Skip to main content
Download PDF

Single-cell and spatial-resolved profiling reveals cancer-associated fibroblast heterogeneity in colorectal cancer metabolic subtypes

Journal of Translational Medicine volume 23, Article number: 175 (2025) Cite this article

Abstract

Background

Colorectal cancer (CRC) presents significant treatment challenges due to its high heterogeneity and complex intercellular interactions. Further exploration of CRC subtypes and interactions among tumor-specific clusters will facilitate the development of personalized treatment strategies.

Methods

Single-cell RNA sequencing and bulk RNA sequencing datasets were integrated to determine CRC metabolic subtypes by hierarchical clustering. The analysis was further extended to cellchat, pseudotime, immune infiltration, and clinicopathological relevance to explore the characteristics of secreted frizzled related protein 2 (SFRP2) + cancer-associated fibroblast (CAF) clusters, and validated by spatial transcriptomics (ST), in vivo experiments, and multiple immunohistochemistry (mIHC).

Results

CRC samples were stably classified into three heterogeneous metabolic subtypes, each exhibiting different microenvironment and CAF heterogeneity, particularly in the distribution of SFRP2 + CAF, which was aligned with metabolic activity. SFRP2 + CAF exhibits high extracellular matrix (ECM) activity and is closely involved in cellular communication, not only promoting the malignant progression of cancer cells but also inducing the differentiation of Tregs. Compared to responders of chemotherapy, the proportion of SFRP2 + CAFs is significantly increased in non-responders. Importantly, mIHC and ST analyses confirm that cancer cells with low expression of agmatinase (AGMAT) can recruit SFRP2 + CAFs, and Treg infiltration surrounding SFRP2 + CAFs was observed. AGMAT combined with oxaliplatin showed the best efficacy in vivo, which may be associated with the inhibition of SFRP2 + CAF infiltration.

Conclusions

Our study identified and described the potential protumor biological properties of SFRP2 + CAFs, and AGMAT may be a valuable target for disrupting their properties.

Introduction

Colorectal cancer (CRC) is the fourth leading cause of cancer-related deaths, and its incidence rate ranks third worldwide. There are 1.9 million new cases and 900,000 deaths each year, and thus, CRC poses a severe threat to the lives and health of the global population [1]. CRC patients are often diagnosed at an advanced tumor stage, missing the window for surgical intervention [2], with a 5-year overall survival (OS) rate of approximately 50%, whereas it drops to roughly 13% for patients with stage IV CRC [3, 4]. In order to improve the resection rate and reduce local recurrence, the oxaliplatin containing chemotherapy regimens, such as CapeOx (capecitabine and oxaliplatin) and FOLFOX (leucovorin, fluorouracil, and oxaliplatin), are the first-line means of CRC adjuvant chemotherapy. However, chemotherapy resistance, especially in advanced CRC, remains a significant challenge [5, 6]. The occurrence and development of drug resistance do not arise solely from individual tumor cells; rather, they are often the result of prolonged interactions among various cellular components. The concept of treatment centered on the tumor microenvironment (TME) provides new strategies to restore chemotherapy sensitivity and increase cure rates.

Cancer cells, stromal cells, immune cells, and other noncellular components (extracellular matrix and cytokines) collectively shape the complex TME [7]. Cancer-associated fibroblasts (CAFs) are the major stromal cells in the TME and secrete various cytokines, metabolites and matrix-remodeling enzymes to influence the TME [7, 8]. CAFs not only impede drug infiltration into the TME but also increase cancer cell resistance through intercellular “crosstalk“ [9, 10]. Additionally, CAFs suppress the maturation of dendritic cells (DCs) and promote the differentiation of regulatory T cells (Tregs), inducing tumor immune evasion [11, 12]. Furthermore, recent research has revealed distinct CAF subtypes with heterogeneous properties through single-cell sequencing (scRNA-seq), suggesting their diversity. However, whether CAF subtypes in CRC induce chemoresistance remains to be further investigated.

In recent years, researchers have attempted to develop molecular classifications and features based on scRNA-seq and transcriptional profiling, and have comprehensively analyzed the “crosstalk” between various cells in the TME, providing a better understanding of CRC heterogeneity [13]. As a metabolically disordered disease, CRC exhibits metabolic heterogeneity that is associated with tumor drug resistance, survival, and genetic mutations [14, 15]. However, it has yet to be clarified whether heterogeneity in different metabolic pathways divides CRC into clinically relevant subgroups.

In this study, through a multiomics approach, we identified three CRC subgroups based on 84 tumor metabolism-related signature scores. The distribution and heterogeneity of CAFs, particularly SFRP2 + CAFs, were aligned with the metabolic activity of three subgroups. We integrated single-cell analysis with experimental validations (in vivo and in vitro) to explore the biological properties of SFRP2 + CAFs, and provided a potential target for TME chemotherapy in CRC.

Materials and methods

Data sources and preprocessing

The colorectal cancer scRNA-seq dataset files of GSE178341, GSE232525 and GSE178318 were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/), including 127 samples from 62 CRC patients [16]. Using the “CreateSeuratObject” package in R software (version 4.2.1), the CRC samples were converted into Seurat objects. The cell selection criteria were as follows: cells containing 200–10,000 features; each cell expressed > 1000 genes; genes shared by more than three cells; and the content of mitochondrial RNA in each cell was < 20%.

The TCGA-COAD RNA-seq data (TPM) and corresponding clinical data were obtained from the TCGA database (https://portal.gdc.cancer.gov/), and the GSE39582, GSE28702, GSE17536, GSE106584 and GSE76092 datasets were obtained from the GEO database. The diagnostic slides in the TCGA-CRC cohort were downloaded from the TCGA database, and three cell types (cancer cells, CAFs, and lymphocytes) were labeled using QuPath-0.3.2. We used the GSE178341 and TCGA-COAD cohorts as the discovery cohort, and the GSE232525, GSE178318, GSE39582, and GSE17536 cohorts as the validation cohort for subsequent investigations.

Hierarchical clustering of CRC patients

Pathway-level metabolic gene set variation analysis (GSVA) based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database was performed using the R Bioconductor package GSVA v1.46 [17].We performed pseudo-bulk analysis and TPM format conversion on the single-cell data GSE178341 [18]. According to the GSVA signature scores, CRC samples from TCGA, GSE39582, and GSE178341 were classified into 3 subgroups at the bulk RNA-seq level (TPM format) using a hierarchical clustering method. According to metabolic activity, we defined the hyperactive group as G1, the hypoactive group as G3, and the mixed (fluctuating between the two subtypes) group as G2.

scRNA-seq data integration and cell type identification

The “NormalizeData” function was used to standardize the expression values. The 3000 highly variable genes (HVGs) were identified through “SelectIntegrationFeatures”, and the “ScaleData” function was used for centering. Then, the “RunPCA” function was used to reduce the dimensionality of the Seurat set of HVGs. Through the “FindNeighbors”, “FindClusters” functions and t-distributed stochastic neighborhood embedding (tSNE) method with parameter settings of PC = 8 and a resolution = 0.3, we achieved effective clustering visualization.

Through a literature review, 7 kinds of marker genes were collected [19]. The “ReactomeGSA” R package was used to analyze the enriched metabolic pathways of the scRNA-seq data [20].

Cell-cell communication analysis and pseudotime analysis

To investigate ligand‒receptor interactions between different cells and determine cell‒cell communication, we utilized the “CellChat” R package to construct a cellular communication network [21].

