This study was conducted at The Fourth Hospital of Shijiazhuang. Two microarray datasets were analyzed to identify differentially expressed genes (DEGs) in endometrial tissue from patients with endometriosis. We included 14 patients with a surgically and histologically confirmed diagnosis of endometriosis who underwent laparoscopic excision during the study period for further validation. Ectopic endometrial samples were collected for mRNA and protein analysis. Eutopic endometrial tissue from the same patient served as the control. Additionally, endometrial samples from 5 healthy participants were obtained to assess target gene expression in normal endometrium. All clinical data were retrieved from the patients’ electronic medical records. Primary ESCs were isolated to investigate the biological function of the candidate gene.

Gene expression datasets were retrieved from the GEO portal and filtered based on data characteristics, experimental design, and sample size. Two microarray datasets (GSE11691 and GSE23339) were selected for DEGs analysis. Transcriptome analysis was performed using R (version 4.4.1) in RStudio (Desktop version, 2024.04.2+764; Boston, MA, USA). Background adjustment was performed using the gcRMA software package [16]. Data quality was assessed with the array QualityMetrics software package (v3.65.0, Bioconductor Project, Boston, MA, USA) [17], and samples of poor quality were excluded from subsequent analyses. Multiple probes corresponding to the same gene were consolidated into a single value by summarizing the median expression levels. DEGs were identified using the limma software package (version 3.52.4, Walter and Eliza Hall Institute of Medical Research, Royal Parade, PV, Australia), with thresholds set at an adjusted p-value 0.05 and $|$Log₂FoldChange$|$ ($|$Log₂FC$|$) $>$1. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were subsequently performed for the identified key genes.

WGCNA was applied to the GSE11691 and GSE23339 datasets to identify critical module genes. Pearson correlation coefficients were calculated to assess gene to gene relationships. A scale-free network was constructed, and an appropriate soft threshold was selected for network formation. The adjacency matrix was then transformed into a topological overlap matrix, and hierarchical clustering was used to generate a clustering tree. Dynamic tree cutting defined the co-expression modules. Modules with a correlation coefficient $>$0.5 with specific traits were considered critical. Finally, a Venn diagram identified the intersecting genes between the DEGs and the critical module genes.

Intersection genes underwent GO enrichment analysis, covering cellular components (CCs), molecular functions (MFs), and biological pathways (BPs), as well as KEGG pathway analysis. The clusterProfiler R package (v4.12.6) was used for statistical analysis.

To identify biological pathways, MFs, and CCs significantly associated with genes whose expression correlated with FABP4, GSEA was performed. First, expression profiles for FABP4 and all other genes were extracted across samples. Spearman correlation coefficients between FABP4 and each gene were then computed. Genes were ranked based on their correlation coefficients, from the highest positive to the lowest negative correlation, and the ranked gene list was subsequently subjected to GSEA.

Endometriotic tissues were collected from 14 Chinese-Han women diagnosed with endometriosis at The Fourth Hospital of Shijiazhuang (mean age: 30.2 years; range: 28–34 years; follicular phase, n = 5; luteal phase, n = 9) between October 2022 and May 2025. The inclusion criteria were as follows: (1) a history of infertility for more than one year; (2) age between 20 and 38 years; (3) a basal serum follicle-stimulating hormone (FSH) level 8 IU/L and a basal serum LH level 10 IU/L prior to controlled ovarian hyperstimulation; and (4) a baseline antral follicle count $>$6. Exclusion criteria comprised: (1) polycystic ovary syndrome or isolated hyperandrogenism, premature ovarian failure, uterine abnormalities (e.g., uterine fibroids, Asherman’s syndrome, adenomyosis, or endometrial polyps), diabetes, hypertension, thyroid disorders, or other hepatic, renal, cardiac, or hematologic diseases; (2) untreated or inadequately managed endocrine or immune disorders; and (3) a history of smoking, alcohol abuse, or substance addiction. All participants were diagnosed with superficial peritoneal endometriosis or deep infiltrating endometriosis and subsequently underwent complete excision of the lesions. None of the patients had received preoperative hormonal therapy. Endometriosis was initially suspected based on clinical or ultrasonographic findings and confirmed through surgical and postoperative pathological examination. Laparoscopic evaluation ruled out other pelvic pathologies that could be confounders. The menstrual phase was determined based on the day of the reproductive cycle and histological analysis of the endometrium.

5 healthy women (mean age: 27.4 years; range 26–30 years; follicular phase, n = 2; luteal phase, n = 3) without endometriosis were recruited in this study, and endometrial samples were obtained using a pipelle device.

Written informed consent was obtained from all participants prior to enrollment in this study. All procedures adhered to the ethical standards of the institutional and national research committees, as well as the Helsinki Declaration and its subsequent amendments. The Bioethics Committee of The Fourth Hospital of Shijiazhuang approved this study (Approval number: 20220039).

