A reference atlas of human cell types is a major goal for the field. Here, we set out to generate single-cell atlases of both gene expression (this study) and chromatin accessibility (Domcke et al., this issue) using diverse human tissues obtained during midgestation.

Contemporary knowledge of the molecular basis of in vivo human development mostly derives from a combination of human genetics, in vivo investigations of model organisms, and in vitro studies of differentiating human cell lines, rather than through direct investigations of developing human tissues. Several challenges have historically limited the study of developing human tissues at the molecular level, including limited access, tissue degradation, and cell type heterogeneity. For this and the companion study (Domcke et al., this issue), we were able to overcome these challenges.

We applied three-level single-cell combinatorial indexing for gene expression (sci-RNA-seq3) to 121 human fetal samples ranging from 72 to 129 days in estimated postconceptual age and representing 15 organs, altogether profiling 4 million single cells. We developed and applied a framework for quantifying cell type specificity, identifying 657 cell subtypes, which we preliminarily annotated based on cross-matching to mouse cell atlases. We identified and validated potentially circulating trophoblast-like and hepatoblast-like cells in unexpected tissues. Profiling gene expression in diverse tissues facilitated the cross-tissue analyses of broadly distributed cell types, including blood, endothelial, and epithelial cells. For blood cells, this yielded a multiorgan map of cell state trajectories from hematopoietic stem cells to all major sublineages. Multiple lines of evidence support the adrenal gland as a normal, albeit minor, site of erythropoiesis during fetal development. It was notably straightforward to integrate these human fetal data with a mouse embryonic cell atlas, despite differences in species and developmental stage. For some systems, this essentially permitted us to bridge gene expression dynamics from the embryonic to the fetal stages of mammalian development.

The single-cell data resource presented here is notable for its scale, its focus on human fetal development, the breadth of tissues analyzed, and the parallel generation of gene expression (this study) and chromatin accessibility data (Domcke et al., this issue). We furthermore consolidate the technical framework for individual laboratories to generate and analyze gene expression and chromatin accessibility data from millions of single cells. Looking forward, we envision that the somewhat narrow window of midgestational human development studied here will be complemented by additional atlases of earlier and later time points, as well as similarly comprehensive profiling and integration of data from model organisms. The continued development and application of methods for ascertaining gene expression and chromatin accessibility—in concert with spatial, epigenetic, proteomic, lineage history, and other information—will be necessary to obtain a comprehensive view of the temporal unfolding of human cell type diversity that begins at the single-cell zygote. An interactive website facilitates the exploration of these freely available data by tissue, cell type, or gene (descartes.brotmanbaty.org).

We used a three-level combinatorial indexing assay (sci-RNA-seq3) to profile gene expression in ~4,000,000 single cells from 15 fetal organs. This rich resource enables, for example, the …