Respiratory function of the plastron in the aquatic bug Aphelocheirus aestivalis (Hemiptera, Aphelocheiridae)

Insects rely primarily on their tracheal system to exchange respiratory gases with the atmosphere (Wigglesworth, 1931). The system consists of air-filled tubes that open to the air through spiracles on the surface of the thorax and abdomen, branch throughout the body, and eventually reach the cells with blind-ended tracheoles. Oxygen (O₂) and carbon dioxide (CO₂) move through the smaller tubes largely by diffusion down gradients in partial pressure (P_O2, P_CO2), although the outer tracheae are connected and can be ventilated by movements of the body (Harrison et al., 2013). Many aquatic insects still use the gas-filled tracheal system, but interface it with snorkels, tracheal gills or air bubbles on their bodies that act not only as an O₂ store, but also as a ‘physical gill’ capable of exchanging O₂ and CO₂ with the water (Seymour and Matthews, 2013). Physical gills are classified into two types: ‘compressible’ or ‘incompressible’, depending on whether the air–water interface is unsupported or supported, respectively (Flynn and Bush, 2008). Compressible gas gills decrease in volume during a dive, because: (1) O₂ is removed by respiration; (2) CO₂ does not replace the O₂ removal as it easily dissolves in water; and (3) nitrogen (N₂) dissolves in the water because its partial pressure (P_N2) increases over the P_N2 of the surrounding water because of the decrease in P_O2 in the gill (Ege, 1915; Rahn and Paganelli, 1968). The limited lifetime of compressible gas gills forces the insects to renew the bubble periodically at the surface. Incompressible gas gills, defined as ‘plastrons’ by Thorpe and Crisp do not collapse, because the air–water interface is supported by hydrophobic cuticular hairs that push against it (Marx and Messner, 2012; Thorpe and Crisp, 1947a). Oxygen can diffuse from the water into the plastron gas space and from there through spiracles into the tracheal system of the insect. The physical gill is permanent, and the insects never have to surface if the water is sufficiently oxygenated.

Plastron respiration has been considered sporadically for about a century (Balmert et al., 2011; Ege, 1915; Flynn and Bush, 2008; Hinton, 1976; Messner and Adis, 1995; Thorpe, 1950). A uniform picture of the anatomy, physiology and physics of this interesting system emerges, indicating that the hydrophobic surface of the plastron matches the hydrostatic pressure with surface tension of the curved menisci on the plastron. Dissolved O₂ taken up by the insect reduces the P_O2 in the plastron gas, but P_N2 does not increase as it does in collapsible gas gills. Instead, plastron P_N2 remains in equilibrium with the water and the total pressure in the plastron gas becomes less than the combined atmospheric and hydrostatic pressures outside the plastron. The decrease in total pressure is roughly equivalent to the decrease in P_O2.

Conclusions about plastron function have been largely theoretical, although good measurements of plastron structure of the river bug Aphelocheirus aestivalis (Hinton, 1976) and rates of O₂ consumption have been measured for several species of plastron bugs (Kölsch and Krause, 2011; Thorpe and Crisp, 1949; Verberk and Bilton, 2015). According to a recent analysis (Seymour and Matthews, 2013), respiration rates of resting bugs range from 27 to 55% of the prediction of resting metabolic rates based on an interspecific allometric analysis of 391 species of insects in general (Chown et al., 2007). Such low values might be related to diffusion limitation through the layer of water adjacent to the plastron. This is the ‘boundary layer’, which is calculated according to the Fick general diffusion equation as the thickness of an imaginary layer of completely stagnant water next to the exchange surface. Thus the measured rates of O₂ consumption could be satisfied with effective boundary layers of approximately 100 to 800 μm thick in air-equilibrated water (Seymour and Matthews, 2013). It is expected that the effective thickness would decrease if the insect swam through the water or lived in moving water. However, there are no direct measurements of boundary layer thickness in any aquatic insect. It is important to know the boundary layer thickness because several models of gas exchange of aquatic insects assume that it is very thin (Chaui-Berlinck and Bicudo, 1994; Chaui-Berlinck et al., 2001; Rahn and Paganelli, 1968).

