Familial dilated cardiomyopathy (DCM) is a significant cause of heart failure and heart transplantation. DCM is typically a monogenic disorder with autosomal dominant (AD) inheritance but autosomal de novo, recessive, X-linked, and mitochondrial patterns have also been observed (1, 2). Although genetic testing is recommended in familial DCM, it is not routinely used in all centers (3, 4). Identifying the underlying cause of DCM helps recognize family members at risk. Patients with a diagnostic variant may have a worse prognosis (5, 6).

The genetic etiology of DCM is more heterogenous than in other cardiomyopathies (7). Disease-causing variants have been identified in over 60 genes that encode a variety of proteins in sarcomeres, nuclear lamina, desmosomes, cytoskeleton, and mitochondria (1, 2). Truncating titin variants (TTNtv) are the most common genetic cause of DCM, and other common genes include LMNA, MYH7, RBM20, TNNT2, TPM1, FLNC, DSP, and DES (1, 8, 9). Currently over half of the patients undergoing comprehensive genetic testing are left without a molecular diagnosis even when patient selection follows strict DCM criteria (10–15).

We aimed to evaluate the diagnostic yield of genetic testing in a real-life setting in patients with a clinical suspicion of DCM, referred for genetic testing from multiple centers around the world. We present the genes in which diagnostic variants were identified and the clinical variables that influenced the likelihood of observing a diagnostic variant.

This study comprised 2,088 patients with clinically suspected inherited DCM. Over half were males (63%, n = 1,308). The mean age at the time of genetic testing was 46.2 years. Patient characteristics and clinical variables are outlined in Table 1.

A diagnostic variant was found in 24.6% of the 2,088 patients. The results in this heterogenous population are comparable to other studies, even though in previous studies the cohorts were selected using stringent DCM diagnostic criteria. Our cohort was selected based on suspected DCM and no supporting clinical data were used in patient selection. This heterogenous cohort is comparable to the real-life patients with suspected DCM referred for genetic testing.

Previous studies with patients fulfilling DCM diagnostic criteria have observed an overall diagnostic yield of 19%–47% (11, 12, 15, 20). In the recent ESC EURObservational Research Programme (EORP) Cardiomyopathy Registry study 33% of the patients had a diagnostic variant (13). The study included patients from 18 countries and strict DCM criteria were used as an inclusion criterion (13). The slightly higher overall diagnostic yield in the EORP study could be partly explained by the results of the genetic testing not being strictly controlled; the pathogenicity of the variant was self-reported by the investigators at the time of inclusion (13).

We observed that positive family history, lower age at diagnosis, paroxysmal AF, arrhythmias at rest, and the absence of LBBB were independently associated with molecular diagnosis. Previously observed factors associated with diagnostic test results have been skeletal myopathy, family history of DCM, low voltage on ECG, absence of hypertension, and absence of LBBB (6, 20, 21). Diagnostic yield has been consistently higher in those with a familial disease (36%–64%) when compared to sporadic DCM (13%–36%) (11, 12, 15, 20, 22). Genetic testing has been generally recommended in familial DCM, but taken together, these findings suggest that this recommendation might be too strict as LP/P variants have been identified in up to a third of those with no affected family members (4, 11, 12, 15, 22).

Even though LBBB is typical in patients with LMNA variants, it is rarer in other genetic etiologies, such as those with TTNtv (23). As the number of patients with LMNA variants was relatively low, the tendency for LBBB was not reflected in the overall association. AF has been associated with diagnostic test results, but the subtype has not been defined (5, 6, 20). We observed chronic AF to be equally common in the groups with and without a diagnostic variant whereas paroxysmal AF was more common in those with a positive genetic test.

Most diagnostic variants were observed in TTN, followed by DSP, LMNA, MYH7, TNNT2, and FLNC. The distribution of variants was similar when compared to previous reports (6, 11, 15). In our cohort, two diagnostic variants in dominant genes were observed in 2.7%. This is consistent with the EORP registry study and the recent cohort study by Stroeks et al., in both studies 2.3% of the patients had at least two diagnostic variants (13, 24). Stroeks et al. observed that patients with multiple variants did neither appear to have more severe disease nor an earlier disease onset (24). By contrast, some case studies have reported an earlier onset and more severe clinical course in patients carrying multiple variants in DCM-associated genes (25–28). As there are no larger studies of patients with multiple variants, it is still not known, how carrying more than one disease-causing variant affects the patient's clinical course. In our cohort, only 14 patients carried more than one diagnostic variant and the patients had different variant combinations. Due to small sample size as well as limited clinical data, we were not able to carry further analysis concerning the possible severity of the phenotype. Even though future research is still needed to assess how having more than one variant may affect the severity and prognosis of the disease, the reported prevalence of more than one variant in multiple cohort studies highlights the importance of comprehensive genetic testing. Regardless of the possible effect of multiple variants on the phenotype, the identification of all pathogenic variants contributing to the disease affects future family screening and cascade genetic testing.

