Sex-specific association of sequence variants in CBS and MTRR with risk for promoter hypermethylation in the lung epithelium of smokers

Abstract

Gene promoter hypermethylation is now regarded as a promising biomarker for the risk and progression of lung cancer. The one-carbon metabolism pathway is postulated to affect deoxyribonucleic acid (DNA) methylation because it is responsible for the generation of S-adenosylmethionine (SAM), the methyl donor for cellular methylation reactions. This study investigated the association of single nucleotide polymorphisms (SNPs) in six one-carbon metabolism-related genes with promoter hypermethylation in sputum DNA from non-Hispanic white smokers in the Lovelace Smokers Cohort (LSC) (n = 907). Logistic regression was used to assess the association of SNPs with hypermethylation using a high/low methylation cutoff. SNPs in the cystathionine beta synthase (CBS) and 5-methyltetrahydrofolate-homocysteine methyltransferase reductase (MTRR) genes were significantly associated with high methylation in males [CBS rs2850146 (-8283G > C), 
OR = 4.9; 95% CI: 1.98, 12.2, P = 0.0006] and low methylation in females [MTRR rs3776467 (7068A > G), OR = 0.57, 95% CI: 0.42, 0.77, P = 0.0003]. The variant allele of rs2850146 was associated with reduced gene expression and increased plasma homocysteine (Hcy) concentrations. Three plasma metabolites, Hcy, methionine and dimethylglycine, were associated with increased risk for gene methylation. These studies suggest that SNPs in CBS and MTRR have sex-specific associations with aberrant methylation in the lung epithelium of smokers that could be mediated by the affected one-carbon metabolism and transsulfuration in the cells.

Abbreviations:

CBS

cystathionine beta synthase

DNA

deoxyribonucleic acid

HBEC

human bronchial epithelial cell

Hcy

homocysteine

LD, linkage disequilibrium; LSC

lovelace Smokers Cohort

MAF

minor allele frequency

MTHFR

methylenetetrahydrofolate reductase

MTRR

methyltransferase reductase

SNP

single nucleotide polymorphisms

SAH

S-adenosylhomocysteine

SAM

S-adenosylmethionine

Introduction

Worldwide, lung cancer is the leading cause of cancer death in men and the second leading cause of cancer death in women (1). Considerable excitement was generated over findings from the national lung screening trial that reported a 20% reduction in mortality from lung cancer with low-dose computed tomography screening compared with standard chest radiography (2). However, the high false-positive rate (>96%) of computed tomography screening substantiates the need for molecular biomarkers of lung cancer risk.

Gene promoter hypermethylation is now regarded as a promising biomarker for the risk and progression of lung cancer. Deoxyribonucleic acid (DNA) methylation in mammalian cells consist of the addition of a methyl group to the five carbon position of the cytosine ring in CpG dinucleotides. Hypermethylation of CpG islands within the promoter region of a tumor suppressor gene can lead to transcriptional repression, an important mechanism in cancer initiation and progression (3). Belinsky et al. (4) showed that prevalence of gene promoter hypermethylation in high-risk smokers was predictive of lung cancer up to 18 months before clinical diagnosis. These epigenetic changes reflect the well-documented field cancerization present throughout the aerodigestive tract of smokers with long-term exposure to tobacco carcinogens (5).

The one-carbon metabolism pathway is necessary for DNA methylation reactions and nucleotide synthesis. Folate and vitamin B12 are used in the conversion of homocysteine (Hcy) to methionine that, in turn, is used in the generation of S-adenosylmethionine (SAM), the primary methyl donor for DNA, RNA and protein methylation reactions in the cell. The vitamin B6 dependent transsulfuration pathway is utilized to clear Hcy by converting it to cystathionine and is influenced by levels of methionine (6,7).

The effects of one-carbon metabolism on DNA methylation has been a focus of investigation for many years. Recently, Stidley et al. (8) found that dietary intake of folate, multivitamins, and leafy green vegetables was protective against gene promoter hypermethylation in the sputum of high-risk smokers from the Lovelace Smokers Cohort (LSC). This observation is consistent with other studies in which manipulation of the level of methyl donors (folic acid, vitamin B12, methionine, etc.) in the diet-affected epigenetic reprogramming in vivo (9) and serum levels of B vitamins (folate and methionine) were associated with gene methylation in white blood cells (10). In addition to the epigenetic effects, B vitamins (including vitamin B6, methionine and folate) measured in serum collected several years prior to cancer diagnosis were also associated with reduced risk for colorectal and lung cancers (11,12). Thus, disregulation of the one-carbon metabolism pathway coupled with genomic instability may be an important mediator of DNA methylation that ultimately contributes to cancer risk.

