A DROSOPHILA MELANOGASTER Strain From Sub-Equatorial Africa Has Exceptional Thermotolerance But Decreased Hsp70 Expression

Drosophila melanogaster has been a key model system for the elucidation of mechanisms of thermotolerance and the evolution of thermotolerance. The heat-shock response was discovered in this model, a finding that eventually led to the discovery of heat-shock proteins, their encoding genes and molecular chaperones (Ritossa, 1996). In addition, the prominence of Drosophila melanogaster as a model system has encouraged numerous studies of its thermal biology and a comprehensive characterization of its thermal limits (David et al., 1983). Against this background, any exception is remarkably distinctive. In the early 1970s, L. Tsacas (C.N.R.S., Gif-Sur-Yvette, France) obtained D. melanogaster from Fort-Lamy (now N’Djamena), Chad, in the central semiarid tropical (Sahel) zone of Africa (latitude 12°N) (J. R. David, personal communication). The climate is seasonally warm, with daytime temperatures averaging above 38°C between March and June. These flies were fertile in continuous culture at 30°C and were eventually shared with investigators in St Petersburg and Moscow, Russia, where they were fertile in continuous culture at 31–32°C. Insofar as D. melanogaster is generally acknowledged to be incapable of continuous culture above 30°C (Parsons, 1973) and numerous attempts to breach this limit through laboratory selection have failed, this finding is remarkable and presents a rare opportunity to understand how a species may evolve enhanced thermotolerance.

In D. melanogaster, typically cultured at 18–25°C, survival at high temperature is normally inversely related to the duration and severity of heat shock, with 1h heat shocks of 38–39°C sufficient to cause 50% mortality (David et al., 1983; Feder and Krebs, 1997). Culture at higher temperatures, either constantly at slightly elevated temperatures (e.g. 28°C) or cycling between typical and still higher temperatures, can increase thermotolerance above these basal levels (Bettencourt et al., 1999; Lansing et al., 2000). This enhanced thermotolerance can ensue both within the lifetime of an individual fly (i.e. phenotypic plasticity, thermal acclimation) or via genetic change among generations (i.e. evolution). Moreover, pretreatment with a mild heat shock (e.g. 1h at 35–37°C) can rapidly increase tolerance of a more severe heat shock; this is termed ‘inducible thermotolerance’. Expression of the heat-shock proteins is one of the primary mechanisms underlying inducible thermotolerance in D. melanogaster. In this species, one such inducible protein, Hsp70, is encoded by five nearly identical genes at two chromosomal loci (87A7, two genes; 87C1, three genes) (for a review, see Feder and Hofmann, 1999). At least eight other members of the hsp70 gene superfamily, both inducible and constitutively expressed, are present in the genome (Burmester et al., 2000; Easton et al., 2000; Rubin et al., 1993). Hsp70 is the most abundant heat-shock protein induced by heat, and manipulation of its abundance (via antisense and engineering of hsp70 copy number) is sufficient for large changes in inducible thermotolerance (Feder et al., 1996; Solomon et al., 1991; Welte et al., 1993). Activation of the heat-shock transcription factor HSF via trimerization, phosphorylation and nuclear localization is a key step in heat-shock protein expression. HSF is negatively regulated through interaction with Ku autoantigen, HSBP1 and Hsp70 family members (Cotto and Morimoto, 1999; Lerman and Feder, 2001; Tuteja and Tuteja, 2000; Zatsepina et al., 2000). When activated, HSF can bind heat-shock response elements (HSEs) in the hsp70 promoter to induce transcription; the number and arrangement of the HSEs affect hsp70 gene expression (Lis and Wu, 1994).

These findings for typical D. melanogaster led us to investigate basal and inducible thermotolerance in the ‘T’ (thermotolerant) strains descended from those D. melanogaster collected in Chad and their patterns of Hsp70 protein accumulation and its transcriptional regulation. Not only can the T strain survive prolonged culture at temperatures above 30°C, it tolerates several high-temperature treatments (36°C for 26h; 38°C for 4h) better and low temperatures more poorly than does Canton-S, a standard wild-type line (Tikhomirova and Belyatskaya, 1980). When subjected to ionizing radiation and heat shock in combination, the T strain is less prone to mutagenesis and repairs radiation injuries more readily than wild-type and Hsp-deficient strains (Tikhomirova, 1980; Tikhomirova et al., 1993). In part, this resistance depends on culture temperature because both adults and oocytes of T flies reared at 25°C (hereafter T25) show intermediate levels of resistance compared with those reared at 32°C (hereafter T32) and Canton-S flies (Tikhomirova and Belyatskaya, 1980; Tikhomirova and Belyatskaya, 1993). Here, we differentiate between basal thermotolerance, which is presumably due to mechanisms other than heat-shock proteins, and inducible thermotolerance, which is due in part to heat-shock proteins, in the T32, T25 and a wild-type strain, and relate these differences to levels and varieties of heat-shock proteins and the regulation of their expression. Although we initially hypothesized that study of the T strain would disclose molecular and evolutionary mechanisms of increased thermotolerance, we report here several unique features of this strain that are apparently associated with decreased inducible thermotolerance.

