Quantifying the acid–base status of dragonflies across their transition from breathing water to breathing air

The changes in blood acid–base balance that occur when an animal transitions from breathing water to breathing air were first established from studies on vertebrates – in particular, the amphibians and air-breathing fish that switch between these two respiratory media during their lifetime (e.g. Erasmus et al., 1970; Garey and Rahn, 1970; Just et al., 1973; Lenfant and Johansen, 1968). However, while numerous invertebrates also move routinely between water and air, the respiratory and acid–base consequences of this behavior have been well characterized only among the decapod crustaceans (e.g. Howell et al., 1973; Truchot, 1975). As both vertebrates and crustaceans are ancestrally aquatic, their transition from water to air represents a shift from their ancestral state to a derived condition. In contrast, although insects are an ancestrally terrestrial group, members of at least 12 of the ∼30 insect orders have independently evolved the ability to live and breathe underwater as juveniles (Balian et al., 2008; Wootton, 1988). Thus, these amphibiotic insects provide a valuable opportunity to learn how terrestrial animals have modified their respiratory systems and acid–base balance to function underwater over evolutionary time, and what physiological changes occur as they transition back to breathing air during their development. Lee et al. (2018) examined how hemolymph total carbon dioxide (TCO₂) content changes during development in the amphibiotic dragonflies (Odonata, Anisoptera). The water-to-air respiratory transition of a dragonfly proceeds in stages. Nymphs begin their life as water breathers, using only their tidally ventilated rectal gill to exchange gases with the environment (Tillyard, 1915). However, immediately before metamorphosis, the late-final instar nymphs become bimodal breathers, developing functional mesothoracic spiracles to breathe air while continuing to breathe water using their rectal gill (Corbet, 1962; Gaino et al., 2007). Finally, the adult dragonflies emerge from their final-instar exuviae, shedding their rectal gill in the process, and begin breathing air exclusively through their spiracles (Tillyard, 1915). Studying the change in hemolymph TCO₂ content across these developmental stages revealed that the water-to-air respiratory transition of aeshnid and libellulid dragonflies is accompanied by a much smaller increase in hemolymph TCO₂ (45–48%) compared with the increase experienced by vertebrates (up to 445%) making the same respiratory transition (Lee et al., 2018). The finding that water-breathing dragonfly nymphs have an unusually high hemolymph TCO₂ was unexpected, and is possibly due to their secondarily derived water-breathing capabilities and, presumably, an unusually low water-convection requirement.

The observation that TCO₂ increases significantly as a dragonfly nymph approaches metamorphosis begs the question: what is happening to the acid–base status of the hemolymph? The hemolymph acid–base status describes the equilibrium between CO₂ partial pressure (P_CO2), bicarbonate concentration [HCO₃⁻] and pH, and can reveal whether these parameters change passively or are defended by active pH regulatory mechanisms across the water-to-air transition. However, there is a paucity of information regarding the acid–base status of insect hemolymph generally, let alone for amphibiotic insects, and the few measurements available in the literature are restricted to a small number of species (Matthews, 2017). Therefore, studying the hemolymph acid–base status of dragonflies not only provides further insight into how amphibiotic insects transition between two respiratory media but also expands our knowledge of insect respiratory and acid–base physiology more generally.

To quantify the hemolymph acid–base status of dragonflies across development, the hemolymph CO₂ solubility and apparent dissociation constant of carbonic acid (pK_app) were measured in aeshnid and libellulid dragonflies at multiple developmental stages, from pre-final instar nymphs to adults. Hemolymph samples were extracted and equilibrated with various levels of P_CO2 using a custom-built microtonometer setup before pH and [HCO₃⁻] were measured. From this information, non-HCO₃⁻ blood buffer lines and P_CO2 isopleths were calculated to describe and compare the acid–base status of dragonfly hemolymph as they develop from water-breathing nymphs to air-breathing adults.

