Hormonal Control of Transmitter Plasticity in Insect Peptidergic Neurons: I. Steroid Regulation of the Decline in Cardioacceleratory Peptide 2 (Cap₂) Expression

Although the transmitter phenotype of most fully mature neurons remains invariant throughout their lifespan, a few have the ability to adjust their transmitter profile postembryonically. Transmitter plasticity in post-mitotic neurons has been best explored in sympathetic neurons, which are usually noradrenergic but, under certain conditions, can be induced to express a cholinergic phenotype (Patterson and Chun, 1977). For example, rat sympathetic neurons innervating peripheral sweat glands possess intense catecholamine histofluorescence and immunoreactivity for catecholamine synthetic enzymes shortly after they arrive at the developing sweat glands. These properties, diagnostic for catecholamines, slowly disappear as the gland matures and are replaced by increases in choline acetyltransferase activity and cholinergic transmission (Landis and Keefe, 1983; LeBlanc and Landis, 1986; Stevens and Landis, 1987; Schotzinger and Landis, 1988). Other mammalian neurons, including ciliary ganglion cells (Landis et al. 1987), superior cervical cells (Potter et al. 1986) and sympathoadrenal derivatives (Anderson et al. 1991), also exhibit transmitter plasticity postembryonically. Such plasticity is not limited to the mammalian nervous system: transmitter variations have also been reported in several invertebrate phyla, including coelenterates (Koizumi and Bode, 1986, 1991), molluscs (Gesser and Larson, 1985) and arthropods (Tublitz and Sylwester, 1990). Thus, there are several well-documented instances of transmitter-switching neurons.

Although phenotypic changes in transmitter expression have been well described, much less is known about the regulatory mechanisms that control this type of neuroplasticity. The most detailed work to date has been performed on the adrenergic-to-cholinergic switch by sympathetic neurons that innervate sweat glands. Cultured sympathetic neurons treated with sweat gland extracts undergo many of the changes seen in vivo, including the induction of choline acetyltransferase activity and a reduction of catecholaminergic properties (Rao and Landis, 1990). Recent biochemical analyses of sweat gland homogenates have isolated a heat-labile, trypsin-sensitive diffusible factor that triggers cholinergic differentiation in cultured sympathetic cells (Rohrer, 1992). Other soluble factors, including cholinergic differentiation factor (CDF; Yamamori et al. 1990), membrane-associated neurotrophic substance (MANS; Wong and Kessler, 1987) and ciliary neurotrophic factor (CNTF; Lin et al. 1989), have also been shown to induce phenotypic changes in transmitter expression in sympathetic neurons.

The purpose of this study was to elucidate the mechanisms controlling transmitter expression in one set of transmitter-switching neurons. We have previously shown that the central nervous system (CNS) of the tobacco hawkmoth, Manduca sexta (L.), contains a set of four identified lateral neurosecretory neurons (LNCs) that alter their transmitter profile in vivo during postembryonic development (Tublitz and Sylwester, 1990). The primary transmitter of the LNCs during embryonic and larval stages is cardioacceleratory peptide 2 (CAP₂), a group of at least three related myoregulatory neuropeptides that modulate gut activity (Broadie et al. 1990; Tublitz et al. 1991, 1992a; Cheung et al. 1992). During metamorphosis, these cells switch their transmitter phenotype: CAP₂ content decreases markedly to less than 2% of its original level and the LNCs begin to produce a second neuropeptide, bursicon, which is responsible for tanning the insect cuticle after each molt (Reynolds, 1983). By the final molt from pupa to adult, the LNCs express high levels of bursicon with no detectible CAP₂ bioactivity. In this report, we focus on the regulatory mechanisms underlying the first part of the transmitter switch by the LNCs, the CAP ₂ decline, and show that the decrease in CAP₂ is under the influence of the insect steroid hormone 20-hydroxyecdysone (20-HE). This work has previously been reported in abstract form (Tublitz and Loi, 1990; Loi and Tublitz, 1991).

