Excitatory and Inhibitory Roles of Central Ganglia in Initiation of the Insect Ecdysis Behavioural Sequence

The initiation of behaviour at appropriate times depends on the selective activation of certain neural networks and inhibition of others. The unique properties of these circuits, including their responsiveness to excitatory or inhibitory influences, are determined by precise patterns of gene expression. The relationships between gene expression and behaviour can be conveniently examined in instances where they play critical roles in developmental and reproductive functions. For example, the initiation of insect behaviours such as wandering and ecdysis is preceded by intense bouts of gene expression under the control of steroid hormones (ecdysteroids), resulting in the production of specific receptors, the synthesis of signalling molecules and the precise timing of their release (Thummel, 1995; Hewes and Truman, 1994; Žitňan et al., 1999).

Behaviours associated with shedding of the old cuticle are crucial for the successful development of the tobacco hornworm Manduca sexta. This behavioural sequence, consisting of pre-ecdysis I, pre-ecdysis II and ecdysis, is initiated by the direct action of peptide hormones from endocrine Inka cells on the central nervous system (CNS) (Žitňan et al., 1996). During the 1–2 days preceding larval ecdysis of M. sexta, a pulse of ecdysteroids controls the expression of specific genes both in Inka cells and in the CNS necessary for the subsequent initiation of these behaviours (Žitňan et al., 1999). Rising ecdysteroid levels induce ecdysis-triggering hormone (ETH) gene expression in Inka cells, while a decline in ecdysteroid levels regulates the late transcriptional activity required for the release of pre-ecdysis-triggering hormone (PETH) and ETH derived from this gene (Žitňan et al., 1999). Meanwhile, elevated ecdysteroid levels induce CNS sensitivity to these peptide hormones (Žitňan et al., 1999) and the fall in ecdysteroid levels induces increased excitability of the brain VM neurons followed by the release of eclosion hormone (EH) (Hewes and Truman, 1994). Blood-borne EH causes the release of PETH and ETH from Inka cells (Ewer et al., 1997; Kingan et al., 1997), which in turn act back on the CNS (Žitňan et al., 1999). Low levels of PETH and ETH subsequently initiate pre-ecdysis I and II. Increased ETH concentrations in the blood lead to activation of the ecdysis neuronal network (Žitňan et al., 1996, 1999), but ecdysis behaviour is initiated after central release of crustacean cardioactive peptide (CCAP) from this network (Gammie and Truman, 1997). Each motor unit of the ecdysis behavioural sequence is driven by a specific central pattern generator (Weeks and Truman, 1984a; Novicki and Weeks, 1995). It is therefore important in the context of the needs of the animal that selective activation and inhibition of these different motor units is provided at the appropriate time and in the appropriate order.

In this paper, we identify specific target ganglia for PETH and ETH, and provide new insights into the mechanisms underlying the switch from pre-ecdysis to ecdysis. We have found that the ecdysis neuronal network is activated by ETH early in pre-ecdysis, but the onset of ecdysis is delayed by inhibition from cephalic and thoracic ganglia. The eventual switch to ecdysis is controlled by an independent clock in each abdominal ganglion, which relieves the inhibition and initiates ecdysis. Our data show that PETH and ETH induce a series of downstream events within the CNS, leading to the subsequent activation of distinct central pattern generators and the performance of the ecdysis behavioural sequence at a specific time. This system provides an ideal opportunity to study mechanisms underlying a centrally patterned series of movements, since each behaviour can be hormonally induced and analyzed in animals in vivo and in the isolated CNS in vitro.

In vivo and in vitro bioassays using intact, ablated or transected central nervous sytems were carried out to determine the target sites for PETH and ETH in pharate fifth-instar larvae of Manduca sexta (L.) 6–8 h before ecdysis. Prior to each surgical procedure, larvae were anaesthetized with CO₂. For the CNS transections, a small incision was made through the ventral cuticle between abdominal segments 4 and 5 or 6 and 7, and the connectives between the respective ganglia were severed using forceps. For the ablation or removal of the brain, a small incision was made using sharp ethanol-sterilized forceps into the frontal part of the head, and the entire brain or proto- and deutocerebrum was cut off using microscissors. Selected groups of these larvae were injected with 50 pmol of one peptide, or injected with one peptide followed by treatment with the other peptide. Different phases of the induced ecdysis behavioural sequence were observed under a stereomicroscope and recorded on videotape. To study the inhibitory role of the cephalic and thoracic segments in delaying the onset of ecdysis, larvae were injected with 50 pmol of ETH and ligated behind the head or the first two abdominal segments. The anterior part was cut off using scissors 10–30 min into pre-ecdysis.

