Inhibition of Homophilic Interactions and Ligand Binding of the Receptor for Advanced Glycation End Products by Heparin and Heparin-Related Carbohydrate Structures

Background: Heparin and heparin-related sulphated carbohydrates inhibit ligand binding of the receptor for advanced glycation end products (RAGE). Here, we have studied the ability of heparin to inhibit homophilic interactions of RAGE in living cells and studied how heparin related structures interfere with RAGE–ligand interactions. Methods: Homophilic interactions of RAGE were studied with bead aggregation and living cell protein-fragment complementation assays. Ligand binding was analyzed with microwell binding and chromatographic assays. Cell surface advanced glycation end product binding to RAGE was studied using PC3 cell adhesion assay. Results: Homophilic binding of RAGE was mediated by V1- and modulated by C2-domain in bead aggregation assay. Dimerisation of RAGE on the living cell surface was inhibited by heparin. Sulphated K5 carbohydrate fragments inhibited RAGE binding to amyloid β-peptide and HMGB1. The inhibition was dependent on the level of sulfation and the length of the carbohydrate backbone. α-d-Glucopyranosiduronic acid (glycyrrhizin) inhibited RAGE binding to advanced glycation end products in PC3 cell adhesion and protein binding assays. Further, glycyrrhizin inhibited HMGB1 and HMGB1 A-box binding to heparin. Conclusions: Our results show that K5 polysaccharides and glycyrrhizin are promising candidates for RAGE targeting drug development.


Introduction
The receptor for advanced glycation end products (RAGE) is a multiligand cell surface transmembrane receptor that regulates normal lung physiology, cell motility, and inflammatory reactions. RAGE and some of its ligands (HMGB1, S100A8/S100A9, and amyloid β-peptide) bind to proteoglycans on the cell surface and these interactions regulate the ligand binding and signaling of the receptor [1,2].
RAGE homodimerisation and oligomerisation occurs in cells [3]. Homophilic interactions of RAGE in living cells is demonstrated with fluorescence resonance energy transfer and bioluminescence resonance energy transfer methods [4,5]. Formation of RAGE homodimers and oligomers in solution Salt concentrations in glycyrrhizin and glucose stocks and in buffer controls were equalized. For PC3 adhesion studies, glycyrrhizin or glucose were dissolved in a NaOH solution as described [38]. Heparin was from Sigma-Aldrich.

Bead Aggregation Assay
Protein A coated bead aggregation mediated by recombinant Ig-fusion proteins was studied as described [33]. In brief, monomeric fluorescent protein A coated beads were incubated with 10 µg/mL of different sRAGE-Ig forms and aggregation was monitored using fluorescence microscopy (Olympus, Tokyo, Japan). Pictures were taken after 60 min bead aggregation. The extent of bead aggregation is represented by the index N t /N 0 . N t and N 0 are the total number of particles at incubation times t and 0, respectively.

Live-Cell Protein-Fragment Complementation Assay (PCA)
Full-length RAGE was cloned to plasmids containing complementary fragments of humanized Gaussia (hGLuc) luciferase and livecell protein-fragment complementation assay was done as described [39]. In brief, white-walled 96-well plates (Perkin Elmer, Waltham, MA, USA) were coated with poly-L-lysine (Sigma), and 10,000 cells per well were seeded. Transfection of the phGLuc-RAGE plasmids (100 ng of DNA per well) was performed after 24 h, while detection of PCA signal was measured 48 h post-transfection. Culture media was changed 30 min before measurements to serum and phenol red-free Dulbecco's Modified Eagle Medium (DMEM) (Invitrogen, Carlsbad, CA, USA). Luminescence was measured with Victor 3 plate reader (Perkin Elmer) following injection of native coelenterazine, the substrate for the GLuc (NanoLight Technology, Pinetop, AZ, USA). At least four replicate wells per condition were analyzed and a minimum of three independent experiments were performed.

