Steroid hormones alter dendritic architecture in many animals, but the exact relationship between dendritic anatomy, synaptic strength, and behavioral expression is typically unknown. In larvae of the mothManduca sexta, the tip of each abdominal proleg (locomotory appendage) bears an array of mechanosensory hairs, each innervated by a planta hair sensory neuron (PH-SN). In the CNS, PH-SN axons make monosynaptic, excitatory nicotinic cholinergic connections with accessory planta retractor (APR) motoneurons. These synapses mediate a proleg withdrawal reflex behavior that is lost at pupation. The prepupal peak of ecdysteroids (molting hormones) triggers the regression of APR dendrites and a >80% reduction in the amplitude of EPSPs produced in APRs by PH-SNs that innervate posterior planta hairs. The present study tested the hypothesis that a decrease in the number of synaptic contacts from PH-SNs to APRs contributes to this synaptic weakening. Pairs of PH-SNs and APRs were fluorescently labeled in larvae and pupae, and the number of indistinguishably close anatomical contacts (putative synapses) was counted by confocal laser scanning microscopy. During APR dendritic regression, the mean number of contacts from posterior PH-SNs decreased by ∼80%, whereas the size of individual contacts did not change detectably and the axonal arbors of PH-SNs did not regress. These results suggest that the steroid-induced regression of motoneuron dendrites physically disconnects the motoneurons from the synaptic terminals of sensory neurons, producing synaptic weakening and the developmental loss of the proleg withdrawal reflex behavior at pupation.
Steroid hormones can profoundly alter the size and complexity of dendritic arbors (for review, seeGarcı́a-Segura et al., 1994; Weeks and McEwen, 1997; McEwen, 2000). For example, in female rats, the rise and fall of estradiol during the estrous cycle regulates the cyclical increase and decrease in dendritic spine density on hippocampal pyramidal cells (for review, see Woolley, 1998). In male rats, androgen levels regulate the size of the dendritic arbors of motoneurons in the spinal nucleus of the bulbocavernosus (Kurz et al., 1986). Natural or experimentally induced changes in steroid hormone levels that alter dendritic architecture often simultaneously alter an animal's behavior; e.g., high estradiol levels at proestrous promote lordosis behavior in females, whereas circulating androgens support copulatory behavior in male rats. It is logical to assume that steroid-induced changes in dendritic arbors are accompanied by changes in synaptic connectivity (Matsumoto et al., 1988; Yankova et al., 2001) that contribute to changes in behavior. However, direct demonstration of this presumed chain of events in any one system is lacking.
During metamorphosis of the moth Manduca sexta, steroid hormones [ecdysteroids, including 20-hydroxyecdysone (20E)] regulate dramatic changes in neurons and behavior (for review, seeLevine and Weeks, 1989; Fahrbach and Weeks, 2002). A “prepupal peak” of 20E at the end of larval life triggers dendritic regression in motoneurons (see Fig. 1 A) (Weeks et al., 1992), which are subsequently respecified for new functions (accompanied by dendritic regrowth) or eliminated by cell death during the pupal stage (Weeks and Truman, 1985; Weeks and Ernst-Utzschneider, 1989; Weeks et al., 1992; Kent and Levine, 1993; Duch and Levine, 2000). Pupation also involves the loss of many larval behaviors, including the proleg withdrawal reflex (Weeks and Jacobs, 1987). This reflex is triggered by the deflection of the mechanosensory hairs located on the tip of each abdominal proleg (locomotory appendage) (see Fig. 1 B) (Wiel and Weeks, 1996). Each hair is innervated by a single planta hair sensory neuron (PH-SN) (Peterson and Weeks, 1988), which makes monosynaptic, excitatory nicotinic cholinergic connections with the accessory planta retractor (APR) motoneuron and related motoneurons. The proleg withdrawal reflex is mediated by these direct sensorimotor synapses plus indirect interneuronal pathways (see Fig. 1 C) (for review, see Weeks et al., 1997).
