We used data from the National Birth Defects Prevention Study (NBDPS), a multisite, population-based, case-control study of major birth defects that included a maternal interview and self-collection of buccal (cheek) cells from each case and control infant and his/her mother and father. Detailed methodology for the NBDPS has been published previously [Rasmussen et al., 2002; Yoon et al., 2001]. Briefly, case infants with selected major birth defects were identified using birth defects surveillance systems at the 10 participating sites. Liveborn control infants without major birth defects were randomly selected from birth certificates or birth hospital data from the same region and time period. Clinical geneticists reviewed data abstracted from medical records using standardized case definitions. Case infants with known chromosomal abnormalities or single gene disorders were excluded. Standardized computer assisted telephone interviews were conducted in English or Spanish between six weeks and 24 months after the estimated date of delivery (EDD). Women were asked about their exposures from three months before conception until delivery. Following completion of the interview, buccal cell collection kits that included cytobrushes for the mother, her child, and the child’s father (two brushes per participant) were mailed. Buccal cell collection initiation varied by site, and samples were requested only from mothers whose interviews were completed after collection began. Institutional Review Boards (IRBs) at the Centers for Disease Control and Prevention (CDC) and each study site have approved the NBDPS.

These analyses included infants of non-Hispanic white or Hispanic mothers with an EDD between October 1, 1997 and December 31, 2003. Race-ethnicity was self-reported by each mother, and infants were analyzed according to their mother’s race-ethnicity. Infants of mothers of other race-ethnicities were not included because of small numbers of case infants (i.e., < 4) with mothers who reported periconceptional smoking and with analyzable buccal cell samples. Samples from mothers were removed from analyses if she reported using an egg or embryo donor. DNA samples from the infant, mother, or both were analyzed; father samples were not included in these analyses. Case infants had gastroschisis with or without other major congenital anomalies, and samples were available only if they were liveborn. Infants diagnosed with limb body wall defects were excluded from these analyses.

Infants and mothers were classified as exposed to periconceptional maternal smoking if the mother reported any smoking at any time in the month before or in the first three months of pregnancy, since gastroschisis occurs during the third and fourth weeks post-fertilization [Sadler and Feldkamp, 2008]. Infants and mothers were classified as unexposed if the mother did not report any smoking in the month before and in the first three months of pregnancy.

Laboratories at each participating site extracted DNA from buccal cells using a variety of methods for samples collected prior to mid-2003 [Rasmussen et al., 2002]. A laboratory atCDC extracted DNA from Georgia participant samples and from all sites after mid-2003 using a modified phenol-chloroform method [Garcia-Closas et al., 2001]. Human genomic DNA (gDNA) yields were assessed by quantitative real-time PCR using TaqMan® Ribonuclease P assays (Applied Biosystems, Foster City, CA). Specimens with DNA concentrations less than 0.1ng/μl were excluded. DNA quality and family relationships were assessed using tetranucleotide short tandem repeats (STRs) as described previously [Gallagher et al., 2011]. DNA samples from inconsistent mother-infant pairs were excluded; consistent pairs and unpaired mothers and infants were included. Positive and negative controls were included in each DNA extraction and quantitation assay.

We analyzed five SNPs in three genes (CYP1A1, CYP1A2, and NAT2) that were selected based on their effect on XME activity [Consensus Human NAT Gene Nomenclature Database; Human CYP Allele Nomenclature Committee Database], their minor allele frequencies [Packer et al., 2006], and assay success in preliminary validation studies. Appendix 1 provides more information on the selected XME gene variants. Genotyping was completed on either gDNA or whole genome amplified (WGA) products from mothers and infants using Pyrosequencing® technology (Qiagen, Valencia, CA). Methods and quality assessment results were described previously [Gallagher et al., 2011]. Replica genotyping was performed on separate days for at least 4% of specimens from each genotyping plate. For each mother-infant pair, SNPs that were inconsistent with Mendelian inheritance were removed from further analyses. Specimens with missing data for one or more SNPs were removed from further analyses. The laboratory at CDC successfully completed external quality assessment (protocols are available upon request).

