Association of MTHFD Gene Polymorphisms and Maternal Smoking With Risk of Congenital Heart Disease: A Hospital-Based Case-Control Study

Background: MTHFD may affect the embryonic development by elevated homocysteine levels, DNA synthesis and DNA methylation, but limited number of genetic variants of MTHFD was focused on the association with congenital heart disease (CHD). This study examined the role of MTHFD and maternal smoking in CHD risk, and investigated their interaction effects in Chinese populations. Methods: A case-control study of 464 mothers of CHD infants and 504 mothers of health controls was performed. The exposures of interest were maternal tobacco exposure, single nucleotide polymorphisms (SNPs) of maternal MTHFD gene. The logistic regression model was used for accessing the strength of association. Results: Mothers exposed to secondhand smoke during three months before pregnancy (adjusted odds ratio [aOR] = 1.56; 95% con�dence interval [CI]: 1.13-2.15) and in the �rst trimester of pregnancy (aOR = 2.24; 95%CI: 1.57-3.20) were observed an increased risk of CHD. Our study also found that polymorphisms of maternal MTHFD gene at rs1950902 (AA vs. GG: aOR = 1.73, 95% CI: 1.01-2.97), rs2236222 (GG vs. AA: aOR = 2.38, 95% CI: 1.38-4.12), rs1256142 (GA vs.GG: aOR = 1.57, 95% CI: 1.01-2.45) and rs11849530 (GG vs. AA: aOR = 1.68, 95% CI: 1.02-2.77) were signi�cantly associated with higher risk of CHD. Furthermore, we found the different degrees of interaction effects between polymorphisms of the MTHFD gene including rs1950902, rs2236222, rs1256142, rs11849530 and rs2236225, and maternal tobacco exposure.


Background
Congenital heart defect (CHD) was often de ned as a structural or functional abnormality of the heart and/or great vessels that were present at birth 1 .Among all recognized structural birth defects, CHD was the most common and severe, with 4 to 10 cases per 1000 live births, which imposed a huge economic burden on the society and family 2 .Although the past few decades have seen a rapidly growing interest in exploring the etiology of CHD, the pathogenesis of most CHD cases remains unknown 3,4 .So far, folate supplementation was the most effective intervention for decreasing CHD 5,6 , while the folate-cycle product homocysteine might affect fetal heart development by disruption of gene methylation, increasing oxidative stress and homocysteinylation of key proteins 5,7 , which indicated that the occurrence of CHD was highly responsive to changes in genes related to maternal folatehomocysteine metabolism.The methylenetetrahydrofolate dehydrogenase (MTHFD) gene, located on chromosome 14q24, encoded the trifunctional enzyme MTHFD (5,10-methylenetetrahydrofolate dehydrogenase 8 , 5,10-methenyltetrahydrofolate cyclohydrolase, and 10-formyltetrahydrofolate synthetase).This enzyme catalyzed three sequential reactions in the interconversion of tetrahydrofolate (THF) to 5,10-methylenetettrahydrofolate (5,10-methylene THF), the crucial substrate for 5methyltetrahydrofolate (5-methlyTHF), which was required for DNA synthesis, DNA repair and provide the methyl donor for regeneration of methionine from homocysteine for subsequent methylation reactions 9,10 .Plausible mechanisms of the MTHFD gene in CHD susceptibility might involve restricted DNA synthesis, high levels of homocysteine, and DNA methylation 5,[11][12][13][14] .The experimental studies had revealed that the MTHFD gene with mutant genotypes expressed less stable in vitro and low active in vivo MTHFD protein 11,12 , which consequently disturbed de novo purine synthesis and impacted DNA synthesis.Moreover, the epidemiologic studies suggested that polymorphisms of the MTHFD gene such as rs2236225 and rs1950902 were closely related to homocysteine levels [14][15][16][17] and DNA methylation 16 , and these data published indicated the plausible association between genetic variants of the MTHFD gene and the risk of CHD.MTHFD R653Q (rs2236225 G→A) and MTHFD R134K (rs1950902 G→A) were the two most well-studied polymorphisms, but previous efforts involved in their associations with CHD risk had yielded con icting or negative results 11,17,18 , which might result from insu cient statistical power and different methods.Notably, previous studies focused mainly on a small number of functional nonsynonymous single-nucleotide polymorphisms (SNPs) of the MTHFD gene with known biochemical phenotypes such as rs1950902 and rs2236225; the other signi cant variants have been largely ignored; thus this study represents both the rst report and replication efforts in Han Chinese populations.
It had been reported that 85% of CHD resulted from a complex interplay of genetic variants and environmental factors 19 .Maternal tobacco exposure in the periconceptional period, de ned as 3 months before pregnancy through the rst trimester of pregnancy, was one of the most common environmental factors affecting abnormal fetal development 20 .Amounts of epidemiologic studies about the associations of maternal tobacco exposure in the periconceptional period involved in active smoking and passive smoking and CHD risk were showed heterogeneous results, indicating that people had different susceptibility to the effects of tobacco exposure [21][22][23][24][25] .Convincing evidence showed that it was clear that particular genotypes of metabolizing systems and DNA repair pathways might modulate the effect leading to varying susceptibility to the CHD of tobacco exposure 26 .The maternal tobacco exposure was observed associations with substantial reductions of folate levels in plasma 27,28 as well as in cord blood 29 , and red blood cells 30,31 , even after correcting for folate intake 32 , which implied the interactive association between maternal tobacco exposure and polymorphisms of the MTHFD in CHD susceptibility.In addition, a recent study suggested that maternal folate levels might partly modify the in uence of maternal tobacco exposure during pregnancy on the DNA methylation of the newborn epigenome and therefore affected embryonic development 33 .Based on the above, we hypothesized that there existed the interaction effects between maternal tobacco exposure and the MTHFD genetic variants, namely that the two factors jointly caused the CHD.In our study, a hospital-based case-control design based on the Han Chinese population was performed with the following objectives: (i) to assess the association of genetic variants of maternal MTHFD gene with risk of CHD in offspring; (ii) to examine whether maternal smoking including active and passive smoking was signi cantly associated with risk of CHD in offspring; and (iii) to analyze the interaction effects between maternal smoking and the MTHFD genetic variants for CHD.