Binding of Amyloid Beta 1-42 Peptide to RAGE
Ig-fusion protein of sRAGE in a concentration of 5 mg/mL was bound to protein A (1 mg/mL, Sigma-Aldrich, St. Louis, MO, USA) coated wells (Maxisorb ELISA plates, Sigma-Aldrich). Unbound proteins were washed away, and the wells were blocked with BSA (Sigma-Aldrich). Biotinylated amyloid β 1-42 peptide (20 nM, American Peptide, Sunnyvale, CA, USA) or biotinylated HMGB1 (20 nM) were bound to the wells in the presence or absence of inhibitors, and unbound ligands were washed away. Bound biotinylated ligands were detected with horse radish peroxidase conjugated streptavidin (Sigma-Aldrich) and peroxidase substrate (Sigma-Aldrich). The color was developed using o-phenylenediamine dihydrochloride peroxidase substrate (Sigma-Aldrich). Absorbance at 490 nm was measured.

RAGE Binding to AGE-BSA
Plastic Maxisorb microwells (Sigma-Aldrich) were coated with 20 µg/mL of AGE-BSA [35] and blocked with BSA. sRAGE-Ig protein (0.1 µg/mL) was incubated in wells in the presence of various amounts of glycyrrhizin. Bound proteins were quantified using horse radish peroxidase conjugated goat anti-human IgG (Sigma-Aldrich). The color was developed with peroxidase substrate. Absorbance was measured at 450 nm.

Cell Adhesion Studies
PC3 cell binding assay was carried out as described [40]. Briefly, Maxisorp (Sigma-Aldrich) wells were coated with RAGE V 1 -Fc, collagen, laminin, or fibronectin [10 µg/mL in phosphate buffered saline (PBS)]. Coated wells were washed with PBS, blocked with 0.5% BSA-PBS, and washed once with PBS. Ten microliters of Iscoves's Modified Dulbecco's Media (IMDM, Invitrogen)-0.5% BSA containing 11× concentration of possible inhibitors was added to wells, and 1 × 10 5 PC3 cells in 100 µL of IMDM-0.5% BSA was added to wells. Cells were adhered for 30 min at 37 • C and unbound cells were washed away with warm IMDM-0.5% BSA (6 times) and bound cells were quantitated with intracellular acidic phosphatase assay [26]. EDTA, heparin, and glucose served as control inhibitors in cell adhesion studies.

HMGB1 Binding to Heparin
Heparin-Sepharose (GE Healthcare, Chicago, IL, USA) affinity chromatography of HMGB1 was carried out as previously described using PBS as a chromatography buffer [26]. HMGB1 was loaded to the column in the presence of different concentrations of glycyrrhizin, the column was washed with PBS, and bound protein was eluted with PBS containing 1.5 M NaCl. In some experiments, HMGB1 or thrombin were loaded onto the column in PBS and eluted with increasing amounts of glycyrrhizin in PBS. HMGB1 in elution fractions was detected with an anti-HMGB1 immunoblot [32], and thrombin was detected in a dot blot assay using Coomassie blue staining. The optical density (OD) values of bands or dots were measured as described [35].
The effect of glycyrrhizin on HMGB1 A-box binding to heparin-Sepharose was studied using heparin-Sepharose precipitation assay. The A-box was diluted with PBS (4 µg/mL) and an equal volume of 50% heparin-Sepharose slurry in PBS was added. Glycyrrhizin or glucose were at concentrations of 0-3 mM. Samples were incubated in rotation at 4 • C for 1 h, heparin-Sepharose was pelleted and washed with PBS. Bound protein was eluted with 30 µL of hot reducing SDS-PAGE sample buffer, and samples were analyzed using an anti-HMGB1 (anti peptide III, [32]) Western blotting assay. The OD values of A-box bands were measured.
Precipitation assays of full length HMGB1 protein. HMGB1 was diluted with PBS (25 µg/mL) and divided to 100 µL aliquots, and various amounts of glycyrrhizin or glucose were added. One-hundred microliters of 50% heparin-Sepharose slurry in PBS was added to aliquots, and the samples were incubated in rotation at 4 • C for 1 h. Heparin-Sepharose was washed once with PBS, and bound proteins were eluted with 30 µL of hot reducing SDS-PAGE sample buffer. The samples were run on SDS-PAGE and the gels were stained with Coomassie blue. The OD values of HMGB1 bands were measured.

