Differences in critical thermal maxima and mortality across life stages of the mealworm beetle Tenebrio molitor

Thermal limits to activity are profoundly important for determining the abundance and geographic distribution of ectothermic animals. Understanding thermal limitation is a topic of growing interest among biologists, primarily because of concerns regarding the effects of global climate change on biodiversity (Deutsch et al., 2008). Our ability to identify the physiological factors that underlie temperature-related distribution patterns is highly important for making predictions about where species will occur and how they will perform in warmer climates (Somero, 2010). Thermal limits to activity are often determined experimentally as the critical thermal maximum (CT_max) and minimum (CT_min) (Klok and Chown, 2003; Folk et al., 2007; Terblanche et al., 2007; Nyamukondiwa and Terblanche, 2009), which are the temperatures at which activity becomes uncontrolled (see Cowles and Bogart, 1944). Although critical thermal limits are widely measured, methods for identifying CT_max and CT_min are highly varied. To determine upper thermal limits, CT_max is commonly measured by identifying the temperature at which behavioral endpoints (loss of righting response, onset of muscle spasms, knock down) occur as temperature increases (reviewed by Lutterschmidt and Hutchison, 1997). As behavioral responses must be identified visually by the experimenter, these measurements of CT_max strongly depend on both the visual clarity of the animal subject and the skill and vigilance of the observer.

Prior work has shown that CT_max measurements depend on a variety of factors, especially those that influence the thermal history of the animal. These include season (Terblanche et al., 2006), hardening and acclimation (Jumbam et al., 2009; Nyamukondiwa and Terblanche, 2010), and aspects of the experimental design such as the rate of temperature increase (Terblanche et al., 2007; Chown et al., 2009). Terblanche and colleagues demonstrated that both the rate of temperature change and the starting temperature significantly impact upper and lower critical limits in tsetse flies (Terblanche et al., 2007). For example, they showed that using faster rates of change and starting at temperatures closer to the thermal limit both result in higher estimates of CT_max and lower estimates of CT_min (Terblanche et al., 2007). The results of that study highlight the importance of experimental design on the outcome of CT_max experiments, and the need to clearly define which experimental conditions are used.

Recently, a technique termed ‘thermolimit respirometry’ was developed (Lighton and Turner, 2004) that allows CT_max to be identified as the point of spiracular and/or motor failure. The technique employs flow-through respirometry to measure CO₂ output from the animal during increasing temperature, typically with simultaneous infrared activity detection to monitor locomotor activity (Lighton and Turner, 2004; Klok et al., 2004; Folk et al., 2007; Lighton, 2007; Stevens et al., 2010). Prior studies have shown that these two independent measures of thermal limitation, spiracular and motor failure, occur simultaneously in adult insects tested under normoxic conditions, and thus represent statistically equivalent estimates of CT_max (Lighton and Turner, 2004; Klok et al., 2004; Folk et al., 2007; Stevens et al., 2010). The technique has several advantages. First, it allows CT_max to be identified directly from respiratory and/or activity data, rather than by visual inspection alone. Additionally, it permits CT_max to be determined without disturbing the animal. This is an important advantage over techniques that require repeated disturbances during the experiment (i.e. assessments of righting response), given that these methods have been shown to produce significantly higher CT_max values compared with assessments of undisturbed animals (Stevens et al., 2010). Because of these advantages, thermolimit respirometry continues to gain popularity as a technique among investigators measuring CT_max, particularly in arthropods.

