The animal study was reviewed and approved by The Committee on the Use of Live Animals for Teaching and Research, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences. All animal experimental methods were performed in accordance with the approved guidelines and regulations.

C57BL/6J wild-type female mice were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. and acclimated to housing conditions for 1 week at a temperature of 22.0 ± 1°C, humidity of 40–60% and a 12-h light/dark cycle with fresh food and water available ad libitum.

At postnatal day 25, mice of comparable body weight were randomly divided into four treatment group (CTRL, DHT, DHT + NC-LV, DHT + Col6a5-LV. n = 6 per subgroup). The mice from DHT, DHT + NC-LV, DHT + Col6a5-LV groups were implanted subcutaneously (sc) with 1.7-cm silicone capsules (size: 1.57 mm inner diameter (ID) × 3.18 mm outer diameter (OD), Dow Corning, United States) containing 5 mg of DHT powder. Control mice (CTRL) received empty capsules (van Houten et al., 2012). Col6a5-LV is a recombinant Col6a5-CRISPR/Cas9 lentivirus, and NC-LV is a control lentivirus. The lentiviruses were delivered by hydrodynamic tail vein injection from the first day of DHT treatment (106 lentivirus particles/mouse/per week) (Chen and Calvisi, 2014).

Mice were sacrificed at the end of the 60-days treatment period. To determine the stage of the estrous cycle, daily vaginal smears were taken and examined 2 weeks before the animals were killed. The body weight of each mouse was measured every 10 days. In addition, at the end of the 60-days treatment period, blood samples were collected by orbital puncture after the mice were anesthetized with isoflurane, and tissues (ovaries, gonadal fat pads) were isolated and fixed overnight in 4% paraformaldehyde (PFA) or soaked in RNAiso Plus (−80°C cryopreservation).

For the IPGTT, female mice were fasted for 16 h with free access to drinking water and intraperitoneally (i.p.) injected with D-glucose (2.0 g/kg body weight). Blood glucose levels were measured before and 15, 30, 60, 90, and 120 min after i.p. glucose injection by using an Accu-Chek glucose monitor (Roche Diagnostics Australia Pty. Ltd., Australia) (n = 6 per group).

Total RNA from tissues and cells was extracted using RNAiso Plus and subjected to qPCR analysis. RNA samples were reverse transcribed into cDNA according to the manufacturer’s instructions (Toyobo, Osaka, Japan). The PCR mixtures contained 10 μl of SYBR Premix Ex Taq II (Toyobo), 1 μl of each primer, 1 μl of cDNA, and 7 μl of DNase-free water to a final volume of 20 μl. The cycle conditions were 1 min at 95°C, followed by 45 cycles at 95°C for 15 s, 65°C for 30 s, and 72°C for 1 min. The primers were designed on the basis of the published sequences. The RNA levels were calculated by the 2⁻ΔΔ^CT method, where CT is the cycle threshold (Livak and Schmittgen, 2001; Table 1 contains the primer sequences).

Ovaries and gonadal fat pads from the Ctrl and DHT treatment groups were sent to Genminix Informatics Ltd., Shanghai, China, for gene expression analysis via the Affymetrix GeneChip^® Mouse Transcriptome Assay 1.0 array (n = 3 per group). The signatures and abundances of RNAs can be found in National Center for Biotechnology Information, Gene Expression Omnibus (NCBI, GEO) under series number GSE171431.

Differential expression between the comparison groups was determined using online tools from the Gene-Cloud of Biotechnology Information (GCBI)¹, based on the SAM method for at least three samples per group. GO and annotation analysis of disordered genes was performed using Metascape², based on Fisher’s exact test, which was represented in a 2 × 2 contingency table. mRNA transcripts were analyzed after using GCBI to filter out genes with detection rates below 50% in the overall expression profile. The mRNAs between the comparison groups exhibited expression fold-changes ≥ 2 and p-values ≤ 0.01, which were considered significant (Huang et al., 2018). Gene set enrichment analysis (GSEA) was conducted on the desktop version of the GSEA software (v.4.1.0).

