Unconventional mechanisms control cyclic respiratory gas release in flying Drosophila

Over the past decades energy consumption and metabolic activity of animals have been vigorously investigated, especially with respect to circulation and respiration. In insects, tremendous progress has been made in identifying respiratory strategies and spiracle function in resting animals, whereas relatively few studies have addressed the mechanisms of respiratory gas exchange in running and flying insects(Bartholomew et al., 1985; Berrigan and Lighton, 1994; Harrison and Roberts, 2000; Herreid II and Full, 1984; Herreid II et al., 1981; Komai, 2001; Lehmann, 2001; Lighton, 1994, 1996; Miller, 1960; Wasserthal, 2001). During rest, many insects close the outer entrances of the tracheal system via spiracles and only exchange oxygen and carbon dioxide during brief periods of time. This behavior results in a discontinuous gas exchange cycle (DGC) that may limit both toxic oxygen levels in body tissues and respiratory water loss from insects living under xeric environmental conditions (Duncan and Byrne, 2002, 2005; Hadley, 1994; Hetz and Bradley, 2005; Lighton, 1994; Miller, 1981; Slama, 1994; Snyder et al., 1995). For example, respirometric recordings in the ant Camponotus have shown that water only leaves the tracheal system when the spiracles open for gas exchange (Lighton, 1992; Lighton and Garrigan, 1995). Similar results were obtained in resting fruit flies Drosophila mimica (Lehmann,2000).

In most running insects, as energetic demands increase the DGC typically ceases, allowing respiratory gas exchange rates to increase likewise(Full and Tullis, 1990; Full et al., 1990; Jensen and Holm-Jensen, 1980;for reviews, see Lighton, 1994, 1996). In the desert ant Pogonomyrmex rugosus, Lighton and Feener(1989) reported a discontinuous breathing pattern while the unrestrained animal was walking voluntarily at constant speed within a respirometric chamber. A similar breathing pattern was found in blowflies Protophormia terraenovaewalking back and forth in a running tube(Berrigan and Lighton, 1994). Although the first study was originally interpreted as a rare example in which the environment constrained breathing behavior to avoid excessive water loss,a recent study on walking energetics in ants suggests that DGC-like respiratory behavior in walking animals may also result from Doppler-effects occurring inside a running tube under flow-through respirometric conditions(Lipp et al., 2005).

In flying insects, flight-specific metabolic rate increases up to 15-fold over resting values and spiracles open in order to allow gas exchange rates to increase (Casey, 1980, 1989; Casey and Ellington, 1989; Harrison and Roberts, 2000; Hedenström et al., 2001; Lehmann et al., 2000; Miller, 1960; Moffatt, 2001; Wasserthal, 2001; Weis-Fogh, 1972). In small animals such as the fruit fly Drosophila, the uptake of oxygen into and the release of carbon dioxide out of the tracheal system are thought to be a diffusion-based process, and thus the spiracle opening area matches the metabolic needs of the animal (Lehmann,2001; Weis-Fogh,1964). Applying the diffusive theory of respiration, adaptive spiracle control in Drosophila may potentially establish constant oxygen partial pressures near atmospheric partial pressure within the tracheae(Lehmann, 2001). Large insects additionally ventilate their tracheal air system to satisfy the increased oxygen needs of four distinct but interacting mechanisms: (i) contraction of the abdomen (abdominal pumping; for example, Harrison and Roberts, 2000; Komai, 2001; Miller, 1960); (ii)potentially, by muscle-induced deformations of large tracheae, as shown in insects breathing under X-ray in a synchrotron(Westneat et al., 2003); (iii)thoracic auto-ventilation resulting from the vibrations of the thorax during wing flapping (Miller, 1966);and (iv) directed Bernoulli suction-ventilation due to differences in static pressure distribution above two thoracic spiracles(Miller, 1966). In the hawkmoth Manduca sexta, for example, auto-ventilation produces pronounced pressure fluctuations in-phase with the 20 ms wing flapping cycle,causing inhalation during the downstroke and exhalation during the upstroke of respiratory gases on a stroke-by-stroke basis(Wasserthal, 2001).

