Variant Interpretation for Dilated Cardiomyopathy (DCM): Refinement of the ACMG/ClinGen Guidelines for the DCM Precision Medicine Study

Abstract

Background -

The hypothesis of the Dilated Cardiomyopathy (DCM) Precision Medicine Study is that most DCM has a genetic basis. The study returns results to probands, and when indicated, to relatives. While both the American College of Medical Genetics and Genomics/Association for Molecular Pathology and ClinGen’s MYH7-cardiomyopathy specifications provide relevant guidance for variant interpretation, further gene and disease specific considerations were required for DCM. To this end, we tailored the ClinGen MYH7-cardiomyopathy variant interpretation framework; the specifications implemented for the study are presented here.

Methods -

Modifications were created and approved by an external Variant Adjudication Oversight Committee. After a pilot using 81 probands, further adjustments were made, resulting in 27 criteria (9 modifications of the ClinGen MYH7 framework and reintroduction of two ACMG/AMP criteria that were deemed not applicable by the ClinGen MYH7 working group).

Results -

These criteria were applied to 2,059 variants in a test set of 97 probands. Variants were classified as benign (n=1,702), likely benign (n=33), uncertain significance (n=71), likely pathogenic (LP; n=12), and pathogenic (P; n=3). Only 2/15 LP/P variants were identified in Non-Hispanic African ancestry probands.

Conclusions –

We tailored the ClinGen MYH7 criteria for our study. Our preliminary data show that 15/97 (15.5%) probands have LP/P variants, most of which were identified in probands of Non-Hispanic European ancestry. We anticipate continued evolution of our approach, one that will be informed by new insights on variant interpretation and a greater understanding of the genetic architecture of DCM.

Clinical Trial Registration -

clinicaltrials.gov; NCT03037632

Introduction

The Dilated Cardiomyopathy (DCM) Precision Medicine Study was launched in 2015 to test the hypothesis that both familial and non-familial DCM have a genetic basis. For this study, DCM is defined as non-ischemic disease and idiopathic, after excluding non-genetic etiologies.¹ The study, which aims to recruit 1,300 individuals with DCM (600 non-Hispanic ethnicity [NHE] African Ancestry, 600 NHE-European Ancestry, and 100 HE; 50% female) and 2,600 relatives, includes a randomized controlled trial to evaluate the effectiveness of an educational guide (Family Heart Talk) on the uptake of and adherence to cardiovascular screening in at-risk relatives. Families are recruited at DCM Consortium sites, by telephone, and mail.²

After exome sequencing in probands and affected relatives, results from analysis of 35 DCM genes (see Supplemental Material) are returned to probands and relatives of probands who are found to carry likely pathogenic or pathogenic (LP/P) variants. As part of the study, periodic screening is recommended for unaffected relatives with LP/P variants. Unaffected relatives who do not carry a known LP or P variant identified in the proband are discharged from surveillance. For at-risk relatives of probands in whom a LP/P variant is not identified, surveillance is recommended. These screening recommendations follow Heart Failure Society of America and the American College of Medical Genetics and Genomics (ACMG)^{3, 4} guidelines and depend on accurate variant interpretation.

The ability to conduct large scale molecular genetic testing has resulted in an increased volume of sequence variants, and, consequently, new challenges in interpretation. This prompted ACMG to revise prior guidance,⁵ resulting in updated standards, jointly with the Association for Molecular Pathology (AMP).⁶ Recognizing that the ACMG/AMP standards serve a broad spectrum of single-gene conditions, the ClinGen⁷ Cardiovascular Domain Working Group (CDWG) tailored ACMG/AMP guidance for cardiovascular disease.⁸ Incorporating the genetic features of cardiomyopathy (phenotypic, locus and allelic heterogeneity, reduced and adult onset penetrance, multiple variants interacting to cause disease), the ClinGen Inherited Cardiomyopathy Expert Panel (CMP-EP), which included two members of this author group (A.M., R.E.H.), published specifications for hypertrophic and dilated cardiomyopathies focused on MYH7.⁹ The MYH7 specifications provided a foundation for the variant interpretation criteria needed to conduct our study; however, further specifications were required to incorporate considerations that are unique to DCM. To this end, we tailored the ClinGen MYH7 variant interpretation framework. The resulting criteria, along with preliminary data, are presented here.

