Skip to main content
Download PDF

Integrating spatial and single-cell transcriptomes reveals the role of COL1A2(+) MMP1(+/-) cancer-associated fibroblasts in ER-positive breast cancer

Cancer Cell International volume 25, Article number: 82 (2025) Cite this article

Abstract

Cancer-associated fibroblasts (CAFs) are highly heterogeneous cells and important components of the breast tumor microenvironment (TME). However, their role and clinical value in ER-positive breast cancer have not been fully clarified. Our study aims to comprehensively characterize the heterogeneity, potential biological functions, and molecular mechanisms of CAFs in ER-positive breast cancer within the tumor microenvironment using multi-omics data, to provide new strategies for the diagnosis and treatment of ER-positive breast cancer patients. In this study, we found that COL1A2(+) MMP1(+) and COL1A2(+) MMP1(-) CAFs were associated with unfavorable prognosis. The dynamic evolution and cell-cell communications of CAFs were analyzed, revealing that COL1A2(+) MMP1(+/-) CAFs show extensive crosstalk with tumor-associated macrophages (TAMs), contributing to an immunosuppressive TME. Moreover, the somatic mutation of TP53 may be a potential indicator for evaluating the infiltration of COL1A2(+) MMP1(+/-) CAFs. Finally, an MRI-based radiomic model was constructed to estimate the abundance of these CAFs. In conclusion, our findings provide a theoretical basis for targeting CAFs and offer a noninvasive approach to evaluate the infiltration level of COL1A2(+) MMP1(+/-) CAFs.

Introduction

Breast cancer has become one of the most common tumors among women globally for decades, posing a serious threat to health and life quality [1,2,3]. As a major type of breast cancer, ER-positive breast cancer (ER+, PR+/-, HER-) accounts for about 70% of all breast cancers [4]. While this type of breast cancer is less aggressive compared to HER2-positive breast cancer (HER2+, ER+/-, PR+/-) or triple-negative breast cancer (ER-, PR-, HER2-), ER-positive breast cancer carries a higher long-term risk of recurrence than these subtypes [5]. Therefore, identifying new therapeutic targets and novel effective prognostic biomarkers is crucial for improving the long-term prognosis of breast cancer patients.

With the accumulating understanding of tumors, the tumor microenvironment has emerged as a key focus for the future of personalized cancer therapy [6,7,8]. In general, the tumor microenvironment (TME) is a complex and heterogeneous ecosystem composed of tumor cells, immune cells, and stromal cells [9]. Interactions among these components contribute to tumor progression, drug resistance, recurrence, and metastasis [10]. As a special component in TME, cancer-associated fibroblasts (CAFs) have been demonstrated to promote the progression of multiple cancers [11,12,13] via plentiful approaches including suppressing the antitumor function of immune cells and remodeling the extracellular matrix, etc [14, 15]. The tumor-promoting role of CAFs has also been observed in breast cancer. Sebastian et al. [16] identified three types of CAFs (myofibroblastic CAFs, ‘inflammatory’ CAFs, and antigen-presenting CAFs) using a syngeneic mouse model and single-cell RNA sequencing, and provided novel ideas for TNBC treatment. Wu et al. [17] reported the heterogeneity of CAFs in TNBC and identified several potential prognostic CAF biomarkers, including CCR7, CD2, SASH3, and TAGAP. Besides, in ER-positive breast cancer, CAFs have been found to promote tumor progression through various mechanisms. This underscores the complexity of CAF roles in ER-positive breast cancer, necessitating further exploration of their interactions with tumor cells and other TME components [18, 19]. However, the role of CAFs in ER-positive breast cancer remains largely underexplored, highlighting an important direction for future research.

The rapid development of sequencing technology has enabled tumor microenvironment analysis at the single-cell level. Compared to traditional transcriptome sequencing, single-cell sequencing significantly enhances transcriptome resolution, providing technical support for understanding cellular-level heterogeneity [8]. However, this technique has limitations. The lack of spatial information restricts oncologists’ ability to study the spatio-temporal dynamics of tumor development. Fortunately, the advent of spatial transcriptomics has addressed this limitation, enabling the capture of spatially resolved cellular information [20]. Combining these two sequencing techniques allows for a comprehensive understanding of the spatio-temporal changes in each component of the tumor microenvironment. Additionally, imaging techniques, such as MRI, CT, and PET, have become essential tools for predicting treatment response, tumor molecular subtype, and prognosis in cancer patients [21]. Therefore, integrating multiple omics and radiomics provides a broader perspective for cancer research, facilitating the identification of novel diagnostic and therapeutic markers.

Herein, we integrated abundant data including single-cell transcriptomics, spatial transcriptomics, radiomics, somatic mutation profiles, and traditional transcriptomics to reveal the heterogeneity of CAFs. Our findings comprehensively analyzed the dynamic evolution of CAFs and confirmed that COL1A2(+) MMP1(+/-) CAFs are associated with poor prognosis in ER-positive breast cancer. Moreover, extensive cell-cell communications between COL1A2(+) MMP1(+/-) CAFs and tumor-associated macrophages (TAMs) were identified by single-cell transcriptomics, and their colocalization was validated by spatial transcriptomics. Finally, we developed a radiomic model for estimating the abundance of COL1A2(+) MMP1(+/-) CAFs. Our study provides a theoretical basis for targeting CAFs and offers a noninvasive approach to evaluate the infiltration level of COL1A2(+) MMP1(+/-) CAFs.

Materials and methods

Data collection

The overall flow chart of this study is illustrated in Fig. 1. This study included abundant data including single-cell RNA profiles, spatial transcriptome data, radiomics, bulk-RNA data, as well as genomic data. Concretely, the expression profiles, clinical outcome, and genomic profiles of TCGA-BRCA dataset were downloaded by ‘TCGAbiolinks’ (v2.26.0) [22] package. The spatial transcriptome data and single-cell RNA data were obtained from GSE243022 and GSE176078 on GEO database, respectively. The radiomic data of TCGA-BRCA patients were acquired by TCIA database (https://dev.cancerimagingarchive.net/).

Fig. 1
figure 1

The overall design of the study. TCGA, The Cancer Genome Atlas; TCIA, The Cancer Imaging Archive; scRNA-seq, single-cell RNA sequencing; GSE, Gene Expression Omnibus series; ST, spatial transcriptome; DCE, Dynamic Contrast Enhancement

Full size image

Data inclusion and preprocessing

All data collected included only ER-positive patients, excluding patients with triple-negative breast cancer and HER2-positive breast cancer. The TCGA-BRCA patients without complete overall survival information were excluded. Finally, a total of 486 ER-positive patients were included in the study, including 474 patients in TCGA-BRCA dataset, 11 ER-positive breast cancer patients in GSE176078, and 1 ER-positive breast cancer patient in GSE243022.

