DNA methylation is differentially associated with environmental cadmium exposure based on sex and smoking status

Abstract

The adverse health effects of cadmium (Cd) are well known in human populations; however, much of what is known about biological mechanisms of Cd comes from in vitro and animal studies. The adverse health outcomes due to high levels of Cd exposure in the population of Mae Sot, Thailand have been extensively characterized. Here, for the first time, this population is being studied in an epigenetic context. The objective of this study was to characterize the association between DNA methylation markers and Cd exposure, taking into account sex and smoking differences, in an adult population at an increased risk of experiencing adverse health outcomes from high body burden of Cd. One hundred and sixty-nine residents from known exposure areas of Mae Sot, Thailand and one hundred residents from non-exposed areas nearby were surveyed in 2012. Urine and blood samples were collected for measurement of urinary Cd (UCd) and DNA methylation of Cd-related markers (DNMT3B, MGMT, LINE-1, MT2A). UCd levels were 7 times higher in the exposed compared to the unexposed populations (exposed median: 7.4 μg/L, unexposed median: 1.0 μg/L, p <0.001). MGMT hypomethylation was associated with increasing levels of UCd in the total population. Sex-specific associations included MT2A and DNMT3B hypomethylation in women and LINE-1 hypermethylation in men with increasing UCd. Upon subanalysis, these associations separated by smoking status. In summary, environmental Cd exposure is associated with gene-specific DNA methylation in a sex and smoking dependent manner.

1. Introduction

Cadmium (Cd) ¹ is a biologically toxic transition metal found in low levels in nature and high levels from anthropogenic sources (Agency for Toxic Substances and Disease Registry (ATSDR), 2008; World Health Organization (WHO), 2010). Non-occupational exposures occur from tobacco smoke and ingestion of contaminated food or water and are associated with kidney damage, osteoporosis, osteomalacia, cardiovascular diseases, and multiple cancers (ATSDR, 2008; WHO, 2010; International Agency for Research on Cancer (IARC), 2012). Cd sequesters in the liver and kidney and replaces calcium in bone (Limpatanachote et al., 2010; Liang et al., 2012). Many studies have provided evidence linking Cd to the induction of cancer (Benbrahim-Tallaa et al., 2007; Benbrahim-Tallaa et al., 2009; Gallagher et al., 2010) and the IARC has classified Cd as a Group 1 human carcinogen (IARC, 2012).

Mae Sot District is located on Thailand’s northwest border and is an area of high Cd contamination due to a long-standing zinc mine. Several subdistricts have Cd levels in soil and rice exceeding Thai standards of 0.15mg/kg of soil and 0.043mg/kg of rice (Zarcinas et al., 2004). These standards reflect levels at which human exposure may result in adverse health effects (Pongsakul P, 1999). Rice is a dietary staple and farmed in contaminated soil; therefore, exposure to Cd in this population is via ingestion of contaminated rice. A study in Mae Sot in 2004 found a subset of residents with urinary Cd (UCd) levels ≥5μg/g creatinine to also have irreversible renal dysfunction, indicated by clinical parameters of renal marker levels (Swaddiwudhipong et al., 2007; Limpatanachote et al., 2010). UCd levels above this level induce preclinical renal dysfunction; therefore, 5μg/g creatinine has been deemed as a human exposure standard by the WHO (WHO JEFCA, 2010).

