Association of maternal multiple serum metal/metalloid concentrations and metabolic gene polymorphisms with offspring skeletal defects: a multicenter case-control study

Abstract

Background

Skeletal defects (SDs) can be caused by various factors, including genetic and environmental factors, as well as gene-environment interactions. This study aimed to investigate the association of maternal serum concentrations of metals/metalloids and metabolic gene polymorphisms with offspring SDs risk, including their potential interactions.

Methods

In this multicenter case-control study, 316 pregnant women with offspring SDs and 361 healthy controls were recruited from 49 hospitals in China. Serum concentrations of aluminum (Al), iron (Fe), zinc (Zn), selenium (Se), strontium (Sr), molybdenum (Mo), and cadmium (Cd) were measured. Genotyping of 63 single nucleotide polymorphisms (SNPs) across 30 metal/metalloid metabolism-related genes was performed on first- and second-trimester blood samples. Associations were assessed using logistic regression and generalized multifactor dimensionality reduction.

Results

Median maternal serum concentrations of Fe, Zn, Sr, and Mo were significantly lower in cases than controls (all P < 0.001), while Cd was higher (P = 0.017). Higher (second/third tertile) concentrations of Fe, Zn, Sr, and Mo were associated with a decreased risk of offspring SDs. Notably, the protective association was strongest for Mo (adjusted odds ratio, aOR = 0.10–0.15). In contrast, higher concentrations of Se and Cd were associated with an increased risk (aOR = 2.04–2.43). No significant associations were found between maternal metabolic SNPs and offspring SDs risk after false discovery rate correction, and no significant gene-metal/metalloid interactions were observed.

Conclusions

The lowered maternal serum concentrations of the essential trace elements Fe, Zn, and Mo, as well as of the non-essential element Sr, were associated with an increased risk of offspring SDs, with Mo showing a particularly strong protective association. Cd showed a positive association, though its interpretation warrants caution. Larger-scale and prospective cohort studies on the effects of maternal serum concentrations of metals/metalloids and genetic polymorphisms on offspring SDs are warranted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-026-09023-2.

Background

Skeletal defects (SDs), encompassing limb malformations and skeletal dysplasia, are among the most prevalent birth defects in China, affecting approximately 1 in 3000 individuals, and ranking second in the incidence of birth defects nationwide [1–3]. Because of the skeletal deformities and disabilities, repeated surgical corrections are sometimes needed and usually result in limited functional recovery, thus SDs seriously affect physical and mental health of children, and impose a heavy burden on the families and the society [4, 5]. Embryonic SDs involve complex gene regulation and pathways and may be caused by genetic and/or environmental factors. A study showed that less than 5% of scoliosis cases could be clearly attributed to a genetic mutation, while most cases of spinal malformations were considered to have a significant environmental cause [6]. Therefore, it is necessary to investigate the role of environmental factors and gene-environment interactions in the etiology of SDs.

Maternal nutrition during pregnancy plays a fundamental role in fetal organogenesis, with trace elements representing the critical modulators of developmental pathways. Several elements, including iron (Fe), copper (Cu), zinc (Zn), selenium (Se), strontium (Sr), and molybdenum (Mo), are essential for metabolic processes, cellular functions, and successful pregnancy [7–10]. Maternal metals/metalloids can cross the placental barrier resulting in fetal exposure [11]. Adequate intake of essential trace elements is critical for fetal development. The World Health Organization has outlined the physiological roles and estimated safe intake ranges for key elements such as Fe, Zn, and Se in human nutrition, highlighting their importance throughout the life course, including pregnancy [12]. A recent systematic review confirms that prenatal exposure to some trace elements such as Zn and Cu appears to be beneficial for child growth and development [13]. Notably, both deficiency and excess of most nutrients follow a U-shaped risk curve, wherein deviations from optimal physiological ranges increase the risk of adverse outcomes [14]. Conversely, toxic heavy metals like lead (Pb), cadmium (Cd), and mercury (Hg) generally exert negative effects on child growth parameters and development [13]. Despite these insights, there have been very few human studies examining the specific relationship between maternal metal/metalloid exposure and the risk of offspring SDs.

Individual susceptibility to metal exposure is modulated by genetic variations in metal-handling proteins. Single nucleotide polymorphism (SNP) in genes such as MT2A (metallothionein 2 A), VDR (vitamin D receptor), and TF (transferrin) can alter metal absorption, distribution, and detoxification, thereby influencing the effective fetal dose [15–18]. For instance, the MT2A rs28366003 variant has been linked to reduced cellular protection against Cd, Zn, Cu, and Pb accumulation [15, 16], while VDR rs10735810 may modify placental Pb retention and the associated risks [17]. However, few studies have explored the relationship between maternal genetic polymorphisms related to metal/metalloid metabolism and the risk of offspring SDs, and there is a lack of studies of the interaction between maternal metabolic gene polymorphisms and metals/metalloids with respect to its effect on the risk of offspring SDs.

Therefore, the present study aimed to investigate the association of maternal serum concentrations of metals/metalloids, aluminum (Al), Fe, Zn, Se, Sr, Mo, and Cd, with the risk of SDs in offspring. Further, we intended to learn if SNPs related to metal/metalloid metabolism are associated with the risk of offspring SDs and if the SNPs interact with maternal serum metal/metalloid concentrations in affecting the risk of offspring SDs.

Methods

Study population

The study was based on a large cohort study that recruited pregnant women from 49 hospitals in China between August 2018 and December 2021, with the goal of recruiting at least 150,000 pregnant women and constructing a corresponding biobank. Participant eligibility required a gestational age of < 14 weeks at enrollment, intention to deliver at a collaborating hospital, and provision of an informed consent. Exclusion criteria included fetal death at enrollment, HIV positivity, and conditions compromising questionnaire compliance. Participants were followed across five time points: first (< 14 weeks), second (22–26 weeks), and third (32–36 weeks) trimesters, delivery, and within 42 days postpartum. At each trimester, questionnaires were administered with investigator guidance, and maternal blood samples (4 mL) were collected in the first two trimesters. Serum samples, separated after 30-minute standing and centrifugation, were stored at -80 °C. Detailed cohort profile is available from our prior publication [19].

