Effect of GSTM2-5polymorphisms in relation to tobacco smoke exposures on lung function growth: a birth cohort study

Pregnancy is a critical period of lung growth and development, making the fetus susceptible to environmental exposures [1]. Indeed, in utero exposure to tobacco smoke has shown detrimental effects on lung health [1‐4]. In utero tobacco smoke exposure may affect morphogenesis and maturation of the lungs, leading to impaired lung health by adulthood [2]. The toxic effect of tobacco smoke exposure might be reduced through metabolizing xenobiotic products, thus minimizing oxidative stress, a physiological event caused by the imbalance between reactive oxygen species and the body’s ability to detoxify these products [5]. However, if detoxification is inefficient or absent within the mother or the child due to genetic polymorphisms, the exposed child may have a decreased ability to remove reactive oxygen species properly. Furthermore, effects of prenatal and early childhood exposure to tobacco smoke may persist through time, thus leading to reduced lung function growth even in adolescence. Additional exposures during adolescence, such as personal smoking, may further affect the growth trajectory in lung function in vulnerable individuals as well; therefore, it is important to control for these factors as well in order to determine the pure joint effects of tobacco smoke exposures and genetic polymorphisms responsible for detoxification.

The glutathione S-transferases (GST) are a superfamily of catalytic proteins responsible for detoxification of xenobiotic compounds, in conjugation with glutathione. There are seven classes of the GST superfamily: alpha, mu, pi, sigma, theta, omega, and zeta [5]. While the effects of glutathione S-transferase mu 1 (GSTM1) deletion have been extensively studied, results of studies examining these effects on various pulmonary outcomes are in conflict [6‐11]. For instance, findings in a study among German schoolchildren aged 9 to 11 years only found a significant interaction between GSTM1 status and in utero smoke exposure on maximum mid expiratory flow (MMEF) levels [10], however Henderson et al. while confirming a detrimental effect of intrauterine tobacco smoke exposure on childhood lung function found no strong evidence of modification by maternal or child GSTM1 genotype [12]. Regarding growth, only one study found an association between GSTM1 gene function and lung function growth, where children with the GSTM1 deletion had slower lung function growth compared to those with the normal genotype [6]. In addition to copy number variation of the GSTM1gene, genetic variation in the form of single nucleotide polymorphisms (SNPs) within other members of the GST family also needs to be considered as it may lead to reduced expression of enzymes that detoxify harmful products that impact lung function growth.

One explanation for conflicting results of studies of the GSTM1 deletion is the failure to consider the impact of SNP variation in the adjacent GSTM2-5 loci. This is highlighted by the study of Breton et al. where GSTM2 was associated with growth in FEV₁ and MMEF; GSTM4 was associated with FEV₁, FVC, and MMEF; GSTM3 and GSTM5 were associated with MMEF; and in utero smoke exposure modified the effect of GSTM2 haplotypes on growth in FEV₁ and FVC [13]. Because of these independent effects of variation within the GSTM2-5 loci on lung health, it is important to take this into account when examining the effect of GSTM1 variation on lung function growth in adolescence.

An alternative explanation for variation between studies in genetic epidemiology is variation in environmental exposure between cohorts [14]. Effects of genetic polymorphisms may be only seen in exposed populations, or vice versa, and in some circumstances, ‘flip-flop’ effects may be present where the effect of alleles on disease outcome is reversed depending on environmental exposure. Environmental exposure may also result in epigenetic effects such as altered DNA methylation that may either mask or synergize with the effects of germline genetic variation on phenotype [15]. For example in utero tobacco smoke exposure has been shown to result in changes in DNA methylation in cord blood DNA [16] that persist until adulthood (unpublished observations).

Thus far, only one study has reported the effects of genetic variation found in GSTM2-5 and their interaction effect with tobacco smoke exposure on lung function levels and growth in late adolescence, where in utero smoke exposure modified the effect of GSTM2 on lung function growth [13]. Since active smoking only had marginally significant effects on FEV₁ and FEV₁/FVC at age 18 in the Isle of Wight (IOW) cohort (unpublished observations), the effect may only be detected in those with a genetic susceptibility, acting through an epigenetic mechanism. To expand these findings, we analyzed data collected in the IOW birth cohort study. Specifically, we explore the interactive relationships between smoking status and individual GSTM diplotypes. Furthermore, we validated whether or not smoke exposures modified the effect of diplotypes on methylation levels of CpG sites within this gene cluster and determined whether or not methylation levels of these CpG sites affected lung function levels at age 18.

