From August 2021 to October 2021, 26 patients with PCOS and 26 control individuals were recruited from the Reproductive Center, Shanghai First Maternity and Infant Hospital (Shanghai, China). Blood samples were obtained from each subject to determine the levels of FGF19 using the human FGF-19 ELISA Kit (ImmunoDiagnostics Limited, Shenzhen, China) according to the manufacturer’s instructions. Subsequently, the correlation between FGF19 and PCOS-related indices was analyzed. PCOS-related indices refer to key clinical and biochemical parameters closely associated with the diagnosis and pathophysiology of PCOS, including reproductive hormones [FSH, LH, estradiol (E2), prolactin (PRL), progesterone, and testosterone] and anti-Müllerian hormone (AMH), which reflect ovarian reserve, hormonal imbalance, and follicular development characteristics in PCOS patients (Li et al., 2022; Kulkarni et al., 2023).

Additionally, 3 PCOS patients undergoing IVF treatment and three controls were recruited to collect follicular fluid samples for subsequent proteomics. The inclusion criteria were as follows: three patients with PCOS and three control individuals who underwent oocyte retrieval at our reproductive center in September 2024 were randomly selected, were under 35 years of age, and underwent first-time oocyte retrieval. Participants aged >35 years, those with a history of oocyte retrieval, and those with a history of recurrent miscarriages were excluded. The participant characteristics, including demographic and clinical features, are presented in Table 1. The study protocol was reviewed and approved by the Institutional Ethics Committee of Shanghai First Maternity and Infant Hospital, and all participants provided informed consent following the national laws and Declaration of Helsinki guidelines.

The follicular fluid samples from each group (n = 3) were sent to Jiahua Yaorui Technology Co., Ltd. (Beijing, China) for proteomic analyses. Briefly, the follicular fluids were thawed at 4 °C, and 100 μL of follicular fluid was incubated in 100 μL of beads (1 mg/mL). After binding at 37 °C for 1 h, 200 μL of wash buffer (10 mM tris(hydroxymethyl)aminomethane [Tris], pH = 7.4; 150 mM potassium chloride; 0.05% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate [CHAPS]) was added. Based on the methods of Liu et al. (Li et al., 2022), the bead incubation, reduction, and alkylation were performed according to the methods described by Li et al. (2022). To the bead precipitate, lysis buffer (0.5% sodium deoxycholate monohydrate [SDC] (w/v), 10 mM Tris (2-carboxyethyl)phosphine hydrochloride [TCEP], 40 mM calcium acetylacetonate [CAA], 50 mM tetraethylammonium bromide [TEAB]) was add. After sonication for 10 min and centrifugation, the supernatant was transferred to a new tube and this step was repeated once. Supernatants from both steps were combined and heated at 95 °C for 10 min in a metal bath. For digestion, 0.5 μg of trypsin (Promega, United States) was used, and cultivated for 16 h at 37 °C. Thereafter, the digested peptide sample was desalted using a 96-well plate (Model PLATE NBE ATLAS 96-Well 2.5 mg) following the instructions provided with the desalting kit.

The prepared samples were used for liquid chromatograph-mass spectrometer (LC-MS) using the Vanquish Neo UHPLC system (Thermo Fisher Scientific, United States). Following dissolution in 0.1% (v/v) aqueous formic acid, peptide samples (200 μg) were injected onto a C18 pre-concentration column (Acclaim PepMap 100, 5 μm, 100Å, Thermo Fisher) for desalting and concentration, as well as then chromatographically resolved using a laboratory-prepared analytical column (17 cm in length, 75 μm in internal diameter). The temperature of sample injector and column was respectively 8 °C and 35 °C. The flow rate was 0.25 mL/min, and the injection volume was 2 μL. The mobile phase consisted of solvents A (0.1% formic acid in water) and B (80% acetonitrile/0.1% formic acid). The gradients of chromatographic separation were as follows: 6.7%–24% B for 11.3 min; 24%–36% B for 5.7 min; 36%–55% B for 1.5 min; 55%–99% B for 0.5 min; as well as 99% B for 5 min. The samples were then analyzed by high-resolution mass spectrometry (Orbitrap Astral, Thermo Scientific) using positive-ion mode electrospray ionization at 2.2 an applied potential. The MS1 scan range was 380–980 m/z, with a resolution of 240,000. The DIA MS2 isolation range was 380–980 m/z, with an isolation window width of 2 m/z, a scan time of 3 ms, an MS2 scan range of 150–2000 m/z, and a normalized collision energy of 25. The automatic gain-control (AGC) setting were 5e5.

