Identification of cancer-associated fibroblast subpopulation and construction of an immunotherapy signature for gastric cancer

Introduction

Gastric cancer (GC) is one of the most common malignant tumors, ranking fifth in incidence and third in mortality worldwide, with more than one million new cases occurring every year (1). Although the incidence of GC has decreased in the past five years, the 5-year overall survival (OS) rate of GC remains unsatisfactory (2). GC is highly invasive and asymptomatic, therefore, most of GC cases are already advanced or even metastatic at the time of diagnosis (1,3). Since patients are prone to develop drug resistance leading to tumor recurrence and metastasis, the therapeutic effect of systemic chemotherapy on advanced GC is not optimistic (4). In recent years, the immune checkpoint blockade (ICB) has become a promising treatment option for many kinds of tumors. However, these treatments are only effective in a minority of GC patients, and most GC patients are primarily resistant to ICB therapy (5). Therefore, there is an urgent need to find new biomarkers to accurately determine which type of GC patients are suitable for ICB therapy.

Clinical response to ICB depends largely on the interaction of tumor microenvironment (TME) and global immunity (6). TME incorporates heterogeneous cell types, including tumor-infiltrating lymphocytes and stromal cells (7). Recent studies have shown that cancer-associated fibroblasts (CAFs) are the main stromal cells in TME and have important influence on the response to immunotherapy (8). CAFs are a group of fibroblasts existing in almost all solid tumors. They are similar to normal fibroblasts (NFs), but compared with NFs, CAFs are larger, with more cytoplasm branches (9). Although fibroblasts have been described decades ago, their role in tumors has not been revealed until now. In non-tumor tissues, NFs are normally quiescent, after tissue damage, fibroblasts are orchestrated to proliferate and increase their synthetic and metabolic activities (10). Activated fibroblasts produce higher levels of extracellular matrix (ECM) components, which are conducive to wound closure and scarring (11). After wound healing, most of the activated fibroblasts undergo apoptosis and ECM returns to tissue homeostasis (12). By contrast, tumors are considered as “wounds that do not heal”, because the wound healing response is not self limiting, and the number of activated fibroblasts does not decrease as in the normal wound healing process (9). As the tumor grows, the number of CAFs increased, which accumulate in the TME. Once CAFs are reprogrammed, they affect most of the hallmark capabilities of cancer cells through paracrine mechanisms or direct contact (13,14). Emerging evidence suggests that CAFs are involved in tumor immune dysfunction and immunotherapy response (15), as CAFs mediate extensive immunosuppression by specifically excluding immune effector cells such as CD8⁺ T cells (not macrophages or CD4⁺ T cells) from tumors (16). Notably, CAFs are considered to be heterogeneous population due to the differences in contributions of different CAFs subpopulations to tumor progression (17). Studies have confirmed that there are three subgroups of CAFs in pancreatic ductal adenocarcinoma, including myofibroblast (myCAF), inflammatory CAF (iCAF), and antigen-presenting CAF (apCAF). Among them, myCAF is an important factor for poor prognosis of patients (18). Zheng et al. (19) constructed a prognosis and immunotherapy prediction risk score model based on CAFs specific genes for colorectal cancer, and the results suggested that patients with high risk scores had poor prognosis and were more prone to be unresponsive to immunotherapy. This suggests that CAFs have potential as predictors of immunotherapy and prognostic stratification; however, the practice in GC is lacking.

Therefore, this study aimed to identify more GC-related CAFs subgroups, explore the impact of these subgroups on GC progression, screen CAFs specific genes with prognostic value to construct a prognostic signature, and predict the response to immunotherapy. Our work will provide reference for the clinical diagnosis, prognosis and treatment of GC. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-996/rc).

Methods

Study objects and data acquisition

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. We downloaded the messenger RNA (mRNA) expression profile data and corresponding clinical information of 375 GC patients from The Cancer Genome Atlas (TCGA, https://tcga-data.nci.nih.gov/tcga/) database, and eliminated the samples with incomplete survival information. A total of 350 GC patients with complete survival information were obtained. In addition, we also downloaded data sets numbered GSE194261, GSE116167, GSE47007 and GSE84437 from Gene Expression Omnibus database (GEO, https://www.ncbi.nlm.nih.gov/geo/). GSE194261 contained 20 fibroblast samples, and GSE116167 contained 24 fibroblast samples. GSE47007 contained 12 diffuse GC samples and 18 intestinal GC samples, while GSE84437 contained 433 GC samples with survival information. The details of the GC samples were shown in https://cdn.amegroups.cn/static/public/tcr-2025-996-1.xlsx.

