Neuronal activity drives FMRP- and HSPG-dependent matrix metalloproteinase function required for rapid synaptogenesis

See allHide authors and affiliations

Sci. Signal.  07 Nov 2017:
Vol. 10, Issue 504, eaan3181
DOI: 10.1126/scisignal.aan3181

Restraining MMPs or surface glypicans may treat FXS

The breakdown of the extracellular matrix, such as by matrix metalloproteinases (MMPs), helps cells to make protrusions and migrate. In neurons, this activity facilitates the growth and development of boutons that enable synaptogenesis, neuronal adaptability, and memory formation. MMP function is abnormally increased in patients with the intellectual disorder fragile X syndrome (FXS), whose neurons exhibit synaptic dysfunction and increased numbers of immature dendritic spines. In the FXS fly model, the abundance of the heparan sulfate proteoglycan receptor Dlp is also increased. Using flies, Dear et al. found that Dlp directly interacted with and recruited MMP1 to the neuronal cell surface, facilitating its secretion and activity underlying FXS-associated neurological phenotypes. Homologs of both exist in humans, suggesting a potential avenue for therapeutic development.


Matrix metalloproteinase (MMP) functions modulate synapse formation and activity-dependent plasticity. Aberrant MMP activity is implicated in fragile X syndrome (FXS), a disease caused by the loss of the RNA-binding protein FMRP and characterized by neurological dysfunction and intellectual disability. Gene expression studies in Drosophila suggest that Mmps cooperate with the heparan sulfate proteoglycan (HSPG) glypican co-receptor Dally-like protein (Dlp) to restrict trans-synaptic Wnt signaling and that synaptogenic defects in the fly model of FXS are alleviated by either inhibition of Mmp or genetic reduction of Dlp. We used the Drosophila neuromuscular junction (NMJ) glutamatergic synapse to test activity-dependent Dlp and Mmp intersections in the context of FXS. We found that rapid, activity-dependent synaptic bouton formation depended on secreted Mmp1. Acute neuronal stimulation reduced the abundance of Mmp2 but increased that of both Mmp1 and Dlp, as well as enhanced the colocalization of Dlp and Mmp1 at the synapse. Dlp function promoted Mmp1 abundance, localization, and proteolytic activity around synapses. Dlp glycosaminoglycan (GAG) chains mediated this functional interaction with Mmp1. In the FXS fly model, activity-dependent increases in Mmp1 abundance and activity were lost but were restored by reducing the amount of synaptic Dlp. The data suggest that neuronal activity-induced, HSPG-dependent Mmp regulation drives activity-dependent synaptogenesis and that this is impaired in FXS. Thus, exploring this mechanism further may reveal therapeutic targets that have the potential to restore synaptogenesis in FXS patients.


Developing and maintaining proper synaptic connectivity requires tightly coordinated intercellular signaling across the synaptic cleft between pre- and postsynaptic cells. This feat is achieved through selective expression of a range of cell adhesion molecules and secreted ligands within the highly specialized extracellular synaptomatrix that separates synaptic partners (1, 2). This cellular interface is highly adaptable (plastic), sculpted by use (activity), and constantly remodeled by secreted enzymes molding the dynamic environment (36). A principal challenge is to understand exactly how these extracellular forces are regulated to control activity-dependent synaptic development. A core family of secreted proteases, called matrix metalloproteinases (MMPs), occupies a key nexus of coordinated cell-to-cell and cell-to-matrix signaling to modulate synaptic architecture and neurotransmission strength (79). Conserved MMP domains include N-terminal secretory signal sequence, the cleavable prodomain, zinc-binding catalytic site, and the closely linked C-terminal hemopexin domain (10). Changes in MMP function regulate synaptogenesis and mediate activity-dependent remodeling, and MMP dysfunction is implicated in numerous neurological disorders including autism, epilepsy, addiction, and schizophrenia (1115). Thus, MMP-dependent control of the extracellular synaptomatrix interface appears critical for normal synaptic development and is implicated in a wide range of synaptic disease states.

The Drosophila glutamatergic neuromuscular junction (NMJ) synapse has been proven as an excellent model to interrogate synaptomatrix questions, with conserved extracellular mechanisms and well-defined activity-dependent synaptogenic processes (2, 1618). Compared to challenges of testing the 24 MMPs in mammalian models, Drosophila has only 2 MMPs, secreted Mmp1 and glycosylphosphatidylinositol (GPI)–anchored Mmp2 (1921), although one Mmp1 isoform is also GPI-anchored (22). Both Mmps regulate motor neuron axon defasciculation, with the matrix molecule frac (faulty attraction) identified as an Mmp2 substrate promoting motor axon targeting (23, 24). Both Mmps also regulate synaptogenesis in multiple contexts, including reshaping dendritic arbors (25, 26) and axonal terminals (27). At the NMJ, these two Mmps coordinate structural and functional synaptogenesis by heparan sulfate proteoglycan (HSPG)–dependent regulation of Wnt/Wingless (Wg) trans-synaptic signaling (7). In the Drosophila model of fragile X syndrome (FXS), aberrant HSPG- and Mmp-dependent Wnt/Wg signaling is causative for synaptogenic defects (28). FXS is an intellectual disability and autism spectrum disorder caused by the loss of fragile X mental retardation protein (FMRP), an mRNA-binding translational regulator (generally a repressor) with roles including activity-dependent synaptogenesis (2931). The mRNAs encoding HSPGs and Mmps are putative FMRP targets, and the corresponding protein products are more abundant in the Drosophila FXS model (28, 29, 32, 33).

HSPGs are particularly intriguing extracellular regulators of activity-dependent MMP localization and function (34, 35). The diverse, multifunctional HSPGs can be fully secreted, membrane-tethered, or transmembrane and consist of a core protein plus heparan sulfate (HS) glycosaminoglycan (GAG) chains. HSPGs are known to function as extracellular signaling platforms linking secreted ligands, receptors, and proteases, and MMPs can also proteolytically cleave HSPGs in a reciprocal HSPG-Mmp relationship (34, 3639). HSPGs are potent regulators of activity-dependent synaptic development and plasticity (4043). The ~17 HSPGs in mammals present considerable experimental challenges, but the Drosophila genome encodes only 5 HSPGs in total, with only 3 synaptic HSPGs (38, 40, 41). At the NMJ, the secreted HSPG perlecan mediates Wg signaling directionality and serves to restrict synaptic structural growth (44), and the transmembrane HSPG syndecan conversely promotes synaptic development (41), whereas the GPI-anchored HSPG Dally-like protein (Dlp) serves as a critical biphasic regulator of Wg signaling to modulate structural and functional synaptogenesis (40, 45, 46). Trans-synaptic Wg signaling also drives fast, activity-dependent synaptogenesis, promoting both structural growth and functional strengthening in response to changes in neural activity (4750). Activity-induced Wg signaling mediates rapid, new synaptic bouton formation (47), but a role for Dlp in this synaptogenic mechanism has not been explored.

Given the role of neuronal activity in controlling FMRP abundance and function (5154), with known requirements in activity-dependent synaptogenesis (28, 5560), we hypothesized a downstream mechanistic link between the membrane-anchored glypican Dlp and the Mmps driving activity-induced synapse modulation. Upon testing this hypothesis, we found that Mmp1, but not Mmp2, is required for rapid, activity-dependent synaptic bouton formation. Using both acute transgenic and ionic depolarization paradigms, we found that Mmp1 is increased in response to heightened activity, with both FMRP and Dlp required for this activity-dependent mechanism. Conversely, Mmp2 is reduced after neuronal stimulation. We found that Dlp colocalization with Mmp1 is enhanced by neuronal activity and that Dlp induces Mmp1 synaptic localization with heightened colocalization in use-activated synapses. We found that both FMRP and Dlp are required for this activity-dependent Mmp1 enhancement, but the FMRP requirement was bypassed by genetically reducing Dlp abundance in the FXS model. Overall, these results show that the HSPG glypican Dlp is a direct, positive regulator of basal and activity-dependent Mmp1 abundance at the synapse. Our study also reveals that inappropriate Mmp1 regulation in this FXS model can be rectified by reducing genetic expression of Dlp, suggesting that future work should examine whether targeting the human glypicans, HS-modifying enzymes, or another HSPG-MMP–paired mechanism might correct neuropathological synaptic defects in FXS patients.