In the present study, the case group was defined as pregnant women of fetuses with isolated, non-syndromic SDs including but not limited to polydactyly, syndactyly, macrodactyly, brachydactyly, split-hand–foot malformation, fatal bone dysplasia, osteogenesis imperfecta, chondrodysplasia, multiarticular contracture, and clubfoot disorders. These defects were identified by prenatal ultrasound during mid-pregnancy and/or further confirmed through corresponding postnatal examinations, pathological anatomy for aborted fetuses, and telephone follow-ups of live births within 42 days. All potential cases were reviewed by a panel of clinical experts (pediatricians and clinical geneticists) who confirmed the diagnosis based on clinical records, imaging reports, and/or surgical notes, following the International Classification of Diseases (ICD-10) codes for congenital malformations of the musculoskeletal system (Q65-Q79). The exclusion criteria were as follows: (1) fetuses with syndrome diseases and chromosomal abnormalities, (2) pregnant women with multiple pregnancies, and (3) mothers with a family history of SDs. The control group was selected from the same cohort. The inclusion criteria were: (1) absence of any significant maternal or fetal conditions, (2) recruitment from the same hospital as the matched case, (3) matched on the gestational week at blood collection, (4) matched on the date of blood sampling. If multiple eligible controls were identified for a case, all controls were retained in the study. Ultimately, a total of 316 SDs cases and 361 controls were included in the nested case-control study, referring to the flow chart of the study (Fig. 1).

Fig. 1.

Flow chart of the study

Determination of maternal serum concentrations of metals/metalloids, quality assurance, and quality control

Maternal serum samples were analyzed by inductively coupled plasma mass spectrometry (ICP-MS) using an Agilent 7500cx ICP-MS system (Agilent Technologies, Wilmington, DE) equipped with a G3160B I-AS integrated autosampler. The quantitative analysis of 22 trace elements, including manganese (Mn), cobalt (Co), nickel (Ni), Cd, lanthanum (La), tungsten (W), Hg, Mo, antimony (Sb), As, Sr, Pb, tin (Sn), chromium (Cr), Fe, Cu, Zn, Se, Al, thallium (Tl), cerium (Ce), and barium (Ba) was performed at the West China School of Public Health in Sichuan University. Out of these 22 metals/metalloids, only 7 fulfilled our criteria of having a detection rate greater than 50%. Specifically, 250 µL aliquot of maternal serum was digested with 1 mL of concentrated nitric acid using a microwave digestion system. Following digestion, the remaining acid was evaporated to near dryness at 80 °C. The digestion vessel was then washed several times with ultrapure water, and the combined rinses were transferred to a graduated flask and brought to a final volume of 5 mL with ultrapure water. A digestion blank was prepared concurrently using the same procedure. The above samples were injected into the ICP-MS instrument for determination and quantified by internal standard curve method, for which 209Bismuth serves as the internal standard for 208Pb, 205Tl, 202Hg, 184 W; 115Indium as the internal standard for 140Ce, 139La, 121Sb, 118Sn, 111Cd, 98Mo, 88Sr, 82Se, 137Ba; and 45Scandium as the internal standard for 75As, 66Zn, 63Cu, 60Ni, 59Co, 57Fe, 55Mn, 52Cr, and 27Al. The samples from the cases and the controls were randomized independently (random sequence was generated by a computer software). In the same analytical runs, the samples were loaded in a mixed manner (cycles of a random case sample next to a random control sample) to avoid the systematic error caused by the continuous loading of samples of the same group.

The conditions for ICP-MS instrument were as follows: RF power 1550 W, carrier gas flow rate 1.05 L/min, sampling depth 8.0 mm, atomizer pump speed 0.10 rps, atomization chamber temperature 2 ℃, extraction lens − 200.0 V, the fourth stage rod deflection voltage − 3.0 V, the eighth stage rod RF power 150 V, energy discrimination 5.0 V. All metal elements were measured using Helium (He) collision mode, with He air velocity 1.0 L/min.

The ICP-MS method was validated in the laboratory for key performance parameters, including the limit of detection (LOD), limit of quantification (LOQ), precision, and accuracy. Specifically, the LOD and LOQ were calculated as 3 times and 10 times the standard deviation, respectively, obtained from repeated measurements of a pooled human serum matrix. Precision was assessed through replicate analyses. Intra-day precision (relative standard deviation, RSD) ranged from 1.08% to 5.68% (n = 8 replicates analyzed in one day), while inter-day precision ranged from 2.33% to 8.79% (n = 8 replicates analyzed over three consecutive days). Method accuracy was evaluated via a spike-and-recovery experiment. Five replicate samples were spiked at each of three concentration levels (LOQ, 2×LOQ, and 5×LOQ), yielding recoveries between 82.2% and 121.5%. Furthermore, the method’s reliability was verified by analyzing a certified reference material (ClinChek^®-Control serum for trace elements, Level I, Recipe, Germany). The validation results are shown in Table S1. All measured values were within the specified quality control ranges. Samples with concentrations below the LOD were assigned the value of LOD/.

To ensure analytical accuracy and reliability, a calibration curve was constructed for each batch. Before injecting the calibration standards, the dilution solvent was injected into the instrument to assess background contamination from the solvent and instrumentation. A mid-level calibration standard, a sample solution, and a quality control sample were analyzed at an approximate frequency of one per twenty test materials, and the responses were required to be within 90–110% of the average value. In addition, the internal standards recovery rates were also observed, and must be within 80–120%. A procedural blank was included in each batch to detect potential background contamination throughout the experiment. Quantification was performed using an internal standard calibration curve, requiring back-calculated concentrations of calibration points to be within ± 15% of the nominal values (± 20% for the lowest level point), with at least 75% of calibration points meeting this criterion. All sample concentrations were required to be within the calibration curve range.

Genotyping

Genomic DNA was extracted from maternal peripheral blood by using whole blood genomic DNA extraction kit (BioTeke, Wuxi, China) following the manufacturer’s instructions. The concentration of DNA from each sample was measured with a QubitTM 4.0 Fluorometer (Thermo Fisher Scientific, USA). Then, the samples were stored at -80 °C.

According to the Single Nucleotide Polymorphism Database (dbSNP Database), 63 SNPs from 30 genes were selected as target regions in this project. The selection of SNPs was based on the following criteria: (1) Previous studies have found their relations with some diseases or with metal/metalloid metabolism levels [18, 20–30]; and (2) The minor allele frequency in Han Chinese was > 0.05. More detailed information about the genes, genetic variations, and genotyping studied is shown in Table S2 of the supplementary appendix. Then, a MultipSeq^® Custom Panel covering above 63 SNPs based on amplicon sequencing technology was designed and developed by iGeneTech Bioscience (Beijing, China). According to the protocol of manufacturer, 40 ng of genomic DNA from each sample was used for amplicon library construction. Then, amplicon libraries were sequenced on a NovaSeq 6,000 (Illumina, United States) with a paired-end 150 sequencing strategy by Novogene Co. Ltd (Beijing, China).

