The G82S Polymorphism Promotes Glycosylation of the Receptor for Advanced Glycation End Products (RAGE) at Asparagine 81

Interaction between the receptor for advanced glycation end products (RAGE) and its ligands amplifies the proinflammatory response. N-Linked glycosylation of RAGE plays an important role in the regulation of ligand binding. Two potential sites for N-linked glycosylation, at Asn25 and Asn81, are implicated, one of which is potentially influenced by a naturally occurring polymorphism that substitutes Gly82 with Ser. This G82S polymorphic RAGE variant displays increased ligand binding and downstream signaling. We hypothesized that the G82S polymorphism affects RAGE glycosylation and thereby affects ligand binding. WT or various mutant forms of RAGE protein, including N25Q, N81Q, N25Q/G82S, and N25Q/N81Q, were produced by transfecting HEK293 cells. The glycosylation patterns of expressed proteins were compared. Enzymatic deglycosylation showed that WT RAGE and the G82S polymorphic variant are glycosylated to the same extent. Our data also revealed N-linked glycosylation of N25Q and N81Q mutants, suggesting that both Asn25 and Asn81 can be utilized for N-linked glycosylation. Using mass spectrometry analysis, we found that Asn81 may or may not be glycosylated in WT RAGE, whereas in G82S RAGE, Asn81 is always glycosylated. Furthermore, RAGE binding to S100B ligand is affected by Asn81 glycosylation, with consequences for NF-κB activation. Therefore, the G82S polymorphism promotes N-linked glycosylation of Asn81, which has implications for the structure of the ligand binding region of RAGE and might explain the enhanced function associated with the G82S polymorphic RAGE variant.

The receptor for advanced glycation end products (RAGE) 2 is a multiligand receptor that binds to carboxymethyl lysineand AGE-modified proteins and lipids but also to more auton-omous ligands including high mobility group box 1 protein (HMGB1), members of the S100/Calgranulin protein family, amyloid ␤-peptide and Mac-1 (1)(2)(3)(4)(5)(6)(7). Binding of ligands to RAGE is mediated by an extracellular region of the receptor comprising an N-terminal variable (V)-domain and two constant (C)-immunoglobulin domains (2,8,9). Interaction between RAGE and ligands recruits diverse signal transduction pathways involving NF-B and MAP kinase activation and consequently the expression of proinflammatory genes (10 -14). RAGE also acts as an endothelial adhesion receptor promoting leukocyte recruitment during inflammation via a direct interaction with the ␤2-integrins Mac-1 or p150,95 expressed by leukocytes (15,16). RAGE contains two potential N-linked glycosylation sites at Asn 25 and Asn 81 (1). It is now well established that RAGE is N-link glycosylated and that some of the added N-glycan is further modified, resulting in an anionic nonsialylated carboxylated N-glycan (17)(18)(19)(20). The nonsialylated carboxylated Nglycan is essential for binding of RAGE to HMGB1, S100A8/A9, and S100A12 in particular, indicating that N-linked glycosylation and/or the modification of the added N-glycan play important roles in the regulation of RAGE-ligand binding (18,21,22).
A naturally occurring polymorphism has been identified that results in a glycine-to-serine substitution at position 82 within the V-domain. This polymorphism occurs with relatively high incidence compared with other RAGE polymorphisms that have been identified (23,24). The G82S mutant RAGE displays enhanced ligand binding to S100A12 and AGE ligands (25,26). Consequently, this RAGE variant is associated with increased NF-B activation and inflammatory gene expression (25,26). In addition, the G82S polymorphism is associated with reduced levels of soluble RAGE (sRAGE) that in a number of diseases magnifies the contribution from RAGE toward inflammation (27)(28)(29)(30)(31). How ligand binding and sRAGE levels are altered by the G82S polymorphism is unknown. The G82S substitution occurs within one of the potential N-linked glycosylation consensus sites, involving Asn 81 . On this basis, we hypothesized that the G82S substitution may influence the glycosylation pattern of RAGE, with consequences for ligand binding and proinflammatory signaling. Here, we describe detailed analysis of the glycosylation of RAGE and identify enhanced glycosylation induced by the G82S polymorphism.
