Endogenous Shv tagged with eGFP is detected in glial cells.

(A) Schematic of eGFP knock-in to the Shv gene generated using CRISPR/Cas9 system (top). Exons are in orange, signal peptide in red. Western blots using antibodies against GFP and Shv confirms the presence of eGFP in Shv. β-tubulin is used as a loading control. (B) Low magnification images of the 3rd instar larval brain showing weak Shv-eGFP signal throughout the brain (left). Scale bar = 100 μM. Zoomed in view of the brain hemisphere and ventral nerve cord with neurons and glia marked by Elav and Repo antibodies, respectively. Asterisks label glial cells with Shv expression. Scale bar for brain and VNC are 10 and 15 μM, respectively. (C) Shv-eGFP can be detected at the NMJ. Glial membrane is marked by membrane targeted tdTomato (driven by glial specific Repo-GAL4) and neuronal membrane labeled by HRP. Zoomed in views show that Shv-eGFP colocalizes with glial membrane (magenta arrow) and synaptic boutons (yellow arrow). Shv-eGFP also weakly labels the muscle. Scale bar = 10 μM in the upper panels, and 2 μM in the lower panels.

Shv in glia is necessary for activity-induced synaptic remodeling.

(A) Tissue-specific knockdown of shv. Reducing Shv in neurons or glia blocks activity-induced synaptic remodeling, but not when shv is knocked down in muscles. Notably, glia-specific knockdown of shv increases the levels of GluRIIC, whereas neuronal Shv knockdown decreases basal GluRIIC. (B) Acute knockdown of shv in glia using the inducible repo-GeneSwitch-GAL4 driver affects basal GluR intensity and abolishes activity-induced synaptic changes. (A) and (B), Scale bar = 2 μm. All values are normalized to unstimulated control and presented as mean ± S.E.M. Control contains TRiP RNAi control vector driven by the indicated driver. Statistics: One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p ≤ 0.05; ## p ≤ 0.01; ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 when comparing stimulated to unstimulated NMJs.

Expression of shv in perineurial and subperineurial glia is required for activity-induced synaptic remodeling.

(A) Diagram of relative membrane position and extension of WG, SPG, and PG at the NMJ. Astrocyte-like glia can be detected in the central brain and ventral nerve chord. (B) Representative images and quantification of synaptic changes following knockdown of shv in glia subtypes. Knockdown of shv in SPG and PG recapitulate the phenotypes of pan-neuronal knockdown, as well as abolish activity-induced synaptic remodeling. Scale bar = 2 μm. All values are normalized to unstimulated control and presented as mean ± S.E.M. Control contains TRiP RNAi control vector driven by the indicated driver. Statistics: One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p ≤ 0.05; ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 when comparing stimulated to unstimulated NMJs.

Glial Shv rescues defective synaptic plasticity observed in shv1 mutant.

(A) Selective expression of shv in glia of shv1 mutants is sufficient to rescue activity-dependent synaptic changes but not basal bouton size. Scale bar = 2μm. (B) Representative mEPSP and eEPSP recordings conducted using HL3 solution containing 0.5 mM Ca2+. Average eEPSP amplitude is shown after nonlinear summation correction. (C) Normalized eEPSP before and following tetanus (10 Hz for 2 min) at the indicated time points. Recordings were done using HL3 solution containing 0.25 mM Ca2+. shv1 showed significantly diminished PTP following stimulation. Expression of shv in glial cells of shv1 rescued PTP. The number of NMJs examined is shown in parentheses. Student’s t-test was used to compare between control and the indicated genotypes. * shows that shv1 displays PTP that is significantly lower than the control (p ≤ 0.05), starting from the indicated time and onwards. All values are mean ± S.E.M. For (A) and (B), one-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p ≤ 0.05; ## p ≤ 0.01; ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; **p ≤ 0.01 when comparing stimulated to unstimulated NMJs.

Glial Shv does not activate integrin signaling but modulates neuronal release of Shv.

