Human gene copy number spectra analysis in congenital heart malformations

Abstract

The clinical significance of copy number variants (CNVs) in congenital heart disease (CHD) continues to be a challenge. Although CNVs including genes can confer disease risk, relationships between gene dosage and phenotype are still being defined. Our goal was to perform a quantitative analysis of CNVs involving 100 well-defined CHD risk genes identified through previously published human association studies in subjects with anatomically defined cardiac malformations. A novel analytical approach permitting CNV gene frequency “spectra” to be computed over prespecified regions to determine phenotype-gene dosage relationships was employed. CNVs in subjects with CHD (n = 945), subphenotyped into 40 groups and verified in accordance with the European Paediatric Cardiac Code, were compared with two control groups, a disease-free cohort (n = 2,026) and a population with coronary artery disease (n = 880). Gains (≥200 kb) and losses (≥100 kb) were determined over 100 CHD risk genes and compared using a Barnard exact test. Six subphenotypes showed significant enrichment (P ≤ 0.05), including aortic stenosis (valvar), atrioventricular canal (partial), atrioventricular septal defect with tetralogy of Fallot, subaortic stenosis, tetralogy of Fallot, and truncus arteriosus. Furthermore, CNV gene frequency spectra were enriched (P ≤ 0.05) for losses at: FKBP6, ELN, GTF2IRD1, GATA4, CRKL, TBX1, ATRX, GPC3, BCOR, ZIC3, FLNA and MID1; and gains at: PRKAB2, FMO5, CHD1L, BCL9, ACP6, GJA5, HRAS, GATA6 and RUNX1. Of CHD subjects, 14% had causal chromosomal abnormalities, and 4.3% had likely causal (significantly enriched), large, rare CNVs. CNV frequency spectra combined with precision phenotyping may lead to increased molecular understanding of etiologic pathways.

structural congenital heart disease (CHD) is the most common form of congenital malformations, affecting 0.8% of live births (21). Other than infection, more children die from CHD in infancy than from all other forms of disease (25). In addition, it is estimated that at least 10% of early miscarriages are a consequence of severe cardiac malformations (10). The causes of congenital cardiac malformations are largely unknown. It is estimated that 18% are due to chromosomal causes or genetic structural abnormalities including trisomies (Trisomy 21, 13, and 18) as well as deletion syndromes; all of these are associated with significant disease risk for CHD (36). A small percentage of congenital cardiac malformations are disorders in which underlying single genes have been discovered such as TBX5 in Holt-Oram syndrome; JAG1 in Alagille syndrome; and PTPN11, SOS1, and KRAS in Noonan syndrome (36). Known environmental risk factors during pregnancy, such as maternal diabetes or prenatal exposure to drugs, viruses, and reduced folate intake account for a small percentage of CHD cases (16, 24). Although our understanding of molecular pathways in cardiac development has grown tremendously in the past few years, the etiology of human and clinically relevant CHD in the majority (∼75%) of cases cannot yet be identified or explained (14, 16).

The widespread use of microarray-based genomic technologies over the past 5–6 yr have implicated copy number variants (CNVs) in numerous disorders such as neuropsychiatric diseases (49), craniofacial phenotypes, cancer, and congenital anomalies including CHD (7, 18, 35, 36). Relative to sequence variations such as single base-pair mutations or single nucleotide polymorphisms (SNPs), rare and large CNVs are hypothesized to confer higher disease risk as entire genes are deleted or duplicated (12, 31). However, poor reproducibility between microarray platforms and the lack of standardized analytical tools highlight the importance of careful filtering in CNV detection studies (37). Nondisease-related copy number polymorphisms (CNPs and/or common CNVs ≥1%) are abundant, as evidenced by the growing Database of Genomic Variants (DGV) (22, 57). Similar to the challenges in the sequence analysis of unique genetic variants, the discovery of rare etiologic CNVs remains a challenge, both because it is more difficult to detect a rare event over another event seen many times and because of the intrinsic low prior probability of there being such a variant at any particular location in the genome in any individual (28).

Recently, an algorithm to clinically interpret CNVs in patients with CHD was described (6). This approach is primarily based on gene content and overlap with known causal CHD syndromes, rather than on CNV inheritance and size (6). We employed a parallel approach in this study and utilized a strict criteria to define “likely causal” duplications or deletions, in well-established human CHD risk genes. We chose 100 CHD risk genes or regions that were supported by published observations in human studies as a means to identify potentially disease-relevant CNVs. A majority of these known CHD risk genes were previously described or could be identified through the CHD WIKI portal (1, 36). In addition, genes associated with recognized causal chromosomal abnormalities in CHD were included, as well as recently identified candidate genes from association studies (see Table 1) (1, 42).

Table 1.

