Retrograde semaphorin–plexin signalling drives homeostatic synaptic plasticity

Abstract

Homeostatic signalling systems ensure stable but flexible neural activity and animal behaviour^1,2,3,4. Presynaptic homeostatic plasticity is a conserved form of neuronal homeostatic signalling that is observed in organisms ranging from Drosophila to human^1,5. Defining the underlying molecular mechanisms of neuronal homeostatic signalling will be essential in order to establish clear connections to the causes and progression of neurological disease. During neural development, semaphorin–plexin signalling instructs axon guidance and neuronal morphogenesis^6,7,8,9,10. However, semaphorins and plexins are also expressed in the adult brain^{11,12,13,14,15,16}. Here we show that semaphorin 2b (Sema2b) is a target-derived signal that acts upon presynaptic plexin B (PlexB) receptors to mediate the retrograde, homeostatic control of presynaptic neurotransmitter release at the neuromuscular junction in Drosophila. Further, we show that Sema2b–PlexB signalling regulates presynaptic homeostatic plasticity through the cytoplasmic protein Mical and the oxoreductase-dependent control of presynaptic actin. We propose that semaphorin–plexin signalling is an essential platform for the stabilization of synaptic transmission throughout the developing and mature nervous system. These findings may be relevant to the aetiology and treatment of diverse neurological and psychiatric diseases that are characterized by altered or inappropriate neural function and behaviour.

Main

Semaphorins are a large family of secreted or membrane-associated signalling proteins and plexins serve as signal-transducing semaphorin receptors^8,9,10. Semaphorin–plexin signalling was initially described as mediating growth cone collapse^8,9. But, semaphorin–plexin signalling is far more diverse^10,11. Notably, semaphorins and plexins continue to be expressed in the mature brain, where their function remains mostly unknown^{12,13,14,15,16}. Semaphorins have been shown to be synaptic signalling proteins, but the activity of semaphorins has been limited to the control of neuroanatomical synapse formation and elimination^15,16,17. Here we demonstrate that semaphorin–plexin signalling achieves retrograde, trans-synaptic control of presynaptic neurotransmitter release and homeostatic plasticity.

We use a well-documented assay to induce presynaptic homeostatic plasticity (PHP), applying a sub-blocking concentration of the glutamate-receptor antagonist philanthotoxin-433 (PhTx; 15 μM) to significantly decrease the amplitude of average miniature excitatory postsynaptic potentials (mEPSPs; 0.3 mM [Ca²⁺]_e) or miniature excitatory postsynaptic currents (mEPSCs; 1.5 mM [Ca²⁺]_e). This postsynaptic perturbation induces a significant increase in presynaptic neurotransmitter release (the quantal content) that offsets the postsynaptic perturbation and restores normal muscle excitation (Fig. 1a–c; raw values and sample sizes for all normalized data are shown in Supplementary Table 1). This offsetting increase in presynaptic neurotransmitter release is characteristic of PHP¹. When we repeated this assay in larvae containing a null mutation in either the sema2b gene (sema2b^C4) or the PlexB gene (PlexB^KG00878), PHP was blocked (Fig. 1a–c; see also Extended Data Fig. 1 for a further description of the gene mutations used in this study). Consistent with this being a loss-of-function phenotype, heterozygous mutations (either sema2b/+ or PlexB/+) have normal PHP (Fig. 1d, e). Remarkably, a double-heterozygous mutant combination of sema2b/+ and PlexB/+ blocks PHP, consistent with both genes acting in concert to drive the expression of PHP (Fig. 1 d, e).

Figure 1: sema2b and PlexB are necessary for PHP.

a, Representative traces (left, 0.3 mM [Ca²⁺]_e; right, 1.5 mM [Ca²⁺]_e) recorded from the indicated genotypes. b, Average mEPSPs and quantal content (QC) for each genotype expressed as the percentage change in the presence of PhTx compared to the baseline (absence of PhTx). c, Data as in b at 1.5 mM [Ca²⁺]_e. d, Representative traces as in a. e, Average mEPSPs and quantal content as in b. f, Representative traces as in a for indicated genotypes alone (GluRIIA^+/+) or for indicated genotypes combined as double mutants with the GluRIIA mutation (GluRIIA^−/−). g, Average data for each genotype compared to the genotype control in the absence of the GluRIIA mutation. **P < 0.01; NS, not significant; two-tailed Student’s t-test, pairwise comparison to genotype control. Data are mean ± s.e.m. and individual data points are shown.

