Intraspecific variation in thermal tolerance and heat shock protein gene expression in common killifish, Fundulus heteroclitus

The pervasive effects of temperature on biochemical and physiological processes are thought to play a fundamental role in shaping the distribution and abundance of organisms. This is particularly true for ectotherms such as fish. Although studies comparing species that are widely divergent phylogenetically have revealed major patterns of thermal adaptation,understanding the role of temperature in establishing fine-scale patterns of thermal tolerance in closely related species or among populations within a species can provide additional insights into the nature of adaptive variation in thermal tolerance (Somero,2002). The relationship between environmental conditions and variation in thermal tolerance, both intraspecifically and between closely related species, has been extensively studied in Drosophila (reviewed by Hoffmann et al., 2003). Comparisons of thermal tolerance of lab-bred flies derived from distinct geographic locations, however, have produced conflicting results. In some cases, geographic comparisons of thermal tolerance traits indicate that differences among populations are consistent with what is predicted by local environmental thermal regimes (Krebs and Loeschcke, 1995; Guerra et al., 1997; Sorensen et al.,2001). Other comparisons where latitudinal variation in thermal tolerance is expected have failed to reveal such differences(Davidson, 1990; Kimura et al., 1994). Similarly, in fish, evidence for intraspecific variation in thermal tolerance is mixed. Many studies have shown that populations of a species sampled from differing thermal environments and acclimated to common temperatures have thermal tolerance limits such that fish from cooler latitudes exhibit lower tolerance limits than their warmwater counterparts(Hart, 1952; McCauley, 1958; Otto, 1973; Fields et al., 1987; Lohr et al., 1996; Strange et al., 2002). On the other hand, some studies have failed to show thermal tolerance differences in populations from thermally contrasting environments(Brown and Feldmeth, 1971; Elliott et al., 1994; Smale and Rabeni, 1995).

Common killifish (Fundulus heteroclitus), inhabit estuaries and salt marshes along the east coast of North America through a latitudinal temperature gradient, and thus have been studied extensively as a model to investigate mechanisms of thermal adaptation. It has been shown that substantial variation exists within the species in morphological, molecular,genetic and physiological traits (reviewed in Powers et al., 1993; Powers and Schulte, 1998; Schulte, 2001). This variation shows significant directional change with temperature/coastal latitude such that two distinct regional subspecies have been suggested - the northern form, Fundulus heteroclitus macrolepidotus, occurring from the Gulf of St Lawrence, Canada to New Jersey, USA, and the southern form, Fundulus heteroclitus heteroclitus, distributed from Virginia, USA to the North-eastern coast of Florida, USA (Morin and Able, 1983). At the extremes of the species' range, northern fish experience temperatures ranging from -1.4°C to 21°C, whereas southern fish encounter temperatures ranging from 7°C to 31°C, and monthly mean temperatures are on average, 13°C higher in the south than in the north at any given time of year [calculated from NOAA NERRS Data, Sapelo Island, GA and Wells Inlet, ME, USA (NOAA NERRS, 2004)]. Clearly, the ability to acclimate to seasonal and daily temperature fluctuations is critical to all killifish populations, but the temperature range over which they must make adjustments is very different between populations and seasons. Early work on the thermal tolerance limits of killifish confirmed that these fish do in fact possess the ability to acclimate and tolerate a wide range of temperatures(Bulger, 1984; Bulger and Tremaine, 1985). This work, however, was performed on a single killifish population from Virginia (southern subspecies) acclimatized to seasonal photoperiod and temperature combinations, and only upper thermal tolerance limits were quantified. These data, although useful in describing patterns of acclimation/acclimatization and tolerance for a single killifish population,provide no information about the nature of, or mechanisms involved in,intraspecific variation in thermal performance.

A number of physiological and biochemical traits that are influenced by temperature and may play important roles in thermal performance of organisms have been proposed (Hochachka and Somero,2002). At the molecular level, many candidate genes have been identified as potential targets of adaptive evolution to temperature (reviewed in Hoffmann et al., 2003; Somero, 2005). In particular,heat shock proteins (Hsps) are thought to play an ecologically and evolutionarily important role in thermal adaptation(Parsell and Lindquist, 1993; Feder and Hofmann, 1999). As molecular chaperones, Hsps interact with proteins that are in their non-native conformation (stress denatured) in such a way that they prevent these proteins from interacting inappropriately with one another (for a review, see Lindquist, 1986; Hightower, 1991; Morimoto, 1998). It has been clearly shown that a species' threshold for Hsp expression is correlated with the levels of thermal stress they naturally experience, and that natural fluctuations in environmental temperatures are sufficient to elicit the heat shock response (Roberts et al.,1997; Tomanek and Somero,1999; Buckley et al.,2001). Taken together, these findings suggest that Hsps are good candidates as ecologically relevant mechanisms used by animals to ameliorate thermal stress and likely have important roles in thermal adaptation.

It has long been known that heat shock proteins are encoded by multiple genes and are assigned to families based on sequence similarity and molecular mass. Two important Hsp families are Hsp90 and Hsp70, each containing several members, some of which are expressed constitutively under normal physiological conditions (i.e. Hscs) and some of which are induced in response to protein-denaturing stress (i.e. Hsps)(Gething, 1997). Members of both the Hsp90 and Hsp70 families are known to be important in folding of nascent polypeptides as well as renaturation of heat damaged proteins(Morimoto and Santoro, 1998). However, our understanding of the true biochemical diversity of the heat shock response in an ecologically relevant context remains limited, because much of the early work addressing this question was performed using one-dimensional gel electrophoresis, which often fails to discriminate among related heat shock proteins within a family. As a result of these technical limitations most of the work that has comprehensively addressed the relationship between the heat shock response, whole-organism thermal tolerance and population distribution and abundance has been performed on a few well-characterized model species (Feder and Hofmann,1999; Sorensen et al.,1999; Sorensen et al.,2001; Sorensen et al.,2005; Michalak et al.,2001) (but see also White et al., 1994; Norris et al.,1995; Tomanek,2005).

