The amino-terminal domains of the ezrin, radixin, and moesin (ERM) proteins bind advanced glycation end products, an interaction that may play a role in the development of diabetic complications.

The presence of advanced glycation end products (AGEs) formed because of hyperglycemia in diabetic patients has been strongly linked to the development of diabetic complications and disturbances in cellular function. In this report, we describe the isolation and identification of novel AGE-binding proteins from diabetic rat kidneys. The proteins were purified by cation exchange and AGE-modified bovine serum albumin (AGE-BSA) affinity chromatography. NH2-terminal and internal sequencing identified the proteins as the NH2-terminal domains of ezrin, radixin, and moesin (ERM proteins). Using BIAcore biosensor analysis, human N-ezrin-(1-324) bound to immobilized AGE-BSA with a KD of 5.3 +/- 2.1 x 10 -7 m, whereas full-length ezrin-(1-586) and C-ezrin-(323-586) did not bind. Other glycated proteins such as AGE-RNase, N in -carboxymethyllysine (CML)-BSA, and glycated human serum albumin isolated from hyperglycemic diabetic sera competed with the immobilized AGE-BSA for binding to N-ezrin, but non-glycated BSA and RNase did not. Thus N-ezrin binds to AGEs in a glycation- and concentration-dependent manner. Phosphorylated ezrin plays a crucial role in cell shape changes, cell attachment, and cell adhesion. The effect of AGE-BSA on ezrin function was studied in a tubulogenesis model in which LLC-PK1 cell tubule formation is dependent on phosphorylated ezrin. Addition of AGE-BSA completely inhibited the ability of the cells to produce tubules. Furthermore, in vitro tyrosine phosphorylation of N-ezrin and ezrin was also inhibited by AGE-BSA. These proteins represent a novel family of intracellular binding molecules for glycated proteins and provide a potential new target for therapeutic intervention in the prevention or treatment of diabetic complications.

Diabetes mellitus is associated with very significant morbidity as a result of long-term complications, including nephropathy, retinopathy, neuropathy, and macrovascular disease. Diabetes is characterized by chronic hyperglycemia, which leads to an acceleration of the Maillard reaction. This is a spontane-ous reaction between glucose and proteins, lipids, or nucleic acids, particularly on long-lived proteins such as the collagens (1). A sequence of biochemical reactions, many of which are still poorly defined, leads to the formation of a range of advanced glycation end products (AGEs). 1 Although research interest has focused predominantly on the measurement and binding interactions of extracellular AGEs, recent studies have indicated that intracellular AGEs are also extremely important (2).
There is substantial evidence for a link between hyperglycemia/AGEs and the pathogenesis of diabetic complications such as retinopathy, nephropathy, neuropathy, and vasculopathy (3). First, levels of AGEs are correlated with the severity of these complications (4). Second, in the non-diabetic mouse, injection of AGEs prepared in vitro leads to histological changes resembling diabetic nephropathy, including mesangial expansion and glomerulosclerosis (5). Finally, pharmacological inhibitors of AGE-dependent pathways retard diabetic complications. Aminoguanidine and related compounds prevent AGE formation (6) and have been shown in experimental models of diabetes to reduce tissue AGE levels and retard the development of neuropathy, retinopathy, and nephropathy (7)(8)(9). Although it is well established that hyperglycemia and the presence of AGEs on proteins contribute to complications, the pathways for this process have not been clearly determined.
Previous studies have shown that AGEs prepared in vitro bind to cultured cells via cell surface receptors or binding proteins (10), resulting in binding and internalization of AGEreceptor complexes and proteolytic processing of the AGE ligand. Binding of AGE ligands to cells can also modulate cell function. For instance, cell activation with increased expression of extracellular matrix proteins, adhesion molecules, cytokines, and growth factors has been described (10). Several AGE-binding proteins have been identified, including RAGE, AGE-R1, AGE-R2, AGE-R3, and macrophage scavenger receptors 1 and 2 (11)(12)(13)(14). Whether all these proteins bind AGEmodified proteins in vivo is not yet clear.
