Mechanosensory-induced behavioural gregarization in the desert locust Schistocerca gregaria

Locusts have the remarkable ability to change between two morphologically,physiologically and behaviourally distinct forms depending on their population density (Uvarov, 1966). This density-dependent phase polyphenism is an extreme form of variation common to many species that migrate, form seasonal herds or are otherwise subject to fluctuating population densities (Dingle,1995). Locusts in the solitarious phase are cryptic in appearance and behaviour, occur at very low population densities and actively avoid each other. Conversely, locusts in the gregarious phase are brightly coloured as nymphs, more behaviourally active and aggregate together into bands, which may ultimately form migrating swarms consisting of millions of individuals(Uvarov, 1966; Simpson et al., 1999). The transition between the two extreme phases is a multistage process, with some changes in morphology and physiology taking several stadia or even generations to complete (Pener, 1991; Applebaum and Heifetz, 1999). By contrast, many behaviours change extremely rapidly. Solitarious locusts acquire most of the behavioural characteristics of gregarious phase locusts within only 4 h of crowding including, critically, the propensity to aggregate and thus reinforce the gregarizing stimuli provided by other locusts(Ellis, 1963; Gillett, 1973; Roessingh and Simpson, 1994; Bouaïchi et al., 1995). These rapid changes make locust phase change a potentially valuable tool in understanding neuronal plasticity and its behavioural consequences.

Olfactory and visual stimulation provided by other locusts promotes behavioural gregarization when presented together but has little effect individually. An extremely potent stimulus, however, is repeated physical contact produced by either touching or jostling, which elicits rapid and full behavioural gregarization in the absence of any other sensory stimuli(Roessingh et al., 1998; Hägele and Simpson,2000). This mechanical stimulation must be directed to the middle or hind legs in order to have any effect, with the hind femur being by far the most effective site at which to elicit gregarization(Simpson et al., 2001). Here,we have analysed what constitutes an adequate gregarization-inducing mechanosensory stimulus and developed a suitable preparation for future experimental investigations of processes in the central nervous system that drive or accompany phase change.

The five goals of the study were: (1) to determine how much of the surface of a hind femur needs to be mechanically stimulated in order to elicit behavioural gregarization; (2) to establish whether there is evidence for mechanosensory specialisation in the hind femur of solitarious locusts in terms of the number or distribution of mechanosensory hairs; (3) to develop a fixed preparation in which gregarization could be elicited in immobilised locusts, either through mechanical stimulation of a hind femur or electrical stimulation of hind leg nerves; (4) to analyse the relative roles of exteroceptive and proprioceptive signals carried in different leg nerves in eliciting gregarization; and (5) to determine whether gregarization is mediated via spiking local interneurones that receive convergent mechanosensory and chemosensory inputs from the hind leg(Newland, 1999) or through a separate mechanosensory pathway. To do this we used a chemosensory stimulus,acetic acid odour, that strongly activates chemosensory neurones without activating mechanoreceptors (Newland,1998).

All locusts (Schistocerca gregaria Forskål) originated from a colony maintained at the Department of Zoology, University of Oxford, UK reared under crowded conditions (500-1000 insects per 56 cm×76 cm×60 cm rearing bin). Solitarious phase locusts were derived from this colony but had been reared under physical, visual and olfactory isolation from other locusts for two or three generations as described in Roessingh et al.(1993). Unless otherwise stated, all experiments used final-instar nymphs.

The behavioural phase state of locusts was measured using an observation arena (41 cm×30 cm), flanked by two 7.5 cm×30 cm backlit chambers separated from the main arena by perforated clear plastic partitions. A group of 20 gregarious-phase final-instar nymphs was placed in one chamber, while the other chamber was left empty. An experimental locust was introduced into the arena via a central hole in its floor and the animal's behaviour recorded for the following 500 s using an event recorder in real time(Roessingh et al., 1993;Simpson et al., 1999, 2001). These data were analysed using the same statistical multiple logistic regression model in SPSS(version 10.0.07) as developed by Simpson et al.(2001), based on observations of 96 gregarious and 96 solitarious final-instar nymphs. A logistic algorithm was used to make an optimally fitting model, which correctly classified 92.7%and 91.7% of gregarious and solitarious locusts, respectively. This algorithm produced an index of solitariousness, termed P_sol, which is the probability that an assayed locust belonged to the solitarious model group. This provided an indicator of behavioural phase state, ranging from 1(for locusts indistinguishable from the solitarious model group) to 0 (when locusts behaved fully like the gregarious model group). In the analyses,locusts with P_sol values of less than 0.2 were judged to have become fully behaviourally gregarious. The P_soldistributions obtained for each experiment were normalised rank transformed to make the data suitable for parametric analyses to compare the effects of different experimental treatments on phase state.

