Regulation of gas exchange and haemolymph pH in the cockroach Nauphoeta cinerea

Animals regulate their ventilation to ensure an adequate supply of O₂ to their tissues while eliminating CO₂. While aquatic animals primarily control their ventilation to extract sufficient O₂ from a milieu with a low capacitance for O₂ and a high capacitance for CO₂, terrestrial animals find atmospheric O₂ readily available and generally control their ventilation to regulate CO₂ elimination (Dejours, 1988). This regulation of internal CO₂ partial pressure (P_CO2) has direct consequences for the acid–base balance of the animal's extracellular body fluids, as CO₂ reacts with water to form carbonic acid. However, most terrestrial animals use this reaction to their advantage, hypoventilating or hyperventilating to adjust the P_CO2 of their body fluids, thereby counteracting short-term fluctuations in acid–base balance. Although the trend towards CO₂-sensitive ventilatory control and elevated body fluid P_CO2 is apparent in most air-breathing animals [e.g. molluscs (Barnhart, 1992), crustaceans (Burggren, 1992) and vertebrates (Reeves, 1977)], the control of gas exchange among insects, and more particularly its role in acid–base balance, is poorly understood.

Insects were among the very first air-breathing animals to invade the terrestrial environment, and by almost any metric they are the most successful. They possess a highly effective respiratory system, which consists of a network of air-filled tubes that branch throughout all parts of their body, opening to the atmosphere through muscular valves called spiracles. Insects display a range of respiratory gas exchange patterns by controlling when they open and close their spiracles and when they use muscular contractions to actively ventilate their tracheal system. Active insects tend to take up O₂ and release CO₂ continuously, but resting insects may display alternating periods of high and low gas exchange, or even protracted periods with no gas exchange (Marais et al., 2005). These patterns are known to change in response to changes in metabolic rate or inspired levels of O₂ and CO₂. Early investigations into the effects of O₂ and CO₂ on insect gas exchange patterns showed that insects in air containing more than 2% CO₂ opened their spiracles continuously and increased ventilation frequency, whereas oxygen levels had to be severely reduced to achieve the same effect (Kitchel and Hoskins, 1935; Miller, 1960a; Miller, 1960b; Wigglesworth, 1935). From this it was concluded that CO₂ rather than O₂ was the more important respiratory stimulus (Miller, 1960a). However, whether insects were controlling their gas exchange in order to maintain the acid–base balance of their haemolymph or to maintain a stable P_CO2 irrespective of body-fluid pH remained unknown. It was suggested initially that insects regulate gas exchange in response to changes in body-fluid pH. This view is supported by experiments on fleas, flies and cockroaches that show both ventilation rate and spiracle aperture increase in response to injections of acid solutions into the insect's haemocoel (Case, 1957; Case, 1961; Snyder et al., 1980; Wigglesworth, 1935). Unfortunately, interpreting these results is difficult because acidifying an insect's haemolymph unavoidably shifts its HCO₃^–↔CO₂ equilibrium towards CO₂, thereby confounding any independent effect of low haemolymph pH with increased haemolymph and intratracheal P_CO2. Attempts to separate the effects of pH and CO₂ have produced mixed results. Studies on cockroaches have shown that irrigating their ventral nerve chord with a range of different acids stimulates abdominal pumping in the whole animal and rhythmic action potentials in isolated nerve chord preparations (Case, 1961; Snyder et al., 1980). Low pH solutions were found to produce greater and more consistent increases in ventilation frequency than solutions with high P_CO2 buffered to a uniform pH (Snyder et al., 1980). But haemolymph pH does not appear to influence gas exchange the same way in grasshoppers, with several studies concluding that no relationship between haemolymph pH and ventilation frequency exists independent of P_CO2 (Gulinson and Harrison, 1996; Harrison, 1989; Krolikowski and Harrison, 1996). These inconsistent results may be due to different chemoreceptor responses in the different insect groups (Harrison, 2001), although this possibility has yet to be investigated. Thus, there remains a gap in our understanding of this fundamental aspect of insect physiology.

In recent years, research on insect respiration has become increasingly concerned with a curious pattern of gas exchange called the discontinuous gas exchange cycle (DGC). It consists typically of three sequentially repeating phases (open, closed and flutter phases) characterised by the activity of the insect's spiracles. Depending on the metabolic rate of the insect, the closed and flutter phases can last for many hours, during which time CO₂ release is minimal. The accumulation of respiratory CO₂ within the insect's haemolymph during the closed and flutter phases lowers its pH (Buck and Keister, 1958; Levy and Schneiderman, 1966). Many researchers have hypothesised that the DGC must confer some adaptive benefit, as it appears to directly conflict with both the insect's acid–base balance and respiratory demands by periodically impeding gas exchange, and yet has evolved independently among many different insect orders (Chown et al., 2006; Marais et al., 2005). Fluctuations in haemolymph pH of ∼0.06 units have been measured in butterfly pupae during DGCs (Hetz and Wasserthal, 1993) and fluctuations of ∼0.05 units were recorded from the grasshopper Taeniopoda eques during periodic ventilation (Harrison et al., 1995). Of the five insect orders known to display typical closed/flutter/open DGC patterns, measurements of intratracheal O₂ levels and haemolymph pH exist only for two: lepidopteran pupae and adult grasshoppers (Harrison et al., 1995; Hetz and Wasserthal, 1993; Hetz et al., 1993; Krolikowski and Harrison, 1996). Comprehensive measurements of haemolymph pH, CO₂ release and intratracheal O₂ levels are currently lacking for any species of insect. In order to elucidate the mechanisms underlying insect gas exchange patterns we measured in vivo haemolymph pH, intratracheal O₂, CO₂ release and ventilation frequency in speckled feeder roaches (Nauphoeta cinerea) during continuous and discontinuous ventilation while exposed to normoxia, hypoxia, hyperoxia and hypercapnia. This study also compares the relative magnitude of the insect's response to inspired levels of O₂ and CO₂, as well as how these responses affect haemolymph pH.

