Rare missense variants outnumber common ones, with 85% of non-synonymous variants displaying a minor allele frequency of less than 0.5% (Abecasis et al., 2012), and 200–300 such alleles per sequenced individual (Bamshad et al., 2011). As each specific variant is present at a very low frequency within the population, the impact on human health is hard to assess. Collectively however, rare variants are thought to be a significant component of the ‘missing heritability’ paradigm and their neglected contribution may explain why only a fraction of inherited diseases are genetically accounted for Maher (2008). Classic GWAS approaches lack the power to correlate trait heritability with rare coding alleles (Auer and Lettre, 2015; Bomba et al., 2017); indeed, they are largely limited to the identification of common variants with small effect sizes, particularly those within regulatory regions (Astle et al., 2016). Yet, the clear inverse correlation between allele frequency and trait impact suggests that rare variants are more likely to be disease-causing (Kryukov et al., 2007; Park et al., 2011). Common variants have undergone purifying selection and are therefore more likely to be benign. In contrast, a higher proportion of rare variants will be functionally damaging as they mostly comprise recent or de novo mutations (Keinan and Clark, 2012) on which selection has not yet acted. Multiple different rare de novo mutations are thought to underlie the genetics of many complex human disorders including schizophrenia, epilepsy, lipid metabolism disorder, and inflammatory disease (Andrews et al., 2013; McClellan and King, 2010). Estimates from the 1000 Genomes Project suggest that 40% of rare missense mutations are damaging compared to 5% of common variants (Abecasis et al., 2010).

While the advent of next-generation sequencing (NGS) has made obtaining human sequence data straightforward and inexpensive, linking genotype to phenotype is far less trivial. Sophisticated computational tools have been produced in order to predict the functional impact of missense variants. Early prediction methods typically utilized a combination of sequence conservation and amino-acid properties while newer tools typically employ ensemble methods that integrate a large number of diverse features using machine learning. Unfortunately, these predictions are not always prognostic of disease severity or outcome. A study of the cystic fibrosis gene CFTR found a poor correlation between predicted functional impact and disease (Dorfman et al., 2010), while in silico classification of rare BRCA1/2 mutations was not predictive of pathogenicity (Ernst et al., 2018). A direct assessment of multiple computational methods, carried out as part of the Critical Assessment of Genome Interpretation, compared phenotypic predictions with an empirical dataset quantifying the ability of SUMO-conjugating enzyme UBE2I variants to rescue the growth of S. cerevisiae (Zhang et al., 2017). While most predictors identified deleterious mutations at conserved positions, they were less accurate at non-conserved sites and gave false-positive predictions.

The most direct way to assess the phenotypic contribution of a missense mutation is through empirical testing. A landmark study introduced 33 de novo missense mutations by random mutagenesis into immune genes and measured the impact on lymphocyte subsets in homozygous mice (Miosge et al., 2015). Strikingly, only 20% of variants predicted by computational methods to be deleterious gave an observable phenotype. The same study found that while approximately 50% of de novo missense mutations within the same species were predicted to be functionally impaired, this compared with only 5% of the variants found between-species. This would suggest that many variants possess near neutral phenotypes not discernible in vivo yet sufficiently impactful to undergo purifying selection. Moreover, it highlights a crucial general question: are predicted deleterious mutations actually often neutral or does phenotypic characterization fail to capture their impact?

We decided to address this question by investigating how naturally occurring variants impact the cytosolic antibody receptor TRIM21, using multiple molecular and cellular assays that independently quantify protein stability, function and phenotype. TRIM21 intercepts incoming antibody-coated pathogens during cellular infection and causes them to be degraded by the proteasome. TRIM21 also activates immune signaling pathways, including NF-κB, although this is tightly regulated to prevent inopportune inflammation. These disparate complex functions are achieved using multiple component domains and by recruiting a range of cofactors. We determined the protein stability, ligand binding, viral neutralization, innate immune signaling, protein expression and dominant-negative potential of rare missense variants in TRIM21 identified from the gnomAD consortium and 1000 Genomes Project (Lek et al., 2016; Sudmant et al., 2015). We characterized variant activity in cells from 1000 Genomes volunteers, by CRISPR gene editing and through a novel reconstitution system that recapitulates endogenous expression and regulation.

To provide a comprehensive description of natural variation in TRIM21, we considered all missense variants from gnomAD, which comprises 125,748 exome sequences and 15,708 genome sequences, including data from the 1000 Genomes Project. Analysis of these variation data revealed a total of 277 rare missense variants, defined as having less than 1% minor allele frequency (MAF) and no common variants (≥1% MAF). Of these rare variants, the highest allele frequency is 0.32% for V50I, with 882 heterozygotes and nine homozygotes. We compared non-synonymous genetic variability within the TRIM21 gene to related TRIMs and gene families with high (HDAC and actin genes) and low (HLAs) levels of conservation. Degree (frequency) of variation was quantified as the number of missense variants per 100 amino-acid positions (AAs) to take account of differing gene size (Figure 1). There was a substantial difference in the frequency of common variants between genes, which were more prevalent in immune genes STING, TRIM20 and TRIM5 than in TRIM21 (Figure 1A). Indeed, TRIM21, is more invariant than HDAC and similar to actin in its fixation of a common allele. Both STING (Mozzi et al., 2015; Yi et al., 2013) and TRIM5 (Sawyer et al., 2005) are thought to be under positive selection, as evidenced by their high dN/dS in interspecies sequence comparisons. In human populations, variants in these genes display geographic variation, suggesting that exposure to pathogens explains the increased frequency of common variants associated with these genes (Figure 1A). TRIM20 is associated with familial Mediterranean fever and the existence of common deleterious geographic variants is thought to be a result of past heterozygous advantage (Manukyan and Aminov, 2016). In contrast, TRIM21 has no common missense variants, consistent with purifying selection. This difference between TRIM5 and TRIM21, despite both detecting incoming viruses and restricting replication, is consistent with their different binding mechanisms. TRIM5 binds directly to retroviral capsids, which are continually evolving, whereas TRIM21 binds viral capsids via attached antibodies, which are conserved within their Fc region. There was noticeably less variation in the frequency of rare variants between genes (Figure 1B). This is consistent with rare human SNP frequencies being significantly impacted by de novo mutation rates (Pritchard, 2001).

Figure 1

Download asset Open asset

Figure 1—source data 1

https://doi.org/10.7554/eLife.48339.003

Download elife-48339-fig1-data1-v1.xlsx

Within the 1000 Genomes Project (2504 individuals), there are 22 rare variants in TRIM21, all of which are heterozygous, although for four of them, there are individuals in gnomAD who are homozygous (E38G, V50I, R57C and Q88K). We focused on the 1000 Genomes panel of variants because immortalized cell lines are available from these volunteers, allowing each haplotype to be tested in its naturally occurring genomic context. We first assessed whether identified variants are likely to perturb TRIM21 activity by utilizing a panel of phenotypic prediction algorithms, including PolyPhen-2 (Adzhubei et al., 2010), SIFT (Sim et al., 2012), CADD (Rentzsch et al., 2019), REVEL (Ioannidis et al., 2016), MetaLR (Dong et al., 2015) and MutationAssessor (Reva et al., 2011) (Supplementary file 1). Of the 22 variants, 21 are missense and one is a premature stop codon. PolyPhen-2 predicts 7/21 variants to be possibly or probably deleterious, SIFT 4/21, while CADD, REVEL and MetaLR predict only one variant, F446I, to be deleterious with the remaining either tolerated or benign (Supplementary file 1). Taken together, there are noticeably fewer predicted deleterious variants in TRIM21 than have been reported for other immune genes (Miosge et al., 2015) or for hominid genes in general (Eyre-Walker and Keightley, 1999). We hypothesized that some of this difference may be due to the increase in the number of genomes now available. To test this, in addition to SIFT and PolyPhen-2 predictions from the current version of Ensembl, we obtained equivalent predictions based on earlier versions of the underlying sequence databases (Materials and methods) and found that this caused both SIFT and PolyPhen-2 to predict a higher proportion of deleterious variants (8/21 and 9/21, respectively).