In this study, we quantified thermal tolerance and investigated the mRNA expression patterns of a variety of isoforms of heat shock proteins (Hsps) in killifish populations in order to address several questions. (1) Are there differences in thermal tolerance between killifish populations that correlate with latitudinal temperature ranges? (2) Is intraspecific variation in thermal tolerance related to differences in the sequence of hsp genes between populations? (3) Is intraspecific variation in thermal tolerance related to differences in the expression patterns of hsp genes? (4) Are differences in hsp expression patterns between populations consistent across multiple hsp genes? By addressing these questions we are able to provide valuable insight into how local adaptation can occur between fish populations from two different environments even when the local environments are themselves highly variable.

Adult killifish of the northern subspecies (Fundulus heteroclitus macrolepidotus Walbaum) were collected from three locations: Hampton, New Hampshire, USA (NH; 42° 54′ 46″ N), Salsbury Cove, Maine, USA(ME; 44° 25′ 54″ N) and Antigonish, Nova Scotia, Canada (NS;45° 37′ 0″ N). Fish of the southern subspecies (Fundulus heteroclitus heteroclitus L.) were also collected from three sites:Brunswick, Georgia, USA (GA; 31° 7′ 31″ N), Whitney Island,Florida, USA (WI; 29° 39′ 34″ N) and Fernandina Beach,Florida, USA (FB; 30° 40′ 51″ N). All collections were made in late spring of 2002 [northern (NH) and southern (GA) thermal tolerance experiments] or 2004 [interpopulation thermal tolerance and heat shock experiments]. Fish were held in 75 l glass aquaria with biological filtration at 20 ppt salinity, 20±2°C, and 12 h:12 h L:D photoperiod for a minimum of 3 weeks before the experimental acclimations described below. Fish were fed TetraMin® fish flakes supplemented with commercial trout chow(PMI Nutrition International, Brentwood, MO, USA) daily to satiation, but were not fed for 24 h prior to experimental trials. Treatment of all experimental animals was in accordance with the University of British Columbia animal care protocol #A01-0180.

Upper and lower thermal acclimation limits were quantified for northern(NH) and southern (GA) killifish populations using chronic thermal tolerance methodology. Following a two-week holding period at 20±0.5°C, 30 fish from each population were subjected to either increasing or decreasing water temperatures of 0.5°C per day. This rate is slow enough to allow the fish's thermal acclimation to keep pace with the temperature change but yet is ecologically realistic (Bennett et al.,1997). The respective chronic thermal maximum or minimum value is taken as the high or low temperature at which 50% morbidity is observed(Fields et al., 1987; Bennett and Beitinger,1997).

Temperature tolerance in killifish populations was determined using the critical thermal methodology (CTM). The critical thermal maximum (CTMax) and critical thermal minimum (CTMin) are typically defined as the upper and lower temperatures, respectively, at which fish lose the ability to escape conditions that will ultimately lead to death(Cox, 1974; Becker and Genoway, 1979; Beitinger et al., 2000). The CTM test chamber consisted of a plastic rectangular water bath(50×35×15 cm) containing 10 individual 1 l plastic test beakers. The water bath was filled with dilute ethylene glycol that could be heated or cooled with an immersion coil connected to a Lauda RM6 benchtop unit, and circulated with a Mag-Drive model 1.5 pump to ensure complete mixing. Each beaker was filled with seawater and individually aerated to maintain oxygen concentrations at saturation and prevent thermal stratification during the trials. Beaker temperatures were monitored with Fisherbrand® NIST certified mercury thermometers (Fisher Scientific, Nepean, ON, Canada) and heating/cooling rates were between 0.28 and 0.33°C min^-1 for all trials.

Loss of equilibrium (LOE) was chosen as our experimental endpoint, and critical thermal maxima and minima were calculated by taking the arithmetic mean of the LOE temperatures for each acclimation group(Cox, 1974; Beitinger et al., 2000). At the end of each trial, fish were weighed (wet mass ± 0.1 g), measured(total length ± 0.1 cm), and returned to their acclimation conditions for recovery. We achieved >95% post-trial survival in all acclimation groups.

We assessed the relationship between acclimation temperature and upper or lower thermal tolerance of northern (NH) and southern (GA) killifish by estimating CTMax and CTMin of fish acclimated to one of seven constant temperature treatments ranging from 2.3°C to 34.0°C. Acclimation temperatures were controlled with Fisherbrand® NIST traceable temperature controllers and Ebo Jager 250 watt submersible heaters. Killifish were acclimated for a minimum of 21 days to each treatment temperature under a 12 h:12 h (L:D) photoperiod and 20 ppt salinity. Three replicate 75 l acclimation tanks per temperature treatment were divided to house 10 northern fish on one side and 10 southern fish on the other. Five fish from each population per acclimation tank were randomly chosen to be in either a CTMax or CTMin trial. In total, 30 fish from each population (N=15, CTMax and N=15, CTMin) were tested from each acclimation group. All CTM trials were run between 10.00 am and 2.00 pm to minimize any effects of daily rhythms in thermal tolerance.

In a second experiment, we explored intraspecific variation in thermal tolerance between replicate northern and southern killifish populations. Three northern populations (NH, ME and NS) and three southern populations (GA, WI and FB) were acclimated to 22±0.25°C under identical experimental conditions to those described above. Thermal tolerance trials and calculations were performed as previously described.

