Respiratory gas levels interact to control ventilatory motor patterns in isolated locust ganglia

Insects use a system of gas-filled trachea to exchange respiratory gases with their external environment. Large insects actively ventilate their tracheal system in order to satisfy their respiratory demands even at rest (Harrison, 1997). The best-studied models for respiration and ventilation regulation are locusts and cockroaches (e.g. Burrows, 1996; Harrison, 1997; Harrison et al., 2013; Matthews, 2018; Miller, 1966). Under increased metabolic demands these insects increase their respiratory gas delivery by increasing their tidal volume by way of recruiting auxiliary respiration muscles (Miller, 1960; Weis-Fogh, 1967) and a higher ventilation rate (reviewed in Harrison, 1997; Matthews, 2018; Miller, 1966). Numerous studies have shown that resting insects increase ventilation rate in response to exposure to experimental hypercapnic/hypoxic conditions (e.g. Gulinson and Harrison, 1996; Krolikowski and Harrison, 1996; Matthews and White, 2011). Studies of the chemosensory mechanisms by which ventilation is controlled have mainly been conducted on intact insects, using manipulation of external gas levels or, alternatively, manipulation of endo-tracheal respiratory gases by means of tracheal system perfusion (e.g. Arieli and Lehrer, 1988; Groenewald et al., 2014; Gulinson and Harrison, 1996; Huang et al., 2014; Hustert, 1975; Matthews and White, 2011; Talal et al., 2018). Hence, the physiological factors that have been suggested to regulate the ventilation rate include endo-tracheal and haemolymph respiratory gas levels (Gulinson and Harrison, 1996; Miller, 1966, also reviewed in Matthews, 2018; Matthews and Terblanche, 2015) and haemolymph pH (Farley and Case, 1968; Snyder et al., 1980).

The neural control of the ventilation rhythm in insects has been extensively studied (reviewed by Burrows, 1996; Miller, 1966; Miller, 1981). In locusts and cockroaches, the source of the ventilation pattern was reported to be an endogenous pacemaker or central pattern generator (CPG) circuit located in the meta-thoracic ganglion complex (comprising in locusts the meta-thoracic ganglion fused with the first three abdominal ganglia; reviewed in Burrows, 1996). In locusts, direct CO₂ perfusion of head and thoracic ganglia in an exposed neural system preparation resulted in increased ventilation rate (Miller, 1960). Moreover, several components of the ventilatory CPG, i.e. interneurons whose activity could change the rhythmic pattern, have been identified in the meta-thoracic ganglion (Pearson, 1980; Ramirez and Pearson, 1989). Bustami and colleagues (Bustami and Hustert, 2000; Bustami et al., 2002) were the first to demonstrate changes in the rate of fictive (in vitro) ventilation using an isolated CNS preparation composed of the thoracic ganglia and the first unfused abdominal ganglion. These studies demonstrated a change in the output of the CPG in response to manipulation of the respiratory gas levels in the gas phase above the saline in which the preparation was bathed. The authors suggested the existence of oxygen and CO₂/pH sensors that modulate the rate of the ventilatory CPG (Bustami et al., 2002). However, despite their important pioneering work, some of the findings were at odds with observations from intact insects. For example, the effect of hypercapnia on the fictive ventilation rate was recorded at 20% CO₂ (Bustami et al., 2002), a considerably higher level than that eliciting hyperventilation in experiments using intact insects (2–3%; Gulinson and Harrison, 1996; Harrison et al., 1995; Hustert, 1975; Matthews and White, 2011). Moreover, the P_CO2 level measured in the meta-thoracic trachea was only ∼2% in post-exercise grasshoppers which exhibited a 6- to 8-fold increase in ventilation rate (Krolikowski and Harrison, 1996).