Known CHD risk genes

Gene	Gene Name	Cytoband	Gene Start	Gene Size	ABI CN Assay #	CHD WIKI	OMIM ID	PubMed ID
ACP6	ACID PHOSPHATASE 6, LYSOPHOSPHATIDE	1q21.1	145585791	23467	Hs00320736_cn		611471	15117819, 19597493
ACTC1	ACTIN, ALPHA, CARDIAC MUSCLE	15q14	32867588	7631		NS	102540	17611253, 17947298
ACVR2B	ACTIVIN A RECEPTOR, TYPE IIB	3p22.2	38470793	38844		NS	602730	20193066
ALDH1A2	ALDEHYDE DEHYDROGENASE 1 FAMILY, MEMBER A2	15q22.1	56032918	112280		NS	603687	19886994
ANKRD1	ANKYRIN REPEAT DOMAIN-CONTAINING PROTEIN 1	10q23.31	92661836	9176		NS	609599	18273862, 20193066
ASXL2	ADDITIONAL SEX COMBS-LIKE 2	2p23.3	25815756	139060			612991	19597493
ATRX	ATR-X GENE	Xq21.1	76647011	281364		S	300032	20193066
BCL9	B-CELL CLL/LYMPHOMA 9	1q21.1	145479805	84834	Hs01608359_cn		602597	15117819, 19597493
BCOR	BCL6 COREPRESSOR	Xp11.4	39795442	46221	Hs02764783_cn	S	300485	15770227
BRAF	V-RAF MURINE SARCOMA VIRAL ONCOGENE HOMOLOG B1	7q34	140080281	190752		S	164757	16474404, 19206169
CBL	CAS-BR-M MURINE ECOTROPIC RETROVIRAL TRANSFORMING SEQUENCE HOMOLOG	11q23.3	118582199	101870			165360	15266616
CFC1	CRYPTIC PROTEIN	2q21.1	131066804	6748		NS, S	605194	11062482
CHD1L	CHROMODOMAIN HELICASE DNA-BINDING PROTEIN 1-LIKE	1q21.1	145180914	53153	Hs00327255_cn		613039	15117819, 19597493
CHD7	CHROMODOMAIN HELICASE DNA-BINDING PROTEIN 7	8q12.2	61753892	188129	Hs01604098_cn	S	608892	15300250
					Hs01362863_cn
CITED2	CBP/p300-INTERACTING TRANSACTIVATOR, WITH GLU/ASP-RICH C-TERMINAL DOMAIN	6q24.1	139735091	2387		NS	602937	16287139
COL2A1	COLLAGEN, TYPE II, ALPHA-1	12q13.11	46653014	31538	Hs00560273_cn	S	120140	20193066
CRELD1	CYSTEINE-RICH PROTEIN WITH EGF-LIKE DOMAINS 1	3p25.3	9950505	11585		NS	607170	12632326
CRKL	V-CRK AVIAN SARCOMA VIRUS CT10 ONCOGENE HOMOLOG-LIKE	22q11.21	19601713	36177	Hs01301005_cn		602007	20494672**
CSDE1	COLD-SHOCK DOMAIN-CONTAINING E1, RNA-BINDING	1p13.2	115061059	41135		S	191510	20193066
EHMT1	EUCHROMATIC HISTONE METHYLTRANSFERASE 1	9q34.3	139725237	125162	Hs00150023_cn	S	607001	16826528, 20193066
ELN	ELASTIN	7q11.23	73080362	41810	Hs03073113_cn	NS, S	130160	12952863
EVC	ELLIS-VAN CREVELD SYNDROME	4p16.1	5763824	103108		S	225500	12571802
EVC2	EVC2 GENE	4p16.1	5615052	146143		S	607261	12571802
FBN1	FIBRILLIN 1	15q21.1	46487796	237414		S	134797	10441597, 18412115
FKBP6	FK506-BINDING PROTEIN 6	7q11.23	72380235	30342	Hs03635913_cn		604839	12952863
					Hs03630484_cn
FLNA	FILAMIN A	Xq28	153230093	26107		S	300017	17190868
FMO5	FLAVIN-CONTAINING MONOOXYGENASE 5	1q21.1	145124461	39085	Hs02744463_cn		603957	15117819, 19597493
FOXC1	FORKHEAD BOX C1	6p25.3	1555679	3449	Hs02241194_cn	S	601090	15654696
FOXH1	FORKHEAD BOX H1	8q24.3	145670316	2210		NS	603621	18538293
FOXL2	FORKHEAD TRANSCRIPTION FACTOR FOXL2	3q22.3	140145755	2736	Hs01045878_cn	S	605597	18642388, 20193066
FOXL2	FORKHEAD TRANSCRIPTION FACTOR FOXL2	3q22.3	140145755	2736	Hs01045878_cn	S	605597	18642388, 20193066
GATA4	GATA-BINDING PROTEIN 4	8p23.1	11599125	55793	Hs01321405_cn	NS	600576	16025100
GATA6	GATA-BINDING PROTEIN 6	18q11.2	18003413	32812	Hs02615249_cn	NS	601656	19666519
GDF1	GROWTH/DIFFERENTIATION FACTOR 1	19p13.11	18840360	27593	Hs07489748_cn	NS	602880	17924340
GJA1	GAP JUNCTION PROTEIN, ALPHA-1	6q22.31	121798443	14129		S	121014	11470490
GJA5	GAP JUNCTION PROTEIN, ALPHA-5	1q21.1	145694955	17153	Hs00597111_cn	NS	121013	15117819
GPC3	GLYPICAN 3	Xq26.2	132497441	449891	Hs00702786_cn	S	300037	10232747, 20193066
GTF2IRD1	GTF2I REPEAT DOMAIN-CONTAINING PROTEIN 1	7q11.23	73506055	148793			604318	12952863
HAND1	HEART- AND NEURAL CREST DERIVATIVES-EXPRESSED 1	5q33.2	153834724	3293			602406	10189962
HEY2	HAIRY/ENHANCER OF SPLIT-RELATED WITH YRPW MOTIF 2	6q22.31	126112424	11684		NS	604674	20193066
HOXA1	HOMEOBOX A1	7p15.2	27099138	3012	Hs00428080_cn	S	142955	16155570
HRAS	V-HA-RAS HARVEY RAT SARCOMA VIRAL ONCOGENE HOMOLOG	11p15.5	522241	3309	Hs00137975_cn	S	190020	17054105
ISL1	ISL LIM HOMEOBOX 1	5q11.2	50714714	11606			600366	20520780
JAG1	JAGGED 1	20p12.2	10566331	36363		NS, S	601920	11152664
KIF3C	KINESIN FAMILY MEMBER 3C	2p23.3	26002958	55989			602845	19597493
KRAS	V-KI-RAS2 KIRSTEN RAT SARCOMA VIRAL ONCOGENE HOMOLOG	12p12.1	25249446	45675		S	190070	16474405, 16474404
LBR	LAMIN B RECEPTOR	1q42.12	223655826	27316		S	600024	20193066
LEFTY1	LEFT-RIGHT DETERMINATION FACTOR 1	1q42.12	224140604	2855			603037	10053005
LEFTY2	LEFT-RIGHT DETERMINATION FACTOR 2	1q42.12	224190925	4618		NS	601877	10053005
MAP2K1	MITOGEN-ACTIVATED PROTEIN KINASE KINASE 1	15q22.31	64466264	104672		S	176872	18042262
MAP2K2	MITOGEN-ACTIVATED PROTEIN KINASE KINASE 2	19p13.3	4041319	33807		S	601263	18042262
MAP3K7IP2	MITOGEN-ACTIVATED PROTEIN KINASE KINASE KINASE 7	6q25.1	149680755	93687		NS	602614	20493459
MAPK1	MITOGEN-ACTIVATED PROTEIN KINASE 1	q11.21–22q11.	20443946	108024	Hs02937892_cn		176948	21127295**
MED13L	MEDIATOR COMPLEX SUBUNIT 13-LIKE	12q24.21	114880763	318763		NS	608771	14638541
MGP	MATRIX GAMMA-CARBOXYGLUTAMIC ACID	12p12.3	14926093	4002		S	154870	9916809, 20193066
MID1	MIDLINE 1	Xp22.2	10373595	388135	Hs02158662_cn	S	300552	12833403, 20193066
					Hs02784563_cn
MLL2	MYELOID/LYMPHOID OR MIXED LINEAGE LEUKEMIA 2	12q13.12	47699024	36350		S	602113	20711175
MYH11	MYOSIN, HEAVY CHAIN 11, SMOOTH MUSCLE	16p13.11	15704492	153896	Hs00358138_cn	NS	160745	16444274
MYH6	MYOSIN, HEAVY CHAIN 6, CARDIAC MUSCLE, ALPHA	14q11.2	22921038	26284		NS	160710	15735645
MYH7	MYOSIN, HEAVY CHAIN 7, CARDIAC MUSCLE, BETA	14q11.2	22951786	22924		NS	160760	21604106, 18159245
NF1	NEUROFIBROMATOSIS, TYPE I	17q11.2	26446120	282701		S	162200	11078559, 20193066
NKX2-5	NK2 HOMEOBOX 5	5q35.2	172591743	3125		NS	600584	9651244
NKX2-6	NK2, DROSOPHILA, HOMOLOG OF, 6	8p21.1	23615909	3957		NS	611770	15649947
NODAL	NODAL, MOUSE, HOMOLOG OF	10q22.1	71862076	9353		NS	601265	19064609
NOTCH1	NOTCH, DROSOPHILA, HOMOLOG OF, 1	9q34.3	138508716	51343	Hs00041764_cn	NS	190198	16025100, 19597493
NOTCH2	NOTCH, DROSOPHILA, HOMOLOG OF, 2	1p12	120255698	158101		S	600275	16773578
NPHP3	NEPHROCYSTIN 3	3q22.1	133882143	41823	Hs02580407_cn	S	608002	19177160
NRAS	NEUROBLASTOMA RAS VIRAL ONCOGENE HOMOLOG	1p13.2	115048600	12438		S	164790	20193066
NSD1	NUCLEAR RECEPTOR-BINDING Su-var, ENHANCER OF ZESTE, AND TRITHORAX	5q35.2–5q35.3	176492685	167135	Hs00053100_cn	S	606681	15742365, 20193066
					Hs00022652_cn
PDGFRA	PLATELET-DERIVED GROWTH FACTOR RECEPTOR, ALPHA	4q12	54790020	69149		NS	173490	20071345
PITX2	PAIRED-LIKE HOMEODOMAIN TRANSCRIPTION FACTOR 2	4q25	111758028	19929			601542	16274491
PPM1K	PROTEIN PHOSPHATASE, PP2C DOMAIN-CONTAINING, 1K	4q22.1	89400555	24357			611065	19597493
PRKAB2	PROTEIN KINASE, AMP-ACTIVATED, NONCATALYTIC, BETA-2	1q21.1	145093308	17445	Hs02605549_cn		602741	15117819
PTPN11	PROTEIN-TYROSINE PHOSPHATASE, NONRECEPTOR-TYPE, 11	12q24.13	111340918	91182		S	176876	17515436
RAB10	RAS-ASSOCIATED PROTEIN RAB10	2p23.3	26110477	103305			612672	19597493
RAF1	V-RAF-1 MURINE LEUKEMIA VIRAL ONCOGENE HOMOLOG 1	3p25.1	12600099	80601	Hs02645733_cn		164760	17603483, 19597493
RAI1	RETINOIC ACID-INDUCED GENE 1	17p11.2	17525511	129979		S	607642	16845274, 20193066
ROR2	RECEPTOR TYROSINE KINASE-LIKE ORPHAN RECEPTOR 2	9q22.31	93524704	227561		S	602337	20193066
RUNX1	RUNT-RELATED TRANSCRIPTION FACTOR 1	21q22.12	35081967	261498			151385	19863549, 19172993
SALL4	SAL-LIKE 4	20q13.2	49833989	18466	Hs00139344_cn	S	607343	12843316
SEMA5A	SEMAPHORIN 5A	5p15.2	9088137	511096	Hs01709772_cn		609297	9464278
SH3PXD2B	SH3 AND PX DOMAINS-CONTAINING PROTEIN 2B	5q35.1	171693107	121025		S	613293	20137777
SHOC2	SUPPRESSOR OF CLEAR, C. ELEGANS, HOMOLOG OF	10q25.2	112713902	49511		S	602775	19684605, 20193066
SLC2A10	SOLUTE CARRIER FAMILY 2 (FACILITATED GLUCOSE TRANSPORTER), MEMBER 10	20q13.12	44771685	26707		S	606145	16550171, 20193066
SOS1	SON OF SEVENLESS, DROSOPHILA, HOMOLOG 1	2p22.1	39062193	138915		S	182530	17143285
SOX7	SRY-BOX 7	8p23.1	10618687	6745	Hs00923277_cn		612202	19606479
STRA6	STIMULATED BY RETINOIC ACID 6, MOUSE, HOMOLOG OF	15q24.1	72258860	23385	Hs01994903_cn	S	610745	17273977
TBX1	T-BOX 1	22q11.21	18124225	26887	Hs01313390_cn	NS, S	602054	14585638
TBX20	T-BOX 20	7p14.3	35208566	51201	Hs04957392_cn	NS	606061	17668378, 19762328
TBX3	T-BOX 3	12q24.21	113592441	13911		S	601621	16892408
TBX5	T-BOX 5	12q24.21	113276117	54513		S	601620	11376442
TDGF1	TERATOCARCINOMA-DERIVED GROWTH FACTOR 1	3p21.31	46594183	4773		NS	187395	18538293, 20193066
TERT	TELOMERASE REVERSE TRANSCRIPTASE	5p15.33	1306286	41876	Hs03078158_cn		187270
TFAP2B	TRANSCRIPTION FACTOR AP2-BETA	6p12.3	50894397	28888	Hs01355864_cn	NS, S	601601	10802654
TGFBR2	TRANSFORMING GROWTH FACTOR-BETA RECEPTOR, TYPE II	3p24.1	30622997	87640			190182	15235604, 15731757
TMEM40	TRANSMEMBRANE PROTEIN 40	3p25.1	12750391	25417	Hs01878707_cn			19597493
VEGFA	VASCULAR ENDOTHELIAL GROWTH FACTOR A	6p21.1	43845930	16271		NS	192240	20420808
WHSC1	WHS CANDIDATE 1 GENE	4p16.3	1842920	110812	Hs02237093_cn		602952	9222965
ZEB2	ZINC FINGER E BOX-BINDING HOMEOBOX 2	2q22.3	144862052	132334		S	605802	11595972
ZFPM2	ZINC FINGER PROTEIN, MULTITYPE 2	8q23.1	106400322	485621		NS	603693	9927675, 10892744
ZIC3	ZINC FINGER PROTEIN OF CEREBELLUM 3	Xq26.3	136476011	5914	Hs02692150_cn	NS	300265	14681828, 10980576

NS, nonsyndromic; S, syndromic, NCBI Build 36.1/hg18.

^**Animal study.

CHD consists of heterogenous anatomy with distinct phenotypic subtypes. The European Paediatric Cardiac Coding (EPCC) System (17) has been cross mapped with the Society of Thoracic Surgeons/ European Association of Cardiothoracic Surgery (STS/EACTS) coding system through the International Society for Nomenclature of Paediatric and Congenital Heart Disease in the creation of the International Pediatric and Congenital Cardiac Code (IPCCC). We characterized cardiac malformations by subphenotyping according to both the EPCC and the STS/EACTS coding systems. We compared 945 CHD cases with a publicly available cohort of 2,026 disease-free primarily pediatric individuals (40). Cases and controls were genotyped on different platforms; therefore, a second cohort of 880 control subjects genotyped on the same platform and within the same facility as the CHD cohort was included in the analysis.

This study represents a quantitative analysis of CNVs in a large population of subjects with precisely phenotyped cardiac malformations involving 100 candidate CHD risk genes. We hypothesized that large rare CNVs that were statistically enriched against two control cohorts would be causal. A strict algorithm was employed to determine if subphenotypes were enriched in gains and losses within 100 recognized CHD risk genes selected based on gene content compared with two control cohorts. Finally, a novel analytical approach, permitting CNV gene frequency spectra to be computed as a proportion of each cohort containing a gain or a loss over the above prespecified regions, was employed to determine phenotype-gene dosage relationships.

METHODS

CHD Case Ascertainment and Confirmation

This study was reviewed and approved in accordance to institutionally approved research [Institutional Review Board (IRB)] protocols by the Children's Hospital of Wisconsin (CHW, Milwaukee, WI). Subjects were consented through the Congenital Heart Disease Tissue Bank (CHDTB) and the Wisconsin Pediatric Cardiac Registry (WPCR), IRB-approved research databases housed at CHW (20, 47). These two biobanks provide DNA samples from cases and family members, detailed maternal environmental exposure data, family history of CHD, and cardiac tissue discards.

Inclusion criteria.

Structural congenital cardiac abnormalities, as identified within the IPCCC, included abnormalities of the following: the atria and atrial septum; atrioventricular valves or atrioventricular septum; cardiac position and connections; chest wall; conduction system; coronary arteries, arterial duct, pericardium, or arteriovenous fistulae; great veins; ventricles or ventricular septum; and ventriculoarterial valves or great arteries.

Exclusion criteria.

All acquired forms of pediatric heart disease in the absence of CHD, and frequent nonpathologic structural variants when no other CHD is present, included: patent foramen ovale, patent ductus arteriosus (PDA) under 30 days of age, PDA in premature infants (<35 wk gestation) and mitral valve prolapse (in the absence of at least mild valve insufficiency).

Note: The presence of a known or suspected chromosomal abnormality or known sequence variant in a CHD risk gene did not preclude participation in the study. In addition, the presence or absence of known environmental exposures did not preclude participation in the study.

Anatomic cardiac malformations were carefully characterized by phenotyping and subphenotyping according to both the EPCC 2011 and the STS/EACTS 2011 coding systems. All phenotypes were initially reviewed by a coding specialist, a surgeon, and a cardiologist. All discrepancies were reconciled by review of source documents including operative notes, echocardiograms, and review of operative surgeon. Anatomic phenotypes and subphenotypes were reported using EPCC 2011 terms, and final confirmatory review of all cases was performed by a single pediatric cardiothoracic surgeon (17). In addition, information regarding additional diagnosis, accompanying conditions, demographics, and a limited number of genetic risk factors was obtained through the Herma Heart Center (HHC) cardiac database at CHW.

Children's Hospital of Philadelphia Control Cohort

DNA samples analyzed in this study were obtained from the whole blood of healthy subjects routinely seen at primary care and well-child clinic practices within the Children's Hospital of Philadelphia (CHOP) Health Care Network. Data using hg18/March 2006/build 36.1 genomic coordinates were downloaded from http://cnv.chop.edu/ (40). High-resolution mapping of copy number variations in 2,026 healthy individuals was performed using the Illumina HumanHap 550 BeadChip (Illumina, San Diego, CA) (40).