For in vitro studies, the isolated CNS was transected between abdominal ganglia 1 and 2 (AG1–2) or each abdominal ganglion was individually isolated before treatment with PETH or ETH (100 nmol l⁻¹). Alternatively, nerve cords were treated with these peptides and transected 10–30 min after the initiation of pre-ecdysis. Burst patterns triggered by these peptides were detected by extracellular recordings with suction electrodes as described by Žitňan et al. (1996, 1999).

To compare different functions of peptides released by Inka cells or the CNS during ecdysis, the desheathed entire CNS or isolated chain of AG1–8 was treated with 100 nmol l⁻¹ of ETH, EH or CCAP in vitro, and induced burst patterns were recorded from dorsal nerves of the abdominal ganglia using suction electrodes. In some cases, only the anterior ganglia (suboesophageal ganglion, SG, and thoracic ganglia 1–3, TG1–3) or posterior CNS (AG1–8) were desheathed on the dorsal side and the remaining ganglia were left intact. These nerve cords were then treated with EH (100 nmol l⁻¹) to determine its ability to induce elevations in cGMP levels and/or ecdysis bursts in intact non-desheathed ganglia.

For immunohistochemical detection of cGMP in the CNS, we followed modified procedures described previously (Ewer et al., 1994; Gammie and Truman, 1999). Using this simplified procedure, no methanol or collagenase treatment to allow antibodies to penetrate the CNS was necessary. Briefly, the CNS was fixed overnight in 4 % paraformaldehyde, washed in phosphate-buffered saline containing 0.5 % Triton X-100 (PBST) and incubated for 2 days with sheep or rabbit antiserum to cGMP (dilution 1:10 000 and 1:2000, respectively) obtained from Drs J. De Vente and H. M. W. Steinbush. Specific binding of these antisera was detected by overnight incubation with horseradish peroxidase (HRP)-labelled donkey anti-sheep IgG (Jackson Immunoresearch Labs, West Grove, PA, USA) or HRP-labelled goat anti-rabbit IgG (American Qualex, San Clemente, CA, USA). HRP was stained with diaminobenzidine (Sigma, St Louis, MO, USA) and hydrogen peroxide, and nerve cords were mounted in glycerol.

Values are presented as means ± S.D.

PETH and ETH induce the two distinct behavioural parts of pre-ecdysis. PETH elicits pre-ecdysis I, which is characterized by strong dorsoventral contractions, while ETH induces pre-ecdysis II, which consists of posterioventral and proleg contractions (Žitňan et al., 1999).

To identify the target ganglia for each peptide, we observed the effects of nerve cord transection on pre-ecdysis I and II contractions in larvae injected with PETH or ETH. Following surgical transection of the nerve cord between AG4 and AG5 (N=10) or AG6 and AG7 (N=10), animals were injected with PETH (50 pmol). Within 4–8 min, the peptide induced strong and regular dorsoventral contractions typical of pre-ecdysis I only in segments posterior to the cut (Fig. 1A). Abdominal segments anterior to the transection displayed strong, but prolonged and irregular, dorsoventral contractions (3–25 s of contraction and 10–30 s of relaxation). This behaviour ended within 35–50 min, at which time abdominal segments anterior to nerve transection remained in permanent dorsoventral contraction, while the posterior abdominal segments maintaining a connection to the terminal abdominal ganglion (TAG, fused AG7,8) were relaxed. These experiments showed that PETH targets all the abdominal ganglia, but that coordinated dorsoventral movements depend on pacemaker activity in the TAG.

Subsequent injection of the same larvae with ETH (50 pmol) induced normal pre-ecdysis II behaviour within 8–12 min. This consisted of strong posterioventral and proleg contractions both anterior and posterior to the transections (Fig. 1B; N=20). All animals (N=20), switched to typical ecdysis peristaltic contractions 25–36 min later, but only in segments anterior to the cut (Fig. 1C). The transition to ecdysis behaviour therefore requires a connection between the abdominal ganglia and the anterior CNS.

To identify specific ganglia targeted by PETH, we examined the effects of nerve cord transection on bursting in the dorsal nerves of the abdominal ganglia in vitro. In the isolated intact CNS, application of PETH (100 nmol l⁻¹) induced rhythmic, synchronous bursting motor patterns typical of pre-ecdysis I within 5–8 min (Fig. 2A; N=10). Transection of the abdominal connectives led to irregular, long-duration bursting of isolated ganglia, whereas typical pre-ecdysis I bursts persisted in the TAG (N=8). Similarly, in nerve cords transected prior to peptide treatment, PETH always induced normal pre-ecdysis I bursting in the TAG, while individually isolated AG3–6 showed only irregular, prolonged bursting (Fig. 2B, N=10). This pattern was similar to the long dorsoventral contractions of transected larvae observed in vivo (Fig. 1A). These experiments showed that all the abdominal ganglia respond to PETH application, but that interneurons projecting from the TAG determine the patterns of pre-ecdysis I bursting in the entire abdominal chain. The pacemaker interneurons providing this input have been identified as IN-402 (Novicki and Weeks, 1995).

