Extracellular heparan sulfate proteoglycans and glycan-binding lectins orchestrate trans-synaptic signaling

Neuronal synapses cross-communicate via orchestrated intercellular glycoprotein signals passed between partner cells, with trans-synaptic ligands traversing the heavily glycosylated extracellular synaptomatrix separating presynaptic and postsynaptic cells. As a key genetic model synapse, the Drosophila glutamatergic neuromuscular junction (NMJ) is large, accessible and genetically malleable (Bai and Suzuki, 2020; Frank et al., 2020). Foundational work in Drosophila has revealed fundamental cellular mechanisms, starting with discovery of the universal rules of inheritance over a century ago. Recent NMJ work has revealed the critical importance of synaptic glycobiology, particularly the roles of secreted carbohydrate-binding lectins and heparan sulfate proteoglycan (HSPG) co-receptors within the extracellular synaptomatrix (Dear et al., 2017; Laaf et al., 2019). These extracellular glycoprotein signaling regulators are required for the tight control of the multiple forms of trans-synaptic ligand signaling, including both Wingless integration (Wnt) and bone morphogenetic protein (BMP) pathways (Kim et al., 2019; Kopke et al., 2020). As a disease model, the Drosophila NMJ has been critical for defining cellular mechanisms and suggesting potential therapeutic treatments. Neurological diseases are accurately recapitulated when orthologous Drosophila genes are mutated, and the accessible NMJ provides an experimentally tractable synaptic system (Menon et al., 2013). In addition, recent work has linked defects in extracellular synaptomatrix glycan mechanisms to numerous heritable disease states, including multiple congenital disorders of glycosylation (CDG), multiple galactosemias and Fragile X syndrome (FXS), the most common inherited form of both intellectual disability (ID) and autism spectrum disorder (ASD) (Friedman et al., 2013; Jumbo-Lucioni et al., 2016; Parkinson et al., 2016).

In this Review, we first briefly describe the Drosophila NMJ and the trans-synaptic signaling pathways that are regulated by membrane-anchored and secreted components of the synaptomatrix. This NMJ synaptic cleft is exquisitely narrow and lacks basal lamina, but is nevertheless dominated by highly organized extracellular glycans forming a dynamic synaptic intercellular interface (Dani et al., 2012; Scott and Panin, 2014). We then focus on the four extracellular HSPGs at this NMJ, including transmembrane, membrane-anchored glypican and secreted HSPGs, which have cell type-specific roles in trans-synaptic signaling that control NMJ architecture and neurotransmission function (Johnson et al., 2006; Kamimura et al., 2019; Ren et al., 2009). We next discuss three secreted lectins, including different C-type lectins and galectins, and their roles in regulating trans-synaptic signaling, synaptic strength and glutamate receptor localization (Rohrbough et al., 2013; Rushton et al., 2012). We then turn to synaptic regulation by extracellular enzymes, including secreted protease, sulfatase, glycase and deacylase enzymes. These secreted synaptomatrix enzymes control trans-synaptic signaling ligand distribution, co-receptor localization and cognate receptor availability at the Drosophila NMJ through synaptic activity-dependent mechanisms (Dani et al., 2012; Dear et al., 2016; Kopke et al., 2017). Finally, we consider the cellular sources of HSPGs and lectins in determining the directional control of NMJ trans-synaptic signaling. We end by discussing the intriguing challenges facing the field in unraveling the many extracellular glycan mechanisms working in concert within the synaptomatrix.

The Drosophila NMJ is an exquisite model for studying synapse structure and function owing to relative stereotypical connectivity, ease of manipulation and genetic malleability of conserved molecular mechanisms (Bai and Suzuki, 2020; Frank et al., 2020; Menon et al., 2013). In each abdominal hemisegment, 36 motoneurons in the ventral nerve cord (VNC) project into a three-layered array of 30 named multinucleate muscles (Fig. 1A). Each NMJ contains large motor axon varicosities (synaptic boutons) that house the molecular machinery for glutamate transmission. Small, budding satellite boutons develop into large, mature boutons containing multiple synapses (Fig. 1A). The commonly used horse radish peroxidase (HRP) antibody recognizes extracellular fucosylated N-glycans on neuronal membranes, and active zone (AZ) sites of synaptic vesicle (SV) fusion contain the presynaptic Bruchpilot (Brp) organizing scaffold, which links SVs and Ca²⁺ channels (Fig. 1A). SV cycling is widely studied with live dynamic fluorescent dye imaging and by electron microscopy ultrastructure (Kopke and Broadie, 2018). In muscle, organizing scaffolds of Discs large 1 (Dlg1), the homolog of mammalian postsynaptic density 95 (PSD-95; also known as DLG4), and two ionotropic glutamate receptor (GluR) classes (GluRIIA- and GluRIIB-containing receptors) (Fig. 1A,B) cluster at postsynaptic sites in the subsynaptic reticulum (SSR) membrane folds. Distinct mechanisms regulate the two GluR classes to modulate NMJ transmission. The precise AZ–GluR juxtaposition enables rapid communication across the synaptic cleft (Fig. 1A,B). The Drosophila glutamatergic NMJ contains many components that are conserved in mammalian glutamatergic synapses, including glycan mechanisms and the trans-synaptic signaling pathways that help direct synapse assembly (Menon et al., 2013; Scott and Panin, 2014).

