Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Understanding molecular mechanisms and predicting phenotypic effects of pathogenic tubulin mutations

Understanding molecular mechanisms and predicting phenotypic effects of pathogenic tubulin mutations Cells rely heavily on microtubules for several processes, including cell division and molecu- lar trafficking. Mutations in the different tubulin-α and -β proteins that comprise microtubules have been associated with various diseases and are often dominant, sporadic and congeni- OPENACCESS tal. While the earliest reported tubulin mutations affect neurodevelopment, mutations are Citation: Attard TJ, Welburn JPI, Marsh JA (2022) also associated with other disorders such as bleeding disorders and infertility. We per- Understanding molecular mechanisms and formed a systematic survey of tubulin mutations across all isotypes in order to improve our predicting phenotypic effects of pathogenic tubulin mutations. PLoS Comput Biol 18(10): e1010611. understanding of how they cause disease, and increase our ability to predict their pheno- https://doi.org/10.1371/journal.pcbi.1010611 typic effects. Both protein structural analyses and computational variant effect predictors Editor: Bert L. de Groot, Max Planck Institute for were very limited in their utility for differentiating between pathogenic and benign mutations. Biophysical Chemistry, GERMANY This was even worse for those genes associated with non-neurodevelopmental disorders. Received: June 20, 2022 We selected tubulin-α and -β disease mutations that were most poorly predicted for experi- mental characterisation. These mutants co-localise to the mitotic spindle in HeLa cells, sug- Accepted: September 28, 2022 gesting they may exert dominant-negative effects by altering microtubule properties. Our Published: October 7, 2022 results show that tubulin mutations represent a blind spot for current computational Peer Review History: PLOS recognizes the approaches, being much more poorly predicted than mutations in most human disease benefits of transparency in the peer review genes. We suggest that this is likely due to their strong association with dominant-negative process; therefore, we enable the publication of all of the content of peer review and author and gain-of-function mechanisms. responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1010611 Copyright:© 2022 Attard et al. This is an open Author summary access article distributed under the terms of the Creative Commons Attribution License, which Filament-like structures, called microtubules, are essential for cells to function, distribute permits unrestricted use, distribution, and material around the cell and organisms, and help cells grow. The building blocks of micro- reproduction in any medium, provided the original tubules are proteins called tubulins, which can rapidly polymerise and depolymerise. author and source are credited. Mutations in tubulin genes can have catastrophic consequences on many different types Data Availability Statement: All relevant data are of cells, leading to diseases such as bleeding defects, female infertility, and disorders provided within the manuscript and its Supporting impairing brain development. However, how these mutations cause disease and whether Information files. they can be predicted is still unknown. We used computational and experimental tech- Funding: This work was supported by the Medical niques to address these issues. First, we compared how disease-causing tubulin mutations Research Council via a Precision Medicine Doctoral PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 1 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Training Programme studentship to TJA and a Career Development Award (MR/M02122X/1) to and ones found in healthy people impact the structure of tubulin. Then, we tested the abil- JAM. JAM is a Lister Institute Research Fellow. ity of available computational predictors to distinguish between these two types of tubulin JPIW is supported by a Wellcome Trust Senior mutations. We found these programs poorly predict tubulin mutations that cause diseases, Research Fellowship (207430). The funders had no limiting their usefulness. Next, we studied disease-causing mutations that were not pre- role in study design, data collection and analysis, dicted by computational methods. We found that these did not prevent tubulin from decision to publish, or preparation of the forming microtubules, indicating these mutations change the function of tubulin without manuscript. inactivating them. Our work presents tubulins as a weakness of current computational Competing interests: The authors have declared predictors, potentially because they fail to consider different ways in which mutations that no competing interests exist. cause disease. Introduction Microtubules are polarised cytoskeletal filaments essential in several cellular processes, ranging from cell division to signalling and transport. They assemble into axons, cilia, and the mitotic spindle, while also providing tracks for microtubule-associated proteins (MAPs) and motors for molecular trafficking [1,2]. Microtubules self-assemble from tubulin-α and -β heterodimers, with the dynamics of their assembly and disassembly being integral to their function [3]. Tubulin-α and -β are ubiquitous in eukaryotes, while related proteins in the FtsZ family show similarity in sequence, structure and function in archaea and bacteria [4,5]. Nine tubulin-α and ten tubulin-β genes have been identified in humans, originating from evolutionary gene duplication events [6]. In the tubulin field, these tubulin paralogues are referred to as isotypes. Between tubulin-α and -β, conserva- tion in sequence and structure is high, especially at interfaces stabilising the heterodimer, and at contacts between tubulin heterodimers across (lateral) and along (longitudinal) protofila- ments [7]. While tubulin-α and -β both bind to GTP, only tubulin-β can hydrolyse it to GDP, with residues on these binding sites amongst the most conserved [8]. GTP hydrolysis enables distinct conformations that mediate the dimer’s ability to be incorporated into microtubules [9]. The C-terminal region makes up the outer surface of the microtubule and so contributes to most of the interactions with MAPs [10]. Furthermore, many differences in amino acid sequences between isotypes occur in this region [11] and in the unstructured, highly negative tail [12]. Tubulin-γ,δ andε are more divergent in sequence than tubulin-α and -β and are involved in the basal bodies of centrioles, rather than being self-assembled into dynamic poly- mers [13,14]. A wide range of genetic disorders–called ’tubulinopathies’–have now been attributed to tubulin mutations. Over 225 pathogenic mutations in human tubulin isotypes have been reported (see Table 1). These findings highlight the importance of understanding tubulin func- tion in different cell types. The first reported tubulin mutations associated with pathogenic phenotypes were found in TUBα1A, TUBβ2B, TUBβ3, and TUBβ4A and resulted in neurode- velopmental defects [15]. Mutations in theγ-tubulin isotype TUBγ1 –which is necessary for microtubule nucleation–also cause a similar neurodevelopmental disorder [16,17]. While mutations in TUBα4A and TUBβ4A have been linked to neurodegenerative disease [18,19], phenotypes outside the nervous system are now emerging. These include TUBβ1 mutations associated with bleeding disorders [20], a link between tubulin-α acetylation and reduced sperm motility [21], and TUBβ8 mutations connected with female infertility due to incorrect meiotic spindle assembly [22]. Despite the large number of identified pathogenic tubulin mutations, our understanding of the molecular mechanisms by which these mutations cause disease remains limited. The large PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 2 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Table 1. Summary of all tubulin isotypes considered for this study. The number of dominant pathogenic mutations identified for each isotype is denoted, as well as their pathogenicity classification. Where necessary, additional comments about the pathogenicity type and references for all mutations are also included. MCM = missense constraint metric (Z-Score obtained from gnomAD); ALS = amyotrophic lateral sclerosis; H-ABC = hypomyelination with atrophy of basal ganglia and cerebellum. Isotype # # MCM Phenotype classification References and comments pathogenic gnomAD TUBα1A 67 680 5.584 neurodevelopmental Various neuropathies [35,79–83] TUBα1B 0 696 5.412 cancer-only Roles in resistance to chemotherapy [84–86] TUBα1C 0 153 2.171 cancer-only Identified as an oncogene [87] TUBα3D 0 343 1.533 non-disease TUBα3E 0 531 -0.296 non-disease Only 1 homozygous mutation known, which was excluded [88] TUBα4A 7 255 3.303 other All linked with ALS [19]. One deletion mutation also identified (excluded) (neurodegenerative) TUBα8 0 201 0.631 non-disease Homozygous 14bp deletion linked with polymicroglia [24] (excluded) TUBβ1 6 268 0.003 other (bleeding Linked to platelet defects [45–47] disorders) TUBβ2A 3 192 5.263 neurodevelopmental Simplified gyral patterning, infantile-onset epilepsy and progressive spastic ataxia [30, 89] TUBβ2B 29 190 5.120 neurodevelopmental Various neuropathies [15,29,55,57,64,90–92] TUBβ3 24 76 4.579 neurodevelopmental Various neuropathies, a few mutations associated with milder phenotypes [15,32,55,57,64,66,93,94] TUBβ4A 38 198 4.262 neurodevelopmental Mostly H-ABC and Hypomyelination, some less severe and a few linked with Dystonia [56,95–112] TUBβ4B 2 201 4.498 other (sensory) Linked heavily with cancer [113]. Two mutations have been linked to a sensorineural neurodegenerative disease impairing vision [18] TUBβ5 7 198 5.625 neurodevelopmental Various neuropathies, mostly severe [28,114–119] TUBβ6 1 74 2.637 neurodevelopmental Linked with non-progressive congenital facial palsy [120] TUBβ8 35 464 1.860 other (female infertility) Linked to oocyte maturation defects [22,23,50–53] TUBβ8B 0 228 -2.372 non-disease TUBγ1 8 57 4.155 neurodevelopmental All linked to cortical development malformations, one of unknown inheritance [16,17] (excluded). Two deletions also cause a lack of centrosome localisation [121] TUBγ2 0 113 2.443 non-disease https://doi.org/10.1371/journal.pcbi.1010611.t001 majority of pathogenic tubulin mutations involve missense changes (i.e. single amino acid resi- due substitutions) and have autosomal dominant inheritance. There are only a few known exceptions, including homozygous null and in-frame deletion mutations in TUBβ8 that can cause an oocyte maturation defect [23], a dominant nonsense mutation that results in a slightly truncated TUBα4A linked to the neurodegenerative disease amyotrophic lateral sclerosis [19], and a recessive intron deletion in TUBα8 in polymicrogyria patients that interferes with splic- ing, producing shorter mRNAs that do not contain exon 2 [24]. The absence of any known protein null mutations in tubulins causing dominant disease is striking. This tells us that the molecular mechanism underlying dominant mutations cannot simply be haploinsufficiency, whereby disease is caused by a complete lack of functional pro- tein produced from the mutant allele. One possibility is that disease is caused by a milder loss of function (i.e. hypomorphic mutations). For example, pathogenic TUBα1A mutations have been reported to disrupt interactions of the nascent protein with tubulin chaperones, which impairs heterodimer formation, suggesting that disease can be caused by a partial loss-of-func- tion [25]. Alternatively, a pathogenic mutation could act via a non-loss-of-function mecha- nism, having a dominant-negative effect or causing a gain of function [26]. This would typically be associated with the mutant protein retaining the ability to incorporate into micro- tubules, which has been observed for many pathogenic tubulin mutations [27–32]. For domi- nant-negative mutations, the incorporation of mutant protein directly or indirectly disrupts PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 3 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations the activity of the wild-type protein [26]. In these cases, the mutant tubulin retains its ability to form a heterodimer and assemble into microtubules before consequently impacting function in some other way, e.g. by perturbing microtubule properties or disrupting interactions with MAPs. For instance, TUBβ3 mutations that alter the charged surface of the microtubule pre- vent molecular motors from binding and thus have profound impacts on cellular transport [33]. Changes in microtubule properties could also induce a gain of function, as has been pro- posed for the pathogenic T178M variant in TUBβ2A and TUBβ3, which has been reported to make microtubules more stable and cause altered microtubule growth dynamics [34]. With the increasing accessibility of sequencing data, novel tubulin variants are being con- tinually discovered [35]. Since it is impractical to test them all experimentally, there is a strong need for computational approaches to identify tubulin mutations most likely to be pathogenic. Many variant effect predictors (VEPs) have been developed in recent years, and some of these are now in widespread use to help identify mutations that potentially have clinical significance [36]. However, the performance of these predictors can vary quite dramatically across different proteins and, to our knowledge, there has been no systematic assessment of their performance on tubulin mutations specifically. Missense mutations that cause pathogenicity by dominant- negative or gain-of-function mechanisms tend to be poorly predicted by most currently avail- able VEPs [37], which could potentially limit their applicability to tubulins. It is therefore important for us to understand to what extent we can rely on computational predictors when assessing tubulin mutations. In this study, we have first performed a systematic survey of known pathogenic missense mutations across all human tubulins and analysed their positions within the three-dimensional structures of tubulin heterodimers. This approach has allowed us to look for patterns in muta- tions across isotypes, structural locations, and phenotypes, in an attempt to obtain insight into the likely molecular disease mechanisms. Next, we assessed the performance of several differ- ent VEPs in distinguishing between pathogenic and putatively benign missense variants, observing that the predictive performance of all tested methods is poor compared to most other proteins. Finally, we have selected pathogenic tubulin mutations that were poorly pre- dicted by computational approaches for experimental characterisation and show that the mutant proteins are able to be incorporated into microtubules, consistent with a likely domi- nant-negative mechanism. Our work suggests that many tubulin pathogenic mutations that act via non-loss-of-function mechanisms cause pathogenic phenotypes that cannot be explained computationally using current methods that rely on sequence conservation or pro- tein structure. Overall, this study highlights the need for a greater understanding of microtu- bule-protein interactions to understand the molecular mechanisms underlying tubulinopathies. Results Survey of tubulin missense mutations First, we compiled as many previously identified pathogenic or likely pathogenic dominant tubulin missense mutations as possible, using online databases [38,39] and extensive literature searching (S1 Table). In addition, we also identified missense variants in tubulin genes observed across >140,000 people from the gnomAD v2.1 database [40]. Given that the gno- mAD dataset comprises mostly healthy individuals without severe genetic disorders, these var- iants are unlikely to cause dominant disease, and we therefore refer to them as "putatively benign". However, we acknowledge that some of these variants could have milder effects, vari- able penetrance, or be associated with late-onset disease. Table 1 shows the numbers of patho- genic and gnomAD missense variants for each tubulin isotype and the associated type of PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 4 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations genetic disease. Somatic mutations in tubulins are also implicated in cancer development; however, these mutations are likely to provide a selective advantage to cancer cells by provid- ing resistance to chemotherapeutic drugs [41–43]. Hence these mutations might obscure our results and have not been included in our study, although we have noted two isotypes with links to cancer in which no other disease-related mutations have been identified yet (Table 1). While pathogenic mutations occur throughout the tubulins, mutations in both tubulin-α and -β show clustering towards the intermediate and C-terminal domains when shown in the con- text of the linear amino acid sequence (Fig A in S1 Text). We also considered the gene-level missense constraint metric (MCM) scores provided by the gnomAD database (Table 1). These are derived from a model based on sequence context to predict the number of expected variants present in a healthy population relative to the number of actual variants observed [44]. They provide a metric for the tolerance of each isotype to mis- sense variation, with higher values representing genes that are more intolerant to amino acid sequence changes. Interestingly, we observed high MCM scores for isotypes linked with neuro- developmental disorders (TUBα1A, TUBβ2A, TUBβ2B, TUBβ3, TUBβ4A and TUBβ5). TUBβ6 is the only exception, and has only one pathogenic mutation reported so far, causing congenital facial palsy. These scores contrast with the much lower scores observed for TUBβ1 and TUBβ8, associated with platelet defects [45–49] and female infertility [22,23,50–53], respectively, which suggest that they are much more tolerant to sequence variation. Overall, our analysis indicates there is a stronger sequence constraint in the human population for tubulin isotypes that contribute to neurodevelopment. This may be due to the selective pres- sure of the process, compared to tubulin isotypes expressed in cells that affect organism fitness to a lesser extent. Protein structural context is of limited utility for explaining tubulin mutation pathogenicity Next, we mapped all variants to three-dimensional structures of tubulins in the Protein Data Bank (PDB). We classified each mutation-carrying residue from every tubulin isotype based upon its structural location: whether it occurs at an intermolecular interface, on the protein surface, or buried in the protein interior. We used a hierarchy to classify structural locations (Fig B in S1 Text). First, given that GTP binding (and hydrolysis for tubulin-β) is essential, we classified residues on the GTP nucleotide binding site as ’GTP binding’. Other tubulin muta- tions occurring at intermolecular interfaces were classified into three categories depending on where they occur in the model structures described in S2 Table: at the intradimer interface, at an interdimeric interface (which we refer to as microtubule contacts, as they support lateral and longitudinal interactions with microtubules), or at interfaces with MAPs. Remaining mutations not at interface positions were then classified depending on whether they were on the protein interior or surface according to relative solvent accessible surface area [54]. Finally, we labelled any residues that were not present in any ordered part of any experimentally derived tubulin structures as ’outside of structure’. These typically occurred in the disordered C-terminal region, or at positions of divergence between the sequences of mapped structures and the tubulin isotype in question. Notably, while several gnomAD variants occurred at ’out- side of structure’ positions, this was true for only a single pathogenic mutation: R307H in TUBβ1. Interestingly, all other tubulin-β isotypes have a histidine at this position, explaining why the residue could not be mapped within our pipeline, as the wild-type amino acid residue needs to be present in the structure. First, we used this classification system to assess whether tubulin mutations are enriched at particular locations. Despite the large dataset size, no significant differences in the locations of PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 5 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 1. Comparison of the distribution of structural locations between pathogenic and putatively benign tubulin variants. The proportion of pathogenic and putatively benign gnomAD mutations in each location type for tubulin-α,β andγ globally (A) and for individual isotypes (B). Mutation totals for each group are shown at the bottom, and only isotypes with at least five pathogenic mutations were included. Fisher’s exact tests were used to compare frequencies between gnomAD and pathogenic mutations considering each family and isotype separately. Asterisks indicate a location class with a significantly higher � � � � � � � � � � proportion of mutations compared to its corresponding group, where p < 0.05, p < 0.01, p < 0.001, p < 0.0001. https://doi.org/10.1371/journal.pcbi.1010611.g001 pathogenic or putatively benign gnomAD mutations were found across tubulin-α isotypes col- lectively (Fig 1A) or in any tubulin-α isotype individually (Fig 1B). This result was surprising given that the link between TUBα1A and disease remains the most well-established of all tubu- lin isotypes, with 67 pathogenic variants linked with various neurodevelopmental defects [35]. Observations were similar for TUBα4A, the only other tubulin-α isotype with dominant path- ogenic mutations (Fig 1B) [19]. Furthermore, pathogenic mutations in both isotypes showed no obvious patterns or spatial clustering when visualised within the protein structure (Figs 2 and C and D in S1 Text). In contrast to tubulin-α, we do observe a significant enrichment of pathogenic tubulin-β mutations relative to gnomAD variants at GTP binding residues and at lateral or longitudinal (i.e. interdimeric) microtubule contacts (Fig 1A), especially in TUBβ2B and TUBβ4A (Fig 1B). Mutations in these isotypes have been extensively linked to neuronal defects, and can be observed to form spatial clusters on tubulin structures (Fig 2). In contrast, putatively benign gnomAD variants are significantly more likely to occur on the protein surface or at ’outside of structure’ positions (Fig 1B). TUBβ3 and TUBβ5 are also linked to similar disease phenotypes, but do not show any notable differences in structural location [15,28,55–57]. Interestingly, we found no significant differences in structural location in tubulin-β isotypes associated with diseases outside the neuronal system (Fig 1B), nor did pathogenic mutations show any clear patterns or clustering at the three-dimensional structure level (Fig 2). We did, however, note that many TUBβ1 variants occurred at ’outside of structure’ residues (Fig 1B). This is due primarily to the greater sequence divergence between TUBβ1 and experimentally determined tubulin structures, as any positions where the TUBβ1 sequence is different than the aligned positions within the available structures will remain unmapped using our approach. The centrosome-localised TUBγ1 has also been associated with neurodevelopmental dis- ease [16,17]. Using the crystal structure of TUBγ1 containing two molecules per asymmetric unit (PDB ID: 3CB2) [58], we observed that 4 out of 8 pathogenic mutations identified so far occur at interfaces responsible for lateral interactions between TUBγ1 monomers (Figs 1 and PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 6 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 2. Visualisation of pathogenic tubulin mutations on three-dimensional structures of tubulin heterodimers. Structures of tubulin isotypes with at least five identified pathogenic mutations. Coloured residues indicate pathogenic mutations according to their location. The position of the tubulin subunit in relation to the microtubule (MT) is noted at the top. Residues in red denote mutations buried in the protein interior, while light cyan residues are on the surface. Green residues indicate mutations occurring at GTP binding interfaces, while ones in magenta highlight residues on intradimer interfaces. Navy blue and yellow residues mark mutations on residues interacting with microtubule associating proteins (MAPs) and other tubulin dimers, respectively. PDB IDs: 6s8k (for all tubulin-α andβ isotypes) and 3cb2 (for TUBγ1). Alternate versions of this figure with the structures rotated 90˚ and 180˚ around the y-axis are provided in Figs C and D in S1 Text. https://doi.org/10.1371/journal.pcbi.1010611.g002 2, labelled in magenta). We suspect these interactions to occur between adjacent TUBγ1 chains withinγ-TURC [59]. Their distribution was not significantly different to gnomAD variants, although we note the small sample size. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 7 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 3. Comparison of predicted changes in protein stability between pathogenic and putatively benign tubulin mutaitons.ΔΔG values representing the predicted change in free energy of folding were calculated with FoldX considering the structure of the monomeric subunit only. Scores are shown for tubulin-α and -β families globally (A) and in isotypes with at least five identified pathogenic mutations (B). Maroon diamonds indicate the meanΔΔG values, and mutation totals for each group are also shown at the bottom. The p-values displayed were obtained via unpaired Wilcoxon tests. https://doi.org/10.1371/journal.pcbi.1010611.g003 Most pathogenic tubulin mutations are not highly disruptive to protein structure Next, we considered the predicted structural perturbations of pathogenic and putatively benign gnomAD missense variants using FoldX [60]. This outputs aΔΔG value, in units of kcal/mol, with positive values indicating that a mutation is likely to destabilise protein struc- ture and negative values indicating predicted stabilisation. Previous work has shown that com- putationally predictedΔΔG values can sometimes show considerable utility for the identification of pathogenic missense mutations, and for understanding likely molecular dis- ease mechanisms [61]. Interestingly, we observe no significant differences between theΔΔG values of pathogenic and gnomAD missense variants for tubulin-α, -β or -γ (Fig 3A). Of the individual isotypes, only TUBβ2B shows significantly higherΔΔG values for pathogenic mutations (p = 0.01), although this would not remain significant when accounting for multiple testing (Fig 3B). We initially usedΔΔG values that only consider the structural impact of variants on the monomer alone, as they are more consistent between structures. However, we also observed very similar results using the fullΔΔG values calculated using the entire complex, including intermolecular interactions, as well as when using absoluteΔΔG values (Fig E in S1 Text). These results suggest that the structural destabilisation is not a primary molecular disease mechanism underlying pathogenic tubulin mutations, and that considering structural impact is not particularly useful for differentiating between pathogenic and benign tubulin variants. Notably, this aligns with recent work showing that the predicted effects on protein stability tend to be much milder in gain-of-function and dominant-negative mutations than for patho- genic mutations associated with a loss of function [37], supporting the idea that most patho- genic tubulin mutations are due to non-loss-of-function effects. Variant effect predictors show poor performance in discrimination between pathogenic and putatively benign tubulin mutations Next, we assessed the abilities of 25 different VEPs to distinguish between pathogenic and putatively benign tubulin missense mutations. A complete set of predictions for every tubulin mutation from all VEPs is provided in S3 Table. To compare the performance of different PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 8 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations VEPs, we first used a metric of predictor performance, known as the receiver operating charac- teristic (ROC) area under the curve (AUC) generated from each VEP over the entire dataset of tubulin missense variants (Fig 4A). Overall, the VEPs performed very poorly. Most had overall AUCs below 0.6, with the top-performing predictor, REVEL, having an AUC of only 0.68. In contrast, a recent study using a very similar methodology found that many VEPs had overall AUCs above 0.8, e.g. REVEL had an AUC of 0.9 for haploinsufficient disease genes, 0.85 for genes associated with a gain-of-function, and 0.83 for genes associated with dominant-nega- tive effects [37]. Thus, even considering that VEPs tend to do worse for non-loss-of-function mutations, the performance we observe here for tubulin mutations is strikingly poor. It is important to note that our analysis is likely to overstate the predictive power of some VEPs. Supervised machine learning approaches underpin most VEPs, and typically use data- sets of known pathogenic and benign variants for training. Since some VEPs are likely to have been trained using some of the tubulin mutations in our evaluation, their performance has a strong possibility of being overstated. This problem is particularly acute for metapredictors, including the top-performing methods in our analysis, REVEL and M-CAP, which combine supervised learning with multiple other predictors as inputs. In contrast, predictors based upon unsupervised machine approaches and those utilising empirical calculations should be