This study was approved by the Ethics Committee of Obstetrics and Gynecology Hospital affiliated Fudan University. Informed written consent was obtained from all the participants. Eligible women who had undergone IVF were recruited from September 2020 to January 2021. PCOS was diagnosed based on the Rotterdam criteria with two of the following: oligo and/or anovulation, polycystic ovarian morphology, and clinical and/or biochemical signs of hyperandrogenism (in this study, hyperandrogenism was defined as T>51ng/dl、the menstrual cycle exceeds 45 days and the number of small follicles visible on both sides of the ovary under ultrasound is ≥12). The control group included women who seek treatment for tubal infertility or male factors, with normal ovarian reserve (regular menstrual cycles, and normal ovarian morphology) who has normal BMI, the menstrual cycle is 28-35 days and the number of small follicles in a unilateral ovary under ultrasound <10. Women with endometriosis, cancer, or other medical disorders that could affect folliculogenesis were excluded.

FF samples were collected from 3 PCOS patients and 3 controls of the participants for cytokine array analysis. Patients included met all the three items of Rotterdam criteria. The clinical, hormonal characteristics were compared between the two subgroups ( Table 1 ). FF samples from an additional 20 PCOS patients and 15 controls of the participants were collected for ELISA validation of the identified cytokines, and the clinical, hormonal characteristics were presented in Table 2 . All subjects underwent controlled ovarian stimulation using the standard IVF antagonist stimulation regimen protocol. FF was collected by transvaginal ultrasound-guided aspiration, 36 h after the administration of recombinant human chorionic gonadotropin. Only clear FF samples with no macroscopic blood contamination were included. After oocyte isolation, the FF samples were centrifuged at 800 g for 10 min to remove pellets. The supernatant was then separated and stored at -80°C for future use.

A cytokine and chemokine array (Proteome Profiler TM Human XL Cytokine Array Kit, R&D Systems, MN, USA) was used to detect the changes in 102 cytokines and chemokines in follicular fluid from PCOS and non-PCOS patients according to the manufacturer’s instructions. Briefly, samples were incubated on the membrane overnight at 4°C on a rocking shaker. The membranes were washed and incubated with a cocktail of biotinylated dectection antibody, and then the membrane was incubated with Strepatividin-HRP and chemiluminescent detection reagents. The chemiluminescent signal on each membrane was collected using an Amershan Imager 600 (GE Healthcare Life Sciences, Pittsburgh, PA, USA). The intensity (Pixel density) of each spot was quantified using HL Image++ (Western Vision Software, Salt Lake City, UT, USA), and corrected for background intensity and normalized to the membrane’s positive control.

The differential abundance of IL-15 was validated by ELISA using FF samples of 20 patients with PCOS and 15 control participants. Concentrations of IL-15 in the FF samples were measured using the commercial human IL-15 ELSIA kits (Multi Science, Hangzhou, China) according to the manufacturer’s instructions.

Three-week-old female C57BL/6 mice (Jiesjie laboratory animal co. LTD, Shanghai, China) were maintained in a 12h light/12 h dark cycle with free access to rodent feed and water. All procedures were carried out followed the guidelines provided by the Fudan University Institutional Animal Ethical Committee. Mice were divided randomly into two subgroups: control was subcutaneously implanted with a placebo and PCOS was subcutaneously implanted with 3mg letrozole (LTZ) pellet (Innovative Research of American, Sarasota, FL, USA) for 21 days. At the end of experiment, all mice were euthanized by intraperitoneal injection with pentobarbital sodium. Ovary was dissected out and fixed by paraformaldehyde to paraffin embedding. The animal study was approved by the Ethics Committee of Fudan University.

Viginal smears of all mice were collected and for determination of estrous cycle, Giemsa staining was used, and stages of estrous cycle were determined microscopically.

Mice were fasted for 12 h before the glucose tolerance tests (GTT) and 4h before the insulin tolerance tests (ITT). Glucose levels were measured by tail vein blood sampling using Accu-Chek Performa blood glucose analyzer (Roche Diagnostics). The mice were intraperitoneally injected with D-glucose (2g/kg body weight) for GTT or insulin (1 IU/kg body weight) for ITT after measurement of fasting glucose levels, and tail samples were collected at 15,30,60,90 and 120 min after the IP injection for glucose level detection. And the level of fasting insulin was measured by ELSIA (Multi Science, Hangzhou, China). And HOMA-IR (homeostasis model assessment of insulin resistance) index was calculated refer to (34).