The relationship between single nucleotide polymorphisms (SNPs) in folate-related genes such as methylenetetrahydrofolate reductase (MTHFR) and lung cancer has been intensively studied but with conflicting results. Although studies show a protective effect of the MTHFR C677T polymorphism in some populations, other studies suggest that this variant is associated with increased lung cancer risk (13–15). Two recent meta-analyses examined the relationship of MTHFR polymorphisms with lung cancer, but did not find evidence of a major role for MTHFR polymorphisms C677T or A1298C (16,17). However, an SNP–folate interaction was identified for C677T with a 28% increased risk in individuals with low folate intake (16). A similar MTHFR C677T–folate interaction was also identified in meta-analyses conducted in head and neck tumor and gastric tumor (16,18).

In colon cancer, MTHFR A1298C was consistently associated with an increased risk for the CpG island methylator phenotype status in two studies (19,20). No such studies have been conducted in lung. Based on the likelihood that dysregulation of one-carbon metabolism-related pathways may affect the propensity for acquiring de novo methylation in the lungs of smokers, we tested the hypothesis that genetic variants comprehensively selected in six key genes (CBS, MTHFR, MTR, MTRR, SCL19A1, TYMS) are associated with susceptibility for smokers to acquire gene-specific promoter methylation detected in sputum. Molecular validation of a significant SNP in the CBS gene, rs2850146 (-8283G > C), was conducted to assess its association with total homocysteine (tHcy) and cystathionine plasma concentrations and with CBS gene expression in primary human bronchial epithelial cell (HBEC) cultures. The association between plasma metabolites and risk for gene methylation was also explored in a subset of LSC members.

Materials and methods

Study population and sample collection

The LSC recruits both current and former smokers aged 40–75 with a minimum of 15 pack years of smoking as obtained from an initial screen. Members are primarily residents of the Albuquerque, New Mexico metropolitan area. Study enrollment began with females in 2001, but has included males since 2004 and is currently enrolling both men and women. Members complete a standard questionnaire covering demographics, smoking history and personal and family health history. Dietary information is collected using the Harvard University Food Frequency Questionnaire. Weight and height are measured during each visit. Members provide both blood and sputum samples and undergo standard pulmonary function testing. Plasma, mononuclear cells and red blood cells are separated using a density gradient centrifugation method (Histopaque®-1077) within 30min after blood samples are drawn. Plasma samples are aliquoted into 5ml tubes and frozen at −80°C. All members sign a consent form, and the Western Institutional Review Board and the University of New Mexico Human Research Review Committee approved this project. This chapter focuses on the 924 non-Hispanic white participants who had 
≥ 15 pack years of smoking as determined from the detailed questionnaire and a complete sputum methylation panel as outlined in the following subsections.

Methylation-specific PCR

Twelve genes (p16, MGMT, DAPK, RASSF1A, PAX5α, PAX5β, GATA4, GATA5, SULF2, PCDH20, DAL1 and JPH3) were selected for analysis of methylation in sputum due to their association with lung cancer in previous and ongoing studies (4,21). DNA was isolated from cytologically validated sputum, modified with bisulfite and used in nested methylation-specific PCR reactions to detect methylated alleles from individual genes as described previously (4,22). Sputum DNA samples from non-Hispanic white participants were assessed for prevalence of methylation in all 12 genes. The primer sequences and the thermo cycling conditions are available upon request.

Candidate gene and tag SNP selection

Six candidate genes (CBS, MTHFR, MTR, MTRR, SLC19A1 and TYMS) were chosen from the one-carbon metabolism and transsulfuration pathways based on their previous associations with lung cancer (14,23–27). Tag SNPs (n = 101) were selected using the pairwise R ² algorithm (R ² > 0.80) based on the phase 2 HapMap CEU database (28). Only SNPs with minor allele frequency (MAF) > 0.05 were included.

