In cancer patients detection of systemic disease is of great importance to obtain prognostic information and to guide therapy. Bone marrow (BM) seems to be a common homing tissue for the early spread of tumor cells from various epithelial tumors; however, verification of the prognostic significance of BM-disseminated tumor cells (BM-DTCs), is restricted to breast cancer so far. These cells may be dormant for a long time, and signals triggering their activation leading to recurrence remain to be characterized. A recent study involving metastatic breast cancer patients reported that the shortest disease-free survival is correlated with cytokeratin (CK)-negative BM aspirates and that CK-positive BM-DTCs correspond to dormant tumor cells. Soluble CK fragments in serum including CK18 and 19 (measured as TPS and CYFRA 21-1, respectively) and caspase-cleaved CK18 are widely used to monitor tumor progression and response to therapy, actually indicating proliferation and/or necrotic/apoptotic cell death. In order to assess the source of the CK fragments, we used determinations of CK18 and caspase-cleaved CK18 fragments in serum samples before and after radical tumor surgery in colon cancer patients. Elevated serum concentrations of CK18 were found to persist in patients with a high incidence of BM-DTCs, and high perioperative levels of caspase-cleaved CK18 fragments were detected in patients with early relapses, respectively. These results indicate that in some patients at increased risk of recurrence disseminated cell populations exist that are responsible for the release of the bulk of CK fragments after removal of the apparently nonmetastatic tumor. In good agreement with the results in metastatic breast cancer patients, release of CK18 or 19 fragments by BM-DTCs seem to indicate disseminated tumor cells mainly in a dormant state, whereas caspase-cleaved CK18 may indicate skipping of this latent phase and early progression. Therefore, caspase cleavage of CKs in intact tumor cells seems to accompany or is involved in the differentiation leading from dormant to progressively active disseminated tumor cells. Release of respective CK fragments would result in an apparent clearing of CK-positive cells in BM, leaving malignant cells that have possibly undergone an epithelial-mesenchymal transition. Micrometastatic cancer cell lines derived from breast cancer patients were found to display loss of epithelial CK8, 18 and 19 as well as ectopic expression of vimentin as in mesenchymal cells. In conclusion, degradation of CKs may represent a marker indicating reactivation of dormant tumor cells in BM.