Heat shock proteins (HSPs), so named because their transcription is up-regulated upon exposure to elevated temperatures, are prime candidates for targets of selection as plants experience increasingly frequent episodes of stress-inducing high temperatures. In this study, we examine genetic variation in HSP expression, its fitness consequences and its pleiotropic effects in Arabidopsis thaliana, a species that grows in environments that are thermally moderate during its growing season (Hoffmann 2002). Our investigation of the functional significance of HSPs and the evolvability of HSP expression in this model organism is intended to provide insight for both crop breeders and those interested in the evolutionary ecology of natural populations in changing environments.

In response to high temperature, plants rapidly produce proteins belonging to five HSP families: Hsp100/ClpB (eukaryotic/Escherichia coli nomenclature), Hsp90, Hsp70/DnaK, Hsp60/GroEL and small HSPs. HSPs are proposed to act primarily as molecular chaperones, refolding misfolded proteins and/or preventing accumulation and/or aggregation of nonfunctional proteins in stressed cells.

Despite their up-regulation in response to elevated temperature, no direct role in survival at high temperature has been demonstrated for most HSPs. This may result in part from genetic redundancy and consequent limits on the utility of reverse genetics for elucidating gene function. The apparently essential nature of HSP function even in nonstressful environments can further complicate the description of function, since complete gene knockouts may be lethal. For three of the four large HSP families, this is the case. Fourteen Hsp70 genes have been identified in A. thaliana. It has proven difficult to alter expression levels and maintain viability (see Larkindale et al. 2005). Hsp90, found in four cytosolic copies, is up-regulated in response to heat, but no direct role in acquired thermal tolerance has been demonstrated (Larkindale et al. 2005). A null mutant of the chloroplast Hsp60β does show increased heat sensitivity, although expression does not appear to be heat induced (Ishikawa et al. 2003).

Here, we focus on the best-studied HSP in plants, AtHsp101 (ClpB1, At1g73410; hereafter Hsp101). This HSP, a member of the Hsp100/ClpB family of proteins, is thought to be involved in disaggregation of misfolded proteins and their hand-off for either refolding by Hsp90 or Hsp70, or degradation (Wang et al. 2004). Three additional ClpB proteins have been identified in A. thaliana: ClpB2 – ClpB4. ClpB2 has recently been re-annotated as encoding a 68.8-kDa protein. There is no evidence of expression of a ClpB-sized mRNA or protein from this locus (Lee et al. 2007). ClpB3 is localized in chloroplasts, while ClpB4 is associated with mitochondria (Rottgers et al. 2002; Lee et al. 2007). Thus Hsp101/ClpB1 is the only cytosolic ClpB in Arabidopsis. ClpB3 and ClpB4 apparently diverged from Hsp101 early in the evolution of life, since their gene sequences are derived from cyanobacterial ClpBs in an ancient split from the lineage leading to Arabidopsis Hsp101, and there appears to have been substantial divergence in function as well (Lee et al. 2007). Indeed, genetic analysis has shown that Hsp101 is absolutely essential for thermal tolerance in both Arabidopsis and maize (Hong & Vierling 2000; Queitsch et al. 2000; Nieto-Sotelo et al. 2002), while ClpB3 and ClpB4 null mutants are indistinguishable from wild type in their ability to acquire thermal tolerance.

Under optimal growth conditions, Hsp101 is present in vegetative tissues in such low-copy number that it is difficult to detect by standard biochemical technique. However, increased Hsp101 transcript and protein production begins essentially instantaneously on exposure of plants to temperatures 5–10 °C above optimal growth temperature and increases dramatically in heat-stressed plants (30 °C and above in Arabidopsis) (Hong et al. 2001; Young et al. 2001). Expression levels increase with the degree of thermal stress until lethal temperatures are reached (Chen et al. 1990). Hsp101 protein expression therefore acts as a virtual thermometer. Hong et al. (2001) show that a Hsp101 null mutant lacks thermal tolerance, but is visually normal under standard (non-heat-stressed) growth conditions. Quantitative analysis of growth and reproductive output were not performed.

