Tetraspanin-3 is an organizer of the multi-subunit Nogo-A signaling complex

Growth inhibitory proteins restrict neurite outgrowth in the adult central nervous system (Giger et al., 2010; Schwab and Strittmatter, 2014; Yiu and He, 2006). Among those, the transmembrane protein Nogo-A (an isoform encoded by RTN4), which is enriched in central nervous system (CNS) myelin (Caroni and Schwab, 1988; Chen et al., 2000), is one of the most potent known inhibitors of regeneration and plasticity after injury (Cafferty and Strittmatter, 2006; Schwab, 2010; Schwab and Strittmatter, 2014; Wahl et al., 2014). Nogo-A restricts neurite outgrowth through two different domains – Nogo-66 (rat, amino acid residues 1026–1091) and Nogo-A-Δ20 (rat, amino acid residues 544–725; part of the ‘amino-Nogo’ domain) (Chen et al., 2000; GrandPré et al., 2000; Oertle et al., 2003; Schwab, 2010). In vivo, full-length Nogo-A might undergo transcytosis or proteolytic cleavage, resulting in the release of active fragments such as Nogo-A-Δ20 (Joset et al., 2010; Kempf et al., 2014; Thiede-Stan and Schwab, 2015). In the developing CNS, Nogo-A-Δ20 promotes the migration of neuronal precursors but inhibits primary brain microvascular endothelial cell migration, as well as the spreading of endothelial cells and fibroblasts (Kempf et al., 2014; Mathis et al., 2010; Oertle et al., 2003; Rolando et al., 2012; Schmandke et al., 2013; Walchli et al., 2013). Although Nogo-A-Δ20 and Nogo-66 signal through different receptors, the activation of RhoA is a common intracellular signal step downstream of both inhibitory domains (Joset et al., 2010; Kempf et al., 2014; Montani et al., 2009; Nash et al., 2009). Nogo-66 signals through a co-receptor complex containing NgR1 (also known as RTN4RL1) and p75 (also known as TNFRSF1B) or TROY; the G-protein-coupled receptor (GPCR) sphingosine 1-phosphate receptor 2 (S1PR2) has recently been identified as the signal-transducing receptor for Nogo-A-Δ20 (Kempf et al., 2014).

Because several ligand–receptor interactions are modulated by co-receptors or specific functional membrane scaffolds, we were interested in components that could exert such a modulation and fine-tuning of the Nogo-A-Δ20-induced S1PR2 signaling. Based on the same yeast two-hybrid screen that has recently been used to reveal S1PR2 as a Nogo-A-Δ20 interaction partner, we identified the four-transmembrane-spanning protein tetraspanin-3 (TSPAN3) as an additional interaction partner of Nogo-A-Δ20. The tetraspanin family comprises 33 mammalian cell-type-specific members (Hemler, 2005). In the CNS, TSPAN3 is expressed in neurons, oligodendrocytes and astrocytes (Tiwari-Woodruff et al., 2001, 2004). In the immune and vascular systems, tetraspanin proteins act as organizers of functional microdomains and multi-subunit receptor complexes (Charrin et al., 2014; Hemler, 2005; Junge et al., 2009; Levy and Shoham, 2005; Mattila et al., 2013). Here, we show that TSPAN3 acts as an organizer of the signaling complex and as a core element for the complex comprising Nogo-A-Δ20 and the S1PR2 G-protein-coupled receptor, ensuring neurite outgrowth and spreading inhibition. TSPAN3 is required for the initial contact of Nogo-A with the cell surface and its assembly into the S1PR2-containing signaling platforms – TSPAN3 facilitates Nogo-A-Δ20-induced S1PR2 clustering, and silencing of TSPAN3 expression abolishes Nogo-A-induced RhoA activation. Potent Nogo-A-Δ20-induced neurite outgrowth and spreading restriction depends on the interaction with tetraspanins and the presence of spatiotemporally coordinated TSPAN3 microdomains. These findings reveal the importance of a tetraspanin co-receptor in Nogo-A–signal-receptor complexes for inhibition of neurite outgrowth. These insights support the idea that tetraspanins have similar modes of action in other receptor complexes that regulate cell adhesion, spreading and migration during development, immune responses and cancer.

TSPAN3 has been identified as an interaction partner of Nogo-A-Δ20 in a yeast two-hybrid screen conducted under very stringent conditions, which also led to the identification of S1PR2 as a signal-transducing receptor for Nogo-A-Δ20 (Kempf et al., 2014). In order to validate the indicated interaction of Nogo-A-Δ20 and TSPAN3, we performed co-immunoprecipitation experiments using NIH 3T3 cells, which endogenously express TSPAN3, as well as S1PR2, and strongly respond to Nogo-A-Δ20 by inhibiting cell spreading and exerting anti-adhesive effects (Joset et al., 2010; Kempf et al., 2014; Oertle et al., 2003). NIH 3T3 cells were incubated with hemagglutinin (HA)-tagged recombinant Nogo-A-Δ20. After lysis and immunoprecipitation with TSPAN3-specific antibodies, HA–Nogo-A-Δ20 was found in the TSPAN3-immunoprecipitated fractions but not in mouse IgG2b isotype control fractions (Fig. 1A).

We were next interested in the interaction of TSPAN3 and Nogo-A-Δ20 at the cell surface. Cell surface staining of living NIH 3T3 fibroblasts that had been incubated with Nogo-A-Δ20 at 4°C showed colocalization of TSPAN3 and Nogo-A-Δ20 in distinct clusters at the cell surface (Fig. 1B). Knockdown of TSPAN3 (30±8.8% of control; Fig. 2C) strongly reduced binding of Nogo-A-Δ20 at the cell surface (Fig. 1C,D; for knockdown quantification, see Fig. 2C). Cells that had been treated with small interfering (si)RNA against TSPAN3 (TSPAN3 siRNA) had significantly less Nogo-A-Δ20 at their surface (41.7±2.7 clusters/cell) compared to control-siRNA-transfected cells (81.6±4.5 clusters/cell, P<0.001; Fig. 1D). In addition, the mean size of Nogo-A-Δ20 surface-bound clusters was significantly smaller after knockdown of TSPAN3 (2.6±0.15 μm² vs 3.2±0.1 μm² in the control, P<0.001; Fig. 1D).

Thus, TSPAN3 is essential for binding of Nogo-A-Δ20 at the cell surface at an early stage of the ligand making contact with the cell surface.

