Studies were conducted on C57BL/6JRj mice aged 7 to 27 days post-natal (dpn) and on adult females that were born at the animal facility from genitors purchased at Janvier Labs (Le Genest St Isle, France) (n=180 mice in total). The day of birth was designed as 0 dpn. Mice were maintained under controlled conditions (12h light/dark cycle) with food (Scientific Animal Food and Engineering (SAFE), A03-10) and water available ad libitum. Ten µg of GnRH antagonist (Ganirelix, Orgalutran^®, N.V. Organon, Puteaux, France) or saline were subcutaneously injected twice on prepubertal female mice at 12 and 13 dpn, as previously described (4). In addition to Ganirelix, 13 dpn mice received an intraperitoneal injection of 5 U.I. human chorionic gonadotropin (hCG) (N.V.Organon), and were killed at 14 dpn. Injections were performed with Hamilton syringes connected with catheters and needles (Phymep, Paris, France). For another study, mice were subcutaneously injected with an androgen receptor antagonist, flutamide (5 mg/kg; F9397, Sigma), dissolved in corn oil (Sigma) every 10 hours starting at 12 dpn until the day of dissection (14 dpn) (total: 4 injections). Adult females were killed on the day of diestrus 1 or 2 determined by vaginal smears, as described (4). Mice were anesthetized with a mix of ketamine (Imalgene^® 1000) and xylazine (Rompun^® 2%) to collect the blood by cardiac puncture. After cervical dislocation, ovaries were either frozen in liquid nitrogen and stored at -80°C for RNA extraction, fixed in 4% paraformaldehyde (PFA) for immunofluorescence studies or immediately used for organotypic cultures. Blood was allowed to clot at room temperature for at least 15 minutes, and then centrifuged at 4°C, 5000 g for 5 minutes to obtain serum. Experiments were performed in accordance with standard ethics guidelines and were approved by Institutional Animal care and Use committee of the University Paris Cité and by the French Ministry of Agriculture (agreement #04015.01).

Single frozen ovaries were lysed with TissueLyser II (Qiagen, Courtaboeuf, France) in RLT buffer from RNeasy mini kit (#74106, Qiagen) containing 1% β-mercaptoethanol, following manufacturer’s protocol. Total RNA extraction was performed on columns and eluted with 30 µl of sterile water. Total RNA (120 to 250 ng) was reverse-transcribed with Superscript II (Invitrogen, Cergy Pontoise, France) and Random Primers (Invitrogen) following manufacturer’s instructions. Primers used for quantitative real-time PCR are listed in Table 1 . Ppid (Cyclophilin D) was used as a reference gene to normalize the data. Real-time PCR was performed with Lightcycler 480 SYBR Green I Master and LightCycler instrument (Roche Molecular Biochemicals, La Rochelle, France), following the MIQE guideline [https://pubmed.ncbi.nlm.nih.gov/19246619/]. Experiments were performed with at least four different ovarian samples/age or treatment group, with each sample run in triplicate as previously described (26).

Ovaries from 13 dpn female mice were placed on cell culture inserts (Millicell #PICM01250, Millipore, Guyancourt, France) on the top of 400 µl of RPMI culture medium without red phenol containing fetuin (Sigma, #F2379), insulin (Sigma, #I0516), transferrin (Sigma, #T8158), sodium selenite (Sigma, #S5261) and BSA (Euromedex, Souffelweyersheim, France) in 24-well plates for 1 hour. Then, the culture medium was replaced with 400 µl of fresh medium containing either or in combination: recombinant FSH (500 ng/ml, Gonal-F, Merck Serono), hydroxyflutamide (10 µM; H4166, Sigma), forskolin (10 µM, F39147, Sigma). After 8 hours of treatment, ovaries were snap-frozen in liquid nitrogen and stored at -80°C for RNA extraction.