The “monocle” R package was used to perform pseudotime analysis on CAF subtypes with reduction_method = “DDRTree” and max_components = 2 [22]. Using the “plot_cell_trajectory” function for ordering cells according to their pseudotime. The “plot_genes_branched_heatmap” function was used for visualization.

LASSO regression analysis and prognostic model construction

The common differentially expressed genes (DEGs) between the G1 group and G3 group were screened in the TCGA and GSE39582 datasets using the R package “limma” (cutoff criteria: adjusted P < 0.05; |log2-fold change| ≥ 0.5). Univariate Cox regression was used to identify potential prognostic DEGs (P < 0.05). Then, we used least absolute shrinkage and selection operator (LASSO) analysis to further identify effective biomarkers [23]. Random forest analysis was carried-out with 1,000 trees, using randomForest R package.

The “glmnet” package was used to calculate the risk score. The “survival receiver operating characteristic (ROC)” package was used to plot the 1-year, 3-year, and 5-year OS curves of CRC patients. We further constructed a nomogram on the basis of the risk score and clinical parameters in the TCGA cohort.

Immune infiltration analysis

The EPIC algorithms and ssGSEA were used to calculate the immune cell infiltration scores of CRC patients in the TCGA and GSE39582 datasets through the R package “IBOR” [24].

Functional enrichment analysis

KEGG pathway enrichment analysis and Gene Ontology (GO) enrichment analysis were performed via DAVID 6.8 (https://david.ncifcrf.gov/). Gene Set Enrichment Analysis (GSEA) was performed using the SangerBox online website (http://sangerbox.com/index.html).

Processing of spatial transcriptomics (ST) data

The CRC spatial transcriptomics data was obtained from Gao’ study (http://www.cancerdiversity.asia/scCRLM/) [25]. To acquire the spatial coordinates of cells, we utilized CellTrek (v0.0.94) R package with default parameters. This tool integrates single-cell data and ST data using machine learning methods, mapping individual cells directly back to their spatial coordinates in tissue slices. We employed the run_kdist function to calculate the spatial k-distance between cancer cells and other cell types.

Cell culture

The human CRC cell lines (HCT116 cell lines) and HEK293T were obtained from the American Type Culture Collection (ATCC). All cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) or RPMI-1640 medium supplemented with 10% FBS and 1% P/S (Thermo Fisher Scientific, #15070063) and incubated at 37 °C in a 5% CO2 incubator.

Plasmid construction and transfection

The expression plasmids for Agmatinase (AGMAT) were constructed from the PCDH-Flag vectors, with empty vector (EV) serving as the control plasmid. The expression efficiency was tested via Western blot. Plasmid construction was conducted with a ClonExpress II One Step Cloning Kit (Vazyme, #C112). The plasmid transfection experiments were conducted with Lipofectamine™ 3000 Transfection Reagent (Thermo Fisher, #L3000-015) according to the manufacturer’s instructions. Then, the AGMAT expression plasmids, the lentiviral packaging plasmid psPAX.2 (Addgene, #12260) and the envelope plasmid pMD2.G(Addgene, #12259) were cotransfected into HEK293T cells. After 48 h, the lentiviruses were used to infect HCT116 cells, and the cells were then screened with puromycin for 3 days.

Other basic experiments

Details of the methods used for RNA extraction, reverse transcription, qPCR, Western blot analysis, and hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining were described in our previous study [26, 27].

Animal experiments

All BALB/c nude mice (4 weeks old, male) were purchased from Guangdong Gem Pharmatech (Foshan, China). The treatment of the mice was carried out in strict accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals. The experimental procedures were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University.

Subcutaneous xenograft models were established by subcutaneous injection of 2 × 106 HCT116 vector cells or HCT116 AGMAT cells into the left axilla of BALB/c nude mice. After 1 week, oxaliplatin (5 mg/kg body weight) was injected intraperitoneally three times a week [28]. After 3 weeks, all the mice were sacrificed. The subcutaneous tumors were removed and weighed. Tumor size was measured every 3 days. Tumour volume was calculated using the following formula: volume = 0.52 × short diameter2 × long diameter.

Patient samples

The study was approved by the Ethics Committee of The First Affiliated Hospital, Sun Yat-sen University (Ethical Approval No.2024 − 656) and conducted in accordance with recognized ethical guidelines of the Declaration of Helsinki. mIHC analyses were performed in tumor tissues from 6 patients with CRC. The baseline characteristics of these 6 CRC patients are provided in Table S7.

Statistical analysis

SPSS 16.0 and GraphPad 8.0 were used for the statistical analyses. Comparisons between two groups of continuous variable data were subjected to unpaired t-test or rank sum test. Survival analysis was performed using the Kaplan‒Meier method and the log-rank test. Two-variable correlation analysis was performed via the Pearson method. p < 0.05 was considered to indicate statistical significance.

Results

Three subtypes based on metabolic pathways according to the scRNA-seq data

To establish a spatial single-cell landscape in CRC, we acquired scRNA-seq data from 137 CRC samples and bulk RNA-seq data from 1318 CRC samples, along with ST data from 3 tissue slices (Fig. 1A).

Fig. 1
figure 1

Three CRC subtypes based on metabolic pathways according to the scRNA-seq data. (A) The workflow of the present study. (B) Heatmap of metabolic pathway-based GSVA analysis of the GSE178341 CRC samples. (C) PCA plot of three different subgroups. (D) t-SNE plots showing the distribution of the three subgroups. (E) The 7 cell types were identified by marker genes. (F) The cell types present in the different subgroups. (G) Heatmap showing the ORs of metaclusters occurring in each group. OR > 1.5 indicates that the clusters are preferred to distribute in the corresponding tissue. (H) ECM signaling pathway enrichment analysis for the three subgroups using the “ReactomeGSA” package. (I) Cell‒cell communication analysis using the “CellChat” package

Full size image

Based on the pathway GSVA enrichment scores of 84 KEGG metabolic pathways (Table S1), 127 tumor samples (GSE178341, pseudo-bulk TPM format) were classified into three heterogeneous subtypes through hierarchical clustering. According to metabolic activity, we defined Group 1 (G1) as hypoactive (n = 29), Group 2 (G2) as mixed (fluctuating between the two subtypes; n = 55), and Group 3 (G3) as hyperactive (n = 43; Fig. 1B). Principal component analysis (PCA) illustrated the existence of three distinct transcriptomic clusters (Fig. 1C).

Ligand‒receptor interactions between CAFs and epithelial/cancer cells

Subsequently, we performed standardization, centralization, and PCA downscaling for cluster visualization of the scRNA-seq data (Figure S1A). Figure 1D shows the t-SNE distribution in three metabolic subgroups. According to the cell gene markers, we classified the cells into 7 clusters: B cells (n = 15510), Macrophages (n = 13828), T cells (n = 13411), epithelial/cancer cells (n = 8201), endothelial cells (n = 2065), CAFs (n = 2674), Monocytes (n = 12906) (Fig. 1E, S1B). Compared with the G2 and G3 groups, the odds ratios (ORs) and proportion of CAFs was relatively high, and the proportion of epithelial/cancer cells was relatively low in the hypoactive G1 group. (Fig. 1F-G).

