Skip to main content
Download PDF

Tumor-associated-fibrosis and active collagen-CD44 axis characterize a poor-prognosis subtype of gastric cancer and contribute to tumor immunosuppression

Journal of Translational Medicine volume 23, Article number: 123 (2025) Cite this article

Abstract

Background

Tumor-associated fibrosis modifies the tumor microenvironment (TME), hinders the infiltration and activity of cytotoxic immune cells, and is a critical pathological process leading to the ineffectiveness of tumor immunotherapy in gastric cancer (GC). However, the specific mechanisms and interventions are yet to be fully explored.

Methods

Our study included 375 gastric cancer samples from TCGA, 1 single-cell RNA sequencing (scRNA-seq) dataset comprising of 15 gastric cancer samples from GEO, 19 cohorts of immunotherapy and 2 GWAS datasets. Consensus clustering identified a gastric cancer subtype characterized primarily by fibrosis, and various methods such as pseudotime analysis, CellChat analysis and Colocalization analysis were used to explore its mechanisms.

Results

A subtype of gastric cancer was identified with poor prognosis, characterized by higher malignancy, drug resistance, and poor immune infiltration, associated with elevated expression of genes related with Extracellular matrix (ECM). Single-cell transcriptome analysis showed active Collagen-CD44 signaling axis between cancer-associated fibroblasts (CAFs) and immune cells in gastric cancer, with ECM-related genes upregulated during tumor progression. The expression of CD44 was significantly elevated in the subtype, associated with poor prognosis and tumor immune suppression in gastric cancer, potentially involved in the recruitment of immunosuppressive cells such as M2 macrophages and regulatory T cells (Tregs) and the upregulation of multiple immune checkpoints including PD-1/PD-L1.

Conclusion

Our study identified a new subtype of gastric cancer, revealing that fibrosis is a critical mechanism driving immune suppression in gastric cancer and emphasizing the central role of the Collagen-CD44 signaling axis. The Collagen-CD44 signaling axis has the potential to serve as a novel therapeutic target for gastric cancer by enhancing immune cell-mediated tumor suppression. By combining it with immune checkpoint inhibitors (ICIs), it may improve the efficacy of immunotherapy for gastric cancer and offer new hope for treatment.

Background

Gastric cancer (GC) is a highly heterogeneous malignancy [1], making it difficult to find broadly effective treatment options. Although some molecular markers for targeted therapy in gastric cancer are available, their positivity rates are low; for example, the positivity rate of HER2 is only 14% in gastric cancer patients [2, 3]. In patients with advanced gastric and esophageal cancers, the efficacy rate of PD-1 therapy is only 15% [4, 5]. Therefore, it is necessary to explore different molecular subtypes of gastric cancer to provide references for designing more personalized treatment plans, with the tumor extracellular matrix (ECM) being a potential therapeutic target.

Tumor-associated fibrosis is a key feature of gastric cancer. Certain cytokines, such as TGF-β and IL-6 [6], promote the differentiation of fibroblasts into cancer-associated fibroblasts (CAFs), with identified subtypes including FAP + and α-SMA+ [7]. These subtypes synthesize and secrete higher levels of ECM components [8], such as fibronectin and type I collagen, leading to fibrosis [9]. The presence of CAFs is strongly correlated with the invasiveness and prognosis of gastric cancer, particularly in scirrhous gastric cancer [10]. Scirrhous gastric cancer is a distinct subtype of advanced gastric cancer, characterized by extensive stromal fibrosis and widespread tumor cell infiltration. In the gastric cancer microenvironment, CAFs can remodel the ECM, influencing tumor behavior [11], for example, by secreting matrix metalloproteinases (MMPs) to regulate tumor cell migration [12].

Tumor-associated fibrosis exerts a profound impact on the tumor immune microenvironment and the efficacy of tumor immunotherapy in terms of both physical stiffness and chemical environment. On one hand, the tumor ECM’s rigidity is approximately 1.5 times that of the surrounding normal tissue [13]. The dense and orderly arrangement of collagen fibers and the abnormal upregulation of adhesion proteins direct T cells into the tumor stroma rather than the tumor cell nests, impeding their access to tumor cell clusters [14]. This also facilitates tumor cell adhesion, migration, and metastasis. Integrin receptors in the ECM are capable of transducing mechanical stimulation signals [6], enhancing the immunosuppressive function of Treg cells via the TGFβ-SOX4 and Hippo signaling pathways, upregulating PD-L1 expression in tumor cells [15], and inhibiting the activity of CD8 + T cells. Additionally, fibrosis increases internal tumor pressure, leading to vascular leakage, drug resistance, and tumor hypoxia [13]. On the other hand, the ECM is not merely a structural framework between cells; many of its components, such as hyaluronic acid (HA) and collagen [16], also function as messengers. Collagen is one of the primary molecules produced during tumor-associated fibrosis, mainly synthesized and secreted by CAFs [16]. The main receptors for collagen include DDR1 and DDR2 from the discoidin domain receptors (DDRs) and integrin receptors [16]. In pancreatic ductal adenocarcinoma (PDAC), collagen exerts a promotional effect on PDAC growth through cCOLI-DDR1-NF-κB-p62-NRF2 signaling [17, 18]. Integrin α11β1 is a stromal cell-specific fibrillar collagen receptor that is overexpressed in CAFs [19]. The loss of α11 expression in the tumor stroma causes collagen reorganization and reduced stiffness [20], while forced expression of β1 integrin significantly stimulates the phosphorylation of Src and extracellular signal-regulated kinase (ERK), leading to enhanced metastatic potential of cancer cells [21]. Additionally, CD44 serves as an additional receptor for collagen, though the downstream signaling triggered by their interaction remains to be studied [22, 23].

In conclusion, a thorough understanding of CAFs and ECM is crucial for the treatment of gastric cancer. Therapies that target CAFs and ECM-related molecules, either as monotherapy or in conjunction with chemotherapy and immunotherapy represents a significant research direction in cancer treatment [24,25,26]. Here, we collected multiple datasets, including samples from The Cancer Genome Atlas (TCGA), scRNA-seq of gastric cancer, 19 immunotherapy cohorts and 2 Genome-Wide Association Studies (GWAS) datasets. A poor-prognosis subtype of gastric cancer characterized by highly expressed ECM-related genes was identified and its potential mechanism was discussed through a series of analytical approaches, including pseudotime analysis, CellChat analysis, and colocalization analysis.

Methods

Study design

We first identified 138 risk genes in gastric cancer as a clustering gene set through univariate Cox regression (OR > 1.2, P < 0.05), and obtained two gastric cancer subtypes through consensus clustering, describing their characteristics. Differential analysis was used to further explore the gene expression characteristics of the two gastric cancer subtypes, finding that ECM-related genes were significantly upregulated in the gastric cancer subtype with poor prognosis. Using single-cell transcriptome data, we further explored the expression and regulatory mechanisms of these genes through pseudotime analysis and Cellchat analysis, discovering significant activity in the Collagen-CD44 signaling axis. Finally, we comprehensively explored the role of CD44 in gastric cancer using GWAS data and immunotherapy cohorts.

Data source

RNA sequencing data (level 3) and clinical information were collected from 375 gastric cancer samples and 32 adjacent non-tumor samples from TCGA database [27] and from 359 normal gastric samples from the Genotype-Tissue Expression Project (GTEx) database [28]. The expression matrices and clinical information of GSE167297, GSE93157, GSE135222, GSE126044, GSE93157, GSE91061, GSE78220, GSE145996, GSE115821, GSE106128, GSE100797, GSE67501, and GSE93157 were obtained from the Gene Expression Omnibus (GEO) database [29], with GSE167297 containing 15 single-cell transcriptome samples, and the rest being samples from immunotherapy cohorts. PRJEB25780 and PRJEB23709 were obtained from the European Nucleotide Archive (ENA) [30], and phs000452 was obtained from Database of Genotypes and Phenotypes (dbGaP) [31]. The Nathanson_2017 and Braun_2020 datasets were based on previous studies [32, 33]. E-MTAB-6270 was obtained from the ArrayExpress database [34]. PRJNA482620 was sourced from the BioProject database [35]. The GWAS dataset ebi − a−GCST90018849 was obtained from the NHGRI-EBI GWAS Catalog [36], and the eQTL dataset eqtl − a−ENSG00000026508 was obtained from the EMBL-EBI eQTL Catalog [37]. In summary, we incorporated bulk transcriptomic and clinical data on gastric cancer from TCGA, single-cell transcriptomic data on gastric cancer from GEO, GWAS data from the Catalog database, and clinical cohorts from immunotherapy studies. By integrating diverse data types from multiple platforms with cross-validating results, we improved the reliability, robustness, and generalizability of the study.

