Course-Control, Metabolism and Wing Interference During Ultralong Tethered Flight in Drosophila Melanogaster

Investigation of ultralong tethered flight in Drosophila touches several topics. (1) Orientation towards a singular visual object (fixation) prevails during early flight and is, therefore, considered as an invariant goal of course-control in a flight’ simulator. However, other behavioural options, including orientation in the opposite direction (anti-fixation), are pre-programmed in Drosophila (Götz, 1980, 1983b; Bülthoff, Götz & Herre, 1982) and might dominate course-control as flight is continued. (2) Object-induced course-control requires continuous visual attention (Wolf & Heisenberg, 1980; Heisenberg & Wolf, 1984), which is hardly expected to last throughout the ‘circadian’ cycle of an insect (Kaiser & Steiner-Kaiser, 1983). (3) The comparatively low carbohydrate consumption during tethered flight in Drosophila (Wigglesworth, 1949) suggests unexpectedly high efficiencies of the flight system (Weis-Fogh, 1972). Incomplete lift production and deviation from the equilibrium between expenditure and regeneration of stored reserve substances may account for this discrepancy. (4) Lift-generating aerodynamic mechanisms of the ‘clap and fling’ type (Weis-Fogh, 1973; Ellington, 1980, 1984a, b; Götz, 1983a) are not required to sustain tethered flight. Omission of the ‘clap and fling’ could increase the apparent efficiency and perseverance of a tethered flight system. (5) Estimates of the perseverance in course-controlled free flight may help us to understand the results of field studies where the dispersal of marked strains was measured as a function of time (Powell & Dobzhansky, 1976).

The experiments shown in Fig. 1 were performed with 3-day-old female flies of wildtype ‘Berlin’. The flight simulator used to investigate object-induced course-control is described in the legend of Fig. 1A. The cold-anaesthetized flies were attached, with dental cement, either to a dorsal holder (Fig. 2) or to a ventral holder (Figs 1, 5) of an opto-electrical device for the evaluation of wing movement. Magnitude and direction of the intended turns of the tethered fly were derived from the difference of the wingbeat amplitudes on either side (Götz, 1968, 1983a). Simulation of free flight was accomplished by servo-controlled rotation of a visual object in a direction opposite to that of the intended turns. The angular position of the object was sampled at a rate of 4s⁻¹. The histograms in Fig. 2 show the relative time spent by the object in 20 sectors of 18° bin-width.

The triggering signals from the opto-electrical device were used to inspect, stroboscopically, the wing movement of the tethered fly, and to take stereo-photographs of the wings in selected phases of the beat cycle. Details are given in the legend of Fig. 1B. A series of photographs was made in the wind tunnel to show the influence of airspeed on the interference of the wings during the transition between upstroke and downstroke.

Uninterrupted tethered flight in Drosophila rarely persists for more than 3 h. This period decreases progressively in successive flights unless the fly is ‘motivated’ by the simulation of free flight. The closed-loop reactions of a single fly in the flight simulator are shown in Fig. 2. Within 39-1 h we obtained 18 successive flights, each of which was initiated by a break of about 22 min for feeding and preening. Subtraction of the breaks results in a total flight period of 32-2h. This time scale applies to the data in Fig. 2. During the first 20 s of a break the fly was fed by means of a tiny scrap of tissue soaked with a 0·5 mol 1⁻¹ sucrose solution and held by a pair of tweezers. Between 10 and 30 h of flight, the uptake of carbohydrate was derived from the difference between the weight of both the tissue and its attachment immediately before and after feeding. The air in the room, and particularly under the hood of the balance, was moistened to keep the humidity close to saturation. The evaporation of solvent from the tissue used for feeding under these conditions amounted to less than 0·03 mg in 20 s. This approximation was obtained in a sham feeding experiment, and resulted in minor corrections to the calculated carbohydrate uptake. The accuracy of the balance (about 0·05 mg) corresponds to an amount of solution containing about 0·009 mg sucrose. The relationship between energy consumption and the duration of the ensuing flight is shown in Fig. 3. The uptake of sucrose as the only nutrient may have led to dietary problems preventing further flight in the simulator. Back on standard medium, the fly survived the experiment for 30 days.

