The transcription factor Sox7 modulates endocardiac cushion formation contributed to atrioventricular septal defect through Wnt4/Bmp2 signaling

Cardiac septum malformations account for the largest proportion in congenital heart defects. The transcription factor Sox7 has critical functions in the vascular development and angiogenesis. It is unclear whether Sox7 also contributes to cardiac septation development. We identified a de novo 8p23.1 deletion with Sox7 haploinsufficiency in an atrioventricular septal defect (AVSD) patient using whole exome sequencing in 100 AVSD patients. Then, multiple Sox7 conditional loss-of-function mice models were generated to explore the role of Sox7 in atrioventricular cushion development. Sox7 deficiency mice embryos exhibited partial AVSD and impaired endothelial to mesenchymal transition (EndMT). Transcriptome analysis revealed BMP signaling pathway was significantly downregulated in Sox7 deficiency atrioventricular cushions. Mechanistically, Sox7 deficiency reduced the expressions of Bmp2 in atrioventricular canal myocardium and Wnt4 in endocardium, and Sox7 binds to Wnt4 and Bmp2 directly. Furthermore, WNT4 or BMP2 protein could partially rescue the impaired EndMT process caused by Sox7 deficiency, and inhibition of BMP2 by Noggin could attenuate the effect of WNT4 protein. In summary, our findings identify Sox7 as a novel AVSD pathogenic candidate gene, and it can regulate the EndMT involved in atrioventricular cushion morphogenesis through Wnt4–Bmp2 signaling. This study contributes new strategies to the diagnosis and treatment of congenital heart defects.


Introduction
Congenital heart defects (CHDs) are the most common form of birth defects, account for~1% of newborns, and are the leading cause of infant death resulting from birth abnormality, with 22.5% deaths of infants 1 . Cardiac malformations which are attributable to aberrant development of the cardiac septum are the most common CHDs 2,3 .
Among them, atrioventricular septal defect (AVSD) covers a spectrum of heart anomalies with a common atrioventricular connection and has an incidence of 4-5.3 per 10,000 live birth [4][5][6] . A key process in atrioventricular septal development is the endothelial to mesenchymal transition (EndMT), which consists of multiple cellular events including the delamination of endocardial cells from the atrioventricular canal (AVC) endocardium, the acquisition of mesenchymal phenotypes and their invasion into the extracellular matrix (ECM) [7][8][9] . Further proliferation and remodeling of these mesenchymal cells, as well as the extracellular matrix, result in the formation of membranous portions of the atrial and ventricular septae, and the generation of thin, pliable valves, creating the partitioned four-chamber heart 10 . However, the underlying genetic etiology of AVSD still remains poorly understood.
The SOX family (Sry-related high mobility group box) of transcription factors have pivotal function in multiple developmental processes 11 . Sox7 is one of the Sox-F gene family (Sox7, Sox17, and Sox18), which plays a critical role in hematopoiesis, angiogenesis, vascular development, and formation [12][13][14][15][16][17] . In addition, global knockout of Sox7 is embryonic lethal from E10.5 with developmentally delayed embryos characterized by dilated pericardial sacs and a failure to remodel yolk sac vasculature 18 . However, the role of Sox7 in mammal heart development still has not been explored.
In this study, we identified an AVSD patient with Sox7 haploinsufficiency using whole exome sequencing in 100 AVSD patients. Then, multiple Sox7 conditional loss-offunction mice models were generated to explore the role of Sox7 in atrioventricular cushion development. Sox7 endocardial deficiency resulted in impaired atrioventricular cushion EndMT process, which led to improper atrioventricular cushion formation and partial AVSD. We also found that Sox7 modulates the atrioventricular cushion EndMT process through Wnt4-Bmp2 signaling. This study provides the first evidence that Sox7 has a critical role in mammal heart development, and contributes new strategies for the diagnosis and treatment of CHDs.