Serum testosterone (T), dehydroepiandrosterone sulfate (DHEAS), luteinizing hormone (LH), follicle-stimulating hormone (FSH) concentrations were measured by corresponding ELISA kits (Sino-UK bio, Beijing, China). Moreover, serum and ovarian tissue homogenates IL-15 concentration was determined through a mice ELISA kit (Multi Science, Hangzhou, China).

Hematoxylin and eosin (H&E) staining was performed for ovary following deparaffinization and rehydration. 5μm sections were stained using hematoxylin followed by eosin staining and subjected to graded alcohol dehydration. The histopathological analysis of the ovary was evaluated by two independent viewers (Yan Liu, Zhi Li) who were blinded to the group information.

Immunohistochemistry (IHC) staining of IL-15 was used an immunohistochemical SP kit (Origene, Rockville, MD, USA) following the manufacturer’s instructions. Briefly, tissue sections were boiled in antigen retrieval buffer. After cooling, sections were incubated with 3% hydrogen peroxide solution at room temperature for 15 min followed with 10% goat serum as an antigen-blocking buffer for 30 min at 37°C. Sections were then incubated with Rabbit antibody against IL-15 (1:100, Affinity Bioscience, OH,USA) overnight at 4°C in a humid chamber. HRP-conjugated secondary antibody was applied for 30 min and visualized with 3’3-diaminobenzideine (DAB). Observations were made using a Nikon Eclipse 80i (Nikon, Tokyo, Japan) microscope. The positive staining areas were measured by Image J software (NIH, USA). Immunofluorescence staining was applied by a multiple fluorescent staining kit (Absin Bioscience, Shanghai China) according to the manufacturer’s instructions. The sections were incubated with primary antibody against IL-15 (1:100), and FSHR (1:500, Service Bio, Wuhan, China). Images were captured using a laser confocal microscope (Leica TCS SP8, Germany).

KGN (human ovarian granulosa cell tumor) cells were obtained from Shandong University and were cultured with DMEM/F12 medium (Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin-streptomycin in a humidified incubator at 37°C with 5% CO2. Mouse primary granulosa cells (mGCs) were isolated from ovaries of 3 weeks age female C57 mice. Ovaries were dip on ice in Lebovitz’s-15 medium (Sigma-Aldrich, USA) with 10% fetal bovine serum and 1% penicillin-streptomycin. A 25-gauge needle was used to separate surrounding adipose and envelope tissues then puncture the ovary to release the GCs. The cell suspension was centrifuged at 100 rpm for 5 min, resuspended with McCoy’s 5A medium (Sigma Aldrich, USA) supplemented with 5% fetal bovine serum and 1% penicillin-streptomycin. To further investigate the involvement of MAPK signaling pathways in IL-15–induced apoptosis and CYP17A1 abundance, the cells were preincubated with or without p38 MAPK inhibitor SB203580 (10 uM, Selleck Chemicals, USA) (35), and JNK inhibitor SP600125 (10 uM, Selleck Chemicals, USA) (36) for 30 minutes before IL-15 (500pg/ml) treatment.

Cell viability was determined by cell-counting kit-8 (CCK8) (Selleck, Shanghai, China). KGN cells and mouse primary GCs were seeded in 96-well plates at 1×103 cells/well in 100μl cell suspension. 10μl CCK-8 reagent was added to each well and then the cells were cultured for 1h at 37°C. Optical density was measured at 450 nm using a microplate reader (BioTeck, USA). Each experiment was carried out in triplicate at least.

The apoptosis level of GCs cells was detected using Annexin V-FITC apoptosis kit (BD Pharmingen, San Diego, CA, USA) according to the manufacturer’s instructions. Experiments were performed by a CytoFLEX flow cytometer (Beckman coulter, Brea, CA, USA) and analyzed by FlowJo 10.0 software (Tree Star, Ashland, USA).

Total RNA was extracted from ovarian tissues and GCs using RNA-Trizol reagent (Invitrogen, Carlsbad, CA, USA). The RNA concentration of all samples was quantified by NanoDrop 1000 (Thermo Fisher Scientific, USA), and was reverse transcribed to cDNA by Reverse Transciption kit (Takara, Dalian, China). Quantitative reverse transcription PCR (RT-PCR) was performed in SYBR Green^® fast qPCR Mix (Takara, Dalian, China) on an ABI 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA USA). Results were normalized using β-actin expression. Primers were designed using Primer-Blast tool by NCBI and present in Supplementary Table 1 . Relative transcription levels were calculated using 2 ^-ΔΔCT methods.

