Heparin-binding Properties of the Amyloidogenic Peptides Aβ and Amylin

Aggregation and deposition of the 40–42-residue amyloid β-protein (Aβ) are early and necessary neuropathological events in Alzheimer’s disease. An understanding of the molecular interactions that trigger these events is important for therapeutic strategies aimed at blocking Aβ plaque formation at the earliest stages. Heparan sulfate proteoglycans may play a fundamental role since they are invariably associated with Aβ and other amyloid deposits at all stages. However, the nature of the Aβ-heparan sulfate proteoglycan binding has been difficult to elucidate because of the strong tendency of Aβ to self-aggregate. Affinity co-electrophoresis can measure the binding of proteoglycans or glycosaminoglycans to proteins without altering the physical state of the protein during the assay. We used affinity co-electrophoresis to study the interaction between Aβ and the glycosaminoglycan heparin and found that the aggregation state of Aβ governs its heparin-binding properties: heparin binds to fibrillar but not nonfibrillar Aβ. The amyloid binding dye, Congo red, inhibited the interaction in a specific and dose-dependent manner. The “Dutch” mutant AβE22Q peptide formed fibrils more readily than wild type Aβ and it also attained a heparin-binding state more readily, but, once formed, mutant and wild type fibrils bound heparin with similar affinities. The heparin-binding ability of aggregated AβE22Q was reversible with incubation in a solvent that promotes α-helical conformation, further suggesting that conformation of the peptide is important. Studies with another human amyloidogenic protein, amylin, suggested that its heparin-binding properties were also dependent on aggregation state. These results demonstrate the dependence of the Aβ-heparin interaction on the conformation and aggregation state of Aβ rather than primary sequence alone, and suggest methods of interfering with this association.

A hallmark of both Alzheimer's disease (AD) 1 and Down's syndrome is the presence of numerous extracellular deposits of the amyloid ␤-protein (A␤), termed senile or neuritic plaques, in the brain parenchyma. In most cases, A␤ is also deposited in the walls of parenchymal and meningeal blood vessels. In these two types of deposits, A␤ exists largely in a fibrillar form consisting of 40 or 42 amino acid monomers aggregated into insoluble filamentous polymers. A␤, which is derived by endoproteolysis from the ␤-amyloid precursor protein (1,2), is by far the major constituent of plaques (3)(4)(5). However, several other plaque-associated proteins have been described, including ␣ 1antichymotrypsin (6), apolipoprotein E (7), the heparan sulfate proteoglycan (HSPG) perlecan (8), serum amyloid P component (9), and complement factors (10).
The order in which these various proteinaceous components are added to the senile plaques is not well understood, but some clues can be obtained from the composition of another type of A␤ deposit found in AD brains. "Diffuse" plaques are composed of A␤ in a particulate but not fibrillar form and do not react with the classic amyloid-staining dyes, Congo red and thioflavin S. Because the brains of young (Ͻ30 year old) Down's syndrome patients almost exclusively contain diffuse plaques, these deposits are believed to be the precursors of the compacted amyloid plaques, which invariably develop in older (Ͼ30 -40 year old) Down's syndrome subjects (11,12).
Snow and colleagues (8,13) have detected immunohistochemically the presence of a specific HSPG, perlecan, in the compacted plaques and cerebrovascular amyloid of Alzheimer's disease brain. In addition, diffuse plaques in the hippocampus and cerebral cortex, but not in the cerebellum, were shown to contain perlecan (13,14). Because compacted plaques are rarely found in the cerebellum even in end stage AD brains (15), it is postulated that HSPGs such as perlecan could play a role in the transition of diffuse A␤ deposits into compacted amyloid. The finding that cortical diffuse plaques in Down's syndrome brain are also perlecan-immunoreactive (13) is consistent with this hypothesis. HSPGs may play a general role in the formation and stabilization of many types of amyloid, since they have also been identified in association with amyloid deposits in virtually all other human amyloid diseases (for review, see Ref. 16).
