A liquid biopsy signature of circulating extracellular vesicles-derived RNAs predicts response to first line chemotherapy in patients with metastatic colorectal cancer

Background Colorectal cancer (CRC) is one of the most threatening tumors in the world, and chemotherapy remains dominant in the treatment of metastatic CRC (mCRC) patients. The purpose of this study was to develop a biomarker panel to predict the response of the first line chemotherapy in mCRC patients. Methods Totally 190 mCRC patients treated with FOLFOX or XEOLX chemotherapy in 3 different institutions were included. We extracted the plasma extracellular vesicle (EV) RNA, performed RNA sequencing, constructed a model and generated a signature through shrinking the number of variables by the random forest algorithm and the least absolute shrinkage and selection operator (LASSO) algorithm in the training cohort ( n = 80). We validated it in an internal validation cohort ( n = 62) and a prospective external validation cohort ( n = 48). Results We established a signature consisted of 22 EV RNAs which could identify responders, and the area under the receiver operating characteristic curve (AUC) values was 0.986, 0.821, and 0.816 in the training, internal validation, and external validation cohort respectively. The signature could also identify the progression-free survival (PFS) and overall survival (OS). Besides, we constructed a 7-gene signature which could predict tumor response to first-line oxaliplatin-containing chemotherapy and simultaneously resistance to second-line irinotecan-containing chemotherapy. Conclusions The study was first to develop a signature of EV-derived RNAs to predict the response of the first line chemotherapy in mCRC with high accuracy using a non-invasive approach, indicating that the signature could help to select the optimal regimen for mCRC patients.