Data on lung function measures were available for 1,121 children (Table 1). No significant differences were found between this subsample and the total cohort (n = 1,456). All SNPs in the GSTM2-5 loci were in HWE and had a MAF ≥ 5% (Table 2). SNPs within each gene formed their own blocks based on linkage disequilibrium (LD) measures, where the r² value was 0.8 or greater. A total of four blocks were generated (Additional file 1: Figure S2). The each block was comprised of the following SNPs: GSTM2 rs574344 and rs12024479; GSTM3 rs1537236, rs7483, and rs7537234; GSTM4 rs668413, rs560018, and rs506008; and GSTM5 rs929166 and rs11807; (Additional file 1: Table S1). All reported p-values were adjusted for multiple testing for an FDR of 0.05, when applicable. Only GSTM3 and GSTM5 show direct associations with gain in FEV1 or FVC (Table 3). Statistical analyses of the individual effects of SNPs within the GSTM2-5 cluster on gain in each lung function outcome yielded no significant findings (Additional file 2: Table S2). When looking at main effects of GSTM2-5 diplotypes by age, only GSTM5 diplotypes contributed to FEV₁ levels at age 10 (Additional file 3: Table S3a). Several interactions between GSTM2-5 diplotypes and smoke exposures were detected, but only at age 18. SHS modified the relationship between GSTM2 and FEV1 and FVC levels (p_interaction = 0.004 and p_interaction = 0.0003, respectively) (Additional file 3: Table S4b); SHS modified the relationship between GSTM3 and FEV1 levels (p_interaction = 0.04) (Additional file 3: Additional file 3: Table S5b; and active smoking modified the relationship between GSTM5 and FEV1 and FVC levels (p_interaction = 0.004 and p_interaction = 0.05, respectively) (Additional file 3: Table S7c).

Results suggest that variation within the GSTM3 and GSTM5 loci had significant impact on lung function outcomes at age 18 after adjustment for confounding. Some diplotypes, but none of their interactions with smoke exposure (in utero, SHS, or active smoking exposure at 18), produced an independent, statistically significant relationship with FEV₁ and FVC, but not with FEV₁/FVC. Additional analyses suggest a multi-stage model. First, DNA methylation was affected by diplotypes conditionally on smoking exposure; second, an altered DNA methylation modified the effect of diplotypes on lung function (acting as modGVs), leading to significant effects of GSTM2 and GSTM5 on FEV₁ and FVC. Path analyses show that methylation at cg06970744 did not lie on the pathway between GSTM5 diplotypes and lung function, further providing evidence that methylation at certain sites can modify the effect of genetic variation on an outcome. Also, methylation at this site appears to be unaffected by nearby SNPs (unpublished observations).

Several GSTM3 diplotypes had strong, negative effects on gain in lung function. SNPs in GSTM3 included a functional SNP (rs7483). Previous studies have found an association with this SNP and Alzheimer’s disease [29, 30]. Breton et al. have previously examined the association of rs7483 with lung function, and similar to the present study, no significant associations were found [13]. GSTM4 had a synonymous coding SNP (rs506008). This SNP produced positive, but insignificant effects on lung function across time and thus more than likely does not make contributions to lung function at the diplotype level. There are no reported associations between rs506008 and any disease outcomes. Although GSTM5 had no functional SNPs, there was a significant lung function improvement in those who possessed the AA_CA diplotype, demonstrating a need to investigate this particular region. Only one of the SNPs (rs11807) was previously reported in the literature, showing a strong association with hypertension [31]; however, no association with lung function had been reported.

Due to previous findings that showed lower lung function outcomes due to various tobacco smoke exposures [2‐4, 32], we assessed the critical period where lung development may be severely impaired by tobacco smoke exposure. No two-way interaction effects of tobacco smoke with GSTM2-5 loci were seen, which is in contrast to findings by Breton et al., who reported interaction effects with GSTM2 and in utero tobacco smoke [13]. This discrepancy may be either due to misclassification of the exposure and/or insufficient sample size in our study or missing replication in the study by Breton et al.[13]. Because questions involving tobacco smoke exposure at age 18 were validated with urinary cotinine levels (Additional file 3: Table S9), there is no suggestion of major misclassification, as those who were active smokers or were exposed to SHS had significantly higher levels of cotinine compared to nonsmokers. The study by Breton et al. included at least one more repeated measurement [13], hence, it is possible that the present study does not have equal statistical power to detect interactions. However, there is a need of replication studies to examine the role of these genes on lung function in conjunction with tobacco smoke exposure, especially in different environments. It is also necessary to consider that other environmental and lifestyle exposures including air pollution and paracetamol (acetaminophen) use [33] may also alter oxidative stress [34] and mask or falsely indicate an effect of related exposures. Air pollution was not controlled in the study by Breton et al.[13], but was reported to modify the effect of a similarly functioning gene (GSS) on lung function growth in another study of this group [33]. Because it has been suggested in the literature that tobacco smoke exposure, especially in utero[35], may epigenetically modify these genes, hence changing the function of these genes, we tested the effect of CpG sites on lung function levels at age 18. Although there were no independent effects of these CpG site methylation levels on lung function outcomes, it is interesting to note that these effects were not seen until joint effects of GSTM2 and GSTM5 diplotypes were taken into account, indicating that DNA methylation may modify the effect of genetic variants on lung function, a mechanism that needs further investigation [15]. These findings may also help explain the lack of consistency between the present study’s findings and the findings of Breton et al.[13]. It may not be the genotype that produces inconsistent results, but rather different DNA methylation levels in different study groups that accounted for the discrepancy.

