Predicting the impact of Lynch syndrome-causing missense mutations from structural calculations

Abstract

Accurate methods to assess the pathogenicity of mutations are needed to fully leverage the possibilities of genome sequencing in diagnosis. Current data-driven and bioinformatics approaches are, however, limited by the large number of new variations found in each newly sequenced genome, and often do not provide direct mechanistic insight. Here we demonstrate, for the first time, that saturation mutagenesis, biophysical modeling and co-variation analysis, performed in silico, can predict the abundance, metabolic stability, and function of proteins inside living cells. As a model system, we selected the human mismatch repair protein, MSH2, where missense variants are known to cause the hereditary cancer predisposition disease, known as Lynch syndrome. We show that the majority of disease-causing MSH2 mutations give rise to folding defects and proteasome-dependent degradation rather than inherent loss of function, and accordingly our in silico modeling data accurately identifies disease-causing mutations and outperforms the traditionally used genetic disease predictors. Thus, in conclusion, in silico biophysical modeling should be considered for making genotype-phenotype predictions and for diagnosis of Lynch syndrome, and perhaps other hereditary diseases.

Fig 1. In silico saturation mutagenesis and thermodynamic stability of MSH2 mutants.

(A) Example of structure-based MSH2 saturation mutagenesis energy calculations shown as a heat-map that represents the change in thermodynamic stability relative to the wild type protein. Due to space limitations, the results are only shown for the first 95 residues (the full dataset is provided in the supporting information S1 File & S2 File). The wild type MSH2 sequence is given below, while the 20 possible residues at each position are shown on the vertical axis. The color bar indicates the magnitude of the change in stability, so that stabilizing mutations are shown in blue, neutral mutations in turquoise, and highly destabilizing mutations in red (mutations that decrease the stability by more than 7 kcal/mol are represented by the same color). (B) A histogram of the number of mutants (vertical axis) plotted against the predicted thermodynamic stability (ΔΔG) (horizontal axis). The mean and variance of the entire dataset, and known disease-linked variants [7] are included in the insert. The arrows mark the ΔΔG values for the set of disease-linked MSH2 mutants included in this study. (C) The positions of 24 selected MSH2 mutants are marked in colors according to the heat-map in (A) on a trace of the MSH2-MSH6 structure (PDB: 2O8E). MSH2 is shown in light gray, MSH6 in dark grey and DNA in orange.

Fig 2. MSH2 mutation leads to reduced steady-state levels due to proteasomal degradation.

(A) The steady-state level of the indicated wild type (WT) or MSH2 mutants expressed in U2OS cells was determined using SDS-PAGE and Western blotting with antibodies to 6His-tag on MSH2 in cultures that were either untreated or treated with the proteasome inhibitor bortezomib (BZ) for 10 hours. β-actin served as a loading control. See also supporting information S1 Fig. (B) Blotting, using antibodies to MSH2, revealed that transfection of wild type 6His-MSH2 led to an overexpression of approximately 4 fold. (C) Steady-state levels of the selected MSH2 variants plotted vs. the calculated ΔΔG values. The error bars indicate S.E.M. (n = 3). The red line corresponds to the fit with the thermodynamic model. The grey line indicates the 25th and 75th percentiles after a bootstrapping procedure, in which we fit random subsets of the data to the thermodynamic model (percentiles taken after 5000 iterations of fitting). The ΔG(WT) estimate and error are derived from the same bootstrapping procedure.

Fig 3. Degradation of MSH2 variants.

(A) The degradation of wild type (WT) MSH2 or the indicated MSH2 mutants in U2OS cells was followed at 37°C in cultures treated with cycloheximide (CHX) for 0, 8, 16 and 24 hours. β-actin served as a loading control. The half-life (t½) of each variant was determined by densitometry of this and longer exposures and is given below along with the standard deviation. As a guide, slowly degraded variants have been boxed in green, rapidly degraded variants in red, while intermediate variants are boxed in yellow. See also supporting information S2 Fig. (B) The half-life for the MSH2 variants plotted vs. the calculated ΔΔG values in categories of < 3 kcal/mol and > 3kacl/mol. Bootstrapping shows that average half-life values for low-ΔΔG variants are higher than for high-ΔΔG variants with a p-value of 0.01 (**). (C) A plot of the calculated thermodynamic stabilities of the mutants (vertical axis) against the degradation rate. The horizontal line marks the median and the bars indicate the spread of the data points. *** indicates p < 0.001.

