Hypothalamic Epac2/AKT/GnRH signalling is associated with metformin-responsive neuroendocrine improvement in a PCOS model Authors: Na Guo, Hongyun Li, Jinhong He, Huanhuan Zhou, Xing Wang, Linlin Yang, Huijuan Ma Published: 2026 ABSTRACT AbstractBackgroundPolycystic ovary syndrome (PCOS) is a complex endocrine-metabolic disorder characterised by hypothalamic-pituitary-gonadal (HPG) axis dysfunction and pathological gonadotropin-releasing hormone (GnRH) hypersecretion. While metformin demonstrates therapeutic efficacy in PCOS, its potential neuroendocrine mechanisms remain incompletely understood. This study aimed to investigate the involvement of hypothalamic signalling in metformin-responsive neuroendocrine regulation in PCOS.MethodsLetrozole-induced PCOS rat models were treated with metformin to assess reproductive and metabolic outcomes. Data-independent acquisition (DIA) proteomics of hypothalamic tissue was performed to identify dysregulated pathways, followed by network analysis to explore candidate regulatory nodes. Mechanistic investigations were conducted in glucosamine-induced insulin-resistant GT1-7 GnRH neurons using pharmacological modulation and siRNA knockdown. In vivo functional relevance was examined using AAV-mediated Epac2 knockdown in the mouse arcuate nucleus.ResultsMetformin improved metabolic and reproductive abnormalities in PCOS rats, including insulin resistance, GnRH/LH hypersecretion, estrous cyclicity disruption, and ovarian morphological alterations. Hypothalamic proteomic and network analyses suggested that metformin-responsive changes were associated with insulin and cAMP-related signalling pathways. In GT1-7 neurons, metformin increased Epac2 expression, restored AKT signalling, and reduced GnRH overproduction under insulin-resistant conditions. Hypothalamic Epac2 knockdown reduced AKT signalling and was associated with systemic insulin resistance, increased GnRH/LH secretion, and preliminary supportive ovarian morphological observations, consistent with a functional contribution of Epac2 to reproductive-metabolic regulation.ConclusionsThese findings support the involvement of hypothalamic Epac2/AKT/GnRH signalling in metformin-responsive neuroendocrine regulation in PCOS. Further studies are required to determine the relative contributions of central and peripheral mechanisms.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13048-026-02212-9.Keywords: Polycystic ovary syndrome, Metformin, Hypothalamus, AKT2, Epac2, Hypothalamic insulin signalling, GnRH Introduction Polycystic ovary syndrome (PCOS) is a prevalent endocrine disorder affecting 8–13% of reproductive-aged women worldwide [1–3]. It is clinically characterised by chronic anovulation, hyperandrogenism, and frequent comorbid metabolic disturbances, leading to significant fertility impairment [1–3]. Advancing the understanding of PCOS pathophysiology is therefore critical, not only for improving individual health and quality of life but also for developing effective strategies to safeguard reproductive health at the population level. Over the past decade, insulin resistance (IR) has emerged as an important pathophysiological feature of PCOS [1, 4, 5]. Although peripheral IR, typically associated with obesity, dyslipidemia, and increased risk of type 2 diabetes, has been extensively investigated, its relationship with reproductive axis imbalance remains incompletely understood [1, 2, 5]. PCOS is characterised by dysregulation of the hypothalamic-pituitary-gonadal (HPG) axis, whereby altered gonadotropin-releasing hormone (GnRH) pulsatility contributes to elevated luteinizing hormone (LH) secretion [2, 3]. Experimental studies further indicate that metabolic and androgen-related cues can alter GnRH neuronal activity [6, 7]. These observations underscore hypothalamic insulin sensitivity as a poorly defined yet therapeutically relevant determinant of GnRH dysregulation in PCOS. Metformin, a first-line insulin-sensitizing agent in PCOS [1, 4, 5, 8], has been reported to serve as an adjunct to aromatase inhibitors for ovulation induction [9], with potential benefits for pregnancy-related outcomes in selected PCOS populations [10]. However, its reproductive mechanism remains poorly defined. The 2023 International PCOS Guideline conditionally recommends metformin for ovulation induction [1], with insufficient mechanistic evidence. This knowledge gap poses a clinical challenge: without understanding how metformin improves reproductive outcomes, treatment response remains unpredictable. Evidence that metformin can access the central nervous system or exert biologically relevant hypothalamic effects suggests potential central involvement [11, 12]. Elucidating this mechanism would provide biological basis for metformin’s reproductive benefits and inform hypothalamus-directed therapeutic development in PCOS. To address this question, we employed hypothalamic proteomic profiling in a letrozole-induced PCOS rat model to explore candidate pathways associated with metformin-responsive neuroendocrine regulation. While our previous work identified a hypothalamic RASA1/Ras/AKT/GnRH axis in GLP-1RA-responsive rescue of PCOS HPG dysfunction [13], the present study was designed to examine whether metformin is associated with a distinct upstream regulatory module within the broader AKT/GnRH framework. Functional validation was performed using in vitro GnRH neuronal models and in vivo AAV-mediated gene manipulation in mice. Our findings suggest the involvement of hypothalamic signalling in metformin-responsive reproductive effects and may help to inform the relationship between peripheral insulin resistance and central GnRH dysregulation in PCOS. The present work provides a mechanistic framework for investigating insulin-related neuroendocrine regulation in PCOS and highlights hypothalamic insulin signalling as an area for further study in ovulatory dysfunction. Methods Animals Female Sprague-Dawley (SD) rats (7 weeks old) were obtained from Hua Fukang Company (Beijing, China) and acclimatized for one week under specific pathogen-free conditions. Animals were maintained on irradiated chow with ad libitum access to water. At 8 weeks of age, experimental rats (n = 19) received daily oral gavage of letrozole (1 mg/kg/day; Cat# MB1137, Meilunbio, Dalian, China) dissolved in 1% carboxymethylcellulose sodium (CMC-Na) solution for three weeks to induce polycystic ovary syndrome (PCOS). Control rats (n = 9, Con) received an equivalent volume of 1% CMC-Na. Successful PCOS induction was confirmed by persistent disruption of the estrous cycle. Following PCOS establishment, PCOS rats were randomly assigned to receive metformin by oral gavage (300 mg/kg/day, Met, n = 9) for a period of six weeks. Letrozole administration was continued during this period to maintain the PCOS phenotype. At study termination, rats were deeply anaesthetised with sodium pentobarbital (50 mg/kg, intraperitoneal injection). Adequate depth of anaesthesia was confirmed by the absence of the pedal withdrawal reflex. Blood was collected via the abdominal aorta, and animals were euthanised by exsanguination under deep anaesthesia. After terminal blood collection, rats were transcardially perfused with PBS followed by 4% paraformaldehyde (PFA) under deep anaesthesia. Death was confirmed by cessation of respiration and heartbeat prior to tissue collection. Blood samples were allowed to clot at room temperature and centrifuged at 3000×g for 15 min at 4 °C. Serum was aliquoted and stored at − 80 °C. For tissue collection, whole brains from three rats per group were immersion-fixed in 4% paraformaldehyde (PFA), and hypothalamic regions were subsequently processed for immunofluorescence staining, while the remaining hypothalamic samples were snap-frozen in liquid nitrogen for subsequent molecular analyses. Ovaries were fixed in 4% PFA for histological evaluation. All animal procedures were conducted in accordance with institutional guidelines and internationally accepted standards for the care and use of laboratory