Upper thermal tolerance and oxygen limitation in terrestrial arthropods

Owing to the realization that many species will be incapable of migrating in response to modern climate change, largely as a consequence of habitat fragmentation (Warren et al.,2001), there is increasing interest both in the ways in which thermal tolerance evolves and the rate at which this can take place(Davis and Shaw, 2001; Hoffmann and Blows, 1993). Building on previous work (Cossins and Bowler, 1987; Prosser,1986; Ushakov,1964), Pörtner (Pörtner, 2001, 2002a; Pörtner et al., 2000) has developed a wide-ranging hypothesis suggesting that in complex metazoans,critical temperatures that affect fitness (i.e. survival and reproduction) are generally not set by cellular level responses (such as loss of protein or membrane function), but are rather set by a transition to unsustainable anaerobic metabolism. Thus, deleterious temperatures (referred to as the pejus by Pörtner) result from insufficient aerobic capacity of mitochondria at low temperatures, and a mismatch between excessive oxygen demand by mitochondria and insufficient oxygen uptake and distribution by ventilation and circulation at high temperatures. In other words, whole-animal metabolism is limited at both low and high temperatures, and this sets limits to animal performance. Seasonal and evolutionary adjustments to temperature are made by alterations of mitochondrial aerobic scope and these changes have concomitant effects on both high and low deleterious temperatures. In consequence, there is an inverse relationship between performance at high and low temperatures when measured in either a population across seasons, or among populations from habitats differing in their thermal regimes (Pörtner, 2001, 2002c; Pörtner et al., 2000). This inverse relationship is also apparent in interspecific comparisons,although there are differences between stenothermal polar species and those from other regions (see Pörtner,2002a).

Pörtner (2001, 2002a) has argued that this oxygen limitation of thermal tolerance applies as much to terrestrial animals as it does to the marine species for which he marshalled most evidence. In support of this proposition, Pörtner(2001, 2002a) points out that less complex organisms, such as eukaryotes and prokaryotes, have high thermal tolerances owing to the simplicity of their organization relative to spiders,scorpions, turtles and endothermic vertebrates which, as metazoans, have a much increased organizational complexity, resulting in a considerable decrease in their thermal tolerances. The unicellular organisms have no need of complex circulatory and gas exchange mechanisms, and therefore oxygen delivery does not set limits to performance.

However, there is a growing body of evidence suggesting that the thermal tolerances of terrestrial species might not be limited by the same mechanisms as those in marine species. The bulk of this evidence comes from insects, in which upper and lower lethal limits are not necessarily related. Rather, these limits are decoupled, such that alterations in low temperature tolerance do not usually result in a change in upper lethal limits. Such a decoupled response has been found in interspecific comparisons at global to regional spatial scales (Addo-Bediako et al.,2000; Chen et al.,1990; Gaston and Chown,1999; Goto et al.,2000), and in intraspecific comparisons at somewhat smaller scales(Hercus et al., 2000; Klok and Chown, 2003; but see also Hoffmann et al., 2002),and has also been documented in the responses of Drosophila species to selection (Gilchrist et al.,1997; Hoffmann et al.,1997). There is also a generally greater acclimation response in lower than in upper lethal temperatures, although the extent of variability in both sets of traits is often small and insufficient for perfect compensation(Chown, 2001; Kingsolver and Huey, 1998; Klok and Chown, 2003). These findings strongly suggest that thermal tolerance in insects, and possibly in other terrestrial ectotherms (see discussion in Chown, 2001; Klok and Chown, 2003), is not limited by oxygen delivery. However, to date no direct test of Pörtner's oxygen limitation hypothesis or any of its predictions has been undertaken for terrestrial ectotherms.

Of the many predictions arising from Pörtner's hypothesis, one of the most significant is that hypoxia should result in a decline in critical temperatures (Pörtner, 2001, 2002b). A test of this prediction, for insects (or other terrestrial ectotherms), initially appears straightforward. However, quite how critical limits should be identified in insects is a potential obstacle. In the work discussed by Pörtner(2001, 2002a), critical limits are reflected in a decline in aerobic scope (or a measure thereof, such as changes in haemolymph O₂ concentration), whereas in insects critical limits are generally measured as knockdown temperature as temperatures are altered(Gibert and Huey, 2001; Huey et al., 1992; Klok and Chown, 2003), or as time to knockdown at a given temperature(Hoffmann et al., 1997). Whilst lethal limits and knockdown temperatures are related to some degree,selection experiments often reveal a large measure of independence, reflecting the fact that these traits are genetically independent(Hoffmann et al., 1997; Berrigan and Hoffmann, 1998; Berrigan, 2000). Even so, it is clear that knockdown methods are not conducive to understanding oxygen demand,nor are they entirely free from observer bias if the onset of muscular spasms must be assessed, as suggested by Lutterschmidt and Hutchison(1997). However, J. R. B. Lighton and R. J. Turner (personal communication; see also Lighton and Turner, 2004) have developed a technique, dubbed thermolimit respirometry, which enables a marriage of both conventional observation of the critical thermal maximum(CT_max) using detection of movement by means of infra-red diodes, and realtime respirometry, which provides a measure of metabolic rate as temperature changes. This method objectively pinpoints the exact temperature where activity ceases and closely links this with changes in CO₂ release patterns.

