RNA polymerase II pausing is essential during spermatogenesis for appropriate gene expression and completion of meiosis

Abstract

Male germ cell development requires precise regulation of gene activity in a cell-type and stage-specific manner, with perturbations in gene expression during spermatogenesis associated with infertility. Here, we use steady-state, nascent and single-cell RNA sequencing strategies to comprehensively characterize gene expression across male germ cell populations, to dissect the mechanisms of gene control and provide new insights towards therapy. We discover a requirement for pausing of RNA Polymerase II (Pol II) at the earliest stages of sperm differentiation to establish the landscape of gene activity across development. Accordingly, genetic knockout of the Pol II pause-inducing factor NELF in immature germ cells blocks differentiation to spermatids. Further, we uncover unanticipated roles for Pol II pausing in the regulation of meiosis during spermatogenesis, with the presence of paused Pol II associated with double-strand break (DSB) formation, and disruption of meiotic gene expression and DSB repair in germ cells lacking NELF.

Fig. 1. Most active genes are differentially expressed during spermatogenesis.

a–c Genome browser views of RNA-seq at Crabp1, a SG-specific gene (a), Spo11, a SC-specific gene (b), and Tssk3, a RS-specific gene (c). d, e Volcano plots showing fold changes in RNA-seq expression versus adjusted p-values, for active genes (n = 19,823). SG compared to SC (d), and SC to RS (e). Dotted lines indicate 1.5-fold cut-off for significantly increased (red) or decreased (blue) genes, with numbers of genes passing these thresholds and p-values < 0.05 indicated on each plot. P-values from DESeq2, (Wald test, adjusted p-value calculated using Benjamin-Hochberg correction). f Heatmap showing relative RNA-seq expression in each cell type using k-means clustering and standard Euclidean distances. Shown are 17,078 genes differentially expressed in (d, e). Cluster 1 n = 3226, Cluster 2 n = 2645, Cluster 3 n = 3601, Cluster 4 n = 4024, Cluster 5 n = 1596, and Cluster 6 n = 1986. g–i Top Gene Ontology terms and significance values for clusters 1 (g), 4 (h) and 6 (i), generated with ClusterProfiler and consolidated using Revigo. Enrichment p-values were calculated by a hypergeometric test, and the p-values were adjusted using a Benjamin-Hochberg correction. j Percent of genes from each cluster that correspond to the indicated RNA biotypes, as compared to all genes. k Distribution of gene length (TSS to TES) for genes in each cluster defined in (f). Line represents median, box represents 25–75th percentile, whiskers represent 1.5X interquartile range. P-values from unpaired, two-sided Mann-Whitney U test are shown for cluster 6 vs. all other clusters. Source data are provided as Source Data Fig. 1.

Fig. 2. Gene expression changes during spermatogenesis are largely transcriptional.

a Example windows for PRO-seq analysis in promoter (TSS to TSS + 150) or gene body (TSS + 250 to TES). b Scatter plot comparing the fold changes between SG and SC in RNA-seq versus PRO-seq gene body signal (TSS + 250 to TES). Spearman’s r = 0.85, and n = 19,823. c Scatter plot as in b, for the fold changes between SC and RS. Spearman’s r = 0.63, and n = 19,823. d Box plot representation of the distribution of fold changes in RNA-seq (empty box) or PRO-seq gene body (TSS + 250 to TES) (dotted box) for the SC to RS transition. Genes are grouped based on RNA-seq status as upregulated (red, n = 1134), unchanged (gray, n = 5880), or downregulated (blue, n = 12,226) as in Fig. 1e. Line represents median, box represents 25–75th percentile, whiskers represent 1.5X interquartile range. P-values from two-sided Wilcoxon signed rank test. e Heatmaps show relative RNA-seq (left, as in Fig. 1f) compared to relative PRO-seq gene body (TSS + 250 to TES) signal (right). n = 17,078. f–h Plots of average relative signal from RNA-seq (solid line and circle) or PRO-seq in gene bodies (dotted line and square) for each cell type, for clusters 1 (f), 4 (g), and 6 (h). Source data are provided as Source Data Fig. 2.

Fig. 3. Analysis of promoter PRO-seq signal reveals that SC selective genes accumulate paused Pol II in SG.

a Heatmaps show the relative promoter PRO-seq signal (TSS to +150, left) as compared with the gene body PRO-seq signal (at right, as in Fig. 2e, right). Genes are ranked as in Figs. 1f, 2e. b–d Plots of average relative signal from PRO-seq within promoter (dotted line with triangles) or gene body (dashed line with squares) regions for each cell type. Shown are data for genes in clusters 1 and 2 (b) cluster 6 (c) or cluster 4 (d). e Genome browser view of RNA-seq and PRO-seq at Fnta, a SC-specific cluster 4 gene that has a prominent peak of promoter PRO-seq in SG. f–g Metagene plots of average PRO-seq reads in each cell type for all protein-coding genes in cluster 4 (f) or cluster 5 (g). Inset y-axis is zoomed-in to show gene body signal from +125 to +1500 nt downstream of the TSS. Source data are provided as Source Data Fig. 3.

