Surgical first line resection samples are frequently accessed to select patients for immune checkpoint inhibitor (ICI) therapy, often based on expression of PD-L1 measured by immunohistochemistry (IHC). However, while PD-L1 expression may enrich for response to ICIs, other immune parameters in the tumor microenvironment will likely contribute to outcome. To assist in the identification of biomarker(s) that might predict response to ICIs, we have analyzed a cohort of pre-treatment NSCLC cases for which second line immunotherapy clinical follow-up data are available. Formalin fixed paraffin embedded (FFPE) tumor samples were analysed (i) by single-plex IHC for CD3, CD8, CD68, CD163 and PD-L1, plus digital image analysis (CellProfilerTM), and (ii) profiled for multidimensional biomarkers using the targeted RNA sequencing and machine-learning platform ImmunoPrism® from Cofactor Genomics. Clinical follow-up data indicated objective response to ICI therapy for 4/18 patients, with average time from initial diagnosis to ICI treatment of 33.5 ± 29.6 months (mean ± SD). While CD68+ macrophage frequencies evaluated by IHC did not differ significantly between responder and non-responder populations, significant increases in T cell numbers (CD3: 2.3-fold; CD8: 2.7-fold; both p<0.05) were observed for the responder population. CD8 T cells were orthogonally measured using the ImmunoPrism assay, and the same significant differences for CD8 were observed. Although a trend towards decreased M2-like CD163+ macrophage/monocyte numbers was apparent by IHC for responders, this was not supported by the Cofactor analysis. When sections from the same FFPE block are analyzed across multiple commercially available platforms, there is high confidence for those signals which show concordance between platforms, such as the increase in CD8 T cell abundance observed for responders in this study. Moving beyond single-analyte biomarkers, a multidimensional biomarker combining immune escape genes and RNA-based immune cell measurements was generated using the ImmunoPrism platform with standard parameters. The resulting biomarker had the following performance characteristics: predictive accuracy, 89%; positive predictive value (PPV), 100%; negative predictive value (NPV), 88%; sensitivity, 50%; and specificity, 100%. A receiver-operating characteristic (ROC) curve was also generated for this putative biomarker, with an area under the curve (AUC) of 0.87. The promising results from this exploratory sample set warrant further investigation in a larger cohort. These data also demonstrate that clinical archives with well-curated demographic and outcome data, such as the FFPE samples analyzed here, provide excellent cohorts for biomarker screening and discovery studies. The application of new multianalyte approaches enables additional signals from the tumor microenvironment to be captured and included for predicting patient response.

Citation Format: Milan Bhagat, Woo Ho Kim, Lorenzo Memeo, Lorenzo Colarossi, Natalie LaFranzo, Steve Daniel, Christopher Womack, Marie Cumberbatch. Immune biomarkers in the tumor microenvironment associated with response in pre-treatment non-small cell lung cancer (NSCLC) samples with second line immunotherapy follow-up data [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3194.