A hind femur was stroked in locusts in which different regions on the anterior surface had been coated with a water-based poster paint (Winsor and Newton, Harrow, UK), leaving only one region with mechanosensory hairs free to move. Ninety-six locusts isolated for three generations were divided between six treatments in which half or three-quarters of the femur was painted(Fig. 1) and two control treatments in which the whole femur was painted and either left unstimulated or stroked as in the other treatments. The painted femora were stroked for approximately 5 s once every 60 s for 4 h using a small paintbrush as described in Simpson et al.(2001). At the end of the 4 h stimulation period, behavioural phase state was assayed.

The anterior faces of detached legs were drawn at 10× magnification using a camera lucida attached to a dissecting microscope, with the location of each tactile hair mechanoreceptor (60-780 μm long; Newland, 1991) marked on the drawing. The numbers of basiconic (bimodal chemosensory and mechanosensory)sensilla on the anterior distal hind femur were also counted, but as these are small (less than 40 μm long) a cast was made of the surface of the femur using transparent nail polish, which made a clear impression of the cuticle and all the sensilla. The casts were viewed and drawn under transmitted light at 100× magnification with the aid of a camera lucida.

Locusts were immobilised ventral side uppermost, with all legs fixed out laterally but with the anterior surface of the hind femur directed upwards,and given the same mechanosensory stimulation regime as the free-moving animals described above.

Locusts were restrained in modelling clay ventral side uppermost as described above. Stimulating electrodes, made from a pair of 50 μm coated steel wires, delivered 5 s of pulses once every 60 s for 4 h. Each 5 s burst consisted of 10 ms square pulses repeated at 50 Hz organised into 200 ms trains separated by 200 ms intervals. This electrical stimulation pattern was designed to simulate the temporal sequence of stimulation produced by stroking a femur with a paintbrush. The stimulating electrodes were placed within either the thorax or the hind femur according to treatment. In each location,a flap was cut in the cuticle and overlying air sacs and trachea were removed or displaced to reveal the selected nerve. The bare tips of the stimulating electrodes were carefully wrapped around the nerve and insulated with a 10:1 petroleum jelly: paraffin oil mix. Nerve 5B1 in the femur was identified by the branches it made onto the extensor tibiae muscle and by the activity of the femoral chordotonal organ (Usherwood et al., 1968), monitored using an AC amplifier and oscilloscope. Nerve 5B2 was identified by the branches it made onto the flexor tibiae muscle(Heitler and Burrows, 1977). The cuticle was replaced as far as possible and the wounds sealed with more petroleum jelly. In one set of experiments, the extensor tibiae muscle was stimulated directly by electrodes inserted into the muscle through the cuticle. The voltage was adjusted so that it just elicited leg movements. Protracted electrical stimulation often led to leg autotomization, but the nerves remained undamaged provided that the hind coxa was immobilised. After 4 h of stimulation, the electrodes were carefully removed and the locusts were rested for 10 min before being assayed for behavioural phase state. Controls where made in which the electrodes were placed but no current was passed. In all experiments, lost haemolymph was replaced with locust saline as necessary.

The thorax was cut open and nerve 5A, 5B or the whole of nerve 5 cut using fine iridectomy scissors. The cuticle was resealed using beeswax and, after a 10 min recovery period, the unrestrained locusts were placed in experimental boxes and a hind femur stimulated with a paintbrush for 4 h as described above.

Locusts were placed in boxes under a fume hood and stimulated at 1 min intervals for 4 h with one of the following treatments: 5 cm³ of the vapour taken from the headspace of a bottle of glacial acetic acid, which was slowly delivered over a hind femur for 5 s using a 5 ml syringe; a 10%dilution of the acetic acid vapour mixed with air; 5 cm³ of air;stroking a hind femur with the fibres and ferrule of a fine paintbrush mounted on a 5 ml syringe containing 5 cm³ of acetic acid vapour that was simultaneously discharged through the brush whilst stroking; or just stroking a hind femur with a brush. After stimulation, the locusts were removed and their behavioural phase state tested.

