All animal procedures were carried out in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes under Monash University Animal Ethics Committee approval. Mice were housed at Monash Medical Centre Animal Facility under a 12-hour dark/light cycle and with food and water available ad libitum. For all experiments except RNA-Seq, timed-mates were set up between Swiss females and transgenic Oct4-eGfp males (OG2; pure 129T2svJ background) (25, 26). Females were checked daily and the presence of a vaginal plug marked as E0.5. At E13.5, pregnant females were culled by cervical dislocation and the uterine horns removed and placed in phosphate buffered saline (PBS). Embryos were removed and euthanised by decapitation and the developmental stage was determined by the time since mating, and fore- and hind-limb morphology. The gonad/mesonephros complex was dissected out of each embryo and the mesonephros removed. Sex of the embryo was identified by the absence or presence of testis cords in the gonads, visualised using an upright dissecting microscope (MZFLIII, Leica, Wetzlar, Germany).

For RNA-Seq, mice lacking the inhibin βA [encoded by Inhba; no activin A (17)] or inhibin alpha subunits [Inha; high activin A (27)] on a C57/Bl6 background were crossed with Oct4-eGfp mice (9, 25, 26). For each line, heterozygous timed mates were set up and fetal gonads from E13.5 and E15.5 embryos collected as above. Tails were collected from each embryo for commercial genotyping (Transnetyx, USA).

E13.5 testes were randomly assigned treatment groups. Testes were cultured on 0.4 µM Millicell cell culture inserts (Merck Millipore, Germany) in 6-well plates with each well containing 1.4 mL media (DMEM/F12, Thermo Fisher Scientific, Waltham, MA, USA; 10% FBS, Bovogen, Keilor East, VIC, Australia; 1% penicillin-streptomycin, Thermo Fisher Scientific) (28). Previous cell culture experiments with mouse postnatal day 6 and 15 Sertoli cells revealed SMADs localise to the nucleus following exposure to 5 ng/mL activin A (29). Further, the human germ cell-like line, TCam-2, is responsive to activin A at 5 ng/mL (30). However, the local concentration of activin A in the fetal testis is unknown, therefore we determined the optimum concentration of activin A by culturing whole testes with 0 (vehicle control; 4 µM HCl), 5, 25, 50 or 100 ng/mL human recombinant activin A (R&D Systems, Minneapolis, MN, USA) and activin-responsive genes measured. Following analysis of changes in activin A-responsive genes, testes were cultured in media containing 50 ng/mL activin A as the optimal dose (described in Results), or 10 µM SB431542 (Sigma-Aldrich, St Louis, MS, USA) and their respective controls (DMSO for SB431542). PBS was placed in the gaps between wells to maintain humidity. E13.5 testes were cultured for 48 hours with a full media change at 24 hours. Following culture, gonads were imaged using bright field and fluorescence using an Olympus IX70 inverted microscope to visualize gross gonad structure and GFP-positive germ cells. Gonads were removed from the membrane, washed in PBS and individually snap-frozen on dry ice for transcript analysis, fixed in 4% paraformaldehyde (PFA) for histological analysis, or dissociated for flow cytometry.

For RNA-Seq experiments, paired testes from one embryo yielded a single biological replicate. From whole gonad cultures (E13.5 + 48h), single testes were a biological replicate. For E13.5 germ cell cultures, 6 - 10 paired testes were pooled. Testis dissociation and isolation of germ and somatic cells were performed as previously described (9). Briefly, testes were dissociated in 0.25% Trypsin-EDTA. Dissociation was halted with media containing 10% FBS. Cells were passed through a 35 µM strainer to obtain a single-cell suspension then centrifuged. Cell pellets were resuspended in 0.4% BSA/PBS and propidium iodide was added for exclusion of non-viable cells. GFP-positive and GFP-negative cells were sorted by Monash FlowCore staff using either an Influx or ARIA Fusion (BD Biosciences) machine. Sorted cells were pelleted, supernatant removed, and stored at -80°C for transcript analyses. Gonocyte cell culture is described below.

This protocol was based on a previously published method using the Click-iT™ Edu Alexa Fluor™ kit (Thermo Fisher Scientific) (31). For EdU incorporation, a final concentration of 20 µM was added to culture media two hours before collection. Then testes were washed in PBS and dissociated in 0.25% Trypsin-EDTA at 37°C for 5 to 10 minutes. Dissociation was halted with DMEM/F12 containing 10% FCS, and the cells were passed through a 35 µm mesh cell strainer to obtain a single cell suspension. Following centrifugation and removal of supernatant, cells were resuspended in 4% PFA and fixed for 15 minutes at room temperature. After 3 washes in permwash (1X saponin-based permeabilisation reagent (Thermo Fisher Scientific) in 1% BSA/PBS) cells were stored in permwash for no more than one week before immunostaining. For all steps, solutions were made up in and washes done with permwash and performed at room temperature.

Cells were centrifuged and resuspended in 5% donkey serum (Sigma) for 15 minutes. Cells were incubated with anti-DDX4 (detection of germ cells; R&D Systems; AF2030; goat polyclonal; 1:100) and anti-DNMT3L (Abcam; ab194095; rabbit polyclonal; 1:200), or anti-SOX9 (Millipore; AB5535; rabbit polyclonal, 1:300) and anti-AMH (Anti-Mullerian Hormone; Santa Cruz; sc6886; 1:300), or anti-DDX4 and anti-SOX9 in combination for 45 minutes. Dissociated mesonephros was used as a negative control for DDX4 staining, and dissociated E13.5 ovaries were used as a negative control for SOX9 staining. Cells were washed twice then incubated 45 minutes with secondary antibodies (Donkey anti-rabbit biotin, Thermo Fisher Scientific; Donkey anti-goat Alexa Fluor 488, Thermo Fisher) diluted 1:300. Cells were washed twice then resuspended and incubated in the Click-iT reaction cocktail containing Alexa Fluor 647 for 30 minutes. Following two washes, samples were incubated with Streptavidin Pacific Blue (Thermo Fisher Scientific; 1:500) for 30 minutes. Cells were washed twice, then resuspended in 300 µL permwash containing 20 µg/mL propidium iodide (Sigma) to measure cellular DNA content. Samples were run on the same day on the BD Biosciences FACS CANTO-II at Monash FlowCore (Monash Health and Translational Precinct (MHTP) Node) and data analysed using FlowJo X.0.7 software (Ashland, OR, USA). Single, intact cells were analysed following gating using forward and side scatter characteristics, and DNA content.

