Entropy sorting of single cell RNA sequencing data reveals the inner cell mass in the human pre-implantation embryo

Abstract

A major challenge with single cell gene expression data is to discern meaningful cellular heterogeneity from technical or biological noise. To address this challenge, we present Entropy Sorting, a mathematical framework that distinguishes genes indicative of cell identity. On synthetic data we demonstrate the removal of noisy signals to reveal a higher resolution of gene expression patterns than commonly used feature selection methods. We then apply Entropy Sorting to human pre-implantation embryo scRNA-seq data. Previous studies failed to identify a distinct signature of the inner cell mass (ICM), suggesting that lineage segregation in the human embryo does not follow the paradigm established in mouse. However, Entropy Sorting identifies a set of highly informative genes that clearly separate the ICM, Trophectoderm, Epiblast and Hypoblast and reveals sequential lineage bifurcations as in mouse. Entropy sorting thus provides a powerful approach for maximising information extraction from high dimensional datasets such as scRNA-seq data.

A new human embryonic cell type associated with activity of young transposable elements allows definition of the inner cell mass

“…More generally, ICM and EPI markers on the one hand and NCC, PrE-markers along with DNA damage response genes on the other, show an antagonistic pattern of expression (Figs 4B and S7B ). Our top ICM markers that are reproduced in other studies [ 3 , 41 ], include NANOGNB, FBXL20, FOXR1, PRR13, and SPIC ( Fig 4B ). In our comparisons of hESCs transcriptomes, these genes are expressed (log2 TPM > 1) in HERVH High cells only.…”

“…This rejects the hypothesis of simultaneous blastocyst lineage specification in human early embryos [ 2 ]. Since we first reported this [ 15 ], the finding has been replicated in recent studies, via diverse methods [ 3 , 41 ].…”

Evidence implicating sequential commitment of the founder lineages in the human blastocyst by order of hypoblast gene activation

“…One report suggested origination of the hypoblast from the epiblast ( Meistermann et al, 2021 ). Using an entropy-based approach for feature selection for published scRNA-seq, we identified an early ICM population from which both the epiblast and hypoblast arise ( Radley et al, 2022 ), as implicated previously ( Stirparo et al, 2018 ).…”

Lineage segregation in human pre-implantation embryos is specified by YAP1 and TEAD1

“…Our data also support the hypothesis that PrE cells differentiate from EPI cells in expanding blastocysts (between B3 on 5 dpf and B4 on 6 dpf) (Meistermann et al , 2021). Their single-cell transcriptomics data and our NANOG analysis also suggest that the ICM (5 dpf) and EPI (6 dpf) are very similar (if not the same), but further characterization of the pluripotent cells is needed to precisely define the stages (Radley et al, 2022).…”

“…According to the one step model, EPI and PrE lineages arise simultaneously with TE cells 5 days post fertilisation (dpf) (Petropoulos, 2016) whereas according to the two step model the first and the second lineage differentiation events appear consecutively (Meistermann et al , 2021) as in mice (Chazaud and Yamanaka, 2016). It is also unclear whether ICM cells segregate into EPI and PrE cells (Radley et al , 2022) as in mice (Chazaud and Yamanaka, 2016) or whether PrE cells differentiate from the EPI cells on 6 dpf (Meistermann et al , 2021). This lack of clarity may have been caused by the heterogeneous development of outbred human embryos in vitro, different laboratory culture conditions and the lack of systematic coupling of morphology with the underlying analysis.…”