Aeshnid and libellulid dragonflies and nymphs were captured in and around ponds at the University of British Columbia, Point Grey campus. They were housed and identified as described previously (Lee et al., 2018). The adult aeshnid dragonflies were identified as Anax junius (Drury 1773) and Aeshna multicolor Hagen 1861. Pre-, early- and late-final instar nymphs of A. junius were also collected and identified. However, only a very few A. multicolor nymphs were collected and, thus, they could not be used for analysis. Libellulid dragonflies were classified as early- and late-final Libellula quadrimaculata Linnaeus 1758 and Libellula forensis Hagen 1861. However, these nymphs could only be identified to species level after hemolymph collection. As the small volumes of hemolymph collected from these nymphs necessitated combining them for analysis, this resulted in some pooled samples being a mix of the two species. Therefore, it was decided to analyze L. quadrimaculata and L. forensis together as early- and late-final Libellula. Unfortunately, adult L. quadrimaculata and L. forensis could not be included in this study because of difficulties in obtaining sufficient hemolymph for analysis.

Hemolymph was obtained from dragonfly nymphs using the same protocol as described in Lee et al. (2018). However, for this study a 22s-gauge 25 μl Hamilton syringe with its needle lumen primed with distilled water was used to extract the hemolymph. Priming the syringe added 1.13 µl of distilled water to the hemolymph sample per extraction, resulting in the sample being diluted by 12%. However, this dilution was unavoidable as priming was necessary to remove the air from the needle lumen, as well as to prevent clogging of the syringe. Hemolymph was collected into an ice-cooled 1.5 ml Eppendorf tube to minimize clotting, and was centrifuged at 1520 g for 10 min to separate any cellular and tissue debris from the supernatant.

Dragonfly adults were restrained on an expanded polystyrene sheet as described previously (Lee et al., 2018), then the 8th abdominal tergite was removed using a razor blade and scissors to expose the underlying hemocoel (Heinrich and Casey, 1978). The gut and heart were pushed to the side using an insect pin, and the hemolymph from the wound was then slowly aspirated into a length of 30-gauge PTFE tubing connected to a 26s-gauge 10 μl Hamilton syringe. The hemolymph was then collected into an Eppendorf tube and processed as described above. Subsequent handling of the hemolymph using a primed syringe resulted in the final analyzed sample being diluted by 7% with distilled water.

To calculate hemolymph CO₂ solubility (α_CO2), 30 μl of hemolymph was collected from an individual insect. This was then centrifuged as described above, then 20 μl of the supernatant was removed and placed in a 6 mm×50 mm test tube together with 20 μl of 0.1 mol l⁻¹ HCl. The hemolymph–HCl solution always had a pH below 3, indicating that all HCO₃⁻ had been converted to CO₂ (Harrison, 1988). A micro-combination pH electrode mounted in a 16 gauge hypodermic needle (MI-414B, Microelectrodes Inc., Bedford, NH, USA) was used for all pH measurements. It was connected to a pH amplifier (FE165, ADInstruments, Colorado Springs, CO, USA) and a PowerLab analog-to-digital converter (PL3504, ADInstruments). The pH electrode was calibrated using three pH standards (pH 4.00±0.01, 7.00±0.01, 10.00±0.01), and pH measurements were recorded using Labchart (v.8.1.10400.0, ADInstruments).

