Green fluorescent protein (GFP) has been used as a reporter molecule for gene expression because it fluoresces green after blue-light excitation. Inclusion of this gene in a vector could allow rapid, nontoxic selection of successfully transduced cells. However, many attempts by our laboratory to isolate stable retroviral producer cell clones secreting biologically active vectors containing either the highly fluorescent S65T-GFP mutant or humanized GFP have failed. Vector plasmids containing various forms of GFP and the neomycin resistance gene were transfected into three different packaging cell lines and fluorescence was observed for several days, but stable clones selected with G418 no longer fluoresced. Using confocal microscopy, the brightest cells were observed to contract and die within a matter of days. RNA slot-blot analysis of retroviral producer supernatants showed no viral production from the GFP plasmid-transfected clones, although all clones derived after transfection with an identical retroviral construct not containing GFP produced virus. Genomic Southern analysis of the GFP-transduced clones showed a much higher probability of rearrangement of the priviral sequences than in the control non-GFP clones. Overall, 18/34 S65T-GFP clones and 17/33 humanized-GFP clones had rearrangements, whereas 2/15 control non-GFP clones had rearrangements. Hence, producer cells expressing high levels of these GFP genes seem to be selected against, with stable clones undergoing major rearrangements or other mutations that both abrogate GFP expression and prevent vector production. These observations indicate that GFP may not be an appropriate reporter gene for gene transfer applications in our vector/packaging system.