Zinc metalloproteinase-mediated cleavage of the human Nogo-66 receptor

The regeneration of central nervous system (CNS) axons following injury is drastically restricted by the presence of inhibitory proteins within myelin and the glial scar of the injury site, thus compromising functional recovery (Yiu and He, 2003). These inhibitors are characterised by their ability as purified or recombinant preparations to inhibit neurite outgrowth and induce the collapse and retraction of nerve growth cones in vitro. Elucidated myelin-associated inhibitors are the membrane proteins Nogo-A (Chen et al., 2000; GrandPre et al., 2000; Prinjha et al., 2000), myelin-associated glycoprotein (MAG) (McKerracher et al., 1994) and oligodendrocyte-myelin glycoprotein (OMgp) (Wang et al., 2002a) and the chondroitin sulphate proteoglycans brevican and brain-specific versican V2 (Niederöst et al., 1999; Schmalfledt et al., 2000). Nogo-A has three inhibitory domains exposed on the surface of myelin oligodendrocytes: a C-terminal 66 residue ectodomain (Nogo-66) and two regions in the amino terminal ectodomain (GrandPre et al., 2000; Oertle et al., 2003). Although structurally dissimilar, Nogo-66, MAG and OMgp bind a mutual neuronal glycosyl-phosphatidylinositol (GPI)-anchored receptor, the Nogo-66 receptor (NgR, also referred to as NgR1) that transduces the inhibitory signal to the cell interior via a transmembrane co-receptor, p75^NTR (Fournier et al., 2001; Wang et al., 2002b; Wong et al., 2002). The receptor(s) for the amino-Nogo inhibitory domains, however, remain elusive. Nogo-66 binding to NgR critically requires an N-terminal domain (residues 27-310 for human NgR) comprising eight leucine-rich repeats (LRR) flanked by N- and C-terminal subdomains (LRRNT and LRRCT, respectively) (Fournier et al., 2002; Liu et al., 2002; Wang et al., 2002a; Barton et al., 2003), whereas binding to p75^NTR additionally requires the C-terminal domain (residues 311-446) (Wang et al., 2002b; He et al., 2003). Like many other GPI-anchored proteins, NgR resides within cholesterol-rich regions of cellular membranes known as lipid rafts which serve as platforms for a number of signal transduction pathways (Simons and Ikonen, 1997; Fournier et al., 2002; Pignot et al., 2003). The raft-association of NgR, however, is not essential for Nogo-66-induced growth cone collapse, although the efficiency of this inhibition is diminished when NgR resides outside rafts (Fournier et al., 2002).

Inhibition of NgR has been achieved using a peptide corresponding to residues 1-40 of Nogo-66 (NEP1-40) that acts as a competitive antagonist of the receptor in vitro and, more importantly, promotes axonal regeneration and functional recovery following CNS injury in vivo (GrandPre et al., 2002; Li and Strittmatter, 2003). In addition, several independent reports have shown that the soluble full length NgR ectodomain lacking the GPI anchor and an artificially C-terminally truncated NgR ectodomain comprising the ligand binding domain alone are potent antagonists of neurite outgrowth inhibition mediated by Nogo-66, MAG and whole myelin in vitro (Domeniconi et al., 2002; Fournier et al., 2002; Liu et al., 2002; Barton et al., 2003; He et al., 2003). However, at present, the cellular mechanisms involved in the regulation of neuronal NgR function at the level of the receptor have yet to be elucidated.

An emerging process for the modulation of membrane-anchored receptor function is their proteolytic cleavage to generate soluble extracellular fragments with the potential to perform agonistic or antagonistic roles distal to the cell surface, a process termed `ectodomain shedding' (Hooper et al., 1997; Peschon et al., 1998). Ectodomain shedding is generally mediated by members of the matrix metalloproteinase (MMP) and a disintegrin and metalloprotease (ADAM) families of zinc metalloproteinases and, as such, can be inhibited by hydroxamate-based compounds (Hooper et al., 1997; Schlondorff and Blobel, 1999). Apart from being upregulated by phorbol esters, ectodomain shedding has recently been shown to be modulated by cellular cholesterol. For example, cleavage of the amyloid precursor protein (APP) and the L1 adhesion molecule by ADAM10, and the interleukin-6 receptor by ADAM10 and ADAM17, can be up-regulated by cellular cholesterol depletion using agents such as methyl-β-cyclodextrin (MβCD) (Kojro et al., 2001; Mechtersheimer et al., 2001; Gutwein et al., 2003; Matthews et al., 2003).

In the present study, we demonstrate that human NgR expressed in human neuroblastoma cells is cleaved to generate a lipid raft-associated C-terminal fragment and a soluble N-terminal fragment containing the entire ligand-binding domain. In common with other shedding mechanisms, the release of the N-terminal fragment was blocked by a hydroxamate-based zinc metalloproteinase inhibitor and upregulated by MβCD treatment. That the N-terminal fragment retained the ability to bind Nogo-66 but failed to associate with the NgR co-receptor p75^NTR suggests that this fragment may antagonise Nogo-66 signaling. In addition, we show that fragments of NgR analogous to those generated in human neuroblastoma cells are detectable within human brain cortex and cerebrospinal fluid (CSF).