A custom-made microtonometer, 3D printed from ABS plastic filament, was used to equilibrate the hemolymph–HCl sample with gas mixtures of known P_CO2 (Fig. 1). The microtonometer was capable of rotating two test tubes simultaneously, and consisted of a 1000:1 geared-down DC electric motor (Micrometal Gearmotor no. 1596, Pololu, Las Vegas, NV, USA) that turned a 24 mm diameter wheel with two grooves around its perimeter. An elastic band in each groove looped around the middle of a test tube, which was loosely clamped in a frame beneath the motor mount. The motor caused the test tube to rotate in place at 42 rpm, allowing the hemolymph–HCl mixture to smear across the inner wall. Mixing of the sample was further enhanced by the passive tumbling of a 7×2 mm PTFE-coated stir bar placed into each test tube. During equilibration, the microtonometer was held at an angle of ∼20 deg from horizontal with the test tubes partly submerged in a water bath that was temperature controlled to 20°C (F33-ME, Julabo, Seelbach, Baden-Württemberg, Germany). Two 0–500 ml min⁻¹ mass flow controllers (MC500SCCM-D/5M, Alicat Scientific, Tucson, AZ, USA) were controlled by gas mixing software (Flow Vision v.1.3.13.0, Alicat Scientific) to combine 99.998% N₂ and 100% CO₂ compressed gases (Praxair, Mississauga, ON, Canada) into a 20 kPa P_CO2 gas mixture flowing at a combined total flow rate of 400 ml min⁻¹ STPD. Flow rates were verified using a Bios DryCal Definer 220-L primary flow meter that had been calibrated using NIST standards (Mesa Laboratories, Inc., Lakewood, CO, USA). The 20 kPa gas mixture was split into four streams using an air-line aquarium manifold (Accuair 4-way aquarium gang valve, J. W. Pet Company Inc., Arlington, TX, USA), and each stream was humidified by bubbling it through a 2 ml volume of 20°C water before being delivered to the surface of the hemolymph–HCl mixture by a length of 21 gauge Tygon tubing. The microtonometer motor was connected to a 6 V DC power supply, and the hemolymph–HCl mixture was equilibrated with the 20 kPa P_CO2 gas mixture for 30 min. The time required for a sample to fully equilibrate was determined empirically using hemolymph solutions. Initial tests recording the pH change of a 70 μl sample of hemolymph showed that it was fully equilibrated with 0.5 kPa P_CO2 within 30 min. As 70 μl was the largest volume and 0.5 kPa was the lowest P_CO2 used in this study, 30 min was deemed sufficient time for full equilibration of all samples. After equilibration, the test tube was removed from the microtonometer and slotted into a holding rack inside the 20°C water bath. The gas line was placed back into the tube to continue flushing the headspace with the same 20 kPa P_CO2 gas mixture, while the top of the tube was loosely plugged with a cotton ball to hold the tube in position and prevent outside air from reaching the hemolymph sample. A 5 μl sample of the CO₂-equilibrated hemolymph–HCl mixture was extracted and analyzed for TCO₂ using the protocol outlined in Lee et al. (2018). Briefly, compressed N₂ from a gas cylinder (99.998% pure, Praxair) was regulated at a flow rate of 20 ml min⁻¹ STPD through a 0–100 ml min⁻¹ mass flow controller (MC100SCCM-D/5M, Alicat Scientific). This N₂ stream then passed through cell A of a two-channel LI-7000 infra-red CO₂ gas analyzer (LI-COR, Lincoln, NE, USA) to provide a constant zero CO₂ reference. It then flowed through a custom-made gas sparging column that consisted of a cylindrical glass chamber (4.7 ml internal volume), bubbling up through an acid solution [1 ml of 0.01 mol l⁻¹ HCl mixed with 1 μl Antifoam 204 (Sigma-Aldrich, St Louis, MO, USA)] to prevent excessive frothing during sample injections. After passing through the acid solution, the N₂ gas and any liberated gaseous CO₂ was passed through cell B of the infra-red CO₂ gas analyzer for analysis. Injecting a 5 μl hemolymph sample into the sparging column through the gas-tight injection port with a PTFE-lined septum resulted in a bolus of CO₂ being liberated and flushed through the infra-red CO₂ gas analyzer. The area-under-the-curve (AUC) function in the statistical software R v.3.5.0 (http://www.R-project.org/) was used to calculate the volume of liberated CO₂ (TCO₂). To ensure the accuracy of these measurements, a series of sodium bicarbonate standard solutions (10, 15, 20, 25 and 30 mmol l⁻¹) was prepared weekly and run before and after each measurement.

The same protocol as above was used for adult dragonflies; however, 60 μl of hemolymph was pooled from two A. junius, while 60 μl was pooled from three A. multicolor, and 48.6 μl of the supernatant was mixed with 1.4 μl of 0.7 mol l⁻¹ NaF. Only 0.5, 1, 3 and 5 kPa P_CO2 was tested because of the limited volume of hemolymph available for analysis in these samples.