Data from control mothers were assessed for Hardy-Weinberg equilibrium by race-ethnicity for each of the five SNPs studied using Chi square tests. Mendelian errors were identified and allele frequencies were calculated using PedCheck Version 1.00 [O’Connell and Weeks, 1998] and PLINK Version 1.07 [Purcell et al., 2007]. Maternal age at delivery, alcohol use, body mass index, obesity, parity, and education were assessed as potential confounders using Chi square tests in non-Hispanic white and Hispanic control mothers separately. Maternal age at delivery was also assessed as a potential effect modifier by completing stratified analyses (< 25 years vs ≥ 25 years). Maternal age at delivery (continuous) was included in the logistic regression models.

Logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) using PASW Statistics 18, Release Version 18.0.0 (SPSS, Inc., 2009, Chicago, IL, www.spss.com). Maternal age-adjusted associations between smoking and gastroschisis were assessed, stratified by race-ethnicity. Maternal age-adjusted associations between maternal or infant XME gene variants and gastroschisis with and without stratification by maternal periconceptional smoking status were assessed separately in non-Hispanic white and Hispanic mothers and infants using dominant or recessive inheritance models. For all analyses, dominant inheritance models were used when assessing CYP1A1*2A, CYP1A2*1C, NAT2*5, and NAT2*6 (i.e., persons who had one or two copies of the variant allele were combined and compared to persons who had zero copies) because small numbers of mothers and infants carrying two copies of the variant allele limited analyses of other inheritance models. Recessive inheritance models were used when assessing CYP1A2*1F (i.e., persons who had two copies of the variant allele were compared to persons who had zero or one copy of the variant allele combined) because small numbers of mothers and infants carrying two copies of the wild-type allele limited analyses of other inheritance models. After stratification, analyses were completed only if there were four or more mothers or infants in each genotype category.

To assess the contribution of having any high risk XME gene variants in the mother and her infant, we also dichotomized combined gene variants from available mother-infant pairs (0 (referent group) or ≥1) for each of the five XME gene variants. These analyses were completed only when DNA was available from both a mother and her infant. If a mother or her infant carried two copies of CYP1A2*1F, the pair was categorized as having a high risk gene variant; for all other variant alleles (i.e., CYP1A1*2A, CYP1A2*1C, NAT2*5, and NAT2*6), if a mother or her infant carried one or two copies of the variant allele, the pair was categorized as having a high risk gene variant.

The interview participation rate was 72% for all mothers of infants with gastroschisis (n=504), and 69% for all mothers of control infants (n=4949). Buccal cell samples were requested from 455 case families and 4251 control families and were submitted for the mother, infant, or both for 47% of families with gastroschisis (n=215), and 43% of control families (n=1834). After excluding families with reported maternal race-ethnicity other than non-Hispanic white or Hispanic, and specimens that did not pass quality control (i.e., STR or SNP results were inconsistent with Mendelian inheritance; DNA quantity was <0.1 ng/μl; data were missing for >1 SNP), samples from 108 non-Hispanic white case families (76 mother-infant pairs; 29 mother only; and 3 infant only), 62 Hispanic case families (36 mother-infant pairs; 22 mother only; and 4 infant only), 1147 non-Hispanic white control families (890 mother-infant pairs; 210 mother only; and 47 infant only), and 337 Hispanic control families (233 mother-infant pairs; 72 mother only; and 32 infant only) were included (Figure 1).

There were some differences in selected maternal demographic and behavioral risk factors for gastroschisis among case and control infants (Table I). Mothers of infants with gastroschisis were younger, less educated, and more likely to be underweight.

Genotype call rates were between 99 and 100 percent for all five variants. The genotype distribution of each variant did not deviate from Hardy-Weinberg equilibrium (P>0.05) in non-Hispanic white or Hispanic mothers of control infants. The minor allele frequencies of each genetic variant in non-Hispanic white and Hispanic control mothers are listed in Appendix 1 and were consistent with reported published frequencies [Chang et al., 2009; Sherry et al., 2001; Swinney et al., 2011].

