GWAS of Folate Metabolism With Gene–environment Interaction Analysis Revealed the Possible Role of Lifestyles in the Control of Blood Folate Metabolites in Japanese: The J-MICC Study

Background The present genome-wide association study (GWAS) aimed to reveal the genetic loci associated with folate metabolites, as well as to detect related gene–environment interactions in Japanese. Methods We conducted the GWAS of plasma homocysteine (Hcy), folic acid (FA), and vitamin B12 (VB12) levels in the Japan Multi-Institutional Collaborative Cohort (J-MICC) Study participants who joined from 2005 to 2012, and also estimated gene–environment interactions. In the replication phase, we used data from the Yakumo Study conducted in 2009. In the discovery phase, data of 2,263 participants from four independent study sites of the J-MICC Study were analyzed. In the replication phase, data of 573 participants from the Yakumo Study were analyzed. Results For Hcy, MTHFR locus on chr 1, NOX4 on chr 11, CHMP1A on chr 16, and DPEP1 on chr 16 reached genome-wide significance (P < 5 × 10−8). MTHFR also associated with FA, and FUT2 on chr 19 associated with VB12. We investigated gene-environment interactions in both studies and found significant interactions between MTHFR C677T and ever drinking, current drinking, and physical activity >33% on Hcy (β = 0.039, 0.038 and −0.054, P = 0.018, 0.021 and <0.001, respectively) and the interaction of MTHFR C677T with ever drinking on FA (β = 0.033, P = 0.048). Conclusion The present GWAS revealed the folate metabolism-associated genetic loci and gene–environment interactions with drinking and physical activity in Japanese, suggesting the possibility of future personalized cardiovascular disease prevention.


INTRODUCTION
Folic acid (FA) is involved in the transfer of one-carbon units of thymidylate, purines, and methionine. 1 Intake of FA is essential for DNA synthesis, stability, repair, and normal cell division, especially during rapid growth such as embryonic development and cancer. 1 The lack of FA in pregnancy is associated with an increased risk of neural tube defects. 2,3omocysteine (Hcy) is an established blood risk marker of atherosclerosis and subsequent cardiovascular diseases (CVDs).5][6][7] The increased risk of dementia and Alzheimer's disease is also associated with elevated Hcy. 8 The heritability of Hcy is estimated to be 47-70%, [9][10][11][12] and gene mutations in Hcy metabolic enzymes are well-known to be related to the levels of Hcy.
Vitamin B 12 (VB 12 ) works as a coenzyme for methionine synthase in the formation of Hcy to methionine.The enzyme 5methyltetrahydrofolate reductase (MTHFR) produces 5-methyltetrahydrofolate from 5,10-methyltetrahydrofolate, which is transferred to Hcy by methionine synthase to form methionine and tetrahydrofolate (Figure 1).In Hcy metabolism, FA deficiency and lack of VB 12 may lead to homocysteinemia, and subsequent vascular inflammation and increased risk of CVDs. 13The conditions of the lack of VB 12 also lead to megaloblastic anemia and neurodegeneration, and to cognitive decline.Moreover, it leads to an increased risk of neural tube defects. 14everal studies have reported associations between genetic polymorphisms related to the folate metabolizing pathway and Hcy metabolism.][17][18] In addition, one recent genome wide association study (GWAS) has revealed several functional variants that affect folate metabolisms in humans, including genomic variants in betainehomocysteine S-methyltransferase, folate hydrolase 1, and cystathionine beta-synthase genes. 19Several GWAS on folate metabolisms have also been conducted, but few studies have examined the interaction between these SNPs and lifestyles, 20 prompting us to investigate these associations to find a way for possible individualized prevention of human disorders related to folate metabolism, such as CVDs, in the future.
This study aimed to investigate the SNPs associated with blood FA, VB 12 , and Hcy levels in Japanese using the GWAS data of the Japan Multi-Institutional Collaborative Cohort (J-MICC) Study and to detect the interactions between genetic polymorphisms and lifestyle variables, such as smoking, alcohol intake, and physical activity (PA), on blood Hcy concentrations, FA, and VB 12 among Japanese.

