The Sphingolipid Receptor S1PR2 Is a Receptor for Nogo-A Repressing Synaptic Plasticity

This study identifies a GPCR, S1PR2, as a receptor for the Nogo-A-Δ20 domain of the membrane protein Nogo-A, which inhibits neuronal growth and synaptic plasticity.


Introduction
Factors inhibiting nerve fiber growth substantially contribute to the limited regenerative capacity of the adult central nervous system (CNS) after injury. They play important roles in stabilizing the complex wiring of the adult CNS of higher vertebrates and in establishing neuronal pathways in the developing nervous system [1,2]. One of the best-studied factors is the membrane protein Nogo-A, which occurs in myelin and certain neurons, inhibiting axonal regeneration and plasticity after CNS injury [3][4][5]. Neutralization of Nogo-A has been shown to enhance axonal growth and compensatory sprouting in the adult spinal cord and brain, as well as to improve functional recovery after CNS injury [4,6]. Recent studies have shown novel important roles of Nogo-A signaling in the repression of synaptic plasticity in mature neuronal networks, indicating an inhibitory potential of Nogo-A far beyond its well-studied restriction of axonal growth [1,[7][8][9][10][11].
Nogo-A exerts its inhibitory effects via two distinct extracellular domains: Nogo-66 (rat amino acid (aa) 1026-1091) and Nogo-A-D20 (rat aa544-725; part of ''Amino-Nogo'') [2,12]. Nogo-66 induces growth inhibition via two membrane proteins, Nogo-66 receptor 1 (NgR1) [13], together with accessory proteins, and paired immunoglobulin-like receptor B (PirB) [14]. By contrast, the molecular identification and characterization of the receptor(s) transducing signals from the inhibitory Nogo-A-D20 domain has failed so far [2]. Nogo-A-D20 has been shown to partially mediate its inhibitory activity by interfering with integrins, but proof of a direct interaction has remained elusive [15]. Here we identified the G protein-coupled receptor (GPCR) sphingosine 1-phosphate receptor 2 (S1PR2) as a functional receptor for the D20 domain of Nogo-A. S1PR2 belongs to the subfamily of five S1PRs [16]. S1PRs are known to be activated by the low molecular weight (MW) lipid ligand sphingosine 1-phosphate (S1P), which exerts diverse receptor-specific effects on various cell types, including regulation of apoptosis, cell motility and cytoskeleton dynamics [16]. In the brain and spinal cord, S1P has been shown to regulate angiogenesis and neurite outgrowth: activation of S1PR1 promotes neurite outgrowth in vitro via G i/o and Rac1, whereas activation of S1PR2 leads to neurite retraction, involving G i/o , G q , or G 12/13 and the RhoA pathway [16][17][18].
In this study we demonstrate that Nogo-A-D20 binds S1PR2 via extracellular receptor loops 2 and 3, which are distinct from the previously described binding site of S1P [19]. Nogo-A-D20 signals through the G protein G 13 , leukemia-associated Rho guanine exchange factor (RhoGEF) LARG and RhoA. Deleting or blocking S1PR2 counteracts Nogo-A-D20-and myelin-mediated inhibition of neurite outgrowth and cell spreading. Acute S1PR2 blockade increases hippocampal and cortical long-term synaptic plasticity similarly to Nogo-A neutralization. These results strengthen the recently proposed physiological role of Nogo-A in restricting synaptic plasticity to stabilize neuronal circuits [1,9]. Further, these data support the paradigm shift for GPCR signaling from the classical ''one ligand -one receptor'' situation towards more dynamic models [20,21].

Nogo-A Binds to S1PR2
The GPCR S1PR2 was identified as a novel receptor candidate of the Nogo-A-D20 domain using a yeast two-hybrid (Y2H) screen of custom-made adult and fetal human brain libraries. In the adult CNS, S1PR2 is mainly expressed in the grey matter ( Figure 1). Hippocampal pyramidal cells, cerebellar Purkinje cells, cortical neurons and spinal motoneurons, as well as retinal ganglion cells are S1PR2-positive ( Figure 1A-1K). Importantly, S1PR2 is also expressed in Nogo-A-D20-responsive cells in vitro including 3T3 fibroblasts and immature cerebellar granule neurons ( Figure S1). To validate the interaction of Nogo-A-D20 ( Figure 2A) and S1PR2, His-tagged Nogo-A-D20 was co-incubated with membranes of S1PR2-overexpressing cells and subsequently immunoprecipitated ( Figure 2C). S1PR2 was specifically detected in immunoprecipitation fractions ( Figure 2C). Vice versa, His-tagged Nogo-A-D20 could be specifically probed in S1PR2 immunoprecipitated fractions, suggesting that the two proteins interact in vitro ( Figure 2D). Co-immunoprecipitation experiments of Nogo-A or S1PR2 from whole mouse brain protein extracts further demonstrated that endogenous S1PR2 interacts with Nogo-A under physiological conditions in vivo ( Figure 2B). To determine the binding affinity, binding of the entire D20-containing extracellular N-terminal domain of Nogo-A (Nogo-A-ext; Figure 2A) to biosensor-immobilized membrane preparations expressing functional full length S1PR2 protein or non S1PR2-expressing control membranes was monitored in real-time using Bio-Layer interferometry (OctetRED). Non-linear fitting revealed that Nogo-A-ext binds to S1PR2 with an apparent equilibrium binding constant (K D ) of ,142 nM ( Figure 2E). The binding affinity was not influenced by the addition of S1P versus vehicle control (MeOH) (K D MeOH ,192 nM; K D S1P ,202 nM; Figure 2F). For a mapping of binding sites, individual extracellular domains (N-terminus and extracellular loops [ECLs]) of S1PR2 were synthesized as peptides and analyzed for binding to Nogo-A-D20 by microscale thermophoresis ( Figure 2G). Nogo-A-D20 was found to bind primarily to ECL2 (K D ,280 nM) and 3 (K D ,350 nM), less strongly to ECL1 (K D ,2 mM) and negligibly to the N-terminus of S1PR2 (K D ,11 mM) ( Figure 2G). Importantly, binding analysis of the other bioactive domain of Nogo-A, Nogo-66, to S1PR2 extracellular domains revealed only unspecific binding in the high micromolar range (K D ECL1 ,46 mM; K D ECL2 ,7 mM; K D ECL3 ,67 mM) or complete absence of binding (N-terminus) ( Figure 2H). Collectively, these data show that Nogo-A-D20 but not Nogo-66 binds to specific extracellular domains of the GPCR S1PR2.

