A molecular handbook for human development.

Abstract

natural selection by comparing features between individuals and species. A comparative approach is also crucial to establish the cellu lar taxonomy that underlies human physiology. Technological advances in single-cell genomics have facilitated the production of numerous cell atlases that, through comparative analysis, define the full set of cells that constitute a system of interest — usually a whole organ. Extending the scope of an atlas from organ to whole organism increases the power of this approach by capturing data across physiological systems. To this end, two papers in Science present the comprehensive molecular characterization of cell types across nearly all organs during human fetal development. They reveal previously unidentified cell subtypes, and define celldifferentiation pathways through analysis of gene expression and chromatin (the DNA–protein complex into which a cell’s genetic material is packaged). The work is a remarkable feat, both technically, in terms of the complexity of the paired data, and because of the scale of the studies, which involved analysis of 15 organs from human fetuses between 72 and 129 days after conception. In the first of the papers, Cao et al. generated gene-expression profiles (transcriptomes) from 4 million single cells across these organs. Analysis of these profiles revealed 77 main cell types, defined with reference to existing singleorgan atlas data. In the second of the papers, Domcke et al. presented an improved method for assessing chromatin accessibility, an analysis that provides insight into how genes are regulated during development. Loosely packaged chromatin regions are thought to be more accessible to regulatory proteins such as transcription factors, and are often involved in regulating gene expression and in establishing and maintaining cell identity. The authors’ approach enabled the analysis of 800,000 cells from the same samples as those used by Cao and colleagues, which led to the identification of 54 of the same cell types. The considerable data collected allowed both groups to define highly expressed ‘marker’ genes and corresponding transcription factors unique to each cell type. The authors also integrated their atlases with existing mouse atlas data, making each a more robust and complete reference. Combining these data sets enables validation of how cell types are characterized in each species and will help researchers to better design experiments that use mouse models to investigate human physiology. Together, the papers constitute a substantial resource, which is openly available on an interactive website (descartes.brotmanbaty.org). The authors developed an analytical framework that led to interesting biological insights, demonstrating the potential of this body of work for making discoveries. They defined previously uncharacterized cell subtypes by comparing expression and chromatin patterns across organs, and they used this multiorgan approach for cell-lineage analysis. Domcke et al. compared cell-lineage diversity across organs and revealed that circulating blood cell subtypes are almost identical, irrespective of the organ from which they were isolated. Conversely, they found that endothelial cells (key components of blood vessel walls) are regulated by many tissue-specific factors and differ by organ. Therefore, in trying to understand functional relationships between these subtypes and other cells during development, tissue context might figure more prominently in endothelial cell variability than it does in other lineages. Cao et al. precisely annotated three subtypes of red blood cell (erythroid) progenitor, each representing a different stage of maturation, and measured their presence across organs. They identified early erythroid progenitors in the adrenal gland, a previously unknown site of erythroid development, which might bridge the developmental switch in the site of production of these cells from fetal liver to bone marrow. Domcke and colleagues also used the paired data sets to assess the relationship between gene expression and its regulation. They identified previously unknown transcription factors specific to discrete developmental stages by analysing transcription-factor binding sites in accessible chromatin regions. Then, on the basis of the relationships between expression of cell-typespecific transcription factors and binding-site availability, they assigned putative functions to these transcription factors as activators or repressors (Fig. 1). These studies represent the next generation of atlas papers. Currently, standard cell atlases characterize a single organ on a molecular level using one data modality. The new work provides a road map for unifying these disparate data sets. A limitation of the work is that the current Developmental biology