Next-generation sequencing identified genetic variations in families with fetal non-syndromic atrioventricular septal defects

Abstract

Atrioventricular septal defects (AVSDs) account for approximately 5% of all congenital heart disease (CHD). About half of AVSDs are diagnosed in cases with trisomy 21 (Down’s syndrome, DS). However, many AVSDs occur sporadically and manifest as non-syndromic. The pathogenesis is complex and has not yet been fully elucidated. In the present study, we applied two advanced applications of next-generation sequencing (NGS) to explore the genetic variations in families with fetal non-syndromic AVSDs. Our study was mainly divided into two steps: (1) low-pass whole-genome sequencing (WGS) was used to detect the genome-wide copy number variations (CNVs) for included subjects; (2) whole-exome sequencing (WES) was used to detect the gene mutations for the subjects without AVSD-associated CNVs. A total of 17 heterozygous de novo CNVs and 19 heterozygous de novo gene mutations were selected, and 15 candidate genes were involved in these variations. Among these heterozygous de novo variations, most have potential pathogenicity for AVSDs, but the others require further investigation to confirm their pathogenicity. Our study not only shows the genetic diversity and the etiological complexity of AVSDs but also shows the rationality and practicability of this sequential genetic detection and analysis strategy.

Introduction

Congenital heart disease (CHD) is the most common congenital malformation seen at birth as well as the most common congenital defect contributing to death in the first year. The prevalence of CHD is approximately 8 per 1000 births worldwide [1]. Atrioventricular septal defects (AVSDs) or atrioventricular (AV) canal defects account for approximately 5% of all CHD [2,3]. This group of defects is caused by abnormal development of endocardial cushions. Meanwhile, endocardial cushions are involved in the formation of the atrial septum, the ventricular septum, and the mitral and tricuspid valves during embryonic development. Therefore, AVSDs are manifested by varying degrees of AV valvular and septal abnormalities, including atrial septal defect (ASD) and ventricular septal defect (VSD), and are classified as “partial, intermediate, or complete”. AVSDs are often associated with other cardiac defects, such as Tetralogy of Fallot (TOF), double outlet right ventricle, and transposition of the great arteries [4,5]. The clinical presentation and prognosis in AVSDs depend on the specific morphology of the defects and the associated anomalies. Untreated patients with AVSDs may present with cyanosis, breathlessness, recurrent respiratory infection, growth retardation, variable heart murmur, or even congestive cardiac failure, pulmonary hypertension, and death in early life [3,4]. Many parents who are diagnosed with fetal AVSDs may choose to terminate the pregnancy to reduce the economic and psychological burdens of their families. At the same time, some parents may seek genetic counseling to assess the genetic defects of the malformed fetuses and the risk of recurrence in their next pregnancies.

About half of AVSDs are diagnosed in cases with trisomy 21 (Down’s syndrome, DS) [6,7]. Deletions on chromosome 21 on a trisomic background may reduce the risk for AVSDs [8]. Some genes may act as susceptibility factors for AVSDs in DS patients, such as CRELD1 gene [9]. However, many AVSDs occur sporadically and manifest as non-syndromic. The pathogenesis is complex and has not yet been fully elucidated.

In recent years, a relationship between subchromosomal anomalies and CHD has been strongly suggested [10]. These subchromosomal anomalies are known as copy number variations (CNVs) and defined as copy number changes, including deletions, duplications, or multiallelic variation events of genomic regions ranging from 1 kilobase (Kb) to several megabases (Mb). CNVs can be identified using chromosomal microarray analysis (CMA), which is based on gene chip technology and limited by probe spacing and density. Recently, several studies have demonstrated the possibility of using low-pass whole-genome sequencing (WGS) to detect CNVs [11,12]. Low-pass WGS is an application of next-generation sequencing (NGS) that can detect genome-wide CNVs, even those beyond the probe’s range of CMA [13]. However, there is very little research on the detection of CNVs in AVSD cases without DS [14]. As another approach of NGS, whole-exome sequencing (WES) has been more and more used to explore the gene mutations of some diseases. However, this research is only just beginning for AVSDs [15].

In our study, we applied NGS to explore the genetic variations in fetuses with non-syndromic AVSDs and normal chromosome karyotypes. Our study was mainly divided into two steps: (1) Low-pass WGS was used to detect the genome-wide CNVs for included subjects; (2) WES was used to detect the gene mutations for the subjects without AVSD-associated CNVs. To exclude benign family genetic factors and to analyze the sources of the meaningful genetic variations, we applied family study, and the same steps were completed on the healthy parents.