Actual measurements of P_O2 in the boundary layer or within plastron gas have not been possible, because the volume of the gas is miniscule. In particular, anatomical data from the river bug A. aestivalis indicate that the plastron hairs are only 3 μm long and the entire gas space between the hairs of the plastron is only about 0.14 mm³ (calculated in this study from data in Hinton, 1976). Despite this problem of size, we were able to measure the P_O2 across the boundary layer and inside the plastron gas of A. aestivalis. The technique involves a modification of O₂-sensing fibre optics (optodes) that have been used to measure P_O2 in collapsible gas gills of other diving insects with collapsible gas gills (Matthews and Seymour, 2006, 2008, 2010). We also investigate the interaction between swimming activity, respiration rate and boundary layer thickness over the surface of the plastron.

The mean body mass of 15 adult A. aestivalis was 43.5±3.2 mg (mean±95% confidence interval, CI). There was no significant difference between 8 males and 7 females (t-test; P=0.39). Mean body length was 9.40±0.19 mm. The surface area of the silhouette from above, without the head and appendages, was 47.5±1.8 mm². Males had a slightly, but significantly, smaller silhouette (45.7 mm²) than females (49.5 mm²) (P=0.016). The bugs were highly dorsoventrally flattened (Fig. 1), so assuming that they were perfectly flat, the total surface area of the plastron was taken as twice the area of the headless silhouette, or 95 mm². Hinton (1976) measured the structure of the plastron of A. aestivalis, concluding that the hairs were 3 μm high, 0.4 µm in diameter and their centres spaced 0.5 μm apart. These data indicate that the air fraction is 0.5. Our measured surface area of 95 mm² multiplied by the thickness of 0.003 mm² gave a total plastron volume of 0.28 mm³ and an air space of 0.14 mm³.

We do not know the total volume of the tracheal plus plastron system accurately, but can estimate the maximum volume by assuming that the density of the air-less body is 1.078 g ml⁻¹, which is the measured density of the aquatic bug Anisops deanei (Matthews and Seymour, 2008). Because A. aestivalis was observed to be negatively buoyant, the total air space is somewhat less than 3.39 mm³ or 8.4% of body volume.

When the bubble on the tip of the optode sheath contacted the plastron, there was an immediate change in P_O2 that progressed toward a relatively stable value (Fig. 2). The speed of this change was related to how fast the bubble was absorbed, with equilibration occurring generally between 1.5 and 4 min. Although it was thought that placing a bubble of nitrogen instead of air on the optode tip would speed up equilibration, it was not always evident in the traces. The most successful traces occurred when equilibration coincided with complete absorption of the bubble and the sheath was contacting the plastron. Sometimes the lowest P_O2 occurred before the bubble was completely absorbed, in which case we waited until it was absorbed and then recorded the value. Data were discarded if equilibration was incomplete when the bubble disappeared totally.

In normoxic water (∼20.7 kPa), the difference in equilibrium P_O2 between the water and the plastron was 10.6±2.5 kPa (mean±CI) in 8 tethered bugs with high convection, 10.2±1.7 kPa in 12 bugs with moderate convection and 18.7±0.9 kPa in 8 bugs in stagnant water (Fig. 3). There was no significant difference between high and moderate convection (P=0.38), but plastron P_O2 in stagnant water was lower than both other treatments (P<0.0001; t-tests). The effect of ambient P_O2 was tested only with moderate convection. The slope of the linear regression for data for 32 points for aquatic P_O2 between 14 and 32 kPa was slightly less than that for equality (P_O2,_plastron=0.82 P_O2,_water−7.3; r²=0.75).

Boundary layer thickness was evaluated from P_O2 transects above the plastron surface in a miniflume by observing where the P_O2 began to decrease from ambient levels in the free water (Fig. 4). This began at approximately 500 μm away from the ventral sternite surface in stagnant water, 200 μm in water flowing through the flume at 92 ml min⁻¹ and 100 μm in water flowing at 256 ml min⁻¹. The velocity of the water over the surface of the bugs is not known precisely, but the mean velocity across the cross section of the empty flume was approximately 2.5 mm s⁻¹ and 5.8 mm s⁻¹, respectively. Transects of the boundary layer on the dorsal side tergites in stagnant water were similar to those on the ventral surface, beginning to decrease at approximately 600 μm away and dropping to 6.3±0.8 kPa (mean±CI) at the surface.

