Glycation Induces Formation of Amyloid Cross-β Structure in Albumin

Amyloid fibrils are components of proteinaceous plaques that are associated with conformational diseases such as Alzheimer's disease, transmissible spongiform encephalopathies, and familial amyloidosis. Amyloid polypeptides share a specific quarternary structure element known as cross-β structure. Commonly, fibrillar aggregates are modified by advanced glycation end products (AGE). In addition, AGE formation itself induces protein aggregation. Both amyloid proteins and protein-AGE adducts bind multiligand receptors, such as receptor for AGE, CD36, and scavenger receptors A and B type I, and the serine protease tissue-type plasminogen activator (tPA). Based on these observations, we hypothesized that glycation induces refolding of globular proteins, accompanied by formation of cross-β structure. Using transmission electron microscopy, we demonstrate here that glycated albumin condensates into fibrous or amorphous aggregates. These aggregates bind to amyloid-specific dyes Congo red and thioflavin T and to tPA. In contrast to globular albumin, glycated albumin contains amino acid residues in β-sheet conformation, as measured with circular dichroism spectropolarimetry. Moreover, it displays cross-β structure, as determined with x-ray fiber diffraction. We conclude that glycation induces refolding of initially globular albumin into amyloid fibrils comprising cross-β structure. This would explain how glycated ligands and amyloid ligands can bind to the same multiligand “cross-β structure” receptors and to tPA.

Protein aggregation in organs and in the circulation is a common aspect of conformational diseases such as Alzheimer's disease, transmissible spongiform encephalopathies, pancreatic islet amyloidosis, and familial amyloidosis (1,2). The insoluble aggregates can form amyloid fibrils with a cross-␤ structure, the amyloid fiber-specific quarternary structure element. Elevated levels of protein aggregates induce neurodegeneration in patients with Alzheimer's disease and transmissible spongiform encephalopathies. In patients with type II diabetes mellitus, pancreatic islet amyloidosis results in ␤-cell damage and impaired insulin secretion. Organ damage results from amyloidosis in patients carrying mutated proteins, e.g. transthyretin amyloidosis (3) and immunoglobulin light chain amyloidosis (4).
Amyloid plaques modified by AGE have been identified in brain tissue of Alzheimer's disease patients (22) and transmissible spongiform encephalopathy patients (23) and in the islets of Langerhans of diabetic patients (24). Furthermore, amyloid protein deposits as well as AGE are identified in plaques associated with dialysis-related amyloidosis, atherosclerosis, and vascular occlusions in patients suffering from diabetes (12,(25)(26)(27). Accelerated fibril formation occurs upon glycation of monomers of ␤-amyloid (A␤) and islet amyloid polypeptide (IAPP), two prototype peptides associated with conformational diseases (24,28). These data indicate that glycation of polypeptides, which have the propensity to condensate into amyloid fibrils, can accelerate cross-␤ structure formation.
Diverse cell types, such as endothelial cells, mesangial cells, and macrophages, express AGE receptors (6,29,30). These receptors are involved in clearance of obsolete proteins (30 -32). In addition, AGE receptors can mediate toxic responses to glycated proteins (33)(34)(35)(36). Interestingly, scavenger receptor class A, scavenger receptor class B type I, CD36, and receptor for AGE are multiligand receptors that bind to AGE as well as to amyloid fibrils (35)(36)(37)(38)(39)(40). Recently, we identified tPA, one of the key regulators in blood fibrinolysis (41), as a new member of this class of multiligand amyloid-binding proteins (42). We showed that cross-␤ structure in denatured proteins and in fibrin fragments is a prerequisite for tPA binding. In a previous study we also showed that after glycation various proteins gain tPA binding capacity (43). Therefore, we tested the hypothesis that glycation results in cross-␤ structure formation.
