The respiratory system of growing caterpillars is challenged in two distinct ways as they develop from hatchlings to fifth instars preparing for pupation. First, across instars, body sizes and tracheal lengths increase substantially. Second, within each instar, animal mass can more than double while major tracheal respiratory system structures, such as spiracles and large tracheae, are fixed in size until molting. To test whether these growth processes result in a decrease in O2 delivery capacity relative to tissue oxygen needs, we exposed feeding Manduca sexta larvae of various ages to decreasing levels of atmospheric O2 and measured their metabolic rate and ability to feed. We found that near the beginning of all instars, M. sexta were able to maintain gas exchange and feed down to approximately 5 kPa O2, indicating that these insects are able to match tracheal O2 delivery to increased metabolic rates across instars. However, gas exchange and feeding of caterpillars nearing the molt were limited at much higher O2 levels (up to 15 kPa O2), suggesting that caterpillars have limited capacities to increase tracheal O2 delivery as O2 consumption rates increase within instars. It seems possible that the safety margin for O2 delivery may disappear completely in the last hours before ecdysis, providing an ultimate if not proximate explanation for the necessity of molting.

The mechanisms and consequences of growth and its regulation are of primary importance to developmental biology and physiology. Growing arthropods gain substantial mass both within and across intermolt periods (instars), so the respiratory system of developing insects must cope with tremendous increases in O2 consumption and CO2 production, as well as increases in tracheal lengths that may challenge diffusive capacity. For example, in the tobacco hornworm caterpillar, Manduca sexta, body masses more than triple within each instar, body lengths increase more than 10-fold from first to fifth instar (K.J.G., unpublished) and total body mass increases 10,000-fold across the larval stages(Goodman et al., 1985).

How can the insect tracheal system compensate for growth and increased gas exchange needs across and within instars? The tracheal system may potentially compensate for increased lengths by either morphological or physiological mechanisms. Across instars, morphological mechanisms of compensation may include increasing tracheal diameters at each molt(Beitel and Krasnow, 2000). In addition, tracheal branching and sprouting are sensitive to PO2(Wigglesworth, 1954; Locke, 1958; Jarecki et al., 1999), and tracheole density increases with age in grasshoppers(Hartung et al., 2005). Thus if tissue PO2 were to decrease due to inadequate O2 delivery, tracheole proliferation could be stimulated, providing an opportunity for tracheal morphology to match O2 delivery needs. Physiological changes in the mechanisms of gas exchange are also possible. Grasshoppers increase their mass-specific tracheal capacities as they grow across instars by increasing convective gas exchange through increases in the rate of abdominal pumping and tidal volume(Greenlee and Harrison,2004a). Caterpillars have traditionally been thought to breath by diffusion (Krogh, 1920), but they may use hemolymph pulsations generated by micro-contractions of the intersegmental muscles of the abdomen to drive convection through the compressible tracheae (Slama,1999; Smits et al.,2000).

In addition to dramatic body size increases across instars, within each instar, caterpillars increase in body mass at least 100% and up to 1000% in the fifth instar (Goodman et al.,1985). Do these size increases correlate with increasing challenges for oxygen delivery? In Drosophila, tracheal diameters do not increase during an intermolt period(Beitel and Krasnow, 2000). Similarly, spiracles are sclerotized and can only increase in size at the molt. Thus, the only likely morphological mechanism to allow increased O2 delivery within an intermolt period is tracheole sprouting(Locke, 1958; Jarecki et al., 1999). Physiological mechanisms, such as increasing convective gas exchange, could compensate for increased O2 demands; however compression of flexible tracheae by growing tissues may impede the animal's ability to increase convection. For example, grasshoppers late in the intermolt period have decreased ability to respond to hypoxia (i.e. critical PO2 increases throughout the instar, Greenlee and Harrison, 2004b). The decreased ability to respond to hypoxia may be caused by a decrease in the volume of the air sacs that become compressed as tissue grows within the sclerotized exoskeleton (Clarke,1957; Greenlee and Harrison,2004a). However, since the caterpillar exoskeleton is only sclerotized at the head, leaving the rest of the body free to expand throughout the instar (Eaton,1988), tissue growth seems less likely to compress the tracheae.