Tobacco hornworms were raised in the Manduca sexta rearing facility at the University of Oregon and were staged as previously described (Tublitz, 1989). Briefly, Manduca sexta larvae were raised on an artificial diet modified only slightly from that of Bell and Joachim (1976). Shortly after hatching, larvae were placed in individual plastic cups, where they remained until wandering, the period during the fifth instar that marks the beginning of metamorphosis. At the onset of wandering, larvae were transferred to individual chambers in wooden blocks for the remainder of the fifth instar. Animals were removed from the blocks 2 days after the larval–pupal molt and placed in marked plastic bags for the duration of adult development. Larvae and pupae were housed in a controlled environmental chamber using a 17 h:7 h L:D regime with a superimposed thermal period (27°C during the light period, 25°C during the dark period) to improve developmental synchrony. Prior to emergence, these adults (called pharate adults) were placed in a large enclosed chamber (50% relative humidity) for breeding purposes.

Staging of larvae and pupae was based on previously described Manduca sexta staging series (Truman and Riddiford, 1974; Weeks et al. 1992), which used developmental events as markers. These include ecdysis from the fourth to the fifth instar (day 0, D₀), the onset of wandering behavior (wandering day 0, W₀) and larval–pupal ecdysis (pupal day 0, P₀). To stage animals more precisely throughout the fifth instar (8 days) and the first 10 days of adult development, animals were staged daily. Unless noted otherwise, we only used gate I animals, those that wandered during the fourth night of the fifth instar (Dominick and Truman, 1984).

LNCs were isolated using a modification of a method described elsewhere (Tublitz and Sylwester, 1990). Briefly, the fourth abdominal ganglion (A4) was removed from animals at various stages, pinned in a small Sylgard-containing Petri dish and immersed in paraffin oil. Each cluster of four LNCs (see inset in Fig. 1) was removed from the ganglion with glass needles using dark-field optics to visualize the cells (Rowell, 1976). Previous work has shown that all four LNCs are identical in terms of peptide content (Tublitz and Sylwester, 1990). Isolated LNC clusters were individually transferred into a small volume of distilled water (2–5 μl) to promote cell lysis (Berlind and Maddrell, 1979). The LNC-containing water droplet was then divided into two equal samples, both of which were frozen at -20°C until needed, usually a period of not more than 7 days. Each cluster was assayed for both bursicon and CAP₂ bioactivities unless otherwise noted. This paper focuses on the CAP₂ results.

CAP₂ levels were quantitatively assayed on an isolated pharate adult Manduca sexta heart according to a previously published protocol (Tublitz and Truman, 1985a). Briefly, a piece of the abdominal heart was attached to a force transducer and superfused with standard physiological saline. Previous studies have determined that heart rate is directly proportional to CAP₂ concentration (Tublitz and Truman, 1985a–d). Each frozen sample (half of a cluster) was thawed and resuspended in 95 μl of physiological saline immediately prior to assaying for cardioacceleratory activity. Data are expressed in terms of CAP units, where 1 unit is equal to the total amount of CAP bioactivity (CAP₁ and CAP₂ combined) measured in the ventral nerve cord of a single pharate adult moth. The limit of resolution of this bioassay is 0.02 CAP units.

The remaining sample from each cluster was assayed for bursicon bioactivity using the adult wing bioassay (Taghert and Truman, 1982a,b; Tublitz and Sylwester, 1990). After thawing, enough saline was added to each sample to bring the final volume up to 10 μl, which was then injected directly into the lumen of the costal vein of the mesothoracic wing that had been removed from a pre-emergent pharate adult. The contralateral wing from the same individual received a control injection of 10 μl of saline. Wing pairs were incubated for 3h at room temperature in a moist environment to prevent dehydration. Following the incubation period, each wing was descaled with paraffin wax and 70 % ethanol to visualize the wing veins. Tanning of the injected wing veins is proportional to bursicon concentration and bursicon activity was determined by visually comparing the tanning level of the control and test injections using an arbitrary scale of 1–5 as previously described (Taghert and Truman, 1982a,b; Tublitz and Sylwester, 1990). Bursicon levels are expressed in terms of bursicon units, where 1 unit is equal to the quantity of bursicon bioactivity in the ventral nerve cord of a pharate adult. The wing bioassay can detect as little as 0.005 bursicon units.

