Joint associations of folate, homocysteine and MTHFR, MTR and MTRR gene polymorphisms with dyslipidemia in a Chinese hypertensive population: a cross-sectional study

Background Dyslipidemia is a well-established risk factor for cardiovascular disease. Serum lipids were affected by several gene polymorphisms, folate, homocysteine and other metabolite levels. We aim to investigate the effects of homocysteine metabolism enzyme polymorphisms (MTHTR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) and their interactions with folate, homocysteine on serum lipid levels in Chinese patients with hypertension. Methods Participants were 480 hypertensive adults that enrolled in September to December 2005 from six different Chinese hospitals (Harbin, Shanghai, Shenyang, Beijing, Xi’an, and Nanjing). Known MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G genotypes were determined by PCR-RFLP methods. Serum folate was measured by chemiluminescent immunoassay, homocysteine was measured by high-performance liquid chromatography, serum lipids parameters were determined by an automatic biochemistry analyzer, low-density lipoprotein was calculated by Friedewald’s equation. Unitary linear regression model was used to assess the associations of gene polymorphisms, folate and homocysteine on serum lipid profiles. Unconditional logistic regression model was applied to test the interactions of folate, homocysteine and gene polymorphisms on dyslipidemia. Results No correlations between gene polymorphisms and homocysteine on serum lipid profiles. Compared with normal folate patients, patients with low folate showed higher odds of hypertriglyceridemia (OR = 2.02, 95 % CI: 1.25-3.25, P = 0.004) and low levels of high-density lipoprotein cholesterol (OR = 1.88, 95 % CI: 1.07-3.28, P = 0.027). Each of four gene polymorphisms (MTHTR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) combined with low folate showed higher odds of hypertriglyceridemia (P for trend: 0.049, 0.004, 0.007 and 0.005, respectively). MTHFR C677T and A1298C with low folate showed higher odds of low levels of high-density lipoprotein cholesterol (P for trend: 0.008 and 0.031). Conclusions Low folate status and homocysteine metabolism gene polymorphisms (MTHTR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) may have a synergistic effect increased the incidence of dyslipidemia in Chinese hypertensive population.


Background
Cardiovascular diseases (CVDs) are still a major cause of morbidity and mortality worldwide, which is expected to remain the same during the foreseeable future [1]. Dyslipidemia is a major risk factor for stroke, coronary artery disease, atherosclerosis and other CVDs [2][3][4]. Sun et al. showed that 64.4 % of rural Chinese adults had at least one type of abnormal lipid concentration [5]. The prevalence of dyslipidemia in Chinese was higher than that in American [6], Canadian [7], and Iranian population [8]. In addition, abnormal folate or homocysteine (Hcy) level was also exacerbated the incidence of CVDs [9,10]. Recently, a large sample of clinical trials showed folate supplements effectively lower Hcy level and reduced the risk of stroke [11]. However, the potential mechanisms of folate, Hcy and serum lipid levels remain to be explored.
Our previous study found Hcy metabolism gene polymorphisms (MTHFR and MTRR) are significantly associated with serum lipid levels [12]. These findings suggest that there is an inherent correlation among folate, Hcy and lipid profiles. Additionally, several studies showed that medical nutrition therapy with folate intake significantly reduced serum lipids levels [13,14]. Thus, we speculate that Hcy-metabolism gene polymorphisms associated with folate and Hcy interactively affect the incidence of dyslipidemia.
Several key enzymes, including methylenetetrahydrofolate reductase (MTHFR) [15], methionine synthase (MTR) [16] and methionine synthase reductase (MTRR) [17] are important in folate and Hcy metabolism and also in methylation reactions. MTHFR converts 5, 10methylenetetrahydrofolate into 5-methyltetrahydrofolate and this reaction provide a methyl for homocysteine into methionine in the catalyzed by MTR [18]. MTR requires vitamin B 12 (cobalamin) as a coenzyme. Over time, the cobalamin(I) cofactor of MTR will be oxidized to cobalamin(II) which leads to inactivation of MTR. And MTRR restore oxidized cobalamin(II) to CH 3 -cobalamin(III) in order to maintain the activity of MTR [17]. MTHFR C677T (rs1801133) is a common mutation of MTHFR and its suggested being an unfavorable factor for cardiovascular diseases [19,20]. Yilmaz et al. showed that the C allele has a protective effect on blood lipid concentration and the T allele has a harmful effect when they are screening MTHFR C677T polymorphism in renal transplant patients [21]. However, another mutation MTHFR A1298C (rs1801131) seemed have no influences on serum lipids levels [12]. In hyperlipidemia patients, MTR A2756G (rs1805087) mutation was associated with a higher level of total cholesterol and low-density lipoprotein cholesterol [22]. MTRR A66G (rs1801394) is a susceptible marker of congenital heart defects [23], but the effect of this mutation on lipid profiles still no conclusion.
We hypothesized that homocysteine metabolism gene polymorphisms in combination with a low folate level or a high Hcy level may affect serum lipid profiles. In this study, we aim to explore the independent associations of folate, Hcy and these four gene polymorphisms (MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) on serum lipid profiles, as well as to evaluate the joint effect of gene, Hcy and genotypes on dyslipidemia in Chinese hypertensive patients.

