Development of respiratory function in the American locust Schistocerca americana I. Across-instar effects

The development of physiological traits is a critical aspect of evolutionary and ecological physiology(Burggren, 1991). Most natural selection probably occurs during the juvenile stages, because young animals generally have much higher rates of mortality than fully developed, adult organisms (insects – Estevez and Gonzalez, 1991; reptiles – Iverson, 1991; arachnids– Tanaka, 1992; birds– Currie and Matthysen,1998; fish – Molles,1999; mammals – Durant,2000). Therefore, knowledge of the physiology of juveniles and developmental patterns of physiological traits is important in understanding ecological success and evolution.

Adult and pupal insects tolerate low levels of oxygen (<5 kPa O₂), at which most mammals would die(Wegener, 1993), and generally have resting critical oxygen partial pressures [P_c; the oxygen partial pressure (P_O2) below which metabolism can no longer be maintained] well below those of vertebrates and perhaps below those of non-tracheate invertebrates(Table 1; Hoback and Stanley, 2001). P_c is the point at which an animal switches from being an oxy-regulator to an oxy-conformer (Yeager and Ultsch, 1989; Portner and Grieshaber, 1993), and thus it defines the safety margin for O₂ delivery (the range of P_O2s across which metabolism is constant; Kam and Lillywhite, 1994). For this reason, organisms with low P_cs tend to tolerate hypoxia better than animals with higher P_cs (Ultsch, 1973, 1974). To our knowledge, P_c has not yet been compared across the developmental stages of an insect.

One prominent hypothesis suggests that as insects become larger, gas exchange becomes more difficult due to increases in tracheal system lengths(Graham et al., 1995; Dudley, 1998). This logic is the basis for the hypothesis that increases in atmospheric O₂during the Paleozoic era facilitated insect gigantism(Graham et al., 1995; Dudley, 1998). Alternatively,larger insects could have greater tracheal diameters or could increase the use of ventilation to compensate for increased tracheal lengths. We tested this hypothesis using developing grasshoppers. If older, larger grasshoppers have smaller safety margins (higher P_cs) for oxygen delivery than younger, smaller grasshoppers, this would support the hypothesis of Graham et al. (1995).

Development of the respiratory system is particularly interesting, because all organisms grow in size and increase total oxygen demand with age. P_c changes during ontogeny will depend on whether the ratio of maximal tracheal conductance to metabolic rate changes with age. Maximal tracheal system conductance (G_max; μmol g^–1 h^–1 kPa^–1) will depend on tracheal system structure (e.g. spiracle size and dimensions and density of tracheae and tracheoles) and function (e.g. the magnitude of convective gas exchange). The scaling exponents for metabolic rates during insect development vary widely and may exceed 1.0 (Casey and Knapp, 1987; Vogt and Appel,1999).

In the present study, we quantify developmental changes in metabolic rate,maximal tracheal system conductance and several aspects of tracheal system function (ventilation frequency and abdominal pumping height, an index of tidal volume) for the American locust Schistocerca americana. We also compare the P_c for oxygen consumption and carbon dioxide emission to verify that our P_c for CO₂ emission truly reflected the P_c for oxygen delivery. In addition,we examine the effect of hypoxia exposure time on the P_cfor CO₂ emission, since we wanted to be sure that our calculated P_c values were not strongly affected by the duration of hypoxia exposure in our major experiment.

Schistocerca americana Drury were reared from eggs in culture at Arizona State University as previously described(Harrison and Kennedy, 1994). For these experiments, it was important to have grasshoppers of known age and body size. We took 200 grasshoppers that hatched on the same day and marked them with paint (Testors, Rockford, IL, USA). We monitored the colony each day, and, when an animal molted, we applied a new color of paint. All animals that molted on a given day had the same paint color, so we knew the age in days of each individual. Experimental animals were randomly selected from the animal care facility in the morning and kept in a lit, 35°C incubator with access to green leaf lettuce for 0.5–6 h before measurements were made. We recorded body mass (M_b) of all animals, weighing them to the nearest 0.001 g using an analytical balance (Mettler Instruments,Hightstown, NJ, USA). For measurements of adult insects, we used only males,but for measures of juveniles, sex was not determined.