Numerous trans-synaptic signaling pathways operate through the synaptomatrix. One cascade involves binding of neuronally secreted Jelly belly (Jeb) ligand to the muscle Anaplastic lymphoma kinase (Alk) receptor, which drives synaptic Ras–MAPK–ERK signaling (Rohrbough and Broadie, 2010; Rohrbough et al., 2013). More conserved BMP ligands act as retrograde signals in Drosophila and mammalian synapses (Regehr et al., 2009; Xiao et al., 2013). At the Drosophila NMJ, muscle-derived Glass bottom boat (Gbb) signals through the presynaptic type II BMP receptor (BMPR) Wishful thinking (Wit) together with the type I BMPR Thickveins (Tkv) or Saxophone (Sax) (Fig. 1B) (Aberle et al., 2002; McCabe et al., 2003). There is also newly reported non-canonical Wit–GluRIIA signaling, which does not involve a BMP ligand (Fig. 1B) (Kamimura et al., 2019). BMPR activation results in phosphorylation of Mothers against decapentaplegic (Mad), and phospho-Mad (P-Mad) regulates transcription of Fragile X mental retardation protein (Fmr1), which is a negative translational regulator of the microtubule-associated protein 1B (MAP1B) protein Futsch (Fig. 1B), as well as the Rac1 guanine-nucleotide exchange factor (GEF) Trio (Ball et al., 2010; Nahm et al., 2013; Zhang et al., 2001). Crimpy (Cmpy)-bound Gbb secreted from boutons regulates transmission strength, providing a directional signaling cue (Fig. 1B) (James et al., 2014). Any pathway mutations that disrupt retrograde Gbb signaling reduce NMJ size, whereas elevated Gbb increases NMJ growth with excess satellite boutons (Aberle et al., 2002; McCabe et al., 2003). Abelson (Abl) kinase and substrate Abl interactor (Abi) link Rac1 GTPase signaling to macropinocytosis-mediated BMPR downregulation to limit Gbb-mediated growth (Kim et al., 2019). Retrograde Gbb signaling results in local, branch-specific regulation of presynaptic growth in a synaptic activity-dependent manner (Fig. 1B) (Berke et al., 2013, 2019).

The most studied trans-synaptic signaling pathway involves secreted glycoprotein Wnt ligands, whose misregulation underlies many disease states (Nusse and Clevers, 2017; Steinhart and Angers, 2018). The first Wnt protein was discovered in Drosophila and named Wingless (Wg; mammalian Wnt-1), with six more subsequently described – Wnt2, Wnt4, Wnt5, Wnt6, Wnt8 (WntD) and Wnt10 (Table 1). Of these, Wg, Wnt2 and Wnt5 have known roles in the Drosophila NMJ (Liebl et al., 2008, 2010; Packard et al., 2002). Wg secreted from presynaptic boutons (Packard et al., 2002) and glia (Kerr et al., 2014) binds to Frizzled-2 (Fz2) receptors in both autocrine (neuron) and anterograde (muscle) signaling to control NMJ bouton number, postsynaptic GluRs and transmission strength (Fig. 1B) (Ataman et al., 2008; Kerr et al., 2014; Packard et al., 2002). Canonical Wnt signaling components are not present at the NMJ (Messéant et al., 2017). Instead, Wg signaling occurs via (1) the presynaptic divergent canonical pathway and (2) the postsynaptic non-canonical pathway of Frizzled nuclear import (FNI) (Fig. 1B). Presynaptically, Wg binds Fz2 and the low-density lipoprotein receptor-related protein (LRP) homolog Arrow (Arr) receptors (Fig. 1B), which act through the glycogen synthase kinase 3β (GSK3β) homolog Shaggy (Sgg) and the MAP1B homolog Futsch to control microtubule stability, regulating new bouton formation (Fig. 1B). Postsynaptically, Wg binding to Fz2–Arr causes internalization of the complex, Shank-dependent cleavage of the Fz2 C-terminus (Fz2-C), and import of Fz2-C fragment into muscle nuclei (see Fig. 3A) (Ataman et al., 2008; Harris et al., 2016). Fz2-C in ribonucleoprotein (RNP) granules, together with synaptic transcripts, are exported for local translation control (Fig. 3A) (Mathew et al., 2005; Packard et al., 2015; Speese et al., 2012). Our focus is on how trans-synaptic signaling pathways are regulated within the extracellular synaptomatrix.

