Two established mouse lines formed the basis of this work: Inha KO, lacking the Inha mature subunit coding sequence (31) and Inhba KO (BA 192), which lacks the Inhba mature subunit coding sequence (33). Both were maintained on a C57/Bl6 background. In addition, we utilised the Inha x OG2 (bearing Oct4-GFP) mouse strain described previously (19). All three lines were maintained by heterozygote breeding, and genotypes were determined by commercial vendor using real-time PCR (Transnetyx, TN, USA). Testes were collected from E13.5, E15.5 and PND 0 testes of Inha WT (wildtype) and KO littermates, E12.5, E13.5, E14.5 and E15.5 Inhba WT and KO littermates, and E15.5 WT and E17.5 WT and KO pups from the Inha x OG2 strain. For timed matings, the presence of a vaginal plug was used to determine embryonic day (E) 0.5 of pregnancy. Embryonic age was determined at the time of collection from fore- and hind-limb development and male gonads identified by the presence of testis cords. Natural birth, occurring on E19.5 – E20.5, was recorded as postnatal day 0 (PND 0). Testes were dissected free from mesonephros or epididymis at the time of collection. Mice were housed at the Monash Medical Centre Animal Facility in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (1997), with a 12hr light/12hr dark cycle with food and water available ad libitum. This study was approved by the Monash University Animal Ethics Committee.

Testes were collected from E13.5 and E15.5 embryos and PND 0 pups. One testis was fixed in Bouin’s solution (Sigma-Aldrich) for 20 - 60 mins depending on age (E13.5: 20 mins, E15.5: 30 mins, PND 0: 60 mins), washed twice in 70% ethanol and stored at 4°C in 70% ethanol. Tissues were processed and embedded in paraffin wax using standard protocols by the Monash University Histology Platform (MHTP node), with 4 µm serial sections cut onto Superfrost™ Plus (Fisher Scientific) slides. The second testis was either fixed in 4% paraformaldehyde in PBS (PFA) (Alfa Aesar, Cat # J61899) for 40-90 mins depending on age (E13.5: 40 mins, E15.5: 60mins, PND 0: 90 mins) or snap-frozen and stored at -80°C for RNA. Incubations and washes were at room temperature (RT) unless otherwise indicated.

Sections on slides were dewaxed in histosol, then rehydrated in a series of decreasing ethanol concentrations (100% to 70%). All washes and incubations were at RT unless otherwise specified. Antigen retrieval was performed by microwaving slides in 50 mM glycine buffer (pH 3.5) for 10 minutes, then allowing them to cool for 20 mins. Tissue sections were blocked using 5% bovine serum albumin (BSA, Sigma-Aldrich, Cat # A7906) diluted in phosphate buffered saline (PBS, made from a 10X PBS stock, Gibco, Cat #70011-044) to prevent non-specific staining. All primary antibodies (DDX4, AF2030, R&D Systems, 1:400, RRID : AB_2277369; SOX9, Santa Cruz, sc-20095, 1:500, RRID : AB_661282; SMA, Sigma-Aldrich, A2547, 1:1000, RRID : AB_476701) were diluted in 1% BSA/PBS and applied overnight at 4°C. Negative control slides lacked primary antibodies. The next day, slides were washed 3 x 5 minutes in PBS, and sections were incubated with appropriate secondary antibodies (Alexa Fluor 488 donkey anti-mouse, Cat # A-21202, RRID : AB_141607; Alexa Fluor 488 donkey anti-goat, Cat # A-11055, RRID : AB_2534102; Alexa Fluor 546 goat anti-rabbit Cat # A-11010, RRID : AB_2534077, Alexa Fluor 546 rabbit anti-goat, Cat# A-21085, RRID : AB_1500595, all Invitrogen, 1:500) for 1 hour, then washed 3 x 5 minutes in PBS. Sections were mounted under coverslips with ProLong™ Gold Antifade Mountant with DAPI (Invitrogen, Cat # P36935) and stored at 4°C prior to image capture.

