A total of 16 women were enrolled in this study. Eight women suffered from PCOS, while the remaining eight were normal control (NC) women. The subjects’ average age and BMI were statistically identical ( Table 1 ). The study was approved by the ethics committee of Shanghai Tongren Hospital, Shanghai Jiao Tong University School of Medicine. All patients provided informed consent. The Rotterdam criteria (18), which include polycystic ovaries, oligo- or anovulation, clinical and/or hyperandrogenism, and exclusion of other causes of hyperandrogenism like androgen-secreting tumors, non-classical congenital adrenal hyperplasia, hyperprolactinemia, and Cushing’s syndrome, were used to diagnose PCOS. When two or more of the three requirements were met, the diagnosis was deemed accurate. The women in the non-PCOS group had regular menstrual cycles every 26–34 days with no signs of hyperandrogenism. Additionally, none of the study participants had used any medications known to impact metabolic or hormonal characteristics within the 3 months prior to the study.

Subcutaneous fat of the abdomen (PCOS, 5; NC, 8) was sampled from women with tubal factor infertility using laparoscopy. Three patients dropped out of the trial. When performing the laparoscopy, all patients were fasting, and the surgeon removed 3–4 g of subcutaneous fat. Biopsies were immediately rinsed in cooled NaCl 0.9% solution, segmented, and snap-frozen as previously reported (9, 18). Until analysis, samples were kept frozen in liquid nitrogen (196°C).

Around 20 mg of subcutaneous fat of the abdomen was added to 500 μl methanol solution (containing 5 μg/ml L-2-chlorophenylalanine as internal standard) and homogenized for 2 min. The supernatant was centrifuged at 13,000 rpm at 4°C for 10 min, and 200 μl was obtained. The same volume of serum was extracted from all samples and mixed evenly to prepare QC (quality control) samples.

The UHPLC system was coupled to Orbitrap/MS (Waters Corp., Milford, MA, USA), equipped with an electrospray ionization source, and operated in positive or negative ionization modes with a mass resolution of 70,000 and an m/z of 200. Using data correlation (dd-MS2, TopN = 10) MS/MS mode, the full-scan quality resolution was 17,500 when m/z was 200, and the scanning range was 100–1,500. The chromatographic conditions were as follows: sample size was 2 μl, column temperature was 25°C, flow rate was 0.35 ml/min, and mobile phases were 0.1% formic acid aqueous solution and 0.1% acetonitrile formic acid solution. The optimized chromatographic gradient was as follows: 0–2 min, 5% in liquid B; 2–10 min, 5%–95% in liquid B; 10–15 min, 95% in liquid B; 15–18 min, 5% in liquid B. Data were obtained in centroid mode using Thermo Excalibur 2.2 software from Thermo Fisher Scientific, Massachusetts, USA.

Serum sample collection was referred to Yan et al. (19). Shanghai Lu-Ming Biotech Company Limited (Shanghai, China) provided an experimental platform and assistance for the targeting amino acid metabolomics analysis. Briefly, a mixture of methanol/water (4:1 by volume) was used to collect 2 × 10⁷ per sample. The sample was quickly stored in liquid nitrogen. Before testing on the machine, the sample was equilibrated to ambient temperature for 30 min. The sample was dispersed using the ultrasonic lysis method. It was then concentrated and centrifuged and then freeze-dried. Finally, a mixture of BSTFA and n-hexane (4:1 by volume) was added to the sample, vortexed vigorously for 2 min, and derivatized at 70°C for 60 min. These samples were analyzed by a gas chromatography system (Thermo Fisher Scientific TSQ 9000, USA). UPLC–ESI–MS/MS was utilized as the analytical method for the quantitative detection of targeted amino acid metabolites.

A steroid-derived human granulosa-like tumor cell line (KGN cell line) was selected for this study because it retains the physiological properties of ovarian granulosa cells (20). In the DMEM/F-12 medium including 10% fetal bovine serum (FBS, Gibco), KGN cells were grown in a humid atmosphere at 37°C and 5% CO₂. The concentrations of metabolites used are listed in the legend. DHEA was added to the medium 4 h prior to metabolite treatment.