Mmp1 is required for rapid, activity-dependent synaptic bouton formation

At the Drosophila NMJ, presynaptic bouton formation occurs rapidly in response to acute neuronal stimulation (47). The initial, immature, transitional “ghost boutons” contain synaptic vesicles and presynaptic markers, such as membrane marker horseradish peroxidase (HRP), but are largely devoid of presynaptic active zones and postsynaptic markers, such as discs large (DLG) (47, 6163). Activity-dependent ghost boutons develop into mature boutons and thus represent the immediate readout for activity-dependent synaptogenesis (47, 63, 64). To acutely increase activity, we genetically expressed the temperature-dependent ion channel dTRPA1 (transient receptor potential cation channel A1) in motor neurons (vglut-Gal4>UAS-dTRPA1) (6569) and then assayed new ghost bouton formation after a transient 1-hour temperature increase (Fig. 1). We used the mmp1Q112* genetic null allele to remove Mmp1 function and the mmp2W307* genetic null allele in trans to chromosomal deficiency Df(2R)BSC132 to remove Mmp2 function (21, 70). In each mmp mutant condition (mmp−/−), we compared four genotypes at both restrictive (18°C) and permissive (30°C) temperatures (a total of eight conditions): the driver controls (vglut-Gal4/+), experimental animals (vglut-Gal4>UAS-dTRPA1), mmp driver controls (vglut-Gal4/+, mmp−/−), and experimental mutants (vglut-Gal4>UAS-dTRPA1, mmp−/−).

Fig. 1 Mmp1 mediates rapid, activity-dependent synaptic bouton formation.

(A) NMJs colabeled for HRP and DLG after dTRPA1 stimulation in the indicated genotypes. White asterisks mark ghost boutons. Scale bar, 5 μm. Higher-magnification images of synaptic boutons are shown below. Scale bar, 2 μm. (B) Quantified ghost bouton number per terminal after dTRPA1 stimulation: vglut-Gal4/+ [number of NMJs (n) = 28; 1.54 ± 0.30], vglut-Gal4>UAS-dTRPA1 (n = 20; 7.4 ± 0.97), vglut-Gal4, mmp1Q112*/mmp1Q112* (n = 24; 1.54 ± 0.34), and vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112* (n = 24; 1.38 ± 0.33). Data are means ± SEM from three replicates. ***P < 0.001 by nonparametric analysis of variance (ANOVA) (Kruskal-Wallis) and Dunn’s multiple comparison posttest. Nonsignificant (P > 0.05) comparisons are not represented for (i) vglut-Gal4/+ versus vglut-Gal4, mmp1Q112*/mmp1Q112*; (ii) vglut-Gal4/+ versus vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112*; and (iii) vglut-Gal4, mmp1Q112*/mmp1Q112* versus vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112*. Temperature controls are shown in fig. S1.

Unstimulated controls (vglut-Gal4/+) or transgenic experimental animals reared at the dTRPA1-restrictive temperature (vglut-Gal4>UAS-dTRPA1) displayed very few ghost boutons (Fig. 1, A and B, and figs. S1, A and B, and S2, A and B), consistent with previous reports (47, 6163). In contrast, shifting vglut-Gal4>UAS-dTRPA1 animals to a dTRPA1-permissive temperature resulted in a significant increase in activity-dependent ghost bouton formation (Fig. 1, A and B, and fig. S2, B and C). We did not observe any difference in ghost bouton number after shifting vglut-Gal4/+ driver controls to the dTRPA1-permissive temperature, confirming that the temperature shift did not contribute to synaptogenesis. We next tested the Mmp1 requirement for ghost bouton formation. In all unstimulated conditions, ghost bouton number was completely indistinguishable between vglut-Gal4/+ control, vglut-Gal4>UAS-dTRPA1, vglut-Gal4/+, mmp1Q112*/mmp1Q112* null, and the vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112* animals (fig. S1, A and B). After acute neuronal stimulation, we did not detect any increase in ghost bouton formation in the absence of Mmp1 (Fig. 1, A and B). There was no detectable difference between the mmp1 null driver control and the mmp1 null dTRPA1-stimulated conditions. As a result, the mmp1 null stimulated condition (vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112*) had significantly reduced bouton numbers compared with the wild-type stimulated condition (vglut-Gal4>UAS-dTRPA1; Fig. 1, A and B).

We next tested whether Mmp2 is similarly required for activity-dependent ghost bouton formation (fig. S2). In all unstimulated conditions, ghost bouton number was indistinguishable between vglut-Gal4/+, vglut-Gal4>UAS-dTRPA1, vglut-Gal4/+, mmp2W307*/Df(2R)BSC132 null, and the vglut-Gal4>UAS-dTRPA1, mmp2W307*/Df(2R)BSC132 animals (fig. S2, A and B). After acute neuronal stimulation, we saw a significant increase in ghost bouton formation in both the vglut-Gal4>UAS-dTRPA1 (control) and vglut-Gal4>UAS-dTRPA1, mmp2W307*/Df(2R)BSC132 (mutant) conditions, with no significant difference detected between the groups (fig. S2, C and D). Consistently, activity-dependent ghost bouton formation in the mmp2 null dTRPA1-stimulated condition was significantly increased relative to both the driver (vglut-Gal4/+) and the mmp2 null [vglut-Gal4/+, mmp2W307*/Df(2R)BSC132] stimulated controls (fig. S2, C and D). Together, these results show that Mmp2 is not detectably involved in ghost bouton formation, whereas Mmp1 is required for this activity-dependent synaptic bouton formation.

Colocalized Mmp1 and Dlp exhibit positive activity-dependent regulation

We next tested whether activity regulates the abundance of synaptic Mmps using both genetic and ionic stimulation paradigms (64, 65, 71, 72). We labeled Mmp1 with well-characterized antibodies, using detergent-free conditions to mark extracellular protein (7, 21, 7375). Mmp1 abundance and spatial distribution were both increased by stimulation (Fig. 2). First, we again used dTRPA1 to transgenically increase activity and found that stimulation caused a significant >40% increase in the abundance of Mmp1 compared to that in matched controls (vglut-Gal4/+; Fig. 2, A and B). We repeated tests with a more restricted Gal4 driver (CcapR-Gal4), which expresses dTRPA1 in only a subset of motor neurons to provide an additional internal control (fig. S3). CcapR-Gal4 drives in the RP3 motor neuron innervating muscle 6 and 7 (NMJ m6/7) but is excluded from the motor neuron innervating muscle 4 (NMJ m4). At the dTRPA1-restrictive temperature (18°C), we found no difference in Mmp1 between driver controls (CcapR-Gal4/+) and CcapR-Gal4>UAS-dTRPA1 conditions at either NMJ m6/7 or m4 (fig. S3, A and B). Mmp1 was significantly increased by >50% at NMJ m6/7 after the heat-induced dTRPA1 channel activation (30°C) in CcapR-Gal4>UAS-dTRPA1 compared to that in the CcapR-Gal4/+ driver control (fig. S3, C and D). However, there was no change in Mmp1 abundance at NMJ m4 after dTRPA1 neuronal stimulation, consistent with the restricted CcapR-Gal4>UAS-dTRPA1 expression pattern (fig. S3, C and D).

Fig. 2 Synaptic Mmp1 is rapidly increased after acute neuronal stimulation.