In this study we use thermolimit respirometry to test directly the prediction of the oxygen limitation hypothesis that hypoxia should result in a decline in the CT_max. We also predicted that if the CT_max is a function of failure in oxygen uptake and distribution by ventilation and circulation, then hyperoxia should lead to an increase in the CT_max. To test these predictions we use two terrestrial species - an isopod, which makes use of pleopodal exopodites for gas exchange, and subsequent transport of oxygen via haemocyanin in the circulatory system (Schmidt and Wägele, 2001), and a tenebrionid beetle, which, like all insects (Chapman, 1998), has an elaborate tracheal system for delivery of oxygen directly to its tissues and cells.

Armadillidium vulgare (Latreille) (Isopoda: Armadillidiidae) (live mass: 0.0621±0.0018 g, mean ± s.e.m., N=10 for each treatment) and Gonocephalum simplex (Fabricius) (Coleoptera:Tenebrionidae) (live mass: 0.0523±0.0080 g, N=10 for each treatment) were collected 24 h prior to experiments from gardens on the University of Stellenbosch main campus during the austral mid-summer. Individuals were housed in the laboratory at 25±1°C with a natural photoperiod (16 h:8 h L:D) and were provided with soil and detritus from their microhabitats. In a pilot trial, CT_max was determined under normoxia by visual observation of knockdown temperatures in a water-jacketed set of Perspex chambers connected to a Grant (Cambridge, UK)LTD20 water bath programmed to increase temperature at 0.25°C min^-1 following initial equilibration for 15 min at 30°C.

For the main experiments, a Sable Systems (Las Vegas, NV, USA) flow-through CO₂ respirometry system was used to record gas exchange characteristics (LiCor 6262 CO₂/H₂O infra-red gas analyser; IRGA) and motor activity (AD1 activity detector; Sable Systems) (see Lighton, 1988). Compressed synthetic air (21% O₂ and balance N₂) was passed through soda lime and Drierite columns to remove CO₂ and H₂O. From there the scrubbed air flowed through a mass flow controller (Sidetrak;Monterey, CA, USA) set to regulate gas flow at 75 ml min^-1 into an automatic baselining system, the 5 mlcuvette containing the animal, and finally the IRGA. Sable Systems DATACAN V software was used for data capture and control of the respirometry system. The cuvette and activity detector were placed inside a waterproof container and immersed in a programmable water bath(Grant LTD20) set to equilibrate the animal at 30°C for 15 min in synchrony with the respirometry system's baseline and gas equilibration procedures. Thereafter the water bath increased the temperature at 0.25°C min^-1 to several degrees past the CT_max (all animals were dead by this point). A 40-SWG copper-constantan thermocouple connected to a Grant Squirrel SQ800 datalogger was used to monitor the cuvette's internal temperature in synchrony with DATACAN V. Owing to the isopods' sensitivity to desiccation, the synthetic air was rehumidified by inserting a LiCor LI610 dew point generator in the stream. At the start of a recording (from 30°C) a dew point of 20°C (2.347 kPa saturation vapour pressure) was set. The dew point was increased to 25°C (3.181 kPa) after the water bath reached 40°C.

Subsequent to the normoxic (21% O₂) CT_maxdetermination, thermolimit respirometry was repeated in three separate trials using hyperoxic air (40% O₂), mildly hypoxic air (10%O₂), and extremely hypoxic air (2.5% O₂). Oxygen concentrations were manipulated using two Sidetrak mass flow control units,providing N₂ and O₂ respectively, and connected in parallel so that they could be adjusted to provide the required concentration at a combined steady flow rate of 75 ml min^-1. Oxygen concentrations were monitored using an AMETEK S3A-II Oxygen Analyzer (Paoli,PA, USA). To allow accurate regulation of 2.5% O_2, the total flow rate was increased to 200 ml min^-1.