Clinical characteristics and genotype distributions
A total of 480 patients were recruited for this study. Since 55 patients missing the data of genotypes or lipids parameters were excluded, a total of 425 subjects were included in our final analysis. The average age of our patients was 56.7 ± 9.9 years, including 185 (43.5 %) men and 240 (56.5 %) women (Table 1). Genotype frequencies in the total population were showed in Table 2. Four gene polymorphisms (MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) in this population had no deviation from the Hardy-Weinberg equilibrium (P values were 0.885, 0.384, 0.937 and 0.400, respectively).

Associations of folate, Hcy and lipid profiles
The associations of serum folate and lipid profiles were showed in Table 3. Although folate may have no influence on serum TC and LDL-C, there was a significant negative correlation between folate and TAG before adjusted covariates (β (SE): -0.11 (0.01), P = 0.029). However, when adjusted several confounding factors this correlation disappeared. Furthermore, we observed a positive correlation between folate and HDL-C (crude β (SE): 0.15 (0.00), P = 0.002; adjusted β (SE): 0.14 (0.00), P = 0.002). We further used folate as a dichotomous variable to assess the influences of low folate on dyslipidemia ( We also analyzed the associations of Hcy and these lipid profiles. However, we had no significant findings (data not show). And high Hcy status had no effect on the incidence of dyslipidemia (data not show).

Associations of genotypes and lipid profiles
The correlations of these four gene polymorphisms and serum lipids levels were showed in Table 5. There was a marginal significance that the MTRR 66GG carriers had a higher serum total cholesterol level compared with 66AA carriers (β (SE): 0.14 (0.09), P = 0.054). We further analyzed the influences of gene polymorphisms on dyslipidemia risk, and the results were not reached significance (data not show).

Interactions of folate, Hcy and genotypes on dyslipidemia
Compared with normal folate and wild-type carriers, those with low folate and mutant genotypes showed higher odds of hypertriglyceridemia (Table 6). In trend test, all these gene polymorphisms (MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) with low folate increased the prevalence of hypertriglyceridemia (interaction of MTHFR C677T and folate: P = 0.049; interaction of MTHFR A1298C and folate: P = 0.004; interaction of MTR A2756G and folate: P = 0.007; interaction of MTRR A66G and folate: P = 0.005, respectively). Table 7 showed the interactions of genotypes and low folate on low levels of high-density lipoprotein cholesterol. Compared with normal folate and MTHFR 677CC carriers, patients with both low folate and 677CT + TT genotype showed higher odds of low levels of HDL-C (OR (95 % CI): 2.67 (1.08-6.60), P = 0.033). Furthermore, low folate patients with the mutant allele of MTHFR A1298C showed a marginal significance higher incidence of low levels of HDL-C (OR (95 % CI): 2.52 (0.97-6.57), P = 0.059) than high folate patients carried 1298AA genotype. These findings were also confirmed by the trend test (interaction of MTHFR C677T and folate: P = 0.008; interaction of MTHFR A1298C and folate: P = 0.031, respectively). However, for interactions of genotypes and low folate on hypercholesterolemia and high levels of LDL-C, we have no significant findings.
We also carried out the analysis of interactions of high Hcy and genotypes with dyslipidemia profiles. Regrettably, no dyslipidemia odds were observed to be associated with high Hcy and mutant genotypes (data not show).