Cockroaches [Nauphoeta cinerea (Olivier 1789)] were purchased from The Herp Shop (Ardeer, VIC, Australia) as final instar nymphs and adults. They were kept in a 60 l plastic tub and fed ad libitum on a diet of carrot and dry cat food. Only adult males were used in experiments. Animals were weighed immediately prior to experimentation. Mean cockroach weight was 432.5±25 mg.

Gas mixtures of O₂ and CO₂ in N₂ were produced using three mass flow controllers (Aalborg, Orangeburg, NY, USA) calibrated for N₂ (0–500 ml min^–1), O₂ (0–500 ml min^–1) and CO₂ (0–50 ml min^–1) connected to high-purity N₂ (>99.99% pure) and standard O₂ (>99.5% pure) pressurised gas cylinders. The mass flow controllers were regulated using the voltage outputs of a DT9853 analogue output module (Data Translation Inc., Marlboro, MA, USA) controlled by a desktop PC. All mass flow controllers were calibrated using an NIST traceable 1–10–500 ml bubble flow meter (Bubble-O-Meter, Dublin, OH, USA). The CO₂/O₂ composition of the gas mixtures was accurate to within <1% of the set value. Flow rates were monitored using the flowmeter of an SS3 subsampler (Sable Systems, Las Vegas, NV, USA) located upstream of the respirometry chamber. The flow rate for all experiments was 500 ml min^–1.

Cockroaches had their wings clipped by 1 cm and were then fixed by their pronotum to a 35×35 mm square of card with melted Investo dental modelling wax (Ainsworth Dental Company Pty Ltd, Marrickville, NSW, Australia). The cockroach was placed on its back within a 36×36×25 mm respirometry chamber. A 5 mm² window cut into the card beneath the cockroach's abdomen was placed above a 3.4×4.2 mm infrared photodiode/LED chip (SFH 9202, Osram Opto Semiconductors, Regensburg, Germany) to detect abdominal pumping movements associated with ventilation. The photodiode was connected to a custom-built circuit that produced a voltage change proportional to the abdomen's distance from the photodiode. This allowed abdominal ventilation frequency (f_V, cycles min^–1) to be measured, but only allowed for a qualitative measure of abdominal amplitude. Cockroaches were exposed to stepwise changes in O₂ (40, 30, 21, 15, 10 and 5% O₂) and CO₂ (0.3, 0.6, 1, 2, 3 and 4% CO₂) with each exposure lasting 15 min and a minimum 1 h between exposure to oxygen or carbon dioxide treatments.

Haemolymph pH was measured using implantable fibre-optic pH optodes (140 μm diameter) connected to a PreSens pH-1 micro optical pH meter (PreSens GmbH, Regensburg, Germany). The pH probes were calibrated at 25°C using seven sodium phosphate pH buffers that varied between pH 5 and 8 in 0.5 pH increments. Optical pH probes show slight cross-sensitivity to the ionic strength of the sample, with high ionic strength mimicking lower pH. Therefore, the pH buffers were made up with an ionic strength of 0.35 using NaCl to approximate the ionic strength of cockroach Ringer's solution and ensure correct operation of the probes. All sodium phosphate buffers were checked using a HI9025C pH meter with combined pH electrode (Hanna Instruments, Woonsocket, RI, USA) calibrated at 25°C with standard pH 4 and 7 buffer solutions.

Simultaneous measurement of haemolymph pH, CO₂ release and f_V was performed in a respirometry chamber constructed from a 50 mm i.d. × 12.5 mm deep polystyrene Petri dish (Fig. 1). The lid of the Petri dish was mounted upside down on an x/y rotational stage (XYR1/M, Thorlabs, Newton, NJ, USA) beneath a S216 dissecting microscope (Olympus Australia, Mt Waverley, VIC, Australia). A 2 mm i.d. tubing inlet and outlet were also mounted through the lid, allowing gas mixes to be flushed through the chamber to the CO₂ analyser. A 34×34×6 mm acrylic platform containing the activity detector was affixed within the Petri dish lid. To secure the cockroach within the chamber, a 6 mm hole was drilled through the base of the Petri dish, slightly off-centre. Investo dental modelling wax (Ainsworth Dental Company Pty Ltd) was melted around the hole. Cockroaches were cold-knocked-out by being placed within a –3°C freezer for 5 min. They were then affixed within the Petri dish using the melted dental wax around the hole, such that the left side of their pronotum was sealed over the hole. The Petri dish containing the cockroach was then inverted and pressed into a ring of Blu-Tack® modelling putty (Bostik, Thomastown, VIC, Australia) running around the inside edge of the Petri dish lid, thus sealing the chamber while orienting the cockroach's abdomen directly above the activity detector. To insert the pH optode into the cockroach's haemocoel, a 25 gauge needle was first used to cut a <0.3 mm diameter hole in the pronotum, slightly to the left of the insect's mid-line. As the main tracheae are visible through the insect's cuticle, they were carefully avoided during this procedure. A micromanipulator (World Precision Instruments, Sarasota, FL, USA) was then used to insert the pH optode through this opening ∼1 mm into the cockroach's haemocoel. A small droplet of haemolymph would occasionally well up through the freshly made incision in the pronotum, but once the pH probe was inserted a clot formed rapidly between the optode and pronotum, preventing further bleeding. Haemolymph pH was measured every 10 s. Following implantation of the pH optode, the cockroach was given a minimum of 2 h to recover while the respirometry chamber was flushed with 21% O₂. After the 2 h recovery period, the insect was then exposed to hypoxia (5% O₂ for 10 min) and then returned to 21% O₂ for a further 1 h. Cockroaches were exposed to stepwise changes in O₂ (40, 30, 21, 15, 10 and 5% O₂ with CO₂ kept constant at 0%) and CO₂ (0.3, 0.6, 1, 2, 3 and 4% CO₂ with O₂ kept constant at 21%), with each exposure lasting 10 min and a minimum of 1 h between exposure to O₂ or CO₂ treatments.