The source of the ecdysteroids in Manduca sexta is the prothoracic glands, located in the first thoracic segment. To prevent the ecdysone-secreting prothoracic glands from affecting the LNCs in the abdominal CNS, animals were ligated with suture thread at the thoracic–abdominal juncture following a 10–30min anesthetization period in iced water. In those cases where suture thread proved unsuitable, animals were also clamped with small hemostats. After ligation, the anterior end of the animal was removed and discarded. Ligated animals were returned to the colony rearing chamber until further manipulation. Ligated animals died over a period, with approximately 95% remaining alive and healthy after 1 week and about 95% of these still living 14 days after ligation. The health of the animals was determined by measuring cuticle coloration and hydration state; those deemed unhealthy were discarded. Ligations were performed early in the day, prior to any changes in LNC peptide levels.

For the single-ganglion implantation experiments, fourth abdominal ganglia (A4) were removed from day 2 fifth-instar caterpillars (D₂ is the second full day after ecdysis from the fourth larval instar) and prepared for implantation by rinsing in sterile physiological saline. Prior to implantation, host animals were anesthetized in ice and/or CO₂ and the tip of the dorsal ‘horn’ was cut off. Every host received a single rinsed ganglion, which was injected with a 1ml sterile syringe through polyethylene tubing (PE 160) into the body hemocoel via the opening in the dorsal horn. Following implantation, the opening in the horn was ligated with 6-0 surgical suture thread, and after post-operative recovery the animal was returned to the colony rearing facility. Hosts were allowed to develop until the equivalent of pupal day 8 (‘P₈’), at which time both implanted and host A4 ganglia were removed. Identification and proper orientation of implanted ganglia on ‘P₈’ were facilitated by tying 6-0 sutures of differing lengths around both anterior and posterior connectives prior to implantation. LNCs from hosts and implanted ganglia were separately dissected, processed and assayed for CAP bioactivity as described above.

The CNS implantation experiments were carried out using the same protocol as the ganglion implantation except that the entire CNS from a day 2 fifth-instar caterpillar was isolated and injected through the dorsal horn into another D₂ larva. Some hosts received a brainless CNS, i.e. a CNS with all ventral ganglia except a supraesophageal ganglion. Hosts were killed on wandering day 1 (W₁) and the A4 ganglion was removed from both the host CNS and the implanted CNS. Dissection of LNCs and bioassay for CAP activity were identical to those for the ganglion implantation experiments.

As a means of blocking the metamorphic effects of the commitment pulse (CP) of 20-HE that occurs in fifth-instar larvae (see Fig. 1A), animals were treated with methoprene, a juvenile hormone agonist (Levine et al. 1986). Methoprene (7S isomer no. 312008, a gift of Dr D. Cerf, Sandoz Crop Protection, Palo Alto, CA) was dissolved in cyclohexane (1mgml⁻¹) and 25 μl was topically applied twice onto the cuticle of feeding larvae, once on day 2 at 17:00h and again on day 3 at 09:00h. A control population of identically staged animals was treated with cyclohexane only. After treatment, control and experimental animals were returned to our animal rearing facility where they remained until the controls reached pupal day 8 (P₈). When controls had developed to P₈, each cluster of four LNCs from ganglion A4 was removed from controls and experimental animals and assayed for bursicon and CAP₂ levels using the wing and heart bioassays, respectively.

