Hexameric Calgranulin C (S100A12) Binds to the Receptor for Advanced Glycated End Products (RAGE) Using Symmetric Hydrophobic Target-binding Patches*

Calgranulin C (S100A12) is a member of the S100 family of proteins that undergoes a conformational change upon calcium binding allowing them to interact with target molecules and initiate biological responses; one such target is the receptor for advanced glycation products (RAGE). The RAGE-calgranulin C interaction mediates a pro-inflammatory response to cellular stress and can contribute to the pathogenesis of inflammatory lesions. The soluble extracellular part of RAGE (sRAGE) was shown to decrease the inflammation response possibly by scavenging RAGE-activating ligands. Here, by using high resolution NMR spectroscopy, we identified the sRAGE-calgranulin C interaction surface. Ca2+ binding creates two symmetric hydrophobic surfaces on Ca2+-calgranulin C that allow calgranulin C to bind to the C-type immunoglobulin domain of RAGE. Apo-calgranulin C also binds to sRAGE using a completely different surface and with substantially lower affinity, thus underscoring the role of Ca2+ binding to S100 proteins as a molecular switch. By using native gel electrophoresis, chromatography, and fluorescence spectroscopy, we established that sRAGE forms tetramers that bind to hexamers of Ca2+-calgranulin C. This arrangement creates a large platform for effectively transmitting RAGE-dependent signals from extracellular S100 proteins to the cytoplasmic signaling complexes.

Calgranulin C (S100A12) is one of 21 members of the S100 family of EF-hand calcium-binding proteins identified to date (1). Members of the S100 family have both intracellular and extracellular functions and participate in diverse cellular processes leading to cell growth and differentiation, cell cycle regulation, transcription, and signal transduction receptor activities (2). This diversity is tailored in part by a distinct pattern of subcellular localization and tissue-specific expression. EF-hand calcium-binding proteins translate the physiological changes in calcium levels into specific cellular responses by undergoing a large conformational change that exposes a binding site recog-nized by downstream effectors, essentially acting as a molecular switch (3,4).
Most S100 proteins are noncovalent homodimers. Some members of the S100 family, such as S100A8 and S100A9, form heterodimers and heterotetramers (5). Higher order homo-oligomers have also been reported (6,7) and are suggested to play an important biological role in ligand targeting (8). The basic structural and functional unit of the S100 proteins is a symmetric dimer comprised of two EF-hand subunits organized into an eight-helix bundle (4). The conformational change triggered by calcium binding asymmetrically affects the two EF-hand motifs of the subunit with the N-terminal EF-hand undergoing a relatively small change in conformation, whereas the change in the C-terminal EF-hand is much larger.
This conformational rearrangement creates a shallow hydrophobic pocket. Structural studies of the ligated forms of S100 proteins showed that this hydrophobic surface can bind short peptides from the respective targets of S100 proteins (9 -13). Surprisingly, despite utilizing a similar binding surface, different peptides bind to individual S100 proteins in different ways. For example, in the complexes determined by using x-ray crystallography involving S100A10 and a peptide from annexin II (10) and S100A11 and a peptide from annexin I (11), the peptide is packed against helices 1 and 4 of S100A10. However, in the complexes determined using NMR involving S100B and the peptide from p53 (13), the p53 peptide makes contacts with helices 3 and 4, and in the complex between S100B and the peptide from NDR (nuclear Dbf2-related) kinase (9), the NDR peptide is mostly packed against helix 4. The observed structural differences suggest that the binding modes may be still different for the full-length S100 protein targets.
The crystal structure of Ca 2ϩ -bound calgranulin C closely resembles structures of the S100/calgranulin family proteins, presenting a dimer of four-helix subunits (14). The calciumbinding EF-hand motifs, helix 1-loop 1-helix 2 and helix 3-loop 3-helix 4 are linked by a short anti-parallel ␤-bridge, a common feature of the EF-hand proteins. Helices 2 and 3 are linked by the hinge region formed by loop 2. The hinge region is poorly conserved among S100 family members suggesting that specificity of target recognition may reside in this area. At low millimolar concentrations of Ca 2ϩ , calgranulin C can also form a hexamer consisting of three symmetrically positioned Ca 2ϩbound dimers (6).
The S100 family members characteristically accumulate at sites of chronic inflammation. The mechanism by which cal-granulin C modulates the course of inflammatory processes is related to its interaction with the receptor for advanced glycated products (RAGE) 2 (15). RAGE is a multiligand member of the immunoglobulin superfamily of cell surface molecules and displays a high degree of sequence homology to the neural adhesion molecule (N-CAM) (16,17). Based on sequence homology, RAGE contains an N-terminal V-type immunoglobulin domain followed by two C-type immunoglobulin domains, a trans-membrane helix and short cytoplasmic tail. The cytoplasmic tail is absolutely essential for intracellular signaling. A RAGE mutant lacking a cytoplasmic tail, dominant negative-RAGE, exhibits a strong dominant negative effect on RAGE-dependent signaling (15,18,19). This observation indicates that signaling molecules associate with the cytosolic domain of RAGE. A recent study revealed that ERK kinase binds directly to a site on the RAGE cytoplasmic tail suggesting its essential role in RAGE-mediated signaling (20).
RAGE was originally identified and characterized based on its ability to bind advanced glycation products (AGEs), adducts formed by glycoxidation that accumulate especially in disorders such as diabetes and renal failure (21). Rather than working as a scavenger receptor to effectively uptake and dispose of AGEs, ligand-active RAGE causes cellular perturbations leading to an inflammatory response (22). The extracellular, naturally occurring splicing variant of RAGE, soluble RAGE (sRAGE) that includes three immunoglobulin domains (17), was shown to alleviate the inflammatory effects resulting from the AGEs-RAGE interaction by sequestering RAGE-activating ligands (23).
At least seven members of the S100 family, calgranulin C, S100A1, S100A4, S100A11, S100A13, S100B, and S100P, were identified to be ligands of RAGE (15,24,25). Calgranulin C binding to RAGE mediates the activation of endothelial cells, macrophages, and lymphocytes, cells central to the inflammatory response. By using specific inhibitors of different signaling pathways, it was found that phospholipase C, protein kinase C, Ca 2ϩ fluxes, calmodulin-dependent kinase II, and mitogen-activated protein kinase/ERK kinase (MEK) are required for extracellular calgranulin C to initiate signal transduction through RAGE (25). Besides S100 family proteins, a small protein shown to be involved in tumor growth, HMGB1, was also identified as a RAGE ligand (26).