PowerPoint slide

We subsequently investigated the long-term maintenance of PHP and the involvement of other semaphorin or Plexin gene family members. Deletion of a non-essential glutamate-receptor subunit (GluRIIA) induces a long-lasting form of PHP¹. We find that long-term PHP is blocked in a sema2b;GluRIIA double mutant as well as in GluRIIA larvae expressing transgenic RNA interference (RNAi) to knockdown PlexB selectively in motor neurons (Fig. 1f, g; see also below). Next, we separately tested the effect of mutations in all of the remaining semaphorin and Plexin genes encoded in the Drosophila genome (Extended Data Fig. 1b). The sema2b and PlexB mutants are the only mutants that show disruption of PHP.

We subsequently performed tissue-specific RNAi and transgenic rescue experiments. Expression of UAS-Sema2b-RNAi in motor neurons (OK371-Gal4) had no effect on PHP, whereas expression in muscle (BG57-Gal4) blocked PHP (Fig. 2a). In addition, expression of UAS-sema2b in muscle rescues PHP in the sema2b-mutant background (Fig. 2a; see Supplementary Table 1 for additional controls). Consistent with these data, we find that sema2b is expressed in muscle and that Sema2b protein, expressed under endogenous promoter sequences, concentrates at postsynaptic membranes (Extended Data Fig. 2). Next, we show that motor neuron-specific expression of UAS-PlexB-RNAi blocks PHP, whereas muscle-specific expression does not (Fig. 2b). We also show that motor neuron-specific expression of a previously characterized UAS-PlexB^DN dominant-negative transgene, lacking the intracellular signalling domain, blocks PHP (Fig. 2b). RNA-sequencing analysis of purified motor neurons demonstrates PlexB expression in motor neurons (data not shown). Finally, motor neuron-specific expression of a PlexB-myc transgene shows that PlexB traffics to the presynaptic nerve terminal (Fig. 2d). Taken together, our data indicate that Sema2b is a ligand originating in the muscle that acts via presynaptic PlexB to drive expression of PHP.

Figure 2: Sema2b and PlexB function as a retrograde trans-synaptic signal.

a, Representative traces recorded from the indicated genotypes (0.3 mM [Ca²⁺]_e). Quantification of mEPSP and quantal content (QC) as in Fig. 1b. Muscle Gal4 indicates muscle-specific BG57-Gal4/+. MN-Gal4 indicates motor neuron-specific OK371-Gal4/+. MN sema2b RNAi is Ok371-Gal4/+;UAS-sema2b-RNAi/+. Muscle sema2b RNAi is BG57-Gal4/+;UAS-sema2b-RNAi/+. Muscle sema2b rescue is sema2b;BG57-Gal4/UAS-sema2b. b, Traces, data are shown as in a. MN PlexB RNAi is OK371-Gal4/+;UAS-PlexB-RNAi/+. MN DN PlexB is Ok371-Gal4/+ expressing UAS-PlexB-DN. Muscle PlexB RNAi is BG57-Gal4/+,UAS-PlexB-RNAi/+. c, Muscle-specific expression (BG57-Gal4/+) of a UAS-Sema2b-CD8-GFP. Anti-HRP (blue) labels the neuronal membrane. Anti-Dlg (pink) labels postsynaptic membranes. d, UAS-PlexB-myc (green) expressed in motor neurons (OK371-Gal4/+). e, Traces for a sema2b mutant in the absence (baseline) and presence of PhTx. Sham incubation is saline without Sema2b. +Sema2b indicates addition of Sema2b protein (100 nM). f, Traces as in e for the PlexB mutant. g, Quantification of the data shown in e. h, Quantification of data shown in f. **P < 0.01; two-tailed Student’s t-test, pairwise to genotype control (a, b) or to wild type (g, h). Data are mean ± s.e.m. and individual data points are shown. Scale bars, 5 μm.