Data Analyses
Analyses were performed using Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA), GraphPad Prism Software 4.0 (GraphPad Software Inc., La Jolla, CA, USA), and PSI-Plot V 7.0 (Poly Software International, Pearl River, NY, USA). Student's t-test was used to compare group means. p-values < 0.05 were considered as significant.

Nature of Homophilic RAGE Interactions
Homophilic interactions of RAGE are mediated by multiple mechanisms involving both extra cellular and cytoplasmic domains [2][3][4][5][6][7]. Wei et al. have described non-covalent homophilic interactions occur between V 1 -domains and dimerization by covalent homophilic bonding via interchain disulphide bridge between C 2 -domains [3]. In addition, homophilic binding of soluble recombinant RAGE protein have been shown to require high concentrations of protein and the presence of divalent metal ions [41,42]. Here, we studied whether the homophilic interactions of recombinant sRAGE occur in low protein concentrations using bead aggregation assay. Furthermore, we studied whether the homophilic interactions of full length RAGE on a living cell surface can be inhibited by heparin.
Recombinant sRAGE-Ig protein and Ig-fragments bearing RAGE domains used in this study are dimeric proteins. sRAGE-Ig and V 1 -Ig induced bead aggregation in protein-A coated bead suspension, suggesting that dimeric RAGE is able to form oligomeric structures in low concentrations (10 µg/mL) and in the absence of other exogenous ligands. The V 1 -domain was necessary and sufficient for  Next, we studied the homophilic interactions of RAGE in living cells using a protein-fragment complementation assay (PCA). In this assay, cells are transfected with full-length RAGE-cytoplasmic Gaussia luciferase-fragment chimeras that become chemiluminescent when dimerised. Amyloid precursor protein (APP) that forms homodimers served as a positive control ( Figure 2A). RAGE self-association occurred in living cells without any exogenous macromolecular ligands. Self-association was partially inhibited by heparin ( Figure 2B). Our results indicate that the soluble dimeric ectodomain of RAGE, at low protein concentration and in the absence of exogenous ligands, can form oligomers. Further, to our knowledge, this is the first time that it has been shown that RAGE self-association in the living cells can be inhibited by heparin. As heparin does not penetrate through the plasma membrane, the inhibition most probably occurs at the cell surface. Xu et al. have shown that sRAGE dimerization is enhanced by heparin and dimerization at the cell surface is inhibited by heparanase treatment [2]. Our results suggest that soluble heparin may interfere with cell surface RAGE-heparan sulphate interactions that lead to dissociation of RAGE homodimer ( Figure 3A). Next, we studied the homophilic interactions of RAGE in living cells using a protein-fragment complementation assay (PCA). In this assay, cells are transfected with full-length RAGE-cytoplasmic Gaussia luciferase-fragment chimeras that become chemiluminescent when dimerised. Amyloid precursor protein (APP) that forms homodimers served as a positive control (Figure 2A). RAGE selfassociation occurred in living cells without any exogenous macromolecular ligands. Self-association was partially inhibited by heparin ( Figure 2B). Our results indicate that the soluble dimeric ectodomain of RAGE, at low protein concentration and in the absence of exogenous ligands, can form oligomers. Further, to our knowledge, this is the first time that it has been shown that RAGE selfassociation in the living cells can be inhibited by heparin. As heparin does not penetrate through the plasma membrane, the inhibition most probably occurs at the cell surface. Xu et al. have shown that sRAGE dimerization is enhanced by heparin and dimerization at the cell surface is inhibited by heparanase treatment [2]. Our results suggest that soluble heparin may interfere with cell surface RAGE-heparan sulphate interactions that lead to dissociation of RAGE homodimer ( Figure 3A).

