M2 macrophages undergoing ferroptosis regulate uterine smooth muscle contraction to facilitate labor onset Authors: Limin Song, Yue Wu, Kunpeng Liu, Wanqiu Xia, Xiaoying Wang, Ye Bu, Huike Yang, Yan Cai Published: 2025 ABSTRACT AbstractLabor onset is a complex process influenced by multiple factors. Although the mechanisms initiating labor remain incompletely understood, accumulating evidence suggests that maternal immune responses serve as critical drivers of this process. Ferroptosis, an iron-dependent form of programmed cell death, has emerged as a potential modulator of the immune microenvironment. But its role in labor onset remains largely unexplored. This study examined ferroptosis occurrence and how it affects cell functions during labor onset. We analyzed multi-cohort transcriptome datasets to identify immune cell infiltration using CIBERSORT and ssGSEA. Meanwhile, with via LASSO regression, we identified four core immune cells including M2 macrophages in labor onset. Differential analysis revealed 49 ferroptosis-related differential genes (FRDGs) in this process. Ferroptosis and iron ion homeostatic pathways were enriched by KEGG and GO analysis. Our ROC model using the key FRDGs for diagnosing labor onset performed excellently. We further identified a significant negative correlation between M2 macrophages and the FRDGs HMOX1 by Spearman. Finally, we performed in vitro and in vivo experiments to confirm the ferroptosis involvement in myometrium and to affect the specific functions of human uterine smooth cells (HUSMCs) and THP1. Several key FRDGs are significantly differentially expressed at both mRNA and protein levels in the labor myometrium. In addition, elevated Fe²⁺, decreased T-GSH, and increased MDA reflected higher levels of ferroptosis during labor. In vitro, we detected a shift towards a pro-inflammatory phenotype induced by ferroptosis in M2 macrophages using flow cytometry and PCR arrays, which enhanced cytokine secretion and promoted uterine smooth muscle contraction. Conversely, inducing ferroptosis in HUSMCs suppressed cytoskeletal remodeling and gap junction protein expression, impairing contraction. The study aims to provide a new theoretical basis for the mechanism of labor onset and a potential therapeutic target for abnormal timing of delivery.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-33523-4.Keywords: Ferroptosis, M2 macrophages, Uterine smooth muscle, Labor onset, Immune microenvironment, InflammationSubject terms: Cell biology, Diseases, Immunology Introduction Labor onset is a finely tuned and complex process, regulated by maternal-fetal interactions, uterine contractions, decidual activation, cervical ripening, and membrane rupture1–6. Uterine contractions play a pivotal role in labor by transitioning the uterus from a quiescent to a contractile state. Although the roles of hormonal signals, such as “functional” progesterone withdrawal7–9, prostaglandin elevation9,10, and mechanical stretch, have been studied with limited understanding11,12, increasing evidence indicates that maternal inflammatory and immune responses are involved in every stage of labor, including the uterus transition13–15. During late pregnancy, a large infiltration of immune cells occurs within the uterine microenvironment, facilitating tissue remodeling and activation of the uterine myometrium through the secretion of inflammatory cytokines, matrix-degrading enzymes, and other mediators16,17. Among these immune cells, macrophages are crucial regulators of both immune tolerance during pregnancy and inflammation during labor18,19. Notably, analyses of human and murine uterine tissues indicate that M2 macrophages predominate in late pregnancy but decrease during labor, supporting their role in maintaining uterine quiescence20,21. In recent years, ferroptosis, a non-apoptotic form of programmed cell death, has attracted significant attention. It is characterized by iron-dependent lipid peroxidation and the accumulation of reactive oxygen species (ROS) and has been recognized as an important link between iron metabolism, oxidative stress, and inflammation22,23. Ferroptosis has been implicated in placental dysfunction in pregnancy-related disorders, such as spontaneous abortion, preeclampsia, gestational diabetes, intrahepatic cholestasis of pregnancy, and spontaneous preterm birth24–27. Recent studies suggest that ferroptosis is closely associated with immune cell function, potentially influencing local tissue homeostasis by promoting macrophage polarization to the M1 phenotype and cytokine secretion28,29. However, the role of ferroptosis in uterine immune regulation during labor onset remains largely unexplored. Specifically, whether M2 macrophages undergo ferroptosis during labor initiation, and whether this process influences the uterine immune microenvironment and smooth muscle contraction, remains a critical research gap. Based on this, we hypothesize that ferroptotic M2 macrophages may induce a shift in the local uterine immune environment from anti-inflammatory to pro-inflammatory. This transition enhances uterine smooth muscle cell activation and contraction, thereby promoting labor onset. In this study, we integrated multiple transcriptomic datasets to conduct an in-depth analysis of immune cell infiltration and key immune cell subsets in the myometrium during labor onset. We also systematically identified ferroptosis-related differential genes (FRDGs) within these datasets. We successfully demonstrated a reduction in M2 macrophage infiltration and an elevated expression of the FRDG heme oxygenase 1 (HMOX1) during labor onset, with a significant negative correlation observed between the two. Additionally, in vitro functional assays were performed to investigate the interactions between ferroptosis-induced M2 macrophages and myometrium. Our findings offer a novel mechanistic insight into the immune regulatory role of ferroptosis