Mesenchymal stem cell therapy ameliorates metabolic dysfunction and restores fertility in a PCOS mouse model through interleukin-10

Background Polycystic ovary syndrome (PCOS) is the most common endocrine and metabolic disorder in reproductive-age women. Excessive inflammation and elevated androgen production from ovarian theca cells are key features of PCOS. Human bone marrow mesenchymal stem cells (BM-hMSC) and their secreted factors (secretome) exhibit robust anti-inflammatory capabilities in various biological systems. We evaluated the therapeutic efficacy of BM-hMSC and its secretome in both in vitro and in vivo PCOS models. Methods For in vitro experiment, we treated conditioned media from BM-hMSC to androgen-producing H293R cells and analyzed androgen-producing gene expression. For in vivo experiment, BM-hMSC were implanted into letrozole (LTZ)-induced PCOS mouse model. BM-hMSC effect in androgen-producing cells or PCOS model mice was assessed by monitoring cell proliferation (immunohistochemistry), steroidogenic gene expression (quantitative real-time polymerase chain reaction [qRT-PCR] and Western blot, animal tissue assay (H&E staining), and fertility by pup delivery. Results BM-hMSC significantly downregulate steroidogenic gene expression, curb inflammation, and restore fertility in treated PCOS animals. The anti-inflammatory cytokine interleukin-10 (IL-10) played a key role in mediating the effects of BM-hMSC in our PCOS models. We demonstrated that BM-hMSC treatment was improved in metabolic and reproductive markers in our PCOS model and able to restore fertility. Conclusion Our study demonstrates for the first time the efficacy of intra-ovarian injection of BM-hMSC or its secretome to treat PCOS-related phenotypes, including both metabolic and reproductive dysfunction. This approach may represent a novel therapeutic option for women with PCOS. Our results suggest that BM-hMSC can reverse PCOS-induced inflammation through IL-10 secretion. BM-hMSC might be a novel and robust therapeutic approach for PCOS treatment. Supplementary Information The online version contains supplementary material available at 10.1186/s13287-021-02472-w.

theca cell androgen production and the expression of enzymes responsible for producing androgens in vitro (13). Recently, anti-in ammatory therapy has been shown to reduce ovarian androgen secretion and induce ovulation in lean, insulin-sensitive women with PCOS (14). These ndings clearly implicate in ammation as an underlying mechanism of ovarian dysfunction even in the absence of insulin resistance in PCOS.
In the last decade, extensive research has focused on the immunosuppressive and anti-in ammatory effects of bone marrow mesenchymal stem cells (BM-hMSC). Several reports suggest that these effects are mediated by secreted factors including interleukin (IL)-10, an anti-in ammatory cytokine (14)(15)(16). These secreted factors are either released following cross-talk with target cells or produced constitutively by BM-hMSC (15,17). As chronic low-grade in ammation is strongly implicated as a driver of pathophysiology in PCOS(18), we hypothesized that interventions involving BM-hMSC or its secreted factors can improve the endocrine and metabolic abnormalities observed in PCOS, as well as fertility outcomes.
Mice receiving letrozole (LTZ), a nonsteroidal aromatase inhibitor, exhibit increases in circulating testosterone due to impaired conversion of testosterone to estrogen (19). These mice exhibit the hallmarks of hyperandrogenism, anovulation, and polycystic ovaries, as well as impaired fertility and the metabolic dysregulation often observed in women with PCOS (19,20). Recently, anti-in ammatory therapy has been shown to reverse key features of the PCOS phenotype in this mouse model (11). Thus, the LTZ-induced PCOS mouse is a useful model to evaluate the effect of BM-hMSC on in ammatory pathophysiological mechanisms in PCOS, and to establish novel stem cell-based therapeutics for PCOS. Because approximately half of women diagnosed with PCOS have excessive adrenal androgen production (21)(22)(23), androgen-producing human adrenocortical-carcinoma cells (H295R) represent a good in vitro model of PCOS.
We analyzed the effects of exposing H295R cells to the BM-hMSC secretome and IL-10 on testosterone production and the expression of genes encoding enzymes involved in androgen biosynthesis. We performed similar experiments on primary cultures of theca cells obtained from women with PCOS. Finally, we examined the effect of intra-ovarian injection of BM-hMSC on in ammation, metabolism, and ovarian and endometrial gene expression, as well as measures of fertility, in LTZ-induced PCOS mice.

