Acid–Base and Respiratory Responses to Hypoxia in the Grasshopper Schistocerca Americana

It is well known that quiescent insects can maintain constant rates of oxygen consumption in extremely low oxygen concentrations. For example, the critical values (the below which oxygen consumption begins to fall) for some resting insects are as follows: Tenebrio molitor pupae, <5 kPa (Gaarder, 1918); adult Aedes aegypti mosquitoes, 3–4 kPa (Galun, 1960); adult Phormia regina flies, 2–5 kPa (Keister and Buck, 1961); adult Termopsis navidensis termites, 2–5 kPa (Cook, 1932) and adult Locusta migratoria, 3–4 kPa (Arieli and Lehrer, 1988). While these data demonstrate that the safety margin for oxygen delivery in resting insects is large, the physiological mechanisms responsible for this remain unclear. One hypothesis is that the conductance (the quantity of gas transferred divided by the partial pressure gradient) of the tracheal system is so high at rest that no response is needed to hypoxia. Alternatively, insects may need to increase the conductance of the tracheal system in proportion with the fall in atmospheric oxygen. If tracheal conductance does increase, by what mechanisms is this accomplished? We investigated these questions by testing the effects of hypoxia on the tracheal physiology of the grasshopper Schistocerca americana.

Together, the tracheal morphology, mechanisms of gas exchange and the neural control of the ventilatory system have been better studied in grasshoppers than in any other insect. Large longitudinal trunks run along each side of the animal connecting all ipsilateral spiracles and branching into a system of air sacs and secondary and tertiary tracheae, further branching into tracheoles which are the sites of gas exchange in the tissues (Weis-Fogh, 1964, 1967). Convective gas exchange in non-flying grasshoppers is accomplished mostly by abdominal pumping, which includes both dorso-ventral contractions and longitudinal telescoping movements (Miller, 1960a; Weis-Fogh, 1967). Abdominal pumping is initiated by a pacemaker in the metathoracic ganglion (Miller, 1960a; Hoyle, 1959) and is synchronized with spiracular opening so that inspiration occurs through the first four pairs of spiracles and expiration through the last six pairs of abdominal spiracles, producing a largely unidirectional flow of air through the grasshopper (McCutcheon, 1940; Weis-Fogh, 1967). Abdominal pumping is stimulated by hypoxia (Miller, 1960a), but ventilatory frequency is reported not to be stimulated by hypoxia until the falls below the point at which the rate of oxygen consumption drops (Arieli and Lehrer, 1988). However, when tracheal is varied at constant tracheal ⁠, ventilatory frequency is inversely related to tracheal ⁠, and evidence suggests that resting grasshoppers in room air actively maintain tracheal at approximately 18 kPa (Gulinson and Harrison, 1996).

The complex respiratory system of an insect can be viewed as two steps in series: (1) a trans-spiracular step, between the large trachea and air across the spiracles; and (2) a tracheolar step, between the secondary and tertiary tracheae and the cells (Harrison, 1997). The conductance at each step can be estimated from the gas flow divided by the gradient for gas diffusion at that step. Technical problems make measurements of the O₂ gradients difficult, but techniques for the measurement of haemolymph and tracheal and CO₂ emission are now established (Harrison, 1988; Gulinson and Harrison, 1996). In the present paper, we investigate the effect of hypoxia on the conductance of the trans-spiracular and tracheolar steps for CO₂ as a first step in understanding the tracheal responses to hypoxia. To address the mechanisms responsible for variations in spiracular conductance, it is necessary to be able to quantify the bulk flow of air accomplished by abdominal pumping. This has been done by Weis-Fogh (1967), who measured the flow of air from the most terminal abdominal spiracles after sealing all the other abdominal spiracles with wax. This procedure is quite invasive and provides poor temporal resolution. We have, therefore, developed a new optical method for quantifying insect convective ventilation from measurements of abdominal volume changes.