(A) Images of NMJs after the denoted stimulation colabeled with HRP and Mmp1 in the indicated genotypes. Heat map shows Mmp1 alone (scale below). White outline, HRP bouton. Scale bars, 2 μm. (B) Quantification of synaptic Mmp1 intensity normalized to matched controls in (i) vglut-Gal4/+ (n = 10; 1.0 ± 0.07) versus vglut-Gal4>UAS-dTRPA1 (n = 13; 1.42 ± 0.1) and (ii) w1118 control (unstimulated, n = 56; 1.0 ± 0.03) versus w1118 stimulated (high [K+], n = 57; 1.54 ± 0.07). **P < 0.01, ***P < 0.001 by unpaired t test with Welch correction (dTRPA1) or Mann-Whitney U test (high [K+]). Data are means ± SEM from three replicates for each experiment. Additional controls are shown in fig. S3.

After the dTRPA1 transgenic trials, we tested Mmp1 abundance and distribution after depolarizing synaptic terminals with KCl (high [K+]) for varying time periods (2 to 60 min). We found that synaptic Mmp1 was highly increased after only 10 min of acute stimulation (Fig. 2, A and B) and therefore used this rapid stimulation paradigm for the remainder of our studies (64). Similar to the above dTRPA1 results, this acute depolarizing stimulation caused a highly significant >50% increase in Mmp1 at the synapse relative to unstimulated controls (Fig. 2, A and B). Although Mmp2 had no detectable role in activity-dependent bouton formation (fig. S2), we also tested whether Mmp2 is regulated by neuronal activity. We labeled extracellular Mmp2 under detergent-free conditions using our previously characterized polyclonal antibody (7) and quantified Mmp2 abundance at the synapse before and after the ionic depolarization paradigm (fig. S4). In contrast to Mmp1, synaptic Mmp2 was significantly decreased by ~35% after acute, high [K+] stimulation (fig. S4, A and B). Together, these results demonstrate that elevated neuronal activity causes concurrent strong Mmp1 increase and Mmp2 reduction at the synapse. Consistent with previous studies, the loss of Mmp2 may be a secondary consequence of increased Mmp1 function (7). The specific increase in Mmp1 supports its requirement in activity-dependent synaptic bouton formation.

We previously established a strong genetic interaction between Dlp and both Mmps at the NMJ (7). We therefore next tested whether Dlp is similarly regulated by activity (fig. S5). Dlp was labeled with a well-characterized antibody (28, 41, 76), as well as a transgenic green fluorescent protein (GFP) tag in the endogenous dlp locus (Dlp::GFP) (77, 78). After an acute increase in activity, NMJs displayed an immediate and striking increase in synaptic Dlp intensity (fig. S5A). With high [K+] depolarization, GFP labeling showed a significant >60% increase in Dlp::GFP, and Dlp labeling consistently revealed a significant >60% increase compared to the unstimulated control condition (fig. S5B). These results indicate that activity induces rapid up-regulation of Dlp at the synapse. In unstimulated terminals, both Mmp1 and Dlp appeared in concurrent domains in a dynamic subset of synaptic boutons (Fig. 3A, arrows), with overlap but also with spatial separation of the two proteins. Upon acute stimulation, both proteins were significantly increased within the same synaptic subdomains (Fig. 3A, top panels, arrows), and the incidence of colocalization became much greater (Fig. 3A, bottom panels, asterisks). Given that HSPGs bind and localize MMPs in other contexts (34, 36, 39), we hypothesized that activity stimulates the colocalization of Dlp and Mmp1 at the synapse, with GPI-tethered Dlp serving to capture and localize the secreted Mmp1.

Fig. 3 Mmp1 and Dlp colocalization in synaptic subdomains is increased by acute activity.

(A) Images of Dlp::GFP NMJs labeled with HRP, GFP, and Mmp1 under basal conditions or after high [K+] stimulation. Dlp::GFP and Mmp1 signals are shown as a heat map with HRP synaptic outlines in white. Arrows indicate overlapping Dlp::GFP and Mmp1 signals. Scale bar, 5 μm. Higher-magnification images of single optical sections are shown below. Scale bar, 1 μm. Asterisks denote overlapping Dlp::GFP and Mmp1, shown pseudocolored in white in the rightmost images. (B) Quantification of both Dlp::GFP and Mmp1 fluorescence intensities normalized to unstimulated controls. (i) “Dlp::GFP”: control (n = 20; 1.0 ± 0.06) versus stimulated (n = 19; 1.48 ± 0.09). (ii) “Mmp1”: control (n = 20; 1.0 ± 0.05) versus stimulated (n = 19; 1.56 ± 0.14). ***P < 0.001 by unpaired t test with Welch correction. (C) Quantification of MCCs. (i) MA (Mmp1/Dlp::GFP) in basal control (n = 27; 0.33 ± 0.03) and after stimulation (n = 27; 0.72 ± 0.03) and (ii) MB (Dlp::GFP/Mmp1) in basal control (n = 27; 0.37 ± 0.03) and after stimulation (n = 27; 0.59 ± 0.03). ***P < 0.001 by unpaired t test (MA) or Mann-Whitney U test (MB). Data are means ± SEM from three replicates. Additional controls are shown in figs. S5 and S6.

To test for this coregulation, we triple-labeled unstimulated and high [K+]–stimulated Dlp::GFP terminals with antibodies against GFP, Mmp1, and HRP. In parallel comparisons, we confirmed a significant ~50% increase in Dlp and Mmp1 after stimulation compared to unstimulated controls (Fig. 3, A and B). Moreover, acute stimulation increased the spatial overlap between Dlp and Mmp1, with specificity confirmed by multiple antibody and imaging controls (Fig. 3C and fig. S6). We assessed co-occurrence with Manders’ colocalization coefficient (MCC) measurements (79, 80). Upon stimulation, the Mmp1/Dlp MCC significantly increased more than twofold compared to unstimulated controls (Fig. 3C). Likewise, the Dlp/Mmp1 MCC showed a similar significant increase in stimulated terminals (Fig. 3C). Thus, with acute stimulation, both Dlp::GFP and Mmp1 MCC values were similarly increased compared to unstimulated controls, demonstrating a rapid, activity-dependent recruitment of both proteins to colocalized synaptic domains. These results have four possible interpretations: (i) Dlp and Mmp1 have unrelated activity-dependent increases within the same dynamic synaptic domains; (ii) secreted Mmp1 drives the activity-dependent Dlp increase; (iii) membrane-tethered Dlp drives the activity-dependent Mmp1 increase, recruiting Mmp1 to synaptic subdomains; or (iv) Dlp and Mmp1 activity-dependent increases are reciprocally codependent.

Dlp localizes Mmp1 in synaptic domains and promotes Mmp1 abundance

To first test the hypothesis that Dlp regulates Mmp1, we altered synaptic Dlp expression and quantified Mmp1 abundance. At the Drosophila NMJ, Dlp serves as a Wg co-receptor in a mechanism dependent on HS-GAG sulfation state, and Dlp also binds the receptor phosphatase leukocyte antigen–related-like (dLAR) through HS-GAG chains to modulate synaptogenesis (40, 41). Complete loss of dlp is embryonic lethal, so we used dlpA187/+ heterozygotes and dlpRNAi (70, 81). To reduce Dlp at the NMJ, we targeted dlpRNAi to both presynaptic (elav-Gal4) and postsynaptic (24B-Gal4) cells (elav-Gal4, 24B-Gal4>UAS-dlpRNAi). Both dlpA187/+ and dlpRNAi significantly reduced synaptic Dlp abundance by >45% (fig. S7, A and B). In parallel, dlpA187/+ significantly reduced Mmp1 intensity by >30% compared to background controls (Fig. 4, A and B), and dlpRNAi likewise significantly decreased Mmp1 intensity by >20% compared to Gal4 driver controls (Fig. 4, A and B). To conversely test whether Dlp overexpression reciprocally increases Mmp1, we postsynaptically overexpressed wild-type Dlp [24B-Gal4>UAS-dlpWT; (82)]. Excess Dlp caused a significant >50% increase in synaptic Mmp1 (Fig. 4, C and D). Moreover, the expanded Dlp domain caused a striking increase in perisynaptic Mmp1 localization (Fig. 4C). Together, these results indicate that Dlp positively regulates Mmp1 intensity and spatial distribution at the synapse.