We identified diagnostic variants also in genes less frequently associated with DCM. Diagnostic variants in PRDM16 were observed in three patients with DCM, two of them had also features of LVNC. Deletions, missense, and frameshift variants in PRDM16 have been associated mostly with LVNC but also with DCM (29, 30). PRDM16 is predicted to be intolerant to LoF with a maximal probability of loss-of-function intolerance (pLI) value of 1.00 in gnomAD. Additionally, in our enrichment analysis LoF variants in PRDM16 were significantly rarer in the gnomAD reference population when compared to our cohort.

One patient with severe left ventricular dysfunction (LVEF <20%) had a diagnostic variant in PLEKHM2. To our knowledge this patient represents the third family with homozygous/compound heterozygous truncation of PLEKHM2. In previous studies the patients with PLEKHM2 variants presented with DCM and features of LVNC, and a family history of SCD (31, 32).

One patient with LP/P variants in CASZ1 and MT-TL1 presented with DCM, short stature, bilateral hearing-loss, low muscle mass, bilateral ptosis, diabetes, and multiorgan dysfunction. Heterozygous LoF variants in CASZ1 have been associated with DCM (33–35). CASZ1 is also involved in 1p36 deletion syndrome, which is characterized with dysmorphic facial features, intellectual disability, developmental delay, hearing loss, seizures, cardiomyopathy, and cardiovascular malformations (36). The pLI score of CASZ1 in gnomAD is 1.00 suggesting that this gene is intolerant for LoF variation. According to the available evidence, LoF can be considered an established disease-mechanism in this gene. One patient had a homozygous truncating variant in GCOM1/MYZAP. DCM caused by homozygous GCOM1/MYZAP variants have been reported by us and one other group (37, 38). The GRINL1A complex transcription unit (CTU) contains the downstream gene POLRM2 and the upstream gene MYZAP, as well as the GCOM1 combined gene that uses exons from the upstream and downstream genes as well as its own exons (39). Due to the complexity of the GRINL1A CTU, a variant in GCOM1 may also affect MYZAP transcripts, thus, variants can be called into different genes within the CTU. Relying on genotype tissue expression data, the affected gene should be called MYZAP.

Consistent with previous studies, most LP/P variants followed an AD inheritance. We observed LP/P variants across 45 genes. Currently over 60 genes have been associated with DCM. Due to lack of clinical data and varying variant interpretation, some of these 60 genes lack reliable evidence (10). Recent studies have aimed to re-assess the genes associated with monogenic DCM (14). Mazzarotto et al. defined 12 genes with strong association with DCM (TTN, LMNA, MYH7, TNNT2, TPM1, DSP, VCL, BAG3, TNNC1, ACTC1, NEXN, and PLN (14). Verdonschot et al. observed that almost 90% of their LP/P variants were in these 12 genes and additional 1.2% of the LP/P variants were in RBM20 and FLNC (20). Stroeks et al. observed that using a gene panel including these 14 genes, an LP/P variant was found in 16.9% of the patients, whereas an extended panel with 48 genes resulted in a diagnostic yield of 17.8% (5). In our cohort 388 (72.7%) LP/P variants were observed in these 12 genes and additionally 41 variants were observed in RBM20 and FLNC. If the patients in our cohort were tested with panels that included only these 14 genes, 19.7% (105/534) of the LP/P variants would have been missed. In our cohort, no variants were observed in VCL or ACTC1. This could be partly explained by the observation that VCL and ACTC1 are more prevalent in pediatric patients (14).

The reassessment of genes has been based on individuals who fulfil strict DCM diagnostic criteria. This does not completely reflect the real-life situation in the clinics as sometimes the phenotype might overlap with other cardiomyopathies. Assigning a specific phenotype might come down to the personal interpretation of the clinician and the imaging modalities available. Previous publications suggest that the number of genes in a panel could be reduced, however, due to difficulties in phenotyping the patient, a larger panel combining different cardiomyopathy genes might be necessary.

Patients with LP/P variants have a worse prognosis, and they experience more adverse events (6, 20). Genetic etiology may affect clinical decision making. In our cohort 16% of those with molecular diagnosis had a LP/P variant in LMNA, RBM20, FLNC, or PLN. Pathogenic variants in these genes have been associated with increased risk of sustained ventricular arrhythmias. The ESC guidelines recognize pathogenic variants in these genes as a risk factor in patients with DCM when considering ICD implantation (18).

Data were based on the requisition forms completed by the ordering provider. Complete data were not available on all patients. No DCM criteria were used to confirm the diagnosis. The main genes associated with DCM were evaluated in all patients, but all patients were not tested with the same panels. LP/P variants identified in 14 patients were incorporated to the overall diagnostic yield but were not disclosed in this study as they represent new genes or founder variants that will be published in the future. In ten of these fourteen patients, the variant was observed in the same gene.

Comprehensive NGS panels provide high diagnostic yield in both pediatric and adult patients with suspected inherited DCM. Genetic etiology is relatively wide, and causative variants were observed in 45 genes. Majority had AD disease, but LP/P variants were also observed in genes associated with recessive, X-linked or mitochondrial diseases, and some had double diagnosis in dominant genes—all supporting the use of genetic panels.