Modulation of ventilatory motor patterns is well exemplified in insect discontinuous gas exchange (DGE). This well-studied insect gas exchange pattern is exhibited during periods of quiescence or at low metabolic rates (reviewed in Chown, 2011; Chown et al., 2006; Contreras and Bradley, 2009; Lighton, 1996; Matthews, 2018; Terblanche and Woods, 2018). It comprises three phases, defined by the spiracle state as reflected by CO₂ emission rate: closed, flutter and open. The transition between the DGE phases is triggered by chemoreceptors that respond to changes in internal respiratory gas levels. It is well established that when moth pupae employ DGE, a build-up in internal CO₂ levels triggers spiracle opening and a bout of rapid gas exchange (Förster and Hetz, 2010; Levy and Schneiderman, 1966a). In a model orthopteran system, a decrease in tissue oxygenation level was also shown to trigger ventilatory movements in order to mix tracheal gaseous contents when spiracles are closed (Huang et al., 2014). Several recent studies have focused on the DGE pattern in actively ventilating insects such as locusts and cockroaches. Ventilation was shown to be tightly coupled to spiracle activity, creating a unidirectional air flow through the main tracheae (Heinrich et al., 2013; Matthews and White, 2011; Talal et al., 2018). It was further demonstrated that the rate of ventilation changes continuously during the open phase (during the active ventilation of the tracheal system in order to exchange respiratory gases), probably in response to changes in the internal gas levels (Matthews and White, 2011; Talal et al., 2018). A number of studies have reported that the duration of the closed phase is extended in response to hyperoxic conditions, even when metabolic rates remained unchanged, thus suggesting interactions between the effects of the different respiratory gases in the regulation of DGE (Levy and Schneiderman, 1966b; Matthews and White, 2011; Talal et al., 2018; Terblanche et al., 2008). Periodic fictive ventilation events, reminiscent of DGE, were demonstrated in the isolated locust thoracic ganglia preparation mentioned above (Bustami and Hustert, 2000).

The current study was therefore aimed at characterizing tentative interactions between the modulatory effects of the respiratory gases and their potential role in the control of ventilation and gas exchange motor patterns in insects. We hypothesized that ventilatory regulation is sensitive to the build-up of CO₂ in the tracheal system, as it is to interacting effects of internal O₂ and CO₂ levels as observed in intact insects (Levy and Schneiderman, 1966b; Matthews and White, 2011; Talal et al., 2018), and indeed in terrestrial animals of various taxa, including mammals (reviewed in Teppema and Dahan, 2010). In order to achieve full control over the experimental gaseous conditions, and to validate the physiological relevance of previous reports, we utilized an isolated in vitro preparation of the locust thoracic ganglia, together with the associated ventral longitudinal tracheal trunks. This allowed delivery of tightly controlled respiratory gas mixtures of various compositions to the isolated ganglia, while simultaneously controlling saline pH and gas levels. Ventilatory motor patterns were continuously monitored by recording efferent discharges under the different experimental conditions.

In this study we used desert locusts (Schistocerca gregaria Forsskål 1775) from our stock population at the Tel Aviv University. These were kept at 33.0±3.0°C, under a 14 h light:10 h dark photoperiod (supplementary radiant heat was supplied during the daytime by incandescent 25 W electric bulbs). Locusts were fed daily with wheat shoots and dry oats ad libitum. All experiments were carried out on males only, 1–3 weeks after adult eclosion.

The locusts were anaesthetized on ice prior to dissection. Following removal of the legs, wings, pronotum and abdomen posterior to the fourth abdominal segment, a longitudinal cut was performed in the cuticle along the dorsal midline of the thorax. The preparation was attached to a Sylgard dish (Sylgard 182 silicon Elastomer, Dow Corning Corp., Midland, MI, USA), the cut was widened and the gut removed. The body cavity was perfused with locust saline (Knebel et al., 2019). Air sacs and fatty tissue covering the ventral nerve cord were removed. The thoracic ganglia chain, comprising the three thoracic ganglia (pro-, meso- and meta-thoracic ganglia) and the first unfused abdominal ganglion with the surrounding tracheal supply were dissected out, pinned in a clean Sylgard dish, dorsal side up, and bathed in locust saline. The two main tracheae were opened from both sides and floated on the saline surface (Fig. 1).

