J. McLaurin, Centre for Research in Neurodegenerative Diseases, Tanz Neuroscience Building, 6 Queen's Park Crescent West, Toronto, Ontario, Canada, M5S 3H2. Tel: + 1 416 978 1035, Fax: + 1 416 978 1878, E-mail: firstname.lastname@example.org
One of the major clinical features of Alzheimer's disease is the presence of extracellular amyloid plaques that are associated with glycosaminoglycan-containing proteoglycans. It has been proposed that proteoglycans and glycosaminoglycans facilitate amyloid fibril formation and/or stabilize these aggregates. Characterization of proteoglycan–protein interactions has suggested that basic amino acids in a specific conformation are necessary for glycosaminoglycan binding. Amyloid-β peptide (Aβ) has a cluster of basic amino acids at the N-terminus (residues 13–16, His-His-Gln-Lys), which are considered critical for glycosaminoglycan interactions. To understand the molecular recognition of glycosaminoglycans by Aβ, we have examined a series of synthetic peptides with systematic alanine substitutions. These include: His13→Ala, His14→Ala, Lys16→Ala, His13His14Lys16→Ala and Arg5His6→Ala. Alanine substitutions result in differences in both the secondary and fibrous structure of Aβ1–28 as determined by circular dichroism spectroscopy and electron microscopy. The results demonstrate that the His-His-Gln-Lys region of Aβ, and in particular His13, is an important structural domain, as Ala substitution produces a dysfunctional folding mutant. Interaction of the substituted peptides with heparin and chondroitin sulfate glycosaminoglycans demonstrate that although electrostatic interactions contribute to binding, nonionic interactions such as hydrogen bonding and van der Waals packing play a role in glycosaminoglycan-induced Aβ folding and aggregation.
Alzheimer's disease is characterized by the presence of amyloid plaques surrounded by dead and dying neurons . The principal component of the plaque is the amyloid-β peptide (Aβ), a 39–43 residue peptide found in normal human tissue and generated as a cleavage product from the larger amyloid precursor protein [2–5]. Aβ in the plaque is in the form of an amyloid fibril, 100 Å in diameter and several micrometers in length . The conformational change required for the conversion of soluble Aβ into amyloid fibrils has been demonstrated to be a nucleation-dependent process [7,8], modulated by pH, Aβ concentration, and the presence of nucleation seeds, such as proteoglycans and apolipoproteins [9–14].
In Alzheimer's disease, at least four types of proteoglycans colocalize with both amyloid plaques and neurofibrillary tangles suggesting that these are important in the formation of these deposits. These include: heparan sulfate proteoglycans [15,16], keratan sulfate proteoglycans , dermatan sulfate proteoglycans , and chondroitin sulfate proteoglycans . In vitro studies demonstrate that proteoglycans and glycosaminoglycans decrease the lag time of Aβ fibril formation and affect plaque turnover and removal [20,21]. The inherent binding of glycosaminoglycans to Aβ has been shown to inhibit the proteolysis of Aβ fibrils , accelerate maximal fibril formation in vitro and attenuate Aβ-induced neurotoxicity .
There is considerable interest in elucidating the specific mechanism by which proteoglycanss interact with proteins to regulate metabolic processes in both normal and disease states. However, the structural heterogeneity of the acidic mucopolysaccharides and the variety of amyloid proteins that are known to bind to these molecules have complicated the elucidation of the molecular details of the amyloid–glycosaminoglycan interactions. Studies to date have focused on the heterogeneity of the glycosaminoglycans that have indicated the importance of sulfate groups [13,14,23]. Heparin binding proteins are well characterized and suggest that clusters of basic residues on the protein surface are necessary for binding to sulfated glycosaminoglycans . Aβ has a cluster of basic amino acids in its N-terminal domain, His-His-Gln-Lys (residues 13–16) that have been suggested to interact with glycosaminoglycans . To explore the molecular recognition of glycosaminoglycans by Aβ, we have made a series of synthetic peptides with systematic alanine substitutions at positions encompassing this site and other basic domains. The effect of these substitutions on glycosaminoglycan binding, fibril formation and aggregation were examined by circular dichroism spectroscopy and electron microscopy. These studies have indicated that glycosaminoglycan binding to Aβ is enhanced by the His-His-Gln-Lys domain but that the fibrillar β-conformation is equally important for Aβ–glycosaminoglycan interactions.