Biotinylation of Cell Surface Proteins-Transient transfected cells expressing WT or mutant forms of RAGE were washed twice with cold PBS and then incubated with PBS containing 1 mg/ml sulfo-NHS-LC-Biotin (Pierce) for 30 min on ice for cell surface biotinylation. The cells were then washed three times with cold PBS and lysed in PBS containing 1% (vol/vol) SDS and 1 ϫ Complete protease mixture to which was added an equal volume of TBS containing 0.4% (vol/vol) Triton X-100. Cell debris was removed by centrifugation at 10,000 ϫ g for 10 min at 4°C, and the supernatant was incubated with NeutrAvidin TM protein beads (Pierce) at 4°C overnight. After three washes with cold PBS, biotinylated cell surface proteins were eluted by boiling for 5 min in SDS-PAGE sample buffer (150 mM Tris, pH 6.8, 5% SDS, 0.08% bromphenol blue, 25% glycerol with 0.01% (vol/vol) 2-mercaptoethanol) and analyzed by Western blotting. As a control, Western blot analysis of the endogenous intracellular protein, Murr-1 was used to confirm that biotinylated RAGE proteins are from the cell surface.
Enzymatic Deglycosylation-Whole cell lysates from transient transfected HEK293 cells were digested with peptide-Nglycosidase-F (PNGase-F; Roche Applied Science). From 10 to 20 g of protein in 20 mM Tris-HCl, pH 7.5, 120 mM NaCl was denatured by addition of SDS to 0.25% then incubated in a boiling water bath for 10 min. Upon cooling, 2 units of PNGase-F in 100 mM phosphate buffer, pH 7.2, containing 25 mM EDTA and 2.5% (vol/vol) Triton X-100 was added and the samples incubated at 37°C for 6 h. Duplicate aliquots of whole cell lysates were digested with endoglycosidase-H (endo-H; Roche Applied Science). Protein was denatured in 0.1 M 2-mercaptoethanol containing 0.1% (vol/vol) SDS at 90°C for 5 min, cooled, and then diluted 1:1.5 in 0.5 M sodium citrate, pH 5.5, containing 1 ϫ Complete protease inhibitor mixture before adding 1 unit of endo-H and incubation at 37°C for 6 h. All glycosidase-treated samples were separated by SDS-PAGE and analyzed by Western blotting or mass spectrometry as described below.
Western Blotting-From 5 to 15 g of cell lysates were analyzed by 12% SDS-PAGE under reducing conditions (32) and Western blotting. RAGE protein was detected using polyclonal goat anti-human RAGE antibody (0.2 g/ml; R&D Systems) and then with appropriate HRP-conjugated rabbit anti-goat antibody (0.2 g/ml; Rockland) and Supersignal CL-HRP (Pierce).
Flow Cytometry-Cell surface expression of WT RAGE and mutant RAGE was verified by flow cytometry analysis. Transient transfected HEK293 cells were washed with cold PBS and detached from the cell culture flask by gentle scraping. Cells were then resuspended at 2 ϫ 10 6 cells/ml in PBS containing 1% BSA. An aliquot of 10 5 cells was stained with or without a goat anti-human RAGE antibody (25 g/ml; R&D systems) and incubated on ice for 30 min. Cells were washed twice by suspension in PBS containing 0.2% BSA, followed by centrifugation at 200 ϫ g for 3 min. Bound antibody was detected by incubation on ice for 30 min with 100 l of Alexa Fluor 488 rabbit anti-goat IgG (17.5 g/ml; Invitrogen). Cells were washed twice and analyzed using a FACScalibur flow cytometer (Becton Dickinson) operated with CELLQuest Pro TM software (Becton Dickinson). For each sample, information on at least 5000 cells was collected.