Cell Cultures-Murine bEnd.3 mouse brain endothelial cells were cultured in Invitrogen Dulbecco's modified Eagle's medium (Invitrogen) enriched with 5% fetal calf serum. For experiments, the cells were seeded in 24-well culture dishes (Corning Inc., Corning, NY) and allowed to attach for 16 h. Before the experiments, the medium was exchanged for human endothelial serum-free medium basal growth medium (Invitrogen). Cell confluency was 50% at the start of the experiments. bEnd.3 cells were incubated in triplicate with 25 M albumin-control:23, 25 M albumin-AGE:23, 12.5 M of positive control IAPP, or water. IAPP is routinely used in our laboratory for its ability to induce excessive loss of cell viability. Amyloid IAPP was obtained after a 3-week room temperature incubation of lyophilized peptide (Pepscan, Lelystad, The Netherlands) dissolved in PBS. After 48 h, the cells were photographed and subsequently incubated for 15 min at 37°C with 10 g ml Ϫ1 Calcein (Molecular Probes, Leiden, The Netherlands). The cells were subsequently washed with PBS and lysed with 100 l of lysis buffer (1% (v/v) Nonidet P-40, 15 mM NaCl, 10 mM Tris-HCl, pH 7.5, 5 mM EDTA). For each sample, 50 l was transferred to a 96-well plate, and Calcein fluorescence was measured. Excitation and emission wavelengths were 485 and 530 nm, respectively.
Transmission Electron Microscopy-For TEM analyses of albumin-g6p:2, albumin-g6p:4, albumin-g6p:23, and the corresponding controls, protein samples were applied to 400-mesh specimen grids covered with carbon-coated collodion films. After 5 min the buffer was removed with filter paper, and the preparations were stained with 1% (m/v) methylcellulose and 1% (m/v) uranyl acetate in water. After washing with water, the samples were dehydrated in a graded series of EtOH and hexanethyldisilazane. Transmission electron micrographs were recorded at 60,000ϫ magnification at 80 kV on a JEM-1200EX electron microscope (JEOL, Tokyo, Japan).
Samples of albumin-glyceraldehyde, -glucose, -fructose, -g6p:86, and -glyoxylic acid and the corresponding controls were prepared slightly differently. The samples were applied to 100-mesh copper grids with carbon-coated Formvar (Merck) and subsequently washed with PBS and H 2 O. The grids were applied to droplets of 2% (m/v) methylcellulose with 0.4% (m/v) uranylacetate pH 4. After a 2-min incubation, the grids were dried on a filter. The micrographs were recorded at 80 kV at 6,000ϫ and 50,000ϫ magnification.
Binding of Congo Red-Binding of Congo red to cross-␤ structure can be detected by measuring absorbance at 530 nm of amyloid structures in solution (45). For this purpose, 2.5 M albumin-AGE and albumincontrol was incubated in multiples of five with 100 M Congo red (Aldrich) in PBS with 10% (v/v) ethanol. Absorbance at 530 nm was recorded for the Congo red-incubated samples, as well as for Congo red background and protein background. Alternatively, binding of Congo red was tested using air-dried samples. For this purpose, 5 l of the albumin solutions after 86 weeks of glycation was applied to a glass coverslip and air-dried. The samples were incubated with Congo red and subsequently washed according to the manufacturer's recommendations (Sigma). The Samples were analyzed on a Leica (Rijswijk, The Netherlands) DMIRBE fluorescence microscope using 596 and 620 nm excitation and emission wavelengths, respectively.
Circular Dichroism Spectropolarimetry-CD spectra of albumin-AGE and of controls were measured on a Jasco (Tokyo, Japan) J-810 CD spectropolarimeter. The CD spectrum of the amyloid prototype A␤ (16 -22), which has all amino acid residues in ␤-sheet conformation (47), was measured as a positive control. Protein and peptide concentrations were 100 g ml Ϫ1 in water. From each sample, ten spectra were measured from 240 to 200 nm, and the relative percentages of the secondary structure elements present were estimated using k2d software (48).
