Abstract P1-06-05: DNA methylation-based classification are mostly concordant with intrinsic subtypes of breast cancer

Abstract

Abstract Backgrounds: Breast cancer (BC) is a heterogeneous disease and usually divided into 5 subtypes according to clinicopathologic features. These subtypes have concordance with mRNA-based subtypes called as "intrinsic subtypes". Further gene expression analysis categorized triple negative BCs into 7 sub-subtypes which relate to clinical behaviors. DNA methylation is one of the epigenetic systems to regulate gene transcription. DNA methylation in CpG islands (CGIs) associated with the gene promoter region affects gene expression levels. Recently genome-wide analyses suggested that methylation status not only in CGIs, but also in CpG shores (within 2000 bps from CGIs) and CpG shelves (2000-4000 bps from CGIs) contribute to gene silencing. Thus, we hypothesized that intrinsic subtypes may possess specific methylation profiles especially in regions surrounding CGIs (islands, shores and shelves). To date, several reports demonstrated each intrinsic subtype has specific methylation patterns in CGIs. They indicate that DNA methylation is promising markers for cancer detection, prediction of therapeutic response and clinical outcomes. However, there were no report that conducted a genome-wide high-resolution epigenetic analysis for BC. Objectives: This study aims to determine whether specific DNA methylation patterns exist in CGIs, shores and shelves in BC cells, and whether the DNA methylation-based classification correlates with intrinsic subtypes, furthermore, to identify the subtype specific epigenetic markers using high-resolution methylation array. Materials and Methods: DNA was extracted from 31 samples of 28 BC cell lines (8 luminal, 18 basal, 2 unclassified) and 3 samples of 3 non-BC cell lines. We performed genome-wide methylation analysis using Illumina Infinium HumanMethylation450 BeadChip. We did unsupervised hierarchical clustering analysis of cell lines using top 3000 variably methylated (VM) loci among cancer cell lines. We also analyzed top 3000 VM CGI-related loci. Gene functional annotation clustering was performed using DAVID. Results: In top 3000 VM loci, CGIs were most frequently observed (43.0%). According to clustering analysis using top 3000 VM loci, cell lines were roughly classified into 4 clusters. There were trends that cell lines of same subtype fell into the same cluster, but not definitely. In case of clustering using top 3000 VM CGI-related loci, cell lines were clearly divided into 2 clusters, high and low methylated clusters. High methylated cluster contained almost all luminal cell lines and low methylated cluster contained most of basal cell lines and all non-BC cell lines. In low methylated cluster, cell lines of the same sub-subtype, such as Basal-like 1 and 2, were clustered closely. Gene functional annotation clustering suggested that genes related to "transcription" and "differentiation" were more frequently regulated by epigenetics. Conclusions: High-resolution methylation analysis revealed methylation-based clusters were concordant with intrinsic subtypes in BC cells. Transcriptional factors and differentiation-related genes could be differently regulated by epigenetics among different subtypes of BC. Citation Format: Natsue Uehiro, Fumiaki Sato, Sunao Tanaka, Fengling Pu, Junji Itou, Shigehira Saji, Masakazu Toi. DNA methylation-based classification are mostly concordant with intrinsic subtypes of breast cancer [abstract]. In: Proceedings of the Thirty-Seventh Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2014 Dec 9-13; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2015;75(9 Suppl):Abstract nr P1-06-05.