Single micrometastatic tumor cells encased in mesenchymal tissues, such as bone marrow (BM), are regarded as suitable targets for adjuvant immunotherapy since they are easily accessible for both immunoglobulins and immune effector cells. However, the antigen profile of such cells, to which antibody therapy might be targeted, cannot be deduced from the antigen pattern of the primary tumor. To evaluate the antigen profile of disseminated cells found in BM aspirates from 20 breast cancer patients, we applied a quantitative immuno-cytochemical double-marker assay and typed for 4 common tumor-associated cell-surface antigens (c-erbB-2, CO17-1A, MUC-1, LewisY). Individual breast cancer cells were identified by F(ab) fragments of the pan-cytokeratin (CK) monoclonal antibody (MAb) A45-B/B3, directly conjugated with alkaline phosphatase, which identified cancer cells as sensitively as the standard APAAP procedure (r = 0.998; p < 0.0001). CK+ cells co-expressed c-erbB-2, CO17-1A, MUC-1 and LewisY in 87%, 78%, 79% and 79% of patients, respectively; however, the frequency of double-positive cells per sample varied considerably. The mean percentage of double-positive cells per total number of CK+ cells was 41% for c-erbB-2 (range 0-92%), 47% for CO17-1A (range 0-75%), 49% for MUC-1 (range 0-67%) and 32% for LewisY (range 0-59%). In 14 of these patients, we used an antibody cocktail to type CK+ cells for the combined expression of all 4 antigens. The antibody cocktail labeled significantly more CK+ cells than each of the single MAbs alone, resulting in a mean of 71% double-positive tumor cells (34-100%). We conclude that expression of tumor-associated cell-surface antigens on micrometastatic cancer cells in BM is heterogeneous, which may limit the efficacy of monovalent immunotherapeutic strategies directed against only one particular antigen. Thus, defining target antigens expressed by the actual target cells emerges as a crucial first step in selecting appropriate therapeutic targets.