Proteome stability, heat hardening and heat-shock protein expression profiles in Cataglyphis desert ants

Protein structural stability is paramount to the maintenance of cell integrity and is highly dependent on temperature. Proteotoxic stressors, such as heat shock, can cause proteins to misfold and denature, resulting in their aggregation and a subsequent loss of biological function and an impairment of normal metabolic processes (Sørensen et al., 2003; Mayer, 2010). Accordingly, the heat-shock response (HSR) was selected for early on in evolutionary time and has been widely conserved across taxa (Moseley, 1997; Feder and Hofman, 1999). It includes the production of molecular chaperones – such as heat-shock proteins (HSPs) – that are involved in the assembly, folding and proper translocation of cellular proteins (Sørensen et al., 2003; Moseley, 1997; Feder and Hofman, 1999). Expression of HSPs depends on the activation of transcriptional heat-shock factors (HSFs). HSF1 trimerises during activation and binds to specific cis-regulatory elements in the genome – known as heat-shock elements (HSEs) – and thus induces the expression of target genes, including molecular chaperones (Sarge et al., 1993; Zatsepina et al., 2000). The constitutive expression of HSPs and their relative levels of induced expression under stressful conditions are associated with an organism's adaptive capacity to tolerate thermal pressures (Evgen'ev et al., 2014). Comparisons among species from contrasting environments have shown that thermophilic species usually have higher constitutive levels of HSPs than do their mesophilic relatives. However, their relative induced expression of HSPs is lower in response to heat stress (Rinehart et al., 2006; Dong et al., 2008; Clack and Peck, 2009; Evgen'ev et al., 2014). Moreover, in thermophiles, HSF activation and the resulting induction of HSP synthesis occur at higher temperatures. Ultimately, proteins of thermophiles are more stable than those of mesophiles, whether as a result of structural changes or as a consequence of large amounts of constitutive HSPs (Evgen'ev et al., 2014). It has been shown that both the duplication of hsp genes and shifts in HSE arrangement and position allow for greater HSR functional plasticity, helping organisms adopt the best-suited response to environmental heat stress (Bettencourt and Feder, 2001; Tian et al., 2010; Nguyen et al., 2016). The use of this system is well developed in ectotherms, which cannot regulate their body temperatures and whose metabolic processes and stability therefore depend on microclimatic conditions (Hazel, 1995; Lin and Somero, 1995; Willmer et al., 2000). Consequently, ectotherms are interesting models with which to explore HSR patterns under conditions of thermal stress (Evgen'ev et al., 2014).

Cataglyphis desert ants present a unique opportunity for investigating the HSR in heat-tolerant terrestrial ectotherms. All Cataglyphis species inhabit arid lands or deserts – they are found in the Sahara, the Near East, the Middle East, the Arabian Peninsula and Central Asia (Lenoir et al., 2009; Boulay et al., 2017). In contrast to many other desert-adapted species, which tend to escape the heat by being nocturnal or crepuscular, Cataglyphis ants take advantage of their high thermal tolerance to remain active even during the hottest parts of the day: most are scavengers and collect the bodies of less-tolerant, heat-stricken arthropods (Harkness and Wehner, 1977; Wehner et al., 1992). These ants are generally equipped with long legs, which increase the distance between their bodies and the soil; this trait reduces exposure of the body core to high ground temperatures, thus improving body cooling, and also enhances running velocity (Marsh, 1985; Sommer and Wehner, 2012). Previous studies have shown that foragers of the Saharan silver ant, Cataglyphis bombycina, produce what was thought to be HSP70 before emerging from the nest, thereby bypassing the need to acclimate to sudden heat exposure (Gehring and Wehner, 1995). However, more recently, it was discovered that the primary heat-induced gene hsp70 has been completely lost in Hymenoptera; instead, two hsc70-4 paralogues are expressed (Nguyen et al., 2016). At present, the ability to express other HSPs and to induce gene expression in response to heat stress remains completely unexplored in Cataglyphis.