SNP genotyping and quality control

SNPs (n = 101) were genotyped using two Illumina GoldenGate genotyping assays and DNA from 924 LSC members. SNPs were genotyped at the University of Utah Genomics Core Facility except for MTHFR SNPs that were genotyped at the University of Southern California Epigenome Center using the same DNA. Those SNPs that completely failed were removed from the analysis (n = 12). Additionally, SNPs (n = 5) and samples (n = 17) with a <90% call rate were also removed. Five SNPs that failed the Hardy–Weinberg Equilibrium test (P < 0.0006, a Bonferroni corrected P value) were also removed from further analysis for genetic association. A total of 79 SNPs in 907 samples with an overall call rate of 99.3% (range from 90.5% to 100.0% across samples and from 92.2% to 100.0% across SNPs) were then analyzed for their association with risk for gene methylation. Inter- and intra-plate duplicate DNA and DNA from a family trio were included on each plate for quality control. The average concordance rate for genotypes between duplicates (n = 36) was 99.1% (range from 93.3% to 100.0%). For a plasma metabolite study, Taqman allelic discrimination assay was used to genotype a single SNP in the CBS gene, rs2850146, in DNA samples from the entire LSC cohort (n = 2165), the New Mexico Lung Cancer Study (n = 350) and HBECs (n = 79).

Plasma metabolite measurements

A random block design was used to examine the association between the CBS genotype for SNP rs2850146 and metabolites within the one-carbon metabolism and transsulfuration pathways in non-fasting plasma samples (n = 86). Plasma samples were selected primarily from LSC individuals with the CBS SNP (rs2850146) genotypes (G/G, G/C and C/C) that were matched on the basis of sex, ethnicity, current smoking status and freezer storage time. Matching ratios were approximately 3:3:1, but were modified if fewer heterozygotes were available. Due to the relatively low MAF of rs2850146 and the lower percentage of males in the LSC, one male sample with homozygous variant alleles for CBS rs2850146 was also identified from the New Mexico Lung Cancer Study (29) and included for metabolite analysis. Removing this sample did not change the assessment of the association between rs2850146 and plasma metabolites, thus this sample was included in the final data analysis (n = 86). Plasma samples were analyzed at the University of Colorado Hematology Division laboratories for a B12, folate and methionine metabolite panel that measured the following nine metabolites: methionine, tHcy, cystathionine, alpha-aminobutyrate, total cysteine, dimethylglycine, methylglycine, glycine and serine. Concentrations of tHcy are similar between fasting and non-fasting plasma samples and there is minimal between-day and within-day variations (5–10% for both) for individuals (30). Plasma metabolites were analyzed by capillary stable isotope dilution gas chromatography/mass spectrometry as described previously (31,32). Laboratory staff was blinded to sample identifications. Inter- and intra-assay replicates were included as controls with the standard deviations for nine metabolites in seven control samples ranging from 1.8% to 6.5% of the average. Plasma serine concentrations were deemed artificially low, possibly due to the use of Histopaque in blood processing, and not analyzed further. Samples from participants with elevated tHcy (>13 uM) and cystathionine (>342nM) values were assayed for methylmalonic acid and, if an elevated methylmalonic acid value (>271nM) was detected, these samples were suspect for vitamin B12 deficiency and were removed from the analysis.

An additional set of plasma samples (n = 63) was also selected from LSC cohort based on the availability of both plasma metabolites and gene methylation data and was combined with the 86 samples mentioned previously for the analysis of the association between plasma metabolites and gene methylation (n = 149).

Gene expression in primary HBECs

Primary HBEC cultures were established from airway biopsies obtained at bronchoscopy from current or former smokers (n = 79). HBECs obtained from 10 heterozygotes and 35 common homozygotes for CBS rs2850146 from male smokers were used to assess the association between rs2850146 and CBS gene expression. Female samples were excluded from the analysis, as were samples without sex designation. No variant homozygotes were identified in the 79 primary HBECs. TaqMan real-time PCR was conducted to measure the CBS gene expression in cDNA using the delta cycle threshold method with β-actin as the endogenous control (33).

Statistical analysis

Demographic and methylation variables were summarized by high and low methylation status. Proportions were used for categorical variables and medians with the (IQR) for continuous variables. Differences in clinical covariates between participants based on methylation status were assessed with Fisher’s exact test and the Wilcoxon two-sample test for categorical and continuous variables, respectively.