S1PR2 Is Internalized upon Nogo-A-D20 Binding
We have shown previously that Nogo-A-D20 is internalized into signaling endosomes upon binding, which results in RhoA activation and growth cone collapse [22]. To investigate whether S1PR2 is co-internalized upon Nogo-A-D20 treatment, cell surface S1PR2 expression was analyzed by immunofluorescence using a custom-made antibody (Figures 3A, S2B, and S2C). Cell surface S1PR2 levels were reduced by ,64% (p,0.001) 30 min after addition of Nogo-A-D20 ( Figure 3B). To confirm this, plasma membranes of 3T3 cells were prepared 15 and 30 min postincubation with Nogo-A-D20 and analyzed for S1PR2 levels by immunoblotting ( Figures 3C and S2A). We found that cell surface S1PR2 levels were reduced by ,77% (p,0.01) and ,70% (p, 0.001) after 15 and 30 min incubation with Nogo-A-D20, respectively, indicating that S1PR2 is internalized upon binding to Nogo-A-D20 ( Figure 3C). Pulse-chase experiments revealed that the majority of internalized Nogo-A-D20 puncta colocalize with S1PR2 as well as with the endosomal marker EEA1 at 15 and 30 min post-incubation with Nogo-A-D20 ( Figure 3D). Ubiquitination of GPCRs is a critical post-translational modification, which is often dispensable for initial receptor endocytosis but important for endosomal trafficking to proteasome/lysosomal degradation pathways [23,24]. S1P has been shown to cause S1PR1 monoubiquitination and, in higher concentrations, polyubiquitination, resulting in subsequent GPCR recycling to the membrane or complete degradation, respectively [25].

Author Summary
Recent studies have demonstrated an important role of Nogo-A signaling in the repression of structural and synaptic plasticity in mature neuronal networks of the central nervous system. These insights extended our understanding of Nogo-A's inhibitory function far beyond its well-studied role as axonal-growth inhibitor. Repression is mediated via two different Nogo-A extracellular domains: Nogo-66 and Nogo-A-D20. Here, we identify the G-protein coupled receptor S1PR2 as a high-affinity receptor for Nogo-A-D20 and demonstrate that S1PR2 binds this domain with sites different from the recently proposed S1P binding pocket. Interfering with S1PR2 activity, either pharmacologically or genetically, prevented Nogo-A-D20-mediated inhibitory effects. Similar results were obtained when we blocked G 13 , LARG, and RhoA, components of the downstream signaling pathway. These findings revealed a strong increase in hippocampal and cortical synaptic plasticity when acutely interfering with Nogo-A/S1PR2 signaling, similar to previous results obtained by blocking Nogo-A. We thus provide a novel biological concept of multi-ligand GPCR signaling in which this sphingolipid-activated GPCR is also bound and activated by the high molecular weight membrane protein Nogo-A. PLOS Biology | www.plosbiology.org S1PR2-ubiquitin conjugates were not detected upon internalization of Nogo-A-D20 as opposed to S1P ( Figure 3E), indicating that Nogo-A-D20 signaling is not permanently terminated in the lysosomal degradation pathway [23][24][25]. These results suggest that S1PR2 is rapidly co-internalized with Nogo-A-D20 into early endosomes upon binding, which is known to be a key step for Nogo-A-D20-mediated growth inhibition [22].
Nogo-A-D20 Signals through G 13 , LARG, and RhoA The G proteins G q , G 12 , and G 13 were shown to interact with S1PR2 and to activate the small GTPase RhoA [16,29]. To determine whether G q , G 12 , or G 13 are implicated in Nogo-A-D20-mediated cell spreading inhibition, we transfected small interfering RNAs (siRNAs) targeting the mRNAs of the G proteins ( Figure S4C and S4D). Downregulation of G 13 but not of G q or G 12 fully rescued cell spreading from ,63% to ,134% on Nogo-A-D20 when compared to the siRNA control (p,0.01) ( Figure 5A). No cumulative effect was observed by co-application of JTE-013, suggesting that G 13 is a key regulator of Nogo-A-D20-mediated effects downstream of S1PR2 ( Figure 5A). Accordingly, inhibition of the Rac1-coupled G i/o protein [16] with Pertussis toxin (PTX) did not have any effect on Nogo-A-D20-mediated cell spreading inhibition ( Figure 5A). To assess whether G 13 is also involved in Nogo-A-D20-mediated inhibition of neurite outgrowth, G 13 was silenced in E19 rat cortical neurons using specific siRNAs ( Figure  S4E and S4F). Knockdown of G 13 but not of G 12 specifically rescued outgrowth from ,68% to ,87% on Nogo-A-D20 when compared to the siRNA control (p,0.05) ( Figure 5B). Taken together, these results demonstrate that G 13 is required for Nogo-A-D20-mediated inhibition of cell spreading and neurite outgrowth in vitro. S1PR2 has been shown to couple via G 12/13 to the RhoGEF LARG to mediate various RhoA-dependent cellular effects [30]. siRNA-mediated downregulation of LARG fully rescued cell spreading from ,63% to ,103% on Nogo-A-D20 when compared to the siRNA control (p,0.01) ( Figures 5A, S4C, and S4G). This is in line with LARG-mediated activation of RhoA reported for other repulsive cues such as S1P (via S1PR2 [30]), semaphorin4D (via PlexinB1 [31]), and repulsive guidance molecule RGMa (via Unc5b [32]).