Identification of 8p23.1 deletion in an AVSD patient
We performed whole exome sequencing in 100 Chinese Han AVSD patients and 474 healthy controls. Based on the criteria mentioned in the section "Materials and methods", we identified 10 rare copy number variants (CNVs) in 13 patients with AVSD finally (Supplementary  Table S1). However, among these 10 rare CNVs, those known genes involved in human development of AVSD 19 , including Gata4, Gata5, Gata6, Nkx2.5, Zic3, Tbx5, Nodal, Cfc1, Lefty2, Acvr2b, Bmp4, Acvr1, Vegf, Col6a1-2, Fbln2, Frzb, Dscam, Creld1, Gja1, or Nr2f2 have not been found. Interestingly, we identified a patient with 8p23.1 deletion which was absent from controls, and this 2.5 Mb region did not contain the known pathogenic Gata4 gene, but contains the Sox7 gene (Fig. 1A). This patient is a 2year-old girl, and her clinical diagnosis is partial AVSD, mitral regurgitation (MR) and tricuspid regurgitation (TR), there are no extra cardiac abnormalities. We also collected blood samples from her healthy parents. Verification by qPCR showed that the 8p23.1 deletion of the AVSD patient is de novo (Fig. 1B). Because there were 12 genes in this 2.5 Mb deletion region, we performed qPCR to detect the expression of these genes in mice embryo hearts at E9.5, and found a relatively high expression level of Sox7, Eri1, Ppp1r3b, Msra, and Mfhas1 (Fig. 1C).

Sox7-deficient mice exhibited abnormal atrioventricular cushions formation
The heart expression patterns of Sox7 in mice and human have not been fully explored previously. In mice, we firstly found that Sox7 was mainly expressed in endocardial cells in the AVC and the delaminating cells forming the cushion mesenchyme at E9.5 ( Fig. 2A, D). At E11.5, Sox7 was mainly expressed in both atrioventricular cushion and outflow tract cushion using immunofluorescence (Fig. 2B, C, E, F). At postnatal stages, Sox7 was highly expressed in the heart by qPCR (Supplementary Fig. S1A). In human embryos, we found that SOX7 proteins had high abundance in the endocardium overlaying atrioventricular cushion, and mesenchymal cells derivated from the endocardial cells by immunofluorescence ( Fig. 2G-L). We also detected Sox7 mRNA expression in several cell lines by qPCR, including human umbilical vein endothelial cell (HUVEC), human vascular smooth muscle cell (hVSMC), mouse aortic endothelial cell (MAEC), and mouse cardiac muscle cell (HL-1). It demonstrated that Sox7 was expressed in MAEC and HUVEC, but hardly expressed in hVSMC and HL-1 cell lines ( Supplementary Fig. S1B). These data indicated that Sox7 was strongly expressed in the developing atrioventricular cushion in both mice and human, suggesting the potential role of Sox7 in cardiac septation.
Previous studies have shown that global and endothelial-specific deletion of Sox7 resulted in embryonic lethality with severely impaired angiogenesis and cardiovascular failure 12,18 . In this study, we firstly inactivated a conditional Sox7 floxed allele in endocardial lineage cells using Nfatc1-Cre (Fig. 3A). We found Nfatc1 Cre;Sox7 fl/fl mice survive after birth. To measure the efficiency of Sox7 inactivation, we tested its expression in Nfatc1 Cre;Sox7 fl/fl and Sox7 fl/fl (littermate control) embryos at E9.5 using in situ hybridization, and found Expression patterns of Sox7 in mice and human embryos. A-F Expression patterns of Sox7 in mice embryos. Sox7 was mainly expressed in endocardial cells in the atrioventricular canal (AVC, arrowheads) and the delaminating cells forming the cushion mesenchyme of mice embryos at E9.5 and E11.5, Sox7 was mainly expressed in both atrioventricular cushion and outflow tract cushion at E11.5 tested by immunofluorescence. G-L Expression patterns of Sox7 in human embryos. SOX7 proteins were strongly expressed in the endocardium overlaying atrioventricular cushion (arrowheads and white dashed lines) and mesenchymal cells derivated from the endocardial cells in the atrioventricular cushion in human embryos at 11S-16S determined by immunofluorescence. A atria, V ventricle, OFT outflow tract. Scale bars: 200 μm.
that Sox7 mRNA was reduced in atrioventricular cushions of Nfatc1 Cre;Sox7 fl/fl embryos compared with those in controls (Fig. 3B). We also collected atrioventricular cushions and hearts of mice embryos at E9.5 and E10.5, respectively, and detected Sox7 mRNA and protein by qPCR and western blot. It showed reduced levels of Sox7 mRNA (by~50%) and SOX7 protein (by~70%) in Nfatc1 Cre;Sox7 fl/fl samples (Fig. 3C, D).