It is generally accepted that blood pK_app is dependent on blood pH in vertebrates and crustaceans (Boutilier et al., 1984). As one of the objectives of this study was to quantify the pK_app of dragonfly hemolymph for comparison with vertebrate and crustacean values, it was necessary to assess whether such a relationship exists in dragonflies. In order to test for a significant effect of pH on pK_app, pK_app was first calculated at each experimental P_CO2 for all measured hemolymph samples from a particular species and developmental stage. The calculated pK_app values for all samples were then plotted against their respective pH on a single graph, and a linear model was fitted to the data to test for a significant relationship between these two variables.

Data were analyzed in R v.3.5.0 (http://www.R-project.org/). A one-way ANOVA was performed to test for any statistical differences between the hemolymph α_CO2 of dragonflies. Linear models were fitted to the pK_app versus pH data to test for any statistical relationships between these variables, and were also fitted to the HCO₃⁻ versus pH data to visualize the non-HCO₃⁻ blood buffer lines and calculate the buffer capacities for the different dragonfly groups. Analysis of covariance (ANCOVA) was performed to test for any statistical differences between the slopes of the non-HCO₃⁻ blood buffer lines of dragonflies. Data are shown as means±s.e.m. unless otherwise stated.

Hemolymph α_CO2 was determined for samples of hemolymph from early- and late-final A. junius nymphs, A. junius and A. multicolor adults, early- and late-final Libellula, as well as for the solution of 0.1 mol l⁻¹ HCl (Table 1). The hemolymph α_CO2 of pre-final A. junius nymphs could not be measured because of a lack of animals; however, statistical analysis showed that there were no significant differences between the hemolymph α_CO2 of any of the dragonflies, nymph or adult, aeshnid or libellulid (one-way ANOVA, F=1.397, d.f.=5, P=0.3). Therefore, the mean of all dragonfly hemolymph α_CO2 was calculated as 0.355±0.005 mmol l⁻¹ kPa⁻¹, and this value was used as the hemolymph α_CO2 of pre-final A. junius.

Measurements of pK_app were made on pre-, early- and late-final A. junius nymphs, A. junius and A. multicolor adults, and early- and late-final Libellula. Pre-final A. junius nymphs, A. multicolor adults, and early- and late-final Libellula nymphs showed a significant relationship between pH and pK_app. Therefore, the equation of the line was calculated for the above four groups, while mean pK_app values were calculated for early- and late-final A. junius nymphs and adults (Table 1).

Comparison of the non-HCO₃⁻ buffer capacities of A. junius (Fig. 2), A. multicolor (Fig. 3) and Libellula spp. (Fig. 4) showed that pre-final A. junius nymphs had a value of 5.4 mmol l⁻¹ pH⁻¹, which was not significantly different from the buffer capacities of both early- and late-final A. junius nymphs (3.6 and 5.8 mmol l⁻¹ pH⁻¹, respectively) (Fig. 5). Anax junius and A. multicolor adults had a buffer capacity of 4.8 and 5.8 mmol l⁻¹ pH⁻¹, respectively, neither of which was significantly different from that of any of the A. junius nymphs. Early-final Libellula nymphs had a buffer capacity that was not significantly different from that of all Aeshnidae (5.3 mmol l⁻¹ pH⁻¹) but this was significantly lower than the value for late-final Libellula (9.9 mmol l⁻¹ pH⁻¹) (Fig. 6), which had the highest buffer capacity of all comparison groups (ANCOVA, F=13.767, d.f.=6, P<0.001).

As a custom-built microtonometer system was used in this study, it was necessary to ensure that our apparatus and CO₂-measurement technique produced accurate values that were comparable to those found by others. The solubility of CO₂ in a 0.1 mol l⁻¹ HCl solution measured here (0.390±0.003 mmol l⁻¹ kPa⁻¹; Table 1) is in excellent agreement with values previously reported for other HCl solutions at 20°C: 0.39 mmol l⁻¹ kPa⁻¹ in 1.0 mol l⁻¹ HCl (Bridges and Scheid, 1982) and 0.38 mmol l⁻¹ kPa⁻¹ in 0.001 mol l⁻¹ HCl (Harrison, 1988).

