Effects of Short-term Exposure to Inhalable Particulate Matter on DNA Methylation of Tandem Repeats

Abstract

There is compelling evidence that particulate matter (PM) increases lung cancer risk by triggering systemic inflammation, and leukocyte DNA hypomethylation. However, previous investigations focused on repeated element sequences from LINE-1 and Alu families. Tandem repeats, which display a greater propensity to mutate, and are often hypomethylated in cancer patients, have never been investigated in individuals exposed to PM. We measured methylation of three tandem repeats (SATα, NBL2, D4Z4) by polymerase chain reaction–pyrosequencing on blood samples from truck drivers and office workers (60 per group) in Beijing, China. We used lightweight monitors to measure personal PM_2.5 (PM with aerodynamic diameter ≤2.5 µm) and elemental carbon (EC, a tracer of PM from vehicular traffic). Ambient PM₁₀ data were obtained from air quality measuring stations. Overall, an interquartile increase in personal PM_2.5 and ambient PM₁₀ levels was associated with a significant covariate-adjusted decrease in SATα methylation (−1.35% 5-methyl cytosine [5mC], P = 0.01; and −1.33%5mC; P = 0.01, respectively). Effects from personal PM_2.5 and ambient PM₁₀ on SATα methylation were stronger in truck drivers (−2.34%5mC, P = 0.02; −1.44%5mC, P = 0.06) than office workers (−0.95%5mC, P = 0.26; −1.25%5mC, P = 0.12, respectively). Ambient PM₁₀ was negatively correlated with NBL2 methylation in truck drivers (−1.38%5mC, P = 0.03) but not in office workers (1.04%5mC, P = 0.13). Our result suggests that PM exposure is associated with hypomethylation of selected tandem repeats. Measuring tandem-repeat hypomethylation in easy-to-obtain blood specimens might identify individuals with biological effects and potential cancer risk from PM exposure.

INTRODUCTION

Particulate Matter (PM), a common airborne pollutant in urban environments, contributes to as many as 3.2 million premature deaths worldwide each year [Lim et al. 2013]. Several epidemiologic investigations have demonstrated the association between PM and lung cancer [Turner et al. 2011; Krewski et al. 2005]. Particles with an aerodynamic diameter <2.5 micrometers (PM_2.5), are especially hazardous to cardiorespiratory health because they can penetrate into the airways and deposit in the respiratory bronchioles and alveoli [Kampa and Castanas 2008]. Each 10-µg/m³ elevation in chronic and acute PM_2.5 has been associated with a ~8% increased risk of lung cancer mortality and a 2.8% increase in PM-related mortality, respectively [Pope et al. 2002; Kloog et al. 2013].

Several mechanistic pathways have been proposed to account for the link between PM and lung cancer, including altered gene expression, systemic inflammatory responses, and oxidative stress (Fig. 1) [Soberanes et al. 2012; Ghio et al. 2012; Delfino et al. 2010; Sava and Carlsten 2012]. PM triggers inflammation and induces the release of reactive oxygen species (ROS), both locally and systemically [Vattanasit et al. 2013; Torres-Ramos et al. 2011; Dagher et al. 2005; Zhao et al. 2013; Dagher et al. 2005], which – in turn – can contribute to epigenetic alterations and genomic instability [Saint-Georges et al. 2009]. DNA methylation is a reversible mechanism of epigenetic regulation that plays a key role in a wide variety of fundamental biological processes by affecting gene expression at the transcriptional level. Both in vitro and in vivo studies suggest that aberrant DNA methylation may lead to abnormal cell cycle arrest, uncontrolled cell proliferation and apoptosis, all of which are known risk factors for carcinogenesis, aggressive tumor neovascularization, angiogenesis, and metastasis [Lou et al. 2013; Valavanidis et al. 2013; Ghio et al. 2012]. In addition, dysregulation of DNA methylation has been linked with overexpression of pro-inflammatory cytokines, especially in leukocytes [Venza et al. 2012; Mishra et al. 2012]. The cross-talk between blood leukocytes and inflammatory cells in the lung enhances chronic inflammation and ROS/reactive nitrogen species (RNS) generation in the lung, which leads to chronic tissue injury and abnormal tumor immunity.

Figure 1.

Proposed conceptual model linking PM exposure, leukocyte DNA hypomethylation and lung cancer. PM, particulate matter; SATα, satellite alpha; ROS, reactive oxygen species; RNS, reactive nitrogen species.

Consistent evidence from in vitro and in vivo studies has shown that exposure to air pollution, including PM, triggers DNA hypomethylation, which can underlie the activation of inflammatory genes and repetitive sequences, including tandem elements and interspersed repeats [De Prins et al. 2013; Tarantini et al. 2009]; Madrigano et al. 2011; Bellavia et al. 2013]. Approximately half of the human genome consists of repetitive elements [Lander et al. 2001], which require DNA methylation to maintain its silencing. The association between decreased repetitive element methylation and exposure to air pollution has been clearly demonstrated in both general population [Baccarelli et al. 2009; Madrigano et al. 2011] and occupationally exposed individuals [Bollati et al. 2007; Tarantini et al. 2009; Pavanello et al. 2010]. However, repetitive elements are controlled through multiple mechanisms, and each element has unique transcription patterns in response to cellular stressors [Meaney and Szyf 2005].

Repetitive elements may encode proteins involved in their replication and insertion into new locations within the genome. These elements influence genome transcriptional output and aberrant epigenetic alteration of the neighboring genes [Byun et al. 2012; Faulkner et al. 2009; Prada et al. 2012]. Dysregulation of repeated elements, e.g. hypomethylation, could be associated with genomic instability and overexpression of oncogenes, and therefore contributes to tumorigenesis [Lamprecht et al. 2010]. Zhu and colleagues have shown that individuals with lower methylation in blood leukocytes repetitive elements, such as the long interspersed nucleotide elements (LINE)-1 are at high risk of developing and dying from cancer [Zhu et al. 2011].

Previous studies have focused exclusively on a limited set of repeated-element sequences in LINE-1 and Alu interspersed repeats. Tandem repeats, a highly represented family of repetitive elements, are hypervariable in mammals. Dysregulation of tandem repeats methylation has been repetitively observed in many types of cancers [Yoshida et al. 2011; Weisenberger et al. 2005; Nishiyama et al. 2005a; Nishiyama et al. 2005b; Tsumagari et al. 2008; Choi et al. 2009]. Various studies have pointed out a distinctive methylation pattern among different tandem repeats dependent on the type of cancer [Choi et al. 2009]. Growing evidence indicates that each tandem repeat may have specific functions and distinct causes of dysregulation, but also that changes in DNA methylation of tandem repeat elements could behave differently than that of interspersed repeats [Nishiyama et al. 2005a; Nishiyama et al. 2005b; Choi et al. 2009]. These findings indicate the need for investigating methylation patterns in different types of tandem repeats.