Streichert and Weeks (1995) showed that the mean amplitude of EPSPs evoked in APRs by PH-SNs decreases significantly (by ∼40–80%, depending on PH-SN location) when APR dendrites regress (see Fig.1 D), although APR input resistance increases. They also provided indirect evidence that the size of PH-SN arbors does not change during this time. These findings suggested that decreased EPSP amplitude was associated with the physical disconnection of APR dendrites from the synaptic terminals of PH-SNs. The present study supported this hypothesis by demonstrating a developmental decrease in the number of close anatomical appositions between PH-SNs and APRs. These findings suggest that steroid-induced dendritic regression plays a key role in synaptic weakening and the elimination of a behavior during postembryonic life. Some results have been published previously in abstract form (Gray and Weeks, 1999).
Materials and Methods
Animals. M. sexta were reared individually on an artificial diet [modified from Bell and Joachim (1976)] under a 17/7 hr light/dark photoperiod and a 27/25°C temperature cycle. Insects were staged using standard developmental markers (see Fig. 1 A). The day of ecdysis (cuticle shedding) to the fifth (final) larval instar is designated day L0, followed by days L1, L2, and L3. The onset of metamorphosis is marked by wandering behavior (day W0), when larvae cease feeding and seek a pupation site. The day of ecdysis to the pupal stage is designated day P0. All comparisons in this study were made between insects on day L3 and day P0.
Fluorescence labeling of sensory and motoneuron pairs. The PH array on each proleg tip is divisible into anterior, middle, and posterior regions. The axon of each PH-SN projects into the ganglion of the same abdominal segment, where the axon terminals form a somatotopic map based on the position of the PH in the array (Peterson and Weeks, 1988). The terminals of posterior PH-SNs remain within the ganglion of the same body segment, whereas anterior and middle PH-SNs send axons into neighboring ganglia and arborize there as well (Peterson and Weeks, 1988; Streichert and Weeks, 1995). We used posterior PH-SNs in the present study because the restriction of their terminals to a single ganglion facilitated complete staining of the entire axonal arbor (see below).
To stain an individual posterior PH-SN, insects were selected on day L2, anesthetized by chilling on ice for 45 min, placed ventral side up in a Petri dish, and immobilized with strips of dental wax. The dish was filled with chilled water to a level just above the spiracles (respiratory openings) to maintain anesthesia. A well of petroleum jelly was formed around the tip of a proleg in abdominal segment 3 (A3) and filled with distilled water (dH2O). A single PH from the most ventral (distal), posterior region of the PH array was cut at its base using iridectomy scissors. After 10 min, the dH2O in the well was replaced with a solution of ∼1% tetramethylrhodamine dextran (molecular weight, 3000; Molecular Probes, Eugene, OR; henceforth termed rhodamine) in dH2O. The water covering the spiracles was removed, and the immobilized larva was held at 4°C for 36–48 hr to reversibly arrest development and movement. The dye and petroleum jelly were then removed, and the larva was returned to standard rearing conditions to continue development.
Animals with a stained posterior PH-SN were selected on either day L3 or day P0. Day L3 larvae were identified by their size and the presence of “frosted frass,” signaling the occurrence of the commitment pulse of 20E (Fig.1 A) (Nijhout and Williams, 1974). Day P0 pupae were used for experiments within 1 hr of completing ecdysis. Insects were sometimes held at 4°C for up to 24 hr to arrest development, before dissection. Animals were anesthetized on ice and opened dorsally under saline (see below), and ganglion A3 was removed with the lateral branch of the ventral nerve (VNL) left long on the side ipsilateral to the stained PH-SN. VNL carries the axons of the two APRs that innervate the ipsilateral retractor muscle and receive monosynaptic EPSPs from ipsilateral PH-SNs (Sandstrom and Weeks, 1996). Ganglia were placed ventral side up in a Sylgard-coated dish containing physiological saline composed of (in mm): 140 NaCl, 5 KCl, 4 CaCl2, 28 glucose, and 5 HEPES, pH 7.4 (Trimmer and Weeks, 1989); they were then desheathed with fine forceps (Weeks and Jacobs, 1987) and observed under a Wild M3C dissecting microscope (Leica Microsystems, Bannockburn, IL) using substage dark-field illumination to visualize APR cell bodies.
