The vesicular SNARE Synaptobrevin is required for Semaphorin 3A axonal repulsion

The guidance of axonal and dendritic growth cones to their specific cellular target is essential for the establishment of functional neural circuits during development (Tessier-Lavigne and Goodman, 1996). Attractive and repulsive external guidance cues bind to receptors, which activates intracellular signaling pathways and reshapes the growth cone, allowing for positive and negative turning, respectively. Several types of endosomal vesicles have been identified in growth cones, but the role of exocytosis and endocytosis in growth and navigation is still largely unknown (Pfenninger, 2009). Exocytosis involves the fusion of intracellular vesicles with the plasma membrane and relies on soluble SNARE proteins located on both vesicular and plasma membranes. The synaptic vesicular SNARE Synaptobrevin 2 (Syb2; also called VAMP2) is the target of several clostridial neurotoxins, including tetanus neurotoxin (TeNT), which blocks neurotransmitter release (Schiavo et al., 2000). Knockout of Syb2 causes an almost complete loss of spontaneous and evoked release of neurotransmitters leading to postnatal death, but has not been associated with a developmental defect in the mouse brain (Schoch et al., 2001). However, there is evidence suggesting that Syb2 may play a role in the development of cultured neurons on specific substrates. Syb2 was recently shown to be required for axon formation in neurons cultured on poly-lysine but not laminin (Gupton and Gertler, 2010). TeNT inhibits calcium-triggered positive turning in neurons grown on an artificial L1 substrate but has no effect on calcium-triggered negative turning in neurons grown on laminin. The later result may suggest that exocytosis is required for attraction, but not repulsion (Tojima et al., 2007), and draw attention to Syb2 as a candidate gene for axonal guidance (Tojima et al., 2011). Previous studies using Semaphorin 3A (Sema3A) have hinted at a complex view. Indeed, repulsion mediated by Sema3A induces massive endocytosis of membrane (Fournier et al., 2000) and of its receptor (Castellani et al., 2004; Piper et al., 2005). After Sema3A-induced endocytosis of its receptor, resensitization requires reexpression and/or recycling of this receptor to the growth cone surface (Piper et al., 2005). Thus, axon guidance may depend on a regulation of both exocytosis and endocytosis.

As alluded to in the previous paragraph, it is not clear whether Syb2 functions in axonal guidance and, if so, how. Here, we took advantage of clostridial neurotoxins and genetic invalidation to demonstrate that Syb2 is required for Sema3A-mediated repulsion, but not for Sema3C-mediated attraction, in vitro and in vivo. We found that Syb2 associates with Neuropilin 1 (Nrp1) and Plexin A1 (PlexA1), two essential components of the Sema3A receptor. This interaction involves Syb2 juxtatransmembrane domain, which suggests a strong biochemical link during vesicular traffic between Syb2 and the Sema3A receptor. We further show that Sema3A inhibits the exocytosis of Syb2. Our findings thus demonstrate that Sema3A repulsion depends on Syb2-dependent vesicular traffic.

To determine the function of Syb2 in axon guidance, we analyzed the guidance responses of cortical explants treated with botulinum neurotoxin D (BoNT/D; Schiavo et al., 2000) and cortical explants from Syb2 knockout mice (Schoch et al., 2001). The explants were embedded in plasma matrix facing an aggregate of cells stably secreting Sema3C, Sema3A, or mock. After 24 h of culture, we analyzed both the ratio of proximal over distal axonal length as well as the turning angles of the axons in the area perpendicular to the gradient (Fig. 1 A; Bagnard et al., 1998).

