Introduction

Endometrial cancer (EC) remains the most prevalent gynecologic malignancy in developed countries, with incidence rates steadily rising due to an aging population and increasing obesity rates [1]. While early-stage disease is often curable, advanced or recurrent EC is associated with poor prognosis, largely driven by therapeutic resistance and metastatic progression [2]. Emerging evidence suggests that the clinical outcome is not solely determined by tumor cell-intrinsic factors but is profoundly influenced by the tumor microenvironment (TME). Among the stromal components, cancer-associated fibroblasts (CAFs) constitute the dominant cell population, orchestrating a desmoplastic stroma that physically and chemically supports tumor aggression [3].

CAFs are a heterogeneous population with diverse functions. They secrete extracellular matrix (ECM) proteins to facilitate invasion, release growth factors (e.g., TGF-β, HGF) to sustain tumor proliferation, and produce cytokines (e.g., IL-6, CXCLs) that modulate local immunity [4]. Single-cell RNA sequencing (scRNA-seq) studies have further stratified CAFs into distinct functional subtypes, primarily myofibroblastic CAFs (myCAFs), which reside near the tumor bed and contribute to matrix stiffening, and inflammatory CAFs (iCAFs), which are located distally and drive immunomodulation [5, 6]. In the context of EC, CAFs have been implicated in promoting epithelial-to-mesenchymal transition (EMT) through the secretion of cytokines such as IL-6 and the activation of JAK/STAT3 signaling, which ultimately fosters tumor progression and angiogenesis [7, 8]. Consequently, isolating and culturing primary CAFs has become the “gold standard” approach for dissecting these complex interactions in vitro.

However, a critical challenge in CAF research is the potential for phenotypic drift during in vitro expansion. The transition from a physiological 3D tissue environment to a rigid 2D plastic surface, coupled with exposure to supraphysiological nutrient and oxygen levels, imposes significant stress on primary cells and can alter their transcriptomic and functional profiles. This phenomenon has been observed in cultured CAFs, which exhibit convergence toward a common phenotype with loss of original heterogeneity by early passages of culture [9]. Currently, there is no consensus on the acceptable limit for passage numbers; while some protocols recommend using only very early passages [10,11,12], practical constraints often necessitate expanding CAFs for weeks—sometimes up to passage 10—to generate sufficient material for high-throughput assays [13, 14]. Despite this common practice, few studies have systematically rigorously benchmarked whether high-passage cells retain the functional and molecular fidelity of the original patient-derived fibroblasts. Using an unverified model could lead to experimental artifacts, particularly in studies involving immune interactions or drug screening.

In this study, we established a primary CAF culture system from human EC tissues and conducted a comprehensive longitudinal analysis from passage 1 (P1) to passage 10 (P10). We integrated morphological observation, functional assays (proliferation, migration, invasion), and bulk RNA sequencing to track the evolutionary trajectory of these cells. Our data reveal that while EC-CAFs maintain their core tumor-promoting functions and myCAF-like matrix remodeling capabilities across passages, they undergo significant transcriptomic reprogramming characterized by a progressive loss of antigen-presenting machinery and an acquisition of inflammatory traits. These findings provide essential guidelines for the appropriate use of primary CAF models in endometrial cancer research.

Materials and methods

Tissue collection and primary CAFs isolation

Fresh endometrial cancer tissues were obtained from four patients undergoing surgical resection at Peoples’ Hospital of Anshun City. None of the enrolled patients received preoperative chemotherapy, radiotherapy, or hormonal therapy. The clinical characteristics of these patients, including age, histological diagnosis, FIGO stage, tumor grade, and other relevant clinical background information, are summarized in Supplementary Table S1. All tissue samples were identified and collected by experienced pathologists to ensure accurate diagnosis. Primary cancer-associated fibroblasts (CAFs) were isolated using a standardized enzymatic digestion and differential adherence method. Briefly, tissues were washed with phosphate-buffered saline (PBS) containing antibiotics, minced into approximately 1 mm³ fragments, and digested with Type I Collagenase (Catalog # C8140, Solarbio, China) at 37 °C for 1 h with constant agitation. The digested suspension was filtered through a 75-µm cell strainer and centrifuged. The cell pellet was resuspended in DMEM/F12 medium (Catalog # PM150312, Procell, China) supplemented with 10% fetal bovine serum (Catalog # A5256701, ThermoFisher, USA) and antibiotics and seeded into culture dishes. To purify CAFs, the differential adherence method was applied: after incubation for 3 h at 37 °C, non-adherent cells (epithelial and immune cells) were removed by changing the medium, leaving the adherent fibroblasts. Cells were passaged when they reached 80–90% confluence. To ensure experimental consistency, CAFs at passage 1 (P1), passage 5 (P5), and passage 10 (P10) were cryopreserved in liquid nitrogen. For subsequent experiments, cells from different passages were thawed and cultured simultaneously.

Cell culture and conditioned medium (CM) preparation

Human endometrial cancer cell lines ISHIKAWA and HEC-1 A were obtained from Procell (Wuhan, China) and cultured in DMEM or McCoy’s 5 A medium, respectively, supplemented with 10% fetal bovine serum and antibiotics. All cells were maintained at 37 °C in a humidified incubator with 5% CO₂.

For conditioned medium (CM) preparation, P1, P5, and P10 CAFs were cultured to approximately 80% confluence, washed with PBS, and incubated in serum-free medium for 48 h. The supernatant was collected, centrifuged to remove debris, filtered through a 0.22-µm filter, and stored at − 80 °C until use. For functional assays, CM was mixed with fresh complete medium at a 1:1 ratio or used directly as indicated.

