HSD11B1 suppresses ferroptosis in endometrial stromal cells through the JUND/IL-10 axis to promote endometriosis progression

Four EMs datasets ( GSE7305 , GSE120103 , GSE23339 , GSE37837 ) were downloaded from the GEO database, and all samples were from human endometrial tissues ( https://www.ncbi.nlm.nih.gov/geo/ ) [ 18 , 19 ]. FRGs were obtained from the FerrDb ( http://www.zhounan.org/ferrdb/current/ ) and MSigDB databases( https://www.gsea-msigdb.org/gsea/msigdb ) [ 20 , 21 ]. The dataset was de-batched and standardized, and the batch effect removal effect was verified by PCA [ 22 – 24 ]. The ferroptosis score of the samples was calculated using the ssGSEA algorithm, and the gene co-expression network was analyzed through WGCNA to screen the module genes significantly related to the ferroptosis score. The DEGs between EMs and normal samples were analyzed using limma packages. The screening conditions were |logFC| > 1 and adG. p < 0.05. The DEGs and WGCNA module genes were intersected to obtain FRDEGs. Five machine learning algorithms, namely LASSO regression, elastic network, Boruta, Bagged Trees and GMM, were used to screen FRDEGs, and the intersection was taken to obtain the Key Genes [ 25 ]. LASSO and elastic network mainly visualize the FRDEGs using the R package glmnet [ 26 ] (version 4.1-8) with parameters set.seed(2024) and family binomial. The Boruta algorithm is used to screen based on FRDEGs through the R package Boruta (version 8.0.0), and the results of the Boruta algorithm are visualized through box plots. The Bagged Trees algorithm is used to screen based on FRDEGs through the R package Caret (version 6.0–94), and the results of the Bagged Trees algorithm are visualized through line graphs. A GMM model is constructed based on FRDEGs using the R packages SimDesign (version 2.18) and mclust (version 6.1.1), and the results of the GMM model are visualized through scatter plots. The intersection of the genes selected by LASSO regression, elastic network regression, Boruta algorithm, Bagged Trees algorithm, and GMM model is taken, and an UpSet graph is drawn for visualization through the R package UpSetR (version 1.4.0). The genes in the intersection of these five algorithms are the key genes.Based on key genes, diagnostic models were constructed using six algorithms such as SVM, random forest, and XGB to evaluate the model performance (ROC curve, PRC curve, etc.) [ 27 , 28 ]. Using the R packages Caret (version 6.0–94) and DALEX (version 2.4.3), six machine learning algorithms were employed to build machine learning models. Six machine learning algorithms were used to construct machine learning models for key genes. Residual distribution graphs, PRC curves, and ROC curves were drawn for each model. Subsequently, the importance of key genes in different machine learning models was investigated, and the Break Down values and SHAP values of each key gene were plotted and visualized using the R package iBreakDown (version 2.1.2).The expression differences and diagnostic efficacy (AUC values) of key genes were verified in the integrated dataset and the validation set ( GSE37837 ). A Logistic regression model was constructed based on key genes, and the model performance was evaluated through nomogram, correction curve and DCA. The diagnostic model of this study was compared with the published EMs model to verify its diagnostic efficacy. The Seurat package [ 29 ] (version 5.0.1) was used to import the expression matrices of all EMs and control samples from the scRNA-seq dataset GSE214411 and create them as Seurat objects. The parameters were set to at least 3 cells with expressed genes and at least 200 genes expressed in each cell. The filtering conditions were set as nCount RNA (Unique Molecular Identifier, UMI) greater than 500 and less than 10,000, nFeature RNA greater than 200, Log10(Genes per UMI) greater than 0.8, and mitochondrial gene proportion (mitoRatio) less than 0.2. Low-quality cells were removed. The scDblFinder package (version 1.16.0) was used to perform double-cell identification on the scRNA-seq dataset and remove double cells. Statistical analysis was conducted on the quality-controlled cells, and the changes in cell numbers before and after filtering were calculated. The sequencing depth of the scRNA-seq dataset GSE214411 was standardized using the ‘NormalizeData’ function with the “LogNormalize” method, and the high-variable genes of the dataset were detected using the “FindVariableFeatures” function’s “vst” method. Then, the data was scaled using the “ScaleData” function to eliminate the influence of sequencing depth. Principal Component Analysis (PCA) was applied to identify significant principal components (Principal Component, PC), and the distribution of standard deviations was visualized using the ‘Elbowplot’ function. 30 principal components were selected for unified manifold approximation and projection (Uniform Manifold Approximation and Projection, UMAP) analysis for dimensionality reduction. The ‘FindNeighbors’ function was used to construct the k-nearest neighbor neighborhood in the Euclidean distance of the base PCA space using the 30 principal components dimension parameters. Import the quality-controlled and standardized single-cell (scRNA-seq) dataset GSE214411 Seurat object using the R package Seurat (Version 4.3.0). To address the potential batch effects between samples, use the “RunHarmony function” in the R package harmony (Version 1.2.0) for data integration and batch effect removal, and divide the cells into different clusters at a resolution of 1.0. Use multiple methods to conduct comprehensive cell type annotation for the single-cell (scRNA-seq) dataset. Use the “SingleR” function in the R package SingleR [ 30 ] (Version 2.0.0) to annotate the cell types of the single-cell (scRNA-seq) dataset based on the reference dataset HumanPrimaryCellAtlasData, and identify the cell types. Through the scType [ 31 ] method, using six datasets from different human and mouse tissues, as well as the Marker genes of each cell downloaded from the CellMarker 2.0 website [ 32 ] ( http://117.50.127.228/CellMarker/ ), perform unsupervised cell clustering and cell annotation; download the Tabula Sapiens- Uterus dataset from the CELLxGENE platform [ 33 ] ( https://cellxgene.cziscience.com/ ) and perform cell identification based on the differentially expressed genes of each cell type in the dataset. Combine the above three cell type annotation methods to annotate the cells in different clusters, and present the proportion of each cell type in each sample in the form of a stacked bar chart to analyze the relationship between cell types and clusters.The single-cell data ( GSE214411 ) were subject to quality control, clustering and annotation, and the expression of key genes in cell types and AUCell scores were analyzed. Identify cell type differential genes (scDEGs), and conduct GO/KEGG enrichment analysis, pseudo-temporal analysis and cell communication analysis. Subtype analysis and GSVA enrichment were conducted on fibroblasts and pericytes. All analyses were based on R software and appropriate statistical methods (such as Wilcoxon test, Spearman correlation analysis, etc.) were used, and the significance threshold was set at p < 0.05. The disease groups of the integrated GEO dataset were divided into two groups, namely High expression and Low expression, based on the expression level of the gene HSD11B1. After conducting difference analysis, all genes were ranked according to the logFC value. Then, the R package clusterProfiler (Version 4.10.1) was used to perform GSEA on all genes in the integrated GEO dataset. The parameters used in this GSEA are as follows: The seed number is 2024. Each gene set contains a minimum of 10 genes and a maximum of 500 genes. Through Molecular Signatures Database (MSigDB) Database access to c2. All. V2024.1. Hs. Symbols. The GMT gene sets, enrichment of GSEA, The screening criterion of GSEA was adj. p  < 0.05, and the P-value correction method was Benjamini-Hochberg (BH). The transcription factor combined with IL-10 was predicted by bioinformatics means. The results of multiple prediction websites were intersected to obtain the transcription factor JUND. Primary normal endometrial stromal cells (NESC) and primary ectopic endometrial stromal cells (EESC) were purchased from the National Collection of Genuine Products for cell culture (Shanghai, China). All cells were cultured in DMEM/F12 medium. 