Nogo-A, -B, and -C Are Found on the Cell Surface and Interact Together in Many Different Cell Types*

Nogo-A, -B, and -C are generated from the Nogo/RTN-4 gene and share a highly conserved C-terminal domain. They lack an N-terminal signal sequence and are predominantly localized to the endoplasmic reticulum (ER). We found the N terminus of endogenous Nogo-A exposed on the surface of fibroblasts, DRG neurons, and myoblasts. Surface-expressed Nogo-A was also present on presynaptic terminals of the neuromuscular junction and on DRG neurons in vivo. Surface biotinylations confirmed the presence of all Nogo isoforms on the surface. To search for proteins that interact with Nogo-A and suggest a function for the large intracellular pool of Nogo-A, immunoprecipitations were performed. Surprisingly, the most predominant proteins that interact with Nogo-A are Nogo-B and Nogo-C as seen with radiolabeled lysates and as confirmed by Western blotting in multiple cell lines. Nogo-A, -B, and -C share a 180-amino acid C-terminal domain with two highly conserved hydrophobic stretches that could form a channel or transporter in the ER and/or on the cell surface.

Reticulons are a family of proteins that share a common C-terminal domain and are particularly abundant in the endoplasmic reticulum (ER) 1 (1). The most recognized reticulon, Nogo-A, is a potent neurite outgrowth inhibitor of high molecular mass (200 kDa) (2)(3)(4) mainly expressed in the nervous system in oligodendrocytes and myelin, and in some neuronal populations (5)(6)(7). Nogo-B (55 kDa), a splice variant of Nogo-A, is ubiquitously expressed in many tissues; it has been suggested to function in vascular remodeling (8) and in apoptosis (9), but its function is far from clear. An alternate promoter is used for expressing the smallest RTN-4 form, Nogo-C (25 kDa), a protein highly expressed in differentiated muscle fiber, al-though there is some expression also in brain, in particular in adult Purkinje cells (5,10). As with the related RTN-1 to RTN-3, there is no known function for Nogo-C.
The C-terminal reticulon domain (about 180 amino acids) for these proteins is evolutionarily conserved and can be found in all eukaryotes (11). In contrast, the N-terminal regions are of very different lengths for the different reticulons and show less conservation (12). For instance, the N-terminal regions of the three Nogo proteins can only be found in higher vertebrates. A recent study shows that reticulons can modulate the activity of BACE1 and affect the secretion of amyloid-␤ peptide (13). However, solid data are sparse, and any common functional role for the C-terminal region of reticulons has not yet been discovered.
All reticulons have two large hydrophobic domains near the C terminus of the protein. In the case of rat Nogo proteins these presumed transmembrane regions are 35 and 36 amino acids each, respectively, long enough to span the membrane twice (2,3). In between the two transmembrane domains, the 66-amino acid loop has been found to bind the Nogo receptor subunit, NgR (14). Another important region for the neurite outgrowth inhibitory function is located in the Nogo-A-specific region in the middle of the protein (15). For this domain to be inhibitory, it naturally must be displayed outside the oligodendrocyte to bind and activate a receptor on the surface of neurons or fibroblasts. Specific high affinity binding of this domain to brain membranes or to live fibroblasts has been shown (15). However, the localization and topology of Nogo-A on plasma membranes have not been clear up to now.
Other characteristics of reticulons include a di-lysine ER retention/retrieval signal at the extreme C terminus and a lack of a signal sequence at the N terminus. All the studied proteins of this family display a reticular pattern that co-localizes with ER markers, at least for a large part of the protein localization, when tested by indirect immunofluorescence (1,3,8,15,16). When the second transmembrane domain of Nogo-A was deleted the protein showed a partial cytoplasmic staining; therefore, this region could be critical for insertion into the ER (1). Many membrane-bound proteins that do not contain a signal sequence use one of their hydrophobic domains and regions around it for signaling entry into the ER (17). In addition, other myelin proteins (MAL and PMP-22) have putative di-lysine ER retention/retrieval signals, yet can be found on the cell surface. Surprisingly, even proteins that are well known ER markers such as calreticulin and calnexin can be found in low amounts on the plasma membrane (18 -20). Furthermore, many studies with potassium channels demonstrate that the ER retention signals are masked once the individual subunits form a functional channel, whereby the bulk of the total individual potassium channel subunits remain in the ER with only a very low amount of functional channel expressed on the cell surface (21).
Many other proteins with ER retention/retrieval signals are located on the cell surface; for example, the glutamate receptors AMPA (22) and NMDA (23), a kainate receptor subunit (24), GABA B receptors (25), and a subunit of a calcium channel (26).
Here we show that not only is Nogo-A found on the surface of multiple cell types, but that Nogo-B and Nogo-C can also be detected on the plasma membrane. In addition, all immunofluorescent studies for Nogo-A were performed with antibodies that are specific for the Nogo-A-specific, N-terminal, and middle portion of this large protein, meaning that different cell types have the ability for expressing this region of Nogo-A on the cell surface facing the extracellular space. Immunoprecipitations of cell lysates done with protein radiolabeling demonstrate that in different cell types the predominant proteins that interact specifically with Nogo-A are Nogo-B and Nogo-C. Both Western blotting after immunoprecipitation and size exclusion chromatography confirm the existence of a high molecular mass complex that includes all three Nogo proteins.