Abundance calculation and the subgroup identification of CAFs

The microenvironment cell populations (MCP) counter algorithm was used to calculate the abundance of CAFs in GC samples, and classical markers (ACTA2, FAP, PDGFRA, PDGFRb, PDPN, THY1 and COL1A1) were used to quantify CAFs. We summarized 235 CAFs characteristic genes from previous studies (17,20). Based on the expression of these genes, we used unsupervised clustering to randomly classify all CAFs cells by “K-mean” method. We further extracted the gene expression profiles of different CAFs clusters and conducted differential gene analysis with NFs to obtain the specific marker genes (SMGs) of each cluster, and the overlapping genes between each cluster were deducted.

Differential expression gene (DEG) screening

The limma function package of R software (version 3.5.2, the same below) was used to analyze the mRNA expression of GC samples (21), and DEGs were screened by the absolute value of Log₂ fold change (Log₂FC) >1 and false discovery rate (FDR) ≤0.05.

Functional enrichment analysis

For the DEGs obtained, we used the “clusterProfiler” function package in the R software to perform enrichment analysis of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes KEGG pathway (22). P value. adjust <0.05 was considered statistically significant. The pathway activation module of the GSCALite online analysis platform (http://bioinfo.life.hust.edu.cn/web/GSCALite/) was used to analyze the pathways involved in the SMGs used to construct prognostic models (23).

Construction of CAFs-related score (CRS) for prognosis

In this study, we constructed different prognostic CRS evaluation models based on the SMGs of each CAFs cluster. Firstly, univariate Cox regression analysis was used to analyze the SMGs in each cluster, and P<0.05 was used as the threshold to screen genes significantly correlated with prognosis of GC. Then, least absolute shrinkage and selection operator (LASSO) Cox regression analysis was applied to screen the optimal genes related to prognosis by glmnet package of R software. These selected genes were used to calculate the CRS of each sample by the following formula:

CRS=∑i=1nCoefi∗Xi[1]

Coefi was the coefficient of each gene calculated by the LASSO Cox model, and X_i was the expression value of each gene. Then the patients were divided into Low CRS group and High CRS group according to the median of CRS.

Survival analysis

The OS rate of High CRS and Low CRS groups was estimated by R software survival package and survivminer package based on Kaplan Meier method, and the significance of the difference in survival rate between the two groups was tested by log rank test. The R software survivalROC package (24) was used to draw the time-dependent receiver operating characteristic (ROC) curve. In addition, the multivariate Cox regression model was used to analyze whether the CRS could predict the survival of GC patients independently of other factors.

Analysis of tumor microenvironment

The StromalScore, ImmuneScore, and TumorPurity were calculated for each sample using the ESTIMATE function package. The ESTIMATE package calculated the ImmuneScore and StromalScore of samples through RNA-seq data, and then evaluated the purity of tumors. The proportion of immune cell infiltration in tumor samples was calculated by CIBERSORT algorithm (25).

Immune checkpoint expression and tumor mutation spectrum

Several important immune checkpoints include CD47, CTLA-4, LAG3, MAGE-A3, PD-1, and PD-L1, and we calculated their expression in different CRS subgroups using R software. In addition, we read MAF files with read.maf, then plotted and compared mutation profiles of two CRS subgroups using the tmb function in the “maftools” package.

Statistical analysis

The Kaplan-Meier method was applied to estimate the OS rate of each group, and the log-rank test was used to further compare the significance of OS differences. The difference in infiltration of immune cells in different groups was compared by Wilcoxon rank sum test, and P<0.05 was considered statistically significant. All statistical analyses were performed by R software (version 3.5.2).

Results

High abundance of CAFs was associated with poor prognosis of GC

CAFs abundance of GC samples in TCGA queue was calculated by MCP-counter algorithm. The samples were grouped according to the median CAFs score, and we found that the survival rate of the high score group was significantly lower than that of the low score group (Figure 1A, P=0.003). Further analysis of Immune score, Stromal score and tumor purity between the high and low CAFs abundance groups showed that the Immune score and Stromal score in the high CAFs group were significantly higher than those in the low CAFs group (Figure 1B,1C), while the tumor purity was lower (Figure 1D).

Figure 1 Prognostic value of CAFs in GC. (A) High abundance of CAFs was associated with poor prognosis of GC. (B-D) The Immune score, Stromal score and tumor purity between high abundance group and low abundance group of CAFs. (E,F) The CAFs were divided into four clusters by unsupervised clustering based on 235 characteristic genes. CAFs, cancer-associated fibroblasts; GC, gastric cancer.