Logistic regression was used to estimate odds ratios and 95% confidence intervals for the association with high and low methylation status and SNPs using an additive inheritance model, with adjustment for age, sex and smoking history. If the MAF was ≤0.10, a dominant model that combined heterozygotes and variant homozygotes was used. Interactions between genotype and sex were assessed through the addition of interaction terms and by stratified analysis. The significance of individual SNP association was assessed using a Bonferroni corrected P value (0.05/79 SNPs, P < 0.0006) with the goal of identifying the most significant SNPs for functional validation.

The association of plasma metabolites with CBS rs2850146 genotype was analyzed using a mixed model, with the genotype as a one-degree of freedom ordered fixed effect and the matched groups as random effects, and fit by sex. Mean levels were obtained by genotype. The association between plasma metabolites and gene methylation was analyzed using a logistic regression model with adjustment for the design variables (age, ethnicity, current smoking status, rs2850146 genotype, sex and rs2850146–sex interaction). The difference in CBS gene expression between homozygotes and heterozygotes of rs2850146 in primary HBECs was examined using the Kriskal–Wallis test and summarized by median levels. Statistical analyses were conducted in SAS 9.2 and PLINK.

Results

Gender differences in prevalence of gene methylation

The prevalence for methylation of the 12-gene panel by sex is shown in Table 1. Men have a higher prevalence of gene methylation compared with women as previously reported (8,34,35). This sex-specific difference was taken into account when defining low and high methylation status. The number of methylated genes in the 12-gene panel was dichotomized into low methylation (0–2 genes methylated for women or 0–3 genes methylated for men) and high methylation (>2 genes methylated for women or >3 genes methylated for men). Characteristics of the LSC members (n = 907) by methylation status are described in Table 2. The gender-specific methylation cutoffs allowed us to remove the effects of sex on methylation and to analyze a similar distribution of high and low methylation among men and women as reflected in Table 2. The entire LSC is largely female and non-Hispanic white (8). The subset used for the association study was restricted to non-Hispanic whites with >75% female participants. The mean age of the low methylation group was significantly lower (55 years) compared with high methylation group (57 years), while the percentage of current smokers was significantly higher in the low methylation group (60%) compared with the high methylation group (50%).

Table I.

Prevalence of methylation in a 12 gene panel in sputum by sex

Number of genes methylated	Females (n = 684)		Males (n = 223)		P
	n	% (Cumulative %)	n	% (Cumulative %)
0	150	21.9 (21.9)	23	10.3 (10.3)
1	148	21.6 (43.6)	38	17.0 (27.3)
2	130	19.0 (62.6)	43	19.3 (46.6)
3	92	13.5 (76.0)	30	13.4 (60.1)
4	68	9.9 (86.0)	33	14.8 (74.9)
5	36	5.3 (91.2)	23	10.3 (85.2)
6	27	4.0 (95.2)	15	6.7 (91.9)
7	18	2.6 (97.8)	10	4.5 (96.4)
8	11	1.6 (99.4)	5	2.2 (98.6)
9	3	0.4 (99.8)	2	0.9 (99.6)
10	1	0.2 (100.0)	1	0.4 (100.0)
					<0.001

P = P value for difference in distribution between genders (X ² test).

Table II.

Selected characteristics of participants in the LSC by methylation status

Variable	High methylationa (n = 345)	Low methylation (n = 562)	P c
Age [median (IQR)]	56.9 (49.8, 65.4)	54.8 (48.3, 63.8)	0.008
Sex (% Female)	74.2	76.2	0.53
Smoking status (% Current)	50.1	60.0	0.005
Pack-years [median IQR (interquartile range)]	37.0 (27.0, 52.0)	35.0 (27.5, 49.5)	0.43
BMIb (%)
<25	30.6	34.1
25–29.9	37.1	38.0
30.0+	32.3	27.9	0.38

^aHigh methylation (>2 genes methylated for women or >3 genes methylated for men) and low methylation (0–2 genes methylated for women or 0–3 genes methylated for men).

^bSamples available with BMI data (high methylation, n = 294; low methylation n = 484).

^cDifferences between cases and controls were assessed by the Fishers exact test for categorical variables and the Wilcoxon two-sample test for continuous variables.

