The Lipid-related Effects of Resveratrol on Human Ectopic Endometrial Stromal Cells and a Rat Model of Endometriosis


 Background: Endometriosis is a complex disease in the field of gynecology that has certain limitations for its interim treatment. Resveratrol has been recently used for the treatment of endometriosis in experimental and clinical studies, but its molecular mechanism remaines elusive. Results: In this study, based on a case-control study, we identified that a decreased BMI and altered lipid profiles were associated with endometriosis patients. We applied resveratrol treatment on human ectopic endometrial stromal cells (HEcESCs) and a rat model of endometriosis. Lipidomics analysis showed that resveratrol altered lipid profiles in HEcESCs, with the sphingolipids Cer and SM increased significantly, while FA and most phospholipids were significantly reduced. Pathway enrichment analysis showed that several lipid-associated signaling pathways could be targeted by resveratrol. Our experiments in a rat model showed that resveratrol reduced the lesion and rectified lipid profiles in rats with endometriosis. In addition, resveratrol treatment significantly increased the expression of PPARα in lesion tissues of model rats and HEcESCs of EMs patients.Conclusion: Our data reveal that the development of EMs is closely related to lipid metabolism, and resveratrol may play a therapeutic role by targeting the lipid metabolism of ectopic endometrial stromal cells in endometrosis. Our study provides valuable insights for understanding the pathogenesis and clinical treatment of endometriosis.

molecules to cells and HDL-C promotes cholesterol e ux from cells (9). Gene expression analysis revealed that LDL receptors were highly expressed in endometrial tissues of patients with DIE (10). ApoA1 mediates the exchange between HDL and chylomicron and functions as a regulator of in ammatory responses (11). These studies indicated that there may be some strong links between the disorders of fat digestion and absorption, and the in ammation in the patients with EMs. We have previously investigated rutine metabolite parameters of EMs patients using enzymatic colorimetric assays or the immune turbidimetric methods, attempting to nd potential indicators that can be used to detect the EM phenotype (12). However, such serum metabolites index seemed to be lack of speci city and sensitivity for the diagnosis of EMs.
Recently, lipidomics analysis has been widely used to assess lipid homeostasis in various tissue samples (13,14). In such studies, elevated levels of SM, PC and TG were identi ed in serum samples of OMA patients (15). Signi cant alterations in SM, PC, TG and PE between the eutopic and ectopic endometrium were also noted in patients with EMs (16,17). As a direct in ltration environment for ectopic endometrium, peritoneal uid in patients with EMs was found to have a decreased PC level (18,19). Such lipidomics analysis data has been mostly targeted to nd biomarkers for the clinical detection of this disease. Although these observations indicate that abnormal lipid distribution may play a major role in pathology related to EM, little attention has been given towards the potential application of lipidomics analysis in evaluating drug e cacy or further elucidating disease mechanisms. Rresveratrol (trans-3,5,4'trihydroxystilbene), a phytoalexin polyphenol found in natural plants or fruits, has previously been highlighted as a potential supplement for the treatment of cancers, cardiovascular disease and EMs (20,21). The pharmacological effects of resveratrol on energy and lipid metabolism have been revealed in animal models or in human eutopic endometrial stromal cells (HESCs) of EMs (22). Resveratrol intervention also led to a decrease in total cholesterol and triacylglycerol concentrations in individuals with dyslipidemia ( 23,24). The lipid-related effects of resveratrol on HEcESCs and animal model of EMs is no known.
In the present study, based on analysis of metabolic indicators in clinical cases, we identi ed that a decreased BMI and abnormal lipid metabolism is strongly associated with the development of endometriosis. Using lipidomics analysis, we evaluated the effect of resveratrol on HEcESCs. Therapeutic effect of resveratrol on EMs model rats were also observed, showing with atennuated lesion size and recti ed lipid pro les. Our study provides valuable insights towards the understanding of pathologenesis of EMs and reveals the potential of resveratrol for the treatment of patients with endometriosis.