Anti-V5, anti-p75^NTR, anti-clathrin, anti-flotillin-1 and human NgR-specific polyclonal antibodies were obtained from Invitrogen, Santa Cruz Biotechnology, Cymbus Biotechnology, BD Transduction Laboratories and R&D Systems, respectively. Recombinant human tissue inhibitors of metalloproteinases were obtained from R&D Systems. To generate a human NgR-specific monoclonal antibody, BALB/C mice were immunised with the full length human NgR ectodomain and hybridomas generated using standard methods (Yokoyama, 1995). A clone expressing an anti-human NgR antibody was identified by ELISA and confirmed by western blot analysis. The epitope for this antibody was identified as residues 328-332 (EPLGL) of human NgR using the PepSPOT procedure according to the manufacturer's instructions (Jerini Peptide Technologies).

To generate NgR-310, the coding region for residues 27-310 of human NgR was amplified together with the V5 tag sequence using an N-terminally V5 tagged human NgR cDNA as template (Pignot et al., 2003). The PCR product was inserted into pSecTag2A (Invitrogen) via a modified `Quick Change' procedure (Geiser, 2001) using the following primers for insertion: Ngins1 (forward primer): 5′-CAG GTT CCA CTG GTG ACG CGG GTA AGC CTA TCC CTA ACC CTC-3′ and Ngins2 (reverse primer): 5′-GAT CCT CTT CTG AGA TGA GTT TTT GTT CAG CGC AGC CCT GCA GGT CAT-3′. HEK-293T cells were transfected with hNgR-310 using FuGENE 6 (Roche) according to the manufacturer's instructions and incubated for 72 hours with Opti-MEM (Invitrogen). To generate AP-Nogo-66, the coding region for rat Nogo-66 was PCR amplified and sub-cloned C-terminally to the coding region of AP into the XhoI/XbaI sites of a modified pAP-tag5 vector (GeneHunter). This modified vector (pNFAP) was generated by inserting a 6-histidine tag oligonucleotide sequence into the SfiI/HindIII site of pAP-tag5 in-frame to the Igκ chain signal peptide to create pNHIS. In a separate cloning, a linker sequence was inserted into the XhoI/XbaI site of pAP-tag5 to create pFAP (linker sequence: coding strand, 5′-TCG AAA TCG AGG GAA GGG AAT TCC GGC ACC GGC GGG CAT CTG TTC TCG AGG AGG AAG CGC TCT CT-3′; complementary strand, 5′-CTA GAG AGA GCG CTT CCT CCT CGA GAA CAG ATG CCC GCC GGT GCC GGA ATT CCC TTC CCT CGA TT-3′). This linker encoded a Factor Xa cleavage site and restored the original multiple cloning site. Finally, the HindIII/XhoI fragment from pNHIS was replaced by the HindIII/XhoI fragment from pFAP to obtain pNFAP. HEK-293T cells were transfected with either the pAP-tag5vector or AP-Nogo-66 as above and stably expressing transfectants selected with 200 μg/ml Zeocin (Invitrogen) followed by cloning by serial dilution. Cells were incubated for 48 hours with OptiMEM and the medium was concentrated through an Amicon Cell 8400 (Amicon). After equilibration to the loading buffer (50 mM NaH₂PO₄ pH 7.4, 300 mM NaCl, 20 mM imidazole) and filtration (0.4 μm), the sample was loaded on a HiTrap Chelating Column using Äktaprime (Amersham). Elution of AP or AP-Nogo-66 was performed using a linear gradient of 20-300 mM imidazole and the peak fractions dialysed against TBS/5% Glycerol.

SH-SY5Y cells stably expressing human NgR containing an N-terminal V5 epitope tag have been described (Pignot et al., 2003). Following incubation in OptiMEM, cells were washed twice with phosphate-buffered saline (PBS, Invitrogen), scraped into PBS and harvested by centrifugation at 100 g for 3 minutes. The cell pellet was resuspended in M-PER (Pierce) containing protease inhibitors (Roche) and incubated at 4°C for 30 minutes prior to clarification by centrifugation at 13,000 g for 5 minutes. Conditioned medium was clarified by centrifugation at 100 g for 3 minutes. Human brain cortex extracted 4 hours post-mortem from a normal adult (aged 50 years) was homogenised at 4°C with T-PER (Pierce) containing protease inhibitors according to the manufacturer's instructions. Human CSF was extracted with informed consent from a patient with degenerative lumbar spine syndrome (aged 69 years). Digestion with peptide N-glycosidase F (PNGase F, Glyko) was performed in 20 mM sodium phosphate pH 7.6, 50 mM EDTA, 5% (w/v) SDS, 5% (v/v) β-mercaptoethanol. Excised gel slices were disrupted in the above buffer using a polytron homogeniser (Kinematica). Samples were boiled for 5 minutes, diluted 5-fold with 1.25% (v/v) Triton X-100 and incubated at 37°C for 16 hours with 1 U PNGase F. For digestion with phosphatidylinositol-specific phospholipase C (PI-PLC, Glyko), cell lysates were incubated at 37°C for 3 hours with 1 U/ml PI-PLC. For treatment with MβCD and brefeldin A (BFA), cells were incubated for 30 minutes with OptiMEM containing 10 mM MβCD or for 6 hours with OptiMEM containing 5 μM BFA.