Cultured and uncultured testes were fixed in 4% PFA for immunofluorescence (IF) analysis. After standard ethanol processing conditions, they were paraffin embedded and sectioned at 4 µM onto Superfrost Plus slides.

Unless stated, all steps were performed at room temperature. Sections on slides underwent dewaxing in histosol, followed by rehydration in a graded ethanol series (100%, 95% and 70% ethanol). Slides were briefly washed in water and incubated for 30 mins at 98°C in Citrate buffer (pH 6; DAKO) for antigen retrieval. After cooling, slides were rinsed in distilled water, washed once in PBS, then sections permeabilised in 0.1% Triton-X-100 (Merck) in PBS for 30 mins. Slides were washed twice in PBS and a wax circle drawn around sections using a PAP pen (Cederlane Laboratories, Burlington, Canada). Non-specific antibody binding was blocked by incubation with 10% donkey serum (Sigma-Aldrich) in 5% bovine serum albumin (BSA)/PBS for 1 hour. The blocking liquid was tapped off and primary antibodies in dual combination were diluted in 1% BSA/PBS and added to sections, with 1% BSA/PBS serving as a control lacking primary antibody. Primary antibodies against the following proteins were used: DNMT3L (Abcam; ab194094; 1:200), VASA (R&D Systems; AF2030; raised in goat; 1:400), VASA (Cell Signalling Technologies; 8761S; raised in rabbit; 1:400), cKIT (R&D Systems; AF1356; 1:100), Laminin (Sigma; L9393; 1:200) and AMH (Santa Cruz; sc6886; 1:200). Slides were incubated overnight at 4°C in a humid chamber. The next day, primary antibodies were removed, and slides washed 3 x 5 minutes in PBS. Secondary antibodies (Donkey anti-Rabbit Alexa Fluor 594, Invitrogen, A21207; Donkey anti-Goat Alexa Fluor 488, A11055) were diluted 1:300 in 1% BSA/PBS and added to sections for 2 hours. Slides were washed one time in 0.1% Triton-X-100 in PBS, then twice for 5 minutes each in PBS. DAPI (Thermo Fisher Scientific) was applied to sections at 5 µg/mL in PBS for 15 minutes. Following three washes in PBS, slides were mounted under glass coverslips with ProLong Gold Anti-fade Mountant (Thermo Fisher Scientific) and set overnight. Imaging was performed by Monash Histology Platform (MHTP node) using the VS120 Slide Scanner (Olympus, Tokyo, Japan) and images were processed using OlyVIA software (Olympus).

Following dissociation of Swiss x Oct4-Gfp E13.5 testes and isolation of gonocytes via FACS, germ cells were counted using a haemocytometer. In each well (48-well plate), 20,000 germ cells were added in 500 µL media (MEM-α, 10% FBS, 1% penicillin-streptomycin) containing 10 µM SB431542, 5 ng/mL activin A, or relevant vehicle control. A lower concentration of activin A was used compared with the whole testis cultures, as cells grown as a monolayer have been demonstrated to be robustly responsive to 5 ng/mL activin A (29, 30). Cells were cultured for 24 hrs in 5% CO₂/air, after which the cells were collected for transcript analysis. Because the germ cells were lightly adherent, media and one PBS wash were collected to avoid losing cells, and 200 µL of 0.1% Trypsin-versene added per well and incubated for approximately 5 minutes or until all remaining cells were detached. Trypsin was quenched with media containing 10% serum, and all contents were transferred to a 1.5 mL tube containing the original media and PBS wash. Cells were centrifuged at 1020 g, supernatant removed, and cell pellets stored at -80°C.

All RNA extractions and on-column DNase treatment were performed using the NucleoSpin RNA XS kit (Machery-Nagel, Germany) according to the manufacturer’s protocol. RNA concentration was quantified using a NanoPhotometer (Implen, Munchen, Germany). RNA was subjected to reverse transcription in a reaction containing 200 Units SuperScript III Reverse Transcriptase (Thermo Fisher Scientific), 50 ng random primers and 500 ng oligo dTs (Promega, Madison, USA) per sample. For whole gonads, 100 ng RNA was added to the reaction. For isolated cell cultures and cells isolated following culture, 40 ng and 15 ng respectively, was used in each cDNA reaction and RNaseOUT Recombinant Ribonuclease Inhibitor (Thermo Fisher Scientific) added to each 20 µL reaction as per the manufacturer’s protocol.

Real time PCR was conducted on the QuantStudio Fast Real-time PCR System at the MHTP Medical Genomics Facility (Clayton, Australia), and data generated using SDS software (Applied Biosystems). Each reaction contained power SYBR Green Master Mix (Thermo Fisher) and specific primer pairs ( Table 1 ; Integrated DNA Technologies, Coralville, IA, USA) facilitating transcript measurements in 384 well plates. Primers pairs were designed to span exon-exon junctions or have pairs separated by an intron where possible. Each cDNA was diluted 1:20 or 1:10 for whole testes and isolated cells respectively. Every sample was measured in triplicate, and amplification of a single product was indicated by detection of a single peak in a melt curve analysis. Data were normalised to the Canx housekeeper gene (33) and analysed using the 2^-ΔCt method.