The expression profiles of TCGA-BRCA patients were transformed into TPM to reduce the effects of sequencing depth and gene length. Seurat (v4.3.0) package was used to performed the quality control for single-cell RNA profiles. Low-quality cells were removed through the following criteria: (1) nFeature_RNA < 250; (2) nCount_RNA < 500; (3) The percent of mitochondria genes > 20%. ‘DoubletFinder’ (v2.0.3) package was conduct to remove potential doublet. Batch effect among samples were removed by ‘harmony’ package. The standardization of data was performed by ‘SCTransform’ (v2). The top 3000 highly variable genes were obtained for resolution selection. Cell clustering and resolution were aided by the ‘Clustree’ (v0.5.0) package. ‘FindAllMarkers’ function in Seurat package was conducted to identify the overexpressed genes of each cluster.

Cell annotation

The cell markers used for annotation were collected by public research and online database [23, 24]. Generally, the principle of cell annotation is to annotate the cell subsets of the major category first. For cell annotation of small categories, large subgroups were extracted first, re-standardized and dimensionality reduced, and then annotated by cell marker. It should be noted that the first clustering of single cells will inevitably cause a small number of cells to be mixed. Therefore, for cells that appear to be mixed in other categories, we re-annotate them through cell markers as their real cell categories in the second annotation.

Digital cytometry

The single-cell expression profiles of cancer-associated fibroblasts were used as single-cell reference matrix to infer the relative cell portion in TCGA-BRCA patients by CIBERSORTx website (https://cibersortx.stanford.edu/index.php). The absolute mode was selected.

Survival analysis

The ‘survminer’ (v0.4.9) package was conduct to the overall survival of patients. The infiltration level of distinct CAFs were estimated by the best cutoff value. P-value < 0.05 was considered as significant.

Enrichment analyses

The gene set enrichment analysis was performed by ‘RunGSEA’ function in ‘SCP’ (v0.4.7.9) package based on gene ontology term (biological function). The HALLMARK enrichment analysis was analyzed by the ‘GSVA’ (v1.46.0) and ‘clusterProfiler’ (v4.6.2) package.

Cell trajectory and developmental analyses

The cell trajectory analysis was performed by ‘SCP’ (v0.4.7.9) package. The slingshot method was used to infer the dynamic gene expression and biological function change among distinct CAFs. Furthermore, the ‘Vector’ [25] package was conduct to analyze the cell developmental direction.

Cell-cell communication

The cell-cell crosstalk between CAFs and other cell types was calculated by the ‘CellChat’ [26] package (v1.6.1). The receptor-ligand pairs derived from ‘Secreted Signaling’ and ‘ECM-Receptor’ were included for analyses. Cell type less than 10 was excluded for the analyses. P-value < 0.05 was thought as significant.

Spatial transcriptome analysis

The downstream analysis of spatial data was conducted by Seurat (v4.3.0). The filtered spatial matrix was standardized by the ‘SCTransform’. The regions were first identified by dimensionality reduction and then confirmed by two experienced pathologists. ‘clusterProfiler’ package was analyzed the differential pathways between tumor region and non-tumor region.

The single-cell matrix we previously annotated was used as reference matrix. ‘MuSiC’ (v1.0.0) and ‘CRAD’ (v1.1) package were used to infer the cell types in each spatial spot. The spatially co-localizing cell types were analyzed by ’mistyR’ (v1.10.0).

Genomic analysis

The high- and low- groups of distinct CAFs were previous described in ‘Survival analysis’ section. The somatic mutation data was analyzed by ‘maftools’ [27] package. Fisher Exact test was used to detect the significant somatic mutations between high- and low- group. P < 0.05 was considered to be significant.

Radiomics analysis

A total of 137 MRI images of TCGA-BRCA patients were first obtained. Subsequently, strict standards are used to further obtain high-quality data: (1) ER-positive patients; (2) Patients with expression profiles. (3) Standard double breast coil on a 1.5T GE whole-body MRI systems were used with T1-weighted DCE-MRIs. Finally, 61 MRI data of ER-positive patients were included for further analysis.

The ’3D Slicer’ software was used to perform fully manual segmentation of the tumors based on DCE-T1 images. Concretely, two experienced radiologists who specialized in breast cancer at least ten years performed segmentation of the tumors together. A more experienced radiologist made a final decision on any disagreement. The plentiful features of images were extracted, including ‘shape’, ‘ngtdm’, ‘glszm’, ‘glrlm’, ‘gldm’, ‘glcm’, ‘fisrtorder’ as well as ‘wavelet-based feature’. Also, all the data were resampled voxel size to make the spatial information of the different images match.

The patients were divided into training and validation sets in a 7:3 ratio. Both features and infiltration score were standardized by Z-score. Pearson correlation analysis was performed, and the 5 candidate features were identified (P < 0.05). The linear model was established by The Least Absolute Shrinkage and Selection Operator. The alpha was set as 1. Finally, the valuable features were selected to construct the radiomic model. The correlation relationship between radiomic score and the infiltration score of CAFs was estimated by Pearson correlation. P < 0.05 was considered as significant.

Immune phenotype analysis

Immune phenotype-related markers were collected by public research [28]. The single sample gene enrichment analyses were performed to calculated the enrichment score of three distinct immune phenotypes (inflamed, excluded, desert). The spearman correlation analyses were utilized to evaluate the correlation between the abundance of COL1A2(+) MMP1(+/-) CAFs and enrichment score of immune phenotypes.

Statistical analyses

All the statistical analyses in this study were conducted by R (v4.2.2). The statistical methods and related details were described in each section of methods.

Results

First annotation of cell types in ER-positive breast cancers

To obtain the high-quality single cell, a total of 11 single-cell RNA profiles of ER-positive breast cancers in GSE176078 were standardized. Low-quality cells and potential doublet were removed by strict procedures and ‘DoubletFinder’ package (Supplementary File: Figure S1 a-c). Subsequently, ‘harmony’ package was used to reduce the batch effect among samples, and the batch effect was significantly removed (Supplementary File: Figure S2 a-b). Finally, a total of 36,413 single cells were identified and were first grouped into 28 distinct cell clusters with the resolutions of 0.9 (Fig. 2a). Further, all cells have been annotated with public research and online database. As shown in Fig. 2b, B cells/plasma, cycling cells, endothelial cells, pericytes, epithelial cells, smooth muscle cells, myeloid cells, T cells, and cancer-associated fibroblasts were identified. Uniform manifold approximation and projection confirmed the distinct cell types were separated. The markers for cell annotation were presented in Fig. 2c and Figure S3. For instance, EPCAM was used to identify epithelial cells, and PECAM1 was the marker for endothelial cells. All markers for the first cell annotation could be found in supplementary File (Table S1). In summary, nine main cell types in ER-positive breast cancers were identified for the subsequent research.

