We discovered naturally occurring mutants of the Hsp83 gene with P element insertions in the gene's proximal promoter region. These mutants occur in three out of 42 populations that we examined, and at modest allele frequencies, not exceeding 7.7 percent. The mutations down-regulate Hsp83 gene expression by 32 percent to 40 percent, depending on the population. They reduce competitive fitness, female fecundity, and longevity. We also found that Hsp83 (and possibly linked loci) strongly influences the expression of cryptic deleterious genetic variation in inbred populations. That is, flies that carry the mutant Hsp83 allele are much more poorly buffered against such variation than flies with the wild-type allele. We found that even mild thermal stress can completely break down the impaired buffering associated with mutant Hsp83. Specifically, inbred fly populations with mutant Hsp83 alleles go extinct at modestly elevated temperatures of 28°C.
To our knowledge, this work is the first to show that naturally occurring regulatory mutations of the Hsp83 gene-a key modulator of the stress response and cellular signaling-can affect reproductive success, reduce longevity, and reduce variation buffering in fruit flies derived from wild populations. A previous study on naturally occurring Hsp83 variants had identified a nonsynonymous deletion mutation in the Hsp83 coding region, and focused on the effects of this mutation on morphological traits, not on fitness components . In addition, this study perturbed variation buffering only through thermal stress exposure, not through inbreeding, which is especially useful to reveal expression of recessive cryptic genetic variation. Our limited sequence polymorphism analysis found only one synonymous mutation in the Hsp83 coding region in the flies we studied, thus making it unlikely that the same deletion mutation stands behind our observations.
Our naturally occurring variants of Hsp83 have fairly mild effects compared to some of the drastic perturbations that earlier laboratory experiments used [7, 11, 16, 22, 26, 32, 33]. For example, the reduced expression of Hsp83 in our mutants did not significantly affect thermotolerance in outbred flies (See additional file 5). Working with these mutants has several benefits. First, it avoids pharmacological manipulation or structural modification of Hsp90, both of which may have unknown side effects, for example, by changing the protein's molecular interaction partners [1, 25]. Second, it avoids the use of mutations with drastic expression effects, such as engineered homozygous loss-of-function mutants [31, 32]. Such mutants can only be maintained as heterozygotes in the laboratory [16, 33]. A continuous inbreeding experiment like the one we performed to reveal deleterious cryptic variation would be difficult with such drastic mutations, partly because homozygote offspring would be lethal. In contrast, the mild variants we found enabled us to ask whether more Hsp83 means better buffering . Finally, they allowed us to study flies derived from natural populations, and to examine whether laboratory findings on Hsp90 apply to such populations.
A potential disadvantage of working with natural populations and the mutants that they contain is that each mutant occurs in a different genetic background, and that this background can influence observations. For example, we found that the genetic background of each geographic population influenced the competitive ability and fecundity of the flies we studied. However, our key observations, for example that Hsp83 can buffer deleterious variation in inbreeding lines, and that Hsp83 mutants show reduced fitness, were consistent across all three genetic backgrounds.
P element insertions are widespread in the Drosophila genome, and they are especially abundant in heat-shock genes [34, 35]. In Japanese natural populations, for example, at least 6 distinct P element insertions exist in the heat shock gene Hsp26, and at least 5 distinct insertions exist in the heat-shock gene Hsp27 at high population frequency . These observations stand in stark contrast to the low frequency of insertions into the Hsp83 gene. They suggest that the low frequency of P-element insertions we observe in Hsp83 are not a consequence of a recent insertion, but that it results from natural selection against mutant Hsp83. Our observation that these insertions are possibly associated with reduced competitive fitness, reduced fecundity, and shorter life span are fully consistent with this suggestion. The question then arises why we observe any individuals that carry P-element insertions in our study populations. One possible answer is that the insertion may be neutral or even beneficial when heterozygous or when at low population frequencies. Our observation that mutant allele frequencies did not decline in a competition assays seeded with only 10 percent of Hsp83
alleles are consistent with this possibility. Moreover, the fitness effect of Hsp83
may depend on the environment to which a population is adapted, as has been observed for P element insertions in other heat shock genes [35, 54]. In addition, some P-element insertions can help maintain tradeoffs between stress resistance and developmental homeostasis of flies living in a changing environment [34, 35, 37]. Unfortunately, our data does not allow us to answer which of these possibilities is responsible for the maintenance of P element insertions at modest frequencies in our study populations.
Our observation that mutations in Hsp83 affect fecundity is consistent with other reports that Hsp83 is involved in reproductive functions. Hsp83 plays a critical role in the process of both oogenesis  and spermatogenesis . For example, Hsp83 RNA is a component of the posterior polar plasm , and Hsp90 protein is required for localization of maternal mRNA to the posterior pole, which is essential for development of germ cells in the Drosophila embryo . Thus, Hsp83 is involved in the molecular pathways responsible for oogenesis and spermatogenesis in D. melanogaster.
Inbreeding increases the fraction of homozygous loci in a genome. Because many alleles are recessive and deleterious in the homozygous state , inbreeding will cause previously cryptic (heterozygous) deleterious variation to be expressed . Such deleterious variation manifests itself as a reduction in one or more fitness components, such as fecundity. In our inbreeding experiments, we found that wild-type Hsp83 flies were better buffered against the deleterious effects of inbreeding than mutant flies. Specifically, wild-type Hsp83 flies from all three geographic populations showed no decline in fecundity after three generations of inbreeding, and only one population showed a small decline (by 14 percent) after four generations. In stark contrast, mutant lines from all three geographic populations showed a significant decline in fecundity of up to 47 percent after between two and four generations. Genetic polymorphisms are widespread in many populations [47, 59], but the incidence of cryptic variation is usually unknown for wild populations [14, 15]. Inbreeding can reveal such variation. Our experiments demonstrate that cryptic variation must be abundant in the populations we studied, because inbreeding reduced fecundity substantially. Moreover, the experiments show that expression of Hsp83 at wild type levels can buffer the damage caused by inbreeding.
The observation that a chaperone can buffer deleterious variation is not unprecedented. Over-expression of GroEL, a molecular chaperone of Escherichia coli, can help overcome the accumulation of deleterious mutations that occur in E. coli strains with high mutation rate . In other words, a chaperone can buffer these organisms against deleterious mutations [17, 20]. Although the sources of deleterious variation-inbreeding and mutation accumulation-and the chaperones-Hsp90 and GroEL-differ in these two organisms, chaperones have the same qualitative effect in both cases. However, the E. coli strain in which these previous experiments were conducted is a laboratory strain. Observations made with this strain need not necessarily transfer to populations in the wild.
Three main observations lead us to think that the buffering capacity of Hsp90 is relevant for wild fly populations. First, the flies in which we identified the effects of the Hsp83 mutation all stem from wild populations. Second, although the buffering capacity of Hsp90 varies with genetic background, significant buffering did occur in all three populations. Third, the regulatory mutation in Hsp83 (and possibly linked loci) is mild in its effect on expression, but strong in its effects on fitness, i.e., on fecundity, longevity, and most importantly, buffering capacity. Mutations that affect these aspects of fitness are likely to be selected against in nature . More specifically, because Drosophila has very large effective population size in the wild , fitness differences much smaller than we detected would be visible to natural selection [14, 59], and would lead to the eventual elimination of mutant alleles with low fitness.