Prior studies using thermolimit respirometry have reported that tracheated arthropods (i.e. most insects) produce a characteristic pattern of CO₂ output during increasing temperature, which includes seven phases: (1) V̇_CO2 remains constant while temperature is maintained at the start temperature (equilibration), (2) V̇_CO2 increases exponentially as temperature increases (ramping), (3) V̇_CO2 reaches a plateau, (4) V̇_CO2 drops abruptly and spiracular activity ceases (mortal fall), (5) V̇_CO2 declines to a low point (valley), (6) V̇_CO2 rises to form a large peak (post-mortal peak) and, finally, (7) V̇_CO2 decays exponentially after death (Lighton and Turner, 2004) (see also Fig. 1). The dramatic reduction in V̇_CO2 variance that occurs during the mortal fall phase signifies the loss of spiracular control, and therefore CT_max. This seven-phase pattern has been described in a variety of insects, including harvester ants (Lighton and Turner, 2004), drosophilid flies (Folk et al., 2007; Lighton, 2007) and tenebrionid beetles (Klok et al., 2004; Stevens et al., 2010). Only isopods have shown a different pattern of CO₂ release, which may result from differences in respiratory morphology between isopods and insects (Klok et al., 2004; Stevens et al., 2010). However, the seven-phase pattern shows little variation in those insects with tracheated respiratory systems gated by spiracular closure muscles (Lighton and Turner, 2004; Klok et al., 2004; Folk et al., 2007; Lighton, 2007; Stevens et al., 2010).

Because this V̇_CO2 pattern is a fundamental component of thermolimit respirometry, accurate determination of CT_max using this technique requires that the phases of the pattern be both well understood and highly repeatable. However, to date there has been no formal investigation of the variability in the V̇_CO2 pattern in the different developmental stages of an insect. Specifically, work is needed to determine the universality of the CO₂ release patterns across development. Additionally, it is not known at what point during this response death occurs under sustained heating, or how lethal temperatures relate to CT_max. Unlike sub-lethal assessments of CT_max (such as knock-down experiments), thermolimit respirometry requires the animal to be exposed to a sustained thermal, which continues even after CT_max is reached. As temperature increases, CT_max will occur first during the ramp, and if heating is sustained beyond this, the animal will later die. Although it is known that death occurs at some point following CT_max under sustained heating, it is not known how closely lethal temperatures and CT_max are related, or which phase of the CO₂ release pattern corresponds to death. We have used thermolimit respirometry to investigate these responses in three life stages of Tenebrio molitor beetles during increasing temperature. Our objectives were (i) to determine the extent to which patterns of CO₂ release during heating are consistent across life stages of T. molitor, (ii) to compare three different methods of analysis for determining CT_max, (iii) to compare CT_max across life stages, and (iv) to map lethal temperatures onto the pattern of CO₂ release for each life stage. The results of this study provide the first comparison of CO₂ release patterns during heating across the developmental stages of an insect, and the first experimental link between mortality and the CO₂ release pattern produced by insects during thermolimit respirometry.

Larval, pupal and adult T. molitor L. beetles from a colony maintained in our laboratory at the University of California, Irvine, were used for these experiments. All life stages were kept together in a glass terrarium containing old-fashioned rolled oats (Quaker, Chicago, IL, USA) to a depth of 3 cm, Brewer's yeast flakes (Lewis Laboratories International, Southport, CT, USA) to supplement nutrition, and a beaker filled with water to increase relative humidity. Beetles were kept at room temperature (25°C) on a 12 h:12 h L:D cycle. Prior to each experimental trial, the beetle subject was weighed on an analytical balance (Mettler Toledo AB104, Riverside, CA, USA) to the nearest 0.1 mg. Live mass (mean ± s.e.m.) was 180.1±8.8 mg for larvae, 167.8±7.1 mg for pupae and 132.3±6.8 mg for adults.

Thermolimit respirometry (Lighton and Turner, 2004) was employed using a flow-through respirometry system to measure CO₂ release from each beetle while the air temperature was steadily increased. Respirometry chambers (2 ml volume) were constructed from plastic syringes and sealed with a single-hole rubber stopper (no. 1). Each chamber was lined with fine mesh to prevent the beetles from escaping through the entrance and exit ports, and to promote turbulent flow of air within the chamber.