The cell body of an APR ipsilateral to the stained PH-SN was impaled with a borosilicate microelectrode filled at the tip with 4% Lucifer yellow CH (Molecular Probes) and backfilled with 0.5m LiCl2 (electrode resistance, ∼30 MΩ). APR was identified by its characteristic cell-body position and time-locked extracellular action potential recorded in the ipsilateral VNL by means of a suction electrode led to a differential, alternating-current-coupled preamplifier (Sandstrom and Weeks, 1996). Lucifer yellow was iontophoresed into APR using 5–10 nA hyperpolarizing current pulses (500 msec, 1 Hz) generated by a Grass 88 stimulator (Grass Instruments, West Warwick, RI) connected to a Getting M5 intracellular amplifier (Getting Instruments, San Diego, CA). APRs were filled for at least 20 min or until action potentials were no longer detected. The electrode was withdrawn, VNL was crushed near the ganglion, and the preparation was left at room temperature for 30–45 min in the dark to permit dye diffusion. The ganglion was then fixed overnight in 4% paraformaldehyde at 4°C, rinsed in PBS for 10 min, mounted on a poly-l-lysine-coated coverslip, dehydrated in an ascending series of ethyl alcohol, cleared in xylene, and mounted ventral side up in DPX mounting medium (Electron Microscopy Sciences, Fort Washington, PA). Slides were stored in the dark at room temperature until observed, usually within 2–4 d.
Confocal imaging. The quality of double fills was first assessed under epifluorescence using a Leitz (Wetzlar, Germany) Laborlux D compound microscope. If staining of the PH-SN and the APR appeared to be successful, the ganglion was then examined on a Zeiss(Oberkochen, Germany) LSM 310 confocal laser scanning microscope. Rhodamine fluorescence (red) was excited with a 568 nm laser line and detected through a 590 nm long-pass filter. Lucifer yellow fluorescence (green) was excited with a 488 nm laser line and detected through a 515–565 nm bandpass filter. To visualize anatomical contacts, ganglia were scanned using a 63× oil immersion objective [numerical aperture (NA), 1.25]; at this magnification, the entirex,y extent of the PH-SN arbor was contained within the field of view (see Fig. 4). Four scans of 1024 × 1024 pixels were averaged for each optical section, which were taken at 0.5 μm intervals. This method produced images with a voxel resolution of 0.2 × 0.2 × 0.9 μm in the x-, y-, and z-planes, respectively. For two-dimensional projections of optical sections, the total depth of tissue sampled was computed as [(number of intervals × 0.5 μm) + 0.9 μm]. To ensure that scans in the two channels were in register, each optical section was scanned in the rhodamine channel and the Lucifer yellow channel before moving to the next optical section. False color was used to identify the rhodamine (red) and Lucifer yellow (green) channels. The scanned volume contained the entire axonal arbor of the PH-SN and the portion of APR dendritic arbor that was coextensive with the PH-SN arbor. Accordingly, all potential sites of contact between a PH-SN and APR were imaged. A subset of putative synaptic contacts (see Fig. 5) was examined at higher magnification (63× oil immersion objective and 3× hard magnification in the software).
The APR dendritic field spans the entire dorsoventral extent of the ganglion (Weeks and Jacobs, 1987) and extends more anteriorly than does the PH-SN arbor (see Fig. 7 A). We did not image the entire APR dendritic arbor in double fills, because only the portion of the arbor that contacts PH-SNs was relevant to this study, and photobleaching of Lucifer yellow became a factor during the prolonged scanning required to visualize the entire APR arbor. In a few cases, the entire APR arbor was scanned using a 40× water immersion objective (NA, 1.2), which provided a larger field of view (see Fig.2).
Identification of putative synaptic contacts. Only ganglia in which the arbors of both the PH-SN and the APR were completely and brightly stained in confocal images were used to count anatomical appositions. Staining was determined to be complete when adjacent segments of the processes of a neuron were continuous within each optical section and between adjacent sections, up to the point of termination of the process; i.e., there were no “gaps.” Staining quality was also readily assessed by comparison with images of the same neurons stained with cobalt chloride (Weeks and Jacobs, 1987; Peterson and Weeks, 1988; Streichert and Weeks, 1995; Sandstrom and Weeks, 1996). Successful staining of APRs was routine, but only a low proportion of PH-SNs were completely and brightly stained (Melville et al., 2003). Accordingly, the number of ganglia containing an acceptably stained pair of neurons was small (n = 4 on day L3;n = 3 on day P0; see Results). Examination of less well stained material (data not shown) supported the conclusions based on ganglia used for quantitative analysis.