Additionally, transcript levels in isolated E13.5 germ cell cultures were measured using the Fluidigm Biomark™ 96x96 Dynamic Array IFC by the MHTP Medical Genomics Facility. Taqman assays (Thermo Fisher Scientific; Table 2 ) were used for amplification of specific transcripts. The geometric mean of two housekeeper genes, Canx and Mapk (33), was used for normalisation of data following a Pearson correlation between the two Ct values (R²>0.92). Data were normalised to the housekeepers, stable across samples, and analysed using the 2^-ΔCt method. Multiple experiments were analysed on the same array, accounting for the remaining samples and Taqman assays which make up the 96x96 array.

RNA-Sequencing was performed on gonocyte and somatic cells isolated from E13.5 and E15.5 Inhba x Oct4-Gfp, and germ cells isolated from E13.5 Inha x Oct4-Gfp wildtype and knockout testes. RNA sample quality was assessed on the Agilent 2100 Bioanalyzer using the Eukaryote total RNA Pico Kit, providing a measure of RNA integrity (RIN). All samples were high quality (RIN 8.4 – 9.9). Double stranded cDNA was prepared from 2-20 ng total RNA using Trio RNA-Seq or RNA-Seq V2 kits and SPIA amplification (Tecan/NuGEN, Leek, The Netherlands). These methods both use full length linear amplification to minimise bias. RNA-Seq libraries were then prepared with unique barcodes to allow multiplexing during sequencing. Illumina single end sequencing was performed on the HiSeq 3000 or NextSeq2000 (Illumina, San Diego, CA, USA). All RNA quality control, library preparation, and sequencing were performed by staff at the MHTP Medical Genomics Facility.

Sequencing from E13.5 and E15.5 Inhba x Oct4-Gfp returned 35-40 million 80 base pair reads. Sequencing from E13.5 Inha x Oct4-Gfp returned 65-85 million 100 base pair reads. The Inhba and Inha datasets were analysed independently of each other. RNA sequencing data were processed and analysed by Monash University Bioinformatics Platform. Sequencing reads were aligned to the Ensembl mouse reference genome GRCm38 (Ensembl release 84) and analysis was performed using the RNAsik pipeline with STAR aligner (34). Differential gene analysis was performed on Degust V4.1.5 (David R. Powell, Monash Bioinformatics Platform), using Limma-Voom (35, 36). Heatmaps were generated using ClustVis (37). RNA-Seq data are available via accession number GSE201520.

Inhba x Oct4-Gfp analysis: Following principal component analysis of the samples, two samples were excluded as outliers: one sample from the E13.5 Inhba KO somatic cell group, and one sample from the E13.5 Inhba WT gonocyte group. Further scrutiny of these samples led us to conclude these may have been contaminated or swapped, and their exclusion was supported following consultation with a bioinformatician (Monash Bioinformatics Platform).

The detection limit was determined by calculating the median of the entire array of counts per million (cpm) values for the datasets. For the entire dataset (germ and somatic cells), the detection limit was determined as 2.2 cpm, while the detection limit for the germ cell only dataset was calculated at 2.391 cpm. Values greater than these were determined as being detectable. Analysis of these data confirmed the purity of the germ and somatic cell populations through absence or presence of Ddx4 (germ cells), and Sox9 and Nr5a1 (somatic cells) ( Supplementary Figure 1A ). Absence of Inhba in knockout animals was confirmed in the somatic cell population with a 4-fold increase measured from E13.5 to E15.5 ( Supplementary Figure 1B ), consistent with previously published data (10, 12).

Differentially expressed genes were identified as having a false discovery rate (FDR) <0.05, a LogFC>0.585 and <-0.585 (i.e. a 1.5-fold change up or down, respectively), and two or more samples across genotypes being greater than the detection limit of 2.391cpm. There were 44 DEGs identified in the E15.5 Inhba KO gonocyte dataset. None were identified in the E13.5 Inhba KO dataset, therefore further analysis was performed to generate a list of transcripts that are altered in E13.5 gonocytes, described below. The data was processed in Degust using Limma-Voom, and the p-value was calculated within the software using the trimmed mean of M-values (TMM) normalised voom-transformed expression levels. Differentially expressed genes at E13.5 were identified using a LogFC> 0.585 and <-0.585, p-value <0.01 and restriction to at least two samples across wildtype and knockout animals being greater than the detection limit of 2.391 cpm. This approach enabled less abundant transcripts to be considered, and it resulted in the identification of 46 DEGs.

Inha x Oct4-Gfp analysis: Two wildtype and knockout littermate pairs were analysed using batch correction. Mitochondrial genes were filtered out as they were highly variable and genes with a minimum of 2 cpm in at least 2 samples included. Differentially expressed genes were identified by FDR<0.05 and LogFC>0.585. This led to the identification of 45 DEGs. Germ cell purity was assessed by the presence of Ddx4 and absence of Sox9 and Nr5a1 ( Supplementary Figure 1C ). Inha genotypes were confirmed in the somatic cell population by qRT-PCR ( Supplementary Figure 1D ).

Gene lists obtained after analysis ( Supplementary Tables 1–3) were submitted to the PANTHER classification system (38, 39) to identify molecular functions, biological processes, and protein classes of the DEGs. A Venn diagram was created following input of DEG lists to JVenn (40).