Study design and recruitment of study participants
For requirements for a speci c article type please refer to the Article Types on any Frontiers journal page.The main characteristics of the participants and research procedure had been described by our previous study 34 .Recruitment was conducted by the Hunan Provincial Children's Hospital (Changsha, Hunan Province, China).A total of 464 CHD patients and their mothers were consecutively enrolled from the Department of Cardiothoracic Surgery between November 2017 and December 2019.The non-CHD patients were from the Department of Child Healthcare in the same hospital during the same time period and were matched to the affected individuals by age and sex.The controls included 504 non-CHD patients who were without any congenital malformations after medical examination and their mothers.All CHD cases were diagnosed using echocardiography and con rmed by surgery.The CHD patients with structural malformations involving another organ system or known chromosomal abnormalities were excluded.Considering the homogenous ethnic background may reduce residual confounding factors from genetic and cultural differences, we only included the Han Chinese descent.We further excluded mothers who achieved pregnancy by assisted reproductive technology including in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) and reporting.Again, mothers who reported a history of depression or other psychiatric disorders or were diagnosed with depression or a psychiatric illness were also excluded.We classi ed the 464 CHD cases into 7 broad categories as described previously 35 .In particular, 28 (6.0%) had conotruncal defects, 360 (77.6%) had septation defects, 11 (2.4%) had left ventricular out ow tract obstruction, 17 (3.7%)had right ventricular out ow tract obstruction, 16 (3.4%)had anomalous pulmonary venous return, 20 (4.3%) had complex CHD, and 12 (2.6%) had other CHD defects (Table S1).
Our study was approved by the ethics committee of the Xiangya School of Public Health of Central South University, and written informed consent was obtained from all mothers.Besides, we have registered this study in the Chinese Clinical Trial Registry Center (registration number: ChiCTR1800016635).