Adult animals were placed in a 26 ml Plexiglas cylinder sealed with rubber stoppers. Inside the chamber, we placed cotton anterior and posterior to the grasshopper to prevent movement and around the animal's head to prevent visual stimulation. Test animals sat in the chamber for 20 min before measurements were made, a procedure that kept ventilatory frequencies similar to those of unrestrained grasshoppers behind one-way mirrors(Gulinson and Harrison, 1996). Each grasshopper was exposed to 11 different levels of P_O2 (21, 16, 13, 9, 7, 5, 3, 2, 1, 0.5 and 0 kPa O₂) for at least 15 min in descending order to prevent increases in metabolic rate associated with prior hypoxic exposure(Greenlee and Harrison,1998).

Gas mixtures were made by diluting dry, CO₂-free air with N₂ (Balston purge-gas generator; Havervill, MA, USA). The ratio of air to N₂ was controlled by a Brooks 5878 mass flow controller and mass flow meters (Brooks Instruments, Hatfield, PA, USA). Gas mixtures were pushed through three identical chambers at 223–326 ml min^–1. Flow from one of these chambers was directed to the reference cells of the gas analyzers. The sample cells of the gas analyzers received output either from the animal chamber or an empty baseline chamber. All three chambers emptied into 60 ml syringes from which excurrent gases were subsampled. The reference stream and the sample stream were pulled through the CO₂ analyzer (LI-6252; Li-Cor, Lincoln, NE, USA) and then the O₂ analyzer (Sable Systems Oxzilla, Las Vegas, NV, USA) by an AMETEK R-1 flow controller (Pittsburgh, PA, USA). Water and CO₂were removed between the two analyzers using columns of drierite and ascarite. The output of both gas analyzers was digitized and recorded (Sable Systems).

Respirometry chambers, as described above, were placed in a 35°C water bath, the preferred field body temperature for most grasshoppers(Uvarov, 1966). First-, third-and fifth-instar and adult grasshoppers were each exposed to 12 different levels of P_O2 (21, 16.2, 13.2, 9.1, 7.5, 6.2,5.1, 3.8, 2.4, 1.3, 0.7 and 0 kPa O₂) for 3–4 min at each P_O2. We counted ventilatory frequency at each P_O2 while viewing the animals through a dissecting microscope. We calculated CO₂ emission rate using equation 3 above.

We determined individual P_c values for gas exchange by statistically identifying the P_O2 at which CO₂ emission or O₂ consumption dropped(Fig. 1, box A). We compared 95% confidence intervals (CI) created using the mean ±(t×s.e.m.) where t=1.96(Zar, 1999). There were three criteria for designating a P_O2 as critical: the 95% CI at the P_c could not overlap and had to be (1) less than the CI of the previous P_O2(Fig. 1, box B), (2) less than the CI of all the previous P_O2s combined(Fig. 1, box C) and (3) greater than or equal to the CI of all subsequent P_O2s. The mean P_O2 between the P_O2 at which CO₂ emission dropped significantly and the next higher P_O2 was designated as the P_c(Fig. 1, box B; 1.3%O₂). The P_c for abdominal pumping frequency was determined for each individual as the first P_O2at which abdominal pumping frequency dropped below that in normoxia and continued to decrease.

Mean values ± s.e.m. are presented for parametric data,and median values are shown for nonparametric data. Statistical analyses were performed using SYSTAT 10.2, with our within-experiment type I error less than or equal to 5%. For our analyses, the adult instar was designated as instar 7. We used repeated-measures analysis of variance (ANOVA) to compare the hypoxia responses of each instar, since individuals were measured at multiple levels of P_O2; N=8 for all treatment groups. All P_c values were statistically analyzed as nonparametric data, since these are discrete variables. We used the Kruskal–Wallis test, a single-factor analysis of variance in SYSTAT 10.2 and also calculated nonparametric multiple comparisons as described in Zar(1999). To determine whether there was a linear effect of instar on P_c, we also compared those values using nonparametric regression, Kendall's robust line-fit method and rank correlation coefficient(Sokal and Rohlf, 1995).