Decreased BMI and altered serum lipid pro les in EMs patients
In a case-control study (n = 205), we assessed the clinical characteristics and serum metabolic pro les in women with or without EMs. 110 patients were histologically con rmed with EMs and 95 EMs-free women served as the control group. According to the phenotypes, the EMs patients were sub-grouped as 79 OMA and 31 DIE. The patient's distribution according to the r-ASRM stage was 46 moderate and 64 severe. Among controls, the indications for surgery were summarized as benign ovarian tumors (41 cases), tubal infertility (13 cases), cervical intraepithelial neoplasia (28 cases), and intrauterine adhesion (13 cases). There were no signi cant differences in age, parity, and gravidity across the different groups.
However, univariate analyses revealed that the BMI in all patients with Ems and in patients with OMA were signi cantly lower than those of control. However, there were no signi cant differences with respect to serum TC, TG, LDL, HDL, Lp(α) and UA levels between EMs and control patients. Strikingly, elevated levels of R-ApoA1/ApoB and decreased levels of R-ApoB/ApoA1 and serum FBG were displayed in all EMs patients as compared with controls (Table. 1). With respect to the EMs phenotypes, serum R-ApoA1/ApoB levels were both signi cantly higher in OMA and DIE patients, as compared to controls. Elevated serum Lp(α) and a decreased FBG levels were observed in OMA patients but not in DIE patients, when compared to controls. There was also signi cant differences in Lp(α) levels between OMA patients and DIE patients (Table. 1). A stepwise logistic regression analysis showed that BMI was associated with a decreased chance, whereas Lp(α) and R-ApoA1/ApoB was associated with an increased risk of OMA. As for DIE, only R-ApoA1/ApoB showed a strong association with the phenotype (Table 2). Diagnostic performances of BMI and serum lipid pro les for EMs were also predicted using ROC analysis. The area under the curve (AUC) for R-ApoA1/ApoB could as a single biomarker of EMs with good speci city and relative low sensitivity (Fig. S1). These data indicated that an abnormal lipid metabolism was strongly associated with the development of EMs.

