Neural dysfunction correlates with heat coma and CT_max in Drosophila but does not set the boundaries for heat stress survival

Thermal tolerance is one of the most important traits defining the biogeographical distribution of ectothermic species (Addo-Bediako et al., 2000; Sunday et al., 2014). This is also the case for insects (Gaston and Chown, 1999; Vorhees et al., 2013), including Drosophila, where tolerance to both low and high temperature shows a high correlation to the current species distributions (Andersen et al., 2015; Jørgensen et al., 2019; Kellermann et al., 2012; Kimura, 2004). In the case of insect cold tolerance, there is a general understanding of the processes causing cold coma and cold mortality (Andersen et al., 2018; Bayley et al., 2018; Koštál et al., 2004; MacMillan and Sinclair, 2011), and many physiological adaptations that underlie differences in cold tolerance between species and populations have been uncovered (Feder and Hofmann, 1999; Overgaard and MacMillan, 2017; Sinclair et al., 2003; Yi and Lee, 2004; Zachariassen, 1985). In contrast, it is generally less clear which physiological perturbations cause heat coma and heat mortality, and accordingly there is a poorer understanding of the adaptations that result in intraspecific and interspecific variations in insect heat tolerance (but see Bowler, 2018; Neven, 2000).

Heat tolerance of insects and other ectotherms is typically measured by recording the onset of characteristic behaviours (or endpoints) during heat exposure. These measures include the loss of equilibrium or righting response, onset of spasms, entry into a comatose state or heat mortality (Cowles and Bogert, 1944; Lutterschmidt and Hutchison, 1997a,b; Terblanche et al., 2011). The term ‘CT_max’ (critical thermal maximum) is frequently and indiscriminately used for all of these endpoints, although the different behavioural traits represent the responses to different intensities or durations of heat stress. Thus, mortality is most often preceded by a progressive loss of motor control (Friedlander et al., 1976; Gladwell et al., 1976; Lutterschmidt and Hutchison, 1997a) and some of the endpoints, such as heat coma, can be reversed if the animal is removed from the heat stress immediately after the endpoint is observed (Fraenkel, 1960; Hamby, 1975; Heath et al., 1971; Martinet et al., 2015; Rodgers et al., 2010; but see O'Sullivan et al., 2017). It can be difficult to discriminate between heat coma and heat death (Larsen, 1943; Mellanby, 1954), as the rate of heat injury accumulation responds strongly to small changes in temperature. Accordingly, slightly longer exposures to high temperatures than those causing coma can result in the accumulation of lethal amounts of heat injury (Bigelow, 1921; Jørgensen et al., 2019; Kingsolver and Umbanhowar, 2018).

There are a number of physiological dysfunctions that have been suggested to cause heat coma in arthropods. These include a mismatch between demand and supply of oxygen to active tissues (described in the hypothesis of oxygen and capacity limited thermal tolerance, OCLTT) (Pörtner, 2001), haemolymph hyperkalaemia, which would impair muscle function (Gladwell, 1976; Gladwell et al., 1976; O'Sullivan et al., 2017), cellular heat injury to the membranes (Bowler, 1981, 2018; Bowler et al., 1973; Hazel, 1995) and breakdown of central nervous function (Hamby, 1975; Larsen, 1943; Prosser and Nelson, 1981; Robertson, 2004). The evidence to support acute heat failure due to oxygen limitations is not strong for terrestrial insects (Klok, 2004; Mölich et al., 2012; Verberk et al., 2015) and there is also limited support for haemolymph hyperkalaemia as the proximal cause of heat coma, as hyperkalaemia was suggested to occur downstream of cellular heat injury in locusts (O'Sullivan et al., 2017). This leaves membrane dysfunction and breakdown of nervous function as the strongest candidate mechanisms underlying heat coma. Silencing of nervous function has been observed in heat-exposed fruit flies and locusts where heat stress causes a spreading depolarisation (SD) in the central nervous system (CNS) (Money et al., 2009; Robertson, 2004; Rodgers et al., 2007). SD is triggered by a failure to maintain ion gradients between the intracellular and extracellular compartments within the CNS, which results in depolarisation of neurons and glial cells and a surge of potassium ions into the extracellular space of the brain, preventing neural activity (Robertson, 2004; Robertson et al., 2020; Spong et al., 2016). Furthermore, studies have shown that interspecific and intraspecific differences in cold coma are highly correlated with the loss of CNS function in insects (Andersen et al., 2018; Robertson et al., 2017). Given the similarity in the behavioural traits of heat and cold coma, it is possible that the onset of heat coma is also caused by CNS failure in insects.

In most insects, heat mortality follows closely after the onset of heat coma (Mellanby, 1954) and the hypothesis about hyperthermic loss of CNS function could therefore also be extended to be the proximal cause of heat mortality. In goldfish, heating either the cerebellum or the water caused similar behavioural responses, which progressed from hyperactivity to coma (Friedlander et al., 1976). A recent study revisited the work of Friedlander et al. (1976); the authors selectively cooled the brain of Atlantic cod while the fish were subjected to heat stress, and found that this resulted in slightly increased heat tolerance (measured as loss of equilibrium) compared with that of controls and instrumented controls (Jutfelt et al., 2019). Accordingly, it appears that controlling the temperature of the CNS can mimic whole-animal exposure to a specific temperature.