Raw mass spectrometry data were converted to mzML format using MSConvert (v3.0.21072) and subsequently analyzed using DIA-NN software (v1.8). The human database was downloaded from Uniport _Homo_sapiens_UP000005640.fasta with 20588 entries. A library-free quantity was used for the database search. Trypsin was used for enzymatic digestion, allowing for no more than two missed cleavages and peptide lengths of less than 7 amino acids. Cysteine carbamidomethylation was among the fixed modifications, whereas methionine oxidation and N-terminal acetylation were considered variable modifications. The mass tolerance for both MS1 and MS2 was set to 20 ppm, with a false discovery rate (FDR) set at 1%, and the match between run (MBR) features was enabled. After searching the sequence database, we used Python (v3.10.4) for quantification data preprocessing, including normalization and missing-value imputation. After that, differential expressed proteins (DEPs) were identified using R software (v4.2.2), with the thresholds of fold change (FC) ≥ 2.0 or FC ≤ 0.5, and p-value <0.05. After that, the identified DEPs were subjected for functional analyses, including Kyoto Encyclopedia of Genes and Genomes (KEGG) as well as Gene Set Enrichment Analysis (GSEA).

Twenty female C57BL/6 mice were procured from Sipefu Biotechnology Co., Ltd. (Beijing, China). These mice were housed in an environment where the temperature was 20 °C–25 °C as well as the humidity was 50 ± 5%, and they were exposed to a 12-h light/dark cycle. Throughout the experiments, mice had unrestricted access to food and water. Following a 3-day acclimation period, all mice were randomly assigned to the control and PCOS groups, with ten mice in each group. Mice in the control group received a daily subcutaneous injection of 0.2 mL of sesame oil at the neck nape and were fed a standard diet. However, the mice in the PCOS group were injected with dehydroepiandrosterone (DHEA, 60 mg/kg/d) in 0.2 mL sesame oil subcutaneously on the nape of the neck daily and fed a 60% high-fat diet.

Starting on experimental day 10, the estrous cycle of mice was monitored daily for 10 consecutive days. Vaginal swab smears were obtained at a fixed time each morning. An appropriate amount of 0.9% normal saline was absorbed using a Pasteur pipette and gently flushed into the vaginal orifice to collect the secretions. The samples were then transferred onto glass slides, stained using a rapid, non-toxic, modified Papanicolaou (PAP) staining kit (Jiancheng Bioengineering Institute, Nanjing, China), and observed under a light microscope to determine the estrous phase based on cellular morphology.

Prior to the experimental endpoint, the mice were a 16-h fasting period (from 5:00 p.m. on the previous day to 9:00 a.m. on the test day) with free access to water for oral glucose tolerance tests (OGTT). Briefly, fasting blood glucose levels were measured in tail vein blood samples at time point 0 (fasting baseline). Subsequently, a glucose solution (20% w/v, 100 μL/10 g) was administered by oral gavage; as well as post-administration glycemic profiles were determined at 30-min intervals up to 120 min via tail vein blood draws. Glucose tolerance was assessed by analyzing blood glucose dynamics over time.

After completion of the experiment, blood samples were collected from each mouse. After sacrifice by cervical dislocation, the ileum, liver, and ovarian tissues were collected from control and PCOS mice for further experiments.

Ovarian tissues were immersed in 4% paraformaldehyde (PFA) and fixed for 24 h. Subsequently, they were dehydrated using a sequence of ethanol solutions of increasing concentrations (70%, 80%, 90%, 95%, and 100%), cleared with xylene, and finally embedded in paraffin. After slicing into 5-μm-thick sections, they were firstly stained with hematoxylin for 5 min, followed by eosin for 3 min. After mounting with neutral resin, stained slides were observed and imaged under a light microscope (Nikon Corporation, Tokyo, Japan).

Ovarian tissue sections were baked at 60 °C for 1 h, dewaxed in xylene, and dehydrated using an ethanol gradient. After rinsing with PBS, antigen retrieval was performed using a sodium citrate-hydrochloric acid buffer. After quenched endogenous peroxidase for 10 min, samples were sealed with 5% goat serum for 1 h at 25 °C. The samples were then incubated with an anti-FGFR4 antibody (1:200, Abcam, Cambridge, UK) overnight at 4 °C, followed by incubation of HRP-conjugated anti-rabbit IgG (1:1000, Proteintech, Wuhan, China) at 37 °C for 60 min. After staining with DAB (Beyotime, Shanghai, China) and hematoxylin, and sealing with neutral resins, the images were observed and photographed under a microscope (Nikon Corporation).