Milwaukee Family Heart Study Control Cohort

Control subjects were drawn from the Milwaukee Family Heart Study (MFHS) in accordance with Medical College of Wisconsin IRB protocols (MCW, Milwaukee, WI). Subjects were ascertained as a hospital-based cohort, referred to the catheterization laboratory for diagnostic coronary angiography. Inclusion criteria were the ability to consent and age >21 yr. The following were considered exclusion criteria: end-stage renal disease, current treatment for a malignancy, and a diagnosis of coronary artery disease or a myocardial infarction at age >69 yr. In addition, we excluded all participants with acute coronary syndrome and significant valvular disease. Individuals with a diagnosis of other cardiac structural abnormalities were excluded based on either the result of echocardiography prior to or as determined during the invasive cardiac procedure.

Genomic DNA Extraction

Genomic DNA for CHD and MFHS cohorts was obtained from peripheral blood using standard protocols for DNA isolation from Roche Diagnostics, Promega Biotech (Wizard), and Qiagen (Gentra Puregene). Purified genomic DNA was resuspended in 1.0 mM Tris HCl pH 8.0 and 0.1 mM EDTA. DNA quality was tested by optical density 260/280 ratios, quantified by UV spectrophotometry using a Nanodrop 2000 (Thermo Scientific, Wilmington, DE). DNA stocks were stored at −80°C, dilutions for microarray analysis were stored at 100 ng/μl at −20°C.

CHD Risk Gene Prioritization and Selection

Genes or regions with previously associated disease/syndrome variants as identified through the CHD WIKI website (searched 01/04/2011 and updated 07/28/2011) and/or supported by previously published observations in human studies were selected (1, 34, 36, 42, 48). These known CHD risk genes are outlined in Table 1.

Briefly, CHD WIKI offers an updated overview of genes implicated in human CHD, obtained by an OMIM search, and complemented with a study of the PubMed literature concerning mutation analysis of candidate genes for congenital heart defects (1). The level of support was defined by inheritance of the mutation (de novo or inherited and segregated with a phenotype) and the association of a variant in the investigated CHD population vs. a normal control population (1). A comprehensive list of 100 CHD risk genes was selected; the vast majority of these selected genes are known to be expressed in the human heart (3, 11, 43, 46, 50, 54). According to CHD WIKI, syndromic genes were defined as congenital heart defects that are associated with a second major malformation (i.e., renal defects, cleft palate, brain malformations), with developmental delay or mental handicap, and/or the presence of dysmorphism.

Genotyping

Genotyping for the CHD and MFHS control cohort was performed with the Affymetrix Genome-Wide Human SNP Array 6.0 (Affymetrix, Santa Clara, CA) as previously described (30, 47). All samples were run in the Advanced Genomics (AGEN) laboratory core at the Children's Research Institute (CRI)/MCW (Milwaukee, WI). A reference genomic DNA control sample, ref 103, supplied by Affymetrix, was run with every batch of subjects (Santa Clara, CA).

CNV Analysis and Quality Control

The CHD subject cohort comprised 1,020 subjects consented through the CHDTB or WPCR. We evaluated the quality and suitability of the subject population for a genetic association study. The population was required to pass copy number analysis quality metrics as seen in Table 2.

Table 2.

Quality control of CHD case and MFHS control cohorts and genotyping data

	Subjects, n		Subjects, n
MFHS Control Cohort		CHD Case Cohort
Starting subjects	950	Starting subjects	1,020
Remaining subjects	880	Remaining subjects	958
QC Exclusions	% Total		% Total
MAPD QC	3.05	MAPD QC	2.35
Segment QC	4.32	Segment QC	2.15
Consent QC	NA	Consent QC	0.10
Sex QC	NA	Sex QC	0.59

Copy number analysis exclusions were as follows: median absolute pairwise difference (MAPD) quality control (QC) ≥0.35, number of copy number polymorphism (CNP) segments ≥250, 1 subject with a status change to his/her consent, and sex tracking QC. Congenital heart disease (CHD) cases were reduced to a final n = 945 after inclusion and exclusion criteria were met.

CNV identification of study subjects required the processing of Affymetrix intensity (CEL) files using Genotyping Console version 3.0.2 (GTC) software as previously described (20, 47). CEL files of subjects with a median absolute pairwise difference >0.35 and a CNV segmentation count ≥250, indicative of poor DNA quality, were excluded from the study.

A final number of 945 CHD subjects and 880 MFHS controls remained in the study after inclusion and exclusion criteria were met.

As summarized in Table 3, the cases and controls were stratified according to age, sex, and race/ethnicity.

Table 3.

CHD case, CHOP, and MFHS control cohort demographics

	CHD Case Cohort	CHOP Control Cohort	MFHS Control Cohort
Race
Caucasian	655	1,320	870
African American	92	694	5
Native American	14		5
Hispanic	90
Asian	26	12
Other	68
Total	945	2,026	880
Sex, %
Female	44.02		36.59
Male	55.87		63.41
Age, yr
Median age	0.62		67.00
Average age	4.03		65.66

CHOP, Children's Hospital of Philadelphia ; MFHS, Milwaukee Family Heart Study.

Copy number state of those subjects who passed quality control thresholds were determined with reference to the GenomeWideSNP_6.hapmap270 file and copy number calls were determined using the Affymetrix GTC segmentation algorithm. To reduce the presence of false positive CNVs, the segmentation algorithm parameters were set to identify only those regions larger than 25 kb comprising at least 25 contiguous markers. It has been shown that CNVs smaller than this are frequently false positive detection (40). In addition, all segments were monitored for degree of overlap with previously identified common CNVs, annotated by the DGV (22, 57).

Using a BED file format (chromosome, gene starting position, gene ending position, gene name), copy number information was drawn from custom gene regions (Table 1) extracted from the processed segment data.

A flowchart for copy number analysis is presented in Fig. 1. A multipurpose Access database (Microsoft, Redmond, WA) served as a central repository for the cohort demographic data as well as the entire experimental set of copy number variant data. Database tables were populated with copy number data from the GTC analysis, detailed demographic data, and the annotated 100 CHD risk gene list (Table 1). Demographic data for CHD cases and MFHS controls were obtained via clinical and consent verification methods. SQL query results included aggregate CNV counts by phenotype or region for both CHD and MFHS controls. Graphical representation of the query results was accomplished using Excel (Microsoft) and R (45). Supplemental Table S1 includes a complete summary of all CNV profiles over the 100 CHD risk gene list for each subject as well as phenotypic and demographic information.¹

Fig. 1.

CNV analysis flowchart from sample to statistics. Blue figures represent software used or a process/task performed. Red figures represent data files.

Overall CNV burden.

The total number of large CNVs throughout the genome was calculated by importing GTC segment files filtered by size (duplication ≥200 kb or deletion ≥100 kb) into an Access database. An external R program further filtered CNVs for all Build 36 annotated genes that did not occur as a CNP, defined as a normal variant (≥1%) in either the CHOP or MFHS control cohorts.

Algorithm for likely causal CNV determination.

A strict algorithm was employed to determine likely causal CNVs. Gains and losses were considered as potentially disease relevant if they fulfilled the following criteria: 1) size: duplication ≥200 kb or deletion ≥100 kb, 2) they did not occur as a CNP, defined as a normal variant (≥1%) in either CHOP or MHFS control cohort, and 3) CNV occurred over a gene region known to be associated with CHD (CHD 100 gene list).

A final step was taken because the MFHS cohort was aged and significantly different from CHD cases. Sex chromosome degradation in peripheral blood appears to be an age-related phenomenon (19). Studies have shown that a strong correlation exists between patient age and loss of the Y chromosome (52). Sex chromosome degradation is easily detected by the segment reports created by GTC because males have only one copy of Chr. X. To optimize the analysis of sex chromosomes, sex-matched references were employed; for X chromosome analysis, only females from all three cohorts were compared (55). Thus male MFHS controls were excluded from X chromosome results in all CNV analyses.

CNV frequency by phenotype.

CNVs fulfilling criteria 1–3 were analyzed for enrichment by subphenotypes.

CNV frequency by gene region.

CNV frequency “spectra” were computed as a proportion of each cohort containing a gain or a loss over the CHD associated gene list.

Complex CNV analysis.

To determine if subjects carried multiple CNVs, large rare CNVs outside of and in addition to the defined set of 100 disease-related CHD genes were screened using criteria 1 and 2 (see Ref. 56).

Confirmatory Studies

CNVs that were identified in the CHD cases were confirmed by either karyotype, FISH analysis, or TaqMan CN real-time quantitative PCR assays (Applied Biosystems). CNVs for one case asterisked in Table 5 was difficult to confirm and is currently pending, due to inconclusive TaqMAN copy number results. A representative set of identified CNVs within the CHOP cohort were previously validated (40), whereas CNVs identified in the MFHS cohort as part of this study were not confirmed. As a means of secondary CNV confirmation of CHD cases, microarray analysis was performed by an independent lab on a number of the CHD study subjects (n = 34). TaqMan copy number reactions (Table 1) were run in triplicate on an ABI HT7900 instrument (Applied Biosystems) under the following cycling conditions: 50°C for 2 min, 95°C for 10 min, then 40 cycles of 95°C for 15 s followed by 60°C for 1 min. Typically ∼20 ng of template genomic DNA was amplified in reaction volumes of 10 μl, as previously described (47). Copy number confirmations were assessed using a calibrator panel of six individuals with known copy number state over the gene of interest and analyzed using Copy Caller software version 1.0 (Applied Biosystems). If parents of subjects with confirmed CNVs were available, their DNA was analyzed to determine if CNVs were inherited or de novo, as noted in Table 5.

Table 5.

