Antibodies against Cyp19a1, CX37/GJA4, GDF-9, LHR, and FSHR were obtained from Abcam (Cambridge, MA, United States). Antibodies against ERK, p-ERK, P38, p-P38, Cleaved-caspase 3, and Cyp19a1were purchased from Cell Signaling Technology (Waltham, MA, United States). DMEM/F12 medium, fetal bovine serum (FBS), penicillin, and streptomycin were purchased from Gibco (Grand Island, NY, United States).

Recombinant mouse irisin (r-irisin) was prepared following our previously published study (Qiao et al., 2016; Ma et al., 2018). In brief, the full-length coding sequence of Fndc5 was amplified from mouse cDNA and subcloned into a pTasy vector. Then, the irisin sequence was amplified from the plasmid using primers containing NdeI and XhoI site sequences. Then, the recombinant irisin plasmid was constructed by ligation with pET28a (+) vector at the NdeI and XhoI sites. After sequencing, to verify the correctness of the cloned irisin cDNA, the plasmids were transferred to E. coli strain Rosetta (DE3). The expression of r-irisin in E. coli was induced, purified by a nickel column, and diluted with PBS (pH7.4).Then, the concentration of the protein was determined. Overall, 30 nM r-irisin upregulated the phosphorylated p38 (p-p38) and phosphorylated ERK1/2 (p-ERK1/2) in 3 T3-L1 cells within 20 min, according to the methods described by R&D company (Qiao et al., 2016).

In total, 21-day-old female C57BL/6 mice purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (China) were housed with four to five animals per cage under a 12-h light/dark cycle and in controlled temperature with free access to water and a diet. After 7 days of acclimatization, female mice were randomly divided into three groups: In the vehicle group, animals were injected daily with corn oil subcutaneously along with a control diet. In the PCOS group, animals were subcutaneously injected daily with DHEA (D4000; Sigma–Aldrich; 6 mg per 100 g, dissolved in corn oil) subcutaneously and were fed with a high-fat diet (HFD). In the PCOS +irisin group, animals were injected daily with DHEA subcutaneously, with 400 μg/kg r-irisin intraperitoneally every other day, and were fed with HFD. Here, the control diet comprised 12% fat, 73% carbohydrates (0% sucrose), and 19% proteins while the HFD comprised 60% fat, 26% carbohydrates (17% sucrose), and 25% protein. The mice were all treated with the above methods for 25 days before being sacrificed for further experiments. During treatment, vaginal smears were taken from the 13th day at the start of the intervention to observe estrous cycles. This project has ethical approval granted by the Medical Ethics Committee of West China Second University Hospital, Sichuan University (approval No. WCSUH21-2019-089).

After the onset of intervention, vaginal smears were taken daily from the 13th day for 10 days to observe the estrous cycle at 9:00 a.m. In brief, 20 μL of 0.9% sterile saline was injected into the vagina using a pipette and gently swabbed repeatedly 3–4 times and then aspirated the saline mixed with the vaginal secretion from the mouse vagina and smeared it evenly on a slide. Then, after air drying at room temperature, the slices were stained with Wright’s solution, and the estrous cycle was determined by microscopic analysis. To evaluate the estrous cycle, the following can be observed: the proestrus is made up of nucleated epithelial cells; the estrus is defined by the predominance of cornified squamous epithelial cells; the metestrus is marked by the appearance of white blood cells and a decrease in cornified squamous epithelial cells relative to the estrus; the diestrus is defined by the predominance of white blood cells.

Mice were anesthetized with an intraoral injection of chloral hydrate solution in a dose of 350 mg/kg body weight. Then, the blood was collected from the orbital venous plexus, and the serum was collected by centrifuging the whole blood at 4°C in 1200 g for 15 min. The serum levels of testosterone (T), follicle stimulating hormone (FSH), and luteinizing hormone (LH) were measured by radioimmunoassay (RIA) (Xin Fan Biotechnology, Shanghai, China).

The mice were fasted for 12 h before the IPGTT experiment and were given oral glucose of 50% concentration by gavage (in dose of 2.0 mg/kg body weight). Then, blood glucose (BG) levels were measured in the tail tip blood by 0 min (defined as fasting blood glucose or FBG), 15 min, 30 min, 60 min, and 120 min after gavage using a portable blood glucose meter (Accu-Chek Performa, Roche Diagnostics). For ITT, the mice were fasted for 6 h before 80 IU/kg body weight insulin (Wanbang Biopharmaceuticals, Shanghai, China) was intraperitoneally injected into each mouse, and BG at each time point was measured via the same method as IPGTT.