We have observed increased binding of AGEs to diabetic rat kidney sections, which are reduced in animals treated with aminoguanidine (15). The major binding site did not appear to be any of the known AGE receptors. These findings led us to undertake the isolation and characterization of novel binding sites for AGE-modified proteins present in the diabetic rat kidney. We report the purification from diabetic rat kidneys of three proteins, i.e. truncated forms of ezrin, radixin, and moesin (ERM). Members of the ERM family of proteins are structurally very similar and are present on cell surface protrusions. They link the actin cytoskeleton of the cell to the plasma membrane and are thought to play a role in cell shape change, motility, and adhesion (16). Our studies identify ERM proteins as novel binding proteins for AGEs, and we hypothesize that the interaction between ERMs and AGEs may play a role in the development of diabetic complications.

EXPERIMENTAL PROCEDURES
With the exception of calpain inhibitor I, which was from ICN Pharmaceuticals, protease inhibitors and actin were purchased from Sigma. Protein concentrations were determined by the DC protein assay (Bio-Rad). Restriction enzymes, DNA grade agarose, and reagents for reverse transcription and PCR were from Promega, Melbourne, Australia. The pProEX-HT prokaryotic expression system was from Invitrogen, and Ni-NTA-agarose was from Qiagen. Oligonucleotides were prepared to PCR grade by Geneworks, Thebarton, Australia.
Preparation and Radiolabeling of AGE-BSA-AGE-modified bovine serum albumin (AGE-BSA), AGE-RNase, and N ⑀ -carboxymethyllysine (CML)-BSA were prepared as described previously (17). The extent of chemical modification of lysine residues was determined using 2,4,6trinitrobenzenesulfonic acid by the method described (18). The extent of modification was 17% for CML-BSA and 77% for AGE-modified proteins. AGE-BSA iodination was performed using the chloramine T method (19) to a specific activity of ϳ10,000 cpm/ng protein. 125 I-AGE-BSA was separated from unbound 125 I with a P6DG desalting gel column (Bio-Rad).
Preparation of AGE-BSA Affinity Column-An AGE-BSA affinity column was prepared by coupling AGE-BSA (5 mg/ml) to Affi-Gel 15 according to the manufacturer's instructions (Bio-Rad). The concentration of ligand bound was ϳ5 mg/ml resin, and a 1.7 ml column was prepared.
Fresh protease inhibitors were added, and the soluble kidney extract (48 ml) was applied to a BioRex-70 cation-exchange column (55 ml) at a flow rate of 0.5 ml/min. The column was washed extensively with 20 mM Tris buffer, pH 7.4, containing 50 mM NaCl and 0.1% Triton X-100 until the absorbance at 280 nm was less than 0.02. Bound protein was eluted with a salt gradient from 200 mM to 1 M NaCl in 20 mM Tris buffer, pH 7.4, containing 0.1% Triton X-100. Fractions were analyzed for AGE-BSA binding activity by slot blotting (see below).
The active fractions were pooled and dialyzed against 20 mM Tris, pH 7.4, containing 50 mM NaCl, 0.1% Triton X-100, and protease inhibitors and applied to the AGE-BSA affinity column. The column was extensively washed and the proteins eluted with a salt gradient. Fractions (1 ml) were collected, and absorbance at 280 nm, salt concentration (osmolarity), and AGE-BSA binding activity were measured.
SDS-PAGE-Purified AGE-BSA binding proteins were analyzed by non-reducing SDS-12.5% PAGE according to the method of Laemmli (20). Proteins were visualized with Coomassie Brilliant Blue R. Molecular masses were estimated from semilogarithmic plots of the migration of standard proteins under reducing conditions run simultaneously.
Ligand and Slot Blotting-Proteins in gels were transferred to nitrocellulose membranes (Osmonics, Westborough, MA) at 20 V for 16 h at 4°C. Nonspecific binding sites were blocked by a 1 h incubation with 20 mM Tris, pH 7.4, containing 50 mM NaCl, 0.1% Tween, and 1% BSA (blocking buffer). The nitrocellulose membrane was probed with 125 I-AGE-BSA (1 ϫ 10 6 cpm/ml) for 2 h at room temperature in blocking buffer in the presence or absence of unlabeled AGE-BSA. The membrane was washed thoroughly and exposed to Kodak BioMax MS film.