Whole E15.5 testis pairs from WT Inha x OG2 embryos were cut in half using a feather knife and placed on 30 mm diameter Millicell^® filters (Merck, Cat # PICM03050) in a 6 well plate, containing 0.6 mL of Dulbecco’s Modified Eagle Medium/F12 + Glutamax (Gibco, Cat # 105650-018) with 1% fetal calf serum (Bovogen) and 1% Pen/Strep (Gibco, Cat # 15670-063). Half testis fragments were cultured at 37°C in 5% CO₂ in 25 ng/ml activin A (R&D Systems, Cat # AF338, RRID : AB_355307), 200 µm mono-2-ethylhexyl phthalate (MEHP; Santa Cruz, Cat # SC396467), 25 ng/ml activin A plus 200 µm MEHP, or appropriate vehicle control (4 μM HCl for activin A, DMSO for MEHP, or HCl plus DMSO, respectively). Testes from a single animal were used for both treatment and control groups. Media (50%) was refreshed daily. After 72 hours, fragments were fixed in Bouin’s for 1 hour and processed for histological analysis, or snap frozen on dry ice and stored at -80°C for transcriptional analyses.

Haematoxylin and immunohistochemistry images were collected using a light microscope (Olympus BX50 microscope with a DP70 camera). Immunofluorescence samples were scanned with a 40x objective, using the Olympus VS120^® Virtual Slide Microscope System at the Monash Histology Platform (MHTP node). Germ cells, Sertoli cells, cord area, and proliferating cells were identified in Inha and Inhba WT and KO mouse testes and in cultured testis fragment sections by immunofluorescence detection of DDX4 (gonocyte cytoplasm), SOX9 (Sertoli cell nuclei), and alpha-smooth muscle actin (SMA; peritubular cell cytoplasm, marking cord perimeter). Non-specific fluorescence signal was removed by setting the baseline exposure for section analysis equal to that of the negative control samples, which displayed no signal. Images were calibrated prior to analysis using ImageJ (RRID : SCR_003070) (34). From each testis cross section, the number of cords per 100 mm², the average cord area in each section (Inha – E13.5: 15-30 cords, E15.5: 25-50 cords, PND 0: 50-100 cords; Inhba – E13.5: 15-30 cords, E15.5: 15-50 cords, PND 0: 15-100 cords), the total section area (mm²), and area of the section occupied by cords (mm²) were measured. Sertoli cell and gonocyte number were recorded for each section; the cross-section area, and area occupied by cords were used to normalise Sertoli and germ cell numbers. DDX4+ cells containing a visible nucleus both inside and outside cords were counted as gonocytes. Bi- and multinucleated gonocytes, identified by a DDX4+ cytoplasm surrounding closely opposed nuclei, were marked as abnormal and each was counted as a single cell.

Inha and Inhba fetal gonad age series and cultured E15.5 testes samples were measured in separate Fluidigm assays using the Fluidigm Biomark™ HD system (96×96 and 48x48 Dynamic Array IFC, respectively; Fluidigm Corporation) by the Monash Health Translation Precinct Medical Genomics Facility. For the Inha age series, whole, paired gonads from Inha WT and KO individual animals at E13.5, E15.5 and PND 0 (n=3/age/genotype) were collected from 2-3 litters. For the Inhba age series, paired gonads from Inhba WT and KO individual animals at E12.5, E13.5, E14.5 and E15.5 (n=3/age/genotype) were collected from 2-3 litters as previously described (19). Transcripts from E15.5 testis halves were measured following 72 hr culture with MEHP (200 µm) or vehicle control (n = 6-7/treatment). RNA extraction and on-column DNase-treatment was performed using the NucleoSpin RNA XS kit (Machery-Nagel) according to the manufacturer’s protocol. RNA quantity was determined using the Nanodrop (Thermo Scientific) and RNA quality was confirmed on a subset of samples using the Agilent 2100 Bioanalyzer (RNA integrity number (RIN) > 8.5). Reverse transcription of 300 ng (whole testes, Inha and Inhba age series) or 70 ng (cultured testis fragments) total RNA was performed using SuperScript™ III Reverse Transcriptase (Invitrogen) with random hexamer oligonucleotides (Promega), according to manufacturer’s guidelines. Fluidigm assays were performed using pre-designed and validated Taqman gene probes (Life Technologies). Raw data were analysed with Fluidigm Real-Time PCR Analysis Software (RRID : SCR_015686, Version 4.1.2). Negative control samples were below the detection limit. Both reference transcripts, Canx and Mapk1 (31), were present at stable levels in all ages and genotypes, and an R value > 0.91 was obtained across all samples indicating they are positively correlated. On this basis, the average of both reference transcript values was used to normalize gene expression data using the ΔΔCT method (35).