Several genes of interest were cloned from killifish tissues including hsc70, hsp70-1 and hsp70-2 (gill, liver and/or spleen), hsp90α (liver) and hsp90β (gill). The intronless hsp70 genes were cloned from both cDNA and genomic DNA, whereas all other genes were cloned from cDNA only. Genomic DNA was isolated from killifish spleens either by proteinase K digestion followed by phenol:chloroform extraction essentially as described elsewhere(Sambrook et al., 1989), or by the salting out method (Medrano et al.,1990).

Total RNA was extracted from either heat shocked or control tissues using the guanidine isothiocyanate method outlined elsewhere(Chomczynski and Sacchi, 1987)using TRIzol® Reagent (Invitrogen Life Technologies, Burlington, ON,Canada). Following isolation, RNA was quantified spectrophotometrically and electrophoresed on an agarose-formaldehyde gel (1% w/v agarose, 16%formaldehyde) to verify RNA integrity. RNA was stored at -80°C. First strand cDNA was synthesized from 5 μg total RNA using oligo(dT₁₈) primer and RevertAid™ H Minus M-MuLV reverse transcriptase as per the manufacturer's instructions (MBI Fermentas Inc.,Burlington, ON, Canada). cDNA was stored at -30°C for up to 1 month or at-80°C for longer-term storage.

Partial hsp90 sequence was obtained using primers determined from conserved regions of European sea bass, Dicentrarchus labrax(accession no. AY395632), Atlantic salmon, Salmo salar (accession no. AF135117) and zebrafish, Danio rerio (accession no. AF042108). The forward primer was 5′-GGA CC(A/C) G(G/C/A)A ACC C(C/T)G A(C/T)G ACA T-3′ and the reverse primer was 5′-CCT G(G/T)G C(C/T)T TCA TGA TCC(T/G)CT CC-3′. All sequences were aligned with ClustalW and primers were designed with the assistance of GeneTool Lite software(www.biotool.com).

Complete sequences of two isoforms of the inducible hsp70 and one isoform of the constitutive hsc70 were obtained from northern killifish (NH). Degenerate primers were initially used to obtain a 1500 bp fragment in the central region of each gene. These primers were designed based on conserved regions of zebrafish hsc70 (accession no. Y11413), Ictalurus punctatus hsp70 (accession no. U22460), Oncorhynchus tschawytscha hsp70 (accession no. U35064), Xenopus laevis hsc70(accession no. X01102) and Gallus gallus hsp70 (accession no. J02579). The sequence of the forward primer (Hsp701F) was 5′ GGA CCA CAC C(C/A)A GCT TGG 3′ and the reverse primer (Hsp701R) was 5′ CGT TIG TGA TIG TGA TCT TGT TC 3′.

Polymerase chain reactions (PCRs) were carried out in a PTC-200 MJ Research thermocycler using Taq DNA polymerase (MBI Fermentas Inc.,Burlington, ON, Canada) and either cDNA (for all genes) or genomic DNA (for hsp70-1 and hsp70-2) isolated as described above. Each PCR consisted of 40 cycles of 30 s at 94°C, 30 s at the primerspecific annealing temperature, and 1 min for every 1,000 bp of expected product at 72°C.

PCR products were electrophoresed on 1.5% agarose gels containing ethidium bromide and bands of appropriate size were extracted from the gels using the QIAEX II gel extraction kit (Qiagen Inc., Mississauga, ON, Canada). Extracted PCR products were ligated into a T-vector (pGEM-T easy; Promega; Fisher Scientific, Nepean, ON, Canada), transformed into heat shock competent Escherichia coli (strain JM109; Promega; Fisher Scientific, Nepean,ON, Canada) and colonies were grown on ampicillin Luria-Bertani (LB) agar plates. Several colonies containing the ligated PCR product were selected and plasmids were isolated from liquid culture using GenElute Plasmid Miniprep kit(Sigma-Aldrich, Oakville, ON, Canada) and sequenced using an ABI 377 automated fluorescent sequencer at York University Molecular Biology core facility(Toronto, ON, Canada), or at the NAPS core facility at the University of British Columbia (Vancouver, BC, Canada). At least three clones of each fragment were sequenced bidirectionally. Consensus sequences for the hsp90 fragments were submitted to GenBank (hsp90α,accession no. DQ202281; hsp90β, accession no. DQ202282).

To determine the complete cDNA sequences for hsc70, hsp70-1 and hsp70-2 genes, isoform-specific nested PCR primers were designed based on the central fragment sequences obtained above, and used for 5′and 3′ rapid amplification of cDNA ends (Smart RACE cDNA amplification kit; BD Bioscience Clontech, Mississauga, ON, Canada). Primer sequences for 5′ RACE of hsp70 were as follows: external primer (for both isoforms) 5′ TTC ACC TCA AA(C/T) ATG CCG TCC 3′; internal primer for hsp70-1 5′ CAT TGC GCT CTCCTC TTT TGC 3′; internal primer for hsp70-2 5′ TTT CTC TCT CCC GTC TTG CC 3′. Primer sequences for 3′ RACE of hsp70 were as follows: external primer (for both isoforms) 5′ AGC CAT GAC CAA GGA CAA CAA 3′;internal primer for hsp70-1 5′ CCA GAG GAG TGC CAC AGA TAG AG 3′, internal primer for hsp70-2 5′ AGG TTT GAG CTG ACG GGA ATC 3′. Primer sequences for 5′ RACE of hsc70 were as follows: external primer 5′ TTT GGC CGG GTG CTG TCA TTG A 3′,internal primer 5′ AAA CCG CCG GCC AAT CAA CC 3′. Primer sequences for 3′ RACE of hsc70 were external primer 5′ CAC TGC TGG AGA TAC TCA TCT TGG TGG G 3′, internal primer 5′ GCG GTG TTC CAC AGA TTG AGG TGA CCT T 3′. Each PCR consisted of 3 min at 94°C followed by 35 cycles of 30 s at 94°C, 30 s at 60°C and 1 min at 72°C, followed by 7 min at 72°C. PCR products were then electrophoresed and cloned as described above.