Immunoprecipitation-WT and G82S mutant RAGE were isolated from whole cell lysates using immunoprecipitation. Biotinylated polyclonal goat anti-human RAGE antibody (R&D Systems) was bound to NeutrAvidin TM protein beads (Pierce) by incubating at 4°C overnight. Whole cell lysate was precleared by incubating with NeutrAvidin TM protein beads (Pierce) at 4°C for 4 h. RAGE protein was then isolated from precleared lysate by incubation with the anti-RAGE antibody bound to the beads at 4°C overnight. After three washes with cold PBS, bound RAGE protein was eluted by boiling for 10 min in SDS-PAGE sample buffer.
Mass Spectrometry-RAGE protein eluted following immunoprecipitation was separated on 12% polyacrylamide gel then stained with SimplyBlue TM Safestain reagent (Invitrogen). RAGE protein bands with appropriate molecular mass were excised from the gel and subjected to in-gel tryptic digestion according to Shevchenko (33). Eluted peptides were dried using a centrifugal vacuum concentrator and resuspended in 2% (vol/ vol) acetonitrile and 0.2% (vol/vol) formic acid in water. All samples were analyzed by HPLC-coupled nanospray tandem mass spectrometry (MS/MS) using an Ultimate 3000 nanoflow HPLC system (Dionex) in-line coupled to the nanospray source of a LTQ Orbitrap XL mass spectrometer (Thermo Scientific). Five l of peptide solution was loaded onto an in-house packed emitter tip column (75-m inner diameter fused silica PicoTip (New Objectives) packed with C-18 material (5-m beads, 100 Å pore size) on a length of 8 -9 cm) and separated by an acetonitrile gradient developed from 2% (vol/vol) acetonitrile and 0.2% (vol/vol) formic acid in water to 55% (vol/vol) acetonitrile and 0.2% (vol/vol) formic acid in water over 15 min. Full mass spectra were acquired in the Orbitrap analyzer at full width at half-maximum resolution of 60,000 in profile mode followed by the acquisition of four data-dependent collision induced dissociation MS/MS spectra in the LTQ ion trap allowing two microscans per fragment spectrum. For protein identification MS/MS spectra were searched against a subset of the NCBI nonredundant sequence data base containing all human amino acid sequences (download Nov. 2009; containing 508,360 entries) using the Mascot search engine. All searches were performed for tryptic and nontryptic cleavages and with mass tolerances of 10 ppm for MS and 0.8 Da for MS/MS. The significance threshold for peptide identification by MS/MS data was p Ͻ 0.01. Extracted ion chromatograms were plotted using the Xcalibur software. The whole isotope cluster of precursors of interest was selected to account for overlapping intensities of isotope peaks of the unmodified and deamidated peptides.
NF-B Nuclear Translocation Assay-10 7 HEK293 cells were transfected with control pcDNA3.1 plasmid (mock) or constructs encoding WT or various mutant forms of RAGE in serum-free DMEM for 6 h. Plasmid DNA was removed and incubation continued in serum-free DMEM for a further 18 h before addition of S100B in PBS to 10 g/ml. Controls received an equivalent volume of PBS. After 30 min, cells were lysed in 140 l of 10 mM HEPES, pH 7.9, containing 10 mM KCl, 0.1 mM EDTA, 0.5 mM DTT, 0.2% (vol/vol) Nonidet P-40, and 1 ϫ Complete protease inhibitor mixture for 15 min on ice, then Dounce-homogenized. Nuclei were pelleted by centrifugation at 10,000 ϫ g for 5 min at 4°C, and the cytosolic supernatant was retained. Nuclei were washed with PBS containing protease inhibitor, then lysed in 25 l of 20 mM HEPES, pH 7.9, containing 400 mM NaCl, 1 mM EDTA, 10% (vol/vol) glycerol, and protease inhibitor by brief sonication. The nuclear extract was centrifuged at 10,000 ϫ g for 8 min at 4°C and then analyzed by SDS-PAGE and Western blotting. Western blots were developed using rabbit anti-human NF-B antibody (1:1500; Sigma) and HRP-conjugated goat anti-rabbit secondary antibody (DAKO) and Supersignal CL-HRP (Pierce). Analysis included detection of actin or tubulin proteins using protein-specific mouse antibodies (both Sigma) and HRP-conjugated rabbit anti-mouse secondary antibody (DAKO). Intensities of NF-B p65 protein bands (normalized to actin bands) in nuclear extracts from S100B-treated cells were compared with those for PBS control-treated cells in three separate experiments. An overall mean value was generated and used to compare differences in response to S100B between mock-and RAGE-trans-fected HEK293 cells using one-way ANOVA. Statistical significance was defined by p Ͻ 0.05.

