Control of optomotor responses by descending deviation detector neurones in intact flying locusts

Flying locusts detect unintended course deviations by means of exteroreceptors of the head (eyes, ocelli, wind-sensitive hairs), and counteract the deviations by optomotor reactions which include corrective steering movements by wings, legs and abdomen, and compensatory head movements (e.g. Möhl and Zamack, 1977; Zamack and Möhl, 1977; Taylor, 1981a,b; Thüring, 1986; Arbas, 1986; Baader, 1988; reviewed by Rowell, 1988; Möhl, 1989; Hensler, 1989). The sensory information about course deviations is coded and transmitted to the motor centres by a population of descending deviation detector neurones (DNs) which project from the brain to the fused abdominal ganglia. Approximately 15 pairs of DNs have been identified physiologically; six of them have also been identified morphologically. DNs are uni- or multimodal, and respond best to course deviations of different directions (Bacon and Tyrer, 1978; Möhl and Bacon, 1983; Rowell and Reichert, 1986; Griss and Rowell, 1986; Hensler, 1988a, 1989; Baader, 1989; Rowell, 1989); a subpopulation additionally codes for movements and/or position of the head (Hensler, 1988a; K. Hensler, in preparation). For three ocellar DNs, the major pathway to wing motoneurones (MNs) has been demonstrated (Reichert and Rowell, 1985, 1986). It involves a class of thoracic interneurones (TTNs) which receive direct input from DNs and synapse onto various subsets of wing motoneurones (MNs).

These data strongly suggest that DNs are involved in course control. However, the connectivity between DNs, TINs and wing MNs has been unravelled in dissected locusts, either at rest or during fictive flight (rhythmic activation of wing MNs in locusts with denervated thoracic ganglia, with a basic pattern similar to that in intact locusts). We know that proprioceptive feedback, as it occurs during normal behaviour, is important in generating the motor output (Wolf and Pearson, 1988), and that reflexes operating in quiescent animals may be changed, or inverted, during active movements (e.g. Skorupski and Sillar, 1986; Bassler, 1988). Hence, the behavioural relevance and functioning during normal behaviour of the pathway from DNs to wing MNs is not proven, and also there is no indication of the importance of single DNs within the entire population. Both questions can only be solved by individual stimulation of DNs in intact flying locusts.

Stimulation during flight behaviour has been achieved for the wind-sensitive TCG (tritocerebral commissure giant) neurone, which indeed elicits steering by wing muscles when activated electrically (Möhl and Bacon, 1983). However, TCG is anatomically a special case: it is easily stimulated individually with an extracellular electrode mounted at the tritocerebral commissure, whereas individual stimulation of all other DNs requires penetration with intracellular electrodes. In a first attempt we tried to reveal effects of single DNs on wing MNs during fictive flight in the type of highly dissected preparations normally used for intracellular recordings from thoracic neurones (Robertson and Pearson, 1982; Rowell and Reichert, 1986; Hensler, 1988a). Neither the stimulation of single DNs nor the correlation of DN activity with muscle activity provided convincing results. The main reason is that steering is expressed as shifts in the relative timing of action potentials in the different wing muscles (Möhl and Zamack, 1977; Zamack and Möhl, 1977; Taylor, 19816; Thüring, 1986; Schmidt and Zamack, 1987; Hensler and Robert, 1990). Generally, these shifts range from 0.5 to 5ms, but during fictive flight the standard deviation of normal fluctuations usually exceeds 5 ms, and may thus obscure the small effect to be expected by stimulating the DN. Another problem is that sequences of fictive flight normally last for only a couple of seconds so that on- and off-effects may mask the influence of the stimulation.

These problems have been overcome in this study with a new preparation requiring only minimal dissection: locusts with intact thoracic box and neck joint are able to move their head and flap their wings in a normal fashion during tethered flight while DNs are recorded or stimulated intracellularly in the neck connectives. Flight sequences are much longer under these conditions, and fluctuations in the relative latency are less pronounced. We show that electrical stimulation of the identified DNs PI(2)5 (Hensler, 1988a) and DNC (Rowell and Reichert, 1986) elicits both considerable steering by wing muscles and compensatory head movements. Both resemble the optomotor behaviour elicited by horizon movements which excite the neurones in question.