Enzyme-linked Immunosorbent Assay-Binding of the cross-␤ structure-marker tPA to albumin-AGE was tested using an ELISA set-up. Albumin glycated for 2, 4, or 23 weeks and corresponding unglycated controls (25 g ml Ϫ1 in coat buffer; 50 mM Na 2 CO 3 /NaHCO 3 pH 9.6, 0.02% m/v NaN 3 , 50 l/well) were covalently immobilized for 1 h at room temperature in 96-well protein Immobilizer plates (Exiqon). Control wells were incubated with coat buffer only. After a wash step with 200 l of PBS, 0.1% (v/v) Tween 20, the plates were blocked with 300 l of PBS, 1% (v/v) Tween 20 for 2 h at room temperature while shaking. The samples glycated for 86 weeks and the controls were coated to Greiner (Frickenhausen, Germany) microlon plates (catalogue number 655092). The wells were blocked with Superblock (Pierce). All of the subsequent incubations were performed in PBS, 0.1% (v/v) Tween 20 for 1 h at room temperature while shaking, with volumes of 50 l/well. After incubation, the wells were washed five times with 300 l of PBS, 0.1% (v/v) Tween 20. Increasing concentrations of tPA were added in triplicate to coated wells as well as to control wells. During tPA incubations of samples incubated for 86 weeks, at least a 123,000-fold molar excess of ⑀-amino caproic acid (10 mM) was added to the solutions. ⑀-Amino caproic acid is a lysine analogue and is used to avoid potential binding of tPA to albumin via its kringle2 domain (49). Polyclonal antibody 385R or monoclonal antibody 374b (American Diagnostica Instrumentation Laboratory, Breda, The Netherlands) and, subse-2 P. G. de Groot and I. Bobbink, unpublished data. quently, SWARPO or RAMPO (Dako Diagnostics, Glostrup, Denmark) were added to the wells at a concentration of 0.3 g ml Ϫ1 . Bound peroxidase-labeled antibody was visualized using 100 l of a solution containing 8 mg of ortho-phenylene-diamine in 20 ml of 50 mM citric acid, 100 mM Na 2 HPO 4 , pH 5.0, with 0.0175% (v/v) H 2 O 2 . Staining was stopped upon adding 50 l of a 2 M H 2 SO 4 solution. Absorbance was read at 490 nm on a V max kinetic microplate reader (Molecular Devices). Background signals from uncoated control wells were substracted from corresponding coated wells.
X-ray Fiber Diffraction-For x-ray fiber diffraction analyses, protein solutions that were extensively dialyzed against water were air-dried at room temperature in Boro glass capillaries. Diffraction measurements were performed on a Nonius CCD diffractometer equipped with a CCD area detector and a graphite monochromator (Bruker-Nonius, Delft, The Netherlands). Sealed tube molybdenum K␣ radiation was used, and capillaries were exposed to the x-ray beam for 16 h. Scattering from air and capillary glass were subtracted from diffraction images using in-house software (VIEW/EVAL, Department of Crystal and Structural Chemistry, Utrecht, The Netherlands).

Glycation of Albumin Induces Formation of Toxic
Aggregates-Albumin is present in the circulation at relatively high concentration (35-50 g liter Ϫ1 ) and is prone to glycation in vivo (50). Glycated albumin is the commonly used prototype protein-AGE adduct in multiligand receptor binding studies in vitro (34,51) and in protein-AGE cell toxicity assays (14,17). Protein-AGE adducts that also occur in vivo can be prepared in vitro by incubating albumin with g6p (44,52). Therefore, we glycated amino groups of bovine serum albumin with g6p for 2, 4, and 23 weeks (albumin-AGE:2, albumin-AGE:4, and albumin-AGE:23, respectively) in a similar way as described previously (44). All of the solutions were clear, and no precipitates were observed. Intense brown staining and AGE-specific autofluorescence was observed for the albumin-AGE preparations (Fig. 1A). Glycation of albumin resulted in formation of dimers and multimers (Fig. 1B). These dimers and multimers likely arise from intermolecular cross-links, such as dialkyldihydro pyrazine lysine-lysine cross-links (53). Aggregation, brown staining, and autofluorescence are characteristic for protein-AGE adducts, indicating that albumin is modified upon g6p incubation. Binding of AGE-specific antibodies confirmed glycation and subsequent AGE formation of our albumin preparations ( Fig. 1, C and D). Albumin-AGE affects cell viability (14,17). We tested the effect of our albumin-AGE preparations on bEnd.3 endothelial cells. Incubation of these cells with albumin-AGE:23 resulted in massive cell death (Fig. 2). Control cells grew normally, whereas prototype amyloid peptide IAPP reduces cell viability (50%). These results demonstrate that albumin-AGE, in contrast to controls, is cytotoxic to the endothelial cells tested, similar to what was observed in previous reports (13,14,16,17). We have used these albumin-AGE preparations as prototype protein-AGE adducts for further structural analyses.
