Endothelial gene regulatory elements associated with cardiopharyngeal lineage differentiation

Abstract

Endothelial cells (EC) differentiate from multiple sources, including the cardiopharyngeal mesoderm, which gives rise also to cardiac and branchiomeric muscles. The enhancers activated during endothelial differentiation within the cardiopharyngeal mesoderm are not completely known. Here, we use a cardiogenic mesoderm differentiation model that activates an endothelial transcription program to identify endothelial regulatory elements activated in early cardiogenic mesoderm. Integrating chromatin remodeling and gene expression data with available single-cell RNA-seq data from mouse embryos, we identify 101 putative regulatory elements of EC genes. We then apply a machine-learning strategy, trained on validated enhancers, to predict enhancers. Using this computational assay, we determine that 50% of these sequences are likely enhancers, some of which are already reported. We also identify a smaller set of regulatory elements of well-known EC genes and validate them using genetic and epigenetic perturbation. Finally, we integrate multiple data sources and computational tools to search for transcriptional factor binding motifs. In conclusion, we show EC regulatory sequences with a high likelihood to be enhancers, and we validate a subset of them using computational and cell culture models. Motif analyses show that the core EC transcription factors GATA/ETS/FOS is a likely driver of EC regulation in cardiopharyngeal mesoderm.

Tbx1stabilizes differentiation of the cardiopharyngeal mesoderm and drives morphogenesis in the pharyngeal apparatus

“…We evaluated the statistical significance between the compared groups using a one-sided paired Mann-Whitney test. Among these regions, we employed an L1-penalized logistic regression model to predict the probability of being an enhancer utilizing ATAC-seq data, as detailed in Aurigemma et al 2024 37 . Briefly, we downloaded the enhancer coordinates (i.e., positive regions with response Y=1) from the VISTA ENHANCER Browser.…”

“…Coordinates for non-enhancers (i.e., negative regions with response Y=0) were randomly generated, matching for cardinality and size those of enhancers. The feature matrix X was constructed with the available epigenetic tracks used by Aurigemma et al 37 (peak coordinates for histone modifications, p300 and CTCF transcription factors), assessing overlaps with both positive and negative regions. The regression model was trained using k-fold cross-validation.…”