Gastric cancer subtype analysis

The association of each gene with overall survival (OS) was assessed using the Cox proportional hazards model. For genes with significant prognostic significance, consensus clustering analysis was performed using ConsensusClusterPlus (v1.54.0). Consensus clustering reduces the effects of noise and outliers, efficiently captures sample heterogeneity, and is suitable for tumor subtyping, providing robustness and reproducibility. Furthermore, Consensus clustering can determine the optimal number of clusters by calculating consistency metrics, such as the CDF curve. A total of 100 repetitions were performed, extracting 80% of the total samples each time, using “hc” as the clustering algorithm with internal linkage by ‘ward.D2’. Principal component analysis (PCA) and clustering heatmaps were analyzed using FactoMineR and pheatmap (v1.0.12), respectively. Clinical staging and GX grading were statistically analyzed using ggplot2 (v3.3.2). Prognostic differences between the two groups were statistically analyzed using the survival and survminer packages, employing the log-rank test, and Kaplan-Meier survival curves were plotted.

IC50 drug sensitivity prediction and immune landscape analysis

Based on the Genomics of Drug Sensitibity in Cancer (GDSC) database [38], we employed pRRophetic to forecast the chemotherapy response of samples. The half-maximal inhibitory concentration (IC50) of the samples was estimated using ridge regression. The distribution of 11 immune cells and the expression heatmap of immune checkpoints in these two samples were plotted utilizing the QUANTISEQ algorithm and pheatmap tool. Based on expression profile data, the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm was used to predict the response of individual samples or specific subtypes to immune checkpoint inhibitors (ICIs). We performed immune scoring using the TIMER algorithm in the “immunedeconv” package.

mRNA differential expression, functional enrichment, and PPI network analysis

We analyzed mRNA differential expression using the Limma package and performed GO functional and KEGG pathway enrichment analysis using the ClusterProfiler package. A protein-protein interaction (PPI) network was constructed using the STRING database [39], and the top 10 hub genes among the differentially expressed genes were identified using the “cytohubba” plugin in Cytoscape. We quantified gene correlations via Spearman correlation analysis, analyzed the association between genes and immune checkpoints and immune cells using the QUANTISEQ algorithm. The resultant associations were visualized in heatmaps generated through ggplot2 and pheatmap.

Single-cell transcriptome data preprocessing and gene expression analysis

The selection of cells for analysis was predicated on the following criteria: cells with more than 1,000 unique molecular identifier (UMI) counts, cells with more than 500 but less than 4,000 unique genes, and cells with less than 25% mitochondrial gene expression in their UMI counts. Cell type annotation was facilitated by the FindCluster function, selecting the resolution parameter (resolution = 0.6) to balance the number of clusters and the data structure. Sankey diagrams and pie charts were plotted based on the proportions of annotated cell types. Expression levels were illustrated via bubble plots, generated with FlexDotPlot. For each cell type, we determined the area under the receiver operating characteristic (ROC) curve (AUC) using the AUC package, then plotted violin plots.

Intercellular communication dynamics analysis

CellChat analysis was conducted to identify signaling pathways mediating interactions between fibroblasts and other cells in the tumor microenvironment (TME). Utilizing the CellChat package, we dissected the gene expression patterns that govern intercellular interactions, providing a comprehensive view of the communication across different cellular contexts. The intensity of intercellular communication was displayed using the netVisual_heatmap function; the main senders and receivers of intercellular communication, as well as their degree of connectivity, were visualized using the netAnalysis_signalingRole_scatter and netVisual_aggregate functions, respectively; the ligand-receptor signaling pathways of intercellular communication were displayed using the netVisual_bubble function.

Pseudotime trajectory analysis

Pseudotime analysis was employed to trace the dynamic differentiation trajectory of fibroblasts transitioning from superficial gastric cancer tissue to deep gastric cancer tissue. The progress also involved tissue stem cells, smooth muscle cells, and chondrocytes. We inferred the pseudotime trajectory of cells using Monocle2, determining the sequential progression of cell development. Dimensionality reduction analysis was performed using the reduceDimension function of Monocle2, leveraging the DDRTree algorithm. The learn_graph function was used to construct a graph model of cell state transitions. The number of nearest-neighbor cells (kNN = 5) and the root node state (root_state) were set as fibroblasts, chondrocytes, smooth muscle cells, and tissue stem cells, respectively. The learn_graph function was used to help identify potential root nodes in the cell state transition graph. The order_cells function was used to order the cells, with parameters optimized for maximum iteration (max_iter = 100) and a defined time step (delta_t = 1). The plot_genes_vs_pseudotime function was used to display the expression patterns of genes along the pseudotime line, with the num_cells parameter set to control the number of cells displayed.

CD44 expression analysis

We analyzed the expression of the CD44 gene using ggplot2 and evaluated its significance using the Wilcoxon test. The cellular expression patterns of CD44 were visualized using the FeaturePlot function of the Seurat package. Prognostic differences between the two groups were analyzed using the survival and survminer packages. Immunohistochemical sections were obtained from the The Human Protein Atlas (HPA) [40]. We plotted the correlation between CD44 expression and tumor immune target expression using the ggstatsplot package, employing Spearman’s rank correlation analysis for non-parametric quantitative relationships. Bayesian colocalization analysis evaluates the probability that CD44 and gastric cancer share the same causal variation. We collected data for eqtl − a−ENSG00000026508 and ebi − a−GCST90018849, performed statistical analysis of GWAS and eQTL data, and calculated the association strength of each locus and the colocalization score to evaluate the overlap between GWAS and eQTL signals. The -log10 (P) values of GWAS and eQTL were plotted on a position map of chromosome 11, showing their distribution across the genome. The truncation value of the colocalization evidence is defined as PP.H4.abf > 0.75. Gene expression profiles and associated clinical data across various tumor types were sourced from a range of public databases. The predictive efficacy of CD44 gene expression levels regarding immunotherapy responses was assessed using ROC curves, with AUC value being calculated for evaluation.

Results

A high-risk subtype of gastric cancer and its clinical parameters, drug sensitivity, and immune characteristics

An initial univariate Cox regression analysis was conducted on sequencing data and clinical data from 375 gastric cancer samples from TCGA, and 138 genes significantly associated with poor prognosis in gastric cancer were identified (OR > 1.2, P < 0.05). Using this gene set, consensus clustering was performed on the 375 gastric cancer samples, resulting in two gastric cancer subtypes: C1 (N = 286) and C2 (N = 89) (Fig. 1A), and a PCA plot was created (Fig. 1B). The study retained feature genes with a variance > 0.1 to plot a heatmap, revealing significant upregulation of many genes in C2 compared to C1 (Fig. 1C). While no notable differences in TNM staging were observed between the subtypes, a higher prevalence of G3 tumors was noted in the C2 group compared to C1, reaching statistical significance (P < 0.05) (Fig. 1D). Moreover, C2 was associated with a markedly inferior prognosis relative to C1, as demonstrated by Kaplan-Meier survival analysis (HR = 1.734, 95%CI: [1.224–2.459], P < 0.01) (Fig. 1E). We selected six common tumor treatment drugs, including paclitaxel, cetuximab, 5-fluorouracil, docetaxel, camptothecin, and cisplatin, to predict drug sensitivity between the two subtypes. The half inhibitory concentration (IC50) of paclitaxel, cetuximab, and 5-fluorouracil was significantly higher in C2 than in C1 (P < 0.0001), indicating that C2 exhibited stronger drug resistance to these three drugs. There was no significant difference in the efficacy of docetaxel, camptothecin, and cisplatin between the two subtypes (Fig. 1F). In terms of immune characteristics, compared to C1, C2 was significantly enriched in CD4 + T cells (P < 0.001), M2 macrophages (P < 0.001), and Treg cells (P < 0.05) (Fig. 1G). Of the eight selected classical immune checkpoints, PD-L2 and TIM-3 were significantly upregulated (P < 0.001) (Fig. 1H). The TIDE algorithm prediction indicated that C2 had a lower response rate to immune checkpoint blockade (ICB) therapy and a significantly higher TIDE score (P < 0.0001) (Fig. 1I).