Albumin-AGE Condensates as Fibrils or as Amorphous Aggregates-To determine the structural characteristics of the glycated albumin, we applied an array of biophysical methods. TEM analyses revealed that after a 2-week glycation period, bundles of unbranched fibrillar aggregates of albumin-AGE:2 are formed, with lengths ranging from 100 to 150 nm (Fig. 3A). Incubation of albumin with g6p for 4 weeks revealed amorphous aggregates (Fig. 3B), whereas after 23 weeks fibrous sheet-like structures were observed, with a length of 100 -300 nm and a diameter of 5-10 nm (Fig. 3C). Aggregates were absent in albumin controls (not shown). The presence of sheetlike fibers and amorphous aggregates in albumin-AGE is accompanied by an increase in ␤-sheet content from 0 to 7% for albumin-AGE:2 and albumin-AGE:4, and up to 18% for albumin-AGE:23, as measured with CD spectropolarimetry. ␤-Sheets were not detected in solutions of albumin control. Noteworthy, the globular fold of human albumin, which shares over 80% amino acid sequence homology with bovine albumin, comprises 69% ␣-helix with 0% ␤-sheet (54).
Next, we tested the binding of specific amyloid markers to albumin-AGE. Albumin-AGE enhanced fluorescent emission of cross-␤ structure-specific dye ThT in solution (Fig. 3D). In addition, when absorbance is recorded for solutions of Congo red-albumin-AGE, an amyloid-specific increase in intensity at 530 nm is observed (Fig. 3E). Multiligand cross-␤ structurebinding protein tPA also bound specifically to all three albumin-AGE preparations (Fig. 3, F and G). The affinity of Congo red, ThT, and tPA for albumin-AGE:23 and the number of binding sites on albumin-AGE:23 for these substrates are higher than those observed for albumin-AGE:2 (Fig. 3, D-G). In contrast, the amorphous aggregates observed for albumin-AGE:4 expose a relatively higher number of ThT-and tPAbinding sites when compared with albumin-AGE:23 (Fig. 3, D  and F). Possibly, more ThT-and tPA-binding sites are exposed on the relatively small and less condensed aggregates, when compared with the albumin-AGE:2 and albumin-AGE:23 fibrils (Fig. 3, A-C).
The gel electrophoresis, TEM, and CD spectropolarimetry data obtained with albumin-AGE preparations indicate that modification of amino groups and cross-linking upon glycation induces conformational rearrangements. Initially globular albumin refolds into amorphous or fibrillar aggregates compris-ing ␤-sheet secondary structure. Moreover, Congo red binding, ThT fluorescence, and tPA binding are indicative for the presence of cross-␤ structure in glycated albumin preparations. The presence of this cross-␤ structure is independent of the macroscopic appearance of albumin as amorphous aggregates or as fibrillar structures.
Albumin-AGE Displays X-ray Fiber Diffraction Specific for Cross-␤ Structure-Binding of cross-␤ structure markers ThT, Congo red, and tPA to albumin-AGE prompted us to analyze the albumin-AGE structure in greater detail by use of x-ray fiber diffraction analyses. Diffraction experiments revealed that albumin-AGE:23 aggregates are built up by a significant amount of crystalline fibers (Fig. 4, A and C). Diffraction patterns of albumin-control:23 are typical for an air-dried globular protein (Fig. 4B). X-ray fiber diffraction experiments with albumin-AGE:2 and albumin-AGE:4 did not reveal diffraction patterns originating from fibrous structures (not shown). Apparently, the fibers and aggregates that we observed with TEM did not (yet) arrange themselves into well ordered crystallites, which is a prerequisite for the diffraction experiment. However, this observation does not exclude the presence of the cross-␤ structure as it was indicated by the Congo red, ThT, and tPA binding studies and by the presence of the ␤-sheet secondary structure.
