Mice were housed at the Monash Medical Centre Animal Facility with a 12h light/dark cycle and food and water available ad libitum. All animal procedures were carried out in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes. This study was approved by the Monash University Animal Ethics Committee. Mice lacking the inhibin alpha subunit (Matzuk et al., 1992), termed Inha KO (C57/Bl6 background), were maintained through heterozygote breeding. Adult Inha KO male mice were collected at postnatal (P) day 48–64. Tails were collected from each animal for commercial genotyping (Transnetyx).

Whole mouse testes were weighed, and fixed in 4% paraformaldehyde (PFA, ThermoFisher Scientific) overnight at 4°C. Testes were washed 3 × 15 min in PBS, cut in half and cryoprotected in 30% (w/v) sucrose for 2–3 days at 4°C. Testis tissues were then embedded in Tissue-Tek O.C.T. Compound (Sakura) and stored at −80°C until sectioned. Frozen sections (7 µm) were cut onto Superfrost Plus slides by Monash Histology Platform staff (MHTP node). For immunofluorescent analyses, frozen sections were air dried for 15–20 min then washed in PBS, 3 × 10 min. Sections were circled with a wax pen (Cedarlane Laboratories), then blocked with 10% (v/v) fetal calf serum (FCS, Bovogen) in 2% (w/v) bovine serum albumin (BSA, Sigma-Aldrich) in PBS for >1 h. Primary antibodies diluted in blocking solution were used to detect the proteins listed in Table 1. Slides were placed in a humid chamber overnight at 4°C. The next day, slides were washed in PBS 3 × 8 min, and secondary antibodies (Invitrogen, donkey anti-goat 488, A11055, RRID:AB_2534102; donkey anti-rabbit 555, A31572, RRID:AB_162543; Jackson ImmunoResearch Labs, donkey anti-rat 647, 712–605–153, RRID:AB_2340694) diluted 1:500 in 2% BSA/PBS were applied to sections and incubated for 2 h at room temperature in a humid chamber. Slides were washed in PBS 3 × 8 min, then sections were counterstained with 5 μg/mL DAPI (Sigma-Aldrich) prior to mounting with Prolong Gold mounting medium (Invitrogen) under glass coverslips. Slides were scanned (Monash Histology Platform, MHTP node) using the Olympus VS.120^® Virtual Slide Microscope System. For Hematoxylin and Eosin (H&E) staining, adult mouse testes were fixed in Bouins solution for 4–5 h, washed in 70% ethanol (v/v) 2 × 5 min, then processed, embedded in paraffin, and sectioned (4 µm) prior to H&E staining (Monash Histology Platform, MHTP node). H&E sections were imaged using an Olympus BX53 microscope with Olympus DP73 camera and Olympus cellSens Software (RRID:SCR_014551).

Centrally located testicular cross sections from 3 independent animals/genotype were assessed for each measured parameter. Three groups of adult seminiferous tubules were defined based on the mouse genotype and the proximity of the tubules to tumours. Tubules from Inha WT mice exhibited complete, normal spermatogenesis, as expected and were designated WT. In Inha KO testes, tubules distant from tumour foci exhibiting grossly ‘normal’ spermatogenesis were designated N, and tubules adjacent to tumours exhibiting spermatogenic deterioration were designated TAT (for tumour associated tubules). The perimeter of intact tubules (identified as those with an undamaged basement membrane) were measured (µm) using Fiji (Version 2.1.0, RRID:SCR_002285) (Schindelin et al., 2012) or QuPath (RRID:SCR_018257) (Bankhead et al., 2017) and counted. The tumour regions lacked intact tubules and were excluded from the quantitative IF analyses. Quantitative analyses of cells expressing proteins of interest was performed using panels of antibodies as follows: A) To assess the number of SSCs and their proliferation, all spermatogonia were identified using SALL4, then SALL4+ cells with GFRA1 and/or KI67 signals were counted. B) To assess the number of progenitor cells, all spermatogonia positive for both GILZ and SOX3 were counted. C) SOX9+ Sertoli cell nuclei were detected and counted using the Positive cell detection command in QuPath. A scaling range for each antibody was established based on settings that yielded no visible signal in the negative controls lacking primary antibody (secondary antibody only). Sertoli cell-only cords with no visible lumen and rete testis were excluded from all analyses.

To generate single-cell suspensions from adult Inha WT and KO testes (without tumours, Inha KO ‘normal’ tubules only), dissected tissues were coarsely minced and washed with PBS, then digested with 1 mg/mL collagenase Type IV (Sigma-Aldrich) in DMEM medium (Gibco) and 50 μg/mL DNase I (Sigma-Aldrich). Tubule fragments were washed in PBS to deplete interstitial and peritubular myoid cells then dissociated with 0.25% trypsin/DMEM in the presence of DNase I and passed through a 70 μm cell strainer. For intracellular staining, cell suspensions were fixed in 4% PFA for 10 min at 37°C then permeabilized in methanol overnight at −20°C. Fixed and permeabilized cells were washed in PBS with 2% FCS then stained with the following antibodies diluted in PBS with 2% FCS: eFluor 450-conjugated anti-KI67 clone SolA15 (Thermo Fisher Scientific, 1:500, 48–5698–82, RRID:AB_11149124) and phycoerythrin (PE)-conjugated anti-KIT clone 2B8 (Thermo Fisher Scientific, 1:250, 12–1171–81, RRID:AB_465812), rabbit anti-mouse EOMES clone 1219A (R&D Systems, 1:1,000) and Alexa 647-conjugated anti-PLZF clone 9E12 (Hobbs et al., 2010). EOMES antibody was detected with Alexa 488-conjugated secondary antibody (Thermo Fisher Scientific and Jackson ImmunoResearch, 1:500). Cells were analysed on an LSR Fortessa (Monash Flowcore) and data processed with FlowJo software.