Methods
Human bone-marrow mesenchymal stem cell culture Human BM-hMSC (Passage 2) were purchased from Lonza, USA (PT#2501). These cells were isolated from the bone marrow of a healthy non-diabetic female donor 32-year-old. The cells were cultured in mesenchymal stem cell growth medium (MSCGM) per the manufacturer's recommended expansion protocol. When the culture reached approximately 80% con uence, cells were trypsinized using a 0.05% trypsin-EDTA solution and serially expanded for use in experiments. Cells were characterized for typical BM-MSC-positive (CD90, CD73, CD105) and negative (CD34, CD11b, CD19, CD45, HLA-DR) cell surface markers using the human MSC analysis kit (BD Stem ow TM , CA, USA cat. no. 562245).
Human adrenocortical-carcinoma cell line culture Human adrenocortical-carcinoma cells (H295R cells) were used as an in vitro cell culture model for androgen production. These cells are commonly used in studies of steroidogenesis and androgen biosynthesis pathways (21)(22)(23). H295R cells were purchased from ATCC (Manassas, VA, USA, cat. no. ATCC ® CRL-2128™) and cultured per the recommended guidelines. Brie y, H295R cells were cultured in asks pre-coated with extracellular matrix (Gibco, USA, cat. no. S-006-100) with DMEM/F12 (Gibco, cat. no. 21041025) and 2.5% Nu-Serum (Corning, USA). The cells were subcultured at a ratio of 1:3 to 1:4 and culture media were changed twice a week.
Human theca cell culture from women with PCOS Human theca interna tissue was collected at the time of oophorectomy (n=2), which was performed as clinically indicated using a protocol (24)(25)(26)(27) approved by the Institutional Review Board of The Pennsylvania State University College of Medicine. Theca cells from PCOS ovarian follicles were isolated and cultured as previously reported (28,29). The follicles were isolated from the ovaries and dissected under a microscope in a dish containing a 1:1 mixture of DMEM and Ham's F12 medium supplemented with 10% fetal bovine serum (FBS). The cleaned theca shells were digested with 0.05% collagenase I, 0.05% collagenase IA, and 0.01% deoxyribonucleic, in medium containing 10% FBS. The isolated cells were cultured in dishes pre-coated with bronectin in a 1:1 mixture of DMEM and Ham's F12 medium containing 10% FBS, 10% horse serum, 2% UltroSer G, 20 nm insulin, 20 nm selenium, 1 μM vitamin E, and antibiotics. Experiments were performed using passage 4 (31-38 population doublings) PCOS theca cells.

Preparation of the BM-hMSC secretome
The secretome was prepared from three to ve passages of BM-hMSC in T75 asks. Media were collected and discarded from the BM-hMSC culture at 80-90% con uence. Cells were then washed three times with phosphate-buffered saline (PBS) for complete removal of serum. Cells were then maintained in DMEM/F12 (Gibco, USA) serum-free media for 24 hours to collect the secretome. After 24 hours, the media were collected, centrifuged at 500g for 5 min at 4°C to remove the cell debris, aliquoted, and stored at -80°C for use in experiments. DMEM/F12 serum-free media without cells were incubated for 24 hours in the T75 cell culture ask to serve as a negative control.
For in vivo experiments, the secretome was collected using the above method, and cultured cells were trypsinized from the asks and counted. The average cell count was 2.25 X 10 6 cells per ask. The collected BM-hMSC media were then aliquoted at a volume calculated based on the cell secretions from 5 X 10 5 cells on average per ovary of each mouse. The media/secretome were concentrated using a vacuum concentrator (Labconco, MO, USA) and stored at -80°C for use in in vivo experiments. Before intra-ovarian injection, the concentrated secretome was reconstituted with PBS to a nal volume of 10 µl per ovary.

Treatment of H295R cells and human PCOS theca cells with the BM-hMSC secretome
H295R cells and human PCOS theca cells were cultured separately on pre-coated six-well plates for 48 hours. Cells were then treated for 24 hours with secretome diluted in basal media (serum-free) at a 1:1 ratio. Cell culture media were replaced with serum-free media or secretome media, and cells were incubated for an additional 24 hours. After the incubation period, cells were collected for analysis of steroidogenesis-related gene expression. Cell culture media was used for hormone quanti cation using an automated chemiluminescence immunoassay system, UniCel DxI 800, Access Immunoassay System (Beckman Coulter Inc., CA, USA) (30).

Treatment of H295R cells with recombinant human IL-10
To investigate the anti-in ammatory effect of the BM-hMSC secretome, we measured the amount of IL-10 secreted by BM-hMSC into the culture media by ELISA (Abcam, Cambridge, MA, USA) following the manufacturer's instructions (17). We explored the effect of IL-10 on steroidogenesis-related gene expression, androgen secretion, and pro-in ammatory marker expression in H295R cells after treatment with 0, 125, 250, or 500 pg/ml recombinant human IL-10 (rhIL-10; R & D Biosystem, Cat No. 217-IL-010).
These concentrations were selected based on the previously reported level of IL-10 secreted by hMSCs (31). H295R cells were then collected for gene expression analysis, and cell culture media were used for measurement of testosterone using an automated chemiluminescence immunoassay system, UniCel DxI 800, Access Immunoassay System (Beckman Coulter Inc., CA, USA) (30) and androstenedione using ELISA (Biovision, CA, USA) (32).