Fig. 4. NELF-B is required for the successful completion of spermatogenesis.

a–d Immunohistochemistry using NELF-B antibody stains NELF-B-positive Sertoli (S) and germ cells (G) in testis from control PND15 (a) and PND24 (c) mice. Quantification confirms 100% positive staining in n = 50 tubules from mice at PND15 and PND24. In tubules from NELF-B cKO mice at PND15 (b) and PND24 (d), the germ cells (G) are devoid of NELF-B protein while the Sertoli (S) cells remain NELF-B positive. Quantification identified n = 10 out of n = 72 tubules (13.89% positive) at PND15, and n = 8 out of n = 50 tubules (16% positive) at PND24. Scale bar, 50 μm. We note that some NELF-B positive germ cells are detected in the NELF-B cKO, which represent cells that escaped Cre-mediated excision of NELF-B. e–h Testis cross sections of PND15 and PND24 control (n = 3, e, g) or NELF-B cKO (n = 3, f, h) mice. PND15 NELF-B cKO mice showed degeneration of the epithelium (asterisk) involving meiotic cells in some areas (f). At PND24, vacuoles are also formed, and the epithelium is disorganized with detached and fewer RS (h). Scale bar, 50 μm.

Fig. 5. NELF-B and TDP-43 are required for the proper maturation of germ cells.

a UMAP representation of all 10 somatic and germ cell type clusters, integrated across the 3 genotypes (n = 24,269 cells). b Cell clusters from (a), separated by genotype. (Control: n = 11,170, NELF-B cKO: n = 6404, TDP-43 cKO: n = 6695 cells). c Numbers of each germ cell type and somatic Sertoli cells per genotype. Cell counts were normalized by the number of mice per genotype (Control: n = 4, NELF-B cKO: n = 2, TDP-43 cKO: n = 3). Source data are provided as Source Data Fig. 5.

Fig. 6. Genes affected in the absence of NELF-B or TDP-43 are critical for SC and RS development.

a, b Volcano plots of fold changes in gene expression in SG and early meiotic cells (pL, L &Z) versus adjusted p-values, for control mice compared to NELF-B cKO (a) or TDP-43 cKO (e) mice. Dotted lines indicate 2-fold cut-off for significantly increased (red) or decreased (blue) genes, with numbers of genes passing these thresholds and p-values < 0.01 indicated on each plot. P-values from DESeq2, (Wald test adjusted p-value calculated using Benjamin-Hochberg correction). c, d Enriched functional categories among downregulated genes in NELF-B (c) or TDP-43 (d) mice, and their significance determined by ClusterProfiler. Enrichment p-values were calculated by a hypergeometric test, and the p-values were adjusted using a Benjamin-Hochberg correction. e, f Representation of downregulated genes from NELF-B cKO (e) or TDP-43 cKO mice (f), within each cluster (from Fig. 1f). A value of 1 indicates representation expected by chance. g Expression of Cluster 4 genes (n = 3916) in scRNA-seq from control, NELF-B cKO, or TDP-43 cKO mice, in indicated cell types. Line represents median, box represents 25–75th percentile, whiskers represent 10–90th percentile. P-values from two-sided Wilcoxon signed rank test. h Genome browser view of wild type PRO-seq signal in each cell type at Spo11. Note that the y axis is truncated at 50 to highlight PRO-seq gene body signal. Source data are provided as Source Data Fig. 6.

Fig. 7. Promoter activity in SG correlates with sites of double-strand breaks in SC.

a Metagene plots of SPO11-oligo reads around the TSS for all DE genes (n = 17,078). b Heatmaps of SG PRO-seq (left, dark blue) or SPO11 oligo reads (right, gray), aligned at TSSs (indicated by arrow) with genes ranked by SG promoter PRO-seq signal (TSS to TSS + 150, n = 17,078). c, d Metagene plots around the TSSs for the top (c) or bottom (d) quartile based on SG PRO-seq reads in promoters with at least 5 reads (n = 14,804, with n = 3701 genes per quartile). PRO-seq signal for SG (blue) is plotted on the left y-axis, SPO11 oligo reads (gray) is plotted with the right y-axis. e Box plots of the distribution of SPO11 reads at each quartile (n = 3701) of SG PRO-seq reads at promoters with at least 5 reads (n = 14,804). Line represents the median, whiskers represent 1.5X interquartile range. P-values from two-sided Wilcoxon signed rank test. f Testis cross sections of PND24 control (n = 2, left) or NELF-B cKO (n = 2, right) mice. L/Z, indicates SC in Leptotene/Zygotene stages, and P, indicates pachytene SCs. Scale bar, 50 μm. g Quantification of SC at Leptotene/Zygotene versus Pachytene stages per tubule. Shown are the mean and range of n = 2 biologically independent animals. P-value is from Unpaired two-sided t-test. Source data are provided as Source Data Fig. 7.

Fig. 8. NELF-B cKO mice have impaired DSB repair.

a SC showing SYCP3 (red), ɣ-H2AX (green) and DAPI (blue) in control (top row) and NELF-B cKO (bottom row). ɣ-H2AX signal is restricted to the sex body (red arrow) in control whereas it persists on autosomes (white arrows) in NELF-B cKO indicating impairment of DSB repair. Scale bars = 20 μm. At right: Quantification of relative ɣ-H2AX area (ɣ-H2AX/DAPI) between the control (n = 40) and NELF-B cKO (n = 52) from three biological replicates. Mean indicated with horizontal line. Error bars represent standard error of the mean. P-value from two-sided t-test. b SC showing SYCP3 (green) and RPA (red) in control (top row) and NELF-B cKO (bottom row); the higher number of RPA foci in NELF-B cKO implies impaired DSB repair. Scale bars = 20 μm. At right: Quantification of the RPA foci numbers at pachytene from control (n = 51) and NELF-B cKO (n = 46) from three biological replicates. Mean indicated with horizontal line. Error bars represent standard error of the mean. P-value from two-sided t-test. Source data are provided as Source Data Fig. 8.