Target-Dependent Retrograde Signaling Mediates Synaptic Plasticity at the Drosophila Neuromuscular Junction

Abstract

Neurons that innervate multiple targets often establish synapses with target-specific strengths, and local forms of synaptic plasticity. We have examined the molecular-genetic mechanisms that allow a single Drosophila motoneuron, the ventral Common Exciter (vCE), to establish connections with target-specific properties at its various synaptic partners. By driving transgenes in a subset of vCE’s targets, we found that individual target cells are able to independently control the properties of VCE’s innervating branch and synapses. This is achieved by means of a trans-synaptic growth factor secreted by the target cell. At the larval neuromuscular junction, postsynaptic glutamate receptor activity stimulates the release of the BMP4/5/6 homologue Glass bottom boat (Gbb). As larvae mature and motoneuron terminals grow, Gbb activates the R-Smad transcriptional regulator pMad to facilitate presynaptic development. We found that manipulations affecting glutamate receptors or Gbb within subsets of target muscles led to local effects either specific to the manipulated muscle or by a limited gradient within the presynaptic branches. While presynaptic development depends on pMad transcriptional activity within the motoneuron nucleus, we find that the Gbb growth factor may also act locally within presynaptic terminals. Local Gbb signaling and presynaptic pMad accumulation within boutons may therefore participate in a “synaptic tagging” mechanism, to influence synaptic growth and plasticity in Drosophila.

Introduction

Neurons often innervate multiple targets, establishing connections where each contact has a characteristic synaptic size, strength, and plasticity. These synaptic specializations may arise in response to retrograde trans-synaptic signals from the postsynaptic cell. The regulation of synaptic transmission through retrograde signaling has been observed widely in vertebrate nervous systems. These signals are often associated with forms of synaptic plasticity involved in learning and memory.

Within the hippocampus, neurons that innervate multiple targets can have unequal neurotransmitter release and physiological plasticity at their different branches (Ali, Deuchars, Pawelzik, & Thomson, 1998; Bi & Poo, 1998; Galvan & Gutierrez, 2017; Maccaferri, Toth, & McBain, 1998; McMahon & Kauer, 1997; Scanziani, Gahwiler, & Charpak, 1998). In these cases, the contacts have distinct levels of electrical activity, with corresponding differences in synaptic strength and plasticity caused by multiple retrograde trans-synaptic signaling systems. Evidence from cultured hippocampal cells indicates that excitatory postsynaptic neurons release the growth factor BDNF to potentiate local presynaptic glutamate release (Schinder, Berninger, & Poo, 2000). Another form of retrograde modulation is observed at synapses made onto inhibitory interneurons. Somatostatin-type GABAergic interneurons use the molecule ELFN1 to modulate release of glutamate from their innervating cells (Sylwestrak & Ghosh, 2012) by influencing the levels of mGluR7 on the corresponding presynaptic terminals (Tomioka et al., 2014).

Here we examine whether local, trans-synaptic regulation of presynaptic development and plasticity occurs at the neuromuscular junctions (NMJ) of Drosophila larvae. Larval bodywall neuromuscular synapses have been widely used as a genetic model system for identifying mechanisms of glutamatergic synapse development and activity-dependent plasticity. The NMJ is easily accessible, which allows analyses of synaptic structure and function, and advanced genetic tools allow for manipulation of pre- and/or postsynaptic cells (Berke & Keshishian, 2008; Harris & Littleton, 2015). Larval NMJs grow rapidly during development, and the resulting size and complexity of the motoneuron terminal arbors are enhanced by presynaptic activity (Budnik, Zhong, & Wu, 1990; Keshishian et al., 1993). During NMJ growth, postsynaptic muscles secrete Gbb (Glass Bottom Boat), a BMP4/5/6 ligand that acts through an R-Smad (Mad) within presynaptic motoneurons. Phosphorylated Mad (pMad) accumulates in presynaptic nuclei, where it serves as a transcriptional regulator for NMJ growth and plasticity (canonical signaling; Berke, Wittnam, McNeill, Van Vactor, & Keshishian, 2013; Marques et al., 2003; McCabe et al., 2003). pMad is also observed in presynaptic boutons, but the function of this separate pool of pMad is less-well understood (so-called non-canonical signaling; Higashi-Kovtun, Mosca, Dickman, Meinertzhagen, & Schwarz, 2010; M. Sulkowski, Kim, & Serpe, 2014).

Here we examined the role of local BMP signaling in regulating the development and plasticity at distinct branches of the same neuron. Our experiments used the ventral common exciter (vCE), a motoneuron that innervates most of the ventral longitudinal muscle fibers of the embryo and larva (Hoang & Chiba, 2001; Landgraf, Bossing, Technau, & Bate, 1997; Lnenicka & Keshishian, 2000; Lnenicka, Spencer, & Keshishian, 2003). The results indicate that retrograde Gbb signaling from individual muscle fibers produces a local, branch-specific regulation of presynaptic development, in an activity-dependent fashion. This local role for Gbb and synaptic pMad complements canonical, pMad-dependent transcriptional regulation in motoneuron nuclei to influence NMJ development. Our results indicate that Gbb influences local, non-canonical signals at presynaptic terminals, which predisposes specific axon branches for growth and plasticity after activating cell-wide genetic programs for growth.

Materials & Methods

Drosophila stocks.

All stocks were maintained at room temperature (21°C), with genetic crosses kept at 25°C. The glutamate receptor (dGluR) and Glass bottom boat (Gbb) RNAi lines were generated by the Transgenic RNAi Project (TRiP; Hu et al., 2017) and were obtained from the Bloomington Stock Center. These were: 1) y v ; ; UAS-RNAi- dGluRC (TRiP #1854), 2) y v ; ; UAS-RNAi-dGluRA (TRiP #2647), and 3) y sc v ; ; UAS-RNAi-Gbb (TRiP #1243). The efficacy of the dGluR knockdown in vivo was characterized using subunit-specific antibodies (see below) when the RNAi transgenes were expressed in all muscles, MF (muscle fiber) 12, or MFs 7/6. The dGluRIIA antibody was also tested on progeny of a cross between a deficiency that eliminates dGluRA and dGluRB (+ ; Df(2L)cl^h4/ CyO, P[w^+mc= Act-GFP]JMR1) and a deficiency specific to dGluRA (+ ; DGluRA^AD9/ CyO, P[w^+mc= Act-GFP]JMR1, both lines were re-balanced from stocks provided by S. Sigrist, Free University, Berlin, Germany). dGluRA receptor activity was decreased by expressing a mutant channel with reduced conductance (w ; UAS-DGluRIIA M614R; DiAntonio, Petersen, Heckmann, & Goodman, 1999), obtained from M. Serpe (NICHD, Bethesda, MD). BMP signaling by Gbb was suppressed in progeny from a cross between w; gbb¹ , UAS-Gbb / CyO, Act-GFP (D. Allen, University of British Columbia, Canada) and b pr cn bw gbb² / CyO, P[w^+mc= Act-GFP]JMR1 (re-balanced from a stock provided by K. Wharton, Brown University, Providence, RI).

