Abstract Introduction: HER2 (ERBB2) is a target for various anti-cancer therapies, including large (Trastuzumab deruxtecan) or small molecules (Neratinib). Quantitative assessment of HER2 expression at the cellular level may improve therapeutic efficacy prediction. While ERBB2 mutations are known oncogenic drivers, their impact on tumor cell HER2 expression is unclear. In this study, we focused on common ERBB2 mutations, and analyzed HER2 expression via AI-powered image analysis in pan-cancer tumor specimens to identify possible mutation-expression correlations. Methods: We analyzed 183,292 pan-cancer samples from AACR GENIE, and 10,967 from TCGA. A set of 156 cases with both ERBB2 mutation status and HER2 4B5 IHC WSIs was obtained (Neogenomics). An AI-powered HER2 analyzer (Lunit SCOPE HER2) was previously developed using 1,133 HER2 IHC-stained WSIs of breast cancer stained with 4B5. Single-cell HER2 intensity was analyzed and aggregated, reported based on breast ASCO/CAP criteria. ERBB2 RNA levels were analyzed for mutations and amplification. Results: We identified 2,735 cases with ERBB2 mutations (2,603/183,292 [1.4%] in GENIE and 132/10,967 [1.2%] in TCGA). These mutations were categorized for analysis based on their highest frequency and previous characterization: exon 20 insertions (ex20ins, including Y772dup (YVMA) and G776InsVC), S310x, and other pathogenic mutations. In the dataset, ex20ins cases were predominantly non-small cell lung cancer (NSCLC, 381/482 [79.0%]) and S310x cases were enriched in urothelial, NSCLC, and breast cancers (269/683 [39.4%], 64/683 [9.4%], and 47/683 [6.9%], respectively). Of the 156 ERBB2-mutated cases with HER2 IHC images, 59 had ex20ins, 24 had S310x, and 73 had other mutations. AI analysis in each case type indicates a significantly higher proportion of HER2 3+ tumor cells in S310x and ex20ins cases than other mutations (median, 7.5% vs. 6.1% vs. 2.8%, p = 0.003). As such, high HER2 expression was observed in S310x and ex20ins cases compared to other mutation cases (HER2 3+ 10/24 [41.7%], 22/59 [37.3%], and 14/73 [19.2%]). Additionally, in NSCLC, ex20ins and S310x cases exhibited a modest enrichment for high HER2 expression relative to other mutation cases (21/56 [37.5%], 3/9 [33.3%], and 5/20 [27.6%]). RNA analyses from TCGA showed higher levels for ex20ins (median 13.6; interquartile range 13.1-13.7 in log2 scale), compared to S310x (12.4; 11.9-13.6) and other mutations (12.6; 12.1-13.3), but lower than that of amplified ERBB2 (15.5; 13.8-17.0). Conclusion: An AI-powered pan-cancer image analysis of HER2 IHC of tumor cells in conjunction with genomic data reveals a positive correlation between ex20ins and S310x ERBB2 mutation and protein expression. This correlation is also seen at the RNA level, but the lesser levels relative to ERBB2 amplified cases suggests the effect may be mediated at the protein level. Citation Format: Taebum Lee, Sangwon Shin, Heon Song, Seonwook Park, Sergio Pereira, Donggeun Yoo, Juneyoung Ro, Jimin Moon, Soo Ick Cho, Changho Ahn, Chan-Young Ock, Siraj M. Ali. Pan-cancer analysis of the influence of ERBB2 alteration on HER2 expression [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 4640.
Read full abstract