Nogo-A-D20-Mediated Inhibition Is Modulated by Exogenous S1P
To determine possible functional interactions of Nogo-A-D20 and S1P at the level of S1PR2, we first investigated whether Nogo-A-D20 itself modulates S1P production. Extra-(EC) and intracellular (IC) S1P levels were quantified in 3T3 and cerebellar granule neuron cultures after a 30 and 60 min stimulation with Nogo-A-D20 ( Figure 5G and 5H). No significant changes compared to control levels were detected, indicating that Nogo-A-D20 had no influence on S1P production under our experimental conditions ( Figure 5G and 5H).
We then addressed the role of endogenous S1P in Nogo-A-D20mediated inhibitory effects. Pharmacological blockade of the S1Pproducing enzymes sphingosine kinase (SphK) 1 and 2 using D,Lthreo-dihydrosphingosine (DHS) [34,35] had no effect on Nogo-A-D20-mediated inhibition of cell spreading, suggesting that SphKs are not downstream elements of Nogo-A-D20-induced inhibition ( Figure 5I). To confirm this result, MEFs isolated from SphK1 2/2 or SphK2 2/2 mice [36] were plated on a Nogo-A-D20 substrate. Similarly to SphK blockade, no differences in cell spreading inhibition were observed ( Figure 5I). ) coimmunoprecipitated with S1PR2 (,40 kDa) and vice-versa in WT but not Nogo-A 2/2 or S1PR2 2/2 brain extracts (BE). If specified, the following controls were used in WT BE instead of the IP antibody to confirm the specificity of the interaction: IgG, control antibody; Ctrl R, resin only control; qAbR, quenched antibody (Ab) resin control. Input loading control: b-Actin (,42 kDa). (C) S1PR2 immunoprecipitated with His-tagged D20 but not heat-inactivated (hi) D20 in S1PR2-overexpressing membranes. Input loading control: S1PR2. (D) His-tagged D20 but not hi D20 immunoprecipitated with S1PR2 in S1PR2-overexpressing membranes. Input loading control: S1PR2. (E) Nogo-A-ext bound specifically to biosensor-immobilized S1PR2overexpressing versus control membranes (K D ,142 nM). A Scatchard plot analysis is shown on the right. (F) 1 mM S1P does not modulate the interaction between Nogo-A-ext and S1PR2 when compared to the methanol (MeOH) vehicle control (MeOH, K D ,192 nM; S1P, K D ,202 nM). A Scatchard plot analysis is shown on the right. (G) Microscale thermophoresis binding analysis of D20 to S1PR2 extracellular domains: ECL2 (K D ,280 nM), ECL3 (K D ,350 nM), ECL1 (K D ,1.7 mM), and N-terminus (K D ,11 mM). Scrambled ECL1 (ECL1-scr) was used as control (K D ,17 mM). Arrows indicate the identified D20-binding loops in S1PR2. (H) Nogo-66 binding to S1PR2 extracellular domains is unspecific: Because S1P is found in fetal bovine serum (FBS)-containing medium [37] used in our experimental conditions, we investigated if serum-derived S1P modulates Nogo-A-D20-mediated inhibition. For this purpose, extracellular S1P was scavenged using the monoclonal anti-S1P antibody Sphingomab [38]. Cell spreading analysis revealed that Nogo-A-D20-induced inhibition was alleviated by ,28% (p,0.05) in the presence of the anti-S1P antibody when compared to the anti-BrdU control ( Figure 5J). To exclude that disinhibition of Nogo-A-D20 signaling by blocking or silencing S1PR2 is mediated by an increased activation of Rac1-coupled S1PR1 through serum-derived S1P, anti-S1P was applied together with JTE-013. No differences could be observed between anti-S1Pand anti-BrdU-treated cells in the presence of JTE-013 ( Figure 5J). Together, these results suggest that S1PR2-mediated inhibition by Nogo-A-D20 occurs independently of S1P but that S1P can modulate Nogo-A-D20-mediated effects. Indeed, addition of S1P to cells resulted in an ,31% (p,0.001) and ,28% (p,0.001) decrease in cell spreading inhibition on a control and Nogo-A-D20 substrate, respectively, when compared to the MeOH + DMSO control ( Figure 5K). These results point to a modulatory function of S1P in Nogo-A-D20-mediated inhibition of cell spreading, presumably by independently activating RhoA-coupled cell surface S1PRs, e.g., S1PR2. Concordantly, S1P has been previously described to modulate cell adhesion and growth of different cell types [18,27,39]. To test this hypothesis, JTE-013 was co-applied with S1P. S1Pinduced inhibition of cell spreading could be significantly reversed on a control and Nogo-A-D20 substrate in the presence of JTE-013 (p, 0.001) ( Figure 5K). Together, these results indicate that S1P can modulate Nogo-A-D20-mediated cell spreading inhibition via S1PR2. However, they also suggest that Nogo-A-D20 acts independently of SphK or S1P.

