To understand the genetic basis and clonal evolution underlying metastatic progression of human breast cancer in vivo, we analyzed the genetic composition of 29 primary breast carcinomas and their paired asynchronous metastases by comparative genomic hybridization and fluorescence in situ hybridization. The mean number of genetic changes by comparative genomic hybridization was 8.7 +/- 5.3 in primary tumors and 9.0 +/- 5.7 in their metastases. Although most of the genetic changes occurred equally often in the two groups, gains of the Xq12-q22 region were enriched in the metastases. According to a statistical analysis of shared genetic changes and breakpoints in paired specimens, 20 of the metastases (69%) showed a high degree of clonal relationship with the corresponding primary tumor, whereas the genetic composition of 9 metastases (31%) differed almost completely from that of the paired primary tumors. In both groups, however, chromosome X inactivation patterns suggested that the metastatic lesions originated from the same clone as the primary tumor. Fluorescence in situ hybridization analysis with probes specific to metastatic clones usually failed to find such cells in the primary tumor sample. In conclusion, detailed characterization of the in vivo progression pathways of metastatic breast cancer indicates that a linear progression model is unlikely to account for the progression of primary tumors to metastases. An early stem line clone apparently evolves independently in the primary tumor and its metastasis, eventually leading to multiple, genetically almost completely different, clones in the various tumor locations in a given patient. The resulting heterogeneity of metastatic breast cancer may underlie its poor responsiveness to therapy and explain why biomarkers of prognosis or therapy responsiveness measured exclusively from primary tumors give a restricted view of the biological properties of metastatic breast cancer.