PowerPoint slide

If Sema2b is a retrograde signal that acts upon the presynaptic PlexB receptor, then it should be possible to reconstitute this retrograde signalling by acute application of Sema2b protein. Purified Sema2b protein was acutely applied to the neuromuscular junction (NMJ) of sema2b mutants following PhTx treatment to induce PHP. We found that Sema2b protein (100 nM) completely restores PHP in the sema2b mutant, but fails to restore PHP in the PlexB mutant (Fig. 2e–h; see Extended Data Fig. 3 for additional controls). In addition, application of Sema2b protein is sufficient to potentiate baseline release, and this effect is also dependent upon PlexB (Extended Data Fig. 3). Finally, a membrane-tethered UAS-sema2b transgene, expressed in muscle, fails to rescue PHP (Extended Data Fig. 4), even though it is concentrated on the postsynaptic membranes (Fig. 2c). Together, these results indicate that Sema2b is a secreted, postsynaptic ligand that acts upon presynaptic PlexB to enable the expression of PHP. We acknowledge the possibility that PlexB could require a presynaptic co-receptor of, as yet, unknown identity.

Given that acute application of Sema2b protein rescues PHP in the sema2b mutant, the failure of PHP in sema2b-mutant larvae cannot be a secondary consequence of altered NMJ development. Nonetheless, Sema2b–PlexB signalling is required for normal NMJ growth. Axon-targeting errors are rare at muscles 6/7, analysed at the third instar larval stage (Extended Data Table 1). We demonstrate that the NMJs in sema2b and PlexB mutants are composed of fewer, larger synaptic boutons (Fig. 3a–d) with no change in total NMJ area (Fig. 3b). The abundance of the active-zone-associated protein Bruchpilot (Brp) is unaltered in the sema2b mutant and the sema2b/+;;PlexB/+ double-heterozygous larvae (Fig. 3e, ‘trans-het’), both of which block PHP (Fig. 1a, d, e). There is a significant decrease in total Brp staining in the PlexB mutant, an effect of unknown consequence (Fig. 3e; see also below). Qualitatively, the ring-like organization of Brp staining was similar across all genotypes, indicative of normal active-zone organization (Fig. 3a, inset, arrows). Finally, there is no consistent difference in synapse ultrastructure across genotypes (Fig. 3f–i). Therefore, the Sema2b–PlexB-dependent control of bouton size may be a separate function of Sema2b–PlexB signalling, analogous to anatomical regulation by semaphorins in mammalian systems^{10,11,12,15,18}.

Figure 3: Altered NMJ growth with normal active-zone number and integrity.

a, Structured illumination microscopy images of NMJ. Inset, single confocal sections; arrows indicate Brp rings. Scale bars, 5 μm and 0.5 μm (inset). b–e, Quantification of morphology; n = 12, except mical^KG n = 10. f, Representative active zones. Scale bar, 70 nm. g–i, Quantification of the ultrastructure. *P < 0.05, **P < 0.01; two-tailed Student’s t-test, pairwise comparison to wild type larvae. Two larvae per genotype; wild-type, n = 16 active zones; sema2b, n = 29 active zones; PlexB, n = 30 active zones; mical, n = 13 active zones. Data are mean ± s.e.m. and individual data points are shown.

PowerPoint slide

PHP occurs through the potentiation of the readily releasable pool (RRP) of synaptic vesicles¹ (see Methods). Application of PhTx induces a doubling of the apparent RRP in wild-type larvae, an effect that is disrupted in both sema2b and PlexB mutants (Fig. 4). Failure to potentiate the RRP is also shown as a failure to maintain the cumulative EPSC amplitude after PhTx application (Fig. 4c–e). We subsequently show a strong genetic interaction with a mutation in the presynaptic scaffolding gene rab3-interacting molecule (rim), a PHP gene¹. Heterozygous mutations in rim, or in sema2b or PlexB have no effect on PHP (Fig. 4h, i). However, double-heterozygous combinations of rim/+ with either sema2b/+ or PlexB/+ strongly impaired the expression of PHP (sema2b/+,rim/+) or abolished PHP (rim/+;;PlexB/+) (Fig. 4h, i). These data do not, however, reflect direct signalling between PlexB and Rim (Extended Data Fig. 3e).