Infusion of 20-HE was carried out using a minor variation of the protocol of Weeks et al. (1992). Following 1h of chilling on ice to anesthetize the animal, day 2 fifth-instar larvae were ligated and implanted with an intact day 2 CNS including the brain. These animals were allowed to recover from the implantation and infused 24h later with 20-HE, which was delivered by syringe pump into the hemolymph via small-bore polyethylene tubing (PE 10) inserted into the dorsal horn (Weeks, 1987). 20-HE, a gift of Dr Janis Weeks, was hydrated in a physiological saline containing no carbohydrates (Ephrussi and Beadle, 1936) at a concentration of 1mgml⁻¹. After spectrophotometric analysis to confirm the concentration (Meltzer, 1971), the 20-HE stock solution was frozen in 50 μl samples and stored at -20°C until needed. 20-HE was infused into animals at a rate of 3.0 μlh⁻¹(3.0 μgh⁻¹) for 6h, a concentration designed to mimic the CP (Dominick and Truman, 1985). Following infusion, the polyethylene tubing was clamped or melted closed and left attached to the horn. Infused animals remained in our animal colony for an additional 2 days, after which the LNCs from ganglion A4 were removed from the host and from the implanted CNS and assayed separately for CAP₂ bioactivity using the isolated heart bioassay.

Previous work has demonstrated that the LNCs undergo a transmitter switch between the last larval instar and midway through adult development (Tublitz and Sylwester, 1990), but the time course of this change was unknown. To identify more precisely when the conversion from CAP₂ to bursicon occurred, single LNC clusters were removed from ganglion A4 using standard techniques (Tublitz and Sylwester, 1990). Every A4 ganglion contains two identical sets of four LNCs whose cell bodies are bilaterally situated on the posterolateral margin of the ganglion and both sets were used for all the experiments described here (Fig. 1, inset). Individual clusters are easily identifiable in situ under a dissecting microscope by the position and opalescent reflectiveness of the LNC somata (Taghert and Truman, 1982a,b; Tublitz and Sylwester, 1990). After dissection, the contents of each LNC cluster were divided into two equal samples, one of which was assayed for CAP bioactivity on an in vitro pharate adult heart (Tublitz and Truman, 1985a–c), while the remaining sample was analyzed for bursicon activity on an isolated adult wing bioassay (Taghert and Truman, 1982a,b; Reynolds, 1983; Tublitz and Sylwester, 1990). Both bioassays are very specific and highly sensitive; the myogenic heart responds in a dose-dependent fashion to CAP₂ application by increasing beat frequency, while samples containing bursicon induce a dose-dependent premature tanning of wing veins when injected into wings isolated from moths just prior to the pupal–adult molt.

Our results, depicted in Fig. 1B, indicate that the level of CAP₂ remains relatively unchanged during days 0–3 (D₀–D₃) of fifth (final) instar larvae, followed by a precipitous drop on wandering day 0 (W₀). W₀ is the day when the caterpillar stops feeding and commences wandering behavior, a period of sustained locomotory activity in preparation for pupation (Dominick and Truman, 1986). The present data show that the LNCs lose approximately 50% of their total CAP₂ content during wandering (Fig. 1B). The drop in CAP₂ levels at W₀ is followed by a second, slower CAP₂ decline during the remainder of larval life, so that, by the second day following pupation (pupal day 2, P₂), CAP₂ activity is below the limit of detection (Fig. 1B). In contrast, bursicon levels in the LNCs are minimal throughout most of the final larval instar, and it is not until the second day of wandering (W₂) that bursicon activity first becomes measurable in physiologically significant amounts (Fig. 1C). From W₂ until P₁₀, bursicon levels in the LNCs continue to increase so that, by P₁₀, bursicon activity in these cells is over 50-fold higher than in larvae.