In the present study, we used a comparative study system of five Drosophila species with pronounced interspecific differences in heat tolerance. The most heat-sensitive species enters heat coma at a temperature 6°C lower than the most tolerant species (in a ramping assay). Similarly, the constant temperature estimated to cause onset of coma after a 1 h exposure is almost 6°C lower in the sensitive species than in the most heat-tolerant species used here (Jørgensen et al., 2019). To investigate the relationship between neural dysfunction and the two behavioural heat stress traits, loss of coordinated movement (T_back/t_back) and onset of heat coma (T_coma/t_coma), we measured DC potentials in the CNS of the five species during heat exposure to record SD as an indication of neuronal failure. These experiments were performed with both gradual heating (a dynamic ramping assay) and a constant (static) heat exposure to examine whether the way heat stress is inflicted affects the relationship between neuronal perturbation and behavioural stress traits. The loss of coordinated movement, onset of heat coma and heat mortality occur in rapid succession in many insects. To examine whether the onset of heat mortality is caused proximately by failure in the CNS, we designed a simple experiment in which we compared the heat sensitivity of flies that were heated over their entire body with that of specimens heated specifically in the head (CNS) or abdomen (visceral tissues). This experiment was performed in three of the Drosophila species and was designed to evaluate whether some body sections (head with primarily neuronal tissue versus abdomen with primarily visceral tissue) were more sensitive to heat stress than others.

Five species of Drosophila (D. immigrans Sturtevant 1921; D. subobscura Collin 1936; D. mercatorum Patterson and Wheeler 1942; D. melanogaster Meigen 1830; and D. mojavensis Patterson 1940) were used in this study (for details on origin, see table 1 in Jørgensen et al., 2019). The least heat-tolerant species, D. immigrans, can survive at 35.4°C for 1 h, while the most heat-tolerant species, D. mojavensis, can survive at 41.2°C for 1 h (Jørgensen et al., 2019) and collectively these five species represent a broad range of heat tolerances within Drosophila. The small size of Drosophila was a challenge to some of the measurements performed in the study, but the comparative system with Drosophila was chosen because of the ease of producing sufficient numbers of similar aged and reared animals for experiments. Flies were reared and maintained under common garden conditions in 250 ml bottles containing 70 ml of oat-based Leeds medium (see Andersen et al., 2015) in a 19°C room with constant light. Maintenance bottles with adults that parented the experimental flies were changed twice a week, and newly eclosed adults from rearing bottles were collected and transferred to fresh vials with fly medium every 1–3 days. Experimental flies were produced by transferring a tablespoon of medium (including eggs) to another 250 ml bottle with 70 ml new medium; 2–4 days post-eclosion, flies were briefly anaesthetised with CO₂, sexed and female flies were moved to new food vials, and allowed to recover from the CO₂ anaesthesia for at least 2 days before measurements began (MacMillan et al., 2017). All experiments were performed on 4–9 day old non-virgin female flies. Females were used instead of males, because of their larger size.

Behavioural heat tolerance was characterised with a ramping and a static assay using the same setup as described in Jørgensen et al. (2019). In this setup, the fly was subjected to homogeneous heat exposure within a glass vial that was submerged in a water tank at a controlled temperature (Fig. 1A). In the ramping assay, temperature was increased by 0.25°C min⁻¹ from 30°C. Two behavioural traits were recorded during this experiment: (1) the temperature (T) at which the fly would lose coordination and fall on its back (T_back) and (2) the temperature at which the fly was completely still (T_coma). T_coma was verified by poking the vial lids with a stick to agitate the flies and check for reflexes. The static assay used a similar setup and method to record knockdown, but instead of the temperature being increased gradually, the bath was maintained at 38°C and the exposure durations (time t) causing loss of coordinated movement (t_back) and heat coma (t_coma) were noted. The ‘static’ assay was only static for 1 h at 38°C, after which the temperature was increased by 0.25°C min⁻¹ to ensure that more heat-tolerant flies would also succumb to heat stress. n=7 flies were measured for each species in each assay, except D. subobscura in the ramping assay (n=6).

Electrophysiological measurements of DC potentials in the CNS (a proxy for nervous function) were carried out as described by Andersen et al. (2018). Filamented borosilicate glass capillaries (1 mm diameter; 1B100-F-4, World Precision Instruments, Sarasota, FL, USA) were pulled to low tip resistance (5–7 MΩ) using a Flaming-Brown PC-84 micro-pipette puller (Sutter Instruments, Novato, CA, USA) and back-filled with 500 mmol l⁻¹ KCl solution. The glass electrodes were connected to a Duo 773 intracellular differential amplifier (World Precision Instruments) using the low impedance channel and probe, and a chlorinated Ag/AgCl wire was used as the reference electrode to ground the preparation. An MP100 data-acquisition system was used to digitalise the voltage output, which was recorded using AcqKnowledge software (Biopac Systems, Inc., Goleta, CA, USA).