PCOS mouse model and intra-ovarian injection of BM-hMSC
Three-week-old female C57BL/6 mice (Charles River, MA, USA) were housed in a vivarium for 1 week under speci c pathogen-free conditions. The animal experiment protocol for this study was approved by the University of Illinois at Chicago Animal Care Committee (UIC ACC). All animal experiments were performed in compliance with the University of Illinois at Chicago's policies and guidelines for use of laboratory animals. At 4 weeks of age, mice (n = 10/group) were subcutaneously implanted with a placebo or 5 mg LTZ pellet (Innovative Research of America, Sarasota, FL, USA), which provides a constant release of LTZ (50 μg/day). Body weight was monitored weekly before and post-implantation. Body weight and insulin resistance (measured by glucose tolerance test, GTT) were used to monitor development of PCOS characteristics.
Five weeks after placebo or LTZ pellet implantation, mice underwent intra-ovarian injection of BM-hMSC via laparotomy. Mice were treated preoperatively with a single dose of buprenorphine (0.1 mg/kg) and were kept under anesthesia with 1−4% inhalation of iso urane during the entire procedure. A single midline incision, less than 25 mm, was made on the skin to access both ovaries via the caudal abdominal cavity. For the BM-hMSC group, cells were injected in both ovaries at a concentration of 5.0 X 10 5 cells per ovary resuspended in 10 µl PBS. For the secretome group, concentrated secretome reconstituted in 10 µL PBS was injected per ovary in both ovaries. For the control group, 10 µl of PBS was injected into both ovaries. The incision was closed by suturing, followed by wiping with a clean disinfectant swab. Two weeks after BM-hMSC engraftment or secretome injection, the mice were anesthetized and gonadal fat pads, brown fat, and ovaries were collected. A portion of the gonadal and brown fat, as well as one ovary, were xed in 4% paraformaldehyde and embedded in para n; the remainder of the tissue and the other ovary was frozen at -80°C for further analysis.

Glucose tolerance test
Glucose tolerance testing was performed on mice 5 weeks after placebo or LTZ pellet implantation and 2 weeks after BM-hMSC engraftment or secretome treatment. Mice were fasted for 16 h (5 p.m. to 9 a.m.), with free access to drinking water, after which they received an intraperitoneal (i.p.) injection of D-glucose (2.0 g/kg body weight). Blood glucose level was measured at 0, 15, 30, 60, 90, and 120 min following glucose injection using a Bayer glucose monitor (Roche Diagnostics Corp, IN, USA).

Indirect calorimetry
Metabolic rate was measured in mice at 11 weeks of age by indirect calorimetry in open-circuit Oxymax chambers, a unit of the Comprehensive Lab Animal Monitoring System (CLAMS; Columbus Instruments, state, USA). Two weeks after BM-hMSC treatment, mice receiving LTZ only or LTZ and treated with BM-hMSC (n=3) were acclimated to calorimetry cages for 2 days before data sampling at 23 o C under 12:12 hours light:dark cycle. Oxygen consumption rate (VO 2 ), carbon dioxide release (VCO 2 ), respiratory exchange ratio (RER), and heat production were measured in individual mice. The horizontal activity was measured on x, y, and z-axes.

Serum hormone measurements
Blood was collected from all the groups by cardiac exsanguination under iso urane anesthesia; serum was separated and stored at −80°C. Serum hormone levels were measured at the University of Virginia Ligand Core Facility. Serum testosterone (T) and estradiol (E2) levels were measured using ELISA. Serum luteinizing hormone (LH) and follicle-stimulating hormone (FSH) levels were measured by radioimmunoassay (RIA). The sensitivities of each assay are 10 ng/dL (T), 3 pg/ml (E2), 3 ng/ml (FSH), and 0.04 ng/ml (LH). Serum cytokines were analyzed in a membrane-based antibody array (Ray Biotech, GA, USA) per the manufacturer's protocol.

Breeding experiments
One week after BM-hMSC engraftment or secretome treatment, 6 mice per group were randomly selected for the breeding experiment. One male C57BL/6 breeder mouse was used for every two female mice. The male and female mice were caged together for 10 days. Mating was determined by the presence of sperm plug in the vagina. Most of the female mice showed a sperm plug within 3 days, and the average number of pups from each female mouse was compared between treatment groups. At the end of the experiment, all delivered pups were counted per group, their body weight was measured, and any morphological anomalies were noted.