Fig. 2
figure 2

First cell annotation in ER-positive breast cancer through single-cell RNA sequencing. a-b The cell clusters (a) and cell types (b) in ER-positive breast cancer tissue demonstrated using the uniform manifold approximation and projection (UMAP); c The density of cells with specific marker were illustrated using UMAP. The color represents the density level of the cells with the marker

Full size image

Identification of four distinct CAFs

A total of 2256 CAFs were extracted for re-standardized in GSE176078, and dimensionality reduction. As illustrated in Fig. 3a-b, four types of CAFs were identified, including SOD3(+) CAFs, COL1A2(+) MMP1(-) CAFs, RPS17(+) CAFs, and COL1A2(+) MMP1(+) CAFs (N = 69). The top 10 genes of each type were identified (P < 0.05, log2FC > 0.5, Table S2). The markers for CAFs annotation were presented in Fig. 3c-d. Interestingly, we noticed COL1A2 was highly expressed in both COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs, whereas MMP1 was only highly expressed in COL1A2(+) MMP1(+) CAFs, indicating MMP1 could be used as a significant biomarker to distinguish these two types of CAFs. Meanwhile, the expression of MMP1 was associated with the overall survival in ER-positive breast cancer patients (P = 0.0012; Figure S4). The characteristics of CAFs need to be investigated to depict their heterogeneity.

Fig. 3
figure 3

Identification of four subtypes of CAFs with distinct clinical outcome. a-b The distinct clusters (a) and types (b) of CAFs were identified in ER-positive breast cancer tissue visualized by UMAP; c The density of CAFs with specific marker were illustrated using UMAP. The color represents the density level of the cells with the marker.; d The markers used for annotation of CAFs. The size of the dot represents the proportion of cells expressing the gene in the corresponding cell cluster. The color of the dot represents the average expression level of the gene in the cell cluster; e-f The K-M curves showing the overall survival rate of high- and low- SOD3(+), COL1A2(+) MMP1(-), RPS17(+) and COL1A2(+) MMP1(+) CAFs

Full size image

CAFs have been confirmed to be a critical role in the prognosis of tumor patients [15]. Therefore, the identification of CAFs related to poor prognosis is essential for prognostic stratification. Digital cytometry was utilized to infer the relative cell abundance of four subgroups of CAFs in TCGA-BRCA. As shown in Fig. 3e-h, high infiltration levels of COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs were associated with poor prognosis in ER-positive patients (PCOL1A2=0.048, PMMP1=0.034), while the abundance of other types of CAFs presented no significant prognostic value (PSOD3=0.18, PRPS17=0.034). These findings suggested that COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs may play an important role in tumor progression.

To further investigate the biological function of distinct types of CAFs, we performed the enrichment analyses based on ‘GO-BP’ and ‘HALLMARKERS’ gene sets in GSE176078. As shown in Fig. 4a, we observed a similar biological function between COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs, while RPS17(+) CAFs and SOD3(+) CAFs exhibited a similar biological function. For instance, the enrichment scores of extracellular matrix organization, cell adhesion as well as collagen fibril organization, which had been confirmed to be associated with immune suppressive microenvironment [29], were presented significantly higher enriched in COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs. These results indicated these two types of CAFs may contribute to tumor development by remodeling the tumor microenvironment. On the side, the enrichment scores of metabolic-related biological processes were highly enriched in RPS17(+) CAFs and SOD3(+) CAFs, suggesting that these two CAFs may regulate their metabolic status to adapt to the tumor microenvironment and support tumor growth by secreting related metabolites. In addition, the malignant pathways epithelial-mesenchymal transition, WNT beta-catenin, and TGF-beta signaling pathways were highly enriched in COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs, revealing the complicated and plentiful approaches of CAFs to promote tumor progression (Fig. 4b). Taken together, these findings preliminarily reveal the heterogeneity of CAFs among patients with ER-positive breast cancer, including the potential biological function and clinical prognostic value of different subgroups of CAFs.

Fig. 4
figure 4

The cell trajectory and heterogeneity of CAFs in tumor microenvironment. a The gene set enrichment analysis showing the biologic function of distinct CAFs. NES: normalized enrichment score; b The hallmark gene set enrichment analysis showing the enriched pathways in distinct CAFs; c The potential cell trajectory and lineages of CAFs in tumor microenvironment; d Two cell lineages of CAFs with the pseudo-time development; e The dynamic cell trajectory analysis confirming the changes of biological function and markers during the evolution of CAFs; f The cell developmental analysis showing the direction of COL1A2(+) MMP1(-) CAFs to COL1A2(+) MMP1(+) CAFs

Full size image

COL1A2 + MMP1-CAFs and COL1A2 + MMP1 + CAFs exert tumor-promoting effects in TME via the intrinsic evolution of CAFs

As a special type of highly phenotypic plastic cell, understanding the intrinsic dynamic lineage of CAFs is essential to characterize its complicated role in TME. Cell trajectory analysis indicated two potential cell lineages existed in CAFs in GSE176078 (Fig. 4c). Concretely, the RPS17(+) CAFs were at the beginning of the cell trajectory, while COL1A2(+) MMP1(-) CAFs (lineage 1) and COL1A2(+) MMP1(+) CAFs (lineage 2) were at the end. SOD3(+) CAFs may serve as an intermediate state for CAFs. About 93.44% of CAFs were involved in lineage 1, and 62.94% of CAFs were involved in lineage 2 (Fig. 4d). Of note, at the end of the trajectory, our previous results confirmed the poor prognostic value of COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs. RPS17(+) CAFs, at the beginning of trajectory, presented a suggestive favorable prognostic value (Fig. 4f, P = 0.066). Our findings implied that dynamic changes in CAFs subpopulations due to intrinsic evolution may lead to different prognostic outcomes for patients with ER-positive breast cancer.

Further, we investigate the dynamic gene expression change during the development of cell trajectory to explore the important differentiation switch genes and their pathways. As illustrated in Fig. 4e, the expression of SERPINE1, ASPN, COL3A1, and SPARC were specifically highly expressed in COL1A2(+) MMP1(-) CAFs, while ISG15 and TMEM158 were highly expressed in COL1A2(+) MMP1(+) CAFs. Consistently, FN1, SDC1, SULF1, CTHRC1 were highly expressed in both COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs. Furthermore, we noticed COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs were mainly involved in ECM-related biological function. However, COL1A2 + MMP1 + CAFs also participated in some immune-related biological processes, including IL-10, negative regulation of immune system, indicating it may play a critical role in the suppressive immune microenvironment. Based on previous GSEA (Fig. 4a) and Pseudo-time analyses, we hypothesized that the COL1A2(+) MMP1(+) CAFs subgroup may have diverged from the COL1A2 + MMP1-CAFs subgroup. Therefore, cell development analysis was conducted to validate our hypothesis. As shown in Fig. 4f, COL1A2(+) MMP1(-) CAFs showed a tendency to develop into COL1A2(+) MMP1(+) CAFs. Taken together, our results reveal the intrinsic lineage evolution of CAFs and the accompanying dynamic gene expression profile changes. These genes may become diagnostic markers for potential new CAFs and potential therapeutic targets.