Blood samples were obtained to determine the levels of FGF15 (i.e., FGF19 in humans), FSH, and testosterone in the control and PCOS mice using the FGF15 ELISA Kit (7TM Antibodies, Jena, Germany), FSH (Rodent) ELISA Kit (Abnova, Taiwan, China), and testosterone ELISA Kit (Abcam), respectively, in accordance with the manufacturer’s protocols.

Procella (Wuhan, China) provided human ovarian granulosa cells (KGN cells), which were cultured in DMEM/F12 (Thermo Fisher Scientific, United States) supplemented with 10% fetal bovine serum (FBS, Thermo Fisher Scientific) and 1% penicillin/streptomycin (Thermo Fisher Scientific). The KGN cells were maintained in an incubator at 37 °C with 5% CO₂.

Cells were transfected using Lipofectamine 3000 (Thermo Fisher Scientific) following the manufacturer’s recommendations. Briefly, si-NRF2-1/2/3, si-negative control (si-NC), oe-FGF19 (pcDNA3.1+-FGF19), and oe-NC (oe-NC, pcDNA3.1+) were designed, synthesized, and provided by Yanzai Biotechnology (Shanghai, China). The sequences of si-NRF2-1/2/3 were shown as follows: si-NRF2-1, GAGAAAGAATTGCCTGTAA; si-NRF2-2, GCTACGTGATGAAGATGGA; si-NRF2-3, GCCCTCACCTGCTACTTTA; and si-NC, TTCTCCGAACGTGTCACGT. A total of 2 × 10⁴ KGN cells per well were seeded in a 24-well plate and incubated overnight. The next day, the culture medium was replaced with a serum-free medium. Then, 4 μg of oe-FGF19 plasmids, 4 μg of oe-NC, 20 nM of si-NC, or 20 nM of si-NRF2 were respectively transfected to KGN using Lipofectamine 3000. Six hours post-transfection, the medium was replaced with a complete medium. Following a further 24-h incubation period, the transfection efficiency was examined by evaluating FGF19 or NRF2 expression using real-time quantitative PCR (RT-qPCR) or Western blotting.

To investigate the role of FGF19 in PCOS, KGN cells were randomly divided into four groups: oe-NC, LY3214996 (an ERK inhibitor), oe-FGF19, and oe-FGF19+ LY3214996. KGN cells in the oe-NC and LY3214996 groups were first transfected with oe-NC plasmid and then treated with DMSO and 10 nM LY3214996. Cells in the oe-FGF19 and oe-FGF19+ LY3214996 groups were first transfected with oe-FGF19 plasmid and then treated with DMSO and 10 nM LY3214996.

Furthermore, we explored the interaction between FGF19 and NRF2 in high-oxidative stress-induced PCOS progression. KGN cells were randomly assigned to PBS, H₂O₂, H₂O₂+si-NC, H₂O₂+si-NRF2, H₂O₂+oe-FGF19, and H₂O₂+si-NRF2+oe-FGF19 groups. Apart from the PBS-treated cells, KGN cells in the rest groups were firstly administrated with H₂O₂ (100 μM) for 4 h, and then transfected with corresponding si-NC, si-NRF2, and oe-FGF19.

KGN cells under different treatments were harvested for determination of cell viability and apoptosis using a cell counting kit-8 (CCK-8, BBI Life Sciences) and an Annexin V-FITC/PI cell apoptosis assay kit (YEASEN, Shanghai, China), respectively. For cell viability, the harvested cells were added with 10 μL CCK-8 reagent. After 2 h of incubation, the absorbance was measured at 450 nm using a microplate reader (Biotek, Vermont, United States).

For cell apoptosis, the cells after centrifugation (1,000 rpm, 5 min) were resuspended in 195 μL Annexin V-FITC binding buffer, stained with 5 μL Annexin V-FITC and 5 μL propidium iodide (PI), and incubated for 20 min at 25 °C in the dark. Flow cytometry (Becton Dickinson) was used for imaging, and apoptosis rates were quantified using the CellQuest software.

