The Institutional Review Board at the University of Iowa approved this study. The original study population consisted of 171 trios from the National Birth Defects Prevention Study (NBDPS). The NBDPS is a population-based, case-control study designed to examine environmental and genetic risk factors for over 30 major birth defects (12). A signed consent form was obtained from parents of case children. Eligible case children for the current study were term infants delivered between October 1, 1997, and December 31, 2005, diagnosed with one or more NBDPS defects (Table I; available at www.jpeds.com), including a PDA (12). Cases were ascertained in Iowa using multisource, active case finding and detailed medical record abstraction. A clinical geneticist confirmed birth defect diagnoses. Cases that had known chromosomal or other defined (single gene, teratogen) diagnoses or known additional congenital heart disease were excluded from the study.

The replication phase of the study included 573 cases and 1200 controls from New York State. The New York State samples were archived newborn blood spot cards from infants with birth weight >2500g (birth weight was used as a surrogate for gestational age as gestational age was unavailable) born between January 1, 1998 and December 31, 2005 in New York State. Cases were identified from the New York State Congenital Malformations Registry (CMR) and controls were a random sample of live births. The New York State CMR is statewide and includes all hospitals that serve children. PDA was reported by the hospital.

Data for cases from the NBDPS were obtained through part of the NBDPS protocol. Mothers of eligible case children were asked to complete a telephone interview that included items about pre- and/or post-conception reports of illness, medications, diet, substance use, and occupation. Following completion of the interview, mothers, fathers and case children were asked to provide buccal cell samples using cytobrushes. Genomic DNA was obtained from these cytobrushes according to standard NBDPS extraction and quality control protocols (13). Cases in the replication study were identified through the New York State Congenital Registry as infants with isolated PDA born to New York residents at 37 weeks or greater gestation. Cases were linked to the New York State Newborn Screening program to obtain blood spot cards for DNA extraction. Controls were randomly selected from the Newborn Screening program. A majority of cases and controls from the NBDPS and New York State Congenital Registry (90% and 100%, respectively) were non-Hispanic, Caucasian individuals.

Seventeen genes, either associated with syndromic forms of PDA or with regulation of normal PDA closure, were selected as candidates (Tables II and III). A particular focus was placed on genes representing prostaglandin, apoptosis, and inflammation pathways. After identifying high-priority candidate genes, single nucleotide polymorphisms (SNPs) in the candidate genes were identified using HapMap data to define the haplotype blocks (http://www.hapmap.org) (14). SNPs were selected based on the likelihood that the SNP altered the function of the gene product, high heterozygosity to enhance power, and as tagging SNPs in haplotype blocks. To enhance power, only those polymorphisms with a minor allele frequency greater than 10% in the Caucasian population were considered for analysis. Ultimately, the SNPs with the highest minor allele frequency near or within a given haplotype block were selected. Association studies were performed on DNA samples using an initial starting panel of 55 SNPs (Table IV) to determine allelic variation in the 17 selected genes. Candidate gene SNPs that were found significant at p < 0.1 in the initial analysis were genotyped in the replication population (NY samples).

Genotyping was performed using TaqMan® Assay-On-Demand™ probes and TaqMan® Assay-On-Design™ probes by Applied Biosystems (Foster City, CA, USA). Following PCR amplification on a PE 9700 thermocycler, the results were read using the Applied Biosystems 7900 HT instrument with Sequence Detection System software (version 2.2, Applied Biosystems, Foster City, CA, USA).

Statistical analysis for the initial 171 NBDPS trios revealed ten SNPs reaching the predetermined cutoff to be included in the replication study (p < 0.1). Four of the ten SNPs were excluded from further analyses, as the chi-square p-values for the test of Hardy-Weinberg equilibrium were less than .01 in unaffected individuals. The remaining six SNPs (Table IV) meeting the predetermined cutoff and with acceptable chi-square p-values for Hardy-Weinberg equilibrium were genotyped in the additional 1773 New York case-control samples.