Although the association between UCd and adverse health outcomes is well-described (Aoshima, 1987; Oh and Lim, 2006); the mechanism by which Cd exerts its biologic effects is unclear. Cd weakly binds to DNA, is indirectly genotoxic and poorly mutagenic, indicating that it does not initiate disease through direct interaction with DNA(Hartwig, 1994; Waalkes, 2000). A number of genes are transcriptionally induced by Cd implicating exposure may initiate disease through epigenetic (Huang et al., 2008; Jiang et al., 2008; Benbrahim-Tallaa et al., 2009). Cd has been shown to affect DNA methylation, yet the direction of effect is unclear. Candidate gene studies have identified DNA methylation markers associated with Cd-induced kidney damage. Hypermethylation of tumor suppressor RASAL1 and renal fibrosis inhibitor KLOTHO was strongly associated with blood and UCd levels as well as renal fibrogenesis(Sun et al., 2012; Zhang et al., 2013). Many studies have implicated epigenetic changes in the process of Cd-mediated tumorigenesis(Huang et al., 2001; Benbrahim-Tallaa et al., 2007). Increased DNA methylation and DNMT3b enzyme activity were reported in human lung fibroblasts after two months of Cd exposure(Jiang et al., 2008). Cd-induced malignant transformation of human prostate and breast cell lines has been observed in vitro(Benbrahim-Tallaa et al., 2007; Benbrahim-Tallaa et al., 2009). Cd-induced gene-specific DNA hypermethylation has been reported for genes involved in cell cycle regulation, DNA repair, apoptosis and proliferation(Wang et al., 2012; Zhou et al., 2012). Hypermethylation can prevent transcription of these genes and promote initiation of disease. These studies show that Cd induces epigenetic alterations, which may fall along the molecular pathway of Cd exposure and adverse health outcomes.

Much of what we know about Cd-induced epigenetic changes comes from animal models or in vitro studies. Only a few key studies regarding methylation have been in human populations. Hossein et al showed that UCd was associated with LINE-1 hypomethylation and decreased DNMT3b expression among women(Hossain et al., 2012). Sanders et al showed patterns of Cd-associated methylation that were distinct from cotinine-associated methylation in newborns and mothers(Sanders et al., 2014). Additionally, sex differences in body burden, health outcomes and epigenetic alterations have been reported in response to Cd exposure. Here, we measured methylation in MT2A (Metallothionein 2A), DNMT3B (DNA methyltransferase 3B) and MGMT (O-6-methylguanine DNA methyltransferase) genes, which have been found to be altered in the presence of Cd in animal models and in vitro studies (Benbrahim-Tallaa et al., 2007; Arita and Costa, 2009; Hossain et al., 2012; Zhou et al., 2012). Methylation of LINE-1, a retrotransposon, was measured as a proxy for genomic stability(Kippler et al., 2013; Sanders et al., 2014). Because body burden of Cd increases with age and high body burden is known to be associated with adverse health outcomes(Zhang et al., 2013), the objective of this study was to determine the sex- and smoking-specific associations between UCd and methylation of selected genes in an adult population to clarify the role of gene methylation in Cd exposure.

2. Materials & Methods

2.1 Study Population

The study area and subjects have been previously described and extensively characterized (Nambunmee et al., 2010; Ruangyuttikarn et al., 2013). Briefly, the study participants consisted of 169 residents living in Mae Ku, Phra Thad Padang, and Mae Tao, known polluted areas of Mae Sot, and 100 participants living the Mae Kasa subdistrict, a non-contaminated area of Mae Sot (Simmons et al., 2005). These areas are all rural, with similar socioeconomic environments. Subjects from polluted areas aged ≥ 40 years and with urinary Cd levels ≥5 μg/g creatinine in the 2007 survey were included in this study (Nambunmee et al., 2010). Exposed residents were 5-year age-matched to residents from the Mae Kasa subdistrict. Each participant was consented and interviewed by trained nurses. The research ethical committee of the Faculty of Medicine, Chiang Mai University approved this study. (Approval No. 004/2012)

2.2 Sample Collection and Processing

Twenty-five mL of morning urine was collected from each participant in Cd-free polyethylene containers and stored at −20°C. Fasting venous whole blood with EDTA and urine samples were frozen and transported on dry ice to the University of Michigan for DNA extraction and Cd measurements.