The T strain flies (see Introduction), maintained at 31–32°C (T32) and 25°C (T25) for many years (see Introduction), were compared with Oregon R wild-type flies. In some cases T32 flies were reared from egg to adulthood at 25°C before analysis. For determinations of thermotolerance and for Hsp70-specific enzyme-linked immunosorbent assay (ELISA), T32 and T25 flies were transported to Chicago and maintained at their normal temperatures for at least one generation before use. These flies were compared with an Oregon R stock (A25) maintained at 25°C in discrete generations for more than 20 years (Bettencourt et al., 1999; Cavicchi et al., 1995; Cavicchi et al., 1989). All other comparisons were with an Oregon R stock maintained at 25°C in Moscow for many years. All flies were reared on a yeast, cornmeal, molasses and agar medium.

Eclosing individuals were sequestered daily and, when 4 days old, were transferred by aspiration to a fresh glass vial containing 8ml of medium. A circle of filter paper above the medium allowed flies access to it but prevented them from adhering to it. Each vial usually contained 16–20 animals, but occasionally contained as few as 10. On the next day, vials were capped with a moistened stopper above a cotton plug, placed in a rack with vials evenly spaced and immersed in a water bath at pretreatment and heat-shock temperatures. Some flies underwent direct exposure to heat shock (37–41°C) for 30min. Others underwent exposure to a pretreatment temperature (36–38°C) for 30min followed by 25°C for 1h, and then underwent heat shock. Thermotolerance was measured as the proportion of flies in a vial that were able to walk 24h after heat shock.

Flies were heat-shocked as above but for different durations depending on the experimental design, immediately frozen in liquid nitrogen and then stored at −80°C until analysis. Pairs of identically treated flies were lysed in 200–400μl of ice-cold 1× Complete Protease Inhibitor (Boehringer-Mannheim Corp) in phosphate-buffered saline by grinding briefly with an ice-cold disposable pestle. Lysates were centrifuged at 14000g for 30min at 4°C, and the protein content of the supernatant was determined (BCA Assay, Pierce Chemical Co., Rockford, IL, USA). Supernatants prepared the same day were diluted to 20μgml⁻¹ protein in ice-cold coating buffer and used to coat micro-well plates (Falcon no. 3915 ProBind) for determination of Hsp70 content by ELISA (Feder et al., 1997; Feder et al., 1996). Plates were left overnight at 4°C to allow proteins to adsorb. After extensive rinsing, bound Hsp70 was detected using a 1:5000 dilution of the Drosophila Hsp70-specific antibody 7FB (Velazquez et al., 1980; Velazquez et al., 1983) coupled to alkaline phosphatase via a secondary antibody (1:1000 rabbit anti-rat IgG; Cappel Organon Teknika) and a tertiary antibody (1:1000 alkaline-phosphatase-conjugated goat anti-rabbit IgG; Sigma). Plates were incubated at 37°C with the phosphatase substrate p-nitrophenyl phosphate (1mgml⁻¹) prepared according to the manufacturer’s instructions (Sigma), and the colored reaction product was measured at 405nm in a micro-plate reader. For at least one replicate of each sample or standard, the primary antibody was omitted to allow correction for non-specific signal. The ELISA signal is proportional to Hsp70 concentration in the lysates and is expressed as a percentage of a standard signal, that for a lysate of D. melanogaster S2 cells in tissue culture that had been exposed to 36.5°C for 1h and to 25°C for 1h before lysis.

Ten salivary glands from third-instar larvae were labeled in 20μl of Schneider’s insect medium without methionine (Sigma) after the addition of 1μl (1.85MBq) of l-[³⁵S] methionine (Amersham) for 1h at 25°C after various treatments. Two-dimensional gel electrophoresis and other procedures applied were as described by O’Farrell et al. (O’Farrell et al., 1977) and Ulmasov et al. (Ulmasov et al., 1992). The position of major Hsps and actin was determined both by autoradiography (see above) and by subsequent staining of gels with silver (Creighton, 1990). Salivary glands were also labeled with ¹⁴C-labeled amino acids, which yielded essentially the same results (data not shown).