animals and were approved by the Laboratory Animal Committee of Hebei General Hospital (Approval No. 2025-DW-131). Estrous cycle monitoring The stage of the estrous cycle was determined through cytological analysis of vaginal smears observed under light microscopy. Vaginal lavage was gently performed using sterile saline introduced through a plastic pipette tip, and the aspirated fluid was transferred onto clean glass slides. The smears were air-dried for 10 min, fixed in absolute ethanol for another 10 min, and stained with 0.1% crystal violet solution (Cat# G1063, Solarbio, Beijing, China) for 1 min. Following a rinse with distilled water, slides were mounted using glycerol gelatin and examined microscopically. Cytological classification of estrous stages was based on the predominant cell types observed: proestrus was characterised by nucleated epithelial cells, estrus by fully keratinized anucleated squamous cells, metestrus by a mixed population of cornified cells and leukocytes, and diestrus by a predominance of leukocytes. Disruption of this normal cyclic pattern was used as a phenotypic marker for successful induction of the PCOS model. Serological testing To evaluate hormonal alterations associated with PCOS and treatment intervention, serum concentrations of key reproductive hormones and insulin were measured using enzyme-linked immunosorbent assay (ELISA) kits, strictly following the manufacturers’ protocols. Testosterone (T, Cat# 582701) and estradiol (E2, Cat# 501890) assays were obtained from Cayman Chemical (Ann Arbor, USA). GnRH was measured using a commercial ELISA kit (Cat# E-EL-0071c, Elabscience, Wuhan, China), for which the manufacturer reported a sensitivity of 9.38 pg/mL, a detection range of 15.63–1000 pg/mL, intra-assay coefficients of variation of 5.02–6.85%, and inter-assay coefficients of variation of 7.58–8.28%. Follicle-stimulating hormone (FSH; Cat# CEA830Ra) and LH (Cat# CEA441Ra) kits were purchased from Cloud-Clone Corp (Wuhan, China), and insulin (Cat# 80-INSRTU-E01) was assayed using kits from ALPCO (Salem, NH, USA). Glucose tolerance assessment To evaluate systemic glucose homeostasis and insulin sensitivity, an intraperitoneal glucose tolerance test (IPGTT) was performed. Following an 8–12-hour overnight fast (with ad libitum access to water), rats were administered a sterile glucose solution (2 g/kg body weight; dissolved in saline) via intraperitoneal (i.p.) injection. Blood glucose concentrations were measured from the tail vein immediately prior to glucose administration (0 min, baseline) and at 15, 30, 60, and 120 min post-injection using a handheld glucometer. The area under the curve (AUC) for glucose over the 120-minute test period was calculated using the trapezoidal rule to quantify overall glucose excursion. Hematoxylin-eosin staining Ovarian tissues were fixed in 4% paraformaldehyde at 4 °C, dehydrated, embedded in paraffin, and sectioned into 5 μm slices for histological analysis. Sections were deparaffinized using an environmentally friendly dewaxing agent, rehydrated through a descending ethanol gradient, and rinsed in tap water. Hematoxylin staining was performed using HD constant staining solution, followed by differentiation in acid alcohol and bluing in an alkaline solution. Subsequently, cytoplasmic components were stained with eosin after brief dehydration in 95% ethanol. Finally, sections were fully dehydrated, cleared in xylene, and mounted with neutral balsam. Histopathological evaluation was conducted under light microscopy to assess morphological alterations associated with PCOS and treatment response. Data-independent acquisition mass spectrometry (DIA-MS) based proteomic profiling Frozen hypothalamic samples were homogenized in SDT buffer and processed for DIA-based proteomic profiling. The DIA-MS analysis included independent hypothalamic samples from the Control group (n = 6), PCOS group (n = 6), and Met group (n = 5), with each sample derived from one independent animal. No pooled samples were used. Peptides were separated by nano-LC and analysed on a Q Exactive HF-X mass spectrometer. A spectral library was generated from DDA runs using Spectronaut™ against the UniProt rat database. Differential analysis To identify differentially expressed proteins (DEPs) between comparison groups (PCOS vs. Con; Met vs. PCOS), normalised peak area data were analysed using a two-tailed Student’s t-test. A protein was classified as a DEP if it showed nominal statistical significance (P < 0.05, unadjusted) and a fold change > 2 or < 0.5. No FDR-adjusted threshold was applied for DEP selection. Therefore, the proteomic analysis should be interpreted as an exploratory, hypothesis-generating screen used to prioritise candidate pathways and proteins for downstream validation. Functional enrichment analysis Gene Ontology (GO) enrichment analyses were conducted to identify the functional characteristics of the DEPs. The analysis was performed using the R package clusterProfiler through the OmicStudio cloud platform (http://www.bioinformatics.com.cn/). Network analysis A comprehensive protein-protein interaction (PPI) network was constructed based on DEPs between PCOS and metformin groups according to our established pipeline [14, 15]. Network modularity analysis was performed using simulated annealing algorithm [16, 17] to identify functional modules within the largest connected component. Functional annotation of each module was determined by the most significantly enriched biological pathways represented by module nodes. Cell culturing The GT1-7 cell line, an immortalized hypothalamic GnRH neuronal model, was purchased from Sigma-Aldrich (Product Number: SCC116). Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Gibco, Thermo Fisher Scientific, Waltham, USA) supplemented with 10% fetal bovine serum (FBS, Gibco), 1% penicillin-streptomycin (Gibco), and 1% glutamine (Gibco). Cultures were maintained at 37 °C in a humidified incubator containing 5% CO₂. Cells were routinely subcultured using trypsin-EDTA (Gibco) when they reached 80–90% confluence. For experimental treatments, GT1-7 cells were seeded into 6-well plates at a density of 3 × 10⁵ cells per well. Upon reaching approximately 80% confluence, cells were treated with one of the following compounds: glucosamine (100 µM, GlcN, Cat# S1635, Beyotime), metformin (1 mM, Met, Cat# S5958, Selleck), SC79 (10 µM, Cat# HY-18749, MCE), ESI-09 (20 µM, Cat# HY-16704, MCE) or 8-pCPT-2′-O-Me-cAMP sodium (20 µM, 007, Cat# HY-107543, MCE). After 24 h of drug exposure, both cells and culture supernatants were collected for downstream analyses. Glucose uptake quantification Cellular glucose uptake was determined by measuring extracellular glucose depletion in culture supernatants using a glucose oxidase-peroxidase assay kit (Cat# E1011, Applygen). Glucose concentrations were measured in 96-well plates by incubating 5 µL supernatant with 195 µL freshly prepared working solution (Reagent R1:R2 = 4:1) at 37 °C for 20 min protected from light. Absorbance was read at 550 nm against daily-generated glucose standards. Glucose uptake was calculated as: [(Initial medium glucose) - (Measured supernatant glucose)]. For normalisation to viable cell mass, parallel cultures were assayed with Cell Counting Kit-8 (CCK-8): cells were incubated with 10% CCK-8 reagent in serum-free medium for 2 h at 37 °C, and absorbance measured at 450 nm. Detection of GnRH To assess GnRH secretion from GT1-7 cells, culture supernatants were collected following 4-hour treatments with GlcN, Met, ESI-09, or 007. GnRH levels in the conditioned media were quantified using a commercial ELISA kit, following the manufacturer’s instructions (Cat# E-EL-0071c, Elabscience, Wuhan, China). AKT2 knockdown by siRNA transfection AKT2 loss-of-function was achieved through RNA interference using GP-transfect-Mate transfection reagent (Genepharma, Shanghai, China). Cells were seeded in 6-well plates at 3 × 10⁵ cells/well in complete medium and incubated overnight (37 °C, 5% CO₂). At approximately 70% confluence, transfection