Abstract
The time of onset of the second cell fate decision in the mouse preimplantation embryo is still unknown. Ohnishi et al. (2014) looked for cell heterogeneity in the ICM at E3.25 that could indicate the time preceding the apparent segregation of PE and EPI at E3.5, but were not able to detect an early splitting transcriptomics event using state-of-the-art clustering techniques. We developed a new clustering algorithm, hierarchical optimal k-means (HOkM), and identified from single cell (sc) transcriptomics microarray data two groups of ICM cells during the 32 to 64 mouse embryo transition: from embryos with less than 34 cells, and more than 33 cells, corresponding to two developmental sub-stages. The genes defining these sub-stages indicate that the development of the embryo to 34 cells triggers a dramatic event as a result of which Bsg is high expressed, the canonical Wnt pathway is activated, Oct4 is stabilized to high expression and the chromatin remodeling program is initialized to establish a very early narve pluripotent state from the preceding totipotency. We characterized our HOkM partition comparing with independent scRNA-seq datasets. It was staggering to discover that from the 3.4360×1010 possible bi-partitions of the E3.25 data of Ohnishi et al. (2014), our HOkM discovered one partition that shares the biological features of the early and late 32 ICM cells of Posfai et al. (2017). We propose that the stabilization of Oct4 expression is a non-cell autonomous process that requires a minimal number of four inner cell contacts acquired during the transition from a homogeneous outer-cell environment to a heterogeneous inner/outer cell environment formed by the niche of a kernel of at least six inner cells covered by trophectoderm.