Fig. 2: Identification and annotation of larval testis cell types from scRNA-seq and comparison to bulk testis RNA-Seq.
From: Dynamic sex chromosome expression in Drosophila male germ cells
a Illustration of the 3rd instar larval testis, showing germline cell types spermatogonia (G); early- (E1o), middle- (E2o), and late- (L1o) primary spermatocytes; four groups of somatic cysts cells (C1–C4), terminal epithelium precursors (T) and pigment cells (P). These cell types are from germline or soma linages (dashed boxes). See abbreviation and color code key used throughout this manuscript. b Consistency of cell type expression profiles. Uniform Manifold Approximation and Projection (UMAP) for each scRNA-seq replicate. Each cell position in multi-dimensional space projected onto the two-dimensional UMAP space and clustered using k-nearest neighbors. c Correlation between whole testis and single cell profiles. Density scatterplot (high density, red and low density: blue); regression line and Spearman rank correlation, ρ) of gene expression from the sum of all cells from the larval testis scRNA-seq (x-axis) versus bulk larval testis RNA-Seq (y-axis). Transcripts Per Million reads +1 (TPM + 1). d Distribution of genes showing testis-biased expression in bulk RNA-Seq using Gene Set Enrichment Scores (y-axis) among cell types (x-axis) defined by L3 testis scRNA-seq. Each bar is a cell type, so sample size is the number of cells in that cell type: G = 1065, E1 = 1267, M1 = 4570, L1 = 2667, C1 = 4064, C2 = 1230, C3 = 910, C4 = 945, T = 1379, P = 868. Boxplots (box = interquartile range (IQR), notch = 95% 736 confidence interval of median, whiskers = ±1.5xIQR). e Patterns of differential gene expression among putative cell types. Only differentially expressed genes are shown (two-sided Wilcoxon rank-sum test, FDR corrected P ≤ 0.01). Gene expression is summarized (Z-score; low: blue, high: yellow) for each differentially expressed gene (rows) for all cell-type-by-replicate (columns). Genes are ordered and grouped (horizontal dashed lines) manually into 16 classes (numbered) based on differential gene expression. Each class is annotated with the number of genes (parentheses) and the key genes used to identify the cell types (Supplementary Dataset 2). f–k Immunofluorescence images showing protein expression of representative genes from the protein trap reporters used for cell type annotation. f p53, g Add1 h bol i rdo j Nlg3, and k Piezo. Each panel (counter clockwise from top left) consists of color (Fas III, 1B1: red, DAPI: blue, and GFP: green) and grayscale (GFP) images of protein trap expression, normalized gene expression values (normalized to maximum expression) plotted by cell type (Each bar is a cell type, so sample size is the number of cells in that cell type: G = 1065, E1 = 1267, M1 = 4570, L1 = 2667, C1 = 4064, C2 = 1230, C3 = 910, C4 = 945, T = 1379, P = 868. Error bars are bootstrapped 95% confidence intervals around the mean), and scRNA-seq gene expression projected onto the UMAP (Z-score; low: blue, high: yellow, see key).