Tissues were collected from decapsulated testes from Inha WT and KO mice aged between P53-55. Three discrete regions were microdissected from Inha KO testes: Inha KO ‘normal’ tubules (N), ‘tumour associated tubules’ (TAT), and tumour (TM) regions. Inha KO ‘normal’ seminiferous tubules were collected from areas distant from tumour regions. Inha KO tumour associated tubules ‘TAT’ and tumour regions ‘TM’, readily identified by the high level of vascularisation, were microdissected and snap frozen. To assess heterogeneity of multiple tumours within a single testis and between testes from the same animal, ‘normal’, ‘TAT’ and multiple ‘tumour’ regions from both testes of a single mouse were analysed.

Testis tissues (18–100 mg) were homogenised using a syringe and 23G needle. RNA was isolated from Inha WT mouse testis tubules (n = 3), and Inha KO ‘normal’ (n = 4), ‘tumour associated tubules’ (n = 4) and tumour regions (n = 6) using the RNeasy Mini Kit (Qiagen) with on-column DNase treatment (RNase-free DNase Set, Qiagen) according to the manufacturer’s instructions. RNA integrity and concentration were assessed using the 2100 Bioanalyser (Agilent) prior to RNA sequencing (RIN values: 8.3–9.3). RNA-Seq was performed using a custom in-house multiplex method, modified from (Grubman et al., 2021) (MHTP Medical Genomics Facility). Briefly, samples were given a unique i7 index (together with UMI) during individual pA priming and first strand synthesis which also adds a template switch sequence to the 5′-end. Samples were then pooled and amplified using P7 and an oligo which binds the template switch sequence. Final library construction was completed by tagmentation and addition of P5 (with i5 index) by PCR. Sequencing was performed on an Illumina NSQ2k run with up to 101 nt SR (cDNA). An 18 nt i7 read contains the 8 nt index and 10 nt UMI and, where required, an 8 nt i5 index read is also generated.

Bioinformatics analysis was performed as follows: Raw fastq files were analysed using the nf-core/rnaseq pipeline (Ewels et al., 2020) using STAR aligner (RRID:SCR_004463) (Dobin et al., 2013) to map the raw reads to GRCm38 (Mus musculus) reference genome. Reads were quantified using Salmon (RRID:SCR_017036) (Patro et al., 2017) producing the raw genes count matrix. Alignment and transcript quantification were completed via the Laxy platform (Perry and Powell, 2020). Raw counts were exported to and analysed with Degust (RRID:SCR_001878) (Powell et al., 2019), a web tool which performs differential expression analysis using limma voom normalisation (Law et al., 2014), producing counts per million (CPM) library size normalisation and trimmed mean of M values (TMM) normalisation (Robinson and Oshlack, 2010) for RNA composition normalisation. This identified differentially expressed genes (DEGs) in WT and Inha KO ‘normal’ (N), tumour associated tubules (TAT) and tumour (TM) regions using the following criteria: false discovery rate (FDR) < 0.05, Abs log fold change (FC) > 1.5X. RNA-seq data are available via GSE236488.

To identify enriched biological themes, we performed DAVID analysis (RRID:SCR_001881) (Huang da et al., 2009a; Huang da et al., 2009b) which requires a list of <3,000 DEGs. To accommodate this requirement, FDR and Abs log FC values were adjusted for each group as follows: WT and KO N; FDR cut-off <0.05, Abs log FC > 1.5X, WT and KO TAT; FDR cut-off <0.01, Abs log FC > 1.5X, WT and KO TM; FDR cut-off <0.0001, Abs log FC > 3X. Ingenuity Pathway Analysis software (IPA, RRID:SCR_008653) (Qiagen, Hilden, Germany) was used to identify downstream Disease and Bio Functions in adult Inha KO N, TAT and TM samples compared to WT (Kramer et al., 2014). RNAseq data from each comparison was uploaded to the Ingenuity Knowledge Base (Genes Only) reference set, using Expr Fold Change range: < −3; > +3, and Expr p-value of 0.01. Significance calculated using the right-tailed Fisher’s exact test.

In mice, genetic deletion of the Inha gene (Inha KO) results in testicular enlargement (Figures 1A,B), and development of focally invasive tumours by approx. 4 weeks of age (Figure 1A, arrow) (Matzuk et al., 1992; Matzuk et al., 1994; Coerver et al., 1996). Although reduced activin A levels during fetal life affect testis cord development (Archambeault and Yao, 2010; Moody et al., 2022), in adulthood elevated activin A bioactivity does not significantly affect average tubule perimeter in grossly ‘normal’ tubule cross-sections from Inha KO mice (Figure 1C). Given that spermatogonia reside on the basement membrane, all cell counts were normalised to tubule perimeter.

To understand the impact of chronically elevated activin A on spermatogonial populations, we first evaluated the incidence of SALL4+ (marks all spermatogonia, SPG), GFRA1+ (SSC-enriched population) and KI67+ (proliferation) cells in adult Inha KO ‘normal’ tubules. The percentage of tubule cross-sections containing no GFRA1+/SALL4+ cells was significantly lower, and those containing 1-2 or ≥3 GFRA1+/SALL4+ cells was significantly higher, compared to wildtype (WT) testes (Figure 1D). The overall abundance of SALL4+ SPG and GFRA1+/SALL4+ SPGs in Inha KO testes was significantly increased (by 50% and 140%, respectively), and each population exhibited a higher incidence of KI67+ cells, indicating enhanced proliferation (Figures 1E, F). Additionally, in WT testes, there was approximately 1 GFRA1+ cell per 21 SPGs, however in the KO testes, there was approximately 1 GFRA1+ cell per 13 SPGs, suggesting a higher proportion of SPG were SSCs in ‘normal’ tubule cross-sections of adult Inha KO mice compared to in WT controls.

