Cutaneous respiration by diving beetles from underground aquifers of Western Australia (Coleoptera: Dytiscidae)

Insects evolved in a terrestrial environment, as indicated by their gas-filled respiratory system (Pritchard et al., 1993). The tracheal system is a network of branching tubes that run from small openings (spiracles) on the outside of the insect's body through to the tissues to bring O₂ into close proximity with the mitochondria and allow CO₂ to move out. O₂ is able to diffuse through air 250,000 times faster than through water (Dejours, 1981) and, in combination with gas convection, the tracheal system of flying insects supplies the highest known mass-specific metabolic rate of any animal (Snelling et al., 2017). However, insect species in several groups, including Ephemeroptera, Plecoptera, Megaloptera, Trichoptera, Odonata, Diptera, Coleoptera and Hemiptera, have become aquatic secondarily, providing challenges for the gas-filled respiratory system (Foottit and Adler, 2009; Hutchinson, 1981).

Solutions to these challenges include air stores, which are bubbles taken underwater from which O₂ can be consumed during a dive, and gas gills, bubbles with a surface exposed to the water through which O₂ can diffuse, thus extending the dive. There are two types of gas gill: compressible and incompressible (plastrons). Compressible gas gills require replenishment at the surface as a result of both O₂ consumption by the insect and N₂ and CO₂ loss from the bubble to the water (Ege, 1915; Jones et al., 2015; Rahn and Paganelli, 1968; Seymour and Matthews, 2013). However, plastrons can be sustained indefinitely because of structures that prevent collapse of the bubble–water interface (Jones et al., 2018; Marx and Messner, 2012; Seymour and Matthews, 2013; Thorpe and Crisp, 1947a).

In addition to air stores and gas gills, many aquatic insects use cutaneous respiration. Some use tracheal gills, including mayfly, damselfly, stonefly and caddisfly larvae, and some small diving beetles with setal tracheal gills; in others, diffusion occurs through the unelaborated body surface, like the fifth instar of the aquatic bug Aphelocheirus, numerous dipteran larvae, and gill-less caddisfly and stonefly larvae (Bäumer et al., 2000; Eriksen, 1986; Kehl and Dettner, 2009; Morgan and O'Neil, 1931; Rasmussen, 1996; Thorpe and Crisp, 1947b; Verberk et al., 2018; Wichard and Komnick, 1974). Species that utilise cutaneous respiration contend with two significant types of resistance to diffusion in the O₂ cascade from the water to the tissues: the boundary layer (a water layer above a respiratory surface deficient in O₂ that provides resistance to diffusion) and the cuticle. In plastron breathers, the boundary layer can provide the greatest resistance in the O₂ cascade (Seymour et al., 2015). Consequently, plastron breathers are restricted to a small size, and have lower metabolic rates than insects in general, with Aphelochirus being the largest known species at 40 mg (Seymour and Matthews, 2013). This size limitation is associated with the relationship between metabolic rate, surface area and boundary layer thickness. The O₂ diffusion rate is, in part, affected by the quotient of surface area over boundary layer thickness, and as the insect becomes larger, the surface area relative to O₂ demand decreases, requiring the insects to have a thinner boundary layer to satisfy O₂ demand or a lower metabolic rate to avoid O₂ limitation, or both.

It is, therefore, anticipated that cutaneously respiring insects, without cuticular elaboration, would have lower metabolic rates and be smaller again than plastron breathers as a result of the additional resistance of the cuticle to O₂ diffusion. This hypothesis could be tested with subterranean aquatic beetles, as several pieces of evidence indicate that they use cutaneous respiration. Observations of two subterranean aquatic beetles indicate they lack plastrons, have little or no gas underneath the elytra to act as an air store and do not use gas gills (Ueno, 1957). However, subterranean aquatic beetles do have rich tracheation of the elytra and tolerate long periods of submergence, and are likely, as in some other subterranean species, to have low metabolic rates (Ordish, 1976; Smrž, 1981; Ueno, 1957). Additionally, the beetles may become trapped in interstitial spaces without air–water interfaces and would be unable to replenish their air stores (Ueno, 1957). Other than a small number of observations of these beetles' behaviour and habits (Ordish, 1976; Ueno, 1957), and quantification of the level of tracheation in the elytra of six subterranean diving beetles (Dytiscidae) (Smrž, 1981), no studies have thoroughly investigated respiration in subterranean aquatic beetles.

In Western Australia, there are more than 100 species of subterranean diving beetles described from approximately 50 isolated calcrete aquifers, representing the largest assemblage of subterranean diving beetles in the world (Balke et al., 2004; Watts and Humphreys, 2009). Each aquifer contains species of different size classes, but all species are less than 5 mm long (Balke et al., 2004; Watts and Humphreys, 2009). Most species (∼75%) within each aquifer have evolved independently from separate surface-dwelling ancestors, with aridification of the Australian continent, after the Pliocene, likely to be the driver of their isolation in aquifers and evolution underground (Cooper et al., 2002; Guzik et al., 2009; Leys et al., 2003). However, there is also evidence that some species diversified underground from a subterranean ancestor (Cooper et al., 2002; Leijs et al., 2012; Leys and Watts, 2008; Leys et al., 2003). The aquifers are completely enclosed and it is likely that much of each aquifer does not have an air interface. The small size, lack of an air interface and apparent lack of setal tracheal gills found in some small surface dytiscids (Kehl, 2014) makes these diving beetles good candidates for cutaneous respiration.