We focused on CAFs in our subsequent analyses primarily because the distribution of three CAF subtypes aligned with the metabolic activity of our three subgroups. CAFs have distinctive feature of possessing ECM remodeling capabilities in the TME [8, 10]. As expected, ReactomeGSA analysis revealed that ECM-related pathways (ECM proteoglycans, collagen biosynthesis) were significantly enriched in the hypoactive G1 group but were least enriched in the hyperactive G3 group (Fig. 1H). Importantly, in the cell‒cell interaction network, CAFs exhibited the greatest interaction weights/strengths with epithelial/cancer cells, suggesting their intimate cell communication (Fig. 1I).

Three subtypes based on metabolic pathways in bulk RNA-seq data

To confirm the universality and accuracy of our metabolic subtypes, we also utilized the same method in both the TCGA-CRC cohort and the GSE39582-CRC cohort. Similar to the results of the scRNA-seq data, all CRC patients in both datasets were consistently classified into three similar metabolic subgroups (hypoactive G1, mixed G2 and hyperactive G3; Fig. 2A; Figure S2A). In both the TCGA-CRC and GSE39582-CRC cohorts, prognosis analysis revealed that patients in the G1 group had the shortest overall survival (OS), and patients in the G3 group exhibited the most favorable prognosis (Fig. 2B, S2B).This trend was consistent in DFS analysis for patients with early-stage disease (Fig. 2C). Table S2 showed the prognostic analysis was not confounded by differences in disease stage and age among three subgroups.

Fig. 2
figure 2

Three CRC subtypes based on metabolic pathways according to bulk RNA-seq data. (A) Heatmap of metabolic pathway-based GSVA analysis of TCGA-CRC samples. (B-C) Kaplan‒Meier curves of OS and DFS among subgroups in the TCGA-CRC cohort. (D) Volcano plots showing DEGs in G1 or G3 samples. (E-F) GO and KEGG enrichment analyses of genes upregulated in the G1 and G3 samples. (G-H) GSEA identified upregulated and downregulated signaling pathways in G1 samples. (I-J) The violin plot illustrates the abundance of CAFs calculated using the ssGSEA method, as well as the abundance of seven immune cell types calculated using the EPIC method. Rank sum test were used to compare two groups and Kruskal-Wallis rank sum test for more than two groups

Full size image

CAF infiltration and the ECM signaling pathway are enriched in the hypoactive G1 group

To determine specific transcriptome patterns, we identified DEGs between the G1 and G3 groups using volcano plots (Fig. 2D, S2C). GO and KEGG enrichment analyses indicated that “cell proliferation,” and the “PI3K/Akt signaling pathway” were enriched in the G1 group. As expected, “immunoglobulin production” and “metabolic pathways” were enriched in the G3 group (Fig. 2E-F, S2D-E). Similarly, the results of GSEA showed that metabolic pathways such as “citrate cycle” and “fatty acid metabolism” were significantly enriched in the G3 group, and “ECM receptor interaction” was enriched in the G1 group (Fig. 2G-H). Somatic variations indicated a greater mutation rate of TP53 in the G1 group and a greater mutation rate of PIK3CA in the G3 group (p < 0.05; Figure S2F). These results suggest that the enrichment of the PI3K/Akt signaling pathway in G1 could be attributed to alternative drivers or broader pathway involvement beyond PIK3CA mutations.

Furthermore, we used the EPIC methods and CAFs related signature [29] ssGSEA scores to determine the abundance of different cell clusters in the TME of the three metabolic subgroups. Consistent with the results of the scRNA-seq data, in both the TCGA and GSE39582 databases, CAFs exhibited the highest cellular abundance in the G1 group and the lowest in the G3 group (Fig. 2I-J, S2G-H).

Key metabolic gene identification and nomogram model construction

Metabolic activity is an independent prognostic factor influencing OS in both the G1 and G3 groups (Table S2-S3). Metabolic activity is regulated by various metabolic enzymes, and we hypothesize that some key metabolic enzymes may contribute to the differences in metabolic activity and OS. We focused on 1687 metabolic enzymes from the KEGG database (Table S4), and identified 691 common differentially expressed metabolic enzymes between G1 and G3 groups in the GSE39582, and TCGA cohorts (|logFC|>0.5, adj P < 0.05). Subsequently, univariate Cox and LASSO regression analyses were performed, and ten DEGs (GPX3, AGMAT, L3HPDH, RIMKLB, DHODH, GLCE, GSR, NOS2, COX6B1, ASRGL1) were included in the LASSO regression model (Figure S3A). In the training cohort, the LASSO regression model demonstrated AUC values of 0.676, 0.722, and 0.739 for predicting 1-, 3-, and 5-year survival, respectively. Although the AUC values for 1-, 3-, and 5-year survival predictions in two validation cohorts were below 0.7, the model remained effective, as patients in the high-risk group consistently exhibited worse prognosis (P < 0.05; Fig. 3A-F). Table S5 demonstrated that the riskscore, similar to other important clinical indicators, had independent prognostic significance. These findings highlight the potential clinical relevance of the model, though further optimization is required to enhance its predictive performance.

Fig. 3
figure 3

Metabolic risk gene identification and nomogram model construction. (A-C) K‒M survival analysis between the low-risk and high-risk groups in the TCGA, GSE39582 and GSE17536 cohorts. (D-F) The receiver operating characteristic (ROC) curves for 1-, 3- and 5-year survival in the TCGA, GSE39582 and GSE17536 cohorts. (G-H) Correlation analysis between the ssGSEA and EPIC scores of CAFs and the riskScore. (I) Representative diagnostic slides of high-risk and low-risk CRC patients in TCGA. Red: cancer cells; yellow: lymphocytes; blue: CAFs. (J) Construction of a risk nomogram in the TCGA cohort. (K) Calibration plots of the nomogram for predicting the 1-year, 3-year and 5-year OS rates. C-index = 0.7535

Full size image

The riskscore showed a statistically positive correlation with CAFs, and representative diagnostic slides from the TCGA-CRC cohort similarly indicated that patients in the high-risk group exhibited increased levels of CAF infiltration (Fig. 3G-I). However, the correlation coefficient (R < 0.3) indicates a weak relationship. This could be due to multiple factors influencing CAF expression, as CAFs may consist of different subpopulations with varying gene expression profiles. Further studies are needed to explore the complex relationship between these genes and CAF clusters in the tumor microenvironment.

Additionally, we incorporated important clinical factors (including T stage, N stage, Stage, Age) and the riskscore to construct a clinically adaptable risk nomogram for predicting the 1-, 3-, and 5-year OS rates of TCGA-CRC patients. (C-index = 0.7535, Fig. 3J-K).

The single-cell transcriptome atlas reveals the heterogeneity of CAFs

CAFs exhibit functional heterogeneity and plasticity in the TME [30]. To determine the specific CAF clusters associated with the metabolic activity of the three subgroups, we further analyzed the distribution of individual CAF clusters. In this study, 2674 CAFs were extracted from GSE178341 for subsequent clustering, and eight CAF clusters were obtained (nCAF_0=968, nCAF_1=517, nCAF_2=456, nCAF_3=363, nCAF_4=131, nCAF_5=110, nCAF_6=82, and nCAF_7=47, Figure S3B). Meanwhile, we obtained the key DEGs to describe the characteristics in the eight CAF clusters (Fig. 4A, S3C). Interestingly, the odds ratios (ORs) and proportion indicated that the abundance of CAF_3 (defined as SFRP2 + CAF) was highest in the G1 group, and it was decreased in the G2 group but was almost absent in the G3 group, perfectly aligning with the trends in metabolic activity. (Fig. 4B-C).