All statistical analyses were performed using GraphPad Prism Software (San Diego, CA, USA). Normal distribution of control and treatment groups was determined using a Shapiro-Wilk or D’Agostino and Pearson normality test. qRT-PCR and flow cytometric data from whole gonad cultures were analysed using an unpaired Student’s t-test for normally distributed data, or a Mann-Whitney test for data that was not normally distributed. For statistical analysis of the isolated cell culture experiments, a paired t-test or a Wilcoxon matched-pairs signed rank test was performed. Data was determined as significantly different when the p-value was less than 0.05.

Levels of transcripts encoding activin and Nodal ligands, signalling machinery and inhibitors were obtained from RNA-Seq analysis of germ and somatic cell populations collected from wildtype Inhba E13.5 and E15.5 testes ( Figure 1B ). These data reveal the complexity and dynamic nature of signalling potential of these selected components of the TGFβ superfamily within the testis during this window of development that is crucial to testis and embryo masculinization.

Inhba and Inhbb, encoding activin A and B subunits, respectively, were detected in somatic cells at both ages ( Figure 1B ), while Inhbc was below the detection limit in all samples (data not shown). Inhba increased 4-fold from E13.5 to E15.5 (73.8 ± 27.1 to 295.7 ± 69.7 cpm), and Inhbb levels were relatively constant (93.3 ± 6.7 cpm and 81.0 ± 10.9 cpm). At E13.5, Nodal was measured in germ cells, but not somatic cells, and it decreased to undetectable levels at E15.5 ( Figure 1B ). The levels of Inhba, Inhbb and Nodal were consistent with previous reports (12, 41). Transcripts encoding the Type 2 receptors for activin A, activin B and Nodal, Acvr2a and Acvr2b, were present in both somatic and germ cells at both ages highlighting the potential for each of these to respond, however Acvr2a was present at higher levels in both ages and cell types ( Figure 1B ). Acvr1b, encoding the type 1 receptor for activin A, activin B and Nodal, was present in both cell types at E13.5 and E15.5, while the transcript encoding the Nodal and activin B receptor, Acvr1c, was present only at low levels in E13.5 somatic cells (8.0 ± 2.3 cpm), indicating that Acvr1b, and not Acvr1c, is the predominant receptor for Nodal actions in germ cells at E13.5. Nodal signalling additionally requires the co-receptor, Cripto, encoded by Tdgf1, also known to antagonise activin A (42); this transcript was detected in E13.5 germ cells only (18.0 ± 1.2 cpm). These results illustrate the potential for Nodal to specifically impact on the germline cells which are exiting their proliferative state. Transcripts encoding the intracellular signalling components required for activin/Nodal signalling, Smad2 and Smad4, were present at both ages in somatic and germ cells, however Smad3 was predominantly detected in the somatic cell samples ( Figure 1B ).

Activin and Nodal inhibitors are also present during fetal testis development, and these would be expected to fine-tune the responsiveness of cells expressing their receptors ( Figure 1A ). Inha, encoding the inhibin α subunit which forms a potent activin A inhibitor when dimerised with an activin β subunit, was detected only in somatic cells at both E13.5 (60.8 ± 8.1 cpm) and E15.5 (85.9 ± 34.7 cpm). Follistatin (Fst) was detected at low levels (<7 cpm) in all samples ( Figure 1B ), consistent with previous studies demonstrating that Fst is only expressed in the fetal ovary compared with the testis (43, 44). The transcript encoding the decoy receptor Bambi (45) was expressed at both ages in somatic and germ cells, with consistently higher levels in somatic cell samples compared with those in germ cells (53.6 ± 2.3 and 37.3 ± 1.7 cpm in somatic cells, and 9.7 ± 1.6 and 9.6 ± 2.4 cpm in germ cells). Transcripts encoding the inhibitory Smad6 and Smad7 were predominantly expressed in the somatic cells, but were also measured in germ cells at both ages. Betaglycan, encoded by Tgfbr3, is a co-receptor for TGFβs which is required for TGFβ2 signalling, and it can inhibit activin A (46). It was highly expressed in somatic cells compared to germ cells at both ages ( Figure 1B ).

There are several Nodal antagonists which could dampen its capacity to compete with activin A. Lefty1 was identified in germ cells at E13.5 ( Figure 1B ; 63.6 ± 8.1 cpm) and undetectable by E15.5, consistent with previous observations (12). Lefty2 was expressed at higher levels in E13.5 germ cells (192.8 ± 49.9 cpm) and dropped to 4.3 ± 1.5 cpm by E15.5. Lefty2 transcripts were also low in the somatic cells at both ages ( Figure 1B ). The Cerberus transcript, encoding another Nodal inhibitor, was below the detection limit in all samples (data not shown). The Nomo/Nicalin complex has been identified as a Nodal antagonist in zebrafish (47), however its roles in the mouse are not known. Transcripts for each component were present in the mouse fetal testis (Nomo1 and Ncln) at E13.5 and E15.5 in both somatic and germ cells ( Figure 1B ), indicating these proteins may also reduce Nodal activity in the fetal testis.

In germ cells lacking activin A (Inhba KO), there were 46 and 44 differentially expressed genes (DEGs) at E13.5 and E15.5, respectively ( Figures 2A, B ; Supplementary Tables 1, 2). At E13.5, there were no DEGs by FDR (<0.05), therefore we utilised p-value (<0.01) and LogFC (>0.585 and <-0.585) to assess any differences between genotypes ( Figure 2A ). There were 21 downregulated, and 25 upregulated DEGs, which were primarily associated with binding and catalytic functions, and cellular processes. The top protein class was identified as being metabolite interconversion enzymes, which convert one small molecule to another (PANTHER, Table 3 ), however, the function of these genes in the testis are unknown. Interestingly, an association between Galnt6 and piRNAs has been identified in an oral squamous cell carcinoma mouse model (48).