Respiration rates of 14 A. aestivalis in the present study were standardized to 20°C from data taken between 20 and 23°C assuming a Q₁₀ of 2.3 (Fig. 5). We could not control the activity of the insects in the respirometer, but the lowest rates were from mainly stationary animals and the highest rates were from bugs that were swimming almost constantly. There was no significant difference between the overall means of the present study and Thorpe and Crisp, but their data appear overall slightly higher (Thorpe and Crisp, 1947b). Our lowest rate was 0.66 nmol s⁻¹ g⁻¹ (wet mass; =32.6 pmol s⁻¹) and the highest 6.73 nmol s⁻¹ g⁻¹ (=296 pmol s⁻¹), giving a maximum factorial metabolic scope of approximately 9.1. By comparison, Thorpe and Crisp's data were 1.67 and 9.92 nmol s⁻¹ g⁻¹, respectively, giving a scope of 5.9. A recent study of metabolic rate in inactive A. aestivalis at 5–15°C extrapolates to 2.62 nmol s⁻¹ g⁻¹ (=100 pmol s⁻¹) at 20°C (Verberk and Bilton, 2015).

The data gathered in this study allow us to evaluate plastron gas exchange in Aphelocheirus aestivalis according to Fick's general diffusion equation: Ṁ_O2=G_O2(P_O2 out−P_O2 in), where Ṁ_O2 is the rate of O₂ uptake of the whole animal (pmol s⁻¹), G_O2 is the O₂ conductance of the gas exchange surface (pmol s⁻¹ kPa⁻¹) and (P_O2 out−P_O2 in) is the difference in P_O2 (kPa) between the free water and the plastron gas. G_O2=K_O2(A/L), where K_O2 is Krogh's coefficient of O₂ diffusion in water (constant at 0.277 pmol s⁻¹ kPa⁻¹ cm⁻¹ at 20°C) (Seymour, 1994), A is the surface area of the air–water interface plastron (constant at 0.95 cm²) and L is the thickness of the boundary layer (cm). The exchange surface area is less than the surface area of the plastron because the interface is supported by hairs that take up some of the area. This is estimated from direct measurements of the plastron morphology for A. aestivalis (Hinton, 1976). Each square mm of the surface has about 4,000,000 hairs that rise perpendicularly from the cuticle and make a right angle turn at the top. The hairs taper, such that the average width of the bent tip on the surface of the plastron is 0.15 μm. Assuming that the hairs form a grid with 2000 rows of hairs in each square mm, the area covered by hairs is 0.15×1000×2000=300,000 μm² (width of hair×length of row×number of rows). This is 30% coverage of hairs, leaving 70% of area between the hairs, or equivalent to 0.66 cm² for the whole insect. Different packing and overlap of hairs or effects of meniscus shape might change this fraction a little, but this seems to be a reasonable approximation. Independently, the free area for exchange in adult A. aestivalis was estimated to be 0.60 cm² (Thorpe and Crisp, 1947b).

The relationships between Ṁ_O2, P_O2 difference and boundary layer thickness indicate that a higher Ṁ_O2 requires a proportionally higher P_O2 difference or an inversely proportional boundary layer thickness (Fig. 6). However, there is a maximum P_O2 difference that apparently can occur without reducing metabolic rate. Although we have not measured the critical P_O2 below which Ṁ_O2 decreases in this species, it is 2 kPa in aquatic waterboatmen Agraptocorixa eurynome (Matthews and Seymour, 2010), which is equivalent to a P_O2 difference of approximately 19 kPa. This seems reasonable because Aphelocheirus is reported to survive P_O2 around 5% of saturation (∼1 kPa) in running water of 15°C (Marten et al., 1994). The mean P_O2 difference in A. aestivalis is close to this limit: 18.7 kPa in stagnant water. Taking the endpoints of the range of Ṁ_O2 measured in this study, namely 32.6 and 296 pmol s⁻¹, and the P_O2 difference of 18.7 kPa, the boundary layer thickness would be 1057 and 116 μm, respectively (these points appear on Fig. 6). It is evident from this analysis that a boundary layer 500 μm thick, as measured in optode transects in stagnant water (Fig. 4) can support Ṁ_O2 up to approximately 70 pmol s⁻¹ before the P_O2 difference reaches 19 kPa. Resting Ṁ_O2 is about half of 70 pmol s⁻¹ (Fig. 4), so resting metabolic rates should be achievable in stagnant water if the P_O2 is above 10 kPa.