complexes were prepared in sterile microcentrifuge tubes: Reagent complex: 6 µL GP-transfect-Mate + 200 µL serum-free medium; siRNA complex: 7.5 µL AKT2-targeting siRNA (sequence: 5′-GCCGCUAUUAUGCCAUGAATT′) + 200 µL serum-free medium. After brief vortexing, complexes were combined and incubated at room temperature for 15 min. Concurrently, culture medium was replaced with 1.6 mL fresh pre-warmed medium per well. The transfection mixture (400 µL) was added dropwise to each well (final siRNA concentration: 25 nM), followed by gentle plate agitation. Following 6 h incubation, medium was replaced with complete growth medium containing experimental compounds for subsequent functional analyses. Quantitative RT-PCR assay Total RNA was isolated from GT1-7 cells using TRIzol reagent (Cat# RC101-01, Vazyme Biotech, Nanjing, China), and RNA quality was assessed using a NanoDrop 2000 spectrophotometer. Samples with A260/A280 ratios between 1.8 and 2.0 were considered acceptable. Reverse transcription was carried out using the FastKing RT Kit with gDNase (Cat# KR116, TIANGEN, China), followed by quantitative PCR using the SuperReal PreMix Plus kit (Cat# FP205, TIANGEN, China) on an ABI 7500 Real-Time PCR system (Thermo Fisher Scientific). The relative expression levels of GnRH and Epac2 were quantified using the 2^−ΔΔCt method, with GAPDH serving as the internal reference gene. The specific primer sequences were as follows: GnRH: Forward 5′-AGCACTGGTCCTATGGGTTG-3′, Reverse 5′-GGGGTTCTGCCATTTGATCCA-3′; Epac2: Forward 5′-TGGATGGGATGCTCTATTCTGG-3′, Reverse 5′-GCGGAGGAAGTTGTCGGTC-3′; GAPDH: Forward 5′-AGGTCGGTGTGAACGGATTT-3′, Reverse 5′-GGGGTCGTTGATGGCAACA-3′. All primers were synthesized by Sangon Biotech (Shanghai, China). Western blot Total protein was extracted from GT1-7 cells, and concentrations were determined using the BCA protein assay kit (Cat# 23227, Thermo Fisher Scientific, Waltham, USA). Protein samples were denatured, separated on 10% SDS-PAGE gels, and transferred onto polyvinylidene difluoride (PVDF) membranes. Membranes were blocked with QuickBlock buffer (Cat# P0252, Beyotime Biotechnology, Beijing, China) for 20 min at room temperature and then incubated overnight at 4 °C with the following primary antibodies: Epac2 (1:1000; Cat# ab193665, Abcam, Cambridge, UK), AKT2 (1:200; Cat# sc-5270, Santa Cruz Biotechnology, Dallas, TX, USA), phospho-AKT2 (Ser474) (1:1000; Cat# 8599, Cell Signaling Technology, Danvers, MA, USA), total AKT (1:1000; Cat# 4685, Cell Signaling Technology, Danvers, MA, USA; recognizing Akt1, Akt2, and Akt3), phospho-AKT (Ser473) (1:1000; Cat# 4060, Cell Signaling Technology, Danvers, MA, USA), and GAPDH (1:2000; Cat# 390035, ZEN-BIOSCIENCE, Chengdu, China). All primary antibodies were diluted in Antibody Dilution Buffer (Cat# P0023A, Beyotime Biotechnology). After washing in TBST, membranes were incubated with HRP-conjugated secondary antibodies (1:5000) for 1 h at room temperature. Detection was performed using enhanced chemiluminescence (ECL), and protein bands were visualised using the Omega Lum C imaging system (Aplegen, USA). Densitometric analysis was conducted using ImageJ software to quantify protein expression levels relative to GAPDH. For western blot presentation, representative cropped images were prepared from the available original chemiluminescent exposure images by selecting the relevant molecular-weight regions. After transfer, membranes were cut according to molecular-weight regions for antibody incubation. Target proteins and their corresponding GAPDH loading controls were detected on membrane strips prepared from the same gel with the same sample set and lane order. Each independent repeat represents a biological replicate from a separate experimental run. The available source images, together with lane annotations, molecular-weight information, non-analysed lanes where applicable, and lanes used for quantification are provided in the Supplementary Western Blot Source Data. A supplementary traceability map links each main figure panel to the corresponding source blot, target protein, loading control, molecular-weight region, lane order, biological repeat, and non-analysed lanes where applicable. Stereotaxic delivery of Epac2-targeting AAV Female C57BL/6 mice (8 weeks old) were anaesthetised with sodium pentobarbital (50 mg/kg, i.p.) and secured in a stereotaxic frame under aseptic conditions. Adequate depth of anaesthesia was confirmed by the absence of the pedal withdrawal reflex, and body temperature was maintained throughout the procedure. Bilateral injections targeting the arcuate hypothalamic nucleus (ARC) were performed using coordinates relative to bregma: anteroposterior − 1.45 mm, mediolateral ± 0.20 mm, and dorsoventral − 5.78 mm from the dura. A recombinant AAV2/9 vector expressing Epac2 shRNA under the U6 promoter and EGFP under the CMV promoter (pAAV-U6-shRNA(Epac2)-CMV-EGFP-WPRE, AAV-Epac2-shRNA; Heyuan Bio, Shanghai, China) or a scrambled control vector with non-targeting shRNA (pAAV-U6-shRNA(NV2)-CMV-EGFP-WPRE, AAV-scrambled; Heyuan Bio, Shanghai, China) was delivered via a glass micropipette at a volume of 30 nL per side. For the expanded AAV knockdown experiment, endocrine and metabolic phenotyping and estrous cycle assessment were performed in n = 5 mice per group. Ovarian ultrasonography and H&E-based ovarian morphology assessment were performed in a subset of animals (n = 3 per group) and were treated as preliminary/exploratory complementary morphological assessments rather than definitive endpoints. Estrous cyclicity in AAV-treated mice was monitored by daily vaginal smear cytology using the same morphological criteria described above for estrous stage classification. Following injection, the micropipette was left in place for an additional 5 min to minimise reflux before slow withdrawal and wound closure. Mice were allowed to recover and housed for 3 weeks to ensure stable viral expression. Thereafter, mice were deeply anaesthetised with sodium pentobarbital (50 mg/kg, i.p.), and deep anaesthesia was confirmed by loss of the pedal withdrawal reflex. Terminal blood was collected via retro-orbital sampling for serum hormone measurements. While under deep anaesthesia, mice were then euthanised by transcardial perfusion with PBS followed by 4% paraformaldehyde (PFA). Death was confirmed by cessation of respiration and heartbeat prior to brain collection. Brains were collected for histological and immunofluorescence analyses, and serum was separated and stored at − 80 °C until use. All procedures were conducted in accordance with institutional guidelines and internationally accepted standards for the care and use of laboratory animals, and were approved by the Animal Ethics Committee of Hebei General Hospital (Approval No. 2025-DW-164). Ovarian ultrasonography Ovarian ultrasonography was performed on anesthetized mice positioned in dorsal recumbency on a heated platform. After depilation and application of acoustic coupling gel, high-resolution micro-ultrasound imaging was conducted using a Vevo2100 system (FUJIFILM VisualSonics, Canada). The probe was suspended 2–3 mm above the right dorsolateral surface and systematically scanned along the longitudinal axis, using the cranial renal pole as an anatomical landmark. Ovarian images were acquired in both longitudinal and transverse planes for subsequent structural assessment and cross-sectional area measurement. Immunofluorescence analysis Following transcardial perfusion with 4% PFA, brains were post-fixed overnight at 4 °C, cryoprotected in 30% sucrose, and sectioned coronally at 20 μm thickness using a cryostat. Serial sections encompassing the ARC were collected for analysis. Viral targeting accuracy was verified by epifluorescence microscopy detecting inherent EGFP signal in ARC regions. For protein localization, free-floating sections underwent antigen retrieval (10 mM sodium citrate, 80 °C, 30 min), blocking in 10% normal donkey serum/0.3% Triton X-100/PBS, and incubation with primary antibodies against Epac2 (1:100, Cat# sc-28326, Santa Cruz Biotechnology, Shanghai, China), AKT2 (1:100, Cat# sc-5270, Santa Cruz Biotechnology, Shanghai, China), pan-AKT / total AKT (t-AKT; 1:200, Cat# 4685, Akt (pan) (11E7) Rabbit Monoclonal Antibody, Cell Signaling Technology, USA), and GnRH (1:50; Cat# sc-25344, Santa Cruz Biotechnology, Shanghai, China) for 24 h at 4 °C. After PBS washes, sections were incubated for 1 h at room temperature with secondary antibodies: Dylight 488-conjugated anti-rabbit IgG (Cat# A23220, Abbkine, Wuhan, China) and Alexa Fluor 549-conjugated goat anti-mouse IgG (Cat# A23310, Abbkine, Wuhan, China), counterstained with DAPI (1 µg/mL). Images were captured using a ZEISS LSM 900 confocal microscope equipped with ZEN software (ZEISS, Germany), and further processed with ZEN Blue software (version 3.4). For quantitative analysis, images were acquired using identical laser power and detector settings across groups. Mean fluorescence intensity was measured within manually defined hypothalamic regions of interest (ROI) using ImageJ, after background subtraction. Statistical analysis All statistical analyses and data visualisation were performed using GraphPad Prism software (version 9). Data distribution was assessed using normality tests available in GraphPad Prism, including the Shapiro–Wilk or Kolmogorov–Smirnov test as appropriate. Homogeneity of variance was evaluated using the Brown–Forsythe test for three-group comparisons or the F test for two-group comparisons. For two independent groups, normally distributed data were analysed using an unpaired Student’s t-test when variances were equal, or Welch’s correction when variances were unequal. Non-normally distributed two-group data were analysed using the Mann–Whitney U test. For three-group comparisons, normally distributed data with homogeneous variances were analysed using ordinary one-way ANOVA followed by Tukey’s multiple-comparisons test. When variances were unequal, Welch’s ANOVA followed by the Games–Howell post hoc test was used. Non-normally distributed three-group data were analysed using the Kruskal–Wallis test followed by Dunn’s multiple-comparisons test. A P value < 0.05 was considered statistically significant. Data are presented as mean ± SEM unless otherwise specified. For all in vivo experiments, each animal was treated as an independent biological replicate. For immunofluorescence quantification, n refers to the number of independent animals per group, and multiple sections or images from the same animal were used only for within-animal averaging. For in vitro experiments, n refers to independent biological experiments performed on separate occasions rather than technical replicate wells. For ELISA, qPCR, and western blot analyses, technical replicates were averaged within each biological replicate before statistical analysis. For DIA-MS proteomics, each sample was derived from one independent animal. For AAV-mediated knockdown experiments, the sample size for each endpoint is specified in the relevant figure legend. Effect sizes were reported for key statistical analyses. For ordinary one-way ANOVA and two-group t-test analyses, η² was reported using the R-squared value provided by GraphPad Prism. For Welch’s ANOVA, approximate partial η² was calculated from the Welch test statistic and degrees of freedom. For Kruskal–Wallis tests, ε² was calculated from the Kruskal–Wallis H statistic, number of groups, and total sample size. Effect-size estimates for key analyses are provided in Supplementary Table S1. Results Metformin improves metabolic dysfunction in letrozole-induced PCOS rats Chronic letrozole administration successfully established a PCOS-like phenotype characterised by marked metabolic and reproductive dysfunction. Metabolically, PCOS rats exhibited progressive weight gain, hyperinsulinemia, elevated HOMA-IR, and impaired glucose tolerance, reflecting systemic insulin resistance (Fig. 1B–G). Metformin intervention (300 mg/kg/day for 6 weeks) produced substantial therapeutic effects despite continued letrozole exposure during the treatment period. Treatment significantly attenuated weight gain, reduced hyperinsulinemia, improved insulin sensitivity (decreased HOMA-IR), and restored glucose tolerance (Fig. 1B–G). Metformin improves metabolic dysfunction in letrozole-induced PCOS rats. A Experimental schematic showing timeline of letrozole-induced PCOS model establishment and metformin intervention protocol; letrozole administration was continued during metformin treatment to maintain the PCOS phenotype. B Body weight progression throughout the 9-week study period (n = 6/group). Metabolic assessment: (C) fasting blood glucose (FBG), (D) fasting insulin (FINS), (E) homeostatic model assessment for insulin resistance (HOMA-IR), (F) IPGTT profiles, and G AUC analysis (n = 5/group). Statistical analysis: Data are presented as mean ± SEM. Group comparisons performed using one-way ANOVA followed by Tukey’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant Metformin improves neuroendocrine dysfunction in letrozole-induced PCOS rats Reproductively, circulating GnRH, T, and LH were significantly elevated with a markedly increased LH/FSH ratio, while E2 levels were significantly decreased and FSH remained relatively preserved (Fig. 2A–F). Vaginal cytology consistently demonstrated persistent diestrus, confirming chronic anovulation (Fig. 2G). Morphologically, ovarian histology demonstrated enlarged ovaries with multiple cystic follicles, representing the classic polycystic morphology (Fig. 2H). Quantitative follicle counts confirmed a decrease in mature follicles and proportion of mature ovarian follicles in the PCOS group (Fig. 2I–J). Concurrently, metformin reduced the elevated GnRH, T, and LH levels, restored E2 levels, and normalised the LH/FSH ratio (Fig. 2A–F). Functionally, metformin restored estrous cyclicity in a significant subset of treated animals (Fig. 2G). Histologically, treated ovaries showed markedly reduced cystic follicles with concomitant increases in mature follicle numbers and proportions, suggesting improved ovarian morphological features (Fig. 2H–J). Metformin improves reproductive dysfunction in letrozole-induced PCOS rats. Hormonal profiling: (A) serum GnRH, (B) T, (C) LH, (D) FSH, (E) LH/FSH ratio, and (F) E2 concentrations (n = 5–7/group). G Representative vaginal cytology images following crystal violet staining (scale bar = 100 μm). Ovarian morphology: (H) representative H&E-stained ovarian sections (scale bar = 1000 μm), (I) quantitative mature follicle counts, and (J) proportion of mature follicles relative to total follicle population (n = 5–6/group). Statistical analysis: Data are presented as mean ± SEM. Statistical tests were selected according to data distribution and variance homogeneity as described in the Methods. Detailed statistical tests and effect-size estimates are provided in Supplementary Table S1. *P < 0.05, **P < 0.01, ***P < 0.001, ns, not significant. Note: Serum GnRH should be interpreted as an indirect endocrine readout because of its rapid degradation and pulsatile secretion, and was considered together with LH, LH/FSH ratio, and hypothalamic GnRH-related findings Because circulating GnRH is rapidly degraded and secreted in a pulsatile manner, serum GnRH was interpreted as an indirect endocrine readout rather than a definitive measure of hypothalamic GnRH secretory dynamics. These findings confirm successful establishment of a clinically relevant PCOS model and support metformin’s beneficial effects on both metabolic dysfunction and reproductive abnormalities. Exploratory quantitative proteomics prioritises hypothalamic Epac2 as a metformin-responsive candidate target in PCOS To investigate candidate molecular targets associated with metformin-responsive neuroendocrine regulation, we performed DIA-MS on hypothalamic tissue from Con, PCOS, and Met rats. Principal component analysis (PCA) of the hypothalamic proteome demonstrated clear segregation between Con and PCOS samples, while Met samples clustered intermediate between