Lipidomics analysis on the HEcESCs upon resveratrol treatment
With permission, we isolated HEcESCs of lesion samples obtained from 8 EMs patients. The cultured primary HEcESCs treated with resveratrol for 48 hours (Res groups), together with those treated with DMSO as controls (Con groups), were subjected to lipid extraction and non-target lipidomics analysis by UPLC-MS. Based on the OSI/SMMS lipid library, 809 qualitative lipid structures were differentially classi ed, mainly including 5 types of glycerophospholipids (PC, PE, PG, PS, PI), 4 types of sphingolipids (SM, Cer, HexCer, Hex2Cer), 3 types of glycerolipids (MG, DG, TG) and FA. 638 lipids under the positive ions model and 313 lipids under the negative ions model were recognized. Among these, 132 lipids were identi ed in both ion models. Comparing the peak value of differential lipids between the Res groups and Con groups, 63 lipids were quanti ed as signi cantly altered candidates upon resveratrol treatment (P < 0.05, FC > 1.5, VIP > 1). Using One-MAP (www.5omics.com), univariate data analysis showed the overall metabolite features with differential variations among samples of the paired groups (Res vs Con). PE  (Fig. 1A). The lipid changes between the paired groups were also showed with a Z-scores plot (Fig. 1B). In particular, the sphingolipids (such as Cer, SM) showed obvious upregulation, and glycerolipids (such as DG, TG), FA, and most of the phospholipids including PC, LPC, PE, LPE, PG, PI, PS, showed signi cant down regulation ( Fig. 1C and Table. S1).
Multivariate statistical analysis showed that there was an obvious separation trend between the Res group and Con group, rstly revealed in the PCA model ( Fig. 2A). The two groups were then signi cantly distinguished in the supervised models PLS-DA and O-PLS-DA ( Fig. 2B and 2C). ROC analysis showed a high quality of the predictive value (AUC = 1) (Fig. 2D). The perturbative index (R2 × 0 = 0.29, Q2 × 0=-0.26) indicated that the model had good predictability and reliability (Fig. 2E). Among all the lipids altered after resveratrol treatment, PI (15:1-16:2) showed the most variation contributing to the separation between the two groups, with a highest VIP (VIP = 2.88) (Fig. 2F).
Resveratrol-mediated changes in lipid-associated signaling pathways The above altered lipid metabolites (under criteria either VIP > 1, P < 0.05 or FC > 1.5) were subjected to the KEGG database for pathway enrichment analysis. As shown in Fig. 3A, the lipidomic alterations upon resveratrol treatment were mostly assigned to the glycerophospholipid metabolism, insulin resistance (IRS) and sphingolipid signaling pathways, among all related pathways (Table. S2 We next examined the effects of resveratrol on the lesion cells obtained from patients. Three groups of HEcESCs treated with or without resveratrol were subjected to the assays for evaluation of cell proliferation, invasiveness and apoptosis. After 48 hours treatment with resveratrol at different concentrations (40 µM and 100 µM), the proliferation capacity of the HEcESCs had decreased in the Res-40 µM group and Res-100 µM group compared with the control groups (Fig. 4A). In the invasiveness assay, the number of cells crossing the matrigel differed signi cantly with resveratrol treatment at a concentration of 40 µM or 100 µM, which were decreased respectively, compared with the control group ( Fig. 4B and 4C). The effects of resveratrol on the apoptosis of HEcESCs were shown in Fig. 4D, both Res-40 µM and Res-100 µM showed signi cant differences (Fig. 4E).
Lipid pro l varys with the severity of endometriosis in a rat model.
We rstly generated a rat model of EMs by autologous transplantation of rat estrus epithelial tissue into the endometrial abdominal wall ( Fig. 5A and 5A') to test the therapeutic effects of resveratrol. Examined after 4 weeks of modelling, successful implants showed EMs-like lesions appearing as vesicular cysts ( Fig. 5A and 5A"), lled with clear or turbid yellow-brown liquid and surrounded by connective tissue and angiogenesis ( Fig. 5A and 5A'''). HE staining showed that the pathological features of implant-derived ectopic endometrium shared similarities to the eutopic endometrium ( Fig. 5B and 5C).
To further evaluate the pathological characteristics of the model rats, we examined the serum metabolites including cholesterol, HDL, LDL and TG of the model rats and the sham group of animals. We evaluated the relationship between lesion size and the lipid levels in the serum. The EMs model animals were classi ed into three levels according to severity (n = 10 in each group): EMs 1: 2 mm 3 ≤ lesion volume < 20 mm 3 , EMs 2: 20 mm 3 ≤ lesion volume < 100 mm 3 , EMs 3: lesion volume ≥ 100 mm 3 . Compared to the sham group, in EMs 1 group, there were no signi cant differences in serum cholesterol We applied animal open eld assay to evaluate the anxiety of rats with EMs that were likely associated with increased stress or pain. This experiment was often used as test for anxiety, exploration, and locomotion and the behavioral responses could be scored by measuring the time spent in the center zone and numbers of center crossing (25). The results indicated that the EMs group showed more anxiety than the control group ( Resveratrol attenuated the lesion size and aberrant lipid pro les of the EMs model rats Resveratrol were applied for intraperitoneal injection in the experimental rats, after 4 weeks, the ectopic endometrial lesions of the animal models treated with or without resveratrol were examined. Compared to the EMs group without resveratrol treatment ( Fig. 5H and 5I), signi cant reduction of lesion sizes were shown in both Res-med groups and Res-high groups (Fig. 5J). The pathological lesions of EMs are typically characterized by the histological accumulation of endometrial epithelial, glandular tubes, and signi cant invasive growth (26). Histochemical staining showed that resveratrol treatment led to both signi cant decreases in glandular tubes and endometrial epithelial thickness in both Res-med groups and Res-high groups (Fig. 5K), compared to the EMs group without resveratrol treatment (Fig. 5K' and 5K").
After resveratrol treatment for 4 weeks, we measured the serum cholesterol (Fig. 6A), HDL (Fig. 6B), LDL ( Fig. 6C), and TG (Fig. 6D) of model rats. Results showed that the levels of cholesterol, HDL, LDL in the Res-med group, and the levels of cholesterol, HDL but not LDL in the Res-high group were signi cantly decreased, compared to the EMs group without such treatment (Fig. 6A-6C). No signi cant changes in TG levels occurred among these groups (Fig. 6D). These data suggested that resveratrol treatment has e cacy to rectify the aberrant lipid pro les in EMs model rats. As the occurrence of EMs had been previously shown to be related to cell adhesion, angiogenesis and apoptosis in a mouse model, we extracted mRNA from lesion tissues to analyze the corresponding molecules such as MMP-2, ICAM-1, VEGF and BCL-2 (27), and evaluated any expression alterations associated with resveratrol. Results showed that the mRNA expression of MMP-2 (Fig. 6E), VEGF (Fig. 6F) and BCL-2 (Fig. 6G), but not ICAM-1 (Fig. 6H) were signi cantly increased in lesion tissues of the EMs group as compared to the Sham group ( Fig. 6E-6H). After resveratrol treatment, the mRNA expressions of MMP-2, VEGF and BCL-2 (Fig. 6G), but not ICAM-1 (Fig. 6H) were signi cantly decreased, compared to the EMs group ( Fig. 6E-6H). These observations indicated that, in addition to the reduction of lesion size upon resveratrol treatment in the model rats, there were also associated decreases in cell invasion and angiogenesis and increased apoptosis.