Sulphated K5 Polysaccharides Inhibit RAGE Binding to Amyloid β-Peptide and HMGB1
Heparin that has a high content of sulphated L-iduronic acid units binds to RAGE and inhibits RAGE-ligand binding and dimerization ( Figure 2B and [1]). Epimerization of carbon C5 of Dglucuronic acid to create L-iduronic acid is not required for RAGE inhibition by sulphated oligosaccharides [43]. Capsular polysaccharide K5 derived from E. coli contains N-acetylglucosamine and glucuronic acid structures [44]. Here, we studied whether semi-synthetic non-epimerized sulphated K5 polysaccharides inhibit RAGE-ligand binding. O-and N-sulphated K5 oligosaccharides were able to inhibit RAGE binding to amyloid β-peptide and HMGB1 (Table 1) at similar concentrations that have previously been shown for heparin [1,45]. The minimum size of oligosaccharides required for strong inhibition was 3100-5800 Da. Sulphated disaccharides derived from chondroitin and dermatan sulphates did not inhibit RAGE-ligand binding (data not shown). These results are in line with the results from previous studies where the polymeric structure of sulphated carbohydrates was shown to be essential for inhibition of RAGE-ligand binding and RAGE dimerization [1,2]. Our results indicate that E. coli derived K5 polysaccharides can be used to design new RAGE targeting drug molecules ( Figure 3A).

Sulphated K5 Polysaccharides Inhibit RAGE Binding to Amyloid β-Peptide and HMGB1
Heparin that has a high content of sulphated L-iduronic acid units binds to RAGE and inhibits RAGE-ligand binding and dimerization ( Figure 2B and [1]). Epimerization of carbon C5 of D-glucuronic acid to create L-iduronic acid is not required for RAGE inhibition by sulphated oligosaccharides [43]. Capsular polysaccharide K5 derived from E. coli contains N-acetylglucosamine and glucuronic acid structures [44]. Here, we studied whether semi-synthetic non-epimerized sulphated K5 polysaccharides inhibit RAGE-ligand binding. O-and N-sulphated K5 oligosaccharides were able to inhibit RAGE binding to amyloid β-peptide and HMGB1 (Table 1) at similar concentrations that have previously been shown for heparin [1,45]. The minimum size of oligosaccharides required for strong inhibition was 3100-5800 Da. Sulphated disaccharides derived from chondroitin and dermatan sulphates did not inhibit RAGE-ligand binding (data not shown). These results are in line with the results from previous studies where the polymeric structure of sulphated carbohydrates was shown to be essential for inhibition of RAGE-ligand binding and RAGE dimerization [1,2]. Our results indicate that E. coli derived K5 polysaccharides can be used to design new RAGE targeting drug molecules ( Figure 3A).

Glycyrrhizin and Heparin Inhibit AGE-RAGE -Interactions
Glycyrrhizin has two non-sulphated D-glucuronic acids in its structure and it inhibits RAGE binding to HMGB1 [46]. The inhibition is mainly due to binding of glycyrrhizin to HMGB1. However, glycyrrhizin shows minimal binding on RAGE. In addition, glycyrrhizin interferes HMGB1 binding to CXCR4, Toll-like receptor 2, and Toll-like receptor 4 [25,28,38]. Glycyrrhizin might be a potential molecule for RAGE targeting drug development because it can be easily modified [21,22]. Here, we tested whether glycyrrhizin affects homophilic interactions of RAGE in cell culture conditions using the PCA assay. Surprisingly, glycyrrhizin inhibited full-length luciferase activity in cells, that is, it was incompatible with the assay, thereby preventing further studies with glycyrrhizin in the PCA assay (data not shown).
AGE-BSA binding to RAGE induces receptor dimerization and signaling [47]. Here, we studied whether glycyrrhizin interferes with binding of RAGE to AGE-BSA. In a microwell binding assay, 300 µM glycyrrhizin inhibited RAGE binding to AGE-BSA coated microwells (Figure 4). The effect of glycyrrhizin on RAGE-AGE interactions was also tested in cell adhesion assay. PC3 cells express endogenously synthetized AGEs on their surface, which mediate cell binding to RAGE [40,48,49]. PC3 cells adhered to RAGE V 1 -domain, but not to C 1 -or C 2 -domains ( [40]; data not shown). Glycyrrhizin at 1 mM concentration inhibited PC3 cell binding to RAGE V 1 -domain, but not to the control proteins fibronectin, laminin, or collagen. Furthermore, heparin inhibited cell binding to both RAGE V 1 -domain and fibronectin, but not to laminin and collagen. In contrast, EDTA inhibited cell binding to fibronectin, laminin, and collagen, but not to RAGE V 1 -domain ( Figure 5). The finding that the cell binding to RAGE occurs independently of divalent cations excludes the role of Mac-1 integrin in the binding [45]. Our results indicate that AGE binding to RAGE can be inhibited by glycyrrhizin and heparin ( Figure 3A).