Next, we sought to empirically determine the impact of natural variants on TRIM21 function and assess the accuracy of phenotypic prediction. The 22 variants found in the 1000 Genomes Project are distributed amongst all 4 TRIM21 domains – the RING, B-Box, coiled-coil and PRYSPRY. Structural data is available for all domains (albeit the coiled-coil is based on a model of TRIM25) and the location of each variant is marked (Figure 2).

Figure 2

Download asset Open asset

We began by focussing on the PRYSPRY as this is the only globular domain in TRIM21 and is responsible for IgG binding. We expressed each of the seven variants and compared their binding of IgG Fc with wild-type by ITC (Figure 3, Figure 3—figure supplement 1). Wild-type PRYSPRY bound IgG with an affinity of 9.4 nM (Figure 3B). Most variants maintained binding affinity within error of wild-type, with K455E being the weakest at 12 nM and A390V and Q470K the strongest at 1.7 and 1.2 nM respectively (Figure 3B). These results are in contrast to predictions, where SIFT categorized 4 of the PRYSPRY variants as deleterious, PolyPhen-2 predicted three to be possibly damaging and the other algorithms only F446I.

Figure 3 with 2 supplements see all

Download asset Open asset

Figure 3—source data 1

https://doi.org/10.7554/eLife.48339.008

Download elife-48339-fig3-data1-v1.xlsx

Importantly, whilst binding was maintained for all variants, we noted significant changes in protein behavior, in particular with F446I. Most variants expressed with a yield of 2 mg/l and could be concentrated to 500 µM. In contrast, F446I gave yields of 10–100 µg/l and could not be concentrated to levels > 30 µM. G440R also had reduced expression at 0.1 mg/l (Figure 3—figure supplement 2). This suggests that while IgG binding was unaffected, PRYSPRY variants could impact function indirectly by altering protein stability. The PRYSPRY domain of TRIM21 is a β-sandwich with a classic globular fold. Several of the variants map to, or are in the vicinity of, the hydrophobic core. To directly determine the impact of each variant on PRYSPRY stability we measured the melting temperature (Tm) of all variants using a Prometheus differential scanning fluorimeter (DSF) (Figure 4A). As the Prometheus DSF uses natural tryptophan fluorescence rather than an extrinsic reporter dye, it avoids the potential issue of changes in dye binding. Measurement of wild-type PRYSPRY reveals it is a moderately stable protein with a Tm of ~48°C. We observed considerable variation in stability between naturally occurring PRYSPRY variants (Figure 4B), in contrast with their similarities in IgG binding. N297H displayed a slightly increased stability of 2°C, while A390V, G440R and F446I were all substantially less stable, consistent with their lower yield from E. coli.

Figure 4

Download asset Open asset

Figure 4—source data 1

https://doi.org/10.7554/eLife.48339.010

Download elife-48339-fig4-data1-v1.xlsx

To assess the accuracy of each prediction algorithm, we compared the empirically determined ΔΔG of natural PRYSPRY variants with the calculated prediction scores. Overall, there was remarkable correlation between observed and predicted effects, particularly for REVEL and MetaLR (Figure 4C). These programs therefore not only identified deleterious variants without calling false positives but also correctly ranked variants in terms of their relative stability. PolyPhen-2, while not the most accurate in this analysis, still nevertheless predicts as its potentially damaging variants those three with the lowest stability. Of all PRYSPRY variants tested, F446I was most unstable with a dramatic reduction in Tm of ~10°C. This result is consistent with the observed poor behavior of this protein in terms of bacterial expression and purification. Moreover, F446I is the most buried of all the variants, with its side chain packed close to the center of the core. Most importantly, this result is consistent with the identification of F446I by CADD, REVEL and MetaLR as the only deleterious TRIM21 variant (Supplementary file 1).

In contrast to the PRYSPRY, both the RING and B-Box domains of TRIM21 are not globular folds but Zn-fingers each of which is stabilized by two Zn ions (Figure 2B). This makes phenotypic prediction of variants in these domains difficult. However, both domains are essential to TRIM21 function. The RING domain of TRIM21 is a highly active E3 ubiquitin ligase but is tightly regulated to prevent constitutive ubiquitination and signaling down the NF-κB pathway. RING function is regulated by the B-Box domain, which is autoinhibitory and blocks E2 enzyme recruitment (Dickson et al., 2018). To investigate whether RING-Box variants have an impact on the ability of TRIM21 to activate NF-κB signaling in response to viral infection, we needed a system for expressing TRIM21 variants in a TRIM21 KO cell line.

While plasmid transfection is a convenient method for over-expression of proteins in mammalian cell lines, viral promoters (e.g. CMV) typically induce expression that is orders of magnitude greater than endogenous levels, potentially masking subtle differences in activity, intrinsic stability and cellular turnover. We therefore sought a system in which we could ectopically express our panel of natural variants at endogenous levels and retain native transcriptional regulation. To this end, we constructed a lentiviral transduction vector utilizing 2 kb of endogenous TRIM21 sequence predicted to contain native promoter elements (Figure 5A). We removed the endogenous copy of TRIM21 from 293Ts by CRISPR-Cas9 and made stable cell lines of each variant by lentiviral transduction. Using our native transduction cassette, we could restore expression of TRIM21 back to endogenous levels in the three different cell lines tested (Figure 5C, Figure 5—figure supplement 1A–C). The expression level was substantially lower than that driven by the SFFV promoter (Figure 5C). Moreover, reconstituted 293T cells upregulated TRIM21 in response to interferon-alpha, similar to unmodified parent cells (Figure 5B).

Figure 5 with 2 supplements see all

Download asset Open asset

Figure 5—source data 1

https://doi.org/10.7554/eLife.48339.014

Download elife-48339-fig5-data1-v1.xlsx

A specific problem with TRIM proteins is their propensity to spontaneously form higher-order structures when over-expressed (Fletcher et al., 2018; Reymond et al., 2001). Over-expression of TRIM21 has previously been shown to cause large rod-shaped structures unlike the diffuse distribution of endogenous protein (Rhodes et al., 2002). We therefore also tested whether our bespoke native-promoter driven system properly recapitulates the ability of TRIM21 to detect incoming antibody-coated virions during infection. Using an mCherry-TRIM21 fusion, we observed good colocalization with AdV5-IgG complexes (Figure 5—figure supplement 2). This colocalization was specific and dependent on the interaction between TRIM21 PRYSPRY and IgG Fc because the H433A mutation in IgG that abrogates TRIM21 binding also prevented colocalization. Comparing the TRIM21 expression levels in all 22 variant stable cell lines revealed that reconstitution was successful in most cases within twofold of wild-type (Figure 5D, Figure 5—figure supplement 1C). The obvious outliers were A390V, G440R and F446I, which were substantially lower in expression. This result is consistent with the behaviour and intrinsic thermostability of these variants as recombinant proteins. Indeed, within the PRYSPRY variants there is a correlation (R² = 0.72) between expression levels and thermostability (Figure 5—figure supplement 1E).