(A) Representative images showing that Shv can be detected extracellularly when expressed using glial-specific driver. Extracellular Shv-HA (Shvextra)is monitored using an antibody against HA under detergent-free staining condition. (B) Expression of Shv using the glial-specific driver does not rescue pFAK levels during unstimulated and stimulated conditions, revealing that glial Shv does not activate integrin. (C) Knockdown of shv in glia upregulated pFAK, whereas upregulation of shv in glia blocked the activity-dependent increase normally seen in control. (D) Expression of shv in shv1mutants show activity-dependent release of Shv by neurons, but not by glia. Shvextra indicates extracellular Shv stained using detergent-free conditions. (E) Knockdown of endogenous Shv-eGFP in neurons or glia did not diminish extracellular presence of Shv-eGFP at the NMJ, suggesting homeostatic regulation of Shv protein level. Scale bar = 2 μm for all panels. All values are normalized to unstimulated control and presented as mean ± S.E.M. One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p ≤ 0.05; ** p ≤ 0.01; ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 when comparing stimulated to unstimulated NMJs.

Glial Shv regulates ambient extracellular glutamate concentration.

(A) Shv expression in glia maintains ambient glutamate concentration at the NMJ. Left panel shows a schematic of the tripartite synapse at the NMJ. iGluSnFR expression in neurons via the GAL4/UAS system can detect extracellular ambient glutamate concentration at the synapse, while glia-specific knockdown or upregulation of Shv is achieved using the LexA/LexAop system. iGluSnFR signals are seen in green and HRP-Cy3 was added to mark neuronal membrane (middle panels). Lower graph shows quantitation of the relative ambient glutamate concentration at the NMJ. Only Shv from glia affects ambient glutamate concentration, Shv from neurons does not. (B) Incubating the NMJ with 2 mM glutamate rescues synaptic remodeling in case of glial Shv knockdown. Vehicle controls represent NMJs dissected in parallel and incubated with HL3 without 2mM glutamate for the same length of time. Scale bar = 2μm. All values are normalized to unstimulated control and presented as mean ± S.E.M. Control contains TRiP RNAi control vector driven by the indicated driver. Statistics: One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p ≤ 0.05; ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001 when comparing stimulated to unstimulated NMJs.

Acute ablation of glia elevated basal GluR levels and disrupted activity-induced synaptic remodeling, but extracellular glutamate incubation is sufficient to compensate for the loss of glia.

(A) Transient rpr expression is sufficient to trigger death of glial cells. Diagram shows the RU486 feeding protocol used to induce glial cell death in 3rd instar larvae. Lower panels show images of the segmental nerves (left) and the NMJ (middle). The zoomed in region of the NMJ (yellow box) is magnified on the right. Glial membrane is marked by CD4-tdGFP, which appears fragmented in the presence of rpr expression, indicating glial cell death. Scale bars indicate size in μm. (B) Images and (C) quantitation of synaptic bouton size and GluR abundance. Incubating the NMJ for 10 minutes with 2 mM glutamate is sufficient to overcome loss of glial cells and restore basal GluR level and activity-induced synaptic remodeling, suggesting a main function of peripheral glia in synaptic plasticity regulation is to maintain a high ambient glutamate concentration. Scale bar = 2μm in (B). All values are normalized to unstimulated control and presented as mean ± S.E.M. Control contains TRiP RNAi control vector driven by the indicated driver. Statistics: One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. ### p ≤ 0.001 when comparing unstimulated samples to unstimulated control. *p ≤ 0.05; ***p ≤ 0.001 when comparing stimulated to unstimulated NMJs.

Efficiency of shv knockdown by RNAi in neurons and glia.

Staining of the larval brain confirms that the RNAi approach effectively reduces Shv-eGFP in the selective cell types. Yellow arrows highlight neurons (Elav positive), magenta arrows point to glial cells (Repo positive). Scale bar = 10 μm.

Integrity of peripheral glial cell membranes.

Representative images showing that knockdown of shv in glia does not alter the gross morphology of peripheral glia at the NMJ. Lower panels show magnified view of glia closely associating with proximal synaptic boutons. Scale bar = 10 μm.

Defective activity-induced synaptic remodeling caused by the loss of neuronal Shv is not rescued by incubation with 2 mM glutamate.

Vehicle controls represent NMJs dissected in parallel and incubated with HL3 without 2 mM glutamate for the same length of time. Scale bar = 2 μm. All values are normalized to unstimulated control and presented as mean ± S.E.M. Control contains TRiP RNAi control vector driven by the neuronal driver. Statistics: One-way Anova followed by Tukey’s multiple comparison test was used to compare between unstimulated control and unstimulated NMJs across genotypes. Student’s t-test was used to compare between unstimulated and stimulated NMJs of the same genotype. # p σ; 0.05 when comparing unstimulated samples to unstimulated control. **p σ; 0.01 when comparing stimulated to unstimulated NMJs.