The fate of the Nogo-A-Δ20–TSPAN3 complex was followed and quantified at different incubation times of Nogo-A-Δ20 (3–40 min) at 37°C; representative maximum projections of single cells are shown in Fig. 1E and supplementary material Fig. S1A. The image stacks of single NIH 3T3 cells revealed strong colocalization of TSPAN3 and Nogo-A-Δ20 at 3 min of incubation (68±1.8% Nogo-A-Δ20 colocalized with TSPAN3, 90±1.3% TSPAN3 colocalized with Nogo-A-Δ20; Fig. 1E,F), especially at the cell periphery along the ruffling membranes. After 7–15 min, Nogo-A-Δ20 was associated with and partially trans-located through wave-like ruffling membranes towards the dorsal cell surface, still colocalized with TSPAN3 (Fig. 1E,F; supplementary material Fig. S1A). At this time, ring-formation and reorganization of the local actin cytoskeleton began, which was associated with Nogo-A-Δ20 and TSPAN3 (supplementary material Fig. S1B), as well as TSPAN3-positive early endosomes (supplementary material Fig. S1C). TSPAN3 colocalized with Nogo-A-Δ20 in these EEA1-positive early endosomes (Fig. 1G). At 25 min, the trafficking of Nogo-A-Δ20 and TSPAN3 started to separate, indicated by declining colocalization (57.1±3.5% Nogo-A-Δ20 colocalized with TSPAN3, 64.9±3.2% TSPAN3 colocalized with Nogo-A-Δ20) (Fig. 1F; supplementary material Fig. S1A). An even lower colocalization at 40 min (27.6±2.0% Nogo-A-Δ20 colocalized with TSPAN3, 31.6±1.8% TSPAN3 colocalized with Nogo-A-Δ20; Fig. 1E,F) was consistent with a colocalization of TSPAN3 with Rab11-positive recycling endosomes, and enrichment of Nogo-A-Δ20 in Lamp-1-positive lysosomal late endosomes (Fig. 1H).

Because Nogo-A-Δ20 inhibits neurite outgrowth and the spreading of fibroblasts (Kempf et al., 2014; Oertle et al., 2003), we investigated whether there were changes in Nogo-A-Δ20-mediated functional effects in the absence of TSPAN3. First, we tested the ability of fibroblasts to spread on Nogo-A-Δ20 or on control substrate after stable lentiviral small hairpin (sh)RNA-mediated knockdown of TSPAN3. TSPAN3 knockdown (Fig. 2A) led to an increased spreading of fibroblasts (20.7%, P<0.001) on Nogo-A-Δ20 [batch-specific inhibitory concentration 50 (IC₅₀) concentration for control cells, 45 pmol/cm²] compared to shRNA control cells, thus reducing the Nogo-A-Δ20 inhibitory effect by about 40% (Fig. 2B). In order to investigate the specific relevance of this effect, we analyzed the effect of knocking down other tetraspanins on Nogo-A-Δ20-mediated inhibition of spreading. NIH 3T3 cells were transfected with siRNAs that targeted the mRNA encoding TSPAN3 or several other tetraspanin family members, which were either chosen owing to a rather close phylogenetic relationship (compared to other TSPAN family members) to TSPAN3 (TSPAN6, TSPAN7) (Hemler, 2001) or owing to their reported expression and functions in the CNS (TSPAN7, CD82, CD9 and TSPAN2) (Bassani and Passafaro, 2012; Mela and Goldman, 2009; Terada et al., 2002; Tole and Patterson, 1993). The siRNA transfections resulted in a reduction of the corresponding mRNA levels down to 23–60% of the controls, quantified by using quantitative reverse transcriptase (qRT)-PCR (Fig. 2C). Representative pictures of NIH 3T3 fibroblasts that had been transfected with siRNA targeting either TSPAN3 or TSPAN7 and plated on control or Nogo-A-Δ20 substrates are shown in Fig. 2D. Although spreading of NIH 3T3 cells was reduced from 100% on a control substrate to 67.2±1.8% on a low Nogo-A-Δ20 concentration (25 pmol/cm²), knockdown of TSPAN3 allowed 82.3±2.4% of the cells to spread (P<0.001; Fig. 2E). Knockdown of TSPAN2, TSPAN6, TSPAN7, CD9 or CD82 failed to counteract the Nogo-A-Δ20-induced spreading inhibition (Fig. 2E). As an additional control, we plated NIH 3T3 cells that had been transfected with siRNA against TSPAN3 or TSPAN7, or with a control siRNA onto culture dishes coated with the recombinant Nogo-A non-inhibitory fragment Nogo-A-Δ21 (Oertle et al., 2003). NIH 3T3 cell spreading was identical in all groups – i.e. unaffected by the TSPAN3 or TSPAN7 knockdown on this control substrate (supplementary material Fig. S2).

We next investigated the Nogo-A-Δ20-induced neurite outgrowth inhibition after TSPAN3 knockdown in embryonic cortical neurons. The mean neurite outgrowth length per neuron was reduced from 100±3.94% on a control substrate to 70.1±2.76% on Nogo-A-Δ20 (25 pmol/cm²) in control-siRNA-transfected cortical neurons, compared to 81.3±3.3% upon knockdown of TSPAN3. (P<0.01; Fig. 2H). Taken together, these results show that TSPAN3 is important for Nogo-A-Δ20-induced inhibition of fibroblast spreading and neurite outgrowth.