2.3 Urinary Cadmium Measurement

A refractometer determined specific gravity of each sample (PAL-10S, Atago Inc.). UCd was measured at the Michigan Department of Community Health as part of the U.S. Centers for Disease Control (CDC) Laboratory Response Network Program. Briefly, urine samples were diluted 1:10 with a diluent composed of 2.0% nitric acid, internal standards and 0.05% Triton X; and Cd concentrations were determined using ICP-MS (DRCII, PerkinElmer, USA). The analytical accuracy using QMEQAS08U urinary standard reference material (Institut national de santé publique du Québec, INSPQ) was 101.1% (n=8). All samples were above the analytical detection limit of 0.15 μg/L.

2.4 DNA Preparation

DNA was extracted from 300μL of whole blood using the QiaAMP DNA Mini Kit (Qiagen, Valencia, CA, USA). DNA concentration was quantified using the Nanodrop (ThermoScientific, Wayne, MI, USA), and stored at −20°C. Sodium bisulfite modification was performed on 500ng of genomic DNA using the EpiTect Bisulfite Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s protocol.

2.5 DNA Methylation

Methylation assays for MT2A, DNMT3B and MGMT, were designed using PyroMark Assay Design 2.0. LINE-1 methylation was measured using a previously published assay (Yang et al., 2004). Bisulfite singleplex PCR amplification was performed using FastStart Taq Polymerase (Roche Diagnostics, Indiana, USA) with primer concentrations of 0.2 mM and 10ng/μL of bisulfite-converted DNA. Fifteen microliters of PCR product was combined with sequencing primer and methylation analysis was conducted using the Pyromark^™ MD System (Biotage) according to manufacturer’s protocol (PyroGold reagents). Four bisulfite and four pyrosequencing controls were generated by mixing unmethylated and methylated control DNA (genomic: EpigenDX; bisulfite-converted: Epitect) to obtain controls with 0%, 30%, 60% and 100% methylation. Each sample plate was run with all controls. If methylation values of controls were incorrect, all samples on plate were re-run. Measurement of all samples for every methylation marker was not possible due to insufficient quantity of extracted DNA.

2.6 Statistical Analysis

Correlations between creatinine (Cr)-adjusted, specific gravity (SG)-adjusted and unadjusted values were calculated using the Spearman method. UCd levels were adjusted by specific gravity to account for all dilution-related variation (Suwazono et al., 2005). Urinary markers were standardized to the median specific gravity of the control population with the following formula: C_c = C[(1.017 – 1)/SG – 1)], where C_c is the SG-corrected UCd, C is the observed Cd level, 1.017 is the median SG of the unexposed population, and SG is the specific gravity of the individual’s urine sample (Meeker et al., 2009). Chi-squared and Mann-Whitney tests were used to compare exposed and unexposed populations. Statistical significance was defined when p<0.05.

Linear regression models were used to determine the associations between UCd and each methylation marker. Methylation changes were standardized to an interquartile range (IQR) increase in UCd levels. Methylation units were in 1% increments. Due to departures from normality, UCd and all methylation markers were log-transformed for regression analysis. Models were interpreted as percent increases in predictor associated with percent changes in outcome. Age, smoking status and occupation were associated with UCd levels and included as covariates. Because methylation was measured in white blood cell DNA, all models were re-run adjusting for white blood cell count. There were no differences in results, and white blood cell count was not included in final models as it was not associated with UCd and methylation markers.

Women have higher body burden of Cd regardless of exposure; therefore, models were stratified by sex. All models were repeated within nonsmokers and smokers separately. Smokers included both former and current smokers.

R-statistical software was used for all analyses (R version 3.0.1).

