Conflicting results have been reported on the use of cytokeratin-19 (CK-19) in the detection of tumor cells in the peripheral blood of patients with solid tumors. We investigated the expression of CK-19 in lung cancer cell lines and in human lung tumor samples using a nested reverse transcriptase (RT)-PCR to determine the sensitivity and specificity of this method. In addition, blood samples of lung cancer patients and healthy controls were analyzed for the presence of CK-19 transcripts. Amplification products were visualized by ethidium bromide staining and radioactive hybridization with a CK-19-specific probe. Application of a previously described nested RT-PCR for the detection of CK-19 resulted in amplification of the processed pseudogene. Therefore, a more stringent RT-PCR was developed by increasing the annealing temperature. RT-PCR amplification products for CK-19 were detected in 38 of 41 lung cancer cell lines. The three negative cell lines were all variant small-cell lung cancer cell lines. Concordant results were observed between CK-19 detection by immunohistochemistry and by RT-PCR. In serial RNA dilution experiments, CK-19 transcripts could be detected in 18 to 80 pg of total cellular RNA in three cell lines and in 60 ng total RNA in one cell line. The nested RT-PCR had the sensitivity of detecting 50 tumor cells in 10(6) peripheral blood mononuclear cells (PBMNC), and CK-19 transcripts were randomly detected in normal PBMNC. This study shows the necessity in processing parallel samples without reverse transcriptase enzyme to avoid amplification of pseudogenes. A serious problem in the detection of tissue-specific transcripts in PBMNC is the detection of illegitimate transcription levels. In conclusion, although CK-19 may be a useful marker for the detection of lung cancer cells, its application for the detection of circulating tumor cells is not recommended.