In this study, we compared the HSRs of two closely related Cataglyphis species – C. bombycina and C. mauritanica – found in natural habitats that contrast in mean annual temperature and temperature seasonality (Fig. 1). Cataglyphis bombycina is found in warm deserts, which are characterised by extremely high summer temperatures and warm winters. This species has one of the highest critical thermal maximum values observed among terrestrial organisms (CT_max=53.6°C; Hoffmann et al., 2013). Workers are active during the hottest parts of the day, when air temperatures may exceed 50°C (Wehner et al., 1992). Their lower limit for foraging activity is influenced by the presence of their main lizard predator, Acanthodactylus dumerili, which retreats to its underground burrow when temperatures becomes unbearable. Workers are thus constrained to forage within a narrow thermal window (46.5–53.6°C). Cataglyphis mauritanica, in contrast, lives in the Atlas Mountains, which are characterised by a cold semi-arid climate, with warm summers and cold winters. This species experiences heat stress during the summer, and their habitat exhibits extreme diurnal and seasonal temperature fluctuations that are associated with altitude (Dillon et al., 2006; Peel et al., 2007). As a result, when foraging, C. mauritanica workers probably face lower but much more variable temperatures than do C. bombycina workers.

First, we compared the species' general heat tolerance and their ability to induce thermotolerance via heat hardening (i.e. the synthesis of thermoprotective molecules following mild heat stress; Feder and Hofmann, 1999) by exposing workers to sublethal and lethal heat shocks. Second, we quantified in vivo protein damage resulting from heat shock by examining protein aggregation levels. Third, we compared the levels of constitutively expressed HSC70 between the two species. Fourth, we characterised the induced expression of several hsp genes – hsf1, hsc70-4h, hsp83 and hsc70-5 – known to be associated with HSR, heat tolerance and protein folding (Sarge et al., 1993; Feder and Hofmann, 1999; Nguyen et al., 2016). Based on current knowledge regarding the molecular mechanisms underlying adaptations to extreme environments (Evgen'ev et al., 2014), we predicted that the more thermophilic species, C. bombycina, would constitutively express molecular chaperones at higher levels than would the more mesophilic species, C. mauritanica. Similarly, we predicted that C. bombycina would show lower levels of induced hsp gene expression, reduced use of heat hardening to increase thermal tolerance, and lower levels of heat-shock-related protein damage.

Colonies of C. bombycina Roger 1859 were collected near Zagora (30°19′56″N, 5°50′18″W), located in the Draa Valley in southern Morocco. Zagora is characterised by a warm desert climate; it experiences extremely high summer temperatures from early spring to late autumn and warm winters (mean summer temperature 33.4°C, altitude 724 m). Colonies of C. mauritanica (Emery 1906) were collected in Ifrane National Park (33°33′22″N, 5°14′15″W), located in the Atlas Mountains of Morocco. Ifrane is characterised by a cold semi-arid climate; it experiences warm summers and cold winters (mean summer temperature 22.7°C, altitude 1730 m; Peel et al., 2007). Colonies were kept in 30×40×10 mm plastic boxes with Fluon^®-coated sides and a thin layer of clean sand on the bottom. For nesting, they were given 16×150 mm glass test tubes; the bottoms contained water, which ants could access via a moist cotton plug. The ant colonies were reared under constant environmental conditions: a temperature of 25°C, a 12 h:12 h light:dark cycle and 60% relative humidity. They were provided with a sugar solution ad libitum and sliced mealworms twice a week. They were kept under laboratory conditions for at least 2 months prior to experimentation. All the experiments were performed on workers hatched and raised in the laboratory (i.e. from the egg to adult stage).