The series of optical sections from each ganglion was imported into Scion Image (Scion Corporation, Frederick, MD). The PH-SN (red) and APR (green) channels from each optical plane were combined into a single frame, and all frames from a ganglion were combined into a stack. Individual frames from each stack were displayed on a monitor and scored by eye for sites at which (1) pixels in the red and green channels overlapped (in which case the pixels appeared yellow) or (2) pixels in the red and green channels abutted with no discernible gap between (see Fig. 5) (Lamotte d'Incamps et al., 1998). We refer to sites of indistinguishably close anatomical contact in single frames as appositions. Three independent observers, who were kept unaware of the identity of the material, marked each apposition in each frame of each stack, without referring to adjacent frames. An apposition was included in the final data set when it was scored by at least two of the three observers; ∼90% of appositions were identified by all three observers. A small number of appositions (∼8%) were eliminated from the final data set when reconstruction of the PH-SN arbor (see below) revealed that these sites lay outside of the PH-SN arbor, resulting from errant pixels in the rhodamine channel. Because of the narrow optical section thickness, appositions were often present in a similarx,y-location in consecutive frames. Counts of appositions were converted to counts of putative synaptic contacts (see Results) by eliminating duplicate counts of the same anatomical contact that was imaged in multiple consecutive frames. Thus, each putative synaptic contact was made up of one or more appositions scored in adjacent frames.
PH-SN morphometrics. The red channel was used to generate a stack of the PH-SN axonal arbor, which was reconstructed in three dimensions using a National Institutes of Health tree-tracing macro written by Dr. D. K. Hartline (University of Hawaii, Honolulu, HI;www.pbrc.hawaii. edu/∼danh/Resources/treetracedoc.html). Total branch length and the number of branch points was measured for each reconstructed PH-SN. Reconstructions were then recombined with thex-, y-, and z-coordinates of the scored contacts to determine the distribution of putative synaptic contacts on the PH-SN arbor.
Statistical analysis. All statistical comparisons were performed using SigmaStat 2.0 statistical software (Jandel Scientific, San Rafael, CA). Differences were considered to be significant whenp < 0.05.
APR regresses while PH-SN axonal arbors remain stable
Previous quantitative studies demonstrated that the APR dendritic arbors in segments A3–A6 regress significantly between days L3 and P0 (Streichert and Weeks 1995; Sandstrom and Weeks, 1998); for these measurements, APRs were stained with cobalt chloride, the complete dendritic arbor was drawn as a two-dimensional projection, and the density of neuronal processes was calculated. In the present experiments, APRs were stained by the intracellular injection of Lucifer yellow. For technical reasons (see Materials and Methods), in ganglia that contained both a stained APR and a stained PH-SN, only the portion of the arbor of the APR that overlapped with PH-SN axon terminals was imaged. Therefore, we did not obtain quantitative measurements of the APR dendritic extent that could be compared directly with previous data. However, to verify that dendritic regression is apparent in Lucifer yellow-stained motoneurons, the complete arbor of the APR was imaged in ganglia that were not used to count appositions with PH-SNs. Similar to observations in cobalt-stained material (Weeks and Jacobs, 1987; Weeks and Ernst-Utzschneider, 1989; Streichert and Weeks, 1995; Sandstrom and Weeks, 1998), the neurites of APR were densely arrayed with fine, high-order processes that resembled dendritic spines (Fig.2 A,B). In other insect neurons, these processes typically bear input synapses (Peters et al., 1985) (see Discussion). The characteristic loss of dendrites between days L3 and P0 was readily apparent in Lucifer yellow-stained APRs (Fig. 2 A,B); most major neurites (except those in posterior neuropil) were retained, whereas the density of smaller, higher-order branches decreased. Likewise, in ganglia used to count appositions, in which only the portion of the arbor of APR located within sensory neuropil was imaged, the dendritic extent of APR was markedly reduced on day P0 (see Figs. 4, 7). These observations confirm that APRs undergo substantial dendritic loss between days L3 and P0 (Jacobs and Weeks, 1990; Streichert and Weeks, 1995; Sandstrom and Weeks, 1998).
