Collagen-mediated pro-tumorigenic MAPK activation drives stromal-immune reprogramming in solid cancers.

Introduction
Collagen is a predominant protein in the human body and the major constituent of connective tissue (1). It is a long fibrous protein composed of three intertwined protein chains in a helical structure, providing collagen with its unique strength and stability. Collagen serves various crucial functions in the body, including structural support, maintaining skin elasticity and firmness, aiding in wound healing, and supporting joint and cartilage health (2, 3). There are multiple types of collagen, such as type I, II, and III collagens (4). These different types of collagen have slightly varying distributions and functions in the human body. The role of collagen in tumors is a complex and diverse topic, dependent on the cancer type, stage, and microenvironment of the tumor. Increasing evidence suggests that collagen may act as significant roles in the growth, metastasis, and invasion of tumors, and the collagen surrounding tumors can impact the diffusion of drugs within tumor tissue, thus affecting treatment outcomes (5–8). While collagen is widely recognized as a pro-carcinogenic factor, it may also have inhibitory effects on tumor growth in certain circumstances.
The impact of collagen on the tumor microenvironment (TME) is a crucial factor in influencing tumor progression (9, 10). Collagen can influence the immune response of tumors by influencing the activity of immune cells in the TME. The size and density of collagen fibers directly affect the external resistance encountered by immune cells during migration, thereby influencing their speed and quantity entering the tissue. Hartmann et al. found that despite the upregulation of T cell chemokines CXCL10 or CXCL4 in pancreatic cancer, T lymphocyte infiltration in the tumor is still inhibited due to the presence of high-density collagen (11). Collagen receptors play an essential role in collagen-mediated immune effects. The LAIR protein family is an important collagen receptor. Several reports indicate that collagen in the extracellular matrix can directly inhibit the cytotoxic activity of T cells through the LAIR1 receptor (7, 12). Matrix stiffness has been shown to impact T cell migration, with T cell activation significantly inhibited when cultured on matrices with increased stiffness, including the inhibition of T cell proliferation and the expression of activity-related cytokines (13). Furthermore, increased stiffness in the stroma can inhibit the cGAS immune signaling pathway in tumor cells, promoting tumor immune evasion and indirectly leading to fewer infiltrating cytotoxic T cells (14). The evidence above suggests that collagen significantly impacts T cells, but the correlation and potential effects of collagen on different types of immune cells remain unclear.
In the current study, we aimed to characterize the clinical and molecular implications of intratumoral collagen deposition in gastric cancer and other cancer types. Using a combination of in silico analysis, in-house clinical cohorts, and functional assays, we sought to evaluate the correlation between collagen levels and different TME cell populations, identify the signaling pathways regulated by collagen, and elucidate the mechanisms by which collagen promotes tumor progression. Our findings provide new insights into the role of collagen in shaping the TME and highlight potential therapeutic targets for human cancers.

Methods

Collection of transcriptome data
Transcriptome data and clinical annotations for all tumor types were retrieved from The Cancer Genome Atlas (TCGA) via the University of California Santa Cruz (UCSC) Xena platform (https://xenabrowser.net/datapages/). Samples with documented overall survival (OS) information were specifically curated for subsequent analysis. Furthermore, a collection of publicly available immunotherapy datasets comprising transcriptome data from cancer patients receiving immunotherapy, including GSE173839 (15), GSE194040 (16), PRJEB25780 (17), GSE135222 (18), GSE78220 (19), and GSE176307 (20), was incorporated into the study. The MEDI4736 dataset was sourced from Dr. Lajos Pusztai’s study (21). By combining the above triple-negative breast cancer (TNBC) datasets (GSE173839, GSE194040, and MEDI4736), the “removeBatchEffect” function in the “limma” package (22) was used to remove batch effects.