If an insect exchanges gases by diffusion, then larger body sizes may lead to increasing problems with gas exchange due to the well-documented exponential decreases in diffusion rates with distance. This idea has contributed to the suggestion that atmospheric oxygen levels may be linked to insect body size (Graham et al.,1995; Dudley,1998). In the present study, we first tested the hypothesis that larger caterpillars are unable to match tracheal O2 delivery capacity to tissue O2 needs during ontogeny. If larger insects have problems with oxygen delivery, then we would predict that larger caterpillars would have higher critical PO2 values(Pc, the PO2 below which metabolism and feeding can no longer be sustained). Dimensional differences are most prominent across instars, so to test this hypothesis we compared caterpillars at the beginning of different instars. In addition, we tested the hypothesis that as caterpillars grow throughout an instar, their oxygen delivery capacity does not match increases in tissue oxygen demand due to lack of plasticity in tracheal structure within an intermolt period. If animals nearing the molt have problems with oxygen delivery, we would predict that caterpillars near the end of an instar would have higher Pc values compared with animals that have recently molted.

### Caterpillars

Manduca sexta L. larvae were reared from eggs (Carolina Biological Supply, Burlington, NC, USA) on a wheat-germ-based artificial diet as previously described (Ojeda-Avila et al.,2003). We reared the caterpillars and conducted our experiments at room temperature (25°C). Animals were observed each day for signs of molting and the instar noted. We placed animals from each instar into one of two treatment groups: early- or late-stage animals; each treatment group had N=8 or 9. Early-stage animals were used the next day after the head capsule slipped and molting was complete. The first and third instars last approximately 3 days, therefore we considered them late-stage when they were 2 days past molting. The fifth instar lasts approximately 10 days, with peak body mass occurring between days 4 and 5. Therefore, to maximize difference in body mass, fifth instar animals were considered to be late-stage when they were 4 days past molting.

### Respirometry

We measured CO2 emission of caterpillars during feeding to maximize metabolic rates and to minimize struggling in our respirometry system. Food was placed in a small dish and weighed using an analytical balance to the nearest 0.0001 g (Mettler AE-240, Columbus, OH, USA). The dish and food were then placed in the respirometry chamber and flushed with CO2-free air (approximately 100 ml min-1) for 20 min to remove any CO2. Food treated in this manner had no measurable CO2 emission when placed in the chamber alone. Additionally, water loss from the food was negligible (<0.7% of the initial food mass) as estimated by placing food alone in the chamber for approximately 2 h, a period longer than any experimental trial. The experimental animal was then weighed in the same manner, placed on the food in the chamber, and allowed to acclimate to the chamber for 20 min. During each trial, caterpillars were exposed to 10 min each of five different gas mixes (21, 15, 10, 5, 3 and 0 kPa O2) generated with a Brooks 5878 mass flow controller and Brooks mass flow meters (Brooks Instruments, Hatfield, PA, USA). Gas mixes were subsampled from a 60 ml syringe and pushed through the chamber by an Ametek R-1 flow controller (Ametek, Pittsburgh, PA, USA) at 18–160 ml min-1stp. The excurrent air stream was directed through a CO2 analyzer (Li-6252 Li-Cor, Lincoln, NE, USA) to measure CO2 fraction in the air and then through an Ametek S3-A oxygen analyzer to verify the gas mixture. Signals from the CO2 and O2 analyzers were digitized and recorded using Sable Systems DataCan software and hardware (Salde Systems International, Las Vegas, NV,USA). We calculated CO2 (μmol g-1 h-1) as:
$\ {\dot{M}}_{\mathrm{CO}_{2}}=\frac{{\dot{V}}_{\mathrm{in}}(F\mathrm{\mathcal{e}}_{\mathrm{CO}_{2}}-F\mathrm{\mathcal{i}}_{\mathrm{CO}_{2}}){\times}2678.58}{M_{\mathrm{b}}},$
1
where in was the upstream flow rate, FeCO2 was the fraction of CO2 in the excurrent airstream, and FiCO2, the fraction of CO2 in the incurrent airstream, was equal to 0, and Mb was body mass (g). Additionally, 2678.58 is the conversion factor used to convert ml g-1 min-1 toμmol g-1 h-1 (1000 μl ml -1, 60 min h-1 and 22.4 μl μmol-1). We converted our measures of CO2 emission (ml s-1) to metabolic energy (Watts)using the Joule equivalence 23.7 kJ (per litre of CO2)-1(Withers, 1992) for a respiratory exchange ratio (RER) of 0.88, that of M. sexta(Alleyne et al., 1997).

In addition to determining metabolic rates, we calculated critical PO2 values(Pc-CO2)(Greenlee and Harrison,2004a). Briefly, we compared 95% confidence intervals around the mean CO2 emission rates at each PO2. A PO2 was considered to be the Pc-CO2 if the mean CO2 at the next higher and across all higher PO2s were significantly higher.