Snap frozen E17.5 paired testes from individual Inha x OG2 WT and KO mice were homogenized in 130 µl ice cold homogenization buffer (0.5% BSA (w/v), 5 mM EDTA in PBS, pH 7.4) in a 5 ml glass tube for 20 seconds using an IKA T10 basic disperser on the highest setting; the probe was rinsed between samples in homogenisation buffer. Samples were centrifuged (3000 rpm, 10 min, 4°C) and supernatants transferred to new 1.5 ml plastic tubes for storage at -80°C until analysis. Steroid hormones were measured for all samples in a single batch using liquid chromatography–tandem mass spectrometry (LC–MS/MS) assay as previously described (19). The limit of detection (LOD) was set at the limit of quantification (LOQ) and represents the lower limit of assay sensitivity. Not detectable values were set at LOQ per analyte/2. The profile of steroids included androstenedione (A4), testosterone (T), 11-keto A4 (11-KA), 11-keto T (11-KT), 3α androstanediol (3α-diol), 3β androstanediol (3β-diol), dihydrotestosterone (DHT) and 11-keto DHT (11-KDHT), mirroring our previous analysis of the Inhba mouse testis at E17.5 (19).

Bulk RNA sequencing was performed on DNase-treated RNA from E13.5, E15.5 and PND 0 Inha WT and KO whole mouse testes (n = 2-3/age/genotype). Samples were sequenced by BGI-Hong Kong Co Ltd) on the DNBSEQ-G400 with PE100. Raw reads were filtered (using SOAPnuke software, BGI) to remove adaptor sequences, contamination and low-quality reads and stored in a FASTQ format. Bioinformatics analysis was performed at the Monash Bioinformatics Platform. Raw FASTQ files were analysed using the RNAsik (v1.5.4) pipeline (36) using STAR aligner (37) with GRCm38 (Mus musculus) reference genome. Reads were quantified using featureCounts (38) producing the raw genes count matrix and various quality control metrics. RNA-seq data are available via accession number: GSE236618. Raw counts were then analysed with Degust (39), a web tool which performs differential expression analysis using limma voom normalisation (40), producing counts per million (CPM) library size normalisation and trimmed mean of M values (TMM) normalisation (41) for RNA composition normalisation. Differentially expressed genes in Inha WT and KO testis at each age were defined as those with a false-discovery rate (FDR) of < 0.05, showing a > 3-fold (absolute log fold-change >1.5) change in expression. A heat map was generated using ClustVis (RRID : SCR_017133) (42). The list of DEGs was submitted to DAVID (RRID : SCR_001881) (43) to identify significantly associated functions. Venn diagrams were created using jvenn (RRID : SCR_016343) (44). Ingenuity Pathway Analysis software (IPA, RRID : SCR_008653) (Qiagen, Hilden, Germany) was used to analyse E13.5, E15.5 and P0 Inha WT and KO RNAseq data. Upstream regulators and gene networks were generated using IPA using stringency criteria p-value ≤0.05 and a Z-score of (inhibition) -3 ≤ Z ≥ 3 (activated). The Z-score is a prediction scoring system of activation or inhibition based upon statistically significant patterns in the dataset and prior biological knowledge manually curated in the Ingenuity Knowledge Base (45).

Statistical analysis for counting and Real time-PCR data was performed using GraphPad Prism software (RRID : SCR_002798). Distribution of data was determined using the Shapiro-Wilk normality test. Statistical differences between WT and KO data were determined using a two-tailed unpaired t-test. Data are expressed as mean ± SD. Statistical significance was set at * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.

Since the absence of activin A results in decreased Sertoli cell and increased gonocyte numbers in newborn (P0) Inhba mouse testes (6), we studied the reciprocal condition of unopposed activin A in Inha WT and KO testes to learn how these cells were affected. Two surprising gonocyte phenotypes were immediately evident in E15.5 Inha KO testes: multinucleated gonocytes (MNGs) ( Figures 1A, B ) and gonocytes outside the cords ( Figures 1D, E ), each of which was only rarely or not observed in wildtype testis sections. The term ‘multinucleated’ refers here to the observation of ≥2 nuclei located within the cytoplasm of a single cell, with bi-nucleated (2 nuclei/cell) germ cells most frequently identified ( Figures 1A (IHC), B (IF)). We quantified the extent of this anomaly at E13.5, directly after sex determination, at E15.5, and at PND 0. Care was taken to discriminate between MNGs and two closely opposed germ cells (see Materials and Methods). Notably, although MNGs were observed in E15.5 KO testes, none were present in E13.5 or PND 0 Inha KO testes ( Figure 1C ), highlighting that a potential window of germ cell vulnerability to elevated activin A occurs before E15.5. The proportion of MNGs observed in E15.5 testes significantly increased from 0.5% in WT to ~ 10% in Inha KO testes (20-fold increase); by PND 0, the percentage of gonocytes scored as MNGs decreased to 3% which did not differ from WT (p = 0.0890) ( Figure 1C ).