Slot blotting was performed similarly to ligand blotting, except the proteins were immobilized directly onto nitrocellulose membranes under vacuum and incubated with 125 I-AGE-BSA (1 ϫ 10 6 cpm/ml) for 2 h at room temperature in blocking buffer in the presence or absence of unlabeled AGE-BSA. Individual slots were counted in a ␥-counter for quantitation of bound radioactivity.
Western Blotting-Proteins in non-reducing SDS-PAGE sample buffer were separated by SDS-12.5% PAGE and transferred to nitrocellulose membranes. Membranes were probed with ezrin monoclonal antibody (BD Transduction Laboratories). Reactive proteins were visualized using Super Signal chemiluminescent substrate (Pierce).
Preparative Electrophoresis for Amino Acid Sequencing-AGE-binding fractions eluted from the AGE-BSA affinity column were pooled and concentrated 5-fold (Centricon, Millipore). Protein (50 g) was incubated at 37°C for 15 min in reducing buffer (0.25 M Tris-HCl, pH 6.8, 1% SDS, glycerol, 0.015% bromphenol blue, and 0.2 M 2-mercaptoethanol), subjected to SDS-12.5% PAGE, transferred to a polyvinylidene difluoride membrane for 16 h at 20 V at 4°C, and stained with 0.025% Coomassie Brilliant Blue R in 40% methanol. Bands were excised and subjected to NH 2 -terminal sequencing of the first 15 amino acids. Sequences obtained were compared with the Swiss-PROT data base using the basic BLAST software program.
To obtain internal sequences, concentrated AGE-BSA binding proteins were reduced with 0.1 M 2-mercaptoethanol for 2 h at 37°C, lyophilized, and resuspended in 70% formic acid. Reduced proteins were cleaved by the addition of cyanogen bromide (CNBr) crystals and incubation under nitrogen in the dark for 18 h. The cleaved proteins were lyophilized to remove CNBr and formic acid residues prior to preparative electrophoresis, transferred to polyvinylidene difluoride membrane, and sequenced as above, using low molecular mass markers to identify the desired bands in the 5-10 kDa range.
Cloning of ERM Constructs-Total RNA was extracted from LIM 2405 human colon cancer cells and reverse transcribed using Super Script II reverse transcriptase (Invitrogen). The resultant cDNA mix was used as template for PCR using Pfu DNA polymerase to prepare cDNAs for ezrin (1-1758 bp), N-ezrin (1-971 bp), and C-ezrin (952-1758 bp). Primers additionally encoded a 5Ј EcoR1 site and a 3Ј HindIII restriction site. The PCR products were doubly digested with EcoR1/ HindIII, isolated from 1% agarose gels, and cloned into the corresponding restriction sites in the expression vector pProEXHT a to express the protein fused to an amino-terminal six-histidine sequence (His 6 ). DNA sequencing confirmed the correct sequences of the inserts.
Expression and Purification of Recombinant ERM Proteins-Competent Escherichia coli (JM109) was transformed with the ezrin, N-ezrin, or C-ezrin expression vectors, and expression was induced with 100 M isopropyl ␤-D-thiogalactopyranoside (IPTG) for 3 h at 37°C. Each recombinant protein was purified by Ni-NTA chromatography. Cleared cell lysate was prepared as described (Qiagen) and applied to a 1-ml Ni-NTA-agarose column, washed thoroughly with high salt wash buffer (20 mM Tris, pH 8.0, containing 300 mM NaCl, 0.1% Tween, 20 mM imidazole, and inhibitors (phenylmethylsulfonyl fluoride, calpain inhibitor 1, leupeptin, and Trasylol)) and also with low salt buffer (20 mM Tris, pH 8.0, containing 150 mM NaCl, 0.1% Tween, 20 mM imidazole, and inhibitors) before eluting with low salt buffer containing 250 mM imidazole. Isolated recombinant proteins were dialyzed to remove imidazole prior to further studies.