Figure 4: Sema2b, PlexB and Mical control the homeostatic potentiation of the RRP.

a, The mical mutants (green) in 0.3 mM [Ca²⁺]_e (left) and 1.5 mM [Ca²⁺]_e (right) and larvae overexpressing a Mical-resistant UAS-Act5C transgene (red) show blocked PHP. Df, deficiency. b, Expression of UAS-mical in motor neurons (MN mical rescue) rescues PHP. mical RNAi in motor neurons blocks PHP. mical RNAi in muscle has no effect (right). c, Representative traces from wild-type (black) and PlexB-mutant (blue) larvae (1.5 mM [Ca²⁺]_e) with or without PhTx (stimulation frequency 60 Hz, 30 stimuli). Cumulative EPSCs and back extrapolation from steady state (red line) is shown below each trace. d, Quantification of the percentage change in RRP (open) and mEPSPs (filled) for each genotype in the presence of PhTx compared to the genotypic baseline in the absence of PhTx. Sample size: wild-type, −PhTx, n = 14; wild-type, +PhTx, n = 9; sema2b mutant, −PhTx, n = 9; sema2b mutant, +PhTx, n = 9; PlexB mutant, −PhTx, n = 10; PlexB mutant, +PhTx, n= 11. e, Quantification of cumulative EPSCs for recordings in d. f, Quantification as in e for wild type −PhTx, n = 6; wild type, +PhTx, n = 6; mical mutant, −PhTx, n = 12; mical mutant, +PhTx, n = 11; and in larvae overexpressing dominant-negative Act5C −PhTx, n = 8 and +PhTx, n = 8. g, Quantification of cumulative EPSCs. h, Representative traces. i, Genetic interactions of sema2b/+ or PlexB/+ with the rim/+. *P < 0.05; **P < 0.01. Data are mean ± s.e.m. and individual data points are shown.

PowerPoint slide

To define how PlexB could modulate the RRP, we tested known downstream signalling elements. We discovered that mical is necessary for PHP (Fig. 4). In Drosophila a single mical gene encodes a highly conserved multi-domain cytoplasmic protein that mediates actin depolymerization, achieved through redox modification of a specific methionine residue (Met44) in actin^7,19. Notably, prior genetic evidence has placed Mical downstream of both PlexA and PlexB signalling during axon guidance²⁰.

An analysis of multiple mical mutations in larvae as well as transgenic rescue animals demonstrates that mical is necessary presynaptically for PHP (Fig. 4a, b, f, g). Mical protein is present presynaptically (Extended Data Fig. 5) and presynaptic expression of a Mical-resistant UAS-Actin5C transgene⁷, which interferes with Mical-mediated actin depolymerization, blocks PHP (Fig. 4a, b). This transgenic protein also concentrates within presynaptic boutons (Extended Data Fig. 5). Additional experiments reveal that the homeostatic expansion of RRP is blocked in mical mutants and when Mical-resistant UAS-Act5 is expressed presynaptically (Fig. 4f, g and Extended Data Fig. 6). We find strong genetic interactions between mical and both the PlexB and rim mutants (Extended Data Fig. 4b–d). Finally, anatomical experiments demonstrate that active zones are normal in the mical mutant, including in both light and electron microscopy experiments (Fig. 3). We propose that Mical enables PlexB-mediated control of the RRP through the regulation of presynaptic actin.

For half a century, evidence has underscored the importance of target-derived, retrograde signalling that controls presynaptic neurotransmitter release¹⁶. Gene discovery, based on forward genetics, indicates that PHP is controlled by the coordinated action of at least three parallel signalling systems (see Extended Data Fig. 7). If our data regarding Sema2b, PlexB and Mical can be generalized, then semaphorin–plexin signalling could represent a platform for retrograde, trans-synaptic, homeostatic control of presynaptic release, thereby stabilizing synaptic transmission and information transfer throughout the nervous systems of organisms ranging from Drosophila to humans.