Gonocytes normally become enclosed within Sertoli cells in cords by E13.5, thus their presence outside cords in Inha WT and KO testes at E15.5 is abnormal. None were observed outside cords at E13.5, suggesting that cord formation begins normally ( Figure 1F ), however ~8% of Inha WT and KO testis germ cells were in the interstitium by E15.5 ( Figure 1F ). At PND 0, no gonocytes were identified outside cords ( Figure 1F ), although the total gonocyte number was markedly reduced ( Figures 2U, V ; 3D). These data demonstrate how elevated activin A could alter embryonic germline fate and severely reduce the number of gonocytes present at the start of postnatal life.

Fetal testis cord development underlies establishment of the germline niche. We next assessed testis cord formation and relative Sertoli and germ cell numbers in fetal mice lacking inhibin (Inha KO), comparing these with testes lacking activin (Inhba KO) in which effects had been previously identified. Sertoli and germ cells were detected using SOX9 and DDX4, respectively. Testis cord borders were discerned based on SOX9 Sertoli cell nuclear staining at E13.5 and E15.5, and using SMA at PND 0 in Inha WT and KO (E13.5 - Figures 2A, B, E, F ; E15.5 - Figures 2I, J, M, N ; and PND 0 - Figures 2Q, R, U, V ) and Inhba WT and KO mouse testes (E13.5 - Figures 2C, D, G, H ; E15.5 - Figures 2K, L, O, P ; and PND 0 - Figures 2S, T, W, X ). The impact of activin deficiency or elevated bioactivity was most readily observed in PND 0 testis sections. At this age, germ cell numbers, cord size and the frequency of cord cross sections were markedly different in testes from both Inha KO and Inhba KO mice ( Figures 2V, X ) compared to WT ( Figures 2R, T ). Inha KO testes had fewer gonocytes, while in the reciprocal condition of reduced activin A, the substantial increase in gonocyte abundance inside Inhba KO cords correlated with previously published data (6).

These outcomes were quantified by measuring cord parameters relative to inferred activin A levels. In Inha KO testes at E13.5, the density of cords in each section was significantly lower compared to WT but did not differ at E15.5 and PND 0 ( Figure 3A ). The average cord area in Inha KO testes was significantly higher than in WT at E13.5, lower at E15.5, but comparable to WT at PND 0 ( Figure 3B ). The percentage of each testis cross section occupied by cords was significantly lower at E13.5 and E15.5 relative to WT, but not different at PND 0 ( Figure 3C ). These data suggest that elevated activin A bioactivity disrupts initial processes relating to fetal testis cord development, however by birth the overall cord dimensions are normal. The impact of reduced levels of activin A is evident at later time points compared to in Inha testes. In Inhba KO testes at E15.5 and PND 0, cord number was significantly lower ( Figure 3A ), the average cord area was higher ( Figure 3B ), and the percentage of testis area occupied by cords was significantly lower ( Figure 3C ). These data are consistent with previous reports showing the absence of activin A progressively leads to a smaller number of cords with an increased cord diameter (6, 12, 46).

We investigated whether these cord phenotypes reflect an impact of activin A on the relative density of germ and Sertoli cells. In Inha KO mice, the density of gonocytes per cord area is reduced at E15.5 and PND 0 ( Figure 3D ), while Sertoli cell density is greater in E13.5 and E15.5 testes ( Figure 3E ). The higher Sertoli cell density at the two earlier ages suggests that Sertoli cell population expansion is advanced during early fetal life in the Inha KO mice. Inhba KO testes had more gonocytes ( Figure 3D ) as previously (6), while in contrast, Sertoli cell density was not significantly different from in WT testes at any age; at PND 0 it was reduced, but not significantly (p = 0.1474) ( Figure 3E ). This latter finding differs from the significant dose-dependent decrease in Sertoli cell number at PND 0 reported previously (6); our previous application of the optical disector method, suited to counting irregular-shaped nuclei such as those of Sertoli cells, is likely to have led to this discrepancy.