The mass spectrometry coupled with gas chromatography (GC-MS) procedure to determine testosterone content in the serum and in the ovary was performed as described in previous papers (4, 26). Briefly, frozen mouse ovaries were introduced in a Lysing Matrix D tube (MP Biomedicals, Illkirch-Graffenstaden, France) with 0.5 ml of cold saline water and 10 μl deuterated methanol internal standard working solution. The sample was homogenized with an FP120-HY-230 Ribolyser (Hybaid, Teddington, Middlesex, UK) for three times of 20 seconds at maximum speed. After centrifugation, the liquid phase with the ceramic spheres was collected in a clean glass tube. The tissue residuals in Lysing Matrix tube were recovered with 3 ml of 1-chlorobutane (HPLC grade). After fast centrifugation, the upper organic phase was collected on conditioned Hypersep SI 500 mg SPE minicolumn (Thermo Scientific, Rockwood, USA). The column and adsorbed material were then washed with ethyl acetate/hexane (6 ml; 1/9, v/v). The second fraction containing testosterone was eluted using ethyl acetate/hexane (4 ml; 1/1, v/v), then evaporated at 60°C to dryness. Testosterone was derivatized with pentafluorobenzoyl chloride (PFBC) (103772-1G, Sigma-Aldrich, Steinheim, Germany). Final extracts were reconstituted in isooctane, then transferred into conical vials for injection into the GC system. A quadrupole mass spectrometer equipped with a chemical ionization source (NCI) and operating in single ion monitoring mode (SIM), was used for detection (HP5973, Agilent Technologies, Massy, France). The linearity of steroid measurement was confirmed by plotting the ratio of the steroid peak response/internal standard peak response to the concentration of steroid for each calibration standard. Lower limit of quantification was 1.15 pg. The intra- and inter-assay variability of the low limit of quantification CVs was 13.3 – 20%.

FSH and LH were simultaneously assayed in 10 µl of serum simplicate of the same serum samples using the Luminex technology with the mouse pituitary magnetic bead panel Milliplex MAP kit (Merck-Milipore, Nottingham, UK) in accordance with the manufacturer’s instructions. Samples were run on a Bioplex-200 instrument (Bio-Rad, Marne-La-Coquette, France) and concentrations were calculated using a-five parameters logistic fit curve (5PL) generated from the standards by the Bio-Plex Manager 6.1 software (Bio-Rad, Marne-La-Coquette, France). The sensitivity of the assays was 32 pg/ml of FSH and 3.2 pg/ml for LH. The inter-assay coefficient of variation was 5.4% for FSH and 3.2% for LH. The intra-assay coefficient of variation was 7.6% for FSH and 5.9% for LH.

For histological analyses, at least five ovaries of 14 dpn mice treated with either flutamide or its vehicle were fixed for 1-2 hours in Bouin’s solution, rinsed in PBS and placed in ethanol 70%. The ovaries were then dehydrated using increasing concentrations of ethanol (70%-95%-100%) and paraffin-embedded using standard protocols. Sections of 5-µm thickness were mounted on glass slides at the frequency of one every three sections, and they were stained with hematoxylin/eosin (HE), using routine procedures. After dehydration in alcohol and mounting in Eukitt (Sigma), slides were scanned using Axio Scan Z1 Zeiss at Pasteur Institute (Histology platform, Johan Bedel). Follicle classification was the same as that described in (14), with some modifications. Briefly, primordial follicles were constituted by a resting oocyte surrounded by a single layer of flattened granulosa cells. Primary follicles contained a growing oocyte with a single layer of cuboidal or mixed flattened/cuboidal granulosa cells. Preantral follicles contained an oocyte surrounded by at least two layers of granulosa cells, and no antrum or spaces between cells. Follicles were classified as early antral when they contained a growing oocyte enclosed by at least two layers of granulosa cells with either scattered areas of follicular fluid or a single antral cavity. We also monitored follicular atresia in preantral and early antral follicles, and searched for oocyte fragmentation, disordered granulosa layers, pycnotic granulosa cells (at least 3/follicle) and hypertrophied theca layer. Follicular counting was performed on every three tissue sections. To avoid repeated counting of the same preantral/antral follicle on several tissue sections, when a follicle was seen for the first time, it was marked and tracked in each subsequent section throughout which it appeared. Only follicles with visible oocytes were counted. Follicle sizes were determined by the measurement of their areas on HE-stained ovarian tissue sections using Orbit image analysis software. Follicles with oocytes exhibiting a visible nucleus were manually outlined and cross-sectional area measurement was calculated by Orbit software, from 6 representative slides for each ovary.

Mouse control ovaries were rapidly collected and fixed for 1 hour in 4% paraformaldehyde (PFA), rinsed in PBS and placed in sucrose 18% and then in a drop of tissue-Tek O.C.T. compound (Sakura) for preparing frozen sections, as described previously (14). Sections of 7-μm thickness were mounted on glass slides. For each antibody tested, all ovarian sections were stained in one run to compare the immunofluorescence staining among the different studied groups. Frozen sections were rehydrated in PBS and incubated for one hour with 10% (wt/v) normal goat serum in PBS. Following removal of the goat serum, sections were incubated overnight at 4°C in a humid chamber with anti-rabbit polyclonal primary antibodies diluted in PBS-BSA 0.5% ( Table 2 ). They were rinsed three times in PBS at RT, and then incubated for one hour at RT in goat anti-rabbit IgG (H+L) secondary antibodies (ref# A21069, lot #2146040, Invitrogen, dilution 1/1000 in PBS-BSA 0.5%). Control sections were incubated with isotype IgG (Vector Laboratories), instead of the respective primary antibodies, according to the manufacturer’s instructions. Ovarian sections were then rinsed in PBS and mounted in an anti-fading medium containing DAPI (Sigma) for observation with a Nikon Eclipse 90i. The images were processed with ImageJ software to establish the % of Ki-67 positive granulosa cells per follicle (from 8 preantral/antral follicles in 4 control and 4 flutamide-treated mice). This was obtained by normalizing the number of Ki-67 positive granulosa cells by the total number of granulosa (DAPI+) cells. Granulosa cell counting was performed using the “Cell Counter” plugin of the software.