In this study, we tested whether Australian subterranean dytiscids do use cutaneous respiration and investigated the hypothesis that cutaneous respiration limits them to a small size (<5 mm). We examined respiration in three subterranean dytiscids: Paroster macrosturtensis (Watts and Humphreys, 2006) and Paroster mesosturtensis (Watts and Humphreys, 2006), which are sympatric sister species from the Sturt Meadows calcrete aquifer; and an independently evolved species Limbodessus palmulaoides (Watts and Humphreys, 2006) from the Laverton Downs aquifer in Western Australia (Guzik et al., 2009, 2011; Watts and Humphreys, 2006). We tested whether their mode of respiration is similar to that of surface-dwelling species that use air stores and a small compressible gas gill (Calosi et al., 2007) and whether they use cutaneous respiration. We carried out closed-system respirometry to derive O₂-uptake measurements in all three species. In P. macrosturtensis, we determined the relationship between aquatic P_O2 and O₂ uptake. Fibre-optic O₂-sensing probes (optodes) were then used to measure O₂ boundary layers surrounding all three subterranean species, and scanning electron microscopy (SEM) showed they lacked structures such as pores or setae that could have respiratory significance. These results indicated that the beetles use cutaneous respiration. Cuticle thickness was measured in the subterranean species and six surface dytiscids to determine the relationship between cuticle thickness and size. The experimental data and measurements were then integrated into a respiration model based on Fick's general diffusion equation to assess conclusions made from experimental data and understand the limitations of cutaneous respiration on beetle size.

Paroster macrosturtensis and P. mesosturtensis were collected during October 2015 and September 2016 from Sturt Meadows Station, and L. palmulaoides was collected during September 2016 from the Laverton Downs calcrete, WA. Beetles were captured with a 75 mm diameter plankton net attached to a fishing rod placed down boreholes into the calcrete aquifers (borehole depth 4.03–9.31 m, water depth 0.50–5.71 m at Sturt Meadows). After capture, beetles were placed in aquifer water in 200 ml plastic containers with some sand, small rocks and plant roots that were retrieved in the plankton net. The containers were maintained between 20 and 28°C and transported back to Adelaide by road before being placed in a constant temperature (CT) room at 24.5–25.5°C in darkness. Beetles were placed into plastic aquaria (22 cm L×13 cm W×14 cm H) with reverse osmosis (RO) water mixed with rock salt (Australian Sea Salt Rocks, Hoyts Food Industries P/L, Moorabbin, VIC, Australia) to a salinity of 19–22 ppt, and a ∼5 mm layer of calcium carbonate sand and some calcrete stones were placed on the bottom of the aquaria. Water was lightly aerated with air pumped at 1–2 bubbles per second from a tube (4 mm i.d.). Beetles were fed 2–3 times a week with either small pieces of freshly killed amphipods (Austrachiltonia australis) or blackworm (Lumbriculus sp.). Beetles were weighed with an analytical balance (Mettler 163, Mettler, Greifensee, Switzerland; precision 0.01 mg) after each experiment or measurement.

Several species of small dytiscids have been identified as having structures such as setae or pores on the surfaces of their bodies that have, or may have, a respiratory function (Kehl and Dettner, 2009; Madsen, 2012; Verberk et al., 2018). To determine whether these structures were present on the subterranean species, we used SEM. Specimens that had been kept in 100% ethanol were air dried, mounted on stubs, sputter coated with platinum, and viewed with a scanning electron microscope (Quanta 450, FEI, Tokyo, Japan).

To understand how the structure of the cuticle relates to O₂ diffusion, we histologically examined thin sections of the cuticle. For sectioning of the ventral and elytral cuticle, the heads of beetles were removed and the abdomens were then placed in electron microscopy fixative (4% paraformaldehyde/1.25% glutaraldehyde in PBS, +4% sucrose, pH 7.2) for at least 24 h. Samples were then washed in buffer (PBS+4% sucrose, 10 min), post-fixed with 2% OsO₄ for 1 h, and passed through an ethanol series (2×70%, 2×90%, 3×100%, 15 min each), then propylene oxide (15 min), a 1:1 mixture of propylene oxide and epoxy resin for ∼3 h, then two changes of 100% resin for ∼5.5 and 16 h, before being placed in fresh resin and polymerized at 70°C for at least 24 h. Sections, 1–2 µm thick, were made through the transverse plane midway along the abdomen using a microtome (Leica EM UC6, Leica Mikrosysteme GmbH, Vienna, Austria), and samples were stained with a 1% Toluidine Blue and 1% borax mixture. Although samples were prepared for transmission electron microscopy, they were viewed under a light microscope. Four cuticle thickness measurements from each sample produced a mean value for that sample. Some beetle specimens had been prepared and stored differently prior to fixation, including freshly killed specimens, those stored in 70% and 100% ethanol, and air-dried specimens. Data were pooled for each species for comparisons among species.

Cuticle samples from freshly killed surface species were also taken and prepared in the same manner as above to produce an allometric relationship of cuticle thickness with mass. Species included were Necterosoma dispar (Germar 1848), Rhantus suturalis (W. S. Macleay 1825), Sternopriscus clavatus Sharp 1882, Platynectes decempunctatus (Fabricius 1775), Onychohydrus scutellaris (Germar 1848) and Hyderodes shuckardi Hope 1838 (Watts and Hamom, 2014). Necterosoma dispar, R. suturalis and S. clavatus were collected near Forreston in the Adelaide Hills, SA, P. decempunctatus from near Strathalbyn, SA, and O. scutellaris and H. shuckardi from near Penola, SA, between February and May 2017. For larger surface species, cuticular samples were taken from the centre of an elytron and abdominal sternite 4, and in smaller species from the centre third of an elytron and sternites 2–4.