Cells at confluency were incubated at 4°C for 1 hour with PBS containing 1 mg/ml biotin 3-sulfo-N-hydroxysuccinimide ester (Sigma) and the reaction quenched by three washes with PBS containing 50 mM glycine. Following quenching, cells were either lysed or incubated at 37°C for various times in OptiMEM. Biotinylated proteins were precipitated from the cell lysate and medium by incubation at room temperature for 2 hours with 20 μg/ml streptavidin coupled to agarose beads (streptavidin-agarose, Sigma) followed by four washes with 10 mM Tris-HCl pH 7.8, 1% (w/v) N-lauroylsarcosine, 100 mM NaCl (TNS buffer). For immunoprecipitation, conditioned medium and cell lysate were pre-cleared by incubation at room temperature for 1 hour with 1% (v/v) protein A-Sepharose (Sigma) followed by incubation at room temperature for 2 hours with 10 μg/ml of either the anti-V5 or anti-p75^NTR antibody in the presence of 1% (v/v) protein A-Sepharose (Sigma) and the immunoprecipitates washed four times with TNS buffer.

Rafts were isolated as described (Walmsley et al., 2003). Briefly, cells at confluency in a 175-cm² flask were scraped into MES-buffered saline (MBS; 25 mM Mes, 150 mM NaCl, pH 6.5) and harvested by centrifugation. Cells were resuspended in 0.5 ml MBS containing 1% (v/v) Triton X-100 and homogenised with 20 strokes in a Dounce homogeniser. The homogenate was made 40% (w/v) with respect to sucrose by the addition of 0.5 ml 80% (w/v) sucrose in MBS and loaded at the bottom of a 5%/30% sucrose gradient. Gradients were centrifuged at 100,000 g for 18 hours in a SW50.1 rotor (Beckman) and subsequently separated into eight fractions, fraction 1 containing the highest density of sucrose.

SDS-gel electrophoresis was performed using NuPAGE 4-12% bistris gels (Invitrogen). Proteins were transferred to polyvinylidene fluoride (PVDF) membranes (Amersham) which were blocked by incubation for 1 hour with Tris-buffered saline/0.1% Tween-20 (TBST) containing 5% dried milk powder followed by an incubation for 1 hour in the same buffer containing either 200 ng/ml anti-V5 antibody, 4 μg/ml anti-p75^NTR antibody, 100 ng/ml anti-clathrin antibody, 250 ng/ml anti-flotillin-1, 2 μg/ml human NgR-specific polyclonal antibody or a 5-fold dilution of supernatant from hybridoma cells expressing the human NgR-specific monoclonal antibody. Membranes were washed in TBST and incubated with the appropriate peroxidase-conjugated secondary antibody. Signals were developed using ECL™ Western Blotting Detection Reagents (Amersham) and Hyperfilm™ ECL (Amersham) according to the manufacturer's instructions.

SH-SY5Y cells expressing NgR were incubated for 48 hours with OptiMEM and NTF-NgR immunoprecipitated from the medium using the anti-V5 antibody in amounts sufficient for the protein to be visualised by colloidal Coomassie Blue staining after SDS-PAGE (Neuhoff et al., 1988). The band corresponding to NTF-NgR was excised along with a control blank gel piece of similar size and digested in-gel with Lys-C (Achromabacter Protease I, WAKO) according to a protocol from the European Molecular Biology Laboratory (EMBL; http://www.embl-heidelberg.de). Prior to digestion, the protocol included the reduction of cystine residues to cysteine, which were then carbamido methylated. The peptides from the Lys-C digestion were extracted from the gel and purified by Ziptip (Millipore) according to the manufacturer's instructions. The samples were applied to a MALDI sample plate and mixed with a solution of α-cyano-4-hydroxy-cinnaminic acid. Matrix assisted laser desorption/ionisation (MALDI) was performed on an MALDI time-of-flight (TOF) mass spectrometer (Voyager-DE STR, Applied Biosystems) in positive reflectron mode with external calibration. The resulting mass spectra were searched for peaks present in the NTF-NgR band but not in the control. Tandem mass spectrometry (MS/MS) was performed using a MALDI TOF/TOF mass spectrometer (4700 Proteomics Analyzer, Applied Biosystems). Air was used in the collision cell for fragmentation.

