|Home | About | Journals | Submit | Contact Us | Français|
Rare variants in >30 genes have been shown to cause idiopathic or familial dilated cardiomyopathy (IDC/FDC), but the frequency of genetic causation remains poorly understood. We have previously resequenced 9 genes in a cohort of IDC/FDC probands for rare variants, and now report resequencing results for 5 more genes with established relationships to DCM.
Blood samples were collected and DNA prepared from 312 patients, 181 with FDC and 131 with IDC. Genomic DNA underwent bidirectional sequencing and DNA of additional family members underwent analysis when a rare variant was identified. We identified rare variants in 34 probands (10.9% overall), including 29 unique protein-altering rare variants and 2 splicing variants that were absent in 246 control subjects (492 chromosomes). These included 12 myosin binding protein C (MYBPC3) variants in 13 (4.2%) probands, 8 alpha myosin heavy chain (MYH6) variants in 10 (3.2%) probands, six tropomyosin (TPM1) variants in six (1.9%) probands, four cardiac troponin C (TNNC1) variants in four (1.3%) probands, and one cardiac troponin I (TNNI3) variant in two (0.6%) probands. Variants were classified as likely or possibly disease-causing in 13 and 20 probands, respectively (33, 10.6% overall). One MYH6 variant was classified as unlikely to be disease-causing.
Rare variants in these five genes likely or possibly caused 10.6% of DCM in this cohort. When combined with our prior resequencing reports, approximately 27% of DCM probands have had possible or likely disease-causing variants identified.
Enormous progress has recently been made to understand the genetic basis of dilated cardiomyopathy (DCM). Most of this effort has been devoted to identifying genetic cause of DCM in families, termed familial dilated cardiomyopathy (FDC). While mutations in more than 30 genes have been described as causative of DCM,1, 2 determining the frequency of mutations in any one gene in a cohort of patients with DCM has been complicated by this dramatic locus heterogeneity. Also, identified mutations are usually unique (‘private’) to that family, and this allelic heterogeneity demands that, minimally, all coding exons and intron/exon junctions of any gene under study be examined to exclude variants relevant for disease. Furthermore, some of the genes that harbor DCM mutations are very large. Hence, examining coding sequence for variations in multiple genes has been costly and labor intensive.
Despite these experimental issues, gaining knowledge of the identities, character and frequencies of mutations in known DCM genes is essential to understand this phenotype that underlies a great deal of heart failure morbidity and mortality. Such knowledge is critical to the emerging practice of cardiovascular genetic medicine,3 as well as to the design of DCM genetic epidemiology studies.
To begin to address this problem, we recently resequenced exonic and intron/exon junctions of 6 DCM genes (MYH7, TNNT2, SCN5A, CSRP3, LDB3, and TCAP) in 313 DCM probands4 with support from the NHLBI’s Resequencing and Genotyping Service (RS&G) (Table 1).4–6 We found 32 variants (10.2%) not identified in 253 control DNAs.4 Even though the reference (wildtype) sequence has been established for these DCM genes, determining if a newly identified variant is disease-causing is challenging. In discovery studies, confirming that a candidate gene indeed harbors a causative mutation is usually resolved by a combination of a significant lod score, segregation of a variant with the disease phenotype in one or more large families, and/or definitive functional studies. However, such approaches, especially functional studies, exceeded the scope of our prior gene survey study and were not included in the initial publication (although some functional studies, e.g., for TNNT27 and LMNA8 have been published and others are underway). Hence, we developed conservative standards to classify these variants as ‘possibly’ or ‘likely’ disease causing.4
To extend these findings, we have now performed bidirectional sequencing of five additional genes in the same cohort, again with RS&G support. These five genes included myosin binding protein C (MYBPC3), α-myosin heavy chain (MYH6), tropomyosin 1 (TPM1), troponin C (TNNC1) and cardiac troponin I (TNNI3). Using the same approach, we present here the variants identified in this study thought to be possibly or likely disease-causing.