A fly was prepared for measurement by gently fastening its ventral side to a bed of wax on a glass cover slide. Using a fine pair of scissors, a small hole was cut in the abdomen between the second- and third-to-last tergites for placement of the reference electrode. Another cut was made along the head midline just posterior to the ocelli to insert the glass recording electrode. The cover slide with the fly was placed onto a Peltier plate pre-set to 30°C, which could be thermoelectrically heated (PE120, Linkam Scientific Instruments, Tadworth, UK), and temperature was monitored continuously using a type K thermocouple (integrated with the MP100 data-acquisition system) placed on top of the wax, adjacent to the head of the fly (Fig. 1B). This heating method was expected to heat the ventral side of the fly homogeneously, but also result in a small temperature gradient from the ventral to the dorsal side. The manufacturer's note on the type K thermocouples indicates a measurement error of ±2°C across the whole range of use, but to minimise measurement error between thermocouples, all thermocouples used in the present study were calibrated to a single reference thermometer in ice slurry, at room temperature and at several temperatures within the temperature range used in the study prior to measurements. The glass electrode and the reference (Ag/AgCl) electrode were placed blindly in their designated holes using micromanipulators, aiming to always use the same angle and depth of tip insertion, and the voltage was zeroed. To test the quality of the preparation, a flow of humidified N₂ was passed over the fly to elicit an anoxic SD. The single depolarisation triggered by anoxia persists throughout the exposure to N₂, but has been found to be completely reversible in Drosophila (Armstrong et al., 2011; Rodríguez and Robertson, 2012) and locusts (Rodgers et al., 2007), and additionally we did not find any difference in the timing of SD in heating experiments with and without prior anoxia treatment. We therefore used this anoxia test to discard preparations that failed to depolarise (suggesting that there was a problem with the electrode placement). This test also gave an indication of the size of depolarisation that could be expected from that particular preparation as this is also dependent on the quality of impalement and location of the recording electrode. If the preparation depolarised ≥20 mV in response to anoxia, the voltage was zeroed again, and the preparation was used for ramping, static or control experiments.

In ramping experiments, the temperature of the thermal stage was increased from 30°C by 0.25°C min⁻¹ and the temperature (at the half-amplitude of the negative DC shift associated with SD) of the first and last SD event (SD_first and SD_last, respectively) along with the number of SD events was recorded. The ramping continued until it was clear that no more depolarisations would occur, which was concluded when the preparation could no longer maintain a stable baseline DC potential (see example traces in Fig. 2). In static heat exposure experiments, the temperature was rapidly increased from 30°C to 38°C (mean heating time: 73 s, ∼6.6°C min⁻¹), and the timing of SD_first and SD_last and the number of depolarisation events were noted as above. The stage was kept at 38°C until no more depolarisations were anticipated (same criterion as in ramping experiments). In preparations for which no depolarisations had occurred during the 1 h exposure (only in D. melanogaster and D. mojavensis), the stage temperature was increased by 0.25°C min⁻¹ after the first hour at 38°C and this heating was continued until depolarisations were measured. Some of the preparations elicited only a single SD event, and accordingly the temperature/time reported was the same for SD_last as for SD_first (see Fig. 2C).

A number of control experiments were conducted to test whether the starting condition at 30°C or the handling of the fly was stressful enough to elicit SDs by keeping a few D. immigrans (the least heat-tolerant species) and D. mojavensis (the most heat-tolerant species) at 30°C for 1 h, but these conditions failed to elicit SDs in either species. The control experiments were concluded by increasing temperature by 1°C min⁻¹ until SD events were observed, leading us to conclude that the preparations were responsive but that the handling and starting conditions (30°C) alone were unable to evoke this response.

To further examine the role of nervous function in heat tolerance, we performed a series of experiments in which we selectively heated the head or the abdomen of flies and compared their survival after 24 h with that of flies that had been heated more uniformly (see Fig. 1C–E). The motivation for this study was to examine whether the head (dominated by nervous tissue) was more heat sensitive than the abdomen (dominated by fat body and intestinal tissue). Only three species (D. subobscura, D. melanogaster and D. mojavensis) were used for these experiments as they represent low, medium and high heat tolerance, respectively. Drosophila subobscura was chosen to represent low heat tolerance rather than D. immigrans, because of its smaller size, which made it more appropriate for the fixation method.

For these experiments, the flies needed to be restrained in a way that allowed one end of the fly to be held closer to the heating stage, and as survival was used as the measure of sensitivity, the fixation method should also allow for the flies to be moved from the heating stage without inflicting injury to the animals. Accordingly, flies were fastened in 200 µl pipette tips, using a device originally designed for haemolymph extraction (MacMillan and Hughson, 2014). With a stream of air, the fly was manipulated headfirst into the pipette tip, and the airflow was blocked once the fly was stuck in the tip (taking care not to injure it). The pipette tip was removed from the device and the tip was cut off just anterior to the head followed by two cuts (one from the dorsal and one from the ventral view of the fly) that were made in roughly a 45 deg angle towards the anterior part of tip (Fig. 1C–E). These angled cuts allowed better contact between the head and the heating stage on the ventral side and room for the thermocouple to measure head temperature on the dorsal side. Using a scalpel, some of the plastic covering the abdomen was gently ‘shaved’ off, while making sure that no holes were made. The tip was then reattached to the air pressure device and the fly was ‘pushed’ until the head protruded from the tip. The area that had been thinned before was now cut away, leaving the abdomen exposed, thereby decreasing the distance to the heating stage on the ventral side (Fig. 1C–E). Another cut was made in the dorsal side of the tip, allowing placement of a micro-thermocouple directly on the dorsal side of the abdomen (here it was often necessary to move the wings to the side; Fig. 1C–E). Flies that were injured (other than severed wings) were discarded. The preparations were used for whole-body heating, selective heating of the head, selective heating of the abdomen or as un-heated controls. Flies were generally heated on the ventral side, but we also tested some flies exposed to whole-body heating from the dorsal side (see Fig. S1).