We then calculated maximal mass-specific tracheal system conductance, Gmax (μmol g-1 h-1kPa-1, Greenlee and Harrison,2004a). Briefly, we assumed that at the PO2 just higher than the Pc-CO2 for CO2, the animals were maximally conducting gases through wide-open spiracles and a fluid-free tracheal system, and that mitochondrial PO2 at the Pc-CO2 is near zero. We converted CO2 to O2 assuming a RER of 0.88 and calculated:
$\ G_{\mathrm{max}}=\frac{{\dot{M}}_{\mathrm{O}_{2}}}{\mathrm{atmospheric}{\\ }P_{\mathrm{O}_{2}}},$
2
where atmospheric PO2 is the PO2 just higher than the Pc-CO2.

### Feeding behavior

Animals began feeding as soon as they were placed on food in the respirometry chamber. We observed them during each hypoxic exposure to determine whether they were continuing to feed, and considered total feeding time to be the acclimation time in the chamber plus the time until the animal stopped feeding. The PO2 at which animals ceased feeding was identified as critical point for feeding(Pc-feeding). Food was weighed after the last hypoxic exposure, and total feeding rate (g h-1) was calculated as the difference in food mass from the beginning to the end of the trial divided by the time that the animals were observed feeding. Animals ate on average 25% of the food offered (range 2–71%).

### Statistics

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%. We analyzed CO2 data using repeated measures analysis of variance (ANOVA), since individual caterpillars were exposed to multiple PO2 values. Feeding rates were compared using ANOVAs with the independent variables being instar and stage. Pc and Pc-feeding 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).

### Body mass and normoxic CO2 emission rate

As animals grew, body mass increased over 2000 times from 1 day to 17 days of age (0.003±0.0005 g and 6.2±0.5 g,respectively; Fig. 1). The increase in body mass from early- to late-stage depended upon the animal's instar (ANOVA,instar × stage interaction: F1,46=24.7, P<0.001; Fig. 1).

Fig. 1.

Body mass of first, third and fifth instar caterpillars at early (filled symbols) and late (open symbols) stages within an instar.

Fig. 1.

Body mass of first, third and fifth instar caterpillars at early (filled symbols) and late (open symbols) stages within an instar.

Whole-animal, normoxic CO2 emission rates increased four orders of magnitude from the youngest to the oldest animals in our study(Fig. 2; 0.229±0.05 to 317±14.9 μmol h-1; ANOVA, F1,46=52.4, P<0.001). CO2 emission scaled with body mass(Mb) to the 0.98 power (log CO2 μmol h-1=0.98 [log Mb] + 1.66, r2=0.97, P<0.001; Fig. 2), and metabolic energy scaled with body mass to the 0.96 power[Watts=0.0071(mass)0.96].

Fig. 2.

Scaling of normoxic CO2 emission rate with body mass (g).

Fig. 2.

Scaling of normoxic CO2 emission rate with body mass (g).

Mass-specific, normoxic CO2 varied differently with stage depending upon the instar (ANOVA, stage × instar interaction, F1,46=17.7, P<0.0001). For instance, first and third instar caterpillars' CO2 emission rates decreased from early to late stage, while late-stage fifth instar caterpillars increased mass-specific CO2 to 60%higher than the early-stage value (Fig. 3).

Fig. 3.

CO2 emission rates in response to decreasing atmospheric PO2 for each instar and stage. Arrows indicate critical PO2 values(Pc-CO2), filled symbols represent early-stage animals, and open symbols represent late-stage animals.

Fig. 3.

CO2 emission rates in response to decreasing atmospheric PO2 for each instar and stage. Arrows indicate critical PO2 values(Pc-CO2), filled symbols represent early-stage animals, and open symbols represent late-stage animals.

### Hypoxic CO2 emission, Pc, maximal tracheal system conductance

In response to decreasing levels of atmospheric O2, caterpillars generally maintained CO2 emission rates at their normoxic levels,until a critical point was reached. However, depending on the instar, early-and late-stage animals responded differently to the decreased PO2 (Fig. 3; repeated measures ANOVA, instar × stage × PO2 interaction, F5,230=7.9, P<0.001). Analyzing each instar separately, we found that the CO2 emission rate response to hypoxia depended strongly on stage (Fig. 3; repeated measures ANOVA, PO2× stage interaction; first instar: F5,75=4.7, P<0.001; third instar: F5,75=5.8, P<0.0001; fifth instar: F5,70=21.4, P<0.0001).

Pc-CO2 also varied with stage depending on the caterpillar's instar (Fig. 4). When we analyzed each instar separately, we found that late-stage third and fifth instar caterpillars had significantly higher Pc values than early stages(Fig. 4, Mann-Whitney U test,third instar: U=8, P<0.01; fifth instar: U=2, P=0.001),while there was no stage effect in the first instars(Fig. 4).

Fig. 4.