Multiple types of tandem repeats with different functions have been described in the human genome. Satellite α (SATα) tandem repeats, composed of 170-bp DNA sequences, represent the main DNA component of every human centromere [Lee et al. 1997]. Hypomethylation of satellite repeats has been shown to be tightly related to chromatin decondensation, as well as chromosomal breakages and abnormalities [Choi et al. 2009; Fabris et al. 2011; Yoshida et al. 2011; Prada et al. 2012]. Aberrant overexpression of satellite repeats is observed in mouse and human epithelial cancer, including lung cancer [Ting et al. 2011]. Non-satellite tandem repeats such as NBL2 and D4Z4, also exhibit cancer-linked hypomethylation in certain types of cancer [Ehrlich 2006; Choi et al. 2009].

Although the significance of epigenetic changes in tandem repeats on lung cancer development has been repeatedly demonstrated, the initial determinants of hypomethylation are not yet well understood [Xiang et al. 2012]. Little research has investigated the interplay among tandem repeats methylation, inflammatory responses, and cellular oxidative stress induced by ambient PM insult. Therefore, investigating the effect of PM exposure on tandem repeats methylation is critical to provide new hypotheses for the increased PM-related lung cancer risk, especially in high risk populations residing in heavily polluted environments.

Beijing has been ranked in the World Development Indicators as one of the 15 cities with the highest levels of air pollution [World Bank 2012]. In Beijing, PM is primarily from traffic sources aggravated by high population density and the recent rapid increase in vehicular traffic [Yu et al. 2011]. Particles transported from industrial sources and windblown dust are also major sources of air pollution in Beijing [Yu et al. 2011]. In China, a 1.63% increase in lung cancer incidence was recorded each year between 1988 and 2005 [Chen et al. 2010]. In 2008, lung cancer incidence and mortality ranked first among all cancers, both in Beijing and nationwide [Chang et al. 2012]. In 2009, 4598 new lung cancer cases were reported among the 7,645,186 Beijing residents [Chen et al. 2013].

We previously reported the methylation levels of SATα, NBL2, and D4Z4 in relation to several airborne toxic metals in truck drivers and office workers in Beijing, China, who were exposed to high levels of PM [Hou et al. 2013]. In this study, we further investigated the association between short-term PM exposure (personal PM_2.5 and ambient PM₁₀) and methylation of these three tandem repeats in peripheral blood leukocytes. Peripheral leukocytes are the first responders to inflammatory mediators released in the lung after environmental insult. DNA methylation regulates leukocyte functions at the transcription level [Kato et al. 2012] and therefore affects the recruitment and migration of leukocytes from the bloodstream into the lung. Blood DNA methylation has been shown as a sensitive marker for acute and short-term changes in the epigenome [Baccarelli et al. 2009; De Prins et al. 2013; Kile et al. 2013; Bellavia et al. 2013]. Therefore, evaluating blood tandem repeats methylation in relation to PM exposure may unveil novel molecular targets for PM effects and also provide novel biomarkers reflecting biological effects from PM exposures [Brennan and Flanagan 2012; Wu et al. 2012; Marsit and Christensen 2013].

To examine the relationship between individual and environmental PM exposure and tandem-repeat methylation, we studied two groups of workers exposed to high levels of PM (truck drivers (n = 60) and office workers (n = 60)) in Beijing. To enhance the power to identify the effects of PM on tandem-repeat methylation, we studied each participant at two different time points, 1–2 weeks apart. We assessed their exposure to air pollution by using personal measures of PM_2.5 and elemental carbon (EC, a surrogate for PM from vehicular traffic), as well as ambient levels of PM₁₀. We hypothesized that individuals exposed to higher levels of PM will have differences in DNA methylation of SATα, NBL2, or D4Z4.

MATERIALS AND METHODS

Study Participants

The Beijing Truck Driver Air Pollution Study included 60 truck drivers and 60 office workers in Beijing, China, who participated between June 15 and July 27, 2008, as previously described [Baccarelli et al. 2011; Hou et al. 2013]. All participants lived in the Beijing metropolitan area and had been working at their current job for at least 2 years [Baccarelli et al. 2011]. Written informed Consent was obtained from all subjects. We examined all participants on two separate days 1–2 weeks apart. Therefore, the study was based on 120 observations for the 60 truck drivers and 120 observations for the 60 office workers, for a total of 240 observations. The truck driver and office worker groups were matched for smoking status and education. A self-administered questionnaire was used to collect detailed information on demographics, lifestyle, and other exposures. Information on time-varying factors, including tea and alcohol consumption and smoking habits, was obtained for both past lifestyles and their use on the examination days.

Personal Exposure Measurements

As previously reported [Baccarelli et al. 2011; Hou et al. 2013], we measured personal PM_2.5 on both examination days by providing the study participants with a lightweight gravimetric sampler that they wore during the 8 hours of work. The sampler was carried in a belt pack with the inlet clipped near the breathing zone. Each sampler setup included an Apex pump (Casella Inc., Bedford, UK), a Triplex Sharp-Cut Cyclone separator (BGI Inc., Waltham, MA), and a 37-mm Teflon filter placed on top of a drain disc and inside a metal filter holder. We kept the filters in atmosphere-controlled conditions before and after sampling and were weighed with a microbalance (Mettler-Toledo Inc., Columbus, OH). A time-weighted average of PM_2.5 concentration was recorded by dividing the change in filter weight before and after sampling by the volume of air sampled. We conducted technical validation tests of personal PM_2.5 measures and found high reproducibility (r = 0.944) of our measurements. The blackness of the filters used to measure PM_2.5 was assessed using an EEL Model M43D Smokestain Reflectometer (Diffusion Systems Ltd., London, UK), by applying standard black-smoke index calculations of the absorption coefficients based on reflectance (ISO 983). We assumed a factor of 1.0 for converting the absorption coefficient to EC mass [Kinney et al. 2000; Janssen et al. 2001], which was then divided by the sampled air volume to calculate average EC exposure concentration (ISO 983). EC is a combustion by-product contained in PM that has been used as a surrogate measure for PM from gasoline and especially diesel-powered motor vehicles [Kinney et al. 2000].