Immunofluorescence (IF) staining and flow cytometry

Immunofluorescence staining was performed as previously described with minor modifications [15]. Briefly, CAFs seeded on coverslips were fixed, permeabilized, blocked, and incubated with antibodies against FAP (1:500, Catalog # ET1704-23, Huabio) or Vimentin (1:50, Catalog # HA720165F, Huabio), followed by DAPI nuclear staining. Images were acquired using a Nikon AXR confocal laser scanning microscope (Nikon, Tokyo, Japan).

For flow cytometry, cells were harvested and stained for Vimentin to assess CAF purity. Apoptosis was evaluated using an Annexin V-FITC/PI Apoptosis Detection Kit (Catalog # 556547, BD Biosciences, USA) according to the manufacturer’s instructions. Data were acquired using a BD LSRFortessa™ cell analyzer and analyzed using FlowJo software.

Cell viability, proliferation, migration, and invasion assays

Cell viability and proliferation assays were performed as previously described with minor modifications [16]. For CCK-8 assays, CAFs or EC cells were seeded into 96-well plates, incubated with CCK-8 reagent (Catalog # AC11L554, life-ilab, China) at the indicated time points, and absorbance was measured at 450 nm. For EdU assays, cells were incubated with EdU working solution, fixed, stained according to the manufacturer’s protocol, and quantified under a fluorescence microscope.

Cell migration was evaluated using wound healing and Transwell assays. For wound healing assays, confluent cell monolayers were scratched with a sterile pipette tip and cultured in serum-free medium or CM. Images were captured at 0 h and 24 h, and wound closure was quantified using ImageJ. For Transwell assays, EC cells were seeded into the upper chamber in serum-free CAF-CM or control medium, while complete medium was added to the lower chamber as a chemoattractant. For invasion assays, membranes were pre-coated with Matrigel (Catalog # 356234, Corning, USA). After incubation, migrated or invaded cells were fixed, stained with crystal violet, and counted in five randomly selected fields.

Senescence-associated β-galactosidase (SA-β-gal) staining

Cellular senescence was evaluated using a Senescence β-Galactosidase Staining Kit (Catalog # C0602, Beyotime, China) according to the manufacturer’s instructions. SA-β-gal-positive areas were quantified using ImageJ from at least three randomly selected fields per group.

Enzyme-linked immunosorbent assay (ELISA)

The concentrations of TNF-α, TGF-β, and IL-6 in the cell culture supernatants were measured using commercial ELISA kits (Catalog # ED-11776; ED-11801; ED-10377, LunChangShuoBiotech) strictly according to the manufacturer’s protocols, as previously described with minor modifications [17]. The optical density (OD) was measured at 450 nm using a microplate reader.

RNA sequencing

Among the four enrolled patients, CAFs derived from three patients, namely Patient 1, Patient 3, and Patient 4, were selected for RNA-seq analysis. These three patients showed relatively comparable clinicopathological characteristics, including the same tumor grade and FIGO stage, thereby reducing potential heterogeneity caused by major clinical differences. For each patient-derived CAF line, cells were collected at P1, P5, and P10 for RNA-seq, resulting in three biological replicates per passage group and a total of nine RNA-seq samples. The patient samples used for RNA-seq are indicated in Supplementary Table S1. All RNA-seq samples were submitted for sequencing in the same batch and processed using the same library preparation protocol and sequencing platform to minimize potential technical variation.

RNA sequencing was performed as previously described with minor modifications [18]. Total RNA was extracted using Trizol Reagent, and RNA purity, concentration, and integrity were assessed using a NanoDrop spectrophotometer, Qubit 4.0 fluorometer, and Agilent 2100/4200 system, respectively. RNA-seq libraries were constructed using 3 µg of total RNA per sample following a standard poly(A)-enriched mRNA-seq workflow. Libraries were sequenced on the Illumina NovaSeq platform to generate 150-bp paired-end reads. Raw reads were filtered using fastp (v0.21.0), and clean reads were mapped to the reference genome using HISAT2 (v2.1.0).

Bioinformatics analysis

Potential batch effects and overall sample distribution were evaluated by principal component analysis (PCA) based on the normalized gene expression matrix. Differential expression analysis was performed using the DESeq2 R package (v1.30.1) [19]. The raw count matrix was rounded to integers, and low-abundance genes (row sums < 10) were filtered out. Differentially expressed genes (DEGs) were identified based on the criteria of |log2Fold Change| ≥ 1 and adjusted P value < 0.01. Functional enrichment analyses, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG), were conducted using the “clusterProfiler” R package [20]. To capture subtle but coordinated pathway changes, Gene Set Enrichment Analysis (GSEA) was performed on the ranked gene list (sorted by log2Fold Change) using the same package [20].

To comprehensively characterize the phenotypic evolution of CAFs, we employed multiple advanced algorithms:

(1) Immune and stromal signature analysis: The abundance of specific functional signatures, including iCAF (e.g., IL6, CXCL1), myCAF (e.g., ACTA2, COL1A1), and MHC-II complex scores, was quantified using Gene Set Variation Analysis (GSVA) [21] with the “ssgseaParam” method.

(2) Phenotype-specific marker validation: Normalized expression data for key genes related to proliferation, senescence, apoptosis, and invasion were extracted and visualized as box plots using ggplot2.

All statistical visualizations, including volcano plots, heatmaps, and Venn diagrams, were generated in the R statistical environment (version 4.3.0).

Cell treatment with H2O2, NAC, and TPCA-1

To further investigate the potential involvement of oxidative stress and NF-κB signaling in the phenotypic changes of CAFs during in vitro expansion, cells were treated with hydrogen peroxide (H2O2), N-acetyl-L-cysteine (NAC), and TPCA-1. H2O2 was used to induce oxidative stress, NAC was used as a reactive oxygen species (ROS) scavenger, and TPCA-1 was used as an inhibitor of NF-κB signaling.