For immunoblotting, after sodium dodecyl sulfate/polyacrylamide electrophoresis (SDS–PAGE) of larval lysate prepared as above, the proteins were transferred to a nitrocellulose membrane (Hybond ECL; Amersham) according the manufacturer’s protocol. Monoclonal antibodies specific to the entire D. melanogaster Hsp70 family (7.10.3) and to Hsp70 alone (7FB, see above) were obtained from Dr Susan Lindquist (The University of Chicago). Immune complexes were detected via chemoluminescence (ECL kit, Amersham) and 3,3′-diaminobenzidine (DAB) (Sigma) with corresponding peroxidase-conjugated anti-rat secondary antibodies.

RNA was prepared by the standard method using 4moll⁻¹ guanidine isothiocyanate (Chomczynsky and Sacchi, 1987). Blotting and northern hybridization with a ClaI-BamHI fragment containing the Drosophila melanogaster hsp70 gene cloned into the BamHI site of pUC13 (McGarry and Lindquist, 1985) was performed (Sambrook and Fritsch, 1989) with slight modifications. Hybridization was overnight at 42°C in 50% formamide; this was followed by two 20min washes in 2×SSC, 0.2% SDS at 42°C, two 20min washes in 1×SSC, 0.2% SDS at 42°C and one 20min wash in 0.2×SSC, 0.2% SDS at 68°C.

Genomic DNA from the different D. melanogaster strains (adult flies) was prepared as described previously (Zelentsova et al., 1986). Genomic DNA (20μg) was used for a typical restriction enzyme digestion. Digests were prepared for hybridization by electrophoresis on 1% agarose gels, denaturation and capillary-blotting onto nylon membranes according to the manufacturer’s protocol. Fixation was by ultraviolet cross-linking with a UV Stratalinker 2400 (Stratagene). Standard high-stringency hybridization and wash conditions were used (Zelentsova et al., 1986). The probe was the same as that used in northern blots.

Flies were frozen and pulverized in liquid nitrogen, and the powder was suspended (1:5) in a buffer containing 20mmoll⁻¹ Hepes, pH7.9, 25% v/v glycerol, 0.42moll⁻¹ NaCl, 1.5mmoll⁻¹ MgCl₂, 0.2mmoll⁻¹ EDTA, 0.5mmoll⁻¹ phenymethylsulfonyl fluoride (PMSF) and 0.5mmoll⁻¹ dithiothreitol and centrifuged at 100000g for 20min. The supernatants were frozen in liquid nitrogen and stored at −70°C. The protein concentration of the extracts was estimated with a modified Lowry method (Ulmasov et al., 1992).

Consensus HSE probe (Wu et al., 1988) was prepared by annealing partially complementary oligonucleotides (ATCCGAGCGCGCCTCGAATGTTCTAGAA and CTCGCGCGGAGCTTACAAGATCTTTTCCA) in 10mmoll⁻¹ potassium phosphate buffer, pH8.2, in the presence of 0.1mmoll⁻¹ NaCl. Single-stranded termini were filled with Klenow polymerase and [³²P]ATP (Sambrook and Fritsch, 1989). For the gel mobility-shift assay, extracts containing 50μg of protein were mixed with 0.5ng of [³²P]HSE in the binding buffer (as described by Mosser et al., 1993). The binding-reaction mixture was incubated at room temperature (20°C) for 20min. Free probe was separated from HSE–HSF complexes by electrophoresis through 5% polyacrylamide gels (Mosser et al., 1993). The gels were dried and exposed to X-ray film (Kodak X-Omat) at −70°C.

To identify bands corresponding to HSF or Ku autoantigen, cell extracts of control or heat-shocked cells were preincubated for 20min either with anti-D. melanogaster Ku antibodies (gift of D. Rio) or with anti-D. melanogaster HSF serum (gift of C. Wu, NIH) before incubation with the labeled HSE probe; the samples were then subjected to electrophoresis.

Genomic DNA from 75 adults per strain was obtained by standard phenol/chloroform extraction and was used as templates for amplification products to be cloned and sequenced. Single-fly DNA for polymorphism screens was prepared from individual flies by homogenizing adults in 50μl of buffer with 0.2μgμl⁻¹ Proteinase K (Gloor et al., 1993). Samples were incubated at 37°C for 30min, heated to 95°C to inactivate Proteinase K and stored at –20°C.