Nogo-A Restricts Long-Term Potentiation via S1PR2 in the Hippocampus and Motor Cortex
Growing evidence suggests that Nogo-A plays an important role in restricting synaptic plasticity [6,9,11]. S1PR2 is expressed in Figure 3. S1PR2 is internalized upon Nogo-A-D20 binding. (A) Representative confocal micrographs of 3T3 cells stained alive (Non-perm) or fixed (Perm) for S1PR2 before (control) and 30 min after D20 treatment at 37uC. (B) Mean fluorescence intensity quantification of the cell surface staining shown in (A). (C) Addition of D20 downregulates cell surface S1PR2 in 3T3 plasma membranes (PM): immunoblot and relative quantification thereof. Loading control: b-Actin. (D) Representative confocal micrographs of 3T3 cells incubated with 1 mM HA-tagged D20 for 1 h at 4uC (pulse), which were then subsequently chased for 15 and 30 min at 37uC. Cells were stained with an anti-HA (D20), S1PR2, or EEA1 antibody (early endosomes). Arrows indicate cell surface-bound D20 (top panel) or colocalization of D20 and S1PR2 in early endosomes (middle and bottom panel).   . Nogo-A-D20 inhibition is mediated via the G 13 -LARG-RhoA signaling axis and can be modulated by exogenous S1P. (A) 3T3 cells transfected with siRNAs against G 12 , G 13 , G q , or Larg, or control (ctrl) siRNA were replated on a Nogo-A-D20 substrate and assessed for cell spreading. G i/o was blocked with Pertussis Toxin (PTX) for which saline was used as control. JTE-013 was co-applied to G 13 -siRNA-treated cells to investigate a cumulative effect. (B) Transfection of DIV4 E19 cortical neurons with siRNA against G 13 but not G 12 similarly rescued Nogo-A-D20induced neurite outgrowth inhibition. (C,D) Nogo-A-D20-induced RhoA activation was assessed in JTE-013-versus DMSO-treated cells (C) or in cells carrying a stable knockdown of S1PR2 (sh-S1pr2) versus control vector (sh-Vec) (D). (E,F) Relative quantification of (C) and (D), respectively. (G,H) Competitive ELISA quantifications of extra-(EC) and intracellular (IC) S1P levels in 3T3 cells (G) and cerebellar granule neurons (H) before and after 30 and 60 min incubation with Nogo-A-D20. (I) Quantification of Nogo-A-D20-mediated cell spreading inhibition in the presence of the SphK-specific blocker D,L-threo-dihydrosphingosine (DHS) or in SphK1 2/2 or SphK2 2/2 MEFs. (J,K) 3T3 cells were plated on a Nogo-A-D20 substrate in the presence of the function blocking anti-S1P antibody Sphingomab (J) or of exogenous S1P (K) and assessed for cell spreading. Co-application of JTE-013 significantly reversed the modulatory effects obtained by S1P (K) but not anti-S1P (J). Anti-BrdU antibody or methanol was used as control in (J) and (K). Data shown are means 6 SEM (n = 3-6 experiments; *p,0.05, **p,0.01, ***p,0.001). doi:10.1371/journal.pbio.1001763.g005 S1PR2 Is a Receptor for Nogo-A PLOS Biology | www.plosbiology.org CA1 and CA3 pyramidal neurons ( Figure 1A and 1B). In order to investigate the role of the Nogo-A/S1PR2 axis in long-term potentiation (LTP), hippocampal slices of WT and Nogo-A 2/2 mice were tested for LTP after acute blockade of S1PR2 using JTE-013. In WT slices, application of JTE-013 resulted in a significant increase in LTP compared with vehicle (DMSO) (,22%; p,0.05) ( Figure 6A). In contrast, no differences in LTP were detected in Nogo-A 2/2 slices treated with JTE-013 or vehicle, suggesting that Nogo-A is required for S1PR2-mediated effects on LTP ( Figure 6B). No differences in input-output (I/O) curves and paired-pulse facilitation (PPF) could be observed by application of JTE-013, suggesting that S1PR2 blockade does not alter baseline synaptic transmission or the properties of presynaptic terminals ( Figure 6C-6F). In order to confirm the specificity of S1PR2, LTP was measured after blockade of the remaining S1PRs ( Figure S6A and S6B). No differences in LTP and PPF could be observed upon application of VPC-23019 or FTY-720, emphasizing the specificity of a functional Nogo-A/S1PR2 interaction ( Figure S6A-S6C). Next, we investigated LTP, baseline synaptic transmission as well as PPF in S1PR2 2/2 versus WT hippocampal slices. No significant changes in LTP, I/O, or PPF could be observed in S1PR2 2/2 versus WT mice ( Figure S6D-S6F) as opposed to acute neutralization of S1PR2. These results mirror those obtained in Nogo-A KO [7] or NgR1 KO [8] mice and suggest that there is a strong drive for genetic compensation in this functionally very important system. [11].
Next, the outcome of a combined neutralization of the ligand Nogo-A by the function-blocking anti-Nogo-A antibody 11c7 [12] and of the receptor S1PR2 by JTE-013 was analyzed. A synergistic effect of the combined treatment as compared to either treatment alone would indicate that additional molecules, e.g., S1P are involved in S1PR2-mediated LTP restriction. A similar increase in LTP for all treated groups when compared to the IgG1 + DMSO control with no difference between the groups was observed ( Figure 6G). To assess the relative contribution of the Nogo-A receptors NgR1 and S1PR2 onto Nogo-A-mediated restriction of synaptic plasticity, we simultaneously blocked both receptors. No significant difference could be observed between application of JTE-013 alone versus the combined application of JTE-013 and of the function-blocking anti-NgR1 antibody ( Figure 6H).
Finally, we investigated the effect of S1PR2 blockade on longterm depression (LTD) in the hippocampus. In line with the results obtained after acute Nogo-A neutralization [7], JTE-013 application did neither modulate LTD induction nor maintenance compared with control conditions ( Figure S6G).
Recent data indicate that Nogo-A also restricts synaptic plasticity in the primary motor cortex [11]. LTP saturation in this region was also significantly increased in JTE-013 versus DMSO-treated slices (,39%; p,0.001) ( Figure 6I). No differences in the I/O curves were observed after S1PR2 blockade, indicating that the JTE-013-mediated increase in synaptic plasticity was not due to alterations in baseline synaptic transmission ( Figure 6J). Together, these results show that Nogo-A represses synaptic plasticity in the hippocampus and motor cortex via S1PR2.

