Assessment of Interactions between PAH Exposure and Genetic Polymorphisms on PAH-DNA Adducts in African American, Dominican, and Caucasian Mothers and Newborns

Polycyclic aromatic hydrocarbons (PAH) are widespread pollutants commonly found in air, food, and drinking water (1). Airborne PAHs (outdoor or indoor) mainly result from combustion of fossil fuels, tobacco, and other organic materials. Generally, emissions from motor vehicles, electricity generation, and residential heating are the major source of PAHs in outdoor urban air, whereas environmental tobacco smoke (ETS) is a major indoor source (2, 3). Benzo[a]pyrene (B[a]P) is a well-studied representative PAH found in environmental mixtures, including air pollution (4). B[a]P is a well-known human carcinogen and has been found to be a transplacental carcinogen in experimental bioassays and produces tumors in the liver, lung, lymphatic tissues, and nervous system of the offspring (5-7).

PAHs bind covalently to DNA to form DNA adducts. As an indicator of DNA damage, carcinogen-DNA adducts represent a critical step in the carcinogenic pathways and thus can be considered an informative biomarker of cancer risk. Associations between PAH-DNA adduct levels and cancer risks have been reported in both experimental and epidemiologic research (8-14). PAH-DNA adducts have also been implicated in neurodevelopmental deficits in children with exposure. Both experimental and epidemiologic studies have shown that fetal levels of PAH-DNA adducts were generally higher than expected, per estimated transplacental dose of PAHs, suggesting increased susceptibility to deleterious outcomes (15-19). The adverse outcomes linked previously to PAH exposure or PAH-DNA adducts in the cohorts under study in the present report include disorders in fetal growth and child development, childhood asthma, and cancer (20-25).

CYP1A1 and CYP1B1 (26, 27) have been shown to play important roles in the metabolic activation of PAHs, whereas PAH detoxification is partially affected by the glutathione S-transferase (GST) genes GSTM1, and GSTT2. Figure 1 displays a diagram that shows the mechanistic relationship between the genes and the formation of the adducts. There are conflicting data on the contribution of genetic variants in these genes and susceptibility to lung cancer partly due to underpowered studies (13-19, 26-32). Still, common genetic polymorphisms in these genes could affect individual susceptibility to adverse effects of environmental air pollution. Previous studies have not explored possible gene-environment interactions in mothers and their newborns in underrepresented groups, such as African Americans and Dominicans. The purpose of this study was to explore gene-environment interactions by assessing the association of DNA adducts with airborne PAH exposure levels among mothers and their newborns in three different ethnic groups with different genotypes (common genetic variants in CYP1A1, CYP1B1, GSTM1, and GSTT2). Only when no gene-environment interaction was observed did we examine possible main effects of genotypes on DNA adducts.

Subjects were chosen from two independent, parallel studies. One currently is being conducted in New York City (NYC) and the details on the study design have been published previously (20). Study subjects included nonsmoking African American and Dominican women residing in Washington Heights, Central Harlem, NYC, and the South Bronx, who were recruited through the obstetrical services of New York Presbyterian Hospital, Harlem Hospital, or satellite clinics between February 1998 and February 2003. Ethnicity was self-identified. The institutional review board of New York Presbyterian Medical Center approved the study, and informed consent was obtained from all study participants. The second cohort study is being conducted in Krakow, Poland, a region known to have higher levels of air pollution than NYC presumably due to coal burning and vehicle emission (21). Details on study design for the Krakow cohort have been published previously (21). Nonsmoking, pregnant women residing in the Srodmiescie and the Krowodrza-Nowa Huta areas were recruited between November 2000 and March 2003 (21). Informed consent was obtained from all subjects, and the study was approved by the ethics committee of the Jagiellonian University.

In both cohorts, pregnant women were eligible if they were not currently smoking, registered at prenatal health care clinics, had lived at the present address for at least a year before the initial interview, were ages ≥18 years, had no history of illicit drug use, pregnancy-related diabetes, or hypertension, and had a valid estimate of gestational age. During the second or third trimesters pregnancy, the women carried a backpack containing a portable personal exposure air monitor during the day and kept it near the bed at night during a consecutive 48-h period for PAH measurements (20, 21).