Glycyrrhizin and Heparin Inhibit AGE-RAGE -Interactions
Glycyrrhizin has two non-sulphated D-glucuronic acids in its structure and it inhibits RAGE binding to HMGB1 [46]. The inhibition is mainly due to binding of glycyrrhizin to HMGB1. However, glycyrrhizin shows minimal binding on RAGE. In addition, glycyrrhizin interferes HMGB1 binding to CXCR4, Toll-like receptor 2, and Toll-like receptor 4 [25,28,38]. Glycyrrhizin might be a potential molecule for RAGE targeting drug development because it can be easily modified [21,22]. Here, we tested whether glycyrrhizin affects homophilic interactions of RAGE in cell culture conditions using the PCA assay. Surprisingly, glycyrrhizin inhibited full-length luciferase activity in cells, that is, it was incompatible with the assay, thereby preventing further studies with glycyrrhizin in the PCA assay (data not shown).
AGE-BSA binding to RAGE induces receptor dimerization and signaling [47]. Here, we studied whether glycyrrhizin interferes with binding of RAGE to AGE-BSA. In a microwell binding assay, 300 μM glycyrrhizin inhibited RAGE binding to AGE-BSA coated microwells (Figure 4). The effect of glycyrrhizin on RAGE-AGE interactions was also tested in cell adhesion assay. PC3 cells express endogenously synthetized AGEs on their surface, which mediate cell binding to RAGE [40,48,49]. PC3 cells adhered to RAGE V1-domain, but not to C1-or C2-domains ( [40]; data not shown). Glycyrrhizin at 1 mM concentration inhibited PC3 cell binding to RAGE V1-domain, but not to the control proteins fibronectin, laminin, or collagen. Furthermore, heparin inhibited cell binding to both RAGE V1-domain and fibronectin, but not to laminin and collagen. In contrast, EDTA inhibited cell binding to fibronectin, laminin, and collagen, but not to RAGE V1-domain ( Figure 5). The finding that the cell binding to RAGE occurs independently of divalent cations excludes the role of Mac-1 integrin in the binding [45]. Our results indicate that AGE binding to RAGE can be inhibited by glycyrrhizin and heparin ( Figure 3A).

Glycyrrhizin Inhibits HMGB1 Heparin Binding
HMGB1 signals to cells via mechanisms that are, at least partly, dependent on RAGE-heparan sulphate interactions [1,27,28,50]. Whether binding of HMGB1 to sulphated glycosaminoglycans is necessary to RAGE-HMGB1 signaling is currently poorly understood [2,51]. Described heparinbinding sites of HMGB1 are located at the linker sequence between A-and B-boxes and at the amino terminal part of the A-box [2,52,53]. Interestingly, soluble HMGB1 A-box, which lacks RAGE binding site but has a heparin-binding site, inhibits cell migration [54][55][56][57]. This suggests that HMGB1 binding to HSPGs might be necessary to RAGE-HMGB1 complex signaling. Here, we studied the effect of