To further lend validity to this study, selection bias was not apparent regarding availability of lung function data. Approximately 95% (1456/1536) of mother-child pairs were enrolled into this study; and those who underwent pulmonary function tests were not significantly different from the total cohort (Table 1). Also, lung function measurements were obtained under standardized conditions in a prospective manner, decreasing the likelihood of information bias. With respect to genotyping data, all SNPs that were shared with Breton et al. were in HWE and had comparable frequencies with Caucasians in their sample [13]. These polymorphisms also agreed substantially in their association with FEV₁ and FVC with our findings (Additional file 1: Figure S2 and Additional file 2: Figure S3). In regard to haplotypes, diplotypes (haplotype-pairs) were used, thus reducing uncertainty and subsequently misclassification of individuals, which is often encountered in haplotype association studies [36]. Also, our population was homogeneous, meaning controlling for population stratification is unnecessary and ensures HWE, which appeared to be an issue for Breton et al.[13], possibly resulting in different findings. While SNPs included in the haplotype analysis of the latter study were in HWE within ethnic groups, they were not in HWE when examining the entire population [13]. This is problematic because inclusion of SNPs that deviate from HWE may produce spurious associations between haplotypes and lung function [37, 38]. In addition, the findings by Breton et al.[13] may be due to population stratification. The authors identified ancestry indicators, controlled these as confounders but did not stratify their analysis by these markers. Nevertheless, population stratification addresses the possibility that haplotypes may reveal different associations in different ethnic/racial strata. This may also account for the lack of agreement of single SNP effects on lung function outcomes between the present study and the Breton et al. study [13].

Despite the strengths of our study, some limitations are present. First, because maternal smoking and active smoking during adolescence is obtained via self-reported questionnaires, misclassification of this exposure is possible. Our previous publications have shown that maternal smoking during pregnancy interacts with the IL13 and IL1RN genes and increases the risk of asthma and wheezing [39, 40]. Hence, ascertainment of maternal smoking seemed to provide valid information. However, maternal smoking × IL13 and IL1RN gene interaction assumes an effect of smoking on these genes. Against that, GSTM2-5 × maternal smoking interactions assume that the genes regulate the toxicity of tobacco smoke, possibly via epigenetic mechanisms. Also, based on results of the methylation analyses in the present study, the joint effect of maternal smoking and GSTM2-5 diplotypes did not influence methylation levels of CpG sites within the GSTM2-5 cluster; however, these methylation levels were captured at age 18 and therefore we cannot ascertain whether or not epigenetic changes had taken place due to this particular exposure nor can we determine that methylation levels at age 18 are representative of early-life methylation profiles. Regarding validation of other smoke exposures, passive smoke exposure and active smoking at age 18 was associated with increased cotinine levels. Also, active smoking at age 18 led to deficits in FEV₁/FVC after adjustments for GSTM5 diplotype, GSTM1 genotype, sex, BMI, height, in utero smoke exposure, and SHS (Additional file 3: Table S16). Reduced FEV₁/FVC has previously been observed in adolescent smokers [41].

Second, not all SNPs used in the Breton et al. study were genotyped in this work. Therefore combinations of SNPs that were not included in the haplotype construction may lead to different results. Variation in FEV₁ in these results may be due to differences in population characteristics between the two studies, residual confounding, or as demonstrated by methylation levels, differences in expression levels of detoxification enzymes produced by this set of genes. Based on the results of this study, methylation levels at age 18 within this gene cluster can be ruled out because analyses revealed no significant effect of methylation on lung function levels at 18. The goal of this study was to expand upon findings by Breton et al. through accounting for methylation [13].

Based on results of this study, it is clear that certain genes within the GSTM2-5 loci explain some differences in lung function in late adolescence; however, joint effects with tobacco smoke exposure were not detected. This finding was validated through examination of joint effects of diplotypes and in utero smoke exposure, which did not significantly alter methylation levels within this gene cluster. Only one study has reported an interaction of maternal smoking with GSTM2-5 genes [13]. A past genetic association study has examined the interactive effects of maternal genotypes on other GST genes [12], but no studies have examined the joint effect of maternal genotype and GSTM2-5 genotypes in their offspring. Future studies are necessary to elucidate conflicting results, taking into account DNA methylation ×gene interactions, interaction with other xenobiotics affecting the glutathione levels, and interaction with the maternal genome.