Single-cell analysis of human ovarian cortex identifies distinct cell populations but no oogonial stem cells

Abstract

The human ovary orchestrates sex hormone production and undergoes monthly structural changes to release mature oocytes. The outer lining of the ovary (cortex) has a key role in defining fertility in women as it harbors the ovarian reserve. It has been postulated that putative oogonial stem cells exist in the ovarian cortex and that these can be captured by DDX4 antibody isolation. Here, we report single-cell transcriptomes and cell surface antigen profiles of over 24,000 cells from high quality ovarian cortex samples from 21 patients. Our data identify transcriptional profiles of six main cell types; oocytes, granulosa cells, immune cells, endothelial cells, perivascular cells, and stromal cells. Cells captured by DDX4 antibody are perivascular cells, not oogonial stem cells. Our data do not support the existence of germline stem cells in adult human ovaries, thereby reinforcing the dogma of a limited ovarian reserve.

Fig. 1. Human ovary and study design.

a A representative ovary cross-section of a 26-year-old GRP showing the thin cortex layer (∼1 mm, used for fertility preservation) harboring clusters of small primordial follicles that form the ovarian reserve. Beneath the cortex is the inner part of the ovary, medulla, with looser connective tissue, denser vasculature, and larger growing follicles. Yellow rectangle is enlarged to highlight the ovarian reserve composed of primordial follicles resting in the cortex. One early growing (primary) follicle is indicated (arrowhead). Red rectangle is enlarged to highlight the more prominent vasculature in the medulla. Scale bars: 2 mm (overview), 250 µm (zoom-in). b Schematic representation of the study. In total, ovarian cortical tissue from 16 GRPs (mean age 26y) and five C-sec patients (mean age 34y) were used. Ovaries were trimmed upon tissue receival to separate the cortex. Cortex was cut into pieces, and the quality was assessed via general histology and viability of GRP follicles was further confirmed with neutral red staining. The samples were cryopreserved. Thawed ovarian cortex was enzymatically digested into single-cell suspension, filtered, and subjected either uncultured to single-cell transcriptome and surface proteome profiling using 10xGenomics technology and cell surface marker screening panel or cultured under oogonial stem cell (OSC) conditions and subjected to single-cell transcriptome analysis using Smart-seq2. In all assays, cells were stained with DDX4 Ab in order to relate the findings to DDX4 Ab+ (suggested OSCs) and Ab− ovarian cell populations. In the final data analysis, cortical cell populations were identified, related to cultured DDX4 Ab+ and Ab− cells, and compared with cells in the inner part of human ovaries and fetal ovarian cell populations. Expression of key markers was validated with immunostainings, qPCR and RNA-FISH. Ab antibody, C-sec cesarean section, GRP gender reassignment patient, qPCR quantitative real-time PCR, RNA-FISH RNA-fluorescent in situ hybridisation, scRNA-seq single-cell RNA-sequencing.

Fig. 2. Transcriptome analysis of unsorted ovarian cells.

Ovarian cortical cells from GRP and C-sec patients were sequenced on a single-cell level and integrated with ovarian medulla cells. a After quality control and filtration, 5715 GRP cells (black) and 6445 C-sec cells (grey) were further analyzed. Non-hierarchical cluster analysis revealed six main clusters, and both GRP and C-sec cells contributed to all clusters to a similar extent. b Main cell cluster identities were identified with the help of gene expression analysis. c Heatmap showing enriched genes per cluster. Clusters were downsampled to max. 300 cells per cluster for better visualization. d The color-coded violin plots show the expression scores of signature genes. e UMAP of merged scRNA-seq data of our unsorted ovarian cortex cells (grey) and the publicly available scRNA-seq data set of antral follicles and pieces of stroma picked from ovarian medulla (colored cells). Previous cluster annotations of both data sets are maintained. f Same UMAP as in e, showing the contribution of our cortex cells (colored) to previously reported ovarian medulla clusters (grey). All identified cortical cell clusters joined the medullary cell clusters except from the oocytes that were unique to the cortex data set. The horizontal bars in violin plots indicate median gene expression per cluster. C-sec Cesarean section patient, CL cluster, G granulosa, GRP gender reassignment patient, I immune, TC theca cell, O oocytes, UMAP uniform manifold approximation and projection.