In contrast to convective flow, diffusion-based respiratory gas exchange mechanisms for flight have in common that the instantaneously measured gas exchange rate is thought to reflect the animal's actual respiratory demands because there is no bulk flow of respiratory gases into and out of the tracheal system (Kestler,1985). Experimentally, oxygen demands and the magnitude of CO₂ release rate from tethered Drosophila can be controlled by flying the animal in a respirometric chamber under visually controlled feed-back conditions (Dickinson and Lighton, 1995). In response to the vertical motion of a visual pattern displayed in a surrounding panorama, the animal modulates the muscle mechanical power output of its asynchronous flight muscles, and thus metabolic rate. When visual lift stimuli are absent, the rate with which the fly releases CO₂ changes only slightly because flight muscle mechanical power output, and thus metabolic rate, is not extensively modulated by the fly's nervous system.

In comparison to previous findings, we here report a novel type of gas release pattern in an insect by demonstrating that instantaneous CO₂ release rate of tethered flying Drosophila may periodically oscillate with large amplitudes, even though the metabolic rate of the animal remains approximately constant. In general, oscillatory gas release patterns in insects flying at constant metabolic rate may result from at least two distinct major mechanisms: ventilation and changes in the gas exchange area of the spiracles. While the first mechanism results from compression of air sacs and tracheae, the second one relies on spiracle control strategies. We thus evaluated the underlying physiological mechanisms of cyclic breathing behavior at an integrative level of investigation by combining (i) CO₂ release measurements of single fruit flies flying in a virtual-reality flight simulator with (ii) video-based in-flight tracking data of abdomen and proboscis movements, and (iii) employing an analytical model for tracheal gas diffusion through spiracles that allows simulations of tracheal CO₂ partial pressure changes and gas release rates at various metabolic rates.

The methods used in this study have been published elsewhere in greater detail (Lehmann, 2001; Lehmann and Dickinson, 1997)and we provide only a brief description here. Female Drosophila melanogaster Meigen (4–7 days old) from a laboratory colony were tethered and flown in a 15 ml flow-through respirometric chamber. Water vapor and CO₂ were removed from room air using a Drierite®/ascarite column, pulled at 1000 ml min^-1 flow rate through the chamber using a mass-flow controller (El-Flow, Bronkhorst, Ak Ruurlo, Netherlands), and subsequently sampled in a gas analyzer (Licor-7000, Licor, Lincoln, USA). The internal filter frequency of the gas analyzer was set to 0.2 s. Data sampling frequency was 125 Hz and wash-out time constant τ of the respirometric chamber was approximately 910 ms (wash-out time =ke^-x/τ). We estimated the time constant by inserting a small tube inside the respirometric chamber, through which we released small amounts of CO₂ similar to those released from a flying fly. Since the wash-out time constant depended exponentially on flow rate, a high flow rate yielded better temporal resolution but produced relatively low maximum gas concentrations, of approximately 1.0 p.p.m. CO₂. The signal-to-noise ratio (SNR) was approximately 18.6 dB. Flight muscle-specific mechanical power output was calculated from measurements of total flight force production, wing stroke amplitudes and stroke frequency according to energetic theory for flapping flight (Ellington,1984; Lehmann and Dickinson,1997). The respirometric chamber was surrounded by a computer-controlled cylindrical array of light-emitting diodes that allowed open- and closed-loop visual stimulation of the flying animal. Similar to previous procedures, we modulated the mechanical power output of the fly's flight muscle by oscillating a visual stripe grating in open-loop vertically around the fly, while the fly actively controlled the azimuth velocity of a single black stripe displayed in the arena (closed-loop conditions; Lehmann and Dickinson, 1997). All experiments were performed at approximately 28°C ambient temperature. For data analysis and modeling we employed self-written software routines in LabTalk (Origin 7.0, Microcal, Northampton, USA).