3. Results

3.1 High correlation among UCd measurements with different correction methods

Correlation coefficients for unadjusted UCd and SG-adjusted UCd and for SG-adjusted and Cr-adjusted UCd were similar (rho=0.95, p<0.0001). The correlation coefficient for unadjusted UCd and Cr-adjusted UCd was slightly lower (rho=0.87, p<0.0001)

3.2 Differences by exposure group and sex

Median levels of UCd were seven times higher in the exposed population than the unexposed population (unexposed [median (range)]: 1.0 μg/L (0.65, 1.60); exposed: 7.4 μg/L (4.4, 10.2); p <0.001). MGMT was the only methylation marker that differed by exposure group. Methylation of MGMT was significantly lower in the Cd exposed population as compared to the unexposed population (p<0.001) (Table 1). The proportion of regular smokers was highest in the Cd exposed population, of which 66% were male; whereas the unexposed population had similar proportions of never, former and regular smokers (Table 2). The proportion of women was higher in the nonsmoking (83%) compared to the smoking group (31%). The majority of the unexposed population was farmers (74%) compared to only fifty percent in the Cd exposed population.

Table 1.

Descriptive Statistics of Study Population

	Unexposed (n=100)	Exposed (n=169)
	Mean (SD)	Mean (SD)	p-value*
Age (years)	61.0 (11.5)	61.2 (11.8)	0.93
	Median (IQR)	Median (IQR)	p-value*
UCd (μg/L)	1.0 (0.65, 1.60)	7.4 (4.4, 10.2)	<0.001
LINE-1 (%)	80.5 (79.0, 81.1)	80.2 (79.4, 81.1)	0.78
MT2A (%)	23.2 (20.9, 25.8)	23.1 (20.8, 25.0)	0.38
DNMT3B (%)	1.4 (1.1, 1.9)	1.4 (1.2, 1.9)	0.78
MGMT (%)	2.3 (1.8, 3.1)	1.7 (1.4, 2.1)	<0.001

^*p-values from Mann-Whitney test;

Abbreviations: UCd (urinary cadmium); n (number of subjects); SD (standard deviation); IQR (Interquartile Range).

Table 2.

Demographics of Study Population

	All (n=269)			Male (n=133)			Female (n=136)
	Unexposed N (%)	Exposed N (%)	p*	Unexposed N (%)	Exposed N (%)	p*	Unexposed N (%)	Exposed N (%)	p*
Smoking
Never	36 (36)	66 (39.1)	0.03	6 (12)	11 (13.3)	0.03	30 (60)	55 (64)	0.64
Former	33 (33)	32 (18.9)		21 (42)	17 (20.5)		12 (24)	15 (17.4)
Regular	31 (31)	71 (42.0)		23 (46)	55 (66.3)		8 (16)	16 (18.6)
Drinker
Never	50 (50)	82 (48.5)	0.23	6 (12)	19 (22.9)	0.05	44 (88)	63 (73.3)	0.13
Former	26 (26)	33 (19.5)		23 (46)	22 (26.5)		3 (6)	11 (12.8)
Regular	24 (24)	54 (32.0)		21 (42)	42 (50.6)		3 (6)	12 (14)
Occupation
Agriculture	74 (74)	85 (50.3)	<0.001	36 (72)	51 (61.4)	0.29	38 (76)	34 (39.5)	<0.001
Other	26 (26)	84 (49.7)		14 (28)	32 (38.6)		12 (24)	52 (60.5)

^*Chi-squared test; N=number of subjects

We compared men and women within each exposure group to determine sex-specific differences. Within the unexposed group, women had significantly higher levels of UCd and DNMT3B methylation compared to men. However, within the Cd exposed group, women had significantly lower levels of UCd andLINE-1 methylation compared to men (Table 3).

Table 3.