Ambient PM₁₀ Data

During the study period, ambient PM₁₀ (PM with an aerodynamic diameter ≤10 µm) data were obtained from the Beijing Municipal Environmental Bureau (http://www.bjepb.gov.cn/air2008/Air.aspx). To estimate the average PM₁₀ level, we used daily averages of PM₁₀ computed from data obtained from 27 monitoring stations throughout Beijing. We used these ambient PM₁₀ data to assess exposure not only on the day blood was collected, but also on previous days in order to examine the distribution of the effect of air pollutants over time, i.e., the lag between the timing of exposure and the response (in our case tandem-repeat methylation). We also obtained daily outdoor temperature and humidity data for Beijing from the National Oceanic and Atmospheric Administration’s online database (http://www.noaa.gov/).

Sample Preparations and DNA Extraction

We used standardized protocols for blood collection and storage. Peripheral blood was collected from each participant on both examination days, and buffy coat was separated within 2 hours and stored locally at −80°C. DNA was extracted from buffy coat by using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI) according to the manufacturer’s instructions. Purified DNA was resuspended in the kit’s hydration solution, quantified, and stored at −20°C until analysis. Blood DNA samples were randomized across plates to limit potential bias from plate effects, and laboratory personnel were blinded to both exposure groups and exposure study.

DNA Methylation Analysis

As previously reported [Hou et al. 2013], 500 ng of DNA was treated with bisulfite using the EZ-96 DNA Methylation-Gold Kit (Zymo Research, Orange, CA) according to the manufacturer’s protocol to determine methylation. Final elution was performed with 30 µl of M-Elution Buffer. DNA methylation was quantified using bisulfite polymerase chain reaction (PCR) and pyrosequencing. The primers and conditions were described previously [Choi et al. 2009]. Briefly, a 30-µl PCR was carried out in 15 µl of GoTaq Hot Start Green Master Mix (Promega, Madison, WI), with 10 pmol of forward primer, 10 pmol of reverse primer, 1 µl of bisulfite-treated genomic DNA, and water. The amplicon size is in the range of 150–250 bp. Pyrosequencing was performed using the PyroMark Q96 MD Pyrosequencing System (QIAGEN, Germantown, MD). The percentage of methylated and unmethylated cytosines was quantified for three CpG sites (cytosine and guanine separated by one phosphate) from SATα, four CpG sites from NBL2, and four CpG sites from D4Z4. The degree of methylation is expressed as the percentage of methylated cytosines divided by the sum of methylated and unmethylated cytosines (%5mC) measured in each sample.

Measurement of Plasma Inflammatory Mediators

We measured plasma inflammatory mediators in all study subjects, using a pre-set Luminex Bio-Plex 27-plex panel (Bio-Rad, Hercules, CA). The multiplex bead-based Luminex Bio-Plex 27-plex panel permits the simultaneous quantification of 27 cytokines from previously stored plasma, includes IL7, IL8, IL12, GM-CSF, IL4, IL5, IL9, IL10, IL13, GM-CSF, IL1β, IL1ra, IL15, IL17, MCP1, MIP1α, MIP1β, PDGF-BB, VEGF, FGF, RANTES, IP-10, eotaxin, GCSF. 27 sets of fluorescently dyed beads were conjugated with target-specific reactants, and triggered biochemical reactions with target biomolecules, which were detected by flow cytometer. The fluorescent output was processed by a high-speed digital signal processor, and cytokine concentrations were automatically calculated by Bio-Plex Manager™ software using a standard curve. Samples were run in triplicates.

Statistical Analysis

The demographic characteristics of truck drivers and office workers were summarized using standard descriptive statistics (Table I). Because our data include two measures for each study participant, we used a mixed-effect regression model to account for the lack of independence between the two measures [McLean et al. 1991]. Our previous work [Hou et al. 2013] on exposure to airborne toxic metals and tandem repeat methylation used the average methylation levels across different positions for each tandem repeat. To avoid the assumptions of homogeneity-of-regression slopes across different CpG sites within a tandem repeat and achieve the best model fit for the entirety of the data presented, here we considered methylation measures on different positions in each tandem repeat as repeated measures. Therefore, in the present investigation, we used a multilevel model instead of a single-level model to incorporate a nested dependent structure. We first evaluated differences between groups (truck drivers vs. office workers) by using the following multilevel mixed-effect regression model:

$${\gamma }_{ijk}={\beta }_0+{\beta }_1{(\text{group})}_{ijk}+{\beta }_2{\mathrm{X}}_{2ijk}+\dots +{\beta }_{\mathrm{p}}{\mathrm{X}}_{\mathrm{p}ijk}+{\nu }_{0k+}{\mu }_{0jk}+e_{0ijk}$$ [1]

where γ_ijk is the methylation level of ith CpG sites at each tandem repeat investigated for the jth subject on the kth examination day (i = 1, 2, 3 for SATα, i = 1, 2, 3, 4 for NBL2 and D4Z4; j = 1,…,n; and k = 1, 2); β₀ is the overall intercept; β₁ is the regression coefficient for the group; β₂…β_p are the regression coefficients for the covariates, which were all fitted as fixed effects and included variables either not matched or not completely matched by design between the two groups, i.e., age, body mass index (BMI), number of cigarettes per examination day, and work hours per day; ν_0k is the random effect at the examination day level; µ_0jk is the random effect at the subject level, and e_0ijk is the random effect at the CpG site level. We did not adjust for the matching variables sex and smoking, because they were nearly perfectly balanced between groups by matching. In contrast, Hou et al. adjusted for matching factors and other potential confounding variables such as the usage of central heating. The different approaches led to minor discrepancy in the results from between-group methylation comparisons [Hou et al. 2013].

Table I.

Characteristics of the Study Participants

Characteristic	Office Workers (n = 60)	Truck Drivers (n = 60)	P valuea
Sex, n (%)
Male	40 (66.7)	40 (66.7)
Female	20 (33.3)	20 (33.3)	1.00
Age [years], mean ± SD	30.3 ± 8.0	33.5 ± 5.7	<0.001
Smoking status n (%)
Never smoked	35 (58.4)	34 (56.7)
Ex-smoker	2 (3.3)	2 (3.3)
Current smoker	23 (38.3)	24 (40.0)	1.00
Cigarettes smoked during the examination time, mean ± SDb	0.5 ± 1.7	2.3 ± 4.2	<0.001b
BMI [kg/m2], mean ± SD	22.8 ± 3.4	24.3 ± 3.2	0.01
Work hours per day, mean ± SDb	7.9 ± 1.3	9.9 ± 1.7	<0.001

^aP values were calculated using Student’s t-tests and Fisher’s exact tests for continuous and categorical variables, respectively, except for the variable indicated in footnote ^b below.