At least three clones per fragment were sequenced in both directions at least twice, and a majority-rule consensus for the full-length cDNA transcript was developed for each isoform. Sequence assembly and analysis were performed using GeneTool Lite and DNAstar (Lasergene) software. Comparison with published sequences in GenBank was made with the BLAST algorithm(Altschul et al., 1997) and multiple alignments were produced using ClustalW(Thompson et al., 1994). Complete cDNA sequences have been deposited into GenBank (hsc70,accession no. DQ202278; hsp70-1, accession no. DQ202279; hsp70-2, accession no. DQ202280).

To determine whether there are any fixed differences in the sequences of hsc70 or hsp70 isoforms between northern and southern populations of killifish that might affect the function of these proteins, a series of isoform-specific PCR primers were developed to amplify the complete coding region of each isoform: hsp70-1 forward 5′ CTC AGA TCT TTT CCA CGT ACT CA 3′, hsp70-1 reverse 5′CTC CAG TAG TGA AAT GAT GCA GT 3′; hsp70-2 forward 5′ CTG AAA GGA AAG TGA GCC AAG ATG 3′, hsp70-2 reverse 5′ TAA ACA GTC CAG GAG ATG AGA GT 3′; hsc70 forward 5′ CCC GGA GAG GTC TGC TGT GT 3′, hsc70 reverse 5′ GGA GGT CTG AGG ATG GAA TGG T 3′. These primers were used to obtain the complete coding regions of these genes from at least three individuals from both NH and GA killifish populations. The sequences of a central fragment of the coding region of each gene (1409 bp)were also determined for an additional five individuals from each population using the following primers: for hsp70-1 forward 5′ CAT GAA CCC CAC CAA CAC AAT C 3′, reverse 5′ CGA CAG CAG ACA CGT TTA GGA 3′; for hsp70-2 forward 5′ CGC GTA CGG TCT GGA CAA AGG C 3′, reverse 5′ GCC CTT CAA GCT CTC GTC GTC CA 3′; for hsc70 the original degenerate primers (HSP701F and HSP701R) were used on cDNA from control fish.

Amino acid sequences were deduced from the nucleotide sequence of each isoform for both northern and southern fish using GeneTool Lite Software. Because there were few fixed differences between populations, only northern killifish sequences were used in the phylogenetic analysis. Protein sequences or deduced amino acid sequences were obtained from GenBank for all the available complete fish hsp/hsc70 genes, and Bos and Homo sequences were used as representative mammalian species. Sequences were aligned using ClustalW and phylogenetic analysis was performed using the neighbor-joining method with pairwise deletion of gaps using MEGA2 software (Kumar et al., 2001). The support for each node was assessed using 1000 bootstrap replicates, and isoforms were named according to their position on the phylogenetic tree.

To determine the threshold induction temperature of heat shock proteins in killifish, we acclimated northern (NH) and southern (GA) killifish to a common temperature of 20°C for 8 weeks as previously described. Groups of six fish per population were sampled directly from the acclimation tank (control)or transferred to one of several acute thermal challenge groups: 30, 31, 32,33, 34, 35 and 36°C [GA only: preliminary experiments with NH killifish acutely transferred from 20°C to 36°C resulted in 100% mortality within 1 h] or to 20°C (handling control) for 2 h. Fish were then transferred back into 20°C recovery tanks for 1 h. We chose a 1-h recovery period based on literature values for eurythermal fish and several pilot experiments indicating that these exposure and recovery times were sufficient to induce changes in gene expression. Following the recovery period, fish were sacrificed by rapid decapitation and the gills were dissected and immediately frozen in liquid nitrogen. All tissues were stored at -80°C until analysis. We elected to examine expression in the gill because preliminary experiments indicated that interindividual variation in hsp gene expression was lowest in this tissue, thus maximizing our ability to detect inter-population differences in gene expression.

Total RNA was extracted using TRIzol® Reagent, quantified spectrophotometrically, and cDNA was synthesized using 5 μg total RNA per sample (as previously described). Gene expression data were obtained using quantitative real-time PCR (qRT-PCR) on an ABI Prism 7000 sequence analysis system (Applied Biosystems Inc., Foster City, CA, USA). Genespecific primers were designed using Primer Express software (version 2.0.0; Applied Biosystems Inc., Foster City, CA, USA) and are reported in Table 1. qRT-PCR reactions were performed using 2 μl cDNA, 4 pmol of each primer and 2× SYBR Green Master Mix (Applied Biosystems Inc., Foster City, CA, USA) to a total volume of 22 μl under the following conditions: 1 cycle of 50°C for 2 min, 1 cycle of 94°C for 10 min, 40 cycles of 95°C for 15 s, 60°C for 1 min. At the end of each qRT-PCR reaction, PCR products were subjected to a melt curve analysis to confirm the presence of a single amplicon. In addition,representative samples were sequenced to verify that the appropriate gene fragments were amplified.

Samples of RNA that had not been reverse transcribed were also subjected to qRT-PCR to detect the possible presence of genomic DNA contamination. For the constitutively expressed genes as well as the control gene elongation factor-1α, genomic DNA contamination was below 1:1024 starting cDNA copies for hsc70, 1:4096 for hsp90β and 1:2048 for EF-1α. For the inducible genes (hsp70-1, hsp70-2 and hsp90α), we considered a sample to be induced when mRNA levels were at least 32-fold greater than background genomic contamination. One highly induced sample was used to develop a standard curve relating threshold cycle to cDNA amount for each primer set. Results were then normalized using elongation factor-1α (EF-1α; accession no. AY430091) as mRNA levels of this gene do not change with heat shock in killifish gills(data not shown).