Identification of CAFs subpopulations

We then explored the heterogeneity of CFAs in GC. The GSE194261 and GSE116167 cohorts were combined after batch effect was removed, and then 235 characteristic genes of CAFs were used for clustering of CAFs (the 235 characteristic genes of CAFs were listed in https://cdn.amegroups.cn/static/public/tcr-2025-996-2.xlsx). The results showed that these cells were randomly grouped into Cluster1, Cluster2, Cluster3, and Cluster4 (Figure 1E,1F). In order to obtain the DEGs of each cluster of CAFs, we extracted the CAFs gene expression matrix of each cluster and then compared it with NFs to obtain DEGs. We found that there were 14 overlap genes among four clusters, after deducted these common genes, the remaining genes were used as the SMGs of each cluster. Finally, we identified 3232 SMGs in Cluster1, 3097 SMGs in Cluster2, 46 SMGs in Cluster3 and 2801 SMGs in Cluster4 (The SMGs of each cluster were listed in https://cdn.amegroups.cn/static/public/tcr-2025-996-2.xlsx).

Characteristics of CAFs subpopulations and their abundance differences across GC types

Although we have revealed four CAFs subpopulations in GC, the molecular characteristics of these subpopulations and their abundance in different types of GC were still unclear. In order to solve this problem more clearly, we analyzed the enrichment pathways and biological functions of SMGs in each cluster. Obviously, based on increased expression of collagen and ECM marker genes (such as COL1A1, COL1A2, COL3A1, COL9A1, COL10A1, COL15A1, HSA 2 and DCN), Cluster1, Cluster2 and Cluster4 were classified as ECM-related CAFs (eCAFs) (Figure 2A). Cluster1 and Cluster2 had higher similarity because they both expressed smooth muscle cell markers (MCAM, TAGLN). In the top five terms of GO analysis (Figure 2B), Cluster1 and Cluster2 were mainly enriched in extracellular matrix and structure organization, urogenital system development, renal system development and kidney development. Cluster4 was also enriched in extracellular matrix and structure organization, urogenital system development, but also involved in the reproductive system and structure development. Cluster3 was mainly enriched in regulation of blood pressure, artery development, negative regulation of osteoblast differentiation and ossification. KEGG enrichment analysis further revealed the intrinsic differences of the four subgroups of SMGs (Figure 2C). We found that, in addition to the ECM-receptor interaction and PI3K-Akt signaling pathways that were commonly enriched in Cluster1 and Cluster2, differentially enriched pathways of these two subgroups also included protein digestion and absorption, focal adhesion and cell cycle, which reflected the inherent difference between Cluster1 and Cluster2. Cluster3 was mainly enriched in TGF-β signaling pathway, aldosterone synthesis and secretion, and proteoglycans in cancer. Cluster4 was mainly enriched in axon guidance, Rap1 signaling pathway, etc. In general, we identified four CAFs subpopulations in GC, and functional enrichment analysis elucidated the intrinsic differences of these four subpopulations. The details of functional enrichment results were listed in https://cdn.amegroups.cn/static/public/tcr-2025-996-3.xlsx and Figure S1.

Figure 2 Analysis of heterogeneous subpopulations of CAFs in GC and their characteristics. (A) Heat map of CAFs signature gene expression in 4 clusters of CAFs. The redder the color, the higher the differential expression fold. (B) The top five Gene Ontology items with significant enrichment of 4 clusters of CAFs. (C) The top five Kyoto Encyclopedia of Genes and Genomes pathways with significant enrichment of 4 clusters of CAFs. (D-G) Expression of the top five SMGs of each CAFs clusters in diffuse and intestinal GC. ns, P>0.05. CAFs, cancer-associated fibroblasts; ECM, extracellular matrix; GC, gastric cancer; SMGs, specific marker genes; TGF-β, transforming growth factor-beta.

In addition, we observed the distribution of four subgroups in intestinal-type and diffuse-type GC, as shown in Figure 2. We selected the top five SMGs with differentially expressed folds in each cluster, and calculated the expression level of these genes in two types of GC from GSE47007 cohort. The results showed that there was no significant difference in the expression of the five SMGs in Cluster1, Cluster2 and Cluster4 between the two types of GC samples (Figure 2D,2E,2G); the expression of KLF5 of Cluster3 in intestinal GC was significantly higher than that in diffuse GC (Figure 2F).