General thermal tolerance was determined using heat-stress experiments. Three temperature treatments were used – 37, 40 and 45°C – which corresponded to mild, medium and severe heat stress. Previous studies (Gehring and Wehner, 1995) showed that HSC70 synthesis continues up to 45°C in the silver ant. For each treatment, three to six groups of 15 workers were used per species. Worker groups were placed in 50 ml glass vials containing moist cotton balls. The vials were then submerged in a digital water bath (SW22, Julabo GmbH, Seelbach, Germany); the temperature inside the vials was monitored using 0.075­ mm diameter thermocouples [Type K Thermocouple (Chromel/Alumel), RS Components Ltd, Glasgow, UK] connected to a digital thermometer (RS Pro RS52 Digital Thermometer, RS Components Ltd). The ants were exposed to the treatment for 3 h. This duration of exposure was previously used to induce a significant HSR in C. bombycina (Gehring and Wehner, 1995). The species' use of heat hardening was assessed by pre-exposing workers to mild heat stress (37°C) for 2 h before they experienced severe heat stress (45°C) for 3 h. In both the thermal-tolerance and heat-hardening experiments, ants that could still move after the treatments were considered to be alive. We used worker survival rates as proxies for levels of thermal tolerance and heat hardening.

To examine the protein damage caused by heat stress, we compared protein aggregation levels between control workers, which spent 3 h at 25°C, and heat-shocked workers, which spent 3 h at 45°C; the workers were immediately flash frozen following treatment. Nine groups of six individuals were used per treatment per species. Protein aggregates were extracted and purified as per Chen et al. (2002) and Rinehart et al. (2006). Briefly, workers were pooled and placed in a chilled extraction buffer consisting of 0.1% Triton X-100, 60 mmol l⁻¹ PIPES, 1 mmol l⁻¹ EDTA, 1 mmol l⁻¹ EGTA, 100 mmol l⁻¹ NaCl, 0.5 mmol l⁻¹ PMSF, 0.75 mg l⁻¹ leupeptin and 0.1 mmol l⁻¹ dl-DTT (Sigma-Aldrich Chemie GmbH, Munich, Germany). They were then homogenised for 3 min at maximum speed in a mixer mill (MM 301, Retsch GmbH, Haan, Germany) with 2.8 mm stainless steel beads. The resulting homogenate was filtered through a 21-gauge needle (Sterican^®, Braun, Kronberg, Germany) to remove large pieces of cuticle and organs. The filtrate was then centrifuged at 680 g for 10 min at 4°C to pellet nuclei and large cell fragments. The supernatant was centrifuged at 35,000 g for 14 min at 4°C (Optima™ L-70K Ultracentrifuge, Beckman Coulter, Chaska, MN, USA) to separate the Triton-soluble fraction from the Triton-insoluble fraction (Chen et al., 2002). To purify the protein aggregates, the Triton-insoluble fraction was twice suspended, then sonicated and pelleted at 17,000 g for 30 min at 4°C. The resulting pellet was resuspended, sonicated and pelleted at 5000 g for 30 min at 4°C. Then, the pelleted protein aggregates were resuspended and assayed for protein content. Protein concentrations were determined using the Bradford procedure (BioRad, Hercules, CA, USA) and BSA as a standard (Rinehart et al., 2006).

Constitutive levels of HSC70 were estimated by western blotting. We extracted total protein from 20 mg of randomly selected control and heat-shocked workers (see above) per species. Protein concentrations were determined using the Bradford procedure (BioRad). A 15 μg sample of the extract was mixed with 6-fold concentrated Laemmli buffer (1% SDS, 5% β-mercaptoethanol and 10% glycerin; Sigma-Aldrich Chemie GmbH) in a 1:6 ratio and kept at 95°C for 5 min. Samples were then diluted in Laemmli buffer and analysed in duplicate by SDS-PAGE on a 10% gel. The gel was stained with Coomassie Blue and the total signal for each lane was quantified using a CCD camera, which established the control loading values for each sample. Proteins were then electrotransferred onto an Amersham Protran 0.45 μm nitrocellulose blotting membrane (GE Healthcare Life Sciences, Little Chalfont, UK). Skimmed milk (1.5%) was used for blocking (1 h, 20°C) and to dilute the antibodies. The membrane was washed three times with TBS-T buffer (10 mmol l⁻¹ Tris, 150 mmol l⁻¹ NaCl and 2% Tween-20) and incubated overnight at 4°C with an anti-HSP70 monoclonal mouse antibody (1/2000 dilution; clone BRM-22, Sigma-Aldrich, Rehovot, Israel). Secondary antibodies (1/10,000 dilution; anti-mouse IgG peroxidase, Sigma-Aldrich) were then added, and the solution was left at 20°C for 2 h. Signal strength was quantified using an Odyssey^® FC Imaging System (Li-Cor^® Bioscience, Lincoln, NE, USA).