Transgene expression was directed to all larval muscles with MHC-GAL4 and to MF 12 with w ; ; 5053A-GAL4 (also referred to as M12-GAL4), which was obtained from L. Griffith (Brandeis University, Waltham, MA). The P-element used to create 5053A-GAL4 landed in the Tey (teyrha-meyrha) gene, known to affect axonal guidance (Inaki, Shinza-Kameda, Ismat, Frasch, & Nose, 2010) though no guidance defects were observed in 5053A-GAL4 / + heterozygotes. Expression of the 5053A-GAL4 line was verified by crossing to a UAS-GFP effector (w; UAS-2xeGFP; UAS-2xeGFP). In addition to MF 12, 5053A-GAL4 also expresses in motoneurons that innervate midline oblique muscles but it does not express in motoneurons that innervate the muscles studied here. The 5053A-GAL4 line was placed into an eag¹ Sh¹²⁰ double mutant background (C.F. Wu, University of Iowa, Iowa City, IA) where multiple K⁺ currents are altered, motoneuron action potential firing is increased, and third instar larval NMJ synapses are expanded (Budnik et al., 1990). The RNAi transgene targeting dGluRC was also expressed in MFs 7 and 6 in an anterior-posterior gradient using the BG487-GAL4 line (yw ; BG487-GAL4). As above, the expression of the BG487-GAL4 line was verified by crossing to a UAS-GFP effector. The experiments with BG487-GAL4 were coupled with a second manipulation of membrane excitability (high-temperature rearing), which increases neuronal excitability in a cell-autonomous fashion (Peng et al., 2007) and expands NMJ size (Berke et al., 2013; Sigrist, Reiff, Thiel, Steinert, & Schuster, 2003; Zhong & Wu, 2004).

Nomenclature.

The neuron we study here has been previously named MNISNb/d-1_S (Choi, Park, & Griffith, 2004). We have chosen to use the name ventral common exciter (vCE) motoneuron to be consistent with the naming of comparable motoneurons in other invertebrate systems (Pearce & Govind, 1993; Stephens & Govind, 1981). The two heterotetrameric glutamate receptor complexes of the musculature are referred to as the dGluRIIA and dGluRIIB receptors, depending on whether they include the A or B receptor subunits. To avoid confusion, the five protein subunits of the glutamate receptors are referred to in this paper as dGluRA, dGluRB, dGluRC (also known as dGluRIII), dGluRD, and dGluRE.

Immunolabeling.

Third instar larvae were fillet-dissected in 1mM Ca saline (140mM NaCl, 5mM KCl, 1mM CaCl₂, 4mM NaHCO₃, 6mM MgCl₂, 5mM TES, 5mM Trehalose, 50mM Sucrose, and pH adjusted to 7.2 with NaOH) and fixed for 1hr in either 4% paraformaldehyde or for 5 min in Bouin’s Fluid (Polysciences, Inc., Warrington, PA). Bodywall preparations were then washed in PBS containing Triton X-100 (TBS: 20mM NaH₂PO₄, 150mM NaCl, 0.3% Triton X-100, pH 7.3) and blocked with TBS containing 1% bovine serum albumin on an orbital shaker. The procedures of antibody staining were as follows: overnight incubation in primary antibody with agitation at 4°C followed by a 1hr TBS wash (room temperature, RT), a 4hr incubation with secondary antibody and a final 1hr PBS wash, both at RT. Presynaptic motoneuron terminals were labeled with a goat anti-HRP antibody (1:200, Cappel Laboratories, Cochranville, PA and Fischer Scientific, Lenexa, KS) and visualized with a peroxidase-conjugated donkey anti-goat antibody (1:200, Jackson ImmunoResearch Laboratories, West Grove, PA) followed by a diaminobenzidine (DAB, Polysciences and Fischer Scientific) reaction (0.05% DAB in PBS with 0.003% H₂O₂). Glutamate receptor subunits were probed with rabbit anti-dGluRC (1:500) (from A. DiAntonio, Washington University, St. Louis, MO) and mouse anti-dGluRIIA (hybridoma 8B4D2, Developmental Studies Hybridoma Bank, DSHB) after Bouin’s fixation. The antibody specific to phosphorylated Mad (pMad, PS1, rabbit, 1:200, from P. ten Dijke, University of Leiden, Leiden, Netherlands) was visualized by combining fluorescent secondary and tertiary antibodies containing the same fluorophore (listed below) when paraformaldehyde was used. When the fixative was Bouin’s Fluid, the tertiary antibody was not necessary. Alexa fluorescent secondary (and tertiary) antibodies were used at 1:1000 (Invitrogen, Carlsbad, CA) and included: Donkey anti-goat 488, Donkey anti-mouse 594, Goat anti-rabbit 647, Donkey anti-rabbit 647, Donkey anti-goat 647, and Donkey anti-mouse 647. DAB-stained preparations were mounted in glycerol and fluorescent preparations were mounted using the Slowfade antifade kit (Invitrogen).

Microscopy.

All imaging data were obtained from ventral longitudinal muscle fibers 7, 6, 13, and 12 in abdominal segments A3 and A4. For experiments with the BG487-Gal4 driver, data was collected from A2 and A3, as the expression was highest in these abdominal segments. Fluorescent images were collected at Yale University using a Zeiss LSM 510 with a 40X Planapochromat 1.3 NA objective and at Truman State University on a Leica DMI 6000B confocal system using 63X and 100X, 1.4 NA objectives. Optical step sizes varied between 0.4 and 1 μm for the different antibody experiments. Control and experimental animals were stained in the same well of a 24 well plate, mounted on the same slide, and imaged with the same acquisition parameters. Deconvolution (and processing) of images containing glutamate receptor subunit immunolabeling was completed with the Iterative Deconvolve 3D and Diffraction_PSF_3D plugins after background subtraction in ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.nih.gov/ij/, 1997–2016.). NMJ size was quantified by counting boutons from anti-HRP-labeled and DAB-stained preparations using bright-field imaging. Fluorescent and grey-scaled images were organized with Adobe Photoshop (Adobe Systems, San Jose, CA). Synaptic boutons were scored and classified as previously described (Johansen, Halpern, Johansen, & Keshishian, 1989; Ruiz-Canada & Budnik, 2006a). Motoneuron branches that innervated bodywall muscles were distinguished by bouton shapes, which remained consistent among a branch yet distinct between branches (Lnenicka & Keshishian, 2000; ; type I_S - 2 to 4μm, IB - 3 to 5μm, II - 1 to 2μm, III - 3 to 5μm; type I and II boutons are predominantly round while type III boutons are oblong). The relative intensity of pMad immunofluorescence of a 5×5 pixel region (corresponding to a 2×2 μm area of interest) for each Type I_S bouton on MF 12 was quantified with the Measure function of ImageJ. Data on synaptic size (the absolute number of boutons) and pMad fluorescence were computed as the mean for each animal and the mean ± SEM for the n animals per genotype or condition. Comparisons were analyzed with two-tailed t tests.