Streichert and Weeks (1995) provided evidence that the axonal arbors of PH-SNs do not regress between days L3 and P0, but they did not quantify this observation specifically for the posterior PH-SNs used in the present study. Therefore, from three-dimensional reconstructions of individual rhodamine-stained, posterior PH-SNs (Fig. 2 C,D), we determined the mean number of branch points and total branch length of the central axonal arbors on days L3 and P0. Figure3 shows that neither measure changed significantly during this period of development. As described previously in cobalt-stained material (Peterson and Weeks, 1988), the central axonal arbors of larval and pupal PH-SNs stained with rhodamine were characterized by relatively smooth lengths of neurite punctuated by varicosities (Fig. 2 C,D, see also Figs.4, 5, 7). In other insect neurons, similar varicosities are presynaptic boutons (Peters et al., 1985). The lack of axonal regression in PH-SNs is expected, given that these neurons are not respecified for new roles during metamorphosis (see Discussion).
These findings support the hypothesis that the axonal arbors of PH-SNs remain stable between days L3 and P0 while the dendritic arbors of their postsynaptic targets, the APR motoneurons, regress.
Identification of putative synaptic contacts between APRs and PH-SNs
We tested the hypothesis that the developmental weakening of the synapses from posterior PH-SNs to APRs is associated with a decrease in the number of synaptic contacts between the presynaptic and postsynaptic neurons. From each ganglion that contained a pair of completely and brightly stained neurons, we acquired a series of confocal optical sections at 0.5 μm intervals that spanned the entire volume of the PH-SN axonal arbor. Thus, all sites of potential anatomical contact between the PH-SN and APR were imaged. The depth of tissue sampled to image the complete PH-SN arbor was 38 ± 4 μm (mean ± SEM) on day L3 (n = 4) and 50 ± 8 μm on day P0 (n = 3).
In each optical plane, the images acquired in the red (PH-SN) and green (APR) channels were combined into a single frame. The frames were scored by multiple observers blind to the identity of the material for appositions in which pixels in the red channel overlapped with, or directly abutted, pixels in the green channel (see Materials and Methods). Some anatomical appositions appeared in multiple adjacent frames (Figs. 4, 5), so the counts of appositions were converted to counts of putative synaptic contacts by eliminating duplicate counts of the same anatomical apposition that appeared in more than one frame (see Materials and Methods). We use the term putative to refer to synaptic contacts because it is not possible to confirm the presence of active zones or functional synaptic transmission based on confocal microscopy (see Discussion). Similar methods have been used to identify putative synaptic contacts in other preparations (e.g., axo-axonic contacts of Ib fibers in cat spinal cord) (Lamotte d'Incamps et al., 1998).
Figure 4 A,B shows representative confocal images from days L3 and P0, depicting the same volume of sensory neuropil (3.4 μm depth) at both developmental stages. Arrows mark putative synaptic contacts between the axonal arbor of a posterior PH-SN and the APR dendritic arbor. More putative synaptic contacts were apparent on day L3 than on day P0 (see quantitative comparisons below). Putative synaptic contacts between PH-SNs and APRs are shown at higher magnification in Figure 5. Figure 5 A shows seven consecutive frames spanning the contact marked by an asterisk in Figure4 A; appositions were scored in four consecutive frames, indicated by overlap of the red and green channels. Figure5 B shows a single frame from a different preparation illustrating a PH-SN varicosity (red) and an APR dendrite (green) that abutted with negligible overlap.
The number of putative synaptic contacts decreases during the larval–pupal transformation, but contact size does not change
Figure 6 A compares the number of putative synaptic contacts between posterior PH-SNs and APRs on days L3 and P0. The mean number of putative contacts decreased significantly from 42 to 9 over this period of development, representing a 79% decrease. Concomitantly, the mean amplitude of EPSPs produced in APRs by posterior PH-SNs is ∼0.30 mV on day L3 and ∼0.05 mV on day P0 (Fig. 1 D) (Streichert and Weeks, 1995) (see Discussion).
