No significant increase in the extra-nucleolar staining produced by anti-UBF antibody is observed ( Fig. 3A and B ), as recently reported by Chao et al . ( 17 ). Previous studies from our and other laboratories have shown that UBF colocalizes with components of the nucleolar transcriptional machinery, namely TBP, TAF I s and RNA polymerase I, when rRNA synthesis is either active or inactive ( 16 , 18 ). We therefore performed double-labeling experiments in cells treated with either cisplatin or trans -DDP. The results show that the redistribution of UBF induced by cisplatin is paralleled by a redistribution of TBP, TAF I s and RNA polymerase I, which continue to co-localize with each other ( Fig. 3 C and D; and data not shown). In contrast, treatment with trans -DDP does not affect the distribution of any of these factors (data not shown). Transcription of ribosomal RNA is blocked in the nuclei of cisplatin-treated cells Since cisplatin causes a redistribution of the RNA polymerase I transcription machinery similar to that observed when rRNA synthesis is inhibited ( 16 ), we next studied the effects of the drug on transcription. HeLa cells were mildly treated with Triton X-100 in order to selectively permeabilize the plasma membrane, and incubated with bromo-uridine 5′-triphosphate (Br-UTP). Subsequently, the incorporated Br-UTP was visualized by immunofluorescence using anti-bromyl antibodies. The results show several hundred fluorescent foci scattered throughout the nucleoplasm of untreated cells ( Fig. 4A ). Similar data has been previously reported and shown to correspond to sites of transcription by RNA polymerase II ( 13 , 19–22 ). Surprisingly, no significant changes in either labeling pattern or signal intensity are observed following exposure of cells to cisplatin ( Fig. 4B and C ), indicating that the drug does not block the activity of RNA polymerase II. Figure 2 View largeDownload slide Cisplatin induces a redistribution of UBF in the nucleolus. HeLa cells were either mock-treated ( A ) or treated with trans -DDP ( B ) or cisplatin ( C ) for 20 h (final concentration 20 µg/ml). Following drug treatment, indirect immunofluorescence was performed using anti-UBF antibodies. The panels depict a superimposition of confocal fluorescent images with the corresponding phase contrast images. In mock-treated cells and in cells treated with trans -DDP, UBF is localized in discrete foci scattered within the nucleolus [(A) and (B), arrows]. In contrast, in cells treated with cisplatin, UBF is detected in large cap-like structures located at the outer surface of the nucleolus [(C), arrows]. Bar, 10 µm. Figure 2 View largeDownload slide Cisplatin induces a redistribution of UBF in the nucleolus. HeLa cells were either mock-treated ( A ) or treated with trans -DDP ( B ) or cisplatin ( C ) for 20 h (final concentration 20 µg/ml). Following drug treatment, indirect immunofluorescence was performed using anti-UBF antibodies. The panels depict a superimposition of confocal fluorescent images with the corresponding phase contrast images. In mock-treated cells and in cells treated with trans -DDP, UBF is localized in discrete foci scattered within the nucleolus [(A) and (B), arrows]. In contrast, in cells treated with cisplatin, UBF is detected in large cap-like structures located at the outer surface of the nucleolus [(C), arrows]. Bar, 10 µm. Figure 3 View largeDownload slide UBF co-localizes with RNA polymerase I in cisplatin-treated cells. Indirect immunofluorescence using anti-UBF antibodies was performed on HeLa cells that were either untreated ( A ) or treated for 20 h with 1 µg/ml cisplatin (from a fresh stock solution dissolved in PBS) ( B ). Note the redistribution of UBF similar to that observed in cells treated with higher concentration of the drug (cf. Fig. 2 C). ( C and D ) Cells treated for 5 h with 20 µg/ml cisplatin (from a fresh stock solution dissolved in PBS) and double-labeled using anti-RNA poly-merase I (C) and anti-UBF (D) antibodies. Bar, 10 µm. Figure 3 View largeDownload slide UBF co-localizes with RNA polymerase I in cisplatin-treated cells. Indirect immunofluorescence using anti-UBF antibodies was performed on HeLa cells that were either untreated ( A ) or treated for 20 h with 1 µg/ml cisplatin (from a fresh stock solution dissolved in PBS) ( B ). Note the redistribution of UBF similar to that observed in cells treated with higher concentration of the drug (cf. Fig. 2 C). ( C and D ) Cells treated for 5 h with 20 µg/ml cisplatin (from a fresh stock solution dissolved in PBS) and double-labeled using anti-RNA poly-merase I (C) and anti-UBF (D) antibodies. Bar, 10 µm. Figure 4 View largeDownload slide In situ labeling of transcription sites with BrUTP. HeLa cells were either untreated ( A and D ), or treated with cisplatin for the indicated time ( B , C , E and F) ; stock solution of cisplatin prepared in PBS. Cells were incubated with BrUTP after a selective permeabilization of the plasma membrane, and the incorporated bromyl residues were visualized by immunofluorescence. (A-C) show that BrUTP is similarly incorporated in the nucleoplasm of treated and untreated cells. In contrast, when cells are incubated with BrUTP in the presence of 100 µg/ml amanitin (which inhibits RNA polymerases II and III), transcription sites are visualized in the nucleoli of untreated cells (D), but no signal is detected in cisplatin-treated cells (E and F). Bar, 10 µm. Figure 4 View largeDownload slide In situ labeling of transcription sites with BrUTP. HeLa cells were either untreated ( A and D ), or treated with cisplatin for the indicated time ( B , C , E and F) ; stock solution of cisplatin prepared in PBS. Cells were incubated with BrUTP after a selective permeabilization of the plasma membrane, and the incorporated bromyl residues were visualized by immunofluorescence. (A-C) show that BrUTP is similarly incorporated in the nucleoplasm of treated and untreated cells. In contrast, when cells are incubated with BrUTP in the presence of 100 µg/ml amanitin (which inhibits RNA polymerases II and III), transcription sites are visualized in the nucleoli of untreated cells (D), but no signal is detected in cisplatin-treated cells (E and F). Bar, 10 µm. To specifically analyse transcription by RNA polymerase I, the cells were incubated with Br-UTP in the presence of 100 µg/ml α-amanitin to inhibit RNA polymerases II and III ( 13 , 16 ). Under these conditions, the labeling is exclusively observed in nucleolar foci of untreated cells ( Fig. 4D ), whereas no labeling is detected in cells treated with cisplatin ( Fig. 4E and F ). To further investigate this apparent inhibition of rRNA synthesis induced by the drug, a nuclear run-on transcription assay was performed. Nuclei from untreated and cisplatin-treated cells were isolated and incubated with [ 32 P]UTP. Total RNA was then isolated and hybridized to 28S rDNA immobilized on nitrocellulose. As depicted in Figure 5 , cisplatin has a clear inhibitory effect on the synthesis of rRNA. Figure 5 View largeDownload slide Run-on transcription assay. HeLa cells were either untreated or treated with 20 µg/ml cisplatin for 3 or 5 h. From these samples, nuclei were isolated and incubated with [α- 32 P]UTP for 30 min. Subsequently, RNA was isolated and separated from unincorporated nucleotides by gel-filtration. The RNA was then denatured and hybridized to a nitrocellulose membrane. As target DNA, a plasmid containing 28S rDNA was linearized, denatured and dot blotted. Figure 5 View largeDownload slide Run-on transcription assay. HeLa cells were either untreated or treated with 20 µg/ml cisplatin for 3 or 5 h. From these samples, nuclei were isolated and incubated with [α- 32 P]UTP for 30 min. Subsequently, RNA was isolated and separated from unincorporated nucleotides by gel-filtration. The RNA was then denatured and hybridized to a nitrocellulose membrane. As target DNA, a plasmid containing 28S rDNA was linearized, denatured and dot blotted. We therefore conclude that in vivo cisplatin affects preferentially the synthesis of rRNA. Discussion Although it has been postulated that cisplatin has deleterious consequences for DNA replication and transcription, very few studies have addressed the direct effects of this drug on human cells. Here we describe that cisplatin causes a redistribution of the RNA polymerase I transcription machinery similar to that observed after inhibition of rRNA synthesis. Furthermore we show that cisplatin blocks preferentially the synthesis of rRNA in vivo . Accurate and specific transcription of rRNA genes by RNA polymerase I requires at least two transcription factors, UBF and SL1 (promoter selectivity factor) ( 23 , 24 ). UBF is a DNA binding protein that binds to the upstream control element of the promoter; it consists of two alternatively spliced polypeptides of 97 and 94 kDa (UBF1 and UBF2, respectively) and was found to be highly conserved among various vertebrates ( 25 , 26 ). UBF interacts with DNA by way of multiple HMG-box domains and in doing so greatly enhances recruitment of SL1, which appears to mediate communication with RNA polymerase I ( 27 , 28 ). Thus, if UBF fails to bind to the promoter, RNA polymerase I cannot initiate transcription. SL1 is a multisubunit complex containing TBP and three TAF I s of 110, 63 and 48 kDa, which are essential to activate rRNA transcription ( 12 ). As the UBF-promoter interaction is highly sensitive to the antagonistic effects of cisplatin-DNA adducts ( 7 , 29 ), it was predicted that the drug could disrupt rRNA synthesis. Our results fully support this view. Importantly, our data show that the DNA adducts induced by the clinically ineffective trans isomer of cisplatin, which are not recognised by UBF ( 7 ), do not interfere with the localization of the RNA polymerase I transcriptional machinery. This strongly suggests that block of rRNA synthesis is involved in the cytotoxicity of cisplatin. Previous work revealed that cisplatin-DNA adducts cause an elongation block during in vitro RNA synthesis by prokaryotic and eukaryotic RNA polymerases ( 30 , 31 ). Additionally, a 2–3-fold decrease in transcriptional level was observed when platinum-modified reporter genes were transfected into human or hamster cells ( 32 ). Cisplatin was also shown to substantially reduce transcription from the mouse mammary tumor virus promoter stably incorporated into mouse cells ( 33 ). Moreover, the basal transcription factor TBP was recently reported to bind selectively and competitively to cisplatin-damaged DNA, thereby inhibiting transcription from an independent and transcriptionally viable template ( 34 ). It is therefore surprising that cisplatin-treated HeLa cells continue to incorporate Br-UTP in the nucleoplasm ( Fig. 3 ). However, it should be emphasized that, in contrast with the previous studies, we have analysed the effects of cisplatin on endogenous transcription levels of the cellular genome. One possibility to reconcile these data is that in vivo rDNA represents a preferential target for cisplatin damage. In fact, cisplatin does not remove UBF from the nucleolus, and the peripheral caps containing the rRNA transcriptional machinery in inactivated nucleoli have been previously shown to co-localize with rDNA ( 16 ). Thus, it seems unlikely that the block of rRNA synthesis is due to hijacking of UBF from the nucleolus to other sites in the nucleus. Alternatively, we favour the view that the high concentration of proteins that bind to cisplatin-damaged DNA (such as UBF and TBP) in the nucleolus is responsible for a higher incidence of cisplatin lesions in rDNA. In this regard it is noteworthy that a yeast strain deprived of Ixr1 (an HMG-domain protein that binds to platinated DNA) was reported to be half as sensitive to cisplatin and accumulated one-third as many platinum-DNA lesions as the wild-type strain ( 35 ). At least two models are consistent with these data. One is that UBF (and possibly also TBP, as part of SL1) shields rDNA from repair enzymes and therefore cisplatin lesions persist and accumulate in the nucleolus, while extra-nucleolar lesions are more efficiently repaired. The other possibility is that binding of UBF and TBP to rDNA facilitates the formation of cisplatin lesions in the nucleolus. In conclusion, our data suggest that a preferential inhibition of rRNA synthesis is likely to be involved in the cytotoxicity of cisplatin. Because ribosomal transcription regulates ribosome production and, consequently, the translation potential of a cell, it is conceivable that deregulation of ribosomal transcription may, in the long term, be an important determinant in neoplastic transformation ( 36 ). By blocking ribosomal transcription, cisplatin may therefore be preferentially cytotoxic to rapidly proliferating transformed cells. Acknowledgements The authors are grateful to Prof. David-Ferreira for support and to Dr Celso Cunha for help in some experiments. We also thank the following laboratories for generously providing reagents used in this study: Dr L. I. Rothblum for providing the anti-UBF antibodies and the plasmid pBS28S, Prof. U. Scheer for anti-polymerase I autoimmune serum, Dr Zomerdijk and Prof. R. Tjian for the anti-TAF I sera, and Dr R. Bravo for anti-TBP polyclonal antibodies. This work was supported by grants from Junta Nacional de Investigação Científica e Tecnológica (Program PRAXIS XXI). P.J. was a recipient of a post-doctoral fellowship from the European Union (Human Capital and Mobility Program). References 1 Pinedo H.M.,  Schornagel J.H.. ,  Platinum and Other Metal Coordination Compounds in Cancer Chemotherapy ,  1996 New York/London Plenum 2 Yang D.Z.,  Wang A.H.J.. ,  Prog. Biophys. Mol. Biol. ,  1997, vol.  1 (pg.  81- 111) 3 Rosenberg B.,  vanCamp L.,  Kirgan T.. ,  Nature ,  1965, vol.  205 (pg.  698- 699) CrossRef Search ADS PubMed  4 Zamble D.B.,  Lippard S.J.. ,  Trends Biochem. Sci. ,  1995, vol.  20 (pg.  435- 439) CrossRef Search ADS PubMed  5 Grosschedl R.,  Giese K.,  Pagel J.. ,  Trends Genet. ,  1994, vol.  10 (pg.  94- 100) CrossRef Search ADS PubMed  6 Chow C.S.,  Whitehead J.P.,  Lippard S.J.. ,  Biochemistry ,  1994, vol.  33 (pg.  15124- 15130) CrossRef Search ADS PubMed  7 Treiber D.K.,  Zhai X.,  Jantzen H.-M.,  Essigmann J.M.. ,  Proc. Natl. Acad. Sci. USA ,  1994, vol.  91 (pg.  5672- 5676) CrossRef Search ADS   8 Loehrer P.J.,  Einhorn L.H.. ,  Annals Int. Med. ,  1984, vol.  100 (pg.  704- 713) CrossRef Search ADS   9 Fey E.G.,  Krochmalnic G.,  Penman S.. ,  J. Cell Biol. ,  1986, vol.  102 (pg.  1654- 1665) CrossRef Search ADS PubMed  10 O'Mahony D.J.,  Xie W.Q.,  Smith S.D.,  Singer H.A.,  Rothblum L.I.. ,  J. Biol. Chem. ,  1992, vol.  267 (pg.  35- 38) PubMed  11 Metz R.,  Kouzarides T.,  Bravo R.. ,  EMBO J. ,  1994, vol.  13 (pg.  3832- 3842) PubMed  12 Comai L.,  Zomerdijk J.C.B.M.,  Beckmann H.,  Zhou S.,  Admon A.,  Tjian R.. ,  Science ,  1994, vol.  266 (pg.  1966- 1972) CrossRef Search ADS PubMed  13 Jackson D.A.,  Hassan A.B.,  Errington R.J.,  Cook P.R.. ,  EMBO J. ,  1993, vol.  12 (pg.  1059- 1065) PubMed  14 Rothblum L.I.,  Parker D.L.,  Cassidy B.. ,  Gene ,  1982, vol.  17 (pg.  75- 77) CrossRef Search ADS PubMed  15 Zatsepina O.V.,  Voigt R.,  Grummt I.,  Spring H.,  Semenov M.V.,  Trendelenburg M.F.. ,  Chromosoma ,  1993, vol.  102 (pg.  599- 611) CrossRef Search ADS PubMed  16 Jordan P.,  Mannervik M.,  Tora L.,  Carmo-Fonseca M.. ,  J. Cell Biol. ,  1996, vol.  133 (pg.  225- 234) CrossRef Search ADS PubMed  17 Chao J.C.,  Wan X.S.,  Engelsberg B.N.,  Rothblum L.I.,  Billings P.C.. ,  Biochem. Biophys. Acta ,  1996, vol.  1307 (pg.  213- 219) 18 Roussel P.,  Andre C.,  Comai L.,  Hernandez-Verdun D.. ,  J. Cell Biol. ,  1996, vol.  133 (pg.  235- 246) CrossRef Search ADS PubMed  19 Wansink D.G.,  Schul W.,  van der Kraan I.,  van Steensel B.,  van Driel R.,  de Jong L.. ,  J. Cell Biol. ,  1993, vol.  122 (pg.  283- 293) CrossRef Search ADS PubMed  20 Pombo A.,  Ferreira J.,  Bridge E.,  Carmo-Fonseca M.. ,  EMBO J. ,  1994, vol.  13 (pg.  5075- 5085) PubMed  21 Carmo-Fonseca M.,  Cunha C.,  Custódio N.,  Carvalho C.,  Jordan P.,  Ferreira J.,  Parreira L.. ,  Exp. Cell Res. ,  1996, vol.  229 (pg.  247- 252) CrossRef Search ADS PubMed  22 Zeng C.,  Kim E.,  Warren S.L.,  Berget S.M.. ,  EMBO J. ,  1997, vol.  16 (pg.  1401- 1412) CrossRef Search ADS PubMed  23 Bell S.P.,  Learned R.M.,  Jantzen H.-M.,  Tjian R.. ,  Science ,  1988, vol.  241 (pg.  1192- 1197) CrossRef Search ADS PubMed  24 Schnapp A.,  Grummt I.. ,  J. Biol. Chem. ,  1991, vol.  266 (pg.  24588- 24595) PubMed  25 O'Mahony D.J.,  Rothblum L.I.. ,  Proc. Natl. Acad. Sci. USA ,  1991, vol.  88 (pg.  3180- 3184) CrossRef Search ADS   26 Hisatake K.,  Nishimura T.,  Maeda Y.,  Hanada K.,  Song C.-Z.,  Muramatsu M.. ,  Nucleic Acid Res. ,  1991, vol.  19 (pg.  4631- 4637) CrossRef Search ADS   27 Jantzen H.-M.,  Admon A.,  Bell S.P.,  Tjian R.. ,  Nature ,  1990, vol.  344 (pg.  830- 836) CrossRef Search ADS PubMed  28 Bazett-Jones D.P.,  Leblanc B.,  Herfort M.,  Moss T.. ,  Science ,  1994, vol.  264 (pg.  1134- 1137) CrossRef Search ADS PubMed  29 Codonyservat J.,  Gimeno R.,  Gelpi C.,  Rodriguez Sanchez J.L.,  Juarez C.. ,  Biochem. Pharmacol. ,  1996, vol.  51 (pg.  1131- 1136) CrossRef Search ADS PubMed  30 Corda Y.,  Job C.,  Anin M.-F.,  Leng M.,  Job D.. ,  Biochemistry ,  1991, vol.  30 (pg.  222- 230) CrossRef Search ADS PubMed  31 Corda Y.,  Anin M.-F.,  Leng M.,  Job D.. ,  Biochemistry ,  1992, vol.  31 (pg.  1904- 1908) CrossRef Search ADS PubMed  32 Mello J.A.,  Lippard S.J.,  Essigmann J.M.. ,  Biochemistry ,  1995, vol.  34 (pg.  14783- 14791) CrossRef Search ADS PubMed  33 Mymryk J.S.,  Zaniewski E.,  Archer T.K.. ,  Proc. Natl. acad. Sci. USA ,  1995, vol.  92 (pg.  2076- 2080) CrossRef Search ADS   34 Vichi P.,  Coin F.,  Renaud J.-P.,  Vermeulen W.,  Hoeijmakers J.H.J.,  Moras D.,  Egly J.-M.. ,  EMBO J. ,  1997, vol.  16 (pg.  7444- 7456) CrossRef Search ADS PubMed  35 Brown S.J.,  Kellett P.J.,  Lippard S.J.. ,  Science ,  1993, vol.  261 (pg.  603- 605) CrossRef Search ADS PubMed  36 Moss T.,  Stefanovsky V.Y.. ,  Progress Nucleic Acid Res. Mol. Biol. ,  1995, vol.  50 (pg.  25- 66) © 1998 Oxford University Press