Next, we characterised populations of SPG isolated from adult WT and Inha KO ‘normal’ testes (without visible tumours) by intracellular staining for PLZF, KIT and EOMES by flow cytometry (Chan et al., 2017; La et al., 2018b). Undifferentiated SPG are PLZF+/KIT- (population a), early differentiating (diff) SPG are PLZF+/KIT+ (b), and late differentiating SPG are PLZF^low/KIT+ (c) (Figures 2A, B). The relative abundance of undiff SPG (a) and early diff SPG (b) was variable and not significantly altered, however late diff SPG (c) were significantly increased (Figure 2B), consistent with normal differentiation from an elevated population of undiff SPG. Primitive SPG are EOMES+ and were identified in the (a) undiff SPG population (Figure 2C) and intriguingly, the proportion of undiff cells that were EOMES+ was significantly decreased in Inha KO testes (Figure 2D). In agreement with our previous data, primitive and undiff SPG were more proliferative, as indicated by KI67 (Figures 2E, F). These data indicate long term exposure to elevated activin A levels leads to altered SSC and spermatogonial populations.

In WT testis cross-sections, GFRA1+ cells are relatively rare and often observed as isolated SPG on the basement membrane. Therefore, the finding of GFRA1+ cell clusters in seminiferous tubules adjacent to tumour regions (termed here ‘TAT’, for tumour associated tubules) was intriguing (Figures 3A–C). Theses clusters were frequently displaced from the basement membrane, and further examination revealed a high proportion of these cells were EOMES+ (Figure 3B, panels 1–2) and LIN28A– or LIN28A low (Figure 3C). EOMES expression identifies a discrete population of A_s-A_pr spermatogonia (Sharma et al., 2019), and LIN28A is reported to be absent or weakly expressed in A_s and may be expressed at either low or high levels in A_pr spermatogonia (La et al., 2018b; Sharma and Braun, 2018; McAninch et al., 2020). Thus, this combination of staining (GFRA1+/EOMES+/LIN28A- or LIN28A low) indicates these clusters are comprised of A_s – A_pr spermatogonia. To establish if SSC clusters result from a failure to differentiate, we applied the progenitor marker SOX3 and the differentiation marker KIT. Both SOX3+ and KIT+ cells were present in TAT tubules (Supplementary Figure S1A, B) demonstrating SSC differentiation was occurring. Next, we reasoned that the relative proportion of SOX3+ cells compared to GFRA1+ SPGs would be significantly reduced if SSC differentiation was impeded, so we quantified SOX3+/GILZ+ cells in WT, Inha KO N and TAT tubules (Figure 3E). In Inha KO ‘normal’ N tubules (Figure 3E, purple bars) compared to WT, the average tubule perimeter was not different, the average number of GFRA1+ cells/1 mm tubule perimeter was significantly increased in Inha KO N tubules, however SOX3+ cell numbers were unchanged. In Inha KO TAT tubules (Figure 3E, orange bars), the average tubule perimeter was reduced compared to WT and KO N, consistent with tubules exhibiting severely disrupted spermatogenesis. The average number of GFRA1+ cells/1 mm tubule perimeter was further increased in TAT tubules, compared to WT and KO N, while SOX3+ cells were also significantly increased. The ratio of SOX3+/GFRA1+ cell was significantly lower in Inha KO N and TAT tubules compared to WT, suggesting SSC differentiation is suppressed in Inha KO testes (Figure 3E). Thus, the presence of clusters of A_undiff resembling primitive SSCs along with a proportional decrease in progenitors indicates SSC differentiation is hindered in TAT tubules.

Previous histological analyses described Inha KO tumours as undifferentiated gonadal stromal tumours derived from cells of the Sertoli cell lineage (Matzuk et al., 1992; Matzuk et al., 1994) that are positive for GATA4 (Sertoli cell nuclei) (Haverfield et al., 2017). To explore the molecular features of these tumours in Inha KO testes, we performed RNAseq on Inha WT tubules, microdissected Inha KO ‘normal’ (N), tumour associated tubules (TAT) and tumour regions (TM) (Figure 4A). The presence of multiple tumours within a single testis (Figure 4B) allowed tumour heterogeneity to be compared across 3 independent animals, as well as within a single animal and a single testis (Figure 4C). A multidimensional scaling (MDS) plot (Figure 4D) identified that all 4 groups were transcriptionally distinct from each other, indicating the microdissection was effective at separating distinct regions. To examine the degree of tumour heterogeneity, the coefficient of determination (R² value) of the global transcriptome (CPM of ∼22,000 genes/sample) was calculated between samples from each group (WT and Inha KO ‘N’, ‘TAT’ and ‘TM’, including single tumours from multiple animals (multiple animals), tumours from both testes from the same animal (single animal, both testes), and tumours within a single testis (single animal, single testis) (Figure 4E). The overall heterogeneity of WT testis samples was low, with R² values between 0.84–0.96. In Inha KO testes, heterogeneity was even lower in ‘normal’ tubules and in tumours found within a single animal (R² values: 0.93–0.99), suggesting a high degree of transcriptional similarity between focal tumours within testes from an individual mouse. The greatest heterogeneity was observed between tumours from multiple animals (R² values: 0.67–0.92). During the course of cancer development, cellular heterogeneity generally increases (Marusyk and Polyak, 2010), therefore this heterogeneity may reflect the varying severity of the tumours analysed.