PCOS and Con groups, suggesting partial normalisation of the disease-associated proteomic profile (Fig. 3A). Exploratory quantitative proteomics shows metformin-responsive proteomic changes in the hypothalamus in PCOS. A Principal component analysis of DIA-based proteomes from hypothalamic tissue of Con (blue), PCOS (red), and Met (yellow) rats. B Volcano plots displaying DEPs in PCOS versus Con (left) and Met versus PCOS (right). Red dots indicate significantly upregulated proteins; blue dots indicate significantly downregulated proteins; gray dots indicate non-significant proteins. C Venn diagram illustrating overlap between PCOS-dysregulated and metformin-responsive protein sets. D Gene Ontology enrichment analysis of metformin-responsive proteins showing top enriched terms in Biological Process (green), Cellular Component (orange), and Molecular Function (blue) categories. DEPs were defined using fold change > 2 or < 0.5 together with nominal P < 0.05, without FDR adjustment Differential expression analysis revealed extensive proteomic remodelling: 210 proteins were significantly dysregulated in PCOS versus control rats, and 301 proteins showed altered expression following metformin treatment compared to untreated PCOS animals (Fig. 3B). Intersection analysis identified 104 overlapping DEPs (Fig. 3C), representing 49.5% of PCOS-altered proteins and 34.6% of metformin-responsive proteins. GO enrichment analysis of these metformin-responsive proteins revealed significant overrepresentation of biological processes related to neurosecretion, synaptic transmission, and intracellular signalling pathways (Fig. 3D). These findings suggest that PCOS pathophysiology is associated with broad alterations in hypothalamic neuronal and neuroendocrine machinery, some of which may be modulated by metformin. Because DEP selection was based on nominal P values without FDR adjustment, these proteomic findings were interpreted as exploratory and used to guide downstream functional validation. Epac2-AKT signalling is associated with metformin-responsive hypothalamic changes To further elucidate the underlying mechanisms, we conducted protein-protein interaction network analysis based on DEPs. The insulin signalling pathway was identified as a prominent module in the PCOS hypothalamic network, suggesting a potential role for hypothalamic insulin-related signalling in neuroendocrine dysfunction (Fig. 4A). In contrast, the metformin network was characterised by a predominant neuroactive module centred on the cAMP signalling pathway (Fig. 4A). Focused analysis of DEPs within these two modules revealed crosstalk between insulin- and cAMP-related signalling pathways (Fig. 4B–C). Hypothalamic Epac2-AKT signalling is prioritised as a metformin-responsive candidate network in PCOS. A Functional decomposition of PCOS and metformin network. Nodes represent functional modules; edge thickness denotes the amount of interactions across distinct modules; (B) Hierarchical clustering heatmap of DEPs in the core neuroactive modules; (C) PPI network architecture showing convergence of insulin and cAMP signalling pathways with Epac2-AKT2 as candidate regulatory components; (D) Representative immunofluorescence staining demonstrating hypothalamic Epac2 and AKT2 expression patterns across pathological (PCOS) and therapeutic (Met) states; (E) Quantitative analysis of hypothalamic Epac2 and AKT2 immunoreactivity supporting their involvement as candidate treatment-responsive signalling components (n = 3 per group). Data are expressed as mean ± SEM and analysed by one-way ANOVA followed by Tukey’s post hoc test. *P < 0.05, **P < 0.01 In the Met versus PCOS comparison, Epac2 (fold change = 3.445, P < 0.01) and AKT2 (fold change = 2.432, P < 0.05) showed treatment-associated increases within these pathways, whereas their baseline PCOS versus Control differences did not reach statistical significance (Epac2: fold change = 0.498, P = 0.126; AKT2: fold change = 0.461, P = 0.084). Accordingly, Epac2 and AKT2 were interpreted as metformin-responsive candidate signalling components rather than primary basal PCOS proteomic abnormalities. In vivo immunofluorescence analysis provided an orthogonal protein-level assessment, showing reduced Epac2 and AKT2 immunoreactivity in PCOS hypothalamus and increased signals following metformin treatment (Fig. 4D–E). Together, the proteomic, network, and immunofluorescence findings support the prioritisation of Epac2 and AKT2 as treatment-responsive candidate signalling components for downstream functional analysis. Metformin is associated with AKT restoration and reduced GnRH overproduction in insulin-resistant GnRH neurons Building on our proteomic findings, we investigated metformin-associated effects on AKT signalling and GnRH regulation in a neuronal cell context. We established an insulin resistance cell model using GnRH-secreting GT1-7 cells treated with GlcN (100 µM), which significantly impaired insulin-stimulated glucose uptake compared to vehicle-treated controls (Fig. 5A–B). Metformin is associated with AKT restoration and reduced GnRH overproduction in insulin-resistant neurons. A Schematic of GlcN-induced IR model in GT1-7 cells; (B) Glucose uptake in control and IR GT1-7 cells with or without metformin (n = 3/group); (C) Representative immunoblots of AKT2, t-AKT, p-AKT2, and p-AKT; (D–G) Quantitative analysis of AKT2, p-AKT2, t-AKT and p-AKT protein levels normalised to loading controls (n = 3/group); (H) GnRH mRNA expression (n = 3/group); (I) GnRH secretion levels (n = 3/group). Representative cropped western blot images are shown. The available source blot images with lane annotations and loading-control correspondence are provided in the Supplementary Western Blot Source Data. Data are presented as mean ± SEM. Statistical analysis: one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001 IR group exhibited markedly reduced expression of both AKT2 and total AKT (t-AKT), along with decreased phosphorylation levels (p-AKT2 and p-AKT), confirming disrupted AKT signalling (Fig. 5C–G). Consequently, GlcN treatment significantly elevated both GnRH transcription and secretion (Fig. 5H–I). Metformin co-treatment restored AKT2 expression and phosphorylation toward control levels in IR neurons, while reducing pathological GnRH hypersecretion (Fig. 5C–I). These results suggest that metformin may improve neuronal insulin-related signalling alongside AKT restoration and may contribute to modulation of hypothalamic GnRH dysregulation under insulin-resistant conditions. AKT signalling contributes to metformin-associated GnRH regulation in hypothalamic neurons We investigated the role of AKT2 in hypothalamic GnRH regulation using GT1-7 cells. Compared with NC-siRNA, AKT2-targeting siRNA (AKT-KD) efficiently reduced both AKT2 and total AKT protein levels (Fig. 6A–C). Functionally, AKT2 silencing significantly increased GnRH mRNA expression (Fig. 6D), suggesting that AKT2 contributes to the restraint of GnRH transcription. AKT contributes to metformin-associated GnRH regulation. A Representative immunoblot showing AKT2/t-AKT expression in AKT2-knockdown GT1-7 cells; (B–C) Quantification of AKT2 and t-AKT protein levels (n = 3/group); (D) GnRH mRNA levels following AKT2 knockdown (n = 3/group); (E–G) AKT2 and t-AKT protein levels in IR GT1-7 cells treated with metformin±AKT2 siRNA (n = 3/group); (H) GnRH mRNA levels in IR cells with metformin±AKT2 knockdown (n = 3/group); (I) GnRH mRNA in SC79-treated IR neurons (n = 3/group); (J) Representative immunofluorescence of hypothalamic t-AKT and GnRH in experimental rats; (K) Quantification of t-AKT and GnRH fluorescence intensity (n = 3/group). Representative cropped western blot images are shown. The available source blot images with lane annotations and loading-control correspondence are provided in the Supplementary Western Blot Source Data. Data are presented as mean ± SEM. Two-group comparisons: unpaired two-tailed Student’s t-test; multiple-group comparisons: one-way ANOVA with post hoc tests. *P < 0.05, **P < 0.01, ***P < 0.001 To determine whether AKT2 mediates metformin’s GnRH-lowering effects under insulin-resistant conditions, we examined GlcN-induced IR GT1-7 cells. Metformin treatment restored AKT2 and total AKT protein levels (GlcN + Met+NC-siRNA vs. GlcN + NC-siRNA; Fig. 6E–G). However, AKT2 knockdown markedly attenuated the metformin-induced AKT restoration (GlcN + Met+AKT-KD vs. GlcN + Met+NC-siRNA; Fig. 6E–G). Critically, while metformin suppressed GnRH mRNA in control siRNA-treated IR cells, this effect was markedly attenuated by AKT2 knockdown (Fig. 6H), supporting an important role for AKT2 in metformin-associated GnRH suppression. Pharmacological AKT activation with SC79 reduced GnRH mRNA levels in IR GT1-7 cells (Fig. 6I), further suggesting that AKT activation can contribute to the restraint of GnRH overproduction. In vivo validation showed a consistent pattern: the hypothalamus of PCOS rats exhibited decreased total AKT alongside increased GnRH compared with controls, while metformin reversed both abnormalities (Fig. 6J–K). Collectively, these findings support the involvement of impaired hypothalamic AKT signalling in GnRH hyperactivity in PCOS and suggest that AKT restoration may contribute to metformin-associated modulation of GnRH secretion. Epac2 contributes to AKT-GnRH signalling regulation in insulin-resistant neurons To examine whether Epac2 may act within AKT-GnRH regulatory signalling, we investigated Epac2’s role in IR GnRH neurons. GlcN-induced IR markedly suppressed both Epac2 mRNA and protein expression in GT1-7 cells, which metformin treatment substantially restored to control levels (Fig. 7A–B). Pharmacological modulation supported a role for Epac2 in this signalling context. The selective Epac2 agonist 007 recapitulated metformin’s effects, reducing GnRH mRNA expression and secretion in IR neurons (Fig. 7C–D). Conversely, the specific Epac2 inhibitor ESI-09 attenuated metformin-associated suppression of GnRH hypersecretion, supporting Epac2’s regulatory involvement (Fig. 7C–D). Epac2 contributes to AKT-mediated GnRH regulation in IR neurons. A–B Epac2 mRNA and protein levels in GlcN-treated GT1-7 neurons. C–D Effects of Epac2 agonist 007 and inhibitor ESI-09 on GnRH expression and secretion. E–G AKT signalling protein abundance following pharmacological Epac2 modulation. H GnRH levels in AKT2-depleted neurons treated with Epac2 agonist. I–J Hypothalamic Epac2 and GnRH immunofluorescence in control, PCOS, and metformin-treated rats with quantitative analysis (n = 3/group). Representative cropped western blot images are shown. The available source blot images with lane annotations and loading-control correspondence are provided in the Supplementary Western Blot Source Data. Statistical analysis: Data are presented as mean ± SEM. Statistical tests were selected according to data distribution, variance homogeneity, and number of groups as described in the Methods. Detailed statistical tests and effect-size estimates are provided in Supplementary Table S1 Epac2 emerged as a candidate upstream regulator of AKT signalling. Both metformin and 007 increased pan-AKT / total AKT (t-AKT) and AKT2 protein abundance under IR conditions, while ESI-09 blunted these metformin-induced increases (Fig. 7E–G). Genetic validation through AKT2 knockdown revealed that 007’s GnRH-suppressive capacity was markedly impaired in AKT2-depleted versus control cells (Fig. 7H), supporting involvement of the Epac2-AKT2-GnRH pathway. In vivo immunofluorescence analysis showed reduced hypothalamic Epac2 immunoreactivity together with intensified GnRH signals in PCOS rats compared with controls (Fig. 7I–J). Metformin treatment restored hypothalamic Epac2 expression while attenuating GnRH immunofluorescence, consistent with an inverse relationship between Epac2 and GnRH in this pathophysiological context. Together, these findings support Epac2 as a metformin-responsive regulatory component associated with AKT activation and GnRH suppression in insulin-resistant hypothalamic neurons. Targeted Epac2 knockdown in the ARC alters AKT–GnRH neuroendocrine signalling in vivo To examine whether Epac2 contributes to hypothalamic AKT-GnRH signalling in vivo, Epac2 was selectively knocked down in the arcuate nucleus (ARC) via stereotaxic AAV-shRNA delivery (Fig. 8A). Robust EGFP signals confirmed accurate targeting and effective transduction (Fig. 8B). Epac2 knockdown markedly reduced Epac2 immunoreactivity, accompanied by a concomitant reduction in AKT2 signal and pan-AKT / total AKT staining within the ARC compared with scramble controls (Fig. 8C–F). Notably, Epac2 depletion was associated with enhanced hypothalamic GnRH immunoreactivity (Fig. 8E–F), supporting involvement of Epac2 in maintaining AKT activity and restraining GnRH hyperactivation. We next examined whether this local signalling disruption translates into systemic reproductive and metabolic abnormalities. Hypothalamic Epac2 knockdown disrupts AKT signalling and enhances GnRH activity. A Schematic of stereotaxic AAV-mediated Epac2 knockdown in the arcuate nucleus. B Representative EGFP fluorescence confirming successful viral transfection. C Immunofluorescence analysis of Epac2 and AKT2 expression in hypothalamic sections. D Quantification of Epac2 and AKT2 fluorescence intensities (n = 3/group). E Immunofluorescence staining of GnRH and total AKT (t-AKT) in hypothalamus. F Quantification of GnRH and total AKT (t-AKT) fluorescence (n = 3/group). Data are presented as mean ± SEM. Statistical significance was determined by two-tailed Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001 AAV-mediated hypothalamic Epac2 knockdown is associated with PCOS-like reproductive and metabolic abnormalities in vivo Hypothalamic Epac2 depletion was associated with key PCOS-like endocrine and metabolic abnormalities. In the expanded AAV cohort (n = 5/group), Epac2-knockdown mice exhibited disrupted estrous cyclicity characterised by persistent diestrus on vaginal smear cytology (Fig. 9A), together with elevated circulating GnRH and LH levels and an increased LH/FSH ratio (Fig. 9B–E). In parallel, Epac2 knockdown was associated with metabolic impairment, characterised by increased body weight, hyperglycaemia, hyperinsulinaemia, and elevated HOMA-IR indices (Fig. 9F–I). Ovarian ultrasonography and H&E-based morphology were assessed in a smaller subset (n = 3/group). These preliminary/exploratory supportive morphological observations showed enlarged ovarian morphology, as reflected by increased ovarian cross-sectional area, together with a reduction in mature follicle number and the proportion of mature follicles (Fig. 9J–M). Collectively, the expanded endocrine, metabolic, and estrous cyclicity data provide the main evidential basis for the AAV phenotype, while the preliminary/exploratory supportive ovarian morphology findings provide complementary observations. Together, these data support a role of hypothalamic Epac2 in maintaining integrated reproductive–metabolic homeostasis and indicate that its depletion is associated with concurrent reproductive endocrine and metabolic abnormalities resembling core features of PCOS. Hypothalamic Epac2 deficiency is associated with PCOS-like reproductive and metabolic abnormalities in vivo. A Representative vaginal smear cytology showing disrupted estrous cyclicity with persistent diestrus in Epac2-knockdown mice (n = 5/group). Reproductive endocrine parameters: (B) circulating GnRH, (C) LH, (D) FSH levels, and (E) LH/FSH ratio (n = 5/group). Metabolic parameters: (F) body weight, (G) fasting blood glucose (FBG), (H) fasting insulin (FINS), and (I) HOMA-IR index (n = 5/group). Ovarian morphology: (J) representative ovarian morphology, (K) quantitative ovarian cross-sectional area analysis, (L) mature follicle number, and (M) proportion of mature follicles (n = 3/group; preliminary/exploratory supportive morphological endpoints). These ovarian morphology data should be interpreted as preliminary/exploratory complementary supportive evidence rather than definitive