Subjects included in the present analysis are mothers with genotype data (n = 160 African Americans, n = 265 Dominicans, and n = 381 Polish Caucasians) and their newborns with genotype data (n = 99 African Americans, n = 154 Dominicans, and n = 294 Polish Caucasians; Table 1). The subset did not differ significantly from the total population with respect to selected demographic and exposure characteristics.

Maternal blood (30-35 mL) was collected within 1 day postpartum, and umbilical cord blood (30-60 mL) was collected at delivery (33). B[a]P-DNA adducts were analyzed in all maternal and umbilical cord blood samples with sufficient DNA quantities. B[a]P is widely used as a representative PAH because concentrations of individual PAHs in the urban setting are highly correlated (33). B[a]P-DNA adducts can serve as a proxy for PAH-DNA adducts (34). B[a]P-DNA adducts in extracted WBC DNA were analyzed using the high-performance liquid chromatography-fluorescence method of Alexandrov et al. (35), which detects B[a]P tetraols. The method has a coefficient of variation of 12% and a lower limit of detection of 0.25 adducts per 10⁸ nucleotides. As in prior analyses, samples below the limit of detection were assigned a value midway between the limit of detection and zero (0.125 adducts per 10⁸ nucleotides).

Sixteen single nucleotide polymorphisms (SNP) from CYP1A1, CYP1B1, and GSTT2 were selected from the SNP500Cancer resource (36). For the first two genes, haplotype-tagging SNPs were selected based on the genomic analysis of these genes (7 in CYP1A1 and 6 in CYP1B1) by the Breast and Prostate Cancer Cohort Consortium (37) using the approach of Stram et al. (38). Three SNPs were chosen from GSTT2, which cover less than 50% of the common genetic variation in this gene. A real-time TaqMan assay for homozygous or heterozygous deletions of GSTM1 was done (39).

For quality-control assessment, results on individual genetic markers with genotyping rates of <75% were removed (including 3 African American mothers, 25 African American newborns, 1 Dominican mother, 56 Dominican newborns, 14 Polish Caucasian mothers, and 82 Polish Caucasian newborns). All tested genetic variants had minor allele (defined as the less common allele in the cohort) frequencies greater than 0.05. A goodness-of-fit test to test for Hardy-Weinberg equilibrium indicated that there were no significant deviations in each population.

As described previously (33), the eight PAH air concentration measures: benz[a]anthracene, chrysene, benzo[b]fluroanthene, benzo[k]fluroanthene, B[a]P, indeno[1,2,3-cd]pyrene, disbenz[a,h]anthracene, and benzo[g,h,i]perylene (40-42) were significantly correlated; thus, a composite PAH variable was computed. This summed measure was dichotomized at the median (2.45 ng/m³ for NYC African Americans, 2.3 ng/m³ for NYC Dominicans, and 16.89 ng/m³ for Polish Caucasians) to obtain a binary PAH exposure, defined as PAH high or PAH low. We first assessed genetic marker × PAH interactions on PAH-DNA adducts. The adduct data were logarithmic transformed to be consistent with the previous studies and to better normalize the distribution. We used multiple linear regression models:

Although ETS was not significantly associated with DNA adducts, it was included in the model as a potential confounder. Other covariates, such as maternal age, dietary PAH, and maternal BMI, were not significantly associated with DNA adducts at the significance level of P ≤ 0.05 and were not included in the model. For genetic markers that did not show significant interactions (P ≤ 0.05), main effects of genotype on adducts were examined, also with multiple linear regression models:

Because the study was limited by the small sample size, we did not routinely correct for multiple comparisons given the fact we tested many interaction effects. Therefore, the findings are suggestive only.