The dissecting microscope, x/y rotational stage and micromanipulator were all mounted on a 300×450×12.7 mm M6 threaded aluminium breadboard (MB3045/M, Thorlabs) to ensure that they would not move during measurement. The Petri dish chamber was enclosed within a foam box that was ventilated with air passing through a radiator/computer fan assembly. The radiator was continuously flushed with water from an F12-ED constant temperature water bath (Julabo GmbH, Seelbach, Germany) set at 25°C, maintaining the temperature within the respirometry chamber at 25±0.5°C.

Voltage outputs from the pH-1 micro (pH and temperature), LiCOR-7000 (ΔCO₂ and ΔH₂O), SS3 (respirometry chamber incurrent flow rate) and activity detector were sampled at 10 Hz using a Powerlab 16/30 analogue to digital converter (ADInstruments, Bella Vista, NSW, Australia) and recorded on a desktop PC running LabChart 6.1.3 (ADInstruments).

The procedure for measuring haemolymph pH was followed as described above, with the exception that after the cockroach was cold-knocked-out and waxed to the bottom of the Petri dish, its head was quickly removed using a razor blade. The neck wound was then sealed with melted wax. Within 1 h of the pH probe being implanted, cockroaches would spontaneously display DGCs. Decapitated cockroaches prepared in this manner all produced DGCs and survived for several days. A minimum 2 h recovery period in 21% O₂ was allowed before manipulation of O₂ or CO₂.

Measurements of intratracheal O₂ were made on both intact and decapitated cockroaches. The cockroaches were measured while they were fixed upside down to a square piece of card 35×35 mm using dental modelling wax. They were then placed within a 36×36×25 mm respirometry chamber beneath a dissecting microscope. Intratracheal O₂ was measured using a 140 μm diameter fibre-optic oxygen optode connected to a TX3 O₂ meter (PreSens GmbH) and recorded at 1 Hz. For intratracheal O₂ measurements made on decapitated cockroaches, an unmodified optode with a 140 μm diameter tip was inserted through a small opening in the side of the respirometry chamber directly into one of the two severed major longitudinal tracheal trunks exposed in the neck wound (Fig. 2) using a three-axis micromanipulator (World Precision Instruments). The optode was then sealed into the tracheal trunk with wax. The opening in the side of the respirometry chamber was also sealed with wax, thus making it airtight. For measurements of intact cockroaches, the 140 μm diameter tip of the optode was slightly too large to be inserted through a spiracle directly into the cockroach's tracheal system. To make a tube small enough to cannulate a spiracle, the tapered end of a 10–200 μl polypropylene (PP) pipette tip was held over a soldering iron set at 160°C. Once the tip had melted into a droplet containing an air bubble, the molten plastic was quickly gripped with tweezers and pulled into a thread. This technique produced a continuous lumen within a tube 90–120 μm in diameter. A razor blade was used to trim a ∼4 mm length of this PP micro-tube, with one end cut at 45 deg. This 4 mm PP micro-tube was then fixed within a ∼2 mm length of 0.8 mm i.d., 1.6 mm o.d. polytetrafluoroethylene (PTFE) tube using cyanoacrylate adhesive (Supa glue, Selleys Pty Ltd, Padstow, NSW, Australia). The optode was mounted within a 15 mm length of PTFE tubing (1.6 mm i.d., 3.2 mm o.d., Cole-Parmer, Vernon Hills, IL, USA), such that the O₂-sensitive tip was inside the tubing ∼1 mm from the end, while the tube behind the optode tip was filled with Araldite epoxy resin (Selleys Pty Ltd). The smaller PTFE tube containing the PP micro-tube was then fitted inside the larger PTFE tube containing the optode (Fig. 3). Thus, the PP and PTFE tube served as a replaceable tip that could be inserted through a cockroach spiracle, connecting the air space around the optode with the cockroach's tracheal system. Once the PP tube had been inserted in through the second abdominal spiracle (Fig. 2), high vacuum silicone grease (Dow Corning, Midland, MI, USA) was applied around the spiracle, creating an airtight seal around the tube within the spiracle, and between the two PTFE tubes. As the relatively low conductance of the PP micro-tube decreased the speed with which the optode responded to changes in O₂, the system's lag was characterised by placing the PP micro-tube and optode in a stream of pure N₂. The rate of O₂ decline was then recorded, allowing the time constant of the probe to be determined and an instantaneous correction applied to the data (Seymour et al., 1998). The time constant (τ; time taken to reach 63% of the final asymptotic value) of the bare optode was 5 s and the time constant for the PTFE tube/optode assemblies was 42±5 s. The instantaneous corrections applied to the optode data using a z-transform produced a mean τ of 13±2 s.

Rates of CO₂ production (ml h^–1 g^–1) and f_V (cycles min^–1) were calculated with LabChart 6.1.3. Statistical analyses were performed using JMP® 8.0.2.2 (SAS Institute Inc., Cary, NC, USA). The effect of CO₂ and O₂ on f_V and pH was analysed using mixed-model ANOVA with individual cockroach identity included as a random effect, followed by Tukey's honestly significant difference (HSD) test. The inclusion of individual cockroach identity accounts for the non-independence of multiple measurements obtained from single individuals. This approach was preferred to repeated-measures ANOVA because in the present study data were not available for all individuals at all treatment levels.

