Associations of Maternal rs1801131 Genotype in MTHFR and Serum Folate and Vitamin B12 with Gestational Diabetes Mellitus in Chinese Pregnant Women

Circumstantial evidence links one-carbon metabolism (OCM) related nutrients, such as folate and vitamin B12, with gestational diabetes mellitus (GDM). However, few studies have evaluated the combined effects of these nutrients with OCM related gene polymorphisms on GDM. This study investigated whether OCM related genetic variants modified the associations of folate and B12 with GDM. Logistic regression was used to estimate odds ratios (ORs) for OCM related nutrients and single nucleotide polymorphisms (SNPs) in genes encoding main OCM related enzymes (MTHFR, MTR, and MTRR) on GDM. Higher folate concentrations were associated with increased GDM risk (OR: 1.59; 95% CI: 1.22, 2.13). However, higher B12 concentrations were associated with reduced GDM risk (OR: 0.76; 95% CI: 0.65, 0.92). Pregnancies with MTHFR rs1801131 G alleles had a significantly lower risk of GDM than pregnancies with T alleles (OR: 0.65; 95% CI: 0.47, 0.91) under the dominant model. The genotype-stratified analysis revealed the association between folate and GDM (OR: 1.66, 95% CI: 1.20, 2.30) or B12 and GDM (OR: 0.80, 95% CI: 0.65, 0.98) was more evident in pregnancies with TT genotype. Higher folate and lower B12 are associated with GDM. Pregnancies with MTHFR rs1801131 TT genotype are more susceptible to OCM nutrient-related GDM.


Introduction
Gestational diabetes mellitus (GDM), defined as glucose intolerance with onset or first recognition during pregnancy, is currently the most common medical complication. The prevalence of GDM varies considerably among countries, ranging from 1.8% to 31% [1]. GDM; (2) SNPs in OCM pathway is associated with GDM and (3) genetic variants in the OCM pathway modify the association between OCM nutrients and GDM.

Study Population
Pregnant women in this research were enrolled in the Gene-Environment-Nutrient-Epigenetics interaction on Maternal and Children health study (GENEMaC) between 2017 and 2018 in Tianjin, China. This cohort was established primarily to investigate gene-environment interactions of maternal arsenic exposure, arsenic metabolism-related nutrients, and gene polymorphisms on offspring's health via epigenetic changes [26]. The research proposal was approved by the Ethics Committee of Tianjin Xiqing Hospital. All participants provided written informed consent before participating in this study.
A total of 1505 pregnancies who attended GDM screening at the Maternal and Child Health Care Hospital of Beichen District during 24-28 gestational weeks were enrolled in this cohort. The inclusion criteria were: (1) age ≥ 18 years, (2) residents of Tianjin with ≥one year of residence, and (3) intent to inhabit Tianjin in the next six years. The exclusion criteria were: (1) prepregnancy diabetes and previous GDM, (2) unable or unwilling to give informed consent or communicate with study staff. Of the 1505 participants, 1464 pregnant women completed the 75-g oral glucose tolerance test (OGTT). We excluded 49 pregnant women who did not have enough blood samples for OCM nutrients and SNPs determination, and 27 pregnant women with covariates missing, resulting in 1388 pregnancies included in the analysis of the association between OCM and GDM. In addition, we analyzed data from 1364 participants with MTHFR rs1801131 genotype data available for gene-nutrient interaction study ( Figure 1). It should be noted that the GDM screening strategy in this area (Beichen District, Tianjin) is divided into two steps. Firstly, fasting plasma glucose (FPG) was used to rule out GDM (FPG < 4.4 mmol/L) and rule in GDM (FPG ≥ 5.1 mmol/L) in the community health centers (primary care providers) during 24-28 gestational weeks. Secondly, for pregnancies with FPG between ≥4.4 and <5.1 mmol/L, GDM diagnosis was performed by the OGTT examination at the Maternal and Child Health Care Hospital of Beichen District (secondary care provider). Therefore, the prevalence of GDM was higher in the present study.

Sample Collection and Covariates Assessment
The fasting blood sample was collected from each pregnancy during the GDM screening. Aliquots of blood sample and serum were obtained and then transferred to Tianjin Medical University for storage in freezers at −80 • C until analysis.
Baseline characteristics concerning an individual's age, ethnicity, education, smoking and drinking habits, height, current and prepregnancy weight, parity, and family history of diabetes were obtained by a structured questionnaire via well-trained interviewers. Ethnicity was defined as Han nationality or Minority nationality. Education level was categorized according to the duration of education. Smoking and drinking were defined as never or ever before and during the pregnancy. Prepregnancy BMI (kg/m 2 ) was estimated as prepregnancy weight (kg) divided by the square of height (m).