The simple structure of the ventral cuticle formed by a number of parallel laminae contrasts with the elytral cuticle. The elytral cuticle in small species consists of a layer of parallel laminae with a layer of soft tissue on the ventral surface. However, in larger species, from P. decempunctatus (28 mg) up, there are haemolymph spaces that are encompassed within the cuticular laminae to varying degrees (Noh et al., 2016). Cuticle thickness was measured on the ventral surface perpendicular to the cuticular laminae from the outside surface through to the soft tissue, and in the elytra from the outside surface including, or excluding, the soft tissue or haemolymph spaces.

If O₂ is taken up from the water through a surface such as the cuticle or a gas gill bubble, O₂ diffusion will be evident by a decrease in P_O2 from the free water through a layer next to that surface, called the boundary layer, and the effective thickness can be measured by the distance from the beginning of the P_O2 decrease to the surface. Boundary layer thickness in the subterranean dytiscids was detected and measured with a fast-response fibre-optic O₂-sensing optode (Sensor model, PSt 1 taper tip, tip diameter <50 µm; meter model TX-3; PreSens, Regensburg, Germany) mounted in a 3D micromanipulator (Seymour et al., 2015). Individuals of the three subterranean species were glued with cyanomethacrylate gel either dorsally or ventrally to a small wire stand that was placed into a clear Perspex container (110 mm L×90 mm W×25 mm H) with 15 mm-deep stagnant water. If placed with the ventral surface upwards, a small amount of glue was used to restrain the legs as movement would disrupt the boundary layer and P_O2 measurements. A dissecting microscope was mounted horizontal and perpendicular to the Perspex container to view the beetle ∼1 cm below the surface. The optode was lowered to the surface of the beetle and progressively raised in 50 µm steps until 200 µm above the surface of the beetle, then in 100 µm steps to 500 µm, and 250 µm steps to 1000 µm. The optode was left for measurements at each position for 30 s, after which a 30 s recording period took place before the optode was moved to the next measurement position. Optodes were calibrated weekly with sodium sulphite and air-equilibrated water at the measurement temperature. In a pilot study, we attempted to detect boundary layers above the cuticle of the large surface dytiscid O. scutellaris. However, no P_O2 differences consistent with the presence of a boundary layer were detected on the elytron, metacoxa, metafemur or sternites. This pilot study provided evidence that optode measurements in the subterranean beetles were unlikely to be caused by changes in sensor readings when approaching the surfaces of the beetles, or by microbial growth.

Two methods of respirometry were used to measure O₂ consumption rate in the beetles. The first involved a closed chamber with water and a small amount of air in which P_O2 was measured with an optode (Sensor model, B2 flat tip; PreSens). Only the O₂ consumption of P. macrosturtensis was measured with this method when it was unknown whether the beetles returned to the surface to collect air. Once it was established that the beetles could remain underwater for long periods, a second method was used involving a closed respirometry chamber, filled with water only, where P_O2 was measured with a Clark-type O₂ electrode. This chamber was subsequently used for P. macrosturtensis, P. mesosturtensis and L. palmulaoides.

The first setup consisted of a glass vial, 41 mm long, 18 mm in diameter with a 7 mm diameter opening. Nylon mesh surrounded the inner sides and bottom of the jar to prevent beetles getting stuck to the glass and allow them to get to the surface if needed. The jar was filled with air-equilibrated RO water with dissolved NaCl at 25°C and 22 ppt salinity. Individual beetles were placed into the chambers in a drop of water, minimising the introduction of bacteria and preventing the beetle from getting stuck on the meniscus. Chambers were then sealed with a chemical stopper, with excess water pushed through a 1 mm hole in the stopper. The chamber was then dried on the outside and weighed with the analytical balance. Afterwards, approximately 0.2 ml of water-saturated air at 25°C was injected into the chamber, pushing excess water out. The chamber was dried and weighed again. A water-filled Pasteur pipette was then inserted into the chamber, through the hole in the stopper. The chamber was then placed into a water bath regulated by a thermocirculator (Thermomix 1419, type 850094, B. Braun, Melsungen, Germany), and left for 4–5 min to allow the temperature to equilibrate. An optode was inserted through the tight-fitting pipette to measure P_O2, effectively sealing the chamber with water. The optode inserted into the bubble, and then water, recorded the initial P_O2 in each medium for 2 min after the P_O2 traces became stable (∼2–4 min). Each chamber was left for approximately 3 h, after which the optode was placed back into the air bubble, and then the water, which was gently mixed producing a uniform distribution of O₂ for final P_O2 measurements (2 min stable trace). The beetle was then removed from the chamber, dried with a paper towel and weighed live. The experimental water was returned to the chamber, which was topped up with fresh RO water and closed. The pipette and optode were inserted into the chamber, which was returned to the water bath. The water-filled chambers were used to assess background respiration and sensor drift and left for approximately 3 h between initial and final P_O2 measurements. If background respiration and drift exceeded ±10% O₂ uptake, data were excluded. Air and water volumes in the chambers were determined from the mass measurements of the chambers with water, and air and water minus beetle wet mass (if present), and the dry mass of the chambers, the chemical stopper seal and mesh.