Materials and methods
Αβ40, Αβ42, Αβ1–28 and alanine substituted peptides (Table 1) were synthesized by solid phase Fmoc-chemistry by the Hospital for Sick Children's Biotechnology Centre (Toronto, Ontario). Peptides were purified by reverse phase HPLC on a C18 µbondapak column and were initially dissolved in 0.5 mL of 100% trifluoroacetic acid (Aldrich Chemicals, Milwakee, WI, USA), diluted in distilled H2O and immediately lyophilized . Peptides were then dissolved in 40% trifluoroethanol (Aldrich Chemicals) in distilled H2O and stored at −20 °C until use. Alternatively, the lyophilized peptides were dissolved in distilled H2O at 10 mg·mL−1 and used immediately.
Heparin (porcine intestinal mucosa, 6000 Da, 2.4 sulfate groups per repeating sugar unit), heparan sulfate (bovine kidney, 7500 Da, 0.5–1 sulfate groups per sugar repeat), keratan sulfate (bovine cornea), chondroitin 4-sulfate (bovine trachea, 8000 Da, 1 sulfate per repeating unit), dermatan sulfate (bovine mucosa, 16 000 Da, 1.2 sulfate groups per repeating unit), and chondroitin 6-sulfate (shark cartilage, 1 sulfate group per repeating unit) were purchased from Sigma Chemicals (St Louis, MI, USA). These glycosaminoglycans were dissolved in distilled H2O at 10 mg·mL−1 and stored at −20 °C until use.
CD spectra were recorded on a Jasco Circular Dichroism Spectrometer Model J-715 (Easton, MD) at 25 °C. Spectra were obtained from 200 to 250 nm, with a 0.5-nm step, 1-nm bandwidth and 1-s collection time per step. Spectra were averaged from five repeat scans. Peptide/glycosaminoglycan ratios were 1 : 1 (w/w) with a final peptide concentration of 10 µm. The effect of the glycosaminoglycans on peptide conformation was determined by adding an aliquot of stock peptide solutions to glycosaminoglycans suspended in 20 mm sodium phosphate buffer pH 7.0 or distilled H2O (approximately pH 6.8). CD spectra were examined immediately after addition of Aβ or over a 96-h time course. The contribution of glycosaminoglycans to the CD signal was removed by subtracting the glycosaminoglycan only spectra. Aβ peptide conformations were determined in 40% trifluoroethanol/H2O, in 20 mm phosphate buffer and distilled H2O under the same conditions.
Peptides were incubated with the various glycosaminoglycans at a 1 : 1 ratio (w/w) in distilled H2O. Alternatively, peptides were incubated for 72 h at a concentration of 10 mg·mL−1 in distilled H2O to produce preformed fibrils. An aliquot of this stock was diluted to 100 µg·mL−1 in distilled H2O and incubated at a 1 : 1 ratio (w/w) with glycosaminoglycans. For negative staining, carbon-coated pioloform grids were floated on aqueous solutions of peptides (100 µg·mL−1). After grids were blotted and air dried, samples were stained with 1% (w/v) phosphotungstic acid, pH 7.0. The peptide assemblies were observed in a Hitachi H-7000 operated with an accelerating voltage of 75 kV.