^bMeasures from 1 study day. Mixed-effect regression was used in the statistical analyses to account for the lack of independence of the repeated measures and computed means, standard deviations, and P values.

SD, standard deviation; BMI, body mass index.

Mixed-effect regression models were also used to evaluate the associations of personal PM_2.5, personal EC, and ambient PM₁₀ with tandem-repeat methylation adjusted for age, sex, BMI, smoking status, number of cigarettes per day, work hours per day, outdoor humidity, and temperature. In order to optimize power, we conducted primary analyses on the association between exposure measures and methylation levels of tandem repeats by fitting these models in all participants combined. We then evaluated associations in office workers or truck drivers separately. The mixed-effect model was:

$${\gamma }_{ijk}={\beta }_0+{\beta }_1{(\text{Exposure})}_{ijk}+{\beta }_2{\mathrm{X}}_{2ijk}+\dots +{\beta }_{\mathrm{p}}{\mathrm{X}}_{\mathrm{p}ijk}+{\nu }_{0k+}{\mu }_{0jk}+e_{0ijk}$$ [2]

where γ_ijk is the methylation level of ith CpG sites at each tandem repeat investigated for the jth subject on the kth examination day (i = 1, 2, 3 for SATα, i = 1, 2, 3, 4 for NBL2 and D4Z4; j = 1,…,n; and k = 1, 2); β₀ is the overall intercept; β₁ is the regression coefficient for the exposure variable (PM_2.5, EC, or PM₁₀); β₂…β_p are the regression coefficients for the covariates, which included age, sex, BMI, smoking status, number of cigarettes smoked/examination day, work hours/day, outdoor humidity, and temperature; ν_0k is the random effect at the examination day level; µ_0jk is the random effect at the subject level, and e_0ijk is the random effect at the CpG site level. Because personal exposure levels are expected to be higher in smokers, we also fit a second set of models in which smoking status and number of cigarettes smoked were excluded from the list of dependent variables. These models generated regression coefficients for air pollution exposures very similar to the full models with no differences in statistical significance.

To explore the timing of the relationship between the exposure and tandem-repeat methylation, we further analyzed ambient PM₁₀ measures, which—because they were available daily throughout the study period—could be used not only to assess exposure on the day of blood collection, but also on previous days (i.e., lags). To this end, we used ambient PM₁₀ data up to 7 days before the blood collection in unconstrained distributed lag models. All tests were two-sided, and P < 0.05 was considered significant.

RESULTS

Characteristics of the Study Participants and Exposure to Air Particulate Matter

The characteristics of the 60 office workers and 60 truck drivers were described previously [Baccarelli et al. 2011; Hou et al. 2013]. Briefly, truck drivers were moderately older than office workers, had higher BMI, smoked more cigarettes during the 8 hours of the study, and worked more hours per day. Mean personal PM_2.5 was 126.8 µg/m³ for truck drivers and 94.6 µg/m³ for office workers (P < 0.001) (Table II). Average personal EC was 17.2 µg/m³ for truck drivers and 13.0 µg/m³ for office workers (P < 0.001). As expected, the levels of ambient PM₁₀, outdoor temperature, and relative humidity on the day of examination did not differ between truck drivers and office workers (Table II). Personal PM_2.5 exposure was higher in smokers compared to nonsmokers among office workers and truck drivers, as well as in all participants combined (Supplementary Table S1). Personal PM_2.5 exposure was also significantly correlated with the number of cigarettes smoked during the 8 hours of monitoring (r = 0.37; P < 0.001). Personal EC exposure was higher in smokers in all participants combined and among truck drivers but not among office workers (Supplementary Table S1). EC was also significantly correlated with the number of cigarettes smoked during the 8 hours of monitoring (r = 0.29; P = 0.005).

Table II.

Level of Personal Exposure to PM_2.5 and Elemental Carbon (EC) During Work Hours and Level of Ambient PM₁₀, Outdoor Temperature, and Humidity on the Examination Days

Time window	Office Workers (60 participants, 2 measures per participant)a								Truck Drivers (60 participants, 2 measures per participant)a
	# Obs	Mean	SD	10th pct	25th pct	Median	75th pct	90th pct	# Obs	Mean	SD	10th pct	25th pct	Median	75th pct	90th pct	P value
Personal PM2.5b (µg/m3) on the examination days, from personal monitors
8 hoursa	120	94.6	64.9	22.4	48.5	86.2	126.6	183.4	119	126.8	68.8	46.3	73.9	116.8	160.5	213.9	<0.001
Personal ECb (µg/m3) on the examination days, from personal monitors
8 hoursa	118	13.0	4.0	7.3	10.1	13.2	15.7	18.4	120	17.2	6.6	9.2	12.9	16.7	20.8	25.5	<0.001
Ambient PM (µg/m3) from ambient monitors on the day of the study
24 hoursc	120	116.7	50.2	60.0	82.0	114.0	141.0	181.0	120	123.5	50.1	72.0	88.0	116.0	150.0	188.0	0.29
Outdoor relative humidity (%) on the day of the study
24 hoursc	120	20.6	2.1	17.5	20.0	21.0	21.5	23.0	120	20.6	2.1	17.0	20.0	21.0	22.0	23.0	0.93
Outdoor temperature (°C) on the day of the study
24 hoursc	120	25.4	2.5	22.0	23.0	26.0	28.0	29.0	120	25.3	2.5	22.0	23.0	26.0	28.0	28.0	0.96

^aEach of the truck drivers (n = 60) and office workers (n = 60) was studied on two different examination days for a total of 120 observations in each of the two groups. Mixed-effect regression was used in the statistical analyses to account for the lack of independence of the repeated measures and to compute P values.

^bMeasured during the work hours of examination days using lightweight personal monitors.

^c24-hour average of ambient PM₁₀, relative humidity, or temperature on the examination days.

^#Obs, number of observations; SD, standard deviation; pct, percentile

DNA Methylation of Tandem-Repeat Sequences in Truck Drivers and Office Workers

Tandem-repeat methylation levels exhibited varying degrees of within-participant correlation between the two examination days, which were separated by 1–2 weeks (Fig. 2). In all participants combined, Pearson’s correlation coefficients (r) were 0.80 for SATα methylation, 0.81 for NBL2 methylation, and 0.63 for D4Z4 methylation. There were no significant differences in methylation levels of SATα, NBL2, and D4Z4 between truck drivers and office workers in unadjusted analyses, as well as in analyses adjusted for age, BMI, number of cigarettes, and work hours/day (Table III). Further analyses of DNA methylation levels at individual CpG sites within each of the sequences did not show any statistically significant differences between truck drivers and office workers (Supplementary Table S2).