Body mass (M_b) increased over 50-fold from hatching to early adulthood (Fig. 2). Metabolic rates increased with body mass (metabolic rate=0.005×M_b^0.77; r²=0.95). Third instar CO₂ emission rate was higher than predicted by the general scaling relationship for Ṁ_CO2 and M_b (Fig. 3).

In adult grasshoppers, the P_c for Ṁ_O2 and the P_c for Ṁ_CO2 did not differ significantly (Table 2; Fig. 4; Mann–Whitney U test, U=32, P=1.0). Respiratory exchange ratios remained near one until the P_c was reached, below which they increased dramatically (Fig. 4 inset; repeated-measures ANOVA, F_7,49=6.1, P<0.001).

Duration of exposure (3 min or 1 h) had no significant effect on the P_c of first-instar grasshoppers(Table 2; Fig. 5). Adults exposed for 1 h at each P_O2 had lower CO₂ emission rates than those animals exposed for 3 min, and longer exposures slightly (but significantly) increased the P_c for Ṁ_CO2(Table 2; Fig. 5; Mann–Whitney U test, U=4.5, P=0.002).

There was a significant interaction between P_O2 and instar in the repeated-measures ANOVA,indicating that the decrease in Ṁ_CO2 in response to hypoxia differed in different instars (F27,252=10.5, P<0.001). As has been previously shown(Greenlee and Harrison, 1998),adult grasshoppers exposed to hypoxia were able to maintain constant CO₂ emission down to a median P_O2 of 1.8 kPa (Fig. 6). However,first-instar grasshoppers appeared to be oxygen limited across P_O2 levels lower than the median P_c of 14 kPa, as metabolic rate dropped virtually continuously with P_O2 below 14 kPa (18 kPa for 1 h measures). Third and fifth instars showed intermediate responses, with median P_c values of 9.7 kPa and 4.5 kPa, respectively(Figs 6, 7). Overall, P_c decreased linearly over sevenfold from hatchling to adult grasshopper (Fig. 7;Kendall's rank correlation coefficient, τ=–0.58, P<0.01).

For all instars, breathing frequency plummeted at some low level of ambient oxygen (Fig. 8). In contrast to the P_c for Ṁ_CO2, the P_c for abdominal pumping was very low across all instars(Table 2; Fig. 7). However, there was still a significant effect of instar on the P_c for abdominal pumping (Kruskal–Wallis test statistic=14.3, P<0.01). Nonparametric multiple comparisons revealed that only third instars and adults differed in their P_c for abdominal pumping, with the third-instar P_c threefold higher than that of adults (q=4.7, P<0.05).

In response to hypoxia, adults more than doubled their ventilatory frequencies, while the first instars showed no change(Fig. 8). The response of abdominal pumping frequency to hypoxia varied significantly with instar, as indicated by a significant interaction between P_O2 and instar (repeated-measures ANOVA, F27,252=10.5, P<0.001). We used analysis of covariance(ANCOVA) to determine which instars had different responses to hypoxia based on the slope of the line created when abdominal pumping was regressed on P_O2 at all P_O2s above the P_c for abdominal pumping (21 kPa to 3 kPa). We limited our test to these P_O2s because, below the P_c, abdominal pumping decreased dramatically. This analysis revealed that the first-instar grasshoppers had no response to hypoxia above the P_c (slope=–0.114, P=0.9; Fig. 8). Pair-wise comparisons of the slopes of the other instars indicated that all were different from the first-instar slope but not different from each other (first instar versus other instars: all P≤0.011; comparisons between third, fifth and adults: all P>0.6). Thus, only the first-instar grasshoppers did not increase breathing frequency in response to hypoxia, and the responsiveness of abdominal pumping frequency to hypoxia was similar from the third instar to the adult stage [abdominal pumping frequency=2.2(P_O2)+42.2; r²=0.12; Fig. 8].

Our index of tidal volume, percent change in abdominal height, was 35 times higher in normoxia for adults compared with first instars(Fig. 9). Juveniles compressed their abdomens less than 1% of maximal height, compared with a 20% abdominal compression in adults (Fig. 9). Because we measured tidal volume at different P_O2s for each instar, we were able to statistically compare only the responses at common P_O2s (21, 5 and 1 kPa). There was a significant interaction effect of instar and P_O2 on tidal volume (repeated-measures ANOVA, P_O2 ×instar, F_2,28=64.8, P<0.001). Adults increased tidal volume almost 50% in response to hypoxia, while juveniles showed no significant change in tidal volume (repeated-measures ANOVA, adults – F_4,28=21.8, P<0.0001; first instar – F_5,35=1.4, P=0.24).