Briefly, CAFs were seeded into culture plates and allowed to attach overnight. For oxidative stress induction, cells were treated with H2O2 (Fu Lin, China) at a final concentration of 200 µM for 24 h. For ROS inhibition, cells were pretreated with NAC (Catalog # HY-B0215, MedChemExpress, China) at a final concentration of 1 mM for 2 h before subsequent treatment or analysis. For NF-κB pathway inhibition, cells were pretreated with TPCA-1 (Catalog # HY-10074, MedChemExpress, China) at a final concentration of 10 µM for 1 h before subsequent treatment or analysis. TPCA-1 was dissolved in DMSO, and the same final concentration of DMSO was used as the vehicle control. After treatment, cells were collected for subsequent analyses, including ROS detection, quantitative real-time PCR, western blotting, immunofluorescence staining.

ROS detection

Intracellular ROS levels were detected using a DCFH-DA ROS assay kit (Catalog # CA1410, Solarbio, China) according to the manufacturer’s instructions. Briefly, CAFs were incubated with DCFH-DA working solution at 37 °C in the dark, washed with PBS, and immediately observed under a fluorescence microscope.

NF-κB pathway analysis

NF-κB pathway activation was evaluated by Western blotting and immunofluorescence staining. For Western blotting, total protein was extracted from CAFs and analyzed using antibodies against NF-κB p65 (Catalog # 10745-1-AP, Proteintech, China) and phospho-NF-κB p65 (Ser536) (Catalog # 80379-2-RR, Proteintech, China).

NF-κB p65 nuclear translocation was assessed by immunofluorescence staining using an NF-κB p65 nuclear translocation assay kit (Catalog # SN371, Beyotime, China) according to the manufacturer’s instructions. Briefly, adherent CAFs were fixed, washed, blocked, and incubated with an anti-NF-κB p65 mouse monoclonal antibody, followed by Cy3-conjugated anti-mouse secondary antibody incubation. Nuclei were counterstained with DAPI, and fluorescence images were acquired using a fluorescence microscope. NF-κB p65 nuclear translocation was evaluated based on the intracellular distribution of red p65 fluorescence and its overlap with DAPI-stained nuclei.

RT-qPCR validation of inflammatory genes

To validate the inflammatory genes identified by RNA-seq, the mRNA expression levels of IL6, CXCL1, CXCL2, and CCL2 were examined by RT-qPCR. Total RNA was extracted using RNAiso Plus (Catalog # 9108, Takara, Japan), and cDNA was synthesized using a reverse transcription kit (Catalog # RR047A, Takara, Japan). qPCR was performed using TB Green Premix Ex Taq (Catalog # RR820A, Takara, Japan). Relative gene expression was normalized to ACTB and calculated using the 2 − ΔΔCt method. The primer sequences, including those for ACTB, are listed in Supplementary Table S2.

Statistical analysis

All statistical analyses were performed using GraphPad Prism (Version 8.0.1) or R software (version 4.3.0). Data are presented as the mean ± standard deviation (SD) from at least three independent experiments. Sample sizes (n), representing the number of independent biological replicates or independent experiments, are indicated in the corresponding figure legends. Before applying parametric statistical tests, data normality was assessed using the Shapiro–Wilk test, and homogeneity of variance was evaluated using Levene’s test. For comparisons between two groups, a two-tailed Student’s t-test was used when the data met the assumptions of normality and equal variance. For comparisons among three or more groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was used for normally distributed data with homogeneous variance. For data involving multiple time points, such as CCK-8 assays, two-way ANOVA followed by Tukey’s or Sidak’s post hoc test was performed, as appropriate. A p-value < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Results

Isolation and identification of primary endometrial cancer-associated fibroblasts

To establish a reliable in vitro model, we isolated cancer-associated fibroblasts (CAFs) from endometrial cancer tissue samples using a standardized digestion and differential adherence method (Fig. 1A). The isolated cells exhibited a typical spindle-shaped, fibroblast-like morphology, which was maintained consistently across passages 1, 5, and 10 (P1, P5, P10) (Fig. 1B). The purity of the isolated CAFs was verified by flow cytometry, which showed that over 99% of the cells in P1, P5, and P10 populations were positive for the mesenchymal marker Vimentin (Fig. 1C). Furthermore, immunofluorescence staining confirmed the abundant expression of Vimentin (Fig. 1D) and the specific CAF marker, Fibroblast Activation Protein (FAP) (Fig. 1E), in virtually all cells. These data confirm the successful isolation of high-purity CAFs that retain their characteristic morphology and marker expression throughout early and late passages.

Fig. 1
Fig. 1
Full size image

Isolation, identification, and morphological characterization of primary endometrial cancer-associated fibroblasts (CAFs). (A) Schematic diagram illustrating the isolation workflow of CAFs from endometrial cancer tissues using mechanical mincing, enzymatic digestion, and a differential adherence method. (B) Representative phase-contrast microscopic images of CAFs at Passage 1 (P1), Passage 5 (P5), and Passage 10 (P10), showing typical spindle-shaped morphology. Scale bar: 50 μm. (C) Flow cytometric analysis of Vimentin expression to verify the purity of isolated CAFs at different passages. (D–E) Immunofluorescence staining of CAFs (P1, P5, P10) for the mesenchymal marker Vimentin (D, magenta) and the specific CAF marker Fibroblast Activation Protein (FAP) (E, red). Nuclei were counterstained with DAPI (blue). Scale bar: 50 μm

Stability of CAFs biological phenotypes across passages

We next investigated whether continuous culturing affects the fundamental biological properties of CAFs. Proliferation assays using EdU staining revealed no significant difference in the percentage of EdU-positive cells among P1, P5, and P10 CAFs (Fig. 2A, B). Similarly, a CCK-8 viability assay indicated comparable growth curves for all three passages over 72 h (Fig. 2C). The migratory capacity of CAFs, assessed by wound healing, also remained stable, with no statistically significant variations observed between passages (Fig. 2D, E). Furthermore, we evaluated apoptosis and senescence, finding no significant differences in the apoptotic rate (Annexin V/PI staining, Fig. 2F) or cellular senescence (SA-β-gal staining, Fig. 2G, H). Furthermore, we evaluated the Senescence-Associated Secretory Phenotype (SASP) of the CAFs. ELISA results indicated that the secretion levels of TNF-α and TGF-β were consistent across passages (Fig. 2I, J). We also attempted to measure IL-6, a common SASP factor, but its concentration was below the detection limit in all three groups (data not shown). Collectively, these results suggest that CAFs maintain a robust and stable functional phenotype in vitro up to passage 10.