The second respirometry setup, used while at Sturt Meadows Station, consisted of four closed water-filled high-density polyethylene chambers with a glass front (0.67 ml, model 1271, Diamond General, Ann Arbor, MI, USA), each with a Clark-type O₂-electrode (model 730, Micro-electrodes, Bedford, NH, USA) (Mueller and Seymour, 2011). Temperature was regulated by circulating water at the experimental temperature through a water bath and water jackets surrounding the respiratory chambers. The electrodes were connected to an O₂-analyser (ReadOx-4H, Sable Systems, Las Vegas, NV, USA), a data acquisition unit (PowerLab ML750, ADInstruments Pty Ltd, Castle Hill, NSW, Australia) and a laptop computer, and were calibrated with sodium sulphite every 1–4 days (zero drifted ≤1% per day) and air-equilibrated water daily. In the back of each chamber was a miniature magnetic stirrer that rotated at ∼2 Hz.

For air-equilibrated Ṁ_O2 measurements, chambers were 2/3 filled with water and the stable P_O2 recorded for 5 min, providing an initial measurement for drift calculation. Drift was assumed to be linear (Seymour and Roberts, 1995). Individual P. macrosturtensis or L. palmulaoides, or four P. mesosturtensis because of their small size, were then pipetted into the open chambers, which were closed and filled with the treatment water, ensuring no air bubbles remained. Beetles were left for approximately 2 h with the stirrer on to maintain homogeneous P_O2 before the beetles were removed and the experimental water put aside. Chambers were again 2/3 filled with air-equilibrated water for a second stable 5 min period to calculate drift over the time beetles were in the chambers. The water that contained the beetles was then returned to the chamber for 30 min to measure background respiration, followed by another stable 5 min period with clean water for drift calculation during background respiration measurement.

Water saturation in boreholes where beetles are found can vary widely both between and within boreholes (Fig. S1). Because of this variation, the relationship between O₂ consumption rate and declining P_O2 was measured in P. macrosturtensis to understand how these conditions might affect metabolism. Individual P. macrosturtensis were exposed to four individual P_O2 treatments in the 0.67 ml chambers, nominally ∼20, ∼15, ∼10 and ∼5 kPa. This was done to avoid excessive build-up of CO₂ and bacterial contamination that would have occurred in one long respirometry run. Beetles were placed into the chambers for 1.5 h, then removed and background respiration measured for 30 min. Prior to and after each measurement, 5 min periods of P_O2 were measured as above for drift determination.

The rate of P_O2 decline (or Ṁ_O2 decline in the air+water chambers) of the background measurement was subtracted from the total respiration rate of the chamber with the beetle in it to determine the beetles' Ṁ_O2. Background respiration was 4.1±3.1% of the total (N=7) in the air+water chambers with P. macrosturtensis, and 6.4±4.4% (N=7) in the water-only chambers, and 20.1±7.1% (N=27) for the critical P_O2 measurements. For P. mesosturtensis, background respiration was 29.6±24.5% (N=5), and for L. palmulaoides it was 40.7±20.6% (N=6).

Activity of the beetles, e.g. crawling or moving legs or body, was assessed in beetles within the water-filled respirometry chambers as the cumulative duration of activity of one or more beetles within the chamber over a 10 min period. A stopwatch was used while observing the beetles and activity was reported as the percentage of time active during the 10 min period.

Wet mass of P. macrosturtensis was recorded from air+water chamber experiments conducted in the lab; however, all other experiments took place at the field station at Sturt Meadows, where a balance was unavailable. There, beetles were dried with silica gel beads and weighed with the analytical balance once back in the lab. Wet mass was reconstructed from dry mass using the water content of living beetles, which was found to be 66.5% of wet mass.

The subterranean beetles hold a small amount of gas within the sub-elytral cavity but are slightly negatively buoyant. The O₂ within this gas was not included in calculations of Ṁ_O2 because it represents only 1% of the O₂ in the water-only chambers and <0.1% in the air+water chambers.

Video recording of the beetles in the boreholes was undertaken to quantify the beetles' activity and pattern of occurrence, which may be related to O₂ levels or resource availability within the boreholes. To record the beetle's activity, we mounted a bore camera (USB Endoscope Camera, 20 m, 14.5 mm diameter, with four LEDs), angled at ∼45 deg downward on a weight on the end of a tape measure. The camera was orientated so the view was not obstructed by the side of the borehole, and was placed at the water surface, at 0.5 m depth and at the bottom of each hole (N=11 boreholes surveyed). Water depth at the bottom varied between holes. The three beetle species found at Sturt Meadows – P. macrosturtensis, P. mesosturtensis and P. microsturtensis – were not distinguished in the video recordings as parallax and distortion made it difficult to discern size and shape differences between species. Recordings at each point within the borehole were made for 600 s and observed using a VLC media player (version 2.2.3, USA). The number of occurrences of beetles was counted. An occurrence counted as a beetle entering the camera view. The proportion of time beetles were active in boreholes was determined by measuring how long beetles were in view of the camera and recording when and how often beetles were active.

Wilcoxon matched pairs, ANOVA and Tukey's post hoc tests were conducted with GraphPad Prism 7.02 (GraphPad Software Inc., La Jolla, CA, USA), and ANCOVA were conducted according to Zar, 1998. Linear, one-phase association, power and polynomial regressions were performed with Excel or GraphPad Prism. Statistics reported are means and 95% confidence intervals.

