Single-cell RNA-Seq reveals cell heterogeneity and hierarchy within mouse mammary epithelia

Abstract

The mammary gland is very intricately and well organized into distinct tissues, including epithelia, endothelia, adipocytes, and stromal and immune cells. Many mammary gland diseases, such as breast cancer, arise from abnormalities in the mammary epithelium, which is mainly composed of two distinct lineages, the basal and luminal cells. Because of the limitation of traditional transcriptome analysis of bulk mammary cells, the hierarchy and heterogeneity of mammary cells within these two lineages remain unclear. To this end, using single-cell RNA-Seq coupled with FACS analysis and principal component analysis, we determined gene expression profiles of mammary epithelial cells of virgin and pregnant mice. These analyses revealed a much higher heterogeneity among the mammary cells than has been previously reported and enabled cell classification into distinct subgroups according to signature gene markers present in each group. We also identified and verified a rare CDH5⁺ cell subpopulation within a basal cell lineage as quiescent mammary stem cells (MaSCs). Moreover, using pseudo-temporal analysis, we reconstructed the developmental trajectory of mammary epithelia and uncovered distinct changes in gene expression and in biological functions of mammary cells along the developmental process. In conclusion, our work greatly refines the resolution of the cellular hierarchy in developing mammary tissues. The discovery of CDH5⁺ cells as MaSCs in these tissues may have implications for our understanding of the initiation, development, and pathogenesis of mammary tumors.

Keywords: CDH5+ cell; biomarker; breast cancer; cell differentiation; development; mammary gland; mammary stem cells (MaSCs); single cell RNA-seq; stem cells.

Figure 1.

Single-cell transcriptome profiling of mouse mammary epithelium. A, schematic illustration of the workflow. The 4th pairs of mouse mammary glands were digested and enriched for single epithelial cells. Basal and luminal cells were sorted out separately and captured on Fluidigm C1 chips for RNA-seq library preparations. Then the libraries were constructed and sequenced, and the data were analyzed. B, FACS result of mouse mammary cells. CD24^MidCD29^Hi cells were sorted out as basal cells (enriched with MaSCs), whereas the CD24^HiCD29^Lo cells were sorted out as luminal cells. C, distribution of basal and luminal cells in virgin and P12 mouse mammary glands. Data are shown as mean ± S.D. D, summary of the numbers of expressed genes with FPKM >1 (left) and FPKM >10 (right) in an individual cell of each cell type.

Figure 2.

Hierarchy and heterogeneity analysis of mouse mammary epithelium. A, PCA of 239 single cells. B, heterogeneity measured as correlation coefficient of individual samples within each group based on gene expression profiles. C, hierarchy clustering of 239 single mouse mammary cells and expression pattern of some selected genes. The single cells are clustered into the main six subgroups based on overall gene expression profiles, and cluster 2 (C2) was further divided into C2A and C2B subclusters. The expression pattern of some typical genes, such as housekeeping (HK) genes, basal (Bas), and luminal (Lum) markers and six reported MaSCs markers are also shown.

Figure 3.

One rare subpopulation of basal cells bears bi-lineage gene signature. A, tSNE distribution of all the single cells. The cells are marked as the given subgroups separately according to hierarchy clustering. B, summary of the numbers of expressed genes with FPKM >1 (left) and FPKM >10 (right) in an individual cell of each cluster. C, expression pattern of some C3 highly expressed genes in the single cells. D, protein–protein interaction network analysis among the C3 highly expressed genes via STRING. E, functional analysis of enriched GO terms of C3 highly expressed genes via DAVID.

Figure 4.

Expression pattern of Cdh5 in mouse mammary gland. A, expression pattern of Cdh5 in the single cells. B, relative mRNA expression levels of Cdh5, Krt14, and Krt8 in mouse basal and luminal cells sorted from 3-month-old virgin and P12 mice. Bas and Lum, are short for basal and luminal cells; 3M and P12, are for mammary cells from 3-month-old virgin and P12 mice. C, summary of flow cytometry analysis of CDH5 expression in mammary basal and luminal lineages. Data are shown as mean ± S.D.

Figure 5.

CDH5⁺ basal cells serve as mouse MaSCs. A, mammosphere assay of CDH5⁺ and CDH5⁻ basal cells sorted from 3-month-old virgin female mice. B, mammary epithelial reconstruction assay. C, top GO terms from GO analysis results of differentially expressed genes (Cdh5⁺ VB versus Cdh5⁺ PB) enriched in Cdh5⁺ VB or PB single cells. D, average expression levels of some differentially expressed genes (Cdh5⁺ versus Cdh5⁻ VB or PB cells) and their enriched functions (yellow and green colors stand for enriched GO terms in Cdh5⁺ and Cdh5⁻ cells, respectively).

Figure 6.

Reconstruction of the mammary epithelial developmental trajectory. A, single cells from virgin and P12 mice could be further divided into five (left panel) and six (right panel) subgroups with separate distinct markers. LBLCs is short for lipid biosynthetic luminal cells; KLCs is short for keratinized luminal cells; SRLCs is short for stimulus-responsive luminal cells; PBCs is short for proliferative basal cells; WRBCs is short for Wnt signaling–responsive basal cells; Mature LCs is short for mature luminal cells; and PLCs is short for proliferative luminal cells. B, Monocle analysis reconstructs the differentiation trajectory of mouse MaSCs in virgin and P12 mice (top), and the pseudo-differentiation trajectory of MaSCs are shown at bottom.

Figure 7.

Differentially expressed genes and related functions during MaSCs differentiation. A–C, differentially expressed genes and related functions during MaSCs differentiation in virgin (A) and P12 mice (B for basal lineage and C for luminal lineage). Left, the expression trends of each cluster of genes; middle, the heatmap of differentially expressed genes along the pseudo-differentiation trajectory of MaSCs; right, enriched GO terms and representative genes of each gene cluster.