Methods

Fly stocks and genetics

For all experiments, the w¹¹¹⁸ strain of Drosophila melanogaster was used as the wild-type control. Male and female larvae were used. Larvae were maintained at 22 °C. When performing rescue, RNAi or overexpression experiments with the Gal4/UAS expression system, progeny were raised at 25 °C. The following Drosophila stocks were used: sema2b^C4 (ref. 21), UAS-sema2b-RNAi (Bloomington stock 28932), UAS-sema2b-TM-GFP (ref. 21), UAS-sema2b flies were a gift from A. Kolodkin (Johns Hopkins University), P{GawB}Sema2b^NP0592 (Kyoto stock 112237), Sema2b^T:Ivir\HA1 (Bloomington stock 65752), P{SUPor-P}PlexB^KG00878 (ref. 20), UAS-Myc-PlexB^EcTM (ref. 21), UAS-Myc-PlexB (ref. 20), PlexA^EY16548 (Bloomington stock 23097), UAS-PlexB-RNAi (ref. 18), PlexA^Df(4)C3 (Bloomington stock 7083), GluRIIA (ref. 22), OK371-Gal4 (ref. 23), rim¹⁰³ (ref. 24), mical^K584 (ref. 6), Df(3R)swp2^mical mical deficiency¹⁹, mical^KG06518 (ref. 6), UAS-mical-RNAi (Bloomington stock 31148), UAS-Mical^mCherry (ref. 25), Mical-GFP Protein Trap (Bloomington stock number 60203), UAS-Act5C^{GFP M44L} (ref.7).

Electrophysiology

Recordings were made from muscle 6 in abdominal segments 2 and 3 of male and female third-instar larvae in current-clamp (0.3 mM [Ca²⁺]_e) or voltage-clamp (1.5 mM [Ca²⁺]_e) mode as indicated, without randomization^24,26,27. Haemolymph-like (HL3) saline was used (70 mM NaCl, 5 mM KCl, 10 mM MgCl₂, 10 mM NaHCO₃, 115 mM sucrose, 4.2 mM trehalose, 5 mM HEPES). Quantal content was calculated by dividing the average EPSP amplitude by the average mEPSP amplitude for each muscle recording (EPSP/mEPSP) and averages were made across muscles for a given genotype. For acute pharmacological induction of PHP, larvae were incubated in philanthotoxin-433 (PhTx; 10–20 μM; Sigma-Aldrich) for 10 min according to previously published methods^24,26. Two-electrode voltage-clamp recordings were done as previously described in 1.5 mM [Ca²⁺]_e HL3 saline²⁴. EPSC analyses were conducted using custom-written routines for Igor Pro 5.0 (Wavemetrics) available as published²⁸, and mEPSPs were analysed using Mini Analysis v.6.0.0.7 (Synaptosoft). Recordings were excluded if the resting membrane potential was more depolarized than −60 mV. Each experiment was repeated for at least two independent crosses. Key experiments demonstrating the blockade of synaptic homeostasis were performed by an independent investigator who was blinded to the genotype. All controls and experimental genotypes were independently replicated for each experiment in each figure. Experimental sample sizes equal to or greater than seven were considered sufficiently powered to detect a blockade in homeostatic plasticity, an effect size of ~80–120% compared to controls.

Anatomical analyses

Third-instar larval preparations (muscles 6/7) were filleted and fixed in 4% paraformaldehyde, washed and incubated overnight at 4 °C with primary antibodies. Secondary antibodies were applied at room temperature for 2 h. The following primary antibodies were used: anti-Myc (1:500, mouse; 9E10 Santa Cruz), anti-GFP 3E6 (1:500, mouse; Life Technologies), anti-NC82 (1:100, mouse; Developmental Studies Hybridoma Bank), anti-HA antibody (1:1,000; rabbit; Cell Signaling Technology) and anti-Dlg (1:10,000, rabbit; ref. 28). Alexa-conjugated secondary (488, 555) antibodies and Cy5-conjugated goat ant-HRP were used at 1:500 (Life Technologies; Molecular Probes). Larval preparations were mounted in Vectashield (Vector) and imaged with an Axiovert 200 (Zeiss) inverted microscope, a 100× Plan Apochromat objective (1.4 NA) and a cooled charge-coupled device camera (Coolsnap HQ, Roper). Slidebook 5.0 Intelligent Imaging Innovations (3I) software was used to capture, process and analyse images. Structured illumination microscopy imaging was performed using the N-SIM Nikon system, consisting of a Nikon Ti-E microscope equipped with a Apo TIRF 100×/1.49 oil objective and an Andor DU897 camera.