During its effector response, TRIM21 neutralizes incoming virions with an efficiency that is proportional to antibody concentration. Previously, we have shown that neutralization is highly sensitive and that activity can be detected when a virion is coated with as few as two antibodies (McEwan et al., 2012). We carried out neutralization experiments on all 22 variants over a range of antibody concentrations (Figure 6A). Neutralization curves were fit to an exponential decay equation and a neutralization constant (K_neut) determined for each variant. Most variants preserved at least wild-type neutralization activity, and some had slightly enhanced activity. Plotting expression levels of all variants against K_neut suggests that while differences in expression impact activity they do not explain all the variation (R² = 0.36) (Figure 6B). The relationship between stability and activity is most clearly seen with PRYSPRY variants G440R and F446I, which have the lowest expression and also the least activity. Given that F446I maintains wild-type levels of binding to IgG, it is likely that the activity defect of this variant is entirely due to its lack of stability. Consistent with this, cells expressing the F446I variant show some neutralization activity when expression is increased by IFN-α pre-treatment (Figure 6—figure supplement 1). Indeed, intrinsic protein stability (Tm) is a reasonable predictor of neutralization activity, with an R² of 0.63 for PRYSPRY variants (Figure 6B). An outlier to this plot is variant A390V, which is significantly more active than its stability would predict. However, A390V is also the variant with the 2^nd highest affinity to IgG, with a > 5 fold improvement over wild-type. It is therefore possible that the reduced stability of A390V is compensated to some degree by its improved IgG binding to maintain activity. Removing A390V improves the correlation between intrinsic stability and activity to give an R² of 0.91.

Figure 6 with 1 supplement see all

Download asset Open asset

Figure 6—source data 1

https://doi.org/10.7554/eLife.48339.017

Download elife-48339-fig6-data1-v1.xlsx

The reduced thermostability of the PRYSPRY variants A390V, G440R and F446I might mean that these variants can function better at the lower temperatures maintained in the respiratory epithelium, which is cooled by inspired air to a few degrees below the core temperature and is a primary site for adenovirus infection (Keck et al., 2000). Conversely, these variants may lose their activity at higher temperatures such as during a fever when the core temperature is increased. Therefore, we tested the ability of these three PRYSPRY variants to mediate virus neutralization at four different temperatures (33°C, 35°C, 37°C and 39.5°C). The result showed that all three PRYSPRY variants functioned better at lower temperatures, particularly at 33°C, where even the F446I variant, which was non-functional at 37°C, showed some neutralization activity (Figure 7A). Variants A390V and G440R both showed slightly more neutralization activity than WT at 33°C (Figure 7A), possibly due to their higher binding affinity for IgG Fc as measured by ITC (Figure 3). Neutralization activity, including that of the WT, declined as the incubation temperature was raised. However, this was much more pronounced in the less thermostable variants. Variant F446I was largely inactive even at 35°C, while A390V and G440R moved from being more active than WT at 33°C to substantially less active at 39.5°C (Figure 7A). These changes in neutralization activity were not accompanied by significant differences in protein expression level at the time of infection. Immunoblot analysis revealed no obvious changes in overall TRIM21 protein expression in cells after 24 hr incubation at the different temperatures (Figure 7B). This suggest that the differences in neutralization activity were due to stabilization of protein fold at lower temperatures and protein denaturation at higher temperatures. Taken together, the data demonstrate that reduced in vitro protein thermostability has a significant impact on cell function at different physiologically relevant temperatures.

Figure 7

Download asset Open asset

Figure 7—source data 1

https://doi.org/10.7554/eLife.48339.019

Download elife-48339-fig7-data1-v1.xlsx

In addition to mediating neutralization, TRIM21 also triggers immune signaling by activating transcription pathways such as NF-κB. We therefore tested whether any of the variants impact on signaling activity during viral infection. Consistent with the neutralization data, we observed that most variants remain active, preserving the ability to detect antibody-coated adenovirus and trigger NF-κB (Figure 8A). Also consistent with neutralization, many variants had slightly increased activity over wild-type. The overall correlation with absolute expression level was similarly low, with an R² of 0.25 (Figure 8B) but correlation within the PRYSPRY was good (R² = 0.97), suggesting that stability and expression of variants can impact signaling capability. Even A390V showed reduced activity, consistent with previous findings that signaling is more sensitive to perturbation than neutralization (Foss et al., 2016).

Figure 8

Download asset Open asset

Figure 8—source data 1

https://doi.org/10.7554/eLife.48339.021

Download elife-48339-fig8-data1-v1.xlsx

While our native expression system recreates endogenous regulation and protein levels, we sought independent confirmation of the results obtained using this approach. An advantage of the 1000 Genomes Project over other, larger, sequence collections is that immortalized lymphoblastoid cell lines (LCLs) were derived from every volunteer. We therefore obtained lines from each of the 22 volunteers with an annotated missense variant. We first sequenced the cell lines and confirmed that each variant is indeed present and heterozygous (Figure 9—figure supplement 1). Next, we chose several variants from our ectopic expression study – including the two B-Box variants R118Q and R121C, premature stop R234* and unstable PRYSPRY variant F446I and investigated their neutralization activity. This allowed assessment of each variant in their naturally occurring genomic background. Despite lower TRIM21 expression than in a comparison wild-type LCL (Figure 9A), each variant LCL displayed remarkably similar neutralization activity (Figure 9E). As the ectopic expression data shows that both R234* and F446I variants are non-functional for neutralization as homozygotes (Figure 6A), this suggests that the remaining wild-type copy of TRIM21 is sufficient to maintain full activity. The data also indicates that non-functional TRIM21 variants do not exert a dominant negative effect, despite the protein being dimeric. To further test this, we established stable lines expressing variants R234* and F446I in wild-type 293T cells that retain endogenous TRIM21 expression (Figure 9B). We observed no reduction in either neutralization (Figure 9F) or signaling (Figure 9I) capability in these cells, consistent with the LCL data. Together this suggests that deleterious TRIM21 mutants will be autosomal recessive.

Figure 9 with 1 supplement see all

Download asset Open asset

Figure 9—source data 1

https://doi.org/10.7554/eLife.48339.024

Download elife-48339-fig9-data1-v1.xlsx

Volunteer LCLs allow functional testing of variants in situ but the available haplotypes in our study are all heterozygous. We therefore chose two variants – B-Box mutant R118Q and premature stop R234* – and carried out CRISPR gene editing to further confirm the validity of our ectopic expression approach. We isolated both heterozygous and homozygous gene edits of each variant (Figure 9C–D) and tested their respective neutralization (Figure 9G) and signaling activities (Figure 9J). For the B-Box variant R118Q, we found that heterozygous and homozygous clones were equally functional in both neutralization and signaling assays. This is consistent with results from both ectopic homozygous and volunteer LCL heterozygous experiments. For the truncated variant R234*, the heterozygous CRISPR clone maintained wild-type neutralization activity in agreement with the volunteer LCL data. Heterozygous R234* also maintained signaling activity, albeit at approximately half that of wild-type, which may be a result of reduced overall TRIM21 expression. Homozygous R234* clones were completely non-functional in both neutralization and signaling, in agreement with ectopic expression data.