Pc-CO2 as a function of instar. The boundaries of each box indicate the 25th and the 75th percentiles,respectively. The line within the each box indicates the median. Error bars mark the 10th and 90th percentiles. Open boxes are early-stage and hatched boxes are late-stage animals. Asterisks denote significant differences between early and late stages.

Fig. 4.

Pc-CO2 as a function of instar. The boundaries of each box indicate the 25th and the 75th percentiles,respectively. The line within the each box indicates the median. Error bars mark the 10th and 90th percentiles. Open boxes are early-stage and hatched boxes are late-stage animals. Asterisks denote significant differences between early and late stages.

Mass-specific maximal tracheal system conductance, Gmax, decreased with instar(Fig. 5; ANOVA, effect of instar, F1,46=11.1, P<0.01). Within each instar, mass-specific Gmax decreased on average 49% from early to late stage (Fig. 5;ANOVA, effect of stage, F1,46=8.1, P<0.01).

Fig. 5.

Absolute Gmax (A) and mass-specific Gmax (B) versus instar. Filled symbols are early-stage caterpillars; open symbols are late-stage caterpillars. Asterisks denote significant differences between early and late stages.

Fig. 5.

Absolute Gmax (A) and mass-specific Gmax (B) versus instar. Filled symbols are early-stage caterpillars; open symbols are late-stage caterpillars. Asterisks denote significant differences between early and late stages.

### Feeding behavior

Three animals did not eat during the experiment and were therefore excluded from these analyses. The remaining 47 animals ate continuously during the experiment until the PO2 became limiting, at which point they raised their heads off the food. Some animals even moved away from their food. Food eaten (g h-1) varied with instar(Fig. 6, ANOVA, effect of instar, F1,43=14.2, P<0.001).

Fig. 6.

Feeding rates (A; food eaten, g h-1) and mass-specific feeding rates (B; food eaten, g g-1 h-1) of caterpillars versus instar for early and late-stage animals. Asterisks denote significant differences between early and late stages.

Fig. 6.

Feeding rates (A; food eaten, g h-1) and mass-specific feeding rates (B; food eaten, g g-1 h-1) of caterpillars versus instar for early and late-stage animals. Asterisks denote significant differences between early and late stages.

When we corrected the amount of food eaten for caterpillar body mass, [food eaten (g) g caterpillar-1 h-1], we found that the amount of food eaten varied with instar depending on the animal's stage(Fig. 6, ANOVA, instar ×stage interaction, F1,43=5.6, P<0.03). All late-stage animals ate less food (g g caterpillar-1 h-1)compared with early-stage-animals; however these decreases were not statistically significant (ANOVAs, effect of stage, first instar: F1,13=4.7, P=0.05; third instar: F1,15=4.3, P=0.056; fifth instar: F1,13=3.2, P=0.099).

Pc-feeding varied with instar(Fig. 7; Scheirer-Ray-Hare extension of the Kruskal-Wallis test, effect of instar, H1,43=11.0, P<0.05). In addition, Pc-feeding decreased across instar for early-stage animals (Kruskal-Wallis test statistic=13.9, P<0.01). When we analyzed each instar separately to test for stage effects, we found that early-stage fifth instars continued to feed at significantly lower PO2 values than late-stage animals (Mann-Whitney U test, U=14, P=0.03).

Fig. 7.

Pc-feedingversus instar. The boundaries of each box indicate the 25th and the 75th percentiles respectively. The line within the each box indicates the median. Error bars mark the 10th and 90th percentiles. Open boxes are early-stage and hatched boxes are late stage animals. Asterisks denote significant differences between early and late stages.

Fig. 7.

Pc-feedingversus instar. The boundaries of each box indicate the 25th and the 75th percentiles respectively. The line within the each box indicates the median. Error bars mark the 10th and 90th percentiles. Open boxes are early-stage and hatched boxes are late stage animals. Asterisks denote significant differences between early and late stages.

The changes in Pc within an instar were much more striking than the differences across instars. Within an instar, early-stage caterpillars had 2–3-fold lower Pc-CO2values compared with late-stage caterpillars, despite the lower mass-specific CO2 emission rates for late-stage first and third instar animals(Figs 3 and 4). These data support the hypothesis that caterpillars within an instar experience a shrinking safety margin, as the increase in tissue oxygen needs is not matched by changes O2 delivery capacity. The rise in Pc during an instar can be explained simply by increased tissue O2 demands with an unchanging tracheal system capacity in the 1st and 3rd instars, since absolute, maximal Gmax does not vary within these instars(Fig. 5). Interestingly,absolute, maximal Gmax increases in the 5th instar, but by less than the increase in absolute gas exchange, leading to the decrease in Pc.