RESULTS
The G82S Mutation Has No Impact on the Extent of Glycosylation of RAGE in HEK293-To compare glycosylation patterns, WT or G82S mutant variant forms of RAGE were expressed from transient transfected HEK293. The molecular masses of RAGE protein in whole cell lysates were initially established by SDS-PAGE and Western blotting with or without PNGase-F or endo-H treatments (Fig. 1A). It was revealed that HEK293 cells express three species of WT and G82S RAGE with molecular masses of 59, 57, and 52 kDa. After PNGase-F treatment of WT and G82S RAGE, bands of 52 and 50 kDa appeared at the expense of the 59-kDa and 57-kDa proteins A, cell lysates containing WT or G82S RAGE without treatment (Ϫ) or after PNGase-F or endo-H treatment (ϩ) were subjected to Western blotting, then detected with RAGE-specific antibody. Untreated WT and G82S RAGE preparations comprise three protein species with molecular masses of 59, 57, and 52 kDa. Following PNGase-F treatment, 52-kDa and 50-kDa protein species were detected in WT and G82S RAGE preparations. A PNGase-F-resistant protein band with a molecular mass of 57 kDa was observed only in WT RAGE preparation. Endo-H treatment revealed 57-, 52-, and 50-kDa protein species in both WT and G82S RAGE preparations. B, cell surface biotinylation of RAGE-expressing HEK293 cells revealed that all of the WT and G82S RAGE protein species were expressed on the cell surface. Cell surface proteins were biotinylated and membrane fractions precipitated with NeutrAvidin beads before detection by Western blotting using anti-human RAGE antibody. Endogenous, intracellular protein Murr-1 was detected only in cell lysate preparations. C, cell surface expression of WT and G82S RAGE on HEK293 cells was confirmed by flow cytometry analysis using RAGE-specific antibody and Alexa Fluor 488 rabbit anti-goat antibody. (Fig. 1A). The data suggest that WT and G82S RAGE are both N-link glycosylated to the same extent. The one point of difference in the glycosylation patterns was that for WT RAGE, some of the 57-kDa protein species remained PNGase-F-resistant. This did not occur in G82S RAGE. After separate treatment with endo-H, protein species with molecular masses of 57, 52, and 50 kDa remained in both WT and G82S RAGE preparations (Fig. 1A). Thus, similarities in glycosylation patterns between the two forms of RAGE extend to the type of glycan involved. In each, the 59-kDa protein species is modified by high mannose and/or hybrid glycans, and the 57-kDa protein species is modified by complex glycans.
RAGE is a cell surface receptor. Accordingly, we compared the protein species comprising WT and G82S RAGE, expressed on the cell surface, by using a cell surface biotinylation assay. This showed that all of the protein species from both WT and G82S RAGE are expressed on the cell surface of HEK293 cells (Fig. 1B). Intracellular Murr-1 protein was not biotinylated but could be detected in cell lysate preparations, confirming that biotinylated RAGE protein species are derived from the cell membrane and are not contaminated by cytosolic proteins. Furthermore, flow cytometry analysis confirmed that both WT and G82S RAGE are expressed on the surface of transfected HEK293 cells (Fig. 1C).