MATERIALS AND METHODS
Antibodies-Most antibodies are described in Ref. 15, although some of the names have been changed. These are the former names with the new names: Bianca ϭ Rb1, Laura ϭ Rb173A, Florina ϭ Rb173B. The sheep antibody was made by immunization with two bacterially produced purified peptides from rat Nogo-A, NiG⌬20 (amino acids 544 -725), and NiG⌬6 (amino acids 763-975). For the production of AR2, one 15-amino acid peptide, AKIQAKIPGLKRKAE, was chosen from murine Nogo-A sequence according to the hydrophilicity/hydrophobicity profile and antigenicity prediction analyses. This peptide was located in the C terminus of the reticular homology domain and did not show potential cross reactivity with other reticulons (data not shown). After synthesis and purification by standard protocols, peptides were conjugated to keyhole limpet hemocyanin and injected into two specific pathogen-free rabbits. Immunization was boosted with three subsequent injections at 14, 28, and 56 days. Terminal bleeds were affinity purified against the immunizing peptide and stored as frozen aliquots. The polyclonal antisera were found to react specifically with Nogo extracts was confirmed by the elimination of staining in the presence of excess immunizing peptide (data not shown).
Cells and Cultures-Production and maintenance of cell lines is described in Refs. 15 and 27. In more detail, 3T3 cells were grown in Dulbecco's modified Eagle's medium with high glucose (Invitrogen, Life Technologies, Inc.) containing 10% fetal bovine serum (Invitrogen, Life Technologies, Inc.) and gentamicin (Invitrogen, Life Technologies, Inc.). The cells were always split the day before an experiment and kept in an incubator at 37°C with 5% CO 2 . A description of the brain oligodendrocyte culture is in Ref. 28. Dissociated DRG neuron cultures were made from the trypsin-and collagenase-treated, triturated DRGs of newborn rats that were plated on poly-L-lysine (PLL) and laminincoated coverslips. The DRG neurons were kept in L15 medium with L-glutamine (Sigma) and supplemented with N1 additives (Sigma), 100 ng/ml NGF (2.5 S NGF, purified from male mouse salivary glands) and gentamicin (Invitrogen, Life Technologies, Inc.). C2C12 myoblast cells were grown in Dulbecco's modified Eagle's medium with high glucose (Invitrogen, Life Technologies, Inc.), 10% fetal bovine serum, and chicken embryo extract.
Immunofluorescent Staining-For surface staining: Cells were either plated onto PLL-coated glass coverslips (3T3 and DRG) or onto chamber slides (C2C12 myoblasts) using the medium described above. The cells were washed once with room temperature PBS, then were placed either on ice (3T3 and C2C12 myoblasts) or between 10 and 14°C (DRG), and washed once with cold PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (Ca/Mg-PBS). Cold primary antibody, diluted in blocking buffer (either 2% goat serum, 0.2% fish skin gelatin in Ca/Mg-PBS or 10% fetal bovine serum in Ca/Mg-PBS), was added to live cells on ice for 30 min. The cells were washed three times with cold Ca/Mg-PBS. Fixation was done with cold 4% PFA, and the cells were placed at room temperature for 10 -30 min. The cells were then washed three times with PBS. The secondary antibody diluted in blocking buffer was added to the cells for 30 min at room temperature. The cells were washed three times with PBS and mounted on slides with Mowiol (10% Mowiol 4 -88 (w/v) (Calbiochem) was dissolved in 100 mM Tris, pH 8.5. with 25% glycerol (w/v) and 0.1% 1,4-diazabicyclo[2.2.2]octane (DABCO) was added as an anti-bleaching reagent).
For intracellular staining: cells were washed with PBS, fixed with 4% PFA for 15 min, permeablilized with 0.3% Triton X-100 for 5-10 min, washed three times with PBS, and then blocked for 30 min in blocking buffer (2% goat serum, 0.2% fish skin gelatin in PBS). Primary antibody was added for 30 min in blocking buffer. Subsequently, cells were washed three times with PBS and incubated for 30 min in secondary goat anti-mouse or anti-rabbit Cy3 (1:3000) in blocking buffer. Coverslips were washed three times with PBS and mounted on slides with Mowiol containing 0.1% DABCO. Images were acquired on a Leica type DM RE microscope using the confocal laser scanning system TCS-SL from Leica. An HCX PL APO 40ϫ/1.25 Ph 3 objective was used.
Immunohistochemistry-For the DRGs: an adult rat was injected with NEMBUTAL (Abbott Laboratories), and perfused transcardially with Ringer's solution followed by 4% PFA in 0.1 M phosphate buffer with 5% sucrose. The lumbar DRGs were removed and put in the same fixative for 2 h at 4°C. After fixation the DRGs were incubated in 0.1 M phosphate buffer with 30% sucrose for 2 days. Embedding was done in Tissue Tek (OCT compound; Zoeterwoude, the Netherlands) that was subsequently frozen in isopentane at Ϫ40°C. 20-m serial cryostat sections were mounted on Superfrost-Plus slides (Menzel-Glaeser; Germany) and frozen at Ϫ20°C. After washing the slides in PBS, they were left in blocking buffer (5% bovine serum albumin, 0.2% Triton X-100 in PBS) for 1 h. Then, the slides were incubated with Rb173A antibody (1:1000) in blocking buffer overnight at 4°C. Slides were washed three times with PBS and incubated in anti-rabbit Cy3 antibody (1:500) (Jackson ImmunoResearch) in blocking buffer for 1 h at room temperature. The slides were then washed three times with PBS and mounted with Mowiol containing 0.1% DABCO.
For the diaphragm: an adult rat was decapitated and the diaphragm was immediately and thoroughly washed with PBS and cut in half. One half of the diaphragm was incubated with Rb173A (1:1000) antibody and the other half with ␤-tubulin (1:500) (Chemicon) in PBS containing 5% bovine serum albumin (blocking buffer) for 5 days at 4°C. The floating tissue was then washed with PBS at 4°C for 4 days and subsequently incubated with anti-rabbit or anti-mouse Alexa546 (1: 300)(Molecular Probes) and anti-bungarotoxin Alexa 488 (1:1000) (Molecular Probes) in blocking buffer for 2 days at 4°C. The tissue was then washed for 2 days with PBS. Confocal imaging was done with the floating tissue placed between a glass slide and a glass coverslip. The microscope and imaging process was as described in the above section. The day after imaging, the same non-permeabilized half of the diaphragm that had been incubated with anti ␤-tubulin antibody was then fixed with 4% PFA in PBS, pH 7.4 for 20 min at 4°C, washed with PBS, permeablilized with 0.2% Triton X-100 in blocking buffer for 1 h and incubated with anti-␤-tubulin antibody (1:500) for 1 h. The tissue was washed with PBS and incubated with goat anti-mouse Alexa 546 (1:300) in permeabilization blocking buffer for 1 h at room temperature. Confocal imaging was done as described above.