In 3D culture systems, Syb2 is not required for axonal growth of both cortical and dorsal root ganglion (DRG) neurons, as shown by BoNT/D and genetic invalidation (Fig. S1). However, we found a striking requirement of Syb2 for Sema3A-induced repulsion, but not for Sema3C-induced attraction in E15 cortical explants (Fig. 1). In the control condition, Sema3C induced an increase of axonal length in the proximal area and an attractive turning of the axons toward the source as described previously (Bagnard et al., 1998). BoNT/D treatment fully cleaved Syb2 (Fig. S1) but did not affect Sema3C-induced attraction (Fig. 1, B–E). On the contrary, Sema3A decreased axonal length in the proximal area and repelled the axons away from the source in control explants, but did not have any effect in BoNT/D-treated explants (Fig. 1, B–E). To further demonstrate the need for a functional Syb2 in Sema3A repulsion but not Sema3C attraction, we performed guidance assays in Syb2^+/+, Syb2^+/−, and Syb2^−/− explants. We found that the absence of Syb2 in −/− cortical explants abolished the repulsive effects of Sema3A, whereas it had no effect on Sema3C-mediated attraction (Fig. 1, F–I). These results demonstrate that Syb2 is required for the repulsion induced by Sema3A but not for growth or the attraction mediated by Sema3C in cortical axons ex vivo.

We then asked whether Syb2 controls the repulsive response to Sema3A by affecting the acute chemotactic responses of axonal growth cones from isolated neurons in a well-defined concentration gradient. We first cultured dissociated embryonic day 15 (E15) cortical neurons grown in microchambers framed by two reservoirs to produce a linear gradient. We then generated a Sema3A or Sema3C gradient and observed axonal growth cone turning 3 h after the onset of the gradient by time-lapse video microscopy, and quantified the results as described previously (Yam et al., 2009). We found that cortical axons respond to Sema3C by turning toward the source whether or not Syb2 was acutely inactivated by BoNT/D (Fig. 2, A and B). However, Sema3A repulsive turning was abolished by treatment with BoNT/D (Fig. 2, A and B). These experiments further show that Syb2 is not necessary for Sema3C-induced attraction, but is essential for Sema3A-induced repulsion of isolated cortical axonal growth cones.

Genetic invalidation of Sema3A induces profound defasciculation of peripheral nerves (Kitsukawa et al., 1997). However, these nerves express both Syb1 and Syb2 (Trimble et al., 1990), which, with Cellubrevin, are functionally interchangeable v-SNAREs (Deák et al., 2006; Liu et al., 2011). Thus, both Sybs1 and -2 would need to be inactivated to obliterate their function in peripheral nerves. To this end, we cultured DRG explants embedded in collagen 3D matrix facing an aggregate of cells secreting Sema3A or mock (Ben-Zvi et al., 2008), and we treated them with BoNT/D, to remove both Syb1 and Syb2. BoNT/D-treated DRG axons were not repelled by Sema3A (Fig. S2, A–C), which confirms our observations in cortical explants. Not surprisingly, in toto neurofilament immunostaining in Syb2^−/− did not reveal a massive defasciculation of primary sensory fibers around the eyes nor in the trunk and limbs (Fig. S2 D), unlike what was observed in Sema3A^−/− and Nrp1^−/− mice (Kitsukawa et al., 1997). In light of the evidence of redundant function of Syb1 and Syb2 in peripheral nerves, these results suggest that the lack of guidance defect in peripheral nerves in Syb2^−/− is likely caused by a compensatory expression of Syb1.

We went on to test if Syb2 inactivation would generate a defect in Sema3A-mediated guidance in the forebrain in vivo. The corpus callosum is a region where Sema3A and Sema3C play a role in axon guidance. Indeed, mutations in the Sema-binding domain of Nrp1, coreceptor of Sema3A with PlexA1 (Takahashi et al., 1999), prevent both Sema3A and Sema3C activation and generate misguidance of Nrp1-positive fibers in the corpus callosum (Hatanaka et al., 2009; Piper et al., 2009). Furthermore, a possible reduction in Nrp1-expressing axons observed in the corpus callosum of Sema3A^−/− mice (Piper et al., 2009) may be caused by defasciculation of Nrp1-positive fibers. Thus, the lack of Sema3A repulsion in Syb2-deficient cortical neurons may generate a defect in Nrp1-positive fibers in the corpus callosum. To investigate this issue, we analyzed the organization of the corpus callosum at E18 in Syb2^+/+ and Syb2^−/− brains by immunostaining. We observed that Nrp1-positive axons presented a disorganization within the corpus callosum in Syb2^−/− E18 embryos (Fig. 3 A), which is reminiscent of the mild phenotype described in Sema3A^−/− mice (Piper et al., 2009), but clearly different from the one observed in Sema3C^−/− mice (Niquille et al., 2009). Nrp1-positive fiber tract is 33% thicker in Syb2^−/− than in Syb2^+/+ embryos (Nrp1 tract thickness as a percentage of total corpus callosum: 40.5 ± 0.97% in Syb2^+/+, 54.2 ± 1.60% in Syb2^−/−, P = 4.18 × 0.10⁻⁶ by Student’s t test) and displayed an overall weaker staining profile in Syb2^−/− as compared with Syb2^+/+ (Fig. 3 B). These results are in favor of a decompaction/defasciculation of Nrp1-positive fibers in the Syb2 knockout.