In AD cerebrovasculature, A␤ amyloid deposits have been ultrastructurally localized to the vascular basement membrane region of capillaries, arterioles, and small arteries (17), where perlecan is a prominent constituent. In addition, one of the diseases linked genetically to ␤-amyloid precursor protein (He-reditary Cerebral Hemorrhage with Amyloidosis, Dutch-type (HCHWA-D)) causes a particularly severe deposition of A␤ in meningocerebral blood vessels (18,19), again suggesting an important role for a vascular basement membrane factor in A␤ deposition.
Previous attempts to characterize the binding between HSPGs and A␤ (20 -23) have been hindered by the unique difficulties of working with synthetic A␤, a highly hydrophobic 40 -42-amino acid peptide that readily precipitates into insoluble aggregates in vitro. Affinity co-electrophoresis (ACE) is an advantageous method to characterize the binding of proteins to proteoglycans or their glycosaminoglycan side chains (24,25). Each component is freely mobile within a highly porous native agarose gel, and no coupling of either component to any matrix, resin, or solid support is required, so all of the potential binding surfaces remain available. Importantly, A␤ has less opportunity to aggregate during the experiment than in many other types of binding assays.
In this paper, we use ACE to characterize the binding of A␤ peptides to heparin (a glycosaminoglycan that is frequently used as a model for tissue heparan sulfates) at physiological pH and ionic strength. Both wild type A␤ and the HCHWA-D disease-causing mutant form were studied. We show that the heparin-A␤ interaction is critically dependent on the secondary structure and aggregation state of A␤ and is potently inhibited by the amyloid-binding dye Congo red. A␤ containing the HCHWA-D mutation binds heparin more readily than wild type peptide due to its increased tendency to form fibrils, not because of a greater affinity for heparin. Finally, we demonstrate that another amyloid-forming subunit, human amylin, also binds to heparin, whereas the nontoxic and non-amyloidogenic rat isoform of amylin does not.

EXPERIMENTAL PROCEDURES
Peptide Preparations and Electron Microscopy-Wild type A␤  peptide was synthesized and HPLC-purified by Dr. D. Teplow (Biopolymer Laboratory, Brigham and Women's Hospital). A␤ 1-40 containing the E22Q "Dutch" mutation (A␤ E22Q ) was made by Dr. D. Chin (University of Missouri, Columbia) and aliquots were HPLC-purified by Dr. D. Walsh (Biopolymer Facility, Brigham and Women's Hospital). Human and rat amylin peptides were purchased from Peninsula Laboratories (Belmont, CA) or Bachem (Torrance, CA). The amino acid sequence of wild type human A␤ 1-40 is DAEFRHDSGYEVHHQKLV-FFAEDVGSNKGAIIGLMVGGVV (see, e.g. Ref. 1); in mutant A␤ E22Q , Glu 22 (underlined) is changed to Gln (19).
For amylin and nonfibrillar A␤ preparations, lyophilized peptides were freshly resuspended in water (initially at 1 mM) and used immediately. Alternatively, fibrillar A␤ was formed by resuspending the lyophilized peptide in water to 1 mM, then adding artificial cerebrospinal fluid (CSF; 150 mM NaCl, 3 mM KCl, 1.7 mM CaCl 2 , 0.9 mM MgCl 2 in 1.5 mM phosphate buffer, pH 6.5) to achieve a final concentration of 100 M A␤ peptide and rocking it for 2 days at room temperature, followed by centrifugation at 16,000 ϫ g for 15 min. The pellet was resuspended in ACE electrophoresis running buffer (50 mM NaMOPSO in 125 mM acetate buffer, pH 7.0; Ref. 24) and sonicated before use. A similar protocol using an ultrafiltrate of CSF showed A␤ 1-40 fibril formation after 48 h at room temperature (26).
Duplicate aliquots of each peptide preparation were collected and the peptide concentrations were determined by amino acid analysis, performed by M. Condron (Brigham and Women's Hospital Biopolymer Laboratory). Simultaneously, aliquots of each peptide preparation were applied to a 200-mesh copper-Formvar/carbon grid (EM Sciences, Gibbstown, NJ), negatively stained with 2% uranyl acetate, and observed with a JEOL JEM 100CX-II electron microscope at 60 kV. Congo red birefringence was determined by the protocol of Castañ o et al. (27).