Interactions of tetraspanins with their partner proteins have been described as being spatially coordinated within nanoscale assemblies (Hemler, 2005), but existing also outside of these areas (Berditchevski and Rubinstein, 2013; Cornea and Conn, 2014; Espenel et al., 2008; Mattila et al., 2013). Although single micrometer-size clusters can be visualized with conventional fluorescence microscopy, quantitative measurements of nanometer-size clusters and single-molecule reorganizations require sub-diffraction-limited spatial resolution, comparable to the dimensions of the receptor-clusters (<300 nm) (Nydegger et al., 2006). TSPAN3 dynamics were visualized by tracking single molecules through an Atto647N-labeled antibody against TSPAN3 at different time points after addition of Nogo-A-Δ20 (0–19 min), at which TSPAN3–Nogo-A-Δ20 interactions were observed (Fig. 1E,F). Atto647N-labeling of the TSPAN3-targeting antibody did not interfere with detection of TSPAN3 in NIH 3T3 cells (supplementary material Fig. S3A). The TSPAN3-targeting antibody epitope binding-site includes extracellular loop (ECL) 1 but not ECL 2 of TSPAN3 (see Materials and Methods). This could explain why live visualization of the cell surface TSPAN3 with the Atto647N-labeled TSPAN3-specific antibody and the (possibly) direct Nogo-A-Δ20 interaction with TSPAN3 in cells was technically possible and successful. The positions of diffraction-limited spots were determined by Gaussian fitting of their point-spread function (PSF) and linked over time for single-particle tracking (Fig. 3A,B). During the imaging between time points at 0 and 19 min, the general cell movement was only minimal [supplementary material Fig. S3B; 5.68±1.29 cell area change (percentage of the cell area at baseline)] and therefore enabled stable imaging. By contrast, at time points later than 20 min, the cell movement increased [supplementary material Fig. S3B, 34.38±5.63 cell area change (percentage of the cell area at baseline)] and therefore, the imaging was stopped before this occurred. The mean diffusion coefficient (as calculated from the mean square displacement, MSD) of TSPAN3 molecules increased significantly upon incubation with Nogo-A-Δ20 at 1–16 min (0.037±0.003 μm²/s at 10–13 min) compared to PBS-treated cells (0.011±0.003 μm²/s, P<0.0001; Fig. 3B). Because we observed that TSPAN3 was located in confined areas at the cell surface, we hypothesized that this increased mobility is due to the Nogo-A-Δ20-dependent release of TSPAN3 molecules from these areas, as also suggested by an increased surface coverage of the tracks (Fig. 3A). To test this hypothesis quantitatively, we followed the motion of individual molecules of TSPAN3 with quantum dots (Qdots) to detect dynamic changes in the mobility of individual molecules after Nogo-A-Δ20 addition (Fig. 3C–G). By plotting the diffusion coefficient D versus the slope of the moment scaling spectrum, S_MSS, of several tracks (Ewers et al., 2005), we found mainly three distinct patterns of motion (Fig. 3C,D) – (i) Molecules confined to areas of ∼500 nm in diameter (blue in Fig. 3C), (ii) freely diffusing molecules (red) and (iii) molecules that alternated between these two states (black). After 3–11 min of Nogo-A-Δ20-addition, fewer molecules remained in the confined state (27±2.4%) than in untreated cells (52.4±3.6%, P<0.05, Fig. 3E). In addition, in Nogo-A-Δ20-treated cells, we found significantly more single molecules that changed their pattern from confined to free motion, often repeatedly (44.1±9.9% vs 8.3±3.3% in the control group, P<0.01, Fig. 3F). Furthermore, TSPAN3 molecules showed an increased interaction with each other upon treatment with Nogo-A-Δ20 (control treatment 23.5±3.1% vs treatment with Nogo-A-Δ20 33.4±3.1% of total tracks, P<0.05; Fig. 3G). These results reveal that Nogo-A-Δ20 binding increases the mobility of TSPAN3 molecules and induces a spatiotemporal reorganization of cell surface TSPAN3 at the nanoscale level.

An important question was whether the Nogo-A-Δ20-induced TSPAN3 reorganization is related to interactions with the Nogo-A-Δ20 signal-transducing receptor S1PR2. In absence of purified recombinant Nogo-A-Δ20, we observed a partial colocalization of TSPAN3 and S1PR2 at endogenous expression levels in Nogo-A-Δ20-responsive fibroblasts and neurons. Patch-like cell surface colocalization of endogenous TSPAN3 and S1PR2 at peripheral regions of the cells were visualized with immunofluorescence in two different Nogo-A-Δ20-responsive cell types, non-neuronal NIH 3T3 fibroblasts and primary embryonic [embryonic day (E)19] cortical neurons (after 4 days in vitro) (Fig. 4A). Furthermore, S1PR2 was co-immunoprecipitated with TSPAN3 from NIH 3T3 cell lysates by using TSPAN3-specific antibodies (Fig. 4B). We then transfected HEK cells with either TSPAN3–MycDDK alone, TSPAN7–MycDDK alone, green fluorescent protein (GFP)-tagged S1PR2 (S1PR2–GFP), or TSPAN3–MycDDK as well as S1PR2–GFP. We also treated cells that had been transfected with TSPAN7–MycDDK and S1PR2–GFP with Nogo-A-Δ20, and cells were subsequently lysed with a mild detergent (Fig. 4C). Co-immunoprecipitation with either an antibody against MycDDK (mouse IgG; lanes 2, 3, 5, 7, 8, 9) or an isotype control (mouse IgG; lanes 4 and 6), and subsequent detection with an antibody against GFP revealed that S1PR2–GFP exclusively co-immunoprecipitated with TSPAN3–MycDDK, and not with TSPAN7–MycDDK or with an isotype control mouse IgG. Lysates 2 and 5 (overexpressing TSPAN3–MycDDK and S1PR2–GFP) correspond to lysates from two independent experiments. The association of S1PR2–GFP with TSPAN-3–MycDDK was further confirmed with mass spectrometry (data not shown), as using a similar approach to that for co-immunoprecipitation and detection on PVDF membrane. Therefore, immunoprecipitation of MycDDK–TSPAN3 from S1PR2–GFP- and TSPAN3-overexpressing HEK cells was subsequently followed (after gel electrophoresis) by mass spectrometry analysis of a Coomassie-stained gel band at ∼65 kDa (the mass of S1PR2–GFP).

On a qualitative level, immunofluorescence also indicated a partial colocalization, in a tripartite complex, of Nogo-A-Δ20, TSPAN3 and S1PR2 in distinct clusters in NIH 3T3 cells that had been incubated with Nogo-A-Δ20 for 15 min at 37°C (Fig. 4D).

To study whether the presence of Nogo-A-Δ20 enhanced complex formation of its receptor S1PR2 with TSPAN3, we used a proximity ligation assay (PLA) (Söderberg et al., 2006), which enables in situ visualization and quantification of protein–protein interactions within a proximity of a minimum of 40 nm (Fig. 4E–G). TSPAN3–S1PR2 interactions were quantified after 0, 10 and 25 min of incubation with Nogo-A-Δ20 (Fig. 4G). The number of foci, indicating interactions of TSPAN3 with S1PR2, at 0 min of Nogo-A-Δ20 incubation (5.2±0.4 PLA signals/cell) increased significantly after 10 min (23.2±3.3 PLA signals/cell, P<0.0001) and decreased after 25 min (13.3±1.6 PLA signals/cell, P<0.001 compared to 10 min, and P<0.05 compared to 0 min; Fig. 4G). By contrast, no PLA signals were obtained upon incubation with Nogo-A-Δ20 (for 0, 10 or 25 min) in NIH 3T3 cells that had been transfected with TSPAN3 siRNA (n=50 cells per group) (supplementary material Fig. S4A). These observations point to a temporally regulated dynamic complex assembly of the signal-transducing Nogo-A-Δ20 receptor S1PR2 and the co-receptor TSPAN3 in presence of Nogo-A-Δ20. The disassembly starts at the time of transition from early to late endocytosis, consistent with the previously described separation along different trafficking routes.