Active metabolic rates require convection to reduce the boundary layer thickness. The effect of convection on effective boundary layer above plastrons has been analysed theoretically, but not experimentally (Flynn and Bush, 2008). Our data indicate that with a thickness of 200 μm, Ṁ_O2 could rise to approximately 175 pmol s⁻¹, and with 100 μm, it could rise above 300 pmol s⁻¹ (Fig. 6), which is about the maximum rate we ever measured (=6.82 nmol s⁻¹ g⁻¹) (Fig. 5). Boundary layers of approximately these thicknesses occurred in the flume measurements when the flow rate was between 92 and 256 ml min⁻¹ (Fig. 4). Unfortunately we do not know the velocity distribution of the water within the hydraulic boundary layer; however, the mean velocity of the water in the flume without the insect was about 2.5 mm s⁻¹ and 5.8 mm s⁻¹, respectively. The bugs can swim faster than this, so it is apparent that simply moving through the water reduces the effective boundary layer enough to prevent diffusion limitation of respiration at the highest levels we measured. An increased mobility of Aphelocheirus was reported in aquaria when the water current was interrupted (Larsén, 1931).

The thicknesses of the boundary layers that we measured and calculated are considerably larger than previously indicated in the literature. Thorpe and Crisp modelled the oxygen diffusion cascade through four layers of the plastron–tracheal system: (1) the external boundary layer; (2) vertically and laterally in the plastron gas; (3) the tracheal system to the ends of the tracheoles; and (4) through the tissues (Thorpe and Crisp, 1947b). They assumed a resting O₂ consumption rate of 6 mm³ h⁻¹, which is equivalent to 74 pmol s⁻¹. For this rate, they calculated a P_O2 drop across several layers of the oxygen cascade: external boundary layer next to the plastron (0.7 kPa), across the plastron to spiracles (0.1 kPa), spiracles to tracheoles (0.3 kPa), through tissues to mitochondria (2.0 kPa), adding to a total of 3.1 kPa for whole oxygen cascade. The obvious problem with their calculations is the gross underestimation of the boundary layer thickness. Our experiments in stagnant water show that a Ṁ_O2 of 74 pmol s⁻¹ would create a boundary layer of 466 μm and a P_O2 difference of 18.7 kPa (Fig. 6), rather than 0.7 kPa. If we ignore Thorpe and Crisp's calculated decrease of 0.7 kPa across the boundary layer, the decrease in the plastron–tracheal and system–tissue layers totals 2.4 kPa. This almost exactly matches the mean plastron P_O2 (2.0 kPa) that we measured in air-equilibrated stagnant water (Fig. 3). We conclude that the calculations of the oxygen cascade within the plastron–tracheal and system–tissue layers calculated by Thorpe and Crisp are reasonable, but the boundary layer is much too thin. Higher metabolic rates due to activity would be expected to increase the P_O2 difference between the plastron and the tracheoles, but would also increase plastron P_O2 by reducing the effective boundary layer thickness.

There is a technical question of how the gas in the optode sheath and applied bubble affected the measurements of plastron P_O2. The volume of air in the optode sheath was calculated from the internal dimensions of the sheath less the volume of the optode as 0.236 mm³. The bubbles were approximately the same outside diameter of the sheath and therefore nominally 0.113 mm³. The total volume of air associated with the optode was therefore approximately 0.349 mm³. By comparison, the total volume of the tracheal and plastron gas is calculated above to be about 10 times larger, at 3.39 mm³. Assuming that the total pressure in the plastron is 80 kPa and equal to that in the tracheal system, and normal barometric pressure is 101.3 kPa, then the potential air volume associated with the pressure difference is up to 0.713 mm³ (3.39×21.3/101.3), which is about six times larger than a 0.113 mm³ bubble. Therefore the absorption of several optode bubbles can be explained by expansion of the sub-atmospheric plastron-tracheal system when the bubble made contact with the plastron gas.

