All animal procedures were approved by the University of Newcastle Animal Care and Ethics Committee (ACEC, approval A2019-907). Mice were housed under a controlled lighting regimen at 21–22°C and supplied with food and water ad libitum.

Transcriptome analysis was conducted using previously published 10x Genomics Datasets (GSE109033 and GSE163027). Detailed information on the generation of these datasets can be retrieved from the original manuscripts (testis dataset: Hermann et al. (2018), culture dataset: Cafe et al. (2021b)). Briefly, both datasets utilised spermatogonia from mice on a C57BL/6J background. Spermatogonia used to create the ‘culture’ scRNA-seq dataset originated from a Rosa26-LacZ mouse line (Jackson Laboratories, stock number 112073), while testis cells were isolated from Rosa26-LacZ mice that also carried an Id4-eGfp transgene (Chan et al., 2014). These transgenes are utilised as markers for downstream analysis and do not otherwise alter the biology of the spermatogonial population (Chan et al., 2014). Spermatogonia in the adult testis dataset were isolated via fluorescence-activated cell sorting (FACS) (CD9^Bright/ID4-eGFP+, >2 replicates), and the presence of bona-fide stem cells was confirmed via transplantation analysis (Hermann et al., 2018). Undifferentiated cultures of spermatogonia were initially established using the THY1+ contingent of undifferentiated spermatogonia from the adult testis, however were maintained for 10 weeks (i.e. 10 passages) using “glycolysis optimised” culture conditions (10% O₂, 5% CO₂) (Helsel et al., 2017a) before being isolated from feeder cells and prepared for scRNAseq (3 replicates). Stem cell content within cultures was also confirmed using spermatogonial transplantation (Cafe et al., 2021b). It should be noted that we have previously demonstrated that THY1+ selection does indeed capture the ID4-GFP+ contingent of spermatogonia from the adult mouse testis (Lord et al., 2018), thus, although different approaches were used for spermatogonial enrichment in the aforementioned studies, the populations being selected for are largely analogous.

“Culture” and “testis” transcriptomes were imported into Seurat (version 4.03) and merged into a single object (Butler et al., 2018). Low quality cells or doublets with unique feature counts less than 200 or more than 6,500, respectively, or with >25% mitochondrial counts, were filtered from the dataset. Data were normalised using the ‘NormalizeData’ function in Seurat, and were integrated using the ‘Harmony’ algorithm (Korsunsky et al., 2019). The “FindVariableFeatures” function was used to identify variable genes for use in principal component analysis. For clustering and UMAP graphing, 10 significant principal components were used (resolution set to 0.5).

Unique cluster markers and differentially expressed genes (DEGs) between selected clusters were determined using the “FindAllMarkers” and “FindMarkers” functions in Seurat, respectively. Qualitative analysis of gene lists was conducted using the DAVID Bioinformatics Resources (V6.8) (Huang et al., 2008; Huang et al., 2009) platform.

Immunoblotting was conducted as described previously (Lord et al., 2015). Protein extraction was conducted on the SSC-enriched THY1+ contingent of spermatogonia from a postnatal day 8 testis, or alternatively, the equivalent population of undifferentiated spermatogonia following ten passages of in vitro culture (Cafe et al., 2021b). Postnatal day 8 testes were used for this purpose to avoid contamination by elongating spermatids, which often elute from the MACS column following THY1+ spermatogonia selection from the adult testis and cannot be excluded from immunoblotting analyses as they can with scRNA-seq. Protein extraction was achieved via a 5 min incubation at 100°C in SDS extraction buffer (2% [w/v] SDS and 10% [w/v] sucrose in 0.1875 M Tris, pH 6.8) containing a protease inhibitor cocktail (Roche). Lysates were resolved on a 4–12% polyacrylamide gel (BioRad) and transferred to a nitrocellulose membrane using standard techniques. Membranes were blocked in 3% bovine serum albumin (BSA) diluted in Tris-Buffered Saline with 0.1% Tween (TBST). Primary antibodies utilised were anti-ubiquitin (Thermo Fisher Scientific, MA1-10035, 1:1000) and anti-α-tubulin (T5168, Sigma Aldrich, 1:4000). Following an overnight incubation in primary antibody (diluted in 1% BSA/TBST), membranes were washed 3 times in TBST and incubated in goat anti-mouse IgG HRP (1/10,000) for 1 h at room temperature. Blots were developed using ECL reagent (GE Healthcare) an imaged on a ChemiDoc Imaging System (BioRad). Densitometric analysis was conducted using the public sector image-processing program ImageJ (version 1.52A; National Institutes of Health), and values were normalised to the loading control (tubulin).

Analysis of scRNA-seq datasets is described in detail above. All additional experiments were conducted using a minimum of three biological replicates. Quantitative data are presented as mean ± SEM. Differences between means was determined statistically using the t-test function of GraphPad Prism 9 software. A value of p < 0.05 was considered to be statistically significant.

