Cleaved β2-Microglobulin Partially Attains a Conformation That Has Amyloidogenic Features*

β2-Microglobulin, a small protein localized in serum and on cell surfaces, can adopt specific aggregating conformations that generate amyloid in tissues and joints as a complication to long-term hemodialysis. We characterize a proteolytic variant of β2-microglobulin (cleaved after Lys58) that as a trimmed form (Lys58 is removed) can be demonstrated in the circulation in patients with chronic disease. An unexpected electrophoretic heterogeneity of these two cleaved variants was demonstrated by capillary electrophoresis under physiological conditions. Each separated into a fast and a slow component while appearing homogeneous, except for a fraction of oxidized species detected by other techniques. The two components had different binding affinities for heparin and for the amyloid-specific dye Congo red, and the equilibrium between the two forms was dependent on solvent conditions. Together with analysis of the differences in circular dichroism, the results suggest that β2-microglobulin cleaved after Lys58 readily adopts two equilibrium conformations under native conditions. In the cleaved and trimmed β2-microglobulin that appearsin vivo, the less populated conformation is characterized by an increased affinity for Congo red. These observations may help elucidate why β2-microglobulin polymerizes as amyloid in chronic hemodialysis and facilitate the search for means to inhibit this process.

␤ 2 -Microglobulin, a small protein localized in serum and on cell surfaces, can adopt specific aggregating conformations that generate amyloid in tissues and joints as a complication to long-term hemodialysis. We characterize a proteolytic variant of ␤ 2 -microglobulin (cleaved after Lys 58 ) that as a trimmed form (Lys 58 is removed) can be demonstrated in the circulation in patients with chronic disease. An unexpected electrophoretic heterogeneity of these two cleaved variants was demonstrated by capillary electrophoresis under physiological conditions. Each separated into a fast and a slow component while appearing homogeneous, except for a fraction of oxidized species detected by other techniques. The two components had different binding affinities for heparin and for the amyloid-specific dye Congo red, and the equilibrium between the two forms was dependent on solvent conditions. Together with analysis of the differences in circular dichroism, the results suggest that ␤ 2microglobulin cleaved after Lys 58 readily adopts two equilibrium conformations under native conditions. In the cleaved and trimmed ␤ 2 -microglobulin that appears in vivo, the less populated conformation is characterized by an increased affinity for Congo red. These observations may help elucidate why ␤ 2 -microglobulin polymerizes as amyloid in chronic hemodialysis and facilitate the search for means to inhibit this process. ␤ 2 -Microglobulin (␤ 2 m) 1 is a small protein (M r 11,729) localized in the circulation and on cell surfaces, where it constitutes the nonpolymorphic light chain of the class I major histocompatibility complex (1). ␤ 2 m can be cleaved C-terminally to Lys 58 (Lys 58 -␤ 2 m) by activated complement component C1s. In serum, the exposed Lys 58 residue of this cleaved molecule is subsequently removed by a carboxypeptidase B-like activity. This generates a cleaved and trimmed ␤ 2 m variant called des-Lys 58 -␤ 2 m (␤ 2 -microglobulin cleaved C-terminally to Ser 57 and lacking Lys 58 ) (2). Wild-type ␤ 2 m is a one-chain protein belonging to the immunoglobulin superfamily characterized by two antiparallel ␤-sheets connected by an internal disulfide bridge in a ␤-sandwich topology without helical structures (3,4). The Lys 58 -␤ 2 m and des-Lys 58 -␤ 2 m variants are disulfide-linked two-chain heterodimers (Fig. 1). des-Lys 58 -␤ 2 m has been demonstrated in sera from patients in hemodialysis treatment (5) and in patients with malignancies and autoimmune and immunodeficiency diseases (2,6,7), but, to our knowledge, analyses of the conformational structure of the cleaved variants have not previously been published. In chronic renal failure, plasma concentrations of ␤ 2 m may increase 10 -70 times despite dialysis, and within 2 years after the onset of dialysis, about 20% of patients suffer from complications due to the formation of insoluble deposits of ␤ 2 m (␤ 2 m amyloid) in tissues and joints (8). The events leading to the formation of amyloid in this and other chronic diseases such as Alzheimer's disease are still largely unknown (9). There appears to be no simple relationship between ␤ 2 m serum concentrations and the extent of ␤ 2 m amyloidosis (10,11). Whereas different proteins and peptides are able to generate amyloid (12), all types of amyloid