For albumin-AGE:23, the 4.7 Å repeat corresponds to the characteristic hydrogen bond distance between individual ␤-strands in ␤-sheets. The 2.3 and 3.3 Å repeats have a preferred orientation perpendicular to the 4.7 Å repeat (Fig. 4D). Combining the 2.3 and 3.3 Å repeats may point to a fiber axis oriented perpendicular to the direction of the hydrogen bonds, with a repeat of ϳ6.8 Å. This dimension corresponds to the length of two peptide bonds and indicates that ␤-strands run parallel to the fiber axis. A similar orientation is found in amyloid fibrils of a peptide from PrP c (55) and in fibers of natural silk (56,57). Despite the fact that in our experiments no magnetic field was used to orient the samples, we still acquire preferred orientation because of the capillary. Our experience with many fibrous samples is that they can form either a film or a pellet in the capillary. In the first case the fiber axis is along the long axis of the capillary, whereas in the latter case the fiber axis is along the surface of the pellet, i.e. perpendicular to the capillary axis. This latter situation is exactly what we found in Fig. 4A where the capillary was placed in a vertical position, and the fiber axis is perpendicular to that. If the a axis is 9.4 Å and the c axis is 6.8 Å, the 2.5 and 6.0 Å repeats can only be indexed as (hkl). This implies an ordered b axis repeat, corresponding to the inter ␤-sheet distance. The 3.8 Å repeat could then be indexed as (201). These observations are in agreement with a unit cell with dimensions a ϭ 9.4, b ϭ 15.7, c ϭ 6.8 Å, ␣ ϭ 90°, ␤ ϭ 90°, ␥ ϭ 90°. These dimensions resemble those of the H1 peptide of PrP c (55). In addition, both the strong 4.7 and 3.8 Å reflections were found in various amyloid fibrils (55,58). Considering all of these obser-vations, it is clear that the albumin-AGE:23 fibers are, at least in part, composed of cross-␤ structure consisting of packed ␤-sheets (Fig. 4E), a characteristic feature of amyloid fibrils.
The presence of highly ordered cross-␤ structure, arranged in crystallites, may explain the relatively high affinity of tPA for albumin-AGE:23, when compared with albumin-AGE:2 and albumin-AGE:4 (Fig. 3, F and G). In crystallites, amino acid stretches in cross-␤ structure, that likely built up the tPA-

FIG. 4. X-ray fiber-diffraction analysis with albumin-AGE:23 fibrils reveals that glycation induces formation of cross-␤-structure.
A, x-ray scattering of albumin-AGE:23. Scattering intensities are color coded on a logarithmic scale and decreases in the order white-gray-black. Scattering from amorphous control albumin is subtracted, as well as scattering from the capillary glass wall and from air. D-spacings and the direction of the fiber axis are given, and the preferred orientations are indicated with arrows. D-spacings are in Å units. B, radial scans of albumin-control:23 and albumin-AGE:23. C, radial scan of albumin-AGE:23 after subtracting background scattering of amorphous precipitated albumin-control:23. Repeats originating from fibrous structure are indicated with their D-spacings (Å). D, tangential scans along the 2 scattering angles, corresponding to the indicated D-spacings. The scans show that the 4.7 Å repeat, which corresponds to the hydrogen bond distance within individual ␤-sheets, and the 6 Å repeat are oriented perpendicular to the 2.3 Å repeat that runs parallel to the fiber axis. E, schematic drawing of the orientation of the cross-␤ structures in albumin-AGE:23 amyloid fibrils. The direction of the peptide bonds runs parallel with the fiber axis. The 4.7 Å inter-␤-strand hydrogen bond distance runs perpendicular to the fiber axis. binding site, may be ordered and fixed in such a way that tPA can easily bind. In contrast, cross-␤ structure in amorphous, although macroscopically fibrous albumin-AGE:2, and in amorphous albumin-AGE:4 is likely less ordered, thereby apparently exposing lower affinity tPA binding sites.