Fig. 2
Fig. 2
Full size image

Characterization of biological phenotypes and stability of CAFs across passages. (A–B) Cell proliferation analysis using EdU incorporation assay. Representative fluorescence images (A) and quantification of the percentage of EdU-positive cells (B) in P1, P5, and P10 CAFs (n = 4 for each group). Red: EdU; Blue: Hoechst 33,342. Scale bar: 100 μm. (C) Cell viability curves of P1, P5, and P10 CAFs measured by CCK-8 assay at 0, 24, 48, and 72 h (n = 4 for each group). (D–E) Migration capability assessed by wound healing assay (n = 4 for each group). Representative images at 0 and 24 h (D) and quantification of the migration rate (E). (F) Apoptosis analysis using Annexin V-FITC/PI flow cytometry. (G–H) Evaluation of cellular senescence by SA-β-gal staining (n = 4 for each group). Representative images (G) and quantification of the percentage of SA-β-gal-positive area (H) in P1, P5, and P10 CAFs. Scale bar: 50 μm. (I–J) Assessment of the senescence-associated secretory phenotype (SASP) (n = 3 for each group). ELISA quantification of TNF-α (I) and TGF-β (J) levels in the supernatants of CAFs across passages. Data are presented as mean ± SD. ns: not significant

CAF-derived conditioned medium promotes endometrial cancer cell proliferation independently of passage number

To explore the paracrine effects of CAFs on tumor cells, we treated two endometrial cancer cell lines, ISHIKAWA and HEC-1 A, with conditioned medium (CM) derived from P1, P5, and P10 CAFs. CCK-8 assays demonstrated that CAF-CM significantly enhanced the viability of both cancer cell lines compared to the control group (P < 0.05), with no significant differences observed between the varying CAF passages (Fig. 3A, B). This pro-proliferative effect was corroborated by EdU assays, which showed a significant increase in DNA synthesis in cancer cells treated with CAF-CM (Fig. 3C–F). Notably, the magnitude of growth promotion was consistent across P1-, P5-, and P10-CM groups, indicating that the ability of CAFs to stimulate tumor cell proliferation is preserved during in vitro expansion.

Fig. 3
Fig. 3
Full size image

CAF-derived conditioned medium (CM) promotes endometrial cancer cell proliferation regardless of passage number. (A–B) CCK-8 viability assays of ISHIKAWA (A) and HEC-1 A (B) cells treated with control medium (Con) or CM derived from P1, P5, and P10 CAFs (n = 4 for each group). (C–F) EdU proliferation assays (n = 4 for each group). Representative fluorescence images of ISHIKAWA (C) and HEC-1 A (E) cells. Quantification of the EdU-positive rate for ISHIKAWA (D) and HEC-1 A (F). Scale bar: 100 μm. Data are presented as mean ± SD. ns: not significant; * P < 0.05, ** P < 0.01, *** P < 0.001 vs. Con

CAF-secreted factors enhance migration and invasion of endometrial cancer cells regardless of passage

We further assessed the impact of CAF-CM on the metastatic potential of endometrial cancer cells. Wound healing assays showed that treatment with P1-, P5-, or P10-CM significantly accelerated the wound closure rate in both ISHIKAWA and HEC-1 A cells compared to controls (Fig. 4A–D). Consistent with this, Transwell migration (Fig. 4E, F, I, J) and invasion assays (Fig. 4G, H, K, L) revealed a marked increase in the number of migrated and invaded cancer cells in the presence of CAF-CM. Importantly, no significant differences were detected in the pro-migratory or pro-invasive capabilities among the different CAF passage groups. These findings confirm that CAFs secrete soluble factors that potently drive tumor cell motility and invasion, and this secretory phenotype remains functionally stable across passages.

Fig. 4
Fig. 4Fig. 4
Full size image

CAF-CM enhances the migration and invasion of endometrial cancer cells. (A–D) Wound healing assays assessing the migratory ability of ISHIKAWA (A, B) and HEC-1 A (C, D) cells treated with different CAF-CMs (n = 4 for each group). (E–L) Transwell migration and invasion assays (n = 4 for each group). Representative images of migrated (E, I) and invaded (G, K) cells. Quantification of migrated cell numbers for ISHIKAWA (F) and HEC-1 A (J), and invaded cell numbers for ISHIKAWA (H) and HEC-1 A (L). Scale bar: 50 μm. Data are presented as mean ± SD. ns: not significant; * P < 0.05, ** P < 0.01, *** P < 0.001 vs. Con

Transcriptomic profiling reveals significant gene expression changes during CAF passaging

