Timing of elevator muscle activity during climbing in free locust flight

The generation of rhythmic motor patterns underlies locomotor behaviour patterns such as walking, swimming and flight. In insects such as locusts, the basic alternating patterns for wing elevation and depression during flight are generated by a neuronal oscillator (e.g. Wilson and Weis-Fogh, 1962; Delcomyn, 1980; Robertson and Pearson, 1985; Stevenson and Kutsch, 1987; Pearson and Wolf, 1987; Wolf and Pearson, 1987) under the influence of afferent information from different sets of proprioceptive sensory organs that can modulate or control the rhythmic output (e.g. Wendler, 1974; Gewecke, 1975; Pearson and Wolf, 1987; Reye and Pearson, 1988; Büschges and Pearson, 1991; Ramirez and Pearson, 1993). This control of the motor pattern is important for flight performance, affecting, for example, lateral stability during flight (Möhl, 1985) and flight steering behaviour (Zarnack and Möhl, 1977; Baker, 1979a, b; Thüring, 1986; Schmidt and Zarnack, 1987).

Sensory input, in general, can influence the timing of both the depressor and elevator muscles during the wingbeat cycle (e.g. Neumann et al., 1983; Möhl, 1985; Horsmann and Wendler, 1985). Furthermore, it has been shown that proprioceptive feedback (in particular from the tegula organs and the wing hinge stretch receptors) plays an important role in timing the onset of elevator activity during flight pattern generation (Wolf and Pearson, 1988; Pearson and Wolf, 1989; Wolf, 1993). As demonstrated in experiments involving tethered flying locusts or flight preparations, the timing of elevator activity during the wingbeat cycle is of significance during flight not only for the generation of a functional motor pattern (Wolf, 1993) but also during particular flight behaviour patterns (indicated by responses related to roll stimuli or by optomotor reactions associated with flight steering, e.g. Baker, 1979a,b; Thüring, 1986).

Although considerable information is available about motor patterns (including elevator patterning) during steady ‘level’ flight, there is little information about how the locust flight system controls ascending and descending flight paths during free flight. Recently, it was shown that in particular flight muscles the number of motor units recruited during the wingbeat cycle depends on the ascent angle (Kutsch et al., 1999). The first evidence that the timing of elevator activity might play a part in regulating the ascent angle during flight was provided by investigations involving freely flying locusts deprived of afferent information from the tegula (Fischer and Ebert, 1999): although the capability for free flight was maintained even when elevator timing was influenced, the production of the additional effective lifting power required for ascending flight was affected.

To understand more about the functional and behavioural relevance of the modulation of elevator activity during free locust flight, we focus in the present study on the patterns of antagonistic activity in several depressor and elevator muscles of the fore- and hindwing, on the corresponding wingstroke variables and on the flight behaviour pattern with respect to the ascent angle. The antagonistic motor patterns during free flight were investigated by telemetric electromyography (e.g. Fischer and Ebert, 1999), and the related behavioural patterns were recorded synchronously on video. Wingstroke variables during free flight were documented by high-speed video.

All experiments were carried out using adult female Schistocerca gregaria Forskål (15–21 days postmoult). The animals were bred at 42 °C and 70 % relative humidity under a 12 h:12 h light:dark cycle in crowded colonies of both sexes. Individual flight sequences were investigated in a room measuring 10 m×5.7 m×2.2 m at 34 °C. The average illumination of the room was 4500 lx. All animals were warmed to approximately 40 °C before the flight experiments. To initiate flight, each animal was placed on a platform, and the majority of flights started with the animal jumping (either spontaneously or in response to hand-clapping or a short wind puff to the cerci). When jumping could not be elicited, flight was initiated by launching the animals vertically into the air (speed at release 2–3 m s⁻¹ ). To avoid any residual effects of the jump and the launch during the observations, the platform and the point of launch were approximately 5 m in front of the observation area. Irrespective of the method of initiating flight (i.e. jump or vertical launching), individual flight sequences ranged from a minimum of 5 s to a maximum of 120 s. Following a launch, the majority of animals performed 4–6 consecutive flight sequences (initiated by spontaneous jump starts from the ground, the walls and the ceiling; the total flight time per animal was, on average, approximately 90 s). Most of the animals crossed the observation area several times when flight activity lasted longer than a few seconds.