Assessment of gonocyte/Sertoli cell ratios provided a clear measure of the physiological importance of activin A on the relationship between Sertoli and germ cells. Throughout development, Sertoli cells are always in close contact with germ cells, producing factors essential to their support and maturation, and each Sertoli cell supports a finite number of spermatogenic cells (47). The gonocyte/Sertoli cell ratio was significantly and progressively reduced in Inha KO testes at E15.5 and PND 0 and was significantly higher in Inhba KO testes at PND 0 ( Figure 3F ). These data provide evidence that activin A bioactivity is an important contributor to establishing the niche that determines the spermatogenic potential of the newborn testis.

To identify the transcriptional differences in the fetal testis arising from unopposed activin A, whole testis RNA sequencing of Inha WT and KO testes was performed. Principle component analysis ( Figure 4A ) illustrated age and genotype differences. Four hundred and sixty-six differentially expressed genes (DEGs) were identified between Inha KO and WT testes at E13.5, 85 at E15.5 and 411 at PND 0 ( Figure 4B ) using an FDR < 0.05 and absolute fold-change ≥ 1.5. A heat map illustrates transcript cohorts that are differentially affected by the absence of inhibin at each age ( Figure 4C ). Amongst the DEGs, only 12 were significantly altered at all three ages including Inha ( Figure 4D ). Transcripts from 7 genes were decreased (Inha, Rhpn1, Pla2g2f, Chdh, Megf10, Aldoc, Pla2g1b), and 5 were increased (Lect1, Gm15181, Trpv3, Ccl17, Pnmt).

To determine whether one particular cell type was affected in the Inha KO testes, the cell-type expression of DEGs was investigated by interrogating a single cell (sc)-RNAseq dataset obtained from E18.5, PND 2 and PND 7 mouse testes (48). Most DEGs were not cell-specific (present in more than one cell type, indicated as ‘not determined’ (ND), Figure 4E ). The remaining transcripts were predominantly expressed in a single cell type ( Figure 4E ): germ cells, Sertoli cells, Leydig cells, peritubular myoid cells, endothelial cells, stromal cells, and innate lymph cells or macrophages. The deduced cellular localisation for all DEGs are listed in Supplementary Table 1 . At E13.5, the majority of DEGs were present in Leydig cells (23.6%), followed by Sertoli cells (11.6%). By E15.5, this pattern was reversed (Sertoli (20%) and Leydig cells (9.4%), and by PND 0 there was a bias towards cell-specific DEGs being expressed in Leydig (4.6%), Sertoli (7.1%) and germ cells (4.1%) ( Figure 4E ).

The DAVID functional clustering tool was applied to the DEGs in Inha KO testes at each age to investigate the potential cellular consequences of excess activin A bioactivity. At E13.5, the DEGs were significantly enriched in functional terms relating to the metabolism and synthesis of sterols, steroids, and lipids ( Figure 5A ). At E15.5, the DEGs were associated with a single functional cluster annotated as secreted proteins and glycoproteins ( Figure 5A ). At PND 0, the top 8 most significant clusters were enriched for terms such as glycoproteins, secreted proteins, ion transport, extracellular matrix and cell junctions ( Figure 5A ).

To identify genes that could be directly responsive to activin A in embryonic testes, we compared all DEGs in the Inha KO whole testes (this study) with DEGs identified in the Inhba KO E15.5 somatic cells (19). In total, 61 transcripts were altered in both datasets ( Figure 5B ) and exhibited reciprocal changes in testes from mice with high and low activin A levels ( Figure 5B ). While these genes provide a signature of altered activin A bioactivity in the fetal testis, we propose that they are strong candidates for being direct targets of activin signalling.