Sox7 was required for EndMT process
A key event in atrioventricular cushion formation is EndMT, which consists of multiple cellular events.  To determine the contribution of Sox7 to EndMT, we bred Sox7 fl/fl mice with endocardial lineage-specific Nfatc1-Cre transgenic driver line, endothelium-specific Tie2-Cre transgenic driver line, and cardiac progenitor cell-specific Nkx2.5-Cre transgenic driver line. At E10.5, we found evidence of delayed development and failure of yolk sac vascular remodeling in Tie2 Cre;Sox7 fl/fl embryos ( Supplementary Fig. S2), confirming previous studies 12 . Furthermore, atrioventricular cushions were populated by mesenchymal cells formed by endocardial EndMT in control hearts, but there were significantly reduced mesenchymal cell numbers in atrioventricular cushions of Nfatc1 Cre;Sox7 fl/fl embryos at E9.5, E10.5, and E11.5 ( Fig.  4A-D), which pointed out disruption of the EndMT process. We also carefully counted the somite of Nfatc1 Cre; Sox7 fl/fl embryos, there was no significant difference between Nfatc1 Cre;Sox7 fl/fl embryos and littermate controls at E9.5 (19.80 ± 0.66 vs. 18.75 ± 0.75; P = 0.33), therefore, the EndMT defects could be explained by a specific effect of loss of Sox7 deficiency in the endocardial cells. In addition, we also found significantly reduced mesenchymal cell numbers in atrioventricular cushions Nkx2.5 Cre;Sox7 fl/fl embryos at E9.5 (Supplementary Fig.  S3A and 3B).
Proper formation of ECM is a crucial step for endocardial EndMT process 20 . Our data showed significantly decreased glycosaminoglycans in atrioventricular cushions of Nfatc1 Cre;Sox7 fl/fl embryos at E9.5 using Alcian Blue staining (Fig. 4E), suggesting that Sox7 deficiency in the endocardium did affect ECM formation. Since proliferation contributes to EndMT and Sox7 is an important regulator of cell proliferation 21 , we examined the effect of Sox7 inactivation on proliferation of endocardial cells using immunofluorescence, and found that the number of KI67-positive endocardial cells in Nfatc1 Cre;Sox7 fl/fl cushions decreased sharply compared with those in controls (Fig. 4F). Similarly, the number of proliferating cell nuclear antigen (PCNA)-positive endocardial cells in Nfatc1 Cre;Sox7 fl/fl cushions was much fewer than those in controls ( Supplementary Fig. S3C). However, apoptosis analysis by TUNEL showed that there was no significant difference of endocardial cell apoptosis between Nfatc1 Cre;Sox7 fl/fl embryos and controls at E9.5, E11.5 and E14.5 ( Supplementary Fig. S3D, E and F). We then studied whether Sox7 promotes proliferation by regulating cell cycle factors, the mRNA expression levels of cell cycle factors (Cyclin a2, Cyclin d2, Cdkn1a, Cdkn2b, and Cdk4) were detected in hearts of Nfatc1 Cre;Sox7 fl/fl embryos and control embryos at E9.5, and we found reduced expression of Cyclin a2 and Cyclin d2 in mutant embryos compared with the control ones ( Supplementary Fig.  S4A). To explore whether Sox7 can regulate cell cycle, we overexpressed Sox7 in mouse embryonic endocardial cell (MEEC) cell lines, and the flow cytometry results suggested the overexpression group had a significant higher proliferation index than that in the control group (Supplementary Fig. S4B and C). These findings indicated that endocardial ablation of Sox7 resulted in inhibited endocardial cells proliferation, which likely contributed to the hypocellular cushion defects in Nfatc1 Cre;Sox7 fl/fl embryos.
To further identify whether Sox7 was required for endocardial EndMT, we also performed in vitro collagen gel assays and calculated the numbers of mesenchymal cells that have migrated from the explant into the matrix after 72 h. We found significantly reduced migration over the top of matrix in Nfatc1 Cre;Sox7 fl/fl AVC explants compared with those in littermate control (Sox7 fl/fl ) AVC explants (27.25 ± 2.66 vs. 81.25 ± 6.46, P < 0.001) (Fig. 5A, B). In particular, we performed 3D reconstruction using NIS-Elements software (Nikon) to show cells that migrated away from explants and invaded the gel visually, and we found significantly reduced migration and invasion in Nfatc1 Cre;Sox7 fl/fl AVC explants compared with those in littermate control (Sox7 fl/fl ) AVC explants (49.75 ± 7.32 vs. 202.5 ± 13.42, P < 0.001) (Fig. 5C, D). These data indicated that Sox7 was required for normal endocardial EndMT process. Impaired EndMT process might elucidate abnormal atrioventricular cushion formation in Sox7-deficient mice.