Bulk transcriptome data analysis
The collagen scores of each patient were calculated based on the collagen genes obtained from our previous study (8, 10) via the single-sample gene set enrichment analysis (ssGSEA) algorithm in the GSVA package. The full list of genes used for collagen score calculation was documented in Supplementary Table 1. The TME characteristics encompassed immunomodulators, levels of tumor-infiltrating immune cells and other stromal cells, and the presence of inhibitory immune checkpoints. The complete methodological framework for TME characterization has been previously described in our published work (23–26). Then, The TCGA and the PRJEB25780 cohorts were utilized to explore the correlations between Collagen scores and these TME characteristics.

Single-cell RNA-sequencing data analysis
The single-cell RNA sequencing (scRNA-seq) datasets of 26 patients with primary gastric carcinoma were obtained from the GSE183904 dataset (27). We implemented rigorous quality control by excluding cells demonstrating either: (1) mitochondrial gene content >10%, (2) <200 detected genes, or (3) >5,000 detected genes (potential doublets). The “RunHarmony” function in the R package harmony (28) was used to mitigate the technical batch effects among individuals and experiments. Dimensionality reduction employed a dual-phase approach: first identifying 4,000 highly variable genes for principal component analysis (29), then selecting the top 30 principal components for t-SNE visualization (30). Cell clustering utilized the shared nearest neighbor clustering (SNN) graph-based algorithm (31) with modularity optimization at resolution=1. This analytical pipeline systematically resolved 96,162 high-quality cells into 35 transcriptionally distinct clusters, enabling comprehensive characterization of the gastric carcinoma ecosystem.
First, for each signature gene, we defined a background set comprising the 100 genes with the most comparable average expression levels. The expression value of each signature gene was then normalized by subtracting the mean expression of its background set. The migration and invasion scores were subsequently calculated by averaging these normalized values across all signature genes. Additionally, using the “AddModuleScore” function in Seurat, we quantified the activity of the MAPK signaling pathway, as well as the migration and extravasation potential of tumor cells, based on previously established gene signatures (32). In addition, the angiogenesis score in endothelial cells was also determined.

Clinical samples
A total of 60 patients with gastric cancer were recruited from The Affiliated Wuxi People’s Hospital of Nanjing Medical University, following ethical approval (No. KY23176). Tumor tissue samples were obtained at surgery. All patients received standard post-surgical adjuvant therapy. Additionally, paraffin-embedded tissue microarrays containing 60 lung cancer samples were procured from the National Engineering Center for Biochip (Outdo Biotech, Shanghai, China) under approval No. SHYJS-CP-1601005.

Histochemistry and immunohistochemistry analyses
Human paraffin-embedded tissues were sectioned at a thickness of 4 µm. The tissue sections were then subjected to Masson trichrome and immunohistochemical (IHC) staining. Standard operating procedures for Masson and IHC staining were as previously described (10, 33). Specifically, Masson staining was performed using a commercial Trichrome Stain Kit (Cat FH115100, FreeThinking, Nanjing, China) in accordance with the manufacturer’s protocol. Primary antibodies utilized were as follows: CD8 (prediluted, Cat. PA067, Abcarta, Suzhou, China), PD1 (prediluted, Cat. PA153, Abcarta), GZMB (1:3000 dilution, Cat. ab255598, Abcam, Cambridge, England), CD56 (prediluted, Cat. PA211, Abcarta), CD19 (prediluted, Cat. GT2128, GeneTech, Shanghai, China), CD86 (1:500 dilution, Cat. ab269587, Abcam), CD163 (prediluted, Cat. ab74604, Abcam), CD31 (1:2000 dilution, Cat. ab182981, Abcam), α-SMA (1:2000 dilution, Cat. ab124964, Abcam), PD-L1 (prediluted, Cat. GT2280, GeneTech), p-ERK (1:1000 dilution, Cat. 4370, Cell Signaling Technology, Danvers, USA), MLH1 (prediluted, Cat. GT2304, GeneTech), MSH2 (prediluted, Cat. GT2310, GeneTech), MSH6 (prediluted, Cat. GT2195, GeneTech), PMS2 (prediluted, Cat. GT2159, GeneTech), and IgG (1:100 dilution, Cat. ab172730, Abcam).
Masson staining evaluation involved determining positively stained area percentages and IHC analysis for most biomarkers were analyzed by determining the rate of positive cells using the HALO software (Albuquerque, NM, USA). Collagen deposition was classified as low (<10% area) or high (≥10% area), with a cutoff of 10% (10). PD-L1 staining was quantitatively assessed based on the combined positive score (CPS) criterion by two senior pathologists. For mismatch repair genes (MLH1, MSH2, MSH6, and PMS2) IHC analysis, two senior pathologists directly determined the positive and negative cases. Supplementary Figure 1 exhibited the positive and negative staining for various markers in gastric cancer and lung cancer.