Despite extensive in vivo studies of the functional role of RAGE, little is known about the structural organization of the receptor molecule and its interaction with physiological ligands. Biochemical characterization of RAGE-ligand interactions was limited to in vitro binding studies, which revealed that the N-terminal V domain of RAGE is involved in binding AGEs (27). It was also shown that the binding of AGEs to RAGE inhibited calgranulin C binding to RAGE (15).
Here we characterized the interaction between calgranulin C and sRAGE using native gel electrophoresis, chromatography, fluorescence and high resolution NMR spectroscopy. We showed that hexameric Ca 2ϩ -calgranulin C binding to sRAGE is mediated by the first C-type immunoglobulin-type domain of tetrameric RAGE, C1. Interaction surface mapping studies reveal that the apo-form of calgranulin C also binds to RAGE with comparatively low affinity, using a surface that is substantially different from the surface participating in the high affinity Ca 2ϩ -calgranulin C-RAGE interaction. These calciumdependent interactions substantiate the idea that Ca 2ϩ binding is the trigger for a molecular switch that transmits structural rearrangements through the trans-membrane domain to the cytosolic tail of RAGE leading to signal transduction through the RAGE receptor.
Expression and Purification of sRAGE and RAGE Domains-Plasmids expressing sRAGE or RAGE domains were transformed into Escherichia coli strain BL21(DE3) Codonϩ (Novagen) for overexpression and purification. Cells were grown to ϳ0.7 A 600 at 37°C in LB containing 35 mg/liter kanamycin, induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside, and grown overnight. Cells were harvested and resuspended in 50 mM Hepes-NaOH buffer (pH 7.0) containing 8 M urea and heat-lysed at 95°C for 10 min. The lysate was centrifuged, and the supernatant loaded onto a His tag column. The column was washed with 50 mM Hepes buffer (pH 7.0), and the protein was allowed to renature on the column prior to eluting with 50 mM Hepes-NaOH (pH 7.0) containing 500 mM imidazole. Fractions containing the eluted protein were pooled, dialyzed into 10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl, concentrated to 0.2-0.5 mM, and loaded onto a Sephadex 200 size exclusion column (Amersham Biosciences) equilibrated with the same buffer. Fractions containing the eluted protein were pooled and concentrated to 0.3-1.0 mM using centricones (Millipore). All concentrations cited for sRAGE and its domains refer to total monomer concentration. Purity was estimated to be Ͼ90% by Coomassie-stained SDS-PAGE. 2 The abbreviations used are: RAGE, receptor for advanced glycation products; sRAGE, soluble RAGE; AGE, advanced glycation product; HSQC, heteronuclear single quantum coherence; PDB, Protein Data Bank; TNS, 6-(ptoluidino)naphthalene-2-sulfonate; ERK, extracellular signal-regulated kinase.
Expression and Purification of HMGB1-Plasmid expressing full-length human HMGB1, pRSF-HMGB1, was transformed into E. coli strain BL21(DE3) Codonϩ (Novagen) for overexpression and purification. Cells were grown to ϳ0.7 A 600 at 37°C in LB containing 35 mg/liter kanamycin and induced for 3 h with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. Cells were harvested and resuspended in 50 mM Tris-NaOH buffer (pH 8.0), 500 mM NaCl, 20 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride and sonicated on ice for 10 min. The lysate was centrifuged, and the supernatant was loaded onto a DEAE-cellulose column (Sigma) equilibrated in the same buffer. The flow-through fraction of supernatant was collected and mixed with solid ammonium sulfate (50% of maximum solubility at 0°C). The supernatant was dialyzed against 20 mM Tris-NaOH (pH 8.0) and loaded on a Q-Sepharose column (Amersham Biosciences) equilibrated with the same buffer. Protein was eluted at 1 ml/min with a linear gradient of 20 mM Tris/HCl (pH 8.0) to 20 mM Tris/HCl (pH 8.0), 1 M NaCl over 30 min. Fractions containing the eluted protein were pooled, dialyzed into NMR buffer (10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl, 0.02% (w/v) NaN 3 ), and concentrated to 0.3-1.0 mM using centricones (Millipore). Purity was estimated to be Ͼ90% by Coomassie-stained SDS-PAGE.
Labeling, Expression, and Purification of Calgranulin C-Human calgranulin C (S100A12; amino acids 1-92) was expressed using pQE60calC (14); this construct contains four additional amino acids (MGGS) at the N terminus. To uniformly label calgranulin C for NMR spectroscopy, pQE60calC was transformed into E. coli strain BL21(DE3) Codonϩ (Novagen). For U-15 N-labeling, cells were grown at 37°C in minimal medium (M9) containing 35 mg/liter kanamycin and 1 g/liter [ 15 N]ammonium chloride as the sole nitrogen source. For U-15 N, 13 C-labeling, cells were grown at 37°C in M9 medium containing 35 mg/liter kanamycin, 1 g/liter [ 15 N]ammonium chloride, and 0.5 g/liter [ 13 C]glucose instead of unlabeled glucose as the sole carbon source. Cells were grown to ϳ0.6 A 600 , induced with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside, grown for 15 h, harvested, and resuspended in 50 mM Tris/HCl (pH 7.5), 300 mM NaCl, prior to lysis. The supernatant was diluted 1:1 with low salt buffer (50 mM Tris/HCl (pH 7.5), 10 mM CaCl 2 ) and clarified prior to loading onto a phenyl-FF column (Amersham Biosciences). Protein was eluted at 1 ml/min with a linear gradient of 50 mM Tris/HCl (pH 7.5), 150 mM NaCl, 5 mM CaCl 2 to 50 mM Tris/HCl (pH 7.5), 100 mM NaCl, 5 mM EDTA over 30 min. Fractions containing the eluted protein were pooled and dialyzed into 20 mM Tris/HCl (pH 8.0). The dialyzed protein was loaded onto an uno_Q1 column (Bio-Rad) and eluted at 2 ml/min with a linear gradient of 20 mM Tris/HCl (pH 8.0) to 20 mM Tris/HCl, 200 mM NaCl (pH 8.0) over 30 min. Fractions containing the eluted protein were pooled, dialyzed into NMR buffer, and concentrated to a final concentration of 0.3-0.5 mM using centricones. All concentrations cited for calgranulin C refer to total monomer concentration. Purity was estimated to be Ͼ95% by Coomassie-stained PAGE.
Protein Pulldown Assay-10 l of 10 M sRAGE in binding buffer (10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl) was mixed with 10 l of 20 M HMGB1 in the same buffer. For calgranulin C, 10 l of 10 M sRAGE in 10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl, 5 mM CaCl 2 , and 50 mM imidazole were mixed with 10 l of 20 M Ca 2ϩ -calgranulin C in the same buffer. 50 mM imidazole was included in the binding buffer to avoid weak interactions between Ca 2ϩ -calgranulin C and Ni 2ϩ beads. After incubating for 10 min, His tag beads (Amersham Biosciences) equilibrated in the binding buffer were added into the protein solution. The beads were washed three times with the binding buffer. The protein absorbed on the His tag beads was analyzed using Coomassie-stained SDS-PAGE.