during labor onset. Specifically, we reveal its critical influence on the immune microenvironment in delivery, as evidenced by the decreased infiltration of M2 macrophages and changes in the expression of associated genes. Materials and methods Bioinformatics analysis We obtained five transcriptomic datasets of myometrium for labor onset (GSE163773, GSE186763, GSE181348, GSE9159, and E-MTAB-3136) from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) and ArrayExpress (https://www.ebi.ac.uk/biostudies/arrayexpress) databases30,31. The sample data obtained from these datasets, including women who underwent term birth by elective cesarean section without labor and term birth by cesarean section following spontaneous labor onset, were used for in-depth analysis and validation. Data processing was conducted using R language (version 4.2.2). Furthermore, we used the sva R package for data merging and batch effect correction. Differential expression analysis was performed between term no labor (TNL) and term labor (TL) groups in the aforementioned datasets. The resulting p-values were adjusted for multiple testing using the false discovery rate (FDR) method. The details of the datasets used in this study are presented in Supplementary Table S1. Firstly, our workers performed immune infiltration analysis on the corrected data using the cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT) and single sample gene set enrichment analysis (ssGSEA)32,33. CIBERSORT was performed using the LM22 signature matrix, which defines 22 human immune cell subtypes, to estimate the relative proportions of immune cells in every sample. Meanwhile, we conducted principal component analysis (PCA) and differential analysis on the infiltration results. And we applied a Lasso regression model to identify characteristic immune infiltrating cells in the myometrium during labor onset. Next, we used the Wilcoxon test to identify differentially expressed genes (DEGs) between TNL and TL groups from the merged data. Selection criteria included p < 0.05 and |logFC| > 1 to identify DEGs. And we retrieved ferroptosis-related genes (FRGs), including Driver, Suppressor, Marker, and Unclassified genes, from the FerrDb V2 database (http://www.zhounan.org/ferrdb/current/)34. We took the intersection of DEGs and FRGs to produce FRDGs. Pathway enrichment analysis of DEGs and FRDGs was conducted using the Kyoto Encyclopedia of Genes and Genomes (KEGG) (https://www.kegg.jp/)35, 36. Gene Ontology (GO) analysis (http://www.geneontology.org/)37 was performed to cluster the FRDGs based on biological processes (BP), cellular components (CC), and molecular functions (MF). Protein-protein interaction (PPI) network analysis of the identified FRDGs was carried out using the String database (https://cn.string-db.org/)38, with a confidence score threshold of 0.7. The interactions between node genes and their relationships were statistically evaluated. We visualised the PPI network using Cytoscape software (version 3.8.2). The research group performed receiver operating characteristic (ROC)39 analysis of key genes and constructed a diagnostic model of myometrial labor onset. In GSE163773 & GSE186763 and GSE9159 & E-MTAB-3136, the corrplot R package was used to evaluate the correlation between core immune cells and key FRDGs. The researchers simultaneously performed Spearman analysis to assess their correlations. R > 0 indicates a positive correlation between the variables and R < 0 indicates a negative correlation between the variables. The absolute value of the correlation coefficient indicates the strength of the relationship: 0–0.3.3 for very weak or no correlation, 0.3–0.5 for weak correlation, 0.5–0.8 for moderate correlation, and 0.8–1.8 for strong correlation. The workflow for this part of the analysis is depicted in Fig. 1. The workflow for this part of the bioinformatics analysis. Sample collection We collected myometrial tissue from the lower segment of the uterus of women undergoing cesarean section from the Fourth Hospital of Harbin Medical University. All participants were primiparous women with singleton pregnancies at term (gestational weeks ≥ 37 weeks) without pregnancy-related complications such as preeclampsia or gestational diabetes, and none had received labor-inducing agents. Samples were categorized into two groups: TNL (term no labor, n = 18) and TL (term in labor, n = 21). Regular uterine contractions were defined as contractions occurring at intervals of less than 3 min with cervical dilation of at least 2 cm40. Tissue samples were rinsed with pre-cooled saline and snap-frozen at − 80 °C within 20 min. The study was approved by the Ethics Committee of the Fourth Affiliated Hospital of Harbin Medical University (Approval No. 2023-YXLLSC-14). All procedures were performed in accordance with relevant regulations. All patients provided written informed consent. Detailed clinical information for the samples is provided in Supplementary Table S2. Cell culture and treatment Immortalized human uterine smooth muscle cells (HUSMCs, Newgainbio, Jiangsu, China) were cultured in DMEM/F12 medium (Gibco, Thermo Fisher Scientific, USA) containing 10% FBS (Procell, Wuhan, China) and 1% penicillin-streptomycin (Procell, Wuhan, China). The cells were treated with 10 µM Ferrostatin-1 (Fer-1, a ferroptosis inhibitor, S7243, Selleck Chemicals, USA) and 1 µM (1 S,3R)-RSL3 (RSL3, a ferroptosis activator, S8155, Selleck Chemicals, USA) to establish a ferroptosis model41. THP-1 (Tohoku Hospital Pediatrics-1, Procell, Wuhan, China), a human monocytic leukemia cell line, was maintained in RPMI1640 medium (MeilunBio, Dalian, China) containing 10% FBS (Procell, Wuhan, China) and 1% penicillin-streptomycin (Procell, Wuhan, China). We inoculated THP-1 cells at a density of 1 × 10⁶ cells/mL into 6-well plates while adding phorbol 12-myristate 13-acetate (PMA, 100 ng/mL, P1585, Sigma-Aldrich, USA) for a 