At E15.5, there were 27 downregulated, and 17 upregulated DEGs ( Figure 2B ), however there was no overlap in DEGs between E13.5 and E15.5. These transcripts were similarly associated with binding and catalytic activity (PANTHER, Table 3 ). The top biological processes were cellular processes, and biological regulation, with the top protein classes identified as cytoskeletal proteins, gene-specific transcriptional regulators, protein binding activity modulators, and transporters (PANTHER, Table 3 ). Within the DEGs, Musashi-1 (Msi1), an RNA-binding protein in the Musashi family of proteins which function in translational regulation, was identified as lower in KO germ cells compared with WT counterparts. Its essential role in governance of postnatal transitions of murine spermatogenesis has been established, and it was previously shown to be expressed in gonocytes (49). There was no overlap in DEGs between E13.5 and E15.5, indicating age-specific responses of germ cells occurred in the absence of activin A.

A recent study identified that activin A promotes a less differentiated transcript profile in the human germ cell-like cell line, TCam-2 (30). To determine if germline differentiation was similarly altered in Inhba KO mice, we examined early and differentiation-associated germ cell transcripts in the Inhba WT and KO RNA-Seq dataset. Early germ cell transcripts Nodal, Tdgf1, Kit, Lefty1, Lefty2, and Nanog were all downregulated between E13.5 and E15.5 in both WT and KO samples, while differentiation markers Nanos2 and Dnmt3l were upregulated ( Supplementary Figure 2 ). We also observed higher expression of piRNA pathway transcripts such as Piwil1, Piwil2 and Piwil4, Dnmt3a, Dnmt3l, Tdrd1, Tdrd9, Mael and Mov10l1 at E15.5 relative to E13.5 in WT and KO samples. This is consistent with the activation of the piRNA pathway and de novo methylation from around E14.5-E15.5 in quiescent germ cells (50); the higher level of piRNA pathway transcripts encoding components such as in our dataset is consistent with the normal progression of developmental events associated with this phenomenon ( Supplementary Figure 2 ). Of these transcripts, Mov10l1 was decreased (p<0.05, Mann-Whitney test) in the KO germ cells at E15.5, however this was not determined to be differentially expressed in the RNA-Seq dataset by FDR and fold change, as presented in Figure 2B . Transcripts associated with pluripotency and differentiation showed no differences in isolated germ cells from Inhba KO testes compared with WT counterparts at either E13.5 or E15.5. The germ cell-specific transcripts, Ddx4 and Pou5f1 (Oct4), were both detected at relatively high levels in germ cells, with Ddx4 increasing 1.6-fold from E13.5 to E15.5 in WT cells.

While Inhba KO germ cells appear to differentiate normally based on classical germ cell markers, a subset of genes was altered, indicating that loss of activin A modulates some aspects of early male germline transcription. However, the significance of the outcomes remains to be determined.

The elevation of activin A levels linked with pre-eclampsia in human pregnancy can occur in the second and third trimesters when male germ cells are mainly quiescent. The inhibin α subunit encoded by Inha, forms a dimer with an INHBA subunit to form the inhibin A protein, a potent inhibitor of activin A. In Inha KO mice, activin A bioactivity is elevated due to the combined absence of inhibitory inhibin proteins, and to the greater availability of INHBA subunits for dimerization to form activin proteins. In wildtype mice, Inhba is detectable from E11.5, with its levels increasing until just after birth (10, 51). As the phenotype of the E13.5 testis appears normal but is significantly different by E15.5 (data not shown), we examined the germ cell transcriptome in Inha KO compared to WT littermates, prior to gross morphology changes. RNA-Seq analysis of germ cells isolated from two independent wildtype and knockout littermate pairs identified 45 DEGs ( Figure 2C and Supplementary Table 3 ; FDR<0.05, LogFC>0.585, <-0.585). Thirty upregulated transcripts included ribosome structural components such as Rps15, Rps25, Rps5, Rplp1 and Rps3. These transcripts are also associated with RNA binding. Pathway analysis revealed that the top molecular functions of the 45 DEGs were binding and catalytic activity, with cellular processes the top associated biological process. Inha KO DEG were associated with translational proteins (primarily the ribosomal structural component transcripts), and gene-specific transcriptional regulators, which included Sox12, Egr1, Etv1 (upregulated), and Prdm10 (downregulated) ( Table 3 ). Interestingly, there were no reciprocal DEGs between Inha E13.5 germ cells and the E13.5 or E15.5 Inhba germ cells ( Figure 2D ). Collectively, these results demonstrate that gonocytes which develop in an environment of altered activin bioactivity are different from their wildtype counterparts, leading us to investigate whether this effect is direct or indirect.

RNA-Seq revealed differences in male germ cell mRNA profiles in mice with altered activin A bioavailability ( Figure 2 ). To test whether activin A can directly affect germ cells, gonocytes isolated from E13.5 testes were cultured for 24 hours in 5 ng/mL activin A or 10 µM SB431542, and appropriate vehicle controls. After 24 hours in culture, germ cells retained Oct4-eGFP expression, as observed by fluorescence microscopy ( Figure 3A ). Transcripts encoding markers of germ cell differentiation were measured in isolated E13.5 gonocytes and first compared with levels in cells cultured for 24 hours in control conditions ( Supplementary Figure 3 ). After 24 hours in culture, the early germ cell marker Kit had declined to 85% of E13.5 levels, and Nodal was at 10% of E13.5 levels. The differentiation marker Nanos2 was moderately increased (1.8-fold), while Dnmt3l, Piwil4, Tdrd1 and Mov10l1 were all higher after 24 hours in culture compared with E13.5 levels (16-, 11-, 6- and 3-fold, respectively). Interestingly, germ cell markers Oct4 and Mvh increased over time ( Supplementary Figure 3 ). The decrease in Nodal and increase in differentiation markers suggests that E13.5 germ cells can autonomously differentiate outside of the somatic environment.