Key rearrangements of TSPAN3 in the plasma membrane at the nanoscale level and of complex formation with S1PR2 at the microscale level occurred mostly during the first 10 minutes of incubation with Nogo-A-Δ20. To study the role of TSPAN3 in the Nogo-A-induced clustering of S1PR2 and the subsequent intracellular signaling events, the receptor clustering of S1PR2 at the cell surface was quantified in control-siRNA- or TSPAN3-siRNA-transfected cells after control treatment or treatment with Nogo-A-Δ20 (Fig. 5A,B). The size of the Nogo-A-Δ20-induced S1PR2 cell surface clusters (1.067±0.05177 μm² with control treatment vs 1.562±0.08260 μm² in Nogo-A-Δ20-treated control-siRNA cells) was strongly decreased by knocking down TSPAN3 (0.9588±0.03730 μm² Nogo-A-Δ20-treated with TSPAN3 siRNA compared to Nogo-A-Δ20-treated siRNA control; P<0.0001; Fig. 5B, for knockdown quantification, see Fig. 2C). Comparing the total S1PR2 cell surface levels, siRNA-mediated knockdown of TSPAN3, compared to transfection with control siRNA, did not significantly change the presence of S1PR2 at the cell surface. However, upon the exposure to Nogo-A-Δ20, the resulting total S1PR2 cell surface levels remained at baseline upon siRNA-mediated knock down of TSPAN3 compared to that upon transfection with control siRNA (supplementary material Fig. S4B). This suggests that TSPAN3 is not generally involved in the transport of S1PR2 to the cell surface but that it does potentially stabilize a dynamic request for more S1PR2 at the cell surface in the presence of Nogo-A-Δ20.

RhoA GTPase activation is a decisive Nogo-A-Δ20-induced intracellular signaling step (Joset et al., 2010; Kempf et al., 2014). The siRNA-mediated downregulation of TSPAN3 resulted in a 58.0±6.3% decrease of Nogo-A-Δ20-induced RhoA-activation compared to that in control-siRNA Nogo-A-Δ20-treated cells (P<0.01; Fig. 5C,D, for knockdown quantification, see Fig. 2C). All these results demonstrate that TSPAN3 is an important functional player for Nogo-A-Δ20-induced S1PR2 clustering and subsequent signal transduction, leading to the inhibitory effects of Nogo-A-Δ20.

Finally, we investigated a potential additive effect of TSPAN3 and S1PR2 interference on the Nogo-A-Δ20-mediated inhibition of fibroblast spreading. The S1PR2-specific blocker JTE-013 (Marsolais and Rosen, 2009) has been shown to significantly counteract Nogo-A-Δ20-mediated inhibition of cell spreading (Kempf et al., 2014). We combined treatment with JTE-013 with TSPAN3-siRNA-mediated knockdown in NIH 3T3 cells that had been plated on Nogo-A-Δ20 substrate (Fig. 5E). Neither the single nor the combined treatments affected fibroblast spreading on the control plastic substrate. Combined treatment with JTE-013 and TSPAN3 knockdown [72.6±2.0% spreading compared to JTE-013 alone (66.2±2.0%) or TSPAN3 siRNA alone (69.3±1.6%), P>0.05; Fig. 5E] did not reveal a significantly enhanced Nogo-A-Δ20-mediated antagonism of the combined treatment. The absence of a strong additive effect of the combined S1PR2 and TSPAN3 blockage is in line with the findings that both proteins exist in the same receptor complex, mediating the Nogo-A-Δ20-induced inhibition of cell spreading in a concerted manner.

After addition of the Nogo-A active fragment Nogo-A-Δ20 to Nogo-A-responsive NIH 3T3 fibroblasts, Nogo-A-Δ20 colocalized with the membrane protein TSPAN3 on the cell surface.

Binding of Nogo-A-Δ20 was strongly reduced after siRNA-mediated knockdown of TSPAN3. Surface membrane TSPAN3 molecules colocalized with Nogo-A-Δ20, which progressively associated with S1PR2. These multi-subunit complexes were subsequently internalized into early endosomes. Knockdown of TSPAN3 reduced activation of the Nogo-A signal transducer RhoA and the inhibitory effects of Nogo-A-Δ20 on cell spreading and neurite outgrowth, showing that the tetraspanin TSPAN3 organizes, as a co-receptor, a signaling platform for Nogo-A-Δ20–S1PR2-mediated inhibition of neurite outgrowth and cell spreading.

Important roles of tetraspanins in the specific binding of the peptide ligands to the cell surfaces have been shown previously – e.g. during viral infection of cells – where distinct tetraspanin proteins can have crucial roles in the specific post-attachment binding phase of the virus to the cell surface, after initial viral attachment to heparin sulfate proteoglycans and/or low-density lipoprotein receptor-related proteins (LRPs) during viral infection of cells (Dahmane et al., 2014; Lindenbach and Rice, 2013; Monk and Partridge, 2012; Morikawa et al., 2007; Nydegger et al., 2006; Scheffer et al., 2013). To gain more insights into the functions of TSPAN3 in Nogo-A-Δ20 binding, we followed TSPAN3 lateral dynamics in the cell membrane at a nanoscopic level with single-particle-tracking. Within minutes after Nogo-A-Δ20 incubation, TSPAN3 molecules exhibited an increased mobility. In addition, Nogo-A-Δ20 recruited originally immobile TSPAN3 molecules and stimulated them to become more mobile and interactive with each other. Mixed phases of mobility and confinement have been documented for other tetraspanins where small clusters can co-diffuse, and exchange tetraspanins and their partner proteins (Charrin et al., 2009; Espenel et al., 2008; Mattila et al., 2013; Termini et al., 2014).

Interestingly, TSPAN3 and the Nogo-A signal-transducing GPCR S1PR2 showed increased colocalization after 10 min of exposure to Nogo-A ligand. Such dynamic multi-subunit receptor complex (Thiede-Stan and Schwab, 2015) assemblies in the presence of a ligand are known for other regulators of neurite growth – e.g. the neurotrophins BDNF and NGF, or myelin associated glycoprotein (MAG) (Ascano et al., 2009; Haapasalo et al., 2002; Marchetti et al., 2013; Vermehren-Schmaedick et al., 2014; Wong et al., 2002). Signal transduction does not only depend on a dynamic association of the partners of the signaling complex but also on the cellular compartment where the signaling platform forms. For instance, the interaction of MAG with gangliosides, or Nogo-66 binding to NgR1 induces p75 and NgR1 translocation to and RhoA concentration in specific microdomains (lipid rafts), which enhances downstream signaling (Fujitani et al., 2005; Vinson et al., 2003; Vyas et al., 2002; Yu et al., 2004).

Ligand–receptor interactions of neurotrophic factors, for example, or interactions with Nogo-A at the plasma membrane are followed by the internalization of the assembled complex and formation of signaling endosomes (Joset et al., 2010; Zweifel et al., 2005). Nogo-A-Δ20 internalization has been shown to be an indispensable step for RhoA activation and growth cone collapse (Joset et al., 2010). Co-internalization of Nogo-A and S1PR2 into early endosomes is detectable after 15 min of incubation (Kempf et al., 2014).