24 h incubation. The medium was then replaced with fresh PMA-containing medium, and IL-4 (20 ng/mL, PeproTech, USA) and IL-13 (20 ng/mL, PeproTech, USA) were added to continue culturing for 48 h to induce M2 macrophage polarization42,43. Subsequently, Fer-1 (10 µM) and RSL3 (0.5 µM) were added to establish the ferroptosis model in M2 macrophages. Real-time quantitative polymerase chain reaction (RT-qPCR) and PCR array Total RNA from tissues and cells was extracted using the Trizol method. cDNA was synthesized from total RNA using the ReverTra Ace qPCR RT Kit (Toyobo, Japan) according to the manufacturer’s instructions. QPCR was performed using the 2×SYBR Green qPCR MasterMix (Seven, Beijing, China) on the LightCycler96 (Roche Diagnostics, Basel, Switzerland), with the reaction conditions set according to the manufacturer’s protocol. β-actin mRNA was used as the internal control to normalize the total RNA content, and the relative expression of the target gene was calculated using the 2−△△CT method. We provided primer sequences in the supplementary material (Supplementary Table S3). After inducing M2 macrophages from THP-1 cells, we treated them with Fer-1 and RL3. The research group used PCR arrays (Wcgene Biotech, Shanghai, China) to detect human cytokine and chemokine gene expression profiles. P < 0.05 and |logFC| > 0 were considered significant differences. Western blot (WB) RIPA lysis buffer and protease inhibitors were mixed at a 100:1 ratio to lyse the cells on ice. The protein concentration was measured using NanoDrop One (Thermo Fisher Scientific, MA, USA). After loading the samples in PAGE gel (YaMei, Shanghai, China), electrophoresis was performed at 80 V for 40 min, followed by 120 V for 1 h. We applied wet transfer to transfer proteins from gels to PVDF membranes. The membranes were blocked with 10% non-fat milk for 1.5 h, and then washed three times with TBST. The membranes were incubated overnight with the primary antibody at 4 °C. The next day, the membranes incubated with the secondary antibody for 1 h. The protein bands were visualized using ECL chemiluminescence reagent (MeilunBio, Dalian, China). β-actin was selected as the internal control. The relative expression of the target protein was calculated as the ratio of the target protein grayscale value to the internal control protein grayscale value. Antibody details are provided in the supplementary material (Supplementary Table S4). Ferrous ion (Fe²⁺) The Fe²⁺ content in tissue samples was measured using Ferrous Iron Colorimetric Assay Kit (Elabscience, Wuhan, China). We weighed the tissue and added 0.1 g of tissue to 0.9 ml of extraction solution and homogenised thoroughly. The sample was centrifuged at 16,000 g for 10 min, and the supernatant was collected for further analysis. According to the manufacturer’s instructions, Chromogenic Solution was added to both the tissue supernatant and standard samples, mixing homogeneously and incubating at 37 °C for 10 min. The experimenter centrifuged samples at 12,000 g for 10 min. 300 µL of the supernatant was added to a 96-well plate. We recorded the absorbance at 593 nm using a microplate reader for statistical analysis. Total glutathione (T-GSH) The T-GSH content in tissue samples was determined using the Total Glutathione Assay Kit (Beyotime, Shanghai, China) according to the manufacturer’s instructions. We homogenised and vortexed the tissue fully in a protein removal solution. The cells were subjected to two freeze-thaw cycles in liquid nitrogen and water bath, followed by incubation at 4 °C for 5–10 min. The samples were centrifuged at 10,000 g for 10 min at 4 °C, and the supernatant was collected. Glutathione reductase and chromogenic substrates were added to reduce all GSSG to GSH. T-GSH content was quantified and statistically analysed by measuring absorbance at 412 nm using a microplate reader. Malondialdehyde (MDA) MDA levels, a byproduct of lipid oxidation, were measured using the MDA Assay Kit (Beyotime, Shanghai, China). Tissue or cell samples were homogenized or lysed in cell lysis buffer on ice, followed by centrifugation at 12,000 g for 10 min. The supernatant was collected, and protein concentration was determined by NanoDrop One (Thermo Fisher Scientific, MA, USA). We prepared the assay reagents and standards according to the instructions. Absorbance was measured at 532 nm using a microplate reader. We indicated MDA level as the concentration of MDA per unit of protein weight in the sample. Immunofluorescence staining (IF) Paraffin-embedded human myometrial tissue Sect. (4 μm) were deparaffinized in xylene and rehydrated through a graded ethanol series. Antigen retrieval was performed in citrate buffer (pH 6.0) using heat-induced epitope retrieval. After cooling to room temperature, the sections were permeabilized with 0.3% Triton X-100 for 10 min and blocked with 5% normal goat serum for 1 h. The slides were then incubated overnight at 4 °C with a primary antibody against CD206 (Cell Signaling Technology, CST, #24595, 1:500). After washing with PBS three times, the sections were incubated with a fluorophore-conjugated secondary antibody for 1 h at room temperature in the dark. Nuclei were counterstained with DAPI. Fluorescence images were acquired using a Pannoramic MIDI scanner (3DHISTECH, Budapest, Hungary) and viewed with Digital SlideViewer 2.5 (3DHISTECH, Budapest, Hungary). For each tissue section, three non-overlapping regions were selected for quantification of CD206 and DAPI fluorescence intensity using ImageJ software. 