Polymerase chain reaction (PCR) of single-fly preparations was performed by adding 2μl of template DNA to buffer (10mmoll⁻¹ Tris-HCl, pH9.0, 50mmoll⁻¹ KCl, 0.1% Triton X-100) with 1.5–3.0mmoll⁻¹ MgCl₂, 0.2mmoll⁻¹ each dNTP, 5pmol of each primer and 1.25units of Taq DNA polymerase (Promega) per 25μl reaction mixture. For PCR amplification of fragments to be cloned and sequenced, 1μl of template DNA from mass preparations was added to buffer (50mmoll⁻¹ KCl, 50mmoll⁻¹ Tris-HCl, pH8.3) with 1.5mmoll⁻¹ MgCl₂, 0.2mmoll⁻¹ each dNTP, 5pmol of each primer and 2.5units of MasterAmp TAQurate DNA polymerase mix (Epicentre Technologies) per 100μl reaction. Reaction conditions for all PCRs were 30 cycles of 1min at 92°C, 1min at 54°C and 1.5min at 72°C. To amplify a portion of the 87A7 locus, primers were: upper, 5′CATCCCAAAAATCTGTAAAGC3′; lower, 5′ACTGTGTTTCTGGGGTTCAT3′. These flank both an H.M.S. Beagle element insertion site and an approximately 140basepair (bp) insertion characteristic of 56H8-type alleles (Bettencourt, 2001), if present. To amplify the hsp70Ba promoter, primers were: upper, 5′GCAAGCAATCATCATCCAAT3′; lower, 5′ACTGTGTTTCTGGGGTTCAT3′. These flank a Jockey element insertion site. Fig.8A displays the sizes of the resultant amplification products, which were resolved via agarose gel electrophoresis. To screen individual flies for the Jockey element, primers were the foregoing hsp70Ba primers plus a Jockey-specific internal primer (lower, 5′AAGAAGACTCAAGCGACACC3′).

For cloning and sequencing, PCR products were amplified from bulk DNA template, gel-purified and/or cleaned with Qiagen spin columns, suspended in water and cloned into the pGEM-T Easy vector (Promega) according to the manufacturer’s recommendations. Individual plasmid clones were prepared with Qiagen Miniprep spin columns, suspended in water and sequenced. Sequencing reactions were performed with ABI Prism cycle sequencing kits (Perkin Elmer) according to the manufacturer’s instructions. Sequencing was conducted on ABI 377 sequencers. All samples were sequenced using primers based on vector sequence. Internal sequencing primers (sequences available on request) were used to provide double-stranded coverage. Sequences were assembled manually and aligned using CLUSTAL X (Jeanmougin et al., 1998). Relevant sequence information has been deposited in GenBank and/or is available on request.

The basal thermotolerance of T32 adults exceeded that of T25 adults, which in turn exceeded that of OR25 adults (Fig.1A). The corresponding LT₅₀ values for 30min heat shock, estimated by linear interpolation, are as follows: OR25, 39.5°C; T25, 39.8°C; T32, 40.3°C. Of the T32 adults, 80% survived a 40°C heat shock, which killed 75% of T25 adults and 100% of OR25 adults.

In contrast, the inducible thermotolerance (quantified as the change in percentage survival with and without pretreatment) of T32 flies was equal to or less than that of T25 and OR25 flies (Fig.1B–D). At the minimum temperature for 100% mortality without pretreatment (40°C for OR25, 40.5°C for T25, 41°C for T32), a 36°C pretreatment improved survival by 50% in OR25 flies, by 25% in T25 flies and by less than 10% in T32 flies. Inducible thermotolerance was not entirely absent in T32 flies; 37°C and 38°C pretreatments both improved survival at 40.5°C (from 25 to 75%) in the T32 strain.

Hsp70 levels after pretreatment and heat shock were likewise lower in the T32 strain than in the other strains (Fig.2). In OR25 flies, Hsp70 levels increased during exposure to high temperature, continued to increase for 1–3h afterwards, depending on temperature, and then decreased. In the T32 strain, the peak Hsp70 level was only 30–50% of the level in OR25 flies. T25 flies were intermediate in this respect. Immunoblots of Hsp70 corroborate this finding (Fig.3). Similarly, amounts of hsp70 mRNA, determined 1h after a 30min exposure to 35°C and 37.5°C, were also greatest in OR25, intermediate in T25 and lowest in T32 (Fig.4). After 39°C heat shock, in contrast, hsp70 mRNA levels in T32 were intermediate to those in the other strains (Fig.4), and Hsp70 was detectable only in the T strains (Fig.3). As with Hsp70 levels, hsp70 mRNA abundance was least after the most intense heat shock in all strains. According to ELISA, the kinetics of Hsp70 variation were similar in the three strains, and the T32 strain did not accumulate more Hsp70 than the other strains at the highest temperature examined (Fig.2, inset). Rearing eggs of the T32 strain to adulthood at 25°C increased Hsp70 levels above those in both the T32 strain reared at 32°C and the T25 strain (Fig.2, arrow in inset).