Histology and immunohistochemistry
Ovaries and fat tissues were collected, xed in 4% paraformaldehyde, and embedded in para n blocks. Tissue sections were stained with hematoxylin-eosin (H&E) and murine anti-UCP-1 (Abcam, MA, USA), followed by detection with a biotin-labeled rabbit anti-rat antibody and staining with the ABC kit (Vector Laboratories, Burlingame, CA, USA). Sample processing and staining were performed at the histology core of the University of Illinois at Chicago (Chicago, IL, USA). Histological analyses were performed using Asperio ImageScope (Leica Biosystem, Wetzlar, Germany).

Immunoblot analysis
Following treatment of H295R cells and human PCOS theca cells with the BM-hMSC secretome, and treatment of mice with BM-hMSC or its secretome, cultured cells and collected ovarian tissue were lysed with RIPA buffer (Cell Signaling, MA, USA) containing protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific Inc., MA, USA) and sonicated at 20 amplitude with 5 sec on and 5 sec off for a 1-minute cycle. Sonicated samples were then centrifuged at 12,000 rpm for 5 min and the supernatant was transferred into separate tubes. The protein concentration of all samples was determined by the Bradford method. For immunoblot analysis, samples containing equal amounts of protein were incubated with 1x gel loading buffer and separated by SDS-PAGE (4-20% criterion, Bio-Rad), then transferred to PVDF membrane using a Trans-blot turbo system (Bio-Rad, Hercules, CA, USA). After protein transfer, blocked membranes were incubated in 1% non-fat dry milk in 1x PBS (0.05% Tween) overnight at 4°C, with primary antibodies against CYP17A1 (ab125022, 1:500 dilution, Abcam), CYP11A1 (ab75497, 1:500 dilution, Abcam), DENND1A (LS-C167356, 1:250, LSBio), VEGFA (ab1316, 1:500 dilution, Abcam), or β-actin (clone AC-15, A5441, 1:5000, Sigma) in 1% non-fat dry milk in 1x PBS with 0.05% Tween overnight at 4°C. After washing, the membrane was incubated with the appropriate HRP-linked secondary antibodies (anti-mouse secondary antibody, cat. no. 7076, 1:5000 or anti-rabbit secondary antibody, cat. no. 7074, 1:3000, Cell Signaling) in 5% non-fat dry milk in 1x PBS with 0.1% Tween at room temperature for 1 hour. The membrane was developed with Trident Femto Western HRP substrate (GeneTex, Irvine, CA, USA) and visualized using the ChemiDoc XRS + molecular imager (Bio-Rad, Hercules, CA, USA). After imaging, membranes were stripped with Restore TM PLUS stripping buffer (Thermo Scienti c, MA, USA) to incubate with another antibody. The signal density of each protein band was quanti ed using Image J software (US National Institute of Health, Bethesda, MD, USA) and normalized against the corresponding β-actin band.
Quantitative RT-PCR RNA was extracted from H295R cells and human PCOS theca cells treated with the BM-hMSC secretome or rhlL-10. RNA was also extracted from fat and ovarian tissues collected from mice treated with BM-hMSCs or the BM-hMSC secretome. RNA extraction was done using TRIzol (Invitrogen, USA) according to the manufacturer's instructions. The concentration and purity of the extracted RNA were checked using a NanoDrop spectrometer (Thermo Scienti c, MA, USA). 1 µg of total RNA was reverse transcribed using RNA to cDNA EcoDry™ Premix (Double Primed) (Takara Bio USA Inc., CA, USA). The reaction mixture was incubated for 1 h at 42°C; incubation was stopped at 70°C for 10 min. Quantitative real-time PCR (qPCR) was performed using the CFX96 PCR instrument and SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) with speci c primers to the target genes in a 20 µL nal reaction volume. The primer sequences are listed in Table S1. Beta-actin was used as a reference gene for sample normalization. The delta-delta threshold cycle (ΔΔCt) method was used to calculate the fold change expression in mRNA level in the samples.

Flow cytometry (FACS) analysis
After treatment with the BM-hMSC secretome or basal media control, H295R cells were analyzed by FACS for proliferation, apoptosis, and in ammatory markers using antibodies against Ki67 antibody (BioLegend, Cat no. 350514), Annexin-V (BioLegend, Cat no. 640919), IL-1β (R&D Systems, Cat no. IC8406A), and TNF-α (BioLegend, Cat no. 502943). In brief, treated cell pellets were harvested and xed/permeabilized with BD cyto x/cytoperm kit reagent (BD Bioscience, CA, USA) for intracellular staining, per the manufacturer's instructions. After centrifugation at 1500 rpm for 5 minutes, a total of 1 X 10 6 cells were resuspended in 200 µl of antibody solution and incubated for 30 min at room temperature in the dark. After washing, the cells were resuspended in PBS with 2% FBS (v/v) for FACS analysis using (BD, Gallios, Flow-cytometer). Data were analyzed using FlowJo software.