To examine ROS production, KGN cells under different treatments were subjected to flow cytometry using a commercial ROS assay kit (Beijing Biao Leibo Technology Co. Ltd., Beijing, China). The cells were pelleted (1,000 rpm, 5 min), stained with 10 μM DCFH-DA in PBS for 30 min at 37 °C. After washing, the cells were resuspended in complete medium and analyzed using flow cytometry.

Total RNA was extracted from cells or tissues under various treatments using the TRIzol reagent (Invitrogen, United States). Subsequently, the RNA was reverse-transcribed into cDNA using a first-strand synthesis kit (TransGen Biotech, China), following the manufacturer’s instructions. After that, the synthetic cDNA was applied for qPCR, with the reaction system including 10 μL 2 × AceQTM Universal SYBR qPCR Master Mix (Vazyme China), 8 μL DNA, 0.5 μL forward/reverse primers (10 μM), as well as sterile water (up to 20 μL). The mRNA expression of these genes was calculated by the 2^−ΔΔCT method with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the housekeeping gene. The sequences of all the primers are listed in Table 2.

Cells subjected to different treatments, along with tissue samples, were lysed in RIPA buffer for total protein extraction. The extracted protein was quantified using a BCA assay kit (Beyotime, Shanghai, China). Following this step, protein samples (20 μg per lane) were electrophoresed on 10% SDS-polyacrylamide gels and subsequently transferred to PVDF membranes. After blocking with 5% skim milk at 37 °C for 1 h, the membranes were incubated with primary antibodies against NRF2 (cat. no. ab89443, 1: 1000, Abcam), FGFR4 (cat. no. 11098-1-AP; 1:1000; Proteintech) and FLAG (cat. no. 80010-1-RR, 1:10000, Proteintech); ERK (cat. no. 11257-1-AP; 1:1000; Proteintech) and BAX (cat. no. 50599-2-Ig; 1:1000; Proteintech) and p-ERK (cat. no. #4695, 1:1000; Cell Signaling Technology) and BCL2 (cat. no. A0208, 1:1000; Abcam), HO1 (cat. no. 82206, 1:1000, Cell Signaling Technology) and GAPDH (cat. no. #5174, 1:20000, Cell Signaling Technology) overnight at 4 °C. After washing, the membranes were incubated with goat anti-mouse/rabbit IgG (H + L)-HRP (cat. no. #7076 or #7074, 1:5000, Cell Signaling Technology) at 37 °C for 2 h. Finally, the immunoreactive bands were detected using a Millipore enhanced chemiluminescence (ECL) detection kit (Beyotime).

Results are reported as mean ± SD of triplicate experiments. Statistical significance was evaluated using SPSS (IBM, United States) and figures were created using GraphPad Prism 9 (GraphPad Software, United States). For two-group comparisons, an unpaired t-test was used, whereas a one-way ANOVA was used for multi-group analyses. Statistical significance was set at P < 0.05.

To determine FGF19 levels in PCOS and the correlation between FGF19 levels and PCOS-related indices, blood samples were collected from PCOS patients. We found that patients with PCOS exhibited significantly higher FGF19 levels than healthy controls (P = 0.0013, Figure 1A). We further analyzed the correlation between FGF19 and PCOS-related indicators. No significant correlation was observed between FGF19 and FSH, LH, E2, PRL, AMH, progesterone, or testosterone levels (P > 0.05; Figure 1B). However, FGF19 levels were positively correlated with AMH, E2, and FSH levels and negatively correlated with LH and PRL levels (Figure 1B). These hormonal correlations suggest a potential interplay between FGF19 signaling and endocrine axes relevant to PCOS pathogenesis, which might influence both metabolic and proliferative phenotypes.

To further explore the molecular pathogenesis of PCOS, follicular fluid samples from PCOS patients and controls were used for proteomic analysis. Principal component analysis (PCA) showed that the samples in the PCOS and control groups were clearly separated (Figure 2A), indicating that all samples could be used for follow-up analyses. In total, 144 DEPs were identified, including 116 upregulated (XPOT, RFC4, EDN2, SF3B4, and PRPSAP2) and 28 downregulated DEPs (RDH11, CPSF2, IGHA2, CPM, and COQ9) (Figure 2B). The heatmap of the identified DEPs showed that the identified DEPs revealed a clear separation between control and PCOS samples (Figure 2C).