Preparation of Human Serum Albumin (HSA) from Diabetic Patient Serum-HSA was isolated from sera of two hyperglycemic patients with HbA1 C values Ͼ9.5% (normal Յ 6%) using a 1-ml Affi-Gel Blue affinity column. HSA was eluted with 1 M NaCl, concentrated, and dialyzed against Hepes-buffered saline (10 mM Hepes, pH 7.4, containing 150 mM NaCl and 0.005% P20) before biosensor experiments.
BIAcore Biosensor Analysis-AGE-BSA binding studies were carried out on each of the isolated recombinant full-length and truncated ezrin proteins using surface plasmon resonance on a BIAcore 2000 biosensor (BIAcore AB).
A CM5 research grade sensor chip was activated with N-ethyl-N-(3dimethylaminopropyl)carbodiimidehydrochloride/N-hydroxysuccinimide (EDC/NHS), and AGE-BSA was injected in coupling buffer over the active surface until a suitable resonance unit (RU ϭ 650) value was obtained (1 RU equates to 1 pg of protein bound/mm 2 of sensor surface). Remaining activated groups were deactivated by injection of ethanolamine. BSA was immobilized onto another channel of the same chip (RU ϭ 1895) to subtract background binding and identify specific binding to glycated moieties in AGE-BSA. Ezrin, N-ezrin, and C-ezrin (150 nM-2.4 M in Hepes-buffered saline) were passed over immobilized AGE-BSA for 240 s at a flow rate of 10 l/min. The infusion was stopped and dissociation was observed over 200 s, during which time Hepes-buffered saline alone was passed over the chip. The binding curves were analyzed using BIA Evaluation 3.0 software. For competition experiments, mixtures of proteins were applied to the chip. In these experiments, varying concentrations of glycated proteins (AGE-BSA, HSA isolated from diabetic sera, AGE-RNase, and CML-BSA) and non-glycated proteins (BSA, RNase, or C-ezrin) were mixed with N-ezrin for at least 30 min before passing over the chip. In other experiments, insulinlike growth factor binding protein 6, (IGFBP-6) (21), an unrelated protein of similar size to the truncated ezrin proteins, was used as a further control for possible nonspecific binding or steric hindrance effects.
Tubulogenesis-Cells from the porcine kidney epithelial cell line LLC-PK1 were cultured in 24-well plates in DMEM containing 10% FCS. Confluent LLC-PK1 cells were stimulated to form tubules by the addition of DMEM/10% FCS containing 50% fibroblast 3T3 spent conditioned medium and type I collagen (120 g/ml). Fibroblast 3T3 spent conditioned medium was harvested from subconfluent 3T3 cultures grown for 3 days. Tubules formed were counted in each well at 24 and 48 h at 40ϫ magnification (22) (23).
In Vitro Phosphorylation-N-ezrin (2.5 g) or ezrin (15 g) were phosphorylated by epidermal growth factor receptor, EGFR (1.62 units), in kinase buffer (20 mM Hepes buffer, pH 7.4, containing 1 mM dithiothreitol, 100 M ATP, 5 mM MgCl 2 , and 5 mM MnCl 2 ) in the presence of BSA or AGE-BSA (30 M) for 30 min at 30°C. The reaction was stopped by the addition of SDS-PAGE sample buffer and boiling, and the products were analyzed by SDS-PAGE. The phosphorylated ezrin substrates were detected by Western blotting using an antiphosphotyrosine monoclonal antibody conjugated to horseradish peroxidase (HRP, BD Transduction Laboratories) and enhanced chemiluminescence detection. Stripping and reprobing with an anti-ezrin antibody verified equal loading of N-ezrin and ezrin. Band intensities were quantitated by densitometry using a microcomputer imaging device (MCID) and software (Imaging Research Inc., St. Catharines, Ontario, Canada). Results were corrected for any differences in loading and expressed as a percentage of control BSA sample (n ϭ 3).

