Abstract
Hsp90 controls dramatic phenotypic transitions in a wide array of morphological features of many organisms. The genetic-background dependence of specific abnormalities and their response to laboratory selection suggested Hsp90 could be an 'evolutionary capacitor', allowing developmental variation to accumulate as neutral alleles under normal conditions and manifest selectable morphological differences during environmental stress. The relevance of Hsp90-buffered variation for evolution has been most often challenged by the idea that large morphological changes controlled by Hsp90 are unconditionally deleterious. To address this issue, we tested an Hsp90-buffered abnormality in Drosophila for unselected pleiotropic effects and correlated fitness costs. Up to 120-fold differences in penetrance among six highly related selection lines, started from an initially small number of flies and rapidly selected for and against a deformed eye trait (dfe), did not translate into measurable differences in any of several tests of viability, lifespan or competitive fitness. Nor were 17 dfe Quantitative Trait Loci (QTL) associated with fitness effects in over 1,400 recombinant lines. Our ability to detect measurable effects of inbreeding, media environment and the white mutation in the selection line backgrounds independent of dfe penetrance suggests that, within the limitations of laboratory tests of fitness, this large morphological change controlled by Hsp90 was selectable independent of strong, correlated and unconditionally deleterious effects--abundant, polygenic variation hidden by Hsp90 allows potentially deleterious alleles to be readily replaced during selection by less deleterious alleles with similar phenotypic effects. Hsp90 links environmental stress with the expression of developmental variation controlling unprecedented morphological plasticity. As outlined here and in the companion paper of this issue, the complex genetic architecture of Hsp90-buffered variation supports a remarkable modularity of Hsp90 effects on quantitative and qualitative phenotypes, consistent with the 'Hsp90 capacitor hypothesis' and standard quantitative genetic models of threshold traits.