Fig. 4
figure 4

Trajectory analysis of CAFs and identification of SFRP2 + CAF. (A) Feature genes of each CAF clusters. (B) The proportion of CAF clusters in subgroups. (C) Heatmap showing the ORs of CAF clusters occurring in each group. (D) Trajectory plot displaying the identified clusters of CAFs extracted via scRNA-seq. (E) Four subgroups based on the DEGs along the pseudotime axis, with corresponding GO-BP annotations on the left. (F) Expression levels of ECM remodeling-related genes were highest in SFRP2 + CAFs. (G) Umap visualization of AUCELL scores for eight CAF clusters. (H-I) K‒M survival analysis between the low-SFRP2 + CAF and high-SFRP2 + CAF groups in the TCGA and GSE39582 cohorts by using the best performing threshold as the cutoff. (J) Top 3 KEGG pathways in 8 CAF clusters

Full size image

Through cell trajectory analysis, we further investigated the chronological order of CAF cluster differentiation, and found that CAF_0 and CAF_6 belong to the early stage of cellular development, and CAF_1, SFRP2 + CAF, and CAF_4 belong to the terminal stage of cellular development (Fig. 4D). Pseudotime DEGs were categorized into four clusters, primarily associated with angiogenesis, cell differentiation, ECM organization, and immune response (Fig. 4E).

The SFRP2 + CAF has protumor biological properties

To further understand the functional characteristics of SFRP2 + CAFs, we obtained DEGs between SFRP2 + CAFs and other CAFs. The results of GO and KEGG analyses indicated that the DEGs were mainly enriched in “extracellular matrix organization”, “collagen fibril organization”, and “immune response” (Figure S3D). Since SFRP2 + CAFs are mainly distributed in the G1 group, which exhibits active ECM remodeling capabilities, we focused on investigating the abundance of ECM remodeling-related genes, including several collagen genes (COL1A1, COL1A2, COL6A2, and COL6A3), proteoglycans (VCAN), and secreted growth factors (ANGPTL2). The results indicated that the expression levels of these genes in SFRP2 + CAF were higher than other CAF clusters (Fig. 4F). The ECM composition can alter the physical properties (stiffness) of the TME and contribute to tumor growth and drug resistance [31]. Furtherly, GSVA enrichment analysis also suggests that SFRP2 + CAF may be involved in drug metabolism activities (Fig. 4J).

Based on the marker genes of SFRP2 + CAFs, we constructed four gene signatures, which included the top 10 DEGs, top 20 DEGs (ranked by LogFC), top 48 DEGs (LogFC > 1.5, adj. p < 0.05), and top 106 DEGs (LogFC > 1.0, adj. p < 0.05) (Table S6). To assess which gene signature most specifically reflects the SFRP2 + CAF cluster, we found that the SFRP2 + CAF cluster achieved the highest AUCELL score when the top 10 DEGs were used as the gene signature (Figure S4A-D). Therefore, we defined the top 10 DEGs of the SFRP2 + CAF as its specific gene signature (Fig. 4G). Kaplan–Meier analysis showed that high infiltration of SFRP2 + CAFs predicted a poor prognosis in the TCGA and GSE39582 datasets (Fig. 4H-I).

Cell-cell interaction analysis between CAF clusters and cancer cells

We calculated the interaction strength for each cell type and signaling pathway. In the outgoing patterns, SFRP2 + CAFs, CAF_1, and CAF_6 were the predominant cell types. In the incoming patterns, cancer cells were the main cell type. Compared to the other signaling pathways, collagen type I signaling contributed most prominently to the outgoing or incoming patterns of CAFs in the TME (Figure S4E-F).

Effect of SFRP2 + CAFs on TME through cell-cell communication

Recent studies have revealed the existence of distinct transcriptional profiles for different CAF clusters in the TME, including myofibroblasts (myCAFs) [32], inflammatory fibroblasts (iCAFs) [32], antigen-presenting CAFs (apCAFs) [33], vascular CAFs (vCAFs) [33]. To compare the characteristics and origins of our CAF clusters, we obtained specific signature genes for myCAFs, iCAFs, apCAFs, and vCAFs (Table S6) and calculated ssGSEA scores for our eight CAF clusters based on these signatures. The results showed that SFRP2 + CAFs mainly had the characteristics of myCAFs and apCAFs (Fig. 5A-D). This comparison provides insight into the functional and biological properties of SFRP2 + CAFs, aligning them with features of known CAF clusters.

Fig. 5
figure 5

The biological properties of SFRP2 + CAFs. (A-D) ssGSEA scores of myCAFs, apCAFs, iCAFs, and vCAFs signatures in 8 CAF clusters. (E) The cell communications between SFRP2 + CAF and cancer cells. (F-G) The main ligand–receptor pairs between SFRP2 + CAFs and cancer cells. (H) GO and KEGG enrichment of top 30 ligand–receptor pairs between SFRP2 + CAFs and cancer cells. (I) The cell communications between SFRP2 + CAF and CD4 + T cells. (J-K) Top 30 ligand–receptor pairs between SFRP2 + CAFs and CD4 + T cells and GO and KEGG enrichment. (L) The cell communications between SFRP2 + CAF and CD4 + T cells. (M) Treg_quan scores between low and high SFRP2 + CAF in TCGA

Full size image

myCAFs can induce malignant progression of tumor cells [32]. In this study, the interaction weights/strengths between SFRP2 + CAFs and cancer cells were the strongest (Fig. 5E). This interaction predominantly occurs between ligands such as COL1A2 and COL1A1 and receptors such as SDC4 and ITGA2_ITGB1 (Fig. 5G and H). The results of GO and KEGG analyses indicated that the top 30 inferred LRs were mainly enriched in “cell proliferation”, “cell migration” and “ECM-receptor interaction” (Fig. 5F and H). These results showed the most intimate cell communication occurred between SFRP2 + CAFs and cancer cells in the TME.

apCAFs, characterized by high expression of MHC class II molecules, directly induce naïve CD4 + T cells to differentiate into Tregs in an antigen-specific manner [34]. Our study showed that the interaction weights/strengths between SFRP2 + CAFs and CD4 + T cells were the strongest, but there was no apparent correlation between the infiltration abundance (Fig. 5I, S4G). Consistently, the results of GO and KEGG analyses indicated that the top 30 inferred LRs between SFRP2 + CAF and CD4 + T cell were mainly enriched in “antigen-presentation via MHC class II”, “T cell differentiation” (Fig. 5J and K). Further, cellChat analysis indicated that SFRP2 + CAFs exhibit the strongest interaction weights/strengths with Tregs (Fig. 5L). Moreover, high SFRP2 + CAF ssGSEA scores were associated with higher Treg infiltration (Fig. 5M). These findings collectively suggest that SFRP2 + CAFs may play a role in inducing Tregs differentiation and infiltration.