The transmitter device (mass 0.3 g, representing 8–12 % of locust body mass) for telemetric acquisition of electromyographic (EMG) data is shown in Fig. 1A. For a description of the circuitry, the attachment of the device to the animal and the receiver system used, see Fischer (1998). One-channel transmission was used to record simultaneously from two ipsilateral muscle antagonists (nomenclature after Snodgrass, 1929). Using three different experimental designs, the forewing muscle activity (1) of the elevator 83 and the subalar depressor 99, (2) of the elevator 90 and 99, and (3) of 83 and the first basalar depressor 97 was investigated. In the hindwing, (1) the elevator 113 and the subalar depressor 129, (2) the elevator 119 and 129, and (3) 113 and the first basalar depressor 127 were examined. Telemetric EMG traces are shown in Fig. 1B. Instantaneous phase values (ϕ) of the specific antagonists were calculated as ϕ_D/E=(D−E)/(D−D) where D−E is the depressor–elevator interval (in ms) and D−D is the cycle length of the particular depressor (in ms).

The free-flight sequences of individual locusts were recorded using a stationary video camera (Panasonic NV-MS4EG; frame frequency 50 Hz, shuttering time 1 ms). The telemetric EMGs were displayed on an oscilloscope (Tektronix D11) and stored on tape (Racal Store 4DS). The oscilloscope screen was filmed by a second video camera (Revue Video 8, frame frequency 50 Hz, frame interval 20 ms). The video outputs of the two cameras were mixed online (Panasonic digital mixer WJ-AVE5) and recorded on a video master tape.

Wingstroke variables of free-flying locusts (without the transmitter) were determined using a different approach with a digital high-speed video system (maximum frame frequency 500 Hz; shuttering time 500 ms; Weinberger Systems, Switzerland).

The master tape recordings and the high-speed video recordings were both screened on a 27 inch Sony Trinitron colour video monitor and transferred to overhead transparencies for analysis (for methods, see Baker and Cooter, 1979a,b). The free-flight variables (see below) were acquired by frame-by-frame analysis (Panasonic digital video recorder AG7355) of the continuous displacement of the specific body and wing silhouettes. For a detailed evaluation of the measurement error with respect to silhouette length and flight variables, see van der Wall (1996) and Fischer (1998). Motor patterns and wingstroke variables of approximately 15 consecutive wingbeat cycles could be acquired from a flight sequence.

Wingbeat frequency f_WB was determined from the time between two consecutive upper reversal points of the wing. Body length and wing length were measured before the experiments began. All variables were considered as ‘instantaneous’, being constant within the given frame interval.

Data analysis and statistical procedures were computer-aided (software KaleidaGraph, StatView and Origin) and follow the criteria described by Sachs (1978). Correlation and linear regression were performed by the method of least square differences at a significance level of P<0.05, with r indicating the coefficient of linear correlation. Two-sample comparisons of the coefficients of regression (i.e. slopes of the regression line) and the intercepts were performed using a t-test (Sachs, 1978) at a significance level of P<0.05. Mean phase values were calculated using circular statistics after the criteria described in Batschelet (1981) and are given as P ± mean angular deviation, with r describing the mean vector.

The relationships between body angle φ (degrees), ascent angle α (degrees) and flight speed v (m s⁻¹ ) were determined during free flight at a temperature of 34 °C. Eight females (aged 18 days postmoult) were investigated. In addition, to assess how the attachment of a 0.3 g transmitter influenced flight variables, each female performed flights with and without a transmitter: four individuals were equipped with a transmitter after control flights; for the remaining four animals, control variables were recorded after flight performance under loaded conditions. In general, each individual was limited to 20 flight sequences (10 controls and 10 loaded) including those flights in which the animal did not cross the camera focus. The data were evaluated on a per locust basis; for a description of the behaviour of the ‘average’ locust, the data from all animals were pooled. The results are shown in Fig. 2.

