Hypoxia-induced compression in the tracheal system of the tobacco hornworm caterpillar, Manduca sexta

Insects breathe using a tracheal respiratory system. The tracheal system consists of a series of branching tubes that decrease in size from their origin at the spiracles, which are valved openings to the atmosphere. In some species, muscular contractions coordinated with opening and closing of the spiracular valves convectively deliver oxygen to the tissues. The tracheal respiratory system is efficient, allowing insects to tolerate very low oxygen levels [<2 kPa partial pressure of O₂ (P_O2) (Harrison et al., 2006)]. Insects from several orders are known to use abdominal pumping to unidirectionally drive air flow through the tracheal system (Harrison, 2009). During hypoxia, abdominal pumping often increases as a compensatory response to increase airflow to tissues (Greenlee and Harrison, 2004a). Older, larger Schistocerca americana grasshoppers rely more on abdominal pumping at low P_O2 levels, suggesting that this behavior is an important response to hypoxia (Greenlee and Harrison, 2004a).

Insects that lack obvious mechanisms for producing major convective airflows would not be expected to be as tolerant of hypoxia. Caterpillars typically do not exhibit abdominal pumping behavior, and although a few mechanisms of producing convective airflows have been proposed [e.g. hemolymph pressure changes (Wasserthal, 1996) and passive suction ventilation (Kestler, 1985)], the relative role of convection versus diffusion in the tracheal system of caterpillars is not well understood. Despite the lack of obvious external ventilatory mechanisms, Manduca sexta L. caterpillars of all ages are surprisingly tolerant of hypoxia. For example, the average oxygen level below which M. sexta metabolism can no longer be sustained, the critical P_O2, was shown to be 5 kPa (Greenlee and Harrison, 2005), comparable to that of species with known mechanisms of convective ventilation. In addition, the largest larvae have been observed to feed in P_O2 levels as low as 1 kPa (Greenlee and Harrison, 2005), suggesting that size-related compensatory mechanisms for gas exchange may exist. During the course of this previous study, we observed that, when exposed to hypoxia, the older, larger caterpillars exhibited movements resembling abdominal pumping. Although this behavior had not been previously documented in non-molting caterpillars, we suspected that the body movements were a respiratory-related response, perhaps functioning to increase convective gas exchange during hypoxic exposure.

Here, we test the hypothesis that the external body movements play a role in gas exchange in M. sexta larvae by stimulating abdominal pumping with hypoxia (<3 kPa P_O2). We used high-speed flow-through respirometry alone and with X-ray imaging to visualize the tracheal system during abdominal pumping. In the first set of trials without X-ray imaging, high-speed flow-through respirometry enabled the resolution of individual expirations of CO₂, which could be correlated with body movements in hypoxia. In other trials, we combined respirometry, X-ray imaging and light video imaging to determine whether the external body movements observed in hypoxia resulted in internal changes in the tracheal system. Because the abdominal pumping movements were only observed in the largest caterpillars, we also tested the hypothesis that the oxygen delivery capacity of larger caterpillars does not match the increase in oxygen demand with increasing body mass.

Manduca sexta larvae were reared from eggs (Carolina Biological Supply, Burlington, NC, USA) with ad libitum access to a wheat-germ-based artificial diet (Ojeda-Avila et al., 2003). Animals were reared at 25°C on 16 h:8 h light:dark cycles and weighed prior to experiments. Caterpillars ranged in age from first to fifth instar (mass range: 0.001–6.95 g) and were within the first half of each instar based on their molting dates, body masses and head widths.