Sema2b ligand generation and application

We used the Drosophila S2 expression system to express the Sema2b-AP ligand²¹ for the bath application of Sema2b ligand to the Drosophila NMJ for in vivo electrophysiological recordings. Drosophila S2 cells (obtained from the Vale laboratory, UCSF for additional source information, see http://flybase.org/reports/FBtc9000006.html). Cells are mycoplasma negative, tested by MycoAlert (2017). We co-transfected UAS-sema2B-AP and actin-Gal4 plasmids in S2 cells using Effectene (QIAGEN) and incubated the cells at 27 °C for four days in serum-free Schneider’s medium. We then collected the medium and diluted the Sema2b-AP ligand in HL3 (0.3 mM Ca²⁺) to a concentration of 100 nM. To record from the NMJ in the presence of bath-applied Sema2b-AP ligand, we prepared the larval fillet as described above and incubated the preparation in Sema2b-AP HL3 for 10 min. To assess Sema2b-AP ligand rescue, we bath-applied HL3 (0.3 mM Ca²⁺) containing both PhTx (15 μM) and Sema2b-AP ligand (100 nM) to the preparation for a 10-min incubation. Next, we removed the PhTx/Sema2b-AP ligand HL3 saline and replaced it with HL3 containing only the 100 nM Sema2b-AP ligand to record in the presence of Sema2b.

Transmission electron microscopy

Transmission electron microscopy samples for all third-instar larvae were prepared and imaged according to methods that have been previously published²⁹.

Data availability

The datasets generated and/or analysed during the current study are available from the corresponding author upon reasonable request.

Extended data figures and tables

Extended Data Figure 1 Mutations in additional semaphorin and Plexin gene family members do not alter the rapid induction of PHP.

a, Gene diagrams indicating the mutations used in this study. Colours match scale bars. b, Quantification of the percentage change in mEPSPs (solid bars) and quantal content (open bars) in the presence of PhTx for the following genotypes: wild-type (w¹¹¹⁸), sema1a^K13702, sema1b^EY21782, sema2a^EY08184, sema5c^MI10577 and PlexA^EY1654. Each genotype was either previously described as, or predicted to be, a strong loss-of-function mutant: sema1a^K13702 (ref. 17), sema1b^EY21782(ref. 30), sema2a^EY08184 (ref. 31), sema5c^MI10577 (ref. 32) and PlexA^EY1654 (ref. 30). All recordings were made in 0.3 mM [Ca²⁺]_e. Data are mean ±s.e.m. *P < 0.05; **P < 0.01; two-tailed Student’s t-test.

Extended Data Figure 2 Evidence of sema2b gene expression in larval muscle.

a, Left, representative images of a sema2b-promoter–Gal4 fusion driving UAS-cd8-GFP at low magnification; images show multiple NMJs in the peripheral musculature, all expressing GFP. Muscles 4, 6 and 7 are labelled. Segmental boundaries are indicated by the horizontal lines and the middle segment is indicated as abdominal segment 3 (A3). Right, a higher magnification image taken of muscle 6/7, the muscles in which all recordings were made in this study, revealing expression of sema2b-promoter–Gal4. The NMJ is labelled with anti-Dlg (pink). Muscle identity is indicated. b, Images were taken at an identical exposure to those in a showing that there is no background GFP immunofluorescence in the absence of the UAS-cd8-GFP reporter as a control. Scale bars, 200 μm (left) and 10 μm (right). c, Data displayed as in a. Expression of sema2b-HA is controlled by endogenous promoter sequences. Scale bars, 200 μm (left) and 10 μm (right). d, Structured illumination, super-resolution microscopy was used to image Sema2b-HA expressed as in c. A single optical section (single plane) is shown revealing close proximity between Sema2b (green; anti-HA) and the presynaptic membrane (purple) labelled with anti-HRP. Scale bar, 2 μm.