Cell lines and cell culture
The HGC27 gastric carcinoma cell line was used as the primary disease-relevant model. The H1299 NSCLC cell line was selected to investigate the potential conservation of the collagen-MAPK signaling axis across different solid tumor types. Human cancer cell lines HGC27 (Cat. KGG3287-1) and NCI-H1299 (Cat. KGG3216-1) were purchased from KeyGEN (Nanjing, China). The vascular endothelial cell line HUVEC (Cat. SC0396) THP1 mononuclear cell line (Cat. SC0071) were purchased from YUCHI Biology (Shanghai, China). HGC27, NCI-H1299, and THP1 cells were cultured in RPMI-1640 medium supplemented with 10% FBS at 37 °C with 5% CO2. HUVEC cells were cultured in endothelial cell medium (Cat. 1001, ScienceCell, California, USA) at 37 °C with 5% CO2. Primary cancer-associated fibroblasts (CAFs) from gastric cancer tissues were extracted as previously described (34). All human cell lines were authenticated by short tandem repeat profiling, and all experiments were conducted in the absence of mycoplasma contamination. To differentiate THP-1 monocytic cells into macrophages, the cells were treated with 200 ng/mL of Phorbol 12-myristate 13-acetate (PMA, Cat. HY-18739). For ERK inhibition, 1 µM Ravoxertinib (Cat. T6511, TargetMol, Shanghai, China) was used.

Western blotting analysis and immunofluorescence
Total protein was extracted from human cells using lysis buffer, followed by SDS-PAGE and Western blotting according to standard protocols. The following primary antibodies were used: p-MEK (Cat. 9154, Cell Signaling Technology; 1:1000), MEK (Cat. A4868, Abclonal; 1:1000), p-ERK (Cat. 9102, Cell Signaling Technology; 1:1000), ERK (Cat. A4782, Abclonal; 1:1000), and Vinculin (Cat. 66305-1-Ig, ProteinTech; 1:5000). Vinculin was used as a loading control for normalization. The subcellular localization of ERK was examined by immunofluorescence using a specific antibody (Cat. A4782, Abclonal; 1:500), and images were acquired with a fluorescence microscope.

In vitro macrophage polarization assay
To induce M2 polarization, THP-1-derived macrophages were stimulated with 20 ng/ml IL-4 (Cat. KGD1203, KeyGEN) for 24 hours. The polarization status was then assessed by flow cytometry, measuring the expression of the M1 marker CD86 and the M2 marker CD163. The following antibodies were used: APC anti-CD86 (Cat. PE-65165, ProteinTech) and PE anti-CD163 (Cat. APC-65169, ProteinTech).

