Decreased SFRP5 Correlated With Excessive Metabolic Inammation in Polycystic Ovary Syndrome Can Be Reversed by Metformin: A Retrospective Case-Control Study

Background: Polycystic Ovary Syndrome (PCOS) is a complex endocrine disorder of heterogeneous nature. Secreted frizzled-related protein (SFRP) 5 is an anti-inammatory adipokine implicated in metabolic homeostasis. We aimed to conrm the correlation between SFRP5, metabolic inammation and PCOS, investigate the predictive value of SFRP5 for PCOS and the involvement of SFRP5 in metformin treated PCOS. Methods: This retrospective case-control study included PCOS and control women (67 vs. 33) for detecting serum SFRP5 and its correlation with metabolic inammation. Predictive value of SFRP5 for PCOS was evaluated by logistic regression and receiver operating characteristic (ROC) analyses. The effects of metformin on SFRP5 and pro-inammatory cytokines and ovulation of PCOS with metabolic abnormality (37 vs. 36) were analysed. Results: Plasma SFRP5 levels were decreased in PCOS (odds ratio: 0.78, 95% condence interval (CI):0.703–0.866, P<0.001) independent of obesity. SFRP5 was negatively associated with IL-6, TNFα, FAI and HOMA-IR. The cut-off point of SFRP5 < 46.13 ng/ml was optimal to identify PCOS with a higher specicity of 96.87% and a relatively lower sensitivity compared to AMH. SFRP5 increased specicity of AMH for predicting PCOS, especially which with relatively decreased AMH (< 4.7 ng/ml). Metformin promoted SFRP5 and decreased leptin, IL-6 and TNFα secretion in PCOS women with metabolic abnormality in a time dependent manner and with improved ovulation rate and pregnancy outcomes. Conclusion: Decreased SFRP5 was associated with metabolic inammation in PCOS and has a potential role for the supplement of AMH in predicting PCOS. The reverse of serum SFRP5 by metformin indicated that SFRP5 participated in the improvment of follicular development by metformin. Further prospective investigations are needed to conrm these preliminary data.


Background
Polycystic ovary syndrome (PCOS) is a complex endocrine disorder characterized by oligo-anovulation (OA), polycystic ovarian morphology (PCOM), hyperandrogenism (HA) and accompanied by metabolic aberrations (adverse lipid pro le, impaired glucose tolerance (IGT), insulin resistance (IR) and type 2 diabetes mellitus (T2DM)). The prevalence of PCOS varies from 6% to 10% according to different diagnostic criteria, making PCOS the most common endocrine disorder among women of reproductive age [1].
The limited knowledge of its etiology and the heterogeneous nature make the de nition, diagnostic criteria and treatment of PCOS to be controversial issues for many years [2][3]. Approximately half of the PCOS women exhibit obesity. Besides, about 30 to 50% of lean PCOS women still have an increased risk of metabolic abnormalities [4,5]. In addition, there are also a portion of PCOS women without abnormal glycolipid metabolism exhibiting HA and ovarian dysfunction. Although biomarkers such as anti-Müllerian hormone (AMH) have been suggested for diagnosis of PCOS, none of them could be used independently [6]. Appropriate recognition and management of PCOS remain challenges.
Increasing evidence supports that chronic low-grade in ammation correlats with HA and IR and emerges as a key contributor to the pathogenesis of PCOS [7]. Secreted frizzled-related protein (SFRP) 5, a member of the Wnt inhibitor SFRP family, is recently identi ed as an anti-in ammatory adipokine implicated in metabolic homeostasis [8]. Wnt signaling plays critical role in reproduction especially in the regulation of follicle maturation and is involved in the pathogenesis of PCOS [9][10]. SFRP5 is found related to the maturation of oocytes and regulation of granulosa cells [11]. Metformin is extensively evaluated in the treatment of PCOS because of its insulin-lowering effect. Moreover, it can also relieve chronic low-grade in ammation, inhibit ovarian steroidogenesis (including androgen production), and increase ovulation rate of PCOS women [12][13][14][15]. However, few studies supported the relationship between SFRP5 and PCOS [16][17]. The correlation of SFRP5 and metabolic in ammation in metformin treated PCOS women is unclear.
The purpose of this study was to con rm the correlation of SFRP5, metabolic in ammation and PCOS, investigate the predictive value of SFRP5 for PCOS. More importantly, we evaluated the effect of metformin on SFRP5, pro-in ammatory cytokines and ovulation rate in PCOS women with metabolic abnormalities to disclose the role of SFRP5 in metabolic in ammation of PCOS.