On the surface membrane, colocalization of TSPAN3 and Nogo-A-Δ20 was detectable at time points preceding internalization, first at ruffling membranes in the cell periphery and subsequently at dorsal circular macropinocytic ruffles. These structures have been shown previously to be induced and to be required for internalization of ligands like epidermal growth factor (EGF), the neurotrophin NGF, or for the re-distribution of integrins (Gu et al., 2011; Hoon et al., 2012; Valdez et al., 2007; Zweifel et al., 2005). TSPAN3 and Nogo-A-Δ20 were first found to be co-associated with local actin rings (7–15 min), which typically accompany macropinocytic events (Fournier et al., 2000; Kaksonen et al., 2000). At later time points (15–25 min), we observed TSPAN3 and Nogo-A-Δ20 in EEA1-positive early endosomes. Interestingly, the internalized complex of TSPAN3 and Nogo-A-Δ20 disassembled at later time points (40 min) during differential late trafficking. These results provide compelling evidence that TSPAN3 accompanies Nogo-A-Δ20 along its multistep route from cell surface binding to fusion with early endosomes. This spatiotemporally coordinated relationship between the peptide ligand and the tetraspanin co-receptor mediates reorganizations and translocations at the nano- and microscale levels, which are accompanied by an increased association with the signal-transducing GPCR.

In cells of the immune and vascular system, tetraspanin-enriched microdomains often serve as origins for subsequent ligand–receptor interactions in multi-subunit receptor complexes (Junge et al., 2009; Mattila et al., 2013) or as scaffolds for GPCR signaling (Little et al., 2004). Knockdown of TSPAN3 decreased the Nogo-A-Δ20-induced S1PR2 clustering at the cell membrane and the subsequent Nogo-A-Δ20-induced RhoA activation. These results point to a crucial role for the tetraspanin in the assembly and activation of the Nogo-A-Δ20–S1PR2 receptor complex.

Both the subcellular distribution of tetraspanins at the microscale level and the nanoscale organization of tetraspanins in microdomains are known to be regulated by palmitoylation of juxtamembrane cysteine residues (Berditchevski et al., 2002; Charrin et al., 2002; Termini et al., 2014; Yang et al., 2002). Pilot experiments with the general protein-palmitoylation inhibitor 2-bromopalmitate show that Nogo-A-Δ20-induced RhoA activation is decreased and that Nogo-A-Δ20-induced spreading inhibition is reduced, accompanied by reduced surface TSPAN3 levels and a lower number of Nogo-A-Δ20 surface clusters (unpublished observations, Thiede-Stan et al.). However, more general and complex effects of palmitoylation inhibition cannot be excluded at present. Analysis of the described experiments in palmitoylation-deficient TSPAN3-positive cells might provide a more detailed understanding of the described issues. Tetraspanin-promoted GPCR-clustering has been shown for the multi-subunit receptor complex of the GPCR Frizzled4, LRP5 and TSPAN12. This receptor complex is activated by the peptide ligand Norrin (as multimers), instead of the classic Wnt ligands. Importantly, TSPAN12 promotes the multimerization of Frizzled4 cooperatively with LRP5, thereby ensuring downstream signaling (Junge et al., 2009). A similar mechanism could be used for TSPAN3, S1PR2 and Nogo-A.

The activation of the small GTPase RhoA and the subsequent modulation of the cytoskeleton is a common effector step for ligands that induce acute neuronal growth cone collapse, restrict neurite outgrowth, and modulate cell spreading and adhesion (Fournier et al., 2003; Niederost et al., 2002; Parsons et al., 2010; Yiu and He, 2006). siRNA- or lentiviral shRNA-mediated TSPAN3 knockdown attenuated Nogo-A-Δ20-induced inhibition of cell spreading and neurite outgrowth, whereas knockdown of various other tetraspanin proteins did not lead to improved fibroblast spreading on a Nogo-A-Δ20 substrate. These results provide evidence for a crucial role of TSPAN3 for Nogo-A function, enabling full execution of the inhibitory effects that are induced by Nogo-A-Δ20.

Our results extend and refine the emerging concept that repulsive and attractive guidance cues, as well a diverse range of other ligands, signal through receptor complexes that comprise several interacting partners (Chao, 2003; Giger et al., 2008; Neufeld and Kessler, 2008; Niehrs, 2012). The complexes enable modulated cell-type- and developmental-stage-specific expression of the receptors, as shown, for instance, for Nogo-66 (Giger et al., 2008; Kempf et al., 2014) or neurotrophin receptors (Bartkowska et al., 2010; Chao, 2003). Thus, control of responses to highly potent repulsive or attractive cues and fine-tuning of the functional outputs are possible (Harrington et al., 2011; Janes et al., 2005; Seiradake et al., 2013). An interesting question for further studies is whether the Nogo-A receptors NgR1 and S1PR2, and their associated proteins and co-receptors are colocalized and activated by a single Nogo-A molecule through the two active sites Nogo-66 and Nogo-A-Δ20, respectively. The possibility of an independent mode of action is supported by the observation that cells which do not express NgR1 are still able to respond to Nogo-A-Δ20 (Fournier et al., 2001; Joset et al., 2010; Kempf et al., 2014; Oertle et al., 2003). Nogo-A-Δ20 can induce effects that are not induced by Nogo-66, such as inhibition of cell spreading, and migration of fibroblasts and primary brain microvascular endothelial cells (Walchli et al., 2013). NgR1 mediates acute growth cone collapse, but it is not required for long-term inhibition of neurite outgrowth (Chivatakarn et al., 2007); in addition, the neuronal growth cone collapse that is induced by Nogo-A-Δ20 depends on protein synthesis, whereas that induced by the Nogo-66 peptide does not (Manns et al., 2014). These observations suggest that Nogo-A receptors could comprise S1PR2, NgR1 and/or PirB, all in combination or separately, similar to the situation seen with neurotrophin or Wnt receptors, for example.

Tetraspanin proteins are organizers of functional microdomains in different cell types. Thereby, they can modulate the precision and specificity of ligand–receptor interactions and of the subsequent signaling cascades in multi-subunit receptor complexes (Hemler, 2005; Junge et al., 2009; Levy and Shoham, 2005; Mattila et al., 2013), notably during adhesion and cell migration (Powner et al., 2011; Termini et al., 2014). They are also involved in the malfunction of these mechanisms during cancer progression (Hemler, 2014). Interestingly, tumor cell migration and invasiveness are suppressed by tetraspanin CD82, but also by S1PR2 through a Rho GTPase-dependent PTEN phosphatase pathway (Lepley et al., 2005; Sanchez et al., 2005; Zhou et al., 2004). Our results exemplify how a tetraspanin, TSPAN3, acts as a modulatory co-receptor to regulate ligand–receptor interaction and signaling, influencing neurite outgrowth, cell spreading and adhesion in response to an important growth-inhibitory CNS protein, Nogo-A.