Descriptive Statistics of Study Population by Sex

	Unexposed (n=100)			Exposed (n=169)
	Male (n=50)	Female (n=50)	p-valuea	Male (n=83)	Female (n=86)	p-valueb	p-valuec (M)	p-valued (F)
	Mean (SD)	Mean (SD)		Mean (SD)	Mean (SD)
Age (years)	62.5 (12.1)	59.6 (10.9)	0.19	62.7 (11.7)	59.7 (11.8)	0.09	0.91	0.99
	Median (IQR)	Median (IQR)		Median (IQR)	Median (IQR)
UCd (μg/L)	0.9 (0.5, 1.2)	1.4 (0.9, 2.0)	<0.001	8.0 (5.2, 10.5)	6.4 (3.6, 9.5)	0.04	<0.001	<0.001
LINE-1 (%)	80.6 (79.0, 81.1)	80.4 (79.0, 81.2)	0.80	80.4 (79.7, 81.4)	79.7 (78.9, 80.7)	0.001	0.33	0.20
MT2A (%)	22.8 (20.7, 24.9)	23.7 (21.5, 26.2)	0.13	22.7 (20.7, 24.4)	23.5 (21.0, 25.5)	0.33	0.96	0.23
DNMT3B (%)	1.3 (0.9, 1.8)	1.7 (1.3, 2.1)	0.008	1.5 (1.2, 1.9)	1.4 (1.2, 1.9)	0.66	0.03	0.07
MGMT (%)	2.4 (2.0, 3.4)	2.2 (1.7, 2.7)	0.22	1.6 (1.3, 2.2)	1.7 (1.4, 2.0)	0.91	0.0002	0.003

Abbreviations: UCd (urinary cadmium); n (number of subjects); SD (standard deviation); IQR (Interquartile Range); M (males); F (females). p-values are all from Mann-Whitney tests for comparisons between males and females within the ^aunexposed group, ^bexposed group, and between exposure groups within ^cmales and ^dfemales

We then compared exposure groups to determine exposure-specific differences by sex. These comparisons revealed exposed men and women had significantly higher levels of UCd and lower MGMT methylation levels compared to their unexposed counterparts. Additionally, exposed men had significantly higher DNMT3B methylation than unexposed men.

3. 3 Urinary cadmium associated with methylation markers in sex-specific manner

Regression analyses adjusted for age, smoking status, and occupation showed distinct associations of methylation markers with UCd (Figure 1). In the total population, MGMT methylation decreased by 16.6% with an IQR increase in UCd levels (p<0.001). Among men, MGMT methylation decreased by 24.7% and LINE-1 methylation increased by 1% with an IQR increase in UCd (p =0.003; p =0.04). Among women, DNMT3B methylation decreased by 8.5% and MT2A methylation decreased by 7.4% with an IQR increase in UCd (p =0.04; p =0.03). Methylation markers were not correlated with each other.

Figure 1. Regression analysis.

Total population (magenta); Nonsmokers only (light blue, dashed); Smokers only (dark blue, long dash); Lines represent standardized beta estimate and 95% confidence interval of percent change in methylation (y-axis) for every unit increase in IQR of UCd. Smokers included both current and former smokers.

3.4 Methylation markers are differentially associated with UCd based upon smoking status

UCd levels between smokers (n=167) (median (95%CI): 4.4μg/L (1.3, 9.0)) and nonsmokers (n=102) (median (95%CI): 3.4μg/L (1.5, 7.6)) were not statistically different. Within each group, those who were exposed had significantly higher UCd levels and lower MGMT methylation than their unexposed counterparts (Supplementary Table 1S).

Regression models revealed significant hypomethylation of MGMT for all non-smokers and non-smoking men and women separately (Figure 1, Supplementary Table 2S). This was similar to the results observed in the total population, although within nonsmokers, hypomethylation of MGMT in women reached statistical significance. Specifically, an IQR increase in UCd was associated with a 25.6% decrease in MGMT methylation among all nonsmokers, whereas MGMT decreased by 31.9% and 7.2% in nonsmoking men and nonsmoking women, respectively. Associations of other methylation markers that were seen in the total population were not found to be significant within nonsmokers.

In contrast, regression analysis within smokers showed separate associations by sex (Figure 1). An IQR increase in UCd was associated with a 0.9% decrease in LINE-1 methylation in men only; whereas in women, this was associated with an 18.1% decrease in both MT2A and DNMT3B methylation (Figure 1).

