SRA

Implantation of the early embryo—the blastocyst—in the uterine wall to establish a pregnancy is remarkably inefficient in humans for reasons that remain largely unexplained1–3. In recent years, the volume of gene expression data from human preimplantation embryos has rapidly accumulated; however, prioritization of these data to discover specific genes that determine successful implantation is significantly hindered by ethical and experimental constraints. Here, we combine clinical morphologic grading with transcriptome analysis of matched trophectoderm (TE) and inner cell mass (ICM) samples to identify specific genes and cell-cell interactions differentially activated in human blastocysts of high and low implantation potential genome-wide. This allowed us to develop the first prioritized list of genes and cell-cell interactions associated with successful implantation. Employing multiple machine learning approaches, we identified TE and ICM genes that distinguish embryos of high and low implantation potential and found that gene expression within the ICM best predicts implantation. Unexpectedly, we discovered that blastocysts of low implantation potential share defects in the formation of the extraembryonic primitive endoderm (PrE) and the PrE-associated extracellular matrix network. Our results support a model in which successful implantation is most strongly influenced by factors and signals from the ICM, and suggest that defective PrE development, in particular, is a common cause of failed implantation in humans. Overall design: Examination of gene expression changes within the embryonic and abembryonic compartments between human blastocysts graded as good morphology (highly likely to implant and generate a successful pregnancy), versus those graded as poor morphology (unlikely to implant in the uterus).