190 Optimizing PD-1 to PD-L1 proximity assays to predict response in patients with advanced melanoma receiving anti-PD-1

Approximately half of patients with advanced melanoma do not respond to anti-PD-1, and as such, biomarkers for predicting response/resistance are desirable. PD-1-PD-L1 proximity in the tumor microenvironment has been proposed a biomarker. The impact which cell type is expressing PD-1 and/or PD-L1, expression intensity, hot spot vs whole slide sampling, distance between these molecules optimize predictive value determined. FFPE slides from pre-treatment specimens treated anti-PD-1-based therapy (Responders [n=34], Non-Responders [n=18]) were stained multiplex immunofluorescence (mIF) panel: Sox10/S100, PD-1, CD8, CD163, FoxP3. Slides imaged analyzed using AstroPath platform. PD-L1 was divided into tertiles low, mid, high intensity expression. density different lineages within given cell, vice versa, area under curve (AUC) receiver operating characteristic each cell-type pairing CD8+ sampling tested against sampling. AUC all cells ranged 0.60-0.62 10-25 um. highest observed when spots selected CD8+FoxP3+PD-1+ close PD-L1+ (AUC 0.74 at 7 um, p=0.001). AUCs slightly increased assessed, e.g., FoxP3+CD8+PD-1mid located 15 um PD-L1midtumor 0.76, p=0.0008), suggesting that assessment may further stratify patients. These metrics consistent if metric inverted. In conclusion, assays can be improved by selecting mIF assay identify PD-L1.