Introduction
As one of the most common tumors in the world, it is estimated that more than 1.9 million new colorectal cancer (CRC) cases and more than 935,000 deaths in 2020 by the International Agency for Research on Cancer (IARC), ranking third in incidence and second in mortality among known cancers [
1]. Though substantial progress has been made in cancer pathogenesis and drug therapy, the clinical prognosis of colorectal cancer is still poor due to tumor metastasis [
2].
In recent years, the tumor microenvironment (TME) has become a research hotspot of mechanisms of tumor biology and drug development. The tumor microenvironment is a complex system that contains a variety of cellular and noncellular components, such as: immune cells, endothelial cells, cancer-associated fibroblasts (CAFs) and cytokines et al. [
3,
4]. More and more evidences indicate that the crosstalk between various stromal components in TME and tumor cells are critical factors affecting tumor growth and metastasis [
5]. CAFs, as one of the most abundant stromal cell types in TME, can remodel the extracellular matrix (ECM), and promote tumor progression through the interactions with tumor cells and immune cells by secreting various growth factors, chemokines and cytokines [
3]. Some studies have identified different types of CAFs at the single-cell level according to unique gene signatures or functions. However, the types and functions of CAFs varies in different studies or tumor types, suggesting that the functional roles of CAFs in TME is complicated and yet has not been clearly explained, so it is worth for further exploration [
6‐
8].
In this research, we analyzed previously published spatial transcriptomics (ST) data and profiled a spatial atlas of TME inside colorectal cancer tissues [
9]. Our work highlights the role of CAFs in colorectal cancer. In detail, we combined spatial transcriptomics with ssGSEA to precisely identify different cell types. We found that there were subsets enriched with inflammatory-cancer-associated fibroblasts (iCAFs) and myo-cancer-associated fibroblasts (mCAFs) in CRC with bioinformatics analysis and experimental verification. Through functional enrichment analysis, the functional role of CAFs in TME were further investigated. In addition, we explored the correlation between CAFs and prognosis by analyzing bulk RNA-sequencing in public dataset. These results promote an in-depth understanding of the functions of CAFs in TME and provide a basis for CRC precision therapy.
Methods
Data sources
The spatial transcriptomics dataset (10X Genomics) was downloaded from a spatial transcriptomics research website (
http://www.cancerdiversity.asia/scCRLM/). The spatial transcriptomics data of two patients were used, one of whom did not receive neoadjuvant chemotherapy treatment (NACT) was named colon1 (ST-P1), and the other received neoadjuvant chemotherapy with partial response (PR) was named colon2 (ST-P3) [
9]. Bulk RNA-sequencing dataset was derived from the COADREAD cohort of The Cancer Genome Atlas (TCGA) and downloaded from UCSC XENA (
https://xena.ucsc.edu/).
Spatial transcriptomics data processing
We used the R package Seurat (v4.1.0) to process space transcriptomics data and used log-normalization to standardize data [
10]. We used functions SelectIntegrationFeatures, FindIntegrationAnchors, and IntegrateData to remove batch effects and integrate Seurat object into a single ST dataset. To reduce the dimensionality of ST data, function RunPCA was performed, then functions FindNeighbors and FindClusters were used to cluster similar ST spots. Different clusters were preliminarily annotated based on hematoxylin–eosin staining (H&E) sections and unsupervised clustering analysis. When we used cell markers to annotate clusters, we found that some clusters highly expressed multiple cell markers, so ssGSEA algorithm [
11] were performed on scoring common cell types based on the average expression matrix of different clusters, and studies have confirmed its robustness in ST [
12,
13].
Enrichment analysis
Differentially expressed genes (DEGs) of cell clusters were identified by the function FindAllMarkers, and top 10 DEGs of each cluster that were ranked according to the log2FC were applied to data visualization. Function clusterProfiler (v3.18.1) [
14] was used for KEGG pathway analysis. ssGSEA (Hallmark Gene sets from Molecular Signatures Database, MSigDB) was conducted with GSVA (v1.38.2) [
15]. We set adj.
p.val < 0.05, |avg. logFC|> 0.5 as cut-off criteria.