Electrophysiology.

All physiological data was collected at room temperature. Wandering third instar larvae were fillet-dissected in 1mM Ca-containing saline and recorded with 3M KCl sharp microelectrodes (35–45MΩ resistance) as described previously (Berke et al., 2013). Stimuli were delivered with suction electrodes either near the cut end of the segmental nerve (for orthograde stimulation of MFs 6, 13, and 12) or at the branch between MFs 13 and 12 (for orthograde stimulation of MF 12 and anterograde stimulation of MFs 13 and 6; the same location as in Lnenicka & Keshishian, 2000). Stimulus protocols were created with the Clampex program within the PClamp 9.0 suite (Molecular Devices, San Jose, CA). Ten second recordings of spontaneous activity from muscles with a stable resting potential (≤ −50mV) were obtained prior to evoking action potential-dependent synaptic activity. Data were collected at 10kHz with a Dagan 8500 two-electrode voltage-clamp amplifier and filtered using a 3kHz cutoff (Gaussian). For quantal analysis, the frequency and amplitude of spontaneous synaptic potentials and the amplitude of evoked potentials were measured in the Clampfit program within PClamp. Quantal content was computed as the mean excitatory junctional potential (EJP) / the mean miniature EJP (mEJP) by simultaneously stimulating both axons whose terminals contain large (type I_B) and small (type I_S) boutons. Physiological data were computed as a mean for each animal and a mean ± SEM for the n animals of each genotype or condition. Comparisons were analyzed with two-tailed t tests. In all figures, p < 0.005 is indicated by ** and p < 0.05 is indicated by *.

Results

The embryonic and larval abdominal neuromuscular system of Drosophila consists of 35 segmentally repeated, singly-identified motoneurons. Nearly all of the bodywall motoneurons innervate only one or two muscle fibers (MFs) per hemisegment (Landgraf et al., 1997; Schmid, Chiba, & Doe, 1999; Schmidt et al., 1997), with characteristic type I_B synaptic boutons. By contrast, the ventral common exciter (vCE) motoneuron innervates multiple ventral muscle fibers, including the four ventral longitudinal MFs 7, 6, 13, and 12 (Supplemental Fig. 1A, B; Lnenicka et al., 2003). The vCE NMJs have a characteristic, intermediate-sized (type I_S) synaptic bouton (see Materials and Methods; Johansen et al., 1989; Ruiz-Canada & Budnik, 2006a; Takizawa, Komatsu, & Tsujimura, 2007).

At larval neuromuscular junctions, elevated neuronal activity during development results in an increase in the size and complexity of the NMJ, evident by more presynaptic boutons and more highly branched axon terminals (Berke et al., 2013; Budnik et al., 1990; Mosca, Carrillo, White, & Keshishian, 2005). The activity-dependent increase in motoneuron terminal arbor size and complexity is accompanied by a corresponding increase in the amplitude and quantal content of excitatory junction potentials (EJPs), and in the frequency of spontaneous synaptic release events. Postsynaptic glutamate receptor activity is required for this developmental plasticity (Sigrist et al., 2003; Sigrist, Thiel, Reiff, & Schuster, 2002). We first used RNA interference (RNAi) to partially knock down glutamate receptors in only one (MF 12) of the four ventral longitudinal vCE targets, leaving the other three fibers unchanged (MFs 7, 6, and 13). This was made possible by driving transgenes with the 5053A-GAL4 driver, whose muscle expression is restricted to MF 12 in all abdominal segments throughout embryonic and larval development (Supplemental Fig. 1C, D, and Supplemental Fig. 2).

Our initial test examined the effect of knocking down the dGluRC subunit in MF 12 using an RNAi transgene (see Materials & Methods for details). The C subunit is common to the two heterotetrameric glutamate receptor complexes found in muscle (dGluRIIA and dGlurIIB), which utilize either of two conducting subunits respectively (dGluRA and dGluRB). The dGluRC subunit is required for both dGluRII receptor complexes in muscle (Featherstone et al., 2005; Marrus, Portman, Allen, Moffat, & DiAntonio, 2004; Qin et al., 2005). Figure 1A–B illustrate knockdown of both dGluRII complexes in 3^rd instar larvae. Immunolabeling of the dGluRC subunit (shown in red) is shown counterlabeled with immunolabeling of the innervating motoneuron terminals (with anti-HRP, in green; the box highlights the region of innervation of MF 12). As expected, the dGluRC subunit immunolabeling was reduced in MF 12 (Fig. 1B2, arrowheads). These results show that it is possible to reduce glutamate receptor expression at just one of the vCE muscle fiber targets, leaving receptor levels at the other contacts unaffected as demonstrated by immunolabeling. The knockdown caused minimal change in vCE NMJ morphology both at the affected synapse and in the flanking, unmanipulated synapses (synaptic bouton counts, mean ± SEM ; 5053A driver control: MF 7/6 – 36.7 ± 1.6, MF 13 – 20.3 ± 0.8, MF 12 – 17.2 ± 0.8; 5053A > dGluRC RNAi: MF 7/6 – 39.3 ± 1.4, MF 13 – 23.5 ± 0.9, MF 12 – 17.7 ± 1.0). The relatively normal anatomy despite the targeted knockdown was desirable, as we intended to test whether reduced postsynaptic glutamate receptor expression affected the motoneuron’s ability to undergo activity-dependent expansion (see below).

Figure 1. Selective manipulation of dGluR subunits using the 5053A-GAL4 driver reduces glutamate receptor function in MF 12.

A1, A2. dGluRC expression is shown for a 5053A-GAL4 driver control 3^rd instar larva. B1, B2. dGluRC localization is shown in a 5053A-GAL4 > dGluRC RNAi 3^rd instar larva. The vCE NMJs on MFs 7&6 and 13 are denoted by arrows. Boxes indicate regions of MF 12 with different levels of subunit expression. Arrow heads in B2 indicate faint immunolabeling in MF 12. Scale bar, 25 μm. C. Voltage recordings from a control (5053A-GAL4, pale blue) and a dGluRC RNAi-expressing larva (5053A-GAL4 > dGluRC RNAi, dark blue) show evoked junctional potentials (EJPs, top), and spontaneous miniature EJPs (mEJPs, bottom) from MF 6 and MF 12. Spontaneous mEJPs, presented at a different scale, show no change in amplitude but a reduced frequency in MF 12. The mEJPs were collected prior to evoking EJPs with nerve stimulation. D. A comparison of EJP amplitude, mEJP frequency, mEJP amplitude, and quantal content of MF 12 and MF 6 in larvae expressing the 5053A-GAL4 driver control (pale blue), larvae expressing RNAi transgenes against dGluRC (medium blue) and dGluRA (dark blue), and larvae expressing a channel-dead version of dGluRA subunit (dGluRIIA^M/R; black). The mEJP frequency and quantal content (mean EJP amplitude/mean mEJP amplitude) are significantly reduced. N ≥ 5 animals and ≥ 13 muscles for each genotype and condition.