Ovaries and gonadal fat pads were dehydrated with a graded alcohol series and embedded in paraffin. Serial 5-μm sections were stained with hematoxylin and counterstained with eosin. The cell size was measured with Image-Pro Plus 6.0 software. Briefly, the perimeter of the stroma area was manually drawn (Supplementary Figure 2). Then, the area was measured, and the number of cells in the area was counted. The ratio of the cell area to the cell number was calculated. Three to five fields of view were measured on each ovary to cover whole ovarian stromal cells within each section (n = 6 per group). The diameters of adipocytes were evaluated by automatically counting bright objects and measuring the mean diameter. Three to five fields of view were measured on each gonadal fat pad to cover adipocytes of different sizes within each section (n = 6 per group).

Immunohistochemistry (IHC) and Immunofluorescence (IF) were carried out on 5-μm paraffin-embedded ovaries sections. After deparaffifinization and rehydration, 3% H₂O₂ was used to block endogenous peroxidase activity for 30 min at room temperature. For non-specific binding inhibition, the sections were blocked with 1% BSA for 1 h and incubated overnight at 4°C with anti-Col6a5 antibody (diluted 1:100, ThermoFisher, PA570781) and anti-PLIN2 antibody (diluted 1:100, Abcam, ab108323). After washing, the sections were incubated with HRP-Goat-anti-Rabbit antibody (diluted 1:200, Abcam, ab97051) or Alexa-488-Donkey anti- rabbit (diluted 1:200, Invitrogen, A21206) for 1 h. IHC staining was visualized using the 3,3’-Diaminobenzidine (DAB) Substrate Kit for Peroxidase (Beyotime, P0203), and the slides were counterstained with hematoxylin. For IF staining, sections were viewed under an Olympus microscope after the nuclei were stained with DAPI (diluted 1:1,000, sigma, D9542). Control sections were immunostained with non-specific IgG to check for non-specific staining.

The TG and TC contents in mice ovarian stromal cells and 3T3-L1 cells were analyzed. The cells were washed with PBS, harvested by trypsinization and then resuspended in 1 ml PBS (∼10⁶ cells). The cell suspension was homogenized by sonication for 5 min. The TG and TC contents were determined using a commercial assay kit according to the manufacturer’s protocol (GPO-PAP, Nanjing Jiancheng Bioengineering Institute, Nanjing, China). The protein concentration was determined using the Pierce^TM BCA Protein Assay Kit.

We used 21-day-old immature female mice for this study. The collection and purification of ovarian stromal cells were performed as described previously (Magoffin and Erickson, 1988; Duleba et al., 1997). Briefly, the ovaries were removed from the animals and dissected from the oviducts and fat under a dissecting microscope. After 60 min of collagenase digestion, stromal cells were purified using discontinuous Percoll gradient centrifugation. The stromal cells were incubated at a density of 10⁴ cells/ml. The cultures were carried out for 48 h at 37°C in an atmosphere of 5% CO₂ humidified air in serum-free McCoy’s 5A culture medium supplemented with 1% antibiotic/antimycotic mix and 10% fetal bovine serum (FBS). All cultures were carried out in the presence of ovine luteinizing hormone (LH) (5 ng/ml).

For transient transfection, stromal cells and NCTC1469 cells were grown to 80% confluence with complete medium and transfected with Col6a5-siRNA (sense 5′–3′ CCAAUUACCUGGGAAUUAATT, antisense 5′–3′ UUAAUUCCCAGGUAAUUGGTT) or negative control siRNA (NC-RNA, sense 5′–3′ UUCUCCGAACGUGUCACGUTT, antisense 5′–3′ ACGUGACACGUUCGGAGAATT) using Lipofectamine 2000 reagent (Invitrogen). At 4–6 h post-transfection, the cells were washed and incubated in conditioned medium (complete medium containing DHT or inhibitors) for 24 h. The extent of Col6a5 knockdown with siRNA was shown in Supplementary Figure 6.