After that, the identified DEPs were subjected to functional analyses, containing KEGG pathways and hallmark pathways. As shown in Figure 2D, the identified DEPs were significantly enriched in “terpenoid backbone biosynthesis,” “VEGF signaling pathway,” “focal adhesion,” “PPAR signaling pathway,” “pantothenate and CoA biosynthesis,” “fatty acid degradation,” as well as “complement and coagulation cascades.” Furthermore, the identified DEPs were also involved in pathways of “bile acid metabolism” (related DEP: PRDX5), “hypoxia,” “oxidative phosphorylation,” “xenobiotic metabolism,” “IL-2-STAT5 signaling,” “apoptosis,” “TNF-α signaling via NF-κB,” “cholesterol homeostasis,” “epithelial mesenchymal transition (EMT),” as well as “mTORC1 signaling” (Figure 2E).

In female mice, sexual maturation is accompanied by distinct estrous cycles under hormonal regulation, reflecting the ovulatory and endocrine functions of the ovaries. After model induction, daily vaginal smears were performed for 10 consecutive days to compare the estrous cycle patterns between the control and PCOS model groups. We observed that all control mice underwent phases of proestrus, estrus, metestrus, and diestrus (Figure 3A). However, PCOS mice exhibited disrupted estrous cyclicity, with some mice persistently remaining in estrus, while others were in diestrus, demonstrating significant heterogeneity in cycle patterns compared to controls (Figure 3B). Subsequently, the OGTT experiment showed that the glucose level in PCOS mice was evidently higher than that in control mice (P < 0.05, Figure 3C), suggesting impaired glucose tolerance in PCOS mice.

Histological examination of ovarian morphology was performed using HE staining. Control mice exhibited normal ovarian architecture with well-organized follicular development at all stages (primordial, primary, secondary, and antral follicles) and intact corpus luteum formation. The follicular structures were clearly demarcated, with distinct cellular morphologies (Figure 3D). In contrast, PCOS mice displayed multiple enlarged antral follicles characterized by sparse granulosa cells that were peripherally displaced, forming large follicular antrums (Figure 3D). Moreover, FGFR4 levels were significantly higher in mice with PCOS than in healthy animals, as demonstrated by IHC staining (Figure 3E).

Further, the associated markers were measured using ELISA, RT-qPCR as well as Western blot. Compared to control mice, the levels of FGF15 and testosterone were significantly (P < 0.05) higher in the blood of PCOS mice, whereas the FSH level was significantly lower (Figure 4A). The ileum and liver tissues were collected for RT-qPCR. In the ileum tissues, a significant elevation in FGF15 mRNA was observed in PCOS mice versus controls (P < 0.05), whereas FXR mRNA levels were comparable between the groups (P > 0.05; Figure 4B). Liver tissue analysis revealed significantly elevated FGFR4 mRNA levels in PCOS mice compared to controls (P < 0.05); however, CYP7a1 and CYP8b1 mRNA levels were not significantly different between the groups (P > 0.05; Figure 4C). Furthermore, oxidative stress-related proteins (NRF2 and HO1), ERK pathway-related proteins, and apoptosis-related proteins (BAX and BCL2) were examined using Western blotting. NRF2 and HO1 expression was markedly (P < 0.05) elevated in PCOS mice compared to that in control mice (Figure 4D). Quantitative analysis revealed a significant increase in BAX and decrease in BCL2 expression in PCOS mice relative to controls (P < 0.05, Figure 4E). Additionally, compared with control mice, the level of p-ERK/ERK was markedly reduced in PCOS mice (P < 0.05; Figure 4E). Concurrent observations of impaired glucose tolerance, altered bile acid metabolism proteins, and hormonal disturbances suggest systemic metabolic-endocrine crosstalk in PCOS pathogenesis.

Due to the high expression of FGF19 in PCOS, we performed in vitro experiments to elucidate the role of FGF19 in PCOS progression. Images of the survival of KGN cells treated with LY3214996 (an ERK inhibitor) or oe-FGF19 are shown in Figure 5A. Cell viability was quantified, and we observed that with an increase in culture time, the viability of KGN cells in different groups increased gradually. After 72 h of culture, KGN viability was significantly reduced by LY3214996 treatment, whereas it was significantly enhanced by transfection of oe-FGF19 compared with the oe-NC group (Figure 5B). However, compared with the FGF19 overexpressed KGN cells, LY3214996 treatment markedly decreased cell viability (Figure 5B). FGF19 overexpression significantly (P < 0.05) reduced the apoptosis of KGN cells; however, LY3214996 treatment markedly (P < 0.05) enhanced apoptosis caused by FGF19 overexpression (Figure 5C).