For ventral whole-body heating, the pipette tip was placed on the Peltier plate (PE120, Linkam Scientific Instruments) with the wide end of the tip at a slightly positive angle, to facilitate closer contact between the heating stage and the ventral side of the head and abdomen (Fig. 1C). When the tip was staged, two micro K type Fine thermocouples (tip diameter 25 µm, KFG-25-100-100, ANBE, Genk, Belgium) were placed on the surface of the head and the abdomen, respectively (Fig. 1C). This method gave a relatively homogeneous heating of the fly when measured on the dorsal side, with a tendency for slightly higher temperatures measured on the head (possibly as a result of closer contact with the Peltier plate). For every sample, the tip was turned 180 deg horizontally, such that the head and abdomen switched location on the heating stage, to minimise any differences in heating across the stage. The transversal temperature gradient that arose from ventral heating was measured in D. mojavensis by gradually moving the thermocouples through the head and abdomen from the dorsal towards the ventral side, in flies that had been killed before the experiment. This transverse difference was recorded at 2.51±0.22°C and did not differ between the head and abdomen (paired t-test, t=0.49, d.f.=5, P=0.64). Similar measurements were made for a few D. melanogaster and D. subobscura, with similar results.

To test heat tolerance, the temperature of the heating stage was quickly increased to the desired test temperature (∼1.5 min), and once the temperature was stable, the fly was left at this condition for 15 min. After heating, temperature rapidly dropped to room temperature (∼1 min) when the thermal stage was turned off. The tip was then removed from the Peltier plate, and the fly was immediately checked for movement. After 15 min, the fly was again checked for movement, released by cutting the tip and then transferred to a 2 ml Eppendorf tube with fly medium in the bottom and air holes in the lid. Flies were checked for movement after 1 day of recovery following the heat exposure (recovery at 19°C), and their status (live/dead) here was used for further analysis. Flies were regarded as ‘dead’ if they were unable to move after the 24 h recovery period.

Selective heating of either the head (Fig. 1D) or abdomen (Fig. 1E) was performed using the same preparation as above, but with the body part to be heated placed on the heating stage while the rest of the body was placed away from the stage. This heating method resulted in large temperature differences between body parts, with heating of the head giving a larger difference than heating of the abdomen (Table 1).

Control experiments were performed to test whether the manipulation of the flies resulted in any mortality. In these experiments, the flies were prepared similarly to flies used for heating, but instead of heat exposure they were kept at room temperature and assessed for survival following the same protocol.

All data analyses were performed in R version 3.5.2 (http://www.R-project.org/). Unless otherwise stated, all results are reported as means±s.e.m., and the critical value for statistical significance was 0.05. As SD_first necessarily must precede (or coincide with) SD_last, and the behavioural trait T_back/t_back was always observed prior to T_coma/t_coma, the measurements of SD_first×T_back/t_back and SD_last×T_coma/t_coma were tested for co-occurrence using a t-test for each species for each assay type (ramp and static). For each combination (measurements compared and assay type), this resulted in five P-values that were then corrected using the Holm–Bonferroni post hoc adjustment with the function p.adjust() in the inbuilt stats-package (http://www.R-project.org/). The correlation between behavioural heat stress traits and onset of SD events was examined between species within assay type using linear regressions with the lm()-function in R. The regression lines were compared with the line of unity (intercept=0, slope=1) with the function linearHypothesis in the Car-package (Fox and Weisberg, 2011).

Neural function during ramping and static heat exposure was examined by measuring negative DC shifts associated with SD in the CNS in the fly head. The temperature (ramp) or time (static) of the first or last SD (SD_first and SD_last, respectively) was then compared with the temperature or timing of two behavioural heat stress traits measured using similar heating protocols: the loss of coordinated movement (T_back/t_back) and onset of heat coma (T_coma/t_coma) (Fig. 2). These experiments were used to examine (1) whether behavioural heat stress traits coincide with signs of neural dysfunction, and (2) whether this putative correlation is affected by the way heat stress is inflicted.