While individually rare, missense variants are the most common form of coding diversity in the human genome. Predicting the effects of rare variants is therefore of considerable importance to human health. However, specific rare variants are difficult, if not impossible, to associate with disease in GWAS-type studies precisely because of their rarity and thus there are very few studies of the properties of disease-causing rare mutants in human proteins (Yue et al., 2005). We have used TRIM21 as a case study both to assess the impact of naturally occurring rare variants and test the accuracy of de novo prediction. Previous reports have noted a tendency for over-prediction, with anticipated defective variants showing no reduction in wild-type activity. An in vivo study of missense mutations in immune genes concluded that predicted false-positive deleterious variants might be due to nearly neutral mutations that do not impact function sufficiently to give a measurable phenotype (Miosge et al., 2015). In our study, we were able to quantify differences in all tested variants suggesting that our assays are sufficiently sensitive.

We observed that while there was some evidence of false-positives, (e.g. with SIFT and PolyPhen-2), there was remarkably accurate assessment of variant phenotypes with CADD, REVEL and MetaLR – all of which correctly predicted only F446I as a loss-of-function variant.

We observed that overprediction is partly alleviated with increasing size of the underlying sequence database. With fewer genomes, some positions might appear conserved not because mutations at these positions are deleterious but because available datasets are too small for all possible allowed substitutions to be sampled. The improvement with increasing number of genomes suggests taxon sampling is an important factor in improving the accuracy of predictions and suggests that it will further improve in the future as more whole genome sequences become available.

Li et al., recently performed a benchmark of predictors using recently characterized variants of clinical interest and all possible single substitutions in PPARG (Li et al., 2018). Our findings are consistent with their conclusions, in that REVEL, CADD and, MetaLR give the most accurate predictions while SIFT, PolyPhen-2 and MutationAssessor perform less well. Li et al., treated each mutation as binary (benign or pathogenic) but the scores reported by predictors are continuous. Detailed phenotyping based on known functions of TRIM21 that we performed here have allowed us to correlate predictions with specific types of disruptions to protein structure and function, and we noted high correlation between pathogenicity predictions and impact on protein stability. Remarkably, the best performing predictors not only accurately identify the most destabilizing mutations but also are able to correctly rank all variants. If this trend generalizes to other globular proteins, pathogenicity predictors could potentially be used to model the effect of mutations on stability, especially considering that computational methods explicitly designed for the latter perform poorly (Thiltgen and Goldstein, 2012).

The fact that many variants do not impact TRIM21 activity is consistent with the concept of protein robustness and the buffering of function in the presence of destabilizing mutation. Under the threshold model of robustness, destabilizing mutations retain wild-type activity while stability remains above a certain level (Tokuriki and Tawfik, 2009). Importantly, we observed more accurate prediction of deleterious variants in the PRYSPRY, a stable globular domain, than in the RING or B-Box – both of which are Zn-finger motifs. PRYSPRY variants A390V, G440R and F446I were predicted to be impactful and had measurably reduced antiviral activity. Meanwhile, RING variant G47R and B-Box variant R121C, although predicted by some algorithms to be deleterious, possessed wild-type function. The reduced prediction accuracy for the RING and B-Box is consistent with the notion of gradient robustness, whereby dynamic or partly disordered proteins with fewer stabilizing interactions have less to lose, in contrast to highly ordered and well-packed structures like the PRYSPRY that are poised to undergo a steep functional decline once their stability threshold is compromised. Domain architectures such as the Zn-finger, where stability is provided by an inorganic cofactor, may represent a third category that has both threshold and gradient behaviour. Mutations in chelating residues would be predicted to be individually catastrophic (threshold robustness) but elsewhere display little effect (gradient robustness). Categorizing protein folds and processing them independently may improve predictions of variant stability. If pathogenicity predictors are more effective at capturing the impact on protein stability rather than function, this may explain the poor correlation observed between observed and predicted affects in regions other than globular domains.

The accuracy of deleterious prediction is complicated by the fact that a damaging mutation cannot be equated with a disease phenotype (Eilbeck et al., 2017). A limitation of this study is that we have only analyzed individual SNPs and not addressed the potential for epistatic effects and co-variance of multiple SNPs either within TRIM21 or in co-factor alleles. In part this is because there are no identified naturally occurring TRIM21 haplotypes with multiple SNPs. We also did not find evidence of epistasis between TRIM21 and its cognate E2 enzymes Ube2W or Ube2N, as these were wild-type in TRIM21 variant haplotypes from the 1000 Genomes dataset. Some haplotypes do carry variant IGHG1 alleles, the predominant IgG subclass, (including K97R, D239E, L241M, A314G, D390N, G396). However, these IGHG1 variants are relatively conservative substitutions, classified as benign/tolerated by PolyPhen and SIFT, and include common allotypes. The A314G variant is potentially of interest as it is located relatively close to the ‘HNHY’ motif that is bound by the TRIM21 PRYSPRY domain. However, individuals possessing both A314G and a TRIM21 variant have a different RING sequence rather than PRYSPRY, arguing against any specific epistasis.

When predicting the relative function of variants, it is important to note that even individual SNPs can be challenging as they may have opposing effects that mitigate against a deleterious outcome. For instance, although PRYSPRY variant A390V was significantly less stable than wild-type and had lower levels of cellular expression, it bound antibody fivefold more tightly and was able to maintain wild-type neutralization activity. This is reminiscent of the ‘stability-activity trade-off’ noted for enzyme mutations, whereby variants that increase catalytic activity do so at a cost in global stability (Tokuriki et al., 2008). Such trade-offs complicate de novo prediction. Nevertheless, our results support the hypothesis that loss of protein stability is both a major cause of disease-causing mutation and a useful correlate (Yue et al., 2005). The intrinsic instability of TRIM21 variants was predicted with some accuracy and gave a good indicator of activity. Thermostability was also a good predictor of temperature dependent function. Of those variants with reduced thermostability, F446I, which was non-functional at 37°C, possessed some neutralization activity at 33°C, whilst A390V and G440R lost more activity than WT at 39.5°C despite their higher affinity for antibody. These results highlight the importance of considering physiological temperature when assessing the impact of altered thermostability on variant function. For instance, as is the case here, function in the respiratory tract can take place at temperatures a few degrees cooler than core body but also at temperatures above 37°C during fever.

TRIM21 antibody receptor function is maintained in most rare missense variants, suggesting that this activity is under purifying selection and therefore important in humans. In support of this, comparing the predictions of more frequent variants (frequency ≥1×10⁻⁵) with rare variants (frequency <1×10⁻⁵) shows that more abundant variants are less likely to be deleterious (Figure 10). Comparison with predictions for random variants also shows that both sets of naturally occurring variants are less deleterious than those arising by chance (Figure 10). The absence of haploinsufficiency, toxic gain-of-function or dominant negative effects in tested natural variants means that any variant that is deleterious will be autosomal recessive. In gnomAD, a large dataset of variation in over 140,000 individuals, there is no evidence for homozygous loss-of-function variants in TRIM21. As the highest frequency of a high-confidence loss-of-function variant is 3.19 × 10⁻⁵, likely fewer than 1 in 982 million individuals are born with no TRIM21 antibody receptor activity. This result suggests it will not be possible to infer the importance of TRIM21 in human biology based on statistical association with common diseases (e.g. as in GWAS).