Information collection
A self-designed questionnaire was used to collect the corresponding information by specially trained investigators.This questionnaire was developed by experts in the eld of CHD research and administered to eligible mothers (test-retest reliability=0.833;Cronbach's alpha=0.782).In the present study, we collected maternal sociodemographic characteristics (i.e., age at pregnancy onset (years) residence location, maternal education level (years) and annual income in the past 1 year (RMB)), history of adverse pregnancy outcomes (i.e., spontaneous abortion, stillbirth, preterm birth, gestational diabetes mellitus, gestational hypertension, and premature rupture of membranes), family history (i.e., consanguineous marriages, and history of congenital heart disease), cold or fever in the periconceptional period, and personal lifestyle and habit in the periconceptional period including drinking alcohol, drinking tea, living near environmental pollution source, dyeing hair or perming, decorating housing and folate use.
The mentioned above information was further con rmed by consulting their Maternal and Child Health Manual and medical records.In China, each pregnant woman will be provided with a Maternal and Child Health Manual, which will record their basic demographic characteristics, behavioral habits, illness, and the results of various medical examinations during pregnancy.We also examined the SNPs of the maternal MTHFD gene, which were described below.

Sequencing of MTHFD gene and genotyping
The MTHFD gene was the candidate gene for the present study.When mothers completed the questionnaires mentioned above, they were asked to provide 3 to 5 milliliters of peripheral venous blood for genotyping.Methods for DNA extraction and genotyping have been described previously 34 .The laboratory technician, who performed the genotyping, retyped and double-checked each sample, and recorded the genotype data, was blinded to whether the samples were from cases or controls.SNP markers were selected using the SNPBrowser™ program (version 3.0) provided by AppliedBiosystems Inc.This program allowed the selection of SNP markers from the HapMap database.For each target gene, tagging SNPs were selected based on the pairwise r 2 ≥ 0.8.However, we excluded these SNPs with minor allele frequencies less than 10% in Caucasians.We imposed a minimum SNP genotyping call rate at the level of 50%, which was applied to ensure data integrity of the individual's genotypes that had been called.And successful rates for SNPlex assays were >96% for 4 SNPs, from 90% to 96% for 1 SNPs.

Statistical analysis
Statistical analysis was performed using R software, version 3.5.0(R Foundation for Statistical Computing).All tests were performed signi cantly for a two-sided P value not exceeding 0.05, except where otherwise speci ed.Additionally, the false discovery rate (FDR) control was used in this study to correct for multiple testing.The statistically signi cant results were those with the false discovery rate P value (FDR_P) < 0.1.Qualitative data were described using frequencies and percentages, and quantitative data were described using means and standard deviations (SDs).Hardy-Weinberg equilibrium (HWE) was tested for the control group (signi cance level at P <0.01).We used a two-phase analytical method based on genetic model selection (GMS) to test associations between SNPs and CHD, and the speci c calculation process had been described previously 36 .The genetic models contained the dominant model (calculated for mutant type homozygote versus wild type homozygotes and heterozygote), the recessive model (calculated for heterozygotes and mutant type homozygotes versus wild type homozygotes), and the addictive model (calculated for wild type homozygotes versus heterozygote versus mutant type homozygote).We classi ed the genetic model into the recessive model if Z HWDTT > c, the dominant model if Z HWDTT < -c, and in the addictive model if otherwise, where we chose c = Φ -1 (0.95) = 1.645.The Pearson χ 2 test was used to compare the differences of nominal variables across groups.And for ordinal categorical variables, Wilcoxon rank sum test was used.Odds ratios (ORs) and their 95% con dence intervals (CIs) were used to show the level of association.Crude ORs were calculated by univariate logistic regression, while adjusted ORs (aORs) were calculated by multivariable logistic regression.We used logistic regression and controlled for potential confounders, to analyze the main effects and interactive effects of the gene-environment interaction of maternal MTHFD gene and smoking experiences for risk of CHD in offspring.Of note, in the present study, we focused only on the risk of total CHD associated with maternal smoking and genetic variants of the MTHFD gene and did not assess the risk of speci c CHD subtypes due to the limited number of sample sizes for these subtypes.