CRS constructed based on Cluster3 SMGs had better prognostic performance

The above data highlighted the heterogeneity among CAFs, suggesting that four CAFs subpopulations were distinct. Whether there were differences in the prognostic performance of these subgroups was of interest to us. We conducted univariate Cox regression analysis on SMGs in each cluster, and then selected 10 genes with lowest P values in each cluster for further screening by LASSO Cox regression analysis (table available at https://cdn.amegroups.cn/static/public/tcr-2025-996-4.xlsx). According to the lambda value, the optimal number of genes in each cluster was determined to be seven, seven, nine and eight respectively (Figure 3A-3D, lambda value was the smallest). The expression levels of these selected genes were weighted with the regression coefficients that calculated by LASSO Cox regression analysis to establish a CRS model for predicting patient prognosis. The CAF-related score based on Cluster3 (CRS3) calculation formula constructed based on the nine SMGs of Cluster3 was as follows:

CRS3=(HSPB2∗0.142268685)+(MN1∗0.103595941)+(PLAC9∗0.041990552)+(NFE2L3∗−0.788027945)+(SIX2∗0.030780208)+(EFNB2∗0.176139800)+(ACVRL1∗0.051193442)+(TIMP1∗0.001719626)+(PTPRO∗−0.107327227)[2]

Figure 3 Evaluation of prognostic value of each cluster of CAFs. (A-D) Screening SMGs with prognostic value for each CAFs cluster by lasso Cox regression analysis. (E,F) The CRS prognostic performance of Cluster1 was tested in GEO and TCGA cohorts respectively. (G,H) The CRS prognostic performance of Cluster2 was tested in GEO and TCGA cohorts respectively. (I,J) The CRS prognostic performance of Cluster3 was tested in GEO and TCGA cohorts respectively. (K,L) The CRS prognostic performance of Cluster4 was tested in GEO and TCGA cohorts respectively. CAFs, cancer-associated fibroblasts; CRS, CAF-related score; GC, gastric cancer; GEO, Gene Expression Omnibus; TCGA, The Cancer Genome Atlas.

The CRS calculation formula of the other three clusters also followed the above pattern. We then calculated CRS values for each sample in the GSE84437 training set and TCGA validation set, and divided patients into High and Low CRS groups based on the median score. Comparing the survival time of the two groups, we found that the CRS3 constructed based on Cluster3 could better distinguish the prognosis of different patients in the two datasets, and GC patients with high CRS scores had worse OS than patients with low CRS scores (Figure 3E-3L). The three clusters related to ECM (Cluster1, Cluster2 and Cluster4) failed to significantly differentiate patient outcomes in the validation set (P>0.05). This indicated that Cluster3 had better prognostic value in GC patients.

To further validate the expression of these genes, we obtained the representative immunohistochemical images from the Human Protein Atlas (HPA) database. As shown in Figure 4A, the expression of HSPB2, MN1, PLAC9, EFNB2, ACVRL1 and TIMP1 was different between GC samples and normal samples. Moreover, the results of GSCALite analysis showed that these nine genes mainly activated the EMT pathway, suggesting that Cluster3 might be a potential subgroup that promoted tumor metastasis by regulating EMT process (Figure S2).

Figure 4 Validation of six gene expression that constructed of CRS3, and immune landscape of different groups of patients based on CRS3 definition. (A) Representative immunohistochemical staining of six candidate genes (HSPB2, PLAC9, EFNB2, MN1, ACVRL1, and TIMP1) in GC tissues and corresponding normal gastric tissues obtained from Human Protein Atlas database (https://www.proteinatlas.org/ENSG00000170276-HSPB2/tissue/stomach#img; https://www.proteinatlas.org/ENSG00000189129-PLAC9/tissue/stomach#img; https://www.proteinatlas.org/ENSG00000125266-EFNB2/tissue/stomach#img; https://www.proteinatlas.org/ENSG00000169184-MN1/tissue/stomach#img; https://www.proteinatlas.org/ENSG00000139567-ACVRL1/tissue/stomach#img; https://www.proteinatlas.org/ENSG00000102265-TIMP1/tissue/stomach#img). (B,C) Differences in immune cell infiltration between high and low CRS3 groups. *, P<0.05; **, P<0.01; ***, P<0.001. CAF, cancer-associated fibroblast; CRS3, CAF-related score based on Cluster3; TME, tumor microenvironment.

Differences in the immune microenvironment of two GC subgroups defined by CRS3

As mentioned above, we reported the excellent prognostic performance of Cluster3, however, the composition of the microenvironment of patients between the high and low CRS3 subgroups was unclear. To better elucidate this question, CIBERSORT was used to calculate differences in infiltration of immune cells between the two groups. As shown in Figure 4B,4C, the relative contents of 12 types of immune infiltrating cells, including Macrophages M0, CD8⁺ T cell, and CD4⁺ T memory cells were significantly different between the two groups with high and low CRS3 scores.