Immunoblots with antibody 7.10, which detects all D. melanogaster Hsp70 family members, suggest that amounts of Hsp70 family members in non-heat-shocked flies are no greater in the T strains than in OR25 (Fig.3). Although the aggregate Hsp70 family signal increases after heat shock, the increased level is less in the T strains than in OR25.

The T32 flies reared at 25°C expressed much more hsp70 mRNA after 35°C heat shock than the T32 flies reared at 31–32°C. This difference was absent or much less evident after 37.5°C and 39°C heat shock (Fig.4).

Two-dimensional gel electrophoresis detected numerous differences between the protein patterns of OR25 and T flies, both in identifiable heat-shock proteins and in other proteins (Fig.5, Fig.6). A 45kDa protein labeled conspicuously after heat shock in T but not in OR25 (Fig.5, arrow c). Several small (22–27kDa) heat-inducible proteins, presumably members of the D. melanogaster small Hsp family (Michaud et al., 1997), also differ in mobility between T and OR25 (Fig.5, arrows 27,b,d,e). The Hsp70 family clearly differed between T and OR25 strains. Both strains expressed Hsp68 (Fig.6), but this expression was more rapid in OR25 (Fig.6C) than in T (Fig.6E). Also, a third (i.e. in addition to Hsp68 and Hsp70) heat-inducible family member was evident in T but not in OR (low MW in Fig.6E). This protein co-migrates with Hsp70 in the isoelectric focusing dimension and has a slightly lower molecular mass than Hsp68. It is also detected by antibody 7FB, which recognizes D. melanogaster Hsp70 but not other D. melanogaster Hsp70 family members (Palter et al., 1986) (data not shown).

Genomic Southern blots (Fig.7), PCR and restriction digest of PCR products (Fig.8) detected several presumptive polymorphisms at both the 87A7 and 87C1 hsp70 loci in both strains. The nucleotide sequence at both loci reveals several noteworthy polymorphisms (Fig.8). The 87A7 locus is polymorphic in natural populations; an approximately 140bp insertion, presumably derived from a transposition event, is upstream of the coding sequence of the hsp70Ab gene in the ‘56H8’ morph but absent from the ‘122’ morph (Goldschmidt-Clermont, 1980). In both the T25 and T32 strains, 391bp upstream of this insertion is an additional 265bp insertion 98% identical to the long terminal repeat (LTR) of a full-length H.M.S. Beagle transposable element (Snyder et al., 1982). In each T strain, 36 individual flies were homozygous for the H.M.S. Beagle element. At the 87C1 locus, a 1.4kilobase (kb) fragment of a Jockey element (Mizrokhi et al., 1985; Priimagi et al., 1988) disrupts the promoter of the hsp70Ba gene in some T strain flies. The first two HSEs (HSEs 1 and 2) are intact, but the Jockey element intervenes between these HSEs and HSEs 3 and 4. Jockey was fixed in 36 T25 flies sampled (all homozygous) and occurred at an allelic frequency of 0.611 in 36 T32 flies sampled.

In electrophoretic mobility-shift assays, three HSE complexes (A–C) are evident and differ in their characteristics (Fig.9). These complexes are no longer detectable when lysates are incubated with a 200-fold molar excess of unlabeled HSE probe (Fig.9A, lane 5). Corresponding supershifts after incubation with anti-HSF antibody (Fig.9A, lane 3) and anti-Ku autoantigen antibody (lane 4) suggest that A is an HSE complex that includes HSF and that B is an HSE complex that includes Ku autoantigen. Neither antibody affected the mobility of complex C. Complexes B and C are constitutively present, except immediately after 40°C heat shock (neither detectable) and 1h after 40°C heat shock (complex B not detectable) in the T strain. In both the OR25 and T strains, the HSF–HSE complex (complex A) is either not detectable or detectable only at low concentrations at 25°C, and occurs at much higher concentrations after both 35°C and 37°C heat shock, corresponding to heat shocks at which Hsp70 level was maximally elevated. No qualitative differences between T and OR25 in complex A are evident that might be correlated with the lower levels of Hsp70 in the T strains. Although the T and OR25 strains differ in complex A after a 40°C heat shock, this is a heat shock that is lethal to some (T) or all (OR25) flies (Fig.1) and so may represent severe damage rather than a regulatory response.