Raw reads were processed using Trimmomatic-v0.38 software to filter out low-quality bases and adapters. Clean reads were aligned to the GRCh38 reference genome using BWA-v0.7.12 software. SNP calling was performed with Samtools-v1.9 and VarScan.v2.4.3 software. The criteria of SNP calling was as follows: site depth > 30×, mapping quality > 20, base quality > 20, and variant allele frequency > 15%. For quality-control assessment, genotyping was repeated in 10% of samples, and the consistency rate was 100%. The overall genotyping error rate was below 0.1%, and call rate was above 99.9%.

Data analysis

Baseline characteristics were compared between case and control groups using the Chi-square test. Baseline characteristics included maternal age, maternal ethnicity, maternal education level, previous pregnancy, pre-pregnancy body mass index (BMI) [weight/(height × height)] (kg/m²), maternal medication use, smoking or environmental tobacco smoke (ETS) exposure of one or both parents, maternal alcohol consumption, maternal folic acid supplementations, maternal residence area, and surrounding environmental exposures (including landfills, factories, radio transmitting stations, or an airport within 3 km distance), etc. The exposure time-range was from 3 months before the last menstrual period to the second trimester of pregnancy. Confounders and covariates were selected based on previous literature [19, 31–34] and biological plausibility, and the same set was used in all primary models for consistency. The Shapiro-Wilk test showed that the metal/metalloid concentrations did not conform to the normality assumption and were therefore presented as a median (interquartile range, IQR). Comparison of the serum metal/metalloid concentrations between groups was analyzed by the Mann-Whitney U test.

Restricted cubic splines (RCS) analysis was used to examine the potential nonlinear relationships between continuous metal/metalloid levels and the offspring SDs. The RCS analyses revealed significant nonlinear associations for five of the six metals (all *P*-values for nonlinearity < 0.05), with the exception of Al (Supplementary Fig. S1). The natural log-transformation of concentrations of seven metals/metalloids were carried out, and the concentrations were assessed as tertiles. We opted for this categorization to (a) allow for the examination of potential non-linear effects without pre-specifying the shape of the relationship, and (b) make the results more interpretable for a clinical audience by presenting risk estimates for “low,” “medium,” and “high” concentration groups. The relationship between maternal serum metal/metalloid concentrations and the risk of offspring SDs was evaluated by binary logistic regression. All the above analyses were performed using the Statistical Package for Social Sciences (SPSS) version 27 software (SPSS Inc., IBM, Chicago, USA).

The PLINK software (http://zzz.bwh.harvard.edu/plink/ld.shtml) was used to evaluate the Hardy-Weinberg equilibrium of control group, computing the minimum gene frequency. Logistic regression analysis was performed to evaluate the relationship between metal/metalloid metabolism SNPs and the risk of SDs. For each SNP, three genetic models (additive, recessive, and dominant) were tested to evaluate their relationships with SD risk. Considering a biallelic genetic marker with alleles “a” (reference) and “A” (risk), and assuming a genetic penetrance parameter γ (γ > 1), the additive model assumes that the risk of disease increases γ-fold for genotype “a/A” and 2γ-fold for genotype “A/A”. Under the recessive model, a γ-fold increase in disease risk occurs only in individuals with the “A/A” genotype. In the dominant model, the presence of either one or two copies of the “A” allele (i.e., genotypes “a/A” or “A/A”) confers a γ-fold increase in disease risk [35].

Generalized multifactor dimensionality reduction (GMDR, version 0.7, University of Virginia, Charlottesville, VA) was employed to perform logistic regression analysis, assessing the effects of maternal gene-metal/metalloid concentration interactions on offspring SDs risk. This approach was selected for its enhanced power and accuracy in analyzing multifactor interactions for ordinal traits [36]. The P value was determined using the permutation test with 1000 replications.

All logistic regression models were adjusted for covariates or potential confounders. The false discovery rate (FDR) of the multiple hypothesis test was corrected. A two-sided P < 0.05 was considered statistically significant.

Results

Major characteristics of the study population

The major characteristics of the study population are summarized in Table 1. There were no statistical differences in previous pregnancy, maternal medication use, parental smoking or ETS exposure, maternal alcohol consumption, maternal folic acid supplementation, and surrounding environment exposure between the two groups (P > 0.05). Cases were found to be younger (78.16% cases vs. 75.90% controls being 25–34 years old, P = 0.002), and have higher proportions of Han ethnicity (80.06% vs. 69.25%, P = 0.001). Conversely, controls had a higher proportion of urban residency (91.69% vs. 83.23%, P = 0.003) and a college or higher education (76.18% vs. 67.09% in cases, P < 0.001), and were more likely to have a pre-pregnancy BMI in the normal range (69.81% vs. 64.24%, P < 0.001).

Table 1.

Characteristics of the study population

Characteristics	Controls (n = 361)	Cases (n = 316)	χ2	P
	No. (%)	No. (%)
Maternal age (years)			12.0041	0.002
≤ 24	38(10.53%)	49(15.51%)
25–34	274(75.90%)	247(78.16%)
≥ 35	49(13.57%)	20(6.33%)
Maternal ethnicity			10.314	0.001
Han	250(69.25%)	253(80.06%)
Others	111(30.75%)	63(19.94%)
Maternal education level			20.007	< 0.001
Primary or lower	5(1.39%)	5(1.58%)
Junior high	24(6.65%)	56(17.72%)
High school	57(15.79%)	43(13.61%)
College or higher	275(76.18%)	212(67.09%)
Previous pregnancy			0.773	0.379
None	156(43.21%)	126(39.87%)
Once or more	205(56.79%)	190(60.13%)
Pre-pregnancy BMI (kg/m2)			17.905	< 0.001
≤ 18.5	74(20.50%)	47(14.87%)
18.5–24	252(69.81%)	203(64.24%)
≥ 24	35(9.70%)	66(20.89%)
Maternal medication use			0.319	0.572
Yes	107(29.64%)	100(31.65%)
No	254(70.36%)	216(68.35%)
Parental smoking or ETS exposure			0.736	0.391
Yes	189(52.35%)	155(49.05%)
No	172(47.65%)	161(50.95%)
Maternal alcohol consumption			1.273	0.259
Yes	97(26.87%)	73(23.10%)
No	264(73.13%)	243(76.09%)
Maternal folic acid supplementation			0.087	0.768
Yes	263(72.85%)	227(71.84%)
No	98(27.15%)	89(28.16%)
Maternal residence area			11.389	0.003
Urban	331(91.69%)	263(83.23%)
Rural	27(7.48%)	46(14.56%)
Surrounding environment exposure			0.007	0.933
Yes	13 (3.6%)	11(3.5%)
No	348 (96.4%)	305(96.5%)

The exposure time-range was from 3 months before the last menstrual period to the second trimester of pregnancy; surrounding environment exposure included landfills, factories, radio transmitting stations, or an airport within 3 km distance

The distribution of serum metal/metalloid concentrations

The LOD, detection rates, and median (IQR) of the serum concentrations of metals/metalloids are presented in Table 2. The proportions of samples below the LOD for each metal in both cases and controls are presented in Supplementary Table S3. Except the relatively low detection rates of Mo (72.67%) and Cd (57.31%), the detection rates of the remaining 5 metals/metalloids were nearly or equal to 100%.