Cockroaches were exposed to hypoxic, hyperoxic and hypercapnic atmospheres (N=10). Insects at rest in 21% O₂ and 0% CO₂ ventilated continuously with regularly spaced abdominal movements (Fig. 4) with a mean frequency of 5.8±1 cycles min^–1 (N=10). Exposure to hypercapnia produced two distinct levels of ventilatory response depending on the level of CO₂. Abdominal ventilation frequency showed a small linear increase as the inspired level of CO₂ increased from 0 to 1%. This was observed in nine of the 10 cockroaches measured but was non-significant (f_V=5.0719×%CO₂+8.7343, R²=0.83; expressed as percent change relative to f_V in 0% CO₂: Δf_V=0.5985×%CO₂–0.0232, R²=0.98). Exposure to atmospheres containing more than 1% CO₂ was associated with a significant linear increase in ventilation frequency (f_V=11.849×%CO₂+3.252, R²=0.99; Δf_V=2.1989×%CO₂–1.5088, R²=0.997; Fig. 5).

Ventilation frequency showed a slight decrease with increasing O₂ levels in gas mixes containing between 10 and 40% O₂, but this was non-significant (f_V=–0.1188×%O₂+8.5955, R²=0.91; Δf_V=–0.0209×%O₂+0.5767, R²=0.74). Ventilation frequency in 5% O₂ was significantly higher than in all other O₂ treatments (Fig. 6) while CO₂ release was effectively continuous (Fig. 7). All cockroaches displayed a protracted apnoeic period upon return to normoxia following hyperventilation in 5% O₂ (Fig. 7).

The cockroaches' haemolymph pH was 7.30±0.03 measured in normoxia (21% O₂ and 0% CO₂; N=10). Exposure to 0–1% CO₂ did not significantly alter haemolymph pH, whereas pH declined significantly in atmospheres containing higher than 2% CO₂ (Tukey's HSD, P=0.05; Fig. 5). Haemolymph pH showed a significant negative correlation with ambient %O₂ (P=0.05; Fig. 6). Hypoxia (5% O₂) caused haemolymph pH to increase by 0.34 units, whereas hyperoxia (40% O₂) resulted in a pH decrease of 0.12 units.

DGCs were displayed by three intact cockroaches in normoxia. In the 25 cycles measured, haemolymph pH declined steadily during the closed/flutter periods of the gas exchange cycle, before rising during the open phase as a burst of CO₂ was released. The mean change in pH was 0.07±<0.01 unit.

Implantation of the optode through an abdominal spiracle was achieved successfully with eight cockroaches. Measurements where silicone grease or haemolymph blocked the PP micro-tube connecting the optode to the tracheal system were easily distinguished by the sudden onset of a flat O₂ trace, and were discarded. Intratracheal O₂ was measured under conditions of 21% (normoxia), 15% and 30% O₂. In normoxia mean intratracheal O₂ was 18.2±0.1% (N=8). The O₂ level fluctuated between ambient and ∼15% O₂ following restricted gas exchange or an increase in activity (Fig. 8). Mean intratracheal O₂ was 25.8±0.2% (N=8) in an atmosphere containing 30% O₂. The level of O₂ in the tracheal system appeared to undergo larger fluctuations in hyperoxia, with restricted gas exchange occurring for longer periods. However, during periods of CO₂ release, intratracheal O₂ still reached ambient levels (Fig. 8). Cockroaches in 15% O₂ had a mean intrtracheal O₂ level of 12.8±0.1% (N=3).

All cockroaches (N=3) began breathing discontinuously within 1 h of decapitation. In normoxia and hyperoxia the discontinuous gas exchange cycles lacked a flutter phase, alternating between periods of no gas exchange and periods of active abdominal ventilation. The characteristics of the ventilation periods were remarkably consistent between all decapitated preparations. Gas exchange periods commenced with a high rate of ventilation (15±0.2 cycles min^–1), which gradually declined asymptotically to a lower rate (9±0.2 cycles min^–1), at which point gas exchange and ventilation would cease (Fig. 9). The ventilation periods tended to increase in duration as the preparation aged, increasing from ∼5 min initially to 9 min over 24 h. Exposure to hypoxia or hyperoxia did not change the frequency of abdominal pumping during the periods of gas exchange, except in 5% O₂ where ventilation occurred continuously and more than doubled in frequency. However, hypoxia reduced the duration of the apnoeic periods between bouts of ventilation, as well as causing the appearance of a flutter phase. The flutter phase was characterised by several brief bursts of CO₂ preceding the ventilatory period. During the flutter period, the cockroach's abdomen would pump at a lower frequency than during the ventilatory phase (Fig. 10), but the activity trace shows a smaller voltage change during these pulsations, possibly indicating that the stroke volume of the abdomen was lower. When the cockroaches were returned to normoxia following continuous ventilation in 5% O₂, there was no extended apnoea. Instead, the periods of gas exchange continued as regularly as before, but with the volume of CO₂ released during each burst greatly reduced for the first two gas exchange periods.

Exposure to hypercapnia changed the characteristics of the ventilation periods, so that instead of a gradual decline in f_V over the course of a burst, ventilation would change in a U shape, beginning with 15 cycles min^–1, dropping to ∼13 cycles min^–1 and then rising back to 15 cycles min^–1 at the end of the burst. Hypercapnia in excess of 4% CO₂ caused abdominal pumping movements to become continuous, but because of the high level of CO₂ it was impossible to determine from the respirometry trace whether gas exchange also became continuous. Upon return to normocapnia following hypercapnic exposure, the cockroach performed a protracted ventilatory period while pH returned to its normoxic, normocapnic level.

A total of 82 DGCs were measured in normoxia. Alternating periods of no gas exchange and active ventilation caused haemolymph pH to fluctuate by 0.11±0.004 units (Figs 9 and 10). Exposure to 30% O₂ had no effect on either cycle duration or haemolymph pH; the duration of the closed phase increased in 40% O₂ (Fig. 10). Exposure to hypoxia caused pH to increase, coincident with the appearance of a flutter phase comprising low-frequency abdominal pumping movements and discrete bursts of CO₂ before the open phase (Fig. 10). Exposure to hypercapnia was associated with a drop in pH.