The temporal association between the timing of the CAP₂-to-bursicon transmitter switch and the onset of metamorphosis suggested that the LNC transmitter switch might be regulated by one or more of the hormones involved in the control of molting in insects. One likely set of candidates are the ecdysteroids, ecdysone and 20-HE; the titers of these two substances in the blood show fluctuations during the last larval instar that are temporally correlated with changes in transmitter type in the LNCs. Ecdysone is synthesized and secreted from the prothoracic glands located in the first thoracic segment. After its release into the blood, it is converted into the biologically more active 20-HE form. There are two, distinct, physiologically important ecdysteroid pulses in fifth-instar Manduca sexta (Bollenbacher et al. 1981; Hoffman, 1983; Riddiford, 1985): the first, known as the commitment pulse (CP), occurs early on D₃ and serves to commit larval tissues to pupal differentiation as well as triggering wandering behavior (Fig. 1A). Actual expression of the pupal phenotype occurs only in response to the second larger ecdysteroid rise, the prepupal peak (PP), which commences on the day after wandering, W₁. To ascertain whether either of these ecdysteroid pulses was responsible for the decline in CAP₂ expression, ecdysteroid levels were experimentally manipulated in several ways.

First, we exposed the LNCs to different portions of the fifth-instar ecdysteroid titer. One simple method of accomplishing this in vivo was to ligate animals at the thoracic–abdominal juncture. This procedure physically isolated the abdominal LNCs from the ecdysteroid-secreting prothoracic glands, located in the first thoracic segment, since the portion of the animal anterior to the ligation, i.e. the head and thorax, was discarded. About 90% of abdomens ligated in this fashion remained alive and healthy for 2–3 weeks. Ligations were performed on D₂, W₀ or P₀ in order to subject LNCs to no ecdysteroids (D₂ ligations), to the CP alone (W₀ ligations) or to the normal two-peak sequence (P₀ ligations; Fig. 2A). For these experiments, ligated animals were allowed to develop until the equivalent of P₈, the stage in normal development when the LNCs acquire maximal bursicon activity. For example, abdomens isolated on D₂ were held for 14 days, those isolated on W₀ for 12 days and those isolated on P₀ for 8 days in order that they could be dissected on the day that would normally correspond to the eighth day of the pupal stage in intact animals. On ‘P₈’, each cluster of four LNCs from ligated animals was removed from ganglion A4 and assayed for CAP₂ content using the isolated heart bioassay. We found that the normal decline in CAP₂ was prevented only by the earliest ligation, that is, on D₂ prior to the CP (Fig. 2B). LNCs removed from animals ligated on W₀ or P₀ exhibited the normal CAP₂ decline (Fig. 2B).

To test whether the ecdysteroids directly regulate the drop in CAP₂, individual ganglia were implanted into host caterpillars at different developmental stages and assayed for CAP₂ content when the hosts reached P₈ (Fig. 3A). A4 ganglia were removed from D₂ larvae, the stage prior to the commitment and prepupal peaks. These ganglia were then individually implanted into the abdominal hemocoel of D₂, W₀ or W₃ caterpillars. These stages were chosen because they exposed the implanted ganglion to different ecdysteroid regimens; e.g. implanted ganglia were exposed to both ecdysteroid peaks (D₂ hosts), the PP alone (W₀ hosts) or neither peak (W₃ hosts). When hosts reached pupal day 8 (P₈), both host and implanted A4 ganglia were removed and the LNCs were dissected. Each cluster was subsequently bioassayed for CAP₂ content. LNCs removed from implanted ganglia never showed a decline in CAP₂ levels regardless of the type of ecdysteroid exposure (Fig. 3B). The normal drop in CAP₂ levels was seen, however, in LNCs from host animals whose nervous systems were intact.

