Glial Synaptobrevin mediates peripheral nerve insulation, neural metabolic supply, and is required for motor function

Peripheral nerves contain sensory and motor neuron axons coated by glial cells whose interplay ensures function, but molecular details are lacking. SNARE‐proteins mediate the exchange and secretion of cargo by fusing vesicles with target organelles, but how glial SNAREs contribute to peripheral nerve function is largely unknown. We, here, identify non‐neuronal Synaptobrevin (Syb) as the essential vesicular SNARE in Drosophila peripheral glia to insulate and metabolically supply neurons. We show that tetanus neurotoxin light chain (TeNT‐LC), which potently inhibits SNARE‐mediated exocytosis from neurons, also impairs peripheral nerve function when selectively expressed in glia, causing nerve disintegration, defective axonal transport, tetanic muscle hyperactivity, impaired locomotion, and lethality. While TeNT‐LC disrupts neural function by cleaving neuronal Synaptobrevin (nSyb), it targets non‐neuronal Synaptobrevin (Syb) in glia, which it cleaves at low rates: Glial knockdown of Syb (but not nSyb) phenocopied glial TeNT‐LC expression whose effects were reverted by a TeNT‐LC‐insensitive Syb mutant. We link Syb‐necessity to two distinct glial subtypes: Impairing Syb function in subperineurial glia disrupted nerve morphology, axonal transport, and locomotion, likely, because nerve‐isolating septate junctions (SJs) could not form as essential SJ components (like the cell adhesion protein Neurexin‐IV) were mistargeted. Interference with Syb in axon‐encircling wrapping glia left nerve morphology and locomotion intact but impaired axonal transport, likely because neural metabolic supply was disrupted due to the mistargeting of metabolite shuffling monocarboxylate transporters. Our study identifies crucial roles of Syb in various glial subtypes to ensure glial‐glial and glial‐neural interplay needed for proper nerve function, animal motility, and survival.


| INTRODUCTION
Motor control is mediated by motoneurons of the central nervous system that convey electrical information in the form of action potentials (APs) along their axons over large distances to peripheral body muscles. At the target muscle, motoneuronal APs induce the release of neurotransmitters (NT) at neuromuscular junctions (NMJs), leading to muscle excitation and contraction (Kuo & Ehrlich, 2015). Peripheral nerves contain the axons of motor-and sensory neurons together with glial cells that separate axons from the surrounding environment and additionally serve essential functions in development, regeneration, neural metabolism, ion homeostasis, AP propagation and the control of excitability (Acton & Miles, 2015;Araque & Perea, 2004;Fields, 2015; H. U. Lee et al., 2013;Pascual et al., 2005;Simard & Nedergaard, 2004;J. M. Zhang et al., 2003;Zuchero & Barres, 2015).
WG resemble the non-myelinating vertebrate Schwann cells that engulf axon bundles to form Remak fibers, and are thought to function in metabolic/ion homeostasis and support AP propagation Leiserson et al., 2000;Li, Russo, & DiAntonio, 2019).