Overall, we found that compared with Inha WT controls, the number of differentially expressed genes (DEGs) in each KO group increased in parallel with tumour proximity (Figure 4F). As expected, Inha was significantly decreased in all Inha KO groups and Inhba (encoding activin A) was increased in both the TAT and TM groups (Supplementary Figure S2). In Inha KO ‘normal’ testis tubules, there were 38 DEGs compared to Inha WT testes (FDR <0.05, absolute fold change ≥1.5, Supplementary Table S1) including key Leydig cell-associated genes (Insl3, Cyp17a1, Hsd17b3, decreased 3 to 5-fold). Using publicly available testis single-cell RNAseq datasets (Hermann et al., 2018; Tan et al., 2020b), we found 37% (14/38) of the DEGs are normally exclusively expressed in Leydig cells and all were decreased, including several members of the kallikrein 1-related peptides (Klk1b22, Klk1b21, Klk1b24, Klk1b27, decreased ∼ 5-fold). The reduced expression of all Leydig cell-specific DEGs in the Inha KO testes most likely reflects their reduced number (Matzuk et al., 1992). DAVID analysis identified the top affected biological processes to be enriched for the terms ‘Zymogen’, referring to the enzymatically inactive precursor of mostly proteolytic enzymes, and ‘Disulphide bond’, both of which include the kallikrein 1-related peptide genes (Figure 4G, purple bars). To further understand features of the KO ‘normal’ regions, Ingenuity Pathway Analyses (IPA) was employed and identified cellular movement, immune cell trafficking, inflammatory response, lipid metabolism and cancer in the top 20 diseases and bio functions (Supplementary Table S2). Importantly, these indicate the pathways influenced by the subtle transcriptional changes occurring in the Inha KO N tubules, which precede their evolution into the more severe phenotypes evident in TAT and TM regions.

The TAT samples comprised seminiferous tubules and interstitial areas bordering tumour regions. In this group, 4,939 DEGs were identified compared to WT control testis (Supplementary Table S1). DAVID analysis revealed significant enrichment for terms including endoplasmic reticulum (ER), ribosomal proteins and mitochondrion (Figure 4G, orange bars). During tumorigenesis, the high proliferation rate of cancerous cells requires increased ER protein assembly, folding and transport, which can induce physiological ER stress (Chen and Cubillos-Ruiz, 2021), and increased mitochondrial metabolism (Lee et al., 2022). Our data showing 30 ribosomal and 5 mitoribosomal transcripts were significantly increased in TAT samples (Supplementary Table S1) provides evidence that organelles in cells near to these testicular tumours have adopted a stress-response phenotype. In agreement with this, IPA analyses, identified Metabolic disease in the top 20 Disease and Bio Function categories (Supplementary Table S2).

In the tumour regions (TM), we identified 12983 DEGs versus WT testis (Supplementary Table S1) associated with UniProt keywords such as spermatogenesis, cell differentiation, motile cilium and metalloprotease (Figure 4G, red bars, DAVID analysis). IPA analysis further identified the top Disease and Bio Functions as Cancer, Organismal Injury, and Abnormalites, and Endocrine System disorder (Supplementary Table S2). Reproductive related categories included Reproductive System Disease and Reproductive System Development and Function (Supplementary Table S2). This confirms the tumour regions assume a transcriptional landscape consistent with disease and loss of spermatogenesis.

To investigate the cellular composition and transcriptional landscape of Inha KO tumours, we used 1) cell-specific transcripts to explore the relative proportion of testis cell types (germline, immune, Leydig, Sertoli, endothelial, and peritubular myoid cells) within the tumours (Figure 5), 2) immunofluorescence to examine the in situ distribution of these cell types (Figure 6), and 3) H&E staining to examine the cytological features and arrangement of cells within the tumours (Figure 7). Cell-specific transcripts, identified from single-cell RNAseq datasets of adult mouse testis (Hermann et al., 2018), allowed us to assess the representation of each cell type within the tumours. As expected, undifferentiated SPG markers (Sall4, Plzf) generally increased (mean CPM) in KO N, and TAT (Supplementary Table S2) consistent with our IF counting data. In the TM samples, these transcripts were elevated and highly variable. While this may reflect germ cell remnants present within these regions, SALL4 staining was low/absent, therefore the significance of these results in the context of the tumours is currently unclear. Germ cell markers (Ddx4, Sycp3) were significantly reduced in tumours, indicating the paucity of germ cells in these areas (Figure 5A). Immune cell markers (Cx3cr1, Csf1r) were increased in TAT and TM groups, consistent with immune cell infiltrates being a common feature (Figure 5B). Intriguingly, Leydig cell transcripts exhibited two distinct profiles; Insl3 and Klk1b21 were strongly decreased in all KO groups compared with WT, in accord with reduced Leydig cell number and/or function, however steroidogenic markers such as Cyp11a1 and Hsd17b3 were increased in TAT and TM groups indicating these regions display a steroidogenic phenotype (Figure 5C). Interestingly, certain Sertoli cell specific transcripts (Sox9, Gata1, Rhox8, Cst12) were increased in TAT, likely reflecting a proportional increase in Sertoli cells due to a decrease in germ cells but were not significantly different in TM regions compared to WT (Figure 5D). Endothelial (Pecam1, Icam2, Kazald1, Ctla2a) (Figure 5E) and peritubular myoid cell (Acta2, Lsp1, Serpine1, Fbxl22) (Figure 5F) transcripts were significantly increased in TAT and TM groups compared to WT. These likely reflect both the increased vascularisation, which is common in testicular germ cell tumours (TGCT) and tumours in general (Silvan et al., 2010; Ribatti and Pezzella, 2021), and an increase in the myoid cells that contribute to the seminiferous tubule basement membrane.

