OBJECTIVETo screen a set of gene markers related to metastasis and prognosis of breast cancer by comparison of gene expression profiles of primary breast cancers with distant metastasis to the cases without distant metastasis within 3 years follow-up, and to explore the clinical significance of those gene expression in prognostic molecular classification of breast cancer patients.

METHODS5 cases with distant metastasis and 5 cases without distant metastasis within 3 years follow-up were used as training cases to compare their gene expression profiles by Oligo microarray hybridization containing 21 329 human functional genes. K-mean supervised cluster was done for 10 training cases and additional 20 testing cases based on the set of differential genes. "Leave-one-out" was used to eliminate useless genes to obtain optimal gene set that was used for prognostic molecular classification of breast cancer patients.

RESULTSThe different genes screened out from gene expression profiling of primary breast cancers with and without distant metastasis could classify breast cancer patients into two sub-groups. All patients with distant metastasis were included in the "poor prognosis group" (7/10), whereas there were no case showing distant metastasis in the "good prognosis group" (0/20), with a statistically significant difference by exact probability test (P =0. 03). In the set of 104 optimal genes, all 5 genes involved in cell adhesion and migration were up-regulated in cases with distant metastasis, all 2 genes related to immune response of host were down-regulated, 11 genes related to cell growth and metabolism were up-regulated and 14 down-regulated, and 15 genes related to cell signal transduction were significantly changed.

CONCLUSIONA set of genes involved in cell adhesion and migration, cell growth and metabolism, immune response mechanism, cell signal transduction were screened out by comparing gene expression profiles of primary breast cancers with and without distant metastasis within 3 years follow-up, showing highlight in prognostic molecular classification of breast cancer patients and hopeful would benefit to choose patient-tailored therapy strategies.