Discussion
Two distinct domains of Nogo-A can induce growth inhibition: Nogo-A-D20 and Nogo-66. Here, we identified the GPCR S1PR2 as the first functional receptor for the inhibitory D20 domain of Nogo-A. S1PR2 fulfills essential key criteria to be a Nogo-A-D20specific receptor: (i) Expression in the CNS as well as in nonneuronal Nogo-A-D20-responsive cells; (ii) high-affinity binding to Nogo-A-D20; (iii) prerequisite for Nogo-A-D20-induced inhibition of cell spreading and neurite outgrowth; (iv) Nogo-A-D20-induced activation of RhoA; (v) restriction of hippocampal and cortical synaptic plasticity.

S1PR2 Is a Receptor for a Lipid and a Protein Ligand
Until very recently, GPCRs were generally thought to be activated by physical and low MW chemical stimuli [40]. However, a few adhesion GPCRs were found to also bind to membrane-bound and matrix ligands via an extended N-terminal region [41,42]. Many of these receptors such as EGF-containing CD97, the first GPCR shown to bind to the cellular ligand decay accelerating factor, are predominantly expressed by immune cells [43]. To our knowledge, Nogo-A is the first mammalian membrane protein shown to bind to and signal through a nonorphan GPCR of the rhodopsin-like family. In contrast to adhesion GPCRs, S1PR2 does not bind Nogo-A-D20 via its Nterminal domain.
The recent characterization of the crystal structure of S1PR1 provided substantial structural information on its activation by S1P [19]. Access of the ligand to the binding pocket from the extracellular space is occluded by the N-terminus and the ECLs, and may be gained from within the membrane [19]. Our data provide strong evidence that Nogo-A-D20 primarily interacts with ECL2 and ECL3 of S1PR2, suggesting a different mechanism of activation compared to S1P. Our results also suggest that S1PR2mediated inhibition by Nogo-A-D20 does not require S1P but can be exogenously modulated by the latter. Although binding of Nogo-A-D20 to S1PR2 does not require S1P, modulation of receptor-specific physiological outputs by binding of the bioactive lipid to its pocket within the membrane may further expand the signaling repertoire of S1PR2. It may also enable fine-tuned cellular responses depending on the ratio of ligands present under given conditions, as recently suggested for the receptor for advanced glycation endproducts (RAGE) [44]. Future biochemical and structural studies will be necessary to address this and show how binding is transferred into ligand-specific G-protein-dependent signaling. Detailed investigations will also need to determine whether the presence of additional receptors, i.e. NgR1, affects the binding properties of Nogo-A-D20 to S1PR2 as described for other multi-receptor systems, e.g., the viral surface glycoprotein gp120 to CD4 and the GPCR co-receptor CCR5 [45]. We could show that Nogo-A interacts with S1PR2 in trans. However, interaction at the surface of the same cell in cis might also be possible, similar to what has been proposed for the Nogo-A-NgR1 interaction in Purkinje cells recently [46]. Yet, such mechanisms have not been proven and their existence needs to be investigated in detail.

