Multi-omics reveals the impact of cancer-associated fibroblasts on the prognosis and treatment response of adult diffuse highest-grade gliomas.

BACKGROUND

Cancer associated fibroblasts (CAF), an important cancer-promoting and immunosuppressive component of the tumor immune microenvironment (TIME), have recently been found to infiltrate adult diffuse highest-grade gliomas (ADHGG) (gliomas of grade IV).

METHODS

Gene expression and clinical data of ADHGG patients were obtained from the CGGA and TCGA databases. Consensus clustering was used to identify CAF subtypes based on CAF key genes acquired from single-cell omics and spatial transcriptomomics. CIBERSORT, ssGSEA, MCPcounter, and ESTIMATE analyses were used to assess the TIME of GBM. Survival analysis, drug sensitivity analysis, TCIA database, TIDE and cMap algorithms were used to compare the prognosis and treatment response between patients with different CAF subtypes. An artificial neural network (ANN) model based on random forest was constructed to exactly identify CAF subtypes, which was validated in a real-world patient cohort of ADHGG.

RESULTS

Consensus clustering classified ADHGG into two CAF subtypes. Compared with subtype B, patients with ADHGG subtype A had a poorer prognosis, worse responsiveness to immunotherapy and radiotherapy, higher CAF infiltration in TIME, but higher sensitivity to temozolomide. Furthermore, patients with subtype A had a much lower proportion of IDH mutations. Finally, the ANN model based on five genes (COL3A1, COL1A2, CD248, FN1, and COL1A1) could exactly discriminate CAF subtypes, and the validation of the real-world cohort indicated consistent results with the bioinformatics analyses.

CONCLUSION

This study revealed a novel CAF subtype to distinguish ADHGG patients with different prognosis and treatment responsiveness, which may be helpful for accurate clinical decision-making of ADHGG.