1 nM AP or AP-Nogo-66 was pre-incubated at room temperature for 1 hour with medium from a 3-hour incubation with equivalent amounts of mock transfected SH-SY5Y or SH-SY5Y cells expressing human NgR and subsequently incubated at 10°C for 2 hours with CHO cells stably expressing human NgR (Pignot et al., 2003). Cells were washed five times with 20 mM HEPES pH 7.4 in Hanks' balanced saline (HH) and fixed for 15 minutes with 3.7% formaldehyde in 20 mM HEPES pH 7.4. Following inactivation of endogenous AP activity by incubation at 65°C for 1 hour in buffer HH, bound AP activity was quantified using 1-Step™ PNPP (Pierce) according to the manufacturer's instructions.

To investigate potential mechanisms for the regulation of NgR function in neuronal cells, we stably expressed human NgR containing an N-terminal V5 tag in the human neuroblastoma cell line SH-SY5Y. NgR was detected by the anti-V5 antibody as a diffuse band of 64 kDa in the cell lysate and a 48 kDa band in both the lysate and the incubation medium, with the latter band increasing in intensity in the medium over the time of the incubation (Fig. 1A). As human NgR has a predicted molecular weight of around 50 kDa and contains six potential N-glycosylation sites, the possibility existed that the 64 and 48 kDa bands corresponded to glycosylated and un-glycosylated forms of NgR, respectively. However, when the 64 and 48 kDa bands were separately excised from the gel and treated with PNGase F to remove N-linked glycans, their molecular weights were reduced to 53 and 39 kDa, respectively (Fig. 1B). Similarly, the molecular weight of the 48 kDa band in the medium was reduced to 39 kDa following PNGase F treatment (Fig. 1B). Thus, the 64/53 kDa band in the cell lysate probably represented glycosylated/deglycosylated full length NgR, whereas the 48/39 kDa band in both the cell lysate and medium corresponded to a glycosylated/deglycosylated fragment of NgR. As this fragment contained the N-terminal V5 tag and was released into the medium, we postulated that it might represent an N-terminal fragment of NgR (NTF-NgR) lacking the C-terminal GPI anchor of the full length receptor. This was confirmed by treating the cell lysates with phosphatidylinositol-specific phospholipase C (PI-PLC), which cleaves within the GPI anchor, followed by deglycosylation with PNGase F (Fig. 1C). PI-PLC treatment resulted in an up-shift in the molecular weight of the 53 kDa band, indicative of the removal of a GPI anchor (Walmsley et al., 2001), but not the 39 kDa band, demonstrating that it indeed lacked a GPI moiety. When detected with the anti-V5 antibody, NTF-NgR had a slightly higher molecular weight than a construct comprising only the ligand-binding domain of human NgR (NgR-310), which also contained an N-terminal V5 tag (Fig. 1D,E). However, unlike NgR-310, NTF-NgR was also recognised by a human NgR-specific monoclonal antibody that possessed an epitope (residues 328-332) proximal to the C-terminus of the NgR ligand-binding domain (Fig. 1D,E). NTF-NgR thus comprised the entire ligand-binding domain of NgR and terminated beyond residue 332. In addition to full length NgR and NTF-NgR, a band of around 28 kDa was detected in the cell lysate by a human NgR-specific polyclonal antibody (Fig. 1E). That this band was not detected in the medium and was not recognised by either the anti-V5 or human NgR-specific monoclonal antibodies suggests that it corresponds to a membrane-anchored C-terminal fragment of NgR (CTF-NgR).

To determine the cellular location of the NgR fragments, cells were treated with a membrane impermeable biotinylation agent and biotinylated proteins precipitated with streptavidin-agarose beads (Fig. 2A). Full length NgR and CTF-NgR, but not the intracellular protein clathrin, were detected in the precipitate and thus were present at the cell surface. Although NTF-NgR was present in the input cell lysate, this fragment was not detected in the precipitate, consistent with NTF-NgR lacking a membrane anchorage domain and hence being secreted into the medium. As NgR is localised within lipid rafts, we next assessed the raft distribution of the NgR fragments (Fig. 2B). Rafts are insoluble in non-ionic detergents at 4°C and can be isolated from detergent soluble material by flotation to low density on a sucrose gradient. When this procedure was performed, clathrin, a typical non-raft resident protein (Walmsley et al., 2003), was detected in the high-density fractions of the gradient (fractions 1-3), whereas flotillin-1, a well-characterised raft marker protein (Bickel et al., 1997), had migrated to fractions of lower density (fractions 4-8). In agreement with previous reports (Higuchi et al., 2003; Vinson et al., 2003), the NgR co-receptor p75^NTR was found to co-migrate with flotillin in the gradient and was thus raft-associated. Likewise, CTF-NgR and the majority of full length NgR were detected in the raft-containing fractions of the gradient, consistent with these proteins possessing a GPI-anchor. As expected for a soluble protein, NTF-NgR was predominantly detected in the clathrin-containing fractions of the gradient. Even so, NTF-NgR was also observed to co-migrate with flotillin-1 in the raft-containing fractions of the gradient.