Fluorescence Titrations-Fluorescence titrations were performed using an LS-55 luminescence spectrometer (PerkinElmer Life Sciences). All measurements were made at 25°C using a 1 ϫ 1-cm quartz cell. The excitation wavelengths were 255 nm for tyrosine fluorescence and 365 or 370 nm for TNS fluorescence; excitation and emission slit widths were 5 and 10 nm, respectively. The fluorescence intensity (F) at a fixed wavelength was corrected for sample dilution upon addition of titrant. Fluorescence data were analyzed using GRAMS (Thermo Galactic) and SPSS (Lead Technologies, Inc.) software.
Ca 2ϩ binding to calgranulin C was measured using two fluorescence probes. For tyrosine fluorescence experiments, a 2.5 M solution of calgranulin C in 10 mM Hepes-NaOH (pH 7.4) was titrated with a stock solution of CaCl 2 (2 mM in distilled water); for TNS fluorescence experiments, a solution containing 2.5 M calgranulin C and 20 M TNS in 10 mM Hepes-NaOH (pH 7.4) was titrated with the stock solution of CaCl 2 . Corrected fluorescence intensity was fit to Equation 1 to estimate the apparent Ca 2ϩ binding constant, K, for both tyrosine and TNS fluorescence titrations, where f1 is the fluorescence at zero Ca 2ϩ concentration; f2 is the fluorescence of the Ca 2ϩ -calgranulin C complex; N is total protein concentration, and W is the total Ca 2ϩ concentration.
C1C2 binding to calgranulin C was measured by performing TNS fluorescence titrations. A solution containing 3 M calgranulin C and 10 M TNS in 10 mM Hepes-NaOH (pH 6.8) plus 400 M CaCl 2 was titrated with a stock solution of C1C2 (44 M in 10 mM Hepes-NaOH (pH 6.8)). Corrected fluorescence intensity was fit to Equation 1 to estimate the apparent C1C2-binding constant K. In this case f1 is the TNS fluorescence at zero C1C2 concentration; f2 is the TNS fluorescence of the [Ca 2ϩ -calgranulin C]-C1C2 complex; W is the concentration of C1C2, and N is the concentration of Ca 2ϩ -calgranulin C.
Native PAGE-Native gels containing 12, 10.5, and 9% acrylamide were prepared in 0.375 M Tris/HCl (pH 8.8) (29). All samples were prepared in 10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl, 4 mM CaCl 2 , and 5% glycerol. Concentrations were 200 M for calgranulin C and 20 M for sRAGE. Binding reactions containing 20 M sRAGE and 5, 10, 15, and 20:1 mole ratios of calgranulin C were allowed to equilibrate for 30 min at 4°C prior to loading on pre-heated (20 min, 200 V) gels. The gels were loaded under full voltage (200 V) and run at 4°C for 3 h at 200 V using standard Tris-glycine running buffer (pH 8.3) supplemented with 4 mM CaCl 2 . Gels were stained using Imperial protein stain (Pierce). The data were processed according to Ref. 29. Molecular weight standards used were the same as those used in the column chromatography experiments. All experiments were repeated three times.
NMR Experiments-NMR experiments were performed on a Bruker Avance spectrometer, operating at a 1 H frequency of 700 MHz and equipped with a cryoprobe. All NMR data were collected at 25°C. Protein samples of calgranulin C, with concentrations ranging from 0.3 to 0.5 mM, were dissolved in NMR buffer (10 mM Hepes-NaOH (pH 6.5), 100 mM NaCl, 0.02% (w/v) NaN 3 ). The solution conditions used in these experiments are comparable with those expected to be found in vivo.
To prepare Ca 2ϩ -calgranulin C, a solution of 1 M CaCl 2 was titrated into a 0.5 mM [U-13 C, 15 N]calgranulin C solution until the molar ratio of metal:protein was 6:1. No changes in the NMR spectrum of the protein were detected at higher molar ratios. Gradient diffusion experiments to obtain rotational correlation times were performed using the pulse sequence described previously (30). To perform the titration experiments with sRAGE, a solution of 0.5 mM [U-15 N]calgranulin C or [U-15 N]Ca 2ϩ -calgranulin C was titrated into 0.2 mM sRAGE in three steps to yield sRAGE to calgranulin C molar ratios of 1:1, 1:6, and 1:10, respectively. Titration experiments with C1C2, V domain, and C2 domain of sRAGE were performed by titrating 1 mM solutions of C1C2, V domain, and C2 domain into 0.5 mM [U- 15 The results of the titration were monitored by 1 H{ 15 N} HSQC. Over the course of titration, the signal to noise ratio of the peaks that did not show any changes was kept constant by adjusting the number of scans. The [U- 15 N]calgranulin C sample did not show any changes in the 1 H{ 15 N} HSQC spectra upon 3-and 10-fold dilutions. The 1 H{ 15 N} HSQC spectrum of [U-15 N]Ca 2ϩ -calgranulin C sample showed a slight decrease in line broadening upon 10-fold dilution. The HNCACB, CBCA-(CO)NH, HNCO, HNCACO, and 15 N-edited nuclear Overhauser effect spectroscopy experiments were collected using previously described sequences. All spectra were processed using TOPSPIN (Bruker, Inc), and assignments were made using CARA (31).
Titrating C1C2 into apo-calgranulin C and Ca 2ϩ -calgranulin C results in free calgranulin C and a large, 200-kDa, complex in an equilibrium characterized by the lifetimes of two states. Because the affinity of apo-calgranulin C to C1C2 is low (K d ϳ140 M), we expect the fast exchange regime, k off Ͼ Ͼ ⌬, between free and bound states of calgranulin C, where k off is the inverse of the lifetime of the bound state and ⌬ ‫؍‬ ⍀ a Ϫ ⍀ b is the chemical shift difference between the free (⍀ a ) and bound (⍀ b ) states. In this regime, general formulas describing the transverse relaxation rate constant R 2 and resonant frequency of the dominant resonance line, ⍀, for the two-site exchange (32) can be simplified as shown in Equations 2 and 3, where p a(b) and R 2a(b) are the populations and the relaxation rate constants for spins in the free (a) and bound (b) states in the absence of chemical exchange.