Discussion
In the present study, we investigated the association of folate, Hcy and several gene polymorphisms (MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) on serum lipid profiles and assessed the interactions of folate, Hcy and genotypes with dyslipidemia. The results showed patients with folate deficiency showed higher odds of hypertriglyceridemia and low levels of HDL-C. Furthermore, these mutant genotypes  combined with low folate may synergistically affect the incidence of hypertriglyceridemia and low levels of HDL-C. However, we haven't found homocysteine directly or indirectly affected serum lipid profiles. The mutation of MTHFR C677T is located in the catalytic domain of the enzyme, cause an alanine to valine substitution at position 222 of the enzyme, and result in a thermolabile enzyme [15,24]. A previous study reported that homozygous mutation of 677TT reduced by approximately 70 % of the mean enzyme activity and the heterozygous mutation of 677CT reduced by approximately 35 % of the mean MTHFR activity [15]. It is no doubt that MTHFR C677T polymorphism was related to elevate plasma Hcy concentration and lower folate level [25][26][27]. Whereas its effect on lipid profiles is no conclusion. Frelut et al. showed MTHFR C677T significantly increased LDL-C level, and this mutant may also associate with higher total cholesterol and triacylglycerols levels, and lower HDL-C level, although it did not reach significance [28]. In polycystic ovary syndrome women, patients with 677CT genotype have higher serum total cholesterol and triacylglycerols than 677CC carriers [29]. In northern Chinese subjects with hyperlipidemia, individual carried MTHFR 677CT + TT genotypes showed higher triacylglycerols and total homocysteine compared with wild-type [22]. However, Chen et al. found no significant association between MTHFR C677T and lipid profiles (TC, TAG, HDL-C and LDL-C) in total population, but in subgroup analysis, the mutant T allele carriers have higher total cholesterol and low-density lipoprotein cholesterol than 677CC group in women [30]. In another Chinese clinical research, both plasma Hcy level and MTHFR C677T mutant have no association with hyperlipidemia or serum lipid levels [31]. In our study, MTHFR C677T may not affect serum lipid levels, and high homocysteine concentration also has no effect on serum lipid levels or dyslipidemia profiles. Furthermore, interaction of MTHFR C677T and Hcy also showed no influence on dyslipidemia. However, there was a trend for the combination of MTHFR 677CT + TT and low folate significantly increased the incidence of hypertriglyceridemia and low levels of HDL-C. Mikael et al. found that MTHFR deficiency affects apolipoprotein levels and leads to lipid deposition, in mice with MTHFR deficiency showed higher plasma TAG level than controls [32]. Based on the negative correlation between MTHFR C677T and folate level [33], and the detrimental effect of folate deficiency on lipid metabolism [34], we speculate that MTHFR C677T polymorphism and folate deficiency interactively increased the prevalence of dyslipidemia.
The mutation of MTHFR A1298C resulted the substitution of adenine to cytosine at nucleotide 1298 position, cause a glutamic acid to alanine substitution at position 429 of the peptide, is located in the regulatory domain named NADPH and S-adenosyl methionine binding site [24,35]. Experimental results show that the A1298C mutation wasn't caused a thermolabile enzyme [35]. Publications about the influence of MTHFR A1298C mutant on blood lipid profiles were relatively small. In a previous study, both MTHFR C677T and MTHFR A1298C polymorphisms not affect total cholesterol level [36]. Additionally, Chang et al. found no association between MTHFR C677T and MTHFR A1298C on blood lipid profiles (TC, TAG, HDL and LDL) [37]. These results were consisted with our findings, MTHFR A1298C independent not associated with serum lipids level and also have no effect on dyslipidemia profiles. However, low folate patients carried MTHFR 1298 AC + CC showed a higher risk of hypertriglyceridemia and low levels of HDL-C compared with the high folate patients with 1298AA genotype, although the latter was marginal significance. An in vitro experiment showed folate has a direct protective effect on LDL oxidation and its doserelated [38]. However, an animal experiment showed folate deficiency did not influence triacylglycerol levels [32]. Therefore, the potential mechanisms of the joint effect between MTHFR A1298C and folate deficiency on dyslipidemia need further research.
In our study, MTR A2756G polymorphism independently not affects serum lipid profile or dyslipidemia. However, the mutant allele with low folate may elevate the incidence of hypertriglyceridemia. The A2756G mutation in MTR that located at position 919 of the protein, results in substitution of glycine for aspartic acid [39]. It's located in a domain of the protein that interacts with S-adenosyl methionine (SAM) and auxiliary proteins that are required for the reductive methylation and reactivation of the vitamin B 12 cofactor, which can be inactivated by oxidation during catalysis [17,40]. Therefore, it is possible that the mutant might impair the binding of SAM and/or auxiliary proteins [40] and increased plasma homocysteine concentration [41]. In hyperlipidemia patients, MTR 2756AG + GG carriers have higher total cholesterol and low-density lipoprotein cholesterol than those 2756AA carriers [22]. Additionally, the prevalence of 2756AG + GG genotypes in the combined hyperlipidemia group (hypercholesterolemia and hypertriglyceridemia) was significantly higher than that in the control group [22]. Regardless of some research found no association between MTR A2756G and folate level [25,27], a study showed serum folate level, MTHFR 677CC + CT and MTR 2756AA genotypes have a significant interaction on total homocysteine concentration in pregnant woman [25]. Furthermore, individuals with low intake of folate, vitamin B 6 and vitamin B 12 , the MTHFR 677 T allele and MTR 2756G allele were associated with a higher risk of breast cancer [42]. These results indicate that the mutation of MTR A2756G with other unfavorable factors may