A TBX5 3′UTR variant increases the risk of congenital heart disease in the Han Chinese population

TBX5 is a vital transcription factor involved in cardiac development in a dosage-dependent manner. But little is known about the potential association of TBX5 3′ untranslated region (UTR) variations with congenital cardiac malformations. This study aimed to investigate the relationship between TBX5 3′UTR variants and risk for congenital heart disease (CHD) susceptibility in two Han Chinese populations, and to reveal its molecular mechanism. The relationship between TBX5 3′UTR variants and CHD susceptibility was examined in 1 177 CHD patients and 990 healthy controls in two independent case–control studies. Variant rs6489956 C>T was found to be associated with increased CHD susceptibility in both cohorts. The combined CHD risk for the CT and TT genotype carriers was 1.83 times higher than that of CC genotype, while the risk for CT or TT genotype was 1.94 times and 2.31 times higher than that of CC carriers, respectively. Quantitative real-time PCR and western blot analysis showed that T allele carriers exhibited reduced TBX5 mRNA and protein levels in CHDs tissues. Compared with C allele, T allele showed increased binding affinity to miR-9 and miR-30a in both luciferase assays and surface plasmon resonance analysis. Functional analysis confirmed that miR-9 and miR-30a downregulated TBX5 expression at the transcriptional and translational levels, respectively. The assays in zebrafish model were in support of the interaction of miR-9/30a and TBX5 3′UTR (C and T allele). We concluded that TBX5 3′UTR variant rs6489956 increased susceptibility of CHD in the Han Chinese population because it changes the binding affinity of two target miRNAs that specifically mediate TBX5 expression.


Introduction
Congenital heart diseases (CHDs) are the most prevalent type of human structural birth defects worldwide, affecting between 19 and 75 per thousand live births [1]. The etiology of CHDs includes both genetic and environmental factors. Extensive model organism studies and linkage analysis of rare Mendelian CHD families have provided insights into the genetic basis of CHD, which results from alterations in transcription factors and genes involved in cardiac development [2][3][4][5]. However, mutations in these genes can only explain a small fraction of CHD cases. The pathogenesis of sporadic cases, the most common form of CHD, remains poorly understood.
Genome-wide association studies (GWASs) suggest that a large number of non-coding variants contribute to the risk of various diseases, including cardiovascular disease [6][7][8][9][10]. Our previous study demonstrated that non-coding variants in genes coding for enzymes that biotransform homocysteine significantly increase CHD risk by regulating gene expression levels [11][12][13]. During embryonic cardiogenesis, many transcription factors including TBX5 [14], GATA4 [15] and NKX2-5 [16] are strictly regulated in a dosage-dependent manner. Mutations in the coding region of TBX5 invariably result in Holt-Oram syndrome (HOS), an inherited disease characterized by upper limb and cardiac deformities [2,17]. Thus, we hypothesized that the functional regulatory variations in TBX5 might be associated with increased CHD risk through altered gene expression or dosage effects.
TBX5 is a transcription factor with well-defined roles in heart and forelimb development. Multiple studies in animal models have confirmed that cardiac development is sensitive to Tbx5 dosage [14,18,19]. Most notably, CHD might result from the haploinsufficiency of Tbx5.
Recently, several non-coding variants of TBX5 in the promoter [20], intron [21] and enhancer regions [22] were reported to be associated with CHDs. Although the 3′ untranslated region (3′UTR) is critically important for microRNAs (miRNAs) binding to regulate gene expression, the potential contribution of mutations in the TBX5 3′UTR to CHDs remains unknown. Our previous study demonstrated that miR-10a and miR-10b significantly repressed TBX5 protein levels by targeting its 3′UTR [23]. In present study, we detected a common variant, rs6489956, in the TBX5 3′UTR that significantly increases CHD risk in a Han Chinese cohort composed of 1 177 CHD patients and 990 healthy controls. We further investigated the molecular mechanism underlying that TBX5 3′UTR variants influence risk for CHD susceptibility.