We used CO₂ emission as an index of oxygen consumption. High-speed, flow-through respirometry was used to measure CO₂ emission of caterpillars at room temperature (24±1°C) with and without the use of synchrotron X-ray imaging at Argonne National Laboratory and North Dakota State University (NDSU), respectively. Trials without X-ray imaging (N=7 caterpillars) used chambers constructed of 60 ml syringes that had been plumbed with flexible tubing. Chamber volume was adjusted based on the animal's body size. Animals were placed in chambers and allowed to acclimate for 20 min. Dry, CO₂-free air (Balston purge gas generator, Parker, Cleveland, OH, USA) was pushed through the chamber at various flow rates (50–5000 ml min⁻¹) based on animal size, using a mass flow controller (MFC-4, Sable Systems, Las Vegas, NV, USA) and mass flow meters (Sierra Instruments, Monterey, CA, USA). The time constant averaged 28±3.3 s. Hypoxic gas mixes (0, 1, 2, 3 or 5 kPa P_O2) were generated by diluting the air stream with N₂. Water was removed from excurrent air using MgClO₄ and the fraction of CO₂ was measured using a Li-Cor 7000 infrared gas analyzer (Li-Cor Biosciences, Lincoln, NE, USA). During experiments with X-ray imaging (N=37 caterpillars), the respirometry system was the same, except that chambers were constructed of optically clear acrylic and X-ray transparent polyimide film (Kapton, DuPont, Wilmington, DE, USA), as previously described (Greenlee et al., 2009). In addition, incurrent gas mixes were generated from O₂ and N₂ tanks and scrubbed of water and CO₂ using a drierite/ascarite/drierite column. CO₂ emission rates were recorded in normoxia pre- and post-beam exposure. All caterpillars were exposed to decreasing levels of atmospheric oxygen followed by a recovery period in normoxia. In experiments without X-ray imaging, animals were exposed to P_O2 levels of 21, 5, 3 and 0 kPa. Experiments with X-ray exposure used P_O2 levels of 21, 2, 1 and 0 kPa.

Caterpillars were shipped overnight from NDSU to the Advanced Photon Source (Argonne National Laboratory, Argonne, IL, USA) and were given ad libitum access to food. Tracheae in live caterpillars were visualized using synchrotron X-ray phase-contrast imaging with a 2× or 5× objective lens and a Cohu-cooled, charge-coupled device video camera as previously described (Socha et al., 2007). Resulting fields of view for each lens were 2.4×1.8 mm and 0.96×0.72 mm, respectively, with corresponding resolutions of approximately 5 and 2 μm. The distance from the sample to the scintillator (which converted X-rays to visible light) ranged from 0.8 to 1 m, and the monochromatic X-ray energy was 25 kev. Video was recorded to mini-DV tapes using a camcorder (TRV-900, Sony, Tokyo, Japan). A metal grid (400 lines per inch) was placed in the beam as a scale for spatial calibration of the X-ray images. Caterpillars were placed in respirometry chambers on a remote-controlled stage that allowed us to move the insect within the beam and to focus on specific body regions, as previously described (Greenlee et al., 2009). Exposure to X-ray radiation in a typical experiment was no longer than 15 min.

To determine whether and how tracheal structure and function changed during breathing in hypoxia, we used the X-ray videos to measure changes in tracheal diameter and compression frequency. X-ray videos were first digitized and converted to image sequences using ImageJ software (National Institutes of Health, Bethesda, MD, USA). Image sequences were viewed and tracheal compression cycles were identified. Each cycle had an observed maximum and minimum tracheal diameter. Three compression cycles from each P_O2 from the abdomen of each caterpillar were analyzed for tracheal diameter (N=43 caterpillars). When tracheae compressed, the frames with the largest and the smallest diameters were selected for measurement. Tracheal diameters were measured using ImageJ, with a measurement uncertainty of 1.2%. For sequences without any compression cycles, we digitized 20 consecutive video frames, with the starting point of the cycle selected arbitrarily. We calculated the fraction of compression by taking the change in diameter as a percent of the maximum diameter and averaging this for the three breaths from each P_O2. Fractional data were arc-sine transformed for statistical analysis. We also counted the number of tracheal compressions in 30 s to calculate compressions per minute (N=22 caterpillars).