Methods

Please see the Supplemental Material for complete methods. The relevant variant curation criteria are summarized (Table 1). The non-pilot variant data used for this analysis, excluding 1,500 unique variants unlikely to change protein function based on predicted Sequence Ontology¹⁰ impacts, are provided in the Supplemental Material. Source code and complete variant and adjudication data are available upon request from the authors. All research participants provided signed informed consent. Institutional Review Board approval was obtained for this study, initially locally at each of the participating clinical sites of the DCM Consortium, and later with study oversight provided by a single Institutional Review Board at the University of Pennsylvania.

Table 1.

DCM Precision Medicine Study Variant Interpretation Criteria

Strength	Specification type	Rule abbreviation	Rule description
Pathogenic criteria
Very strong	Reintroduced ACMG criterion for DCM gene specification	PVS1	Null variant in LMNA or SCN5A
Strong	none	PS1	Same amino acid change as a previously established pathogenic variant
	none	PS2	De novo (paternity confirmed) in a patient with disease and no family history
	none	PS3	Well-established in vitro or in vivo functional studies supportive of a damaging effect
	none	PS4	Prevalence of variant in affected individuals is significantly increased compared to controls OR variant identified in ≥15 probands with consistent phenotypes OR variant identified in ≥10 confirmed unlreated probands with consistent phenotypes
	none	PP1_Strong	Variant segregates in ≥7 meioses in DCM genes with established evidence
	Strength modification, DCM gene specification	PVS1_Strong	Null variant in FLNC, BAG3,and TTN A-band
Moderate	DCM gene specification	PM1	Located in hotspot (For DCM, RBM20 exon 9, amino acids 634, 636, 637, 638)
	DCM gene specification, study specific	PM2	Absent from gnomAD or at extremely low frequency in all gnomAD non-founder populations with available data
	Reintroduced ACMG criterion, DCM gene specification	PM3	For recessive disorders, detected in transwith a pathogenic variant (rarely observed with TNNI3 and TNNT2 variants)
	none	PM4	In-frame deletions/insertions of any size in a non-repeat region or stop-loss variants
	none	PM5	Different missense change at same amino acid residue previously established as pathogenic
	DCM specific	PM6	De novo without confirmation of paternity (parental clinical data required)
	DCM gene specification	PVS1_Moderate	Null variant in gene with evidence supporting loss of function as disease mechanism (VCL, PLN, DSP)
	none	PS4_Moderate	Variant identified in ≥6 probands with consistent phenotypes
	none	PP1_Moderate	Variant segregates in ≥5 meioses in DCM genes with established evidence
Supporting	none	PP1	Variant segregates in ≥3 meioses in DCM genes with established evidence
	Method specification	PP3	Computational evidence support a deleterious effect (REVEL score >0.7)
	none	PS4_Supporting	Variant identified in ≥2 probands with consistent phenotypes
Benign
Stand alone	none	BA1	Variant allele frequency is >0.1% in any gnomaAD non-founder population
	DCM gene specification	BP1_Stand_Alone	Missense variant in TTN
Strong	DCM gene specification, study specific	BS1	Variant allele frequency is >0.05% in any gnomAD non-founder population
	none	BS3	Well established in vitroor in vivo functional studies show no damaging effect
	none	BS4	Non-segregation in affected members of a family
Supporting	none	BP2	Observed as a compound heterozygote (in trans) or double heterozygote in genes with overlapping function
	Method specification	BP4	Multiple lines of computational evidence suggest no impact on gene or gene product (REVEL score <0.15)
	Study specific	BP7	Unlikely to affect protein function based on calculated Sequence Ontology terms for the transcript(s) of interest.

Results

Pilot Study

A pilot study was conducted with the first 81 probands enrolled in the study. Of these, 40 (49%) carried only likely benign (LB) or benign (B) variants. The rest carried VUS (n=30, 37%) and LP variants (n=11, 14%). Three challenges were identified from this exercise: limited gene-specifications for predicted loss-of-function (LOF) variants, evaluation of a variant’s absence or rarity in reference populations, and use of computational data. Further adjustments, as described below, were approved by the Variant Adjudication Oversight Committee (VAOC; Chair, GJ; Members, W.B, M.D., J.G-F, H.R).