Association of SNPs with gene methylation

The association of 79 SNPs from six one-carbon metabolism genes with gene methylation was initially assessed with adjustment for age, sex, current smoking status and pack years. One SNP, rs3776467 (7068A > G), in the MTRR gene was associated with gene methylation in the LSC (P = 0.004) (Table 3). We further examined the SNP–sex interaction for the association with gene methylation due to the sex-specific difference in gene methylation prevalence observed in the LSC and conducted analyses stratified by sex (Supplementary Table 1). A highly significant interaction between SNP and sex was seen for CBS SNP rs2850146 (P for interaction < 0.0001, Table 3). Stratified analysis by sex found that CBS rs2850146 was associated with a 3.9-fold increased risk for gene methylation in males, but with a decreased risk for gene methylation in females. The association between MTRR rs3776467 and the risk for gene methylation was predominantly found in females but not in males (Table 3).

Table III.

Sex-specific association for CBS (rs2850146) and MTRR (rs3776467) with gene methylation in the LSC

SNP	Genotypesa	n (genotypes)	OR (95% CI)	P
CBS rs2850146b (−8283G > C )c	GG/GC/CC
Males		186/35/2	3.92 (1.83, 8.42)	0.0004
Females		580/101/3	0.63 (0.40, 0.999)	0.049
All		766/136/5	1.03 (0.71, 1.50)	0.86
MTRR rs3776467d (7068A > G)c	AA/AG/GG
Males		121/90/10	0.92 (0.57, 1.47)	0.72
Females		394/225/35	0.64 (0.49, 0.85)	0.002
All		515/345/45	0.71 (0.56, 0.89)	0.004

Adjusted for age, pack years of smoking, current smoking status and sex (in ‘All’ analysis).

^aReported as measured on Top Illumina Strand.

^bAssociation based on dominant model.

^cPositions are relative to the start codon; ‘-’ indicates position 5′ of start codon.

^dAssociation based on additive model.

In order to examine whether the dietary intake of folate, alcohol, green leafy vegetables and multivitamins, that affected the risk for gene methylation in a previous study (8), confounded the sex-specific association between the top two SNPs and the risk for gene methylation, we further restricted the analysis to those members who had completed a food frequency questionnaire and had values for all of these dietary factors. We compared the associations across four models that adjusted for the basic four demographic variables (age, sex, current smoking status and pack years) plus different sets of dietary variables (Supplementary Table 2). There was little change for the sex-specific association for CBS rs2850146 and MTRR rs3776467 with risk for gene methylation across the different models.

A secondary stratification analysis by dietary folate status (high (≥1000 mcg) versus low (<1000 mcg) did not identify any additional SNPs with significant associations with risk for gene methylation (data not shown). No analysis was conducted to stratify the population by both sex and dietary folate intake status due to the significantly reduced sample size in subgroup analysis.

Association of CBS rs2850146 with plasma metabolites

Cystathionine beta synthase catalyzes the conversion of Hcy to cystathionine, thus diverting Hcy from conversion to methionine and hence to SAM (36). Therefore, we examined the relationship of the CBS rs2850146 genotype with tHcy and cystathionine concentrations in plasma samples from LSC members by sex. The concentrations of tHcy and the ratio of tHcy/cystathionine increased with the number of variant alleles (C) of rs2850146 in a dose-response manner in males, but not in females (Table 4). Because methionine intake can affect tHcy concentrations and the rate of transsulfuration (7), the association between rs2850146 and plasma metabolites was also analyzed in a model that adjusted for concentrations of methionine. Methionine did not change the association for tHcy and tHcy/cystathionine ratio by rs2850146 genotype.

Table IV.

The association of CBS -8283G > C (rs2850146) variants with tHcy and plasma cystathionine

Sex	Metabolite	Genotypea	n	Mean	Std dev	Min	Max	P b	P c
Female	tHcy (μM)	GG	27	6.56	2.15	3.00	12.80
		GC	30	6.26	1.72	2.60	11.80
		CC	10	7.37	1.88	4.80	11.00	0.49	0.50
	Cystathionine (μM)	GG	27	0.17	0.07	0.08	0.32
		GC	30	0.18	0.08	0.07	0.43
		CC	10	0.23	0.12	0.09	0.54	0.28
	tHcy/Cystathionine	GG	27	42.66	18.94	16.61	108.79
		GC	30	40.47	16.59	15.70	89.62
		CC	10	36.71	12.05	12.80	57.30	0.50	0.53
Male	tHcy (μM)	GG	10	6.25	1.24	4.50	8.80
		GC	6	8.03	1.56	6.10	9.90
		CC	3	11.90	3.32	8.40	15.00	0.0004	0.0008
	Cystathionine (μM)	GG	10	0.17	0.09	0.07	0.38
		GC	6	0.15	0.05	0.09	0.23
		CC	3	0.16	0.06	0.11	0.22	0.59	0.27
	tHcy/Cystathionine	GG	10	42.82	16.51	18.73	70.37
		GC	6	60.47	25.78	31.14	102.22
		CC	3	89.43	51.92	37.67	141.51	0.015	0.0383

^aMinor allele (C) of rs2850146 reported as measured on the Top Illumina Strand.