Experiments were performed at 24–30 °C on adult Locusta migratoria of either sex, from a laboratory culture kept on a light: dark cycle of 14h:8h.

The legs of a locust were removed and a small piece of balsawood (6 mm×5 mm×1mm) was waxed onto the pterothoracic sternum, its edges touching the ventral insertions of M97 and M127 on either side (first basalar muscles of fore- and hindwings, respectively=direct wing depressors; the insertions are recognized as hairless patches). Electromyograms were recorded from M97 and M127 on both sides, using steel pins, about 2 mm long, inserted through the edges of the balsa block into the muscles for 100–200 μm. The animal was waxed at the lateral walls of the pronotum into the U-shaped part of a holder (Fig. 1A). Pro- and pterothorax were fixed to each other with lateral bridges of beeswax. The posterior part of the pronotum which projects over the mesothoracic wing hinges was removed, and a 3 mm×3 mm window was cut in the remaining posteriodorsal part. The anterior third of the pronotum remained untouched to leave the neck joint intact. The neck connectives were exposed by removing overlying tissue, including the anterior portion of the gut (the remaining posterior portion was ligatured), connective tissue, fat body and both neck muscles M61 which lie directly over the connectives. Sometimes the dorsal longitudinal neck muscles M51 were damaged by the dissection, but the other 13 pairs of neck muscles remained intact, and no obvious change in head movements was observed (for anatomy and nomenclature of neck muscles see Shepheard, 1974; Honegger et al. 1984).

The stimulus device used for the characterization of sensory inputs to DNs is described elsewhere in more detail (Hensler, 1988a). Here we give only a brief account. The animals were positioned in front of an artificial semi-horizon, a hemisphere with a diameter of 70 mm (Rowell and Reichert, 1986). Its lower half was painted black. The translucent upper half was painted with black dots, and illuminated by a small tungsten bulb from the rear. Visual input from behind the animal was prevented by covering the rear halves of the compound eyes with black enamel. Rotational course deviations were mimicked by horizon movements around the roll, yaw and pitch axes by means of servo-motors. Wind of up to 4 ms⁻¹ could be blown onto the locust’s head through a tube (diameter 12 mm) in the middle of the horizon.

Active head rolling (rotation around the longitudinal body axis) was measured using a capacitance device (Sandeman, 1968). This measured, between two lateral sensors, the position of a steel pin glued vertically to the head (Fig. 1A). Passive head rolling was imposed under servo control using the pin as a lever. The servo was mounted on a manipulator, and a fork connected to its axis could be engaged with and disengaged from the lever while an experiment was in progress.

The neck connectives were manipulated with a glass hook onto a rectangular platform (1.5mm×2.5mm) made of copper and galvanically coated with gold (Fig. IB). A row of three pegs, about 300 μm in height, protruded vertically from both the anterior and the posterior edges. The distance between neighbouring pegs in a row was 300-400gm. The connectives were placed between the lateral pegs, with the median pegs between them, and were subsequently fixed with tiny droplets of tissue glue (histoacryl blue, Braun, Melsungen, FRG). This allowed stable intracellular recordings during active and passive head roiling of up to 30°, and while normal flight activity was performed at the usual frequency of 15–20 Hz.

Neurones were penetrated with glass microelectrodes containing 5 % Lucifer head rolling was imposed under servo control using the pin as a lever. The servo was mounted on a manipulator, and a fork connected to its axis could be engaged with and disengaged from the lever while an experiment was in progress.

The neck connectives were manipulated with a glass hook onto a rectangular platform (1.5 mm×2.5 mm) made of copper and galvanically coated with gold (Fig. IB). A row of three pegs, about 300 μm in height, protruded vertically from both the anterior and the posterior edges. The distance between neighbouring pegs in a row was 300–400 μm. The connectives were placed between the lateral pegs, with the median pegs between them, and were subsequently fixed with tiny droplets of tissue glue (histoacryl blue, Braun, Melsungen, FRG). This allowed stable intracellular recordings during active and passive head rolling of up to 30°, and while normal flight activity was performed at the usual frequency of 15–20 Hz.