Next, we visualised the distribution of key cell types across the same tumour region using immunofluorescence (Figures 6A–D). Representative images show three tumours of different sizes and severity: low (L), medium (M) and severe (S) grade tumours (outlined white dotted lines). Tumours were classified based on the following criteria: visible tumour size together with the absence of normal spermatogenic architecture, specifically, the loss GCNA + germ cells and the presence of cell infiltrates. Although Inha KO tumours are thought to be derived from Sertoli cells, unexpectedly, SOX9+ cells, normally marking Sertoli nuclei, were less abundant in tumour regions and very low in the severe tumours despite being readily detected in Inha KO ‘normal’ and TAT tubules, (Figure 6A).

Although smaller tumours contained some germ cell remnants, more developed tumours did not contain germ cells (GCNA+). As outlined above, clusters of GFRA1+ cells were visible within the tumours and in tubules bordering the tumour regions (Figure 6A, arrowheads). SOX3, marking spermatogonial progenitors (Figure 6B, white arrows), was detected in Inha KO ‘normal’ tubules and also in tubules containing clusters of SSCs; these observations show these cells have the capacity to differentiate, albeit at a reduced rate. SPG and early meiotic germ cells (GILZ+) exhibited a similar staining pattern to GCNA and were not readily detected in the tumour regions (Figure 6B). The tubule basement membrane, outlined by LAMININ, appeared thicker in TAT and tumour regions than KO ‘normal’, consistent with our RNAseq data. In humans, the equivalent structure, the lamina propria, is thicker in azoo- and oligospermic patients (de Kretser et al., 1975; Sato et al., 2008), and in Leydig cell tumours (Charfi et al., 2012), providing an example of a feature common to these somatic cell tumours in mice and human testicular disorders.

The steroidogenic enzyme CYP11A1 exhibited remarkably intense staining in all tumours, including in low grade tumours, suggesting the steroidogenic phenotype is present early during tumour initiation (Figure 6C). Consistent with the formation of haemorrhagic tumours (Figure 4B) and our RNAseq data (Figure 5E), we observed increased levels and frequency of PECAM+ endothelial cells/vasculature around the tumour boundary and within severe grade tumours (Figure 6C). To support tumour expansion, blood vessel growth (angiogenesis) is induced via the secretion of various factors including FGFs and VEGF (Sakurai and Kudo, 2011), both of which were increased transcriptionally in the TM group (Figure 7B). Finally, immune cells were detected using the pan-leukocyte marker CD45. The frequency of elongated CD45⁺ cells observed was increased at the perimeter of seminiferous tubules close to the tumour boundary, around the tumour borders, and inside the tumours (Figure 6D, arrows) consistent with our RNAseq data showing Ptprc elevation.

Histological analyses on H&E-stained sections allowed us to examine the cellular and tissue structure of TAT and tumour regions in more detail and provided additional insights (Figures 7A,B). The classification of tumours which appear as low, medium, and severe grade (as described in Materials and Methods) provides clues to the events involved in tumour progression. In low and some medium severity tumours, tightly packed tumour cells were surrounded by a distinct border, likely the LAMININ+ basement membrane (observed in Figure 6B), providing evidence that the tumour originates from a cell/s within the tubule. As the tumour progresses, the border is less well-defined, and the tumour cells appear to infiltrate the interstitial space. In severe tumours, there is a lack of both tubules and basement membrane (Figure 6B, LAMININ staining), and the increased numbers of red blood cells (RBCs) is consistent with haemorrhagia (Figure 7A). Our interpretation based on these H&E images and LAMININ staining is that the tumours initiate within the tubule and proliferate to the extent that the tubule becomes enlarged, leading to eventual breakdown of the basement membrane. Subsequently the tumour cells infiltrate the local region causing progressive degeneration of surrounding tubules, loss of spermatogenesis, and increased vascularisation. Next, a histological comparison of tumour cells with Sertoli cells from adult Inha WT and KO tubules was performed (Figure 7B). Sertoli cell nuclei in Inha WT and KO N tubules (arrowhead, white fill) were identifiable based on their shape and position, parallel or perpendicular to the basement membrane. Depending on the plane of the section, characteristic nuclear indentations, and tripartite nucleoli were also visible. In Inha KO TAT tubules with incomplete spermatogenesis, Sertoli cell nuclei with a normal appearance were observed, however they often appeared displaced from the basement membrane. Nuclei within the tumour regions (arrowhead, yellow fill) did not display typical Sertoli cell characteristics, appearing rounded with chromatin granules arranged along the nuclear membrane (see inset), more closely resembling fetal Sertoli cells (arrowhead, blue fill) (Figure 7B).

Overall, our RNAseq, immunofluorescence and H&E data were in accord: tumour progression results in a loss of germ cells and mature Sertoli cells, accompanied by an increase in the proportion of peritubular myoid cells, vascular endothelial cells and immune cells. Intriguingly, tumour cell nuclei more closely resemble fetal compared to adult Sertoli cell nuclei and exhibit evidence of an increasingly altered steroidogenic profile with tumour progression.