Altogether, these results further support the notion that Syb2 is required for Sema3A-mediated repulsion both in vitro and in vivo. We did not find any requirement for Syb2 in Sema3C-mediated attraction, which suggests the potential involvement, if any, of another traffic pathway.

In vitro and in vivo studies have shown that the Sema3A receptor is composed of Nrp1 and PlexA1 (Takahashi et al., 1999). To determine the role of Syb2 in Nrp1/PlexA1 traffic, we first tested whether or not Syb2, Nrp1, and PlexA1 interact in developing neurons. We reasoned on the basis that Syb2 interacts with other vesicular proteins including Synaptophysin (Syp) and V-ATPase (Edelmann et al., 1995; Galli et al., 1996) that are sorted together into synaptic vesicles in mature neurons. Here, we found that Syb2 coimmunoprecipitated Nrp1 and PlexA1 in E15 brain extracts, but not in adult brain extracts (Fig. 4, A and B). In addition, Syb2 coimmunoprecipitated Syp, but not L1, N-cadherin, or Syntaxin 1 (Stx1), a target SNARE partner of Syb2 in neurotransmitter release, which suggests that Syb2–Nrp1–PlexA1 interaction is not a mere result of detergent solubilization (Fig. 4 A). Tetanus insensitive–vesicle associated membrane protein (TI-VAMP)/VAMP7, another v-SNARE expressed in neurons, did not coimmunoprecipitate Nrp1 and coimmunoprecipitated only small amounts of PlexA1. Furthermore, Stx1 did not coimmunoprecipitate Nrp1 or PlexA1, which provides additional evidence for the specificity of the interaction with Syb2 (Fig. S3 A). The reverse experiment using Nrp1 and PlexA1 as bait failed to coimmunoprecipitate Syb2, most likely because of the fact that the available tested antibodies could not recognize the tripartite complex because the Sema3A receptor has a specific conformation when bound to Syb2. To circumvent this problem and further identify which domain of Syb2 mediates the interaction with Nrp1 and PlexA1, we reconstituted the complex in heterologous HEK-293T cells by expressing tagged Nrp1, PlexA1, and Syb2. GFP-Syb2, bearing GFP at the N-terminal cytosolic end, or Syb2–Super Ecliptic Phluorin (Syb2-SEP), bearing a pH-sensitive GFP at the C-terminal luminal extremity, were cleaved or not with cotransfected wild-type or mutant TeNT light chain, respectively (Fig. 4 C). The interaction between Nrp1, PlexA1, and full-length Syb2 was demonstrated by coimmunoprecipitation using each of the respective tags (Fig. 4 D). Although this interaction was maintained in the TeNT-cleaved Syb2-SEP, it was abolished in the TeNT-cleaved GFP-Syb2, showing that Syb2 interacts with Nrp1 and PlexA1 through its transmembrane and linker domain (Fig. 4 D). Additionally, Syb2 and Nrp1 strongly colocalized in intracellular membranes, particularly early endosomes (EEA1 staining) and clathrin-coated pits (AP2 labeling), when coexpressed with PlexA1 in Cos7 cells (Fig. S3 B). These results suggest that (a) the recognition between Syb2 and Nrp1/PlexA1 is highly specific, and the v-SNARE is part of a complex excluding its t-SNARE partner Stx1; and (b) Syb2 and the Sema3A receptor interact and traffic together in endocytic and/or recycling vesicles. Moreover, several single transmembrane domain proteins, including Nrp1 and Syb2, were shown to share similar structural features in their transmembrane regions susceptible to induce homo- or heterodimerization (Hubert et al., 2010). Our results showing that Syb2 interacts with Nrp1/PlexA1 through its juxtatransmembrane domain suggest that this region may play a pivotal role in sorting together Syb2, Nrp1, and PlexA1. Additional work will be required to further characterize the function of these transmembrane domains in the co-sorting of Syb2 and Sema3A receptors.