Unless stated otherwise, saline was bubbled with pure nitrogen at a flow rate of 25 ml min⁻¹ using a mass-flow controller (FMA-2617A, Omega Engineering, Inc., Stamford, CT, USA). This created an anoxic saline (verified using an IMP-PSt1 P_O2 optode and a Microx TX3, micro-fibre oxygen transmitter, PreSens, Regensburg, Germany) in order to prevent oxygen diffusion to the nervous system and to ensure that tissue oxygenation was limited to tracheal supply only.

We manipulated CO₂ and O₂ levels in the main longitudinal tracheae, which supply the thoracic ganglia with respiratory gases (Burrows, 1980; Fig. 1B). Respiratory gas mixtures were perfused through a glass capillary directly into the tracheae at a flow rate of 5 ml min⁻¹. We used certified gas mixtures (79% N₂/21% O₂; 87% N₂/7% CO₂/6% O₂; 87.5% N₂/3.5% CO₂/9% O₂; 94% N₂/6% O₂; pure N₂) and two mass-flow controllers (FMA-2616A, Omega Engineering Inc.) to achieve the desired experimental gas compositions. At an accuracy/repeatability of 0.2% of full scale (20 ml min⁻¹ for the FMA-2616A), the error was <0.2% and 0.1% for O₂ and CO₂ concentrations, respectively. Following glass capillary insertion (usually at the anterior end near the pro-thoracic ganglion) (Fig. 1B), the tracheal opening was pushed under the saline, resulting in collapse of the tracheal walls against the capillary tube and its sealing. In order to ensure perfusion of the gas mixtures through both ventral tracheae (which are connected through air sacs under each thoracic ganglion) and to make sure that the perfused gases leave the tracheal system only from the far end, the tracheal opening contralateral to the capillary-connected one was also pushed under the saline surface and collapsed shut. The preparation temperature was monitored and maintained at 24±1°C.

We used custom-made suction electrodes to record extracellularly the ventilation motor pattern activity from the median nerves of the pro-thoracic and meta-thoracic ganglia, which innervate the first thoracic spiracle closer muscle and the ventilatory and spiracle closer muscles of the 4th abdominal segment, respectively (Fig. 1A). Data were acquired (5000 samples s⁻¹ for each channel) and stored on a computer for off-line analysis using a four-channel differential amplifier (Model 1700, A-M Systems, Sequim, WA, USA) and an Axon Digidata 1440A A-D board with Axo-Scope software (Molecular Devices, Sunnyvale, CA, USA).

Each experiment started 30 min following the insertion of the electrodes and the gas-perfusion capillary. Each preparation was exposed to ‘control’ gaseous conditions (see below) for 30 min before tracheal gas composition was changed for 30 min, followed by an additional 30 min exposure to the initial (control) gas composition. Endo-tracheal oxygen or carbon dioxide levels were first manipulated separately while the other gas was kept constant. Next, several additional gas level combinations were tested to elucidate the interaction between tracheal levels of the two respiratory gases. The initial and final exposures to control conditions served to verify the tissue viability and rule out any potential effects of the long exposure to dry tracheal contents.

In order to rule out the possibility that our findings reflected the tissue response to anoxic saline conditions, we compared fictive ventilation rates in anoxic saline with those recorded from preparations that were bathed in hypoxic (6% O₂ in N₂) and normoxic (21% O₂ in N₂) saline (achieved by bubbling the respective gas mixes into the saline). These experiments were carried out while perfusing the trachea with 6% O₂ and 3.5% CO₂ (in N₂).

Extracellular recordings were analysed using DataView 10.6 (W. J. Heitler, University of St Andrews, Scotland). Bursts of action potentials were identified and their rate (the fictive ventilation rhythm frequency) was determined from the last 10 min of each 30 min data recording session (no unit identification or sorting was performed). Following normality tests (Kolmogorov–Smirnov and Shapiro–Wilk), we used paired t-tests to compare the fictive ventilation rates at the manipulated and control respiratory gas levels. One data set failed the normality tests, so log₂ transformation was applied to this to allow the use of the paired t-test as above. We compared the effect of different saline oxygen levels on the ventilation rate using one-way ANOVA, followed by Bonferroni post hoc test. Statistical analyses were performed using SPSS 20.0 (IBM).