Collagen coating and functional assays
Prior to cell culture, plates were pre-coated with Type I collagen (Cat. A1048301, Gibco, Thermo Fisher Scientific, MA, USA) using a method standardized in our laboratory and consistent with the manufacturer’s protocol (35). The final concentration used in vitro assays was 10 μg/cm2. Cells were seeded onto culture plates for collagen stimulation for 24 hours.
Cell proliferation was assessed by CCK-8 assay. Briefly, cells were seeded in 96-well plates at 5 × 103 cells/mL (100 μL/well) and cultured for 24 hours. After adding 10 μL of CCK-8 reagent (Cat. KGA9310, KeyGEN), the plates were incubated for 1 hour, and the absorbance at 450 nm was measured with a microplate reader. Cell migration and invasion were evaluated using Transwell chambers, uncoated or pre-coated with Matrigel, respectively. Following digestion with 0.25% trypsin, 5 × 104 cells in 200 μL serum-free medium were plated in the upper chamber, while the lower chamber was filled with 600 μL medium containing 10% FBS. After 24 hours, cells that had traversed the membrane were fixed with 4% paraformaldehyde, stained with 0.2% crystal violet, and quantified by counting three random 100× fields. Apoptosis was analyzed using an Annexin V-FITC/PI Kit (Cat. KGA1102, KeyGEN) per the manufacturer’s instructions. In accordance with the kit’s recommendation, only early apoptotic cells were enumerated to avoid overlap with late apoptotic and necrotic populations. For the tube formation assay, HUVECs (3 × 104 cells/well) were seeded onto a solidified layer of Matrigel (50 µL/well, Cat. KGL5101, KeyGEN) in a 96-well plate and incubated for 6 hours. Tube formation was assessed by counting tubular structures from three random 100× microscopic fields.

Tumor-bearing mouse model and drug treatment
All experimental procedures used 5-week-old male 615 mice supplied by Hangzhou Ziyuan Animal Co., Ltd. (Hangzhou, China). Animals were housed in a specific pathogen-free environment with a 12-h light/dark cycle, controlled temperature (20–24 °C), and ad libitum access to food and water. All mouse studies were approved by the Laboratory Animal Ethics Committee of Wuxi People’s Hospital (No. DL2024008). The murine gastric cancer cell line MFC (Cat. KGG2227-1, KeyGEN Biotech, Nanjing, China) and fibroblast cell line 3T3 (Cat. KGG1305-1, KeyGEN Biotech) were cultured in Dulbecco’s minimum essential medium supplemented with 10% FBS at 37 °C with 5% CO2. All cell lines were free from mycoplasm. A mouse model of gastric cancer was established by subcutaneous inoculation of approximately 5 × 106 tumor cells, either alone or pre-mixed with fibroblasts at a 1:1 ratio, into each 615 mouse. Tumor dimensions were measured every 2–3 days using calipers, and the volume was calculated as (length × width2)/2. When the average volume of control tumors (without fibroblasts) reached approximately 100 mm3, mice bearing tumors derived from the cell-fibroblast mixture were randomly divided into two groups. One group received daily oral gavage of phosphate-buffered saline (PBS) as a vehicle control, while the other was treated with 20 mg/kg Ravoxertinib (Cat. T6511, TargetMol) administered orally each day.
On day 21 post-treatment initiation, tumors were excised from anesthetized animals, documented, and weighed. The harvested tumor tissues were then processed for subsequent Masson and IHC staining. Primary antibodies utilized were as follows: α-SMA (1:2000 dilution, Cat. ab124964, Abcam), p-ERK (1:1000 dilution, Cat. 4370, Cell Signaling Technology), Ki67 (1:1000 dilution, Cat. ab15580, Abcam), CD163 (1:500 dilution, Cat. ab182422, Abcam), and CD31 (1:2000 dilution, Cat. ab182981, Abcam).

Statistical analysis
All statistical analyses and graphical representations were performed using R language (v4.0.2) and GraphPad Prism (v6.0). Group differences were analyzed with the Student’s t-test or Mann-Whitney test for two groups, and with one-way ANOVA or Kruskal-Wallis test followed by multiple comparisons for multiple groups. Associations between categorical variables were evaluated using the chi-square or Fisher’s exact test, while correlations were evaluated by Pearson’s or Spearman’s method. The diagnostic performance of candidate biomarkers was assessed by receiver-operating characteristic (ROC) curve analysis, with the area under the curve (AUC) quantifying their specificity and sensitivity. For survival data, the prognostic significance of categorical variables was assessed using the log-rank test. A P-value of less than 0.05 was considered statistically significant.