Resveratrol induces PPARα expression in both HEcESCs and EMs model rats
Resveratrol has been previously shown to stimulate PPARα activation that suppresses the transcriptional activity of metabolic genes involved in energy and lipid metabolism homeostasis in endothelial cells (28,29). We analyzed the mRNA levels of PPARα in HEcESCs and in ectopic endometrial tissues of the model rats upon resveratrol treatment. Results showed that mRNA expression of PPARα was signi cantly increased in either HEcESCs (Fig. 6I) or model rats (Fig. 6J). The protein levels of PPARα were analyzed using the ectopic endometrial tissues of model rats (EMs) and the lesion samples obtained from the model rats were treated with either medium or high dosage of resveratrol. An increased PPARα expression was detected in the lesion tissues of model rats treated with resveratrol, compared to the untreated EMs groups ( Fig. 6K and 6L). These observations suggested that resveratrol treatment had resulted in lesion attenuation in model rats and that apoptosis in HEcESCs might occur via PPARα activation.

Discussion
EMs is a refractory disease that affects approximately 10% of women of reproductive age and up to 50% of women with infertility. Immune de ciency, heightened oxidative stress, and systemic chronic in ammation have been considered as critical facilitators in disease progression (30). The heterogeneity of the disease, having different stages and phenotypes, makes timely and accurate diagnosis of EMs a considerable clinical challenge. In the present study, we evaluated metabolite pro les as risk factors and potential biomarkers for EMs phenotypes. In doing so, we highlight the critical involvement of the lipid metabolism in the progression of EMs. This may help identify patients at risk of developing this disease and aid in treatment decisions based on lipid pro les.
Resveratrol has been shown to inhibit the development of EMs using a nude mouse model where it reduced the invasiveness of eutopic endometrial stromal cells (31). Our study showed that the treatment of resveratrol led to the inhibition of cell proliferation and invasiveness and the promotion of apoptosis in HEcESCs. We further demonstrated that upon resveratrol treatment, glycerolipids such as FA, DG and TG and most phospholipids showed signi cant downregulation, particularly those involved in the cholesterol metabolism and insulin resistance pathways (Fig. 7). Sphingolipids such as SM and Cer have been demonstrated to have inhibitory effects on colon cancer, suppressing cell proliferation (32,33). In our experiments of resveratrol treatment in HEcESCs, increased SM and Cer were also observed along with the inhibition of cell proliferation (Fig. 4A and Fig. 7). These lipids are key components of the plasma membrane and other cellular compartments that integrate into many biological processes such as those of signaling pathways, wound healing and anti-in ammation. Such lipidomic alterations may form a dynamic network contributing to pathologies associated with EMs.
PPARα responds to fatty acid signals derived from dietary lipids, pathogenic lipoproteins or essential fatty acid metabolites and thereby controls both the lipid metabolism and in ammation (34). Resveratrol stimulated PPARα activation has been reported to be associated with an increased phosphorylation of AMPK in human glomerular endothelial cells (35). In our study, lipidomics analysis of HEcESCs treated with resveratrol also showed a signi cant activation of PPARα, probably through an up-regulated AMPK signaling and PGC1 pathway. Resveratrol mediated down-regulation of DAG might directly activate IRS/PI3K-AKT pathways (Fig. 7). PPARα-mediated changes in the FA and AMPK pathways eventually resulted in regulating lipid transport genes, such as ApoA1 and ApoA2 (Fig. 7). Therefore, resveratrol might act as an agonist for PPARα and interplay with lipid-associated pathways, together contributing to recovery from EMs. Further investigation is required to con rm the possible role of PPARα as a molecular target for the treatment of EMs.
As a chronic in ammatory disease, both the onset and recovery phases of EMs may be closely related to the status of lipid metabolites. Lipids can function in tissue remodelling and act to maintain homeostasis during in ammatory processes (36). Lp(α) acts as an acute phase protein with a pro-in ammation role and is active in the modulation of tissue repair in cases of injury (9). ApoA1 has anti-in ammatory properties and also can act as a phase protein involved in wound healing (37,38). Both ApoA1 and ApoB are involved in cholesterol tra c (39). Our study revealed that imbalanced R-ApoA1/ApoB may function as a speci c suppressor of in ammatory responses in EMs cases which is also present as a risk factor to facilitate the abnormal survival of endometrial tissue. In addition, an increased serum Lp(α) level as an independent risk factor for the OMA phenotype may indicate the critical effect of in ammation and proatherosclerosis towards the development of EMs. In the model rats, serum levels of cholesterol, HDL and LDL showed signi cant increases in a lesion size-dependent manner, and subsequent signi cant decreases upon resveratrol treatment. These data manifested a critical involvement of lipid metabolites in EMs and the therapeutic e cacy of resveratrol targeting of the lipid metabolism.