In developing neurons, Syb2 is transported from the Golgi apparatus to the surface of dendrites, being afterward transcytosed to the axon, where it mediates exocytosis (Sampo et al., 2003). Therefore, inactivation of Syb2 in developing neurons could lead to a defective polarization of Nrp1/PlexA1 and/or defective intracellular traffic of Nrp1/PlexA1. The first hypothesis is contradicted by the expression of Nrp1 and PlexA1 at the cell surface in Syb2^−/− neurons (Fig. S3 C). When axons encounter an isotropic concentration of Sema3A, they collapse and retract due to massive endocytosis of the membrane and of the Nrp1/PlexA1 receptor in growth cones (Fournier et al., 2000; Piper et al., 2005). To test if Syb2 is involved in Sema3A-induced endocytosis, we measured axonal length and PlexA1 localization in dissociated cortical neurons in the presence or absence of Syb2 upon Sema3A treatment. Sema3A led to a 40% reduction of axonal length of wild-type axons, but had no effect on axonal length in the absence of Syb2 in both Syb2^−/− or BoNT/D-treated cortical neurons (Fig. 5, A–C). Furthermore, Sema3A led to both depletion of PlexA1 in growth cones and a 60% increase of its concentration in the cell body in the presence of Syb2, but had no effect in Syb2^−/− and BoNT/D-treated cortical neurons (Fig. 5 D–F; and Fig. S3 D). Thus, these data further demonstrate that axons lacking a functional Syb2 are insensitive to Sema3A and suggest that Syb2 is required for the endocytosis, retrograde transport of Nrp1/PlexA1 receptor upon Sema3A activation, and signaling downstream of the Sema3A receptor.

In turn, activation of the Sema3A receptor may be expected to regulate Syb2-dependent secretion. As shown earlier, Sema3A induces the collapse of growth cones, thus reducing cell surface. Decrease in growth cone cell surface complicates proper quantification of the frequency of exocytic events by optical imaging. Therefore, we reconstituted Syb2 and Sema3A receptor in a heterologous system that is not prone to collapse. We imaged Syb2-SEP exocytosis in Cos7 cells in which Nrp1 and PlexA1 were coexpressed to reconstitute Sema3A receptor. We found that control Fc treatment did not affect exocytosis, whereas the number of exocytic events decreased by 50% upon addition of Sema3A (Fig. 5, G and H; and Video 1). As a control, we found no effect of Sema3A on Syb2 exocytosis when we expressed a mutant of Nrp1 lacking its cytosolic domain (Valdembri et al., 2009). This result reveals that Sema3A signaling through Nrp1 and PlexA1 inhibits Syb2 exocytosis.