To examine the species' ability to induce the expression of the target hsp genes in response to heat stress, total RNA was extracted from control workers, heat-shocked workers (see above) or heat-shocked workers after 1 h respite time at 25°C. We used three groups of six randomly selected workers per treatment per species. Samples were pooled by treatment and placed in 1 ml of chilled Trizol reagent (Invitrogen, Carlsbad, CA, USA). They were then homogenised for 3 min at maximum speed in a mixer mill (MM 301, Retsch GmbH) with 2.8 mm zirconium oxide beads (Bertin Technologies, Montigny-le-Bretonneux, France). Total RNA was extracted according to the manufacturer's instructions. RNA was quantified with a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific Bvba, Gent, Belgium). The 260 nm/280 nm absorbance ratios for the samples were between 1.8 and 2.0, indicating sufficient RNA purity.

After DNase treatment (amplification-grade DNAse I, ThermoFisher Scientific, Carlsbad, CA, USA), 500 ng of total RNA was retrotranscribed using a RT-mix kit (SuperScript^® III Reverse Transcriptase, ThermoFisher Scientific). Primers were designed according to quality guidelines for qPCR to detect specific target genes (Thornton and Basu, 2011; Table S1). Quantitative PCR (qPCR) was performed using an ABI StepOnePlus Real-Time PCR System (ThermoFisher Scientific). We used 20 μl reaction volumes containing 2 ng of template cDNA, 8 nmol l⁻¹ of total primer and 10 μl of SYBR^® (Premix Ex Taq™ II, Takara Bio USA Inc., Mountain View, CA, USA). The following temperature program was used: an initial incubation step at 95°C for 2 min, then 40 cycles at 95°C for 15 s, and finally an annealing step followed by an extension step, each at 60°C for 60 s. Melt curve analysis confirmed the presence of a single amplicon. Control qPCR reactions were performed using RNA extract as a template to confirm the absence of genomic DNA contamination. Relative changes in gene expression were calculated using the ΔΔCT method (Livak and Schmittgen, 2001) and Rest software v.2.0.13 (Pfaffl et al., 2002). We screened for several commonly used housekeeping genes; the gene set we used for standardisation [i.e. β-actin and eukaryotic elongation factor-1 (eEF-1) ] was determined using both Normfinder (Andersen et al., 2004) and GeneNorm (Vandesompele et al., 2002).

Survival rates in the thermal tolerance and heat-hardening experiments, as well as levels of protein aggregation in response to heat stress, were compared using one-way analyses of variance (ANOVA). Multiple comparisons among pairs of means were carried out using Tukey post hoc tests. Because 100% of C. mauritanica workers survived in the 37 and 40°C treatments (meaning residuals did not meet the assumption of normality; see Results), we also tested for an association between survival and temperature (37, 40 and 45°C) for each species using a non-parametric Spearman's rank correlation test. Before analysing the gene expression patterns, the relative expression values were log transformed to meet assumptions of normality. The significance of treatment on gene upregulation in each species was tested using ANOVA followed by Tukey post hoc tests. Relative changes in gene expression between species were analysed with Student’s t-tests.