Across instars, larger caterpillars did not appear to have more difficulty than smaller caterpillars in responding to hypoxia. There was no significant effect of instar on Pc-CO2, while Pc-feeding actually decreased in the older/larger instars(Figs 4, 7). Thus tracheal conductances matched or exceeded changes in oxygen consumption rates across instars.

### CO2 emission rate in normoxia

CO2 emission rates scaled with body mass to the 0.98 power(Fig. 2), a finding similar to that for some lepidopterans throughout ontogeny (cecropia moths, Hyalophora cecropia, 0.9, Schroeder and Dunlap, 1970; M. sexta, 1.0, Dahlman and Herald,1971; and brown-tailed moths, Euproctis chrysorrhoea,1.0, Migula, 1974) but not others (corn earworms, Heliothis zea, 0.7, Edwards, 1970; tent caterpillars, Malacosoma neustria, 0.8, Migula, 1974; and M. sexta, 0.8, Alleyne et al.,1997). Our finding that CO2 varies strongly within an instar suggests that the variation among scaling coefficients reported for caterpillars could be due to variation in the age within an instar of the caterpillars studied.

Variation in age within an instar may contribute strongly to the variation in mass scaling of metabolic rate during ontogeny across insect species. For example, mass-specific metabolic rate was higher for late-stage fifth instar caterpillars than for early stage caterpillars of the same instar, while an opposite pattern was observed for the first and third instars(Fig. 3). Late-stage fifth instar caterpillars had increases in absolute CO2 emission (630%from early-stage) that were double the increases in body mass (360% from early-stage). A possible explanation is that preparation for wandering and pupation increase metabolic rate in fifth instars relative to other instars. The high late-instar metabolic rates of fifth instars contribute to the reduced safety margin for O2 delivery observed for this group. Our results indicate that future studies of the ontogenetic scaling of gas exchange and, probably physiological processes in general, must carefully control for age within instar.

### Hypoxia response and Pc forṀCO2

#### Across-instar effects

Newly molted caterpillars were able to maintain CO2 emission down to low PO2 levels (as low as 5 kPa). These results are comparable to those found for other insects(Keister and Buck, 1974; Wegener and Moratzky, 1995; Greenlee and Harrison, 1998; Emekci et al., 2002; Greenlee and Harrison, 2004a). The constant Pc-CO2 across instars results from a matching of maximal tracheal conductance to metabolic rate changes, as both mass-specific Gmax and CO2 emission rate decrease by about 50% from first to fifth instar(Fig. 5B). These patterns are not a general phenomenon in insects, as Pc in grasshoppers decreases and absolute Gmax increases during ontogeny(Greenlee and Harrison,2004a).

CO2 emission rates at low PO2levels were (for a short time at least) maintained at relatively high levels. In anoxia, all animals were able to maintain CO2 emission rates at 30% of their normoxic values for the 10 min trial(Fig. 8). These high ratios of anoxic CO2/normoxic CO2 suggest possible use of anaerobic metabolism during extreme hypoxia/anoxia. In support of this hypothesis, caterpillar intersegmental muscles contain considerable quantities of lactate dehydrogenase (Gade,1975). The high ratios of anoxic CO2/normoxic CO2(Fig. 8) are also likely due to washout of CO2 from the tissues as hypoxia/anoxia promote maximal spiracular opening (Case,1956), evacuation of fluid from the tracheoles(Wigglesworth, 1931) and reduced tissue PCO2 values(Greenlee and Harrison 1998).

Fig. 8.

CO2 in anoxia divided by CO2in normoxia as a function of instar. Filled symbols are early-stage caterpillars; open symbols are late-stage caterpillars.

Fig. 8.

CO2 in anoxia divided by CO2in normoxia as a function of instar. Filled symbols are early-stage caterpillars; open symbols are late-stage caterpillars.

#### Within-instar effects

CO2 emission rates and feeding rates of caterpillars late within each instar were clearly limited at higher PO2values than their earlier counterparts as evidenced by the much higher Pc-CO2 and Pc-feeding in late-stage caterpillars. Early-stage caterpillars had lower Pc-CO2 values compared with late-stage caterpillars, despite the generally lower mass-specific CO2emission rates for late-stage animals (Figs 3 and 4). These data support the hypothesis that caterpillars within an instar experience a shrinking safety margin, as changes in O2 delivery capacity do not parallel changes in tissue oxygen needs. In fact, mass-specific Gmaxdecreased by nearly 50% from early to late-stage within each instar(Fig. 5B), whereas within-instar variation in mass-specific CO2 emission rates was different for each instar (Fig. 3). First instar animals showed a similar decrease in both mass-specific CO2 and Gmax, while third instar animals only had a 10% decrease in mass-specific CO2, and fifth instar animals actually had a 60% increase in mass-specific CO2 from the beginning to the end of an instar.