Throughout development, the grasshopper respiratory system improved in its ability to respond to hypoxia, as indicated by the increased responsiveness of ventilation frequency (Fig. 8)and tidal volume (Fig. 9) to hypoxia; together, these lead to a fourfold increase in the mass-specific capacity of the tracheal system to conduct oxygen (G_max; Fig. 10). Older grasshoppers were also much more able to tolerate hypoxia, as the P_cfor metabolic rate decreased sevenfold during ontogeny(Fig. 7). The increased tracheal system conductance with age may have been in preparation for a switch in the principal mode of locomotion from hopping to flight. In addition, the increased responsiveness to hypoxia of the tracheal system of older grasshoppers may be related to a general increased use of convective gas exchange in larger insects. Furthermore, our finding that development strongly affected the ability of grasshoppers to respond to ventilatory system challenges may mean that juveniles would be more sensitive to hypoxic environments encountered within restricted burrows, at high altitudes or during discontinuous gas exchange(Lighton, 1996).

We used CO₂ emission as our index of metabolic rate in the across-instar comparisons because we could measure fractional CO₂content of the excurrent air stream quickly and more accurately than if we measured O₂, especially for the smallest grasshoppers. This approach raises the concern that CO₂ emission might not reflect the P_c for aerobic metabolism. For example, the P_c for Ṁ_CO2 might be lower than the P_c for Ṁ_O2 if Ṁ_CO2 was elevated at low P_O2s due to anaerobiosis or CO₂ washout from body tissues due to hyperventilation. In our study, despite persistent elevation of Ṁ_CO2 at very low P_O2s, adult grasshoppers had identical P_c values for Ṁ_CO2 and Ṁ_O2(Table 2; Fig. 4). The P_c for Ṁ_CO2 is probably also close to the P_c for Ṁ_O2 in juveniles,because juveniles exhibited much lower ratios of CO₂ emission in anoxia to CO₂ emission in normoxia(Fig. 11; ANOVA F_3,28=15.2, P<0.001). Therefore, we conclude that the P_c for Ṁ_CO2 provided a reasonable approximation of the P_c for aerobic metabolism.

Interestingly, the RERs that we measured(Fig. 4 inset) were higher than those measured for resting Schistocerca gregaria (0.82; Krogh and Weis-Fogh, 1951). One possible explanation for the discrepancy is that fuel use was different. Our grasshoppers were fed on green leaf lettuce, kale and dried wheat germ,whereas the grasshoppers in the prior study were fed green cabbage and dried grass. Alternatively, our grasshoppers could have been in a stressed state. However, the normoxic, acute metabolic rates we measured (after 20 min chamber acclimation time) were similar to those measured over 1 h, indicating that grasshoppers either were still stressed after 1 h or were settled within 20 min. Typically, grasshopper ventilation frequencies return to normal 20 min after a disturbance (Gulinson and Harrison, 1996). Perhaps the difference was due to inherent differences between the two species.

