Abstract Background: Contactin-4 (CNTN4) has been identified as an inhibitory immune checkpoint protein and mainly expressed on tumor cells. It interacts with APP present in T cells, leading to a reduction in T cell proliferation, activation, and responsiveness to tumors. In this study, we applied artificial intelligence (AI)-powered analyzers to examine the expression of CNTN4 and its association with PD-L1 in various solid tumors. Methods: The AI-powered immunohistochemistry (IHC) analyzers were developed using pan-cancer whole slide images (WSIs) stained with 18 different IHC types. Pathologists annotated cells and tissues on the WSIs. The universal model was trained by annotations from IHC WSI, which could detect tumor cells with four grade intensities (strong, moderate, weak, and negative) and segment invasive cancer area. Finally, the AI analyzer can count the tumor cells in an invasive cancer area and calculate the ratio based on intensity. The AI analyzer was applied to the pan-cancer tissue microarray dataset (n=795) consisting of 18 cancers to validate the analyzer. The dataset was stained with rabbit monoclonal antibody (clone EPR 8735) and PD-L1 22C3 antibody, respectively. The expression levels of CNTN4 and PD-L1 were calculated with H-score (range 0 to 300) or tumor proportion score (TPS). Results: When positivity was defined as moderate-to-strong staining intensity in ≥ 30%, 50%, 75%, and 90%, 42.4% (337/795), 25.5% (203/795), 11.4% (91/795), and 5.2% (41/795) of samples were CNTN4 positive, respectively. In case of moderate-to-strong staining in ≥ 50% tumor cells, CNTN4 expression were most frequently observed in hepatocellular carcinoma (63.9%), endometrial carcinoma (50.0%), stomach cancer (43.9%), pancreatic cancer (40.0%), and prostate cancer (35.4%). Notably, 10% of cases with an H-score < 200 (75/757) showed a range of PD-L1 expression (PD-L1 TPS ≥ 1%), while all of the samples with an H-score ≥ 200 (n=38) were PD-L1 TPS <1%. Conclusions: This study describes the CNTN4 expression levels in various tumors using an AI-powered IHC analyzer. CNTN4 tends to be expressed in opposition to PD-L1, making it a viable immune therapeutic target in cancers that do not conventionally express PD-L1. Citation Format: Seunghwan Shin, Jimin Moon, Soo Ick Cho, Chan-Young Ock, Bu-Nam Jeon, Gyeongyeon Kim, Hyunkyung Yu, Miyoung Cha. Assessment of CNTN4 expression and its association with PD-L1 across 18 various cancers using an artificial intelligence-powered immunohistochemistry analyzer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 553.
Read full abstract