Fig. 1
figure 1

(A) Consensus clustering identified two gastric cancer subtypes, C1 and C2, from gastric cancer samples; (B) PCA plot of the two gastric cancer subtypes; (C) Heatmap using feature genes shows differences in gene expression between the C1 and C2 groups; (D) Differences in tumor staging and malignancy grading between C1 and C2; (E) Kaplan-Meier curves for overall survival of C1 and C2 groups; (F) Comparison of drug sensitivity to paclitaxel, cetuximab, 5-fluorouracil, docetaxel, camptothecin, and cisplatin between C1 and C2; (G) Heatmap shows the proportion of immune cells calculated by the Quantlseq algorithm between C1 and C2; (H) Heatmap shows the expression of eight classical immune checkpoints in C1 and C2; (I) TIDE algorithm prediction comparing the responsiveness to immunotherapy between the C1 and C2 groups

Full size image

ECM-related hub genes in high-risk gastric cancer subtype

In a subsequent inquiry into the molecular underpinnings of the C2 subtype, a differential analysis juxtaposing C1 and C2 subtypes was conducted, identifying 1248 genes significantly upregulated in C1 and 16 genes significantly downregulated (|log2FC| > 2, P < 0.05) (Fig. 2A). KEGG and GO enrichment analysis on the upregulated genes revealed that 136 genes were significantly enriched in ECM-related pathways (Fig. 2B). The top 10 hub genes were selected using cytohubba plugin: COL1A1, COL1A2, COL3A1, COL5A1, FN1, POSTN, FBN1, LUM, COL4A1, and COL6A3 (Fig. 2C). Except for COL4A1, the remaining genes showed high correlations with each other (Fig. 2D), suggesting that they may have common or similar regulatory mechanisms and spatial locations. To verify whether the immune characteristics of C2 were related to ECM, we analyzed the relationships between these 10 genes and immune checkpoints and immune cell infiltration. The results showed that these 10 genes were significantly positively correlated with the expression of seven immune checkpoints, except for SIGLEC15 (P < 0.05), with the strongest correlations observed with TIM-3. Except for COL4A1, the correlations were all above 0.5 (Fig. 2E), and they were also significantly positively correlated with M1, M2 macrophages, and Treg cell infiltration (P < 0.01) (Fig. 2F). These results align with the phenotype of C2, suggesting that the abnormal overexpression of these genes may be drivers or markers of poor prognosis and immune exclusion phenotype in C2.

Fig. 2
figure 2

(A) Volcano plot showing the differentially expressed genes between C1 and C2; (B) KEGG and GO enrichment analyses reveal that upregulated genes are highly enriched in ECM-related pathways; (C) Top 10 core genes in the pathways identified through PPI and Cytoscape; (D) Correlation analysis between the Top 10 core genes; (E) Heatmap showing the correlation between the Top 10 core genes and immune checkpoints; (F) Heatmap showing the correlation between the Top 10 genes and immune cells

Full size image

Single-cell transcriptome atlas of gastric cancer and ecm-related hub genes expression in different cells

To establish the relationship between ECM and cells in tumor tissues, 15 single-cell samples of diffuse gastric cancer within the GEO database were selected. Post rigorous cell quality control, 2502 cells from normal tissues, 8714 cells from superficial gastric cancer tissues, and 9269 cells from deep gastric cancer tissues were obtained. Through meticulous annotation, the following cell types were identified: epithelial cells, NK cells, B cells, dendritic cells, T cells, endothelial cells, monocytes, tissue stem cells, neutrophils, granulocyte-monocyte progenitors (GMP), macrophages, common-myeloid progenitors (CMP), fibroblasts, CD34-pre B cells, CD34 + pro B cells, chondrocytes, hepatocytes, neurons, CD34 + hematopoietic stem cells, induced pluripotent stem cells, bone marrow cells and progenitors (BM & Prog.), smooth muscle cells, and promyelocytes (Fig. 3A). We calculated the number of each type of cell in normal, superficial, and deep gastric cancer tissues. For immune cells, the proportion of B cells was significantly lower in tumor tissues (superficial & deep) compared to normal tissues, and the proportion of T cells was significantly higher in deep tissues compared to superficial tissues. The proportion of myeloid cells, including monocytes, macrophages, neutrophils, and dendritic cells, was significantly higher. Additionally, the proportion of epithelial cells was lower in tumor tissues and deep tissues compared to normal and superficial tissues, while the proportion of mesenchymal cells was higher, indicating the occurrence of epithelial-mesenchymal transition (EMT) (Fig. 3B). The expression of ECM-related hub genes in cells was subsequently explored. The 10 ECM-related hub genes were mainly co-localized in tissue stem cells, fibroblasts, smooth muscle cells, and chondrocytes. COL1A1, COL1A2, and COL3A1 had the highest expression proportions and average expression levels, with COL4A1 being highly expressed in endothelial cells (Fig. 3C). These results are consistent with the correlation analysis among ECM-related hub genes. Using these 10 genes to construct a gene set for AUC scoring, it was found that the AUC scores were highest in fibroblasts, chondrocytes, and smooth muscle cells in normal, superficial, and deep tissues (Fig. 3D).

Fig. 3
figure 3

(A) UMAP plots of single cells in gastric cancer, showing a total of 23 cell types; (B) Bar and pie charts showing the distribution of cell clusters among normal cells, tumor cells, superficial cells, and deep cells; (C) Bubble plot showing the expression of Top 10 core genes in different cell clusters; (D) UMAP plots and violin plot visualizing the AUC scores of the Top 10 genes in cell clusters

Full size image

Cellchat analysis reveals active collagen-cd44 signaling axis communication in gastric cancer

Subsequently, we analyzed intercellular communication within gastric cancer tissues. Among all pathways identified by the CellChat package, the Collagen pathway exhibited the highest communication intensity, with the primary signal senders being fibroblasts, chondrocytes, smooth muscle cells, tissue stem cells, and nerve cells, while the signal receivers were more diverse (Fig. 4A). Cluster analysis of cellular expression revealed that the expression patterns of fibroblasts, chondrocytes, smooth muscle cells, and tissue stem cells all belonged to pattern 2 (Fig. 4B), exhibiting the same expression pattern for both signal reception and transmission. Active signal reception pathways included NOTCH, FGF, MPZ, CD46, EGF, EDN, PDGF, OSM, PERIOSTIN, VCAM, NEGR, BMP, NT, AGT, PACAP; active signal transmission pathways included MHC-II, SPP1, GALECTIN, VISFATIN, ITGB2, CCL, VEGF, BAFF, RESISTIN, EGF, IL1, GRN, OSM, IL6, SEMA7, ALCAM, CD86, VTN, CD80, VEGI, SEMA6, SN, DESMOSOME, IL10, and PVR (Fig. 4C). Further analysis of the shallow and deep layers of cells revealed that in the shallow layer, the highest communication intensity was observed between tissue stem cells and macrophages, tissue stem cells and epithelial cells, fibroblasts and macrophages, and fibroblasts and epithelial cells. In the deep layer, significant communication was noted between fibroblasts, chondrocytes, smooth muscle cells, tissue stem cells, and nerve cells with epithelial and nerve cells, aligning with the performance of the collagen signaling pathway, indicating that the collagen signaling pathway occupies a significant proportion among all pathways, particularly concentrated in the deep-layer cells (Fig. 4D). Within the shallow layer, macrophages exhibited the greatest signal reception, outpacing other cells, whereas chondrocytes led in signal transmission. Shifting focus to the deep layer, T cells predominated in signal reception, succeeded closely by macrophages and monocytes. Signal transmission was most robust in fibroblasts, chondrocytes, and smooth muscle cells. Comparatively, the deep layer demonstrated heightened cellular communication intensity(Fig. 4E). Our study produced chord diagrams for the collagen signaling pathways in both cellular groups. In the shallow layer, the primary signal senders for collagen signaling were fibroblasts, chondrocytes, smooth muscle cells, and tissue stem cells. In the deep layer, the primary signal senders for collagen signaling were fibroblasts, chondrocytes, smooth muscle cells, tissue stem cells, and nerve cells (Fig. 4F), with higher signal intensity observed in the deep layer. Focusing on fibroblasts, chondrocytes, smooth muscle cells, and tissue stem cells as the signal senders, the study identified significant communication pathways with other cells (prob. > 0.3, P < 0.01), finding that the Collagen-CD44 signaling axis had the highest intensity among all four cell types. COL1A2 predominantly acted as the ligand in chondrocytes, smooth muscle, and fibroblasts, while COL6A2 was identified in stem cells, both interacting with CD44 as the receptor(Fig. 4G).