Figure 10

Download asset Open asset

Figure 10—source data 1

https://doi.org/10.7554/eLife.48339.026

Download elife-48339-fig10-data1-v1.xlsx

All data generated or analysed during this study are included in the manuscript and supporting files. Previously published data from the 100 Genomes Project (2015; http://www.internationalgenome.org/data#download) and the Genome Aggregation Datatbase (2016; https://gnomad.broadinstitute.org/downloads) was used as part of this work.

1. 1. Abecasis GR
   2. Altshuler D
   3. Auton A
   4. Brooks LD
   5. Durbin RM
   6. Gibbs RA
   7. Hurles ME
   8. McVean GA
   9. 1000 Genomes Project Consortium

   (2010) A map of human genome variation from population-scale sequencing

   Nature 467:1061–1073.

   https://doi.org/10.1038/nature09534

   • PubMed
   • Google Scholar
2. 1. Abecasis GR
   2. Auton A
   3. Brooks LD
   4. DePristo MA
   5. Durbin RM
   6. Handsaker RE
   7. Kang HM
   8. Marth GT
   9. McVean GA
   10. 1000 Genomes Project Consortium

   (2012) An integrated map of genetic variation from 1,092 human genomes

   Nature 491:56–65.

   https://doi.org/10.1038/nature11632

   • PubMed
   • Google Scholar
3. 1. Adzhubei IA
   2. Schmidt S
   3. Peshkin L
   4. Ramensky VE
   5. Gerasimova A
   6. Bork P
   7. Kondrashov AS
   8. Sunyaev SR

   (2010) A method and server for predicting damaging missense mutations

   Nature Methods 7:248–249.

   https://doi.org/10.1038/nmeth0410-248

   • PubMed
   • Google Scholar
4. 1. Andrews TD
   2. Sjollema G
   3. Goodnow CC

   (2013) Understanding the immunological impact of the human mutation explosion

   Trends in Immunology 34:99–106.

   https://doi.org/10.1016/j.it.2012.12.001

   • PubMed
   • Google Scholar
5. 1. Astle WJ
   2. Elding H
   3. Jiang T
   4. Allen D
   5. Ruklisa D
   6. Mann AL
   7. Mead D
   8. Bouman H
   9. Riveros-Mckay F
   10. Kostadima MA
   11. Lambourne JJ
   12. Sivapalaratnam S
   13. Downes K
   14. Kundu K
   15. Bomba L
   16. Berentsen K
   17. Bradley JR
   18. Daugherty LC
   19. Delaneau O
   20. Freson K
   21. Garner SF
   22. Grassi L
   23. Guerrero J
   24. Haimel M
   25. Janssen-Megens EM
   26. Kaan A
   27. Kamat M
   28. Kim B
   29. Mandoli A
   30. Marchini J
   31. Martens JHA
   32. Meacham S
   33. Megy K
   34. O'Connell J
   35. Petersen R
   36. Sharifi N
   37. Sheard SM
   38. Staley JR
   39. Tuna S
   40. van der Ent M
   41. Walter K
   42. Wang SY
   43. Wheeler E
   44. Wilder SP
   45. Iotchkova V
   46. Moore C
   47. Sambrook J
   48. Stunnenberg HG
   49. Di Angelantonio E
   50. Kaptoge S
   51. Kuijpers TW
   52. Carrillo-de-Santa-Pau E
   53. Juan D
   54. Rico D
   55. Valencia A
   56. Chen L
   57. Ge B
   58. Vasquez L
   59. Kwan T
   60. Garrido-Martín D
   61. Watt S
   62. Yang Y
   63. Guigo R
   64. Beck S
   65. Paul DS
   66. Pastinen T
   67. Bujold D
   68. Bourque G
   69. Frontini M
   70. Danesh J
   71. Roberts DJ
   72. Ouwehand WH
   73. Butterworth AS
   74. Soranzo N

   (2016) The allelic landscape of human blood cell trait variation and links to common complex disease

   Cell 167:1415–1429.

   https://doi.org/10.1016/j.cell.2016.10.042

   • PubMed
   • Google Scholar
6. 1. Auer PL
   2. Lettre G

   (2015) Rare variant association studies: considerations, challenges and opportunities

   Genome Medicine 7:16.

   https://doi.org/10.1186/s13073-015-0138-2

   • PubMed
   • Google Scholar
7. 1. Bamshad MJ
   2. Ng SB
   3. Bigham AW
   4. Tabor HK
   5. Emond MJ
   6. Nickerson DA
   7. Shendure J

   (2011) Exome sequencing as a tool for mendelian disease gene discovery

   Nature Reviews Genetics 12:745–755.

   https://doi.org/10.1038/nrg3031

   • PubMed
   • Google Scholar
8. 1. Bomba L
   2. Walter K
   3. Soranzo N

   (2017) The impact of rare and low-frequency genetic variants in common disease

   Genome Biology 18:77.

   https://doi.org/10.1186/s13059-017-1212-4

   • PubMed
   • Google Scholar
9. 1. Bottermann M
   2. Lode HE
   3. Watkinson RE
   4. Foss S
   5. Sandlie I
   6. Andersen JT
   7. James LC

   (2016) Antibody-antigen kinetics constrain intracellular humoral immunity

   Scientific Reports 6:37457.

   https://doi.org/10.1038/srep37457

   • PubMed
   • Google Scholar
10. 1. Bottermann M
    2. Foss S
    3. Caddy SL
    4. Clift D
    5. van Tienen LM
    6. Vaysburd M
    7. Cruickshank J
    8. O'Connell K
    9. Clark J
    10. Mayes K
    11. Higginson K
    12. Lode HE
    13. McAdam MB
    14. Sandlie I
    15. Andersen JT
    16. James LC

    (2019) Complement C4 prevents viral infection through capsid inactivation

    Cell Host & Microbe 25:617–629.

    https://doi.org/10.1016/j.chom.2019.02.016

    • PubMed
    • Google Scholar
11. 1. Cunningham F
    2. Achuthan P
    3. Akanni W
    4. Allen J
    5. Amode MR
    6. Armean IM
    7. Bennett R
    8. Bhai J
    9. Billis K
    10. Boddu S
    11. Cummins C
    12. Davidson C
    13. Dodiya KJ
    14. Gall A
    15. Girón CG
    16. Gil L
    17. Grego T
    18. Haggerty L
    19. Haskell E
    20. Hourlier T
    21. Izuogu OG
    22. Janacek SH
    23. Juettemann T
    24. Kay M
    25. Laird MR
    26. Lavidas I
    27. Liu Z
    28. Loveland JE
    29. Marugán JC
    30. Maurel T
    31. McMahon AC
    32. Moore B
    33. Morales J
    34. Mudge JM
    35. Nuhn M
    36. Ogeh D
    37. Parker A
    38. Parton A
    39. Patricio M
    40. Abdul Salam AI
    41. Schmitt BM
    42. Schuilenburg H
    43. Sheppard D
    44. Sparrow H
    45. Stapleton E
    46. Szuba M
    47. Taylor K
    48. Threadgold G
    49. Thormann A
    50. Vullo A
    51. Walts B
    52. Winterbottom A
    53. Zadissa A
    54. Chakiachvili M
    55. Frankish A
    56. Hunt SE
    57. Kostadima M
    58. Langridge N
    59. Martin FJ
    60. Muffato M
    61. Perry E
    62. Ruffier M
    63. Staines DM
    64. Trevanion SJ
    65. Aken BL
    66. Yates AD
    67. Zerbino DR
    68. Flicek P

