Activity Affects Dendritic Shape and Synapse Elimination during Steroid Controlled Dendritic Retraction in Manduca sexta

Introduction

The postembryonic acquisition of behavior requires alterations in neuronal circuitry, which comprise specific changes in neuronal structure, membrane properties, and synaptic connectivity. Insect metamorphosis is a compelling example for dendritic and synaptic remodeling as larval and adult behaviors place distinct demands on the CNS (Consoulas et al., 2000; Tissot and Stocker, 2000). Because metamorphosis is controlled by ecdysteroids [including 20-hydroxyecdysone (20HE)], holometabolous insects have proven useful model systems to analyze steroid effects on dendritic shape (Weeks and Truman, 1986; Truman, 1990; Levine et al., 1995, 1996; Levine and Weeks, 1996). Steroids also affect dendritic outgrowth in hypothalamic slice cultures (Toran-Allerand et al., 1999), cyclical alterations in dendritic spine density on rat hippocampal pyramidal cells during the estrous cycle (Woolley, 1998), and the maturation of the song-control system in birds (Bottjer and Arnold, 1997). Because dendritic shape is affected by activity in many systems (McAllister, 2000; Cline, 2001; Wong and Gosh, 2002; Libersat and Duch, 2004), a possible functional interplay between hormonal cues and neural activity seems logical, but only few studies have addressed this interaction (Woolley and McEwen, 1994; Woolley et al., 1997). An attractive hypothesis is that activity might serve as a modulator of rigid, nonflexible hormone actions.

During the metamorphosis of the moth, Manduca sexta, many larval motoneurons are remodeled to serve a new function in the adult (Consoulas et al., 2000). A surge in the 20HE titer during the prepupal peak (see Fig. 1C) triggers dendritic (see Fig. 1A) and axonal regression as demonstrated by hormonal manipulation experiments (see Fig. 1B) (Weeks and Truman, 1985, 1986; Levine and Weeks, 1989; Fahrbach and Weeks, 2002; Knittel and Kent, 2002). During the loss of larval behaviors, dendritic regression is accompanied by larval synapse elimination (Streichert and Weeks, 1995; Gray and Weeks, 2003). During pupal life, the formation of adult dendrites serves the integration into newly formed networks (Duch and Levine, 2000). Such changes in dendritic shape have been analyzed quantitatively throughout metamorphosis in the individually identifiable flight motoneuron, MN5 (Libersat and Duch, 2002).

During late larval life, structural changes of MN5 are accompanied by a defined sequence of motor behaviors. Five days before pupal ecdysis, wandering starts with the cessation of feeding (day W0). This is followed by crawling and burrowing at W1 and W2 and a phase of reduced movement that leads into a quiescent period underground at W3 and W4 (Reinecke et al., 1980; Dominick and Truman, 1984a,b). Pupal ecdysis occurs at W4 (Reinecke et al., 1980; Truman et al., 1980; Miles and Weeks, 1991). By means of chronically implanted extracellular electrodes at selected nerves in intact, freely moving, and normally developing animals, we demonstrated recently that the activity of MN5 is regulated in a stereotypical manner during the wandering stages (see Fig. 1C,D), and that stage-specific activity patterns induce axonal outgrowth (Duch and Mentel, 2003).

This study tested whether dendritic shape was also affected by neural activity. The results showed that steroid-induced dendritic regression and larval synapse elimination were modulated by increased neuronal activity.