For C1C2 titrated into Ca 2ϩ -calgranulin C, we observed differential broadening of the subset of the peaks characteristic of the intermediate to slow exchange regime, k off Յ ⌬, between free and bound states. Our data analysis followed the formalism developed in Ref. 33. According to this analysis, observed differential broadening is consistent with k off rate values of ϳ10 s Ϫ1 . This value is significantly larger than an estimate of 0.1 s Ϫ1 derived from the equilibrium binding constant, K d ϭ 100 nM, assuming a diffusion limit for k on ϭ 10 6 M Ϫ1 s Ϫ1 . This inconsistency can be because of the multistep binding process of Ca 2ϩcalgranulin C to C1C2 when predominantly dimeric Ca 2ϩ -calgranulin C forms a hexamer to bind effectively to sRAGE.
Homology Modeling-Homology modeling of the apo-calgranulin C structure was performed using the Swiss-Model optimization mode prepared by Swiss-PDB Viewer (34). Sequence homology between calgranulin C and S100A6 was based on alignment of conserved residues and the homology of secondary structures as revealed by NMR experiments. Energy minimization was performed using GROMOS96 with successive application of 200 cycles of steepest descent and 300 cycles of conjugate gradient minimizers. The quality of the resultant structure was estimated by the PROCHECK program (35). 88% of the calgranulin C residues fall into the most favorable region of the Ramachandran plot and 12% to the allowed region. A backbone atom root mean square deviation of 1.1 Å was calculated by comparing the model to the S100A6 crystal structure (PDB code 1K9P (36)).

RESULTS
Bacterially Expressed sRAGE Is Biologically Functional-To structurally characterize the calgranulin C-RAGE interaction, we overexpressed the soluble fraction of human RAGE (sRAGE) and its isolated domains, separately and tandemly in bacteria (Fig. 1a). The ability of bacterially expressed sRAGE to bind to two physiologically relevant RAGE ligands, calgranulin C (15) and HMGB1 (26), was assessed by using pulldown assays. Both ligands were shown to bind to sRAGE with nanomolar Hexameric Calgranulin C Binds to RAGE FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 affinity (15,37). His-tagged sRAGE-ligand complexes were immobilized on Ni 2ϩ beads, eluted, and analyzed by using SDS-PAGE. According to our results (Fig. 1b) both Ca 2ϩ -calgranulin C and HMGB1 formed stable complexes with sRAGE indicating that bacterially expressed sRAGE is properly folded to perform biological function.
sRAGE and Calgranulin C Are Multimeric Proteins-sRAGE and its isolated tandem C1C2 domains exist in solution in two oligomeric forms. During purification, both sRAGE (33 kDa) and C1C2 domains (23 kDa) (data not shown) eluted in two peaks from a gel filtration column (Fig. 2a), one containing a large oligomer with a molecular mass exceeding 1 MDa and the other containing a multimer of sRAGE or C1C2 domains with molecular masses of ϳ130 and ϳ100 kDa, respectively. When the peak corresponding to the large oligomer of sRAGE is collected and rerun over the column, we again elute two peaks, indicating that the two species are in reversible equilibrium (data not shown). In addition, sRAGE runs at ϳ140 kDa on native polyacrylamide gels (Fig. 2b). From these observations we concluded that both sRAGE and C1C2 domains form tetramers.
To further prove that these constructs exist in an oligomeric form, we measured the proton NMR transverse relaxation time for both sRAGE (8 ms) and C1C2 domains (10 ms). The data suggest that the molecular mass of sRAGE and the C1C2 domains are ϳ150 and ϳ110 kDa, respectively, consistent with our conclusion that these species form tetramers in solution.
In contrast, the isolated V and C2 domains of sRAGE are monomeric. According to our preliminary NMR experiments (Supplemental Material), the V and C2 domains of sRAGE are well folded in solution, suggesting that it is the C1 domain that promotes sRAGE oligomerization. Both the individually expressed C1 domain and the tandem VC1 construct had very low solubility and were not used in our experiments.
Ca 2ϩ -calgranulin C is known to form dimers (14), and hexamers have been identified in crystal structures (6) and by using dynamic light scattering (6). Purified Ca 2ϩ -calgranulin C (10 kDa) eluted in two peaks from a gel filtration column (Fig. 2a), corresponding to molecular masses of ϳ50 and ϳ16 kDa. The 3:1 ratio of molecular weights suggests that these species are hexamers and dimers. The larger molecular weight species exhibited a trailing boundary characteristic of a dissociating system. The low apparent molecular weights may result from the oblate ellipsoidal shape of calgranulin C (axial ratio ϳ2:1), which increases the frictional forces acting on the molecule as it migrates through the gel, causing it to elute at later times (38).
Ca 2ϩ -calgranulin C hexamers and dimers were also evident on native polyacrylamide gels, with apparent molecular masses of ϳ102 and ϳ33 kDa, respectively (Fig. 2b). The high apparent molecular weights for Ca 2ϩ -calgranulin C once again may result from the nonspherical shape of the molecule, which increases the frictional forces acting on the molecule, reducing its migration through the gel matrix. In both the chromatographic and electrophoretic experiments, the ratio of the higher molecular weight species to that of the lower molecular weight species is ϳ3:1, and the nonideal hydrodynamic properties predict anomalously low and high molecular weights, respectively. These observations are consistent with previous observations (6) and support our conclusion that Ca 2ϩ -calgranulin C forms dimers and hexamers in solution.
Ca 2ϩ Binding to Apo-calgranulin C Promotes Hexamerization-To investigate the importance of Ca 2ϩ binding to calgranulin C as a molecular switch for RAGE-mediated signaling, we used the fluorescence probe TNS to characterize the binding of Ca 2ϩ to apo-calgranulin C. TNS is useful for studying protein-protein interactions because it is weakly fluorescent in water but exhibits intense fluorescence upon binding to a protein hydrophobic surface (39). An increase in TNS fluorescence intensity upon Ca 2ϩ and Zn 2ϩ binding to both S100A1 and S100B has been observed (40). TNS fluorescence intensity increased strongly as Ca 2ϩ was titrated into a solution of apocalgranulin C, indicating that a hydrophobic surface on calgranulin C is exposed upon Ca 2ϩ binding (Fig. 3b). The apparent dissociation equilibrium constant estimated from this experiment was ϳ0.12 mM. This value was substantially higher than the equilibrium constant estimated from the intrinsic tyrosine fluorescence study (ϳ0.02 mM; Fig. 3, a and b, insets).