Synaptic strength can be influenced by both the number of contacts between presynaptic and postsynaptic neurons and the size of each synaptic contact; larger contacts may contain more active zones and release more neurotransmitter (Grantyn et al., 1984). Having demonstrated a developmental decrease in the number of putative synaptic contacts between PH-SNs and APRs (Fig. 6 A), we tested whether the size of the contacts changed. Optical section thickness was constant (0.5 μm), so the number of adjacent frames in which a putative synaptic contact appeared provided a rough estimate of the size of the contact. The number of adjacent frames in which individual contacts appeared ranged from 1 to 7, with a mean of 1.74 ± 0.12 (SEM) frames on day L3 (n = 167) and 1.90 ± 0.30 frames on day P0 (n = 28). These values did not differ significantly (Student's t test). Conversion of these figures to physical dimensions provides an estimate of the mean size of putative synaptic contacts (measured along thez-axis) of 1.25–1.35 μm (range, 0.5–4.4 μm). These values are similar to the sizes of contacts observed in thex,y-plane (Fig. 5) and appropriate for insect central synapses (see Discussion).
We also compared the proportion of contacts that fell into different size categories at the two developmental stages. Figure6 B shows that the distribution of contact size was similar on days L3 and P0; most contacts appeared in only one or two adjacent frames, and >90% of the contacts appeared in three or fewer frames. The proportion of putative synaptic contacts in each size category did not differ significantly on days L3 and P0 (Fig.6 B). Thus, the number of putative synaptic contacts between PH-SNs and APRs decreased between days L3 and P0, but the size of the contacts did not change detectably.
We did not image the complete APR dendritic arbor in ganglia containing a stained PH-SN (see above), so we did not reconstruct the three-dimensional structure of the motoneuron and determine the spatial distribution of putative synaptic contacts from PH-SNs. Qualitatively, most putative synaptic contacts from posterior PH-SNs occurred on the high-order branches of APR and not on large neuritic trunks (Figs. 4,5). However, because we obtained a complete image of each PH-SN arbor, it was possible to reconstruct the sensory neuron arbors in three dimensions and to map the locations of putative synaptic contacts with APR. Figure 7 A shows representative reconstructions of posterior PH-SNs, displayed as two-dimensional projections, with the portion of the APR dendritic arbor that occupied the same volume of neuropil superimposed. The size of the APR dendritic arbor that was coextensive with the PH-SN axonal arbor was substantially reduced on day P0. Figure 7 B shows the same PH-SN reconstructions as in Figure 7 A, with putative synaptic contacts with APR marked. As documented above (Fig.3), the size of PH-SN axonal arbors appeared similar on days L3 and P0. In this example, there were 43 putative synaptic contacts from the PH-SN to APR on day L3 and four putative synaptic contacts from the PH-SN to APR on day P0. The loss of APR dendrites in posterior neuropil (Fig. 2 B) was reflected in the dearth of putative synaptic contacts in the posterior region of the PH-SN arbor on day P0.
Previous studies demonstrated that the steroid-induced regression of the dendritic arbor of motoneuron APR is associated with a decrease in the amplitude of monosynaptic EPSPs from PH-SNs and the developmental loss of a proleg withdrawal reflex (Jacobs and Weeks, 1990; Streichert and Weeks, 1995). The present study used double fluorescence confocal microscopy to test the hypothesis that physical disconnection of APR dendrites from the synaptic terminals of PH-SNs contributes to the synaptic weakening and behavioral loss. The dendritic morphology of APRs stained with Lucifer yellow on days L3 and P0 (Figs. 2, 4) was indistinguishable from that of APRs stained with cobalt at these stages (Fig. 1 A) (Streichert and Weeks, 1995; Sandstrom and Weeks, 1998). In particular, the region of APR dendritic arbor that overlaps with PH-SN axon terminals was regressed (Fig. 7 A) (Jacobs and Weeks, 1990). The present study (Fig. 3; also see Figs. 2 C,D, 7 B) improved on previous indirect evidence that PH-SN axonal arbors do not regress at pupation (Streichert and Weeks, 1995). Thus, fluorescently stained material viewed by confocal microscopy corroborated previous observations that PH-SN arbors are structurally stable while the dendrites of proleg motoneurons regress at pupation.