Schistocerca americana Drury were reared from eggs in culture at Arizona State University as previously described (Harrison and Kennedy, 1994). We used only adult females at least 1 week past the last moult in all experiments, as their larger abdomens and greater haemolymph volumes facilitated our measurements. All animals were unfed but provided with water for at least 12 h prior to measurements. Air temperature was 25.4 °C (range 24.4–26.3 ºC). Immediately prior to all experiments, the wings were clipped 2 mm from the base to provide a clear view of the abdomen, and the animals were weighed.

Metathoracic spiracles were cannulated using a 30 mm section of heat-stretched Intramedic PE-60 tubing (Clay Adams, Parsippany, NJ, USA) sealed to the cuticle with hot glue (Gulinson and Harrison, 1996). Animals were placed in a 10 cm length of Plexiglas tubing which served as the respirometry chamber (18.25 mm i.d., 26 ml volume); the spiracular cannula exited the chamber via a small hole drilled in the chamber wall. We restrained grasshoppers in the chamber by tightly enclosing the head of the animal with cotton and a metal mesh helmet anchored in the rubber stopper which sealed one end of the chamber. This prevented the animals from moving, but enabled the abdomen to pump freely. The cotton covering the eyes prevented visual input, which helped keep the grasshopper calm and kept ventilatory frequencies statistically identical to those of unrestrained grasshoppers (Gulinson and Harrison, 1996).

Gas mixtures were generated with a Brooks 5878 mass flow controller and Brooks mass flow meters (Brooks Instruments, Hatfield, PA, USA). Gas mixtures (21,10, 5 or 2 kPa O₂, balance N₂) were perfused through the chamber at a rate of 200 ml min⁻¹. Partial pressures of air were calculated from the percentage of oxygen in the mixture and the total barometric pressure, which was measured daily to the nearest 13 Pa. After 30 min of exposure to a gas mixture, two 30 μl tracheal gas samples were withdrawn sequentially from the spiracular cannula using a 100 μl Hamilton gas-tight syringe (Gulinson and Harrison, 1996). We injected these samples into a Varian 3400 gas chromatograph and gas chrom MP1 column (Varian Analytical Instruments, Walnut Creek, CA, USA) for analysis of CO₂, as previously described (Gulinson and Harrison, 1996). We then quickly removed the rubber stopper and attached grasshopper from the chamber and, within 5 s, a small incision was made in the ventral neck cuticle and a 50 μl haemolymph sample was taken using a 100 μl Hamilton gas-tight syringe, as described previously, keeping CO₂ loss statistically insignificant (Harrison, 1988; Gulinson and Harrison, 1996). We analyzed 45 μl of this sample for total carbon dioxide content (⁠⁠, mmol l⁻¹) using a Varian 3400 gas chromatograph as described by Boutilier et al. (1985).

The pH of the remaining 5 μl of haemolymph was measured using a new technique for the measurement of pH in small samples without CO₂ loss. Haemolymph samples were expelled from the gas-tight syringe into a 4 mm length of PE-50 tubing placed snugly on the end of a KCl–agar bridge made of PE-10 tubing held in place by a clamp attached to a micromanipulator. We constructed Hinke-type pH-sensitive electrodes using methods described by Harrison et al. (1990). We viewed the pH microelectrode through a microscope and lowered it into the haemolymph sample using another micromanipulator. The voltage difference between the haemolymph and the calomel reference electrode was measured using a Biologic IS-100 dual electrometer and amplified using a Biologic VF 102 dual microelectrode amplifier (Biologic, France).

We exposed grasshoppers to test gases (21, 10, 5 and 2 kPa O₂, balance N₂) in ascending or descending order for 40 min. During the last 10 min, we concurrently measured the carbon dioxide emission rate (⁠⁠, μmol g⁻¹ h⁻¹) and convective ventilation (ml min⁻¹). Grasshoppers were secured in the respirometry chamber exactly as described above, except that animals were not cannulated, and no hole was drilled in the chamber.