Collectively, these results suggest that Sema3A signaling requires Syb2-dependent traffic. This is in good agreement with the role of Stx1B in Sema3A repulsion (Kabayama et al., 2011). Our results further agree with the dual role of Syb2 in exocytosis and endocytosis of synaptic vesicles (Deák et al., 2004). Given that Syb2 interacts with clathrin assembly lymphoid myeloid leukemia protein (CALM), a clathrin adaptor (Koo et al., 2011), it is tempting to speculate that Syb2 is required for the endocytosis of the Sema3A receptor in growth cones. Then, as shown here, Syb2 and the Sema3A receptor would be sorted together into endosomal vesicles because of an interaction mediated by their transmembrane domains. Sema3A further inhibited Syb2-dependent exocytosis when we reconstituted Sema3A receptor in nonneuronal cells. On this basis, we consider that the model proposing that attraction relies on Syb2 exocytosis, whereas repulsion relies on endocytosis (Tojima et al., 2011), needs to be revised. We instead propose that the molecular mechanism of Sema3A repulsion is based on fine tuning of the opposing processes of Syb2-dependent endocytosis and exocytosis (Fig. 5 I). Sema3A would induce a local disequilibrium at the site facing the Sema3A source by simultaneously triggering endocytosis and inhibiting exocytosis. Such a combination of locally increased endocytosis and decreased exocytosis would prevent progression in the direction of the source, favor turning of the growth cone in the opposite direction, and resensitize it to Sema3A (Piper et al., 2005). It is possible that TI-VAMP’s role in Netrin1 attraction (Cotrufo et al., 2011) may also be based on a change of exocytosis/endocytosis equilibrium, further generalizing the principle unraveled here. Therefore we propose that a main function of Syb2 in developing neurons is to traffic Sema3A receptor, allowing for its endocytosis, traffic away from the Sema3A source, and reexpression at the cell surface by exocytosis. It now remains to be tested whether the primary role of Syb2 resides in endocytosis or exocytosis of the Sema3A receptor, and how Sema3A signaling controls Syb2 exocytosis.

The following antibodies were used: Nrp1 (Western blotting [WB], 1/1,000 [Millipore]; immunofluorescence [IF], 1/500; [R&D Systems]), PlexA1 (WB, 1/1,000; IF, 1/500; Abcam), tubulin (IF, 1/10,000; Developmental Studies Hybridoma Bank), AP2 (WB, 1/1,000 [Abcam]; IF, 1/1,000; Thermo Fisher Scientific), L1 (a gift from F. Rathjen, Max-Delbrück-Centrum für Molekulare Medizin, Berlin, Germany; WB, 1/1,000; IF, 1/200), Syb2 (69.1 clone, a gift from R. Jahn, Max Planck Institute, Göttingen, Germany; WB, 1/10,000; IF, 1/1,000), Syntaxin1 (HPC1 clone, a gift from C. Barnstable, Yale University, New Haven, CT; WB, 1/10,000), Syp (7.2 clone, a gift from R. Jahn, Max Planck Institute, Goettingen, Germany; WB, 1/500), control IgG (Sigma-Aldrich), TI-VAMP (158.2; WB, 1/1,000), Tau (Millipore; IF, 1/1,000), EEA1 (Applied Biosystems; IF, 1/500), FLAG (Sigma Aldrich), GFP (Roche) and RFP (Millipore). DAPI was from Invitrogen.

We are grateful to T. Südhof (Stanford University, Stanford, CA) for the generous gift of the Syb2 knockout mouse. E13 and E15 embryos were extracted from Syb2 strain or Swiss mice, and DRG and cortical explants were embedded in front of stably transfected Sema3A, Sema3C, or mock-secreting cells, in 3D matrices, as described previously (Bagnard et al., 1998; Ben-Zvi et al., 2008). In brief, DRG and cortex were extracted in cold d-PBS and GBSS + glucose, respectively. DRGs were placed on a cushion of collagen gel in front of an aggregate of mock or Sema3A-secreting cells before the application of a second cushion of collagen containing 100 ng/ml h-NGF (Euromedex). Blocks of cortex (∼200 µm in diameter) were positioned next to an aggregate of mock-, Sema3A-, or Sema3C-secreting cells in 20 µl of chicken plasma supplemented with 0.8 U thrombin (Sigma-Aldrich). Neuronal culture medium (Neurobasal 1×, 2% B27, 1% l-glutamine, and 1% penicillin/streptomycin) was added after polymerization of the matrices. Cultures were fixed after 24 h of culture and immunostained for tubulin to analyze axonal lengths and surfaces as described previously (Bagnard et al., 1998).