We were also concerned that the use of short-term exposures to hypoxia would not reflect steady-state conditions. However, calculated P_c values for Ṁ_CO2 did not differ between first instar animals exposed for three minutes or one hour to the test P_O2s(Table 2; Fig. 5). Adult grasshoppers exposed to hypoxia for 1 h did have much lower CO₂ emission rates and a slightly higher P_c compared with acutely exposed animals (Fig. 5;Table 2). We attribute the elevated Ṁ_CO2 for the 3 min exposures in adults to transient increases in CO₂ emission that occurred in response to each drop in atmospheric P_O2 (Fig. 1). The increased CO₂ emission may be attributed to two events. First, it has been noted that grasshoppers may physically struggle to escape when exposed to lower oxygen(Hochachka et al., 1993; Wegener, 1993). Secondly,exposure to hypoxia causes an increase in ventilation (enhancing CO₂ loss) relative to metabolic rate (CO₂ production),which secondarily causes a drop in hemolymph P_CO2 and total CO₂ content(Greenlee and Harrison, 1998). An alternative explanation for the transient increase in CO₂emission is that acid metabolites, such as lactate, accumulated in the tissues as P_O2 decreased(Hochachka et al., 1993). Acid metabolite accumulation would decrease hemolymph pH and hence increase internal P_CO2, enhancing CO₂elimination. However, it is unlikely that acid metabolites accumulated in the hemolymph because we have direct evidence that hemolymph pH increases during hypoxia and bicarbonate concentration is higher than expected at lower P_O2s(Greenlee and Harrison, 1998). Thus, the washout of CO₂ from the hemolymph and tissues as the animal moved towards a new, lower, steady-state internal P_CO2 is likely to be the major cause of the transient increase in Ṁ_CO2 upon exposure to hypoxia. It is worth noting that juveniles showed no transient increase in Ṁ_CO2upon exposure to hypoxia (Fig. 5), again consistent with our conclusion that they lack a ventilatory response to hypoxia.

Interestingly, a transient rise in CO₂ emission upon exposure to a lower P_O2 is observable even at relatively high P_O2s where ventilation rates were unaffected by hypoxia (Fig. 1). One possibility is that ventilation increased transiently in response to hypoxia and then lowered in response to reduced internal P_CO2 levels by the time we measured it. Alternatively, the transient rise in Ṁ_CO2 could be due to transient changes in nonventilatory mechanisms for increasing tracheal conductance (e.g. increasing spiracular opening or decreasing tracheolar fluid levels).

Critical points are commonly determined in physiological research, and a variety of methods have been proposed for statistical analysis of critical point data (Nickerson et al.,1989; Yeager and Ultsch,1989). However, in our study, and many others, critical points were determined by exposing individuals to a range of environmental conditions, an experimental design that should be analyzed using repeated-measures approaches (Potvin et al., 1990). Unfortunately, published methods for determining P_cs have ignored the statistical problems with a repeated-measures design (Yeager and Ultsch, 1989). In the present study, we developed a method that allowed us to assign a P_c value to each animal and then use this as a data point for further statistical analysis. This method may be generally useful for investigators who wish to compare critical points across groups of animals in studies where the parameters are repeatedly measured for each animal.

Our technique could be criticized because the animal's resting Ṁ_CO2 value has a strong effect on the P_c, and visual inspection of Fig. 6 shows a very large drop in Ṁ_CO2 for all instars at ∼2–3 kPa. To address this issue and compare several methods of determining P_c, we first analyzed the individual data files and identified the P_O2 at which Ṁ_CO2dropped below the 75th, 50th or 25th percentile of normoxic Ṁ_CO2(Fig. 7, lines C, D and E,respectively). Of these lines, the 75th and 50th percentiles result in a significant decrease of P_c with instar (Kendall's rank correlation coefficient, 75th – τ₃₂=–0.65, P<0.01; 50th – τ₃₁=–0.32, P<0.01). Only 14 animals had Ṁ_CO2 that dropped below 25% of normoxic metabolic rate and, of these, eight were first-instar juveniles. There was no effect of instar on these P_cvalues (median=0 kPa). We also used paired t-tests to determine the first P_O2 at which mean Ṁ_CO2 decreased significantly below the mean Ṁ_CO2 in normoxia(Fig. 7, line F). P_c determined in this way decreased approximately sixfold with instar. The fact that most of these methods yield linear decreases of P_c with age supports our conclusion that P_c falls with age/size in S. americana. However,the fact that there is no age effect on the P_O2at which Ṁ_CO2drops to 25% of its normoxic rate suggests that even the youngest grasshoppers may be able to sustain some aerobic metabolism at very low atmospheric P_O2. Alternatively, the very low Ṁ_CO2 measured at low air P_O2 may represent anaerobic metabolism.

In general, as grasshoppers grew, their hypoxia tolerance increased, as shown by the decreased P_c for Ṁ_CO2 with age. This pattern is similar to that seen in the lobster Nephrops norvegicus, where adults have a lower P_c for O₂ consumption compared with larvae(Spicer, 1995). Adult and older juvenile crayfish, Procambarus clarkii, respond to hypoxia by increasing ventilation frequency, whereas larval ventilation frequency decreases in hypoxia (Reiber,1997). To our knowledge, P_c has not been compared throughout ontogeny for any other insect species, so it is unclear whether this pattern will prove to be widespread throughout the arthropods.