When flies were exposed to gradually increasing temperature in a ramp protocol, there were clear interspecific differences in the temperatures where the behavioural heat stress traits were observed. For example, the least heat-tolerant species (D. immigrans) showed loss of coordination (T_back) at 35.22±0.45°C and went into heat coma (T_coma) at 38.69±0.25°C, while the most heat-tolerant species (D. mojavensis) reached T_back at 43.01±0.24°C and T_coma at 45.11±0.34°C, giving the species system a range of T_back of 7.8°C and T_coma of 6.4°C. Similarly, the temperatures at which SD events were observed gave interspecific differences of 7.4°C for SD_first and 6.5°C for SD_last between the least and most tolerant species (again, D. immigrans and D. mojavensis). Generally, we found that the temperature of T_back and T_coma coincided with perturbation of nervous function as indicated by SD_first and SD_last, respectively (Fig. 3A). Comparison of the temperatures where SD_first and T_back were observed revealed that the two measures could only be separated statistically in D. mercatorum, as SD_first occurred 2.4°C higher than T_back (P=0.022, t-test with Holm–Bonferroni post hoc correction). For the comparison of SD_last and T_coma, the only significant difference was found in D. subobscura, where SD_last preceded T_coma by 2°C (P=0.030, t-test with Holm-Bonferroni post hoc correction). However, we caution that the means of heating differed between the behavioural trait experiments and the neurological experiments, and that this could be a source of experimental noise (see Materials and Methods and Discussion for further arguments). To test whether there was a general co-occurrence of behavioural traits and neurological events, we performed linear regressions on the mean temperatures of SD_first against T_back and SD_last against T_coma (Fig. 3B,C). Both regressions were significant (P=0.043 and 0.008), yielded high coefficients of determination (R²=0.79 and 0.93), and neither of them was significantly different from the line of unity (P=0.699 and 0.089). The regression analysis indicated that across species there were generally only small differences between the temperatures at which behavioural and neurological collapse were observed.

During constant heat exposure (38°C; Fig. 4), we recorded the timing of SD events and behavioural heat stress traits and again we found these behavioural and neurological measures to coincide. Note that for some species we started to increase the temperature by 0.25°C min⁻¹ after 1 h of exposure, but that all measures are reported in minutes of exposure. There were clear interspecific differences in the duration that the nervous system could uphold function under heat exposure, and this duration increased with species’ heat tolerance (as indicated by observations of behavioural heat stress traits). However, the least tolerant species in terms of neuronal failure in the static assay (D. subobscura) was the second least tolerant when assessed for behavioural traits (D. immigrans was the least tolerant according to this term, as in the ramping assay) (Fig. 4A). It was not possible to separate the timing of SD_last from t_coma in any of the species, and in the comparison of SD_first and t_back, only D. melanogaster showed a significant difference between the timing of behavioural and neurological traits, with a delayed coma onset for D. melanogaster relative to SD_last (P=0.004, t-test with Holm–Bonferroni post hoc correction) (Fig. 4A). A linear regression on the mean time of SD_first against t_back for each species showed a high, significant correlation (R²=0.89, P=0.016; Fig. 4B), while the correlation between SD_last and t_coma was slightly weaker (R²=0.74; Fig. 4C) and was not significantly different from 0 (P=0.062). When the regression lines were compared with the line of unity, none of them were significantly different, again suggesting that across the species system there was generally an overlap between the exposure durations that resulted in behavioural and neurological traits.

Examination of the DC potential measurements showed considerable variance between preparations. Some preparations were characterised by only eliciting a single SD event (meaning that SD_first and SD_last occurred at the same time/temperature; Fig. 2C) while other preparations showed multiple (2–30) SD events (see examples in Fig. 2). Comparison between the ramping and constant heat exposures showed that single SD events were much more prevalent during the ramping heat exposure (40% of individuals showed single SD, n=35) than in the constant heat exposure (9% showed single SD, n=29) (see Fig. S2). Furthermore, when the constant heat exposure for 1 h was followed by a ramping increase in temperature, flies would mostly elicit just a single SD (66%, n=6). All five species were able to display both single and repeated SD events and in roughly the same proportion [2–4 preparations of each species (out of 7) showed a single SD during ramping]. The number of SD events observed in ‘multiple’ SD events also differed with heat exposure assay. In static assays, preparations with multiple SDs elicited 11.38±1.56 SD events while preparations with multiple SDs during ramping assays only had 5.95±1.12 SD events (two sample t-test, t=2.83, d.f.=43.15, P=0.007).

As heat coma and heat death often occur in close succession, we performed an experiment designed to investigate and compare the heat sensitivity of the head (site of nervous function measurements from the first experiment) and the abdomen (consisting more of visceral tissues) (see Fig. 1C–E). This test involved restraining flies in pipette tips, and non-heated controls for handling showed 0% mortality for D. subobscura and D. melanogaster, and 13% mortality for D. mojavensis after 24 h (n=14/16/39, respectively). For these experiments, the temperature estimated to cause 50% mortality in the flies 24 h after heat exposure (LT₅₀) was used to compare heat sensitivity between body parts.

Both whole-fly and selective heating showed that the heat-tolerant D. mojavensis had higher values of LT₅₀ than the moderately heat-tolerant D. melanogaster, which in turn also had higher values of LT₅₀ than the heat-sensitive D. subobscura (Fig. 5). When the whole fly was heated simultaneously, we did record differences between head and abdominal temperature (measured topically on the dorsal side), but these differences were generally less than 2°C (see Table 1 and Fig. S1). In experiments using selective heating of either the head or abdomen, the flies were characterised by considerably larger regional differences in temperature (ΔT ranging from 3.35 to 10.19°C depending on the species and body part heated; see Table 1).