The crosstalk between CAFs and other cell types in TME

Cell-cell communication in TME is an important approach to tumor progression. Therefore, we next investigate the crosstalk between CAFs and other cell types in GSE176078. T cells and myeloid cells were regrouped and annotated. Other cell types mixed in the two groups were reverted to their true cell types, and cells without specific markers were annotated as unknown. Finally, all cell types in ER-positive breast cancer were annotated (Fig. 5a). The markers we used were illustrated in Fig. 5b. Further, the correlation analysis was performed to validate the accuracy of cell annotation. As shown in Fig. 5c, epithelial cells, stromal cells, immune cells were obviously separated.

Fig. 5
figure 5

The cell-cell communications between CAFs and other components in tumor microenvironment. a UMAP-based dimensionality reduction map colored by cell cluster annotations; b The marker used for cell annotations. The size of the dot represents the proportion of cells expressing the gene in the corresponding cell cluster. The color of the dot represents the average expression level of the gene in the cell cluster. pct.exp: percentage of expression; c The heatmap showing the correlation among cell types. The color represents the correlation coefficient; d-g The pathway network of COLLAGEN (d), FN1(e), LAMININ(f), THBS (g) in tumor microenvironment. The thickness of the lines represents the strength of the communication; h Dot plot showing ECM-related signaling from CAFs to TAMs and Tregs. Colors represent the probability of communication. commun.prob: communication probability; i Dot plot showing ECM-related signaling from TAMs and Tregs to CAFs; j Dot plot showing secreted signaling communications from CAFs to endothelial cells

Full size image

We first analyzed the ECM-related receptor-ligand pairs in TME. ECM signaling communications including collagen signaling pathway (Fig. 5d), FN1 signaling pathway (Fig. 5e), LAMIN signaling pathway (Fig. 5f), and THBS signaling pathway (Fig. 5g) were significantly observed in COL1A2 + MMP1-CAFs and COL1A2 + MMP1 + CAFs, which was consistent with previous results. Specifically, Receptor-ligand pairs such as COL1A1-CD44, COL1A2-CD44, and FN1-CD44 showed a high communication probability between COL1A2(+) MMP1(-) CAFs and tumor-associated macrophages (TAMs, Fig. 5h). The same phenomenon has been observed between CAFs and regulatory T cells (Treg, Fig. 5h). Additionally, COL1A2 + MMP1-CAFs may communicate specifically with TAM and Treg via THBS4-CD47 and COMP-CD47. Also, COL1A2(+) MMP1(+) CAFs may communicate specifically with TAM and Treg via THBS4-CD47 and COMP-CD47. COL1A2(+) MMP1(+) CAFs may specifically interact with TAM and Treg via THBS3-CD47. On the other hand, we observed the TAM could also interact COL1A2(+) MMP1(-) CAFs and COL1A2 (+) MMP1(+) CAFs via FN1 signaling (Fig. 5i). This suggested COL1A2(+) MMP1(+/-) CAFs may interact with TAM to form feedback that continuously promotes tumor progression. Next, we explored the interaction between CAFs and endothelial cells. As shown in Fig. 5j, the VEGF signaling was significantly observed in COL1A2(+) MMP1(+/-) CAFs and endothelial, suggesting they may promote angiogenesis in TME. Notably, chemokine-related signals were observed in CAFs (RPS17(+) CAFs, and COL1A2(+) MMP1(+) CAFs) and endothelial cells, indicating that they could regulate endothelial cell migration and reshape TME. All in all, our results reveal the potential role of CAFs in TME, providing therapeutic targets for targeting the communication between CAFs and TAM, Treg, and endothelial cells.

COL1A2 + MMP1+/- CAFs have colocalization relationship with TAM

The crosstalk between CAFs and TAMs was preliminarily confirmed by our results. Next, we investigated their spatial relationship using spatial transcriptomics in ER-positive breast cancer in GSE243022. The spots on HE-stained sections were grouped into five distinct regions (Fig. 6a). We further annotated these regions as tumor-rich and tumor-scarce areas with assistance from experienced pathologists (Fig. 6b). To confirm the accuracy of these annotations, we conducted differential gene expression analysis between the regions, followed by enrichment analysis (Fig. 6c-d). As anticipated, malignant pathways such as the cell cycle and estrogen signaling pathways were significantly enriched in the tumor-rich region, whereas ECM-receptor interaction and focal adhesion pathways were enriched in the tumor-scarce region. Subsequently, single-cell RNA sequencing data was used as a reference to infer cell types in the spatial matrix. We then calculated the importance scores for cell types associated with the spatial distribution of COL1A2(+) MMP1(-) CAFs and COL1A2(+) MMP1(+) CAFs. As shown in Fig. 6e, the most critical cell types associated with COL1A2(+) MMP1(-) CAFs distribution were COL1A2(+) MMP1(+) CAFs, Tregs, and TAMs. Similarly, the cell types most closely linked to COL1A2(+) MMP1(+) CAFs distribution included COL1A2 + MMP1- CAFs, TAMs, and pDCs (Fig. 6f). The relative spatial distributions of COL1A2(+) MMP1(-) CAFs, Tregs, TAMs, COL1A2(+) MMP1(+) CAFs, and pDCs are illustrated in Fig. 6g-k. These findings indicate a close spatial association between the two CAF subtypes. In summary, our results suggest that COL1A2(+) MMP1(+/-) CAFs and TAMs are co-localized within the tumor microenvironment, supporting our earlier cell communication analysis.

Fig. 6
figure 6

Exploration of the colocalization relation by spatial transcriptome. a UMAP confirmed the five distinct regions of spots; b Tumor region was identified though dimensionality reduction and experienced pathologists; c Heatmap showing the differential-expression genes between tumor-rich and tumor-scarce tissues; d Dot plot showing the enriched pathways in tumor-rich and tumor-scarce tissues; e The chart plot showing the importance score of cell types related to the distribution of COL1A2(+) MMP1(-) CAFs; f The chart plot showing the importance score of cell types related to the distribution of COL1A2(+) MMP1(-) CAFs; g-k The relative cell location of COL1A2(+) MMP1(-) CAFs (g), Treg (h), TAM (i), COL1A2(+) MMP1(+) CAFs (j), and pDC(k) in spatial position on the pathological section

Full size image

Somatic mutations related to the infiltration level of COL1A2 + MMP1-CAFs and COL1A2 + MMP1 + CAFs

Next, we intended to explore the factors which influenced the infiltration of CAFs in TCGA-BRCA. We firstly analyzed the correlation between clinical characteristics and CAFs infiltration. As shown in Tables S3 and S4, the races and tumor stage showed no significant correlation with the abundance of COL1A2 + MMP1-CAFs (P = 0.538, P = 0.469) and COL1A2 + MMP1 + CAFs (P = 0.276, P = 0.936). Interestingly, somatic mutations have been demonstrated to play a critical role in the infiltration of cells [30]. Therefore, we hypothesized that somatic mutations may influence the infiltration of CAFs. To identify the most important somatic mutations related to the infiltration of COL1A2(+) MMP1(+/-) CAFs, we grouped the patients into high and low COL1A2(+) MMP1(-) CAFs groups as previously described. We observed that the somatic mutations of TP53 (P < 0.001) and ADGRG4 (P < 0.01) were closely associated with the low COL1A2(+) MMP1(-) CAFs group (Fig. 7a). Interestingly, a high level of co-occurrence and mutually exclusive patterns was observed in the high COL1A2(+) MMP1(-) CAFs group (Fig. 7b), whereas no significant patterns were noted in the low COL1A2(+) MMP1(-) CAFs group (Fig. 7c).

