Particulate Matter, DNA Methylation in Nitric Oxide Synthase, and Childhood Respiratory Disease

Study population. This study was nested in the ongoing CHS (McConnell et al. 2006). We sampled children from the 5,341 kindergarteners and first-graders who were enrolled in the study in 2002. A subset of 940 non-Hispanic white and Hispanic white children who had buccal samples collected, genetic data available, and Fe_NO measurements collected were selected for DNA methylation analysis, as described previously (Breton et al. 2011).

Parents or legal guardians provided written informed consent for all study subjects. The Institutional Review Board of the University of Southern California approved this study.

Buccal sample collection and processing. Children were provided with two toothbrushes and instructed to brush their teeth with the first one. They were instructed to gently brush the buccal mucosa with the second toothbrush. The brush was then placed in a leak proof container that was filled with an alcohol-based fixative. Children then swished liquid throughout their mouths and expelled the fluid into a container. Most buccal-cell specimens were collected at school under the supervision of study staff. The remaining specimens were collected at home and sent to us by mail.

Buccal-cell suspensions were centrifuged at 2,000 × g on the day they were received in the laboratory. The pellets were stored frozen at –80°C until used for DNA extraction, at which time they were resuspended and incubated in 600 μL lysis solution from a PUREGENE DNA isolation kit (#D-5000; GENTRA, Minneapolis, MN) containing 100 μg/mL proteinase K overnight at 55°C. DNA extraction was performed according to manufacturer’s recommendations. The DNA samples were resuspended in the hydration solution (GENTRA) and stored at –80°C.

Selection of CpG methylation loci. We examined CpG loci located in NOS1, NOS2A, and NOS3 (Chan GC et al. 2005; Chan Y et al. 2004; Frain et al. 1989; Shi et al. 1998; Tarantini et al. 2009; Wei et al. 2002) (Figure 1). Two CpG loci in a nuclear hormone receptor (NHR) regulatory sequence in exon 2 of NOS1 were chosen for analysis, based on evidence that this NHR site contributes to the regulation of NOS1 transcription (Wei et al. 2002). Seven CpG sites in three regions within the NOS2A gene were selected for DNA methylation analysis. Two loci were chosen in a non-CpG island region (positions 1 and 2) of the promoter because the promoter was previously shown to be inversely related to iNOS (alias: NOS2A) mRNA expression (Chan GC et al. 2005). A third CpG site located in a non-CpG island between exon 1 and exon 2 (position 3) was chosen to correspond to a previously investigated site by Tarantini et al. (2009). The third region included four CpG sites (positions 4–7) in a locus located in a CpG island that was chosen because it spans transcription factor binding sites conserved in the mammal alignment (http://genome.ucsc.edu/) but whose methylation status had not previously been investigated. For NOS3, we examined two CpG loci located in the positive regulatory domain in the proximal promoter, which has a binding affinity with Sp1/Sp3, an important transcription factor involved in NOS3 transcription (Karantzoulis-Fegaras et al. 1999).

Polymerase chain reaction (PCR) primers targeting these loci were developed using MethPrimer software (Li and Dahiya 2002). Primers were designed not to overlap with any repeated elements and/or single-nucleotide polymorphism (SNP) sites and the specificity of the primer sequence was confirmed using in silico PCR [see Supplemental Material, Table S1 (https://doi.org/10.1289/ehp.1104439)].

DNA methylation. Laboratory personnel performing DNA methylation analysis were blinded to study subject information. Bisulfite conversion of 1 μg of genomic DNA extracted from buccal mucosal cells was performed with the EZ-96 DNA Methylation-Gold Kit™ (Zymo Research, Orange, CA), according to the manufacturer’s recommended protocol. Final elution was performed with 40 μL M-elution buffer. Bisulfite-converted DNA was stored at –70°C until further use. Methylation analyses were performed by bisulfite-PCR. Pyrosequencing assays were performed using the HotMaster Mix (Eppendorf, Hamburg, Germany) and the Pyrosequencing (PSQ) HS 96 Pyrosequencing System (Biotage AB, Uppsala, Sweden) (Jirtle and Skinner 2007) as described in previous work (Byun et al. 2009). Hct116 cell line DNA was used as control DNA and placed on each of 10 plates run on the PSQ. Percent coefficient of variation across the plates ranged from 2.1% to 5.5% for NOS2A and 1.1% for NOS3. The output from Pyrosequencing is reported as a percent of DNA methylation at each CpG locus. As a quality control check to estimate the bisulfite conversion efficiency, we placed duplicate genomic DNA samples on each bisulfite conversion plate to estimate the internal plate variation of bisulfite conversion and the Pyrosequencing reaction. Conversion efficiency was > 95%. We also added universal PCR products amplified from cell-line DNA on each Pyrosequencing plate to check the run-to-run and plate-to-plate variation in performing Pyrosequencing reactions. In addition, the pyrogram peak pattern from every sample was checked to confirm the quality of reaction.