In order to assess how periods of in vitro culture influence gene expression in the undifferentiated spermatogonia population, a scRNA-seq comparison was conducted using previously published datasets comprised of adult mouse spermatogonia (Hermann et al., 2018) (from herein referred to as the “testis” dataset), and our own undifferentiated adult mouse spermatogonia that had been subjected to 10 weeks of in vitro culture (i.e. passage 10) (Cafe et al., 2021b) (herein referred to as the “culture” dataset). A merged dataset was created using Seurat (Butler et al., 2018), which contained 6,650 cells from the testis (enriched for spermatogonia but also containing some testicular somatic cells), and 784 undifferentiated spermatogonia from culture (Figure 1A; Supplementary Figure S1A). Each cell from the testis dataset had an average of 12,908 unique molecular indices (UMIs) and 3,840 median genes per cell, while the culture dataset had an average of 19,193 UMIs and 4,255 median genes per cell.

In order to negate batch effects, the Harmony algorithm (Korsunsky et al., 2019) was utilised for integration of the two datasets. This method was selected based on previously published comparative studies that reported superior outcomes using Harmony, when compared to other data integration approaches (Tran et al., 2020). Indeed, following Harmony integration, it could be appreciated that undifferentiated spermatogonia from the ‘culture’ dataset aligned with the undifferentiated spermatogonia from the ‘testis’ dataset (Supplementary Figure S1A), creating a population that was distinct from the differentiating germ cells and somatic cells, as expected (Figure 1A; Supplementary Figure S1A).

Unsupervised clustering projected onto Uniform Manifold Approximation and Projection (UMAP) analysis plots revealed 18 populations in the Harmony integrated dataset (Supplementary Figure S1B). Based on the expression of known markers for spermatogonial sub-populations and somatic cells in the testis (which were used to classify cell populations in the original ‘testis’ dataset (Hermann et al., 2018), and in other previously published mouse testis datasets (Law et al., 2019)), identities were assigned to these 18 clusters, resolving nine broader populations of cells: undifferentiated spermatogonia (encompassing four sub-populations of cells on the spectrum from SSC to progenitor spermatogonia, see Supplemental Dataset one and Supplementary Figure S1C for marker expression), early differentiating spermatogonia, differentiating spermatogonia, late differentiating spermatogonia/early spermatocytes, peritubular myoid cells, endothelial cells, macrophages, innate lymph, and a fibroblast cell cluster that is primarily comprised of SNL 76/7-4 mouse embryonic fibroblast feeder cells that were carried through in the ‘culture’ scRNAseq dataset (Figure 1A). A complete list of cluster markers is provided in Supplemental Dataset 1, and feature plots depicting expression of a subset of these cluster markers are provided in Supplementary Figure S1C.

Given the discrepancy between cell numbers in the original testis and culture datasets, and the known capacity for such discrepancies to create challenges for efficient data integration and clustering (Forcato et al., 2020), we elected to focus on the undifferentiated spermatogonia population specifically (circled in Figure 1A), given that differentiating spermatogonia and somatic cells are not a component of the ‘culture’ dataset and are not of interest for addressing the specific aims of this manuscript. Re-clustering of the Harmony-integrated undifferentiated spermatogonia population was performed Supplementary Figure S2A), revealing 12 populations of cells (Figures 1B, C). Effective clustering was confirmed via heatmap analysis (Figure 1D), and a subset of cluster markers (and the number of cells per cluster) are listed in Figure 1D. A complete list of unique cluster markers is provided in Supplementary Dataset S1. To assign identities to spermatogonial sub-populations, expression of known SSC markers (Etv5, Bcl6b, and Gfra1), progenitor markers (Sox3 and Neurog3), and markers of differentiation (Kit and Stra8) were assessed (Figure 1E; Supplementary Figure S2B; Supplementary Dataset S1) (Hermann et al., 2018). Based on the elevated expression of SSC markers, and absence of expression of progenitor and differentiating spermatogonia markers, clusters 0, 1, 2, 4 and 11 were categorised as SSCs (Figure 1E; Supplementary Figure S2B). It was noted that progenitor spermatogonia fell in to two ‘sub-categories’ in which they either displayed dual expression of Sox3 and Neurog3 (denoted ‘Progenitor A’, capturing clusters 5, 8 and 9), or expressed Sox3 only, along with increased expression of Stra8 (denoted ‘Progenitor B’, capturing clusters 3, 6, 7 and 10) (Figure 1E; Supplementary Figure S2B). This categorisation aligns with SOX3+/NEUROG3+ and SOX3+/NEUROG3- subsets of progenitors that have been previously identified in the testes using antibody-based techniques (Nakagawa et al., 2021).