affect serum lipid profiles. Our study showed a borderline significant that MTRR 66GG elevated serum TC level. Difference with our results, a previous study suggested that individual with MTRR 66GG genotype have lower serum TC and LDL-C level than 66AA carriers in Chinese hypertensive patients [12]. The mutation of MTRR A66G is the substitution of A for G at nucleotide 66 with a substitution of an isoleucine by a methionine. It has been suggested that the mutation is located in the putative flavin mononucleotide-binding domain of the MTRR enzyme that interacts with MTR [17] and thus disrupt the binding of MTRR to the MTR-cobalamin-complex, thereby decreasing the rate of homocysteine remethylation [39]. Our results showed MTRR 66AA patients with low folate had a 3.4-fold risk of hypertriglyceridemia and a 4.7-fold of high levels of LDL-C. Numerous studies showed MTRR A66G was a risk factor for congenital heart defects [23,43,44]. But the effect of this mutation on blood lipid profiles was lack of research. So we believe that the joint associations of MTRR A66G and folate with dyslipidemia are worthy to be further study.
Folate plays an important role in one-carbon metabolism. It serves as a catalytic substrate for the transfer of one-carbon units. When fed folate deficiency or methyl donor deficiency diets, rats showed lower hepatic folate, phosphocholine (PC) storage and higher plasma homocysteine, glycine, serine and threonine concentrations [45]. Furthermore, compared with control animals, rats with folate or methyl donor deficiency showed several protein regulation disorder and abnormal gene expression in liver [45]. Additionally, folate deficiency will elevate liver triacylglycerols level in spontaneously hypertensive rats [46]. These studies indicate that folate deficiency will damage the hepatic lipid homeostasis. Although short-term folate deficiency has weak impact on lipid metabolism, a long-term folate deficiency showed a marked hepatic lipid accumulation in rats [45,47]. In our study, folate was associated with higher serum TAG and lower serum HDL-C level, patients with low folate showed higher odds of hypertriglyceridemia and low levels of HDL-C. Similar to our findings, Kim et al. found serum folate levels were significantly and positively correlated with HDL cholesterol and negatively with TAG, although the latter showed borderline significance [48]. In homocysteine metabolism, folate as a methyl donor that provides methyl for the synthesis of S-adenosyl methionine (SAM), SAM is a key intermediate in PC synthesis [49]. Decreased SAM level will increase S-adenosyl homocysteine (SAH), a competitive inhibitor of many methyltransferase reactions. So an important function of folate is maintenance the cellular SAM and SAH concentrations [34]. Therefore, folate deficiency will reduce PC synthesis resulting in accumulation of hepatic TAG that synthesis from phosphatidylethanolamine N-methyltransferase (PEMT) derived PC [34]. Furthermore, folate deficiency also reduced PEMT activity and choline kinase expression, and induced the expression of genes involved in hepatic lipid synthesis [34]. In oral contraceptive-treated rats, those with low dietary folate showed lower plasma HDL level and higher LDL level than control groups [50]. Another study showed higher serum folate level was associated with a lower level of LDL-C, higher levels of HDL-C and a lower LDL-C-C/HDL-C-ratio [51]. Elevated level of LDL is related to pathogenesis of atherosclerotic vascular disease whereas HDL has antioxidant effect that beneficial for against atherosclerosis and antiinflammatory [3]. High Hcy was also suggested inducing hepatic cholesterol biosynthesis and lipid accumulation [52]. The liver is the main sites for lipids and Hcy metabolism. Hypomethylation associated with hyperhomocysteinemia was related to lipid accumulation in tissues, elevated Hcy increased SAH level thus inhibited PEMT and lower the production of PC from phosphatidylethanolamine [49]. An animal experiment showed a significant positive correlation between plasma Hcy and triacylglycerols in female mice [32].
Additionally, studies showed that serum lipid profiles were also regulated by vitamin B 6 and vitamin B 12 . Fat oxidation requires carnitine and the endogenous synthesis of carnitine mainly occurs in the liver and kidneys and requires lysine, methionine and vitamin B 6 [53]. Furthermore, vitamin B 6 plays an important role in the desaturation and elongation of fatty acids, the mobilization of unsaturated fatty acids from triglycerides to phospholipids, and methylation of phospholipids [54]. Low vitamin Low HDL-C: serum HDL-C < 1.04 mmol/L; High HDL-C: serum HDL-C ≥ 1.04 mmol/L c Adjusted for sex, age, clinical centers, height and weight B 6 level was reported to impair carnitine synthesis and hence alter the lipid profiles. However, vitamin B 6 supplementation was found to reduce the different components of the lipid profile by~10 %, and significantly lowering TC and HDL-C level [53]. Vitamin B 12 is a crucial constituent of the one carbon cycle and plays an important role in the homocysteine metabolism. It's also a cofactor for the mitochondrial enzyme methyl malonyl CoA mutase and regulates the rate of long chain fatty acyl-CoA transfer into the mitochondria. Vitamin B 12 deficiency leads to the accumulation of fatty acids in the cytosol thereby influencing lipid metabolic pathways [55]. Adaikalakoteswari et al. reported that vitamin B 12 deficiency increased serum triglycerides and HDL level in Europeans and Indians with type 2 diabetes [56]. An animal experiment also showed the adverse effect of vitamin B 12 deficiency on plasma lipid profiles and liver fatty acid levels [57]. However, Wasilewska et al. found no association between B vitamins (folate, vitamin B 6 or vitamin B 12 ) and lipid profiles in any examined group [58]. Therefore, we speculate that B vitamins, some metabolites and genotypes may synergistically affect the hepatic lipid metabolism and serum lipid profiles. However, the molecular mechanism of these metabolites under conditions of different gene polymorphisms on lipid profiles is not fully understood, and need for more extensive research.