Fig. 3. Analysis of DDX4-expressing somatic cells of the ovarian cortex.

Somatic ovarian cortical cells expressing DDX4 were manually pooled for analysis of transcriptional profile. a Heatmap showing representative marker genes of OSCs and pluripotent cells by cell cluster. b Feature and violin plots highlighting all somatic cells expressing DDX4 transcript. Expression is mainly localized to the oocyte cluster (pink circle in feature plot) and dispersedly in 15 cells from other clusters (labeled with * for clarity). The 15 somatic cells expressing DDX4 were manually selected to form their own cluster (ddx4* in violin plot). c Violin plots displaying the expression scores of reported OSC marker genes (DPPA3, DAZL, PRDM1), germline and/or pluripotency-associated genes (POU5F1, NANOG, TFAP2C) and oocyte marker genes (GDF9, ZP3, OOSP2, FIGLA) across all cell clusters. Ddx4* cluster does not show differential expression of these markers. The horizontal bars in violin plots indicate median gene expression per cluster. G granulosa, I immune, O oocytes, UMAP uniform manifold approximation and projection.

Fig. 4. Transcriptome analysis of sorted DDX4 Ab+ and Ab− ovarian cortex cells.

Ovarian cortex cells were sorted into DDX4 Ab+ or Ab− populations and sequenced on a single-cell level. a After quality control and filtration, 5479 of Ab+ and 6690 of Ab− cells were available for downstream analysis. Non-hierarchical cluster analysis revealed that 82.5% of the DDX4 Ab+ cells form a distinct cluster while all other clusters mainly consisted of DDX4 Ab− cells. b Differential gene expression analysis revealed seven main cell types. DDX4 Ab− cells mainly contributed to stroma, endothelial cells, granulosa cells, T-cells, monocytes, and oocytes. DDX4 Ab+ cells made up the cluster identified as perivascular cells. c Violin plot showing the expression of DDX4 among the different clusters (left) as well as among the sorted Ab+ and Ab− populations (right). d Comparison of sorted scRNA-seq data to single-cell transcriptome analysis of cultured DDX4 Ab+ and Ab− ovarian cells. Top 25 upregulated genes of cultured DDX4 Ab+ and Ab− cells were superimposed as expression score on to our uncultured sequenced DDX4 Ab+ and Ab− cells. The upregulated genes from cultured DDX4 Ab+ cells (scale bar depicting low expression in grey and high expression in red) are highly expressed in the perivascular and endothelial clusters, whereas the upregulated genes in cultured DDX4 Ab− cells (scale bar depicting low expression in grey and high expression in blue) are not found in these clusters. UMAP uniform manifold approximation and projection.

Fig. 5. Validation of perivascular localization of DDX4 Ab signal.

a Feature plots displaying expression of MCAM (green) and RGS5 (red), both found among top highly expressed genes in DDX4 Ab+ cells, and their co-expression (yellow) on a transcriptional level. Both markers are highly expressed in cells of the perivascular cluster. Scale depicting low expression in dark blue and high expression in green, red, and yellow. b Immunostaining of human ovarian tissue sections with MCAM and RGS5, co-stained with the endothelial marker CD31, showed a distinct staining of cells surrounding endothelial cells of blood vessels. In a consecutive section, MCAM is co-expressed in DDX4 Ab+ cells. Oocytes staining brightly positive for DDX4 Ab+ are marked with white asterisks. As negative control, primary antibodies were omitted. DAPI (blue) was used as nuclear counterstain. Orange rectangles in last column (scale bars: 200 µm) demarcate the zoomed-in area (scale bars: 50 µm). Staining results were screened in three histological sections from two patients and a representative image is shown; all visible vessels or follicles were stained positive for MCAM/RGS5/DDX4 or DDX4 only, respectively. c RNA-FISH on ovarian cortex sections from a GRP donor showing the localization of DDX4 transcript in oocytes but not in blood vessel cells. Red rectangles are demarcating the zoomed-in areas containing follicles (F) or blood vessels (V) further highlighted with white dashed circles or lines, respectively. Nuclei were counterstained using Hoechst (blue). Positive control shows expression of Ubiquitin C. As negative control, probes targeting the bacterial DabB gene were used. Scale bars: 200 µm (zoomed-out), 100 µm (zoomed-in). RNA-FISH was performed on three histological sections of three donors, respectively, and representative images are shown. NC negative control, PC positive control.