Results

Clinical relevance of collagen deposition in gastric cancer
Given that the exact clinical value of collagen deposition has not been well defined, we checked the differential expressions and prognostic values of various collagens in gastric cancer, one of cancer types with intermediate collagen deposition levels in solid tumors (10). In the TCGA-STAD cohort, compared with para-tumor tissues, most collagen molecules were notably increased in cancer tissues (Figure 1A), and was correspondingly higher in tumor samples (Figure 1B). Nearly 50% of these collagen molecules were associated with unfavorable clinical outcomes (Figure 1C). Furthermore, a composite collagen score, derived from all collagen molecules, was significantly elevated in tumors and also predicted poorer prognosis (Figure 1D). Moreover, collagen deposition was associated with advanced T stage and TNM stage (Supplementary Figure 2). We also examined the predictive value of collagen deposition in gastric cancer in the PRJEB25780 cohort. As shown in the results, most collagen molecules were up-regulated in tumor tissues from these non-responders compared with responders to immunotherapy (Figure 1E). Interestingly, collagen score was significantly related to the response to immunotherapy and the predictive value of collagen score was even close to PD-L1 expression and T cell inflamed score (Figures 1F, G). We also validated these results in the in-house cohorts. We detected the total collagen level by the Masson staining, and the results exhibited that total collagen level was remarkably increased in tumor tissues (Figures 1H, I). In addition, high collagen level was linked with poor prognosis in the validated cohort (Figure 1J). Overall, collagen was significantly accumulated in gastric tumor tissues and associated with critical clinical outcomes.

Correlations between collagen and microenvironmental cell fractions in gastric cancer
Next, we checked the potential correlations between collagen deposition and the features of the TME. Although expression patterns of MHC molecules, immunostimulators, chemokines, and their receptors did not differ markedly between gastric cancer tissues with low versus high collagen scores (Figure 2A), we further investigated the cellular composition of the TME given its inherent complexity. Using the MCP-counter algorithm, we estimated the relative abundance of immune and stromal cell populations in each sample. Pearson correlation analyses demonstrated significant associations between collagen scores and specific TME cell subsets. Collagens score was positive correlated with fibroblasts, endothelial cells, and monocytic lineage (Figure 2B). Subsequent analysis of macrophage polarization revealed a specific positive correlation with the M2 subset, but not with M1 macrophages (Figures 2C, D). Moreover, the results from the PRJEB25780 dataset also confirmed the positive correlations between collagen score between fibroblasts, endothelial cells, and M2 macrophage (Supplementary Figures 3A–D). For further validation, we conducted Masson staining and IHC analysis of various cell subsets (Figure 2E), illustrating positive correlations between collagen and M2 macrophages, endothelial cells, and fibroblasts (Figures 2F–H, Supplementary Figure 4). To conclusion, these results uncovered the positive correlations between collagen and M2 macrophages, endothelial cells, and fibroblasts, which might account for the oncogenic role of collagen in cancer.

Correlations between collagen and crucial molecular events in gastric cancer
We also evaluated the associations between collagen and crucial molecular events in gastric cancer. Although collagen score was associated with the immunotherapeutic responses, collagen scores exhibited no significant correlations with PD-L1 expression and tumor mutational burden (TMB) in the TCGA dataset (Figures 3A, B). Also, collagen levels were not significantly various across microsatellite instability (MSI) status subgroups in the TCGA dataset (Figure 3C). In the PRJEB25780 dataset, collagen score were not associated with most immune checkpoints and PD-L1 expression (Supplementary Figures 5A–D). In the in-house cohort, no significant differences in collagen distribution were observed based on mismatch repair (MMR) status (Figures 3D, E). No obvious correlation was found between collagen distribution and PD-L1 expression as well (Figures 3F, G). To understand the potential molecular mechanisms of collagen in cancer, GSEA linked high collagen scores to activation of ERK signaling pathways (Figure 3H, Supplementary Table 2), consistent with a strong positive correlation between collagen levels and p-ERK expression in the in-house cohort (Figures 3I, J). These findings underscore collagen deposition as a key regulator of both immunosuppressive TME remodeling and oncogenic signaling in cancer.