Mutational Analysis Suggests That the G82S Polymorphism Promotes N-Linked Glycosylation at Asn 81 -To investigate further how RAGE is glycosylated, we produced mutant forms of RAGE where Asn 25 and/or Asn 81 was substituted by Gln, thereby eliminating potential N-linked glycosylation sites. In addition, the G82S mutation was introduced in combination with an N25Q mutation. Surface expression of these mutant forms of RAGE was again confirmed by flow cytometry analysis (data not shown). Moreover, cell surface biotinylation assays showed that all of the protein species comprising mutant forms of RAGE were expressed on the cell surface, indicating that N-linked glycosylation is not necessary for cell surface trafficking of RAGE (Fig. 2). The molecular masses of protein species comprising WT and mutant forms of RAGE are summarized in Table 1. The N25Q/N81Q mutant, which does not contain any N-linked glycosylation sites, comprises two protein species with molecular masses of 52 and 50 kDa. This infers an additional (post-translational) modification to RAGE, which adds ϳ2 kDa to the molecular mass of RAGE. In the N25Q mutant, an additional species of 57 kDa is present, as well as the 52-and 50-kDa species. This suggests that N-linked glycosylation can occur at Asn 81 . It has been previously shown that Asn 25 is favored for N-linked glycosylation in WT RAGE (34). Combining the data from our mutation analyses suggests that reciprocal promotion of glycosylation occurs in the absence of an alternative glycosylation site. When the Asn 81 site is eliminated, the unglycosylated 50-kDa species is absent, suggesting that more efficient N-linked glycosylation of Asn 25 occurs. Finally, in the N25Q/G82S mutant, 57-and 52-Da species are present. Comparing component protein species of the N25Q mutant and N25Q/G82S mutant reveals that the unglycosylated 50-kDa species is missing in the N25Q/G82S mutant. This indicates that N-linked glycosylation at Asn 81 is promoted by the G82S mutation, when Asn 25 is eliminated.
Differences in Asn 81 Glycosylation between WT and G82S Mutant RAGE-High resolution HPLC-coupled LTQ Orbitrap mass spectrometry of tryptically digested WT-and G82S RAGE was performed to establish whether N-linked glycosylation of Asn 81 is promoted in G82S RAGE. To address this question, the extent of Asn 81 to Asp 81 conversion (deamidation) was measured after hydrolytic deglycosylation of RAGE by PNGase-F (35)(36)(37). To account for nonenzymatic deamidations, we always compared the ratio of tryptic peptides containing deamidated Asn versus the unmodified form before and after PNGase-F treatment. In tryptic digests of WT RAGE, Asn 81 was detected in the peptide (VLPN 81 GSLFLPAVGIQDEGIFR) at m/z 1121.615 of the doubly charged ion. However, a peptide at m/z 1122.107 was also detected that may represent a naturally or nonenzymatically deamidated Asn 81 (Fig. 3). Both peptides were chromatographically well resolved with the deamidated form eluting ϳ15 s later. High mass accuracy Orbitrap measurement and collision-induced dissociation-based peptide fragmentation unambiguously discriminated the unmodified peptide from the deamidated species with significant Mascot ion scores of 101 and 82, respectively (supplemental Fig. 1). Following PNGase-F treatment, the ratio of peak intensities of  Table 1. deamidated (Asp 81 ) versus unmodified (Asn 81 ) peptide was significantly increased as shown by the extracted ion chromatograms of signals between m/z 1121.6 and 1124.2 which covers the isotope clusters of both peptide species (Fig. 3). The increase of deamidation by PNGase-F treatment of WT RAGE indicates that a subpopulation of Asn 81 is glycosylated, as recently reported (22). The situation for WT RAGE contrasts with data from G82S RAGE, where neither the tryptic peptide (VLPN 81 SSL-FLPAVGIQDEGIFR) nor its deamidated form was detected at the expected m/z values of peptide ions. However, prior PNGase-F treatment revealed a strong signal for the deamidated form of the peptide at m/z 1137.117 of the doubly charge species. The deamidated peptide was identified significantly by a Mascot search, resulting in an ion score of 95 (supplemental Fig. 2). No unmodified peptide was detected in the PNGase-F treated G82S RAGE protein. These data confirm that the Asn 81 site is fully glycosylated in G82S RAGE with significant amounts of N-linked glycan. Our data clearly show that the G82S polymorphism promotes N-linked glycosylation at Asn 81 .