Jordan, Peter; Carmo-Fonseca, Maria

doi: N/Apmid: N/A

Cis-diammininedichloroplatinum(II) (cisplatin or cis-DDP) is a DNA-damaging agent that is widely used in cancer chemotherapy. Cisplatin crosslinks DNA and the resulting adducts interact with proteins that contain high-mobility-group (HMG) domains, such as UBF (upstream binding factor). UBF is a transcription factor that binds to the promoter of ribosomal RNA (rRNA) genes thereby supporting initiation of transcription by RNA polymerase I. Here we report that cisplatin causes a redistribution of UBF in the nucleolus of human cells, similar to that observed after inhibition of rRNA synthesis. A similar redistribution was observed for the major components of the rRNA transcription machinery, namely TBP, TAFIs and RNA polymerase I. Furthermore, we provide for the first time direct in vivo evidence that cisplatin blocks synthesis of rRNA, while activity of RNA polymerase II continues to be detected throughout the nucleus. The clinically ineffective trans isomer (trans-DDP) does not alter the localization of either UBF or other components of the RNA polymerase I transcription machinery. These results suggest that disruption of rRNA synthesis, which is stimulated in proliferating cells, plays an important role in the clinical success of cisplatin.

Pehrson, John R.;Fuji, Reina N.

doi: 10.1093/nar/26.12.2837pmid: 9611225

Abstract Histone macroH2A is an unusual core histone that contains a large non-histone region, and a region that resembles a full length H2A. We examined the conservation of this novel structural arrangement by cloning chicken macroH2A cDNAs and comparing them to their rat counterparts. The amino acid sequences of the two known macroH2A subtypes are >95% identical between these species despite evolutionary separation of ∼300 million years. The H2A region of macroH2A is completely conserved, and thus is even more conserved than conventional H2A in these species. The origin of the non-histone domain was examined by comparing its sequence to proteins found in bacteria and RNA viruses. These comparisons indicate that this domain is derived from a gene that originated prior to the appearance of eukaryotes, and suggest that the non-histone region has retained the basic function of its ancestral gene. Introduction Core histones are among the most evolutionarily conserved proteins in eukaryotes. This conservation is presumably the result of the critical role that nucleosomes play in DNA packaging and gene regulation. We discovered a new type of core histone, macroH2A (mH2A), in rat liver nucleosomes ( 1 ). The N-terminal third of mH2A is 64% identical to a full length H2A. MacroH2A also contains a large region that does not resemble any other known histone ( Fig. 1 ). This large non-histone region distinguishes mH2A from all other known core histones. Sequences from mH2A cDNAs and reactions with mH2A-specific antibodies established the existence of two distinct mH2A proteins in mammalian tissues ( 1 , 2 ). These subtypes are called mH2A1.1 and mH2A1.2, and they differ from one another in only one region ( Fig. 1 ). The nucleotide sequences of the cDNAs that encode these subtypes are identical in both their coding and non-coding regions except for one short segment in the non-histone region ( 1 , 2 ). This indicates that these subtypes are produced from the same gene by alternate splicing; this has been confirmed by cloning the rat mH2A1 gene (unpublished results). Subtype specific functions are suggested by studies that showed that mH2A1.1 and 1.2 proteins have distinct patterns of expression during development and in different adult organs ( 2 ). In rat liver, we estimated that there is one mH2A for every 30 nucleosomes ( 1 ). We recently identified a third mH2A subtype that is produced from a separate gene that we call mH2A2 (unpublished results). The unusual structure of mH2A suggests that it is functionally distinct from conventional H2As. Consistent with this possibility, we recently showed that mH2A is preferentially concentrated in the inactive X chromosome of female mammals ( 3 ). This association suggests that mH2A participates in the transcriptional silencing of this chromosome. In the present work we sought to identify the regions of mH2A that are most directly involved in its function(s) by identifying regions that are highly conserved in evolution. We cloned and sequenced chicken mH2A cDNAs, and compared them to those previously known from the rat. These two species separated in evolution ∼300 million years ago ( 4 ) and prior to the appearance of X chromosome inactivation, which occurred only in mammals ( 5 ). We also examined the origins of the H2A and non-histone regions by comparing their sequences to known proteins. Materials and Methods Cloning and sequencing of chicken macroH2A cDNAs A chicken liver cDNA library ( 6 ) was screened with the non-H2A region of a rat mH2A1.1 cDNA. A positive plaque was identified, and the insert was cloned into pBluescript KS+ (Stratagene). Nested deletions of the mH2A insert were generated using exonuclease III digestion ( 7 ). Subclones were sequenced ( 8 ) using the Sequenase DNA sequencing kit (US Biochemicals). Both strands were sequenced except for two small segments of the 3′ non-coding region. Separate sequences were generated incorporating either dGTP or dITP as a substitute for dGTP. Reactions with dITP were treated with terminal deoxynucleotidyl transferase to reduce artefacts associated with the use of dITP (US Biochemicals). This chicken mH2A cDNA was missing the region of non-identity between mH2A1.1 and 1.2, and therefore encodes a truncated mH2A protein. Attempts to confirm the expression of this truncated mH2A gave ambiguous results. The polymerase chain reaction (PCR) was used to amplify mH2A cDNAs that contain the region that was missing from the cDNA clone discussed above. One ng of cDNA from chicken liver, brain, spleen or muscle served as a template and amplification was achieved by 30 cycles of 1 min at 95°C, 30 s at 50°C and 2 min at 72°C. The reactions were carried out in a standard reaction buffer (Promega) containing 1.5 mM MgCl 2 , and Tli DNA polymerase (Promega) was used to minimize mutations during amplification. The primer sequences were ggaattccAAGAAGCAGGGAGAAGT and ggaattccACAAACTCCTTGCCGCC; these sequences include sites for restriction nuclease cleavage used in cloning the products. A small amount of the PCR products was radiolabeled using T4 polynucleotide kinase and run on a 6% denaturing acrylamide gel for analysis ( 7 ). The remaining DNA was cloned into pBluescript KS+ and sequenced. Both strands of two independent clones were sequenced for each reported sequence. The major products of these PCRs contained the region that was missing from the original chicken cDNA. Many of the PCRs also produced a minor product of the size expected for cDNAs that lack this region (see liver sample in Fig. 2 ); the predicted size for such a product is 164 bp. Attempts to reamplify and clone this fragment were unsuccessful. Figure 1 View largeDownload slide Diagram of the structure of mH2A subtypes. ++ indicates a lysine-rich region that resembles part of the C-terminal domain of histone H1, Zip indicates a region that resembles a leucine zipper, and the gray region shows the location of the region that is different between mH2A1.1 and 1.2 ( 1 , 2 ). The region C-terminal to the lysine-rich region is referred to as the non-histone region (residues 160–367 of rat mH2A1.1). Accession numbers for rat mH2A1.1 and 1.2 are M99065 and U79139, respectively. Figure 1 View largeDownload slide Diagram of the structure of mH2A subtypes. ++ indicates a lysine-rich region that resembles part of the C-terminal domain of histone H1, Zip indicates a region that resembles a leucine zipper, and the gray region shows the location of the region that is different between mH2A1.1 and 1.2 ( 1 , 2 ). The region C-terminal to the lysine-rich region is referred to as the non-histone region (residues 160–367 of rat mH2A1.1). Accession numbers for rat mH2A1.1 and 1.2 are M99065 and U79139, respectively. Antibody production and purification The non-histone region of rat mH2A1.1 (residues 160–367) was expressed in Escherichia coli (strain BL-21) as a glutathione transferase fusion protein using the expression vector pGEX-2TK ( 9 , 10 ). The fusion protein was purified using glutathione agarose beads ( 10 ). Antibodies against the fusion protein were raised in chickens. IgY was prepared from egg yolks ( 11 ) and immuno-affinity purified ( 2 ). Western blot analysis Frozen chicken liver was obtained from Pel-Freez Biologicals. Adult and embryonic chicken blood were from Hy-Vac laboratories. The nuclei were isolated and digested with micrococcal nuclease ( 12 ). An equal volume of 2× SDS sample buffer was added to the digests, they were run in SDS polyacrylamide gels and western blots were performed ( 2 ). Phylogenetic analysis of H2A protein sequences Histone H2A protein sequences were obtained from the histone sequence database ( 13 ) ( http://www.nhgri.nih.gov/DIR/GTB/HISTONES/ ). The 71 H2A sequences previously used to construct a phylogenetic tree of H2As ( 14 ) were aligned along with the H2A region of mH2A using Clustal W ( 15 ) ( http://www.dot.imgen.bcm.tmc.edu:9331/multi-align/multi-align.html ). The aligned sequences were analyzed using Joseph Felsenstein's phylogeny inference package PHYLIP ( http://www.bioweb.pasteur.fr/seqanal/interfaces/phylip-uk.html ); distance measures were calculated from the aligned protein sequences with PROTDIST using maximum likelihood estimates based on the Dayhoff PAM matrix, and the phylogenetic tree was produced from the distances matrix by NEIGHBOR using Saitou and Nei's ‘Neighbor Joining Method’. The phylogenetic tree was viewed with TreeView 1.5 ( 16 ) ( http://taxonomy.zoology.gla.ac.uk/rod/treeview.html ). Figure 2 View largeDownload slide PCR amplification of the regions of non-identity of chicken mH2A1.1 and 1.2. Primers that flank the region that is different between mH2A1.1 and 1.2 were used in PCRs that used cDNA from chicken brain (lane Br) or liver (lane Li) as a template. Reaction products were labeled with 32 P and run on a denaturing acrylamide gel. Lane Sd, end-labeled Msp I digested pBR322 DNA. Numbers on the left indicate the length of selected marker fragments. Figure 2 View largeDownload slide PCR amplification of the regions of non-identity of chicken mH2A1.1 and 1.2. Primers that flank the region that is different between mH2A1.1 and 1.2 were used in PCRs that used cDNA from chicken brain (lane Br) or liver (lane Li) as a template. Reaction products were labeled with 32 P and run on a denaturing acrylamide gel. Lane Sd, end-labeled Msp I digested pBR322 DNA. Numbers on the left indicate the length of selected marker fragments. Results Identification and sequencing of chicken mH2A cDNAs A chicken liver cDNA library was screened with a cDNA fragment from the non-histone region of rat mH2A1.1. A positive clone contained a 1720 bp insert that was completely sequenced and found to be highly homologous to rat mH2A. However, this chicken cDNA was missing the region that is different between the two known rat mH2A subtypes, mH2A1.1 and 1.2 (gray region in Fig. 1 ). To obtain these missing sequences, segments of chicken cDNAs that contain them were amplified by PCR. Two prominent products were made when cDNA from chicken liver, brain, spleen or muscle was used as a template for the PCR; the results with liver and brain cDNA are shown in Figure 2 . Sequencing of these products revealed that they contain the missing regions of mH2A1.1 and 1.2. Surprisingly, the nucleotide sequences of these regions are nearly identical to those of the rat: 96% for mH2A1.1 and 98% for mH2A1.2 ( Fig. 3 ). The nucleotide sequences of the PCR products outside this conserved region are identical to the original chicken cDNA and differ from the homologous rat sequences at 25 of 114 positions (parts of these sequences are shown in Fig. 2 ). This demonstrates that these PCR products were made from chicken cDNA, and not trace contaminants of rat cDNAs. The exceptional conservation of the nucleotide sequences in the region that is different between mH2A1.1 and 1.2 does not occur in other regions. Overall, the nucleotide sequences of the coding regions of rat and chicken mH2A1.1 are 83% identical. The non-coding regions are highly divergent. Figure 3 View largeDownload slide Nucleotide sequences of the regions that are different between mH2A1.1 and 1.2. Sequences of the products of the PCRs shown in Figure 2 are compared to the regions that are different between rat mH2A1.1 and 1.2. The complete sequences of chicken PCR#1 (accession no. AF058445) and chicken PCR#2 (accession no. AF058446) are in GenBank. The sequence labeled Chicken trunc. is from the chicken liver cDNA that is missing this region (accession no. AF058444). (|) indicates identities and (−) indicates a gap. Figure 3 View largeDownload slide Nucleotide sequences of the regions that are different between mH2A1.1 and 1.2. Sequences of the products of the PCRs shown in Figure 2 are compared to the regions that are different between rat mH2A1.1 and 1.2. The complete sequences of chicken PCR#1 (accession no. AF058445) and chicken PCR#2 (accession no. AF058446) are in GenBank. The sequence labeled Chicken trunc. is from the chicken liver cDNA that is missing this region (accession no. AF058444). (|) indicates identities and (−) indicates a gap. Figure 4 View largeDownload slide Comparison of rat and chicken mH2A subtypes. Amino acids are indicated by their one letter code. The sequences for mH2A1.1 are shown and the region of non-identity is in bold. The region of non-identity of mH2A1.2 is shown below. (|) indicates identities, (:) indicates conservative substitutions, (.) indicates less conservative substitutions and (−) indicates a gap. Figure 4 View largeDownload slide Comparison of rat and chicken mH2A subtypes. Amino acids are indicated by their one letter code. The sequences for mH2A1.1 are shown and the region of non-identity is in bold. The region of non-identity of mH2A1.2 is shown below. (|) indicates identities, (:) indicates conservative substitutions, (.) indicates less conservative substitutions and (−) indicates a gap. Comparison of chicken and rat mH2A subtypes The amino acid sequences of chicken mH2A1.1 and 1.2 were deduced from the cDNA sequences and are compared to rat mH2As in Figure 4 . Both subtypes have been remarkably conserved over their entire length. Overall, the conservation is 95% for mH2A1.1 and 96% for mH2A1.2. The H2A region (residues 1–122) is completely conserved, and thus is even more conserved than conventional H2A in these species; conventional chicken and mammalian H2As differ from one another at two to four positions depending on which subtypes are compared ( 17 ). The complete conservation of the H2A region of mH2A is consistent with its evident role in nucleosome core formation ( 1 ). The non-histone region of mH2A (residues 160–367 in rat mH2A1.1; Fig. 1 ) is also exceptionally conserved: 93% identical for mH2A1.1 and 95% for mH2A1.2. This degree of conservation approaches that of the core histones H2A and H2B [98 and 97% identical, respectively, between chickens and rats ( 17 )], and is significantly greater than most proteins. This suggests that most of the non-histone region is involved in specific interactions with nuclear components or is unusually constrained by its three-dimensional structure. Expression of chicken mH2A Western blots were used to examine the expression of mH2A proteins in chicken liver and blood. Antibodies against the non-histone region of rat mH2A1.1 detected two proteins in chicken liver nuclear extracts with electrophoretic mobilities virtually identical to rat mH2A1.1 and 1.2 ( Fig. 5 , lane 2). The relative intensity of the mH2A bands in chicken liver was lower than in rat liver. However, the level of chicken mH2A may be underestimated since this antisera was raised against rat mH2A. In addition, chicken liver contains nucleated red blood cells which contribute chromatin, but have little or no mH2A (see below). Figure 5 View largeDownload slide Western blot detection of mH2A in chicken liver and blood. Proteins were extracted from nuclei and stained using an affinity purified antibody specific for the non-histone region of rat mH2A1.1. Nuclei were prepared from: 1, rat liver; 2, chicken liver; 3, 5-day embryonic chicken blood; 4, 8-day embryonic chicken blood; 5, 14-day embryonic chicken blood; and 6, adult chicken blood. The bands below mH2A are histone H1, which crossreacted with this antiserum for unknown reasons. Loadings were adjusted to equalize amount of core histone present in each lane. Figure 5 View largeDownload slide Western blot detection of mH2A in chicken liver and blood. Proteins were extracted