To investigate the scaling of tracheal structures with body size in larval M. sexta, we used synchrotron X-ray imaging to create high-resolution projection images of the tracheal system of euthanized animals. Caterpillars within the first half of each instar of a range of sizes (0.0019 to 1.591 g) were euthanized with ethyl acetate, chilled to stabilize residual internal motion, and warmed to room temperature before X-ray imaging. Animals were placed upright in polyimide tubing on the movable stage. Because the field of view was smaller than the animal, multiple images were required to image entire caterpillars. For each specimen, composite images were created first by linearly translating the sample and capturing X-ray images that successively overlapped by ~30%, and then stitching them together using a custom MATLAB program (MathWorks, Natick, MA, USA). Images were captured using a CCD camera (Sensicam, Cooke Corporation, Romulus, MI, USA), which provided higher spatial resolutions than those achievable with the live video system. We directly measured the width of major tracheae in X-ray images (N=16 caterpillars) using ImageJ. Specific tracheae were recognized using morphological characters identified by Eaton (Eaton, 1988).

Means ± s.e.m. are presented throughout. Statistical analyses were performed using SPSS version 19 (IBM, Armonk, NY, USA) and SigmaPlot version 11 (Systat Software, San Jose, CA, USA). P-values <0.05 were considered to be statistically significant.

To test the effect of P_O2 on the ratio of CO₂ peaks to abdominal pumps, repeated-measures ANOVA (RM-ANOVA) was used. Because each animal was exposed to multiple P_O2 levels, P_O2 was the within-subjects factor and mass was a covariate.

To test for differences in CO₂ emission rates, log-transformed data were subjected to RM-ANOVA with treatment as the within-subjects factor (levels include pre-beam normoxia, normoxia with X-ray, 2 kPa P_O2, 1 kPa P_O2, anoxia and recovery in normoxia) and body mass as a covariate. We then calculated the linear regression between CO₂ emission and body mass for each within-subjects factor and compared 95% confidence intervals (CI) of slopes.

To assess the relationships between the fractional change in tracheal diameter and tracheal compression frequency with body mass, we used two approaches. First, data were log transformed and analyzed using RM-ANOVA with P_O2 as the within-subjects repeated factor and log-transformed body mass as the covariate. Second, because only large animals exhibited abdominal pumping, we binned the data. We classified animals by size, as either large (>0.2 g) or small (<0.2 g) and used RM-ANOVA with P_O2 as the within-subjects factor, size as the between-subjects factor and mass as the covariate.

To determine the relationship between tracheal diameter and body mass, data were log transformed and subjected to linear regressions of each trachea on body mass. Because there may be correlations between tracheae, we also grouped tracheae according to body segment and used a multivariate general linear model with log-transformed body mass as a covariate. Tracheae from the head and thorax were in one group, while abdominal tracheae from the first four segments were grouped. In addition, because each caterpillar had multiple tracheal measurements, we used RM-ANOVA with trachea as the within-subjects factor and mass as a covariate. We used three methods to determine whether the multiple comparisons were significant. First, we corrected the P-values for 15 multiple comparisons using the Bonferroni correction (1–α/k, where α=0.05 and k is the number of comparisons), decreasing the significance level to 0.003. Because the Bonferroni correction for multiple comparisons is very conservative and known to reduce power (Benjamini and Hochberg, 1995; Narum, 2006), we used two other methods. As our data are dependent and have a large number of comparisons, we used the false discovery rate (FDR) method, which results in an adjusted significance level of 0.033, and a modified FDR method, which results in a significance level of 0.015 (Benjamini and Hochberg, 1995; Benjamini and Yekutieli, 2001; Narum, 2006).