Three additional arguments indicate that the effect of the optode gas was minor. First, the application of the bubble required about 2 min to reach equilibrium (Fig. 2). During this time, the respiration rate of a resting bug consumes about 0.19 mm³ of oxygen, which is more than the estimated maximum of 0.14 mm³ of oxygen in the entire optode–plastron–tracheal system when the optode was applied. (This is calculated assuming that the P_O2 in the 3.39 mm³ plastron–tracheal system is 2 kPa, yielding 0.07 mm³ of O₂ plus 0.349 mm³ of gas in the bubble and optode sheath at a P_O2 of 21 kPa, yielding another 0.07 mm³ of O₂.) Second, the traces appeared to plateau when the rate of consumption equalled that entering the plastron from the water (Fig. 2), so equilibrium was reached at the same level as occurred before the optode bubble was applied. Third, by the time equilibrium was reached, the size of the bubble had disappeared and the sheath contacted the plastron. In a separate set of experiments, we measured the O₂ conductance of naked bubbles on the end of the optode (our unpublished data). Conductance was high on large bubbles, but decreased to practically zero when the sheath contacted a gas-impermeable surface. This indicates that the P_O2 represented that in the plastron and was not significantly influenced by aquatic P_O2.

Diving insects that use a plastron are under severe constraints because of the limitations of gas exchange inherent in this mechanism. There are several lines of evidence for this idea. First, true plastron insects are generally small, because scaling principles show that metabolic rate generally increases faster than surface area as insects increase in size, reaching a limit set by surface area (Seymour and Matthews, 2013). This applies to species that never have to surface to replenish the bubble, including some beetles (Madsen, 2008, 2012). A. aestivalis and some other species of Aphelocheirus may be the largest plastron insects at approximately 40 mg body mass (Xie and Liu, 2014). Second, resting metabolic rates of true plastron insects are in the region of one-quarter to one-half of the rates predicted from most insects of the same body size and temperature (Seymour and Matthews, 2013). Third, plastron insects tend to occur in well-oxygenated, flowing water (Hutchinson, 1981), which reduces the effective boundary layer thickness and its resistance to diffusion. In particular, A. aestivalis live typically in moving streams (Basu et al., 2013; Hoffmann, 2008; Miguélez and Valladares, 2010). In stagnant water, the measured boundary layer thickness in A. aestivalis is about 500 μm, which is thin enough to satisfy a bug with respiratory demands twice the level of the minimum that we measured. Therefore, resting bugs could withstand P_O2 levels in stagnant water at approximately half of air-saturation. Even if the insects find themselves in more severely O₂-depleted stagnant water, they can survive by moving, either to reduce the boundary layer, to enter water near the surface with higher P_O2 or to obtain O₂ from the surface of aquatic vegetation (Pedersen and Colmer, 2012). If O₂ becomes very low, A. aestivalis moves near the surface and can even voluntarily emerge from the water (Thorpe, 1950). Finally, the critical thermal maximum of A. aestivalis decreases in aquatic hypoxia (P_O2=5 kPa), whereas that of a comparable air-breathing bug Ilyocorus cimicoides does not (Verberk and Bilton, 2013, 2015).

Aphelocheirus aestivalis (Fabricius 1794) Hemiptera were captured in small rivers in the northern part of Germany (Warnow River as well as in the northern run off of the Schweriner See). The summer water temperature was 22 to 23°C and oxygen saturation was between 70 and 90%. Free water flow rate was approximately 0.1 to 0.4 m s⁻¹. The bottom of the rivers was covered with gravel and small stones and was similar to the biotopes of Aphelocheirus described elsewhere (Larsén, 1931; Ussing, 1910). The population density was very high and up to 20 adult individuals could be caught on a patch of 30×20 cm. They were maintained in small holding aquaria (∼50 litres), which were aerated and filtered. They were fed living red blood worms (Chironomus sp.) and small gammarid crustaceans (Hyalella sp. and Gammarus sp.), which form part of their natural diet (Lemb and Maier, 1996).