For each experiment, the number of samples per age/treatment group is indicated in the legend of the figure. Data were analyzed using Prism 6 (version 6.0, GraphPad Software). Depending on experimental setting, statistical analyses were performed by Student t-test (one parameter and two groups), one-way ANOVA (one parameter and more than two groups) when data showed normal distribution as evaluated by the Shapiro-Wilk test, or non-parametric Mann-Withney (two groups) or Kruskal-Wallis test (more than two groups). For analyses of at least two parameters, two-way ANOVA was used. Data are shown as means ± SEM. A P value < 0.05 was considered as significant.

We specified prepubertal levels of testosterone in ovaries and serum, by selecting females at different periods before puberty: 7 dpn (neonatal), 12 and 14 dpn (infantile), 17 and 21 dpn (early juvenile) and 28 dpn (late juvenile) ( Figure 1A ). We used the GC-MS technology which is one of the most reliable methods to measure sex steroids. The results showed that intra-ovarian testosterone is produced as early as the neonatal period, reaching approximately 1 to 9 pg/ovary between 7 dpn and 21 dpn. It was lower at these developmental stages than in juvenile and adult females, reaching about 30 pg/ovary and 62 pg/ovary, respectively ( Figure 1B ). Testosterone was detected in the serum as early as 7 dpn (about 22 pg/ml), and it increased to about 48 pg/ml at 14 dpn ( Figure 1C ). Circulating testosterone levels further increased in late juvenile and adult females, reaching about 125-150 pg/ml ( Figure 1C ).

We then studied the expression of AR and by using RT-qPCR, we detected Ar transcripts throughout the prepubertal period ( Figure 1D ). We observed that Ar expression progressively increased from 7 to 21 dpn ( Figure 1D ). To specify AR protein expression pattern in the mouse ovary at mini-puberty, we performed immunofluorescence studies with a specific anti-AR antibody. We found AR expression in the nucleus of granulosa cells of follicles from the primary stage onwards and in thecal cells, but not in oocytes ( Figure 1E ). Variable levels of AR staining were observed in granulosa cells, suggesting different androgen receptivity between cells. No AR staining was observed in the population of primordial follicles ( Figure 1E ).

LH levels are elevated during mini-puberty in the mouse (1, 2, 4), but whether they promote androgen synthesis through their action on their biosynthesis pathways (schematized in Figure 2A ) is not known. We, thus, sought whether LH could already drive testosterone synthesis and AR expression at this stage. We used a mouse model with pharmacological reduction of gonadotropin levels by the GnRH receptor antagonist Ganirelix, at a stage when LH levels surge between 12 and 14 dpn (4, 26). We supplemented these mice with the exogenous gonadotropin hCG to replace LH ( Figure 2B ). Ganirelix injection leads to a reduction by ~80% and ~95% of circulating FSH and LH levels, respectively (4). It induced a ~85% decrease in intra-ovarian testosterone contents in comparison with control mice injected with saline ( Figure 2C ). The supplementation of Ganirelix-treated mice with hCG restored intra-ovarian testosterone abundance to that of control mice ( Figure 2C ). In line with these findings, the analyses of transcript levels of several components of the androgen biosynthesis pathway, i.e., Star (steroid acute regulatory protein), Cyp11a1, Cyp17a1, and Hsd3b1, showed that their relative intra-ovarian levels, except that of Hsd3b1, were downregulated in Ganirelix-treated females as compared with those in controls. The relative levels of Cyp11a1 and Cyp17a1 returned to those of control ovaries and those of Star became higher upon co-treatment with hCG, suggesting that mini-pubertal LH plays an important role in the synthesis of testosterone by regulating the levels of these transcripts ( Figures 2D–G ). The relative abundance of intra-ovarian Ar mRNAs showed a ~20% increase in Ganirelix-treated females as compared with that in controls, suggesting that it was repressed by gonadotropins. The co-treatment with hCG restored Ar transcript abundance to that of control females ( Figure 2H ). Overall, these data suggest that LH regulates androgen signaling during mini-puberty.