Heparin Tracer Preparation-Tyramine was added to the reducing ends of unbleached heparin molecules (Grade I, porcine intestinal mucosa, Sigma) via a reductive amidation reaction (28). This method adds Յ1 tyramine per heparin molecule and does not disturb the chemistry of the heparin along its chain length. The tyramine moiety was iodinated by the IODOGEN method (Pierce), and the heparin was then size-fractionated by gel filtration. The low molecular weight (LMW) fraction (Յ 6 kDa) was used to minimize the number of potential binding sites per heparin molecule. Unlabeled LMW-heparin was from Sigma.
Affinity Co-electrophoresis-Horizontal 1% (w/v) agarose gels were cast with two combs essentially as described (24,25), producing a set of nine rectangular wells and a perpendicularly oriented slot (Fig. 1). Serial dilutions of each peptide preparation were prepared in ACE electrophoresis running buffer (see above) and mixed with an equal volume of molten 2% agarose, cast into the nine wells, and allowed to set. The slot was loaded with 125 I-LMW-heparin, which was subjected to electrophoresis through the protein-containing zones. After drying the gels, the positions of bands were measured using a PhosphorImager 400A and ImageQuant software (Molecular Dynamics). Staining the gels with Coomassie Brilliant Blue showed that A␤ itself did not noticeably migrate from its original position during the electrophoresis times used (not shown).  (25). In general, better fits were obtained with n ϭ 2, suggesting either positive cooperativity in binding, or perhaps the fact that some A␤ fibrils are lost on the walls of tubes and pipette tips in the process of making serial dilutions, causing A␤ concentrations to be somewhat overestimated at the highest dilutions. Protein concentrations are given in units of molarity of monomer, and were determined by amino acid analysis. Values for K d were calculated using these units of molarity; a calculation based on the molarity of polymerized fibrils would yield a much lower K d (see "Discussion").
Solution Phase Binding Assay-Preparations of fibrillar or nonfibrillar A␤ were incubated for 1 h at room temperature with 125 I-LMWheparin in ACE electrophoresis buffer. In some cases, serial dilutions of Congo red or of unlabeled LMW-heparin were included; these solutions were made fresh for each experiment. The mixtures were then transferred to nitrocellulose by vacuum filtration through a dot-blot apparatus, followed by one rinse with phosphate-buffered saline. The filter was dried and the bound radioactivity retained on the filter was counted directly in a ␥ counter (Cobra II AutoGamma, Packard Instruments). Background radioactivity in control wells containing 125 I-LMW-heparin tracer alone were subtracted from the values for A␤-containing wells before analysis. Values from six control wells containing only A␤ and heparin tracer were averaged to provide a value for maximum A␤heparin binding.

Fibrillar, but Not Non-fibrillar, A␤ Binds Heparin-ACE
experiments to measure the binding of A␤ to heparin were initially attempted using aliquots of lyophilized wild type A␤ 1-40 that were freshly resuspended in water, diluted, and assayed immediately. Samples were incorporated into the nine lanes of a 1% agarose gel, and 125 I-LMW-heparin was subjected to electrophoresis through those lanes (Fig. 1A). At neutral pH, the mobility of A␤ is much less than that of heparin (data not shown), so we expected a complex of heparin and A␤ to have a mobility significantly less than that of free heparin. Assuming this is the case, any binding of A␤ to heparin should have been revealed as a series of electrophoretic bands that were progressively retarded with increasing A␤ concentration (Fig. 1B). In contrast, no effect of A␤ on heparin mobility was seen at peptide concentrations up to 243 M ( Fig. 2A).
