The receptor for advanced glycation end products (RAGE) is a cellular binding site for amphoterin. Mediation of neurite outgrowth and co-expression of rage and amphoterin in the developing nervous system.

The receptor for advanced glycation end products (RAGE), a newly-identified member of the immunoglobulin superfamily, mediates interactions of advanced glycation end product (AGE)-modified proteins with endothelium and other cell types. Survey of normal tissues demonstrated RAGE expression in situations in which accumulation of AGEs would be unexpected, leading to the hypothesis that under physiologic circumstances, RAGE might mediate interaction with ligands distinct from AGEs. Sequential chromatography of bovine lung extract identified polypeptides with Mr values of ≈12,000 (p12) and ≈23,000 (p23) which bound RAGE. NH-terminal and internal protein sequence data for p23 matched that reported previously for amphoterin. Amphoterin purified from rat brain or recombinant rat amphoterin bound to purified sRAGE in a saturable and dose-dependent manner, blocked by anti-RAGE IgG or a soluble form of RAGE (sRAGE). Cultured embryonic rat neurons, which express RAGE, displayed dose-dependent binding of I-amphoterin which was prevented by blockade of RAGE using antibody to the receptor or excess soluble receptor (sRAGE). A functional correlate of RAGE-amphoterin interaction was inhibition by anti-RAGE F(ab′) and sRAGE of neurite formation by cortical neurons specifically on amphoterin-coated substrates. Consistent with a potential role for RAGE-amphoterin interaction in development, amphoterin and RAGE mRNA/antigen were co-localized in developing rat brain. These data indicate that RAGE has physiologically relevant ligands distinct from AGEs which are likely, via their interaction with the receptor, to participate in physiologic processes outside of the context of diabetes and accumulation of AGEs.

Incubation of proteins or lipids with aldose sugars results in nonenzymatic glycation and oxidation (1)(2)(3)(4)(5)(6)(7). Following formation of the reversible early glycation products, Schiff bases and Amadori products, further complex molecular rearrangements result in irreversible advanced glycation end products (AGEs). 1 Factors favoring nonenzymatic glycation include delayed protein turnover, as in amyloidoses, accumulation of macromolecules with high lysine content, and situations with elevated glucose levels, as in diabetes. AGE formation occurs during normal aging, and at an accelerated rate in diabetes, in which their accumulation in the plasma and vessel wall has been speculated to underlie the pathogenesis of vasculopathy (1,2,4).
One of the principal means through which AGEs impact on cellular elements is through interaction with cellular binding proteins. Although there are several possible cell-associated polypeptides with which AGEs might interact (8,9), our work has focussed on the receptor for AGEs (RAGE), as its expression in endothelium, vascular smooth muscle, mononuclear phagocytes, and the central nervous system suggests strategic loci for interaction with the glycated ligands (10,11). The potential pathophysiologic relevance of AGE-RAGE interaction was emphasized by studies demonstrating that blockade of RAGE prevented multiple AGE-induced perturbations of cellular functions. For example, AGE-stimulated mononuclear phagocyte migration and activation, endothelial expression of vascular cell adhesion molecule-1, and increased monolayer permeability were prevented by blocking AGE interaction with RAGE (12,13). Consistent with the latter data in cell culture, in vivo studies have shown RAGE to mediate the early and rapid removal of AGEs from the intravascular space, AGE induction of vascular oxidant stress and vascular hyperpermeability in diabetic animals and rodents infused with AGEs (14 -17). 2 These data emphasize a potential role for AGE-RAGE interactions in pathologic states, and have led us to assess RAGE expression in vasculopathies, especially that associated with diabetes.
During the course of tissue surveys to assess RAGE distribution, it became evident that expression of the receptor occurred in early development, especially in the central nervous system; i.e. in situations distinct from those in which AGEs might be present. RAGE is a member of the immunoglobulin superfamily of cell surface molecules (19), and the extracellular domain consists of one putative "V" type followed by two putative "C" type domains, bearing closest homology to neural cell adhesion molecule (19). In view of previous studies with other immunoglobulin-like receptors, such as intercellular adhesion molecule-1, in which one receptor has several pathophysiologically relevant ligands (20), we considered the hypothesis that AGEs might be accidental ligands for a receptor which had other functions. By analogy with other immunoglobulin superfamily members, we speculated that RAGE might participate in cell-cell or cell-matrix interactions, or, perhaps, might function as a cytokine or growth factor receptor. Toward this end, we sought to define putative natural ligands for RAGE which were not AGEs. In this study we report the identification of amphoterin as a ligand for RAGE: amphoterin binds RAGE with higher affinity than AGEs; RAGE serves as the binding site for amphoterin on rat embryonic cortical neurons; amphoterin-RAGE interaction promotes neurite outgrowth in cell culture; and, RAGE and amphoterin are co-expressed in developing rat neuroepithelium. Although future in vivo studies will be required to determine the contribution of amphoterin-RAGE interaction in neuronal development, these studies provide a first step in identifying novel functions for RAGE distinct from its role as a receptor for AGEs in the vasculature.

EXPERIMENTAL PROCEDURES
Preparation of AGEs-Bovine serum albumin (fraction V; Sigma) was glycated by incubation with glucose (0.5 M) for 6 weeks at 37°C. Glycated protein was characterized based on fluorescence and binding to cultured endothelial cells and mononuclear phagocytes, as described previously (10). Control, nonglycated albumin consisted of the same initial preparations of albumin incubated at 37°C in the same manner, except that no aldose was present.