The investigation of ecological and geographical variation in heat-shock protein expression, begun by Evgenev and colleagues (Evgenev et al., 1978; Evgenev et al., 1987; Lyashko et al., 1994; Ulmasov et al., 1993; Ulmasov et al., 1988; Zatsepina et al., 2000) and now widespread (Feder and Hofmann, 1999), clearly establishes that the magnitude, kinetics, threshold and molecular diversity of Hsp expression are correlated with the prevailing levels of stress that species, populations and developmental stages naturally undergo (Feder, 1999b; Feder and Hofmann, 1999). In addition, the work of Evgenev and colleagues suggests that species from high-temperature climates have higher constitutive levels of Hsp70 family members than related species from more moderate climates. At least two major questions are still unresolved: how do natural selection and other evolutionary processes create and maintain this covariation between the heat-shock response and environment, and which genetic elements (e.g. cis- and trans-regulatory elements, gene copy number, coding sequence) does evolution manipulate to achieve this covariation? The answers to these questions may be most evident from intraspecific variation because variation among conspecifics is presumably the raw material upon which selection acts. To this end, we compared the T strain of D. melanogaster with another strain of the same species and found many noteworthy differences. Indeed, differences between the T strains and the Oregon R wild-type strain are much greater in some respects (e.g. in protein patterns revealed by two-dimensional electrophoresis, Fig.5) than between different species of Drosophila (Sinibaldi and Storti, 1982).

At the outset, the extraordinary thermotolerance of the T strain was evident from its ability to thrive above temperatures previously considered to be the limit for continuous culture of this species (Parsons, 1973). This exceptional ability clearly extends to tolerance of acute high-temperature exposure (i.e. basal thermotolerance), in which the T strain in the present study outperforms other Drosophila melanogaster strains and compares favorably with many cactophilic Drosophila species (Krebs, 1999; Krebs and Loeschcke, 1995a; Krebs and Loeschcke, 1995b; Parsons, 1979; Stratman and Markow, 1998). In other strains and species, mechanisms of basal thermotolerance may include enhanced stability of cellular proteins through modification of primary structure and synthesis of thermoprotective osmolytes such as trehalose and sorbitol, constitutively expressed molecular chaperones, homeoviscous adaptation (i.e. adjustment in the lipid constituents of cells) and regulated depression of cellular function to moderate energy requirements. Although only semi-quantitative, the various determinations of constitutively expressed molecular chaperones reveal no obvious differences between T and OR25 flies (Fig.3, Fig.5, Fig.6); if anything, levels are lower in T flies than in OR25 flies. Investigation of the other mechanisms of basal tolerance in the T strain, in which they are presently unknown, may well be fruitful.

In contrast, in terms of inducible thermotolerance, the T strain is unremarkable if not comparatively poor. Heat-shock protein expression is a principal mechanism of inducible thermotolerance in D. melanogaster (Feder et al., 1996; Solomon et al., 1991; Welte et al., 1993). Correspondingly, the T strain exhibits both lower amounts of most Hsps after heat shock and slower kinetics of Hsp70 and Hsp68 expression than the OR25 strain. The only obvious exceptions to this conclusion are an apparent low-molecular-mass Hsp70 and several putative small Hsps (Fig.5), the latter identified exclusively by electrophoretic mobility and heat inducibility. Small Hsps, related to α-crystallins, clearly protect against stress (including high temperature) in other species, and their encoding genes have evolved in response to environmental temperature in D. melanogaster (Frydenberg et al., 1999). Their role in thermotolerance has not been confirmed in D. melanogaster, however. In any event, decreased inducible thermotolerance and increased expression of small Hsps would seemingly be inconsistent. The low-molecular-mass Hsp70 (Fig.6) is detected both by an antibody specific for all Hsp70 family members (7.10) and by one specific for D. melanogaster Hsp70 (7FB), and its electrophoretic mobility resembles that of Protein 38 in Fig.2B of Buzin and Petersen (Buzin and Petersen, 1982). Indeed, Buzin and Peterson detected isoforms of Hsp70 considerably more numerous than the Hsp70-encoding genes, and they hypothesized that this was due to alternative post-translational modifications. In the T strain, however, this particular Hsp70 is far more abundant than in other strains (Buzin and Petersen, 1982; Palter et al., 1986). Whether this Hsp70 is functional and the basis for its low molecular mass are presently unknown.