Conformational transitions and fibril structure of Aβ1–28 mutant peptides
Because synthesis of Aβ40 and Aβ42 peptides with substitution in the proposed glycosaminoglycan-binding site is a difficult task, we compared the interactions of Aβ1–28 and Aβ40/42 with various glycosaminoglycans. No differences could be detected between Aβ1–28 and Aβ40 in their transition to β structure and subsequent fibril morphology in the presence and absence of all glycosaminoglycans examined (data not shown). As we have previously reported , Aβ42 in the presence of glycosaminoglycans was more highly aggregated than both Aβ1–28 and Aβ40. Therefore, all alanine substitution studies were examined using Aβ1–28 peptides. Substitutions of the charged amino acids that may be involved in Aβ lateral aggregation and glycosaminoglycan binding resulted in variations in both secondary and fibrous structures. Circular dichroism spectroscopy demonstrated that wild-type Aβ1–28, His14→Ala, Arg5His6→Ala were random when initially dissolved in distilled H2O, pH 6.8 (Fig. 1). Ageing of these peptides resulted in transition from random to β sheet conformation as has been previously reported for the wild-type peptide [27–29]. The Arg5His6→Ala mutant was indistinguishable from the wild-type peptide suggesting that these residues are not involved in the conformational transition to β structure (Fig. 1). In contrast, substitution of His13→Ala results in a random secondary structure in distilled H2O pH 6.8, which failed to undergo a conformational change even following extensive incubation (Fig. 1B). These results suggest that His13 is essential for the random coil to β sheet transition, as this single substitution is sufficient to increase the lag phase associated with a β-structural transition. The Lys16→Ala substitution resulted in a partially random conformation that readily underwent a transition to β structure upon incubation (Fig. 1D). The His13His14Lys16→Ala triple substituted peptide had a β structure at all time points (Fig. 1E). Treatment with 100% trifluoroacetic acid has been shown to disaggregate both Aβ40 and Aβ42 . Repeated treatment of His13His 14Lys16→Ala with trifluoracetic acid did not affect the secondary structure. These results demonstrate that removal of three positive charges from the central region of Aβ1–28 favours the formation of β structure. The stability appears to be side-chain-dependent as complete substitution of the His13 His14Lys16 residues to Gly resulted in destabilization of the β structure .
Negative stain electron microscopy demonstrated that substitutions in Aβ1–28 altered the Aβ fibre morphology. Wild-type Aβ1–28 formed numerous long fibres with occasional lateral aggregation under neutral conditions, pH 6.8 (Fig. 2A). The resulting fibres displayed a diameter of 70–100 Å and were several micrometers long. In agreement with the CD data, His13→Ala did not assemble into fibres after 3 days of incubation (Fig. 2B); only small aggregates and fine protofilaments could be detected after 14 days (Fig. 2B). The protofibrils were 30–60 Å in diameter and 60–100 nm in length. These results are similar to those previously seen when His13 was substituted with Asp in Aβ11–25  and further emphasize the role of His13 in folding and fibril assembly. The His14→Ala peptide formed fibres similar to wild-type but with much decreased length suggesting that this residue is not critical for fibril formation (Fig. 2C). The relative density of fibres formed by His14→Ala was less than that detected for wild-type, which suggests that although the structures of the fibres are similar to wild-type, the His14→Ala mutation may alter the kinetics of fibril formation. The Lys16→Ala substituted peptide displayed only sparse protofilaments and globular aggregates suggesting that Lys16 may be important for the formation of mature fibres but not nucleation (Fig. 2D). These results are in contrast to those seen by Kirschner and coworkers  who found that Lys16→Ala substitutions promoted stacking of Aβ11–25 β sheets in slab-like assemblies. This discrepancy is probably due to the altered packing of the shorter peptide fragments. The triple alanine substituted peptide, His13His14Lys16→Ala, formed numerous fibres of varying lengths with a diameter of 30 Å similar to that of protofilaments (Fig. 2E). This observation is consistent with previous reports for a double substituted peptide, His13His-14→Ala, which increased fibrillogenesis and decreased solubility  and the His13His14Lys16 to Gly substituted peptide, which only forms protofilaments . The Arg5His-6→Ala substituted peptide formed individual fibres, which were indistinguishable from the wild-type peptide, except that the periodicity of the helical twisting was more pronounced (Fig. 2F). The periodicity varied from 75 to 100 nm suggesting that the packing between protofilaments may vary slightly.