Because altered levels of transcripts important for lipid and steroid metabolism are key outcomes of unopposed activin A in E13.5 Inha KO testes, Ingenuity Pathway Analysis was employed to evaluate the broader implications on this and other signalling and cellular activities. The top 15 cellular pathways affected in the E13.5 testes are identified ( Figure 6A ), and 11 of these (73%) are predicted to be inhibited (a z-score < -2). The top 4-most significantly affected pathways relate to cholesterol biosynthesis, while others relate to synthesis of zymosterol, glycocortioids and mineralcorticoids and the LXR/RXR, GPRS, CREB and S-100 family signalling pathways ( Figure 6A ). Network analysis ( Figure 6B ) indicates robust downregulation of genes involved in lipid and steroid metabolism, including Insl3, Cyp17a1, Cyp11b2, Cype1a2, amongst other transcripts expressed exclusively in fetal testis Leydig cells, including Crhr and Mc2r (49). These results demonstrate that emergence of steroidogenic Leydig cells, in either number and/or function, is disrupted in E13.5 Inha KO testes. Ingenuity Pathway Analysis identified candidate upstream regulators that influence the levels of DEGs in E13.5 Inha KO testes ( Figure 6C ). Based on z-scores, there was a strong bias towards inhibitory factors, with 7 activating and 34 inhibitory (90%) ( Figure 6C , Supplementary Table ; z-score of < -3 and > 3). Supplementary Table 2 contains the complete list of potential upstream regulators, their activation status and target genes.

These analyses indicate that the most affected cellular processes in E13.5 Inha KO testes relate to Leydig cell function ( Figure 4E ) and impact on steroid/lipid biosynthesis and metabolism ( Figures 5A , 6A ). Levels of key steroidogenic enzyme transcripts involved in the Δ4 classical steroidogenesis pathway, the predominant pathway in rodents, are presented in Figure 7 . At E13.5, transcripts produced in fetal Leydig cells encoding Star, Cyp11a1, Hsd3b1 and Cyp17a1 were robustly detected in WT testes but were barely detected in Inha KO ( Figure 7A ). In contrast, there were no differences in transcripts encoding steroidogenic enzymes Hsd17b1, Hsd17b3 and Srd5a1, which are exclusively expressed in fetal Sertoli cells at this age ( Figure 7A ). In Inha KO testes at E15.5, most of steroidogenic transcripts were unaffected except for Star (reduced) and Hsd17b1 (increased). At PND 0, steroidogenic enzyme transcripts produced in Leydig cells were not significantly different except for Hsd3b1 (decreased) whereas the activin A-responsive steroidogenic enzymes expressed in Sertoli cells, Hsd17b1 and Hsd17b3 (19, 48), were significantly increased.

To assess the outcomes of these transcriptional differences in Inha KO testes, several steroids were measured in E17.5 mouse testes using mass-spectrometry. Androstenedione (A4), testosterone and DHT were each significantly lower in Inha KO testes compared to their WT counterparts ( Figure 7B ), indicating that the overall level of steroidogenesis was reduced. The ratios of particular steroids highlighted key affected steroid biosynthesis steps ( Figure 7B ): the elevated T/A4 ratio in Inha KO testes was consistent with increased Sertoli cell expression of testosterone synthesising enzymes Hsd17b1 and 3 ( Figure 7A ), while the lower DHT/T ratio indicates 5alpha-reduction of testosterone is diminished. These results provide evidence that elevated activin A transiently impacts on Leydig cell gene transcription ( Figure 4E ), reducing the production of androstenedione (A4) ( Figure 7C ) at E13.5, whereas at later time points, the capacity for Sertoli cells to convert A4 into testosterone (T) is elevated.

We previously identified that levels of the 11-keto steroids 11-KA4 and 11-KT are increased when activin A are decreased (in Inhba KO testes) at E17.5 (19), suggesting their synthesis was responsive to activin A bioactivity. During the conversion of A4 to the 11-ketoandrogens, CYP11B1 converts A4 to the first 11-keto derivative, 11-OHK4 ( Figure 7C ). 11-OHK4 is then converted to 11-KA4 by HSD11B2, and 11-KA4 is converted to 11-KT by AKR1C6/AKR1CL ( Figure 7C ). Importantly, Cyp11b1 was barely detected in Inha KO testes at all ages but Hsd11b2 levels were not altered ( Figure 7D ), and Hsd17b2 was not detected (data not shown). We were unable to confidently measure 11-KA4 in E17.5 testes however 11-KT, measurable in WT testes, was undetectable in Inha KO testes ( Figure 7E ) indicative of reduced 11-keto androgen production in testes with unopposed activin. Overall, these data suggest that elevated activin A levels in early fetal life have a major impact on testicular steroidogenesis, affecting first fetal Leydig cells, then Sertoli cells ( Figure 7A ). The overall outcome is that levels of both testosterone and DHT, androgens crucial to testicular development and fetal masculinisation are reduced.