Simultaneous extracellular recordings of the median nerves of the pro-thoracic ganglion and the third fused abdominal ganglion in our in vitro preparation revealed robust fictive rhythmic ventilation-related motor patterns. The recorded pattern was characterized by perfect burst alternation (Fig. 2) that in the intact insect would result in alternation between activity of the anterior and posterior spiracle muscles and the ventilatory muscles. The in vitro frequency recorded in our control conditions was ∼5 bursts min⁻¹ at 24±1°C.

Although we used only the last 10 min of each 30 min exposure to the experimental conditions for our data analysis, changes in the efferent discharge in response to changing tracheal gas levels were typically evident within 2–5 min (Fig. 2). Decreasing tracheal oxygen level to 2% resulted in a small but significant increase in the fictive ventilation rate (Fig. 3A; t₆=5.702, P=0.001; CO₂ kept at 0%). However, no response was recorded at 10% or 4% O₂ (t₆=0.672, P=0.527 and t₅=0.882, P=0.418, respectively).

Preliminary experiments in which the tracheal perfusion comprised 21% O₂ did not result in a response to elevated (up to 7%) CO₂ levels (data not shown). Consequently, we studied the effect of tracheal CO₂ on the ventilatory pattern while maintaining the tracheal O₂ level at 6% (which was above the recorded oxygen threshold for inducing increased ventilation rate; Fig. 3A), similar to haemolymph values measured in a continuously ventilating insect (P. G. D. Matthews, personal communication). Overall, the effect of changing the CO₂ levels was much more pronounced compared with the relatively limited O₂ effect (Fig. 3B). The threshold for the CO₂ effect was found to be between 2% and 3.5%. A 2-fold increase in the fictive ventilation rate was recorded at 3.5% and at 5% CO₂ (t₅=2.582, P=0.049 and t₅=4.614, P=0.006, respectively), while 7% CO₂ resulted in a 4-fold increase (t₅=6.558, P=0.001). As noted, there was no increase in the ventilation rate when the endo-tracheal CO₂ level was 2% (t₅=0.999, P=0.364; Fig. 3B). Tissue viability during the experiment was confirmed by the almost-identical burst frequencies at the same (control) gaseous exposure at the beginning and end of each experiment (t₁₇=0.002, P=0.99 and t₁₈=0.65, P=0.53 for the O₂ and the CO₂ manipulation experiments, respectively).

While neither 2% CO₂ nor ≥4% O₂ elicited a change in the fictive ventilation rate when each gas was tested separately (Fig. 3A,B), an interaction between the effects of the two gases was evident when applying combinations of different gas levels. In a separate set of measurements (N=7), changing tracheal gas levels from 3.5% CO₂/6% O₂ to 2% CO₂/2% O₂ resulted in an increase in the ventilation rate from 6.4±0.8 to 15.4±2.3 bursts min⁻¹ (t₆=3.385, P=0.015). Similarly, when tracheal CO₂ level was maintained at 3.5% (above our recorded threshold for hyperventilation; Fig. 3B), changing the O₂ level from 6% to 3% resulted in a 2-fold increase in fictive ventilation rate (t₇=3.656, P=0.008, following log₂ transformation to reach normality; Fig. 3C), whereas a significant decrease was recorded at 9% O₂ (t₆=4.245, P=0.005, paired t-test; Fig. 3C).

Saline perfusion with nitrogen resulted in anoxic saline and prevented oxygen diffusion from the saline surface to the submerged ganglia. In order to rule out a possible artifact resulting from the anoxic saline in which the preparation was bathed, we carried out a complementary set of experiments in which we changed the saline oxygen level by means of perfusion of different gas mixtures (pure N₂, 6% O₂ in N₂ and 21% oxygen in N₂), while the main trachea was constantly perfused with 6% O₂ and 3.5% CO₂. We found that the saline oxygen level indeed affected the ventilation rate (Fig. 3D). In normoxic saline, the ventilation rate was significantly lower than that in hypoxic and anoxic saline (F_2,33=7.57, P=0.002, one-way ANOVA; Fig. 3D), as 3.5% CO₂ failed to elicit hyperventilation (rates were no different from those for 0% CO₂; Fig. 3B).