Thermal tolerance data sets were analyzed by analysis of covariance(ANCOVA) with length or mass as covariates. Corrected CTM values differed by no more than 0.1°C from actual values; therefore, data sets from both the thermal tolerance and heat shock experiments were analyzed by multiple analysis of variance (ANOVA) with population, acclimation group, and/or heat shock temperature as factors without statistical adjustment for body size. Simple linear regression (SLR) and polynomial regressions were used to explore and model the statistical relationship between CTMax or CTMin of killifish and acclimation temperature. All data met the assumptions of normality, and data were log transformed where necessary to meet assumptions of homogeneity of variance. When interaction terms were not significant, post-hoccomparisons were performed among the groups with the Student-Newman-Keuls multiple range test (SNK MRT). If the interaction terms were significant, the data were separated and analyzed independently using one-way ANOVA. All statistical decisions were based on P=0.05.

Using laboratory acclimation studies, we identified acclimation ranges and thermal tolerance scopes for killifish from northern (NH) and southern (GA)populations. Chronic thermal maximum and minimum experiments revealed pronounced differences in survival between northern and southern killifish populations (Fig. 1). Survival data were fit to a third order regression for each population, which predicted the chronic thermal maxima to be 36.4°C for northern fish and 38.2°C for southern fish. Chronic thermal minima experiments, however, revealed no difference between northern and southern populations with both populations surviving until the water ultimately froze at -1.1°C.

Acclimation temperature had a substantial effect on both CTMaxima and minima in northern (NH) and southern (GA) killifish populations such that CTM values increased with increasing acclimation temperature in both killifish populations (Fig. 2). Simple linear regressions of CTMax and CTMin on acclimation temperature for northern and southern killifish populations were highly significant (SLR, P<0.001, r²=0.958 (CTMax) and 0.858 (CTMin),for northern fish and SLR, P<0.0001, r²=0.935(CTMax) and 0.910 (CTMin), for southern fish)(Table 2). Critical thermal maxima increased by 0.41°C (northern fish) and 0.36°C (southern fish)for every 1.0°C increase in acclimation temperature, whereas CTMin increased by 0.28°C (northern fish) and 0.35°C (southern fish) for every 1.0°C increase in acclimation temperature.

Although simple linear regressions, as reported above, are traditionally used to model the relationship between acclimation temperature and thermal tolerance, we also used multiple regression techniques to determine the best-fit relationship for our data. Second order regression models(Fig. 2) increased the coefficients of determination (r²) for each regression with the greatest improvements made in the CTMin models(Table 2). By using these second order regression models, we were able to explain 95% or more of the variation in critical thermal limit by variation in acclimation temperature in all four models.

There were significant differences within northern (NH) and southern (GA)killifish populations in mean length and mass among acclimation groups(Table 3). When CTM values were adjusted by analysis of covariance (ANCOVA) using either total length (cm) or wet mass (g) as covariates, the corrected CTM values differed by no more than 0.1°C from the actual values (Table 3). Therefore, no statistical adjustments of thermal tolerance values were necessary for either length or mass, and we used only the actual measured critical thermal tolerance values for all subsequent data interpretation and comparisons.

CTMax and CTMin were analyzed using two-way ANOVA with population and acclimation group as factors. Two-way ANOVA revealed a significant effect of population and acclimation group, as well as a significant interaction term for both the CTMax and CTMin data sets (P<0.001 for all comparisons). One-way ANOVA followed by post-hoc tests revealed that within a killifish population, CTMax were significantly different at each acclimation temperature with two exceptions: critical thermal maxima for southern fish acclimated to 7.2°C and 12.4°C, and critical thermal maxima for southern fish acclimated to 32.1°C and 34.0°C were not statistically significantly different from one another (SNK MRT, P=0.242 and P=0.709, respectively)(Fig. 2A). Critical thermal minima responses within a population of killifish were also significantly different at each acclimation temperature with the exception of the three lowest acclimation temperature groups. There were no significant differences in the CTMin of northern fish acclimated to 2.3°C, 7.2°C and 12.4°C (P=1.000 for all comparisons). Similarly, no differences in CTMin within the southern population acclimated to 2.3°C, 7.2°C and 12.4°C were found (P-values between 0.544 and 1.000 for all comparisons; Fig. 2B).

Between-population comparisons revealed that the CTMax of southern fish within a temperature acclimation group was always significantly higher than the CTMax of northern fish (P<0.001 for all comparisons)(Fig. 2A). The CTMin of southern fish within an acclimation group were also significantly higher than that of northern fish, except at acclimation temperatures of 12.4°C or lower, where the CTMin converged at the freezing point of brackish water(P-values between 0.544 and 1.000 for all comparisons)(Fig. 2B).

In order to determine if thermal tolerance differences between northern and southern killifish were consistent across multiple populations within a subspecies, we quantified CTMax and CTMin in samples of killifish from six different populations (three of the northern subspecies and three of the southern subspecies) acclimated to a common temperature of 22°C(Fig. 3). Within a subspecies,killifish populations did not differ in CTMax (P-values ranged from 0.060 to 0.928) or in CTMin (P-values ranged from 0.092 to 0.921). Between subspecies, however, southern killifish always had significantly higher CTMax (P<0.001 for all comparisons) and CTMin(P<0.001 for all comparisons) than the northern forms(Fig. 3A,B).