The absence of activin A (in Inhba KO testes) has been reported to predominantly affect somatic cells, in particular the Sertoli cells (12, 19), whilst MEHP predominantly impacts Leydig cell functions in rodents (27, 50). Nevertheless, due to the common germ cell phenotypes of increased MNG numbers and gonocytes present in the interstitium at E15.5 following MEHP exposure (27) and reported here for Inha KO testes, we hypothesised that activin and MEHP regulate common pathways and genes. E15.5 mouse testes fragments were cultured for 72 hrs with activin A (25 ng/ml), MEHP (200 µm) or both factors. Immunofluorescence images showed that culture with activin A and MEHP, individually or combined, resulted in increased frequency of MNGs within cords and gonocytes outside cords ( Figure 8A ). Quantification showed neither activin A or MEHP alone altered the total density of gonocytes, however there was a modest but statistically significant increase in cultures with both factors present ( Figure 8B ). The proportion of gonocytes outside cords was increased by activin A and robustly enhanced by MEHP ( Figure 8C ). Interestingly, testis fragments exposed to both activin A and MEHP did not have a higher proportion of gonocytes outside cords ( Figure 8C ). Fragments cultured with activin A or MEHP, or both combined, displayed a striking increase in the proportion of multinucleated gonocytes (MNGs) ( Figure 8D ).

To test for relationships in transcriptional regulation by activin and MEHP, transcript levels measured in whole fetal testes with different activin bioactivity levels were compared with those in E15.5 testis fragments cultured for three days with MEHP. Known activin-responsive transcripts (in Inhba testes; (19)) that were also identified as DEGs in whole Inha KO vs WT testes (this study), and selected germ cell markers, including markers of spermatogonial stem cells (SSCs), were examined.

Amongst the activin-responsive somatic cell transcripts, several that are expressed in fetal and immature Sertoli cells were identified as significantly altered in all three groups ( Figure 9 ). Amh, encoding Anti-Mullerian Hormone, was significantly decreased at multiple ages in both Inha and Inhba mouse testes and in MEHP cultures, while in contrast, levels of Ccl17, encoding the chemokine CCL17, were positively regulated by activin A (i.e. increased in Inha KO and decreased in Inhba KO testes). Ccl17 was also significantly increased by MEHP ( Figure 9 ). Gja1 was significantly lower in E13.5 Inha KO testes, E15.5 Inhba KO testes and in MEHP-treated samples ( Figure 10 ). Also known as Cx43, Gja1 encodes an activin A-regulated gap junction protein, GJA1, crucial for postnatal blood-testis barrier formation and important for Sertoli cell maturation that is affected by in vivo DEHP exposure (51). In fetal mouse testes lacking activin A (Inhba KO), the Sertoli cell-expressed enzymes Hsd17b1 and Hsd17b3, which convert A4 to T, are reduced while in the contrasting condition of unopposed activin A (Inha KO testes), only Hsd17b1 was higher at E15.5. Hsd17b1 was not altered by MEHP exposure, however Hsd17b3 was significantly reduced. Sel1l3, which encodes a protein of unknown function and poorly characterised distribution, and Serpina5, encoding a serine proteinase inhibitor produced in Sertoli and other cell types, were both increased in testes with elevated activin A (Inha KO), decreased in the absence of activin A (Inhba KO), and reduced following MEHP treatment. The Csf1 transcript, encoding cytokine colony stimulating factor 1 (CSF1), is more broadly expressed, being produced in interstitial cells including the vasculature, macrophages, Leydig and peritubular myoid cells (reviewed in (52, 53)). Csf1 was increased significantly in MEHP-treated testis fragments but increased only modestly at E13.5 in Inha KO testes ( Figure 9 ). Overall, these findings indicate that the dynamic range of activin A signalling differs between target genes and that many activin-regulated transcripts of functional importance to fetal testis development that are synthesized in Sertoli cells can be altered by exposure to MEHP.

To account for the reduction in gonocytes, measures of germ cell transcripts were normalised to the germ cell marker, Ddx4. Amongst the cohort of SSC-specific transcripts examined, regulation by activin was limited to the PND 0 Inha KO testis, with Bcl6b and Rara significantly elevated and Sohlh1 significantly decreased ( Figure 9 ). Exposure to MEHP elevated Bcl6b, T, and Ngn3, each of which normally increase normally between E15.5 and PND 0, which suggests MEHP actions can advance the trajectory of germ cell development.