Spatial and Functional Distribution of MYBPC3 Pathogenic Variants and Clinical Outcomes in Patients With Hypertrophic Cardiomyopathy

Supplemental Digital Content is available in the text.


Introduction
Familial hypertrophic cardiomyopathy (HCM) is an autosomal dominant condition, and pathogenic variants in cardiac myosin binding protein C (protein abbreviation, MyBP-C, encoded by the gene, MYBPC3) are the most common cause. 1 MyBP-C is a sarcomeric protein that binds both actin and myosin and regulates cardiac contractility by modulating myofilament sliding velocity. 2,3 Because a large number of unique MYBPC3 variants have been associated with HCM, small, single-center cohorts have had limited capacity to systematically analyze genotype-phenotype relationships, particularly given the marked variability in penetrance of MYBPC3-associated HCM. [4][5][6][7] Resolving these gaps in knowledge will be critical to further personalized risk assessment and management of patients with HCM.
Most MYBPC3 pathogenic variants are frameshift, nonsense, or splice-site variants that result in premature termination codons (PTCs). PTC-containing transcripts are targeted for degradation through nonsense mediated RNA decay (NMD), and hence may cause disease through allelic loss of function (resulting in reduced levels of MyBP-C). Consistent with allelic insufficiency, we and others have shown a ~40% reduction in MyBP-C in heart tissue from HCM patients, 8,9 due to a rate-limiting reduction in MYBPC3 mRNA. 10 These studies support the hypothesis that truncating variants in MYBPC3 likely exert a similar primary effect, independent of the specific variant locus. However, comparative analyses across the full genotypic and phenotypic spectrum of truncating variants have not been possible due to the small size of previously-available cohorts. Distinct from truncating MYBPC3 variants, non-truncating pathogenic variants (including missense and short in-frame deletions/insertion variants) account for ~15% of MYBPC3 HCM. The mechanism(s) of MYBPC3 non-truncating pathogenic variants are largely unknown, and it is unclear whether phenotypic expression or clinical outcomes are different in patients carrying missense variants. 7,11 A greater understanding of the diseasecausing mechanism(s) of non-truncating MYBPC3 pathogenic variants through functional analyses could improve adjudication of variant pathogenicity and expand the pool of clinically actionable gene test results.
Here, we utilize the largest registry of combined genetics and clinical data for HCM to date, the Sarcomeric Human Cardiomyopathy Registry 1 (SHaRe), to generate an adjudicated and comprehensive compendium of MYBPC3 variation, analyze regional variation within MYBPC3, and correlate clinical phenotypes. We find that pathogenic truncating variants are homogeneously distributed throughout the gene, in contrast to non-truncating MYBPC3 pathogenic variants that cluster in specific protein domains. Disease severity is highly variable in MYBPC3 HCM, and we show that this variability is largely independent of variant location or the specific truncating or non-truncating variant based on both disease severity metrics and clinical outcomes. Finally, we experimentally test functional effects of non-truncating pathogenic variants in the identified variant-enriched domains and identify a subset that exhibit allelic loss of function.

Methods
The methods used are described for purposes of replicating the study procedure. Individual patient data will not be made available for purposes of reproducing the results. The study was independently approved by the institutional review board at each center. A detailed methods section is available in the Supplemental Data. (LV) wall thickness was greater in the relatively small subset of pediatric patients with nontruncating variants, but was similar in other age groups. LV ejection fraction was similarly elevated at a young age in both groups and declined similarly in later age groups. Left atrial diameter progressively increased to a similar extent with increasing age in both truncating and non-truncating groups (with the single exception of the smaller sized group of non-truncating variant patients at age >60; N=17). The distributions of maximum wall thickness, left atrial size, and age of diagnosis among non-truncating and truncating pathogenic variant cases are shown in Figure 1A-C. Time to event analysis for composite adverse outcomes revealed no difference between patients with non-truncating or truncating pathogenic variants ( Figure 1D), and this result was not different when including probands only (Supplemental Figure 1A). There were similarly no differences between the groups for heart failure or ventricular arrhythmia composite outcomes (not shown).

Pathogenic Variants
If truncating MYBPC3 variants cause allelic insufficiency as their primary consequence, then the location of the variant within the gene would not be expected to influence the disease severity.
To test this, we categorized truncating MYBPC3 pathogenic variants into quartiles by 5' to 3' location and compared morphologic markers of severity and adverse outcomes. We found no statistically significant difference in maximum wall thickness or age-adjusted left atrial diameter among these groups (Figure 2A-B). Composite adverse events were also similar when stratified by variant location quartile ( Figure 2C) or by truncating variant type (Supplemental Figure 1B).  Figure 2D). Additionally, adverse events were similar in each founder population compared to patients with non-founder truncating variants ( Figure 2E).