Consistent with N-linked glycosylation, Asn 25 was not covered by any of the detected peptide species from WT or G82S RAGE. However, following PNGase-F treatment only low intensity MS signals for the deamidated peptide (AQN 25 ITAR) at m/z 387.708 of the doubly charged ion were detected, with Mascot ion scores Ͻ44 for MS/MS spectra (data not shown). Accordingly, MS evidence for Asn 25 glycosylation in WT and G82S RAGE was inconclusive.
N-Linked Glycosylation Influences the Outcome of RAGE-S100B Ligand Interaction-Asn 81 is located in close proximity to a hydrophobic cavity in the RAGE V-domain. Hydrophobic interactions involving residues within this cavity are critical for RAGE-S100B protein ligand binding (38) and are those most likely to be influenced by Asn 81 glycosylation. Any effect of N-linked glycosylation on the outcome of the RAGE-S100B ligand interaction was considered through measures of NF-B activation. NF-B p65 protein was present in cytoplasmic extracts from mock-transfected HEK293 cells and those expressing WT and mutant forms of RAGE. These levels of NF-B p65 were not affected by S100B treatment (supplemental Fig. 3).
Consistent with NF-B activation, S100B treatment increased the levels of NF-B p65 in nuclear extracts from transfected HEK293 cells expressing WT or G82S RAGE (Fig. 4). Cells expressing N25Q mutant RAGE showed no significant difference in nuclear NF-B p65 levels from mock-control cells or between S100B-treated and PBS control-treated cells. In contrast, S100B treatment of cells expressing the N81Q mutant caused NF-B activation (Fig. 4). These data indicate that N-linked glycosylation of Asn 25 is required for S100B binding to WT RAGE. However, S100B treatment also induced NF-B activation through the N25Q/G82S mutant, whereas this RAGE ligand had no effect when acting through N25Q/N81Q mutant RAGE (Fig. 4). It appears that the glycosylation of Asn 81 also affects the hydrophobic interaction(s) between S100B ligand and RAGE.

DISCUSSION
The G82S polymorphism of RAGE has important functional implications because it is associated with enhanced ligand binding, reduced sRAGE production, and consequently an enhanced receptor signaling (25)(26)(27)(28)(29). Nevertheless, it is not well understood how this RAGE polymorphism affects ligand binding and soluble RAGE production. There are two potential N-linked glycosylation sites within the ligand binding domain of RAGE, at Asn 25 and Asn 81 . The inclusion of the G82S polymorphism within the Asn 81 site suggested to us that this polymorphism might affect RAGE glycosylation.
Previously, it has been shown that Asn 25 in WT RAGE is always modified with fully processed N-linked glycan, whereas Asn 81 is not favored for N-linked glycosylation (34). More recent analysis indicates that Asn 81 is modified (22). Our data concur with the glycosylation at Asn 25 in WT RAGE and suggest that this is also the case for G82S RAGE. In particular, an N81Q mutant was N-link glycosylated, and tryptic peptides including Asn 25 were not detected from either WT or G82S RAGE by MS. However, we were unable to confirm that a deamidated version of this peptide could be generated by PNGase-F treatment. This anomaly in MS analysis is striking in its similarity between WT and G82S RAGE.