Why does mass-specific Gmax decrease within an instar?If the tracheal system was fixed within an instar (same spiracle and main tracheal dimensions, no tracheole sprouting), then absolute Gmax should be constant, but mass-specific Gmax would fall due to the rise in tissue mass with constant tracheal morphology. This pattern is observed for the first and third instars (Fig. 5), suggesting that the within-instar rise in Pc-CO2 for the third instars can be explained by rising animal oxygen needs without an increase in tracheal system oxygen delivery capacity. In the fifth instars,however, caterpillars increase absolute Gmax 192%(Fig. 5A) from early to late-stage, indicating a within-instar enhancement of oxygen delivery capacity. This increase in Gmax could be due to an increased use of convection or tracheole sprouting. The increase in absolute Gmax is far less than the 630% rise in absolute CO2 emission rate, leading to the rise in Pc-CO2 within the fifth instar.

### Feeding behavior

Caterpillars were able to continue feeding, in most cases without interruption, to very low PO2 values(Fig. 7). The oldest caterpillars in our study ate slightly more food (g g-1h-1, Fig. 6)compared to fifth instar M. sexta in another study using the same diet and temperature (approximately 0.06 g g-1 h-1 over a 4 h period, Kingsolver and Woods,1997). However, the ontogenetic pattern of increasing feeding rates with age that we observed follow those of previous researchers(Fig. 6). For example, zebra caterpillars (Melancha picta) feeding on sugarbeet leaves increased feeding rates (leaf area/day) with age more than 70-fold(Capinera, 1979). Growing armyworms (Mamestra configurata, Bailey,1976), saltmarsh caterpillars (Estigmene acrea, Capinera, 1978), and cotton bollworms (Helicoverpa zea, Huffman and Smith, 1979) also increased leaf feeding area per day with age. In addition, the mass-specific feeding rates of our caterpillars decreased 60-fold from hatching to fifth instar (Fig. 6), a finding similar to that observed in other lepidopteran larvae (Slansky Jr, 1993).

### Possible mechanisms of hypoxia tolerance

The low and similar values of Pc-CO2 and Pc-feeding suggest that caterpillars early in the intermolt periods have very large safety margins for O2 delivery. The existence of these large safety margins for O2 delivery in the absence of any outward signs of increasing ventilation begs the question, what is the mechanism for the large safety margin for O2 delivery? One possible explanation for their hypoxia tolerance is that these insects have`overbuilt” tracheal systems (larger diameters than needed) and simply tolerate substantial variation in tissue PO2during exposure to hypoxia. Another potential explanation is that the head movements during feeding are also used for ventilation, possibly creating convective gas exchange by compressing tracheae as the hydrostatic skeleton changes shape with each rhythmic movement(Westneat et al., 2003). Hemolymph pressure changes could also be driven by contractions of intersegmental (Slama, 1999)or heart (Smits et al., 2000)muscles that would generate convective gas flow and increase during hypoxia. Lastly, caterpillars may increase spiracular opening and/or remove tracheolar fluid to increase gas exchange(Wigglesworth, 1981).

### Implications

The decreased Pc values and Gmaxwithin an instar support the hypothesis that O2 may play a role in triggering larval to larval molting in insects (Figs 4 and 5). Also, since the late-stage animals we measured were at least one day away from molting, the effect of within-instar development on Pc and Gmax may be even more striking as animals more closely approach ecdysis. We hypothesize that as animals near the end of an instar,internal PO2 decreases, providing a signal for the initiation of the molting pathway. The signaling cascade is currently thought to begin with a decline in ecdysteroids, which stimulates release of pre-ecdysis-triggering hormone (PETH)(Zitnan and Adams, 2000) and ecdysis-triggering hormone (ETH) from epitracheal glands(Zitnan et al., 1996). Eclosion hormone (EH) is then released by ETH and a positive feedback system occurs between EH and ETH (Ewer et al.,1997). However, it is unclear what the initial trigger of the pathway is. In one investigation, Nijhout(1975) found that in M. sexta molting is initiated by achievement of a critical body size, a finding potentially consistent with internal gas tensions being important as a trigger. The involvement of O2 in the signal transduction pathway to molting is supported by the findings of Greenberg and Ar(1996) who found that in mealworms the number of molts was inversely proportional to PO2. The Pc-CO2in late-stage animals was approximately 15 kPa, indicating that a small safety margin for O2 delivery still exists at this time. However, the late-stage animals we measured were at least one day away from molting, with approximately 50% more mass to be gained(Goodman et al., 1985). Thus,it seems possible that the safety margin for O2 delivery may disappear completely in the last hours before ecdysis, providing an ultimate if not proximate explanation for the necessity of molting.