TBX5 3′UTR variant rs6489956 significantly increased CHD susceptibility in Han Chinese
In our study, a total of ten variants in TBX5 3′UTR were identified. Among them, four variants had minor allele frequency (MAF) 40.05 and were chosen for genotyping in 288 cases and controls of the Shandong group. Among them, the prevalence of variant rs6489956 (c.* 1101C4T) was significantly different between the CHDs and the control subjects. This variant was then selected for further validation in the other two independent case-control cohorts (Supplementary Table S3).
As the homozygous TT genotype was very rare, we applied the dominant model of inheritance to enhance the statistical power of the study, which combined homozygous TT alleles with heterozygous CT alleles to compare with the wild-type CC alleles in the association studies. Logistic regression analyses demonstrated heterozygote CT and homozygote TT subjects had a , as well as 272 CHDs and 384 controls in the Shanghai cohort (OR = 1.75, 95% CI = 1.24-2.46, P = 0.0013). The combination of these two independent studies showed a 1.83-fold increased CHD risk when the dominant model is used for TBX5 rs6489956 (P = 3.62 × 10 − 8 ) in a total number of 1 177 CHDs patients and 990 controls (Table 1), with an adjusted OR of 1.94 (95% CI = 1.52-2.47, P = 8.21 × 10 − 8 ) and 2.31 (95% CI = 1.01-5.30, P = 0.033) for the CT and TT genotypes, respectively. Therefore, The variant rs6489956 was significantly associated with an increased risk of CHD in two cohorts separately and together. Similarly, the allele analysis showed that carriers of the susceptible T allele exhibited a higher risk of CHD than carriers of the C allele in both cohorts, resulting in a 1.86-fold combined higher susceptibility of CHD (95%CI = 1.50-2.30, P = 2.01 × 10 − 8 ). All genotype frequencies were in accordance with the Hardy-Weinberg expectation among the control subjects (P40.05).
The TBX5 c.*1101C4T affected TBX5 mRNA transcription and translation Since the TBX5 c.*1101C4T (rs648995) variant is located in the TBX5 3′UTR, we assumed that this  (Figure 1b). Although this study measured TBX5 mRNA and protein levels in heart tissue samples from subjects who have already been born, which might be different from embryonic TBX5 levels, our results demonstrated that TBX5 transcription and translation levels were correlated with the variant rs6489956 genotypes. Altered TBX5 mRNA expression associated with the variant rs6489956 was further confirmed by in vitro luciferase assays in different cell lines. Consistent with the tissue sample results, the plasmid containing the mutant-T allele showed a 26%, 28% and 31% reduction in luciferase activity when compared with the wild-type C allele in HEK 293T, HCM and H9C2 cells, respectively (Po0.01) ( Figure 1c). Therefore, both in vivo and in vitro experimental evidence clearly supports the hypothesis that the TBX5 c.*1101C4T variant affected the transcription of TBX5.
miR-9 and miR-30a inhibited TBX5 expression by interacting with the TBX5 c.*1101C4T variant In the flanking region of the TBX5 3′UTR variant rs6489956, there are 16 software-predicted miRNAs including five upstream miRNAs, seven downstream miRNAs and four on-site miRNAs (Figure 2a). It was found that only 2 out of these 16 miRNAs, namely, miR-9 and miR-30a, can cause a significant disparity in the expression of the two allelic reporters in HEK 293T cells co-transfected with the psiCHECK-C/T plasmid ( Figure 2b). The endogenous expression of miR-9 and miR-30a was detected in all of the 30 human cardiac samples ( Figure 3a). We also observed that miR-30a binds to TBX5 mRNA carrying the T allele with a higher affinity compared with that of the C allele (5.10 × 10 − 7 mol l −1 [M] vs 1.23 × 10 − 6 M) (Figure 2e). For the upstream miR-9, although the variant rs6489956 appears not to directly affect miR-9 binding (Supplementary Figure S1), it might impair the miRNAs complex that interacts with the TBX5 motif containing variant rs6489956, as previously described [25].
The specific inhibition of miR-9 and miR-30a for TBX5 caused by the 3′ UTR variant c.*1101C4T was demonstrated by a psiCHECK luciferase reporter assay as well as the rescue experiments from miRNA inhibitors. The luciferase gene expression of psiCHECK-C and psiCHECK-T were both inhibited in a linear manner with increased miRNAs concentration, and the miRNAs displayed a more pronounced inhibition effect on psiCHECK-T ( Figure 2c). On the other hand, the allelic expression disparity was diminished by the use of each specific inhibitor for miR-9 or miR-30a individually or in tandem ( Figure 2d).