Conclusion
This comprehensive study shows that the development of EMs is strongly correlated to lipid metabolism, and resveratrol may play a therapeutic role by targeting the lipid metabolism of ectopic endometriotic endometrial stromal cells. Our study provides valuable insights for understanding the pathogenesis and clinical treatment of endometriosis.

Participants
Patients who were diagnosed having EMs, requiring surgical treatment and referred to the general gynecology department of the Women's Hospital, Zhejiang University School of Medicine, were included in the study. Written informed consent was obtained from each patient before study inclusion. A total of 205 women with EMs (110 cases) and without EMs (95 cases) were enrolled. Indications for surgery in the EMs group were as follows: pelvic mass, history of infertility, pelvic pain, and failed analgesics. Study inclusion criteria of the EMs group included: (1) Age ≤ 40 years; (2) histologically proven EMs; (3) restriction of samples to those of moderate or severe disease (stages 3 and 4) according to the r-ASRM Classi cation; (4) restriction to OMA or DIE phenotypes (SUP, OMA and DIE phenotypes are frequently associated with each other), with the nal phenotypic diagnosis of EMs designated according to the worst lesion, as per a previous study (6). Exclusion criteria were: (1) women with only a SUP phenotype; (2) irregular menstrual cycles; (3) those with history of metabolite diseases such as diabetes, obesity, cardiovascular disease or thyroid disease etc; (4) those with a history of autoimmune or in ammatory diseases; (5) pregnancy; (6) those having had hormone treatment such as oral contraceptives, GnRH analogues or any other hormone treatment during the previous 3 months before the study.
Women ≤ 40 years old requiring surgical treatment, but without any evidence of EMs, were recruited as controls during the same period. Detailed history, a thorough physical examination of the abdominopelvic cavity and sonography screenings were performed by the designated experts for every patient. Control patients presenting with dysmenorrhea or tenderness in the pelvic area or a mass in ovary or those with history of metabolite diseases such as diabetes, obesity, cardiovascular disease or thyroid disease were excluded.

Measurement of serum metabolites
Weight and height was determined for all patients. The BMI was calculated as weight (kg) divided by the square of height (m 2 ). Venous blood samples were obtained from each patient at baseline after an overnight fast of 12 hours. Cases and controls also received standard laboratory testing. The concentrations of serum fast blood glucose (FBG), TC, total TGs, HDL and LDL were measured using an enzymatic colorimetric assay. Lp(α), Serum ApoA1 and ApoB levels were measured using the immune turbidimetric method. Uric acid (UA) was measured using uric acid enzymatic methods. The metabolite pro les were performed on ABBOTT ARCHITEC c16000 (Chicago). The intra-and inter-assay coe cients of variation for all measurements were 5% and 10%, respectively. The ratios of these markers including the ratio of TG to HDL (R-TG/HDL), the ratio of TC to HDL (R-TC/HDL), the ratio of LDL to HDL (R-LDL/HDL), the ratio of ApoB to ApoA1 (R-ApoB/ApoA1) and the ratio of ApoA1 to ApoB (R-ApoA1/ApoB), were calculated.

