Background
Recently, the degree of methylation aberrancy, or methylation burden has shown inverse correlation with tumor immunogenicity measured by gene expression profile (GEP) reflecting activated immune cells across multiple cancer types.1 Previously, an artificial intelligence (AI)-powered immune phenotype classification based on spatial tumor-infiltrating lymphocyte (TIL) analysis in hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) was correlated with activated immune GEP. In this study, we investigated correlation between methylation burden and AI-based immune phenotype in The Cancer Genome Atlas (TCGA) Pan-Cancer Atlas dataset.
Methods
Analyzed data including H&E-stained WSIs, relevant gene expression profiles, and methylation array data were obtained from TCGA database (22 tumor types, N = 6243). Methylation burden was calculated by β-score of the TCGA dataset.1 Lunit SCOPE IO, an AI-powered spatial TIL analyzer developed using 16,443 WSIs with pathologists’ annotation, was applied to rate Inflamed, Immune-Excluded, and Immune-Desert Score (IS, IES, and IDS, respectively) which were defined by the proportion of area with TIL density in the cancer area (iTIL) and cancer stroma (sTIL).
Results
Overall, methylation burden showed inverse correlation with iTIL, sTIL, and total TIL densities (r =-0.08, -0.11, and -0.18 respectively, p < 0.001). According to the immune phenotype score, IS or IES having high iTIL or sTIL, respectively, were negatively associated with methylation burden, while IDS having low total TIL was positively associated in either merged pan-cancer analysis (r = 0.22) or each subgroup analysis of 21 tumor types (r = 0.04 ~ 0.59, (table 1) with statistical significance. According to the immune subtype (C1 to C6), methylation burden was highest in C4 and C5, in the context of the subtypes showing significantly higher IDS compared to the others (15.9, p < 0.001).
Conclusions
The degree of methylation aberrancy in cancer is inversely correlated with TIL infiltration in the tumor microenvironment assessed by AI-based H&E analysis.
Reference
Park C, Jeong K, Park JH, et al. Pan-cancer methylation analysis reveals an inverse correlation of tumor immunogenicity with methylation aberrancy. Cancer Immunology Immunotherapy. 2021;70(6):1605–1617.
Abstract 1446 Table 1
Correlation between methylation burden and AI-based immunohistological variables according to tumor types