Fig. 4
figure 4

(A) Heatmaps showing the communication intensity of intercellular pathways; (B) Heatmaps depicting the distribution of cell patterns in cellular interactions and signaling pathways; (C) Sankey diagrams presenting the patterns of different cells in cellular communication, where the width of the flowing lines represents the contribution of signaling pathways to different cells; (D) Heatmaps comparing the differences in communication intensity between superficial and deep tumor tissues; (E) Activity levels of different cell populations as signal senders and receivers during cellular communication. The color of the dots represents different cell populations, and the size of the dots is proportional to the inferred number of ligands and receptors, with the x-axis and y-axis representing the activity intensity of the cell population as signal senders and receivers, respectively; (F) Chord diagrams clearly illustrating the connections and intensity of collagen signaling pathways in intercellular communication. The arcs on the circumference represent different cell types, while the chord lines connecting different arcs represent communication links between cells, with the thickness of the chord lines indicating the strength of communication; (G) Bubble plots revealing the interactions between different cell populations and the correlation between ligand-receptor pairs in cellular communication. The x-axis lists different signaling pathways, and the y-axis shows the correlation of communication between tissue stem cells, fibroblasts, chondrocytes, and smooth muscle cells as signal senders with other cells

Full size image

Cell trajectory analysis of ECM-secreting cells

To further investigate the four cell types with high expression of collagen family genes (fibroblasts, chondrocytes, smooth muscle cells, tissue stem cells), we performed pseudotime analysis on these cells, mapping their differentiation trajectory over time. Three nodes and seven cell clusters were identified, and the distribution of cells from the normal group, shallow layer group, and deep layer group, along with the four cell types, was marked on the trajectory map. Cells from the deep layer group and fibroblasts were primarily located at the end of the pseudotime axis, while cells from the shallow layer group and tissue stem cells were mainly distributed at the beginning of the pseudotime axis (Fig. 5A). We then explored changes in gene expression along the pseudotime axis. APOE, GPC3, CCL2, CFD, LUM, THBS4, C7, CCDC80, EFEMP1, and TIMP1 were significantly upregulated as pseudotime increased, while ADIRF, MYL9, MT1A, ACTA2, ACTG2, TPM2, TAGLN, PPP1R14A, RERGL, BCAM, PLN, SNCG, CRIP1, RGS5, NDUFA4L2, TINAGL1, MYH11, and MCAM were significantly downregulated (Fig. 5B). Our study also presented expression distribution maps (Fig. 5C) of selected genes and expression trajectory maps (Fig. 5D).

Fig. 5
figure 5

(A) Trajectory maps depicting the differentiation trajectories of cells from different depths of the tumor, and of tissue stem cells, fibroblasts, chondrocytes, and smooth muscle cells. The differentiation paths of these cell types over time are shown in the figure; (B) Heatmap showing changes in gene expression levels along the differentiation process. The x-axis represents the psuedotime, and the color change represents changes in the relative intensity of gene expression; (C) The expression levels of specific marker genes over differentiation time in tumor tissues of different depths and different cell types. The x-axis shows the progression of the psuedotime, while the y-axis measures the level of gene expression; (D) The dynamic trajectories of the expression levels of eight marker genes over time

Full size image

CD44 is highly expressed in gastric cancer and Associated with poor prognosis and immunosuppression

We analyzed the expression of CD44 in different samples. CD44 expression in tumors was significantly higher than in the normal group (P < 0.0001) (Fig. 6A), with significantly higher expression in the C2 subtype than in the C1 subtype (P < 0.0001) (Fig. 6B). Immunohistochemical sections from the Human Protein Atlas (HPA) also showed abnormal upregulation of CD44 in gastric cancer tissues (Fig. 6C). CD44 was widely expressed across various cell types (Fig. 6D), with the highest expression observed in monocytes and macrophages (Fig. 6E). The correlation between CD44 expression and the expression of various immune cells and immune checkpoints was significant. CD44 expression was positively correlated with PDCD1LG2, HAVCR2, CD274, TIGIT, PDCD1, LAG3, and CTLA4 (P < 0.001), and also positively correlated with infiltration of B cells, neutrophils, M1 macrophages, M2 macrophages, and Treg cells (P < 0.01) (Fig. 6F). Colocalization analysis of GWAS data for gastric cancer and CD44 revealed that they share common genetic loci (PP.H4.abf > 0.75) (Fig. 6G). Kaplan-Meier survival analysis in gastric cancer samples showed that the group with high CD44 expression had significantly worse prognosis (P < 0.05) (Fig. 6H). Finally, using the pROC package, we predicted the immune response of different tumor groups based on CD44 expression levels. The AUC values for predictions on the melanoma dataset GSE145996 and the head and neck squamous cell carcinoma dataset GSE93157 reached 1, while predictions for two KIRC datasets were less effective, with the AUC for the Braun_2020 dataset being 0.496 and for the GSE67501 dataset being 0.357.

Fig. 6
figure 6

(A) Box plots comparing the differences in CD44 expression levels between tumor tissues (n = 375) and normal tissues (n = 391). (B) Differences in CD44 expression levels between C1 (n = 286) and C2 (n = 89). (C) Immunohistochemistry images comparing the expression patterns of CD44 in the superficial and deep layers of tumor tissues. (D) UMAP plots showing the distribution of CD44 expression levels in superficial and deep cells of tumor tissues, with color intensity positively correlated with CD44 expression levels. (E) Violin plot comprehensively displaying the differences in CD44 expression across various cell types at different depths of tumor tissues. (F) Scatter plots revealing the correlation between CD44 and other immune checkpoint genes and immune cells. (G) Colocalization analysis of eQTL-GWAS for CD44 and gastric cancer. The labeled SNP is the lead SNP, and other SNPs are colored according to their LD with the lead SNP. (H) K-M curves for the high CD44 expression group and low CD44 expression group. (I) Bar gragh of AUC shows the ability of CD44 expression levels predicting the effectiveness of various tumor immunotherapies

Full size image

Discussion

First, through clustering analysis of 375 TCGA gastric cancer samples, we distinguished two gastric cancer subtypes (C1 and C2). The C2 subtype exhibits higher malignancy, poorer prognosis, and significant resistance to chemotherapy and immunotherapy. Just as we introduced in the Background, these malignant manifestations of the C2 subtype are inseparable from the significant upregulation of ECM-related genes in the C2 subtype, particularly collagen genes COL1A1, COL1A2, and COL3A1, which are implicated in ECM restructuring and modulation of the TME [41, 42]. However, in addition to this, there are still some results worthy of further discussion.

In clinical tumor treatment, the expression level of immune checkpoints is a critical indicator for determining whether a cancer patient is suitable for ICIs. In this study, we observed that certain immune checkpoints were significantly upregulated in the C2 subtype (poor-prognosis gastric cancer subtype), including HAVCR2 (TIM-3) and PDCD1LG2 (PD-L2) (Fig. 1H). However, we found that the TIDE score of the C2 subtype was significantly higher than that of the C1 subtype, indicating a lower immunotherapy response rate in the C2 subtype (Fig. 1I). There are two main reasons for this: First, the C2 subtype exhibited significant enrichment of M2 macrophages and Treg cells (Fig. 1G), which hindered T-cell-mediated tumor cell killing through various mechanisms, leading to tumor immune evasion. For instance, M2 macrophages can limit T-cell function by secreting immunosuppressive cytokines and promoting T-cell exhaustion [43, 44], as well as by forming sustained interactions with T cells that hinder their migration toward tumor cells [45, 46]. Treg cells can suppress T-cell activity through various mechanisms, including inhibition of co-stimulatory signals, depletion of interleukin-2 (IL-2), secretion of suppressive cytokines, and metabolic regulation [47,48,49]. Second, the C2 subtype exhibited significantly higher levels of fibrosis (Fig. 2B). As mentioned in the Background, extensive fibrosis can lead to tumor immune evasion by blocking CD8 + T-cell infiltration, thereby rendering immunotherapy ineffective. This suggests that for gastric cancers characterized by extensive fibrosis, the effectiveness of immunotherapy is limited, providing new insights for clinical treatment. This also reaffirms that it is limited to rely solely on immune checkpoint expression levels to predict the suitability of ICIs for specific patient populations. This is because the efficacy of ICIs is influenced by various factors, including tumor genomics, host genetic factors, TME characteristics, and gut microbiota composition [50]. Therefore, when assessing the suitability of cancer patients for immunotherapy, it is necessary to consider other biomarkers and the complex characteristics of the TME in addition to immune checkpoint expression, to comprehensively evaluate patients’ immune status and therapeutic potential [51]. We believe that the tumor-associated fibrosis level examined in this study has significant potential as an indicator for evaluating the response rate to ICIs in gastric cancer and as a therapeutic target to enhance the efficacy of ICIs treatment in gastric cancer, which is one of the critical implications of our work.