Fig. 4 Synaptic Dlp positively and bidirectionally regulates secreted Mmp1 abundance.

(A and B) Images of NMJs (A) and quantification of Mmp1 fluorescence intensity (B) from the denoted dlp reduction conditions compared to matched controls colabeled with HRP and Mmp1. Mmp1 signal intensity is shown as a heat map with HRP synaptic outlines in white. Scale bars, 2 μm. Fluorescence intensity was normalized to that from matched genetic controls: (i) w1118 (n = 34; 1.0 ± 0.04) versus dlpA187/+ (n = 37; 0.67 ± 0.04), and (ii) elav-Gal4,24B-Gal4/+ control (n = 28; 1.0 ± 0.07) versus elav-Gal4,24B-Gal4>UAS-dlpRNAi knockdown (n = 32; 0.77 ± 0.05). *P < 0.05, ***P < 0.001 by Mann-Whitney U test. (C and D) As in (A) and (B), but from the denoted dlp overexpression conditions. Scale bar, 2 μm. 24B-Gal4/+ (n = 51; 1.0 ± 0.03), 24B-Gal4>UAS-dlpWT (n = 42; 1.55 ± 0.09), and 24B-Gal4>UAS-dlp−HS (n = 34, 1.13 ± 0.06). ***P < 0.001 by nonparametric ANOVA (Kruskal-Wallis) with Dunn’s multiple comparison posttest. The nonsignificant (P > 0.05) comparison for 24B-Gal4/+ versus 24B-Gal4>UAS-dlp−HS is not shown. Data are means ± SEM from at least three independent replicates for each experiment. Abundance of Dlp in each of the genetic manipulations is shown in fig. S7.

HSPG interactions are often mediated through HS-GAG chain binding, which can anchor interactors in close proximity, influencing both diffusion and clustering (83, 84). In addition, however, HSPG core proteins also have well-characterized binding functions (38, 46, 8587). Therefore, we next sought to test whether Dlp-dependent Mmp1 regulation is mediated through the Dlp core protein, HS-GAG chains, or both. To test Dlp HS-GAG chain requirements, we overexpressed an HS-deficient Dlp (UAS-dlp−HS), in which all five serine GAG attachment sites have been mutated to alanine (46). We reasoned that if the Dlp-Mmp1 interaction is mediated by the Dlp core protein, then overexpressing HS-deficient Dlp should phenocopy the synaptic Mmp1 increase caused by overexpressing wild-type Dlp (Fig. 4, C and D). We confirmed that the 24B-Gal4>UAS-dlpHS overexpression was comparable to the 24B-Gal4>UAS-dlpWT condition, with very high synaptic expression in both cases (fig. S7C). However, in stark contrast to the striking Mmp1 increase with wild-type Dlp overexpression, Dlp−HS overexpression did not cause any change in synaptic Mmp1 abundance compared to the Gal4 driver control (24B-Gal4/+; Fig. 4, C and D). When the activity-induced Mmp1 increase caused by wild-type Dlp was compared to Dlp−HS, there was a significant decrease (Fig. 4, C and D). These results strongly suggest that Dlp promotes synaptic Mmp1 localization through HS-GAG chain interaction.

Despite the lack of Mmp2 involvement in this synaptogenic mechanism, we also tested whether synaptic Mmp2 may be regulated by Dlp, independently or downstream of the Mmp1 changes. Dlp is a direct Mmp2 proteolytic substrate (70), and Mmp1 and Mmp2 are reciprocally coregulated at the NMJ synapse under both basal conditions and after neuronal stimulation (Fig. 2 and figs. S3 and S4) (7). To test whether Dlp regulates synaptic Mmp2, we again used dlpA187/+ heterozygotes to reduce Dlp and postsynaptic Dlp overexpression (24B-Gal4>UAS-dlpWT) to increase Dlp (fig. S7). We quantified Mmp2 abundance at the NMJ compared to matched controls under both conditions (fig. S8). We did not observe any change in synaptic Mmp2 in the dlpA187/+ heterozygous condition compared to w1118 genetic background control (fig. S8, A and B). Postsynaptic Dlp overexpression caused a subtle ~10% decrease in Mmp2 abundance at the NMJ (fig. S8, A and B). These data show that Dlp has a minor influence on synaptic Mmp2 but functions as a strong, bidirectional regulator of Mmp1 at the synapse.

Dlp is required for the rapid, activity-dependent increase of synaptic Mmp1

Because Dlp and Mmp1 are both increased by acute neuronal activity and Dlp regulates synaptic Mmp1 abundance, we next hypothesized that Dlp is required for the activity-induced Mmp1 increase at the synapse. To test this hypothesis, we genetically manipulated Dlp and quantified Mmp1 after acute, high [K+]–induced depolarization (Fig. 5). Supporting the above results, we found an activity-dependent increase in Mmp1 abundance in both w1118 background and Gal4 driver controls, with significant increases in all comparisons. Reducing Dlp expression abolished any detectable Mmp1 increase in response to the acute activity stimulation (Fig. 5, A and B). We found instead a small ~15% decrease in synaptic Mmp1 abundance in the stimulated dlpA187/+ heterozygotes relative to unstimulated dlpA187/+ controls (Fig. 5B). The dlpRNAi condition yielded very similar results, with a small, nonsignificant ~12% reduction in synaptic Mmp1 after the high [K+] stimulation (Fig. 5B). Conversely, the activity-stimulated Mmp1 abundance was significantly increased by a further ~65% with Dlp overexpression (Fig. 5, C and D). The basal amount of Mmp1 at the synapse was increased by Dlp overexpression (Fig. 4), yet there was still a strong increase in synaptic Mmp1 abundance after acute depolarizing stimulation, consistent with Dlp promoting the activity-dependent increase in Mmp1. Together, these results show that Dlp is required for the activity-dependent Mmp1 increase at the synapse.

Fig. 5 Dlp mediates activity-dependent regulation of synaptic Mmp1.

(A and B) Images of NMJs (A) and quantification of Mmp1 fluorescence intensity (B) from the indicated dlp reduction conditions treated with or without high [K+] and colabeled for HRP and Mmp1. Mmp1 signal intensity is shown as a heat map with HRP synaptic outlines in white. Scale bar, 2 μm. Quantified Mmp1 fluorescence intensity from stimulated conditions (bar graphs) was normalized to unstimulated controls (red lines). (i) w1118 control (n = 17; 1.0 ± 0.08) versus w1118 stimulated (n = 19; 1.34 ± 0.06); (ii) dlpA187/+ control (n = 19; 1.0 ± 0.03) versus dlpA187/+ stimulated (n = 21; 0.85 ± 0.06); (iii) elav-Gal4,24B-Gal4/+ control (n = 26; 1.0 ± 0.04) versus elav-Gal4,24B-Gal4/+ stimulated (n = 24; 1.34 ± 0.06); and (iv) elav-Gal4,24B-Gal4>UAS-dlpRNAi control (n = 25; 1.0 ± 0.06) versus elav-Gal4,24B-Gal4>UAS-dlpRNAi stimulated (n = 18; 0.88 ± 0.06). (C and D) As in (A) and (B), but from the dlp overexpression conditions. Scale bar, 2 μm. (i) 24B-Gal4/+ control (n = 21; 1.0 ± 0.06) versus 24B-Gal4/+ stimulated (n = 18; 1.39 ± 0.07); (ii) 24B-Gal4>UAS-dlpWT control (n = 21; 1.0 ± 0.03) versus 24B-Gal4>UAS-dlpWT stimulated (n = 16; 1.64 ± 0.16); and (iii) 24B-Gal4>UAS-dlp−HS control (n = 23; 1.0 ± 0.05) versus 24B-Gal4>UAS-dlp−HS stimulated (n = 13; 1.19 ± 0.08). (B and D) For stimulated versus unstimulated pairwise comparisons (red asterisks), significance was determined by Mann-Whitney U tests, and significance across genotypes (black) was determined by unpaired t tests (B) or nonparametric ANOVA (Kruskal-Wallis; D) with Dunn’s multiple comparison posttest, as indicated by *P < 0.05, **P < 0.01, and ***P < 0.001. Nonsignificant (P > 0.05) comparisons for (i) stimulated versus unstimulated elav-Gal4,24B-Gal4>UAS-dlpRNAi (B), (ii) stimulated 24B-Gal4/+ versus stimulated 24B-Gal4>UAS-dlpWT (D), and (iii) stimulated 24B-Gal4/+ versus stimulated 24B-Gal4>UAS-dlp−HS (D) are not shown. Data are means ± SEM from three independent replicates.