The existence and location of the ventilatory CPG in insects have been known for more than half a century (reviewed in Ayali and Lange, 2010; Burrows, 1996; Miller, 1966). However, our knowledge of the mechanisms related to gas sensing is still very limited. The findings by Bustami et al. (2002) provided the first direct evidence of the existence of oxygen chemoreceptors within the locust CNS. However, limited control over the precise gas levels close to the site of the respiratory CPG in their experiments obscured a significant response to certain physiologically relevant gas conditions. We developed a novel in vitro preparation that allowed us to examine the mechanisms of chemosensory modulation of the endogenic ventilatory CPG(s) while controlling the immediate respiratory gaseous environment. We achieved this by using direct tracheal perfusion of respiratory gas mixtures varying in composition, while controlling saline oxygenation level.

The robust and consistent alternation between the bursts of action potentials recorded from the median nerves of different ganglia throughout our experiments supports the notion that the generation and control of the rhythmic ventilation-related motor patterns are hard-wired in the insect CNS (i.e. a system of CPGs), and result in a functional unidirectional air flow in the intact insect. Furthermore, our findings clearly demonstrate the existence of CO₂- and O₂-sensing mechanisms in the thoracic ganglia, and their ability to manipulate the fictive ventilation-related output of the CPG. Throughout our experiments there was a clear time lag between a change in the perfused gas mixture and the observed change in ventilation rate (see Fig. 2). This lag was probably only the result of a combination of the flow rate and the length of the tubing used for supplying the respiratory gas mixtures.

The most interesting result in the current study is that of the interaction between the respiratory gases in their effect on the fictive ventilation rate. We have shown that the threshold level (or the sensitivity) for the modulatory effect of each gas depends on the level of the other. For example, the fictive ventilation rate doubled when the tracheal oxygen level was decreased from 6% to 3% at 3.5% CO₂, whereas at 0% CO₂ a significant (and milder: ∼20%) response was recorded only when the tracheal oxygen level was lowered to 2%. Moreover, ventilation rate at 2% O₂/2% CO₂ (15.4±2.3 bursts min⁻¹) was 3-fold higher than the rate when each gas was kept at 2% separately, while the other was at sub-threshold (for CO₂; Fig. 3A) or supra-threshold (for O₂; Fig. 3B) values. Notwithstanding the low oxygen threshold reported here (Fig. 3A), intact locusts hyperventilate when the ambient oxygen level is reduced to ∼10% (Bustami et al., 2002; Greenlee and Harrison, 1998). This apparent difference between the in vitro preparation and intact insects could be explained by the aforementioned interaction between modulatory effects of the two respiratory gases. A 2% oxygen threshold was recorded at 0% CO₂ (Fig. 3A) whereas the endo-tracheal CO₂ level of the intact insects is 2–5% (Gulinson and Harrison, 1996; Harrison et al., 1995; Levy and Schneiderman, 1966a). We show that at 3.5% CO₂, a significant increase in fictive ventilation was recorded when tracheal oxygen was lowered from 9% to 6% (Fig. 3C) to a rate 2-fold higher than normoxic values (Fig. 3A). The combined effects of the respiratory gases have also been found in intact cockroaches, in which the ventilation rate increased significantly more when exposed simultaneously to hypoxia and hypercapnia compared with exposure to the same O₂/CO₂ condition separately (Matthews and White, 2011).