10% fetal bovine serum and 1× penicillin-streptomycin were added to the cell culture medium. The cells were maintained at 37℃ and 5% CO 2 in a humidified incubator. All the cells were purchased from the National Collection ofAuthenticated Cell Cultures( https://www.cellbank.org.cn/ ). RNA was extracted using the Trizol method. For every 1 × 10⁷ cells, 1 mL of Trizol was added for resuspend. After chloroform oscillation and stratification, the cells were centrifuged at 2000 RPM at 4℃ for 20 min to obtain the supernatant. Isopropyl alcohol was added to precipitate RNA. After washing with 75% ethanol, it was dissolved in 20 µl DEPC. The RNA concentration (OD260/280 = 1.8-2.0) was detected and diluted to the working concentration. Subsequently, the reverse transcription reaction was carried out according to the instructions of the reverse transcription kit. After qPCR was completed, with GAPDH as the internal reference, the relative expression level was analyzed by the 2-△△ct method. Dissolve 30 mg of BSA in 1.2 ml of the prepared solution and dilute it to 0.5 mg/ml with PBS when in use. Preparation of BCA working solution: Mix reagent A and B at a ratio of 50:1. Add the standard and the sample to be tested to a 96-well plate, with 200 µl of BCA working solution added to each well. Incubate at 37℃ for 20–30 min. Use an enzyme-linked immunosorbent assay (ELISA) reader to detect A562nm, and calculate the concentration based on the standard curve. Prepare 10% and 12.5% separation gels according to the instructions of the kit. Pour the 5% concentrated gel onto the upper edge of the glass plate, insert the comb to solidify, add 1× electrophoresis buffer, add 5 µl of Marker and sample protein to the Wells for electrophoresis and membrane transfer, seal with a 5% skimmed milk powder shaking table for 1–2 h, wash with TBST three times, and incubate the primary antibody at 4 ℃ overnight. After the secondary antibody was incubated and developed, the developing solution covered the band, and the image was exposed. Cells were inoculated in 96-well plates at a density of 5 × 10 4 cells per well, and 10mL of CCK-8 solution was added to each well. The optical density (OD) value at 450 nm was determined using an ELISA reader to detect the cell survival rate. The iron ion level in cells was accurately evaluated using the FenroOrange detection kit and confocal microscopy. After washing the treated cells three times with PBS, FerroOrange fluorescent probes were added to each well. The cells were incubated in the dark at room temperature for 20 min, and then the iron ion levels within the cells were observed using a confocal microscope. Discard the cell culture supernatant, use a cell scraper to scrape off the cells, then transfer the cells to a centrifuge tube with a pipette. Add 0.5 ml of the extract and mix well for 2 min. After that, break the cells to form a suspension and take 0.1 ml of the sample in a 1.5 ml centrifuge tube. Add the corresponding reagents in sequence according to the steps, cover the lid, mix well with a vortex mixer, and then treat in a 95 ℃ water bath for more than 40 min. After taking it out, cool it with running water, centrifuge at 4000 revolutions per minute for 10 min, and then measure the absorbance at 530 nm. Scan the empty plate of the microplate. Accurately draw 0.25 milliliters of the famous tube reaction solution and add it to a new 96-well plate. Use an microplate reader to measure the absorbance of each well. The collected cells are washed 1 to 2 times with PBS, then the precipitated cells are collected by low-speed centrifugation, and finally the cells are suspended with PBS buffer. Then, use ultrasonic disruption to break the cells. Take 0.1 ml of the cell suspension, add 0.1 mL of reagent and mix well. Centrifuge at 3500 rpm for 10 min and take the supernatant for detection. The supernatant obtained after treating different samples with reagent I shall be operated in accordance with the operation table. Mix well, let it stand for 5 min, and measure the absorbance at 405 nm. Prepare all types of cells to be tested. Dilute DCFH-DA with serum-free culture medium at a ratio of 1:1000 to make the final concentration 10µM. For adherent cells, remove the cell culture medium, add an appropriate volume of diluted DCFH-DA, incubate at 37℃ in the dark for 30 min, wash the cells three times with serum-free cell culture medium, and then digest and collect for the next step. For the suspended cells, centrifuge to collect the cell precipitate, resuspend it with an appropriate amount of diluted DCFH-DA, incubate at 37℃ in the dark for 30 min, gently pipette every 3–5 min to ensure full contact between the probe and the cells, and then wash the cells three times with serum-free cell culture medium to remove the DCFH-DA that has not entered the cells. Samples were analyzed using the flow cytometer ThermoFisher Attune NxT(Thermo Scientific, Waltham, MA, USA), and data were analyzed using Flowjo 10.5.3 software (TreeStar, shland, OR, USA). Accurately assess the lipid peroxidation levels in cells using the Liperlo detection kit and confocal microscopy. After washing the treated cells three times with PBS, add Liperfluo fluorescent probes (1MM) to each well. Incubate the cells in the dark at room temperature for 30 min. After the incubation was completed, the cells were washed three more times with PBS, and the lipid peroxide levels were observed using confocal microscopy. Cells cultured in 6-well plates were fixed in PBS solution containing 2.5% glutaraldehyde for 24 h. After washing in 0.1 M PBS, the cells were treated with tannic acid buffered with 0.1% Millipore-filtered dimethyl arsenate, fixed with 1% osmium buffered, and stained with 1% Millipore-filtered dioxyuranium acetate. After dehydration and embedding, the samples were incubated in an oven at 60℃ for 24 h. Digital images were obtained using a transmission electron microscope. Mitochondrial density was quantitatively analyzed by mageJ software. Use websites such as genomatix/JASPAR to search for the binding sites of the promoter region of JUND and IL-10. Then, the promoter region after the binding site mutation was inserted into the 5 ‘end of the luciferase reporter gene, denoted as IL-10-Pro-MUt, and the corresponding wild type was denoted as IL-10-pro-WT. The effect of overexpression of IL-10 on the expression of the luciferase reporter gene was detected. The 3’utr region of the target gene JUND was constructed to the 3’ end of the firefly luciferase reporter gene, and co-transfected into cells with the JUND expression vector and the sea renal luciferase reporter gene vector. Two substrates, luciferin, were added successively. The regulatory effect of JUND on the target gene was confirmed by detecting the expression level of the luciferase reporter gene to verify the targeted interaction between JUND and IL-10. Cross-linked cells with 1% formaldehyde at 37℃ for 15 min, then neutralized with 125 mM glycine for 5 min, washed with PBS, scraped cells and stored at -80℃. After the cell precipitate was lysed in an ice bath with SDS lysis buffer containing PMSF, ultrasonic fragmentation was performed. The supernatant was taken after centrifugation of the lysis buffer and diluted to 2 mL with ChIP Dilution buffer, with 200 µL reserved as the Input control. After the samples were pre-cleared by Protein A/G Agarose, AR antibody or IgG control was added and incubated overnight at 4℃. Subsequently, the complexes were precipitated with Protein A/G Agarose and washed. After the complex was eluted by Elution Buffer, NaCl was added and de-cross-linked at 65℃ for 4 h, and then DNA was purified by protease K treatment. The purified DNA samples were detected according to the qPCR experimental method. Rinse the required tissue with PBS and then add it to 4% paraformaldehyde for fixation overnight. Discard 4% paraformaldehyde, rinse the tissue three times with distilled water, and dehydrate it with alcohol of different concentrations following the procedure :50% alcohol for 1 h, 75% alcohol for 1 h, 85% alcohol for 1 h, 