The Col6a5 sgRNA CRISPR/Cas9 All-in-One Lentivector with the target sequence TGTTTGCCCCAAACATGACA (K4947208, Applied Biological Materials (ABM) Inc., Vancouver, Canada) was cotransfected with the packaging plasmids pHelper1.0 and pHelper2.0 into 293T cells to prepare the recombinant lentiviral vector Col6a5-LV. Scrambled sgRNA CRISPR/Cas9 All-in-One Lentivector was used as the control lentivirus [NC-LV including the target sequence GCACTCACATCGCTACATCA (K010, ABM)]. These vectors were concentrated to 10⁷ lentivirus particles/ml. The concentration of lentivirus particles was measured with a commercial ELISA kit (Beijing Biodragon lmmunotechnologies Co., Ltd., Beijing, China).

3T3-L1 preadipocytes [American Type Culture Collection (ATCC)] were allowed to grow to 60% confluence in 6-well plates, and then the cells were infected with lentivirus (1 ml Col6a5-LV plus 1 ml complete medium). Twelve hours post-infection, the medium was replaced with fresh complete medium. Three to four days later, the infected cells were cultured with 10 μg/ml puromycin for 1 week for selection. Control cell lines were generated in the same way as described above with the control lentivirus (NC-LV). The extent of Col6a5 knockdown with lentivirus was shown in Supplementary Figure 6.

3T3-L1 preadipocytes were allowed to grow to 80% confluence in complete medium (high glucose Dulbecco’s modified Eagle’s medium (Gibco), 10% FBS and 1% antibiotic/antimycotic mix). Then, the 3T3-L1 preadipocytes were seeded in 6-well plates and stimulated with 0.5 mM 1-isobutyl-3-methylxanthine (IBMX), 1 mM dexamethasone, and 10 μg/ml insulin for 2 days. DMEM was then replenished with the addition of 10 μg/ml insulin for 2 days, and then the medium was replaced with DMEM without insulin for 2 days (Li et al., 2016).

NCTC1469 cells were routinely cultured in growth medium (DMEM with 10% FBS and 1% penicillin-streptomycin solution) in a humid incubator at 37°C and 5% CO₂. Fat-overloading induction of the cells was performed when the cells grew to 80% confluence; the growth medium was removed, and the cells were divided into two groups: the control group and the model group. The control group was cultured in growth medium, and the model group was cultured in high-FBS medium (DMEM with 50% FBS and 1% penicillin-streptomycin solution) for 24 h to induce steatosis (Liu and Huang, 2016).

For steroid treatment experiments, DHT (in 100% EtOH) at the appropriate concentration (0.1, 1, 10, and 100 μM) or EtOH vehicle was added to the culture medium (to a final EtOH concentration of 0.00001%) (Mereness et al., 2015). Stromal cells and NCTC1469 cells were treated with DHT for 24 h. 3T3-L1 cells were treated for 6 days.

To explore which stages of lipid accumulation in cells were affected by DHT, the cells were treated with stage-specific inhibitors, simvastatin (an inhibitor of cholesterol synthesis), ezetimibe (an inhibitor of cholesterol uptake), and 4-phenylbutyric acid (4-PBA, an inhibitor of endoplasmic reticulum stress). Inhibitors (in 100% EtOH) at the appropriate concentration or EtOH vehicle was added to the culture medium (to a final EtOH concentration < 0.001%). The doses were as follows: simvastatin (0.005, 0.05, and 0.5 μM), ezetimibe (0.3, 3, and 30 μM) and 4-PBA (0.01, 0.1, and 1 μM) (Yu et al., 2006; Zhang et al., 2012; Tung et al., 2015). Stromal cells and NCTC1469 cells were treated with inhibitor for 24 h, and 3T3-L1 cells were treated for 6 days.