Decapitated cockroaches all displayed DGCs. During the periods of no gas exchange, intratracheal O₂ levels fell continuously, but rose rapidly to a plateau level during periods of ventilation (Fig. 11). Mean intratracheal O₂ in normoxia was 16.8%.

Exposure to hypercapnic, hyperoxic and hypoxic atmospheres demonstrate that respiratory gas exchange in the cockroach is stimulated by O₂ and CO₂ in a manner comparable to most other air-breathing animals: both hypoxia and hypercapnia stimulate ventilation, but small increases in ambient CO₂ levels effect far greater increases in ventilation frequency than equivalent decreases in O₂ (Fig. 12). Thus, while O₂ stimulates ventilation only under conditions of moderate to extreme hypoxia (i.e. <10% O₂), low levels of hypercapnia (i.e. >1% CO₂) dramatically increase ventilation frequency (Figs 5 and 6). It can be seen that ventilation is regulated according to the accumulation of intratracheal CO₂, as the mean O₂ level within the tracheal system of an active cockroach remains close to 18% in normoxic air and intratracheal O₂ levels do not reach the hypoxic trigger point. These observations lead to the general conclusion that, under normal atmospheric conditions of 21% O₂ and 0.03% CO₂, the regulation of gas exchange depends primarily on the level of CO₂ within the insect.

The role of CO₂ as the primary stimulus driving ventilation is also clearly apparent during exposure to hypoxia and hyperoxia. The cockroaches' ventilation proved to be insensitive to hyperoxia, with ambient O₂ levels of 30–40% causing no significant changes in abdominal pumping frequency compared with normoxia (Fig. 6). However, the significant reduction in haemolymph pH in atmospheres containing 40% O₂ requires comment. A previous study on moth pupae provides evidence that the CO₂ sensitivity of their spiracles is depressed by high intratracheal levels of O₂, elevating the CO₂ threshold that causes them to open (Schneiderman, 1960). This behaviour would explain the observed pH decrease, as a transient depression of ventilation frequency at the onset of hyperoxic exposure causes intratracheal CO₂ to increase until it once again stimulates spiracular opening. From then on, the cockroach resumes ventilation at the same frequency as before, but with higher intratracheal levels of O₂ and CO₂, as well as a lower haemolymph pH. The resulting surfeit of intratracheal O₂ is apparently of little consequence to these animals; a finding that is at odds with recent assertions that insects regulate their gas exchange specifically to maintain low intratracheal O₂ levels (Bradley, 2007). However, the results presented here show, perhaps not surprisingly, that O₂ influences gas exchange more through its absence than by its presence. As such, cockroaches respond to hypoxia by hyperventilating, increasing their ventilation frequency fivefold in atmospheres containing only 5% O₂ (Figs 6 and 7). While this behaviour is necessary to maintain adequate O₂ uptake in hypoxia, it has the unintended side effect of increasing the rate of CO₂ clearance above its rate of production, resulting in internal hypocapnia. The end result is that, upon their return to normoxia following hypoxic hyperventilation, cockroaches display an extended apnoeic period to correct their internal CO₂ deficit (Fig. 7). Although hypoxia alone increases ventilatory drive, compared with the effect of hypercapnia, it fails to stimulate ventilation to the same degree. For example, in 5% O₂, ventilation frequency increased fivefold above resting, whereas 4% CO₂ caused ventilation frequency to increase sevenfold. It is possible that the stimulatory effect of hypoxia is counteracted in part by the decrease in internal CO₂ due to hyperventilation. To test this, cockroaches (N=4) were exposed simultaneously to hypoxia and hypercapnia (5% O₂ and 3% CO₂). Under these conditions, mean ventilation frequency was 54±2 cycles min^–1, compared with 32±15 cycles min^–1 in 5% O₂ or 40±12 cycles min^–1 in 3% CO₂. This experiment indicates that the effects of hypoxia and hypercapnia are synergistic, as together they drive greater increases in ventilation frequency than either could alone.

Exposing cockroaches to selected levels of O₂ and CO₂ demonstrates four critical points necessary for understanding insect gas exchange: (1) small increases in ambient CO₂ levels elicit greater increases in ventilation than equivalent decreases in O₂; (2) hypoxia stimulates ventilation whereas hyperoxia has no significant effect; (3) hypoxia stimulates ventilation independently, regardless of CO₂ levels; and (4) apnoeic periods occur when internal CO₂ levels fall below a threshold value and when internal O₂ levels are not so low as to stimulate ventilation. It is interesting to note that the conclusions presented here on cockroach ventilation are essentially the same as those of Haldane and Priestly in their classic paper examining lung ventilation in humans (Haldane and Priestley, 1905). Although the respiratory systems of mammals and cockroaches are separated by hundreds of millions of years of independent evolution, natural selection under the same physiochemical constraints associated with breathing air has caused them to converge on essentially identical mechanisms to regulate their gas exchange: ventilation is stimulated by internal CO₂ levels except when hypoxia demands hyperventilation.