Study subjects
This study is part of the J-MICC study.The J-MICC study is one of the largest genome cohort studies in Japan, conducted at 13 independent universities and research institutions whose main goal is to detect gene-environment interactions mainly for cancer prevention. 21The recruitment for participants began in 2005 and ended in March 2014, with a total of 92,610 participants nationwide. 22At the baseline survey, the volunteer participants aged 35-69 completed a self-administered questionnaire and provide blood samples after informed consent.We collected blood samples in a 7-mL vacuum tube for serum and a 7-mL EDTA-2Na-containing vacuum tube for plasma and buffy coat.DNA was extracted from buffy coat and provided for genotyping.In the present study, we measured and analyzed plasma Hcy, FA, and VB 12 levels of 2,263 participants from eight independent study sites of the J-MICC study (Chiba, Aichi Cancer Center, Shizuoka, Daiko, Kyoto, Okazaki, Saga and Kagoshima), who participated in the study from the year of 2005 to 2012.
In the replication phase, we used a single data set consisting of 572 participants from the Yakumo Study. 23The Yakumo Study is an epidemiological study conducted annually on health checkup examinees aged ≥39 years who resided in Yakumo-cho, Hokkaido, Japan since the year of 1982. 23,24We asked the 593 health check-up examinees residing in Yakumo-cho, Hokkaido to complete a self-administered questionnaire and provide blood samples after explanation of the aim and the procedure of the study, of whom 572 (96.5%) provided the consent for genetic testing and participated in the study.
Informed consent was obtained from all participants in this study.The protocol of this study was approved by the ethics committees of the Nagoya University Graduate School of Medicine (approval number: 253) and by each participating institution.All research procedures were conducted according to the Ethical Guidelines for Human Genome and Genetic Sequencing Research and the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan.

Sample measurement and phenotype definition
In the J-MICC Study, the plasma sample of participants was stored at −80°C until measurements were performed.The serum Hcy levels were measured with EIA (enzyme immune assay) using JCA-RX20 autoanalyzer in a laboratory (SRL Inc., Hachioji, Japan) in Yakumo Study, and plasma Hcy levels were measured with LC-MS/MS (liquid chromatography-tandem mass spectrometry) in the J-MICC Study; serum FA and VB 12 levels in Yakumo Study and plasma FA and VB 12 levels in the J-MICC Study were measured with EIA in the same laboratory.Measurements of Hcy   [25][26][27][28] suggesting that the results of these measurements are highly reliable.The clinically normal laboratory values are >4.0 ng/mL for plasma FA, 180-914 pg/mL for plasma VB 12 , and 3.7-13.5nmol/mL for plasma Hcy.In the Yakumo Study, the serum samples of the study participants were measured using the same methods.
Genotyping, imputation and quality control DNA samples were automatically extracted from the buffy coat using a BioRobot M48 Workstation (QIAGEN group, Tokyo, Japan).Genotyping for the discovery phase has been performed using the Illumina HumanOmniExpressExome ver1.umich.edu/wiki/Minimac3)software based on the 1,000 Genomes Project cosmopolitan reference panel (phase 3).After the genotype imputation, variants with MAF <0.05 and r 2 < 0.3 were excluded, leaving 6,288,024 variants for final analyses.In the discovery phase, the data for 2,263 subjects of 8 study areas (about 16% of the entire GWAS subjects), whose plasma folate levels were available, were used.In the replication phase, we determined the genotypes of all the discovered SNPs other than the SNP of NOX4 (NADPH Oxidase 4 ) by the TaqMan real-time polymerase chain reaction (PCR), using the StepOnePlus™ real-time PCR system (Thermofisher, Waltham, MA, USA).As the genotyping for Nox4 rs2289125 was technically difficult and unsuccessful, we adopted Nox4 rs10830278 as a surrogate marker, which was in tight linkage with rs2289125 (DA = 0.9792, r 2 = 0.8685).For the NOX4 rs10830278 polymorphism, we determined the genotypes using PCR with confronting two-pair primers (PCR-CTPP). 31The primers used (and the thermal cycler conditions) for NOX4 rs10830278 were as follows: F1: CTA TTA GGT TGA GCC ATA TAA AAT GGC TGA TT, R1: TCA TGT TGT CAC AAA TGG CAG GA, F2: ATG AGA TTA TAA AAG GGG CCA AGA ACT G and R2: TTG AAT CAT ATA GAT TGG TAG ATC AGA AAC AGT CAA AT (initial denaturation at 95°C for 10 min, followed by 30 cycles of 95°C for 1 min, 58°C for 1 min, and 72°C for 1 min, with a final extension of 72°C for 5 min).The representative gel for the genotyping is shown in eFigure 1.We also confirmed the genotyping results for rs10830278 using PCR-CTPP were completely replicated by TaqMan real-time PCR (primer-probe used: TaqMan SNP Genotyping, SNP ID: C_3223929_10).

