Exploring digenic inheritance in arrhythmogenic cardiomyopathy

Arrhythmogenic cardiomyopathy (ACM) is a genetic disorder in which the ventricular myocardium is progressively replaced by fibro-fatty tissue. Since this occurs predominantly in the right ventricle, the disease is also known as arrhythmogenic right ventricular cardiomyopathy (ARVC). ACM is associated with progressive heart failure and severe ventricular arrhythmias, often leading to sudden death, especially in young people and athletes [1]. About half of the affected individuals harbor mutations in one of the five genes of the cardiac desmosome (PKP2, JUP, DSP, DSG2, DSC2), of which mutations in PKP2 are most common. Desmosomes are intercellular junctions that confer strong cell–cell adhesion and provide a mechanical connection between cardiomyocytes. Therefore, desmosomal defects can have deleterious effects on tissue integrity. In addition, desmosomal proteins play an important role in signaling and regulation of cell proliferation and differentiation [2]. In ACM, pathogenic mechanisms include suppression of Wnt signaling and activation of the Hippo pathway [3] leading to adipogenesis. Beside desmosomal genes, mutations in eight additional genes (DES, PLN, TGFB3, CTNNA3, LMNA, TMEM43, RYR2, and TTN) have been found to cause ACM [1]. Recently, two studies reported FLNC and CDH2 as possible novel causative genes for ACM [4, 5]. In most patients, ACM is inherited in an autosomal dominant mode with reduced penetrance (not all individuals with a causal mutation develop ACM) and variable expressivity (the severity and the nature of the symptoms may vary between affected individuals, even if they have the same causal mutation).

Autosomal dominant mutations have only been identified in up to 60% of all ACM patients [1] suggesting the existence of unknown mechanisms such as higher genetic heterogeneity, modifier genes, or cross talk between genetic background and environmental factors [6]. In fact, three loci have been mapped in ACM linkage studies, for which the causal gene has not yet been identified: ARVD3 (OMIM %602,086) at 14q12–32.3 [7], ARVD4 (OMIM %602,087) at 2q32.1–32.3 [8], and ARVD6 (OMIM %604,401) at 10p14-p12 [9]. In addition, the frequency of variants associated with ACM has been found to be much higher than expected given the phenotype prevalence in the general population, suggesting that a high number of these variants are not monogenic causes of ACM [10]. In fact, recent reports have suggested digenic inheritance as an alternative disease mechanism of ACM [11‐14]. In digenic inheritance the presence of two variants in two different genes is required for the manifestation of a clinical phenotype; in the absence of one of these variants, the other variant might be benign. For example, Xu et al. screened 198 ACM patients for variants in the desmosomal genes. Of the 38 patients in which PKP2 variants were detected, additional variants in PKP2 itself (compound heterozygosity) were identified in nine patients; variants in other desmosomal genes (digenic inheritance) were identified in 13 patients. Related family members harboring a variant in just one of these genes were unaffected by ACM. The authors concluded that the disease was caused by compound heterozygosity or digenic inheritance in these patients [12]. Rasmussen et al. investigated 12 families with variants in DSG2. In three of these families, additional variants were identified in the DSP gene in affected family members. Only individuals with both variants in DSG2 and DSP were affected by ACM, leading the authors to conclude that low penetrance of desomosmal variants in ACM patients may also be explained by digenic inheritance [13]. Cooper et al. proposed that digenic inheritance may occur as a result of variants in two genes encoding different subunits of the same protein (complex); two proteins that interact functionally; are a receptor-ligand pair; are a target gene and transcription factor; or compromise the same regulatory, biosynthetic, or degradative pathway [11]. Digenic inheritance is distinct from modifier genes: in digenic inheritance, both variants individually usually do not lead to disease, whereas in modifier genes one pathogenic variant is enhanced by a putatively contributing variant of unknown significance [15]. Non-genetic factors known to influence ACM penetrance are age, male sex and intense physical activity [16, 17].

We performed whole exome sequencing on two families, in which diagnostic tests have identified a PKP2 mutation in affected and healthy individuals. Assuming a digenic mode of inheritance, we determined all genes, where in addition to the observed PKP2 variant a second causal variant was expected to be present in either the affected individuals or the PKP2 carriers. Filtering and prioritizing these genes, we determined four candidate genes in the first, and eleven candidate genes for digenic inheritance in the second family.