Figure 2.

Scatterplots of the within-participant correlations in blood tandem-repeat methylation between the two examination days. Circles and crosses represent truck drivers and office workers, respectively.

Table III.

Average DNA Methylation of Tandem-Repeat Sequences in Office Workers and Truck Drivers

	DNA Methylation (%5mC)
Marker	Office Workers (60 participants, 2 measures per participant)a		Truck Drivers (60 participants, 2 measures per participant)a		P value
	Mean	(SD)	Mean	(SD)
Unadjusted
SATα	73.9	(15.1)	73.1	(15.1)	0.17
NBL2	82.2	(5.7)	82.6	(5.7)	0.34
D4Z4	66.7	(6.5)	67.3	(6.5)	0.52
Adjusted for age, BMI, number of cigarettes, and work hours/dayb
SATα	79.6	(15.3)	79.7	(15.3)	0.93
NBL2	86.1	(6.0)	87.0	(6.1)	0.11
D4Z4	69.6	(7.6)	70.8	(7.8)	0.26

^aEach of the truck drivers (n = 60) and office workers (n = 60) was studied on 2 different examination days for a total of 120 observations in each of the two groups. Mixed-effect regression was used in the statistical analyses to account for the lack of independence of the repeated measures and to compute means, standard deviations (SDs), and P values.

^bAdjusted means were computed by holding covariates fixed at their average values. Office workers and truck drivers were matched by sex and smoking status (never, former, current). Only unmatched variables were adjusted for in the regression models.

SD, standard deviation; BMI, body mass index.

Association of Personal PM_2.5, Personal EC, and Ambient PM₁₀ with Methylation Levels of Tandem Repeats

In covariate-adjusted analyses of all participants combined, personal PM_2.5 exposure exhibited a significant negative association with SATα methylation. An interquartile increase in personal PM_2.5 exposure was associated with a 1.35%5mC decrease in SATα methylation (P = 0.01) in a model adjusted for age, sex, BMI, smoking status, number of cigarettes smoked during the 8 hours of the study, work hours/day, and outdoor humidity and temperature on the examination days. The negative correlation between personal PM_2.5 exposure and SATα methylation was found in both groups, although it was more pronounced and statistically significant (P = 0.02) only in truck drivers (Table IV).

Table IV.

Effects of Airborne Particulate Matter on DNA Methylation of Tandem-Repeat Sequences^a

Marker	Exposure	All Participants (120 participants, 2 measures per participant)b			Office Workers (60 participants, 2 measures per participant)b			Truck Drivers (60 participants, 2 measures per participant)b
		β	(SE)	P value	β	(SE)	P value	β	(SE)	P value
SATα	Personal PM2.5c	−1.35	(0.54)	0.01	−0.95	(0.83)	0.26	−2.34	(0.94)	0.02
	Personal ECc	−0.37	(0.49)	0.45	−0.84	(0.90)	0.36	−0.54	(0.68)	0.43
	Ambient PM10d	−1.33	(0.51)	0.01	−1.26	(0.79)	0.12	−1.44	(0.75)	0.06
NBL2	Personal PM2.5c	−0.05	(0.46)	0.92	0.23	(0.71)	0.75	−0.88	(0.84)	0.30
	Personal ECc	0.03	(0.40)	0.94	−0.44	(0.77)	0.57	0.09	(0.58)	0.88
	Ambient PM10d	−0.09	(0.44)	0.84	1.04	(0.66)	0.13	−1.38	(0.63)	0.03
D4Z4	Personal PM2.5c	0.74	(0.91)	0.42	0.36	(1.43)	0.80	1.02	(1.73)	0.56
	Personal ECc	0.15	(0.80)	0.85	0.36	(1.54)	0.81	−0.83	(1.17)	0.49
	Ambient PM10d	−0.13	(0.87)	0.88	0.88	(1.36)	0.52	−1.34	(1.33)	0.32

^aEstimates (βs) and standard errors (SEs) adjusted for age, sex, BMI, smoking status, number of cigarettes smoked during the 8 hours of the study, work hours/day, and outdoor humidity and temperature on the examination days. Estimates are scaled to an interquartile range increase in exposure.

^bEach of the truck drivers (n = 60) and office workers (n = 60) was studied on two different examination days for a total of 120 observations in each group. Mixed-effect regression was used in the statistical analyses to account for the lack of independence of the repeated measures and to compute β coefficients, standard errors (SEs), and P values.

^cMeasured during the work hours of examination days using lightweight personal monitors.

^d24-hour average measure of ambient PM₁₀ on the examination days

Ambient PM₁₀ levels also showed a significant negative association with SATα methylation; an interquartile increase in ambient PM₁₀ levels was associated with a covariate-adjusted 1.33%5mC decrease in SATα methylation in all participants combined (P = 0.01). The negative correlation between ambient PM₁₀ exposure and SATα methylation was found in both groups, although it was moderately more pronounced and borderline statistically significant (P = 0.06) only in truck drivers (Table IV). Personal EC exposure did not show any significant associations with SATα methylation (Table IV).

In all participants combined, NBL2 and D4Z4 showed no associations with the levels of personal PM_2.5, personal EC, or ambient PM₁₀ (Table IV). However, ambient PM₁₀ levels showed a significant negative association with NBL2 methylation in truck drivers but not in office workers: an interquartile increase in ambient PM₁₀ levels was associated with a covariate-adjusted 1.38%5mC decrease in NBL2 methylation among truck drivers (P = 0.03).

To complement the results of the covariate-adjusted analyses, we present scatterplots of the raw data and unadjusted correlations corresponding to the significant findings described above (Fig. 3). Consistent with the covariate-adjusted analyses, all the unadjusted correlations were negative, although the degree of significance varied: in all participants combined, an interquartile increase in personal PM_2.5 was associated with a nonsignificant unadjusted 0.32%5mC decrease in SATα methylation (P = 0.45, Fig. 3A); among truck drivers, an interquartile increase in personal PM_2.5 was associated with a nonsignificant unadjusted 0.55%5mC decrease in SATα methylation (P = 0.33, Fig. 3B); in all participants combined, an interquartile increase in ambient PM₁₀ was associated with an unadjusted 1.00%5mC decrease in SATα methylation (P = 0.05; Fig. 2, Fig. 3C); and in truck drivers, an interquartile increase in ambient PM₁₀ levels was associated with an unadjusted 1.05%5mC decrease in NBL2 methylation (P = 0.07; Fig. 3D).