Culture of human ectopic endometrial stromal cells
With permission of the patients, lesion tissues from the 8 patients whose intraoperative r-ASRM scores were all endometrial stage 3/4 were sampled under sterile conditions and kept in cold DMEM/F-12 (Gibco) with 1:100 Penicillin-Streptomycin Liquid (Beyotime) for subsequent cell culture. The tissues were digested with 0.2% type I collagenase (Solarbio) in 37 O C for 1.5 hours and then hand ltered using a 70 µM cell lter (BD Falcon). Cells were cultured in DMEM/F-12 with 10% FBS (Gibco) and 1% Penicillin-Streptomycin Liquid in an incubator (Esco) at 37 O C, 5% CO2. The cells were passaged when the density of primary cells had reached more than 75%. Resveratrol (Selleck) was initially dissolved in DMSO (Sangon) to make 20 mM and 8 mM mother uids, and then diluted to the working concentrations of 100 µM and 40 µM, respectively.

Lipidomics analysis
Sample preparation and detection: Eight groups of the cultured primary HEcESCs were divided into two parts. One was treated with 100 µM resveratrol and the other was treated with solvent only. Culturing was for 48 hours where about 1 × 10 6 cells were collected and subjected for the following lipidomic analysis. All samples were prepared according to previously described techniques (22). An UHPLC system was used to coordinate an electrospray ion source using a Q Exactive-HF MS system (Thermo) which was used for lipid pro ling (UPLC-MS). Chromatographic conditions: Flow rate was 0.26 mL/min while column temperature was 55 O C. The mobile phases consisted of (A) 60% acetonitrile/H 2 O with 10 mM ammonium and (B) Isopropanol: acetonitrile = 9:1 (with 10 mM ammonium format). We applied positive and negative mode linear gradients to detect the subjects, respectively. Mass spectrometry was performed using a Thermo Q ExactiveTM benchtop Orbitrap mass spectrometer equipped with heated ESI source in ESI positive and negative modes (Thermo).
Data Processing: All assay raw data were collected using Xcalibur data acquisition software (Thermo). The data, including m/z-values, retention times, and peak areas, were extracted using LipidSearch software (Thermo). All of the detected lipids were quanti ed using the Thermo TraceFinderEFS software (version 3.2). The lipid molecules were named by reference to the LipidMaps website. We enabled One-MAP (www.5omics.com) software to support comprehensive metabolic data analysis. Multivariate statistical analysis was performed online. This included hierarchical clustering analysis, Pearson correlation heat maps, Z-score plot, Volcano plot, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), orthogonal partial least squares discriminant analysis (OPLS-DA), construction of a receiver operating characteristic curve (ROC) univariate, and the construction of a permutation plot and a variable importance in projection (VIP) plot.

Cell proliferation, invasiveness and apoptosis assays
Proliferation: Cultured HEcESCs, grown to the logarithmic growth stage, were digested with 0.25% Trypsin-EDTA and re-suspended. 100 µL of 4 × 10 4 cells/mL suspension was inoculated into 96-well plates (Corning) for 24 hours. Four experimental groups, each with triplets, were prepared as follows: blank (100 µL culture medium); control (DMSO); Re-40 µM (resveratrol at a concentration of 40 µM); Re-100 µM (resveratrol at a concentration of 100 µM). After 48 hours treatment, 10 µL CCK8 (Solarbio) solution was added and incubated for another 4 hours. The cell viability was measured with a BioTek Synergy 1 plate reader (BioTek) and calculated. Modeling: Prior to any surgery, the estrous cycle stages of female rats were examined using vaginal biopsy samples. Attrition cells, showing as irregular keratinocyte like cells and gathered together on the slides, was considered to be an indicator of a mature estrous stage for e cient EMs modeling. Rats having a 4-5 days estrous cycle and two consecutive estrus cycles were then selected for surgery. The animals were anesthetized using 45 mg/kg by intraperitoneal injection of 3% pentobarbital (BIOCAM) sodium and operated under strict aseptic conditions at a room temperature of 28-30 O C. Rat estrus epithelial tissue with a 0.8 × 0.8 cm 2 endometrium was auto-transplanted into the endometrial abdominal wall. Welfare nursing was provided after the operation. Ten rats were also selected for a placebo operation to serve as sham. The animals were fed regularly for 4 weeks.
Examination: A laparotomy was performed 4 weeks after the surgery. The rats were euthanized and the laparotomy was performed to measure size of the implant. Modeled rats were recorded and the lesion volume was calculated using the following formula: V = a × b 2 /2 (a represents the broadest transverse diameter of the lesion, b represents the vertical diameter line) and V ≥ 2 mm 3 was considered as a successful model.