To evaluate abdominal pumping during flight of Drosophila, we developed an automatic tracking technique that allowed simultaneous recordings of abdominal length and width while the animal was flying inside the respirometric chamber. For this purpose, we marked the underside of the fly's abdomen with four small droplets of commercial yellow fluorescent dye and illuminated these markers using a UV-light emitting diode. While measuring wing kinematics and carbon dioxide release, we recorded in-flight movements of the abdomen using a conventional 50 Hz video camera (frame rate=20 ms). Subsequently, the positions of the four fluorescent markers were automatically tracked using a commercial video analysis program (MaxTraq, Innovision,Columbiaville, MI, USA). In-flight extensions of the fly's proboscis were derived by analyzing light intensity changes on the video images in a rectangular region (ROI) in front of the animal's head. To allow easy adjustments of ROI size and shape for each fly, we employed self-written software developed under Visual C++ and Matrox Imaging Library (Matrox,Quebeck, Canada).

Fig. 1B shows a typical flight sequence of a tethered flying fruit fly in which the animal exhibits both non-cyclic gas release (gray) and periodical oscillation of CO₂ release (blue), while producing only small-scale fluctuations in flight-muscle specific mechanical power output. The data apparently show that gas release oscillations occur randomly in a flight sequence and often in bursts of up to 18 cycles in a row (Figs 1C, 2A,B). The frequency of the respiratory cycles is approximately 0.37±0.055 Hz (mean ± s.d., 180 cycles, N=12 flies; for single flies, mean s.d. = 0.097 Hz; Fig. 2C,D). On average, gas release rate during oscillatory breathing behavior is not significantly different from flight sequences yielding non-oscillatory CO₂ release patterns (t-test, P>0.05, `cyclic' CO<sub>2</sub> release = 26.5±1.5 ml g<sup>-1</sup> flight muscle h<sup>-1</sup>, `non-cyclic' = 26.6±1.2 ml g^-1 h^-1, N=19 flight sequences, 401 s total flight time), suggesting that oscillatory breathing does not result from a pay-off of anaerobic debt during elevated locomotor activity. Moreover, Fig. 1C shows that oscillatory gas exchange patterns in Drosophila mostly occur in flight sequences without visual stimulation (section III, Fig. 1C), during which the temporal changes in metabolic rate are negligible and total flight force remains close to hovering force production (flight force/body weight =0.93±0.20, body mass = 0.99±0.22 mg, means ± s.d., N=8 out of 12 flies; Fig. 1C, sections II and III). We rarely observed oscillatory gas release patterns near maximum locomotor capacity of approximately 1.6 times the hovering flight force, suggesting that cyclic breathing is not needed to satisfy oxygen demands at maximum metabolic rate. Experiments in which we changed the order of the presented stimulus conditions revealed that cyclic gas release does not occur as a consequence of previously applied open- or closed-loop feedback conditions. To validate the relationship between locomotor output of the animal and gas release statistically, we calculated the temporal cross-correlation coefficient r between flight muscle mass-specific mechanical power output and CO₂ release rate (Fig. 3A,D,G). Superficially, the small coefficients show no preferred phase lag between both parameters, implying that none of the temporal fluctuations during CO₂ release oscillations are correlated with the animal's changes in metabolic rate (maximum r=0.07 at phase lag =–0.08s).

To tackle the significance of abdominal pumping as a source of the measured CO₂ fluctuations in Drosophila, we monitored the geometry(length and width) of the abdomen during flight using an automatic video tracking technique (Fig. 3B). Although the data that have been reconstructed from single video images show systemic changes in abdominal geometry, the changes were quite small and consistently below approximately 100 μm(Fig. 3E). The temporal cross-correlation coefficient r, averaged over six flight sequences all exhibiting pronounced and long-lasting gas release oscillations (sequence length = 141±77 s, mean ± s.d.), shows that superficially none of the abdominal length changes appear to be correlated with the cyclic changes in gas exchange rate (maximum r=0.06 at phase lag = –0.44 s, Fig. 3H).