Intracellular lipid accumulation was determined by Oil Red O staining. After removal of the culture media, the cells were washed twice with PBS, fixed with 4% PFA, and stained with Oil Red O (six parts 0.6% Oil Red O dye in isopropanol and four parts water) for 30 min. After rinsing three times with distilled water, the cells were photographed under a microscope.

To quantify lipid accumulation, lipids and Oil Red O were dissolved in isopropanol, and absorbance was measured by a microplate spectrophotometer at 500 nm. The percentage of Oil Red O-stained material relative to that in control wells was calculated as the absorbance at 500 nm (sample)/the absorbance at 500 nm (control) (Li et al., 2016).

The serum of mice treated with or without DHT were collected, and DHT levels were measured with a mouse DHT ELISA kit (Jiangsu Enzyme Biotechnology, MB-3267A). The intra- and interassay error rates among all assays were less than 10 and 15%, respectively.

Results are expressed as mean ± SEM. Duncan’s multiple range test for individual comparison repeated measures ANOVA was used to analyzed body weight and serum glucose. Mann-Whitney U-test (a non-parametric statistical approach) was used to compare the mean DHT concentration, ovarian stromal cells areas, adipocyte diameters, mRNA levels, lipid concentration, TG concentration, and TC concentration between both groups and Fisher’s exact test (a non-parametric statistical approach) was used to compare the rates of irregular cycles between both groups in SPSS 21.0 package for Windows. P-values < 0.05 were considered significant.

To demonstrate the effects of DHT, parameters representing reproduction and metabolism were measured. The DHT levels in DHT-treated mice were significantly increased compared with those in placebo-treated mice (Figure 1A). In the late stage of the treatment, DHT-treated mice had higher body weights than placebo-treated mice (Figure 1B). The blood glucose levels of DHT-treated mice were higher than those of placebo-treated mice at 30 min after i.p. glucose injection (Figure 1C). DHT-treated mice had fewer regular cycles of 3–5 days than placebo-treated mice (Supplementary Figure 1A), resulting from prolong and stagnant estrus cycle, suggesting that the DHT-treated mice were acyclic (Figure 1D).

The ovaries of DHT-treated mice showed no corpora lutea but some cyst-like antral follicles (Figure 1E and Supplementary Figure 1B). The stromal area and the diameters of the adipocytes was larger in DHT-treated mice than in placebo-treated mice (Figure 1F and Supplementary Figures 3, 4). These results suggest that DHT may induce the accumulation of lipid droplets in the stroma of ovaries and adipocytes of gonadal fat pads.

The enriched gene ontology (GO) terms comprised 114 bioprocesses in ovaries. The top 10 enriched terms sorted in descending order based on −log(P)-values referred to the regulation of lipid metabolic processes (GO: 0019216) and lipid catabolic processes (GO: 0016042) (Figure 2A, left). The top 10 enriched terms sorted in descending order based on the numbers of changed genes also included lipid metabolic processes (GO: 0006629) (Figure 2A, right). Gene enrichment analysis also identified a significant decrease in metabolic processes including fatty-acyl-CoA metabolic process, long-chain-fatty-acyl- CoA metabolic process, neutral-lipid metabolic process, very-long-chain-fatty-acid metabolic process, and membrane-lipid catabolic process (Supplementary Figure 5).

The enriched GO terms contained 99 bioprocesses in gonadal fat pads. The top 10 enriched terms sorted based on − log(P)-values referred to the regulation of monocarboxylic acid metabolic processes (GO: 0032787), fatty acid metabolic processes (GO: 0006631), the positive regulation of fatty acid biosynthetic processes (GO: 0045723), fatty acid biosynthetic processes (GO: 0006633), the regulation of fatty acid biosynthetic processes (GO: 0042304), the positive regulation of fatty acid metabolic processes (GO: 0045923), and lipid biosynthetic processes (GO: 0008610) (Figure 2B, left). It is worth emphasizing that when based on the number of changed genes, the lipid metabolic process was the top enriched term (GO: 0006629) (Figure 2B, right). Gene enrichment analysis also identified a significant decrease in fatty-acyl-CoA metabolic process, very-long-chain-fatty-acid metabolic process, long-chain-fatty-acyl-CoA metabolic process, neutral-lipid catabolic process, and increase in lipid import into cell (Supplementary Figure 5).