miR-9 and miR-30a inhibited TBX5 expression at the transcriptional and translational levels independently
It was determined that both miR-9 and miR-30a were present at detectable levels in cardiac tissues (Figure 3a), and the expression of the two miRNAs was 3.7 to 4.9 times higher in cardiac tissues from CHD cases compared with that of cardiac tissues from the non-CHD controls (Figure 3b). The cardiac tissue samples of CHD cases were collected from children (average 4.71 years old), and the non-CHD controls were collected from adults (average 32.7 years old). Despite a potential miRNAs expression disparity between adults and children, such a large difference in miR-9 and miR-30a expression might still represent the potential contributions of the two miRNAs in CHD development, which would require more robust studies to confirm in the future.
Because miR-9 and miR-30a had similar inhibition patterns on the luciferase expression vector with the TBX5 3′UTR (Supplementary Figure S2), we measured whether TBX5 transcription or translation in cultured HEK 293T cells could be regulated by miR-9 and miR-30a. MiR-9 reduced the transcription of TBX5 mRNA by 57% (Figure 3c, left), and the miR-9 inhibitor up-regulated TBX5 mRNA expression 1.88fold (Figure 3d, left). However, neither miR-30a nor its inhibitor had any effect on TBX5 mRNA expression (Figure 3c and d, left). Both miR-9 and miR-30a could decrease TBX5 translation (Figure 3c, right). Additionally, miR-9 or miR-30a inhibitors could increase TBX5 protein expression separately or together (Figure 3d, right). These results suggested that miR-9 primarily regulates endogenous transcription of TBX5, and that miR-30a regulates translation of the TBX5 mRNA. Therefore, the mechanisms by which miR-9 and miR-30a regulate TBX5 were confirmed both in vitro and in vivo.