In each experimental animal, the body angle φ was significantly correlated with the ascent angle α for control flights and for flights with the transmitter attached (see insets of Fig. 2A,B, P<0.05 in all individual regressions).The slopes of the two ‘average’ regressions did not differ significantly between the different conditions: φ increased by 1.087 degree degree⁻¹ of α when climbing in control flights (r=0.965, N=37, P<0.05; Fig. 2A) and by 0.95 degree degree⁻¹ of α under transmitter load (r=0.948, N=36, P<0.05; Fig. 2B; for equations, see Table 1). The range of instantaneous ascent angles observed during the flights under both conditions was comparable (loaded: −3 to 36 °; controls: −10 to 40 °). However, a significant difference (P<0.05) was found in the intercepts, indicating a systematic decrease in the body angle (Δφ, Fig. 2B) when a transmitter was attached. This angle, deduced from the regressions the body angle, e.g. during level flight (i.e. α=0 °), was φ=12.5 ° in control animals, whereas the same animals exhibited a lower body angle (φ=5.2 °) during level flight under loaded conditions.

In seven females, flight speed was not significantly correlated with the ascent angle α during control flight (P>0.05), but in one animal, the flight speed decreased significantly (P<0.05) during climbing (Fig. 2C). For flight under loaded conditions, flight speed was not correlated with α in five females (P>0.05, Fig. 2D), while two individuals exhibited a significant increase in speed when climbing, and one animal exhibited a significant decrease in flight speed during climbing (P<0.05). The mean flight speed for the population under control conditions was 4.48±0.49 m s⁻¹ (mean ± S.D., N=37); when flying with the transmitter, the mean flight speed v was slightly but significantly lower (v=4.28±0.28 m s⁻¹, N=36, P<0.05). However, only in two of the eight individuals was a strong decrease in flight speed observed when loaded with a transmitter (animals 3 and 6, P<0.05; see Fig. 2C,D).

These findings indicate that the attachment of a 0.3 g transmitter systematically influenced the values of particular flight variables such as the body angle; however, the findings do not provide sufficient evidence to support the assumption that the functional relationships between the flight variables investigated here were changed substantially.

In 12 females (18 days postmoult, without a transmitter), the instantaneous wingbeat frequencies f_WB were determined by high-speed video and were related to the ascent angle. Each female performed 20 flights (including flights that were out of the camera focus). The results are shown in Fig. 3. In nine of the 12 locusts (i.e. in 75 % of the animals investigated), no significant correlation between f_WB and ascent angle was observed (P>0.05). A significant increase in f_WB during climbing was found in the remaining three animals (animals 5, 8 and 10 in Fig. 3, P<0.05). The mean value of f_WB for all animals investigated at 34 °C was 24.4±1.07 Hz (N=44, mean ± S.D.).

In eight young adult females, the stroke amplitude Ψ and the phase of wing elevation within the downstroke cycle (ϕ_el) of the hindwing and the forewing were determined and plotted in relation to the ascent angle during climbing flight (Fig. 4). The animals used were younger (11–13 days postmoult) than the animals investigated in the previous sections; their mean wingbeat frequency f_WB was 23.2±1.12 Hz (N=28, mean ± S.D.). The use of younger animals enhanced the numbers of frames per wingbeat cycle obtained using high-speed video recording (an average of 22 frames per cycle compared with an average of 20 frames per cycle in adult mature females when recording at 500 frames s⁻¹ ). Individuals were evaluated only when data from both the hindwing and the forewing could be acquired.

In four of the eight females, the stroke amplitude of the forewing Ψ_fw increased significantly during climbing (P<0.05, Fig. 4A). A significant increase in the stroke amplitude of the hindwing Ψ_hw was found in five of the eight animals (P<0.05, Fig. 4B). However, in only two animals (animals 3 and 5, compare Fig. 4A,B) could an increase in both Ψ_fw and Ψ_hw be observed.

In four of the eight individuals, the phase of forewing elevation ϕ_el increased significantly in relation to the ascent angle (P<0.05, Fig. 4C), and a significant increase in ϕ_el of the hindwing was found in four animals (Fig. 4D). An increase in ϕ_el in both the forewing and hindwing was observed in three of the eight individuals (animals 1, 2 and 3, see Fig. 4C,D).

The relationships between the wingstroke variables for all eight females are summarised in Table 2. The data for all individuals were pooled to obtain a mean slope and intercept for each wing; the regressions (P<0.05; for equations, see Table 1) indicate (a) that the increase in stroke amplitude with increasing ascent angle α was more marked in the forewing than in the hindwing (the slopes are significantly different, P<0.05, N=32, but not the intercepts), and (b) that there was no difference in ϕ_el between the two wings (the slopes and intercepts are not significantly different in the hindwing and forewing, P>0.05).