Glycyrrhizin Inhibits HMGB1 Heparin Binding
HMGB1 signals to cells via mechanisms that are, at least partly, dependent on RAGE-heparan sulphate interactions [1,27,28,50]. Whether binding of HMGB1 to sulphated glycosaminoglycans is necessary to RAGE-HMGB1 signaling is currently poorly understood [2,51]. Described heparinbinding sites of HMGB1 are located at the linker sequence between A-and B-boxes and at the amino terminal part of the A-box [2,52,53]. Interestingly, soluble HMGB1 A-box, which lacks RAGE binding site but has a heparin-binding site, inhibits cell migration [54][55][56][57]. This suggests that HMGB1 binding to HSPGs might be necessary to RAGE-HMGB1 complex signaling. Here, we studied the effect of glycyrrhizin on HMGB1 heparin binding. First, we tested whether A-box binds to heparin in a similar manner to the full-length HMGB1. The A-box bound to a heparin-Sepharose in column chromatography and was eluted with the same NaCl concentration (~0.7 M) as full-length HMGB1 (data not shown). This suggests that the A-box lacking heparin binding cationic amino acid residues in a linker sequence is still able to bind to heparin.
Next, we tested whether glycyrrhizin can inhibit heparin binding in heparin-Sepharose chromatography. The excess of glycyrrhizin prevented HMGB1 binding to the chromatography column ( Figure 6A). Further, a heparin-Sepharose column bound HMGB1 was eluted with a glycyrrhizin gradient at the same concentration of glycyrrhizin as the control protein thrombin ( [58]; data not shown). Next, we studied the effect of glycyrrhizin on heparin binding of full-length HMGB1 and A-box using a heparin-Sepharose precipitation assay. Glycyrrhizin inhibited binding of both the full-length HMGB1 and the A-box ( Figure 6B,C). These results indicate that the A-box and the full length HMGB1 bind to heparin in a similar manner, and that their binding to heparin is inhibited by millimolar concentrations of glycyrrhizin. It is tempting to speculate that glycyrrhizin might inhibit RAGE signaling via interfering HMGB1-heparan sulphate interactions ( Figure 3B). glycyrrhizin on HMGB1 heparin binding. First, we tested whether A-box binds to heparin in a similar manner to the full-length HMGB1. The A-box bound to a heparin-Sepharose in column chromatography and was eluted with the same NaCl concentration (~0.7 M) as full-length HMGB1 (data not shown). This suggests that the A-box lacking heparin binding cationic amino acid residues in a linker sequence is still able to bind to heparin.
Next, we tested whether glycyrrhizin can inhibit heparin binding in heparin-Sepharose chromatography. The excess of glycyrrhizin prevented HMGB1 binding to the chromatography column ( Figure 6A). Further, a heparin-Sepharose column bound HMGB1 was eluted with a glycyrrhizin gradient at the same concentration of glycyrrhizin as the control protein thrombin ( [58]; data not shown). Next, we studied the effect of glycyrrhizin on heparin binding of full-length HMGB1 and A-box using a heparin-Sepharose precipitation assay. Glycyrrhizin inhibited binding of both the full-length HMGB1 and the A-box ( Figure 6B,C). These results indicate that the A-box and the full length HMGB1 bind to heparin in a similar manner, and that their binding to heparin is inhibited by millimolar concentrations of glycyrrhizin. It is tempting to speculate that glycyrrhizin might inhibit RAGE signaling via interfering HMGB1-heparan sulphate interactions ( Figure 3B).
Our results show that RAGE dimerization occurring in living cells can be inhibited by heparin, RAGE ligand binding can be inhibited by sulphated K5 polysaccharides, and RAGE-AGE interactions and HMGB1 binding to heparin can be inhibited by glycyrrhizin. We conclude that sulphation of D-glucuronic acid and L-iduronic acid structures offer a promising strategy for the development of RAGE targeting molecules. Funding: A.R. was funded by the Foundation of the Finnish Anti-Tuberculosis Association.

Conflicts of Interest:
The authors declare no conflict of interest. Our results show that RAGE dimerization occurring in living cells can be inhibited by heparin, RAGE ligand binding can be inhibited by sulphated K5 polysaccharides, and RAGE-AGE interactions and HMGB1 binding to heparin can be inhibited by glycyrrhizin. We conclude that sulphation of D-glucuronic acid and L-iduronic acid structures offer a promising strategy for the development of RAGE targeting molecules.