The dGluRC subunit knockdown in MF 12 had significant physiological effects. It reduced the amplitude of evoked excitatory junction potentials (EJPs), the frequency of spontaneous miniature EJPs (mEJPs), and the computed quantal content of the vCE at that muscle fiber (Figure 1). Each of these changes imply a presynaptic change in response to a postsynaptic manipulation. By contrast, the reduction of glutamate receptors in MF 12 had little effect on vCE’s excitatory neurotransmission onto the neighboring unmanipulated target MF 6 (Fig. 1, bottom traces; Fig. 1D, mEJP frequency was reduced by GluRA RNAi). This suggests a branch-specific regulation of presynaptic release, dependent on local postsynaptic glutamate receptor levels.

Of the two current-carrying glutamate receptor subunits (dGluRA or dGluRB), dGluRA plays a distinct role in the retrograde control of presynaptic development and plasticity (DiAntonio et al., 1999; Petersen, Fetter, Noordermeer, Goodman, & DiAntonio, 1997; Sigrist et al., 2000; Sigrist et al., 2002; M. Sulkowski et al., 2014). We therefore examined how an RNAi knockdown of the dGluRA subunit alone, or alternately the expression of a channel-dead, dominant-negative variant of dGluRA (dGluRA^M/R; DiAntonio et al., 1999) affected the vCE’s NMJ development. First, we verified the ability of an RNAi transgene to effectively knock down the dGluRA subunit in all muscles using the pan-muscular MHC-GAL4 driver (data not shown; expression pattern shown in Supplemental Fig. 2A). We then found that the targeted reduction of dGluRA in MF 12, specifically affecting the dGluRIIA complex, caused the same, predominantly local physiological effects as was seen with the dGluRC knockdown that affects both glutamate receptor complexes (Fig. 1D). The amplitude of evoked EJPs, the computed mean quantal content, and mEJP frequencies were all reduced in MF 12. We failed to detect a decrease in quantal size (mEJP amplitude) in any of the tested situations, suggesting that the effects are presynaptic. As noted above with the dGluRC knockdown, only mEJP frequency on MF 6 was significantly reduced by dGluRA RNAi expression in MF 12.

Postsynaptic glutamate receptor function is required for local, presynaptic activity-dependent growth plasticity

Mutations affecting potassium channels (such as eag¹ Sh¹²⁰) cause elevated neuromuscular activity, leading to significantly expanded NMJs, as indicated by the number of synaptic boutons and the complexity of branches (Berke et al., 2013; Budnik et al., 1990; Mosca et al., 2005). We quantified the number of vCE boutons on MF 12 in response to elevated NMJ activity, compared to the number found on neighboring vCE target muscles. We first examined the effect of knocking down all dGluRII receptors in MF 12 by expressing the dGluRC RNAi, as described above. Figure 2A shows that the expression of the dGluRC RNAi in MF 12 of a hyperactive eag Sh larvae significantly suppressed growth plasticity in the MF 12 branch of the vCE. We also noted that plasticity was affected, albeit to a lesser extent in the adjacent branch to MF 13. By contrast, the vCE branches onto MFs 7 and 6 were unaffected, and expanded to a similar size as those of the eag Sh, 5053A-GAL4 driver control.

Figure 2. Activity-dependent growth plasticity of vCE synapses at MF 12 is strongly suppressed by knockdown of the dGluRC or dGluRA subunits.

A. The eag¹ Sh¹²⁰ hyperactivity mutations expand the NMJs on MFs 7&6, 13, and 12 (dark blue bar), as compared to the 5053A-GAL4 driver control (light blue bar). The dGluRC subunit is required for expression of both the dGluRIIA and the dGluRIIB glutamate receptor complexes in muscle (Featherstone et al., 2005; Marrus et al., 2004; Qin et al., 2005). Knockdown of the dGluRC subunit (black bar) in MF 12, targetting both receptor complexes, suppresses activity-dependent expansion most strongly at that muscle, with less suppression observed in the adjacent MF 13. B. An RNAi knockdown of dGluRA subunit in MF 12 (black bars), targetting the dGluRIIA receptor complex, suppresses growth plasticity only in that muscle fiber in eag Sh animals. N ≥ 9 animals for all genotypes.

Knockdown of the dGluRA subunit suppressed activity-dependent NMJ expansion solely at MF12 (Fig. 2B), a more local effect than observed for the dGluRC knockdown. We did not find that dGluRA knockdowns suppressed NMJ plasticity on any of the other vCE targets that were examined. These results are consistent with the idea that synaptic contacts made by a neuron onto different targets may have distinct degrees of activity-dependent plasticity, dependent on a target-specific retrograde signal from the postsynaptic cell. The results also agree with previous findings that dGluRIIA receptor complexes are both necessary and sufficient for neural activity to enhance NMJ size during development.

To rule out the possibility that our postsynaptic manipulations somehow affected presynaptic excitability at the MF 12 branch, which as a result would affect NMJ expansion, we determined the frequency of spontaneous EJPs, as a measure of motoneuron hyperactivity. Both eag Sh double mutants and manipulations of the Sh channel alone increase the frequency of spontaneous EJPs (Budnik et al., 1990; Mosca et al., 2005). We found that large, spontaneous EJPs still occurred in MF 12 of eag Sh mutants, despite the dGluRC subunit knockdown (5053A-GAL4 driver: 0.65 ± 0.03 Hz; eag Sh, 9.93 ± 1.61 Hz; eag Sh, 5053A-GAL4 > dGluRC RNAi: 7.72 ± 0.47 Hz; n ≥ 3 animals and 9 muscles per genotype; p=0.24 compared to eag Sh and p=0.035 compared to the 5053A control). This indicates that despite elevated presynaptic activity, NMJ expansion was suppressed predominantly at the VCE branch to the muscle where glutamate receptors had been knocked down. The observations are consistent with a local activity-dependent control of presynaptic growth by trans-synaptic signals from a postsynaptic target.