In normoxia, the pattern of CO₂ emission of caterpillars was continuous and irregular (Fig. 1A,B, supplementary material Movie 1). In addition, external body movements were not correlated with CO₂ peaks (mean ratio=0.27±0.1; Fig. 2). As atmospheric P_O2 decreased, caterpillar abdomens began rhythmically contracting, and two types of pumping movements were observed. One resembled traditional abdominal pumping, where the abdominal segments shorten dorso-ventrally in concert along the entire length of the abdomen (supplementary material Movie 1). The other type of pumping resulted in dorso-ventral compression of each abdominal segment in series from the thorax to the posterior of the caterpillar. For the second type of pumping, we counted one series of contractions as one movement. Correlation of external body movements and CO₂ emission increased as atmospheric P_O2 decreased (RM-ANOVA, F_2,18=12.06, P<0.001; Fig. 1A,C, Fig. 2, supplementary material Movie 1). Responses to exposures of either 5 or 3 kPa P_O2 did not differ, and caterpillar CO₂ emission became cyclic and was nearly perfectly correlated with external body movement (mean ratio of CO₂ peaks to external body movements=1.01±0.11; Fig. 2).

To determine whether the external body movements we observed were driving compressive deformations of the tracheal system, we used the same method while recording X-ray video of caterpillars breathing in normoxia and hypoxia. CO₂ emission rates varied with treatment (RM-ANOVA, F_1,21=7.51, P<0.02; Fig. 3) and body mass (RM-ANOVA, F_1,21=24.94, P<0.01). Bonferroni-corrected post hoc tests showed that CO₂ emission rate only varied due to the effects of hypoxia; there was no effect of X-ray exposure on CO₂ emission rate. Regression lines of CO₂ emission versus log mass were not significantly different between pre-beam normoxic CO₂ emission and normoxic X-ray exposure (95% CIs for slope: pre-beam normoxia 0.39–1.19, X-ray normoxia 0.30–1.16; for y-intercept: pre-beam normoxia 1.49–1.81, X-ray normoxia 1.35–1.69; Fig. 3A). CO₂ emission rate varied with P_O2 treatment (RM-ANOVA, F_1,30=2.76, P<0.001) and with mass (RM-ANOVA, F_1,30=58.76, P<0.001). Mass scaling of CO₂ emission rate did not vary with P_O2, as evidenced by overlapping 95% CI of slopes: 2 kPa PO₂, 95% CI=0.52–1.33; 1 kPa P_O2, 95% CI=0.71–1.31; anoxia, 95% CI=0.76–1.24; recovery, 95% CI=0.54–1.18; Fig. 3).

In normoxia, the X-ray video showed that tracheae translated with external body movements, but tracheal diameters did not change (supplementary material Movie 2). During hypoxia exposure, the external body movements of large caterpillars were also highly correlated with CO₂ emission peaks. Furthermore, the external body movements during hypoxia co-occurred with the compression of many tracheal tubes (Fig. 4, supplementary material Movie 1). Fractional change in tracheal diameter in hypoxia, but not normoxia or anoxia, was significantly correlated with body mass (Fig. 5) at 1 and 2 kPa P_O2 (RM-ANOVA, P_O2 × mass interaction: F_2,80=3.57, P<0.04). Regressing fractional change of tracheal diameter in both 1 and 2 kPa P_O2 on body mass revealed significant logarithmic relationships [1 kPa: y=0.36(mass)^0.46, F_1,47=54.4, P<0.001; 2 kPa: y=0.29(mass)^0.39, F_1,40=22.31, P<0.001; Fig. 5]. However, because smaller caterpillars (mass <0.2 g) did not exhibit external body movements or tracheal compressions, we decided to bin the data into two groups and determine the effects of mass within each group. There was a significant interaction between P_O2 and size, indicating that smaller caterpillars responded differently to hypoxia than larger ones (RM-ANOVA, F_2,76=3.27, P<0.05). Within each size class (i.e. small or large) there was no effect of body mass (P=0.18), suggesting that there is a size threshold for tracheal compression (Fig. 5).