To investigate the subcellular site for the generation of NTF-NgR, cells were treated with BFA (Fig. 3A), which blocks protein transport from the ER to the Golgi (Lippincott-Schwartz et al., 1989). BFA treatment slightly increased the mobility of full length NgR in the lysate, possibly because the N-glycans on this form of NgR lacked Golgi-specific modifications. In addition, BFA not only blocked the release of NTF-NgR into the medium, but also prevented the generation of NTF-NgR within the cell. The cleavage of NgR to generate NTF-NgR thus occurred in a post-ER compartment. To establish whether the cleavage of NgR constituted a classical `shedding' process, cells were surface biotinylated at 4°C and then incubated at 37°C for various time periods followed by the precipitation of biotinylated proteins with streptavidin-agarose (Fig. 3B). Biotinylated full length NgR in the lysate decreased in intensity during incubation with the concomitant appearance and increase in intensity of biotinylated NTF-NgR in the medium. Thus, full length NgR at the cell surface can indeed undergo proteolysis to yield NTF-NgR, which is subsequently shed into the medium.

To elucidate the C-terminus of NTF-NgR, this fragment was immunoprecipitated from conditioned medium using the anti-V5 antibody, resolved by SDS-PAGE and stained with Coomassie Blue (Fig. 4A). The band corresponding to NTF-NgR was excised from the gel and subjected to MALDI-TOF mass spectrometry analysis following in-gel digestion with the endoproteinase Lys-C, which cleaves C-terminal to lysine residues (Fig. 4B) (Keil, 1992). As a control, a blank gel slice of comparable size to that of the NTF-NgR band was excised and processed in parallel. Three peaks were detected in the NTF-NgR band that were absent in the control sample which were predicted to correspond to Lys-C derived fragments encompassing residues 278-343 of human NgR (Fig. 4B). Identification of a peak corresponding to a Lys-C derived fragment containing residues 39-277 of human NgR was not expected due to its mass being too high for efficient extraction from the gel. Similarly, the mass of the fragment corresponding to residues 423-447 of human NgR could not be predicted as it contained the GPI moiety. An additional peak was detected at 1396.72 that was absent in the control sample and was predicted to correspond to residues 344-358 of human NgR with the sequence ASVLEPGRPASAGNA (Fig. 4C). The identity of this peptide was confirmed by MS/MS analysis which yielded fragment peaks consistent with the following subunits: AS, ASV, ASVLEPGR, PGRPASAGNA and PASAGNA (Fig. 4D). As this peptide possessed a C-terminal Ala358 residue, it cannot have been derived by cleavage of NgR with Lys-C. This therefore indicated that NTF-NgR terminated at Ala358.

To determine the proteases involved in the generation of NTF-NgR, cells were incubated with inhibitors of the main classes of protease (Fig. 5A), namely, E-64 for cysteine proteases, AEBSF for serine proteases, pepstatin for aspartic proteases, the hydroxamate-based compound PKF226-967 for zinc metalloproteinases (Fig. 5B) and NH₄Cl for lysosomal proteases. Only PKF226-967 inhibited the release of NTF-NgR into the medium, thus attributing the action of zinc metalloproteinases to this process. The level of intracellular NTF-NgR, however, was unaffected by PK226-967 treatment, perhaps because this compound failed to permeate the cell sufficiently to inhibit intracellular processing or the intracellular NTF-NgR detected was generated prior to the treatment. To determine the zinc metalloproteinases involved in the cleavage of NgR, we employed tissue inhibitors of metalloproteinases (TIMPs) as these exhibit distinct inhibition profiles against different zinc metalloproteinases (Baker et al., 2002). TIMP-2 and TIMP-3 potently inhibit all types of MMPs, whereas TIMP-1 is a poor inhibitor of most membrane-type (MT) MMPs. Furthermore, TIMP-1 inhibits ADAM10, whereas TIMP-3 inhibits several ADAMs including tumor necrosis factor-alpha (TNFα) converting enzyme (TACE or ADAM17) (Amour et al., 1998; Amour et al., 2000). When the cells were treated with recombinant human TIMPs (Fig. 5C), TIMP-3, and to a lesser extent TIMP-2, inhibited the release of NTF-NgR into the medium, but no effect was observed for TIMP-1. This pattern of inhibition suggests that the zinc metalloproteinases involved in NgR cleavage probably belong to the MT-MMP group, but it cannot be ruled out at this stage that ADAMs other than ADAM10 may also be involved. As the proteolytic processing of NgR involved a zinc metalloproteinase, we next investigated whether this process had additional properties common to other shedding mechanisms. We therefore treated cells with the cholesterol extraction agent MβCD (Fig. 5D), which has been shown to up-regulate the ectodomain shedding of several proteins (Kojro et al., 2001; Mechtersheimer et al., 2001; Gutwein et al., 2003; Matthews et al., 2003). Although MβCD treatment did not affect the amount of full length NgR and NTF-NgR in the cell lysate, an increase was observed in the amount of NTF-NgR shed into the medium (Fig. 5D). In addition to NTF-NgR, a band probably corresponding to the full length NgR ectodomain was detected in the medium from cells treated with MβCD. PKF226-967 inhibited the MβCD-mediated increase in the amount of released NTF-NgR, implicating the action of zinc metalloproteinases in this process, but, on the contrary, had no effect on the amount of the released full length NgR ectodomain. Thus, the release of the full length NgR ectodomain involved a mechanism distinct to that for NTF-NgR. One possibility is that MβCD treatment induced a cleavage within the GPI anchor of NgR by a phospholipase-like enzyme, a process that has also been reported for another GPI anchored protein, the cellular prion protein, following MβCD treatment of SH-SY5Y cells (Parkin et al., 2004).