Air was pushed across two columns of silica gel and one column of Ascarite to remove water and CO₂, respectively, and through a mass flow controller (0–200 ml model, Side-trak, Sierra Instruments, Monterey, CA, USA) to maintain the air flow at 100 ml min^–1. The air stream then entered a temperature-controlled cabinet (Sable Systems International, Las Vegas, NV, USA), where it passed through a coil of copper tubing (160 cm length) to ensure thermal equilibration of the air before entering the respirometry chamber where the insect was held. The air leaving the chamber then exited the temperature-controlled cabinet and entered an infrared CO₂ analyzer (LiCor, Model LI-6251, Lincoln, NE, USA).

The air temperature was manipulated using a temperature-controlled cabinet wired to a PELT-5 temperature controller (Sable Systems International). The temperature of the airstream flowing across the beetle was measured using a thermocouple wire inserted into the tubing at the exit of the respirometry chamber, and monitored using a TC-2000 thermocouple meter (Sable Systems International). The outputs from the thermocouple meter and CO₂ analyzer were connected to a laptop computer via a data acquisition interface (UI2 Interface, Sable Systems International).

Recording was initiated as air was pushed through an empty chamber to establish baseline CO₂ values for 1–2 min. Recording was paused as the beetle was placed into the respirometry chamber, then resumed after 2 min. At this point the temperature profile programmed into the PELT-5 was initiated. The profile began with a 20 min equilibration period at 25°C, followed by a temperature ramp at 0.5°C min^–1 for 50 min, a 10 min period at 50°C, and finally a decrease in temperature at –10°C min^–1 until the chamber returned to 25°C. Although 0.5°C min^–1 is likely much greater than rates of temperature increase experienced by these insects under field conditions (see Terblanche et al., 2011), these experiments were designed to detect physiological differences in CT_max among life stages rather than to mimic ecological conditions. Therefore, the fastest rate of change that did not produce a lag effect [see Lighton and Turner for discussion of lag effects (Lighton and Turner, 2004)] was selected. Temperature (°C) and CO₂ concentration (p.p.m.) were recorded once per second for the duration of each trial using Expedata Software (Expedata PRO, version 1.3.4, Sable Systems International). This procedure was repeated in beetles at each life stage.

V̇_CO2 data were used exclusively to determine CT_max values in this study. Infrared activity detection was not used as we made comparisons between mobile and immobile life stages. We did not wish to use a form of analysis for two of the life stages (larvae and adults) that could not be used for the third (pupae). Therefore, we elected to use the V̇_CO2 patterns to determine CT_max in all three life stages. Although prior studies using thermolimit respirometry have been conducted exclusively in adult insects, these studies have shown that V̇_CO2 and activity measurements produce equivalent estimates of CT_max in tracheated insects tested under normoxia (Lighton and Turner, 2004; Folk et al., 2007; Klok et al., 2004; Stevens et al., 2010).

All three life stages of T. molitor were used to determine the point of death during heating in order to map lethal temperatures onto the patterns of CO₂ release. To do this, we assessed recovery in individual beetles from each life stage after exposure to temperature profiles similar to those used for CT_max trials (25–50°C at 0.5°C min^–1). The same respirometry protocol used during CT_max experiments was employed, with the exception that the temperature increase was stopped at a specific, predetermined temperature between 46 and 50°C. Each beetle was then kept inside the respirometry chamber at room temperature (25°C) for 30 min for recovery. Beetles were scored as alive if a pre-ramping V̇_CO2 release pattern resumed within 30 min of returning to 25°C, or as dead if a pattern of controlled V̇_CO2 release did not resume. This procedure was repeated for multiple individual beetles of each life stage at each test temperature. To determine the relationship between the lethal temperatures and CT_max, we determined the percentage of individuals for which CT_max fell within the range of lethal temperatures for each life stage.