Air pollutants. Air pollution data were obtained from central monitoring sites in each study community operated by local air pollution control agencies in conformance with U.S. Environmental Protection Agency (EPA) monitoring requirements (U.S. EPA 2012). Each community contained a single central-site monitor. At each monitoring site, 24-hr average measurements PM_2.5 were obtained daily or every third day. In addition, hourly PM_2.5 measurements were collected at selected community air monitoring sites. Continuous hourly average measurements were made for PM₁₀. When pollution data were not available for certain days, the gaps were filled by using data from nearby monitors provided that the monitors were not more than 7 km apart and the measurements from the monitors were reasonably well correlated (0.5 < r² < 0.95, depending on site and season) with each other. Daily 24-hr averages of PM_2.5 and PM₁₀ were extracted for the buccal cell collection date to calculate cumulative average exposure levels 7 days, 1 month, 6 months, and 1 year before the date of DNA collection.

Assessment of covariates. Race/ethnicity, physician diagnosis of asthma, current wheeze, allergy, annual family income, parental education, and exposure to in utero and secondhand tobacco smoke (SHS) were obtained through annual written questionnaires completed by the parents. Children were considered to have asthma if their parents reported on the questionnaire that their child had ever been physician-diagnosed; questionnaires were collected at the school visit during which the child’s buccal cell sample was collected. Children were considered to have current wheeze if their parents reported on the same questionnaire that their child had wheezed in the previous 12 months. History of respiratory allergy was assessed by whether the child had a problem with sneezing or runny nose when he or she did not have a cold or the flu. Height and weight were measured on the day of collection. Age- and sex-specific percentiles based on the Centers for Disease Control and Prevention (2011) body mass index (BMI) growth charts were used to categorize BMI.

Statistical analysis. Descriptive analyses were first conducted to examine the distribution of DNA methylation in NOS genes and air pollutants by subject characteristics. Spearman correlations were used to study the pairwise correlations of percent methylation between different CpGs in the same gene. For genes with multiple CpG loci measured, percent DNA methylation at individual loci as well as average percent DNA methylation were analyzed. For the NOS2A CpG island, only the average of the percent methylation of four loci was evaluated.

To investigate the association between air pollutants and percent DNA methylation, we fitted beta regression models adjusted for age, sex, race/ethnicity, experimental plate (for Pyrosequencing reactions), town of residence, month of DNA collection, and parental education. Additional adjustment for asthma, allergy, and wheeze were evaluated, because presence of disease may affect methylation status. Asthma status was retained a priori in the model; however, further adjustment for allergy and wheeze status did not change the regression results and were dropped from final models. Beta regression was used to address the non-normal distribution of DNA methylation values, which are bounded by 0 and 1 and in many cases heavily skewed toward one end or the other (Ferrari and Cribari-Neto 2004). Beta regression requires the data to be between zero and one, therefore a shrinkage method was applied to force the zeros to be positive as follows: 0.999999 × (meth value – 0.5) + 0.5. Because beta regression is modeled on the logit scale, we transformed the results to a linear scale centered around the mean values for covariates in order to interpret the results as an absolute difference in percent methylation in the average individual for a given unit change in pollutant exposure. We scaled the results to a change in methylation per 5-µg/m³ increase in PM_2.5 exposure because this was very close to the interquartile range in the population (4.8 µg/m³). One subject with an extremely high methylation value for NOS2A CpG island average was considered an outlier and removed from analyses.

We also used beta regression models to model differences in methylation levels by asthma and wheeze status in this population. A priori variables included for adjustment were age, sex, race/ethnicity, experimental plate, allergy, and town. Within asthmatics, stratified analyses by medication use, which can be used as an indicator of severity of asthma, were also conducted. For analyses restricted to asthmatics (n = 133), plate and town were dropped from models given the small sample size and large degrees of freedom required by these variables.

Sensitivity analyses were also conducted to evaluate whether the inclusion of children with asthma or wheeze influenced our results. Thus, the analyses of air pollutants and DNA methylation were rerun in a population in which children with asthma and wheeze were removed (n = 704).

All tests assumed a two-sided alternative hypothesis, a 0.05 significance level, and were conducted using SAS/STAT® software, version 9.2 (SAS Institute Inc., Cary, NC) and in the R programming language, version R2.12.2 (R Project for Statistical Computing, Vienna, Austria).