Fractions from the heparin column were screened for RAGE binding activity following their adsorption to MaxiSorp microtiter wells (Nunc, Naperville, IL) in bicarbonate/carbonate buffer (pH 9.6) for 16 h at 4°C. After blocking excess binding sites in the wells with phosphate-buffered saline containing bovine serum albumin (1%), a binding assay with 125 I-RAGE was performed. For these studies, RAGE was purified to homogeneity from bovine lung, as described (10), and radioiodinated by the lactoperoxidase method using Enzymobeads (Bio-Rad) according to the manufacturer's instructions. The final specific radioactivity of the tracer was 1,000 cpm/ng, and 125 I-RAGE was Ͼ95% precipitable in trichloroacetic acid (20%), migrated as a single band on SDS-PAGE (comigrating with the unlabeled protein), and bound AGEs. 125 I-RAGE was added to wells previously coated with fractions from the heparin column for 3 h at 37°C in phosphate-buffered saline containing bovine serum albumin (0.2%). At the end of the incubation period, wells were washed rapidly (10 s/wash and 0.3 ml/wash) with Hank's balanced salt solution (Life Technologies, Inc.). Bound radioactivity was eluted with NaCl (2 M) and counted in a ␥-counter (Pharmacia LKB, Gaithersburg, MD). Fractions containing RAGE binding activity, identified in the 0.5 M NaCl eluate, were dialyzed versus phosphate-buffered saline containing octyl-␤-glucoside (0.1%; final pH 7.4) and applied to Affi-Gel 10 (Bio-Rad), to which had been bound purified bovine RAGE (2 mg/ml resin), according to the manufacturer's instructions. After 16 h of incubation at 4°C, the resin was washed with 10 bed volumes of buffer (phosphate-buffered saline containing octyl-␤-glucoside, 0.1%), and eluted by high salt (NaCl, 2 M). Fractions containing binding activity for 125 I-RAGE, based on the competitive binding assay above, were dialyzed versus phosphate-buffered saline containing octyl-␤-glucoside (0.1%) and subjected to nonreduced SDS-PAGE (10%). Protein on the gel was either visualized by silver staining (Bio-Rad), or gels were sliced (2 mm each) and individual slices were eluted by incubation in acetate buffer (pH 8.0) for 4 h at 4°C. The mixture was then centrifuged to pellet debris, and supernatants were tested for RAGE binding activity.
Two positive fractions were identified, eluted from the gel, and subjected to nonreduced SDS-PAGE (12%). Protein on the gel was visualized by either silver or Coomassie Blue staining to identify bands. Two protein bands with M r Ϸ 23,000 and Ϸ 12,000 were visualized and sequence analysis performed as follows.
Amino-terminal Sequence Analysis-The gel bands were eluted from the SDS-PAGE gels using previously described methods (21). This method was modified by increasing buffer volumes to accommodate the larger gel volume, and adding a final wash with two 0.1-ml aliquots of guanidine (5.0 M), urea (5.0 M), trifluoroacetic acid (0.2%), acetonitrile (10%), and Zwittergent 3-08 (1.0%) (Calbiochem, San Diego, CA) to insure that protein was completely washed from the filter. Automated Edman degradation was carried out using either an HP-G1005A sequencer (Hewlett Packard Analytical Instruments, Palo Alto, CA) or a 477A sequencer (Perkin Elmer-Applied Biosystems, Foster City, CA).
Internal Sequence Analysis-To obtain internal sequence, the gel bands were treated as above for elution, except that the extraction buffer had half the usual amount of SDS (21). One microgram of endoproteinase Lys-C (Boehringer Mannheim) was added and the sample incubated overnight. Separation of the fragments was carried out by HPLC on a Vydac C 18 column. A linear gradient of 10 -40% acetonitrile in trifluoroacetic acid (0.1%) over 35 min was used to elute the fragments. Fractions were collected at 2-min intervals and absorbance monitored at 214 nm. Fractions that corresponded to chromatographic peaks were then subjected to sequence analysis. In order to obtain additional internal sequence, the eluate from the RAGE-Affi-Gel 10 column was purified by HPLC using a 1% per minute gradient from 10 to 60% acetonitrile in trifluoroacetic acid (0.1%) to elute the material from a Vydac C 4 column. Fractions were collected at 1-min intervals. Initially, one-fourth of the available sample was run and peaks identified by NH 2 -terminal sequence analysis. The remainder of the material was then run and the 23-kDa containing fraction digested with cyanogen bromide to obtain the internal sequence. To digest with cyanogen bromide (Pierce) the 23-kDa containing fraction was diluted 1:1 with 100 mg/ml cyanogen bromide in HCl (0.2 N) and incubated overnight. The digest was fractionated by microbore HPLC (Michrom Bioresources, Auburn, CA) on a 1 mm ϫ 50-mm PLRP-S column (Polymer Laboratories, Ltd., Church Stretton, UK). The gradient utilized was 2% per minute from acetonitrile (5-75%) in trifluoroacetic acid (0.1%) and fractions were collected at half-minute intervals. Absorbance was monitored at 214 nm and fractions that corresponded to chromatographic peaks were then subjected to sequence analysis.
Data Base Search-The NH 2 -terminal sequence was used to query the PIR and SWIS-PROT protein data bases using the Genetics Computer Group Sequence Analysis Software package (22).