Baseline characteristics of study population
After considering the inclusion criteria, we nally recruited 464 CHD cases and their parents into the case group and 504 health infants and their parents into the control group.Comparisons of baseline characteristics across groups were summarized in Table 1.Our study showed that there were statistically signi cant differences between two groups for the following characteristics: maternal education level (years), annual income in the past 1 year (RMB), history of adverse pregnancy outcomes, consanguineous marriage, history of congenital malformations in family, cold or fever in the periconceptional period, and personal lifestyle and habit in the periconceptional period including drinking alcohol, drinking tea, living near environmental pollution source, dyeing hair or perming and folate use (all P values < 0.05).Thus, these factors were adjusted when accessing the association of maternal tobacco exposure, the genetic variants of the maternal MTHFD gene, and their interactions with the risk of CHD in offspring.The exposure occurred in the periconceptional period.
d P < 0.05 was considered to indicate a statistically signi cant difference.
Maternal tobacco exposure and risk of CHD in offspring Table 2 showed the association between maternal smoking and the risk of CHD in offspring.The prevalence rate of active smoking in 3 months before pregnancy in our controls (2.0%) was lower than the smoking rate among Chinese women in China Adult Tobacco Survey Report in 2015 (2.7%) 37 .None of the mothers in cases and controls reported active smoking in the rst trimester of pregnancy.Mothers who reported active smoking in 3 months before pregnancy had an increased risk of CHD in offspring compared with the controls (P <0.001), but this association was not independent of potential confounders (P = 0.052).After adjustment for baseline data, mothers exposed to secondhand smoke in 3 months before pregnancy were observed an increased risk of CHD in offspring (aOR = 1.56; 95% CI: 1.13-2.15).Additionally, the risk of CHD in offspring was signi cantly higher among mothers who were exposed to secondhand smoke in the rst trimester of pregnancy (aOR = 2.24; 95% CI: 1.57-3.20).Abbreviations: CI = con dence interval.
a Data presented as number (percentage) unless otherwise indicated.
b Adjusted for maternal education level (years), annual income in the past 1 year (RMB), history of adverse pregnancy outcomes, consanguineous marriage, history of congenital malformations in family, cold or fever in the periconceptional period, and personal lifestyle and habit in the periconceptional period including drinking alcohol, drinking tea, living near environmental pollution source, dyeing hair or perming and folate use.Genotypes Frequencies of SNPs and the results of HWE tests and GMS The genotype frequencies for each SNP of the maternal MTHFD gene and the results of HWE tests were summarized in Table S2.The HWE tests showed that the genotype frequencies of the 5 SNPs of maternal MTHFD gene in the control group were all within HWE (all P values >0.01).The results of the GMS of each SNP were presented in Table S3.The genetic models of SNPs including rs1950902, rs2236225, rs2236222 and rs11849530 were all classi ed into the addictive model since Z HWDTT > -1.645 and Z HWDTT < 1.645.The genetic model of rs1256142 was classi ed into the dominant model because of Z HWDTT < -1.645.We initially ascertained the genetic models of overall SNPs of the maternal MTHFD gene, which was used for accessing the association between each SNP and risk of CHD in offspring based on the corresponding genetic model.