Morphologic Severity, Adverse Events, and Variability in Phenotype
HCM is known to have broad variance in phenotypic severity across individuals. This variance in expressivity has been thought to be due to heterogeneity of effect size of underlying pathogenic variants, the influence of background genetic variation (i.e. genetic modifiers), and clinical comorbidities. 14-17 Taking advantage of the founder populations, we compared variances across these subgroups each carrying identical pathogenic variants. As shown in the histogram plot of maximum wall thickness in Figure 2F, the 4 founder populations demonstrate similar variance (mean of standard deviations 5.96 + 0.79 mm) compared to the remainder of the truncating variant population (standard deviation 5.98 mm, p=NS). Taken together, these findings indicate that truncating variants likely exert a similar primary effect, and the marked variance in disease phenotype among truncating variant patients is caused by additional genetic and non-genetic factors, independent of the driving MYBP3 variant.

Pathogenic Variants in HCM
MYBPC3 truncating variant types in SHaRe patients consisted of 110 unique insertion/deletion variants, 55 unique nonsense variants, and 69 unique splice variants (Supplemental Table 1).

Pathogenic variants on Myofilament Incorporation and Degradation Rate
While strong evidence supports allelic insufficiency is the primary mechanism across the spectrum of truncating MYBPC3 variants, the mechanism(s) of non-truncating MYBPC3 pathogenic variants has not been resolved. We hypothesized that some non-truncating MYBPC3 pathogenic variants may also cause loss of function, but through lack of normal protein localization or structural stability rather than reduced expression. Therefore, we first tested whether exogenously expressed MyBP-C with non-truncating pathogenic variants incorporates normally into the myofilaments. We expressed FLAG-epitope labeled MyBP-C with or without pathogenic non-truncating variants in neonatal rat ventricular myocytes (NRVMs) and analyzed localization by immunofluorescence. We found that MyBP-C containing representative C3 or C6 domain non-truncating variants localized normally to the sarcomere A bands while MyBP-C containing C10 domain non-truncating variants was essentially absent from the myofilaments  Table 2). In contrast, most pathogenic variants in the C3 and C6 domains resulted in MyBP-C protein half-lives that were not significantly different from wild-type MyBP-C, though the Arg502Trp variant resulted in a modest 36% shorter protein half-life compared to wild-type (p=0.04). Paradoxically, the Arg810His variant resulted in a 44% prolonged MyBP-C protein half-life compared to wild-type (P=0.008).