Single-cell transcriptome analysis delved deeper into the expression of ECM-related genes in different cell types. We found that fibroblasts, chondrocytes, smooth muscle cells, and tissue stem cells are the cells primarily expressing ECM-related genes, with a significantly higher proportion in tumor tissues than in normal tissues. CAFs have been thoroughly introduced in the Background. Chondrocytes may influence the phenotype and behavior of tumor cells by secreting specific ECM components, such as chondroitin and aggrecan. These molecules can regulate cell signaling pathways, including the Wnt/β-catenin and Hippo pathways, thereby affecting tumor cell proliferation, migration and invasion [52]. Smooth muscle cells can secrete smooth muscle actin (SMA) and other ECM proteins, which contribute to forming the structural framework of the TME [53]. Tissue stem cells may promote ECM deposition and the formation of supportive structures for tumors by differentiating into fibroblasts or other mesenchymal cells. Additionally, factors secreted by tissue stem cells, such as hepatocyte growth factor (HGF) and insulin-like growth factor (IGF), may directly promote the survival and proliferation of tumor cells [54]. In normal, superficial, and deep tissues, fibroblasts, chondrocytes, and smooth muscle cells had the highest AUC scores, further supporting their dominant role in ECM-related gene expression.

CellChat analysis underscored that the Collagen-CD44 signaling axis plays an important role in communication between these cells, particularly in deeper gastric cancer tissues. This may result from increased hypoxia in the deeper TME, where hypoxia-inducible factors (HIF) facilitate fibroblast activation and proliferation [55]. We also observed that the Collagen-CD44 signaling axis primarily mediates interactions between fibroblasts and various immune cells. This is because the interaction between CD44 and collagen fibers serves as a key pathway for the directed migration of immune cells. However, ECM remodeling, including the reorganization of collagen fiber, is observed in gastric cancer and many other types of tumors [56,57,58]. This reorganization inhibits cytotoxic immune cells, such as CD8 + T cells, from approaching the tumor parenchyma, instead redirecting them into the tumor stroma, thereby enabling tumor immune escape [59, 60]. Furthermore, entering the stroma brings immune cells closer to fibroblasts. In this context, the Collagen-CD44 signaling axis recruits tumor-associated macrophages, which secrete inflammatory factors like IL-4 and TGF-β that further stimulate fibroblasts to produce significant amounts of collagen [61, 62]. This feedback loop ultimately leads to tumor desmoplasia, hypoxia, immune escape, and other detrimental biological processes. Additionally, the interaction between CD44 and collagen may bolster tumor cell resilience and invasive capabilities by activating downstream signaling pathways such as the PI3K/Akt and MAPK pathways [63, 64].

Through pseudotime analysis, we identified three key nodes and seven cell clusters, encapsulating the spectrum of cell differentiation stages. Fibroblasts and deep group cells are primarily distributed at the end of the pseudotime axis, while tissue stem cells and superficial group cells are mainly distributed at the beginning of the pseudotime axis. This distribution pattern suggests that fibroblasts may play a more critical role in the late stages of tumors, while tissue stem cells may be more active in the early stages. As the pseudotime axis progresses, we observed a series of changes in gene expression. For example, APOE, GPC3, and CCL2 genes were significantly upregulated at the end of the pseudotime axis, while ADIRF, MYL9, and MT1A genes were significantly downregulated. Studies have found that APOE can regulate K1 cell cytokine expression through the PI3K/Akt/NF-κB pathway, promoting the polarization of TAMs from the M0 type to the M2 type [65]. APOE + macrophages promote tumor EMT through IGF1-IGF1R interactions, reshaping the TME [66]. Glypican-3 (GPC3) is upregulated in the CAF subgroup of advanced GC and is associated with poor prognosis in GC patients [67]. GPC3 binds to molecules such as Wnt signaling proteins and growth factors [68], activating β-catenin and thereby increasing the expression of ECM components such as collagen, fibronectin, and laminin [69]. CCL2 interacts with CCR2 to increase the expression of matrix metalloproteinase (MMP)-9, which cleaves various ECM proteins to regulate ECM remodeling [70, 71]. THBS4 is an extracellular calcium-binding glycoprotein that is overexpressed in gastrointestinal solid tumors, interacting with the cell surface or components of the ECM, mediating tissue remodeling, and participating in wound healing. Elevated THBS4 protein levels have been associated with poor clinicopathological features in gastric cancer [72]. Tissue inhibitor of metalloproteinase-1 (TIMP-1) is an important regulator of ECM turnover, and its overexpression is mediated by the selective overactivity of the pro-fibrotic TGF-β1/SMAD3 pathway [73]. The expression of the aforementioned upregulated genes positively correlates with the distribution of fibroblasts along the pseudotime axis. Therefore, these genes may be key contributors to gastric cancer-associated fibrosis.

Finally, we conducted a comprehensive exploration of the expression and role of CD44 in gastric cancer. In addition to its interaction with collagen, CD44 can also contribute to tumor immune suppression and promote tumor growth through various other mechanisms. Studies have shown that CD44 + gastric cancer patients have a worse prognosis [74], which is consistent with our findings. In gastric cancer, patients with high CD44 expression exhibit significantly reduced infiltration of CD4 + T cells, B cells, and dendritic cells in the TME [75], while its expression level is associated with an increased number of M2 macrophages [76, 77]. In clear cell renal cell carcinoma, the expression of CD44 positively correlates with the density of tumor-associated macrophages, and inhibiting CD44 can reduce the infiltration of Treg cells in the tumor, thereby improving the efficacy of tumor immunotherapy [78, 79]. In lung adenocarcinoma, CD44 expression is positively correlated with PD-L1 expression, suggesting that in addition to regulating the abundance of various immune cells, CD44 may also promote immune escape of tumor cells by regulating PD-L1 expression [80]. Therefore, CD44 plays a multifaceted role in tumor immune suppression in gastric cancer. It promotes immune suppression by affecting the infiltration of various immune cells, such as T cells, Treg cells, tumor-associated macrophages, and regulating the expression of immune checkpoints. In addition to its role in tumor immunity, CD44 is also a classic marker of cancer stem cells. It promotes malignant phenotypes of tumor cells, such as proliferation, migration, invasion, resistance, and metastasis, through interactions with various signaling molecules. For example, overexpression of CD44 has been shown to promote cell proliferation and migration in gastric cancer cells, and its downregulation significantly inhibits tumor growth [81]. In HER2 + gastric cancer, the binding of hyaluronic acid (HA) and CD44 leads to resistance of gastric cancer cells to trastuzumab [82]. Blocking the CD44s/STAT3 crosstalk can inhibit peritoneal metastasis of gastric cancer [83]. In summary, CD44 promotes gastric cancer progression through various mechanisms. Therefore, therapeutic strategies targeting CD44 may provide new directions for the treatment of gastric cancer.

However, our research acknowledges inherent limitations. Primarily, the single-cell gene expression analysis was confined to analyzing expression within a single group only. Although the results were validated in multiple datasets, we cannot preclude the potential for error. The constrained sample size of individual cells in the pseudotime analysis may have led to biased results. Additionally, the limited scope of the immunotherapy cohort, coupled with the modest AUC values across datasets, suggests that the correlation between CD44 expression and the efficacy of immunotherapy merits further exploration and substantiation. Finally, more advanced algorithms, such as Metaheuristic Optimization Algorithms, can be used in certain analytical methods [84]. For example, the.

PPI used to select core ECM-related genes can be optimized using Metaheuristic Optimization Algorithms to uncover key protein nodes and modules, predict new protein-protein interactions, and assist in gene selection.