Adult grasshoppers exposed to hypoxia were able to maintain constant CO₂ emission down to a P_O2 of 1.8 kPa (Figs 6, 7). Adult grasshoppers had such a low P_c because they increased both breathing frequency and abdominal pumping height in response to hypoxia (Figs 6, 7, 8, 9). Together, the increases in abdominal pumping frequency and tidal volume resulted in a fourfold increase in G_max from first-instar to adult grasshoppers(Fig. 10; ANOVA effect of instar, F_1,30=31.6, P<0.001). Interestingly,while G_max increased strongly with age when using Ṁ_CO2 measures and P_c values from the 3 min exposures to hypoxia, G_max was similar for first instars and adults when analyzed using 1 h exposures (Fig. 10). This pattern occurred because the 1 h exposure to hypoxia substantially decreased Ṁ_CO2 in adults but not first instars (Fig. 5)and P_c increased more dramatically in adults(Table 2). The physiological bases to the disparity between the 1 h and 3 min patterns for G_max are unknown without information on internal gas level variation during ontogeny. Based on our present data, the most likely explanation seems to be that Ṁ_CO2 in adults exposed to short-term hypoxia was elevated by changes in internal P_CO2 that did not occur in juveniles (Figs 5 and 11 suggest greater CO₂ washout occurs during hypoxia for adults). When CO₂emission occurs at rates equivalent to metabolic CO₂ production (as presumably happens when Ṁ_CO2 is measured over 1 h), G_max is size invariant. This suggests that first instars were able to achieve similar gas exchange rates despite lower tidal volumes and ventilatory frequencies, presumably because their smaller size enhances diffusive gas exchange.

In contrast to the pattern seen in older grasshoppers, first-instar animals had no apparent ventilatory response to hypoxia. They showed no change in ventilation frequency (Fig. 8)or tidal volume (Fig. 9), and Ṁ_CO2 dropped,though not always significantly, with every decrease in atmospheric P_O2 (Fig. 6). Conceivably, first instars could have responded to hypoxia by increasing diffusive gas exchange (e.g. opening spiracles or removing tracheolar fluid). To test this idea, we calculated the ratio of Ṁ_CO2 (μmol min^–1) to abdominal pumping frequency. If younger grasshoppers were using more diffusive gas exchange, then the μmol CO₂ g^–1 breath^–1 should be higher in first instars and should increase during hypoxia. Indeed, in first-instar grasshoppers, this index of gas exchange per breath was higher than that in older instars (Fig. 12),supporting the hypothesis that first instars were more reliant on diffusion than older grasshoppers. Alternatively, first instars could simply have had higher internal and expired P_CO2 levels. However, there is no evidence that first instars responded to hypoxia by opening spiracles or removing tracheolar fluid, as the μmol CO₂g^–1 breath^–1 was constant as air P_O2 decreased until P_O2s dropped below the P_cfor abdominal pumping (Fig. 12).

Why do first-instar grasshoppers lack a ventilatory response? One hypothesis is that younger grasshoppers have not developed the complete neural circuitry for responding to hypoxia. This idea is supported by the findings of Miller and Mills (1976), who noted that first- and second-instar S. gregaria also lack synchronization between spiracular valve activity and abdominal pumping. The adult pattern of synchronization between abdominal pumping and spiracular valve activity appears sporadically in the third instars(Miller and Mills, 1976). In our study, third-instar animals demonstrated an adult-like ventilatory response to hypoxia (Fig. 8). Alternatively, first-instar grasshoppers may have not yet developed the capacity to sense oxygen. This hypothesis could be tested by looking for members of the HIF-1α oxygen-sensing cascade(Lavista-Llanos et al., 2002)and comparing the expression patterns between instars. Interestingly,expression of HIF-1α and HIF-1β has been found to decrease with juvenile development in mice, the opposite of what we would predict for grasshoppers (Madan et al.,2002). Finally, first-instar grasshoppers may lack the large air sacs found in older animals, reducing the capacity for convective gas exchange.