95% alcohol for 1 h, and anhydrous ethanol I and II each for 1 h. The anhydrous ethanol was mixed with xylene in a 1:1 ratio. The tissue was transferred to this solution and soaked for 30 min, then placed in xylene I and xylene for 30 min each. The tissue was transferred from xylene II to Wax No. 1 and placed in an oven at 65℃ for 1 h, then transferred to Wax No. 2 and placed in an oven at 65℃ overnight. The next day, it was transferred to a 65℃ oven with No. 3 wax for 1 h. The wax block was cut into 4-micron paraffin tissue sections using a slicer. The paraffin sections were spread out in water at 42℃ and lifted with a slide. Then, they were baked in a constant temperature oven at 60℃ for 2 h. Place the required paraffin sections at 65℃ for 1 h, then restore to room temperature. Xylene and anhydrous ethanol are mixed in a 1:1 ratio. The slices are transferred into the liquid for 7 min, and then transferred to xylene I and xylene I for 7 min each. Rehydrate the slices according to the following procedures: 100% alcohol for 3 min, 95% alcohol for 3 min, 85% alcohol for 3 min, 75% alcohol for 3 min, 50% alcohol for 3 min, and ddH 2 O for 3 min. The slices were placed in the EDTA repair solution for 25 min and then washed with PBS. Absorb the moisture with absorbent paper. Drip 5% sheep serum into the tissue and avoid light for 30 min. Discard the sheep serum, add the primary antibody, and incubate in a wet box at 4℃ overnight. Discard the primary antibody, wash with PBS for 5 min three times, add the secondary antibody dropwise, incubate at warm temperature for 30 min, aspirate and discard the secondary antibody, wash with PBS for 3 min three times. Add DAB solution in the dark, incubate for 10 min, and stop staining with ddH 2 O. Su added wood essence to stain the cell nuclei, incubated for 10 min, ddH 2 O stopped staining, and then placed in water for bluing for about 1 h. After that, dehydration and sealing were carried out. During laparoscopic surgery, ectopic endometrial tissue samples were collected from 15 patients with endometriosis. Control samples were taken from normal endometrial tissue from 15 women with tubal factors associated with infertility. All procedures were approved by the Human Ethics Committee of Lanzhou University First Hospital. We obtained 6-8-week-old female BALB/c mice without specific pathogens from the Laboratory Animal Center of Lanzhou University Medical College and fed them for one week. Subcutaneously inject estradiol (100 µg/kg body weight) once a week until the end of the experiment. On the day of modeling, prepare the media containing PBS on both sides of the uterine horns of mice, and remove the fat and mesangial tissues, etc. The uterus of the donor mouse was incised and cut into small pieces of about 1–2 millimeters. There were 10 recipient mice in each group, and the ratio of donor mice to recipient mice was 1:2. The uterine fragments were divided into several parts of the same weight and injected into the abdominal cavity of the recipient mice. Two weeks later, the mice were sacrificed, and the endometrial tissues of the normal control group mice and the abdominal ectopic endometrial tissues of the EMS group mice were taken for further experiments. All animal experiments were approved by the First Hospital of Lanzhou University Research Ethics Committee. The Prussian blue-stained images were processed and quantified using Image processing software, and the data were statistically analyzed using GraphPadPrism9.0.0 software. Normality and continuity tests were conducted on all data before differential comparison. Independent sample t-test was used for data comparison between the two groups, and Spearman analysis was used for correlation analysis. Results with a p value < 0.05 were considered statistically significant (* p  < 0.05.** p  < 0.01.*** p  < 0.001), while ns showed no statistically significant difference. All data were expressed as mean plus standard deviation, and each experiment was repeated at least three times.

To further determine the key biomarkers of ferroptosis used for the diagnosis of EMs, after taking the intersection of FRGs collected from the FerrDb database and the MSigDB database, a total of 50 ferroptosis-related genes (Table S1) were obtained. WGCNA analysis was performed on the genes with the top 50% difference in the EMs samples. The genes with the top 50% difference in the cluster tree were clustered and the grouping information was marked (Fig. S1). Based on this, a co-expression network was constructed, and 13 modules were obtained (Fig. S1). Among them, modules MEpurple, MEblue, and Module MEtan were significantly positively correlated with the ferroptosis score (Fig.  1 A), and were used as Module Genes for subsequent analysis. After integrating the difference analysis of the GEO dataset and taking the intersection, 10 differentially expressed FRDEGs were obtained: HSD11B1, SERPINE1, FKBP5, ALDH1A3, FN1, MAP3K8, CCDC102B, CEL, RAMP1, EMX2 (Fig.  1 B).Four key genes were screened out through multiple algorithms, and their diagnostic value was verified by using various machine learning models (Fig. S2). As shown by the inverse cumulative distribution map of residuals and the box plot, the residuals of the XGB model were the lowest (Fig. S3). Based on five-fold cross-validation, the PRC and ROC curves were plotted. The results showed that the diagnostic efficacy of the XGB model was the best (Fig.  1 C). The results of univariate logistic regression indicated that HSD11B1, FKBP5, and CCDC102B significantly promoted the occurrence of the disease, while EMX2 significantly inhibited the occurrence of the disease (Fig. S4). The results of multivariate logistic regression indicated that HSD11B1 was an independent diagnostic factor (Fig.  1 D). The Nomogram results indicated that the expression level of HSD11B1 had a significantly higher utility for the EMs diagnostic model than that of other variables (Fig.  1 , E and F). The accuracy is above 0.9. Fig. 1 Screening of key genes and bioinformatics analysis. A Correlation analysis results between the gene clustering modules of the top 50% variance and the ferroptosis score. Color indicates correlation. Red indicates positive correlation and blue indicates negative correlation. The darker the color, the greater the absolute value of the correlation; the lighter the color, the smaller the absolute value of the correlation. B The heat map of FRDEGs in the integrated GEO dataset. The blue samples are normal samples and the red samples are endometriosis samples. The colors on a heat map represent the expression levels. The redder the color, the higher the expression level; the bluer the color, the lower the expression level. C ROC curves of six machine learning algorithms. D Forest plots of the four Key Genes included in the multivariate logistic regression models in the EMs diagnostic model. E Key Genes in the EMs diagnostic model. F Linear Predictor in the integrated GEO dataset. The horizontal axis represents the false positive rate (FPR), and the vertical axis represents the true positive rate (TPR). The closer the AUC is to 1, the better the diagnostic effect is. When the AUC is above 0.9, there is relatively high accuracy Screening of key genes and bioinformatics analysis. A Correlation analysis results between the gene clustering modules of the top 50% variance and the ferroptosis score. Color indicates correlation. Red indicates positive correlation and blue indicates negative correlation. The darker the color, the greater the absolute value of the correlation; the lighter the color, the smaller the absolute value of the correlation. B The heat map of FRDEGs in the integrated GEO dataset. The blue samples are normal samples and the red samples are endometriosis samples. The colors on a heat map represent the expression levels. The redder the color, the higher the expression level; the bluer the color, the lower the expression level. C ROC curves of six machine learning algorithms. D Forest plots of the four Key Genes included in the multivariate logistic regression models in the EMs diagnostic model. E Key Genes in the EMs diagnostic model. F Linear Predictor in the integrated GEO dataset. The horizontal axis represents the false positive rate (FPR), and the vertical axis represents the true positive rate (TPR). The closer the AUC is to 1, the better the diagnostic effect is. When the AUC is above 0.9, there is relatively high accuracy To determine the expression of key genes in various types of cells, we analyzed the single-cell sequencing dataset of normal endometrial tissues and EMs tissues [ 17 ]. Through quality control and batch effect correction, we obtained 67,827 qualified cells and annotated and integrated the cell types (Fig. S5). Ultimately, we identified 14 cell clusters, including Ciliated Cells, Endothelial Cells, Epithelial Cells, Fibroblasts, and Leukocytes. Macrophages, Mast Cells, Megakaryocyte-Erythrocyte Progenitors (MEP), Mesenchymal Stem Cells MSC, Neutrophils, Natural Killer Cells (NK Cells), pericytes, Stem Cells, T Cells And draw the clustering maps of each sample cell (Fig. 2 A). Then, the cell clustering plots of the EMs samples and the Control samples were drawn respectively (Fig. 2 , B and C), as well as the accumulation bar plots of 14 cell types in all samples (Fig. 2 D). In the single-cell dataset, the AUCell score indicated that the key genes were mainly active in fibroblasts and pericytes (Fig. 2 , E to G). Fig. 2 Single-cell analysis of key genes. A to C Cell clustering UMAP of different cell types in the scRNA-seq dataset. The UMAP map of cell clustering of different cell types in EMs samples and normal samples. D Bar chart of cell proportions in different samples. E and F . AUCell analysis of the gene set of Key Genes in the scRNA-seq dataset. The redder the color is, the higher the AUCell value is. G AUCell plots of high and low AUC value groups. Violin plots of K. AUCell values in 14 cell types. The horizontal axis represents the cell type, and the vertical axis represents the AUCell value Single-cell analysis of key genes. A to C Cell clustering UMAP of different cell types in the scRNA-seq dataset. The UMAP map of cell clustering of different cell types in EMs samples and normal samples. D Bar chart of cell proportions in different samples. E and F . AUCell analysis of the gene set of Key Genes in the scRNA-seq dataset. The redder the color is, the higher the AUCell value is. G AUCell plots of high and low AUC value groups. Violin plots of K. AUCell values in 14 cell types. The horizontal axis represents the cell type, and the vertical axis represents the AUCell value To further confirm the high expression of HSD11B1 we examined the expression level of HSD11B1 primary NESC. In PCR and WB results showed that HSD11B1 was significantly highly expressed in primary EESC compared with NESC (Fig.  3 , A and B). To explore the role of HSD11B1 in ferroptosis, we induced ferroptosis in EESC cells using Erastin and observed its effect on ferroptosis by overexpression or knockdown of HSD11B1. The results showed that compared with the control group, Erastin treatment reduced cell survival rate, GSH level and SLC7A11/GPX4 protein expression, while increasing Fe 2 + content, lipid peroxidation and MDA level, and led to mitochondrial damage, volume reduction, cristae reduction, membrane density increase. After overexpression of HSD11B1 in EESCs treated with Erastin, the cell survival rate, GSH level and SLC7A11/GPX4 protein expression were all the same, and the Fe 2 + content, lipid peroxidation and MDA level decreased, resulting in a reduction in the degree of mitochondrial damage. Overexpression of HSD11B1 can reverse the above-mentioned Erastin-induced ferroptosis phenotype. To further confirm the role of HSD11B1 in ferroptosis, we knocked down HSD11B1 in EESC, and the results showed that compared with the control group, the cell survival rate and GSH level of EESC treated with Erastin after knockdown of HSD11B1 decreased significantly, while the content of Fe 2 + , lipid peroxidation and MDA level increased significantly, the protein expression of SLC7A11/GPX4 decreased, and the degree of mitochondrial damage increased (Fig.  3 , C to I). Knockdown of HSD11B1 aggravates these changes. In conclusion, HSD11B1 inhibits the ferroptosis phenotype. Fig. 3 HSD11B1 is highly expressed in EMs and inhibits ferroptosis of stromal cells. A The RT-qPCR results indicated that HSD11B1 was highly expressed in EESC. B The WB results indicated that HSD11B1 highly expressed. C to I in EESC. The experimental results showed that compared with the control group, Erastin treatment reduced cell survival rate, GSH level and SLC7A11/GPX4 protein expression, while increasing Fe²⁺ content, lipid peroxidation and MDA level, and led to mitochondrial damage. Including volume reduction, ridge reduction and membrane density increase. J Effects of HSD11B1 overexpression and knockdown on relative cellular ATP levels. (* p  < 0.05, ** p  < 0.01) HSD11B1 is highly expressed in EMs and inhibits ferroptosis of stromal cells. A The RT-qPCR results indicated that HSD11B1 was highly expressed in EESC. B The WB results indicated that HSD11B1 highly expressed. C to I in EESC. The experimental results showed that compared with the control group, Erastin treatment reduced cell survival rate, GSH level and SLC7A11/GPX4 protein expression, while increasing Fe²⁺ content, lipid peroxidation and MDA level, and led to mitochondrial damage. Including volume reduction, ridge reduction and membrane density increase. J Effects of HSD11B1 overexpression and knockdown on relative cellular ATP levels. (* p  < 0.05, ** p  < 0.01) To determine the impact of high expression of HSD11B1 on endometriosis, the expression of all genes in the integrated GEO dataset and the involved biological processes were studied through gene set enrichment analysis (GSEA). The results showed that in the integrated GEO dataset, the gene associated with ferroptosis, highly expressed in EMs and upregulated by HSD11B1 was IL-10 (Fig.  4 A). Subsequently, we conducted PCR and WB experiments for verification. The results showed that compared with NESC, IL-10 was significantly highly expressed in EESC (Fig.  4 , B and C). To further determine the effect of HSD11B1 regulating IL-10 on ferroptosis, we knocked down or overexpressed HSD11B1 and IL-10 when using Erastin to induce ferroptosis in EESC. The results indicated that overexpression of HSD11B1 could inhibit Erastin-induced ferroptosis, manifested as increased cell survival rate and GSH levels, decreased Fe 2 + content, lipid peroxidation and MDA levels, upregulated expression of SLC7A11/GPX4, and alleviated mitochondrial damage. However, when IL-10 was knocked down, cell survival rate and GSH level decreased, while Fe 2 + content, lipid peroxidation and MDA level increased. A reduction in the expression levels of SLC7A11 and GPX4 was observed, accompanied by aggravated mitochondrial damage (Fig.  4 , D to J). In contrast, knockdown of HSD11B1 significantly enhanced Erastin-induced ferroptosis. Conversely, IL-10 overexpression increased cell survival, elevated GSH levels, and upregulated the expression of SLC7A11 and GPX4. It also led to a reduction in Fe 2 + accumulation, lipid peroxidation, and MDA content, thereby mitigating mitochondrial damage (Fig.  5 , A to G). These findings collectively suggest that HSD11B1 modulates ferroptosis resistance in EESCs via an IL-10-dependent signaling pathway. Fig. 4 HSD11B1 promotes the expression of IL-10 and inhibits ferroptosis. A Venn diagram screened out genes that were differentially expressed in EMs, associated with ferroptosis, and regulated by HSD11B1. B The WB results showed that, compared with NESC, IL-10 was significantly highly expressed in EESC. C PCR results showed that, compared with NESC, IL-10 was significantly highly expressed in EESC. D Erastin reduces cell survival rate. Overexpression of HSD11B1 can exacerbate the decline in cell survival rate, and knockdown of IL-10 can increase cell survival rate. E Erastin reduces the GSH level of cells. Overexpression of HSD11B1 can aggravate the decrease of GSH level, and knockdown of IL-10 can increase the GSH level. F Erastin increases the Fe 2+ level in cells. Overexpression of HSD11B1 can promote the increase of Fe 2+ level, and knockdown of IL-10 can reduce the Fe 2+ level. G Erastin increases the MDA level in cells. Overexpression of HSD11B1 can promote the increase of MDA level, and knockdown of