To address the first challenge, a very strong criterion for pathogenicity for LOF variants in LMNA and SCN5A (PVS1) and a strong criterion for pathogenicity for LOF variants in FLNC, BAG3, and TTN A-band (PVS1_Strong) were added. The moderate criterion (PVS1_Moderate) was applied to LOF variants in VCL, PLN, and DSP. These modified criteria were applied considering available evidence associated with DCM. LOF has been established as a disease mechanism for LMNA-DCM and other LMNA-associated diseases.¹¹ Supporting this are large multicenter studies demonstrating a substantially increased odd ratio (OR) in DCM cases versus controls (OR = 99.7).¹² Furthermore, in the latest release of gnomAD (v2.1.1), the LMNA pLI (probability of being loss-of-function intolerant) was established to be 1.00. For SCN5A, LOF variants are pathogenic for Brugada syndrome.^{13, 14} Also, a large multicenter study demonstrated a significantly increased odds ratio of 16.5 for LOF variants in SCN5A in DCM cases versus controls.¹² In addition, the p.Arg219His variant, identified in DCM and arrhythmias,¹⁵ was studied in Xenopus oocytes, leading to a proton leak but not sodium current alterations. While no segregation has been demonstrated for LOF in primary DCM, supporting segregation data has been published in a family with sick sinus syndrome, and studies in the human cell line tsA-201 with heterologously expressed mutant sodium channels showed LOF properties of reduced or no sodium current density in conjunction with gating modulations.¹⁶ For this gene, the gnomAD pLI score is 0.91. We also note that when a truncating SCN5A variant is identified in a proband with DCM a search for an arrhythmia phenotype is also warranted.

LOF variants in FLNC, BAG3, and the A-band of TTN were assigned strong criteria for pathogenicity. First, for FLNC, a LOF mechanism has been reported¹⁷ and the gnomAD pLI = 1.00. Due to the relative gene size of FLNC and LMNA, the pLI calculation appears to be reliable. However, in smaller genes, where variants are less prevalent in the DCM population although absent from controls, the pLI may not reflect the actual contribution of LOF variants to disease. For BAG3, a gene comprising 4 coding exons, despite LOF been known as a disease mechanism,^{18, 19} the gnomAD pLI = 0.62. In the case of TTN, the coding sequence leads to the generation of alternative splicing isoforms,²⁰ with the heart expressing two major isoforms (N2B from transcript NM_003319.4 and N2BA from transcript NM_001256850.1) that incorporate the Z-line, I-band, A-band, and M-line.²⁰ The N2B and N2BA isoforms diverge mainly for the incorporation in the long cardiac isoform N2BA of a large portion of I-band coding exons, similar to the skeletal muscle isoform N2A derived from transcript NM_133378.4.²⁰ It is known that LOF variants affecting the TTN A-band (defined as chr2:179483218–179400709, NM_001267550.1, GRCh37/hg19) are overrepresented in DCM,^{21, 22} establishing LOF as a mechanism of disease.¹² However, due its large size, potential LOF variants are differently distributed across the various functional domains between DCM cases and controls.²³ The high prevalence in the control population of LOF variants in TTN regions other than the A-band led to a calculated gnomAD pLI = 0.00; however, the calculation does not discriminate between the various transcripts and protein domains and thus does not accurately reflect the deleterious role of LOF variants in the A-band.²⁴

LOF variants in VCL, PLN, and DSP were assigned a moderate degree of pathogenicity, although for a DSP variant in this category, the variant must have been observed in at least one unrelated DCM case. We implemented this criterion for VCL and DSP LOF variants based on their enrichment in DCM cases versus controls (OR_VCL = 21.3; OR_DSP = 41.0).¹² In addition, the constraint metrics for DSP support a LOF mechanism (pLI = 1.00). For PLN, a trend for enrichment of LOF variants was observed in DCM cases versus controls (OR_PLN = 13.8).¹² In addition, PLN LOF variants have been shown to be pathogenic and described in both DCM and HCM).^25–29 For both PLN and VCL, the constraint metrics from gnomAD are below threshold (PLN, pLI = 0.45; VCL pLI = 0.05), however, the PLN coding sequence has only one exon, possibly being too small a target to reliably detect population levels of LOF variants. Haploinsufficiency of VCL, although enriched in DCM, may represent a low penetrant event. However, we suggest that additional data are needed to confirm our preliminary VCL observations for LOF.

When evaluating PM2, Wilson’s score confidence intervals could not be calculated if there were no genotype calls at the variant position in both exomes and genomes, which could occur due to either the absence of the variant in all populations or lack of coverage. To allow using PM2 in the former scenario, the criterion was adjusted so it was met if the allele count was sufficient (120,000) in neighboring nucleotides, coverage was sufficient, and a mix of exome and genome data was available in gnomAD.