Gas mixtures were drawn through the chamber and through a magnesium perchlorate column (to remove water), a LI-6252 CO₂ analyzer (Li-Cor, Lincoln, NE, USA), an Ascarite column (to remove CO₂) and an AMETEK S3A/I oxygen analyzer (AMETEK, Pittsburgh, PA, USA) at 147 ml min⁻¹ ATP by a downstream AMETEK R-1 pump. The output of the gas analyzers was digitized and recorded (Sable Systems, Las Vegas, NV, USA). We calculated the CO₂ emission rate by multiplying the flow rate and the expired CO₂ fraction since incurrent CO₂ was zero and the volumes of CO₂ and water vapour produced by the grasshopper were negligible relative to the flow rate of air through the chamber. Water production for grasshoppers at this temperature is reported to be 0.0147 ml min⁻¹ (Loveridge and Bursell, 1975; Prange, 1990); CO₂ production was measured in this study and was less than 0.4 % of the flow rate (see Results).

The Plexiglas respirometry chamber was placed within an optical bench for measurement of the volume of air moved by abdominal pumping (Fig. 1). The respirometry chamber was surrounded by an array of three silicon solar cells (Radio Shack, USA). Three 12 V 55 W quartz–halogen H3 light bulbs (Blazer International Corp., Franklin Park, IL, USA) were mounted at different heights and each was powered by its own low-noise, single-output linear open-frame direct-current power source (Sola 12 V, 10 A regulated supply, Newark Electronics, Chicago, IL, USA) with line regulation (0.1 % for 10 % change), load regulation (0.1 % for 90 % change) and ripple control (0.1 % P−P_max PARD). The light beams were diffused using convex lenses (Edmund Scientific, Barrington, NJ, USA) and polarized and directed by mirrors towards the solar cells. One light beam travelled from directly above the animal, casting a ‘top-view shadow’ of the abdomen onto a solar cell placed directly beneath the chamber. A second light beam travelled horizontally towards the animal, casting a ‘side-view shadow’ of the abdomen onto a solar cell placed vertically beside the chamber. The third solar cell was placed at 45 ° to the other two solar cells and was used to record the length of the abdomen. The light beam for length recording travelled through a metal plate into which was machined a 2 mm×10 mm rectangular slit, then onto a mirror, and finally past the grasshopper to the solar cell, which was covered with a piece of cardboard with a slit dimensionally identical to the one in the metal plate. We mounted the lights, mirrors and solar cell array on a 61 cm×30.5 cm optical bench (Edmund Scientific, Barrington, NJ, USA; Fig. 1) and covered the entire structure with an opaque box to prevent interference from outside light.

Mean values ± S.E.M. are shown throughout. Unless otherwise stated, statistical analysis was performed using SYSTAT (Wilkinson, 1989), with our within-experiment, type I error less than or equal to 5 %. Haemolymph acid–base status and tracheal gas data were analyzed by linear regression (GLM, general linear model), since different individuals were used and atmospheric was manipulated in a graded fashion. Plots of residuals versus ambient yielded lines with slopes and intercepts equal to 0, indicating that no assumptions of linear regression were violated (Sokal and Rohlf, 1995). Regression analyses performed including terms were non-significant, lending further support for the use of linear regression. Ventilation data were analyzed using repeated-measures analysis of variance (ANOVA), since individuals were measured at multiple values.

Haemolymph pH became increasingly alkaline as ambient decreased from 21–2 kPa (Fig. 5). The total CO₂ in the haemolymph did not change significantly with reductions in inspired (Fig. 5). The calculated value for haemolymph decreased nearly threefold when ambient decreased 10-fold (Fig. 5). The variation in tracheal appeared to decrease as inspired decreased; however, further statistical analysis showed no significant heterogeneity in variances (Scheffe box test; Sokal and Rohlf, 1995; F_3,8=1.88, 0.05<P<0.1). Tracheal also decreased strongly in response to hypoxia (Fig. 6), and the gradient from haemolymph to trachea, calculated within individuals, was more than halved by hypoxia (Fig. 6).

Ventilatory frequency, tidal volume and convective ventilation were not affected by the order of exposure to test gases as analyzed by a multi-factor ANOVA (Table 1). Since the order of test gas exposure did not affect ventilatory frequency, tidal volume or convective ventilation, data from the ascending and descending protocols were pooled for the subsequent statistical analysis of these variables. In response to a 10-fold reduction in ambient ⁠, ventilatory frequency more than doubled, tidal volume did not change significantly and convective ventilation increased significantly by a. proximately fourfold (Table 1; Fig. 7).