IL-10 can reduce the MDA level. H Erastin increases the fluorescence intensity of lipid ROS in cells. Overexpression of HSD11B1 can promote the increase of lipid ROS fluorescence intensity, and knockdown of IL-10 can reduce lipid ROS fluorescence intensity. I Erastin reduces the protein levels of SLC7A11 and GPX4. Overexpression of HSD11B1 can significantly reduce the expression levels of SLC7A11 and GPX4, and knockdown of IL-10 can increase the expression levels of SLC7A11 and GPX4. J Relative cellular ATP levels under Erastin treatment with co-intervention of HSD11B1 overexpression and IL-10 knockdown. K Erastin increases mitochondrial damage in cells. Overexpression of HSD11B1 can increase the degree of mitochondrial damage, and knockdown of IL-10 can alleviate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) HSD11B1 promotes the expression of IL-10 and inhibits ferroptosis. A Venn diagram screened out genes that were differentially expressed in EMs, associated with ferroptosis, and regulated by HSD11B1. B The WB results showed that, compared with NESC, IL-10 was significantly highly expressed in EESC. C PCR results showed that, compared with NESC, IL-10 was significantly highly expressed in EESC. D Erastin reduces cell survival rate. Overexpression of HSD11B1 can exacerbate the decline in cell survival rate, and knockdown of IL-10 can increase cell survival rate. E Erastin reduces the GSH level of cells. Overexpression of HSD11B1 can aggravate the decrease of GSH level, and knockdown of IL-10 can increase the GSH level. F Erastin increases the Fe 2+ level in cells. Overexpression of HSD11B1 can promote the increase of Fe 2+ level, and knockdown of IL-10 can reduce the Fe 2+ level. G Erastin increases the MDA level in cells. Overexpression of HSD11B1 can promote the increase of MDA level, and knockdown of IL-10 can reduce the MDA level. H Erastin increases the fluorescence intensity of lipid ROS in cells. Overexpression of HSD11B1 can promote the increase of lipid ROS fluorescence intensity, and knockdown of IL-10 can reduce lipid ROS fluorescence intensity. I Erastin reduces the protein levels of SLC7A11 and GPX4. Overexpression of HSD11B1 can significantly reduce the expression levels of SLC7A11 and GPX4, and knockdown of IL-10 can increase the expression levels of SLC7A11 and GPX4. J Relative cellular ATP levels under Erastin treatment with co-intervention of HSD11B1 overexpression and IL-10 knockdown. K Erastin increases mitochondrial damage in cells. Overexpression of HSD11B1 can increase the degree of mitochondrial damage, and knockdown of IL-10 can alleviate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) Fig. 5 HSD11B1 promotes the expression of IL-10 and inhibits ferroptosis. A Knockout of HSD11B1 can increase cell survival rate, while overexpression of IL-10 can reduce cell survival rate. B Knockdown of HSD11B1 can increase GSH levels, while overexpression of IL-10 can decrease GSH levels. C Knockdown of HSD11B1 can reduce the level of Fe 2+ , while overexpression of IL-10 can increase the level of Fe 2+ . D Knockdown of HSD11B1 can reduce the MDA level, while overexpression of IL-10 can increase the MDA level. E Knockdown of HSD11B1 can reduce the fluorescence intensity of lipid ROS, while overexpression of IL-10 can increase the fluorescence intensity of lipid ROS. F Knockdown of HSD11B1 can significantly increase the expression levels of SLC7A11 and GPX4, while overexpression of IL-10 can reduce the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of HSD11B1 silencing and JUND overexpression. H Knockdown of HSD11B1 overexpression can alleviate the degree of mitochondrial damage, while overexpression of IL-10 can aggravate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) HSD11B1 promotes the expression of IL-10 and inhibits ferroptosis. A Knockout of HSD11B1 can increase cell survival rate, while overexpression of IL-10 can reduce cell survival rate. B Knockdown of HSD11B1 can increase GSH levels, while overexpression of IL-10 can decrease GSH levels. C Knockdown of HSD11B1 can reduce the level of Fe 2+ , while overexpression of IL-10 can increase the level of Fe 2+ . D Knockdown of HSD11B1 can reduce the MDA level, while overexpression of IL-10 can increase the MDA level. E Knockdown of HSD11B1 can reduce the fluorescence intensity of lipid ROS, while overexpression of IL-10 can increase the fluorescence intensity of lipid ROS. F Knockdown of HSD11B1 can significantly increase the expression levels of SLC7A11 and GPX4, while overexpression of IL-10 can reduce the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of HSD11B1 silencing and JUND overexpression. H Knockdown of HSD11B1 overexpression can alleviate the degree of mitochondrial damage, while overexpression of IL-10 can aggravate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) To explore how HSD11B1 regulates IL-10, we selected the IL-10 transcription initiation site and used UCSC’s Jaspar TFBS hub to predict transcription factors. We found that the transcription factor JUND could potentially regulate IL-10 (Fig.  6 A). To verify the expression of JUND, we conducted PCR and WB experiments. The results showed that compared with NESC, JUND was significantly highly expressed in EESC (Fig.  6 , B and C). Subsequently, in order to further explore the interaction between JUND and IL-10, we employed PCR and dual-luciferase reporter gene assays. The results indicated that JUND transcriptively activated IL-10 (Fig.  6 , D and E). In addition, we adopted the ChIP-qPCR experiment, and the results indicated that the enrichment degree of JUND in the IL-10 promoter region was relatively high (Fig.  6 F). The above experimental results indicate that JUND can transcriptionally activate IL-10. Fig. 6 IL-10 is transcribed and activated by JUND. A IL-10 binding motifpredicted by JASPAR. B WB results showed that compared with NESC, JUND was significantly highly expressed in EESC. C PCR showed that compared with NESC, JUND was significantly highly expressed in EESC. D PCR results indicated that the expression level of IL-10 in EESC decreased after knockdown of JUND, while the expression level of IL-10 in EESC increased after overexpression of JUND. E Insert the promoter region after the binding site mutation into the 5 ‘end of the luciferase reporter gene and mark it as IL-10-pro-mut, and the corresponding wild type is marked as IL-10-pro-wt. The results indicated that after the mutation of the binding site, the effect of JUND on luciferase disappeared. F CHIP-qPCR indicated that the enrichment degree of JUND in the IL-10 promoter region increased. G to H WB and PCR indicated that the expression level of JUND in EESC increased after overexpression of HSD11B1. I and J WB and PCR indicated that the expression level of JUND in EESC decreased after knockdown of HSD11B1. K The PCR results showed that when HSD11B1 was overexpressed and JUND was knocked down, the expression level of IL-10 decreased, and when HSD11B1 was knocked down and JUND was overexpressed, the expression level of IL-10 increased. (** p  < 0.01) IL-10 is transcribed and activated by JUND. A IL-10 binding motifpredicted by JASPAR. B WB results showed that compared with NESC, JUND was significantly highly expressed in EESC. C PCR showed that compared with NESC, JUND was significantly highly expressed in EESC. D PCR results indicated that the expression level of IL-10 in EESC decreased after knockdown of JUND, while the expression level of IL-10 in EESC increased after overexpression of JUND. E Insert the promoter region after the binding site mutation into the 5 ‘end of the luciferase reporter gene and mark it as IL-10-pro-mut, and the corresponding wild type is marked as IL-10-pro-wt. The results indicated that after the mutation of the binding site, the effect of JUND on luciferase disappeared. F CHIP-qPCR indicated that the enrichment degree of JUND in the IL-10 promoter region increased. G to H WB and PCR indicated that the expression level of JUND in EESC increased after overexpression of HSD11B1. I and J WB and PCR indicated that the expression level of JUND in EESC decreased after knockdown of HSD11B1. K The PCR results showed that when HSD11B1 was overexpressed and JUND was knocked down, the expression level of