Materials and methods

Subject enrollment

The study subjects were fetuses with non-syndromic AVSDs diagnosed by fetal echocardiography and confirmed by post-mortem autopsy in Beijing Obstetrics and Gynecology Hospital, China. Fetuses with identified chromosomal karyotype abnormalities or extracardiac malformations were excluded. Umbilical cord blood samples were collected from prenatal samples, and fetal tissues were collected from abortuses. Meanwhile, peripheral blood samples were collected from the parents. All samples sent to the MyGenostics medical laboratory (Beijing, PRC) for analysis. The study was approved by the ethics committee of the hospital. Informed consent for storage and subsequent analysis was obtained from all parents.

DNA library construction

The RelaxGene Blood DNA System (Tiangen Biotech, Beijing, PRC) and the Universal Genomic DNA Kit (CWBiotech, Beijing, PRC) were used to extract genomic DNA from the blood and tissue samples, respectively. The quality and concentration of the genomic DNA were evaluated by Nanodrop 2000 (Thermo Fisher, MA, USA). The ratio of A260/280 was between 1.8 and 2.0, and the concentration was greater than 30 ng/μL. The genomic DNA was broken into fragments of 100-500 base pairs (Bp) using the Covaris S220 DNA sonication system (Covaris, MA, USA). The fragments were end-polished, adenylated, and ligated with adaptors in turn. Proper reaction systems and cycles of polymerase chain reaction (PCR) amplification were carried out using the GeneAmp PCR System 2720 (Applied Biosystems, CA, USA) for enrichment of ligated DNA fragments. All enzymes and buffers were from MyGenostics (Baltimore, MD, USA). All operations were carried out according to the manufacturers’ recommendations. The final DNA library products were quantitatively detected using NanoDrop 2000 and 1% agarose gel electrophoresis. The concentration of normal DNA library products was greater than 30 ng/μL, and the ratio of A260/280 was between 1.8 and 2.0. The main bands of the DNA library fragments were about 280-400 Bp.

Low-pass WGS and data analysis

NGS was carried out on the Nextseq 500 system (Illumina, CA, USA) to generate 150 Bp paired-end reads (a target depth of 0.6×) for each prepared DNA library according to the manufacturer’s recommendations. Reads were aligned to the National Center for Biotechnology Information human reference genome build 37 (HG19) using Burrows-Wheeler Aligner (version 0.7.10) [16]. Quality control and removal of duplicated reads were carried out using Picard (picard-tools-1.119). Finally, the mapped reads were produced. The exact CNV breakpoint sequences were calculated using the binary segmentation algorithm to determine candidate CNV regions and the copy ratio. A CNV was defined as a deletion or a duplication when its average copy ratio did not exceed 0.75 or was not less than 1.25, respectively. To assess the clinical importance of the detected fetal CNVs and the potential relationship with AVSDs, we selected the CNVs containing the AVSD-associated genes (described in detail below). Finally, the selected CNVs were compared to the databases of known pathogenic or likely pathogenic variations and the general population databases of CNVs (Database of Genomic Variants, DGV) [11,17].

WES and bioinformatics analysis

For exome capture of the prepared DNA library, a GenCap Enrichment Kit (Baltimore, MD, USA) was used according to the manufacturer’s recommendations. NGS was performed using the Nextseq 500 system (Illumina, CA, USA) to generate 150 Bp paired-end reads and cover at least 98% of the exome (an average depth of 200×) for each sequenced sample. A Burrows-Wheeler Aligner was used to align the raw data to HG19 and Picard was used to sort and mark the duplicated reads. Then, local realignment, base quality score recalibration, single nucleotide polymorphism calling, and short insertion/deletion calling were performed using the Genome Analysis Toolkit (version 3.7) software tools [18]. Variants were first prioritized based on their frequency in the 1000 Genomes Project (1000 g 2015aug_all), Exome Sequencing Project (ESP6500, ExAC_ALL, ExAC_EAS) and an inhouse database of 800 healthy Chinese Han adults, with rare (minor allele frequency < 0.05) variants receiving priority [19,20]. Variants in AVSD-associated genes (described in detail below) were selected for further analysis and annotated by different bioinformatics tools: The Sorting Intolerant Form Tolerant (SIFT), PolyPhen-2, Mutation Taster, GERP++ [21-24].