morphological proof. Data are presented as mean ± SEM. Statistical significance was determined by two-tailed Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001. Note: Circulating GnRH values should be interpreted as indirect endocrine readouts because of pulsatile secretion and rapid degradation, and were considered together with downstream endocrine readouts and hypothalamic GnRH immunofluorescence Given the rapid degradation and pulsatile secretion of GnRH, circulating GnRH in the AAV experiment was interpreted as an indirect endocrine readout and considered together with LH, LH/FSH ratio, and hypothalamic GnRH immunofluorescence rather than as a standalone measure of GnRH secretory dynamics. Discussion PCOS is characterised by HPG axis dysfunction, yet the lack of neuroendocrine-targeted therapies has confined clinical management to symptomatic relief, including progestin for menstrual irregularities, ovulation induction for infertility, and lifestyle interventions for metabolic dysfunction [1–3]. This approach often necessitates recurrent or lifelong treatment, highlighting the need to better understand neuroendocrine mechanisms relevant to HPG axis regulation. Metformin, though approved primarily for peripheral insulin sensitization in PCOS [1, 4, 5, 8], demonstrates reproductive benefits through incompletely understood mechanisms. Beyond peripheral effects, emerging evidence suggests that metformin can access the central nervous system and exert biologically relevant brain effects. Prior pharmacokinetic and experimental studies have reported metformin in cerebrospinal fluid and brain regions, including the hypothalamus [11], and a recent study further showed that centrally administered metformin inhibited brain Rap1, reduced hyperglycemia, and activated a subset of neurons in the ventromedial hypothalamus [12]. These observations support the biological plausibility of central involvement, although they do not establish direct hypothalamic specificity in our own experimental system. Using integrated approaches—hypothalamic proteomics, GnRH neuronal models, and AAV-mediated gene manipulation—we identified a molecular pathway associated with metformin-responsive hypothalamic signalling via the Epac2/AKT axis in PCOS. Consistent with prior studies [18–20], metformin improved metabolic parameters (reduced weight gain, fasting insulin, HOMA-IR) and reproductive outcomes (normalised LH/FSH ratio, restored estrous cyclicity, increased mature follicles) in PCOS rats. Beyond its known suppressive effect on GnRH neuronal firing [6, 7], our study suggests that metformin is associated with reduced hypothalamic and circulating GnRH levels in PCOS. These parallel changes are consistent with a potential link between hypothalamic signalling and systemic neuroendocrine improvement. To investigate candidate molecular targets associated with metformin-responsive neuroendocrine regulation, we performed hypothalamic proteomic analysis and constructed protein-protein interaction networks for PCOS and metformin-treated conditions. The insulin signalling pathway was identified as a prominent module in the PCOS hypothalamic network, suggesting a potential role for hypothalamic insulin-related signalling in neuroendocrine dysfunction. Within the metformin-responsive network, AKT2 showed treatment-associated modulation, prompting systematic investigation of its role in GnRH regulation. AKT2, a predominant neuronal AKT isoform [21], was prioritised within the metformin-responsive signalling context and further examined by IF and functional validation. IF analysis showed reduced hypothalamic AKT2 and pan-AKT / total AKT signals in PCOS rats, both of which were increased after metformin treatment. Because AKT2 did not reach statistical significance in the PCOS versus Control DIA proteomic comparison, this result should be interpreted as IF-supported and functionally supported evidence within a treatment-responsive signalling framework, rather than as a proteomics-defined basal differential protein. AKT-related signals showed an inverse relationship with GnRH expression across treatment groups. Functional validation further supported this relationship: AKT2 knockdown attenuated metformin-associated GnRH suppression, whereas pharmacological AKT activation reduced GnRH expression. Collectively, these findings suggest that AKT restoration may contribute to metformin-associated modulation of pathological GnRH activity under insulin-resistant conditions. Our previous study demonstrated that GLP-1RA improved PCOS-related neuroendocrine dysfunction, in part, through modulation of the hypothalamic RASA1/Ras/AKT/GnRH axis [13], indicating that AKT/GnRH regulation itself is not entirely novel in this disease context. The present study therefore does not propose a new downstream AKT/GnRH axis, but instead suggests a distinct candidate upstream regulatory node involving Epac2. Notably, in the GLP-1RA-focused dataset reported previously, RASA1 was significantly suppressed in the GLP-1RA versus PCOS comparison, whereas Epac2 did not show a statistically significant change (P = 0.12), suggesting that Epac2 was not a statistically prioritized signal in that context. Taken together, these findings raise the possibility that different upstream signalling modules may converge on a common AKT/GnRH regulatory framework in a treatment-dependent manner. However, the present study does not establish whether Epac2 acts upstream of, parallel to, or independently from Ras signalling, and this remains an open question for future investigation. Beyond AKT, proteomic profiling identified Epac2 as a prominent metformin-responsive candidate node. Epac2 showed a marked treatment-related increase in the Met versus PCOS comparison, whereas its PCOS versus Control proteomic difference did not reach statistical significance. Therefore, Epac2 should be interpreted primarily as a treatment-responsive candidate target rather than as a proteomics-defined basal PCOS differential protein. IF analysis and insulin-resistant GT1-7 cell experiments provided complementary support, showing reduced Epac2-related signals under PCOS or insulin-resistant conditions and restoration after metformin treatment. Pharmacological Epac2 activation reduced GnRH expression and secretion, whereas Epac2 inhibition attenuated metformin-associated GnRH suppression. Together, these findings support Epac2 as a metformin-responsive candidate regulatory component with functional relevance in neuroendocrine regulation, rather than as a definitively proven or indispensable mediator of metformin efficacy. Beyond its established role in AKT phosphorylation via Rap1 [22–24], we identified a potentially important role of Epac2 in regulating total AKT protein levels. In PCOS rat hypothalamus, Epac2 downregulation paralleled decreased AKT2 and total AKT expression, which were coordinately restored by metformin. Insulin-resistant cell models recapitulated this pattern, where metformin corrected Epac2 and AKT deficiencies while normalizing AKT phosphorylation, supporting restoration of insulin signalling. Direct manipulation showed that Epac2 activation enhanced AKT2 and total AKT protein abundance, while Epac2 inhibition decreased both. These findings suggest a previously underexplored role of Epac2 in regulating AKT abundance in addition to phosphorylation status, potentially through ubiquitin-mediated protein stability mechanisms [25–27]. Additional in vivo support came from AAV-mediated hypothalamic Epac2 knockdown, which was associated with reduced hypothalamic AKT levels, peripheral insulin resistance (elevated HOMA-IR), increased hypothalamic and circulating GnRH levels, and LH hypersecretion with an elevated LH/FSH ratio. In the revised study, the AAV knockdown cohort for endocrine and metabolic phenotyping, together with estrous cyclicity assessment, was expanded to n = 5 mice per group, which improved the interpretability of these in vivo comparisons. Ovarian imaging and H&E-based histological assessments were performed in a smaller subset