Findings from the aeshnid dragonflies (Figs 2, 3 and 5) show that at any given P_CO2, the hemolymph [HCO₃⁻] of late-final A. junius nymphs is always higher than that of the pre- and early-final A. junius nymphs, and is instead very similar to that of the air-breathing adults. This is in good agreement with previous measurements of hemolymph TCO₂ in these species, which showed that the TCO₂ of late-final A. junius nymphs was not significantly different from that of the air-breathing adults (Lee et al., 2018), indicating that the hemolymph acid–base status of late-final nymphs is more like that of an air breather rather than that of a water breather. This can be explained by the observation that late-final instar nymphs of both aeshnid and libellulid dragonflies are not exclusively water breathers, but develop the ability to breathe air even while continuing to use their rectal gills (Corbet, 1962; Gaino et al., 2007; de Pennart and Matthews, 2019). The data also show that the hemolymph pH of late-final nymphs is more alkaline at any given P_CO2 compared with that of early-final nymphs (Fig. 5). A higher hemolymph [HCO₃⁻] and pH at a given P_CO2 may indicate a metabolic compensation in response to a respiratory acidosis (Truchot, 1975); however, without knowing the in vivo hemolymph P_CO2 and pH from late-final nymph and adult dragonflies, it is difficult to assess whether the air-breathing life stages maintain a constant pH or allow it to change during the respiratory transition. Future studies will aim to obtain in vivo measurements of P_CO2 and pH across dragonfly development in order to determine whether air-breathing dragonfly late-final nymphs and adults exist in a fully or partially compensated state of respiratory acidosis, relative to early-final nymphs.

Despite the increase in hemolymph [HCO₃⁻] during the respiratory transition of dragonflies, the non-HCO₃⁻ blood buffer capacity of A. junius did not show any statistical differences between developmental stages (Fig. 5). This indicates that the hemolymph of water-breathing pre- and early-final nymphs, the bimodally breathing late-final nymphs, and air-breathing adults is equally resistant to changes in respiratory CO₂. However, this is not the case for the Libellula nymphs (Figs 4 and 6), which show a significant increase in buffer capacity as the nymphs transition from water-breathing early-final nymphs to bimodally breathing late-final nymphs. This indicates that the hemolymph of late-final Libellula is more resistant to respiratory acidosis than that of early-final nymphs. The non-HCO₃⁻ buffer capacity is typically attributed to the presence of proteins and organic acids (Harrison, 2001; Levenbook, 1950b), but without measuring how the composition and concentration of various proteins and other organic molecules change during development in dragonflies, the basis for the difference in non-HCO₃⁻ buffer capacity of Libellula nymphs remains unknown. That libellulid dragonfly nymphs change their hemolymph chemistry differently to the aeshnid dragonflies is in agreement with the large differences in TCO₂ recorded between these families at all developmental stages: adult L. quadrimaculata and L. forensis dragonflies possess a hemolymph TCO₂ that is significantly higher than that of all aeshnid life stages, both nymph and adult (Lee et al., 2018).