(μmol g⁻¹ h⁻¹) was affected by the order of exposure to gases and by (Table 1; Fig. 8). When analyzed separately, neither ascending nor descendin.g protocol animals experienced a significant change i.n in response to decreased ambient ⁠. While mean at 21 kPa was significantly higher in animals with prior exposure to hypoxia (F. ig. 8, Tukey HSD multiple comparisons, P=0.001), mean at all other values did not differ between animals exposed to ascending or descending levels of (Fig. 8, Tukey HSD multiple comparisons; 10 kPa, P=1.0; 5 kPa, P=0.61; 2 kPa, P=0.97).

However, two of the animals had very high calculated expired values relative to the other animals (6–7 kPa in room air) and low conductances and were, therefore, excluded from further statistical analyses of expired and both conductance variables on the basis of Dixon’s test for outliers (Sokal and Rohlf, 1995). According to this test, these two animals were not outliers for any of the other variables measured (ventilatory frequency, tidal volume, convective ventilation and ⁠), so these animals were not excluded from any other statistical analyses. With these two outliers removed, expired decreased strongly as ambient decreased, with expired at 2 kPa being 27 % of the value for animals in room air (Fig. 9B). Spiracular conductance increased nearly fourfold over the 10-fold decrease in ambient (Fig. 10). Mean tracheolar conductance nearly doubled as inspired decreased, although the effect was not significant by ANOVA (Fig. 10).

The effects of ambient on calculated expired and measured tracheal were compared using analysis of covariance (ANCOVA) since there was no significant difference in the slopes of these lines (Fig. 9A, GLM, F_1,70=2.62, P=0.110). The intercepts of these lines also did not differ (ANCOVA, F_1,69=0.21, P=0.65), indicating that these independent measurements of tracheal yield similar values. The slope of the haemolymph versus ambient regression line differed from slopes of regression lines calculated for expired and tracheal versus ambient (Fig. 9A; GLM, haemolymph versus expired ⁠, slope effect F_1,70=3.97, P=0.05; GLM, haemolymph versus tracheal ⁠, slope effect F_1,64=6.79, P=0.011).

Convective ventilation is one of the most important parameters used in investigating any respiratory system. Our optical method of recording abdominal volume changes allows the first continuous estimates of tidal volume and convective ventilation in an unimpeded tracheal system.

Because of the chamber design, we were able simultaneously to perform respirometry and manipulate ambient gases. This method should be widely applicable to insects of various sizes and shapes and may, perhaps, be used during tethered flight if the animal can be restrained in the chamber and the size of the insect relative to the photo-sensing system is maintained. We have successfully measured convective ventilation in a juvenile S. americana grasshopper (0.16 g, abdomen length 9.51 mm) within our apparatus, demonstrating that this apparatus could be used without alteration over an order of magnitude size range. However, to record volume changes of an insect with differing abdominal morphology, the shape of the abdomen must be assessed, and the relationship between shadow area and abdominal volume determined empirically.

We were not certain that the dimensional changes in abdominal volume during breathing would match the dimensional variation among individuals under dynamic conditions, an assumption implicit in our calibration method. In addition, our method ignores ventilation due to movements of the neck and prothoracic regions, which, according to Miller (1960a), can account for up to 14 % of convective ventilation. It is also important to note that our technique actually measures abdominal volume changes, not convective ventilation. Conceivably, there might not be either sufficient time between breaths or a low enough spiracular resistance (especially at high ventilatory frequencies) for convective ventilation to match abdominal volume changes. However, the close match between calculated expired and measured tracheal across a variety of ventilatory frequencies (Fig. 9B) strongly supports the hypothesis that our optical method accurately measured convective ventilation. At 21 kPa O₂, our measurement of convective ventilation of 0.78±0.15 ml min⁻¹ (mean ± S.E.M., Fig. 7, mean mass for the S. americana used in this study was 1.04 g) was very similar to that measured by Weis-Fogh (1967, 0.7 ml g⁻¹ min⁻¹) for S. gregaria using a completely different method, further supporting the accuracy of this method. The response of ventilatory frequency to hypoxia was also similar in this study to that found by Arieli and Lehrer (1988).