Evaluation of lifestyle information
The lifestyle information was collected using a self-administered questionnaire by well-trained interviewers at the timing of baseline survey.3][34] Smoking habits were categorized as never, former, and current smokers, and pack-years of smoking were counted as well.Alcohol habits were categorized as never, former, and current drinkers, and the amount of alcohol consumption (g/day) was also estimated.Information on PA were collected at baseline survey of J-MICC Study using the questionnaire for the duration per episode in recent years and types of activity.The assigned intensity (in metabolic equivalents) was set based on International Physical Activity Questionnaire as follows 35,36 : labor work, 4.5; walking, 3.0; mild exercise (not breathtaking), 3.4; breathtaking hard exercise (talkable), 7.0; breathtaking hard exercise (untalkable), 10.0; standing, 2.0; and sitting, 1.5.The total amount of PA was calculated as the sum of the products of duration and intensity for each type of activity in metabolic equivalents + hour/day.PA was dichotomized based on the distribution of the metabolic equivalents + hour/day, which is the well-known unit for metabolic equivalent, where subjects were coded as 1 if their PAs were more than or equal to 33 percentile, and coded as 0 otherwise.

Statistical analysis
We examined the associations of the SNPs with the quantitative traits of plasma FA, VB 12 , and Hcy using the EPACTS software (http://genome.sph.umich.edu/wiki/EPACTS).The associations of SNPs with the natural logarithms of plasma FA, VB 12 , and Hcy as continuous variables were tested using linear regression.For covariates to be adjusted, gender, age, and the first five principal components were included.All 6,288,024 variants with the MAF greater than (or equal to) 0.05 were considered.Manhattan and Q-Q Plots were generated using the 'qqman' function in R (https:// cran.r-project.org/web/packages/qqman/index.html).Lead SNPs (or variants) were defined as those SNPs (or the variants) that reached the minimum P-values in each genetic locus, defined as the positions on the chromosome identified by the cytogenetic banding of the chromosome. 37The β coefficient for the gene-environment interaction was estimated based on linear regression with the multiplicative product term for the interaction of the number of minor alleles and lifestyles as binary variables.The GxEs of GWAS identified SNPs with lifestyles of smoking, alcohol consumption and physical activity on blood levels of folate metabolites (Hcy, FA and VB12) were examined.
For the analyses of GWAS, the genome-wide significance levels were set at P < 5 × 10 −8 , and suggestive levels were set at P < 1 × 10 −6 ; for the rest of the analyses, statistical significance levels were set at P < 0.05, where adjustments for multiple comparisons were not applied due to the exploratory nature of the analyses. 38

Study characteristics and the genome-wide association study of folate metabolites
The characteristics of the study participants were shown in GWAS of Folate Metabolism in Japanese Table 1.In the J-MICC Study, median plasma Hcy concentration were higher in men than in women (8.8; interquartile range [IQR], 7.2-10.8nmol/mL vs 6.8; IQR, 5.8-8.2nmol/mL, respectively; P < 0.001), but median plasma FA and VB 12 were higher in women than in men (855; IQR, 730-1,010 pg/mL vs 800; IQR, 700-945 pg/mL, respectively; P < 0.001).

Gene-environment interaction analyses
We also examined the gene-environment interactions of MTHFR C677T (rs1801133) with lifestyles (Table 3).Although the previously reported interaction of MTHFR C677T and smoking habits on plasma Hcy was only marginal (β = 0.029, P = 0.089), a significant interaction of MTHFR C677T and drinking habits on plasma FA, and the interaction of MTHFR C677T and PA on To validate these findings, we examined the corresponding interactions in an independent dataset from the Yakumo Study (Table 3).While only statistically marginal, we recapitulated the trend of the interactions of MTHFR 677 T=T with current drinking on FA and with PA on Hcy ( β = 0.074, P = 0.054, and β = −0.054,P = 0.100, respectively).
We also examined these observed interactions in a combined data set of J-MICC and Yakumo Study.A significant interaction between MTHFR C677T and drinking habits on plasma FA ( β = 0.037, P = 0.019 for the interaction of MTHFR C677T and ever drinking, β = 0.037, P = 0.020 for the interaction with current drinking), and the interaction between MTHFR C677T and PA on plasma Hcy were observed (β = −0.043,P = 0.005).(Table 3).In addition, the exhaustive examination of gene-environment interactions between the lead SNPs (and the surrogate, Nox4 rs10830278) found in the present GWAS and the lifestyles of smoking habits, drinking habits, and PA revealed several statistically significant associations, although most of the interactions based on smoking, drinking, or PA as continuous variables failed to reach any statistical significance, which may worth verifying in future studies (eTable 1).