free from this bias. Therefore, given the performance of the unsupervised predictor DeepSe- quence, ranking third overall, we likely consider it to be the most reliable predictor of tubulin mutation pathogenicity, consistent with its top-ranking performance in a recent study [62]. However, even DeepSequence only achieves an AUC of 0.63 for tubulin mutations here, com- pared to well over 0.8 for all disease-associated proteins tested in that study. Next, we compared the AUCs calculated for individual isotypes across all VEPs, consider- ing isotypes with at least 10 pathogenic mutations. We found that pathogenic mutations in the tubulin-β isotypes TUBβ2B, TUBβ3 and TUBβ4A, which are all associated with neurodevelop- mental diseases, were predicted better compared to TUBβ8, associated with oocyte maturation defects (Fig 4B). Therefore, we classified tubulins into two groups based upon observed disease phenotypes: neurodevelopmental and other (as classified in Table 1). We observed a signifi- cantly better performance on isotypes linked to neurodevelopmental diseases (Fig 5A). Fig 4. Assessment of VEP performance for identification of pathogenic tubulin mutations. (A) ROC AUC values for each VEP across all tubulin-α and -β isotypes with at least one identified pathogenic mutation, colour coded according to predictor category. (B) Distribution of ROC AUC values across all VEPs for isotypes with at least 10 identified pathogenic missense mutations. Dashed line indicates the performance of a random predictor. Maroon diamonds indicate the mean area. https://doi.org/10.1371/journal.pcbi.1010611.g004 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 9 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 5. Comparison of VEP performance on pathogenic mutations in tubulin genes associated with neurodevelopmental vs other disease phenotypes. (A) Distribution of ROC AUC values across all VEPs for isotypes with mutations linked with neurodevelopmental or other disease phenotypes. The p-value stated was obtained via a paired Wilcoxon test. Maroon diamonds indicate the mean area. (B) ROC AUC values for each VEP in isotypes with mutations associated with neurodevelopmental or other disease phenotypes. Dashed line indicates the performance of a random predictor. https://doi.org/10.1371/journal.pcbi.1010611.g005 Interestingly, when we compare the performance on neurodevelopmental vs other disorders across the individual VEPs (Fig 5B), we observe that metapredictors and supervised VEPs, like REVEL, M-CAP and VEST4, outperformed all other VEPs on isotypes linked to neurodeve- lopmental disease but showed a drastic decrease in performance on other phenotypes. In con- trast, unsupervised DeepSequence shows very similar performance between the two groups. This strongly suggests that that certain VEPs have likely been overfitted in their training against the neurodevelopmental mutations. Given its overall performance in our analyses and unsupervised nature, we currently rec- ommend DeepSequence for predicting the effects of tubulin mutations, although we empha- sise that its predictive utility is still relatively limited. Therefore, we have produced DeepSequence predictions for every possible amino acid substitution across most tubulin iso- types and have provided them as a resource (S4 Table). We have also calculated optimal thresholds for DeepSequence using the closest point to the top left corner of our ROC curves. Based upon this, we suggest that DeepSequence scores lower than -5.89 are likely to be patho- genic in isotypes linked with neurodevelopmental phenotypes, and lower than -4.83 for iso- types linked with other phenotypes. The most poorly predicted tubulin pathogenic mutations likely act via non- loss-of-function mechanisms To further examine the contributing factors behind the limited performance of VEPs in detect- ing pathogenic tubulin mutations, we sought to identify the pathogenic mutations predicted most poorly by current approaches. To do this, we developed a ranking method to examine mutations across all the VEPs used in this study. First, all predictions of gnomAD and patho- genic tubulin mutations were amalgamated. We then transformed all scores from individual VEPs to be on the same scale for comparison. Next, for each VEP, cumulative distribution ranks were computed for normalisation within a window partition across the combined data- set. Finally, for each mutation, these normalised ranks were averaged across all VEPs to gener- ate a new metric we termed mean cumulative distribution (MCD), provided in S5 Table. Essentially the MCD provides a single value for each mutation representing how damaging it is predicted to be across all of the VEPs we used in our analysis, ranging from 0 for mildest to PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 10 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 1 for most disruptive. MCD replicated the results we observed for the VEP analysis, with the most significant differences between pathogenic and gnomAD values being observed in the same isotypes (TUBβ2B, TUBβ3, TUBβ4A) that showed higher ROC values (Fig F in S1 Text). Next, we used the MCD scores to identify pathogenic tubulin mutations that were most poorly predicted by the VEPs (Table 2). These mutations generally involve substitutions between chemically similar amino acids. In particular, mutations from valine to isoleucine at position 353 in three different isotypes–TUBα1A, TUBβ5 and TUBβ8 –are among the worst predicted, with the variant in TUBβ8 having the lowest MCD score of any pathogenic tubulin mutation. Phenotypically, in TUBα1A and TUBβ5, V353I is related to the malformations of cortical development [15], while the mutation is associated with female infertility in TUBβ8, although whether the mutation is causing the issue is not shown [50]. Consequently, we experimentally tested whether mutations in the 353 position for TUBα1 and TUBβ8 enabled tubulin incorporation into microtubules. As V353I introduces a mild physicochemical change, as with most of the poorly predicted pathogenic tubulin mutations, we also tested a more disruptive mutation in both isotypes where the hydrophobic valine is changed to a positively charged arginine. V353R has to our knowledge not been observed in humans (nor is it possible with only a single nucleotide substitution), but it allows us to com- pare an apparently mild substitution to a more disruptive substitution at the same position. We transiently transfected the fluorescently tagged tubulin mutants into HeLa cells and ana- lysed whether they were incorporated into the mitotic spindle, where microtubules can be clearly observed and where spindle architecture relies on MAPs. N-terminally tagged mCherry-TUBα1 incorporated into microtubules with the aid of a linker; however mCherry- TUBβ8 did not. We therefore tested a C-terminally tagged TUBβ8-GFP, which did successfully incorporate. Table 2. The most poorly predicted pathogenic tubulin mutations. For each of the VEPs used in this study, cumula- tive distribution ranks were generated in a combined dataset of putatively benign and pathogenic tubulin variants. MCD scores for each variant were calculated by averaging the ranks from each predictor. Pathogenic variants with the lowest MCD scores are shown here, representing those most poorly predicted by VEPs. Isotype Mutation MCD Score TUBß8 V353I 0.188 TUBa1A I238V 0.215 TUBa1A I219V 0.215 TUBß8 I4L 0.221 TUBa1A D127E 0.245 TUBß8 A352S 0.247 TUBß1 R307H 0.253 TUBß8 M330I 0.280 TUBß5 V353I 0.280 TUBa1A S54N 0.300 TUBß8 R2K 0.311 TUBa1A A270S 0.326 TUBa1A V353I 0.335 TUBa1A I188L 0.341 TUBß3 M323V 0.342 TUBß8 F200L 0.352 TUBß8 V175M 0.354 TUBa1A I5L 0.357 TUBa1A V409I 0.371 TUBa1A R214H 0.372 https://doi.org/10.1371/journal.pcbi.1010611.t002 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 11 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Both wild-type TUBα1 and TUBβ8 isotypes strongly incorporated into the mitotic spindle in HeLa cells, co-localising with tubulin (Fig 6A and 6B and 6G in S1 Text). The V353I mutants also showed incorporation that was not significantly different from wild type for both isotypes. This is also consistent with previous work showing that V353I in TUBβ5 can still form a heterodimer and be incorporated into microtubules [28]. Interestingly, however, the V353R mutant showed severe reduction in incorporation for TUBα1, but no significant effect for TUBβ8. We also tested whether the tubulin mutations affected mitotic spindle morphology and measured the mitotic spindle length (Fig 6C). Notably, while TUBα1 V353I did not show reduced incorporation into microtubules, it did lead to significantly shorter spindles, suggest- ing the mutation may ultimately interfere with correct microtubule dynamics or interactions with microtubule-associated proteins. Other mutants showed no significant difference from wild type. We also confirmed that spindle aspect ratios were similar between wild-type and mutant tubulin constructs, indicating spindle scaling is conserved (Fig H in S1 Text). Our results clearly suggest a possible molecular mechanism underlying genetic disease in the TUBα1A V353I mutation: the mutant can incorporate into microtubules along with wild- type protein, but appears to have a dominant-negative effect on spindle organisation of micro- tubules. In contrast, the V353R mutation, which we predict to be more severe at a protein structural level, shows reduced incorporation and thus has no effect on spindle morphology. This mirrors the recent observation that dominant-negative and gain-of-function mutations tend to be more structurally mild than loss-of-function mutations [37]. None of the tested TUBβ8 mutations show any phenotypic effect. It is interesting to note that V353 residues in tubulin-α and -β are not at precisely equivalent positions in a sequence alignment or in the structures of tubulin-α and -β (Fig I in S1 Text). V353 is a surface residue in tubulin-α, close to the longitudinal interdimeric interface, whereas in tubulin-β, it is more buried in the protein interior, but close to the intradimer interface, and thus it is unsurprising that mutations at V353 could have different molecular effects in each isotype. It is possible that the molecular effects of V353 mutations in TUBβ8 are too subtle to detect in our experimental system. Alternatively, the evidence that V353I is a true pathogenic mutation in TUBβ8 is quite limited. It was observed in a heterozygous state in a single patient with infertility, but it is unknown if either of her parents also had the mutation [50]. Thus, another possible explana- tion for the poor computational prediction of TUBβ8 V353I is that it is not really pathogenic. Discussion Tubulinopathies have previously been associated with defects in cortical brain development [15,57,63–66]. However, several pathogenic tubulin mutations reported in recent years are not linked to neurodevelopmental diseases, and thus the phenotypical profile of ’tubulinopathies’ needs redefining. As the mechanisms by which tubulin mutations give rise to these disease phenotypes are generally unknown, the extensive repertoire of newly identified mutations now at our disposal provides an opportunity to study pathogenicity globally, across the tubulin family. The strong conservation in tubulin sequence and structure across isotypes [12], and the lethality of tubulin mutations in yeast [67] might lead to the assumption that many pathogenic mutations would cause protein instability and consequently a loss of function. However, this appears not to be the case, with predictions of changes in protein stability showing very little ability to discriminate between pathogenic and putatively benign mutations, with the excep- tion of a slight trend observed for TUBβ2B. There is some previous evidence for a loss-of-func- tion mechanism underlying some pathogenic tubulin mutations, e.g. the R264C mutation in TUBα1A perturbs chaperone binding and reduces the extent of heterodimer assembly, which PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 12 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 6. Effects of V353 mutants on microtubules and mitotic spindles in cells. (A) Representative live-cell images of mitotic HeLa cells after being transfected with fluorescently tagged wild type and mutant tubulin-α1B and tubulin-β8 constructs (green) and incubated with SiR-tubulin (red) (B) Wide linescans across the whole mitotic spindle were taken to measure fluorescence intensity of tubulin across the mitotic spindle compared to total tubulin. For each cell, two-sided Spearman correlations were calculated between both signal intensities. (C) Spindle length measurements for the cells transfected with fluorescently tagged wild-type and mutant tubulin-α1B and tubulin-β8 and measured in (B). For each construct, median values are displayed with 95% confidence intervals. � � � � � � � Asterisks indicate Kruskal-Wallis test (with post-hoc Dunn) significance values, , P < 0.001; , P < 0.0001. Bars, 5μm. https://doi.org/10.1371/journal.pcbi.1010611.g006 has been suggested to account for its pathogenic effects [25]. However, it appears that the over- all role of destabilisation in driving tubulin pathogenicity must be very limited. Tubulin mutations can act via dominant-negative or gain-of-function mechanisms, where the mutant protein incorporates into the microtubule lattice and instead disrupts its function. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 13 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations These effects have been reported in tubulins before and include mutations in TUBα1A [27] and others across different tubulin-β isotypes [28–32]. Once they are incorporated into micro- tubules, these mutations may alter biophysical microtubule properties or disrupt interactions with MAPs. The latter model includes interactions with non-motor MAPs, like NuMA and PRC1 [31,32], and the impairment of motor protein binding and, subsequently, cellular trans- port [33]. The disruption of tubulin-MAP interactions may be an especially pertinent mecha- nism given that mutations on protein-protein interfaces are more likely to be pathogenic [68]. Here, multiple copies of mutant tubulin incorporated into microtubule can compound a mutation’s effect. Also, some tissues express multiple tubulin-β isotypes, especially in the developing brain [69]. Here, redundancy may mitigate any loss-of-function in a specific iso- type, but cells struggle to cope with dominant-negative or gain-of-function effects. The classic definition of the dominant-negative effect involves the mutant protein disrupt- ing the activity of the wild-type protein, either directly or indirectly, whereas the phenotypic effects of gain-of-function mutations are due to the mutant protein doing something different than the wild-type protein [26]. For most proteins that form relatively small complexes, classi- fication of a mutation as being associated with to either mechanism should usually be straight- forward. However, given the large size of microtubule assemblies and their diverse functional roles, such classification may be ambiguous for tubulins. It depends on whether disease is due to reduced microtubule function (dominant negative), or whether altered microtubule proper- ties are driving disease (gain of function). While here we have referred to TUBα1 V353I, which incorporates into microtubules and reduces spindle length, as dominant negative, it is possible that its pathogenic effects might be better described as gain of function. These assays give limited data because of their lack of sensitivity. Further experimental work and more quantitative assays are required to understand the effect of these types of tubulin mutations in the cellular and molecular context. The pathogenic mutations predicted most poorly tend to involve substitutions between similar amino acids, with hydrophobicity and charges largely maintained between wild-type and mutant residues. This observation likely reflects the fact that most predictors incorporate some manner of amino acid substation matrix, thus predicting more minor effects for variants that do not change a residue’s charge or hydrophobic state. In contrast to more drastic muta- tions that tend to be better predicted, these mutations may be less detrimental to folding and microtubule assembly. Indeed, this difference is likely to explain why we observed a weaker co-localisation to the mitotic spindle in the TUBα1 V353R mutant compared to V353I. In this study, the performance of supervised predictors, and especially metapredictors like REVEL and M-CAP, highlights the issue of circularity present in these types of VEPs, which are typically trained using published pathogenic and benign variants [70]. There is a strong likelihood that some of these mutations are in training sets for these predictors. These biases would lead metapredictors and supervised methods to predict mutations in isotypes linked with neurodeve- lopmental diseases as pathogenic correctly, while mutations in other isotypes would be predicted poorly. This may explain why metapredictors performed so much worse on such mutations, espe- cially when compared to unsupervised methods like DeepSequence and PROVEAN. We judge DeepSequence to be the best performing VEP, considering consistency across isotypes and phenotypes. Unlike supervised methods, it does not use a labelled training set. DeepSequence can also make predictions based on the evolutionary conservation of entire sequences at once as opposed to no more than a few sites of interest, setting it apart from many other VEPs [65]. Benchmarking several VEPs against deep mutational scanning data also highlighted DeepSequence as the best pathogenicity predictor [62]. We, therefore, recom- mend using DeepSequence over metapredictors like REVEL and M-CAP to achieve more accurate predictions of tubulin variant pathogenicity. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 14 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations This study presents the tubulin family as a blind spot for current phenotype predictors, especially in identifying mutations that result in non-neurodevelopmental pathogenic pheno- types. Predictor biases due to their training sets introduce circularity and could limit predic- tions. The diversity of pathogenic phenotypes necessitates further study to understand the molecular mechanisms underlying their pathogenicity. Experimental approaches to dissect the effect of these tubulin pathogenic mutations on microtubule dynamics, microtubule assembly and microtubule interactions with associated proteins would contribute to a better molecular understanding of tubulinopathies [71–73]. These additional data could refine current pheno- type predictors and increase their performance. For non-neurodevelopmental phenotypes, pathogenic effects initially appeared to be iso- type-specific, but this could be linked to where these isotypes are expressed and to what level, or other unknown factors. For example, pathogenic mutations in TUBβ8 are associated to female infertility, but males carrying these mutations would presumably present a healthy phe- notype [22]. Importantly, many tubulin mutations are currently reported as linked to a disease or phenotype, but the evidence supporting causality is limited. It will be important to demon- strate whether the mutation is causative of pathogenesis or a consequence of other changes to cell homeostasis or genome stability. A pragmatic experimental approach is therefore essential to define the mechanism of pathogenesis associated with tubulin mutations. It will then be pos- sible to develop strategies to treat tubulin-related disorders using a personalised medicine approach. Methods Mutation datasets A thorough literature search was first performed to identify as many pathogenic or likely path- ogenic missense mutations across the tubulin family as possible. While the recently released tubulin mutation database [39]–as well as others such as OMIM [74] and ClinVar [38]–were excellent starting points, mutations from other publications (especially ones reported recently) were included (all detailed in Table 1). The literature search also uncovered a few other muta- tions that were incompletely dominant, homozygous, or of an unknown heritability, but these were omitted from the study. Any mutations described as de novo were assumed to be hetero- zygous and hence also dominant. As the underlying mechanisms with pathogenic mutations leading to cancer are likely to be more complex (and distinct), somatic mutations linked with such phenotypes were excluded. A set of putatively benign missense mutations for each isotype was retrieved from gnomAD v2.1 [40]. We excluded mutations present in both pathogenic and gnomAD datasets and considered only isotypes with more than one pathogenic mutation. Protein structural analyses Three-dimensional models of tubulin structures were taken from the protein data bank (PDB) on 2020.05.27, using the first biological assembly from each entry to represent its quaternary structure. We searched for polypeptide chains with at least 70% sequence identity to human tubulin isotype over a stretch of at least 50 amino acid residues. This allowed structural analysis using related structures for some isotypes for which structures are not available. Importantly, we only considered those residues where the amino acid in the structure, as well as both adja- cent residues, are the same as in the sequence of the isotype of interest. Unless otherwise stated, all structures that satisfy this threshold were considered. A hierarchy was implemented to determine where a residue occurred on the tubulin structure concerning its function (Fig B in S1 Text). First, we looked through all structures to highlight residues interacting with GTP (or its analogues). Then, model structures were used (S2 Table) to determine whether the PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 15 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations remaining residues occur on interfaces between the corresponding subunit in the tubulin dimer or with MAPs and other tubulin dimers in the microtubule in that specified order. Resi- dues not present on any interface were mapped to the highest resolution structure within the dataset. These mappings were used to classify residues according to their location, separating residues on the surface (relative solvent accessible surface area >0.25) from ones at the protein interior. Once every residue was labelled, their distribution was compared between mutations in our pathogenic and gnomAD datasets. Finally, all mutations in tubulin-α or -β were consid- ered before moving on to specific isotypes with at least 5 identified pathogenic missense muta- tions, also using PyMOL [75] to confirm that these patterns could also be observed qualitatively. To facilitate further analysis of tubulin mutations, we have provided a list of all amino acids in tubulin-α and-β that are involved in protein-protein or protein-ligand interac- tions (S6 Table). Structural predictions were then obtained from FoldX [60] for pathogenic and gnomAD variants. Each mutation was mapped onto the highest resolution structure from the dataset above. Here, theΔΔG values were calculated considering the monomer only, as well as the whole biological assembly, in the same manner as previously described [37]. All FoldX pre- dictedΔΔG values are provided in S7 Table. Computational variant effect predictions Scores from VEPs were generated using the same pipeline as described previously [62]. Though most of the methods used failed to produce scores for the entire combined dataset, the percentage of missing scores from each predictor did not exceed 7.8%, except for S3D-PROF and SNPs&GO-3D, which required PDB structures as inputs. For S3D-PROF and SNPs&GO- 3D, structures with the closest sequence identity were used, which have the following PDB IDs: 5iyz (TUBα1A), 5lxt (TUBα4A), 4tv9 (TUBβ1), 5nm5 (TUBβ2A and TUBβ8), 4i4t (TUBβ2B), 4lnu (TUBβ3), 5jqg (TUBβ4A and TUBβ4B) and 4zol (TUBβ5). For plotting ROC curves, mutations from the pathogenic dataset were labelled as true posi- tives, with gnomAD variants marked as true negatives. The plots themselves, as well as the AUC, were calculated using the PRROC package in R. Due to predictors PROVEAN, DeepSe- quence, SIFT4G and BLOSUM operating via inverse metrics (i.e. variants predicted to be more damaging have lower scores), all scores were multiplied by -1 before plotting curves and calculating AUC values. Finally, ROC AUC values were calculated globally across the whole dataset for each VEP first before generating specific scores of each VEP for individual isotypes. To calculate the MCD rankings, the datasets containing prediction scores for gnomAD and pathogenic mutations were combined. Then, the cumulative distribution function was applied to the collective scores obtained from each predictor using the cume_dist function from the DPLYR package on a per-VEP basis. Each value represents the proportion of scores that are less than or equal to the one obtained for that particular mutation from a specific VEP. Then, the cumulative distribution scores from all VEPs were averaged for each mutation to generate the MCD score. Molecular cloning The tubulin genes human TUBα1B (NM_006082.3) or TUBβ8 Class VIII (NM_177987.2) were synthesised by Genewiz. The TUBα1B gene were cloned into pBABE- blasticidin con- taining an N-terminal mCherry tag [76]. An additional linker region (with sequence 5’- ATGAGGAGGGGCGCTGCCGATAGGGAAACTGAGAGGCTCCCCGGCGCA- CAAGGTCCGTGCAGTGCGGTCAGTGCGGCCAGCTCCACATTGGCCGCAG- PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 16 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations CAGCGGCCCCTCGTGCTCGGGCGACCGCTGCCGCGTCCACCCTCAGCGC- CACCGCCCTCGAG-3’) was also inserted between the N-terminal mCherry tag and the start of the tubulin sequence. The TUBβ8 gene was cloned into a pBabe-blasticidin contanining a C-terminal GFP. Mutagenesis was performed using the Agilent Quickchange II mutagenesis kit, as per the manufacturer’s instructions. While the TUBα1B sequence was used rather than TUBα1A, where the V353I disease mutations has been identified, these are nearly identical, with only two minor amino acid sequence differences between them (G versus S at residue 232 and T versus S at residue 340), neither being close to V353 in the three-dimensional structure. Cell biology and fluorescence imaging HeLa cells were cultured and maintained in a humidified 5% CO atmosphere at 37˚C in DMEM supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco) and 10% Fetal Bovine Serum (Gibco). Cells were plated on 35 mm glass-bottom microwell dishes (ibidi) for live-cell imaging. Lipofectamine 3000 reagent (Invitrogen) was used for transient transfec- tions according to manufacturer guidelines. 24–48 hours after transfection, cells were incu- bated with 20–50 nM SiR-tubulin dye (SpiroChrome) for 1–2 hours and transferred to L-15 medium (ThermoFisher Scientific) before imaging. Images were taken using a Nikon Ti2 live imaging microscope were stored on OMERO [77], which was used to manually measure spin- dle length and width. ImageJ was then used to process the acquired images [78]. A region of interest was drawn over the spindle length observed through the Cy5 channel (visualising SiR- tubulin), and intensity profiles were generated for this channel and the GFP or mCherry chan- nel (depending on the construct). These profiles were then background subtracted using the average of the intensity profiles of two thinner lines on either side of the spindle. For each cell, a two-sided Spearman’s correlation coefficient was calculated between the fluorescent and SiR-tubulin intensity profile. OMERO was also used to calculate spindle length and width, where the length was divided by width to obtain calculate the spindle aspect ratio. Plots were visualised using GraphPad Prism 9, which was also used to carry out Kruskal-Wallis tests (with post-hoc Dunn). Supporting information S1 Text. Supplemental figures. File containing additional figures complementary to the anal- ysis shown in the main text of this manuscript. (DOCX) S1 Table. List of tubulin mutations. Full description of all pathogenic mutations compiled in this study (including phenotypes and references) as well as a list of all gnomAD mutations used. (XLSX) S2 Table. Structures used for location assignments. All model structures used for the hierar- chy implemented for assigning structural locations for tubulin residues. (CSV) S3 Table. Variant effect predictions. Complete set of raw predictions obtained from all the variant effect predictors used in this study for every tubulin mutation. (CSV) S4 Table. DeepSequence outputs. DeepSequence predictions for every possible missense mutation for all tubulin isotypes considered in this study. (CSV) PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 17 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations S5 Table. Mean cumulative distribution scores. Ranking of mean cumulative distribution (MCD) scores for all pathogenic tubulin mutations. (CSV) S6 Table. Tubulin residue interactions. All tubulin residues found to be interacting with GTP (or its analogues), MAPs, or with other tubulin subunits (interdimer or intradimer). (CSV) S7 Table. Predicted effects on protein stability for tubulin mutations. AllΔΔG values pre- dicted using FoldX using both the structure of the monomeric subunit (ddG_fold) and the full biological assembly (ddG_full). (CSV) S1 Data. Multiple sequence alignment. Multiple sequence alignment of all tubulin-α andβ isotypes considered in this study. (PDF) Acknowledgments We thank Benjamin Livesey for help with generating the VEP outputs. Author Contributions Conceptualization: Julie P. I. Welburn, Joseph A. Marsh. Data curation: Joseph A. Marsh. Formal analysis: Thomas J. Attard. Investigation: Thomas J. Attard. Supervision: Julie P. I. Welburn, Joseph A. Marsh. Writing – original draft: Thomas J. Attard. Writing – review & editing: Julie P. I. Welburn, Joseph A. Marsh. References 1. Brouhard GJ, Rice LM. Microtubule dynamics: an interplay of biochemistry and mechanics. Nat Rev Mol Cell Bio. 2018; 19(7):451–63. https://doi.org/10.1038/s41580-018-0009-y PMID: 29674711 2. Janke C, Magiera MM. The tubulin code and its role in controlling microtubule properties and functions. Nat Rev Mol Cell Biol. 2020; 21(6):307–26. https://doi.org/10.1038/s41580-020-0214-3 PMID: 3. Mitchison T, Kirschner M. Dynamic Instability of Microtubule Growth. Nature. 