Mechanical stimulation of different regions on the anterior surface of the femur produced behavioural gregarization in the majority of locusts after 4 h(Figs 1, 2).

An analysis of variance (ANOVA) indicated significant differences between the treatment and control groups (F_7,88=10.55, P<0.001; Fig. 2). The control locusts in which the whole femur was painted and then left unstimulated remained highly solitarious (median P_sol=0.99, which was comparable with the model solitarious group; Fig. 1H). The paint formed an effective barrier to the mechanical stimulation of tactile hairs, as most locusts remained solitarious when the entire hind femur was painted and then stroked (median P_sol=0.90; not significantly different from the painted and unstimulated controls; Figs 1G, 2). Conversely, strong behavioural gregarization was elicited by stimulating free mechanoreceptors on either just the upper or lower halves of the femoral surface [median P_sol<0.025 (Fig. 2), with 83-92% of the experimental locusts having P_sol<0.2 (Fig. 1A,B)]. Stimulation of a smaller region, one-quarter of the femoral surface, also produced significant behavioural gregarization for three of the four tested regions (Figs 1C-E, 2) but was not as strong a stimulus: a greater proportion of locusts failed to show gregarious behaviour;fewer locusts (<60%) had P_sol<0.2(Fig. 1), and median P_sol values were consequently higher (0.1-0.36; Fig. 2). The lower distal quadrant alone stood out as a wholly ineffective site for tactile gregarization (Figs 1F, 2).

Mechanical stimulation of the hind femur in locusts that were unable to move failed to elicit behavioural gregarization after 4 h(Fig. 3). There was only a small, though significant, change in P_sol values in immobilised stroked locusts, where the median P_sol value was 0.93, compared with 0.98 for the control treatments in the painted leg experiments described above (t- test; equal variance not assumed;d.f.=32.04, N=101, P<0.05). P_solvalues of <0.2, consistent with fully gregarious behaviour, were found in only 13% of stroked restrained locusts.

Solitarious final-instar nymphs had 29% more sensilla on the anterior(outward) surface of a hind femur than did gregarious nymphs (mean ± s.e.m., 135±6.1 compared with 105±5.4; Fig. 4; significant interaction term of ANOVA shown in Table 1). Both phases had similar numbers of tactile hairs on the hind tibiae and tarsi as well as on all the segments of the front and middle legs(Fig. 4A). Adult solitarious locusts had 26% more tactile hairs on their hind femora than did gregarious adults (177±4.6 and 140±6.3 sensilla, respectively; Fig. 5A; significant interaction term of ANOVA shown in Table 1). Adult solitarious locusts had significantly fewer tactile hairs on the hind tarsus, front femur and front tarsus than gregarious phase adults (Fig. 5A), in contrast to final-instar nymphs. Solitarious adult locusts also had fewer sensilla on the front tibiae, but this was marginally non-significant(Fig. 5A). Adult locusts of both phases had similar numbers of sensilla only on the hind tibiae.

The hind femora of solitarious locusts are proportionately longer than those of gregarious locusts (Dirsh,1953), but this was not significantly correlated with the number of sensilla in either nymphs or adults [General Linear Model (GLM) analyses using femur length or length² (proportionate to area) as covariate;data not shown].

The additional sensilla of solitarious locusts were not evenly distributed over the hind femora (Figs 4B, 5B,C). There were 60% more sensilla on the most distal dorsal region and 36% more sensilla on the most proximal ventral region, but similar numbers on the distal ventral and proximal dorsal surfaces in final-instar nymphs (post-hoc t-tests;significance as in Fig. 4B). A similar distribution of sensilla was found on the hind femora of adults(Fig. 5B). Most hairs were located on the longitudinal ridges along the dorsal, latero-dorsal,latero-ventral and ventral surfaces of the femur and thus formed dense aggregations in solitarious locusts (Fig. 5C shows locations of hairs in adults). Solitarious locusts had significantly fewer bimodal basiconic sensilla on the distal femur than did gregarious locusts (57±2.9 and 66±1.1 sensilla, respectively; Fig. 4C).