Conclusion

Our study identified a new subtype of gastric cancer, revealing that fibrosis is a critical mechanism driving immune suppression in gastric cancer and emphasizing the central role of the Collagen-CD44 signaling axis. The Collagen-CD44 signaling axis has the potential to serve as a novel therapeutic target for gastric cancer by enhancing immune cell-mediated tumor suppression. By combining it with ICIs, it may improve the efficacy of immunotherapy for gastric cancer and offer new hope for treatment.

Data availability

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding authors.

Abbreviations

TME:

Tumor microenvironment

GC:

Gastric cancer

TCGA:

The Cancer Genome Atlas

scRNA seq:

Single-cell RNA sequencing

GEO:

Gene Expression Omnibus

CAFs:

Cancer-associated fibroblasts

ECM:

Extracellular matrix

Tregs:

Regulatory T cells

ICIs:

Immune checkpoint inhibitors

MMPs:

Matrix metalloproteinases

OR:

Odds ratio

GDSC:

Genomics of Drug Sensitivity in Cancer

IC50:

The half-maximal inhibitory concentration

TIDE:

Tumor Immune Dysfunction and Exclusion

PPI:

Protein-protein interaction

UMI:

Unique molecular identifier

AUC:

Area under the receiver operating characteristic curve

ROC:

Receiver operating characteristic

PCA:

Principal component analysis

EMT:

Epithelial-mesenchymal transition

GWAS:

Genome-Wide Association Studies

HPA:

The Human Protein Atlas

HR:

Hazard ratio

GMP:

Granulocyte-monocyte progenitors

CMP:

Common-myeloid progenitors

BM & Prog.:

Bone marrow cells and progenitors

EMT:

Epithelial-mesenchymal transition

KIRC:

Kidney renal clear cell carcinoma

References

  1. Joshi SS, Badgwell BD. Current treatment and recent progress in gastric cancer. CA Cancer J Clin. 2021;71:264–79.

    Article  PubMed  PubMed Central  Google Scholar 

  2. Janjigian YY, Werner D, Pauligk C, Steinmetz K, Kelsen DP, Jäger E, Altmannsberger HM, Robinson E, Tafe LJ, Tang LH, et al. Prognosis of metastatic gastric and gastroesophageal junction cancer by HER2 status: a European and USA International collaborative analysis. Ann Oncol. 2012;23:2656–62.

    Article  CAS  PubMed  Google Scholar 

  3. Zhao D, Klempner SJ, Chao J. Progress and challenges in HER2-positive gastroesophageal adenocarcinoma. J Hematol Oncol. 2019;12:50.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Fuchs CS, Doi T, Jang RW, Muro K, Satoh T, Machado M, Sun W, Jalal SI, Shah MA, Metges JP, et al. Safety and Efficacy of Pembrolizumab Monotherapy in patients with previously treated Advanced gastric and gastroesophageal Junction Cancer: phase 2 clinical KEYNOTE-059 trial. JAMA Oncol. 2018;4:e180013.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Kim ST, Cristescu R, Bass AJ, Kim KM, Odegaard JI, Kim K, Liu XQ, Sher X, Jung H, Lee M, et al. Comprehensive molecular characterization of clinical responses to PD-1 inhibition in metastatic gastric cancer. Nat Med. 2018;24:1449–58.

    Article  CAS  PubMed  Google Scholar 

  6. Karakasheva TA, Lin EW, Tang Q, Qiao E, Waldron TJ, Soni M, Klein-Szanto AJ, Sahu V, Basu D, Ohashi S, et al. IL-6 mediates cross-talk between Tumor cells and activated fibroblasts in the Tumor Microenvironment. Cancer Res. 2018;78:4957–70.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Biffi G, Tuveson DA. Diversity and Biology of Cancer-Associated fibroblasts. Physiol Rev. 2021;101:147–76.

    Article  CAS  PubMed  Google Scholar 

  8. Chen Y, McAndrews KM, Kalluri R. Clinical and therapeutic relevance of cancer-associated fibroblasts. Nat Rev Clin Oncol. 2021;18:792–804.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Chen X, Song E. Turning foes to friends: targeting cancer-associated fibroblasts. Nat Rev Drug Discov. 2019;18:99–115.

    Article  CAS  PubMed  Google Scholar 

  10. Fuyuhiro Y, Yashiro M, Noda S, Kashiwagi S, Matsuoka J, Doi Y, Kato Y, Muguruma K, Sawada T, Hirakawa K. Myofibroblasts are associated with the progression of scirrhous gastric carcinoma. Exp Ther Med. 2010;1:547–51.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Xu G, Zhang B, Ye J, Cao S, Shi J, Zhao Y, Wang Y, Sang J, Yao Y, Guan W, et al. Exosomal miRNA-139 in cancer-associated fibroblasts inhibits gastric cancer progression by repressing MMP11 expression. Int J Biol Sci. 2019;15:2320–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Liu D, Shi K, Fu M, Chen F. Melatonin indirectly decreases gastric cancer cell proliferation and invasion via effects on cancer-associated fibroblasts. Life Sci. 2021;277:119497.

    Article  CAS  PubMed  Google Scholar 

  13. Najafi M, Farhood B, Mortezaee K. Extracellular matrix (ECM) stiffness and degradation as cancer drivers. J Cell Biochem. 2019;120:2782–90.

    Article  CAS  PubMed  Google Scholar 

  14. Xiao Z, Tan Y, Cai Y, Huang J, Wang X, Li B, Lin L, Wang Y, Shuai X, Zhu K. Nanodrug removes physical barrier to promote T-cell infiltration for enhanced cancer immunotherapy. J Control Release. 2023;356:360–72.

    Article  CAS  PubMed  Google Scholar 

  15. Kieffer Y, Hocine HR, Gentric G, Pelon F, Bernard C, Bourachot B, Lameiras S, Albergante L, Bonneau C, Guyard A, et al. Single-cell analysis reveals fibroblast clusters linked to Immunotherapy Resistance in Cancer. Cancer Discov. 2020;10:1330–51.

    Article  CAS  PubMed  Google Scholar 

  16. Theocharis AD, Skandalis SS, Gialeli C, Karamanos NK. Extracellular matrix structure. Adv Drug Deliv Rev. 2016;97:4–27.

    Article  CAS  PubMed  Google Scholar 

  17. Aguilera KY, Huang H, Du W, Hagopian MM, Wang Z, Hinz S, Hwang TH, Wang H, Fleming JB, Castrillon DH, et al. Inhibition of Discoidin Domain Receptor 1 reduces collagen-mediated tumorigenicity in pancreatic ductal adenocarcinoma. Mol Cancer Ther. 2017;16:2473–85.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Su H, Yang F, Fu R, Trinh B, Sun N, Liu J, Kumar A, Baglieri J, Siruno J, Le M, et al. Collagenolysis-dependent DDR1 signalling dictates pancreatic cancer outcome. Nature. 2022;610:366–72.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Naci D, Vuori K, Aoudjit F. Alpha2beta1 integrin in cancer development and chemoresistance. Semin Cancer Biol. 2015;35:145–53.

    Article  CAS  PubMed  Google Scholar 

  20. Navab R, Strumpf D, To C, Pasko E, Kim KS, Park CJ, Hai J, Liu J, Jonkman J, Barczyk M, et al. Integrin α11β1 regulates cancer stromal stiffness and promotes tumorigenicity and metastasis in non-small cell lung cancer. Oncogene. 2016;35:1899–908.

    Article  CAS  PubMed  Google Scholar 

  21. Ishimoto T, Miyake K, Nandi T, Yashiro M, Onishi N, Huang KK, Lin SJ, Kalpana R, Tay ST, Suzuki Y, et al. Activation of transforming growth factor Beta 1 signaling in gastric Cancer-associated fibroblasts increases their motility, via expression of rhomboid 5 homolog 2, and ability to Induce Invasiveness of Gastric Cancer cells. Gastroenterology. 2017;153:191–e204116.

    Article  CAS  PubMed  Google Scholar 

  22. Misra S, Heldin P, Hascall VC, Karamanos NK, Skandalis SS, Markwald RR, Ghatak S. Hyaluronan-CD44 interactions as potential targets for cancer therapy. Febs j. 2011;278:1429–43.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Skandalis SS, Karalis T, Heldin P. Intracellular hyaluronan: importance for cellular functions. Semin Cancer Biol. 2020;62:20–30.