Discussion
Despite genetic variants in MYBPC3 being the most common cause of familial HCM, identifying genotype-phenotype correlations has been elusive, due to the large number of individual pathogenic variants and small numbers of patients previously available to study from single centers. Here, we harness the largest cohort of genotyped HCM patients to comprehensively describe MYBPC3 genetic variation and associated clinical phenotypes.
A convergent theory of allelic insufficiency from truncating MYBPC3 variants has emerged from human tissue, rodent, and iPSC model systems. 8,10,[20][21][22] Reduction in MyBP-C relative to myosin alters sliding velocities as actin-myosin sliding reaches the C-zones, where MyBP-C is specifically present, resulting in a more rapid contractile deceleration toward peak force development. 2,8,10,23 However, clinical-genetics data to confirm this theory have been notably absent. Our findings of a homogeneous distribution of HCM-causing truncating variants throughout MYBPC3, similar phenotypic severity across spatial quartiles in the coding sequence, and similar adverse event rates support the theory that disease results from a biologically similar loss of function mechanism across truncating variants, as opposed to dominant negative consequences from truncated MyBP-C protein (which has not been detectable in human heart or cellular models 9, 10, 12 ). Furthermore, we found that 4 founder populations, with distinct truncating variants and sizable numbers in SHaRe, exhibit similar disease severity and adverse event rates as compared to non-founder truncating variant patients. This result extends findings from a single site investigation of the Netherlands founder cohort 24 , and counters smaller series that have suggested less pathogenic effects in truncating variant founder cohorts. 25,26 A major implication of these results is that patients with truncating MYBPC3 variants would likely derive similar benefit from targeted treatment approaches irrespective of the specific location of the truncating variant.
We further leveraged the truncating variant founder populations in SHaRe to investigate the variability in expressivity in HCM. HCM exhibits vast genetic and phenotypic heterogeneity, which has been a major challenge in determining genotype-phenotype relationships. 27 We found that patients with founder variants had a similar distribution of phenotypic features and clinical outcomes as non-founder HCM patients with truncating variants. This finding suggests that the variability in disease phenotype among MYBPC3 truncating variant carriers is not dictated solely by the primary pathogenic variant. An important implication of this finding is that additional genetic and non-genetic modifiers likely account for the broad variance in phenotypic severity among patients with MYBPC3 HCM.
We also demonstrate that MYBPC3 non-truncating pathogenic variants, accounting for 15% of MYBPC3 pathogenic variants, generally had a similar phenotypic effect as truncating variants. Minor differences between the groups included a modestly greater proportion of pediatric diagnoses in the non-truncating group and modestly reduced prevalence of LVOT obstruction. However, maximal LV wall thickness across all other age groups, and adverse event rates were highly similar.
Because non-truncating variants are robustly adjudicated in SHaRe, we were able to identify strong evidence of domain clustering. We then demonstrated that a subgroup of nontruncating pathogenic variants (those in the C10 domain) renders the resultant mutant protein susceptible to rapid degradation, resulting in a loss of function mechanism similar to truncating variants. In contrast, we show no destabilization in the majority of C3 and C6 domain mutant proteins, which integrate normally in myofilaments. The C3 variant Arg502Trp alters the electrostatic properties of the domain, but how this alteration affects MyBP-C function is not known. 28 In engineered heart tissue, overexpression of the C3 mutant Gly531Arg (not present in SHaRe), caused hypercontractility at low calcium levels and was not able to rescue MyBP-C knock-out tissues. 29 Further study is required to fully elucidate the impact of C3 and C6 pathogenic variants on contractile function.
In contrast to the clustering evident for pathogenic non-truncating variants, VUS's in MYBPC3 were relatively common in the SHaRe cohort (N=148, 87% of all unique MYBPC3 non-truncating variants). Accurate prediction of pathogenicity of sarcomere VUS's is a major challenge for interpretation of genetic testing results and determination of the suitability for cascade testing in family members. Although we confirmed enrichment of non-truncating pathogenic variants in specific MyBP-C domains, as also shown in an independent cohort by Walsh and colleagues 30  Several limitations to our study should be considered. This was a retrospective, observational study. Although we analyzed by far the largest cohort of HCM patients with MYBPC3 pathogenic variants to date, the study may be underpowered to detect small differences in phenotype severity or adverse events between groups. In addition, we analyzed pathogenic variant carriers in groups based on variant type and location, but further subdivision to individual pathogenic variants was only feasible for the founder subpopulations. As such, differences in effect size for specific pathogenic variants, particularly in the case of the non-truncating variants, could still exist. Both the SHaRe population and gnomAD populations predominantly consist of individuals from European ancestry. Although these attributes lend confidence to the calculation of the odds ratios for HCM-associated versus common population variants reported here, the results are not necessarily representative of genetic variation in other ancestries. Relatedly, the SHaRe population has a greater proportion of patients with HCM with truncating founder variants due to inclusion of certain European sites (The Netherlands, Italy). Lastly, we strategically focused experimental testing of non-truncating pathogenic variants to the impact on protein stability and only examined a subset of representative variants. Future work will be needed to further resolve the functional effects of pathogenic non-truncating MYBPC3 variants that do not destabilize the protein structure and extending these analyses more comprehensively across MYBPC3 non-truncating variants.
In conclusion, we leverage the largest cohort of patients with MYBPC3 pathogenic variants to date to develop a compendium of benign, pathogenic, and uncertain MYBPC3 variants and identify genotype-phenotype correlations. Our results demonstrate that phenotypic severity and clinical outcomes are similar across the range of MYBPC3 pathogenic variant carriers, without obvious associations based on the location of truncating variants, founder or non-founder truncating variant carriers, or truncating versus non-truncating variants. These findings highlight the need to identify additional background genetic and non-genetic modifiers that influence the broadly variable HCM disease phenotype. In addition, we show that nontruncating pathogenic variants cluster in particular MyBP-C domains, with those variants in the C10 domain exhibiting protein destabilization leading to loss of function, in contrast to a second subset exhibiting normal myofilament incorporation and stability.     To determine whether non-truncating MYBPC3 pathogenic variants alter protein stability, NRVMS were transduced with adenoviral constructs expressing wild-type (WT) control and non-truncating mutant MyBP-C. Cyclohexamide was administered at 0, 30 minutes, 1 hour, 3 hours, 6 hours, and 12 hours to inhibit protein synthesis and MyBP-C was measured (see Methods). Data from two or more independent experiments performed in quadruplicate were fit to a first order exponential decay curve. The same control data (from FLAG-labeled wild type expressed MyBP-C) is depicted on each graph (A-C). A-B. C3 and C6 mutant MyBP-C demonstrates similar degradation rates as control. C. C10 mutant MyBP-C demonstrates rapid degradation compared to control. Data is represented as mean + standard error of the mean. The calculated half-lives with 95% confidence intervals are shown in Table 2.