These data indicate that either A␤ and heparin do not bind under the conditions of the assay, or that binding could not be detected because the mobility of the A␤-heparin complex is not sufficiently different from that of heparin alone. To test the latter possibility, we reversed the roles of heparin and peptide in the experiment (25). In this case, heparin was cast into the nine parallel lanes, and 125 I-A␤ introduced into a single slot perpendicular to the lanes. Because heparin is more mobile than A␤, the peptide-containing slot was placed between the heparin lanes and the anode (Fig. 1C). In this experiment, any binding of heparin to A␤ should have been revealed as a series of electrophoretic bands whose mobility was progressively increased by increasing heparin concentrations. However, as shown in Fig. 2B, no change in the mobility of 125 I-A␤ was seen at heparin concentrations up to 10 mg/ml (Ͼ1.5 mM).
As before, the only way we could have failed to detect actual binding of heparin and A␤ in this experiment is if the mobility of the heparin-peptide complex was not significantly different from that of the labeled species, in this case A␤. However, it is not possible for the mobility of the heparin-A␤ complex to be indistinguishable from both that of free A␤ and that of free heparin, as A␤ and heparin have mobilities that are very different from each other. Thus, the negative results in Fig. 2, A and B, together indicate that freshly resuspended A␤ and heparin do not bind each other (K d Ն 0.25 mM).
Because the close association of heparan sulfate with A␤ deposits in AD brain still suggested an important interaction between the two molecules, we next considered the possibility that the secondary structure of polymerized fibrillar A␤ creates a heparin-binding epitope. To aggregate A␤ into fibrils, aliquots of lyophilized A␤ 1-40 were resuspended initially in water and then diluted into artificial CSF and rocked for 48 h at room temperature. Electron microscopy of the pelleted precipitate confirmed the formation of straight, unbranched filaments approximately 10 nm in diameter (Fig. 2D), similar to the structure and dimensions of A␤ fibrils purified from Alzheimer's plaques (29,30). Furthermore, after staining with Congo red, A␤ fibrils showed birefringence under polarized light (not shown). In contrast, electron microscopy of the freshly resuspended nonfibrillar A␤ preparation demonstrated large amorphous clumps (Fig. 2C), indicating the presence of aggregated but not fibrillar A␤. Upon reaction with Congo red, the nonfi-brillar A␤ was not birefringent (not shown).
Increasing concentrations of the A␤ fibrils were then loaded into the lanes of an ACE gel and tested for heparin binding. In sharp contrast to what was observed with the freshly resuspended peptide, low micromolar concentrations of fibrillar A␤ completely retarded the mobility of the heparin tracer (Fig.  2E). To confirm that the retardation was due to a specific binding interaction and not to nonspecific physical blockage of the heparin by a dense network of fibrils, we demonstrated that the binding of 125 I-LMW-heparin to A␤ fibrils could be completely competed away by excess unlabeled LMW-heparin (Fig. 2F).
Several independently aggregated fibril preparations showed electrophoretic retardation profiles very similar to that shown in Fig. 2E. To measure binding affinity, we calculated retardation coefficients, R, from each of several ACE gels. The results from five experiments were fit to the equation R ϭ R ϱ /(1 ϩ (K d /[protein]) 2 ) (25, 31), which yielded an average K d of 1.31 Ϯ 0.10 M (Fig. 3; this value is expressed in units of molarity of A␤ monomer (see "Experimental Procedures")).
As an independent test of the results obtained by ACE, a solution-phase binding assay was also used. Increasing concentrations of fibrillar or nonfibrillar A␤ were incubated with 125 I-LMW-heparin in solution, after which the mixtures were applied to a nitrocellulose membrane by vacuum filtration and counted to measure the retained (protein-bound) heparin. Again, fibrillar but not non-fibrillar A␤ bound heparin with a low micromolar affinity (not shown).
Congo Red Blocks the Binding of Heparin to A␤ Fibrils-Congo red is a histochemical dye used for the detection of amyloids of all types in tissue sections. The binding of Congo red to A␤ has been postulated to depend in part on fibrillar A␤ structure (32). We tested whether Congo red could inhibit the binding of heparin to fibrillar A␤. Equal aliquots of fibrillar A␤ 1-40 were incubated with increasing concentrations of Congo red, and then loaded into the lanes of an ACE gel. Compared with the control lanes of Congo red alone (Fig. 4A, lane 7) and fibrils alone (lane 8), a dose-dependent inhibition of heparin binding to the fibrils was observed (lanes 1-6).
