Computational prediction of protein subdomain stability in MYBPC3 enables clinical risk stratification in hypertrophic cardiomyopathy and enhances variant interpretation

Abstract

Purpose: Variants in MYBPC3 causing loss of function are the most common cause of hypertrophic cardiomyopathy (HCM). However, a substantial number of patients carry missense variants of uncertain significance (VUS) in MYBPC3. We hypothesize that a structural-based algorithm, STRUM, which estimates the effect of missense variants on protein folding, will identify a subgroup of HCM patients with a MYBPC3 VUS associated with increased clinical risk.

Methods: Among 7,963 patients in the multicenter Sarcomeric Human Cardiomyopathy Registry (SHaRe), 120 unique missense VUS in MYBPC3 were identified. Variants were evaluated for their effect on subdomain folding and a stratified time-to-event analysis for an overall composite endpoint (first occurrence of ventricular arrhythmia, heart failure, all-cause mortality, atrial fibrillation, and stroke) was performed for patients with HCM and a MYBPC3 missense VUS.

Results: We demonstrated that patients carrying a MYBPC3 VUS predicted to cause subdomain misfolding (STRUM+, ΔΔG ≤ -1.2 kcal/mol) exhibited a higher rate of adverse events compared with those with a STRUM- VUS (hazard ratio = 2.29, P = 0.0282). In silico saturation mutagenesis of MYBPC3 identified 4,943/23,427 (21%) missense variants that were predicted to cause subdomain misfolding.

Conclusion: STRUM identifies patients with HCM and a MYBPC3 VUS who may be at higher clinical risk and provides supportive evidence for pathogenicity.

Fig. 1. Patients with a MYBPC3 VUS identified as deleterious by STRUM (STRUM+) are associated with an increased risk for adverse hypertrophic cardiomyopathy (HCM)-related outcomes.

Selection within Sarcomeric Human Cardiomyopathy Registry (SHaRe) of patients with HCM carrying a single MYBPC3 missense variant of uncertain significance (VUS) is shown on the left. One hundred five patients carry a single MYBPC3 missense VUS, covering 77 distinct MYBPC3 VUS. Kaplan–Meier curves, median event free survival (years), and hazard ratio with corresponding 95% confidence interval (CI) reveal that patients carrying a STRUM+ MYBPC3 VUS (red) exhibited a higher rate of adverse HCM-related outcomes (overall composite) compared with patients carrying a STRUM- variant (black).

Fig. 2. Patients with a MYBPC3 variant of uncertain significance (VUS) identified as deleterious by STRUM (STRUM+) exhibit clinical outcomes similar to patients with a MYBPC3 pathogenic variants.

Selection within Sarcomeric Human Cardiomyopathy Registry (SHaRe) of patients with hypertrophic cardiomyopathy (HCM) and a single MYBPC3 pathogenic variant (MYBPC3 -Path-all) and patients with HCM without a sarcomere gene variant after clinical genotype analysis (Sarc−) is shown on the left. Kaplan–Meier curves, median event free survival (median survival), and hazard ratio with corresponding 95% confidence interval (CI) reveal patients carrying a STRUM+ MYBPC3 VUS (red) exhibited overall composite outcomes similar to MYBPC3-Path-all patients (blue, p value 0.5036). Whereas, patients carrying a STRUM- variant (black) exhibited a lower rate of adverse HCM-related outcomes (overall composite) similar to Sarc- patients (gray).

Fig. 3. STRUM is complementary to CardioBoost.

Results of computational analysis for each unique MYBPC3-Benign (gray triangles, n = 110) and MYBPC3-Path (red circles, n = 19) variant. (a) STRUM. (b) CardioBoost. Mean and SEM for each group depicted. The cutoff for deleterious variants for STRUM was ΔΔG ≤ −1.2 kcal/mol. The cutoff for deleterious variants for CardioBoost (CardioBoost +) was a probability score > 0.90; this is graphed as 1-CardioBoostScore < 0.10. C3 pathogenic variants are depicted in open circles in (a) and (b). (c) Statistical analysis of computational method utilized here in STRUM (Fig. 3), CardioBoost (Fig. 3), SIFT (Figure S6), and PolyPhen-2 (Figure S6) is shown including odds ratio, 95% confidence interval (CI), sensitivity, and specificity. (d) Using the same patient selection criteria in the Sarcomeric Human Cardiomyopathy Registry (SHaRe) detailed in Fig. 1, patients with hypertrophic cardiomyopathy (HCM) and a MYBPC3 missense variants of uncertain significance (VUS) were analyzed by CardioBoost. CardioBoost (+) was a probability score > 0.90, CardioBoost VUS ≥ 0.10 and ≤0.90, and CardioBoost (−) < 0.10. Of the 105 patients analyzed by STRUM 19 were CardioBoost (+). Kaplan–Meier curves reveal that patients carrying a CardioBoost (+) MYBPC3 VUS (red) exhibited higher rates of adverse HCM-related outcomes (overall composite) than patients carrying a CardioBoost (−) MYBPC3 VUS (black); however, the null hypothesis could not be excluded, p value 0.0945. This remains true when comparing patients carrying a MYBPC3 VUS that is CardioBoost (+) (red), CardioBoost (VUS) (gray), and CardioBoost (−) (black) (p value 0.2534).

Fig. 4. Structural analysis of pathogenic missense MYBPC3 variants.

MyBP-C (the protein encoded by MYBPC3) domains C3, C6, and C10 were structurally modeled using I-TASSER^– (PyMOL,cartoon, green). Wild-type residues that are affected by missense pathogenic variants are depicted in red (PyMOL, sticks). (a) For the C3 domain, the I-TASSER model is aligned with an available NMR structure (2mq0.pdb, blue, PyMOL cartoon). Pathogenic variants within C3 largely cluster in a surface-exposed region. (b) C6 domain and (c) C10 domain pathogenic variants do not cluster within a specific region of the domain. (d) Results of STRUM analysis for MYBPC3 pathogenic and benign variants within C3, C6, and C10 are shown, with mean and SEM for each group depicted. Graph is labeled to indicate variants predicted to be deleterious.