RESULTS
To purify and identify the AGE binding sites in the diabetic kidney, a kidney extract was prepared and chromatographed on a cation exchange BioRex-70 column. A broad peak of AGE-BSA binding activity eluted between 0.2 and 1.0 M NaCl (Fig.  1A). The active fractions (21-52) were pooled and dialyzed prior to application to an AGE-BSA affinity column. One major peak of AGE-BSA binding activity was eluted with 200 mM NaCl (Fig. 1B, 17-25). SDS-PAGE of this peak demonstrated a triplet of bands sized at 38 -45 kDa (Fig. 1C).
When the electrophoresed proteins were transferred to nitrocellulose and subjected to ligand blotting, all three bands bound 125 I-AGE-BSA ( Fig. 2A, lane 1). The binding of 125 I-AGE-BSA to the triplet was completely inhibited by 300-fold excess unlabeled AGE-BSA ( Fig. 2A, lane 2) but not by 300-fold excess unmodified BSA (Fig. 2A, lane 3), showing that the binding was specific for glycated BSA.
The binding affinity of the isolated proteins for AGE-BSA was assessed by a slot blot competition experiment. With a constant protein amount (10 g) bound to nitrocellulose membranes and nonspecific binding blocked with BSA, slots were incubated with 125 I-AGE-BSA alone (2 nM) or in the presence of increasing concentrations of unlabeled AGE-BSA (10 nM-10 M). Half-maximal binding was observed at a concentration of 3 ϫ 10 Ϫ8 M (Fig. 2B), which is similar to binding affinities described for other AGE receptors (11)(12)(13).
To identify the purified AGE-BSA binding-proteins, aminoterminal amino acid sequencing was performed on each of the three bands. The amino acid sequences were PK(T/P)I(S/ N)VRVTTMDAEL. Comparison with rodent sequences on the Swiss-PROT data base using a basic BLAST software program revealed that these sequences were 100% homologous to the QKQ). However, no evidence was obtained for the presence of the next CNBr fragment (aa 319/322 onwards). Thus, the isolated proteins that bind AGE-BSA are the NH 2 -terminal ϳ320 amino acids of the ERM proteins, which is consistent with the size of the proteins seen by SDS-PAGE.
To define regions of ERM proteins involved in binding AGEs, the human cDNA for ezrin was cloned into the expression vector pPROEX-HT a . Three constructs with NH 2 -terminal His 6 tags were produced, i.e. full-length ezrin (aa 1-586), N-ezrin (aa 1-324), and C-ezrin (aa 323-586). All three constructs were expressed in E. coli and purified by Ni-NTA-agarose chromatography.
The isolated recombinant proteins were characterized by SDS-PAGE (Fig. 3A), Western blotting with an anti-ezrin antibody, which recognizes an epitope in the aa 391-515 region of the protein (Fig. 3B), and ligand blotting with 125 I-AGE-BSA (Fig. 3C). The anti-ezrin antibody recognized recombinant ezrin and C-ezrin but not N-ezrin, whereas 125 I-AGE-BSA bound to N-ezrin but not ezrin or C-ezrin, showing that the binding site for AGE-BSA is located between amino acids 1-324 and is masked in the full-length ezrin molecule, possibly by intramolecular interaction between the NH 2 and COOH termini (24).
Recombinant ezrin and N-ezrin were subjected to the same isolation procedure of cation exchange and AGE-BSA affinity chromatography as that employed for the diabetic kidney. Although both proteins bound and eluted from the cation exchange column at a similar salt concentration as that of the kidney ERM proteins, only N-ezrin bound and was eluted from the AGE-BSA affinity column (data not shown), confirming that full-length ezrin does not bind AGE-BSA.