DATACAN V analysis software was used to extract thermolimit respirometry data from the recordings of every individual at the four O₂concentrations. Prior to data analyses the temperature data were combined and aligned with the CO₂ release and activity data from the Sable Systems recordings, using the time stamp of the instruments.

The critical thermal maximum was defined dually in terms of the species'motor activity, monitored by the AD1, and respiratory breakdown, based on V̇_CO2(Fig. 1). From activity data the CT_max, for both the isopods and the beetles, was recorded as the last temperature where a movement was detected by the AD1(Fig. 1). In the isopods, the respiratory signal that corresponded closely with the activity-based CT_max point was a brief spike in the CO₂emission. Because of their diffusion-based gas exchange via their pleopodal exopodites, the isopods gave generally smooth CO₂recordings and the characteristic spike at the onset of CT_max could be easily identified(Fig. 1). In the beetles,spiracular activity resulted in rapidly fluctuating CO₂ emissions at temperatures preceding the CT_max. However, their respirometry CT_max spike can be distinguished from preceding spikes by a smooth CO₂ decline following the CT_max onset, signifying the complete cessation of spiracular activity (Fig. 1). This cessation of spiracular activity therefore corresponds closely with the overall cessation of motor activity measured by the infrared activity detector.

In addition to the CT_max, we also measured the temperature of maximum metabolic activity (T_MetMax). This was the point at which CO₂ production reached its peak, and there was an inflection in the curve (indicated on Fig. 1). The deleterious temperature range was between the T_MetMax and the CT_max, where metabolic rate declined with increasing temperature (Fig. 1).

Least-squares linear regression was used to determine the relationship between logV̇_CO2and temperature, and the regression coefficients of these lines were compared among O₂ treatments using analysis of covariance (ANCOVA). Analyses of variance (ANOVA) were used to compare maximum V̇_CO2 and CT_max at the four O₂ concentrations in each of the test species.

From the starting temperature, 30°C, the animals experience benign thermal conditions, and thus metabolic rates remained stable, with no evidence of discontinuous gas exchange cycles in the beetle. At increasing temperatures the metabolic rates reach a maximum point analogous to the optimum range, as defined by Pörtner (2001, 2002a)(Fig. 1). This point, the T_MetMax, indicates a definite inflection in the metabolic rate curve from which the metabolic rate declines markedly for a few degrees -the deleterious temperature range - before another break appears,characterised by a brief CO₂ spike in the recording (Figs 1, 2A, 3A) signifying the respirometry CT_max. The animals did not recover once they reached the post-CO₂ decline (Fig. 1) and we regarded them as functionally dead. After death there is a brief rise in CO₂ emission as residual CO₂ is released from the body at the increasing temperature, followed by a steady and smooth decline until CO₂ emission rates approach zero.

Pilot trials indicated that during normoxia both the respirometry and activity CT_max values were identical to the knockdown CT_max estimated visually for both species (Isopod 44.4±0.4°C, mean ± s.e.m., ANOVA F_(1,27)=0.001, P>0.98; beetle 48.8±0.3°C ANOVA F_(1,18)=1.017, P>0.32). Therefore, identifying the CT_maxvalues based on activity recordings and on respirometry recordings was straightforward and consistent for both Armadillidium vulgare and Gonocephalum simplex.

In A. vulgare, the CO₂ traces were smooth(Fig. 2), as might be expected for a species with no means of physically regulating gas exchange. The regression coefficients of the rate-temperature relationships decreased with declining oxygen concentration, from 40% to 21% O₂, but were idiosyncratic thereafter (Table 1). At 2.5% O₂, the large majority (7 out of 10) of the individuals investigated showed a decline in metabolic rate(Fig. 2C) with increasing temperature, suggesting that this species was under considerable stress and is a metabolic conformer (sensuHerreid, 1980), with metabolic rate declining in response to hypoxia. This was reflected in the significant decline of T_MetMax with declining oxygen concentration(Table 1). There was a decrease in both activity and respirometry CT_max with declining oxygen concentration, at least at and below normoxia(Table 2). At values between 21% and 40% O₂ there was no change in CT_max(Table 2).