Similar experiments were performed to ascertain which ganglia are targeted by ETH for the initiation of pre-ecdysis II. Exposure of isolated AG1–6 to ETH (100 nmol l⁻¹) led to strong bursting in the ventral nerves (Fig. 2C; N=12) corresponding to posterioventral and proleg contractions. However, unlike the experiments described above with PETH, typical pre-ecdysis II bursts were induced in individually isolated abdominal ganglia (N=8) after the transection of all connectives (Fig. 2D). Therefore, each abdominal ganglion appears to contain the entire neural circuitry necessary for pre-ecdysis II and is not dependent on pacemaker interneurons projecting from distant ganglia. These results also showed that the target sites for ETH are located in each abdominal ganglion, which is consistent with in vivo experiments described above.

While low concentrations of Inka cell peptides trigger pre-ecdysis I and II, higher levels of ETH are necessary to trigger the switch from pre-ecdysis to ecdysis (Žitňan et al., 1999). A previous report showed that the brain plays an essential role in the transition to ecdysis (Novicki and Weeks, 1996). By performing experiments in vivo and in vitro, we have found that both the brain and the SG are important in ETH-induced ecdysis.

We examined the roles of the cephalic ganglia in the activation of ecdysis by removing the brain of the larvae or by severing the connectives behind the SG prior to injection of the peptide. Most debrained animals (17 out of 20; 85 %) failed to make the transition to ecdysis following ETH injection (50 pmol), and transection of the connectives posterior to the SG prevented ecdysis in all larvae (N=10). To determine the areas of the brain involved in the activation of ecdysis, partial ablations were performed. When the protocerebrum and deutocerebrum were ablated, leaving one or both tritocerebral lobes intact, most animals (10 out of 14; 71 %) initiated ecdysis behaviour at the expected time. These experiments show that the tritocerebrum and SG are critical in facilitating the transition to ecdysis in vivo.

We also examined the involvement of the anterior ganglia in the transition to ecdysis in isolated nerve cords. In these experiments, the connectives were transected posterior to the brain, SG or TG3 prior to exposure to ETH (100 nmol l⁻¹). In 25 out of 26 cases, brainless nerve cords switched from pre-ecdysis to ecdysis after the normal interval 41.64±4.2 min (Fig. 3A). However, nerve cords transected posterior to the SG showed ecdysis in only two out of eight experiments (in 45 and 49 min). Removal of TG1–3, leaving intact a chain of AG1–8, prevented the transition to ecdysis in all experiments (N=8). These results suggest that the action of ETH on the SG, and in some cases on TG1–3, is sufficient to activate the ecdysis circuitry in vitro. Experiments with individually isolated abdominal ganglia showed that motor units driving ecdysis are present in each abdominal ganglion (Fig. 3B; see below).

The activation of the ecdysis circuitry is associated with the elevation of cGMP levels in a network of 27/704 neurons (Ewer et al., 1994; Ewer and Truman, 1997). Since PETH and ETH activate different phases of the ecdysis behavioural sequence (Žitňan et al., 1996, 1999), we examined the timing of elevation of cGMP levels in the CNS under natural conditions and after exposure to these peptides in vivo and in vitro.

Under natural conditions, we observed cGMP immunoreactivity in CNS ganglia 6–8 h before ecdysis, well before the release of peptides from the Inka cell (N=17). Notably, cGMP levels were elevated in the SG, TG1–3 and AG8, but not in AG1–7 (Fig. 4A). During the first 30 min of pre-ecdysis, cGMP staining in neurons and axons of the SG and TG1–3 clearly increased (N=9). At approximately 30 min into pre-ecdysis, cGMP staining appeared in the 27/704 neurons of AG1–7 (N=10). At approximately 45 min into pre-ecdysis and lasting until ecdysis onset, strong cGMP immunoreactivity was observed in the entire 27/704 network and weaker staining was detected in the abdominal neurosecretory cells L2,3 that produce diuretic peptides and kinins (N=16; Chen et al., 1994).