Purification of Rat Amphoterin, Rat RAGE, and Preparation of Antisera-Amphoterin was purified from the brains of 10-day-old Wistar rats as described (23) by sequential chromatography on heparin-Sepharose and Affi-Gel Blue (Bio-Rad). A single band, M r Ϸ 30,000, was identified on nonreduced SDS-PAGE (12%), eluted and subjected to NH 2 -terminal sequence analysis and identified as amphoterin. Rat RAGE was purified to homogeneity from lung powder (Sigma) using similar methods described for bovine RAGE (10). Immunoreactivity with anti-RAGE IgG (see below) and NH 2 -terminal and internal sequencing indicated identity to the predicted sequence for rat RAGE based on that deduced from the rat cDNA. 3 Based on M r Ϸ 32,000 and comparison with RAGE purified from bovine lung, rat lung RAGE also most likely represents the amino-terminal two-thirds of the molecule (i.e. the extracellular domain; Ref. 19) and is identified as soluble RAGE or sRAGE. Rat sRAGE bound AGE albumin in a dose-dependent and specific manner analogous to bovine sRAGE (data not shown). Polyclonal antiserum to rat RAGE was prepared in rabbits by standard procedures (24), and monospecific immune IgG was purified by chromatography on Protein A (Schleicher and Schuell) according to the manufacturer's instructions. F(abЈ) 2 fragments were prepared from IgG using a kit from Pierce according to the manufacturer's instructions. Nonimmune IgG and F(abЈ) 2 fragments were similarly made from sera derived from rabbits not immunized with RAGE. Polyclonal antiserum to rat amphoterin was prepared in chickens and immune IgG was purified by ammonium sulfate precipitation (25). Enzyme-linked immunosorbent assay for AGE antigen was performed with affinity purified anti-AGE antibody as described previously (14).
Cloning and Expression of Rat Amphoterin-A baculovirus expression plasmid coding for rat amphoterin was prepared as follows. A DNA fragment coding for amphoterin was obtained from rat lung cDNA by polymerase chain reaction (GeneAmp, Perkin-Elmer); primers used were 5Ј-CTAAACATGGGCAAAGGAGAT-3Ј and 5Ј-CGTAGAAC-CAACTTATTCATC-3Ј. Sequencing of the polymerase chain reaction product by the dideoxy chain termination method (26) revealed the same sequence as that published for rat brain amphoterin (27). The polymerase chain reaction product was subcloned into the pCR TM II vector (Invitrogen, San Diego, CA) and the EcoRI fragment of the resultant plasmid was cloned into the pBacPAK8 vector (Clontech, Palo Alto, Ca) under control of AcMNPV polyhedrin promoter. Baculovirus expression of recombinant rat amphoterin was performed by co-transfecting the plasmid pBacPAK8/amphoterin with a linearized BacPAK6 viral DNA (Clontech) into Spodoptera frugiperda (Sf9) cells according to the manufacturer's instructions. Recombinant plaques were identified and purified by their ␤-galactosidase-negative phenotype. Recombinant amphoterin was then purified to homogeneity by sequential chromatography on heparin-Sepharose and Affi-Gel blue. Recombinant amphoterin was identified by immunoblotting with monospecific anti-rat amphoterin IgG and mobility on SDS-PAGE.
Binding of Radiolabeled Amphoterin to Purified RAGE-Rat RAGE (2.5 g) dissolved in bicarbonate/carbonate buffer (pH 9.6) was incubated for 16 h at 4°C in MaxiSorp (Nunc) microtiter wells, wells were washed four times with washing buffer (phosphate-buffered saline containing Tween 20, 0.02%), and blocked with phosphate-buffered saline containing bovine serum albumin (1%). Purified rat brain amphoterin and purified recombinant amphoterin were radiolabeled with Na 125 I by the lactoperoxidase method, as above, to final specific activities of 4 ϫ 10 4 and 5 ϫ 10 4 cpm/ng of protein, respectively. Tracers were Ͼ95% precipitable in trichloroacetic acid (20%) and migrated on SDS-PAGE as single bands with M r Ϸ 30,000. Wells with adsorbed RAGE were incubated for 2 h at 37°C with the indicated concentration of 125 Iamphoterin, dissolved in phosphate-buffered saline containing bovine serum albumin (1%), either alone or in the presence of an 100-fold excess of unlabeled material. Following the incubation period, wells were washed four times rapidly in washing buffer, as above, and eluted with NaCl (2 M) for 10 min. Specific binding was defined as the total binding minus binding in the presence of excess unlabeled material (i.e. nonspecific binding) and analyzed by the methods of Klotz and Hunston (28). Where indicated, wells were preincubated with anti-RAGE IgG or nonimmune IgG prior to incubation with 125 I-amphoterin (2-3 nM) or 125 I-amphoterin was preincubated with various concentrations of rat sRAGE for 2 h at 37°C in phosphate-buffered saline containing bovine serum albumin (1%) prior to incubation with RAGE adsorbed to the wells. In certain experiments, wells were preincubated with AGE albumin or native albumin for 2 h at 37°C prior to the addition of 125 I-amphoterin.