Ca 2ϩ -calgranulin C dimers in crystal form have a 2:1 molar ratio of Ca 2ϩ to calgranulin C (14), which corresponds to four tightly bound Ca 2ϩ ions per dimer. Ca 2ϩ -calgranulin C hexamers require the binding of six additional Ca 2ϩ ions (6). We speculate that the binding of these additional Ca 2ϩ ions leads to conformational changes in calgranulin C that facilitate TNS binding and a corresponding further increase in TNS fluorescence. The larger apparent dissociation constant estimated from the TNS fluorescence binding experiment therefore likely results from two classes of Ca 2ϩ -binding sites with different affinities as follows: a comparatively high affinity class of Ca 2ϩ sites that binds to dimers and a comparatively low affinity class whose binding is linked to the formation of hexamers (6). Ca 2ϩ -Calgranulin C Binds to the C-type Immunoglobulin Domain of RAGE-Calgranulin C was shown to bind to sRAGE with an affinity of ϳ90 nM (15) in vitro. Competition assays showed that the binding was inhibited by a 50:1 molar excess of two alternate ligands of sRAGE, AGE albumin and amyloid ␤-peptide. Based on the result that AGEs bind to the V domain of RAGE (27). it was assumed that calgranulin C also binds to the V domain. We analyzed the binding of Ca 2ϩ -calgranulin C to the isolated V domain and did not observe any change in TNS fluorescence. To resolve an apparent contradiction with previous results (15,27) we suggest that the inhibition of calgranulin C binding to sRAGE by alternative ligands is because of structural rearrangements in sRAGE rendering the calgranulin C-binding site unavailable, rather than direct competition for the same binding site.
Similarly, we analyzed the binding of Ca 2ϩ -calgranulin C to the isolated C2 domain and did not observe any change in TNS fluorescence. In contrast, the interaction between Ca 2ϩ -calgranulin C and the construct containing tandem C-type immunoglobulin domains, C1C2, resulted in a large increase in the TNS fluorescence intensity as the concentration of C1C2 was increased (Fig. 3c). We concluded that the interaction between sRAGE and Ca 2ϩ -calgranulin C is mediated by the C1 domain. No increase in the TNS fluorescence was observed when either C1C2 was added to apo-calgranulin C or when Ca 2ϩ was added to the solution of C1C2. Therefore, the change in TNS fluorescence is because of the binding between C1C2 and Ca 2ϩ -calgranulin C. The dissociation constant estimated for this interaction is Ͼ70 nM, consistent with the reported value of 90 nM (15).
Stoichiometry of Ca 2ϩ -Calgranulin C-sRAGE Complexes-We used native PAGE to estimate the stoichiometry of Ca 2ϩ -calgranulin C-sRAGE complexes in solution (Fig.  2b). Because the dissociation constant for the formation of Ca 2ϩ -calgranulin C-sRAGE complexes is submicromolar, and the concentrations of Ca 2ϩ -calgranulin C and sRAGE were Ͼ10 M, stable complexes should be evident on a native gel. However, these estimates are highly qualitative because of inaccuracies in determining molecular weights of calgranulin C oligomers.
At calgranulin C to sRAGE molar ratios of 5:1 and greater, several species with higher molecular weights than either component are evident (Fig. 2b). Coomassie staining defined a  FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6 broad zone consistent with a dissociating system corresponding to a molecular mass range of ϳ300 to ϳ550 kDa. The largest species, observed at molar ratios of Ն10:1, barely enters the gel and corresponds to a molecular mass of ϳ640 kDa; however, this band lies outside the zone of staining and was not evident in all gels. At the highest molar ratios examined within the zone of staining, two bands with apparent molecular masses of ϳ545 and ϳ300 kDa were evident. These species are consistent with a tetramer of sRAGE bound to four hexamers and one hexamer of calgranulin C, respectively. At lower molar ratios, the upper limit of molecular weight decreases as expected, because a molar ratio of at least 6:1 is required to form the fully ligated complex. The zone of staining therefore likely represents a distribution of sRAGE bound to one or more hexamers of calgranulin C at molecular masses above 140 kDa and free sRAGE and calgranulin C hexamers and dimers below 140 kDa.

Sequence-specific Backbone Resonance Assignments of Apoand Ca 2ϩ -loaded Calgranulin C-Atomic resolution informa-
tion about the interaction surface between calgranulin C and sRAGE can be obtained by using heteronuclear NMR spectroscopy (33,41). To facilitate the study of the binding between sRAGE and calgranulin C, we first had to assign the chemical shifts of apo-and Ca 2ϩ -bound calgranulin C. Backbone assignments of human calgranulin C were made using uniformly labeled 15 N, 13 C-labeled protein, overexpressed in E. coli using a T5 expression system (42). The purified protein has a primary sequence of 92 amino acids with the first methionine being removed by bacterial aminopeptidase. Sequential assignments of the apo-form of calgranulin C were achieved using HNCA/ HN(CO)CA, HNCO/HN(CA)CO, and HN(CA)CB/HNCO-CACB pairs of triple resonance NMR experiments (43) so that connectivities could be traced through three independent, through-bond pathways. Overall, 96% of the 1 H, 13 C, 13 CO, and 15 N backbone resonances were sequentially assigned (Fig. 4a). The incompleteness of the assignments can be traced to broadening of the peaks because of amide proton exchange processes in the loop structures. Similar broadening of the amide protons in loop stretches was observed for other members of S100 family (44 -46). Despite these gaps in backbone assignments, the majority of backbone resonance assignments are complete for the apo-form of calgranulin C, providing site-specific reporters for ligand binding.
The secondary structure of apo-calgranulin C shown in Fig. 5 is based on the PECAN algorithm (47) that utilizes both the amino acid sequence and NMR chemical shifts to determine elements of protein secondary structure. The secondary structure of apo-calgranulin C closely matches the structures of S100A1, S100A6, and S100B for which high resolution structures are known (36,48,49). The primary structures of calgranulin C, S100A1, S100A6, and S100B are highly similar, with 40% sequence identity. The secondary structure data from the NMR studies, in combination with sequence analysis, suggests that the overall fold of calgranulin C is similar to that of S100A1, S100A6, and S100B, retaining all characteristic features of the S100 family.
Ca 2ϩ -loaded calgranulin C was prepared by titrating Ca 2ϩ into a solution of apo-calgranulin C. Structural changes were monitored using 1 H{ 15 N} HSQC experiments (Fig. 6a). At a mole ratio of Ca 2ϩ to calgranulin C of 1:1, many of the 1 H-15 N correlations broadened below the base line; however, most of these peaks reappeared at a mole ratio of 2:1. Broadening of the 1 H{ 15 N} HSQC peaks at the mid-point of titration with Ca 2ϩ has been also observed for other S100 family proteins (46,50). This result is related to the Ca 2ϩ ion exchange between two binding sites on the intermediate time scale leading to the spectral broadening. Further increases in calcium concentration resulted in a slight uniform broadening of the peaks in the 1 H{ 15 N} HSQC spectrum, consistent with predominantly two calcium-binding sites per subunit of calgranulin C.