Table 2.

The maternal serum concentrations of metals/metalloids

Elements	LOD (ng/mL)	Concentration ≥ LOD, n (%)	Median (IQR)			P a
			Total participants (n = 677)	Controls (n = 361)	Cases (n = 316)
Al	4.33	676(99.85%)	112 (64–172)	109 (61–166)	115 (71–183)	0.116
Fe	12.20	677(100%)	1334 (1025–1683)	1396(1152–1725)	1190(794–1582)	< 0.001*
Zn	15.90	677(100%)	836 (745–947)	858(782–965)	801 (712–919)	< 0.001*
Se	9.88	676(99.85%)	98 (83–117)	98 (81–119)	99 (85–117)	0.382
Sr	0.75	677(100%)	28 (22–34)	29 (23–36)	26 (21–32)	< 0.001*
Mo	0.51	492(72.67%)	0.79(< 0.51–1.21)	0.94(0.72–1.22)	0.54(< 0.51–1.11)	< 0.001*
Cd	0.54	388(57.31%)	0.62(< 0.54–0.95)	0.54(< 0.54–0.95)	0.66(< 0.54–0.95)	0.017*

*P < 0.05

^aP values for the Mann‒Whitney U test between case and control group

There was no difference in the serum concentrations of Al and Se between the controls and the cases (P > 0.05). However, the median concentrations of Fe, Zn, Sr, and Mo in the cases were lower than those in the controls (P < 0.001). In addition, the median concentration of Cd in the cases was higher than that in the controls.

The relationship between maternal serum metal/metalloid concentrations and the risk of offspring SDs

The associations between maternal serum metal/metalloid concentrations and the risk of SDs in offspring are shown in Table 3. Taking the first-tertile as the reference and after covariate adjustment, the logistic regression results showed that the second- and/or third-tertile of maternal serum concentrations of Fe, Zn, Sr, and Mo were associated with decreased risk of offspring SDs. The association was clearly strongest for Mo (second-tertile: aOR = 0.10, 95% CI: 0.06–0.16; third-tertile: aOR = 0.15, 95% CI: 0.09–0.25). The second- and/or third-tertile of maternal serum concentrations of Se and Cd were associated with the increased risk of offspring SDs. The difference between crude and adjusted ORs were very small, showing that the adjustments for potential confounders only marginally influenced the results.

Table 3.

The associations between maternal serum metal/metalloid concentrations and the risk of offspring SDs

Elements	Concentration levels	Controls	Cases	cOR (95% CI)	P	aOR (95% CI)a	P
		No. (%)	No. (%)
Al	First-tertile	139(38.50%)	87(27.53%)	Reference		Reference
	Second-tertile	105(29.09%)	118(37.34%)	1.69(1.06–2.71)	0.028	1.57(0.95–2.60)	0.078
	Third-tertile	117(32.41%)	111(35.13%)	1.66(1.03–2.65)	0.036	1.41(0.86–2.33)	0.173
Fe	First-tertile	80(22.16%)	143(45.25%)	Reference		Reference
	Second-tertile	141(39.06%)	87(27.53%)	0.42(0.26–0.67)	< 0.001	0.38(0.23–0.62)	< 0.001*
	Third-tertile	140(38.78%)	86(27.22%)	0.41(0.25–0.68)	< 0.001	0.36(0.21–0.61)	< 0.001*
Zn	First-tertile	84(23.27%)	142(44.94%)	Reference		Reference
	Second-tertile	142(39.34%)	89(28.16%)	0.42(0.26–0.67)	< 0.001	0.38(0.23–0.62)	<0.001*
	Third-tertile	135(37.40%)	85(26.90%)	0.53(0.32–0.88)	0.013	0.47(0.27–0.79)	0.005*
Se	First-tertile	129(35.73%)	89(28.16%)	Reference		Reference
	Second-tertile	107(29.64%)	125(39.56%)	2.22(1.38–3.57)	0.001	2.21(1.34–3.65)	0.002*
	Third-tertile	125(34.63%)	102(32.28%)	1.78(1.09–2.89)	0.021	2.04(1.21–3.43)	0.008*
Sr	First-tertile	103(28.53%)	121(38.29%)	Reference		Reference
	Second-tertile	119(32.96%)	114(36.08%)	0.80(0.51–1.26)	0.342	0.75(0.46–1.21)	0.238
	Third-tertile	139(38.50%)	81(25.63%)	0.49(0.30–0.80)	0.004	0.41(0.25–0.69)	< 0.001*
Mo	First-tertile	50(13.85%)	175(55.38%)	Reference		Reference
	Second-tertile	170(47.09%)	58(18.35%)	0.09(0.06–0.15)	< 0.001	0.10(0.06–0.16)	< 0.001*
	Third-tertile	141(39.06%)	83(26.27%)	0.14(0.08–0.22)	< 0.001	0.15(0.09–0.25)	< 0.001*
Cd	First-tertile	180(49.86%)	109(34.49%)	Reference		Reference
	Second-tertile	67(18.56%)	95(30.06%)	2.24(1.37–3.66)	< 0.001	2.43(1.45–4.10)	< 0.001*
	Third-tertile	114(31.58%)	112(35.44%)	1.47(0.94–2.29)	0.089	1.46(0.91–2.34)	0.114

*P < 0.05

^aaOR, adjusted odds ratio. Logistic regression was used to calculate odds ratios and 95% CIs; all models were adjusted for maternal age (continuous), maternal ethnicity, maternal education level, previous pregnancy, pre-pregnancy BMI, maternal medication use during pregnancy, smoking or environmental tobacco smoke (ETS) exposure, maternal alcohol consumption in early pregnancy, maternal folic acid supplementation, maternal residence area, and surrounding environment exposure

The relationship between maternal gene polymorphisms and the risk of offspring SDs

Among the 63 SNP loci, the genotypic distributions of four SNPs, including SNP rs11191453, rs3740390, rs3957356, rs606548, did not conform to the Hardy-Weinberg equilibrium (see Supplementary Appendix, Table S4), and these four SNPs were therefore excluded from subsequent analyses.