Evaluation of computational data (PP3 and BP4) was performed manually for this pilot using a holistic approach. To harmonize computational data evaluation, a method based on the REVEL score³⁰ was introduced. Cutoffs were developed for this study on the basis of the publication (which estimated the sensitivity and specificity of each potential cutoff using out-of-bag predictions based on 6,182 HGMD disease mutations and 123,706 rare exome sequence variants)³⁰ and made these data publicly available (https://sites.google.com/site/revelgenomics/). On the basis of these data, variants meeting BP4 must have a REVEL score below 0.15, a cutoff estimated to falsely predict no deleterious effect for ~5% of pathogenic variants while correctly identifying ~55% of neutral variants (Figure 1). Variants meeting PP3 must have a REVEL score greater than 0.7, which would correctly identify ~58% of pathogenic variants and falsely predict a deleterious effect for only ~5% of neutral variants. When comparing calls made before REVEL implementation to those that would have been made after, we found that 5/6 variants below the 0.15 cutoff had been manually classified as meeting BP4. Among the 13 variants that had a REVEL score greater than 0.7, 12 had been previously classified as meeting PP3. Of 26 variants with a REVEL score between 0.15 and 0.7, four had been manually classified as meeting BP4, 14 met PP3, and eight met neither. Although variants with a score between 0.15 and 0.7 would not receive an automatic PP3 or BP4 evaluation using the REVEL method, manual review of the data and clinical judgment can be applied if application of the PP3 or BP4 criteria would lead to a clinically significant difference in variant classification.

Figure 1.

Automated REVEL score Utilization. Cutoffs for BP4/PP3 were based on the estimated sensitivity and specificity reported in Figure S1 of the original publication reproduced here. Comparison of holistic evaluations of variants in the pilot phase with automatic evaluations for the same variants based on the selected REVEL score cutoffs are shown above the blue line.

Results Using DCM Variant Interpretation Criteria Adaptations

Our modifications resulted in 27 criteria for variant interpretation in DCM as summarized (Table 1); full documentation is also provided (Supplemental Material). Using this framework, curation was performed between July 2018-April 2019 for 2,059 variants in an additional set of 97 probands (Supplemental Material). Variants were classified as B (n=1,702), LB (n=33), VUS (n=71), LP (n=12), and P (n=3). 216 (11%) were not adjudicated (low impact variants for which BP7 was met); 22 (1%) were classified as Low Quality.

Among the 97 probands, P variants were identified in 3%, LP in 12%, and VUS in 46%, (Table 2). Most LP/P variants were in TTN (n=8; Table 3). Variants reached LP/P classifications based on case (n=1, PS4; n=3, PS4_Moderate; n=3, PS4_Supporting) and segregation data (n=2, PP1_Strong; n=2, PP1), allele frequency (n=15, PM2), predicted loss of function (n=1, PVS1; n=11, PVS1_Strong), protein length change (n=1, PM4), hotspot location (n=1, PM1), and computational evidence (n=2, PP3). LP/P variant types included stop gained (n=6/15, 40%), frameshift (n=6/15, 40%), in-frame deletion (n=1/15, 7%), and missense (n=2/15, 13%). Of the 15 LP/P variants, 13 were found in individuals of non-Hispanic European ancestry. More than one VUS and above variant were identified in 20 (21%) probands; however, combinations including more than one LP/P variant were not observed in this preliminary sample.

Table 2.

Most Severe Variant Among 97 Probands with DCM, by Ethnicity and Ancestry

	Benign/Likely Benign	Uncertain Signficance	Likely Pathogenic	Pathogenic	Total
	N (Row %)
Non-Hispanic, White	14 (29.2%)	21 (43.8%)	10 (20.8%)	3 (6.3%)	48
Hispanic, White	0 (0%)	2 (100%)	0 (0%)	0 (0%)	2
Non-Hispanic, African Ancestry	23 (48.9%)	22 (46.8%)	2 (4.3%)	0 (0%)	47
Hispanic, African Ancestry	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0
TOTAL	37 (38.1%)	45 (46.4%)	12 (12.4%)	3 (3.1%)	97

Row percentages may not add to exactly 100% due to rounding

Table 3.