To understand the nature of peak broadening at Ca 2ϩ to calgranulin C molar ratios above 2:1 we performed gradient diffusion NMR experiments (30), which allowed us to measure molecular diffusion coefficients to estimate the molecular weight of calgranulin C. The diffusion coefficient for apo-calgranulin C was found to be (1.25 Ϯ 0.04) ϫ 10 Ϫ10 m 2 s Ϫ1 consistent with the dimeric form of the protein in solution (21 kDa). Ca 2ϩ -bound calgranulin C has a diffusion coefficient of (0.93 Ϯ 0.07) ϫ 10 Ϫ10 m 2 s Ϫ1 . This 30% decrease in the diffusion coefficient can be due to two factors as follows: a conformational change in calgranulin C caused by Ca 2ϩ complexation and/or the appearance of higher oligomeric states of Ca 2ϩ -calgranulin C. The total areas of the apo-calgranulin C (based on our homology model) and Ca 2ϩ -calgranulin C (PDB code 1E8A) molecular surfaces are 6496 and 6609 Å 2 , respectively. Such a difference, amounting to a 3% increase in total surface area for the Ca 2ϩ -bound state, cannot solely account for a 30% increase in the apparent molecular weight suggesting that larger oligomers of Ca 2ϩ -calgranulin C are formed in solution. This result is consistent with our observation of calgranulin C hexamers in solution.
Because addition of calcium resulted in large chemical shift changes in the NMR spectrum of calgranulin C, it was necessary to reassign the resonances of Ca 2ϩ -calgranulin C a priori using the same set of triple resonance experiments. Ca 2ϩ -calgranulin C is a homodimer, and 85 of the expected 93 correlations in the 1 H{ 15 N} HSQC were observed and assigned (Fig. 6a). The eight missing peaks, Lys 24 through Ser 31 , were presumed to be broadened because of intermediate exchange and are located in the first loop of calgranulin C.
The large chemical shift changes seen upon adding calcium to apo-calgranulin C indicate substantial structural perturbations. It is known that in S100A1, S100A6, and S100B, helix 3 undergoes a significant rearrangement upon calcium binding (4,50). The signature of this rearrangement is the formation of a strong hydrogen bond between the amide proton of Glu 66 and the side chain carbonyl group of Asp 62 in helix 3 seen in the crystal structure of Ca 2ϩ -calgranulin C. Upon adding calcium, the amide proton of Glu 66 undergoes a large downfield chemical shift, which is consistent with the formation of this hydrogen bond.  . Primary sequences and secondary structure homologies between S100 proteins. The extent of helices 1-4 in apo-calgranulin C is shown above the primary sequence. Sequence number is indicated on the right. Homologous residues involved in target binding are highlighted in red. Residues implicated in binding additional Ca 2ϩ during hexamerization are highlighted in blue. GenBank TM accession numbers for the shown sequences are NM005621 (Hsu S100A12), NM174651 (Bta S100A12), NM005621 (Hsu S100A11), NM002966 (Hsu S100A10), NM014624 (Hsu S100A6), NP006262 (Hsu S100A1), and NM006272 (Hsu S100B).  (Figs. 4b and 6b). Titrating calgranulin C with individually expressed V and C2 domains did not result in any changes in chemical shifts of [U-15 N]calgranulin C with or without Ca 2ϩ . On the other hand, titrating calgranulin C with both full-length sRAGE and C1C2 resulted in almost identical changes in chemical shifts and line broadening. This supports our assertion that calgranulin C binding is mediated by the first C-type domain of RAGE, C1.

Chemical Shift Changes upon Forming the Apo-calgranulin C-sRAGE Complex-Structurally
Because of the limited solubility of sRAGE, we used only C1C2 to characterize the interaction between sRAGE and calgranulin C in detail. The calgranulin C-C1C2 complexes have a large (Ͼ200 kDa) molecular weight and, probably, are invisible by NMR. Nevertheless, they can be studied indirectly by observing the changes in the position and line width of the free calgranulin C chemical shifts because of the exchange with the calgranulin C-C1C2 complex (see "Experimental Procedures"). Chemical exchange detected by NMR spectroscopy is generally characterized by three regimes: fast exchange when the characteristic exchange rate kЈ ex is much larger than the changes in chemical shifts between free and bound forms ⌬; intermediate exchange when ⌬ Ϸ k ex ; and slow exchange when ⌬ is much larger than k ex (41,51,52). In the fast exchange regime, even in the case of large protein-protein complexes, we should see gradual changes in the chemical shifts of the calgranulin C nuclei affected by protein-protein binding. In the slow and intermediate exchange regime, we will see differential broadening of the calgranulin C NMR signals originating from the affected nuclei (33,53).
Titrating C1C2 into [U-15 N]apo-calgranulin C resulted in small gradual changes in a subset of apo-calgranulin C amide proton and nitrogen chemical shifts (Fig. 4b). These changes are consistent with the exchange between free and complexed forms of apo-calgranulin C occurring on the fast NMR time scale. Based on these changes, we estimated that the equilibrium-binding constant of C1C2 to apo-calgranulin C is about 140 M.
The structural elements of apo-calgranulin C affected by the binding of C1C2 include loop 1 (Gly 25 -Thr 29 ), the N terminus of helix 4 (Asp 72 -Phe 76 ), and the C terminus (His 88 -Lys 93 ). Loop 1 and the N terminus of helix 4 of one subunit and the C terminus of the other subunit define each symmetric binding site. Loop 1 together with helices 1 and 2 form the first EF-hand motif. We mapped these elements onto a model structure for apo-calgranulin C (Fig. 7a). The apo-S100 structure is highly conserved among S100 members (1,4). Using homology modeling and energy minimization, a model of apo-calgranulin C was generated on the basis of the apo-S100A6 (36) structure, the alignment of helices determined by NMR methods, and the alignment of conserved amino acids. Fig. 7a demonstrates that all amino acids greatly affected by sRAGE binding are located on two symmetric faces of the apo-calgranulin C dimer providing evidence for two unique, contiguous binding sites for sRAGE. The interaction map suggests that the chemical shift perturbations largely reflect a contact surface, not a global conformational change. As expected, in the absence of Ca 2ϩ , the interaction surface of apo-calgranulin C is highly negatively charged suggesting that the binding energy is predominantly defined by electrostatic interactions. Indeed, increasing the Chemical Shift Changes upon Forming the Ca 2ϩ -Calgranulin C-C1C2 Complex-Titrating C1C2 into [U-15 N]Ca 2ϩ -calgranulin C led to a dramatic decrease in the intensity of the NMR signal (Fig. 6b). At a molar ratio of Ca 2ϩ -calgranulin C to C1C2 of ϳ6:1, the NMR signal of free Ca 2ϩ -calgranulin C largely disappeared. At higher molar ratios we observed differential broadening of a subset of the amide peaks from the 1 H{ 15 N} HSQC spectrum of Ca 2ϩ -calgranulin C. Because the molecular weight of the Ca 2ϩ -calgranulin C-C1C2 complex is greater than 200 kDa, we expect the signals arising from the bound state of Ca 2ϩ -calgranulin C to be absent from the spectrum. This result is consistent with hexameric calgranulin C binding to sRAGE. When Ca 2ϩ -calgranulin C is present in excess, the lines of the free form are broadened because of the exchange with the bound state. Because resonances at the binding interface are expected to experience the largest chemical shift changes, this effect is used to map binding interactions (33,53,54).