As a second explanation for our experimental data, we considered whether cyclic gas exchange is due to changes in haemolymph pressure caused by any ventilatory function of the proboscis. In Drosophila, the proboscis is relatively large and its volume amounts to approximately 15–20% of the head's volume (∼0.2 mm³; Demerec, 1965). Fruit flies regularly extend and retract their proboscis during flight, a behavior that might enlarge (inhalation) or compress (exhalation) the large paired frontal,postgenal and postocular air sacs in the fly's head(Lehmann et al., 2000). We quantified voluntary proboscis movements by mapping light intensity changes of the video images in a defined region shown by the red box in Fig. 3C. Cross-correlation analysis between the derivative of proboscis motion and CO₂ release shows the following. (1) In single flies, up to 80% of the variance in tracheal CO₂ release fluctuations can be assigned to PER activity. The mean correlation coefficient amounts to 0.31±0.24 (mean ± s.d., N=6 sequences, 3 flies, Fig. 3I), which is at least fivefold higher than the maximum temporal correlation coefficient between CO₂ release and abdominal movements (0.31 vs 0.06). (2)There is a small temporal shift of the maximum cross correlation value with respect to zero phase lag, indicating that the proboscis starts moving approximately 0.6 s before the fly changes its gas release rate(ΔL, Fig. 3I).

Besides the employment of convective strategies for breathing, oscillatory gas release in diffusion based respiratory systems may also result from changes in spiracle opening area, similar to the mechanism causing the discontinuous gas exchange cycle in a resting insect(Lighton, 1994; Miller, 1981; Slama, 1994; Snyder et al., 1995). In this scenario, the proboscis extension reflex (PER) would only be correlated with CO₂ release rather than being the primary cause for its oscillatory behavior. In insects, spiracle muscle control is CNS-mediated but the muscle activity also depends on local concentrations of respiratory gases (for a review, see Nikam and Khole,1989). In the locust Schistocerca gregaria, for example,the mesothoracic closer muscle is innervated by two excitatory motorneurons and a peripherally located neurosecretory cell(Swales et al., 1992). Due to their small size it is difficult to record electrically from spiracle muscles in flying Drosophila. The same holds for direct measurements of total spiracle opening area, because the mesothoracic spiracle is partly covered during flight by the beating haltere. For this reason, we cannot reject per se the hypothesis that the CNS synchronously opens and closes multiple thoracic spiracles via excitatory motorneurons.

However, even assuming that no CNS activity is involved in spiracle control and that each spiracle functions autonomously, gas release may start to oscillate in cases where several spiracles stochastically synchronize their opening activities. In flying Drosophila, total spiracle opening area matches the respiratory exchange area of the diffusive path exactly to the actual needs, and thus tracheal partial pressures for CO₂ and oxygen are stabilized in a narrow range between 1.3–1.4 kPa and 19.8–19.9 kPa, respectively(Lehmann,2001). It seems thus likely that spiracles open and close around a threshold value for tracheal partial gas pressures. Inspired by these observations, we analytically modeled gas release for each of Drosophila's four large thoracic spiracles,employing a simple diffusive model for the control of tracheal CO₂partial pressure (see Materials and methods, Fig. 4A). Using the `fruit fly-inspired' characteristics of spiracle function, and depending on the parameter settings, each model spiracle is able to produce small-scale fluctuations in tracheal CO₂ partial pressure at constant metabolic rate autonomously. Consequently the rate at which muscular CO₂enters the distal endings of the model tracheae (tracheoles) changes only slightly (slope of blue line, Fig. 4B). As long as the four model spiracles open and close out of phase, the total release rate of CO₂ is non-cyclic (gray, Fig. 4C). If certain conditions are met, however, opening activity randomly synchronizes, producing large oscillations of gas release similar to the pattern observed in the flying fruit fly (blue, Figs 1B, 4C). This result is independent of the initial tracheal partial pressure(Pt_CO2) for the simulation that was shown by selecting 20 different randomly distributed starting values for Pt_CO2(Fig. 4D).