Study participants
This retrospective study included PCOS and control women of Chinese Han population visiting the reproductive medical center of the First A liated Hospital of Chongqing Medical University from January to July 2018. PCOS patients ful lled the Rotterdam criteria containing at least two of the following three characteristics: (1) OA, (2) clinical and/or biochemical signs of HA, and (3) PCOM. Exclusion criteria were other etiologies for HA. Women with regular menstrual cycles were recruited as the controls. None of them had any known disease, were smokers or took any hormonal or other insulin-modifying therapy during the preceding 2 months. The PCOS women were divided into two subgroups according to the thresholds of body mass index (BMI) and AMH [6] indicated by previous studies: (1) BMI <25 and BMI ≥25; (2) AMH < 4.7 ng/ml and AMH ≥4.7 ng/ml. The serum level of SFRP5 and anthropometric and biochemical parameters of PCOS were detected and their association were analysed .
The other PCOS women complicated with IGT/IR and without other infertile factors and accepted no treatment in the previous six months were included for analyzing the role of SFRP5 in metformin treated PCOS and were divided into two groups. Group I was pretreated with metformin (1,000-1,500 mg/day; Conquer, Chongqing, China) for at least three months. Group II accepted no metformin treatment. Both groups were treated with clomiphene (150 mg/day for 5 consecutive days on menstrual cycle day 5; Codal-Synto, Limassol, and Cyprus). Human chorionic gonadotropin (hCG), 10,000 U (Livzon, Zhuhai, China) was administered intramuscularly whenever the diameter of the dominant follicle reached ≥18 mm and the estradiol level reached 150-200 pg/mL/follicle. Regular sexual intercourse was advised 24 h after the hCG day and afterward. If there were more than 3 dominant follicles, contraception was requested. The ovulation rates per cycle, days for follicular development, endometrial thickness, number of dominant follicles (>14 mm), serum estradiol on the day of HCG, accumulated HCG positive rate, accumulated clinical pregnancy rate, miscarriage rate and multiple pregnancy rate were investigated.

Anthropometric and biochemical measurements
Plasma glucose and hemoglobin A1c (HbA1c) levels were measured by the glucose-oxidase method and anion-exchange high-performance liquid chromatography, respectively. Total cholesterol, high-density lipoprotein cholesterol (HDL-C), triglycerides (TGs), and low-density lipoprotein cholesterol (LDL-C) were determined enzymatically using an auto analyzer (Hitachi, Tokyo, Japan). Serum insulin, total testosterone, follicle-stimulating hormone (FSH), luteinizing hormone (LH) and sex hormone-binding globulin (SHBG) levels were assayed by an automated chemiluminescence system on the UniCel Dxl 800 (Beckman Coulter, Fullerton, CA, USA) between day 2 and 5 of the menstrual cycle. Insulin resistance index (HOMA-IR) and Free androgen index (FAI) were calculated as following respectively: fasting glucose Oral glucose tolerance test (OGTT) The baseline blood samples were obtained after an overnight fasting of 12 h. Then blood samples were obtained at 0.5, 1 and 2 h for the measurements of glucose and insulin after dranking 75 g of glucose.
Measurement of human SFRP5, TNFα, IL-6 and leptin serum concentrations Fasting blood samples were analyzed for SFRP5, TNFα, IL-6 and leptin after 0, 1 and 3 months of metformin treatment in Group I using enzyme-linked immunosorbent assays (ELISAs) (USCN Life Science and CUSABIO, Wuhan, PR China) with a sensitivity of 0.60 ng/ml, 1.95 pg/ml, 1.95 pg/ml and 0.06 ng/ml, respectively.