Written, informed consent was obtained from all subjects, and the Institutional Review Board at the Oregon Health & Science University approved the project. The study included 312 probands (311 subjects from the prior cohort4 and one additional DCM case), 290 Caucasians, of whom seven were of Hispanic descent; 16 African-Americans, three Asians and three Native Americans/Alaskan Natives, and used methods of clinical categorization of FDC versus IDC as previously described.9 These 312 probands included 298 of the 304 previously described9 in detail. Families with confirmed and probable histories of familial disease were classified as having FDC; those with histories consistent with possible FDC or a negative family history were classified as having IDC.4, 9 In cases classified as confirmed FDC, the patient and at least one closely related relative had IDC, defined as left ventricular enlargement (LVE) accompanied by systolic dysfunction upon exclusion of other detectable causes of DCM, as previously described.4, 9
The 5 genes reported herein were chosen following the same selection and analysis strategy described in our 6 gene resequencing report.4 Briefly, using the published FDC literature available in early 2005, we estimated the number of exons needed to be sequenced in order to find a single mutation in each gene, based upon the numbers of exons in that gene and an estimate of the frequency of that gene in DCM, in an effort to conserve resequencing resources at the NHLBI.4 We selected genes in that study with a threshold of sequencing up to 1000 exons to identify one putative disease-causing variant (as shown in Table 1 of that report).4 The prior estimates for TNNC110 (1500 exons), TNNI311 (1600 exons) 11 and TPM112 (1667 exons), were used for this study’s gene selection, as previously reported,4 as well as estimates from single studies for MYH613 (estimated at 907 exons), and MYBPC314 (estimated at 1591 exons).
Genomic DNA was extracted from whole blood as previously reported.4 Bidirectional sequencing was conducted for the following five genes: MYBPC3, myosin binding protein C, NM_000256.3, chr11:47,309,533-47,330,829, MYH6, α-myosin heavy chain, NM_002471.2, chr14:22,921,039-22,947,322, TPM1, tropomyosin 1, NM_001018004.1, NM_001018005.1, NM_001018006.1, NM_001018007.1, NM_001018008.1, NM_001018020.1, spanning chr15:61121891–61151166, TNNC1, troponin C, NM_003280.1, chr3:52,460,158-52,463,098, and TNNI3, cardiac troponin I, NM_000363.3, chr19:60,354,948-60,360,912. All exons and intron/exon boundaries were PCR amplified by standard methods at SeattleSNPs under contract with the NHLBI resequencing service.
Samples from probands identified by the resequencing service as carriers of protein-altering variants, as well as any available samples from their relatives were sequenced in our laboratory for confirmation and segregation analysis. Nucleotide changes were only evaluated if they were absent from all 246 (186 Caucasian, 23 Yoruban, 19 Asian and 18 Hispanic) control samples analyzed at the resequencing center. Any possibly disease-causing nucleotide alterations identified in African-American samples were further evaluated in an additional 167 control African-American DNA samples in our laboratory, for a total of 190 controls of African descent (380 chromosomes). As per our prior report,4 nucleotide changes were considered possibly disease-causing if they predicted a change in a conserved amino acid, a frameshift, a premature truncation, or a mis-splicing event and were absent in ethnically matched normal controls. Nucleotide changes were considered likely disease-causing if they met the above criteria and also segregated with disease in multiple affected individuals within a family or were identified in multiple unrelated probands, or had previously been reported as causative of DCM. Nucleotide changes that did not segregate with disease were considered unlikely to be associated with IDC or FDC.
Haplotype analysis was performed with the program Cocaphase15 for variants identified in multiple subjects to rule out founder effects. All individuals with the mutations in question were of Caucasian origin, thus haplotype estimation was carried out on Caucasian samples only. Markers at each gene locus with minor allele frequencies >0.1 in the whole sample were selected for analysis.
The bidirectional sequencing of five genes known to cause DCM was completed for DNA specimens from 312 unrelated probands with IDC or FDC. Protein-altering variants, none of which were present in 246 control specimens, were identified in 34 of 312 probands (10.9%) (Table 2). Most unique variants were missense mutations (29/31; 94%) and altered highly conserved amino acids; two were predicted to affect splicing. Some had previously been reported with DCM or hypertrophic cardiomyopathy (HCM) (Table 2).16–21
Of the five genes examined, the most rare variants were identified in the gene encoding myosin binding protein C (MYBPC3), with variants identified in 13 of the 312 probands (4.2%). All MYBPC3 variants were considered possibly or likely disease-causing (Table 2). One mutation (Ala833Thr; A.8), which was observed to segregate with disease, was also found in two additional families in our cohort (A.7 and A.9), none of whom were known to be related and who were geographically remote to one another; haplotype sharing was identified near to the mutation, suggesting that this may be a founder mutation. This same variant had been previously reported in an individual with familial HCM.19 In that report, the variant was identified in the proband with HCM as well as in his brother and father, both who had mild cardiac hypertrophy. Two probands (A.11 and A.12) had novel, possibly disease-causing splicing variants. Tissue samples were not available from either proband to assess the impact of these variants on mRNA splicing. One of the subjects with a splicing variant (A.12) also carried a MYBPC3 Gly1260Asp variant in trans (A.13). DNA from relatives was not available in either case with splicing variants and therefore segregation could not be assessed. The Gly5Arg variant (A.1) was previously reported in a patient with early onset HCM harboring this and another MYBPC3 variant;16 however, segregation was not assessed in that report. DNA from other affected relatives was not available to assess segregation, and we therefore classified the Gly5Arg variant as possibly disease-causing. The Arg272Cys variant (A.3) was considered possibly disease-causing due to lack of segregation data. The Gly490Arg variant (A.4), which has been reported with HCM,16, 18 was considered possibly disease causing, as segregation with DCM could not be assessed.
