Metastasis of human head and neck cancer is a multistep and highly heterogeneous process requiring activation and deactivation of multiple and specific genes. To identify these genes, we established highly metastatic head and neck squamous cell carcinoma (HNSCC) cell lines from poorly metastatic HNSCC cells through in vivo selection using a lymph node metastatic mouse model. The very close genetic relationship between these highly metastatic cell lines and the parental cell line provided an excellent model for differential gene expression analysis using cDNA microarrays. Comparison of 6 cell lines established individually from the lymph node metastases with their poorly metastatic parental cell line revealed 33 differentially expressed genes. Some of these genes are involved in cellular signal transduction and matrix modeling. Differences in expression of members of the tumor necrosis factor, interleukin, caspase, and matrix metalloproteinase families were also examined. We found that two upregulated genes participated in the NF-kappaB regulatory pathway. Furthermore, differences in gene expression between six cell lines derived from primary tumors and six cell lines derived from lymph node metastases in the mouse model were analyzed statistically. Tissue growth factor-beta and tumor necrosis factor-related genes showed significantly altered expression in cells derived from lymph node metastases as compared with cells derived from primary tumors, suggesting that the differential growth advantage of metastatic cells requires more aggressive responses to their environment, such as a lymph node tissue.