Surface Biotinylation-3T3 cells were plated the day before the experiment. Plates were placed on ice and washed five times with cold PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (Ca/Mg-PBS). A solution of cold 0.2 mM Sulfo-NHS-LC-LC biotin (Pierce) in Ca/Mg-PBS was incubated on the cells for 1 h. After washing three times with cold Ca/Mg-PBS the biotin was quenched for 10 min with cold Ca/Mg-PBS containing 20 mM glycine. Cells were lysed in either radioimmune precipitation assay buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) or RSB-Nonidet P-40 buffer (10 mM Tris pH 7.5, 10 mM NaCl, 1.5 mM MgCl 2 , 1% Nonidet P-40) containing 1 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Applied Science). The cells were scraped from the plate and centrifuged for 10 min at 10,000 ϫ g. The supernatant was incubated either with streptavidin gel (Pierce) or with preclearing goat anti-mouse IgG and IgM antibody (Jackson ImmunoResearch) for 2 h rotating at 4°C. The samples incubated with streptavidin gel were then washed five times with cold Ca/Mg-PBS, aspirated with a 27gauge needle, and SDS loading buffer was added. They were boiled for 5 min before the supernatant was loaded on a 10% SDS-PAGE. The Western blot was probed with Rb1 antibody (1:25,000). The samples incubated with preclearing antibody had protein G beads (Pierce) added, and were rotated for 2 h at 4°C. The supernatant was moved to a new tube and they were precleared again with Sheep IgG (Sigma) and rabbit serum (Vector Laboratories, Inc.) for 2 h followed by 2 h of incubation with protein G beads. The supernatant was moved to a new tube and incubated with the antibodies of interest while rotating overnight at 4°C. Protein G beads were then added for 2 h at 4°C. The beads were then washed one time with RSB-Nonidet P-40 buffer, three times with Buffer A (10 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 0.2% Nonidet P-40), two times with Buffer B (10 mM Tris, pH 7.5, 500 mM NaCl, 2 mM EDTA, 0.2% Nonidet P-40) and one time with Buffer C (10 mM Tris, pH 7.5). The beads were then aspirated with a 27-gauge needle, SDS loading buffer was added, they were boiled for 5 min, and loaded on a 14% SDS-PAGE. The blot was probed with neutravidin-HRP (1:20,000) (Pierce). 35 S-labeling and Immunoprecipitations-With radioactivity: CHO-NogoA and 3T3 cells were plated the day before the experiment. Primary oligodendrocytes were grown in differentiation medium for 5 days prior to the experiment. Cells were washed with warmed medium and incubated on a rocker at 37°C for 1 or 2 h with a solution containing DME without methionine and cysteine (Invitrogen) and with 450 Ci/ml of Trans 35 S-LABEL (ICN). All subsequent steps were done on ice or at 4°C. The cells were washed with PBS and lysed with CHAPS buffer (50 mM NaH 2 PO 4 , pH 8, 150 mM NaCl, 0.5% CHAPS) with 1 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Applied Science) for 30 min. The cells were scraped, and the solution was centrifuged for 10 min at 10,000 ϫ g. The supernatant was put into a new tube, and a preclearing goat anti-mouse IgG and IgM antibody (Jackson ImmunoResearch) was added. These tubes were rotated for 2 h, then protein G beads were added, and the samples were again rotated for 2 h. This preclearing step was repeated. Subsequently, the antibodies of interest were added, and the samples were rotated overnight. Protein G beads were added to each sample, and the tubes were rotated for 2 h, washed as detailed in the surface biotinylation protocol, aspirated with a 27-gauge needle, incubated with SDS loading buffer and run on a 14% SDS-PAGE. The gel was subsequently dried for 1 h, and then imaged on a Phosphorimager (the gel with CHO samples) or were put on film (the gels with 3T3 and oligodendrocyte samples).
Without radioactivity: cells were plated the day before the experiment. The experiment in Fig. 8 did not involve any transfection, but for Figs. 9 and 10 cells were transiently transfected with the indicated plasmids using Lipofectamine Plus (Invitrogen). The day after transfection cells were washed, incubated with RSB-Nonidet P-40 lysis buffer (10 mM Tris, pH 7.5, 10 mM NaCl, 1.5 mM MgCl 2 , 1% Nonidet P-40 containing 1 mM phenylmethylsulfonyl fluoride and protease inhibitor mixture (Roche Applied Science)), scraped, and centrifuged at 10,000 ϫ g for 10 min at 4°C. The supernatant was put into a new tube, the antibodies of interest were added, and the samples were processed as detailed above in the radiolabeled samples.
Size Exclusion Gel Filtration Column-Ten 150-mm plates were transfected with a plasmid containing rat Nogo-C using Lipofectamine Plus according to the manufacturer's description. After 24 h, the cells were scraped into cold PBS and centrifuged at 1000 ϫ g for 10 min at 4°C. The cells were lysed in 5 ml of ice-cold RSB-Nonidet P-40 lysis buffer (see above) and centrifuged for 10 min at 10,000 ϫ g at 4°C. 140 mM NaCl was added to the extract. A Superdex 200 26/60 column was equilibrated with extraction buffer (150 mM NaCl) on an FPLC system from Amersham Biosciences. 5 ml of cell extract was separated with a flow of 1.8 ml/min, and after 40 min the fraction collector started to collect 1.8-ml fractions. 30 l of fractions 20 -45 were run on a 4 -12% NuPage gel (Invitrogen) and analysed by Western blotting using the AR2 antibody. Fraction 20 corresponds to 36 ml after sample injection.