Binding of Radiolabeled Amphoterin to Rat Cortical Neurons-Cortical neurons were isolated from the brains of 17-day-old embryonic Wistar rats according to established methods (23). Cells (10 5 cells/well) in Dulbecco's minimal essential medium (Life Technologies, Inc.), containing fetal bovine serum (10%; Gemini, Calabasas, CA) were plated in 96-well plates (Corning) previously coated with poly-L-lysine (50 g/ well, Sigma). The cell population was Ͼ90% neuronal based on immunostaining with anti-neurofilament antibody (Sigma). Two days later, cells were washed extensively in phosphate-buffered saline, fixed in paraformaldehyde (2%; fixation was required to prevent detachment of cells during the binding assay), and again washed four times with phosphate-buffered saline. Fixed cultures were then placed in phosphate-buffered saline containing bovine serum albumin (1%) and incubated for 2 h at 37°C with 125 I-amphoterin either alone or in the presence of an 100-fold excess of unlabeled amphoterin. Wells were then washed 4 times with phosphate-buffered saline and cell-associated 125 I-amphoterin was eluted with phosphate-buffered saline containing Nonidet P-40 (1%). Where indicated, cultured neurons were preincubated with anti-RAGE IgG, nonimmune IgG, AGE albumin, or native albumin, or 125 I-amphoterin was preincubated with sRAGE, as above. Nonspecific binding was Ϸ25% of total binding, and binding data were analyzed as above. In order to approximate equilibrium binding conditions, the incubation time was chosen to allow 125 I-amphoterin binding to cultured neurons to reach an apparent maximum, and the washing time was sufficiently short so that Ͻ5-10% of bound ligand would dissociate. Amphoterin produced endogenously was removed from the cell surface during the extensive washing procedure prior to fixation of neurons and performance of the binding assay.
Assessment of Neurite Outgrowth-Eight-chamber slides (Nunc Lab-Tek) were coated with either purified brain amphoterin (20 g/ml), recombinant amphoterin (10 g/ml), or poly-L-lysine (50 g/ml) for 18 h at 37°C. Cortical neurons were isolated from the cerebral hemispheres of day 17 rat embryos, plated on the latter precoated slides, and incu-bated for 18 h at 37°C in Dulbecco's minimal essential medium containing bovine serum albumin (0.5%). Slides were subsequently fixed with paraformaldehyde (4%) containing Nonidet P-40 (0.1%) and stained with anti-tubulin antibody (Sigma) according to the manufacturer's instructions. Where indicated, isolated neuronal cells in medium containing bovine serum albumin (0.5%) were preincubated for 1 h at 4°C with anti-RAGE F(abЈ) 2 or nonimmune F(abЈ) 2 , or amphoterincoated wells were pretreated for 1 h at 37°C with sRAGE in medium containing bovine serum albumin (0.5%) prior to the addition of the cortical neurons.
In Situ Hybridization for RAGE and Amphoterin mRNA-In situ hybridization for RAGE was performed according to previously published methods (11). Cortical neurons were prepared as described above, fixed with paraformaldehyde (4%) containing Nonidet P-40 (1%), or rat brains were harvested, fixed in formalin (10%), and thin sagittal sections were prepared. Digoxigenin-labeled riboprobes were transcribed from the plasmid B379 -2A containing a 1406-base pair fragment of bovine RAGE (nucleotides 1-1406) cloned into the EcoRI site of pBluescript II SK. Antisense probe was transcribed from the T3 promoter with the plasmid linearized by XhoI. Control sense probe was transcribed from the T7 promoter with the plasmid linearized by XbaI. Transcription was performed using an RNA labeling kit (Boehringer Mannheim) with digoxigenin-UTP by in vitro transcription with either T3 or T7 polymerase. The digoxigenin-labeled RNA probe was hybridized to cellular mRNA for 16 h at 55°C and detected with anti-digoxigenin alkaline phosphatase-conjugated antibody over 16 h at 4°C. Antibody was visualized after 6 h incubation with X-phosphate and nitro blue tetrazolium salt. In situ hybridization for rat amphoterin was performed using a plasmid containing 936 base pairs of rat amphoterin spanning nucleotides 32-967. The antisense probe was transcribed from the SP6 promoter with the plasmid linearized by XbaI. The sense probe was transcribed from the T7 promoter with the plasmid linearized by HindIII. Transcription, labeling, and detection were carried out as above and the antibody visualized for 10 h with X-phosphate and nitro blue tetrazolium salt.
Immunohistochemistry for RAGE and Amphoterin-Cortical neurons, prepared as described above, were fixed with paraformaldehyde (2%) or rat brains, harvested and prepared as above, were fixed in formalin (10%). Immunohistochemistry for RAGE was performed using rabbit anti-RAGE IgG and nonimmune rabbit IgG as above and immunohistochemistry for amphoterin was performed using chicken antiamphoterin IgG and nonimmune chicken IgG. Peroxidase-conjugated goat anti-rabbit IgG and rabbit anti-chicken IgG (Sigma) were used, respectively, as secondary antibodies, according to the manufacturer's instructions.

RESULTS
Purification of Ligands for RAGE from Bovine Lung-Detergent extract of bovine lung acetone powder was applied to heparin-Sepharose and the column was step-eluted with ascending concentrations of salt (Fig. 1A). Fractions were evaluated for the presence of RAGE binding activity following adsorption to microtiter wells using a competitive binding assay measuring specific binding of bovine 125 I-RAGE. Only material eluting from the heparin column at an NaCl concentration of 0.5 M bound RAGE, and the latter fractions (50 -100) were pooled, dialyzed, and applied to a column with immobilized RAGE (Fig. 1B). Following extensive washing, the column was eluted with high salt buffer (2 M NaCl), and fraction numbers 10 -20 demonstrated the capacity to bind RAGE. The latter pool of active fractions from the immobilized RAGE column was subjected to nonreduced SDS-PAGE. Two protein bands were visualized on the gel by silver staining, corresponding to approximate M r values of 12,000 and 23,000. Material eluted from these bands was re-run on SDS-PAGE, appeared homogenous by silver staining with M r Ϸ 23,000 and M r Ϸ 12,000 (Fig. 1C, p23 and p12, respectively) and bound 125 I-RAGE, based on subsequent gel elution (Fig. 1D). Because of limited amounts of purified protein which could be prepared, the current studies are limited to further characterization of the Ϸ23,000 polypeptide which binds to RAGE.