4. Discussion

The toxic effects of environmental Cd exposure were studied with the outbreak of Itai-Itai disease in Japan in response to extreme Cd exposure from ingestion of contaminated rice. Cd has since been shown to have a variety of adverse effects on human health including proximal tubular renal dysfunction, bone disorders, and cancer (Simmons et al., 2005; Limpatanachote et al., 2010; Satarug et al., 2010; Swaddiwudhipong et al., 2010). However, the pathway by which Cd induces these health outcomes is unknown.

Our study shows the expected differences in UCd between exposure groups. The exposed group had significantly higher levels of UCd compared to the unexposed group. Although the unexposed population had detectable amounts of UCd, body burden of Cd increases with age and these levels are consistent with unexposed populations above age 40 from similar geographic areas. Residents over the age of 50 living in non-polluted areas in Japan had urinary Cd levels of 1.6 μg/L in men and 1.8 μg/L in women (Suwazono et al., 2000). Residents with a median age of 50 living in non-polluted areas of China had urinary levels of 2.5 μg/g creatinine (Zhang et al., 2013). This is higher than our unexposed population considering the correlation between creatinine-adjusted and specific gravity-adjusted UCd measurements. Therefore, our unexposed group provides a good control to measure Cd exposure in demographically similar populations.

We report the expected sex-specific differences in UCd within exposure groups. Women, especially those with low iron stores, tend to have higher body burdens of Cd as compared to males, regardless of exposure levels (Vahter et al., 2002; Satarug et al., 2004). This is seen in our study as women living in the unexposed area had higher levels of UCd compared to men in the same area. However, within those living in high Cd-exposed areas, women had lower levels of UCd compared to men living in the same area. The large proportion of male smokers living in Cd-exposed areas may explain this observation. UCd measurements are representative of total body burden from all exposure sources, including both ingestion of contaminated rice and inhalation of tobacco smoke. Toxicologically, Cd absorption through lungs is much higher compared to that of the gut(IARC, 1993). Therefore, exposed men may have a larger body burden due to a combination of exposures.

We also report differences in methylation markers across exposure groups for each sex. Expectedly, exposed men and women had higher UCd levels than their unexposed counterparts; however, they had lower levels of MGMT methylation. MGMT hypomethylation was associated with an increase in UCd in the total population and men alone. Although MGMT methylation was not significant in women, the trend was similar. MGMT is involved in DNA repair from oxidative stress. Cd is hypothesized to induce toxic effects via induction of reactive oxidative species (Khojastehfar et al., 2015; Matovic et al., 2015). Reduced methylation of this gene promoter may reflect induction to repair damage from oxidative stress due to Cd exposure. As a biomarker, methylation of this gene may be used as an indicator of oxidative stress in Cd-exposed populations.

LINE-1 hypermethylation was associated with an increase in UCd in men. This corroborates previous literature showing men tend to have higher methylation of LINE-1, even in the presence of Cd exposure (Zhang et al., 2011). However, this association is opposite to that found in a previous study, although it’s likely due to the fact that the subjects in that study were all women with much lower levels of Cd exposure (Hossain et al., 2012). Our findings validate previous evidence of LINE-1 hypermethylation in newborn boys prenatally exposed to Cd (Kippler et al., 2013), and extend this evidence to an adult male population.

MT2A hypomethylation was associated with increasing UCd in women only. This finding is new and has not been reported previously in the literature. MT2A encodes metallothionein, a protein that binds free Cd and sequesters it in the liver and kidney. Women tend to have higher body burden of Cd compared to men due to higher gut absorption rates(Nishijo et al., 2004) and, this may point to a potential mechanism of increased MT2A expression by which women may sequester larger amounts for longer periods of time. As a biomarker, MT2A methylation may be useful in determining actual body burden of Cd in women.