The implantation of micro pH optodes within the cockroaches' haemolymph reveals that changes in ventilation frequency caused by hypoxia and hypercapnia have significant effects on haemolymph pH. In normoxia, cockroaches maintained a haemolymph pH of 7.30±0.03 (Figs 5 and 6). Breathing atmospheres containing up to 1% CO₂ did not cause this pH to change. Although there was a non-significant trend towards higher ventilation frequencies with increasing CO₂ (Fig. 5), an increase in ventilation frequency, amplitude or some combination of the two is the only plausible mechanism to explain a stable pH in the face of increasing hypercapnia. By increasing the convective ventilation of their tracheal system, cockroaches would maintain a constant in the face of a decreased P_CO2 gradient between their tissues and tracheal system, and thus maintain a stable haemolymph pH. However, in hypercapnic atmospheres containing more than 1% CO₂, ventilation rate increased significantly, while haemolymph pH dropped (Fig. 5). These changes in haemolymph pH and ventilation frequency in hypercapnia reveal that cockroaches maintain an intratracheal CO₂ level between ∼1 and 2%. The significant decrease in haemolymph pH associated with a rapid increase in ventilation when breathing gas mixes containing 2% or more CO₂ demonstrates that this level of CO₂ exceeds that normally maintained within the cockroaches' tracheal system, such that no increase in ventilation can reduce their internal CO₂ back to their desired level. Further confirmation that intratracheal CO₂ levels are close to 2% in normoxia can be found in the experiment where cockroaches were forced to hyperventilate in hypoxic atmospheres (5% O₂) containing up to 4% CO₂ (Fig. 13). The loss of internal CO₂ due to excessive ventilation in hypoxia caused the cockroach's haemolymph pH to become alkaline relative to its usual normoxic level. However, this pH disturbance was fully reversed in an atmosphere containing 2% CO₂, suggesting that this is the usual intratracheal CO₂ level. An intratracheal CO₂ level regulated at ∼2% is in very close agreement with ∼1.5% measured for the grasshopper Schistocirca americana (Gulinson and Harrison, 1996). Similarly, the spiracles of the atlas moth Attacus atlas remain open when their tracheal system is perfused with 1.5% CO₂ (Förster and Hetz, 2009), and the spiracles of the flea Xenopsylla cheopis remain open in 2% CO₂ (Wigglesworth, 1935).

The internal regulation of a relatively constant internal CO₂ level is also apparent from experiments where cockroaches hyperventilated when exposed to severe hypoxia (5% O₂). Hyperventilation caused a rapid initial increase in , which then declined asymptotically to a constant level, while haemolymph pH increased by 0.39±0.04 units (Fig. 6). Assuming that the constant approximates the cockroach's metabolic rate, then the increase in over and above this level must be due to the liberation of CO₂ stored in the cockroach's haemolymph. This assumption is supported by the rapid rise in haemolymph pH. Therefore, by integrating the area below the burst and using the stable as the baseline, it was estimated that the cockroaches' lost 0.071±0.01 ml of CO₂ in total from their haemolymph, or 0.146±0.02 ml g^–1 body mass (Fig. 7). When normoxia was restored, cockroaches stopped ventilating almost immediately and closed their spiracles, preventing gas exchange. During this apnoeic period, haemolymph pH dropped steadily as internal levels of CO₂ increased. Once haemolymph pH approached or exceeded its pre-hypoxia level, ventilation and gas exchange would resume as before (Fig. 7). The respiratory alkalosis induced by hypoxic hyperventilation, and the subsequent apnoea, show that cockroaches maintain their haemolymph in a state of compensated respiratory acidosis. It also demonstrates that a constant pH and/or P_CO2 is defended by the regulation of gas exchange.

Insects display a range of respiratory patterns by varying when they open their spiracles and when they generate convective airflow through their tracheal system. In this study, N. cinerea displayed two distinct patterns of gas exchange: a continuous pattern consisting of a train of small CO₂ ‘breaths’ spaced at regular intervals (Fig. 4) and a DGC consisting of extended breath-hold periods punctuated by periods of active ventilation (Fig. 9). Cockroaches ordinarily display DGCs when they are quiescent, but in this study decapitation was necessary to elicit this respiratory pattern consistently. The DGC has long been considered so unusual a gas exchange pattern that its occurrence demands special explanation. It was initially proposed that the long periods of spiracular closure typical of the DGC may have evolved to reduce respiratory water loss [the hygric hypothesis (Buck and Keister, 1955)]. More recently, other hypotheses have been put forward suggesting that DGCs may facilitate gas exchange in hypoxic or hypercapnic burrows [the chthonic hypothesis (Lighton and Berrigan, 1995)], or reduce intratracheal O₂ levels to protect against oxidative damage [the oxidative damage hypothesis (Bradley, 2000)]. While the hygric hypothesis is based on the assumption that respiratory water loss is lower during DGCs compared with other patterns of gas exchange, both the chthonic and oxidative damage hypotheses are based on assumptions regarding how insects regulate their gas exchange patterns in response to varying levels of atmospheric and intratracheal O₂ and CO₂. Therefore, the results of the present investigation can be used to test the validity of these assumptions.

The chthonic hypothesis is based on the assumption that an insect can facilitate gas exchange when exposed to ambient hypoxia or hypercapnia by employing the closed phase of the DGC to increase intratracheal hypoxia and hypercapnia above ambient levels, thereby maximising the partial pressure gradients driving O₂ and CO₂ between the insect and its environment when the spiracles eventually open (Lighton and Berrigan, 1995). However, it is apparent from experimental manipulation of ambient O₂ and CO₂ that a subterranean or chthonic origin of DGCs is unlikely. Rather than favouring the emergence of DGCs, both hypoxic and/or hypercapnic conditions stimulate hyperventilation, directly opposing the emergence of cyclic or periodic gas exchange patterns. Similar findings have been reported for beetles that switch from a discontinuous to continuous pattern of gas exchange when exposed to hypoxia (Chappell and Rogowitz, 2000; Chown and Holter, 2000), while the grasshopper Taeniopoda eques abandons DGCs in atmospheres containing more than 2% CO₂ (Harrison et al., 1995). Ambient hypoxia and hypercapnia have also been shown to prevent the occurrence of DGCs in Samia cynthia moth pupae (Terblanche et al., 2008). All of these examples demonstrate that insects possess essentially the same respiratory sensitivities to low O₂ (below approximately 10% O₂) and high CO₂ (above 2% CO₂) during DGCs as during continuous gas exchange. This illustrates the fundamental problem with the chthonic hypothesis, specifically that using periodic apnoea to enhance gas exchange when exposed to deleterious atmospheres requires the insect to suffer even greater intratracheal hypoxia and hypercapnia than if it simply abandoned the DGC altogether. A corollary of this issue is that for DGCs to have evolved in a subterranean atmosphere, insects must also possess a respiratory system capable of distinguishing whether hypoxia or hypercapnia is of ambient or intratracheal origin and then respond in one of two contradictory ways: ambient hypercapnia or hypoxia would need to inhibit gas exchange to produce the DGC, whereas high intratracheal CO₂ and low intratracheal O₂ produced by routine aerobic activities must promote gas exchange. This scenario is paradoxical. Following the above lines of reasoning, and the consistent responses of other insects exposed to hypoxia and hypercapnia, we conclude that the chthonic hypothesis cannot explain the origin or maintenance of DGCs in insects.

