A tumor immune microenvironment-related lncRNA signature for the prognosis and immunotherapeutic sensitivity prediction in colorectal cancer.

Abstract

3543 Background: As important molecules in the CRC tumor microenvironment (TME), long non-coding RNAs (lncRNAs) regulate the functions of tumor infiltrating immune cells and sculpt the tumor immune microenvironment (TIME), resulting in difference in survival and response to immunotherapy among CRC patients. However, challenges remain in selecting TIME related lncRNAs (TIME-lncRNAs) of prognosis value and stratifying CRC patients for immunotherapy. Here, the aim of our study was to develop a CRC TIME-lncRNAs signature to provide survival and immunotherapy response predictions. Methods: Gene expression profiles and clinical information of CRC cases (n = 1807) were collected from 7 datasets and divided into training cohort (n = 519) and two testing cohorts (n = 595 and 693, respectively). Utilizing gene expression data of 97 immune cell lines and 61 CRC cell lines, differential expression analysis was used to identify TIME-lncRNAs. Univariate Cox regression and LASSO regression analysis were used to establish a TIME-lncRNAs signature to predict the prognosis of CRC patients. To further investigate the model, multivariate Cox regression, lncRNA-mRNA regulation analysis, gene enrichment analysis and immune infiltration analysis were carried out. The immunotherapy response predicting ability of the model was verified with an independent immunotherapy dataset. Results: Integrating the expression profiles of 10 TIME-lncRNAs, the model stratified CRC patients into low and high-score groups. Patients of the low score group had significantly prolonged survival in both training (hazard ratio (HR) = 2.63, 95% confidence interval (CI) = 1.9-3.63, P < 0.001) and testing cohorts (testing cohort 1: HR = 1.6, 95% CI = 1.19-2.16, P = 0.002; testing cohort 2: HR = 1.64, 95% CI = 1.19-2.26, P = 0.002), while higher tumor purity and less pro-tumor immune cells infiltration were also observed in the low score group. Further investigation showed that both genes differentially expressed between different groups and mRNAs regulated by 10 lncRNAs of the signature were enriched in immune-related and immunotherapy-related pathways. Multivariate Cox regression indicated that the TIME-lncRNAs signature was an independent prognosis factor. Validated with external immunotherapy dataset, the signature provided distinct predictions for patients’ responses to PD-L1 inhibitor therapy, suggesting cases of high score group could benefit more from immunotherapy. Conclusions: Based on the expression of 10 lncRNAs in the TIME, the signature makes predictions for patients’ survival and immunotherapy responses, which could help in stratifying CRC patients for immunotherapy at the bedside.