Pre-ecdysis I induced by PETH in pharate larvae (50 pmol; N=8) or in isolated nerve cords (100 nmol l⁻¹; N=7) was not associated with an elevation in cGMP level. However, application of 50 pmol of ETH in vivo (N=8) or of 100 nmol l⁻¹ ETH in vitro (N=7) induced strong cGMP immunoreactivity in all the neurons and axons described above during the first 10–15 min of pre-ecdysis (Fig. 4B). This strong immunoreactivity persisted until the onset of ecdysis (N=10). Similar staining occurred after ETH treatment of debrained nerve cords (N=10; Fig. 4C), but not in isolated abdomens (N=8) or in the isolated AG1–8 in vitro (N=7). These results show that the action of ETH on the SG is sufficient to induce a complete elevation of cGMP levels followed by ecdysis.

We also tested the cGMP response to ETH (50 pmol) injection in larvae that had had their abdomen ligated 10 min into pre-ecdysis. The CNS of these ligated abdomens was dissected out and processed for cGMP immunoreactivity just after the onset of ecdysis, which was considerably accelerated, as described below. In six out of 11 experiments, all the abdominal ganglia showed strong elevation of cGMP levels in 27/704 neurons and axons, but the remaining five abdominal nerve cords showed weaker staining confined to the cell bodies. These results indicate that the weaker elevation of cGMP levels in the abdominal ganglia may be sufficient for the activation and onset of ecdysis.

Injection of pharate larvae with ETH (50 pmol) leads to pre-ecdysis within 5–8 min, and ecdysis behaviour follows after a very consistent delay (41±2.6 min) (N=15; Fig. 5; Table 1). Despite this delay, we found that ETH actually activates the ecdysis neuronal network during early pre-ecdysis, but that the onset of edysis is delayed by inhibition from the cephalic and thoracic segments. Removal of inhibition by ablation of these segments greatly reduces the delay in the onset of edysis.

We demonstrated this by injecting larvae with ETH (50 pmol), then removing the head by neck ligation (applied between thoracic segments 1 and 2) at various intervals following the initiation of pre-ecdysis. When neck ligation was applied 30 min into pre-ecdysis, a premature switch to ecdysis occurred within 10–20 s in all larvae (N=10; Table 1). This manipulation thus reduced the delay to the onset of edysis by approximately 10 min, suggesting that inhibition emanates at least partly from the cephalic ganglia. If we applied neck ligation earlier, at 20–25 min into pre-ecdysis, the larvae immediately switched to ecdysis for approximately 1–4 min, but then reverted to weak pre-ecdysis-like movements in 11 out of 15 experiments. The remaining four larvae in this experimental group as well as eight additional animals neck-ligated 15 min into pre-ecdysis failed to continue in normal pre-ecdysis and did not initiate ecdysis (Table 1). However, if a second ligation (applied between abdominal segments 1 and 2) followed 5–15 min later, immediate and strong ecdysis movements occurred in the isolated abdomens. These results suggest that inhibition from the thoracic ganglia also contributes to the delay in the onset of ecdysis.

The role of thoracic inhibition was further examined by performing ligations between abdominal segments 1 and 2 at various times after the initiation of pre-ecdysis. Abdominal ligation 20–30 min into pre-ecdysis (N=11) caused a premature switch to ecdysis behaviour within 1–2 min (Table 1). Earlier abdominal ligations also caused premature ecdysis, but with some delay. Animals ligated 15 min into pre-ecdysis switched to ecdysis within 1–9 min (N=12), while ligation at 10 min into pre-ecdysis led to ecdysis within 5–14 min (N=19). These animals therefore performed pre-ecdysis for only 15–24 min (18±4.1 min) before the initiation of ecdysis contractions, which is much shorter than the approximately 40 min interval observed consistently in intact larvae injected with ETH (Fig. 5). Animals ligated 5 min into pre-ecdysis (N=10) invariably failed to perform ecdysis (Table 1). From these results, we conclude that injected ETH activates the ecdysis neuronal network in the abdominal ganglia during the first 10–15 min of pre-ecdysis, which correlates with the appearance of cGMP immunoreactivity in AG1–7 (Fig. 4B).

However, inhibition from both cephalic and thoracic segments delays the onset of edysis.