The mean phase relationship between the hindwing and forewing (i.e. the interval between downstroke onset in the two wings) was 0.16±0.03 (mean ± angular deviation; r=0.956, N=32), with the hindwing leading. This phase relationship did not change significantly with increasing ascent angle α (P>0.05, data not shown).

To measure changes in the motor pattern related to the ascent angle α during free climbing flight, telemetric recordings were made from several pairs of muscle antagonists in both wing segments. In different experimental sets, each pair of antagonists was investigated in 10 adult mature females (18–20 days postmoult) at a temperature of 34 °C. Each individual performed 10 flight sequences (including those not in the camera focus), and a further 10 flights were conducted after a 1 h break followed by warming-up.

For one individual, the antagonistic patterns of the first basalar depressor (127) and the tergosternal elevator (113, Fig. 5A) are shown in relation to the corresponding flight variables (Fig. 5B,C). During flight, this particular animal increased its ascent angle α continuously. The phase difference ϕ_127/113 between the recorded activity of the depressor (127) and the elevator (113) increased with increasing values of α. This indicates that, in this flight sequence, the onset of elevator (113) activity was delayed with respect to the onset of depressor (127) activity when the animal performed climbing flight.

For both wings, the phase relationships between the antagonistic muscle pairs in relation to the ascent angle α are shown in Fig. 6. The relationships obtained for each of the 10 animals tested in each experimental set are shown by insets in Fig. 6. To monitor the motor behaviour of the ‘average’ locust, the data for the 10 individuals tested were pooled to obtain a mean slope and intercept of the regression for each pair of antagonists (main parts in Fig. 6A,B; for equations, see Table 1).

For the antagonists of the hindwing, the phase values in relation to the ascent angle α are shown in Fig. 6A: the phase values of the hindwing elevators 119 (Fig. 6Ai) and 113 (Fig. 6Aii) increased significantly with increasing α (P<0.05). This indicates that both elevator muscles were delayed in their onset with respect to the subalar 129 (depressor) cycle when the animals performed ascending flight. The phase relationship ϕ_129/119 was in general lower than ϕ_129/113 (the intercepts were significantly different, P<0.05): this means that muscle 119 is activated earlier during the wingbeat cycle than the synergist 113.

Comparable relationships were found for the forewing elevators 90 (Fig. 6Bi) and 83 (Fig. 6Bii) with respect to the subalar (99) depressor: both elevator synergists showed a significant delay in activation when the ascent angle was increased (P<0.05). However, the shift in ϕ_99/90 was weak compared with that in ϕ_99/83 (slopes significantly different, P<0.05, Table 1). The intercepts were significantly different (P<0.05). This indicates that elevator 83 was activated later during the wingbeat cycle than elevator 90.

In addition, phase values of wing elevators (83, 113) were determined with respect to the first basalar depressors of each wing segment (97, 127). The results are shown in Fig. 6Aiii and Biii. In the hindwing, ϕ_127/113 increased significantly (P<0.01) with increasing α (Fig. 6Biii). The delay in 113 activation in relation to 127 activation was greater than with respect to 129 (slopes significantly different between ϕ_127/113 and ϕ_129/113, P<0.05). This further indicates that the delay between the activation of the synergists 129 and 127 decreased with increasing ascent angle; both muscles are expected to be activated nearly synchronously at an ascent angle of approximately 40 ° (Fig. 6Aiii).

Considering the homonomous muscles in the forewing (elevator 83 and basalar 97), comparable results were obtained (Fig. 6Aiii,Biii): activity in 83 was delayed with respect to activity in 97 when the animals increased their ascent angle α. However, ϕ_97/83 values were, in general, below those found for ϕ_99/83 (intercepts significantly different, P<0.0.5). This indicates that, with respect to 83, the first basalar (97) is activated later than the subalar (99) muscle within the wingbeat cycle. A scheme for the activation of the antagonists examined in these experiments with respect to climbing flight was deduced from Table 1 and is shown in Fig. 7.

