A Genomic Perspective on Early Embryo Development in Cattle

Abstract

In vitro embryo production (IVP) has revolutionized the cattle industry, but it still faces efficiency limitations compared to in vivo embryo production. With a major impact on it, a proper transcriptome profile is crucial for an embryo to undergo pre-implantation development and become developmentally competent. Several research groups have been investigating highly dynamic gene expression regulation during early embryo development and its relationship to embryonic phenotypes. While it has been fundamental to improve our understanding of the genetics underlying cattle embryo development and identify potential phenotypic markers associated with developmental competence, many aspects have not been fully elucidated. Moreover, there are few cattle studies focusing on novel gene expression and the development of viable embryo selection methods. Here, our goals were to identify potential markers associated with the acquisition of developmental competence in in vitro-produced cattle embryos and to develop an automated method for predicting competent embryos. In our first study, we identified over three thousand unknown loci being expressed during different stages of oocyte and embryos, from which almost one third (63.67%; 1,931/3,033) had coding potential. Interestingly, 102 novel genes with greater transcriptional activity at the eight-cell stage were highly co-expressed (|r|>0.85, P<1x10-8) with genes involved in pluripotency maintenance and embryo development processes. It included a novel gene (ENSBTAG00000068261), which reduced blastocyst formation when its function was disturbed via CRISPR-Cas9 gene editing (P=1.55x10-7). Our second study identified several morphokinetic features associated with embryo development groups (P<0.05). Specifically, fast-cleaved embryos with fewer abnormal events were more prone to be developmentally competent, in which embryos reaching the blastocyst stage went through synchronous cleavage, a higher number of cells at 42 and 48 hours post-fertilization (hpf), and faster timing in reaching eight-cell stages than embryos arrested in earlier stages. By integrating morphokinetic features up to 72hpf into our machine learning (ML) model, we achieved high performance in predicting developmental competence. It displayed an overall accuracy of 69.23% (18/26), with local accuracies of 42.86% (6/14) for competent and 100.00% (12/12) non-competent embryos. Moreover, it had 100% precision in identifying competent embryos, a recall of 42.86%, and an F1-score of 60.00%. In conclusion, this work provides important findings underlying gene expression architecture on developmental competence acquisition, and for the first time in cattle studies, we propose an automated ML model using non-invasive morphokinetic markers to predict and select potentially competent embryos up to 3 days of culture, which will allow for a better understanding of this highly complex process and pave ways for predictive models in embryo implantation or pregnancy outcome development.