We here used the accessibility of Drosophila peripheral nerves and their sheathing with distinct glial cell layers to address the question how glial v-SNAREs support peripheral nerve function and locomotion. We report that TeNT-LC expression in peripheral glial cells severely disrupts nerve morphology and the axonal transport of synaptic material. It furthermore caused motoneuronal hyperactivity and paralysis, typical effects of TeNT intoxication in higher organisms. Unlike neurons, where TeNT-LC cleaves nSyb and arrests neurotransmission, we show that in glial cells these effects were caused by the interference with non-neuronal Syb which is cleaved at low efficacy. We furthermore link the functional requirement for Syb to distinct glial subtypes: TeNT-LC/Syb-RNAi expression in axon encircling wrapping glia (WG) left locomotion intact, but impaired the axonal transport of synaptic cargo similar to knocking down Basigin, an essential component for neural metabolic supply. Moreover, WG TeNT-LC/Syb-RNAi expression blocked the integration of F I G U R E 1 Pan-glial TeNT-LC expression impairs locomotion, disrupts nerve integrity, and results in motoneural hyperactivity. (a,b) Sketches of dissected larvae depicting segmental nerves radiating from the brain (a) containing glia ensheathed efferent motoneuron axons and afferent sensory neuron axons (b, left; right: nerve cross section according to Stork et al. (2008)). (c,d) Analysis of larval locomotion (crawling) with example tracks of individual animals over 85 s (c) and quantification across all animals for distances travelled over time (85 s; d, left) and comparison of final distances travelled by third instar larvae of the indicated genotypes (d, right): Repo-GAL4 (black), UAS-TeNT-LC (grey) and Repo>TeNT-LC (red). See also Videos 1-3. (e,f) Examples images of nerves of segments A2-A4 labelled with antibodies specific for the neuronal membrane (HRP; axonal morphology, e, left) and BRP (e, right) and quantification of axonal area, nerve morphology and BRP spots per nerve (f) in third instar larvae of Repo-GAL4 (black), UAS-TeNT-LC (grey) and Repo>TeNT-LC (red) animals. Dashed lines in (e, BRP) indicate nerve area and arrow head (e, axonal morphology) indicates defasciculated single axon. (g,h) Representative mEPSP (spont.) and eEPSP (evoked) traces (g) and quantification of mEPSP/eEPSP amplitudes, % of recorded cells displaying normal activity (single eEPSP in response to single stimulation) and the number of response peaks per stimuli in Repo-GAL4 (black), UAS-TeNT-LC (grey) and Repo>TeNT-LC (red) animals. Asterisks indicate postsynaptic response peaks (eEPSP). Scale bars: (c) 1 cm; (e) 10 μm; (g) mEPSP: 1 s, 2 mV. eEPSP: 25 ms, 5 mV. Statistics: parametric one-way analysis of variance (ANOVA) test, followed by Tukey's multiple comparison test except for (f, nerve morphology) where a non-parametric Kruskal-Wallis test was performed. ***p ≤ .001; **p ≤ .01; *p ≤ .05; n.s. (not significant) p > .05. Number of animals: (c,d) Repo-GAL4: six larvae; UAS-TeNT-LC: nine larvae; Repo>TeNT-LC: 11 larvae; (e,f) Repo-GAL4: 17 nerves, six larvae; UAS-TeNT-LC: 18 nerves, six larvae; Repo>TeNT-LC: 18 nerves, six larvae; (g,h) Repo-GAL4: eight cells, four larvae; UAS-TeNT-LC: seven cells, four larvae; Repo>TeNT-LC: eight cells, four larvae. All panels show mean ± SEM F I G U R E 2 Syb but not nSyb is targeted by TeNT-LC in glial cells. (a) Sketch of a vesicle and target membrane schematically depicting the SNARE proteins and the v-SNARE attack point of TeNT-LC. (b) Sequence alignment of Drosophila nSyb (magenta) and Syb (blue) isoforms A SNARE motifs. The TeNT-LC cleavage site (QF, red box) is highlighted. Alignment performed using Clustal Omega. Gonnet PAM250 matrix used to compare sequence substitutions: * identical aa,: score > 0.5,. score < 0.5, "gap" score below 0. (c) SDS-PAGE analysis of in vitro TeNT-LC cleavage assay. T4-lysozyme fusion proteins of Drosophila nSyb (magenta, left) and Syb (blue, right) were expressed, purified and cleavage was tested by adding TeNT-LC for the indicated durations (see methods for details). Asterisks indicate main cleavage product. (d-g) Example images