Genes associated with lipid metabolic processes that were altered in DHT-treated mouse ovaries were selected and analyzed (Figure 2C). Compared with placebo-treated mice, DHT-treated mice showed decreased expression of Adora1, Ephx2, Idh1, Scd1, StAR, Akr1c18, Cyp11a1, Hao2, Gm2a, Pik3c2g, Ptgfr, and Mgst2 and increased expression of Fabp3, Kit, Oca2 and Sult1e1.

Altered genes associated with lipid metabolism processes in the gonadal fat pads of DHT-treated mice were selected and analyzed (Figure 2D). Compared with placebo-treated mice, DHT-treated mice showed decreased expression of Elovl3, Kit, Mlxipl, Akr1c14, and Pnpla3 and increased expression of Cyp1a1, Hpgd, Lep, Anxa1, Abcd2, Fads3, Lpgat1, and Mgst3.

Gene expression was analyzed via the transcriptome array. According to the stricter selection criteria [fold-change ≥ 2, p ≤ 0.05 and false discovery rate (FDR) ≤ 10%], 66 mRNAs in the ovaries (Figure 2E, red circle) and 115 mRNAs in the gonadal fat pads (Figure 2E, green circle) were dysregulated in DHT-treated mice compared with placebo-treated mice. When comparing the dysregulated mRNAs in the ovaries and gonadal fat pads, we found that there were only six overlapping genes: Col6a5, Ptgr, Ptgr1, Kit, Sult1e1, and C7 (Figure 2E).

Our verification showed that Col6a5 was highly expressed in both the ovaries and gonadal fat pads of DHT-treated mice compared with placebo-treated mice (Figure 3A), which was consistent with previous mRNA array results. Additionally, IHC staining showed that Col6a5 was expressed mainly in ovarian stromal cells rather than theca cells and granulosa cells (Figure 3B and Supplementary Figure 7). From the above, we suggest that Col6a5 may be an important factor during the effect of DHT on ovarian stromal cells and adipocytes.

To test our hypothesis, we used Col6a5-CRISPR/Cas9 recombinant lentivirus to knockdown Col6a5 expression in mice by tail vein injection. In the mice treated with DHT and injected with Col6a5-LV, the Col6a5 mRNA levels were significantly decreased in both the ovaries and gonadal fat pads (Figure 3D). Compared with the NC-LV group, Col6a5-LV ameliorated the hypertrophy of ovarian stromal cells and adipocytes in DHT-treated mice (Figure 3E and Supplementary Figures 3, 4). The hypertrophy of ovarian stromal cells in DHT-treated mice was associated with excess lipid accumulation, as indicated by the expression of the lipid droplet-specific marker adipose differentiation-related protein adipophilin/perilipin-2 (PLIN2). From the immunochemistry results for PLIN2, we found that DHT treatment increased the accumulation of lipid droplets in ovarian stroma (Figure 3C) and this effect could be alleviated by the inhibition of Col6a5 expression (Figure 3F). In addition, PLIN2 was also expressed in granulosa cells of antral follicles, especially after DHT treatment, this effect could also be alleviated by the inhibition of Col6a5 expression, which is consistent with the results in stromal cells (Figures 3C,F).

Oil Red O staining was used to identify and quantify oil droplets. DHT treatment significantly increased the formation of oil droplets in ovarian stromal cells at 1 μM. However, Col6a5-siRNA treatment relieved the effect of DHT and dramatically reduced lipid accumulation in the cells (Figure 4A).