On average, children were 9 years old, two-thirds were of Hispanic white ethnicity, 14% had been diagnosed with asthma by the time of buccal cell collection, 11% had current wheezing reported by their parents, and 56% had a parentally-reported history of respiratory allergies (Table 1). Cumulative average PM_2.5 and PM₁₀ were right-skewed, median PM_2.5 ranged from 11.5 to 15.8 µg/m³, and median PM₁₀ ranged from 27.0 to 33.8 µg/m³ (Figure 2). Distributions of cumulative averages are shown in Supplemental Material, Table S2 (https://doi.org/10.1289/ehp.1104439), and the correlations for PM_2.5 and PM₁₀ are shown in Supplemental Material, Table S3.

Distribution of percent methylation of NOS loci is shown in Table 2. In general, the non-CpG island loci in NOS1, NOS2A, and NOS3 were methylated > 75%. The proportion of methylation at the CpG locus in NOS2A position 3 (located in a non-CpG island between exon 1 and 2) was 51%, whereas CpG island loci in NOS2A were largely unmethylated (< 2.5%). Within each gene, the multiple CpG loci measured were not strongly correlated, with correlation coefficients < 0.5 for all but NOS1 positions 1 and 3 [see Supplemental Material, Tables S4 and S5 (https://doi.org/10.1289/ehp.1104439)].

PM_2.5 was negatively associated with percent DNA methylation in the non-CpG island regions of NOS2A (Table 3). In beta regression models, cumulative average PM_2.5 exposure windows were significantly associated with lower percent methylation in positions 1 and 2, although the associations for position 1 had smaller p-values than for other positions. A 5-µg/m³ increase in acute (7 day, 1 month) and long-term (6 months, 1 year) cumulative averages of PM_2.5 was associated with a 0.2% [95% confidence interval (CI): –0.32, –0.07] to 1% (95% CI: –1.61, –0.56) lower DNA methylation at position 1. In addition, cumulative averages of PM_2.5 were associated with lower methylation of NOS2A at position 2 for 1-month and 1-year averages and at position 3 for 7-day and 6-month averages. In contrast, PM_2.5 exposure was associated with small but significantly higher in percent DNA methylation of the average of positions 4–7 in the NOS2A CpG island region (Table 3).

Because these analyses were conducted in all children, including those with a diagnosis of asthma or wheeze, we conducted a sensitivity analysis in which we estimated the associations between PM_2.5 and NOS2A methylation in a population restricted to children without diagnosis of asthma or wheeze (n = 720). Associations were consistent with those for all children combined [see Supplemental Material, Table S6 (https://doi.org/10.1289/ehp.1104439)].

PM_2.5 was also positively associated with percent DNA methylation in both positions measured in NOS3. A 5-µg/m³ increase in acute (7 day) and long-term (1 year) cumulative average PM_2.5 levels was associated with 0.6% (95% CI: 0.13, 0.99) to 2.8% (95% CI: 1.77, 3.75) higher absolute level of percent DNA methylation in the two loci (Table 3). PM₁₀ showed associations with DNA methylation in the NOS genes that were smaller in magnitude compared with PM_2.5 and often lacked statistical significance [see Supplemental Material, Table S7 (https://doi.org/10.1289/ehp.1104439)].

DNA methylation in NOS2A was investigated with respect to current childhood wheeze. Children with current wheeze had 0.21% (95% CI: –0.22, 0.64) higher methylation in NOS2A position 1, 0.39% (95% CI: –0.84, 0.07) lower methylation in the NOS2A position 2 and 0.25% (95% CI: –1.07, 0.58) lower methylation in the NOS2A position 3 compared with children without wheeze. No direct association between DNA methylation and asthma overall was observed. However, among 133 children with asthma, children with current wheeze only (n = 54) and children taking asthma medications (n = 66) had lower DNA methylation in the NOS2A promoter region than did asthmatic children without a recent history of wheeze or medication use, respectively. Children using asthma medication had 0.49% (95% CI: –1.42, 0.45), 1.72% (95% CI: –2.77, –0.67), and 1.37% (95% CI: –2.79, 0.06) lower DNA methylation in NOS2A positions 1, 2, and 3, respectively, compared with asthmatic children not taking medication. Asthmatic children with current wheeze had 0.28% (95% CI: –1.22, 0.66), 0.90% (95% CI: –2.01, 0.20), and 0.93% (95% CI: –2.39, 0.54) lower DNA methylation in NOS2A positions 1, 2, and 3, respectively, compared with asthmatic children without wheeze.