The identification of the Nogo-A-Δ20 interaction partner TSPAN3 was based on the same yeast two-hybrid screen as described previously for the identification of the signal-transducing receptor S1PR2 (Kempf et al., 2014). Recent studies have demonstrated that a classic yeast two-hybrid approach is well suited to screen for TSPAN protein interactions. Thus, tetraspanins have been used previously as prey in yeast two-hybrid assays (Bhave et al., 2013; Chen and Enns, 2015; Liu et al., 2009), as well as a bait (Tiwari-Woodruff et al., 2001). In our screen, the number of different baits per prey was divided by the total number of times that prey was isolated in our screens. This normalization step accounts for the abundance of the prey in the cDNA libraries screened (for details on our method, please see Koegl and Uetz, 2008; Mohr and Koegl, 2012). All our hits were observed multiple times under high stringency conditions. It has also been observed that libraries that express protein fragments have a greater chance of containing prey fragments that lack ‘difficult’ protein regions, such as transmembrane domains. Consequently, they might detect interactions that are lost when using arrays of full-length proteins. Importantly, the Nogo-A-Δ20–TSPAN3 interaction that was observed by using a yeast two-hybrid assay has been verified by us with several independent methods.

NIH 3T3 Swiss and NIH fibroblasts (American Type Culture Collection) were cultivated in Dulbecco's modified Eagle's medium (DMEM; Invitrogen, 61965-026) and supplemented with 10% newborn calf serum (Invitrogen). Embryonic day (E)19 rat cortical neurons were prepared as described previously (Lee et al., 2008) and cultivated in Neurobasal medium (Invitrogen, 21103-049) containing B27 supplement, 25 mM glucose, 1 mM glutamine, 50 U/ml penicillin and 50 μg/ml streptomycin. HEK293T cells (American Type Culture Collection) were maintained in DMEM containing 10% fetal calf serum (Invitrogen, 21885-025).

Co-immunoprecipitation was performed with lysates of NIH 3T3 fibroblasts, incubated with HA-tagged Nogo-A-Δ20. Briefly, cells were lysed with a buffer containing a mild detergent (1% CHAPSO), which maintains tetraspanin second-level interactions (Berditchevski and Rubinstein, 2013; Wakabayashi et al., 2009) [50 mM Tris-HCl pH 7.2, 150 mM NaCl, 10 mM MgCl₂, 1% CHAPSO and Complete Mini EDTA-free protease inhibitor cocktail tablets (Roche)]. Cell debris was removed by centrifuging, and the solubilized membranes in the supernatant were incubated with mouse anti-TSPAN3 antibodies (Ptglab 60049-1-Ig; 2 μg/500 μl protein lysate) or isotype control mouse IgG₂b antibodies (Life Technologies, MG2B00) at 4°C overnight. The next day, a pre-clearing step was performed – 50 μl of pre-blocked protein G beads (Pierce; in 1% BSA in lysis buffer) was added to the samples for 30 min on a rocking wheel at 4°C. After centrifugation, the samples were incubated with protein G beads (50 μl/500 μl lysate) for 2 h on a rocking wheel at 4°C. Beads were washed three times, and bound proteins were eluted and prepared for SDS-PAGE analysis and transferred onto a PVDF membrane. Overnight incubation at 4°C was performed with either rat anti-HA (Roche, 11867423001, 0.5 μg/ml) or human IgG F(abʹ)₂ fragment anti-S1PR2 (AbD Serotec custom-made HuCAL antibody AbD14533.1 targeting extracellular S1PR2 ECL2, 1.86 μg/ml). The following horseradish peroxidase (HRP)-conjugated secondary antibodies were used: goat anti-rat (Thermo Scientific, PA1-28779, 0.1 μg/ml) or goat anti-human IgG Fab-specific (AbD Serotec, PA1-28779, 1 μg/ml).

Co-immunoprecipitation with the tagged S1PR2, TSPAN3 or TSPAN7 protein was performed with HEK293T cells. Cells that had been cultivated in 10-cm dishes were transiently transfected with Lipofectamine 3000 transfection reagent (Life Technologies, L3000-008) and 2.5 µg TrueORF cDNA plasmids (TSPAN3-MycDDK and TSPAN7-MycDDK, OriGene MR203238 and OriGene MR203045, respectively) and/or S1PR2-GFP plasmid (cloned by Michael Arzt, Brain Research Institute Zurich). At 48 h after-transfection, cells were treated with 500 nM Nogo-A-Δ20 for 10 min at 37°C and used for co-immunoprecipitation. Briefly, cells were lysed in 1% NP40 buffer (20 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 1% NP-40) containing Complete Mini EDTA-free protease inhibitor cocktail tablets (Roche) and then centrifuged to clarify the lysate. The supernatants were pre-cleared by incubating with 50 µl of protein A/G agarose beads (Pierce, 20421; in 1% BSA in lysis buffer) for 30 min on a rocking wheel at 4°C. After centrifugation, the samples were incubated with mouse anti-DDK antibodies (OriGene, TA50011; 2.5 μg/500 μl protein lysate) or isotype control mouse IgG antibodies (Thermo Scientific, 31903) at 4°C overnight. The next day, protein-A/G–agarose beads were added to the samples (50 μl/500 μl lysate) for 2 h on a rocking wheel at 4°C. Immune-complex-captured beads were washed three times with lysis buffer and eluted in non-reducing SDS-PAGE sample buffer. The proteins were transferred onto a PVDF membrane. Overnight incubation at 4°C was performed with a chicken anti-GFP antibody (Abcam, ab13970, 10 mg/ml). The rabbit anti-chicken HRP-conjugate (Abcam, ab97140, 1 mg/ml) was used as a secondary antibody. Western blot analysis was performed with SuperSignal West Femto Chemiluminescence Detection kit (Pierce).

The NIH 3T3 fibroblast spreading assay was performed as described previously (Kempf et al., 2014; Oertle et al., 2003). Four-well plates (Greiner) were coated with the indicated Nogo-A-Δ20 concentrations and incubated at 4°C overnight. Transient knockdown was performed using TSPAN3 ON-TARGET plus siRNA SMARTpool Dharmacon (Thermo Scientific). A total of 3×10⁴ Swiss NIH 3T3 were transfected with 50 nM of siRNA using the DharmaFECT 3 transfection reagent in serum-free DMEM (Gibco) 24 h after plating. Transfection efficiency was determined by transfecting a fluorescently tagged non-targeting siGLO-red RISC-free siRNA. Cells that had been transfected with the siCONTROL non-targeting siRNA pool were used as negative control. Quantification of the knockdown of the respective mRNAs was performed by using qRT-PCR. Silencing of TSPAN3 by using lentiviral transduction of shRNA constructs was achieved with a Mission EGFP_shRNA clone against TSPAN3 (Sigma-Aldrich). EGFP_shRNA non-targeting control particles were used as negative control. Lentiviral shRNA transduction of NIH 3T3 fibroblasts was performed according to the manufacturer's instructions and in the presence of polybrene (8 μg/μl, Santa Cruz). At post-transduction day 4, positive clones were selected in geneticin-containing medium (Invitrogen, DMEM 61965-026; Life Technologies, Geneticin G418, 300 μg/ml). Quantification of the knockdown at the protein level was achieved by western blotting. Each spreading experiment was performed at least three times in four replicate wells. Spreading experiments, including pharmacological blockage of S1PR2, were performed with pre-treatment of cells with 100 nM JTE-013 (Tocris Bioscience, preparation according to the manufacturer's instructions), as described previously (Kempf et al., 2014).