External validation of SFRP2 + CAFs

In order to validate the stability of the SFRP2 + CAFs, we selected and integrated GSE232525 and GSE17831 as a validation cohort, and performed standardization, centralization, and PCA downscaling for cluster visualization (Fig. 6A, S5A-B). 1476 CAFs were extracted for subsequent clustering, and five CAF clusters were obtained (nCAF_0=456, nCAF_1=398, nCAF_2=290, nCAF_3=272, nCAF_4=60). Meanwhile, we obtained the key DEGs to describe the characteristics in the five CAF clusters, and defined CAF_1 as SFRP2 + CAF cluster (Fig. 6B). Consistently, through calculating ssGSEA scores for five CAF clusters, the results also showed that SFRP2 + CAFs mainly had the characteristics of myCAFs and apCAFs (Fig. 6C and F). Interestingly, we found a significant increase in the proportion of SFRP2 + CAFs in patients with chemotherapy resistance (Fig. 6G and H). These results suggested that SFRP2 + CAFs may contribute to chemotherapy resistance in CRC patients.

Fig. 6
figure 6

AGMAT was negatively correlated with SFRP2 + CAF infiltration and External validation. (A) The t-SNE plots of 7 cell types in the validation cohort. (B) Feature genes of each CAF clusters in the validation cohort. (C-F) ssGSEA scores of myCAFs, apCAFs, iCAFs, and vCAFs in 5 CAF clusters. (G-H) SFRP2 + CAF ssGSEA scores of responder and non-responder in GSE28702 and GSE106584. (I) Dot plot showing RIMKLB, AGMAT, and GPX3 gene expression across the 7 cell subtypes in GSE178341. (J) The proportion of CAF clusters between low- and high-AGMAT groups. (K) Correlation analysis between AGMAT expression and ssGSEA scores of SFRP2 + CAFs in TCGA and GSE39582 cohorts. (L) Spatial cell charting and AGMAT expression of ST_CRC1. (M) Spatial cell charting of ST_CRC2 and ST_CRC3 using CellTrek. (N) Boxplot showing the average k-distance from different cell types to cancer cells in three ST_CRC samples. (O) AGMAT expression level between oxaliplatin-resistant and oxaliplatin-sensitive groups. (P) Oxaliplatin sensitivity analysis between low- and high-AGMAT groups in TCGA cohort. (Q) The mRNA expression of AGMAT was detected by qPCR in HCT116 cells treated with oxaliplatin. (R) Cytotoxicity of oxaliplatin and/or AGMAT in HCT116 cells. Unpaired t-test were used to compare two groups. *p<0.05, *p<0.01, *p<0.001

Full size image

AGMAT was negatively correlated with SFRP2 + CAF infiltration and involved in oxaliplatin resistance

To identify key metabolism-related risk genes associated with SFRP2 + CAF infiltration (GPX3, AGMAT, L3HPDH, RIMKLB, DHODH, GLCE, GSR, NOS2, COX6B1, ASRGL1), we performed a random forest analysis in the GSE39582 cohort. The results indicated that GPX3, AGMAT, and L3HPDH are the top three genes most significantly affecting SFRP2 + CAF infiltration (Figure S5C-D). Analysis of the GSE178341 and Validation cohort scRNA datasets revealed that GPX3 and L3HPDH are predominantly expressed in CAFs, while AGMAT is mainly expressed in epithelial/cancer cells (Fig. 6I, S5E). Considering the close association between cancer cells and SFRP2 + CAFs, we subsequently selected and explored the function of AGMAT in cancer cells.

We observed that high expression of AGMAT was significantly associated with a reduced proportion of SFRP2 + CAFs (Fig. 6J). Consistently, in the TCGA and GSE39582 cohorts, the AGMAT expression was negatively correlated with the abundance of SFRP2 + CAFs (p < 0.05) and ECM remodeling-related genes (Fig. 6K, S5F). However, since the R values in the TCGA cohort are relatively small, indicating a weak correlation, further methods are needed to validate this relationship. Furthermore, the results of spatial transcriptomic (ST) data of three CRC patients revealed that SFRP2 + CAFs are primarily distributed within the malignant regions, particularly in those tumor regions with low expression of AGMAT (Fig. 6L-N).

Considering the high infiltration of SFRP2 + CAFs in chemotherapy non-responders, it remains unclear whether AGMAT expression in tumor cells induces chemotherapy resistance by influencing SFRP2 + CAF infiltration. We found that AGMAT was highly expressed in the treatment-resistant group, although further validation in larger cohorts is needed (Fig. 6O). Drug sensitivity analysis revealed that high AGMAT expression predicted increased sensitivity to oxaliplatin compared to low AGMAT expression (Fig. 6P).

The qPCR results showed that the expression levels of AGMAT in HCT116 cells decreased (Fig. 6Q) with oxaliplatin (10 μm) stimulation (12 h, 24 h). However, Cytotoxicity experiments indicated that overexpression of AGMAT did not significantly alter the cytotoxic effect of oxaliplatin in HCT116 cells (Fig. 6R).

AGMAT inhibits SFRP2 + CAF infiltration to increase the therapeutic sensitivity to oxaliplatin in vivo

We hypothesized that AGMAT might alter SFRP2 + CAF infiltration in the TME, thereby affecting the efficacy of oxaliplatin. We used HCT116EV cells and HCT116AGMAT cells to establish subcutaneous tumorigenesis models in nude mice (Figure S5G). The results showed that, compared to those in the EV group, the tumor volume and weight were decreased in the AGMAT group and the EV + oxaliplatin group. The combination of AGMAT overexpression and oxaliplatin had a more significant antitumor effect (Fig. 7A-C). Immunohistochemistry analysis revealed that, compared to that in the EV group, the positive rate of SFRP2 + CAFs in the tumor nests significantly decreased after AGMAT overexpression. Additionally, the intensity of Ki-67-stained tumor cells significantly decreased after AGMAT overexpression combined with oxaliplatin treatment (Figs. 7D and 8).

Fig. 7
figure 7

AGMAT inhibits SFRP2 + CAF infiltration to increase the therapeutic sensitivity to oxaliplatin. (A) Mice tumor growth curves. (B-C) Images of subcutaneous tumors in nude mice and the tumor weights (g) of the four groups. (D) HE staining and IHC staining of SFRP2 and ki-67 in subcutaneous tumors. (E) The mIHC images of indicated SFRP2 + CAF (yellow), AGMAT + cancer (red), CAFs (green) and FOXP3 + Treg (blue) in CRC samples. Quantification of fluorescence intensity (right) of SFRP2 and FOXP3. The yellow arrow indicates ACTA2 + SFRP2 + CAF, and the blue arrow indicates FOXP3 + Treg. Unpaired t-test were used to compare two groups. *p<0.05, *p<0.01, *p<0.001

Full size image
Fig. 8
figure 8

Schematic showing that AGMAT inhibits SFRP2 + CAF infiltration to increase the therapeutic sensitivity to oxaliplatin

Full size image

SFRP2 + CAF was verified by clinical CRC samples

Six clinical CRC samples were selected for multiplex IHC staining of 4 markers. Figure 7E showed that AGMAT was mainly expressed in tumor cells. For patients with low expression of AGMAT, there was significant infiltration of SFRP2 + CAFs. Concurrently, we observed a significant infiltration of FOXP3 + Tregs surrounding SFRP2 + CAFs (Fig. 7E).

Discussion

To improve personalized therapy for CRC, we need to enhance our understanding of the clinical subtypes and TME heterogeneity of CRC. In this study, we divided CRC into three clinical subtypes based on the metabolic transcriptome in multiple databases. scRNA-seq provides valuable information about intracellular RNA and cell clusters in TME [35, 36]. This technology allowed us to determine at single-cell resolution that SFRP2 + CAFs were predominantly present in the hypoactive CRC subtype, which exhibited high ECM activity and was closely involved in cellular communication. AGMAT may be a valuable target to inhibit SFRP2 + CAFs infiltration and provide strategies for the chemotherapy of CRC (Fig. 8).