The experiments revealed species-specific differences in the relationship between LT₅₀ estimates during whole-animal heating and selective heating. For D. mojavensis, heating the abdomen (and maintaining the head at a lower temperature, ΔT=4.79±0.29°C) did not change the LT₅₀ compared with the abdominal LT₅₀ when the whole fly was heated (LT₅₀ was 0.35°C higher with selective heating of the abdomen but the estimates have overlapping 95% confidence intervals; Fig. 5A). Thus, for D. mojavensis, LT₅₀ was the same irrespective of whether the head was kept cool or warm during heating of the abdomen. When the head of D. mojavensis was heated selectively (with the abdomen considerably cooler: ΔT=10.19±0.36°C), LT₅₀ increased by 2.33°C compared with that of flies experiencing whole-animal heating (non-overlapping 95% confidence interval; Fig. 5B). Thus, a higher head temperature was needed to evoke mortality in D. mojavensis when the abdomen was relieved of heat stress.

Performing the experiments on D. melanogaster, we observed slightly smaller differences between body parts than in D. mojavensis, both when the head was selectively heated (abdomen maintained at a lower temperature, ΔT=9.16±0.41°C) and when the abdomen was heated (head kept cooler, ΔT=4.6±0.22°C). For D. melanogaster, we found LT₅₀ to increase when applying selective heating on the abdomen (LT₅₀ was 2.59°C higher; Fig. 5A) and the head (LT₅₀ was 3.77°C higher; Fig. 5B), compared with LT₅₀ resulting from whole-fly heating. Accordingly, maintaining one end of a D. melanogaster at a lower temperature than the other increases heat tolerance of the fly.

In experiments with D. subobscura, the temperature differences between body parts were smaller than for the other two species. Selectively heating the abdomen made the abdomen 3.35±0.28°C warmer than the head but did not change the LT₅₀ of the abdomen when compared with that of whole-fly heating (LT₅₀ was 0.13°C lower for the selective heating, probably attributed to the shape of the survival curve, but with overlapping 95% confidence intervals). When selectively heating the head, resulting in a 6.44±0.28°C colder abdomen, head LT₅₀ increased by 1.87°C compared with that of whole-animal heated flies. Thus, a higher head temperature was also needed to evoke mortality in D. subobscura when the abdomen was relieved of heat stress.

Interspecific and intraspecific differences in heat tolerance have been demonstrated for Drosophila in multiple studies (Castañeda et al., 2015; Jørgensen et al., 2019; Kellermann et al., 2012; Kimura, 2004; Overgaard et al., 2014; Stratman and Markow, 1998). These differences have often been measured using the onset of reversible behavioural traits such as loss of coordinated movement and entry into heat coma, or by measuring heat-induced mortality in animals exposed to high temperatures (Lutterschmidt and Hutchison, 1997a). However, it is still unclear which physiological perturbations are the proximate causes of the different heat tolerance endpoints (but see Robertson, 2004; Rodgers et al., 2010), and this has been particularly difficult to discern because of the close proximity of the endpoints at high temperatures. Multiple physiological mechanisms have been suggested as the proximate cause of heat mortality, including oxygen transport limitations, protein denaturation, loss of membrane integrity or ion homeostasis, and mitochondrial dysfunction (Bowler, 2018; Davison and Bowler, 1971; Gladwell, 1976; Pörtner, 2001; Somero, 1995). The endpoint prior to mortality, the onset of heat coma, has instead been suggested to be caused by either muscular or nervous failure (Bowler, 1963; Gladwell et al., 1976; Robertson, 2004), which could also be linked to membrane damage. In locusts exposed to increasing temperature, ventilation failed concurrently with an abrupt surge in extracellular [K⁺] within the CNS and this disturbance has been linked to a drop in DC potential, which is a reliable marker of SD in the CNS (Robertson, 2004; Rodgers et al., 2007). Once the locust was returned to benign temperatures, extracellular [K⁺] surrounding the neurons returned to baseline levels, and the motor pattern ventilation resumed (Rodgers et al., 2007, 2010).

To our knowledge there have been no comprehensive comparative studies investigating species differences in CNS function at high temperature and the aim of this study was to examine the role of the nervous system in relation to heat tolerance in five Drosophila species. The temperatures at which two behavioural traits [loss of motor control (T_back) and loss of motor function (T_coma)] were observed were compared with the temperature of neuronal failure (SD) as assessed by electrophysiological measurements of DC potentials in the fly brain during ramping heat exposure, and likewise the timing of SD and behavioural traits (t_back and t_coma) during constant heat exposure. These experiments generally revealed a good correlation between the failure of motor control/function and neuronal failure; however, it is unclear whether failure of the CNS also causes heat mortality. Thus, we designed an experiment to test the sensitivity to heat exposure on different parts of the fly body to further examine whether the nervous system could be limiting heat stress survival.

Measurements of SD (i.e. large negative shifts in DC potential) during both ramping and static assays showed that, overall, perturbation of nervous function correlated well with the two behavioural heat stress traits (T_back/t_back and T_coma/t_coma) (Figs 3, 4). Onset times and temperatures of heat coma were similar to the values previously reported in the five species, measured in similar heat tolerance assays (Jørgensen et al., 2019). The rate of heat injury accelerates extremely quickly at high temperature in these species (Q₁₀ of heat injury accumulation rate is often >10,000; Jørgensen et al., 2019) and even small differences in exposure temperature (or time) can separate tolerance and death during heat exposure. Unfortunately, it was not possible to perform measurements of both SD and behavioural traits on the same individual and the differences in experimental protocols between behavioural and neurological experiments are likely to introduce some noise in the comparison between these experiments. For example, the loss of motor function (T_back/t_back and T_coma/t_coma) was assessed in untethered flies in glass vials with a homogeneous temperature, whereas SD measurements required the flies to be fastened and furthermore a hole was cut in the head and abdomen to insert measurement electrodes (Fig. 1). It is possible that the more invasive preparation required for SD measurements could affect heat tolerance, and we also observed a considerable internal thermal gradient in the fly (sometimes more than 2°C) when using the Peltier plate for heating. Considering these sources of variation, it would be unexpected to find a perfect correlation between the two types of experiment, which is also why we used two types of heating (ramp and static) and multiple species, to allow for more ‘axes’ of comparison between the neurological and behavioural traits. Despite these experimental challenges, we found patterns of association between loss of motor control and the occurrence of SD events in the CNS (Figs 3 and 4).

