Identification of conserved gene expression features across human and murine mammary tumors

4841 Breast cancer has been shown to consist of not just a single disease but several distinct diseases or subtypes identified by microarray analysis. To study breast cancer, researchers have spent the last couple of decades developing a number of mouse models to mimic human breast cancer. How well these models represent human breast tumor phenotypes is still unclear. We have used gene expression analysis with DNA microarrays to address this question. We have characterized more than 13 different murine mammary models and compared these data to the expression patterns from primary human breast tumors. Hierarchical clustering analysis of the murine samples showed that the WAP- myc , MMTV- neu , MMTV- PyMT , WAP- In t3, WAP- Tag and C3(1)- Tag tumors were highly correlated within each model. Other models, including the WAP- T 121 , MMTV- Wnt1 , Brca1 conditional, and DMBA-induced tumor model, did not show this consistency and gave rise to tumors with expression profiles consistent with the expansion of multiple cell types or different cell types of origin. A combined clustering analysis of the murine tumors with 184 human breast tumors showed many shared expression features. These features included a proliferation signature, an interferon-regulated pattern, and patterns reflective of the presence of immune cells, adipocytes and fibroblasts. Like human breast tumors, murine tumors could be categorized according to their presumed cellular origins; the C3(1)- Tag , two independent Brca1 deficient models, and DMBA-treated models displayed expression characteristics of human basal-like breast tumors. In contrast, the MMTV- neu , MMTV- PyMT , and WAP- myc models shared features with human luminal breast tumors including the high expression of GATA3 and XBP1 . Single mouse models did not reproduce the entire expression pattern seen in a specific human subtype; rather portions of a subtype’s expression profile were captured/represented by different murine models. The shared patterns of expression present between mice and humans provides a common framework for the direct comparison and integration of animal models with human breast tumors.