Electrical stimulation of the whole of metathoracic nerve 5 within the thoracic cavity produced full gregarization in most immobilized locusts(Fig. 6A; median P_sol=0.13, with 62% of assayed locusts having P_sol<0.2). Electrical stimulation of nerve 5A alone was wholly ineffective at eliciting gregarization(Fig. 6B; median P_sol=0.98). While stimulation of nerve 5B(Fig. 6C) elicited some change in phase state, it was less effective than stimulating the whole of nerve 5(median P_sol=0.51, with just 37% of locusts having P_sol<0.2). Sham-operated controls, in which stimulating wires were placed around nerve 5 in the thorax but no current passed, had a median P_sol of 0.86, and only 0.6% of control locusts had a P_sol<0.2 (Fig. 6H). An ANOVA of all the thoracic nerve 5-stimulated data and sham-operated controls indicated a significant effect of treatment (normalized ranked data, F_3,54=4.56, P=0.006; Nvalues for each group and the results of Dunnet's two-tailed post-hoctests against the control group are given in Fig. 6).

Stimulating both branches of nerve 5B together in the femur(Fig. 6D,J) produced a similar degree of gregarization to that produced by stimulating nerve 5B within the thorax (median P_sol=0.66; Fig. 6D), but stimulating either nerve 5B1 (Fig. 6E) or nerve 5B2 (Fig. 6F) alone was much less effective. Stimulation of nerve 5B1 alone produced a small significant behavioural change, but the median P_sol was still 0.85 compared with a median value of 1.00 for sham-operated femoral controls (Fig. 6E,I,J), whereas electrical stimulation of nerve 5B2 (Fig. 6F) or of the extensor tibiae muscle alone(Fig. 6G) was entirely ineffective at eliciting any behavioural change. The small change in behavioural phase state produced by stimulation of nerve 5B1 was similar to the effect seen when femora of restrained animals were stimulated mechanically. An ANOVA of the normalized ranked data for animals stimulated in the femur indicated a significant effect of the different stimulation regimes on phase state (F_4,61=4.02, P=0.006; Nvalues for each group and the results of Dunnet's two-tailed post-hoctests against the sham-operated control group are given in Fig. 6). Autotomization of the hind leg was a common occurrence when nerves either in the leg or thorax were repeatedly stimulated. We could see no relationship between the incidence of autotomization and change in behavioural phase state, which depended solely on the nerve stimulated.

The data from the electrical stimulation experiments suggested that sensory inputs from more than one branch of metathoracic nerve 5 may be needed to elicit gregarization, as stimulating the whole of nerve 5 was more effective than stimulating nerve 5B. Free-moving locusts in which nerve 5A had been cut,however, gregarized normally following 4 h of stroking the hind femur on the operated side (Fig. 7B; median P_sol=0.18). Conversely, cutting nerve 5B completely prevented mechanosensory-induced gregarization from occurring(Fig. 7C), and the data were similar to a control group in which the whole of nerve 5 had been severed prior to mechanical stimulation (Fig. 7A). A second control group, in which the femur contralateral to an entirely cut nerve 5 was stroked, showed normal gregarization(Fig. 7D). These data indicate that the operation on the insect itself did not prevent, or promote, the propensity for phase transition. An ANOVA of normalized ranked data showed a significant effect of the different treatments on the resulting behavioural phase state (F_3,32=4.00, P=0.016; the significances of Dunnet's two-tailed post-hoc tests against the ipsilaterally stimulated control group are shown in Fig. 7).

Puffing a stream of air over a hind femur elicited little behavioural response in the locusts. By contrast, the odour of acetic acid, particularly at the stronger concentration, evoked vigorous withdrawal reflexes similar to those evoked by tactile stimulation but with less adaptation. Neither air(Fig. 8A) nor acetic acid odour(Fig. 8B,C), however, elicited any behavioural gregarization; median P_sol values were 0.99 for air-treated animals, 0.97 for the 10% strength acetic acid odour and 0.96 for the full-strength acetic acid odour. Fewer than 12.5% of locusts had P_sol values less than 0.2. Stroking a hind femur in conjunction with acetic acid odour (Fig. 8E), however, produced strong behavioural gregarization, with a median P_sol value of 0.10 compared with a value of 0.15 for those just mechanically stimulated(Fig. 8D). The acetic acid odour therefore did not inhibit or enhance gregarization. An ANOVA of normalized ranked data confirmed the significance of the results(F_4,35=4.60, P=0.004; the significances of Dunnet's post-hoc tests using the air-stimulated animals as the control group are indicated in Fig. 8).