Generally, the characteristics of behavioural heat stress traits follow a progressive loss of motor control, from first hyperactivity, through loss of coordinated movement and spasms to the onset of heat coma or heat stupor, where the animal is unresponsive (Cossins and Bowler, 1987; Heath and Wilkin, 1970; Lutterschmidt and Hutchison, 1997a). For our experiments, it follows that the two behavioural traits T_back/t_back and T_coma/t_coma are bound in a way such that T_back/t_back will occur at a lower temperature than (or prior to) T_coma/t_coma. Similarly, the first SD must precede the last SD, unless only a single SD event is observed (in which case the first and last SD are the same). It is therefore tempting to conclude that SD_first is linked to T_back/t_back and likewise SD_last is linked to T_coma/t_coma but because of the noise in the data we merely conclude that it is likely that the two closely occurring behavioural traits (T_back/t_back and T_coma/t_coma) are linked to the simultaneously occurring SD events (SD_first and SD_last, respectively) (Figs 3, 4). For example, we cannot rule out that other cell types (e.g. muscle) fail concurrently with the nervous system and therefore also influence the behavioural traits. Further research is still required to establish a direct causal relationship between neurological function and behavioural heat stress traits. That said, we found that there is a close relationship between the SD events and behavioural traits across species with variable thermal tolerance and this was seen in both dynamic and static tests. Further, it has previously been shown that heat hardening increases the temperature for onset of SD events and loss of motor function (flight) proportionally in locusts (Robertson et al., 1996). The relationship between behavioural traits and nervous dysfunction has also been examined at low temperatures in different species of Drosophila, where the temperature of cold coma onset is also highly correlated with the temperature of SD in the CNS of Drosophila (Andersen and Overgaard, 2019; Andersen et al., 2018). However, similar to our heat experiments, it is difficult to determine specifically how SD_first and SD_last events are linked to loss of motor control (T_back) or loss of movement (T_coma). Importantly, there is no association between cold-induced SD events and cold mortality as insects can survive cold in a ‘comatose’ state for long periods of time (MacMillan and Sinclair, 2011; Overgaard and MacMillan, 2017).

The present study found that single SD events (instead of multiple events) were more prevalent in ramping experiments than during static heat exposure (Fig. S2). Additionally, if multiple SD events occurred in the preparation, the total number of SD events was significantly lower in flies exposed to the ramping heat assay than to the static assay. In hyperthermic locusts, single continuous SD events that persist until the heat exposure is removed are the most prevalent, but repetitive SD events have been observed in locusts treated with ouabain (Rodgers et al., 2009; Spong et al., 2014) and in hyperthermic brain slices from immature rats (Wu and Fisher, 2000). Contrary to hyperthermia, which is thought to lead to accumulation of potassium in the extracellular space of the nervous system, ouabain limits K⁺ clearance through its inhibition of Na⁺/K⁺-ATPase (Rodgers et al., 2009). According to Rodgers et al. (2009), the repetitive SD events are caused by transient surges in extracellular [K⁺] in the interstitial space that result from imbalances between accumulation and clearance of potassium. A speculative explanation for the increased prevalence of single SD events in ramps could be that when temperature is gradually increased, the mitigation of the physiological conditions resulting in SDs (high extracellular [K⁺] in the space surrounding the CNS) cannot keep up as heat stress increases exponentially (Jørgensen et al., 2019), resulting in a total silencing of the CNS. Conversely, static exposure may allow the fly to remove some of the potassium that has accumulated in the extracellular space. This could relieve the condition causing the SD event and temporarily restore some nervous function until a new SD event occurs, when K⁺ clearance is surpassed by accumulation (Rodgers et al., 2010).

The representative DC potential traces shown in Fig. 2 all advance toward (more) negative values throughout the heat exposure. In measurements of a conventional membrane potential, such a decrease would be interpreted as a more polarised membrane, which opposes the depolarisation towards 0 mV that would be expected in a failing nervous system with damaged membranes and collapsed ion gradients. Here, it is important to emphasise that the DC potential measured in this study is actually made up of two membrane potentials, as the measuring electrode penetrates both the basolateral and adglial membranes of the perineural cells of the blood–brain barrier (Robertson et al., 2020). If the adglial membrane facing the nervous system depolarises (as a result of the increased interstitial [K⁺] during an SD event), but the basolateral membrane (facing the haemolymph) remains negatively polarised, the net DC potential across the two membranes will hyperpolarise, which is seen as a negative DC shift (Robertson et al., 2020; Schofield and Treherne, 1984). Thus, the observation that the DC potential does not approach zero suggests that the adglial membrane maintains its integrity as the ion gradient disruption is kept within the interstitium of the nervous system, and accordingly that the functionality of the blood–brain barrier is intact.