We next extended our analysis by knocking down dGluRC in a second subset of vCE targets to demonstrate the generality of these effects. The BG487-GAL4 driver expresses transgenes only in MFs 7 and 6, and it does so in an anterior-posterior gradient (Budnik et al., 1996 and Supplemental Fig. 2). In addition, we increased membrane excitability using a different, non-genetic method (elevated-temperature rearing; see Materials and Methods). When the dGluRC subunit RNAi was expressed with the BG487-GAL4 driver, dGluRC immunolabeling was reduced in an anterior-posterior gradient as expected, with the strongest suppression occurring in segments A2 and A3 (Fig. 3A). Control larvae (BG487-GAL4/+) reared at an elevated temperature (30°C vs. RT 21°C) showed a significant increase in total boutons on MFs 7, 6, and 13 (Fig. 3B1). When the dGluRC subunit RNAi was expressed in MFs 7 and 6, the number of boutons on those muscles was reduced in comparison to the control levels. The number of boutons on the neighboring MF 13 remained elevated (Fig. 3B1), indicating that the suppression was local to the muscle fibers where glutamate receptors were knocked down. These results are consistent with the above observations found for receptor knockdown in MF 12 and support the view that each branch of this motoneuron is subject to independent trans-synaptic control of activity-dependent growth plasticity.

Figure 3. The RNAi-mediated knockdown of dGluRC in MFs 7 and 6 suppresses temperature-dependent NMJ plasticity.

A. When the dGluRC subunit is knocked down using the BG487-GAL4 driver (targeting both glutamate receptor complexes), the resulting dGluRC immunolabeling shows an anterior-posterior gradient of suppression, strongest in abdominal segments A2 and A3, and lower in posterior segments A4 and A5. Scale bar, 20μm. The second row of images presents a region of higher resolution area, as indicated by the box. Scale bar, 5 μm. B. Quantification of NMJ arbor size: B1. Total bouton counts on MFs 7, 6 and 13 in BG487-GAL4 controls (BG487/+) reared at room temperature (21°C) and 30°C. B2. The total number of synaptic boutons on MFs 7 and 6, as compared to MF 13 in BG487 > dGluRC RNAi animals under similar rearing conditions. B3. The effect of expressing the dGluRC RNAi with BG487-GAL4 on temperature-dependent vCE terminal expansion. N ≥ 13 animals for all genotypes and conditions.

Knockdown of glutamate receptors in a single muscle fiber reduces local presynaptic pMad expression

Activity-dependent plasticity of NMJs depends on retrograde signaling by a trans-synaptic BMP ligand (Gbb) and by presynaptic nuclear transcription regulated by the downstream R-SMAD transcription factor Mad (Berke et al., 2013). The trans-synaptic BMP signal activates the canonical, presynaptic BMP signaling pathway. The ligand Gbb activates the phosphorylation and endocytosis of type I and II BMP receptors from presynaptic terminal membranes, followed by their transport to the nucleus within a signaling endosome, where they phosphorylate Mad. This is known as the canonical pathway (Ball et al., 2010; Marques et al., 2003; Smith, Machamer, Kim, Hays, & Marques, 2012). While the role of nuclear pMad has been extensively studied, a non-canonical role of presynaptic pMad remains a debated question. Sulkowski et al. (2014) showed that local appearance of pMad within presynaptic boutons correlated with the postsynaptic expression and activity of the dGluRIIA receptor complex. After pan-muscular dGluRA manipulations, pMad levels within boutons were reduced or eliminated independently from the nuclear pool of pMad (M. Sulkowski et al., 2014).

We therefore examined presynaptic pMad localization within vCE boutons following the postsynaptic reduction of glutamate receptors in MF 12. Figure 4 shows presynaptic pMad immunolabeling in the 5053A-GAL4 control (Fig. 4A1–4), followed by RNAi knockdown of the dGluRC subunit (Fig. 4B1–4), as well as by the dominant-negative suppression of the dGluRIIA receptor conductance (using dGluRA^M/R; Fig. 4C1–4). The small boxes indicate the pMad image location of the enlargements (Fig. 4A3, B3, C3), from which the quantification highlights the loss of pMad immunolabeling in MF 12 boutons compared to the adjacent MF13 (Fig. 4A4, B4, C4). This reduction is striking, when compared to the unaltered neighboring MFs 13 and 7 and 6. The level of pMad immunolabeling was estimated by examining average pixel intensity in histograms from Type I_S boutons (see Methods). Our observations confirm those of Sulkowski et al. (2014). In addition, the results show that the retrograde effect on presynaptic pMad levels is specific to the vCE branch projecting to the muscle fiber with reduced receptor expression. Intriguingly, pMad labeling remained unaffected in the NMJ arbor of the peptidergic motoneuron that innervates MF 12 with type III boutons (arrows, Fig. 4).

Figure 4. Accumulation of presynaptic pMad is reduced by knocking down dGluRC and by inhibiting dGluRA conductance.

Third instar 5053A-GAL4 driver control (A1–4), 5053A-GAL4 > dGluRC RNAi (B1–4), and 5053A-GAL4 > dGluRIIA^M/R (C1–4) larvae were double-labeled with antibodies against HRP (A1, B1, and C1, green) and pMad, the activated and phosphorylated form of Mad (A2, B2, and C2, red). Scale bar, 20 μm. Enlarged regions (corresponding to the box) of pMad immunofluorescence show the differential localization among motoneuron terminals (A3, B3, and C3). Scale bar, 5 μm. The ventral longitudinal MFs 7, 6, 13, and 12 are shown. A4, B4, C4. pMad fluorescence was quantified in Type I_S boutons of MF 13 and MF 12 and is presented as the average pixel intensity (See Methods). We note that in control animals, the mean intensities of pMad labeling on MF 12 were greater than those on MF 13, but the difference was not significant (Fig. A4). We also observed that the manipulation of dGluR subunits does not affect pMad immunolabeling of the peptidergic motoneuron (with Type III boutons; Gorczyca et al., 1993) at MF 12 (arrow). Scale bar, 20 μm.

The NMJs of gbb mutant larvae are substantially reduced in size, and they are similar to those lacking either BMP receptors or following suppression of pMad signaling at the nucleus (Berke et al., 2013). We therefore tested whether expression of Gbb in a subset of the vCE targets would locally rescue the mutant gbb phenotype. We expressed the wild type Gbb cDNA in MF 12 of gbb mutants, and compared NMJ size at the ventral longitudinal muscle targets of the vCE target. Figure 5A1–3 shows NMJs on MFs 13 and 12 in the 5053A-GAL4 driver control (Fig. 5A1), the NMJs of the UAS-Gbb effector control, in a gbb mutant background (Fig. 5A2), and finally the NMJs where the Gbb wild-type transgene was driven exclusively in MF 12 by 5053A-GAL4 (Fig. 5A3). Wild-type Gbb rescued NMJ size specifically at MF 12, with the adjacent MF 13 NMJ remaining diminished in size. Figure 5B quantifies motoneuron terminal growth, showing that the expression of Gbb affected the vCE branch to MF 12, and partially rescued both the type II NMJs (5053A-GAL4 control, gbb control, Gbb rescue: 63 ± 6, 28 ± 4, 42 ± 5) as well as the muscle-specific type III peptidergic neuron (5053A-GAL4 control, gbb control, Gbb rescue: 18 ± 2, 8 ± 1, 14 ± 3). By contrast, there was little to no rescue observed for the vCE branches on any of the adjacent muscle fibers, suggesting that the trans-synaptic Gbb diffusion is limited and that presynaptic rescue was restricted to motoneuron branches that directly receive Gbb.