Bugs were removed from the holding aquarium, blotted lightly with facial tissue and weighed to within 0.1 mg on a balance (model 1201 Sartorius GmbH; Göttingen, Germany). They were photographed from above, against a grid and the length and area of the silhouette was determined with ImageJ (Version 1.47, Wayne Rasband, NIH, USA). Legs of the bugs were removed from the photographs to produce a body outline. Each photograph was converted into a binary image where particle analysis was used to determine the surface area of the bugs' silhouettes. Proportion of surface area attributed to the head was removed as the plastron does not occur there.

The bugs were inverted and attached to a bent wire stand with a tiny drop of cyanomethacrylate adhesive on the thorax between the two wing buds. A cloth mesh, about 12 mm square with six openings per centimetre, was used to restrain the legs by cutting four to five strands in a central slit and pulling the mesh over the abdomen toward the head. The anterior end of the mesh was secured above the head on two wires attached to the stand. The wire stands were placed on an inverted beaker in a small aquarium (30L×20W×16H cm), so that the bugs were approximately 1 cm below the water surface. The aquarium had a sintered glass gas diffuser on one side that stirred the water and equilibrated it with selected gas mixtures. Pure oxygen and nitrogen were mixed with mass flow controllers [models 1179A (200 ml min⁻¹) and 1259B (1000 ml min⁻¹), MKS Instruments, Andover, MA, USA] and a custom-made control unit. The surface of the water was covered with a polystyrene raft that cut off most of the exposure to air except for an 8-cm-diameter opening above the insects. It was possible to control the gas flow through the diffuser to create different convective environments around the insect described qualitatively as ‘stagnant’ when the diffuser was stopped after equilibration, ‘moderate convection’ with a gas flow rate of 460 ml min⁻¹, and ‘high convection’ at 920 ml min⁻¹. The system was allowed to equilibrate for 30 min between P_O2 treatments, 10 min between convection treatments and 10 min between boundary layer measurements. P_O2 was measured continuously in the water with one optode, while P_O2 in the plastron gas was measured with another optode.

The oxygen-sensing, glass fibre-optic optodes were fast-response, tapered tipped fibres 140 µm in diameter and 20 mm of exposed length (sensor model B2, meter model TX-3 with thermal compensation, PreSens Precision Sensing GmbH, Regensburg, Germany). The optode output was logged at 1 s intervals by Presens software in the Microsoft Windows environment. The optodes were mounted in a Pasteur pipette held in a 3D micromanipulator above the aquarium. They were calibrated with the manufacturer's software by equilibrating them in a separate chamber with pure nitrogen (Air Liquide, UN 1066) and atmospheric air, taking into account the barometric pressure. The calibration was done with temperature compensation. For measurement of P_O2 inside the plastron, an optode was equipped with a 5 mm polyethylene tubing sheath (OD=0.60 mm, ID=0.25 mm) positioned such that the optode tip was inside the sheath about 1 mm from the end (Fig. 1 inset). The sheath was sealed to the optode at the proximal end with a thick layer of petroleum jelly.

The protocol for measuring plastron P_O2 involved submerging the optode under the water and placing a bubble of air or N₂ on the end of the sheath with a 50 μl gas-tight syringe (Hamilton Bonaduz, Switzerland). The bubble joined the air inside the optode sheath. The optode was then lowered until the bubble pushed against the ventral plastron of the bug, immediately over a spiracle. Successful measurements were characterised by immediate shrinkage of the bubble, which indicated that a gas contact was made between the bubble and the gas in the plastron. This was accompanied by a rapid change in optode P_O2, particularly if the composition of the initial bubble on the optode was quite different from that in the plastron gas (Fig. 2). The initial volume of the bubble was important to control, in order to match the rate of bubble shrinkage to the rate of equilibration of the system. Bubbles that were initially 0.60 mm in diameter functioned well. As equilibration approached, the optode sheath was gradually lowered until it eventually rested almost flat on the plastron such that the gas around the optode was maximally exposed to plastron gas and minimally exposed to the ambient water. This situation did not last long, because the air connection between the plastron and optode quickly broke. This is the reason why continuous measurement of plastron P_O2 was not possible.