Of the pollutants evaluated, PM_2.5 was consistently and significantly associated with DNA methylation levels in NOS genes, primarily NOS2A and NOS3. PM₁₀ showed similar associations with DNA methylation in these genes as compared with PM_2.5 but the magnitude of association was weaker. No statistically significant associations were observed for NOS1 (results not shown).

In this study, PM_2.5 was associated with differences in level of percent DNA methylation in NOS genes, suggesting that NO homeostasis may be influenced by air pollution. NOS DNA methylation was also associated with wheeze and medication use, a proxy for active disease, in children with asthma.

Particulate matter is associated with increased expression of iNOS (Ulrich et al. 2002). A decrease in DNA methylation in the promoter of NOS2A—the gene encoding iNOS—that increases transcription of the gene is one hypothesis to explain such an observed association. A recent study by Tarantini et al. provided preliminary evidence in support of such a theory (Baccarelli et al. 2009; Tarantini et al. 2009). In that study, NOS2A DNA methylation level decreased after a 3-day shift in a steel mill; however, methylation was not associated with measured PM₁₀ levels in the workers (Baccarelli et al. 2009; Tarantini et al. 2009). In the present study, we estimated short- and long-term effects of both PM₁₀ and PM_2.5 on DNA methylation of several CpG loci in promoter and nonpromoter regions of NOS2A. PM_2.5 was more strongly associated with lower NOS2A promoter methylation than PM₁₀. We observed a significant association of 7-day cumulative average PM_2.5 with percent DNA methylation in the same locus that Tarantini et al. (2009) evaluated (NOS2A Position 3), although stronger associations were observed for short and long-term cumulative averages at the two NOS2A promoter CpG loci more proximal to the transcription start site.

Interestingly, although PM_2.5 was associated with lower methylation in the NOS2A promoter, it was also associated with higher methylation in the CpG island in the body of the gene. Regulation of iNOS activity is complex, and the inducibility of iNOS by cytokines is known to involve DNA methylation, histone H3 lysine 9 methylation, and MeCP2 binding at the promoter (Chan GC et al. 2005). Our observation that PM_2.5 was associated with higher DNA methylation in the promoter suggests that iNOS expression might also have been increased in association with PM_2.5 exposure. However, little is known about the functionality of DNA methylation in the downstream CpG island, and recent evidence suggests that methylation in CpG islands can in some cases be associated with increased expression rather than decreased expression (Bogdanovic et al. 2011).

Regulation of NOS3 is similar to that of NOS2A in that increased DNA methylation in the promoter is associated with decreased promoter activity (Chan Y et al. 2004). However, the association between PM_2.5 and NOS3 methylation that we observed was opposite the association between PM_2.5 and the NOS2A promoter. Air pollution was associated with higher DNA methylation in the NOS3 promoter, which should lead to reduced transcriptional activity and lower NO production. This is consistent with evidence that cigarette smoke extract reduces NOS3 activity, protein, and mRNA levels in pulmonary artery endothelial cells (Su et al. 1998). Moreover, Ten Broeke et al. (2006) observed that while iNOS expression was associated with increased Fe_NO in a mouse model of asthma, eNOS overexpression attenuated airway inflammation and hyperresponsiveness in allergic asthma. PM_2.5-induced increases in DNA methylation of NOS3 may contribute to exacerbation of inflammation and hyperresponsiveness via this mechanism. The pathophysiology of asthma may be determined by a delicate balancing of the production of NO by NOS isoforms. This theory is further supported by experimental evidence that demonstrated a simultaneous increase in iNOS expression and decrease in eNOS expression in response to endotoxin challenge in intact rat lungs (Ermert et al. 2002).

Among all children, children with wheeze had lower levels of DNA methylation in two of the three loci in the NOS2A promoter than did children without wheeze. Among children with asthma, NOS2A methylation was lower among those with active wheeze and those currently taking medication compared with asthmatic children without wheeze or medication use, respectively. One explanation for these results is that NO production may be up-regulated via an epigenetic mechanism in allergic airway disease. This explanation would be consistent with findings previously reported on DNA methylation in ARG1 and ARG2 genes, also key players in NO homeostasis, in relation to exhaled nitric oxide (Breton et al. 2011). However, these results are based on a small sample size and the conclusions should be interpreted with caution.