RESULTS
Antibodies That Recognize Nogo-To study the cellular localization of Nogo-A and to discover proteins that interact with it, different antibodies generated against specific regions of the protein were used for immunofluorescent and immunoprecipitation experiments (Fig. 1A). Two rabbit polyclonal antiserum, a sheep polyclonal antisera, and two mouse mAbs specifically recognize Nogo-A and not the spliced variant Nogo-B or the small protein, Nogo-C, which uses an alternate promoter. The N-terminal 172 amino acids of Nogo-A are identical to the N terminus of Nogo-B, and therefore, one rabbit polyclonal antiserum, Rb1, recognizes both Nogo-A and Nogo-B. The AR2 antibody, a rabbit polyclonal antiserum with the epitope in the common C-terminal region of the protein is able to recognize all three Nogo proteins. The Western blot shows that all antibodies are specific for the epitopes they should recognize (Fig. 1B). In all lysates investigated there were two Nogo-B bands, one at around the apparent molecular mass of 42 kDa and one that ran slightly lower (see the * in Fig. 1B). The lower band may represent a cleavage product or a modified form of Nogo-B.
The predicted molecular mass for rat Nogo-A is 139 kDa and rat Nogo-B is 39 kDa (according to the website: ca.expasy.org/ tools/protparam.html). However, on standard SDS-PAGE, Nogo-A migrates at around 200 kDa and Nogo-B around 55 kDa. The slower migration of Nogo-A has been noted before in the literature and was attributed to the high number of charged amino acids (3). However, we have found that Nu-PAGE 4 -12% gradient gels run with MOPS buffer (Invitrogen) show Nogo-A and Nogo-B running closer to their predicted molecular mass of 140 and 40 kDa, respectively. A difference of protein migration on a standard SDS-PAGE compared with the NuPAGE system has been described for ␣-synuclein (29), a protein, similar to Nogo-A, which has many acidic residues. In all figures with gels or Western blots arrows point to the Nogo proteins, and the figure legends detail the gel type that was used.
Nogo-A Is Found on the Surface of Live Cells by Immunofluorescence-Staining of the cell surface of live, cultured cells were done with 3T3 fibroblasts, DRG neurons, and C2C12 myoblasts. All cells were incubated with Nogo-A-specific antibodies either on a metal sheet on ice (for 3T3 fibroblasts and myoblasts) or at 10 -14°C temperatures (for DRG cultures that appeared to come off the coverslip when put directly on the ice-cold metal sheet). Protein internalization is blocked at temperature below 19°C (30); therefore, our staining method should only show surface-bound antibody, and capping of the antibody should not occur, especially for the case of 3T3 cells and myoblasts that were kept at 0°C. Either the Nogo-Aspecific rabbit polyclonal antiserum Rb173A ( Fig. 2A) (27), and this can clearly be seen in the Western blot (Fig. 1B). The Nogo-A surface staining was punctate and was not uniform around the perimeter of the cells. In particular, the spots were often, but not always, localized to regions where the cells were contacting one another through long extended segments. An antibody against ␤-tubulin and also non-immune rabbit serum showed no detectable staining, confirming the intactness of the cells and the specificity of the Nogo-A staining (data not shown). Intracellular staining for Nogo-A was extremely bright and reticular as would be expected (Fig. 3).
Non-uniform distribution of surface Nogo-A was noticeable also with cultured rat DRG neurons (Fig. 2C). In this case the staining was brighter, and the spots seemed larger, but more polarized than with the 3T3 cells. There was usually more staining on one side of the cell compared with another.
Immature skeletal muscle is known to express Nogo-A (6, 31). We, therefore, studied immature C2C12 myoblasts and found that Nogo-A can be detected on the plasma membrane of these cells using a Nogo-A-specific antibody (Fig. 2D). As with the other two cell types, myoblasts also showed a punctate staining, although it was more evenly distributed around the cell. In conclusion, endogenous Nogo-A can be detected on the surface of different cell types and this staining has a punctate, non-uniform pattern.

Immunfluorescence Studies Show Nogo-A on the Surface of Neurons in Rat
Tissue-Nogo-A staining was also detected on the surface of DRG neurons, and on motoneuron fibers at the presynaptic site of the neuromuscular junction in the diaphragm of adult rats. Fig. 4 shows intracellular, reticular staining (see asterisk) as well as strong staining of the surface of the DRG neuron (see arrowheads). The surface staining on the DRG neuron from tissue covers much more surface and is not punctate as the surface staining of DRG neurons that were grown in culture (compare Fig. 4 to Fig. 2C). The diaphragm of adult rats was rapidly dissected, washed thoroughly and incubated in blocking buffer with Rb173A antibody or anti-␤-tubulin antibody followed by secondary antibody and anti-bungarotoxin (Btx). This resulted in strong staining for Nogo-A at the neuromuscular junctions (Fig. 4). No detergent or fixative was used during the staining of this floating, intact diaphragm, and no ␤-tubulin staining was detected. However, when the same tissue was fixed and permeablized, ␤-tubulin staining was detected. Therefore, the tissue in the experiment had been intact and was not leaky confirming that the Nogo-A staining is surface Nogo-A. Nogo-A was presynaptic as it did not overlap with bungarotoxin, a marker for the postsynaptic acetylcholine receptors. Therefore, the surface staining of Nogo-A is not only found in cell culture, but can also be detected in animal tissue in vivo.