from nuclei and stained using an affinity purified antibody specific for the non-histone region of rat mH2A1.1. Nuclei were prepared from: 1, rat liver; 2, chicken liver; 3, 5-day embryonic chicken blood; 4, 8-day embryonic chicken blood; 5, 14-day embryonic chicken blood; and 6, adult chicken blood. The bands below mH2A are histone H1, which crossreacted with this antiserum for unknown reasons. Loadings were adjusted to equalize amount of core histone present in each lane. Neither mH2A1.1 nor 1.2 were detected in nuclear extracts from adult chicken blood ( Fig. 5 , lane 6). Nuclei from blood of 5- and 8-day embryos contained a low level of mH2A1.2, but no mH2A1.1 ( Fig. 4 , lanes 3 and 4); at these stages of development, immature red cells are abundant in blood ( 18 ). In 14-day embryos most red blood cells have reached developmental maturity ( 18 ) and neither mH2A subtype was detected in nuclei isolated from blood at this stage ( Fig. 5 , lane 5). These results indicate that mH2A is expressed in the erythrocyte lineage, but is absent or present at very low levels in mature erythrocytes. Using conventional core histones as an internal standard, we estimated that the mH2A content of mature chicken red blood cells is at least 20-fold lower than chicken liver; this estimate is based on western blots of serial dilutions of the chicken liver nuclear extracts. Comparisons of the non-histone and H2A regions of mH2A to known proteins The non-histone region of mH2A is homologous to a protein encoded by a gene found in some bacteria. One of these genes was discovered serendipitously in the bacteria Alcaligenes eutrophus ( 19 ), and we identified a homologous gene in the genomic database of E.coli . Alignment of the non-histone region of mH2A1.1 with these bacterial proteins is shown in Figure 6A . The homology extends across nearly the entire length of the bacterial proteins, though the bacterial proteins lack a region corresponding to the first 30 amino acids of the non-histone region. The non-histone region of mH2A1.1 is 34 and 30% identical to the A.eutrophus and E.coli proteins, respectively. We found that the non-histone region is also homologous to part of a protein of some positive-strand RNA viruses. Interestingly, the region of homology corresponds to a segment of ∼100 amino acids which is the most conserved region between the alphaviruses and rubella virus ( 20 , 21 ). An alignment of this region of the alphavirus sindbis virus and rubella virus to the corresponding region of mH2A1.1 is shown in Figure 6B . The non-histone region is 24 and 25% identical to this domain of sindbis and rubella viruses, respectively. These two viral sequences are 36% identical to one another and are ∼40% identical to the bacterial proteins ( Table 1 ). Thus, all of these proteins appear to be related to one another. Figure 6 View large Download slide View large Download slide Comparison of the non-histone region of mH2A to bacterial and viral proteins. ( A ) Alignment of the complete non-histone region of rat mH2A1.1, residues 160-367, to proteins from A.eutrophus (accession no. L36817), and E.coli (accession no. 1787283). ( B ) Alignment of residues 209–312 of the non-histone region of mH2A1.1 with residues 1364–1459 of the non-structural polyprotein of Sindbis virus (P03317 Swiss-Prot) and residues 833–938 of the non-structural polyprotein of Rubella virus (P13889 Swiss-Prot). Identical residues are boxed, and (−) indicates a gap. Figure 6 View large Download slide View large Download slide Comparison of the non-histone region of mH2A to bacterial and viral proteins. ( A ) Alignment of the complete non-histone region of rat mH2A1.1, residues 160-367, to proteins from A.eutrophus (accession no. L36817), and E.coli (accession no. 1787283). ( B ) Alignment of residues 209–312 of the non-histone region of mH2A1.1 with residues 1364–1459 of the non-structural polyprotein of Sindbis virus (P03317 Swiss-Prot) and residues 833–938 of the non-structural polyprotein of Rubella virus (P13889 Swiss-Prot). Identical residues are boxed, and (−) indicates a gap. Figure 7 View largeDownload slide Phylogenetic tree analysis of the H2A region of mH2A with other H2As. The H2A region of mH2A was aligned with 71 H2A sequences previously used in a phylogenetic analysis ( 14 ). A phylogenetic tree was constructed from these aligned sequences (Materials and Methods). Only selected representatives of the major groups are shown. The complete tree was essentially identical to one previously published ( 14 ). The accession numbers of the H2As shown are: Rat, A02591; Rat testis, X59962; Chicken, V00413; Xenopus , M21287; Mouse H2A.X, X58069; Strongylocentrotus purpuratus , X06642; Drosophila melanogaster , S10094; Caenorhabditis elegans , X15633; S.cerevisiae , V01304; Schizosaccharomyces pombe , X05220; Pea, JQ1183; T.thermophila , L18892; Rat macroH2A, U79139; Chicken macroH2A, AF058444; P.falciparum , M86865; Rat H2A.Z, M37584; Chicken H2A.F, V00414; T.thermophila H2A.hv1, X15548; S.pombe H2A.Pht1, S52560; Leishmania , X60054. Figure 7 View largeDownload slide Phylogenetic tree analysis of the H2A region of mH2A with other H2As. The H2A region of mH2A was aligned with 71 H2A sequences previously used in a phylogenetic analysis ( 14 ). A phylogenetic tree was constructed from these aligned sequences (Materials and Methods). Only selected representatives of the major groups are shown. The complete tree was essentially identical to one previously published ( 14 ). The accession numbers of the H2As shown are: Rat, A02591; Rat testis, X59962; Chicken, V00413; Xenopus , M21287; Mouse H2A.X, X58069; Strongylocentrotus purpuratus , X06642; Drosophila melanogaster , S10094; Caenorhabditis elegans , X15633; S.cerevisiae , V01304; Schizosaccharomyces pombe , X05220; Pea, JQ1183; T.thermophila , L18892; Rat macroH2A, U79139; Chicken macroH2A, AF058444; P.falciparum , M86865; Rat H2A.Z, M37584; Chicken H2A.F, V00414; T.thermophila H2A.hv1, X15548; S.pombe H2A.Pht1, S52560; Leishmania , X60054. The H2A region of mH2A almost certainly arose from an H2A gene. We examined the relationship of the H2A region to known H2As by constructing a phylogenetic tree ( Fig. 7 ). This analysis did not reveal a close link between the H2A region and any conventional H2A or H2A variant. The nearest link is to the H2A of the malaria parasite Plasmodium falciparum . Discussion The remarkable conservation of both mH2A1.1 and 1.2 between chickens and rats indicates that the basic function(s) of these mH2A subtypes have been conserved during the 300 million years of evolution that separate birds and mammals. Our recent studies showing that mH2A is preferentially concentrated in the inactive X chromosome of female mammals suggest that mH2A is involved in transcriptional silencing ( 3 ). Since X chromosome inactivation arose in the mammalian lineage after mammals separated from birds ( 5 ), the specific role of mH2A in X-inactivation is most likely an adaptation of a pre-existing mH2A function. One interesting possibility is that mH2A participates in gene silencing of autosomal regions in both birds and mammals, and that this function was adapted to X-inactivation in mammals. This possibility is consistent with the finding that mH2A is present in autosomes of both male and female mammals ( 3 ). The observation that mH2A is absent or at very low levels in mature chicken erythrocytes indicates that it is not involved in the transcriptional silencing that occurs in these cells. Table 1 View largeDownload slide Percentage identity of non-histone region of mH2A1.1, and bacterial and viral homologues. Table 1 View largeDownload slide Percentage identity of non-histone region of mH2A1.1, and bacterial and viral homologues. The complete conservation of the H2A region of mH2A between birds and mammals is interesting since this region is only 64% identical to a conventional rat H2A ( 1 ). This suggests that the differences between conventional H2A and the H2A region of mH2A are functionally significant, a possibility consistent with the observation that some core variants are functionally distinct from their conventional counterparts. One example is H2A.Z, an H2A that is 59% identical to conventional H2A ( 22 ). Recent studies showing that H2A.Z is essential in Drosophila ( 23 ) and Tetrahymena thermophila ( 24 ) indicate that it has important function(s) that cannot be carried out by conventional H2A. Another example is CENP-A, a 17 kDa centromere-specific protein ( 25 ) that co-purifies with mononucleosomes ( 26 ) and has a 93 amino acid domain that is 62% identical to histone H3 ( 27 , 28 ). This H3-like domain can localize to centromeres ( 28 ), showing that a variant core histone domain can be targeted to a specific chromosomal region. The non-histone region of mH2A appears to have originated from a gene that existed prior to separation of eubacteria and eukaryotes. This is indicated by the existence of eubacterial proteins that are homologous to the non-histone region. The degree of homology ( Table 1 ) is similar to the average of 37% identity observed for 57 enzymes conserved between eukaryotes and eubacteria ( 29 ), suggesting that the basic function of the non-histone region and these prokaryotic homologues is very similar. Unfortunately, the function of the bacterial homologues is not known. A potential clue to the function of the non-histone region comes from its homology to a domain found in RNA viruses ( Fig. 6 ). Although the function of this viral domain is not known, it is the most conserved sequence between sindbis virus and rubella virus ( 20 , 21 ). In sindbis virus the domain is part of a protein that associates with viral RNA and is required for the synthesis of negative-strand RNA ( 30 , 31 ). One interesting possibility in terms of mH2A function is that this domain binds RNA; it seems less likely that this viral domain binds DNA since these viruses have no DNA intermediates. There is mounting evidence that nuclear RNAs can participate in regulating chromatin structure and function, possibly through a direct interaction between RNA and chromatin ( 32 ). The best described example of this is X chromosome inactivation, which involves a nuclear RNA from the Xist gene. The preferential association of mH2A with the inactive X chromosome could involve an interaction of the non-histone region with Xist RNA. The mH2A gene appears to have formed by the linking of an H2A gene to a non-histone gene. The divergence of the H2A region from conventional H2As suggests that the mH2A gene was formed relatively early in eukaryotic evolution. A phylogenetic analysis of the H2A region of mH2A with other known H2As suggests that the H2A region branched from the H2A phylogenetic tree just prior to the branching of plants and animals. Assuming that this separation corresponds to the formation of the primordial mH2A gene, this analysis suggests that mH2A could potentially be present in many eukaryotes including yeast. However, a search of the complete genome of the yeast Saccharomyces cerevisiae failed to find any sequence that resembles mH2A. It is possible that mH2A was lost in the evolution of this yeast. Alternatively, the phylogenetic analysis could be misleading if the H2A region of mH2A had a period where it evolved more rapidly than conventional H2As. Such a period could have occurred shortly after the formation of mH2A. At that point the H2A region would probably not have been constrained like a conventional H2A, and would not have acquired specialized structures related to mH2A function. In this scenario the mH2A gene may have appeared more recently than indicated by the phylogenetic analysis. MacroH2A appears to be the only known core histone that contains a large domain derived from a non-histone gene. The H3 variant CENP-A contains a highly divergent 47 amino acid N-terminal domain. However, this domain is the same size as the corresponding N-terminal region of H3, and like the conventional N-terminus of H3, is rich in basic amino acids ( 28 ). Thus, it seems likely that this domain evolved from the N-terminus of a conventional H3. Although there may be other examples of core histones that became linked to a non-histone gene, clearly they are rare. This is not surprising given the extreme constraints imposed on the structures of the core histones. The high conservation of mH2A structure seen in this study indicates that this combination of core histone and non-histone domains has acquired valuable functions. Acknowledgements We thank Colin MacNeill and John Burch for providing chicken cDNAs and assisting with the cDNA library, Chhaya Dharia for preparation and purification of antibodies, Carl Costanzi for the phylogenetic analysis and reading the manuscript, Leslie Taylor for technical assistance, William Kaelin, Jr for providing the plasmid pGEX-27 and Mike Hall for providing embryonic chicken blood. This work was supported by NIH grant GM49351 and a grant from Merck. References 1 Pehrson J. R.,  Fried V. A.. ,  Science ,  1992, vol.  257 (pg.  1398- 1400) CrossRef Search ADS PubMed  2 Pehrson J. R.,  Costanzi C.,  Dharia C.. ,  J. Cell. Biochem. ,  1997, vol.  65 (pg.  107- 113) CrossRef Search ADS PubMed  3 Costanzi C.,  Pehrson J. R.. ,  Nature ,  1998  in press 4 McLaughlin P. J.,  Dayhoff M. O.. ,  Atlas Protein Seq. Struct. ,  1969, vol.  4 (pg.  39- 46) 5 Solari A. J.. ,  Sex Chromosomes and Sex Determination in Vertebrates ,  1994 Boca Ratan CRC Press 6 Kirchgessner T. G.,  Heinzmann C.,  Scenson K. L.,  Gordon D. A.,  Nicosia M.,  Lebherz H. G.,  Lusis A. J.,  Williams D. L.. ,  Gene ,  1987, vol.  59 (pg.  241- 251) CrossRef Search ADS PubMed  7 Sambrook J.,  Fritsch E. F.,  Maniatis T.. ,  Molecular Cloning: A Laboratory Manual ,  1989 2nd ed. Plainview, NY Cold Spring Harbor Laboratory Press 8 Sanger F.,  Nicklen S.,  Coulson A. R.. ,  Proc. Natl. Acad. Sci. USA ,  1977, vol.  74 (pg.  5463- 5467) CrossRef Search ADS   9 Kaelin W. G. J.,  Pallas D. C.,  DeCaprio J. A.,  Kaye F. J.,  Livingston D. M.. ,  Cell ,  1991, vol.  64 (pg.  521- 532) CrossRef Search ADS PubMed  10 Smith D. B.,  Johnson K. S.. ,  Gene ,  1988, vol.  67 (pg.  31- 40) CrossRef Search ADS PubMed  11 Gassmann M.,  Thömmes P.,  Weiser T.,  Hübscher U.. ,  FASEB J. ,  1990, vol.  4 (pg.  2528- 2532) PubMed  12 Pehrson J. R.. ,  Proc. Natl. Acad. Sci. USA ,  1989, vol.  86 (pg.  9149- 9153) CrossRef Search ADS   13 Baxevanis A. D.,  Landsman D.. ,  Nucleic Acids Res. ,  1997, vol.  25 (pg.  272- 272) CrossRef Search ADS PubMed  14 Thatcher T. H.,  Gorovsky M. A.. ,  Nucleic Acids Res. ,  1994, vol.  22 (pg.  174- 179) CrossRef Search ADS PubMed  15 Thompson J. D.,  Higgins D. G.,  Gibson T. J.. ,  Nucleic Acids Res. ,  1994, vol.  22 (pg.  4673- 4680) CrossRef Search ADS PubMed  16 Page R. D. M.. ,  Comput. Appl. Biosci. ,  1996, vol.  12 (pg.  357- 358) PubMed  17. Wells D.,  McBride C.. ,  Nucleic Acids Res. ,  1989, vol.  17  Suppl(pg.  311- 346) CrossRef Search ADS   18 Bruns G. A.,  Ingram V. M.. ,  Phil. Trans. R. Soc. Lond., Ser. B.: Biological. Sciences ,  1973, vol.  266 (pg.  225- 305) CrossRef Search ADS   19 Valentin H. E.,  Zwingmann G.,  Schonebaum A.,  Steinbuchel A.. ,  Eur. J. Biochem. ,  1995, vol.  227 (pg.  43- 60) CrossRef Search ADS PubMed  20 Dominguez G.,  Wang C.-Y.,  Frey T. K.. ,  Virology ,  1990, vol.  177 (pg.  225- 238) CrossRef Search ADS PubMed  21 Gorbalenya A. E.,  Koonin E. V.,  Lai M. M.-C.. ,  FEBS Lett. ,  1991, vol.  288 (pg.  201- 205) CrossRef Search ADS PubMed  22 Hatch C. L.,  Bonner W. M.. ,  Nucleic Acids Res. ,  1988, vol.  16 (pg.  1113- 1124) CrossRef Search ADS PubMed  23 van Daal A.,  Elgin S. C. R.. ,  Mol. Biol. Cell ,  1992, vol.  3 (pg.  593- 602) CrossRef Search ADS PubMed  24 Liu X.,  Li B.,  Gorovsky M. A.. ,  Mol. Cell. Biol. ,  1996, vol.  16 (pg.  4305- 4311) CrossRef Search ADS PubMed  25 Earnshaw W. C.,  Rothfield N.. ,  Chromosoma ,  1985, vol.  91 (pg.  313- 321) CrossRef Search ADS PubMed  26 Palmer D. K.,  O'Day K.,  Wener M. H.,  Andres B. S.,  Margolis R. L.. ,  J. Cell Biol. ,  1987, vol.  104 (pg.  805- 815) CrossRef Search ADS PubMed  27 Palmer D. K.,  O'Day K.,  Trong H. L.,  Charbonneau H.,  Margolis R. L.. ,  Proc. Natl. Acad. Sci. USA ,  1991, vol.  88 (pg.  3734- 3738) CrossRef Search ADS   28 Sullivan K. F.,  Hechenberger M.,  Masri K.. ,  J. Cell Biol. ,  1994, vol.  127 (pg.  581- 592) CrossRef Search ADS PubMed  29 Doolittle R. F.,  Feng D.-F.,  Tsang S.,  Cho G.,  Little E.. ,  Science ,  1996, vol.  271 (pg.  470- 477) CrossRef Search ADS PubMed  30 LaStarza M. W.,  Lemm J. A.,  Rice C. M.. ,  J. Virol. ,  1994, vol.  68 (pg.  5781- 5791) PubMed  31 Wang Y.,  Sawicki S. G.,  Sawicki D. L.. ,  J. Virol. ,  1994, vol.  68 (pg.  6466- 6475) PubMed  32 Willard H. F.,  Salz H. K.. ,  Nature ,  1997, vol.  386 (pg.  228- 229) CrossRef Search ADS PubMed  © 1998 Oxford University Press