To assess general thermal tolerance, we measured the survival rates of C. bombycina and C. mauritanica workers exposed to mild (37°C), medium (40°C) and severe (45°C) heat stress (Fig. 2). Cataglyphis bombycina workers showed a slight, but non-significant decrease in survival rate in response to mild and medium heat stress (mean±s.d. survival rate: 93±4.4% and 88.7±7.7%, respectively), while all the C. mauritanica workers in these heat-stress treatments survived (mean±s.d. survival rate: 100±0% for both temperatures). However, severe heat stress resulted in a dramatic decrease in worker survival in both species (mean±s.d. survival rate for C. bombycina: 33±8.9% and C. mauritanica: 45.5±13.51%; one-way ANOVA: F_7,40=69.91, P<0.0001; Tukey post hoc test, P<0.05). Consistent with this result, Spearman's rank correlation tests showed a significant negative association between survival rate and temperature within each species (C. bombycina: r_s=−0.774, n=12, P=0.003; C. mauritanica: r_s=−0.868, n=10, P=0.002). Worker survival did not differ between species (P>0.05).

There was no evidence of heat hardening in C. bombycina. Rather, treatment of workers appeared to decrease survival (13.3±4.4%; P<0.05). In contrast, in C. mauritanica, treated workers had significantly higher survival rates (control: 45.5±13.51% versus treated: 65.9±10.6%; P<0.05), suggesting heat hardening occurred in this species (Fig. 2).

We examined the protein damage resulting from heat stress by comparing protein aggregation levels between control and heat-shocked workers. Control workers had low levels of protein aggregation (mean±s.d., C. bombycina: 1.39±1.04 ng µg⁻¹ total protein extract; C. mauritanica: 1.42±0.54 ng µg⁻¹; Fig. 3). When exposed to heat stress, both species showed increased levels of protein aggregation (C. bombycina: 5.01±2.24 ng µg⁻¹; C. mauritanica: 9.53±2.37 ng µg⁻¹; F_3,56=23.23, P<0.0001; Tukey post hoc test: P<0.05). Protein aggregation levels in heat-shocked workers were significantly higher in C. mauritanica than in C. bombycina (P<0.001).

The concentration of HSC70 in the protein extract was assessed by western blotting and corrected for the total amount of protein loaded per sample using Coomassie Blue signal strength. The results are expressed as the ratio of HSC70 to total protein. Higher constitutive levels of HSC70 were detected in C. bombycina than in C. mauritanica (Fig. 4).

The target hsp genes showed different expression patterns in response to heat shock. We found that hsf1 expression was not upregulated (Fig. 5A), nor did it differ between the two species. In contrast, hsp83 and hsc70-5 expression was upregulated in both species (C. bombycina: F_6,14=6.64, P<0.002; C. mauritanica: F_6,14=31.15, P<0.0001; Tukey post hoc tests: P<0.05 for both species; Fig. 5B,C) but no interspecific difference was found (two-tailed t-test, P>0.07 for both genes). In the case of hsc70-4 h1, gene expression was significantly upregulated in C. mauritanica (P<0.01) but not in C. bombycina (between-species difference: t=3.85, d.f.=4, P<0.02; Fig. 5D). Finally, hsc70-4 h2 expression was upregulated in both species (P<0.01), and the species differed in their expression patterns (t=3.67, d.f.=4, P<0.03; Fig. 5E).

Compared with related mesophilic species, thermophilic species are expected to have higher constitutive expression but lower induced expression of HSP-related genes (Evgen'ev et al., 2014). The results of our study support this prediction: in the highly thermophilic C. bombycina, constitutive expression of HSC70 was higher than in C. mauritanica, but induced expression of hsc70-4 h1 and hsc70-4 h2 in response to heat shock was lower. Furthermore, C. bombycina workers suffered less protein damage upon heat shock and exhibited lower rates of heat hardening. Although the strength of inference that can be drawn from comparison of two species may be limited (Garland and Adolph, 1994), our study strongly supports that the HSR is highly correlated to the species’ thermal environment.