This work was supported by NSF IBN-9985857 to J.F.H., EPA U91616501 to K.J.G., NSF IBN-0206678 to K.J.G. and J.F.H., and a Sigma Xi Grant In Aid of Research to K.J.G. We would like to thank Jeff Hazel, Mike Quinlan, Ron Rutowski and Glenn Walsberg for their thoughtful comments on this manuscript. We would also like to thank Art Woods for his help with caterpillar rearing and Kristina Egbert for rearing, weighing and collecting metabolic data from the caterpillars.

Alleyne, M., Chappell, M. A., Gelman, D. B. and Beckage, N. E. (
1997
). Effects of parasitism by the braconid wasp Cotesia congregata on metabolic rate in host larvae of the tobacco hornworm, Manduca sexta.
J. Insect Physiol.
43
,
143
-154.
Bailey, C. G. (
1976
). A quantitative study of consumption and utilization of various diets in the Bertha armyworm, Mamestra configurata (Lepidoptera: Noctuidae).
Can. Entomol.
108
,
1319
-1326.
Beitel, G. J. and Krasnow, M. A. (
2000
). Genetic control of epithelial tube size in the Drosophila tracheal system.
Development
127
,
3271
-3282.
Capinera, J. L. (
1978
). Consumption of sugarbeet foliage by saltmarsh caterpillar Lepidoptera-Arctiidae.
J. Econ. Entomol.
71
,
661
-663.
Capinera, J. L. (
1979
). Zebra caterpillar(Lepidoptera, noctuidae) foliage consumption and development of larvae on sugarbeet.
Can. Entomol.
111
,
905
-909.
Case, J. F. (
1956
). Carbon dioxide and oxygen effects on the spiracles of flies.
Physiol. Zool.
29
,
163
-171.
Clarke, K. U. (
1957
). On the role of the tracheal system in the post-embryonic growth of Locusta migratoria L.
Proc. Roy. Entomol. Soc. Lond. A
32
,
67
-79.
Dahlman, D. L. and Herald, F. (
1971
). Effects of parasite, Apanteles congregatus, on respiration of tobacco hornworm, Manduca sexta larvae.
Comp. Biochem. Physiol.
40
,
871
-880.
Dudley, R. (
1998
). Atmospheric oxygen, giant Paleozoic insects and the evolution of aerial locomotor performance.
J. Exp. Biol.
201
,
1043
-1050.
Eaton, J. L. (
1988
).
Lepidopteran Anatomy
. New York: Wiley-Interscience.
Edwards, L. J. (
1970
). Oxygen consumption by the corn earworm, Heliothis zea.
Ann. Entomol. Soc. Am.
63
,
773
-776.
Emekci, M., Navarro, S., Donahaye, E., Rindner, M. and Azrieli,A. (
2002
). Respiration of Tribolium castaneum(Herbst) at reduced oxygen concentrations.
J. Stored Prod. Res.
38
,
413
-425.
Ewer, J., Gammie, S. C. and Truman, J. W.(
1997
). Control of insect ecdysis by a positive-feedback endocrine system: Roles of eclosion hormone and ecdysis triggering hormone.
J. Exp. Biol.
200
,
869
-881.
1975
). Zur stoffwechselphysiologischen spezialisierung von insektenmuskeln.
1974
,
258
-261.
Goodman, W. G., Carlson, R. O. and Nelson, K. L.(
1985
). Analysis of larval and pupal development in the tobacco hornworm (Lepidoptera: Sphingidae), Manduca sexta.
Ann. Entomol. Soc. Am.
78
,
70
-80.
Graham, J. B., Dudley, R., Aguilar, N. M. and Gans, C.(
1995
). Implications of the late Palaeozoic oxygen pulse for physiology and evolution.
Nature
375
,
117
-120.
Greenberg, S. and Ar, A. (
1996
). Effects of chronic hypoxia, normoxia and hyperoxia on larval development in the beetle Tenebrio molitor.
J. Insect Physiol.
42
,
991
-996.
Greenlee, K. J. and Harrison, J. F. (
1998
). Acid-base and respiratory responses to hypoxia in the grasshopper, Schistocerca americana.
J. Exp. Biol.
201
,
2843
-2855.
Greenlee, K. J. and Harrison, J. F. (
2004a
). Development of respiratory function in the American locust, Schistocerca americana I. Across-instar effects.
J. Exp. Biol.
207
,
497
-508.
Greenlee, K. J. and Harrison, J. F. (
2004b
). Development of respiratory function in the American locust, Schistocerca americana II. Within-instar effects.