Decreased BMP signaling in AVC of Sox7-deficient embryos Then, we screened altered signaling pathways underlying the hypocellular atrioventricular cushion phenotype of Nfatc1 Cre;Sox7 fl/fl embryos using an RNA-sequencing (RNA-seq) detection in microdissected AVC tissues of E9.5 Nfatc1 Cre;Sox7 fl/fl embryos and controls (Sox7 fl/fl ). The transcriptome analysis yielded 370 differentially expressed genes of which 218 were upregulated and 152 downregulated ( Fig. 6A; Supplementary Fig. S5A). Functional analysis of the differentially expressed genes identified several Gene Ontology (GO) categories, and the most highly represented GO term was Response to X-ray, followed by Retinal rod cell, BMP signaling pathway involved in heart development, activation of meiosis, and endocardial cushion morphogenesis ( Fig. 6B; Supplementary Fig. S5B). Ingenuity pathway analysis figured out 15 altered pathways ( Fig. 6C; Supplementary Fig. S5C).
Next, we utilized whole-mount in situ hybridization (WISH) and immunofluorescence to further test the expression alterations of those markers which related to atrioventricular cushion development selected from the RNA-seq results. We found that the expression of Snai1 which is a key regulator of EndMT was downregulated in Nfatc1 Cre;Sox7 fl/fl AVCs (Fig. 7A), and we also found increased expression of VE-cadherin which was directly repressed by Snai1 in EndMT in AVCs of Nfatc1 Cre; Sox7 fl/fl embryos compared with that in controls at E9.5 ( Supplementary Fig. S6A). However, another EndMT regulator, Notch1, was not changed in Nfatc1 Cre;Sox7 fl/fl AVCs compared with that in controls (Fig. 7A). We also examined the transcription factors Sox9 and Msx1, which are expressed in the atrioventricular cushion mesenchyme and crucial for EndMT process 22,23 . It showed that both of them were reduced in Nfatc1 Cre;Sox7 fl/fl AVCs compared with those in controls ( Fig. 7A; Supplementary Fig.  S6B). Moreover, in agreement with the Alcian Blue staining data (Fig. 4E), the expression of Has2 which is responsible for deposition of the major glycosaminoglycan of the atrioventricular cushion, was also decreased in atrioventricular cushions of Nfatc1 Cre;Sox7 fl/fl embryos compared with that in controls (Fig. 7A).
Pathway analysis of the RNA-seq data identified significant reduction of the BMP/TGF-beta signaling pathway in Nfatc1 Cre;Sox7 fl/fl AVCs ( Fig. 6C; Supplementary  Fig. S5C). We further revealed that the expression of Bmp2 in atrioventricular cushion myocardium was downregulated in Nfatc1 Cre;Sox7 fl/fl AVCs by WISH (Fig. 7B), but another factor Tgfβ-2 which is also expressed in atrioventricular cushion myocardium was grossly normal in Nfatc1 Cre;Sox7 fl/fl AVCs compared with controls ( Supplementary Fig. S6B). In particular, Tbx2, a direct target of Bmp2 signaling pathway during atrioventricular cushion development 24 , was also reduced in Nfatc1 Cre;Sox7 fl/fl AVCs (Fig. 7B). As well known, the Bmp2 signaling pathway works through phosphorylation   and nuclear localization of the Smad1/5/8 complex. Interestingly, we also found that phosphorylated-Smad1/ 5/8 (pSmad1/5/8) levels in atrioventricular cushions of E10.5 Nfatc1 Cre;Sox7 fl/fl embryos were decreased compared with those in controls using immunofluorescence (Fig. 7D). However, Bmp2 is mainly expressed in AVC myocardium, we then pondered over how endocardial Sox7 regulated the AVC myocardium Bmp2 expression. Wnt and Bmp, two critical signaling pathways that have already been shown to regulate EndMT in AVC endocardium and myocardium, respectively 25 . Therefore, we overexpressed Sox7 in the MEEC line to explore several Wnt ligands expressions, and it showed that only Wnt4 was significantly changed (Supplementary Fig. S7A and B). Moreover, Wnt4, which is expressed in endocardium, can regulate Bmp2 expression in AVC myocardium during atrioventricular cushion development 26 . So, we measured the Wnt4 expression in Nfatc1 Cre;Sox7 fl/fl embryos, it showed that Sox7 deficiency reduced Wnt4 expression in endocardium, as measured by both WISH and immunofluorescence (Fig. 7B, C). In addition, Tbx20, an upstream transcription factor of Bmp2 remained unchanged in Nfatc1 Cre;Sox7 fl/fl AVCs (Fig. 7B), indicating that Sox7 regulated Bmp2 signaling independent of Tbx20. WISH analysis of another transcription factor Pitx2, whose transcript levels were significantly altered in Nfatc1 Cre;Sox7 fl/fl AVCs by RNA-seq, showed no changes between Nfatc1 Cre;Sox7 fl/fl embryos and controls ( Supplementary Fig. S6B). All these findings indicated that Sox7 affected endocardial EndMT process and atrioventricular cushion development via Wnt4 and Bmp2 signaling.