Neurones were penetrated with glass microelectrodes containing 5 % Lucifer Yellow (Stewart, 1978) in the tip and 0.5 mol l⁻¹ lithium acetate in the shaft (resistance: 30–80 MΩ). PI(2)5 and DNC were identified physiologically using the criteria described by Hensler (1988a) and Rowell and Reichert (1986), respectively. The appropriate tests included horizon movements, passive head movements, ocellar stimulation and frontal wind.

DNs were penetrated at least 6–7 mm from their spike initiation zones which appear to be in the brain (Rowell and Reichert, 1986; Hensler, 1988a). Consequently, stimulation was not easily achieved by the injection of depolarizing current because the high electrode resistance in combination with the short length constants of DNs allowed excitation of neither the spike initiation zone nor of the axon itself. Stimulation was possible only by taking advantage of a post-inhibitory rebound effect: the axon was hyperpolarized with a current of 10–15 nA for 10–30 min. A burst of action potentials was elicited when the hyperpolarization was released. The frequency within the burst could reach more than 200 Hz when switching directly to depolarizing current. The doses of current depended on the quality of the penetration and on individual properties of the animal. Care had to be taken as this procedure easily led to serious damage of the axon if the hyperpolarization was too strong and/or the stimulation was repeated too often.

All stimulation experiments were performed in complete darkness to prevent interference from visual feedback caused by the resulting head movement.

Tethered, flying locusts reacted to horizon rolling with compensatory head rolling, and with shifts of relative latency (for a definition see Fig. 2) in the first basalar muscles M97 and M127 (Fig. 2; see also Taylor, 1981a,b; Thüring, 1986; Schmidt and Zarnack, 1987; Hensler and Robert, 1990; for changed timing in other wing muscles, see Möhl and Zarnack, 1977; Zarnack and Möhl, 1977). Shifts of latency in M127 were in the opposite direction to those in M97, and generally scatter was more pronounced.

Both PI(2)5 and DNC (Fig. 3A) responded with bursts of action potentials to horizon rolling in the preferred direction (Fig. 3B, C). For PI(2)5 this was horizon rolling to the side opposite to the axon (thus simulating course deviation to the same side as the axon) (Fig. 3B). The preferred direction of DNC is horizon rolling to the same side as the axon (thus simulating course deviation to the side opposite to the axon) (Fig. 3C). The response properties of both neurones are described in detail elsewhere [PI(2)5, Hensler, 1988a; DNC, Rowell and Reichert, 1986].

Electrically induced bursts of action potentials (200±30Hz for 200–300 ms) in either PI(2)5 or DNC had no effects on wing muscles in non-flying locusts. During flight activity, however, considerable shifts of relative latency were elicited in both M97 and M127 (Fig. 4). The shifts were of similar sign and intensity to those caused by horizon rolling of about 40° (see Fig. 2). The maximum shift was not reached immediately but developed within the first 100 ms after onset of the stimulation, i.e. within two wingbeat cycles (Fig. 5). On average, 50% of the effect was reached within the first 30ms, i.e. well within one wingbeat cycle. Scatter of the individual values prevented exact determination of the minimum latency, but the data suggest that it is in the range of 10–20 ms. Spiking activity in DNs ceased immediately when they were again hyperpolarized, but after-effects were observed in M97 and M127 which lasted for 3–5 wingbeats (Fig. 4).

This shows that PI(2)5 and DNC are both involved in the control of corrective steering via short-latency pathways, and that single DNs are capable of producing considerable steering effects. However, single DNs are not designed to control steering alone, as the system is not suited to transmit for long the high-frequency bursts evoked by electrical stimulation. These evoked bursts consist of 40–60 action potentials at 200±30Hz, which are unphysiologically high rates compared with bursts evoked by horizon rolling: these contain 10-50 action potentials at less than 50 Hz (Fig. 3B, C). The motor effect of electrical stimulation declined dramatically, due to fatigue, when bursts of more than 300 ms were repeated more often than once per second. A moderate decline of the response was seen for bursts exceeding 200ms (Fig. 5, point on extreme right).