Within seminiferous tubules, SSCs respond to exogenous growth factors that regulate their self-renewal and differentiation. One example is GDNF, secreted by Sertoli cells in a cyclic manner to regulate spermatogonial stem cell proliferation (Meng et al., 2000; Sharma and Braun, 2018). The observation of GFRA1+ SPG clusters in tubules adjacent to tumours lead us to hypothesise that Inha KO tumours secrete factors that promote SSC self-renewal and maintenance. To corroborate this, and potentially identify new factors that affect signalling important to SSC biology, we profiled growth factor transcript expression in Inha WT and KO testes. Our initial analyses comprised 53 growth factors including TGFβs, BMPs, FGFs, IGFs, PDGFs, and VEGFs, and following exclusion of those below the level of detection, 35 growth factors remained. The PCA plot showed growth factor transcript expression progressively diverges from Inha WT to KO TM (Figure 8A, black arrow). The heatmap revealed a suite of growth factor transcripts that increase from WT tubules to Inha KO tumours (Figure 8B), including Inhba (activin A), consistent with previously published data showing activin A levels are elevated in these mice (Matzuk et al., 1994). We found Gdnf, Igf1 and Fgf2 transcripts, involved in maintaining SSCs in an undifferentiated state (Meng et al., 2000; Kubota et al., 2004; Ishii et al., 2012; Sharma and Braun, 2018), were significantly increased in TAT and/or TM regions (Figure 8C), as were Bmp4 and Inhba which promote SSC differentiation in vitro (Nagano et al., 2003). Given the increased numbers of GFRA1+ SPGs observed in our model, and the elevation of transcripts important for both SSC self-renewal and differentiation, we surmise that the relative stoichiometry of growth factors is critical in determining cell fate outcomes.

Importantly for SSC biology, secretion of FGFs (4, 5 and 8) by lymphatic endothelial (LE) cells located near the vasculature, regulate SSC density and fine tune their self-renewal and differentiation in normal mouse testes (Kitadate et al., 2019). However, transcripts encoding Fgf4, 5 and 8 (plus others) were below the level of detection in all samples. Of the remaining FGF transcripts, 7 were detected, and their expression was variable in the Inha KO TAT and TM regions; we measured increased levels of Fgf2, 11, 13, and 18 transcripts, and decreased levels of Fgf10, 12 and 14 compared to WT (Figure 8B). Fgf2 and Fgf18 encode secreted canonical FGFs, which bind to and activate FGFRs, however Fgf11 and 13 transcripts encode non-signalling, intracellular proteins (Ornitz and Itoh, 2015). In testes, FGF2 is a key factor in promoting SSC expansion in vitro, and a lack of FGF2 leads to dysregulated sperm production and altered sperm morphology and function (Saucedo et al., 2018), while relatively little is known about the function of FGF18. Since Fgf10, 12 and 14 are predominantly expressed in germ cells (Hermann et al., 2018), their decrease likely reflects the loss of germ cells in these regions.

Receptors related to GDNF, IGF and FGF signalling were detectable and generally increased in the TAT and TM samples signifying the potential for these signalling pathways to be active (Figure 8D). Downstream targets of FGF (Stat1, 3, 5, Foxo1, Etv4), GDNF (Bcl6b, Egr2, Egr3) and IGF (Igf1, Trp53, Casp8) signalling were increased in tumours, indicating pathway activation (data not shown). GDNF receptor transcripts Gfra1 and Ret were present at low levels, and Gfra1 transcripts were increased in N and TAT samples compared to WT, consistent with our counting data. Both IGF1 receptors, Igf1r and Insr were detected and FGF2 receptors, Fgfr1 - Fgfr4, exhibited varying expression levels. Fgfr1 was most highly expressed and increased uniformly in relation to the tumour proximity, however Fgfr2 - 4 were expressed at low levels and were highly variable. These data indicate that progression towards a tumour phenotype results in increased production of numerous growth factors including several of known importance in SSC maintenance, some of which are not exclusive to a particular cell type.

Our analyses showed the ‘normal’ Inha KO testis tubules contain a higher proportion of undiff SPG and an expanded pool of GFRA1+ SPGs (∼2-fold higher), indicating increased SSC abundance. In addition, both cell types were more proliferative than in WT controls. The increase in GFRA1+ cells was surprising given that in vitro studies have demonstrated SPG treated with a 50 ng/mL activin A promotes their differentiation (Nagano et al., 2003). Although activin A serum (Matzuk et al., 1994) and transcripts (our data) are elevated in adult (4–6 week old) Inha KO testes, it is highly likely that within the complex in vivo cellular environment, these levels may not be robust enough to elicit differentiation of SPG. Indeed, it is possible that in culture 50 ng/mL activin A represents a supraphysiological level that far exceeds the elevated levels of activin A in even the Inha KO mice. Clusters of primitive SPG and a decreased proportion of progenitors in tubules adjacent to tumour regions indicated that the tumour cells provide a localised microenvironment that supports SSC self-renewal and impedes their differentiation. In agreement, transcriptional analyses demonstrated tumour areas were enriched for transcripts encoding numerous growth factors including several fundamental to SSC self-renewal (Gdnf and Fgf2). In total, over 20 additional growth factors transcripts with the potential to influence SSC biology were also increased, including those shown to promote SSC differentiation in vitro (Inhba, Bmp4) (Nagano et al., 2003; Li et al., 2014). Because SSCs reside within a complex changing cellular environment and are exposed to numerous growth factors simultaneously, the relative stoichiometry of growth factors may be critical to determining stem cell fate. For example, can we identify a concentration at which GDNF overrides BMP4 differentiation cues in SSCs?