Extended Data Figure 3 Effects of exogenous application of Sema2b protein on baseline transmission.

a, Raw data and analysis of additional control genotypes for the electrophysiological analysis of the effects of application of exogenous Sema2b protein (0.3 mM [Ca²⁺]_e). Wild type (wt; n = 5), wt + sham (n = 12), wt + Sema2b (n = 23), PlexB + sham (n = 7), PlexB + Sema2B (n = 6). b, A silver-stained protein gel of supernatant collected from S2 cells transfected with both Actin-Gal4 and UAS:Sema2b-AP, or the Actin-Gal4 plasmid alone (sham). The red box highlights that the Sema2b-AP ligand is present at the correct size when both plasmids were transfected together, but absent when the Actin-Gal4 plasmid is transfected alone (sham). Bottom, BSA standards. c, Representative traces (0.3 mM [Ca²⁺]_e). d, Raw data (0.2 mM [Ca²⁺]_e) for indicated genotypes without (filled bars) and with (open bars) application of exogenous Sema2b protein (100 nM protein). Application of Sema2b protein causes a 40% increase in quantal content in controls (n = 6) and this effect is blocked in larvae that overexpress a PlexB dominant-negative transgene in motor neurons (OE MN PlexB DN) (control, n = 6; OE MN PlexB DN, n = 14). e, Effects of applying Sema2b protein to the rim¹⁰³ mutant (compare rim baseline to rim with Sema2b). Experiments were performed in 0.4 mM [Ca²⁺]_e to achieve comparable levels of absolute baseline vesicle release to the experiments shown in a. Sema2b protein has no effect on mEPSP amplitudes, but potentiates both the average EPSP and quantal content in rim¹⁰³, demonstrating a significant (P < 0.05) potentiation of release. Sema2b protein rescues the blockade of PHP observed in the rim¹⁰³-null mutant. Application of PhTx reduces mEPSP amplitudes in rim¹⁰³ (P < 0.01) and there no significant (n.s.) increase in quantal content resulting in average EPSP amplitudes that are smaller than baseline. When Sema2b is co-applied with PhTx (rim + PhTx + Sema2b), the homeostatic potentiation of quantal content is significantly potentiated (P < 0.01) consistent with a rescue of PHP. (rim¹⁰³ baseline, n = 6; rim¹⁰³ + Sema2b, n = 13; rim¹⁰³ + PhTx, n = 7; rim¹⁰³ + PhTx + Sema2b n = 10). Data are mean ±s.e.m. *P < 0.05; **P < 0.01; two-tailed Student’s t-test.

Extended Data Figure 4 Genetic interactions.

a, Averaged mEPSP and quantal content in the absence and presence of PhTx for the indicated genotypes. Both genotypes (A and B) expressed the membrane-tethered UAS-sema2b-GFP (UAS-sema2b^TM-GFP) in muscle (BG57-GAL4). Expression of membrane-tethered UAS-Sema2b-GFP has no deleterious effects on neurotransmission or the expression of PHP in control larvae with a heterozygous mutation in the sema2b gene (sema2b/+) (n = 8 without PhTx and n = 8 with PhTx; genotype A). Muscle expression of membrane-tethered UAS-sema2b-GFP in the sema2b homozygous mutant background failed to rescue PHP (n = 8, n = 9; genotype B). This is in contrast to the observation that expression of wild-type UAS-sema2b in muscle fully restores PHP in the sema2b mutant background (Fig. 1). We conclude that a membrane-tethered Sema2b protein is unable to signal to the presynaptic terminal without being secreted from the postsynaptic membranes. These data are consistent with Sema2b being a secreted ligand, originating in muscle, for the induction and expression of PHP. b–d, Averaged mEPSPs and quantal content in the absence and presence of PhTx for the indicated genotypes. Heterozygous mutations in rim/+ (n = 8, n = 8; without and with PhTx, respectively), PlexB/+ (n = 9, n = 9) and mical^K584/+ (n = 8, n = 8; note mical^K584 shortened to mical^K5 in the figure) show normal PHP following PhTx-dependent inhibition of mEPSP amplitudes. Double-heterozygous combinations of rim/+ with mical^K584/+ (n = 9, n = 11) or PlexB/+ with mical^K584/+ (n = 8, n = 13) results in complete blockade of PHP. These genetic interactions indicate that mical, rim and PlexB all participate in a common process that is directly required for PHP. Data are mean ±s.e.m. *P < 0.05; **P < 0.01; two-tailed Student’s t-test.