The CRS3 could serve as a valuable marker for immunotherapy response

Next, we sought to evaluate the immunotherapy response in patients in the high and low CRS3 groups. Currently, tumor mutation burden (TMB) was considered a potential biomarker to predict the therapeutic response to immunotherapy (26). A mainstream claim that TMB could be assessed as a biomarker of immunotherapy response sensitivity was that more mutations generated more tumor neoantigens recognized by T lymphocytes. When the PD-1/PDL1 axis was blocked, it might lead to a stronger anti-tumor immune response (27). As shown in Figure 5A,5B, the mutation rates of TTN, TP53 and MUC16 in the low CRS3 group were higher than those in the high CRS3 group. The TMB of the low CRS3 group was also significantly higher than that of the high CRS3 group (Figure 5C), and we found that the prognosis of the high TMB group was significantly better than that of the low TMB group (Figure 5D). In addition, the expression of five checkpoint molecules CTLA-4, LAG3, MAGE-A3, PD-1 and PD-L1 was significantly lower in the High CRS3 group than in the low CRS3 group (Figure 5E-5J). In general, high expression of immune checkpoint molecules and high TMB levels might indicate a better response to immunotherapy of patients in low CRS3 group.

Figure 5 CRS3 for prediction of immunotherapy in GC patients. (A,B) Waterfall plot of somatic mutations in patients with high CRS3 and low CRS3. (C) Differences in TMB between high and low CRS3 patients. (D) Higher TMB levels have better prognosis. (E-J) Differential expression of six immune checkpoints between high and low CRS groups. ns, P>0.05; *, P<0.05; **, P<0.01; ***, P<0.001. CAFs, cancer-associated fibroblasts; CRS3, CAF-related score based on Cluster3; GC, gastric cancer; TMB, tumor mutation burden.

Discussion

Accumulating evidence suggested that CAFs were key factors in the progression of multiple tumors due to their roles in remodeling the ECM, promoting tumor immune escape and tumor metastasis (28). In GC, CAFs were associated with poor prognosis of patients (29). CAFs-derived miR-214 promoted metastasis and invasion by regulating the expression of target gene FGF9 to induce epithelial-mesenchymal transition (EMT) in GC cells (30). Targeted inhibition of CAFs activity could suppress the proliferation, migration and chemotherapy resistance of GC cells induced by CAFs (31). Although some studies have explored CAFs in GC, and constructed a prognosis prediction model of GC based on CAFs signature genes (32), as a heterogeneous cell population, our understanding of different subpopulations of CAFs and their roles in tumors was still limited. Accurate identification of the characteristics of subgroups of CAFs would be helpful to understand the role of CAFs in the development of GC and provide appropriate treatment reference.

Cluster analysis showed that CAFs could be divided into four groups based on 235 characteristic gene expression in GC microenvironment, which was similar to the results of Li et al. (29). They used Single-cell RNA sequencing to identify four CAFs subpopulations in the GC microenvironment, and classified them as myofibroblasts, pericytes, ECM, and iCAFs. However, iCAFs were not observed in our study, which might be due to the heterogeneity of GC samples. This also indicated that there might be potential CAFs subgroups that have not been revealed in the currently reported studies, which were not limited to the above types. In our present study, functional enrichment analysis was applied to describe the intrinsic differences among the four subgroups. Although the DEGs of Cluster1, Cluster2 and Cluster4 were both enriched in GO items such as extracellular matrix and structure organization, the DEGs of Cluster1 were also enriched in protein digestion and absorption, human papillomavirus infection, and cell cycle pathways, the DEGs of Cluster2 were enriched in focal adhesion, arrhythmogenic right ventricular cardiomyopathy, axon guidance pathways, the DEGs of Cluster4 were enriched in Rap1 signaling pathway and regulation of actin cytoskeleton pathway. By contrast, the DEGs of Cluster3 were mainly enriched in TGF-β signaling pathway, aldosterone synthesis and secretion and proteoglycans in cancer, etc. The TGF-β signaling pathway has been found to be dysregulated in many cancers, including gastrointestinal cancer (33). In GC cells, TGF-β was often shown to be upregulated to activate the TGF-β/Smad pathway. TGF-β bound to TGF-β receptors I and II to sensitize phosphorylated Smad 2/3, which further activated Smad 4 to transfer TGF-β/Smad signals into the nucleus to promote EMT (33,34). In addition, activated TGF-β/Smad signaling has also been shown to confer resistance to drugs such as trastuzumab and cisplatin in GC cells (35,36). Therefore, we speculated that the poor prognosis of patients in the high CRS3 group might be related to the activation of TGF-β signaling pathway to promote tumor metastasis and drug resistance.