Figure 5. The expression of Gbb in MF 12 of heteroallelic gbb mutants locally rescues NMJ growth.

A. Shown are NMJs on MFs 12 and 13 of the 5053A-GAL4 driver control (A1), a gbb; UAS-Gbb effector control (A2), and the gbb; 5053A-GAL4 > Gbb (A3) experimentally rescued larvae. The NMJs on MF 12 in rescued larvae show an expansion when compared to the non-rescued, neighboring MF 13 for these genotypes. Scale bar, 20 μm. B. Synaptic boutons on MFs 7, 6, 13, and 12. The total number of boutons, the number of large, type I_B boutons, and the number of smaller, vCE type I_S boutons are shown for the two controls and experimental larvae. Counts for the Type II (octopaminergic) and Type III (peptidergic) terminals are given in the Results. N ≥ 7 animals for all genotypes.

Physiological responses in the rescued vCE target muscles were also larger, consistent with the local rescue of NMJ growth (Fig. 6). Recordings were obtained from the 5053A-GAL4 control, the gbb mutant control, and the MF 12-specific rescued animals after stimulation of the peripheral nerve (shown schematically in Fig. 6A). We first stimulated action potentials in all axons of the ISNb nerve (ISNb stimulus in Fig. 6A) and recorded from MF 12. EJP amplitude, mEJP spontaneous frequency, and the calculated quantal content were significantly and selectively rescued from the reduced responses seen in gbb mutants, while the mean mEJP amplitude was unaffected, indicating no change to the postsynaptic quantal size (Fig. 6B). Next, we stimulated the ISNb nerve branch just distal to its bifurcation to MF 12 (vCE stimulus in Fig. 6A). This type of stimulation produced back-propagating action potentials in the vCE axon. We observed smaller amplitude EJPs in MFs 6 and 13, indicating the absence of a genetic rescue at those synapses (n ≥ 3 animals and 4 muscles per genotype; p=0.53 and 0.38 compared to gbb). Only at MF 12 was there a rescue of the mutant physiological phenotypes.

Figure 6. Expression of Gbb in MF 12 of heteroallelic gbb mutants locally rescues neurotransmission.

A. The diagram shows the locations of the stimulating and recording electrode placements in a hemisegment of a larval bodywall. Stimuli were delivered to the cut end of the peripheral nerve close to its innervation of the musculature (SNb stimulus) to excite all motoneurons of the SNb nerve branch. Stimuli were also delivered directly to MF 12 by an electrode placed adjacent to the muscle fiber’s nerve entry point (vCE stimulus; see Results). The electrode arrangement matches that used previously to distinguish between synaptic potentials on MFs 12 and 13 (Lnenicka & Keshishian, 2000; Lnenicka et al., 2003). B. Intracellular recordings obtained from MFs 12, 13, and 6 are quantified from within the same hemisegment after using the SNb stimulus. EJP amplitude and quantal content are rescued selectively on MF 12 by expression of a Gbb cDNA, while mEJP frequency is slightly rescued. The data were from 5053A-GAL4 driver control larvae (pale blue), the gbb mutant effector control (gbb, UAS-Gbb, medium blue), and the rescued larvae (gbb, 5053A-GAL4 > Gbb, dark blue). N ≥ 5 animals for each genotype and muscle.

Lastly, we examined whether the NMJ expansion due to elevated activity in eag Sh mutants can be locally suppressed by expressing an RNAi transgene against Gbb in MF 12. This experiment is the complement of the muscle-specific rescue of gbb mutants. The knockdown of Gbb reduced the fluorescent immunolabeling of pMad in vCE terminals on MF 12 when compared to the branch on the adjacent MF 13 (Figs. 7A,B; showing anti-HRP and pMad immunolabeling; n = 4 hemisegments, 3 animals), F. By contrast, Fig. 7C shows the expected NMJ expansion at MF 13 and MFs 7/6. These results are consistent with the hypothesis that NMJ growth and plasticity is under local, retrograde trans-synaptic control through the action of a non-canonical BMP signaling cascade.

Figure 7. Knockdown of Gbb in MF 12 locally suppresses activity-dependent growth plasticity of the vCE NMJ.

A, B. RNAi knockdown of Gbb in MF 12 reduces the accumulation of pMad in boutons at that muscle fiber as compared to the adjacent and unmanipulated MF 13 (eag Sh mutant background). The image in A shows the NMJs on MFs 13 and 12, immunolabeled with anti-HRP. The image in B shows the corresponding pMad immunolabeling. Several I_S boutons on MF 13 and MF 12 are indicated in each image with arrowheads. scale bar, 5μm. C. The expansion of vCE NMJs in the hyperactive eag Sh mutants is suppressed by the knockdown of Gbb in MF12 but not at the adjacent MF 13 or MFs 7/6. The dashed lines indicate the bouton count in the 5053-GAL4 control for comparison. N ≥ 9 animals for all genotypes.

Discussion

Our results show that the vCE motoneuron of Drosophila can independently alter the physiology, growth, and developmental growth plasticity of synapses made onto specific synaptic partners. The target-specific plasticity that we describe depends on functional postsynaptic glutamate receptors and by retrograde trans-synaptic Gbb signaling from the corresponding postsynaptic cell (summarized in Fig. 8 for experiments with MF 12).

Figure 8. Summary of the roles of the postsynaptic cell in regulating local, presynaptic growth and plasticity.

A. Muscle expression of 95053A-GAL4 is limited to MF 12 (green), one of the four ventral longitudinal muscle targets of the vCE motoneuron. This driver is used to express various transgenic manipulations in MF 12 and served as a control strain. B. Knockdown of the glutamate receptors (green) results in a loss of pMad localization in glutamatergic boutons on MF 12. pMad accumulation (red) remains normal at the other NMJs of the vCE motoneuron, on MFs 7, 6, and 13. C. eag Sh mutations lead to hyperexcitability and activity-dependent expansion of NMJs on ventral longitudinal muscles, except on MF 12 when Gbb is knocked down (green). Expansion occurrs normally on the unaffected MFs 7, 6, and 13. D. gbb mutations reduces NMJ size, while growth of the vCE branches innervating MF 12 is selectively rescued by providing a wild-type copy of Gbb (green). The NMJs on 7, 6, and 13 remained diminished in size.