The dorsal surfaces of the head, pronotum and elytra of P. macrosturtensis and P. mesosturtensis were very similar (Fig. 1A,B). Both species had very sparse setae with the dominant structures being hexagonal reticulations on all three surfaces viewed. Limbodessus palmulaoides also had sparse setae, but the pronotum and elytra were smooth with no visible reticulations, and the head showed irregular hexagonal reticulations (Fig. 1C). The ventral surface of P. macrosturtensis was also viewed and was similar to the dorsal surface.

Cuticle structure varied between dorsal and ventral surfaces, as well as between the elytral cuticles of different species. The ventral cuticle consisted of a number of parallel cuticular laminae next to the soft tissue of the abdomen in all species (Fig. 2E,F). However, the elytral cuticle in the small dytiscids (subterranean species, and S. clavatus and N. dispar) consisted of a layer of parallel laminae with a thin layer of soft tissue on the ventral surface of the elytra (Fig. 2D). Tracheae were visible within the soft tissue, which in the sections were small circular structures that correspond with the tracheae seen in the elytra of living beetles. In the larger species (O. scutellaris, H. shuckardi, P. decempunctatus and R. suturalis), haemolymph spaces were seen between a dorsal and ventral layer of cuticular laminae separated by trabeculae, pillar-like structures consisting of the chitinous laminae, except in P. decempunctatus, where they were absent or poorly developed (Fig. 2A–C; Noh et al., 2016; Van de Kamp and Greven, 2010). Observations of beetle elytra in another study suggest the tracheae are within the haemolymph spaces (Iwamoto et al., 2002).

The cuticle thickness of all surface dytiscids was significantly greater than that in the subterranean species (Table 1, Fig. 3). The dorsal and ventral thickness data were pooled to produce an allometric regression of cuticle thickness. Dorsal measurements that excluded the soft tissue and haemolymph spaces were used. These values were the most relevant to O₂ diffusion through the cuticle because of the presence of tracheae within the soft tissue and probably within the haemolymph spaces. The pooled data of both surface and subterranean species showed that cuticle thickness increased with mass to the exponent 0.31±0.05 (Fig. 3).These data were not phylogenetically corrected because of the small number of taxa used in the analysis. The ability to detect phylogenetic signal in a trait is much lower where analyses have fewer than 20 taxa (Blomberg et al., 2003). Additionally, a phylogenetically corrected regression may differ from the observed regression, which would result in diffusion models based on the corrected data being less representative of the actual system.

O₂ boundary layers were found on both the dorsal and ventral surfaces of P. macrosturtensis, P. mesosturtensis and L. palmulaoides (Fig. 4A). To reduce variation associated with small differences in the P_O2 of the water in which the beetles were placed, the difference (ΔP_O2) between the maximum P_O2 value recorded in each P_O2 trace and the P_O2 value at each measurement position was calculated (Fig. 4B). To approximate the effective boundary layer thickness, a one-phase association was applied to the ΔP_O2 data (Fig. 4B) where ΔP_O2=Y₀+(Plateau−Y₀)×(1−exp^(−k×X)), where Y₀ is the y-intercept, and the plateau was constrained to 0 representing the P_O2 of the surrounding water. The effective boundary layer thickness was defined as where the ΔP_O2 (Y) is 5% different from zero where Y₀ equals 100%. The boundary layer thickness (X) was calculated by rearranging the one-phase association, where X=[ln(Y/Y₀)]/−k (excluding the plateau which equals 0; Fig. 4B). For P. macrosturtensis, Y₀=−4.995 and k=0.004363 (dorsal, R²=0.80), and Y₀=−7.576 and k=0.003739 (ventral, R²=0.88), for L. palmulaoides, Y₀=−9.067 and k=0.004175 (dorsal, R²=0.88), and Y₀=−7.222 and k=0.004689 (ventral, R²=0.92), and for P. mesosturtensis Y₀=−5.312 and k=0.006659 (dorsal, R²=0.73), and Y₀=−4.113 and k=0.006561 (ventral, R²=0.81). In P. macrosturtensis, the dorsal and ventral boundary layers were 687 µm and 801 µm, respectively; in L. palmulaoides, the dorsal and ventral boundary layers were 718 µm and 639 µm, respectively; and in P. mesosturtensis, the dorsal and ventral boundary layers were 450 µm and 457 µm, respectively. The P_O2 within the boundary layer of P. macrosturtensis on the ventral surface was significantly lower than that within the layer at the dorsal surface at distances less than 400 µm from the surface of the beetles (two-way ANOVA, P<0.0001, Tukey's post hoc, P≤0.01). In L. palmulaoides, the P_O2 within the ventral boundary layer was significantly higher than that for the dorsal surface from 200 µm to the surface (Tukey's post hoc, P≤0.01). In P. mesosturtensis, there are no differences between the dorsal and ventral boundary layers until at the surface where the ventral P_O2 was significantly higher than the dorsal P_O2 (Tukey's post hoc, P≤0.01).

Mass-specific Ṁ_O2 was not significantly different between all three subterranean species in the water-only chambers (Table 2; ANOVA). However, P. macrosturtensis in the air+water chambers had a significantly higher mass-specific metabolic rate than L. palmulaoides (Table 2; ANOVA, P<0.05, Tukey's post hoc). Whole-animal Ṁ_O2 was significantly different between all species and experimental setups except for L. palmulaoides and P. macrosturtensis in the water-only chambers (Table 2; ANOVA, P<0.0001, Tukey's post hoc). Mean percentage of time that individual beetles were active in the water-only chambers over a 10 min period was 15% in P. macrosturtensis (N=8) and 3% in L. palmulaoides (N=6). Mean percentage of time that one or more beetles were active with P. mesosturtensis within the chambers was 1% (N=2). On occasion, beetles clung to the magnetic flea in the chamber; however, while not moving, these beetles may not have had a resting metabolism.