We next investigated whether androgens could regulate follicular growth in mini-pubertal ovaries, which contain many preantral and early antral follicles. Because androgens are produced by the ovary during this period, we adopted a pharmacological approach inhibiting their action with the AR antagonist flutamide. We subcutaneously administered flutamide twice daily from 12 to 14 dpn ( Figure 3A ), and we analyzed follicular growth at the end of the treatment by evaluating the number of follicles present at this stage belonging to each stage, i.e., primordial, primary, preantral and early antral follicles, by morphometric studies. Examination of tissue sections revealed no evident gross abnormalities in flutamide-treated mice when compared with controls ( Figure 3B ). The results showed that the number of primordial and primary follicles was not significantly affected by flutamide treatment ( Figure 3C ). However, it significantly increased the number of preantral follicles by about 20% compared to controls, and it had no effect on the number of early antral follicles ( Figure 3D ). Measurement of preantral and antral follicle areas revealed that the size of these two follicular populations did not differ between the two groups ( Figure 3E ), even when the analyses considered the 5 largest follicles of each category (data not shown). The morphological analysis of preantral and early antral follicles did not reveal the presence of atretic follicles (see the criteria in materials and methods).

To determine whether the treatment could affect granulosa cell proliferation in preantral and antral follicles, we performed in situ analyses of the proliferating cell marker Ki-67 in the ovaries of control and flutamide-treated females ( Figure 3F ). It was present in granulosa cells of virtually all growing follicles in both groups. However, quantification of the % of Ki-67 positive granulosa cells in preantral and antral follicles indicated that these follicles had a lower proliferation rate in flutamide-treated mice than in control mice (38.9% of Ki-67 positive granulosa cells in controls, versus 25.6% in flutamide-treated mice; P=0.0046) ( Figure 3G ). As androgens have also been shown to promote or suppress granulosa cell apoptosis depending on studies (27, 28), we sought whether flutamide treatment could have some effects on this process in mini-pubertal ovaries. Consistent with our morphological observations that there is no follicular atresia at this stage, we found no apoptotic cells in control ovaries, as revealed by the absence of cleaved caspase-3 immunodetection ( Supplementary Figure 1 ). We observed few apoptotic granulosa cells in preantral and antral follicles of flutamide-treated mice (1-2 apoptotic granulosa cells/ovarian section) ( Supplementary Figure 1 ), suggesting that the treatment has very minor effect on granulosa cell apoptosis. As preantral follicles could undergo atresia following an autophagy-related process independent of apoptosis (29), we also evaluated the expression of two autophagy markers, i.e. the protein p62 (the sequestosome 1 abbreviated as SQSTM1) and the autophagy-related gene 7 protein (ATG-7) by immunofluorescence on serial ovarian sections (30). In peripubertal ovaries used as controls, we observed p62/SQSTM1 expression in granulosa cells and oocytes of both preantral and antral follicles ( Supplementary Figure 2 ). There was a weak ATG-7 staining in oocytes and thecal cells of growing follicles ( Supplementary Figure 2 ). We observed that preantral follicles undergoing follicular atresia, as suggested by their weak AMH and Ki-67 expression, displayed increased expression of ATG-7 in granulosa cells and oocytes and either a decrease or an increase in P62/SQSTM1 expression in granulosa cells.

In mini-pubertal ovaries, p62/SQSTM1 was expressed in granulosa cells and oocytes of growing follicles. ATG-7 was essentially detected in oocytes of growing follicles and it was barely detectable in thecal or granulosa cells. AMH was present in growing follicles located at the periphery of the ovary, and weak or absent from the preantral/early antral stages located toward the center of the ovary, as previously shown in mice and rats at this stage (26, 31). We recently demonstrated that the high FSH levels at mini-puberty mediate the loss of AMH expression and promote the expression of Cyp19a1 in these follicles, therefore suggesting their FSH receptivity (26). Importantly, we found no difference in the expression of autophagy markers and AMH in preantral and early antral follicles between the control and flutamide-treated groups, despite the marked decrease in the % of Ki-67 positive cells in the flutamide-treated group, as compared with the control group ( Supplementary Figure 2 ).