To investigate the role of CNS failure in heat mortality, we designed an experiment to estimate heat sensitivity of the head and the abdomen when either the whole fly was heated or one body part was selectively exposed to a higher temperature than the rest of the fly. If CNS failure at high temperatures is the main cause of heat mortality, then we would expect that maintaining the head at a lower temperature than the abdomen should also lower mortality. Conversely, if the head was heated selectively, we would expect mortality to occur at the same temperature as when the whole fly was heated. Manipulations of body compartment temperatures have previously been used successfully in crayfish (Bowler, 1963), goldfish (Friedlander et al., 1976) and Atlantic cod (Jutfelt et al., 2019) to investigate the heat sensitivity of either heat coma or heat mortality. To our knowledge, this is the first attempt to study this question in small insects such as Drosophila.

Using the experimental setup with a fly tethered in a pipette tip, we found clear differences in heat tolerance (measured as LT₅₀) between species, such that the desert species D. mojavensis was more heat tolerant than the cosmopolitan D. melanogaster, which in turn was more heat tolerant than the temperate D. subobscura. This finding is entirely consistent with the other behavioural heat stress traits measured in the present study and with findings from previous studies (Jørgensen et al., 2019; Kellermann et al., 2012). The tethering of the flies was not in itself invasive as attested by no mortality of D. subobscura and D. melanogaster controls, and low mortality of D. mojavensis controls. Experiments with selective heating of the abdomen and head suggest interspecific differences in body part sensitivity (Fig. 5). All three species showed increased heat tolerance of the head when the abdomen was simultaneously kept at a lower temperature (i.e. heating only the head; Fig. 1D). This suggest that the head may not be the most heat-sensitive body part (Fig. 5B). When the head was maintained at a lower temperature (abdomen was heated; Fig. 1E), the species differed in response (Fig. 5A). Drosophila subobscura and D. mojavensis maintained a similar LT₅₀ for the abdomen when only the abdomen was heated compared with heating of the whole animal, suggesting that the abdomen is a heat-sensitive body part in these two species. Drosophila melanogaster showed a different response as LT₅₀ increased in flies when only the abdomen was heated (i.e. a similar response to when the head was selectively heated). This suggests that for D. melanogaster, both body parts are injured through heat exposure and that the damage may be additive such that it is the total amount of accumulated injury that determines heat tolerance. Overall, these experiments showed that the head was not a particularly heat-sensitive region and the higher LT₅₀ values in flies with selective heating of the head suggest that neuronal tissue can survive some degrees beyond the temperature causing SD events.

The increase in LT₅₀ for flies with selective heating of the head supports the notion that SD may be an adaptive mechanism to protect the organism during stress (Robertson, 2004; Robertson et al., 2020; Rodgers et al., 2010). For example, Robertson et al. (2020) speculated that the reversible silencing of the nervous system is beneficial to maintain function in other physiological systems when benign conditions return, and in hyperthermic locust, early onset of SD in the CNS was associated with shorter recovery time of CNS function when the preparation was returned to benign conditions (Rodgers et al., 2007). We observed in multiple cases that flies used for the LT₅₀ experiments would enter a heat coma (they were completely unresponsive immediately following heat exposure), but they would later resume movement and often recover normal behaviour. Likewise, we observed in the initial behavioural trait assays that flies removed from the heat immediately after T_coma/t_coma had been observed would recover subsequently. Together, these data indicate that SD events are not directly associated with mortality and that nervous failure is not a proximal cause of heat death. Nevertheless, thermal sensitivity of the nervous system could impose a critical challenge to fitness if critical behaviours, such as escape responses, are impaired at stressful temperatures (Montgomery and Macdonald, 1990).

In conclusion, this study shows clear interspecific differences in the extent (time/temperature) that the flies can tolerate heat stress. We found that loss of nervous function is likely to be the main cause of the characteristic loss of coordinated movement and coma onset that are classically used to assess heat tolerance in insects (CT_max). The temperature and time span from when the most heat-sensitive species suffered from neural failure to when the CNS of the most heat-tolerant species succumbed was large, inviting further studies to investigate adaptations in the CNS to alter heat sensitivity. Our experimental conditions did not allow for simultaneous measurements of neuronal function and behaviour and we can therefore not conclude whether it is the first or last SD event that is the cause of specific behavioural traits. Likewise, it is possible that neuronal failure in other ganglia or the failure of other cell types (e.g. muscle) could play a role for these behavioural tests. Using selective heating, we also showed that the head (mainly neuronal tissue) is not particularly heat sensitive compared with other parts of the body. Thus, entry into (reversible) coma and heat mortality are likely to be caused by different physiological processes and loss of brain function is not the proximal cause of heat death. However, it is also important to appreciate that even small disturbances in nervous function at less stressful temperatures could mean the difference between life and death to an unrestrained animal in nature if its escape response is retarded by nervous dysfunction.

We would like to thank Kirsten Kromand for animal care, Niels U. Kristiansen for help with preparation of thermocouples and Mads K. Andersen for help with the experimental setup and for valuable discussions of the results and experimental design.