E15 cortical neurons from the SWISS strain were dissected and dissociated before being plated at 10 µl of 6 × 10⁶ cells/ml in loaded medium (MEM 1×, 10% SVF, 1% l-glutamine, and 1% penicillin/streptomycin) in the observation area of hydrophobic chambers (Ibidi) precoated with poly-l-lysine and laminin (10 µg/ml each). Neurons were allowed to attach overnight before loading the neuronal culture medium with or without 30 nM BoNT/D. At day in vitro 2 (DIV2), 3 µg/ml of guidance cues were added or not added in the top chamber, and neurons were imaged for 3 h. Turning angles were measured as described previously (Yam et al., 2009).

E15 cortical neurons from Syb2 or SWISS strains were dissected and dissociated before being plated in loaded medium, on glass coverslips precoated with poly-L-lysine and laminin (10 µg/ml each, overnight in d-PBS and 3 h in Neurobasal medium, respectively). After 3 h of attachment, culture medium was changed to neuronal medium with or without 30 nM BoNT/D. At DIV2, 1 or 3 µg/ml Sema3A-Fc or plain media was added on cortical culture in fresh neuronal medium and incubated for 3 h before fixation with 4% paraformaldehyde (PFA)–4% sucrose and immunostaining.

We are grateful to V. Castellani (Université Claude Bernard Lyon, Villeurbanne, France) for PlexinA1-FLAG cDNA. E15 brain or DIV1 transfected HEK-293T cells were lysed in lysis buffer (TSE-1% Triton X-100, 1% Complete [Roche]) at 4°C. After 20 min of centrifugation at 13,000 g, total proteins were quantified with Bradford assays. 1 mg of total proteins was incubated overnight at 4°C with 5 µg of antibody and 150 µg of protein G Dynabeads (Applied Biosystems), prewashed with lysis buffer. After removal of supernatant and three washes of beads with lysis buffer, proteins were eluted in SDS buffer (loading buffers were obtained from Applied Biosystems) and incubated for 10 min at 95°C. Eluted proteins were finally loaded on 4–12% BisTris NuPAGE gel (Applied Biosystems). The gel was run and transferred to nitrocellulose membrane for Western blot staining. The membrane was blocked by incubation with TBS-1% Tween-20, 5% milk before incubation of primary antibody in TBS-1% Tween-20, 5% milk. After three washes, fluorescent secondary antibodies TBS-1% Tween-20 were incubated, and protein staining was revealed with the Odyssey system (LI-COR Biosciences).

DIV2 E15 neurons or Cos7 cell cultures were fixed for 20 min with 4% PFA; quenched for 20 min in PBS 1× and 50 mM NH₄Cl; permeabilized 4 min in PBS 1× and 0.1% Triton X-100; and blocked for 30 min in PBS 1× and 2.5% fish gelatin. Primary antibodies were incubated either overnight at 4°C or for 2 h at room temperature, in PBS 1× and 0.125% fish gelatin. After washes, secondary antibodies were incubated for 45 min at room temperature before mounting in Prolong medium (Applied Biosystems). For surface staining, DIV2 E15 neurons were processed as described previously without the permeabilization step.

E18 brains were fixed overnight in 4% PFA, washed in PBS 1×, and embedded in 3% low-melt agarose for vibratome sectioning. Slices were first permeabilized for 3 h in PBS 1×, 1% Triton X-100, and 0.25% fish gelatin before incubation of primary antibodies for 2 d at 4°C in PBS 1×, 0.1% Triton X-100, and 0.125% fish gelatin. Slices were washed three times for 2 h, then secondary antibodies were incubated overnight at 4°C before mounting of slices in Vectashield medium (Vector Laboratories).

E12 embryos were fixed overnight in 4% PFA; washed in PBS 1×; dried in methanol series before being whitened 3 h in 80% methanol, 5% DMSO, and 1% H₂O₂ (Huber et al., 2005); and permeabilized for 3 h in PBS 1×, 1% Triton X-100, and 5% milk. Embryos were incubated with 1/100 2H3 primary antibodies in PBS 1× and 1% Triton X-100 for 2 d at room temperature, washed three times for 3 h, and incubated with Cy3-goat anti–mouse secondary antibodies and DAPI for 1 d at room temperature.