Pehrson, John R.; Fuji, Reina N.

doi: N/Apmid: N/A

Histone macroH2A is an unusual core histone that contains a large non-histone region, and a region that resembles a full length H2A. We examined the conservation of this novel structural arrangement by cloning chicken macroH2A cDNAs and comparing them to their rat counterparts. The amino acid sequences of the two known macroH2A subtypes are >95% identical between these species despite evolutionary separation of ∼300 million years. The H2A region of macroH2A is completely conserved, and thus is even more conserved than conventional H2A in these species. The origin of the non-histone domain was examined by comparing its sequence to proteins found in bacteria and RNA viruses. These comparisons indicate that this domain is derived from a gene that originated prior to the appearance of eukaryotes, and suggest that the non-histone region has retained the basic function of its ancestral gene.

Zaphiropoulos, Peter G.

doi: 10.1093/nar/26.12.2843pmid: 9611226

Abstract RT-PCR amplification of P450 2C6 from rat liver, using primers in opposite orientations of exon 6, resulted in PCR products containing segments of exons joined at non-consensus splice sites. Moreover, many of the PCR products identified were composed of not only a single region containing exonic segments joined at non-consensus splice sites but, instead, of several repeats of the non-canonically joined region. To investigate whether these PCR products represent pre-existing molecules or are generated during the amplification process, the liver cDNA template was replaced by a plasmid containing the P450 2C6 cDNA. Surprisingly, PCR products containing repeats of non-canonically joined exonic segments were again revealed. In some cases the position of this non-canonical joining was a sequence of one or two identical nucleotides; however, there were also a number of products lacking any nucleotide identity at the position of joining. DNA nicking and/or DNA damage is thought to favour recombination during PCR, probably by misalignment of incomplete DNA strands; however, the presence of multiple repeats of the recombined region in the PCR products identified suggests a certain repetitiveness of the underlying mechanism. It is therefore proposed that these products result from a template switching event that occurs several times during a single polymerization step, following a rolling circle model of DNA synthesis. Introduction The splicing process has a unique capability of accurately removing intronic sequences from pre-mRNAs. There is a general sequence conservation at the position of intron removal with the most pronounced feature being the terminal intronic dinucleotides. Nearly all introns start with GT and end with AG, with the exception being a small number of introns that follow an AT to AC dinucleotide rule ( 1 , 2 ). Moreover, by mutagenesis analysis it has been shown that splicing can also occur between additional pairs of terminal dinucleotides GC to AG, GT to AT, AT to AA and AT to AG ( 3–5 ). Interestingly, the sequence of the terminal dinucleotides does not dictate which spliceosomal system is being used. Both the U2- and the U12-dependent splicing systems can be active on GT to AG or AT to AC introns ( 5 ). Additionally, it has been found that the order of the exons that are spliced together may not always follow the same order as in genomic DNA. Transcripts containing 3′ exons joined together with 5′ exons have been found in several cases, as well as single exons joined head-to-tail ( 6–12 ). It has been proposed that these transcripts may represent circular RNA molecules, although the possibility of some being linear trans -spliced species has not been formally excluded. In vitro , with the use of splicing extracts, the detection of exonic RNA molecules with aberrant electrophoretic patterns, indicative of their circular nature, has been accomplished ( 13 , 14 ). Most recently, ribozymes have been engineered in a way that allows production of circular exons ( 15 ). In this report evidence is presented that, during the efforts to detect RNA transcripts that might not follow the canonical rules for exon joining, RT-PCR products containing segments of exons joined at non-consensus splice sites were revealed. However, these products do not represent pre-existing RNA molecules, but are rather generated during the PCR amplification process. Moreover, the mechanism that produces these species appears to be an inherent template switching ability of the polymerase itself. Materials and Methods PCR analysis Total RNA from rat liver was isolated by the guanidinium isothiocyanate method ( 16 ) and subjected to reverse transcription using random hexamers or oligo dT primers as described before ( 12 ). PCR amplifications were performed using Perkin Elmer's model 2400 or 480 thermocyclers, for 1 min at 94°C, 1 min at the desired annealing temperature and 1 min at 72°C, in a total volume of 50 µl. The annealing temperature was identical to or at the most 3°C lower than the melting temperature of the oligos. Five units of Taq polymerase were used in a buffer containing 50 mM KCl, 10 mM Tris, pH 9.0, 0.1% Triton X-100, 2 mM MgCl 2 with 0.2 mM of each dNTP. For the nested amplifications, 1 µl from the initial amplification reaction was directly used. The PCR products were cloned after ligation with the pGEM T vector (Promega). Sequencing analysis of individual clones, that were randomly picked, was performed with dye-dideoxy nucleotides using the facilities of Cybergene AB (Huddinge, Sweden). The inserts of the clones were fully sequenced ensuring that the expected PCR primers were present at the ends. Sequencing comparisons were performed using the GCG program of the University of Wisconsin. P450 2C6 plasmid construct The P450 2C6 cDNA clone in pGEM T vector that was used in the experiments of Figure 2 was generated by PCR amplification of rat testicular cDNA with primers 2C6 1F and 2C6 9R. Sequencing established that there was a nucleotide substitution (T→C) in exon 7, at position 32 of Figure 2 . This represents a normal mistake made by DNA polymerases, and could easily have occurred during the 40 cycles of PCR amplification that were performed in order to produce detectable amounts of that cDNA. This clone was digested with restriction enzyme Pvu II, producing one fragment containing mainly insert sequences and another containing the remaining of the vector and then purified through a Wizard DNA Clean up kit (Promega), prior to the amplifications. Table 1 View largeDownload slide Primers used in the PCR amplification experiments Table 1 View largeDownload slide Primers used in the PCR amplification experiments Results Identification of RT-PCR products containing segments of P450 2C6 exons joined at non-consensus splice sites It has previously been proposed that the process of exon skipping might be related with the process of joining exons with an order different to that of genomic DNA ( 10 , 12 ). For that purpose the pattern of expression of P450 2C6 was investigated in rat liver, the tissue that is known to express the highest amounts of that P450. Using RT-PCR analysis, with an exon 1 sense and an exon 9 antisense primers ( Table 1 ), the major product revealed was the canonically spliced mRNA of nine exons. Moreover, several exon skipped mRNAs were also detected: form A had skipped exons 2, 3, 4 and 5, form B had skipped exons 6 and 7 and form C had skipped all of these six exons ( Fig. 1A ). Using sets of primers in both orientations of exon 6 ( Table 1 ), efforts were initiated in order to identify RNAs that might be composed of the skipped exons. Indeed a large number of PCR products was generated with the use of these primers ( Fig. 1B ). However, none of these products was exclusively composed of exons that were spliced at their expected splice sites. There were always exonic segments joined at positions bearing no similarity to a canonical splice site consensus. Moreover, in many cases, these products were not composed of a single non-canonically joined exonic region but of several repeats of that region (see Table 2 for a schematic representation). In one case these repeats reached the number of four. All of the cloned and sequenced RT-PCR products from the liver amplification are shown in Figure 1C . When the same exon 6 primers were used with cDNA from rat testis, a tissue that expresses much lower amounts of P450 2C6 than the liver, only a single PCR product was detected. This product contained a segment of exon 8 joined to a segment of exon 3 at a position with no similarity to a splice consensus. Moreover, PCR products containing incomplete exons joined together had previously been detected during amplification of P450 2C18 from human epidermis with primers in opposite orientation of a single exon ( 12 ). In addition, when amplification of P450 2C18 was performed using exon 1 sense and exon 9 antisense primers, PCR products containing non-canonically joined exonic segments were also identified (data not shown). PCR amplification of cloned P450 2C6 also results in products containing repeats of a recombined sequence To investigate whether the identified RT-PCR products represent pre-existing RNA molecules or were generated during the RT-PCR amplification process, a cloned P450 2C6 cDNA was used for PCR with the same sets of exon 6 sense and antisense primers, as before. Surprisingly, the cDNA plasmid allowed the production of PCR products that had the same characteristics as the ones detected during RT-PCR, namely that several repeats of a non-canonically joined exonic region were present ( Fig. 2A ). However, none of these products was identical to the ones previously detected during the RT-PCR. When low amounts of plasmid cDNA were used for this amplification, recombination at only a single position was observed; however products containing varying numbers of repeats of this recombined region were identified. When higher amounts of cDNA were used, recombination at several different positions was detected ( Fig. 2B ). These findings suggest that the more starting material, the higher the probability for such an event to occur. Moreover, this observation is also consistent with the fact that cDNA from testis (a tissue that expresses minimal amounts of 2C6) produced only a single type of recombined product, while cDNA from liver (a tissue that expresses much higher amounts of 2C6) produced a variety of products that had recombined at different positions. The finding that plasmid DNA allows the production of PCR products containing recombined sequences strongly disfavors the possibility that the non-canonically joined products, previously detected during RT-PCR, could represent pre-existing RNA molecules or might have been generated during the reverse transcription step. In addition, the absence of any consensus sequence or of extensive sequence identity at the positions of joining suggests that there may not be a critical sequence preference for these events to occur. In line with this apparent ‘random selection’ of the recombination junction is the identification of distinct PCR products in each of the amplification experiments that involved the sets of the exon 6 primers. Discussion The observation that during the PCR process, products can be generated containing segments that are not contiguous in the unamplified molecules has been noted before ( 17–19 ). This in vitro recombination activity of the polymerase has been attributed to misalignment of prematurely terminated strands with homologous regions in the other strand. The misaligned strand would thus serve as a primer for the generation of a chimeric molecule during the subsequent elongation step. Moreover, DNA nicks would favor this in vitro recombination because they would allow the production of a population of incomplete strands that would tend to anneal at regions of partial homology ( 20 ). More recently, this model of in vitro recombination during PCR has been extended by the observation that recombination at homologous regions can also take place during a single primer extension step ( 21 ). Thus, during the elongation step, the polymerase along with the extending strand has the capability to switch to another template, at a region of partial homology, therefore generating a recombined molecule. Figure 1 View large Download slide View large Download slide View large Download slide ( A ) RT-PCR analysis using primers 1F and 9R on rat liver RNA. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, RT-PCR products after 30 cycles of amplification with added liver RNA; 3, same as in lane 2 but without addition of RNA. Form B mRNA (skipped exons 6 and 7) cannot be clearly seen as a distinct band on the gel; however, it has been identified as a cloned product. ( B ) RT-PCR analysis using primers 6f and 6r and subsequently primers 6F and 6R on rat liver RNA. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, RT-PCR products after 30 cycles of initial and, subsequently, 15 cycles of nested amplification with added liver RNA; 3, same as in lane 2 but without addition of RNA. ( C ) Schematic representation of the structure of the cloned and sequenced RT-PCR products, and sequence of the non-canonically joined regions. The exons preceding and following the non-canonically joined regions (middle lines) are shown above and below, respectively. Bold letters indicate nucleotides that are identical in the junctions and small letters indicate nucleotides that are absent in the non-canonically joined molecules. Figure 1 View large Download slide View large Download slide View large Download slide ( A ) RT-PCR analysis using primers 1F and 9R on rat liver RNA. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, RT-PCR products after 30 cycles of amplification with added liver RNA; 3, same as in lane 2 but without addition of RNA. Form B mRNA (skipped exons 6 and 7) cannot be clearly seen as a distinct band on the gel; however, it has been identified as a cloned product. ( B ) RT-PCR analysis using primers 6f and 6r and subsequently primers 6F and 6R on rat liver RNA. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, RT-PCR products after 30 cycles of initial and, subsequently, 15 cycles of nested amplification with added liver RNA; 3, same as in lane 2 but without addition of RNA. ( C ) Schematic representation of the structure of the cloned and sequenced RT-PCR products, and sequence of the non-canonically joined regions. The exons preceding and following the non-canonically joined regions (middle lines) are shown above and below, respectively. Bold letters indicate nucleotides that are identical in the junctions and small letters indicate nucleotides that are absent in the non-canonically joined molecules. This type of template switching recombination, also termed copy choice recombination, represents one of the two major models proposed to explain genetic exchange, the other being the breaking and rejoining model. Although there has been overwhelming support for this latter model as the major mechanism of genetic recombination ( 22 , 23 ), there is still evidence for a role of copy choice recombination in some biological processes. For example, several polymerases including reverse transcriptase and Escherichia coli polymerases PolI and PolIII have been shown to either switch templates at regions of extensive homology or to slip between small direct repeats in vitro ( 24–28 ). In vivo , copy choice recombination events between small repeats has been proposed to occur in yeast and E.coli ( 29 , 30 ). Moreover, template switching is considered to be the prevailing mechanism of recombination in RNA viruses ( 31 ). Table 2 View largeDownload slide Schematic representation of the positions of the exon 6 primers in the P450 2C6 cDNA and general structure of the PCR products identified Table 2 View largeDownload slide Schematic representation of the positions of the exon 6 primers in the P450 2C6 cDNA and general structure of the PCR products identified The recombined PCR products that have been identified in this report contain, in some cases, one or two identical nucleotides (in fact, only A or T) at their recombination junction. However, there were also a number of products with no sequence identity at all in this position. Therefore, partial homology appears not to be a prerequisite for these events to take place, although it is conceivable that it might enhance the frequency of occurrence of such an event. What then would be a plausible mechanism that could rationalize the production of such molecules? If, for example the polymerase pauses at certain sites during one PCR cycle, then during the subsequent cycle the prematurely terminated strands might, by misalignment, serve as primers to generate recombined molecules. However, in order to account for the identification of PCR products that contained several identical repeats of the recombined sequence, one would have to assume that premature termination and subsequent misalignment would keep occurring at the same positions during a number of cycles, using longer and longer misaligned molecules as primers, and this process could even take place at positions of no sequence identity. This interpretation becomes even more unlikely, since, at the same time, recombination events at different positions would also be occurring, resulting therefore in the generation of mixed molecules, i.e., molecules containing several DNA regions that had recombined at different positions. To account for the fact that repeats of the same recombined region were always present in the multi-repeat containing molecules, a more plausible scenario might be outlined as follows: during a single elongation step, the extending polymerase may switch template or, even more simply, just slip back to another position of the original template. This slippage event could occur several times during the elongation, because the newly synthesized strand, being complementary to the original template, would stabilize the loop structure that had been generated on the slipped template ( Fig. 3 ). Therefore, following a rolling circle model of DNA synthesis, several identical repeats of the recombined region could be produced. In fact, this rolling circle model of template switching provides a reasonable interpretation that is consistent with the finding of the presence of the same type of repeat and not of combinations of different types of repeats in the multi-repeat containing PCR products. Figure 2 View large Download slide View large Download slide ( A ) PCR analysis using primers 6f and 6r and subsequently primers 6F and 6R on a cloned 2C6 cDNA plasmid. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, PCR products after 30 cycles of initial and, subsequently, 20 cycles of nested amplification with 10 −6 × 5 ng of added 2C6 plasmid DNA; 3, same as in lane 2 but with 10 −4 × 5 ng of added 2C6 plasmid DNA; 4, same as in lane 2 but with 10 −2 × 5 ng of added 2C6 plasmid DNA; 5, same as in lane 2 but with 5 ng of added 2C6 plasmid DNA; 6, same as in lane 2 but with no addition of 2C6 plasmid DNA. ( B ) Schematic representation of the structure of the cloned and sequenced PCR products from lane 3 (labeled with ′) and lane 4 (labeled with *), and sequence of the non-canonically joined regions (same legend as in Fig. 1C ). Figure 2 View large Download slide View large Download slide ( A ) PCR analysis using primers 6f and 6r and subsequently primers 6F and 6R on a cloned 2C6 cDNA plasmid. The individual lanes on the agarose gel electrophoresis are as follows: 1, molecular weight markers; 2, PCR products after 30 cycles of initial and, subsequently, 20 cycles of nested amplification with 10 −6 × 5 ng of added 2C6 plasmid DNA; 3, same as in lane 2 but with 10 −4 × 5 ng of added 2C6 plasmid DNA; 4, same as in lane 2 but with 10 −2 × 5 ng of added 2C6 plasmid DNA; 5, same as in lane 2 but with 5 ng of added 2C6 plasmid DNA; 6, same as in lane 2 but with no addition of 2C6 plasmid DNA. ( B ) Schematic representation of the structure of the cloned and sequenced PCR products from lane 3 (labeled with ′) and lane 4 (labeled with *), and sequence of the non-canonically joined regions (same legend as in Fig. 1C ). Figure 3 View largeDownload slide Proposed mechanism for the generation of PCR products containing several identical repeats of a recombined region. The polymerase initiates DNA synthesis at a specific priming sequence of the original template ( A ) and ( B ). At some nucleotide position during the elongation process the polymerase with the extending strand slips to a downstream nucleotide position of the same template and keeps the elongation process ( B ) and ( C ). The loop structure of the template molecule, that has been generated by the slippage event, is maintained by the complementarity with the newly synthesized strand (C). This growing strand displaces the previously synthesized strand during several rounds of DNA synthesis ( D ) resulting in a molecule, containing a large number of identical repeats ( E ). During the subsequent PCR amplification steps, there will be a tendency to amplify molecules having a single recombined region, since a number of internal priming sites will be present in the multi-repeat containing molecules (E). However if the rolling circle synthesized DNA contains a large number of repeats, PCR products with more than one repeat could easily be identified. Figure 3 View largeDownload slide Proposed mechanism for the generation of PCR products containing several identical repeats of a recombined region. The polymerase initiates DNA synthesis at a specific priming sequence of the original template ( A ) and ( B ). At some nucleotide position during the elongation process the polymerase with the extending strand slips to a downstream nucleotide position of the same template and keeps the elongation process ( B ) and ( C ). The loop structure of the template molecule, that has been generated by the slippage event, is maintained by the complementarity with the newly synthesized strand (C). This growing strand displaces the previously synthesized strand during several rounds of DNA synthesis ( D ) resulting in a molecule, containing a large number of identical repeats ( E ). During the subsequent PCR amplification steps, there will be a tendency to amplify molecules having a single recombined region, since a number of internal priming sites will be present in the multi-repeat containing molecules (E). However if the rolling circle synthesized DNA contains a large number of repeats, PCR products with more than one repeat could easily be identified. Non-homologous recombination events, in a cell-free system, have also been observed with replicable RNA molecules ( 28 ). However, the mechanism proposed to rationalize these findings was suggested to be a splicing-like transesterification process that is guided by the secondary structure of the RNA molecule and not that of a template switching activity of the replicase enzyme. On the other hand, since these experiments involved amplification by replicase in the order of 10 12 , followed by RT and then a two-step PCR, it is still conceivable that some of the recombined molecules detected may represent products of copy choice events. In summary, evidence has been presented in this report that, in vitro , Taq polymerase has the capability to promote recombinational events that appear to follow a copy choice mechanism and, moreover, can take place at positions of no detectable homology. Given the accumulated evidence that template switching is a recombination process that may occur with a number of different polymerases, at homologous regions, it could be speculated that polymerases, in general, might have an inherent capability to slip, at a low frequency, to a different template or to a different position in the same template, even at non-homologous regions. Although the exponential increase of newly synthesized DNA during the PCR process facilitated the detection of such recombined molecules, the frequency of occurrence of these copy choice events is apparently quite low. If such phenomena do indeed occur in vivo , considerably more effort would be required to unambiguously identify them. Acknowledgements This study was supported by grants from the Swedish Natural Science Research Council, the foundations Lars Hiertas Minne and Åke Wiberg and the Karolinska Institute. References 1 Hall S.L.,  Padgett R.A.. ,  Science ,  1996, vol.  271 (pg.  1716- 1718) CrossRef Search ADS PubMed  2 Tarn W.-Y.,  Steitz J.A.. ,  Cell ,  1996, vol.  84 (pg.  801- 811) CrossRef Search ADS PubMed  3 Parker R.,  Siliciano P.G.. ,  Nature ,  1993, vol.  361 (pg.  660- 662) CrossRef Search ADS PubMed  4 Scadden A.D.J.,  Smith C.W.J.. ,  EMBO J. ,  1995, vol.  14 (pg.  3236- 3246) PubMed  5 Dietrich R.C.,  Incorvaia R.,  Padgett R.A.. ,  Mol. Cell ,  1997, vol.  1 (pg.  151- 160) CrossRef Search ADS PubMed  6 Nigro J.M.,  Cho K.R.,  Fearon E.R.,  Kern S.E.,  Ruppert J.M.,  Oliner J.D.,  Kinzler K.W.,  Vogelstein B.. ,  Cell ,  1991, vol.  64 (pg.  607- 613) CrossRef Search ADS PubMed  7 Cocquerelle C.,  Daubersies P.,  Majérus M.-A.,  Kerckaert J.-P.,  Bailleul B.. ,  EMBO J. ,  1992, vol.  11 (pg.  1095- 1098) PubMed  8 Cocquerelle C.,  Mascrez B.,  Hétuin D.,  Bailleul B.. ,  FASEB J. ,  1993, vol.  7 (pg.  155- 160) PubMed  9 Capel B.,  Swain A.,  Nicolis S.,  Hacker A.,  Walter M.,  Koopman P.,  Goodfellow P.,  Lovell-Badge R.. ,  Cell ,  1993, vol.  73 (pg.  1019- 1030) CrossRef Search ADS PubMed  10 Zaphiropoulos P.G.. ,  Proc. Natl. Acad. Sci. USA ,  1996, vol.  93 (pg.  6536- 6541) CrossRef Search ADS   11 Bailleul B.. ,  Nucleic Acids Res. ,  1996, vol.  24 (pg.  1015- 1019) CrossRef Search ADS PubMed  12 Zaphiropoulos P.G.. ,  Mol. Cell. Biol. ,  1997, vol.  17 (pg.  2985- 2993) CrossRef Search ADS PubMed  13 Braun S.,  Domdey H.,  Wiebauer K.. ,  Nucleic Acids Res. ,  1996, vol.  24 (pg.  4152- 4157) CrossRef Search ADS PubMed  14 Pasman Z.,  Been M.D.,  Garcia-Blanco M.A.. ,  RNA ,  1996, vol.  2 (pg.  603- 610) PubMed  15 Mikheeva S.,  Hakim-Zargar M.,  Carlson D.,  Jarrell K.. ,  Nucleic Acids Res. ,  1997, vol.  25 (pg.  5085- 5094) CrossRef Search ADS PubMed  16 Chomczynski P.,  Sacchi N.. ,  Anal. Biochem. ,  1987, vol.  162 (pg.  156- 159) CrossRef Search ADS PubMed  17 Pääbo S.,  Irwin D.M.,  Wilson A.. ,  J. Biol. Chem. ,  1990, vol.  265 (pg.  4718- 4721) PubMed  18 Jansen R.,  Ledley F.D.. ,  Nucleic Acids Res. ,  1990, vol.  18 (pg.  5153- 5156) CrossRef Search ADS PubMed  19 Meyerhans A.,  Vartanian J.-P.,  Wain-Hobson S.. ,  Nucleic Acids Res. ,  1990, vol.  18 (pg.  1687- 1691) CrossRef Search ADS PubMed  20 Marton A.,  Delbecchi L.,  Bourgaux P.. ,  Nucleic Acids Res. ,  1991, vol.  19 (pg.  2423- 2426) CrossRef Search ADS PubMed  21 Odelberg S.J.,  Weiss R.B.,  Hata A.,  White R.. ,  Nucleic Acids Res. ,  1995, vol.  23 (pg.  2049- 2057) CrossRef Search ADS PubMed  22 Meselson M.,  Weigle J.J.. ,  Proc. Natl. Acad. Sci. USA ,  1961, vol.  47 (pg.  857- 868) CrossRef Search ADS   23 Tease C.. ,  Nature ,  1978, vol.  272 (pg.  823- 824) CrossRef Search ADS PubMed  24 Papanicolaou C.,  Ripley L.S.. ,  J. Mol. Biol. ,  1989, vol.  207 (pg.  335- 353) CrossRef Search ADS PubMed  25 Madsen C.S.,  Ghivizzani S.C.,  Hauswirth W.W.. ,  Proc. Natl. Acad. Sci. USA ,  1993, vol.  90 (pg.  7671- 7675) CrossRef Search ADS   26 Negroni M.,  Ricchetti M.,  Nouvel P.,  Buc H.. ,  Proc. Natl. Acad. Sci. USA ,  1995, vol.  92 (pg.  6971- 6975) CrossRef Search ADS   27 Canceill D.,  Ehrlich S.D.. ,  Proc. Natl. Acad. Sci. USA ,  1996, vol.  93 (pg.  6647- 6652) CrossRef Search ADS   28 Chetverin A.B.,  Chetverina H.V.,  Demidenko A.A.,  Ugarov V.I.. ,  Cell ,  1997, vol.  88 (pg.  503- 513) CrossRef Search ADS PubMed  29 Gordenin D.A.,  Malkova A.L.,  Peterzen A.,  Kulikov V.N.,  Pavlov Y.I.,  Perkins E.,  Resnick M.A.. ,  Proc. Natl. Acad. Sci. USA ,  1992, vol.  89 (pg.  3785- 3789) CrossRef Search ADS   30 D'Alencon E.,  Petranovic M.,  Michel B.,  Noirot P.,  Aucouturier A.,  Uzest M.,  Ehrlich S.D.. ,  EMBO J. ,  1994, vol.  13 (pg.  2725- 2734) PubMed  31 Lai M.M.C. ,  Microbiol. Rev. ,  1992, vol.  56 (pg.  61- 79) PubMed  © 1998 Oxford University Press

Zaphiropoulos, Peter G.

doi: N/Apmid: N/A

RT-PCR amplification of P450 2C6 from rat liver, using primers in opposite orientations of exon 6, resulted in PCR products containing segments of exons joined at non-consensus splice sites. Moreover, many of the PCR products identified were composed of not only a single region containing exonic segments joined at non-consensus splice sites but, instead, of several repeats of the non-canonically joined region. To investigate whether these PCR products represent pre-existing molecules or are generated during the amplification process, the liver cDNA template was replaced by a plasmid containing the P450 2C6 cDNA. Surprisingly, PCR products containing repeats of non-canonically joined exonic segments were again revealed. In some cases the position of this non-canonical joining was a sequence of one or two identical nucleotides; however, there were also a number of products lacking any nucleotide identity at the position of joining. DNA nicking and/or DNA damage is thought to favour recombination during PCR, probably by misalignment of incomplete DNA strands; however, the presence of multiple repeats of the recombined region in the PCR products identified suggests a certain repetitiveness of the underlying mechanism. It is therefore proposed that these products result from a template switching event that occurs several times during a single polymerization step, following a rolling circle model of DNA synthesis.