Conclusion and limitations
Our study demonstrated that there was a joint effect of homocysteine-metabolism gene polymorphisms (MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G) and low folate on dyslipidemia in Chinese hypertensive patients. However, the underlying mechanism is not clear. Therefore, we believe that future researches could be more effective to elucidate these findings in a wider population.
Our study had several limitations. Firstly, this was a cross-sectional study that we can't make a prediction about the incidence of dyslipidemia in the future, and the prevalence-incidence bias may also exist. Furthermore, our study sample was relatively small, and it may lack of effect when performed subgroup analysis. Next, serum lipid profiles may modify by vitamin B 6 and vitamin B 12 , and we didn't determine these B vitamins. Additionally, the limitation of generalization in our study was the study population only contained hypertensive adults, so these results need to be verified in a healthy population. In addition, serum lipid profiles may relate to dietary habits that we have not included.

Participants and procedures
This study was conducted using data collected in a previous study [59]. This was a multicenter, randomized, double-blind controlled trial in hypertensive Chinese adults. Details regarding "Study subjects", "Randomization and double blinding", "Data collection procedures", and "Laboratory tests" have been previously described [59]. Totally, 480 patients with mild or moderate hypertension were recruited from six hospitals in different Chinese regions (Ha'erbin, Shanghai, Shenyang, Beijing, Xi'an, and Nanjing) from September to December 2005. All six hospitals were certified as clinical pharmacology centers by the State Food and Drug Administration in China. This study was approved by the Ethics Committee of Peking University First Hospital, Beijing, China. The purpose and procedures of the study were carefully explained to all participants, and written informed consent was obtained from each participant.

Anthropometric measurements
The participants were invited to our clinic center at 8 am the next day after fasting overnight. After 60 min resting in a supine position, supine blood pressures were measured by using a mercury sphygmomanometer. No alcohol, cigarette smoking, coffee or tea was taken before the measurements. Blood pressure was measured by trained nurses. According to World Health Organization (WHO) standardized criteria, SBP was recorded at the appearance of sounds (first Korotkoff sounds) and DBP was recorded at the disappearance of sounds (fifth Korotkoff sounds). Three consecutive measurements were taken on the arms with 30s interval between replicates. If the difference between the measurements was more than 4 mm Hg, the patient was asked to rest for 5 min, and then repeated the measurements. In all of our analysis, the average of three consecutive blood pressure readings was used. Height was measured without shoes to the nearest 0.1 cm on a portable stadiometer. Weight was measured in light indoor clothing without shoes to the nearest 0.1 kg. Body mass index (BMI) was calculated as weight (kilograms)/ height (meters) squared.