Previous works have established that HS-GAG chains often mediate Dlp functions and binding interactions (41, 46, 82). Therefore, we next tested whether an HS-GAG chain–deficient Dlp (UAS-dlp−HS) interferes with the activity-dependent synaptic Mmp1 induction (Fig. 5, C and D). Although the endogenous Dlp with intact HS-GAG chains was still present, we hypothesized that the HS-GAG chain–deficient Dlp should dominantly dampen activity-induced Mmp1 elevation. After acute, high [K+]–induced depolarization, we observed increased synaptic abundance of Mmp1 in both 24B-Gal4/+ driver control and 24B-Gal4>UAS-dlpWT overexpression conditions, with a significant increase in both conditions (Fig. 5, C and D). However, the synaptic Mmp1 abundance in the activity-stimulated 24B-Gal4>UAS-dlp−HS synapses was only slightly increased (by 19%) compared to the matched unstimulated controls, a much weaker albeit still significant response (Fig. 5, C and D). In sharp contrast, the activity-induced Mmp1 increase was significantly reduced at 24B-Gal4>UAS-dlp−HS NMJ synapses compared with the wild-type Dlp overexpression condition (19% versus ~65%, respectively). Together, these results support the conclusion that both basal and activity-induced Mmp1 regulation by Dlp at the synapse is directly mediated through HS-GAG chain–binding interactions in the extracellular space.

At NMJs lacking Mmp1 (mmp1 null mutants), basal Dlp abundance is reduced and Dlp is more spatially restricted. Thus, Mmp1 reciprocally serves as a positive regulator of synaptic Dlp (7). We consistently found above that Dlp is required for the activity-dependent Mmp1 increase at the synapse (Fig. 5). However, there still might be a reciprocal Mmp1 requirement for the activity-dependent increase in Dlp abundance. To test for this putative coregulation, we next genetically reduced Mmp1 abundance using two independent approaches and then quantified Dlp intensity after acute, high [K+]–induced depolarization (fig. S9). Consistent with the above results, control animals displayed a significant >30% increase in synaptic Dlp after the acute stimulation (fig. S9, A and B). Using the combinatorial pre- and postsynaptic mmp1RNAi knockdown (elav-Gal4, 24B-Gal4>UAS-mmp1RNAi), synaptic Mmp1 was reduced by ~70%, yet we still observed the significant increase in Dlp after depolarizing stimulation (fig. S9, C and D). We validated these results in a trans-heteroallelic mmp1 mutant condition (mmp1Q112*/mmp1Q273*) and again observed a similar, maintained >30% increase in synaptic Dlp after acute stimulation (fig. S9, A and B). These results indicate that Mmp1 is not reciprocally required for the activity-dependent Dlp increase at the synapse. Together, the data suggest that Mmp1 is downstream of Dlp in this activity-dependent mechanism.

Synaptic Dlp bidirectionally determines proteolytic function at the NMJ

After discovering that synaptic Mmp abundance and distribution are dependent on Dlp (Figs. 4 and 5), we next assayed whether these changes correlate with synaptic proteolytic activity. We used in situ zymography at the NMJ to measure the metalloproteinase-dependent conversion of a dye-quenched fluorogenic gelatin (DQ-gelatin) and quantified fluorescence changes in both the dlp reduction and overexpression conditions used above (33, 88). Although this method cannot differentiate the contributions of different proteases, it provides a live readout of net enzymatic activity through quantifiable fluorescence changes. Consistent with the above results, dlpA187/+ heterozygotes showed a significant ~35% reduction in proteolytic activity at the synapse compared to w1118 controls (Fig. 6, A and B). Moreover, Dlp overexpression (24B-Gal4>UAS-dlpWT) resulted in a significant ~50% increase in proteolytic activity at the synapse (Fig. 6, A and C). We also tested whether the Dlp-mediated increase in proteolytic function required HS-GAG chains as above (24B-Gal4>UAS-dlp−HS). These NMJs still displayed a significant ~60% increase in proteolytic activity compared to the matched driver control (24B-Gal4/+), with no detectable difference between DlpWT and Dlp−HS overexpression conditions (Fig. 6, A and C). Together, we conclude that Dlp abundance is a strong determinant of proteolytic gelatinase activity at the synapse, consistent with the above measurements of Dlp-dependent synaptic Mmp1 abundance, with the enzymatic function reliant upon the Dlp core protein.

Fig. 6 Dlp positively and bidirectionally regulates proteolytic activity at the synapse.

(A) Images of NMJs from the indicated dlp reduction and overexpression conditions colabeled for HRP (red) and in situ zymography activity (green) compared to matched genetic controls. Gelatinase activity is shown as a heat map with HRP synaptic outlines in white. Scale bar, 2 μm. (B) Quantified in situ zymography fluorescence intensity normalized to controls for the dlp reduction: w1118 control (n = 35; 1.0 ± 0.06) and dlpA187/+ heterozygote (n = 33; 0.66 ± 0.04). Significance was determined by unpaired t test with Welch correction, as indicated by ***P < 0.0001. (C) Quantified in situ zymography fluorescence intensity normalized to controls for the dlp overexpression conditions: 24B-Gal4/+ control (n = 35; 1.0 ± 0.05), 24B-Gal4>UAS-dlpWT (n = 18; 1.5 ± 0.09), and 24B-Gal4>UAS-dlp−HS (n = 20; 1.6 ± 0.18). **P < 0.01 by nonparametric ANOVA (Kruskal-Wallis) and Dunn’s multiple comparison posttest. Data are means ± SEM from three independent replicates.

FMRP regulation of activity-dependent synaptic Mmp1 requires Dlp function

Dlp is up-regulated in the Drosophila model of FXS (dfmr1 null mutant), Mmp1 and Dlp genetically interact with FMRP, and independently correcting either Dlp or Mmp1 in dfmr1 mutants ameliorates FXS-associated synaptogenic defects (13, 28, 33). In conjunction with our above findings, this led us to hypothesize a coordinated FMRP-HSPG-Mmp axis converging in the synaptomatrix to enable activity-dependent synaptogenesis. To test this hypothesis, we first quantified synaptic Mmp1 in dfmr1 null mutants (89). Mmp1 was significantly increased by >50% in dfmr1 animals relative to controls (Fig. 7, A and B). Consistently, we detected a parallel >40% increase in synaptic Dlp at dfmr1 null synapses, a similar significant increase compared to controls (fig. S10, A and B). Our working theory proposes that Dlp positively regulates synaptic Mmp1, and we therefore hypothesized that increased Dlp is causative for the corresponding Mmp1 increase in the FXS model. To test this, we removed one copy of dlp in the dfmr1 null background to reduce Dlp abundance and then tested whether this restored normal synaptic Mmp1 intensity (Fig. 7, A and B). As expected, Dlp expression was significantly reduced at dlpA187/+, dfmr1/dfmr1 synapses compared to the FXS model, with Dlp signal intensity comparable to wild-type controls (fig. S10, A and B). Moreover, the 50% Mmp1 increase observed in the FXS model was fully corrected in dlpA187/+, dfmr1/dfmr1 synapses, with Mmp1 signal intensity now indistinguishable from controls (Fig. 7, A and B). Therefore, dlp coremoval in the FXS disease model is sufficient to restore normal synaptic Mmp1 abundance.