Nogo-A, -B, and -C Are on the Cell Surface of Fibroblasts as Detected by Biochemical Methods-To confirm the cell surface localization of Nogo with a completely different method, 0.2 mM of a membrane-impermeable biotin compound, Sulfo-NHS-LC-LC biotin, was added to live cultured cells to label surface proteins. 3T3 cells were chilled on ice to inhibit internalization of biotinylated proteins. After labeling, the cells were lysed and streptavidin was used to precipitate the biotinylated surface proteins. The sample was divided equally in half, and the proteins were run on SDS-PAGE in parallel with a fraction of the sample from the total lysate. The Rb1 antibody that recognizes Nogo-A and Nogo-B was used to probe half of the Western blot and an antibody against the clathrin heavy chain, an abundant purely intracellular protein, was used to probe the other half (Fig. 5A). 3T3 cells highly expressed Nogo-B on the cell surface. Nogo-A was present in low amounts on the surface. Clathrin heavy chain was not biotinylated, thus showing the integrity of the plasma membrane at the time of biotinylation.
To study the presence of Nogo-C on the plasma membrane a reverse experiment to that described above was performed. 3T3 cells were biotinylated on ice, lysed, and immunoprecipitated with different Nogo antibodies and control antibodies. Immunoprecipitates were run on a gel, and probed with streptavidin-HRP. The very low endogenous levels of Nogo-C were not easily detected with the AR2 antibody. However, Nogo-A, -B, and -C were biotinylated, and thus were all on the cell surface of Nogo-C-transfected cells (Fig. 5B). The antibody that recognizes the very N terminus of Nogo-A and Nogo-B immunopre-

FIG. 3. Intracellular staining of Nogo-A in different cell lines.
3T3 cells, C2C12 myoblasts and DRG neurons were fixed, permeabilized, incubated with an anti-Nogo antibody (Rb173A for the 3T3 cells and DRG neurons and 11C7 for the C2C12 myoblasts), and then detected with a goat anti-rabbit Cy3 antibody. Note the reticular pattern evident in the 3T3 and C2C12 myoblasts. The scale bar represents 10 m. cipitated biotinylated Nogo-A and Nogo-B, but not Nogo-C, and antibodies that are specific for Nogo-A confirmed the presence of Nogo-A on the plasma membrane. Immunoprecipitated clathrin heavy chain was never detected with streptavidin-HRP even at very long exposure times. When the blot was stripped and reprobed with the antibody recognizing clathrin heavy chain, the protein was clearly detected, indicating that the antibody against the clathrin heavy chain was functional and immunoprecipitated intracellular clathrin heavy chain (Fig. 5B). Therefore, the biotinylated Nogo proteins are on the cell surface and are not biotinylated because of rupture of the cell membrane. Combined with the immunofluorescent staining, the biotinylation experiments provide strong evidence that Nogo-A makes its way to the cell surface in low amounts and in addition show that Nogo-B and Nogo-C are also on the plasma membrane.
Nogo-A, -B, and -C Interact and Form a High Molecular Mass Complex-We next wanted to discover binding partners of Nogo-A to see if these could give us any clues as to how the protein traffics to the cell surface, and if there could be an intracellular function for this large pool of Nogo-A remaining in the ER. CHO cells, either wild-type (CHO-WT) or stably overexpressing rat Nogo-Amyc (CHO-NogoA), were labeled with [ 35 S]methionine and -cysteine, and the lysate was immunoprecipitated with different Nogo-specific and control antibodies. As shown in Fig. 6, all Nogo antibodies precipitated large amounts of Nogo-A from the stable CHO-Nogo-A cells, whereas the mouse IgG and rabbit serum did not. Small, but detectable, amounts of endogenous Nogo-A could be detected in the CHO-WT cells and this Nogo-A was not immunoprecipitated with the Myc antibody. Endogenous Nogo-B was pulled-down by the Rb1 antibody in both cell types. Interestingly, not many proteins were detected in every lane from the Nogo-A immunoprecipitated samples that were not nonspecific proteins found also in the negative control lanes: Only two proteins, one slightly above 56 kDa and one slightly below 29 kDa, were detected. These potential Nogo-A interactors were pulled down by all of the Nogo-A-specific antibodies (11C7 ϩ 7B12, Rb173A, Rb173B, and S544). The protein at around 56 kDa is most likely Nogo-B, since it ran at exactly the same molecular mass as the band that is seen in the lane where antibody Rb1 was used for immunoprecipitation of Nogo-A and Nogo-B. The ϳ29-kDa protein would be around the right size for Nogo-C and, as was shown by further immunoprecipitations with Nogo-C recognizing antibodies and by immunoprecipitations combined with Western blots, is most likely Nogo-C (see below). This ϳ29-kDa protein was also immunoprecipitated by the Rb1 antibody in CHO-WT cells. This experiment suggests that Nogo-A interacts with Nogo-B and Nogo-C, independently of species preference because rat Nogo-A pulls down endogenous hamster Nogo-B and Nogo-C.