Cos7 cells were transfected with an Amaxa system according to the manufacturer’s instructions (Lonza). HEK-293T cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions.

Syb2-SEP time-lapse video imaging was performed in modified Krebs-Ringer-Hepes buffer (140 mM NaCl, 4.7 mM KCl, 2 mM MgCl₂, 1 mM CaCl₂, 10 mM Hepes, and 5.5 mM glucose, pH 7.4) at 37°C with an inverted epifluorescence microscope (DMI6000B; Leica) modified to use laser illumination. The 488-nm argon ion laser beam was sufficiently expanded to get near uniform illumination. Fluorescent excitation and collection were performed with 63×/1.6 NA Plan-Apochromat oil-immersion objective lenses. Images were acquired every 500 ms, with an integration time of 100 ms using a highly sensitive EM charge-coupled device camera (Cascade 512B; Roper Scientific) controlled by MetaMorph software (Molecular Devices).

For explants assays, images were taken with a 2.5× objective lens on microscope (DMRD; Leica) controlled by MetaVue software (Molecular Devices). In vitro and in vivo immunofluorescence images were taken with, respectively, oil-immersion 40× or 63× and dry 10× or 20× objective lenses in a confocal microscope (TCS SP5; Leica) controlled the Leica Application Suite. In toto embryo images were taken with a 0.8× dry objective lens in a stereomicroscope (LUMAR; Carl Zeiss) and assembled with CombineZ software. Axon lengths and surfaces were analyzed using MetaMorph software (Molecular Devices). Axon turnings quantifications were performed with ImageJ.

Fig. S1 demonstrates that Syb2 is not involved in axonal growth in both E15 cortical and E13 DRG neurons. Fig. S2 shows that Syb2 is necessary for Sema3A-mediated repulsion but not Sema3A-induced growth inhibition in E13 DRG; E12 Syb2 knockout embryos do not display major defect in peripheral nerves, likely because of compensation by Syb1. Fig. S3 shows control immunoprecipitation experiments (TI-VAMP and Syntaxin1 as baits) in E15 brain; the colocalization of Syb2, Nrp1, and endocytic markers; the surface labeling of Nrp1 and PlexA1 in Syb2 knockout E15 cortical neurons; and labeling related to the graph in Fig. 5 F. Video 1 displays the effect of Sema3A on Syb2-exocytosis imaged with Syb2-SEP in a Cos7 cell.

We are grateful to Lydia Danglot, Alessandra Pierani, Marie-Christine Simmler, and David Tareste, for critical reading of the manuscript. We are grateful to Alessandra Pierani, Patricia Gaspar, and the Galli laboratory for helpful discussions.

Work in our group was funded in part by grants from Institut National de la Santé et de la Recherche Médicale, the Association Française contre les Myopathies (AFM), the Association pour la Recherche sur le Cancer (ARC), the Mairie de Paris Medical Research and Health Program, the Fondation pour la Recherche Médicale (FRM), the Ecole des Neurosciences de Paris (ENP; to T. Galli). K. Zylbersztejn was supported by a Bourse de Docteur Ingénieur (BDI) fellowship from the Centre National de la Recherche Scientifique and a doctoral fellowship from the FRM. M. Petkovic was supported by the ENP PhD program. A. Burgo was supported by FRM and Mairie de Paris. G. Serini was supported by Telethon and Associazione Italiana per la Ricerca sul Cancro (AIRC). M. Deck was supported by the ARC, and S. Garel was supported by a starting grant from the City of Paris and the European Young Investigators program.

Author contributions: K. Zylbersztejn and T. Galli conceived and designed the experiments, analyzed data, and wrote the paper. K. Zylbersztejn, M. Petkovic, A. Burgo, and M. Deck performed experiments. S. Garel, S. Marcos, E. Bloch-Gallego, F. Nothias, D. Bagnard, T. Binz, and G. Serini helped to set up the project, provided reagents, and co-wrote the paper.