In the present study, the onset of elevator muscle activity during the wingbeat cycle was investigated telemetrically with respect to the ascent angle during free locust flight. In the majority of individuals tested, the wing elevator muscles examined were delayed in their activation with respect to the synchronously recorded wing depressor muscle. Corresponding to the alterations in motor pattern, an increase in the phase of wing elevation (i.e. delayed wing elevation) and/or an increase in the stroke amplitude was observed. These findings suggest that the relative timing of elevator activity during flight not only contributes to a functional motor pattern (e.g. Wolf, 1993) but also plays a functional role in regulating the ascent angle during climbing flight.

The application of unsteady aerodynamic calculations to the flapping flight of insects such as locusts has enabled a basic understanding of the relationships between wing kinematic variables and the aerodynamic forces produced by the wing movement (Send, 1992, 1994; Zarnack and Send, 1994; Zarnack, 1997). In locust flight, the main implications of this theory are as follows. (1) The functional task of the complex wing movements, i.e. alternating up and down movement, cyclic changes in the wing angle of attack, and wing pro- and remotion along the longitudinal body axis (‘lagging’), is the generation of thrust not lift. (2) The generation of lift during the wing beat relies on the steady wing angle of attack and on the speed of the air stream faced by the wing profile, i.e. the flight speed. However, (3) wingstroke amplitude and wingbeat frequency have only weaker influences on lift production. Applying the theory of unsteady aerodynamics, some functional implications can be deduced for a transition from level to climbing flight: in general, climbing requires additional lift production, i.e. the output of additional effective lifting power (e.g. Fischer and Ebert, 1999). Since lift generation depends mainly on the steady wing angle of attack during flight, additional lift can be generated by increasing the air speed faced by the wing profile. Theoretically, this can be achieved by increasing flight speed.

Our findings during free flight show that no significant correlation between f_WB and the ascent angle was present in the majority of animals (Fig. 3; see also Baker et al., 1981). This means that the generation of additional lift during climbing is, apparently, not achieved by increasing the wingbeat frequency (see topic 3 above). Approximately half the animals tested during free flight increased the stroke amplitude of the hindwing or the forewing during climbing (Fig. 4). In the majority of the animals investigated, flight speed was not correlated with the ascent angle (Fig. 2).

Since increasing the stroke amplitude causes a small but significant increase in lift during tethered flight (Wortmann and Zarnack, 1993), we conclude that the additional lift required for climbing flight is generated, at least partly, by an increase in the stroke amplitude of a particular wing.

As shown in Table 2, a delay in wing elevation with respect to the ascent angle was also observed in animals that did not synchronously increase the stroke amplitude of that particular wing (although stroke amplitude and phase of elevation in at least one of the wings were positively correlated in three of eight animals investigated). The most likely explanation is that stroke amplitude and the phase of wing elevation during the wingbeat cycle may have different tasks and, thus, may rely on different control mechanisms during flight. It is unclear whether elongating the duration of the downstroke by shifting the phase of wing elevation results in an increase in flight speed.

Body angle and ascent angle are strongly correlated, suggesting a significant role for the body angle in the production of additional lift (Fig. 2; see also Baker et al., 1981). This is confirmed by reports showing that adjusting the locust body angle reproducibly manipulates lift generation (Wortmann and Zarnack, 1993). However, it is not clear which variables control the body angle during free flight.

In our study, the phase value between the two wings did not change with changing ascent angle. As reported previously, relative lift production depends on an appropriate delay between the two flapping wing pairs (maximum lift is generated at Δt=8 ms in tethered flying locusts; Wortmann and Zarnack, 1993). This indicates that the additional lift production required for climbing flight is not, apparently, generated by a phase shift between the hind- and the forewing, since changes in this phase value are expected to reduce lift production in general.

Appropriate timing of the elevator muscles during the wingbeat cycle seems to be an important functional task during flight pattern generation. This becomes evident from the fact that, in particular, elevator activation can be influenced by proprioceptive input (e.g. by signals from tegula organs and wing hinge stretch receptors; Wolf and Pearson, 1988; Pearson and Wolf, 1989) and that the timing of elevator activity onset is affected after sensory deafferentation (Büschges and Pearson, 1991; Büschges et al., 1992; Wolf, 1993). The timing of elevator activity determines the stroke amplitude of the wing (e.g. Hedwig and Becher, 1998). Our suggestion that the timing of elevator activity plays a functional role in regulating the ascent angle during climbing flight is further supported by recent findings during free locust flight. Although the general capability for free flight is maintained, ablation of the tegula organs not only affects the elevator timing (Fischer and Ebert, 1999) but is also responsible for impaired flight behaviour (Gee and Robertson, 1998): the power output of the flight system was reduced, and the range of ascent angles observed after tegula removal was significantly decreased (Fischer and Ebert, 1999).

