Senescence-circadian interplay stratifies patient prognosis and reveals immune remodeling heterogeneity in colorectal cancer.

Background: Colorectal cancer (CRC) exhibits substantial biological and prognostic heterogeneity that is not fully captured by conventional clinicopathological staging, and the interplay between the cellular senescence and circadian dysregulation in CRC remains insufficiently defined. Methods: We integrated bulk transcriptomic and clinical data from public CRC cohorts to construct a senescence-circadian interplay score (SCore). Senescence-related and circadian-related gene sets were intersected with CRC differentially expressed genes to identify candidate genes, from which a four-gene random survival forest model was established. The model was evaluated in the TCGA-COAD/READ training cohort and externally validated in GSE12945 and GSE39582. We further characterized clinicopathological associations, prognostic independence, nomogram performance, pathway enrichment, consensus molecular subtype distribution, immune landscape features, and predicted drug sensitivity. Single-cell RNA sequencing dataset was used to localize SCore-associated programs and to examine T-cell communication and differentiation dynamics. In addition, RT-qPCR in CRC cell lines, immunohistochemical validation in 120 paired CRC and adjacent non-tumor tissues, and NOX4-centered knockdown experiments were performed to provide orthogonal experimental support. Results: A higher SCore was consistently associated with poorer overall survival across cohorts and retained prognostic value when integrated with clinicopathological variables. High-SCore tumors were characterized by enrichment of oxidative stress, extracellular matrix remodeling, focal adhesion, and invasion-related programs, together with computationally inferred immune dysfunction/exclusion-associated features. Single-cell analyses localized SCore-associated signals to a T-cell-centered context, where cell-cell communication, pseudotime dynamics, and functional-state remodeling converged. Notably, MIF-(CD74+CXCR4) signaling emerged as a prominent interaction axis. Consistent with the transcriptomic findings, immunohistochemistry confirmed higher expression of NOX4, CXCL1, CDKN2A, and SIX1 in CRC tissues than in paired adjacent non-tumor tissues. Functionally, NOX4 knockdown reduced intracellular ROS, attenuated multiple inflammatory/immunoregulatory mediators, lowered PD-L1 protein expression, and suppressed migratory capacity, supporting a NOX4-associated redox/inflammatory component within the broader SCore-linked state. Conclusion: SCore provides biologically interpretable transcriptomic framework for CRC risk stratification. The NOX4-centered data provide functional support for one component of this state, whereas the broader immune and circadian implications remain hypothesis-generating.