Biosensor analysis of the binding of recombinant N-ezrin to immobilized AGE-BSA was performed (Fig. 4). Concentrationdependent binding of N-ezrin to AGE-BSA was observed with rapid association and dissociation kinetics. Kinetic analysis of the specific binding curves showed the K D for the reaction is 5.3 Ϯ 2.1 ϫ 10 Ϫ7 M. Recombinant N-radixin also bound to immobilized AGE-BSA with a K D of 1.8 Ϯ 0.57 ϫ 10 Ϫ6 M, which is consistent with all three truncated ERM proteins being isolated from diabetic kidneys on the basis of their ability to bind to AGE-BSA.
The specificity of the reaction was further shown by the ability of excess AGE-BSA or AGE-RNase to compete with immobilized AGE-BSA for the binding of N-ezrin, whereas no inhibition was observed with the non-glycated BSA or RNase (Fig. 5A). When CML-BSA or HSA isolated from the sera of diabetic patients were added, they also competed with the immobilized AGE-BSA for binding to N-ezrin (Fig. 5, A and B). C-ezrin did not bind to immobilized AGE-BSA but inhibited the ability of N-ezrin to bind (Fig. 5C). IGFBP-6, an unrelated protein of similar molecular weight to the truncated recombinant ezrin, had no effect on N-ezrin binding to AGE-BSA (data not shown).
ERM proteins can be phosphorylated in vitro, and two phosphorylation sites (Tyr-145 and Thr-567 in ezrin) are conserved within the three ERM proteins (25). As tyrosine-phosphorylated ezrin is essential for many of its functions, the effect of AGE-BSA on in vitro tyrosine phosphorylation of recombinant human N-ezrin and ezrin by epidermal growth factor receptor was measured. Phosphorylation of N-ezrin by EGFR was completely blocked by AGE-BSA, whereas full-length ezrin phosphorylation was inhibited by 57% (Fig. 6).
To assess whether AGEs affect the function of ezrin, studies were performed using a tubulogenesis model. The porcine kidney epithelial cell line LLC-PK1 forms cysts in monolayer. When exposed to fibroblast 3T3-conditioned medium containing hepatocyte growth factor (HGF) in the presence of collagen 1, tubules form after 24 h (Fig. 7) (22). This type of tubulogenesis is dependent on tyrosine phosphorylation of ezrin by HGF. The addition of glycated BSA, but not non-glycated BSA, significantly inhibited HGF-induced tubulogenesis (Fig. 7). DISCUSSION We have identified a novel interaction between AGE-modified proteins and the NH 2 -terminal regions of the ERM proteins (ezrin, radixin, and moesin), which belong to the erythrocyte protein 4.1 superfamily. The best-known function of ERM proteins within the cell is as a linker between the cytoplasmic tail of membrane proteins and cytoplasmic actin filaments (16). Additionally, ERM proteins have other functions, including regulation of Rho kinase and other signaling molecules such as focal adhesion kinase (FAK) and phosphatidylinositol 3-kinase (PI3K) and modulation of membrane ion transport proteins (NHE1 and 3) (26).
ERM proteins have three structural domains as follows: (i) an N-domain (aa 1-320), which is the most highly conserved (Ͼ85%) compared with the other members of the superfamily; (ii) an ␣-helical mid-region; and (iii) a C-domain, which has regions of positively charged amino acids and a consensus sequence for F-actin binding located in the last 35 amino acids (16). Binding studies with recombinant ezrin constructs re- vealed that the binding site for AGE-BSA was located in the N-domain. Many plasma membrane ligands (such as CD44 (27), CD43, and intercellular adhesion molecule (ICAM) 1, 2 and 3 (28)) bind via their cytoplasmic domains to the N-domain of ERM proteins. Recent studies show that phosphatidyl inositol 4-phosphate (PIP), phosphatidyl inositol 4,5-biphosphate (PIP 2 ), Rho GTPase, Rho GDP dissociation inhibitor (GDI) (29), syndecan 2 (30), and the sodium-hydrogen exchanger regulatory factor (NHERF), previously known as ezrin-binding protein 50 or EBP50 (31), also bind to the N-domains of ERM proteins. Besides the F-actin binding site in the C-domain, two additional actin binding sites have also been described in the N-domain, although the physiological relevance of these sites is not yet clear (32).