We also determined the delay from ecdysis activation to its onset during natural behaviour. Under our laboratory conditions, larvae initiated natural ecdysis 60–70 min after the initiation of pre-ecdysis (N=9; Table 1; Fig. 5). If abdominal ligation was applied 35–40 min into pre-ecdysis (N=8), a switch to ecdysis occurred within 1–16 min (Table 1; Fig. 5). Thus, these larvae switched to ecdysis 35–53 min into pre-ecdysis (40±7.5 min), which was approximately 15–30 min earlier than in intact control larvae. Ecdysis in all ligated larvae always started with complete proleg retraction and anterior peristaltic movements of abdominal segment 3, followed by strong ventral contractions and peristaltic movements of segments 2 and 1. Posterior segments (4–8) were subsequently recruited in this behaviour. Larvae ligated 15 and 30 min into pre-ecdysis failed to show ecdysis (N=9).

We examined the role of inhibition in delaying ETH-induced ecdysis using the isolated CNS. Similar to in vivo experiments, the isolated entire CNS exposed to ETH (100–150 nmol l⁻¹) switches to ecdysis 38–47 min after the initiation of pre-ecdysis (42.3±2.9 min; N=17). This delay was reduced (35.2±2.5 min) upon nerve cord transection 10–20 min into pre-ecdysis, applied either between AG1 and AG2 (N=18) or between each abdominal ganglion (N=13; Fig. 3B). If transections were performed 5 min into pre-ecdysis, all preparations failed to exhibit ecdysis (N=8). As in ligated larvae, transected abdominal nerve cords and individually isolated abdominal ganglia initiated ecdysis bursts in anterior ganglia (AG3,4) and within several minutes these bursts subsequently occurred in more posterior ganglia (AG5–8).

These experiments suggest that ETH activates an independent clock in each abdominal ganglion responsible for the initiation of ecdysis at the appropriate time even if these ganglia are individually isolated as early as 10 min into pre-ecdysis (Fig. 3B). Therefore, the neurons controlling ecdysis are activated by ETH approximately 10 min after the onset of pre-ecdysis, and the CNS is capable of the switch to ecdysis 5 min later. However, this transition is delayed by inhibitory input from the cephalic and thoracic ganglia. The effect of removing these ganglia in vitro was less pronounced than in vivo. This difference could be due to the lack of sensory input to the isolated CNS.

EH and CCAP have been implicated as downstream signals released upon ETH action on the CNS (Gammie and Truman, 1997, 1999). To clarify the respective roles of ETH, EH and CCAP in the regulation of ecdysis, we used the same concentration of these peptides (100 nmol l⁻¹) to compare their effects on the activation and onset of ecdysis in the desheathed entire CNS and AG1–8 in vitro. Since manual desheathing of the entire CNS caused spontaneous pre-ecdysis and ecdysis bursts in three cases, the activity of all nerve cords was observed for the first 10–15 min before application of peptide. Nerve cords displaying spontaneous pre-ecdysis bursting were discarded.

ETH invariably induced strong pre-ecdysis and ecdysis bursts in all desheathed CNS preparations (N=10). The latency to onset of these burst patterns was shorter than observed in the intact non-desheathed CNS; pre-ecdysis occurred in 1–6 min (4.2±1.8 min) followed by ecdysis bursts in 28–36 min (32.4±2.8 min). Characteristic ecdysis bursts were recorded in all ganglia tested (AG3–7), and strong elevation of cGMP levels was observed in the entire network of neurons and axons described above. However, in nerve cords consisting of only AG1–8, ETH failed to induce ecdysis bursts and elevation of cGMP levels during 60–90 min of recording (N=7).

Exposure of the entire desheathed CNS to EH induced an elevation in cGMP levels and ecdysis bursts, but these responses were quite variable and usually confined to the anterior ganglia (Figs 6, 7). Ecdysis bursts occurred in eight out of 11 experiments within 20–52 min of EH application (30.6±11.9 min). However, only three nerve cords showed these bursts in all abdominal ganglia (Fig. 6A), while five CNS preparations produced ecdysis only in AG3 or AG3,4 (Fig. 6B) and three CNS preparations completely failed to initiate ecdysis. In the three nerve cords showing ecdysis bursting in all abdominal ganglia, cGMP staining was strong in all neurons 27/704 and their axons (Fig. 7A). In other nerve cords, strong elevation in cGMP levels was observed in the SG and TG1–3, but weak or very weak responses were detected in the abdominal ganglia. Thus, despite exposure of the entire desheathed CNS to EH, strong cGMP staining was usually restricted to the anterior CNS.