    (2019) Ensembl 2019

    Nucleic Acids Research 47:D745–D751.

    https://doi.org/10.1093/nar/gky1113

    • PubMed
    • Google Scholar
12. 1. Demaison C
    2. Parsley K
    3. Brouns G
    4. Scherr M
    5. Battmer K
    6. Kinnon C
    7. Grez M
    8. Thrasher AJ

    (2002) High-level transduction and gene expression in hematopoietic repopulating cells using a human immunodeficiency [correction of imunodeficiency] virus type 1-based lentiviral vector containing an internal spleen focus forming virus promoter

    Human Gene Therapy 13:803–813.

    https://doi.org/10.1089/10430340252898984

    • PubMed
    • Google Scholar
13. 1. Dickson C
    2. Fletcher AJ
    3. Vaysburd M
    4. Yang JC
    5. Mallery DL
    6. Zeng J
    7. Johnson CM
    8. McLaughlin SH
    9. Skehel M
    10. Maslen S
    11. Cruickshank J
    12. Huguenin-Dezot N
    13. Chin JW
    14. Neuhaus D
    15. James LC

    (2018) Intracellular antibody signalling is regulated by phosphorylation of the fc receptor TRIM21

    eLife 7:e32660.

    https://doi.org/10.7554/eLife.32660

    • PubMed
    • Google Scholar
14. 1. Dong C
    2. Wei P
    3. Jian X
    4. Gibbs R
    5. Boerwinkle E
    6. Wang K
    7. Liu X

    (2015) Comparison and integration of deleteriousness prediction methods for nonsynonymous SNVs in whole exome sequencing studies

    Human Molecular Genetics 24:2125–2137.

    https://doi.org/10.1093/hmg/ddu733

    • PubMed
    • Google Scholar
15. 1. Dorfman R
    2. Nalpathamkalam T
    3. Taylor C
    4. Gonska T
    5. Keenan K
    6. Yuan XW
    7. Corey M
    8. Tsui LC
    9. Zielenski J
    10. Durie P

    (2010) Do common in silico tools predict the clinical consequences of amino-acid substitutions in the CFTR gene?

    Clinical Genetics 77:464–473.

    https://doi.org/10.1111/j.1399-0004.2009.01351.x

    • PubMed
    • Google Scholar
16. 1. Eilbeck K
    2. Quinlan A
    3. Yandell M

    (2017) Settling the score: variant prioritization and mendelian disease

    Nature Reviews Genetics 18:599–612.

    https://doi.org/10.1038/nrg.2017.52

    • PubMed
    • Google Scholar
17. 1. Ernst C
    2. Hahnen E
    3. Engel C
    4. Nothnagel M
    5. Weber J
    6. Schmutzler RK
    7. Hauke J

    (2018) Performance of in silico prediction tools for the classification of rare BRCA1/2 missense variants in clinical diagnostics

    BMC Medical Genomics 11:35.

    https://doi.org/10.1186/s12920-018-0353-y

    • PubMed
    • Google Scholar
18. 1. Eyre-Walker A
    2. Keightley PD

    (1999) High genomic deleterious mutation rates in hominids

    Nature 397:344–347.

    https://doi.org/10.1038/16915

    • PubMed
    • Google Scholar
19. 1. Fletcher AJ
    2. Vaysburd M
    3. Maslen S
    4. Zeng J
    5. Skehel JM
    6. Towers GJ
    7. James LC

    (2018) Trivalent RING assembly on retroviral capsids activates TRIM5 Ubiquitination and Innate Immune Signaling

    Cell Host & Microbe 24:761–775.

    https://doi.org/10.1016/j.chom.2018.10.007

    • PubMed
    • Google Scholar
20. 1. Foss S
    2. Watkinson RE
    3. Grevys A
    4. McAdam MB
    5. Bern M
    6. Høydahl LS
    7. Dalhus B
    8. Michaelsen TE
    9. Sandlie I
    10. James LC
    11. Andersen JT

    (2016) TRIM21 immune signaling is more sensitive to antibody affinity than its neutralization activity

    The Journal of Immunology 196:3452–3459.

    https://doi.org/10.4049/jimmunol.1502601

    • PubMed
    • Google Scholar
21. Book

    1. Gasteiger E HC
    2. Gattiker A
    3. Duvaud S
    4. Wilkins MR
    5. Appel RD
    6. Bairoch A

    (2005)

    The Proteomics Protocols Handbook

    Walker JM, editors. Springer.

    • Google Scholar
22. 1. Ioannidis NM
    2. Rothstein JH
    3. Pejaver V
    4. Middha S
    5. McDonnell SK
    6. Baheti S
    7. Musolf A
    8. Li Q
    9. Holzinger E
    10. Karyadi D
    11. Cannon-Albright LA
    12. Teerlink CC
    13. Stanford JL
    14. Isaacs WB
    15. Xu J
    16. Cooney KA
    17. Lange EM
    18. Schleutker J
    19. Carpten JD
    20. Powell IJ
    21. Cussenot O
    22. Cancel-Tassin G
    23. Giles GG
    24. MacInnis RJ
    25. Maier C
    26. Hsieh CL
    27. Wiklund F
    28. Catalona WJ
    29. Foulkes WD
    30. Mandal D
    31. Eeles RA
    32. Kote-Jarai Z
    33. Bustamante CD
    34. Schaid DJ
    35. Hastie T
    36. Ostrander EA
    37. Bailey-Wilson JE
    38. Radivojac P
    39. Thibodeau SN
    40. Whittemore AS
    41. Sieh W

    (2016) REVEL: an ensemble method for predicting the pathogenicity of rare missense variants

    The American Journal of Human Genetics 99:877–885.

    https://doi.org/10.1016/j.ajhg.2016.08.016

    • PubMed
    • Google Scholar
23. 1. James LC
    2. Keeble AH
    3. Khan Z
    4. Rhodes DA
    5. Trowsdale J

    (2007) Structural basis for PRYSPRY-mediated tripartite motif (TRIM) protein function

    PNAS 104:6200–6205.

    https://doi.org/10.1073/pnas.0609174104

    • PubMed
    • Google Scholar
24. 1. Jinek M
    2. Chylinski K
    3. Fonfara I
    4. Hauer M
    5. Doudna JA
    6. Charpentier E

    (2012) A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity

    Science 337:816–821.

    https://doi.org/10.1126/science.1225829

    • PubMed
    • Google Scholar
25. 1. Keck T
    2. Leiacker R
    3. Heinrich A
    4. Kühnemann S
    5. Rettinger G

    (2000)

    Humidity and temperature profile in the nasal cavity

    Rhinology 38:167–171.