Our calculations of spiracular and tracheolar conductances have a number of assumptions and limitations which should be admitted. For our measurements of tracheolar conductance of ⁠, we assume that haemolymph is representative of cellular ⁠. One would expect cellular to be higher than haemolymph ⁠, since cells are the site of CO₂ production. In addition, CO₂ produced in the cells may move via the haemolymph to enter the tracheal system at other points. To the extent that haemolymph underestimates cellular ⁠, we are overestimating tracheolar CO₂ conductance. However, since carbonic anhydrase occurs in cells but not in haemolymph for most insects (Darlington et al. 1985), CO₂ exchange is likely to occur between the cells and tracheae but not between the haemolymph and tracheae, especially during dynamic conditions. Supporting this assumption, calculated whole-body is similar to haemolymph for resting grasshoppers (Harrison, 1988).

For our calculation of spiracular conductance, we used expired values calculated according to equation 5, which were very similar to tracheal values measured directly via the metathoracic spiracle (Fig. 9). We suspected that expired would be higher than the measured from thoracic tracheal samples, since inspiration occurs via the thoracic spiracles. It is possible that cannulation impeded inspiration, elevating tracheal to levels equivalent to expired of uncannulated animals. However, cannulation does not affect resting ventilatory frequency (Gulinson and Harrison, 1996), suggesting that this is not the case. Our data suggest that the geometry of the insect tracheal system is such that thoracic and expired do not differ appreciably in resting grasshoppers except, perhaps, under conditions of very high convective ventilation (see below).

Is it possible that our 30 min of exposure to test values did not allow us to examine steady-state conditions? Total tracheal volume is 500–950 μl in Schistocerca gregaria (Weis-Fogh, 1964; Harrison, 1989). Assuming tidal flow and complete mixing, at the slowest convective ventilation rate (21 kPa: 780 μl min⁻¹) and the largest possible tracheal volume (950 μl), the system would be 99 % equilibrated in 5.8 min. These assumptions are conservative, since the system is not tidal (Miller, 1960b), but at least partially unidirectional, which would cause complete flushing to occur considerably faster. Thus, the 30 min equilibration time should have been more than ample to allow tracheal levels to reach a steady state. Since the haemolymph pool is large (approximately 500 μl in adult female S. americana; Harrison, 1988) and haemolymph lacks carbonic anhydrase, the slowest-reacting CO₂ pool would be the haemolymph, and it is possible that haemolymph would not reach a steady state after 30 min even if tracheal were equilibrated. Preliminary experiments indicated a transient elevation in the rate of carbon dioxide emission upon exposure to hypoxia, consistent with a small decrease in animal CO₂ content in response to hypoxia.

However, visual inspect.ion of the traces indicated that such elevations in disappeared within minutes, suggesting that the washout of CO₂ from the grasshopper is rapid and complete in well under 30 min.

Hypoxia causes a pronounced alkalization of the haemolymph (Fig. 5), which is partly due to the reduction in haemolymph associated with the hyperventilation induced by hypoxia (Figs 5, 6, 7). However, a Davenport analysis (Fig. 11) demonstrates that haemolymph pH and bicarbonate values observed under hypoxic conditions are much higher than predicted given the measured decrease in haemolymph ⁠. The process that leads to greater extracellular alkalization than predicted by ventilatory changes alone was not examined in this study. One hypothesis is that grasshoppers attempt to conserve intracellular pH by transferring net acid from the haemolymph compartment to the intracellular compartment when intracellular pH is increased by declining during hypoxia exposure. Although we have no direct evidence to support this suggestion, intracellular pH has been shown to be regulated independently of haemolymph pH when body temperature varies (Harrison, 1988). An alternative hypothesis is that hypoxia could inhibit acid–base transport processes either in general and/or in the gut epithelia. For example, apparent base secretion by the midgut lumen (Harrison et al. 1992) could be highly sensitive to hypoxia.