Few studies have investigated HSP synthesis under heat-stress conditions in ants. For those that have, HSP70 synthesis has largely been the focus (Gehring and Wehner, 1995; Maisov et al., 2007). However, Nguyen et al. (2016) recently showed that the primary inducible gene involved in HSR in Drosophila – hsp70 – has been completely lost in Hymenoptera, which instead employs two hsc70-4 paralogues. Like Gehring and Wehner (1995), we found that HSC70 is constitutively expressed in C. bombycina. We also found comparatively moderate levels of HSC70 in C. mauritanica under control conditions. High constitutive expression of molecular chaperones has been reported in a number of ectotherms living under stressful conditions (e.g. elevated temperature, high salinity), including molluscs (Dong et al., 2008), echinoderms (González et al., 2016), dipterans (Rinehart et al., 2006; Garbuz et al., 2008) and lizards (Zatsepina et al., 2000). The sustained production of HSC70 by C. bombycina and C. mauritanica is probably an adaptation for coping with heat. We expect elevated constitutive expression of HSC70 to occur across the entire Cataglyphis genus, as all its members are thermophilic.

To the best of our knowledge, only three studies have explored the expression of HSP-related genes in response to heat stress in ants. Beside the elevated constitutive level of HSC70 in C. bombycina, Gehring and Wehner (1995) also showed that HSC70 synthesis carries on at temperatures as high as 45°C in this species. In contrast, this synthesis is abolished above 40°C in the wood ant Formica polyctena, a mesic species closely related to Cataglyphis that inhabits the northern part of the Paleartic region. The second study investigated hsp60, hsp75 and hsp90 expression in response to temperature in the wood ant Formica cinerea and found no significant effects (Ślipiński et al., 2015). The third study compared heat-shock-related expression patterns of several hsp genes between Pogonomyrmex barbatus, an ant living in hot and arid climates, and Aphaenogaster picea, an ant living in cool woodlands (Nguyen et al., 2016). These authors showed that expression of hsc70-4, hsp83 and hsp40 was upregulated in response to heat stress and the species differed significantly in relative expression levels. More specifically, upregulation occurred at higher temperatures in P. barbatus. In contrast, induced expression of hsc70-5, hsc70-3 and hsp60 was unaffected by heat stress. Our findings add to the knowledge about HSR in ants. They show that, in Cataglyphis, hsf1 expression is not induced by heat shock. This result suggests that ambient concentrations of cytosolic HSF1 are sufficient to induce the expression of hsp genes under conditions of heat stress. Consistent with this hypothesis, in previous studies, differences in HSF1 activation and binding to HSEs, as opposed to upregulation, are crucial to the constitutive and induced expression of HSPs (Zatsepina et al., 2000). In response to heat shock, hsp83 expression was moderately upregulated in both C. bombycina and C. mauritanica. Upregulation of the hsp90 gene family in response to heat stress is well documented in ectotherms (Rinehart et al., 2007; Choi et al., 2014; González et al., 2016; Nguyen et al., 2016) and, based on our findings, also appears to be conserved in Cataglyphis. We also observed an increase in hsc70-5 expression in response to heat shock. This result was somewhat unexpected as this cognate has not been reported to demonstrate heat-inducible expression in the other ant species studied to date (Ślipiński et al., 2015; Nguyen et al., 2016). The search for conserved HSEs in the promoter region of hsc70-5 in Cataglyphis, as well as in other Formicinae, might yield further insights into whether this cognate is a unique tool used by Cataglyphis ants to increase proteome stability.