Genetic variants of maternal MTHFD gene and risk of CHD in offspring
The association between each maternal SNP of the MTHFD gene and the risk of CHD in the Han Chinese population was shown in Table 3.The univariate analyses suggested that there were statistically signi cant differences for the genetic variants at rs1950902 (AA vs. GG: P = 0.006; the addictive model: P = 0.002), rs2236222 (GG vs. AA: P = 0.001; the addictive model: P = 0.001) and rs1256142 (GA vs. GG: P = 0.035) Interactions between maternal smoke exposure and MTHFD gene for risk of CHD We modestly identi ed the four polymorphisms including rs1950902, rs2236222, rs1256142, and rs11849530 with signi cant main effects on CHD risk in the Han Chinese population.Moreover, though rs2236225 was not observed a signi cant main effect on CHD risk, this polymorphism was the most extensively studied one.Thus, we kept the 5 SNPs of the MTHFD gene including rs1950902, rs2236222, rs1256142, rs11849530 and rs2236225 for the interactions analysis.Interactions between maternal SNPs of MTHFD gene in the corresponding genetic model and maternal smoke exposure on CHD risk were summarized in Table 4.Our results showed there were statistically signi cant interaction effects between active smoking in 3 months before pregnancy and genetic variants of maternal MTHFD gene at rs1950902, rs2236225 and rs11849530.Speci cally, mothers with GG/GA genotypes at rs1950902 (the addictive model: aOR = 1.44, 95% CI: 1.02-2.04),AA/GA genotypes at rs2236225 (the addictive model: aOR = 2.15, 95% CI: 1.14-4.05)and GG/GA genotypes at rs11849530 (the addictive model: aOR = 2.10, 95% CI: 1.21-3.64)generated a 1.44-fold, 2.10-fold and 2.15-fold increased CHD risk when they smoked in 3 months before pregnancy, respectively.Additionally, maternal passive smoking was also observed interaction effects with MTHFD gene at rs1950902, rs2236225, rs2236222, rs11849530 and rs1256142 in the Han Chinese population.To be speci c, the mothers with GG/GA genotypes at rs1950902 (the addictive model: aOR = 1.22,95% CI: 1.07-1.40),AA/GA genotypes at rs2236225 (the addictive model: aOR = 1.38, 95% CI: 1.12-1.70),GG/GA genotypes at rs11849530 (the addictive model: aOR = 1.35, 95% CI: 1.13-1.61)or AA genotype at rs1256142 (the dominant model: aOR = 1.26, 95% CI: 1.06-1.50)had signi cantly higher CHD risk in offspring when they were exposed to secondhand smoke in 3 months before pregnancy.Moreover, when mothers carried the AA/GA genotypes at rs1950902 (the addictive model: aOR = 1.38, 95% CI: 1.18-1.60),AA/GA genotypes at rs2236225 (the addictive model: aOR = 1.62, 95% CI: 1.29-2.04),GG/GA genotypes at rs2236222 (the addictive model: aOR = 1.38, 95% CI: 1.11-1.72),GG/GA genotypes at rs11849530 (the addictive model: aOR = 1.64, 95% CI: 1.34-2.00)or AA genotype at rs1256142 (the dominant model: aOR = 1.53, 95% CI: 1.26-1.86),exposure to secondhand smoke in the rst trimester would increase the susceptibility to suffer from a CHD-affected delivery.