To measure the degree of line broadening because of complex formation we used the normalized cross-peak heights (54) shown in Equation 4, where the cross-peak heights and the average values of the cross-peak heights of Ca 2ϩ -calgranulin C are defined in the absence of C1C2, H 0 and H 0 , and at a given Ca 2ϩ -calgranulin C to C1C2 ratio, H RAGE and H RAGE , respectively. In this case, large positive values of ⌬ indicate the resonances that are significantly broadened compared with the average value. Small positive and negative values of ⌬ indicate that the resonances are either broadened to the same extent or less than the average.
We plotted the normalized cross-peak heights of the 1 H{ 15 N} HSQC spectrum of calgranulin C at a stoichiometric ratio of Ca 2ϩ -calgranulin C to C1C2 of 10:1 for the assigned residues in Ca 2ϩ -calgranulin C (Fig. 8). This stoichiometric ratio was chosen as a compromise between the high signal to noise ratio of the NMR signal and the large differential broadening effect. Four stretches of the primary calgranulin C sequence are strikingly affected by C1C2 binding as follows: Thr 1 -Leu 3 and Leu 7 -Gly 9 in helix 1; Gly 31 and Leu 35 -Thr 37 in helix 2; Asn 42 -Ile 44 in loop 2, and; Ala 80 -Thr 88 in the C terminus of helix 4 and the tail of calgranulin C (Fig. 8). These secondary structural elements form two contiguous, symmetric surfaces on the Ca 2ϩ -calgranulin C dimer, each surface consisting of helix 2, loop 2, and helix 4 of one subunit and helix 1 of the other (Fig. 7b). Importantly, the identified structural elements are located on the outside surface of the hexameric form of Ca 2ϩ -calgranulin C (Fig. 9). We propose that these structural elements identify a molecular surface directly involved in the calgranulin C-sRAGE interaction, thus substantiating our observation that it is the calgranulin C hexamers that bind to sRAGE in solution.
The binding site for C1C2 on Ca 2ϩ -calgranulin C is predominantly polar and hydrophobic, decorated by a negatively charged patch formed by Glu 5 and Glu 9 from helix 1 (Figs. 7b and 10), and is vastly different from that of apo-calgranulin C (Fig. 7a), which is dominated by negative charges. In Ca 2ϩcalgranulin C, the calcium ions are coordinated by the peptide carbonyls of Ser 18 , Lys 21 , His 23 , and Thr 26 , and carboxylate group of Glu 31 located in loop 1 and carboxylate groups of Asp 61 , Asn 63 , Asp 65 , Glu 72 , and peptide carbonyl group of Gln 67 located in loop 3 (14). Loops 1 and 3 undergo a change upon calcium binding, leading to the dramatic repositioning of helix 2 and 3. It also creates a shallow hydrophobic pocket formed by helices 3 and 4 and loop 2. Thus, Ca 2ϩ binding leads to a restructuring of the protein surface and redistribution of the surface charge that affects primarily the area(s) involved in C1C2 binding.
Helix 4 and the C-terminal tail are the only common elements between the apo-calgranulin C and Ca 2ϩ -calgranulin C-binding sites for C1C2. The relative position of helix 4 is conserved in both structures, possibly explaining its involvement with both the apo-and Ca 2ϩ -calgranulin C-binding sites.
Among the residues that show significant differential broadening, but are not on a continuous Ca 2ϩ -calgranulin C-C1C2 interaction surface, one region seems important; residues Ala 62 and Gln 64 exhibit strong differential broadening and yet are located in loop 3 away from the identified binding site (Figs. 7b and 8). We propose that this broadening is because of preferential formation of hexamers of Ca 2ϩ -calgranulin C upon binding to C1C2. Using the crystal structure of hexameric calgranulin C (6), we can distinguish between the molecular surfaces responsible for hexamerization and those responsible for direct binding to sRAGE (Fig. 9). In the crystallographic study (6), it was shown that Ca 2ϩ -calgranulin C can exist as a hexamer arranged as a trimer of dimers, with each dimer having only  (14)). Structural elements involved in creating the two symmetric, contiguous C1C2-binding surfaces (red) include helix 2, loop 2, and helix 4 of one subunit and helix 1 of the other subunit. Loop 3 may be involved in hexamerization and the associated binding of additional Ca 2ϩ . Yellow spheres represent Ca 2ϩ ion-binding sites on dimeric calgranulin C. minimal differences from that of previously solved Ca 2ϩ -calgranulin C dimers (14). The root mean square deviation in the main chain atoms between the dimeric subunits of the hexamer (PDB code 1GQM) and the Ca 2ϩ -calgranulin C dimer (PDB code 1E8A) is less than 0.5 Å (6). In the crystal, the hexamer is formed by coordinating additional Ca 2ϩ through the side chain oxygens of Gln 58 and Gln 64 from one dimer and through bidentate bonding to the carbonyl of Glu 55 from the adjacent dimer. In addition, Glu 66 from the first dimer moves closer to the new calcium-binding site. These changes, affecting loop 3, may result in the observed broadening of the backbone amides of Ala 62 and Gln 64 .

DISCUSSION
Identifying the interaction surface between C1C2 and Ca 2ϩcalgranulin C provides an opportunity to understand the interactions of RAGE receptor with one of its physiological ligands. RAGE interacts with a wide range of ligands leading to diverse cellular responses from cytokine secretion and increased cellular oxidant stress to cell survival, differentiation, and proliferation (55). RAGE binds to AGEs, calgranulins, HMGB1, and ␤-sheet fibrils (15) that structurally have very little in common. RAGE is classified as a pattern recognition receptor (56), underscoring its ability to bind to classes of molecules rather than individual ligands. We showed that unlike AGEs, which bind to the V domain of RAGE (27), calgranulin C binds to the C1C2 tandem construct. The presence of the C1 domain is required for binding because the isolated C2 domain does not interact with calgranulin C. Because most multiligand receptors consist of more than one domain, the concept of one domain-one ligand may provide an explanation to the paradox of multiligand molecular recognition.