Figure 3.

Scatterplots of the crude correlations between particulate matter exposures and tandem-repeat methylation for the four significant findings from adjusted multivariable analyses. Circles and crosses represent observations from the first and second examination days, respectively. Unadjusted and adjusted regression estimates obtained from mixed-effect regression models are shown. Mixed-effect regression was used to account for the lack of independence of the repeated measures and to compute β coefficients, standard errors (SEs), and P values. Adjusted estimates were computed from models fitting the following independent variables: age, sex, BMI, smoking status, number of cigarettes smoked during the 8 hours of the study, work hours/day, and outdoor humidity and temperature on the examination days. Estimates are scaled to an interquartile range (IQR) increase in exposure.

To examine the timing of the relationship between PM exposure and tandem-repeat methylation, we used unconstrained distributed lag models to further analyze ambient PM₁₀. PM₁₀ measures were available daily throughout the study period and could be used not only to assess exposure on the day of blood collection, but also on the previous days (i.e., lags). As reported in Table V, results using the distributed lag model showed that the effects of PM₁₀ observed on examination days were distributed across several lags 3–6 days before the day of blood collection. For instance, the negative correlation between ambient PM₁₀ and SATα methylation was observed between 3-day and 6-day lags, with the most negative estimate corresponding to the 6-day lag (Table V).

Table V.

Results of Distributed-Lag Models Examining the Timing of the Effects of PM₁₀ Before the Examination Days on DNA Methylation of Tandem-Repeat Sequences^a

Marker	Exposure	All Participants (120 participants, 240 examination days)			Office Workers (60 participants, 120 examination days)			Truck Drivers (60 participants, 120 examination days)
		β	(SE)	P-value	β	(SE)	P-value	β	(SE)	P-value
SATα	Ambient PM10 (1-day lag)	0.35	(0.41)	0.39	1.31	(0.58)	0.026	−0.64	(0.59)	0.27
	Ambient PM10 (2-day lag)	0.36	(0.43)	0.41	−0.92	(0.64)	0.16	1.20	(0.59)	0.045
	Ambient PM10 (3-day lag)	−0.34	(0.35)	0.33	0.16	(0.48)	0.74	−1.46	(0.51)	0.005
	Ambient PM10 (4-day lag)	−0.42	(0.42)	0.33	−1.02	(0.57)	0.07	0.21	(0.63)	0.74
	Ambient PM10 (5-day lag)	−0.23	(0.44)	0.61	0.38	(0.68)	0.57	−0.88	(0.61)	0.16
	Ambient PM10 (6-day lag)	−0.50	(0.48)	0.30	−1.57	(0.67)	0.021	0.14	(0.69)	0.84
	Ambient PM10 (7-day lag)	0.12	(0.43)	0.79	0.18	(0.60)	0.77	−0.84	(0.64)	0.19
NBL2	Ambient PM10 (1-day lag)	−0.09	(0.15)	0.58	0.10	(0.26)	0.71	−0.24	(0.20)	0.23
	Ambient PM10 (2-day lag)	0.20	(0.16)	0.22	0.10	(0.29)	0.72	0.20	(0.19)	0.30
	Ambient PM10 (3-day lag)	−0.06	(0.13)	0.65	0.03	(0.21)	0.90	−0.24	(0.17)	0.15
	Ambient PM10 (4-day lag)	0.33	(0.16)	0.041	0.48	(0.25)	0.06	0.07	(0.21)	0.73
	Ambient PM10 (5-day lag)	−0.09	(0.17)	0.58	−0.06	(0.31)	0.85	−0.29	(0.21)	0.16
	Ambient PM10 (6-day lag)	0.03	(0.18)	0.85	−0.23	(0.30)	0.46	0.31	(0.23)	0.18
	Ambient PM10 (7-day lag)	0.09	(0.16)	0.60	0.09	(0.27)	0.74	−0.12	(0.22)	0.60
D4Z4	Ambient PM10 (1-day lag)	−0.16	(0.27)	0.56	−0.19	(0.39)	0.62	−0.11	(0.40)	0.78
	Ambient PM10 (2-day lag)	0.40	(0.29)	0.17	0.64	(0.43)	0.14	0.25	(0.41)	0.55
	Ambient PM10 (3-day lag)	0.07	(0.23)	0.77	0.00	(0.32)	0.99	0.24	(0.35)	0.49
	Ambient PM10 (4-day lag)	−0.45	(0.28)	0.11	−0.64	(0.38)	0.09	−0.37	(0.44)	0.40
	Ambient PM10 (5-day lag)	0.24	(0.30)	0.42	−0.10	(0.45)	0.83	0.58	(0.43)	0.18
	Ambient PM10 (6-day lag)	−0.28	(0.32)	0.39	−0.23	(0.45)	0.61	−0.59	(0.49)	0.23
	Ambient PM10 (7-day lag)	−0.14	(0.28)	0.63	0.04	(0.40)	0.92	−0.05	(0.44)	0.92

^aEstimates adjusted for age, sex, BMI, smoking status, number of cigarettes smoked during the eight hours of the study, work hours/day, and outdoor humidity and temperature on the examination days. For each marker, seven lag-day PM₁₀ variables were evaluated at the same time in one model.

^a Estimates (βs) and standard errors (SEs) adjusted for age, sex, BMI, smoking status, number of cigarettes smoked during the 8 hours of the study, work hours/day, and outdoor humidity and temperature on the examination days. Estimates are scaled to an interquartile range (IQR) increase in exposure.

^bEach of the truck drivers (n = 60) and office workers (n = 60) was studied on 2 different examination days for a total of 120 observations in each of the two groups. Mixed-effect regression was used in the statistical analyses to account for the lack of independence of the repeated measures and to compute β coefficients, standard errors (SEs), and P values.

Correlation of Tandem-Repeat Methylation with Inflammatory Markers

We analyzed DNA methylation of tandem repeats in relation to plasma levels of 27 cytokines measured using a premade platform (Bio-Plex Cytokine 27-plex Pro Human Cytokine; Bio-Rad, Hercules, CA). As shown in Supplementary Table S3, 9 of the 27 cytokines were negatively correlated with SATα methylation at a false discovery rate (FDR) lower than 0.05. Neither NBL2 nor D4Z4 methylation showed positive correlations with cytokine levels at the same FDR threshold (Table VI).

Table VI.