Data analysis was performed using Expedata analysis software (version 1.3.4, Sable Systems International). V̇_CO2 data were first copied into a new channel for analysis, baseline corrected, and converted from p.p.m. to μl CO₂ h^–1. Corrected data were analyzed to determine CT_max using methods modified from Lighton and Turner (Lighton and Turner, 2004). Lighton and Turner propose identifying the inflection point in the absolute difference sum (ADS), which is the cumulative sum of the absolute differences between adjacent V̇_CO2 data points. This inflection point is suggested to be an objective method for identifying the point at which short-term variability in the V̇_CO2 data declines abruptly, indicating the cessation of spiracular activity by the insect. This technique has the advantage of being highly repeatable, and as objective as may be possible with these types of data. However, in employing these analyses we discovered that this method does not always identify the ‘last breath’ produced by the insect, especially in the pre-adult life stages. Therefore, we compared CT_max values identified using three distinct methods (see Fig. 2). We first identified CT_max as the ADS inflection point described above. In brief, the ADS inflection was determined by selecting the 10 min interval surrounding the expected CT_max point, and fitting a linear regression to these ADS data points. We identified the 10 highest consecutive residuals of this regression, and the mean temperature of these residuals was recorded as the ADS CT_max (see Fig. 2B). Second, to determine the temperature at which voluntary control of the spiracular muscles failed, we identified CT_max as the mean temperature of the final 10 s before the cessation of spiracular activity (CSA). This point was located using the ‘differentiate’ and ‘square’ functions in Expedata to magnify differences between spiracular activity and electrical noise. The point at which this differentiated trace leveled to zero was then marked as the point of spiracular failure, and the final 10 data points prior to this point were selected. The mean temperature of these data points was recorded as the CSA CT_max (see Fig. 2C). Both of these CT_max measurements were completed using an Expedata macro (supplementary material Table S1). Lastly, we identified the temperature at the pre-mortal valley (PMV), a metric that has been favored by previous authors estimating CT_max with thermolimit respirometry in T. molitor (Stevens et al., 2010). The PMV was measured in Expedata using a nadir search function to find the lowest 10 consecutive points between the ramping phase and the post-mortal peak. The mean temperature of these points was recorded as the PMV CT_max (Fig. 2A).

Analysis of variance (ANOVA) was used to compare each of the three metrics (ADS, CSA, PMV) across life stages, and Tukey's HSD post hoc tests were used to separate statistically different groups. Prior to analyses, data were assessed for normality and equality of variance using Shapiro–Wilk and Barlett's tests, respectively. These analyses were completed in SAS (v. 9.2, SAS Institute, Cary, NC, USA). Data are presented as means ± s.e.m., and significance was set at P=0.05. For mortality assays, percentage survival at each test temperature was calculated and plotted separately for larvae, pupae and adults. The temperature at which death occurred during heating for each life stage was identified as the range of temperatures at which survival was less than 100% but greater than 0%. This range of lethal temperatures was mapped onto the pattern of CO₂ release for each life stage to compare the temperature at which death occurred with CT_max. A two-proportion z-test (Sheskin, 2004) was used to determine whether the proportion of surviving animals at each test temperature was significantly different between the life stages.

We compared the patterns of CO₂ release in larval, pupal and adult life stages of T. molitor using thermolimit respirometry. Each animal was acclimated for 20 min at 25°C before temperature was increased from 25 to 50°C at 0.5°C min^–1. The actual measured rate of temperature change was compared across life stages using analysis of variance (ANOVA), which confirmed that the ramp rate was identical for all three groups at 0.48±0.01°C min^–1 (P=0.960). At all three life stages, T. molitor showed an increase in V̇_CO2 with temperature until a maximum V̇_CO2 was reached (Figs 1, 3). Maximal V̇_CO2 was typically followed by a drop in V̇_CO2, i.e. the CO₂ valley. In adults, this drop was deep and quite evident (Fig. 3C), while in larvae and pupae this feature was less evident and in fact often absent (Fig. 3A,B). All three life stages showed a subsequent burst of CO₂ as temperature continued to increase, followed by an exponential decline in V̇_CO2 to near-zero values (Fig. 3).