Case reports of likely causal CNVs

Subject	Subphenotype	100 CHD Gene Region	Exon(s)	LOSS_GAIN	Cytoband	CNV Start (Build 36, hg18)	CNV Size, kb	Markers, n	Inheritance	Gene Names on CNV Segment (100 CHD Genes in boldface)
1	AS (valvar)	ACP6 BCL9 CHD1L FMO5 GJA5 PRKAB2	all	Loss	1q21.1	144643813	1654	684		NBPF11 FAM108A3 PRKAB2 FMO5 CHD1L BCL9 ACP6 GJA5 GJA8 GPR89B NBPF11
2	AS (valvar)	CHD1L FMO5 PRKAB2	all	Gain	1q21.1	144943150	418	280		PRKAB2 FMO5 CHD1L
		NSD1	all	Gain	5q35.2–5q35.3	175269980	1777	735		THOC3 FAM153B C5orf25 KIAA1191 ARL10 HSPC111 HIGD2A CLTB FAF2 RNF44 PCDH24 GPRIN1 SNCB EIF4E1B TSPAN17 UNC5A HK3 UIMC1 ZNF346 FGFR4 NSD1 RAB24 PRELID1 MXD3 LMAN2 RGS14 SLC34A1 PFN3 F12 GRK6 PRR7 DBN1 PDLIM7 DOK3 DDX41 FLJ10404 TMED9 B4GALT7
3	AS (valvar)	FOXC1	all	Loss	6p25.3–6p25.2	94649	2539	2130		DUSP22 IRF4 EXOC2 HUS1B FOXQ1 FOXF2 FOXC1 GMDS C6orf195 MYLK4
4	ASD-SEC	MYH11	all	Loss	16p13.11–16p12.3	15186307	2903	1521		MPV17L C16orf45 KIAA0430 NDE1 MYH11 C16orf63 ABCC1 ABCC6 NOMO3 LOC339047 XYLT1
5	ASD-SV	GATA4	all	Loss	8p23.1	11390744	304	213		BLK GATA4 NEIL2
6	AVC (partial)	GATA4 SOX7	all	Loss	8p23.1	8055434	3844	3235		PRAGMIN CLDN23 MFHAS1 THEX1 PPP1R3B TNKS MSRA UNQ9391 RP1L1 C8orf74 SOX7 PINX1 XKR6 MTMR9 AMAC1L2 FAM167A BLK GATA4 NEIL2 FDFT1 CTSB CTSB DEFB137 DEFB136 DEFB134
7	AVC (partial)	GDF1	all	Gain	19p13.11	18763592	378	163		UPF1 GDF1 LASS1 COPE DDX49 HOMER3 SFRS14 ARMC6 SLC25A42 TMEM161A MEF2B
8	AVC unbalanced + AVSD with ventricular imbalance	MID1	5′ UTR-i1	Gain	Xp22.2	10714630	509	265		MID1 HCCS ARHGAP6 AMELX
9	AVSD with TOF	CRKL TBX1	all	Gain	22q11.21	17953160	1838	1106		SEPT5 GP1BB TBX1 GNB1L C22orf29 TXNRD2 COMT ARVCF C22orf25 DGCR8 HTF9C RANBP1 ZDHHC8 RTN4R DGCR6L RIMBP3 ZNF74 SCARF2 KLHL22 MED15 PI4KA SERPIND1 SNAP29 CRKL AIFM3 LZTR1 THAP7 P2RX6 SLC7A4
10	CoA	ACP6 BCL9 CHD1L FMO5 GJA5 PRKAB2	all	Gain	1q21.1	144812585	1480	678		PRKAB2 FMO5 CHD1L BCL9 ACP6 GJA5 GJA8 GPR89B GPR89C NBPF11 LOC728912
11	CoA	NOTCH1	all	Loss	9q34.3	138377108	229	105		DNLZ CARD9 SNAPC4 SDCCAG3 PMPCA INPP5E SEC16A C9orf163 NO TCH1
12	DILV	HRAS	all	Gain	11p15.5	354390	256	62		B4GALNT4 PKP3 SIGIRR TMEM16J PTDSS2 RNH1 HRAS LRRC56 C11orf35 RASSF7 KIAA1542 IRF7 MUPCDH
13	DORV	SEMA5A	i8-3′ UTR	Gain	5p15.31–5p15.2	7119715	2152	1769		ADCY2 C5orf49 FASTKD3 MTRR SEMA5A
14	EBSTEIN'S	FKBP6	5′ UTR-i8	Gain	7q11.23	72073034	330	28		TRIM74 STAG3L3 NSUN5 TRIM50 FKBP6
15	HLHS	EHMT1	all	Gain	9q34.3	139701521	264	142		EHMT1 CACNA1B
16	HLHS	FKBP6	5′ UTR-i8	Gain	7q11.23	72052197	348	34	de novo	POM121 NSUN5C TRIM74 ST AG3L3 NSUN5 TRIM50 FKBP6
17	HLHS	GATA4	all	Gain	8p23.1	11049252	1438	755	unknown	XKR6 MTMR9 AMAC1L2 FAM167A BLK GATA4 NEIL2 FDFT1 CTSB DEFB137 DEFB136 DEFB134 DEFB130 ZNF705D DUB3 FAM86B1 DEFB130
		SOX7	all	Gain	8p23.1	8055434	2992	2584	unknown	PRAGMIN CLDN23 MFHAS1 THEX1 PPP1R3B TNKS MSRA UNQ9391 RP1L1 C8orf74 SOX7 PINX1 XKR6
18	HLHS	MYH11	all	Gain	16p13.11	14846829	1414	640	inherited	NOMO1 NPIP PDXDC1 NTAN1 RRN3 MPV17L C16orf45 KIAA0430 NDE1 MYH11 C16orf63 ABCC1 ABCC6 NOMO3
19	Other, Cardiac	CRKL TBX1	all	Gain	22q11.21	17161534	2634	1575		DGCR6 PRODH DGCR2 DGCR14 TSSK2 GSC2 SLC25A1 CLTCL1 HIRA MRPL40 C22orf39 UFD1L CDC45L CLDN5 SEPT5 GP1BB TBX1 GNB1L C22orf29 TXNRD2 COMT ARVCF C22orf25 DGCR8 HTF9C RANBP1 ZDHHC8 RTN4R DGCR6L RIMBP3 ZNF74 SCARF2 KLHL22 MED15 PI4KA SERPIND1 SNAP29 CRKL AIFM3 LZTR1 THAP7 P2RX6 SLC7A4
20	PA, VSD	ACP6 BCL9 CHD1L FMO5 GJA5 PRKAB2	all	Gain	1q21.1	144812585	1480	678		PRKAB2 FMO5 CHD1L BCL9 ACP6 GJA5 GJA8 GPR89B GPR89C NBPF11 LOC728912
21	Subaortic stenosis	CRKL	all	Gain	22q11.21	19093207	699	626		ZNF74 SCARF2 KLHL22 MED15 PI4KA SERPIND1 SNAP29 CRKL AIFM3 LZTR1 THAP7 P2RX6 SLC7A4
22	Subaortic stenosis	HRAS	all	Gain	11p15.5	339238	271	63		B4GALNT4 PKP3 SIGIRR TMEM16J PTDSS2 RNH1 HRAS LRRC56 C11orf35 RASSF7 KIAA1542 IRF7 MUPCDH
23	DORV	FOXL2 NPHP3	all	Gain	3q22.1–3q26.1	131972967	32134	19750		PIK3R4 ATP2C1 ATP2C1 ASTE1 NEK11 NUDT16 MRPL3 CPNE4 ACPP DNAJC13 ACAD11 CCRL1 UBA5 NPHP3 TMEM108 BFSP2 CDV3 TOPBP1 TF SRPRB RAB6B
										C3orf36 SLCO2A1 RYK AMOTL2 ANAPC13 CEP63 KY EPHB1 PPP2R3A MSL2L1 PCCB ST AG1 TMEM22 NCK1 IL20RB SOX14 CLDN18 DZIP1L A4GNT DBR1 ARMC8 TXNDC6 MRAS FAM62C CEP70 FAIM PIK3CB FOXL2 C3orf72 LOC389151 MRPS22 COPB2 RBP2 RBP1 NMNAT3 CLSTN2 TRIM42 SLC25A36 SPSB4 ACPL2 ZBTB38 RASA2 RNF7 GRK7 ATP1B3 TFDP2 GK5 XRN1 ATR PLS1 TRPC1 PCOLCE2 PAQR9 SR140 CHST2 SLC9A9 C3orf58 PLOD2 PLSCR4 PLSCR2 PLSCR1 PLSCR5 ZIC4 ZIC1 AGTR1 CPB1 CPA3 GYG1 HLTF HPS3 CP TM4SF18 TM4SF1 TM4SF4 WWTR1 COMMD2 RNF13 RNF13 PFN2 TSC22D2 SERP1 EIF2A SELT C3orf44 SIAH2 CLRN1 CLRN1 MED12L GPR171 P2RY14 GPR87 P2RY13 P2RY13 P2RY12 IGSF10 AADACL2 AADAC SUCNR1 MBNL1 TMEM14E P2RY1 RAP2B LOC152118 SGEF DHX36 GPR149 MME PLCH1 C3orf33 SLC33A1 GMPS KCNAB1 SSR3 TIPARP LEKR1 CCNL1 VEPH1 PTX3 C3orf55 SHOX2 RSRC1 MLF1 GFM1 LXN RARRES1 MFSD1 IQCJ SCHIP1 IL12A IFT80 SMC4 TRIM59 KPNA4 ARL14 PPM1L B3GALNT1 NMD3 C3orf57 OTOL1 SI SLITRK3 BCHE ZBBX SERPINI2 WDR49 PDCD10 SERPINI1 GOLIM4 EVI1 EVI1 MDS1 ARPM1 MYNN LRRC34 LRRIQ4 LRRC31 SAMD7 SEC62 GPR160 PHC3 PRKCI SKIL CLDN11 SLC7A14 RPL22L1 EIF5A2 SLC2A2 TNIK PLD1 FNDC3B GHSR TNFSF10 AADACL1 ECT2 SPATA16 NLGN1 NAALADL2 TBL1XR1 KCNMB2 ZMAT3 PIK3CA KCNMB3 ZNF639 MFN1 GNB4 ACTL6A MRPL47 NDUFB5 USP13 PEX5L TTC14 CCDC39 FXR1 DNAJC19 SOX2 ATP11B DCUN1D1 MCCC1 LAMP3 MCF2L2 B3GNT5 KLHL6 KLHL24 YEATS2 MAP6D1 PARL ABCC5 HTR3D HTR3C HTR3E
										EIF2B5 DVL3 AP2M1 ABCF3 ALG3 ECE2 CAMK2N2 ECE2 PSMD2 EIF4G1 FAM131A CLCN2 POLR2H THPO CHRD EPHB3 MAGEF1 VPS8 C3orf70 EHHADH MAP3K13 TMEM41A LIPH SENP2 IGF2BP2 C3orf65 SFRS10 ETV5 DGKG CRYGS TBCCD1 DNAJB11 AHSG FETUB HRG KNG1 EIF4A2 RFC4 ADIPOQ ST6GAL1 RPL39L RTP1 MASP1 RTP4 SST RTP2 BCL6 LPP TPRG1 TP63 LEPREL1 SENP2 IGF2BP2 C3orf65 SFRS10 ETV5 DGKG CRYGS TBCCD1 DNAJB11 AHSG FETUB HRG KNG1 EIF4A2 RFC4 ADIPOQ ST6GAL1 RPL39L RTP1 MASP1 RTP4 SST RTP2 BCL6 LPP TPRG1 TP63 LEPREL1
24	TOF	ACP6 BCL9 GJA5	all	Gain	1q21.1	145250193	1678	471		BCL9 ACP6 GJA5 GJA8 GPR89B GPR89C NBPF11 LOC728912 PPIAL4 NBPF14 NBPF10 NBPF15 NBPF16
		CHD1L FMO5 PRKAB2	all	Gain	1q21.1	144643813	600	223		NBPF11 LOC728912 FAM108A3 PRKAB2 FMO5 CHD1L
25	TOF	CRKL	all	Loss	22q11.21	18710744	1085	673		RIMBP3 ZNF74 SCARF2 KLHL22 MED15 PI4KA SERPIND1 SNAP29 CRKL AIFM3 LZTR1 THAP7 P2RX6 SLC7A4
26	TOF	HOXA1	all	Gain	7p15.2–7p15.1	26113744	4718	3324		NFE2L3 HNRNP A2B1 CBX3 SNX10 SKAP2 HOXA1 HOXA2 HOXA3 HOXA4 HOXA5 HOXA6 HOXA7 HOXA9 HOXA10 HOXA11 HOXA13 EVX1 HIBADH TAX1BP1 JAZF1 LOC402644 CREB5 KIAA0644 CPVL CHN2 PRR15 WIPF3 SCRN1 FKBP14 PLEKHA8 C7orf41 ZNRF2 NOD1 C7orf24 GARS CRHR2 INMT FLJ22374
		TBX20	all	Gain	7p14.3–7p14.2	32897122	3321	2145		KBTBD2 FKBP9 NT5C3 RP9 BBS9 BMPER NPSR1 DPY19L1 TBX20 HERPUD2 SEPT7 EEPD1
27	TOF	MYH11	all	Gain	16p13.11	14805290	1455	642		NOMO1 NPIP PDXDC1 NTAN1 RRN3 MPV17L C16orf45 KIAA0430 NDE1 MYH11 C16orf63 ABCC1 ABCC6 NOMO3
28*	TOF	TERT	all	Loss	5p15.33	80069	2948	1893		PLEKHG4B LOC389257 CCDC127 SDHA PDCD6 LOC116349 EXOC3 SLC9A3 CEP72 TPPP ZDHHC11 BRD9 TRIP13 NKD2 SLC12A7 SLC6A19 SLC6A18 TERT CLPTM1L SLC6A3 LPCAT1 MRPL36 NDUFS6 IRX4 IRX2 C5orf38
29	TRI-AT	MAPK1	all	Gain	22q11.21–22q11.22	20264556	447	243		UBE2L3 YDJC CCDC116 SDF2L1 PPIL2 YPEL1 MAPK1 PPM1F TOP3B
30	TRI-AT	NSD1	e24-3′ UTR	Gain	5q35.3	176656286	330	133		NSD1 RAB24 PRELID1 MXD3 LMAN2 RGS14 SLC34A1 PFN3 F12 GRK6 PRR7 DBN1 PDLIM7 DOK3 DDX41 FLJ10404 TMED9 B4GALT7
31	Truncus arteriosus	MAPK1	all	Loss	22q11.21–22q11.22	20055986	1237	863		HIC2 RIMBP3B RIMBP3C UBE2L3 YDJC CCDC116 SDF2L1 PPIL2 YPEL1 MAPK1 PPM1F TOP3B VPREB1 ZNF280B ZNF280A PRAME
		GATA6	all	Gain	18q11.2	17749666	308	168		GATA6
32	Truncus arteriosus	SALL4	all	Loss	20q13.2	49428074	1839	1357		NFATC2 ATP9A SALL4 ZFP64 TSHZ2
33	VSD (perimembranous)	CRKL	all	Gain	22q11.21	19389671	406	451		PI4KA SERPIND1 SNAP29 CRKL AIFM3 LZTR1 THAP7 P2RX6 SLC7A4
		ACP6 BCL9 CHD1L FMO5 GJA5 PRKAB2	all	Loss	1q21.1	144723763	1574	683		NBPF11 LOC728912 FAM108A3 PRKAB2 FMO5 CHD1L BCL9 ACP6 GJA5 GJA8 GPR89B GPR89C NBPF11
34	VSD (perimembranous)	GATA4 SOX7	all	Loss	8p23.1	8027361	4456	3349		PRAGMIN CLDN23 MFHAS1 THEX1 PPP1R3B TNKS MSRA UNQ9391 RP1L1 C8orf74 SOX7 PINX1 XKR6 MTMR9 AMAC1L2 FAM167A BLK GATA4 NEIL2 FDFT1 CTSB CTSB DEFB137 DEFB136 DEFB134 DEFB130 ZNF705D DUB3 FAM86B1 DEFB130
35	VSD (perimembranous)	GATA6	all	Gain	18q11.1–18q11.2	16795645	6118	3683		ROCK1 ESCO1 SNRPD1 ABHD3 MIB1 GATA6 CTAGE1 RBBP8 CABLES1 C18orf45 RIOK3 C18orf8 NPC1 ANKRD29 LAMA3 TTC39C CABYR OSBPL1A IMPACT HRH4 ZNF521 SS18 PSMA8 TAF4B KCTD1 AQP4 AQP4 CHST9