Eight unique α-myosin heavy chain (MYH6) variants were identified in 10 probands (Table 2). Six of the probands carried three likely disease-causing variants (Ala1004Ser, Ala1440Pro and Asp1826Asn). The Ala1004Ser variant (B.3, B.4 and B.5) occurred at a highly conserved site, was previously reported in a sporadic DCM case20 and was therefore considered likely disease-causing. The likely disease-causing Ala1440Pro (B.7) occurred at a highly conserved amino acid and segregated with DCM. The Asp1826Asn variant (B.9 and B.10), caused by two G>A substitutions in cis at positions 25743 and 25744, were identified in two probands. The affected sister of one of these probands was found to carry the same alteration, and thus the variant (B.9) was classified as likely disease-causing. The second proband, who also carries the Asp1826Asn variant (B.10), has sporadic DCM, and was also reported to have an in-frame G474_D475insQ LMNA variant (family R in Parks et al.)5 and a TCAP Pro141Ala variant (denoted as D.3 and categorized as unlikely to be disease causing in Hershberger 2008)4. Two MYH6 variants (B.1 and B.8) were identified in the same individual, a female diagnosed with DCM at 8 years old who was previously reported as carrying the disease-causing TNNT2 Lys210del mutation (B.7 in Hershberger 2009)7. The remaining MYH6 nucleotide variants were considered possibly disease-causing (B.2, B.6), or unlikely to be disease-causing because of lack of segregation (B.11).
We screened the exons from seven different transcripts at the TPM1 locus, encoding the tropomyosin 1 gene. Six different novel protein-altering variants were identified among six probands (Table 1), and all of the nucleotide changes were considered possibly or likely disease-causing. With the exception of Ser16Ile, all of the TPM1 variants were located in exons found to be present in the TPM1κ isoform22, which unlike other TPM1 isoforms has been reported to be expressed uniquely in cardiac tissue. Two novel variants, Lys15Asn and Ile92Thr, (C.1 and C.4) segregated with DCM in the respective families of each proband and were considered likely disease-causing. The Ala277Val variant (C.6), was identified in a severely affected IDC patient requiring transplant at age 13. This proband was previously reported as carrying a likely disease-causing Glu244Asp TNNT2 variant shown to cause decreased calcium sensitivity.7 Segregation could not be assessed in this family; thus the Ala277Val TPM1 variant was considered possibly disease-causing.
Four private protein-altering variants were identified in TNNC1 encoding troponin C, in four probands. One of the variants identified (D.2) occurred at a conserved site, segregated with disease and was considered likely disease-causing. The Tyr5His variant (D.1) occurred in an individual with early onset DCM known to carry a possibly disease-causing Arg1045Cys variant in MYH7.4 Segregation could not be assessed and therefore the Tyr5His variant has been categorized as possibly disease-causing. The individual with variant D.3 (also found to carry MYBPC3 Pro910Thr denoted as A.10), carries a TNNC1 Asp145Glu variant previously reported in a male with a family history of HCM. This variant was shown to increase Ca2+ sensitivity of force recovery.21 The Asp145Glu variant is novel for DCM and segregation data is lacking for this individual; therefore, we categorized this case as possibly disease-causing.
The Asp180Gly nucleotide alteration was identified in TNNI3, encoding cardiac troponin I, in two unrelated probands (variants denoted E.1 and E.2). The variant was predicted to change a highly conserved amino acid; however, no family DNA specimens were available in either case, and the variant was therefore considered possibly disease-causing.
Twenty-one of 181 (11.6%) probands categorized as having FDC and 12 of 131 (9.2%) probands categorized as having IDC carried possibly or likely disease-causing variants.