    Article  CAS  PubMed  Google Scholar 

  24. Huang J, Zhang L, Wan D, Zhou L, Zheng S, Lin S, Qiao Y. Extracellular matrix and its therapeutic potential for cancer treatment. Signal Transduct Target Ther. 2021;6:153.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Guan WL, He Y, Xu RH. Gastric cancer treatment: recent progress and future perspectives. J Hematol Oncol. 2023;16:57.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Sleeboom JJF, van Tienderen GS, Schenke-Layland K, van der Laan LJW, Khalil AA, Verstegen MMA. The extracellular matrix as hallmark of cancer and metastasis: from biomechanics to therapeutic targets. Sci Transl Med. 2024;16:eadg3840.

    Article  CAS  PubMed  Google Scholar 

  27. The Cancer Genome Atlas(TCGA.) [https://portal.gdc.cancer.gov]

  28. Genotype-Tissue Expression Project. (GTEx) [https://commonfund.nih.gov/GTEx]

  29. Gene Expression Omnibus(GEO.) [https://www.ncbi.nlm.nih.gov/geo/]

  30. European Nucleotide Archive(ENA.) [https://www.ebi.ac.uk/ena/browser/home]

  31. Database of Genotypes and Phenotypes(dbGaP.) [https://www.ncbi.nlm.nih.gov/gap/]

  32. Nathanson T, Ahuja A, Rubinsteyn A, Aksoy BA, Hellmann MD, Miao D, Van Allen E, Merghoub T, Wolchok JD, Snyder A, Hammerbacher J. Somatic mutations and Neoepitope Homology in Melanomas treated with CTLA-4 blockade. Cancer Immunol Res. 2017;5:84–91.

    Article  CAS  PubMed  Google Scholar 

  33. Braun DA, Hou Y, Bakouny Z, Ficial M, Sant’ Angelo M, Forman J, Ross-Macdonald P, Berger AC, Jegede OA, Elagina L, et al. Interplay of somatic alterations and immune infiltration modulates response to PD-1 blockade in advanced clear cell renal cell carcinoma. Nat Med. 2020;26:909–18.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. ArrayExpress. [https://www.ebi.ac.uk/biostudies/arrayexpress]

  35. BioProject. [https://www.ncbi.nlm.nih.gov/bioproject/]

  36. NHGRI-EBI GWAS Catalog. [https://www.ebi.ac.uk/gwas]

  37. EMBL-EBI eQTL Catalogue. [https://www.ebi.ac.uk/eqtl/]

  38. Genomics of Drug Sensitibity in Cancer(GDSC.) [https://www.cancerrxgene.org/]

  39. STRING. [https://string-db.org/].

  40. The Human Protein Atlas(HPA.) [https://www.proteinatlas.org/]

  41. Orimo A, Gupta PB, Sgroi DC, Arenzana-Seisdedos F, Delaunay T, Naeem R, Carey VJ, Richardson AL, Weinberg RA. Stromal fibroblasts present in invasive human breast carcinomas promote tumor growth and angiogenesis through elevated SDF-1/CXCL12 secretion. Cell. 2005;121:335–48.

    Article  CAS  PubMed  Google Scholar 

  42. Kuperwasser C, Chavarria T, Wu M, Magrane G, Gray JW, Carey L, Richardson A, Weinberg RA. Reconstruction of functionally normal and malignant human breast tissues in mice. Proc Natl Acad Sci U S A. 2004;101:4966–71.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Han Q, Shi H, Liu F. CD163(+) M2-type tumor-associated macrophage support the suppression of tumor-infiltrating T cells in osteosarcoma. Int Immunopharmacol. 2016;34:101–6.

    Article  PubMed  Google Scholar 

  44. Wu H, Jiang N, Li J, Jin Q, Jin J, Guo J, Wei X, Wang X, Yao L, Meng D, Zhi X. Tumor cell SPTBN1 inhibits M2 polarization of macrophages by suppressing CXCL1 expression. J Cell Physiol. 2024;239:97–111.

    Article  CAS  PubMed  Google Scholar 

  45. Long KB, Collier AI, Beatty GL. Macrophages: key orchestrators of a tumor microenvironment defined by therapeutic resistance. Mol Immunol. 2019;110:3–12.

    Article  CAS  PubMed  Google Scholar 

  46. Peranzoni E, Lemoine J, Vimeux L, Feuillet V, Barrin S, Kantari-Mimoun C, Bercovici N, Guérin M, Biton J, Ouakrim H, et al. Macrophages impede CD8 T cells from reaching tumor cells and limit the efficacy of anti-PD-1 treatment. Proc Natl Acad Sci U S A. 2018;115:E4041–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Paluskievicz CM, Cao X, Abdi R, Zheng P, Liu Y, Bromberg JS. T Regulatory Cells and priming the suppressive Tumor Microenvironment. Front Immunol. 2019;10:2453.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Shitara K, Nishikawa H. Regulatory T cells: a potential target in cancer immunotherapy. Ann N Y Acad Sci. 2018;1417:104–15.

    Article  CAS  PubMed  Google Scholar 

  49. Ohue Y, Nishikawa H. Regulatory T (Treg) cells in cancer: can Treg cells be a new therapeutic target? Cancer Sci. 2019;110:2080–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Havel JJ, Chowell D, Chan TA. The evolving landscape of biomarkers for checkpoint inhibitor immunotherapy. Nat Rev Cancer. 2019;19:133–50.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Borcherding N, Kolb R, Gullicksrud J, Vikas P, Zhu Y, Zhang W. Keeping tumors in check: a mechanistic review of clinical response and resistance to Immune Checkpoint Blockade in Cancer. J Mol Biol. 2018;430:2014–29.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Levental KR, Yu H, Kass L, Lakins JN, Egeblad M, Erler JT, Fong SF, Csiszar K, Giaccia A, Weninger W, et al. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell. 2009;139:891–906.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Chatterjee P, Martin KA. A Concept of Athero-Oncology: Tumor-Like smooth muscle cells drive atherosclerosis. Circulation. 2024;149:1899–902.

    Article  PubMed  Google Scholar 

  54. Olive KP, Jacobetz MA, Davidson CJ, Gopinathan A, McIntyre D, Honess D, Madhu B, Goldgraben MA, Caldwell ME, Allard D, et al. Inhibition of hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer. Science. 2009;324:1457–61.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Cowman SJ, Koh MY. Revisiting the HIF switch in the tumor and its immune microenvironment. Trends Cancer. 2022;8:28–42.

    Article  CAS  PubMed  Google Scholar 

  56. Lee CJ, Jang TY, Jeon SE, Yun HJ, Cho YH, Lim DY, Nam JS. The dysadherin/MMP9 axis modifies the extracellular matrix to accelerate colorectal cancer progression. Nat Commun. 2024;15:10422.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Waters JA, Robinson M, Lujano-Olazaba O, Lucht C, Gilbert SF, House CD. Omental Preadipocytes stimulate Matrix Remodeling and IGF Signaling to Support Ovarian Cancer Metastasis. Cancer Res. 2024;84:2073–89.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  58. Liu T, Huang C, Sun L, Chen Z, Ge Y, Ji W, Chen S, Zhao Y, Wang M, Wang D, Zhu W. FAP(+) gastric cancer mesenchymal stromal cells via paracrining INHBA and remodeling ECM promote tumor progression. Int Immunopharmacol. 2024;144:113697.

    Article  PubMed  Google Scholar 

  59. Fuller AM, Pruitt HC, Liu Y, Irizarry-Negron VM, Pan H, Song H, DeVine A, Katti RS, Devalaraja S, Ciotti GE et al. Oncogene-induced matrix reorganization controls CD8 + T cell function in the soft-tissue sarcoma microenvironment. J Clin Invest. 2024;134(11):e167826.

  60. Yuan Z, Li Y, Zhang S, Wang X, Dou H, Yu X, Zhang Z, Yang S, Xiao M. Extracellular matrix remodeling in tumor progression and immune escape: from mechanisms to treatments. Mol Cancer. 2023;22:48.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  61. Chen T, Cao Q, Wang Y, Harris DCH. M2 macrophages in kidney disease: biology, therapies, and perspectives. Kidney Int. 2019;95:760–73.

    Article  CAS  PubMed  Google Scholar 

  62. Nguyen HN, Jeong Y, Kim Y, Kamiya M, Kim Y, Athar H, Castaldi PJ, Hersh CP, Menon JA, Wong J, et al. Leukemia inhibitory factor (LIF) receptor amplifies pathogenic activation of fibroblasts in lung fibrosis. Proc Natl Acad Sci U S A. 2024;121:e2401899121.