CRC exhibits significant molecular heterogeneity [37]. With the widespread application of various omics technologies, researchers have proposed multiple CRC molecular classifications based on immune infiltration, epigenetic modifications, and gene expression, among which the consensus molecular subtype (CMS) proposed in 2015 has been the most influential [38, 39]. Metabolic reprogramming is a common phenomenon in the development of tumors. However, no metabolic classification of CRC has been applied for clinical practice so far. In this study, CRC samples were divided into three clinically relevant metabolic subgroups—hypoactive G1, mixed G2, and hyperactive G3, which were validated across multiple datasets. Survival analysis indicated that hypoactive G1 tumors had the worst prognosis, and hyperactive G3 tumors had the best prognosis. Interestingly, we found that CAFs were most abundant in G1, least abundant in G3, and intermediate in G2, perfectly aligning with the trends in metabolic activity. This consistency led us to hypothesize that CAFs may play a pivotal role in bridging metabolic activity and the tumor microenvironment, warranting further investigation into their function and mechanisms.

CAFs represent the most common stromal cells in solid tumors [40, 41], and include diverse highly plastic clusters that exhibit phenotypic heterogeneity in response to changes in the TME [42,43,44]. Widely reported CAF subtypes include myCAFs [32], iCAFs [32], apCAFs [33], vCAFs [33]. However, a consensus regarding the number and nomenclature of CAF clusters has not been reached. In recent years, Sathe et al. utilized scRNA-seq data to confirm the presence of SPP1 + CAFs in metastatic CRC, which supported CRC growth in an immunosuppressed metastatic niche in the liver [45]. The interaction between Microfibril associated protein 5 (MFAP5) + CAFs and myeloid cells promotes CRC malignant progression [46]. In addition to their tumor-promoting effects, CD146 + CAFs, Caveolin 1 (CAV1) + CAFs, and platelet derived growth factor receptor alpha (PDGFRα) + CAFs have been identified as tumor-suppressive CAF clusters in breast cancer [47]. In this study, we identified a novel SFRP2 + CAF cluster that is primarily distributed in the hypoactive G1 tumors. There is evidence suggesting that SFRP2 is associated with the malignant progression of CRC, and ECM synthesis is abnormally active in adjacent regions with high SFRP2 expression [48, 49]. However, there is still a lack of relevant reports on the role of SFRP2 + CAFs in the CRC microenvironment.

Through integrating single-cell analysis along with in vivo and in vitro experiments, we explored the biological properties of SFRP2 + CAFs. Our analysis revealed that SFRP2 + CAFs exhibited high ECM activity, and may represent a terminally differentiated CAF cluster, characterized by features of both myCAFs and apCAFs. SFRP2 + CAFs engaged in intimate cellular communication with cancer cells. In addition, we found that SFRP2 + CAFs promote differentiation and infiltration of Tregs, and the results of mIHC also confirmed a significant infiltration of FOXP3 + Tregs surrounding SFRP2 + CAFs. Interestingly, the expression abundance of SFRP2 + CAFs was higher in the chemotherapy-resistant group compared to the chemotherapy-sensitive group. This suggests that SFRP2 + CAFs are involved in chemotherapy resistance, potentially through their high ECM activity and ability to remodel the extracellular matrix [50].

Considering the specific distribution of SFRP2 + CAFs in metabolic subtypes, we hypothesized that key metabolic enzymes might influence the specific recruitment of SFRP2 + CAFs. Using random forest analysis of ten metabolism-related risk genes, we identified the top three genes most significantly influencing SFRP2 + CAF infiltration. Since only AGMAT was primarily expressed in epithelial/cancer cells, we selected AGMAT for further functional exploration in cancer cells. AGMAT is involved in arginine catabolism, and Mossmann et al. recently reported that low expression of AGMAT in liver cancer induces arginine accumulation, thereby promoting tumor development [51]. Additionally, AGMAT promoted nitric oxide (NO) release by upregulating the expression of inducible nitric oxide synthases (iNOS), ultimately promoting the malignant proliferation of lung adenocarcinoma cells [52]. However, the specific roles of AGMAT in CAF infiltration and chemotherapy resistance are poorly understood. In this study, we found that AGMAT overexpression did not affect the cytotoxic effect of oxaliplatin in vitro, but the combination of oxaliplatin and AGMAT overexpression resulted in a stronger anticancer effect in vivo which may be associated with the inhibition of SFRP2 + CAF infiltration. mIHC and spatial transcriptomic analyses of CRC samples also confirmed that tumor cells with AGMAT low expression can recruit SFRP2 + CAF infiltration.

Conclusions

This study revealed that different CRC subtypes exhibit distinct transcriptional features and TME heterogeneity depending on their metabolic characteristics. Our results provide comprehensive information on the biological properties of SFRP2 + CAFs, with a focus on how AGMAT can inhibit their infiltration in the TME to enhance the therapeutic sensitivity to oxaliplatin. Targeted inhibition of SFRP2 + CAFs may offer a new strategy to overcome chemoresistance in CRC patients.

Data availability

The colorectal cancer scRNA-seq dataset files of GSE178341, GSE232525 and GSE178318 were downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/). The CRC RNA-seq data (TPM) and corresponding clinical data were obtained from the TCGA database (https://portal.gdc.cancer.gov/), and the GSE17536, GSE106584, GSE39582, GSE28702 and GSE76092 datasets were obtained from the GEO database. The diagnostic slides in the TCGA-CRC cohort were downloaded from the TCGA database. The CRC spatial transcriptomics data was obtained from Gao’ study (http://www.cancerdiversity.asia/scCRLM/). All data supporting the conclusions are available from the authors on reasonable request.

Abbreviations

CRC:

Colorectal cancer

KEGG:

Kyoto Encyclopedia of Genes and Genomes

GSVA:

Gene set variation analysis

ECM:

Extracellular matrix

CAFs:

Cancer-associated fibroblasts

TME:

Tumor environment

ST:

Spatial Transcriptomics

References

  1. Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer statistics 2020: GLOBOCAN estimates of incidence and Mortality Worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71:209–49.

    Article  PubMed  Google Scholar 

  2. Sinha R. Colorectal cancer. Clin Radiol. 2021;76:870.

    Article  CAS  PubMed  Google Scholar 

  3. Siegel RL, Wagle NS, Cercek A, Smith RA, Jemal A. Colorectal cancer statistics, 2023. CA Cancer J Clin. 2023;73:233–54.

    Article  PubMed  Google Scholar 

  4. Li X, Wu Y, Tian T. TGF-beta signaling in metastatic colorectal Cancer (mCRC): from underlying mechanism to potential applications in Clinical Development. Int J Mol Sci. 2022;23(22):14436.

  5. McQuade RM, Stojanovska V, Bornstein JC, Nurgali K. Colorectal Cancer chemotherapy: the evolution of treatment and New approaches. Curr Med Chem. 2017;24:1537–57.