Compression frequency showed a pattern similar to that of compression fraction (Fig. 6). Compression frequency in hypoxia varied differently with P_O2 level, depending on the mass of the animal (RM-ANOVA, P_O2 × mass interaction: F_2,22=4.36, P<0.03; Fig. 6). The smallest animals had the lowest compression frequencies regardless of P_O2 treatment. There were no compressions at 21 kPa P_O2 and few in anoxia. However, at 1 and 2 kPa P_O2 there was a significant power relationship between compression frequency and body mass (2 kPa: F_1,13=10.69, P<0.01; 1 kPa: F_1,21=14.87, P<0.01; Fig. 6). Because the smallest caterpillars did not exhibit this behavior, we again analyzed the grouped data. There was a significant effect of size on compression frequency, with larger animals having higher frequency than smaller animals (F_1,16=6.28, P<0.03). Within each group, there was no effect of body mass (P=0.996), suggesting that there is a threshold for tracheal compression (Fig. 6).

To determine how tracheal structure varied with body size, we measured the diameter of major tracheae throughout the head, thorax and abdomen of caterpillars of a range of sizes (N=16). We analyzed these data in two ways. First, tracheae were divided into two groups for multivariate analysis: head and thoracic tracheae were grouped, and abdominal segments 1–4 were grouped. Tracheae from abdominal segments 5–10 were excluded, as the most posterior abdominal segments from the largest animals could not be visualized with our imaging setup. Four out of 15 tracheae measured showed statistically significant correlations with body mass (Table 1, Figs 7, 8). Slopes of log-transformed data ranged from 0.04 in the dorsal trachea of the fourth abdominal segment to 0.32 in the metathoracic dorsal trachea (Table 1). RM-ANOVA showed a significant interaction between trachea and body mass, indicating that not all tracheae were similarly correlated with mass (F_14,42=3.41, P<0.01). Parameter estimates from this analysis indicated significant correlations between body mass and several of the tracheae (Table 1). Bonferroni-corrected P-values showed that there was no relationship between tracheal diameter and mass. FDR and modified FDR-corrected P-values showed that some tracheae scaled positively with mass. Between both analyses, tracheae in the head consistently did not scale with body mass.

Here, we describe the first observations of abdominal pumping being used for gas exchange in the caterpillar M. sexta. These external body movements during hypoxia resulted in substantial tracheal compression, which appeared to be the driving force for gas exchange, as evidenced by the high correlation between CO₂ emission peaks and body movements (Figs 1, 2). Tracheal compression frequency and fractional change in tracheal diameter both increased as P_O2 level decreased. In addition, compression frequency and fractional change in diameter increased with body mass. Interestingly, smaller caterpillars did not employ abdominal pumping in response to hypoxia, nor did they exhibit tracheal compression. These size trends suggest that larger caterpillars require greater amounts of convection to satisfy gas exchange needs.

Metabolic rate, as indicated by the rate of CO₂ emission, scaled proportionally with body mass, across the range of masses used for metabolic measures (Fig. 3A). The scaling coefficients are consistent with previous work, showing that across instars, metabolic rates of M. sexta larvae scale with body mass^0.98 (Greenlee and Harrison, 2005) or with body mass^0.94 (Sears et al., 2012). Larval and pre-pupal silkworms, Bombyx mori, also have metabolic scaling coefficients near 1 [0.96–1.49 (Blossman-Myer and Burggren, 2010)]. Ontogenetic scaling relationships are typically more variable than the commonly described allometric scaling of mammalian metabolic rates (Glazier, 2005). One proposed explanation for higher scaling coefficients is that increased epidermal cell growth prior to molting requires increased oxygen consumption (Blossman-Myer and Burggren, 2010). Clearly, more work is needed to fully understand the underlying parameters that determine metabolic scaling in juvenile insects.

