Low birth weight (LBW) refers to the weight of a newborn being less than 2500 g (Kramer, 1987) and occurs in 16% of all live-births worldwide (deOnis et al., 1998). Birth weight is regulated by two major processes: duration of gestation and intrauterine growth rate. Preterm delivery (PTD) of babies born with less than 37 weeks of gestation is responsible for one-third to two-thirds of infants with LBW (Arifeen et al., 2000; Martin et al., 2007). PTD and LBW are independent risk factors for fetal and infant mortality. However, the causes of PTD and LBW are not clear. Multiple factors may contribute to the development of LBW and/or PTD, including genetic and environmental factors, and other specific maternal-fetal characteristics (e.g., demographic, obstetric, nutritional factors, and maternal morbidity during pregnancy) (Kramer, 1987). Based on twin studies, the heritabilites of low birthweight (<2500 g) and PTD (<37 gestational weeks) are estimated to be 37–42 and 36%, respectively (Clausson et al., 2000).

One of the important environmental factors in birth outcomes is maternal tobacco smoking. Smoking exposure may reduce mean birth weight, and increase the risk of PTD and intrauterine growth restriction (Asmussen and Kjeldsen, 1975; Asmussen, 1977;, Nilsen et al., 1984; Ronco et al., 2005; Goldenberg and Culhane, 2007; Haram et al., 2007). Part of the effect of smoking on birth outcomes may be due to the metabolism of the tobacco compound PAH (polycyclic aromatic hydrocarbons) (Perera et al., 2005; Tsui et al., 2008; Wu et al., 2010). PAH is a carcinogenic compound that is detoxified in a two-stage process. PAHs are converted into procarcinogen, which is then conjugated into excretal metabolites. The conjugation is catalyzed by the product of the gene GSTT1 (Glutathione S-transferase theta 1), which belongs to the theta class of GSTs. The class members, GSTT1 and GSTT2, are located on human chromosome 22. The genes are approximately 50 kb away from each other, with a GSTT pseudo gene (GSTTP1) located between them. GSTT1 and GSTT2 share 55% amino acid sequence identity and have detoxification roles (Coggan et al., 1998), which make the GSTT1/GSTT2 region an interesting candidate for explaining smoking-induced adverse effects.

A small deletion in GSTT1 is the only reported common copy number variant (CNV) in the region. A null genotype (homozygous deletion) of GSTT1 modifies the effect of maternal smoking on birth outcomes. Smokers have infants with lower mean birth weight compared to non-smokers; however, the reduction varies according to the GSTT1 copy number. Mean birth weight decreases dramatically in smokers with the GSTT1 null genotype compared to smokers with the GSTT1 wild-type genotype (Wang et al., 2002; Wu et al., 2007; Grazuleviciene et al., 2009; Aagaard-Tillery et al., 2010). It is unclear whether other CNVs may exist in the GSTT1/GSTT2 region.

The goal of this study was to explore whether additional CNVs are distributed in GSTT1/GSTT2 and to investigate their associations with smoking and birth outcomes. First, we detected CNVs in the GSTT1/GSTT2 regions. Next, we examined the association of GSTT1/GSTT2 CNVs with birth outcomes, stratifying by smoking status. Finally we investigated the relationship of GSTT1/GSTT2 CNVs with smoking.

Linkage and SNP association studies have identified some genetic factors that are associated with adverse birth outcomes, including GSTT1. However, few CNV studies have been conducted for LBW and PTD outcomes. We focused on examining CNVs in the region from GSTT1 to GSTT2 and investigating their association with adverse birth outcomes and smoking.

Our finding that smoking may be associated both preterm birth and birth weight is consistent with other reports (Horta et al., 1997; Chan et al., 2001).

After identifying five CNV regions in GSTT2, GSTTP1, and GSTT1, we noted the complex association patterns with preterm birth and birth weight, which may be related to LD among the CNVs. The GSTT2 duplications GSTT2^a and GSTT2^b), GSTTP1 deletion and GSTT1^b deletion are in partial LD (see Table A1). The GSTT2^a deletion decreased the mean birth weight in infants of smoking mothers and is more likely to be present in smokers, but the weight change was of borderline statistical significance. In contrast, the GSTT2^b deletion increases the birth weight (Table 3B: 120.2 g, p = 0.08) in infants of non-smoking mothers and is more common in non-smokers (Table 5: odds ratio of smoking vs. non-smoking = 0.68, p = 0.20), but again the weight change was of borderline statistical significance. Interestingly, neither of the GSTT2 deletions was associated with preterm birth or “very preterm” birth (<35 weeks) (Windham et al., 2000; Danileviciute et al., 2012), but both GSTT2 duplications appear to be associated with increased risk of birth before 35 weeks in non-smokers, and associated with an even higher risk when smoking and the duplications were considered jointly. These findings suggest that GSTT2 may have a role in detoxification of tobacco, although it is also possible that these results are explained by LD with other polymorphisms or multiple testing effects. CNVs in GSTT2 may influence the metabolism of nicotine, and exacerbate the toxic effects of smoking.