Another approach to studying 20-HE effects is to manipulate the titers of another hormone, juvenile hormone (JH). JH, a sesquiterpenoid, is an essential hormonal factor in the larval-to-pupal transformation in Manduca sexta because its absence during the CP is the crucial signal that triggers the onset of metamorphosis. If JH levels remain elevated during the CP, fifth-instar caterpillars will molt into supernumerary sixth-instar larvae, not into pupae (Riddiford, 1985). To explore further the relationship between the CP and transmitter expression in the LNCs, animals were treated with 7S-methoprene, a pharmacological juvenile hormone agonist (Levine et al. 1986). As detailed in the Materials and methods section, experimental animals were exposed to 7S-methoprene dissolved in cyclohexane late in the afternoon of D₂ and again in the morning of D₃ to ensure elevated hemolymph levels of this agonist throughout the duration of the endogenous CP (Fig. 4A). Controls were treated with a equivalent volume of cyclohexane. After the final treatment, both experimental and control animals were returned to our animal colony where they remained for 12 days, at which time the controls had reached P₈ and the experimentals were 3–5 days into a supernumerary sixth-instar larva. Analysis of CAP₂ and bursicon levels in LNC clusters from the methoprene-treated animals revealed that these cells were predominantly CAP-containing, whereas the same cells removed from the controls primarily expressed bursicon with CAP₂ levels being below the limit of detection (Fig. 4B). Thus, treatment with methoprene blocked the normal transmitter changes from occurring in the LNCs.

The extent of the nervous system required for the drop in CAP₂ levels in the LNCs was tested by removing the complete CNS, i.e. brain+ventral nerve cord, from a D₂ fifth-instar larva and implanting it in its entirety into another D₂ caterpillar (Fig. 5A). The procedure for this experiment was the same as described above for the ganglion implantation experiments. In this case, the host larvae contained two complete central nervous systems each with its own set of LNCs. A parallel set of D₂ animals received D₂ CNSs without brains, i.e. containing all neural components between the subesophageal and terminal abdominal ganglia. Following surgery and recovery, host animals were maintained in standard rearing conditions until the day after wandering, W₁, the stage in untreated controls when CAP levels have already undergone a major decline (Fig. 1). Individual LNC clusters were removed from the A4 ganglion of the host and the implanted CNS and separately assayed for CAP content. Animals that received an intact CNS exhibited a normal CAP₂ decline in both implanted and host LNCs (Fig. 5B). In contrast, a drop in CAP₂ was not seen in the implanted LNCs dissected from a brainless CNS, although the LNCs of their hosts did undergo the normal CAP₂ decrease. These data indicate that the decline in CAP₂ in the LNCs at wandering occurs only if the LNCs are able to receive neural information from the brain.

To examine more directly the role of the CP in down-regulating LNC CAP₂ levels, we devised an experimental protocol that enabled us to control exogenous 20-HE levels. The experimental design is depicted in Fig. 6A. D₂ caterpillars were ligated and the isolated abdomens were implanted with a complete D₂ CNS, including the brain, using the procedures described in Materials and methods. Each of these preparations consisted of an isolated abdomen containing the host’s abdominal CNS (A1–A8) plus a complete CNS from the donor. Implantation of an intact CNS was necessary because the ganglion and CNS implantation results showed that intact neural connections between the LNCs and the brain were required for the CAP₂ decline. The day following implantation, some of the ligated and implanted abdomens received a 6h infusion of 20-HE at a concentration designed to approximate closely the CP (refer to Materials and methods for infusion and dosage details). Infused and uninfused abdomens were maintained until the equivalent of W₀, at which time ganglion A4 LNCs were isolated from both the implanted and the host CNS and were assayed separately for CAP₂ content. 20-HE infusion induced a CAP₂ decline in implanted LNCs that was equivalent to the normal CAP₂ drop measured in untreated controls (Fig. 6B). In contrast, CAP ₂ levels in LNCs from the host CNS, when exposed to an identical 20-HE treatment, were unaffected. CAP₂ levels also remained unaltered, i.e. high, in those preparations that were ligated but not infused with 20-HE (Fig. 6B). Hence, exogenous 20-HE induced a major reduction in CAP₂ levels in the LNCs, but only in preparations containing a complete CNS.