The results are consistent with examples where retrograde trans-synaptic signaling regulates local, presynaptic physiological development and plasticity. Similar phenomena occur in the cerebral cortex, hippocampus, cerebellum, Ia sensory afferent connections, and the cochlear nucleus (partially reviewed in Blackman, Abrahamsson, Costa, Lalanne, & Sjostrom, 2013; Pelkey & McBain, 2008). Local control is also found in invertebrates (for examples, see Davis & Murphey, 1993; Laurent & Sivaramakrishnan, 1992; Muller & Nicholls, 1974). Several mechanisms have been proposed to link local presynaptic function with signals from their target cells: BDNF (Schinder et al., 2000), endocannabinoids (Beierlein, Fioravante, & Regehr, 2007), and ELFN1 (Sylwestrak & Ghosh, 2012; Tomioka et al., 2014). Our data reveals a fourth mechanism: retrograde trans-synaptic signaling, mediated by the BMP/TGF-β growth factor glass bottom boat (Gbb).

Our experiments used molecular-genetic tools to manipulate subsets of target muscles in order to compare effects both within and between animals. We took advantage of the fact that 1) most bodywall muscle fibers receive glutamatergic inputs from one motoneuron that is fiber-specific, as well as by a common exciter motoneuron that innervates wide swaths of the musculature, and 2) that larval motoneurons form stereotyped neuromuscular junctions with readily identifiable synaptic boutons and function (Budnik et al., 1990; Keshishian et al., 1993; Lnenicka et al., 2003). Using the GAL4-UAS bipartite expression system, we manipulated glutamate receptor expression and trans-synaptic Gbb signaling in various genetic backgrounds. The results show that the vCE can adjust the size and function of individual synapses in response to retrograde Gbb signaling and to a lesser extent, postsynaptic receptor function.

Evidence of a role for glutamate receptors in local synaptic development and plasticity

At the larval NMJ, there are two postsynaptic heterotetrameric glutamate receptor complexes (dGluRIIA and dGluRIIB). The complexes have in common three subunits (dGluRC, dGluRD, and dGluRE) that help traffic the corresponding 4^th subunit (dGluRA for the dGluRIIA complex or dGluRB for the dGluRIIB complex) to the postsynaptic membrane (Featherstone et al., 2005; Marrus et al., 2004; Qin et al., 2005). The receptor complexes are permeable to both Na and Ca (Featherstone et al., 2005; Jan & Jan, 1976; Marrus et al., 2004; Peled, Newman, & Isacoff, 2014; Petersen et al., 1997; Qin et al., 2005; Schuster et al., 1991). The dGluRIIA and dGluRIIB receptors differ with respect to their ligand binding characteristics, and importantly, their ability to influence presynaptic development in a retrograde fashion (Petersen et al., 1997). In particular, changes in the level or function of the dGluRIIA receptor (but not the dGluRIIB receptor) triggers the homeostatic regulation of presynaptic glutamate release (DiAntonio et al., 1999). dGluRIIA receptors are also required for NMJ growth plasticity in response to motoneuron activity (Sigrist et al., 2002). Finally, elevated dGluRA subunit expression can directly expand the NMJ (Sigrist et al., 2000), suggesting that dGluRIIA receptor activity mediates one or more retrograde signals (M. Sulkowski et al., 2014).

We found that dGluRIIA receptors influence the function and morphology of specific branches of the vCE motoneuron. The RNAi-mediated knockdown of either dGluRC or dGluRA, or the dominant-negative suppression of dGluRA receptor conductance in MF 12, reduced local glutamate release with a marginal effect on the neighboring MF 6 (Fig. 1, mEJP frequency was reduced). RNAi transgenes targeting dGluRA blocked the expected activity-dependent growth plasticity at MF 12 in eag Sh mutants (Fig. 2). This was also seen with RNAi transgenes for dGluRC, which primarily affected MF12, but to a lesser extent also affected the adjacent MF13, but not MF7 and 6 (Fig. 2) Thus we cannot rule out the possibility of a modest presynaptic spread of growth suppression. Lastly, the knockdown of dGluRC in MFs 7/6 caused a specific, local suppression of growth plasticity that showed no presynaptic spread (Fig. 3). While the mechanisms linking glutamate receptor activity to presynaptic growth are still unclear, recent work suggests that release of the Gbb ligand is downstream of dGluRIIA receptor activity, as the loss of the dGluRA subunit or the loss of dGluRIIA receptor permeability to Na/Ca correlates with a reduction in presynaptic BMP signaling within presynaptic boutons (M. Sulkowski et al., 2014). Our results emphasize the local nature of this retrograde control. For example, when the dGluRII receptors were manipulated, the level of pMad in neighboring Type III boutons was unaffected, perhaps due to the peptidergic nature of this motoneuron (Gorczyca, Augart, & Budnik, 1993; Ruiz-Canada & Budnik, 2006b). The results also suggests that postsynaptic receptor composition may influence and tune synaptic drive, as occurs at vertebrate synapses (Choquet, 2018).

Evidence for a non-canonical action of retrograde BMP signaling in local synapse development

Retrograde Gbb acts through presynaptic BMP receptors and downstream Smad transcription factors, but after this step, the signaling pathway bifurcates. On the one hand, the activated receptors can be transported in a retrograde fashion (Smith et al., 2012) to activate a pool of the Mad transcription factor in the cell body (the canonical pathway; Ball et al., 2010; Berke et al., 2013; Kang et al., 2014; Marques et al., 2003; McCabe et al., 2003; Summerville et al., 2016; Zhang et al., 2017). Canonical BMP signaling is required for NMJ development and plasticity. On the other hand, BMP receptor expression can be modulated locally to influence pMad in synaptic boutons, independent from signaling in the cell body (non-canonical signaling; Deshpande et al., 2016; Eaton & Davis, 2005; Fuentes-Medel et al., 2012; Higashi-Kovtun et al., 2010; Merino et al., 2009; M. Sulkowski et al., 2014; Vanlandingham et al., 2013). In these cases, NMJ phenotypes were correlated with altered pMad in presynaptic boutons yet normal or oppositely changed levels within motoneuron nuclei. The relationship between canonical and non-canonical signaling is complicated by the potential for multiple forms of Gbb and distinct aspects of receptor activation (M. J. Sulkowski et al., 2016).

Our results indicate that branch-level, synapse-specific signaling acts in concert with retrograde transport and canonical signaling. We observed a decrease in local, presynaptic pMad levels on branches opposite to muscle fibers where either glutamate receptors were knocked down or where a channel-dead receptor was expressed (Fig. 4D) The data is consistent with a model in which synaptic BMP signaling and pMad activity provides a level of control beyond that of canonical, transcription-dependent signaling. According to this view, canonical, transcriptional signaling permissively activates genetic programs to enable NMJ growth plasticity, while non-canonical, presynaptic signaling fine-tunes the degree of synaptic plasticity at the local, branch-specific level. In this model, the nuclear pool of pMad must be active before non-canonical signaling can influence NMJ development or plasticity.