Materials and Methods

Animals. Manduca sexta (L) obtained from a laboratory culture were reared on artificial diet (Bell and Joachim, 1976) under a long-day photoperiod regime (17/7 hr light/dark cycle) at ∼26°C. Both chronological and morphological criteria were used for staging of animals (Nijhout and Williams, 1974; Bell and Joachim, 1976; Dominick and Truman, 1984a). For the prepupal stages, W0 signifies the first day of wandering, and W1-W4 the remaining larval days (see Fig. 1C). Pupal ecdysis occurs at day W4. Dendritic regression of MN5 starts at W2 and is continuous until the third day of pupal life, stage P2 (Duch and Levine, 2000; Libersat and Duch, 2002). Chronic extracellular recordings can be conducted throughout the life of the fifth larval instar (V) until the first day of pupal life (P0), because the wires are lost during the pupal molt (see below). Therefore, this study examined animals during the last 3 d of larval life, W2 to P0. Stage W3 was defined by the eversion of the head revealing a characteristic triangle on the posterior edge of the head. Stage W3 lasted ∼24 hr and was subdivided by chronological criteria into three substages [W3 early (W3e), W3 mid (W3), and W3 late (W3l)] of equal duration. Depending on how long the development from stage W3 to W4 lasted, these substages were adjusted into three subsequent periods, each of which lasted between 6 and 10 hr. W3 never lasted <18 hr, and animals that remained in stage W3 for longer than 30 hr were regarded to be developmentally delayed and discarded. Stage W4 was defined by the appearance of the brown bar on the dorsal metathoracic segment and lasted ∼24 hr (Truman et al., 1980). Stage W4 was subdivided into stage W4 early (W4e), W4 mid (W4), and W4 late (W4l). W4e and W4 were defined as subsequent substages of equal duration covering the period between the appearance of the brown bar and the first sign of cuticular shrinkage on the first abdominal segment [anterior shrinkage (Truman et al., 1980); corresponding to W4a (Consoulas et al., 1996)] and lasted ∼8 hr each. W4l was the remaining time until pupal ecdysis occurred, lasted 4-8 hr, and corresponded to stage W4b as published by Consoulas et al. (1996). Motor activity was recorded continuously from animals that developed from W2e to W3l or W4e. Then, chronic extracellular stimulations were conducted during stage W4 or W4l. Animals were put on ice for further dissection 5 hr after the onset pupal ecdysis.

Chronic extracellular recordings. Bipolar hook electrodes for chronic implantations were manufactured from steel wires of 25 μm diameter insulated to the tip. Two wires were aligned parallel to each other and fixed on a coverslip by complete coverage with Kwik-Cast/Kwik-Sil two component silicon glue (World Precision Instruments, Sarasota, FL). Subsequently, the silicon and the insulation at the very tips were removed only on the upper side of the wires, and the half-blank tips were bent to form fine hooks of ∼100 μm length, with the blank part on the inside. The silicon cover of the electrode wires was 3 cm long, and the wires were left an additional 30 cm long to connect them to an amplifier (model 1700; A-M Systems, Carlsborg, WA). Extracellular signals were differentially amplified 1000-fold, digitized with an analog-to-digital converter (Digidata 1320A; Axon Instruments, Foster City, CA) and recorded continuously for several days at a sampling frequency of 10 kHz with pClamp8 software (Axon Instruments) on a personal computer. The continuous data trace was broken into subsequent periods of 1 hr to minimize file size (137 MB/hr).

Electrode implantations. Animals were anesthetized by chilling on ice for 3 hr. The following electrode implantations were conducted in the cold room at 4°C to minimize bleeding and to prevent movement. Animals were placed ventral side up in a large sylgard dish. The four anterior-most legs were held with lassos made from stainless steel wire, and the animal was slightly stretched to visualize mesothoracic nerve 1 from the outside. A large trachea served as landmark to record the same site of nerve 1 in different animals (Duch and Mentel, 2003). A fine cut of ∼200-300 μm length was made just above the nerve into the larval cuticle. Then, the nerve was lifted with a glass hook, and the electrode was placed under the nerve with a micromanipulator. Subsequently, the electrode was slightly lifted to remove surrounding hemolymph. The nerve was fixed to the hooks with a 1:3 mixture of Kwik-Cast/Kwik-Sil that hardened within 20 sec to completely insulate the hooks and the nerve. Then, the electrode was pushed back through the incision into the thorax, the wound was closed with tissue glue, and the silicon part of the electrode was glued to the lateral side of the animal toward the abdomen. Implanted animals were placed into cylindrical wooden chambers of 10 cm length and 2 cm diameter as in normal laboratory culture conditions. These were placed into a faraday cage to decrease electrical noise. Recordings were stable throughout the 3 d between stage W2 and larval-to-pupal ecdysis. Animals attempted complete ecdysis but were unable to shed the larval cuticle at the insertion site of the silicon electrode. Therefore, recordings were terminated after 2 hr of ecdysis behavior. Staging of animals by external criteria was conducted every 8 hr.

At the recording site, nerve 1 contained the axons of the motoneurons MN1-MN5, the larger diameter axon of a motoneuron with its soma located in the midline of the prothoracic ganglion and the axons of six smaller motoneurons, two of which were located in the prothoracic ganglion and four of which were located in the mesothoracic ganglion, and the very thin axons of an octopaminergic ventral unpaired median neuron located in the mesothoracic ganglion (Duch and Mentel, 2003).