To our knowledge, this is the first description of abdominal pumping in lepidopteran larvae in response to low oxygen. We analyzed both the frequency of tracheal compression and the fractional change in tracheal diameter as a function of body size and P_O2. The fractional change in tracheal diameter that occurred in hypoxia scaled logarithmically with body mass (scaling coefficient 0.46 or 0.39 depending on P_O2; Fig. 5), with the largest number of caterpillars that compressed their tracheae occurring in 1 kPa P_O2. Compression frequency also showed a statistically significant relationship with body mass, scaling with mass^0.31 or mass^0.5 depending on P_O2 (Fig. 6). In vertebrates, breathing frequency and tidal volume scale with mass^−0.25 and mass¹, respectively (Peters, 1983). Our comparable parameters, compression frequency and fractional change in tracheal diameter, did not scale with body mass in that way. An interspecific comparison of grasshoppers found mass scaling relationships of ventilation frequency (0.23 in hypoxia) and tidal volume (0.71 in hypoxia) similar to our findings here (Greenlee et al., 2007).

Although we were able to fit statistically significant regressions to both parameters as a function of body mass, there was a distinct lack of response in the smallest caterpillars. Similar to our current findings, young grasshoppers also did not increase abdominal pumping in response to hypoxia. In S. americana, the critical P_O2 values for abdominal pumping were low across all instars and were lower than the critical P_O2 for CO₂ emission, indicating that the low oxygen does not limit muscle contraction (Greenlee and Harrison, 2004b). Indeed, young M. sexta larvae have critical P_O2 values for CO₂ emission similar to the largest, oldest caterpillars [5 kPa P_O2 (Greenlee and Harrison, 2005)], but here we observed abdominal pumping and tracheal compression only in the largest larvae (Figs 5, 6). Interestingly, these large specimens were still able to maintain abdominal pumping movements in atmospheres as low as 1 kPa P_O2 (Figs 5, 6). When we analyzed data from animals larger than 0.2 g that exhibited the pumping behavior, we found no significant mass scaling relationship with compression frequency or fractional change in diameter. One possibility is that the range of body masses analyzed needs to be increased. However, taken together with the findings in small grasshoppers, these data strongly suggest that there is a size threshold for this behavior. This may indicate that the largest caterpillars are unable to meet oxygen demands associated with their massive growth without abdominal pumping. To conclusively determine whether the lack of response to hypoxia in the youngest juveniles is a general developmental respiratory pattern for insects, many more species need to be tested.

Because we found that only large caterpillars exhibited abdominal pumping in hypoxia, we suspected that they had reduced oxygen delivery capacity. We analyzed tracheal investment using the diameter of individual tracheae. Based on geometric similarity, isometric scaling of a structure's diameter is predicted to scale with body mass^0.33 (Calder, 1981). Using multivariate analyses, four of the tracheae showed significant correlations with mass, and all of these exhibited scaling coefficients that were not different from 0.33, as evidenced by the overlap in the 95% CI. Using regressions generated by RM-ANOVA, many more tracheae scaled positively with body mass. Because many of the measured tracheae exhibited nearly significant correlations with mass, increasing the sample size may reveal significant scaling relationships (Table 1). Interestingly, the tracheae that exhibited significant mass scaling were all located in the thoracic and abdominal segments. Tracheae in the head did not scale with mass regardless of the type of analysis or multiple comparison adjustment, suggesting that tracheae in the head have a large safety margin for oxygen delivery in the younger, smaller animals (Fig. 8, Table 1). Alternatively, this would explain the increased use of convection observed in the older, larger animals. Whole-animal tracheal volume in M. sexta, measured by water displacement, scaled isometrically with body mass across the third, fourth and fifth instars (Callier and Nijhout, 2011), suggesting that differences in regional tracheal system investment may vary dramatically, which highlights the need for better measures of regional tracheal volumes. For example, our measures only account for changes in major tracheae, whereas measures using water displacement could include growth of smaller structures that would be more likely to increase and proliferate during development. In addition, the water displacement method may result in overfilling of tracheae and, thus, overestimation of tracheal volumes. In support of our current findings, however, another study using a similar X-ray technique to estimate sizes of tracheal system structures found that developing S. americana grasshoppers show differences in regional tracheal investment, with scaling coefficients for tracheae of 0.128 in the head and 0.067 in the thorax (Greenlee et al., 2009). Because there are trade-offs between tracheal investment and non-respiratory structures, one possible explanation for the lack of scaling relationships found in the head is that other tissues or organs, such as the gut, incur greater functional demands with increasing age, relative to the tracheal system. Alternatively, tracheae in the head may need to be proportionally smaller for some other functional reason, perhaps related to hemolymph volume demands or muscular constraints. This also fits with our finding that larger animals respond to low P_O2 levels by increasing convection and provides a mechanism for the finding that animals become hypoxic as they near the end of an instar. Overall, our results suggest that the abdominal pumping/tracheal compression behavior of larger caterpillars is a size-based, compensatory mechanism for gas exchange in times of decreased oxygen availability.