Adult T. molitor produced the expected pattern of CO₂ release (Fig. 1, Fig. 3C) described by several studies using thermolimit respirometry in insects. This pattern is characterized by seven phases: equilibration, ramping, plateau, mortal fall (CT_max), post-mortal valley, post-mortal peak and exponential decay (Lighton and Turner, 2004) (Fig. 1). Prior work (Stevens et al., 2010) that used thermolimit respirometry to assess CT_max in adult T. molitor reported the same pattern of CO₂ release from these animals.

By contrast, the patterns of CO₂ release produced by T. molitor larvae and pupae differed from those of the adults in two ways. First, CT_max estimates (determined using the ADS and CSA methods) occurred during the post-mortal peak phase in larvae and pupae, rather than during the pre-mortal valley as in adults (Fig. 3A,B). Additionally, the smooth and deep decline in V̇_CO2 (i.e. the pre-mortal valley) prior to the post-mortal peak seen in adults (Fig. 3C) did not appear in larvae and pupae (Fig. 3A,B). While a decrease in V̇_CO2 was evident in some pre-adult animals, spiracular control was maintained throughout this phase, and instead CT_max occurred during the first half of the post-mortal peak. In other words, spiracular failure consistently occurred during different phases of the V̇_CO2 pattern in pre-adult and adult life stages of this species. Nonetheless, the cessation of spiracular control (CT_max) was clearly evident in the CO₂ traces of all three life stages.

Second, the pattern of CO₂ release for larvae and pupae differed from that of the adults in terms of the size and shape of the post-mortal peak. In adults, this sharp peak was invariably greater in amplitude than maximal V̇_CO2, and occurred after CT_max and the valley phase (Fig. 3). By contrast, the post-mortal peak in larvae and pupae appeared as a gradual rise in V̇_CO2 that was typically of equal or only slightly greater amplitude than maximal V̇_CO2(Fig. 3A,B). While much less pronounced in larvae and pupae, the post-mortal peak was present in all the animals we tested.

CT_max has been defined (Lighton and Turner, 2004) as the temperature at which spiracular control ceases. We used Expedata to determine CT_max using this criterion as the average temperature of the point during which spiracular failure occurred for each animal. We compared three methods for estimating this point: (1) the inflection point in the ADS residuals, (2) the 10 s interval prior to the CSA and (3) the low point in the PMV. ANOVA was used to compare these metrics among the life stages, followed by Tukey post hoc tests to distinguish between statistically different groups. We found that CT_max determined using all three methods was significantly higher in adult T. molitor than in either larvae or pupae (P<0.001), but statistically identical between larvae and pupae (Table 1). When comparing the three methods for estimating CT_max with an ANOVA, we found that CT_max values varied significantly with analytical method in pupae (P=0.002), while values were marginally different across methods in larvae (P=0.049), and identical across methods for adults (P=0.626). According to post hoc tests, in both larvae and pupae the ADS estimate was identical to the CSA estimate. By contrast, the PMV estimate was significantly lower than the ADS in both larvae and pupae, and was significantly lower than the CSA estimates in pupae. This last result is primarily due to the fact that the pupae (and a small percentage of larvae) did not display a clear pre-mortal valley like the adults. Instead, spiracular activity continued steadily in these animals until the post-mortal peak (Fig. 3B). As a result, PMV estimates occurred well in advance of ADS and CSA measurements in pupae, and may not actually represent a corresponding point to the PMV values measured in adults. By contrast, CT_max values did not differ across analytical methods in adults, as spiracular failure occurred simultaneously with the pre-mortal valley in this life stage.

During data analyses, it was observed that the ADS method, unlike the CSA method, did not consistently identify the point of spiracular failure in this species, particularly in the pre-adult life stages. We quantified this by determining the percentage of animals for which ADS and CSA estimates of CT_max differed by more than 0.5°C. We found that differences greater than 0.5°C occurred in 48% of the animals we measured, the majority of which were larvae and pupae. In all cases for which the ADS and CSA differed by more than 0.5°C, the ADS value was lower than the CSA value.