Fig. 7 FMRP regulation of the activity-dependent Mmp1 enhancement requires Dlp.

(A and B) Images of NMJs (A) and quantification of Mmp1 fluorescence intensity (B) from the denoted genotypes colabeled with HRP and Mmp1. Mmp1 intensity is shown as a heat map; white outlines mark synaptic HRP. Scale bar, 2 μm. Quantification of Mmp1 fluorescence intensity was normalized to w1118 control (n = 49; 1.0 ± 0.03); dfmr150M/50M (n = 35; 1.57 ± 0.12); and dlpA187/+, dfmr150M/50M (n = 22; 1.07 ± 0.07). (C and D) As in (A) and (B), but treated with or without high [K+]. Scale bar, 2 μm. Red line, unstimulated control. (i) w1118 control (unstimulated, n = 13; 1.0 ± 0.04) versus w1118 stimulated (n = 15; 1.51 ± 0.1); (ii) dfmr150M/50M control (unstimulated, n = 15; 1.0 ± 0.07) versus dfmr150M/50M stimulated (n = 13; 1.02 ± 0.11); and (iii) dlpA187/+, dfmr150M/50M control (unstimulated, n = 14; 1.0 ± 0.04) versus dlpA187/+, dfmr150M/50M stimulated (n = 17; 1.55 ± 0.16). For stimulated versus unstimulated pairwise comparisons, significance was determined by Mann-Whitney U tests, as indicated by **P < 0.01 and ***P < 0.001 (red asterisks). In (B) and across stimulated genotypes in (D), significance was determined by nonparametric ANOVA (Kruskal-Wallis) with Dunn’s multiple comparison posttest, as indicated by *P < 0.05, **P < 0.01, and ***P < 0.001 (black asterisks). Nonsignificant (P > 0.05) comparisons for (i) w1118 versus dlpA187/+, dfmr150M/50M (B), (ii) stimulated versus unstimulated dfmr150M/50M (D), and (iii) stimulated w1118 versus stimulated dlpA187/+, dfmr150M/50M (D) are not shown. Data are means ± SEM from at least three independent replicates.

Finally, we asked whether FMRP is required for the activity-induced, Dlp-mediated synaptic Mmp1 enhancement. To test this hypothesis, we again stimulated synapses with high [K+] and quantified synaptic Mmp1 (Fig. 7, C and D). After stimulation, wild-type synapses displayed a significant >50% increase in Mmp1 compared to unstimulated controls (Fig. 7, C and D). Conversely, there was no change in synaptic Mmp1 abundance after stimulation in dfmr1 null mutants (Fig. 7, C and D). Incorporating this into our working model predicts that the FMRP-dependent, activity-induced increase in Mmp1 is mediated through Dlp. Therefore, we hypothesized that Dlp reduction in dfmr1 null animals would restore activity-induced Mmp1 regulation. Consistently, removing a single copy of dlp in the FXS model was sufficient to fully restore the activity-induced Mmp1 increase after activity stimulation (Fig. 7, C and D). Similar to the stimulated controls (w1118), synaptic Mmp1 was significantly increased by >50% in the stimulated dlpA187/+, dfmr1/dfmr1 condition compared to unstimulated dlpA187/+, dfmr1/dfmr1 synapses (Fig. 7, C and D). Accordingly, stimulated dlpA187/+, dfmr1/dfmr1 synapses now displayed a significant increase compared to stimulated dfmr1 homozygous null mutants (Fig. 7D). Thus, dlp coremoval in the FXS disease model restores activity-induced Mmp1 regulation back toward the wild-type condition. Collectively, these findings provide the previously missing mechanistic link in the FMRP-Dlp-Mmp1 pathway, leading us to propose that Dlp misregulation mediates the disease-associated Mmp1 dysfunction causal for the FXS model’s synaptogenic defects that characterize the disease.


We report here that Mmp1, but not Mmp2, is required for rapid, activity-dependent synapse development (ghost bouton formation) in Drosophila (47, 64). Although both Drosophila Mmps act to restrict NMJ growth over developmental time (7), there is a clear differential requirement in fast, activity-dependent de novo synaptic bouton formation. Thus, distinct Mmp-dependent pathways control basal versus activity-induced synaptogenesis. Similarly separable roles have been shown for parallel molecular mechanisms. For example, the actin regulator cortactin is required for fast, activity-dependent ghost bouton formation but has no detectable role in basal synaptic bouton maturation (90). Moreover, high [K+] depolarizing activity stimulation quickly restricts microRNA functions with exclusively activity-dependent roles (91, 92). Our results suggest that Mmps (including proteolytic substrates or effectors) differentially control synaptic bouton formation in distinct time frames, revealing context-specific roles in long-term development versus acute, activity-dependent synaptogenesis. Consistent with the selective Mmp1 requirement, we also find that neuronal stimulation only increased the extracellular localization and abundance of Mmp1, whereas that of Mmp2 was reciprocally suppressed by acutely increased activity. This bidirectional coregulation by neuronal activity may reflect Mmp class interactions, given that Mmp2 restricts synaptic Mmp1 abundance, spatial distribution, and synaptogenesis requirements (7). This likely represents an important extracellular regulatory mechanism at the synapse, controlling the function of Mmp-dependent synaptomatrix outputs.

Both thermogenic dTRPA1 stimulation and the much more acute, high [K+]–induced depolarization induce a rapid increase in synaptic Mmp1 abundance and Mmp1-dependent new bouton formation, suggesting that activity-dependent Mmp1 function enables synaptogenesis. Given the relatively rapid responses to high [K+] stimulation (10 min), the synaptic Mmp1 increase is most likely a posttranscriptional mechanism occurring directly at the synapse. Multiple molecular changes induced by acute, high [K+] are cycloheximide-independent (64). We hypothesized that the Mmp1 increase occurs through extracellular regulation and identified a key interaction with the membrane-tethered HSPG Dlp. We observed that synaptic abundance of Dlp rapidly increased with acute neuronal stimulation, and tight spatial colocalization with Mmp1 in synaptic subdomains was heightened by increased activity. Our data demonstrate that Dlp is a strong, positive regulator of synaptic Mmp1, with Dlp bidirectionally determining Mmp1 abundance at the synapse. As a GPI-anchored glypican, Dlp can interact with many extracellular molecules to bring Mmp1 substrates within close proximity. In parallel, Mmp2 proteolytically cleaves Dlp to regulate Wg signaling in the Drosophila ovary (70), and a similar role likely occurs at the synapse, where Mmp2 spatially confines Dlp in the synaptic domain (7). Thus, Mmp2-dependent Dlp processing may antagonize the Dlp-Mmp1 interaction, which would necessitate reciprocal coregulation of Mmp1 and Mmp2.