A key type of extracellular regulators of trans-synaptic signaling are the HSPGs, which are defined by the attachment of heparan sulfate (HS) glycosaminoglycan (GAG) chains to specific serines of a core protein (Fig. 2A) (Grobe, 2014; Sarrazin et al., 2011; Zhang et al., 2018). HS-GAG chains are linear polysaccharides of repeated disaccharide units (uronic acid, glucosamine) that are formed by the addition of xylose, establishing a tetrasaccharide linker to serine, to which N-acetylglucosamine (GlcNAc) is then added (Fig. 2A). The nascent HS-GAG chain contains both glucuronic acid (GlcA) and GlcNAc, which can be further modified by numerous Golgi enzymes (Fig. 2A); these include the N-deacetylase/N-sulfotransferase (NDST) enzymes (Drosophila Sulfateless; Sfl), the HS C5-epimerase (Drosophila Hsepi), and multiple sulfotransferases that add sulfates to GlcA (Hs2st) and GlcNAc (Hs6st, Hs3st-A and -B) (Dani et al., 2012; Grobe, 2014). In the resulting HS-GAG chain, modified N-sulfated domains alternate with unmodified N-acetylated GlcA or GlcNAc domains (Fig. 2A) (Sarrazin et al., 2011). In Drosophila, knockdown of the muscle-specific Sfl reduces muscle size, whereas global loss of Sfl interferes with Wg signal transduction (Kamimura et al., 2013). At the NMJ, sfl mutants exhibit a decrease in spontaneous SV fusion rates, but elevated evoked transmission, with the appearance of enlarged postsynaptic pockets, similar to a wg-loss phenotype (Ren et al., 2009). In Hs6st mutants, extracellular Wg levels at the NMJ are elevated and downstream postsynaptic FNI signaling is increased, and the mutants show increased NMJ bouton numbers and decreased transmission (Dani et al., 2012). These findings demonstrate that the HSPG sulfation state is important for both localization of Wg ligand and downstream NMJ signaling, which is crucial for synapse development and synaptic activity-dependent plasticity, and ultimately for coordinated movement behavior. The number of sugar residues, uronic acids and sulfates can vary enormously in HSPGs (Fig. 2A), resulting in extensive heterogeneity (polydispersity) and diverse ligand-binding specificities that control trans-synaptic signaling, as discussed below.

HSPGs can be grouped by location (Figs 2B and 3) into glycosylphosphatidylinositol (GPI)-anchored glypicans, transmembrane syndecans and the more disparate secreted HSPGs, classically including Agrin, perlecans and type XVIII collagen (Sarrazin et al., 2011). Glypicans have an ectodomain containing 12 conserved cysteine residues and multiple GAG insertion sites. There are six mammalian glypicans (GPC1–GPC6) compared to just two in Drosophila; Dally and Dally-like protein (Dlp). Only Dlp occurs at the NMJ (Figs 2B and 3). Glypicans function as co-receptors altering ligand–receptor distribution and binding, with both stimulatory and inhibitory effects (Capurro et al., 2014). Dally and Dlp stabilize extracellular Wg (Kirkpatrick et al., 2004). In addition, the Dlp co-receptor has the dual function of repressing short-range signaling and activating long-rang signaling dependent on its levels, Wg ligand level and Fz2 receptor level (the so-called ‘exchange factor model’; Yan et al., 2009). At the NMJ, Dlp surrounds boutons (Fig. 3B) to control bouton number, synaptic AZ formation and transmission strength (Fig. 3A) (Friedman et al., 2013; Johnson et al., 2006; Kamimura et al., 2019). Dlp binds the receptor protein tyrosine phosphatase (RPTP) leukocyte common antigen-related (Lar) to antagonize Enabled signaling (Fig. 3A) (Johnson et al., 2006; Kim et al., 2019). The NMJ abundance of Dlp is decreased in Hs6st mutants (Dani et al., 2012), and Dlp expression and function is synaptic activity dependent (Dear et al., 2017). Dlp is a Wg co-receptor that modulates Wg binding to Fz2, and therefore downstream FNI signaling (Fig. 3A) (Dear et al., 2016, 2017). Defects in the Dlp regulation of Wg signaling have been shown to be causative in FXS, CDG and galactosemia disease models (Friedman et al., 2013; Jumbo-Lucioni et al., 2014, 2016; Parkinson et al., 2013, 2016). Drosophila also has a single transmembrane Syndecan (Sdc), which contains several ectodomain GAG-insertion sites (Fig. 2B). Sdc has been reported on both presynaptic and postsynaptic sides of the NMJ, with binding to Lar promoting formation of new synaptic boutons (Fig. 3A,B) (Johnson et al., 2006; Nguyen et al., 2016; Fox and Zinn, 2005). Both Dlp and Sdc are upregulated in the Drosophila FXS disease model, causing expanded NMJ architecture and elevated transmission function consistent with elevated Wg signaling (Friedman et al., 2013).