Bulk RNA-seq data analysis
We downloaded TCGA colorectal cancer cohort data (COADREAD) for transcriptome analysis. We selected the top 10 DEGs in the cluster as cluster-specific geneset, and calculated the scores of genesets in the transcriptomic data through GSVA, median was set as cut-off value. R packages survival (v3.2-10), survminer (0.4.9) were used for survival analysis. In addition, ImmuneScore and StromalScore were calculated using ESTIMATE algorithm, and their correlation with specific cell type was analyzed by correlation analysis, p value < 0.05 as statistically significant.
Pseudotime analysis
In order to analyze the cell type differentiations in the tumor microenvironment, monocle2 (v2.18.0) [
16] was used for trajectory analysis to find the transitional relationships among different clusters. The plot_genes_in_pseudotime function was applied to discover transitional changes in gene expression levels among different clusters.
Immunofluorescence assay
Specific proteins expression of distinct cancer-associated fibroblasts within colorectal cancer were analyzed using Immunofluorescence staining. Paraffin-embedded colorectal cancer tissues (n = 2) were made slices, heat mediated antigen retrieval with TRIS–EDTA (pH = 8) for 20 min and then blocked with 5% BSA for 30 min at room temperature. After incubated with anti-PDGFRA (rabbit, 1:100, HuaBio, ET1702-49) and anti-RGS5 (rabbit, 1:300, Affinity, Cat. #: DF4417) separately at 4℃ overnight, the sections were washed with PBS for three times, then 488 conjugated goat anti-rabbit IgG (1:800 dilution, HA1121) and 594 conjugated goat anti-rabbit IgG (1:800 dilution, HA1122) were separately added and incubated for 1 h. Finally, DAPI was applied to stain cell nucleus. Sections without incubation with primary antibody or secondary antibody were used as control (without any light).
Immunohistochemical analysis
Immunohistochemical analysis was carried out to explore the specific protein expression in distinct tissues. Paraffin-embedded colorectal cancer or para-carcinoma tissues (n = 5) were applied in this project. Sections were firstly treated with boiling TRIS–EDTA (pH = 8) to repair antigen for 20 min. Then endogenous peroxidases were wiped off. Thereafter, tissues were incubated with 5% BSA for 30 min. Anti-PDGFRA (rabbit, 1:100, HuaBio, ET1702-49) was added in sections at 4 °C overnight. After washed with PBS for three times and probed with HRP conjugated compact polymer system for 30 min, DAB was used as the chromogenic agent. Hematoxylin was used to dye the cell nucleus.
Statistical analysis and visualization
All statistical analysis were based on R (v4.0.5), and data visualization was performed on R packages Seurat (v4.1.0), ggplot2 (v3.3.5), ggsignif (v0.6.1), pheatmap (v1.0.12) and ggstatsplot (v0.9.1) [
17].
Discussion
Currently, targeted drugs combined with surgery has largely reduced the mortality of colorectal cancer, however tumor metastasis and drug resistance are still important factors leading to poor prognosis of CRC patients [
8], so new therapeutic strategies are waiting to be discovered. Previous studies pay more attention to the biology of tumor cells, nevertheless, emerging research gradually noticed the important roles of tumor microenvironment in tumorigenesis, metastasis and drug resistance [
5,
30], while CAFs as the main stromal components in the microenvironment, can affect the tumor growth and metastasis through various mechanisms [
3]. In this research, we utilized spatial transcriptome data and public TCGA cohort dataset to elucidate CAFs functions and its interactions with the microenvironment, hoping to promote the development of drug treatment strategies, a workflow was drew to clearly show the overview of our project (Fig.
6).