The relationship between maternal gene polymorphisms and the risk of offspring SDs was analyzed, and the results for SNPs having associations with P < 0.05 and SNPs considered to have potential biological significance are presented in Table 4. The following SNPs were associated with an increased risk of offspring SDs: SNP rs12767543 in AS3MT gene, SNP rs1713420 in PNP gene, SNP rs233804 in SLC39A8 gene, and SNP rs33916661 in SLC7A5 gene. SNP rs7517826 in GCLM gene was associated with a decreased risk of offspring SDs. There was no association between the remaining 55 selected SNPs and the risk of offspring SDs (see Supplementary Appendix, Table S5). The aforementioned associations were not statistically significant after the FDR correction (Table 4).

Table 4.

The relationship between maternal gene polymorphisms and the risk of SDs

Gene	dbSNP_ID	Model	Genotype	Controls	Cases	cOR (95% CI)	P	aOR (95% CI) a	P	FDR-BH P-Value
				No. (%)	No. (%)
AS3MT	rs12767543	Dominant	G/G	113(31.30%)	78(24.68%)	1		1
			G/A-A/A	248(68.70%)	238(75.32%)	1.39(0.99–1.95)	0.05671	1.51(1.06–2.15)	0.024*	0.5089
		Recessive	G/G-G/A	284(78.67%)	246(77.85%)	1		1
			A/A	77(21.33%)	70(22.15%)	1.05(0.73–1.51)	0.7957	1.25(0.85–1.84)	0.2671	0.765
		Log-additive	-	-	-	1.16(0.94–1.44)	0.1702	1.28(1.02–1.61)	0.03504*	0.4869
AS3MT	rs3740393	Dominant	G/G	209(57.89%)	157(49.68%)	1		1
			G/C-C/C	152(42.11%)	159(50.32%)	1.39(1.03–1.89)	0.03265*	1.36(0.99–1.86)	0.05938	0.5089
		Recessive	G/G-G/C	331(91.69%)	297(93.99%)	1		1
			C/C	30(8.31%)	19(6.01%)	0.71(0.39–1.28)	0.2516	0.62(0.33–1.16)	0.135	0.765
		Log-additive	-	-	-	1.16(0.91–1.48)	0.2216	1.12(0.87–1.45)	0.3735	0.6754
PNP	rs1713420	Dominant	A/A	298(82.55%)	243(76.90%)	1		1
			A/G-G/G	63(17.45%)	73(23.10%)	1.42(0.97–2.07)	0.06792	1.5(1.01–2.24)	0.04359*	0.5089
		Recessive	A/A-A/G	358(99.17%)	308(97.47%)	1		1
			G/G	3(0.83%)	8(2.53%)	3.1(0.82–11.79)	0.0969	3.04(0.77-12.00)	0.1122	0.7479
		Log-additive	-	-	-	1.44(1.03–2.02)	0.0351*	1.51(1.06–2.15)	0.02423*	0.4869
PNP	rs1760940	Dominant	A/A	291(80.61%)	243(76.9%)	1		1
			A/C-C/C	70(19.39%)	73(23.1%)	1.25(0.86–1.81)	0.2385	1.3(0.88–1.91)	0.1896	0.6319
		Recessive	A/A-A/C	360(99.72%)	311(98.42%)	1		1
			C/C	1(0.28%)	5(1.58%)	5.79(0.67–49.81)	0.1099	5.93(0.67–52.35)	0.1094	0.7479
		Log-additive	-	-	-	1.3(0.92–1.84)	0.1356	1.35(0.94–1.95)	0.1055	0.4869
PNP	rs3790064	Dominant	A/A	295(81.72%)	257(81.33%)	1		1
			A/G-G/G	66(18.28%)	59(18.67%)	1.03(0.7–1.51)	0.8966	1(0.67–1.51)	0.9868	0.9879
		Recessive	A/A-A/G	358(99.17%)	311(98.42%)	1		1
			G/G	3(0.83%)	5(1.58%)	1.92(0.45–8.09)	0.375	1.52(0.32–7.21)	0.5995	0.8935
		Log-additive	-	-	-	1.07(0.75–1.52)	0.7285	1.03(0.71–1.5)	0.885	0.9482
SLC39A8	rs10014145	Dominant	A/A	293(81.16%)	240(75.95%)	1		1
			A/G-G/G	68(18.84%)	76(24.05%)	1.36(0.94–1.97)	0.09881	1.32(0.89–1.93)	0.1638	0.6144
		Recessive	A/A-A/G	359(99.45%)	314(99.37%)	1		1
			G/G	2(0.55%)	2(0.63%)	1.14(0.16–8.16)	0.8938	0.66(0.08–5.13)	0.689	0.8935
		Log-additive	-	-	-	1.34(0.94–1.9)	0.1088	1.27(0.88–1.84)	0.2065	0.5632
SLC39A8	rs233804	Dominant	C/C	311(86.15%)	253(80.06%)	1		1
			C/A-A/A	50(13.85%)	63(19.94%)	1.55(1.03–2.33)	0.03495*	1.57(1.03–2.40)	0.03711*	0.5089
		Recessive	C/C-C/A	360(99.72%)	313(99.05%)	1		1
			A/A	1(0.28%)	3(0.95%)	3.45(0.36–33.34)	0.2845	2.12(0.21–21.71)	0.5259	0.869
		Log-additive	-	-	-	1.55(1.05–2.28)	0.02681*	1.55(1.03–2.33)	0.03602*	0.4869
SLC7A5	rs33916661	Dominant	A/A	273(75.62%)	220(69.62%)	1		1
			A/G-G/G	88(24.38%)	96(30.38%)	1.35(0.96–1.9)	0.08031	1.48(1.04–2.11)	0.0299*	0.5089
		Recessive	A/A-A/G	356(98.61%)	310(98.10%)	1		1
			G/G	5(1.39%)	6(1.90%)	1.38(0.42–4.56)	0.5994	1.71(0.50–5.85)	0.3932	0.8411
		Log-additive	-	-	-	1.32(0.96–1.79)	0.08349	1.44(1.04–1.99)	0.02758*	0.4869
GCLM	rs41303970	Dominant	G/G	259(71.75%)	236(74.92%)	1		1
			G/A-A/A	102(28.25%)	79(25.08%)	0.85(0.6–1.2)	0.3525	0.76(0.53–1.09)	0.1299	0.6144
		Recessive	G/G-G/A	349(96.68%)	310(98.41%)	1		1
			A/A	12(3.32%)	5(1.59%)	0.47(0.16–1.35)	0.1594	0.41(0.14–1.22)	0.1093	0.7479
		Log-additive	-	-	-	0.83(0.61–1.11)	0.2099	0.74(0.54–1.02)	0.0677	0.4869
GCLM	rs7517826	Dominant	C/C	229(63.43%)	204(64.56%)	1		1
			C/A-A/A	132(36.57%)	112(35.44%)	0.95(0.7–1.31)	0.7616	0.89(0.64–1.24)	0.4984	0.8054
		Recessive	C/C-C/A	340(94.18%)	306(96.84%)	1		1
			A/A	21(5.82%)	10(3.16%)	0.53(0.25–1.14)	0.1046	0.44(0.20–0.99)	0.04691*	0.7479
		Log-additive	-	-	-	0.89(0.69–1.16)	0.3959	0.83(0.63–1.10)	0.1921	0.5632