We analyzed linkage disequilibrium (LD) patterns and haplotype structures of CYP1A1, CYP1B1, and GSTT2 genes using software Haploview (ref. 43, http://www.broad.mit.edu/mpg/haploview/) in African Americans, Dominicans, and Caucasians, respectively. Pairwise LD between any two genetic markers was measured by standardized LD coefficient D′ and visualized through Haploview. Haplotype blocks were defined based on the method of Gabriel et al. (44), where a pair of genetic markers is defined to be in “strong LD” if the one-sided upper 95% confidence bound on D′ is 0.98 and the lower bound is above 0.7. Haplotype frequencies of defined haplotype blocks were estimated also using Haploview.

Because haplotypes composed of genetic markers (with or without function) can sometimes provide greater power for detection of effect than the analysis of single-genetic marker, we examined haplotypes in CYP1A1 and CYP1B1 (45). As haplotype phase is usually unknown, inferring the most likely pair of haplotypes and then treating them as if they could be directly observed would result in information loss and biased confidence intervals for variable estimates (46). To account for unphased haplotypes, we modeled the probabilities of the possible haplotype pairs per subject using the generalized linear model. The haplotype analyses adjusted for environmental covariates were conducted with the R package haplo.stats (47). We first assessed haplotype × PAH interactions on logarithm-transformed PAH-DNA adduct levels; when no interaction was observed, we tested main effects of haplotypes.

Table 1 provides demographic and exposure characteristics of mothers and their newborns. There were no significant differences between the subset analyzed in this study and the total population with respect to the selected demographic and exposure characteristics. Maternal and newborn PAH-DNA adducts and prenatal PAH exposure were not significantly different between African Americans and Dominicans. Polish Caucasians had more than 10-fold higher PAH exposure than NYC African Americans and Dominicans, consistent with the higher levels of air pollution in Poland (20). Figure 2 displays scatter plots to show the correlations between the DNA adduct levels of the mothers and their newborns in three ethnic groups separately. Significant correlations were observed in all three ethnic groups. The correlations between the maternal or newborn DNA adduct levels and the continuous PAH exposure levels are significant in Caucasians (correlation = 0.205 between the newborn DNA adducts and the continuous PAH with P < 0.001 and correlation = 0.116 between the maternal DNA adducts and the continuous PAH with P = 0.030) but not in African Americans and Dominicans, which is consistent with the previous studies.

The genotype frequency distributions of the tested common genetic variants are summarized for African Americans, Dominicans, and Caucasians for mothers and newborns separately (Supplementary Tables S1 and S2).⁴

Assuming a dominant genetic model (minor allele considered as the risk allele) and with a 1 df association test, we identified several significant interactions between maternal genetic markers and PAHs on maternal PAH-DNA adducts, interactions between maternal genetic markers and PAHs on newborn PAH-DNA adducts. No interaction was observed between newborn genetic markers and PAHs on newborn PAH-DNA adducts in the three populations separately (Table 2).

In African Americans, significant interactions between maternal genotypes of CYP1B1-06 (rs4646903) and GSTT2-01 (rs1622002) and PAHs were observed on maternal PAH-DNA adducts. The interaction effect between GSTT2-01 and PAHs is plotted in Fig. 3A for illustration purposes. It shows that maternal adducts are lower in mothers with genotypes AA/AG at GSTT2-01 than in mothers with wild-type homozygote GG within the high PAH exposure group, but adducts are higher within the low PAH exposure group (interaction coefficient β = −0.44; P = 0.01; n = 127).

In Dominicans, a significant interaction was observed between maternal genotype CYP1B1-03 (rs2617266) and PAHs (interaction coefficient β = −0.30; P = 0.04; n = 165) on maternal PAH-DNA adducts (Fig. 3B). Maternal adducts were lower in mothers with genotypes TT/TC at CYP1B1-03 than in mothers with homozygous wild-type CC within the high PAH exposure group but were higher within the low PAH exposure group.

More genetic marker × PAH interactions were observed in Caucasians. Interestingly, among Caucasians, [CYP1A1-14 (rs2606345), was observed in both mother and newborn to interact with PAHs on newborn PAH-DNA adducts and in the same direction. The interaction effect between newborn CYP1A1-14 and PAHs (interaction coefficient β = 0.29; P = 0.020; n = 288) on newborn adducts is displayed in Fig. 3C. The newborn adducts were higher in newborns with genotypes GG/GT at CYP1A1-14 than newborns with homozygous wild-type TT within the high PAH exposure group but were lower within the low PAH exposure group. Moreover, among Caucasians, maternal CYP1A1-83 and another maternal SNPs of CYP1B1 gene [CYP1B1-74 (rs162560)] were observed to interact with PAHs on maternal PAH-DNA adducts. No genetic markers × PAH interaction remained significant after correction for multiple comparisons among three populations.

