DESCRIPTION (provided by applicant): Alterations in DNA methylation are hallmarks of cancer cells, and epigenetic markers are increasingly viewed as having great potential for diagnosing, classifying and prognosticating cancers and cancer precursor lesions. Thus, a technical challenge is to develop and apply efficient and high-coverage methods to profile DNA methylation genome-wide in human cancers and in the normal precursor tissues of these cancers. We have developed such a method, called MSNP, to characterize DNA methylation genome-wide using Affymetrix single nucleotide polymorphism DNA microarrays. In addition to profiling gains and losses of net DNA methylation (GOM, LOM), a particular strength of MSNP is that it also queries allele-specific DNA methylation (ASM). Here we hypothesize that MSNP can be further developed and optimized as a high resolution method to reveal differences in methylation patterns not only between cancer and normal tissues, but also between normal tissues and the early atypical or dysplastic precursor tissues which eventually give rise to cancers. In this collaborative R21 proposal, with an experienced team of investigators from the Institute for Cancer Genetics and the Departments of Pathology and Biostatistics, we will advance the methodology and applications of MSNP in several ways. Aim 1 is to optimize MSNP for very high density Affymetrix 1.8M (6.0 array) SNP chips, vetting this method by profiling net and allele-specific DNA methylation in human breast cancers and normal breast epithelium. The results of will be verified by independent assays, including high throughput bisulfite sequencing. In this Aim we will develop bioinformatics approaches for tumor class prediction from MSNP data, and develop formats for data annotation and data sharing. Aim 2 is to miniaturize the MSNP method so that high quality genetic and epigenetic data can be obtained from the small amounts of genomic DNA available from laser-capture microdissection (LCM) or manual microdissection (MM). We will establish conditions allowing ASM and net DNA methylation to be determined using genomic DNA obtained by LCM or MM from normal and cancerous breast epithelium. In this aim we will particularly evaluate breast cancer precursor lesions, namely atypical duct epithelial hyperplasias (ADH), which are associated with a high risk for subsequent breast cancer development. Aim 3 is to correlate MSNP data with expression profiling data, to determine whether MSNP can produce a list of candidate DNA sequences, both promoter-associated and non- promoter-associated, in the human genome that may act as novel methylation-sensitive regulatory elements controlling gene expression in normal and cancer tissues.

Relevance of this research for public health There has been much recent progress in using the tools of molecular biology to classify human cancers, but even with our current best classification schemes many of the common cancer types, including breast cancer, show a rate of disease progression and response to therapy that can be difficult to predict. Patterns of DNA methylation across the genome of cancer cells have the potential to provide new and more informative ways of classifying cancers. The new technological approach, called MSNP, which we will optimize for very high resolution genomic analysis in this project, is an efficient way to scan for patterns of DNA methylation, and we expect that information from MSNP will lead to better ways to classify human cancers, including breast cancer, and to an increased understanding of the pathogenesis of these cancers.