Hu and colleagues [9
] used SAGE (serial analysis of gene expression) technology to compare gene expression levels in tumors derived from the mouse p53 null mammary transplant model and human breast tumors and found that approximately 72 transcripts were dysregulated in both. Down-regulation of specific cytokines (LIF, IL6, CXCL1, and CCL2) was identified in mouse and human tumor samples as well as similar changes in the expression of genes involved in apoptosis, proliferation, and differentiation. These data highlight some of the genetic similarities between mouse and human breast cancer and support the use of GEM models as an appropriate resource for better understanding human breast cancer.
Extensive amounts of high-throughput gene expression array data have been generated for human breast tumors and similar data have been emerging for GEM mammary cancer models. Comparing array data between species has been extremely challenging for numerous technical and biologic reasons. For instance, cellular compositions of human and GEM tumors may vary with human breast tumors often containing substantially more stromal components than the mouse tumors. Additionally, comparing data across different array platforms where gene and probe compositions are not identical raises important statistical challenges that are beyond the scope of this review but that are considered further elsewhere [10
However, studies by Herschkowitz and colleagues [8
] and our laboratory [7
] have revealed that gene expression signatures associated with individual GEM models of mammary cancer express genetic profiles that are similar to particular subtypes of human of breast cancer (Table ). Herschkowitz and colleagues [8
] used gene expression profiling to compare 13 different mouse models of mammary cancer with human breast tumor data sets. Analyses of gene expression stratified GEM and DMBA-induced mammary cancer models into five groups: normal mammary gland, tumors with mesenchymal characteristics, basal/myoepithelial, luminal, and tumors with mixed characteristics.
Shared genetic and genomic features between mouse models of mammary cancer and human breast tumor subtypes
Several GEM models display luminal features with expression of Gata3, luminal keratins K8/18, and the luminal tumor-defining gene XBP1
, including tumors arising from MMTV-Neu, MMTV-PyMT, WAP-Myc, and WAP-Int3 models [8
]. However, unlike a substantial portion of human luminal tumors, these models are ER-
. This suggests that, in these mouse models, GATA3 (a gene that is coexpressed with ER in human tumors) may better identify the luminal subtype than ER expression alone. This also indicates that GATA3 expression is not sufficient to activate ER expression [11
]. However, several GEM models develop ER+
tumors, including the p53fp/fp
WAP-Cre conditional knockout [14
], the p53-/-
], MMTV-Wnt1 [16
], and the MMTV-tTA-TAg-ERα conditional ER expression [17
]. Our laboratory has demonstrated that the ER+
tumors from the p53fp/fp
Wap-Cre conditional knockout model segregate with luminal type A human tumors (A.M. Michalowski, T. Qiu, C. Kavanaugh, E. Lee, D. Medina, X. Xu, C. Deng, J. Powell, J. Shih and J. Green, unpublished data), whereas the status of the other tumor models remains to be determined. These results raise the important issue that the method of generating a GEM model (knockout versus promoter-driven transgenic) is critical for determining which cancer progenitor cells may be targeted for transformation leading to basal or luminal phenotypes.
Another major subtype of breast cancer has been defined by the expression of basal cell characteristics, including tumors referred to as triple-negative (ER-
, progesterone receptor-negative [PR-
], and ErbB2-
), which generally have a poor prognosis [18
]. Several mouse models of mammary cancer display basal-type characteristics and cluster with human basal-type tumors, including Brca1+/-
, irradiated [19
, TgMMTV-Cre, p53+/-
], and some DMBA-induced tumors [7
]. In particular, of the mouse models studied, the expression signature of the C3(1)Tag GEM model appears to most closely correlate to the human basal-like subtype [8
]. Because T antigen binds to and inactivates tumor suppressor proteins Rb and p53, SV40-T/t-antigen expression induces an aggressive and genetically relevant oncogenic phenotype. In addition, Ki-ras amplification and overexpression seem to spontaneously occur during tumor progression in the C3(1)/Tag tumors and accelerate the tumor phenotype [22
]. Importantly, both human basal-like and C3(1)Tag tumors have amplifications of human chromosome 12p12 (mouse chromosome 6), a region that contains Ki-ras [8
Comparing gene expression array data across species is likely to reveal additional potentially important gene signatures involved in tumorigenesis. For instance, tumor samples from the MMTV-Wnt1 mouse model express genes represented in both basal and luminal subtypes [8
] and cross-species analyses by Herschkowitz and colleagues [8
] classified MMTV-Wnt1 mammary tumors as 'normal mammary gland'. Interestingly, other studies suggest that MMTV-Wnt1 tumors also express markers associated with progenitor cell lineages [24
], making it a potentially useful model for isolating tumor progenitor cells.
Cross-species analyses also identified a new 'claudin-low' tumor subtype [8
]. Claudin-low tumors are characterized by reduced expression of genes involved in tight junctions and cell-cell adhesion. Mouse models comprising this category include those with spindle-like morphology, including some DMBA-induced tumors, Brca1co/co
, Tg MMTV-Cre, p53+/-
tumors, and tumors from p53 null transplant model. Whether reanalysis of human breast tumor data sets uncovers the claudin-low gene signature and differentiates responses to particular treatments will be of great interest.
Despite the ability of hierarchical clustering to indicate important specific similarities between GEM models and human subtypes of breast cancer, significant differences in gene expression exist, even between mouse and human tumors considered to represent the same subtype of mammary cancer. These distinctions must be considered when applying models to the study of human breast cancer. Nonetheless, identifying the similarities in gene expression and genetic pathways through a cross-species approach highlights important conserved molecular changes involved in oncogenesis.