Effect of amino-acid substitutions on the interaction of Aβ with glycosaminoglycans
We have previously demonstrated that both heparin and chondroitin sulfates are able to induce an instantaneous random to β structure transition in Aβ40/42 . We observed that Aβ1–28 wild-type undergoes a random to β sheet transition in the presence of both heparin and dermatan sulfate when incubated in both 20 mm phosphate pH 7.0 and distilled H2O pH 6.8 (Fig. 3A). His13→Ala, His14→Ala and Lys16→Ala formed β structure upon interaction with both heparin and dermatan sulfate (Fig. 3). These results suggest that removal of a single positive charge in the putative glycosaminoglycan-binding site does not abolish glycosaminoglycan binding and the resultant-structural transition. The binding of His13His14Lys16→Ala substituted peptide to heparin resulted in a slight shift in the minimum away from 218 nm (Fig. 3E) but did not enhance the β structure observed with this peptide. The shift away from 218 nm may represent a slight change in the tertiary structure of Aβ upon binding to heparin. Similarly, dermatan sulfate did not enhance the β structure detected with the triple alanine substituted peptide (Fig. 3E). As the addition of glycosaminoglycans did not result in an increased β structure of His13His14Lys16→Ala peptide, this indicates that fibril nucleation was not enhanced. Substitution of charged residues away from the putative glycosaminoglycan-binding site, Arg5His6→Ala, did not alter the β-structural transition in the presence of both heparin and dermatan sulfate. These results suggest that Arg5, His6 are not critical for the β-structural transition associated with fibril nucleation and for glycosaminoglycan-binding.
In order to assess the effects of charge residue substitutions on glycosaminoglycan-mediated fibril formation, Aβ1–28 and all substituted peptides were incubated in the presence of heparin and dermatan sulfate at a 1 : 1 ratio (w/w) for 3 days at 37 °C. All peptides were diluted from stock solutions containing 40% trifluoroethanol into distilled H2O pH 6.8, to ensure that they were random monomers with the exception of His13His14Lys16→Ala as described above. Wild-type Aβ1–28 formed fibres of varying length, which were laterally aggregated in the presence of dermatan sulfate (Fig. 4A). Similar results were detected in the presence of heparin, heparan sulfate, chondroitin 4-sulfate and chondroitin 6-sulfate (data not shown). The Arg5His6→Ala peptide was indistinguishable from the wild-type Aβ1–28 peptide in the presence of all glycosaminoglycans. These results further emphasize that Arg5 and His6 residues are not critical for both fibre formation and glycosaminoglycan binding. Although fibres were not detected for the His13→Ala peptide alone, in the presence of dermatan sulfate fibres had a length of greater than 300 nm with a periodicity of 100 nm (Fig. 4B). His13 →Ala formed fibres in the presence of all glycosaminoglycans examined, yet the overall density was significantly lower than those formed in the presence of wild-type Aβ1–28. In contrast, His14→Ala formed many laterally aggregated fibrils in the presence of dermatan sulfate (Fig. 4C). As was demonstrated for wild-type peptide, the fibres appeared predominantly as ribbons, although some fibril intertwining was observed. In the presence of dermatan sulfate, Lys16→Ala fibres were of varying lengths, exhibited a ribbon morphology and a helical twisting with a pitch of 50–100 nm (Fig. 4D). These findings suggest that both His13 and Lys16 are important for both the formation and lateral aggregation of Aβ fibrils. The His13His14Lys16→Ala peptide formed truncated fibres present in the absence of glycosaminoglycans, upon incubation with heparin or heparan sulfate fibres were laterally aggregated or present as small aggregate masses (data not shown). In the presence of chondroitin sulfate glycosaminoglycans, the density of His13His14Lys16→Ala fibres was greater than control with some lateral aggregation. These data suggest that in the absence of the His-His-Gln-Lys domain, glycosaminoglycans are able to aggregate fibres, but to a much lesser extent.