Experiments with non-desheathed nerve cords confirmed previous observations (Gammie and Truman, 1999) that EH does not penetrate the neural sheath and fails to induce any activity in the intact CNS within 60 min of incubation (N=8). Therefore, we desheathed only certain anterior or posterior ganglia to determine whether the action of EH on desheathed ganglia induced an elevation in cGMP levels and ecdysis in intact ganglia. When desheathing was performed only in AG1–8, ecdysis occurred in six out of 11 experiments, but the onset of the behaviour was extremely variable (range 22–95 min; 53.6±25.5 min). Similar to experiments described above, only three nerve cords showed strong ecdysis bursts and cGMP staining in all abdominal ganglia. In the remaining three nerve cords, ecdysis was recorded only in AG1–4, and five CNS preparations failed to initiate ecdysis during 60–90 min of recording. These nerve cords showed only weak to very weak cGMP immunoreactivity in the abdominal ganglia. Surprisingly, all SG and TG1–3 displayed strong cGMP staining even if their neural sheath had not been removed (Fig. 7B).

If the brain, SG and TG1–3 were desheathed leaving AG1–8 intact (N=8), EH failed to activate ecdysis bursts during 60–90 min, but strong cGMP staining was observed in 27/704 neurons of all non-desheathed abdominal ganglia (Fig. 7B). Extended incubation of one of these nerve cords for approximately 2 h eventually resulted in ecdysis bursts in AG3,4. These experiments showed that the action of EH on desheathed ganglia may result in the activation of ecdysis in intact ganglia through an elevation in cGMP levels, but imply that this activation is indirect and probably requires some downstream signalling events.

Treatment of the desheathed abdominal nerve cord (AG1–8) with EH resulted in weak synchronous bursts similar to pre-ecdysis I within 1–5 min, followed by ecdysis bursts 10–73 min later (39.2±18.1 min) in 15 out of 18 experiments. However, only five of these nerve cords showed ecdysis bursts in all abdominal ganglia; in 10 nerve cords, ecdysis was recorded only in AG3 or AG3,4, while posterior ganglia showed pre-ecdysis-like synchronous bursts (Fig. 6B). In three cases, no ecdysis activity was observed after 80 min of exposure. Only five nerve cords showed strong cGMP staining in all abdominal ganglia; in the remaining experiments, the cGMP response was stronger in anterior ganglia (mostly in AG1,2) and weak or undetectable in the more posterior abdominal ganglia (Fig. 7C).

An earlier study suggested that central release of CCAP from abdominal neurons 27/704 provides an immediate signal for the initiation and performance of ecdysis behaviour (Gammie and Truman, 1997). However, we observed that CCAP treatment of the entire desheathed CNS did not induce typical ecdysis bursts. Instead, the abdominal ganglia of these nerve cords showed prolonged synchronous bursts or irregular bursting patterns (Fig. 8A). Subsequent removal of the brain, SG, TG1 and TG2 had no effect, but when the connectives between AG1 and AG2 were severed after 10–20 min of exposure to CCAP, typical ecdysis motor bursts occurred immediately in all the abdominal ganglia posterior to the cut in eight out of 10 experiments (Fig. 8A). Interestingly, in the nerve cord containing TG3 and AG1–8, CCAP induced ecdysis bursts in AG1–3, mixed pre-ecdysis-like and ecdysis bursts in AG4–6 and pre-ecdysis-like bursts in the AG7 (Fig. 8B). This indicates that inhibitory factor(s) in the SG-TG1–3 suppress the ecdysis circuitry even after the release of CCAP. Indeed, CCAP treatment of isolated AG2–8 induced strong and typical ecdysis bursts in 3–6 min (N=10). CCAP never induced an elevation in cGMP levels.

The initiation of each phase of the ecdysis behavioural sequence requires the action of PETH and ETH on specific central ganglia. Novicki and Weeks (1993, 1995) showed that the primary pacemaker for the pre-ecdysis I motor pattern resides in the TAG, but our experiments indicate that all abdominal ganglia respond to PETH exposure. Isolation of anterior abdominal ganglia from the TAG led to prolonged dorsoventral contractions in vivo or increased uncoordinated bursting activity in dorsal nerves of isolated abdominal ganglia in vitro. Therefore, motoneurons MN2,3 causing these contractions (Miles and Weeks, 1991; Novicki and Weeks, 1995) may be controlled in part by PETH-activated pre-motor interneurons in each abdominal ganglion. Nevertheless, the primary interneurons IN-402 residing in the TAG are necessary for the normal rhythm to occur (Novicki and Weeks, 1995). In contrast, each abdominal ganglion appears to contain the entire ensemble of ETH-sensitive circuitry necessary for the performance of pre-ecdysis II since individually isolated abdominal ganglia produce bursts in ventral nerves similar in duration and frequency to those in the intact nerve cord. We conclude that all abdominal ganglia respond to PETH and ETH and, therefore, presumably contain receptors for both peptides. Our data showed the necessary role of cephalic ganglia in the ETH-mediated activation of the ecdysis network. The importance of the brain for ecdysis was reported previously by Novicki and Weeks (1996), who found that most debrained animals were incapable of switching to ecdysis following injection of EH. More recent work has demonstrated that EH acts initially through the release of ETH from Inka cells, and that the resulting high levels of ETH in the haemolymph trigger ecdysis through direct action on the CNS (Žitňan et al., 1996; Ewer et al., 1997; Kingan et al., 1997). Our in vivo experiments show that the brain tritocerebral lobes and SG are required for ETH-induced ecdysis behaviour. The failure of most debrained larvae to initiate ecdysis could be explained by the activity of some peripheral factor, which may suppress the initiation of ecdysis in vivo. However, experiments on isolated, debrained nerve cords showed that the action of ETH on the SG is sufficient to induce ecdysis in virtually all cases. These data imply that ETH may trigger ecdysis by acting on downstream elements located in the SG in the absence of the EH-producing VM neurons. These hypothetical structures and factors remain to be identified.