In this study we investigated the genetic cause of ACM in two families using whole exome sequencing. Since all affected and some unaffected individuals were known to harbor PKP2 variants, we investigated whether a second gene was involved in a digenic inheritance pattern, with the second gene either causing ACM in the affected individuals together with PKP2, or compensating the effect of the PKP2 variants in the carriers. We identified 74 and 212 genes in families 1 and 2, respectively, which carried variants consistent with the mode of digenic inheritance. To obtain results that can be easily interpreted, we restricted our analysis to genes either associated with ACM or related to PKP2. In family 1 we identified four genes that are annotated with the same biological process as PKP2. In family 2 we identified the ACM associated gene TTN and ten genes related to PKP2 through a shared biological process or protein interactions (see Fig. 2).

Of the four PKP2-related Fam1 genes, the genes homologous to DAG1, TCF25, and CTBP2 have been linked to cardiomyocyte proliferation or heart development in mice and, in case of TCF25, also in human. DAG1 and TCF25 negatively regulate heart development, while a knock out of CTBP2 leads to a lethal malformation of the heart in mice. A variant in DAB2IP has been associated with coronary heart disease in two studies, indicating that this gene might also play a crucial rule for the normal functioning of the heart. It has been reported that β-dystroglycan, a protein product of DAG1, directly binds to the Hippo pathway effector Yap to inhibit cardiomyocyte proliferation in mice [42]. In particular, the Hippo pathway and DGC cooperatively regulate tissue growth in mouse hearts after injury. Yap and the Hippo pathway have been directly implicated in ACM pathogenesis [30]. TCF25 (previously named NULP1) was suggested as a transcription factor that negatively regulates the serum response factor (SRF). SRF controls muscle differentiation and cellular growth and regulates cardiac genes. SRF over-expression has been shown to cause cardiomyopathy and cardiac hypertrophy in mice. Therefore, TCF25 may function as a transcriptional repressor of SRF in human heart development [43]. DAB2IP acts as a tumor suppressor gene, and is inactivated by methylation in prostate and breast cancers. A genome-wide association study found the rs7025486 variant in DAB2IP associated with coronary heart disease, which was replicated in a second study [44]. CTBP2 encodes two proteins, a transcriptional repressor and a major component of synaptic ribbons. Silencing the homologous Ctbp2 gene in mice causes defects in heart morphogenesis and results in early embryonic lethality [45]. Ctbp2-null mice show similar axial truncation phenotypes as mice with mutations in some Wnt target genes, suggesting that CTBP2 may be a regulator of Wnt-mediated gene expression [45]. Indeed, CtBP2 acts as corepressor of C/EBPα, an early regulator of adipogenesis, and target of the Wnt signaling pathway [46]. Furthermore, Sox6 has been found to bind Ctbp2 to repress the fibroblast growth factor 3 [47] and Sox6 to regulate the cardiac myocyte development in mice [48]. Although none of the variants in these four genes are predicted to be deleterious and the variant in DAG1 is even predicted to stabilize protein structure, they could nevertheless affect protein stability, flexibility, and interaction with the other binding partners. However, in the present study we could not find any indication that these genes may act together with PKP2 to cause the ACM phenotype in the affected individuals of Fam1.

Since TTN is a known ACM associated gene, it is a likely candidate in the second family. TTN encodes for titin, the largest human protein with isoforms ranging from about 27.000 to 36.000 amino acids. Titin is functionally linked to the desmosome (and thereby to PKP2), since titin filaments are a key component of sarcomeres and connect to the transitional junction at the intercalated disk [8]. In a cohort of 38 ACM families, Taylor et al. identified novel TTN mutations in 18% of the families [8]. In addition to ACM, TTN has been associated with dilated, hypertrophic, and restrictive cardiomyopathy [49]; its association with hypertropic cardiomyopathy, however, is still under debate [50]. The affected patient Fam2.II.1 and his father both harbor two rare heterozygous missense variants that are predicted deleterious. The Gln24857His variant is located in titin’s only kinase domain at a conserved position and is predicted to destabilize protein structure, while Arg23483His is located in one of the 132 Fn3 domains. Therefore, the Gln24857His variant is more likely to impair titin function than the Arg23483His variant, even though disease causal variants in repetitive titin domains have been reported [8]. Studies with transfected cell lines have shown that heterozgyous mutations, in contrast to homozygous mutations, still allow for functional sarcomeres but may alter the organizational characteristics and impair the normal cardiac function [49]. These findings agree well with the hypothesis that either one or both of these variants alter the structure of titin and only lead to ACM in combination with the PKP2 mutation.