Diagnosis of GDM
According to the diagnostic criteria recommended by the Ministry of Health of China, all the pregnancies underwent a GDM screening using a 75-g OGTT during 24-28 gestational weeks [27]. The Chinese diagnostic criteria agree with the International Association of Diabetes and Pregnancy Study Groups (IADPSG). Accordingly, a diagnosis of GDM can be made if one or more of the following glucose values are evaluated: fasting plasma glucose (FPG) ≥ 5.1 mmol/L, 1-h plasma glucose (1-h PG) ≥ 10.0 mmol/L, 2-h plasma glucose (2-h PG) ≥ 8.5 mmol/L.

Sample Collection and Covariates Assessment
The fasting blood sample was collected from each pregnancy during the GDM screening. Aliquots of blood sample and serum were obtained and then transferred to Tianjin Medical University for storage in freezers at −80 °C until analysis.
Baseline characteristics concerning an individual's age, ethnicity, education, smoking and drinking habits, height, current and prepregnancy weight, parity, and family history of diabetes were obtained by a structured questionnaire via well-trained interviewers. Ethnicity was defined as Han nationality or Minority nationality. Education level was categorized according to the duration of education. Smoking and drinking were defined as never or ever before and during the pregnancy. Prepregnancy BMI (kg/m 2 ) was estimated as prepregnancy weight (kg) divided by the square of height (m).

Determination of OCM Related Nutrients
Determination of OCM nutrients was performed on maternal serum as previously described [26]. In brief, folate, and B 12 concentrations were measured using an automated chemiluminescence immunoassay system (Architect-i2000SR Analyzer; Abbott Diagnostics, Chicago, USA). Hcy levels were determined using an automatic biochemical analyzer with an enzymatic cycling method (Dirui CS-T300; Dirui, Changchun, China).

Statistical Analysis
The folate/B 12 was calculated as folate divided by B 12 according to our previous study [17]. Since the skewed distribution of folate, B 12 , Hcy, and folate/B 12 , these OCM indicators were reported as median (interquartile range, IQR). The baseline characteristics of the study participants were summarized using descriptive statistics (n [%] for categorical variables, median [IQR] for continuous variables). Wilcoxon Mann-Whitney U test was applied to determine the differences for continuous variables with skewed distribution. Chi-square test or Fisher's exact test were used to examine the differences for categorical variables. Spearman correlation was performed to evaluate monotonic relationships between OCM indicators and glucose levels of the 75-g OGTT. OCM indicators, including folate, B 12 , Hcy, and folate/B 12 , were evaluated as continuous variables and categorical variables. Logistic regression was applied to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for OCM indicators and SNPs on GDM. The restricted cubic spline (RCS) regression model with four knots was used to evaluate the potential nonlinear relationship among serum folate, B 12 , Hcy, folate/B 12 , and GDM risk. Potential maternal confounders, including age, ethnicity, education, drinking, smoking, parity, family history of diabetes, and prepregnancy BMI, were adjusted in all models. In addition, serum OCM indicators (folate, B 12 , and Hcy) were mutually adjusted in estimating the association of OCM indicators with GDM. When evaluating the relationship between different SNPs and GDM, both maternal characteristics and serum folate, B 12 , and Hcy were adjusted.
Since MTHFR rs1801311 was significantly associated with GDM under the dominant and additive models (see the Results section), associations between OCM indicators and GDM were reanalyzed under the stratification of rs1801311 genotypes. This allows us to target the subpopulation for intervention. Due to the small sample size of rs1801311 GG genotype, pregnant women were divided into GG/TG and TT groups according to the dominant model for further stratified analysis. Multiplicative and additive interactions were also performed to identify whether the effect of OCM indicators on GDM would be different in different genotype subgroups. A detailed description of these models could be found in our previous study [32]. Briefly, multiplicative interaction was assessed via the p-value (p interaction ) of a cross-product interaction term of the OCM indicators and the rs1801311 genotype in a multiple logistic regression model. Additive interaction was evaluated through the relative excess risk due to interaction (RERI) using a multiple logistic regression model, and its 95% CI was computed with bootstrapping [33]. To estimate the overall association of the OCM nutrients with GDM stratified by rs1801311 genotype, the Bayesian kernel machine regression (BKMR) with 10,000 iterations was also employed. A more detailed description of this model can be found in our previous study [26]. Briefly, BKMR combines Bayesian and statistical learning methods to flexibly model the individual and joint effects of OCM mixtures on GDM using a Gaussian kernel function [34]. Results from these models could be used to (1) provide the exposure-response relationship for each OCM indicator on GDM when other indicators are fixed at their median; (2) evaluate the association of an IQR increase in a single OCM indicator on GDM when all the other indicators are fixed at either the 25th, 50th, or 75th percentile.
All statistical analyses were performed using R (version 4.0.2; R Project for Statistical Computing). BKMR was implemented with the R packages "bkmr" (version 0.2.0). A p-value < 0.05 was considered to be statistically significant.