Among these four subgroups, we found that Cluster3 was significantly associated with the prognosis of GC patients. Based on the nine SMGs of Cluster3, we constructed a GC prognostic model CRS3, which could not only effectively distinguish the prognosis of different GC patients, but also predict the response to immunotherapy. One of the core genes constituting CRS3, HSPB2, is a member of the heat shock protein family and has been shown to be involved in the regulation of tumor cell apoptosis in recent years. Oshita et al. (37) found that HSPB2 inhibited breast cancer cell apoptosis by suppressing the activation of caspase-8. However, HSPB2 also inhibited pancreatic cancer cell proliferation (38). These studies indicate that HSPB2 plays different roles in different cancer tissues, but the role of HSPB2 in GC has not been reported yet and needs to be further explored. NFE2L3, also known as NRF3, was a transcription factor with cancer-promoting effects (39). In GC tissues, knockdown of NFE2L3 expression resulted in the inhibition of N-cadherin, vimentin and Snail expression, suggesting that NFE2L3 might be involved in EMT in GC cells (40). Moreover, knockdown of EFNB2 expression in pancreatic ductal adenocarcinoma not only enhanced p53/p21-mediated G0/G1 arrest to inhibit cancer cell proliferation, but also blocked EMT to reduce cancer cell migration and invasion (41). In addition, high expression of TIMP1 was associated with poor prognosis of renal cell carcinoma, and silencing TIMP1 expression inhibited the progression of renal cell carcinoma through EMT (42). These studies suggested that Cluster3 might be a potential pro-metastatic phenotype (pro-meCAFs). In fact, whether Cluster3 could promote EMT in GC still needs to be further explored in vitro and in vivo.

In this study, we found that the GC samples in the low CRS3 group had a higher proportion of immune effector cell infiltration, including CD8⁺T cells, NK cells resting and CD4⁺T memory cells. Emerging studies have shown that these cells are an important part of anti-tumor immunity and can recognize and destroy tumor cells through multiple pathways (43,44). However, the anti-tumor effect of T lymphocytes is impaired by the high expression of immune checkpoints (45). In the low CRS3 group, the levels of five immune checkpoints (CTLA-4, LAG3, MAGE-A3, PD-1 and PD-L1) were significantly upregulated, which might be conducive to the formation of an immunosuppressive microenvironment. These data suggested that patients with low CRS3 scores, high levels of immune checkpoints and immune effector cell infiltration, might benefit from anti-PD1 immunotherapy (46).

Nevertheless, as with many studies based on public datasets, this study has certain limitations. First, it is a retrospective analysis based entirely on publicly available transcriptomic datasets, which may be subject to inherent biases. Second, the findings were derived exclusively through bioinformatics methods without experimental validation. In particular, the prognostic value of CRS3 and its potential as a predictor for immunotherapy response require further in vivo and mechanistic studies. Therefore, future work should incorporate experimental and clinical validation to confirm the biological significance and clinical utility of the identified CAF subpopulations and marker genes of Cluster3.