We also used the solution-phase binding assay (above) to confirm the interaction of Congo red with fibrillar A␤. In the ACE gels, an inhibition of heparin-protein binding is most readily visualized when the protein concentration in the lanes containing no inhibitor is high enough to completely retard the migration of heparin (see, e.g., Fig. 4A, lane 8). However, using the solution-phase binding assay, we were able to test a range of A␤ concentrations to show that the IC 50 for Congo red inhibition of heparin binding increases with increasing protein concentration (Fig. 4B). When similar concentrations of A␤ were compared in the two assays, ACE gels showed a higher IC 50 for Congo red than the solution-phase assay (Fig. 4, A  versus B, middle trace). This may be due to the migration of the negatively charged Congo red out of the well during ACE electrophoresis, decreasing the total amount of drug in the lane.
HCHWA-D Mutant A␤ Fibrils Have an Affinity for Heparin Similar to That of Wild Type Fibrils-HPLC-purified A␤ 1-40 containing the E22Q (HCHWA-D) substitution was either freshly resuspended from a lyophilized aliquot or aggregated into fibrils as described above. Serial dilutions of each preparation were loaded into the lanes of ACE gels and tested for binding to 125 I-LMW-heparin. The freshly resuspended peptide did not bind to heparin up to peptide concentrations of 15.2 M (Fig. 5A, lane 1), whereas 5.7 M fibrillar A␤ E22Q almost completely retarded the mobility of heparin (Fig. 5B, lane 4).
From autoradiograms of ACE gels such as that in Fig peptide concentration (Fig. 5C). Values of K d were calculated as described above, and results from several independent preparations of fibrillar A␤ E22Q yielded an average K d of 3.45 Ϯ 0.08 M. This is in the low micromolar range, similar to the value measured for heparin binding to wild type A␤ fibrils prepared by the same method.
In contrast, when A␤ was subjected to a milder aggregation protocol, we observed a distinct difference in the heparin-binding behavior of wild type and mutant peptides. In this case, each peptide was neutralized, incubated for 48 h at 4°C at a concentration of 1 mM in water and then lyophilized. Immediately after resuspension in water, the wild type A␤ remained in a non-heparin binding state (data not shown). However, after identical treatment, the mutant peptide was able to bind heparin (Fig. 6A) and showed an affinity similar to that of preformed A␤ E22Q fibrils. We interpreted this result as indicating that water-aggregated A␤ E22Q adopted a structure similar to that of wild type A␤ that had been aggregated in artificial CSF (i.e. fibrils with substantial ␤-pleated sheet conformation). Further evidence for the presence of this conformation came from the fact that the interaction of heparin with the treated A␤ E22Q was blocked by Congo red in a dose-dependent manner (Fig.  6B). In addition, the ability of these A␤ E22Q samples to bind heparin was reversed by overnight incubation in (1,1,1,3,3,3)hexafluoroisopropanol (HFIP) for 72 h and lyophilization (Fig.  6C). HFIP is a solvent that is thought to promote or stabilize ␣-helices (33), and that can also reverse the neurotoxicity of fibrillar A␤ in cell cultures (34).