Northern (NH) and southern (GA) killifish were sampled from the same geographic location in two separate years and acclimated to 22°C(2002-thermal tolerance experiment, Fig. 2; 2004-intraspecific experiment, Fig. 3). There were no significant differences in CTM values within a population sampled in either year (compare Table 3 with Fig. 3).

The degenerate primers used in this study allowed us to obtain three distinct heat shock protein 70-related transcripts from killifish gills. Phylogenetic analysis of the complete amino acid sequences of these transcripts (Fig. 4) revealed that one was highly similar to other fish hsc70 sequences, whereas the remaining two transcripts were similar to other fish hsp70sequences. The putative hsc70 from killifish was cloned from both control and heat shock samples, whereas the putative hsp70transcripts were found only in cDNA isolated from fish exposed to heat shock,confirming their identification as constitutive and inducible transcripts,respectively. The two hsp70 transcripts did not group together phylogenetically, but instead each grouped with a distinct sequence from Xiphophorus maculatus. We tentatively named the hsp70transcripts hsp70-1 and hsp70-2, following the convention established for X. maculatus.

In order to determine whether there were any fixed differences in hsp sequences between northern and southern killifish populations that might affect their function, we obtained the complete coding sequences of hsp70-1, hsp70-2 and hsc70 from a sample of individuals from each population (∼10 individuals from each of the NH and GA populations). There were no fixed differences between populations in hsp70-2, and only a single fixed difference between populations in hsp70-1, which did not result in a change in the amino acid sequence. There were two fixed differences between populations in hsc70, one of which was silent,and one of which resulted in a change from serine in the southern population to threonine in the northern population at amino acid position 98.

Despite the high conservation of these sequences at the amino acid level,there was substantial silent polymorphism in these genes, most of which was found in the southern population, consistent with the pattern observed for other genes (e.g. Bernardi et al.,1993). There were 14 polymorphic sites in hsp70-1, all of which were found in the southern population, whereas the sample from the northern populations exhibited no variation among individuals. Similarly, hsp70-2 had five polymorphic sites, only one of which was found in the northern population. by contrast, there were three silent polymorphic sites in hsc70 (in addition to the two fixed differences), two of which were present in both populations, and one of which was found only in the northern population.

Prior to heat shock treatment, there were no differences in control and handling control mRNA levels in the gill between northern (NH) and southern(GA) killifish for any of the three hsp70 genes measured. The mRNA levels for hsc70 (constitutive isoform) in northern fish did not change with heat shock (Fig. 5A). Southern fish, however, had elevated hsc70 mRNA levels in all heat shock groups, and this increase was significant in the 32°C group. As a result, southern fish had significantly higher hsc70 mRNA at all heat shock temperatures when compared to northern fish (P-values ≤0.019 for all comparisons).

The patterns of mRNA expression differed substantially between hsp70-1 and hsp70-2 (inducible isoforms). Levels of hsp70-1 increased gradually with increasing heat shock temperatures in both populations, and there was no difference between populations in the magnitude of this progressive induction(Fig. 5B). The induction profile for hsp70-2 was more typical of an inducible gene with a clear onset temperature of expression (Fig. 5C). Both northern and southern killifish demonstrated a significant elevation above control values in mRNA levels(T_on) for hsp70-2 at 33°C. However, northern fish had significantly higher levels of hsp70-2 at 33°C, 34°C and 35°C than southern fish (P<0.05 for all comparisons).

Both northern and southern killifish populations showed a gradual increase in hsp90α expression (inducible isoform) with increasing heat shock temperature, and the overall magnitude of induction did not differ between populations (Fig. 6A). T_on for hsp90α, however, did differ between populations, with significant induction occurring at 30°C in southern fish(P=0.021), and at 32°C in northern fish (P=0.012). Levels of hsp90β (constitutive isoform) did not change with heat shock in southern killifish. Northern killifish had slightly elevated hsp90β levels at 31-33°C. Two-way ANOVA for hsp90β mRNA levels, with population and heat shock temperature as factors, revealed a significant effect of heat shock temperature(P=0.001) and population (P=0.037) and no significant interaction (P=0.675) with southern individuals having overall higher hsp90β mRNA levels than northern fish. Post-hoc tests,however, revealed a significant difference in hsp90β mRNA levels between populations only in control fish.

From the data presented here, it is clear that killifish are impressive eurytherms and can be acclimated to an exceedingly wide range of environmental temperatures. Both northern (NH) and southern (GA) killifish were able to survive temperatures approaching the freezing point of brackish water for several days and had chronic upper lethal limits of 36.4 and 38.2°C,respectively (Fig. 1). As a result, fish from a southern population (GA) had a chronic scope (chronic maximum - minimum) of 39.3°C (Fig. 1), which is larger than any previously published value for fishes(reviewed by Beitinger et al.,2000). Furthermore, both northern and southern killifish had CTMax that approached or exceeded 42°C and CTMin of approximately -1.1°C(Table 3), which are among the highest and lowest values, respectively, measured in fishes (reviewed by Beitinger et al., 2000). Of the available thermal tolerance scopes (CTMax - CTMin), only three of the most eurythermal fish have scopes that meet or exceed those of the killifish: the Amargosa pupfish, Cyprinodon navadensis(Feldmeth, 1981), the sheepshead minnow, C. variegates(Bennett and Beitinger, 1997),and the Atlantic stingray, Dasyatis sabina(Fangue and Bennett, 2003). The combined thermal acclimation and tolerance attributes of both northern and southern killifish populations encompass the entire range of daily and seasonal temperatures they naturally experience and may allow killifish to move freely between dynamic thermal microhabitats.