Graubert, Timothy A.;Hug, Bruce A.;Wesselschmidt, Robin;Hsieh, Chih-Lin;Ryan, Thomas M.;Townes, Tim M.;Ley, Timothy J.

doi: 10.1093/nar/26.12.2849pmid: 9611227

Abstract The random insertion of transgenes into the genomic DNA of mice usually leads to widely variable levels of expression in individual founder lines. To study the mechanisms that cause variegation, we designed a transgene that we expected to variegate, which consisted of a β-globin locus control region 5′ HS-2 linked in tandem to a tagged human β-globin gene (into which a Lac-Z cassette had been inserted). All tested founder lines exhibited red blood cell-specific expression, but levels of expression varied >1000-fold from the lowest to the highest expressing line. Most of the variation in levels of expression appeared to reflect differences in the percentage of cells in the peripheral blood that expressed the transgene, which ranged from 0.3% in the lowest expressing line to 88% in the highest; the level of transgene expression per cell varied no more than 10-fold from the lowest to the highest expressing line. These differences in expression levels could not be explained by the location of transgene integration, by an effect of β-galactosidase on red blood cell survival, by the half life of the β-galactosidase enzyme or by the age of the animals. The progeny of all early erythroid progenitors (BFU-E colony-forming cells) exhibited the same propensity to variegate in methylcellulose-based cultures, suggesting that the decision to variegate occurs after the BFU-E stage of erythroid differentiation. Collectively, these data suggest that variegation in levels of transgene expression are due to local, integration site-dependent phenomena that alter the probability that a transgene will be expressed in an appropriate cell; however, these local effects have a minimal impact on the transgene's activity in the cells that initiate transcription. Introduction When fragments of DNA containing one of the human β-like globin genes are randomly integrated into the genome of transgenic mice, the result is generally very low level expression of the transgene in a small percentage of transgenic lines ( 1–6 ); however, these transgenes are appropriately expressed at the correct developmental stage and specifically in red blood cells, suggesting that the tissue and development-specific information required for targeted expression lies within or near the genes themselves ( 1–6 ). The very low levels of expression seen in a small fraction of the animals prompted a search for DNA elements that were required for more proper, high levels of expression in a higher percentage of the animals. The first clue that these elements might exist came from the work of Tuan et al. ( 7 , 8 ) and Groudine and colleagues ( 9 ), who showed that erythroid-specific DNase I hypersensitive sites lie upstream from the globin gene cluster, and may contribute to organization of the erythroid domain of the cluster. The finding of an Hispanic patient with γδβ-thalassemia who carried a deletion spanning a region 6–30 kb upstream from the ε-globin gene suggested that this region might be critical for activity of the β-globin locus ( 10 ). Grosveld and colleagues ( 11 ) indeed showed that linkage of this upstream region to globin genes in transgenic mice greatly increased the percentage of mice that expressed the transgene, and also increased the level of output of the linked globin genes at each integration site, so that transgenes demonstrated ‘integration site-independent, copy number-dependent’ expression. However, more recent studies have suggested that linkage of single hypersensitive sites to globin genes does not eliminate integration site-specific variation ( 12–19 ). Recent studies by Whitelaw and Martin and colleagues ( 20–25 ) have suggested mechanisms by which transgenic variegation may occur. In their studies, globin regulatory sequences were linked with a Lac-Z reporter gene, and transgenic animals were produced. Collectively, these studies have shown that the variation in output seen among different integration sites can largely be explained by different percentages of expressing red blood cells in the transgenic animals ( 20 ). Their results also suggested that addition of a locus control region element to a globin transgene seems to increase expression by increasing the probability that a high percentage of red blood cells will express the transgene, not by increasing the level of expression per cell ( 21 , 22 ). These studies complement work that has been performed in tissue culture cells, where we and others have observed that LCR elements seem to act by increasing the probability that an individual integration event will be transcriptionally productive, not by increasing the output of the promoter linked to the LCR element ( 21 , 26–28 ). In this report, we linked human 5′ HS-2 to a human β-globin transgene that contained a Lac-Z cassette within the β-globin gene. Previous reports suggested that expression of this transgene would be variegated ( 12 , 16 , 19 ). We created 10 transgenic founder lines with this transgene, and indeed found this to be the case. The basis for this wide variation in transgene expression was found to be primarily due to alterations in the percentages of transgene-expressing cells within each founder line; however, small differences in the output of the transgenes were also noted. We observed that the ‘decision’ to variegate must be made after the BFU-E stage of erythroid differentiation, suggesting that the regulatory elements in the transgene must interact with stage-specific information in the developing erythroid cell, and that individual integration sites may have differing abilities to respond to that information. Materials and Methods Production of transgenic mice The transgene was assembled in pUC19, and contains the 1.9 kb Kpn I- Pvu II human 5′ HS-2 fragment (previously shown to contain all known HS-2 activity; 16 ), upstream from the marked β-globin transgene ( Fig. 1 ). The 5′ part of the human β-globin gene (an Hpa I- Nco I fragment extending from −800 to +46) was fused to a 3.0 kb Nco I- Bgl II fragment obtained from pLacD. The 3′ part of the β-globin gene consisted of a 2.8 kb Bam HI- Xba I fragment that contains the 3′ end of exon 2, intron 2, exon 3 and 3′ flanking sequence. 5′ HS-2 was inserted in the genomic (5′→3′) orientation with respect to the transgene. The transgene was isolated from the plasmid vector backbone by cleavage with Xho I and Sal I, gel purified, concentrated on a DEAE mini column (Elutip, Schleicher and Schuell), and resuspended in injection buffer (5 mM NaCl, 5 mM Tris pH 7.5, 0.25 mM EDTA). Pronuclear injections of fertilized eggs derived from inbred C57Bl/6 mice, or from B6XC3H mice, were performed. Lines A, B and C were made in the pure Bl/6 background, and lines D-J were made in the hybrid background. Founders were identified by Southern blotting of tail DNA. All 10 founder mice passed the transgene through the germline. Only mice that were F1 or beyond were used for analysis in subsequent studies. Fluorescence in situ hybridization (FISH) The entire plasmid containing the β-globin Lac-Z transgene (pTR159) was nick-translated with Bio-11-dUTP (Sigma) for in situ hybridization. Procedures for hybridization, washing, blocking, detection and amplification were described previously ( 29 , 30 ) with minor changes. After denaturation for 15 min at 37°C, the biotinylated probe (at 15 µg/ml) was mixed in 0.1 mg/ml salmon sperm DNA, 50% formamide, 2× SSC and 10% dextran sulfate, and hybridized to chromosome spreads for 16 h at 37°C. After hybridization, the slides were washed three times in 4× SSC, 0.1% Tween 20 for 5 min each time at 45°C, followed by one wash in 0.1× SSC, 0.1% Tween 20 for 5 min at 45°C. Sites of hybridization were detected with avidin conjugated with FITC (Molecular Probes, Eugene, OR) followed by one round of amplification. Metaphase spreads were counterstained with 100 ng/ml propidium iodide, mounted with 90% glycerol and 2.3% DABCO, then scanned, analyzed and photographed on an Olympus BX-60 fluorescent microscope. Preparation of metaphase spreads Skin fibroblasts from the transgenic and non-transgenic mice were cultured to obtain metaphase preparations. Fibroblasts were cultured in Dulbecco's Modified Eagle Medium with 10% fetal calf serum and penicillin-streptomycin. Cells were harvested for metaphase preparation by standard methods. Preparation of peripheral blood or whole animals for determination of β-galactosidase activity Peripheral blood was obtained from mice between 1 and 24 months of age by retro-orbital bleeding. For X-gal staining, 2 µl of blood was fixed in 1 ml of 2% formaldehyde/0.2% glutaraldehyde in phosphate buffered saline (PBS) at 4°C for 10 min, washed twice with PBS and then stained in 100 µl of a mixture containing 1 mg/ml X-gal, 4 mM potassium ferricyanide, 4 mM potassium ferrocyanide and 2 mM MgCl 2 at room temperature for 2–4 h. The reaction was stopped by adding 1 ml of PBS and cooling to 4°C; 100–200 µl of cells were then prepared for examination using a cytospin apparatus. For quantitative determination of β-galactosidase activity, 3 µl of peripheral blood was lysed in 300 µl of extraction buffer (50 mM Tris pH 8.0, 1 mM DTT and 1% Triton X-100). Extracts were vigorously vortexed, and then 50 µl of lysate was used to normalize for hemoglobin content by measuring the OD 405 . Thirty µl of normalized lysate was mixed with 270 µl of ONPG cocktail, exactly as described ( 43 ). Reactions were incubated at room temperature for 4–24 h. In each assay, samples were performed in duplicate, and 0–200 defined units of Escherichia coli β-galactosidase (Boehringer) were assayed to ensure that all samples were measured in the linear range of the assay (OD 420 measurements of 0.0–1.0). Reactions were stopped by adding 500 µl of dH 2 0 and immediately measured on a spectrophotometer at OD 420 . For determination of β-galactosidase expression in whole organs, mice were anesthetized and the left ventricle was cannulated. The animals were perfused exhaustively (≥5 min, until distinct organ blanching was observed), and then organs were harvested and frozen at −70°C. The organs were thawed in extraction buffer at 4°C, disrupted with a tissue grinder, and then sonicated. Cleared supernatants were evaluated for total protein content using the BioRad assay. Equal amounts of total protein were then evaluated using the ONPG assay; substrate conversion for the highest expressing tissue (peripheral blood) was always performed in the linear range of the assay. All assays were performed in duplicate. Flow cytometry Reticulocytes were enumerated on a FACScan (Becton Dickinson) after staining with acridine orange using standard techniques ( 31 ). Reticulocytosis was induced with a single subcutaneous injection of phenylhydrazine (Sigma, St Louis, MO) (60 mg/kg), as previously described ( 32 ). Reticulocytopenia was induced by a single exposure of the mice to 750 cGy from a 137 Cs source. Flow cytometric examination for β-galactosidase activity was performed using the FACS-FDG staining protocol developed by Fiering et al. and Nolan et al. ( 33 , 34 ). Basically, 2 µl of peripheral blood was mixed with 1 ml of FACS buffer (PBS containing 10 mM HEPES pH 7.0, 4% fetal calf serum and 300 µM chloroquine). One hundred µl of cells in this mixture were warmed to 37°C and then osmotically loaded by adding 100 µl of fluorescein di-β-d-galactopyranoside (FDG) (Molecular Probes, Eugene, OR) diluted 1:10 in dH 2 O, with vigorous mixing. After 30 s of osmotic shock at 37°C, 1.2 ml of FACS buffer at 4°C was added, and the reaction was allowed to proceed for exactly 5 min on ice. The reaction was stopped by adding 30 µl of 1 mM phenylethyl β-d-thiogalactopyranoside to the reaction. Flow cytometric analysis of the gated red cell population was then performed immediately. The duration of substrate loading and enzymatic conversion were carefully controlled to ensure that assays were performed in the linear range. Cell concentration was kept low (<10 6 cells/ml) in order to minimize transfer of fluorescein from Lac-Z + cells to Lac-Z − cells. One hundred thousand events were collected per sample in order to increase the statistical significance of positive events detected in the low expressing lines. Methylcellulose culture of bone marrow progenitor cells Bone marrow cells were flushed from mouse femurs. After washing once with PBS, 5 × 10 4 cells were plated in M3230 methylcellulose (Stem Cell Technology, Vancouver, Canada) supplemented with erythropoietin (3 U/ml), c-kit ligand (2.5 ng/ml) and IL-3 (2.5 ng/ml). After 3 days (for CFU-E) or 7 days (for BFU-E) of culture at 37°C, colonies were counted and then overlayed with X-gal reagents. Colonies were scored and photographed after an additional 24 h incubation at 37°C. Results The β-Lac-Z transgene and production of founder animals To study the molecular basis of variegation, we wished to create a transgene that would be highly variegated, but expressed in most integration sites, and that would allow us to measure transgene output in individual cells. We therefore assembled the transgene shown in Figure 1 . A previously characterized human β-globin LCR 5′ HS-2 fragment was linked upstream in tandem to a human β-globin transgene that contained a Lac-Z cassette. The sequences between the initiation codon and the Bam HI site of the β-globin gene were removed, and the Lac-Z cassette was inserted in this site. This transgene contains the known β-globin regulatory sequences in the 5′ flanking region, in IVS-2 and 3′ to the gene. Previous studies suggested that the output of this transgene should vary widely from one integration site to the next ( 12 , 16 , 19 ). Ten founder lines were created with this transgene. Lines A, B and C were made in inbred C57Bl/6 mice, and lines D-J were made in a hybrid C3H × C57Bl/6 background. Southern blot analysis revealed that the copy numbers of the transgene per haploid genome ranged from 4 to 167; all of the transgenes were arranged as head-to-tail concatamers in the genome, and all had non-rearranged transcription units by Southern analysis (data not shown). Figure 1 View largeDownload slide FISH analysis of the integrated transgene. A diagram of the HS-2/β-globin-Lac-Z transgene is shown at the bottom of this figure. The entire DNA fragment was used as a probe for FISH of cells derived from skin fibroblasts of each of the 10 founder lines (A–J). The location of each transgene integration site is shown with an arrowhead. An interpretation of the sites of integration is provided in Table 1 . Figure 1 View largeDownload slide FISH analysis of the integrated transgene. A diagram of the HS-2/β-globin-Lac-Z transgene is shown at the bottom of this figure. The entire DNA fragment was used as a probe for FISH of cells derived from skin fibroblasts of each of the 10 founder lines (A–J). The location of each transgene integration site is shown with an arrowhead. An interpretation of the sites of integration is provided in Table 1 . To define the chromosomal integration sites of the transgene in all 10 founder lines, the entire transgene was used as a probe for FISH analysis. The transgene did not hybridize with DNA sequences in non-transgenic mice (data not shown). Hybridization signals were detected in all 10 founder lines, and the intensity of these signals correlated with Southern blot analysis ( Fig. 1 ). Each integration site was unique. Only one of the 10 founder lines exhibited transgene insertion in the centromere; this line demonstrated very low expression of the transgene ( Table 1 ). Five other transgenic lines exhibited low level expression, but none of these had centromeric transgene insertion sites. Table 1 View largeDownload slide Characteristics of transgenic lines Table 1 View largeDownload slide Characteristics of transgenic lines Transgene expression was highly variable and depended on the integration site Peripheral blood was obtained from young adult animals (ranging from 2 to 4 months of age) from all 10 founder lines and wild type controls, and lysates from red blood cells were prepared and normalized for hemoglobin content. The total β-galactosidase activity in these lysates was then measured using a quantitative ONPG assay. Different quantities of lysates were assayed to be certain that all measurements were performed in the linear range of the assay. Mice derived from all 10 founder lines were assayed simultaneously on at least three separate occasions. The results from one representative analysis are shown in Table 1 (and Fig. 5A , in graphic form). Two of the founder lines expressed very high levels of β-galactosidase activity in the peripheral blood (lines A and G), two lines contained intermediate levels of activity (lines C and E), and six lines contained very low or undetectable levels of β-galactosidase activity (lines B, D, F, H, I and J). The total variation in expression of β-galactosidase was >1000-fold using the quantitative assay. Transgene expression did not correlate positively with transgene copy number ( Table 1 ). Because the variation in transgene expression was extreme among the 10 founder lines, we wanted to be certain that this β-globin transgene appropriately targeted the adult red blood cell compartment. Therefore, young adult mice from the high expressing and intermediate expressing transgenic lines were examined for β-galactosidase activity in all organs. Peripheral blood lysates were prepared from adult animals ranging from 8 to 12 months of age, and then the animals were extensively perfused with PBS to reduce the amount of blood present in all peripheral organs. Lysates of all organs were made and normalized for total protein content. β-galactosidase activity was then determined in each organ using the ONPG