We found that, among all the hsp genes examined, hsc70-4 showed the greatest variation in expression in response to heat shock. This pattern was also observed in P. barbatus and A. picea (Nguyen et al., 2016). We also discovered differences in induced expression between species. First, while expression of both hsc70-4 paralogues was upregulated in C. mauritanica, expression of only one (hsc70-4 h2) was upregulated in C. bombycina. Consequently, in the latter species, hsc70-4 h1 might have evolved a housekeeping function to enhance proteome stability. This difference was not surprising, as the hsc70-4 family shows member duplication in various insects, suggesting a wide array of functional shifts have taken place over the course of evolution. In the mosquito Culex quinquefasciatus, differences in paralogue inducibility were inferred from the absence of HSEs in the promoter region of one copy of the gene, which is consistent with a housekeeping function; the other copy might remain heat inducible (Nguyen et al., 2016). Our data suggest the scenario may be similar in C. bombycina. Although expression of hsc70-4 h2 was induced at high levels in both species, the relative increase was higher in C. mauritanica. The relatively lower levels of induced expression in C. bombycina were expected and make sense given the high constitutive HSC70 concentrations observed. Indeed, a trade-off between the constitutive and induced expression of HSP70 has been documented in a number of organisms living under stressful conditions (Rinehart et al., 2006; Dong et al., 2008; Clark and Peck, 2009; Zatsepina et al., 2016). This trade-off seems at least partially conserved in Cataglyphis: compared with the more mesophilic C. mauritanica, the more thermophilic C. bombycina had relatively higher constitutive expression of HSC70 but relatively lower induced expression of the target hsp genes. This shift towards higher constitutive expression in C. bombycina may have been selected for by the narrow range of extremely high temperatures that workers experience upon leaving the nest to forage in the Sahara desert. In C. mauritanica, the shift towards induced expression might allow workers to cope with the less intense but more variable temperatures of the Atlas Mountains. Overall, these results suggest key differences exist in Cataglyphis HSRs, probably as a result of the contrasting environments the species inhabit.

Proteotoxic stressors such as heat denature proteins, causing them to aggregate because their insoluble residues become exposed (Chen et al., 2002). As predicted, levels of protein aggregation were significantly higher in heat-shocked C. mauritanica workers than in heat-shocked C. bombycina workers. This result is consistent with our western blot analyses, which show that constitutive expression of HSC70 is greater in the more thermophilic C. bombycina than in C. mauritanica. A similar positive correlation between levels of constitutively expressed molecular chaperones and proteome stability has been previously documented in the Antarctic midge, Belgicana antartica, which also tolerates extreme environmental conditions (Rinehart et al., 2006).

As predicted, in the heat-hardening experiment, the survival rate of C. mauritanica workers was significantly higher, strongly suggesting the species uses heat hardening to induce thermotolerance. This finding is consistent with our gene expression results, which showed that C. mauritanica workers strongly upregulated the expression of the two hsc70-4 cognates in response to heat stress. In contrast, in C. bombycina, only hsc70-4 h2 expression was upregulated, and to a lesser extent than in C. mauritanica. As the upregulation of hsp70 expression is closely tied to survival and heat hardening (Feder and Hofmann, 1999; Bahrndorff et al., 2009), our results suggest that C. mauritanica can acquire a high level of induced thermotolerance while C. bombycina cannot. In the latter species, a 2 h pre-treatment at 37°C increased mortality of workers subsequently exposed to a severe heat stress (3 h at 45°C). It has been shown that pre-exposure to temperatures above the range inducing thermotolerance can cause thermosensitivity instead (Yocum and Denlinger, 1993). Exposure to such critical temperatures results in a rather quick mortality among individuals and limits protein synthesis to molecular chaperones (Yocum and Denlinger, 1993). Contrary to these expectations associated with thermosensitivity, we did not observe any significant mortality of C. bombycina workers after 3 h exposure to 37 or 40°C. Furthermore, we detected ongoing transcriptional activity of non-molecular chaperone-related genes such as housekeeping genes (β-actin, eEF-1) and hsf1 (Fig. 5A) after 3 h at 40°C (see Results, ‘Heat stress and gene upregulation’). Therefore, and regardless of the highly thermophilic nature of C. bombycina workers whose CT_max reaches 53.6°C (Hoffmann et al., 2013), it is unlikely that exposure to 37°C is stressful enough to induce thermosensitivity or impair the HSR. Rather, the lower survival rate of C. bombycina workers after pre-treatment probably stems from the combination of prolonged exposure to above-optimal temperatures and lessened HSC70 up-regulation.