are ultrastructurally similar and bind the anionic dye Congo red in a characteristic fashion (13). The deposits also always contain noncovalently bound amyloid P component (9,14,15) and glycosaminoglycans, especially heparan sulfate, dermatan sulfate, and chondroitin sulfate proteoglycans (16,17). For many amyloid-forming proteins or peptides, some proportion of mutated or aberrantly processed molecules appears to be necessary for the generation of amyloid (9). ␤ 2 m-amyloid, however, appears mainly to contain intact, nonmodified ␤ 2 m molecules (14,18,19), although cleaved (20,21), deamidated (22,23), and glycated (24) ␤ 2 m may also be found in serum and urine from hemodialysis patients. Some reports also indicate the presence of minor amounts of ␤ 2 m that has undergone limited lysinespecific proteolysis in the deposits (21). The generation of ␤ 2 m amyloid may depend on a combination of a sustained increase in serum ␤ 2 m concentration and subtle changes in the molecule. We and others have recently characterized how solvent and/or pH manipulations can induce variant conformations of wild-type ␤ 2 -microglobulin including long-lived structured intermediates that may be on the folding pathway to amyloid (25)(26)(27)(28). 2 Using capillary electrophoresis (CE), we were able to demonstrate the existence of two wild-type ␤ 2 m conformations in equilibrium in the presence of acetonitrile and trifluoroethanol and the increased affinity of one of the conformations for the amyloid-specific dye Congo red (27,28). Recent data from other groups support the notion that this folding intermediate may be on the pathway to amyloid formation (25). In addition, we observed by CE that purified Lys 58 -␤ 2 m is also heterogeneous in the absence of organic solvent (27). In the present study, we characterize the Lys 58 -␤ 2 m and des-Lys 58 -␤ 2 m molecules using circular dichroism, CE, and affinity CE with heparin and Congo red. We find evidence for the existence of two conformers in cleaved ␤ 2 m and demonstrate that the variant, less populated conformation has an increased affinity for heparin and Congo red. We conclude that Lys 58 -cleaved ␤ 2 m is less conformationally constrained than wild-type ␤ 2 m and exists in two distinct and stable conformations under physiological conditions. The variant conformation resembles the organic solventinducible conformation of wild-type ␤ 2 m that may be an intermediate on the folding pathway to amyloid.
Analytical and Preparative Capillary Electrophoresis-A Beckman P/ACE 2050 instrument equipped with sample cooling was used for CE. Electrophoresis buffer was 0.2 M tricine/NaOH, pH 7.65, unless noted otherwise. Detection was by UV absorbance at 200 nm, and the separation tube was a 50-m-inner diameter, uncoated, fused silica capillary of 37-or 57-cm total length with 30 or 50 cm to the detector window. Separations were carried out at 10 kV. Data were collected and processed by the Beckman system Gold software. The capillary cooling fluid was thermostatted at 25°C. The capillary was rinsed after electrophoresis for 1 min with each of the following: 0.1 M NaOH, water, and electrophoresis buffer. The ␤ 2 m samples and the marker peptide were analyzed in the dilutions given in the figure legends. In affinity experiments, various concentrations of ligands (heparin, other glycosaminoglycans, or Congo red) were added to the electrophoresis buffer from stock solutions of 5-10 mg/ml glycosaminoglycans or 0.2 mg/ml Congo red in electrophoresis buffer. In some experiments, NaCl was added at various concentrations to assess the influence of ionic strength on the binding interactions. Preparative affinity CE followed by immunodetection was performed with 1 mg/ml LMW heparin present in the electrophoresis buffer. A sample consisting of 0.3 mg/ml Lys 58 -␤ 2 m was injected for 12 s (corresponding to a sample volume of ϳ11 s), and the instrument was programmed to change outlet vial when the two peaks of Lys 58 ␤ 2 m that were separated well in the presence of heparin were calculated to emerge at the capillary end. Outlet collecting vials contained 25 l of electrophoresis buffer or 25 mM ammonium bicarbonate (when collecting samples for mass spectrometry). The material was processed for immunodetection after 10 runs by dotting 8 l of each of the two resulting fractions onto nitrocellulose. After air drying, the nitrocellulose was blocked with 2% Tween 20 for 5 min and reacted with rabbit anti-␤ 2 m (Dako, Glostrup, Denmark; code no. A0072) at a dilution of 1:500 for 2 h at room temperature. After washing, bound primary antibody was visualized using an alkaline phosphatase-labeled swine anti-rabbit Ig (Dako; code no. D0306) at 1:1,000 for 1 h at room temperature followed by nitro blue tetrazolium staining.