To examine whether NTF-NgR is proficient in ligand binding, this fragment was immunoprecipitated from conditioned medium using the anti-V5 antibody and incubated with rat Nogo-66 fused at the N-terminus to human placental alkaline phosphatase (AP-Nogo-66) (Fig. 6A). The amount of AP-Nogo-66 associated with the immunoprecipitate of medium from SH-SY5Y cells expressing NgR was significantly higher (Student's t-test: P<0.005) than that in the immunoprecipitate of medium from mock transfected SH-SY5Y cells, demonstrating that NTF-NgR can bind Nogo-66. Human placental alkaline phosphatase (AP) alone did not bind significantly to immunoprecipitated NTF-NgR. Likewise, binding of AP-Nogo-66 to the surface of CHO cells stably expressing human NgR (Pignot et al., 2003) was significantly reduced (P<0.005, Student's t-test) when the ligand was pre-incubated with medium from SH-SY5Y cells expressing NgR compared to pre-incubation with medium from mock transfected SH-SY5Y cells (Fig. 6B). No significant binding of AP alone was detected with either of the pre-incubation conditions. These results are in accordance with NTF-NgR containing a functional ligand-binding domain and thus competing with membrane bound NgR for Nogo-66. Using a co-immunoprecipitation approach, we next determined whether NTF-NgR associated with the co-receptor of NgR, p75^NTR (Fig. 6C). Along with endogenous human p75^NTR, full length NgR and CTF-NgR was detected in the immunoprecipitate of the anti-p75^NTR antibody, but not in the immunoprecipitate of an unrelated IgG. NTF-NgR, although present in the input lysate, was not detected in the anti-p75^NTR immunoprecipitate, indicating that, unlike full length NgR and CTF-NgR, this fragment failed to associate with p75^NTR.

To corroborate our results in vivo, samples of human adult cortex extracted 4 hours post mortem and cerebrospinal fluid from a live adult were subjected to western blot analysis using the human NgR-specific polyclonal antibody (Fig. 7). NgR was detected as 64, 48 and 28 kDa bands in crude cortical homogenate, a band pattern identical to that for NgR expressed in SH-SY5Y cells. More importantly, the 48 kDa band was also detected in CSF, consistent with it representing a soluble fragment of NgR analogous to NTF-NgR. A minor band of 53 kDa was also detected in CSF, which may represent a further proteolytic product of NgR.

As the mutual receptor for three structurally diverse neurite outgrowth inhibitory molecules, NgR represents an attractive target for therapies designed to improve the regenerative capacity of the CNS. Indeed, several inhibitors of NgR signaling have already been characterised. NEP1-40, a peptide containing the first 40 residues of the Nogo-66 domain, binds NgR and blocks the binding of Nogo-66 but fails to activate the receptor, thus behaving as a receptor antagonist (GrandPre et al., 2002; Li and Strittmatter, 2003). Conversely, soluble forms of NgR that either comprise the ligand binding domain alone or the full length ectodomain lacking the GPI-anchor constitute NgR ligand antagonists as they are proficient in binding ligands but incapable of signaling (Domeniconi et al., 2002; Fournier et al., 2002; Liu et al., 2002; Barton et al., 2003; He et al., 2003). Both types of antagonist significantly block the neurite outgrowth inhibitory effect of CNS myelin in vitro and, at least in the case of NEP1-40, improve functional recovery in a rodent model of spinal cord injury. However, even though exogenous inhibitors of NgR have been identified, endogenous mechanisms for the regulation of the receptor remain unknown. The present study was therefore focused on identifying potential pathways involved in the regulation of NgR in neuronal cells. To this end, we have demonstrated that human NgR expressed in human neuroblastoma cells is constitutively proteolytically processed by the action of a zinc metalloproteinase to generate a soluble N-terminal fragment that is proficient in binding the NgR ligand Nogo-66, but not the NgR co-receptor p75^NTR. More importantly, we detected NgR fragments in the human brain cortex and CSF analogous to those generated in neuroblastoma cells, demonstrating that NgR proteolysis occurs within the human nervous system. Our findings therefore provide for the first time a potential endogenous mechanism for the regulation of NgR.