Statistical analysis
Comparisons between groups were made by one-way ANOVA with Tukey's post hoc test or Student's ttests. All data are presented as mean ± standard deviation (SD). A difference between groups with * p<0.05, * * p<0.005, or * * * p<0.0005 was considered statistically signi cant.
BM-hMSC secretome decreases steroidogenesis-related gene expression and androgen production in H295R cells We previously reported that CYP17A1, CYP11A1, and DENND1A, key genes for ovarian androgen biosynthesis, are upregulated in PCOS-theca cells compared with healthy theca cells (27,33). Hence, we evaluated the effect of the BM-hMSC secretome on the expression of these genes using our in vitro model. Secretome treatment resulted in signi cant downregulation of CYP17A1 (0.56 ± 0.02 fold) and DENND1A (0.37 ± 0.05 fold) gene expression in H295R cells compared with media-treated cells (Fig. 1d-f). However there was no signi cant decrease in CYP11A1 gene expression (0.82 ± 0.05 fold, p=0.127). We con rmed these ndings at the protein level using immunoblot analysis, which showed that CYP17A1 (0.84 ± 0.02 fold) and DENND1A (0.26 ± 0.01 fold) were signi cantly decreased in secretome-treated H295R cells compared with the control group, while no change was observed in CYP11A1 (0.97 ± 0.02 fold, p=0.29; Fig. 1g We investigated the effects of steroidogenesis-related gene inhibition by the BM-hMSC secretome on testosterone secretion.We explored whether steroidogenesis-related gene inhibition by the BM-hMSC secretome affected testosterone secretion. Compared with the media control group (474.6±27.5 ng/dL), secretome treatment suppressed testosterone secretion in H295R cells (267.7±4.0 ng/dL) (Fig. 1l). Testosterone secretion was also suppressed in human PCOS theca cells (146.4±13.4 ng/dL) (Fig. 1m) compared with the media control group (214.7±11.8 ng/dL). In summary, our data indicate that BM-hMSC secreted factors inhibit androgen production.
BM-hMSC reverse the metabolic phenotypes in an LTZ-induced PCOS mouse model Next, we evaluated the potential therapeutic effects of BM-hMSC in vivo by injecting BM-hMSC into the ovaries of the well-established LTZ-induced PCOS mouse model (19). Five weeks after LTZ implantation, the mice in the PCOS group were signi cantly heavier (21.1 ± 0.25 grams) compared with age-matched control mice (19.3 ± 0.60 grams) that had received placebo pellets (Fig. 3a, b). Since PCOS women have insulin resistance and impaired glucose tolerance (45), we also performed a glucose tolerance test (GTT) and measured energy expenditure in PCOS mice before (5 weeks after LTZ or placebo) and 2 weeks after BM-hMSC engraftment (7 weeks after LTZ or placebo). Interestingly, PCOS mice treated with BM-hMSC exhibited a normal glucose tolerance pro le compared with untreated PCOS mice (Fig. 3c-e). Moreover, we found that the untreated PCOS group had lower energy expenditure, based on a signi cant difference in thermogenesis, compared with PCOS mice treated with BM-hMSC (Fig. 3f-i).
The increase in thermogenesis in BM-hMSC-treated PCOS mice encouraged us to further evaluate fat metabolism in treated versus control PCOS mice. A previous study revealed the role of brown fat cells in the regulation of total energy expenditure (46). A process called "browning," which refers to the transition of white fat into brown fat, is associated with upregulation of UCP-1 (47). Therefore, we stained white fat tissues collected from BM-hMSC-treated and untreated PCOS mice with UCP-1 and found greater proportions of brown-like fat cells, suggesting increased browning of white fat, in the BM-hMSC-treated group (Fig. 3j). At the molecular level, the browning process is regulated by several genes that control multiple aspects of mitochondrial activity, such as Pgc-1α, Cidea, and Prdm-16 (48). qPCR con rmed our UCP-1 immunohistochemistry results and showed signi cant upregulation of Ucp1 (21.00 ± 0.67 fold), Pgc1a (2.61 ± 0.08 fold), Cidea (3.78 ± 0.31 fold), and Prdm16 (5.15 ± 0.19 fold) gene expression in white fat collected from the BM-hMSC-treated PCOS mice compared with the untreated PCOS mice (Fig. 3k-n).
We also explored marker expression levels in brown fat tissue, and found that BM-hMSC treatment increases brown fat-related marker expression even in brown fat tissue (Fig S4). These results suggest that BM-hMSC can regulate adipose tissue metabolism by ameliorating in ammation and promoting brown fat formation.