Purified Ϸ23-kDa polypeptide (p23) was subjected to NH 2terminal and internal sequence analysis. The NH 2 -terminal sequence of p23 had 17 identifiable residues of the first 20. The 17 identified residues matched exactly the NH 2 terminus of bovine amphoterin (Table I). Five internal sequence peptides generated by Lys-C and cyanogen bromide cleavage were found (Table I). In total, the residues identified from the NH 2 -terminal and five internal peptides (including tentative assignments) matched 65 of 68 residues in amphoterin. All of the inconsistent amino acid assignments occurred in one fragment (number 3), possibly the result of a contaminating peptide that copurified with fragment 3 on the HPLC). Each of the internal peptides generally aligned with amphoterin in a location that would be predicted by the cleavage used to generate the peptide (Lys-C peptides follow lysine residues and cyanogen bromide peptides follow methionine). These data indicated that p23 and amphoterin were identical and suggested the hypothesis that RAGE might be a cell surface acceptor site for amphoterin.
Amphoterin Binds to RAGE-Microtiter wells with adsorbed RAGE bound 125 I-rat brain amphoterin in a dose-dependent and saturable manner, with K d ϭ 6.4 Ϯ 1.0 nM, and a capacity of 46.7 Ϯ 2.4 fmol/well ( Fig. 2A). To further characterize the interaction of amphoterin with RAGE, rat amphoterin was expressed recombinantly using a baculovirus expression system and was purified from neonatal rat brain. In this case, adsorbed RAGE bound 125 I-recombinant amphoterin in a similar manner, with K d of 10.24 Ϯ 2.84 nM, and a capacity of 43.01 Ϯ 4.53 fmol/well (data not shown). Binding of 125 I-rat brain amphoterin to RAGE-coated wells was specific, as demonstrated by dose-dependent inhibition on addition of excess sRAGE (Fig. 2B) or by preincubation of immobilized RAGE with blocking antibody to RAGE (Fig. 2C). In contrast, nonimmune IgG was without effect (Fig. 2C), and wells coated with an irrelevant polypeptide, such as albumin, in place of RAGE showed no specific binding of 125 I-amphoterin (data not shown). Binding was also inhibited by excess amounts of AGE albumin, whereas native albumin was without effect (Fig. 2D). Similar results were observed using recombinant radiolabeled amphoterin (data not shown).

FIG. 1. Purification of RAGE binding proteins by sequential chromatography on heparin-Sepharose (A) and RAGE affinity chromatography (B), followed by preparative SDS-PAGE (C and D).
A, bovine lung powder was extracted with octyl-␤-glucoside and chromatographed on heparin-Sepharose. Specific binding of 125 I-RAGE alone (total) or 125 I-RAGE in the presence of an 100-fold excess of unlabeled RAGE (nonspecific binding) and OD 280 of the fractions was determined. Specific 125 I-RAGE binding (total minus nonspecific) is shown. B, fractions with RAGE binding activity, corresponding to those eluted at an NaCl concentration of 0.5 M were subsequently chromatographed on Affi-Gel 10 to which had been bound purified RAGE. After extensive washing with equilibration buffer, the resin was eluted with NaCl (2 M) and positive fractions were identified based on their ability to bind 125 I-RAGE. C, fractions with RAGE binding activity were subjected to preparative SDS-PAGE (nonreduced, 12%) followed by silver staining. D, lanes of the SDS gel identical to that in panel C were cut into 1-mm pieces, gel elution was performed as described, and RAGE binding activity was determined.
to be highly expressed in embryonic rat neurons, and that amphoterin promotes neurite outgrowth (23). As these data suggest a physiologic role for amphoterin in neurons early in development, we sought to establish whether RAGE would be expressed as well. Immunostaining with anti-rat RAGE IgG displayed the antigen in embryonic cortical neurons (Fig. 3A), whereas cultures incubated with nonimmune IgG were negative (Fig. 3B). Similarly, in situ hybridization revealed the presence of RAGE mRNA in E17 neuronal cells using the antisense probe (Fig. 3C), whereas hybridization with sense probe was negative (Fig. 3D). Although neural cell adhesion molecule and RAGE share some sequence homology, as both are members of the immunoglobulin superfamily, there is no evidence that antibodies to either cross-react, and the pattern of neural cell adhesion molecule immunostaining on cultured rat cortical neurons was distinct from that for RAGE (data not shown). Additional data supporting the specificity of RAGE immunostaining was its disappearance on addition of sRAGE (data not shown).
Since neonatal cortical neurons expressed RAGE, it was important to determine if they bound amphoterin, and if this was mediated by interaction with RAGE. Radioligand binding studies with 125 I-amphoterin were performed on cortical neurons isolated from neonatal rat brain (E17) and cultured on poly-L-lysine-coated wells after brief fixation in paraformaldehyde. 125 I-Rat brain amphoterin bound specifically to cultured Atn 112-122 Atn 127-38 Cyanogen bromide fragments 4 Atn 13-29 Atn 75-81 The following abbreviations were used: Atn, bovine amphoterin published sequence, numbers indicate the residues shown; X, no residue identifiable at that cycle; parenthesis, tentative assignment.

FIG. 2. Binding of amphoterin to immobilized RAGE: dose dependence (A), effect of sRAGE (B), blocking antibody to RAGE (C), and AGE albumin (D).
