Inflammatory response signature score model for predicting immunotherapy response and pan-cancer prognosis

Shuzhao Chen,Mayan Huang,Limei Zhang,Qianqian Huang,Yun Wang,Yang Liang
DOI: https://doi.org/10.1016/j.csbj.2023.12.001
2023-12-06
Abstract:Background: Inflammatory responses influence the outcome of immunotherapy and tumorigenesis by modulating host immunity. However, systematic inflammatory response assessment models for predicting cancer immunotherapy (CIT) responses and survival across human cancers remain unexplored. Here, we investigated an inflammatory response score model to predict CIT responses and patient survival in a pan-cancer analysis. Methods: We retrieved 12 CIT response gene expression datasets from the Gene Expression Omnibus database (GSE78220, GSE19423, GSE100797, GSE126044, GSE35640, GSE67501, GSE115821 and GSE168204), Tumor Immune Dysfunction and Exclusion database (PRJEB23709, PRJEB25780 and phs000452.v2.p1), European Genome-phenome Archive database (EGAD00001005738), and IMvigor210 cohort. The tumor samples from six cancers types: metastatic urothelial cancer, metastatic melanoma, gastric cancer, primary bladder cancer, renal cell carcinoma, and non-small cell lung cancer.We further established a binary classification model to predict CIT responses using the least absolute shrinkage and selection operator (LASSO) computational algorithm. Findings: The model had high predictive accuracy in both the training and validation cohorts. During sub-group analysis, area under the curve (AUC) values of 0.82, 0.80, 0.71, 0.7, 0.67, and 0.64 were obtained for the non-small cell lung cancer, gastric cancer, metastatic urothelial cancer, primary bladder cancer, metastatic melanoma, and renal cell carcinoma cohorts, respectively. CIT response rates were higher in the high-scoring training cohort subjects (51%) than the low-scoring subjects (27%). The five-year survival rates in the high- and low score groups of the training cohorts were 62% and 21%, respectively, while those of the validation cohorts were 54% and 22%, respectively (P < 0·001 in all cases). Inflammatory response signature score derived from on-treatment tumor specimens are highly predictive of response to CIT in patients with metastatic melanoma. A significant correlation was observed between the inflammatory response scores and tumor purity. Regardless of the tumor purity, patients in the low score group had a significantly poorer prognosis than those in the high score group. Immune cell infiltration analysis indicated that in the high score cohort, tumor-infiltrating lymphocytes were significantly enriched, particularly effector and natural killer cells. Inflammatory response scores were positively correlated with immune checkpoint genes, suggesting that immune checkpoint inhibitors may have benefited patients with high scores. Analysis of signature scores across different cancer types from The Cancer Genome Atlas revealed that the prognostic performance of inflammatory response scores for survival in patients who have not undergone immunotherapy can be affected by tumor purity. Interleukin 21 (IL21) had the highest weight in the inflammatory response model, suggesting its vital role in the prediction mode. Since the number of metastatic melanoma patients (n = 429) was relatively large among CIT cohorts, we further performed a co-culture experiment using a melanoma cell line and CD8 + T cell populations generated from peripheral blood monocytes. The results showed that IL21 therapy combined with anti-PD1 (programmed cell death 1) antibodies (trepril monoclonal antibodies) significantly enhanced the cytotoxic activity of CD8 + T cells against the melanoma cell line. Conclusion: In this study, we developed an inflammatory response gene signature model that predicts patient survival and immunotherapy response in multiple malignancies. We further found that the predictive performance in the non-small cell lung cancer and gastric cancer group had the highest value among the six different malignancy subgroups. When compared with existing signatures, the inflammatory response gene signature scores for on-treatment samples were more robust predictors of the response to CIT in metastatic melanoma.
What problem does this paper attempt to address?