The Human Isoform of Amylin Binds Heparin-The presence of heparan sulfate has been demonstrated immunohistochemically in amyloid deposits from most human amyloidotic diseases (for review, see Ref. 16). One well known example is the islet amyloid of type II diabetes (35). In this disease, the 37residue amyloid subunit protein, called amylin, is the major constituent of the extracellular amyloid deposits that surround pancreatic islet cells. In vitro, the human isoform of amylin readily forms fibrils that are toxic to cultured islet cells (36). However, the six amino acid substitutions in the rat isoform (Fig. 7A) prevent both its assembly into fibrils and its cytotoxicity (36). We used ACE to test the heparin-binding properties of the human and rat isoforms of amylin. Each was freshly resuspended in water from lyophilized aliquots. ACE profiles of 125 I-LMW-heparin binding to HPLC-purified human and rat amylin are shown in Fig. 7 (B and C). Human amylin bound to heparin at peptide concentrations as low as 200 nM, whereas rat amylin did not bind to heparin even at 16.7 M. Amylin peptides obtained from two different sources (Bachem and Peninsula Laboratories) gave identical results (data not shown). Only the human isoform was birefringent after Congo red staining (not shown), indicating its aggregation and ␤-pleated conformation. Congo red blocked the amylin-heparin interaction in a dose-dependent manner (Fig. 7D), and the heparinbinding structure could be removed by centrifugation at low speed (Fig. 6E). DISCUSSION Heparan sulfate proteoglycans are integral and invariant components of parenchymal and vascular amyloid deposits in Alzheimer's disease and other amyloidoses. However, characterization of the binding interaction between HSPG and A␤ has been hindered by the unique difficulties of working with A␤, a 40 -42-residue amphipathic peptide that readily precipitates into insoluble aggregates in vitro. Since monomers, short oligomers and fibrils of A␤ have different biochemical properties and are likely to have different roles in AD pathology, methods used to study the interaction of A␤ and HSPG should affect as little as possible the aggregation state of the A␤ peptides.
In this paper, we have used ACE in native agarose gels at physiological pH and ionic strength to evaluate heparin-A␤ binding and explore its dependence on A␤ primary and secondary structure.
Examination of the primary amino acid sequence of A␤ re- veals that residues 12-17 (VHHQKL) fit the proposed consensus for a linear heparin-binding site (XBBXBX, where B is a basic residue) (37). However, from our experiments using up to 350 M nonfibrillar A␤ 1-40 , we conclude that in the structural context of nonfibrillar A␤, VHHQKL is an inactive heparinbinding sequence. Brunden et al. (20) reached a similar conclusion by showing that the corresponding synthetic hexapeptide does not bind to a heparin affinity column at neutral pH.
In contrast to nonfibrillar A␤, fibrillar A␤ readily bound to heparin. Our measurements, using ACE, show that the affinity of heparin for wild type A␤ 1-40 fibrils is in the very low micromolar range. Importantly, this value was calculated using the molarity of A␤ monomers, as determined by amino acid analysis; a calculation based on the molarity of fibrils would yield a much lower K d . The number of monomers per fibril is unknown and is probably highly variable in vivo; however, a hypothetical fibril containing 10 4 monomers with a measured K d of 2 M would have a K d of 200 pM when expressed in units of fibril concentration. Our results appear to be consistent with those of Gupta-Bansal et al. (38).
We used low molecular weight heparin in our tests of binding to A␤ fibrils to decrease the possibility of multiple A␤-binding sites along the heparin chain length. Future experiments using defined heparin fragments of progressively lower molecular weight may help to determine the minimum region on a hep- arin chain required for fibril binding. Additionally, we used fully formed A␤ fibrils to exclude any simultaneous effects of heparin on A␤ aggregation and to measure only the binding affinity of heparin for a characterized population of A␤. With the recent isolation of stable protofibrillar intermediates in the pathway of A␤ fibrillogenesis (39,40), it is now possible to compare the heparin-binding properties of A␤ monomers, low molecular weight oligomers, protofibrils, and other intermediates. We favor the hypothesis that endogenous HSPGs (especially their long, sulfated heparan sulfate chains) initially may bind to one or more of these intermediates, creating a template for further A␤ aggregation.
The specific amino acids on the surface of amyloid fibrils that mediate the heparin-A␤ interaction remain to be identified. In the structural context of fibrillar A␤, the linear VHHQKL epitope within an A␤ monomer might adopt a conformation that enables it to bind heparin. Alternatively, basic residues from multiple A␤ monomers that are brought together on the fibril surface may constitute the heparin-binding site. Examples of proteins in which positively charged heparin-binding "patches" are formed from basic residues contributed by multiple protein domains or subunits include type I collagen and basic fibroblast growth factor (reviewed in Ref. 41).