The CHO immunoprecipitations were done with cells stably overexpressing Nogo-A, and we wanted to test whether Nogo-A would also interact with Nogo-B and Nogo-C when all proteins were endogenously expressed. 3T3 fibroblasts and oligodendrocytes both express Nogo-A, and the same radiolabeling experiments were performed with these cells (Fig. 7, A and B, respectively). In all cases Nogo-A was robustly immunoprecipitated with Nogo-A antibodies. The AR2 antibody, which recognizes Nogo-A, -B, and -C precipitated all three proteins. Again, the Nogo-A specific antibodies co-immunoprecipitated proteins that were exactly the same molecular mass as the proteins pulled down by AR2, highly suggesting that Nogo-B and Nogo-C were interacting with Nogo-A. The background, probably because of proteins binding to the agarose beads or to the non-variable regions of the antibody, was high in both cells types despite two rounds of preclearing. Both the nonimmune rabbit serum and the ␤-tubulin antibody immunoprecipitates contained many of these nonspecific proteins. However, a specific strong band representing ␤-tubulin was clearly detected in FIG. 5. Nogo-A, -B, and -C are detected on the surface of 3T3 cells by surface biotinylation. A, 3T3 cells were biotinylated on ice, lysed, and precipitated with streptavidin-HRP gel. Samples were split in half, run on separate lanes on SDS-PAGE, and Western-blotted. One-half of the blot was probed with an antibody to the intracellular protein clathrin heavy chain and the other half with Rb1 for Nogo-A and -B. B, 3T3 cells were transfected with Nogo-C, surface-biotinylated on ice, lysed, and immunoprecipitated with Nogo antibodies and control antibodies. Samples were run on SDS-PAGE and Western-blotted. The membranes were probed with neutravidin-HRP. After stripping the proteins from the blot, the blot was reprobed with an antibody against clathrin heavy chain, and the result from this is shown in the bottom right corner .   FIG. 4. Confocal microscope images of DRG and neuromuscular junction from adult rat tissue. The DRG was fixed, sectioned, incubated with 11C7, and detected with goat anti-mouse Cy3. The fresh diaphragm from a non-perfused rat was removed, thoroughly washed with PBS, cut in half, incubated either with Rb173A or ␤-tubulin and then stained with goat anti-mouse or anti-rabbit Alexa 546 (red) and anti-bungarotoxin Alexa 488 (green). The non-permeablized tissue that was incubated with ␤-tubulin was then fixed, permeablized, and incubated with anti-␤-tubulin antibody followed by incubation with goat anti-mouse Alexa 546 (red). Cell surface Nogo-A staining of the DRG neuron is shown with arrowheads, and intracellular Nogo-A staining is marked with an asterisk. The scale bars are 20 m.
the ␤-tubulin antibody lanes in both gels slightly above the 56-kDa marker, and for both controls in both gels, Nogo-A, -B, and -C were not detected (Fig. 7, A and B).
The Nogo-C bands were detected in equal, strong amounts in all lanes from Nogo-A immunoprecipitations performed in oligodendrocytes (Fig. 7B). However, in 3T3 cells the Nogo-C protein bands that were co-immunoprecipitated with Nogo-Aspecific antibodies were of much lower intensity than with the AR2 antibody that directly recognizes Nogo-A, -B, and -C (Fig.  7A). This could suggest that the complex of Nogo-A, -B, and -C that forms in different cell types could be composed of different ratios of the three proteins. Unfortunately, because the immunoprecipitation with Nogo-A-specific antibodies also pulls down Nogo-A that may not be interacting with Nogo-B and Nogo-C, quantitating the ratio of each Nogo protein in the complex is not possible with this assay.
The experiments with radiolabeling did not definitively show that the two lower molecular mass proteins that co-immunoprecipitate with Nogo-A are Nogo-B and Nogo-C. Therefore, to determine whether the interacting proteins were really Nogo-B and Nogo-C, immunoprecipitation was combined with Western blotting. CHO Nogo-A cells, 3T3 cells, the CG4 oligodendrocyte line (undifferentiated or differentiated) and NGF-primed or unprimed PC12 cells were lysed with CHAPS or octyl-␤-glucoside as detergents, immunoprecipitated with the Nogo-A-specific S544 antibody and Nogo proteins were detected with the Nogo-A/B-specific antiserum Rb1. In all four cell types, Nogo-A co-immunoprecipitated Nogo-B (Fig. 8). The detergent in the lysis buffer made a large difference in the amount of Nogo-B that could be co-immunoprecipitated. The interaction of the two Nogo proteins was retained in the relatively weak detergent CHAPS, but the stronger octyl-␤-glucoside detergent disrupted most of this interaction.
To test whether the lowest molecular mass protein that was immunoprecipitated after radiolabeling was Nogo-C, immunoprecipitations with the sheep Nogo-A-specific antibody S544 were done with 3T3 cells expressing Nogo-A with a FLAG tag on the N terminus and Nogo-C with a Myc tag on the C terminus. A large amount of Nogo-B and a very small amount of Nogo-C co-immunoprecipitated with Nogo-A when the S544 antibody was used (Fig. 9). When the immunoprecipitation was done with an antibody against the FLAG tag, Nogo-A co-immunoprecipitated Nogo-B, but Nogo-C was not present. When the Myc antibody was used, Nogo-B co-immunoprecipitated with the Myc-tagged Nogo-C; however, Nogo-A was not present. The reason that Nogo-A and Nogo-C did not co-immunoprecipitate when the N-and C-terminal tags were used for the precipitations may be caused by the low level of Nogo-C in the complex in this cell line making the protein level of either Nogo-A or Nogo-C undetectable by Western blot after the immunoprecipitation. However, the lack of the co-immunoprecipitation could also be caused by interference of the Nogo-A, -B, -C complex interaction because the antibodies binding the tags interfere with the stability of the complex. In either case, this result shows that the interaction of Nogo-A, -B, and -C is most likely not caused by a nonspecific aggregation of the three proteins, but more likely by a specific formation of a distinct complex.
To test whether Nogo-A can interact with itself, endogenous Nogo-A from mouse was immunoprecipitated from 3T3 cells transfected with Myc-tagged rat Nogo-A. It is possible to detect the species-specific proteins because of their slightly different molecular masses; mouse Nogo-A runs at ϳ210 kDa whereas rat Nogo-A migrates closer to 200 kDa. As shown in Fig. 10, the endogenous mouse Nogo-A did not co-immunoprecipitate the transfected rat Nogo-A. Previous experiments demonstrated that the Myc antibody successfully co-immunoprecipitated rat Nogo-A with endogenous hamster Nogo-B and Nogo-C (Fig. 5), and another experiment showed that endogenous mouse Nogo-A co-immunoprecipitated transfected rat Nogo-C (Fig.  10). Therefore, it is unlikely that the rat Nogo-A protein does not pull-down the mouse Nogo-A protein because of a species difference. This result shows that the interaction between the Nogo-A, -B, and -C proteins is specific and is not due, for example, to aggregation by the common C-terminal domain.