Antibody Absorption-Protein G-Sepharose 4FF (Amersham Bio-sciences, Inc.) was washed in electrophoresis buffer diluted 1:10 in water, and 40 l of 1:1 suspensions were transferred to 500-l microcentrifuge tubes. Monoclonal antibodies against ␤ 2 m (BBM.1 and L368) were diluted to 2 mg/ml and 4 l of each together with 80 l of phosphate-buffered saline were added to separate tubes with the washed protein G-Sepharose. After 10 min at room temperature, the matrices were washed twice with 450 l of dilute electrophoresis buffer, and 2 l of ␤ 2 m sample (wild-type or variants) (0.2 mg/ml) and marker peptide (0.03 mg/ml) were added to individual tubes containing bound antibodies or control tubes without antibody. After overnight incubation at 4°C, the matrices were resuspended and centrifuged at 16,000 ϫ g for 2 min. Aliquots of the supernatants were finally analyzed by CE. Data Handling in Migration Shift Affinity CE-Peak appearance times (t) were equalized with respect to variations in electro-osmotic flow and shifts in buffer dielectric constant from run to run by subtracting the inverse appearance time of the added marker peptide from the inverse ␤ 2 m peak appearance times. This figure is proportional to the electrophoretic mobility (31). Triplicate sets of control runs consistently had a relative S.D. of the normalized 1/t below 0.5%. Mobility shifts were then expressed as the difference between the normalized 1/t values at various concentrations (c) of additive. Linearization of the plot for binding constant estimation took place according to the equation: (31,32). Binding constants were also derived from direct nonlinear curve fits in plots of ⌬(1/t) as a function of c (GraphPadPRISM, v. 2.01; GraphPad Software, Inc., San Diego, CA).
RP-HPLC and MS-␤ 2 m masses were verified by MALDI-MS on a Voyager Elite instrument (PerSeptive Biosystems, Framingham, MA) with delayed ion extraction using the dried-droplet method with sinnapic acid as sample matrix (33). RP-HPLC separation of individual components of wild-type ␤ 2 m and Lys 58 -␤ 2 m took place at 1 ml/min using a 60 min linear gradient of 28 -49% acetonitrile in 0.1% (v/v) aqueous trifluoroacetic acid on an analytical C18 column (5 m, 4.6 ϫ 250 mm) from Vydac (Hesperia, CA). Peaks were detected at 210 and 280 nm, collected manually, and dried down in a vacuum evaporator. The material for MS from capillary electrophoresis was collected during 44 repetitive runs in electrophoresis buffer without additions using a protein concentration of 0.6 mg/ml. Before MS, these samples were micropurified to remove buffer salts by passage through a reversephase microcolumn prepared by placing a Poros 50 R2 matrix (PerSeptive Biosystems) in an Eppendorf GelLoader tip with an extended outlet (33).