Chronic stimulation of motoneurons. To determine the effects of spiking activity, stimulation of nerve 1 was conducted with an A-M Systems stimulus generator (model 2100). Individual pulses were set to 0.1 msec duration and 8 V amplitude for antidromic in vivo stimulation of the motoneurons MN1-MN5 during normal development. These pulse parameters induced no current spread and were also used in previous studies for antidromic identification of the motoneurons MN1-MN5 in the isolated ganglion preparation. By simultaneous intracellular recording in animals with chronically implanted extracellular electrodes, we have shown previously (Duch and Mentel, 2003) that each extracellularly applied stimulus resulted in one spike in MN5. Therefore, the stimulation protocol resembled the induced spiking activity in MN5. Stimulation amplitude and duration were just above threshold for inducing spikes in the large-diameter motoneuron axons, which makes it unlikely that other axons were also stimulated (Duch and Levine, 2000, 2002). To test the role of motoneuron activity for their structural remodeling nerve 1 was stimulated for 12 hr with bursts of 1 sec duration, 10 Hz frequency, and an interburst period of 10 sec. Individual pulses within each burst were of 0.1 msec duration and 8 V amplitude.

Intracellular identification of MN5. In all preparations, chronic recordings and stimulations were followed by intracellular recordings and staining of MN5. Animals were dissected carefully, dorsal side up, to prevent destruction of the nerve with the electrode implant. The mesothoracic ganglion was placed on a wax-coated platform, a suction electrode was placed on nerve 1 far more proximal than the site of the chronic electrode, and MN5 was impaled as described previously (Duch and Levine, 2000).

Intracellular recording and staining and synaptotagmin immunocytochemistry. A glass microelectrode (20-25 MΩ with dye solution) was filled with 7% neurobiotin in 2 m potassium acetate to fill the MN5 intracellularly by injection of 3 nA depolarizing current for 30 min. Before dye fills impaled motoneurons, MN5s were recorded in situ in bridge mode (Axoclamp 2B) from their somata to measure resting membrane potential, input resistance, and spontaneously occurring postsynaptic potentials (PSPs) resulting from spontaneous spiking of unidentified neurons presynaptic to MN5. The ganglia were fixed with 7°C cold 4% paraformaldehyde for 2 hr and rinsed for 1 hr in PBS. Then, dehydration in an ascending ethanol series (50, 70, 90, 100%; 15 min each) was followed by lipid extraction in three steps. First, an ethanol/methylsalicylate mixture (1:1) was applied for 10 min. This was followed by 10 min methylsalicylate and 10 min methylsalicylate/xylene (1:1) to increase antibody permeability and reduce background staining. Subsequently, these steps were conducted in reverse order. After rehydration in a descending ethanol series, ganglia were bathed in 0.5% Triton X-100 PBS for 2 hr. If immunolabeling was performed, the ganglia were preincubated in 10% normal goat serum, 0.5% Triton X-100 PBS (0.1 m) for 1 hr, incubated in primary antibody (rabbit anti-synaptotagmin) for 2 d, and washed in PBS for 1 hr. The synaptotagmin antibody was a gift from Dr. L. P. Tolbert (ARL Division of Neurobiology, University of Arizona, Tucson, AZ). It had been raised against the unique N terminus of Manduca sexta synaptotagmin and has been shown to specifically detect Manduca synaptotagmin (Dubuque et al., 2001). The neurobiotin staining was visualized by coupling to cyanine 3 (Cy3)-streptavidin (1:750 streptavidin-Cy3 in 0.05% Triton X-100 PBS). In case of previous treatment with primary antibodies, Cy5-coupled mouse anti-rabbit (Jackson ImmunoResearch, West Grove, PA) secondary antibody was coapplied in 1:200 concentration. After washing in PBS for 1 hr, the preparation was dehydrated, thereafter cleared in a mixture of ethanol and methylsalicylate (1:1) for 10 min, and mounted in 100% methylsalicylate. Acquisition of confocal image stacks was done immediately.

Confocal microscopy. Digital images were captured on a Leica (Bensheim, Germany) TCS SP2 scanning confocal microscope. All scans were conducted with a Leica HCX PL APO CS 40× oil immersion objective (numerical aperture, 1.25). Optimal physical resolution with the 40× oil lens used is 0.1 × 0.1 × 0.28 μm. At 3.5× zoom and a z-step increment of 0.3 μm, the resulting voxel resolution was 0.1 × 0.1 × 0.3 μm, thus matching the optimal physical resolution. However, optical resolution is also limited to approximately half excitation wavelength but may further be degraded to some degree by inhomogeneous tissue properties. All images were obtained in simultaneous acquisition mode (i.e., the emitted fluorescence light of both synchronously excited dyes was divided by an acoustic-optical beam splitter to be detected by two separate photomultipliers). Thus, no image misalignment occurred as a result of an error in scan mirror positioning. Images were further processed using Amira (TGS, San Diego, CA) and Corel Draw 8 software (Corel, Ottawa, Ontario, Canada). All original images shown are the projections of all optical planes of a given stack into one focal plane to visualize all dendrites of MN5 in one two-dimensional image.