Oct4 and Mvh transcripts were unaffected by activin A exposure, however SB431542 resulted in a significant decrease in Oct4 (0.8-fold). Nodal and Lefty2 levels were also unaffected by activin A exposure, however both were lower in SB431542-treated cells, consistent with previous reports (12). Kit was significantly lower following activin A exposure (0.85-fold), and significantly higher following SB431542 (1.75-fold), and Tdgf1, encoding the Nodal co-receptor, was significantly reduced by activin A ( Figure 3A ). The mRNA encoding the Nodal inhibitor Lefty2 is a known activin A-responsive gene, demonstrated in mouse embryonic stem cells and P19 embryonic carcinoma cells (52, 53), and in human TCam-2 cells (30). This responsiveness was also demonstrated here in isolated gonocytes, with a 1.61-fold increase in Lefty2 following activin A exposure ( Figure 3A ).

In addition, treatment of E13.5 gonocytes with activin A resulted in a more differentiated transcript profile, with significant elevation of Nanos2, Piwil4, Dnmt3l and Mov10l1. Further, SB431542 decreased Nanos2, consistent with whole gonad culture, and Piwil4, while Kit increased These results demonstrate that gonocytes respond directly to activin A and the inhibition of its pathway in culture ( Figure 3A ).

Two transcripts, Musashi-1 (Msi1) and Solute carrier family 43 member 3 (Slc43a3), identified as DEGs in the RNA-Seq data from E15.5 activin A knockout mouse testes ( Figure 2B ), were investigated in these samples. Following exposure to SB431542, Msi1 was significantly decreased to 0.86-fold of control levels in E13.5 gonocytes after 24 hours, but it was not affected by activin A ( Figure 3B ). In Inhba WT germ cells, Msi1 normally increases 10-fold between E13.5 and E15.5. This was reduced to a 4-fold increase between E13.5 and E15.5 in Inhba KO germ cells, resulting in a significant difference in expression levels between wildtype and knockout germ cells at E15.5 (60% decrease, Figure 3C ). In the isolated somatic cells of Inhba WT and Inhba KO testes, examined using RNA-seq, the level of Msi1 recorded was greater than in germ cells (>20 cpm; data not shown) (9) but was not different between genotypes. Thus, Msi1 appears to be a germ cell-specific activin A target gene, a conclusion supported by the results in the E13.5 isolated germ cell cultures in which Msi1 was significantly decreased following activin/Nodal/TGFβ inhibition, and that it was significantly reduced in germ cells of Inhba KO animals at E15.5.

Slc43a3, originally identified as an equilibrative nucleobase transporter, has also been identified as influencing fatty acid flux (54, 55) but its function in the testis is unknown. Slc43a3 was lower in the E15.5 Inhba KO germ cells compared to WT ( Figures 2B , 3C ), and was significantly higher in activin A-treated gonocytes (1.26-fold of controls) ( Figure 3B ) suggesting that it is upregulated by activin A directly in germ cells. While Slc43a3 was not altered following SB431542 exposure; this may be due to a difference between the chronic absence of activin A in the Inhba KO mouse and acute inhibition in these cultures via SB431542. It is also important to consider that the germ cells may be developmentally different, or that Slc43a3 transcript may be relatively stable and therefore not reduced within the 24-hour window examined in the isolated E13.5 germ cells.

After determining that gonocytes can directly respond to activin A and SB431542 through altered gene expression, we cultured whole testes to assess the outcome of altered signalling on germ cells within their somatic niche. We first performed a dose-response, to determine the optimal concentration of activin A. E13.5 testes were cultured with 5, 25, 50 or 100 ng/mL activin A for 48 hours and compared with control samples cultured in the vehicle. Levels of known activin A-induced somatic cell transcripts, Hsd17b1, Ccl17 and Serpina5 (9), were monitored to determine the optimal dose at which responses were evident. Hsd17b1 was significantly higher in testes exposed to 25, 50 and 100 ng/mL, while Ccl17 and Serpina5 were significantly higher in testes exposed to 50 and 100 ng/mL of activin A, when compared with vehicle controls ( Figure 4 ). Because all three transcripts were increased following exposure to at least 50 ng/mL activin A, this concentration was chosen for subsequent experiments.

E13.5 testes were cultured with 50 ng/mL activin A or with the activin/Nodal/TGFβ inhibitor, SB431542, which blocks ligand access to the Type 1 receptors, ALK4, ALK7 and ALK5 (56). Testes were photographed immediately after collection at E13.5 and after 48 hours of culture. Testis cords were easily observed in the E13.5 testes, and after 48 hours of culture in either vehicle, the cords appeared elongated, contained GFP-positive germ cells, and were grossly of the shape normally observed in vivo at E15.5 ( Figure 5A ). In contrast, after 48 hours in culture the effectiveness of inhibitor treatment was evident based on the appearance of cords that were fatter and appear stunted, compared with the DMSO controls ( Figure 5A ), previously demonstrated by Miles and colleagues in cultures beginning at E12.5 (12). Cords in testes cultured with activin A were grossly similar to control testes but appeared to be slightly thinner. Activin A target gene transcripts were measured by qRT-PCR. Ccl17, Serpina5, Hsd17b1 and Gja1 (encoding gap-junction protein Connexin 43, expressed in Sertoli cells) were significantly higher than in corresponding control samples following activin A exposure (5.8-, 3.5-, 4.5- and 1.7-fold, respectively; Figure 5B ), and significantly lower in SB431542-treated testes (0.18-, 0.3-, 0.04- and 0.63-fold of control) ( Figure 5C ), confirming the efficacy of these treatments and demonstrating a dose-dependency of these transcript levels as previously reported in vivo (9). Cldn11, also encoding a component of Sertoli cell tight junctions, decreased in post-pubertal rat Sertoli cell in vitro cultures following activin A exposure (22). The finding that Cldn11 was significantly lower in activin A-treated fetal testes (0.36-fold), and significantly increased in SB431542-treated testes (3.3-fold) ( Figures 5B, C ) indicates that the responsiveness of these genes to activin bioactivity is likely to be conserved through the Sertoli cell lifespan.

