Abstract
BACKGROUND: Delayed detection of recurrence significantly contributes to colorectal cancer (CRC) mortality, underscoring the need for robust prognostic biomarkers. Although extrachromosomal circular DNA (eccDNA) is a known oncogenic driver, its prognostic utility in CRC remains largely unexplored. METHODS: In this 6-year prospective cohort study, full-length eccDNA profiling of 153 plasma samples was performed using Nanopore sequencing. Differential eccDNA signatures between recurrence (R, n = 20) and non-recurrence (NR, n = 133) patients enabled construction of predictive models for recurrence and mortality. Functional validation of eccDNAs was conducted in HCT116 cells. RESULTS: Compared to NR patients, R patients exhibited enrichment of eccDNAs derived from chromosome 9, shorter median eccDNA lengths, and reduced variability in eccDNA length. All 4.9-5.0 kb eccDNAs derived from CKM, while other eccDNAs showed a strong genomic distribution correlation between groups (Spearman's ρ = 0.73). Promoter-derived eccDNAs were enriched in R patients, particularly from the promoter of CARD9 (eccPromoter-CARD9, 10.4-fold increase). Overexpression of eccPromoter-CARD9 significantly promoted CRC cell proliferation and migration. R patients exhibited elevated eccDNAs harboring the hsa-mir-374c cluster in plasma and tissues, and their corresponding miRNAs demonstrated exceptional diagnostic accuracy in CRC-related TCGA cohorts. An eccDNA-based random forest classifier achieved superior recurrence prediction accuracy (AUC > 0.8), correlating with shorter time-to-recurrence (HR = 3.79) and elevated CA125 and CEA levels. Additional eccDNA-based models effectively predicted recurrence-associated mortality (AUC ≥ 0.93). CONCLUSIONS: The plasma eccDNA landscape may serve as an early and powerful non-invasive biomarker for CRC prognostication, optimizing risk stratification and guiding personalized treatment.