Discussion
It is well known that the mutations in TBX5 gene coding regions have a role in syndromic CHD such as Holt-Oram syndrome [2,[30][31][32][33][34][35][36]. However, the role of TBX5 in sporadic isolated cases of CHD has not been extensively explored. Experimental evidence based on animal models has confirmed that Tbx5 has vital roles in cardiac development, and that cardiac development is highly sensitive to Tbx5 dosage [14,18,37,38]. Therefore, we believed that functional regulatory variants of TBX5 might contribute to the etiology of sporadic isolated CHDs through a dosage effect. In this study, we investigated all common SNPs in the TBX5 3′UTR region and focused on the potential functional variant that we thought might modulate TBX5 expression by its interaction with miRNAs. The functional variant TBX5 c.*1101C4T (rs6489956) was found to significantly increase CHD risk in Han Chinese by interacting with miR-30a and miR-9 (OR = 1.83. 95% CI = 1.48-2.28, P = 3.62 × 10 − 8 in the dominant model). Although we first postulated that a functional variant in the TBX5 3′UTR may increase the risk of CHD, a recent study indirectly supported our findings by reporting that a TBX5 enhancer variant causes isolated congenital heart disease [22]. Because our study was only conducted among the Han Chinese subjects, such relationships between variants in the TBX5 3′UTR and the risk of CHD need further exploration in other ethnic groups. However, our study elucidated that the sporadic CHDs could be attributed to the combined contribution of the common Figure 3 The functional validation of the candidate miRNAs. (a) RT-PCR showed that miR-9 and miR-30a were expressed in heart tissues. These tissues originated from the outlet samples of TOF patients during cardiac operation. (b) miR-30a and miR-9 expression levels in 30 CHDs heart tissue samples were 3.73 and 4.91 times higher than the controls, respectively. The actual values compared to the internal reference gene levels for each miRNA were as follows: miR-30a = 0.096 ± 0.071, miR-9 = 0.0034 ± 0.0005; (c) TBX5 mRNA expression in HEK 293T cells was downregulated by transfection with miR-9 rather than miR-30a (left), while the presence of both miRNAs inhibited TBX5 protein concentration according to the western blot analysis results (right). (d) TBX5 mRNA concentration increased in the cells transfected with miR-9 inhibitor, but did not change following transfection of miR-30a inhibitor (left). In contrast, TBX5 protein concentration was up-regulated with transfection of the two miRNAs inhibitors. NC represents as negative control.
Feng Wang et al. 7 regulatory variations in the TBX5 gene, as well as the epigenetic factors such as miR-30a or miR-9. The possible mechanism of underlying the association of TBX5 non-coding variants with increased CHD risk may be explained by the sensitive nature of the dosage effect of TBX5 on heart development. While severe loss-of-function mutations in TBX5 coding regions led to both heart and forelimb phenotypes, non-coding variants were associated with compartmentalized phenotypes, which might reflect the differing sensitivities of the impacted tissues to the altered TBX5 expression level. This contention is supported by the fact that the TBX5 intronic variant rs7312625 was reported to be associated with altered PR intervals (intervals between P wave and R wave in electrocardiogram). In African Americans, and intronic variant rs3825214 was shown to reduce lone atrial fibrillation risk in Han Chinese [39,40]. The tissue- (f) EGFP sensors were coinjected with mCherry control as indicated. miR-9/30a precursor injection reduced EGFP levels in EGFP-3′UTR sensor (third column), while mCherry levels were unchanged. In the EGFP-3′ UTR-T sensor (fourth column), more significant reduction in EGFP was noted.
TBX5 3′UTR variant increases the risk of CHD 8 specific regulation of different non-coding regions could be one possible explanation. In both embryonic and adult human heart tissues, TBX5 expression can be observed throughout the epicardium and myocardium, including the free walls and septa of all four cardiac chambers, while its expression in the endocardium is limited to the left ventricle [41]. However, in mouse [42] and chicken [43], Tbx5 mRNA expression is restricted to the atria and left ventricle. Consistent with these expression patterns, HOS patients with TBX5 mutations may have a number of significant heart defects, such as septal defects and certain complex malformations (tetralogy of Fallot (TOF), APVR and hypoplastic left heart) [42,44]. In our study of sporadic CHD cases, we observed that rs6489956 is correlated with a series of cardiac abnormalities; septal defects were statistically most common (n = 830, P = 9.70 × 10 − 6 ). Additionally, we speculated that the failure to detect any correlation between variant rs6489956 and APVR and outflow tract malformation susceptibility might be due to the limited and insufficient sample size.