Western blotting revealed that NRF2 and HO1 protein expression was downregulated in the LY3214996-treated KGN cells, whereas it was markedly upregulated in the FGF19 over-expressed KGN cells compared to the non-NC-transfected KGN cells (Figure 5D). Relative to the FGF19 over-expressed KGN cells (P < 0.05), LY3214996 treatment markedly downregulated their expression (Figure 5D). Additionally, relative to oe-NC-transfected KGN cells, FLAG and FGFR were upregulated after FGF19 overexpression (P < 0.05); however, LY3214996 treatment had no significant effect on their expression relative to the oe-FGF19 group (P > 0.05, Figure 5E). BAX expression was significantly downregulated after FGF19 overexpression (P > 0.05), but LY3214996 treatment markedly upregulated its expression caused by FGF19 overexpression (P < 0.05; Figure 5E). Nonetheless, the expression trends of BCL2 and p-ERK/ERK were reciprocally related to the BAX levels in all treatment groups. (Figure 5E).

To further explore oxidative stress in PCOS, KGN cells were treated with H₂O₂ and transfected with si-NRF2 or oe-FGF19. NRF2 expression remained unchanged between the control and si-NC groups (P > 0.05) but was significantly restrained by both si-NRF2-1 and si-NRF2-3 transfection, with si-NRF2-3 demonstrating superior knockdown efficiency (Figure 6A). Therefore, si-NRF2-3 was selected to construct KGN cells with NRF2 silencing for further study. The viability of KGN cells after different treatments was examined using the CCK-8 assay. The viability of KGN cells in the different groups increased with increasing culture time (Figure 6B). Relative to the control cells, H₂O₂ induction evidently decreased cell viability (P < 0.05). NRF2 knockdown significantly reduced cell viability, whereas FGF19 overexpression markedly (P < 0.05) increased cell viability stimulated by H₂O₂ (Figure 6B). The apoptotic rate of KGN cells inversely correlated with cell viability across the treatment groups (Figure 6C). Furthermore, we measured ROS levels in different KGN cells under high oxidative stress conditions. Relative to the control conditions, H₂O₂-treated KGN cells exhibited significantly higher ROS production (P < 0.05). NRF2 knockdown further increased the ROS levels induced by H₂O₂ (P < 0.05), whereas FGF19 overexpression had the opposite effect (Figure 6D).

Finally, Western blotting was performed to examine the levels of FLAG, FGFR4, BAX, BCL2, NRF2, HO1 and p-ERK/ERK in KGN cells treated with different methods under high oxidative stress. For FLAG-FGF19 and FGFR4, their protein expression showed no significant changes (P > 0.05) among the PBS, H₂O₂, H₂O₂+si-NC, and H₂O₂+si-NRF2 groups, whereas FGF19 overexpression significantly upregulated their expression relative to the control KGN. In the cells with NRF2 knockdown and FGF19 overexpression, FLAG-FGF19 was evidently upregulated while FGFR4 was significantly downregulated compared to the FGF19 overexpressed KGN (P < 0.05, Figure 7). Based on the apoptosis-related proteins, H₂O₂ administration significantly enhanced BAX and decreased BCL2 levels (P < 0.05), and NRF2 knockdown potentiated these effects, further elevating BAX and reducing BCL2 levels beyond the H₂O₂-induced changes (P < 0.05; Figure 7). However, compared to the H₂O₂+si-NRF2 group, FGF19 overexpression downregulated BAX expression and upregulated BCL2 expression (Figure 7). The expression of NRF2 and HO1 was significantly higher after H₂O₂ administration compared to the control KGN, and NRF2 silencing markedly decreased their expression caused by H₂O₂ (P < 0.05); however, FGF19 overexpression had the opposite effect on NRF2 knockdown (Figure 7). In addition, the level of p-ERK/ERK was lower in the H₂O₂ group than in the control KGN cells (P < 0.05), and NRF2 knockdown did not significantly alter the p-ERK/ERK levels caused by H₂O₂ (P > 0.05, Figure 7). Nonetheless, FGF19 overexpression significantly elevated p-ERK/ERK levels under high oxidative stress relative to H₂O₂-induced KGN (P < 0.05; Figure 7).