Matrix metalloproteinases are involved in tissue remodelling and have been detected in the fetal testis (57, 58). Exposure of the human gonocyte-like seminoma cell line, TCam-2, to activin A increased both MMP2 transcript and protein levels (59). Therefore, Mmp2 was also assessed as a potential activin A target in the mouse fetal testis. Activin A exposure did not alter Mmp2 transcript in fetal mouse testes, however SB431542 significantly decreased 0.43-fold of controls ( Figures 5B, C ). Mmp2 may not be solely upregulated by activin A, as its decrease following SB431542 exposure could be due to the inhibition of TGFβs or Nodal. Alternatively, Mmp2 synthesis could have already been at the highest level normally reached by activin A stimulation by the levels present at E13.5. Opposing regulation of Ccl17, Serpina5, Hsd17b3, Gja1 and Cldn11 by activin A and SB431542 demonstrates the effectiveness of each in culture, while extending our knowledge of how transcripts encoding extracellular matrix components are regulated in the fetal gonad.

To assess Sertoli cells, Sox9 transcription was measured following whole gonad culture with activin A or SB431542. Interestingly, Sox9 transcript was significantly lower in activin A-treated gonads (0.76-fold) and significantly higher (1.33-fold) in SB431542-treated gonads ( Figures 5B, C ). This was consistent with our RNA-sequencing analysis of fetal somatic cells from Inhba KO mice (data not shown) which collectively suggests that Sox9 transcription or turnover may be modulated by activin A.

Testes in fetal mice lacking activin A have a reduced proportion of proliferative Sertoli cells (10, 19), and E12.5 testes exposed to SB431542 for 72 hours exhibited a five-fold decrease in Sertoli cell proliferation (12). To assess the effects of activin A and SB431542 on cell proliferation in cultured whole testes, Edu-incorporation followed by flow cytometry was employed. Fetal Sertoli cells, detected by AMH immunostaining, comprised 14% of the total cell population after 48 hours in culture with vehicle controls (HCl, DMSO); testes exposed to activin A had a significantly higher proportion of fetal Sertoli cells, with a 1.5-fold increase to 21% of AMH-positive cells. Conversely, SB431542 exposure significantly reduced the proportion of Sertoli cells to 9% (0.65-fold of DMSO levels) ( Figure 5D ). Consistent with this, we observed a significant increase in the proportion of EdU-positive Sertoli cells following activin A exposure (1.7-fold), demonstrating that activin A increased Sertoli cell proliferation, and a decrease following SB431542 exposure (0.37-fold), demonstrating decreased Sertoli cell proliferation ( Figure 5E ). In addition, the mean fluorescent intensity (MFI) of AMH in AMH-positive cells was also measured as an indication of relative protein levels; Flow cytometric analysis revealed that SB431542 significantly reduced the AMH MFI ( Figure 5F ), and this was confirmed in sections of SB431542-treated testes analysed using immunofluorescence staining ( Figure 5G ).

The Oct4-Gfp transgene allowed visualisation of germ cells by fluorescent microscopy after culture. Based on GFP localisation, germ cells appeared restricted to the cords ( Figure 5A ). Levels of germ cell transcripts, assessed by qRT-PCR, were compared between E13.5 whole testes and testes cultured for 48 hours in vehicle. These were also examined against the RNA-seq data of wildtype E13.5 and E15.5 gonocyte populations isolated from Inhba x Oct4-Gfp mice. Early germ cell transcripts Kit, Nodal, Nanog and Tdgf1 were lower in testes after 48 hours in culture compared with E13.5 testes, and the differentiation markers Nanos2, Dnmt3a, Dnmt3l, Mov10l1, Piwil2, Piwil4, Tdrd9 and Tdrd1, normally upregulated by E15.5, were all increased after 48 hours in culture. These findings were consistent the changes measured by RNA-Seq ( Figure 6A ) and demonstrate the suitability of the culture system for investigating effects on germ cell development.

Treatment of E12.5 testes with 10 µM SB431542 for 72 hours previously resulted in an increased proportion of germ cells escaping mitotic arrest, with a 4-fold increase (3% to 14%) in germ cells incorporating EdU (12). This indicates that blocking ALK4/5/7 signalling has a robust effect on mitotic arrest. To assess the window of vulnerability of germ cells to this disruption, and to assess whether the proportion of germ cells in this sub-population was sustained, we investigated whether E13.5 testes were similarly susceptible to SB431542 treatment, an age when most germ cells have already entered mitotic arrest. In parallel, we sought to determine whether exposure to exogenous activin A would influence germ cell numbers or proliferation. For these studies, EdU-incorporation and flow cytometry were employed detect proliferating MVH-positive (germ) cells after 48 hours in culture. Germ cells comprised approximately 20% of the total cell population in the cultured testes. There were no significant differences in this value between treatment and control samples after 48 hours in culture, but there was a trend to fewer germ cells following activin A treatment (0.86-fold, p=0.0719) ( Figure 6B ). The proportion of Edu⁺ germ cells in the SB431542 treatment group was increased (2% Edu⁺, compared with controls, 0.5% Edu⁺) ( Figure 6C ). This was statistically significant, and indicates that a small proportion of germ cells in E13.5 testes retain the capacity to escape mitotic arrest. Moreover, together the observations that SB431542 diverts a greater proportion of the germ cell population from mitotic arrest at E12.5 compared to E13.5, indicate that there is a window at around E12.5 during which inhibiting AKL4/5/7 can divert germ cells from their normal entry into mitotic arrest.