AVSD-associated gene list

In order to identify potential candidate CNVs and gene mutations associated with AVSDs, we compiled a list of 375 human genes with a putative role in the development of AVSDs using Phenolyzer. We used the disease or phenotype terms “heart septal defect”, “heart ventricular septal defect”, “heart atrial septal defect”, “atrioventricular canal/septal defect”, “endocardial cushion defect”, and selected the “seed genes” sorted by Phenolyzer (Supplementary Table 1) [25]. Also, we added 21 other genes by consulting the related published literature (including human and animal studies) on the candidate genes associated with AVSDs (Supplementary Table 2).

Variation validation

The selected CNVs were validated using quantitative real-time PCR (qPCR), and amplification levels were calculated with the 2-ΔΔCT method. The selected gene mutations were validated using Sanger sequencing. Primer pairs were designed by the Realtime PCR tool from Integrated DNA Technology or Primer3 (v.0.4.0), verified by primer BLAST or UCSC In-Silico PCR. The SYBR Premix Ex Taq II PCR reagent kit (TaKaRa Bio, Dalian, PRC) was used for qPCR reactions, and the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, CA, USA) was used for the Sanger sequencing reactions. All operations were done according to the manufacturers’ recommendations.

Results

Basic characteristics of the study subjects

We recruited 50 non-syndromic AVSD families from Beijing Obstetrics and Gynecology Hospital in China; each family comprised one non-syndromic AVSD fetus and two healthy parents. All the couples were non-consanguineous and terminated the pregnancy at midterm. All the fetuses had normal chromosomal karyotypes, and were without extracardiac malformations. The phenotypic characteristics of these AVSD fetuses are presented in Table 1.

Table 1.

Phenotypic characteristics of the 50 AVSD fetuses

Sex	n (%)
Male	28 (56.0)
Female	22 (44.0)
AVSD type	n (%)
Partial	21 (42.0)
Intermediate	11 (22.0)
Complete	18 (36.0)
Associated cardiac defects	n (%)
Yes	27 (54.0)
No	23 (46.0)

Abbreviations: AVSD, atrioventricular septal defect.

CNVs detected by low-pass WGS

In total, 1,736 CNVs were detected from the 50 AVSD fetuses. Seventeen de novo CNVs containing 10 AVSD-associated genes (candidate genes) were selected from these CNVs. These CNVs were derived from 17 AVSD fetuses, and none of them were carried by the healthy parents and included in the DGV. These CNVs ranged from 127 Kb to 4.29 Mb. Two CNVs larger than 1 Mb were the known pathogenic CNVs (pCNVs) for chromosome 22q11.2 deletion syndrome and chromosome 22q13.3 deletion syndrome respectively (Supplementary Figure 1). The others were smaller than 1 Mb and validated by qPCR (Supplementary Table 3). All the 17 CNVs contained whole or a part of the exons of their candidate genes (Table 2).

Table 2.

De novo CNVs containing AVSD-associated genes

Candidate gene	Fetus ID	AVSD type	Associated cardiac defects	Cytoband	Start-end	Length	Type
NOTCH2	4	Partial	TA, PA	1p12p11.2	120524783-120904419	379.64 Kb	Dup, het, intragenic
	17	Complete	TGA, PS	1p12p11.2	120563920-120936695	372.78 Kb	Dup, het, intragenic
	45	Intermediate	DORV	1p12p11.2	120597708-120904419	306.71 Kb	Dup, het, intragenic
COL11A1	16	Complete	None	1p21.1	103361276-103582736	221.46 Kb	Dup, het, intragenic
	30	Partial	None	1p21.1	103319157-103743271	424.12 Kb	Del, het, whole gene
	32	Partial	None	1p21.1	103403979-103856289	452.31 Kb	Del, het, intragenic
NIPBL	1	Partial	IAA	5p13.2	36891413-37044984	153.57 Kb	Del, het, intragenic
	8	Intermediate	None	5p13.2	36891312-37054895	163.58 Kb	Del, het, intragenic
EHMT1	12	Complete	TGA, RAA	9q34.3	140203637-141023198	819.56 Kb	Dup, het, whole gene
	15	Intermediate	None	9q34.3	140481413-140707091	225.68 Kb	Del, het, intragenic
NR2F2	9	Complete	TGA, DORV, PS, PA	15q26.2	96777378-96923621	146.24 Kb	Dup, het, whole gene
	21	Complete	None	15q26.2	96786595-97311581	524.99 Kb	Dup, het, whole gene
COL6A1/2	2	Partial	TGA, PA	21q22.3	47389758-47576705	186.95 Kb	Dup, het, whole gene
	49	Partial	None	21q22.3	47378452-47612768	234.32 Kb	Del, het, whole genes
TBX1	38	Complete	TGA, DORV, PS	22q11.21	18939748-21721712	2.78 Mb	Del, het, whole gene
SHANK3	5	Complete	CAT	22q13.31q13.33	46933489-51219152	4.29 Mb	Del, het, whole gene
SMC1A	29	Partial	AS	Xp11.22	53363770-53490937	127.00 Kb	Dup, het, whole gene