of nerves of segments A2-A4 labelled with antibodies specific for the neuronal membrane (HRP; axonal morphology, d,f, left) and BRP (d,f, right) and quantification of axonal area, nerve morphology and BRP spots per nerve (e,g) in third instar larvae of the indicated genotypes. Dashed lines in d,f indicate nerve area. (h,i) Representative mEPSP (spont.) and eEPSP (evoked) traces (h) and quantification of mEPSP/eEPSP amplitudes, % of recorded cells displaying normal activity (single eEPSP in response to single stimulation) and the normalized number of response peaks per stimulus (i) in Repo-GAL4 (black), UAS-Syb-RNAi (grey) and Repo>Syb-RNAi (red) animals. Asterisks indicate postsynaptic response peaks (eEPSP). Scale bars: (d,f) 10 μm; (h) mEPSP: 1 s, 2 mV. eEPSP: 25 ms, 5 mV. Statistics: parametric one-way analysis of variance (ANOVA) test, followed by Tukey's multiple comparison test except for nerve morphology (e,g) where a non-parametric Kruskal-Wallis test was performed. ***p ≤ .001; *p ≤ .05; n.s. (not significant) p > .05. Number of animals: (d,e) Repo-GAL4: 15 nerves, six larvae; UAS-nSyb-RNAi: 15 nerves, six larvae; Repo>nSyb-RNAi: 15 nerves, six larvae; (f,g) Repo-GAL4: 19 nerves, six larvae; UAS-Syb-RNAi: 18 nerves, six larvae; Repo>Syb-RNAi: 19 nerves, six larvae; (h,i) Repo-GAL4: 9 cells, five larvae; UAS-Syb-RNAi: 10 cells, five larvae; Repo>Syb-RNAi: 10 cells, five larvae. All panels show mean ± SEM. See also Figure S1 monocarboxylate transporters (MCTs) that energetically supply neurons at the glia/neuronal interface, consistent with their Syb mediated insertion. In contrast, interference with Syb functionality (by TeNT-LC/Syb-RNAi expression) in subperineurial glia (SPG) interfered with the formation of nerve-insulating SJs which disrupted nerve morphology, impaired axonal transport and caused animal paralysis.

| Expression of TeNT-LC in glial cells causes paralysis and disrupts nerve morphology and function
We wondered whether TeNT-LC targeted to glial cells of peripheral Drosophila nerves affected motor function. We expressed TeNT-LC (Sweeney et al., 1995) in all glial subtypes using the pan-glial driver Repo-GAL4 (Sepp, Schulte, & Auld, 2001) and investigated its influ- Electrical stimulation of the innervating nerve in control animals evokes a single AP that reliably triggered a single evoked excitatory postsynaptic potential in the muscle (eEPSP; seen in 8/8 Repo-GAL4 driver control and 7/7 UAS-TeNT-LC construct control cells; Figure 1g,h, evoked). In contrast, upon pan-glial TeNT-LC expression, a single stimulus triggered additional eEPSPs and thus motoneuronal hyperactivity (seen in 10/10 cells; Figure 1g,h), a typical hallmark of TeNT intoxication in higher organisms. While the number of eEPSPs elicited per stimulus was increased, the average amplitudes of the first eEPSPs did not differ from control cells (Figure 1g,h), indicating that glial TeNT-LC expression did not disrupt synaptic transmission per se. Thus, we report that TeNT-LCmediated interference with SNARE-dependent processes in glial cells leads to paralysis, disrupts nerve integrity, impairs axonal transport, and causes motoneural hyperexcitability.

| Essential glial functions are mediated by Syb but not nSyb
We next asked which v-SNARE might be targeted by TeNT-LC in glial cells ( Figure 2a). Tetanus pathology is attributed to Synaptobrevin-2/ VAMP2 cleavage in mammalian interneurons (Schiavo et al., 2000).
We thus performed an in vitro cleavage assay to test whether TeNT-LC cleaves both v-SNAREs. To allow for recombinant production of the proteins, and to facilitate the visualization of cleavage products by SDS-PAGE analysis, the membrane anchors of Drosophila Syb and nSyb were replaced by a lysozyme sequence from T4 phage (T4L).
Syb-T4L and nSyb-T4L fusion proteins were expressed in Escherichia coli, purified and cleavage was tested by adding TeNT-LC for various durations (see methods for details). After 6 hr of incubation, nSyb was almost completely cleaved by TeNT-LC, while only a small proportion of Syb was cut ( Figure 2c, major cleavage products are indicated by red asterisks), in line with previous results (Sweeney et al., 1995).