1984; 312(5991):237– 42. https://doi.org/10.1038/312237a0 PMID: 6504138 4. Battaje RR, Panda D. Lessons from bacterial homolog of tubulin, FtsZ for microtubule dynamics. Endocr Relat Cancer. 2017; 24(9):T1–T21. https://doi.org/10.1530/ERC-17-0118 PMID: 28634179 5. Dyer N. Tubulin and its prokaryotic homologue FtsZ: a structural and functional comparison. Sci Prog. 2009; 92(Pt 2):113–37. https://doi.org/10.3184/003685009X461431 PMID: 19697711 6. Findeisen P, Muhlhausen S, Dempewolf S, Hertzog J, Zietlow A, Carlomagno T, et al. Six subgroups and extensive recent duplications characterize the evolution of the eukaryotic tubulin protein family. Genome Biol Evol. 2014; 6(9):2274–88. https://doi.org/10.1093/gbe/evu187 PMID: 25169981 7. Lowe J, Li H, Downing KH, Nogales E. Refined structure of alpha beta-tubulin at 3.5 A resolution. J Mol Biol. 2001; 313(5):1045–57. https://doi.org/10.1006/jmbi.2001.5077 PMID: 11700061 8. Kobayashi T. Dephosphorylation of tubulin-bound guanosine triphosphate during microtubule assem- bly. J Biochem. 1975; 77(6):1193–7. PMID: 1225903 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 18 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 9. Mitchison TJ. Localization of an Exchangeable Gtp-Binding Site at the Plus End of Microtubules. Sci- ence. 1993; 261(5124):1044–7. https://doi.org/10.1126/science.8102497 PMID: 8102497 10. Wang Z, Sheetz MP. The C-terminus of tubulin increases cytoplasmic dynein and kinesin processivity. Biophys J. 2000; 78(4):1955–64. https://doi.org/10.1016/S0006-3495(00)76743-9 PMID: 10733974 11. Burns RG, Surridge C. Analysis of beta-tubulin sequences reveals highly conserved, coordinated amino acid substitutions. Evidence that these ’hot spots’ are directly involved in the conformational change required for dynamic instability. FEBS Lett. 1990; 271(1–2):1–8. https://doi.org/10.1016/0014- 5793(90)80359-q PMID: 2226794 12. Chakraborti S, Natarajan K, Curiel J, Janke C, Liu J. The emerging role of the tubulin code: From the tubulin molecule to neuronal function and disease. Cytoskeleton. 2016; 73(10):521–50. https://doi.org/ 10.1002/cm.21290 PMID: 26934450 13. Kollman JM, Merdes A, Mourey L, Agard DA. Microtubule nucleation by gamma-tubulin complexes. Nat Rev Mol Cell Biol. 2011; 12(11):709–21. 14. Stathatos GG, Dunleavy JEM, Zenker J, O’Bryan MK. Delta and epsilon tubulin in mammalian devel- opment. Trends in Cell Biology. 2021; 31(9):774–87. https://doi.org/10.1016/j.tcb.2021.03.010 PMID: 15. Bahi-Buisson N, Poirier K, Fourniol F, Saillour Y, Valence S, Lebrun N, et al. The wide spectrum of tubulinopathies: what are the key features for the diagnosis? Brain. 2014; 137:1676–700. https://doi. org/10.1093/brain/awu082 PMID: 24860126 16. Brock S, Stouffs K, Scalais E, D’Hooghe M, Keymolen K, Guerrini R, et al. Tubulinopathies continued: refining the phenotypic spectrum associated with variants in TUBG1. Eur J Hum Genet. 2018; 26 (8):1132–42. https://doi.org/10.1038/s41431-018-0146-y PMID: 29706637 17. Poirier K, Lebrun N, Broix L, Tian G, Saillour Y, Boscheron C, et al. Mutations in TUBG1, DYNC1H1, KIF5C and KIF2A cause malformations of cortical development and microcephaly. Nat Genet. 2013; 45(6):639–+. https://doi.org/10.1038/ng.2613 PMID: 23603762 18. Luscan R, Mechaussier S, Paul A, Tian GL, Gerard X, Defoort-Dellhemmes S, et al. Mutations in TUBB4B Cause a Distinctive Sensorineural Disease. Am J Hum Genet. 2017; 101(6):1006–12. https://doi.org/10.1016/j.ajhg.2017.10.010 PMID: 29198720 19. Smith BN, Ticozzi N, Fallini C, Gkazi AS, Topp S, Kenna KP, et al. Exome-wide Rare Variant Analysis Identifies TUBA4A Mutations Associated with Familial ALS. Neuron. 2014; 84(2):324–31. https://doi. org/10.1016/j.neuron.2014.09.027 PMID: 25374358 20. Stoupa A, Adam F, Kariyawasam D, Strassel C, Gawade S, Szinnai G, et al. TUBB1 mutations cause thyroid dysgenesis associated with abnormal platelet physiology. Embo Mol Med. 2018; 10(12). https://doi.org/10.15252/emmm.201809569 PMID: 30446499 21. Bhagwat S, Dalvi V, Chandrasekhar D, Matthew T, Acharya K, Gajbhiye R, et al. Acetylated alpha- tubulin is reduced in individuals with poor sperm motility. Fertil Steril. 2014; 101(1):95–104 e3. 22. Chen BB, Wang WJ, Peng XD, Jiang HF, Zhang SZ, Li D, et al. The comprehensive mutational and phenotypic spectrum of TUBB8 in female infertility. Eur J Hum Genet. 2019; 27(2):300–7. https://doi. org/10.1038/s41431-018-0283-3 PMID: 30297906 23. Feng RZ, Yan Z, Li B, Yu M, Sang Q, Tian GL, et al. Mutations in TUBB8 cause a multiplicity of pheno- types in human oocytes and early embryos. J Med Genet. 2016; 53(10):662–71. https://doi.org/10. 1136/jmedgenet-2016-103891 PMID: 27273344 24. Abdollahi MR, Morrison E, Sirey T, Molnar Z, Hayward BE, Carr IM, et al. Mutation of the Variant alpha-Tubulin TUBA8 Results in Polymicrogyria with Optic Nerve Hypoplasia. Am J Hum Genet. 2009; 85(5):737–44. https://doi.org/10.1016/j.ajhg.2009.10.007 PMID: 19896110 25. Tian G, Kong XP, Jaglin XH, Chelly J, Keays D, Cowan NJ. A pachygyria-causing alpha-tubulin muta- tion results in inefficient cycling with CCT and a deficient interaction with TBCB. Mol Biol Cell. 2008; 19 (3):1152–61. https://doi.org/10.1091/mbc.e07-09-0861 PMID: 18199681 26. Backwell L, Marsh JA. Diverse Molecular Mechanisms Underlying Pathogenic Protein Mutations: Beyond the Loss-of-Function Paradigm. Annual Review of Genomics and Human Genetics. 2022; 23 (1):null. https://doi.org/10.1146/annurev-genom-111221-103208 PMID: 35395171 27. Aiken J, Buscaglia G, Aiken AS, Moore JK, Bates EA. Tubulin mutations in brain development disor- ders: Why haploinsufficiency does not explain TUBA1A tubulinopathies. Cytoskeleton (Hoboken). 2020; 77(3–4):40–54. https://doi.org/10.1002/cm.21567 PMID: 31574570 28. Breuss M, Heng JIT, Poirier K, Tian GL, Jaglin XH, Qu ZD, et al. Mutations in the beta-Tubulin Gene TUBB5 Cause Microcephaly with Structural Brain Abnormalities. Cell Rep. 2012; 2(6):1554–62. 29. Cederquist GY, Luchniak A, Tischfield MA, Peeva M, Song Y, Menezes MP, et al. An inherited TUBB2B mutation alters a kinesin-binding site and causes polymicrogyria, CFEOM and axon PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 19 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations dysinnervation. Hum Mol Genet. 2012; 21(26):5484–99. https://doi.org/10.1093/hmg/dds393 PMID: 30. Cushion TD, Paciorkowski AR, Pilz DT, Mullins JGL, Seltzer LE, Marion RW, et al. De Novo Mutations in the Beta-Tubulin Gene TUBB2A Cause Simplified Gyral Patterning and Infantile-Onset Epilepsy. Am J Hum Genet. 2014; 94(4):634–41. https://doi.org/10.1016/j.ajhg.2014.03.009 PMID: 24702957 31. Ti SC, Pamula MC, Howes SC, Duellberg C, Cade NI, Kleiner RE, et al. Mutations in Human Tubulin Proximal to the Kinesin-Binding Site Alter Dynamic Instability at Microtubule Plus- and Minus-Ends. Dev Cell. 2016; 37(1):72–84. https://doi.org/10.1016/j.devcel.2016.03.003 PMID: 27046833 32. Tischfield MA, Baris HN, Wu C, Rudolph G, Van Maldergem L, He W, et al. Human TUBB3 Mutations Perturb Microtubule Dynamics, Kinesin Interactions, and Axon Guidance. Cell. 2010; 140(1):74–87. https://doi.org/10.1016/j.cell.2009.12.011 PMID: 20074521 33. Niwa S, Takahashi H, Hirokawa N. beta-Tubulin mutations that cause severe neuropathies disrupt axonal transport. Embo J. 2013; 32(10):1352–64. 34. Park K, Hoff KJ, Wethekam L, Stence N, Saenz M, Moore JK. Kinetically Stabilizing Mutations in Beta Tubulins Create Isotype-Specific Brain Malformations. Front Cell Dev Biol. 2021; 9:765992. https:// doi.org/10.3389/fcell.2021.765992 PMID: 34869359 35. Hebebrand M, Huffmeier U, Trollmann R, Hehr U, Uebe S, Ekici AB, et al. The mutational and pheno- typic spectrum of TUBA1A-associated tubulinopathy. Orphanet J Rare Dis. 2019; 14. https://doi.org/ 10.1186/s13023-019-1020-x PMID: 30744660 36. Livesey BJ, Marsh JA. Interpreting protein variant effects with computational predictors and deep mutational scanning. Disease Models & Mechanisms. 2022; 15(6). https://doi.org/10.1242/dmm. 049510 PMID: 35736673 37. Gerasimavicius L, Livesey BJ, Marsh JA. Loss-of-function, gain-of-function and dominant-negative mutations have profoundly different effects on protein structure. Nat Commun. 2022; 13(1):3895. https://doi.org/10.1038/s41467-022-31686-6 PMID: 35794153 38. Landrum MJ, Lee JM, Benson M, Brown GR, Chao C, Chitipiralla S, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res. 2018; 46(D1):D1062–D7. https:// doi.org/10.1093/nar/gkx1153 PMID: 29165669 39. Pham CL, Morrissette NS. The tubulin mutation database: A resource for the cytoskeleton community. Cytoskeleton. 2019; 76(2):186–91. https://doi.org/10.1002/cm.21514 PMID: 30667171 40. Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alfo ¨ ldi J, Wang Q, et al. Variation across 141,456 human exomes and genomes reveals the spectrum of loss-of-function intolerance across human pro- tein-coding genes. bioRxiv. 2019:531210. 41. Huzil JT, Chen K, Kurgan L, Tuszynski JA. The roles of beta-tubulin mutations and isotype expression in acquired drug resistance. Cancer Inform. 2007; 3:159–81. PMID: 19455242 42. Nami B, Wang ZX. Genetics and Expression Profile of the Tubulin Gene Superfamily in Breast Cancer Subtypes and Its Relation to Taxane Resistance. Cancers. 2018; 10(8). https://doi.org/10.3390/ cancers10080274 PMID: 30126203 43. Parker AL, Teo WS, McCarroll JA, Kavallaris M. An Emerging Role for Tubulin Isotypes in Modulating Cancer Biology and Chemotherapy Resistance. Int J Mol Sci. 2017; 18(7). https://doi.org/10.3390/ ijms18071434 PMID: 28677634 44. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016; 536(7616):285–91. https://doi.org/10.1038/ nature19057 PMID: 27535533 45. Burley K, Westbury SK, Mumford AD. TUBB1 variants and human platelet traits. Platelets. 2018; 29 (2):209–11. https://doi.org/10.1080/09537104.2017.1411587 PMID: 29333906 46. Kunishima S, Kobayashi R, Itoh TJ, Hamaguchi M, Saito H. Mutation of the beta 1-tubulin gene associ- ated with congenital macrothrombocytopenia affecting microtubule assembly. Blood. 2009; 113 (2):458–61. https://doi.org/10.1182/blood-2008-06-162610 PMID: 18849486 47. Kunishima S, Nishimura S, Suzuki H, Imaizumi M, Saito H. TUBB1 mutation disrupting microtubule assembly impairs proplatelet formation and results in congenital macrothrombocytopenia. Eur J Hae- matol. 2014; 92(4):276–82. https://doi.org/10.1111/ejh.12252 PMID: 24344610 48. Lecine P, Italiano JE Jr., Kim SW, Villeval JL, Shivdasani RA. Hematopoietic-specific beta 1 tubulin participates in a pathway of platelet biogenesis dependent on the transcription factor NF-E2. Blood. 2000; 96(4):1366–73. PMID: 10942379 49. Wang D, Villasante A, Lewis SA, Cowan NJ. The mammalian beta-tubulin repertoire: hematopoietic expression of a novel, heterologous beta-tubulin isotype. J Cell Biol. 1986; 103(5):1903–10. https:// doi.org/10.1083/jcb.103.5.1903 PMID: 3782288 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 20 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 50. Chen BB, Li B, Li D, Yan Z, Mao XY, Xu Y, et al. Novel mutations and structural deletions in TUBB8: expanding mutational and phenotypic spectrum of patients with arrest in oocyte maturation, fertiliza- tion or early embryonic development. Hum Reprod. 2017; 32(2):457–64. https://doi.org/10.1093/ humrep/dew322 PMID: 27989988 51. Feng RZ, Sang Q, Kuang YP, Sun XX, Yan Z, Zhang SZ, et al. Mutations in TUBB8 and Human Oocyte Meiotic Arrest. New Engl J Med. 2016; 374(3):223–32. https://doi.org/10.1056/ NEJMoa1510791 PMID: 26789871 52. Huang LL, Tong XH, Luo LH, Zheng SX, Jin RT, Fu YY, et al. Mutation analysis of the TUBB8 gene in nine infertile women with oocyte maturation arrest. Reprod Biomed Online. 2017; 35(3):305–10. https://doi.org/10.1016/j.rbmo.2017.05.017 PMID: 28652098 53. Xiang JJ, Wang W, Qian CF, Xue JY, Wang T, Li HB, et al. Human oocyte maturation arrest caused by a novel missense mutation in TUBB8. J Int Med Res. 2018; 46(9):3759–64. https://doi.org/10.1177/ 0300060518778638 PMID: 29877102 54. Levy ED. A Simple Definition of Structural Regions in Proteins and Its Use in Analyzing Interface Evo- lution. J Mol Biol. 2010; 403(4):660–70. https://doi.org/10.1016/j.jmb.2010.09.028 PMID: 20868694 55. Cushion TD, Dobyns WB, Mullins JGL, Stoodley N, Chung SK, Fry AE, et al. Overlapping cortical mal- formations and mutations in TUBB2B and TUBA1A. Brain. 2013; 136:536–48. https://doi.org/10.1093/ brain/aws338 PMID: 23361065 56. Erro R, Hersheson J, Ganos C, Mencacci NE, Stamelou M, Batla A, et al. H-ABC syndrome and DYT4: Variable expressivity or pleiotropy of TUBB4 mutations? Movement Disord. 2015; 30(6):828– 33. https://doi.org/10.1002/mds.26129 PMID: 25545912 57. Romaniello R, Arrigoni F, Panzeri E, Poretti A, Micalizzi A, Citterio A, et al. Tubulin-related cerebellar dysplasia: definition of a distinct pattern of cerebellar malformation (vol 27, pg 5080, 2017). Eur Radiol. 2017; 27(12):5093–. 58. Rice LM, Montabana EA, Agard DA. The lattice as allosteric effector: Structural studies of alpha beta- and gamma-tubulin clarify the role of GTP in microtubule assembly. P Natl Acad Sci USA. 2008; 105 (14):5378–83. 59. Wieczorek M, Urnavicius L, Ti SC, Molloy KR, Chait BT, Kapoor TM. Asymmetric Molecular Architec- ture of the Human gamma-Tubulin Ring Complex. Cell. 2020; 180(1):165–+. 60. Guerois R, Nielsen JE, Serrano L. Predicting changes in the stability of proteins and protein com- plexes: A study of more than 1000 mutations. J Mol Biol. 2002; 320(2):369–87. https://doi.org/10. 1016/S0022-2836(02)00442-4 PMID: 12079393 61. Gerasimavicius L, Liu X, Marsh JA. Identification of pathogenic missense mutations using protein sta- bility predictors. Sci Rep. 2020; 10(1):15387. https://doi.org/10.1038/s41598-020-72404-w PMID: 62. Livesey BJ, Marsh JA. Using deep mutational scanning to benchmark variant effect predictors and identify disease mutations. Molecular Systems Biology. 2020; 16(7):e9380. https://doi.org/10.15252/ msb.20199380 PMID: 32627955 63. Fallet-Bianco C, Laquerriere A, Poirier K, Razavi F, Guimiot F, Dias P, et al. Mutations in tubulin genes are frequent causes of various foetal malformations of cortical development including microlissence- phaly. Acta Neuropathol Com. 2014;2. https://doi.org/10.1186/2051-5960-2-69 PMID: 25059107 64. Oegema R, Cushion TD, Phelps IG, Chung SK, Dempsey JC, Collins S, et al. Recognizable cerebellar dysplasia associated with mutations in multiple tubulin genes. Hum Mol Genet. 2015; 24(18):5313–25. https://doi.org/10.1093/hmg/ddv250 PMID: 26130693 65. Riesselman AJ, Ingraham JB, Marks DS. Deep generative models of genetic variation capture the effects of mutations. Nat Methods. 2018; 15(10):816–+. https://doi.org/10.1038/s41592-018-0138-4 PMID: 30250057 66. Romaniello R, Zucca C, Arrigoni F, Bonanni P, Panzeri E, Bassi MT, et al. Epilepsy in Tubulinopathy: Personal Series and Literature Review. Cells. 2019; 8(7). https://doi.org/10.3390/cells8070669 PMID: 67. Schatz PJ, Solomon F, Botstein D. Isolation and Characterization of Conditional-Lethal Mutations in the Tub1 Alpha-Tubulin Gene of the Yeast Saccharomyces-Cerevisiae. Genetics. 1988; 120(3):681– 95. https://doi.org/10.1093/genetics/120.3.681 PMID: 3066684 68. Livesey BJ, Marsh JA. The properties of human disease mutations at protein interfaces. PLOS Computational Biology. 2022; 18(2):e1009858. https://doi.org/10.1371/journal.pcbi.1009858 PMID: 69. Hausrat TJ, Radwitz J, Lombino FL, Breiden P, Kneussel M. Alpha- and beta-tubulin isotypes are dif- ferentially expressed during brain development. Dev Neurobiol. 2021; 81(3):333–50. https://doi.org/ 10.1002/dneu.22745 PMID: 32293117 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 21 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 70. Grimm DG, Azencott CA, Aicheler F, Gieraths U, MacArthur DG, Samocha KE, et al. The Evaluation of Tools Used to Predict the Impact of Missense Variants Is Hindered by Two Types of Circularity. Hum Mutat. 2015; 36(5):513–23. https://doi.org/10.1002/humu.22768 PMID: 25684150 71. Krajka V, Vulinovic F, Genova M, Tanzer K, Jijumon AS, Bodakuntla S, et al. H-ABC- and dystonia- causing TUBB4A mutations show distinct pathogenic effects. Sci Adv. 2022; 8(10):eabj9229. https:// doi.org/10.1126/sciadv.abj9229 PMID: 35275727 72. Leca I, Phillips AW, Hofer I, Landler L, Ushakova L, Cushion TD, et al. A proteomic survey of microtu- bule-associated proteins in a R402H TUBA1A mutant mouse. Plos Genet. 2020; 16(11):e1009104. https://doi.org/10.1371/journal.pgen.1009104 PMID: 33137126 73. Ti SC, Alushin GM, Kapoor TM. Human beta-Tubulin Isotypes Can Regulate Microtubule Protofila- ment Number and Stability. Dev Cell. 2018; 47(2):175–90 e5. 74. Online Mendelian Inheritance in Man, OMIM. [Internet]. McKusick-Nathans Institute of Genetic Medi- cine, Johns Hopkins University (Baltimore, MD). 2019. Available from: https://omim.org/. 75. Schrodinger LLC. The PyMOL Molecular Graphics System, Version 2.3.2. 2015. 76. Cheeseman IM, Desai A. A combined approach for the localization and tandem affinity purification of protein complexes from metazoans. Science’s STKE: signal transduction knowledge environment. 2005; 2005(266):pl1. https://doi.org/10.1126/stke.2662005pl1 PMID: 15644491 77. Goldberg IG, Allan C, Burel J-M, Creager D, Falconi A, Hochheiser H, et al. The Open Microscopy Environment (OME) Data Model and XML file: open tools for informatics and quantitative analysis in biological imaging. Genome Biology. 2005; 6(5):R47. https://doi.org/10.1186/gb-2005-6-5-r47 PMID: 78. Rueden CT, Schindelin J, Hiner MC, DeZonia BE, Walter AE, Arena ET, et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinformatics. 2017; 18(1):529. https://doi.org/10. 1186/s12859-017-1934-z PMID: 29187165 79. Aiken J, Buscaglia G, Bates EA, Moore JK. The alpha-Tubulin gene TUBA1A in Brain Development: A Key Ingredient in the Neuronal Isotype Blend. J Dev Biol. 2017; 5(3). 80. Boissel S, Fallet-Bianco C, Chitayat D, Kremer V, Nassif C, Rypens F, et al. Genomic study of severe fetal anomalies and discovery of GREB1L mutations in renal agenesis. Genet Med. 2018; 20(7):745– 53. https://doi.org/10.1038/gim.2017.173 PMID: 29261186 81. Gardner JF, Cushion TD, Niotakis G, Olson HE, Grant PE, Scott RH, et al. Clinical and Functional Characterization of the Recurrent TUBA1A p.(Arg2His) Mutation. Brain Sci. 2018; 8(8). https://doi.org/ 10.3390/brainsci8080145 PMID: 30087272 82. Keays DA, Tian G, Poirier K, Huang GJ, Siebold C, Cleak J, et al. Mutations in alpha-tubulin cause abnormal neuronal migration in mice and lissencephaly in humans. Cell. 2007; 128(1):45–57. https:// doi.org/10.1016/j.cell.2006.12.017 PMID: 17218254 83. Poirier K, Keays DA, Francis F, Saillour Y, Bahi N, Manouvrier S, et al. Large spectrum of lissence- phaly and pachygyria phenotypes resulting from de novo missense mutations in tubulin alpha 1A (TUBA1A). Hum Mutat. 2007; 28(11):1055–64. https://doi.org/10.1002/humu.20572 PMID: 17584854 84. Martello LA, Verdier-Pinard P, Shen HJ, He LF, Torres K, Orr GA, et al. Elevated levels of microtubule destabilizing factors in a Taxol-resistant/dependent A549 cell line with an alpha-tubulin mutation. Can- cer Res. 2003; 63(6):1207–13. PMID: 12649178 85. Poruchynsky MS, Kim JH, Nogales E, Annable T, Loganzo F, Greenberger LM, et al. Tumor cells resistant to a microtubule-depolymerizing hemiasterlin analogue, HTI-286, have mutations in alpha- or beta-tubulin and increased microtubule stability. Biochemistry-Us. 2004; 43(44):13944–54. https://doi. org/10.1021/bi049300+ PMID: 15518543 86. Yang CPH, Verdier-Pinard P, Wang F, Lippaine-Horvath E, He LF, Li DS, et al. A highly epothilone B- resistant A549 cell line with mutations in tubulin that confer drug dependence. Mol Cancer Ther. 2005; 4(6):987–95. https://doi.org/10.1158/1535-7163.MCT-05-0024 PMID: 15956256 87. Wang J, Chen W, Wei WW, Lou JY. Oncogene TUBA1C promotes migration and proliferation in hepa- tocellular carcinoma and predicts a poor prognosis. Oncotarget. 2017; 8(56):96215–24. https://doi. org/10.18632/oncotarget.21894 PMID: 29221200 88. Alazami AM, Patel N, Shamseldin HE, Anazi S, Al-Dosari MS, Alzahrani F, et al. Accelerating Novel Candidate Gene Discovery in Neurogenetic Disorders via Whole-Exome Sequencing of Prescreened Multiplex Consanguineous Families. Cell Rep. 2015; 10(2):148–61. https://doi.org/10.1016/j.celrep. 2014.12.015 PMID: 25558065 89. Sferra A, Fattori F, Rizza T, Flex E, Bellacchio E, Bruselles A, et al. Defective kinesin binding of TUBB2A causes progressive spastic ataxia syndrome resembling sacsinopathy. Hum Mol Genet. 2018; 27(11):1892–904. https://doi.org/10.1093/hmg/ddy096 PMID: 29547997 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 22 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 90. Amrom D, Tanyalcin I, Verhelst H, Deconinck N, Brouhard GJ, Decarie JC, et al. Polymicrogyria with dysmorphic basal ganglia? Think tubulin! Clin Genet. 2014; 85(2):178–83. https://doi.org/10.1111/cge. 12141 PMID: 23495813 91. Guerrini R, Mei D, Cordelli DM, Pucatti D, Franzoni E, Parrini E. Symmetric polymicrogyria and pachy- gyria associated with TUBB2B gene mutations. Eur J Hum Genet. 2012; 20(9):995–8. https://doi.org/ 10.1038/ejhg.2012.21 PMID: 22333901 92. Jaglin XH, Poirier K, Saillour Y, Buhler E, Tian G, Bahi-Buisson N, et al. Mutations in the beta-tubulin gene TUBB2B result in asymmetrical polymicrogyria. Nat Genet. 2009; 41(6):746–52. https://doi.org/ 10.1038/ng.380 PMID: 19465910 93. MacKinnon S, Oystreck DT, Andrews C, Chan WM, Hunter DG, Engle EC. Diagnostic Distinctions and Genetic Analysis of Patients Diagnosed with Moebius Syndrome. Ophthalmology. 2014; 121(7):1461– 8. https://doi.org/10.1016/j.ophtha.2014.01.006 PMID: 24612975 94. Minoura I, Takazaki H, Ayukawa R, Saruta C, Hachikubo Y, Uchimura S, et al. Reversal of axonal growth defects in an extraocular fibrosis model by engineering the kinesin-microtubule interface. Nat Commun. 2016; 7. 95. Arai-Ichinoi N, Uematsu M, Sato R, Suzuki T, Kudo H, Kikuchi A, et al. Genetic heterogeneity in 26 infants with a hypomyelinating leukodystrophy. Hum Genet. 2016; 135(1):89–98. https://doi.org/10. 1007/s00439-015-1617-7 PMID: 26597493 96. Blumkin L, Halevy A, Ben-Ami-Raichman D, Dahari D, Haviv A, Sarit C, et al. Expansion of the spec- trum of TUBB4A-related disorders: a new phenotype associated with a novel mutation in the TUBB4A gene. Neurogenetics. 2014; 15(2):107–13. https://doi.org/10.1007/s10048-014-0392-2 PMID: 97. Carvalho D, Santos S, Martins B, Marques FP. TUBB4A novel mutation reinforces the genotype-phe- notype correlation of hypomyelination with atrophy of the basal ganglia and cerebellum. Brain. 2015;138. https://doi.org/10.1093/brain/awu242 PMID: 25168210 98. Curiel J, Bey GR, Takanohashi A, Bugiani M, Fu XQ, Wolf NI, et al. TUBB4A mutations result in spe- cific neuronal and oligodendrocytic defects that closely match clinically distinct phenotypes. Hum Mol Genet. 2017; 26(22):4506–18. https://doi.org/10.1093/hmg/ddx338 PMID: 28973395 99. Ferreira C, Poretti A, Cohen J, Hamosh A, Naidu S. Novel TUBB4A mutations and expansion of the neuroimaging phenotype of hypomyelination with atrophy of the basal ganglia and cerebellum (H- ABC). Am J Med Genet A. 2014; 164(7):1802–7. https://doi.org/10.1002/ajmg.a.36526 PMID: 100. Hamilton EM, Polder E, Vanderver A, Naidu S, Schiffmann R, Fisher K, et al. Hypomyelination with atrophy of the basal ganglia and cerebellum: further delineation of the phenotype and genotype-phe- notype correlation. Brain. 2014; 137:1921–30. https://doi.org/10.1093/brain/awu110 PMID: 24785942 101. Hersheson J, Mencacci NE, Davis M, MacDonald N, Trabzuni D, Ryten M, et al. Mutations in the auto- regulatory domain of beta-tubulin 4a cause hereditary dystonia. Ann Neurol. 2013; 73(4):546–53. 102. Isakov O, Lev D, Blumkin L, Celniker G, Leshinsky-Silver E, Shomron N. Crowdfunding Effort Identi- fies the Causative Mutation in a Patient with Nystagmus, Microcephaly, Dystonia and Hypomyelina- tion. J Genet Genomics. 2015; 42(2):79–81. https://doi.org/10.1016/j.jgg.2014.12.004 PMID: 103. Kancheva D, Chamova T, Guergueltcheva V, Mitev V, Azmanov DN, Kalaydjieva L, et al. Mosaic dom- inant TUBB4A mutation in an inbred family with complicated hereditary spastic paraplegia. Movement Disord. 2015; 30(6):854–8. https://doi.org/10.1002/mds.26196 PMID: 25772097 104. Lohmann K, Wilcox RA, Winkler S, Ramirez A, Rakovic A, Park JS, et al. Whispering dysphonia (DYT4 dystonia) is caused by a mutation in the TUBB4 gene. Ann Neurol. 2013; 73(4):537–45. https:// doi.org/10.1002/ana.23829 PMID: 23595291 105. Lu Y, Ondo Y, Shimojima K, Osaka H, Yamamoto T. A novel TUBB4A mutation G96R identified in a patient with hypomyelinating leukodystrophy onset beyond adolescence. Hum Genome Var. 2017; 4:17035. https://doi.org/10.1038/hgv.2017.35 PMID: 28791129 106. Miyatake S, Osaka H, Shiina M, Sasaki M, Takanashi J, Haginoya K, et al. Expanding the phenotypic spectrum of TUBB4A-associated hypomyelinating leukoencephalopathies. Neurology. 2014; 82 (24):2230–7. https://doi.org/10.1212/WNL.0000000000000535 PMID: 24850488 107. Pizzino A, Pierson TM, Guo Y, Helman G, Fortini S, Guerrero K, et al. TUBB4A de novo mutations cause isolated hypomyelination. Neurology. 2014; 83(10):898–902. https://doi.org/10.1212/WNL. 0000000000000754 PMID: 25085639 108. Purnell SM, Bleyl SB, Bonkowsky JL. Clinical exome sequencing identifies a novel TUBB4A mutation in a child with static hypomyelinating leukodystrophy. Pediatr Neurol. 2014; 50(6):608–11. https://doi. org/10.1016/j.pediatrneurol.2014.01.051 PMID: 24742798 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 23 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 109. Pyle A, Smertenko T, Bargiela D, Griffin H, Duff J, Appleton M, et al. Exome sequencing in undiag- nosed inherited and sporadic ataxias. Brain. 2015; 138:276–83. https://doi.org/10.1093/brain/awu348 PMID: 25497598 110. Sagnelli A, Magri S, Farina L, Chiapparini L, Marotta G, Tonduti D, et al. Early-onset progressive spas- tic paraplegia caused by a novel TUBB4A mutation: brain MRI and FDG-PET findings. J Neurol. 2016; 263(3):591–3. https://doi.org/10.1007/s00415-016-8020-8 PMID: 26810722 111. Simons C, Wolf NI, McNeil N, Caldovic L, Devaney JM, Takanohashi A, et al. A de novo mutation in the beta-tubulin gene TUBB4A results in the leukoencephalopathy hypomyelination with atrophy of the basal ganglia and cerebellum. Am J Hum Genet. 2013; 92(5):767–73. 112. Vanderver A, Simons C, Helman G, Crawford J, Wolf NI, Bernard G, et al. Whole Exome Sequencing in Patients with White Matter Abnormalities. Ann Neurol. 2016; 79(6):1031–7. https://doi.org/10.1002/ ana.24650 PMID: 27159321 113. Wang WW, Zhang HX, Wang XM, Patterson J, Winter P, Graham K, et al. Novel mutations involving beta I-, beta IIA-, or beta IVB-tubulin isotypes with functional resemblance to beta III-tubulin in breast cancer. Protoplasma. 2017; 254(3):1163–73. 114. Bowling KM, Thompson ML, Amaral MD, Finnila CR, Hiatt SM, Engel KL, et al. Genomic diagnosis for children with intellectual disability and/or developmental delay. Genome Med. 2017;9. 115. Isrie M, Breuss M, Tian GL, Hansen AH, Cristofoli F, Morandell J, et al. Mutations in Either TUBB or MAPRE2 Cause Circumferential Skin Creases Kunze Type. Am J Hum Genet. 2015; 97(6):790–800. https://doi.org/10.1016/j.ajhg.2015.10.014 PMID: 26637975 116. Leonard NJ. A second patient with MCA/MR syndrome with multiple circumferential skin creases. Am J Med Genet. 2002; 112(1):91–4. https://doi.org/10.1002/ajmg.10665 PMID: 12239728 117. Madrigal I, Rabionet R, Alvarez-Mora MI, Sanchez A, Rodriguez-Revenga L, Estivill X, et al. Spectrum of clinical heterogeneity of beta-tubulin TUBB5 gene mutations. Gene. 2019; 695:12–7. 118. Ulucan H, Koparir E, Koparir A, Karaca E, Emre R, Gezdirici A, et al. Circumferential skin folds and multiple anomalies: confirmation of a distinct autosomal recessive Michelin tire baby syndrome. Clin Dysmorphol. 2013; 22(2):87–90. https://doi.org/10.1097/MCD.0b013e32835cd5df PMID: 23324645 119. Wiszniewski W, Gawlinski P, Gambin T, Bekiesinska-Figatowska M, Obersztyn E, Antczak-Marach D, et al. Comprehensive genomic analysis of patients with disorders of cerebral cortical development. Eur J Hum Genet. 2018; 26(8):1121–31. https://doi.org/10.1038/s41431-018-0137-z PMID: 29706646 120. Fazeli W, Herkenrath P, Stiller B, Neugebauer A, Fricke J, Lang-Roth R, et al. A TUBB6 mutation is associated with autosomal dominant non-progressive congenital facial palsy, bilateral ptosis and velo- pharyngeal dysfunction. Hum Mol Genet. 2017; 26(20):4055–66. https://doi.org/10.1093/hmg/ddx296 PMID: 29016863 121. Leask A, Stearns T. Expression of amino- and carboxyl-terminal gamma- and alpha-tubulin mutants in cultured epithelial cells. J Biol Chem. 1998; 273(5):2661–8. https://doi.org/10.1074/jbc.273.5.2661 PMID: 9446570 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 24 / 24 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png PLoS Computational Biology Public Library of Science (PLoS) Journal