Fully secreted HSPGs at the NMJ include perlecan and Carrier of wingless (Cow) (Fig. 2B). In Drosophila, perlecan is encoded by terribly reduced optic lobes (trol), which binds to multiple growth factors and basement membrane components, including laminin and collagen IV (Kamimura et al., 2013). At the postsynaptic side of the NMJ, Trol resides close to the muscle surface (Fig. 3A,B), more specifically in the extracellular lumen of the SSR as revealed by immunoelectron microscopy (Kamimura et al., 2013). Null trol mutants have smaller muscles, reduced SSR and structurally abnormal NMJ boutons, with a decreased frequency and/or amplitude of spontaneous SV fusion events and reduced levels of GluRIIA receptors (Kamimura et al., 2013). Many of these defects are phenocopied in wg mutants; consistently, the postsynaptic FNI pathway is reduced in trol mutants (Fig. 3A), and genetically reinstating FNI signaling corrects most NMJ phenotypes (Kamimura et al., 2013). Conversely, presynaptic Wg signaling is reported to be increased in trol mutants, with genetically correcting presynaptic signaling able to suppress new bouton formation (Kamimura et al., 2013). Thus, Trol is hypothesized to balance bidirectional Wg signaling pre- and post-synaptically, primarily by sequestering Wg near the SSR (Fig. 3A, arrow). The most recently identified secreted HSPG is Cow (Chang and Sun, 2014), which contains a signal peptide (SP), as well as conserved Kazal, thyroglobulin and EF-hand Ca²⁺-binding domains (Fig. 2B). Both Kazal and thyroglobulin domains are predicted proteinase inhibitors (Chang and Sun, 2014). Local Ca²⁺ depletion from synaptic activity by extracellular Cow could potentially affect either its localization or function (Fig. 3A,B). Indeed, Ca²⁺ induces a large conformational change in the mammalian Cow homolog testican-2 (SPOCK2) (Table 1) (Nakada et al., 2001, 2003). As synaptic activity regulates Wg in a Ca²⁺-dependent manner (Ataman et al., 2008), a Ca²⁺-dependent conformational change could regulate Cow–Wg interactions in the extracellular synaptomatrix (Fig. 3A). Secreted Cow binds directly and specifically to Wg in the extracellular environment (Chang and Sun, 2014), suggesting a possible Ca²⁺-dependent regulation of Wg extracellular movement.

Wnt ligand hydrophobicity severely limits extracellular movement between cells (Routledge and Scholpp, 2019). A role for exosomes in Wnt extracellular transport was first reported for Wg at the NMJ (Korkut et al., 2009). In addition, actin-based filopodial cytonemes have been reported to transport mammalian Wnts between cells, as well as Wg in Drosophila (Stanganello et al., 2015). Depletion of the HSPG glypicans Dally or Dlp significantly reduces the expansion of cytonemes, with cytonemes only rarely detected in dally dlp double mutants (González-Méndez et al., 2017). These HSPGs mediate Wg signaling by stabilization and/or lateral diffusion of the extracellular ligand (Fig. 3A) (Yan et al., 2009). Wg also co-purifies with lipophorins, Drosophila homologs of lipoproteins, and lipophorin knockdown reduces Wg ligand extracellular movement (Panáková et al., 2005). Another factor that transports Wg is Secreted wingless-interacting molecule (Swim), which binds Wg to facilitate extracellular movement between cells (Mulligan et al., 2012). In mammals, secreted afamin similarly enhances Wnt secretion and/or its solubility (Mihara et al., 2016; Naschberger et al., 2017). Likewise, secreted Cow also binds to Wg to increase its extracellular mobility (Fig. 3A), a function that is conserved in human testican-2 (Table 1) (Chang and Sun, 2014; Yang et al., 2016). In sfl mutants, Cow lacks properly sulfated HS chains and fails to bind and transport Wg (Chang and Sun, 2014). At the NMJ, Cow secreted from presynaptic boutons negatively regulates Wg signaling to limit both bouton formation and neurotransmission strength (Fig. 3A,B) (Kopke et al., 2020). Cow-null mutants phenocopy increased presynaptic Wg, in that they show elevated NMJ bouton and synapse numbers, with greater evoked synaptic current amplitudes and SV fusion frequency. Furthermore, expression of a membrane-tethered Wg variant prevents these cow phenotypes, indicating that Cow mediates signaling of only secreted Wg (Kopke et al., 2020). These observations are all consistent with Cow sequestering Wg away from its neuronal secretion source to limit signaling at the synapse (Figs 3A and 5A). Thus, multiple extracellular HSPGs act as key signaling ligand co-receptors and transport agents in the NMJ synaptomatrix, working in concert with other secreted regulators that prominently include glycan-binding lectins.