Consistent with previous suggestions of local thermal adaptation, there were differences in thermal tolerance between killifish populations from different geographic locations. On average, southern (GA) killifish had critical thermal maxima that were 1.5°C higher than northern (NH) fish across all acclimation temperatures (Fig. 2A). This 1.5°C difference between populations was also maintained for critical thermal minima except at the three lowest acclimation temperatures (Fig. 2B). These differences in thermal tolerance between killifish populations are small relative to their large thermal acclimation ability, but comparisons within species for many physiological traits often show small overall intraspecific variation compared to the variation seem among species or higher taxa(Feder et al., 2000).

The difference in CTM values between northern and southern subspecies was consistent across multiple sampling sites: southern (GA, WI and FB) killifish populations had critical thermal tolerance values that were significantly higher than northern (NH, ME and NS) populations(Fig. 3). These differences in thermal tolerance were also maintained across samples obtained from the same locations in NH and GA in different years (2002 and 2004). Taken together,these data suggest that the critical thermal limits for each killifish subspecies are an intrinsic property of that subspecies and are constant from year to year. However, additional work on laboratory-reared killifish will be necessary to confirm this suggestion and to rule out the possibility of maternal or developmental effects. Killifish thermal tolerance limits vary in a direction consistent with that predicted for fish that have undergone localized adaptation to habitat temperatures. These functional differences between killifish subspecies are in agreement with a variety of work by others suggesting that the physiological specializations and genetic variation between subspecies are likely to be adaptive responses to temperature or some other factor correlated with latitude (reviewed by Powers and Schulte, 1998; Schulte, 2001).

The heat shock response is thought to be important for adaptation of organisms to their thermal environment(Feder and Hofmann, 1999; Hoffmann et al., 2003; Somero, 2005). A number of characteristics of the heat shock response could, in principle, respond to thermal selection, including: (1) the functional efficiency of the heat shock proteins themselves, (2) the onset temperature (T_on) at which hsp expression is induced, and (3) the magnitude of hsp expression under either basal or induced conditions. Although there is some evidence for each of these mechanisms in natural populations or following laboratory selection (Tomanek and Somero, 1999; Michalak et al., 2001; Feder et al.,2002; Sorensen et al.,2005) little is known about whether similar patterns are observed across multiple isoforms within a species, and few studies have assessed the relative roles and generality of these mechanisms.

We found no differences in the amino acid sequences of hsp70-1 or hsp70-2 within or between populations of killifish, and there was only a single highly conservative substitution between populations in hsc70 (ser to thr at amino acid 98). These data strongly suggest that changes in the functional efficiency of the 70 kDa heat shock proteins have not been involved in the evolution of differences in thermal tolerance between northern and southern killifish populations.

The onset temperature of hsp induction (T_on)and magnitude of hsp expression exhibited a variety of patterns among isoforms when compared between northern and southern killifish populations. Among the three inducible genes we examined (hsp90α, hsp70-1 and hsp70-2), hsp90α had a lower T_on in southern populations, but populations did not differ in the magnitude of induction, whereas hsp70-2 did not differ in T_on between populations, but was induced to a greater magnitude by heat shock in northern fish than in southern fish. By contrast,neither T_on nor the magnitude of expression differed between populations for hsp70-1. The observation that each isoform exhibited a different pattern of expression between populations suggests that differences in expression are not due to a global factor that affects all hsps, such as differences between populations in the stability of the total protein pool or overall rates of protein or mRNA turnover. Instead, this complex pattern of expression suggests that differences in mRNA levels between populations result from gene-specific mechanisms such as differences in transcription as a result of promoter sequence variation, or differences in mRNA stability as a result of sequence variation in the 5′ or 3′untranslated region in a particular hsp gene(McGarry and Lindquist, 1986; Petersen and Lindquist,1988).

The lower T_on for hsp90α we observed in southern killifish (Fig. 6A) is in marked contrast to the results of most other studies, which have generally found that organisms from warmer environments induce Hsps at a higher temperature than closely related organisms from colder environments(Huey and Bennett, 1990; Dietz and Somero, 1992; Fader et al., 1994; Gehring and Werner, 1995; Hofmann and Somero, 1995; Hofmann and Somero, 1996; Tomanek and Somero, 1999; Tomanek and Somero, 2000). This discrepancy might be explained by the fact that our study examined mRNA levels, while most previous studies have focused on protein levels(Feder and Hofmann, 1999; Tomanek and Somero, 1999; Tomanek and Somero, 2000; Buckley et al., 2001). However,if this were the case, we would have to postulate a decoupling of transcription and translation for hsp90α. Alternatively, the lower T_on for hsp90α could reflect an anticipatory response in southern killifish. Southern fish frequently experience peak water temperatures greater than 30°C, whereas water temperatures exceeding 30°C are rare in northern habitats. There is some support from experiments in Drosophila for maximum rather than mean environmental temperature as the important environmental thermal feature structuring adaptive thermal responses(Davidson, 1988; Anderson et al., 2003). An anticipatory upregulation of hsp90α could allow southern fish to protect critical components of their protein pool in the face of high environmental temperatures.

The greater magnitude of hsp70-2 upregulation in northern killifish (Fig. 5C) is consistent with the hypothesis that these fish are more sensitive to thermal stress. The magnitude of the heat shock protein response is typically proportional to the severity of the heat shock and associated protein damage(e.g. see DiDomenico et al.,1982). Similar to the results presented here, lines of Drosophila selected for high-temperature resistance have decreased expression of Hsp70 in response to heat shock compared to control lines,suggesting that a sub-lethal heat exposure is less stressful for heat adapted populations, thus leading to less cell damage and an overall smaller magnitude of stress protein induction (Sorensen et al., 1999; Sorensen et al.,2001). However, the hypothesis of differential thermal sensitivity is not supported by the results with hsp70-1 mRNA levels, which exhibited no differences between populations in either T_onor the magnitude of induction (Fig. 5B). Possible explanations for the differences in response between the two isoforms include differential sensitivity of these two isoforms to the denatured protein pool or differences in specificity between the hsp70-2 response and the hsp70-1 response.