Correlation of Tandem-Repeat DNA Methylation with Plasma Levels of 27 Cytokines

	SATα				NBL2				D4Z4
Cytokine	β	SE	P value	FDR	β	SE	P value	FDR	β	SE	P value	FDR
EOTAXIN	−0.345	0.365	0.348	0.47	−0.202	0.673	0.765	0.892	0.477	0.296	0.11	0.269
FGF_BASIC	−0.272	0.138	0.051	0.107	0.267	0.285	0.351	0.892	0.123	0.152	0.418	0.538
GM_CSF	−0.185	0.107	0.086	0.154	−0.017	0.216	0.938	0.938	0.069	0.109	0.525	0.621
G_CSF	−0.122	0.094	0.196	0.294	0.379	0.19	0.049	0.892	0.138	0.094	0.143	0.297
IFN_G	−1.452	0.595	0.016	0.049	1.304	1.231	0.292	0.892	1.423	0.607	0.021	0.157
IL_10	−0.018	0.006	0.007	0.032	−0.006	0.014	0.669	0.892	0.002	0.007	0.783	0.844
IL_12_P70	0.047	0.124	0.705	0.762	−0.068	0.211	0.748	0.892	0.08	0.08	0.32	0.5
IL_13	−0.115	0.082	0.164	0.261	−0.031	0.174	0.859	0.892	0.047	0.086	0.588	0.662
IL_15	−0.025	0.013	0.052	0.107	−0.013	0.026	0.632	0.892	0.009	0.014	0.529	0.621
IL_17	−0.248	0.09	0.007	0.032	−0.059	0.19	0.757	0.892	0.087	0.103	0.401	0.538
IL_1B	−0.033	0.012	0.008	0.032	0.04	0.026	0.123	0.892	0.027	0.013	0.037	0.166
IL_1RA	−1.856	0.691	0.008	0.032	0.683	1.454	0.639	0.892	0.981	0.71	0.169	0.305
IL_2	−0.033	0.071	0.64	0.72	0.028	0.137	0.84	0.892	0.07	0.072	0.333	0.5
IL_4	−0.036	0.012	0.002	0.032	0.027	0.025	0.277	0.892	0.025	0.012	0.051	0.167
IL_5	−0.061	0.024	0.01	0.035	0.01	0.049	0.837	0.892	0.036	0.024	0.138	0.297
IL_6	−0.06	0.039	0.124	0.21	0.018	0.077	0.813	0.892	0.032	0.037	0.395	0.538
IL_7	−0.154	0.054	0.005	0.032	0.193	0.113	0.091	0.892	0.153	0.057	0.008	0.11
IL_8	−0.139	0.047	0.004	0.032	0.058	0.097	0.552	0.892	0.106	0.046	0.023	0.157
IL_9	−0.115	0.505	0.82	0.852	−0.282	0.927	0.762	0.892	−0.055	0.377	0.884	0.884
IP_11	−1.247	2.596	0.632	0.72	2.265	5.097	0.658	0.892	2.772	2.693	0.305	0.5
MCP_1_MCAF	−0.342	0.148	0.023	0.061	0.178	0.285	0.534	0.892	0.253	0.127	0.048	0.167
MIP_1A	0.019	0.04	0.638	0.72	0.019	0.075	0.803	0.892	0.073	0.033	0.029	0.157
MIP_1B	−0.592	0.279	0.036	0.088	0.313	0.526	0.554	0.892	0.651	0.238	0.007	0.11
PDGF_BB	−8.466	4.446	0.059	0.115	7.201	9.139	0.432	0.892	6.8	4.79	0.158	0.305
RANTES	−19.952	17.723	0.263	0.373	13.553	35.181	0.701	0.892	30.901	16.827	0.069	0.186
TNF_A	−0.068	0.423	0.873	0.873	−0.481	0.703	0.495	0.892	−0.069	0.288	0.813	0.844
VEGF	−0.071	0.1	0.478	0.615	−0.111	0.187	0.553	0.892	0.155	0.08	0.056	0.167

SE, standard error; FDR, false discovery rate.

DISCUSSION

In the present investigation of truck drivers and office workers in Beijing, China, who were exposed to high levels of PM, we found significant associations between both personal PM_2.5 and ambient PM₁₀ levels and lower levels of SATα methylation in DNA from blood. The associations were more pronounced in truck drivers than in office workers. There were no associations between EC exposure and tandem repeat methylation.

Air pollution has been consistently associated with adverse health outcomes, including lung cancer, in several epidemiologic studies [Krewski et al. 2005; Vineis et al. 2006; Merlo et al. 2010; Raaschou-Nielsen et al. 2010]. Although the mechanisms underlying air pollution-related lung carcinogenesis remain to be fully clarified, many carcinogens and co-carcinogens found in tobacco smoke, such as polycyclic aromatic hydrocarbons, transition metals, benzene, and aldehydes, are also present in PM [Feng et al. 2006].

Our study investigated a group of healthy individuals exposed to high levels of PM, well above the levels that have been documented to cause lung carcinogenicity [Pope et al. 2004]. We found that SATα methylation measured in blood leukocytes was negatively associated with personal PM_2.5 and ambient PM₁₀ levels. Personal exposure to EC, used as a tracer of PM from vehicular traffic, was not significantly associated with tandem-repeat methylation. Taken together, these findings indicate that SATα methylation is sensitive to overall particle levels, as traced by personal PM_2.5 and ambient PM₁₀, rather than specifically to PM from vehicular traffic. Consistent with this conclusion, we found negative correlations of personal PM_2.5 and ambient PM₁₀ with SATα methylation among truck drivers and office workers, although they were moderately stronger in truck drivers. Our previous study found a positive association between NBL2 methylation and concentrations of Si and Ca in truck drivers, and a positive association between SATα methylation and concentrations of S in office workers. The opposite directions of associations in two sets of pollutants suggest that the patterns of epigenetic regulation of tandem repeat are not only element-specific, but also exposure-specific. We did not find any significant association between D4Z4 methylation and PM exposures. A subgroup association of ambient PM₁₀ was observed with NBL2 methylation. Consistent with our previous finding, SATα and NBL2 methylation demonstrate a better sensitivity than D4Z4 methylation to the overall particle and elemental components. Whether D4Z4 and NBL2 methylation in DNA from blood specimens or other tissues is sensitive to PM exposure should be further evaluated in future investigations using larger samples of exposed individuals.