If the three Nogo proteins form a discrete complex, then they should co-elute after filtration on a size exclusion column. As shown in Fig. 11, Nogo proteins were found in high molecular mass complexes when the lysate from 3T3 cells transfected with Nogo-C were run on a Superdex 200 column. Fractions 32-35 had very high molecular mass complexes of ϳ560 kDa and contained all three Nogo proteins confirming that these proteins interact together. However, Nogo-A alone could be found in fractions 20 -25 with much larger complexes of ϳ670 kDa. The discrete pattern of the Nogo proteins found in these fractions underline the specificity of the interactions, and again shows that the proteins are not aggregating together. Nogo-A in the very high molecular mass fractions was not complexed to Nogo-B and Nogo-C, and therefore, is probably interacting with other proteins.

DISCUSSION
This study provides evidence that the Nogo-A protein is on the cell surface of different cell types expressing endogenous Nogo-A and that Nogo-B and Nogo-C can also be found on the cell surface. The proteins that most vigorously interact with Nogo-A in many different cell types are Nogo-B and Nogo-C. The interaction of these three proteins together is not caused by unspecific aggregation, but is forming a specific complex, almost certainly with functional consequences.
As shown by immunofluorescence with a variety of antibodies as well as by cell surface biotinylations the plasma membranes of cultured oligodendrocytes, neurons, fibroblasts, and myoblasts have Nogo-A with the large N terminus region facing the extracellular space. Therefore, the name reticulon and the early studies that found reticulons to be localized to the ER because of the ER retention signal give a wrong impression: Nogos are also plasma membrane molecules, which, however, does not exclude a second, intracellular function. Many myelin proteins have ER retention signals but are located on the cell surface. Potassium channels, GABA receptors, AMPA recep- tors, NMDA receptors, kainate receptors, and glutamate receptors are functional complexes only when located on the plasma membrane; however, they have large intracellular ER pools, and only a fraction of the total cellular amount of these proteins are found on the cell surface (22)(23)(24)(25)(26).
The present results also show that the large N-terminal region of Nogo-A is located outside cells. In addition, the Nterminal region of Nogo-B was recently shown exposed on the surface of endothelial and smooth muscle cells (8). Therefore, some initial studies depicting the topology of Nogo-A with the N terminus facing the cytoplasm give an erroneous or incomplete interpretation of the topology of Nogo-A on the surface of the cell (3,32). Nogo-A topology is not entirely clear. When the plasma membrane is damaged by incubation with detergents such as saponin or digitonin, under conditions where the ER membrane is left intact, N-terminal Nogo-A is strongly reticular indicating that the N terminus of Nogo-A is in the cytoplasm when the proteins is in the ER (15). Therefore, the N-terminal region of Nogo-A should also be intracellular when the protein is on the plasma membrane. How does Nogo-A have the N-terminal extracellular when it seems to be cytoplasmic in the ER? One possibility is that the protein is able to adopt more than one topology. One topological form has the N terminus in the cytoplasm and the other has the N terminus extracellular. Many other proteins are known to adopt multiple topologies, FIG. 7. Endogenous Nogo-A co-immunoprecipitates Nogo-B and Nogo-C in 3T3 fibroblasts and primary oligodendrocytes. A, 3T3 cells were labeled with [ 35 S]methionine and -cysteine, lysed, and immunoprecipitated with various antibodies against Nogo-A or with control antibodies. Samples were run on 14% SDS-PAGE gel, the gel was dried, and then placed on x-ray film for detection of radioactivity. B, primary oligodendrocytes were grown in culture for 5 days until cells were noticeably differentiated. The same procedure as for 3T3 cells was used for the immunoprecipitation and the detection of proteins.
FIG. 8. Nogo-B co-immunoprecipitates with Nogo-A in different cell types and this interaction is dependent on the lysis buffer. CHO-Nogo-A cells, 3T3 fibroblasts, undifferentiated or differentiated CG4 oligodendrocyte line cells, and unprimed or NGF-primed PC12 cells were either lysed in buffer containing the detergent CHAPS or the detergent octyl-␤-glucoside (OG). Lysates were immunoprecipitated with the S544 Nogo-A-specific antibody. Samples were run on an SDS-PAGE, Western-blotted, and probed with the Rb1 antibody that recognizes Nogo-A and Nogo-B.
FIG. 9. Nogo-B and Nogo-C interact with Nogo-A. Plasmids containing FLAG-tagged Nogo-A and Myc-tagged Nogo-C were transiently transfected into 3T3 cells. After 24 h cells were lysed, and proteins were immunoprecipitated with different antibodies. The control Ab was ␤-tubulin. Samples were run on 14% SDS-PAGE, Western blotted, and probed with the AR2 antibody that recognizes Nogo-A, -B, and -C.
FIG. 10. Transiently expressed Nogo-A does not co-immunoprecipitate with endogenous mouse Nogo-A. 3T3 cells were transfected with a plasmid that expresses rat Nogo-A containing a C-terminal Myc epitope. The cells were lysed, and the lysate was immunoprecipitated with the indicated antibodies. Samples were run on an SDS-PAGE, Western-blotted, and probed with 11C7. such as: P-glycoprotein, ductin, cytochrome P-450s, microsomal epoxide hydrolase (review in Ref. 33), the prion protein (34), hepatitis B, and C virus envelope proteins (35)(36)(37), lactose permease (38,39), and phenylalanine permease (40). Phospholipid composition was found to direct the topology of phenylalanine permease. Other than this hint, there are no other possible mechanisms known for how these proteins are able to achieve multiple topologies.
The Nogo proteins have no signal sequence and may belong, therefore, to the growing group of unconventional proteins that are either secreted or brought to the plasma membrane by alternative trafficking pathways. These unconventional proteins include the secreted interleukin-1␤, many members of the FGF family and CNTF. There are also transmembrane proteins that are transported unconventionally that include the ATP-binding cassette (ABC) transport superfamily proteins, cystic fibrosis transmembrane protein (CFTR) and P-glycoprotein (41,42). The translocation through the membrane for CFTR and PgP occurs in an unusual manner (43)(44)(45). How proteins traffic in a nonclassical pathway to the cell surface is largely unknown.