Another unexpected result of this study is that the two species suffered equal mortality rates (approximately 60%) when exposed to severe heat stress (Fig. 2). Given that C. bombycina workers constitutively express more HSC70, one would predict that this species should be able to tolerate higher body temperatures. One tentative explanation may involve a combination of worker morphology and experimental design. In C. bombycina, the high constitutive levels of HSC70 could adaptively complement cuticle morphology in generating resistance to the heat stress induced by sunlight under natural conditions. Indeed, recent studies have shown that the dorsal side of C. bombycina workers (but not of C. mauritanica workers) is covered by a unique array of prism-shaped hairs that reflect incident sunlight and are responsible for the species' silver colour (Fig. 1) (Shi et al., 2015; Willot et al., 2016). The hairs are thermally protective in the visible and near-infrared range of the electromagnetic spectrum, reducing the absorption of sunlight energy and leading to lower internal body temperatures in workers. Sensitive tissues (e.g. central neural tissue) that are protected by such hairs might not need as high a level of constitutive molecular chaperones as ventral organs, which are directly exposed to infrared radiation emanating from the hot ground. Therefore, we cannot exclude the possibility that the high levels of constitutive HSC70 detected in whole-body extracts from C. bombycina stem from the localised accumulation of HSC70 in specific tissues. Moreover, hairs have elevated transmissivity in the mid-infrared range, allowing radiation to bypass the ant's protective coat more readily at those wavelengths (Shi et al., 2015). As a consequence, we may have observed similar mortality rates in the two species because our experimental design did not reproduce the natural balance of incoming solar energy. Instead, homogeneous infrared radiation came from all directions, bypassing C. bombycina's protective coat and hitting critical organs. A tissue-specific comparison of HSC70 constitutive expression in C. bombycina might reveal key differences in heat tolerance strategies.

Finally, in all our experiments the ants were exposed to a 3–5 h heat stress. This corresponds to somewhat artificial conditions, as C. bombycina foragers frequently use thermal refuges in order to dump excessive internal heat in the wild (Wehner et al., 1992). The length of exposure to heat stresses in our experiments was chosen to take into account the full extent of the HSR and compare results with existing data (Gehring and Wehner, 1995). We showed that in both species of Cataglyphis studied, mRNA of up-regulated genes following heat shock abruptly decreased after 1 h respite time (Fig. 5B–E). We are therefore confident that the lessened hsc70 mRNA induction observed in C. bombycina does not result from an experiment in which 4 h at 40°C would allow significant mRNA decay.

In summary, this study demonstrates that the more thermophilic ant species, C. bombycina, has higher constitutive expression of HSC70 and lower induced expression of hsp genes in response to heat stress than the more mesophilic ant species C. mauritanica. Consequently, C. bombycina workers show increased protein stability when exposed to heat stress but are less prone to acquiring induced thermotolerance by means of heat hardening. These results confirm that HSR patterns are highly representative of the species’ thermal environment. More generally, our findings provide insights into the physiological plasticity needed to cope with fluctuating or novel temperature regimes. For example, global warming might prove to be an even greater challenge for species with less-plastic HSP systems than for species that are better equipped to cope with fluctuating body temperatures.

We thank E. Goormaghtigh and G. Vandenbusshe for their advice concerning the protein aggregation experiments and the use of their laboratory. We are grateful to R. Soin and A. Nazih for their help with the RTq-PCR analyses and the western blots. Thanks to J. Pearce for her language editing services.