Haplotype analysis was consistent with a possible founder effect in the shared mutations for the MYBPC3 Ala833Thr and MYH6 Asp1826Asn variants, where haplotype sharing was present. However, the high degree of linkage disequilibrium between the observed variations at the MYBPC3 locus makes it difficult with this limited dataset to decipher if this is a potential founder effect or a chance finding.
This study, when combined with our prior reports,4–6 provides the most extensive resequencing data to date in one of the largest and most well characterized cohorts of probands with DCM.9 We used Sanger-based sequencing, still the gold-standard for sequencing sensitivity and specificity, and have now established approximately 27% of putative genetic cause within this cohort of DCM patients: 10.6% from the current study of five genes (MYBPC3, MYH6, TPM1, TNNC1, TNNI3), 10.2% from the prior resequencing study of six other genes4 (MYH7, TNNT2, SCN5A, CSRP3, LDB3, and TCAP) (Table 1), 5.9% from a study of the LMNA gene5 encoding lamins A and C, and 1.0% from a resequencing study of PSEN1 and PSEN2,6 encoding presenilins 1 and 2. Approximately 3% of these probands have been found to carry multiple variants in the same or different DCM genes.
The genetics of DCM vary significantly from that of other genetic cardiomyopathies, most strikingly with >30 genes reported to be involved, but with each accounting for only a small fraction of our cohort (0.3% – 5.9%), as shown in this and our prior resequencing studies (Table 1). In contrast, mutations in 2 genes, MYH7 and MYBPC3, collectively account for approximately 40–45% of HCM (or 80–90% of detectable genetic cause when a genetic cause can be identified, which has estimated to occur in 42%23 to 65%24 of HCM cases). Similarly, 3 genes, PKP2 (plakophilin 2), DSP (desmoplakin), and DSG2 (desmoglein 2) account for 40–50% of arrhythmogenic right ventricular dysplasia/cardiomyopathy (ARVD/C).25 Further, while HCM and ARVD/C are largely genetic disorders of genes encoding sarcomeric or desmosomal proteins, respectively, genes encoding proteins of exceedingly diverse function have been implicated to cause genetic DCM. For example, LMNA, to date the most frequent genetic contributor of DCM, encodes a structural protein of the inner nuclear membrane with as yet undefined disease pathways. Other genes implicated to cause DCM range from sarcomeric proteins, transcription factors, Z-disc proteins, channel proteins (including Na+, Ca++, and K+), and most recently a protein of the RNA spliceosome.26
These findings have several implications. One key insight is that despite the large number and variety of genes involved, a unitary, homogenous DCM phenotype results. Careful phenotyping by our group and others has distinguished ‘DCM with prominent conduction system disease’ as the only sub-phenotype of non-syndromic DCM, observed principally with mutations in LMNA and SCN5A in a small fraction of cases.1, 2 Very few other Mendelian conditions show this degree of locus heterogeneity while manifesting one homogeneous phenotype. Examples include retinitis pigmentosa or hereditary deafness, encompassing >40 genes each.27
Furthermore, how this homogeneous DCM phenotype occurs from such a broad array of genetic cause remains largely unexplained. While our colleagues have suggested a ‘final common pathway’ for DCM,28 we suggest that the DCM phenotype results from several different genetic injury pathways. This multi-locus pathway hypothesis resulting in a final DCM phenotype will need to have the gene network injury pathways and their respective injury mechanisms elucidated to lay this important question to rest.