Variant Details for Likely Pathogenic and Pathogenic Variants, by Ethnicity and Ancestry

Adjudication	Ethnicity, Ancestry	Gene	HGVSc	HGVSp	Impacts
Pathogenic	Non-Hispanic, White	LMNA	NM_170707.3:c.673C>T	NP_733821.1:p.Arg225Ter	Stop Gained
		TNNT2	NM_000364.3:c.650_652delAGA NM_001001430.2:c.629_631delAGA	NP_000355.2:p.Lys217del NP_001001430.1:p.Lys210del	Inframe Deletion
		TTN	NM_001267550.1:c.61876C>T NM_133378.4:c.54172C>T	NP_001254479.1:p.Arg20626Ter NP_596869.4:p.Arg18058Ter	Stop Gained
Likely Pathogenic	Non-Hispanic, White	BAG3	NM_004281.3:c.1417C>T	NP_004272.2:p.Arg473Ter	Stop Gained
		FLNC	NM_001458.4:c.4926_4927insACGTCACA	NP_001449.3:p.Val1643ThrfsTer26	Frameshift Variant
		FLNC	NM_001458.4:c.1948C>T	NP_001449.3:p.Arg650Ter	Stop Gained
		LMNA	NM_170707.3:c.949G>A	NP_733821.1:p.Glu317Lys	Missense Variant
		RBM20	NM_001134363.1:c.1906C>T	NP_001127835.1:p.Arg636Cys	Missense Variant
		TTN	NM_001267550.1:c.90844delA NM_133378.4:c.83140delA	NP_001254479.1:p.Thr30282ArgfsTer8 NP_596869.4:p.Thr27714ArgfsTer8	Frameshift Variant
		TTN	NM_001267550.1:c.88703_88704delAC NM_133378.4:c.80999_81000delAC	NP_001254479.1:p.His29568LeufsTer7 NP_596869.4:p.His27000LeufsTer7	Frameshift Variant
		TTN	NM_001267550.1:c.66804_66807delGAAG NM_133378.4:c.59100_59103delGAAG	NP_001254479.1:p.Lys22269HisfsTer10 NP_596869.4:p.Lys19701HisfsTer10	Frameshift Variant
		TTN	NM_001267550.1:c.64915C>T NM_133378.4:c.57211C>T	NP_001254479.1:p.Arg21639Ter NP_596869.4:p.Arg19071Ter	Stop Gained
		TTN	NM_001267550.1:c.71307_71310dupTGAC NM_133378.4:c.63603_63606dupTGAC	NP_001254479.1:p.Ser23771Ter NP_596869.4:p.Ser21203Ter	Frameshift Variant
	Non-Hispanic, African Ancestry	TTN	NM_001267550.1:c.83515C>T NM_133378.4:c.75811C>T	NP_001254479.1:p.Arg27839Ter NP_596869.4:p.Arg25271Ter	Stop Gained
		TTN	NM_001267550.1:c.66931_66932delAA NM_133378.4:c.59227_59228delAA	NP_001254479.1:p.Lys22311ValfsTer3 NP_596869.4:p.Lys19743ValfsTer3	Frameshift Variant

Discussion

We tailored the ACMG and ClinGen MYH7 variant interpretation criteria to DCM. Previous publications report a ~27% detection rate for LP/P variants in DCM,³¹ however, the detection rate for LP/P variants using our criteria so far has only been ~16%; a significant proportion of these variants are predicted LOF. Significant due diligence was done when applying PVS1 rules for this study. Variants affecting the initiation codon are not automatically assumed to cause haploinsufficiency due to the possible rescue mechanism of a downstream methionine. Also, variants predicted to lead to premature termination codon or predicted to affect mRNA stability occurring in the last exon and last 55nt of the second-last exon possibly escaping nonsense mediated decay were excluded from PVS1 application. Similarly, invariant splice site variants would be considered as leading to a null variant if a new in-frame splice junction is not predicted and the variant affects all biologically functional transcripts known to be active in the heart and altered in DCM. These principles are consistent with published guidance for interpretation of LOF variants.³²

Compared to a decade ago, concluding pathogenicity in Mendelian disorders requires a higher level of evidence.³³ This approach strengthens the degree of certainty when reporting variants. While our approach adds gene-specificity, a limitation not unique to this study is that variant adjudication is a probabilistic process. Achieving precision medicine in DCM requires large data sets and functional studies that demonstrate pathogenicity. Further evaluation of these criteria by an independent group, such as the ClinGen CMP-EP, could serve the community by generating international consensus on variant interpretation for DCM.