Like AGE-BSA, ligands such as CD44, NHERF, and the Rho GDP dissociation inhibitor bind to N-ezrin but do not bind to full-length ezrin (33). In full-length ERM proteins, the N-and C-domains associate by a high affinity intramolecular interaction that masks the binding sites for these ligands and actin (24). Low-angle shadowing electron microscopy recently reported for radixin revealed a compact globular structure that is changed on activation by phosphorylation to an elongated structure with two globular domains linked by a thin, filamentous region (34). In the experiments presented here, full-length ezrin did not bind to AGE-BSA by ligand blot, surface plasmon resonance, or affinity chromatography (data not shown), suggesting that the AGE-BSA binding site on N-ezrin may also be masked by interaction with C-ezrin. This hypothesis was confirmed by the ability of C-ezrin to block the binding of N-ezrin to AGE-BSA.
Because AGE-BSA only binds to recombinant NH 2 -terminal ezrin or truncated ERMs isolated from diabetic kidneys, ERM activation or proteolysis would have to occur in vivo for AGE binding. Although every care was taken to inhibit proteolytic degradation of proteins during the isolation (i.e. addition of protease inhibitors and purification at 4°C), it remains possible that truncated ERMs isolated from diabetic kidney may be products of the isolation procedure. However, both full-length ezrin and truncated ERMs were detected by ezrin immunoblotting in the diabetic kidney extract, whereas recombinant fulllength ezrin bound to the cation-exchange column but not to the AGE-BSA affinity column, with no evidence of ezrin proteolysis during the isolation steps (data not shown). There is some evidence that proteolysis and activation of ERMs occur in vivo (29,(35)(36)(37)(38), which would make the proteins accessible for binding to glycated proteins.
When bovine endothelial cell monolayers polarize and crawl in response to injury, ezrin proteolytic breakdown products accumulate. The ezrin degradation was found to follow a transient increase in intracellular calcium and could be blocked by a specific calpain I inhibitor (35). In situ experiments with intact gastric glands treated with calcium ionophore also show rapid hydrolysis of ezrin (36). Activation of platelets with thrombin or calcium ionophore caused rapid proteolysis of platelet ezrin and moesin (37). A primary event of cell activation by external stimuli such as high glucose or AGEs is a rise in intracellular calcium, and chronic hyperglycemia is also associated with a decrease in calcium efflux from cells leading to sustained elevation of basal calcium levels in diabetes (39,40). These observations, together with the results presented in this paper, suggest that the occurrence of high glucose and the glycated proteins present in patients with diabetes create an intracellular environment with high calcium levels that may lead to proteolyzed ERM proteins capable of interacting with intracellular glycated proteins and cause derangements in cellular function.
Activation of full-length ERM proteins is incompletely understood, but phosphorylation of specific residues, dimerization, or phospholipid (phosphatidyl inositol 4,5-biphosphate in particular) binding may be involved (29,38,41). ERM proteins can be phosphorylated in vitro, and two phosphorylation sites (Tyr-145 and Thr-567 in ezrin) are conserved within the three ERM proteins (25,42). The addition of epidermal growth factor to human epidermoid cancer cells (A-431) leads to tyrosine phosphorylation of ezrin and the recruitment of ezrin and actin to membrane projections (43). Inclusion of a tyrosine kinase inhibitor completely blocks both tyrosine phosphorylation of ezrin and its translocation to the membrane (44).
We investigated whether EGFR-induced ezrin tyrosine phosphorylation was changed by the addition of AGE-BSA and showed that AGE-BSA inhibited tyrosine phosphorylation of N-ezrin by EGFR. The fact that phosphorylation of full-length ezrin by EGFR was partially inhibited by AGE-BSA suggests that, although AGE-BSA does not bind to full-length ezrin, phosphorylation by EGFR may open up the protein to permit binding of AGE-BSA with subsequent inhibition of further phosphorylation.