Wingbeat frequency and room temperature were measured throughout the experiment. The means and standard deviations are 196 ±24 Hz and 24·4±0·7°C, respectively. After 32·2 h of flight the wings have made about 45 million half-strokes. The lift produced during tethered flight is not sufficient to support all of the body weight (Götz & Wandel, 1984). However, the perseverance of flight in Drosophila is impressive and is likely to explain the enormous distances covered by some active flies of a population (Powell & Dobzhansky, 1976).

Continuous course-control, another prerequisite of locomotor efficiency, is even more sensational in the present experiment. The histograms in Fig. 2, obtained during the first 30 h of flight, show the time course of the preferred orientation with respect to a grey vertical bar. A maximum close to the middle of the abscissa prevails in the histograms of the first 24 h. This proves that a course towards a visual landmark can be maintained with astonishing persistence. Fixation of the bar as the primary target appears to be an attempt to approach a landing site. The failure to arrive at the target may have led to a number of tentative deviations from the original course during the last 8h of the experiment. Orientation towards a visual target is obviously not a ‘rigid’ reflex in Drosophila. The ability to explore different options, or to focus attention on different segments of the visual field, has been established in closed-loop experiments with two or more targets (Götz, 1980, 1983b; Wolf & Heisenberg, 1980; Bülthoff et al. 1982; Heisenberg & Wolf, 1984). The results of the present experiment show the ‘flexibility’ of orientation in the fly.

Holding the bar in a selected angular position of the visual field requires continuous attention and efforts to overcome fluctuation and drift imparted by the beating wings. The fly counteracts the escape of the visual object from the selected position by intended ‘body saccades’ which were produced at an average rate of about 5 s⁻¹ (K. G. Götz, unpublished results). The fluctuation of the bar about a selected position on the screen of the flight simulator in Fig. 1A demonstrates the impressive frequency, speed and strength of the course-correcting manoeuvres in Drosophila. The manoeuvres resemble the actions of a driver under the stress of heavy traffic. With regard to this analogy it is most astonishing that the attention and effort of the fly were not significantly reduced during the 32 h of flight in the simulator. With the exception of a brief period of idle flight in the last hour of the experiment, the coursecorrecting manoeuvres have been neither discontinued nor modified. A ‘circadian cycle’ of attention or activity (Kaiser & Steiner-Kaiser, 1983) was not observed in the present experiment.

Using a quasisteady aerodynamic approach, Weis-Fogh (1972, 1973) calculated the mean aerodynamic power requirements per body weight for hovering in one of the larger species of Drosophila. The approach, as well as the result of about 2·4WN’¹, have been convincingly challenged by Ellington (1984a), who showed that lift production in a hovering insect is almost entirely due to non-steady rotational mechanisms. Evaluation of these mechanisms suggested, however, that estimates of power requirements would not deviate too much from those obtained under quasisteady conditions. Using Ellington’s approach, Laurie-Ahlberg et al. (1985) calculated power requirements per body weight of 1·7WN⁻¹ for hovering in Drosophila melanogaster. Similar results can be expected in other species of this genus.

The proportion of flight muscle weight to body weight found by Chadwick & Gilmour (1940) in Drosophila repleta is 0·16. This appears to be a reasonable estimate for the different species of Drosophila. The expected power requirements for hovering related to the weight of the flight muscles thus amounts to 1·7/ 0·16WN⁻¹, or about 11 WN⁻¹. The maximum weight-specific mechanical power output of vertebrate striated muscles is at least twice as high. Calculations by WeisFogh & Alexander (1977), Pennycuick & Rezende (1984) and Ellington (1985) suggest 26—40 W N⁻¹ as a reasonable upper limit of the mechanical power output of the asynchronous fibrillar flight muscles in insects. Although direct evidence concerning the efficiency of these muscles at wingbeat frequencies of about 200 Hz is missing, the calculated mechanical power output for hovering in Drosophila appears to be well below the physiological limit of the flight muscles. The availability of surplus power facilitates energy-consuming manoeuvres such as the sudden transition between hovering and fast forward flight (Götz, 1983a ; Götz & Biesinger, 1983).