*P < 0.05

^aaOR, adjusted odds ratio. Logistic regression was used to calculate odds ratios and 95% CIs; all models were adjusted for maternal age (continuous), maternal ethnicity, maternal education level, previous pregnancy, pre-pregnancy BMI, maternal medication use during pregnancy, smoking or environmental tobacco smoke (ETS) exposure, maternal alcohol consumption in early pregnancy, maternal folic acid supplementation, maternal residence area, and surrounding environment exposure

Effects of interactions between maternal genotypes and serum metal/metalloid concentrations on the risk of offspring SDs

The metals/metalloids (Se and Cd) and the SNPs (rs12767543, rs1713420, rs233804, and rs33916661) that were associated with the increased risk of offspring SDs, were included in the GMDR analysis. The results of GMDR analysis for the effect of the interaction between the SNPs and the metals/metalloids on the risk of offspring SDs are shown in Table 5. For gene-metal/metalloid interactions, five interaction combination models were analyzed, but no statistical significance was observed.

Table 5.

GMDR analysis for interactions between maternal SNPs and specific metal/metalloid concentrations in affecting the risk of offspring SDs

Model	Training Bal. Acc	Testing Bal. Acc	Sign Test (P-Value)	CV Consistency
rs1713420 Cd	0.5683	0.5568	7(0.1719)	10/10
rs12767543 Se Cd	0.5815	0.5421	6(0.3770)	7/10
rs12767543 rs1713420 Se Cd	0.6032	0.5324	6(0.3770)	10/10
rs12767543 rs1713420 rs33916661 Se Cd	0.6203	0.4881	4(0.8281)	7/10
rs12767543 rs1713420 rs233804 rs33916661 Se Cd	0.6446	0.4798	3(0.9453)	10/10

Training Bal. Acc Training balanced accuracy, Testing Bal. Acc Testing balanced accuracy, CV Consistency Cross validation consistency

Discussion

Genetic and environmental factors, as well as their combinations, play a crucial role in the occurrence of birth defects. However, there are very few reports on the role of maternal microenvironmental factors in the pathogenesis of offspring SDs. In this case-control study, we observed that the serum concentrations of Fe, Zn, Sr, and Mo in the cases were lower than those in the controls, and the serum concentration of Cd in the cases was higher than that in the controls. The second- and third-tertile concentrations of Se and the second-tertile concentration of Cd in pregnant women were associated with an increased risk of offspring SDs. The second- and third-tertile concentrations of Fe, Zn, and Mo, and the third-tertile concentration of Sr in pregnant women were associated with a decreased risk of offspring SDs. The association was particularly strong for Mo. Maternal SNP rs12767543 of AS3MT, SNP rs1713420 of PNP, SNP rs233804 of SLC39A8, and SNP rs33916661 of SLC7A5 were nominally associated with an increased risk of offspring SDs before FDR correction. Maternal SNP rs7517826 of GCLM was nominally associated with a decreased risk of offspring SDs before FDR correction. However, no significant association between any of the selected loci and the risk of offspring SDs was found after FDR correction. In addition, no statistically significant maternal gene-metal/metalloid interaction was observed in respect of its effect on the risk of offspring SDs.

Previous studies showed that Mo nanoparticles could reduce mitochondrial membrane potential through reactive oxygen species, leading to cell cycle arrest, inducing DNA damage [37], and resulting in weight loss in pregnant mice, a decrease in the number of mice fetus, and an increase in fetal skeletal abnormalities [38]. However, the present study showed that Mo concentration in the case group was lower than that in the control group, and the lower maternal concentrations of Mo were associated with a remarkably large increased risk of offspring SDs. Specifically, when compared to mothers in the lowest Mo tertile (the reference group), those in the second and third tertiles had significantly reduced odds of having offspring with SDs (aOR = 0.10 and 0.15, respectively), indicating a strong inverse association. This magnitude of association was notably stronger than for the other elements studied and was unexpected, given the limited prior human evidence linking maternal Mo status to fetal SDs.

Fe can coordinate with ascorbate to promote differentiation of skeletal mesenchymal stromal cells, regulate histone methylation in human pluripotent stem cells-mesenchymal stromal cells (hPSCs-MSCs), and increase the self-renewal and osteochondrogenic capacity of hPSC-MSCs. During embryonic development, all cartilage, bone, and bone marrow of the limbs and tailbone originate from the lateral lamellar mesoderm. MSCs are derived from embryonic mesodermal progenitor cells [39, 40]. Maternal Fe deficiency constitutes a risk factor for adverse pregnancy outcomes, encompassing greater maternal morbidity, low birthweight, preterm birth, and intrauterine growth restriction [41]. Systemic unbalance in Fe concentrations negatively affects bone health [42]. Nevertheless, to the best of our knowledge, there has been no previous study providing clinical data about any association between maternal Fe levels and offspring SDs. In this report, we provide the first evidence in humans that decreased maternal serum concentrations of Fe may be associated with the increased risk of offspring SDs.

Zn deficiency in the body may lead to DNA damage, oxidative stress, apoptosis, and organ destruction in cell components via reactive oxygen species [43]. Dietary intakes of Zn and Se, the ratios of Zn -to- Cu and Se -to- Cu, and the use of Zn and Se supplements (both throughout pregnancy and in the first trimester) have all been found to have significant inverse associations with congenital heart diseases (CHDs) [44]. Zn plays a crucial role in bone metabolism by synergizing bone induction with immune regulation. Beyond directly promoting osteogenic differentiation of progenitor cells, Zn²⁺ has been shown to induce a transition of macrophages to the pro-healing M2 phenotype via the PI3K/Akt/mTOR pathway [45]. This creation of a favorable immunomodulatory microenvironment further amplifies Zn²⁺’s osteogenic effects, establishing a reciprocal mechanism that significantly enhances bone regeneration. This is consistent with our results that the lowered maternal serum concentrations of Zn were associated with the increased risk of offspring SDs.