In addition to the genes described here in more detail, there are other promising candidates for the second causal gene in Fam2 (see Table 2). For example, of the PKP2-related genes, NOTCH2 and SCN5A harbor one of two compound heterozygous variants that have been reported to cause congenital heart disease and isolated conduction disease, respectively [51, 52]; DMD harbors a neutral hemizygous variant in the affected Fam2 individual, a gene where recessive variants can cause muscle dystrophy; DAG1 and MKKS are associated with recessive diseases, yet the variants in the affected individuals are heterozygous; IRF1 is associated with non-cardiac diseases; DSC1, a desmosomal gene not associated with heart disease, harbors a common missense variant that is predicted neutral. Since all of these genes are interesting candidates for follow up studies, it would be interesting to test whether the same or other rare variants in our candidate genes can be identified in a large cohort ACM patients, both in patients carrying desmosomal mutations or other ACM related mutations as well as in genetically unsolved cases.

Digenic inheritance has previously been reported as a disease-causal mechanism for ACM, however, these studies have focused on desmosomal genes. As a result, there are no reports of digenic inheritance in ACM with PKP2 and a non-desmosomal gene such as TTN.

We are aware of limitations in our study. Both families have relatively few members, which resulted in a large set of variants and genes as possible candidates for digenic inheritance. Only 85% of the exome was covered with at least 10X. Since we required a minimum coverage of 10X to accept a variant call, 15% of the exome could not be investigated. However, coverage at the ACM genes was well above average, so it is unlikely that variants were missed in these genes. Due to the large number of candidate genes in each family, we restricted our analysis to ACM or PKP2-related genes, potentially removing causative digenic genes with unknown associations. Even though the TTN variants in Fam2 were confirmed by Sanger sequencing, no functional validation of the variant effects was performed. Therefore, it still needs to be shown if TTN or any of the candidate genes in Fam1 truly cause ACM together with PKP2 in the respective family. A functional validation could be performed based on induced pluripotent stem cell (iPSC) models from the ACM-affected individuals, where the PKP2 variant or the TTN/Fam1 variants are reversed [53]. In the iPSC derived cardiomyocytes the effect of the genetic variants could be investigated by comparing fat accumulation and cell electrophysiology to the double mutant cells. Other cell models that could be employed for validation are progenitor cells (they differentiate easily in vitro), non-contractile cardiac mesenchymal stromal cells (ideal for studying lipid metabolism), or primary or immortalized cardiomyocytes (enable investigation of gap-junctions and ion-channels) [54]. Yet, even if successful, such experiments would demonstrate the mode of effect in the respective family, while general conclusions about the role of TTN/Fam1 genes in ACM could not necessarily be drawn from them. To evaluate the roles of these genes in ACM more generally, other ACM patients carrying desmosomal variants could be checked for rare variants in the respective genes. Unfortunately, we currently do not have additional ACM patients for testing and the NCBI Sequence Read Archive does not contain public whole exome or whole genome sequence data of ACM patients. The search for genes with a digenic effect is a considerable challenge since variants in both relevant genes do not necessarily have a pathogenic effect when occurring individually [11]. The functional or structural change caused by the variant in either protein may be subtle, and may for example lead to a change at a protein binding affinity or a change in gene expression. Therefore, standard criteria usually applied to evaluate the likelihood of variant pathogenicity like rarity and computational predictions might not be well suited. Consequently, we did not exclude variants based on these criteria, yet in the absence of functional validation and more appropriate models, we prioritized and discussed our results according to these methods. We would like to recall that we did not distinguish between variants that were present in the affected and not in the carriers and variants present in the carriers but not in the affected, since we were interested in genes whose function might differ between affected individuals and carriers due to the variants. However, we point out that of the 17 variants listed in Table 2, only three (Ile339Val in MKKS, His558Arg in SCN5A, and Ser321Leu in DROSHA) are present in the carriers and not in the affected individuals, suggesting that our strategy of prioritizing based on rarity and predicted pathogenicity is appropriate. Finally, we acknowledge the possibility that more than two genes could be involved in the pathogenesis (oligogenic inheritance) or, contrarily, that environmental factors could influence the penetrance of the PKP2 variants without other genetic variants having an effect on the development of ACM. However, in our study, the main non-genetic disease modulators (age, sex, and physical exercise) are not sufficient to explain the different phenotypic expression in affected individuals and carriers in the two analyzed families (see Table 1). In Fam1, both the carrier Fam1.III.1 and the affected Fam1.III.2 are male and are close in age, and carrier Fam1.III.1 and the affected Fam1.III.3 are both physically active. In Fam2, both the affected Fam2.II.1 and the carrier Fam2.II.2 are male, physically active, and relatively close in age.