Objective Cancer-associated fibroblasts (CAFs), a major component of cancer stroma, can confer aggressive properties to cancer cells by secreting multiple factors. Their phenotypes are stably maintained, but the mechanisms are not fully understood. We aimed to show the critical role of epigenetic changes in CAFs in maintaining their tumour-promoting capacity and to show the validity of the epigenomic approach in identifying therapeutic targets from CAFs to starve cancer cells.

Design Twelve pairs of primary gastric CAFs and their corresponding non-CAFs (NCAFs) were established from surgical specimens. Genome-wide DNA methylation and H3K27me3 analyses were conducted by BeadArray 450K and ChIP-on-Chip, respectively. Functions of potential a therapeutic target were analysed by inhibiting it, and prognostic impact was assessed in a database.

Results CAFs had diverse and distinct DNA methylation and H3K27me3 patterns compared with NCAFs. Loss of H3K27me3, but not DNA methylation, in CAFs was enriched for genes involved in stem cell niche, cell growth, tissue development and stromal–epithelial interactions, such as WNT5A, GREM1, NOG and IGF2. Among these, we revealed that WNT5A, which had been considered to be derived from cancer cells, was highly expressed in cancer stromal fibroblasts, and was associated with poor prognosis. Inhibition of secreted WNT5A from CAFs suppressed cancer cell growth and migration.

Conclusions H3K27me3 plays a crucial role in defining tumour-promoting capacities of CAFs, and multiple stem cell niche factors were secreted from CAFs due to loss of H3K27me3. The validity of the epigenetic approach to uncover therapeutic targets for cancer-starving therapy was demonstrated.