A, dose-dependence. Rat brain amphoterin was prepared and purified as described and radiolabeled with 125 I. Microtiter wells were coated with purified RAGE (2.5 g/well) and a binding assay was performed as described by adding 125 I-amphoterin alone (total binding) or in the presence of 100-fold excess unlabeled amphoterin (nonspecific). Specific binding of 125 I-amphoterin is plotted versus free/ added 125 I-amphoterin. Parameters of binding for rat brain amphoterin were K d ϭ 6.4 Ϯ 1.0 nM with capacity of 46.7 Ϯ 2.4 fmol/well. B, effect of sRAGE. 125 I-Amphoterin (3 nM) was preincubated with the indicated concentration of sRAGE for 2 h, and then binding to RAGE immobilized on microtiter wells was studied as above. Maximal binding was specific binding observed in the absence of added sRAGE. C, effect of anti-RAGE IgG. RAGE-coated microtiter wells were preincubated with anti-RAGE IgG or nonimmune IgG for 2 h and then a radioligand binding assay performed with 125 I-amphoterin and excess unlabeled amphoterin as above. D, effect of AGE albumin. RAGE-coated wells were preincubated with AGE albumin or native albumin for 2 h and then a radioligand binding assay with 125 I-amphoterin was performed as above.
neurons in a dose-dependent and saturable manner, with K d ϭ 8.8 Ϯ 2.4 nM and a capacity of 28.8 Ϯ 2.8 fmol/well (Fig. 4A). The binding of 125 I-recombinant amphoterin was very similar, with K d ϭ 8.09 Ϯ 1.60 nM and capacity of 34.66 Ϯ 2.34 fmol/well (data not shown). Binding of 125 I-rat brain amphoterin to cortical neurons was dependent on interaction with RAGE; addition of sRAGE blocked binding in a dose-dependent manner (Fig. 4B) and pretreatment of cells with increasing doses of anti-RAGE IgG also prevented binding, whereas nonimmune IgG was without effect (Fig. 4C). Binding of 125 I-rat brain amphoterin to cortical neurons was also inhibited in the presence of AGE albumin, whereas native albumin was without effect (Fig. 4D). Similar results were observed using recombinant radiolabeled amphoterin (data not shown).
RAGE-Amphoterin Interaction Promotes Neurite Outgrowth in Vitro-Previous studies suggested a role for amphoterin in neuronal development, based on induction of neurite outgrowth in vitro. We sought to determine if amphoterin interaction with RAGE mediated the latter effect. Cortical neurons plated on amphoterin, or poly-L-lysine substrates demonstrated neurite outgrowth (Fig. 5, a and g, respectively). When amphoterincoated wells were pretreated with sRAGE, a dose-dependent inhibition of neurite-outgrowth occurred (Fig. 5b, sRAGE at 50 g/ml; and Fig. 5c, sRAGE at 5 g/ml). Further evidence implicating amphoterin-RAGE interaction in neurite outgrowth was inhibition in the presence of F(abЈ) 2 prepared from anti-RAGE IgG. Pretreatment of the neurons with anti-RAGE F(abЈ) 2 inhibited neurite formation by neurons in amphoterincoated wells in a dose-dependent manner (Fig. 5e, anti-RAGE F(abЈ) 2 at 40 g/ml; and Fig. 5f, anti-RAGE F(abЈ) 2 at 4 g/ml) whereas pretreatment with nonimmune F(abЈ) 2 was without effect (Fig. 5d). To demonstrate that blocking access to RAGE FIG. 3. Immunostaining and in situ hybridization for RAGE protein and mRNA, respectively, in E17 rat cultured neuronal cells. Cortical neuronal cells were isolated from E17 rat embryos and cultured for 2 days on poly-L-lysine (50 g/ml)-coated dishes. Cells were then fixed with paraformaldehyde (2%) for immunostaining or paraformaldehyde (4%) with Nonidet P-40 (0.1%) for in situ hybridization studies. Fixed cells were stained with anti-rat RAGE IgG (a) or nonimmune IgG (b). In situ hybridization was performed with digoxigenin-labeled RAGE riboprobes and detected with alkaline phosphatase-conjugated anti-digoxigenin antibody. Panel c, antisense probe, and panel d, sense probe. Scale bar, 50 m. Cortical neuronal cells were isolated from E17 embryos as described and cultured for 2 days on poly-L-lysine-coated wells (1 ϫ 10 5 cells/well). After cells were fixed with paraformaldehyde (2%), a radioligand binding assay was performed with 125 I-amphoterin as described. The parameters of specific binding of 125 I-amphoterin were: K d ϭ 8.8 Ϯ 2.4 nM with capacity of 28.8 Ϯ 2.8 nM. B, effect of sRAGE. 125 I-amphoterin (3 nM) was preincubated with the indicated concentration of sRAGE for 2 h, and then binding to cultured neurons was studied as above. Maximal binding was specific binding observed in the absence of added sRAGE. C, effect of anti-RAGE IgG. Cultured cortical neuronal cells were preincubated with anti-RAGE IgG or nonimmune IgG for 2 h and then a radioligand binding assay performed with 125 I-amphoterin and excess unlabeled amphoterin as above. D, effect of AGE albumin. Cultured cortical neurons were preincubated with AGE albumin or native albumin for 2 h and then a radioligand binding assay with 125 I-amphoterin was performed as above.
specifically inhibited neurite outgrowth observed on amphoterin-coated substrates, parallel experiments were performed with neurons plated on poly-L-lysine, which provides a suitable substrate for neurite outgrowth. Neurite outgrowth on poly-Llysine-coated wells was not inhibited by sRAGE (Fig. 5h) or by anti-RAGE F(abЈ) 2 (Fig. 5i).