Total protein was extracted using ice-cold radio-immunoprecipitation assay lysis buffer (Cwbio) containing a phosphatase inhibitor and a protease inhibitor cocktail (both from Roche). Protein from each sample was electro-phoresed in a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel and then transferred onto a nitrocellulose blot. After 1 hour of blocking with 5% nonfat milk, the blot was incubated overnight at 4°C with antibodies against phospho-JNK (Thr183/Tyr185; 1:1000), and total JNK (1:1000), phospho-p38 MAPK (Thr180/Tyr182; 1:1000), total p38 MAPK (1:1000). On the second day, the blot was washed and then incubated with the respective secondary antibody conjugated to horseradish peroxidase (1:1500) for 1 hour. An enhanced chemiluminescent detection system (Millipore) was used to detect the bands with peroxidase activity. A G-Box iChemi Chemiluminescence image capture system (Syngene) was used to visualize the bands. The same blot was also probed with α-tubulin (1:1000) as internal controls. All antibodies were from Cell Signaling Technology (CST, Massachusetts, USA).

Microarray datasets 155489 and GSE106724 were downloaded from Gene Expression Omnibus and collected using the following platforms: GPL20795 (HiSeq X Ten) and GPL21096 (Agilent-062918 Human lncRNA array V4.0).The raw data was converted to a recognizable format by GEO2R (https://www.ncbi.nlm.nih.gov/geo/geo2r/).

Statistical analysis was performed with GraphPad Prism version 8.0 software (GraphPad, San Diego, CA, USA), and data were presented as the mean± the standard error of the mean (SEM)or medians with interquartile ranges. Student’ t test of two independent samples was used for the data of normal distribution, and the Wilcoxon rank-sum test of two independent samples was used for the data of non-normal distribution. Pearson’s linear regression was used for correlation analysis to investigate the relationship among IL-15 and serum Testosterone concentration. Spearman’s or pearson’s correlation analysis were used to investigate the relationship among the relative expression of IL-15 or IL-2rg(IL-15r)and CYP17A1 in PCOS patients. The one-way ANOVA with Tukey’s multiple comparison post-hoc test was used for multiple groups. The post hoc statistical power value is expressed as “R”. A P<0.05 was considered statistically significant.

In order to determine the alteration of cytokine and chemokine in the follicular fluid of PCOS women, we recruited 3 PCOS and 3 non-PCOS patients, whose clinical characteristics are illustrated in Table 1 . In the PCOS patients, the BMI and serum LH, T and AMH level, and the ratio of LH/FSH were significantly higher, while the age, serum FSH and E2 levels were similar between the two groups. According to the findings of cytokine array, we found that chemokine MPO, IL-1α, Kallikrein3, IL-15, MCP-1, and IL-8 were significantly increased in PCOS group ( Figure 1A and Supplementary Figure 1A ). As previously reported, IL-15 is a regulator of obesity related pathological changes (37, 38), and can modulate insulin sensitivity (39). Whereas, obesity and insulin resistance (IR) are typical pathological changes in PCOS, we speculated that IL-15 plays an important role in PCOS. Then, the similar results were observed in FF collected from 20 PCOS and 15 non-PCOS patients ( Figure 1C ), and their clinical characteristics are illustrated in Table 2 . Non-PCOS patients matched the BMI of PCOS patients, while the LH level of PCOS patients tends to be higher、the FSH level and E2 level tend to be lower and the T level is significantly higher compared with non-PCOS patients, which were also consistent with the clinical endocrine characteristics of PCOS. Furthermore, Pearson correlation analysis showed that the FF IL-15 concentration was positively correlated with serum T (r=0.2466, P = 0.0024, Figure 1D ). And relative expression of IL-15 was positively correlated with CYP17A1 in cumulus granulosa cells of PCOS patients and age-matched control (r=0.7904, P=0.0254, Figure 1E ) and relative expression of IL-2rg (IL-15r) was positively correlated with CYP17A1 in cumulus granulosa cells of PCOS patients and non-PCOS women (r=0.5849, P=0.0458, Figure 1F ) by using bioinformatics analysis.