Fig. 7
figure 7

The somatic mutations were associated with the infiltration of COL1A2(+) MMP1(+/-) CAFs. a Forest plot demonstrating focal somatic single nucleotide variant (SNV) is significantly different between patients in the COL1A2(+) MMP1(-)-CAFs high and low infiltration level groups. OR > 1, higher SNV frequency for patients in the high infiltration level group. OR < 1, higher SNV frequency in low infiltration level group; b-c Landscape of gene co-mutations in patients with high (b) and low (c) abundance of COL1A2(+) MMP1(-) CAFs; d Forest plot demonstrating focal somatic single nucleotide variant (SNV) is significantly different between patients in the COL1A2(+) MMP1(+)-CAFs high and low infiltration level groups; e-f Landscape of gene co-mutations in patients with high (e) and low (f) abundance of COL1A2(+) MMP1(+) CAFs

Full size image

We applied the same method to classify patients based on the infiltration of COL1A2(+) MMP1(+) CAFs. As shown in Fig. 7d, high infiltration of COL1A2(+) MMP1(+) CAFs was associated with somatic mutations in TP53 (P < 0.001) and APOB (P < 0.01), while low abundance of COL1A2(+) MMP1(+) CAFs correlated with somatic mutations in CDH1 (P < 0.001) and PIK3CA (P < 0.01). Additionally, co-occurrence and mutually exclusive patterns were observed in the high COL1A2(+) MMP1(+) CAFs group, but no significant patterns were found in the low COL1A2(+) MMP1(+) CAFs group (Fig. 7e-f). In summary, these findings highlight the complex relationship between somatic mutations and the infiltration of two types of poor prognostic CAFs, which may provide novel insights for predicting CAF infiltration.

Establishment of a radiomic model for predicting the infiltration of CAFs in ER-positive breast cancer patients

The above results suggested the prognostic value of COL1A2(+) MMP1(+/-) CAFs. Interestingly, CAFs could shape inflamed, excluded, or desert immune phenotypes in TME [28]. The correlation between the infiltration of COL1A2(+) MMP1(+/-) CAFs and reported immune phenotypes showed that COL1A2(+) MMP1(+/-) CAFs was highly related to excluded and desert immune phenotypes (Figure S5A-C). Thus, we aimed to establish a correlation between the infiltration of COL1A2(+) MMP1(+/-) CAFs and non-invasive detection. Radiomic profiles were derived from TCGA-BRCA patients in TCIA database, whose transcriptome profiles were previously described. A total of 61 ER-positive patients with dynamic contrast enhanced (DCE)-T1 sequences were grouped into a training cohort (N = 43) and a validation cohort (N = 18). The workflow was illustrated in Fig. 8a. The tumor region was manually segmented by two experienced radiologists together (Fig. 8b). Subsequently, 851 features were extracted and the Pearson correlation analysis was performed to identify features significantly related to the infiltration of COL1A2(+) MMP1(+/-) CAFs. And five valuable features were identified (P < 0.05). LASSO algorithm was applied and three features were obtained for constructing a linear regression model (Fig. 8c-d). Finally, the radiomic score was calculated as followed: Radiomic score = 0.228379423* original.shape.Sphericity- 0.228999483*original.firstorder.10Percentile − 0.011557297*wavelet-LHL.firstorder.Median. The Pearson correlation analyses showed the radiomic score was positively correlated with the abundance of COL1A2(+) MMP1(+/-) CAFs in both training cohort (P = 0.035, R = 0.5, Fig. 8e) and validation cohort (P = 0.023, R = 0.35, Fig. 8f). Taken together, our findings demonstrated that the macroscopic effects of COL1A2(+) MMP1(+/-) CAFs could be observed by radiomics. Furthermore, the radiomic model we constructed could serve as a promising tool for evaluating the abundance of COL1A2(+) MMP1(+/-) CAFs.

Fig. 8
figure 8

The radiomic model based in MRI was constructed for predicting the abundance of COL1A2(+) MMP1(+/-) CAFs in ER-positive breast cancer. a Schematic diagram of the radiomics analysis process; b An example of segmentation for gross tumor volume on DCE-T1 images; c Distribution of coefficients for variables in the LASSO regression; d Parameter tuning plot in least absolute shrinkage and selection operator (LASSO) regression; e-f Pearson’s correlations between z-score normalized abundance of COL1A2(+) MMP1(+/-) CAFs measured by the transcriptome and the fitted value by the linear regression radiomics model in the training cohort (e) and validation cohort (f)

Full size image

Discussion

In this study, we identified four subtypes of CAFs in ER-positive breast cancer through single-cell RNA sequencing. Among them, COL1A2(+) MMP1(-) and COL1A2(+) MMP1(+) CAFs were associated with poor prognosis and exhibited highly similar biological functions within the TME. We also explored the dynamic evolution of CAFs to reveal the characteristics of each subtype. Cell-cell communication and spatial transcriptomics suggested critical crosstalk between COL1A2(+) MMP1(+/-) CAFs and TAMs. Additionally, we analyzed potential somatic mutations linked to the infiltration of these CAFs. Finally, a radiomic model was established to estimate the abundance of target CAFs. In previous studies, researchers utilized spatial techniques to identify CAF subtypes, with certain subtypes associated with tumor prognosis [31, 32]. Unlike them, we integrated spatial techniques with other methods to identify CAFs related to patient prognosis. To our knowledge, this is the first study to characterize the role of CAFs in ER-positive breast cancer using multi-omics strategies, including spatial transcriptomics and radiomics. Our research provides pioneering insights into the pro-tumor effects and potential clinical applications of CAFs in ER-positive breast cancer.