Co-localization of Messenger RNA and Protein for Amphoterin and RAGE in Developing Rat Brains-These data indicated that the interaction of amphoterin with cultured rat neurons is mediated, at least in large part, by RAGE, and that neurite outgrowth is a consequence of amphoterin binding to RAGE. As a first step in understanding the potential implications of our findings for the developing nervous system in vivo, studies to identify RAGE and amphoterin in brains of E17 as well as 5-and 17-day-old rats (P5 and P17, respectively) were performed. In situ hybridization studies revealed the co-localization of RAGE and amphoterin mRNA in the cerebral cortex of E17 rats ( Fig. 6: a, E17 RAGE antisense; and b, E17 amphoterin antisense). Sense control experiments were negative for RAGE and amphoterin in E17 cerebral cortex ( Fig. 6; c, E17 RAGE sense; and d, E17 amphoterin sense). Similar co-localization of RAGE and amphoterin mRNA was demonstrated in P5 cerebral cortex (Fig. 6: e, P5 RAGE antisense; and f, amphoterin antisense). Sense controls for P5 RAGE and amphoterin were negative (data not shown). RAGE and amphoterin mRNA were also co-localized in developing areas of the hippocampus in P5 rats ( Fig. 6: g, P5 RAGE antisense; and h, P5 amphoterin antisense) as well as the developing cerebellum of P17 rats ( Fig. 6: i, P17 RAGE antisense; and j, P17 amphoterin antisense). In all cases, sense controls were negative (data not shown). In P17 cerebellum, the pattern of RAGE and amphoterin mRNA localization was similar in the granular layer, but the intensity of the RAGE mRNA signal was detected to a slightly greater degree in the Purkinje layer (Fig. 6i, arrow).
Immunohistochemistry studies further supported the co-localization of amphoterin and RAGE in the developing nervous system of the rat. RAGE and amphoterin protein were both present in the cerebral cortex of P5 rat ( Fig. 7: a, staining with anti-RAGE IgG; (c, nonimmune rabbit IgG revealed no staining) and b, staining with anti-amphoterin IgG (d, nonimmune chicken IgG was negative)). Higher magnification views of the P5 cerebral cortex revealed that while the cell bodies of the developing neurons were intensely positive for RAGE and amphoterin, staining of the developing axonal processes was even more dramatic (Fig. 7, e and f, respectively, thick arrow). Similar results were observed in the hippocampus of P5 rats (Fig. 7, g and i, anti-RAGE IgG; and h and j, anti-amphoterin IgG).
Taken together, these data from in situ hybridization and immunohistochemistry studies suggest that cells likely to express RAGE and amphoterin in developing rat brain are in close proximity, potentially allowing RAGE-amphoterin interaction to mediate neurite outgrowth. DISCUSSION We initially anticipated that cellular binding proteins/receptors for AGEs would be analogous to the collagen-like heterotrimeric scavenger receptors for acetylated low density lipoprotein on mononuclear phagocytes which mediate cellular uptake of modified lipoproteins (29). In fact, recent studies have shown that such scavenger receptors can interact with AGEs (9). However, RAGE, the first AGE cellular binding protein to be characterized in detail, was most analogous to immunoglobulin-like receptors. Other functional properties of RAGE suggested it was not an effective scavenger: infused AGEs, removed from the circulation by endothelial RAGE, were, in large part, transported by a transcytotic mechanism to the subendothelium where they became associated with matrix and smooth muscle cell elements (16). Consistent with these data, following infusion of AGE albumin, induction of oxidant stress, assessed by appearance of malondialdehyde epitopes in the vasculature, was present in endothelium, subendothelium, and smooth muscle cells (14). Furthermore, since the interaction of AGEs with RAGE enhanced monocyte chemotaxis and activation, and increased expression of endothelial vascular cell adhesion molecule 1, it was plausible that under physiologic conditions, natural or non-AGE ligands might interact with RAGE to mediate, for example, adhesive functions critical for normal development (12,13,30).
As lung is a rich source of RAGE, we considered this tissue a logical place in which to identify natural ligands of RAGE. The data presented in this study demonstrate that two polypeptides, with M r values of Ϸ12,000 and Ϸ23,000 on SDS-PAGE, bind AGEs. Whereas the Ϸ12-kDa polypeptide has an unique NH 2 -terminal sequence and its initial characterization is still under way, the Ϸ23-kDa polypeptide proved to be identical to amphoterin based on extensive protein sequence data. Although the more rapid migration of p23 purified from bovine lung is somewhat different from the M r reported for amphoterin purified from rat brain (Ϸ30,000; Ref. 23), in view of sequence identity and recognition of p23 by anti-peptide antisera made to either amino acids 2-21 or 94 -101 (data not shown), it is most likely that p23 arises from cleavage of amphoterin during the tissue extraction or purification procedure. Consistent with this view, Northern analysis of rat lung RNA using amphoterin cDNA showed a single band, which is similar to that reported for amphoterin mRNA in rat brain (data not shown).