In a separate set of measurements, we compared mortality during heating among life stages by removing the insects from the temperature ramp at a precise, predetermined temperature. Each insect was scored for survival based on whether pre-CT_max spiracular activity resumed within 30 min of the return to 25°C. From these data, we calculated percentage survival for each life stage at each test temperature, and identified the range of temperatures at which recovery was not possible for each life stage (temperatures between 0 and 100% survival) (Fig. 4). Survival was 100% for all life stages at 46.5°C and below. The range of lethal temperatures for adults spanned a higher range (47.5–49°C) than that for larvae (46.5–48°C) or pupae (46.5–48.5°C). To assess whether lethal temperatures differed with life stage, we compared the proportion of surviving animals from each life stage at each test temperature using a two-proportion z-test, which showed that survival was significantly greater in adults at 48.5°C than in either larvae or pupae at that temperature (P<0.050; N=4 for larvae, N=4 for pupae, N=10 for adults) (Fig. 5).

With the CT_max and mortality data combined, we were able to discern a clear association between the range of lethal temperatures and CT_max for each life stage. More than 90% of the beetles tested displayed a CT_max that occurred within 1°C of the range of lethal temperatures for its respective life stage, with 50% occurring just below the lethal range and 50% occurring within this range (Fig. 5). We also found that lethal temperatures spanned the period between CT_max and the zenith of the post-mortal peak in more than 77% of the individuals tested (Fig. 3).

This paper is the first to report distinct patterns of CO₂ release in different life stages of a single insect species during thermolimit respirometry. Adult T. molitor displayed a pattern of CO₂ release during thermolimit respirometry that has been well described in the literature (Lighton and Turner, 2004; Klok et al., 2004; Folk et al., 2007; Stevens et al., 2010). However, we found that the CO₂ release pattern produced by larvae and pupae differed from that of adults in terms of both when spiracular failure occurred and the characteristics of the post-mortal peak. Most notable was that the point of spiracular failure in pre-adult T. molitor occurred during the post-mortal peak phase rather than during the CO₂ valley as it did in adults.

This study is the first to use thermolimit respirometry in pre-adult insects. Prior studies have used this technique in adults and found remarkably little variation in the pattern of CO₂ release across the species that have been examined. The V̇_CO2 pattern reported for harvester ants [Pogonomyrmex rugosus and P. californicus (Lighton and Turner, 2004)], drosophilid flies [Drosophila melanogaster (Folk et al., 2007; Lighton, 2007)] and tenebrionid beetles [Gonocephalum simplex (Klok et al., 2004); T. molitor (Stevens et al., 2010)] all resemble each other closely and match the result found for adult T. molitor in this study. Patterns of CO₂ release have also been reported during thermolimit respirometry for two non-insects, the isopod species Armadillidium vulgare (Klok et al., 2004) and Porcellio scaber (Stevens et al., 2010). In isopods, gases are exchanged through pleopodal exopodites, and circulated to the tissues by respiratory pigments in the hemolymph (Wright and Ting, 2006). This differs markedly from insects, in which gases are delivered directly to the tissues via the air-filled tracheal system. During rising temperature, isopods showed a smooth rise in CO₂ release until CO₂ plummeted at CT_max, followed by a broad post-mortal peak (Klok et al., 2004; Stevens et al., 2010). This variation in pattern between insects and isopods has been suggested to result from differences in respiratory morphology, and to date has been the only example of variation in V̇_CO2 patterns during heating.