Understanding molecular mechanisms and predicting phenotypic effects of pathogenic tubulin mutations

Download PDF
24 pages

Loading...

Page 2

Loading...

Page 3

Loading...

Page 4

Loading...

Page 5

Loading...

Page 6

Loading...

Page 7

Loading...

Page 8

Loading...

Page 9

Loading...

Page 10

Loading...

Page 11

Loading...

Page 12

Loading...

Page 13

Loading...

Page 14

Loading...

Page 15

Loading...

Page 16

Loading...

Page 17

Loading...

Page 18

Loading...

Page 19

Loading...

Page 20

Loading...

Page 21

Loading...

Page 22

Loading...

Page 23

Loading...

Page 24

 
/lp/public-library-of-science-plos-journal/understanding-molecular-mechanisms-and-predicting-phenotypic-effects-JWpm0kfI2J

References (119)

Publisher
Public Library of Science (PLoS) Journal
Copyright
Copyright: © 2022 Attard et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are provided within the manuscript and its Supporting Information files. Funding: This work was supported by the Medical Research Council via a Precision Medicine Doctoral Training Programme studentship to TJA and a Career Development Award (MR/M02122X/1) to JAM. JAM is a Lister Institute Research Fellow. JPIW is supported by a Wellcome Trust Senior Research Fellowship (207430). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.
ISSN
1553-734X
eISSN
1553-7358
DOI
10.1371/journal.pcbi.1010611
Publisher site
See Article on Publisher Site