We observed a previously unreported association between lowered maternal Sr levels and the increased risk of offspring SDs. As Sr is generally considered nonessential, this finding warrants careful interpretation. Evidence supporting a potential biological role includes Sr’s chemical similarity to calcium, its ability to incorporate into bone, and experimental data showing Sr’s role in modulating bone metabolism by inhibiting resorption and promoting bone formation and angiogenesis [46, 47]. However, there is no established evidence that Sr is essential for mammalian fetal development. High-dose Sr exposure has been shown to cause skeletal abnormalities in animal models [48]. But we found no indications of such an effect among humans in the presently studied exposure range. Possibly, the observed association of the low maternal serum Sr levels with the increased risk of offspring SDs might not reflect a direct causal effect but could instead serve as a proxy for other nutritional deficiencies, broader maternal undernutrition, or dysregulated mineral metabolism. Whether Sr plays an active role in fetal skeletal development or merely acts as a biomarker of nutritional status thus remains unclear. Further investigations on its placental transport, interactions with essential minerals, and effects of its relevant low levels on human pregnancy are warranted. Overall, in the present study, the lowered maternal serum concentrations of Fe, Zn, Sr, and Mo were associated with the increased risk of offspring SDs. While previous studies using animal models showed that excessive exposure to these metals/metalloids could lead to developmental abnormalities. It might be that the amount of these metals/metalloids in the blood samples from our included subjects did not reach a harmful threshold. Of course, these observations warrant more research on this matter.

Se is an essential trace element for living organisms and acts mainly through the synthesis of selenoprotein [49]. While adequate Se status is generally associated with higher bone mineral density and lower osteoporosis risk in adults [50, 51], our findings suggest a more complex relationship between maternal Se levels and the risk of offspring SDs. In this study, we found no difference in median maternal serum concentration of Se between the case group and the control group. However, the second- and/or third-tertile serum concentrations of Se were associated with an increased risk of offspring SDs. The biological basis for this association remains unclear. It may reflect unknown threshold effects, interactions with other elements (e.g., Hg, given their documented antagonism [52]), or confounding from unmeasured factors. Further large-scale and prospective cohort studies are needed to clarify this matter and to determine whether this association reflects causality, interaction, or residual confounding.

A recent Chinese cohort study showed that higher placental Cd levels correlated with both preeclampsia risk and fetal growth restriction [53]. Another study has further elucidated the impact of Cd at the maternal-fetal interface. Cd exposure may impair fetal development by inducing placental metallothionein synthesis, which reduces Zn bioavailability and promotes fetal Zn deficiency. Notably, Cd(II) levels in newborns may also contribute to neurodevelopmental alterations, including cognitive impairments [54]. It was found that the offspring of female zebrafish exposed to Cd had smaller ceratohyal cartilage angle [55, 56]. The possible causes for the effects of Cd include vitamin D deficiency, elevated levels of inflammation, and inhibition of osteoblast differentiation, maturation and mineralization [57–59]. In the present work, we extended previous studies by showing that maternal median serum Cd concentration was higher in the case group than that in the control group, and the maternal Cd concentrations showed a positive association with the increased risk of offspring SDs. The association between the elevated maternal Cd levels and the offspring SDs risk might reflect Cd’s complex biological interactions with functions of divalent metals, such as Fe and Zn. It has been found that Fe deficiency increases intestinal absorption of Cd through upregulation of the divalent metal transporter DMT1 [60]. Similarly, Cd competes with Zn for cellular uptake via shared transporters and can disrupt Zn homeostasis by inducing the Zn exporter ZnT1 [61]. Consequently, the observed positive association between maternal serum Cd levels and offspring SDs in our study might be partially mediated by or co-occur with suboptimal maternal Fe and Zn statuses. Thus, the association could partly be driven by underlying deficiencies of essential metals. Residual confounding from smoking, a major source of Cd exposure, may also impact the association. While a causal effect could not be ruled out, those alternative explanations warrant caution in interpretation.

A study in an Arab population with a case group of 33 congenital anomalies found that household and industrial exposure to Al and Cd during pregnancy was not associated with congenital limb anomalies [62]. However, only four congenital limb anomalies were included in the case group, which was not enough to draw reliable conclusions on the relationship. Another study in North Carolina examined the relationship between Cd levels in private wells and the rates of birth defects including reduction defects of the upper and lower limbs [63], and found that the heavy metal levels in the environment were not associated with congenital limb defects in the region. This is basically inconsistent with our findings. The reason for the inconsistency may be that the previous study measured the heavy metal levels in private wells, while our study measured serum metal/metalloid concentrations in the pregnant women, directly reflecting individual maternal metals/metalloids concentrations. A recent study in Japan used logistic regression analysis and quartile g-computation methods to analyze the relationship between exposure to metals/metalloids and congenital limb abnormalities in middle and late pregnancy, and found that maternal exposure levels of Cd, Se, Pb, Hg, and Mn were not associated with the risk of congenital limb abnormalities in infants [34]. In the present study, we found that maternal serum concentrations of Fe, Zn, Sr, and Mo were negatively associated with the risk of offspring SDs, and maternal serum concentration of Cd was positively associated with that risk. Generally, it is assumed that deficiency of essential metals/metalloids during pregnancy only has a small effect on the fetus, due to the principle of fetal nutrient priority. It is worth noting that our findings do not contradict the principle of fetal nutrient priority, but suggest that the suboptimal maternal metal/metalloid status in some Chinese pregnant women may affect the fetal environment in ways relevant to skeletal development. Overall, there are very few studies on the relationship between maternal metal/metalloid levels and the risk of offspring SDs in terms of population-based epidemiological data, and more research is needed.

Furthermore, after the necessary FDR correction, no significant associations were observed between the selected maternal metabolic gene polymorphisms and the risk of offspring SDs. In contrast to the prevailing paradigm that developmental anomalies often arise from gene-environment interactions, our analysis found no statistically significant evidence that maternal metal exposure and the selected SNPs interacted to influence the risk of offspring SDs. This null finding was consistent across all five interaction models tested. This result appears to contradict the evidence supporting the interaction model in other contexts. For instance, previous studies have demonstrated significant interactions between metals and genetic polymorphisms in relation to inflammatory markers like C-reactive protein [64] and complement C3 [65]. Interactions between estrogen receptor-α polymorphisms and lean tissue mass affected bone mineral density in menopausal women [66], and polymorphisms in genes like N6AMT1 could modify the metabolic response to arsenic exposure during pregnancy [67]. In general, there have been limited studies on the effects of maternal gene-environment interactions on offspring SDs. The null finding in the present study could stem from limited statistical power for possible subtle interactions, the etiological heterogeneity of the composite skeletal abnormality phenotype, or the involvement of genetic factors beyond the selected SNPs. Consequently, a true interaction might be absent, undetectable by our study design, or manifest only in specific, unmeasured genetic or phenotypic contexts.