    • PubMed
    • Google Scholar
26. 1. Keeble AH
    2. Khan Z
    3. Forster A
    4. James LC

    (2008) TRIM21 is an IgG receptor that is structurally, Thermodynamically, and kinetically conserved

    Proceedings of the National Academy of Sciences 105:6045–6050.

    https://doi.org/10.1073/pnas.0800159105

    • PubMed
    • Google Scholar
27. 1. Keinan A
    2. Clark AG

    (2012) Recent explosive human population growth has resulted in an excess of rare genetic variants

    Science 336:740–743.

    https://doi.org/10.1126/science.1217283

    • PubMed
    • Google Scholar
28. 1. Kryukov GV
    2. Pennacchio LA
    3. Sunyaev SR

    (2007) Most rare missense alleles are deleterious in humans: implications for complex disease and association studies

    The American Journal of Human Genetics 80:727–739.

    https://doi.org/10.1086/513473

    • PubMed
    • Google Scholar
29. 1. Lek M
    2. Karczewski KJ
    3. Minikel EV
    4. Samocha KE
    5. Banks E
    6. Fennell T
    7. O'Donnell-Luria AH
    8. Ware JS
    9. Hill AJ
    10. Cummings BB
    11. Tukiainen T
    12. Birnbaum DP
    13. Kosmicki JA
    14. Duncan LE
    15. Estrada K
    16. Zhao F
    17. Zou J
    18. Pierce-Hoffman E
    19. Berghout J
    20. Cooper DN
    21. Deflaux N
    22. DePristo M
    23. Do R
    24. Flannick J
    25. Fromer M
    26. Gauthier L
    27. Goldstein J
    28. Gupta N
    29. Howrigan D
    30. Kiezun A
    31. Kurki MI
    32. Moonshine AL
    33. Natarajan P
    34. Orozco L
    35. Peloso GM
    36. Poplin R
    37. Rivas MA
    38. Ruano-Rubio V
    39. Rose SA
    40. Ruderfer DM
    41. Shakir K
    42. Stenson PD
    43. Stevens C
    44. Thomas BP
    45. Tiao G
    46. Tusie-Luna MT
    47. Weisburd B
    48. Won HH
    49. Yu D
    50. Altshuler DM
    51. Ardissino D
    52. Boehnke M
    53. Danesh J
    54. Donnelly S
    55. Elosua R
    56. Florez JC
    57. Gabriel SB
    58. Getz G
    59. Glatt SJ
    60. Hultman CM
    61. Kathiresan S
    62. Laakso M
    63. McCarroll S
    64. McCarthy MI
    65. McGovern D
    66. McPherson R
    67. Neale BM
    68. Palotie A
    69. Purcell SM
    70. Saleheen D
    71. Scharf JM
    72. Sklar P
    73. Sullivan PF
    74. Tuomilehto J
    75. Tsuang MT
    76. Watkins HC
    77. Wilson JG
    78. Daly MJ
    79. MacArthur DG
    80. Exome Aggregation Consortium

    (2016) Analysis of protein-coding genetic variation in 60,706 humans

    Nature 536:285–291.

    https://doi.org/10.1038/nature19057

    • PubMed
    • Google Scholar
30. 1. Li J
    2. Zhao T
    3. Zhang Y
    4. Zhang K
    5. Shi L
    6. Chen Y
    7. Wang X
    8. Sun Z

    (2018) Performance evaluation of pathogenicity-computation methods for missense variants

    Nucleic Acids Research 46:7793–7804.

    https://doi.org/10.1093/nar/gky678

    • PubMed
    • Google Scholar
31. 1. Liu X
    2. Jian X
    3. Boerwinkle E

    (2013) dbNSFP v2.0: a database of human non-synonymous SNVs and their functional predictions and annotations

    Human Mutation 34:E2393–E2402.

    https://doi.org/10.1002/humu.22376

    • PubMed
    • Google Scholar
32. 1. Liu H
    2. Naismith JH

    (2008) An efficient one-step site-directed deletion, insertion, single and multiple-site plasmid mutagenesis protocol

    BMC Biotechnology 8:91.

    https://doi.org/10.1186/1472-6750-8-91

    • PubMed
    • Google Scholar
33. 1. Maher B

    (2008) Personal genomes: the case of the missing heritability

    Nature 456:18–21.

    https://doi.org/10.1038/456018a

    • PubMed
    • Google Scholar
34. 1. Manukyan G
    2. Aminov R

    (2016) Update on pyrin functions and mechanisms of familial mediterranean fever

    Frontiers in Microbiology 7:456.

    https://doi.org/10.3389/fmicb.2016.00456

    • PubMed
    • Google Scholar
35. 1. McClellan J
    2. King MC

    (2010) Genetic heterogeneity in human disease

    Cell 141:210–217.

    https://doi.org/10.1016/j.cell.2010.03.032

    • PubMed
    • Google Scholar
36. 1. McEwan WA
    2. Hauler F
    3. Williams CR
    4. Bidgood SR
    5. Mallery DL
    6. Crowther RA
    7. James LC

    (2012) Regulation of virus neutralization and the persistent fraction by TRIM21

    Journal of Virology 86:8482–8491.

    https://doi.org/10.1128/JVI.00728-12

    • PubMed
    • Google Scholar
37. 1. Miosge LA
    2. Field MA
    3. Sontani Y
    4. Cho V
    5. Johnson S
    6. Palkova A
    7. Balakishnan B
    8. Liang R
    9. Zhang Y
    10. Lyon S
    11. Beutler B
    12. Whittle B
    13. Bertram EM
    14. Enders A
    15. Goodnow CC
    16. Andrews TD

    (2015) Comparison of predicted and actual consequences of missense mutations

    PNAS 112:E5189–E5198.

    https://doi.org/10.1073/pnas.1511585112

    • PubMed
    • Google Scholar
38. 1. Mozzi A
    2. Pontremoli C
    3. Forni D
    4. Clerici M
    5. Pozzoli U
    6. Bresolin N
    7. Cagliani R
    8. Sironi M

    (2015) OASes and STING: adaptive evolution in concert

    Genome Biology and Evolution 7:1016–1032.

    https://doi.org/10.1093/gbe/evv046

    • PubMed
    • Google Scholar
39. 1. Park JH
    2. Gail MH
    3. Weinberg CR
    4. Carroll RJ
    5. Chung CC
    6. Wang Z
    7. Chanock SJ
    8. Fraumeni JF
    9. Chatterjee N

    (2011) Distribution of allele frequencies and effect sizes and their interrelationships for common genetic susceptibility variants

    PNAS 108:18026–18031.

    https://doi.org/10.1073/pnas.1114759108

    • PubMed
    • Google Scholar
40. 1. Pritchard JK

    (2001) Are rare variants responsible for susceptibility to complex diseases?