The oxidative damage hypothesis was proposed following the observation that moth pupae performing DGCs maintained a low, stable intratracheal O₂ level of ∼4% for the duration of the flutter phase, irrespective of whether they were exposed to atmospheres containing 6 or 50% oxygen (Hetz and Bradley, 2005; Levy and Schneiderman, 1966). Because the DGC restricts O₂ entry to the tracheal system during both the closed and flutter phases, it is possible that this gas exchange pattern is a mechanism to limit the exposure of the insect's tissues to high levels of O₂, thereby protecting them from excessive oxidative damage. However, diapausing pupae are a highly modified insect life stage, and the mean intratracheal levels of O₂ and CO₂ regulated by a pupa may not be typical of most other active insect life stages. Until now, lack of data has prevented any comparison. Here we show that the mean intratracheal O₂ level is 18.2±0.1% measured in intact cockroaches, and 16.8% in the longitudinal tracheal trunks in decapitated preparations displaying DGCs. Even during the closed phase of a DGC, intratracheal O₂ never fell lower than 10% and rarely dropped below 13%. Cockroaches do not match the pattern observed in moth pupae: low and stable levels of intratracheal O₂, either during DGCs or when active, do not occur (Figs 8 and 9). Likewise, the duration of DGCs displayed by western lubber grasshoppers (Taenepodia eques) and American cockroaches (Periplaneta americana) were unaffected by hyperoxic atmospheres containing up to 40–60% O₂, indicating that, under these conditions, intratracheal O₂ levels never decreased to the hypoxic trigger point during the closed phase of the cycle (Harrison et al., 1995; Woodman et al., 2008). In these insects, too, stable and low intratracheal O₂ levels apparently do not occur. Measured levels of O₂ in the locusts' tracheal system [18.6±0.6% in a normoxic atmosphere (Gulinson and Harrison, 1996)] are also the same as those recorded from cockroaches in the present study. Given the exceptionally close agreement between these two intratracheal O₂ measurements made on insects belonging to different orders, it would appear that insects do not seek to avoid near-ambient levels of O₂ within their tracheal systems, contrary to the assumption that gas exchange patterns function to limit O₂ uptake (Bradley, 2007). However, it should be noted that O₂ levels within the insects' tissues are likely to be lower and more stable than those recorded within the tracheal system.

The current adaptive hypotheses cannot explain why DGCs have evolved among such a diverse range of insects. Therefore, explanations for the origin and maintenance of this pattern of gas exchange must be sought elsewhere. A non-teleological approach can be found in the mechanistic hypotheses that explain DGCs as an emergent property of the insect's respiratory system. These theories are based on the assumption that the DGC may not be adapted for any particular purpose; rather, it is a pattern that arises spontaneously due to interactions between the regulatory mechanisms controlling gas exchange, but only under particular circumstances (Chown et al., 2006; Chown and Holter, 2000). The one condition common to all insects displaying DGCs is that they are inactive, either as quiescent adult insects or diapausing pupae. This then begs the question: what happens when an insect becomes inactive that causes them to breathe periodically rather than continuously? One obvious change is that the insect's metabolic rate decreases at rest. However, if the same respiratory control mechanism responsible for producing individual breaths during continuous ventilation continued to operate when an insect became quiescent and its respiratory demand decreased, then individual breaths would still occur, but less frequently, as gas exchange would continue to be triggered by the ∼2% CO₂ threshold. Instead, quiescent insects display DGCs that are characterised by large fluctuations in intratracheal CO₂, intratracheal O₂ and haemolymph pH due to rhythmically occurring apnoeas punctuated by bouts of vigorous gas exchange. This difference in stability is indicative of changes in the underlying respiratory control mechanism driving gas exchange. In the case of the decapitated cockroaches, this change in regulatory control can be linked directly to the absence of a head and, consequently, the absence of the cephalic ganglion.