Shedding mediated by zinc metalloproteinases appears to occur at, or near, the cell surface (Hooper et al., 1997). In this regard, we found that the generation of NTF-NgR was abolished when trafficking to the cell surface was blocked by BFA. Accordingly, proteolysis of NgR was shown to occur at the cell surface as demonstrated by the release of biotinylated NTF-NgR into the medium. The presence of NTF-NgR in the lysate, however, suggests that either NgR proteolysis can also occur intracellularly or the released NTF-NgR binds to the cell surface and/or is internalised. Certainly, soluble NgR has been previously shown to associate with cell surface NgR and this interaction required only the ligand-binding domain of the receptor (Fournier et al., 2002; Barton et al., 2003). However, the fact that NTF-NgR failed to be biotinylated does not support the presence of this fragment on the cell surface. In addition, biotinylated NTF-NgR released into the medium did not appear in the lysate over the course of a 2 hour incubation, thus arguing against the internalisation of this fragment. Consequently, NTF-NgR detected in the lysate probably arises from the proteolysis of NgR in the secretory pathway. As the presence of NTF-NgR in the lysate was abolished by BFA treatment, the cleavage of NgR must be instigated after the receptor is trafficked from the ER compartment but prior to it reaching the cell surface. This scenario is similar to that for the release of APP by α-secretase, where cleavage occurs both at the cell surface and in the trans-Golgi compartment of the secretory pathway (Parvathy et al., 1999; Skovronsky et al., 2000).

In addition to NTF-NgR, a further fragment of NgR was detected in cell lysate with the human NgR-specific polyclonal antibody. That this represented a C-terminal fragment of NgR was evidenced by the fact that it failed to be recognised by monoclonal antibodies against the N-terminal V5 tag or the 328-332 region of NgR. Furthermore, the presence of CTF-NgR on the cell surface but not in the medium and its localisation within lipid rafts is consistent with it containing the C-terminal GPI-anchor of NgR. It is therefore conceivable that CTF-NgR corresponds to the C-terminal product of the cleavage that releases NTF-NgR. Along with CTF-NgR and full length NgR, we also observed some NTF-NgR in the lipid raft fractions of the sucrose gradient. This cannot be due to the contamination of these fractions with non-raft components as they did not contain clathrin, a typical non-raft resident protein (Walmsley et al., 2003). One possibility for this raft localisation may be the interaction of NTF-NgR with raft associated full length NgR. However, as NTF-NgR was not present at the cell surface where rafts are predominantly found (Simons and Ikonen, 1997), the most likely explanation is that intracellular NTF-NgR interacted with full length NgR following lysis of the cells with Triton X-100.

To identify the C-terminus of NTF-NgR, we employed MALDI-TOF and tandem mass spectrometry following digestion of immunoprecipitated NTF-NgR with Lys-C. In addition to detecting peaks for the predicted products of the Lys-C cleavage of NgR, we also identified a peptide that terminated at Ala358. This is two residues N-terminal to Lys360, the predicted C-terminal residue from the Lys-C digestion of NgR. As we detected neither the whole predicted Lys-C peptide (344-360) nor any other Lys-C-peptides C-terminal to Lys360, we surmised that NTF-NgR terminated at Ala358. This result was consistent with the detection of NTF-NgR with the human NgR-specific monoclonal antibody that recognised residues 328-332 of NgR.

Akin to other shedding mechanisms (Schlondorff and Blobel, 1999), the release of NTF-NgR was inhibited by a hydroxamate-based inhibitor of zinc metalloproteinases, PKF226-967. More specifically, the lack of inhibition of this process by TIMP-1 excludes the involvement of soluble MMPs and ADAM10, whereas inhibition by TIMP-2 and TIMP-3 is suggestive of the action of MT-MMPs (Baker et al., 2002). However, as TIMP-3 was a potent inhibitor of NFT-NgR release, the possibility still exists that members of the ADAM family other than ADAM-10 may also participate in the cleavage of NgR. Certainly, that the shedding of TNFα has been found to be mediated by the ADAM member TACE and the MMP member MMP-7 (Peschon et al., 1998; Haro et al., 2000) demonstrates that the shedding of an individual protein can be mediated by members of more than one family of zinc metalloproteinases. In this respect, in common with the substrates of ADAM10 and ADAM17 (Kojro et al., 2001; Mechtersheimer et al., 2001; Gutwein et al., 2003; Matthews et al., 2003), NTF-NgR generation was upregulated by cholesterol depletion with MβCD and this up-regulation was inhibited by PKF226-967. MβCD treatment also induced the release of a protein, which from the molecular weight, presumably corresponded to the full length NgR ectodomain. Unlike the up-regulation of NTF-NgR generation, this induced release of NgR failed to be inhibited by PKF226-967 and hence occurred via a zinc metalloproteinase-independent mechanism. This finding is corroborated by a recent report demonstrating that, within the same cell line as that used in the present study, MβCD treatment induced the release of the ectodomain of another neuronally expressed GPI-anchored protein, the cellular prion protein (Parkin et al., 2004). Here, the authors concluded that this release most probably occurred by the action of a phopholipase-like enzyme that cleaved within the GPI-anchor, thus releasing the full length ectodomain into the medium.