The effects of metabolites were assessed using the Cell Counting Kit-8 assay (CCK-8, Beyotime, China). KGN cells were reseeded in 96-well plates at 5,000 cells/well. Each well received 10 l of CCK-8 reagent. Cells were then incubated for 1 h at 37°C. Absorbance was measured at a wavelength of 450 nm using a microplate reader (Thermo Fisher Scientific, USA). Cells were assessed at 0, 24, and 48 h.

A flow cytometry study of the Annexin V-fluorescein isothiocyanate (FITC) against propidium iodide (PI) assay was used to identify apoptosis (BD Pharmingen, CA, USA). A total of 3 × 10⁵ KGN cells were reseeded in each well of six-well plates. Cells were trypsinized, washed twice with cold phosphate-buffered saline for 15 min at room temperature, and used to incubate cells with 5 µl of Annexin v-FITC and 5 µl of PI each well.

Serum samples extracted from KGN culture supernatants were tested in duplicate using testosterone (T₀), estrogen (E₂), and progesterone (P₄) ELISA kits (CUSABIO, China) in accordance with the directions provided by the manufacturer.

Using the appropriate kits, adenosine triphosphate (ATP) concentrations were assessed following the extraction and quantification of proteins from KGN (Beyotime, China).

Data were obtained using the Thermo Xcalibur 2.2 software (Thermo Scientific, SAN Jose, USA). Peak calibration and extraction were performed using the Compound Discovery Software (Thermo Fisher Scientific). Data tables were imported into Simca-P 13.0 for multivariate statistical analysis. Unsupervised principal component analysis (PCA) was used to assess the overall trend of separation between these samples. Differential metabolites were screened using partial least square, discrimination analysis (PLSDA). According to the PLSDA model, variables whose importance was greater than 1.0 in the projection (VIP) value were selected, and SPSS Statistics 18.0.0 was used for the two-tailed Student’s t test, and a p-value < 0.05 was considered statistically significant. Multiple test adjustments were made using Bonferroni correction. To identify these potential biomarkers, the accurate ion mass was entered into the Human Metabolome Database (HMDB), Metlin, MoNA, and MassBank databases to match accurate molecular weight and automatically search for MS1/MS2 fragment ions. Finally, to determine the structure of the compound, we used our internal standard metabolite library, matching the exact mass, fragment ion mass, and retention time. Furthermore, metabolic pathway enrichment analysis was performed using the KEGG database (https://www.metaboanalyst.ca/). The area under the receiver operating characteristic (ROC) curve, i.e., AUC, was used to evaluate the diagnostic ability of biomarkers. The metabolite interaction network analysis was conducted using the IPA online database.

Patients’ AT was drawn for metabonomics analysis to examine the metabolic profiles of the NC and PCOS groups. The findings of the PCA for the NC and PCOS groups ( Figures 1A, B ) revealed a pattern of aggregation within groups and segregation between groups. The plots of the PCOS group were very different from those of the NC group, according to the PLS-DA model, which revealed significant metabolite differences between the two groups ( Figures 1C, D ). Additionally, a continuing analysis of the PLS-DA model was carried out ( Figures 1E, F ). In the PCOS group, a total of 107 distinct metabolites were found (VIP >1, p < 0.05).