Of the potential confounders assessed, only maternal age at delivery (continuous) and maternal education (≤12 years or >12 years) were found to be associated with the XME gene variants (Appendix 2). Because maternal age and maternal education are correlated and young maternal age at delivery is an established risk factor for gastroschisis [Rasmussen and Frias, 2008], we included only maternal age at delivery in the models. Among non-Hispanic white and Hispanic control mothers included in these genetic analyses, 20.1% and 9.8%, respectively, reported smoking in the month before pregnancy or during the first trimester. Nearly identical, elevated maternal age-adjusted ORs were observed for gastroschisis risk and exposure to maternal periconceptional smoking in both racial-ethnic groups; however, the finding was statistically significant only in non-Hispanic white mothers (aOR=2.07, 95% CI 1.33-3.23, P<0.01) (Table II).

A suggestive maternal-age adjusted association of NAT2*6 with gastroschisis was observed in Hispanic mothers (aOR=1.88, 95% CI 1.04-3.39, P=0.04) and their infants (aOR=1.93, 95% CI 0.96-3.88, P=0.07) (Table III). An age-adjusted association of NAT2*6 with gastroschisis was not observed in non-Hispanic white mothers or their infants and adjusted associations of CYP1A1*2A, CYP1A2*1C, CYP1A2*1F, and NAT2*5 with gastroschisis were not observed in mothers of either race-ethnicity or their infants (Table III). Similar results were observed in analyses stratified by maternal age at delivery (data not shown).

After stratifying by smoking status, a suggestive maternal age-adjusted association of NAT2*6 with gastroschisis continued to be observed in Hispanic non-smoking mothers (aOR=2.17, 95% CI 1.12-4.19, P=0.02) and their infants (aOR=2.11, 95% CI 1.00-4.48, P=0.05); no association was observed in Hispanic smoking mothers (Table IV). No statistically significant age-adjusted associations of NAT2*6 with gastroschisis were observed in non-Hispanic white smoking or non-smoking mothers or their infants (Table IV). A suggestive maternal age-adjusted association of CYP1A1*2A with gastroschisis was observed in non-Hispanic white smoking mothers (aOR=0.38, 95% CI 0.15-0.98, P=0.05) that was not observed in their infants or in non-Hispanic white non-smoking mothers or their infants (Table IV). No associations of CYP1A1*2A with gastroschisis were observed in Hispanic non-smoking mothers or their infants (Table IV). No statistically significant age-adjusted associations were observed between CYP1A2*1C, CYP1A2*1F or NAT2*5 and gastroschisis (Table IV).

A suggestive maternal age-adjusted association of NAT2*6 with gastroschisis was observed in non-Hispanic white (aOR=3.41, 95% CI 1.25-9.31, P=0.02) and Hispanic (aOR=3.31, 95% CI 1.42-7.75, P=0.01) non-smoking mother-infant pairs when comparing those pairs carrying one or more high risk gene variant to those pairs with no high risk gene variant (Table V). A statistically significant adjusted association of NAT2*6 with gastroschisis was not observed in non-Hispanic white smoking mother-infant pairs (Table V). No statistically significant associations were observed in non-smoking mother-infant pairs of either race-ethnicity for the other four gene variants and were not observed in non-Hispanic white smoking mother-infant pairs for three of the four gene variants with sufficient numbers (Table V).

The elevated effect estimates observed for gastroschisis risk in Hispanic mothers and their infants who carried one or two copies of NAT2*6 (Table III) are biologically plausible because the resulting decrease in NAT2 activity [Consensus Human NAT Gene Nomenclature Database] leads to increased susceptibility to the toxic effects of the intermediates formed in phase I reactions. NAT2*6 has not been reported in previous studies to be associated with gastroschisis.

Analyses of CYP variants were stratified by maternal periconceptional smoking status because CYP1A1 and CYP1A2 are induced by exposure to cigarette smoke [Gunes and Dahl, 2008]. We expected individuals carrying CYP1A1*2A to be more susceptible to the toxic effects of chemicals in cigarette smoke because this variant has been reported to increase enzyme activity [Georgiadis et al., 2005] and lead to increased toxic intermediates; however, mothers carrying this variant who smoked periconceptionally appeared to be less likely to have an infant with gastroschisis (Table IV). The CYP1A1*2A fetal variant has been reported to play a protective role for oral cleft risk in children whose mothers were exposed to secondhand tobacco smoke during the first trimester [Chevrier et al., 2008]. Kurahashi and colleagues [Kurahashi et al., 2005] reported a protective effect of the maternal variant for hypospadias risk in the offspring of Japanese mothers (smoking and non-smoking); however, there was no interaction effect. In our study, this was the only variant that had a suggestive modifying effect on maternal periconceptional smoking. CYP1A1*2A has not been reported in previous studies to be associated with gastroschisis.