    Article  CAS  PubMed  Google Scholar 

  6. Cao Y, Tang H, Wang G, Li P, Song Z, Li W, et al. Targeting survivin with Tanshinone IIA inhibits tumor growth and overcomes chemoresistance in colorectal cancer. Cell Death Discov. 2023;9:351.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Heichler C, Scheibe K, Schmied A, Geppert CI, Schmid B, Wirtz S, et al. STAT3 activation through IL-6/IL-11 in cancer-associated fibroblasts promotes colorectal tumour development and correlates with poor prognosis. Gut. 2020;69:1269–82.

    Article  CAS  PubMed  Google Scholar 

  8. Belhabib I, Zaghdoudi S, Lac C, Bousquet C, Jean C. Extracellular matrices and Cancer-Associated fibroblasts: targets for Cancer diagnosis and therapy? Cancers (Basel). 2021; 13.

  9. 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 

  10. Linares J, Sallent-Aragay A, Badia-Ramentol J, Recort-Bascuas A, Mendez A, Manero-Ruperez N, et al. Long-term platinum-based drug accumulation in cancer-associated fibroblasts promotes colorectal cancer progression and resistance to therapy. Nat Commun. 2023;14:746.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Varveri A, Papadopoulou M, Papadovasilakis Z, Compeer EB, Legaki AI, Delis A, et al. Immunological synapse formation between T regulatory cells and cancer-associated fibroblasts promotes tumour development. Nat Commun. 2024;15:4988.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Huang TX, Tan XY, Huang HS, Li YT, Liu BL, Liu KS, et al. Targeting cancer-associated fibroblast-secreted WNT2 restores dendritic cell-mediated antitumour immunity. Gut. 2022;71:333–44.

    Article  CAS  PubMed  Google Scholar 

  13. Isella C, Brundu F, Bellomo SE, Galimi F, Zanella E, Porporato R, et al. Selective analysis of cancer-cell intrinsic transcriptional traits defines novel clinically relevant subtypes of colorectal cancer. Nat Commun. 2017;8:15107.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Sedlak JC, Yilmaz OH, Roper J. Metabolism and colorectal Cancer. Annu Rev Pathol. 2023;18:467–92.

    Article  CAS  PubMed  Google Scholar 

  15. Peng X, Chen Z, Farshidfar F, Xu X, Lorenzi PL, Wang Y, et al. Molecular characterization and clinical relevance of metabolic expression subtypes in human cancers. Cell Rep. 2018;23:255–69. e4.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Pelka K, Hofree M, Chen JH, Sarkizova S, Pirl JD, Jorgji V, et al. Spatially organized multicellular immune hubs in human colorectal cancer. Cell. 2021;184:4734–e5220.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Chaffin M, Papangeli I, Simonson B, Akkad AD, Hill MC, Arduini A, et al. Single-nucleus profiling of human dilated and hypertrophic cardiomyopathy. Nature. 2022;608:174–80.

    Article  CAS  PubMed  Google Scholar 

  19. Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol. 2018;36:411–20.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Griss J, Viteri G, Sidiropoulos K, Nguyen V, Fabregat A, Hermjakob H. ReactomeGSA - efficient multi-omics comparative pathway analysis. Mol Cell Proteom. 2020;19:2115–25.

    Article  CAS  Google Scholar 

  21. Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan CH, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12:1088.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Trapnell C, Cacchiarelli D, Grimsby J, Pokharel P, Li S, Morse M, et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol. 2014;32:381–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Duan J, Soussen C, Brie D, Idier J, Wan M, Wang YP. Generalized LASSO with under-determined regularization matrices. Sig Process. 2016;127:239–46.

    Article  Google Scholar 

  24. Zeng D, Ye Z, Shen R, Yu G, Wu J, Xiong Y, et al. IOBR: Multi-omics Immuno-Oncology Biological Research to Decode Tumor Microenvironment and signatures. Front Immunol. 2021;12:687975.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Wu Y, Yang S, Ma J, Chen Z, Song G, Rao D, et al. Spatiotemporal Immune Landscape of Colorectal Cancer Liver Metastasis at single-cell level. Cancer Discov. 2022;12:134–53.

    Article  CAS  PubMed  Google Scholar 

  26. Wen X, Qin J, Zhang X, Ye L, Wang Y, Yang R, et al. MEK-mediated CHPF2 phosphorylation promotes colorectal cancer cell proliferation and metastasis by activating NF-kappaB signaling. Cancer Lett. 2024;584:216644.

    Article  CAS  PubMed  Google Scholar 

  27. Di Y, Zhang X, Wen X, Qin J, Ye L, Wang Y, et al. MAPK signaling-mediated RFNG phosphorylation and nuclear translocation restrain Oxaliplatin-Induced apoptosis and Ferroptosis. Adv Sci (Weinh). 2024;11:e2402795.

    Article  PubMed  Google Scholar 

  28. Di Y, Jing X, Hu K, Wen X, Ye L, Zhang X, et al. The c-MYC-WDR43 signalling axis promotes chemoresistance and tumour growth in colorectal cancer by inhibiting p53 activity. Drug Resist Updat. 2023;66:100909.

    Article  CAS  PubMed  Google Scholar 

  29. Bagaev A, Kotlov N, Nomie K, Svekolkin V, Gafurov A, Isaeva O et al. Conserved pan-cancer microenvironment subtypes predict response to immunotherapy. Cancer Cell. 2021; 39: 845 – 65 e7.

  30. Feng B, Wu J, Shen B, Jiang F, Feng J. Cancer-associated fibroblasts and resistance to anticancer therapies: status, mechanisms, and countermeasures. Cancer Cell Int. 2022;22:166.

    Article  PubMed  PubMed Central  Google Scholar 

  31. Shen Y, Wang X, Lu J, Salfenmoser M, Wirsik NM, Schleussner N et al. Reduction of liver metastasis stiffness improves response to Bevacizumab in Metastatic Colorectal Cancer. Cancer Cell. 2020; 37: 800 – 17 e7.

  32. Affo S, Nair A, Brundu F, Ravichandra A, Bhattacharjee S, Matsuda M, et al. Promotion of cholangiocarcinoma growth by diverse cancer-associated fibroblast subpopulations. Cancer Cell. 2021;39:866–82. e11.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Zhang M, Yang H, Wan L, Wang Z, Wang H, Ge C, et al. Single-cell transcriptomic architecture and intercellular crosstalk of human intrahepatic cholangiocarcinoma. J Hepatol. 2020;73:1118–30.

    Article  CAS  PubMed  Google Scholar 

  34. Huang H, Wang Z, Zhang Y, Pradhan RN, Ganguly D, Chandra R et al. Mesothelial cell-derived antigen-presenting cancer-associated fibroblasts induce expansion of regulatory T cells in pancreatic cancer. Cancer Cell. 2022; 40: 656 – 73 e7.

  35. Kanton S, Boyle MJ, He Z, Santel M, Weigert A, Sanchis-Calleja F, et al. Organoid single-cell genomic atlas uncovers human-specific features of brain development. Nature. 2019;574:418–22.

    Article  CAS  PubMed  Google Scholar 

  36. Zhai Y, Zhang J, Huang Z, Shi R, Guo F, Zhang F, et al. Single-cell RNA sequencing integrated with bulk RNA sequencing analysis reveals diagnostic and prognostic signatures and immunoinfiltration in gastric cancer. Comput Biol Med. 2023;163:107239.

    Article  CAS  PubMed  Google Scholar 

  37. Cancer Genome Atlas N. Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487:330–7.

    Article  Google Scholar 

  38. Guinney J, Dienstmann R, Wang X, de Reynies A, Schlicker A, Soneson C, et al. The consensus molecular subtypes of colorectal cancer. Nat Med. 2015;21:1350–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Huang TX, Fu L. The immune landscape of esophageal cancer. Cancer Commun (Lond). 2019;39:79.