Multi-ligand/Multi-receptor Cross-Talk
The classic ''one ligand-one receptor'' paradigm has recently been challenged by an increasing number of multi-ligand/multireceptor interactions, which could be identified in different biological systems, adding another level of complexity for finetuning of cellular responses [20]. Examples include neurotrophin receptors, Wnt receptors, and receptors for axonal guidance molecules such as plexins and neuropilins [20]. We propose that the D20 domain of Nogo-A binds to S1PR2 and the Nogo-66 loop to NgR1 and/or PirB, resulting in the formation of a multi-site/ multi-ligand receptor complex. NgR1 and PirB can also interact with ligands other than Nogo-A, thereby increasing the dynamics of signal transduction [6,9]. Additional Nogo-A co-receptors and downstream signaling components potentially located within or attached to these multi-receptor complexes might further amplify Nogo-A-mediated inhibitory effects. It was recently demonstrated that canonical GPCR signaling also occurs from endosomes for, e.g., the Wnt receptor Frizzled [47] and the b2-adrenoceptor [48]. Along this line, the Nogo-A-D20/S1PR2 complex is co-internalized into endosomes, from which signaling may be sustained. Currently, the concerted action and downstream trafficking of all these receptor components is still poorly understood, in particular in vivo. Future studies will need to assess whether all Nogo-A (co-)receptors are found within the same complex or in different membrane microdomains, and how the receptor composition varies between different cell types, developmental stages, and pathophysiological conditions.

Interfering with Nogo-A/S1PR2 Signaling Increases Synaptic Plasticity
Nogo-A stabilizes neuronal networks by restricting CNS plasticity [2,9]. Acute neutralization of Nogo-A or NgR1 in hippocampal slices was shown to induce an increase in LTP at CA3-CA1 synapses [7]. On the other hand, conventional knockouts of Nogo-A, PirB, or NgR1 do not show significant modulations in LTP, presumably due to compensatory mechanisms [7,8,10,11,49]. This is well in line with the lack of LTP modulation observed in S1PR2 2/2 mice. A novel transgenic rat model in which Nogo-A expression was silenced but not completely ablated by using a synthetic anti-Nogo-A microRNA leaving the genomic locus intact showed a significant increase in LTP in the hippocampus as well as in the motor cortex [11]. This underlines the strong drive for genetic compensation after complete ablation of components within this functionally very important system. Our present findings revealed an increase in hippocampal and cortical LTP when acutely interfering with S1PR2 signaling by JTE-013. Notably, no JTE-013-mediated increase in hippocampal LTP was observed in Nogo-A 2/2 mice, underlining the plasticity-restricting role of Nogo-A/S1PR2 signaling independently of S1P. Indeed, CA3-CA1 LTP was shown to be independent of SphK/S1P receptor signaling [50]. Interestingly, the blockade of both Nogo-A receptors NgR1 and S1PR2 does not show an additive effect on LTP potentiation, suggesting that both receptor-evoked responses induced by different domains of Nogo-A converge onto the same signaling pathways. However, detailed mechanisms and kinetics by which Nogo-A/S1PR2-NgR1 modify synaptic plasticity remain to be analyzed.

Conclusion
Our finding that the GPCR S1PR2 binds two structurally unrelated molecules, a low MW sphingolipid and the high MW membrane protein Nogo-A, by distinct sites contributes to and extends the paradigm shift from a classical linear model of GPCR signaling towards a more dynamic model with shared components and intramolecular cross talks [51,52]. It will be important to understand to which extent S1P affects signaling induced by Nogo-A and vice-versa. Detailed high-resolution structural characterization of the receptor in complex with S1P, Nogo-A, or both will be required to unravel the mechanistic properties of these two signaling systems. Furthermore, the cell-specific interplay of S1PR2 with known receptors and co-receptors for Nogo-A needs to be determined in detail with regard to their corresponding physiological effects. This information will be the basis for the design of novel molecular tools to better understand the roles of Nogo-A/S1PR2 signaling for CNS plasticity and repair.

Animals
All animal experiments were performed with the approval of and in strict accordance with the guidelines of the Zurich Cantonal Veterinary Office. All efforts were made to minimize animal suffering and to reduce the number of animals required. S1pr2 2/2 (B6.129S6-S1pr2 tm1Rlp ) mice were produced by targeted mutagenesis as described previously [28] and backcrossed to C57BL/6 background.

Tissue Preparation and Cell Culture
Total myelin protein extracts were prepared from the brains and spinal cords of adult Wistar rats as described previously [12]. Swiss 3T3 (ATCC), NIH 3T3 cells (ATCC), and HEK293T cells (ATCC) were maintained in DMEM containing 10% neonatal calf serum (Invitrogen). Postnatal (P5-8) cerebellar granule neurons were prepared as described previously [12]. Embryonic day (E) 19 rat cortical neurons were prepared as described previously [8]. Primary MEFs were isolated and immortalized as described previously [54]. Each primary fibroblast culture was isolated from a single E9.5 S1pr2 2/2 or WT littermate mouse.
Silencing of S1pr2 by retroviral transduction of shRNA constructs was done by using phoenix helper-free retrovirus producer lines with pSIR delta HRCG U6 for the generation of helper-free retroviruses as described below [55]. The following shRNA construct targeting S1pr2 mRNA transcript was used: ACCAAGGAGACGCTGGACATG [56]. Empty vector was used as control. Quantification of the respective mRNA knockdown was performed by qRT-PCR. Quantification of protein knockdown was performed by FACS analysis.
Recombinant protein Nogo-A-D20 (rat aa544-725) was purified as described previously [12]. Briefly, BL21/DE3 Escherichia coli were transformed with the pET28 expression vector (Novagen) containing His-/T7-or His-/HA-tagged Nogo-A-D20 and cultured at 37uC to reach an OD of 0.6 AU. Protein expression was induced by addition of 1 M IPTG for 2 h at 30uC. Fusion proteins were purified using Co 2+ -Talon Metal Affinity Resin (Takara Bio Inc.). Nogo-A-ext (rat aa1-979) was cloned into the KpnI and XhoI restriction sites of the pEXPR-IBA5 expression vector and the recombinant protein was purified from transiently transfected HEK293T cells using Strep-tactin chromatography (IBA).