Abbreviations: CNV, copy number variation; AVSD, atrioventricular septal defect; TA, tricuspid atresia; PA, pulmonary valve atresia; TGA, transposition of the great arteries; PS, pulmonary stenosis; DORV, double outlet right ventricle; IAA, interruption of aortic arch; RAA, right aortic arch; CAT, common arterial trunk; AS, aortic stenosis; Dup, duplication; Del, deletion; het, heterozygous.

We retrieved the 10 candidate genes in DECIPHER. Seven genes (NOTCH2, NIPBL, EHMT1, NR2F2, TBX1, SHANK3, SMC1A) are contained in the CNVs (including deletions and duplications) detected in patients with septal defects (including AVSD, ASD, VSD) (Supplementary Table 4). In our study, 8 de novo deletions and 9 de novo duplications were selected. Among these duplications, 5 duplications contained the whole candidate genes, and the others contained part of exons of the candidate genes (intragenic duplications).

Gene mutations detected by WES

Thirty-three AVSD fetuses and their healthy parents were included for WES. A total of 6,713 high-quality, rare, and nonsynonymous variants were detected from these AVSD fetuses, and there were 138 variants in the AVSD-associated genes. Among them, 7 candidate genes (C5ORF42, COL11A1, COL6A2, GATA6, GLI3, HSPG2, LRP2) were relatively enriched for de novo variants at least 2 AVSD fetuses carried the de novo variants in the same gene). Nineteen de novo heterozygous variants in these genes were selected (Table 3), and these variants were derived from 14 AVSD fetuses. All variants were validated by Sanger sequencing (Supplementary Table 5).

Table 3.

Rare nonsynonymous de novo variants in 7 AVSD-associated genes

Fetus ID	AVSD type	Associated cardiac defects	Candidate gene	Nucleotide changes	Amino acid changes	dbsnp147	Damaging predict*	GERP++
23	Partial	None	COL11A1	652-5->TT	Splicing	rs749687230	-	-
40	Complete	TGA	COL11A1	3266C>T	P1089L	rs373734529	Yes	Conserved
13	Complete	CAT	COL6A2	499G>A	G167S	rs115957676	Yes	Conserved
3	Partial	None	COL6A2	679G>A	D227N	rs35881321	No	Conserved
43	Intermediate	TGA, PS, AS	COL6A2	2798G>A	R933H	rs374384263	Yes	Nonconserved
20	Partial	TGA, PA	C5ORF42	8746G>A	A2916T	rs369585190	Yes	Conserved
25	Partial	None	C5ORF42	6443A>G	N2148S	rs150999024	No	Nonconserved
28	Complete	TGA, PS	C5ORF42	608A>G	Y203C	rs144969169	Yes	Conserved
3	Partial	None	GLI3	169G>A	A57T	rs775586921	No	Conserved
3	Partial	None	GLI3	164G>A	R55K	rs764332121	Yes	Conserved
20	Partial	TGA, PA	GLI3	169G>A	A57T	rs775586921	No	Conserved
20	Partial	TGA, PA	GLI3	164G>A	R55K	rs764332121	Yes	Conserved
7	Complete	TGA, PS	LRP2	9937G>A	D3313N	-	Yes	Conserved
41	Complete	TGA	LRP2	9914G>A	R3305H	rs3213760	Yes	Conserved
10	Intermediate	None	GATA6	43G>C	G15R	rs116262672	Yes	Conserved
44	Intermediate	DORV, TAPVC	GATA6	551G>A	S184N	rs387906816	No	Nonconserved
14	Partial	TGA, IAA	HSPG2	2008G>A	V670I	rs147810145	No	Conserved
27	Intermediate	None	HSPG2	10589G>A	R3530Q	rs200062985	Yes	Conserved
44	Intermediate	DORV, TAPVC	HSPG2	2057T>C	L686P	-	Yes	Conserved