Our research has many advantages. Firstly, to the best of our knowledge, this is the first study to evaluate the association between specific maternal serum metal/metalloid concentrations and the risk for a wide range of SDs in offspring. Moreover, this study was made possible by designing a SD study within the large prospective pregnancy cohort with a systematically established biobank, which provided pre-diagnosis serum samples for biomarker assessment. Secondly, the 7 metals/metalloids with levels above the LOD were detected in blood samples of pregnant women in the first and second trimester, which allowed us to conduct a reliable assessment of internal exposure levels. Thirdly, we did not distinguish subjects with occupational exposure from those with non-occupational exposure, and given that the exposure levels were assessed at the individual level, the results can be fairly well generalized to all women in the studied serum metal/metalloid concentration ranges. Finally, we explored the effect of interaction between maternal metabolic gene polymorphisms and metal/metalloid concentrations on the risk of offspring SDs, and reviewed the findings on the effect of the interaction between environment factors and genetics on health. However, the present study also has some limitations. Firstly, the sample size was small, so subgroup grouping could not be carried out. Secondly, we only measured maternal serum metal/metalloid concentrations at one point during pregnancy. Thirdly, due to the limit of the detection method, of the 22 metals/metalloids analyzed, only 7 metals/metalloids with levels above the limit of detection were included in the association analysis. Fourthly, the exclusion of cases with a family history of SDs, while improving phenotypic homogeneity, might have reduced the representation of strong genetic factors among cases and thus limited the power to detect gene-environment interactions. Fifthly, maternal serum Al concentrations we report are higher than internationally published reference values for healthy populations [68]. We acknowledge that this finding may be subject to technical limitations: blood samples were collected using BD Vacutainer^® red-top tubes rather than tubes specifically recommended for trace element analysis (e.g., dark-blue-top tubes). Studies have demonstrated that rubber stoppers or tube walls of standard blood collection tubes may release Al, leading to sample contamination and falsely elevated results. Therefore, Al concentrations we report may not reflect true levels in our participants. We have chosen to report these data to provide a cautionary note for future research, underscoring necessity of strictly employing contamination-free pre-analytical procedures when conducting trace element analyses. Concurrently, we cannot entirely exclude the possibility that these elevated concentrations have genuine biological relevance for this pregnant population, a question that warrants further investigation using rigorously quality-controlled methods. In addition, a further limitation is the potential for unmeasured confounders due to the lack of data on certain covariates. Most notably, the absence of information on fish consumption and Hg levels prevented us from fully accounting for their mutual influence and its potential impact on the model estimates. Besides, we found some SNPs with borderline significant associations with the risk of offspring SDs before FDR correction, and no such significant associations were found after FDR correction. These are preliminary results that need to be validated in future independent samples. In the future, a larger population study and subgroup analysis will be conducted. Samples from pre-pregnancy and multiple time-points during pregnancy should be analyzed to detect the dynamic changes in metal/metalloid concentrations and metabolism, and to distinguish between pre-pregnancy exposure and exposure during different stages of pregnancy.

Conclusions

This study provides novel epidemiological evidence linking maternal trace element status with fetal skeletal development. We demonstrate, for the first time in a human population, that lower maternal serum concentrations of the essential elements Fe, Zn, and Mo, as well as of the non-essential element Sr, are associated with an increased risk of SDs in offspring, whereas Cd shows a positive association. Notably, the protective association between the maternal serum concentration of Mo and the offspring SDs risk is exceptionally strong, which was unexpected and highlights Mo as a priority for future investigation. In contrast, we found no significant contribution from the selected maternal metabolic gene polymorphisms or their interactions with serum metal concentrations within this cohort. These findings shift the focus from toxicity of high exposures to the potential importance of maternal nutritional adequacy, even for non-essential elements like Sr. As the first study to report these specific associations, our results require confirmation from independent and larger prospective cohorts. Further research is warranted to elucidate the underlying biological mechanisms.

Abbreviations

SDs

Skeletal defects

Al

Aluminum

Fe

Iron

Zn

Zinc

Se

Selenium

Sr

Strontium

Mo

Molybdenum

Cd

Cadmium

SNP

Single nucleotide polymorphism

aOR

Adjusted odds ratio

CI

Confidence interval

Cu

Copper

Pb

Lead

Hg

Mercury

MT2A

Metallothionein 2A

VDR

Vitamin D receptor

TF

Transferrin

ICP-MS

Inductively coupled plasma mass spectrometry

As

Arsenic

Mn

Manganese

Co

Cobalt

Ni

Nickel

La

Lanthanum

W

Tungsten

Sb

Antimony

Sn

Tin

Cr

Chromium

Tl

Thallium

Ce

Cerium

Ba

Barium

He

Helium

LOD

Limit of detection

LOQ

Limit of quantification

BMI

Body mass index

ETS

Environmental tobacco smoke

IQR

Interquartile range

RCS

Restricted cubic splines

GMDR

Generalized multifactor dimensionality reduction

FDR

False discovery rate

hPSCs-MSCs

Human pluripotent stem cells-mesenchymal stromal cells

CHD

Congenital heart disease

Authors’ contributions

P. Y., C. D., and N. L. developed the study design and drafted the manuscript. L. L. and M. W. assisted in conducting the experiments and analyzing the data. H. K., Y. D., Y. L. and Z. L. assisted in preparing samples and extracting DNA. P. Y., N. L., Y. W., and J. Z. participated in reviewing, editing, and revising the manuscript. All authors have read and approved the final manuscript.

Funding

This work was supported by the National Key Research and Development Program of China (grant numbers 2022YFC2703700, 2022YFC2703702, 2022YFC2703300, 2022YFC2703302).

Data availability

The SNP data obtained through next-generation sequencing analysis for this study have been deposited in the European Variation Archive (EVA) at EMBL-EBI under accession number PRJEB78551.

Declarations

Ethics approval and consent to participate

All the participants signed an informed consent form. This research was approved by the Ethics Committee of Sichuan University, China (Protocol ID: K2017045), and followed the tenets of the Declaration of Helsinki.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.