Extensive evidence indicates that decreased methylation of repetitive elements in cancer tissues is causally involved in carcinogenesis [Ehrlich 2009; Wolff et al. 2010]. Initial findings also indicate that repetitive element methylation measured in blood leukocyte DNA might provide a biomarker for cancer risk, thus supporting epigenetic analyses of blood samples. For instance, in a recent study of 722 elderly participants from the Normative Aging Study cohort, Zhu et al. showed that repetitive element hypomethylation in blood leukocyte DNA was related to higher incidence of and mortality from cancer, including lung cancer [Zhu et al. 2011]. A recent study by Friso et al. showed that low levels of DNA methylation in blood, which is largely repetitive element methylation, was associated with cancer risk [Friso et al. 2013]. In addition, the sensitivity of peripheral blood DNA methylation profile after PM exposure suggests that blood DNA could serves as an appropriate biomarker to monitor effects of PM on the epigenome [Baccarelli et al. 2009; De Prins et al. 2013; Kile et al. 2013].

Tandem repeats are a distinct family of repetitive elements that, although extensively present in the human genome and involved in cancer etiology, have not yet been studied in human investigations of environmental carcinogens. Relative to most regions of the genome, tandem repeats display a greater propensity to mutate and variable methylation may influence their mutagenicity rates. Our results suggest that different tandem repeats have differential sensitivity to DNA methylation. In particular, we found the strongest associations between PM exposure and lower methylation of SATα. This finding is consistent with the biological behavior of different types of tandem repeats in carcinogenesis. Mutation rates in tandem repeats have been shown to be highly dependent on the structure and length of each specific, repetitive element [Brinkmann et al. 1998]. Similarly, the patterns of epigenetic regulation of each type of tandem repeat are specific to each element [Meaney and Szyf 2005]. In fact, although the three tandem repeats evaluated in our study have all been shown to contribute to chromosomal instability if dysregulated, they are controlled through different mechanisms and have different transcription patterns in response to cellular stressors [Meaney and Szyf 2005]. In particular, the SATα family of tandem-repeated sequences, which are found in all normal human centromeres, has been shown to be methylated via the action of the DNA methyltransferase 3B (DNMT3B), which plays a critical role in maintaining chromosomal stability [Ehrlich, Sanchez et al. 2008]. A constitutive centromeric protein, centromere protein C (CENP-C), has been shown to interact specifically with DNMT3B and to recruit it to methylate SATα repeats [Gopalakrishnan, Sullivan et al. 2009]. A recent study based on next-generation digital gene expression analysis in cancer cells, including lung cancer, showed that satellite hypomethylation is associated with massive expression of noncoding RNAs (ncRNAs) that may have active roles in dysregulated cell growth and cancer progression [Ting, Lipson et al. 2011]. It remains to be determined whether the relation between PM exposure and SATα DNA hypomethylation is mediated by effects on the DNMT3B/CENP-C interaction and by increased ncRNA expression.

Our experimental data showed that the decreased methylation of SATα, but not of NBL2 and D4Z4, was correlated with higher levels of several cytokines in plasma. This finding demonstrates that SATα hypomethylation is associated with both increased immunoregulatory and pro-inflammatory cytokines levels. However, our results are observational and therefore not sufficient to explain their roles as inflammatory mediators after PM exposure. Future mechanistic studies are needed to unravel the mechanism of oxidative stress and inflammatory responses in the hypomethylation of the SATα sequences and subsequent lung carcinogenesis.

DNA methylation is a reversible mechanism of epigenetic regulation and can reflect the dynamic gene-environmental interaction. Reduced blood DNA methylation has been repetitively reported after short-term PM exposure[Dioni, Hoxha et al. 2011; Bind, Baccarelli et al. 2012; Bellavia, Urch et al. 2013; Kile, Fang et al. 2013], suggesting that this condition may affect the epigenome. In our study, we evaluated DNA methylation of tandem repeats only in relation to personal PM_2.5 levels of exposure measured on the same day of the examinations. In addition, because ambient PM₁₀ data were available throughout the study period, we used distributed lag models to explore the timing of the relationship between PM₁₀ exposure and tandem-repeat methylation.

In Table V, the reduction in SATα methylation associated with the 3-day lag was highly significant in truck drivers. For office workers, the negative estimates corresponded to both the 4-day and 6-day lags. We observed increased SATα methylation with marginal significance in the 1-day and 2-day lags for the office worker group and the truck driver group respectively. Whereas our lag model results do not provide conclusive evidence as to the temporal relationships linking PM exposure to SATα methylation, the fluctuation in methylation level across the exposure period suggest a relatively high degree of variability in DNA methylation. However, these results appear to indicate that the effect of PM exposures on tandem-repeat methylation is not solely due to the PM levels on the day of blood collection but rather, PM exposure may be distributed over the preceding several days. The most significant reduction in SATα methylation corresponded to 3-day and 6-day lag in the truck drivers and office workers, respectively. This result suggests that the difference in DNA methylation response between two groups might reflect the magnitude of environmental exposure.

As mentioned above, DNA hypomethylation is usually considered a reversible epigenetic change. However, the consequences of hypomethylation, such as uncontrolled inflammation responses, ROS production, abnormal cell cycle arrest, enhanced blood-to-lung signaling, and genomic instability may lead to irreversible alterations and are associated with both the duration and intensity of exposure.

Given the slow clearance rate of PM from the respiratory system, combined with the fact that the effects of chronic PM pollution exist even at very low exposure, it is plausible that PM induced inflammation, ROS damage, and tissue damage are cumulative [Brunekreef and Holgate 2002; Daniels, Dominici et al. 2000]. These events and changes are important risk factors for lung carcinogenesis. In addition, inflammation and ROS stress exacerbates DNA hypomethylation. As those stimuli continues, the methylation and stress/damage in the lung act in a positive feedback manner. As such, capturing this reversible epigenetic change associated with short-term PM exposure may represent a target for early intervention, which is especially important in lung cancer research due to the lack of early stage biomarker.

As previously noted [Baccarelli, Barretta et al. 2011], although we matched subjects for smoking status and education and used multivariable models to control for potential confounders, we cannot exclude residual confounding from unmeasured variables. One limitation of the study is that we did not measure the differential blood counts. Also, the averaging of ambient PM₁₀ from many stationary monitors might not accurately estimate personal outdoor exposure to particles. Simulation studies have shown that the error introduced by using data from stationary monitors is highly unlikely to bias away from the null hypothesis, but rather is expected to underestimate the air pollution [Zeger, Thomas et al. 2000]. In addition, serial measures of ambient PM concentrations have been shown to be representative of variations in personal exposures [Wilson and Brauer 2006], particularly in the presence of high ambient PM levels [Avery, Mills et al. 2010].

In summary, our results indicate that tandem-repeat hypomethylation in SATα can be detected in an easily obtainable DNA source, such as blood leukocytes. Future studies are warranted to determine whether PM-induced SATα blood hypomethylation is associated with risk of lung cancer.