Synaptic pMad is likely part of a complex system involving receptor dynamics and endosomal processing/shuttling, that ‘tags’ axonal branches for growth, maturation, and plasticity. The molecular mechanisms known to regulate local BMP signaling in presynaptic boutons are diverse: the dynamics of nuclear importins, interactions with other trans-synaptic and cytoplasmic signaling systems, recycling of BMP receptors through endocytic and endosomal regulation, RNA binding protein activity, phospholipid composition, and cytoskeletal regulation (Banerjee & Riordan, 2018; Banerjee, Venkatesan, & Bhat, 2017; Deshpande et al., 2016; Eaton & Davis, 2005; Fuentes-Medel et al., 2012; Heo et al., 2017; Higashi-Kovtun et al., 2010; Huang et al., 2016; Kim et al., 2019; Li et al., 2016; Liu et al., 2014; Merino et al., 2009; Piccioli & Littleton, 2014; Politano et al., 2019; M. J. Sulkowski et al., 2016; Summerville et al., 2016; Vanlandingham et al., 2013; Zhang et al., 2017; Zhao et al., 2015). Local, retrograde effects are also possible through muscle-specific regulation of glutamate receptor composition and Gbb production/release (this study; also see Halstead et al., 2014; M. Sulkowski et al., 2014).

At this point, it is unknown how the release of Gbb from a single muscle and subsequent local activation of BMP receptors might alter the NMJ growth, maturation, and plasticity of individual axonal branches. A direct effect of Gbb on BMP receptor expression, perdurance, or signaling has not yet been described. However, Gbb signaling influences macropinocytosis in vitro, and this process does influence the expression of a type I BMP receptor (Kim et al., 2019). It is also unknown whether retrograde Gbb signaling influences the delivery or release of Gbb from motoneurons, which has striking effects on NMJ growth (James & Broihier, 2011; James et al., 2014). BMP signaling is necessary for the rapid budding of new boutons during motoneuron activity (Piccioli & Littleton, 2014), so it will be interesting to identify whether acute Gbb signaling facilitates the budding process, potentially by stabilizing BMP receptors, modulating the local cytoskeleton, or by influencing membrane dynamics.

Local, non-canonical signaling at specific synapses might also potentiate the capture of molecules needed for growth plasticity as they are transported along the axon from the cell body. Several lines of evidence support this view (summarized in Fig. 8 for MF 12 data). First, pMad immunolabeling was reduced in motoneuron terminals synapsing onto MF 12 when either glutamate receptor or Gbb expression was experimentally reduced in that muscle, yet pMad immunolabeling throughout the remainder of the vCE branches remained normal (Figs. 4 and 7). Second, pMad localization in peptidergic type III boutons on MF 12 was retained, even when all pMad labeling on glutamatergic boutons was absent (Fig. 4). Third, the local change in pMad immunolabeling correlated with a local reduction in NMJ growth, glutamate release, and activity-dependent plasticity, but only (or predominantly) at the manipulated muscle fiber. Fourth, the rescue of Gbb expression in MF 12 rescued both NMJ growth and physiology only on that muscle. While our data cannot differentiate a role for presynaptic pMad from that of BMP receptor activity or other molecular processes, the observations indicate that non-canonical, local signaling directs products of transcriptional activity to facilitate the synaptic growth of individual axonal branches.

Canonical retrograde BMP signaling has an early developmental critical period for NMJ growth (predominantly during the first larval instar), after which signaling no longer exerts a positive growth effect (Berke et al., 2013). Non-canonical signaling may therefore be influential after this early period of transcription, where a reduction in non-canonical signaling might lead to NMJ phenotypes if it is reduced during the second and third larval instars, when most NMJ growth occurs (Keshishian et al., 1993). Conversely, reducing non-canonical signaling before or during the transcriptional critical period (in the embryo or early first instar) would presumably have no effect. We find that Gbb acts local to a branch and that activated pMad does not spread widely to vCE terminals contacting adjacent muscle fibers. One mechanism that fulfills both characteristics is the regulation of vesicular cargo, perhaps by altering vesicle dwell time, similar to observations of peptidergic vesicles (Rao, Lang, Levitan, & Deitcher, 2001). Further exploration of this phenomenon should provide insight into how presynaptic neurons strengthen only a subset of their synaptic connections. The powerful genetic tools afforded by Drosophila and the wealth of exploration may even help elucidate mechanisms used during ‘synaptic tagging and capture’ (Frey & Morris, 1997), a key process involved in memory formation in vertebrates.

Supplementary Material

Supp FigS1

Supplemental Figure 1. Selective expression of transgenes in one postsynaptic target (MF 12) of the ventral common exciter (vCE) motoneuron. A. The innervation of ventral longitudinal muscle fibers (MFs 7, 6, 13, and 12) as visualized with anti-HRP immunolabeling. The four ventral muscle fibers are innervated by synaptic terminals with distinct bouton sizes, characteristic of different motoneurons (glutamatergic Type I_B and I_S, and octopaminergic Type II; peptidergic Type III is not shown, but see Fig. 5). Scale bar, 25 μm. Reproduced from Chang and Keshishian (1996), with permission. B. Axon terminals of the ventral common exciter motoneuron, vCE, synapse onto the four ventral longitudinal muscle fibers, MF 12, MF 13, and MFs 7&6. Scale bar, 25 μm. Lucifer yellow dye-fill reproduced from Lnenicka and Keshishian (2000), with permission. C. and D. The 5053A-GAL4 driver expresses transgenes selectively in MF 12 throughout development. C. A DIC image of a stage 16 embryonic bodywall showing 5053A-driven β-Galactosidase expression in MF 12. Scale bar, 4 μm. D. 5053A-driven green fluorescent protein (GFP) expression in MF 12 of an early second instar larva. Neuromuscular terminals (red) are revealed by anti-HRP immunolabeling. Arrows indicate synaptic boutons in the cleft between MFs 7 and 6 and on MFs 13 and 12. 5053A-GAL4 also drives expression in a few motoneurons that innervate the ventral oblique muscles (indicated by fluorescent axon branches at left, below MF 6). Scale bar, 25 μm.

Supp figS2

Supplemental Figure 2. The three GAL4 lines used in this study express GFP in distinct patterns in third instar larvae. A. The Myosin Heavy Chain MHC-GAL4 driver expresses transgenes in all bodywall muscles beginning in late stage embryos. The driver is not expressed within the central nervous system (CNS; inset). B. The 5053A-GAL4 driver expresses transgenes selectively in abdominal MF 12s with uniform intensity along the anterior-posterior axis of larvae. The 5053A-GAL4 driver also expresses in a subpopulation of neurons within the CNS (inset). C. The BG487-GAL4 driver expresses transgenes in MFs 7 and 6 in an anterior-posterior gradient, without neuronal expression (inset). Abdominal segments 2 and 3 (A2 and A3) show consistent expression in MFs 7 and 6.