It is unclear whether gastroschisis risk is influenced more by maternal or fetal genes or both equally. We observed suggestive adjusted associations between NAT2*6 and gastroschisis for Hispanic and non-Hispanic white non-smoking mother-infant pairs. The suggestive associations that were consistently observed in our analyses between NAT2*6 and gastroschisis in Hispanic families have not been reported previously. Although the variant has not been previously reported to be associated with gastroschisis, it has been associated with cleft lip with or without cleft palate [Lie et al., 2008], including reports of having a modifying effect on the association between maternal smoking and orofacial clefts [Shi et al., 2007].

In our study, CYP1A1*2A was the only variant that acted as an effect modifier for maternal periconceptional smoking and gastroschisis. The effects we observed in mothers and infants who were not exposed to periconceptional smoking could be due to interactions of NAT2*6 with other exposures. Our data were analyzed separately for each race-ethnicity because of large differences in allele frequencies, which limited our ability to assess interactions. Further sub-classification of the Hispanic population was not completed, and genetic admixture within this population might have an impact on our results [Martinez, 1998]. Maternal and infant genotypes were not adjusted for each other when analyses were completed separately which could be a potential source of confounding. Other limitations included the use of self-reported maternal race-ethnicity, which was used to classify the infant race-ethnicity, and the use of self-reported smoking that did not include data on level of smoking or secondhand smoking exposures. These exploratory analyses were completed with limited numbers of families and by reporting results without correcting for multiple testing we can provide more liberal data that can better inform future studies.

Strengths of our study included the assessment of data from a large population-based, case-control study of risk factors for birth defects with both genetic and environmental exposure data and standardized case definitions.

This study focused on a small number of XME genes because of limited DNA quantity and stringent quality control. Other gene variants in the XME pathway might affect gastroschisis risk through their effect on smoking behavior (e.g. CYP2A6 [Tyndale and Sellers, 2002]) or important detoxification reactions (e.g., glutathione S-transferase [Garlantezec et al., 2012]). Other exposures might also explain, in part, the contradictory associations with maternal smoking observed in our study. CYP1A1 and CYP1A2 are inducible by a large number of common exposures in addition to cigarette smoke, such as cruciferous vegetables [Vistisen et al., 1992], caffeine [Tantcheva-Poor et al., 1999], and charcoal-grilled food [Kall and Clausen, 1995]. Oral contraceptives [Abernethy and Todd, 1985] and apiaceous vegetables [Peterson et al., 2006] inhibit enzyme activity. NAT2 metabolizes a wide range of drugs, including isoniazid (antituberculotic), hydralazine (antihypertensive), sulfonamides (antibacterials), and caffeine [Daly, 2003; Kawamura et al., 2005].

We believe this is the first report of CYP1A1*2A as a possible protective variant against gastroschisis in the offspring of women who smoke during the periconceptional period and also the first report of a suggestive association between NAT2*6 and gastroschisis risk for Hispanic non-smoking mothers and their infants. Although the sample size is small, to our knowledge, this is the largest case-control study examining genetic and non-genetic risk factors for gastroschisis that has been completed to date. Five previous studies of genetic risk factors for gastroschisis included no more than 57 case families (whereas we included 170 case families) [Cardonick et al., 2005; Feldkamp et al., 2012; Komuro et al., 2001; Lammer et al., 2008; Torfs et al., 2006]. It is challenging to conduct genetic epidemiologic analyses on such a rare birth defect, especially one that disproportionally affects younger mothers who typically have lower participation in biospecimen collection. We feel the value of these exploratory analyses is to inform studies that can build upon these methodologies, resources and results. Future studies are needed to confirm our findings with these gene variants and to investigate other exposures or other XME genes and exposure to periconceptional smoking.