Although functional phenotypes appeared stable, we performed RNA sequencing to detect potential molecular alterations underlying CAF passaging. Volcano plots illustrated substantial changes in gene expression, with the highest number of differentially expressed genes (DEGs) observed between P10 and P1 (Fig. 5A–C). Interestingly, P5 and P10 CAFs shared a high degree of similarity, as indicated by the relatively low number of DEGs between them (Fig. 5H) and their clustering in Principal Component Analysis (PCA) (Fig. 5G). Heatmaps of the top 50 DEGs further highlighted distinct expression patterns, where P1 CAFs clustered separately from the P5 and P10 groups (Fig. 5D–F). Notably, the comparison between P10 and P5 yielded very few differentially expressed genes (DEGs), indicating that the transcriptomic profile stabilizes after the initial passages (Fig. 5H). In contrast, a substantial number of DEGs were identified when comparing either P5 or P10 to the primary P1 CAFs. Given that the major molecular shifts occurred early and persisted, we focused our subsequent analysis on the intersection of the ‘P5 vs. P1’ and ‘P10 vs. P1’ datasets. Venn diagram analysis identified a core signature of 1,300 common genes (Fig. 5I). This gene set, representing the stable transcriptional alterations associated with in vitro expansion, was utilized for the downstream functional enrichment analyses. These data suggest that while gross phenotype is maintained, CAFs undergo significant transcriptomic reprogramming early in culture (P1 to P5), stabilizing thereafter.

Fig. 5
Fig. 5
Full size image

Transcriptomic profiling reveals gene expression changes during CAF passaging. (A–C) Volcano plots displaying differentially expressed genes (DEGs) for the comparisons: P5 vs. P1 (A), P10 vs. P1 (B), and P10 vs. P5 (C). Red dots indicate upregulated genes, and blue dots indicate downregulated genes. (D–F) Heatmaps showing the expression patterns of the top 50 DEGs for each comparison group, ranked by adjusted P-value. (G) Principal Component Analysis (PCA) plot demonstrating the clustering of P1, P5, and P10 samples. (H) Bar chart summarizing the number of upregulated and downregulated DEGs (padj < 0.01, |Fold Change| > 2) for each comparison. (I) Venn diagram illustrating the overlap of DEGs between the “P10 vs. P1” and “P5 vs. P1” comparisons, identifying a core set of 1300 genes modified during expansion

Functional enrichment and subtype evolution analysis of CAFs in vitro

To understand the biological implications of the transcriptomic shifts, we performed functional enrichment analyses. Gene Ontology (GO) and KEGG pathway analyses of the common DEGs highlighted an enrichment in cell cycle-related processes, including “nuclear division,” “chromosome segregation,” and “DNA replication” (Fig. 6A, B), suggesting a modulation of proliferative machinery at the transcriptional level. To further capture the global trends in pathway regulation, GSEA was performed (Fig. 6C). Interestingly, key signaling pathways involved in fibroblast maintenance, such as “TGF-beta signaling” and “FoxO signaling,” did not show statistically significant enrichment (P.adjust > 0.05), implying that these intrinsic regulatory networks remain relatively stable during in vitro expansion. This finding aligns with our earlier observation of conserved biological phenotypes. In contrast, a distinct cluster of immune-related and infection-response pathways—including “Graft-versus-host disease,” “Type I diabetes mellitus,” “Phagosome,” and “Tuberculosis”—exhibited significant negative enrichment (NES < 0, P.adjust < 0.05). Although ostensibly related to specific diseases, these pathways share a common core of genes involved in antigen processing and presentation (e.g., MHC-II complex). Their collective downregulation strongly suggests a specific loss of immune-interacting potential in passaged CAFs.

Consistent with this, subtype analysis revealed a striking evolution in CAF signatures. The “inflammatory CAF” (iCAF) score increased significantly from P1 to P10 (Fig. 6E), mirroring the upregulation of inflammatory cytokines shown in the heatmap (Fig. 6D). In parallel, the “myCAF Matrix Score” remained relatively stable throughout the culture period (Fig. 6F). This stability aligns with the consistent expression of matrix markers (e.g., COL1A1, ACTA2) and corroborates our functional findings that CAF-mediated invasion support is preserved. Conversely, the “MHC-II Presentation Score” dropped sharply after P1 (Fig. 6G), driven by the downregulation of HLA genes (Fig. 6D). Taken together, these results indicate a molecular transition where CAFs lose their antigen-presenting phenotype and acquire a pro-inflammatory state, while maintaining their core myofibroblastic capabilities.

Fig. 6
Fig. 6
Full size image

Functional enrichment and subtype evolution analysis. (A–B) Gene Ontology (GO) enrichment (A) and KEGG pathway analysis (B) performed on the common DEGs shared by P5 and P10 CAFs. The size of the dots represents the gene count, and the color indicates the adjusted P-value. Key pathways highlighted in the text are marked in red. (C) Gene Set Enrichment Analysis (GSEA) summary showing Normalized Enrichment Scores (NES) for selected pathways in passaged CAFs relative to P1. Pathways with significant negative enrichment (NES < 0, FDR < 0.05) are highlighted in red (e.g., Graft-versus-host disease, Type I diabetes mellitus), while fibroblast-related pathways that did not reach statistical significance (FDR > 0.05) are highlighted in blue (e.g., TGF-beta and FoxO signaling). (D) Heatmap displaying the expression of selected marker genes associated with inflammatory CAFs (iCAFs), myofibroblastic CAFs (myCAFs), and MHC-II antigen presentation. (E–G) Single-sample GSEA (ssGSEA) scores comparing the iCAF Score (E), myCAF Matrix Score (F), and MHC-II Presentation Score (G) across P1, P5, and P10 groups

Validation of phenotype-associated gene expression changes

Finally, we validated key genes associated with CAF phenotypes. Consistent with the global transcriptomic data, proliferation markers (MKI67, PCNA, MCM2) were significantly downregulated in P5 and P10 compared to P1 (Fig. 7A), despite the functional stability observed in Fig. 2. Senescence-associated genes (CDKN2A, CDKN1A, GLB1) showed a progressive upregulation (Fig. 7B), indicating the onset of molecular senescence. Apoptosis-related genes showed mixed trends, with a significant decrease in the anti-apoptotic marker BCL2 (Fig. 7C). Additionally, markers related to invasion and matrix remodeling exhibited dynamic changes; notably, FN1 (Fibronectin) and VIM (Vimentin) were upregulated in later passages (Fig. 7D), which may compensate for other changes to maintain the functional support of cancer cell invasion. This validation confirms that CAFs undergo specific molecular adaptations, including a drift towards senescence and inflammation, while preserving their key tumor-promoting functions.