Li et al. firstly divided CAFs into two distinct subpopulations (CAF-A and CAF-B) based on the single-cell sequencing data in CRC. CAF-A was related to ECM remodeling, while CAF-B was similar to myo-fibroblasts. However, the functional roles of these two types of CAFs remain unclear [
19]. In recent years, studies have classified CAFs into iCAFs and mCAFs according to their functions in pancreatic cancer, prostate cancer and triple negative breast cancer [
6,
7,
31]. We also identified the existence of these two types of CAFs in CRC with fundamental experiment. An increasing number of research reveal that CAFs through various approaches such as receptor activation, cytokine and cytotoxic production, to inhibit NK cells [
3,
27,
32], and we also found that NK cells were decreased in iCAFs-enriched cluster, which was consistent with these studies. In addition, we detected that iCAFs and macrophages were co-enriched in the same subcluster by using bioinformatics analysis. Zhang et al. reported that CAFs can inhibit NK cells proliferation by regulating tumor-associated macrophages (TAM) [
29], suggesting the extensively interactions between iCAFs and immune cells in TME. Besides, the relationships of chemotherapy drugs and tumor microenvironment were also examined, and we found that the proportion of iCAFs increased while the number of NK cells and monocytes decreased in patient who underwent neoadjuvant chemotherapy (colon2). Since iCAFs can inhibit immune cells in TME, the decrease of immune cells may be related to iCAFs [
27]. Therefore, we speculated that the increase of iCAFs in patient who underwent chemotherapy might be an important mechanism for the occurrence of drug resistance. In the future, more fundamental experiments should be conducted to elucidate the underlying mechanism.
Although epithelial-mesenchymal transformation (EMT) has been widely studied, we still have limited understanding of the mechanism of this phenomenon [
33]. Our study indicated that iCAFs up-regulated EMT signature, while tumor cells lacked of an EMT marker, suggesting that EMT may be driven by fibroblasts rather than tumor cells [
19,
33,
34]. In addition, metabolic reprogramming is also an important mechanism for tumor progression. More and more studies have shown that CAFs can promote tumor metastasis through the metabolic crosstalk with tumor cells or other stromal cells in TME [
28,
35]. Despite the known glycolysis and amino acid metabolism play important roles in TME, there are emerging evidences demonstrate that lipid metabolism is also essential for tumor development [
36,
37]. This is consistent with our ssGSEA pathway analysis. Furthermore, when we investigated the relationship between chemotherapy and metabolism, we discovered opposite trends between fatty acid metabolism pattern and other metabolism patterns in iCAFs-enriched cluster, and we speculated this phenomenon might lead to drug resistance. Therefore, targeted therapy toward iCAFs as well as its metabolites in tumor microenvironment is likely to be an effective way to inhibit tumor metastasis.
By correlating our spatial transcriptomics results to a public database, we confirmed that iCAFs was associated with immune infiltration and clinical outcomes in CRC. As main stromal cells in tumor microenvironment, CAFs can be detected in almost all solid tumors, thus targeting iCAFs can include the vast majority of patients, which is an ideal choice for colorectal cancer treatment.
There are still some limitations in this project. In this study, spatial transcriptome data were used to profile the heterogeneity of TME. Although this emerging technology can provide spatial location information of cells and facilitate the identification of cell types, due to the fact that 10X Genomics platform contains 1–10 cells in each spatial spot [
38], the accuracy of spatial transcriptome data is lower than that of single-cell sequencing, so more efforts needed to improve the resolution of ST. Therefore, aiming to annotate cell types of each cluster more precisely, ssGSEA algorithm were applied in this project. Compared with morphological regions annotation, each cluster had been accurately annotated, which once again confirmed that ssGSEA algorithm is feasible for ST. Hence, the combination of ST and ssGSEA can provide a more comprehensive understanding of the cell types contained in the subpopulation, so as to discover the co-enriched cell types and the potential interactions between themselves. In addition, the small sample in our study may also be a possible limitation. And our project concerned on the functions of iCAFs in TME, the value of mCAFs was not fully recognized, so ongoing efforts are required to explore the functions of mCAFs.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.