Correlation of Oncotype Dx With Image Features and Molecular Characteristics of Breast Cancer

OBJECTIVES:

• To characterize differences in tumor morphology and molecular features that exist in good- and poor-prognosis breast cancer as determined by Oncotype Dx assay (a reverse transcriptase-PCR-based assay).
• To develop a computer-aided design (CAD)-based system of analysis of digitized histological images that would perform as well as Oncotype Dx assay in stratifying estrogen receptor-positive (ER+) breast cancer into prognostic categories.
• To identify new therapeutic targets in the PI3-kinase signaling pathway for a subset of poor-prognosis ER+ breast cancers.

OUTLINE: Patients who had Oncotype Dx assay (a reverse transcriptase-PCR-based assay) performed for their breast cancer are identified by review of medical records. Diagnostic H&E stained tumor tissue samples are reviewed by a pathologist. Relevant pathologic features of the tumor (size, stage, grade, estrogen receptor, progesterone receptor, and HER2 staining) and linked Oncotype Dx score are recorded.

The H&E stained tumor tissue samples are scanned to create high-resolution digital images. The images from each tissue sample are subjected to image analysis algorithms to identify sets of image features that most clearly separate tumors with low recurrence scores from those with high recurrence scores.

Unstained paraffin sections from each tumor tissue sample, when available, are processed using IHC to analyze PI3-kinase signaling pathway (PTEN expression). Staining for other components of the PI3-kinase signaling pathway, phospo-AKT, and other proteins of interest is also performed. DNA is extracted from the samples and subjected to pyrosequencing analysis on exons 9 and 20 of PI3KCA. Sequencing of other relevant potential oncogenes, such as AKT, is also performed. Statistical analysis is performed to look for a correlation between Oncotype Dx scores and the presence of PI3KCA mutations and/or loss of PTEN expression.