The role of CAFs in tumors has remained controversial [33]. For example, Hutton et al. [34] identified CD105-negative CAFs in both human and murine tissues and confirmed their antitumor role in pancreatic cancer. Additionally, tertiary lymphoid structures, correlated with a favorable prognosis for cancer patients, were shown to be driven by a specific type of CAF [35]. Conversely, increasing evidence supports the pro-tumor role of CAFs. Li et al. [36] reported that inflammatory and extracellular matrix CAFs could promote gastric cancer development by interacting with T cells and M2 macrophages, respectively. The above research highlights the high heterogeneity of CAFs, emphasizing the need for a systematic exploration of CAFs in breast cancer. In our study, we identified four types of CAFs: SOD3(+) CAFs, RPS17(+) CAFs, COL1A2(+) MMP1(-) CAFs, and COL1A2(+) MMP1(+) CAFs. RPS17(+) CAFs were found to evolve into COL1A2(+) MMP1(-) or COL1A2(+) MMP1(+) CAFs via an intermediate stage (SOD3(+) CAFs). Interestingly, survival analyses confirmed the favorable prognostic value of RPS17(+) CAFs and the association of COL1A2(+) MMP1(-) and COL1A2(+) MMP1(+) CAFs with poor prognosis, indicating that CAFs promote tumor progression through intrinsic evolution. Thus, COL1A2(+) MMP1(+/-) CAFs could serve as potential biomarkers for prognostic stratification in ER-positive breast cancer. Upregulation of ECM-related pathways has been reported as a key characteristic of myofibroblastic CAFs(myCAFs) [37, 38]. And COL1A2(+) MMP1(+/-) CAFs showed a high enrichment score in ECM-related pathways, suggesting they could represent myCAFs. Similarly, Croizer et al. [39] found that ECM-myCAFs polarize TREM2(+) macrophages, forming an immunosuppressive microenvironment in breast cancer. Our study also observed extensive cell communication between COL1A2(+) MMP1(+/-) CAFs and TAMs. ECM-related ligand-receptor pairs, such as THBS4/THBS3-CD47, FN1-CD44, and COL1A2/COL1A1-CD44, have been reported to play an immunosuppressive role in the TME [40, 41], implying a potential way in which communication between CAFs and TAMs leads to the formation of immunosuppressive microenvironments. Notably, consistent with previous study, our results showed significant correlation between COL1A2(+) MMP1(+/-) CAFs and excluded, desert immune phenotypes [28]. Furthermore, positive feedback was observed between TAMs and COL1A2(+) MMP1(+/-) CAFs, and colocalization analysis supported this hypothesis. Overall, our study reveals crosstalk between CAFs and TAMs, providing new insights into the formation of immunosuppressive tumor microenvironments. Importantly, the above ligand-receptor pairs may serve as potential therapeutic targets for patients with ER-positive breast cancer.

The origin of CAFs has been confirmed by a large number of published studies [42, 43]. Therefore, we aimed to explore the critical factors related to the infiltration level of CAFs instead of their origin. Li et al. [30] found that the abundance of γδ T cells could be evaluated by somatic mutations as well as mutational patterns. Interestingly, we observed that somatic mutations in TP53 were associated with low infiltration of COL1A2(+) MMP1(-) CAFs. In contrast, a high abundance of COL1A2(+) MMP1(+) CAFs was closely correlated with somatic mutations in TP53. The accumulation of somatic mutations in TP53 may lead to decreased infiltration of COL1A2(+) MMP1(-) CAFs and increased infiltration of COL1A2(+) MMP1(+) CAFs, which may serve as a new indicator affecting CAF infiltration patterns. Consistently, the co-recurrence and mutually exclusive patterns showed the same tendency in both high COL1A2(+) MMP1(-) and high COL1A2(+) MMP1(+) groups. Our study provides a novel insight into estimating CAF abundance and reveals potential links between somatic mutations and the immune microenvironment.

Noninvasive strategies based on radiomics have been developed in multiple cancers [44,45,46]. These successful applications have accelerated the development of radiomics in breast cancer. Yu et al. [47] developed an MRI-based machine learning model to predict axillary lymph node metastasis status and disease-free survival in breast cancer. Shi et al. [48] established an MRI-based radiomic model to predict treatment response to neoadjuvant chemotherapy. However, the precise association between radiomic features and CAF infiltration has not been confirmed. In this study, we identified three valuable radiomic features based on DCE-T1 images using Pearson correlation and the LASSO algorithm. Due to the limited and rare sample size, we opted for a conservative linear model. Surprisingly, we found that the infiltration of COL1A2(+) MMP1(+/-) CAFs could be detected by radiomic tools, such as DCE-T1 images. It is important to note that our aim was not to build an effective predictive model but rather to explore the association of image features with the tumor microenvironment. Taken together, our findings provide new insights and a theoretical basis for the application of imaging omics in the tumor microenvironment, offering new ideas for the diagnosis and treatment of breast cancer patients.

However, our study also has some limitations. Although we combined multi-scale data, the reliability of the analysis should be validated across multiple different types of sequencing data. Due to the limited sample size, the CAF subtypes identified in this study require validation in larger patient cohorts. In vivo and in vitro experiments are needed to further validate our results in subsequent research. For instance, we plan to use multiplex immunohistochemistry to validate the colocalization relationship between CAFs and TAMs. Additionally, more external datasets are needed to illustrate the prevalence of prognostic differences in CAFs. Although we have stated that our purpose was not to establish an effective radiomic model without external datasets, we will collect a large number of MRI-based images to address this issue in the future.

Data availability

All the datasets in this study were obtained from public database including GSE176078(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE176078), GSE243022 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE243022), TCGA database (https://portal.gdc.cancer.gov/), and The Cancer ImagingArchive database (https://dev.cancerimagingarchive.net/). All data generated or analyzed during this study are included in this article/Additional files.

Abbreviations

CAFs:

Cancer-associated fibroblasts

ER:

Estrogen receptor

TME:

Tumor microenvironment

ECM:

Extracellular matrix

TAMs:

Tumor-associated macrophages

Tregs:

Regulatory T cells

DCE:

Dynamic contrast enhanced

TCIA:

The cancer imaging archive

TCGA:

The cancer genome atlas

References

  1. Arnold M, et al. Current and future burden of breast cancer: global statistics for 2020 and 2040. Breast. 2022;66:15–23. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.breast.2022.08.010

    Article  PubMed  PubMed Central  Google Scholar 

  2. Bray F, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024;74:229–63. https://doiorg.publicaciones.saludcastillayleon.es/10.3322/caac.21834

    Article  PubMed  Google Scholar 

  3. Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA Cancer J Clin. 2024;74:12–49. https://doiorg.publicaciones.saludcastillayleon.es/10.3322/caac.21820

    Article  PubMed  Google Scholar 

  4. Nolan E, Lindeman GJ, Visvader JE. Deciphering breast cancer: from biology to the clinic. Cell. 2023;186:1708–28. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.cell.2023.01.040

    Article  PubMed  CAS  Google Scholar 

  5. Pedersen RN, et al. The incidence of breast Cancer recurrence 10–32 years after primary diagnosis. J Natl Cancer Inst. 2022;114:391–9. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/jnci/djab202

    Article  PubMed  CAS  Google Scholar 

  6. Wang Q, et al. Single-cell transcriptome sequencing of B-cell heterogeneity and tertiary lymphoid structure predicts breast cancer prognosis and neoadjuvant therapy efficacy. Clin Transl Med. 2023;13:e1346. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/ctm2.1346

