A transcriptomic readout of chromosomal instability reveals genome-stress states beyond proliferation in colorectal cancer

Background. Chromosomal instability (CIN) is a defining feature of colorectal cancer (CRC), yet its quantification is limited by the availability of DNA copy-number data, and most existing expression-based CIN signatures largely recapitulate proliferation. We aimed to develop and validate a DNA copy-number-anchored transcriptomic index that measures CIN itself. Methods. A DNA-level CIN_index was built in TCGA-COAD by integrating copy-number breakpoint burden, fraction of genome altered and arm-level gain/loss events (PCA). A 38-gene mRNA-CIN model was derived by WGCNA and XGBoost/Boruta feature selection with cross-validated model comparison. External validation used reference CIN/MMR phenotypes in GSE39582, GSE30540, GSE13294 and GSE18088, head-to-head comparison against CIN70 and proliferation signatures, survival meta-analysis, and pathway, immune, single-cell, DepMap and pan-cancer analyses. HCT116 cells were perturbed with the MPS1 inhibitor Reversine and assayed by flow cytometry, micronucleus staining and RNA-seq re-scoring. Results. The mRNA-CIN score recapitulated DNA-level CIN in TCGA (r = 0.750). It discriminated the reference CIN phenotype in GSE39582 (AUC 0.838, 95% CI 0.797–0.876) while CIN70 and proliferation performed at random (AUC 0.464/0.455; DeLong P = 1.6 x 10 ^− 28 ), tracked MSS/pMMR status across cohorts (AUC 0.83–0.90), and retained performance after proliferation adjustment. mRNA-CIN-high tumors showed replication-stress/DNA-repair programs and an immune-low phenotype reproduced across cohorts, including MSS-only tumors. The score showed no independent prognostic association (meta-analysis per-SD HR 1.02, 95% CI 0.92–1.13), supporting a measurement-tool rather than prognostic-biomarker interpretation. Single-cell and DepMap analyses resolved CIN-associated malignant epithelial states and prioritized HCT116 (low) and SW480 (high) models. Experimentally, Reversine induced dose- and time-dependent abnormal ploidy. At 48 h, micronucleus formation increased with dose, while RNA-seq showed a dose-dependent increase in the mRNA-CIN score despite suppression of CIN70 and proliferation-associated programmes. Conclusions. A copy-number-anchored 38-gene mRNA-CIN score provides a transferable transcriptomic readout of CIN-associated states across public cohorts and experimental perturbation, outperforming proliferation-based signatures. It provides a practical CIN measurement tool and a framework for selecting cellular models to study CIN-associated phenotypes and therapeutic vulnerabilities.