Baseline Characteristics
The demographic characteristics and OCM related nutrient concentrations of the study participants are shown in Table 1. Of the 1388 pregnancies, 274 (19.7%) were diagnosed with GDM. Pregnancies with GDM were more likely to be older and multiparous and have higher prepregnancy BMI and a family history of diabetes than non-GDM pregnancies. The median (IQR) levels of serum folate, B 12 , Hcy, and folate/B 12 were 9.4 (6.2-14.6) ng/mL, 271 (214-337) pg/mL, 5.0 (4.5-6.0) µmol/L, and 35.1 (23.9-49.2), respectively. Compared with non-GDM women, subjects with GDM had significantly higher folate levels, lower B 12 levels, and corresponding higher folate/B 12 . No significant differences were observed in Hcy levels between the two groups.

Associations between OCM Related Gene Polymorphisms and GDM
The genotypes of the 12 SNPs were in Hardy-Weinberg equilibrium (p > 0.05). Among these SNPs in the OCM pathway, the genotypic distribution of the MTHFR rs1801131 SNP (TT, TG, and GG) was significantly different between the two groups (Table S1). Figure S3 shows the associations between OCM related SNPs and GDM under three genetic models. After adjustment for maternal characteristics and serum OCM indicators, MTHFR rs1801131 was associated with GDM in the dominant and additive models but not in the recessive model. Compared with pregnancies with TT genotype, pregnancies with TG (OR: 0.68; 95% CI: 0.49, 0.96) and GG (OR: 0.30; 95% CI: 0.07, 1.33) genotype had lower odds of GDM after adjustment for multiple covariates in the logistic regression analysis (Table 3). In the analysis of both mutant genotypes (GG/TG) under the dominant model, pregnancies with G alleles had a significantly lower risk of GDM than pregnancies with T alleles (OR: 0.65; 95% CI: 0.47, 0.91). However, pregnancies with GG genotype did not have a significantly lower risk of GDM than did pregnancies with TG/TT genotypes under the recessive model. In addition, an increased copy of the G allele was found to be associated with a lower risk of GDM (OR: 0.66; 95% CI: 0.48, 0.89) under the additive model (Table 3).

Combined Effects of OCM Indicators and rs18011311 Genotypes on GDM
The genotype-stratified analysis revealed that an IQR increase in maternal serum folate was associated with higher odds of GDM (OR: 1.66, 95% CI: 1.20, 2.30) among pregnancies with MTHFR rs1801131 TT genotype (Table 4). Similarly, the associations between serum B 12 and GDM were more evident in pregnant women with MTHFR rs1801131 TT genotype (OR: 0.80, 95% CI: 0.65, 0.98) after adjustment of maternal characteristics and serum folate and Hcy concentrations. To further estimate whether the association between OCM indicators and GDM was modified by MTHFR rs1801131 genotypes, the interactions on the multiplicative and additive scales were evaluated. However, no significant interactions on the multiplicative and additive scales between OCM indicators and GDM were observed ( Table 4).
The exposure-response functions of the three OCM indicators on GDM stratified by MTHFR rs1801131 genotypes are shown in Figure 2. Among pregnancies with rs1801131 GG/TG genotype, folate showed increasing association with GDM when B 12 and Hcy were fixed at their median levels. However, B 12 and Hcy displayed decreasing associations with GDM when the other two OCM indicators were fixed at their median levels, respectively ( Figure 2A). Figure 2B showed the association of each OCM indicator with GDM when the single OCM indicator increased an IQR, where all of the other indicators are fixed at 25th, 50th, or 75th percentiles. However, there were no significant associations of folate, B 12 , and Hcy with GDM among pregnant women with rs1801131 GG/TG genotype. Among pregnancies with rs1801131 TT genotype, folate and Hcy showed increasing association with GDM. However, BKMR identified an individual U-shaped association between B 12 concentrations and GDM ( Figure 2C) while holding folate and Hcy at their median concentrations. As shown in Figure 2D, an IQR increase in folate was associated with a 0.07-unit (95% CI: 0.03, 0.12) increase in GDM risk when B 12 and Hcy were fixed at their 50th percentile values (similar results were obtained when B 12 and Hcy were fixed at their 25th and 75th percentile values). In contrast, an IQR increase in B 12 was associated with a 0.07-unit (95% CI: −0.10, −0.03) decrease in GDM risk when folate and Hcy were fixed at their median values (similar results were obtained when folate and Hcy were fixed at their 25th and 75th percentile values). There was no significant association between Hcy and GDM when folate and B 12 were fixed at different percentile values.