Multiple variations of the analytical model parameters show that the likelihood of the occurrence of oscillatory gas release depends at least on two parameters: (i) the ratio between muscular partial pressure and the spiracle threshold values for that gas(Fig. 5A,B) and (ii) the ratio between maximum conductance of each model spiracle and the simulated conductance of the cytoplasm between muscle tissue and the tracheoles(Fig. 5C,D). The first prerequisite implies that cyclic CO₂ release only occurs when the tracheal partial pressure for CO₂ is allowed to increase when spiracles are held predominantly closed, and to decrease when the spiracles are predominantly open over time. The analytical model thus predicts that oscillatory release patterns disappear when tracheal partial pressure for CO₂ is consistently below or above the spiracle threshold value for this tracheal gas (black traces, Fig. 5B). At high metabolic rates, for example, all four model spiracles stay predominantly open (between 0.5 and 1.0G_S,max) because the tracheal partial pressure of the respiratory gas remains above the opening threshold value(Pm_CO2/T_s≥1.1, Fig. 5A,B). Under these conditions, any changes in gas release rates reflect changes in metabolic rate or differences in partial pressure between the tracheal system and the ambient air, but not changes in the spiracle's own diffusive exchange area. In contrast, the changes in gas release pattern are more subtle when changing the ratio between cytoplasm and maximum spiracle conductance(Fig. 5C). At the given parameter settings, small ratios below 0.1 suppress the likelihood of cyclic release patterns because CO₂ gets completely removed out of the tracheal system even at the leaky (`predominantly closed') spiracle state. In contrast, higher conductance ratios between 0.2 and 0.8 consistently produce pronounced temporal fluctuation in gas release rate. With increasing G_S,max/G_c ratio >0.1, however, the amplitude of the data's main FFT frequency component (i.e. equal to spiracle hysteresis δ) simultaneously decreases, and the broader distribution of frequency components in the time domain indicates that the gas release rates become increasingly noisy (Fig. 5D).

The data presented in the present study suggest that cyclic gas release patterns during constant locomotor activity in flying Drosophilamight result from at least three distinct respiratory mechanisms: (i)ventilation caused by the proboscis extension reflex, (ii) synchronized spiracle activity mediated by the CNS which we could not validate experimentally, and (iii) stochastic synchronization of autonomously working spiracles, as suggested by analytical modeling. Although PER ventilation and spiracle synchronization behavior might result in similar breathing patterns,their functional relevance for the animal is quite different. PER ventilation produces convective flow and should thus increase (decrease) tracheal partial pressure of oxygen (CO₂) in the flying fruit fly, whereas as spiracle synchronization behavior should not affect local partial pressure for tracheal gases. PER mediated ventilation thus compares to the most common mechanism for ventilation in insects, i.e. abdominal pumping (for a review,see Wigglesworth, 1972). In contrast to other insects, however, the small cross-correlation coefficients below approximately 0.06 (Fig. 3H) suggest that abdominal pumping is not the primary source for cyclic gas release patterns in flying Drosophila. Moreover, it is also unlikely that mechanical deformations of the thoracic box(auto-ventilation) cause cyclic breathing patterns, as mentioned in the Introduction, for two reasons: firstly, the length changes of the thorax between upstroke and downstroke of the wings in Drosophila are relatively small and amount to not more than approximately 2% of the overall thoracic length (Chan and Dickinson,1996). Secondly, wing stroke frequency in Drosophila is approximately 200 Hz and thus 540 times higher than CO₂ cycling frequency of approximately 0.37 Hz(Lehmann and Dickinson, 1998). Instead, the high cross-correlation coefficient between tracheal CO₂ release fluctuations and PER in conjunction with the temporal timing between both events (0.6 s phase advance of PER) strongly suggests a functional relevance of PER for respiratory gas exchange. This view on respiration in the fruit fly is also supported by the finding that the variance of PER frequency in single flies and among different flies is notably small (single fly = 2.6% of 0.37 Hz, multiple flies = 0.8% of 0.37 Hz),suggesting an internal neuronal pacemaker for the PER breathing rhythm rather than random activity of the proboscis' retractor and extensor muscles. Alternatively, we should also take into consideration that PER-mediated convection simply occurs as a by-product of proboscis extensions, serving a different yet unknown task during flight. In conclusion, our analysis offers a novel and quite unconventional mechanism for reinforcing respiratory gas exchange rates in an insect. Considering the similarities in morphology,proboscis-induced ventilation might become increasingly important for larger flies that rely on convective flow into and out of their tracheal system during elevated locomotor performance.