10 flow cytometry To detect polarisation markers, we carried out flow cytometry analysis of THP-1-derived M2 macrophages treated with Fer-1 and RSL3.The cells were collected and resuspended in PBS containing 1% BSA. Cells were first incubated at room temperature for 10 min with human Fc receptor blocking reagent (Purified Anti-Human CD16 Antibody, Elabscience, Wuhan, China). Subsequently, the cells were incubated with fluorescently labeled CD206 antibody (APC anti-human CD206 Antibody, Biolegend, San Diego, CA, USA) at 4 °C for 30 min in the dark. After staining, the cells were washed with FACS buffer (1% BSA in PBS) and analyzed using the NovoCyte flow cytometer (Agilent Technologies, Santa Clara, CA, USA). Our group utilised FlowJo software (version 10.9.0, FlowJo LLC, Ashland, USA) to process the data. Antibody details are provided in the supplementary material (Supplementary Table S4). 11 F-actin staining To visualise the cytoskeletal changes, we stained HUSMCs with phalloidin. Cells were first fixed with 4% paraformaldehyde for 20 min and then permeabilised with 0.5% Triton X-100 for 10 min. The cells were stained with Fluor 594-conjugated phalloidin (Absin, Shanghai, China) for 40 min. Finally, the cells were counterstained with 4’,6-diamidino-2-phenylindole for 2 min to stain the cell nuclei. We acquired images using a Leica DM4B fluorescence microscope (Leica Microsystems, Germany). 12 collagen gel contraction assay The contraction ability of HUSMCs was assessed using a collagen gel-based assay kit (CBA-201, Cell Biolabs, San Diego, CA). Cells were digested and resuspended at a concentration of 5 × 10⁶ cells/mL. The cell suspension was mixed with the collagen gel working solution at a 1:4 ratio on ice. A final volume of 500 µL containing 0.5 × 10⁶ cells was added to each well of a 24-well plate for collagen polymerization. After 1 h of incubation at 37 °C, 1 mL of culture medium was added, and the cells were incubated at 37 °C, 5% CO₂ for 2 days to generate contraction force. Fer-1 (10 µM) and RSL3 (1 µM) were added to the medium to induce ferroptosis inhibition and induction. Additionally, HUSMCs were co-cultured with induced M2 macrophages or ferroptotic M2 macrophages at a 1:1 ratio for the collagen contraction assay. After 2 days, the collagen matrix was detached from the well wall using a sterile needle. The gel area at 0 h, 1 h, 1.5 h, 2 h, and 24 h was measured using ImageJ software and analyzed. 13 statistical analysis Data were analyzed with GraphPad Prism 9.0 and R software (version 4.2.2). For comparisons between two groups: if the data met both normality and homogeneity of variance assumptions, a t-test was used. If the data met normality but failed the homogeneity of variance test, a Welch’s t-test was applied. For data that did not meet normality, the non-parametric Wilcoxon rank-sum test was used. For comparisons among multiple groups: if the data met both normality and homogeneity of variance, one-way ANOVA was performed. If the data met normality but failed the homogeneity of variance assumption, Welch’s one-way ANOVA was used. For non-normally distributed data, the Kruskal-Wallis test was applied. All experiments were conducted with at least three replicates. Statistical significance was defined as P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.001). Results Immune infiltration analysis and core immune cell identification during labor onset To assess immune cell infiltration during delivery, we initially performed immune infiltration analysis on the batch-corrected GSE163773 (TNL, n = 5; TL, n = 5) and GSE186763 (TNL, n = 6; TL, n = 11) datasets using the CIBERSORT algorithm. The results revealed that 21 out of 22 immune cell types infiltrated the myometrium (Fig. 2A, Supplementary Figure S1A-B). PCA clustering and differential analysis of immune cell infiltration showed significant differences in immune infiltration patterns between the TL group and the control TNL group (Fig. 2B). Differential analysis identified 12 immune cell types with significantly altered infiltration levels (p < 0.05), including a significant increase in monocytes and neutrophils in the TL group, and a significant decrease in M2 macrophages and activated NK cells (Fig. 2C, Supplementary Figure S1C). Furthermore, we applied Lasso regression for determining characteristic immune cells in the myometrium in labor onset, resulting in the identification of six key cell types: T cells follicular helper, activated NK cells, M2 macrophages, activated dendritic cells, resting mast cells, and eosinophils (Fig. 2D-E). The intersection of differential immune cells and characteristic immune cells identified four core immune cells potentially relevant to labor onset: activated NK cells, M2 macrophages, activated dendritic cells, and resting mast cells (Fig. 2F). Immune infiltration analysis and core immune cell identification during labor onset. (A) Heatmap of immune infiltration analysis in the batch corrected GSE163773 (term no labor, TNL, n = 5; term labor, TL, n = 5) and GSE186763 (TNL, n = 6; TL, n = 11) datasets using the CIBERSORT algorithm. (B) PCA clustering of the immune infiltration analysis in GSE163773 and GSE186763 datasets. (C) The histograms displaying the distribution of 21 types of immune cell in TNL and TL groups. Different colors represent different immune cells. (D)-(E) Lasso regression was used to select characteristic immune cells. (F) Venn diagram of differentially immune infiltration cells and characteristic immune cells from lasso regression. Additionally, immune infiltration analysis was performed using the ssGSEA method. The results showed significant differences (p < 0.05) in the levels of 20 out of 28 immune cell types between TL and TNL group. Specifically, TL group exhibited significantly increased levels of monocytes, neutrophils, macrophages, and mast cells, while the levels of effector memory CD8 + T cells and immature B cells were decreased (Supplementary Figure S1D-E). To further validate the immune infiltration results, we obtained datasets GSE9159 (TNL, n = 3; TL, n = 3) and E-MTAB-3136 (TNL, n = 24; TL, n = 18) from the GEO and ArrayExpress databases. The CIBERSORT algorithm identified 10 immune cell types with significantly different infiltration levels. The significant differences in activated