Sox7 transcriptionally regulates Wnt4-Bmp2 signaling
In an attempt to explore the specific modulation between Sox7 and Bmp2 signaling, we overexpressed Sox7 in MEEC, then the expressions of Wnt4, Bmp2 and its target gene Tbx2 were examined. The mRNA levels of Wnt4, Bmp2, and Tbx2 were significantly increased in MEEC (Supplementary Fig. S7A). To determine whether Smad nuclear localization was altered by Sox7 overexpression, we therefore performed pSmad1/5/8 immunostaining assay in MEEC, and found that Sox7 overexpression significantly increased pSmad1/5/8 expression and nuclear localization (Fig. 8A). Interestingly, nuclear colocalization of SOX7 and pSmad1/5/8 was also found in the endocardium overlaying atrioventricular cushions of human 13S embryos ( Supplementary  Fig. S7C). These data suggested that Sox7 regulated pSmad1/5/8 accumulation in the nucleus. Sequence analysis showed that there were several Sox7-binding sites in the promoter regions of Wnt4 and Bmp2 (Fig. 8B). Luciferase assays revealed that Sox7 could activate the promoter of both Wnt4 and Bmp2 in MEEC (Fig. 8B).
We then performed ChIP-qPCR, and found that Sox7 bound to the promoter regions of Wnt4 and Bmp2 in both adult mice hearts and MEEC cell lines ( Fig. 8C and Supplementary Fig. S7D). Specifically, Sox7 bounds to the first and third Sox7-binding site in the promoter of Bmp2 and the second Sox7-binding site in the promoter of Wnt4 in adult mice hearts, and Sox7 bounds to the first, third, and sixth Sox7-binding site in the promoter of Bmp2 and the first Sox7-binding site in the promoter of Wnt4 in MEEC cell lines (Fig. 8C and Supplementary Fig. S7E). These data indicated that Wnt4 and Bmp2 might be Sox7 direct downstream target genes. Bmp2 is an important regulator of mesenchymal cell proliferation during endocardial cushion formation 27,28 , so we examined the mesenchymal cell proliferation by KI67 staining, and found that mesenchymal cell proliferation was depressed in Nfatc1 Cre;Sox7 fl/fl AVCs compared with that in controls at E10.5 ( Supplementary Fig. S7F).
To determine whether Wnt4-Bmp2 signals were required for the regulation of Sox7 on EndMT process, we performed rescue experiments using the collagen gel EndMT assay. We cultured AVC explants isolated from E9.5 Nfatc1 Cre;Sox7 fl/fl embryos and controls, and found that the migration and invasion by transformed mesenchymal cells in Nfatc1 Cre;Sox7 fl/fl AVC explants were inhibited compared with those in controls (Fig. 8D). Nevertheless, the addition of recombinant WNT4 or BMP2 protein was able to partially rescue the EndMT defect caused by Sox7 deficiency, and blocking BMP2 activities with Noggin attenuated rescue of impaired EndMT by WNT4 protein (Fig. 8D). Taken together, these data implicated that Sox7 regulated EndMT process through Wnt4-Bmp2 signaling axis (Fig. 8E).