Correction for chromatic aberration. To correct for chromatic aberration, a neuron was filled with biotin and double labeled with Cy3- and Cy5-coupled streptavidin (Jackson ImmunoResearch). An image stack was obtained as described above. The misalignment between both structures was measured and applied onto later acquired image stacks of equally treated specimens (Wouterlood et al., 1998).

Three-dimensional reconstructions of dendritic structure and morphometric analysis. To obtain quantitative data on dendritic morphology, the arborizations of MN5 were reconstructed from confocal image stacks. Confocal image stacks were acquired as described above and analyzed with Amira software. A semiautomatic tool to create three-dimensional reconstructions in the Amira environment with higher accuracy compared with conventional automatic or manual method was recently developed and described in detail by Schmitt et al. (2004). Briefly, a semiautomatic reconstruction procedure is realized by adapting the snake algorithm (Kass et al., 1988) to trace approximately tubular neuronal structures in confocal image stacks. User interaction is needed to define branch point hierarchy, whereby the algorithm automatically determines the midlines and diameters of interconnecting link segments at a user-predefined step size. This step size was set to 0.5 μm in all reconstructions conducted. The resulting tubular reconstruction of the neuronal tree contains precise information on length, diameter, orientation, and branching pattern. To access exact surface and volumetric measures, an adapted geodesic active contour algorithm (Caselles et al., 1997) is used to fit the idealized tubular reconstruction to actual image data, resulting in precise boundary definition.

Analysis of labeled presynaptic protein distribution along motoneuronal dendritic surfaces. The exact surface definitions obtained by fitting the tubular reconstructions to the actual image data with an adapted geodesic contour algorithm allow analysis of labeled molecule distribution along neuronal surfaces in submicron resolution (Schmitt et al., 2004). Briefly, all anti-synaptotagmin label image intensities more distant than 0.3 μm from the dendritic surface were set to zero. The remaining synaptotagmin label was located within 0.3 μm from the dendritic surface of MN5 and was used to calculate the mean staining density around each patch of the triangulated surface reconstruction. The calculated synaptotagmin staining intensities can be visualized as a color code by projecting each patch of the surface reconstruction onto the individual cylindrical compartments. This enables analysis of geometrical relationship between staining distribution and the morphology of the neuron. Warmer colors represent highest synaptotagmin staining intensity localized in the terminals of unidentified presynaptic neurons, which are located 0.3 μm or closer to the dendritic surface of the motoneuron (see Fig. 5). Because synaptic protein markers like synapsin 1 and synaptotagmin can be used to mark the sites of synapses in the insect CNS, and electron micrographs have shown that presynaptic vesicles are found in distances up to 0.2-0.3 μm from the postsynaptic membrane (Peters et al., 1985; Watson and Schurmann, 2002), this offers a method to assess putative synapse distribution throughout entire dendritic trees on the light microscopy level (Schmitt et al., 2004) (J. F. Evers and C. Duch, unpublished data).

Statistics. For paired statistical comparison of two independent groups, one-sided paired Student's t tests were conducted. All morphometric parameters obtained from pairs of stimulated and nonstimulated motoneurons, with each pair derived from the same animal were tested in this way. Physiological parameters like the amplitude and the frequency of postsynaptic potentials derived from two different experimental groups (stimulated and nonstimulated animals) were tested by one-sided unpaired Student's t tests. Differences were judged statistically significant at p ≤ 0.01. t tests were conducted with Microsoft (Seattle, WA) Excel. To test for statistically significant differences within more than two experimental groups, one-way ANOVA was conducted (significance level, p ≤ 0.01). If ANOVA revealed significant differences within the data, the individual experimental groups were compared by post hoc comparisons between means by the Newman-Keuls test (Zar, 1984) (significance level, p ≤ 0.05). ANOVA and post hoc Newman-Keuls tests were conducted with Statistica (StatSoft, Tusla, OK).

Solutions. Saline consisted of the following (in mm): 140 NaCl, 5 KCl, 4 CaCl₂, 28 d-glucose, 5 HEPES, pH was adjusted to 7.4 using 1 m NaOH.