With continued use of our criteria and an ever-expanding literature body on variant interpretation, cases will emerge that provide opportunities to refine this framework. For example, we recognize that restricting pathogenicity to the A-band of TTN excludes potentially relevant regions that can be identified with more granular analyses.^{34, 35} We have the infrastructure to support ongoing, thoughtful reanalysis and refinement of our criteria. At study completion, a re-evaluation will be undertaken, incorporating best practices and any new data. In addition, there is potential for reassessment of low-quality variants and those that were not adjudicated. Variant reclassifications may result from this process, and, following published principles for researchers returning results,³⁶ should this occur while the study is active, our protocol has provisions for re-contacting participants.

DCM genetic testing results almost always impact families. In probands, LP/P results may directly affect treatment. For example, DCM patients with LP/P variants in arrhythmogenic genes (e.g., LMNA, FLNC) should have an implantable cardioverter defibrillator placed earlier.³ Negative testing results neither exclude the possibility that the proband’s condition has a genetic etiology, nor relieves relatives from risk.

Variant reclassifications also have an impact. The highest-impact reclassifications involve downgrading of LP/P variants. In this scenario, where at-risk relatives would have been offered cascade testing, those found not to carry the variant would have been informed in error that no additional cardiac surveillance is recommended. We note, however, that in the case of LP variants a 10% error rate is assumed,⁶ participants in this category are informed that no additional surveillance is needed beyond what would routinely be recommended for the general population. The counseling conversation also includes discussion of symptoms that should be reported to their physician. Despite these caveats, relatives who do not carry a familial LP or P variant may be discharged from surveillance. When instead of downgraded, variants are upgraded to LP/P, cascade testing is made possible, without which at-risk relatives would have had only the option of cardiac screening. Reclassifications involving changes to and from LB/B and VUS have less clinical impact: cardiac screening should continue in at-risk relatives in every case.^{3, 4}

Although we prefer a conservative approach for return of results in this study, understanding the genetic architecture of DCM will require substantial further research. Indeed, a proportion of DCM may be oligogenic,³⁷ but the proportion following such a mechanism remains uncertain. We note that a high rate of VUSs in BRCA genes were reclassified as LB or B, although in that study only 6.4% (2868 of 44,777) of unique variants were reclassified;³⁸ whether these data predict a similar eventual resolution of VUSs observed in DCM also remains uncertain. Augmented methods for variant interpretation allowing the assessment of criteria consistent with oligogenic inheritance will be needed for disorders with complex genetics, including DCM. Such methods may provide a pathway to evaluate variants that currently do not exceed a VUS classification.

Cohort diversity is essential to understanding the genetic architecture of complex inherited disorders³⁹, including DCM. Our study aims to evaluate the genetic foundation of DCM, regardless of family history, racial, or ethnic background. Our sample of 97 probands included 47 individuals of Non-Hispanic, African Ancestry; actionable variants were identified in only 2 of these participants (4.3% vs. 27.1% in those of Non-Hispanic European ancestry). Although a variant may reach a pathogenic classification based on data from a single family, in reality most pathogenic variants, particularly missense variants, reach such classification only after combining evidence from multiple families and resources. Our study is contributing to the evidence base needed to adjudicate variants in non-European ancestry populations, but at the same time depends on currently existing evidence to classify variants in individuals of diverse backgrounds. Although we present these preliminary results for the purpose of detailing the development and performance of our variant adjudication and return of results framework, our findings echo the call for ongoing attention and inclusion of non-European ancestry families in genetics research.^{40, 41}

Conclusion

The DCM Precision Medicine Study has developed an approach to variant classification by adapting ACMG/AMP and ClinGen guidelines. Following modifications that were deemed necessary after a pilot study of 81 probands, the new framework was applied to a test set of 97 probands, and locked for continued use in this DCM research study. Our preliminary data shows that 15.5% of DCM probands had LP/P variants, and most of these were identified in individuals of Non-Hispanic European ancestry. Our results highlight the value of ethnically and racially diverse family units in research, but also illustrate the challenges in variant interpretation for genetically complex disorders.