Conclusions

In conclusion, we identified four CAFs subsets in the GC microenvironment by unsupervised clustering based on the expression of 235 characteristic genes. Cluster was defined as a pro-meCAFs group with prognostic and immunotherapy predictive value. Several specific marker genes of Cluster3, such as NFE2L3, EFNB2, SIX2, etc., had the potential to become targets for targeting CAFs to inhibit GC progression.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-996/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-996/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-996/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Ferlay J, Soerjomataram I, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2018;68:394-424. [Crossref] [PubMed]
2. Rawla P, Barsouk A. Epidemiology of gastric cancer: global trends, risk factors and prevention. Prz Gastroenterol 2019;14:26-38. [Crossref] [PubMed]
3. Zhao L, Liu Y, Zhang S, et al. Impacts and mechanisms of metabolic reprogramming of tumor microenvironment for immunotherapy in gastric cancer. Cell Death Dis 2022;13:378. [Crossref] [PubMed]
4. Smyth EC, Nilsson M, Grabsch HI, et al. Gastric cancer. Lancet 2020;396:635-48. [Crossref] [PubMed]
5. Fuchs CS, Doi T, Jang RW, 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. [Crossref] [PubMed]
6. Spitzer MH, Carmi Y, Reticker-Flynn NE, et al. Systemic Immunity Is Required for Effective Cancer Immunotherapy. Cell 2017;168:487-502.e15. [Crossref] [PubMed]
7. Jiang Y, Li L, Li Y, et al. Neddylation Regulates Macrophages and Implications for Cancer Therapy. Front Cell Dev Biol 2021;9:681186. [Crossref] [PubMed]
8. Tian L, Long F, Hao Y, et al. A Cancer Associated Fibroblasts-Related Six-Gene Panel for Anti-PD-1 Therapy in Melanoma Driven by Weighted Correlation Network Analysis and Supervised Machine Learning. Front Med (Lausanne) 2022;9:880326. [Crossref] [PubMed]
9. Fozzatti L, Cheng SY. Tumor Cells and Cancer-Associated Fibroblasts: A Synergistic Crosstalk to Promote Thyroid Cancer. Endocrinol Metab (Seoul) 2020;35:673-80. [Crossref] [PubMed]
10. Sliker BH, Campbell PM. Fibroblasts Influence the Efficacy, Resistance, and Future Use of Vaccines and Immunotherapy in Cancer Treatment. Vaccines (Basel) 2021;9:634. [Crossref] [PubMed]
11. Shabestani Monfared G, Ertl P, Rothbauer M. An on-chip wound healing assay fabricated by xurography for evaluation of dermal fibroblast cell migration and wound closure. Sci Rep 2020;10:16192. [Crossref] [PubMed]
12. Distler JHW, Györfi AH, Ramanujam M, et al. Shared and distinct mechanisms of fibrosis. Nat Rev Rheumatol 2019;15:705-30. [Crossref] [PubMed]
13. Bhome R, Goh RW, Bullock MD, et al. Exosomal microRNAs derived from colorectal cancer-associated fibroblasts: role in driving cancer progression. Aging (Albany NY) 2017;9:2666-94. [Crossref] [PubMed]
14. Labernadie A, Kato T, Brugués A, et al. A mechanically active heterotypic E-cadherin/N-cadherin adhesion enables fibroblasts to drive cancer cell invasion. Nat Cell Biol 2017;19:224-37. [Crossref] [PubMed]
15. Kumar V, Donthireddy L, Marvel D, et al. Cancer-Associated Fibroblasts Neutralize the Anti-tumor Effect of CSF1 Receptor Blockade by Inducing PMN-MDSC Infiltration of Tumors. Cancer Cell 2017;32:654-668.e5. [Crossref] [PubMed]
16. Ford K, Hanley CJ, Mellone M, et al. NOX4 Inhibition Potentiates Immunotherapy by Overcoming Cancer-Associated Fibroblast-Mediated CD8 T-cell Exclusion from Tumors. Cancer Res 2020;80:1846-60. [Crossref] [PubMed]
17. Galbo PM Jr, Zang X, Zheng D. Molecular Features of Cancer-associated Fibroblast Subtypes and their Implication on Cancer Pathogenesis, Prognosis, and Immunotherapy Resistance. Clin Cancer Res 2021;27:2636-47. [Crossref] [PubMed]
18. Hu B, Wu C, Mao H, et al. Subpopulations of cancer-associated fibroblasts link the prognosis and metabolic features of pancreatic ductal adenocarcinoma. Ann Transl Med 2022;10:262. [Crossref] [PubMed]
19. Zheng H, Liu H, Ge Y, et al. Integrated single-cell and bulk RNA sequencing analysis identifies a cancer associated fibroblast-related signature for predicting prognosis and therapeutic responses in colorectal cancer. Cancer Cell Int 2021;21:552. [Crossref] [PubMed]
20. Elyada E, Bolisetty M, Laise P, et al. Cross-Species Single-Cell Analysis of Pancreatic Ductal Adenocarcinoma Reveals Antigen-Presenting Cancer-Associated Fibroblasts. Cancer Discov 2019;9:1102-23. [Crossref] [PubMed]
21. Ritchie ME, Phipson B, Wu D, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 2015;43:e47. [Crossref] [PubMed]
22. Yu G, Wang LG, Han Y, et al. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 2012;16:284-7. [Crossref] [PubMed]