Discussion
AVSD is a genetically heterogeneous defect. Several syndromes are associated with AVSD, including Down syndrome, Noonan syndrome, Holt-Oram syndrome, and 8p23.1 deletion syndrome 6,29-31 . We used whole exome sequencing to screen out possible pathogenic CNVs in 100 AVSD patients and 474 healthy controls. We identified an AVSD patient harbored a de novo 8p23.1 deletion, it did not contain the GATA4 gene which is considered as a likely candidate gene for the heart defects of 8p23.1 deletion patient but contained the Sox7 gene. Similarly, Wat et al. 32 indicated that the haploinsufficiency of Sox7 may increase severity of the cardiac phenotype in 8p23.1 deletion syndrome patient with GATA4 deletion. We also performed qPCR to detect the expression of genes within the 2.5 Mb deletion in mice embryo hearts at E9.5, and found a relatively high expression level of Sox7, Eri1, Ppp1r3b, Msra, and Mfhas1. However, there was no report on the relationship between Eri1 or Mfhas1 and cardiovascular disease/development. Wei-J et al. reported that Ppp1r3b polymorphisms were associated with serum LDL-C levels, the risk of coronary artery disease and ischemic stroke in the Southern Chinese Han population 33 . The lack of Msra exacerbates cardiovascular disease phenotypes driven by increased oxidative stress 34 . Nevertheless, only Sox7 is related to cardiovascular development 12,16,17 . Therefore, we considered Sox7 as the candidate gene for the AVSD patient in our study. The role of Sox7 in mammal heart development has not been explored. In humans, Sox7 was expressed in the heart, lung, brain, tongue, vertebrae, and liver of the 8week human embryo 35,36 . We showed here for the first time that SOX7 proteins were strongly expressed in the endocardium overlaying atrioventricular cushion in early human embryos (4-6 weeks). A previous study by deleting Sox7 in pan-endothelium with Tie2-Cre indicated an essential role of Sox7 in angiogenesis 12 . However, these Sox7 mutant embryos died before E10.5. In our study, we deleted Sox7 specifically in endocardium and its mesenchymal progeny using the Nfatc1 cre driver line, which developed normal yolk sac vasculature and survived till to postnatal stage, allowing us to intensively study the role of Sox7 in atrioventricular cushion morphogenesis. Histological analysis revealed that the EndMT process of Nfatc1 Cre;Sox7 fl/fl and Nkx2.5 Cre;Sox7 fl/fl mice embryos were disrupted. Nfatc1 was expressed in the developing heart of mice embryos at E7.75-E9.0 37 , occurred before our proliferation and apoptosis analysis, which allowed us to explore the role of Sox7 in endocardial cells proliferation and apoptosis at E9.5. We found Sox7 may promote cell proliferation by upregulating cell cycle factors Cyclin a2 and Cyclin d2, and the reduced cell proliferation is at least one of the reasons that resulted in formation of hypocellular endocardial cushions in Nfatc1 Cre;Sox7 fl/fl embryos. In addition, we found incomplete fusion or developmental delay of contact of the developing ventricular septum with the cardiac cushion in Nfatc1 Cre;Sox7 fl/fl embryos at E13.5, which was similar to a previous study of Nipbl haploinsufficiency mouse model 38 , and we also found defects in closure of the ASD at E14.5 in Nfatc1 Cre;Sox7 fl/fl mice. Taken together, all these data pointed out the abnormal formation of cardiac septation in Sox7-deficient mice.
The regulation of atrioventricular cushion formation consists of a series of complex and molecular factors. The BMP signaling pathway, particularly Bmp2, is a well-known paracrine signal secreted from myocardium for EndMT and post-EndMT mesenchymal cell proliferation during endocardial cushion formation 26,39 . Signaling crosstalk between endocardium and myocardium has been showed essential for atrioventricular cushion EndMT process. It has been reported that Jagged1-Notch1 signaling in endocardial cells can induce the expression of endocardial Wnt4, which subsequently acts as a paracrine factor to upregulate Bmp2 expression in the adjacent AVC myocardium to trigger EndMT 26 . Wnt4 is a member of the WNT family of secreted molecules that function in a paracrine manner to affect several developmental changes, and Bmp2 is also the downstream target gene of Wnt4 during ovary development 40 . In this study, we found Bmp2 signaling was downregulated in Nfatc1 Cre;Sox7 fl/fl AVCs at E9.5 by both RNA-seq and WISH. Furthermore, we firstly showed that Sox7 deficiency reduced endocardial Wnt4 expression, luciferase and ChIP-qPCR assays identified Wnt4 as a direct downstream target of Sox7. In addition, WNT4 or BMP2 treatment was able to partially rescue EndMT defect caused by Sox7 deficiency, and blocking BMP2 activities with Noggin attenuated rescue of impaired EndMT by WNT4. It will be interesting to investigate how endocardial Wnt4 regulates AVC myocardial Bmp2 expression in the future.
(see figure on previous page) Fig. 8 Sox7 transcriptionally regulated Wnt4-Bmp2 signaling. A Immunofluorescence of pSmad1/5/8 in Sox7-overexpression and wild type MEEC. B The schematic diagram of Sox7-binding sites in the promoter regions of Bmp2 and Wnt4, and the effect of Sox7 overexpression on the luciferase activity of Bmp2 and Wnt4 promoter in MEEC (n = 3). C Chromatin immunoprecipitation (ChIP) assay of adult mice hearts by ChIP-qPCR which showed nucleic acid electrophoresis of genomic DNA fragments after ChIP using anti-Sox7 antibody compared to control isotype for Bmp2 and Wnt4 (n = 4), the binding sites were highlighted by red boxes. D In vitro 3D collagen gel EndMT assays showed that WNT4 or BMP2 recombinant protein additions was able to rescue EndMT defect induced by Sox7 deficiency, and blocking BMP2 activities with Noggin attenuated above rescue caused by WNT4 protein (n = 3). Quantitative analysis of the number of transformed mesenchymal cells under each treatment. Data are means ± SEM. *P < 0.05, **P < 0.01. Scale bars: 200 μm. E Schematic illustration of how Sox7 regulates atrioventricular cushion development through Wnt4-Bmp2 signaling. Sox7 expression in AVC endocardium drives EndMT via Wnt4 and Bmp2 activation in AVC endocardium (red) and myocardium (yellow), respectively, in E9.5 WT heart, and a four-chamber heart is formed eventually at E14.5. Sox7 deficiency downregulates Wnt4 and Bmp2 expression, which disrupts normal EndMT process, leading to abnormal formation of atrioventricular cushion at E14.5. During early atrioventricular cushion development, Sox7 is mainly expressed in the AVC endocardium, it can activate endocardial Wnt4 expression directly, which then subsequently acts as a paracrine factor to upregulate Bmp2 expression in the adjacent AVC myocardium to trigger EndMT via Tbx2 and Snail, Bmp2 also regulates EndMT through promoting Sox9, Msx1, and Twist1 expression in the atrioventricular cushion mesenchyme.