In viewing UMAP plots that depict the dataset origins of each cell, it was clear that although the undifferentiated spermatogonia do largely congregate together, there are a number of clusters that are preferentially occupied by cells from either testis or culture datasets. For instance, 90% of cells in cluster SSC_0 originated from the testis dataset, whereas >95% of cells in SSC_11 and >70% of cells in SSC_1 originated from the culture dataset (Figures 1C; Figure 2A). Also interesting was the observation that progenitor spermatogonia from culture only occupied the ‘B’ sub-class (predominantly cluster Progenitor_B3), while progenitors from the testis occupied both ‘A’ and ‘B’ clusters ((Figures 1C; Figure 2A). In focusing on the distribution of SSCs specifically amongst clusters; >40% of testis SSCs occupied the SSC_0 cluster, followed by 27% in SSC_2, and 19% in SSC_4 (Figure 2B). For SSCs from culture, >45% occupied the SSC_1 cluster, followed by 23% in SSC_2, 18% in SSC_4 (Figure 2B).

As a whole, the expression of a broad range of SSC markers (Etv5, Gfra1, Lhx1, Bcl6b, Ret) was robust across all SSC clusters (Figure 2C), noting that expression of Shisa6 (Tokue et al., 2017) was below levels of detection (Figure 2C), and Id4 levels could not reliably be assessed given that spermatogonia from the testis, but not from culture, came from an Id4-eGfp transgenic mouse line (Helsel et al., 2017b), with transgene expression causing an artefactual increase in expression in the testis dataset (Martin and Wang, 2011). It is interesting to note, however, that cluster SSC_0 exhibited a small but significant (p < 0.001) increase in expression of a subset of genes that have been reported to play a role in SSC maintenance, including; Etv5 (Wu et al., 2011), Gfra1 (He et al., 2007), Ret (He et al., 2007), Sall4 (Chan et al., 2017), Chd4 (Cafe et al., 2021b), and Plvap (Nakagawa et al., 2021) (Figure 2C; Supplementary Dataset S1). In considering that 49 spermatogonia from culture (representing 6.8% of total cells from the culture dataset) reside within the SSC_0 cluster, one could postulate that these may be the SSCs that colonise recipient testes post-transplantation. Indeed, when cells from these cultures were transplanted, they produced an average of 270 colonies per 100,000 cells (Cafe et al., 2021b); equating to ∼2.2–5.4% of cells possessing regenerative capacity when factoring in a colonisation efficiency of 5–12% (Nagano et al., 1999; Ogawa et al., 2003).

To gain further understanding on the effects that in vitro culture may be eliciting on undifferentiated spermatogonia over time, and thus mechanisms that may underpin the temporal loss of SSCs (Helsel et al., 2017a), a list of DEGs was generated using the ‘FindMarkers’ function in Seurat, comparing the ‘SSC_0’ and ‘SSC_1’ clusters, given that these were predominated by SSCs from the testis and culture datasets, respectively. Using this strategy, 924 DEG’s were identified with an adjusted p value < 0.05 and a Log2-foldchange value > 0.25 (Figure 3A; Supplementary Dataset S1). Among these DEGs, 644 were downregulated in SSC_1, while 280 were upregulated. Gene ontology (GO) analysis was performed to identify enriched biological processes and pathways within these gene lists. A number of key themes were identified in genes that were upregulated in the ‘SSC_1’ cluster predominated by cells from culture, including oxidative phosphorylation (e.g. Cox6a1, Cox7C, Ndufb7), protein folding (e.g. Hsp90ab1 and Hsp90b1), and sterol biosynthesis/lipid metabolism (e.g., Cyp51, Hmgcr) (Figure 3B; Supplementary Dataset S1). Biological processes and pathways that were enriched in the list of downregulated DEGs included chromatin modification (e.g., Chd4, Chd7), ubiquitin mediated proteolysis (e.g., Psmd4, Psmd2), spermatogenesis (e.g., Tex19, Sycp1), and DNA repair (e.g. Rad51, Brca2) (Figure 3C; Supplementary Dataset S1).

Finally, to demonstrate the utility of our dataset, we utilised an independent set of undifferentiated spermatogonia populations to provide ‘proof-of-concept’. These populations were again isolated either directly from the testis (THY1+) or from primary cultures of spermatogonia that were established from the THY1+ population and maintained for a period of 10 weeks. Given that ‘Ubiquitin mediated proteolysis’ and ‘Ubiquitin-dependent catabolism’ were identified to be downregulated in culture (Figure 3C; Supplementary Dataset S1), we elected to establish whether there is indeed impaired clearance of ubiquitinated proteins by the proteasome in cultured spermatogonia (Figures 3D, E). Immunoblotting analysis using an α-ubiquitin antibody revealed a striking accumulation of ubiquitinated proteins within lysates from cultured spermatogonia. This was particularly the case for proteins of a higher molecular weight (>75 kDa), albeit, an increase in ubiquitination was also evident in low molecular weight proteins (Figure 3D). Densitometry analysis was performed on the entire complement of proteins, identifying a statistically significant >2-fold increase of ubiquitinated proteins in cultured spermatogonia (Figure 3E, p < 0.05, n = 3, normalised to tubulin). These results clearly demonstrate the biological relevance of the scRNA-seq dataset we have generated, which has identified disruption of several integral pathways for cellular function and longevity and thus provides a platform for future investigation into adapting cell culture conditions for safe, long-term maintenance of SSCs.