Nonstandard Abbreviations and Acronyms:

ACMG

American College of Medical Genetics and Genomics

AMP

Association of Molecular Pathology

CDWG (of ClinGen)

Cardiovascular Domain Working Group

ClinGen

Clinical Genome Resource

CMP-EP (of ClinGen)

Cardiomyopathy Expert Panel

DCM

dilated cardiomyopathy

HE

Hispanic ethnicity

LOF

loss of function

LP

likely pathogenic

NHE

non-Hispanic ethnicity

OR

odds ratio

P

pathogenic

VUS

variant of uncertain significance

Appendix: DCM Consortium institutions and personnel participating in this study.

• Study Principal Investigator and Co-Investigators:

  The Ohio State University: Ray E. Hershberger (Principal Investigator), Daniel D. Kinnamon, Ana Morales, Elizabeth Jordan

  Nationwide Children’s Hospital: Julie M. Gastier-Foster

  University of Washington: Wylie Burke, Deborah J. Bowen, Deborah A. Nickerson, Michael O. Dorschner.

• DCM Consortium Clinical Site Principal Investigators and Clinical Site Other Significant Contributors (OSC). The following clinical sites and individuals contributed to the submission of R01HL128857 as Site Principal Investigators (Site PI) or as Other Significant Contributors (OSC):

  The Ohio State University: Garrie Haas (Site PI), William Abraham (OSC), Philip Binkley (OSC), Ayesha Hasan (OSC), Jennifer Host (OSC), Brent Lampert (OSC), Sakima Smith (OSC)

  Tufts University School of Medicine: Gordon Huggins (Site PI), Dr. Huggins also served as study co-principal investigator; David DeNofrio (OSC), Michael Kiernan (OSC)

  University of Washington: Daniel Fishbein (Site PI), Richard Cheng (OSC), Todd Dardas (OSC), Wayne Levy (OSC), Claudius Mahr (OSC), Sofia Masri (OSC), April Stempien-Otero (OSC)

  University of Maryland: Stephen Gottlieb (Site PI)

  Stanford University Medical Center: Matthew Wheeler (Site PI), Euan Ashley (OSC)

  Medstar Health Research Institute, Washington Hospital Center: Mark Hofmeyer (Site PI)

  Cleveland Clinic Lerner College of Medicine: Wilson Tang (Site PI), Randall Starling (OSC), Rocio Moran (OSC)

  University of Pennsylvania: Anjali Owens (Site PI), Kenneth Marguilies (OSC), Thomas Cappola (OSC), Lee Goldberg (OSC), Susan Brozena (OSC), J. Rame (OSC), Rhondalyn McLean (OSC)

  University of Mississippi Medical Center: Charles Moore (Site PI), Matthew deShazo (OSC), Robert Long (OSC)

  South Miami Hospital: Francisco Jimenez Carcamo (Site PI), Hakop Hrachian-Haftevani (OSC)

  The Methodist Hospital Research Institute: Barry Trachtenberg (Site PI), Guhu Ashrith (OSC), Arvind Bhimarahj (OSC), Jerry Estep (OSC)

• The following clinical site was added following approval of NHGRI supplemental funding but prior to initiation of enrollment:

  University of Arizona: Nancy Sweitzer (Site PI)

• The following clinical sites were added following study activation:

  Inova Heart and Vascular Institute: Palak Shah (Site PI)

  University of Nebraska Medical Center: Brian Lowes (Site PI)

  Louisiana State University: Frank Smart (Site PI)

  Emory University: Alanna Morris (Site PI)

  Northwestern University: Jane Wilcox (Site PI)

  New York University Langone Medical Center: Stephan Pan (Site PI)

  Washington University: Gregory Ewald (Site PI)

  University of Michigan: Keith Aaronson (Site PI)

  University of California – Los Angeles: Jessica Wang (Site PI)

  University of Alabama – Birmingham: Salpy Pamboukian (Site PI)

  Medical University of South Carolina: Daniel Judge (Site PI)

  Cedars-Sinai Medical Center: Evan Kransdorf (Site PI)

  University of Texas Southwestern Medical Center: Sonia Garg (Site PI)

• Study Consultants

  Gail P. Jarvik, University of Washington

  Eden R. Martin, University of Miami Miller School of Medicine

  Heidi Rehm, Harvard Medical School

• National Institutes of Health Program Personnel

  National Heart, Lung, and Blood Institute: Patrice Desvigne-Nickens, James Troendle, Yi-Ping Fu.

  National Human Genome Research Institute: Lucia Hindorff