SUMMARY:The suitability of "real-time" quantitative reverse transcriptase polymerase chain reaction (RT-PCR) for the detection of isolated carcinoma cells in bone marrow was investigated by evaluating the expression of cytokeratin (CK)7, CK8, CK18, CK19, and CK20 in 17 gastrointestinal cancer cell lines, 64 control bone marrow specimens from noncancer patients, and 30 bone marrow specimens from patients with gastric or colorectal cancer. RT-PCR products for CK8 and CK18 were detected in all cancer cell lines, but only 16, 5, and 11 cell lines provided evidence for CK19, CK7, and CK20 transcription. Variable numbers of bone marrow specimens from noncancer patients demonstrated background transcription of CK8 (78.1%), CK18 (95.3%), CK19 (35.9%), CK20 (29.6%), and CK7 (16.7%). Maximal background transcription for CK8, CK18, and CK19 ranged from 52.2 to 56.1 copies/10 3 copies glyceraldehyde-3-phosphate dehydrogenase (GAPDH), the corresponding values of 0.06 and 0.76 copies for CK7 and CK20 being distinctly lower. When maximal background values were used as a threshold value to define positivity in tumor cell dilution experiments, sensitivity levels of one tumor cell in 10 4 bone marrow cells were determined for CK7 and CK20 RT-PCR assays. Maximal background expression values of the different CKs as obtained in the control series were exceeded once (CK20), twice (CK18 and CK19), and 18 times (CK7) in bone marrow specimens from cancer patients, with none of these specimens exceeding the maximal background expression value of CK8. We conclude that RT-PCR for CK8, CK18, and CK19 cannot be recommended for the detection of isolated tumor cells in bone marrow of cancer patients. On the other side, the limited number of gastric and colorectal cancer cell lines expressing CK7 and CK20 indicates that assay sensitivity for these CKs might be limited because of their selective expression by carcinoma cells. (Lab Invest 2001, 81:1351-1361.