The deletions in GSTTP1 were in high LD with GSTT1^b (22,695,592–22,725,333 bp) (D′ = 0.89) and in moderate LD with two duplications in GSTT2 (D′ = 0.62). We did not observe significant effects of GSTTP1 and GSTT1^b deletions on birth weight in full term controls, but they were associated with increased risk of birth before 35 weeks in both non-smokers and smokers, with the odds ratio slightly higher when the deletions were considered jointly with smokers. This suggests that GSTTP1 and GSTT1^b deletions are associated with reduced birth weight through very preterm birth, but the effect on birth weight was not detected when including full term infants. We also found that GSTTP1 and GSTT1^b deletions were less likely to occur in smokers. Smoking could be a proxy for some other linked yet unknown behavior or may be independently associated with these deletion CNVs.

We compared our CNV results in GSTT1 and GSTT2 with published findings for the previously described GSTT1 CNV, which was reported to modify the effect of smoking on birth weight (Wang et al., 2002; Wu et al., 2007; Grazuleviciene et al., 2009; Aagaard-Tillery et al., 2010). Smokers with the null genotype of GSTT1 had infants with lower mean birth weight than those with the wildtype genotype of GSTT1. The deletion's physical genomic position (22,698,742–22,706,887 bp on chromosome 22) was not covered by the HumanHap660 chip resulting in a marker gap between 22,697,104, and 22,713,954 bp in our arrays. However, our GSTT1^b deletion encompasses the reported GSTT1 null genotype region. It is notable that the frequency of the homozygous GSTT1^b deletion CNV in our data was highly consistent with the previously reported frequency of the GSTT1 null genotype in Denmark (15%) (Saadat et al., 2001). In contrast to most previous studies, we were able to examine the influence of GSTT1^a and GSTT1^b deletions on fetal growth in women with full term births, which means that we were able to examine the effects on birth weight more or less independently of effects on gestational age. We did not find significant association of GSTT1 deletions with birth weight in full term birth, nor did we observe interactions of genotypes and smoking status on birth weight in controls. This may be due to the relatively small number of mothers who smoked during pregnancy, or due to the smoking measurement being a dichotomous measure. When contrasted with earlier reports of a GSTT1—smoking interaction impacting birth outcomes, the current study's lack of evidence for an interaction may be attributed to differences in confounders included in models, variability of smoker definitions, and this study's mothers being categorized as non-smokers based on the pregnancy time period only.

A single SNP (rs1622002) in GSTT2 has been associated with metabolism of major tobacco carcinogen PAH (Wang et al., 2008), but this SNP was not included in the present study's genotype panel.

There are several epidemiologic limitations in our study. Smoking was recorded as any smoking present during pregnancy, which did not consider the influence of smoking in different trimesters. Secondly, regressions were adjusted for some demographic factors, but excluded adjustments for psychosocial factors. Statistical evaluations were not corrected for multiple testing which tempers the conclusions that can be drawn about associations, but the uncorrected p-values do suggest a prioritization of findings to pursue in future studies.

In addition to the epidemiological issues, CNV calls are notoriously error-prone. It is not feasible to validate specific CNV calls within the context of a large-scale association study. However, for any given CNV with calling error rates independent of phenotype, those errors should only reduce power and not cause bias in the association measure (Zheng et al., 2012). We have no reason to expect CNV calling biases in the current study, so we feel that our results are quite robust. Moreover, the fact that our frequency for the GSTT1^b homozygous deletion was an almost perfect match to the frequency reported in this population for the known deletion was very reassuring. We did not compare our PennCNV CNV calls to any other algorithm in this study, although we did so previously in a more explicit algorithm comparison (Zheng et al., 2012).

Despite these limitations, our study is able to shed light on CNVs in smoking-associated adverse birth outcomes. Our candidate gene analysis in GSTT1/GSTT2 and identification of five CNVs that appear to be associated with birth weight and/or birth before 35 weeks is novel. The distribution and LD pattern of these CNVs are more complicated than expected, which may impact our understanding of the relationship between the GSTT1 deletion and adverse birth outcomes. Follow-up genetic association studies of birth weight need to include multiple CNVs in this region instead of single polymorphisms. Additional studies including molecular evidence will be needed to validate the detected CNVs and replicate these results.