In Silico Analysis of Novel Titin Non-Synonymous Missense Variants Detected by Targeted Next-Generation Sequencing in a Cohort of Romanian Index Patients with Hypertrophic Cardiomyopathy

Titin, encoded by titin gene (TTN), is the largest human protein and a key component of sarcomere, involved in sarcomere assembly, mechano-sensing, and signal transduction^1,2,3. Considering its essential role in both structure and function of cardiac sarcomere, it is not surprising that TTN mutations are involved in the pathogenesis of various cardiomyopathies. Of note, due to its vast size, TTN is particularly prone to variation, sequence variants being relatively common in the general population, being found in up to 6% of reference subjects^4,5,6,7. Indeed, almost 25.000 unique TTN variants have been identified in the gnomad database (https://gnomad.broadinstitute.org).

The advent of high-throughput next generation sequencing (NGS) technology has undeniably enhanced the efficiency of genetic testing for inherited cardiac conditions, by enabling the identification of both novel disease-related genes and novel causative variants in already known genes. However, the increased genetic information is encumbered by challenges related to accurate interpretation and classification of hundreds, or even thousands of variants detected in a NGS run. In spite of clear standards and guidelines available, the definite clinical classification of sequence variants is not always possible, the majority of detected variants in cardiogenetic panels being still classified as variants of unknown significance (VUS)^8,9. Although VUS reclassifying is of paramount importance, with a substantial impact on the clinical management of patients and their relatives, it is hindered by the time and costs required to gain additional evidence, such as allele segregation within large pedigrees, or in vivo and/or in vitro functional assessment.

Our group recently reported a number of rare variants detected by targeted NGS in core and putative genes associated with hypertrophic cardiomyopathy (HCM) in a cohort of Romanian adult probands¹⁰. For the majority of identified mutations, the clinical significance is yet to be established. In the current study, we focused explicitly on novel TTN nonsynonymous missense variants identified in our cohort. Herein we present our strategy for prioritization these variants for subsequent experimental investigation to enable a definite classification.

While the role of TTN as a causative gene is largely acknowledged for dilated cardiomyopathy (DCM)^{21,22,23,24,25}, its implication in HCM is not so well established. Especially, studies have focused on the assessment and clinical interpretation of TTN truncating mutations²⁶, overlooking missense variants despite the fact that they are the most commonly observed. Accordingly, these variants are the most challenging in terms of defining their clinical significance²⁷, requiring expensive and time-consuming additional studies, such as in vivo/in vitro functional studies and sequencing of identified variants in family members. Hence, limiting the number of mutations to be further analysed, but without missing some of the disease-causing ones, is of paramount importance. Herein we present our strategy to prioritize TTN variants detected in our HCM cohort.

The general characteristics of our study cohort were reported in detail earlier, with an average age at enrolment falling in the fifth decade of life, and with male predominance¹⁰. No statistically significant differences were found in terms of clinical presentation or general characteristics between TTN positive and TTN negative individuals.

The screening for TTN variations showed 17 TTN coding variants with AF < 0.1% in reference populations, the vast majority being nonsynonymous missense variants (70%, n=12), of which over half (n=7) were novel (Table 2) and were the subject of further in silico analysis. As detailed in our prior study¹⁰, all the 7 variants were classified as VUS; none of the patients harbouring one of the 7 novel TTN variants carried a LP/P in other HCM-associated genes.

In silico prediction is an important step in assessment of novel detected sequence variants, as it is one of the criteria recommended for variant interpretation by ACMG/AMP²⁰. Polyphen²⁸, SIFT²⁹, MutationTaster³⁰ CADD³¹, and PROVEAN³² are the most frequently used algorithms in daily practice, with 79% concordance for pathogenic variants and only 33% for benign variants³³. After extensive documentation of various computational tools that can identify those missense variants most likely to have a pathogenic effect, we chose to use as first line the M-CAP score¹⁵. M-CAP is a highly sensitive pathogenicity classifier for rare missense variants in the human genome, designed to misclassify no more than 5% of pathogenic variants while significantly reducing VUS number. As opposed to the aforesaid algorithms which misclassified 26 to 38% of known pathogenic mutations, M-CAP has been proved to outperforms existing methods at all thresholds and correctly dismissed 60% of rare missense VUS detected in a typical genome at 95% sensitivity¹⁵. Three out of the 7 TTN variants in our study were predicted to be LB, thus reducing the list of VUS to be further analysed to 4 (57%) (Figure 2). c.30392G>T, c.25185G>T, c.2518G>T, and c.49G>T entered TITINdb prediction, with retrieval of Q(SASA) values and mCSM score.

Solvent accessibility surface area (SASA) is a critical attribute of proteins for determining their folding and stability. It was defined in early ‘70s as the surface described around a protein by a hypothetical centre of a solvent sphere in contact with the van der Waals surface of the molecule³⁴. Based on SASA values, amino acid residues of a protein can be classified as “buried” or “exposed”. Laddach et al. showed that disease associated mutations tend to be located to residues with a significantly lower Q(SASA)¹⁶. Indeed, by mapping single amino acid variants (SAVs) on a cu-rated database of human protein structures, Savojardo and colleagues³⁵ found that disease related SAVs are less accessible to solvent than those involved in polymorphisms, suggesting that pathogenicity is more frequently associated to the buried quality than to the exposed one.

Moreover, the mCSM algorithm within TITINdb enabled the assessment of SAVs on protein stability. Accordingly, it has been evidenced that disease associated single-point mutations were predicted to be significantly more destabilising than neutral ones¹⁶.

In our study, all 4 affected residues except p.Ala840Ser (c.2518G>T) had Q(SASA) values ≤0.3 indicating that they were buried, and that the respective variants might cause disease through disruption of the underlying domain. Additionally, computational saturation mutagenesis performed by mCSM predicted destabilizing effects for the same 3 sequence variants, strengthening the hypothesis derived from the Q(SASA) analysis that the respective altered coding sequences could lead to disease by protein stability changes.

Finally, 3 TTN missense variants (c.30392G>T, c.2518G>T, and c.49G>T) were designated as likely function-impacting and considered for further experimental studies.

Herein we presented our strategy to prioritize the novel TTN missense variants detected in a cohort of HCM patients. By applying various in silico tools, we restricted the list of VUS to be further investigated to those most likely to be disease-associated, and thus reducing the need to perform expensive and time-consuming additional studies.