Fibroblast-enriched senescence-associated transcription and measurement-context-dependent FAP–matrix covariation in colorectal cancer

Abstract Bulk tumour transcriptomes mix cellular composition, senescence-associated transcription and extracellular-matrix signals. We examined the compartmental distribution of senescence-associated transcription in colorectal cancer and its relationship to fibroblast activation protein (FAP)-linked matrix covariation. Patient-paired compartments were analysed in two single-cell RNA-sequencing cohorts, followed by FAP-specific models, expression-matched null signatures, target-purged bulk composition analyses, colon tumour proteomics and three-cohort spatial assessment. Source-available SenMayo scores were higher in fibroblast than epithelial compartments in GSE132465 (23 of 23 patients; median paired difference 0.466; exact P = 2.38 × 10 − 7 ) and GSE166555 (9 of 10; median difference 0.149; P = 0.0195). A 111-gene common-core sensitivity retained the direction in both cohorts. Within GSE132465 tumour fibroblasts, patient-cluster bootstrap estimates did not support additional enrichment among FAP-detected cells: β = 0.0239 (95% confidence interval [CI], − 0.0175 to 0.0675) for SenMayo and β = 0.0340 (95% CI, − 0.0374 to 0.1095) for the senescence-associated secretory phenotype. FAP–matrix correlations were positive in both single-cell cohorts, although direct contrasts did not establish that they were larger than FAP–SenMayo correlations. In 380 TCGA tumours, FAP13–matrix4 covariation remained positive after target-purged joint composition adjustment (partial ρ = 0.466; 95% CI, 0.363 to 0.556), but its magnitude varied across proxy definitions. Spatial common-core effects were positive in GSE280315 and the Valdeolivas cohort but mixed or negative in GSE334323. Together, the data localize senescence-associated transcription preferentially to fibroblast compartments while separating this finding from FAP detection and matrix abundance. Establishing durable cellular senescence or causal stromal signalling will require orthogonal validation.