Background: To examine the usefulness of green fluorescent protein (GFP) mice for studying the interactions between normal cells and tumor cells in a host, we used a melanoma model in such “green” mice [C57BL/6-Tg(CAG-EGFP)1Osb mice]. Mice were given a subcutaneous injection of B16-F10 cells, and the resultant primary tumors were removed. Then cells from individual tumors were cultured. Results: The proportion of EFGP + cells was determined by fluorescence-activated cell sorting (FACS) and was 6.8% ± 3.2% (mean ± s.d.) on day 1 of culture, 0.6% ± 0.3% on day 2, and 0.02% ± 0.01% at day 7. In all cases, isolated cells grew at a constant rate, but fluorescence decreased over time and became undetectable on day 14. Cells were tested using PCR for the presence of an EGFP-specific sequence, and results were negative in all cases, thus indicating that the cells did not harbor the host's reporter gene. Cells were also tested for the presence of EGFP mRNA, which was consistently detected for 22 days after the start of culture. The tumorogenicity of the cultured cells was confirmed in GFP mice injected with cells from a selection of cultures. Conclusions: In a melanoma model in GFP mice, the detection of “green” cells in tumors was not equivalent to the detection of host-derived cells. Such “masking” was caused by a transient, but lasting, transfer of EGFP mRNA from the host's normal cells to tumor cells. Thus, an analysis of tumors postmortem by techniques that yield only a single snapshot can lead to incorrect interpretations and erroneous conclusions.