    Article  CAS  PubMed  Google Scholar 

  63. El Azreq MA, Naci D, Aoudjit F. Collagen/β1 integrin signaling up-regulates the ABCC1/MRP-1 transporter in an ERK/MAPK-dependent manner. Mol Biol Cell. 2012;23:3473–84.

    Article  PubMed  PubMed Central  Google Scholar 

  64. Wu X, Cai J, Zuo Z, Li J. Collagen facilitates the colorectal cancer stemness and metastasis through an integrin/PI3K/AKT/Snail signaling pathway. Biomed Pharmacother. 2019;114:108708.

    Article  CAS  PubMed  Google Scholar 

  65. Huo R, Zhao R, Li Z, Li M, Bin Y, Wang D, Xue G, Wu J, Lin X. APOE expression in papillary thyroid carcinoma: influencing tumor progression and macrophage polarization. Immunobiology. 2024;229:152821.

    Article  CAS  PubMed  Google Scholar 

  66. Liu H, Gao J, Feng M, Cheng J, Tang Y, Cao Q, Zhao Z, Meng Z, Zhang J, Zhang G, et al. Integrative molecular and spatial analysis reveals evolutionary dynamics and tumor-immune interplay of in situ and invasive acral melanoma. Cancer Cell. 2024;42:1067–e10851011.

    Article  CAS  PubMed  Google Scholar 

  67. Li D, Wang Y, Shi C, Fu S, Sun YF, Li C. Targeting GPC3(high) cancer-associated fibroblasts sensitizing the PD-1 blockage therapy in gastric cancer. Ann Med. 2023;55:2189295.

    Article  PubMed  PubMed Central  Google Scholar 

  68. Zhou F, Shang W, Yu X, Tian J. Glypican-3: a promising biomarker for hepatocellular carcinoma diagnosis and treatment. Med Res Rev. 2018;38:741–67.

    Article  CAS  PubMed  Google Scholar 

  69. Li Y, Hong J, Jung BK, Oh E, Yun CO. Oncolytic ad co-expressing decorin and wnt decoy receptor overcomes chemoresistance of desmoplastic tumor through degradation of ECM and inhibition of EMT. Cancer Lett. 2019;459:15–29.

    Article  CAS  PubMed  Google Scholar 

  70. Tang CH, Tsai CC. CCL2 increases MMP-9 expression and cell motility in human chondrosarcoma cells via the Ras/Raf/MEK/ERK/NF-κB signaling pathway. Biochem Pharmacol. 2012;83:335–44.

    Article  CAS  PubMed  Google Scholar 

  71. Huang H. Matrix Metalloproteinase-9 (MMP-9) as a Cancer Biomarker and MMP-9 biosensors: recent advances. Sens (Basel). 2018;18(10):3249.

  72. Katzendorn O, Peters I, Dubrowinskaja N, Moog JM, Reese C, Tezval H, Faraj Tabrizi P, Hennenlotter J, Lafos M, Kuczyk MA, Serth J. DNA methylation in INA, NHLH2, and THBS4 is associated with metastatic disease in renal cell carcinoma. Cancers (Basel). 2021;14(1):39.

  73. Duch P, Díaz-Valdivia N, Ikemori R, Gabasa M, Radisky ES, Arshakyan M, Gea-Sorlí S, Mateu-Bosch A, Bragado P, Carrasco JL, et al. Aberrant TIMP-1 overexpression in tumor-associated fibroblasts drives tumor progression through CD63 in lung adenocarcinoma. Matrix Biol. 2022;111:207–25.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  74. Hou W, Kong L, Hou Z, Ji H. CD44 is a prognostic biomarker and correlated with immune infiltrates in gastric cancer. BMC Med Genomics. 2022;15:225.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  75. Wang P, Zhu Y, Jia X, Ying X, Sun L, Ruan S. Clinical prognostic value of OSGIN2 in gastric cancer and its proliferative effect in vitro. Sci Rep. 2023;13:5775.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  76. Bonnin E, Rodrigo Riestra M, Marziali F, Mena Osuna R, Denizeau J, Maurin M, Saez JJ, Jouve M, Bonté PE, Richer W, et al. CD74 supports accumulation and function of regulatory T cells in tumors. Nat Commun. 2024;15:3749.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  77. Harada K, Dong X, Estrella JS, Correa AM, Xu Y, Hofstetter WL, Sudo K, Onodera H, Suzuki K, Suzuki A, et al. Tumor-associated macrophage infiltration is highly associated with PD-L1 expression in gastric adenocarcinoma. Gastric Cancer. 2018;21:31–40.

    Article  CAS  PubMed  Google Scholar 

  78. Ma C, Komohara Y, Ohnishi K, Shimoji T, Kuwahara N, Sakumura Y, Matsuishi K, Fujiwara Y, Motoshima T, Takahashi W, et al. Infiltration of tumor-associated macrophages is involved in CD44 expression in clear cell renal cell carcinoma. Cancer Sci. 2016;107:700–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  79. Ma J, Wu R, Chen Z, Zhang Y, Zhai W, Zhu R, Zheng J. CD44 is a prognostic biomarker correlated with Immune infiltrates and metastasis in Clear Cell Renal Cell Carcinoma. Anticancer Res. 2023;43:3493–506.

    Article  CAS  PubMed  Google Scholar 

  80. Zhang C, Wang H, Wang X, Zhao C, Wang H. CD44, a marker of cancer stem cells, is positively correlated with PD-L1 expression and immune cells infiltration in lung adenocarcinoma. Cancer Cell Int. 2020;20:583.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  81. Chen S, Zhang G, Liu Y, Yang C, He Y, Guo Q, Du Y, Gao F. Anchoring of hyaluronan glycocalyx to CD44 reduces sensitivity of HER2-positive gastric cancer cells to trastuzumab. Febs j. 2024;291:1719–31.

    Article  CAS  PubMed  Google Scholar 

  82. Jin Y, Wang C, Zhang B, Sun Y, Ji J, Cai Q, Jiang J, Zhang Z, Zhao L, Yu B, Zhang J. Blocking EGR1/TGF-β1 and CD44s/STAT3 crosstalk inhibits peritoneal metastasis of gastric Cancer. Int J Biol Sci. 2024;20:1314–31.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  83. Deng H, Gao J, Cao B, Qiu Z, Li T, Zhao R, Li H, Wei B. LncRNA CCAT2 promotes malignant progression of metastatic gastric cancer through regulating CD44 alternative splicing. Cell Oncol (Dordr). 2023;46:1675–90.

    Article  CAS  PubMed  Google Scholar 

  84. Abualigah L. Metaheuristic Optimization Algorithms: optimizers, analysis, and applications. Elsevier; 2024.

Download references

Acknowledgements

Not applicable.

Funding

This work was supported by the grant from Nanjing Health Science and Technology Development Special fund project (YKK22209).

Author information

Authors and Affiliations

  1. The Second School of Clinical Medicine, Nanjing Medical University, Nanjing, 211166, China

    Yingqi Yang, Haohan Sun, Hongkai Yu, Luyao Wang & Chang Gao

  2. The Fourth School of Clinical Medicine, Nanjing Medical University, Nanjing, 211166, China

    Haokun Mei

  3. Department of Gastroenterology, Sir Run Run Hospital, Nanjing Medical University, Nanjing, 211100, China

    Xiaomeng Jiang

  4. School of Nursing, Nanjing Medical University, Nanjing, 211166, China

    Minghui Ji

Authors
  1. Yingqi Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Haohan Sun
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Hongkai Yu
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Luyao Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Chang Gao
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Haokun Mei
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Xiaomeng Jiang
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Minghui Ji
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Yingqi Yang: Formal analysis, Investigation, Data curation, Methodology, Supervision, Writing - original draft. Haohan Sun: Writing - review & editing. Hongkai Yu: Writing - review & editing. Luyao Wang: Writing - review & editing. Chang Gao:Writing - review & editing. Haokun Mei: Visualization. Xiaomeng Jiang: Administration. Minghui Ji: Conceptualization.

Corresponding authors

Correspondence to Xiaomeng Jiang or Minghui Ji.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Yang, Y., Sun, H., Yu, H. et al. Tumor-associated-fibrosis and active collagen-CD44 axis characterize a poor-prognosis subtype of gastric cancer and contribute to tumor immunosuppression. J Transl Med 23, 123 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06070-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06070-9

Keywords

Journal of Translational Medicine

ISSN: 1479-5876

Contact us