Abbreviations: AVSD, atrioventricular septal defect; TGA, transposition of the great arteries; PA, pulmonary valve atresia; PS, pulmonary stenosis; CAT, common arterial trunk; AS, aortic valve stenosis; DORV, double outlet right ventricle; TAPVC, total anomalous pulmonary venous connection.

^*Yes: at least 2 bioinformatics tools suggest damaging or probably damaging or possibly damaging (SIFT, PolyPhen-2, Mutation Taster); No: 2 or 3 bioinformatics tools suggest benign (SIFT, PolyPhen-2, Mutation Taster).

Two genes, COL11A1 and COL6A2, were contained in de novo CNVs derived from another 3 and 2 AVSD fetuses, respectively. In the WES group, 2 AVSD fetuses had rare nonsynonymous variants in the COL11A1 gene; one was an exonic splicing variant, and the other (P1089L) was highly conserved and predicted to be damaging. Three AVSD fetuses had rare nonsynonymous variants in the COL6A2 gene, 2 variants (G167S and R933H) were predicted to be damaging.

Two AVSD fetuses had the same highly conserved, de novo, compound heterozygous mutations in GLI3 gene; however, only one variant (R55K) was predicted to be damaging. The other variant (A57T) was predicted to be benign.

Three AVSD fetuses had rare nonsynonymous variants in the HSPG2 gene, and 2 variants (R3530Q and L686P) were highly conserved and predicted to be damaging, and one of them (L686P) was a novel variant. Three AVSD fetuses had rare nonsynonymous variants in the C5ORF42 gene, and 2 variants (A2916T and Y203C) were highly conserved and predicted to be damaging.

Rare nonsynonymous variants in another 2 genes (GATA6 and LRP2) were carried by 2 AVSD fetuses for each gene. Except for one de novo variant (S184N) in the GATA6 gene, the other 3 de novo variants were highly conserved and predicted to be damaging. Among the 4 variants, one variant (D3313N) was novel.

Discussion

Embryologically, human cardiac septation takes place in the first 8 weeks of pregnancy. After primary heart tube looping, endocardial cushions (superior, inferior, and two lateral cushions) are formed at the AV junction as a result of a critical process, endothelial to mesenchymal transition. Subsequently, the two lateral endocardial cushions develop and divide the AV canal into two separate AV orifices and contribute to the formation of the mitral valve and tricuspid valve. A deficiency in these processes will lead to a common AV annulus and a common AV valve. Meanwhile, the superior and inferior endocardial cushions extent and close the atrial septum primum and the interventricular foramen, but a deficiency in these processes will lead to an ostium primum defect and an inlet VSD just below the AV valves (membranous VSD). In partial AVSD, there is an isolated ostium primum defect or an inlet VSD, and two separate AV orifices and AV valves. In complete AVSD, besides an isolated ostium primum defect and an inlet VSD, there is a common AV annulus and a common AV valve. Intermediate AVSD refers to the situation between the partial type and complete type, in which there is an atrial septum primum and an inlet VSD, but two separate AV orifices [26,27].

This study was designed to detect the genetic variations associated with non-syndromic AVSDs. To cover the meaningful variations as far as possible, we used two applications of NGS to achieve it, low-pass WGS for the genome-wide CNVs, and WES for the gene mutations. NGS is an advanced technology used to detect genetic variations with unprecedented resolution. Although the application of low-pass WGS is not widely used for CNV detection, it was confirmed to have an equivalent effectiveness for detection of pCNVs compared with CMA, and besides, it can detect CNVs beyond the probe’s range of CMA [11]. The specificity of detected deletions and duplications larger than 100 Kb was 100%, even using a read depth of 0.2× [28]. In our study, we chose a more accurate read depth (0.6×) to detect CNVs. We finally selected 17 de novo CNVs containing AVSD-associated genes, and all small CNVs (larger than 100 Kb but smaller than 1 Mb) were validated by qPCR, with a very high credibility. WES is a cost-effective, high-depth DNA sequencing strategy to detect DNA variations in the coding regions that may alter protein function. Not only can it detect common variations, it can also find low frequency variations, and rare variations. In our study, we used WES to detect gene mutations with an average sequencing depth of 200× and finally selected 19 de novo, high-quality, rare, and nonsynonymous variants in 7 AVSD-associated genes.