P ^b is the block design matching on age, ethnicity, smoking status and freezer storage time.

P ^c is the block design matching on age, ethnicity, smoking status, freezer storage time and adjustment for plasma methionine levels.

Association of rs2850146 with CBS expression in HBECs

CBS rs2850146 is located approximately 4kb upstream of the CBS transcriptional start site. This SNP is in high linkage disequilibrium (LD) (>0.80) with only one other SNP (rs28360502, r ² = 0.826 in 1000 Genome Pilot1 CEU population) within 500kb upstream and downstream of rs2850146 (see Figure 1). The functional potential for these two SNPs was assessed by searching the FuncPred: Functional SNP Prediction module of the SNPinfo Web Server (37). Rs2850146 was predicted to affect the binding sites of several transcription factors including androgen receptor (data not shown). Because our results showed that rs2850146 genotype was associated with reduced conversion of tHcy to cystathionine in males, we tested whether the variant allele of rs2850146 reduces CBS gene expression. Only HBEC lines from male patients were analyzed due to the limited number of lines collected from female patients (n = 1 and 7 for heterozygous and homozygous, respectively). The CBS expression in 10 heterozygotes was compared with that observed in 35 common homozygotes. Expression of CBS in HBEC cultures heterozygous for rs2850146 was reduced on average by 45%, although this difference did not reach statistical significance (P = 0.13) in this relatively small sample set (see Figure 2).

Fig. 1.

Plot of SNPs in linkage disequilibrium with rs2850146 (large diamond). (A) Only one other SNP is in LD (R ² > 0.8) with rs2850146 within 500kb upstream and downstream. (B) Plot of rs2850146 and SNPs within 10kb indicates lack of known SNPs in CBS promoter region. Solid line indicates recombination rates.

Fig. 2.

CBS SNP rs2850146 (-8283G > C) correlates with CBS gene expression. Box plots show the range of expression in HBECs with common homozygous alleles of rs2850146 compared with HBECs with heterozygous alleles. Solid line indicates median expression; diamond indicates mean expression.

Association of plasma metabolites and risk for gene methylation

Because CBS rs2850146 was associated with both tHcy concentration and risk for gene methylation, we hypothesized that tHcy may be associated with risk for gene methylation in the LSC members. A total of 149 members with available metabolite panel and gene methylation data were used to analyze the association between plasma tHcy and risk for gene methylation. tHcy and methionine concentrations in plasma were associated with an increased risk for gene methylation. A one unit (1 μM) increase of tHcy and methionine concentrations in plasma was associated with ORs of 1.20 (95% CI = 1.02, 1.41) and 1.06 (1.01, 1.13) for gene methylation for tHcy and methionine, respectively. Interestingly, a significant association between plasma dimethylglycine and risk for gene methylation was also identified with one unit (1 μM) increase of dimethylglycine associated with 50% increased risk (OR = 1.50, 95% CI = 1.13−2.26, P = 0.006).

Discussion

This is the first study to comprehensively evaluate the relationship between common variants in six candidate genes involved in one-carbon metabolism and transsulfuration pathways and risk for gene methylation in a large population of current and former smokers. Our findings support a sex-specific role for genetic variants in CBS and MTRR as predictors for the acquisition of gene promoter methylation in exfoliated cells in smokers’ sputum. The variant allele of rs2850146 in the CBS gene correlates with higher tHcy in males, suggesting a potential functional effect of this SNP or another unknown SNP in high LD on the circulating concentrations of plasma metabolites necessary for DNA methylation.