Together, the rapid response of Hsp101 to temperature changes and its absolute requirement for survival in extreme heat stress suggest that Hsp101 will be an essential component of plant adaptation to thermal stress in future climates. Across all population studies of HSPs in animals, the evidence suggests that the HSP expression response to thermal stress evolves in natural populations (Krebs & Loeschke 1995; Otsuka et al. 1997; McColl & McKechnie 1999; Bettencourt et al. 2002; Frydenberg et al. 2003). However, the evolutionary responses in animals are not simple and suggest that there are both costs and benefits to HSP expression. These include effects on longevity and fecundity that result in the evolution of reduced HSP induction in populations that experience chronic thermal stress (reviewed in Sørensen et al. 2003). We know almost nothing about HSP adaptation in plants.

Whether or not Hsp101 contributes to an adaptive response to increased thermal stress clearly depends on how Hsp101 expression influences fitness. In addition, we need to know that Hsp101 expression is evolvable. Evolvability has been defined as the ability to acquire and maintain new, potentially adaptive phenotypes through genetic change (after Wagner 2005 and Hansen 2006). The conventional approach to assessing evolvability has been to measure heritability within populations, defined as the ratio of additive genetic variance to total genetic variance (Fisher 1930). However, heritability measures often depend strongly on environmental conditions, and response to selection depends on an interaction between the complexities of genetic effects (additive vs. interactive) and mating structure (Whitlock et al. 1995). In addition, genetic correlations between traits, caused by pleiotropy and linkage, can constrain or eliminate adaptive response despite high heritabilities (Roff 1997), and their values can change nonproportionally with selection (Roff & Mousseau 1999). Finally, the standing genetic variation underlying heritabilities may not be qualitatively similar to the variation that underlies adaptive evolution, consisting hypothetically of mildly deleterious mutations in mutation–selection balance that contribute little to adaptive evolution (Barton & Keightly 2002). Indeed, Nordborg et al. (2005) showed that much of the observed DNA sequence polymorphism in A. thaliana is best explained in this way. Therefore, we suggest assessing evolvability by examining extant genetic variation among populations. This approach is of particular value when a cline or contrast in environmental factors is of interest as an agent of natural selection across the geographical areas sampled. Used in this way, variation among populations quantifies a sort of realized evolvability. Furthermore, it tells us how much the evolvability, together with selection and drift, has been capable of producing genetic differentiation among populations, taking into account both the availability of mutationally derived trait variation and any negative pleiotropic constraints on its evolution. In this study, we assess evolvability of Hsp101 expression by comparing a set of wild-collected genotypes of A. thaliana from contrasting portions of the species’ range.

In plants, despite considerable progress at the biochemical and genetic levels, we know little about naturally occurring variation in HSP or any resulting phenotypic or fitness effects in any species. Barua & Heckathorn (2006) demonstrate an interaction between light and temperature in field Hsp70 induction responses in Solidago altissima (Asteraceae). Manitasevic et al. (2007) quantify Hsp70 and Hsp90 expression variation in response to season and habitat in Iris pumila (Iridaceae). Knight & Ackerly (2003) provide the only published support that HSP expression evolves in natural plant populations, showing differential responses of coastal and desert congeneric species in small HSP expression in field and controlled environments. Queitsch et al. (2002) used pharmacological suppression of Hsp90 function to uncover developmental abnormalities in A. thaliana. They suggest that dysfunction in Hsp90 expression can uncover hidden genetic variation on which selection might act in periods of high stress, hypothesizing that HSPs may serve as ‘evolutionary capacitors’. Despite the broad interest generated by this hypothesis, there is little information on among-population, within-species variation in HSP expression in plants.

Functionally, our understanding of the integrated phenotypic effects of HSP and their evolvability is limited. Functional genomic studies strongly suggest that pleiotropy links multiple plant functions, sometimes in surprising ways. That HSPs seem to be associated with stress responses in general provides a strong argument that we should expect pleiotropic effects for genes associated with HSP expression and function, since they are likely to influence multiple plant pathways and traits. Pleiotropic effects can include trade-offs and hence represent constraints on adaptive responses (Waxman & Peck 1998; Tonsor et al. 2005). In animal systems, chronic exposure to high temperature has been associated with reduced high temperature HSP response, implying a cost to chronic HSP expression (Bettancourt et al. 1999). Functional studies comparing HSP-functional wild types with HSP-knockout mutants can provide a direct estimate of the fitness effects of HSP functionality and benefit/cost. When such studies also measure multiple traits known to be of ecological importance, key pleiotropic effects can be described. Comparative studies of naturally occurring variation in HSP expression and correlated variation in ecologically important traits can elucidate the extent to which multiple plant functions covary and have co-evolved with HSP expression. The results can guide strategies for plant improvement, and provide increased understanding of the mechanistic basis of plant metabolism and development in a thermally varying world.