In the past, several researchers have attempted to model respiratory processes. At the single cell level, Thumfort et al.(2000) recently modeled oxygen diffusion by computer simulation in three dimensions and applied this model to a case study. They found that generating a one-dimensional representation of the three-dimensional surface of the cell is a close approximation to the more complex three-dimensional model with systematic differences below 10%. Research on diffusive models of the entire tracheal system of insects was pioneered by a study of Weis-Fogh(1964), who developed an extensive set of equations for steady state diffusion in an isotropic tissue,tracheal gas transport and exchange (air–tube diffusion). Weis-Fogh, for example, also considered different topologies of diffusive systems and estimated their effect on gas exchange. Later, Kestler(1985) also modified these models towards ventilation and focused on different gas exchange models describing respiratory flux between the tracheoles and mitochondria. Lehmann(2001) applied Kestler's model to diffusive gas exchange and respiratory water loss through the spiracle in tethered flying Drosophila and demonstrated how the water loss rate can be used to derive total spiracle opening area in the animal in vivo. Snyder et al.(1995) proposed an elaborate model for cyclic ventilation in insects that also covers the three phases of cyclic ventilation. The authors basically reported that volume expansion of the trachea, and not an increased cross-sectional area of the spiracles per se, is the important adaptation to normobaric hypoxia. However,although all elegant, none of the elaborate studies above have considered the potential complex temporal interactions in diffusive gas exchange between multiple spiracles.

The simplified analytical model of tracheal diffusion presented here offers an alternative explanation for the experimental data achieved in the flying fruit fly. The model proposes that the temporal switching between non-cyclic and cyclic breathing patterns can be explained by phase transitions between(i) times during which opening activity of the four autonomously working thoracic spiracles in the fly synchronize (cyclic breathing) and (ii) times at which the spiracle-controlled total diffusive areas are temporally out-of-phase (non-cyclic breathing). Interestingly, recent data on respiration in resting ants Camponotus seem to support this scenario also occurring during the discontinuous gas exchange cycle(Lipp et al., 2005). Multiple studies have shown that resting ants typically release a single peak of CO₂ during the DGC's opening phase (`O'-phase; Lighton, 1992, 1994, 1996; Lighton and Feener, 1989),which is consistent with the idea of synchronized spiracle opening activity,assuming that multiple spiracles participate in gas exchange. In comparison,the recent study on Camponotus gas release measured with high-temporal resolution showed that resting animals also release CO<sub>2</sub> as multiple `O'-peaks within a single DGC cycle(Lipp et al., 2005). One explanation for the latter finding could be that it is a consequence of desynchronized opening activity of at least two spiracle muscles, which compares to the non-cyclic gas release pattern in our flying insect. The temporal distribution of the gray areas in Fig. 1B,C shows that in flying Drosophila, the desynchronized spiracle opening condition seems to be more common and, in addition, not all tested flies have shown cyclic breathing patterns. In most recordings of flying fruit flies, instantaneous gas exchange rate thus matches the actual metabolic needs of the animal, as reported previously (Lehmann,2001).

One of the most interesting predictions of our analytical model is that cyclic CO₂ release patterns do not necessarily require large changes in diffusive area of a single spiracle during control behavior, as indicated by the small (±3%) tracheal partial pressure changes in the example shown in Fig. 4B. Under certain conditions, these small changes in diffusive area, in conjunction with concomitant small changes in tracheal partial pressure, appear to be sufficient to modulate total CO₂ release rates of up to ±50%peak-to-peak of the mean value, when all four model spiracles open and close at similar phase, as shown in Fig. 4C. The most critical prerequisite for the occurrence of CO₂ release cycling of the analytical model seems to be that tracheal pressure of CO₂ decreases (increases) when the model spiracle opens between 50% and 100% (0 and 50%) of the maximum diffusive area,as mentioned above. If metabolic rate causes tracheal partial pressure for CO₂ to increase above the spiracle opening threshold(T_s), cyclic respiration vanishes and the modeled instantaneous CO₂ release rate matches instantaneous metabolic activity. This observation could potentially explain why, in the experiments performed with flying Drosophila, oscillatory gas release patterns only occurred at flight forces and metabolic rates well below maximum locomotor capacity.