As P. macrosturtensis was exposed to progressively lower P_O2, Ṁ_O2 declined, as indicated by a linear regression (solid orange line in Fig. 5), and a decline in the upper bounds of Ṁ_O2 measured at a given P_O2 (dashed black line, N=11 beetles, dry mass 1.08±0.13 mg, calculated wet mass 3.23±0.4 mg).

Eight of 11 boreholes surveyed with the borescope contained beetles. Water depth in the boreholes with beetles was 2.7±1.7 m (range 0.50–5.71 m). The mean number of beetle occurrences over 10 min was 4.8±4.4 (N=8) at the surface, 1.7±1.1 (N=6) at 50 cm water depth, and 2.1±2.4 (N=8) at the bottom of the boreholes. Two boreholes had total depths of 0.50 and 0.65 m and these records are included in the bottom category. The number of beetle occurrences at the surface was significantly higher than at the bottom (Wilcoxon matched pairs non-parametric test, P<0.05). The mean percentage of time individual beetles spent active while in the view of the camera, either swimming or crawling in boreholes, was 71±7% (N=50 observations determined from six boreholes), spending an average of 9.9±7.8 s in the view of the camera with 3.08 bouts of activity during this time.

This study shows that small subterranean diving beetles rely entirely on cutaneous respiration, in which O₂ diffuses from the water through the O₂ boundary layer to the beetles' surface and then through the cuticular laminae. On the ventral surface, O₂ diffuses directly into the tissues, but on the dorsal surface of the elytra, it goes into the thin layer of soft tissue or the tracheoles within that tissue, as well as the sub-elytral cavity. In all three species, the sub-elytral cavity is gas filled, so O₂ entering this space could diffuse to the spiracles on the dorsal side of the abdomen and into the tracheal system or diffuse directly through the tergites. SEM showed the tergites are not hardened like the elytra or sternites.

Unlike in surface-dwelling dytiscids, the gas within the sub-elytral cavity does not appear to be exchanged with the air or water directly. These beetles rarely go to the surface and break the meniscus with the tip of the abdomen, and were never observed to have a small gas gill bubble at the tip of the abdomen, like in surface species. Additionally, P. macrosturtensis and P. mesosturtensis have tightly locked elytra (Watts and Humphreys, 2006). Consistent with the hypothesis that the beetles use cutaneous respiration, we observed the beetles could survive at least 12 days submergence without access to air, and they are slightly negatively buoyant. Another subterranean beetle, Phreatodytes relictus (Noteridae), lacks gas under the elytra completely (Ueno, 1957), showing that respiratory involvement of this space is not necessary.

The strongest support for cutaneous respiration in P. macrosturtensis, P. mesosturtensis and L. palmulaoides is the presence of O₂ boundary layers surrounding the beetles, indicating O₂ diffusion into the surface of their bodies (Fig. 4). Boundary layer thickness on the dorsal surface is similar to that on the ventral surface in P. mesosturtensis, 450 and 457 µm, respectively. However, in P. macrosturtensis and L. palmulaoides, the dorsal layers (687 and 718 µm, respectively) differ from the ventral layers (801 and 639 µm, respectively; Fig. 4). The difference in thickness between the dorsal and ventral boundary layers is unclear, but may be linked to differences in the cuticle thickness more widely across the body despite the similarities measured in thin sections. The thinner boundary layer found around P. mesosturtensis is due to this species' smaller size. In experiments, glue holding the beetles to the stand would increase O₂ demand through the other surfaces that are not covered, thus increasing the boundary layer thickness. Under natural conditions, the boundary layers are likely to be thinner, because the beetles are often moving (mean 3.08 bouts of activity over an average of 9.9 s while in view of the borescope), and, at least at Sturt Meadows, there appears to be some slow water movement within the aquifer that would help ventilate the boundary layer.

Differences between the model and experimental results arise when considering O₂ uptake per unit surface area. Ṁ_O2 is 36.5 pmol s⁻¹ cm⁻² in P. macrosturtensis, 24.6 pmol s⁻¹ cm⁻² in P. mesosturtensis and 26.4 pmol s⁻¹ cm⁻² in L. palmulaoides. These values result in a similar boundary layer thickness for P. mesosturtensis and L. palmulaoides; however, the experiments show the boundary layer is thinner in P. mesosturtensis (Fig. 4). The difference may be linked to the smaller size of P. mesosturtensis, in which diffusion is closer to radial than linear as assumed in Fick's diffusion equation for the other species. Nevertheless, despite the similarity in size, L. palmulaoides is likely to be more tolerant of low P_O2 compared with P. macrosturtensis because of a better ratio of surface area to Ṁ_O2 and a thinner cuticle (Table 1, Fig. 6A).