NK cells, M2 macrophages, activated dendritic cells, and resting mast cells between TL and TNL groups were validated (Supplementary Figure S1F and Figure S2A). PCA clustering results also distinguished the majority of TL and TNL samples (Supplementary Figure S2B). Based on these findings, the team selected these four immune cell types as core immune cells for subsequent analysis. Ferroptosis-related genes and pathways enriched in the transcriptomic profile of myometrium during labor onset The working group merged data from GSE163773 and GSE186763 and eliminated batch effects (Fig. 3A, Supplementary Figure S1A-B). Differential expression analysis identified a total of 1416 differentially expressed genes. KEGG analysis based on an unbiased set of DEGs indicated that ferroptosis is one of the key biological processes during labor onset (Supplementary Figure S2C). On this basis, we intersected these with 516 ferroptosis-related genes FRGs from the FerrDb V2 database, resulting in the identification of 49 FRDGs, including 32 upregulated and 17 downregulated genes (Fig. 3B-C). Pathway enrichment analysis using the KEGG database revealed that these genes were primarily enriched in the ferroptosis pathway (Fig. 3D, Supplementary Figure S2D). Functional clustering of FRDGs based on the GO database revealed their involvement in oxidative stress, cellular metal ion transport and iron ion homeostasis (Fig. 3E). These results suggest that ferroptosis plays a critical role in the process of labor onset. Ferroptosis-related genes and pathways enriched in the transcriptomic profile of uterine myometrium during labor onset. (A) Box plots showing the expression profiles before and after normalization. (B) Venn diagram of differentially expressed genes (DEGs) from two GEO datasets (GSE163773 and GSE186763) and ferroptosis-related genes (FRGs) from the FerrDb V2 database. (C) Heatmap of 49 ferroptosis-related differential genes (FRDGs) intersected from DEGs and FRGs. (D) Pathway enrichment analysis using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database of 49 FRDGs. (E) Functional clustering of FRDGs based on the Gene Ontology (GO) database. Ferroptosis-related proteins constructed a PPI network and diagnostic model for labor onset We performed PPI network analysis on the 49 identified FRDGs utilising the String database, resulting in a protein network interaction map (Fig. 4A). This network includes 23 node genes and 34 relationship pairs, with the number of relationship pairs among node genes calculated (Fig. 4B). The differential expression of these 23 key FRGs was visualised with a volcano plot (Fig. 4C). The interaction network was visualized using Cytoscape software (Fig. 4D). Through this analysis, we discovered genes promoting delivery, such as interleukin 6 (IL6), interleukin 1 beta (IL1B), vascular endothelial growth factor A (VEGFA), and prostaglandin-endoperoxide synthase 2 (PTGS2), were located at the core nodes of the network10,44–46. Classic FRGs, including HMOX1, solute carrier family 40 member 1‌ (SLC40A1), transferrin receptor (TFRC), acyl-CoA synthetase long chain family member 4 (ACSL4), and solute carrier family 39 member 14 (SLC39A14), were also identified as key node genes. Ferroptosis-related proteins construct a protein–protein interaction network and diagnostic model for the onset of labor. (A) The protein-protein interaction (PPI) network analysis on the 49 FRDGs using the String database. (B) Calculating the number of relationship pairs between node genes of PPI network. (C) A volcano plot of the differential expression of 23 key FRGs. (D) The interaction network by using Cytoscape software. (E) 16 FRDGs were identified significantly differentially expressed in the validation dataset GSE181348 (TNL, n = 10; TL, n = 15). (F) Receiver operating characteristic (ROC) curve analysis for 12 key genes. (G) Constructing a diagnostic model for labor onset in the uterine myometrium. To validate these findings, we grabbed the dataset GSE181348 from the GEO database (TNL, n = 10; TL, n = 15). Differential expression analysis of the 23 previously identified FRDGs demonstrated that 16 genes remained significant difference (Fig. 4E). Subsequently, receiver operating characteristic (ROC) curve analysis was performed for 12 key genes, including iron metabolism-related genes (TFRC, HMOX1, SLC40A1, SLC39A14, HAMP), inflammation-related genes (IL6, IL1B, IL33), and FRGs (GPX4, ACSL4, PTGS2, HSPB1) (Fig. 4F). Following these analyses, we chose TFRC, HMOX1, SLC40A1, and glutathione peroxidase 4 (GPX4) to construct a diagnostic model for labor onset in the myometrium. This model exhibited excellent classification ability between labor and non-labor samples, with an area under the curve (AUC) of 0.987 and a 95% confidence interval (CI) of 0.940–1.000 (Fig. 4G). Correlation between core immune cells and key FRGs during labor onset In GSE163773 and GSE186763, the correlation between core immune cells and key FRDGs was assessed (Fig. 5A). Using Spearman method, the correlation between M2 macrophages and key FRDGs in the TL group was analyzed separately (Supplementary Figure S2E). The results revealed varying degrees of negative correlation between M2 macrophages and SLC39A14, TFRC, ACSL4, IL1B, IL6, and HMOX1, while a weak positive correlation was observed with SLC40A1 (Supplementary Figure S2E). Notably, when analyzing both TL and TNL groups together, only HMOX1 showed a significant moderate negative correlation with M2 macrophages (Fig. 5B). This finding was also verified in the validation datasets GSE9159 and E-MTAB-3136 (Supplementary Figure S3). In addition, we found that HMOX1 is significantly correlated with FRGs (IL1B, IL6, SLC39A14, ACSL4, TFRC, SLC40A1) in TL and TNL groups (Fig. 5C). These results suggest that HMOX1 may play an important role in labor onset by regulating the ferroptosis process in M2 macrophages. Correlation between core immune