Medical records were obtained from the Shijiazhuang Obstetrics and Gynecology Hospital. Each tissue sample was divided for total RNA extraction, protein extraction, and cell isolation.

Paraffin-embedded tissue sections (4 µm thick) were deparaffinized through sequential xylene immersion (2 $\times$ 10 min) and rehydrated in a graded ethanol (cat.no.BP2818-500, Thermo Fisher Scientific, Waltham, MA, USA) series (100%, 95%, and 70%, 3 min each). After thorough rinsing in PBS (Phosphate-Buffered Saline; pH 7.4, cat.no.18912014, Thermo Fisher Scientific, Waltham, MA, USA), antigen retrieval was performed by heating the sections in 10 mM sodium citrate buffer (pH 6.0, cat.no.S5770-50 ML, Sigma-Aldrich, St. Louis, MO, USA) at 95 °C for 20 min in a water bath. Following cooling to room temperature (RT) and washing with PBS, endogenous peroxidase activity was quenched with 3% H₂O₂ (cat.no. 033323-AP, Thermo Fisher Scientific, Waltham, MA, USA) for 15 min at RT.

Sections were blocked with 5% normal goat serum (cat.no.C0265, Beyotime Biotechnology, Shanghai, China) in PBS with 0.3% Triton X-100 (cat.no. X100-100ML, Sigma-Aldrich, St. Louis, MO, USA) for 1 h at RT, then incubated overnight at 4 °C with primary antibody (rabbit anti-FABP4 polyclonal antibody, 1:200 dilution, cat.no. ab13979, Abcam, Waltham, MA, USA). After three 10 min washes with TBST (Tris-buffered saline with 0.1% Tween 20), sections were incubated with horseradish peroxidase (HRP)-conjugated goat anti-rabbit secondary antibody (1:500 dilution, cat.no. #7074, Cell Signaling Technology, Inc., Danvers, MA, USA) for 2 h at RT.

Following three additional TBST washes (5 min each), immunoreactivity was visualized using a DAB substrate kit (cat.no. #8059, Cell Signaling Technology, Inc. Danvers, MA, USA) with a development time of 3–5 min. Sections were counterstained with Mayer’s hematoxylin for 1 min, dehydrated through graded alcohol series, cleared in xylene, and mounted with mounting medium.

Images were captured using microscopy. Two pathologists independently evaluated staining intensity (0: negative; 1: weak; 2: moderate; 3: strong) and the percentage of positively stained cells (0–100%). An H-score (range: 0–300) was calculated by multiplying the staining intensity by the percentage of positive cells.

ESCs were isolated from endometrial tissue with established methods [18, 19]. Briefly, tissue samples were minced and then digested in DMEM/F12 with type I collagenase (2.5 mg/mL, cat.no. C1-22-1G, Sigma-Aldrich, St. Louis, MO, USA) and DNase I (15 U/mL, cat.no. D7291, Sigma-Aldrich, St. Louis, MO, USA) at 37 °C for 1 h. Debris was removed by filtration through a 40 µm nylon cell strainers. Cells were collected by centrifugation at 400 $\times$g for 10 min and cultured in DMEM/F12 supplemented with 10% fetal bovine serum (FBS) (cat.no. E510008, Sangon Biotech Co., Ltd, Shanghai, China), penicillin (100 U/mL) and streptomycin (100 µg/mL) (cat.no. E607011-0100, Sangon Biotech Co., Ltd, Shanghai, China). Upon reaching confluence, cells were dissociated with 0.25% trypsin (cat.no. E607002-0100, Sangon Biotech Co., Ltd, Shanghai, China), harvested, and re-plated in 6 well plates at a density of 2 $\times$ 10⁵ cells/well. Cell purity was assessed by immunofluorescence staining for vimentin (stromal cells), cytokeratin (epithelial cells), and CD45 (leukocytes). Stromal cell purity exceeded 95%, as confirmed by positive vimentin staining and negative staining for cytokeratin and CD45.

The full length of 396 bp FABP4 coding sequence was cloned into the pcDNA3.1 plasmid between BamHI and XbaI sites to generate the FABP4 overexpression vector. The plasmid was transfected into primary ESCs by electroporation, and the expression of FABP4 was examined by immunoblotting 48 h after transfection.

Proteins were extracted from cell samples using Radioimmunoprecipitation Assay (RIPA) buffer (cat.no.J63306-AP, Thermo Fisher Scientific, Waltham, MA, USA), separated on 4–12% Bis-Tris polyacrylamide gels (cat.no. NP0321BOX, Thermo Fisher Scientific, Waltham, MA, USA), and transferred to polyvinylidene difluoride (PVDF) membranes (cat.no. IPVH00010, Millipore, Burlington, MA, USA). Membranes were blocked with 5% fat-free milk and incubated overnight at 4 °C with a primary antibody against FABP4. After washing with TBST, membranes were incubated with an HRP-conjugated secondary antibody diluted in TBST containing 5% fat-free milk (cat.no.P0216-300g, Beyotime Biotechnology, Shanghai, China). Protein bands were visualized using a SuperSignal West Femto Maximum Sensitivity Substrate kit (cat.no. 34094, Thermo Fisher Scientific, Waltham, MA, USA), with $\beta$-actin as the loading control.