The model can be adjusted to make predictions about O₂ limitation with increasing beetle size. This is achieved by taking the average relationship between length, width and height of the three subterranean diving beetles and increasing the size of a hypothetical beetle (assuming a lozenge shape; Tables S1, S2 and Fig. S2). This model assumes circulated water and cuticle thickness increase with mass according to the equation determined in Fig. 3. Ṁ_O2 scaling began with a resting metabolic rate of 0.2 nmol h⁻¹ at 0.04 mg up to 24.2 nmol h⁻¹ and 4.82 mg (5 mm), which is approximately the size of the largest known subterranean dytiscid. Increments in metabolic rate up to 10-fold represent the potential metabolic scope of the beetles. There is little effect of size on the P_O2 at the surface of the cuticle (Fig. 6Bi). A thicker boundary layer would have a greater effect; however, given the animals would be moving when metabolic rates are high, this would not limit the system. Cuticle thickness has a greater effect on internal P_O2 (Fig. 6Bii). Although the estimated resting metabolism does not become limited at the maximum beetle size, the metabolic scope is reduced. A beetle such as P. mesosturtensis would be able to increase metabolic rate more than 10 times above resting levels; however, P. macrosturtensis and L. palmulaoides have a calculated metabolic scope of only 4–5 times resting, which is similar to the narrower metabolic scopes recorded between rest and terrestrial activity in several beetle species (Bartholomew and Casey, 1977; Rogowitz and Chappell, 2000).

When P. macrosturtensis was exposed to declining P_O2, both metabolic scope and Ṁ_O2 tended to decrease, as shown by the linear decline in Ṁ_O2 as well as a linear decline in the upper Ṁ_O2 values recorded (Fig. 5). Each point in Fig. 5 represents the mean Ṁ_O2 of the beetle while it was in the chamber. However, because the beetles' activity cannot be controlled, they may be more or less active during the experiments, producing a range of mean Ṁ_O2 at a given P_O2. Fig. 7 shows the experimental results can be explained by varying levels of activity during each experiment, where maximum metabolic rate declines linearly with P_O2 and is therefore O₂ limited, but resting metabolic rate remains unaffected. Experimentally measured values below the calculated values may indicate a lower resting metabolic rate than anticipated or that resting metabolic rate is affected at lower P_O2. The model results in Fig. 7 would differ if the proportion of the experiment in which the beetles were active changed or the metabolic scope changed. However, if maximum metabolic rate is O₂ limited and the beetles' activity is variable, the same general trend would occur.

A considerable reduction in metabolic scope could limit particularly larger beetles' ability to successfully occupy and utilise a subterranean habitat. Reduction in metabolic rate has been suggested as an adaptation to stressful environments, enabling a greater allocation of resources to growth, reproduction and development (Chown and Gaston, 1999). If the metabolic scope becomes narrower in larger beetles, this could also limit the resources available for growth, reproduction and development. Additionally, declines in aquatic O₂ levels reduce the beetles' metabolic scope. The largest known subterranean dytiscids, Limbodessus magnificus and Limbodessus hahni at 4.8 mm long, equivalent to 4.3 mg (Fig. 6B; Balke et al., 2004; Watts and Humphreys, 2009), may represent the upper size limit of these beetles which still have a functional metabolic scope. If small subterranean dytiscids are found to have a greater metabolic scope than larger species, this would support the idea that a reduction in metabolic scope associated with cutaneous respiration limits subterranean beetle size.

The metabolic rate of subterranean dytiscids is lower than that of resting insects generally, and of plastron breathing insects (Fig. 8). In plastron breathers, their low metabolic rate is associated with the significant resistance to O₂ diffusion that the boundary layer provides (Seymour and Matthews, 2013). The subterranean beetles have, in addition to the boundary layer, resistance of the cuticle, which corresponds with a further reduced metabolism. However, there are other factors that could contribute to a lowered metabolic rate. Subterranean dytiscids have reduced wings and are unable to fly (Watts and Humphreys, 2006), and insects that undertake low-energy activities have lower resting metabolism than those that do high-energy activities like flying (Reinhold, 1999). Low metabolism has also been associated with low and variable O₂ levels in subterranean aquatic isopods and amphipods (Hervant et al., 1998; Malard and Hervant, 1999), and resource limitation in subterranean environments (Hüppop, 1985). It is unclear whether the beetles in the Sturt Meadows and Laverton aquifers are resource limited, but they are exposed to variable O₂ levels. In the Sturt Meadows aquifer, O₂ saturation ranges from ∼50% to 100% in boreholes known to have beetles (Fig. S1), and O₂ saturation can vary by more than 40% within an individual borehole. This variability within and between boreholes has also been recorded in other aquifers containing dytiscids (Watts and Humphreys, 2006).

The cuticle thickness of the three subterranean species in this study was between 6.3 and 8.4 µm, and was similar between the species despite the size differences (Table 1). These measurements give an indication of the thickness of two major surfaces of the beetles, the elytra and ventral sternites. However, cuticle thickness varies across the body, including overlap of sternites, and the edges of the elytra over the edge of the abdomen, which could influence O₂ diffusion. The similarity in cuticle thickness between the subterranean species may indicate a trade-off between being thin enough to facilitate cutaneous respiration and thick enough to provide sufficient strength and integrity to allow the beetles to function within their environment without physical damage (Lane et al., 2017). Surface dytiscids may require thicker cuticles to resist damage from high-energy environments and to protect them from predators. This hypothesis is supported by surface species S. clavatus and N. dispar having thicker than expected cuticles given their mass (Fig. 3), and potentially why the surface dytiscid Deronectes aubei, which has a relatively thick cuticle given its size (∼50 µm; Kehl and Dettner, 2009), uses setal tracheal gills. The tracheal gills reduce resistance to O₂ diffusion, which would be much greater through the unelaborated cuticle, while still maintaining its strength.