Although the regression coefficients of the relationships between V̇_CO2 and temperature differed with O₂ concentration, albeit in inconsistent directions (Table 1), the majority of the respiratory and thermal parameters in G. simplexdiffered from A. vulgare. Excluding extreme hypoxia, T_MetMax did not differ with %O₂. The CO₂ traces also showed similar, marked fluctuations, suggesting spiracular opening and closing, probably as a consequence of animal activity(Fig. 3). At 2.5%O₂, the CO₂ trace was generally much smoother,corresponding to reduced activity levels(Fig. 3C) and maximum spiracle opening to ensure sufficient O₂ uptake at that low O₂concentration. Moreover, metabolic rates were significantly lower than those at the other concentrations, as was the regression coefficient of the rate temperature relationship (Table 1). In other words, oxygen delivery only becomes problematic when the concentration declines from 10% to 2.5% O₂. The respirometry CT_max was unaffected by oxygen concentration, although the beetles tended to maintain very low activity levels at 2.5% O₂,resulting in a lower, but probably artefactual, activity CT_max (Table 2).

If oxygen limitation of thermal tolerance is a unifying principle at temperatures above freezing (Pörtner, 2001, 2002a), then the predictions of the hypothesis should apply to all ectothermic animals, or at least to the vast majority of them. In this particular instance, CT_maxshould decline with hypoxia and increase with hyperoxia. The two terrestrial species we investigated showed very different responses to changing oxygen partial pressures. In the isopod there was a significant decline in CT_max with hypoxia, whereas there was no significant change in CT_max in the tenebrionid beetle. Moreover, it was clear that there was a considerable effect of hypoxia on maximum metabolic rates in the isopod, whereas the effect was noticeable only at 2.5%O₂ in the tenebrionid, and then did not cause a pronounced decline in the CT_max. This difference can readily be explained by the dissimilarity of the gas exchange and transport systems of the two species. Terrestrial isopods such as A. vulgare exchange gasses by means of pleopod exopodites, from where oxygen is transported to the areas where it is required by haemolymph, containing the respiratory pigment haemocyanin (Ruppert and Barnes,1994; Schmidt and Wägele,2001). In this respect A. vulgare is functionally similar to many of the marine species investigated by Pörtner and his colleagues(e.g. Frederich and Pörtner,2000; Mark et al.,2002; Peck et al.,2002; Sommer et al.,1997; van Dijk et al.,1999). Moreover, it is also similar to the terrestrial species from which evidence has been cited in support of the oxygen limitation hypothesis (Pörtner,2002a). Spiders and scorpions generally exchange gasses using book lungs and circulation is via the haemolymph using respiratory pigments (Bridges et al., 1997; Burmester, 2002; Paul et al., 1987), whilst in the terrestrial vertebrate models that Pörtner discussed, gas exchange and circulation are also based on two separate systems. It is no surprise in animals using a two-stage system for gas exchange and its subsequent delivery that at high temperatures a mismatch between excessive oxygen demand by mitochondria and insufficient oxygen uptake and distribution should arise(Pörtner, 2001, 2002a).

By contrast, insects have a very different system of gas exchange and delivery. In general, oxygen moves along a pathway from the spiracles through the main tracheal tubes (via convection or diffusion) to the tracheoles, where it diffuses to the mitochondria(Buck, 1962; Kestler, 1985; Nation, 2002). Some diffusion from the large tracheae to the surrounding tissues or haemolymph also takes place, although it probably accounts for no more than 25% of the total because of the small partial pressure difference between the tracheal lumen and the surrounding tissues (Schmitz and Perry,1999). Carbon dioxide does not follow the same route in the opposite direction. Rather, it is thought to enter the tracheal system at all points from the tissues and haemolymph, where it is buffered as bicarbonate(Bridges and Scheid, 1982; Schmitz and Perry, 1999). Thus, O₂ delivery is highly efficient. Indeed, most insects can be considered metabolic regulators in the sense that oxygen consumption remains constant with declining O₂ concentration until a critical oxygen tension of about 5-10% O₂ is reached (generally lower in adults)(Loudon, 1988). Only at oxygen tensions below this critical point does metabolic rate decline. However, many species remain apparently unaffected at low oxygen partial pressures,maintaining gas exchange and activity below these levels(Holter and Spangenberg, 1997; Chown and Holter, 2000). This suggests that oxygen limitation is unlikely for insects under a wide variety of conditions, and would certainly not limit thermotolerance, at least not under the conditions generally experienced in terrestrial habitats. Our data on the tenebrionid beetle G. simplex certainly support this idea, and indicate that oxygen limitation of thermal tolerance does not take place:hypoxia has no effect on CT_max.