Amyloid Albumin Is Formed Irrespective of the Glycating Agent-From the above listed observations it is clear that modification of amino groups of albumin with g6p induces formation of the amyloid cross-␤ structure. The next question we addressed was whether triggering of refolding of globular albumin into an amyloid fold was a restricted property of g6p or whether amyloid formation occurs irrespective of the glycating agent. Albumin solutions were incubated for 86 weeks at 37°C with g6p, DL-glyceraldehyde/NaCNBH 3 , ␤-D(Ϫ)-fructose, D(ϩ)glucose, glyoxylic acid/NaCNBH 3 , or without glycating agent (PBS or PBS/NaCNBH 3 ). After 86 weeks, suspensions of albumin-glyceraldehyde and albumin-fructose and clear solutions of albumin-glucose and albumin-glyoxylic acid were light yellow/brown, albumin-g6p:86 was a clear and dark brown solution, whereas clear solutions of albumin incubated without glycating agent were colorless. AGE formation was confirmed by autofluorescence measurements using AGE-specific excitation/emission wavelengths (Fig. 5A), binding of monoclonal anti-AGE antibody 4B5 (Figs. 1C and 5B), and binding of polyclonal anti-AGE antibody (Figs. 1D and 5C). As expected, albumin-glyoxylic acid did not show an autofluorescent signal because of the fact that nonfluorescent carboxymethyl-lysine is formed. Glycation of albumin with various glycating agents is accompanied with differences in macroscopic appearance, as observed with TEM (Fig. 5, D-R). Incubation of albumin with glucose resulted in irregularly shaped sheets with varying sizes, together with free and clustered granular precipitates (Fig. 5, D-F). Albumin-glyceraldehyde appeared as clustered granules and sporadic occurring linear rods, 250 -400 nm in length (Fig. 5, G-I). With fructose, large and condensed amorphous aggregates are observed (Fig. 5, J-L). In addition, curved and linear fibrillar structures are present (Fig. 5, J and K). Glycation with glyoxylic acid resulted mainly in large branched chains of globular aggregates (Fig. 5, M and N). The albumin PBS control appeared as a rather uniformly distributed fine amorphous precipitate (Fig. 5O). The albumin-NaCNBH 3 control solution contained long and unbranched fibrillar structures that are occasionally slightly curved (Fig. 5, P-R). Surprisingly, on a polyacrylamide gel, the fine precipitates observed with the PBS control and the fibrils obtained with the PBS/NaCNBH 3 control co-migrated with freshly dissolved albumin, at a size corresponding to a molecular mass of 60 kDa (not shown). Similar to what was observed for albumin-AGE:23 (Fig. 1B), all five incubations with glycating agents resulted in albumin multimers that hardly entered the gel (not shown). We conclude that albumin multimers formed during glycation are kept together by AGE cross-links, whereas precipitates observed with the controls are easily monomerized upon boiling in SDS. The observation that albumin-glyoxylic acid/NaCNBH 3 appears as multimers on a polyacrylamide gel indicates that in addition to carboxymethyl-lysine, at least some other nonfluorescent AGE structures are formed that are involved in albumin cross-linking.
The autofluorescence data and the binding of AGE-specific antibodies listed above show that various carbohydrates and carbohydrate derivatives can lead to similar AGE structures. Using g6p as starting point for AGE formation, we showed that albumin adopted amyloid properties similar to those observed in well studied fibrils of A␤ and IAPP. Therefore, we tested for the presence of amyloid structures in the albumin-AGE adducts obtained with alternative carbohydrates and derivatives.
We measured fluorescence of albumin-AGE preparations that were incubated with Congo red (Fig. 6, A-I) and of albumin-AGE-ThT solutions (Fig. 6J). Albumin-g6p, albumin-fructose, and albumin-glyceraldehyde/NaCNBH 3 gave a Congo red fluorescent signal similar to signals of the amyloid core peptide of IAPP and A␤ (Fig. 6, A-E). With albumin-glucose, a uniformly distributed pattern of fluorescent precipitates is observed (Fig.  6F). With albumin-glyoxylic acid/NaCNBH 3 and buffer controls, hardly any fluorescence is observed (Fig. 6, G-I). Incubation of albumin with glyceraldehyde/NaCNBH 3 , glucose, or fructose resulted in an increased fluorescent signal of ThT (Fig.  6J). No specific amyloid-ThT fluorescence was measured for albumin-glyoxylic acid and buffer controls. These ThT and Congo red fluorescence data show that in addition to albumin-g6p, albumin-glyceraldehyde/NaCNBH 3 , albumin-glucose, and albumin-fructose have amyloid-like properties. To further substantiate these findings, we tested for binding of amyloid-specific serine protease tPA in an ELISA. Recently, we showed that tPA binds to prototype amyloid peptides human A␤  and human IAPP (42). Therefore, we used tPA binding to these two peptides as a positive control. The enzyme bound specifically to albumin-g6p, albumin-glyceraldehyde/NaCNBH 3 , albumin-glucose, and albumin-fructose (Fig. 6, K and L). No tPA binding is observed for albumin-glyoxylic acid/NaCNBH 3 and buffer controls. These tPA binding properties correlate with observed Congo red and ThT fluorescence. The cross-␤ structure detected by ThT fluorescence, Congo red stain, and tPA binding did not result in an x-ray diffraction pattern specific for amyloid fibers. Apparently, aggregates comprising cross-␤ structure did not arrange themselves in crystalline fibers large enough to display x-ray diffraction as with albumin-g6p:23.
