www. landesbioscience. com Cell Cycle 167 inhibition. Some traits were clearly specific to the genetic background, suggesting that previously hidden genetic variation was responsible. Indeed, this was demonstrated for the two traits investigated in detail; both could be enriched in a population through selection on underlying genetic variation. Similar to the results of Sollars et al., selection over several generations resulted in near fixation of an eye and a wing phenotype and, after many rounds of selection, expression of the traits did not require a continued reduction in Hsp90. However, both traits were as readily inherited through males as through females, in contrast with the data of Sollars et al. Furthermore, Rutherford and Lindquist conducted the litmus test for the genetic basis of the traits: outcrossing the selected flies exhibiting high trait penetrance to unselected lines heterozygous for the Hsp90 mutation, thereby halving the enriched predisposing genetic polymorphisms. Flies receiving two wild-type Hsp90 alleles did not display the trait under physiological conditions, whereas a significant number of their heterozygous siblings with reduced Hsp90 function exhibited the phenotype. That is, continued expression of the trait when Hsp90 function was normal required the enrichment of underlying genetic variation. Should the trait have been epigenetically inherited, it would not have disappeared in outcross progeny with wild-type Hsp90 levels. Rutherford and Lindquist did not propose that all Hsp90-dependent phenotypes require hidden genetic variation, only that many likely do and that the two studied in detail surely did. It must also be noted that the phenotype examined by Sollars et al. does not have a purely epigenetic origin, as it depends on the presence of the predisposing mutant krüppel allele. Studies with Arabidopsis in the Lindquist lab also revealed a wide variety of traits when Hsp90 was reduced. Some of these were due to genetic variation, but others were non-propagatable in isogenic lines, indicating that Hsp90 buffers normal development from the destabilizing effects of stochastic processes. 8 Sollars et al. now add epigenetic inheritance to the known panoply of Hsp90-dependent phenomena, previously including altered developmental stability and buffered genetic variation. Given the remarkable variety of traits produced upon the reduction of Hsp90 in both Drosophila and Arabidopsis, it is of great interest to determine the relative contribution of epigenetic versus genetic mechanisms, and, indeed, the capacity of the two to work in concert. Here we hypothesize a mechanism to explain the epigenetic inheritance of the ectopic outgrowth phenotype reported by Sollars et al. and the as yet undefined link between Hsp90 and trithoraxgroup proteins, drawing together several different lines of investigation from a variety of sources. First, the absence of the locus verthandi results in 55% ectopic outgrowth in the F1 generation of flies containing the KrIf-1 mutation, while the Hsp90 alleles and trithoraxgroup proteins produce a significantly smaller penetrance (5–15%). So how might mutations at vtd affect phenotype so dramatically and in a heritable manner? One clue might be the unusual nature of this locus. As deletion alleles of vtd dominantly suppress polycomb mutations, vtd was originally defined as a trithorax gene. 9 Trithorax and polycomb genes maintain patterns of Hox developmental gene expression, with trithorax-group proteins generally functioning as activators of transcription and polycomb-group proteins generally functioning as repressors. 3 The vtd locus has been cytogenetically mapped near to the centromeric heterochromatin of …