After the mice were killed by cervical dislocation, the ovaries of each mouse were dissected and fixed in 4% paraformaldehyde for 24 h. According to the standard histological procedures, the maximum section (5 μm thick) of ovarian tissue in each sample was stained with hematoxylin and eosin. The number of different stages of follicles (primordial, primary, antral, and cystic follicles) was counted. To evaluate the effects of irisin on the expression of follicular development and atresia-related proteins, immunohistochemistry was conducted using the above paraffin sections. After routine dewaxing and hydration process, we incubated the tissue sections with 3% H₂O₂ solution for 10 min to remove endogenous catalase and then blocked the sections with 5% goat serum in 0.01 M PBS. After blocking, tissue was incubated with the primary antibody, and we used horseradish peroxidase (HRP)-labeled secondary antibody for further incubation, and diaminobenzidine (DAB) reagent was used for colorization. Hematoxylin was used for counterstaining the nuclei. Finally, gradient alcohol-xylene was used for tissue dehydration and transparency before the sections were mounted with neutral resin. The results of H&E staining and IHC were observed under an ordinary light microscope (Olympus CX33). The primary antibodies used here were as follows: anti-cleaved-caspase3 (1:400, 9664, CST), anti-GDF-9(1:2000, ab254323, abcam), anti-CX37/GJA4 (0.5–1 μg/mL, ab254323, abcam), and anti-P450 aromatase (Cyp19a1) antibody (1:200, ab18995, abcam).

In total, 400 mg of fecal pellets were collected from each mouse in cryogenic vials immediately after being discharged before sacrifice. The fecal samples were snap-frozen with liquid nitrogen and were stored at −80°C before analysis. Fecal DNA extraction and 16S rRNA gene sequencing were carried out by Biomarker Technologies Co., Ltd. (Beijing, China) using the Pacbio sequencing platform. In briefly, fecal microbial DNA was extracted from mouse fecal samples using the TGuide S96 Magnetic Soil/Stool DNA Kit (Tiangen Biotech Co., Ltd., Beijing, China), according to the manufacturer’s instructions. The quantity and purity of the extracted DNA were controlled using NanoDrop 2000 (Thermo Scientific, United States). The integrity and size of the isolated DNA were evaluated using agarose gel electrophoresis. The full length of bacteria 16 s RNA gene was PCR-amplified using barcoded primers 27F_(16S-F) 5’-AGRGTTTGATYNTGGCTCAG-3′ and 1492R_(16S-R) 5’-TASGGHTACCTTGTTASGACTT-3′. Then, the PCR products were purified, quantified, and homogenized for constructing the sequencing library. The library repair kit (SMRTbell Template Prep Kit provided by PacBioperform) was used for damage repair, end repair, and adaptor attachment on a PCR instrument. Then, AMpure PB magnetic beads were used to recycle the final library, and a Sequel II sequencer was used for sequencing. After full length 16S rRNA sequences were obtained, the raw sequences were merged and quality filtered, and duplicates were removed by FLASH, Trimmomatic, and UCHIME, respectively. The resulting sequences were then aligned against the Greengenes database of 16S rRNA gene sequences. Reads were then measured by quantitative insights into microbial ecology (QIIME) analysis. The analyses of gut microbiota including alpha diversity, beta diversity, and difference significance analysis among groups were conducted using PD whole-tree, principal co-ordinate analysis (PCoA), and ANOVA, respectively.

In total, 50 mg of feces for each sample were used for metabolite extraction and were subjected to liquid chromatography-mass spectrometry (LC–MS), which was also carried out on Waters Xevo G2-XS QT of high-resolution mass spectrometer with Waters Acquity UPLC HSS T3 column (1.8 μm 2.1 × 100 mm) by Biomarker Technologies Co., Ltd. (Beijing, China), following the ordinary experimental protocol. Ultra performance liquid chromatography (UPLC) software (Waters ACQUITY UPLC HSS T3 C18, United States) was used to quantify the metabolite concentrations. The LC–MS data were analyzed by a supervised orthogonal partial least-squares discriminant analysis (OPLS-DA) clustering method. Variable Importance in Projection (VIP) derived from the OPLS-DA analysis was used to screen a variable, and a criterion of VIP value>1, p ≤ 0.01, and FC>1 was marked as a discriminating metabolite in this study. Biological pathway analysis was performed based on LC–MS data using MetaboAnalyst 4.0. The impact value threshold calculated for pathway identification was set at 0.1. Other instruments used in this study are shown in Supplementary Table S1.

All analyses were performed using SPSS (IBM SPSS Statistics 25). For unordered categorical variables and ordinal categorical variables, we used the χ² test and non-parametric test, respectively. For continuous variables, mean ± SD (standard deviation) was used, and for others, we used median ± quartile range. For the comparison among the three groups, a one-way analysis of variance (ANOVA) was conducted, followed by Tukey’s test. Pearson correlation analyses were used for normal distribution, or Spearman correlation analyses were conducted. The significance level was set at p < 0.05. Immunohistochemistry and immunohistochemistry Western Blot data were collected and analyzed by Image J.