Discussion
Convincing evidence implies that periconceptional intake of folic acid leads to a 40% to 60% reduction in the risk of a CHD-affected delivery, which makes investigating the association between genetic polymorphisms in genes related to folate metabolism and CHD risk an attractive pursuit 38 .The MTHFD gene plays a key role in the folate-homocysteine metabolism pathway, encoding a single protein with three catalytic properties crucial for DNA synthesis, DNA repair, and methylation reactions.The epidemiologic and experimental studies indicate the plausible association between genetic variants of the MTHFD gene and the risk of CHD [11][12][13][14]39 , but previous efforts on the association have yielded controversial results 40 and are limited to a small number of functional nonsynonymous polymorphisms.
This is therefore the rst study to explore the other variants in the coding region of the MTHFD gene on CHD risk and represent replication efforts in Han Chinese populations.Furthermore, folate-homocysteine metabolism may partly modulate the effect leading to varying susceptibility to the CHD of tobacco exposure 26,33 .Thus this study also seeks to examine the interactions of maternal tobacco exposure and polymorphisms of the MTHFD gene on CHD risk, which may help to provide new clues for future etiological research and intervention of CHD.
In the present study, four genetic variants of maternal MTHFD gene including rs1950902, rs2236222, rs1256142 and rs11849530 were revealed to have signi cant associations with an increased risk of CHD in case-control studies based on the Han Chinese population (Figure .1 B).The information from the public databases and the literature revealed unequivocally these SNPs were all within the coding region; one was functional nonsynonymous polymorphism with a known biochemical phenotype and the other were synonymous ones 8,11 .Notably, the well-studied polymorphism rs1950902 G→A (MTHFD G401A) leads to an arginine (G allele) to a lysine (A allele) substitution, lying within the dehydrogenase/cyclohydrolase domain of MTHFD protein.The mutant genotypes of the MTHFD gene at rs1950902 may in uence the stability of the enzyme and change the catalytic activity 17 , causing disturbance of the folate-homocysteine metabolism pathway and therefore may alter folate or homocysteine levels.It was reported that rs1950902 was signi cantly related to elevated plasma homocysteine and reduced folate levels 16,39 .These observations suggested that MTHFD rs1950902 could affect embryonic development employing restricted DNA synthesis and a high level of homocysteine.Considering the results of the present study and the phenotypes related to rs1950902, we postulate that this polymorphism plays a role in fetal heart development.One previous study 41 , however, reported negative results that the signi cant association between MTHFD 1950902 and tetralogy of Fallot were not observed.Due to our limited sample size of each subtype of CHD cases, we didn't analyze the association between speci c CHD subtypes and genetic variants of maternal MTHFD gene.Besides, the present study modestly found the synonymous polymorphisms within the intronic region including rs2236222, rs1256142 and rs11849530 were associated with increased CHD risk, and the plausible mechanism of these polymorphisms increasing CHD susceptibility was likely to affect codon usage and translational e ciency 42 .
Particularly, it was worth mentioning that MTHFD rs2236225 (MTHFD R653Q), the most investigated genetic variant of the MTHFD gene, leads to an arginine (G allele) to glutamine (A allele) substitution in the MTHFD protein.Convincing evidence suggested that mutant type protein of MTHFD gene at rs2236225 caused signi cant DNA synthesis restriction 11 and increased homocysteine levels 14,16,39 .In addition, the speci c CHD subtypes such as atrial septal defects 13 , tetralogy of Fallot 43 and aortic stenosis 11 had been reported signi cant associations with this polymorphism.In this present study, however, we did not observe the signi cant association between MTHFD rs2236225 and CHD risk, which may partly be explained by the different susceptibility of folate-homocysteine imbalance to every subtype of CHD and the relatively low proportion of the speci c high susceptibility to subtypes in our study 5 .
Overall, some of the polymorphisms involved in our study had not been con rmed before, and literature involved in the association between genetic variants of maternal MTHFD gene and CHD was still lack.It needs further and clearer evidence to gure out the mechanism.
We also examined the association between maternal smoking including active and passive smoking and the risk of CHD in offspring.Findings from the present study suggested that mothers who reported passive smoking at home or in the workplace 3 months before pregnancy and the first trimester of pregnancy were observed a 1.56-fold and 2.24-fold increased CHD risk, respectively, which was basically consistent with a recent meta-analysis 40 .Obviously, the maternal passive tobacco exposure in the first trimester of pregnancy was shown more harmful than in 3 months before pregnancy, which may be partly explained by the fact that the former was the sensitive period for fetal heart development.Additionally, a great number of previous epidemiologic studies supported that periconceptional active smoking was associated with risk of CHD in offspring 40 .However, we modestly did not observe that maternal active smoking in 3 months before pregnancy could increase CHD risk after adjusted potential confounders, and even none of our subjects reported active smoking during pregnancy.The causes for the insigni cant association between maternal active smoking in 3 months before pregnancy and CHD risk may be due to our limited sample size and the subjective smoking records.There was a possibility that pregnant smokers underreported their smoking and such potential misclassi cation might lead to underestimation of the impact of maternal active smoking on CHD.The behavior was attributed to medical and societal pressures that made pregnant women reluctant to report their smoking activities 44 .
Our results also showed the different degrees of interaction effects between polymorphisms of MTHFD gene including rs1950902, rs2236222, rs1256142, rs11849530 and rs2236225 and maternal tobacco exposure (Figure .1 B).As shown in the gure, active smoking before pregnancy seemed more harmful interacted with the SNPs of the MTHFD gene on CHD risk, compared with passive smoking before pregnancy or in the rst trimester.Of note, among these polymorphisms, MTHFD rs2236225 was relatively the most effective one to modify the association between maternal tobacco exposure and CHD risk (aOR=2.15).However, studies concerning the interactions of SNPs in folate-related genes and maternal tobacco exposure were lack, Hobbs 45 indicated that the combined effect of elevations in maternal homocysteine, smoking, and the MTHFR 677C>T polymorphism increased the risk of having a CHD-affected pregnancy (aOR=11.8).Possible mechanisms by which the MTHFD gene interacted with maternal tobacco exposure to increase CHD susceptibility include elevated serum homocysteine, increased DNA methylation, and DNA synthesis restriction.It was well-studied that the maternal tobacco exposure was observed associations with substantial reductions of folate levels [27][28][29]32 , elevated levels of homocysteine 46 in cord blood, and altered global genomic methylation. Th polymorphisms of the MTHFD gene also showed a close relation to plasma homocysteine and DNA methylation. Hece, the mutant type of MTHFD protein and periconceptional tobacco exposure jointly caused the plasma homocysteine elevated, and high levels of homocysteine consequently affected fetal heart development by disruption of gene methylation, increasing oxidative stress and homocysteinylation of key proteins (Figure .1 A).Additionally, a recent study suggested that maternal folate-homocysteine metabolism may partly modify the in uence of maternal tobacco exposure during pregnancy on the DNA methylation of newborn epigenome and therefore affected embryonic development 33 , which provided indirect evidence to support our ndings.Nevertheless, speci c mechanisms are unclear and need further research.
The limitations of the study need to be addressed.First, based on a case-control study, information on maternal tobacco exposure was collected through self-reported interviews, and therefore recall bias inevitably had to be taken into consideration.To reduce recall bias to some extent, the exposure information was further con rmed by consulting their Maternal and Child Health Manual and medical records.Second, despite adjusting many confounders, potential confounding factors cannot be entirely ruled out.Third, there are so many genes that are also involved in cardiovascular development, but we only focused on the MTHFD gene.Fourth, considering population strati cation bias in epidemiologic studies, we recruited the participants restricted to the Han Chinese ethnicity, and further work was needed to estimate the effect of the MTHFD gene and maternal tobacco exposure in CHD risk within other populations.Fifth, sample size limitations prevented us from examining speci c CHD subtypes.