cells and key ferroptosis-related genes during labor onset. (A) The correlation between core immune cells and key FRDGs in GSE163773 and GSE186763 by corrplot R package. (B) HMOX1 showed a correlation with M2 macrophages in TL and TNL groups by Spearman. (C) The correlation between HMOX1 and key FRDGs in GSE163773 and GSE186763 by Spearman. Ferroptosis levels upregulation in human uterine smooth muscle tissue during labor onset Next, we examined the levels of ferroptosis in human uterine smooth muscle tissue collected clinically at labor onset. Upon statistical analysis of the clinical data, we observed no significant differences between the TNL and TL groups in maternal age, before pregnancy BMI, delivery BMI, gestational week, newborn weight, newborn length, iron metabolism-related parameters (hemoglobin), and the levels of monocytes, which are precursors to macrophages (Supplementary Table S2). Subsequently, we examined the changes in mRNA and protein levels of key FRGs by RT-qPCR and WB. We found that during labor onset, mRNA expression of HMOX1, TFRC, ACSL4, SLC39A14, and NFE2-like bZIP transcription factor 2 (NRF2) was elevated, while SLC40A1 and GPX4 expression were decreased (TNL, n = 15; TL, n = 15) (Fig. 6A). NRF2, a key transcription factor inducing HMOX1 expression in monocytes/macrophages, was also detected47. Further protein expression analysis revealed consistent results, with increased expression of HMOX1, TFRC, and ACSL4, and decreased expression of SLC40A1 and GPX4 (Fig. 6B, Supplementary Figure S4A-E). Given that the hallmark changes of ferroptosis involve lipid peroxidation and disrupted iron metabolism, we measured levels of Fe²⁺, T-GSH, and MDA (TNL, n = 15; TL, n = 15). We found that, compared to TNL group, TL group had increased Fe²⁺ levels, decreased T-GSH levels, and elevated MDA content (Fig. 6C-E). These findings further confirm the upregulation of ferroptosis during labor onset. Ferroptosis levels upregulation in human uterine smooth muscle tissue during labor onset and regulate phenotypic changes in HUSMCs. (A) The mRNA levels of key ferroptosis genes by RT-qPCR (HMOX1, TFRC, ACSL4, SLC39A14, NRF2, SLC40A1 and GPX4) in human myometrium of TNL and TL groups (TNL, n = 15; TL, n = 15). (B) The protein levels of key ferroptosis genes by WB (HMOX1, TFRC, ACSL4, SLC40A1 and GPX4) in human myometrium of TNL and TL groups. Original blots are presented in Supplementary Figure S5-S9. (C)-(E) The levels of Fe²⁺, T-GSH, and MDA in human myometrium of TNL and TL groups (TNL, n = 15; TL, n = 15). (F) The expression of the key gap junction protein CX43 in HUSMCs treated with DMSO, Ferrostatin-1 (Fer-1) and RSL3 (n = 3 in every group). Original blots are presented in Supplementary Figure S10. (G)-(H) Phalloidin staining and statistics of F-actin (red) in HUSMCs treated with DMSO, Fer-1 and RSL3 (n = 3 in every group). Scale bar: 50 μm. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.001. Ferroptosis regulates phenotypic changes in HUSMCs In vitro experiments, we further investigated the impact of ferroptosis on HUSMCs phenotypes. As HUSMCs activation precedes contraction, we first examined the effects of ferroptosis on functional gap junctions and cytoskeletal remodeling48. The results showed that, compared to Fer-1 treatment, RSL3 significantly reduced the expression of the key gap junction protein connexin 43 (CX43) during labor (Fig. 6F, Supplementary Figure S4F). Subsequently, changes in actin filaments, stained with Phalloidin, were used as indicators of cytoskeletal remodeling. Fer-1 treatment induced the transformation of HUSMCs into a triangular shape, representing the myocyte morphology during labor (Fig. 6G). In contrast, RSL3 treatment caused actin filaments to remain parallel along the length of the cells, even aggregating, which inhibited cytoskeletal remodeling and resulted in a compact morphology of HUSMCs (Fig. 6G-H). These findings suggest that the inhibition of ferroptosis facilitates HUSMCs activation into the contractile state, while ferroptosis occurrence inhibits this process. Ferroptotic M2 macrophages enhance the contractility of HUSMCs Based on our preliminary data analysis, FRGs showed a significant negative correlation with M2 macrophages during labor onset. Initially, we performed IF of human myometrial tissue sections for the M2 macrophage marker CD206. The results showed that CD206 was significantly decreased in the TL group compared with the TNL group (Fig. 7A, Supplementary Figure S4G). These findings led us to hypothesize that ferroptosis occurring in M2 macrophages may release pro-inflammatory mediators that promote HUSMCs contraction. To test this hypothesis, THP-1 cells were differentiated into M2 macrophages and treated with either Fer-1 or RSL3. We examined the M2 macrophage marker CD206 by flow cytometry. The results showed that Fer-1 promoted high expression of CD206 in M2 cells, while RSL3 induced low expression of CD206 in M2 cells (Fig. 7B-C). It indicates that ferroptosis inhibits M2 macrophage polarization. Furthermore, the human cytokine and chemokine PCR array results showed that in M2 macrophages treated with RSL3, the expression of proinflammatory genes such as IL1B, IL16, C-X-C motif chemokine ligand 1 (CXCL1), and C-X-C motif chemokine ligand 2 (CXCL2) was upregulated. However, treatment with Fer-1 suppressed the expression of immune cell recruitment-related genes (CCL2, CCL3)49,50, thereby reducing immune cell aggregation (Fig. 7D-F, Supplementary Figure S4H). Ferroptotic M2 macrophages enhance the contractility of HUSMCs. (A) Immunofluorescence staining of the M2 macrophage marker CD206 in human myometrial tissue sections from the TNL and TL group (n = 3 in every group). Scale bar: 100 μm. (B) M2 macrophage marker CD206 was detected by flow cytometry in M2 macrophages induced from THP-1 cells and treated with either Fer-1 or RSL3 (n = 3 in every group). (C) Statistical analysis of CD206 flow cytometry data. (D)-(F) The human cytokine and chemokine PCR array results of M2 macrophages treated with Fer-1 and RSL3 (DMSO and Fer-1 groups, n = 3; RSL3 group, n = 4). (G) A collagen gel contraction assay in HUSMCs treated with DMSO, Fer-1, RSL3, M2 macrophages and ferroptotic M2 macrophages. (H) Statistical analysis of the collagen gel contraction assay (n = 4 in every group). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.001. We next evaluated the direct effects of Fer-1, RSL3, and ferroptotic M2 macrophages on HUSMC contractility by the collagen gel contraction assay (Fig. 7G–H). The results showed that, at 1, 1.5, and 2 h after collagen matrix separation from the well wall, the area of collagen gel in HUSMCs treated with Fer-1 was significantly reduced compared to HUSMCs under ferroptosis conditions. However, we noticed that the differences observed after 24 h were not significant, suggesting that inhibition of ferroptosis only accelerated the activation and contraction process of HUSMCs. Importantly, co-culture with ferroptotic M2 macrophages significantly promoted HUSMC contraction compared with direct RSL3 treatment of HUSMCs alone (Fig. 7G–H). Taken together, these results suggest that during labor onset, ferroptotic M2 macrophages are more likely to promote HUSMC contraction indirectly by releasing proinflammatory mediators, rather than ferroptosis occurring within HUSMCs themselves playing the predominant role. This mechanism may represent an important regulatory pathway underlying labor onset. Discussion Currently, the mechanisms underlying labor onset remain incompletely understood. Several single-cell sequencing studies of pregnancy tissues have indicated that inflammatory immunity plays a crucial role in the cascade of labor events5,40,51.Macrophages, as one of the key immune cells, primarily exert their effects through secreted products, including matrix metalloproteinases, IL1B, IL6, ‌tumor necrosis factor-α‌ (TNF-α), and nitric oxide, which upregulate uterine activation proteins and influence the uterus during labor52–54. However, the precise molecular mechanisms involved remain to be explored. This study, based on multi-center transcriptomic data, combined with tissue samples and in vitro functional experiments, innovatively elucidates the potential role of ferroptosis in M2 macrophages during labor onset. The study found significant enrichment of FRGs (HMOX1, TFRC, SLC40A1, SLC39A14, ACSL4) in the myometrium during labor onset. Among them, HMOX1 showed a negative correlation with M2 macrophages both before and after labor onset, leading us to hypothesize that HMOX1 is a key gene in the ferroptosis of M2 macrophages. Previous studies have shown that in macrophages, NRF2 acted as a transcription factor, and when cells undergo oxidative stress, NRF2 translocates from the cytoplasm to the nucleus, binding to stress response elements on HMOX1 to activate its transcription47. Our study confirms the high expression of NRF2. The elevated HMOX1 accelerates heme degradation, generating a large amount of Fe2+ and triggering the Fenton reaction. We speculate that this is one of the mechanisms for ferroptosis in M2 macrophages (Figure 8). A recent study has indicated that in proliferative diabetic retinopathy, HMOX1 expression and M2 macrophage infiltration are associated with ferroptosis55, which corroborates our findings in terms of molecular mechanisms. Ferroptosis-driven M2 macrophages secrete proinflammatory cytokines to modulate uterine smooth muscle contraction and facilitate labor onset. The figure was generated by Figdraw. Given that the validated FRGs are involved in multiple biological processes, changes in their expression alone cannot provide definitive evidence of ferroptosis. Therefore, we further examined elevated levels of Fe2+, glutathione depletion, and increased lipid peroxidation, which further indicate the activation of ferroptosis in the myometrium. These findings are consistent with previous studies that highlight the close association between oxidative stress, iron metabolism imbalance, and labor onset, suggesting that ferroptosis may serve as a crucial link between disturbances in iron metabolism and local inflammation56–59. Yun Zhou’s team demonstrated that iron overload induces macrophage polarization toward a pro-inflammatory phenotype through the ROS/acetyl-p53 pathway, which indirectly supports our findings60. However, the study by Jun-Kai Kao and colleagues reported that iron overload induces the polarization of THP-1 monocyte-derived macrophages toward the M2 phenotype61. We do not consider this to be in conflict with our findings, as their research primarily focuses on the effects of chronic iron overload. This study is the first to reveal that M2 macrophages may be the key regulatory cell population for ferroptosis during labor. The number of M2 macrophages decreases because of ferroptosis. Roger Pique-Regi’s working group explored the single-cell atlas of the myometrium during labor and found that the proportion of M2 macrophages, which are involved in wound healing, significantly decreases during labor13. Our immune infiltration analysis and flow cytometry results consistently support this dynamic change. We observed that under ferroptosis induction, M2 macrophages exhibit a decrease in CD206 expression. Proinflammatory cytokines, such as IL1B and CXCL1, were significantly upregulated. These findings suggest a shift of M2 macrophages toward an M1-like phenotype. This shift enhances local inflammatory responses, promoting myometrial activation and contraction. It is consistent with the theory proposed by Xiaopei Hao that ferroptotic M2 macrophages drive their conversion to an M1-like phenotype29. Moreover, Fer-1 treatment inhibited the expression of immune cell recruitment-related genes (CCL2, CCL3) in M2 macrophages. This suppression facilitates tissues in reducing the