Circular Dichroism-Spectra were recorded on a Jobin Yvon Dichrograph mark V purged with nitrogen. The CD signal was calibrated with (ϩ)-camphor-10-sulfonic acid (1 mg/ml; Merck) in a 1-mm cell, assuming an ⑀ of 2.36 M Ϫ1 cm Ϫ1 at 291.0 nm. The ratio between the negative band at 191.5 nm and the positive band at 291.0 nm was in the range of 2.07-2.10. A spectral bandwidth of 1.0 and 2.0 nm in the near and far UV, respectively, was used. Samples were contained in cylindrical cells with nominal path length of 0.005-0.5 cm, and the actual path length of the 0.005-and the 0.010-cm cells was determined interferometrically. For all CD measurements, the total optical density of sample, solvent, and cell was Ͻ1.0. All measurements were made at room temperature (24 Ϯ 1°C). Spectra (180 -250, 225-250, and 240 -325 nm) were collected at a final ␤ 2 m/Lys 58 -␤ 2 m/des-Lys 58 -␤ 2 m concentration of about 1 mg/ml (85 M) in 0.1 M phosphate buffer, pH 7.1. Final smoothed CD spectra, averaged from at least three scans, were obtained after baseline subtraction. The CD is expressed as ⌬⑀ (M Ϫ1 cm Ϫ1 ) normalized to the molar concentration of peptide bond.

RESULTS
Probing ␤ 2 m and Cleaved ␤ 2 m Conformation by Circular Dichroism-The structure of ␤ 2 -microglobulin and variants thereof generated by limited proteolysis is shown in Fig. 1A. The three purified components were analyzed by CD, and the results (Fig.  1B) indicate that significant differences in the conformation of ␤ 2 m are induced in response to cleavage of the molecule. The differences in the spectra in the near-UV region (240 -325 nm) suggest changes of the environment of aromatic side chains, i.e. changes in the tertiary folding of the ␤ 2 m chain. The two cleaved forms clearly deviate from the fingerprint of the wild-type ␤ 2 m structure in this region of the spectrum. The far-UV region (180 -250 nm) provides quantitative information on the overall secondary structure content of the protein (34), and differences in the spectra in this region for the three forms are also observed (Fig.  5). The characteristic spectrum of wild-type ␤ 2 m is compatible with ϳ50% ␤-structures (strands and turns), 50% random structure, and practically no helix structures as reported previously (4,27). The changes in the far-UV spectrum of the cleaved forms, most pronounced for the Lys 58 -␤ 2 m species, are similar to the changes in the CD spectra of wild-type ␤ 2 m obtained in 10 -15% trifluoroethanol or 20 -30% acetonitrile where there is a mixture of the native fold and a solvent-induced species with less ␤-structure and more helix structure (27). Thus the far-UV spectra of Fig. 5 are compatible with the presence of a mixture of differently folded species in the Lys 58 -␤ 2 m preparation. Although it is not possible to quantitatively correct for their individual contributions to the spectra, the observations indicate that the two structures are substantially different.
Electrophoretic Separation of ␤ 2 m and Its Cleaved Variants-When the three purified ␤ 2 m species were separated by CE, more than one peak was observed in both the Lys 58 -␤ 2 m and the des-Lys 58 -␤ 2 m preparations (Fig. 2B), i.e. a smaller and more slowly migrating peak appeared after the main peaks. This was most pronounced in the Lys 58 -␤ 2 m preparation (Fig.  2B, 2), where this slow fraction constituted ϳ25% of the total sample independently on the sample dilution from 1.3 to 0.13 mg/ml (data not shown). Wild-type ␤ 2 m (Fig. 2B, 1) appeared homogenous in the CE analysis, and none of the samples were heterogeneous by SDS-PAGE except for faint bands around M r 25,000 ( Fig. 2A). However, it was noted that the cleaved forms migrated more slowly in SDS-PAGE than wild-type ␤ 2 m.