The functional role of activating a particular wing muscle in the complex three-dimensional wing movement is difficult to evaluate. Together with flight muscles that were not recorded in our study, the relative activity of the antagonists investigated is reported not only to determine the downstroke interval and, hence, the phase of wing elevation and stroke amplitude (e.g. Hedwig and Becher, 1998), but also to control the wing angle of attack (Wilson and Weis-Fogh, 1962; Pfau, 1977, 1978, 1982) that is co-responsible for thrust production (for an overview, see Zarnack, 1997). The motor pattern recorded during free flight therefore represents a ‘superimposed image’ with respect to at least two functional tasks (i.e. climbing, see this paper) and the control of lateral stability (Möhl, 1985). Alterations in the pattern due to steering and corrective steering (e.g. Robert, 1988) cannot be excluded completely; however, our study focused on straight forward flights, and flights with any visible sign of directional changes within the horizontal plane were excluded from evaluation.

In this study, the activity onsets of the first (113, 83) and second (119, 90) tergosternal elevator muscles were examined with respect to the onsets of the subalar (129, 99) and first basalar (127, 97) depressor muscles in the meso- and metathoracic segments. Our findings indicate that ascent-dependent phase changes are different for the particular antagonistic muscles investigated (Figs 6, 7). In many previous studies, locust flight motor patterns have been monitored by recording from a single pair of antagonists (127/113, e.g. Büschges and Pearson, 1991; 97/83, e.g. Wolf, 1993; 129/119 and 90/99, e.g. Stevenson and Kutsch, 1987). When focusing on only one pair of antagonists, general conclusions, for example how strongly sensory input modulates motor pattern generation (Wolf and Pearson, 1987; Stevenson and Kutsch, 1987), might be misleading (i) because, during the wingbeat cycle, the timing of the muscles investigated differs, in general, from that of other hemisegmental synergists, and (ii) because the timing of a particular muscle may be influenced strongly by proprioceptive input, whereas the timing other synergists, although in the same wing segment, may be only weakly influenced (Möhl, 1993).

Given the present state of knowledge, we cannot conclude from our free-flight investigation that there is an altered ‘basic’ mechanism of flight pattern generation during tethered flight since (i) the basic relationships between the free-flight variables match the findings and their aerodynamic implications found in tethered studies (e.g. Zarnack and Wortmann, 1989; Wortmann and Zarnack, 1993; Zarnack, 1997) and (ii) the motor pattern is, in general, comparable with previous reports in tethered flight (e.g. Wolf and Pearson, 1987; Stevenson and Kutsch, 1987; Pearson and Wolf, 1989; Büschges and Pearson, 1991; Hedwig and Becher, 1998).

In contrast to the well-documented behaviour of migrating locust swarms with respect to environmental variables in the southern hemisphere (for a review, see Farrow, 1990), there is little field information about individual flyers (Waloff, 1972; Baker and Cooter, 1979a,b; Baker et al., 1981). The relationships between body angle, wingbeat frequency, flight speed and ascent angle match the findings of previous field observations. This is of particular interest (i) because the field observations were performed on L. migratoria (e.g. Baker et al., 1981) compared with our study of S. gregaria, and (ii) because the locusts observed in the field had become airborne at least 30 min before the data were acquired (Baker et al., 1981), whereas the majority of the animals investigated in our study had been flying for only a few seconds before data acquisition. This suggests that the relationships between free-flight variables investigated in both types of study are (i) comparable for both genders of both locust species and (ii) independent of the duration of flight (migrating ‘long-term’ flight or ‘short-time’ flights of approximately 30 s).

We are grateful to A. Büschges (Universität Köln) for a critical reading of the manuscript and to G. Wendler (Universität Köln) for valuable discussion. We wish to thank an anonymous referee for his constructive criticism and valuable suggestions. We thank especially C. Graef (Universität Köln) for her assistance. This project was supported by the DFG (Ku 240/17-1,2).