Given that PM_2.5 was associated with lower DNA methylation in the NOS2A promoter, epigenetic regulation of NOS2A may provide one biological mechanism by which PM_2.5 affects respiratory health outcomes in children. Epigenetic regulation of other genes has also been implicated in association with air pollution and asthma. Children in a community with high levels of ambient air pollution (Fresno, CA) had increased methylation of the FOXP3 locus, a locus important in Treg-cell function and asthma morbidity, compared with children residing in a low-pollution community (Palo Alto, CA) (Nadeau et al. 2010). In a cohort of children living in New York City, exposure to PAHs was associated with DNA methylation in the ACSL3 gene, a gene associated with asthma symptoms in children (Perera et al. 2009). Emerging evidence suggests a link between air pollution exposures, alterations in DNA methylation levels and downstream health outcomes.

Although associations between air pollution and iNOS methylation have been observed across studies, the magnitudes of associations for single CpG methylation loci are small and thus their biological relevance is unclear. Nevertheless, we previously reported that in children with the highest 10th percentile of iNOS methylation (> 56.6%), PM_2.5 exposure was significantly associated with higher Fe_NO levels, whereas no association was seen at lower methylation levels (Salam et al. 2012). NOS2A genetic and epigenetic variations and short-term PM_2.5 acted synergistically in association with Fe_NO levels in our population, suggesting that a small change in DNA methylation at one CpG locus may have functional significance. However, regulation of Fe_NO is complex and future studies warrant the joint evaluation of genetic and epigenetic variations and air pollution exposure on phenotype expression.

We have observed particulate matter-associated differences in DNA methylation in buccal cells. Our associations are similar in direction of effect and magnitude to those of Tarantini et al. (2009) despite the fact that different cell types were used. However, the implications of DNA methylation changes for health outcomes are far from clear and are likely to differ by cell type and disease pathophysiology. For example, Chan GC et al. (2005) demonstrated variable levels of inducibility of iNOS dependent on cell type investigated. Ideally, methylation changes in NOS genes should be evaluated in the most appropriate tissue for the outcome of interest in order to study functional consequences.

In evaluating childhood wheeze and asthma, buccal mucosal cells—an aerodigestive tract epithelium—were used for DNA methylation analysis. Previous studies have demonstrated that buccal epithelium can serve as a surrogate tissue for the lung when measuring DNA methylation (Bhutani et al. 2008). In addition, studies comparing expression profiles in buccal, bronchial, and nasal epithelial cells in response to tobacco smoke have demonstrated striking similarities, providing further support for use of buccal cells as a useful surrogate for respiratory tract cells (Boyle et al. 2010; Sridhar et al. 2008). However, buccal cells are a mixed population of epithelial cells and leukocytes. Thus, a shift in cell populations caused by air pollutant exposure, rather than by differences in methylation, could explain our observed results.

Buccal-cell DNA was collected after diagnosis of asthma and wheeze, so we cannot address whether DNA methylation in NOS genes precedes or is the result of childhood respiratory disease. Similarly, we cannot rule out the possibility that asthma medications may have altered DNA methylation in the asthmatic children. However, the observed trends showing that children with wheeze had intermediate levels of DNA methylation, compared with healthy children and children with active asthma, support the conclusion that DNA methylation levels in NOS genes may play a role in asthma pathogenesis. Further investigation in a longitudinal study design that collects DNA samples at or near birth would clarify these lingering concerns and is currently underway.

We evaluated four cumulative average exposure levels of PM_2.5 and PM₁₀ before the date of DNA collection. Our choice to evaluate short-term cumulative averages was based on previous evidence of an association between cumulative pollutant exposure and Fe_NO, the product of the genetic pathway of interest (Berhane et al. 2011), and on the knowledge that the half-life of buccal cells is only a few weeks. However, long-term exposure effects on DNA methylation may also be possible in buccal cells if exposure affects the basal or progenitor cells from which the surface epithelium is formed, or if methylation marks, once altered, are heritably passed on from one generation of cells to the next. The shorter cumulative averages are more likely to have increased measurement error in exposure assessment which may explain the weaker statistical results. The chronic exposure estimates are more robust and representative of typical and steady state exposures.

Twelve distinct CpG loci in three NOS genes, and two air pollutants, were evaluated in this study. Given the multiple tests that were conducted, some associations may be attributable to chance. Last, parental report of physician-diagnosed asthma was used in our study, and concern has been raised that parental report might not reflect physician diagnosis. To investigate this potential bias, we reviewed medical records of 172 children with asthma and found that 95.9% had a definite or probable diagnosis of asthma (Salam et al. 2007).

In this study, we provide evidence that DNA methylation levels in NOS genes vary by PM_2.5 exposure levels and possibly wheeze during childhood. These results add to a small but growing body of literature suggesting epigenetic regulation of certain genes as potentially important biological mechanisms underlying the health effects of air pollution exposures.