IL-10 decreased, and when HSD11B1 was knocked down and JUND was overexpressed, the expression level of IL-10 increased. (** p  < 0.01) To explore whether HSD11B1 affects the expression level of IL-10 by regulating the transcription factor JUND, after knockdown or overexpression of HSD11B1 and JUND, we observed the expression of IL-10. The PCR experiment results showed that, compared with the control group, when HSD11B1 was overexpressed and JUND was knocked down simultaneously, the expression level of IL-10 decreased significantly. Conversely, when HSD11B1 was knocked down and JUND was overexpressed simultaneously, the expression level of IL-10 rebounded (Fig.  6 , G to K). This indicates that HSD11B1 can regulate the transcriptional activation of IL-10 by JUND. To explore the regulatory effects of JUND and IL-10 on ferroptosis, we used the ferroptosis inducer Erastion to knock down or overexpress JUND and IL-10 respectively, and observed the changes of ferroptosis in EESC. The results showed that overexpression of JUND could inhibit Erastin-induced iron apoptosis, increase cell survival rate and GSH level, reduce Fe 2+ content, lipid peroxidation and MDA level, upregulate the expression of SLC7A11/GPX4 protein, and alleviate mitochondrial damage. However, knockdown of IL-10 can partially reverse the protective effect of JUND, reducing cell survival rate and GSH level, increasing Fe 2+ content, lipid peroxidation and MDA level, decreasing the expression of SLC7A11/GPX4 protein, and aggravating mitochondrial damage (Fig.  7 , A to G). Conversely, knockdown of JUND aggravates erastin-induced ferroptosis, while overexpression of IL-10 can partially salvage this phenotype, manifested as increased cell survival rate and GSH level, decreased Fe 2+ content, lipid peroxidation and MDA level, upregulated expression of SLC7A11/GPX4 protein, and alleviated mitochondrial damage (Fig.  8 , A to G). The above results indicate that JUND transcriptional activation of IL-10 regulates ferroptosis in EESC. Fig. 7 JUND transcriptional activation of IL-10 regulates ferroptosis. A Erastin reduces cell survival rate. Overexpression of JUND can exacerbate the decline in cell survival rate, and knockdown of IL-10 can increase cell survival rate. B Erastin reduces the GSH level of cells. Overexpression of JUND can aggravate the decrease of GSH level, and knockdown of IL-10 can increase the GSH level. C Erastin increases the Fe 2+ level in cells. Overexpression of JUND can promote the increase of Fe 2+ level, and knockdown of IL-10 can reduce the Fe 2+ level. D Erastin increases the MDA level in cells. Overexpression of JUND can promote the increase of MDA level, and knockdown of IL-10 can reduce the MDA level. E Erastin increases the fluorescence intensity of lipid ROS in cells. Overexpression of JUND can promote the increase of lipid ROS fluorescence intensity, and knockdown of IL-10 can reduce lipid ROS fluorescence intensity. F Erastin reduces the protein levels of SLC7A11 and GPX4. Overexpression of JUND can significantly reduce the expression levels of SLC7A11 and GPX4, and knockdown of IL-10 can increase the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of JUND overexpression and IL-10 silencing. H Erastin increases mitochondrial damage in cells. Overexpression of JUND can increase the degree of mitochondrial damage, and knockdown of IL-10 can alleviate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) JUND transcriptional activation of IL-10 regulates ferroptosis. A Erastin reduces cell survival rate. Overexpression of JUND can exacerbate the decline in cell survival rate, and knockdown of IL-10 can increase cell survival rate. B Erastin reduces the GSH level of cells. Overexpression of JUND can aggravate the decrease of GSH level, and knockdown of IL-10 can increase the GSH level. C Erastin increases the Fe 2+ level in cells. Overexpression of JUND can promote the increase of Fe 2+ level, and knockdown of IL-10 can reduce the Fe 2+ level. D Erastin increases the MDA level in cells. Overexpression of JUND can promote the increase of MDA level, and knockdown of IL-10 can reduce the MDA level. E Erastin increases the fluorescence intensity of lipid ROS in cells. Overexpression of JUND can promote the increase of lipid ROS fluorescence intensity, and knockdown of IL-10 can reduce lipid ROS fluorescence intensity. F Erastin reduces the protein levels of SLC7A11 and GPX4. Overexpression of JUND can significantly reduce the expression levels of SLC7A11 and GPX4, and knockdown of IL-10 can increase the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of JUND overexpression and IL-10 silencing. H Erastin increases mitochondrial damage in cells. Overexpression of JUND can increase the degree of mitochondrial damage, and knockdown of IL-10 can alleviate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) Fig. 8 JUND transcriptional activation of IL-10 regulates ferroptosis. A Knockout of JUND can increase cell survival rate, while overexpression of IL-10 can reduce cell survival rate. B Knockdown of JUND can increase GSH levels, while overexpression of IL-10 can decrease GSH levels. C Knockdown of JUND can reduce the level of Fe 2+ , while overexpression of IL-10 can increase the level of Fe 2+ . D Knockdown of JUND can reduce the MDA level, while overexpression of IL-10 can increase the MDA level. E Knockdown of JUND can reduce the fluorescence intensity of lipid ROS, while overexpression of IL-10 can increase the fluorescence intensity of lipid ROS. F Knockdown of JUND can significantly increase the expression levels of SLC7A11 and GPX4, while overexpression of IL-10 can reduce the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of JUND overexpression and IL-10 silencing. H Knockdown of JUND overexpression can alleviate the degree of mitochondrial damage, while overexpression of IL-10 can aggravate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) JUND transcriptional activation of IL-10 regulates ferroptosis. A Knockout of JUND can increase cell survival rate, while overexpression of IL-10 can reduce cell survival rate. B Knockdown of JUND can increase GSH levels, while overexpression of IL-10 can decrease GSH levels. C Knockdown of JUND can reduce the level of Fe 2+ , while overexpression of IL-10 can increase the level of Fe 2+ . D Knockdown of JUND can reduce the MDA level, while overexpression of IL-10 can increase the MDA level. E Knockdown of JUND can reduce the fluorescence intensity of lipid ROS, while overexpression of IL-10 can increase the fluorescence intensity of lipid ROS. F Knockdown of JUND can significantly increase the expression levels of SLC7A11 and GPX4, while overexpression of IL-10 can reduce the expression levels of SLC7A11 and GPX4. G Relative cellular ATP levels under Erastin treatment following co-intervention of JUND overexpression and IL-10 silencing. H Knockdown of JUND overexpression can alleviate the degree of mitochondrial damage, while overexpression of IL-10 can aggravate the degree of mitochondrial damage. (* p  < 0.05, ** p  < 0.01) To further confirm the involvement of HSD11B1, JUND, and IL-10 in endometriosis, we examined their expression profiles in both an established EMs animal model and clinical specimens. Quantitative PCR and western blot analyses revealed markedly elevated expression levels of HSD11B1, JUND, and IL-10 in ectopic endometrial lesions compared to controls. Immunohistochemical staining further confirmed the spatial enrichment of these targets within ectopic tissues (Fig.  9 , A and B). These data indicate a potential regulatory role of HSD11B1, JUND, and IL-10 in the pathogenesis of EMs. To extend these findings to human disease, eutopic endometrial tissues from healthy individuals and ectopic lesions from EMs patients were analyzed. Consistent with the animal model, PCR, western blot, and IHC analyses demonstrated robust upregulation of HSD11B1/JUND/IL-10 in ectopic lesions (Fig.  9 , C and D). Notably, HSD11B1 exhibited prominent differential expression, underscoring its potential as a diagnostic biomarker and therapeutic target in EMs. Fig. 9 HSD11B1/JUND/IL-10 is highly expressed in EMs. A and B The results of WB and IHC experiments indicated that HSD11B1/JUND/IL-10 was highly expressed in the ectopic endometrium of the EMs mouse model. C and D WB and IHC experiments indicated that HSD11B1/JUND/IL-10 was highly expressed in the ectopic endometrium of EMs patients. (** p  < 0.01) HSD11B1/JUND/IL-10 is highly expressed in EMs. A and B The results of WB and IHC experiments indicated that HSD11B1/JUND/IL-10 was highly expressed in the ectopic endometrium of the EMs mouse model. C and D WB and IHC experiments indicated that HSD11B1/JUND/IL-10 was highly expressed in the ectopic endometrium of EMs patients. (** p  < 0.01)