We found that ETH (100 nmol l⁻¹) invariably induces strong cGMP staining in a subset of neurons in isolated intact or debrained nerve cords, and that this is coupled to the subsequent occurrence of ecdysis. In contrast, application of EH (100 nmol l⁻¹) to the desheathed CNS induces quite variable elevations in cGMP levels and ecdysis. Anterior ganglia usually show stronger staining than posterior ganglia, and no augmentation of these responses is obtained following exposure to even higher levels of EH (200–300 nmol l⁻¹; N=6; D. Žitňan, unpublished observations). Independent experiments showed that approximately the same concentrations of high-performance liquid chromatography (HPLC)-purified native EH (100 nmol l⁻¹) evoked ecdysis bursts only in seven out of 18 preparations and a similarly weak cGMP response in isolated desheathed abdominal nerve cords (Gammie and Truman, 1999). In contrast, we found that prolonged action of EH on desheathed ganglia induces an elevation in cGMP levels in intact ganglia. These data suggest that, although EH is able to induce limited cGMP production and ecdysis bursts, its action on CCAP neurons is not direct and probably requires some additional downstream signalling events.

We propose that ETH acts on multiple targets in the cephalic ganglia. In addition to possible effects on the VM cells, its primary targets are probably other neurons and structures in the tritocerebrum and SG, which induce a cGMP production and activate ecdysis neurons in all the abdominal ganglia.

We found that the cephalic and thoracic ganglia provide an inhibitory influence on the abdominal ecdysis network that delays the initiation of this behaviour in pharate larvae. Although our studies indicate that inhibitory input from the SG and TG1–3 probably delays the onset of edysis, the neurons responsible for this function remain to be identified. Cells 27/704 in the SG and in TG1–3 produce cGMP 6–8 h prior to the onset of the ecdysis behavioural sequence, but do not show CCAP immunoreactivity (Davis et al., 1993; Ewer et al., 1994). It is possible that these cells produce an inhibitory factor delaying the release of CCAP from 27/704 homologues in AG1–7, which show cGMP staining much later, after the initiation of pre-ecdysis. Interneurons 704 in the SG and TG1–3 are particularly good candidates for this role, since their axons project posteriorly throughout the entire nerve cord and make contact with each pair of 27/704 cells in the abdominal ganglia (Ewer et al., 1994; Klukas et al., 1996). This may be a mechanism whereby subsequent elevations in cGMP level first activate inhibitory neurons in the SG and TG1–3, which delays the release of CCAP from later-activated ecdysis neurons in AG1–7.

Our ligation experiments provided evidence that removal of the head and thorax during early pre-ecdysis greatly accelerates the onset of edysis. This effect was much more pronounced in vivo than in isolated nerve cords. We suggest that this difference may be explained by the absence of peripheral sensory input into the isolated nervous system, which may be an important signal for the switch from pre-ecdysis to ecdysis. Upon the attainment of critical internal pressures and complete separation of the old and new cuticles, the sensory input arising from the body wall and gut may relieve the inhibition and send stimulatory commands to the ecdysis network in the abdominal ganglia to release CCAP. Our experiments further showed that CCAP does not trigger typical ecdysis bursts in the entire desheathed CNS. Normal ecdysis activity in the abdominal ganglia occurs after the inhibitory ganglia have been removed by transection. Therefore, the switch from pre-ecdysis to ecdysis probably requires actions that regulate both the removal of inhibition in the cephalic and thoracic ganglia and the release of CCAP in all abdominal ganglia. This mechanism of transition to ecdysis is not limited to the larval stage, since decapitation accelerated the onset of eclosion in pharate adults of M. sexta (Ewer and Truman, 1997) and Drosophila melanogaster (Baker et al., 1999).