More detailed studies of amphoterin-RAGE interaction, performed with larger amounts of amphoterin purified from rat brain or produced recombinantly with baculovirus, showed that FIG. 5. Neurite outgrowth assays and the effect of RAGE blockade. Eight-chamber wells were coated with either amphoterin (20 g/ ml, panels a-f) or poly-L-lysine (50 g/ml, panels g-i) for 18 h. Cortical neuronal cells were isolated from E17 rat embryos as described and fixed with paraformaldehyde (4%) containing Nonidet P-40 (0.1%) and stained with monoclonal anti-tubulin antibody. a-c, effect of sRAGE. Amphoterin-coated wells and neuronal cells were pretreated with sRAGE or bovine serum albumin for 1 h at 37°C. a, neurite outgrowth on amphoterin-coated wells alone; b, in the presence of sRAGE (50 g/ml); or c, in the presence of sRAGE (5 g/ml). d-f, neurite outgrowth in the presence of anti-RAGE F(abЈ) 2 or nonimmune F(abЈ) 2 was assessed: d, in the presence of nonimmune F(abЈ) 2 (40 g/ml) or anti-RAGE F(abЈ) 2 (e, 40 g/ml; or f, 4 g/ml). g-i, neurite outgrowth on poly-L-lysine coated wells and the effect of RAGE blockade. g, neurite outgrowth on poly-L-lysine alone or in the presence of sRAGE (50 g/ml, h) or anti-RAGE F(ab)Ј 2 (40 g/ml, i). Scale bar, 50 m.
binding of amphoterin to RAGE was specific, saturable, and of higher affinity than that previously reported for AGEs (K d of Ϸ6 nM for amphoterin versus K d of Ϸ50 nM for AGE albumin). Domains in amphoterin mediating interaction with RAGE appear to be unrelated to AGE-like epitopes, as enzyme-linked immunosorbent assay of amphoterin preparations showed no detectable AGE antigen and anti-AGE IgG had no effect on amphoterin binding to RAGE (data not shown), although this antibody has been previously shown to block interaction of AGE albumin and AGEs immunoisolated from diabetic plasma with RAGE (14).
Amphoterin was first identified by Rauvala and Pihlaskari (23), as an Ϸ30-kDa polypeptide selectively and highly expressed in the developing rat central nervous system. Cultured embryonic neurons plated on amphoterin-coated matrices formed neuritic processes, suggestive of their differentiation, a critical step in the overall design and maturation of the nervous system. RAGE expression in cell bodies and axonal processes was first observed in the bovine nervous system in motor neurons and in certain populations of cortical neurons (11), and has been extended in this study to include the developing rat central nervous system. Consistent with these data, embryonic rat cortical neurons which have been show to produce amphoterin also expressed RAGE. Neuronal RAGE was functional, as indicated by its capacity to bind amphoterin and to mediate amphoterin-induced neurite outgrowth.
Previous studies have identified potential binding sites for amphoterin distinct from RAGE. Salmivirta and colleagues FIG. 6. In situ hybridization studies co-localize RAGE and amphoterin mRNA in the developing rat nervous system. Sections from E17, P5, and P17 developing rat brain were harvested and prepared as described above. In situ hybridization was performed as indicated in the text in E17 cerebral cortex: a, RAGE antisense; b, amphoterin antisense; c, RAGE sense, and d, amphoterin sense, P5 cerebral cortex; e, RAGE antisense; and f, amphoterin antisense, P5 hippocampus: g, RAGE antisense, and h, amphoterin antisense and P17 cerebellum: i, RAGE antisense; and j, amphoterin antisense. In all other cases, sense controls for RAGE and amphoterin were negative (data not shown). Scale bar, 1 mm (panels a, b, c, d, g, h, i, and j); scale bar, 500 m (panels e and f). (31) showed that amphoterin binds to syndecam, a cell surface proteoglycan containing both heparan sulfate and chondroitin sulfate glycosaminoglycan chains, in cultured mouse mammary epithelial cells. However, anti-syndecam antibodies did not detect specific staining in neural tissues. At the mRNA level, Northern blots of RNA from mouse forebrain hybridized with a syndecam cDNA showed a band of 4.5 kilobases; the significance of the latter was unclear, as it is considerably larger than that detected in epithelial cells (2.6 and 3.4 kilobases) and has not been further characterized in the literature to date. In other studies, Mohan et al. (32) presented evidence that sulfoglycolipids (immunoreactive with the monoclonal antibody HNK-1) may interact with amphoterin in the developing nervous system based on solid phase binding assays. The present studies identify RAGE as a cellular binding site for amphoterin, based on binding and functional data, and provide the strongest evidence, thus far, for a putative amphoterin receptor in the developing central nervous system.
Expression of amphoterin has also been demonstrated in transformed cells (33). Parkkinen and colleagues (33) showed that C6 glioma cells, HL-60 promyelocytes, U937 promonocytes, HT1080 fibrosarcoma cells, and B16 melanoma cells produced amphoterin. These investigators also demonstrated that amphoterin strongly enhanced the rate of plasminogen activation and promoted the generation of surface-bound plasmin by both tissue-type and urokinase-type plasminogen activators, suggesting a role for amphoterin in invasive neoplastic lesions (18). Future studies will be required to determine if RAGE modulates any of these properties of amphoterin in the biology of neoplasia.
In summary, these data suggest possible roles for RAGE under circumstances in which the presence of AGEs is unlikely. Specifically, in the rat embryonic nervous system, RAGE is highly expressed, and co-localizes, at the level of antigen and mRNA, with the presence of amphoterin. As blocking access to RAGE, with either excess sRAGE or anti-RAGE IgG/F(abЈ) 2 , prevents amphoterin-induced neurite outgrowth in cell culture, it is tempting to speculate that RAGE mediates the potential role of amphoterin in neuronal development. These observations provide a first step in characterizing a novel aspect of the biology of RAGE, and emphasize the importance of future in vivo studies to address the physiologic significance of amphoterin-RAGE interaction, with respect to the development and maturation of the central nervous system.