Due to the lack of elaborate data for Drosophila's respiratory system, including muscular partial pressure estimates, gas conductance of the cytoplasm and flow conditions inside the tracheae, it is difficult to model gas release using exact physiological values(Kestler, 1985). The proposed analytical model, including all parameter settings, should thus be considered as a rough hypothesis on how tracheal gas release can potentially be shaped by stochastic spiracle opening and closing processes. Moreover, our simplified analytical model makes some inherent assumptions about spiracle control strategies of the living organism (e.g. binary random function for spiracle opening/closing behavior). The results derived from the analytical model,including any comparison with data recorded in the flying animal, should thus be treated with caution. Nevertheless, considering all limitations and problems of our simplified analytical model for insect respiration, the coincidence between the data produced by the simulation and the flying fly is marked, and thus might highlight a fundamental inherent property of spiracle function in diffusion-based tracheal systems of small insects.

In sum, this study proposes that periodically oscillating gas release patterns in flying Drosophila might result from at least two unconventional respiratory mechanisms: firstly, the proboscis appears to serve as a pumping organ for ventilation, and secondly, gas release oscillations may come about by synchronized opening activity of the large thoracic spiracles similar to the DGC. Interestingly, in the fruit fly both mechanisms are thought to have little significance for flight muscle function, and ventilation is apparently not required to satisfy the high oxygen demands even at maximum locomotor capacity (Weis-Fogh,1964; Lehmann,2001). Instead, a possible advantage of the proboscis extension reflex for tracheal ventilation might be to actively promote local oxygen supply of the animal's head. In dipteran flies, the retina and optic lobes may require at least 20% of the resting metabolic rate due to the highly specialized and large visual system of flies(Laughlin, 1987). Since there are no spiracles in the head, respiratory gases must pass through small tracheae inside the approximately 80 μm diameter neck connective(Demerec, 1965). If correct,the hypothetical benefit of the proboscis-induced ventilation for breathing might be to circumvent this bottleneck for diffusive respiration, in order to ensure evacuation of CO₂ from – and the supply of oxygen to– the fly's brain. Eventually, this behavior might balance tracheal partial pressures of respiratory gases locally within the body compartments of Drosophila during certain flight conditions.

List of symbols and abbreviations

 
• DGC

  discontinuous gas exchange cycle

 
• dNa

  outflow of gas molecules through the open spiracles

 
• dNt

  number of gas molecules entering the tracheal system from the muscle

 
• IFM

  indirect flight muscle

 
• G_c

  gas conductance of the cytoplasm

 
• G_S

  instantaneous gas conductance of the spiracle opening

 
• G_S,max

  maximum spiracle opening area

 
• k

  Boltzmann constant

 
• Ṁ _CO2,tot

  total CO₂ release rate

 
• Ṁ _M

  metabolic rate of the flight muscle

 
• Ṁ _S

  gas flux through the spiracle

 
• Ṁ _T

  temporal flux of CO₂ molecules entering the tracheoles of the tracheal system

 
• Pa_CO2

  partial pressure of CO₂ in ambient air

 
• PER

  proboscis extension reflex

 
• Pm_CO2

  partial pressure of CO₂ in the muscle

 
• Pt_CO2

  partial pressure of CO₂ in the trachea

 
• r

  cross-correlation coefficient

 
• rnd

  equally distributed random function

 
• SNR

  signal-to-noise ratio

 
• T

  temperature

 
• t

  normalized time

 
• t

  time

 
• T_s

  tracheal partial pressure threshold value

 
• Vt

  tracheal volume

 
• ΔL

  cross-correlation temporal phase shift between data sets

 
• δ

  response time of the model spiracle to changes in tracheal gas concentration (spiracle hysteresis)

 
• τ

  time constant

We would like to acknowledge Simon Pick for his help with the analytical model, Markus Mronz for developing video imaging software routines and Harald Wolf for critically reading a previous version of the manuscript. This work was funded by the Biofuture grant 0311885 of the German Federal Ministry for Education and Research to F.O.L.