^*Inconclusive TAQMAN results (see Subject 28). Boldface indicates confirmed genes. “Unknown” means one parental DNA was unavailable. “Other cardiac” phenotype (case 19) is double-chamber right ventricle (DCRV).

Statistical Analysis

Since the expected incidence is very small (typically <5%) tests based on a normality assumption would be incorrect, therefore a one-tailed Barnard exact test was used for all comparisons of proportions of CNVs (8). A P ≤ 0.05 without adjustment is used for significance. A custom R program was used to calculate the P value and checked using Cytel StatXact (Cytel, Cambridge, MA) (15). StatXact was also used to calculate power. With a sample of 810, and a CNV incidence of 4.3%, we would have at least 90% power to detect a significant difference from 0.0196 (the CNV incidence of CHOP cohort's 39/2,026). We have given other power calculations for possible scenarios of subphenotypes (Fig. 2). We see that in an n = 100 sample group we would have ≥80% power if we had an 8% CNV incidence. For a cohort of 200 we would have ≥80% power to detect a difference of 6% CNV incidence.

Fig. 2.

Sample size (n) and copy number variant (CNV) proportion (fraction), required to detect difference from 0.0196 (CHOP control CNV fraction) at an alpha = 0.05, power at least 80%. CHOP, Children's Hospital of Philadelphia.

This figure demonstrates the sample size required (x-axis) with power of at least 80% under varying CNV proportions (y-axis) when the control cohort is 0.0196 (CHOP control CNV proportion) at an alpha = 0.05.

RESULTS

Phenotypes of CHD Study Subjects

Subjects diagnosed with congenital heart malformations (n = 945) and phenotyped in accordance with the EPCC terms were categorized into the 40 cardiac subphenotypes listed in Table 4 (17). The five largest phenotypes represented were as follows: hypoplastic left heart syndrome (HLHS) 14.8%, ventricular septal defect (VSD perimembranous) 7.7%, tetralogy of Fallot (TOF) 7.7%, coarctation of the aorta (CoA) 7.0%, and atrioventricular canal complete (AVC complete) 5.0%. The majority of subjects were represented by individual subphenotypes most of which contained <5.1% of the total CHD cohort.

Table 4.

CHD cohort by subphenotypes

Diagnoses	Subjects	% of Total	Diagnoses	Subjects	% of Total
Aorto-pulmonary window + Patent Ductus Arteriosus (PDA)T21	5	0.53	Mitral Valve Stenosis (MS, subvalvar, parachute)22q	6	0.63
AVSD + TOF (AVSD + TOF)T21	7	0.74	Other, CardiacT21, 22q	18	1.90
Arrhythmias (Congenital Heart Block, Long QT, WPW)	7	0.74	Pulmonary Atresia (PA)
Aortic Stenosis (Valvar)T	31	3.28	- IVS-T21	18	1.90
Atrial Septal Defect Secundum (ASD-SEC)T21	47	4.97	- VSD-22q	34	3.60
Atrial Septal Defect Sinus Venosus (ASD-SV)	13	1.38	PAPVR	12	1.27
A-V Canal Complete (AVC Complete)T21	48	5.08	Pulmonary Stenosis (Valvar)	9	0.95
A-V Canal Intermediate (AVC Intermediate)T21	7	0.74	Shone's	8	0.85
A-V Canal Partial (AVC Partial)T21	17	1.80	Subaortic stenosisT21	12	1.27
A-V Canal Unbalanced + AVSD with ventricular imbalanceT21	14	1.48	Supravalvar aortic stenosis (supravalvar AS)	4	0.42
Cardiomyopathy (DILATED)	13	1.38	Total Anomalous Venous Connection (TAPVC)	15	1.59
Cardiomyopathy (HYPERTROPHIC)	4	0.42	Tetralogy of Fallot (TOF)T21, 22q	73	7.72
Chest Wall	4	0.42	Transposition of Great Arteries (TGA)
Coarctation of the Aorta (CoA)T	66	6.98	- IVS -	21	2.22
Coronary Arteries (COR ART)	10	1.06	- VSD -	21	2.22
Double Inlet Left Ventricle (DILV)	19	2.01	Tricuspid Atresia (TRI-AT)	29	3.07
Double Outlet Right Ventricle (DORV)22q	41	4.34	Truncus Arteriosus (TA)22q	29	3.07
Ebstein's Anomaly (EBSTEINS)	9	0.95	Vascular ring and PA slingT21, 22q	14	1.48
Hypoplastic Left Heart Syndrome (HLHS)T	140	14.81	VSD inletT21	4	0.42
Interrupted Aortic Arch (IAA)22q	11	1.16	VSD multiple + muscular	10	1.06
L-TGA	7	0.74	VSD perimembranousT21, 22q	73	7.72
Dilated Ascending Aorta (MARFAN)	8	0.85	VSD subarterialT21	7	0.74

The following individual phenotypes were included in the “other cardiac” subphenotype category: single ventricle other, absent left pulmonary artery (LPA), absent pulmonary valve, aorto-left ventricular tunnel, bicuspid aortic valve (BAV), cor triatriatum, double-chamber right ventricle (DCRV), left ventricular aneurysm, tricuspid regurgitation, true cleft of mitral leaflet (without AVSD). Superscripts were used to denote phenotypes where causal chromosomal as shown abnormalities were observed (see results, where T21 = Trisomy 21, 22q = 22qDS, and T = Turner's Syndrome). PAPVR, partial anomalous pulmonary venous return; VSD, ventricular septal defect.

Subjects With Recognized Causal Chromosomal Abnormalities

We ascribed 135 subjects to known CHD-related chromosomal abnormalities [T21 (n = 80), T18 (n = 1), 22qDS (n = 42), Turner (n = 8), William's (n = 3), and XXX (n = 1)] (36, 44). The syndromes and their associated phenotypes were as follows: T21: aorto-pulmonary window with PDA n = 2; AVSD + TOF n = 5; ASD-SEC n = 4; AVC complete n = 35; AVC intermediate n = 5; AVC partial n = 2; AVC unbalanced + AVSD with ventricular imbalance n = 1; other cardiac n = 1; pulmonary atresia (PA), IVS n = 1; subaortic stenosis n = 1; TOF n = 6; vascular ring + PA sling n = 1; VSD (inlet) n = 2; VSD (perimembranous) n = 13 and VSD (subarterial) n = 1, T18: TOF n = 1, 22qDS: DORV n = 1; IAA n = 4; mitral stenosis, subvalvar, parachute + mitral stenosis n = 1; other cardiac n = 1; PA, VSD n = 10; TOF n = 9; truncus arteriosus n = 12; vascular ring + PA sling n = 1 and VSD (perimembranous) n = 3, Turner: aortic stenosis (valvar) n = 1; CoA n = 4 and HLHS n = 2, mosaic Turner: CoA n = 1, William's: supravalvar aortic stenosis and XXX: PA, IVS.

Overall CNV Burden

The total number of large CNVs (≥100 kb loss, ≥200 kb gain) throughout the genome were similar in both CHD and MFHS cohorts. When subjects with chromosomal abnormalities such as Trisomy 21 and 18, Turner, 22qDS, William's, and XXX were excluded, a significant number of the CHD cohort, 567 out of 810, carried a large rare CNV over a gene somewhere in their genome, while in the MFHS control cohort, this number was 391 of 880. Gains were twofold more common than losses in both cohorts despite the requirement to be twice as long (Fig. 3).

Fig. 3.

Total CNV burden by cohort. Standard box-and-whiskers plot for the distribution of large rare CNV segment count per subject in each of 4 cases: congenital heart disease (CHD) vs. Milwaukee Family Heart Study (MFHS) and gains vs. losses. Boxes represent the 1st and 3rd quartiles of each distribution, thick horizontal lines represent the median value, circles represent outliers, or the CHD cohort, major syndromes would significantly skew the distribution, so those subjects were excluded, leaving 810 syndrome-free subjects. Trisomy 21 and 18, Turner, 22qDS, William's and XXX chromosomal abnormalities were therefore excluded.

CHD Case Reports

Likely etiologic large, rare CNVs were identified in 35 CHD subjects. Table 5 summarizes the complete list of CHD subjects with CNVs over the known CHD risk gene regions (excluding the 135 subjects with known CHD-related chromosomal abnormalities). Three HLHS subjects (cases 16, 17, and 18) were studied for inheritance, a gain over FKBP6 was found to be a de novo event, a gain involving GATA4 and SOX7 was not present in one parent and the status of the other parent was unknown, and the MYH11 gain was inherited. Table 5 reports all of the known genes within each CNV segment, including our selected 100 CHD-associated genes.

Statistical Analysis of CNVs

Subphenotype analysis.

