A limited set of transcriptional programs define major cell types

Abstract

The ENCODE project has produced a collection of RNA sequencing experiments from many cell lines and bulk tissues that constitutes an extensive catalogue of the expression programs utilized in the human body. However, the relationship between the transcriptomes of tissues and those of the constituent primary cells, and how these impact tissue phenotypes has not been well established. Here we have produced RNA sequencing data for a number of primary cells from ten human body locations. The analysis of this data, together with additional epigenetic data from a total of 146 primary cells, indicates that many cells in the human body belong to five major cell types of similar transcriptional complexity: three, epithelial, endothelial, and mesenchymal, are broadly distributed across the human body acting as components for many tissues and organs, and two, neural and blood cells, are more anatomically localized. Based on gene expression, these redefine the basic histological types by which tissues have been traditionally classified.We have identified genes whose expression is specific to these cell types, and have estimated the relative proportion of the major cell types in human tissues using the transcriptional profiles produced by the GTEx project. The inferred cellular composition is a characteristic signature of tissues and reflects tissue morphological heterogeneity and histology. We identified changes in cellular composition in different tissues associated with age and sex and found that departures from the normal cellular composition correlate with histological phenotypes associated to disease. This transcriptionally based classification of human cells provide a new view of human biology and disease.2 Transcriptional profiles reflect cell type, condition and function. In tissues and organs, they are monitored in RNA extracted from millions to billions of cells (11 6 − 10 9 ) 1 , likely including multiple cell types. As a consequence, the transcriptional profiles obtained from tissue samples represent the average expression of genes across heterogeneous cellular collections, and gene expression differences measured in bulk tissue transcriptomes may thus reflect changes in cellular composition rather than changes in the expression of genes in individual cells. Single-cell RNA sequencing (scRNA-seq) has indeed revealed large cellular heterogeneity in many tissues and organs 2 , and the Human Cell Atlas (HCA) project 3 has been recently initiated with the aim of defining all human cell types and to infer the cellular taxonomy of the human body.As a step in that direction and to bridge the transcriptomes of tissues with the transcriptomes of the constituent primary cells, and to understand how these impact tissue phenotypes, we have generated bulk expression profiles of 53 primary cell lines isolated from ten different anatomical sites in the human body.These profiles include long and short strand-specific RNA-seq, and RAMPAGE data (Fig. 1a, Table S1-4). Major cell types in the human bodyClustering of the primary cell...

The impact of sex on gene expression across human tissues

“…Male breast samples diagnosed with gynecomastia (N = 69), a male-specific condition involving an enlargement or swelling of breast tissue, exhibited more epithelial cells and fewer adipocytes than non-diagnosed samples (PEpithelial cells = 1.91e-14, PAdipocytes = 1.71e-14, Fig. 4C), as previously reported (22). Gynecomastia is characterized by altered levels of steroid hormones and their receptors ( 23), as well as TRHmediated prolactin levels (24); among other genes, TRH expression in breast differed between males with and without gynecomastia (Fig.…”

Perspectives on ENCODE

“…An important part of ENCODE 3 is that the regulatory mapping efforts have now been integrated and synthesized into the first version of an encyclopedia, highlighting a registry of 0.9 million cCREs in human and 0.3 million cCREs in mouse. Details can be found in the accompanying ENCODE paper 8 and companion papers in this issue and other journals [9][10][11][12][13][14] .…”

Evolutionary cell type mapping with single-cell genomics

“…An extension of the candidate marker gene profiling is the analysis of genome-wide gene expression using bulk transcriptomics. Pioneering studies in cell type bulk transcriptomics provided systematic cell classifications schemes [17], revealed the hierarchical structure of cell type transcriptomes [18], and enabled the first attempts at building phylogenies of closely related cell types based on their gene expression profiles [19]. However, these enrichment strategies are necessarily limited to cell lines [17,[20][21][22] -virtually non-existent in the vast majority of organisms-or homogeneous cell populations isolated manually or by FACS-sorting [10,23,24] which require dedicated methods and do not ensure purity.…”

“…Pioneering studies in cell type bulk transcriptomics provided systematic cell classifications schemes [17], revealed the hierarchical structure of cell type transcriptomes [18], and enabled the first attempts at building phylogenies of closely related cell types based on their gene expression profiles [19]. However, these enrichment strategies are necessarily limited to cell lines [17,[20][21][22] -virtually non-existent in the vast majority of organisms-or homogeneous cell populations isolated manually or by FACS-sorting [10,23,24] which require dedicated methods and do not ensure purity. High-throughput single-cell RNA sequencing methods (scRNA-seq) overcome many of these limitations, ultimately facilitating the minimally biased sampling and molecular characterization of thousands of single-cells, and setting the stage for in silico reconstruction of cell type repertoires in species that were so far difficult to study.…”

“…By analogy, cell type taxonomies remain Linnean in the sense that they do not incorporate evolutionary considerations -whether or not implying the existence of staggered ranks, e.g. broad cell type identities [17]. In fact, cell type classifications very often include hierarchical tree representations based on, for example, transcriptome similarities.…”