Abstract

Cells rely heavily on microtubules for several processes, including cell division and molecu- lar trafficking. Mutations in the different tubulin-α and -β proteins that comprise microtubules have been associated with various diseases and are often dominant, sporadic and congeni- OPENACCESS tal. While the earliest reported tubulin mutations affect neurodevelopment, mutations are Citation: Attard TJ, Welburn JPI, Marsh JA (2022) also associated with other disorders such as bleeding disorders and infertility. We per- Understanding molecular mechanisms and formed a systematic survey of tubulin mutations across all isotypes in order to improve our predicting phenotypic effects of pathogenic tubulin mutations. PLoS Comput Biol 18(10): e1010611. understanding of how they cause disease, and increase our ability to predict their pheno- https://doi.org/10.1371/journal.pcbi.1010611 typic effects. Both protein structural analyses and computational variant effect predictors Editor: Bert L. de Groot, Max Planck Institute for were very limited in their utility for differentiating between pathogenic and benign mutations. Biophysical Chemistry, GERMANY This was even worse for those genes associated with non-neurodevelopmental disorders. Received: June 20, 2022 We selected tubulin-α and -β disease mutations that were most poorly predicted for experi- mental characterisation. These mutants co-localise to the mitotic spindle in HeLa cells, sug- Accepted: September 28, 2022 gesting they may exert dominant-negative effects by altering microtubule properties. Our Published: October 7, 2022 results show that tubulin mutations represent a blind spot for current computational Peer Review History: PLOS recognizes the approaches, being much more poorly predicted than mutations in most human disease benefits of transparency in the peer review genes. We suggest that this is likely due to their strong association with dominant-negative process; therefore, we enable the publication of all of the content of peer review and author and gain-of-function mechanisms. responses alongside final, published articles. The editorial history of this article is available here: https://doi.org/10.1371/journal.pcbi.1010611 Copyright:© 2022 Attard et al. This is an open Author summary access article distributed under the terms of the Creative Commons Attribution License, which Filament-like structures, called microtubules, are essential for cells to function, distribute permits unrestricted use, distribution, and material around the cell and organisms, and help cells grow. The building blocks of micro- reproduction in any medium, provided the original tubules are proteins called tubulins, which can rapidly polymerise and depolymerise. author and source are credited. Mutations in tubulin genes can have catastrophic consequences on many different types Data Availability Statement: All relevant data are of cells, leading to diseases such as bleeding defects, female infertility, and disorders provided within the manuscript and its Supporting impairing brain development. However, how these mutations cause disease and whether Information files. they can be predicted is still unknown. We used computational and experimental tech- Funding: This work was supported by the Medical niques to address these issues. First, we compared how disease-causing tubulin mutations Research Council via a Precision Medicine Doctoral PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 1 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Training Programme studentship to TJA and a Career Development Award (MR/M02122X/1) to and ones found in healthy people impact the structure of tubulin. Then, we tested the abil- JAM. JAM is a Lister Institute Research Fellow. ity of available computational predictors to distinguish between these two types of tubulin JPIW is supported by a Wellcome Trust Senior mutations. We found these programs poorly predict tubulin mutations that cause diseases, Research Fellowship (207430). The funders had no limiting their usefulness. Next, we studied disease-causing mutations that were not pre- role in study design, data collection and analysis, dicted by computational methods. We found that these did not prevent tubulin from decision to publish, or preparation of the forming microtubules, indicating these mutations change the function of tubulin without manuscript. inactivating them. Our work presents tubulins as a weakness of current computational Competing interests: The authors have declared predictors, potentially because they fail to consider different ways in which mutations that no competing interests exist. cause disease. Introduction Microtubules are polarised cytoskeletal filaments essential in several cellular processes, ranging from cell division to signalling and transport. They assemble into axons, cilia, and the mitotic spindle, while also providing tracks for microtubule-associated proteins (MAPs) and motors for molecular trafficking [1,2]. Microtubules self-assemble from tubulin-α and -β heterodimers, with the dynamics of their assembly and disassembly being integral to their function [3]. Tubulin-α and -β are ubiquitous in eukaryotes, while related proteins in the FtsZ family show similarity in sequence, structure and function in archaea and bacteria [4,5]. Nine tubulin-α and ten tubulin-β genes have been identified in humans, originating from evolutionary gene duplication events [6]. In the tubulin field, these tubulin paralogues are referred to as isotypes. Between tubulin-α and -β, conserva- tion in sequence and structure is high, especially at interfaces stabilising the heterodimer, and at contacts between tubulin heterodimers across (lateral) and along (longitudinal) protofila- ments [7]. While tubulin-α and -β both bind to GTP, only tubulin-β can hydrolyse it to GDP, with residues on these binding sites amongst the most conserved [8]. GTP hydrolysis enables distinct conformations that mediate the dimer’s ability to be incorporated into microtubules [9]. The C-terminal region makes up the outer surface of the microtubule and so contributes to most of the interactions with MAPs [10]. Furthermore, many differences in amino acid sequences between isotypes occur in this region [11] and in the unstructured, highly negative tail [12]. Tubulin-γ,δ andε are more divergent in sequence than tubulin-α and -β and are involved in the basal bodies of centrioles, rather than being self-assembled into dynamic poly- mers [13,14]. A wide range of genetic disorders–called ’tubulinopathies’–have now been attributed to tubulin mutations. Over 225 pathogenic mutations in human tubulin isotypes have been reported (see Table 1). These findings highlight the importance of understanding tubulin func- tion in different cell types. The first reported tubulin mutations associated with pathogenic phenotypes were found in TUBα1A, TUBβ2B, TUBβ3, and TUBβ4A and resulted in neurode- velopmental defects [15]. Mutations in theγ-tubulin isotype TUBγ1 –which is necessary for microtubule nucleation–also cause a similar neurodevelopmental disorder [16,17]. While mutations in TUBα4A and TUBβ4A have been linked to neurodegenerative disease [18,19], phenotypes outside the nervous system are now emerging. These include TUBβ1 mutations associated with bleeding disorders [20], a link between tubulin-α acetylation and reduced sperm motility [21], and TUBβ8 mutations connected with female infertility due to incorrect meiotic spindle assembly [22]. Despite the large number of identified pathogenic tubulin mutations, our understanding of the molecular mechanisms by which these mutations cause disease remains limited. The large PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 2 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Table 1. Summary of all tubulin isotypes considered for this study. The number of dominant pathogenic mutations identified for each isotype is denoted, as well as their pathogenicity classification. Where necessary, additional comments about the pathogenicity type and references for all mutations are also included. MCM = missense constraint metric (Z-Score obtained from gnomAD); ALS = amyotrophic lateral sclerosis; H-ABC = hypomyelination with atrophy of basal ganglia and cerebellum. Isotype # # MCM Phenotype classification References and comments pathogenic gnomAD TUBα1A 67 680 5.584 neurodevelopmental Various neuropathies [35,79–83] TUBα1B 0 696 5.412 cancer-only Roles in resistance to chemotherapy [84–86] TUBα1C 0 153 2.171 cancer-only Identified as an oncogene [87] TUBα3D 0 343 1.533 non-disease TUBα3E 0 531 -0.296 non-disease Only 1 homozygous mutation known, which was excluded [88] TUBα4A 7 255 3.303 other All linked with ALS [19]. One deletion mutation also identified (excluded) (neurodegenerative) TUBα8 0 201 0.631 non-disease Homozygous 14bp deletion linked with polymicroglia [24] (excluded) TUBβ1 6 268 0.003 other (bleeding Linked to platelet defects [45–47] disorders) TUBβ2A 3 192 5.263 neurodevelopmental Simplified gyral patterning, infantile-onset epilepsy and progressive spastic ataxia [30, 89] TUBβ2B 29 190 5.120 neurodevelopmental Various neuropathies [15,29,55,57,64,90–92] TUBβ3 24 76 4.579 neurodevelopmental Various neuropathies, a few mutations associated with milder phenotypes [15,32,55,57,64,66,93,94] TUBβ4A 38 198 4.262 neurodevelopmental Mostly H-ABC and Hypomyelination, some less severe and a few linked with Dystonia [56,95–112] TUBβ4B 2 201 4.498 other (sensory) Linked heavily with cancer [113]. Two mutations have been linked to a sensorineural neurodegenerative disease impairing vision [18] TUBβ5 7 198 5.625 neurodevelopmental Various neuropathies, mostly severe [28,114–119] TUBβ6 1 74 2.637 neurodevelopmental Linked with non-progressive congenital facial palsy [120] TUBβ8 35 464 1.860 other (female infertility) Linked to oocyte maturation defects [22,23,50–53] TUBβ8B 0 228 -2.372 non-disease TUBγ1 8 57 4.155 neurodevelopmental All linked to cortical development malformations, one of unknown inheritance [16,17] (excluded). Two deletions also cause a lack of centrosome localisation [121] TUBγ2 0 113 2.443 non-disease https://doi.org/10.1371/journal.pcbi.1010611.t001 majority of pathogenic tubulin mutations involve missense changes (i.e. single amino acid resi- due substitutions) and have autosomal dominant inheritance. There are only a few known exceptions, including homozygous null and in-frame deletion mutations in TUBβ8 that can cause an oocyte maturation defect [23], a dominant nonsense mutation that results in a slightly truncated TUBα4A linked to the neurodegenerative disease amyotrophic lateral sclerosis [19], and a recessive intron deletion in TUBα8 in polymicrogyria patients that interferes with splic- ing, producing shorter mRNAs that do not contain exon 2 [24]. The absence of any known protein null mutations in tubulins causing dominant disease is striking. This tells us that the molecular mechanism underlying dominant mutations cannot simply be haploinsufficiency, whereby disease is caused by a complete lack of functional pro- tein produced from the mutant allele. One possibility is that disease is caused by a milder loss of function (i.e. hypomorphic mutations). For example, pathogenic TUBα1A mutations have been reported to disrupt interactions of the nascent protein with tubulin chaperones, which impairs heterodimer formation, suggesting that disease can be caused by a partial loss-of-func- tion [25]. Alternatively, a pathogenic mutation could act via a non-loss-of-function mecha- nism, having a dominant-negative effect or causing a gain of function [26]. This would typically be associated with the mutant protein retaining the ability to incorporate into micro- tubules, which has been observed for many pathogenic tubulin mutations [27–32]. For domi- nant-negative mutations, the incorporation of mutant protein directly or indirectly disrupts PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 3 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations the activity of the wild-type protein [26]. In these cases, the mutant tubulin retains its ability to form a heterodimer and assemble into microtubules before consequently impacting function in some other way, e.g. by perturbing microtubule properties or disrupting interactions with MAPs. For instance, TUBβ3 mutations that alter the charged surface of the microtubule pre- vent molecular motors from binding and thus have profound impacts on cellular transport [33]. Changes in microtubule properties could also induce a gain of function, as has been pro- posed for the pathogenic T178M variant in TUBβ2A and TUBβ3, which has been reported to make microtubules more stable and cause altered microtubule growth dynamics [34]. With the increasing accessibility of sequencing data, novel tubulin variants are being con- tinually discovered [35]. Since it is impractical to test them all experimentally, there is a strong need for computational approaches to identify tubulin mutations most likely to be pathogenic. Many variant effect predictors (VEPs) have been developed in recent years, and some of these are now in widespread use to help identify mutations that potentially have clinical significance [36]. However, the performance of these predictors can vary quite dramatically across different proteins and, to our knowledge, there has been no systematic assessment of their performance on tubulin mutations specifically. Missense mutations that cause pathogenicity by dominant- negative or gain-of-function mechanisms tend to be poorly predicted by most currently avail- able VEPs [37], which could potentially limit their applicability to tubulins. It is therefore important for us to understand to what extent we can rely on computational predictors when assessing tubulin mutations. In this study, we have first performed a systematic survey of known pathogenic missense mutations across all human tubulins and analysed their positions within the three-dimensional structures of tubulin heterodimers. This approach has allowed us to look for patterns in muta- tions across isotypes, structural locations, and phenotypes, in an attempt to obtain insight into the likely molecular disease mechanisms. Next, we assessed the performance of several differ- ent VEPs in distinguishing between pathogenic and putatively benign missense variants, observing that the predictive performance of all tested methods is poor compared to most other proteins. Finally, we have selected pathogenic tubulin mutations that were poorly pre- dicted by computational approaches for experimental characterisation and show that the mutant proteins are able to be incorporated into microtubules, consistent with a likely domi- nant-negative mechanism. Our work suggests that many tubulin pathogenic mutations that act via non-loss-of-function mechanisms cause pathogenic phenotypes that cannot be explained computationally using current methods that rely on sequence conservation or pro- tein structure. Overall, this study highlights the need for a greater understanding of microtu- bule-protein interactions to understand the molecular mechanisms underlying tubulinopathies. Results Survey of tubulin missense mutations First, we compiled as many previously identified pathogenic or likely pathogenic dominant tubulin missense mutations as possible, using online databases [38,39] and extensive literature searching (S1 Table). In addition, we also identified missense variants in tubulin genes observed across >140,000 people from the gnomAD v2.1 database [40]. Given that the gno- mAD dataset comprises mostly healthy individuals without severe genetic disorders, these var- iants are unlikely to cause dominant disease, and we therefore refer to them as "putatively benign". However, we acknowledge that some of these variants could have milder effects, vari- able penetrance, or be associated with late-onset disease. Table 1 shows the numbers of patho- genic and gnomAD missense variants for each tubulin isotype and the associated type of PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 4 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations genetic disease. Somatic mutations in tubulins are also implicated in cancer development; however, these mutations are likely to provide a selective advantage to cancer cells by provid- ing resistance to chemotherapeutic drugs [41–43]. Hence these mutations might obscure our results and have not been included in our study, although we have noted two isotypes with links to cancer in which no other disease-related mutations have been identified yet (Table 1). While pathogenic mutations occur throughout the tubulins, mutations in both tubulin-α and -β show clustering towards the intermediate and C-terminal domains when shown in the con- text of the linear amino acid sequence (Fig A in S1 Text). We also considered the gene-level missense constraint metric (MCM) scores provided by the gnomAD database (Table 1). These are derived from a model based on sequence context to predict the number of expected variants present in a healthy population relative to the number of actual variants observed [44]. They provide a metric for the tolerance of each isotype to mis- sense variation, with higher values representing genes that are more intolerant to amino acid sequence changes. Interestingly, we observed high MCM scores for isotypes linked with neuro- developmental disorders (TUBα1A, TUBβ2A, TUBβ2B, TUBβ3, TUBβ4A and TUBβ5). TUBβ6 is the only exception, and has only one pathogenic mutation reported so far, causing congenital facial palsy. These scores contrast with the much lower scores observed for TUBβ1 and TUBβ8, associated with platelet defects [45–49] and female infertility [22,23,50–53], respectively, which suggest that they are much more tolerant to sequence variation. Overall, our analysis indicates there is a stronger sequence constraint in the human population for tubulin isotypes that contribute to neurodevelopment. This may be due to the selective pres- sure of the process, compared to tubulin isotypes expressed in cells that affect organism fitness to a lesser extent. Protein structural context is of limited utility for explaining tubulin mutation pathogenicity Next, we mapped all variants to three-dimensional structures of tubulins in the Protein Data Bank (PDB). We classified each mutation-carrying residue from every tubulin isotype based upon its structural location: whether it occurs at an intermolecular interface, on the protein surface, or buried in the protein interior. We used a hierarchy to classify structural locations (Fig B in S1 Text). First, given that GTP binding (and hydrolysis for tubulin-β) is essential, we classified residues on the GTP nucleotide binding site as ’GTP binding’. Other tubulin muta- tions occurring at intermolecular interfaces were classified into three categories depending on where they occur in the model structures described in S2 Table: at the intradimer interface, at an interdimeric interface (which we refer to as microtubule contacts, as they support lateral and longitudinal interactions with microtubules), or at interfaces with MAPs. Remaining mutations not at interface positions were then classified depending on whether they were on the protein interior or surface according to relative solvent accessible surface area [54]. Finally, we labelled any residues that were not present in any ordered part of any experimentally derived tubulin structures as ’outside of structure’. These typically occurred in the disordered C-terminal region, or at positions of divergence between the sequences of mapped structures and the tubulin isotype in question. Notably, while several gnomAD variants occurred at ’out- side of structure’ positions, this was true for only a single pathogenic mutation: R307H in TUBβ1. Interestingly, all other tubulin-β isotypes have a histidine at this position, explaining why the residue could not be mapped within our pipeline, as the wild-type amino acid residue needs to be present in the structure. First, we used this classification system to assess whether tubulin mutations are enriched at particular locations. Despite the large dataset size, no significant differences in the locations of PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 5 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 1. Comparison of the distribution of structural locations between pathogenic and putatively benign tubulin variants. The proportion of pathogenic and putatively benign gnomAD mutations in each location type for tubulin-α,β andγ globally (A) and for individual isotypes (B). Mutation totals for each group are shown at the bottom, and only isotypes with at least five pathogenic mutations were included. Fisher’s exact tests were used to compare frequencies between gnomAD and pathogenic mutations considering each family and isotype separately. Asterisks indicate a location class with a significantly higher � � � � � � � � � � proportion of mutations compared to its corresponding group, where p < 0.05, p < 0.01, p < 0.001, p < 0.0001. https://doi.org/10.1371/journal.pcbi.1010611.g001 pathogenic or putatively benign gnomAD mutations were found across tubulin-α isotypes col- lectively (Fig 1A) or in any tubulin-α isotype individually (Fig 1B). This result was surprising given that the link between TUBα1A and disease remains the most well-established of all tubu- lin isotypes, with 67 pathogenic variants linked with various neurodevelopmental defects [35]. Observations were similar for TUBα4A, the only other tubulin-α isotype with dominant path- ogenic mutations (Fig 1B) [19]. Furthermore, pathogenic mutations in both isotypes showed no obvious patterns or spatial clustering when visualised within the protein structure (Figs 2 and C and D in S1 Text). In contrast to tubulin-α, we do observe a significant enrichment of pathogenic tubulin-β mutations relative to gnomAD variants at GTP binding residues and at lateral or longitudinal (i.e. interdimeric) microtubule contacts (Fig 1A), especially in TUBβ2B and TUBβ4A (Fig 1B). Mutations in these isotypes have been extensively linked to neuronal defects, and can be observed to form spatial clusters on tubulin structures (Fig 2). In contrast, putatively benign gnomAD variants are significantly more likely to occur on the protein surface or at ’outside of structure’ positions (Fig 1B). TUBβ3 and TUBβ5 are also linked to similar disease phenotypes, but do not show any notable differences in structural location [15,28,55–57]. Interestingly, we found no significant differences in structural location in tubulin-β isotypes associated with diseases outside the neuronal system (Fig 1B), nor did pathogenic mutations show any clear patterns or clustering at the three-dimensional structure level (Fig 2). We did, however, note that many TUBβ1 variants occurred at ’outside of structure’ residues (Fig 1B). This is due primarily to the greater sequence divergence between TUBβ1 and experimentally determined tubulin structures, as any positions where the TUBβ1 sequence is different than the aligned positions within the available structures will remain unmapped using our approach. The centrosome-localised TUBγ1 has also been associated with neurodevelopmental dis- ease [16,17]. Using the crystal structure of TUBγ1 containing two molecules per asymmetric unit (PDB ID: 3CB2) [58], we observed that 4 out of 8 pathogenic mutations identified so far occur at interfaces responsible for lateral interactions between TUBγ1 monomers (Figs 1 and PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 6 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 2. Visualisation of pathogenic tubulin mutations on three-dimensional structures of tubulin heterodimers. Structures of tubulin isotypes with at least five identified pathogenic mutations. Coloured residues indicate pathogenic mutations according to their location. The position of the tubulin subunit in relation to the microtubule (MT) is noted at the top. Residues in red denote mutations buried in the protein interior, while light cyan residues are on the surface. Green residues indicate mutations occurring at GTP binding interfaces, while ones in magenta highlight residues on intradimer interfaces. Navy blue and yellow residues mark mutations on residues interacting with microtubule associating proteins (MAPs) and other tubulin dimers, respectively. PDB IDs: 6s8k (for all tubulin-α andβ isotypes) and 3cb2 (for TUBγ1). Alternate versions of this figure with the structures rotated 90˚ and 180˚ around the y-axis are provided in Figs C and D in S1 Text. https://doi.org/10.1371/journal.pcbi.1010611.g002 2, labelled in magenta). We suspect these interactions to occur between adjacent TUBγ1 chains withinγ-TURC [59]. Their distribution was not significantly different to gnomAD variants, although we note the small sample size. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 7 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 3. Comparison of predicted changes in protein stability between pathogenic and putatively benign tubulin mutaitons.ΔΔG values representing the predicted change in free energy of folding were calculated with FoldX considering the structure of the monomeric subunit only. Scores are shown for tubulin-α and -β families globally (A) and in isotypes with at least five identified pathogenic mutations (B). Maroon diamonds indicate the meanΔΔG values, and mutation totals for each group are also shown at the bottom. The p-values displayed were obtained via unpaired Wilcoxon tests. https://doi.org/10.1371/journal.pcbi.1010611.g003 Most pathogenic tubulin mutations are not highly disruptive to protein structure Next, we considered the predicted structural perturbations of pathogenic and putatively benign gnomAD missense variants using FoldX [60]. This outputs aΔΔG value, in units of kcal/mol, with positive values indicating that a mutation is likely to destabilise protein struc- ture and negative values indicating predicted stabilisation. Previous work has shown that com- putationally predictedΔΔG values can sometimes show considerable utility for the identification of pathogenic missense mutations, and for understanding likely molecular dis- ease mechanisms [61]. Interestingly, we observe no significant differences between theΔΔG values of pathogenic and gnomAD missense variants for tubulin-α, -β or -γ (Fig 3A). Of the individual isotypes, only TUBβ2B shows significantly higherΔΔG values for pathogenic mutations (p = 0.01), although this would not remain significant when accounting for multiple testing (Fig 3B). We initially usedΔΔG values that only consider the structural impact of variants on the monomer alone, as they are more consistent between structures. However, we also observed very similar results using the fullΔΔG values calculated using the entire complex, including intermolecular interactions, as well as when using absoluteΔΔG values (Fig E in S1 Text). These results suggest that the structural destabilisation is not a primary molecular disease mechanism underlying pathogenic tubulin mutations, and that considering structural impact is not particularly useful for differentiating between pathogenic and benign tubulin variants. Notably, this aligns with recent work showing that the predicted effects on protein stability tend to be much milder in gain-of-function and dominant-negative mutations than for patho- genic mutations associated with a loss of function [37], supporting the idea that most patho- genic tubulin mutations are due to non-loss-of-function effects. Variant effect predictors show poor performance in discrimination between pathogenic and putatively benign tubulin mutations Next, we assessed the abilities of 25 different VEPs to distinguish between pathogenic and putatively benign tubulin missense mutations. A complete set of predictions for every tubulin mutation from all VEPs is provided in S3 Table. To compare the performance of different PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 8 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations VEPs, we first used a metric of predictor performance, known as the receiver operating charac- teristic (ROC) area under the curve (AUC) generated from each VEP over the entire dataset of tubulin missense variants (Fig 4A). Overall, the VEPs performed very poorly. Most had overall AUCs below 0.6, with the top-performing predictor, REVEL, having an AUC of only 0.68. In contrast, a recent study using a very similar methodology found that many VEPs had overall AUCs above 0.8, e.g. REVEL had an AUC of 0.9 for haploinsufficient disease genes, 0.85 for genes associated with a gain-of-function, and 0.83 for genes associated with dominant-nega- tive effects [37]. Thus, even considering that VEPs tend to do worse for non-loss-of-function mutations, the performance we observe here for tubulin mutations is strikingly poor. It is important to note that our analysis is likely to overstate the predictive power of some VEPs. Supervised machine learning approaches underpin most VEPs, and typically use data- sets of known pathogenic and benign variants for training. Since some VEPs are likely to have been trained using some of the tubulin mutations in our evaluation, their performance has a strong possibility of being overstated. This problem is particularly acute for metapredictors, including the top-performing methods in our analysis, REVEL and M-CAP, which combine supervised learning with multiple other predictors as inputs. In contrast, predictors based upon unsupervised machine approaches and those utilising empirical calculations should be free from this bias. Therefore, given the performance of the unsupervised predictor DeepSe- quence, ranking third overall, we likely consider it to be the most reliable predictor of tubulin mutation pathogenicity, consistent with its top-ranking performance in a recent study [62]. However, even DeepSequence only achieves an AUC of 0.63 for tubulin mutations here, com- pared to well over 0.8 for all disease-associated proteins tested in that study. Next, we compared the AUCs calculated for individual isotypes across all VEPs, consider- ing isotypes with at least 10 pathogenic mutations. We found that pathogenic mutations in the tubulin-β isotypes TUBβ2B, TUBβ3 and TUBβ4A, which are all associated with neurodevelop- mental diseases, were predicted better compared to TUBβ8, associated with oocyte maturation defects (Fig 4B). Therefore, we classified tubulins into two groups based upon observed disease phenotypes: neurodevelopmental and other (as classified in Table 1). We observed a signifi- cantly better performance on isotypes linked to neurodevelopmental diseases (Fig 5A). Fig 4. Assessment of VEP performance for identification of pathogenic tubulin mutations. (A) ROC AUC values for each VEP across all tubulin-α and -β isotypes with at least one identified pathogenic mutation, colour coded according to predictor category. (B) Distribution of ROC AUC values across all VEPs for isotypes with at least 10 identified pathogenic missense mutations. Dashed line indicates the performance of a random predictor. Maroon diamonds indicate the mean area. https://doi.org/10.1371/journal.pcbi.1010611.g004 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 9 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 5. Comparison of VEP performance on pathogenic mutations in tubulin genes associated with neurodevelopmental vs other disease phenotypes. (A) Distribution of ROC AUC values across all VEPs for isotypes with mutations linked with neurodevelopmental or other disease phenotypes. The p-value stated was obtained via a paired Wilcoxon test. Maroon diamonds indicate the mean area. (B) ROC AUC values for each VEP in isotypes with mutations associated with neurodevelopmental or other disease phenotypes. Dashed line indicates the performance of a random predictor. https://doi.org/10.1371/journal.pcbi.1010611.g005 Interestingly, when we compare the performance on neurodevelopmental vs other disorders across the individual VEPs (Fig 5B), we observe that metapredictors and supervised VEPs, like REVEL, M-CAP and VEST4, outperformed all other VEPs on isotypes linked to neurodeve- lopmental disease but showed a drastic decrease in performance on other phenotypes. In con- trast, unsupervised DeepSequence shows very similar performance between the two groups. This strongly suggests that that certain VEPs have likely been overfitted in their training against the neurodevelopmental mutations. Given its overall performance in our analyses and unsupervised nature, we currently rec- ommend DeepSequence for predicting the effects of tubulin mutations, although we empha- sise that its predictive utility is still relatively limited. Therefore, we have produced DeepSequence predictions for every possible amino acid substitution across most tubulin iso- types and have provided them as a resource (S4 Table). We have also calculated optimal thresholds for DeepSequence using the closest point to the top left corner of our ROC curves. Based upon this, we suggest that DeepSequence scores lower than -5.89 are likely to be patho- genic in isotypes linked with neurodevelopmental phenotypes, and lower than -4.83 for iso- types linked with other phenotypes. The most poorly predicted tubulin pathogenic mutations likely act via non- loss-of-function mechanisms To further examine the contributing factors behind the limited performance of VEPs in detect- ing pathogenic tubulin mutations, we sought to identify the pathogenic mutations predicted most poorly by current approaches. To do this, we developed a ranking method to examine mutations across all the VEPs used in this study. First, all predictions of gnomAD and patho- genic tubulin mutations were amalgamated. We then transformed all scores from individual VEPs to be on the same scale for comparison. Next, for each VEP, cumulative distribution ranks were computed for normalisation within a window partition across the combined data- set. Finally, for each mutation, these normalised ranks were averaged across all VEPs to gener- ate a new metric we termed mean cumulative distribution (MCD), provided in S5 Table. Essentially the MCD provides a single value for each mutation representing how damaging it is predicted to be across all of the VEPs we used in our analysis, ranging from 0 for mildest to PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 10 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 1 for most disruptive. MCD replicated the results we observed for the VEP analysis, with the most significant differences between pathogenic and gnomAD values being observed in the same isotypes (TUBβ2B, TUBβ3, TUBβ4A) that showed higher ROC values (Fig F in S1 Text). Next, we used the MCD scores to identify pathogenic tubulin mutations that were most poorly predicted by the VEPs (Table 2). These mutations generally involve substitutions between chemically similar amino acids. In particular, mutations from valine to isoleucine at position 353 in three different isotypes–TUBα1A, TUBβ5 and TUBβ8 –are among the worst predicted, with the variant in TUBβ8 having the lowest MCD score of any pathogenic tubulin mutation. Phenotypically, in TUBα1A and TUBβ5, V353I is related to the malformations of cortical development [15], while the mutation is associated with female infertility in TUBβ8, although whether the mutation is causing the issue is not shown [50]. Consequently, we experimentally tested whether mutations in the 353 position for TUBα1 and TUBβ8 enabled tubulin incorporation into microtubules. As V353I introduces a mild physicochemical change, as with most of the poorly predicted pathogenic tubulin mutations, we also tested a more disruptive mutation in both isotypes where the hydrophobic valine is changed to a positively charged arginine. V353R has to our knowledge not been observed in humans (nor is it possible with only a single nucleotide substitution), but it allows us to com- pare an apparently mild substitution to a more disruptive substitution at the same position. We transiently transfected the fluorescently tagged tubulin mutants into HeLa cells and ana- lysed whether they were incorporated into the mitotic spindle, where microtubules can be clearly observed and where spindle architecture relies on MAPs. N-terminally tagged mCherry-TUBα1 incorporated into microtubules with the aid of a linker; however mCherry- TUBβ8 did not. We therefore tested a C-terminally tagged TUBβ8-GFP, which did successfully incorporate. Table 2. The most poorly predicted pathogenic tubulin mutations. For each of the VEPs used in this study, cumula- tive distribution ranks were generated in a combined dataset of putatively benign and pathogenic tubulin variants. MCD scores for each variant were calculated by averaging the ranks from each predictor. Pathogenic variants with the lowest MCD scores are shown here, representing those most poorly predicted by VEPs. Isotype Mutation MCD Score TUBß8 V353I 0.188 TUBa1A I238V 0.215 TUBa1A I219V 0.215 TUBß8 I4L 0.221 TUBa1A D127E 0.245 TUBß8 A352S 0.247 TUBß1 R307H 0.253 TUBß8 M330I 0.280 TUBß5 V353I 0.280 TUBa1A S54N 0.300 TUBß8 R2K 0.311 TUBa1A A270S 0.326 TUBa1A V353I 0.335 TUBa1A I188L 0.341 TUBß3 M323V 0.342 TUBß8 F200L 0.352 TUBß8 V175M 0.354 TUBa1A I5L 0.357 TUBa1A V409I 0.371 TUBa1A R214H 0.372 https://doi.org/10.1371/journal.pcbi.1010611.t002 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 11 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Both wild-type TUBα1 and TUBβ8 isotypes strongly incorporated into the mitotic spindle in HeLa cells, co-localising with tubulin (Fig 6A and 6B and 6G in S1 Text). The V353I mutants also showed incorporation that was not significantly different from wild type for both isotypes. This is also consistent with previous work showing that V353I in TUBβ5 can still form a heterodimer and be incorporated into microtubules [28]. Interestingly, however, the V353R mutant showed severe reduction in incorporation for TUBα1, but no significant effect for TUBβ8. We also tested whether the tubulin mutations affected mitotic spindle morphology and measured the mitotic spindle length (Fig 6C). Notably, while TUBα1 V353I did not show reduced incorporation into microtubules, it did lead to significantly shorter spindles, suggest- ing the mutation may ultimately interfere with correct microtubule dynamics or interactions with microtubule-associated proteins. Other mutants showed no significant difference from wild type. We also confirmed that spindle aspect ratios were similar between wild-type and mutant tubulin constructs, indicating spindle scaling is conserved (Fig H in S1 Text). Our results clearly suggest a possible molecular mechanism underlying genetic disease in the TUBα1A V353I mutation: the mutant can incorporate into microtubules along with wild- type protein, but appears to have a dominant-negative effect on spindle organisation of micro- tubules. In contrast, the V353R mutation, which we predict to be more severe at a protein structural level, shows reduced incorporation and thus has no effect on spindle morphology. This mirrors the recent observation that dominant-negative and gain-of-function mutations tend to be more structurally mild than loss-of-function mutations [37]. None of the tested TUBβ8 mutations show any phenotypic effect. It is interesting to note that V353 residues in tubulin-α and -β are not at precisely equivalent positions in a sequence alignment or in the structures of tubulin-α and -β (Fig I in S1 Text). V353 is a surface residue in tubulin-α, close to the longitudinal interdimeric interface, whereas in tubulin-β, it is more buried in the protein interior, but close to the intradimer interface, and thus it is unsurprising that mutations at V353 could have different molecular effects in each isotype. It is possible that the molecular effects of V353 mutations in TUBβ8 are too subtle to detect in our experimental system. Alternatively, the evidence that V353I is a true pathogenic mutation in TUBβ8 is quite limited. It was observed in a heterozygous state in a single patient with infertility, but it is unknown if either of her parents also had the mutation [50]. Thus, another possible explana- tion for the poor computational prediction of TUBβ8 V353I is that it is not really pathogenic. Discussion Tubulinopathies have previously been associated with defects in cortical brain development [15,57,63–66]. However, several pathogenic tubulin mutations reported in recent years are not linked to neurodevelopmental diseases, and thus the phenotypical profile of ’tubulinopathies’ needs redefining. As the mechanisms by which tubulin mutations give rise to these disease phenotypes are generally unknown, the extensive repertoire of newly identified mutations now at our disposal provides an opportunity to study pathogenicity globally, across the tubulin family. The strong conservation in tubulin sequence and structure across isotypes [12], and the lethality of tubulin mutations in yeast [67] might lead to the assumption that many pathogenic mutations would cause protein instability and consequently a loss of function. However, this appears not to be the case, with predictions of changes in protein stability showing very little ability to discriminate between pathogenic and putatively benign mutations, with the excep- tion of a slight trend observed for TUBβ2B. There is some previous evidence for a loss-of-func- tion mechanism underlying some pathogenic tubulin mutations, e.g. the R264C mutation in TUBα1A perturbs chaperone binding and reduces the extent of heterodimer assembly, which PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 12 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations Fig 6. Effects of V353 mutants on microtubules and mitotic spindles in cells. (A) Representative live-cell images of mitotic HeLa cells after being transfected with fluorescently tagged wild type and mutant tubulin-α1B and tubulin-β8 constructs (green) and incubated with SiR-tubulin (red) (B) Wide linescans across the whole mitotic spindle were taken to measure fluorescence intensity of tubulin across the mitotic spindle compared to total tubulin. For each cell, two-sided Spearman correlations were calculated between both signal intensities. (C) Spindle length measurements for the cells transfected with fluorescently tagged wild-type and mutant tubulin-α1B and tubulin-β8 and measured in (B). For each construct, median values are displayed with 95% confidence intervals. � � � � � � � Asterisks indicate Kruskal-Wallis test (with post-hoc Dunn) significance values, , P < 0.001; , P < 0.0001. Bars, 5μm. https://doi.org/10.1371/journal.pcbi.1010611.g006 has been suggested to account for its pathogenic effects [25]. However, it appears that the over- all role of destabilisation in driving tubulin pathogenicity must be very limited. Tubulin mutations can act via dominant-negative or gain-of-function mechanisms, where the mutant protein incorporates into the microtubule lattice and instead disrupts its function. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 13 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations These effects have been reported in tubulins before and include mutations in TUBα1A [27] and others across different tubulin-β isotypes [28–32]. Once they are incorporated into micro- tubules, these mutations may alter biophysical microtubule properties or disrupt interactions with MAPs. The latter model includes interactions with non-motor MAPs, like NuMA and PRC1 [31,32], and the impairment of motor protein binding and, subsequently, cellular trans- port [33]. The disruption of tubulin-MAP interactions may be an especially pertinent mecha- nism given that mutations on protein-protein interfaces are more likely to be pathogenic [68]. Here, multiple copies of mutant tubulin incorporated into microtubule can compound a mutation’s effect. Also, some tissues express multiple tubulin-β isotypes, especially in the developing brain [69]. Here, redundancy may mitigate any loss-of-function in a specific iso- type, but cells struggle to cope with dominant-negative or gain-of-function effects. The classic definition of the dominant-negative effect involves the mutant protein disrupt- ing the activity of the wild-type protein, either directly or indirectly, whereas the phenotypic effects of gain-of-function mutations are due to the mutant protein doing something different than the wild-type protein [26]. For most proteins that form relatively small complexes, classi- fication of a mutation as being associated with to either mechanism should usually be straight- forward. However, given the large size of microtubule assemblies and their diverse functional roles, such classification may be ambiguous for tubulins. It depends on whether disease is due to reduced microtubule function (dominant negative), or whether altered microtubule proper- ties are driving disease (gain of function). While here we have referred to TUBα1 V353I, which incorporates into microtubules and reduces spindle length, as dominant negative, it is possible that its pathogenic effects might be better described as gain of function. These assays give limited data because of their lack of sensitivity. Further experimental work and more quantitative assays are required to understand the effect of these types of tubulin mutations in the cellular and molecular context. The pathogenic mutations predicted most poorly tend to involve substitutions between similar amino acids, with hydrophobicity and charges largely maintained between wild-type and mutant residues. This observation likely reflects the fact that most predictors incorporate some manner of amino acid substation matrix, thus predicting more minor effects for variants that do not change a residue’s charge or hydrophobic state. In contrast to more drastic muta- tions that tend to be better predicted, these mutations may be less detrimental to folding and microtubule assembly. Indeed, this difference is likely to explain why we observed a weaker co-localisation to the mitotic spindle in the TUBα1 V353R mutant compared to V353I. In this study, the performance of supervised predictors, and especially metapredictors like REVEL and M-CAP, highlights the issue of circularity present in these types of VEPs, which are typically trained using published pathogenic and benign variants [70]. There is a strong likelihood that some of these mutations are in training sets for these predictors. These biases would lead metapredictors and supervised methods to predict mutations in isotypes linked with neurodeve- lopmental diseases as pathogenic correctly, while mutations in other isotypes would be predicted poorly. This may explain why metapredictors performed so much worse on such mutations, espe- cially when compared to unsupervised methods like DeepSequence and PROVEAN. We judge DeepSequence to be the best performing VEP, considering consistency across isotypes and phenotypes. Unlike supervised methods, it does not use a labelled training set. DeepSequence can also make predictions based on the evolutionary conservation of entire sequences at once as opposed to no more than a few sites of interest, setting it apart from many other VEPs [65]. Benchmarking several VEPs against deep mutational scanning data also highlighted DeepSequence as the best pathogenicity predictor [62]. We, therefore, recom- mend using DeepSequence over metapredictors like REVEL and M-CAP to achieve more accurate predictions of tubulin variant pathogenicity. PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 14 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations This study presents the tubulin family as a blind spot for current phenotype predictors, especially in identifying mutations that result in non-neurodevelopmental pathogenic pheno- types. Predictor biases due to their training sets introduce circularity and could limit predic- tions. The diversity of pathogenic phenotypes necessitates further study to understand the molecular mechanisms underlying their pathogenicity. Experimental approaches to dissect the effect of these tubulin pathogenic mutations on microtubule dynamics, microtubule assembly and microtubule interactions with associated proteins would contribute to a better molecular understanding of tubulinopathies [71–73]. These additional data could refine current pheno- type predictors and increase their performance. For non-neurodevelopmental phenotypes, pathogenic effects initially appeared to be iso- type-specific, but this could be linked to where these isotypes are expressed and to what level, or other unknown factors. For example, pathogenic mutations in TUBβ8 are associated to female infertility, but males carrying these mutations would presumably present a healthy phe- notype [22]. Importantly, many tubulin mutations are currently reported as linked to a disease or phenotype, but the evidence supporting causality is limited. It will be important to demon- strate whether the mutation is causative of pathogenesis or a consequence of other changes to cell homeostasis or genome stability. A pragmatic experimental approach is therefore essential to define the mechanism of pathogenesis associated with tubulin mutations. It will then be pos- sible to develop strategies to treat tubulin-related disorders using a personalised medicine approach. Methods Mutation datasets A thorough literature search was first performed to identify as many pathogenic or likely path- ogenic missense mutations across the tubulin family as possible. While the recently released tubulin mutation database [39]–as well as others such as OMIM [74] and ClinVar [38]–were excellent starting points, mutations from other publications (especially ones reported recently) were included (all detailed in Table 1). The literature search also uncovered a few other muta- tions that were incompletely dominant, homozygous, or of an unknown heritability, but these were omitted from the study. Any mutations described as de novo were assumed to be hetero- zygous and hence also dominant. As the underlying mechanisms with pathogenic mutations leading to cancer are likely to be more complex (and distinct), somatic mutations linked with such phenotypes were excluded. A set of putatively benign missense mutations for each isotype was retrieved from gnomAD v2.1 [40]. We excluded mutations present in both pathogenic and gnomAD datasets and considered only isotypes with more than one pathogenic mutation. Protein structural analyses Three-dimensional models of tubulin structures were taken from the protein data bank (PDB) on 2020.05.27, using the first biological assembly from each entry to represent its quaternary structure. We searched for polypeptide chains with at least 70% sequence identity to human tubulin isotype over a stretch of at least 50 amino acid residues. This allowed structural analysis using related structures for some isotypes for which structures are not available. Importantly, we only considered those residues where the amino acid in the structure, as well as both adja- cent residues, are the same as in the sequence of the isotype of interest. Unless otherwise stated, all structures that satisfy this threshold were considered. A hierarchy was implemented to determine where a residue occurred on the tubulin structure concerning its function (Fig B in S1 Text). First, we looked through all structures to highlight residues interacting with GTP (or its analogues). Then, model structures were used (S2 Table) to determine whether the PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 15 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations remaining residues occur on interfaces between the corresponding subunit in the tubulin dimer or with MAPs and other tubulin dimers in the microtubule in that specified order. Resi- dues not present on any interface were mapped to the highest resolution structure within the dataset. These mappings were used to classify residues according to their location, separating residues on the surface (relative solvent accessible surface area >0.25) from ones at the protein interior. Once every residue was labelled, their distribution was compared between mutations in our pathogenic and gnomAD datasets. Finally, all mutations in tubulin-α or -β were consid- ered before moving on to specific isotypes with at least 5 identified pathogenic missense muta- tions, also using PyMOL [75] to confirm that these patterns could also be observed qualitatively. To facilitate further analysis of tubulin mutations, we have provided a list of all amino acids in tubulin-α and-β that are involved in protein-protein or protein-ligand interac- tions (S6 Table). Structural predictions were then obtained from FoldX [60] for pathogenic and gnomAD variants. Each mutation was mapped onto the highest resolution structure from the dataset above. Here, theΔΔG values were calculated considering the monomer only, as well as the whole biological assembly, in the same manner as previously described [37]. All FoldX pre- dictedΔΔG values are provided in S7 Table. Computational variant effect predictions Scores from VEPs were generated using the same pipeline as described previously [62]. Though most of the methods used failed to produce scores for the entire combined dataset, the percentage of missing scores from each predictor did not exceed 7.8%, except for S3D-PROF and SNPs&GO-3D, which required PDB structures as inputs. For S3D-PROF and SNPs&GO- 3D, structures with the closest sequence identity were used, which have the following PDB IDs: 5iyz (TUBα1A), 5lxt (TUBα4A), 4tv9 (TUBβ1), 5nm5 (TUBβ2A and TUBβ8), 4i4t (TUBβ2B), 4lnu (TUBβ3), 5jqg (TUBβ4A and TUBβ4B) and 4zol (TUBβ5). For plotting ROC curves, mutations from the pathogenic dataset were labelled as true posi- tives, with gnomAD variants marked as true negatives. The plots themselves, as well as the AUC, were calculated using the PRROC package in R. Due to predictors PROVEAN, DeepSe- quence, SIFT4G and BLOSUM operating via inverse metrics (i.e. variants predicted to be more damaging have lower scores), all scores were multiplied by -1 before plotting curves and calculating AUC values. Finally, ROC AUC values were calculated globally across the whole dataset for each VEP first before generating specific scores of each VEP for individual isotypes. To calculate the MCD rankings, the datasets containing prediction scores for gnomAD and pathogenic mutations were combined. Then, the cumulative distribution function was applied to the collective scores obtained from each predictor using the cume_dist function from the DPLYR package on a per-VEP basis. Each value represents the proportion of scores that are less than or equal to the one obtained for that particular mutation from a specific VEP. Then, the cumulative distribution scores from all VEPs were averaged for each mutation to generate the MCD score. Molecular cloning The tubulin genes human TUBα1B (NM_006082.3) or TUBβ8 Class VIII (NM_177987.2) were synthesised by Genewiz. The TUBα1B gene were cloned into pBABE- blasticidin con- taining an N-terminal mCherry tag [76]. An additional linker region (with sequence 5’- ATGAGGAGGGGCGCTGCCGATAGGGAAACTGAGAGGCTCCCCGGCGCA- CAAGGTCCGTGCAGTGCGGTCAGTGCGGCCAGCTCCACATTGGCCGCAG- PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 16 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations CAGCGGCCCCTCGTGCTCGGGCGACCGCTGCCGCGTCCACCCTCAGCGC- CACCGCCCTCGAG-3’) was also inserted between the N-terminal mCherry tag and the start of the tubulin sequence. The TUBβ8 gene was cloned into a pBabe-blasticidin contanining a C-terminal GFP. Mutagenesis was performed using the Agilent Quickchange II mutagenesis kit, as per the manufacturer’s instructions. While the TUBα1B sequence was used rather than TUBα1A, where the V353I disease mutations has been identified, these are nearly identical, with only two minor amino acid sequence differences between them (G versus S at residue 232 and T versus S at residue 340), neither being close to V353 in the three-dimensional structure. Cell biology and fluorescence imaging HeLa cells were cultured and maintained in a humidified 5% CO atmosphere at 37˚C in DMEM supplemented with 100 U/ml penicillin and 100 μg/ml streptomycin (Gibco) and 10% Fetal Bovine Serum (Gibco). Cells were plated on 35 mm glass-bottom microwell dishes (ibidi) for live-cell imaging. Lipofectamine 3000 reagent (Invitrogen) was used for transient transfec- tions according to manufacturer guidelines. 24–48 hours after transfection, cells were incu- bated with 20–50 nM SiR-tubulin dye (SpiroChrome) for 1–2 hours and transferred to L-15 medium (ThermoFisher Scientific) before imaging. Images were taken using a Nikon Ti2 live imaging microscope were stored on OMERO [77], which was used to manually measure spin- dle length and width. ImageJ was then used to process the acquired images [78]. A region of interest was drawn over the spindle length observed through the Cy5 channel (visualising SiR- tubulin), and intensity profiles were generated for this channel and the GFP or mCherry chan- nel (depending on the construct). These profiles were then background subtracted using the average of the intensity profiles of two thinner lines on either side of the spindle. For each cell, a two-sided Spearman’s correlation coefficient was calculated between the fluorescent and SiR-tubulin intensity profile. OMERO was also used to calculate spindle length and width, where the length was divided by width to obtain calculate the spindle aspect ratio. Plots were visualised using GraphPad Prism 9, which was also used to carry out Kruskal-Wallis tests (with post-hoc Dunn). Supporting information S1 Text. Supplemental figures. File containing additional figures complementary to the anal- ysis shown in the main text of this manuscript. (DOCX) S1 Table. List of tubulin mutations. Full description of all pathogenic mutations compiled in this study (including phenotypes and references) as well as a list of all gnomAD mutations used. (XLSX) S2 Table. Structures used for location assignments. All model structures used for the hierar- chy implemented for assigning structural locations for tubulin residues. (CSV) S3 Table. Variant effect predictions. Complete set of raw predictions obtained from all the variant effect predictors used in this study for every tubulin mutation. (CSV) S4 Table. DeepSequence outputs. DeepSequence predictions for every possible missense mutation for all tubulin isotypes considered in this study. (CSV) PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 17 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations S5 Table. Mean cumulative distribution scores. Ranking of mean cumulative distribution (MCD) scores for all pathogenic tubulin mutations. (CSV) S6 Table. Tubulin residue interactions. All tubulin residues found to be interacting with GTP (or its analogues), MAPs, or with other tubulin subunits (interdimer or intradimer). (CSV) S7 Table. Predicted effects on protein stability for tubulin mutations. AllΔΔG values pre- dicted using FoldX using both the structure of the monomeric subunit (ddG_fold) and the full biological assembly (ddG_full). (CSV) S1 Data. Multiple sequence alignment. Multiple sequence alignment of all tubulin-α andβ isotypes considered in this study. (PDF) Acknowledgments We thank Benjamin Livesey for help with generating the VEP outputs. Author Contributions Conceptualization: Julie P. I. Welburn, Joseph A. Marsh. Data curation: Joseph A. Marsh. Formal analysis: Thomas J. Attard. Investigation: Thomas J. Attard. Supervision: Julie P. I. Welburn, Joseph A. Marsh. Writing – original draft: Thomas J. Attard. Writing – review & editing: Julie P. I. Welburn, Joseph A. Marsh. References 1. Brouhard GJ, Rice LM. Microtubule dynamics: an interplay of biochemistry and mechanics. Nat Rev Mol Cell Bio. 2018; 19(7):451–63. https://doi.org/10.1038/s41580-018-0009-y PMID: 29674711 2. Janke C, Magiera MM. The tubulin code and its role in controlling microtubule properties and functions. Nat Rev Mol Cell Biol. 2020; 21(6):307–26. https://doi.org/10.1038/s41580-020-0214-3 PMID: 3. Mitchison T, Kirschner M. Dynamic Instability of Microtubule Growth. Nature. 1984; 312(5991):237– 42. https://doi.org/10.1038/312237a0 PMID: 6504138 4. Battaje RR, Panda D. Lessons from bacterial homolog of tubulin, FtsZ for microtubule dynamics. Endocr Relat Cancer. 2017; 24(9):T1–T21. https://doi.org/10.1530/ERC-17-0118 PMID: 28634179 5. Dyer N. Tubulin and its prokaryotic homologue FtsZ: a structural and functional comparison. Sci Prog. 2009; 92(Pt 2):113–37. https://doi.org/10.3184/003685009X461431 PMID: 19697711 6. Findeisen P, Muhlhausen S, Dempewolf S, Hertzog J, Zietlow A, Carlomagno T, et al. Six subgroups and extensive recent duplications characterize the evolution of the eukaryotic tubulin protein family. Genome Biol Evol. 2014; 6(9):2274–88. https://doi.org/10.1093/gbe/evu187 PMID: 25169981 7. Lowe J, Li H, Downing KH, Nogales E. Refined structure of alpha beta-tubulin at 3.5 A resolution. J Mol Biol. 2001; 313(5):1045–57. https://doi.org/10.1006/jmbi.2001.5077 PMID: 11700061 8. Kobayashi T. Dephosphorylation of tubulin-bound guanosine triphosphate during microtubule assem- bly. J Biochem. 1975; 77(6):1193–7. PMID: 1225903 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 18 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 9. Mitchison TJ. Localization of an Exchangeable Gtp-Binding Site at the Plus End of Microtubules. Sci- ence. 1993; 261(5124):1044–7. https://doi.org/10.1126/science.8102497 PMID: 8102497 10. Wang Z, Sheetz MP. The C-terminus of tubulin increases cytoplasmic dynein and kinesin processivity. Biophys J. 2000; 78(4):1955–64. https://doi.org/10.1016/S0006-3495(00)76743-9 PMID: 10733974 11. Burns RG, Surridge C. Analysis of beta-tubulin sequences reveals highly conserved, coordinated amino acid substitutions. Evidence that these ’hot spots’ are directly involved in the conformational change required for dynamic instability. FEBS Lett. 1990; 271(1–2):1–8. https://doi.org/10.1016/0014- 5793(90)80359-q PMID: 2226794 12. Chakraborti S, Natarajan K, Curiel J, Janke C, Liu J. The emerging role of the tubulin code: From the tubulin molecule to neuronal function and disease. Cytoskeleton. 2016; 73(10):521–50. https://doi.org/ 10.1002/cm.21290 PMID: 26934450 13. Kollman JM, Merdes A, Mourey L, Agard DA. Microtubule nucleation by gamma-tubulin complexes. Nat Rev Mol Cell Biol. 2011; 12(11):709–21. 14. Stathatos GG, Dunleavy JEM, Zenker J, O’Bryan MK. Delta and epsilon tubulin in mammalian devel- opment. Trends in Cell Biology. 2021; 31(9):774–87. https://doi.org/10.1016/j.tcb.2021.03.010 PMID: 15. Bahi-Buisson N, Poirier K, Fourniol F, Saillour Y, Valence S, Lebrun N, et al. The wide spectrum of tubulinopathies: what are the key features for the diagnosis? Brain. 2014; 137:1676–700. https://doi. org/10.1093/brain/awu082 PMID: 24860126 16. Brock S, Stouffs K, Scalais E, D’Hooghe M, Keymolen K, Guerrini R, et al. Tubulinopathies continued: refining the phenotypic spectrum associated with variants in TUBG1. Eur J Hum Genet. 2018; 26 (8):1132–42. https://doi.org/10.1038/s41431-018-0146-y PMID: 29706637 17. Poirier K, Lebrun N, Broix L, Tian G, Saillour Y, Boscheron C, et al. Mutations in TUBG1, DYNC1H1, KIF5C and KIF2A cause malformations of cortical development and microcephaly. Nat Genet. 2013; 45(6):639–+. https://doi.org/10.1038/ng.2613 PMID: 23603762 18. Luscan R, Mechaussier S, Paul A, Tian GL, Gerard X, Defoort-Dellhemmes S, et al. Mutations in TUBB4B Cause a Distinctive Sensorineural Disease. Am J Hum Genet. 2017; 101(6):1006–12. https://doi.org/10.1016/j.ajhg.2017.10.010 PMID: 29198720 19. Smith BN, Ticozzi N, Fallini C, Gkazi AS, Topp S, Kenna KP, et al. Exome-wide Rare Variant Analysis Identifies TUBA4A Mutations Associated with Familial ALS. Neuron. 2014; 84(2):324–31. https://doi. org/10.1016/j.neuron.2014.09.027 PMID: 25374358 20. Stoupa A, Adam F, Kariyawasam D, Strassel C, Gawade S, Szinnai G, et al. TUBB1 mutations cause thyroid dysgenesis associated with abnormal platelet physiology. Embo Mol Med. 2018; 10(12). https://doi.org/10.15252/emmm.201809569 PMID: 30446499 21. Bhagwat S, Dalvi V, Chandrasekhar D, Matthew T, Acharya K, Gajbhiye R, et al. Acetylated alpha- tubulin is reduced in individuals with poor sperm motility. Fertil Steril. 2014; 101(1):95–104 e3. 22. Chen BB, Wang WJ, Peng XD, Jiang HF, Zhang SZ, Li D, et al. The comprehensive mutational and phenotypic spectrum of TUBB8 in female infertility. Eur J Hum Genet. 2019; 27(2):300–7. https://doi. org/10.1038/s41431-018-0283-3 PMID: 30297906 23. Feng RZ, Yan Z, Li B, Yu M, Sang Q, Tian GL, et al. Mutations in TUBB8 cause a multiplicity of pheno- types in human oocytes and early embryos. J Med Genet. 2016; 53(10):662–71. https://doi.org/10. 1136/jmedgenet-2016-103891 PMID: 27273344 24. Abdollahi MR, Morrison E, Sirey T, Molnar Z, Hayward BE, Carr IM, et al. Mutation of the Variant alpha-Tubulin TUBA8 Results in Polymicrogyria with Optic Nerve Hypoplasia. Am J Hum Genet. 2009; 85(5):737–44. https://doi.org/10.1016/j.ajhg.2009.10.007 PMID: 19896110 25. Tian G, Kong XP, Jaglin XH, Chelly J, Keays D, Cowan NJ. A pachygyria-causing alpha-tubulin muta- tion results in inefficient cycling with CCT and a deficient interaction with TBCB. Mol Biol Cell. 2008; 19 (3):1152–61. https://doi.org/10.1091/mbc.e07-09-0861 PMID: 18199681 26. Backwell L, Marsh JA. Diverse Molecular Mechanisms Underlying Pathogenic Protein Mutations: Beyond the Loss-of-Function Paradigm. Annual Review of Genomics and Human Genetics. 2022; 23 (1):null. https://doi.org/10.1146/annurev-genom-111221-103208 PMID: 35395171 27. Aiken J, Buscaglia G, Aiken AS, Moore JK, Bates EA. Tubulin mutations in brain development disor- ders: Why haploinsufficiency does not explain TUBA1A tubulinopathies. Cytoskeleton (Hoboken). 2020; 77(3–4):40–54. https://doi.org/10.1002/cm.21567 PMID: 31574570 28. Breuss M, Heng JIT, Poirier K, Tian GL, Jaglin XH, Qu ZD, et al. Mutations in the beta-Tubulin Gene TUBB5 Cause Microcephaly with Structural Brain Abnormalities. Cell Rep. 2012; 2(6):1554–62. 29. Cederquist GY, Luchniak A, Tischfield MA, Peeva M, Song Y, Menezes MP, et al. An inherited TUBB2B mutation alters a kinesin-binding site and causes polymicrogyria, CFEOM and axon PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 19 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations dysinnervation. Hum Mol Genet. 2012; 21(26):5484–99. https://doi.org/10.1093/hmg/dds393 PMID: 30. Cushion TD, Paciorkowski AR, Pilz DT, Mullins JGL, Seltzer LE, Marion RW, et al. De Novo Mutations in the Beta-Tubulin Gene TUBB2A Cause Simplified Gyral Patterning and Infantile-Onset Epilepsy. Am J Hum Genet. 