To examine the effect of the alanine substitutions on glycosaminoglycan-induced aggregation of Aβ, we incubated all peptides in order to preform fibres. The various peptides took 2–14 days of incubation to form fibres in distilled H2O, pH 6.8; the quickest being the Arg5His6→Ala peptide in 2 days (Fig. 5A) and the slowest was the His13→Ala peptide, which took 14 days to form a few truncated fibres (Fig. 6A). Once fibres were formed, all peptides were incubated in the presence of heparin and dermatan sulfate overnight at a 1 : 1 ratio (w/w) and examined by negative stain electron microscopy. Preformed Arg5His6→Ala fibres (Fig. 5A) were readily aggregated into fibres and bundles by both heparin (Fig. 5B) and dermatan sulfate (Fig. 5C). Similar to our previous results on Aβ40/42, chondroitin sulfate glycosaminoglycans were more effective at aggregating Arg5His6→Ala fibres than heparin glycosaminoglycans. These results are similar to those observed for the wild-type peptide except that the extent of lateral aggregation was slightly less than that seem for Arg5His6→Ala (data not shown). Similarly, the His13His14-Lys16→Ala peptide was already present in a β structure, and therefore the addition of glycosaminoglycans enhanced aggregation and precipitation of large fibrillar aggregates (Fig. 6). At higher magnification, the aggregates demonstrated that heparin (Fig. 6F) and dermatan sulfate (data not shown) were able to stack the fibres into sheets of up to six fibres thick compared with peptide alone (Fig. 6C). The His14→Ala peptide fibres and aggregation pattern were similar to wild-type peptide (data not shown). His13→Ala peptide formed a few short fibres after 14 days of incubation (Fig. 6A); the addition of heparin (Fig. 6D) resulted in an increase in the number of fibres that were detected, but without apparent lateral aggregation. In contrast, dermatan sulfate incubation resulted in a significant increase in both the density of His13→Ala fibres and degree of lateral aggregation (data not shown). For Lys16→Ala, once fibres were formed (Fig. 6B), both heparin (Fig. 6E) and dermatan sulfate were able to laterally aggregate these fibres. In all cases examined, chondroitin sulfates were much more efficient than heparin glycosaminoglycans at laterally aggregating Aβ1–28 wild-type and alanine substituted peptides (Fig. 5). These data suggest that once Aβ forms fibrils, the interaction with glycosaminoglycans may be elicited through a structural motif that does not rely solely on the His-His-Gln-Lys sequence but may involve other charge residues or nonionic interactions such as van der Waals and hydrophobic interactions.
Amyloid deposition is a factor in numerous disorders including rheumatoid arthritis, Alzheimer's disease and familial amyloid polyneuropathy (reviewed in ). Although the precursor proteins in the various diseases do not share sequence homology, the morphology and structure of the resultant amyloid fibres are similar suggesting a common pathway of fibril formation . The fibrils are of varying lengths, unbranched, 70–120 Å in diameter and have a characteristic β-sheet conformation and protofilament organization. Molecules, such as glycosaminoglycans, that affect nucleation and propagation of these fibres, may act by stabilizing the β structure and inter–protofilament interactions. In the case of the glycosaminoglycans, one possibility is that these molecules bind to the region of His13–Lys16 of Aβ between two neighbouring protofilaments and thereby stabilize the tertiary structure of Aβ fibrils. Our results demonstrate that substitution of these basic residues with alanine results in alterations in the rate of formation and structure of Aβ fibrils but does not effect Aβ–glycosaminoglycan binding. This would suggest that the cluster of basic residues may aid the kinetics of glycosaminoglycan binding, but that they are not critical. These results also suggest that ultimately glycosaminoglycan binding is more sensitive to alterations in the secondary structure of Aβ. The contribution of a conformational motif to Aβ–glycosaminoglycan interactions is supported by in vitro studies using affinity co-electrophoresis to examine Aβ and amylin peptides in the presence of heparin , fluorescence spectroscopy to examine amylin in the presence of perlecan , and immunoglobulin light chain interactions with heparin and chondroitin sulfates .