Discussion
In this gene-nutrient interaction study, we evaluated the combined effects of OCM related nutrients and gene polymorphisms on GDM in a Chinese pregnancy cohort for the first time. We found that serum folate concentrations were positive, whereas serum B12 concentrations were negatively associated with the risk of GDM. Notably, we found

Discussion
In this gene-nutrient interaction study, we evaluated the combined effects of OCM related nutrients and gene polymorphisms on GDM in a Chinese pregnancy cohort for the first time. We found that serum folate concentrations were positive, whereas serum B 12 concentrations were negatively associated with the risk of GDM. Notably, we found that the MTHFR rs1801131 TT genotype was significantly associated with an increased risk of GDM. Moreover, we found homozygous in pregnant women for the MTHFR rs1801131 TT genotype, higher folate, and lower B 12 were more obviously associated with increased GDM risk.
Folic acid is widely used to prevent birth defects, with a recommended daily intake of 400 micrograms from prepregnancy until 12 weeks of pregnancy in many countries [35]. Emerging evidence suggests that periconceptional higher folate intake is associated with higher GDM risk [8,9]. However, inconsistent findings were found for prepregnancy habitual intakes of folate in the Nurse's Health Study [11]. Although folic acid intake evaluated via the questionnaire may not accurately reflect folate levels in pregnant women, serum and red blood cell folate levels have also been associated with GDM [10,16]. Our preliminary study indicated that higher folate levels in mid-pregnancy can slightly increase maternal GDM risk [17]. In the present study, we expanded our findings with a large sample size. We found that serum folate levels are positively correlated with blood glucose levels and significantly associated with GDM risk in a dose-response manner. Our findings were consistent with recently published results [16,36], which indicated that higher maternal folate during pregnancy is associated with increased GDM risk.
In contrast to folate, we found a significantly negative correlation between serum B 12 levels and FPG. However, weak but not significantly positive correlations between B 12 , 1-h PG, and 2-h PG were observed in the present study ( Figure S1). In addition, we found that the risk of GDM decreased with the increase of B 12 levels. This, in turn, suggested that lower B 12 was related to a higher risk of GDM (Table 2). Our findings were in line with the results of some previous studies, in which lower B 12 was associated with a higher risk of GDM [13][14][15][16]. However, Chen et al. reported a positive relationship between B 12 levels and GDM risk in a prospective study from Shanghai, China [10]. The reasons for the conflicting results are unclear. It was reported that the level of serum B 12 decreased gradually with the progress of the pregnancy [37]. In our study, the median concentration of mid-pregnancy serum B 12 was 271 pg/mL, which is lower than Chen's report in early pregnancy (405.93 pg/mL). In addition, we found that the relationship between B 12 and GDM was nonlinear ( Figure S2B). When the serum B 12 level reached about 400 pg/mL, the risk of GDM did not decrease with the increase of B 12 levels. This may partly explain the differences between our study and Chen's study. Since the proportion of pregnant women with serum B 12 levels greater than 400 pg/mL (14.3%) was small, we did not observe a significantly positive association between B 12 and GDM at higher serum B 12 both in the logistic regression model and RCS model. Therefore, future studies should be performed in pregnancies with a wide range of serum B 12 to investigate the dose-response association between B 12 and GDM.
Hcy is a surrogate marker for folate and B 12 insufficiency. In this study, we confirmed that Hcy was negatively correlated with folate and B 12 . Although Hcy was negatively correlated with postprandial blood glucose, we did not find a significant association between Hcy and GDM in different statistical models (Table 2 and Figure S2C). Nevertheless, we discovered that folate/B 12 is a sensitive index to evaluate the relationship between OCM nutrients and GDM. A higher folate/B 12 value represents a relatively high folate level or a relatively low B 12 level. In this study, we found that pregnancies with GDM have significantly higher folate/B 12 values. Simultaneously, this index was positively associated with blood glucose levels and GDM risk. Although Chen's study has shown that folate/B 12 is not related to GDM [10], our findings suggested that the balance of folate and B 12 may be necessary for the health of pregnant women. In addition, we also found that the relationship between folate/B 12 and GDM was nonlinear ( Figure S2D). This also partly explains the difference between our and Chen's results in Shanghai, China [10].