We have demonstrated here that the pattern of CO₂ release during heating described in prior studies of adult insects can vary in other life stages within the same species. The source of these differences in CO₂ release across development remains unclear, however. Variation may be related to developmental differences in either the structure of the tracheal system or control of the spiracles. Mölich and Lighton have found that the post-mortal peak is abruptly truncated in adult Drosophila if, during thermolimit respirometry, the incurrent gas is changed from air to pure nitrogen shortly after the post-mortal rise in CO₂ output begins (Mölich and Lighton, 2007). They interpreted this to mean that the formulation of CO₂ in the post-mortal peak requires oxygen and thus is a product of aerobic respiration. We have obtained the same result using adult mosquitoes (A.S.V., E. M. Gray and T.J.B., unpublished). It is possible that the post-mortal peak results from run-away aerobic respiration in the mitochondria at the point of, or just after, death of the organism. In their recent paper, Mölich and colleagues discuss the origin of this phenomenon, which they refer to as mitochondrial ‘hyperthermic overdrive’ (Mölich et al., 2012). They explain that hyperthermic overdrive may originate from the breakdown of the mitochondrial membrane under high temperatures, thereby increasing the permeability of the inner mitochondrial membrane to protons and causing the uncoupling of oxidative phosphorylation (Mölich et al., 2012). This process could result in the appearance of a post-mortal burst of CO₂ as the gas diffuses out of the open tracheal system, particularly in insects with well-developed tracheal systems. For example, adult insects such as flying mosquitoes, tenebrionid beetles and foraging ants can be highly active and often have well-developed tracheal systems. Therefore, the CO₂ produced in these insects can readily diffuse out of the tracheal system if the spiracles are open. Pupal life stages are similarly developing the adult tracheal system and have a well-developed capacity for gas exchange. By contrast, larvae tend to be less active and do not experience the extremes of aerobic activity found in the adults. It may be, therefore, that the larvae have fewer mitochondria and/or a less extensive tracheal system, and that the CO₂ formed when the mitochondria become more permeable at high temperature does not readily escape as a ‘post-mortal peak’.

We found that CT_max differed across life stages in T. molitor, with adult beetles showing significantly higher CT_max compared with larval and pupal beetles (P<0.001). A variety of studies have reported differences in thermal tolerance among insect life stages, although the majority of this work examined cold tolerance rather than heat tolerance (reviewed in Bowler and Terblanche, 2008). Studies that have investigated the effects of ontogeny on heat tolerance have typically shown that tolerance declines in later stages. Authors have attributed this trend to increasing mobility across ontogeny, allowing adults to compensate behaviorally for changes in temperature and thus avoid dangerous extremes (Bowler and Terblanche, 2008; Marais et al., 2009). However, there are exceptions to this generality, such as a study in Drosophila buzzatii which demonstrated that variation in temperature tolerance across life stages does not always proceed chronologically, with heat resistance being highest in pupae, followed by eggs, first instar larvae and finally third instar larvae (Krebs and Loeschcke, 1995). In their recent review, Bowler and Terblanche cautioned against generalizations regarding the effects of ontogeny on heat resistance, as these relationships vary significantly across studies (Bowler and Terblanche, 2008). Our finding that CT_max is highest in adult T. molitor could reflect the interaction of multiple factors, such as the microclimate conditions experienced by each life stage and the degree to which behavioral compensation can be used to avoid extreme temperatures (Bowler and Terblanche, 2008).

The CT_max values we obtained for T. molitor differ from those reported previously for this species. Stevens and colleagues used both the ADS method and the PMV method to estimate CT_max in T. molitor, and reported values of 44.0 and 44.9°C, respectively (Stevens et al., 2010). These values are approximately 4°C lower than those found using the ADS and PMV methods in our study. However, these differences are to be expected given the differences in our experimental design, notably the rate of temperature change. While Stevens and colleagues elected to use a ramp rate of 0.25°C min^–1 (Stevens et al., 2010), here we used a rate of 0.50°C min^–1. Terblanche and colleagues have shown that higher ramp rates produce higher estimates of CT_max (Terblanche et al., 2007). Differences in ramp rate of about 0.2°C min^–1 resulted in CT_max values that differed by as much as 3.5°C, depending on the start temperature. Therefore, the observed differences in CT_max values reported for T. molitor by Stevens and colleagues (Stevens et al., 2010) and in the present study are to be expected. Nonetheless, it is important to acknowledge that we cannot rule out the possibility that population differences may be a source of variation in the CT_max values reported in the two studies.