Statistical analysis
All statistical analyses were performed using SPSS 19.0 software (SPSS Inc., Chicago, IL, USA) and MedCalc 18.11.3 software (Acacialaan, Ostend, Belgium). Normally and non-normally distributed continuous data are presented as the mean±standard deviation (SD) or medians (25th and 75th percentiles). Subjects with missing data were excluded. For normally distributed continuous data, comparisons among groups were performed by ANOVA and an unpaired t test as appropriate. For nonnormally distributed continuous data, we used Kruskal-Wallis ANOVA followed by pairwise comparisons or a Mann-Whitney U test to compare differences in the variables in normal controls and different PCOS subgroups and a Dunn-Bonferroni test for post hoc comparisons. Partial correlations between SFRP5 and other variables were calculated by controlling for the covariates. Simple and multivariable linear regression analyses examined the associations of SFRP5 with other variates.
Stepwise logistic regression analyses were adjusting for age, BMI, AMH, AUCg, AUCins, HOMA-IR and FAI were conducted to assess the effects of SFRP5 in the diagnoses PCOS. The variables were log transformed to conform to normal distribution assumptions. The Box-Tidswell transformation (x*ln(x)) of the predictor x was used to assess the linearity in the logit. A receiver operating characteristic (ROC) analysis was performed to assess the diagnostic value of SFRP5 as a marker of PCOS. The maximum Youden index was used to determine the optimal sensitivity and speci city, as well as the corresponding cut-off value. P values of less than 0.05 were considered statistically signi cant.

Circulating SFRP5 levels and its association with clinical and biochemical characteristicsin PCOS
A total of 91 PCOS patients and 108 control women satis ed the inclusion criteria. However, only 67 PCOS and 33 control women with consent and available blood samples were enrolled. Thirty-one of PCOS were with PCOM, HA and OA, twenty-three with HA and OA and thirteen with PCOM and OA.
No obvious difference of SFRP5 level was found between normal weight (BMI<25) and overweight /obese (BMI ≥25) PCOS patients, although BMI, WHR, HOMA-IR and FINS levels were signi cantly higher in the latter. Furthermore, SFRP5 was still lower in normal-weight PCOS patients than BMI-matched control patients, P<0.001.

Metformin promoted SFRP5 and decreased leptin, IL-6 and TNFα protein secretion in PCOS women with metabolic abnormalities
There was no difference of baseline demographic and clinical characteristics between Group I (N=37) and Group II (N=36) ( Table 5). After treated with metformin for 1 and 3 months, serum SFRP5 of Group I signi cantly increased than that before administration (45.60±8.148 and 63.92±8.24 vs 32.16±5.71 ng/ml, P<0.001). Conversely, the secretion of leptin, IL-6 and TNFα were decreased after 1 and 3 months' treatment (P<0.001) (Figure 3). 4. Pretreatment with metformin had a positive effect on the ovulation rate and pregnancy outcomes of PCOS women with metabolic abnormalities Before clomiphene treatment, the parameters of glucose metabolism in Group I were all adjusted to normal by metformin (data not shown). The ovulation rate, number of dominant follicles (P<0.001), serum estradiol level on the day of HCG administration and accumulated clinical pregnancy rate (P<0.05) in Group I was signi cantly higher than that of Group II. The duration for follicular development in Group I was signi cantly shorter than that of Group II (P<0.01). There was no statistical signi cance in endometrial thickness, accumulated HCG positive rate and miscarriage rate between the two groups (Table 6).