One counterargument might be that our assessments of CT_max were undertaken over short timescales and that the deleterious limits only become apparent over longer periods (Pörtner, 2001, 2002a). In other words, had we continued our trials for a longer period, oxygen limitation of thermal tolerance would have become apparent. In our view this is unlikely. Insects are known to be highly responsive to changing oxygen availability and can compensate for longer-term changes in gas concentrations in both physiological and developmental ways (Frazier et al.,2001; Loudon,1988; Wigglesworth,1935). Even in flying insects, flight metabolism is generally only sensitive to ambient oxygen partial pressures below 10% O₂(Harrison and Lighton, 1998; Joos et al., 1997). Thus, it seems unlikely that a mismatch between oxygen supply and demand sets thermal limits in insects. In addition, in contrast to stenothermal subtidal marine habitats, microclimates can and do reach upper lethal temperatures for terrestrial arthropods (e.g. Neargarder et al., 2003; Roberts and Feder,1999; Wehner et al.,1992), and approach lethal limits in intertidal environments(Helmuth et al., 2002).

The question of what accounts for thermal limits in insects then remains. At high temperatures, thermal limits are probably the consequence of differing responses to heat injury at the cellular level. High temperature injury generally results from disruption of membrane structure, and problems associated with protein folding. This causes a breakdown in the function of membranes, especially synaptic membranes, alterations in the cell microenvironment, DNA lesions and perturbation of protein structure(Feder, 1999; Somero, 1995). The importance of responses to cellular level thermal damage for whole-organismal survival is clearly reflected in geographic variation in heat shock protein expression associated with variation in environmental temperatures to which the organisms are exposed (Bettencourt et al.,2002; Dahlhoff and Rank,2000; Neargarder et al.,2003; Sørensen et al.,2001). In addition, thermal resistance of the nervous system has been correlated with concentrations of membrane polyunsaturated fatty acids(PUFAs) in vertebrates (Logue et al.,2000; Hulbert,2003), and could also play a significant role in insects. In turn,membrane damage affects development, neural functioning, muscular contraction and several other processes at higher organizational levels(Denlinger and Yocum, 1998). Such determinants of CT_max at the cellular level in insects would make the CT_max independent of oxygen availability, as we have observed. Moreover, the absence of a response of CT_max to hyperoxia in our experiments also suggests that cellular level damage sets the CT_max - no amount of improved oxygen delivery can alter the value upwards. Intriguingly, this was also the case in the isopod, suggesting that a combination of higher level organizational constraints and cellular level resistance to thermal injury might set the upper thermal limit (see also Sokolova and Pörtner,2003).

However, it seems that insufficient aerobic capacity of mitochondria at low temperature might well be important in setting lower critical limits. In freeze-tolerant Pringleophaga marioni (Lepidoptera, Tineidae)caterpillars there is a precipitous decline in metabolic rate at the critical thermal minimum (Sinclair et al.,2004). Moreover, in both honey bees and Drosophila,decreasing temperature results in a steady decline in the resting potential of flight muscle neurons. The critical thermal minimum appears to be the temperature at which the Na⁺/K⁺-ATPase pump can no longer maintain nerve cell polarisation to a level where action potentials could be produced (Hosler et al.,2000). Thus, lack of ATP owing to insufficient aerobic metabolism in the mitochondria might well set lower limits in insects. This is in keeping with Pörtner's hypothesis(Pörtner, 2001), and also supports the idea that mechanisms underlying the higher and lower critical thermal limits in tracheated terrestrial ectotherms are different(Chown, 2001).

Our results therefore suggest that oxygen limitation of upper thermal tolerance is unlikely in insects, and probably also in other tracheated arthropods. They also suggest an explanation for the observation that, in insects, upper and lower lethal limits are generally decoupled(Chown, 2001), whereas this is not the case in most marine species(Pörtner, 2001). Thus,the question remains as to whether oxygen limitation of thermal tolerance can be considered a unifying general principle. This depends very much on one's perspective. From an entomological perspective, oxygen limitation is not pervasive: most terrestrial animal species are tracheated arthropods with two sets of wings (Samways, 1994). In this regard, terrestrial species are very different to those from marine environments. However, from a phylogenetic perspective it is clear that the members of most other higher taxa, with their two-stage oxygen delivery systems, probably face oxygen limitation of thermal tolerance. At present, too few species have been examined to verify these ideas, but clearly Pörtner's hypothesis makes several testable predictions that would cast considerable light on the apparent dichotomy.

John Lighton and Robbin Turner are thanked for insightful discussions concerning the interpretation of thermolimit respirometry and for providing access to unpublished information. Sue Jackson and Richard Mercer and two anonymous referees provided useful and insightful comments on an earlier version of the manuscript. C.J.K. was supported by a National Institutes of Health Grant AI-5245601, B.J.S. by a New Zealand Science and Technology Post-Doctoral Fellowship, and S.L.C. by National Institutes of Health Grant AI-5245601 and by National Research Foundation Grant GUN 2053804.