The neurite outgrowth assay with E19 rat cortical neurons was performed as described previously (Kempf et al., 2014; Lee et al., 2008). The transient knockdown was achieved as described above for the NIH 3T3 fibroblasts using TSPAN3 ON-TARGET plus siRNA SMARTpool Dharmacon (rat-specific; Thermo Scientific). In outgrowth experiments, the 4-well plates were pre-coated with 0.3 mg/ml poly-l-lysine (PLL; Sigma-Aldrich) for 1 h at 37°C, before coating with the different substrates. For the neurite outgrowth, the cells were cultured for 24 h at 37°C under 5% CO₂, then fixed with 4% paraformaldehyde (PFA) and stained for βIII tubulin. The mean neurite length refers to the mean total length of all neurites per cell (Kempf et al., 2014). Each experiment was performed at least three times in four replicate wells. Statistical analysis was performed in GraphPad Prism5 using a one-way ANOVA test followed by a Bonferroni post-hoc test or by using an unpaired Student's t-test.

At 72 h post-transfection, total mRNA was isolated from siRNA-transfected cells using the RNeasy Micro Kit (Qiagen) followed by an RT-PCR reaction. Tetraspanin-specific primer probes were purchased from Eurofins (TSPAN3 FW 5′-GTTGGCCTTTGCAGCTATTCAGC-3′, REV 5′-ACTCCGTCTGCACAACACGATG-3′; TSPAN2 FW 5′-TTGCGGGTCTCACGATCTTTGG-3′, REV 5′-TCGCGTGAGTTCCGTATTGCAC-3′; TSPAN6 FW 5′-GCACCTTTGGTTGTTTCGCTACC-3′, REV 5′-AGTGTCAGAAACATCGCGTACAG-3′; TSPAN7 FW 5′-TGGAATTGCGTTCTCCCAGTTG-3′, REV 3′-ACTGATTGGCAGTGATGAACCG-5′; CD9 FW 5′-AGTTACGGAGCAAGGATGAACCC-3′, REV 5′-TATGCCACAGCAGTCCAACG-3′; CD82 FW 5′-GCTGGTGTTGCTGTCATTGAGC-3′, REV 5′-AGGCTGAGGGTGGAGAAATAGGAC-3′). qRT-PCR was performed as described previously using the LightCycler 480 System (Roche). Silencing of target genes was determined by normalizing target gene expression to GAPDH expression.

NIH 3T3 cells and rat E19 cortical neurons [4 days in vitro (DIV)] were fixed with 4% PFA for 15 min, washed, permeabilized with 0.1% Triton X-100 and blocked with 2% goat serum or 2% fish skin gelatin. For primary antibody incubation, the following antibodies were used: anti-S1PR2 (AbD Serotec, 2.8 μg/ml), anti-TSPAN3 (Ptglab, 3.3 μg/ml), anti-HA (Roche, 0.4 μg/ml), rabbit anti-EEA1 (Cell Signaling, 3.3 μg/ml), rat anti-Lamp-1 (Becton Dickinson, 5 μg/ml). The following secondary antibodies were used: Alexa488-conjugated goat anti-mouse IgG (Invitrogen; 1 μg/ml), Cy3-conjugated goat anti-rabbit IgG (Invitrogen, 1 μg/ml), Alexa488-conjugated goat anti-rat IgG (Invitrogen, 1 μg/ml), Biotin SP-conjugated AffiniPure goat anti-human IgG F(abʹ)₂-fragment-specific (Jackson ImmunoResearch Laboratories, 2 μg/ml), Cy3-conjugated Streptavidin (Jackson ImmunoResearch Laboratories, 2 μg/ml), Cy5-conjugated donkey anti-mouse (Jackson Immuno, 1.3 μg/ml), Alexa488–phalloidin (Invitrogen, 1:300) or Alexa594–phalloidin (Invitrogen, 1:300). The mCherry-tagged Rab11 construct was a gift from the group of Kurt Ballmer-Hofer (Paul Scherrer Institute, Switzerland).

For cell surface immunocytochemical detection of TSPAN3 or Nogo-A-Δ20–HA, NIH 3T3 cells were incubated with anti-HA (1 μg/ml) and anti-TSPAN3 (3.3 μg/ml) antibodies in serum-free medium containing 0.02% sodium azide for 45 min on ice. Cells were washed with serum-free medium and fixed with 1.0% PFA. After blocking (with 2% fetal calf serum, 2% horse serum, 0.1% fish gelatin, 0.1% casein), cells were incubated with Alexa488-conjugated goat anti-rat IgG and Cy3-conjugated goat anti-mouse IgG or Cy5-conjugated donkey anti-mouse. Coverslips were mounted with mounting medium (Dako). Images were acquired on a microscope (DM RE; Leica) using a confocal scanning system (SP2, SP5; SP8 Leica) using 63×1.4 NA or 100× Plan-Chromat objectives. Images were processed to equivalent extents with the use of Photoshop (Adobe, CS5).

Quantification of Nogo-A-Δ20 cell surface binding after incubation with 500 nM Nogo-A-Δ20–HA in serum-free medium at 4°C was assessed in siRNA non-target control-treated or TSPAN3-siRNA-treated cells. Cell surface associated Nogo-A-Δ20 clusters (minimum cluster size 0.2 μm) were automatically quantified in three-dimensional reconstructions of imaged stacks (single confocal slices at 0.18–0.22 μm) of single cells with the surface module of Imaris 7.4.2 (Bitplane). TSPAN3 and Nogo-A-Δ20–HA colocalization after different Nogo-A-Δ20–HA incubation times [in minutes, 3, 7, 15 (10 min Nogo-A-Δ20–HA+5 min serum-free medium), 25 (10 min Nogo-A-Δ20–HA+15 min serum-free medium), 40 (10 min Nogo-A-Δ20–HA+30 min serum-free medium)] was analyzed in three-dimensional reconstructions of imaged stacks (single confocal slices at 0.18–0.22 μm) of single cells with the colocalization module of Imaris. Representative pictures showing top views of single cell maximum projections with all confocal slices either of single channels, both channels or xz and yz side views of both channels. Data analysis was performed by using GraphPad Prism5 performing statistical analysis using a one-way ANOVA test followed by a Bonferroni post-hoc test. Data are given as the mean value±s.e.m.