Conclusions
The current ndings observed that maternal polymorphisms of the MTHFD gene at rs1950902, rs2236222, rs11849530, and rs1256142 were signi cantly associated with the risk of CHD in offspring.In addition, a positive association between maternal passive smoking in the periconceptional period and risk of CHD was found.Furthermore, our results showed the different degrees of interaction effects between polymorphisms of the MTHFD gene including rs1950902, rs2236222, rs1256142, rs11849530 and rs2236225, and maternal tobacco exposure.It seemed that MTHFD rs2236225 was relatively the most effective one to modify the association between maternal tobacco exposure and CHD risk among these polymorphisms, and maternal active smoking was more harmful interacted with SNPs of the MTHFD gene on CHD risk, compared with passive smoking.However, considering the complexity of the mechanism and the limitation of sample size, more studies in different ethnic populations with a larger sample and prospective designs were required to con rm our ndings.

cP
< 0.05 was considered to indicate a statistically signi cant difference.

Figures
Figures

Figure 1 Possible
Figure 1

Table 1 .
Baseline characteristics in case and control groups a a Data presented as number (percentage) unless otherwise indicated.bTheWilcoxon rank-sum test method was used; otherwise, the χ 2 test was used.c

Table 2 .
Maternal smoking and risk of congenital heart defects in offspring a

Table 4 .
Interactions between SNPs of MTHFD gene and maternal smoking detected by logistic regression Adjusted for maternal education level (years), annual income in the past 1 year (RMB), history of adverse pregnancy outcomes, consanguineous marriage, history of congenital malformations in family, cold or fever in the periconceptional period, and personal lifestyle and habit in the periconceptional period including drinking alcohol, drinking tea, living near environmental pollution source, dyeing hair or perming and folate use.
a b FDR_P < 0.1 was considered to indicate a statistically signi cant difference.