Cuticle thickness of the subterranean beetles is within the range of that of other cutaneously respiring insects. Fifth instar Aphelocheirus aestivalis (20 mg) respire through a 45 µm thick cuticle (Thorpe and Crisp, 1947b), and the diffusion distance of the tracheal gills of lestid damselfly larvae is estimated at 10–20 µm (Eriksen, 1986). In D. aubei, diffusion distance through the respiratory setae is <1 µm, like in the tracheal gills of Trichoptera larvae (Kehl and Dettner, 2009).

As dytiscids become larger, the internal structure of the elytra changes (Fig. 2). In the subterranean and small surface dytiscids, the elytra consists of a layer of cuticular laminae with a layer of soft tissue on the ventral surface containing the tracheae (Fig. 2D). However, in the larger dytiscids, haemolymph spaces occur within the cuticular laminae separated by pillar-like trabeculae (Fig. 2A–C). The spaces may help lighten the elytra, assisting with flight, and the trabeculae provide additional mechanical strength (Ni et al., 2001; Van de Kamp and Greven, 2010). The tracheae, which appear to be within the haemolymph spaces (Iwamoto et al., 2002), would allow diffusion of respiratory gases throughout the elytra. This may also allow O₂ diffusion into the tracheal system through the cuticle from the water, or from the sub-elytral cavity, the latter being more likely because of the ventral cuticular laminae being thinner than the dorsal cuticular laminae. Some beetles have air sacs within the elytra (Chen and Wu, 2013; Gokan, 1966); however, it is unclear whether they are present in the larger dytiscids.

The subterranean beetles in this study differ from surface dytiscids that use air stores and gas gills and return to the surface regularly. However, they also differ from submergence-tolerant surface dytiscids because they lack structures like respiratory setae or pores (Fig. 1). The submergence-tolerant dytiscid D. aubei have ∼20 µm long, spoon-shaped setae on their body, which have tracheoles running to the base and probably into the setae themselves (Kehl and Dettner, 2009). These setae, which are found at a density of 5900 mm⁻² on the elytra, allow O₂ diffusion from the water into the tracheal system through the thin cuticle of the setae (<1 µm) traversing the ∼50 µm thick elytral cuticle (Kehl and Dettner, 2009). In other submergence-tolerant surface dytiscids, punctures, small openings in the cuticle surface from which structures emerge (Wolfe and Zimmerman, 1984), and other small openings have been identified as potentially having respiratory roles. Individual punctures or openings/pores are 7–43 µm² at a density of 3400–14,000 mm⁻². Similar structures of this size and density are absent on the subterranean dytiscid species (Fig. 1).

Observations with the borescope in boreholes show there were more occurrences of beetles near the water's surface. This may be associated with a higher P_O2 near the water's surface as a result of O₂ diffusion from the atmosphere, a general trend shown in boreholes known to have beetles (Fig. S1). However, O₂ levels can be inverted within a borehole, with higher O₂ levels being deeper down (Watts and Humphreys, 2006). Beetles may also be congregating near the water's surface because insects and detritus fall into the water, providing a food source. On two occasions in the laboratory, P. macrosturtensis was observed to temporarily crawl out of the water completely, associated with a strong odour above the water. This behaviour is likely to be secretion grooming, which surface dytiscids undertake to reduce bacterial growth with antimicrobial chemicals and to increase wettability of the cuticle (Dettner, 1985). It may also result in the beetles being closer to the surface. Whether smaller species like P. mesosturtensis undertake this behaviour is unclear as they easily become stuck to the meniscus and are unable to dive. However, maintaining a clean cuticle would be important for ensuring optimal cutaneous respiration by reducing O₂ consumption caused by microbial growth, and encouraging convection of the boundary layer.

Most subterranean dytiscids are from the subfamily Hydroporinae in which members are generally <5 mm long and weigh <5 mg (Miller and Bergsten, 2016). Small size is useful for both evolving cutaneous respiration and occupying intermediate habitats between terrestrial and subterranean environments where beetles inhabit small interstitial spaces. These include water-filled gravel where surface water interfaces with ground water or where surface water dries up and ground water is used as a refuge (Kato et al., 2010; Leys et al., 2010; Leys and Watts, 2008). Other groups of dytiscids are likely not to have made the transition to the subterranean habitats because of difficulties of fitting into small spaces, limited resource availability for larger species and propensity to disperse under adverse conditions (Leys and Watts, 2008). Many aquatic insects and in particular small species are likely to benefit to some degree from cutaneous respiration (Vlasblom, 1970). The ability of these subterranean dytiscids to rely solely on this mode of respiration and at relatively high temperature has enabled this group to adapt to and survive in the aquifer environment. Given the considerable number of independent incursions into the subterranean environment (Cooper et al., 2002; Leys et al., 2003), cutaneous respiration in these dytiscids has contributed to the largest radiation of subterranean diving beetles in the world (Balke et al., 2004).

We thank Flora, Peter and Paul Axford for providing access and accommodation at the Sturt Meadows pastoral property, Chris Watts from the South Australian Museum for assisting with species identification, Lyn Waterhouse and Lisa O'Donovan from Adelaide Microscopy, William Humphreys from the Western Australian Museum for water quality data, Sally Maxwell and Silvia Clarke from DEWNR and Thomas Nelson and Qiaohui Hu from Adelaide University for assisting with collecting beetles and maintaining them, and two anonymous reviewers whose comments and suggestions helped improve this manuscript.