Pan-cancer exploration of correlations between collagen and immune & molecular features
Based on the above findings in gastric cancer, we further validated the correlations between collagen and M2 macrophage, angiogenesis, and the MAPK signaling pathways in other solid cancers. In the TCGA database, we found that collagen score was positively correlated with M2 macrophage level and angiogenesis activity in most solid cancer types, especially carcinomas (Figures 4A, B). In addition, one in-house lung cancer cohort was used for validation. Notably, the collagen levels were positively related to the positive rates of both CD163 and CD31, the markers for M2 macrophage and vascular endothelial cells (Figures 4C–E). Also, we explored the correlation between collagen and the activity of the MAPK signaling pathway, and the results exhibited that collagen was positively correlated with the activity of the MAPK signaling pathway, especially lung cancer (Figure 4F). Moreover, these results were also validated in the in-house lung cancer cohort by detecting the total collagen level by the Masson staining and the positive rate of p-ERK by the IHC staining (Figures 4G, H). The predictive value of collagen for immunotherapy was also checked in more solid cancer types. The results revealed that high collagen score predicted the resistance to immunotherapy in most cancer types suitable for immunotherapy, including melanoma, urothelial cancer, triple-negative breast cancer, and non-small cell lung cancer (Supplementary Figures 6A–D). Overall, the correlations between collagen and immuno-suppressive TME and immunotherapeutic resistance were conserved in pan-cancer.

Oncogenic role of collagen was dependent on the MAPK signaling pathway in various cells
Given the tight correlations between collagen deposition and the MAPK signaling pathway, we speculated that collagen-mediated tumor cells aggressiveness, macrophage M2 polarization, and angiogenesis were based on the activity of the MAPK signaling pathway. We first checked the activities of the MAPK signaling pathway in various cell types in tumor tissues, and the results showed that the MAPK signaling pathway was conserved in various cells, especially macrophages and vascular endothelial cells (Figures 5A, B). We also validated the expression and phosphorylated levels of ERK, the critical molecule of the MAPK signaling pathway, which was expressed in all cell types, including tumor cells H1299 and HGC27, CAFs, endothelial cells HUVEC, and macrophage THP1 (Figure 5C). Next, we checked the effects of collagen on the MAPK signaling pathway in stromal cells, including HUVEC and THP1 cells. Collagen enhanced the phosphorylated levels of MEK and ERK and promoted the nuclear translocation of ERK in HUVEC and THP1 cells (Figures 5D, E). The scRNA-seq analysis indicated that macrophages with high MAPK activities exhibited increased M2 phenotype (Figure 5F). Additionally, collagen stimulation significantly elevated p-p38 levels in H1299 and HGC27 cells, consistent with earlier observations. By contrast, no notable p-p38 activation was detected in THP-1 or HUVEC cells following collagen treatment (Supplementary Figure 7), further support the cell-type-specific activation of p38 and reinforce the conclusion that the classical MEK/ERK pathway acts as the common downstream signaling cascade activated by collagen across the relevant cellular models. In vitro assays indicated that collagen promoted the M2 polarization of THP1 cells, but the ERK inhibitor Ravoxertinib could reversed the promoting effects (Figure 5G). Moreover, the MAPK activities were associated with angiogenesis activities in the scRNA-seq analysis and the ERK inhibitor Ravoxertinib inhibited the collagen-mediated angiogenesis of HUVEC cells (Figures 5H, I). Given that the collagen is essentially derived from CAFs, we also evaluated the cellular effects of ERK inhibition on CAFs. The results showed that Ravoxertinib inhibited collagen expression and the migration of CAFs (Figures 5J, K).
In addition, we also evaluated the effects of collagen on the MAPK signaling pathway in tumor cells. We found that collagen enhanced the phosphorylated levels of MEK and ERK and promoted the nuclear translocation of ERK in H1299 and HGC27 tumor cells (Figures 6A, B). Based on the scRNA-seq analysis, we found that the activities of the MAPK signaling pathway were related to the proliferation and aggressiveness potentials in tumor cells (Figure 6C). Further in vitro assays indicated that collagen promoted the proliferation, migration, and invasion of H1299 and HGC27 tumor cells, but the ERK inhibitor Ravoxertinib could reversed the promoting effects (Figures 6D, E). In addition, the ERK inhibitor Ravoxertinib also enhanced the collagen-mediated inhibition of tumor cells apoptosis (Figure 6F). Totally, the inhibition of the MAPK signaling pathway suppressed collagen-mediated cancer progression by regulating both tumor cells and stromal cells.