We studied the molecular mechanism underlying AR actions on follicle growth in the mini-pubertal ovary by analyzing the relative expression levels of a factor inhibiting cell cycle progression, i.e., p27Kip1(Cdkn1b), which is an inhibitor of the cyclin dependent kinases (32). The relative abundance of Cdkn1b transcripts increased by about 20% in flutamide-treated females compared with controls ( Figure 3H ). We also analyzed the relative expression levels of factors promoting cell proliferation, including Cyclin D1 (Ccnd1) and Cyclin D2 (Ccnd2), which are required for cell progression through the G1 phase of the cell cycle (32, 33). The relative expression levels of Ccnd1 were not altered by flutamide treatment ( Figure 3I ), unlike those of Ccnd2 showing significant up-regulation ( Figure 3J ). In situ studies with a specific Cyclin D2 antibody showed that this protein was specifically expressed in granulosa cells of primary and preantral follicles located at the periphery of the ovary and not in preantral and antral follicles located in the center, in controls as in flutamide-treated females ( Supplementary Figure 3 ). Taken together, these data suggest that androgens could have some effects on follicular growth by modulating granulosa cell proliferation during mini-puberty.

We investigated whether AR signaling could contribute to follicular maturation, as reported in the adult ovary (16). We, thus, investigated by RT-qPCR the possible changes induced by flutamide treatment in the relative expression of key genes involved in steroidogenesis and folliculogenesis, i.e., Cyp19a1, Cyp17a1, Amh, Fshr and Ar. Interestingly, the relative abundance of both Fshr and Cyp19a1 mRNAs decreased by about 2-fold following flutamide treatment compared with controls ( Figures 4A, B ), while that of the other studied transcripts did not significantly change, including that of Amh, consistent with our in situ studies ( Figures 4C–E , Supplementary Figure 2 ). To determine whether the alteration in Cyp19a1 and Fshr mRNA levels resulted from a possible change in gonadotropin levels induced by flutamide, we measured LH and FSH serum levels. However, neither FSH nor LH circulating levels were impacted by flutamide treatment during mini-puberty ( Figures 4F, G ), suggesting that the observed alterations in ovarian gene expression rely on direct AR-mediated regulation in the mini-pubertal ovary.

We next investigated whether the AR-mediated regulation of Cyp19a1 expression observed in this study implies FSH signaling, and in particular the regulation of the abundance of Fshr, which is a known AR target in the ovary (17–19). We carried out organotypic cultures with mini-pubertal ovaries to determine the relative expression levels of Fshr and Cyp19a1 in response to treatment with hydroxyflutamide (HF) (used instead of flutamide because it is already active in blocking the AR pathway, unlike flutamide which needs to be metabolized in vivo) or with a high FSH concentration mimicking that of mini-puberty ( Figure 5A ) (4). In ovaries cultured for 8 hours under control conditions, Fshr mRNA abundance remained stable while that of Cyp19a1 mRNA dropped sharply compared to uncultured ovaries, indicating that FSH stimulation may be required for maintaining basal levels of Cyp19a1 mRNA ( Figures 5B, C ). The treatment by HF, either alone or combined with FSH, down-regulated Fshr expression by about 50% compared to control ovaries, showing that HF treatment efficiently counteracted the actions of androgens produced by ovarian explants ( Figure 5B ). In contrast, HF had no effect on the relative expression of Cyp19a1, which remained at the same low levels seen in untreated control ovaries ( Figure 5C ). However, it significantly suppressed the FSH-induced elevation in Cyp19a1 abundance, decreased by about 2.5-fold in the group of HF/FSH-treated ovaries compared to the FSH-treated group ( Figure 5C ).

It is generally accepted that FSH up-regulates the expression of the Cyp19a1 gene upon binding to its receptor leading to activation of several down-stream signaling pathways among which adenylate cyclase/PKA plays a major role (34) ( Figure 5D ). We then sought to determine whether the observed HF-induced down-regulation of Cyp19a1 in FSH-treated ovaries may result from a decreased activity of FSH signaling resulting from the decrease in Fshr transcripts, rather than a direct effect on Cyp19a1 gene expression. To this end, we used forskolin (FSK) to activate the adenylate cyclase, and under this condition we analyzed the effect of HF on the relative levels of Fshr and Cyp19a1 expression. In agreement with our previous results, HF treatment alone decreased by about 2-fold Fshr expression and had no effect on that of Cyp19a1 ( Figures 5E, F ). The treatment with FSK alone also decreased Fshr expression by about 1.5-fold compared with the control group, and it significantly increased that of Cyp19a1, as expected ( Figures 5E, F ). Importantly, addition of HF to FSK did not decrease the relative abundance of Fshr and Cyp19a1 transcripts, unlike addition of FSH ( Figures 5B–F ). Taken together, these findings support the idea that the androgens would regulate Cyp19a1 expression by acting up-stream FSH-mediated activation of adenylate cyclase.