In order to verify the role of IL-15 in the pathogenesis of PCOS, the well-established LTZ-induced PCOS mouse model was confirmed. The mice in the PCOS group were significantly heavier compared with age-matched control mice that had received placebo pellets ( Figure 2A ). Similar results were observed in the weight of ovary and fat pads in PCOS group ( Supplementary Figure 1C, D ). PCOS mice had significantly reduced glucose tolerance and insulin sensitivity ( Figures 2B, C ). And we tested the fasting blood glucose of the mice and calculated HOMA-IR. The results showed that the PCOS mice had obvious insulin resistance compared with the control group ( Figure 2D and Supplementary Figure 1E ). PCOS mice also displayed an irregular estrous cycle ( Supplementary Figure 1F ). The levels of serum T, DHEAS and LH were significantly higher in PCOS mice (P<0.05) compared with the control, but there is no significant difference in serum FSH levels ( Figure 2E and Supplementary Figure 1G ). Atresia and cystic dilated follicles were significantly increased and the number of corpora lutea was decreased in the PCOS mice compared with the control mice ( Figures 2F–H ), indicating similar ovarian dysfunction to the clinical presentation of PCOS.

Consistently, we found that IL-15 mRNA and protein expression were elevated significantly in ovarian of PCOS mice compared with control, while the serum IL-15 level was similar between two groups ( Figures 3A–C ). Moreover, the IL-15 mRNA expression was elevated significantly in adipose tissues of PCOS mice compared with control, suggesting that IL-15 may play a role in chronic inflammation of adipose ( Figure 3D ). IHC results confirmed that IL-15 was overexpressed in PCOS mice compared with control ( Figure 3E ), and colocalized with FSHR, a GCs marker ( Figure 3F ). This data suggested that IL-15 may be involved in the pathogenic role of GCs in PCOS.

Due to the significant association of IL-15 with the clinic pathological characteristics in PCOS women and the colocalization between IL-15 and GCs in PCOS mice, we examine the effects of IL-15 on GCs growth and apoptosis. Our data showed that IL-15 suppressed proliferation of KGN cells and the intervention of JNK inhibitor SP600125 and p38 MAPK inhibitor SB203580 alleviated this effect of IL-15 ( Figure 4A ). The results obtained by flow cytometry analysis are also consistent with CCK-8 assay ( Figure 4B ). These results suggested that IL-15 decreased proliferation, whereas increased apoptosis in GCs.

To delineate the role of IL-15 in GCs steroidogenesis and the state of inflammation, we cultured the KGN cells in the presence or absence of IL-15 with or without inhibitors. Treatment of KGN cells with IL-15 for 4 h resulted in significantly increases the mRNA levels of cytochrome P450, family 17, subfamily A, polypeptide 1 (CYP17A1), which converts progesterone to 17α-hydroxyprogesterone and androstenedione ( Figures 4C, D ). Interestingly, we found that these two inhibitors have inhibitory effects on the increase in CYP17A1 expression induced by IL-15. Consistent to previous studies that IL-15 function as a proinflammatory cytokine (40), IL-15 significantly increased mRNA level of IL-1b and Ifng and the above two inhibitors have a tendency to reverse this effect ( Figured 4E, F ). These data suggested that IL-15 might play a role in the pathogenesis of PCOS by increasing production of androgen hormones and sustaining inflammation state in GCs, and these effects may be related to the activation of the signaling pathways p38 MAPK and JNK. Therefore, we carried out the detection of these two signaling pathway molecules and their phosphorylation levels on different treated KGN cells by Western Blot. The results of WB showed that IL-15 treatment increased the phosphorylation level of JNK and P38 MAPK in the KGN cell line compared with DMSO alone ( Figure 4G ).

IL-15 suppressed proliferation of primary GCs cells at dose dependent manner ( Figures 5A, B ). In addition, IL-15 promoted apoptosis of primary mGCs at dose dependent manner ( Figure 5C ).

Consistent with previous study, IL-15 significantly increased mRNA levels of CYP17A1, while decreased mRNA levels of CYP19A1 and FSHR ( Figures 5D–G ). IL-15 also increased proinflammatory cytokine mRNA levels of Ifng and Tnfa ( Figures 5H, I ). The effects of IL-15 on the steroidogenic activity of primary GCs were evaluated. The level of DHEAS and progesterone in the medium secreted from primary GCs for 24 h in the presence of IL-15 were measured. The concentration of DHEAS were significantly increased after incubation of the cells with IL-15 ( Figure 5J ), while the concentration of progesterone remained unaltered ( Figure 5K ). These data confirmed the results obtained from the human GCs cell line KGN.