Staining APRs with Lucifer yellow was routine, whereas staining PH-SNs with rhodamine via the hair shaft had a low success rate that limited the number of double fills in which contacts could be counted. Only brightly and completely stained neurons were used for quantitative analysis (see Materials and Methods). The mean number of putative synaptic contacts between a posterior PH-SN and an APR decreased from 43 on day L3 to nine on day P0 (Fig. 6 A), whereas the size of individual contacts did not change detectably (Fig.6 B). The 79% decrease in the number of putative synaptic contacts parallels an 83% decrease in the mean amplitude of the EPSPs at these synapses (Streichert and Weeks, 1995). Because the input resistance of APR increases by 40% during this period (Streichert and Weeks, 1995), the actual decrease in synaptic current may exceed 83%. Furthermore, the loss of posterior dendrites in APR (Figs. 1 A, 2 B,7 A) should preferentially eliminate PH-SN inputs that are electrotonically the most distant from the cell body of the APR (Fig. 7) and that make a lesser contribution to EPSP amplitude. Nevertheless, EPSP amplitude is severely reduced on day P0. The correspondence between the developmental decrease in the number of putative synaptic contacts and EPSP amplitude supports the idea that these two features are related.
For technical reasons (see Materials and Methods) we used only posterior PH-SNs. However, because EPSP amplitude decreases significantly for PH-SNs located in all regions of the array (Streichert and Weeks, 1995), we expect that the contact number likewise decreases for all PH-SNs.
Validity of the analysis
These results raise two major issues of interpretation. The first concerns the identification of indistinguishably close contacts observed at the light microscopic level as synapses, and the second concerns synaptic strength. Several lines of evidence support the designation of appositions between PH-SNs and APRs as putative synapses. First, the size and shape of the contacts resemble central synapses in other systems. Putative synapses were identified as indistinguishably close or overlapping profiles in images with a pixel resolution of 0.2 μm, with the mean contact length in thez-axis estimated to be 1.25–1.35 μm (see Results). This size is consistent with other insect sensorimotor synapses (e.g., output synapses from locust wing hinge stretch receptor neurons to wing motoneurons) (Peters et al., 1985) and central synapses in general (for review, see Edwards, 1995). Most contacts occurred en passant from PH-SN varicosities onto high-order dendrites of an APR (Fig. 5). Axonal varicosities (boutons) typically contain output synapses in insects [cockroach (Blagburn et al., 1985), locust (Watson and Burrows, 1985), Manduca (Sun et al., 1997)] and other species (DeRiemer and Macagno, 1981; Glanzman et al., 1989; Tse et al., 1991; Magoski and Bulloch, 1997). Output synapses may also be made by fine, smooth processes (Watson and Burrows, 1982; Granzow et al., 1985;Walmsley et al., 1987; Lamotte d'Incamps et al., 1998). Close appositions observed at the light microscopic level have been confirmed to be synapses by electron microscopy in a number of cases (Muller and McMahan, 1976; Pilowsky et al., 1990; Sorra and Harris, 1993; Buhl et al., 1994; Gonchar et al., 2002). In summary, the morphology of appositions between PH-SNs and APRs is entirely consistent with synaptic contacts, an assumption that can be tested in future experiments.