Recombinant protein Nogo-A-Δ20 (rat, amino acid residues 544–725) was purified as described previously (Joset et al., 2010; Kempf et al., 2014). Briefly, BL21/DE3 Escherichia coli were transformed with the pET28 a+expression vector (Novagen) containing His- or T7-, or His- or HA-tagged Nogo-A-Δ20 and cultured at 37°C to reach an OD₆₀₀ of 0.6–1.0 arbitrary units. Protein expression was induced by addition of 1 mM isopropyl β-d-1-thiogalactopyranoside for 2 h at 30°C. Fusion proteins were purified using HiTrap TALON crude column (GE Healthcare), and the batch-specific inhibitory concentration 50 (IC₅₀-value) was determined with a fibroblast spreading assay at increasing concentrations Nogo-A-Δ20.

Control-siRNA-transfected and TSPAN3-siRNA-transfected NIH 3T3 cells were plated for 2 h and then treated for 20 min with 1 μM Nogo-A-Δ20–T7. Cells were fixed, permeabilized, blocked and stained with an anti-RhoA-GTP antibody (NewEast Biosciences, 2.5 μg/ml), detected with an Alexa488-conjugated goat anti-mouse antibody (1 μg/ml) and co-stained with Alexa594–phalloidin. A total of 135 cells from three independent experiments were imaged per group, and automatized quantification of mean intensity RhoA-GTP levels per cell was performed with ImageJ [National Institutes of Health (NIH)].

PLA-mediated detection of the S1PR2–TSPAN3 interaction was performed as previously described (Söderberg et al., 2006) using in situ PLA Duolink Detection kit orange (Olink Bioscience), according to the manufacturer's instructions for Duolink labeling, blocking and detection protocol. PLA Probe anti-mouse PLUS reagent (Olink) was used to directly label the primary mouse anti-TSPAN3 antibody, and PLA Probe anti-goat MINUS reagent (Olink) was used to label the secondary goat anti-human F(abʹ)₂ antibody. Briefly, NIH 3T3 fibroblasts were incubated with 500 nM Nogo-A-Δ20–T7 for different incubation times (0, 10, 25 min). Cells were fixed with 2% PFA for 15 min, washed, permeabilized with 0.1% Triton X-100 and blocked with 2% goat serum or 2% fish skin gelatin for 30 min, and additionally with Duolink blocking solution for a further 30 min at 37°C. Subsequently, cells were incubated with primary antibodies [PLUS-labeled mouse anti-TSPAN3, 4 μg/ml and human F(abʹ)₂ fragment anti-S1PR2, 4 μg/ml] in PLA probe diluent for 1 h at 37°C. Samples were washed according to the manufacturer's instructions, and incubation with the secondary goat anti-human F(abʹ)₂ antibody was performed for 1 h at 37°C. The following washing steps, hybridization of connecter oligonucleotides, enzymatic ligation of DNA strands through a subsequent addition of two other circle-forming DNA oligonucleotides, rolling circle signal amplification and mounting were performed as indicated in the manufacturer's instructions. Cells were co-stained with Alexa488–phalloidin (Invitrogen, 1:300). A total of 75 cells from three independent experiments were imaged per group by using an automatized method. Control PLA experiments were performed with TSPAN3-siRNA-transfected cells (for the transfection method, see In vitro bioassays, above).

Anti-TSPAN3 antibodies were covalently labeled with Atto647N (Atto-Tec), according to the manufacturer's instructions and as described previously for nanobody labeling (Ries et al., 2012). Briefly, the anti-TSPAN3 antibody was dialyzed into 0.2 M NaHCO₃ pH 8.2 in a mini-dialysis unit (Thermo Scientific). The amine-reactive fluorophore [N-hydroxysuccinimidyl (NHS)-ester] Atto647N was dissolved in DMSO (10 mg/ml) and added to fivefold molar excess to the antibodies. The labeling reaction was performed for 1 h at room temperature, and excess dye was removed using buffer exchange into 1× PBS with a Zeba Spin desalting column (Thermo Scientific). NIH 3T3 fibroblasts were treated with 500 nM Nogo-A-Δ20–T7 or equal volumes of 1× PBS for a maximum of 27 min. Cells were grown in DMEM without Phenol Red in order to reduce background fluorescence (Gibco, 11880-028). Live-cell single-molecule tracking was performed with a custom-built microscope, as described previously (Ries et al., 2012). The temperature was stabilized at 37°C, and image acquisition started at 2 min before (baseline) and during treatment in blocks at 5000 frames (25-ms exposure time per frame). Localization analysis was performed as described previously (Schoen et al., 2011). Coordinates of single molecules were determined for each frame and linked in consecutive frames to obtain trajectories. The mean square displacement (MSD) was plotted versus time, and the diffusion coefficient D determined by linearly fitting the points 2 to 4 (MSD=4 Dt). For mean D values, tracks in >32 frames were used.

Qdot imaging was conducted with Qdot^®-655 goat F(abʹ)₂ anti-mouse IgG conjugates (1 μM stock solution, Life Technologies), and Qdots were prepared in a similar manner to that described before (Saint-Michel et al., 2009). Briefly, Qdots were mixed with anti-TSPAN3 antibodies (1 mg/ml, Ptglab) in 1× PBS for 20 min at a 3:1 dilution, blocked with 10% casein solution and centrifuged at 5000 g for 3 min prior to use. Qdot–anti-TSPAN3 complexes (2 μg/ml in DMEM without Phenol Red, Gibco, 11880-028) were visualized using a wide-field LX Leica microscope and a Hamamatsu EM-CCD. The image acquisition rate was 440 ms per frame. Single-molecule tracking of TSPAN3 molecules was performed by using image series at time points between 3 and 11 min upon treatment with Nogo-A-Δ20 or control and by using ImageJ (NIH) with the ParticleTracker (MOSAIC group, ETH Zurich). Tracks of at least 50-frame lengths were measured, and their diffusion patterns were quantified as confined (presence within an area of 500-nM diameter for a minimum of 50 consecutive frames), Brownian (freely diffusing) or mixed (alternating between confined and Brownian movements). Interactions of two particles were defined when two tracks overlapped for at least 2 frames. Released diffusion behavior was defined for particles exhibiting, first, for at least 50 frames, confined and then mixed or Brownian diffusion. The diffusion coefficients D and the slopes of the moment scaling spectrum S_MSS of representative Qdot-labeled TSPAN3 tracks were calculated and plotted against each other, as described previously (Ewers et al., 2005; Sbalzarini and Koumoutsakos, 2005). Statistical analysis using a one-way ANOVA test followed by a Bonferroni post-hoc test was performed by using GraphPad Prism5. Data are given as mean values±s.e.m.

We thank Oliver Weinmann for technical assistance, Anissa Kempf and Anna-Sophia Wahl (Brain Research Institute, University of Zurich) for helpful comments; Michael Arzt (Brain Research Institute, University of Zurich) for providing the S1PR2-GFP construct and sharing unpublished data results with us, Kurt Ballmer-Hofer and Philipp Berger (Paul Scherrer Institute, Villigen, Switzerland) for providing the Rab11-mCherry construct.