To evaluate how the bubble used to attach the optode to the plastron was affected by diffusion to or from the water, the sheathed optode was set up in an aquarium, just like the real experiment, except that the insect was replaced by a horizontal piece of solid styrene plastic. A bubble of atmospheric air was placed on the sheath and exposed to aquarium water equilibrated to a P_O2 of 31.7 kPa. This was performed with different-sized bubbles, different levels of convection and a range of exposure to the water (whole bubble and bubbles contacting the plastic to several extents, including those similar to the experimental contact with the plastron when the bubble was about to disappear).

All measurements were made at water P_O2=20 kPa, nominally. A tapered optode without sheath was fixed to a micromanipulator and advanced so that the tip touched the surface of the plastron. After equilibration, it was pulled away in steps of 0.05 mm until the tip was 0.2 mm away and then in 0.10 mm steps until it reached 1.0 mm (outside the boundary layer). Boundary layer curves were obtained from the ventral (sternite 4; above the spiracle) and dorsal (tergite 2, about 30% from the midline) plastrons in stagnant water, from the ventral plastrons in the aquarium with moderate convection generated by the gas diffuser, and from ventral plastrons in a miniflume at two rates of water flow. The flume was milled from a piece of acrylic plastic 100 mm long×39 mm wide×35 mm high. The dimensions of the water course were 80 mm long, 20 mm wide and 16–19 mm deep (depending on flow rate). The first 28 mm contained a fabric screen and drinking straws to smooth the flow, followed by 52 mm of free water with the insect in the middle. A small pump from the aquarium flowed water through the flume and the flow rate was measured by collecting water in a graduated cylinder over a known period.

Closed-system respirometry was used to measure rate of oxygen uptake from individual bugs. This involved a 20 ml glass vial sealed with a thick rubber stopper designed to be pierced by hypodermic needles. The thin end of a Pasteur pipette was pushed through the stopper to create a tight fit. The vial was filled with aquarium water without any visible bubbles, the bug was admitted and then the vial was stoppered under water, which pushed a column of water into the pipette. With the vial lying horizontally, an optode was inserted into the pipette so that the tip was inside the chamber. There was little space between the optode covering and the glass, and this was filled with an approximately 5 cm length of water, so that diffusion was negligible. The chamber water was well mixed at the beginning of the run, and was mixed from time to time by the swimming insects. Nevertheless, at the end of each 60 min run, the vial was tilted repeatedly which caused the bug to fall back and forth, thus quickly mixing the water for the final measurement. A straight line was plotted between the P_O2 of the mixed water initially and finally in the chamber to obtain the rate of oxygen uptake according to the equation: Ṁ_O2=Vβ(ΔP_O2/ΔT) where Ṁ_O2 is the metabolic rate (µmol s⁻¹), V is the volume of the chamber (ml), β is the oxygen capacitance of the water (e.g. 0.013 µmol ml⁻¹ kPa⁻¹ at 23°C) and (ΔP_O2/ΔT) is the rate of change of P_O2 (kPa s⁻¹). Two blank runs were made at the beginning and end of the day on each chamber and the background respiration (∼17% of the total respiration) was subtracted from the bug respiration. A second set of measurements was performed on six bugs at 20°C in a separate constant volume respirometry system. A glass syringe was equipped with a Clark-type oxygen electrode (E5047, Radiometer, Kopenhagen, Denmark) connected to a custom made amplifier. The cathode was covered with two 25 µm Teflon membranes in order to reduce the oxygen consumption of the electrode itself. The electrode was pushed forward in the syringe to reduce the volume of water to 2.65 ml. This small volume ensured a complete mixing during the measurement. The whole setup was submerged in a water bath kept at 20°C. The oxygen consumption rate of the aquarium water itself was determined to be less than 2 pmol s⁻¹ (0.05 nmol g⁻¹ s⁻¹ for a 40 mg Aphelocheirus). There was no significant difference between the two techniques, so the data were combined.

We appreciate the technical assistance of Maik Kunert who helped with technical problems and Lisa Schilha and Ramona Voss who helped to catch Aphelocheirus in the field. The removal of the animals complied with the laws of the German federal state of Mecklenburg-Vorpommern.