Binding Assays
Immobilization-based binding assays were performed on an Octet Red Instrument (fortéBIO). Recombinant S1PR2 and control membrane preparations (Millipore) were immobilized on amine-reactive biosensors (25 mg/ml; fortéO) in HBSN running buffer (BIAcore) supplemented with 10 mM MgCl 2 . Nogo-A-ext protein was serially diluted and allowed to bind the saturated biosensor tips for 15 min at 1,000 rpm at 30uC. For experiments including S1P, 1 mM S1P was added together with Nogo-A-ext. Methanol was used as vehicle control. The binding response was normalized for baselines differences between runs and binding affinities (K D ) were calculated from a nonlinear fit according to the double-reference subtraction method in GraphPad Prism5 (GraphPad software). Data shown are the average of three to five experiments per condition.
Immunoprecipitation was performed with Nogo-A-D20 and S1PR2 membrane preparations using the His Protein Interaction Pull-Down kit following the manufacturer's instructions (Pierce). Heat-inactivated Nogo-A-D20 was used as control.
In Vitro Bioassays 3T3 fibroblast spreading assays and P5-8 cerebellar granule neurons neurite outgrowth assays were performed as described previously [12]. Briefly, four-well plates (Greiner) were coated with 40 pmol/cm 2 Nogo-A-D20 or 5 mg/cm 2 myelin at 4uC overnight. Nogo-66 Fc was used at a concentration of 500 nM and Aggrecan at 1,000 ng/ml. In outgrowth experiments, wells were precoated with 0.3 mg/ml for 1 h at 37uC before addition of the different substrates. 3T3 cells were plated at 7,000 cells per cm 2 for 1 h at 37uC and 5% CO 2 , fixed with 4% paraformaldehyde (PFA) and stained with Phalloidin-Alexa-488. Mouse P5-8 cerebellar granule neurons were plated at 7.5610 4 cells per cm 2 , cultured for 24 h at 37uC and 5% CO 2 , fixed with 4% PFA and stained with anti-bIII tubulin. Each experiment was performed at least three times in four replicate wells. Spreading was quantified manually in a blinded manner and mean neurite length was quantified using the MetaMorph software (Molecular Devices). The mean neurite length is referred to as the mean total length of all neurites per cell. 3T3 cells were classified as spread cells if they bear at least two lammelipodial processes longer than one cell body diameter. Round cells were classified as non-spread. Data were normalized to baseline and plotted as average 6 standard error of the mean (SEM). Cells were imaged with a Leica DM5500B microscope equipped with HCX PL FL Dry 106/0.3 and 206/0.5 objectives in a semi-automated way. 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. All inhibitors were used at a concentration of 100 nM if not elsewhere specified.

Internalization Assays and Flow Cytometry Analysis
Plasma membranes of 3T3 cells were prepared as described before [61] and after treatment with 1 mM T7-tagged Nogo-A-D20. Nogo-A-D20 internalization assays were performed as described previously after treatment of 3T3 cells with 1 mM HAtagged Nogo-A-D20 [22]. Briefly, 3T3 cells were incubated with 1 mM Nogo-A-D20 for 1 h on ice (pulse) and subsequently chased for 15 and 30 min at 37uC. Flow cytometry-based quantification of S1PR2, G13, and LARG expression on 3T3 cells and CGNs, respectively, was done in a BD FACSCalibur.

Ubiquitination Assay
3T3 cells were starved in serum-free medium for 24 h. 1 mM S1P or Nogo-A-D20, respectively, was added to 3T3 cells for 60 min. Isolation of ubiquitinated protein fractions was done using UbiCapture-Q (Enzo Life Sciences). Finally, western blot analysis was performed to detect S1PR2 and ubiquitin.

S1P Quantification
3T3 cells or CGNs were cultured up to 80%-85% confluence in 15 cm dishes and serum starved for 24 h prior to the experiment. Nogo-A-D20 was added to the cells at a concentration of 1 mM. After 15, 30, and 60 min, 3T3 cells and CGNs were lysed in 400 ml lysis buffer (20 mM PIPES, 150 mM NaCl, 1 mM EGTA, 1% v/v Triton X-100, 1.5 mM MgCl 2 , pH 7.4). Lysates were frozen immediately at 280uC. Protein concentration was measured and cell lysates (1:10 in delipidized human sera) were analyzed with the Echelon S1P ELISA kit according to the manufacturer's instruction. Serum free cell culture medium was directly diluted 1:10 in delipidized human serum and subsequently analyzed with the S1P ELISA kit.