Inhibition of the MAPK signaling pathway suppressed collagen-mediated cancer progression in vivo
To further assess the effects of inhibition of the MAPK signaling pathway on cancer progression, we designed an in vivo assay. We mixed mouse 3T3 fibroblasts and MFC gastric cancer cells for subcutaneous injection to simulate more collagen deposition in the TME, and Ravoxertinib was injected intraperitoneally to block the MAPK signaling pathway (Figure 7A). As the results shown, mixed with fibroblasts and Ravoxertinib treatment did not change mouse weight (Figure 7B). However, mixed with fibroblasts significantly accelerated tumor growth, both tumor volume and tumor weight were enhanced remarkably in mice, but Ravoxertinib inhibited tumor progression (Figures 7C–E). Furthermore, histological staining of major organs (heart, liver, spleen, lung, and kidney) combined with liver and renal function analyses demonstrated that Ravoxertinib therapy was well-tolerated (Figures 7F, G). Histological analysis was performed to check the levels of collagen, α-SMA, p-ERK, Ki67, CD163, and CD31. The results showed that mixed with fibroblasts significantly enhanced collagen deposition and the expression of α-SMA, p-ERK, Ki67, CD163, and CD31 in tumor tissues, but Ravoxertinib therapy reversed these elevations (Figure 7H). In summary, these findings suggested that inhibition of the MAPK signaling pathway suppressed collagen-mediated cancer progression in vivo.