BM-hMSC normalize the adipokine pro le in adipose tissue in an LTZ-induced PCOS mouse model
Weight gain associated with LTZ-induced PCOS is partially due to white fat expansion (19). The expansion of gonadal fat in our LTZ-induced PCOS mice was marked by characteristic morphologic enlargement of fat cells detected by H&E staining (Fig S5a). Remarkably, the average size of adipocytes in the BM-hMSC-treated PCOS mice was signi cantly smaller than that in the untreated PCOS mice (Fig   S5b), approaching the normal size range of adipocytes.
White fat adipocyte expansion is usually associated with an increase in leptin that correlates inversely with adiponectin levels (49). Studies have shown that adiponectin is a pivotal adipokine that can reverse PCOS metabolism (50), acting as a humoral factor that regulates fat homeostasis by establishing crosstalk between white and brown fat cells (51). To explore such cross-talk in our PCOS mouse model, we measured gene expression of leptin and adiponectin in brown adipose tissue as well as white gonadal fat using qPCR. Treatment with BM-hMSC upregulated adiponectin and downregulated leptin expression, thus normalizing the ratio of leptin to adiponectin in brown fat tissue compared with the untreated group (Fig S5c-e). Similar ndings were observed in the white gonadal fat (Fig S5f-h), highlighting the ability of BM-hMSC to normalize fat metabolism in our PCOS mouse model.

Serum hormone analysis
To assess the endocrine status following BM-hMSC engraftment, total serum hormone levels in BM-hMSC-treated and untreated PCOS animals were measured. Serum T levels were signi cantly higher in the untreated PCOS group versus healthy controls, with no signi cant difference in the BM-hMSC-treated group (p=0.797). Furthermore, there were no changes in serum estrogen levels among the three groups.
However, LH was signi cantly lower in the PCOS group than healthy controls, and LH levels decreased after BM-hMSC engraftment in PCOS mice. In addition, FSH levels were lower in the PCOS group compared with healthy controls and increased after BM-hMSC treatment, though the change was not statistically signi cant (Fig S6a-d).
BM-hMSC treatment reverses endometrial abnormalities in an LTZ-induced PCOS mouse model PCOS imparts abnormalities in endometrial tissue, such as the thickening of endometrium epithelial cells and aberration of steroid receptor gene expression (52)(53)(54). Consequently, we analyzed endometrial tissue in BM-hMSC-treated versus untreated PCOS mice. The endometrial tissue of the PCOS group showed abnormal thickness (Fig S7a) and the AIB1 gene, known to be elevated in the PCOS endometrium (54), showed signi cant alterations in the PCOS group endometrium. These abnormalities were reversed in the BM-hMSC-treated PCOS group (Fig S7b). Similarly, steroid receptor genes AR and ERβ trended higher in the PCOS group and this was reversed after BM-hMSC treatment (Fig S7c, d).
Interestingly, the proliferation marker Ki-67 was signi cantly upregulated in BM-hMSC-treated PCOS mice compared with the untreated PCOS group (Fig S7e). Additionally, several in ammatory regulator genes such as IL6, IL16, CCL2, and TNF-α were higher in the PCOS group endometrium compared with normal control endometrium, and these changes were signi cantly reversed after BM-hMSC treatment (Fig S7f-i). These results suggest that intra-ovarian injection of BM-hMSC reversed various alterations in PCOS endometrium, at least partially, by normalizing steroid hormone receptors and in ammatory cytokine gene expression. These changes likely improved the quality of endometrium and contributed to the improved reproductive outcomes in the PCOS mice after BM-hMSC treatment.