The presence of Nogo on the plasma membrane has important biological significance. Nogo-A on the surface of oligodendrocytes and myelin in the adult CNS inhibits neurite growth and regeneration (3, 46 -48) and restricts CNS plasticity (47). The exposed N-terminal region of Nogo-B is important for regulating vascular remodeling by influencing endothelial and smooth muscle cell migration and by promoting cell adhesion (8). The N-terminal region of Nogo-A and -B must interact with receptors for these processes to occur. There are three different known receptor binding regions on Nogo (15): (i) The 66 amino acids between the two large transmembrane domains at the C terminus of all Nogo proteins bind the Nogo-66 receptor, NgR, inhibiting neurite outgrowth (14,32). (ii) There is a Nogo-Aspecific region that has a high affinity binding site for a yet unidentified receptor that not only inhibits neurite outgrowth but also the spreading of fibroblasts (15). Active Nogo-A fragments bind with high affinity to brain membranes and live 3T3 cells. (iii) There is a region at the common N-terminal region of Nogo-A and -B that inhibits fibroblast spreading (15), and alters endothelial and smooth muscle migration and adhesion (8) and binds to an unknown receptor. Many signaling molecules are known to have multisubunit receptor complexes; this is probably also the case for Nogo.
We show that Nogo proteins are preferred mutual interaction partners, regardless of whether the proteins are endogenous or overexpressed, and this interaction occurs in fibroblasts, CHO cells, oligodendrocytes, and PC12 cells. Because Nogo-A does not interact with itself, the interaction of Nogo-A, -B and -C is specific and most likely not due to aggregation. We predict that there would be a certain stoichiometry for the Nogo complex. That the three proteins co-elute in a limited number of fractions from a gel filtration column implies that the proteins interact to form a complex of a precise molecular mass. Nogo-A was found, in addition, in fractions from the gel filtration column with very high molecular mass devoid of Nogo-B and Nogo-C, suggesting that there must be Nogo-A-containing complexes with additional proteins. Work is currently being done to find other Nogo-A-interacting proteins.
The staining pattern for Nogo-A on the cell surface is punctate in all three cell types studied. Similarly, Acevedo et al. (8) found Nogo-B surface staining also to be punctate on the surface of endothelial cells. Many receptor complexes cluster on the plasma membrane in a punctate pattern, and their localization is often caused by specific plasma membrane microdomains. The discovery of Nogo-B as an inhibitor of vascular remodeling was based on its enrichment in caveolae and/or lipid rafts (8). Clustering of Nogo with other neurite outgrowth inhibitors on the plasma membrane, such as myelin-associated glycoprotein (MAG) and oligodendrocyte myelin glycoprotein (OMgp) might be important for transmitting an optimal neurite outgrowth inhibitory signal to the neuron. One reason that Nogo-A antibody treatment in vivo is so successful may be that antibody binding to surface Nogo-A not only blocks the interaction of Nogo-A with the neuronal receptor(s), but the other inhibitory proteins in the same microdomain would also lose their binding to the neuron by steric hindrance or by endocytosis of the entire inhibitory domain into the oligodendrocyte.
What could the function for the Nogo-A, -B, -C complex be? The hydrophobic regions at the C termini of the proteins are highly conserved evolutionarily, even among other reticulons. The two transmembrane domains that are 35 and 36 amino acids, respectively could be traversing the membrane once or looping through the membrane twice. Protein complexes that form by interactions among proteins that are spliced variants and have conserved transmembrane domains include ion channels, pores, and transporters (49,50). The N-terminal regions of both Nogo-A and -B have a putative calcium binding site (27). Therefore, we speculate that Nogo proteins could form a channel or transporter that may regulate calcium. The variable length N termini of reticulon subunits is also a property frequently found in the spliced variant subunits that form ion channel complexes. If the Nogo complex is operating as a channel, it could be functional in the ER, in another intracellular organelle and/or on the plasma membrane. Indeed, it may be that reticulons could form a family of channels where the different N-terminal regions of the proteins would alter the channel function or operation. This could be why different reticulons are up-or down-regulated during different periods of development and in different cell types. For example, there is a down-regulation of Nogo-A in Purkinje cells and in skeletal muscle during differentiation (5). In the muscle of patients with FIG. 11. Size exclusion gel filtration of 3T3 cell lysates demonstrate that Nogo-A, -B, and -C co-elute in high molecular mass fractions. The lysate from 3T3 cells transfected with rat Nogo-C was run over a Superdex 200 gel filtration column, and 1.8-ml fractions were collected. Thirty microliter samples from the fractions were run on a 4 -12% NuPAGE gradient gel. The gel was Western-blotted and probed for Nogo-A, -B, and -C. Based on molecular mass standards, fraction 20 corresponds to complexes of ϳ670 kDa, fraction 32 is ϳ560 kDa, and the last fraction shown, number 42, is ϳ400 kDa.
ALS there is a down-regulation of Nogo-C while there is a concomitant up-regulation of Nogo-A (10). For RTN-3, the expression for different spliced variants is highly regulated during development (12). Also, in the Nogo-A knock-out mouse, we found an up-regulation of Nogo-B (51).
In conclusion, our results suggest that Nogo proteins are multifunctional proteins. There is most likely an evolutionarily "old" function that developed from the ancient, highly conserved reticulon domain. Reticulons can create protein complexes with the different splice variants as subunits, and these most likely function in the ER and possibly also on the cell surface. The evolutionarily "new" function for Nogo would be from the more newly evolved N-terminal domain. The N terminus of Nogo-A and -B contains an active site that influences cell migration and in the middle region of Nogo-A is a site that inhibits neurite growth. These active regions of Nogo, when presented outside of the cell, are involved in the regulation of nervous system plasticity and repair and in cell motility and migration of non-neuronal cells.