However, after prolonged incubation also Syb cleavage became apparent (>6 hr; Figure 2c). Thus, TeNT-LC cleaves both v-SNAREs in vitro, albeit with largely different efficiency.
To assess whether the observed phenotypes following glial TeNT-LC expression depended on the functional loss of nSyb or Syb, we tested whether pan glial (using Repo-GAL4) knockdown of either protein using specific RNAi-lines phenocopied TeNT-LC expression

| Syb disruption in subperineurial glia interferes with septate junction formation which causes paralysis, nerve disruption, and impaired axonal transport
We next speculated how the impairment of Syb functionality caused these phenotypes and whether all glial subtypes are affected (see Figure 1b, right, for subtypes). SPG enwrap axons and WG (Stork et al., 2008 and Figures 1b and 4a). By using rL82-GAL4/Gli-GAL4 (Auld, Fetter, Broadie, & Goodman, 1995;Sepp & Auld, 1999) mediated expression of membrane associated GFP (UAS-mCD8::GFP; Figure 4b,c), we also discovered that SPG extend to fully cover the NMJ in 35% of cases, in line with previous results (Auld et al., 1995;Brink, Gilbert, & Auld, 2009;Fuentes-Medel et al., 2009;Kerr et al., 2014;Sepp & Auld, 1999). In the course of this study, we also found two rare adult escapees of TeNT-LC expressing SPG which were also severely paralyzed, showed almost no movement and were even unable to reorientate to an upright stance when falling on their back (Video 15).
These data clearly implicate a necessity of Syb in SPG to support nerve integrity and axonal transport to finally allow proper locomotion.
Additionally, electrophysiological recordings of larvae expressing TeNT-LC or Syb-RNAi in SPG revealed occasional aberrant motoneuronal responses after AP-evoked stimulation ( Figure S2e-h) but these effects were much weaker than upon pan-glial expression. This might be due to the additional contribution of other glial cell types (e.g., wrapping glia, (Li et al., 2019)) or due to a weaker expression strength of the rL82-GAL4 driver compared to Repo-GAL4 (Yildirim, Petri, Kottmeier, & Klambt, 2019).
Finally, we can exclude that TeNT-LC or Syb-RNAi expression kills SPG cells as their visualization by simultaneous CD8::GFP expression showed an intact GFP coverage of investigated nerves similar to control animals ( Figure S2i-l).
SJ formation depends on the delivery of key components (including NrxIV) to the glial surface likely by exocytosis (Babatz et al., 2018;Tiklova, Senti, Wang, Graslund, & Samakovlis, 2010) and we hypothesized that this might be driven by Syb. We therefore tested whether TeNT-LC/Syb-RNAi expression additionally blocks other relevant exocytic processes that for example, establish a compensatory evolutionary ancient barrier (noted in nrxIV mutants [Babatz et al., 2018]) and/or release gliotransmitters like ATP or NPY that suppress neural excitation (Acton & Miles, 2015;Carlsen & Perrier, 2014;Coco et al., 2003;Pascual et al., 2005;Redman & Silinsky, 1994;Schwarz et al., 2017;J. M. Zhang et al., 2003). Mammalian astrocytes control neural excitability by secreting ATP and NPY via VAMP7 and VAMP3, the latter of which is a TeNT-LC target (Galli et al., 1998;Schwarz et al., 2017;Verderio et al., 2012). Peptidergic signaling from Drosophila SPG likely also controls neural excitability, because SPG knockdown of the signaling protein Wingless (Wnt) affected synaptic transmission at the NMJ (Kerr et al., 2014). We therefore probed whether Syb-knockdown in SPG caused an accumulation of peptidergic vesicles using the marker atrial natriuretic peptide (ANF (Rao, Lang, Levitan, & Deitcher, 2001). Indeed, we observed an 11-fold accumulation of such vesicles upon Syb-RNAi expression in SPG (Figure S3h,i), suggesting that peptidergic release from SPG is also mediated by Syb and might contribute to nerve function.