Discussion
As an anti-in ammatory adipokine, SFRP5 has been identi ed implicated in metabolic homeostasis. In this study, SFRP5 levels were negatively associated with incidence of PCOS, HA, IR and in ammation independent of obesity. Plasma SFRP5 level might have a potential role for the supplement of AMH in predicting PCOS. Metformin promoted SFRP5 protein secretion and decreased pro-in ammatory factors in PCOS women with improved ovulation rate and pregnancy outcomes.
Consistent with the ndings of Hu, we found serum SFRP5 levels were decreased in PCOS in Chinese population [17], which was opposite to the ndings of Almario in American population [16]. One difference was that we used Rotterdam criteria to include PCOS women, whereas Almario used the NIH criteria. The other difference was that the reference women reported by Almario were fatter (BMI: 29.0 ± 6.3 kg/m 2 ) than our Chinese controls (BMI: 20 (19.8, 21)). And they were older and with higher cholesterol level than their PCOS women. Because Chinese women have an increased risk of metabolic problems at a low BMI (≥24) [18], the control women included in our study have a relatively lower BMI. As BMI can not discriminate between fat mass and muscle mass and cannot represent abdominal adiposity, we detected WHR and found no relationship between SFRP5 and WHR. Consistent with the ndings of Almario and other study [19], serum SFRP5 level was independent of obesity. Intensive exploration showed SFRP5 conditionally exerted an anti-in ammatory effect in response to an obesity-associated microenvironment.
So the microenvironment rather than the mass of adipocytes plays a more important role in the regulation of SFRP5.
We detected higher level of IL-6 and TNF-and FAI in PCOS, which were correlated with SFRP5. In ammation is closely interrelated with HA in ovarian microenvironment. HA stimulated cumulus cells secreting more in ammatory cytokines as well as oocyte maturation. In turn, the direct exposure of ovarian theca cells to pro-in ammatory stimuli also increases androgen production [20]. Our ndings suggest SFRP5 is involved in the interaction of in ammations and HA. We demonstrated that FINS, 2-h INS, AUCI and HOMA-IR were signi cantly higher in PCOS, which indicated the association of IR and PCOS. Moreover, 2-h INS and 2-hBG were higher in normal-weight PCOS women than normal control indicating the metabolic aberration independent of obesity in PCOS. Serum SFRP5 was negatively correlated with FINS, 2-h INS, AUCI and HOMA-IR. The negative correlation between SFRP5 and IR has been determined in T2DM patients and animal models [21]. Caloric restriction improved insulin sensitivity and was associated with a signi cant increase of SFRP5. Proin ammatory cytokines and mediators closely correlate with insulin resistance through interfering with insulin signal transduction. Whether SFRP5 responds to hyperinsulinemia or the in ammatory conditions accompanied with IR need to be considered.
AMH, correlating with PCOM, has been proposed as a biomarker for the diagnosis of PCOS. However, AMH can be in uenced by HA, obesity and in ammation. Moreover, the con icting results of AMH's effect on the growth of primordial follicles [22] and the heterogeneity of the PCOS phenotypes led to current controversy on the value of AMH for predicting PCOS among an unselected group of women. Our results suggest that SFRP5 could predict PCOS with a higher speci city in total PCOS population. However, the sensitivity of SFRP5 was lower than AMH. Combination of AMH and SFRP5 improved the sensitivity and speci city of the two biomarkers individually. In the subgroup of PCOS with lower AMH (< 4.7 ng/ml), SFRP5 still had a higher speci city. We proposed that SFRP5 worked as a speci c biomarker for PCOS as complement to AMH especially in phenotype without PCOM.
The current therapeutic options for infertile PCOS women using clomiphene citrate (CC) et al are always associated with drug resistance, especially in PCOS women with metabolic abnormalities. The effects of metformin on PCOS related IR, in ammation, ovarian steroidogenesis and ovulation have been con rmed. We found that metformin promoted SFRP5 protein secretion and decreased leptin, IL-6 and TNFα protein secretion of PCOS women in a time dependent manner. Meanwhile, our results identi ed that metformin combined with CC had positive effects on the ovulation rate and pregnancy outcomes of PCOS women with metabolic abnormalities. Accordingly, a similar conclusion was drawn in the 2017 Cochrane Review [23]. Our results suggest that SFRP5 related anti-in ammatory process is involved in the improvement of ovulation of PCOS by metformin.
Conclusions SFRP5 was decreased in PCOS and associated with metabolic in ammation. SFRP5 has a potential role for the supplement of AMH in predicting PCOS. SFRP5 participates in the anti-in ammatory effect of metformin in improving follicular development. As this is a retrospective study with limited cases, a welldesigned randomized controlled intervention trial is required to eliminate potential bias. And the molecular mechanism of SFRP5's anti in ammatory role in PCOS should be further investigated in vitro and in animal model.

Availability of data and materials
The data generated during the current study are available from the corresponding author on reasonable request.