Part of the distinctive thermal phenotype of the T strain is due to its thermal history while in laboratory culture. The T25 strain is intermediate to OR25 and T32 in basal thermotolerance, in inducible thermotolerance, in Hsp70 expression (present study) and in several other aspects of resistance to heat and ionizing radiation (Tikhomirova, 1980; Tikhomirova and Belyatskaya, 1993; Tikhomirova et al., 1993). Laboratory evolution at the respective culture temperatures of the T25 and T32 strains is clearly sufficient to generate such differences in thermal phenotype and the heat-shock response (Bettencourt et al., 1999; Gilchrist and Huey, 1999; McColl et al., 1996; McKechnie et al., 1998; Sorensen et al., 1999), and the two strains differ in hsp70Ba allele frequency. Within a generation, culture at different temperatures (i.e. thermal acclimation) can also affect diverse traits in D. melanogaster, and these effects can persist for multiple generations (Crill et al., 1996; Hercus and Hoffmann, 2000). Thermal acclimation can affect the threshold temperature for induction of a heat-shock response (for a review, see Lerman and Feder, 2001), but several attempts to demonstrate a large effect in D. melanogaster (Bettencourt et al., 1999; Lerman and Feder, 2001) have failed.

This work adds to a growing list of Drosophila studies suggesting that, under certain conditions, evolution at high temperatures leads to decreased expression of Hsp70, a seemingly paradoxical outcome given the role of Hsp70 in thermotolerance in Drosophila. D. melanogaster undergoing laboratory natural selection at 28°C (henceforth ‘Cavicchi 28°C lines’) and D. buzzatii (a cactophilic species) reared for 6h each day at 38.2°C both express less Hsp70 than 25°C controls (Bettencourt et al., 1999; Sorensen et al., 1999). Similarly, subtropical D. buzzatii collected from a low (i.e. warm) elevation express less Hsp70 than conspecifics collected from a high (i.e. cool) elevation (Sorensen et al., 2001). We suggest, as have others (Bettencourt et al., 1999; Krebs and Feder, 1997a; Krebs and Feder, 1997b; Krebs and Feder, 1998b; Krebs et al., 1998; Lansing et al., 2000; Sorensen et al., 2001; Sorensen et al., 1999), that this pattern is due (i) to the low levels of Hsp70 present in wild-type D. melanogaster transiently exposed to temperatures that these high-temperature lines chronically experience, (ii) to deleterious consequences of these low Hsp70 levels and (iii) to natural selection to reduce these low Hsp70 levels, which reduces Hsp70 levels at all temperatures as a correlated response to selection (Fig.10).

Hsp70 is normally undetectable in D. melanogaster cells and whole D. melanogaster cultured at 25°C and below (Velazquez et al., 1980; Velazquez et al., 1983), which are typical temperatures for D. melanogaster maintenance in the laboratory. Exposure to slightly higher temperatures (e.g. 28°C) results in HSF activation (Lerman and Feder, 2001) and low, but non-zero, concentrations of Hsp70 (Lindquist, 1980). The constitutive presence of Hsp70 can be harmful in at least three ways: by acting as a sink for substrate and cellular machinery that could be devoted to the synthesis of other proteins (Krebs and Feder, 1998a), as an impediment to normal processing and/or degradation of non-native proteins (Dorner et al., 1988; Dorner et al., 1992; Johnston et al., 1998; Newnam et al., 1999) and as an inappropriately regulated extra- and intracellular participant in diverse signaling pathways involving cell growth and death (Jaattela, 1999a; Jaattela, 1999b; Xie et al., 1999). Indeed, constitutive expression or overexpression of Hsp70 in D. melanogaster reduces growth, development and, potentially, fitness (Feder et al., 1992; Krebs and Feder, 1997a; Krebs and Feder, 1998b; Krebs and Loeschcke, 1994; Roberts and Feder, 2000), except in the context of inducible thermotolerance, where it clearly enhances fitness (Feder, 1999a; Roberts and Feder, 1999; Roberts and Feder, 2000). Strains evolving at moderately high but constant temperatures realize only the deleterious consequences of Hsp70, but none of the benefits, because these strains never encounter the extremely high temperatures against which Hsp70 is protective. In response, natural selection may have acted to reduce the entire Hsp70 versus temperature norm of reaction (Fig.10), which results in decreased Hsp70 levels after exposure to temperatures commonly used in heat-shock studies. Interestingly, soil arthropods populating soils contaminated with high levels of heavy metals (another inducer of Hsp70) also evolve decreased Hsp70 expression (Kohler et al., 2000). Whether natural selection can result in an Hsp70 versus temperature norm of reaction that includes both suppression of Hsp70 expression at moderately high temperatures and increased Hsp70 expression at typical heat-shock temperatures or whether the levels of Hsp70 at moderately high and heat-shock temperatures are inextricably linked by a constraint of the Hsp70 autoregulatory apparatus (Lindquist, 1993) remains to be determined.