Discussion
In the current research, we systematically characterized the clinical and molecular implications of intratumoral collagen deposition in solid cancers, with a focus on gastric cancer. Our findings revealed that elevated collagen levels are significantly associated with poor prognosis. Through comprehensive spatial and molecular analyses, we identified a strong positive correlation between collagen deposition and the abundance of vascular endothelial cells and M2-polarized macrophages within the TME. Furthermore, pathway enrichment analysis demonstrated that collagen-mediated activation of the MAPK signaling pathway plays a pivotal role in promoting tumor cell aggressiveness, angiogenesis, and M2 macrophage polarization. These results collectively highlight the multifaceted role of collagen in shaping the TME and driving tumor progression.
The TME constitutes a complex ecosystem that encompasses a diverse array of stromal and immune cells, extracellular matrix (ECM) components, and signaling molecules (36–38). Among these, collagen, as a major ECM component, has been increasingly recognized for its role in modulating cellular behaviors and signaling pathways (39, 40). Our study adds to this growing body of evidence by uncovering that collagen deposition is closely linked to the recruitment and functional polarization of vascular endothelial cells and M2 macrophages. Specifically, the positive correlation between collagen and M2 macrophages suggests that collagen may facilitate the establishment of an immuno-suppressive TME, which is known to promote tumor immune evasion and progression (41, 42). Similarly, the association with vascular endothelial cells underscores the role of collagen in supporting angiogenesis, a critical process for tumor growth and metastasis (35, 43).
One of the most intriguing results of our study is the identification of the MAPK signaling pathway as a key downstream effector of collagen-mediated TME remodeling. The MAPK signaling pathway is well-known for its involvement in cell proliferation, survival, and differentiation (44, 45). Our data suggest that collagen deposition activates MAPK signaling in tumor cells, endothelial cells, and macrophages, thereby driving multiple pro-tumorigenic processes. For instance, MAPK activation in tumor cells may enhance their invasive potential (46), while in endothelial cells, it may promote angiogenesis (47, 48). In macrophages, MAPK signaling has been implicated in M2 polarization (49), which is consistent with our observation of increased M2 macrophage abundance in high-collagen tumors. Notably, analysis of gastric cancer scRNA−seq data revealed that tumor cells, macrophages, and endothelial cells with high MAPK activity frequently co-upregulated several integrin subunits, specifically ITGB1, ITGB5, and ITGAV (Supplementary Figures 8A–D). These subunits form part of known collagen-binding integrins (e.g., α1β1, α2β1, αvβ5) and have been previously linked to MAPK pathway activation (50–52). For instance, a study by Wu et al. showed that COL4A2 promotes glioblastoma vascularization by activating MAPK-ERK signaling through ITGA1/ITGB1 receptors on tumor-associated endothelial cells (52). Based on these observations, we propose ITGB1, ITGB5, and ITGAV as candidate receptors that may mediate collagen−induced MAPK signaling in the gastric cancer TME, positioning them as important targets for future functional validation. These findings provide a mechanistic link between collagen and the functional reprogramming of TME components.
The strong association between collagen deposition and poor clinical outcomes highlights its potential as a prognostic indicator in gastric cancer. Our results suggest that patients with high intratumoral collagen levels may benefit from more aggressive therapeutic strategies. Additionally, the identification of the MAPK signaling pathway as a central mediator of collagen’s effects opens new avenues for targeted therapy. For example, MAPK inhibitors, either alone or in combination with existing therapies (53, 54), could be explored to disrupt collagen-mediated TME remodeling and improve patient outcomes. Furthermore, targeting collagen itself or its downstream effectors may represent a novel therapeutic strategy to counteract tumor progression (34, 55).
Despite the insights our study provides into collagen’s role in gastric cancer, several limitations must be noted. First, although the in vivo experiment demonstrated that the MAPK inhibitor Ravoxertinib effectively counteracted the tumor-promoting effect of fibroblast co-injection, the experimental design lacked a control group receiving Ravoxertinib monotherapy. Consequently, the present data cannot conclusively rule out a concomitant general growth-inhibitory effect of Ravoxertinib on MFC cells themselves, which warrants further investigation to delineate the precise mechanism of action. Secondly, the generalizability of our findings to other solid cancers remains to be explored. In addition, our co-transplantation model could not isolate the specific role of collagen from other factors secreted by CAFs. However, supporting in vitro evidence confirms that purified collagen alone is sufficient to drive tumor−promoting phenotypes via the MAPK signaling pathway. Moreover, the collagen receptors were various in different cell types, how these collagen receptors linked to the MAPK signaling pathway was perplexing. Future studies should explore whether similar collagen-driven mechanisms exist in other cancers, and whether targeting these pathways holds broader therapeutic relevance.

Conclusion
In conclusion, our study elucidates the clinical and molecular significance of intratumoral collagen deposition in gastric cancer and other solid cancers. We demonstrate that collagen accumulation is associated with an immunosuppressive and pro-angiogenic TME, driven in part by MAPK signaling pathway activation. These findings not only enhance our understanding of the complex interplay between ECM components and TME dynamics but also provide a rationale for developing novel therapeutic strategies targeting collagen and its downstream signaling pathways. Future research should focus on translating these insights into clinical applications to improve outcomes for patients with gastric and other solid cancers.