Tactile stimuli provide a potent drive for behavioural phase transition in locusts, but their efficacy differs depending on the site that is stimulated. The most effective location is the anterior (outward-facing) surface of a hind femur (Simpson et al., 2001). We show that mechanical stimuli directed to most regions on this surface in free-moving solitarious locusts can elicit full behavioural gregarization.

Stimulation of half of the femoral surface is more effective at inducing gregarization than stimulating just one quarter. This suggests that in order to elicit gregarization reliably there may be a minimum area over which stimulation must occur but that there is no specific region on a femur that must be touched. The only ineffective site is the ventral distal region, which has few tactile hairs.

A striking finding of this study is that solitarious phase locusts have, on average, 30% more tactile hairs on their hind femora than do gregarious locusts. By contrast, solitarious locusts have similar numbers or fewer tactile hairs on other hind leg segments and on the front and middle legs. The hind femora have a considerably larger surface area than any other leg segment and because of their orientation during walking or standing present themselves as large targets for mechanosensory stimuli. The additional tactile hairs of solitarious locusts are concentrated on the most prominent or protruding parts of the hind femur, in the regions most likely to be contacted or jostled by other locusts. These data suggest that there is some functional specialization of the hind femora of solitarious locusts towards detecting mechanosensory gregarizing stimuli. Nevertheless, although touch-induced gregarization is largely restricted to the hind femora, this cannot simply be a consequence of the number of mechanoreceptors present on different legs. A hind femur has more tactile hairs than any other leg segment, but it has fewer tactile hairs on one quarter of its anterior surface than on the whole of the front or middle femora; yet, stimulating just these hind leg sensilla is sufficient to produce gregarization, whereas stroking the whole of the front or middle femora has little gregarizing effect.

The legs of locusts also possess dual modality basiconic sensilla, which contain a single mechanosensory and several chemosensory afferent neurones. The mechanosensory afferents from basiconic sensilla are more sensitive to touch than those from tactile hairs, responding best to short, intermittent stimuli and adapting strongly on repeated stimulation(Newland and Burrows, 1994). There are 14% fewer basiconic sensilla on the distal femora of solitarious locusts than gregarious locusts. Despite this, solitarious locusts still have 27% more mechanosensory afferents overall in this region because they have 41%more tactile hairs.

Tactile hairs fall into at least two response classes, with distinct angular deflection activation thresholds: low-threshold hairs have phaso-tonic response characteristics that enable them to respond to repeated stimuli,whereas high-threshold hairs adapt completely within a few cycles of stimulation (Newland, 1991). It is unknown whether the change in mechanoreceptor numbers affects one class of tactile hair more than the other and it is possible that the physiological properties of the mechanosensory neurones themselves could also differ between phases. We would predict that phaso-tonic mechanosensory neurones, whose long-lasting response properties mean that they are able to signal repeated contacts, should be more effective in inducing gregarization. This suggestion is supported by the reduction in the number of basiconic sensilla on the femora of solitarious locusts. The highly phasic response properties and rapid adaptation of basiconic mechanosensory afferents makes it unlikely that they could provide a suitable gregarizing signal.