Our study and those of others (57) showed that even unligated sRAGE oligomerizes, and this oligomerization is mediated by self-association of the C1 domain. There are examples of other receptors, e.g. tumor necrosis factor superfamily of receptors (58,59) and the erythropoietin receptor (60), that form homo-oligomers in the inactive state. Our experiments suggest that sRAGE forms a tight tetramer that may bind up to four hexamers of calgranulin C. We propose that full-length RAGE receptor also forms a homotetramer.
Oligomerization of RAGE receptor provides an explanation for its ability to bind a diverse set of protein ligands possessing AGE modifications in vitro. For example, the AGE-modified amino acid, carboxymethyllysine, binds to the V domain of RAGE receptor with millimolar affinity, 3 whereas AGE-modified albumin binds to RAGE with nanomolar affinity in these assays (61). To reconcile these observations, we propose that the clustering of V domains in tetrameric RAGE promotes cooperative binding of AGE-modified albumin. The increase in binding affinity of albumin over that of carboxymethyllysine is because of the diversity of the side chain residues that bind with affinities usually ascribed to nonspecific interactions. Because both AGE-modified proteins and calgranulin C bind to different domains of RAGE, these two ligands may work synergistically to amplify pro-inflammatory signaling through RAGE (62).
We showed that the apo-form of calgranulin C binds to sRAGE with a low (140 M) affinity, and we identified three structural elements responsible for this interaction as follows: the N and C termini and loop 1 of calgranulin C. This interaction likely takes place inside the cells where the concentration of calcium is on the order of 100 nM (63) and calgranulin C is, preferentially, in the apo-form. Intracellular RAGE was also found, although the significance of calgranulin C-RAGE interactions inside cells remains unknown (64). However, given the 3 A. Shekhtman, unpublished results. ) indicate substantial perturbation of residues involved in C1C2 binding. Gray bars represent unassigned residues. Middle, substantial line broadening for residue Asn 43 (N43) upon complex formation. Free [U-15 N]Ca 2ϩ -calgranulin C is shown in black. Ca 2ϩ -calgranulin C to C1C2 at a molar ratio of 10:1 is shown in gray. Bottom, absence of substantial line broadening for residue Glu 40 (E40) upon complex formation. Free [U-15 N]Ca 2ϩ -calgranulin C is shown in black. Ca 2ϩ -calgranulin C to C1C2 at a molar ratio of 10:1 is shown in gray.
comparatively weak dissociation constant and the fact that the binding is mediated primarily by electrostatic contacts, this interaction is probably nonspecific and may not have physiological relevance.
In contrast to apo-calgranulin C, the restructuring of the protein surface and redistribution of the surface charge upon calcium binding results in an ϳ1000-fold increase in calgranulin C binding affinity (140 M versus 90 nM), underscoring the importance of Ca 2ϩ binding as a molecular switch. Four elements of the calgranulin C structure are affected by C1C2 binding: the N and C termini of helix 1; the C terminus of helix 2; loop 2; and the C terminus of helix 4 and the tail of calgranulin C. These secondary structural elements form two contiguous, symmetric surfaces on the Ca 2ϩ -calgranulin C dimer, each surface consisting of helix 2, loop 2, and helix 4 of one subunit and helix 1 of the other (Fig. 7b).
Our characterization of the sRAGE-Ca 2ϩ -calgranulin C complex suggests that Ca 2ϩ -calgranulin C binds to sRAGE in a highly oligomerized form. Based on our studies we propose that Ca 2ϩ -calgranulin C binds predominantly as a hexamer. The identified sRAGE-binding site lies on the outer surface of the hexamer, giving the hexamer the capacity to make multiple interactions with the extracellular domain of RAGE (Fig. 9). In our experiments the calcium concentration was 2 mM, which is close to the extracellular concentration, and therefore supports the proposed hexamerization of calgranulin C. Intracellular calcium may also reach low millimolar concentrations during specific activation events (65). In this case, calcium-dependent calgranulin C-RAGE activation during intracellular translocation may play a role in cell physiology or pathology (64).
The interaction surface on Ca 2ϩ -calgranulin C for C1C2 binding strongly resembles the binding site for annexin II binding to S100A10 (10) and for annexin I binding to S100A11 (11,66) (Fig. 10). In these complexes, a short peptide from annexin forms an amphiphilic helix upon binding to S100 protein. Most of the hydrophobic residues of annexin I or II are in contact with helix 4 and loop 2, whereas the hydrophilic residues of the peptide are involved in hydrogen bonding with helix 1. All these regions of Ca 2ϩ -calgranulin C show significant differential broadening upon complexation with sRAGE. There is a high degree of conservation of hydrophobic and hydrophilic residues across the primary structures of S100 family members (Fig. 5), which may explain their ability to effectively interact with RAGE. A detailed comparison of the Ca 2ϩcalgranulin C-binding site to that of S100A10 in complex with annexin II reveals that Glu 4 and Glu 8 of Ca 2ϩ - FIGURE 9. Molecular surface of calgranulin C hexamer (PDB code 1GQM (6)) affected by sRAGE binding. a, calgranulin C hexamer (PDB code 1GQM (6)) showing the symmetry of the binding sites for RAGE (colored regions). b, calgranulin C hexamer looking down the 3-fold symmetry axis; calgranulin C dimer-binding sites for RAGE are shown in color. FIGURE 10. Comparison between S100A10-annexin II and Ca 2؉ -calgranulin C-RAGE-binding sites. a, S100A10 complexed with annexin II (light blue) from the crystal structure PDB code 1BT6 (10). Associated residues of S100A10, Glu 6 , Glu 10 , and Asn 84 involved in peptide binding are in pink. b, electrostatic surface of Ca 2ϩ -calgranulin C dimer showing areas of positive (blue) and negative (red) potential. c, interaction map showing the C1C2 binding surface (red) on Ca 2ϩ -calgranulin C dimer.
calgranulin C are in the same conformation as in the S100A10 complex. In the S100A10-annexin II complex homologous Glu 9 and Glu 13 form hydrogen bonds with the target peptide. We suggest that the binding of Ca 2ϩ -calgranulin C and C1C2 closely resembles that of the S100A10-annexin II complex. There is increasing evidence that receptor oligomerization provides a general mechanism for signal transduction in cytokine receptors. Signal transduction by receptor trimerization has been suggested for the neural cell adhesion molecule L1, which is related to RAGE (67). We propose that hexamerization of calgranulin C facilitates further oligomerization of the RAGE receptor. Engagement of hexameric calgranulin C by tetrameric RAGE then creates a large signaling platform to effectively amplify intracellular signal transduction.