It seems to be accepted that aerobic metabolism of carbohydrates is the only significant source of flight energy in Drosophila. Neither fat (Wigglesworth, 1949) nor protein (Chadwick, 1947) contributes significantly to metabolism during flight (for other insects see Sacktor, 1975). Table 1 shows the metabolic power input per body weight derived from the flight-specific uptake of different sugars, consumption of oxygen or depletion of stored glycogen in several species. A conversion factor of 16 J mg⁻¹ carbohydrate, or 20 J mlO₂⁻¹, has been used to determine the total input of chemical energy. The metabolic power required to sustain tethered flight in Drosophila is about 12 times the resting level, or about 92 % of the total consumption of chemical energy per time interval (Chadwick & Gilmour, 1940). The mechanochemical efficiency in Table 1 is the proportion of the expected mechanical power requirements of about 1·7 W N⁻¹ to the flight-specific metabolic power input during tethered flight. This proportion is the upper limit for the actual mechanochemical efficiency of the flight muscles. During tethered flight in still air, Drosophila melanogaster produces only between 27 % (three flies, Götz & Wandel, 1984) and 44% (14 flies, Götz, 1968) of the force required to support the body weight. An average force production of 41 % of the body weight is likely to require, as a first approximation, a similar fraction of the expected mechanical power of 1·7WN⁻¹. (The Rankine-Froude relationship P_indℝ F^3/2 between the induced power and the resulting force might be used to improve the approximation.) The factor of 0·41, tentatively applied to the maximum efficiencies in Table 1, yields, for the fruitfly, actual efficiencies between 0·03 (line 3) and 0·10 (line 2). The results correspond to the efficiencies for the hovering insects Apis (0·05), Bombus (0·06) and Eristalis (0·08) determined by Ellington (1984a, 1985).

The present data on carbohydrate uptake are shown in Fig. 3 and Table 1, line 1. The results were obtained under unusual conditions. (1) The fly was held in an attentive state requiring a high rate of course-correcting manoeuvres. The freedom to select a visual target in a flight simulator seems to be an incentive for continuous flight, but has almost no effect on at least two of the force-controlling wingbeat parameters: the average wingbeat amplitude is negligibly diminished ( — 5·9 ± 0·4 %, 360 measurements) and the average wingbeat frequency is not significantly changed ( — 0·2±3-6%, 480 measurements) under these conditions. (2) In contrast to the investigation of the metabolic requirements during the first hours of flight (Table 1, lines 2-6), in the period between 10 and 30 h we measured the uptake of a 0·5 mol 1⁻¹ sucrose solution. The time scale eliminates the breaks for feeding and preening. We assumed that the equilibrium between consumption and regeneration of stored fuel reserves can be established within 10 h of almost continuous flight. As expected, the metabolic power input obtained under these conditions (line 1) is almost equivalent to the metabolic turnover derived from oxygen consumption during the earliest period of flight (line 3).

The variability of force production under conditions of tethered flight and the sex-, age-and satiation-dependent fluctuation of body weight among the different strains of Drosophila do not seem to explain all of the discrepancies in Table 1. Comparison of the data on metabolic power input (lines 2 and 3) suggests that food deprivation and flight to exhaustion prior to measurements of sugar uptake and flight reactivation in Wigglesworth’s experiments have left substantial reserves of chemical energy. The oxygen consumption in Laurie-Ahlberg’s experiment, a respiratory quotient of about 1 during flight, and the absence of significant oxygen debt after flight (Chadwick, 1947), point to an apparent supply of two-thirds of the initial power requirements by mobilization of stored fuel reserves. However, differences in the methods used may be sufficient to explain the different results. For example, in the present study we did not check the uptake of sugar during the first 10 h of flight.

The assumption of undiminished lift production during tethered flight, and a low power input derived from the parsimonious carbohydrate uptake at the beginning of this flight (line 2), have nourished earlier speculation about an unexpectedly high efficiency of the flight system in Drosophila melanogaster. The present data suggest a mechanochemical efficiency of about 0·04—0·07, which does not require unusual savings in the production and elastic storage of mechanical energy. However, some latitude for conclusions remains until the metabolic power requirements of a freely hovering fruitfly can be directly determined (C. P. Ellington & K. E. Machin, in preparation), and until the corresponding wingbeat parameters can be compared with those available for tethered flight (Götz, Hengstenberg & Biesinger, 1979).