Regardless of its ultimate significance (sensu Mayr), the lower levels of Hsp70 expression in the T strains must have a mechanistic basis. Features determining the concentration of Hsp70 in D. melanogaster cells include hsp70 copy number (Feder et al., 1996; Solomon et al., 1991; Welte et al., 1993), chromatin structure (Wu, 1980), transcription (Li et al., 1996; Mason and Lis, 1997; Morimoto et al., 1994), RNA processing (Lindquist, 1993; Yost et al., 1990), message stability (Petersen and Lindquist, 1989) and translation (Hess and Duncan, 1996; Zapata et al., 1991) and the sequestration and degradation of protein (Feder et al., 1992). As in prior studies (Bettencourt et al., 1999), we find no unequivocal evidence for genes other than the typical five hsp70 genes in the T strain (data not shown). At the level of transcriptional activation, the T strain exhibits the same temperature-sensitivity of HSF–HSE interaction as the OR25 strain (Fig.9). As in other organisms (Zatsepina et al., 2000), HSF constitutively forms additional complexes, one with Ku autoantigen and one of unknown composition, but both complexes are similarly abundant in both the T and OR25 strains. Presumably, therefore, the lower Hsp70 concentrations in the T strain have their basis downstream of HSF activation and binding to HSEs; indeed, the sizes of restriction fragments and PCR products indicate polymorphism in the hsp70 loci, both in the strains examined here and worldwide (Bettencourt, 2001). One discrete candidate mechanism is disruption of the hsp70 promoters by transposable elements, leading to altered transcription. It has been shown that both H.M.S. Beagle and Jockey elements may affect the expression of other genes whose promoters they disrupt (Kimbrell et al., 1988; Kimbrell et al., 1989; White and Jacobson, 1996a; White and Jacobson, 1996b). When present in the hsp70Ba promoter of the T strain, the Jockey element intervenes between the proximal and distal pairs of HSEs. Although the two proximal HSEs constitute the minimal Hsp70 promoter, the distal HSEs contribute to heat-inducible hsp70 transcription (Lee et al., 1992; Topol et al., 1985). The H.M.S. Beagle element is in a region of the 87A7 locus whose impact on hsp70 transcription is unclear. Nonetheless, the H.M.S. Beagle sequence includes putative enhancer-like elements that may affect transcriptional regulation at some distance from the promoter (Kimbrell et al., 1988; Kimbrell et al., 1989). Transposable elements may be a significant evolutionary force in D. melanogaster via their alteration of the pre-existing genome, and it is tempting to conclude that these elements reduce Hsp70 levels in the T strain, which natural selection has then favored in the native environment of the strain. Whether disruption of one or two hsp70 genes out of five would be sufficient to reduce the overall transcription of hsp70 mRNA is not known, however, and whether the H.M.S. Beagle and Jockey elements actually disrupt transcription as hypothesized requires further study.

The heat-shock genes and the proteins they encode are among the most ancient and highly conserved known, so much so that they have been considered useful in defining the phylogeny of major species groups in Drosophila (Bettencourt and Feder, 2001; Drosopoulou et al., 1996; Molto et al., 1994; Sinibaldi and Storti, 1982) and even phyla and kingdoms (Feder and Hofmann, 1999). Against this background, the variation among populations of a single species (i.e. T versus Oregon R) is remarkable. Evidently, adaptation via natural selection is sufficiently strong to overcome even the immense phylogenetic inertia of the heat-shock response.

In Moscow, the research was supported by Russian Grants for Basic Science 0004-48-285 and 0004-32-243 and by a grant from the Russian Foundation to Support Fundamental Science (grant 99-04-49222). In Chicago, the research was supported by an International Supplement to NSF grant IBN97-23298 and by IBN99-86158. HHMI graduate fellowships supported B.R.B. and D.N.L. We thank Jean David for clarifying the provenance of the T strain and H. Dai, A. Helin and P. Michalak for comments on the manuscript.