Natural killer cell transcriptomic expression and prediction of survival after immune checkpoint blockade across cancers.

2518 Background: Preclinical and clinical evidence has suggested the role of natural killer (NK) cells in tumor immunity and prognosis across various cancer types, but their significance during immune checkpoint blockade (ICB) treatment is poorly understood. This study investigated the impact of tumor-infiltrating NK cells, surrogated by the RNA expression of genes related to NK cells in the tumor microenvironment, on the outcomes of the patients who undergo ICB, using real-world, pan-cancer data. Methods: We analyzed RNA sequencing data of 395 immune-related genes from 514 patients with various cancers included in the Study of Personalized Cancer Therapy to Determine Response and Toxicity (NCT02478931). After excluding 25 patients ineligible for survival analysis, we defined two distinctive cohorts: patients who received ICB (ICB cohort, N = 217) and those who did not (non-ICB cohort, N = 272). Among the 395 immune-related genes, 43 were selected as NK-related genes according to the Human Protein Atlas. Patients in each cohort were clustered into two groups based on the NK-related gene expression. The associations between the clusters and the clinical outcomes, including overall survival (OS) and progression-free survival (PFS), were analyzed using univariate and multivariate analyses. In the multivariate analysis, cancer types, line of immunotherapy, positive programmed-death ligand 1 immunohistochemistry (PD-L1 IHC, ≥ 1%), high tumor mutational burden (TMB, ≥ 10/Mb), and microsatellite instability (MSI) were adjusted. Results: The ICB cohort (N = 217) was divided into two clusters (hot vs. cold), characterized by general abundance and paucity of NK-related gene transcripts (N = 101 and 116, respectively). The clusters were not significantly associated with histology, positive PD-L1 IHC, high TMB, or MSI. Those in the hot cluster demonstrated significantly longer overall survival (OS) after starting ICB compared to those in the cold clusters in univariate analysis (hazard ratio [HR] and 95% confidence interval [CI]: 0.65 [0.45-0.92], p = 0.015) and multivariate analysis (HR and 95% CI: 0.57 [0.34-0.87], p = 0.010). The cluster was not significantly associated with PFS. The non-ICB cohort (N = 272) was similarly divided into two clusters (hot vs. cold), with the characteristics of generally high and low NK-related gene RNA expressions. (N = 114 and 158, respectively). However, in the non-ICB cohort, patients in the hot clusters did not demonstrate significantly prolonged OS compared with those in the cold cluster either with univariate or multivariate analysis (HR and 95% CI: 0.93 [0.65-1.32], p = 0.67 and 0.97 [0.76-2.01], p = 0.90 respectively). Conclusions: Transcriptomic expression of NK-related genes in tumor tissue independently and significantly predicted longer survival after ICB treatment, which implies a role of tumor infiltrating NK cells in immunotherapy outcome.