Human subjects
Our study recruited 100 AVSD patients in Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine and Shanghai Children's Medical Center whose diagnoses were confirmed by echocardiography, cardiac catheterization examinations, computed tomography, and other operation recordings. 474 healthy controls were also enrolled.
Whole exome sequencing analysis and quantitative realtime polymerase chain reaction validation We performed whole-exome sequencing in 100 AVSD patients and 474 healthy controls. The DNA was sequenced using the Illumina HiSeq 2500 platform at a commercial provider (Shanghai Biotechnology Co, Ltd., Shanghai, China). The data of CNVs were filtered. We distinguished common CNVs from rare CNVs by comparing the results with the known CNVs in the Database of Genomic Variants (DGV, http://dgv.tcag.ca/) and Online Mendelian Inheritance in Man (OMIM, http://omim.org). Rare CNV segments were identified based on the following criteria: (1) CNV > 500 kb, but <5 Mb in size; (2) average depth of sequencing >54; (3) Zscore of average depth ≥2.0; (4) present at <0.1% frequency or not found in the DGV (http://dgv.tcag.ca/). Finally, we identified a patient with 8p23.1 deletion which was absent from controls, and we also collected blood samples from her healthy parents, they were verified by quantitative real-time polymerase chain reaction (qPCR) at a commercial provider (Genesky Biotechnologies Inc., Shanghai). Moreover, we performed qPCR to detect the expression of genes within the 8p23.1 deletion in mice embryo hearts at E9.5. Primers used to amplify genes within the 8p23.1 deletion region are shown in Supplementary Table S2.

