Targeted Sequencing and Meta-Analysis of Preterm Birth

Abstract

Understanding the genetic contribution(s) to the risk of preterm birth may lead to the development of interventions for treatment, prediction and prevention. Twin studies suggest heritability of preterm birth is 36-40%. Large epidemiological analyses support a primary maternal origin for recurrence of preterm birth, with little effect of paternal or fetal genetic factors. We exploited an "extreme phenotype" of preterm birth to leverage the likelihood of genetic discovery. We compared variants identified by targeted sequencing of women with 2-3 generations of preterm birth with term controls without history of preterm birth. We used a meta-genomic, bi-clustering algorithm to identify gene sets coordinately associated with preterm birth. We identified 33 genes including 217 variants from 5 modules that were significantly different between cases and controls. The most frequently identified and connected genes in the exome library were IGF1, ATM and IQGAP2. Likewise, SOS1, RAF1 and AKT3 were most frequent in the haplotype library. Additionally, SERPINB8, AZU1 and WASF3 showed significant differences in abundance of variants in the univariate comparison of cases and controls. The biological processes impacted by these gene sets included: cell motility, migration and locomotion; response to glucocorticoid stimulus; signal transduction; metabolic regulation and control of apoptosis.

Fig 1. Manhattan Plot of Significant Variants.

The 13,000 variants from the targeted exome library and 11,000 variants from the haplotype block library were compared for difference in abundance in the cases versus the controls. The figure shows a Manhattan plot of all variants across 22 autosomes with the vertical axis being the -logP value from the statistical test for association, with the threshold line (-logP 1.3) indicating p-value of 0.05. There were 205 and 168 variants that significantly differed in abundance in cases versus controls from the exome and haplotype block libraries respectively.

Fig 2. Meta-analysis and analytical pipeline:

The genes harboring variants in each patient were analyzed by gene set enrichment using the MSig database C2 collection of gene sets [43]. The significant gene sets for each patient were combined into a binary association matrix. The iBBiG algorithm extracts modules of gene sets and patient subsets from the data matrix. The modules are represented by different colors. Fisher’s exact test was used to identify modules with significant differences in the number of cases and controls.

Fig 3. Network analysis of cases and controls.

Cases are labeled by letter “c” in the significant modules. (A) Network of all modules of the exome library and patients, with the patients from significant Modules 4 and 8 highlighted in red and blue, respectively. (3B) Network of all modules of the haplotype block library and patients, with the patients from significant Modules 2, 8 and 9 highlighted in light blue, green and orange, respectively.

Fig 4. Network of modules and their gene sets.

(A) Network output showing all 10 modules from the exome library and the genes contained in each module. The two significant modules are displayed as insets. Inset a1 and a2 display the genes of E8 and E4 respectively. (B) Network output showing all 10 modules from the haplotype block library and the genes contained in each module. Insets b1, b2 and b3 show the genes of H2, H9 and H3 respectively.

Fig 5. Ontology groups.

Diagram showing the clusters of terms from the Gene Ontology analysis for biological processes related to preterm birth. Gene Ontology Database terms for biological processes shown in clusters A thru I are detailed in S3 Table.