of animals (n = 3/group) and are therefore interpreted consistently as preliminary/exploratory supportive, not definitive, morphology endpoints. This loss-of-function approach supports the functional relevance of hypothalamic Epac2 in reproductive-metabolic regulation, but it does not test whether Epac2 restoration is required for metformin-mediated improvement under PCOS conditions. Accordingly, the main evidential weight is placed on endocrine, metabolic, and estrous cyclicity endpoints, while the ovarian morphology findings provide preliminary/exploratory complementary observations that are directionally consistent with these functional abnormalities. Overall, the observed phenotypes should be interpreted as key PCOS-like endocrine and metabolic features rather than a complete recapitulation of the PCOS phenotype. The overall phenotypic concordance between Epac2 knockdown mice and letrozole-induced PCOS rats across molecular, hormonal, metabolic, and supportive ovarian readouts is consistent with involvement of the hypothalamic Epac2/AKT/GnRH axis in neuroendocrine dysfunction in PCOS (Fig. 10). Proposed involvement of hypothalamic Epac2/AKT/GnRH signalling in metformin-responsive neuroendocrine regulation in PCOS It should be noted that while our functional studies showed metformin-associated changes in hypothalamic Epac2 and AKT expression, proteomic screening between PCOS and control groups revealed only trending differences that did not reach statistical significance. This limitation should be considered when interpreting the pathway-level findings. Accordingly, Epac2 and AKT2 should be interpreted as suggestive, treatment-responsive candidate targets rather than statistically validated primary differential proteins in the basal PCOS state. In contrast, Epac2 showed a marked change in the Met versus PCOS comparison (fold change = 3.445), supporting its prioritisation as a metformin-responsive candidate target for further functional investigation. At the same time, convergent evidence from hypothalamic proteomic/network analysis, targeted immunofluorescence, metformin intervention in insulin-resistant GT1-7 neurons, AKT2 modulation, pharmacological Epac2 modulation, and AAV-mediated Epac2 perturbation supports involvement of the Epac2/AKT pathway in GnRH regulation. Together, these findings are consistent with the possibility that metformin-responsive hypothalamic insulin-related signalling contributes to neuroendocrine regulation in PCOS, while still warranting cautious interpretation. Despite the mechanistic insights provided by this study, several methodological limitations should be considered. First, metformin was administered systemically by oral gavage. Therefore, the present design cannot distinguish whether the observed hypothalamic Epac2/AKT/GnRH changes resulted from direct central actions of metformin, secondary consequences of improved peripheral insulin resistance, or both. Improved whole-body insulin sensitivity may have contributed, at least in part, to the normalisation of hypothalamic signalling observed in the PCOS model. In the current study, hypothalamic metformin concentrations were not measured, intracerebroventricular metformin administration was not performed, and peripheral insulin-correction or central blockade control groups were not included. The study was also not designed to determine the temporal sequence between peripheral metabolic improvement and hypothalamic signalling correction. Future follow-up studies should therefore use dedicated experimental designs incorporating hypothalamic metformin quantification, time-course sampling, intracerebroventricular metformin delivery, peripheral insulin-correction control groups, and central pharmacological blockade to define the relative contributions of central and peripheral mechanisms more rigorously. Second, the DIA-MS proteomic analysis was exploratory and based on nominal P values without FDR-adjusted thresholds. Thus, the proteomic screen may be affected by false-positive findings and may also fail to detect basal intergroup differences when effect sizes are modest, sample heterogeneity is present, or statistical power is limited. Accordingly, Epac2 and AKT2 should be interpreted as treatment-responsive candidate signalling components rather than statistically validated primary basal PCOS proteomic abnormalities. Targeted immunofluorescence, western blotting, pharmacological modulation, siRNA knockdown, and AAV-mediated perturbation were used as complementary validation approaches; these assays may be more sensitive for detecting selected pathway-related changes, but they do not replace fully powered, FDR-controlled proteomic validation. In addition, the GT1-7 cell model and AAV-mediated Epac2 knockdown model cannot fully reproduce the multicellular hypothalamic environment or the full endocrine-metabolic complexity of PCOS. AAV-mediated Epac2 knockdown supports the functional relevance of hypothalamic Epac2, but the present experiment tested the consequences of Epac2 loss rather than whether blocking Epac2 under PCOS plus metformin conditions abolishes or attenuates the therapeutic effects of metformin. Thus, Epac2 restoration should be viewed as a metformin-responsive candidate mechanism rather than an established requirement for metformin efficacy. Future rescue-loss studies should include PCOS + metformin + central Epac2 knockdown, or PCOS + metformin + Epac2/AKT pathway blockade, followed by assessment of hypothalamic AKT-GnRH signalling, GnRH/LH/LH-FSH ratio, insulin resistance, estrous cyclicity, and ovarian morphology. Such studies will be required to determine whether hypothalamic Epac2/AKT signalling is necessary for metformin-responsive endocrine, metabolic, and neuroendocrine improvement. Finally, ovarian ultrasonography and H&E-based ovarian morphology in the AAV experiment were assessed in a smaller subset of animals (n = 3/group) and should consistently be interpreted as preliminary/exploratory supportive observations rather than definitive morphological evidence. Circulating GnRH should also be interpreted cautiously as an indirect endocrine readout because of its rapid degradation and pulsatile secretion pattern. In summary, our findings support the involvement of hypothalamic Epac2/AKT/GnRH signalling in metformin-responsive neuroendocrine regulation in PCOS. These results may help to connect metformin-responsive metabolic and reproductive changes within a common mechanistic framework and are consistent with potential involvement of hypothalamic insulin-related signalling in neuroendocrine homeostasis. However, the present study does not distinguish whether the observed hypothalamic changes reflect direct central actions of metformin, secondary consequences of peripheral metabolic improvement, or a combination of both. Further studies are needed to determine the relative contributions of central and peripheral mechanisms and to define the extent of hypothalamic involvement. Conclusions This study provides evidence supporting the involvement of hypothalamic Epac2/AKT/GnRH signalling in metformin-responsive neuroendocrine regulation. Our findings are consistent with potential hypothalamic involvement in metformin-responsive reproductive-metabolic homeostasis, although direct hypothalamic specificity cannot be established in the present study. Genetic perturbation experiments are further consistent with a functional contribution of hypothalamic Epac2/AKT/GnRH signalling to neuroendocrine homeostasis, but pathway necessity for metformin efficacy remains to be tested. The extent of direct central metformin action also remains to be defined. These findings highlight hypothalamic insulin-related signalling as a potential area for further investigation in the context of PCOS pathophysiology and treatment. Overall, the present study provides a mechanistic basis for future work examining how central and peripheral pathways may jointly contribute to the therapeutic effects of metformin in PCOS. Supplementary Information Supplementary Material 1.