We found that Dlp functions are context-dependent, with synaptic Dlp coordinating signaling mechanisms converging in the synaptomatrix. Consistently, synaptic defects observed in mutant conditions that also concurrently display misregulated Dlp are remediated by restoring Dlp expression back toward the wild-type condition (7, 28). HSPGs consist of a core protein linked to sulfated GAG chains, with both components contributing to ligand interactions (38, 46, 8387). We found that an HS-GAG chain–deficient Dlp was unable to efficiently retain Mmp1 at the synapse or enhance activity-dependent control of Mmp1 abundance. We suggest that this defect arises from loss of direct binding between Mmp1 and Dlp HS-GAG chains. Protease-HSPG interactions are bidirectional, with HS-GAG chains regulating protease localization, activity, and inhibition, whereas proteases conversely mediate HSPG proteolytic processing and turnover through cleavage of core proteins (3439, 93). We suggest that the Dlp-Mmp1 interaction is mediated through HS-GAG chains through a nonproteolytic mechanism (7). We further suggest that the synaptic colocalization of Dlp and Mmp1 is mediated through Dlp HS-GAG chains, with Dlp recruiting essential Mmp1 function driving activity-dependent bouton formation. With in situ zymography proteolytic function tests at the synapse (33, 88), we consistently found that reduced Mmp function correlated with Dlp depletion and increased function with Dlp overexpression. However, enzymatic activity appears primarily dependent on interaction with the Dlp core protein. Note that this assay reflects net proteolytic function, with likely context-dependent contributions by a variety of synaptic proteases.

Our data from the Drosophila disease model of FXS indicate that Mmp1 is constitutively increased and that activity-dependent Mmp1 enhancement is lost but that both defects are prevented when Dlp is suppressed. The mouse FXS model mirrors effects observed in Drosophila: Mouse MMP-9 is similarly up-regulated, and pharmacological and genetic inhibition of MMP-9 restores synaptic development (13, 94, 95). This suggests that the HSPG-Mmp mechanism identified here may also be conserved in mammals. The FMRP mRNA-binding translational repressor is required for activity-dependent synaptogenesis and is itself directly up-regulated by neuronal activity (5154). Moreover, HSPGs are predicted FMRP targets (32), and synaptic HSPGs are increased in the Drosophila FXS model (28). Thus, FMRP might directly regulate mmp1 and/or dlp translation in an activity-dependent mechanism. HSPG Dlp functions as a bifunctional Wg co-receptor during normal synaptogenesis, with impaired Wg trans-synaptic signaling causative in FXS disease-associated synaptogenic defects (28, 40). Mmp1, Dlp, and Wg are at least partially interdependent, and all three are disrupted in the FXS disease state, which could therefore collectively impede the activity-induced Mmp1 enhancement characterizing the FXS condition, dependent on the Dlp interaction. It will be of great interest in our future studies to dissect the differential activity requirements of this FMRP-Wg-Dlp-Mmp1 mechanism.

How do FMRP, Dlp, and Mmp1 intersect in activity-dependent synaptogenesis? Dlp is a positive regulator of Mmp1 and is required for activity-dependent increases in Mmp1 abundance and localization. Why then is the activity-induced Mmp1 increase not present in the FXS disease model despite the more abundant synaptic Dlp? It seems counterintuitive that dfmr1 mutants display increased Dlp, yet the activity-induced Mmp1 increase was completely abolished, and this defect could be corrected by reducing Dlp. One possibility is a ceiling effect in the FXS condition, in which maximal amounts of Dlp or Mmp1 no longer permit changes in response to activity. We feel that this is unlikely because Mmp1 was significantly increased by Dlp overexpression, yet an activity-induced Mmp1 increase persisted. Another possibility is that the relative abundance of Dlp and Mmp1 control their interactions. Consistently, Dlp has biphasic functions as both a positive and a negative regulator of Wg trans-synaptic signaling, depending on relative abundance of Wg ligand, Frizzled-2 receptor, and Dlp co-receptor (40, 46). Is an unknown effector(s) regulated by FMRP modulating the Dlp-Mmp1 interaction? We feel that this is most likely, because numerous activity-dependent changes can increase Dlp in the synaptomatrix. As a core signaling platform, Dlp could interact with an unknown protein(s) in Mmp1 secretion or sequestration. In any case, it is apparent that the FMRP loss impedes Dlp and Mmp1 interaction in response to neuronal activity and that reducing Dlp in the FXS model restores this core activity-dependent intersection, enabling appropriate control of Mmp1 induction at the synapse.


Drosophila genetics

All stocks were maintained on a standard medium at 25°C, except for lines used to manipulate neuronal activity with dTRPA1 (see below). The following Gal4 driver lines were obtained from the Bloomington Drosophila Stock Center (BDSC): pan-neuronal driver elav-Gal4 (#8760), selective neuronal driver CcapR-Gal4 (#39292), glutamatergic neuron driver vglut-Gal4 (#26160), and muscle-specific driver 24B-Gal4 (#1767). The dual neuron and muscle driver elav-Gal4, 24B-Gal4 was created using standard genetic recombination (96). For the dTRPA1 studies, vglut-Gal4 and CcapR-Gal4 were crossed to the heat-activated UAS-dTRPA1 (BDSC #26263) cation channel transgene to stimulate activity within motor neurons (6569). Genetic lines used to manipulate Mmps include (i) mmp1Q112* (BDSC #59380) genetic null loss-of-function allele, (ii) mmp1Q273* hypomorphic allele caused by a point-mutation early stop codon, (iii) UAS-mmp1RNAi (97), (iv) mmp2W307* genetic null loss-of-function allele, and (v) chromosomal deficiency Df(2R)BSC132 (BDSC #9410) that removes the entire mmp2 gene (21, 70). Standard recombination was used to create vglut-Gal4, mmp1Q112*; vglut-Gal4, mmp2W307*; UAS-dTRPA1, mmp1Q112*; and UAS-dTRPA1, Df(2R)BSC132 lines, which were then crossed to generate vglut-Gal4> UAS-dTRPA1, mmp1Q112*/mmp1Q112* and vglut-Gal4> UAS-dTRPA1, mmp2W307*/Df(2R)BSC132, respectively. The genetic lines used to manipulate Dlp expression include (i) dlpA187, a 26-nucleotide reading frameshift deleting the GAG attachment domain, GPI-anchor, and part of the cysteine-rich region (81); (ii) UAS-dlpRNAi [Vienna Drosophila Resource Center (VDRC) #10299]; (iii) wild-type UAS-dlpWT [BDSC #9160; (82)]; and (iv) UAS-dlp−HS, with all five GAG attachment sites (Ser625, Ser629, Ser631, Ser643, and Ser686) mutated to alanines (46, 87). The Dlp::GFP line used was y1,w*;dlpMI04217-GFSTF.1 [BDSC #60540; (77, 78)] containing an EGFP-FlAsH-StrepII-TEV-3xFlag insertion within a dlp coding intron, shared by both annotated transcripts, resulting in expression of Dlp tagged with an internal GFP tag incorporated into the endogenous dlp locus. The dfmr150M deletion allele is a full loss-of-function null mutant (89). Standard genetic recombination was used to generate dlpA187, dfmr150M/TM6, Hu-GFP (28), which was subsequently crossed to dfmr150M/TM6, Tb to produce dlpA187/+, dfmr150M/50M animals. Genetic controls included the w1118 genetic background, as well as Gal4 drivers alone crossed into the w1118 background.

Activity manipulations

A high [K+] depolarization paradigm was performed for 10 min using 90 mM KCl in 1.8 mM CaCl2 saline (64). Unstimulated controls received a 10-min mock treatment with standard physiological saline (see below). For dTRPA1 studies, the pan–motor neuron driver vglut-Gal4 (OK371, BDSC #26160) (98100) and selective neuronal driver CcapR-Gal4 (BDSC #39292) (101103) were crossed to heat-activated UAS-dTRPA1 cation channel transgene (BDSC #26263) to stimulate activity (6569). Animals were raised at the dTRPA1-restrictive temperature (18°C) until the wandering third instar stage and then transferred to pretemperature-treated apple juice agar plates for 1 hour at the dTRPA1-permissive temperature (30°C). Controls included vglut-Gal4/+ or CcapR-Gal4/+ (± temperature shift); vglut-Gal4/+, mmp1Q112*/ mmp1Q112* or vglut-Gal4/+, mmp2W307*/Df(2R)BSC132 (± temperature shift); vglut-Gal4>UAS-dTRPA1 and CcapR-Gal4>UAS-dTRPA1 (no temperature shift); and vglut-Gal4>UAS-dTRPA1, mmp1Q112*/mmp1Q112* or vglut-Gal4>UAS-dTRPA1, mmp2W307*/Df(2R)BSC132 (no temperature shift).