Several studies have shown that ERM proteins and, in particular, ERM phosphorylation, play a crucial role in cell shape changes, cell attachment, and cell adhesion. Suppressing the expression of all three ERM proteins inhibited cell-cell and cell-substrate adhesion, microvilli disappeared, and collapsed actin fibers were observed (45). In studies where ezrin phosphorylation was reduced (similar to our results with AGE-BSA), normal ezrin-mediated cross-linking between the plasma membrane and the actin cytoskeleton was broken. Instead, ezrin remained in the cytoplasm despite ligand induction, and normal cellular responses such as cell shape change and adhesion/attachment were lost or reduced (44,46). Furthermore, two tyrosines of ezrin, which are phosphorylated in response to HGF or EGFR, are essential for cell motility and tubulogenesis (22). Transfection of LLC-PK1 cells with ezrin mutated at Y145F and Y353F inhibited cell migration and tubulogenesis. Rearrangements of the actin cytoskeleton underlie cell motility and the formation of long tubules as shown by blockage with cytochalasin B (22). Because AGE-BSA inhibits both tubulogenesis, for which phosphorylated ezrin is an effector, and in vitro phosphorylation of N-ezrin, we hypothesize that glycated proteins may modulate ERM function by inhibition of ERM phosphorylation.
The significance of ERM protein binding to AGEs in the development of diabetic complications remains to be established. AGE-BSA and AGE-RNase produced in vitro are highly modified molecules (ϳ77% modified), and, thus, may not reflect endogenous AGEs. The degree of protein modification is extremely variable in vivo (determined by ambient hyperglycemia, oxidative stress, protein identity, and turnover), and rates of glycosylation with intracellular sugars such as fructose, glucose-6-phosphate, and glyceraldehyde 3-phosphate are much greater than those achieved extracellularly with glucose (47). Indeed, endogenously glycated HSA isolated from human diabetic sera as well as CML-BSA competed with the immobilized AGE-BSA for binding to N-ezrin. N ⑀ -carboxymethyllysine is a modification that occurs in vivo (1).
ERM proteins are present at all sites of diabetic complications, and we expect that AGE binding might inhibit ERM functions involving interactions with the N-domain of ERM proteins. These functions include cross-linking between plasma membrane proteins and the actin cytoskeleton and disruption of signal transduction pathways involving phosphorylated tyrosine residues of ERM proteins and other signaling molecules. Thus, the envisioned result of high levels of glycated proteins on ERM responses would be actin disorganization and disconnection from ligand/receptor stimulation and, consequently, abnormal cell-cell adhesion, cell-extracellular matrix attachment, and irregular cell shape. Indeed, there have been several reports of reorganization or disassembly of actin filaments in response to high glucose or AGEs (48,49), and actin rearrangement has also been shown in vivo in experimental diabetes (50).
Actin cytoskeleton disassembly is a prominent feature of diabetic nephropathy (48) and is particularly evident and important in the podocyte or glomerular epithelial cells. These cells differentiate into a special structure called the glomerular filtration apparatus, which includes the foot processes and slit diaphragms with highly organized actin bundles. These glomerular epithelial cells create a pathway for the glomerular filtrate (51). Ezrin is concentrated along the apical membrane of the foot processes in association with podocalyxin, NHERF2, and actin (52,53), and this complex is essential for proper foot process organization. Disruption of this complex leads to the loss of foot process architecture, as seen in diabetes, with effacement of the slit pores and subsequent increases in glomerular permeability and proteinuria (51,53,54).
The cofactor role of ERMs in the regulation of the sodium/ hydrogen exchanger 3 (NHE-3) is also relevant to diabetic complications. We have reported previously that inhibitors of AGE formation as well as cariporide (an inhibitor of a sodium/ hydrogen exchanger) can attenuate the development of diabetic complications in a rat model (55). It remains to be determined if this relates to a central role of ERM proteins that not only bind AGEs but also NHERF 1 and 2 (56) and, thus, modulate sodium/hydrogen exchangers.
ERM proteins are novel intracellular binding proteins for glycated proteins. This interaction may modulate a range of processes involved in the development of diabetic complications. Understanding this interaction may thus provide a new avenue for therapeutic intervention and the development of organ-protective agents in diabetes.