Southern killifish had generally higher basal levels of the two constitutive mRNAs (hsc70 and hsp90β) than northern killifish. If these mRNA levels are indicative of differences in the standing protein pool, these differences could be protective and are thus consistent with the greater thermal tolerance of southern fish. Work in poeciliid fishes(diIorio et al., 1996)suggests that higher constitutive levels of Hsc70 protein may be as or more important for thermal tolerance than changes in the inducible genes. However,there is some evidence in Drosophila and Arabidopsis that enhanced levels of Hsps under basal conditions can be deleterious(Krebs and Feder, 1997; Krebs and Feder, 1998; Zatsepina et al., 2001; Sung and Guy, 2003), and thus increased levels of the constitutive Hsps in killifish in unstressed fish could have a negative effect.

Basal levels of heat shock proteins, particularly Hsp90, can also affect the induction of heat shock protein genes (for a review, see Voellmy, 2004). Heat shock protein induction occurs via the binding of a transcription factor,the heat shock factor (HSF). Under unstressed conditions, HSF is present as a protein complex that includes Hsp90, Hsp70 and other chaperone molecules. This complex does not bind DNA effectively, and thus inducible heat shock genes are transcribed almost undetectably under unstressed conditions. During thermal stress, the chaperone proteins dissociate from the HSF protein complex and bind to unfolded proteins within the cell, releasing HSF from inhibition(Wu, 1995). Based on this mechanism, we would predict that southern killifish populations, which have a higher level of hsp90β mRNA under unstressed conditions, would require a higher level of thermal stress to remove the repression of HSF by Hsp90, and thus would have a higher T_on and lower magnitude of expression at a given temperature for all inducible heat shock proteins. This was not the case. In fact, the only difference we observed in T_on between populations was a lower T_on for hsp90α in southern fish, in direct contrast to the predictions of this model. However, northern fish had a greater magnitude of hsp70-2 induction, which is consistent with model predictions.

Although hsc70 and hsp90β are considered to be constitutively expressed, we observed statistically significant increases in the expression of both of these genes in response to heat shock. It has been shown that some fish have the ability to upregulate hsc70 levels with increasing acclimation temperatures as well as with heat shock(Deane and Woo, 2005), whereas other research has shown a pattern more typical of a constitutive isoform with no change in hsc70 levels with heat shock(Yamashita et al., 2004; Ojima et al., 2005). The pattern of hsc70 and hsp90β induction differed between killifish populations. Southern killifish exhibited a substantial increase hsc70 levels in response to heat shock whereas no change was observed in northern fish. The upregulation of hsc70 expression by southern killifish in response to heat shock may suggest an important role for Hsc70 in handling protein damage associated with daily fluctuations in environmental temperatures, and is consistent with the greater thermal tolerance exhibited by killifish from southern populations. In contrast, there was a small but statistically significant elevation of hsp90β in northern fish but not southern fish following heat shock, which is not obviously consistent with a hypothesis of thermal adaptation.

The modulation of hsp mRNA expression patterns involving multiple isoforms from several Hsp families has been suggested as one mechanism ectotherms use to maintain flexibility in thermal phenotype in response to changing thermal environments (Hightower,1991; Hochachka and Somero,2002). The combinations of Hsps expressed, however, vary widely between organisms and the reason for this variation in protein expression is unknown. Often, this variation reflects both the evolutionary histories of the species and the recent thermal acclimation conditions encountered by that organism (White et al., 1994; Hofmann and Somero, 1995; Roberts et al., 1997; Tomanek and Somero, 2002; Tomanek, 2005) suggesting that substantial adaptive variation exists in the heat shock response. To our knowledge, only a single study has addressed the mRNA expression profiles of multiple hsp genes from several gene families in fish(Ojima et al., 2005). This work was performed in an immortalized rainbow trout gonadal fibroblast cell line and only a single acclimation temperature and heat shock temperature treatment was evaluated. Even with this simple experimental design, however,the authors demonstrate gene-specific variation in hsp mRNA levels. Results from microarray studies in fish exposed to constant or cycling environmental temperatures (Podrabsky and Somero, 2004) or during cold acclimation(Gracey et al., 2004) show complex gene expression signatures that involve many gene classes known to be associated with thermal tolerance. However, these studies, and the study reported here, have examined these processes at the mRNA level, and mRNA levels are not necessarily predictive of the behaviour of the protein pool. Recently, ³⁵S-labelling of newly translated proteins followed by two-dimensional gel electrophoresis in turban snails exposed to heat shock,showed the induction of over 30 proteins from several Hsp families with varying patterns among isoforms (Tomanek,2005). Whereas it is not yet known whether these proteins are coded by different genes or represent post-transcriptional modifications, it is clear that these protein variants are important and could contribute to the phenotypic plasticity seen in eurythermal organisms. Although correlative studies such as those of Tomanek (Tomanek,2005) and the current study cannot directly establish a causal link between the patterns of hsp isoform expression and whole organism thermal tolerance, these studies provide critical evidence of the underappreciated diversity of the patterns of hsp expression in natural populations, their relationship to differences in whole-organism thermal tolerance, and their possible role in the establishment of biogeographical patterns.

This work was financially supported through a Natural Sciences and Engineering Research Council of Canada (NSERC) discovery grant to P.M.S. N.A.F. is supported by Izaak Walton Killam and UBC University Graduate predoctoral fellowships. We are grateful to Anne Dalziel, and Drs Anne E. Todgham, Jeff Richards and Charles Darveau for constructive editorial comments on this manuscript.