The CHD cohort, even after excluding genes involved in the known CHD-related chromosomal abnormalities, was enriched in large, rare CNVs involving CHD risk genes, where 35 of 810 subjects carried such a CNV (P ≤ 0.05 vs. both CHOP with 39 of 2,026 and MFHS with 14 of 880). Breaking this cohort into subgroups by specific phenotype often resulted in groups too small for statistical significance. Different subdivision schemes may achieve nominal significance. The entries in Table 6 where the frequency of CNV was significantly (P ≤ 0.05) different from the CHOP and MFHS cohorts are marked with a double asterisk. The CHD cohort, after excluding known causal chromosomal abnormalities, showed a frequency of CNV at 4.3%, and a power calculation is performed in Fig. 2 showing the difficulty in detecting a difference from the control's 1.9%. For subgroups of 10–25 individuals, the power to detect a difference from 1.9% (CHOP) required a proportion of 30 and 17%, respectively. Phenotypes showing significant (P ≤ 0.05) enrichment of large CNV events were aortic stenosis (valvar), AV canal (partial), AVSD with TOF, subaortic stenosis, TOF, and truncus arteriosus. Although HLHS was the most common phenotype in the CHD case cohort, this phenotype did not demonstrate significant large rare CNV enrichment.

Table 6.

CNV frequency by subphenotype

	Totals Including Causal Chromosomal Abnormalities				Totals Excluding Causal Chromosomal Abnormalities
Phenotype/Subphenotype	Subjects	Subjects with CNV Loss	Subjects with CNV Gain	Subjects with CNV Loss or Gain	Subjects	Subjects with CNV Loss	Subjects with CNV Gain	Subjects with CNV Loss or Gain
CHOP Cohort	2,026	19 (0.94)	20 (0.99)	39 (1.92)	2,026	19 (0.94)	20 (0.99)	39 (1.92)
MFHS Cohort	880	3 (0.34)	11 (1.25)	14 (1.59)	880	3 (0.34)	11 (1.25)	14 (1.59)
CHD Cohort	945	66 (6.98)	110 (11.64)	172 (18.20)**	810	12 (1.48)	23 (2.84)	35 (4.32)**
Turner	8	8 (0.84)	1 (0.10)	8
Trisomy18 (T18)	1	0 (0.00)	1 (0.10)	1
Trisomy21 (T21)	80	0 (0.00)	80 (8.35)	80
Williams	3	3 (0.31)	0 (0.00)	3
XXX	1	0 (0.00)	1 (0.10)	1
22qDS	42	42 (4.38)	1 (0.10)	42
Aorto-pulmonary window + PDA	5	0 (0.00)	2 (40.00)	2 (40.00)**	3	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	2			2	2
22qDS	0				0
AVSD + TOF (AVSD + TOF)	7	0 (0.00)	6 (85.71)	6 (85.71)**	2	0 (0.00)	1 (50.00)	1 (50.00)**
Trisomy21	5		5	5
22qDS	0			0
Aortic Stenosis (Valvar)	31	3 (9.68)	1 (3.23)	4 (12.90)**	30	2 (6.67)	1 (3.33)	3 (10.00)**
Turner	1	1		1
Trisomy21	0			0
22qDS	0			0
Atrial Septal Defect Secundum (ASD-SEC)	47	1 (2.13)	4 (8.51)	5 (10.64)**	43	1 (2.33)	0 (0.00)	1 (2.33)
Trisomy21	4		4	4
22qDS	0			0
Atrial Septal Defect Sinus Venosus (ASD-SV)	13	1 (7.69)	0 (0.00)	1 (7.69)	13	1 (7.69)	0 (0.00)	1 (7.69)
Trisomy21	0			0
22qDS	0			0
A-V Canal Complete (AVC Complete)	48	0 (0.00)	35 (72.92)	35 (72.92)**	13	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	35		35	35
22qDS	0			0
A-V Canal Intermediate (AVC     Intermediate)	7	0 (0.00)	5 (71.43)	5 (71.43)**	2	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	5		5	5
22qDS	0			0
A-V Canal Partial (AVC Partial)	17	1 (5.88)	3 (17.65)	4 (23.53)**	15	1 (6.67)	1 (6.67)	2 (13.33)**
Trisomy21	2		2	2
22qDS	0			0
A-V Canal Unbalanced + AVSD with ventricular imbalance	14	0 (0.00)	2 (14.29)	2 (14.29)**	13	0 (0.00)	1 (7.69)	1 (7.69)
Trisomy21	1		1	1
22qDS	0			0
Coarctation of the Aorta (CoA)	66	6 (9.09)	2 (3.03)	8 (12.12)**	61	1 (1.64)	1 (1.64)	2 (3.28)
Turner	5	5	1	5
Trisomy21	0			0
22qDS	0			0
Double Inlet Left Ventricle (DILV)	19	0 (0.00)	1 (5.26)	1 (5.26)	19	0 (0.00)	1 (5.26)	1 (5.26)
Trisomy21	0			0
22qDS	0			0
Double Outlet Right Ventricle (DORV)	42	1 (2.38)	3 (7.14)	4 (9.52)**	41	0 (0.00)	2 (4.88)	2 (4.88)
Trisomy21	0			0
22qDS	1	1	1	1
Ebstein's Anomaly (EBSTEINS)	9	0 (0.00)	1 (11.11)	1 (11.11)	9	0 (0.00)	1 (11.11)	1 (11.11)
Trisomy21	0			0
22qDS	0			0
Hypoplastic Left Heart Syndrome (HLHS)	140	2 (1.43)	5 (3.57)	7 (5.00)**	138	0 (0.00)	4 (2.90)	4 (2.90)
Turner	2	2		2
Trisomy21	0			0
22qDS	0			0
Interrupted Aortic Arch (IAA)	11	4 (36.36)	0 (0.00)	4 (36.36)**	7	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	0			0
22qDS	4	4		4
Mitral Valve Stenosis (MS, subvalvar, parachute)	6	1 (16.67)	0 (0.00)	1 (16.67)	5	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	0			0
22qDS	1	1		1
Other, Cardiac	18	1 (5.56)	2 (11.11)	3 (16.67)**	16	0 (0.00)	1 (6.25)	1 (6.25)
Trisomy21	1		1	1
22qDS	1	1		1
Pulmonary Atresia (PA)					16	0 (0.00)	0 (0.00)	0 (0.00)
-IVS-	18	0 (0.00)	2 (11.11)	2 (11.11)**
Trisomy21	1		1	1
XXX	1		1	1
22qDS	0			0
-VSD-	34	10 (29.41)	1 (2.94)	11 (32.35)**	24	0 (0.00)	1 (4.17)	1 (4.17)
Trisomy21	0			0
22qDS	10	10		10
Subaortic stenosis	12	0 (0.00)	3 (25.00)	3 (25.00)**	11	0 (0.00)	2 (18.18)	2 (18.18)**
Trisomy21	1		1	1
22qDS	0			0
Supravalvar AS	4	3 (75.00)	0 (0.00)	3 (75.00)**	1	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	0			0
Williams	3	3		3
22qDS	0			0
Tetralogy of Fallot (TOF)	73	11 (15.07)	10 (13.70)	21 (28.77)**	57	2 (3.51)	3 (5.26)	5 (8.77)**
Trisomy18	1		1	1
Trisomy21	6		6	6
22qDS	9	9		9
Tricuspid Atresia (TRI-AT)	29	0 (0.00)	2 (6.90)	2 (6.90)	29	0 (0.00)	2 (6.90)	2 (6.90)
Trisomy21	0			0
22qDS	0			0
Truncus Arteriosus (TA)	29	14 (48.28)	1 (3.45)	14 (48.28)**	17	2 (11.76)	1 (5.88)	2 (11.76)**
Trisomy21	0			0
22qDS	12	12		12
Vascular ring and PA sling	14	1 (7.14)	1 (7.14)	2 (14.29)**	12	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	1		1	1
22qDS	1	1		1
VSD inlet	4	0 (0.00)	2 (50.00)	2 (50.00)**	2	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	2		2	2
22qDS	0			0
Ventricular Septal Defect (VSD perimembranous)	73	5 (6.85)	15 (20.55)	19 (26.03)**	57	2 (3.51)	2	3 (5.26)
Trisomy21	13		13	13
22qDS	3	3		3
Ventricular Septal Defect (VSD subarterial)	7	0 (0.00)	1 (14.29)	1 (14.29)	6	0 (0.00)	0 (0.00)	0 (0.00)
Trisomy21	1		1	1
22qDS	0			0

^**Significance over both CHOP and MFHS controls (P ≤ 0.05). Four patients had both gains and losses but are only counted once in the column “Subjects with CNV Loss or Gain”. The following subphenotypes contained 0 subjects with a CNV and were therefore removed from the table: Arrhythmias (Congenital Heart Block, Long QT, WPW), 7; Cardiomyopathy (DILATED), 13; Cardiomyopathy (HYPERTROPHIC), 4; Chest Wall, 4; Coronary Arteries (COR ART), 10; L-TGA, 7; Dilated Ascending Aorta (MARFAN), 8; Partial Anomalous Pulmonary Venous Return (PAPVR), 12; Pulmonary Stenosis (Valvar), 9; Shone's, 8; Total Anamolous Pulmonary Venous Connection (TAPVC; infracardiac, intracardiac, mixed, supracardiac), 15; Transposition of Great Arteries (IVS), 21; (VSD), 20; and Ventricular Septal Defect (VSD multiple + muscular), 10 (n = 161 total).

CNV gene frequency analysis and gene enrichment.

In addition, CNV frequency “spectra” were computed as a proportion of each cohort containing a gain or a loss over 100 CHD genes of interest (Fig. 4). (Spectra for individual CHD subphenotypes with statistically higher CNV frequencies are represented in Fig. 5.) The frequency of genes with gain or loss was compared with both control cohorts and significantly enriched genes are listed in Table 7. In addition, Supplemental Table S1 includes a complete summary of all CNV profiles over the 100 CHD risk gene list for each CHD subject, and a heatmap (Supplemental Fig. S1) illustrates the clustering of various groups of multiple subjects who share contiguous blocks of deleted or duplicated genes.

Fig. 4.

CNV frequency spectrum. Note that any gene showing 0% CNV frequency in all 3 cohorts was omitted from this figure due to space considerations. Vertical error bars drawn represent 1 SD from the mean in the estimated sampling distribution. From this visualization it is clear that gains over gene FKBP6 on chromosome 7 occur in all 3 cohorts, while losses of the same gene are only seen in the CHD cohort, implying a loss could cause CHD.

Fig. 5.

CNV frequency spectra of significantly enriched phenotypes. Note that any gene showing 0% CNV frequency in all 3 cohorts was omitted from this figure due to space considerations. Vertical error bars drawn represent 1 SD in the estimated sampling distribution. Significantly enriched phenotypes included: aortic stenosis (valvar), atrioventricular canal (partial), atrioventricular septal defect (AVSD) with tetralogy of Fallot (TOF), subaortic stenosis, TOF, and truncus arteriosus.

Table 7.