From the ThT fluorescence, Congo red stain and tPA binding data, it is clear that inducing amyloid properties in albumin is not an exclusive property of g6p. Apparently, a spectrum of glycating agents, comprising g6p, glucose, fructose, glyceraldehyde/NaCNBH 3 , and likely more has the capacity to trigger the switch from a globular native fold to the amyloid cross-␤ structure fold upon their covalent binding to albumin. With the experimental conditions applied, glyoxylic acid seems to be the first exception. Albumin-AGE aggregates are formed upon incubation with glyoxylic acid/NaCNBH 3 , as determined with anti-AGE antibodies, TEM analysis, and SDS-PAGE. These mainly carboxymethyl-lysine modifications are, however, not accompanied with typical amyloid characteristics, as is observed with the other albumin-AGE adducts. Cross-linking following AGE formation may be the driving force for amyloid formation seen with g6p, glucose, fructose, and glyceraldehyde/NaCNBH 3 .

DISCUSSION
The biophysical and biochemical data for albumin-AGE presented in this study show that glycation of albumin with various glycating agents can induce exposure of neo-epitopes for the amyloid markers ThT, Congo red, and tPA. The TEM data and the x-ray crystallography data obtained with g6p-glycated albumin together show that amyloid fibrils can be formed that contain the cross-␤ structure. These structural characteristics are similar to those reported for prototype amyloid fibrils of disease-related peptides and proteins A␤, IAPP, PrP, and transtherytin, which can form spontaneously (3,42,55,59,60). Moreover, our data are in agreement with previous observations that glycation increases the rate of cross-␤ structure formation of A␤  and IAPP (24,28). Therefore, we conclude that glycation has the potential to induce refolding of globular proteins into cross-␤ structure. Although the covalent binding of carbohydrates is most likely the main reason for the observed effects on cross-␤ structure formation, we cannot com- FIG. 5. Autofluorescence and anti-AGE antibody binding of albumin-AGE:86 and TEM analyses. A, autofluorescence of 86-weeks glycated albumin indicating that fluorescent carbohydrate-albumin adducts are formed after incubation with glyceraldehyde/NaCNBH 3 , glucose, and fructose. The excitation and emission wavelengths were 380 and 445 nm, respectively. Background fluorescence of buffer and nonglycated control albumin are subtracted. Averaged values over three independent measurements are shown with error bars. B and C, AGE-specific monoclonal anti-albumin-g6p antibody 4B5 (44) (B) and polyclonal anti-fibronectin-g6p antibody 2 (C) bind to glyceraldehyde/NaCNBH 3 -, glucose-, fructose-, and glyoxylic acid/NaCNBH 3 -modified albumin but not to albumin controls. D-R, TEM analyses of albumin glycated for 86 weeks and corresponding buffer controls. The scale bars are shown. Incubation of albumin with glucose resulted in irregularly shaped sheets with varying size, together with free and clustered granular precipitates (D-F). Albumin-glyceraldehyde/NaCNBH 3 appeared as clustered granules and sporadic occurring linear rods, 250 -400 nm in length (G-I). With fructose, large and condensed amorphous aggregates are observed (J-L). In addition, curved and linear fibrillar structures are present (J and K). Glycation with glyoxylic acid/NaCNBH 3 resulted mainly in large branched chains of globular aggregates (M and N). The albumin PBS control appears as a rather uniformly distributed fine amorphous precipitate (O). The albumin-NaCNBH 3 control solution contains long and unbranched fibrillar structures that are occasionally slightly curved (P-R).
pletely exclude a catalytic effect for the small drop in pH that was observed during the glycation reaction.