Stimulation of PI(2)5 in quiescent locusts caused head rolling away from the axon of the stimulated neurone (Fig. 6). The head movement also contained yaw and pitch components but these were weak compared with roll. The latency between stimulus and head movement was about 30 ms. The delay between activity in neck MNs and detectable head movements is known to range between 10 and 20 ms (K. Hensler, unpublished results; A. Baader, personal communication); hence, DN excitation elicited the first action potential in neck motoneurones (MNs) after 10–20 ms. PI(2)5 frequencies of about 200 Hz normally led to head rolling of 5–10°, a similar range to compensatory head movements elicited by horizon rolling of 30–40°.

The direction of evoked head rolling coincided with the direction of compensatory head rolling caused by horizon movements in the preferred direction of the stimulated PI(2)5 (Figs 2, 3). For example, horizon rolling to the right, excited the right PI(2)5 and elicited compensatory head rolling to the right. Electrical stimulation of the same PI(2)5 also caused head rolling to the right. Analogous results have been observed for DNC (K. Hensler, unpublished results; A. Baader, personal communication), i.e. stimulation of the left DNC, which responds to horizon rolling to the right, was seen to elicit head rolling to the right. Thus, PI(2)5 and DNC are both involved in the control of compensatory head rolling via a pathway of short latency.

Recognizable head movements were only elicited in quiescent locusts. Active head movements performed by the locust during flight or other behaviour suppressed the electrically evoked head movements (Fig. 6B). Even in averaged responses (up to five single events) no head movement was detected under these circumstances.

We show for the first time in intact flying locusts that the identified descending deviation detector neurones (DNs) PI(2)5 and DNC are capable of eliciting motor activity which closely resembles the optomotor responses following simulated course deviations around the roll axis (Figs 2, 4). This strongly indicates that PI(2)5 and DNC are directly involved in corrective course control. Similar motor effects have been observed for other visual DNs which have not been identified morphologically (K. Hensler, in preparation). In the same context Baader (1989; and reviewed by Rowell, 1989) describes DNs which evoke head rolling and rudder-like movements of abdomen and hindlegs. Abdominal movements were never recorded when stimulating PI(2)5, although the projections of PI(2)5 into the fused abdominal ganglion (Fig. 3A) might provide subthreshold input to the abdominal motor system. Baader (1989) also observed antennal movements upon stimulation of some DNs. Although it is not clear whether PI(2)5 and DNC provide input to the antennal motor sytem, both send blebbed branches into the deutocerebra! neuropile containing the arborizations of antennal MNs (Fig. 3A; Hensler, 1988a; Griss and Rowell, 1986; antennal MNs, Bauer, 1987). Action potentials which we evoked electrically in axons propagate in both directions from their point of origin and, in principle, the motor effects shown here could be caused by secondary, descending neurones driven by ascending action potentials. However, the short latency between stimulus and motor responses (<20ms; Figs 5, 6) suggests direct effects of both PI(2)5 and DNC. Secondary (i.e. postsynaptic) descending neurones cannot be ruled out, however, because a descending neurone of unknown function has been shown to be inhibited when PI(2)5 is stimulated electrically (K. Hensler, in preparation).

One could argue that our results allow only restricted conclusions as was investigated only one pair of about 10 pairs of wing muscles per segment. However, shifts in the relative latency of M97 over the range ±4 ms correlate well with torque developed around the roll-yaw axis (Hensler and Robert, 1990), and torque is the ultimate measure of the strength of steering.

Our results do not reveal the pathway from DNs to motoneurones (MNs), but for ocellar DNs (including DNC) the pathway includes a population of intercalated thoracic interneurones (TINs) which coordinate steering in subsets of wing MNs (Reichert and Rowell, 1985,1986). No neurones postsynaptic to PI(2)5 have been identified. However, direct connections between PI(2)5 and wing MNs, as reported for ocellar DNs (Rowell and Pearson, 1983; Simmons, 1980) and for TCG (Tyrer, 1981), appear to be excluded (K. Hensler, unpublished results). After subtracting from the short overall latency of 10–20 ms (Fig. 5) (i) the conduction time for DN action potentials from the electrode to the mesothoracic ganglion (about 3 ms; K. Hensler, unpublished results), (ii) the conduction time and synaptic delays for MN action potentials from the ganglion to the wing muscles (about 3 ms; K. Hensler, unpublished results; A. Baader, personal communication), and (iii) synaptic delays for at least two synapses (about 2 ms), 2–12 ms remain. This does not allow for many neurones to be intercalated.