Extended Data Figure 5 Synaptic localization of Mical and Act5C.

a, Transgenic expression of UAS-Act5C^M44L (middle; green in merge on the right) localizes throughout the presynaptic terminal marked with anti-HRP (left; magenta). Scale bar, 5 μm. b, The transgenic expression of UAS-mical-mCherry (green), used to rescue PHP in the mical mutant, localizes to the presynaptic boutons of motor neurons labelled with anti-HRP (magenta). Scale bar, 10 μm. c, Image of a Mical protein trap (see Methods) showing endogenous localization of Mical protein in the postsynaptic muscle and enrichment at the NMJ, which is labelled with anti-HRP (magenta). Projections in the z plane (y–z or x–z planes) indicate the presence of Mical protein within the presynaptic bouton of the protein trap. d, To selectively image presynaptic Mical–GFP in the protein-trap background, UAS-mical-RNAi was selectively expressed in muscle (BG57-GAL4) in the protein-trap background, greatly reducing muscle Mical–GFP and revealing strong presynaptic Mical–GFP originating from the protein trap at the endogenous gene locus. The NMJ is defined by anti-HRP (magenta, left) and by anti-Dlg (magenta, right).

Extended Data Figure 6 Representative data showing that homeostatic expansion of the RRP fails in sema2b and mical mutants and following expression of mutant UAS-Act5C.

Representative traces and graphs indicating cumulative EPSCs and back extrapolation from steady state (red line) for the indicated genotypes. a, Data are shown for wild type. b, Data are shown for sema2b. c, Data are shown for mical^KG. d, Data are shown for larvae expressing UAS-DN-Act5C. We note that sema2b mutants have a baseline synaptic transmission defect, with a smaller baseline initial EPSC and correspondingly smaller RRP compared to both wild-type larvae and PlexB mutants (Fig. 4). Because there is no change in the number of active-zone-associated vesicles in sema2b mutants (Fig. 3), this defect must reflect a change in the allocation of vesicles to the RRP at baseline that parallels the failure to homeostatically potentiate the RRP during PHP. However, because PlexB mutants also block PHP without a change in baseline RRP, it seems unlikely that there is a causal link between reduced baseline RRP and failed PHP in the sema2b mutant.

Extended Data Figure 7 A schematic of retrograde, trans-synaptic Sema2b–PlexB signalling.

Signalling is schematized in the context of other mechanisms that have been recently demonstrated to be necessary for PHP. Sema2b–PlexB signalling (red) is a coherent trans-synaptic, retrograde signalling system that is conveyed, via Mical, to modify presynaptic actin and potentiate the readily releasable synaptic vesicle pool. Other genes have been shown to be necessary for PHP, but none can be connected into a coherent, trans-synaptic signalling cascade. In brown, presynaptic Deg/ENaC channels are inserted into the presynaptic plasma membrane, causing sodium leak and potentiation of presynaptic calcium influx through presynaptic calcium channels (Ca_V2.1)³³. In blue, two components residing in the synaptic extracellular matrix have been implicated in PHP. The α2δ3 auxiliary subunit of the presynaptic calcium channel is necessary for PHP³⁴. The matrix-derived signalling protein Endostatin, a cleavage product of the collagen homologue Multiplexin, is also necessary for PHP³⁵. In orange, the innate immune receptor, peptidoglycan-recognition protein (PGRP), is essential for PHP²⁹. Signalling downstream of PGRP is hypothesized to reach the neuronal nucleus and could thereby mediate the long-term maintenance of PHP. A major task for the future will be to define how these diverse signalling mechanisms participate in a coordinated response that rapidly, accurately and persistently regulates presynaptic neurotransmitter release following disruption of postsynaptic glutamate receptor function. P, phosphorylation; TF, transcription factor; Ub, ubiquitination.

Extended Data Table 1 Axon guidance defects for type 1s and type 1b motor neurons

PowerPoint slides

PowerPoint slide for Fig. 1

PowerPoint slide for Fig. 2

PowerPoint slide for Fig. 3

PowerPoint slide for Fig. 4