Table 3 displays main effects for those genetic markers that did not show genetic marker × PAH interactions. In African Americans, mothers with genotypes +/+ or +/− at genetic marker GSTM1-02 have significantly higher PAH-DNA adducts than mothers with genotypes −/−. African American newborns whose mothers had genotypes TT/TA at CYP1B1-66 or genotypes AA/AG at CYP1B1-74 had significantly higher newborn PAH-DNA adducts, whereas Dominicans newborns whose mothers had genotypes TT/TA at CYP1B1-66 or genotypes AA/AG at CYP1B1-74 had significantly lower PAH-DNA adducts. In Caucasians, newborns with genotypes CC/CT at CYP1A1-06 or genotypes TT/TC at CYP1A1-15 had significantly higher newborn PAH-DNA adducts.

Pairwise LD analysis shows that the same haplotype block with six CYP1A1 SNPs was defined in the three populations for CYP1A1 gene (Supplementary Fig. S1). Similar LD patterns were observed in mothers and newborns in the three populations. The estimated haplotype frequencies of the defined haplotype block of the CYP1A1 gene for mothers and newborns in each population are displayed in Table 4. The LD structure of the CYP1B1 gene differed substantially between Caucasians and African Americans and Dominicans but this is not surprising because of the differences in population genetic histories of the different groups and the selection bias of choosing SNPs based on the Caucasians in Breast and Prostate Cancer Cohort Consortium (37). Pairwise LD analysis suggests that four CYP1B1 SNPs [CYP1B1-66 (rs162549), CYP1B1-06 (rs1056837), CYP1B1-05 (rs1056836), and CYP1B1-74 (rs162560)] could form one haplotype block (D′ > 0.9) in African Americans, whereas three CYP1B1 SNPs (CYP1B1-06, CYP1B1-05, and CYP1B1-74) could form one haplotype block in Dominicans. Two CYP1B1 SNPs [CYP1B1-04 (rs10012) and CYP1B1-03 (rs2617266)] formed another haplotype block in both African Americans and Dominicans. Among Caucasians, all six CYP1B1 SNPs defined one haplotype block (Supplementary Fig. S2). The estimated haplotype frequencies of each defined haplotype block of the CYP1B1 gene are displayed in Table 5. No strong haplotype block was defined for the GSTT2 gene (Supplementary Fig. S3) in any population.

A significant interaction between maternal haplotypes of the defined haplotype block of the CYP1A1 gene and PAHs on newborn PAH-DNA adducts was observed in Caucasians (P = 0.03). More specifically, for Caucasians, newborns whose mothers had haplotype CGTCGG had higher PAH-DNA adducts than those whose mothers had the most common haplotype GATCTA within the high PAH exposure group, but newborns whose mothers had haplotype CGTCGG had lower PAH-DNA adducts than whose mothers had the most common haplotype GATCTA within the low PAH exposure group. A significant main effect of maternal haplotype on maternal PAH-DNA adducts was observed in Dominicans (P = 0.02). Dominican mothers with haplotype CGTTGG had higher PAH-DNA adducts than Dominican mothers with the most common haplotype GATCTA. Based on the regression model described above (47), among Dominican mothers with high PAH exposure level and mean ETS level, those that had haplotype CGTTGG had an average DNA adduct level of 0.203, whereas those with the common haplotype GATCTA had an average DNA adduct level of 0.160 (P = 0.02). A significant main effect of newborn haplotype on newborn PAH-DNA adducts was observed in Caucasians (P = 0.03). Caucasian newborns with haplotype CGTCGG had lower PAH-DNA adducts than Caucasians newborns with the most common haplotype GATCTA.