In the low-pass WGS group, 10 AVSD-associated genes were involved in 17 de novo CNVs derived from 17 AVSD fetuses. All CNVs contained the whole or a part of the exons of their candidate genes, causing the dosage changes of the genes or functional changes of the proteins. Among these CNVs, there were 8 deletions and 9 duplications (4 duplications were intragenic duplications). In humans, deletion (such as COL11A1, NIPBL, EHMT1, COL6A1, COL6A2, TBX1, SHANK3 gene in our study) can lead to haploinsufficiency and a loss-of-function change of an important gene, and this is very similar to those caused by heterozygous mutations within the coding region of the gene. Duplication of the whole gene (such as EHMT1, NR2F2, COL6A1, COL6A2, SMC1A in our study) can cause triplication of the gene that could cause a similar but milder clinical phenotype resulting from the deletion [29]. However, intergenic duplication (such as the COL11A1, NOTCH2 gene in our study) may lead to gene disruption or fusion, resulting in loss of gene function, and then cause a similar clinical phenotype to the deletion [30]. Except for 3 genes (COL11A1, COL6A1, COL6A2), the other 7 genes are contained in the CNVs detected in patients with septal defects (including AVSD, ASD, VSD) in DECIPHER. Also, 8 genes (NOTCH2, COL11A1, NIPBL, EHMT1, NR2F2, TBX1, SHANK3, SMC1A) are dominant pathogenic genes, and heterozygous or haploinsufficient variations of these genes can cause syndromes which have CHD phenotypes, including AVSD, ASD and VSD, suggesting a potential relationship between these CNVs with phenotypes. Seven genes (NOTCH2, COL11A1, NIPBL, NR2F2, COL6A1, COL6A2, TBX1) have been thought to play a role in the normal development of the AV canal, endocardial cushions, or AV valves, according to some molecular studies and animal models. Mutations in these genes could cause AVSD formation (Table 4) [31-42]. The other 3 genes (EHMT1, SHANK3, SMC1A) have not been reported to play a direct role in heart development or CHD formation, but the variations of them have been reported in some AVSD cases, suggesting the need for more research in this area [34,36,37].

Table 4.

The association of gene variations with AVSD

Candidate gene (OMIM ID)	CHD-associated syndromes caused by heterozygous or haploinsufficient variations	Association with AVSD
NOTCH2 (600275)	Alagille syndrome 2; Hajdu-Cheney syndrome	Chick Notch2 initiates the signaling cascades that delimits the non-chamber AV canal regions, causes the progressive restriction of Bmp2 and Tbx2 expression to within the developing AV canal [31].
COL11A1 (120280)	Stickler syndrome, type II	Murine Col11a1 can express in AV valve and involved in AV valve development and maintenance [32,33].
NIPBL (608667)	Cornelia de Lange syndrome 1	30% Cornelia de Lange syndrome patients have CHD, including AVSD, ASD, VSD [34]. Nipbl ± mice can exhibit the phenotypes of Cornelia de Lange syndrome 1, septal defects were especially common [35].
EHMT1 (607001)	Kleefstra syndrome 1	41% Kleefstra syndrome patients have CHD, including VSD, ASD; EHMT1 de novo mutation was reported in an AVSD patient [36,37].
NR2F2 (107773)	Congenital heart defects, multiple types, 4	Nr2f2 is expressed in the endocardium and the epicardium; Nr2f2 mutant mice exhibit a spectrum of cardiac defects (including AVSD) resulting from the disruption of endocardial cushion development in a dosage-sensitive fashion [38]. Rare variants in NR2F2 gene were reported in AVSD patients [39].
COL6A1 (120220)	-	Collagen VI is expressed in the AV cushions in human and mouse heart, plays a role in valve and septal differentiation; overexpression or insufficient expression of COL6A1 could cause AVSD formation [2,40].
COL6A2 (120240)	-	Collagen VI is expressed in the AV cushions in human and murine heart, plays a role in valve and septal differentiation; overexpression or insufficient expression of COL6A2 could cause AVSD formation [2,40].
TBX1 (602054)	Chromosome 22q11.2 deletion syndrome	Tbx1 regulates SHF progenitor cell status during heart tube elongation, its failure results in a spectrum of morphological defects affecting the cardiac poles, including AVSD [41,42].
SHANK3 (606230)	Chromosome 22q13.3 deletion syndrome	Patient 253,900 with 86.55 Kb duplication containing SHANK3 gene at 22q13.33 has AVSD in DECIPHER.
SMC1A (300040)	Cornelia de Lange syndrome 2	30% Cornelia de Lange syndrome patients have CHD, including AVSD, ASD, VSD [34].
C5ORF42 (614571)	-	C5orf42 -/- mice exhibit multiple CHD, including AVSD, VSD; its mutation disrupts ciliogenesis and cilia transduced Hedgehog signaling, and the Hedgehog signaling is required in the SHF for AV septation [43,44].
GLI3 (165240)	Pallister-Hall syndrome; Greig cephalopolysyndactyly syndrome	GLI3 is a transcription factor that functions in the Hedgehog signaling [44].
LRP2 (600073)	-	LRP2 acts as a receptor of Hedgehog signaling, Lrp2 -/- mice result in abnormal development of the SHF [45].
GATA6 (601656)	Atrioventricular septal defect 5; Atrial septal defect 9; Tetralogy of Fallot	Gata6 is expressed in the endocardial cushions, atrial and ventricular myocardium, atrioventricular valve leaflets, and a heterozygous missense mutation in the gene was identified in an AVSD patient [49].
HSPG2 (142461)	-	HSPG2 is expressed in the basal surface of myocardium and endocardium, plays a role in the earliest stages of formation of the endocardial cushions [50].