Figure 2A illustrates the results as a volcano plot of all metabolites. Red and green colors were used to represent elevated and decreased metabolites, respectively. Figure 2B depicts a heat map of the various metabolites identified by LC–MS. The differential metabolites could be loosely divided into various super-classes as per the enriched metabolite set clustering analysis ( Figure 2C ), such as organic acids, fatty acyls, nucleic acids, and main classes such as amino acids and peptides ( Figure 2D ). The biomarker pathway enrichment analysis used the KEGG and MetaboAnalyst databases. According to Figures 2E, F , these differential metabolites were mainly enriched in inflammatory diseases, beta-oxidation, mitochondrial beta-oxidation, etc. It is well known that PCOS is an inflammatory, systemic, lifestyle endocrinopathy (21). The mitochondrion is a vital organelle that regulates energy production necessary for cellular survival, and mitochondrial malfunction has been implicated as a possible pathogenesis-inducing factor for PCOS, according to a previous study (22–24). Consequently, the experiments confirmed that the altered metabolites might play important functions in PCOS.

Research has shown that granulosa cell death may be the fundamental cause of follicular atresia. Granulosa cell proliferation and apoptosis determine the fate of a follicle (25). The levels of plasma amino acids in PCOS women have been demonstrated to be significantly out of balance (26), given that L-tyrosine and L-leucine may recover regular menstrual cycle and ovulation in PCOS (27). Hence, we chose amino acids such as DL-tryptophan, L-lysine, L-histidine, L-tyrosine, L-phenylalanine, and isoleucine as candidate metabolites to explore their potential biological roles. The ROC analysis uncovered that DL-tryptophan, L-lysine, L-histidine, L-tyrosine, L-phenylalanine, and isoleucine demonstrated a strong capacity for PCOS diagnosis, with AUC values of 0.900, 0.825, 0.900, 0.850, 0.925, and 0.925, respectively ( Figures 3A–F ). This suggested that the amino acids may assist in making a clinical diagnosis. Further studies need to be conducted to confirm the functions of the amino acids. The primary mechanism of follicular atresia in PCOS was found to be increase in granulosa cell apoptosis (28). Cell viability assay ( Figures 4A–F ) was performed, which revealed that L-phenylalanine and isoleucine could significantly promote the proliferation of cells in cultured KGN ( Figures 4E, F ), and the two amino acids were chosen for an in-depth study.

KGN cells were pretreated with DHEA to imitate the pathophysiological state of PCOS and simulate the hyperandrogenic milieu in order to further investigate the physiological roles of the two amino acids (29). L-Phenylalanine ( Figure 4G ) and isoleucine ( Figure 4H ) significantly reversed the growth inhibition of DHEA-treated KGN cells. To delve deeper into this finding, apoptosis of KGN cells was discovered using flow cytometry. Isoleucine significantly recovered the apoptosis of DHEA-treated KGN cells but not in the L-phenylalanine group ( Figure 5 ).

A creative pathway analysis was performed on various metabolites (IPA), which identified numerous linked signaling pathways, including the ones that have been shown to be closely related to PCOS such as pathways related to mitochondrial dysfunction, ROS, PI3K/AKT, MAPK, and Wnt signaling ( Figure 6A ). The mitochondrion is a crucial organelle that controls energy production required for cellular survival. According to earlier reports (22), PCOS pathogenesis could be exacerbated by mitochondrial dysfunction (23, 24). After DHEA treatment, mitochondrial function indicator ATP level decreased markedly ( Figure 6B ), indicating the possibility of mitochondrial malfunction. Isoleucine replenishment reduced the DHEA-induced mitochondrial dysfunction ( Figure 6B ). Additionally, ELISA was used to measure the T₀, E₂, and P₄ in each group. Compared with the control group, the DHEA group had higher T₀, E₂, and P₄ concentrations. However, compared with the DHEA group, the levels of T₀ and P₄ but not E₂ were suppressed in the DHEA+ isoleucine group ( Figures 6C–E ). Consistently, we deduced that DHEA could lead to mitochondrial dysfunction and sexual hormone disturbances in granulosa cells, which were ameliorated by isoleucine treatment. To further clarify the role of isoleucine in PCOS, we detected the levels of intracellular amino acids by GC–MS/MS-targeted amino acid metabolism analysis. Notably, the reduced level of isoleucine could be significantly detected in PCOS ( Figure 7A ). Taken together, these results indicated that isoleucine was the potential target for PCOS treatment.