Both in vivo and in vitro assays on the function of variant rs6489956 indicated that the T allele led to reduction in TBX5 gene expression. We further identified two miRNAs, miR-9 and miR-30a, that interacted with this variant. This is not surprising, because miRNAs are predicted to function in the posttranscriptional regulation of nearly one third of all human genes [45]. It was confirmed that miR-30a did not affect the concentration of TBX5 mRNA, but reduced its protein concentration, while miR-9 downregulated both transcription and translation of TBX5. It has been extensively reported that miRNAs help to regulate cardiac growth and remodeling, and contribute to specific myocyte properties. miRNAs are implicated in numerous cardiovascular diseases, such as heart hypertrophy, heart failure, atherosclerosis, acute myocardial infarction, VSD, TOF and arrhythmias [46][47][48][49][50][51][52]. Thus, when the expression of the developmentally important transcription factor, TBX5 was disturbed during a sensitive time window of development, it increases the likelihood of inducing CHD. As a new model trigger, the interactions between miRNAs and TBX5 regulatory variant increase the CHD risk through the downregulated TBX5, either at the posttranscriptional or translational level.
Because miRNAs are vital factors that mainly target 3′UTR sequences of mRNA, we suspected that miR-NAs might have important roles in CHDs. Both bioinformatic predictions and functional analyses showed that both miR-30a and miR-9 have a higher affinity to the minor T allele. We further confirmed these results using luciferase assays, quantitative RT-PCR, western blot analysis and whole-embryo microRNA sensor assay in zebrafish. Recently, many studies have implicated miRNAs in various cardiovascular diseases. For example, miRNA-1-2 knockout mice have decreased levels of the cardiac transcription factor Hand2, which results in heart malformations (VSD) similar to those observed in Hand2-deficient mice [53,54]. Comparisons of malformed heart tissues from CHD patients with healthy controls have demonstrated dysregulation of miRNAs. However, to our knowledge, no evidence has shown that variations of miRNAs sequence, DNA binding motif or even altered miRNAs levels can directly cause CHD. Our study initially demonstrated that certain miRNAs can inhibit the expression of important transcription factors for cardiac development through the alternative binding sites and variable binding affinities, which eventually increase CHD risk. We also found that the levels of miR-30a and miR-9 were 3.72 and 4.91 times higher in CHD children compared with non-CHD adult controls. If it is true that miR-30a and miR-9 expression level changes are minimal after the heart has formed, our data suggest that miR-30a and miR-9 are likely contributors to the occurrence of CHDs. Because miRNAs are surprisingly stable and detectable in the circulating blood stream, they could serve as novel biomarkers for molecular diagnostics, as well as the promising therapeutic targets after they are validated by further studies. Therefore, future investigations on the interaction of CHD-associated regulatory variants in the master transcription factors and its target miR-NAs will open a door for the improved CHD diagnosis and therapy.
In summary, our study demonstrated a significant effect of TBX5 3′UTR variant rs6489956 on the increased susceptibility of CHD in two independent Han Chinese populations. It was well known that both Tbx5 overexpression and insufficiency result in heart defects in developing embryos. In view of the decreased TBX5 mRNA levels detected in the heart tissues of CHD patients with the CT and TT genotypes, the individuals with the T allele are susceptible to develop CHD due to the TXB5 deficiency. Functional studies clearly showed that the consequent repression of TBX5 expression on the post-transcriptional level resulted from an enhanced binding affinity of the T allele to miRNAs. Our data highlighted an important function of the underlying genetic variations in the 3′UTR of TBX5 in CHDs, providing new insight into risk assessment for these common birth defects.