Abbreviations: AVSD, atrioventricular septal defect; CHD, congenital heart disease; AV, atrioventricular; ASD, atrial septal defect; VSD, ventricular septal defect; SHF, second heart field.

In the WES group, 7 AVSD-associated genes were involved in 19 de novo variants derived from 14 AVSD fetuses. Interestingly, the variations of 2 genes (COL11A1 and COL6A1) were detected both in the low-pass WGS group and the WES group, and there were total of 10 variations (including CNVs and gene mutations), suggesting the important roles of collages in heart development [2,32,33,40]. Notably, the COL6A1 gene is mapped to the DS’s obligate region of chromosome 21, the same as the COL6A2 gene, and AVSD is a common feature of DS. These 2 genes encode the collagen VI a1 and a2 chains, respectively. The collagen VI a3 chain is encoded by the COL6A3 gene which is located at chromosome 2. Normally, these 3 chains are assembled in a 1:1:1 stoichiometric ratio. Overexpression or insufficient expression of one gene could result in an inappropriate collagen VI chain secretion and a functional abnormality of collagen VI, and may have a role in the pathogenesis of AVSDs [2,40]. Three genes (C5ORF42, GLI3, LRP2) are involved in hedgehog signaling, and hedgehog signaling is required in the second heart field (SHF) [43-45]. Molecular events (such as Hedgehog signaling, BMP signaling, and T-box gene family signaling) in the SHF cardiac progenitors, which are located dorsal to the primary heart tube, can drive the processes of heart tube elongation and AV septation [31,41,42,44-46]. Failure of these processes could result in a spectrum of morphological defects affecting the cardiac poles, including outflow tract defects and AVSDs [41]. In our study, a total of 9 de novo variants were detected in the C5ORF42, GLI3 and LRP2 genes, and most of them were predicted to be damaging and highly conserved. Among these heterozygous variants, 2 AVSD fetuses had the same compound heterozygous mutations (R55K and A57T) in the GLI3 gene. Although only one variant (R55K) was predicted to be damaging, GLI3 is a dominant pathogenic gene for Pallister-Hall syndrome and Greig cephalopolysyndactyly syndrome, both of which have CHD phenotypes, and heterozygous mutations in the GLI3 gene may be associated with AVSDs [47,48]. Both the C5ORF42 and LRP2 genes are recessive pathogenic genes, and the contribution of the heterozygous variants in the two genes to AVSD phenotypes is uncertain and needs further study. Another 2 AVSD-associated genes, GATA6 and HSPG2, were involved in 5 heterozygous de novo variants. Both of them play a role in the development of endocardial cushions [49,50]. The GATA6 gene has been identified as a dominant pathogenic gene for multiple CHD, including AVSD5, ASD9, and TOF. The heterozygous variants in GATA6 gene are likely to be the cause of the fetal phenotypes. Although haploinsufficient variations of HSPG2 gene have been considered a possible cause of heart defects in patients with chromosome 1p36 deletion syndrome, heterozygous mutant mice did not exhibit significant heart defects [50-52]. The potential pathogenicity for AVSDs of the heterozygous variants in HSPG2 gene is not yet clear.