Results

Discussion

During the metamorphosis of amphibians (Alley and Omerza, 1998) and holometabolous insects (Consoulas et al., 2000; Tissot and Stocker, 2000), reutilization of motoneurons for new adult behaviors is characterized by the retraction of axonal processes from degenerating larval muscles followed by outgrowth on the newly formed adult muscles and by retraction of larval dendrites followed by the maturation of the adult dendritic architecture. In holometabolous insects, dendritic regression is dependent on hormonal cues (Weeks et al., 1992; Weeks, 1999; Williams and Truman, 2004). Manduca proleg motoneurons are one of the rare systems in which it has been shown that dendritic retraction is accompanied by synapse elimination and, thus, the dismantling of larval motor circuits (Streichert and Weeks, 1995; Gray and Weeks, 2003). The present study supports this view by showing that regression of MN5 dendrites went along with a marked reduction in the number of putative input synapses, most likely resembling the dismantling of the larval crawling circuits.

During normal development, larval circuit dismantling is accompanied by a phase of reduced movement after the animals burrowed underground awaiting pupation (Reinecke et al., 1980; Dominick and Truman, 1994a). During this phase, the spiking activity of MN5 ceases significantly (Duch and Mentel, 2003). We used chronically implanted extracellular electrodes to test whether changes in the normal decrease in spiking activity in MN5 would interfere with dendritic regression and larval circuit dismantling. This approach had two important advantages. First, we were able to manipulate the activity of a subset of identified motoneurons during selected periods of development only, whereas all other neurons were unaffected and the animals developed normally. In genetic model systems, this is only possible by the expression of mutant alleles that cause hyperexcitability or hypoexcitability in subsets of neurons and can be controlled in a temporal manner (Rubin, 1988). Second, we could manipulate spiking activity on the basis of our knowledge of the alterations in activity during normal development (Duch and Mentel, 2003). The results clearly showed that stimulating MN5 during a developmental time when it usually showed only few spikes did not stop but significantly decreased dendritic regression and synapse elimination. This agreed with steroid control of dendritic regression (Weeks et al., 1992; Knittel and Kent, 2002), but it also clearly showed that activity significantly decreased both dendritic regression and synapse elimination. The possibility that motoneuron activity might, in turn, be controlled by steroid hormones or downstream peptide cascades is discussed previously (Duch and Mentel, 2003). However, it appeared unlikely that steroid control of dendritic regression might entirely be mediated via changes in activity, because stimulation experiments did not stop but reduced dendritic regression, even at higher stimulation frequencies (data not shown).

Dendritic regression and the effects of induced spiking activity were quantified by geometric reconstructions of the dendritic field of MN5. The effects of activity on synapse elimination were analyzed physiologically and anatomically. The results demonstrated that amplitude and frequency of spontaneously occurring PSPs, as recorded from the soma of MN5, were significantly increased in stimulated neurons. Because neither physiological parameters like spike shape and input resistance nor geometric parameters like dendritic diameters were affected by the stimulation, increased PSP amplitude most likely resembled increased synaptic input. This was further supported by anatomical results showing more immunocytochemically labeled synaptotagmin profiles in close proximity to the dendritic surface of the stimulated compared with the control MN5. Electron microscopy in other insects demonstrated that the distribution of synapsin I-protein and synaptotagmin-protein is useful in detecting synaptic loci at the light microscopic level (Watson and Schurmann, 2003). Paired with the precise surface definitions of our geometric neuron reconstructions, synaptotagmin profiles can be used to estimate the distribution of putative input synapses throughout the entire dendritic tree of MN5 on the light microscopy level (Schmitt et al., 2004). We do not claim that every synaptotagmin label within 0.3 μm of the dendritic surface resembles an input synapse, but rather that a close spatial association between the dendritic surface and synaptic proteins indicates a high likelihood for a synaptic connection. Corresponding to the physiological results, this “synaptic contact probability” was strongly increased in stimulated versus control motoneurons. This suggested that activity decreased steroid-induced synapse elimination in the CNS.