2014; 94(4):634–41. https://doi.org/10.1016/j.ajhg.2014.03.009 PMID: 24702957 31. Ti SC, Pamula MC, Howes SC, Duellberg C, Cade NI, Kleiner RE, et al. Mutations in Human Tubulin Proximal to the Kinesin-Binding Site Alter Dynamic Instability at Microtubule Plus- and Minus-Ends. Dev Cell. 2016; 37(1):72–84. https://doi.org/10.1016/j.devcel.2016.03.003 PMID: 27046833 32. Tischfield MA, Baris HN, Wu C, Rudolph G, Van Maldergem L, He W, et al. Human TUBB3 Mutations Perturb Microtubule Dynamics, Kinesin Interactions, and Axon Guidance. Cell. 2010; 140(1):74–87. https://doi.org/10.1016/j.cell.2009.12.011 PMID: 20074521 33. Niwa S, Takahashi H, Hirokawa N. beta-Tubulin mutations that cause severe neuropathies disrupt axonal transport. Embo J. 2013; 32(10):1352–64. 34. Park K, Hoff KJ, Wethekam L, Stence N, Saenz M, Moore JK. Kinetically Stabilizing Mutations in Beta Tubulins Create Isotype-Specific Brain Malformations. Front Cell Dev Biol. 2021; 9:765992. https:// doi.org/10.3389/fcell.2021.765992 PMID: 34869359 35. Hebebrand M, Huffmeier U, Trollmann R, Hehr U, Uebe S, Ekici AB, et al. The mutational and pheno- typic spectrum of TUBA1A-associated tubulinopathy. Orphanet J Rare Dis. 2019; 14. https://doi.org/ 10.1186/s13023-019-1020-x PMID: 30744660 36. Livesey BJ, Marsh JA. Interpreting protein variant effects with computational predictors and deep mutational scanning. Disease Models & Mechanisms. 2022; 15(6). https://doi.org/10.1242/dmm. 049510 PMID: 35736673 37. Gerasimavicius L, Livesey BJ, Marsh JA. Loss-of-function, gain-of-function and dominant-negative mutations have profoundly different effects on protein structure. Nat Commun. 2022; 13(1):3895. https://doi.org/10.1038/s41467-022-31686-6 PMID: 35794153 38. Landrum MJ, Lee JM, Benson M, Brown GR, Chao C, Chitipiralla S, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res. 2018; 46(D1):D1062–D7. https:// doi.org/10.1093/nar/gkx1153 PMID: 29165669 39. Pham CL, Morrissette NS. The tubulin mutation database: A resource for the cytoskeleton community. Cytoskeleton. 2019; 76(2):186–91. https://doi.org/10.1002/cm.21514 PMID: 30667171 40. Karczewski KJ, Francioli LC, Tiao G, Cummings BB, Alfo ¨ ldi J, Wang Q, et al. Variation across 141,456 human exomes and genomes reveals the spectrum of loss-of-function intolerance across human pro- tein-coding genes. bioRxiv. 2019:531210. 41. Huzil JT, Chen K, Kurgan L, Tuszynski JA. The roles of beta-tubulin mutations and isotype expression in acquired drug resistance. Cancer Inform. 2007; 3:159–81. PMID: 19455242 42. Nami B, Wang ZX. Genetics and Expression Profile of the Tubulin Gene Superfamily in Breast Cancer Subtypes and Its Relation to Taxane Resistance. Cancers. 2018; 10(8). https://doi.org/10.3390/ cancers10080274 PMID: 30126203 43. Parker AL, Teo WS, McCarroll JA, Kavallaris M. An Emerging Role for Tubulin Isotypes in Modulating Cancer Biology and Chemotherapy Resistance. Int J Mol Sci. 2017; 18(7). https://doi.org/10.3390/ ijms18071434 PMID: 28677634 44. Lek M, Karczewski KJ, Minikel EV, Samocha KE, Banks E, Fennell T, et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature. 2016; 536(7616):285–91. https://doi.org/10.1038/ nature19057 PMID: 27535533 45. Burley K, Westbury SK, Mumford AD. TUBB1 variants and human platelet traits. Platelets. 2018; 29 (2):209–11. https://doi.org/10.1080/09537104.2017.1411587 PMID: 29333906 46. Kunishima S, Kobayashi R, Itoh TJ, Hamaguchi M, Saito H. Mutation of the beta 1-tubulin gene associ- ated with congenital macrothrombocytopenia affecting microtubule assembly. Blood. 2009; 113 (2):458–61. https://doi.org/10.1182/blood-2008-06-162610 PMID: 18849486 47. Kunishima S, Nishimura S, Suzuki H, Imaizumi M, Saito H. TUBB1 mutation disrupting microtubule assembly impairs proplatelet formation and results in congenital macrothrombocytopenia. Eur J Hae- matol. 2014; 92(4):276–82. https://doi.org/10.1111/ejh.12252 PMID: 24344610 48. Lecine P, Italiano JE Jr., Kim SW, Villeval JL, Shivdasani RA. Hematopoietic-specific beta 1 tubulin participates in a pathway of platelet biogenesis dependent on the transcription factor NF-E2. Blood. 2000; 96(4):1366–73. PMID: 10942379 49. Wang D, Villasante A, Lewis SA, Cowan NJ. The mammalian beta-tubulin repertoire: hematopoietic expression of a novel, heterologous beta-tubulin isotype. J Cell Biol. 1986; 103(5):1903–10. https:// doi.org/10.1083/jcb.103.5.1903 PMID: 3782288 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 20 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 50. Chen BB, Li B, Li D, Yan Z, Mao XY, Xu Y, et al. Novel mutations and structural deletions in TUBB8: expanding mutational and phenotypic spectrum of patients with arrest in oocyte maturation, fertiliza- tion or early embryonic development. Hum Reprod. 2017; 32(2):457–64. https://doi.org/10.1093/ humrep/dew322 PMID: 27989988 51. Feng RZ, Sang Q, Kuang YP, Sun XX, Yan Z, Zhang SZ, et al. Mutations in TUBB8 and Human Oocyte Meiotic Arrest. New Engl J Med. 2016; 374(3):223–32. https://doi.org/10.1056/ NEJMoa1510791 PMID: 26789871 52. Huang LL, Tong XH, Luo LH, Zheng SX, Jin RT, Fu YY, et al. Mutation analysis of the TUBB8 gene in nine infertile women with oocyte maturation arrest. Reprod Biomed Online. 2017; 35(3):305–10. https://doi.org/10.1016/j.rbmo.2017.05.017 PMID: 28652098 53. Xiang JJ, Wang W, Qian CF, Xue JY, Wang T, Li HB, et al. Human oocyte maturation arrest caused by a novel missense mutation in TUBB8. J Int Med Res. 2018; 46(9):3759–64. https://doi.org/10.1177/ 0300060518778638 PMID: 29877102 54. Levy ED. A Simple Definition of Structural Regions in Proteins and Its Use in Analyzing Interface Evo- lution. J Mol Biol. 2010; 403(4):660–70. https://doi.org/10.1016/j.jmb.2010.09.028 PMID: 20868694 55. Cushion TD, Dobyns WB, Mullins JGL, Stoodley N, Chung SK, Fry AE, et al. Overlapping cortical mal- formations and mutations in TUBB2B and TUBA1A. Brain. 2013; 136:536–48. https://doi.org/10.1093/ brain/aws338 PMID: 23361065 56. Erro R, Hersheson J, Ganos C, Mencacci NE, Stamelou M, Batla A, et al. H-ABC syndrome and DYT4: Variable expressivity or pleiotropy of TUBB4 mutations? Movement Disord. 2015; 30(6):828– 33. https://doi.org/10.1002/mds.26129 PMID: 25545912 57. Romaniello R, Arrigoni F, Panzeri E, Poretti A, Micalizzi A, Citterio A, et al. Tubulin-related cerebellar dysplasia: definition of a distinct pattern of cerebellar malformation (vol 27, pg 5080, 2017). Eur Radiol. 2017; 27(12):5093–. 58. Rice LM, Montabana EA, Agard DA. The lattice as allosteric effector: Structural studies of alpha beta- and gamma-tubulin clarify the role of GTP in microtubule assembly. P Natl Acad Sci USA. 2008; 105 (14):5378–83. 59. Wieczorek M, Urnavicius L, Ti SC, Molloy KR, Chait BT, Kapoor TM. Asymmetric Molecular Architec- ture of the Human gamma-Tubulin Ring Complex. Cell. 2020; 180(1):165–+. 60. Guerois R, Nielsen JE, Serrano L. Predicting changes in the stability of proteins and protein com- plexes: A study of more than 1000 mutations. J Mol Biol. 2002; 320(2):369–87. https://doi.org/10. 1016/S0022-2836(02)00442-4 PMID: 12079393 61. Gerasimavicius L, Liu X, Marsh JA. Identification of pathogenic missense mutations using protein sta- bility predictors. Sci Rep. 2020; 10(1):15387. https://doi.org/10.1038/s41598-020-72404-w PMID: 62. Livesey BJ, Marsh JA. Using deep mutational scanning to benchmark variant effect predictors and identify disease mutations. Molecular Systems Biology. 2020; 16(7):e9380. https://doi.org/10.15252/ msb.20199380 PMID: 32627955 63. Fallet-Bianco C, Laquerriere A, Poirier K, Razavi F, Guimiot F, Dias P, et al. Mutations in tubulin genes are frequent causes of various foetal malformations of cortical development including microlissence- phaly. Acta Neuropathol Com. 2014;2. https://doi.org/10.1186/2051-5960-2-69 PMID: 25059107 64. Oegema R, Cushion TD, Phelps IG, Chung SK, Dempsey JC, Collins S, et al. Recognizable cerebellar dysplasia associated with mutations in multiple tubulin genes. Hum Mol Genet. 2015; 24(18):5313–25. https://doi.org/10.1093/hmg/ddv250 PMID: 26130693 65. Riesselman AJ, Ingraham JB, Marks DS. Deep generative models of genetic variation capture the effects of mutations. Nat Methods. 2018; 15(10):816–+. https://doi.org/10.1038/s41592-018-0138-4 PMID: 30250057 66. Romaniello R, Zucca C, Arrigoni F, Bonanni P, Panzeri E, Bassi MT, et al. Epilepsy in Tubulinopathy: Personal Series and Literature Review. Cells. 2019; 8(7). https://doi.org/10.3390/cells8070669 PMID: 67. Schatz PJ, Solomon F, Botstein D. Isolation and Characterization of Conditional-Lethal Mutations in the Tub1 Alpha-Tubulin Gene of the Yeast Saccharomyces-Cerevisiae. Genetics. 1988; 120(3):681– 95. https://doi.org/10.1093/genetics/120.3.681 PMID: 3066684 68. Livesey BJ, Marsh JA. The properties of human disease mutations at protein interfaces. PLOS Computational Biology. 2022; 18(2):e1009858. https://doi.org/10.1371/journal.pcbi.1009858 PMID: 69. Hausrat TJ, Radwitz J, Lombino FL, Breiden P, Kneussel M. Alpha- and beta-tubulin isotypes are dif- ferentially expressed during brain development. Dev Neurobiol. 2021; 81(3):333–50. https://doi.org/ 10.1002/dneu.22745 PMID: 32293117 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 21 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 70. Grimm DG, Azencott CA, Aicheler F, Gieraths U, MacArthur DG, Samocha KE, et al. The Evaluation of Tools Used to Predict the Impact of Missense Variants Is Hindered by Two Types of Circularity. Hum Mutat. 2015; 36(5):513–23. https://doi.org/10.1002/humu.22768 PMID: 25684150 71. Krajka V, Vulinovic F, Genova M, Tanzer K, Jijumon AS, Bodakuntla S, et al. H-ABC- and dystonia- causing TUBB4A mutations show distinct pathogenic effects. Sci Adv. 2022; 8(10):eabj9229. https:// doi.org/10.1126/sciadv.abj9229 PMID: 35275727 72. Leca I, Phillips AW, Hofer I, Landler L, Ushakova L, Cushion TD, et al. A proteomic survey of microtu- bule-associated proteins in a R402H TUBA1A mutant mouse. Plos Genet. 2020; 16(11):e1009104. https://doi.org/10.1371/journal.pgen.1009104 PMID: 33137126 73. Ti SC, Alushin GM, Kapoor TM. Human beta-Tubulin Isotypes Can Regulate Microtubule Protofila- ment Number and Stability. Dev Cell. 2018; 47(2):175–90 e5. 74. Online Mendelian Inheritance in Man, OMIM. [Internet]. McKusick-Nathans Institute of Genetic Medi- cine, Johns Hopkins University (Baltimore, MD). 2019. Available from: https://omim.org/. 75. Schrodinger LLC. The PyMOL Molecular Graphics System, Version 2.3.2. 2015. 76. Cheeseman IM, Desai A. A combined approach for the localization and tandem affinity purification of protein complexes from metazoans. Science’s STKE: signal transduction knowledge environment. 2005; 2005(266):pl1. https://doi.org/10.1126/stke.2662005pl1 PMID: 15644491 77. Goldberg IG, Allan C, Burel J-M, Creager D, Falconi A, Hochheiser H, et al. The Open Microscopy Environment (OME) Data Model and XML file: open tools for informatics and quantitative analysis in biological imaging. Genome Biology. 2005; 6(5):R47. https://doi.org/10.1186/gb-2005-6-5-r47 PMID: 78. Rueden CT, Schindelin J, Hiner MC, DeZonia BE, Walter AE, Arena ET, et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinformatics. 2017; 18(1):529. https://doi.org/10. 1186/s12859-017-1934-z PMID: 29187165 79. Aiken J, Buscaglia G, Bates EA, Moore JK. The alpha-Tubulin gene TUBA1A in Brain Development: A Key Ingredient in the Neuronal Isotype Blend. J Dev Biol. 2017; 5(3). 80. Boissel S, Fallet-Bianco C, Chitayat D, Kremer V, Nassif C, Rypens F, et al. Genomic study of severe fetal anomalies and discovery of GREB1L mutations in renal agenesis. Genet Med. 2018; 20(7):745– 53. https://doi.org/10.1038/gim.2017.173 PMID: 29261186 81. Gardner JF, Cushion TD, Niotakis G, Olson HE, Grant PE, Scott RH, et al. Clinical and Functional Characterization of the Recurrent TUBA1A p.(Arg2His) Mutation. Brain Sci. 2018; 8(8). https://doi.org/ 10.3390/brainsci8080145 PMID: 30087272 82. Keays DA, Tian G, Poirier K, Huang GJ, Siebold C, Cleak J, et al. Mutations in alpha-tubulin cause abnormal neuronal migration in mice and lissencephaly in humans. Cell. 2007; 128(1):45–57. https:// doi.org/10.1016/j.cell.2006.12.017 PMID: 17218254 83. Poirier K, Keays DA, Francis F, Saillour Y, Bahi N, Manouvrier S, et al. Large spectrum of lissence- phaly and pachygyria phenotypes resulting from de novo missense mutations in tubulin alpha 1A (TUBA1A). Hum Mutat. 2007; 28(11):1055–64. https://doi.org/10.1002/humu.20572 PMID: 17584854 84. Martello LA, Verdier-Pinard P, Shen HJ, He LF, Torres K, Orr GA, et al. Elevated levels of microtubule destabilizing factors in a Taxol-resistant/dependent A549 cell line with an alpha-tubulin mutation. Can- cer Res. 2003; 63(6):1207–13. PMID: 12649178 85. Poruchynsky MS, Kim JH, Nogales E, Annable T, Loganzo F, Greenberger LM, et al. Tumor cells resistant to a microtubule-depolymerizing hemiasterlin analogue, HTI-286, have mutations in alpha- or beta-tubulin and increased microtubule stability. Biochemistry-Us. 2004; 43(44):13944–54. https://doi. org/10.1021/bi049300+ PMID: 15518543 86. Yang CPH, Verdier-Pinard P, Wang F, Lippaine-Horvath E, He LF, Li DS, et al. A highly epothilone B- resistant A549 cell line with mutations in tubulin that confer drug dependence. Mol Cancer Ther. 2005; 4(6):987–95. https://doi.org/10.1158/1535-7163.MCT-05-0024 PMID: 15956256 87. Wang J, Chen W, Wei WW, Lou JY. Oncogene TUBA1C promotes migration and proliferation in hepa- tocellular carcinoma and predicts a poor prognosis. Oncotarget. 2017; 8(56):96215–24. https://doi. org/10.18632/oncotarget.21894 PMID: 29221200 88. Alazami AM, Patel N, Shamseldin HE, Anazi S, Al-Dosari MS, Alzahrani F, et al. Accelerating Novel Candidate Gene Discovery in Neurogenetic Disorders via Whole-Exome Sequencing of Prescreened Multiplex Consanguineous Families. Cell Rep. 2015; 10(2):148–61. https://doi.org/10.1016/j.celrep. 2014.12.015 PMID: 25558065 89. Sferra A, Fattori F, Rizza T, Flex E, Bellacchio E, Bruselles A, et al. Defective kinesin binding of TUBB2A causes progressive spastic ataxia syndrome resembling sacsinopathy. Hum Mol Genet. 2018; 27(11):1892–904. https://doi.org/10.1093/hmg/ddy096 PMID: 29547997 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 22 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 90. Amrom D, Tanyalcin I, Verhelst H, Deconinck N, Brouhard GJ, Decarie JC, et al. Polymicrogyria with dysmorphic basal ganglia? Think tubulin! Clin Genet. 2014; 85(2):178–83. https://doi.org/10.1111/cge. 12141 PMID: 23495813 91. Guerrini R, Mei D, Cordelli DM, Pucatti D, Franzoni E, Parrini E. Symmetric polymicrogyria and pachy- gyria associated with TUBB2B gene mutations. Eur J Hum Genet. 2012; 20(9):995–8. https://doi.org/ 10.1038/ejhg.2012.21 PMID: 22333901 92. Jaglin XH, Poirier K, Saillour Y, Buhler E, Tian G, Bahi-Buisson N, et al. Mutations in the beta-tubulin gene TUBB2B result in asymmetrical polymicrogyria. Nat Genet. 2009; 41(6):746–52. https://doi.org/ 10.1038/ng.380 PMID: 19465910 93. MacKinnon S, Oystreck DT, Andrews C, Chan WM, Hunter DG, Engle EC. Diagnostic Distinctions and Genetic Analysis of Patients Diagnosed with Moebius Syndrome. Ophthalmology. 2014; 121(7):1461– 8. https://doi.org/10.1016/j.ophtha.2014.01.006 PMID: 24612975 94. Minoura I, Takazaki H, Ayukawa R, Saruta C, Hachikubo Y, Uchimura S, et al. Reversal of axonal growth defects in an extraocular fibrosis model by engineering the kinesin-microtubule interface. Nat Commun. 2016; 7. 95. Arai-Ichinoi N, Uematsu M, Sato R, Suzuki T, Kudo H, Kikuchi A, et al. Genetic heterogeneity in 26 infants with a hypomyelinating leukodystrophy. Hum Genet. 2016; 135(1):89–98. https://doi.org/10. 1007/s00439-015-1617-7 PMID: 26597493 96. Blumkin L, Halevy A, Ben-Ami-Raichman D, Dahari D, Haviv A, Sarit C, et al. Expansion of the spec- trum of TUBB4A-related disorders: a new phenotype associated with a novel mutation in the TUBB4A gene. Neurogenetics. 2014; 15(2):107–13. https://doi.org/10.1007/s10048-014-0392-2 PMID: 97. Carvalho D, Santos S, Martins B, Marques FP. TUBB4A novel mutation reinforces the genotype-phe- notype correlation of hypomyelination with atrophy of the basal ganglia and cerebellum. Brain. 2015;138. https://doi.org/10.1093/brain/awu242 PMID: 25168210 98. Curiel J, Bey GR, Takanohashi A, Bugiani M, Fu XQ, Wolf NI, et al. TUBB4A mutations result in spe- cific neuronal and oligodendrocytic defects that closely match clinically distinct phenotypes. Hum Mol Genet. 2017; 26(22):4506–18. https://doi.org/10.1093/hmg/ddx338 PMID: 28973395 99. Ferreira C, Poretti A, Cohen J, Hamosh A, Naidu S. Novel TUBB4A mutations and expansion of the neuroimaging phenotype of hypomyelination with atrophy of the basal ganglia and cerebellum (H- ABC). Am J Med Genet A. 2014; 164(7):1802–7. https://doi.org/10.1002/ajmg.a.36526 PMID: 100. Hamilton EM, Polder E, Vanderver A, Naidu S, Schiffmann R, Fisher K, et al. Hypomyelination with atrophy of the basal ganglia and cerebellum: further delineation of the phenotype and genotype-phe- notype correlation. Brain. 2014; 137:1921–30. https://doi.org/10.1093/brain/awu110 PMID: 24785942 101. Hersheson J, Mencacci NE, Davis M, MacDonald N, Trabzuni D, Ryten M, et al. Mutations in the auto- regulatory domain of beta-tubulin 4a cause hereditary dystonia. Ann Neurol. 2013; 73(4):546–53. 102. Isakov O, Lev D, Blumkin L, Celniker G, Leshinsky-Silver E, Shomron N. Crowdfunding Effort Identi- fies the Causative Mutation in a Patient with Nystagmus, Microcephaly, Dystonia and Hypomyelina- tion. J Genet Genomics. 2015; 42(2):79–81. https://doi.org/10.1016/j.jgg.2014.12.004 PMID: 103. Kancheva D, Chamova T, Guergueltcheva V, Mitev V, Azmanov DN, Kalaydjieva L, et al. Mosaic dom- inant TUBB4A mutation in an inbred family with complicated hereditary spastic paraplegia. Movement Disord. 2015; 30(6):854–8. https://doi.org/10.1002/mds.26196 PMID: 25772097 104. Lohmann K, Wilcox RA, Winkler S, Ramirez A, Rakovic A, Park JS, et al. Whispering dysphonia (DYT4 dystonia) is caused by a mutation in the TUBB4 gene. Ann Neurol. 2013; 73(4):537–45. https:// doi.org/10.1002/ana.23829 PMID: 23595291 105. Lu Y, Ondo Y, Shimojima K, Osaka H, Yamamoto T. A novel TUBB4A mutation G96R identified in a patient with hypomyelinating leukodystrophy onset beyond adolescence. Hum Genome Var. 2017; 4:17035. https://doi.org/10.1038/hgv.2017.35 PMID: 28791129 106. Miyatake S, Osaka H, Shiina M, Sasaki M, Takanashi J, Haginoya K, et al. Expanding the phenotypic spectrum of TUBB4A-associated hypomyelinating leukoencephalopathies. Neurology. 2014; 82 (24):2230–7. https://doi.org/10.1212/WNL.0000000000000535 PMID: 24850488 107. Pizzino A, Pierson TM, Guo Y, Helman G, Fortini S, Guerrero K, et al. TUBB4A de novo mutations cause isolated hypomyelination. Neurology. 2014; 83(10):898–902. https://doi.org/10.1212/WNL. 0000000000000754 PMID: 25085639 108. Purnell SM, Bleyl SB, Bonkowsky JL. Clinical exome sequencing identifies a novel TUBB4A mutation in a child with static hypomyelinating leukodystrophy. Pediatr Neurol. 2014; 50(6):608–11. https://doi. org/10.1016/j.pediatrneurol.2014.01.051 PMID: 24742798 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 23 / 24 PLOS COMPUTATIONAL BIOLOGY Understanding and predicting tubulin mutations 109. Pyle A, Smertenko T, Bargiela D, Griffin H, Duff J, Appleton M, et al. Exome sequencing in undiag- nosed inherited and sporadic ataxias. Brain. 2015; 138:276–83. https://doi.org/10.1093/brain/awu348 PMID: 25497598 110. Sagnelli A, Magri S, Farina L, Chiapparini L, Marotta G, Tonduti D, et al. Early-onset progressive spas- tic paraplegia caused by a novel TUBB4A mutation: brain MRI and FDG-PET findings. J Neurol. 2016; 263(3):591–3. https://doi.org/10.1007/s00415-016-8020-8 PMID: 26810722 111. Simons C, Wolf NI, McNeil N, Caldovic L, Devaney JM, Takanohashi A, et al. A de novo mutation in the beta-tubulin gene TUBB4A results in the leukoencephalopathy hypomyelination with atrophy of the basal ganglia and cerebellum. Am J Hum Genet. 2013; 92(5):767–73. 112. Vanderver A, Simons C, Helman G, Crawford J, Wolf NI, Bernard G, et al. Whole Exome Sequencing in Patients with White Matter Abnormalities. Ann Neurol. 2016; 79(6):1031–7. https://doi.org/10.1002/ ana.24650 PMID: 27159321 113. Wang WW, Zhang HX, Wang XM, Patterson J, Winter P, Graham K, et al. Novel mutations involving beta I-, beta IIA-, or beta IVB-tubulin isotypes with functional resemblance to beta III-tubulin in breast cancer. Protoplasma. 2017; 254(3):1163–73. 114. Bowling KM, Thompson ML, Amaral MD, Finnila CR, Hiatt SM, Engel KL, et al. Genomic diagnosis for children with intellectual disability and/or developmental delay. Genome Med. 2017;9. 115. Isrie M, Breuss M, Tian GL, Hansen AH, Cristofoli F, Morandell J, et al. Mutations in Either TUBB or MAPRE2 Cause Circumferential Skin Creases Kunze Type. Am J Hum Genet. 2015; 97(6):790–800. https://doi.org/10.1016/j.ajhg.2015.10.014 PMID: 26637975 116. Leonard NJ. A second patient with MCA/MR syndrome with multiple circumferential skin creases. Am J Med Genet. 2002; 112(1):91–4. https://doi.org/10.1002/ajmg.10665 PMID: 12239728 117. Madrigal I, Rabionet R, Alvarez-Mora MI, Sanchez A, Rodriguez-Revenga L, Estivill X, et al. Spectrum of clinical heterogeneity of beta-tubulin TUBB5 gene mutations. Gene. 2019; 695:12–7. 118. Ulucan H, Koparir E, Koparir A, Karaca E, Emre R, Gezdirici A, et al. Circumferential skin folds and multiple anomalies: confirmation of a distinct autosomal recessive Michelin tire baby syndrome. Clin Dysmorphol. 2013; 22(2):87–90. https://doi.org/10.1097/MCD.0b013e32835cd5df PMID: 23324645 119. Wiszniewski W, Gawlinski P, Gambin T, Bekiesinska-Figatowska M, Obersztyn E, Antczak-Marach D, et al. Comprehensive genomic analysis of patients with disorders of cerebral cortical development. Eur J Hum Genet. 2018; 26(8):1121–31. https://doi.org/10.1038/s41431-018-0137-z PMID: 29706646 120. Fazeli W, Herkenrath P, Stiller B, Neugebauer A, Fricke J, Lang-Roth R, et al. A TUBB6 mutation is associated with autosomal dominant non-progressive congenital facial palsy, bilateral ptosis and velo- pharyngeal dysfunction. Hum Mol Genet. 2017; 26(20):4055–66. https://doi.org/10.1093/hmg/ddx296 PMID: 29016863 121. Leask A, Stearns T. Expression of amino- and carboxyl-terminal gamma- and alpha-tubulin mutants in cultured epithelial cells. J Biol Chem. 1998; 273(5):2661–8. https://doi.org/10.1074/jbc.273.5.2661 PMID: 9446570 PLOS Computational Biology | https://doi.org/10.1371/journal.pcbi.1010611 October 7, 2022 24 / 24

Journal

PLoS Computational Biology – Public Library of Science (PLoS) Journal

Published: Oct 7, 2022

There are no references for this article.