Fig. 7
Fig. 7
Full size image

Transcriptomic validation of phenotype-associated marker genes. (A–D) Analysis of the variance stabilizing transformed (VST) expression levels of key genes related to Proliferation (A), Senescence (B), Apoptosis (C), and Invasion (D) in P1, P5, and P10 CAFs based on the RNA-seq dataset. Box plots represent the normalized expression values. Data are presented as mean ± SD. ns: not significant; * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

ROS/NF-κB signaling contributes to passage-associated inflammatory drift during CAF expansion

To experimentally validate whether the passage-associated transcriptomic changes identified by RNA-seq reflected a biologically interpretable stress/inflammatory program, we examined intracellular ROS production, NF-κB activation, and representative inflammatory mediators during CAF expansion. ROS staining showed a progressive increase in intracellular ROS levels from P1 to P10 CAFs (Fig. 8A). Immunofluorescence staining further showed enhanced P65 nuclear localization in P5 and P10 CAFs compared with P1 CAFs, suggesting increased NF-κB activation during serial passage (Fig. 8B). Consistently, western blot analysis showed increased P65 phosphorylation in later passages (Fig. 8C), and quantification confirmed a significant increase in the p-P65/P65 ratio from P1 to P10 CAFs (Fig. 8D). RT-qPCR analysis further demonstrated that the RNA-seq-upregulated inflammatory mediators IL6, CXCL1, CXCL2, and CCL2 were progressively increased during in vitro expansion (Fig. 8E). These results indicate that late-passage CAFs undergo a passage-associated inflammatory transcriptional drift accompanied by ROS accumulation and NF-κB activation.

To determine whether this late-passage inflammatory phenotype could be attenuated by targeting ROS or NF-κB signaling, P10 CAFs were treated with NAC or the NF-κB inhibitor TPCA-1. ROS staining showed that NAC markedly reduced ROS accumulation in P10 CAFs, while TPCA-1 treatment also attenuated the overall stress-associated phenotype (Fig. 8F). Immunofluorescence staining showed that both NAC and TPCA-1 reduced P65 nuclear localization in P10 CAFs (Fig. 8G). Western blot analysis further showed decreased P65 phosphorylation after NAC or TPCA-1 treatment (Fig. 8H), and quantification confirmed a significant reduction in the p-P65/P65 ratio (Fig. 8I). In parallel, qRT-PCR analysis showed that NAC and TPCA-1 both downregulated IL6, CXCL1, CXCL2, and CCL2 expression in P10 CAFs (Fig. 8J). These findings suggest that ROS accumulation and NF-κB activation contribute to the inflammatory drift of late-passage CAFs.

We next examined whether oxidative stress was sufficient to induce a similar inflammatory phenotype in early-passage CAFs. P1 CAFs were treated with H₂O₂ to mimic oxidative stress, with or without TPCA-1. ROS staining showed that H₂O₂ increased intracellular ROS levels, whereas combined treatment with TPCA-1 reduced the H₂O₂-associated stress response (Fig. 8K). Immunofluorescence staining showed that H₂O₂ enhanced P65 nuclear localization, while TPCA-1 attenuated this effect (Fig. 8L). Western blot analysis showed increased P65 phosphorylation after H₂O₂ treatment, which was partially suppressed by TPCA-1 (Fig. 8M), and quantification confirmed corresponding changes in the p-P65/P65 ratio (Fig. 8N). RT-qPCR analysis further showed that H₂O₂ upregulated IL6, CXCL1, CXCL2, and CCL2 expression in P1 CAFs, whereas TPCA-1 significantly reduced the induction of these inflammatory mediators (Fig. 8O). Together, these gain- and loss-of-function validation experiments support a ROS/NF-κB-dependent mechanism underlying the inflammatory transcriptomic drift observed during CAF expansion. Thus, although core CAF-associated cellular functions were relatively maintained across passages, RNA-seq revealed a biologically meaningful stress-responsive inflammatory program in late-passage CAFs.

Fig. 8
Fig. 8Fig. 8
Full size image

ROS/NF-κB signaling contributes to passage-associated inflammatory drift during CAF expansion. (A) Representative ROS staining images of P1, P5, and P10 CAFs. (B) Representative immunofluorescence images showing P65 localization in P1, P5, and P10 CAFs. (C, D) Western blot analysis and quantification of p-P65, total P65, and p-P65/P65 ratio in P1, P5, and P10 CAFs (n = 3 for each group). (E) RT-qPCR analysis of IL6, CXCL1, CXCL2, and CCL2 expression in P1, P5, and P10 CAFs (n = 3 for each group). (F) Representative ROS staining images of P10 CAFs treated with NAC or TPCA-1. (G) Representative immunofluorescence images showing P65 localization in P10 CAFs treated with NAC or TPCA-1. (H, I) Western blot analysis and quantification of p-P65, total P65, and p-P65/P65 ratio in P10 CAFs treated with NAC or TPCA-1 (n = 3 for each group). (J) RT-qPCR analysis of IL6, CXCL1, CXCL2, and CCL2 expression in P10 CAFs treated with NAC or TPCA-1 (n = 3 for each group). (K) Representative ROS staining images of P1 CAFs treated with H₂O₂ or H₂O₂ plus TPCA-1. (L) Representative immunofluorescence images showing P65 localization in P1 CAFs treated with H₂O₂ or H₂O₂ plus TPCA-1. (M, N) Western blot analysis and quantification of p-P65, total P65, and p-P65/P65 ratio in P1 CAFs treated with H₂O₂ or H₂O₂ plus TPCA-1 (n = 3 for each group). (O) RT-qPCR analysis of IL6, CXCL1, CXCL2, and CCL2 expression in P1 CAFs treated with H₂O₂ or H₂O₂ plus TPCA-1 (n = 3 for each group). Data are presented as mean ± SD. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001