Based on histological features such as SSC clusters, spermatogenic deterioration, and increased transcriptional dysregulation in tubules bordering tumour regions, we propose that tumour-derived growth factors influence a localised microenvironment ∼1-3 tubules wide surrounding the tumours. Therefore, the increased numbers of SSC within ‘normal’ tubules, distant from the tumours was surprising. Below we discuss three potential tumour independent mechanisms of actions based on our findings that could result in the increased frequency of SSCs within Inha KO ‘normal’ tubules (summarised in Figure 9A) and speculate that one or more of these contributes to the observed phenotype.

In the first model (a), Sertoli cell niche-driven: The roles of activin A on Sertoli cell proliferation and maturation throughout testis development are well documented (Boitani et al., 1995; Mendis et al., 2011; Nicholls et al., 2012; Itman et al., 2015; Haverfield et al., 2017). In Inha KO testes, Sertoli cell numbers are higher than in WT (Haverfield et al., 2017), and adult Inha HET testes support an increased (44%) daily sperm production (Itman et al., 2015). Because each Sertoli cell supports a fixed number of germ cells, and their postnatal complement determines SSC numbers in adult mice (Orth et al., 1988; Oatley et al., 2011), we considered the relationship between the activin driven increase in Sertoli cell number and GFRA1+ SPGs. As expected, both SOX9+ Sertoli and GFRA1+ cells were elevated in Inha KO testes, however the GFRA1+/SOX9+ cell ratio (Supplementary Figure S3A) remained significantly higher (∼50%) in Inha KO testes compared to WT, indicating that Sertoli cell numbers do not solely contribute to SSC numbers in this model. The increased Sertoli cell population likely aids in supporting the expanded stem cell pool.

In the second model (b), Germ cell catchup/regeneration: Germ cell depleted tubules observed in P16 Inha HET testes were resolved by P28 (Itman et al., 2015), indicating that germ cell loss due to increased activin A bioactivity in developing testes induces a germline regenerative process. Our unpublished observations indicate that germ cell numbers are reduced by approx. 50% in neonatal Inha KO mouse testes compared to controls. In adult Inha KO ‘normal’ tubules with ongoing spermatogenesis, the full complement of germ cells is visible. We therefore propose establishment of SSCs in juvenile life set the size of the stem cell pool for adult life and there is an ongoing regenerative response by SSCs. In damaged testicular tissue following busulfan-induced germline depletion, SSCs exhibit molecular characteristics different from homeostatic SSCs, including increased expression of cell cycle genes, and activation of the PI3K/AKT and mTORC1 signalling pathways (La et al., 2018a; Kitadate et al., 2019; La et al., 2022). The undifferentiated spermatogonia in Inha KO testes were more frequently proliferative (Ki67+), and cell cycle transcripts (Ccnd2, Mki67, Kif4, Top2a, Smarca5, Hells) were elevated, consistent with a regenerative SSC phenotype (La et al., 2022) (Supplementary Table S2). In contrast, as a readout of P13K/AKT activity, we found FOXO1 was nuclear in SPG of Inha KO N and TAT tubules, indicative of a homeostatic environment ((Goertz et al., 2011; La et al., 2022); Supplementary Figure S3B). Additionally, transcripts encoding several regenerative SSC markers were undetectable (e.g., Eomes, Pax7, Plaur) (Sharma et al., 2019; La et al., 2022), and EOMES⁺ cells were decreased as a proportion of the PLZF + population, compared to in Inha WT testes. Thus, SSCs in Inha KO ‘normal’ tubules exhibit some features of regenerative SSC but lack others.

In the third model (c), activin A-driven: CCND2 (Cyclin D2) acts synergistically with CCNE (Cyclin E) to induce SSC proliferation without the requirement for self-renewal factors such as GDNF and FGF (Lee et al., 2009). The cell cycle transcript Ccnd2 was 1 of only 9 transcripts elevated in Inha KO ‘normal’ tubules (Figure 4F; Supplementary Figure S2) and is activin A regulated in several tissues, including in embryonic mouse testes lacking activin A (Inhba KO) (Park et al., 2005; Mendis et al., 2011; Pauklin and Vallier, 2013). By sc-RNAseq, Ccnd2 is predominantly expressed in undiff SPG, but it is also produced in endothelial, Leydig and peritubular myoid cells in adult testes (Hermann et al., 2018), therefore we propose activin A may act as a niche factor, promoting SSC self-renewal via Cyclin D2. Intriguingly, transplantation of SSCs overexpressing both CCND2 and CCNE1 produced tubules consisting of spermatogonial clusters and an absence of haploid cells, resembling germ cell tumours (Lee et al., 2009) and similar to TAT tubules. In contrast, in vitro studies have shown activin A promotes SSC differentiation (Nagano et al., 2003; Tan et al., 2020a). Nevertheless, our in vivo results suggest that elevated activin A levels may contribute to increased SSC renewal in Inha KO ‘normal’ tubules via upregulation of Ccnd2.