Fig. 6. Screening of ovarian cortex cells for surface proteins.

The surface of ovarian tissue cells was screened for the expression of 242 CD markers by flow cytometry. a Dot plot showing the log median fluorescence intensity (MFI, x axis) and the percentage of CD-positive cells (y axis) for those 43 CD markers that were repeatedly found to be present on ovarian tissue cells (left, CD marker expression values of repeat number three are shown as representatives). When co-stained with DDX4 Ab, seven of the markers were highly expressed (red) and seven were absent or lowly expressed (blue) on DDX4 Ab+ cells when compared with the DDX4 Ab− cell population. Co-expression of DDX4 and the 14 markers identified as high or low in DDX4 Ab+ cells is shown in representative contour plots (right). X axes display intensity of DDX4 Ab signal, y axes display intensity of CD marker signal. b Feature plots showing expression scores of the seven cell surface markers highly expressed in DDX4 Ab+ cells (scale bar depicting low expression in grey and high expression in red) and the seven markers lowly expressed in DDX4 Ab+ cells (scale bar depicting low expression in grey and high expression in blue) overlaid with our sorted scRNA-seq data. While markers that were found to be highly expressed in DDX4 Ab+ cells on a protein level are mainly expressed in perivascular cluster, CD markers lowly expressed in DDX4 Ab+ cells show a more disperse expression pattern. c FACS dot plots showing the co-expression of MCAM and CD9, two markers brightly expressed on DDX4 Ab+ cells, in regard to DDX4 Ab+ and Ab− cell populations. Around 3.67% of ovarian cortex cells were double positive for MCAM and CD9 (in red) and 5.48% are DDX4 Ab+. Around 80% of MCAM/CD9 double positive cells co-stained with DDX4 Ab (2.94% of MCAM⁺/CD9⁺/DDX4⁺ cells in total). Flow analysis was repeated in two independent experiments. d Unfiltered lysate of ovarian cortex cells cytospinned and stained for MCAM (green), CD9 (red), and DDX4 (magenta) showing co-expression of all three markers on blood vessels. As negative control, primary antibodies were omitted. DAPI (blue) was used as nuclear counterstain. Scale bars: 50 µm. Immunostaining was performed in duplicates in two independent experiments and a representative image is shown. All visible vessels stained positive for MCAM/CD9/DDX4. FACS fluorescence-activated cell sorting, MFI median fluorescence intensity, NC negative control, UMAP uniform manifold approximation and projection.

Fig. 7. Investigation of potential germline stem cells in the adult human ovarian cortex.

Data sets of unsorted and sorted ovarian cortex cells were used individually for integration with scRNA-seq data set of human fetal tissue cells. a UMAP plot showing sample origin of fetal (black) and adult cells (C-sec data, grey). b UMAP showing nine clusters representing different cell types annotated based on differential gene expression analysis. The same clusters as in adult tissue were found, as well as three new ones: mitotic fetal germ cells, retinoic acid responsive FGCs, and meiotic FGCs. c Table listing cell numbers from adult and fetal data sets per annotated cluster. One cell of C-sec sample, one cell of DDX4 Ab+ sample and two cells of DDX4 Ab− sample clustered with mitotic FGCs of the fetal sample (in red). These four cells were manually clustered together and studied further with violin plots. d Violin plots displaying expression scores of known pluripotency, germline and oocyte markers in the different FGC clusters in addition to adult oocyte cluster and the four adult ovarian cells clustering with mitotic FGCs. C-sec Cesarean section patient, FGCs fetal germ cells, GRP gender reassignemnt patient, NA non-annotated, UMAP uniform manifold approximation and projection.