Targeting a candidate stromal immune checkpoint: a cellular senescence signature identifies COMP⁺ cancer-associated fibroblasts associated with CD8⁺ T-cell exclusion via a predicted COMP–CD47 axis in colorectal cancer

INTRODUCTION: Resistance to immunotherapy in colorectal cancer (CRC) is closely linked to an immunosuppressive tumour microenvironment (TME), in which cancer-associated fibroblasts (CAFs) are key regulators. Cellular senescence and the senescence-associated secretory phenotype (SASP) are increasingly recognised as drivers of CAF activity, yet their systematic prognostic and immunomodulatory value in CRC remains poorly defined. METHODS: Using transcriptomic and clinical data from the TCGA-COADREAD cohort, we screened differentially expressed cellular senescence-related genes (DE-CSRGs) and constructed a prognostic signature, externally validated in independent GEO cohorts (GSE39582). Multi-omics analyses-functional enrichment, immune-infiltration estimation, single-cell and spatial transcriptomics, and ligand-receptor inference-were integrated to characterise the associated biology, and findings were validated at the protein level by multiplex immunofluorescence and immunohistochemistry in an independent pMMR CRC cohort. RESULTS: We established and validated a five-gene senescence signature (COMP, MMP3, MMP1, EPHX4, PLCD4). After adjustment for tumour purity, stromal content and CMS subtype, the signature remained independently prognostic and was enriched in the mesenchymal (CMS4) subtype. The high-risk group showed stromal remodelling, epithelial-mesenchymal transition and an immunosuppressive phenotype with CD8⁺ T-cell exclusion. Single-cell analysis localised COMP to a myofibroblastic CAF subset exhibiting a non-inflammatory senescence phenotype (up-regulated CDKN2A/CDKN1A/SERPINE1 with suppressed IL6/NF-κB SASP). Spatial transcriptomics and tissue staining confirmed spatial co-localisation of COMP⁺ CAFs with reduced CD8⁺ T-cell density. Ligand-receptor inference nominated the thrombospondin-family COMP-CD47 pair as a predicted interaction associated with, but not proven to cause, this immune-excluded phenotype. CONCLUSION: We describe an immunosuppressive CRC subtype defined by a senescence signature and associated with COMP⁺ CAFs. We propose the predicted COMP-CD47 axis as a candidate stromal immune-checkpoint that warrants functional validation, rather than an established therapeutic target.