Our regression analysis found that DNMT3B is inversely associated with UCd in women. DNMT3B encodes a DNA methyltransferase involved in de novo methylation. Previous studies evaluated the relationship between DNMT3b protein expression and Cd, finding expression to be inversely associated with Cd in women and positively associated with Cd in prostate and lung cell lines (Takiguchi et al., 2003; Benbrahim-Tallaa et al., 2007; Jiang et al., 2008; Hossain et al., 2012). Our finding of hypomethylation of DNMT3B with increasing levels of UCd in women is consistent with DNMT3b protein overexpression. However, we observed that both men and women with higher levels of UCd exhibited higher levels of methylation of DNMT3B, suggesting reduced expression. These contrasting findings are indicative that DNMT3B methylation may be a marker of a specific type of exposure to Cd. UCd is a measure of total body burden and can reflect multiple routes of Cd exposure, such as smoking and diet, which may have distinct biological effects. Recently, cotinine-associated methylation patterns have been shown to be distinct from Cd-associated patterns (Sanders et al., 2014). This is likely due to the large number of toxic compounds found in cigarette smoke that can alter methylation (ATSDR, 2008) as well as the mixture of metals likely found in Mae Sot. To address this, we conducted a subanalysis within smokers and nonsmokers to validate our findings.

Stratification by smoking status revealed differential associations between methylation markers and UCd. In nonsmokers, MGMT hypomethylation was associated with increasing levels of UCd. This association persisted in nonsmoking men and women separately, although there were only 17 nonsmoking men. However within smokers, LINE-1 hypermethylation was associated with increasing UCd in men only, while MT2A and DNMT3B hypomethylation was associated with increasing UCd in women only. All these associations were found in the total population and separate out by smoking status. Although samples sizes are small, especially by sex, these findings suggest that specific methylation markers may be important depending upon the source of Cd exposure. LINE-1, MT2A and DNMT3B seem to be important in tobacco-related Cd exposure, while MGMT may play a role in other routes of exposure, including dietary intake.

The findings presented here are critical in contributing to future epigenetic studies on Cd exposure in humans, which should be attentive to addressing the limitations of the present study. The subset analysis of smoking was useful in identifying markers that may be indicators of dietary versus tobacco exposures. However, stratification by sex within exposure and smoking groups led to small sample sizes, limiting generalizability of the results. Additionally, our exposed smokers were subjected to Cd via both tobacco and diet. Our intent in conducting these subanalyses was to provide preliminary evidence of differential methylation by smoking. Nevertheless, larger studies separating smokers and nonsmokers by sex and type of exposure in Cd-exposed populations are necessary to support the evidence presented here. Future studies may be able to achieve a large enough sample size and include cotinine measurements for tobacco exposure to separate the effects of multiple Cd sources.

Gene expression was not measured in this study; however, the methylation markers presented here are intended as molecular biomarkers of Cd exposure rather than indicators of mechanisms. Environmental exposures may not always be reflective of body burden since the exposure must be inhaled/ingested to contribute to internal dose. These methylation markers are associated with UCd, and can be used to assess dysfunction preceding a Cd-associated adverse health outcome.

It is well known that exposure to Cd likely includes co-exposures to other heavy metals commonly found with Cd, such as arsenic, copper and lead, are also associated with epigenetic changes (Cheng et al., 2012); however, these were not measured in this study. Our current study identified methylation markers relevant to Cd exposure and future studies measuring multiple metals for association with these markers in this population are currently underway.

1.5 Conclusions

Our study takes advantage of an extensively characterized population in Mae Sot, Thailand with a wide range of urinary Cd levels to validate DNA methylation markers of Cd exposure in a human population. DNMT3B, MGMT, LINE-1 and MT2A were chosen based on in vitro and in vivo evidence of their association with Cd. This is the first study to be able to look at such exposure ranges within the context of DNA methylation in a population-based study. Studies such as these are a logical first step to identifying the mechanisms of Cd toxicity and may identify useful biomarkers that indicate underlying health consequences of Cd exposure.