Many studies have reported that insects tolerate exposure to hypoxia or anoxia with little or no detectable oxygen debt (Park and Buck, 1960; reviewed by Keister and Buck, 1974). Our data suggest that exposure. to hypoxia does have some effect on metabolism, since the values of animals in room air previously exposed to hypoxia were elevated by approximately 45 % relative to animals without prior hypoxic exposure (Fig. 8). One. possible explanation for this effect is that the increase in represents a classic oxygen debt, associated with clearance of lactate produced by grasshopper leg muscles (Gade, 1975; Zebe and McShaw, 1957; H. arrison et al. 1991) in response to hypoxia. If so, the we measured under hypoxic conditions may not accurately reflect the oxygen consumption rate since CO₂ buffering of metabolic acid is expected to occur under these conditions. Arieli and Lehrer (1988) measured oxygen consumption of Locusta migratoria grasshoppers and found that the rate of oxygen consumption decreased at values of less than 4 kPa. Thus, if CO₂ buffering of metabolic acid occurred, it is most likely to have occurred at 2 kPa O₂ and not at 5 kPa O₂.

Since resting grasshoppers maintain a high tracheal (Gulinson and Harrison, 1996), and prior studies have suggested that ventilatory frequency does not increase until approaches 5 kPa in grasshoppers (Arieli and Lehrer, 1988), it was conceivable that S. americana would simply tolerate the fourfold lower oxygen gradient from air to mitochondria in response to a switch from 21 to 5 kPa O₂ without any active change in tracheal function. In contrast to this view, our study demonstrates that S. americana grasshoppers respond to hypoxia with increases in the conductance of the tracheal system at each progressive decrease in ambient (Figs 10, 12). Our study also demonstrates that the conductance of the tracheal system of S. americana increases because of increases in both spiracular and tracheolar conductances (Figs 10, 12). Between 21 and 5 kPa O₂, the increases in spiracular and tracheolar conductances are similar (Figs 10, 12); whereas at 2 kPa O₂, it is primarily spiracular conductance which increases further. Since the rate of oxygen consumption falls in this range (Arieli and Lehrer, 1988), our data are consistent with the hypothesis that tracheolar conductance in this range (<5 kPa O₂) cannot be appreciably increased and becomes limiting for oxygen delivery. As ambient drops from 21 to 2 kPa, the haemolymph-to-air CO₂ conductance increases approximately half as much as required for perfect compensation for the hypoxia; thus, the maintenance of constant metabolic rates in the face of severe hypoxia is probably also facilitated by tolerance of lower cellular values.

Our results strongly suggest that convection is the major mechanism by which spiracular conductance is increased during hypoxic exposure. If diffusion played a large role in spiracular gas exchange, the expired calculated according to equation 5 would have been higher than that measured directly from the trachea. In fact, calculated expired and directly measured thoracic tracheal were statistically identical (Fig. 9). In addition, at least between 21 and 5 kPa O₂, the measured rise in spiracular conductance was linearly related with a slope of approximately 1 to the rise in spiracular conductance calculated using expired values derived from equation 5 (Fig. 13), further supporting the argument that the rise in spiracular conductance in response to hypoxia is almost completely due to an increase in bulk flow. At 2 kPa O₂, the spiracular conductance calculated from measurements of thoracic. tracheal gases was very high relative to that calculated from and convective ventilation (five times greater, Fig. 13B), owing to very high calculated spiracular conductances for a few animals with near-zero thoracic tracheal values. Since inspiration occurs via the thoracic spiracles, it is possible that for some animals the convective ventilation is so high during severe hypoxia that the of thoracic tracheal gases closely approximates ambient ⁠.

Previous studies have suggested that in room air a considerable proportion (50 %) of the CO₂ emission of quiescent grasshoppers can be independent of abdominal pumping (Harrison, 1997). For the two animals excluded from the statistical analysis, calculated expired values w. ere very high and convective ventilations were low, yet values were normal. These data, together with previous studies (Harrison, 1997; Harrison et al. 1995), suggest that adequate gas exchange can occur when visible abdominal pumping is absent in resting grasshoppers. In these animals, other mechanisms, such as convection associated with miniature ventilations (Hustert, 1975) or diffusion, may be important.

