Metastasis is increasingly gaining scientific attention, while the traditional focus on the primary tumor is still predominant. This can be easily observed in several experimental models as well as in current attempts to select therapy targets for systemic cancer after mutational or molecular profiling of primary tumors. The shared characteristic of such primary tumor-centric concepts is the exclusion of selective adaptation outside the primary tumor. They generally hold that most metastasis-associated molecular traits and therapy targets are acquired or mutated within the primary tumor and then transmitted to disseminating tumor cells. By direct ex-vivo analysis of disseminated cancer cells we found little support for these hypotheses. In general, we document frequently early dissemination and observe that disseminated cancer cells (DCCs), which we isolated from bone marrow or lymph nodes, and primary tumor cells display disparate changes on all levels of genomic resolutions, including point mutations, allelic losses, and genome wide chromosomal rearrangements, providing evidence for ectopic selective adaptation. As DCCs were isolated at the time of surgery or months to years after, they are representative (i) for cells that have disseminated until resection of their source or (ii) for cells that have ectopically survived at least for the period between resection of the primary tumor and bone marrow sampling. DCCs, detected by epithelial markers such as cytokeratin or EpCAM, in bone marrow and lymph nodes are extremely rare (frequency 10-5 - 10-6) but associated with metastatic relapse and poor outcome and therefore apparently comprise the metastatic precursor cells. While the genomic characterization of DCCs has generated important insights into the time point of dissemination and provided evidence of selection and mutation outside the primary tumor, it will be important to describe their phenotype and determine the molecular mechanisms of ectopic survival and outgrowth.