The marked accumulation of mutant A␤ peptides in the basement membranes of meningeal and cerebral blood vessels in HCHWA-D amyloidosis led us to speculate that this mutant A␤ might have a higher affinity for vascular basement membrane HSPGs. Indeed, the HCHWA-D mutation (Glu 3 Gln) removes one negative charge from the A␤ 1-40 peptide, which could theoretically enhance its binding to sulfated HS chains. As measured by ACE, however, the heparin-binding properties of the mutant peptide were similar to those of wild type peptide for both the fibrillar (binding) and nonfibrillar (non-binding) forms. ACE experiments using the mutant A␤ E22Q did demonstrate that the mutant peptide assumed a heparin-binding conformation more readily than the wild type. This finding is consistent with previous studies which showed the enhanced ability of A␤ E22Q to stably self-associate in vitro (42)(43)(44)(45). Two additional lines of evidence support ␤-sheet structure as a feature essential for heparin-binding. First, the amyloid-binding dye Congo red blocked the interaction; and second, incubation of A␤ E22Q in the ␣ helix-promoting solvent HFIP reversed its ability to bind heparin.
Amylin is the major protein subunit of pancreatic amyloid deposits in patients with type II diabetes. Lorenzo et al. (36) demonstrated that full-length human amylin formed amyloidlike fibrils in vitro and was toxic to cultures of pancreatic islet cells. The six amino acid substitutions in the rat isoform of amylin (three of them conservative and three that introduce prolines in place of alanines and serines) blocked both the fibril formation and toxicity of amylin (36). Because HSPGs have been detected immunohistochemically in the pancreatic amyloid deposits of type II diabetes (35), we investigated the binding of amylin to heparin. Despite the lack of a linear consensus heparin-binding sequence (i.e. XBBXBX; Ref. 37), freshly resuspended human amylin from two different sources avidly bound to heparin. In contrast to A␤, we found that freshly resuspended human amylin bound to heparin immediately. This may be due either to a novel linear heparin-binding epitope or, more probably, due to the strong tendency of human amylin to assume an insoluble ␤-sheet conformation in water. The ability of Congo red to block the interaction, the removal of the heparin-binding species by brief low speed centrifugation, and the lack of heparin binding by rat amylin all support the latter hypothesis.
In Alzheimer's disease, the pathological consequences of A␤ binding to HSPGs could include protection from proteolytic degradation and clearance (38,46), and/or creation of a favorable surface for further A␤ deposition. Cell surface proteoglycans may also play a direct role in mediating the neurotoxic effects of A␤ (47). These various observations raise the possibility that pharmacologic inhibition of the HSPG-A␤ interaction could be a means of therapeutic intervention in AD. For example, Congo red is known to block the neurotoxicity of A␤ (47,48), and we have shown here that Congo red also inhibits A␤ binding to glycosaminoglycans. Congo red analogues that are able to penetrate the blood-brain barrier might ameliorate the neurotoxicity of aggregated A␤ as well as prevent the formation or maturation of new A␤ deposits. As a first step, it would be interesting to determine whether Congo red analogues retard the development of cerebral amyloidosis and subsequent AD neuropathology in a transgenic mouse model of Alzheimer's disease (49,50). Oral administration of other small sulfated or sulfonated compounds were highly effective in reducing the progression of inflammation-associated AA amyloidosis in an experimental mouse model (51), so this approach seems feasible. Finally, the major HSPG of adult brain parenchyma is probably glypican (52), a cell-surface HSPG expressed by many populations of neurons. Our preliminary experiments indicate that both glypican and perlecan can bind fibrillar A␤ in ACE assays. 2 In future, we plan to use ACE both to refine the characterization of the heparin-binding conformation of A␤ as well as to more thoroughly characterize the binding of intact HSPGs such as glypican and perlecan to fibrillar and nonfibrillar A␤.