Peripheral blood cell signatures as novel breast cancer biomarkers

B13

Background: In 2007, over 40,000 women died from breast cancer in the United States alone. Detecting breast cancer at an early stage, or when the tumor is small, dramatically increases chances for survival, but current technologies, including mammography, have limited sensitivity and specificity. Our approach has been to use peripheral blood cells (PBCs) as a means of early breast cancer detection. PBCs act as biosensors by responding to both physiological changes and environmental exposures. These responses can be detected by microarray analysis as specific gene expression patterns, or signatures, that reflect the presence of a tumor. Methods: We used Affymetrix M430 2.0 Gene Chips to analyze gene expression in peripheral blood cell samples collected from either transgenic mice with spontaneous mammary adenocarcinomas (n= 15) or tumor-free, matched control mice (n= 15). In addition, we used Affymetrix Human U133 Plus 2.0 Gene Chips to analyze gene expression in peripheral blood cell samples collected from patients diagnosed by biopsy with either invasive breast cancer (n= 15) or a benign breast abnormality (n= 15). Expression data was subject to binary regression analysis, including leave-one-out crossvalidation, as well as validation in an independent sample set. Optimal cutoffs for predicted probabilities were calculated using receiver-operator characteristic curves. Results: We identified a 250-probe signature associated with the presence of spontaneous tumors in a transgenic mouse model of breast cancer. This signature is capable of accurately distinguishing tumor-bearing mice from tumor-free controls in a leave-one-out crossvalidation. The signature was validated in an independent sample set, where it performed with an overall accuracy of 100%. We have also developed a human breast cancer signature, which is able to accurately distinguish patients with invasive breast cancer from those with benign breast abnormalities in a leave-one-out crossvalidation. Conclusions: Our findings indicate that gene expression signatures can be derived from PBCs and can be used to not only distinguish between breast cancer patients and disease-free individuals, but also to predict cancer status. We describe breast cancer signatures derived from both mouse and human peripheral blood cells, demonstrating for the first time that this phenomenon is common to both species, suggesting that mouse models represent an opportunity to complement human biomarker studies, as well as a system in which mechanisms of tumor and PBC interaction may be revealed. These results support the concept of using blood-derived signatures as clinical diagnostic tests, although further investigation is needed to establish the specificity of such tests. We believe this is an approach to the development of clinical biomarkers for accurate, effective and noninvasive breast cancer detection.