Lectins are a large family of carbohydrate-binding proteins with diverse cellular roles (Gabius et al., 2016; Nio-Kobayashi, 2017), including crucial functions within the nervous system (Velasco et al., 2013; Higuero et al., 2017; Motohashi et al., 2017) and specifically at synapses (Singhal et al., 2012; McMorran et al., 2016; Karakatsani et al., 2017). Lectins are defined by the presence of one or two carbohydrate-binding domains (CBDs) (Fig. 4A,B), which bind to specific glycan targets (Modenutti, et al., 2019). Many lectins are secreted to bind glycoproteins and glycolipids in the extracellular space (Gabius et al., 2016), and thereby mediate intercellular communication, including at the Drosophila NMJ (Figs 4 and 5). At the NMJ, the C-lectin Mind-the-gap (MTG) is secreted from synaptic boutons to regulate postsynaptic GluRs (Rohrbough et al., 2007). C-lectins are defined by Ca²⁺-binding (Fig. 4A,B), suggesting a local Ca²⁺ depletion owing to synaptic activity might be sensed by extracellular C-lectins to affect glycan-binding function (Fig. 5A) or punctate localization (Fig. 5B). MTG organizes glycan distribution within the NMJ extracellular synaptomatrix that surrounds synaptic boutons, and so generates the highly patterned material that defines the synaptic cleft (Rushton et al., 2009, 2012). This function is presumably mediated by MTG binding to specific carbohydrates in glycan targets (Fig. 4B), but this has not been demonstrated, and the nature of the putative glycan binding partners in entirely unknown (Fig. 5A). MTG regulates trans-synaptic signaling through the presynaptically secreted Jeb ligand, whose binding to the postsynaptic Alk receptor drives Ras–MAPK–ERK signaling to negatively regulate neurotransmission function (Fig. 5A) (Rohrbough and Broadie, 2010; Rohrbough et al., 2013). Based on earlier findings, we performed a transgenic RNAi screen of other known lectins to systematically map their requirements for NMJ synaptic structure and function. This screen identified roles for Lectin-GalC1 and Galectin at the NMJ synapse (Dani et al., 2012).

Like MTG, Lectin-GalC1 is a C-lectin with Ca²⁺-binding activity, which interacts with extracellular β-galactoside glycans through its single CBD domain (Fig. 4A,B) (Keller and Rademacher, 2019). Lectin-GalC1 has a well-conserved signal peptide (SP) that drives vesicular secretion into the extracellular synaptomatrix surrounding synaptic boutons (Fig. 5A). RNAi knockdown of Lectin-GalC1 strengthens NMJ neurotransmission, demonstrating that LGC1 normally limits presynaptic function (Dani et al., 2012). Like Lectin-GalC1, galectins bind to β-galactoside glycans through CBDs, with two repeat domains, each containing a CBD, present in Drosophila Galectin (Fig. 4A,B). In mammals, there are 15 known galectins that bind multiple glycoconjugates (secreted glycoproteins and cell surface receptors) to control intercellular signaling (Vokhmyanina et al., 2012; Bum-Erdene et al., 2016). In neurons, mammalian galectin-1 regulates axonal growth, whereas galectin-3 binds to integrins, laminins and fibronectins, and galectin-4 controls axon-glia interactions (Díez-Revuelta et al., 2010, 2017; Yang et al., 2017). Importantly, mammalian galectin-3 has been directly implicated in Wnt-mediated intercellular signaling (Shimura et al., 2005; Song et al., 2009). Drosophila Galectin has highest sequence conservation with secreted galectin-4 (Table 1) (Pace et al., 2002), whose functions in multiple studied cell types include lipid raft stabilization, trafficking and intercellular adhesion (Cao and Guo, 2016). Galectin-4 is secreted from axons and interacts with the GPI-anchored cell adhesion molecule contactin-1 (Díez-Revuelta et al., 2017). Importantly, contactin-1 is a presynaptic responder for Wnt signaling, and the closely-related Drosophila Cortactin acts as a regulator of activity-dependent synaptic plasticity controlled by Wg signaling (Alicea et al., 2017). These mechanistic links suggest that Galectin (galectin-4) is a conserved extracellular synaptomatrix regulator of Wg (Wnt-1)-signaling in activity-dependent synapse modulation (Fig. 5A).