In this study, we addressed four hypotheses aimed at a broad characterization of evolvability in Hsp101 expression in the model plant A. thaliana. The hypotheses:

The adaptive nature of the observed genetic variation in Hsp101 expression is revealed by the observed relationship between latitude of origin and magnitude of leaf Hsp101 content. Interestingly, at low latitudes, where average annual temperatures are highest, leaf Hsp101 content is actually lowest. This seems at first counter-intuitive, given the observation that Hsp101 enhances survival at high temperature. It is unlikely that Hsp101 expression has direct negative effects on fitness at low latitudes. It is required for survival at high temperatures (Hong & Vierling 2000) and enhances fitness at cool temperatures and under mild thermal stress (this study). Together, these results suggest that adaptation of Hsp101 expression to the local thermal environment cannot be driven simply by the average annual temperature of the habitat, unless Hsp101 expression actually interferes with growth and reproduction in warmer low-latitude habitats. We suggest that it is more likely to be the pleiotropic effects of Hsp101 expression that drive the evolved relationship between latitude and Hsp101 expression. We observed two important pleiotropic effects of Hsp101 tissue content that support this idea.

First, we observed a highly significant exponentially decaying relationship between Hsp101 content and allocation to roots. As Hsp101 content increases, allocation to roots declines asymptotically to a level beyond which further increases in Hsp101 do not appear to further suppress root growth. Our results are corroborated in two additional studies. Nieto-Sotelo et al. (2002) showed this same suppressive effect on hypocotyl growth in corn seedlings by comparing a null Hsp100 mutant to wild type. More recently, McClellan et al. (2007) showed that increases in Hsp101 content mediated by a fungal endophyte decrease root growth in Arabidopsis thaliana. Thus, induction of high levels of Hsp101 may exacerbate the effects of drought stress by decreasing the plant’s water uptake capacity through root growth suppression. We suggest that this Hsp101–root growth relationship may represent an adaptive trade-off when the species is found in conditions where both high temperature and low water availability coincide. Exposure to high temperature is often accompanied by water stress (Moffat 2002), and their combined effects have been long known to be synergistically negative (Savage & Jacobson 1935; Rizhsky et al. 2004). Furthermore, our results imply a negative pleiotropic effect of Hsp101 on water use. Combined high thermal stress and drought stress result in lowered expression of loci associated with thermal tolerance alone, also suggesting negative pleiotropic effects of some elements of acquired thermal tolerance on drought tolerance (Rizhsky et al. 2004).

These experiments conducted either short-term measures of hypocotyl growth response (Nieto-Sotelo et al. 2002 and McClellan et al. 2007) or long-term up-regulation of Hsp101 expression and lifetime measures of root allocation (this study). The duration of root growth effects resulting from up-regulation of Hsp101 for short periods of time and the frequency, timing and duration of such high temperature effects in natural populations remain unclear. If this species experiences temperature extremes of short duration in the middle to the end of the growing season when the root system is well established, there may be little effect on root growth. However, if temperatures climb early in the plant’s life cycle and remain high, there could be substantial effects on the root : shoot ratio with consequent difficulty in maintaining water balance.

An additional pleiotropic effect related to plant water status was observed in the higher transpiration rates of the null mutant. This result further supports the hypothesis that chronically up-regulated Hsp101 expression influences the plant’s water uptake capacity. It also suggests a second fitness trade-off associated with Hsp101 expression: the increased thermal tolerance benefit associated with the Hsp101 biochemical function of remediating temperature-induced misfolding and aggregation of proteins likely comes at the cost of reduced ability to resist high temperature stress through evaporative cooling. This hypothesis deserves further investigation.