Perhaps the most striking finding of the current study is that the hemolymph acid–base status of dragonfly hemolymph across their water-to-air transition lies midway between the blood bicarbonate values that occur across the lifecycle of a vertebrate amphibian: water-breathing dragonfly nymphs have much higher hemolymph [HCO₃⁻] than water-breathing tadpoles, but the terrestrial adult dragonflies have a [HCO₃⁻] that is far lower than that of air-breathing frogs (Figs 5 and 6). Despite these differences in [HCO₃⁻], it can be shown using the in vivo P_CO2 value of 0.9 kPa measured from early-final aeshnid nymphs that the pH of this nymph's hemolymph must be ∼7.8, essentially indistinguishable from the blood pH of a water-breathing tadpole (Erasmus et al., 1970) or crab (Truchot, 1975). Thus, aeshnid dragonfly nymphs appear to be maintaining a hemolymph pH which is not dissimilar from that of other aquatic animals, but are maintaining a higher internal [HCO₃⁻] and P_CO2. Unfortunately, the in vivo pH or P_CO2 measurements necessary to calculate libellulid in vivo pH are currently lacking. However, hemolymph TCO₂ has been measured in early-final L. quadrimaculata nymphs, and this is a reasonable proxy for [HCO₃⁻], given that even at a P_CO2 of 5 kPa, the current data indicate that 93% of the TCO₂ exists as HCO₃⁻. Using this 19.6±1.0 mmol l⁻¹ value (Lee et al., 2018), the in vivo pH of the early-final Libellula nymphs in this study would be ∼7.9, only slightly more alkaline than estimated for A. junius. However, this is substantially higher than the mean hemolymph pH of 7.58±0.074 measured from ‘final’ instar Libellula julia nymphs (Rockwood and Coler, 1991), which could set the lower bound for a reasonable in vivo hemolymph pH for a libellulid nymph.

Calculating the non-HCO₃⁻ buffer capacities from amphibians and crustaceans shows that these values change very little between the water-breathing and air-breathing life stages in both lineages. In Rana catesbeiana, the blood buffer capacity for both tadpoles and frogs is ∼9.5 mmol l⁻¹ pH⁻¹ (Erasmus et al., 1970), while for Carcinus maenas crabs moving between water and air, the buffer capacity is also similar (∼13 to 13.3 mmol l⁻¹ pH⁻¹; Truchot, 1975). The stability of these buffer values indicates that the compensation of a respiratory acidosis during the transition from water breathing to air breathing is mainly achieved by increasing HCO₃⁻ in the blood (Truchot, 1975), rather than by changing the non-HCO₃⁻ buffer system. Anax junius and adult A. multicolor show the same trend, indicating that these dragonflies also mainly rely on metabolic compensation during their water-to-air transition. Libellula nymphs, however, appear to use both the HCO₃⁻ and the non-HCO₃⁻ buffer systems to defend hemolymph pH in the face of elevated hemolymph P_CO2, which is unlike what has been observed in vertebrates and crustaceans.

Comparison of the non-HCO₃⁻ buffer capacities of dragonflies with those measured from other insects shows that they are noticeably lower. The highest buffer capacity of late-final Libellula is only two-thirds of the average value found in grasshoppers, locusts, moth pupae and fly larvae (Harrison, 2001). It may be that these other flying species lack the impressive autoventilation ability of dragonflies (Weis-Fogh, 1967), resulting in larger changes in internal P_CO2 between rest and activity. In addition, many of these species also display discontinuous gas exchange patterns (Buck and Friedman, 1958; Harrison et al., 1995) or live in hypercapnic environments (Levenbook, 1950a), both of which are expected to elevate their hemolymph P_CO2 intermittently. Under such circumstances, the hemolymph would require higher buffer capacity to prevent changes in pH due to the respiratory CO₂, and this may explain the comparatively high non-bicarbonate buffer values in these species.

