Immune Cell Densities Predict Response to Immune Checkpoint-Blockade in Head and Neck Cancer.

Abstract

Pre-ICB tumor tissue sections were obtained from 9 responders (complete response, partial response, or stable disease) and 11 non-responders (progressive disease) classified via RECISTv1.1. A custom multi-immunofluorescence (mIF) staining assay was designed, optimized, and applied to characterize tumor cells (pan-cytokeratin), T cells (CD4, CD8), B cells (CD19, CD20), myeloid cells (CD16, CD56, CD163), dendritic cells (LAMP3), fibroblasts (alpha-Smooth Muscle Actin), proliferative status (Ki67) and immunoregulatory molecules (PD1). Spatial metrics were compared among groups. Serial tissue sections were scored for TLS in both H&E and mIF slides. A machine learning model was employed to measure the effect of these metrics on achieving a response to ICB (SD, PR, or CR). A higher density of B lymphocytes (CD20+) was found in responders compared to non-responders to ICB (p=0.022). A positive correlation was observed between mIF and pathologist identification of TLS (R2= 0.66, p-value= <0.0001). TLS trended toward being more prevalent in responders to ICB (p=0.0906). The presence of TLS within 100 um of the tumor was associated with improved overall (p=0.04) and progression-free survival (p=0.03). A multivariate machine learning model identified TLS density as a leading predictor of response to ICB with 80% accuracy. Immune cell densities and TLS spatial location within the tumor microenvironment play a critical role in the immune response to HNSCC and may potentially outperform CPS as a predictor of ICB response.