The LNCs of the tobacco hawkmoth, Manduca sexta, because of their location, size and accessibility, have been a model system for investigations into the neurobiology of peptide-containing neurons (Taghert and Truman, 1982a,b; Tublitz and Sylwester, 1988, 1990). One outcome of these studies has been the discovery of the unexpected transmitter plasticity exhibited by these neurons. The initial work on these cells by Taghert and Truman (1982a,b) unequivocally showed that the LNCs in late larvae and adults contained bursicon, the classic insect tanning hormone (Reynolds, 1983). That these cells in larvae also expressed CAP₂ was first demonstrated immunocytochemically (Tublitz and Sylwester, 1988) and later confirmed biochemically (Tublitz and Sylwester, 1990). This latter report also showed that bursicon levels were low in LNCs removed from larvae and high in LNCs dissected from adults. It was concluded from these experiments that the LNCs underwent a qualitative alteration in transmitter expression from CAP₂ to bursicon during metamorphosis.

The current study shows that the transmitter switch by the LNCs from CAP₂ to bursicon is a multi-step process commencing with a decline in CAP₂ at the onset of metamorphosis (Fig. 1B). CAP₂ levels in the LNCs drop precipitously on W₀, with the LNCs losing more than 50% of their accumulated CAP₂. The timing of this transmitter depletion is temporally correlated with wandering behavior, a complex series of motor activities, including increased locomotion, which in nature enables the caterpillar to crawl off its host plant and burrow underground. One of the activities associated with wandering is ‘gut-emptying’ behavior, when the alimentary canal is completely cleared of undigested materials in preparation for metamorphosis. Recent work has demonstrated that ‘gut-emptying’ is accomplished in part by a CAP₂-mediated increase in gut activity (Edwards et al. 1990; Tublitz et al. 1992a,b). Because the LNCs in larvae are strongly CAP-immunopositive (Tublitz et al. 1991, 1992a), it has recently been proposed that the LNCs are the source of the CAP₂ released at wandering. Our measurements of the time course of the drop in the LNC content of CAP₂ (Fig. 1B) provide important additional support for the hypothesis that the LNCs are the major source of CAP₂ at wandering.

The primary goal of this study was to identify the mechanisms responsible for the reduction in CAP₂ expression in the LNCs. The data presented in this paper implicate the CP of 20-HE as the causal factor in this process. The results of the ligation experiments (Fig. 2) and the endocrine manipulations using a juvenile hormone agonist (Fig. 4) demonstrate the existence of a temporal relationship between the CP and the CAP₂ decline in the LNCs. Although this evidence by itself is not conclusive, it does indicate that manipulation of the endocrine milieu, specifically the CP, is sufficient to alter significantly the normal progression of transmitter expression in the LNCs. A more definitive test of the hypothesis that the CP triggers the CAP₂ drop is furnished by the 20-HE infusion experiments (Fig. 6). The results of these experiments demonstrate that 20-HE, infused at levels equivalent to that of the endogenous CP, is capable of reducing CAP₂ levels in the LNCs. These data, taken together, provide strong evidence that the CP is the proximate cause of the CAP₂ release by the LNCs at wandering.

The nature of the role played by the CP was also explored. In particular, we were interested in determining whether 20-HE acted directly on the LNCs. On the basis of the work presented here, it is unlikely that all the effects of the CP on the LNCs can be accounted for by a direct steroid action on the LNCs. This conclusion is based in part on the results of the D₂-to-D₂ ganglion implantation experiments (Fig. 3). A decline in CAP₂ levels was detected only in the host LNCs and not in the LNCs removed from the implanted CNS, although both host and implanted LNCs were exposed to the endogenous CP (Fig. 3B). The observation that the LNCs from the host CNS exhibited a normal decline in CAP₂ suggested that intact neural connections were required in order for the CP to be effective, and this hypothesis was tested by the CNS implantation experiments. The results from those experiments unequivocally demonstrated that the normal decline in CAP₂ in the LNCs occurred only if the LNCs were in direct neural contact with the brain (Fig. 5).