A mutagenesis approach also supported that N-linked glycosylation of Asn 81 was possible in WT RAGE. For example, glycosylation at Asn 81 occurred when Asn 25 was eliminated. In addition, a comparison showed there were similar glycosylated protein species (i.e. 57 and 52 kDa) comprising the N81Q and the N25Q mutants that each retained a single, but different Asn for N-linked glycosylation. However, the presence of an additional nonglycosylated 50-kDa component to the N25Q mutant suggests that the glycosylation of Asn 81 is less efficient amid WT RAGE protein sequence.
We also utilized a combination of MS analysis and PNGase-F digestion to confirm that Asn 81 is modified by N-linked glycosylation. Specifically for WT RAGE, our results are consistent with a recent report that described the glycosylation of Asn 81 to involve variable, partially processed glycan (22). Here also, our results indicate that the variable amount of carbohydrate added to Asn 81 is not guaranteed in WT RAGE. Tryptic peptides derived from WT RAGE were consistent with the presence of both unmodified and deamidated forms of Asn 81 . Although we found evidence for the deamidated Asn 81 peptide prior to PNGase-F treatment, suggesting some spontaneous deamidation, our data also showed that the deamidated Asn 81 peptide increased following PNGase-F treatment, indicating that a subpopulation of Asn 81 is glycosylated as reported previously (22). Consequently, we conclude that Asn 81 in WT RAGE may or may not be glycosylated.
A comparison with G82S RAGE extended the detail of RAGE glycosylation, suggesting that the glycosylation of Asn 81 is in fact promoted by the naturally occurring G82S polymorphism. Similar to WT RAGE, the absence of Asn 25 in the G82S polymorphic RAGE variant allowed N-linked glycosylation at Asn 81 . However, further data indicated that Asn 81 might provide a key point of difference between the glycosylation patterns of WT and G82S RAGE. For example, a comparison of the N25Q mutant (in otherwise WT RAGE) and the combination N25Q/G82S mutant revealed that the additional 50-kDa unmodified protein component of the N25Q mutant is missing in the combination mutant. This provided further evidence for less efficient glycosylation of Asn 81 within WT RAGE and conversely the promotion of Asn 81 glycosylation within G82S RAGE. In fact, the overall analysis of G82S RAGE-based mutants showed no evidence of variation in the use of the Asn 81 glycosylation site. Consequently, in G82S RAGE, glycosylation of Asn 81 is always seen. Again this was verified using a combination of PNGase-F digestion and MS analysis to identify N-linked glycosylation sites in G82S RAGE. In contrast to WT protein, Asn 81 -containing tryptic peptides derived from G82S RAGE were not detected. However, deamidated Asn 81 peptides were present following PNGase-F treatment. Combined, our data show that the G82S polymorphism promotes N-linked glycosylation of Asn 81 , while having no effect on Asn 25 glycosylation. A consequence is that in the G82S polymorphic variant of RAGE, Asn 81 is always glycosylated.
From our data there appears to be contrasting use of Asn 81 between WT and the G82S polymorphic variant of RAGE. The published evidence from RAGE produced in bacteria (i.e. lacking N-linked glycosylation) indicates that the G82S polymorphism causes a local change around the mutation site and a more global destabilization of the protein structure, with increased flexibility of the V-domain (39). Thus, substitution of Gly 82 by Ser in G82S RAGE appears to "relax" protein conformation, thereby promoting Asn 81 glycosylation. A key question is how N-linked glycosylation of Asn 81 might translate into the increase in ligand binding and decrease in sRAGE production associated with G82S RAGE. . S100B-induced, NF-B p65 activation in RAGE-transfected HEK293 cells. Transient transfected HEK293 cells expressing WT or various mutant forms of RAGE were exposed to S100B (S) or PBS (P). Negative control cells were transfected with pcDNA3.1 (mock). A, nuclear extracts from transfected cells were analyzed by Western blotting using antibodies specific for NF-B p65, actin, or tubulin as indicated. Increased levels of nuclear NF-B p65 expression following S100B treatment indicate NF-B activation. For individual forms of RAGE, the levels of