Drosophila Galectin has been shown to aid motor axon recognition of muscle synaptic targets, with a transgenic overexpression screen revealing NMJ targeting errors (Kurusu et al., 2008). At the NMJ, secreted Galectin is present in the extracellular space surrounding synaptic boutons within the synaptomatrix (Fig. 5A). Galectins multimerize and form complexes that crosslink their glycosylated ligands, giving rise to a dynamic extracellular lattice that regulates diffusion and compartmentalization of ligands and their receptors (Stewart et al., 2017; Modenutti et al., 2019). This extracellular Galectin lattice also regulates the distribution and signaling of cell surface receptors, including cadherins and integrins (Cao and Guo, 2016; Higuero et al., 2017). The repeat dimer CBDs are critical for this cross-linking function (Fig. 4A,B), as the affinity of the galectin lattice for glycoprotein targets is proportional to the number of N-glycans and their branching, which is mediated by the Golgi N-acetylglucosaminyltransferases (Laaf et al., 2019). Although lectins are secreted into the extracellular synaptomatrix at the NMJ, it is not yet known whether they interact with signaling ligands (e.g. Jeb, Gbb and Wg) or their respective receptors and/or co-receptors, either by direct binding or more indirect interactions (Fig. 5A). The outstanding questions include the number of synaptic lectins, their pre- versus post-synaptic distribution, their division into structural or functional roles at the NMJ, functions in regulating the numerous trans-synaptic signaling cascades, as well as other glycan-binding roles at the NMJ synapse. In addition to Galectin, there are four other uncharacterized FlyBase genes that contain the Galectin signature of two concanavalin A-like lectin and/or glucanase domains (Fig. 4B) (Klyosov and Traber, 2012). Likewise, there are additional predicted C-lectin genes in the annotated Drosophila genome that resemble MTG and Lectin-GalC1. Further work is needed to fully investigate the roles of these synaptic lectins at the NMJ.

Secreted enzymes have vital roles in regulating trans-synaptic signaling through modulating HSPGs and lectins. One critical class are the matrix metalloproteinases (Mmps), which are regulated by tissue inhibitors of metalloproteinases (Timps) and cleave secreted and membrane targets (Chan et al., 2020). Drosophila contains two Mmps, secreted Mmp1 and GPI-anchored Mmp2, as well as a single secreted Timp (Fig. 5A) (Dear et al., 2016). All three factors co-dependently localize in the synaptomatrix, with Mmp1 and Mmp2 both restricting NMJ size and strength, and muscle-derived Timp limiting presynaptic architecture and SV cycling (Shilts and Broadie, 2017). Timp inhibits Mmp proteolysis, and thus determines Mmp enzyme dynamics around synaptic boutons (Fig. 5A). Rapid, activity-dependent bouton formation depends on secreted Mmp1 (Dear et al., 2017). Intriguingly, loss of both Mmps restores normal synapses in a reciprocal suppression mechanism (Fig. 5A) (Dear et al., 2016). Both Mmps control trans-synaptic signaling of Wg through Dlp, with Mmp loss resulting in misregulated Dlp localization, whereas genetically correcting Dlp level restores NMJ structure and function in mmp nulls (Dear et al., 2016). Dlp is a proteolytic target of at least Mmp2 (Fig. 5A) (Wang and Page-McCaw, 2014). Timp also restricts trans-synaptic Gbb signaling, and genetically correcting Gbb level in timp nulls alleviates NMJ defects (Shilts and Broadie, 2017). Similarly, inhibition of Mmp in timp nulls restores Gbb signaling and normal synaptic phenotypes. Binding of Mmp1 to Dlp HS chains recruits Mmp1 to NMJ boutons (Fig. 5A) (Dear et al., 2017). Aberrant Mmp synaptic function has been implicated in the Drosophila FXS model, with Mmps regulating Dlp to restrict trans-synaptic signaling, and NMJ defects in the disease model being restored by Mmp inhibition and genetic reduction of Dlp (Friedman et al., 2013; Siller and Broadie, 2011). Synaptic activity-dependent Mmp1 activation lost in the FXS model is restored by reducing Dlp levels, indicating that activity-induced Dlp-dependent control of Mmp generates defective NMJs in the FXS disease state (Dear et al., 2017).

