Intracellular Recordings From Intact Locusts Flying Under Closed-Loop Visual Conditions

The principal neural events underlying corrective steering have been un-ravelled: sensory information on course deviations (coming from the eyes, ocelli and wind-sensitive hairs on the head) and proprioceptive information caused by head movements are coded in a class of descending deviation detector neurones (DDNs). 15 DDNs have already been identified. Most of them are multimodal and respond in a direction-specific manner to deviations from course (e.g. Rowell and Reichert, 1986; Hensler, 1988,1992). Information from DDNs is further processed and transmitted to motoneurones via a class of thoracic interneurones (Reichert and Rowell, 1986; Rowell and Reichert, 1991; Hensler, 1992).

Corrective steering stabilizes the locust’s orientation in space. A new orientation in space is acquired by a behaviour which I call ‘voluntary’ steering (this does not imply a conscious act by the locust). ‘Voluntary’ course deviations produce the same sensory inputs as involuntary course deviations; hence, the ‘autopilot’ has to be disabled during voluntary steering otherwise corrective steering would restore the initial flight path. In principle, voluntary steering could be accomplished in two ways: (1) by modulating the neurones of the ‘autopilot’ in order to produce the ‘voluntary’ course change, or (2) by controlling moto-neurones through a parallel neuronal pathway, while the ‘autopilot’ is temporarily suppressed or overridden. To investigate this, intracellular recordings of steering neurones are required during the performance of ‘voluntary’ steering. The key problem of such an investigation is to distinguish between ‘voluntary’ and corrective steering in an experimental set up.

The usual set up for making intracellular recordings of steering neurones used either reduced (Rowell and Reichert, 1986) or intact (Hensler and Rowell, 1990) locust preparations, which were fixed in front of an artificial horizon. The former animals performed fictive flight (i.e. flight muscles were denervated, but wing motoneurones were active with a motor pattern resembling that of normal flying), the latter performed normal flight behaviour. Under such open-loop visual conditions, corrective steering is easily induced by rotating the artifical horizon in order to simulate course deviations for the locust (Rowell and Reichert, 1986; Hensler, 1988, 1992), but ‘voluntary’ steering cannot be determined unequivo-cally. It can only be detected when the locust adopts and maintains a new orientation with respect to its visual surroundings, either spontaneously or in response to an attractive or repellent stimulus. This situation may be simulated in a so-called closed-loop compensator: this set up has been used for the investigation of visual orientation in flies (e.g. Götz, 1987) and recently also in locusts (Robert, 1988). The present report introduces a method which enables intracellular recordings to be made from flight neurones of intact locust preparations flying in a closed-loop compensator. In the set up used here, the feedback loop for visual inputs signalling rotation around the roll axis (longitudinal body axis) was closed (in principle, the same method works for rotation around the yaw and pitch axes).

Locusts {Locusta migratoria) were tethered on a flight balance in front of an artificial semihorizon (diameter 15 cm; the posterior halves of the eyes were covered). The flight balance consisted of a holder connected to a rotatable metal rod fixed on ball bearings. The axis of rotation corresponded to the longitudinal body axis of the animal (Fig. 1). The rod was held in place by an elastic element (a strip of Sylgard), which restricted rotation to less than 1° when the locust produced maximum torque. A small permanent magnet was fixed to the metal rod in order to record (as a measure of torque) minute rotations using a Hall detector. Flight behaviour was induced and maintained by a frontal airstream of about 2ms⁻¹. Turbulence was reduced by blowing the airstream through a bundle of pipes (diameter of each pipe 4mm; diameter of the whole bundle 7cm; length 6cm). Course deviations were simulated by rolling the artificial horizon (e.g. horizon rolling to the left simulated roll deviation of the locust to the left).

In freely flying locusts a steering manoeuvre produces torque, which reorients the animal in space. This mechanism was simulated in the closed-loop compensa-tor by electronically integrating the torque signal and setting the roll orientation of the artificial horizon in proportion to the resulting voltage level (Fig. 1). Accordingly, a pulse of torque resulted in a new orientation of the artifical horizon and constant torque produced constant rotation of the artificial horizon. The locust’s incentive to steer was controlled by feeding a bias voltage into the integrating element. The bias voltage caused horizon rotation with an angular velocity of 90° s⁻¹ when steering effort ceased. Transient course deviations were simulated by feeding short voltage pulses into the integrating element (in the open-loop mode, the horizon position was controlled by a function generator).