In this study, we applied NGS to explore the genetic variations in 50 non-syndromic AVSD families. For a more comprehensive exploration of genetic variations associated with non-syndromic AVSDs, we conducted an advanced detection and analysis strategy. First, we applied family study which was helpful in finding out the meaningful de novo genetic variations deriving from the AVSD fetuses, and in better understanding the potential causes of these sporadic, non-syndromic AVSDs. Second, we compiled an AVSD-associated gene list of 396 human genes by retrieving Phenolyzer and by reviewing the literature, and these genes are thought to have a potential relationship with septal defects or AVSDs. The genetic variations in these genes are more likely to be associated with AVSDs. Third, we applied two excellent applications of NGS to detect the genetic variations. Low-pass WGS was used to detect the genome-wide CNVs for 50 non-syndromic AVSD families, and WES was used to detect whole-exome mutations for 33 non-syndromic AVSD families without AVSD-associated CNVs. Both of the two methods are beneficial to the discovery of more meaningful genetic variations. Fourth, we systematically searched the related databases (such as DECIPHER, OMIM) and the published literature to explore the relationship between these candidate genes and AVSDs and to assess the potential pathogenicity of these de novo heterozygous genetic variations. As far as we know, there is no similar study.

There are two important findings from our study. First, it shows the genetic diversity and the etiological complexity of AVSDs. Although half of the AVSDs are associated with trisomy 21, many AVSDs occur sporadically and without a clear cause. So, we chose the fetuses with non-syndromic AVSDs and normal chromosome karyotypes as our study subjects, applied a reasonable and comprehensive strategy to explore the genetic variations associated with the phenotypes in addition to chromosomal karyotype abnormalities. In the low-pass WGS group, we ultimately selected 17 heterozygous de novo CNVs. According to the American College of Medical Genetics standards and guidelines for interpretation and reporting of CNVs, 2 CNVs are associated with the known syndromes, and can be defined as pCNVs; the other CNVs can be defined as likely pCNVs, because the heterozygous/haploinsufficient variations or overexpression of the candidate genes in these CNVs have been reported to be associated with AVSDs or CHD-associated syndromes which have septal defect phenotypes [17]. In the WES group, we finally selected 19 de novo mutations, and all of the candidate genes are important functional genes in the normal development of the heart, especially for endocardial cushions. The heterozygous variants in 4 genes (COL11A1, COL6A2, GLI3, GATA6) have the potential pathogenicity that lead to the occurrence of AVSDs. The pathogenicity of the other mutations is uncertain and needs further study. Second, our study shows the rationality and practicability of this sequential genetic detection and analysis strategy, especially for the diseases with undefined pathogenic mechanisms and genetic bases. In clinical work, when the traditional genetic testing methods (such as chromosomal karyotype analysis) can’t determine the genetic defects associated with diseases, meaningful CNVs and gene mutations should be considered. We can choose some cost-effective detection methods (such as low-pass WGS and WES) to get more genetic information about the diseases, and we can use the related databases and published literature to select the pathogenic or likely pathogenic variations. The strategy can help us to make a more accurate genetic diagnosis, providing a theoretical basis for individualized prenatal diagnosis and genetic counseling.

In summary, we applied two advanced applications of NGS, low-pass WGS and WES, to explore the genetic variations in families with fetal non-syndromic AVSDs. A total of 17 heterozygous de novo CNVs and 19 heterozygous de novo gene mutations were selected by using a sequential genetic detection and analysis strategy. Fifteen candidate genes were involved in these variations, and all of them have demonstrated an association with AVSDs. Among these heterozygous de novo variations, most have potential pathogenicity for AVSDs, but the others require further investigation to define their pathogenicity. The functional validation of these genetic variations wasn’t the focus of our study, and the number of included subjects was somewhat small, so these were the shortcomings of our study to be improved on in the future.

Disclosure of conflict of interest

None.