NF-B p65 expression were variably affected by S100B treatment. The presence of actin (42 kDa) but not tubulin (51 kDa) indicates that nuclear samples were not contaminated by cytosolic proteins. Results are representative of three independent experiments. B, quantitative measures of NF-B activation in RAGE-transfected cells. Values for the nuclear presence of NF-B p65 in each sample were normalized relative to nuclear actin. The figure shows mean values (n ϭ 3) for the ratio of normalized nuclear NF-B p65 present in S100B-treated cells to normalized nuclear NF-B p65 present in PBS control-treated cells. Statistical comparisons with the same ratio in mock-transfected cells are shown (one-way ANOVA; *, p Ͻ 0.05). S100B treatment induced significant NF-B p65 activation in transfected cells expressing WT and G82S RAGE, as well as N81Q or N25Q/G82S mutant forms of RAGE. Error bars, S.E. Some insight can be gained from the structure of the ligand binding domain of RAGE (40,41). The extracellular domain of RAGE contains three immunoglobulin-like domains, including a variable V-domain followed by two constant C-domains, each with a set of conserved cysteine residues (1,42). In the solved structure the V-domain and first C-domain of RAGE form a single structural unit, whereas the second C-domain moves independently from the V-and C-domain unit (40). Furthermore, a hydrophobic cavity (Ile 26 , Ala 28 (38,41), whereas the cationic residues provide points of electrostatic interaction for ligand binding. In fact, the distribution of the cationic residues determines the extent of their contribution to the electrostatic interaction. Those cationic residues located in proximity of the hydrophobic cavity (Lys 37 , Lys 43 , Lys 44 , Arg 48 , Lys 52 , Arg 98 , and Lys 104 ) appear crucial for ligand binding whereas others, on the opposite side with regard to the cavity, have less impact (41). Consequently, RAGE binding to ligand is dependent on hydrophobic and/or electrostatic interactions. The relative contribution from each is ligand dependent. In the RAGE structure, Asn 81 is strategically located adjacent to the cationic region and in close proximity of the hydrophobic cavity. In contrast, Asn 25 is located on the opposite side of the RAGE molecule (Fig. 5). The inference is that N-link glycosylated Asn 81 could play a sentinel role in RAGE ligand binding, controlling the access and binding of ligand to the hydrophobic cavity.
This possibility was considered in assays of RAGE binding to S100B protein ligand. Of relevance is that RAGE binding to S100B ligand is mediated by hydrophobic interaction(s) and involves residues within the hydrophobic cavity (38). As such, these interactions are those most likely to be influenced by the glycosylation of Asn 81 . We established any influence of glycosylation on this hydrophobic interaction(s) with measures of RAGE-mediated NF-B activation. This approach accounted for the signaling capacity of the various mutant forms of RAGE but clearly did not distinguish any form of RAGE that bound S100B without signaling. Nevertheless, these assays establish that S100B can induce NF-B activation through WT and G82S RAGE and the N81Q mutant. A comparative lack of NF-B activation via the N25Q mutant highlights that Asn 25 glycosylation is required for S100B binding/signaling. Given the position of glycosylated Asn 25 in the RAGE structure, this suggests an impact of Asn 25 glycosylation on protein structure, a possibility that requires further investigation. Further comparisons that show S100B signals through the N25Q/G82S mutant (but not via the N25Q mutant) suggest that glycosylation of Asn 81 can compensate for any lack of Asn 25 glycosylation but impacts similarly on hydrophobic interactions between RAGE and S100B ligand. This effect is consistent with the possibility that regions of RAGE involving glycosylated Asn 25 and Asn 81 interact, as previously suggested (22).  (42). The RAGE V-domain contains a hydrophobic cavity (orange) bordered by cationic residues (yellow) and a flexible region (green). The N-linked glycosylation sites, at Asn 25 and Asn 81 , are also shown (purple). The Asn 81 site is located adjacent to the hydrophobic cavity that mediates S100B binding (42), whereas the Asn 25 site is located on the opposite side of the RAGE molecule.