The Genomic Architecture of Pregnancy Loss

点击翻译按钮获取中文摘要

ABSTRACT Pregnancy loss (PL) occurs in approximately 15% of clinically recognized pregnancies and only 30% of conceptions result in a live birth, yet little is known about genomic predictors of PL beyond large chromosomal aberrations. While it is likely that non-genetic etiologies and common variants underlie a component of PLs, we propose here to disentangle the mutational spectrum of rare and de novo variation contributing to non-viability. We overcome traditional barriers to genomic studies of PL, namely insufficient power, low-resolution technologies, and reductive statistical approaches, by establishing a Fetal Genomics Consortium (FGC) comprised of 21 international sites. Our team includes leading expertise in maternal-fetal medicine, statistical genetics, genomics, technology and algorithm development, structural variation, and in vivo CRISPR modeling. We will apply high-throughput genome sequencing (WGS) at the Broad Institute and external datasets as a frontline strategy and perform analyses of at least 2,500 PL trios. Our cohort will include the PL continuum, including at least 2,000 fetal demise trios from 20-42 weeks gestation and 500 recurrent pregnancy loss trios in couples with at least two previous losses at any gestational age. We will combine, process, analyze, and interpret pathogenic variation and return clinically relevant results to families, while prioritizing a subset of unsolved cases with complex fetal anomalies for long-read WGS and de novo assembly (AIM 1). We will then explore novel genomic predictors of PL and compare the genomic architectures of the developmental continuum from early PL to later onset developmental disorders (AIM 2). These studies will apply novel analytic methods to integrate all classes of genomic variation, mutation rates, relative risk estimates and measures of evolutionary constraint for each gene in the genome to interrogate the ‘intolerome’. To improve discovery power in PL, we will leverage massive population-scale datasets (>2M genomes), and the aggregation of >200,000 cases from ongoing developmental disorder studies. These cohorts are accessible and already being analyzed by our FGC groups, including the Broad Institute Center for Mendelian Genetics, Gabriella Miller Kids First sequencing center, gnomAD, All of US, the Undiagnosed Disease Network, and autism and neurodevelopmental disorder consortia studies. AIM 2 will integrate computational models of coding and noncoding constraint into a statistical framework to identify novel genes and loci associated with PL, and prioritize variants for in vivo CRISPR lethality screening in mouse embryos (AIM 3). This FGC proposal is thus poised to transform our understanding of the genomic predictors of PL. We will evaluate meticulously phenotyped PL families with emerging technologies, population- scale datasets and developmental disorder cohorts using uniform bioinformatic and statistical approaches. Our analyses will deliver clinically meaningful results to current families, and our functional modeling will inform interpretation of variation incompatible with human development for future PL families.