Immunocytochemistry confocal imaging

Imaging was performed on wandering third instars at NMJs m4 and m6/7. Staged larvae were dissected in physiological saline [128 mM NaCl, 2 mM KCl, 4 mM MgCl2, 70 mM sucrose, 5 mM Hepes, and 0 mM CaCl2 (pH 7.2)], fixed with 4% paraformaldehyde (PFA) plus 4% sucrose in phosphate-buffered saline (PBS) for 15 min, washed three times with PBS, and then incubated with primary antibodies diluted in PBS overnight at 4°C. Preparations were then washed three times with PBS, incubated with secondary antibodies diluted in PBS for 2 hours at room temperature, washed three times with PBS, and then mounted in Fluoromount-G (Electron Microscopy Sciences). For colocalization studies, a 1-hour blocking step [0.5% bovine serum albumin (BSA) in PBS] was used, with primary and secondary antibody incubations containing 0.5% BSA in PBS. The following primary antibodies were used: Alexa Fluor 488–conjugated goat α-HRP (1:200; The Jackson Laboratory, 123-545-021), Cy3-conjugated goat α-HRP (1:250; The Jackson Laboratory, 123-165-021), Cy5-conjugated goat α-HRP (1:200; The Jackson Laboratory, 123-605-021), rabbit α-GFP (Abcam, ab290), rabbit α-Mmp2 (1:500), mouse α-Mmp1 [1:1:1 at 1:10; Developmental Studies Hybridoma Bank (DSHB), 3B8D12, 3A6B4, and 5H7B11], mouse α-DLG (1:200; DSHB, DLG1), and mouse α-Dlp (1:5; DSHB, 13G8). The following secondary antibodies were used (all 1:500; Invitrogen): goat α-mouse (Alexa Fluor 488 and Alexa Fluor 568), Alexa Fluor 488–conjugated donkey α-mouse, Alexa Fluor 488–conjugated donkey α-rabbit, and goat α-rabbit (Alexa Fluor 488 and Alexa Fluor 568). All labeling was performed detergent-free [extracellular labeling only (73, 74)], except for structural studies using mouse α-DLG, which required addition of 0.1% Triton X-100 to both primary and secondary antibody incubations. NMJ Z-stacks were acquired with a Zeiss LSM 510 META laser scanning confocal microscope using 40×/1.4 (ghost boutons) and 63×/1.4 (all other experiments) Plan Apochromat oil immersion objectives.

Synaptic in situ zymography

NMJ in situ zymography assays were performed as previously described (33, 88). Briefly, live larval preparations were dissected, exposing the NMJ in physiological saline, and then immediately submerged in the zymography buffer [50 mM tris-HCl, 150 mM NaCl, 5 mM CaCl2, and 0.2 mM NaN3 (pH 7.6)] containing fluorescein-conjugate DQ porcine gelatin [500 μg/ml; Molecular Probes (Life Technologies), D-12054]. Preparations were incubated for 45 min at room temperature, kept stationary, and protected from light. Preparations were then briefly rinsed three times in PBS, fixed with 4% PFA in PBS for 30 min, and washed again three times in PBS. Preparations were then incubated with Cy3-conjugated goat α-HRP (1:250; The Jackson Laboratory, 123-165-021) for 1 hour at room temperature. Preparations were finally washed again three times in PBS and mounted in Fluoromount-G (Electron Microscopy Sciences). NMJ Z-stacks were acquired with a Zeiss LSM 510 META laser scanning confocal microscope using a 488-nm argon laser to excite the fluorescein substrate and imaged with a 63×/1.4 Plan Apo oil immersion objective.

Quantification and statistical analyses

All experiments were independently performed at least three times, with all comparisons performed blind. All analyses were done on staged and sized-matched animals. For each trial, three to five animals per condition were assayed, with NMJs from both left and right segments (A3 to A4) analyzed. For all intensity comparisons, images were obtained with the same confocal settings and quantified in parallel using NIH (National Institutes of Health) ImageJ software (104). Intensity measurements were made with the HRP signal delineated from Z-stack areas of maximum projections. Muscle intensity was taken from areas adjacent to the HRP-marked NMJ and subsequently subtracted from the NMJ intensity measurement. NMJ and background region selection excluded regions that overlapped with obstructions, such as the trachea. For high [K+] stimulation studies involving multiple genotypes, each stimulated condition was normalized to its own unstimulated genotype control. Confocal settings were maintained constant for unstimulated and stimulated conditions within each genotype. All cross-compared genotypes within each experiment were processed together and imaged at the same time. Ghost boutons were quantified as HRP-positive and DLG-negative varicosities emanating from the main NMJ arbor. DLG-negative was defined as <2 SDs below the mean. Contrast and brightness were applied uniformly using ImageJ. Colocalization analyses were performed using ZEN image processing software (Zeiss). Single slices (<1 μm) from the middle of a Z-stack were analyzed for each NMJ. Measurements of full width at half maximum (FWHM) for axial resolution were 840 nm (FWHM488) and 880 nm (FWHM555). HRP was used as a guide to create a region of interest. Thresholds were manually set on the basis of background intensity values. Manders’ overlap coefficients and weighted MCCs were used (80). All images were filtered and processed in parallel in ImageJ before being exported to Adobe Photoshop. ImageJ versions used were 1.46r, 1.51h, and 1.51p (104, 105). All reported statistical comparisons were always performed using Instat3 software (GraphPad Software). Mann-Whitney U tests and ANOVAs (Kruskal-Wallis) were used for all the nonparametric comparisons. ANOVA tests were used for all data sets of three or more comparisons, followed by appropriate post hoc analyses as stated in the figure legends. All data are presented in figures as means ± SEM, and n represents the total number of NMJs analyzed from at least three independent replicates. Significance is indicated in the figures as follows: *P < 0.05, **P < 0.01, and ***P < 0.001.


Fig. S1. Temperature controls for dTRPA1 activity-induced synaptic bouton formation.

Fig. S2. Mmp2 is not required for activity-dependent synaptic bouton formation.

Fig. S3. Mmp1 is rapidly and specifically increased after dTRPA1 neuronal stimulation.

Fig. S4. Synaptic Mmp2 is rapidly reduced after acute neuronal stimulation.

Fig. S5. Synaptic Dlp is rapidly increased after acute neuronal stimulation.

Fig. S6. Synaptic Mmp1 and Dlp imaging controls at the NMJ terminal.

Fig. S7. Synaptic Dlp changes with bidirectional dlp genetic manipulations.

Fig. S8. Synaptic Mmp2 changes with bidirectional dlp genetic manipulations.

Fig. S9. Activity-dependent synaptic Dlp increase occurs in the absence of Mmp1.

Fig. S10. Synaptic Dlp in FXS disease model is restored by single-copy dlp coremoval.


Acknowledgments: We thank the BDSC (Indiana University) and VDRC ( for essential stocks used in this study. We thank the DSHB (University of Iowa) for providing critical antibodies used in this study. We are particularly grateful to X. Lin (Fudan University and Cincinnati Children’s Hospital Medical Center) for the UAS-dlp−HS mutant transgene and D. Bohmann (University of Rochester Medical Center) for the UAS-mmp1RNAi line. We are also grateful to C. Doll and T. Kennedy for manuscript input and the entire Broadie Laboratory for helpful insights. Funding: This work was supported by NIH grants MH096832 and MH084989 (to K.B.). Author contributions: M.L.D. performed all the experiments except for the zymography assays, which were performed by J.S. K.B. and M.L.D. jointly conceived the experiments and co-wrote the manuscript, and K.B. oversaw all the studies. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article