Our findings are notable because it appears that PAH-DNA adduct formation could be modulated by common genetic variants in key genes, CYP1A1, CYP1B1, GSTT2, and GSTM1. Interaction effects between genetic markers and PAH were observed in African American, Dominican, and Caucasian mothers and their newborns, but more interaction effects were observed in Caucasians than in African Americans and Dominicans (Table 2). In Fig. 3, we graphically presented some of the interactions observed in the respective ethnic groups. In African Americans, maternal adducts were lower in mothers with genotypes AA/AG at GSTT2-01 than in mothers with homozygous wild-type GG within the high PAH exposure group but were higher within the low PAH exposure group. In Dominicans, maternal adducts were lower in mothers with genotypes TT/TC at CYP1B1-03 than in mothers with homozygous wild-type CC within the high PAH exposure group but were higher within the low PAH exposure group. In Caucasians, newborn adducts were higher in newborns whose mothers have genotypes CC/CT at CYP1A1-14 than in newborns with homozygous wild-type TT within the high PAH exposure group but were lower within the low PAH exposure group. A haplotype by PAH interaction effect PAH × CYP1A1 haplotype was also observed but in Caucasians only. Although genetic polymorphisms were detected to interact with PAHs on PAH-DNA adducts in the three ethnic groups, with the limited sample sizes, we consider the current findings as suggestive only and warranting further study.

Interactions between PAHs and genetic polymorphisms in CYP1A1 were observed only in Caucasians. This may be due to the much larger sample size and the much higher PAH exposure levels in the Caucasian group. Among Caucasians, maternal CYP1A1-14 significantly interacted with PAH exposure on both maternal and newborn adducts and maternal CYP1A1-83 significantly interacted with PAHs on maternal DNA adducts. These observations suggest that the CYP1A1 gene may be an important effect modifier in Caucasians. Whether a similar conclusion can be drawn in other ethnic groups needs to be investigated by independent studies with bigger sample sizes. We note that the CYP genotypes would in theory be less predictive for the adduct levels than the enzyme levels measured in the blood. However, we did not obtain data on enzyme levels. We also note that in several cases different interaction effects were observed on maternal DNA adducts and newborn DNA adducts when the same genetic markers were investigated. This may be due to the fact that the mothers and babies have different genotypes. Moreover, the enzyme systems responsible for metabolism, detoxification, and DNA repair are immature in the fetus so that one might not expect to see the same effect of the genotype in mothers and newborns. More research efforts are needed to examine these relationships.

Main effects of the genetic polymorphisms on PAH-DNA adducts were also observed. One noticeable finding is the protective effect of GSTM1-02 deletion in African Americans, which is contradictory to some reports that GSTM1-02 deletion increases cancer risk (48). More research effort is needed to replicate this finding in African Americans. A strength of the study is the ability to explore gene-environment interaction using haplotype-tagging genetic markers, adducts, and environmental monitoring. However, it is possible that the effects we observed in this pilot study could be due to other variants that were untested but are in strong LD with the tested variants or, alternatively, chance. Moreover, because of the difference in population genetic histories of the different groups and the selection bias of choosing SNPs based on the Caucasians in Breast and Prostate Cancer Cohort Consortium (37), further work is required to confirm these findings in replication studies and, if replicated, fine mapping to determine variants that could be causally defined by corroborative laboratory analyses.

Our results are also notable because we interrogated common genetic variants in candidate genes in three distinct populations and observed different underlying genetic structures. For instance, the frequencies of the genetic polymorphisms among Caucasians are significantly different from those in African Americans and Dominicans. LD analysis also shows that heterogeneity in LD structures among the three ethnic groups is present, especially for CYP1B1 gene. Although the haplotype structure is similar in the three ethnic groups for CYP1A1 gene, the haplotype frequencies are different. Thus, in looking at genetic main effects or gene-environment interactions, researchers should be aware of the heterogeneity between ethnic groups and try to perform studies within relatively homogeneous populations to avoid spurious associations from population stratification (49). In conclusion, further studies in distinct populations will not only help to identify loci in different LD structures that are associated with gene-environment interactions but also to highlight possible variants that could confer increased risk according to population structure.