assay. All assays were performed in linear range. For each founder line, the level of β-galactosidase in the peripheral blood lysate was defined as 100%. The normalized values of the perfused organ lysates are shown in Figure 2 . In all of these founder lines, the highest level of activity was found in peripheral blood, followed by the spleen and bone marrow (which both contain large numbers of erythroid precursors). Smaller amounts of β-galactosidase activity (<5% of peripheral blood levels, in most cases) were detected in all other organs. The low level of β-galactosidase activity in non-erythroid tissues is most likely accounted for by residual RBC contamination following organ perfusion. Within the bone marrow itself, we detected Lac-Z staining only in cells that had the morphology of erythroid precursors and mature erythroid cells (data not shown). These results show that the transgene was appropriately targeted to the erythroid compartment of adult animals, and that the mutation introduced into the β-globin transgene did not alter its targeting properties. Variations in transgene expression are predominantly due to variations in the percentages of expressing red blood cells To determine whether the wide variation in expression of the transgene was due to differences in the percentages of red blood cells that expressed the transgene or due to differences in output of the transgene in all of the red blood cells, we obtained peripheral blood from young adult animals from all 10 founder lines simultaneously. We either fixed these cells and stained them with X-gal reagents, or introduced the biochemical substrate FDG into red cells by osmotic shock and measured β-galactosidase activity using a flow cytometric method. The two independent techniques yielded results that were virtually identical. The results of the X-gal staining of peripheral blood are shown in Figure 3 , and the results of flow cytometric analyses in Figures 4 and 5B . The two founder lines with the largest amounts of β-galactosidase activity in peripheral blood lysates (lines A and G), demonstrated very high percentages of red blood cells containing detectable amounts of enzymatic activity (78 and 88%, respectively). The two intermediate expressing lines (C and E) demonstrated intermediate percentages of FDG + red blood cells (14 and 12%, respectively). The six lines with very low levels of enzyme activity in peripheral blood demonstrated very small percentages of expressing red blood cells in the periphery (0.3–1.8%; Fig. 4 ). The plots shown in Figure 4 present β-galactosidase activity on the x -axis and an irrelevant parameter on the y -axis, to make it possible to see the mean level of fluorescence per cell in the lines that have only a very small percentage of FDG + cells. The percentage of positive red cells in each founder line, and the mean fluorescence intensity of the expressing cells, are summarized in Figure 5B . From these data, it is evident that the mean fluorescence intensity per positive cell (a function of the number of β-galactosidase molecules per cell) ( 34 ) varies only 10-fold between the lowest expressing line and the highest line. The percentage of positive cells varies by >250-fold, suggesting that most of the variegation can be best explained by differences in the percentage of cells that express the transgene in the peripheral blood. We also detected a correlation between the percentage of positive cells and the mean fluorescence intensity of the positive cells ( R = 0.8). Figure 2 View largeDownload slide β-galactosidase expression in organ lysates from founder lines. Individual mice from founder lines A, C, E and G were obtained, and the anesthetized mice were exhaustively perfused with PBS using standard techniques. The organs were then harvested and lysates were made and normalized for protein content. Equal amounts of protein were evaluated for total β-galactosidase content with the ONPG assay. The amount of β-galactosidase activity was defined as 100% in the peripheral blood lysate. The levels of detectable β-galactosidase in the other organs is represented as a percentage of that value. Note that the highest level of expression in every animal is in the peripheral blood, followed by the bone marrow and spleen. Non-hematopoietic organs invariably contain small or nearly undetectable amounts of β-galactosidase activity, indicating that the transgene is correctly targeted to hematopoietic tissues in all founder lines tested. Figure 2 View largeDownload slide β-galactosidase expression in organ lysates from founder lines. Individual mice from founder lines A, C, E and G were obtained, and the anesthetized mice were exhaustively perfused with PBS using standard techniques. The organs were then harvested and lysates were made and normalized for protein content. Equal amounts of protein were evaluated for total β-galactosidase content with the ONPG assay. The amount of β-galactosidase activity was defined as 100% in the peripheral blood lysate. The levels of detectable β-galactosidase in the other organs is represented as a percentage of that value. Note that the highest level of expression in every animal is in the peripheral blood, followed by the bone marrow and spleen. Non-hematopoietic organs invariably contain small or nearly undetectable amounts of β-galactosidase activity, indicating that the transgene is correctly targeted to hematopoietic tissues in all founder lines tested. The β-galactosidase enzyme has a long half life in adult red blood cells, and has no adverse effects on red cell survival To be certain that our measured transgene activity was not affected by alterations in red cell sub-populations, or by altered red blood cell half life due to transgene expression, we performed complete blood counts on at least two independent animals from each founder line, and determined the reticulocyte counts from at least two independent animals from each founder line. The hemoglobin content in the peripheral blood from all examined animals was within normal limits, and the reticulocyte counts, as determined both by manual counting, and by flow cytometry, were <3% for all examined animals (data not shown). To determine whether the detected β-galactosidase was present at higher levels in very young red cells (reticulocytes), we induced hemolytic anemia in several low to intermediate expressing animals by treating the animals with phenylhydrazine, which induced reticulocyte counts of 50–80% 3 days after administration of the drug (data not shown). The percentage of β-galactosidase cells in the peripheral blood, and the Lac-Z content per red cell were unaltered by phenylhydrazine-induced hemolysis (data not shown). In addition, we irradiated (750 cGy) several high expressing and intermediate expressing animals, and then measured the reticulocyte counts and percentages of positive cells 1 week later. In all cases, no reticulocytes were detected in the irradiated animals. However, the percentage of positive cells was unchanged from the pre-irradiation values (data not shown). In sum, these data suggest that the Lac-Z protein has a very long half life in red blood cells (perhaps as long as the red cell itself), and that high level expression of β-galactosidase in red blood cells does not alter their half life in vivo . Minimal changes in transgene expression as a function of time Whitelaw and colleagues have previously reported that globin transgenes can exhibit age-dependent silencing ( 35 ). We selected three lines that permitted accurate measurements of β-galactosidase activity in peripheral blood, and followed mice for up to 2 years. Levels of transgene expression and the frequency of positive cells were similar from animal to animal, and revealed little or no reduction in levels of expression as a function of time (Fig. 6). Many of the other animals and additional founder lines were followed for >1 year, and similar levels of transgene expression were noted in all animals at all time points measured (data not shown). In particular, 11 animals were studied for up to 1 year for line B, which demonstrated centromeric insertion of the transgene. All of these animals had similar patterns of expression at all time points tested (data not shown). Although these data differ from the findings of Whitelaw and colleagues ( 35 ), there are a number of technical differences that make a meaningful comparison between the two studies difficult. In the prior study, age-dependent silencing was most striking in the lines utilizing an embryonic (ξ) promoter. Thus, silencing could be interpreted as an appropriate development switch. However, five of five founder lines containing a transgene with a mini LCR and a human β-globin promoter driving a Lac-Z cassette also exhibited significant transgene silencing by 8 weeks of age. Since we did not analyze mice younger than 4 weeks of age, it is possible that we failed to detect early silencing in some lines. However, since lines A and G in our study express the transgene in >80% of red blood cells from the earliest time point tested, early silencing is not a universal feature of β-globin transgenes. In any case, silencing during adult life seems to us to be a more appropriate endpoint for analysis of lines expressing adult globin transgenes. Figure 3 View largeDownload slide X-gal staining of peripheral blood from adult animals in all 10 founder lines. Peripheral blood was obtained from mice between 2 and 5 months of age for all 10 founder lines (and a non-transgenic control, K). Peripheral blood cells were stained with the X-gal reagent for 2 h at room temperature, and then cytospin preparations were made and photographed. Note that the intensity of X-gal staining of red blood cells is approximately the same for X-gal + cells in all 10 transgenic founder lines, but that the percentage of positive cells varies widely from one founder line to the next. Data is shown from one representative experiment. These results were repeated at least twice with independent groups of mice, and were found to be virtually identical in every case. Figure 3 View largeDownload slide X-gal staining of peripheral blood from adult animals in all 10 founder lines. Peripheral blood was obtained from mice between 2 and 5 months of age for all 10 founder lines (and a non-transgenic control, K). Peripheral blood cells were stained with the X-gal reagent for 2 h at room temperature, and then cytospin preparations were made and photographed. Note that the intensity of X-gal staining of red blood cells is approximately the same for X-gal + cells in all 10 transgenic founder lines, but that the percentage of positive cells varies widely from one founder line to the next. Data is shown from one representative experiment. These results were repeated at least twice with independent groups of mice, and were found to be virtually identical in every case. Figure 4 View largeDownload slide Flow cytometric determination of β-galactosidase expression in the red blood cells of all 10 founder lines. Peripheral blood was obtained from all 10 founder lines and a non-transgenic control (K) on the same day. The FDG reagent was introduced into red cells using osmotic shock for 30 s, and then conversion of the fluorescent substrate was allowed to proceed for 5 min at 4°C. Flow cytometric analysis was then immediately performed. An irrelevant parameter is shown on the x -axis, and green fluorescence (a reflection of β-galactosidase activity) on the y- axis. The data is plotted in two dimensions so that the Lac-Z activity in founder lines with a very small percentage of positive events can be clearly seen off the baseline. Note that the percentage of FDG + cells correlates precisely with the percentage of X-gal positive cells shown in Figure 3 , and that the mean fluorescence intensity of the positive cells varies little from one founder line to another. Data is shown from one representative experiment. These results were repeated at least twice with independent groups of mice, and were found to be virtually identical in every case. Figure 4 View largeDownload slide Flow cytometric determination of β-galactosidase expression in the red blood cells of all 10 founder lines. Peripheral blood was obtained from all 10 founder lines and a non-transgenic control (K) on the same day. The FDG reagent was introduced into red cells using osmotic shock for 30 s, and then conversion of the fluorescent substrate was allowed to proceed for 5 min at 4°C. Flow cytometric analysis was then immediately performed. An irrelevant parameter is shown on the x -axis, and green fluorescence (a reflection of β-galactosidase activity) on the y- axis. The data is plotted in two dimensions so that the Lac-Z activity in founder lines with a very small percentage of positive events can be clearly seen off the baseline. Note that the percentage of FDG + cells correlates precisely with the percentage of X-gal positive cells shown in Figure 3 , and that the mean fluorescence intensity of the positive cells varies little from one founder line to another. Data is shown from one representative experiment. These results were repeated at least twice with independent groups of mice, and were found to be virtually identical in every case. Figure 5 View largeDownload slide Graphical representation of β-galactosidase expression in the 10 founder lines. ( A ) The adjusted ONPG values in the peripheral blood for all 10 founder lines is shown. This data was extrapolated from the data presented in Table 1 . ( B ) The data from the flow cytometric analysis is presented in graphical form. The data is presented from the highest to the lowest expressing line. The bar graphs represent the percentage of FDG + cells in each founder line, and the superimposed line represents the absolute mean fluorescence of the FDG + cells present within the peripheral blood of each line. Note that the total ONPG activity in the peripheral blood is highly correlated with the percentage of positive cells determined by FACS. Also, note that the average level of β-galactosidase per cell (reflected by mean fluorescence) varies only 10-fold from the lowest to the highest expressing founder line. Figure 5 View largeDownload slide Graphical representation of β-galactosidase expression in the 10 founder lines. ( A ) The adjusted ONPG values in the peripheral blood for all 10 founder lines is shown. This data was extrapolated from the data presented in Table 1 . ( B ) The data from the flow cytometric analysis is presented in graphical form. The data is presented from the highest to the lowest expressing line. The bar graphs represent the percentage of FDG + cells in each founder line, and the superimposed line represents the absolute mean fluorescence of the FDG + cells present within the peripheral blood of each line. Note that the total ONPG activity in the peripheral blood is highly correlated with the percentage of positive cells determined by FACS. Also, note that the average level of β-galactosidase per cell (reflected by mean fluorescence) varies only 10-fold from the lowest to the highest expressing founder line. The ‘decision’ to variegate occurs after the BFU-E stage of erythroid differentiation We next wished to determine whether the heterocellular pattern of transgene expression was due to the fact that different erythroid progenitors had different propensities to express the transgene, or whether each early erythroid progenitor would give rise to progeny with or without the potential to express the transgene. If the decision to variegate occurred before the BFU-E stage, then all of the cells within a given BFU-E colony would be expected to either express the transgene or not express it; the percentage of Lac-Z + colonies should reflect the percentage of positive red blood cells in the periphery. If the decision was made after the BFU-E stage, then all colonies should have the same pattern of transgene expression; the percentage of positive cells within an individual colony should reflect the percentage of positive cells in the peripheral blood ( Fig. 7 ). Figure 6. View largeDownload slide Minimal changes in the levels of transgene expression as a function of time. Several different F1 and F2 progeny from the transgenic founder lines had peripheral blood examined at various ages by flow cytometry and ONPG determinations on peripheral blood lysates. The percentage of FDG + cells in the peripheral blood of each animal is plotted as a function of time. Notice that the mice exhibit little or no change in the frequency of positive cells as a function of time. Figure 6. View largeDownload slide Minimal changes in the levels of transgene expression as a function of time. Several different F1 and F2 progeny from the transgenic founder lines had peripheral blood examined at various ages by flow cytometry and ONPG determinations on peripheral blood lysates. The percentage of FDG + cells in the peripheral blood of each animal is plotted as a function of time. Notice that the mice exhibit little or no change in the frequency of positive cells as a function of time. We obtained bone marrow from founder lines that demonstrated high, intermediate or low levels of transgene expression, and obtained peripheral blood for Lac-Z staining at the same time. The bone marrow was plated in methylcellulose cultures under conditions that favored the development of erythroid colonies. After 7 days of culture, the methylcellulose plates were treated with X-gal stain, allowed to develop overnight at 37°C, and then photographed and scored. Representative results are shown in Figure 8 . No X-gal + cells were found in the peripheral blood or the erythroid colonies of non-transgenic mice. The frequency of BFU-E, and the appearance of BFU-E colonies, was the same for all founder lines (not shown). The percentage of X-gal positive cells within individual BFU-E reflected