Endometriosis (EMs) is a chronic, estrogen-dependent gynecological disorder affecting approximately 10% of women of reproductive age, with a global patient population exceeding 190 million [ 1 – 3 ]. Characterized by the ectopic implantation of endometrial-like tissue, EMs often leads to pelvic pain, dysmenorrhea, and infertility, substantially compromising patients’ quality of life. Despite its high prevalence and clinical burden, current diagnostic approaches remain invasive and often delayed, underscoring the urgent need for reliable non-invasive biomarkers and targeted therapeutic strategies. Ferroptosis, a recently characterized form of regulated cell death, is distinguished by iron-dependent lipid peroxidation and the catastrophic collapse of membrane integrity [ 4 – 6 ]. Accumulating evidence implicates ferroptosis in the pathophysiology of various diseases, including cancer, neurodegeneration, and inflammatory disorders [ 7 – 10 ]. In the context of EMs, aberrant iron metabolism has emerged as a defining feature of the disease microenvironment. Studies have demonstrated excessive iron accumulation in the peritoneal fluid, ectopic lesions, and activated macrophages of EMs patients, suggesting a potential link between iron overload and disease progression [ 11 , 12 ]. Furthermore, ferroptosis may modulate key cellular behaviors—such as migration, invasion, proliferation, and inflammatory responses—of endometrial cells [ 13 ]. However, the molecular mechanisms and regulatory networks governing ferroptosis in EMs remain largely elusive. 11β-Hydroxysteroid Dehydrogenase Type 1 (HSD11B1) is a membrane-associated enzyme comprising extracellular, transmembrane, and intracellular domains [ 14 ]. It is broadly expressed across tissues such as adipose, liver, ovary, and endometrium, and has been implicated in metabolic and inflammatory diseases [ 15 ]. Functionally, HSD11B1 catalyzes the interconversion of inactive cortisone to active cortisol, thus playing a pivotal role in local glucocorticoid metabolism [ 16 ]. In EMs, elevated glucocorticoid signaling and receptor expression have been associated with enhanced survival and inflammation in ectopic lesions [ 17 ], implying a potential regulatory role for HSD11B1. However, the involvement of HSD11B1 in EMs pathogenesis, particularly in the context of ferroptosis, remains unexplored. In this study, we employed integrative machine learning approaches and single-cell transcriptomic analysis to systematically investigate ferroptosis-related genes in EMs. Through the construction of diagnostic models, HSD11B1 was identified as a key molecular candidate with high diagnostic value. Functional validation in vitro and in vivo revealed that upregulated HSD11B1 promotes the transcriptional activation of JUND, which enhances IL-10 expression and suppresses ferroptosis in endometrial stromal cells, thereby facilitating disease progression. Our findings highlight HSD11B1 as a potential non-invasive biomarker and a promising therapeutic target for the clinical management of EMs.

Our findings reveal a previously unrecognized mechanism whereby HSD11B1 promotes the progression of endometriosis by suppressing ferroptosis. Mechanistically, HSD11B1 enhances the transcriptional activity of JUND, leading to upregulated IL-10 expression and a ferroptosis-resistant microenvironment that favors ectopic lesion survival and proliferation. This HSD11B1/JUND/IL-10 axis represents a critical regulatory pathway contributing to EMs pathogenesis. Importantly, the consistent overexpression of HSD11B1 in both animal models and clinical specimens highlights its potential as a diagnostic biomarker. These insights not only deepen our understanding of ferroptosis resistance in EMs but also lay the groundwork for the development of HSD11B1-targeted therapeutic strategies aimed at disrupting this pathogenic signaling cascade. The integration of machine learning approaches into the study of EMs holds immense promise for addressing longstanding clinical challenges, including delayed diagnosis, reliance on invasive procedures, and the lack of personalized therapeutic strategies. In this study, we employed five widely used machine learning algorithms to screen FRDEGs, revealing a robust ferroptosis-related molecular signature in EMs. Among them, the GMM outperformed traditional hard clustering methods such as K-means by capturing the intrinsic heterogeneity of endometrial lesions through probabilistic modeling of overlapping subgroups. To evaluate the diagnostic utility of the identified key genes, we constructed predictive models using six machine learning frameworks. Comparative assessments based on residual distribution plots, precision-recall curves, and ROC curves demonstrated that our models, particularly the XGB-based classifier, outperformed three previously reported EMs diagnostic models (Figure S6) [ 34 – 36 ]. These results underscore the superior diagnostic efficacy of the screened genes and their potential to inform the development of non-invasive early detection tools. Single-cell transcriptomic analysis further localized the expression of HSD11B1 to fibroblasts and pericytes within ectopic lesions, suggesting its involvement in both fibrosis and neovascularization. Given that HSD11B1 catalyzes the conversion of inactive cortisone to active cortisol [ 37 ], we speculate that it may promote TGF-β/Smad pathway activation, stimulate extracellular matrix remodeling, and enhance vascular stability, thereby fostering lesion persistence and expansion. These cellular functions of HSD11B1 align with its potential role as a disease driver and therapeutic target. Ferroptosis, an iron-dependent form of regulated cell death driven by lipid peroxidation, has emerged as a critical pathway implicated in the pathophysiology of EMs [ 38 ]. The periodic hemorrhage characteristic of ectopic lesions leads to erythrocyte lysis and local iron overload, triggering ROS production via the Fenton reaction and initiating ferroptotic stress [ 39 ]. Decreased activity of GPX4, a pivotal antioxidant enzyme that prevents ferroptosis by reducing lipid peroxides, has been reported in ectopic lesions, resulting in accumulation of cytotoxic products such as malondialdehyde and 4-hydroxynonenal (4-HNE) [ 40 – 42 ]. Our data-driven identification of a ferroptosis-associated gene module comprising ten genes (HSD11B1, SERPINE1, FKBP5, ALDH1A3, FN1, MAP3K8, CCDC102B, CEL, RAMP1, EMX2) enabled systematic modeling and ranking via multiple algorithms, including LASSO and elastic net regression. Four key genes (HSD11B1, FKBP5, CCDC102B, EMX2) emerged as robust diagnostic indicators, with HSD11B1 demonstrating particularly strong independent predictive power. The superior performance of the XGB model was validated through rigorous cross-validation and multi-metric evaluation. HSD11B1, also known as 11β-hydroxysteroid dehydrogenase type 1, regulates intracellular cortisol levels and modulates inflammation [ 43 ]. In the context of EMs, high HSD11B1 expression may facilitate M2 macrophage polarization, suppressing immune clearance and enabling ectopic cell survival. Previous reports have linked HSD11B1 to impaired dendritic cell maturation, further implicating it in immune evasion [ 44 ]. Consistently, our findings revealed significant HSD11B1 upregulation in EESCs, as confirmed by transcriptomic and protein analyses. Functionally, overexpression of HSD11B1 in EESCs suppressed Erastin-induced ferroptosis, suggesting a protective role. Gene set enrichment analysis of the GEO-integrated dataset identified IL-10 as a downstream effector significantly associated with HSD11B1 expression. IL-10, a potent anti-inflammatory cytokine, has recently been implicated in ferroptosis