BM-hMSC restore fertility in an LTZ-induced PCOS mouse model
To explore the effect of BM-hMSC treatment on reproductive function, we rst analyzed ovarian morphology in BM-hMSC-treated versus untreated PCOS mice. Ovaries from untreated PCOS mice displayed typical PCOS characteristics, including lack of corpora lutea and antral follicles compared with untreated normal control mice (Fig. 4a). After intra-ovarian engraftment of BM-hMSC in both ovaries of PCOS mice, normal ovarian morphology was partially restored, including the reappearance of corpora lutea and antral follicles, as well as well-ordered stroma that was morphologically similar to that of the normal control group. These morphological changes suggest that BM-hMSC engraftment improved the pathological changes in PCOS ovaries and may potentially restore ovulation in PCOS mice (Fig. 4a).
Next, we performed a breeding experiment to test BM-hMSC's treatment capacity to restore fertility in our PCOS mouse model. We found that healthy control mice had a higher rate of fertility (80%) than the subfertile PCOS group (10%). Interestingly, the pregnancy rate in BM-hMSC-treated PCOS mice was restored to a rate equal to that of the control group (Fig. 4b).
We also counted the number of delivered pups in all experimental groups. As shown in Figure 4d-g, we found that most PCOS mice were infertile. The number of pups delivered in the BM-hMSC-treated PCOS group was equivalent to the number delivered in the control group. Moreover, the average number of pups delivered per mouse in the BM-hMSC-treated PCOS group (5.5 ± 1.1) was signi cantly higher than that in the untreated PCOS group (0.8 ± 1.7; Fig. 4g). We also found no signi cant differences in the average body weight of the delivered pups between the control group and BM-hMSC-treated PCOS group at 0, 5, and 10 postnatal days (Fig. 4h). Notably, we did not observe any apparent morphological abnormalities in any pups during the study period. We further tested the effect of injecting the BM-hMSC secretome in our PCOS mouse model (Fig S8). Fertility of PCOS mice was restored in secretome-treated animals.
These results suggest that either BM-hMSC or its secretome can restore impaired fertility in an LTZinduced PCOS mouse model with no detectable abnormalities in the delivered newborns.
BM-hMSC restore normal ovarian gene expression in an LTZ-induced PCOS mouse model PCOS abnormalities include enhanced androgen production and altered ovarian angiogenesis (55). We next examined the effect of BM-hMSC treatment on these abnormalities in our PCOS mice in vivo, to validate our in vitro data on the secretome effects on ovarian steroidogenic gene and in ammation marker expression. Mouse ovarian tissues from BM-hMSC-treated and untreated animals were analyzed for RNA and protein levels of steroidogenesis and angiogenesis markers. Cyp17a1 gene expression was signi cantly elevated in PCOS ovaries (13.73 ± 5.78 fold), which was signi cantly reversed after BM-hMSC treatment (1.22 ± 0.20 fold; Fig. 5a). Cyp19a1 (0.14 ± 0.02 fold) and Fshr (0.03 ± 0.01 fold) gene expression were lower in PCOS group ovaries, consistent with the prior characterization of the LTZinduced PCOS mouse model (19); levels of both genes signi cantly increased after BM-hMSC treatment (Cyp19a1: 0.93 ± 0.11 fold, Fshr: 0.80 ± 0.09 fold; Fig. 5b, c). Previous studies reported an abnormal increase in ovarian angiogenesis in PCOS (55, 56); thus, we measured gene expression of angiogenesis marker Vegfa in untreated and BM-hMSC-treated PCOS mice. Gene expression of Vegfa was elevated in the PCOS group (7.50 ± 3.69 fold) and decreased after BM-hMSC treatment (0.48 ± 0.29 fold; Fig. 5d).
Immunoblot analysis supported these results, where CYP17A1 was higher in the PCOS group (2.21 ± 0.14 fold) and signi cantly decreased after BM-hMSC treatment (1.37 ± 0.12 fold; Fig. 5e, f). Moreover, VEGF-A protein expression was elevated in PCOS mice (1.28 ± 0.15 fold) and decreased after BM-hMSC treatment (1.08 ± 0.38 fold), although the change was not statistically signi cant (Fig. 5e, g). Taken together, our in vivo results suggest that BM-hMSC treatment can inhibit androgen synthesis and angiogenesis consistent with the effect of the BM-hMSC secretome on H295R cells.

BM-hMSC secretome improves metabolic and reproductive phenotypes in LTZ-induced PCOS mice
Our in vitro and in vivo data suggest that the favorable effects of BM-hMSC engraftment likely occur in a paracrine fashion via secreted humoral factors in the BM-hMSC secretome. To explore the paracrine effect of BM-hMSC in the LTZ-induced PCOS mouse model, we delivered the BM-hMSC secretome by direct intra-ovarian injection into the ovaries of mice, and various metabolic and reproductive parameters were assessed. Analysis of white fat demonstrated a signi cant reduction in the size of fat cells in the secretome-treated PCOS group compared with the untreated PCOS group (Fig S8a), and the expression of UCP1 in brown fat tissue was also signi cantly higher in the secretome-treated PCOS group compared with untreated PCOS group (Fig S8b).
Morphological comparison of ovaries among the groups of mice by H&E staining revealed that secretome-treated PCOS ovaries had more antral follicles compared with untreated PCOS group ovaries (Fig S8c). Importantly, secretome treatment also restored fertility in PCOS mice compared with the untreated PCOS control group (Fig S8d, e). These results con rm the ability of the BM-hMSC secretome to reverse metabolic and reproductive abnormalities in the LTZ-induced PCOS mouse model and suggest that the positive effects of BM-hMSC are primarily mediated via paracrine action of the BM-hMSC secretome.
Taken together, our data suggests that intra-ovarian injection of BM-hMSC reduces in ammation by increasing the expression of anti-in ammatory mediators such as IL-10 and its receptor in the ovary, and circulating IL-10, IFN-γ, and TIMP-2 in serum, while decreasing pro-in ammatory mediators such as IL-6, IL-1β, CCL2, and CD11c gene expression in periovarian adipose tissue.

