Metastatic lymphadenopathy from carcinogen-induced oral cavity squamous cell carcinoma (OSCC) heralds a worse prognosis for affected patients. Towards defining the molecular basis of tumor cell aggressiveness and metastasis, we analyzed a new C57BL/6 mouse OSCC model with a directed analysis of defined molecular pathways and now extend this work to an unbiased microarray based interrogation. We generated six mouse cell lines from 7,12-dimethylbenz(α)anthracene (DMBA) induced unique primary oral cancers. The cell lines were designated as mouse oral cancer (MOC1, 2, 7, 10, 22 and 23) and 5 of 6 (excluding MOC23) formed tumors when transplanted into wild-type (WT) mice. MOC23 did form tumors in immunodeficient RAG2-/- mice. Analyses of tumor growth in WT mice showed that MOC2, MOC7 and MOC10 grew more rapidly and almost always metastasized to regional lymph nodes, while MOC1 and MOC22 grew much slower and never metastasized. Towards defining molecular aberrations associated with the dichotomous growth patterns, known pathways (STAT3, NF-KB, AKT, etc) were assessed and aggressive growth was correlated with ERK1/2 activation. Further, phosphorylated ERK1/2 directly controlled CD44 expression- a molecule whose expression is associated with epithelial to mesenchymal transition (EMT) and putative cancer stem cells. Inhibiting MEK, the upstream activator of ERK1/2, led to decreased CD44 expression and promoter activity and also decreased cellular migration and invasion. Conversely, enforced activation of MEK1 led to enhanced CD44 expression and promoter activity. CD44 was required for the aggressive growth since reduction of CD44 levels significantly attenuated in vitro migration and in vivo tumor formation of MOC2 and MOC10 cell lines. These results in the mouse model were extended to freshly resected human OSCC where a strict relationship between ERK1/2 phosphorylation and CD44 expression was also identified. We have extended this work via an unbiased Illumina microarray based interrogation of the 6 cell lines, normal mouse oral keratinocytes and a cell line derived from a MOC2 bearing metastatic lymph node. Principal component analysis strictly clustered the aggressive cell lines together, whereas the indolent cell lines in general clustered with normal oral keratinocytes. Significance Analysis of Microarrays (SAM) revealed 158 genes upregulated and 77 genes downregulated in aggressive versus indolent tumors. The former gene set included several known genes associated with metastasis, e.g. MUC1, and interestingly also identified a novel panel of transcription factors and cell surface molecules. Finally, we used Gene Set Enrichment Analysis (GSEA) comparing the mouse aggressive/metastasis gene set with a previously described microarray dataset from patients with OSCC with and without lymphatic metastasis (Mendez et al., Clin. Can. Res., 17:2466-73). The mouse aggressive/metastatic signature showed strong concordance with expression data from the metastatic human tumors (normalized enrichment score of 1.44 with a nominal enrichment p-value of 0.008 and a False Discovery Rate of 1.4 %). Thus, molecular analysis of tumor aggressiveness in a novel immunocompetent mouse model was used to (1) identify CD44 as an essential mediator of the aggressive tumor-promoting effects of ERK1/2 and (2) define a genetic signature of metastasis that was concordant with gene expression profiles of metastatic human OSCC.