On the basis of these data, we propose the following model illustrated in Fig. 7: the larval brain in Manduca sexta contains a set of steroid-activated neurons that drive a population of interganglionic descending interneurons. These descending interneurons project posteriorly down the entire length of the nerve cord and synapse directly onto the LNCs in each ganglion. Activation of the brain neurons by the CP stimulates the descending interneurons which, in turn, excite the LNCs and cause them to release CAP₂ at wandering. The proposed model is attractive because of its parsimony, because all its neuronal elements are known to be present in Manduca sexta and because it is testable. The existence of CP-sensitive neurons in the insect brain affecting neuronal activity in the ventral nerve cord was first postulated by Dominick and Truman (1986) during their investigations into the hormonal control of wandering behavior in Manduca sexta larvae. They proposed a model in which the CP initiates wandering behavior by suppressing the activity of a set of pre-motor brain neurons, which, in the absence of 20-HE, normally inhibit the central neurons that generate the wandering motor program. These neurons described by Dominick and Truman are potential presynaptic elements in the LNC circuit, not only because of their steroid sensitivity but also because they are activated at the same time as the LNCs. Our model also postulates the existence of a second class of neurons, a set of descending interganglionic neurons with processes in the Manduca sexta ventral nerve cord. The presence of such neurons in Manduca sexta has recently been documented using a combination of intracellular dye-filling and confocal microscopy (Mesce and Klukas, 1991). Although the connectivity between individual descending interganglionic interneurons and the LNCs is not yet known, it is plausible that some of these descending neurons are presynaptic to the LNCs, since stimulation of the interganglionic connectives in adult moths elicits a barrage of excitatory postsynaptic potentials in the LNCs and other CAP-containing neurons (N. J. Tublitz, unpublished results). Although we favor this model for the reasons stated above, it should be noted that other hypothetical models can be constructed to fit the existing data. In one alternative model, 20-HE acts directly on the LNCs in conjunction with the descending input from the brain. Studies are currently under way to distinguish between these two models.

In this paper, we have demonstrated that the initial step in the CAP₂-to-bursicon transmitter switch by the LNCs, the decline in CAP₂ levels, is hormonally regulated by the CP of the insect steroid hormone 20-HE. The developmental decline in CAP activity in these cells appears to be a two-step process: a release of CAP₂ followed by suppression of its further synthesis. Our data show the steroid dependence of the initial peptide release step but provide no evidence about the relationship between 20-HE and the down-regulation of CAP synthesis. It would not be surprising, however, if CAP synthesis were suppressed by 20-HE since steroid-dependence of peptide expression has been documented in several neuroendocrine systems (Alexander and Miller, 1982; Gross et al. 1987). Steroid-induced effects on transmitter phenotype have also been described in embryonic neurons during embryogenesis in invertebrates (Witten and Truman, 1991a,b) and in vertebrates (Anderson, 1989; Anderson and Axel, 1986). This study provides evidence that steroids control transmitter expression in mature, functional neurons. Whether the molecular mechanisms underlying steroid-dependent transmitter plasticity are the same for developing and mature neurons must await future studies.

We are grateful to Dr Janis Weeks for her advice and for the use of her equipment for some of the experiments described here. We also thank Drs Weeks and J. W. Truman for their permission to modify and use their ecdysteroid figure in this paper. This manuscript was greatly improved by the thoughtful comments of Clement Cheung, Barbara Gordon-Lickey and Terry Takahashi. This work is supported by grants from NIH, NSF, the Alfred P. Sloan Foundation and the Medical Research Foundation of Oregon.