PCOS mice gained more weight than control mice, and the body weights of r-irisin-treated PCOS mice were not significantly different from PCOS mice (Figure 1A). To investigate the effect of irisin on glucose metabolism in PCOS mice, we performed the IPGTT and ITT trials. In the IPGTT test, we found that PCOS mice showed a significant increase in blood glucose (BG) after intraperitoneally given with glucose solution at each time point compared with control mice, but irisin treatment did not effectively improve the situation (Figure 1B). In the ITT test, we neither found that the BG of PCOS mice was significantly altered compared with the control mice after receiving insulin treatment, nor found a significant difference in the BG level in irisin-treated PCOS mice compared with PCOS mice. Only after 120 min of injecting insulin, higher BG in the PCOS + irisin group mice compared with mice in the PCOS group was found (Figure 1C). These results indicated that irisin did not improve IR and glucose metabolism in PCOS mice.

Through observation of the estrous cycle of mice, we found that PCOS mice showed estrous cycle derangement, which showed near stagnation in the proestrus phase, whereas in control phase, each phase of the estrous cycle appeared more regularly. When treated with r-irisin, the appearance of each phase was partly restored in PCOS mice (Figure 2A). Similar to the phenotype of human PCOS, the ovary index (calculated by ovary weight/body weight of the same individual mouse) and uterus index (calculated by uterus weight/body weight of the same individual mouse) in PCOS mice increased significantly compared with control mice, while r-irisin treatment reduced the ovary index in PCOS mice, though the uterus index was not changed obviously (Figures 2B,C). Through H&E staining of ovary tissue sections, we found only a significant decrease in primordial follicles in PCOS mice ovaries. Although no statistical differences were presented, PCOS mice had an increased percentage of large cystic-like follicles compared with controls. In contrast, irisin treatment groups showed a reduction in cystic-like follicle number compared with PCOS mice. To be specific, the large cyst-like follicles which have very thinly and loosely arranged granulosa cells inside (commonly 2–3 layers) without oocyte can be regarded as specially atresia follicles and are considered as characteristic ovarian morphological changes in some PCOS mice models (Ryu et al., 2019) (Figures 2D,E).

In addition, by measuring serum steroid hormones using radioimmunoassay, we observed that PCOS mice had higher levels of testosterone (T), and a significant increase was also observed in LH/FSH ratio than control mice. However, there was no significant change in the level of LH and FSH between PCOS and the control groups. After being treated with r-irisin, the level of T went down, but the LH and FSH levels were unexpectedly increased, accompanied by a slight decrease in LH/FSH ratio without statistical differences when compared with PCOS mice (Figures 2F–I). These results indicated that exogenous irisin supplementation ameliorated the reproductive endocrine-related pathological phenotypes in PCOS mice.

To further explore the mechanism by which r-irisin altered the ovarian pathological phenotype of PCOS mice, we used immunohistochemistry to examine the expression of some proteins involved in follicle development or atresia. The results demonstrated that the expression of cleaved-caspase-3, the marker for follicle atresia, in the ovarian granulosa cells was significantly increased in PCOS mice compared with the control mice; however, the expression of GDF-9 in oocyte, a positive regulator of follicle development, and the expression of the bridge protein CX37/GJA4 in granulosa cells, which mediates the connection and communication between oocyte and granulosa cells, were significantly reduced in PCOS ovaries compared with the control. After r-irisin treatment, expression of cleaved-caspase-3 in PCOS ovary granulosa cells was decreased, and the expression of GDF-9 in oocytes and CX37/GJA 4 in granulosa cells was restored (Figures 3A–C). There was also a significant increase of cyp19a1 (aromatase) expression in r-irisin-treated mice ovaries (Figure 3D). These results suggested that r-irisin improving the ovarian pathological phenotype in PCOS mice may be related to its participation in the regulation of follicle development and atresia.