Hcy is cleared through remethylation and transsulfuration in mammalian cells (36). CBS regulates the first and irreversible step of the transsulfuration pathway in which Hcy is converted to cystathionine (6). Transsulfuration occurs predominately in the liver and kidney, but CBS expression is also found in other tissues including normal lung (38,39). Importantly, in our study, concentrations of tHcy and the ratio of tHcy to cystathionine were associated with the rs2851046 genotype highlighting a CBS-metabolite-methylation model and substantiating a role for CBS in aberrant DNA methylation. In this model, a functional CBS SNP impairs CBS gene function leading to an increase in concentration of tHcy that further influences aberrant DNA methylation patterns. Several lines of evidence support this model. Mice deficient in CBS have an altered Hcy remethylation pathway with corresponding decreases in betaine-Hcy methyltransferase activity and increases in 5,10-methylenetetrahydrofolate reductase activity (40). The elevation of plasma Hcy level by a dietary and/or genetic approach in CBS deficient mice was significantly associated with global DNA hypomethylation in a tissue-specific manner (41,42). Moreover, high plasma Hcy levels were associated with promoter hypermethylation of essential genes in the pathogenesis of other disease including postmenopausal osteoporosis, atherosclerosis, and breast and colorectal cancers (43–47).

A clear mechanism explaining how Hcy might influence epigenetic reprogramming in cells remains elusive. Several studies assessing the association between Hcy and global DNA hypomethylation proposed that an increase in Hcy level reverses the S-adenosylhomocysteine (SAH) hydrolase reaction, increases SAH levels and reduces cellular methylation capacity via inhibition of DNA methyltransferases (36,41,42). Although this hypothesis provides a reasonable explanation for the global hypomethylation due to the inhibition of the maintenance methylation during DNA replication, it does not explain de novo promoter hypermethylation in tumor suppressor genes upon DNA damage.

Studies from our group and others have shown that extensive DNA damage could be responsible for acquisition of promoter methylation of tumor suppressor genes during carcinogenesis (48–50). Thus, the increased propensity for acquiring gene promoter hypermethylation associated with elevated plasma Hcy levels in smokers is more likely to result from a complex combination of events in lung cells, including tissue-specific SAM/SAH ratios, DNA damage in gene promoters, impaired fidelity in DNA replication and repair and the disruption of normal epigenetic regulatory machinery upon DNA damage. Although such a scenario is clearly speculative, an in vitro malignant transformation model or in vivo tumorigenesis model using CBS deficient human epithelial cells or animals could be used to eventually delineate the biological mechanism.

The sex-specific association of risk for gene methylation and tHcy with CBS rs2850146 is not entirely unexpected because the expression of CBS is regulated by male hormone in mice and humans (51,52). CBS expression and activity were reduced in castrated mice, but were not affected in ovariectomized mice. Moreover, renal CBS activity was slightly higher in females compared with males (52). Testosterone can also reduce CBS expression in the human LNCaP prostate adenocarcinoma cell line (51). In functional prediction models, SNP rs2850146 and its surrounding sequence are predicted to lie within transcription factor binding sites including the androgen receptor, suggesting a potential role of this SNP in disrupting the androgen regulation of CBS expression. Genetic variants in the CBS promoter have not been well characterized even in the 1000 Genome Project. Thus, we cannot exclude the possibility that the SNP rs2850146 may be in LD with other undefined SNPs with functional relevance for androgen regulation.

A limitation of this study is that the association of plasma metabolites and methylation is measured in a cross-sectional analysis and does not provide evidence for a causal relationship of plasma metabolites concentrations on DNA methylation. Longitudinal studies using plasma samples collected from multiple visits with corresponding DNA methylation profiles could help substantiate a causal relationship of metabolite concentrations on risk for methylation. The MAF of the CBS SNP rs2850146 varies across populations according to the DbSNP database (Europeans, MAF = 0.06; African Americans, MAF = 0.12 and Chinese, MAF = 0.01). Thus, more studies are needed to examine the association of this SNP with methylation in other high-risk smoking populations.

In this study, three plasma metabolites (tHcy, methionine and dimethylglycine) in the one-carbon metabolism pathway are associated with increased risk for gene methylation in exfoliated cells from the lungs of smokers. These data implicate metabolites as risk markers for methylation in the pulmonary epithelium of high risk smokers. The association of SNPs with serum metabolites and risk for methylation found in this study emphasizes the intricate mechanisms that potentially regulate methylation via one-carbon metabolism. A further understanding of the complex mechanisms regulating metabolic pathways to prevent aberrant DNA methylation could ultimately reduce cancer progression.