The decrease in mass-specific metabolic rate with size is well known among animals (Schmidt-Nielsen,1984; West et al.,1997). However, the increase in mass-specific gas-exchange capacity with age was less expected (Fig. 10). One possibility is that the increased tracheal system conductance with age in these grasshoppers facilitates flight, an adult activity that requires a 40–50-fold increase in O₂consumption above basal rates (Krogh and Weis-Fogh, 1951). Additionally, larger insects may generally be more able to cope with hypoxia because they have evolved mechanisms, such as abdominal pumping, to facilitate convective gas exchange, as may be necessary for adequate oxygen delivery in large insects(Weis-Fogh, 1964).

In contrast to the decrease in P_c for Ṁ_CO2 throughout ontogeny, the P_c for abdominal pumping was very low across all instars. At each instar, abdominal pumping continued even after Ṁ_CO2 began to drop. The large disparity between the P_c for Ṁ_CO2 and the P_c for ventilation frequency strongly suggests that insects are able to preferentially shut down certain body systems in response to hypoxia. Additionally, O₂ delivery to all parts of the respiratory system may be better than to other parts of the body, since the tracheal supply to the pacemaker neuron is substantial.

Interestingly, when we calculated P_c as the P_O2 at which Ṁ_CO2 was lower than 50% of the normoxic rate, this line overlapped the P_cfor abdominal pumping (Fig. 7). Does this correlation mean that the P_c for abdominal pumping represents the `true' critical point for oxygen delivery? This strongly depends on what is meant by `true' critical point. Any significant drop in Ṁ_CO2could be due to either a direct oxygen limitation of chemical reactions or to an oxygen-sensing mechanism that reduces oxygen-consuming behavioral or physiological functions (for example, hypoxia induces fatigue in humans under conditions in which oxygen is not believed to be limiting to muscle mitochondrial function). It is likely that animals have several P_c values that vary with the function being observed; an exciting test would be to determine the P_O2 at which other functions, such as feeding, protein synthesis or locomotion, are affected.

In this study, we partially tested the idea that larger insects have more difficulty with gas exchange. A negative scaling of the safety margin for oxygen delivery is one possible prediction derived from the hypothesis that atmospheric hyperoxia in the late Paleozoic facilitated insect gigantism(Graham et al., 1995; Dudley, 1998). Clearly, the results from this study provide no support for this prediction. In fact, our finding that smaller grasshoppers had a much higher P_c for gas exchange (Figs 6, 7) was exactly the opposite of what would be predicted if larger insects had decreased safety margins for gas exchange. However, our experiments did not distinguish between the effects of body size and development; to do so, the safety margin for oxygen delivery must be compared across individuals of different species at the same developmental stage. Additionally, a better test of whether gas exchange becomes more challenging for larger insects may be to examine the scaling of P_c for gas exchange for insects at their maximal Ṁ_O2. Indeed, Erythemis simplicicollis dragonflies are oxygen limited during flight in normoxia,evidence that safety margins for oxygen delivery are small during activity(Harrison and Lighton,1998).

It is generally thought that larger insects must use convective gas exchange to achieve adequate oxygen delivery(Weis-Fogh, 1964), and our data strongly suggest that this is true within the developmental stages of S. americana (Figs 8, 9, 13). Therefore, an alternative prediction derived from the hypothesis that larger insects have more difficulty with gas exchange is that larger insects may use compensatory mechanisms to overcome diffusion limitations. If so, we might expect the largest extant insects to exhibit maximal use of convective gas exchange. Increased use of convection might allow larger insects to have more responsive respiratory systems and control over diffusive water loss(Kestler, 1985). Examining the scaling of air sac volumes and use of convection across insects of various sizes would allow testing of this hypothesis.

Support for this project was provided by NSF IBN-9985857 to J.F.H., NSF IBN-0206678 to K.J.G. and J.F.H., and a Sigma Xi GIAR to K.J.G. We would also like to thank three anonymous reviewers, Michelle Fay, Jeffrey Hazel, Joanna Henry, Paul Kestler, Michael Quinlan, Brenda Rascón, Ron Rutowski,Glenn Walsberg, Art Woods and especially Scott Kirkton for helpful comments on this manuscript.