Together with comparable results for the wind-sensitive DN TCG (Möhl and Bacon, 1983; see Introduction) these data lead us to propose that DNs in general control steering by wing muscles via thoracic interneurones like those described for ocellar DNs by Reichert and Rowell (1985, 1986; see Introduction).

PI(2)5 and DNC are both involved in the control of compensatory head rolling (Fig. 6; A. Baader, personal communication). Investigations on the similarly organized neck muscle system of crickets reveal that head rolling is brought about by asymmetrical activity in dorsoventral neck muscles (Hensler and Honegger, 1985; Hensler, 1986) which are thought to be homologous with dorsoventral wing muscles (Honegger et al. 1984). At present the connectivity pattern between DNs and neck MNs is not known, but there is evidence that it is organized like the pathway to wing MNs described in the previous section, (i) The latency between the onset of the stimulation and the head movement is 10–20 ms (Fig. 6A, and calculation in Results), (ii) A neurone has been identified in the prothoracic ganglion of locusts which is excited direction-specifically by horizon movements and which produces head rolling by coordinating neck muscles. It also receives rhythmic input during flight activity (Hensler, 1990). This input/output pattern closely resembles that of TINs which are intercalated between ocellar DNs and wing MNs (Reichert and Rowell, 1985, 1986). (iii) Several prothoracic neurones with similar motor function to that described in ii have been identified in another orthopteran species, the cricket Gryllus campestris (Hensler, 19886).

Electrical excitation of DNs elicits steering by wings only during flight activity, but head movements are evoked only in quiescent locusts (Fig. 6B). Two explanations are possible: first, active head movements are caused by activity in many premotor neurones coordinating neck MNs (as found for active head rolling during eye cleaning in crickets, Hensler, 1988b), and artificially increased activity in only one of them might be masked by the motor effects of the others. However, the same argument should apply to wing MNs if this were the only explanation. A second, more likely, explanation is that the flow of information from DNs to neck MNs may be modulated or suppressed by unknown neural pathways which might not operate in quiescent locusts. Support for this hypothesis comes from measurements performed on locusts flying under visual closed-loop conditions which showed that the gain of compensatory head movements following simulated course deviation is highly variable and may change markedly within seconds. In addition, compensatory head movements are often suppressed or dominated by active head movements (Hensler and Robert, 1990).

Steering by head, wings and abdomen seems to be controlled by homologous neural systems. First, DN axons give off blebbed branches (presumed output regions) into the motor neuropile of most, if not all, neuromeres along their way (Fig. 3A; Griss and Rowell, 1986; Hensler, 1988a). Second, electrical stimulation of DNs can evoke movements of wings, head, abdomen and antennae (Figs 4,6; Baader, 1989), although not every DN excites MNs in all segments with the same intensity [e.g. PI(2)5 does not move the abdomen]. Third, the stimulus–response latencies are similar for muscles of wings (Fig. 5) and neck (Fig. 6; and calculation in Results) and the muscles moving the abdomen (Baader, 1989). Fourth, the neuronal organization underlying steering by wing MNs and head movements seems to involve TTNs of similar properties (Reichert and Rowell, 1985, 1986; Hensler, 1990).

PI(2)5 and DNC, with mutually contralateral axons, produce the same motor effects. In the mesothoracic ganglion both DNs cause increased excitation of MNs ipsilateral to the PI(2)5 axon; in the pro- and metathoracic ganglia MNs are more excited ipsilateral to the DNC axon (Fig. 7). Hence, PI(2)5 and DNC may interact in correcting roll and yaw deviation because, like most DNs, they have the same directional specificity for roll and yaw. It is not yet understood how they interact during pitch deviation because PI(2)5 is excited by pitch-up and DNC is excited by pitch-down.

This study was supported by a grant of the Swiss National Foundation to CHFR.