Our results are in contrast to those from tracheae in adult beetles (Kaiser et al., 2007) and air sacs in growing grasshoppers (Greenlee et al., 2009), which scale hypermetrically with mass. A study using tracheal diameters from larval, wandering Drosophila melanogaster did not scale with body mass (Henry and Harrison, 2004), although the range of masses was too small to infer scaling relationships. Interestingly, in developing grasshoppers, tracheal scaling of two dorsal transverse tracheae exhibited isometric growth in diameter, but hypermetric growth in length (Harrison et al., 2005). Perhaps tracheae in the head of M. sexta larvae increase in length rather than diameter. Alternatively, proliferation of tracheae smaller than what we could visualize with synchrotron X-ray imaging could help insects to match oxygen delivery needs as body size increases. These data indicate that measures of tracheal diameter alone are not sufficient to determine how tracheae grow during juvenile development and suggest that measures of length changes of tracheae in M. sexta, which routinely grow by 50% of their initial body length within an instar (K.J.G., personal observation), would be useful. Together, these results highlight the need for more studies of tracheal system scaling and suggest that patterns of tracheal investment in larval, holometabolous insects may differ from that of adults and of hemimetabolous insects.

Although caterpillars may not experience hypoxia routinely in nature, it is possible that tracheae are occluded during molting, resulting in localized hypoxia. During experimental hypoxia exposure, body contractions were coordinated to drive gas exchange, as evidenced by the high synchrony between CO₂ emission peaks and both external body movements and tracheal compressions (Fig. 1). The close similarity between abdominal pumping behavior and that observed during molting (Chapman, 1998) provides circumstantial evidence that oxygen limitation may be a signal for ecdysis in insects. Ecdysis, the act of removing the old cuticle, involves three types of abdominal movement: pumping, rotation and peristalsis. During the 20 min ecydysial phase in crickets, abdominal contractions help to extricate the insect from its old cuticle (Carlson, 1977). In pharate moths, eclosion hormone stimulates a central pattern generator that results in abdominal rotations and peristaltic bursts (Truman and Sokolove, 1972). In larval M. sexta, two pairs of motor neurons on each abdominal ganglion are responsible for activating the compression muscles of each segment that contract during ecdysis (Novicki and Weeks, 2000). It is unknown whether these ecdysis-triggering neurons also respond to hypoxia. In grasshoppers, some abdominal neurons are sensitive to hypoxia and increase firing rate in response to low oxygen (Bustami et al., 2002). Finally, in Drosophila, oxygen-sensitive neuronal cells were discovered that are required for larval ecdysis and adult eclosion (Morton et al., 2008). Together, these studies provide evidence for the hypothesis that hypoxia response observed in the older caterpillars mimics their molting behavior.

We would like to thank Jordan Boe, Art Woods, Andy Toth, Elizabeth Stanford, Elizabeth Nyberg and Kathryn Jackson for assistance with data collection, Ashley Zondervan for data analysis, and two anonymous reviewers for their helpful comments.