But why did BRP also accumulate when SPG function was perturbed (Figures 4f,g,j,k; Figure S2c,d)? We hypothesized that this may be caused by an indirect interference with WG function as the morphological nerve disruption upon SPG manipulation might lead to a loss of the usual WG-axon contact (which would then in turn interfere with metabolic supply). To test this, we expressed TeNT-LC in SPG together with the genomically GFP-tagged MCT1 construct (predominantly expressed in WG, Figure S4e,f) and compared the overlap of MCT1::sfGFP signal with that of the axonal area (measured via HRPstaining). While SPG TeNT-LC expression did not change the number of MCT1::sfGFP spots seen per nerve, the strongly increased nerve area caused a largely decreased MCT1::sfGFP density (MCT1::sfGFP/ μm 2 nerve; Figure S5f,g). Thus, while MCT1::sfGFP is present in similar amounts upon SPG TeNT-LC (and possibly upon Syb-RNAi/Nrx-F I G U R E 5 Legend on next page. IV-RNAi) expression its mal-distribution due to morphological nerve disruption may hamper axonal energy supply thereby causing axonal BRP accumulation.

| DISCUSSION
Glial cells are essential to nervous system function in a multitude of ways, including neural metabolic supply, NT uptake, nerve development and -regeneration, compartmental nerve insulation for AP propagation and ion homeostasis (Edgar et al., 2009;Edgar et al., 2004;Fields, 2015;Funfschilling et al., 2012;Griffiths et al., 1998;Kottmeier et al., 2020;Y. Lee et al., 2012;Leiserson et al., 2000;Li et al., 2019;Machler et al., 2016;Pellerin & Magistretti, 1994;Simard & Nedergaard, 2004;Volkenhoff et al., 2015;Zuchero & Barres, 2015).  et al., 1995)). In contrast, cleavage of Syb required much longer incubations. The reason for this remarkable difference is not clear, especially since the cleavage site (and most of the SNARE motif) is conserved between both proteins (Figure 2b/ Figure S1). This could indicate that the long-term expression of TeNT-LC we are using in our study might suffice to cleave a large fraction of available Syb.
Accordingly, direct SJ disruption in SPG (via NrxIV knockdown) caused nerve disintegration, impaired locomotion but no motoneural hyperactivity. Notably, a recent study also reported on hyperactivity and nerve swellings upon glial knock-down of the salt-inducible kinase 3 (SIK3), a pathway controlling glial K + and water homeostasis (Li et al., 2019) and effects were partially restored upon re-expression in WG or SPG. Thus the observed phenotypes in our study may be related to imbalance ion and water homeostasis which is controlled by more than one glial type. Another interesting possibility is that one of the other glia subtypes (e.g., WG and/or perineurial glia) releases a cargo that limits motoneural excitability. Interesting candidates are purinergic or peptidergic vesicles (which we found to accumulate) and whose release from glia of the mammalian central nervous system reduces neural excitability (Carlsen & Perrier, 2014;Pascual et al., 2005;Schwarz et al., 2017;J. M. Zhang et al., 2003). Further research will be necessary to investigate whether similar functions are conserved in the peripheral nervous system and across species.
Our study provides new insights into the multitude of SNAREdependent processes in peripheral glial cells to ensure glial and motoneuronal function and locomotion. Future investigation of molecular underpinnings (including the identification of the relevant target SNAREs, co-factors, cues and triggers for the localized fusion of cargo-vesicles) will provide further mechanistic insight and may clarify whether disruptions of these processes contribute to diseases of peripheral motor control, including tetanus.