Tumour cells can originate from progenitor cells that fail to differentiate appropriately (i.e., testicular germ cell tumours (TGCT) (Rajpert-De Meyts, 2006)), or they may arise via reversion to a more immature state. Inha KO male mice develop focally invasive gonadal sex cord-stromal tumours, thought to originate from Sertoli cells. Because these tumours initiate from foci, we considered that elevated activin A levels in fetal and postnatal life sustains or modifies a precursor somatic cell population, preserving it in an inappropriate developmental state that manifests as a tumour in adulthood (Figure 9B). Indeed, the presence of cell nuclei resembling fetal Sertoli cell nuclei surrounded by an intact basement membrane in low grade tumours suggests the tumour mass is derived from cells initially present within the tubules (i.e., Sertoli or germ cells). To determine if these tumours possess a signature consistent with immature Sertoli cells, we searched for Sertoli cell-specific transcripts selectively expressed in either undifferentiated or differentiated Sertoli cells based on sc-RNAseq data (Hermann et al., 2018)) and prior knowledge (Haverfield et al., 2015). Amh, marking immature Sertoli cells, was not detectable in our RNAseq dataset, while levels of mature Sertoli cells markers, Gata1 and Cldn11, were not different compared to WT, indicating the tumour cells had not acquired an immature Sertoli cell signature. These findings are limited by the small number of genes fitting the criteria, and lineage tracing studies and/or single-cell RNAseq analyses would address this.

The combination of transcriptional and histological analyses provided new insights into Inha KO tumour formation and their molecular and cellular characteristics. Small, low-grade tumours are initially surrounded by an intact basement membrane, while larger more ‘advanced’ tumours appear to invade the interstitial space, engulfing nearby tubules, such that remnants of intact tubules were frequently observed within severe tumour regions. This may occur due to tubule basement membrane breakdown, angiogenesis and hyperproliferation/infiltration of tumour cells. There is a signature consistent with elevated levels of several cell types (immune, endothelial, and peritubular myoid cells) and the deficiency in germ cells manifests as spermatogenic failure in these tumours.

Interestingly, Inha KO tumours exhibited changes to hallmark proteins and transcripts associated with Sertoli and adult Leydig cells indicative of cell identity changes that can occur within a tumour environment. Although the tumours were GATA4+ (Haverfield et al., 2017), Sertoli cell-specific transcripts were not increased (Figure 5D) and SOX9 protein, a marker of both immature and mature Sertoli cells, was not detected (Sharpe et al., 2003). As Sox9 transcripts were not altered compared to WT, this suggested post-transcriptional and/or post-translational mechanisms may limit SOX9 levels in these tumours. SUMOylation (small ubiquitin-like modifier) is a key mechanism by which SOX9 proteins are repressed (Williams et al., 2020), and we measured elevated Sumo2 and Sumo3 transcript levels in TAT and TM regions, suggesting this is one mechanism by which SOX9 protein expression may be disrupted (Supplementary Figure S2). Whether the loss of SOX9 protein is a signature of tumorigenic Sertoli cells remains to be elucidated. Additionally, transcripts encoding steroidogenic enzymes (Cyp11a1, Hsd3b1, Cyp17a1, Hsd17b1, Hsd17b3, Cyp19a1) and protein (CYP11A1) found in adult Leydig cells were elevated in both Inha KO TAT and tumour regions when compared to Inha KO ‘normal’ tubules, leading us to suggest cells in these regions acquire an atypical steroidogenic phenotype. Metaplasia refers to the replacement of one differentiated somatic cell type with another, a response typical of chronic exposure of cells to an irritant, and a known precursor to certain cancers (Giroux and Rustgi, 2017). In Inha testis tumour regions, the combination of low SOX9 protein levels and increased expression of steroidogenic components allude to cell identity changes consistent with this process (Figure 9B).

In adult testes, appropriate steroid production relies on a functional Leydig cell population. Adult Inha KO mice have reduced Leydig cell numbers (Matzuk et al., 1992) (Figure 7B; Supplementary Figure S3C); this is supported by RNAseq data showing an approximately 5-fold reduction in Leydig cell transcripts (Insl3, Klk1b21) in Inha KO ‘normal’ tubules, TAT tubules and tumours. Steroidogenic enzyme transcripts and protein (listed above) were elevated in Inha KO TAT tubules and tumours compared to Inha KO ‘normal’ tubules and IPA analyses identified lipid metabolism to be altered in Inha KO testes (N, TAT and TM samples, Supplementary Table S2). The observation that tubules are enriched for steroidogenic markers as they transition towards a tumorigenic phenotype, regardless of tumour size and severity, was intriguing. The enzymes CYP11A1, HSD3B1, and HSD17B1/B3 catalyse conversion of cholesterol to testosterone, while CYP19A1 catalyses androstenedione and testosterone to estrone and estradiol respectively, providing considerable scope for dysregulated steroid production. How reduced Leydig cell numbers, together with altered expression of steroidogenic enzymes impacts on the steroid milieu in Inha KO adult testes is currently under investigation.

The origin of these tumours remains open to further study. It is known that Sertoli cell development and proliferation are regulated by several hormones including activin A and FSH, (reviewed in (Meroni et al., 2019)), both of which exhibit elevated serum levels in adult Inha KO mice (Matzuk et al., 1992). Consistent with a role in driving Sertoli cell proliferation and tumour formation, disruption of luteinising hormone (LH) and/or FSH signalling ablates or reduces tumour size (Kumar et al., 1996; Haverfield et al., 2017). Our histological data showing an accumulation of tumour cells within low severity tumour tubules suggests the tumours are likely derived from Sertoli cells.

In summary, the data presented here show the condition of chronically elevated activin A can modify the adult testis SSC pool. Inha KO somatic cell tumours lack SOX9, a key hallmark feature of adult Sertoli cells and exhibit an unexpected steroidogenic profile, while providing a microenvironment that supports SSC self-renewal. Our identification that these tumours produce growth factors of potential importance to SSC biology highlight its value for studies of SSC regulation. Further examination of this model may help explain the aetiology of testis stromal cell tumours (Ulbright et al., 2000), and we propose that cells derived from tumours may be employed as feeders to help improve in vitro SSC culture that can be relevant for therapeutic purposes (Tran et al., 2022).