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  7. Tang S, et al. Metabolic heterogeneity and potential immunotherapeutic responses revealed by Single-Cell transcriptomics of breast Cancer. Apoptosis. 2024;29:1466–82. https://doiorg.publicaciones.saludcastillayleon.es/10.1007/s10495-024-01952-7

    Article  PubMed  CAS  Google Scholar 

  8. Wang G, et al. Identification of the tumor metastasis-related tumor subgroups overexpressed NENF in triple-negative breast cancer by single-cell transcriptomics. Cancer Cell Int. 2024;24:319. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12935-024-03505-z

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  9. Harris MA, et al. Towards targeting the breast cancer immune microenvironment. Nat Rev Cancer. 2024;24:554–77. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41568-024-00714-6

    Article  PubMed  CAS  Google Scholar 

  10. Wang X, Almet AA, Nie Q. The promising application of cell-cell interaction analysis in cancer from single-cell and Spatial transcriptomics. Semin Cancer Biol. 2023;95:42–51. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.semcancer.2023.07.001

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  11. Wang H, et al. POSTN(+) cancer-associated fibroblasts determine the efficacy of immunotherapy in hepatocellular carcinoma. J Immunother Cancer. 2024;12. https://doiorg.publicaciones.saludcastillayleon.es/10.1136/jitc-2023-008721

  12. Wu T, et al. Targeting HIC1/TGF-beta axis-shaped prostate cancer microenvironment restrains its progression. Cell Death Dis. 2022;13:624. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41419-022-05086-z

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  13. Gu X, et al. Lactate-induced activation of tumor-associated fibroblasts and IL-8-mediated macrophage recruitment promote lung cancer progression. Redox Biol. 2024;74:103209. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.redox.2024.103209

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  14. Chhabra Y, Weeraratna AT. Fibroblasts in cancer: unity in heterogeneity. Cell. 2023;186:1580–609. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.cell.2023.03.016

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  15. Lavie D, Ben-Shmuel A, Erez N, Scherz-Shouval R. Cancer-associated fibroblasts in the single-cell era. Nat Cancer. 2022;3:793–807. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s43018-022-00411-z

    Article  PubMed  PubMed Central  Google Scholar 

  16. Sebastian A, et al. Single-Cell transcriptomic analysis of Tumor-Derived fibroblasts and normal Tissue-Resident fibroblasts reveals fibroblast heterogeneity in breast Cancer. Cancers (Basel). 2020;12. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/cancers12051307

  17. Wu X, et al. Integrated analysis of single-cell RNA-seq and bulk RNA-seq unravels the heterogeneity of cancer-associated fibroblasts in TNBC. Aging. 2023;15:12674–97. https://doiorg.publicaciones.saludcastillayleon.es/10.18632/aging.205205

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  18. Diep CH, et al. Progesterone receptor signaling promotes Cancer associated fibroblast mediated tumorigenicity in ER + Breast Cancer. Endocrinology. 2024;165. https://doiorg.publicaciones.saludcastillayleon.es/10.1210/endocr/bqae092

  19. Reid SE, et al. Cancer-associated fibroblasts rewire the Estrogen receptor response in luminal breast cancer, enabling Estrogen independence. Oncogene. 2024;43:1113–26. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41388-024-02973-x

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  20. Williams CG, Lee HJ, Asatsuma T, Vento-Tormo R, Haque A. An introduction to Spatial transcriptomics for biomedical research. Genome Med. 2022;14:68. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13073-022-01075-1

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  21. Lambin P, et al. Radiomics: the Bridge between medical imaging and personalized medicine. Nat Rev Clin Oncol. 2017;14:749–62. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/nrclinonc.2017.141

    Article  PubMed  Google Scholar 

  22. Colaprico A, et al. TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data. Nucleic Acids Res. 2016;44:e71. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/nar/gkv1507

    Article  PubMed  CAS  Google Scholar 

  23. Wu SZ, et al. A single-cell and spatially resolved atlas of human breast cancers. Nat Genet. 2021;53:1334–47. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41588-021-00911-1

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  24. Hu C, et al. CellMarker 2.0: an updated database of manually curated cell markers in human/mouse and web tools based on scRNA-seq data. Nucleic Acids Res. 2023;51:D870–6. https://doiorg.publicaciones.saludcastillayleon.es/10.1093/nar/gkac947

    Article  PubMed  CAS  Google Scholar 

  25. Zhang F, Li X, Tian W. Unsupervised inference of developmental directions for single cells using VECTOR. Cell Rep. 2020;32:108069. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.celrep.2020.108069

    Article  PubMed  CAS  Google Scholar 

  26. Jin S, et al. Inference and analysis of cell-cell communication using cellchat. Nat Commun. 2021;12:1088. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41467-021-21246-9

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  27. Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28:1747–56. https://doiorg.publicaciones.saludcastillayleon.es/10.1101/gr.239244.118

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  28. Zheng S, et al. Tumor battlefield within inflamed, excluded or desert immune phenotypes: the mechanisms and strategies. Exp Hematol Oncol. 2024;13:80. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40164-024-00543-1

    Article  PubMed  PubMed Central  Google Scholar 

  29. Yuan Z, et al. Extracellular matrix remodeling in tumor progression and immune escape: from mechanisms to treatments. Mol Cancer. 2023;22:48. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12943-023-01744-8

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  30. Li J, et al. Multiomics profiling reveals the benefits of gamma-delta (gammadelta) T lymphocytes for improving the tumor microenvironment, immunotherapy efficacy and prognosis in cervical cancer. J Immunother Cancer. 2024;12. https://doiorg.publicaciones.saludcastillayleon.es/10.1136/jitc-2023-008355

  31. Mi H, et al. Spatial architecture of Single-Cell and vasculature in tumor microenvironment predicts clinical outcomes in Triple-Negative breast Cancer. Mod Pathol. 2024;38:100652. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.modpat.2024.100652

    Article  PubMed  Google Scholar 

  32. Cords L et al. Cancer-associated fibroblast phenotypes are associated with patient outcome in non-small cell lung cancer. Cancer Cell 42, 396–412 e395. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ccell.2023.12.021 (2024).

  33. Kennel KB, Bozlar M, De Valk AF, Greten FR. Cancer-Associated fibroblasts in inflammation and antitumor immunity. Clin Cancer Res. 2023;29:1009–16. https://doiorg.publicaciones.saludcastillayleon.es/10.1158/1078-0432.CCR-22-1031

    Article  PubMed  CAS  Google Scholar 

  34. Hutton C et al. Single-cell analysis defines a pancreatic fibroblast lineage that supports anti-tumor immunity. Cancer Cell 39, 1227–1244 e1220. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ccell.2021.06.017 (2021).