CHD-associated gene regions significantly enriched with large, rare CNVs

	Gains, %			Losses, %			Enriched For
Gene	CHD	CHOP	MFHS	CHD	CHOP	MFHS	Gains	Losses
PRKAB2	0.4	0.1	0.0	0.2	0.1	0.2	√
FMO5	0.4	0.1	0.0	0.2	0.1	0.2	√
CHD1L	0.4	0.1	0.0	0.2	0.1	0.2	√
BCL9	0.3	0.1	0.0	0.2	0.1	0.2	√
ACP6	0.3	0.1	0.0	0.2	0.1	0.2	√
GJA5	0.3	0.1	0.0	0.2	0.1	0.2	√
FKBP6	0.3	0.1	0.3	0.3	0.0	0.0		√
ELN	0.0	0.0	0.0	0.3	0.0	0.0		√
GTF2IRD1	0.0	0.0	0.0	0.3	0.0	0.0		√
GATA4	0.1	0.0	0.0	0.3	0.0	0.0		√
HRAS	0.3	0.0	0.0	0.0	0.1	0.0	√
GATA6	0.3	0.0	0.0	0.0	0.0	0.0	√
RUNX1	8.5	0.0	0.0	0.0	0.0	0.0	√
CRKL	0.4	0.0	0.3	4.2	0.1	0.0		√
TBX1	0.2	0.0	0.2	4.4	0.1	0.0		√
ATRX	0.2	0.0	0.0	1.9	0.0	0.0		√
GPC3	0.2	0.0	0.0	1.9	0.0	0.0		√
BCOR	0.2	0.0	0.0	1.9	0.0	0.3		√
ZIC3	0.2	0.1	0.0	1.9	0.0	0.0		√
FLNA	0.2	0.1	0.0	1.9	0.0	0.0		√
MID1	0.2	0.0	0.3	2.1	0.0	0.3		√

The statistical test applied was the Barnard's exact test. Of our 100 candidate genes, 21 were found to be significantly enriched for CNVs (null hypothesis rejected P ≤ 0.05 in both cohorts: CHD vs. CHOP and CHD vs. MFHS, see boldface). We used the full cohorts for genes in autosomal chromosomes, and only the female portion for any genes on chromosomes (Chr.) X or Y. This leaves 322/880 for MFHS, 416/945 for CHD, and an estimated 1,013/2,026 for CHOP, usable for testing on the Chr. X genes.

Numerous genes were identified as significantly enriched (P ≤ 0.05 against both control cohorts), including losses, FKBP6, ELN, GTF2IRD1, GATA4, CRKL, TBX1, ATRX, GPC3, BCOR, ZIC3, FLNA and MID1, and gains, PRKAB2, FMO5, CHD1L, BCL9, ACP6, GJA5, HRAS, GATA6, and RUNX1. These genes are identified in Table 7.

The authors recognize that syndromic forms of congenital heart disease are relatively well understood; therefore, genes in chromosomal abnormalities known to be causally related to CHD were intentionally kept on the 100 candidate CHD risk gene list to contrast with CNVs found elsewhere. For instance, haploinsufficiency of the genes associated with William's Syndrome, FKBP6, ELN, and GTF2IRD1, identified the three William's Syndrome patients in the study (1). Losses of the TBX1 and CRKL genes are associated with 22qDS and were observed in deleted subjects (32, 53). Turner syndrome subjects carrying losses on the chromosome X genes involving MID1, BCOR, ATRX, GPC3, ZIC3, and FLNA were identified, as well as a female subject (XXX) who was identified with gains over these chromosome X gene regions. In addition, duplications involving RUNX1 were primarily Trisomy 21 subjects.

Gains at 1q21.1 including PRKAB2, FMO5, CHD1L, BCL9, ACP6, and GJA5 were significantly enriched in this study; however, losses that were observed in both control cohorts as well as the CHD cohort were not. Interestingly, gains at 1q21.1 were previously reported in isolated sporadic TOF (18). In our case cohort we observed one subject (case 24) with TOF (2 contiguous CNVs, 0.6 and 1.6 Mb), one subject (case 20) with PA-VSD (1.5 Mb), and another (case 10) with CoA (1.5 Mb). One complex subject (case 2) with AS valvar and Shone's had a shorter gain (418 kb) involving only PRKAB2, FMO5, and CHD1L in conjunction with a 1.8 Mb gain at 5q35.2, which included the NSD1 gene.

Chromosome 8p23.1 deletions involving GATA4 were enriched and have been reported as a cause of complex congenital heart defects and diaphragmatic hernia (51). These included subjects with AVC partial (case 6, 3.8 Mb loss), VSD perimembranous (case 34, 4.5 Mb loss), and ASD-SV (case 5, 304 kb loss).

Three subjects had gains involving the HRAS gene. The first was found in a complex subject with coarctation of the aorta: in addition to a 284 kb duplication involving the HRAS gene the subject had Turner syndrome. The remaining two gains (case 12, 256 kb; case 22, 271 kb) were found in subjects with DILV and subaortic stenosis, respectively (Table 5). Cardiovascular malformations are known to be related to Ras/MAPK pathway syndromes, and previous literature findings have reported associations of HRAS mutations in Costello Syndrome and with the subaortic stenosis phenotype (29). These gains involving HRAS appear to expand phenotypes related to the Ras/MAPK pathway.

Enriched CNVs identified in Table 7 are previously reported or can be found in the Database of Chromosomal Imbalance and Phenotype in Humans Using Ensembl Resources (DECIPHER) with the exception of the gains involving GATA6. One of the three gains involving GATA6 was in a subject with Trisomy 18 with TOF. The remaining two subjects with CNV gains involving GATA6 were 1) a subject (case 31) with truncus arteriosus with a complex CNV over two CHD genes of interest, a 308 kb gain including GATA6, and a 1.2 Mb 22q11.2 distal deletion involving MAPK1 (losses in the distal region of 22q11.2 have previously been reported in subjects with truncus arteriosus) (2), and 2) a subject (case 35) with VSD perimembranous with two neighboring 6.1 and 6.9 Mb gains involving a gain on GATA6. Although sequence variants in GATA6 have been previously found to be associated with cardiac outflow tract defects (27), these gains have not been reported and suggest possible GATA6 triple sensitivity to conotruncal defects.

Collapsing groups of phenotypes by recognized causal chromosomal abnormalities.

To increase statistical power, a strategy for summing cohorts was employed; subphenotypes associated with T21, 22qDS, and Turner Syndrome (see Tables 4 and 6) were collapsed into three groups, respectively (33). We hypothesized collapsing subphenotypes into genetically related groups would increase power to detect additional related CNVs by phenotype. The three collapsed groups each demonstrated significant enrichment (P ≤ 0.05) of additional CNVs compared with both control cohorts (see Table 8 - Enriched Syndrome Genes and Fig. 6 - Spectra). Large, rare CNVs were significantly more frequent (P ≤ 0.05) in the groups of T21 subphenotypes and included gains involving GATA6 and RUNX1 and losses involving GATA4, SOX7, TBX1, and CRKL. Likewise, collapsing the HLHS, CoA, and AS (valvar) subphenotypes, which made up the Turner syndrome group, indicated significant gains involving the 1q21.1 gene regions, enriched losses involving the Chr. X genes, as well as gains involving GATA4, SOX7, EHMT1 (case 15), and HRAS and losses involving FOXC1 (case 3) and NOTCH1 (case 11). Although the T21 and 22qDS subclasses share some overlap of phenotypes (other cardiac, TOF, vascular ring/PA sling, and VSD perimembranous), it is interesting to note that the 22qDS grouping also included gains involving the 1q21.1 genes as well as GATA6 and RUNX1. Significant CNV losses within the 22qDS subclasses involved TBX1 and CRKL. All CNVs identified through the collapsed phenotypes are listed in Table 8 and are reported in DECIPHER.

Table 8.

Enriched syndrome genes

		22q Like				T21 Like				Turner Like				CHOP		MFHS
Chr.	Gene		Gain, %		Loss, %		Gain, %		Loss, %		Gain, %		Loss, %	Gain, %	Loss, %	Gain, %	Loss, %
1	ACP6	√	0.67		0.33		0.27		0.27	√	0.42		0.42	0.05	0.05	0.00	0.23
1	BCL9	√	0.67		0.33		0.27		0.27	√	0.42		0.42	0.05	0.05	0.00	0.23
1	CHD1L	√	0.67		0.33		0.27		0.27	√	0.84		0.42	0.05	0.05	0.00	0.23
1	FMO5	√	0.67		0.33		0.27		0.27	√	0.84		0.42	0.05	0.05	0.00	0.23
1	GJA5	√	0.67		0.33		0.27		0.27	√	0.42		0.42	0.05	0.05	0.00	0.23
1	PRKAB2	√	0.67		0.33		0.27		0.27	√	0.84		0.42	0.05	0.05	0.00	0.23
6	FOXC1		0.00		0.00		0.00		0.00		0.00	√	0.42	0.00	0.00	0.00	0.00
8	GATA4		0.00		0.33		0.00	√	0.55	√	0.42		0.00	0.00	0.00	0.00	0.00
8	SOX7		0.00		0.33		0.00	√	0.55	√	0.42		0.00	0.00	0.00	0.00	0.00
9	EHMT1		0.00		0.00		0.00		0.00	√	0.42		0.00	0.00	0.00	0.00	0.00
9	NOTCH1		0.00		0.00		0.00		0.00		0.00	√	0.42	0.00	0.00	0.00	0.00
11	HRAS		0.00		0.00		0.27		0.00	√	0.42		0.00	0.00	0.05	0.00	0.00
18	GATA6	√	1.00		0.00	√	0.55		0.00		0.00		0.00	0.00	0.00	0.00	0.00
21	RUNX1	√	7.36		0.00	√	22.25		0.00		0.00		0.00	0.00	0.00	0.00	0.00
22	CRKL		0.67	√	13.38		1.10	√	3.57		0.00		0.00	0.00	0.05	0.34	0.00
22	TBX1		0.33	√	14.05		0.55	√	3.85		0.00		0.00	0.00	0.10	0.23	0.00
X	ATRX		0.00		0.00		0.27		0.00		0.00	√	8.99	0.00	0.00	0.00	0.00
X	BCOR		0.00		0.00		0.27		0.00		0.00	√	8.99	0.00	0.00	0.00	0.31
X	FLNA		0.00		0.00		0.27		0.00		0.00	√	8.99	0.10	0.00	0.00	0.00
X	GPC3		0.00		0.00		0.27		0.00		0.00	√	8.99	0.00	0.00	0.00	0.00
X	MID1		0.00		0.00		0.27		0.00		0.00	√	10.11	0.00	0.00	0.31	0.31
X	ZIC3		0.00		0.00		0.27		0.00		0.00	√	8.99	0.10	0.00	0.00	0.00

Boldface indicates significant values.

Fig. 6.

CNV frequency spectra of collapsed phenotypes by syndrome. Note that any gene showing 0% CNV frequency in all 3 cohorts was omitted from this figure due to space considerations. Vertical error bars drawn represent 1 SD in the estimated sampling distribution. Turner phenotypes, T21 phenotypes, and 22qDS phenotypes.

Additional findings of note include a gain involving TBX20 and loss involving SALL4. Three losses including TBX20 have been previously reported in subjects with CHD (ASD and VSDs) (26, 38). We identified a subject (case 26) with TOF with a 3.3 Mb gain involving TBX20 and an adjacent 4.7 Mb gain involving HOXA1, which has been reported in DECIPHER. Finally, we report a subject (case 32) with truncus arteriosus with a 1.8 Mb loss over the SALL4 gene, which has not been previously reported. This segment included a loss over NFATC2, a regulator of cardiac transcription factors but was not included in our 100 gene list because likely causal variants have not previously been reported in humans in this gene(9).

Distribution of CNVs by subject.

To characterize CHD study subjects with an approach more typically used in clinical genetics, CNVs were separated by size (whether or not they would be cytogenetically visible) and then the CHD WIKI site was employed to determine if remaining CNVs should be classified as involving a “syndromic” (two or more clinical features) or a “nonsyndromic” gene (1). Cytogenetically visible CNVs (category A) included chromosomal abnormalities ≥3 Mbps. This category contained subjects with Trisomy 21, 18; Turner; and XXX syndrome and represented ∼9% of the CHD cohort. Category B, contributing 6% to the overall CNV distribution, were those subjects with a CNV over a “syndromic-associated” CHD gene as reported by CHD WIKI (1). This subset contained 22qDS subjects (n = 42) with losses over the TBX1 gene, William's Syndrome subjects, all with a phenotype of supravalvar aortic stenosis (n = 3) with losses over the ELN, GTF2IRD1, and FKBP6 genes. The “nonsyndromic” segment (category C) representing 1% of the CHD cohort was also defined by the CHD WIKI portal. Six CHD case subjects, contributing 1% to the total, had a CNV over one of the 100 CHD-associated genes; however, their category was considered unknown. Category E represented individuals with no CNV over our predefined 100 CHD risk gene list. An individual could only fit into one category where D>A>B or C (see Fig. 7).