In general, the shedding of membrane-anchored receptors represents a mechanism to down-modulate their function as it leads to a reduction in the amount of functional receptors at the cell surface and releases an ectodomain that, even if it retains the capacity to bind ligands, can no longer signal due to the lack of a cytoplasmic domain (Schlondorff and Blobel, 1999). In the case of NgR, deletion mutagenesis analysis has clearly demonstrated that the binding of ligands to the receptor critically requires the LRRNT, LRR and LRRCT sub-domains of the receptor (Fournier et al., 2002; Liu et al., 2002; Wang et al., 2002a; Barton et al., 2003). Deletion of any two of the eight repeats in the LRR domain is sufficient to abolish ligand binding. On the other hand, the C-terminal domain of NgR, although not essential for ligand binding, is critical for signaling (Domeniconi et al., 2002; Fournier et al., 2002). Accordingly, this domain is necessary for the binding of NgR to its co-receptor, p75^NTR, which transduces the inhibitory signal to the inside of the cell (Wang et al., 2002b; He et al., 2003). Based on these observations, a soluble NgR ligand antagonist, referred to as NgREcto (Fournier et al., 2002) or NgR310 (He et al., 2003) and analogous to NgR-310 in the present study, has been described comprising the ligand binding domain of NgR alone. This protein binds NgR ligands, inhibits the binding of ligands to cell surface NgR and blocks neurite outgrowth inhibition mediated by NgR ligands or CNS myelin in vitro. Similarly, in the present study, we have shown that NTF-NgR, as expected for a fragment containing an intact ligand-binding domain, binds Nogo-66 and inhibits the binding of this ligand to cells expressing NgR. Although NTF-NgR retained part of the C-terminal domain of NgR, we found that this fragment, like NgR310 (He et al., 2003), also failed to associate with p75^NTR. This demonstrates that NTF-NgR, even though it can bind NgR ligands, would be incapable of signaling through p75^NTR. NTF-NgR would therefore be predicted to function as an antagonist of NgR ligands. However, the question remains whether the level of NTF-NgR in the human nervous system would be sufficient for this antagonistic function to be regarded as physiologically relevant. Interestingly, the other product from the cleavage of NgR, CTF-NgR, was found bound to p75^NTR. This finding is not in agreement with a previous report where a soluble construct consisting of the p75^NTR ectodomain fused to alkaline phosphatase failed to bind to CHO cells expressing the C-terminal domain of NgR, although binding was observed to cells expressing full length NgR (Wang et al., 2002b). One explanation for this discrepancy is that although the soluble C-terminal domain of NgR may indeed have a low affinity for p75^NTR, an interaction between CTF-NgR and p75^NTR would be greatly facilitated by their mutual localisation on the plasma membrane and, more importantly, within lipid rafts. Furthermore, if NgR undergoes cleavage when associated with p75^NTR, it is possible that the resulting CTF-NgR may be slow to dissociate from p75^NTR. This interaction would not be expected to result in constitutive inhibitory signalling as the expression of the C-terminal domain of NgR in primary neurons fails to result in spontaneous neurite outgrowth inhibition (Fournier et al., 2002). In addition, the finding that a soluble form of the C-terminal domain of NgR had no effect on neurite outgrowth inhibition mediated by Nogo-66 implies that CTF-NgR would be unlikely to antagonise Nogo-66 signalling (Fournier et al., 2002). However, as noted above, given that discrepancies can occur between findings based on soluble and membrane-anchored proteins, an antagonistic function for CTF-NgR cannot as yet be ruled out.

Of note is the recent finding that p75NTR undergoes proteolysis mediated by TACE, one of the main members of the ADAM family, which results in the release of an N-terminal fragment containing the binding domain for nerve growth factor (Weskamp et al., 2004). Whether this process and the cleavage of NgR are inter-related remains to be seen. Certainly, the proteolysis of NgR and its co-receptor highlights a potential role for ectodomain shedding in the regulation of neurite outgrowth inhibition mediated by NgR ligands.

We thank Aurélie Verles for the production of AP and AP-Nogo-66, Prof. Volker Dietz and Dr Heike Kuenzel (Universitätsklinik Balgrist, Zurich) for the provision of human CSF and the Institute of Pathology (University of Basel) for the provision of human adult cortex. We also thank Judith Schäfer for her excellent technical support in the generation of monoclonal antibodies.