Characterization of the Components of Lys 58 -␤ 2 m Resolved by CE-Two different monoclonal antibodies against human ␤ 2 m (L368 and BBM.1) immobilized on protein G-Sepharose removed the two peaks of des-Lys 58 -␤ 2 m and Lys 58 -␤ 2 m as well as the single peak representing wild-type ␤ 2 m when added to samples before CE (results not shown). This argued against the possibility that impurities or other 200 nm-absorbing components in the sample aside from ␤ 2 m molecules contributed to the observed peaks. Also, both peaks in both the des-Lys 58 -␤ 2 m and the Lys 58 -␤ 2 m samples absorbed at 280 nm and were degraded by trypsin treatment (results not shown). Masses of the proteins measured by MALDI-MS were within 0.02% of the theoretical masses of the ␤ 2 m forms based on their amino acid sequences. No evidence of dimers or higher order polymers was seen in the mass spectra. However, two species in both the ␤ 2 m and the Lys 58 -␤ 2 m preparations differing by ϳ16 mass units (16.8 and 18.5) were separated by RP-HPLC using a shallow gradient (Fig. 3). The higher mass form (about 25% of the total amount) eluted earlier than the main component. In the MS analyses, some dissociation of the two polypeptide chains of the Lys 58 -␤ 2 m heterodimer was observed (data not shown). The A-chain (residues 1-58) was identified by its close agreement with its theoretical mass in both RP-HPLC purified peaks. The 16-unit mass addition thus resided in the B-chain that differed by this number in the two purified peaks. The observations suggest that the difference between the two HPLC-resolvable components was due to molecules with an oxidized C-terminal methionine residue (Met 99 ). This would add 16 mass units to the molecular mass. However, it was unlikely that the two peaks of Lys 58 -␤ 2 m observed in the CE analysis (Fig. 3B, inset) corresponded to the heterogeneity revealed by the RP-HPLC analysis (Fig. 3B). First, wild-type ␤ 2 m had the same amount of the putative oxoform in the HPLC analysis as Lys 58 -␤ 2 m but did not display heterogeneity in the CE analysis (Fig. 3A,  inset). Second, when the two peaks that were resolved by RP-HPLC were collected separately and reanalyzed by CE, both components gave two peaks exactly as observed in the Lys 58 -␤ 2 m starting material (data not shown).

Lys 58 -␤ 2 m Components Bind Heparin Differently and Are
Immunochemically Identical-To collect and immunochemically characterize each of the CE-resolved fast (f) and slow (s) peaks of Lys 58 -␤ 2 m, it was necessary to enhance their spacing. We found that addition to the electrophoresis buffer of heparin or heparan sulfate, highly sulfated glucosaminoglycans that are found associated with all types of amyloid (16,17), substantially increased the selectivity of the separation (Fig. 4). Heparin prolonged the appearance time of the s peak, indicating complex formation between heparin and the Lys 58 -␤ 2 m s species. Using this approach for preparative CE, it was possible to separately collect enough of the components of Lys 58 -␤ 2 m to probe them with anti-␤ 2 m antibodies. As shown in Fig. 4, material from both peaks subsequently reacted with a polyclonal anti-␤ 2 m antibody. This confirmed that ␤ 2 m was present in both peaks; thus, the apparent heterogeneity of the sample was not due to contaminating, unrelated proteins. When collecting the two peaks together in CE runs without a buffer modifier, the only components that were detectable by MS were the unmodified and the oxidized Lys 58 -␤ 2 m species that were observed in the RP-HPLC analyses (Fig. 3).
The heparin binding activity of the two Lys 58 -␤ 2 m species was quantitated by a series of affinity CE experiments (Fig. 5). As noted in the preparative affinity CE experiments (Fig. 4), the changes in the migration profile with heparin in the electrophoresis buffer were most pronounced for the s component of Lys 58 -␤ 2 m that shifted to longer appearance times as compared with the f component and with the marker peptide (Fig. 5A, M). The longer peak appearance times in the presence of heparin indicate that the s component of Lys 58 -␤ 2 m reversibly binds to the anionic heparin ligand during electrophoresis and thereby get a net increase in electrophoretic mobility, leading to a retardation of this peak. The predominant peak (f) was affected in the same way, but only at the highest concentration of heparin (Fig. 5A). The consistent shift of the whole s peak with increasing heparin additions to the buffer shows that it represents a homogeneous population of molecules with respect to heparin binding affinity.