    The American Journal of Human Genetics 69:124–137.

    https://doi.org/10.1086/321272

    • PubMed
    • Google Scholar
41. 1. Rentzsch P
    2. Witten D
    3. Cooper GM
    4. Shendure J
    5. Kircher M

    (2019) CADD: predicting the deleteriousness of variants throughout the human genome

    Nucleic Acids Research 47:D886–D894.

    https://doi.org/10.1093/nar/gky1016

    • PubMed
    • Google Scholar
42. 1. Reva B
    2. Antipin Y
    3. Sander C

    (2011) Predicting the functional impact of protein mutations: application to Cancer genomics

    Nucleic Acids Research 39:e118.

    https://doi.org/10.1093/nar/gkr407

    • PubMed
    • Google Scholar
43. 1. Reymond A
    2. Meroni G
    3. Fantozzi A
    4. Merla G
    5. Cairo S
    6. Luzi L
    7. Riganelli D
    8. Zanaria E
    9. Messali S
    10. Cainarca S
    11. Guffanti A
    12. Minucci S
    13. Pelicci PG
    14. Ballabio A

    (2001) The tripartite motif family identifies cell compartments

    The EMBO Journal 20:2140–2151.

    https://doi.org/10.1093/emboj/20.9.2140

    • PubMed
    • Google Scholar
44. 1. Rhodes DA
    2. Ihrke G
    3. Reinicke AT
    4. Malcherek G
    5. Towey M
    6. Isenberg DA
    7. Trowsdale J

    (2002) The 52 000 MW ro/SS-A autoantigen in sjögren's syndrome/systemic lupus erythematosus (Ro52) is an interferon-gamma inducible tripartite motif protein associated with membrane proximal structures

    Immunology 106:246–256.

    https://doi.org/10.1046/j.1365-2567.2002.01417.x

    • PubMed
    • Google Scholar
45. 1. Richardson CD
    2. Ray GJ
    3. DeWitt MA
    4. Curie GL
    5. Corn JE

    (2016) Enhancing homology-directed genome editing by catalytically active and inactive CRISPR-Cas9 using asymmetric donor DNA

    Nature Biotechnology 34:339–344.

    https://doi.org/10.1038/nbt.3481

    • PubMed
    • Google Scholar
46. 1. Sawyer SL
    2. Wu LI
    3. Emerman M
    4. Malik HS

    (2005) Positive selection of primate TRIM5alpha identifies a critical species-specific retroviral restriction domain

    PNAS 102:2832–2837.

    https://doi.org/10.1073/pnas.0409853102

    • PubMed
    • Google Scholar
47. 1. Schindelin J
    2. Arganda-Carreras I
    3. Frise E
    4. Kaynig V
    5. Longair M
    6. Pietzsch T
    7. Preibisch S
    8. Rueden C
    9. Saalfeld S
    10. Schmid B
    11. Tinevez JY
    12. White DJ
    13. Hartenstein V
    14. Eliceiri K
    15. Tomancak P
    16. Cardona A

    (2012) Fiji: an open-source platform for biological-image analysis

    Nature Methods 9:676–682.

    https://doi.org/10.1038/nmeth.2019

    • PubMed
    • Google Scholar
48. 1. Sim NL
    2. Kumar P
    3. Hu J
    4. Henikoff S
    5. Schneider G
    6. Ng PC

    (2012) SIFT web server: predicting effects of amino acid substitutions on proteins

    Nucleic Acids Research 40:W452–W457.

    https://doi.org/10.1093/nar/gks539

    • PubMed
    • Google Scholar
49. 1. Sudmant PH
    2. Rausch T
    3. Gardner EJ
    4. Handsaker RE
    5. Abyzov A
    6. Huddleston J
    7. Zhang Y
    8. Ye K
    9. Jun G
    10. Fritz MH
    11. Konkel MK
    12. Malhotra A
    13. Stütz AM
    14. Shi X
    15. Casale FP
    16. Chen J
    17. Hormozdiari F
    18. Dayama G
    19. Chen K
    20. Malig M
    21. Chaisson MJP
    22. Walter K
    23. Meiers S
    24. Kashin S
    25. Garrison E
    26. Auton A
    27. Lam HYK
    28. Mu XJ
    29. Alkan C
    30. Antaki D
    31. Bae T
    32. Cerveira E
    33. Chines P
    34. Chong Z
    35. Clarke L
    36. Dal E
    37. Ding L
    38. Emery S
    39. Fan X
    40. Gujral M
    41. Kahveci F
    42. Kidd JM
    43. Kong Y
    44. Lameijer EW
    45. McCarthy S
    46. Flicek P
    47. Gibbs RA
    48. Marth G
    49. Mason CE
    50. Menelaou A
    51. Muzny DM
    52. Nelson BJ
    53. Noor A
    54. Parrish NF
    55. Pendleton M
    56. Quitadamo A
    57. Raeder B
    58. Schadt EE
    59. Romanovitch M
    60. Schlattl A
    61. Sebra R
    62. Shabalin AA
    63. Untergasser A
    64. Walker JA
    65. Wang M
    66. Yu F
    67. Zhang C
    68. Zhang J
    69. Zheng-Bradley X
    70. Zhou W
    71. Zichner T
    72. Sebat J
    73. Batzer MA
    74. McCarroll SA
    75. Mills RE
    76. Gerstein MB
    77. Bashir A
    78. Stegle O
    79. Devine SE
    80. Lee C
    81. Eichler EE
    82. Korbel JO
    83. 1000 Genomes Project Consortium

    (2015) An integrated map of structural variation in 2,504 human genomes

    Nature 526:75–81.

    https://doi.org/10.1038/nature15394

    • PubMed
    • Google Scholar
50. 1. Thiltgen G
    2. Goldstein RA

    (2012) Assessing predictors of changes in protein stability upon mutation using self-consistency

    PLOS ONE 7:e46084.

    https://doi.org/10.1371/journal.pone.0046084

    • PubMed
    • Google Scholar
51. 1. Tokuriki N
    2. Stricher F
    3. Serrano L
    4. Tawfik DS

    (2008) How protein stability and new functions trade off

    PLOS Computational Biology 4:e1000002.

    https://doi.org/10.1371/journal.pcbi.1000002

    • PubMed
    • Google Scholar
52. 1. Tokuriki N
    2. Tawfik DS

    (2009) Stability effects of mutations and protein evolvability

    Current Opinion in Structural Biology 19:596–604.

    https://doi.org/10.1016/j.sbi.2009.08.003

    • PubMed
    • Google Scholar
53. 1. Yi G
    2. Brendel VP
    3. Shu C
    4. Li P
    5. Palanathan S
    6. Cheng Kao C

    (2013) Single nucleotide polymorphisms of human STING can affect innate immune response to cyclic dinucleotides

    PLOS ONE 8:e77846.

    https://doi.org/10.1371/journal.pone.0077846

    • PubMed
    • Google Scholar
54. 1. Yue P
    2. Li Z
    3. Moult J

    (2005) Loss of protein structure stability as a major causative factor in monogenic disease

    Journal of Molecular Biology 353:459–473.

    https://doi.org/10.1016/j.jmb.2005.08.020

    • PubMed
    • Google Scholar
55. 1. Zennou V
    2. Perez-Caballero D
    3. Göttlinger H
    4. Bieniasz PD

    (2004) APOBEC3G incorporation into human immunodeficiency virus type 1 particles

    Journal of Virology 78:12058–12061.

    https://doi.org/10.1128/JVI.78.21.12058-12061.2004

    • PubMed
    • Google Scholar
56. 1. Zhang J
    2. Kinch LN
    3. Cong Q
    4. Weile J
    5. Sun S
    6. Cote AG
    7. Roth FP
    8. Grishin NV

    (2017) Assessing predictions of fitness effects of missense mutations in SUMO-conjugating enzyme UBE2I

    Human Mutation 38:1051–1063.

    https://doi.org/10.1002/humu.23293

    • PubMed
    • Google Scholar

Author details

1. Jingwei Zeng

   MRC Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   Conceptualization, Data curation, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Contributed equally with

   Greg Slodkowicz

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-9802-6019

2. Greg Slodkowicz

   MRC Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Contributed equally with

   Jingwei Zeng

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6918-0386

3. Leo C James

   MRC Laboratory of Molecular Biology, Cambridge, United Kingdom

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   For correspondence

   lcj@mrc-lmb.cam.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-2131-0334

• 1,278

  views

• 119

  downloads

• 9

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.