Discussion
In this study, we report a signi cant inhibitory effect of the BM-hMSC secretome on steroidogenesis gene expression, in ammation, and androgen production in H295R cells, as well as in primary cultures of theca cells from women with PCOS. Additionally, our in vivo experimental data showed that intra-ovarian engraftment of BM-hMSC is capable of correcting several PCOS-related metabolic abnormalities in a mouse model of PCOS. While this LTZ-induced PCOS mouse model is infertile (19), we demonstrated that BM-hMSC treatment was able to restore fertility and treated mice delivered healthy pups. Interestingly, similar improvements in metabolic and reproductive endpoints were achieved with injection of BM-hMSC secretome, suggesting that most, if not all, of BM-hMSC effects in this model are paracrine in nature.
Chronic in ammation plays an important role in PCOS pathogenesis (32). BM-hMSC engraftment signi cantly reduced several in ammatory markers in PCOS mouse ovaries. Reports have demonstrated a positive feedback loop between in ammation and androgen production (11,12,32), suggesting that androgen synthesis and in ammation could be reciprocally self-propagated. Up-regulation of CYP17A1 gene expression through oxidative stress, which is a known stimulator of in ammation (63, 64), also demonstrates a positive feedback loop in PCOS. Surprisingly, while there was signi cant suppression of androgen production in vitro after BM-hMSC secretome treatment, we did not see a difference in serum testosterone levels between the untreated PCOS and BM-hMSC-treated PCOS groups. This could be attributed to the episodic nature of steroid hormone secretion. Furthermore, our ndings may highlight a limitation of the chemically (LTZ)-induced PCOS model, which primarily relies on the induction of higher testosterone accumulation via marked supraphysiological inhibition of its aromatization (19). Key ovarian steroidogenic genes as Cyp17a1 were upregulated in the PCOS group, and signi cantly suppressed by BM-hMSC treatment. The effect of engrafted BM-hMSC on ovarian cells could occur via cell-to-cell contact or paracrine effects through secreted humoral factors.
IL-10 is an important immune-suppressive and anti-in ammatory cytokine that is key to several human disorders, including PCOS. Recent reports showed signi cantly lower serum levels of IL-10 in PCOS women compared with age-and BMI-matched healthy controls (42). BM-hMSC secrete physiologically relevant quantities of IL-10 (31,39,40), which we con rmed in the BM-hMSC used in this work (Fig. 2a).
We showed that IL-10 treatment signi cantly downregulates steroidogenesis and in ammatory gene expression as well as suppresses androgen production by H295R cells (Fig. 2f). In vivo, BM-hMSC treatment signi cantly increased IL-10 expression in ovarian tissue and its serum concentration in PCOS mice. These results suggest that BM-hMSC can ameliorate PCOS-induced in ammation through IL-10 secretion, and IL-10-overexpressing BM-hMSC might be a novel and robust therapeutic approach for PCOS treatment (Fig. 6k).
Recently, two reports have described the utility of tail-vein injected MSCs to reverse some PCOS immunephenotypes (65, 66). In these studies, MSCs inhibited T cell proliferation, decreased in ammation in vitro and in vivo, and enhanced ovarian function in a PCOS animal model. However, these studies lacked translational fertility, reproductive, and metabolic outcomes data. Moreover, many reports have described that, after systemic infusion, stem cells are trapped in the lungs with limited in vivo persistence (67-69). In contrast, local engraftment of hMSCs directly into the target organ, as we describe here, can initiate the reparative process in a more robust manner for cell homing and effective tissue repair (70). Additional research is needed to evaluate the utility of intra-ovarian engraftment of BM-hMSC as a potentially promising approach for the treatment of PCOS-associated infertility in women.

Conclusions
Polycystic ovary syndrome (PCOS) is the most common disease in women. Many PCOS patient typically shows excessive in ammation and infertility. In this study, we present a stem cell-based therapy to restore fertility in PCOS condition. We report that BM-hMSC secretome inhibit steroidogenesis, in ammation, and androgen production in both H295R cells and primary cultured theca cell. We also report the therapeutic e cacy of mesenchymal stem cells (BM-hMSC) in PCOS mouse models. Our study shows that BM-hMSC can treat PCOS-related characteristics, including infertility. This approach may represent a novel therapeutic option for women with PCOS.