Studies on the acid–base physiology in insects often make the assumption that the CO₂ solubility and dissociation constants measured from the hemolymph of one species are sufficiently similar to be substituted for those of another (e.g. Gulinson and Harrison, 1996; Harrison, 1989). However, in the absence of additional data, this assumption had not been well tested. This study is the first to provide these constants for an amphibiotic insect lineage, in an attempt to fully describe the hemolymph acid–base status of dragonflies without relying on constants derived from distantly related species. The data presented here indicate that hemolymph α_CO2 does not change appreciably during a dragonfly's water-to-air transition. As hemolymph temperature was held constant during these measurements, it follows that the only other parameter that could influence CO₂ solubility, hemolymph ionic strength or composition (Boutilier et al., 1984), did not change sufficiently between the aquatic and terrestrial life stages to alter solubility. Measurement of the composition of L. quadrimaculata hemolymph shows that its salt concentration (expressed as an equivalent NaCl solution) increases from 170 mmol l⁻¹ in the final-instar nymph to 247 mmol l⁻¹ in the mature adult (Nicholls, 1983), which would cause CO₂ solubility to decrease by a trivial 0.007 mmol l⁻¹ kPa⁻¹ (Heisler, 1986). Comparison of α_CO2 values calculated from these two previous hemolymph salt concentrations using the equations given by Heisler (1986) and more recently Stabenau and Heming (1993), with the mean α_CO2 values for early- and late-final Libellula nymphs, shows very good agreement: the calculated value for unspecified ‘final’ instar nymphs is 5.2% higher than the measured value for early-final nymphs, but the calculated value for the adult dragonfly is within 0.5% of the measured late-final nymph's value. A hemolymph molarity of 206 mmol l⁻¹ measured from Aeshna grandis and 195 mmol l⁻¹ from Aeshna cyanea nymphs, some of which were ‘final instar’ (Sutcliffe, 1962), give α_CO2 values that are only 0.8% and 1% higher, respectively, than that reported here for early final A. junius nymphs. The measured α_CO2 values in the hemolymph of dragonfly nymphs and adults are also in good agreement with values determined for grasshoppers and locusts, with Melanoplus bivittatus having a hemolymph α_CO2 of 0.34 mmol l⁻¹ kPa⁻¹ at 20°C (see Table 1), which is similar to values determined for vertebrate plasma (Harrison, 1988). Thus, the current data support the general trend that plasma/hemolymph α_CO2 is similar across a wide variety of animal lineages for a given temperature and ionic strength.

The pK_app of carbonic acid calculated from early-, late-final and adult A. junius in this study (6.23–6.27) brackets that obtained from M. bivittatus (6.24 at 20°C; Harrison, 1988), and the range of pK_app calculated from pre-final A. junius, and early- and late-final Libellula also overlaps with these values. These results show that the pK_app of dragonfly hemolymph is similar across water- and air-breathing life stages, as well as between families. However, it is surprising that pH does not appear to have a consistent effect on pK_app in the above dragonfly groups, as data from vertebrates indicate either a non-linear (Boutilier et al., 1984) or a negative linear relationship (Iversen et al., 2012) between pH and pK_app. Three of the A. junius groups do not show pH dependence of pK_app, and while pre-final A. junius, adult A. multicolor, and early- and late-final Libellula show a statistically significant effect of pH on pK_app, only early-final Libellula shows the same negative trend as described previously (Boutilier et al., 1984; Iversen et al., 2012). These results are in contrast to those reported previously from vertebrates, as well as with the observation that pH does not affect pK_app in grasshoppers and locusts (Harrison, 1988). However, while the relationship between pK_app and pH may be statistically significant for the data presented here, it remains to be seen whether it is biologically relevant.

The number of insects that have adapted to living and breathing underwater is vast, yet we are only just beginning to understand the respiratory and acid–base physiology of these animals. The aquatic nymphs of both dragonfly families studied here show very similar non-HCO₃⁻ buffer capacities, although [HCO₃⁻] is higher in the libellulid nymphs. Upon their transition to bimodal gas exchange, these groups diverge, with both showing additional increases in [HCO₃⁻], but with the aeshnid dragonflies maintaining their buffer capacity, while the buffer capacity of late-final libellulid nymphs undergoes a dramatic increase. This variation in the hemolymph non-HCO_3­⁻ buffer capacities between two families in the Anisoptera, and the large differences seen between dragonflies and other terrestrial insects, shows that this parameter varies widely within Insecta according to the insect's respiratory physiology, habitat and behavior. Finally, it remains to be seen whether the elevated P_CO2 and [HCO₃⁻] observed in the hemolymph of secondarily aquatic dragonfly nymphs, relative to that of ancestrally water-breathing animals, is a trait specific to anisopterans, or whether this is a pattern that will be seen in water-breathing insects in general or even other secondarily aquatic animals.

We thank the University of British Columbia Botanical Garden and Dolph Schluter for allowing us to collect dragonflies from in and around their ponds. We would also like to thank the two reviewers, whose constructive criticism greatly improved this paper.