23. Qin XJ, Lin X, Xue G, et al. CXCL10 is a potential biomarker and associated with immune infiltration in human papillary thyroid cancer. Biosci Rep 2021;41:BSR20203459. [Crossref] [PubMed]
24. Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics 2000;56:337-44. [Crossref] [PubMed]
25. Newman AM, Liu CL, Green MR, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 2015;12:453-7. [Crossref] [PubMed]
26. Hellmann MD, Ciuleanu TE, Pluzanski A, et al. Nivolumab plus Ipilimumab in Lung Cancer with a High Tumor Mutational Burden. N Engl J Med 2018;378:2093-104. [Crossref] [PubMed]
27. Lee J, Kim EH. Mechanisms underlying response and resistance to immune checkpoint blockade in cancer immunotherapy. Front Oncol 2023;13:1233376. [Crossref] [PubMed]
28. Ji X, Zhu X, Lu X. Effect of cancer-associated fibroblasts on radiosensitivity of cancer cells. Future Oncol 2017;13:1537-50. [Crossref] [PubMed]
29. Li X, Sun Z, Peng G, et al. Single-cell RNA sequencing reveals a pro-invasive cancer-associated fibroblast subgroup associated with poor clinical outcomes in patients with gastric cancer. Theranostics 2022;12:620-38. [Crossref] [PubMed]
30. Wang R, Sun Y, Yu W, et al. Downregulation of miRNA-214 in cancer-associated fibroblasts contributes to migration and invasion of gastric cancer cells through targeting FGF9 and inducing EMT. J Exp Clin Cancer Res 2019;38:20. [Crossref] [PubMed]
31. Zheng S, Wang J, Ding N, et al. Prodrug polymeric micelles integrating cancer-associated fibroblasts deactivation and synergistic chemotherapy for gastric cancer. J Nanobiotechnology 2021;19:381. [Crossref] [PubMed]
32. Ucaryilmaz Metin C, Ozcan G. Comprehensive bioinformatic analysis reveals a cancer-associated fibroblast gene signature as a poor prognostic factor and potential therapeutic target in gastric cancer. BMC Cancer 2022;22:692. [Crossref] [PubMed]
33. Yang T, Huang T, Zhang D, et al. TGF-β receptor inhibitor LY2109761 enhances the radiosensitivity of gastric cancer by inactivating the TGF-β/SMAD4 signaling pathway. Aging (Albany NY) 2019;11:8892-910. [Crossref] [PubMed]
34. Chen ZL, Qin L, Peng XB, et al. INHBA gene silencing inhibits gastric cancer cell migration and invasion by impeding activation of the TGF-β signaling pathway. J Cell Physiol 2019;234:18065-74. [Crossref] [PubMed]
35. Zhou X, Men X, Zhao R, et al. miR-200c inhibits TGF-β-induced-EMT to restore trastuzumab sensitivity by targeting ZEB1 and ZEB2 in gastric cancer. Cancer Gene Ther 2018;25:68-76. [Crossref] [PubMed]
36. Zhu QL, Li Z, Lv CM, et al. MiR-187 influences cisplatin-resistance of gastric cancer cells through regulating the TGF-β/Smad signaling pathway. Eur Rev Med Pharmacol Sci 2019;23:9907-14. [Crossref] [PubMed]
37. Oshita SE, Chen F, Kwan T, et al. The small heat shock protein HspB2 is a novel anti-apoptotic protein that inhibits apical caspase activation in the extrinsic apoptotic pathway. Breast Cancer Res Treat 2010;124:307-15. [Crossref] [PubMed]
38. Yu Z, Wang H, Fang Y, et al. Molecular chaperone HspB2 inhibited pancreatic cancer cell proliferation via activating p53 downstream gene RPRM, BAI1, and TSAP6. J Cell Biochem 2020;121:2318-29. [Crossref] [PubMed]
39. Qian J, Huang C, Zhu Z, et al. NFE2L3 promotes tumor progression and predicts a poor prognosis of bladder cancer. Carcinogenesis 2022;43:457-68. [Crossref] [PubMed]
40. Wang X, Li Y, Fang Z, et al. Elevated expression of NFE2L3 promotes the development of gastric cancer through epithelial-mesenchymal transformation. Bioengineered 2021;12:12204-14. [Crossref] [PubMed]
41. Zhu F, Dai SN, Xu DL, et al. EFNB2 facilitates cell proliferation, migration, and invasion in pancreatic ductal adenocarcinoma via the p53/p21 pathway and EMT. Biomed Pharmacother 2020;125:109972. [Crossref] [PubMed]
42. Shou Y, Liu Y, Xu J, et al. TIMP1 Indicates Poor Prognosis of Renal Cell Carcinoma and Accelerates Tumorigenesis via EMT Signaling Pathway. Front Genet 2022;13:648134. [Crossref] [PubMed]
43. Alonso R, Flament H, Lemoine S, et al. Induction of anergic or regulatory tumor-specific CD4(+) T cells in the tumor-draining lymph node. Nat Commun 2018;9:2113. [Crossref] [PubMed]
44. Dong Y, Wan Z, Gao X, et al. Reprogramming Immune Cells for Enhanced Cancer Immunotherapy: Targets and Strategies. Front Immunol 2021;12:609762. [Crossref] [PubMed]
45. Chen YP, Zhang Y, Lv JW, et al. Genomic Analysis of Tumor Microenvironment Immune Types across 14 Solid Cancer Types: Immunotherapeutic Implications. Theranostics 2017;7:3585-94. [Crossref] [PubMed]
46. Lu Y, He W, Huang X, et al. Strategies to package recombinant Adeno-Associated Virus expressing the N-terminal gasdermin domain for tumor treatment. Nat Commun 2021;12:7155. [Crossref] [PubMed]