J. Exp. Biol.
207
,
509
-517.
Hartung, D., Kirkton, S. D. and Harrison, J. F.(
2005
). Ontogeny of tracheal system structure: a light and electron-microscopic study of the metathoracic femur of the American locust, Schistocerca americana.
J. Morphol.
in press.
Huffman, F. R. and Smith, J. W. (
1979
). Bollworm-peanut foliage consumption and larval development Lepidoptera,Noctuidae.
Environ. Entomol.
8
,
465
-467.
Jarecki, J., Johnson, E. and Krasnow, M. A.(
1999
). Oxygen regulation of airway branching in Drosophila is mediated by branchless FGF.
Cell
99
,
211
-220.
Keister, M. and Buck, J. (
1974
). Respiration:some exogenous and endogenous effects on rate of respiration.
Physiol. Insecta
6
,
469
-509.
Kingsolver, J. G. and Woods, H. A. (
1997
). Thermal sensitivity of growth and feeding in Manduca sextacaterpillars. In
The Physiology of the Insecta
, vol.
6
(ed. M. Rockstein), pp.
469
Krogh, A. (
1920
). Uber gasdiffusion in den tracheen.
Pfleugers Archiv feur die gesamte physiolgie des menschen und der tiere
179
,
95
-112.
Locke, M. (
1958
). The co-ordination of growth in the tracheal system of insects.
Q. J. Microsc. Sci.
99
,
373
-391.
Migula, P. (
1974
). The effect of temperature on the rate of oxygen consumption by different life stages of Malacosoma neustria L. and Euproctis chrysorrhoea L. (Lepidoptera).
Ekologia Polska
22
,
173
-194.
Nijhout, H. F. (
1975
). A threshold size for metamorphosis in the tobacco hornworm, Manduca sexta (L.).
Biol. Bull.
149
,
214
-225.
Ojeda-Avila, T., Woods, H. A. and Raguso, R. A.(
2003
). Effects of dietary variation on growth, composition, and maturation of Manduca sexta (Sphingidae: Lepidoptera).
J. Insect Physiol.
49
,
293
-306.
Schroeder, L. and Dunlap, D. G. (
1970
). Respiration of cecropia moth (Hyalophora cecropia L.) larvae.
Comp. Biochem. Physiol.
35
,
953
-957.
Slama, K. (
1999
). Active regulation of insect respiration.
Ann. Entomol. Soc. Am.
92
,
916
-929.
Slansky, F., Jr (
1993
). Nutritional ecology:the fundamental quest for nutrients. In
Caterpillars: Ecological and evolutionary constraints on foraging
(ed. N. E. Stamp and T. M. Casey), pp.
29
-91. New York: Chapman and Hall.
Smits, A. W., Burggren, W. W. and Oliveras, D.(
2000
). Developmental changes in in vivo cardiac performance in the moth Manduca sexta.
J. Exp. Biol.
203
,
369
-378.
Wegener, G. and Moratzky, T. (
1995
). Hypoxia and anoxia in insects: microcalorimetric studies on two species (Locusta migratoria and Manduca sexta) showing different degrees of anoxia tolerance.
Thermochimica Acta
251
,
209
-218.
Westneat, M. W., Betz, O., Blob, R. W., Fezzaa, K., Cooper, W. J. and Lee, W.-K. (
2003
). Tracheal respiration in insects visualized with synchrotron x-ray imaging.
Science
299
,
558
-560.
Wigglesworth, V. B. (
1931
). The extent of air in the tracheoles of some terrestrial insects.
Proc. Roy. Soc. Lond. B
109
,
354
-359.
Wigglesworth, V. B. (
1954
). Growth and regeneration in the tracheal system of an insect, Rhodnius prolixus(Hemiptera).
Q. J. Microsc. Sci.
95
,
115
-137.
Wigglesworth, V. B. (
1981
). The natural history of insect tracheoles.
Physiol. Entomol.
6
,
121
-128.
Withers, P. C. (
1992
).
Comparative Animal Physiology
. New York: Saunders College Publishing.
Zar, J. H. (
1999
).
Biostatistical Analysis
. Upper Saddle River, NJ: Prentice-Hall.
Zitnan, D. and Adams, M. E. (
2000
). Excitatory and inhibitory roles of central ganglia in initiation of the insect ecdysis behavioural sequence.
J. Exp. Biol.
203
,
1329
-1340.
Zitnan, D., Kingan, T. G., Hermesman, J. L. and Adams, M. E.(
1996
). Identification of ecdysis-triggering hormone from an epitracheal endocrine system.
Science
271
,
88
-91.