Abstract
Non-Small Cell Lung Cancer (NSCLC) is the leading cause of cancer death globally, and new immunotherapies developed and under development targeting PD-1/PD-L1 checkpoint inhibition require accurate patient selection to assure good clinical outcome. PD-L1 immunohistochemistry is the current biomarker assay used for patient selection, but still imprecise in predicting therapy response. Exploring this issue, we performed computational tissue analysis of PD-L1 immunostaining in procured NSCLC tissues (n = 50) using the Merck KGaA anti-PD-L1 clone MKP1A07310. Staining patterns and PD-L1 cut-off points were interrogated using relevant cancer immune-surveillance biomarkers. Groups with high PD-L1 expression levels (above 25/50% staining cut-off points) were enriched for a biomarker profile in the tumor-nest and microenvironment indicating escape from host-immunity, as represented by increased numbers of cells positive for CD8 and Granzyme B (immune-effectors), FOXP3 (immune-suppressive), and CD68 (P < 0.05). Manual analysis of PD-L1 staining patterns identified tumors with an immune-induced reactive pattern relevant for immunotherapy that would ordinarily be excluded by the arbitrary 25% staining threshold (P < 0.05). Conversely, some cases with completely or predominantly immune-independent constitutive PD-L1 staining patterns that indicate insensitivity to immunotherapy may have been incorrectly selected using this staining cut-off point criterion. Therefore, we propose differentiation of reactive vs constitutive PD-L1 staining patterns to improve the accuracy of this biomarker assay in selecting NSCLC patients for PD-1/PD-L1 immunotherapy.