A second key issue in interpreting our results is the strength of individual synaptic contacts between PH-SNs and APRs. The amplitude of an EPSP reflects the number of release sites and the size of the synaptic current produced at each release site, which is in turn determined by the amount of neurotransmitter released, the number of receptors, and other factors (for review, see Edwards, 1995). The size of contacts between PH-SNs and APRs did not change detectably between days L3 and P0 (Fig. 6 B), but more subtle alterations in synapse structure that could affect transmission (e.g., a decrease in the number of active zones or in the size of the postsynaptic density) are undetectable at the light microscopic level. It is well established that synaptic strength and synaptic morphology can be regulated independently (Tsujimoto et al., 1990; Stewart et al., 1996; for review, see Davis and Goodman, 1998), so the developmental weakening of synapses from PH-SNs to APRs could be accomplished entirely by electrophysiological changes rather than by altering the number of contacts. However, analysis of EPSP time-to-peak and the intrinsic electrical properties of APRs suggest that the developmental decrease in EPSP amplitude results from a decrease in binomial n(i.e., the number of contacts or release sites) rather than from other factors (Streichert and Weeks, 1995). Evidence against the possibility that a change in synaptic release properties from PH-SNs is involved in synaptic weakening comes from heterochronic mosaic animals, in which PH-SNs are maintained in the larval state by treating their somata with juvenile hormone, whereas APRs undergo normal dendritic regression (Streichert and Weeks, 1995). The EPSP amplitude decreases between days L3 and P0 at these heterochronic mosaic synapses, suggesting that genomic actions of 20E on PH-SNs are not required for synaptic weakening. On the postsynaptic side, the possibility that 20E affects the electrophysiological properties of APR, such as the number or density of ACh receptors, can be investigated in cell culture (Melville et al., 2003; and our unpublished observations).
Although such factors could contribute to synaptic weakening, the existing anatomical and electrophysiological data support the hypothesis that the massive reduction in the number of physical contacts between PH-SNs and APRs (Fig. 6 A) plays a predominant role in the developmental reduction in EPSP amplitude. Fewer contacts should produce less total synaptic current, and hence a smaller EPSP. Relationships between the number of synaptic contacts and synaptic strength at monosynaptic connections have been amply demonstrated in other systems [crayfish (Nakagawa and Mulloney, 2001), mollusc (Glanzman et al., 1989), frog (Kuno et al., 1971;Grantyn et al., 1984), goldfish (Korn et al., 1981), and cat (Malenka and Nicoll, 1997)]. By this scenario, the predominant locus for the weakening of synapses from PH-SNs to APRs is postsynaptic, because of the steroid-induced regression of APR dendrites.
Relationship to neural circuits for behavior
The excitatory synapses from PH-SNs to APRs constitute the direct pathway in the neural circuit for the larval proleg withdrawal reflex (Fig. 1 C) and are the major source of synaptic drive to proleg retractor motoneurons during the reflex (Jacobs and Weeks, 1990;Streichert and Weeks, 1995). Hence, the steroid-induced weakening of these synapses has direct behavioral relevance in contributing to the weakening and loss of the proleg withdrawal reflex at the end of larval life. After pupation, some APRs are eliminated by programmed cell death, whereas the others undergo steroid-induced dendritic regrowth, receive new synaptic inputs, and are respecified for new functional roles (Weeks and Ernst-Utzschneider, 1989; Sandstrom and Weeks, 1998; Lubischer et al., 1999). Dendritic regression and synaptic loss extricates APRs from outmoded larval circuits and prepares the motoneurons to be incorporated into new circuits for the next life stages. The PH-SNs are believed to die after pupation (Streichert and Weeks, 1995) and so are not expected to undergo regression.
Our findings represent the first demonstration in a single system that a steroid-induced change in dendritic architecture is associated with a change in the strength of monosynaptic connections and a change in behavior. Steroid hormones have pervasive effects on dendritic architecture and behavior (see Introduction), but the demonstration of specific links between hormones, synapses, and behavior is difficult. Many fundamental issues regarding steroid effects on neurons and behavior can be readily addressed in the experimentally accessible insect nervous system (for review, see Fahrbach and Weeks, 2002). Application of this level of analysis to the vertebrate nervous system should reveal intriguing insights into the regulation of sexually differentiated and other behaviors by steroid hormones.
This work was supported by National Institutes of Health Grant R01 NS23208 (J.C.W.). We thank Dr. Daniel Hartline for advice on three-dimensional reconstruction, Dr. David Lenzi for service as a blinded observer, and Dr. William M. Roberts, Dr. Johnathan Melville, M. Jade Zee, and Ari Winbush for helpful suggestions on this manuscript.
Correspondence should be addressed to Janis C. Weeks, 1254 Institute of Neuroscience, University of Oregon, Eugene, OR 97403-1254. E-mail:.
J. R. Gray's present address: Department of Biology, University of Saskatchewan, 112 Science Place, Saskatoon, SK S7N 5E2 Canada.