To further examine the relevance of this pathway to fetal germ cell differentiation in these whole fetal testis cultures, key markers were measured by qRT-PCR. Early germ cell marker transcripts Nodal, Kit and Oct4 were not different following activin A treatment, and amongst key transcripts normally upregulated between E13.5 and E15.5 (Nanos2, Dnmt3l, Dnmt3a, Mov10l1, Tdrd1, Tdrd9, Piwil2 and Piwil4), only Mov10l1 was affected and was 11% lower than in the control sample. However, the germ cell marker Mvh was reduced by 12% following activin A treatment ( Figure 6D ).

E13.5 testes exposed to 10 µM SB431542 exhibited a less-differentiated transcript profile. Nodal is highly expressed at E13.5 in germ cells and decreases to <20% by E15.5 ( Figure 1B ). After 48 hours of culture with SB431542, Nodal was downregulated to 54% of the control level ( Figure 6E ). Nodal upregulates its own expression (32, 41), and because SB431542 blocks Nodal signalling through ALK4/5/7 inhibition, this downregulation of Nodal was expected, and consistent with findings from Miles and colleagues (12). The early germ cell marker Kit was significantly higher following SB431542 exposure (1.22-fold, compared with controls; Figure 6E ). While Kit is also expressed in somatic cells, the Inhba KO RNA-Seq data shows that at E13.5, Kit is predominantly expressed in germ cells (196 ± 5 cpm vs 36 ± 4 cpm in somatic cells; Figure 7A ) suggesting that the increase in Kit is most likely due to an effect of activin/Nodal/TGFβ inhibition on germ cells.

In SB431542-treated testes, the gonocyte differentiation marker Nanos2 was reduced to 70% of controls, and several PIWI/piRNA pathway components, which normally increase by E15.5 when germ cells are quiescent, were also reduced. The de novo DNA methyltransferases Dnmt3a and Dnmt3l were both reduced by 14% to 86% of control levels in SB431542-treated gonads compared to controls. Similarly, Piwil4 was reduced to 76% of controls, but there was no change in Mov10l1, Tdrd1, Tdrd9 or Piwil2 levels. The germ cell markers Oct4 and Mvh were reduced to 69% and 93% of control levels following SB431542 exposure, however there was no change in germ cell numbers ( Figure 6E ). Collectively, these changes indicate a modest transcriptional response of these genes to activin/Nodal/TGFβ inhibition.

Because the early germ cell marker Kit was upregulated, and the differentiation marker Dnmt3l was downregulated following SB431542 exposure, immunofluorescence staining for these two markers was performed on E13.5 testes, and on the activin A and SB431542 treatment samples. KIT was co-localised with MVH in E13.5 germ cells, corresponding with transcript data, but was not detectable in SB431542-treated testes ( Figure 7B ), despite transcript up-regulation. Dnmt3l values in germ and somatic cells at E13.5 are below 1 cpm and increase in germ cells to 115 ± 50 cpm at E15.5 ( Figure 7A ). By immunofluorescence, DNMT3L was not detectable in any MVH-positive germ cells at E13.5 but was detected in the nucleus of germ cells after 48 hours of culture in every treatment group, consistent with its normal upregulation by E15.5 ( Figure 7C ). There were no obvious differences between activin A- or SB431542-treated testes compared with their respective controls. DNMT3L appeared to be heterogeneously distributed, with bright and dim staining present in individual germ cell nuclei ( Figure 7C ), however flow cytometry revealed no difference in DNMT3L-positve germ cells ( Figure 7D ) or its MFI between treatment groups ( Figure 7E ). Further scrutiny of the data did not reveal any distinct “bright” or “dim” populations, nor differences in their distribution across treatment groups.

These data suggest that inhibition of activin/Nodal/TGFβ activity in E13.5 testes cultured for 48 hours results in a less-differentiated germ cell transcript profile. Considering that a small subpopulation of germ cells escaped mitotic arrest in SB-treated gonads ( Figure 6C ), it is possible that the changes observed in the transcript profiles may reflect only the small population of germ cells that have not yet entered quiescence.

After documenting the impact on fetal germ cells of chronic activin A disruptions in transgenic mouse models and demonstrating that isolated germ cells in culture can respond directly to activin A, we wanted to extend our knowledge of how exogenous activin A and SB431542 exposures each affect the germ cells within the intact testis environment. Whole E13.5 testes were dissociated after 48h culture with activin A or SB431542, and the gonocytes isolated by FACS for transcript analysis. Known activin A target genes were analysed in the isolated somatic cells to confirm the effectiveness of activin A and SB431542 treatments in the cultures. These results were consistent with the previous whole gonad cultures ( Supplementary Figure 4 ). Transcript analysis of isolated gonocytes after whole testis culture with activin A revealed no changes in the early germ cell (Kit), or differentiation (Dnmt3l, Nanos2, Mov10l1, Piwil4 or Dnmt3a) markers. SB431542 exposure resulted in significantly increased Kit levels (2-fold), consistent with whole testis analysis and isolated germ cell cultures ( Figure 8 ). Interestingly, SB431542-exposure did not result in decreased levels of the differentiation marker transcripts Dnmt3l, Nanos2, Piwil4 or Dnmt3a. Unexpectedly, Mov10l1 was significantly increased (1.4-fold; Figure 8 ). These data suggest that, while gonocytes can directly respond to perturbed activin A and TGFβ signalling, the effects are minimized whilst they reside in an intact somatic environment.