The aerodynamic interference between the two wings at the end of the upstroke can be used to establish lift-inducing circulation around the wing profile before the downstroke begins. The exploitation of wing interference by a ‘clap and fling’ mechanism was discovered in Encarsia (Weis-Fogh, 1973): the wings of this wasp clap together, remain together for a while, and then fling open with their trailing edges still in contact. The same has since been observed in a number of insects including Drosophila melanogaster (Ellington, 1980, p. 70). Once ‘clap and fling’ is established in free flight, the method shown in Fig. IB can be used to resolve, under conditions of tethered flight, some details of this mechanism.

Fig. 4 illustrates two modifications of ‘clap and fling’ in the fruitfly which occur under conditions of both free flight (Ellington, 1984a, pp. 72, 100, 1984b; C. P. Ellington & K. E. Machin, in preparation), and tethered flight (Götz, 1983a, fig. 1. 2—3). The modifications relate to mutual wing contact in the ‘wings-up’ phase of the beat cycle where the longitudinal wing axes are held in a position parallel to the ventrodorsal body axis of the fly. Ellington thoroughly discussed the aerodynamic consequences of these modifications. We conjectured that during ‘squeeze’, and possibly during ‘peel’, a backward momentum is imparted to the air between the wings which is likely to contribute to the thrust of a fly cruising with almost horizontal body posture, and to the lift of a fly hovering with almost vertical body posture (Götz & Biesinger, 1983).

The series of stereophotographs on the left of Fig. 5 shows the modified ‘clap and fling’ under counterlight illumination. The investigation of the effect of airspeed on ‘clap and fling’ in the wind tunnel requires sidelight illumination. The stereophotographs in the middle and on the right of Fig. 5 relate to airspeeds of 0 m s⁻¹ and 1ms¹, respectively. Two effects of increased airspeed have been established in a number of experiments with three flies. (1) The upstroke requires about one-half rather than one-third of the wingbeat cycle (Nachtigall, 1979; Miyan & Ewing, 1985), and (2) the ‘clap and fling’ mechanism degenerates into a ‘near clap and fling’ (Ellington, 1984a, pp. 71, 108) lacking mutual wing contact. Neither of these effects can easily be explained by direct action of airflow on the beating wings: it is the fly which seems to adjust the kinematics of wing interference to the requirements of flight. The control of the ‘clap and fling’ mechanism in Drosophila is explicitly seen in experiments where the flies tried to follow horizontal displacements of an artificial visual environment. The abdominal deflection towards the inner side of the intended curve (Götz et al. 1979; Zanker, 1986) is accompanied by a similar deflection of the bisector of the wing cords during the ‘squeeze’. The corresponding deflection of thrust is likely to contribute to the course-control response of the fly (K. G. Götz, in preparation).

The lift-enhancing effect of the vigorous ‘clap and fling’ in Drosophila seems to be essential for the support of the body weight of the hovering fly. Cruising at non-zero airspeed facilitates the induction of circulatory lift. The observed decrease in wing interference by transition to a ‘near clap and fling’ is tentatively ascribed to a compensating influence of the altitude-control system. However, the tethered fly in the present experiments lacks much of the sensory feedback which is normally used to control the lift (Götz, 1968, 1983a; Götz & Biesinger, 1983; Götz & Wandel, 1984; David, 1985). It is thus conceivable that the increase in lift production by wing interference is not sustained over extended periods of flight. The tethered fly could eventually omit the ‘clap and fling’. This would allow us to derive the power requirements for the forced acceleration of the air between the interacting wings in Drosophila from the conjectural savings in metabolic energy. To test the possibility, the wing interference was observed at irregular intervals of the experiment in Fig. 2. Contrary to expectation, there was no significant decline of the ‘clap and fling’ within 32 h of tethered flight. The marvellous wings of the fly survived about 23 million ‘peels’ at a rate of about 200 s⁻¹ without noticeable damage.

It is a pleasure to thank Mr R. Biesinger who helped to watch a fly for 39 h, to supervise the measurements and to evaluate the data. Mrs U. Winz and Mr R. Zorn assisted efficiently with the preparation of the manuscript. Valuable suggestions came from Drs A. Borst, C. P. Ellington and R. Hengstenberg and Mr J. Zanker.