We compared three distinct methods for estimating CT_max based on thermolimit data, two of which have been used in prior studies. The ADS method, which is based on identification of the inflection point in the residuals of the absolute difference sum, has previously been favored by authors because of its perceived objectivity. In their original paper describing thermolimit respirometry, Lighton and Turner explain that the inflection point for each ADS corresponds to the point at which spiracular control ceases (Lighton and Turner, 2004). Although this is often the case, we discovered that estimations of CT_max using the ADS method did not consistently identify spiracular failure in this species, particularly in the pre-adult life stages. Therefore, we sought to identify the CSA more directly by measuring CT_max as the point where the V̇_CO2 differential becomes flat. Our results showed that although the ADS method produced slightly lower estimates of CT_max than the CSA method in larvae and pupae, the two methods produced similar, low-variability estimates of CT_max. Importantly, the difference we found in CT_max between adult and pre-adult life stages was evident using either method. Therefore, we conclude that the ADS and CSA methods are both appropriate for analysis of thermolimit data in any life stage.

By contrast, we found that the PMV was not a reliable measure of CT_max in pre-adult life stages of T. molitor. Both larvae and pupae lack a true pre-mortal valley phase, in which CO₂ declines to a smooth valley preceding the post-mortal peak. Although a small decline in V̇_CO2 can be observed in these stages, the lack of a clear nadir at this point results from the fact that spiracular activity continues throughout this phase until the post-mortal peak is reached. Therefore, we found that estimates of CT_max using this technique in the pre-adult phases often preceded the point of spiracular failure by 1–2°C, and hence did not represent a corresponding value to the PMV recorded in adults. We conclude that although the PMV method is useful for analysis in T. molitor adults, it may not be appropriate for comparisons of CT_max across developmental stages.

We examined the point at which death occurred during heating in T. molitor to determine how mortality related to the metrics we used to estimate thermal limitation. We determined percentage survival at each test temperature to identify the range of lethal temperatures, i.e. temperatures at which survival fell between 0 and 100%, in each life stage. Our results showed that the lethal temperature range was closely related to CT_max within each life stage. The range of lethal temperatures was higher in adults than in larvae and pupae, which was consistent with our finding that CT_max was statistically higher in adults than in the earlier life stages. We also found that the vast majority of CT_max values occurred within 1°C of the lethal temperatures measured for each life stage, demonstrating that heat death strongly correlates with CT_max in this species. Additionally, despite differences in V̇_CO2 patterns and CT_max across life stages, we found that the range of lethal temperatures spanned the same phases of the V̇_CO2 pattern in all three life stages, between CT_max and the zenith of the post-mortal peak in most animals. We interpret this to mean that heat death occurs between CT_max and the post-mortal peak during heating in this species, regardless of life stage.

This is the first time that CT_max and mortality have been directly compared in one species under identical experimental conditions. Our results show that under sustained heating regimes that surpass lethal limits, death shortly follows CT_max in these animals. The close association between these two endpoints demonstrates that the cessation of spiracular activity successfully identifies the point of physiological limitation during heating in this insect, rather than a behavior associated with thermal stress.

We conclude that thermolimit respirometry can be used to measure CT_max in different developmental stages of insects exhibiting variable patterns of CO₂ release. CT_max, when measured as the loss of spiracular control, does correlate with lethal temperatures in this species under sustained thermal ramping, and thus represents physiological failure due to heat stress. We caution that the CO₂ release patterns obtained using this method can vary even within a species, particularly across developmental stages, and thus care needs to be taken when interpreting results obtained using thermolimit respirometry. Finally, we confirm that measuring thermal limitation as the cessation of spiracular activity provides an estimate of CT_max that closely approximates the actual point of death.

We thank John Lighton for his guidance with our thermolimit respirometry experiments and analysis of the V̇_CO2 data. We also thank Erica Heinrich and anonymous reviewers for helpful comments on earlier versions of this manuscript.