Flight performance was monitored electromyographically from the first basalar musles of both forewings (M97). Head rolling was recorded using a capacitative device (Hensler, 1988). DDNs were recorded intracellularly from the neck connectives through a small window cut into the posteriodorsal portion of the pronotum. The anterior part of the gut was removed to expose the neck connectives, which were then fixed with tissue glue onto a supporting metal platform. Locusts were otherwise intact and were capable of performing normal flight activity and normal head movements. Details of the preparation method are described elsewhere (Hensler and Rowell, 1990).

The results shown in this paper are taken from DNC, a DDN that has already been described under open-loop visual conditions (Rowell and Reichert, 1986). DNC was phasically excited by course deviations towards the side contralateral to the axon (in the following this is called the ‘preferred direction’) and phasically inhibited by simulated course deviations towards the side ipsilateral to the axon (the ‘antipreferred direction’). It was tonically excited by frontal wind. The anatomy of DNC has been described by Griss and Rowell (1986).

Fig. 2 shows the activity pattern of the DNC with its axon in the left-hand connective during an episode of closed-loop flight. Initially the horizon was stabilized at a roll angle of about 15° right of the normal orientation. This stabilization resulted from active steering because the horizon would otherwise have rotated at 90° s⁻¹ as a result of the bias fed into the horizon-controlling servo. DNC spiked at 11 Hz during this period. Simulated deviation of 65° to the left, i.e. in the antipreferred direction of the DNC under investigation (Fig. 2, left arrowhead), induced steering which restored the normal position within 400ms. DNC did not spike during this period except for one action potential (AP) shortly before the normal orientation was regained. An overshoot of the steering reaction caused slight deviation to the right. Shortly after the overshoot had passed its maximum (a point not easily definable), there was a burst of APs in DNC (40Hz for 150 ms; asterisk in Fig. 2). This burst seemed to be of central origin since it was neither evoked by visual input (during this period the visual image moved in the antipreferred direction) nor could any other obvious external input be correlated with it. After 1 s of stable, straight flight, a course deviation of 60° to the right was simulated, i.e. in the preferred direction of the DNC under investigation (Fig. 2, right arrowhead). This induced a burst of APs (45 Hz, 150 ms) followed by spiking at about 10 Hz until the normal flight course had been attained after about 800 ms. Note that the peak deviation was associated with the failure of one AP in the left M97. The relative timing of APs in the left and right M97 plays an important role in steering (Mohl and Zarnack, 1977; Hensler and Robert, 1990). The head did not perform compensatory movements during the entire sequence, apart from a sudden roll of 13° to the right about 15 ms after the initial spike of the burst marked with the asterisk.

Under open-loop visual conditions, electrical stimulation of DNC induces shifts in the timing of wing motoneurones (Hensler and Rowell, 1990). Fig. 3 shows that such stimulation produces torque that relocates the horizon. Hence, the burst in Fig. 2 (asterisk) probably participated in the restoration of the normal flight course, which was stabilized almost perfectly afterwards. Because of the back-ground of other horizon movements, the electrically induced roll manoeuvre in Fig. 3 was difficult to detect in individual observations. The overall tendency, however, was clearly discernible. Stimulation of DNC could be expected to have only a small effect, as DNC is merely one neurone of a bundle of at least 15 pairs of DDNs involved in course control (Hensler, 1992). In parallel with the production of torque, electrical stimulation of the left DNC induced head rolling to the left (Fig. 3; Hensler and Rowell, 1990, described the same effect, but their statement in the first paragraph of p.200 mistakenly exchanges left and right). The head rolling in Fig. 2 (asterisk) could not have been initiated by the corresponding burst in DNC as it was directed to the right.

In summary, I have shown that it is possible to record intracellularly from neck connectives of locusts flying under nearly natural conditions, i. e. steering causes reafferent visual input. This method is suited to the investigation of the neural control of ‘voluntary’ steering, as the flight path is the only relevant measure that can be used to determine whether a steering movement is corrective or the basis of a ‘voluntary’ manoeuvre. Correlation of neurone activity with the flight path should reveal neurones involved in this behaviour. As far as visual responses in the closed-loop situation are concerned, DNC behaved qualitatively in the same way as it did under open-loop visual conditions during flight (Hensler and Rowell, 1990) and in resting locusts (Rowell and Reichert, 1986). The same holds for DNI, another DDN (K. Hensler, unpublished result). Most interesting was the occurrence of the ‘spontaneous’ burst (asterisk in Fig. 2), which may have been evoked by central input to DNC. Central excitation, inhibition and/or modulation seem to be a prerequisite for the induction of ‘voluntary’ steering by DNC or other DDNs.

This study was supported by a grant from the Swiss National Foundation to C. H. F. Rowell, whom I thank for critically reading the manuscript. Karin Tritt accomplished the painstaking task of correcting the English.