Resveratrol treatment and evaluation
Resveratrol was dissolved in 35% DMSO for intraperitoneal injection in rats, while the Sham group and EMs group were injected with the same amount of the solvent (0.9% NaCl + 35% DMSO). Thirty rats with successful modelling were divided into three groups randomly: EMs group (n = 10), Res-med group (n = 10, resveratrol dose = 15 mg/Kg/d), Res-high group (n = 10, resveratrol dose = 45 mg/Kg/d). The rats of four groups were administered continuously for 28 days. Lesions were examined (as above method). Lesion tissues and blood were sampled before and after resveratrol treatment for evaluation.

HE staining
Lesion tissues were xed in 10% formalin and dehydrated with a gradient of alcohol for para n slicing.
The sections were processed according to a standard protocol for staining with Hematoxylin and Eosin (Solarbio). Images were taken under a light microscope (Nikon) and pathological features were analyzed.
Detection of serum TC, TG, HDL and LDL of rat models Whole blood (500 µL) was collected and centrifuged to detect TC, TG, HDL and LDL and analyzed by fully automatic biochemical analyzer (Toshiba FR120). The following Detection kit (Beijian) were used: total cholesterol measurement kit (CHOD-PAP method); low-density lipoprotein cholesterol measurement kit (direct method-protective reagent method); high-density lipoprotein cholesterol measurement kit (Direct method-selective inhibition method); Triglyceride kit (GPO-PAP).

Western blot
Total proteins were extracted using a standard protein lysis buffer. The protein concentration was determined using a BCA kit (Gene Ray). The standard curve was made according to the absorption value of standard liquid, and the concentration of protein was measured and calculated. Samples were subjected to SDS-PAGE and transferred to a polyvinylidene uoride membrane. Membranes were immunoblotted with rabbit anti-PPARα (1:500, Proteintech) and mouse anti-Actin (1:1000, Goodhere Biotechnology Co, AB-M-M001). Detection of proteins was performed using the ChemiLucentTM ECL detection reagents (Millipore, WBKLS0500). Images were taken using the chemiluminescence imaging system (Clinx Science Instruments).

The open eld assay
The open eld experiment was carried out in a market equipment (open square box, 2 m × 2 m × 50 cm) which was equipped with an infrared camera (CCTVLENS). Experiments were performed in the standard manner. The rats were placed in the test room to acclimatize for 2 hours, and then each one placed in the same orientation when entering into the market. The test time of each rat was 5 minutes. The trajectory and movements of the rats was tracked using Video Track 3.10 software for subsequent analysis of parameters such as the movement time in the central area and the number of entrances into the central area. In between each experiment the eld was cleaned with 70% ethanol to eliminate the odour of the previous rat.

Statistics
Continuous characteristics were presented as Means ± SD for normal variables, and Median (Q1-Q3) for abnormal variables. The difference between control groups and the EMs group for different phenotypes was tested by ANOVA and T-test for normal variables, Kruskal-Wallis test and Wilcoxon test for abnormal variables. Categorical variables were given as N (%) and the chi-square test was applied to compare the distributions across different groups. Multivariable logistic models were applied to assess the association of Metabolite indicators with EMs in general, and with its two phonotypes more speci cally.
As potential risk factors, all above interesting metabolite markers, plus general information including age, BMI, history of delivery and abortion, were entered into the initial model. A step-wise selection method, with a signi cance level of 0.05 required to allow a variable into the model (SLE = 0.05), and 0.10 for a variable to remain in the model (SLS = 0.10), was then adopted to identify the nal model which contains the best subset of the potential risk factors. Finally, for each single variable, as well as for the union of variables in the nal model, ROC curves were performed to determinate the diagnostic value where Youden index (sensitivity + speci city-1) was used to select best cut-off point. Statistical analyses were conducted using SAS, version 9.4 (SAS Institute, Cary, NC). The difference between the two groups was analyzed using a T test (Graphpad Prism5). Differences were considered signi cant at a p-value of < 0.05, marked *P < 0.05, **P < 0.01, ***P < 0.001. The experimental and required specimen collections were reviewed and approved by the Ethics Committee of Women's Hospital School of Medicine Zhejiang University, and all the participants signed their written informed consent before the study. All animal experiments were approved by the Zhejiang University Experimental Animal Welfare Ethics Review Committee.

Consent for publication
The manuscript has been approved by all authors.

Availability of data and material
The datasets used or analysed during the current study are available from the corresponding author on reasonable request.