Study subjects
We analyzed samples from two independent case-control groups. The Shandong group consisted of 905 CHD patients and 606 matched controls recruited from the Cardiovascular Disease Institute, General Hospital of Jinan Military Command (Jinan, Shandong Province, China) between August 2009 and September 2011. The Shanghai group was composed of 272 CHD cases and 384 matched controls enrolled in the Children's Hospital of Fudan University (Shanghai, China) between August 2010 and November 2011. All subjects were unrelated ethnic Han Chinese. All of the controls were non-CHD outpatients from the same geographical area who were matched to the affected individuals in terms of age and sex over the same period. CHDs patients who were considered to be syndromic and not isolated CHDs, or who had a positive family history of CHD in a first-degree relative (parents, siblings and children), were excluded from the study.
To screen for genetic variants in the 3′UTR of TBX5 gene, we selected a subset of 32 subjects for resequencing. Half of these subjects were CHD patients, and half were controls. Fifty-three cardiovascular tissue samples were obtained with informed consent from CHD patients who had undergone cardiac All study protocols were reviewed and approved by the medical ethics committee of Children's Hospital of Fudan University, and written consent was obtained from the parents of subjects and/or the subjects themselves before study enrollment.

Variant identification and genotyping
Genomic DNA was isolated from the peripheral venous blood using conventional regents. TBX5 3′UTR fragments were amplified using PCR to amplify a 1757-bp fragment from 114793401 to 114791645 (NM_000192) in 32 unrelated individuals randomly selected for variant screening by sequencing. Direct dye terminator sequencing of the PCR products was carried out using the ABI Prism BigDye system according to the manufacturer's instructions (Applied Biosystems, Foster City, CA, USA). Selected single-nucleotide polymorphisms (SNPs) with minor allele frequency (MAF) 40.05 were genotyped using SNaPshot analysis (ABI). In the first stage of the present study, a total of 288 CHD cases and the same number of controls of the Shandong group were collected and genotyped in October 2010. The variants associated with CHD risk were chosen for further case-control study in the following Shandong and the Shanghai samples. The samples were sequenced and genotyped on an ABI 3730 automated sequencer and analyzed using Lasergene 7.0 and Peakscanner 1.0, respectively. All of the PCR and sequencing primers were listed in Supplementary Table S2.

Quantitative real-time PCR
Total RNA was extracted from human cardiovascular tissue samples preserved in RNAlater (Qiagen, Valencia, CA, USA) and converted to cDNA using random hexamers, oligo(dT) primers and PrimeScript RT reagent Kit (TaKaRa, Tokyo, Japan). Quantitative Real-time PCR was performed with SYBR Premix Ex Taq (TaKaRa) using a StepOne Real-Time PCR system (ABI) with β-actin as an internal reference gene. Each reaction was performed in triplicate. The primers were described in Supplementary Table S2.

Plasmid
To construct the reporter plasmids, TBX5 3′UTR containing either the C or T allele was amplified from genomic DNA and subcloned into the downstream region of Renilla luciferase using XhoI and NotI restriction sites in psiCHECK-2 vector (Promega, Madison, WI, USA). Primers were listed in Supplementary Table S2. All recombinant clones were verified by DNA sequencing.
Human wild-type TBX5 ORF fused to PCMV6-XL4 was purchased from Origene Technologies (Beijing, China). The mutant TBX5 was generated by the the Quick Change Site-Directed Mutagenesis Kit (Stratagene). Constructs were linearized using PvuII and the capped and polyA tailed mRNA was synthesized with mMessage mMACHINE T7 Ultra Kit (Ambion, Life Technologies) for the zebrafish microinjection.

Cell culture and luciferase reporter assays
Human embryonic kidney 293T (HEK 293T), human cardiac myocytes (HCM) and rat cardiac myocyte (H9C2) cells were TBX5 3′UTR variant increases the risk of CHD 10 grown in Dulbecco's Modified Eagle's Medium supplemented with 10% fetal bovine serum. HEK 293T, HCM or H9C2 cells were seeded in 24-well culture plates at the density of 4.0 × 10 5 , 2.0 × 10 5 and 1.0 × 10 5 per liter, respectively. Twenty-four hours later, the cells were transfected with 400 ng of each TBX5 reporter plasmid using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). To screen the predicted candidate miRNAs, HEK 293T cells were co-transfected with 25 ng of psiCHECK-C/T plasmid and 75 ng of miRNA expression vector in 96-well culture plates. To further explore the interaction between TBX5 3′UTR and selected candidate miRNAs, HEK 293T, HCM or H9C2 cells were co-transfected with 50 ng psiCHECK-C/T plasmid as well as 0 ng, 50 ng, 100 ng, 150 ng or 200 ng of miRNA expression vector. For small interfering RNA assays, HEK 293T cells were co-transfected with 100 ng psiCHECK-C/T plasmid and 100 nM miRNA inhibitors (RiboBio, Guangzhou, China) in 24-well culture plates.
Cells were lysed 24 h after transfection for luciferase assays that employed the Dual Luciferase Reporter Assay System (Promega) according to the manufacturer's instructions. Three independent transfection experiments were performed, and each luciferase assay was conducted in triplicate.

