FLP, like Cre, is a frequently employed site-specific recombinase. Because wild-type FLP (wtFLP) is thermolabile, a thermostable FLP mutant (FLPe) has been developed for efficient recombination of FLP in studies using mammalian cells and animals. FLPe and wtFLP have been compared in multiple assays in vitro and in vivo, and in mouse genetics, FLPe has been shown to be very effective like Cre. Here we show an adenovirus vector (AdV) system to be valuable for quantitative measurements of the enzyme activity in mammalian cells and, using this system, precisely compare activities of wtFLP and FLPe. Unexpectedly, we found that the recombination efficiency of FLPe enzyme was lower on a molar basis than that of wtFLP even at 37 degrees C and, consequently, that the higher recombination yield per transduced AdV genome expressing FLPe compared to wtFLP was due not to inherently higher enzyme activity, but rather to higher steady-state levels of FLPe by its thermostability. Therefore, trying to increase FLPe levels further, we generated a "humanized" FLPe (hFLPe) gene with codon usage optimized for mammals. hFLPe produced about 10-fold more FLPe enzyme in transfection experiments than FLPe, as expected. However, hFLPe-expressing AdV was unstable and could not be prepared without deletion, suggesting that a subtle deleterious effect of FLP on 293 cells may exist. With hFLPe-expressing AdV thus unavailable, of the AdV constructs tested, AdV-expressing FLPe yielded the most recombined targets, despite the lower recombination efficiency of FLPe per enzyme molecule compared with that of wtFLP. We found hFLPe to be valuable for plasmid transfection, and its properties are probably suitable for experiments involving cell lines and transgenic mice.