Fig 4. Stabilizing rapidly degraded MSH2 variants by mutation or lowered temperature.

(A) The degradation of wild type (WT) MSH2 or the indicated MSH2 mutants in U2OS cells was followed at 37°C in cultures treated with cycloheximide (CHX) for 0, 8, 16 and 24 hours. β-actin served as a loading control. The half-life (t½) of each variant was determined by densitometry of this and longer exposures and is given below along with the standard deviation. Note that MSH2 variants with lower ΔΔG values appear more stable. (B) The degradation of wild type (WT) MSH2 or the indicated MSH2 mutants in U2OS cells was followed at 29°C in cultures treated with cycloheximide (CHX) for 0, 8, 16 and 24 hours. β-actin served as a loading control. Those MSH2 mutants that displayed a temperature dependent degradation are boxed (red, heat sensitive; blue, cold sensitive). The half-life (t½) of each variant was determined by densitometry and is given below along with the standard deviation. See also supporting information S3 Fig. (C) Those MSH2 mutants that displayed a temperature-dependent degradation are marked on a trace of the MSH2-MSH6 structure, red, heat sensitive; blue, cold sensitive. MSH2 is shown in orange, MSH6 in grey and DNA in yellow.

Fig 5. Functional analyses of MSH2 variants.

(A) The MSH2 variants were analyzed for interaction with MSH6. Endogenous MSH6 was immunoprecipitated and the precipitated material analyzed by SDS-PAGE and Western blotting using antibodies to the 6His-tag on the MSH2 variants. Input samples (5%) were included as a control. α-tubulin served as a loading control. To obtain sufficient amount of the MSH2 variants the cells were treated with the proteasome inhibitor bortezomib (BZ) for 6 hours before harvest. See also supporting information S4 Fig. (B) The calculated thermodynamic stabilities of the test mutants (vertical axis) are plotted towards categories of normal MSH6 interaction or reduced MSH6 interaction. The horizontal line marks the median and the bars indicate the spread of the data points. *** indicates p < 0.001. (C) U2OS cells stably transfected with either vector (upper row) or wild type 6His-tagged siRNA resistant MSH2 (lower row), were transfected with control siRNA or siRNA to endogenous MSH2 and treated with either 0 or 100 nM MNNG. The surviving colonies were stained with crystal violet. (D) The siRNA-mediated knock-down of endogenous MSH2 is shown by Western blotting using antibodies to MSH2 and the 6His-tag on the recombinant (siRNA-resistant) MSH2. α-tubulin served as a loading control. Note that the recombinant variants are not overexpressed compared to the endogenous MSH2 (E) The survival of the U2OS cells, stably transfected with siRNA to endogenous MSH2, was monitored in response to increasing amounts of the alkylating agent MNNG. Note that cells containing wild type MSH2 fail to survive, whereas the vector control cells survive. The tested MSH2 variants display an intermediate MNNG sensitivity. The error bars indicate the S.E.M. (F) Estimate of the stability of MSH2 assuming that the wild type and MSH2-L187P protein can be used as approximations for fully folded and non-folded protein, respectively. The red line shows the fit to the thermodynamic model.

Fig 6. Pathogenicity predictions of MSH2 mutations.

(A) Distributions of FoldX ΔΔG scores for MSH2 variants tested in this work with short (red), intermediate (yellow), and long (light green) half-life (t½), common variants found in ExAC (dark green), known non-pathogenic (blue) and pathogenic (purple) variants, and recently identified pathogenic variants (RIPV, magenta) [30]. Dots indicate the mean ΔΔG score for each group, and the bars indicate the standard error of the mean. (B) Receiver operating characteristics (ROC) curves for the selected predictors of MSH2 variant pathogenicity: co-variation (red), FoldX ΔΔG (yellow), Rosetta ΔΔG (green), PROVEAN (cyan), SIFT (blue), and PolyPhen2 (purple). Accuracy is assessed on known pathogenic and non-pathogenic variants according to Houlleberghs et al., 2016. As the area under the curve (AUC) indicates, while all predictors show reasonable performance, ΔΔG and co-variation provide considerably higher accuracy. (C) The distribution of FoldX ΔΔG scores vs. the allele frequency of MSH2 mutations found in the Exome Aggregation Consortium (ExAC) database. Note that alleles that are found at a high frequency in the population, and are therefore unlikely to be pathogenic, appear stable (display low ΔΔG values). The three horizontal lines correspond to quartiles. For further information, refer to S6 Fig.