the percentage of positive cells within the peripheral blood. Similar data were obtained for CFU-E (not shown). These data strongly suggest that integration site-dependent variegation occurs after the BFU-E stage of differentiation ( Fig. 7 ), since virtually all BFU-E exhibited the same potential to either express or not express the transgene in their progeny. Figure 7 View largeDownload slide Potential models for transgene variegation in BFU-E colonies. Patterns of Lac-Z + cells in BFU-E colonies are presented for two different models of variegation. If the decision for the transgene to be expressed or not expressed occurs prior to the BFU-E stage of erythroid differentiation, some colonies will be negative (depicted here as red cells) and others should exhibit Lac-Z staining (depicted as blue cells) in all of the BFU-E progeny. If the decision to variegate occurs after the BFU-E stage of differentiation, then all BFU-E colonies should exhibit both positively and negatively staining red blood cells. The percentage of positive cells within the BFU-E should reflect the percentage of positive cells in the peripheral blood. Figure 7 View largeDownload slide Potential models for transgene variegation in BFU-E colonies. Patterns of Lac-Z + cells in BFU-E colonies are presented for two different models of variegation. If the decision for the transgene to be expressed or not expressed occurs prior to the BFU-E stage of erythroid differentiation, some colonies will be negative (depicted here as red cells) and others should exhibit Lac-Z staining (depicted as blue cells) in all of the BFU-E progeny. If the decision to variegate occurs after the BFU-E stage of differentiation, then all BFU-E colonies should exhibit both positively and negatively staining red blood cells. The percentage of positive cells within the BFU-E should reflect the percentage of positive cells in the peripheral blood. Discussion In the present study, we have examined the mechanism of variegated transgene expression at cellular and molecular levels. Using a construct designed to direct expression of the Lac-Z reporter to the erythroid compartment of transgenic mice, we found that the different levels of transgene expression among lines was mostly related to the percentage of Lac-Z + red cells in the peripheral blood, while differences in the level of Lac-Z expression per cell accounted for a much smaller proportion of the variegation. In other words, our data suggest that a stochastic mechanism is responsible for much of this transgene's variegation, and it supports recent studies that have addressed the mechanisms involved in the production of variegation ( 36–38 ). Several studies have shown that enhancers act to increase the probability of expression of a linked promoter, not the transcriptional activity of that promoter ( 21 , 26–28 ). Collectively, these data and ours argue that the probability of transgene expression largely depends upon integration site. A recent alternative model (which is at odds with our data), suggests that β-globin enhancers act to increase the level of transgene expression per cell in a graded (rather than a binary) fashion ( 39 ). These authors based their conclusions on data derived from fetal liver cells, rather than mature adult red blood cells, making comparisons of the data difficult, because of the hemoglobin switch that occurs in that compartment. Figure 8 View largeDownload slide Lac-Z staining of BFU-E derived from a low, medium and high expressing transgenic mouse. Bone marrow cells from a mouse from a low expressing line (line B), an intermediate expressing line (line C) and a high expressing line (line A) were obtained and plated in methylcellulose under conditions that favored erythroid colony formation (see Materials and Methods). All of the colonies on the plate were overlayed with X-gal stain on day 10 after plating, and examined for Lac-Z + cells. Peripheral blood obtained from the mouse at the time of harvest was obtained, and X-gal staining was performed on it as well. The peripheral blood stains are shown in ( A )–( D ), and a representative BFU-E colony from each bone marrow plating experiment are shown in ( E )–( H ). Virtually all of the BFU-E colonies on an individual plate had the same pattern of Lac-Z expression. (A) and (E) are from a wild-type mouse; (B) and (F) are from line B (low); (C) and (G) are from line C (intermediate); and (D) and (H) are from line A (high expressing). Note that the percentage of Lac-Z + cells in the peripheral blood is closely mimicked by the percentage of Lac-Z + cells within each BFU-E, suggesting that the decision to express the transgene occurs after the BFU-E stage of differentiation. Figure 8 View largeDownload slide Lac-Z staining of BFU-E derived from a low, medium and high expressing transgenic mouse. Bone marrow cells from a mouse from a low expressing line (line B), an intermediate expressing line (line C) and a high expressing line (line A) were obtained and plated in methylcellulose under conditions that favored erythroid colony formation (see Materials and Methods). All of the colonies on the plate were overlayed with X-gal stain on day 10 after plating, and examined for Lac-Z + cells. Peripheral blood obtained from the mouse at the time of harvest was obtained, and X-gal staining was performed on it as well. The peripheral blood stains are shown in ( A )–( D ), and a representative BFU-E colony from each bone marrow plating experiment are shown in ( E )–( H ). Virtually all of the BFU-E colonies on an individual plate had the same pattern of Lac-Z expression. (A) and (E) are from a wild-type mouse; (B) and (F) are from line B (low); (C) and (G) are from line C (intermediate); and (D) and (H) are from line A (high expressing). Note that the percentage of Lac-Z + cells in the peripheral blood is closely mimicked by the percentage of Lac-Z + cells within each BFU-E, suggesting that the decision to express the transgene occurs after the BFU-E stage of differentiation. In our model, we did not observe significant changes in the pattern of transgene expression over the life span of mice derived from several different founder lines. Within a single line, the results of ONPG, X-gal, and FACS-gal assays performed on cohorts of mice (or serially in individual mice) aged from 2 to 24 months did not differ significantly in most cases. The interpretation of this experiment might have changed if the expression of X-gal protein affected the survival of red cells. We therefore showed that reticulocyte counts and hemoglobin levels were normal in all animals. In addition, we noted that the heterocellular pattern of Lac-Z expression remained consistent in animals after induction of hemolysis (causing an increase in reticulocytes) or irradiation (causing a decrease in reticulocytes). The β-galactosidase protein, therefore, must be long-lived within red cells. We therefore suggest that β-galactosidase expression does not perturb red cell kinetics, and that determination of Lac-Z activity in peripheral blood accurately reflects the stochastic nature of transgene expression in circulating red cells of all ages. The site of integration of the transgene into the genome differed among all 10 lines we analyzed. In contrast to other studies ( 40 , 41 ), low level transgene expression was not specifically associated with centromeric chromosomal integration in our study. Centromeric integration may therefore be sufficient to cause transgene silencing, but it is not necessary, and it is not the only mechanism that can cause it. In vitro analyses of erythroid development in our mice revealed that the decision to activate the transgene occurs after the progenitor stage (i.e. post-BFU-E, post-CFU-E). For example, within a single BFU-E colony from an intermediate expressing line, some of the cells stain positively for Lac-Z activity and some do not; all colonies have the same appearance. Thus, the heterocellular pattern of Lac-Z activity evident in peripheral blood is recapitulated in individual erythroid colonies derived from BFU-E and CFU-E of transgenic mice. These results are reminiscent of the pattern of HbF expression by normal adult erythroid bursts. Previous studies have shown that individual BFU-E progenitors are capable of giving rise to subclones containing only HbA, or a mixture of HbA and HbF ( 42 ). These data are best explained by a stochastic model, in which a probability function governs whether the progeny of adult BFU-E express HbF. The similarity between these observations and our own suggests that similar mechanisms may be responsible for the activation of the HbF program as well as our transgene. There are, however, two important differences between the systems. First, transgene-expressing red blood cells were evenly distributed throughout the entire BFU-E colony in our experiments. In contrast, HbF expression appears to be ‘all or none’ within subclones of a BFU-E. Thus, HbF variegation is much ‘coarser’ (occurring in large clusters of cells) than the variegation we observed. The second phenotypic difference is that all of the colonies derived from an individual animal in our studies had the same pattern of Lac-Z staining, while human BFU-E derived from a single patient were quite heterogeneous with respect to HbF staining. These findings suggest that activation of this transgene appears to be a stochastic event late in erythroid development, well beyond the BFU-E and CFU-E stages. The decision to express HbF, on the other hand, probably occurs just a few cell divisions beyond the BFU-E stage and within a narrow temporal ‘window’. These findings have significance for transgenic experiments in general. Integration site-dependent variegation of transgene expression complicates the interpretation of experiments where the direct transcriptional output of different constructs is being compared. However, if the phenotype under analysis is a complex event downstream of transgene expression (e.g. development of tumors in mice with targeted expression of oncogenes), then line to line variegation may not be a critical problem. Our data suggests that low level transgene expression in an organ may represent high level expression in a small number of cells. Therefore, in the example given, all cells that express the transgene should share the same risk for development of a phenotype (i.e. neoplastic transformation), but transgenic lines would differ in the number of cells per tissue at risk for development of the phenotype. In summary, our analysis of transgenic lines expressing an HS-2/β-globin/Lac-Z transgene supports a model in which differences in transgene expression are largely due to differences in the frequency of expressing cells, rather than differences in the level of expression per cell. Our data do not suggest that low level expression is frequently associated with centromeric chromosomal integration, or that age-dependent silencing is a universal property of globin transgenes. We have provided evidence that activation of this transgene is probably a stochastic event that occurs late in erythroid development, much like the decision to express HbF in adult red cells. The mechanisms that underlie these stochastic decisions are currently unknown, but of importance for the correct interpretation of many experiments involving transgenic mice. Acknowledgements This work was supported by NIH grants DK38682, DK 49786 and CA 49712 (T.J.L.). The authors thank Pam Goda for excellent care of our animals and Nancy Reidelberger for preparation of the manuscript. References 1 Chada K.,  Magram J.,  Raphael K.,  Radice G.,  Lacy E.,  Costantini F.. ,  Nature ,  1985, vol.  314 (pg.  377- 383) CrossRef Search ADS PubMed  2 Magram J.,  Chada K.,  Costantini F.. ,  Nature ,  1985, vol.  315 (pg.  338- 340) CrossRef Search ADS PubMed  3 Townes T.M.,  Lingrel J.B.,  Chen H.Y.,  Brinster R.L.,  Palmiter R.D.. ,  EMBO J. ,  1985, vol.  4 (pg.  1715- 1723) PubMed  4 Magram J.,  Niederreither K.,  Costantini F.. ,  Mol. Cell. Biol. ,  1989, vol.  9 (pg.  4581- 4584) CrossRef Search ADS PubMed  5 Behringer R.R.,  Hammer R.E.,  Brinster R.L.,  Palmiter R.D.,  Townes T.M.. ,  Proc. Natl. Acad. Sci. USA ,  1987, vol.  84 (pg.  7056- 7060) CrossRef Search ADS   6 Kollias G.,  Hurst J.,  deBoer E.,  Grosveld F.. ,  Nucleic Acids Res. ,  1987, vol.  15 (pg.  5739- 5747) CrossRef Search ADS PubMed  7 Tuan D.,  London I.M.. ,  Proc. Natl. Acad. Sci. USA ,  1984, vol.  81 (pg.  2718- 2722) CrossRef Search ADS   8 Tuan D.,  Solomon W.,  Li Q.,  London I.M.. ,  Proc. Natl. Acad. Sci. USA ,  1985, vol.  82 (pg.  6384- 6388) CrossRef Search ADS   9 Forrester W.C.,  Thompson C.,  Elder J.T.,  Groudine M.. ,  Proc. Natl. Acad. Sci. USA ,  1986, vol.  83 (pg.  1359- 1363) CrossRef Search ADS   10 Driscoll M.C.,  Dobkin C.S.,  Alter B.P.. ,  Proc. Natl. Acad. Sci. USA ,  1989, vol.  86 (pg.  7470- 7474) CrossRef Search ADS   11 Grosveld F.,  van Assendelft G.B.,  Greaves D.R.,  Kollias G.. ,  Cell ,  1987, vol.  51 (pg.  975- 985) CrossRef Search ADS PubMed  12 Ryan T.M.,  Behringer R.R.,  Martin N.C.,  Townes T.M.,  Palmiter R.D.,  Brinster R.L.. ,  Genes Dev. ,  1989, vol.  3 (pg.  314- 323) CrossRef Search ADS PubMed  13 Curtin P.T.,  Liu D.,  Liu W.,  Chang J.C.,  Kan Y.W.. ,  Proc. Natl. Acad. Sci. USA ,  1989, vol.  86 (pg.  7082- 7086) CrossRef Search ADS   14 Fraser P.,  Hurst J.,  Collis P.,  Grosveld F.. ,  Nucleic Acids Res. ,  1990, vol.  18 (pg.  3503- 3508) CrossRef Search ADS PubMed  15 Talbot D.,  Philipsen S.,  Fraser P.,  Grosveld F.. ,  EMBO J. ,  1990, vol.  9 (pg.  2169- 2178) PubMed  16 Caterina J.J.,  Ryan T.M.,  Pawlik K.M.,  Palmiter R.D.,  Brinster R.L.,  Behringer R.R.,  Townes T.M.. ,  Proc. Natl. Acad. Sci. USA ,  1991, vol.  88 (pg.  1626- 1630) CrossRef Search ADS   17 Liu D.,  Chang J.C.,  Moi P.,  Liu W.,  Kan Y.W.,  Curtin P.T.. ,  Proc. Natl. Acad. Sci. USA ,  1992, vol.  89 (pg.  3899- 3903) CrossRef Search ADS   18 Ellis J.,  Talbot D.,  Dillon N.,  Grosveld F.. ,  EMBO J. ,  1993, vol.  12 (pg.  127- 134) PubMed  19 Caterina J.J.,  Ciavatta D.J.,  Donze D.,  Behringer R.R.,  Townes T.M.. ,  Nucleic Acids Res. ,  1994, vol.  22 (pg.  1006- 1011) CrossRef Search ADS PubMed  20 Robertson G.,  Garrick D.,  Wu W.,  Kearns M.,  Martin D.,  Whitelaw E.. ,  Proc. Natl. Acad. Sci. USA ,  1995, vol.  92 (pg.  5371- 5375) CrossRef Search ADS   21 Walters M.C.,  Fiering S.,  Eidemiller J.,  Magis W.,  Groudine M.,  Martin D.I.K.. ,  Proc. Natl. Acad. Sci. USA ,  1995, vol.  92 (pg.  7125- 7129) CrossRef Search ADS   22 Walters M.C.,  Magis W.,  Fiering S.,  Eidemiller J.,  Scalzo D.,  Groudine M.,  Martin D.I.K.. ,  Genes Dev. ,  1996, vol.  10 (pg.  185- 195) CrossRef Search ADS PubMed  23 Magis W.,  Fiering S.,  Groudine M.,  Martin D.I.K.. ,  Proc. Natl. Acad. Sci. USA ,  1996, vol.  93 (pg.  13914- 13918) CrossRef Search ADS   24 Garrick D.,  Sutherland H.,  Robertson G.,  Whitelaw E.. ,  Nucleic Acids Res. ,  1996, vol.  24 (pg.  4902- 4909) CrossRef Search ADS PubMed  25 Sutherland H.G.E.,  Martin D.I.K.,  Whitelaw E.. ,  Mol. Cell. Biol. ,  1997, vol.  17 (pg.  1607- 1614) CrossRef Search ADS PubMed  26 Moon A.,  Ley T.J.. ,  Proc. Natl. Acad. Sci. USA ,  1990, vol.  87 (pg.  7693- 7697) CrossRef Search ADS   27 Moreau P.,  Hen R.,  Wasylyk B.,  Everett R.,  Gaub M.P.,  Chambon P.. ,  Nucleic Acids Res. ,  1981, vol.  9 (pg.  6047- 6068) CrossRef Search ADS PubMed  28 Weintraub H.. ,  Proc. Natl. Acad. Sci. USA ,  1988, vol.  85 (pg.  5819- 5823) CrossRef Search ADS   29 Lichter P.,  Cremer T.,  Borden J.,  Manuelidis L.,  Ward D.C.. ,  Hum. Genet. ,  1988, vol.  80 (pg.  224- 234) CrossRef Search ADS PubMed  30 Pinkel D.,  Straume A.T.,  Gray J.W.. ,  Proc. Natl. Acad. Sci. USA ,  1986, vol.  83 (pg.  2934- 2938) CrossRef Search ADS   31 Schmitz F.J.,  Werner E.. ,  Cytometry ,  1986, vol.  7 (pg.  439- 444) CrossRef Search ADS PubMed  32 Cynshi O.,  Shimonaka Y.,  Higuchi M.,  Imai N.,  Suzuki H.,  Togashi M.,  Okamoto M.T.,  Hirashima K.. ,  Br. J. Haematol. ,  1990, vol.  76 (pg.  414- 419) CrossRef Search ADS PubMed  33 Nolan G.P.,  Fiering S.,  Nocolas J.-F.,  Herzenberg L.A.. ,  Proc. Natl. Acad. Sci. USA ,  1988, vol.  85 (pg.  2603- 2607) CrossRef Search ADS   34 Fiering S.N.,  Roederer M.,  Nolan G.P.,  Micklem D.R.,  Parks D.R.,  Herzenberg L.A.. ,  Cytometry ,  1991, vol.  12 (pg.  291- 301) CrossRef Search ADS PubMed  35 Robertson G.,  Garrick D.,  Wilson M.,  Martin D.I.K.,  Whitelaw E.. ,  Nucleic Acids Res. ,  1996, vol.  24 (pg.  1465- 1471) CrossRef Search ADS PubMed  36 Milot E.,  Strouboulis J.,  Trimborn T.,  Wijgerde M.,  de Boer E.,  Langeveld A.,  Tan-Un K.,  Vergeer W.,  Yannoutsos N.,  Grosveld F.,  Fraser P.. ,  Cell ,  1996, vol.  87 (pg.  105- 114) CrossRef Search ADS PubMed  37 Boyes J.,  Felsenfeld G.. ,  EMBO J. ,  1996, vol.  15 (pg.  2496- 2507) PubMed  38 McDevitt M.A.,  Fujiwara Y.,  Shivdasani R.A.,  Orkin S.H.. ,  Proc. Natl. Acad. Sci. USA ,  1997, vol.  94 (pg.  7976- 7981) CrossRef Search ADS   39 Guy L.-G.,  Kothary R.,  Wall L.. ,  Nucleic Acids Res. ,  1997, vol.  25 (pg.  4400- 4407) CrossRef Search ADS PubMed  40 Festenstein R.,  Tolaini M.,  Corbella P.,  Mamalaki C.,  Parrington J.,  Fox M.,  Miliou A.,  Jones M.,  Kioussis D.. ,  Science ,  1996, vol.  271 (pg.  1123- 1125) CrossRef Search ADS PubMed  41 Dobie K.W.,  Lee M.,  Fantes J.A.,  Graham E.,  Clark A.J.,  Springbett A.,  Lathe R.,  McClenaghan M.. ,  Proc. Natl. Acad. Sci. USA ,  1996, vol.  93 (pg.  6659- 6664) CrossRef Search ADS   42 Stamatoyannopoulos G.,  Kurnit D.M.,  Papayannopoulou T.. ,  Proc. Natl. Acad. Sci. USA ,  1981, vol.  78 (pg.  7005- 7009) CrossRef Search ADS   43. Sambrook J.,  Fritsch E.F.,  Maniatis T.. ,  Molecular Cloning: A Laboratory Manual ,  1989 Cold Spring Harbor, NY Cold Spring Harbor Laboratory Press(pg.  16.66- 16.67) © 1998 Oxford University Press