Because of the involvement of MTHFR, MTR, and MTRR genes with the OCM pathway and the evidence that maternal folate and B 12 imbalance during pregnancy increase GDM risk, we evaluated the influence of SNPs of these genes on the etiology of GDM in our study. The preliminary results showed that pregnancies with GDM were more prone to have a genotype TT for MTHFR rs1801131 (Table S1). After adjustment for multiple covariates, including maternal characteristics and OCM nutrients, MTHFR rs1801131 was found to be associated with GDM in the dominant model. Compared with the TT genotype, the GG/TG genotypes of rs1801131 were associated with a significantly lower risk of GDM before and after adjustment for multiple covariates (Table 3). Under the additive model, the presence of one or two copies of the G allele was associated with a reduced GDM risk. Our novel findings suggested that the minor G allele represents a protective factor in GDM. In turn, the TT genotype of MTHFR rs1801131 is a risk factor for GDM. Furthermore, we did not find a significant association between MTHFR rs1801133 and GDM, which agrees with the findings of previous studies [11,24,25].
In order to evaluate the effect modification of genetic variants on the associations of OCM indicators with GDM, a stratified analysis was performed. The results indicated that the associations between OCM indicators and GDM are heterogeneous in different genotypic groups. Pregnant women with TT genotype of MTHFR rs1801131 were more susceptible to folate and B 12 related GDM (Table 4). These findings were also supported by the joint association of the OCM nutrients with GDM stratified by rs1801311 genotype using the BKMR model ( Figure 2B,D), which showed significant associations of folate and B 12 with GDM among pregnancies with rs1801131 TT genotype. By estimating the nonlinearity of the exposure-response function in the BKMR model, serum folate showed a positive association with GDM both in GG/TG and TT subgroups. However, the association between B 12 and GDM was different between the two groups (Figure 2A,C). These findings also confirmed the RCS results ( Figure S2A and Figure 2B), which showed a linear relationship between folate and GDM but a nonlinear relationship between B 12 and GDM.
The mechanisms of increased GDM risk with excess folate and low B 12 are not well studied. Nutrients in the OCM pathway are interrelated, and disturbances in one nutrient will affect the status of others. The MTHFR rs1801131 G allele is associated with reduced enzyme activity. Therefore, pregnant women with the TT genotype can typically reduce 5, 10-methylene-THF to 5-methyl-THF. The reduction of 5, 10-methylene-THF by MTHFR is physiologically irreversible [38]. However, B 12 deficiency will impede OCM by trapping folate as 5-methyl-THF since the MTR/MTRR mediates methyl group donation from 5methyl-THF to Hcy requires B 12 as a cofactor [39]. The deficiency of B 12 will lead to an accumulation of 5-methyl-THF. This also explains why the relationship between folate, B 12 , and GDM is more dependent on the genotype of rs1801131 (Table 4 and Figure 2). More recently, an animal study indicated that dams that feed higher levels of 5-methyl-THF during pregnancy gained significantly more weight than dams that provide folic acid [40]. Their findings suggested that high-dose 5-methyl-THF exposure may play a role in the development of metabolic disease. On the other hand, OCM disturbance (enzyme activity or vitamins imbalance) substantially impacts the epigenome [41], such as DNA methylation, which may relate to GDM [42].
Our study has several limitations. First, we studied the relationship between OCM nutrients and GDM by expanding the sample size. However, the nature of the crosssectional baseline study limited us to determining the causality of OCM nutrient imbalance and GDM. Nevertheless, the genetic models can be used to explore causal relationships among OCM related SNPs and GDM. Second, twelve SNPs were genotyped in the present study. Other SNPs related to the OCM pathway are encouraged to further explore the effect modification of genetic variants on the association of OCM nutrients with GDM. Third, total folate was determined in the present study, and we cannot distinguish between 5-methyl-THF and other forms of folate. Further confirmatory studies and mechanistic investigations are required to verify the potential role of 5-methyl-THF in GDM.

Conclusions
Our study demonstrated that higher folate and lower B 12 , as well as MTHFR rs1801131, may be independent risk factors for GDM. In addition, pregnancies with rs1801131 TT genotype are more susceptible to OCM related GDM. To our knowledge, this is the first epidemiologic study to use a gene-nutrient interaction approach to evaluate the combined effects of OCM related nutrients and gene polymorphisms on GDM risk. More importantly, our findings potentially lead to practically feasible GDM prevention via individualized intervention in the future.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Tianjin Xiqing Hospital for studies involving humans (xqyyll-2020-07, 7 August 2020).

Informed Consent Statement:
Written informed consent has been obtained from all participants involved in this study.

Data Availability Statement:
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.