The reduction in the number of basiconic sensilla also means a reduction in the population of contact chemosensory neurones, since these sensilla are bimodal. This decrease in the number of contact chemosensory neurones contrasts with an increase in the number of antennal olfactory sensilla reported in solitarious Locusta migratoria(Greenwood and Chapman, 1984). Contact chemosensory stimuli are not thought to be essential in the early stages of gregarization (Hägele and Simpson, 2000), although the olfactory stimulation provided by other locusts in conjunction with appropriate visual stimulation can induce rapid behavioural gregarization (Roessingh et al., 1998). Our use of a strong chemosensory stimulus in the acetic acid odour experiment, which activates contact chemosensory neurones on the legs, was designed to test another hypothesis, specifically whether the gregarizing stimulus from tactile hairs passes through spiking local interneurones in the metathoracic ganglion. The experiment exploited the known convergence of contact chemosensory neurones and mechanosensory neurones from both basiconic sensilla and tactile hairs onto these interneurones(Burrows and Newland, 1994; Newland, 1999; Newland et al., 2000). The great majority of spiking local interneurones that receive an exteroceptive mechanosensory input also receive monosynaptic chemosensory inputs(Newland, 1999; Rogers and Newland, 2002). If the signal to gregarize passes solely from the tactile hair afferents to these spiking local interneurones it should have been possible to produce behavioural phase change by activating the same interneurones through a purely chemosensory stimulus (by using the acetic acid odour). As this stimulus had no effect on behavioural phase state, we conclude that either there must be a separate pathway that processes the tactile signals responsible for eliciting gregarization or that there must be a convergence with other signals, not provided by chemosensory stimulation, downstream from the spiking local interneurones. Tactile hair afferents are known to make parallel monosynaptic connections onto some classes of intersegmental interneurones(Laurent and Burrows, 1988a; Newland, 1990), non-spiking interneurones (Laurent and Burrows,1988b), motor neurones(Laurent and Hustert, 1988)and perhaps onto as yet unidentified interneurones. Spiking local interneurones are important components in the pathway organising the leg avoidance reflexes elicited by mechanical or chemosensory stimuli, but we show that this pathway cannot by itself drive behavioural phase change.

Behavioural gregarization could not be induced by mechanical stimulation of restrained locusts that were unable to move their hind legs, indicating that the effective gregarizing stimulus is not purely exteroceptive. It must combine exteroceptive stimuli with another mechanosensory signal that is produced when free-moving animals are stimulated with a paintbrush. To help determine the nature of this additional signal we carried out electrical nerve stimulation and ablation experiments.

Electrical stimulation of metathoracic nerve 5, which carries nearly all the sensory and motor neurones that innervate the hind leg distal to the coxa,reliably elicited behavioural gregarization in fully restrained locusts. This argues against motivational state (e.g. `stress') being a factor in the failure to elicit behavioural gregarization in restrained locusts through mechanical stimulation. It suggests instead that electrical stimulation of nerve 5 activates a necessary component of the normal mechanosensory stimulus that is missing when restrained locusts are touched. Electrical stimulation of proximal nerve 5 was generally more effective than stimulation applied more distally. One possible explanation for this could be that current leak from the electrodes placed around the whole of nerve 5 reached the metathoracic ganglion and directly stimulated central neurones integral to initiating phase change. This seems unlikely, however, because in other experiments the electrodes on nerves 5A and 5B were similar distances from the ganglion, yet stimulation of nerve 5A was entirely ineffective at eliciting gregarization. Electrical stimulation of nerve 5 provides us with a powerful fixed preparation with which to initiate and monitor the effect of behavioural gregarization in the central nervous system.

It seems likely that the component necessary to induce behavioural phase change that is missing during mechanical stimulation of restrained locusts or during chemosensory stimulation by acetic acid odour is proprioceptive input from the basal leg joints. This could relate either to the resting position of the leg or could signal the medial displacement of a hind leg towards the body caused by pressure from a paintbrush or another locust. If the latter, this suggests that for gregarization to occur locusts need to be not just touched but jostled strongly enough to produce limb or body displacement. This jostling, if it pushes the femur towards the body, will produce a movement about the complex thoraco-coxal joint with mostly remotion and levation components. The femoro-tibial joint moves in a plane oblique to the direction of the major force produced by touching with a paintbrush and is unlikely to be strongly stimulated, but there may be twisting forces out of the movement plane about the femoro-tibial joint. Movements of the leg that are part of the usual movement pattern of locomoting locusts should not produce proprioceptive stimuli that elicit gregarization, otherwise locusts would risk gregarizing during usual activity.

Moving the leg in the absence of tactile exteroceptive input, for example by stroking a femur when all the exteroceptors have been immobilised with paint, fails to elicit gregarization, ruling out the possibility that the gregarizing stimulus is entirely proprioceptive. Similarly, gregarization induced by sensory feedback from motor responses elicited by the mechanical stimulus in free-moving animals (e.g. avoidance reflexes; Siegler and Burrows, 1986) can be excluded because locusts do not gregarize when the extensor tibiae muscle is stimulated electrically so that the tibia moves or when basiconic sensilla are stimulated with acetic acid odour, which elicits vigorous leg withdrawal responses.