Associations of folate metabolites with the reported loci
We also examined the associations of each of folate metabolites (Hcy, FA, and VB12) with previously reported loci based on GWAS catalogue (trait ID: EFO_0004578, EFO_0005111 and EFO_0004620).As a result, statistically significant associations of rs1801133 (in the MTHFR gene on chromosome 1), rs12085006, and rs1999594 (both on chromosome 1) with blood Hcy levels and that of rs1801133 with blood FA levels were observed (eTable 3).

DISCUSSION
Polymorphisms of MTHFR, DPEP1, NOX4, and FUT 2 have been previously reported to be associated with FA, VB 12 , and Hcy levels. 20,40,41However, no GWAS of the folate metabolic pathways has been reported in Japanese.The present GWAS suggested that polymorphisms of MTHFR, DPEP1, CHMP1A, FUT 2, and NOX4 may play an important role in the regulation of plasma Hcy levels in Japanese.Hcy remethylation requires a cosubstrate 5-methyltetrahydrofolate to form methionine. MTHFR catalyzes the reduction of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate. [42][43][44] The homozygotes of the thermolabile MTHFR 677T allele are shown to have reduced MTHFR enzyme activity, 45 which leads to higher blood Hcy levels and subsequent elevated risks of various vascular/atherosclerotic diseases. 67][48] NOX4 is expressed in endothelial cells, cardiomyocytes, and vascular smooth muscle cells, 49 and elevated expressions were reported in patients with hypertension, atherosclerosis, heart failure, and stroke. 50Upregulation of NOX4 may also contributes to Hcy-mediated apoptosis of endothelial cells. 51PEP1 is a kidney membrane enzyme that is highly expressed in the proximal convoluted tubules. 52Although their roles in Hcy metabolism remain unclear, they may be involved in renal handling and metabolism of Hcy and cysteine, a precursor of Hcy.DPEP1 was previously associated with Hcy in one analysis. 48e also found a reported association between CHMP1 polymorphism and plasma Hcy in humans. 47CHMP1A is involved in protein transport, and is essential for the proliferation and maintenance of neural progenitor cells. 53UT 2 (rs602662, rs492602, and the FUT2 haplotype) was associated with VB 12 levels, which confirmed previously reported findings.19,54,55 Another non-synonymous SNP in the FUT2 gene rs10447781 was strongly associated with VB 12 levels.FUT2 variants reportedly reduce H-type antigen production and function and decrease the risk of VB 12 malabsorption due to Helicobacter pylori infection and associated gastritis.56 In addition, FUT2 variants increase the secretion of fucosylated glycoprotein of gastric intrinsic factors required for VB 12 absorption.57 The functionality of MTHFR rs1801133 was already well demonstrated in previous reports 45 ; the rs2289125 SNP of NOX4 is located in the open chromatin in most tissues and thus may associate with transcription regulation. 46 The r1126464 SNP of DPEP1 and the rs1047781 of FUT2 is located in the nonsynonymous region and may exert effect through alteration of DPEP1 protein structure.Some of the SNPs found in the present GWAS in association with blood Hcy, FA, and VB12 compared with previously found SNPs in the GWAS were partially the same (rs1801133 of MTHFR and rs2289125 of NOX4) but other SNPs were different SNPs in the same genes/genomic loci.For example, the and rs71374191 of the DPEP1=CHMP1A locus found in association with blood Hcy levels were different from those reported in European GWAS (rs7130284 of NOX4 and rs154657 of DPEP1).47 Although the present study found the rs1047781 of FUT2 as the top significant SNP for VB 12 , the two SNPs of rs602662 and rs492602 of FUT2 are non-polymorphic in Japanese.The different SNPs found to be GWAS significant in association with blood folate metabolites may suggest the results of different evolutionary pressure between ethnicities and may lead to population-specific prevention strategies against diseases related to folate metabolism disorders, such as atherosclerosis and CVDs.58,59 In previous studies, a positive interaction between MTHFR C677T and smoking habits on plasma Hcy was reported.18,[60][61][62][63] The present study revealed the marginal interaction in the same direction, although it did not reach statistical significance.According to the Organisation for Economic Co-operation and Development (OECD) health statistics (https://stats.oecd.org/), the rate of male smokers is higher in Japan (around 30%) compared with those in other OECD member countries, such as United States or United Kingdom (around 10-20%).The interaction with smoking habits for MTHFR and vascular diseases risk may be explained by a similar mechanism via elevated Hcy.63 Considering the different prevalence of smokers between countries, further investigation with sufficiently large sample sizes in each country's population is needed to clarify the interaction of MTHFR C677T with smoking habits on blood Hcy levels.