Other secreted enzymes have similar functions at the NMJ. For instance, loss of a secreted sulfatase (Sulf1) that controls the sulfation state of Dlp and Sdc elevates trans-synaptic signaling of Wg and Gbb, which results in increased postsynaptic GluR density and elevated neurotransmission strength (Dani et al., 2012). Likewise, the loss of α-N-acetylgalactosaminyltransferases regulating O-glycosylation at the NMJ elevates synaptic assembly to increase neurotransmission (Dani et al., 2014), although the target glycosylated proteins have yet to be identified. The secreted protein Notum (also known as Wingful because it is a negative feedback regulator of Wg) was initially suggested to modify Dlp HS chains (Giráldez et al., 2002), or else act as a GPI anchor phospholipase to release Dlp (bound to Wg) from cell surfaces (Kreuger et al., 2004; Traister et al., 2008). Although the crystal structures of Drosophila and human Notum confirm binding to Dlp, they suggest that Notum acts as a carboxylesterase that removes the palmitoleic acid of palmitoylated Wg to deacylate the signaling ligand (Fig. 5A) (Kakugawa et al., 2015). Wg palmitoylation potentiates binding to the Fz2 receptor, and wg mutants that lack this palmitoylation show severely reduced Wg–Fz2 signaling (Janda et al., 2012; Tang et al., 2012). At the NMJ, Notum is secreted from postsynaptic muscle and glia to negatively regulate Wg signaling (Fig. 5A) (Kopke et al., 2017; Kopke and Broadie, 2018). Notum-null mutants show upregulated levels of extracellular Wg ligand and postsynaptic FNI signal transduction. The resulting misregulation of downstream NMJ synaptic assembly causes defects in NMJ structure and function that phenocopy elevated Wg signaling (Fig. 5A). Consistent with this, synaptic phenotypes are suppressed by genetically correcting Wg levels (Kopke et al., 2017). Interestingly, Notum and the secreted HSPG Cow work together to negatively regulate trans-synaptic Wg signaling (Fig. 5A), thereby limiting both NMJ size and neurotransmission strength (Kopke et al., 2020).

Notum cleaves the palmitoleic acid of Wg in an HSPG-assisted mechanism, with Dlp presumably providing an organizing scaffold for Notum and Wg in the extracellular synaptomatrix (Fig. 5A). This hypothesis could be tested with a Dlp lacking HS chains (Dlp^-HS) (Yan et al., 2009) to assay Notum localization near Wg. Sdc promotes synaptic bouton formation, and Dlp restricts AZ number (Fig. 3A) (Johnson et al., 2006). Dlp overexpression enhances the elevated bouton number in sdc mutants, suggesting an interaction between HSPGs. Both Dlp and Sdc bind to the Lar receptor, but Dlp has greater affinity, so a pre-bound Sdc–Lar complex might stimulate NMJ growth before its inhibition by Dlp–Lar binding (Fig. 3A) (Johnson et al., 2006). This proposed mechanism could provide a time- and/or HSPG concentration-dependent switch from the formation of synaptic boutons to the generation of functional presynaptic active zones (Fig. 3A). Moreover, HSPG HS chain composition may be a vitally important determinant of NMJ function (Fig. 3A). Nascent GlcA/GlcNAc chains can be substantially modified between the N-sulfated and N-acetylated domains (Fig. 2A) (Sarrazin et al., 2011). Specific Wnts differentially associate with N-sulfated domains to control the signaling ligand distribution within the extracellular space (Mii et al., 2017), including Wg ligand levels and localization at the Drosophila NMJ (Kamimura et al., 2013). Testican-2 (the mammalian homolog of Cow) blocks MMP inhibition to activate proteolytic enzyme activity (Table 1) (Nakada et al., 2001, 2003). Thus, Cow may similarly function to promote synaptic Mmp activity, which in turn is predicted to restrict Wg trans-synaptic signaling (Fig. 5A) (Dear et al., 2016). This hypothesis could be tested by overexpressing Cow and performing in situ zymography (an enzymatic activity assay) to measure in vivo NMJ proteolytic activity (Shilts and Broadie, 2017). If Cow does promote synaptic Mmp function, an increase in Mmp reporter fluorescence and decreased Wg signaling are both predicted (Dear et al., 2017). These experiments are needed, but have not yet been conducted.