Antibody information:

Rabbit anti-FABP4 polyclonal antibody (1:1000 dilution, cat.no. ab13979, Abcam, Waltham, MA, USA).

HRP-conjugated goat anti-rabbit secondary antibody (1:3000 dilution, cat.no. #7074, Cell Signaling Technology, Danvers, MA, USA).

HRP-conjugated mouse anti-$\beta$-actin monoclonal antibody (1:3000 dilution, cat.no. sc-542972, Santa Cruz Biotechnology, Dallas, TX, USA).

Cell viability was assessed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. A total of 2 $\times$ 10³ ESCs were seeded into each well of 96-well plates and allowed to adhere overnight. After treatment, the cells in each well were incubated with 20 µL of MTT (5 mg/mL, cat.no. 475989-1GM, Sigma-Aldrich, St. Louis, MO, USA) for 4 h. The supernatants were carefully removed from each well and dimethyl sulfoxide (DMSO, cat.no. 20-139, Sigma-Aldrich, St. Louis, MO, USA) was added. Absorbance was measured at 570 nm.

Cells were harvested by gentle trypsinization, washed once with ice-cold PBS (pH 7.4), and pelleted by centrifugation at 400 $\times$g for 5 min at 4 °C. The supernatant was carefully aspirated, and cells were resuspended in 1$\times$ Annexin V binding buffer (10 mM HEPES, 140 mM NaCl, 2.5 mM CaCl₂, pH 7.4) at a density of 1 $\times$ 10⁶ cells/mL.

Cells were incubated with 5 µL of Annexin V-FITC (Cat. No. 640906, BioLegend, San Diego, CA, USA) for 15 min at RT in the dark. Prior to analysis, 10 µL of propidium iodide (PI) solution (50 µg/mL) was added to discriminate late apoptotic or necrotic cells. Samples were acquired within 1 h of staining using a BD FACSCanto II flow cytometer (BD biosciences, Franklin Lakes, NJ, USA), with a minimum of 20,000 events recorded per sample. Fluorescence was detected using: FITC (Annexin V): 488 nm excitation, 530/30 nm emission filter. PI: 488 nm excitation, 585/42 nm emission filter. Results were analyzed using FlowJo software (v10.4.1, Tree Star, Inc. Ashland, OR, USA).

Mitochondrial function was evaluated using a Seahorse XF24 Analyzer (Agilent Technologies, Santa Clara, CA, USA) and the Seahorse XF Cell Mito Stress Test Assay (cat.no. 103016-100, Agilent Technologies, Santa Clara, CA, USA). Briefly, 1 $\times$ 10⁴ cells were seeded into 24-well plates and allowed to adhere overnight. OCR was measured following sequential injections of oligomycin, carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP), and a combination of rotenone and antimycin A, according to the manufacturer’s protocol.

The cells were fixed using 4% formaldehyde at RT for 10 minutes, followed by permeabilization on ice for an additional 10 minutes with PBS containing 0.5% Triton-X100. After a 30 min incubation with blocking buffer composed of PBS, 0.1% Tween 20, and 1% BSA, cells were incubated overnight at 4 °C with diluted antibodies: Alexa Fluor 488 labeled VIM antibody (1:300 dilution, cat.no. ab154207, Abcam, Waltham, MA, USA), Alexa Fluor 488 labeled mouse anti-cytokeratin monoclonal antibody (1:200 dilution, cat.no. sc-57004 AF488, Santa Cruz Biotechnology, Dallas, TX, USA), or FITC labeled mouse anti-CD45 monoclonal antibody (1:200 dilution, cat.no. sc-53201 FITC, Santa Cruz Biotechnology, Dallas, TX, USA). After three washes, slides were mounted with an antifade mounting medium containing DAPI (cat.no. H-1200-10, Vector Laboratories, Inc., Newark, CA, USA), and images were captured using fluorescence microscope (ECLIPSE Ti2, Nikon, Melville, NY, USA).

Data was analyzed using R (version 4.4.1). Statistical significance between the two groups was evaluated by using two-tailed Student’s t-test, and results with a p-value 0.05 were considered significant. For comparisons among three groups, one-way analysis of variance (ANOVA) followed by post-hoc tests was applied.

This study had several limitations: (1) Our findings are based on bioinformatic analyses and in vitro cellular experiments; in vivo studies with animal models are necessary to validate the role of FABP4 in endometriosis. (2) Although this study confirms the role of FABP4 in the regulation of mitochondrial function, its potential effects on angiogenesis and immune responses require further investigation.