多模态真实世界数据揭示结直肠癌对TOP1i临床应答的预测因素并优化ADC策略

Real-world data (RWD) encompassing diverse treatment regimens offers a valuable opportunity to identify predictive biomarkers of response and resistance across broad patient populations. Although data from clinical trials with antibody-drug conjugate (ADC) therapies remains limited, analysis of large standard-of-care cohorts offers valuable insight into chemotherapy resistance. For example, irinotecan-a widely used topoisomerase I inhibitor (TOP1i) in metastatic colorectal cancer (CRC) regimens-serves as a model to infer potential resistance to ADCs that utilize TOP1i payloads and the feasibility of combination strategies. We analyzed ConcertAI PT360® electronic health records linked to Caris Life Sciences genomic data for 810 CRC patients with available clinical responses and pre-treatment samples to investigate irinotecan response and resistance mechanisms. Patients were classified into responder (R; n = 241), non-responder (NR; n = 308), acquired resistance (AR; n = 181, transition from R to NR), and stable disease (SD; n = 80). Response to irinotecan was associated with significantly better overall survival (p < 0.0001), with median OS of 106, 87, 73, and 47 months for R, SD, AR, and NR, respectively. Response classifications and transcriptomic data were independent of covariates such as ECOG score, diagnostic stage, ethnicity, sex, or age. Most patients were microsatellite stable with low tumor mutational burden. Gene expression and mutation profiling revealed upregulation of mucins ( MUC5AC , MUC2 ) and enrichment of KRAS mutations in NR, indicating a mucinous CRC subtype resistant to irinotecan. Conversely, TOP1 gene amplification and increased expression were more frequent in R. Pathway analysis indicated higher inflammation and pre-existing immune activity in R/AR versus NR, especially B cell immunity. Gene signatures of tertiary lymphoid structures were also elevated in R/AR, further linking to B-cell immunity. Longitudinal samples revealed enhanced anti-tumor immune signatures post treatment, including dendritic cell activation, upregulated antigen presentation, and elevated interferon-gamma signaling. Additionally, we utilized an innovative in silico CRISPR knockout machine learning model on curated RWD to identify potential novel targets to overcome irinotecan resistance. This study leveraged a rigorously curated multi-modal RWD cohort of CRC patients treated with irinotecan to elucidate mechanisms of clinical response and resistance. Enrichment of immune activation pathways in irinotecan responders suggests that baseline tumor microenvironment may predict outcome, while increased immune signatures post-treatment indicate that combining immunotherapy could enhance responses. These findings provide insights into the biomarkers associated with systemic TOP1i and may inform further clinical development of TOP1i-ADCs.