Graubert, Timothy A.; Hug, Bruce A.; Wesselschmidt, Robin; Hsieh, Chih-Lin; Ryan, Thomas M.; Townes, Tim M.; Ley, Timothy J.

doi: N/Apmid: N/A

The random insertion of transgenes into the genomic DNA of mice usually leads to widely variable levels of expression in individual founder lines. To study the mechanisms that cause variegation, we designed a transgene that we expected to variegate, which consisted of a β-globin locus control region 5′ HS-2 linked in tandem to a tagged human β-globin gene (into which a Lac-Z cassette had been inserted). All tested founder lines exhibited red blood cell-specific expression, but levels of expression varied >1000-fold from the lowest to the highest expressing line. Most of the variation in levels of expression appeared to reflect differences in the percentage of cells in the peripheral blood that expressed the transgene, which ranged from 0.3% in the lowest expressing line to 88% in the highest; the level of transgene expression per cell varied no more than 10-fold from the lowest to the highest expressing line. These differences in expression levels could not be explained by the location of transgene integration, by an effect of β-galactosidase on red blood cell survival, by the half life of the β-galactosidase enzyme or by the age of the animals. The progeny of all early erythroid progenitors (BFU-E colony-forming cells) exhibited the same propensity to variegate in methylcellulose-based cultures, suggesting that the decision to variegate occurs after the BFU-E stage of erythroid differentiation. Collectively, these data suggest that variegation in levels of transgene expression are due to local, integration site-dependent phenomena that alter the probability that a transgene will be expressed in an appropriate cell; however, these local effects have a minimal impact on the transgene's activity in the cells that initiate transcription.