Blood sample collection and laboratory methods
After 10-12 h of fasting, a venous blood sample was obtained from each participant. Serum or plasma samples were separated within 15 min of collection, and were analyzed within 30 min or stored at -80°C for later analysis. Blood samples collected at baseline were used for the measurement of homocysteine (Hcy), folate and lipids levels. Serum lipids including total cholesterol (TC), triacylglycerols (TAG) and high density lipoprotein cholesterol (HDL-C) were performed in the six study center laboratories, using standard reagents and an automatic biochemistry analyzer. Low density lipoprotein cholesterol (LDL-C) was calculated by Friedewald's equation. Plasma Hcy concentration was determined in duplicate by high-performance liquid chromatography. The intra-and inter-assay coefficients of variation were 3.5 % and 4.2 %, respectively. Serum folate was determined by chemiluminescent immunoassay using a Beckman Coulter ACCESS Immunoassay System (Beckman-Coulter Canada Inc., Mississauga, Canada). The intraand inter-assay coefficients of variation were 2.3 % and 3.7 %, respectively. All sample collection and tests were performed in an identical manner following the same standard protocol.

DNA extraction and genotyping
All participants were requested to provide 2 ml peripheral whole blood, which was collected in ethylenediaminetetraacetic acid (EDTA) and stored at -20°C. Genomic DNA was extracted from the cell pellet in whole blood by the QIAamp Blood Kit (Qiagen, Valencia, California, USA) and evaluate for quality by 1 % agarose gel electrophoresis. TaqMan probe technique was used for detecting gene polymorphisms of the Hcy pathway in the central laboratory. The polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) method was applied to detect the MTHFR C677T, MTHFR A1298C, MTR A2756G and MTRR A66G genotypes. Universal reaction conditions for each genotyping are as follows: 4 ng dried DNA, 0.08 mL 40 assay locus-specific probe, and 2.0 mL TaqMan universal polymerase chain reaction (PCR) master mix made to a final volume of 4 mL with 1.92 mL of sterile water. Plus the main parameters for the PCR-RFLP of the four single nucleotide polymorphisms (SNPs) are shown in Table 8. The amplified PCR products were separated on 3 % agarose gel. To ensure the accuracy of the genotype, each sample had twice genotyping for the SNP in our present study by two independent researchers. Genotyping call rate for assessments of all genetic variants was ≥ 98 % in this study. After excluding samples that consistently failed, we selected 10 % of total samples for replication, and concordance of 100 % was repeated for all samples.

Thresholds of dyslipidemia, low folate and high Hcy
According to Chinese Guidelines on Prevention and Treatment of Dyslipidemia in Adults [60], we defined the four types of dyslipidemia as: hypercholesterolemia (serum total cholesterol level > 5.18 mmol/L); hypertriglyceridemia (serum triacylglycerols level > 1.70 mmol/L); Low levels of high-density lipoprotein cholesterol (serum HDL-C level < 1.04 mmol/L); high levels of lowdensity lipoprotein cholesterol (serum LDL-C level > 3.37 mmol/L). In addition, low folate was considered to be serum folate level lower than 10 nmol/L [61], and high Hcy was defined as homocysteine concentration higher than 15 μmol/L [62,63].

Statistical analysis
Statistical analyses were using the IBM SPSS software package (version 19.0 for windows; IBM Inc. Armonk, NY, USA). The results for categorical variables (i.e. sex, clinical centers and genotypes) are presented as number and percentage of cases. Continuous variables (i.e. age, height, weight, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), folate, Hcy, TC, TAG, HDL-C and LDL-C) were given as the mean ± standard deviation. The means for continuous variables in the two groups were compared using Student's t tests, and the prevalence of categorical variables was compared using χ 2 tests. Hardy-Weinberg equilibrium for genotypic frequencies of four genes was assessed with the χ 2 test and Fisher's exact test. Unitary linear regression model was used to access the associations of folate, Hcy and gene polymorphisms on serum lipids levels. Unconditional logistic regression was performed to estimate the independently effects of folate, Hcy status and genotypes on dyslipidemia risk. The interactions of folate, Hcy and genotypes on the incidence of dyslipidemia were also adopted the unconditional logistic regression model and used trend test for further verification. Since the number of cases was small, we adopted the stepwise logistic regression method (Backward: Wald). All analysis were adjusted potential confounding factors, including sex, age, clinical