Our nomenclature of likely or possibly disease-causing implies that the variant alone causes or could cause disease, reflecting common usage in rare-variant classical Mendelian disease. An alternative descriptor would be disease-associated, a term more commonly used to describe risk alleles in genome association studies aimed to determine genetic cause in multifactorial, polygenic disease. For a genetically heterogeneous Mendelian disease such as DCM, we favor the disease-causing nomenclature even though we recognize that our emerging data suggests that two (or more) nonsynonymous rare variants appear to be causative of DCM, suggesting a more complex genetic model in some cases. Also, even though we sequenced 246 control DNAs (492 chromosomes), it is possible that some of the variants reported herein as possibly disease-causing may still be rare benign or neutral mutations that do not directly account for disease. For that reason we have taken a conservative approach to categorizing these variants as ‘possibly’ disease-causing. Conversely, we also recognize that variants excluded from this analysis because they were identified at low frequency in control DNAs could still be relevant for DCM pathogenesis. Further, some variants had been previously identified as disease-causing HCM mutations that were identified in this study in probands with DCM. Considerable locus and allelic heterogeneity of sarcomeric genes for DCM and HCM has been reported previously, (see29 for review). For example, one mutation in TNNT2 identified in our prior study in a DCM proband and previously reported in HCM4 had a mixed picture at functional studies7. Another report showed DCM, HCM and restrictive phenotypes with the same sarcomeric variant in one large family, although the mechanisms, whether genetic or environmental, of this phenotypic variability remain to be discovered.30
Assessing the significance of a novel nonsynonymous rare variant in a patient with DCM but no affected family members presents special challenges. This is because the most powerful clinical genetics finding to establish that a specific rare variant is disease-causing is the segregation of that variant with the disease phenotype in multiple affected family members (and ideally, in multiple families), accompanied by the absence of the variant in multiple unaffected family members. This is a stringent but critical standard for resequencing data to be translated into clinical care. Determining which of the variants reported herein might be useful for genetic counseling of probands and/or family members regarding DCM risk is particularly challenging. While specific criteria for diagnostic molecular genetics laboratories vary, most diagnostic standards would follow the general outline of our approach to categorize variants: rare, nonsynonymous variants identified in genes established to have mutations that cause DCM would be considered at least as possibly disease-causing, and in the absence of segregation, prior reports, or other supporting data, predictive testing of family members would usually not be recommended. However, if one of these possibly disease-causing variants reported herein would be observed in additional DCM probands undergoing diagnostic testing, in most cases the variants would be assigned as likely disease-causing. Functional studies of the variant in gene targeting experiments in animals or in heterologous cell systems could also increase the evidence that a rare variant is disease-causing, but such approaches have not been conducted by diagnostic molecular genetics laboratories for DCM-associated genes. Hence, all of these issues will need to be resolved with research-based resequencing studies of much larger cohorts of DCM patients and their family members, regardless of whether they appear to have familial or sporadic DCM. These studies should also include longitudinal follow up, accompanied by large numbers of reference DNAs and combined with functional studies.
As in our prior study,4 the fraction of likely and possibly disease-causing mutations was similar for probands whether categorized as FDC or IDC. This finding, although preliminary in nature, again suggests that a significant proportion of what has been categorized as idiopathic DCM may rather have a rare variant genetic basis. If replicated in much larger cohorts of patients with IDC, that is, apparently sporadic DCM, this finding would impact our approach to determining DCM etiology in a fundamental way, bringing diagnostic molecular genetics to the clinic for thousands of patients with DCM of unknown cause.3
Family data was not available in 19 of the 34 cases so we were unable to assess segregation of the variant with disease. We have not sequenced two very large DCM genes (titin and dystrophin) in order to apply our resequencing resources to greater numbers of smaller genes; however, dystrophin (DMD) resides on the X chromosome and family history information in our DCM cohort supports autosomal inheritance in most cases.9 We have sequenced only a fraction of DCM genes. However, based on the available literature, we selected genes for study thought more likely to be represented in genetic DCM. Less expensive screening methods (D-HPLC, SSCP) that might have screened greater numbers of genes with equivalent resources are less sensitive and may have missed gene variation relevant to DCM that could be identified with Sanger-based sequencing methods. We also note that we only examined coding sequences and intron/exon boundaries for variation, and hence additional genetic variation including copy number variants, or variation in regulatory areas (promoters, 5’ and 3’ untranslated regions) or introns of these five genes would not have been detected. The difficulties of assessing variants in sporadic DCM have been noted above.
The nonsynonymous rare variants identified in these five genes implicated in FDC and IDC account for only a small fraction of the underlying genetic cause. Taken with prior data from this cohort and others, the genetic landscape of DCM is distinctive from that of other known genetic cardiomyopathies. Future studies in larger cohorts, ideally with family members screened for disease, with many genes sequenced and accompanied by functional studies, will be needed to confirm and extend these findings.
We thank the many families and referring physicians for their participation in the Familial Dilated Cardiomyopathy Research Project, without whom these studies would not have been possible.
This work was supported by NIH awards RO1-HL58626 (Dr Hershberger) and 5 M01 RR000334. Resequencing services were provided by the University of Washington, Department of Genome Sciences, under U.S. Federal Government contract number N01-HV-48194 from the National Heart, Lung, and Blood Institute.
Journal Subject Codes: Genetics of cardiovascular disease; clinical genetics; myocardial cardiomyopathy disease
Conflict of Interest Disclosure: No conflicts.