The number of informative families for the TDT analysis, risk alleles, odds ratios (OR) with 95% confidence intervals, and p values for the six SNPs genotyped in both the NBDPS and New York case-control populations are presented in Table V. As TDT analysis only includes trios in which at least one parent is heterozygous for the specified SNP, the number of informative families in TDT analysis varied and was substantially smaller than the total sample (n=171 trios). For three of the SNPs (rs3761846, rs130021, and rs2238777), the allele conferring risk in the NBDPS trios is reversed in the New York State case-control sample (indicated by an odds ratio less than one in the New York case-control sample).

Three of the six SNPs achieved significance after correction for multiple comparisons (p < 0.01; Table V) in the NY case-control population. Additionally, rs934328 in TGFBR2 was found to be significant after correction for multiple comparisons in both the NBDPS and NY case-control populations (Table V). Importantly, the risk allele was the same for this SNP in both the NBPDS trio data and the New York State case-control data.

The initial analysis found several SNPs to be associated with the occurrence of PDA in term infants using a cutoff of p<0.1, and three of these SNPs (rs934328 in TGFBR2, rs3761846 in TRAF1, and rs130021 in CREBBP) were found to be significant after applying a Bonferroni correction in the replication analyses, although two of these SNPs (rs3761846 in TRAF1 and rs130021 in CREBBP) had a reversal of the risk allele in the New York case-control sample. In theory and occasionally in practice, a reversal of the risk allele can occur if the populations are of different ancestry. This is possible in the current study, although it is more likely that these represent false positives, especially given that both samples were predominantly non-Hispanic Caucasian.

Odds ratios for the risk alleles were similar in magnitude to odds ratios in other studies of common pediatric disease. The SNP found to be significant in both samples, rs934328 in TGFBR2, occurs in an intronic region and thus may not have a functional role, but may be a surrogate for an etiologic variant in the gene that is in linkage disequilibrium with rs934328. TGFBR2 was identified for the current study based on biological plausibility, as transforming growth factor-beta receptor (TGFBR) genes are potent developmental regulators of cell proliferation and differentiation. Variation within TGFBR2 could potentially alter cell cycle progression leading to phenotypic malformations in target tissue types. Certain mutations in TGFBR2 cause Loeys-Dietz syndrome, a syndrome characterized by a variety of clinical features, including thoracic aortic aneurysm and PDA (20).

Further investigation through additional genotyping or sequencing is required to ascertain whether other variants in linkage disequilibrium with the SNP are driving the association found in the current study. Additionally, the candidate gene analysis, a hypothesis-driven approach, is limited in that only genes were included with plausible biological function for contributing to PDA. However, a genome-wide approach, which is hypothesis-free, would allow for detection of variants throughout the genome, including genes for which function remains largely unknown.

Limitations of this study include the modest number studied in the family-based analysis, the high degree of heterogeneity likely in PDAs, and the lack of any connection to environmental covariates. An additional limitation is the variability of surveillance data due to the documentation or lack thereof of PDAs in the medical records and effects of non-participation on the representativeness of the sample. It is recognized that the identification of false-positives is always possible when performing multiple genetic testing, however a conservative method was used to correct for multiple comparisons. Such positive results need to be followed up with analysis of additional SNPs in and around the particular gene based on haplotype structure. Additional cases of PDA in term newborns need to be tested as well, as large sample sizes are required to evaluate the implication of the effect of a genetic variant in a disease appropriately . Major strengths of the current study include the use of TDT analysis in the family cohort and an independent replication cohort to test positive signals from the initial analysis.

PDA is a common congenital heart defect that affects thousands of term and preterm newborns each year. Determining the genetic contributions to PDA may provide opportunities to identify the pathophysiologic mechanisms that are amenable to prevention or intervention and improve the care of newborns with PDA. Our analyses support a genetic pathogenesis for term PDA. These findings suggest strong associations between variations in several genes and the occurrence of PDA in term infants. Further genetic investigation in larger sample sets and utilizing more comprehensive gene coverage (i.e. sequencing or genome-wide approaches) could help to identify other modifier genes of PDA and could further elucidate the molecular mechanisms responsible for this complex trait.