Validation of miRNA regulation
HEK 293T cells were transfected with 2 μg of miRNA expression vector or 100 nM of miRNA inhibitors in 6-well culture plates. Cells were harvested and lysed 48 h after transfection in RIPA buffer and assayed for the relative protein concentrations of intracellular TBX5 (rabbit anti-TBX5 polyclonal antibody; Abcam) and β-actin (Kangchen Bio-tech) using western blot analysis. RNA was also purified from these cells for quantitative RT-PCR analysis of TBX5 mRNA levels.
Surface plasmon resonance SPR analysis was conducted using the ProteOn XPR36 protein interaction array system (Bio-Rad, Hercules, CA, USA). Biotinylated miRNAs were immobilized to an individual channel of the streptavidin-modified sensor chip at a fixed concentration of 400 nM. Single-stranded RNA (GenePharma, Shanghai, China) harboring 32 bp of 3′UTR wild-type-C allele RNA or 3′UTR mutant-T-allele was diluted into a variety of concentrations in PBST buffer (10 mM Na 2 HPO 4 , 150 mM NaCl and 0.005% Tween 20, pH 7.4) and flowed over the biotinylated miRNAs. At the end of each cycle, 5 mM NaOH was applied to regenerate the sensor surface. Data were converted by BIA evaluation software.

Zebrafish, microinjection and imaging
The morpholino (MO) antisense oligonucleotide MO-Tbx5a (5′-GAAAGGTGTCTTCACTGTCCGCCAT-3′) was purchased from Gene Tools (Philomatch, OR, USA). Human miR-30a, mir-9 and a negative control were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Zebrafish were raised under standard conditions at 28.5°C. Each 1-2-cell stage embryo was injected with a constant injection of 5 ng MO, 100 pg TBX5 mRNA and 100 pg miRNAs using a microinjector (Narishige, Japan). Twelve hours post-injection, the dead embryos were removed, leaving only viable embryos that were used for further analysis. Consistent with the previously published studies [18,55], all live embryos were divided into the four categories according to their heart morphologies. At 48-h post fertilization (hpf), images were acquired with an Olympus stereomicroscope microscope or Leica TCS-SP5 LSM confocal microscope. For confocal imaging analysis of zebrafish embryos, they were anesthetized with egg water/0.16 mg ml −1 tricaine/1% 1-phenyl-2-thiourea (Sigma-Aldrich, St Louis, MO, USA) and embedded in 0.6% low melting agarose. Confocal imaging analysis was performed using Imaris software. Two transgenic zebrafish lines: Tg(vmhc:eGFP) and Tg(vmhc: mCherry-NTR) were used as described in previous work [55]. Whole-embryo microRNA sensor assay in zebrafish was carried out as described previously [56].

Statistical analysis
The Hardy-Weinberg equilibrium test was performed on controls using the χ 2 -test. Differences in demographic features and allelic or genotypic frequencies between the CHD cases and controls were evaluated using Pearson's χ 2 -test. To evaluate associations between the genotypes and CHD risk, ORs and 95% CIs were calculated by unconditional logistic regression analysis with adjustments for age and sex. In multiple comparison testing, we employed the Bonferroni correction. The significance level was adjusted to P = 0.0025 given the four SNPs being tested for 5 genetic models in 288 cases and the same number of controls.
Data were calculated as the mean ± s.d. Significant differences between the test and control groups were analyzed by means of an independent t-test with Bonferroni correction. The Po 0.05 was defined as statistical significance. Statistical analyses were conducted with SPSS 16.0 software (SPSS, Chicago, IL, USA). analysis of human data. SSJ provided the microRNAs expression constructs. FW, DL, JYZ, XYY, FZ and LJ contributed to the design and interpretation of the experiments. FW and DL wrote the manuscript and all authors commented and approved the manuscript.