The femur and basal leg joints contain many proprioceptors that could provide the necessary signal (Hustert et al., 1981; Bräunig et al.,1981; Bräunig,1982; Mücke,1991). Fig. 9summarises the major sense organs of the hind leg served by different nerves. As full behavioural gregarization could be elicited by stimulation of nerve 5 alone, signals from leg proprioceptors whose axons run in nerves 2 and 3B(Bräunig and Hustert,1985) can be excluded from a critical role in monitoring gregarizing stimuli, unless the relevant signals can be carried through several redundant pathways. The greater efficacy of electrically stimulating the whole of nerve 5 in driving gregarization compared with stimulating any of its branches separately initially suggested a role for proprioceptive neurones with axons in nerve 5A, such as from two (sub)coxal multipolar sensilla and hair rows sensitive to movements of the coxa(Bräunig et al., 1981; Bräunig, 1982; Fig. 9). Cutting nerve 5A,however, does not reduce the efficacy of touch stimulation, excluding a necessary role for these particular proprioceptors. It is possible, however,that the medial displacement of a hind leg during stimulation with a paint brush, which should strongly stimulate nerve 5A sensory neurones, could contribute to the proprioceptive component of the gregarizing signal and explain the slightly stronger gregarizing effect of electrical stimulation of the whole of nerve 5.

Electrical stimulation of nerve 5B is an effective gregarizing stimulus,yet stimulation of either nerve 5B1 or 5B2 alone is not effective in eliciting robust gregarization despite there being no sensory structures innervated by nerve 5B prior to its bifurcation into 5B1 and 5B2. Stimulating both nerve 5B1 and 5B2 together is almost as effective as stimulation of the whole of nerve 5B, suggesting that the relevant signals must be carried by neurones within nerves 5B1 and 5B2. Nerve 5B1 innervates tactile hairs on the anterior face of the femur, nerve 5B2 innervates those on the posterior face of the femur and, via its first branch (the lateral nerve; Heitler and Burrows, 1977),the row of tactile hairs on the ventral ridge of the femur(Fig. 9). It would be expected therefore that the major exteroceptive input driving gregarization would come via nerve 5B1. This exteroceptive input must be combined with proprioceptive signals, at least part of which should come from nerve 5B2. Both nerves 5B1 and 5B2 supply proprioceptive sensilla on the trochanter,including two campaniform sensilla fields each(Hustert et al., 1981; Bräunig, 1982), which monitor strain on the cuticle. Nerve 5B1 also innervates the trochanteral hair plate and hair row, which monitor movements of the femur. Any of these proprioceptors, alone or in combination, could monitor the displacement of the leg that accompanies appropriate touch stimulation and thus contribute to the gregarizing stimulus. We expect that at least one of the proprioceptive afferents must travel in nerve 5B2, which would explain why only stimulation of both nerves, or the conjoined nerve 5B, is effective in eliciting gregarization. Trochanteral campaniform sensilla are the only known basal leg joint proprioceptors in nerve 5B2, strongly implicating them in mediating proprioceptive phase change stimuli. There are a number of multipolar sensilla innervated via the lateral nerve branch of nerve 5B2, some of which are sensitive to twisting movements of the femoro-tibial joint(Bässler, 1977; Williamson and Burns, 1978; Matheson and Field, 1995),which could also respond to medial displacement of the hind femur(Fig. 9). As with exteroceptors, it is also possible that long-term solitarization is accompanied by changes in the number or physiological properties of proprioceptors.

The neuronal circuits that integrate mechanosensory gregarizing stimuli should combine exteroceptive signals from the anterior surface of a hind femur with a specific proprioceptive signal that naturally results from the inward displacement of the leg on contact with another locust. These neuronal circuits are the first central elements of the pathway that initiates the rapid and widespread neuronal plasticity that underlies behavioural phase change. We show that the inputs to these neuronal circuits come from tactile hairs on the hind femora and from other leg receptors and that this input system is itself modified by the process of phase change.

This work was supported by a grant from the BBSRC (UK) to M. Burrows, T. Matheson and S. J. Simpson.