  35. Rodriguez AB, et al. Immune mechanisms orchestrate tertiary lymphoid structures in tumors via cancer-associated fibroblasts. Cell Rep. 2021;36:109422. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.celrep.2021.109422

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  36. Li X, 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. https://doiorg.publicaciones.saludcastillayleon.es/10.7150/thno.60540

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  37. Morgan A, et al. Medical biology of Cancer-Associated fibroblasts in pancreatic Cancer. Biology (Basel). 2023;12. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biology12081044

  38. Chai C, et al. Single-cell transcriptome analysis of epithelial, immune, and stromal signatures and interactions in human ovarian cancer. Commun Biol. 2024;7:131. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s42003-024-05826-1

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  39. Croizer H, et al. Deciphering the Spatial landscape and plasticity of immunosuppressive fibroblasts in breast cancer. Nat Commun. 2024;15:2806. https://doiorg.publicaciones.saludcastillayleon.es/10.1038/s41467-024-47068-z

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  40. Murphy-Ullrich JE, Sage EH. Revisiting the matricellular concept. Matrix Biol. 2014;37:1–14. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.matbio.2014.07.005

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  41. Alsina-Sanchis E, et al. Endothelial RBPJ is essential for the education of Tumor-Associated macrophages. Cancer Res. 2022;82:4414–28. https://doiorg.publicaciones.saludcastillayleon.es/10.1158/0008-5472.CAN-22-0076

    Article  PubMed  CAS  Google Scholar 

  42. Bagger MM, et al. Evidence of steady-state fibroblast subtypes in the normal human breast as cells-of-origin for perturbed-state fibroblasts in breast cancer. Breast Cancer Res. 2024;26:11. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13058-024-01763-3

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  43. Hu D, et al. Cancer-associated fibroblasts in breast cancer: challenges and opportunities. Cancer Commun (Lond). 2022;42:401–34. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/cac2.12291

    Article  PubMed  Google Scholar 

  44. Zyla J, et al. Combining Low-Dose Computer-Tomography-Based radiomics and serum metabolomics for diagnosis of malignant nodules in participants of lung Cancer screening studies. Biomolecules. 2023;14. https://doiorg.publicaciones.saludcastillayleon.es/10.3390/biom14010044

  45. Jia PF, Li YR, Wang LY, Lu XR, Guo X. Radiomics in esophagogastric junction cancer: A scoping review of current status and advances. Eur J Radiol. 2024;177:111577. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ejrad.2024.111577

    Article  PubMed  Google Scholar 

  46. Alabi RO, Elmusrati M, Leivo I, Almangush A, Makitie AA. Artificial Intelligence-Driven radiomics in head and neck cancer: current status and future prospects. Int J Med Inf. 2024;188:105464. https://doiorg.publicaciones.saludcastillayleon.es/10.1016/j.ijmedinf.2024.105464

    Article  Google Scholar 

  47. Yu Y, et al. Development and validation of a preoperative magnetic resonance imaging Radiomics-Based signature to predict axillary lymph node metastasis and Disease-Free survival in patients with Early-Stage breast Cancer. JAMA Netw Open. 2020;3:e2028086. https://doiorg.publicaciones.saludcastillayleon.es/10.1001/jamanetworkopen.2020.28086

    Article  PubMed  PubMed Central  Google Scholar 

  48. Shi Z, et al. Erratum for: MRI-based quantification of intratumoral heterogeneity for predicting treatment response to neoadjuvant chemotherapy in breast Cancer. Radiology. 2023;308:e239021. https://doiorg.publicaciones.saludcastillayleon.es/10.1148/radiol.239021

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

Figure 1 was created with FigDraw (ID: PYYTT8f45f).

Funding

This work was supported by National Natural Science Foundation of China (Grant Nos. 82304025, 82172827 and 82303857) Tianjin Science and Technology Foundation (23JCQNJC00930).

Author information

Author notes

  1. Zhi-Hao Yu and Huan-Ling Xu contributed equally to this work.

Authors and Affiliations

  1. The First Department of Breast Cancer, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Huan-Hu-Xi Road, He-Xi District, Tianjin, 300060, China

    Zhi-Hao Yu, Huan-Ling Xu, Shuo Wang, Gui-Xin Wang, Yao Tian, Zhao-Hui Chen, Xin Wang, Xu-Chen Cao & Yue Yu

  2. Key Laboratory of Cancer Prevention and Therapy, Tianjin, 300060, China

    Zhi-Hao Yu, Huan-Ling Xu, Shuo Wang, Gui-Xin Wang, Yao Tian, Zhao-Hui Chen, Xin Wang, Xu-Chen Cao & Yue Yu

  3. Tianjin’s Clinical Research Center for Cancer, Tianjin, 300060, China

    Zhi-Hao Yu, Huan-Ling Xu, Shuo Wang, Gui-Xin Wang, Yao Tian, Zhao-Hui Chen, Xin Wang, Xu-Chen Cao & Yue Yu

  4. Key Laboratory of Breast Cancer Prevention and Therapy, Ministry of Education, Tianjin Medical University, Tianjin, 300060, China

    Zhi-Hao Yu, Huan-Ling Xu, Shuo Wang, Gui-Xin Wang, Yao Tian, Zhao-Hui Chen, Xin Wang, Xu-Chen Cao & Yue Yu

  5. Department of General Surgery, Jiangmen Maternity and Child Health Care Hospital, Jiangmen, 529000, Guangdong Province, China

    Huan-Ling Xu

  6. Immunology Department, Key Laboratory of Immune Microenvironment and Disease (Ministry of Education), Tianjin Medical University, Tianjin, 300070, China

    Ying-Xi Li

  7. Department of General Surgery, Tianjin Key Laboratory of Precise Vascular Reconstruction and Organ Function Repair, Tianjin Medical University General Hospital, Tianjin General Surgery Institute, Tianjin, 300052, China

    Yao Tian, Wen-Bin Song & Long He

Authors
  1. Zhi-Hao Yu
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Huan-Ling Xu
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Shuo Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Ying-Xi Li
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Gui-Xin Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Yao Tian
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Zhao-Hui Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Wen-Bin Song
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Long He
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Xin Wang
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Xu-Chen Cao
    View author publications

    You can also search for this author inPubMed Google Scholar

  12. Yue Yu
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

YY, CXC, and YZH: conception and design. XHL, and WS: writing, review, and/or revision of the manuscript. LYX, TY, WGX, CZH, SWB, HL, and WX: administrative, technical, and material support. All the authors approved the final version of the manuscript.

Corresponding authors

Correspondence to Xu-Chen Cao or Yue Yu.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All co-authors have consented the version of manuscript for publication.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

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

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

Yu, ZH., Xu, HL., Wang, S. et al. Integrating spatial and single-cell transcriptomes reveals the role of COL1A2(+) MMP1(+/-) cancer-associated fibroblasts in ER-positive breast cancer. Cancer Cell Int 25, 82 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12935-025-03705-1

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12935-025-03705-1

Keywords

Cancer Cell International

ISSN: 1475-2867

Contact us