In addition to the potential negative plieotropic effects discussed above, our study provides considerable evidence indicating that the production of HSPs is costly. First, ecotypes from more northern or high-elevation habitats have steeper induction response curves across our experimental temperature gradient than ecotypes from southerly, lower elevation sites, suggesting that plants exposed to chronic high temperatures evolve other mechanisms for coping with thermal stress. This could be both because HSP synthesis and activity require energy and/or because of the considerable nitrogen demand imposed by HSP expression (Heckathorn et al. 1996; Feder & Hofman 1999). Together these results suggest HSPs are a costly ‘rapid response’ system, both energetically and in terms of fitness. Long-term exposure to elevated temperatures leads to the evolutionary adjustment of thermal tolerance through other, presumably less costly mechanisms in animals (Sørensen et al. 2001). Further studies in which precise measurement of the thermal and hydrological environment of multiple populations can be correlated with the induction response of Hsp101 will give valuable insight into the evolutionary engineering of optimal thermal tolerance responses.

Despite earlier work establishing that development and germination in a normal thermal environment occur regardless of the presence of a functional Hsp101 gene (Hong & Vierling 2001; Nieto-Sotelo et al. 2002), this study shows more subtle but biologically significant negative fitness effects of the loss of Hsp101 function. In the Columbia genotype, loss of Hsp101 function results in significantly reduced fitness. Since this is true whether the plants experience the control thermal regime (15/23 °C day/night) or the Hsp101 induction environment (15/23 °C day/night +2−3 h at 35 °C mid-day 3× per week), Hsp101 appears to have important functions both at high temperatures (Hong & Vierling 2001) and in normal growing conditions.

The thermal tolerance function of Hsp101 at high temperature has also been observed in yeast and bacteria. For example, Saccharomyces cerevisiae with deletion of the gene encoding the Hsp101 homologue, Hsp104, fail to grow when exposed to extreme stress conditions (e.g. high temperature or excessive concentrations of ethanol; Sanchez & Lindquist 1990; Sanchez et al. 1992). This phenotype is believed to arise from the fact that Hsp104 is required to re-activate unfolded, aggregated proteins, which may accumulate and become toxic under these conditions (Parsell et al. 1994).

The importance of Hsp101 to fitness in normal growth environments, as seen here, was not predicted. Previous studies using Hsp deletion genotypes in yeast and Escherichia coli homologues report no effect on cell growth in the normal growth environments (Sanchez & Lindquist 1990; Squires et al. 1991; Sanchez et al. 1992). In contrast, our results show that even the low amounts of Hsp101 in vegetative tissues detected under normal growth conditions have a significant impact on fitness in the Columbia background. The enhanced fitness effect seen in Columbia was not significant in the Landsberg erecta genomic background. ‘Wild type’ Ler does not appear to express much, if any, Hsp101 in the normal thermal environment (and has substantially lower fitness than wt Col). The difference in Hsp101 content between null mutant Lerhot1-3 and the wild-type Ler is therefore quite small. Landsberg erecta (Ler) was derived from the original Laibach Landsberg through mutagenesis, and is likely to contain multiple mutations. It is possible that one of these mutations affects Hsp101 expression.

Although thermal tolerance involves a number of systems besides HSPs, notably abscisic acid, salicylic acid and probably calcium (Larkindale et al. 2005), our use of the Hsp101 null mutants and corresponding wild types means that we can reject those elements of the thermal tolerance response as direct causes for the significant phenotypic differences between genotype pairs containing functional forms of Hsp101 vs. those that harbour mutant alleles. Nevertheless, genetic differences between Col and Ler in these additional pathways could, through genetic background specific pleiotropic effects, mediate the differences between Ler and Col in Hsp101 effects observed in this study.

The level of Hsp101 expression in the two laboratory lines, Columbia and Landsberg, differed significantly, both in a benign thermal environment, and when up-regulated upon repeated exposure to 35 °C. In the future, it will be valuable to clarify which gene(s) control(s) this differential expression. The effect may arise from genetic differences in the composition and activation of the Hsp101-promotor, by altered rates of protein turn-over, or through some other element(s) of the signal–response pathway. In summary, although the key regulatory agent or pathway remains unknown, it is likely that the differing effects in the two genetic backgrounds are the result of the very limited up-regulation of Hsp101 expression in the Ler background compared to the induction response seen in the Col background.