As methods of cancer diagnosis and treatment improve, interest in metastatic brain tumors continues to increase. In the present study, we attempted to characterize genetically the dynamic changes occurring during brain metastasis formation by DNA microarray, and attempted to compare these findings with histological observations. Lewis lung carcinoma cells were injected into C57BL/6Ncrj mice carotid arteries. The mice were sacrificed at days 1-9 after injection. We performed histological observation and genome-wide expression profiling using a DNA microarray. In histological observation, tumor cells were observed in capillary vessels at day 1 after injection. At day 3, the tumor cells had begun to proliferate. At day 6, the metastatic foci showed "perivascular proliferations". Next, we performed a pairwise comparison of gene expression microarray data from day 1 to day 9 after injection. The first major change occurred between Phase Two and Phase Three. When hierarchical clustering was performed between different samples using the 867 genes, they could be classified into identical clusters for days 1 and 2, identical clusters for day 3 to day 5, and identical clusters for day 6 to day 9. For time course analysis, we extracted 623 genes by the pairwise comparison. By using the quality threshold (QT) nonhierarchical clustering method, we identified 37 expression patterns. These patterns can be separated into eight clusters by using the k-means method. The microarray results reported here strongly suggest that a large number of genes exhibit a spike pattern, which is tantamount to phase-specific expression.