Fig. 7.

Distribution of CNVs in CHD cohort. Type A represents cytogenetically visible chromosomal abnormalities (≥3 Mbp), type B are those subjects with a CNV over a syndromic-associated CHD gene as reported by the CHD WIKI portal, type C are those recognized through CHD WIKI as nonsyndromic, type D are CNVs with an unknown category, and type E represents subjects with no CNV over our predefined 100 CHD-associated genes. An individual can only fit into 1 category where D>A>B or C. Numbers are rounded to the nearest percentage.

Complex CNVs.

Four basic mechanisms are involved in the generation of a majority of CNVs: deletion, duplication, inversion, and related combinations (56). We were interested if CHD subjects were at increased risk for carrying multiple CNVs. In the current study, 125 CHD subjects were defined as complex (methods). We identified 100 of those with known CHD-associated syndromes (T21, 59; T18, 1; 22qDS, 31; Turner, 6; William's, 2; XXX Syndrome, 1). Of the remaining 25, 24 contained likely causal CNVs for CHD as outlined in Table 5, whereas one subject contained a nonconfirmed CNV over a CHD-associated gene. Three complex subjects had CNVs on different chromosomes over two of our CHD associated genes of interest: subjects 2, 31, and 33 (Table 5). In addition, two subjects from the CHD cohort were both syndromic with their additional CNV over a second gene of interest: a Turner syndrome subject had a gain involving the HRAS gene and a 22qDS subject had an additional CNV involving a gain over the MAPK1 gene.

It is interesting to note that applying the “complex” criteria to the MFHS control cohort also identified 10 subjects from the controls that met the complex analysis requirements. These subjects had gains over the genes FKBP6, MYH11, TERT, TBX1, CRKL, SH3PXD2B, and losses over the 1q21.1 gene region and MID1.

DISCUSSION

CHD is a complex disease with demonstrated genetic etiology in a subset of patients. CNVs, viewed as an evolutionary driving force for new gene function resulting in improved survival and/or adaption to new environments and disease, contribute the largest component of natural human variation between any two individuals; indeed, CNVs contribute significantly more to inter-individual variation than SNPs (35, 39, 41). There is a broad range of CNV lengths. In this study we focused on large CNVs that can be detected with high accuracy and are relatively straightforward to confirm. It has previously been estimated that ∼65–80% of individuals have a large CNV (≥100 kb) and approximately three to seven CNV segments per individual (56). The average number of CNVs per subject in our CHD cohort supports these previous observations (Fig. 3). It is apparent that as CNV data continue to grow, the development of higher-resolution approaches will permit smaller CNV detection with better accuracy. This will potentially lead to additional disease association discoveries (23). However, data suggest that common CNVs (CNPs) are likely to be lower penetrance risk factors, whereas rare CNV variants are more likely to carry highly penetrant disease risk factors (13).

Significant challenges remain in CNV disease-association studies at both the platform and analysis levels (37). The relationship between phenotype and gene dosage is complex. Our study represents a comprehensive data curation and filtering of CNVs involving 100 recognized CHD risk genes detected in a large, anatomically phenotyped CHD population. We employed a strict algorithm to determine frequencies of CNVs involving regions that encompassed these CHD risk genes. The algorithm employed was very similar to a recent recommendation by Breckpot et al. (6) for determining if CNVs detected in CHD patients are clinically relevant; herein we performed a comparison against two different control populations and an analysis primarily based on known chromosomal abnormalities and gene content rather than more commonly used CNV detection approaches that prioritize by size. CNVs over these predefined gene regions were then used to search for relationships between cardiac phenotype and gene dosage.

The novel analytical approach described herein identified known causal chromosomal abnormalities (including T21, T18, 22qDS, Turner, William's, and XXX Syndromes), which represent 14% of CHD subjects in this study, similar to previous observations (36). Overall, this descriptive study suggests that (after excluding well-established causal chromosomal abnormalities) large, rare CNVs in 100 well-defined CHD risk genes confers significant risk of CHD and is likely etiologic in 4.3% of CHD cases, similar to previous observations (6). Cardiac subphenotypes showing the most significant (P ≤ 0.05) enrichment of large CNV events were aortic stenosis (valvar), AV canal (partial), AVSD with TOF, subaortic stenosis, TOF, and truncus arteriosus. CNV frequency spectra analysis identified enriched genes (P ≤ 0.05): losses: FKBP6, ELN, GTF2IRD1, GATA4, CRKL, TBX1, ATRX, GPC3, BCOR, ZIC3, FLNA, and MID1; and gains: PRKAB2, FMO5, CHD1L, BCL9, ACP6, GJA5, HRAS, GATA6, and RUNX1. 1q21.1 gains were enriched in subjects with conotruncal defects and coarctation of the aorta. 8p23.1 losses were enriched in subjects with septal defects and gains involving HRAS were observed in subaortic stenosis and DILV. Cardiovascular malformations are known to be related to Ras/MAPK pathway syndromes and previous literature findings have reported associations of HRAS mutations resulting in increased hRAS signaling with the subaortic stenosis phenotype. Other common phenotypes occurring in patients with hRAS mutations (also known as Costello syndrome) are cardiac hypertrophy (usually typical hypertrophic cardiomyopathy) and arrhythmia (usually supraventricular tachycardia, especially chaotic atrial rhythm/multifocal atrial tachycardia or ectopic atrial tachycardia) (43). Although DILV sometimes are associated with pulmonary stenosis we have not found any previous reports of hRAS mutations linked to this phenotype. Thus, our data appear to expand phenotypes related to the Ras/MAPK pathway.

We hypothesized that CNV frequency spectra combined with detailed anatomic classes would define the impact of gene dosage in etiologic molecular pathways. One set of clues when searching for genetic causes of CHD is given by the enrichment of CHD cases in various recognized causal chromosomal abnormalities such as T21, T18, 22qDS, Turner syndrome, and William's Syndrome (36). For instance, three of four (75%) subjects in our study with supravalvar aortic stenosis had deletions involving FKBP6, ELN, and GTF2IRD1 genes; all of these subjects had William's syndrome. In Turner syndrome, the incidence of CHD can be as high as 50% and include phenotypes such as BAV, CoA, ASD-VSD partially anomalus pulmonary vena cava, and HLHS, but these data vary (4). The specific cause for CHD in patients with Turner syndrome is currently unknown; several genes have been implicated but for the most part do not quite match Turner syndrome phenotypes or have only been associated with the syndrome by animal models (4, 5). In the present study, eight Turner syndrome subjects were easily identified from the total CHD cohort by CNV frequency spectra analysis. In these subjects, copy number losses were present on all six of the Chr. X genes (MID1, BCOR, ATRX, GPC3, ZIC3, and FLNA) that were selected as CHD-associated from our list of 100 genes. Two out of eight Turner cases in the study had HLHS, five had coarctation of the aorta, and one had aortic stenosis (valvar). Turner syndrome-associated phenotype percentages for the CHD cohort were in good agreement with published reports (4).

To test for additional CNV gene enrichment with increased power, subphenotypes associated with T21, 22qDS, and Turner Syndrome were collapsed in these groups, respectively. The three collapsed groups of subphenotypes each demonstrated enrichment (P ≤ 0.05) in additional CNVs compared with both control cohorts. Large, rare CNVs significantly increased (P ≤ 0.05) in the groups of T21 subphenotypes included gains over GATA6 and RUNX1 and losses over GATA4, SOX7, TBX1 and CRKL. Likewise, collapsing the HLHS, CoA, and AS (valvar) subphenotypes that made up the Turner syndrome group indicated significant gains over the 1q21.1 gene regions, enriched losses over the Chromosome X genes, as well as gains over likely etiologic genes such as GATA4, SOX7, EHMT1, and HRAS and losses over FOXC1 and NOTCH1. Although the T21 and 22qDS collapsed groups share some overlap of phenotypes, it is interesting to note that the 22qDS grouping also included gains over the 1q21.1 genes, as well as GATA6 and RUNX1. Significant CNV losses within the 22qDS subclasses were over TBX1 and CRKL. Incorporating gene dosage with detailed phenotyping into current molecular cardiogenesis models may allow future models of development to fine tune and increase our understanding of etiologic pathways.

By narrowing our focus on a select set of 100 well-known CHD risk genes, we limited the study by design. We focused on large rare CNVs in the current study; therefore, smaller CNVs have not yet been examined. Furthermore, the number of subjects per phenotype was small because of detailed anatomic groupings; collapsing into fewer groups (with larger n) according to developmental models would increase power and may permit identification of additional enriched genes. An additional foreseeable limitation was that CNVs may have been enriched in genes, but because the analysis required statistical significance with two control cohorts (where CNVs were not confirmed and may have been inflated and manifested as false positives), the study may not have been sufficiently powered to detect smaller but true differences.

To our knowledge, this is the first paper to curate a large and diverse CHD population with regard to subphenotype and CNV frequency by gene region. This appears to be a useful approach to visualize and eventually, given sufficient numbers, to quantify relative risk of CNVs for specific subphenotypes. Broadening to encompass the entire genome and performing the copy number spectra analysis at higher resolution should identify additional candidate genes in CHD. The ability to quantify risk of particular cardiac malformations by gene dosage should offer insight into critical molecular pathways impacted during human cardiogenesis. Furthermore, overlaying CNV data and details of resulting cardiac phenotype with known functional pathways of cardiogenesis should lead to increased understanding of the molecular etiology of heart malformations.

GRANTS

This study was supported in part by National Institutes of Health (NIH) Grant NIH1R21HD-060309-01 (A. Tomita-Mitchell), the CRI (A. Tomita-Mitchell and M. E. Mitchell), the Department of Surgery (A. Tomita-Mitchell, J. S. Tweddell, M. E. Mitchell), and in part by NIH Grant NIH5R01HL-089655 (U. Broeckel).

DISCLOSURES

A. Tomita-Mitchell, M. E. Mitchell, and C. A. Struble have conflicts of interest (significant financial interest in Aria Diagnostics, a molecular diagnostics company). M. Hidestrand and M. A. Goetsch have conflicts of interest (financial interest in Aria Diagnostics, a molecular diagnostics company).

AUTHOR CONTRIBUTIONS

Author contributions: A.T.-M., D.K.M., S.E.H., D.P.B., U.B., A.N.P., J.S.T., and M.E.M. conception and design of research; A.T.-M., D.K.M., C.A.S., M.E.T., K.D.S., M.H., S.E.H., M.A.G., P.M.S., D.P.B., U.B., A.N.P., and M.E.M. analyzed data; A.T.-M., D.K.M., C.A.S., M.E.T., K.D.S., M.H., P.M.S., D.P.B., A.N.P., J.S.T., and M.E.M. interpreted results of experiments; A.T.-M., D.K.M., M.E.T., K.D.S., M.A.G., and P.M.S. prepared figures; A.T.-M., D.K.M., M.E.T., M.H., P.M.S., and M.E.M. drafted manuscript; A.T.-M., D.K.M., C.A.S., M.E.T., K.D.S., M.H., S.E.H., M.A.G., P.M.S., D.P.B., U.B., A.N.P., and M.E.M. edited and revised manuscript; A.T.-M., D.K.M., C.A.S., M.E.T., K.D.S., M.H., S.E.H., M.A.G., P.M.S., D.P.B., U.B., A.N.P., J.S.T., and M.E.M. approved final version of manuscript; D.K.M., D.P.B., and U.B. performed experiments.