To generate cDNA encoding UAS-Syb-QFVW, the sequence of Syb Isoform A was amplified from cDNA clone SD05285 (obtained from DGRC). Point mutations were generated using the following primers: Syn-RA-5 0 -QFVW:

| Image acquisition, processing, and analysis
Confocal microscopy was performed with a Leica SP8 microscope (Leica Microsystems, Germany). Images were acquired at room temperature. Confocal imaging was done using a 63 × 1.4 NA oil immersion objective with a zoom of 1.8 and z-step size of 0.25 μm. All confocal images were acquired using the LAS X software (Leica Microsystems, Germany). Images from fixed samples were taken from nerve bundles exiting the ventral nerve cord of segments A2-A4 (see Figure 1a for depicted ROI) except for Figure 5h,i, S4e-j and S5f,g where nerves crossing muscle four where imaged. Confocal stacks were processed with the ImageJ software (http://rsbweb.nih.gov/ij/).
Quantifications of axonal spot numbers and areas were performed following an adjusted manual (Andlauer & Sigrist, 2012), briefly as follows. The signal of a HRP-647 antibody was used as template for a mask, restricting the quantified area to the shape of the nerve. The original confocal stacks were converted to maximal projections and a mask of the axonal spots was created by applying a threshold to remove irrelevant lower intensity pixels. The threshold was adjusted manually and individually to every image to detect all axonal spots.
The segmentation of single spots was done semi-automatically via the command "Find Maxima" embedded in the ImageJ software and by hand with the pencil tool and a line thickness of 1 pixel. To remove high frequency noise a Gaussian blur filter (0.5 pixel Sigma radius) was applied. The processed picture was then transformed into a binary mask using the same lower threshold value as in the first step. This binary mask was then projected onto the original unmodified image using the "min" operation from the ImageJ image calculator. Spot numbers and areas were determined using the "Analyze particles" function (particle size >2 pixels) embedded into ImageJ. For line profiles, a line (170 pixels) was drawn at the same position in the HRP (axonal morphology) and GFP channel of an image using the straight line function in ImageJ. Line profiles were then plotted using the "plot profile" function of ImageJ. To measure the axonal (HRP) or lamellar (GluRIID) area, the signal of the HRP-647 or the GluRIID antibody was used as template for a binary mask. The area was then quantified using the "wand tool" to select the area and "measure" function embedded into ImageJ to quantify the area. To measure the intensity per nerve, the signal of the HRP-647 antibody was used to define the region of interest, and the mean intensity within this region of interest was then measured in the channel of interest using the "measure" were processed using MATLAB R2016a as follows: first, the whole stack containing the three channels (BRP, GluRIID, HRP) was loaded using the command "imread" in a loop iterating through the first to last image of the stack. Then, the three channels were separated into three stacks for the following permutation operation. Using the command "permute," the second and third dimension were switched separately in each stack. Lastly, each of the three permuted stacks was written to a file using the command "imwrite" in a loop iterating through all images of the stack, resulting in three .tif-files (each for one channel) showing the orthogonal view of the imaged nerve. The respective code is available upon request. Orthogonal views of each channel were then merged using ImageJ.
Images for figures were processed with ImageJ software to enhance brightness using the brightness/contrast function. For axonal morphology images, the HRP-647 signal was converted into a pseudo-color image using the command "Lockup Table>fire" in ImageJ.
Starting from this posterior incision, a cut was made along the length of the larva extending beyond the head pin. The cuticle was pinned down twice on either side. The intestines and trachea were cut at the posterior and held firmly with a forceps as the remaining connections to the body were cut before being fully removed, taking care not to pull on the preparation. The brain was held slightly raised above the preparation and the segmental nerves cut without touching the underlying muscle, before finally removing the brain. The Sylgard block and completed larval preparation was placed in the recording chamber which was filled with 2 ml HL3 (plus 0.4 mM CaCl 2 , 10 mM MgCl 2 ). Recordings were performed at room temperature in current clamp mode at muscle six in segments A2/A3 as previously described

| Quantification and statistical analysis
Data were analyzed using Prism (GraphPad Software, CA). Per default two-sided Student's t test was performed to compare the means of two groups unless the data were non-normally distributed (as assessed D'Agostino-Pearson omnibus normality test) in which case they were compared by a two-tailed Mann-Whitney U test. For comparison of more than two groups, one-way analysis of variance (ANOVA) tests were used, followed by a Tukey's multiple comparison test. In the case