aggregation of inflammatory immune cells, thereby attenuating the inflammatory effects and promoting tissue healing. These findings support the theory that M2 macrophages play a role in maintaining uterine quiescence20. Interestingly, we found that inducing ferroptosis in HUSMCs alone suppressed cytoskeletal remodeling and the expression of gap junction protein (CX43). In contrast, ferroptotic M2 macrophages significantly enhanced HUSMCs contraction through the secretion effect. This suggests that M2 macrophages may regulate labor onset by remodeling the inflammatory microenvironment via ferroptosis, rather than through ferroptosis in smooth muscle cells themselves. To date, there have been no studies specifically investigating the impact of ferroptosis induction on HUSMCs phenotype changes. However, a study has proposed that RSL3 promotes the transition of vascular smooth muscle cells from a contractile to a synthetic phenotype, with reduced expression of contraction-related proteins, which provides some support for our findings62. These discoveries offer new insights into how ferroptosis, oxidative stress, and local immune dysregulation collaboratively influence labor onset. Indeed, this study has several limitations. Initially, further validation using animal models is essential to explore the in vivo dynamics of M2 macrophage ferroptosis. In addition, we acknowledge that CIBERSORT, which relies on the LM22 signature matrix derived from peripheral blood and provides only relative cell proportions, may not fully reflect tissue-resident immune cells in the myometrium, thereby imposing certain limitations on the interpretation of the immune infiltration results. Furthermore, the feasibility of targeting ferroptosis as a potential intervention for abnormal labor timing, such as preterm birth, requires additional validation through preclinical and clinical investigations. In conclusion, this study suggests that ferroptotic M2 macrophages may regulate the local immune microenvironment, promoting uterine smooth muscle contraction and thereby initiating labor. This finding not only enriches our understanding of labor mechanisms but also provides potential new therapeutic targets for regulating labor timing and preventing preterm birth. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors wish to acknowledge the researchers whose work has been referenced in this study. We also gratefully acknowledge the support from Xiantao Academic (https://www.xiantaozi.com/) and Figdraw (https://www.figdraw.com/) in the preparation of figures. Abbreviations Ferroptosis-related differential genes Differentially expressed genes Ferroptosis-related genes uterine smooth muscle cells Term non-laboring Term laboring Cell-type identification by estimating relative subsets of RNA transcripts Single sample gene set enrichment analysis Malondialdehyde Total glutathione Reactive oxygen species Heme oxygenase 1 Solute carrier family 40 member 1 Transferrin receptor Acyl-CoA synthetase long chain family member 4 Solute carrier family 39 member 14 NFE2-like bZIP transcription factor 2 Glutathione peroxidase 4 Connexin 43 Vascular endothelial growth factor A Author contributions Limin Song: Conceptualization, data curation, formal analysis, investigation, project administration, writing – original draft. Yue Wu: Data curation, formal analysis, project administration, writing – review and editing. Kunpeng Liu: Data curation, project administration, methodology. Wanqiu Xia: Data curation, project administration, validation. Xiaoying Wang: Investigation, data curation. Ye Bu: Investigation, supervision. Yan Cai and Huike Yang: Conceptualization, formal analysis, funding acquisition, resources, supervision. All authors contributed to manuscript revision, read, and approved the submitted version. Funding information This research was funded by Graduate Practice Innovation Project of Harbin Medical University (YJSCX2023-287HYD). Data availability The datasets analysed during the current study are available in the GEO repository (https://www.ncbi.nlm.nih.gov/geo/, GSE163773, GSE186763, GSE181348, GSE9159), ArrayExpress (https://www.ebi.ac.uk/biostudies/arrayexpress, E-MTAB-3136). The FerrDb V2 database (http://www.zhounan.org/ferrdb/current/), KEGG (https://www.kegg.jp/), GO (http://www.geneontology.org/), and the String database (https://cn.string-db.org/) are utilized in this study. We stored the raw data from the PCR array on the Figshare platform for open access (https://doi.org/10.6084/m9.figshare.29635565.v1). The data that support the findings of this study are also available from the corresponding author upon reasonable request. Declarations Competing interests The authors declare no competing interests. Ethics statement The studies involving human participants were reviewed and approved by the Ethics Committee of the Fourth Affiliated Hospital of Harbin Medical University (Approval No. 2023-YXLLSC-14). All procedures were performed in accordance with relevant regulations. All patients provided written informed consent. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Huike Yang, Email: huikeyang@hrbmu.edu.cn. Yan Cai, Email: caiyan318@126.com. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets analysed during the current study are available in the GEO repository (https://www.ncbi.nlm.nih.gov/geo/, GSE163773, GSE186763, GSE181348, GSE9159), ArrayExpress (https://www.ebi.ac.uk/biostudies/arrayexpress, E-MTAB-3136). The FerrDb V2 database (http://www.zhounan.org/ferrdb/current/), KEGG (https://www.kegg.jp/), GO (http://www.geneontology.org/), and the String database (https://cn.string-db.org/) are utilized in this study. We stored the raw data from the PCR array on the Figshare platform for open access (https://doi.org/10.6084/m9.figshare.29635565.v1). The data that support the findings of this study are also available from the corresponding author upon reasonable request.