Cancer-associated fibroblasts reveal aberrant DNA methylation across different types of cancer

Abstract

Background: Cancer-associated fibroblasts (CAFs) are essential components of the tumor microenvironment and play a critical role in cancer progression. Numerous studies have identified significant molecular differences between CAFs and normal tissue-associated fibroblasts (NAFs). In this study, we isolated CAFs and NAFs from liver tumors and conducted a comprehensive analysis of their DNA methylation profiles, integrating our finding with data from studies on other cancer types.

Results: Our analysis revealed that several CAF samples exhibited aberrant DNA methylation patterns, which corresponded with altered gene expression levels. Notably, DNA methylation at liver CAF-specific CpG sites was linked to survival outcomes in liver cancer datasets. An integrative analysis using publicly available datasets from various cancer types, including lung, prostate, esophageal, and gastric cancers, uncovered common epigenetic abnormalities across these cancers. Among the consistently altered CpGs were cg09809672 (EDARADD), cg07134930 (HDAC4), and cg05935904 (intergenic). These methylation changes were associated with prognosis across multiple cancer types.

Conclusion: The activation of CAFs by the tumor microenvironment seems to be associated with distinct epigenetic modifications. Remarkably, similar genomic regions tend to undergo hypomethylation in CAFs across different studies and cancer types. Our findings suggest that CAF-associated DNA methylation changes hold potential as prognostic biomarkers. However, further research and validation are necessary to develop and apply such signatures in a clinical setting.

Keywords: Biomarker; Cancer; Cancer-associated fibroblasts; CpG; DNA methylation; Epigenetic; Hepatocellular carcinoma; Human; Survival.

Fig. 1

Aberrant DNA methylation in fibroblasts of liver cancer. A Liver fibroblasts were isolated from cancer tissue (CAFs) or from tumor-free neighboring tissue (NAFs) and analyzed on EPIC bead chips. The heatmap depicts DNA methylation at 2,134 CpGs with at least 20% mean methylation difference between NAFs and CAFs. Hierarchical clustering showed two groups of CAFs, which were referred to as CAF^low (orange) and CAF^high (red). B The principal component analysis (PCA) of the 10,000 most variable CpGs showed that CAF^high clustered apart from CAF^low and NAFs. C Scatterplot comparing the mean beta values of the CAF^high group versus NAFs. Significant differentially methylated CpGs are highlighted (mean DNAm difference > 20%; limma adjusted p-values < 0.05). D Gene ontology enrichment („biological process”) of CpG sites with significant differential DNAm between CAF^high and NAFs (DE = number of differentially methylated genes, FDR = false discovery rate)

Fig. 2

Gene expression differences between CAFs and NAFs from liver. A Heatmap of 891 genes with at least fourfold expression difference between the groups of NAFs and CAFs. Hierarchical clustering showed the same classification of CAF^low (orange) and CAF^high (red), as observed for DNAm. B Principal component analysis of the 500 most variable genes. C Volcano plot comparing the CAF^high group with NAFs. Highlighted are significantly different expressed genes (adjusted p-values < 0.1). D Differential gene expression was compared with differential methylation in CAF^high versus NAFs. Each CpG site in promoter regions (TSS1500 and TSS200) was paired with the associated genes. Highlighted are significantly differentially expressed and methylated genes

Fig. 3

Selection of potential DNA methylation biomarkers for CAFs in liver cancer. A Principal component analysis of DNA methylation profiles (316,641 CpGs) in NAFs, CAFs with public datasets of various other cell types. Our NAFs and CAFs clustered closely to fibroblasts of other studies. B The selection of candidate CpGs was performed with CimpleG [19] on a reduced number of CpGs, that showed at least 20% mean methylation difference between NAFs and CAFs. C Heatmap of DNAm of eight candidate CpGs that were selected to discern CAFs from other cell types. The results of the selection dataset are depicted here. D To investigate if DNA methylation at these eight candidate CpGs is associated with overall survival, we used the TCGA data of hepatocellular carcinoma [38]. Kaplan–Meier analysis of the 25th percentile of patients with the lowest DNA methylation at these sites versus other patients revealed significant results for three CpGs (cg24106661; cg07046030; and cg23256480)

Fig. 4

Aberrant DNA methylation in fibroblasts associated with various types of cancer. A, B Principal component analysis of the 10,000 most variable CpGs in the dataset containing NAFs and CAFs from lung (GSE68851) [24], esophagus (GSE97687)[27], prostate (GSE115413 and GSE86258) [13], and stomach cancer (GSE117087, GSE194259 and GSE97686) [27, 28]. The samples clustered primarily according to the tissue in dimensions 1 and 2 (A), whereas they were separated into NAFs and CAFs by the fourth dimension (B). C Heatmap of 36 hypomethylated and 4 hypermethylated sites in CAFs versus NAFs, which were significantly differentially methylated in at least 4 of the tissues/cancers (mean methylation difference between NAFs and CAFs > 10%; limma adjusted p-values < 0.05). D Box plots of DNA methylation levels for all samples for the four differently methylated hypomethylated sites shared by all five tissues (adjusted p values are based on the limma differential methylation analysis)

Fig. 5

DNA methylation of cancer-associated fibroblasts is indicative for overall survival. Kaplan–Meier plots with overall survival for TCGA DNA methylation data of five different cancers: kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), low grade glioma (LGG), liver hepatocellular carcinoma (LIHC), and uveal melanoma (UVM). Patients were stratified by the 25th percentile of lowest DNA methylation at the three CAF-associated CpGs (cg09809672 in EDARADD, cg07134930 in HDAC4, and cg05935904 without gene-association). Hypomethylation at these CpGs seems to be associated with higher CAF-content and shorter overall-survival