    Article  PubMed  Google Scholar 

  40. Hsiao YW, Chi JY, Li CF, Chen LY, Chen YT, Liang HY, et al. Disruption of the pentraxin 3/CD44 interaction as an efficient therapy for triple-negative breast cancers. Clin Transl Med. 2022;12:e724.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Lavie D, Ben-Shmuel A, Erez N, Scherz-Shouval R. Cancer-associated fibroblasts in the single-cell era. Nat Cancer. 2022;3:793–807.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Biffi G, Oni TE, Spielman B, Hao Y, Elyada E, Park Y, et al. IL1-Induced JAK/STAT signaling is antagonized by TGFbeta to shape CAF heterogeneity in pancreatic ductal adenocarcinoma. Cancer Discov. 2019;9:282–301.

    Article  PubMed  Google Scholar 

  43. Sahai E, Astsaturov I, Cukierman E, DeNardo DG, Egeblad M, Evans RM, et al. A framework for advancing our understanding of cancer-associated fibroblasts. Nat Rev Cancer. 2020;20:174–86.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Chen B, Chan WN, Xie F, Mui CW, Liu X, Cheung AHK, et al. The molecular classification of cancer-associated fibroblasts on a pan-cancer single-cell transcriptional atlas. Clin Transl Med. 2023;13:e1516.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Sathe A, Mason K, Grimes SM, Zhou Z, Lau BT, Bai X, et al. Colorectal Cancer metastases in the liver establish immunosuppressive spatial networking between Tumor-Associated SPP1 + macrophages and fibroblasts. Clin Cancer Res. 2023;29:244–60.

    Article  CAS  PubMed  Google Scholar 

  46. Peng Z, Ren Z, Tong Z, Zhu Y, Zhu Y, Hu K. Interactions between MFAP5 + fibroblasts and tumor-infiltrating myeloid cells shape the malignant microenvironment of colorectal cancer. J Transl Med. 2023;21:405.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Brechbuhl HM, Finlay-Schultz J, Yamamoto TM, Gillen AE, Cittelly DM, Tan AC, et al. Fibroblast subtypes regulate responsiveness of luminal breast Cancer to Estrogen. Clin Cancer Res. 2017;23:1710–21.

    Article  CAS  PubMed  Google Scholar 

  48. Suwatthanarak T, Tanjak P, Suwatthanarak T, Acharayothin O, Thanormjit K, Chaiboonchoe A, et al. Exploring extracellular matrix and prostaglandin pathway alterations across varying resection margin distances of right-sided colonic adenocarcinoma. BMC Cancer. 2023;23:1202.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Li S, Fang W, Zheng J, Peng Z, Yu B, Chen C et al. Whole-transcriptome defines novel glucose metabolic subtypes in colorectal cancer. J Cell Mol Med. 2023;28(5):e18065.

  50. Caligiuri G, Tuveson DA. Activated fibroblasts in cancer: perspectives and challenges. Cancer Cell. 2023;41:434–49.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Mossmann D, Muller C, Park S, Ryback B, Colombi M, Ritter N, et al. Arginine reprograms metabolism in liver cancer via RBM39. Cell. 2023;186:5068–e8323.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. Zhu HE, Yin JY, Chen DX, He S, Chen H. Agmatinase promotes the lung adenocarcinoma tumorigenesis by activating the NO-MAPKs-PI3K/Akt pathway. Cell Death Dis. 2019;10(11):854.

Download references

Acknowledgements

The authors would like to thank the researchers who provided open access to the raw data.

Funding

This project was supported by grants from the National Key Research and Development Plan (2022YFC3401000), National Natural Science Foundation of China (82022037, 82203152, 82303614), Guangdong Basic and Applied Basic Research Foundation (2021B1515230009, 2022A1515111079), Key Research and Development Plan of Guangdong Province (2020B0101030006) and China Postdoctoral Science Foundation (2021M703734, 2022T150760, 2023T160750).

Author information

Author notes

  1. Youpeng Wang, Xingfeng Qiu and Qinghai Li contributed equally to this work.

Authors and Affiliations

  1. Center of Hepato-Pancreato-Biliary Surgery, The First Affiliated Hospital, Sun Yat-sen University, 58 Zhongshan 2nd Road, Guangzhou, Guangdong, 510080, China

    Youpeng Wang & Qi Zhou

  2. Department of General Surgery, Hui Ya Hospital of The First Affiliated Hospital, Sun Yat-sen University, Huizhou, Guangdong, 516081, China

    Qi Zhou

  3. Department of Gastrointestinal Surgery, The First Affiliated Hospital, Sun Yat-sen University, 58 Zhongshan 2nd Road, Guangzhou, Guangdong, 510080, China

    Qinghai Li, Jiale Qin, Xiang Zhang, Xiangqiong Wen, Weiling He & Yuqin Di

  4. Department of Biochemistry, Zhongshan School of Medicine, Sun Yat-sen University, Guangzhou, Guangdong, 510080, China

    Lvlan Ye

  5. Department of Gastrointestinal Surgery, School of Medicine, Xiang’an Hospital of Xiamen University, Xiamen University, Xiamen, Fujian, 361000, China

    Xingfeng Qiu, Xingxiang Huang & Weiling He

  6. Molecular Diagnosis and Gene Testing Center, The First Affiliated Hospital, Sun Yat-sen University, 58 Zhongshan 2nd Road, Guangzhou, Guangdong, 510080, China

    Yuqin Di

  7. Center for Translational Medicine, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, Guangdong, 510080, China

    Ziyang Wang

Authors
  1. Youpeng Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Xingfeng Qiu
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Qinghai Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Jiale Qin
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Lvlan Ye
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Xiang Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Xingxiang Huang
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Xiangqiong Wen
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Ziyang Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Weiling He
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Yuqin Di
    View author publications

    You can also search for this author inPubMed Google Scholar

  12. Qi Zhou
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Youpeng Wang: Writing – original draft, review & editing, Visualization, Methodology. Jiale Qin, Qinghai Li, and Xingfeng Qiu: Data curation, Methodology, Project administration. Lvlan Ye, Xiang Zhang, Xingxiang Huang and Ziyang Wang: Writing – review & editing, Methodology. Qi Zhou: Writing – review & editing, Methodology. Yuqin Di and Weiling He: Writing – review & editing, Methodology, Project administration.

Corresponding authors

Correspondence to Weiling He, Yuqin Di or Qi Zhou.

Ethics declarations

Ethics approval and consent to participation

The experimental procedures were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (SYSU-IACUC-2024-001000). The study was approved by the Ethics Committee of The First Affiliated Hospital, Sun Yat-sen University (Ethical Approval No.2024 − 656). Since this study has no potential harm to the included cases, we applied for exemption from the informed consent process.

Consent for publication

All authors agree to publish this study.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Additional information

Publisher’s note

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

Electronic supplementary material

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

Wang, Y., Qiu, X., Li, Q. et al. Single-cell and spatial-resolved profiling reveals cancer-associated fibroblast heterogeneity in colorectal cancer metabolic subtypes. J Transl Med 23, 175 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06103-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12967-025-06103-3

Keywords

Journal of Translational Medicine

ISSN: 1479-5876

Contact us