By performing 16S RNA sequencing of mice feces, the gut microbiome data were acquired. Through principal component analysis (PCA), we found that there was no significant difference in alpha diversity in the gut microbiome (PD whole tree Index) (Figure 4A), whereas there were significant differences in beta diversity among the three groups (vehicle, PCOS, and PCOS + irisin groups). From the perspective of the beta diversity, principal coordinate analysis (PCoA) based on the unweighted UniFrac distance revealed that samples from three groups of mice formed distinct clusters significantly far apart (Figure 4B). We then analyzed and compared the composition of the gut microbiota of the three groups of mice. From the species relative abundance bar chart, we found that, at the phylum level, Firmicutes, Bacteroidetes, Epsilonbacteraeota, and Proteobacteria were in largest proportions in the three groups of mice; at the level of the class, Clostridia, Bacteroidia, Bacilli, and Erysipelotrichia were in largest proportions in the three groups; at the level of order, Clostridiales, Bacteroidales, Lactobacillales, and Erysipelotrichales accounted for the largest proportion in the three groups; at the level of family, Lactobacillaceae, Erysipelotrichaceae, and Muribaculaceae occupied the largest proportion in the three groups; at the genus level, Lactobacillus, Blautia, and Fecalibaculum accounted for the largest proportion in the three groups; at the level of species, Fecalibaculum_rodentium, uncultured_bacterium_f_Muribaculaceae, Lactobacillus_reuteri, and Blautia_coccoides accounted for the largest proportion in the three groups. We can also observe that the gut microbiota of the three groups of mice did not differ significantly in species composition characteristics, but the proportion of different species varied in the classification level of each species and became the basis for the significant differences in microbial beta diversity (Figures 4C–H).

Using the Wilcoxon analysis, the heat maps of differential abundance between the vehicle versus PCOS and PCOS versus PCOS + irisin group were constructed, respectively, and further showed that DHEA and HFD exposure significantly changed the abundance of bacteria at the different species classification levels, while r-irisin treatment also changed the abundance of multiple strains in the gut flora of PCOS mice at the multispecies level (Figures 5A,B). As these differences are more common at the genus level, we extracted the strains with abundance differences at this level for ANOVA and found that r-irisin reversed the decrease of the abundance of Odoribacter and the increase of Eisenbergiella and Dubosiellad in PCOS mice caused by DHEA and HFD exposure at this level (Figure 5C). Furthermore, through LEfSe analysis, using LDA score > 2.0 as the screening criterion, we also found that the three groups of mice had their remarkable signature intestinal bacteria at the genus level with statistical differences (Figure 5D). Furthermore, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was performed using the predicted gene abundance to understand the functional contribution of gut microbial ecology. The results showed that r-irisin treatment might involve in the regulation of genes related to bile acid secretion, plant secondary metabolite biosynthesis, hormone stimulation, hormone biosynthesis, and ovarian steroid biosynthesis pathways (Figures 5E,F).

The fecal untargeted metabolomics was carried out to further understand the alterations of the gut chemical environment in PCOS mice and PCOS mice treated with r-irisin. Principal component analysis (PCA) of the metabolite profiling from the three groups of mice is shown in Figure 6A. By differential analysis of metabolites according to the differential generation criteria in methods, 1,579 differential metabolites were obtained from the PCOS group and the vehicle group (control), of which 855 increased and 724 decreased (see volcano plot of Figure 6B). We then plotted the top 30 and the top 9 differential metabolites with the largest -log10 p-value in abundance heatmap and boxplots, respectively. Among these 22 significantly increased metabolites, there were 2-(Phenylethenyl)-1,3-dioxolane, Androsta-1,4,6-triene-3,17-dione, and methallenestril, and among the eight significantly decreased metabolites, there were PG(i-19:0/LTE4), 3,4-dihydroxyphenylvaleric acid, and 8-iso-PGA1 (Figures 6C,D). Furthermore, the correlation analysis between differential metabolites and gut microbiota showed that the relative abundance of metabolites AB.Chminaca, Lysosulfatide, and X7.Keto.dehydroepiandrosterone were positively correlated with the relative abundance of the genus Eisenbergiella and Family _ XIII _ AD3011 _ group (Figure 6E). Meanwhile, the pathway enrichment analysis of differential metabolites showed that several key metabolic pathways including α-linolenic acid metabolism, sulfur metabolism, and β-alanine metabolism were affected by the exposure to DHEA + HFD (Figure 6F).

Through non-targeted metabolomics studies, we found 667 differential metabolites in the feces between PCOS and PCOS + irisin mice, including 378 increased metabolites and 299 decreased metabolites (Figure 7A). Similarly, we extracted the top 30 and top 9 differential metabolites with the smallest p-value and plotted into the abundance heat map and boxplot, respectively, and found that after the r-irisin treatment, methallenestril and PS (22:5 (4Z, 7Z, 10Z, 13Z, 16Z) / LTE4) in the fecal metabolites were significantly lower compared with the PCOS group, suggesting that r-irisin can reverse the levels of these two metabolites toward healthy control mice (Figures 7B,C). Moreover, the correlation analysis showed that the two different metabolites were consistent with the changes in the genus level Eisenbergiella and Dubosiellad (positive correlation, Figure 7D). Additionally, pathway enrichment analysis of differential metabolites illustrated r-irisin-altered pathways including sphingolipid metabolism, pyrimidine metabolism, and sulfur metabolism (Figure 7E).