A large number of polypeptides can be forced to adopt nonnative conformations upon fragmentation into separate domains or even small peptides and by exposure to denaturants or high temperature (8, 42, 47, 55, 60 -64). Structural details obtained after these treatments have led to the idea that every polypeptide has the intrinsic propensity to adopt the cross-␤ structure tertiary/quarternary fold (42,(65)(66)(67)(68)(69)(70). We now show that glycation can induce refolding of albumin into cross-␤ structure conformation. Therefore, our observations add new evidence to the aforementioned intriguing hypothesis that refolding into cross-␤ structure is a general phenomenon rather than an exclusive property of the group of about 30 known proteins involved in conformational diseases (71). The fact that virtually every polypeptide can be prone to glycation because of the presence of amino groups in lysine and arginine residues and at the N terminus suggests that glycation of other proteins than albumin also will induce cross-␤ structure formation.
In this study, we demonstrate that glycation induces cross-␤ structure formation. However, the mechanism underlying gly-cation-induced amyloid formation remains unknown. Modification of amino groups at the solvent exposed site of polypeptides may stimulate refolding from a globular state to a fibrillar state in two ways: (i) covalent binding of carbohydrates to lysine and arginine residues alters their microenvironment in such a way that the polypeptide (partly) unfolds and (ii) intramolecular or intermolecular AGE-bridged cross-links exert mechanical stress on the polypeptide, thereby inducing local or global unfolding. Sequences of the polypeptide that were hidden in the globular state may now become solvent-exposed. This may facilitate new contacts between amino acid stretches. Unfolding of native structure elements and introduction of new contacts may stimulate formation of cross-␤ structure, the first step toward amyloid fibril formation. In this way, AGE act as denaturants and as zippers that bring together sequences that have the propensity to fold into cross-␤ structure.
Cellular responses to protein-AGE adducts are mediated by AGE receptors (33)(34)(35)(36). Interestingly, these multiligand receptors are also receptors for amyloid fibrils (35)(36)(37)(38)(39)(40). Protein-AGE adducts and amyloid fibrils lack amino acid sequence homology. Our data now show that amyloid fibrils and glycated FIG. 6. Albumin-AGE has amyloid properties irrespective of the glycating agent. A-I, Congo red fluorescence of air-dried albumin preparations. Gray, red fluorescence; black, background. Fluorescence was measured with albumin incubated with g6p (A), fructose (B), glyceraldehyde/NaCNBH 3 (C), the amyloid core peptide of human IAPP (D), A␤ (E), albumin incubated with glucose (F), glyoxylic acid/NaCNBH 3 (G), PBS (H), or PBS/NaCNBH 3 (I). J, thioflavin T-amyloid fluorescence was measured in solution with the indicated albumin preparations. K and L, binding of amyloid-binding serine protease tPA to albumin preparations was assayed using an ELISA set-up. tPA binds to albumin-glucose, albumin-fructose, and albumin-glyceraldehyde/NaCNBH 3 but not to albumin-glyoxylic acid/NaCNBH 3 and albumin-buffer controls (K). tPA also binds to positive controls albumin-g6p:86, A␤, and IAPP (L). polypeptides can share the cross-␤ structure. Thus, the presence of cross-␤ structure as the common denominator between compounds capable of interacting with multiligand receptors is suggestive. This would imply that AGE receptors, i.e. scavenger receptor class A, scavenger receptor class B type I, CD36, and receptor for AGE (30 -32), are in fact cross-␤ structure receptors. Recently, we found that also multiligand bindingprotein tPA binds to protein-AGE adducts (43), as well as to polypeptides built up by cross-␤ structure (42). If the cross-␤ structure is a unique recognition signal, it should have unique structural features that are absent in globular polypeptides composed of ␣-helix, ␤-sheet, and random coil. In addition, multiligand receptors and cross-␤ structure-binding proteins such as tPA may also share a common structural or sequential motif necessary for interaction with cross-␤ structure.
Binding of misfolded cross-␤ structure comprising polypeptides to tPA and multiligand receptors on cells may lead to proteolysis and the internalization and clearance of these polypeptides. We propose to term this clearance mechanism the cross-␤ structure pathway.