This study observed a positive interaction between MTHFR C677T and drinking habits on blood FA (ie, as the T allele of MTHFR C677T and drinking amount increase, they have multiplicative interactive effects on the increment of FA), and a negative interaction between MTHFR C677T and PA on blood Hcy (ie, as the T allele of MTHFR C677T and calories spent by GWAS of Folate Metabolism in Japanese PA increase, they have multiplicative interactive effects on the reduction of Hcy) in Japanese.MTHFR 677T=T is a genotype similar to ALDH2 Lys=Lys, and individuals with MTHFR 677T=T may have avoidant and deterrent responses to alcohol consumption due to excessive savings of Hcy in those with MTHFR 677T=T, which may explain the possible gene-environment interaction between drinking habits and MTHFR 677T allele on blood FA levels. 64This might act against alcohol dependence and lead to attenuated FA reduction in those with MTHFR 677T=T genotype in drinkers, which is consistent with the results of previous studies. 20,65In the present study, however, no significant association between MTHFR 677T=T genotype and drinking behavior was observed.Considering that the recent meta-analysis revealed no association of MTHFR 677T=T and alcohol dependence, the effect of MTHFR genotype on drinking behavior might be relatively limited. 66In addition, the reduced blood FA levels in subjects with MTHFR 677 T=T genotype in the present study are in accordance with the previous report, 67 underscoring the requirements for more folate intake in those with MTHFR 677 T=T genotype than in those with other genotypes.The influence of MTHFR polymorphism on the association between PA and blood Hcy levels has been an issue of interest to researchers in recent decades. 65,68,69This study might add some beneficial evidence for possible personalized prevention of atherosclerosis/ CVDs based on genetic information in the near future.The detailed mechanisms of the reduction in plasma Hcy levels with exercise are not yet well described, which may be explained by mechanisms of protein turnover and the betaine pathway. 70,71The existence of different distributions of lifestyle factors, as well as different genotype distributions of SNPs, may lead to discovery of different GxEs and disease prevention measures specific to each population.Further investigations with sufficiently larger sample sizes in each population should be expected to clarify these associations.
The strength of the present study would be that this GWAS was conducted using only population-based cohort data from Japanese population with relatively large sample size, so it is possible that gene-environment interactions might be easier to detect without adjustment in Japanese population, where lifestyle components are relatively homogeneous.
Our study has several limitations.First, all lifestyle factors are self-reported and thus subject to bias, which is generally bias toward null.Second, we investigated the gene-environment interaction based only on the SNPs detected from GWAS results of folates (Hcy, FA, and VB 12 ) as outcomes.Although there is an alternative method to examine gene-environment interactions comprehensively based on GWAS data using the ProbABEL, the gene-environment interaction detection software in GWAS, we adopted a candidate SNP approach using GWAS detected SNPs because GWAS data using current folate measurement data (n = 2,263) are somewhat underpowered to detect robust geneenvironment interactions.In addition, the cut-off level for the dichotomization of PA was determined to be 33% based on our sensitivity analysis for different cut-off levels of PA, which was mainly due to the exploratory context of the present study.Further investigations with sufficiently large sample sizes for this potentially intriguing interaction of MTHFR 677T=T with PA on blood Hcy levels should be warranted.

Conclusion
The present GWAS revealed the significance of folate metabolism-associated SNPs in MTHFR, DPEP1, CHMP1, and NOX4 genes also in Japanese.In addition, the interactions of these SNPs with lifestyles of smoking, drinking and PA on blood folate metabolite concentrations may pave the way for the possible personalized prevention of atherosclerotic diseases in Japanese, as well as in world populations.Further investigations are needed to confirm the results of the present study.

Table 1 .
Characteristics of participants in the Japan Multi-Institutional Collaborative Cohort (J-MICC)

Table 3 .
The interactions of MTHFR C677T rs1801133 and lifestyles on plasma folate metabolites of homocysteine (Hcy)/folic acid (FA)/vitamin B 12 (VB 12 ) with lifestyle in the Japan Multi-Institutional Collaborative Cohort (J-MICC) and Yakumo study a Adjusted for age and sex.