Discussion

Primary culture of cancer-associated fibroblasts (CAFs) serves as the cornerstone for elucidating tumor-stroma interactions; however, the reliability of these models over prolonged passage remains a critical yet often overlooked issue. While standard protocols allow for the isolation of CAFs from various solid tumors, few studies have rigorously benchmarked whether high-passage cells retain the functional and molecular fidelity of the original patient-derived fibroblasts [3]. Our study provides a comprehensive longitudinal characterization of endometrial cancer-derived CAFs (EC-CAFs), revealing a striking dichotomy: while the cells retain core tumor-supporting phenotypes and myofibroblastic phenotypes up to passage 10, they undergo profound transcriptomic reprogramming. Specifically, we identified a molecular trajectory characterized by the preservation of myCAF signatures coupled with a progressive loss of antigen-presenting capacity and an acquisition of inflammatory traits. These findings reconcile the “functional consistency” often observed in basic invasion assays with the “molecular drift” suspected in genomic studies.

An important consideration in this study is the rationale for not including P0 cultures and for selecting P1, P5, and P10 as representative passages. P0 cultures were not used for RNA-seq or functional assays because, in our experimental workflow, they represented the initial mixed primary cultures immediately after tissue digestion and seeding rather than a purified CAF population. During primary culture, we observed a small number of cobblestone-like non-fibroblast cells in P0 cultures, which were likely endothelial or epithelial cells. The presence of these contaminating cells could potentially interfere with CAF-specific transcriptomic profiling and functional assays. Therefore, CAFs were further purified through differential adherence and passaging. After the first passage, the cultures showed a more uniform spindle-shaped fibroblast-like morphology, and P1 was considered the earliest experimentally reliable CAF passage for downstream analyses. Accordingly, P1, P5, and P10 were selected to represent early, intermediate, and late stages of in vitro CAF expansion, respectively. This design enabled us to evaluate whether primary EC-CAFs maintain functional consistency during short-term culture and to characterize their molecular and phenotypic evolution during prolonged in vitro expansion, an issue that has been emphasized in CAF research because in vitro culture and passage can influence CAF phenotypes.

A key observation in our study is the functional stability of CAFs in promoting EC cell proliferation, migration, and invasion across passages. Mechanistically, this aligns with our transcriptomic data showing that the “myCAF Matrix Score” and the expression of canonical matrix genes (COL1A1, ACTA2, FN1) remained stable. Current consensus classifies CAFs into two primary functional states: myofibroblastic CAFs (myCAFs), driven by TGF-β signaling, and inflammatory CAFs (iCAFs), driven by IL-1/JAK signaling [22, 23]. Our GSEA results indicated that the TGF-β signaling pathway was not significantly altered during expansion. Since myCAFs are the primary effectors of matrix stiffening and physical remodeling that facilitate cancer cell invasion [24], the preservation of this signaling axis explains why high-passage CAFs continue to support metastasis-related phenotypes efficiently. Furthermore, it has been established that the rigid plastic substrate of 2D culture can constitutively activate mechanotransduction pathways, such as YAP/TAZ, which stabilizes the myofibroblastic phenotype [25]. This suggests that for research focusing on physical tumor-stroma interactions or invasion mechanisms, passaged EC-CAFs (up to P10) remain a biologically relevant and reliable model.

In contrast to the stability of the myCAF phenotype, we observed a dramatic downregulation of MHC-II-related genes (HLA-DRA, HLA-DRB1) immediately after passage 1. This phenomenon likely reflects the absence of extrinsic immune pressure in vitro. Single-cell analyses in pancreatic and breast cancers have identified a subpopulation of “antigen-presenting CAFs” (apCAFs) capable of modulating T-cell responses [5, 26]. These apCAFs express MHC-II molecules and can activate CD4+ T cells in an antigen-specific manner. However, the expression of MHC-II molecules on fibroblasts is typically inducible by interferon-gamma (IFN-γ) derived from lymphocytes [5, 27]. In our sterile culture system, devoid of immune cells and their corresponding cytokines, CAFs likely undergo a “use-it-or-lose-it” downregulation of the antigen presentation machinery. This rapid loss of immunogenicity has significant implications, as it suggests that cultured CAFs may no longer recapitulate the immune-interactive landscape of the original tumor [4, 28]. Consequently, our data strongly caution against using passaged CAFs for immunotherapy research or for investigating CAF-mediated T-cell exclusion.