There is some evidence that tracheolar conductance for CO₂ increased in response to hypoxia, on the basis of the significant heterogeneity of slopes observed when comparing the ambient versus haemolymph and the ambient versus expired regressions (Fig. 9) with the nonsignificant (P=0.052) doubling of calculated tracheolar conductance (Fig. 10). The most likely explanation for these data is that the fluid levels in the grasshopper tracheoles decrease in response to hypoxia as reported by Wigglesworth (1931) for Ceratophyllus fasciatus fleas, adult Aedes argenteus mosquitoes, Tenebrio molitor mealworm larvae and Blattella germanica cockroaches. However, the increase in tracheolar conductance could also be due to increased convection between the large longitudinal trachea and the tracheoles, driven by the abdominal pressure pulsations. Tests of these alternative mechanisms will require observations or measurement of fluid levels in grasshopper tracheoles, the repetition of experiments such as these in animals for which tracheolar fluid levels can be observed or the use of methods to differentiate between the importance of diffusion and convection in tracheolar conductance (such as the use of varying inert gases or barometric pressure).

The spiracular conductance for oxygen is likely to be very similar to that measured for CO₂ since, at least in these experiments, spiracular conductance was primarily convective. Tracheolar conductance is likely to occur at least partially by diffusion, first in the gas phase down the blind-ended tracheoles (within which oxygen will move 1.2 times faster than CO₂) and then through tracheolar fluids and tissue (within which oxygen will move 36 times slower than CO₂; Krogh, 1919). Most authors have assumed that the conductance of oxygen from the tracheoles to the tissues will be much lower than the conductance of CO₂ (Buck, 1962; Kestler, 1985; Wigglesworth, 1983), given the much lower solubility of oxygen than of CO₂ in tissues. However, at 2 kPa O₂, haemolymph was 1.2 kPa (and cellular may be higher), while the maximum tracheal under these conditions is 2 kPa O₂ and, thus, the tracheolar conductance for oxygen cannot be less than 60 % of that for CO₂ under these conditions. At least under severe hypoxic conditions, therefore, transport of oxygen to the mitochondria must occur mostly in the gas phase. However, if the fluid levels in the tracheae of grasshoppers vary with as they do in other terrestrial insects (Wigglesworth, 1983), then the tracheolar conductance for O₂ may be much more affected by variation than the tracheolar conductance for CO₂.

Two possible reasons for the high tolerance of insects to extremely low ambient oxygen concentrations can be suggested. The first possibility is that hypoxia tolerance has evolved to allow insects to survive rare, but potentially deadly, exposure to low oxygen levels such as might occur when an insect is temporarily trapped in a small space. Alternatively, the tolerance to hypoxia might simply reflect an adaptation of the insect tracheal system to high metabolic rates during activity. For example, at 35 °C, both jumping and feeding grasshoppers consume oxygen at rates over 10 times that of an unfed, quiescent grasshopper at 22 °C (Harrison et al. 1991; Harrison and Fewell, 1995), and in-flight oxygen consumption rates can be 40–50 times greater than that of a quiescent grasshopper (Krogh and Weis-Fogh, 1951). To meet the gas exchange requirements of activity, insects in general, and grasshoppers in particular, have evolved the capacity to greatly increase the conductance of the tracheal system. Thus, the ability of these animals to maintain constant oxygen consumption rates in the face of very low oxygen availability probably reflects (1) the low metabolic rates of resting insects relative to those occurring in active insects, (2) the evolution of tracheal morphology and ventilatory mechanisms to exchange gases at some of the highest rates known in the animal kingdom during activity and (3) the ability of insects to rapidly modulate tracheal conductance.

Support for this research was provided by NSF IBN-9317784 to J.F.H. We would like to thank Eric Wilkinson and Katie Krolikowski for their efforts in the design and construction of the optical ventometer, B. J. Behm for his rendering of the apparatus and Stephen Roberts and two anonymous reviewers for helpful comments on the manuscript.