Further evidence that glycosaminoglycan binding may be precipitated through a structural motif rather than a specific sequence arises from comparison of Aβ with other β-structured glycosaminoglycan-binding proteins. Midkine is a 13-kDa heparin-binding polypeptide that promotes neurite outgrowth and neuronal survival. Midkine has two structural domains consisting of three antiparallel β strands, which have been suggested to bind to heparin as a head to head dimer . The most well characterized β-structured glycosaminoglycan binding protein is basic fibroblast growth factor. Mutations of basic residues in basic fibroblast growth factor have demonstrated that high affinity heparin binding critically depends on an intact 3D structure of the growth factor rather than on specific heparin-binding sequence domains [39–41]. In fact, conformational alterations in basic fibroblast growth factor were shown to inhibit heparin binding. Site-directed mutagenesis and titrating calorimetry permitted an analysis of the energetic contributions of individual residues in the binding of heparin to basic fibroblast growth factor . These studies demonstrated that pure electrostatic interactions only contribute 30% of the binding free energy and that more specific nonionic interactions such as hydrogen bonding and van der Waals packing contribute to the majority of the free energy for the binding of basic fibroblast growth factor to heparin or heparan sulfate. The residues, Asn27, Thr121, Gln123 and Gln134 form a pocket in basic fibroblast growth factor, which represent another 33% of the binding free energy between basic fibroblast growth factor and heparin. The remaining 37% of the binding free energy is attributed to nonionic interactions with charged amino-acid side-chains, such as Lys125, which may interact through the aliphatic portion of the side-chain. The low contribution of basic residues to heparin binding was confirmed by NaCl elution curves that were generated on heparin–Sepharose affinity columns . Similarly, hydrophobic and nonionic interactions also control heparin binding to anti-thrombin . Finally, glycosaminoglycan chains have been shown to interact and form aggregated mesh works. This interaction is dependent on hydrophobic binding and, by extrapolation, indicates that glycosaminoglycans may favour binding domains consisting of hydrophobic interactions with other molecules . These data support the involvement of residues other than the classical cluster of basic residues for glycosaminoglycan binding. Similar interactions can be envisioned for Aβ–glycosaminoglycan interactions, because elimination of the classical His-His-Gln-Lys region did not abolish glycosaminoglycan-induced aggregation of His13His-14Lys 16→Ala peptide.
In Alzheimer's disease, binding of Aβ peptides to glycosaminoglycans has been implicated in plaque formation, competence and persistence, as well as attenuation and induction of Aβ-associated neurotoxicity [13,22,35,45]. These various phenomena raise the possibility of designing therapeutic interventions that would limit Aβ deposition and toxicity. Understanding the mechanism whereby Aβ binds to both cell surface and basement membrane associated glycosaminoglycans will aid in the development of more specific treatments. The ability of Aβ to bind to glycosaminoglycans through a β-structured motif suggests that development of compounds that stabilize Aβ in a monomeric, random structure would ablate the formation of plaques and possibly Aβ-induced toxicity. This strategy is further supported by the dysfunctional folding mutant His13→Ala, which we have shown to be the critical residue in Aβ folding and therefore represents a viable target for anti-amyloid agents that would be able to control the earliest stage of the amyloid pathway. Although this study has evaluated Aβ interactions with glycosaminoglycans, these observations have significant implications for other amyloid disorders whose amyloidogenic proteins share a common fibril structure. The common β-structure motif of amyloid fibrils may explain the universal colocalization of glycosaminoglycans to amyloid deposits. Our results also demonstrate that the His-His-Gln-Lys is an essential region for amyloid formation, therefore drugs that target this region may be effective at blocking the β-structural transition necessary for fibril nucleation.
The authors would like to thank Dr N. Wang at the Hospital for Sick Children's Biotechnology Centre for the synthesis of all peptides used in this study. This work was supported by grants from the Ontario Mental Health Foundation (J. M. and P. E. F.), University of Toronto's Dean's Fund (J. M.), Banting Research Foundation (J. M.), the Scottish Rite Charitable Foundation (P. E. F.) and Neurochem Inc. (P. E. F). P. E. F. would like to acknowledge the support of the Alzheimer Society of Ontario. J. M. would like to acknowledge the Kevin Burke Memorial Amyloid Fund.