The apparent discrepancy between preserved tumor-promoting functions and extensive transcriptional remodeling should be interpreted in the context of the specific functional programs affected during passaging. Although RNA-seq identified a substantial number of DEGs across passages, these molecular changes were not uniformly distributed across all CAF functions. Instead, the most prominent alterations were related to immune-associated and inflammatory transcriptional programs, including the loss of MHC-II-related genes and the acquisition of an iCAF-like inflammatory signature. In contrast, key myofibroblastic and matrix-remodeling features, including the myCAF matrix score and canonical matrix-related genes, were relatively preserved. This may explain why passaged EC-CAFs retained their ability to promote EC cell proliferation, migration, and invasion despite marked transcriptomic remodeling. Therefore, our findings do not suggest complete molecular stability across passages; rather, they indicate that selected core tumor-supporting functions can remain phenotypically stable while immune- and inflammation-related molecular programs undergo substantial culture-associated divergence. Although no significant differences were observed among P1, P5, and P10 CAFs in proliferation, migration, apoptosis, or senescence assays, these negative functional findings should not be interpreted as a lack of biological significance. Instead, they define an important boundary of CAF stability during in vitro expansion. The preserved phenotypes indicate that core tumor-supporting functions, particularly those related to proliferation and invasion-associated stromal activity, can be maintained up to P10 under standard culture conditions. However, the accompanying transcriptomic remodeling demonstrates that this apparent phenotypic stability does not reflect complete molecular stability. This distinction has practical implications: passaged EC-CAFs may remain suitable for studies focusing on physical tumor-stroma interactions and invasion-related assays, but they should be used cautiously for investigations of immune regulation, inflammatory signaling, or antigen presentation. Thus, the translational relevance of our findings lies in defining both the utility and the limitations of passaged patient-derived CAF models.

Concurrently, we noted a significant upregulation of the iCAF score and inflammatory cytokines (IL6, CXCL1, CXCL2) in later passages. The etiology of this inflammatory drift in vitro presents an intriguing contrast to microenvironmental cues in vivo. For instance, Schwörer et al. recently demonstrated that in the hypoxic tumor core, iCAFs are induced via HIF-1α stabilization [29]. In our study, however, cells were maintained under standard atmospheric oxygen (~ 21%), which represents a hyperoxic environment compared to the physiological hypoxia typical of the tumor stroma. It is well-established that such supraphysiological oxygen levels induce cumulative oxidative stress and DNA damage in cultured fibroblasts, a phenomenon famously termed ‘culture shock‘ [30]. This chronic stress can activate the NF-κB signaling pathway and promote a SASP-like inflammatory program characterized by the induction of IL6 and chemokine-associated transcripts [31].

Importantly, our experiments provide experimental support for this interpretation. We observed a progressive increase in intracellular ROS levels from P1 to P10 CAFs, accompanied by enhanced P65 phosphorylation and nuclear translocation. Moreover, the RNA-seq-upregulated inflammatory mediators IL6, CXCL1, CXCL2, and CCL2 were also progressively increased during in vitro expansion. Functionally, H₂O₂ treatment in P1 CAFs partially mimicked the late-passage inflammatory phenotype by increasing ROS accumulation, NF-κB activation, and inflammatory gene expression, whereas NF-κB inhibition by TPCA-1 attenuated these effects. Conversely, treatment of P10 CAFs with the ROS scavenger NAC or TPCA-1 reduced P65 activation and suppressed the expression of inflammatory mediators. These results suggest that ROS accumulation and NF-κB activation are not merely speculative explanations, but experimentally supported contributors to the passage-associated inflammatory transcriptional drift observed in late-passage CAFs.

Based on our validation experiments, we propose that the iCAF-like signature observed in P10 cells represents a convergent phenotype: while in vivo iCAFs may be driven by hypoxia, in vitro iCAFs are likely driven by oxidative culture stress, resulting in a similar inflammatory output via distinct upstream mechanisms. It has been documented that mechanical stress and DNA damage accumulation during in vitro proliferation can activate NF-κB and STAT3 pathways, driving fibroblasts towards a senescence-associated secretory phenotype (SASP) or an iCAF-like state [32, 33]. Although our SA-β-gal staining did not show overt cellular senescence, the transcriptomic upregulation of CDKN1A (p21) and CDKN2A (p16) suggests that the cells are entering a “pre-senescent” molecular state, characterized by transcriptional activation of inflammatory mediators. This observation is consistent with findings that prolonged culture of primary fibroblasts inevitably leads to a shift towards pro-inflammatory transcriptomes [34].

Notably, although RNA-seq and qPCR analyses showed that IL6 mRNA was upregulated in P5 and P10 CAFs compared with P1 CAFs, IL-6 protein levels in the unconcentrated conditioned medium were below the detection limit of the ELISA assay under our experimental conditions. This discrepancy may reflect the sensitivity threshold of the ELISA assay, the selected conditioned-medium collection time point, cytokine secretion kinetics, or post-transcriptional regulation affecting translational efficiency and protein stability [35]. Therefore, IL6 should be interpreted primarily as a transcriptional indicator of inflammatory activation rather than direct evidence of increased IL-6 protein secretion in the present study. This artifactual inflammation could potentially confound drug screening results, particularly for agents targeting cytokine signaling pathways or inflammation-driven progression in endometrial cancer [13].

Our study has certain limitations. First, the number of patient-derived CAF lines was relatively limited, which is a common challenge in studies using primary clinical specimens. Future studies with larger patient cohorts will be necessary to further validate the generalizability of these findings. Second, our transcriptomic analysis was based on bulk RNA-seq, which might mask rare subpopulations that persist or perish during passaging [36]. Future studies employing single-cell RNA-seq could further resolve these subpopulation dynamics. Third, this study was conducted in a standard 2D culture system; it remains to be seen whether 3D organoid co-culture models could better preserve the original CAF heterogeneity and immune features, as suggested by recent protocols [37, 38]. Finally, we did not perform in vivo co-injection experiments to verify whether CAFs at different passages exert differential effects on tumor growth in animal models, although our in vitro functional assays suggest that their tumor-supporting capacity remains consistent regardless of passage number.

In conclusion, our work demonstrates that primary EC-CAFs are resilient models that maintain their core myofibroblastic and tumor-promoting capabilities during in vitro expansion. However, they are not static entities; they evolve towards an inflammatory, immune-inert phenotype. Researchers must exercise “model awareness”: while P5-P10 CAFs are suitable for studying tumor invasion and matrix remodeling, rigorously purified early-passage CAFs or more complex co-culture systems are preferable for investigating immune modulation and stromal immunogenicity.