Histology detection
Mouse tissues were fixed with 4% paraformaldehyde in PBS, paraffin embedded, sectioned at 4-μm intervals, and processed. Hematoxylin and eosin (H&E) and Alcian Blue staining were performed using standard procedures. Five embryos were analyzed for H&E and Alcian Blue staining each.

AVC explant cultures
A 1.5 mg/ml solution of rat-tail collagen type I (BD Biosciences) was dispensed into four-well microculture dishes and allowed to solidify inside a 37°C, 5% CO 2 incubator. Collagen gels were washed several times with Opti-MEM. Subsequently, the wells were filled with Opti-MEM containing 1% fetal calf serum, 1% insulintransferrinseleniun (ITS, Invitrogen), and antibiotics, and incubated overnight. AVC explants from E9.5 embryonic hearts were carefully dissected in ice cold PBS, cut open and placed with the endocardium facing down onto the collagen matrixes after removing the excess medium. Following placement, the explants were left to attach for 24 h at 37°C in 5% CO 2 . Medium and specific treatments were added to the dishes, and explants were cultured for 3 days, then all cells that migrated away from explants and invaded the gel were counted. For 3D reconstruction of the E9.5 explants, images were taken of several optical sections per explant and reconstructions were performed using NIS-Elements software (Nikon). For rescue experiments, Bmp2 (200 ng/ml) (R&D systems), Wnt4 (250 ng/ml) (Proteintech), and Noggin (200 ng/ml) (Novoprotein) were added to the explant culture. After fixation in 4% paraformaldehyde (30 min room temperature), antigens were detected using Phalloidin-TRITC (1:100, Yeasen Biotech) and nuclei counterstained with DAPI and analyzed using a Leica SP5 confocal microscope. 4-6 explants were analyzed for each group.
Cell culture and transfection HUVEC, hVSMC, MAEC, HL-1, and MEEC cell lines were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China) and cultured in the DMEM medium (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen), penicillin (100 units/ml), and streptomycin (100 μg/ml). Cells were incubated at 37°C in a humid atmosphere with 5% CO 2 . The transient transfection was performed with Fugene HD transfection reagent (Promega) according to the manufacturer's protocol for adherent cells.
Flow cytometry cell cycle analysis MEEC cells were seeded into six-well plates and transfected with control or wild-type Sox7 plasmids. Cells were trypsinized and collected after 48 h. Cells were washed with PBS, fixed in ice-cold 70% ethanol and kept at 4°C overnight. The following day, cells were resuspended in PBS, centrifuged, and stained with propidium iodide. Flow cytometry analysis was performed using CellQuest software (BD Biosciences, USA). The proliferation index = (G2/M + S)/(G2/M + S + G1/G0).

Transcriptome analysis
AVCs dissected from Nfatc1 Cre;Sox7 fl/fl and control (Sox7 fl/fl ) embryos at E9.5 were flash frozen in RNAlater TM solution (Invitrogene), 10 AVCs were pooled per replicate. The RNA quality was checked by Bioanalyzer 2200 (Aligent) and the RIN > 6.0 is suitable for RNA sequencing. The cDNA libraries were constructed for each pooled RNA sample using the NEBNext ® Ultra™ Directional RNA Library Prep Kit for Illumina according to the manufacturer's instructions.

In situ hybridization
Whole mount and section in situ hybridization were conducted as described previously 42,43 . Details of probes are shown in Table S3.

Quantitative real-time polymerase chain reaction
Total RNA from cardiac tissues and cell lines was isolated using TRIzol reagent (Life Technologies), and converts to cDNA using the PrimeScript™ RT Master Mix (Takara). For qPCR, SYBR Green qPCR master mix (Takara) was used following the protocol supplied by the manufacturer. Primers for qPCR are shown in Table S4.

Luciferase assay
Transfections and dual luciferase reporter assays were carried out as described previously 44 . The luciferase reporters of Bmp2-luc and Wnt4-luc were constructed by inserting the conserved Sox7-binding sites 45 in the promoter regions of the Bmp2 and Wnt4 into the pGL3promoter plasmid (Promega).

ChIP-qPCR assays
ChIP-qPCR assays of MEEC line and mice cardiac tissues were performed using the chromatin immunoprecipitation assay kit (Millipore) and EpiQuik Tissue Chromatin Immunoprecipitation Kit (Epigentek) following the protocol supplied by the manufacturer. Primers used for ChIP-qPCR are listed in Supplementary Table S5.

Statistics
All data are represented as mean (SD) of at least three independent experiments unless otherwise stated in the figure legend, and statistical analysis was performed using the Student's t test or one-way or two-way ANOVA followed by the Bonferroni's multiple comparison test with GraphPad Prism version 5.0 software. P < 0.05 was considered statistically significant.