The migration shifts observed after the addition of LMW heparin and heparan sulfate in similar affinity CE experiments were comparable, whereas smaller shifts were seen with corresponding milligram/milliliter concentrations of chondroitin sulfate (data not shown). No changes in the peak patterns were observed with additions of up to 0.5 mg/ml (1.8 mM) galactose-6-sulfate (data not shown) or 0.5 mg/ml dextran sulfate (data not shown). Decreasing magnitudes of migration shifts were observed with increasing ionic strength of the electrophoresis buffer (data not shown).
Migration shift data are summarized in Fig. 5B. The total LMW heparin concentration was used for the value of c, the concentration of ligand. The des-Lys 58 -␤ 2 m sample showed binding, especially of the minor s peak, but showed negligible migration shifts with heparin (data not shown) and is not included in the data. The data obtained with ␤ 2 m and Lys 58 -␤ 2 m (Fig. 5B) could be fitted (r 2 Ͼ 0.99) to a one site binding hyperbola. The inset of Fig. 5B shows the linearized data. The slopes of the lines are a measure of heparin affinity. Whereas the affinities for heparin of wild-type ␤ 2 m and the fast component of Lys 58 -␤ 2 m are very similar (K d ϭ 1.6 and 2.2 mM, respectively), the slow Lys 58 -␤ 2 m component has a higher affinity (K d ϭ 0.6 mM).
Shift of Lys 58 -␤ 2 m Peak Equilibrium in Organic Solvent-We found that Lys 58 -␤ 2 m samples that had been subjected to RP-HPLC purification and were resolubilized in water almost completely converted to the slow component detected by CE. Over time, this distribution slowly reverted to the normal distribution of the two forms with the fast peak as the major species (Fig. 6).
Slow Component of des-Lys 58 -␤ 2 m Binds Congo Red-With addition of the amyloid-specific dye CR to the electrophoresis buffer, we used CE to assess the binding to the cleaved ␤ 2 m variant des-Lys 58 -␤ 2 m that circulates in chronic disease (Fig.  7). Whereas both components of this molecule bind heparin poorly, it was readily observed that the s component of des-Lys 58 -␤ 2 m interacted more strongly with CR than the f species. In addition, a late peak (marked with an asterisk, Fig. 7), which we interpret as representing the CR-des-Lys 58 -␤ 2 m complex, increased proportionally with the concentration of CR. This late complex was absent in Congo red affinity CE analyses of marker or wild-type ␤ 2 m alone (data not shown). Because of the stable complex formation, it was not possible to calculate a binding constant for the interaction between the s conformation of des-Lys 58 -␤ 2 m and Congo red based on peak migration  Fig. 2B legend, except that 1 mg/ml LMW heparin was added to the electrophoresis buffer. Ten runs were performed, during which the two peaks (fast (f) and slow (s)) were collected separately by changing the position of the capillary outlet corresponding to the calculated time of peak appearance at the capillary end. For the fraction collection, the capillary outlet was placed in a vial containing a 25-l volume of electrophoresis buffer. The dot blot shows the presence of ␤ 2 m in the collected fractions (f and s); as a control, buffer alone (c) was dotted. Eight l of dots were applied and examined for reactivity with a rabbit anti-␤ 2 m antibody visualized by alkaline phosphatase-labeled swine anti-rabbit IgG.
shifts, but complex formation was clearly already detectable at 1-2 M CR (Fig. 7). DISCUSSION The CE and CD analyses of ␤ 2 m variants generated by limited proteolysis strongly suggest that these molecules exist in two distinct conformations that are in equilibrium at neutral pH in physiological buffer. Although the ␤ 2 m variants (as well as wild-type ␤ 2 m) contain both unmodified and modified molecules (probably oxidized on the C-terminal Met 99 residue), the heterogeneity demonstrated by CE was not caused by a fraction of oxidized molecules. Due to the resolution of the CE analysis and the higher affinity of one of the conformations for heparin, the distribution between the two conformations was readily quantitated under different conditions, and we found that the less abundant s conformation is favored under more hydrophobic conditions. We recently found that such conditions induce a loss of native conformation and the emergence of a specific conformational variant or a partly structured intermediate of wild-type ␤ 2 m as well (27).