Immunohistochemistry
Immunohistochemistry was performed as described previously [62]. Briefly, animals were transcardially perfused with Ringer's solution, followed by 4% PFA. Prior to staining, sections were treated with 0.2% glutaraldehyde and 50 mM Tris-glycine (pH 8.0). After antigen retrieval via microwaving three times for 10 s at 600 W, the sections were treated with Kryofix (Merck) for 10 min followed by 0.3% Triton X-100 for 10 min. S1PR2 was detected with AbD14533.1 and corresponding secondary antibodies.
3T3 cells and CGNs were fixed with 4% PFA for 15 min, washed, and permeabilized with 0.1% Triton X-100. After blocking with 2% goat serum, cells were first incubated with AbD14533.1 and detected using Cy3-conjugated Streptavidin.
Blockade of S1PR2 was achieved by incubation of acute slices with JTE-013, blockade of S1PR1, 3, 4, 5 with FTY-720 and blockade of S1PR1, 3 with VPC-23019, respectively. The inhibitors were dissolved in DMSO and freshly added at a final concentration of 5 mM, 1 mM, and 0.1 mM, respectively, to the carbogenated ACSF. The DMSO overall concentration in the ACSF was kept at 0.01%. As control DMSO alone was added. In order to compare the data with previous experiments silicon tubing was used, and pre-washed with ACSF containing BSA (0.1 mg/ml). The slices were pre-incubated for 1 h (or 10 min for the experiments in which JTE-013 and 11c7 were combined) with the inhibitor or DMSO as control in an incubation chamber maintaining a constant flow of the solution. During the experiments the inhibitor was also around. For the electrophysiological recordings, the perfusion rate in the recording chamber was constantly kept at 1.5 ml/min.
After placing the slices in a submerged recording chamber field, excitatory postsynaptic potentials (fEPSPs) were recorded in the stratum radiatum of the CA1 region with a glass micropipette (resistance 3-15 MV filled with 3 M NaCl at a depth of ,100 mm. Monopolar tungsten electrodes were used for stimulating the Schaffer collaterals at a frequency of 0.1 Hz. Stimulation was set to elicit a fEPSP with a slope of ,40%-50% of maximum for LTP recordings. After 20 min baseline stimulation LTP was induced by applying theta-burst stimulation (TBS), in which a burst consisted of four pulses at 100 Hz. These were repeated 10 times in 200 ms intervals (5 Hz). Three such trains were used to induce LTP at 0.1 Hz. Basic synaptic transmission and presynaptic properties were analyzed via I/O measurements and paired pulse facilitation. The I/O measurements were performed either by application of a defined value of current (25-250 mA in steps of 25 mA) or by adjusting the stimulus intensity to a certain current eliciting a fiber volley (FV) of desired voltage. Paired pulse facilitation was performed by applying a pair of two stimuli in different inter-stimulus-intervals (ISIs), ranging from 10, 20, 40, and 80 to 160 ms. Data were collected, stored, and analyzed with LABVIEW software (National Instruments). The initial slope of fEPSPs elicited by stimulation of the Schaffer collaterals was measured over time, normalized to baseline, and plotted as average 6 SEM.
Motor cortex. For LTP measurements in the motor cortex [11], coronal slices containing the forelimb are of M1 (1-2 mm anterior to the bregma [63]), and were prepared from adult Sprague Dawley rats (180-220 g). JTE-013 concentrations were used according to protocols used for hippocampal slices and added to the ACSF: 126 mM NaCl, 3 mM KCl, 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 1 mM MgSO 4 , 2 mM CaCl 2 , and 10 mM glucose, bubbled with a 95% O 2 , 5% CO 2 mixture at 3360.5uC). To allow optimal JTE-013 penetration, responses were recorded from the slice superface of layer II/III within M1. Basic synaptic transmission was analyzed with I/O analysis. I/O measurements were conducted by applying a value of current, which elicited a minimal (threshold) evoked response (0.2-0.3 mV). I/O curves were obtained by averaging field potential peak amplitudes of three responses to stimuli of two, three, four, and five times the threshold response. To elicit the maximum amplitude that could be evoked, we used a stimulation intensity of 256 threshold [64,65]. For baseline measurements, stimulus intensity was adjusted to produce responses 40%-50% of the maximum amplitude. For data analysis, we computed the amplitude of the field potential response because it serves as a measure of the population excitatory synaptic response [65], reflects a monosynaptic current sink [66], and correlates well with the intracellular excitatory postsynaptic response evoked in this pathway [67]. Measurement of the field potential slope, as routinely used, e.g., in the hippocampus, has not been used for neocortical field potential responses due to the interference of the response's initial part by variable nonsynaptic components [68]. After 20 min of baseline stimulation, focal and transient reduction of y-aminobuturic acid-A (GABA) inhibition at the recording site was produced by applying bicuculline methiodide (3.3 mM, Tocris Bioscience) from a micropipette by touching the tip to the slice surface within 100 mm of the recording microelectrode for 15-60 s. The pipette was removed when the amplitude of test responses increased 50%-100% of baseline [66]. Immediately after bicuculline application, LTP was attempted by delivering TBS at double baseline stimulation intensity. LTP induction was attempted by using TBS, which consisted of 10 trains of stimuli at 5 Hz. Each train was composed of four pulses at 100 Hz. This sequence was delivered every 10 s for a total of five presentations. LTP was recorded for at least 20 min after it reached a stable plateau. TBS was induced until responses were saturated. Pathways were considered saturated if the difference between two subsequent states of LTP were not significantly different [68]. Maximum LTP was calculated as a percentage of baseline, plotted as average 6 SEM and analyzed by a Student's t-test. Data were collected, stored, and analyzed with LABVIEW (National Instruments) and MATLAB (MathWorks) software.