Double heterozygote null cow/+; notum/+ mutants exhibit synergistic synaptic defects, indicating Notum and Cow function together to negatively regulate Wg signaling (Fig. 5A) (Kopke et al., 2020). This raises the question of why the NMJ needs multiple extracellular negative Wg regulators? One reason could be the separable functions of Wg regulation based on cellular sources (i.e. neuron, muscle or glia). For example, Notum is secreted from muscle and glia, whereas Cow is only secreted from neurons (Kopke et al., 2017, 2020). Accordingly, muscle-targeted Notum knockdown leads to changes in synaptic bouton number and neurotransmission strength, whereas glia-targeted Notum knockdown only results in changes in the NMJ structure without any functional phenotype (Kopke et al., 2017). Although Cow and Notum both negatively regulate bouton number and neurotransmission strength, only Cow regulates developing satellite boutons, whereas only Notum regulates bouton segregation, with both of these phenotypes associated with Wg overexpression (Kopke et al., 2017, 2020). Consistent with this, earlier studies showed that cellular ligand source is important for trans-synaptic signaling function (Kerr et al., 2014; Packard et al., 2002). For example, neuronal- versus glial-derived Wg regulates distinct NMJ properties, with blocking of neuronal Wg resulting in decreased synaptic bouton number, but no change in SV fusion frequency (Packard et al., 2002), whereas inhibiting glia-derived Wg has no effect on NMJ bouton number, but increases the SV fusion frequency (Kerr et al., 2014). Moreover, Gbb ligand secreted from motor neurons only is Crimpy-tagged based on cellular source (Fig. 1B), with neuron-derived Gbb regulating neurotransmission strength, and muscle-derived Gbb regulating NMJ growth and bouton formation (James et al., 2014). Further studies are needed to determine whether the full complement of different HSPGs and lectins are cell type specific, or interact with cell type-specific binding partners at NMJ synapses.

As discussed here, there are four known HSPGs (Dlp, Sdc, Trol and Cow) and three lectins (Mtg, Lectin-GalC1 and Galectin) working in concert at the Drosophila NMJ. Similar extracellular HSPG co-receptors and glycan-binding lectins operate at the mammalian NMJ, including both secreted and membrane-anchored classes. A major function of these extracellular regulators in the synaptomatrix is to control multiple trans-synaptic signaling pathways (e.g. Jeb–Alk, Gbb–Wit/Sax/Tkv, Wg–Fz2/Arr), but they have additional roles that impact synaptic architecture and neurotransmission strength. Similar BMP and Wnt signaling pathways operate at the mammalian NMJ, but their extracellular regulation mechanisms remain to be investigated. A critical question is whether extracellular regulator functions are overlapping, synergistic or antagonistic. To answer this key question, expression patterns need to be better tested with available genetic tools, and multiply mutant NMJ phenotypes assayed; for example, in trans-heterozygote combinations to probe for nonallelic noncomplementation. Presynaptic boutons may present transmembrane Sdc and secrete Cow and Mtg, and postsynaptic muscles express Dlp (and probably Sdc) and secrete Trol and Galn. The persistent uncertainties about their subsynaptic localization need to be tested in side-by-side comparisons. The cellular sources are likely important, but much work is needed to be able to elucidate NMJ interactions in time and space. Extracellular signaling interactions are also modified by many different synaptically secreted enzymes [e.g. Mmps (and Timp), Sulf1 and Notum] co-regulating the multiple trans-synaptic signaling pathways. In cow mutants, Wg accumulation phenocopies presynaptic Wg elevation, but should also negatively regulate postsynaptic Wg loss. This needs to be assayed directly. Membrane-tethered Wg prevents cow synaptic defects, but appears to be less efficient at signaling and should eliminate all postsynaptic signaling. However, tethered Wg may reach across the narrow synaptic cleft, or else signal via exosomes or cytonemes. Loss of Cow (or Notum) causes NMJ structure and function defects that phenocopy neuronal Wg overexpression because they work together. It is important to test whether Cow interacts with other HSPGs or lectins to establish the correct Wg distribution, both basally and in response to synaptic activity. Notum is a negative-feedback inhibitor of Wg signaling; high Wg induces Notum to turn off signaling. At the NMJ, Fz2-C binding to notum RNA would allow the postsynaptic cell to respond to increased Wg signaling by upregulating translation of Notum, thus generating a negative-feedback loop. These are just a few of the many critical research avenues waiting to be explored in this burgeoning NMJ model.