Studies of wild-type human ␤ 2 m in solution using NMR  Fig. 3). The entire peak material was collected, dried down in a SpeedVac, and resolubilized in 10 l of water. Shown are repeated 8-s injections of this material at 100, 128, and 142 min after resolubilization, as indicated. The electrophoresis buffer contained 1 mg/ml low molecular weight heparin. Analysis conditions were otherwise as described in the Fig. 2B legend. techniques (4) and three-dimensional structure elucidation based on x-ray crystallography of ␤ 2 m in complex with major histocompatibility complex class I molecules (35) show that ␤ 2 m at pH 5.4-8.0 is a pure ␤-sheet protein that assumes a defined, moderately tightly packed structure of two antiparallel ␤-sheets formed around the Cys 25 -Cys 80 disulfide bridge. Lys 58 is part of a 5-residue loop between ␤-strands that is either very rigid or relatively slowly fluctuating between two or more structures (4). ␤ 2 m contains no obvious linear consensus sequences (clusters of basic amino acids) (36) for heparin binding, but it is conceivable that a break in the strained loop between Lys 58 and Asp 59 with the possibility of rotation around the disulfide bond may facilitate alternative conformations, one of which has an increased binding affinity for heparin. The consequences of non-native conformations for ␤ 2 -microglobulin function in cellular immunity remain to be elucidated, but a changed conformation is likely to influence protein-protein interactions between ␤ 2 -microglobulin and the major histocompatibility complex class I heavy chain and thereby influence the presentation of major histocompatibility complex-associated peptides. Conformational changes are supported by the observations of CE heterogeneity and by the changes in the CD spectra as well as by the overall different mobility in SDS-PAGE of the variants as compared with native ␤ 2 m (Refs. 2 and 7; Fig. 2A). Only by CE, however, is it possible to quantitatively separate the two species. The discovery that ␤ 2 m and its proteolytic variant Lys 58 -␤ 2 m are able to bind heparin is interesting because many of the peptides and proteins that have been identified as major components of different types of amyloid have been shown to bind heparan sulfate or chondroitin sulfate (37)(38)(39)(40). Amyloid-like material has been observed in vitro after simply mixing wild-type ␤ 2 m with heparan sulfate and/or serum amyloid P component (41). The K d values for the interactions with heparin estimated in the present study are maximum values because the total LMW heparin concentration was used in the plots. It is not known whether a specific subfraction of heparin (heparin is heterogenous with respect to chain length, sulfation, and disaccharide composition (29,42)) is responsible for binding. However, the binding is weak and ionic strength-dependent, i.e. involves electrostatic interactions. The presence of glycosaminoglycans in ␤ 2 m amyloid is an indication of the abnormal conformation and/or polymerization of the deposited wild-type ␤ 2 m (14,19). A key finding of the present study is that Congo red binds specifically and strongly to the conformational variant existing under native conditions in des-Lys 58 -␤ 2 m. CR resembles heparin/heparan sulfate in being a sulfated molecule with known affinity for amyloid. Taken together, however, our results indicate that anionic groups are not responsible for the binding of CR to the des-Lys 58 -␤ 2 m s conformer. The CE and CD data agree with the notion that the des-Lys 58 -␤ 2 m variant conformation is similar to the organic solvent-induced variant conformation of wild-type ␤ 2 m that also displays considerably increased affinity for CR and that may be an intermediate on the pathway to insoluble amyloid formation (27)(28)(29). A fraction of cleaved ␤ 2 m may thus be able to present differently structured intermediates and thereby instruct wild-type ␤ 2 m to malfold and precipitate as amyloid. Although many steps in this process remain to be elucidated, the approach presented here will be helpful in further analyses of the conformational intermediates and oligo-molecular com-plexes that together form amyloid or amyloid precursors and thus in the discovery of means to inhibit these processes.