Isolation of individual abdominal ganglia 10–15 min into pre-ecdysis leads to the initiation of ecdysis approximately 20 min later. These data indicate that some independent endogenous clock within each abdominal ganglion regulates the timing of the onset of edysis. ETH and EH are apparently involved in the activation of the ecdysis network in the abdominal ganglia and perhaps in the stimulation of inhibitory neurons in the SG and TG1–3, but the mechanisms regulating the switch from pre-ecdysis to ecdysis are unknown.

We propose a model for activation of pre-ecdysis I, pre-ecdysis II and ecdysis by Inka cell and CNS peptides (Fig. 9). The ecdysis behavioural sequence is initiated by the release of low levels of EH from the proctodeal nerves into the haemolymph, which induce PETH and ETH secretion from Inka cells. PETH acts on all abdominal ganglia to activate the neural circuitry specific for pre-ecdysis I (Fig. 9). Several neurons controlling this behaviour have been identified: interneurons IN-402 located in the TAG control rhythmic bursting of paired motoneurons MN2,3 in all abdominal ganglia and corresponding dorsoventral contractions (Miles and Weeks, 1991; Novicki and Weeks, 1993, 1995). Pre-ecdysis II (Fig. 9) begins 15–20 min later as ETH activates additional neurons in each abdominal ganglion to produce strong posterioventral and proleg contractions. Each ganglion regulates this behaviour independently upon exposure to ETH. However, the function and identity of most of the neurons and all the neurotransmitters involved in the regulation of pre-ecdysis I and II remain to be determined.

Activation of the ecdysis network is accomplished by the action of ETH on the cephalic ganglia, which induces cGMP synthesis in CCAP-producing neurons 27/704 in AG1–7. The protocerebral VM neurons containing EH appear to increase the excitability of this network through an elevation in cGMP levels (Ewer et al., 1997; Gammie and Truman, 1999), but they are not necessary in this process. Our data showed that ETH requires the tritocerebrum and/or SG to activate the ecdysis circuitry during the early phases of pre-ecdysis. However, the onset of edysis is delayed by the inhibitory activity of a putative ecdysiostatic factor from the SG and TG1–3 (Fig. 9). This factor may be produced by 27/704 homologues in these ganglia, which do not produce CCAP and show elevated cGMP levels hours before the onset of the behaviour. The switch from pre-ecdysis to ecdysis is regulated by an independent process in each abdominal ganglion and consists of two steps: (i) the removal of inhibition from the SG and TG1–3; and (ii) the release of CCAP from abdominal neurons. Centrally released CCAP controls the initiation and performance of the ecdysis motor program (Gammie and Truman, 1997). This neuropeptide may modulate the activity of MN2,3, ISMM, PPR and APR motoneurons (Fig. 9), which control muscle contractions specific for ecdysis movements (Weeks and Truman, 1984a,b). Perhaps interneurons identified as excitors of the APR (Sandstrom and Weeks, 1991) may also be important for regulation of this motoneuron activity during ecdysis. After the old cuticle has been shed, sensory input probably reactivates inhibitory neurons to stop ecdysis movements.

The discovery and functional analysis of Inka cells resulted in the identification of two peptide hormones (PETH and ETH) that orchestrate the ecdysis behavioural sequence. Physiological experiments with these peptides led to precise definitions of pre-ecdysis I, II and ecdysis and to the localization of the regions of the CNS required for the activation and inhibition of these motor programs. The identification of ETH-related peptides and corresponding genes, which control the ecdysis behavioural sequence in Bombyx mori and Drosophila melanogaster (Adams and Žitňan, 1997; Park et al., 1999; D. Žitňan, L. Hollar, I. Žitňanová, P. Takác?, M. E. Adams, in preparation), indicates that the model described here may be applicable to other insects and perhaps to most arthropods.

Many stereotypical behavioural sequences occur in other invertebrate and vertebrate animals, e.g. egg-laying in molluscs, mating and courtship rituals in fishes and nest-building in birds (Camhi, 1984; Scheller et al., 1984). The extent to which peptide hormones are involved in the initiation and coordination of such behaviours is not completely understood. The experimental system described here could advance our understanding of the mechanistic principles governing general aspects of animal behaviour.

We thank Inka Žitňanová and Dr Peter Takác? for critical reading of the manuscript and helpful comments. This work was supported by grants from the National Institute of Health (AI 40555), the National Science Foundation (IBN 9514678) and Vedecká Grantová Agentúra (95/5305/800).