The interacting effects of respiratory gases are also apparent when insects perform DGE. The CO₂ threshold level for triggering the open phase is elevated from 5% to 8% in lepidopteran pupae exposed to hyperoxia (Levy and Schneiderman, 1966b). In addition, lower haemolymph pH levels (a proxy for higher haemolymph and tracheal P_CO2) were measured in cockroaches under hyperoxic conditions just prior to the initiation of the open phase (Matthews and White, 2011). Likewise, lower endo-tracheal oxygen levels triggered the flutter phase in locusts under various levels of hypoxia (Matthews et al., 2012), possibly as a result of hypoxia-induced hyperventilation and haemolymph CO₂ washout (Matthews and White, 2011). It was also shown that both locusts and moth pupae performing DGE exhibit a longer interburst phase under hyperoxic conditions compared with controls, despite their comparable CO₂ accumulation rates during spiracle closure, suggesting a higher CO₂ threshold for triggering the open phase in hyperoxia (Talal et al., 2018; Terblanche et al., 2008). This synergistic interaction hints at a potential cellular oxygen-sensing mechanism which is mediated by CO₂ and/or pH levels.

Respiratory regulation by a cellular O₂/CO₂-sensing mechanism has never been studied in insects. In contrast, research progress has been made in other terrestrial animals, and mammals in particular. Findings in this study indicate a similar response in actively ventilating insects to that found in mammals (and other vertebrates), which increase their ventilation rate and/or tidal volume in response to local hypoxia or hypercapnia (Teppema and Dahan, 2010). In contrast to insects, tissue gas exchange in vertebrates is mediated by the circulatory system. Arterial blood oxygen, CO₂ and pH levels are monitored by the glomus cells, located in the aortic and the carotid bodies, the output of which modulates the ventilation rate (reviewed in Costa et al., 2014; Kumar and Prabhakar, 2012). There is evidence for interactions between the effects of oxygen and CO₂ on peripheral chemoreceptors in mammals (reviewed in Teppema and Dahan, 2010). Interestingly, the synergistic effect of tracheal hypoxia and hypercapnia on the respiratory CPG in locusts reported here is of similar shape to the previously reported response of the cat carotid body to changing gas partial pressures in the arterial blood (Hornbein et al., 1961) (Fig. 4). Additionally, hyperoxic and hypocapnic perfusions of the carotid sinus in dogs inhibit carotid body activity (Blain et al., 2009). Reduced sensitivity of the locust CPG was also observed in hypocapnia (Fig. 3A) or hyperoxia (Fig. 3D; Bustami et al., 2002). Hence, it is likely that the basic mechanisms underlying respiratory gas chemosensing in mammalian carotid and aortic bodies and in the insect ventilatory CPG are similar, despite their independent evolution.

Another novel insight offered by the current study is a role for haemolymph chemosensing in ventilatory CPG modulation. While the initial motivation for manipulating saline oxygenation level was to account for possible effects of anoxic saline, we were able to show that the ventilatory response to a tracheal level of 3.5% CO₂ is maintained when saline O₂ is increased from 0% to 6%, but abolished altogether in air-saturated saline (Fig. 3D). These manipulated O₂ levels are within the P_O2 range reported for the intact insect (4–10 kPa in a flight muscle of resting hawkmoth: Komai, 1998; ∼6.3±2.3 kPa in the haemolymph of a lepidopteran pupa during continuous gas exchange: P. G. D. Matthews, personal communication; see also Lehmann et al., 2019). This result could explain the previously reported high CO₂ threshold (20%) in a study which used a thoracic ganglia preparation bathed in hyperoxic saline (50% O₂; Bustami et al., 2002).

In conclusion, the current study provides important novel insights into insect central gas chemosensing and its control of ventilation. It also indicates an intriguing convergence with regulatory mechanisms in phylogenetically distinct terrestrial vertebrates. However, the location, cellular mechanism and number of sensing sites are still unknown. The relatively tractable insect preparation, together with rapidly developing genetic tools, could help in revealing the basis of a respiratory gas-sensing mechanism in insects and thus shed light on vertebrate and even human systems, for which some controversies and gaps still exist.

We thank Dr Daniel Knebel and Izhak David for helping us with locust surgical procedures and with the electrophysiological set up. We also thank two anonymous reviewers for their constructive comments, which helped improve an early version of the manuscript.