The intracellular TG and TC levels were also quantified. DHT significantly increased the TC content in stromal cells compared with that in those not treated with DHT. However, both the TG and TC levels were significantly decreased in response to Col6a5-siRNA transfection of DHT-treated stromal cells (Figure 4B).

In addition, the TC content of stromal cells was much higher than the TG content, suggesting that TC may be the main component of the lipids that accumulate in response to DHT. Consistent with this, various concentrations of simvastatin (an inhibitor of cholesterol synthesis) and ezetimibe (an inhibitor of cholesterol uptake) were found to reduce the intracellular lipid content of DHT-treated stromal cells. Endoplasmic reticulum stress can induce cholesterol metabolism disorder, and 4-PBA (an inhibitor of endoplasmic reticulum stress) also reduced the lipid content of DHT-treated cells (Figure 4C).

We further investigated whether DHT affects the expression of lipid metabolism-related genes in stromal cells. Mttp promotes lipid transfer to the extracellular space; Cd36 promotes lipid transfer to the intracellular space; Add1 and Chop promote fat synthesis and lipid accumulation; StAR is a limiting enzyme of cholesterol mobilization; P450scc induces progesterone biogenesis; 3β-HSD and 17β-HSD promote testosterone biogenesis; Sf1 promotes StAR expression; Adx1 inhibits Sf1 action; and Gata6 promotes P450scc and 3β-HSD expression.

We found that DHT led to higher Cd36 and Chop mRNA levels than placebo treatment, while Col6a5-siRNA significantly suppressed the increase in Chop (Figure 4D). However, DHT significantly induced the expression of StAR, 3β-HSD, and their transcription factor Adx1. Compared with the control siRNA (NC-siRNA), Col6a5-siRNA significantly suppressed the change in the expression of these genes in DHT-treated cells. In addition, the expression levels of P450scc were decreased, and the expression of 17β-HSD was increased by Col6a5-siRNA in the DHT-treated cells, although DHT treatment alone did not alter the levels of these factors (Figure 4E).

DHT treatment at 1 μM significantly increased the accumulation of oil droplets in adipogenic 3T3-L1 cells. However, this effect was dramatically alleviated after transfection with Col6a5-LV (Figure 5A). The TC content was decreased after treatment with DHT, while there was no significant difference in 3T3-L1 cells transfected with Col6a5-LV (Figure 5B).

While the level of TC was higher than that of TG in stromal cells, the TC content was similar to the TG content in adipogenic 3T3-L1 cells. Moreover, simvastatin and ezetimibe could not suppress the excess lipid accumulation induced by DHT, suggesting that TC may not be the main component of the lipids that accumulate in response to DHT. However, 4-PBA reduced the intracellular lipid content in DHT-treated adipogenic 3T3-L1 cells (Figure 5C). This implies that DHT may dysregulate lipid metabolism in adipocytes by inducing endoplasmic reticulum stress.

DHT also influenced the expression of lipid metabolism-related genes in adipogenic 3T3-L1 cells. As shown in Figure 5D, DHT significantly induced the expression of Mttp, Cd36, Add1, and Chop, and the increases in the expression of Mttp and Chop were significantly attenuated after transfection with Col6a5-LV (Figure 5D).

The effect of DHT on lipid accumulation was cell type-specific, since there was no increase in lipid droplets in the DHT-treated NCTC1469 cells at a dose similar to that used to treat stromal cells and 3T3-L1 cells (Figure 6A). As an overnourished cell model, NCTC1469 cells exhibited an increase in lipid droplets when cultured in 50% FBS (v/v) medium, and Col6a5-siRNA relieved this effect (Figure 6B). Inhibition of Col6a5 expression commonly affected intracellular lipid accumulation (Figure 6C). However, 50% FBS treatment did not significantly alter the expression of Mttp, Cd36 and Add1, the expression of Twist2, a transcription factor of Add1, was even decreased, while Col6a5 siRNA could significantly increase the expression of Mttp, Cd36 and Add1 (Figure 6D).