The spontaneous occurrence of DGCs in decapitated cockroaches, and their response to hypoxia and hypercapnia, reveals fundamental properties of this pattern. Firstly, it is apparent that decerebration unmasks a periodic gas exchange pattern that is hardwired into the cockroaches' ventilatory neural networks. Secondly, the occurrence of this pattern in a decapitated preparation indicates that pattern generators within the cockroaches' thoracic or abdominal ganglia generate this ventilatory periodicity. A central pacemaker located in the metathoracic ganglion has previously been identified from electrophysiological studies on the cockroach P. americana (Farley et al., 1967). This suggests that a tonic input from the cockroach's brain to the thoracic ganglia is necessary to produce a continuous pattern of breathing, whereas the absence or reduction of this input leads to discontinuous gas exchange. This situation is analogous to that found in vertebrates, where in vitro brainstem–spinal cord preparations spontaneously display periodic ventilatory activity, but only in the absence of a functional pons or midbrain (Fong et al., 2009). The nature of the input from the cephalic ganglion to the thoracic ganglia can be hypothesised from a third observation: the decapitated preparation does not respond to hypocapnia or hypercapnia, but does respond to hypoxia. This suggests that the CO₂ and/or pH chemoreceptors that normally modulate the ventilatory pattern are present in the cockroaches' head (see Miller, 1960a), whereas O₂ chemoreceptors are located in the thorax or abdomen. Thus, in hypoxia, the open and closed phases of the DGC continue to occur at regular intervals in the absence of an intratracheal hypercapnic trigger (as indicated by increasingly alkaline haemolymph; Fig. 10). The flutter phase is then superimposed on this basic pattern, emerging only if intratracheal O₂ falls below the hypoxic threshold before the spontaneous initiation of the next open phase (Fig. 10). This shows that the flutter phase is not an intrinsic component of the DGC, but an independent phenomenon that serves to correct intratracheal hypoxia.

Although there are clear differences between the DGCs displayed by intact and decapitated insects (i.e. sensitivity to CO₂), the decerebrated preparation is informative insofar as it shows that a periodic breathing pattern can occur spontaneously without requiring the presence of other mechanisms, such as chemosensory hysteresis. This finding is important for explaining the origin of the periodicity that characterises the DGC. It has long been known that the open and flutter phases of the DGC are triggered in response to changing intratracheal levels of O₂ and CO₂: the flutter phase is a graded response to intratracheal hypoxia whereas the open phase is a threshold all-or-nothing response induced by hypercapnia (Levy and Schneiderman, 1958). But precisely what determines the duration of the open phase, and what causes it to terminate, is a longstanding question that is fundamental for understanding the DGC, but which has thus far gone unanswered. Previously, models describing the behaviour of DGCs have been forced to assume a large temporal lag between changing intratracheal P_CO2 and the chemoreceptors stimulating gas exchange, thus driving the system to intratracheal hypocapnia during the open phase which is then followed by a corrective apnoeic period, i.e. the closed phase. But as of yet, no evidence has been put forward to demonstrate that this delay exists (Förster and Hetz, 2009). It has also been hypothesised that because the tracheal system can deliver O₂ at far higher rates than are required for oxidative metabolism at rest, the closed phase is initiated and maintained by a combination of high intratracheal O₂ and low intratracheal CO₂ so the spiracles remain closed to reduce internal O₂ to ‘safe’ levels (Bradley, 2007). However, this does not explain why DGCs occur, because low intratracheal CO₂ is sufficient to induce spiracle closure; the presence of high intratracheal O₂ during the open phase has no independent effect on the spiracles, except in the presence of low CO₂ (i.e. the spiracles will only close in the presence of both low CO₂ and high O₂). This being the case, gas exchange should continue with discrete breaths, not periodic bursts, each breath causing intratracheal CO₂ to drop below the hypercapnic threshold level while intratracheal O₂ remains high.

There is reason to believe that the spontaneous ventilatory periods observed in the decapitated cockroaches are the fundamental pattern underlying the emergence of the DGC. These gas exchange cycles appear to be an intrinsic property of thoracic central pattern generators that rhythmically produce ventilatory bursts of regular duration and frequency (e.g. Figs 10 and 11). Assuming an intrinsic ventilation pattern of fixed duration explains the tendency for insects to display burst phases of consistent duration during DGCs while their interburst duration may vary in response to respiratory conditions (i.e. Lighton and Garrigan, 1995; Woodman et al., 2008). However, while the ventilatory periods appear to begin and end spontaneously in decerebrated cockroaches, in intact insects it is generally accepted that the open phase is initiated when intratracheal CO₂ levels cross a hypercapnic threshold (Harrison et al., 1995; Levy and Schneiderman, 1966). It is plausible that crossing this threshold initiates the all-or-nothing response of the thoracic central pattern generators and a protracted ventilatory period of fixed duration then ensues. The long duration of this burst would be sufficient to produce intratracheal hypocapnia, increasing the time required for intratracheal CO₂ to accumulate to the hypercapnic threshold, resulting in a corrective apnoea and delaying the occurrence of the next ventilatory period. This would produce the basic periodicity of the DGC. In insects with high spiracular conductance and low respiratory demand, this automatic ventilatory burst is more than sufficient to remove accumulated CO₂ while increasing intratracheal O₂ to near-ambient levels (e.g. Figs 9 and 10). However, experimental evidence shows that the situation described above is overly simplistic. In intact insects, a multitude of afferent inputs must further modify the basic bursting pattern depending on the insect's respiratory demands and ambient conditions. For example, progressively blocking a moth pupa's spiracles gradually lengthens the duration of its open phase (Hetz, 2007; Moerbitz and Hetz, 2010), indicating that the ventilatory burst may be extended. Clearly, a more complete understanding of how and why DGCs occur could be gained from investigating the intrinsic and extrinsic factors that affect the thoracic central pattern generators.

This study reveals that cockroaches control their gas exchange to regulate intratracheal levels of CO₂ and O₂ in manner comparable to most other air-breathing animals. Under normal conditions, ventilation frequency is varied to maintain a constant internal CO₂ level and haemolymph pH. Intratracheal P_O2 is not regulated so long as it remains above a hypoxic threshold. This control is abandoned and DGCs emerge when regulation of gas exchange switches from the insect's brain to central pattern generators in the thoracic and abdominal ganglia. The experiments presented in this paper all point toward DGCs being symptomatic of reduced or absent brain activity (Matthews and White, 2011), unmasking a central pattern generator that is further modified by afferent inputs.

Our thanks go to the anonymous reviewers whose comments improved this paper; R. S. Seymour, who read an earlier version of this paper; and S. K. Hetz, for providing advice on constructing the movement detector circuit.