In vivo stimulation of nerve 1 led to antidromic action potentials in MN5 and few other efferent neurons (see Materials and Methods) (Duch and Mentel, 2003). Because most efferent insect neurons contain no, or only very few, central output synapses (Burrows et al., 1989; Watson and Schurmann, 2003) decreased dendritic regression in MN5 was most likely caused by intrinsic activity and not by synaptic activity, although we cannot fully exclude the possibility of increased synaptic activity in the motor networks resulting from the stimulation experiments. The mechanism by which increased activity mediated decreased dendritic regression remains to be investigated. Although at current speculative, it seems logical that common cell biological mechanisms might be co-opted by hormonal signals and activity-dependent programs (Eaton and Davis, 2003). Antidromically induced spiking activity in the developing MN5 induced calcium elevations in the soma and possibly also in large parts of the dendrites, depending on developmental changes in the ion channel distribution (Duch and Levine, 2002). Alterations in somatic calcium concentration can affect gene transcription (Hardingham et al., 1997; West et al., 2002), which in turn can affect dendritic shape. In cortical slice culture, dendritic shape is dependent on CaM kinase IV activity and the cAMP-responsive element-binding protein (Redmond et al., 2002). Therefore, activity and steroid hormone action might converge on the level of gene transcription.

The observation that dendritic regression was reduced, but no dendritic growth was induced, by activity contrasted with earlier results on stimulation-induced axonal outgrowth of MN5 (Duch and Mentel, 2003). This is not surprising because axons and dendrites clearly differ in their patterns of growth and in their molecular composition (Ramakers et al., 2001; Nikonenko et al., 2003; Tashiro et al., 2003). Similarly, hormonal manipulation studies have shown that axonal and dendritic regression occurred independently from each other in heterochronic mosaic animals (Knittel and Kent, 2002), suggesting a different set of signals for regressive events in the periphery nervous system and the CNS. These findings also concur with the events occurring during Wallerian degeneration in the mouse, where synaptic degeneration may be mediated by different mechanisms than axon and soma degeneration (Gillingwater and Ribchester, 2001).

The elimination of previously functional synaptic connections has been demonstrated in the vertebrate (Purves and Lichtman, 1980; Nguyen and Lichtman, 1996; Colman et al., 1997; Sanes and Lichtman, 1999; Chen and Regehr, 2000) and in the invertebrate CNS and peripheral nervous system (Sonea and Rheuben, 1992; Streichert and Weeks, 1995; Truman and Reiss, 1995; Consoulas and Levine, 1998; Duch et al., 2000; Eaton et al., 2002; Gray and Weeks, 2003). However, the mechanisms that regulate circuit dismantling are not well understood (Goda and Davis, 2003). Many studies in the CNS and the periphery nervous system support the conclusion that synapse elimination is mediated by activity-dependent competition mediated through the postsynaptic cell (Katz and Shatz, 1996; Sanes and Lichtman, 1999). The postsynaptic cell as intermediary is further supported by differential synapse elimination of a single motor neuron at the vertebrate neuromuscular junction, arguing for local control at each muscle fiber (Keller-Peck et al., 2001). For Manduca proleg motoneurons, it has been suggested recently that steroid-induced dendritic regression physically disconnected dendrites from the synaptic terminals of sensory neurons, producing synaptic weakening (Gray and Weeks, 2003). In contrast, this study suggests at least in part a differential effect of induced activity on dendrites versus synapses, because synapse elimination was decreased in all branch orders above 10, whereas dendritic regression was decreased only in high branch orders above 30. Although the mechanisms for either process remain speculative at present, the results strengthen the role of the postsynaptic cell for synapse elimination because alterations were induced by stimulating the postsynaptic neuron.

In wandering Manduca larvae, activity-dependent modulation of dendritic regression and synaptic weakening is most likely of behavioral importance. Despite the ongoing degeneration of the larval target muscle of MN5, neuromuscular transmission and muscle contraction remain functional until pupal life (Duch et al., 2000), and MN5 participates in crawling, burrowing, and ecdysis behavior (Duch and Mentel, 2003). In case of delayed wandering, motoneuron activity will remain higher compared with normal development. This increased activity will then most likely delay dendritic regression and synapse elimination so that the adequate performance of larval behavior will not be impaired too early by an inflexible hormone program. Activity-dependent modulation of steroid-induced developmental programs was also shown for programmed cell death (PCD). Motoneuron PCD could be delayed by ligating connectives in Manduca (Fahrbach and Truman, 1987) or by forcing emerging flies to show prolonged ecdysis behavior (Kimura and Truman, 1990).

In conclusion, our findings represent the first demonstration of activity-dependent modulation of steroid-induced dendritic regression and synapse elimination. Chronic extracellular recordings and stimulation make holometabolous insects suitable to further elucidate the concerted interplay between hormonal signals and activity-dependent mechanisms for structural and physiological changes of motoneurons.