regulation, although its precise mechanisms remain incompletely understood. To elucidate the regulatory cascade, we investigated the transcriptional activation of IL-10. Motif analysis of the IL-10 promoter region identified JUND as a candidate transcription factor. The research by Wu et al. [ 45 ] was the first to directly demonstrate that IL-10 activates the STAT3 signaling pathway by binding to its receptor IL-10R, thereby upregulating the DLK1/AMPK/ACC axis, reprogramming lipid metabolism and reducing lipid reactive oxygen accumulation, thereby protecting oligodendrocyte progenitor cells from ferroptosis. This discovery provides direct experimental evidence for the IL-10-IL-10R-STAT3 axis in regulating ferroptosis. Further, STAT3, as a key regulatory factor in ferroptosis, has been widely proven to be active in various tumor models. STAT3 mediates ferroptosis resistance by transcriptionally activating the expression of core protective proteins such as SLC7A11, GPX4, and FTH1, while STAT3 inhibitors can reverse this effect. Based on the above evidence, we speculate that in endometrial stromal cells, IL-10 may activate STAT3 through IL-10R, upregulate the expression of molecules related to the ferroptosis defense system, thereby inhibiting ferroptosis and promoting the survival of endometriosis lesions. Consistent with this, both IL-10 and JUND were upregulated in EESCs. Dual-luciferase reporter assays and ChIP-qPCR confirmed that JUND directly binds to the IL-10 promoter and activates its transcription. These findings identify a novel HSD11B1/JUND/IL-10 axis that modulates ferroptosis in EMs.This study conducted immunohistochemical staining on complete endometriosis lesion tissue sections, aiming to present the overall expression profile of HSD11B1/JUND/IL-10 throughout the lesion microenvironment. It is noteworthy that, in addition to stromal cells, positive expression of HSD11B1 was also observed in epithelial cells. Although this study focused on the role of interstitial cell ferroptosis in disease progression, the expression of HSD11B1 in epithelial cells may reflect the adaptive regulation of local glucocorticoid metabolism in the lesion. This phenomenon suggests that the role of HSD11B1 in endometriosis may be cell type-specific, and the specific function of HSD11B1 in epithelial cells and its contribution to the disease process still need further research to clarify.The microenvironment of endometriosis lesions is characterized by chronic inflammation, local hypoxia, and estrogen enrichment. These factors may act synergistically to regulate the expression of HSD11B1 at the transcriptional and post-transcriptional levels. In terms of inflammatory regulation, studies have shown that the inflammatory factor TNF-α can significantly induce the mRNA and protein expression of HSD11B1 in endometriosis primary stromal cells, while inhibiting the expression of HSD11B2, thereby promoting the synthesis and action of local cortisol [ 46 ]. This indicates that the local inflammatory response in the lesion not only exists passively but may also form a positive feedback loop by upregulating HSD11B1, exacerbating the local immunosuppressive microenvironment. In terms of hypoxia regulation, the ectopic endometrium in the early stage of implantation is in an ischemic hypoxic state, which can induce the expression of hypoxia-inducible factor HIF-1α. Although there is no direct report of HIF-1α regulating HSD11B1, studies based on other disease models suggest that the hypoxia signaling pathway may participate in the transcriptional activation of HSD11B1 by binding to the hypoxia response elements in the promoter region of HSD11B1. This mechanism is worthy of further exploration in endometriosis. In terms of estrogen regulation, endometriosis is an estrogen-dependent disease, and the increased expression of aromatase in the lesion leads to a significantly higher concentration of estradiol than normal endometrium. Epigenetic changes have been proven to be able to affect the expression patterns of genes including HSD11B1, suggesting that the estrogen/ER signaling may indirectly participate in the transcriptional regulation of HSD11B1 by recruiting co-activators or altering chromatin states. In conclusion, the inflammatory, hypoxic, and high-estrogen microenvironment of endometriosis lesions may synergistically drive the abnormal upregulation of HSD11B1. In the future, it is necessary to systematically analyze the transcriptional regulatory networks of NF-κB, HIF-1α, ER, etc. in the promoter region of HSD11B1 to discover new targets for blocking this pathway. JUND, a component of the AP-1 transcription factor family, is known to regulate genes involved in redox homeostasis and iron metabolism, including GPX4 and SLC7A11 [ 47 – 49 ]. Cortisol, as an endogenous ligand of the glucocorticoid receptor (GR), can indirectly regulate the transcriptional activity of JUND through the classical GR signaling pathway [ 50 ]. Tang et al. [ 51 ] discovered through transcriptomic analysis that cortisol treatment significantly downregulated the expression of AP-1 family members jund and fosl1, and this effect could be blocked by the GR antagonist RU-486. Their further experiments confirmed that after AP-1 agonist ASLAN003 induction, cortisol treatment significantly downregulated the expression of jund and fosl1 induced by it, proving that cortisol regulates JUND transcription through GR. Mechanistically, activated GR can regulate gene transcription through two ways, directly binding to the glucocorticoid response element (GRE) in the promoter region of the target gene, or through a “restraint mechanism” interacting with the JUND-containing AP-1 complex to regulate its transcriptional activity. It is notable that JUND has a unique interaction with the glucocorticoid signaling in the AP-1 family. Marti’s classic study found that JUND can activate transcription in cells lacking functional GR through multiple GREs, while JUNC or JUNB do not have this effect, suggesting that there is a special regulatory mechanism between JUND and the GR signaling pathway. Additionally, Breslin et al. [ 52 ] confirmed that the GR promoter itself contains an AP-1 site, and this site preferentially binds to JUND rather than other Jun family members, further supporting the bidirectional regulation between GR and JUND. Based on the above evidence, we propose the potential mechanism of the HSD11B1-JUND regulatory axis. HSD11B1 catalyzes the generation of cortisol, which binds to GR, and activated GR can directly bind to the GRE in the promoter region of JUND to regulate its transcription, and can also regulate its transcriptional activity through interaction with the JUND-containing AP-1 complex. In our study, genetic manipulation of JUND expression modulated ferroptosis sensitivity in EESCs, JUND knockdown abrogated the ferroptosis resistance conferred by HSD11B1 overexpression, while IL-10 overexpression partially rescued the pro-ferroptotic phenotype. Collectively, these results confirm that IL-10 acts downstream of JUND, and that HSD11B1 exerts its anti-ferroptotic effects via JUND-mediated IL-10 transcription. Finally, validation in both murine models and clinical samples demonstrated consistent overexpression of the HSD11B1/JUND/IL-10 axis in ectopic lesions, corroborating its translational relevance. To our knowledge, this is the first study to delineate a mechanistic link between HSD11B1 and ferroptosis in EMs via JUND-dependent IL-10 activation.

This study identifies HSD11B1 as a ferroptosis-related gene critically involved in the pathogenesis of EMs. Its elevated expression not only provides a promising non-invasive diagnostic biomarker but also reveals a novel molecular cascade HSD11B1/JUND/IL-10 that regulates ferroptosis and lesion progression. These findings offer a theoretical framework for the development of targeted therapies that restore ferroptotic sensitivity in EMs and highlight the utility of integrative bioinformatics and experimental strategies in unraveling complex disease mechanisms. Future investigations will focus on expanding clinical cohorts and performing in vivo functional validation to facilitate clinical translation of these insights.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3.