Structural properties of Aβ protofibrils stabilized by a small molecule

Materials and Methods

Materials and General Methods. Purified WT Aβ(1-40), with amino acid sequence DAEFRHDSGY EVHHQKLVFF AEDVGSNKGA IIGLMVGGVV, was obtained from the Keck Biotechnology Center at Yale University (New Haven, CT). Proline replacement mutants were obtained unpurified from either the Keck Center or from the Stanford University Protein and Nucleic Acid Facility (Palo Alto, CA). N-terminally biotinylated WT Aβ(1-40) was obtained from the Keck Center. Analogs were purified on a C3 reverse-phase column, and the identity of the material was confirmed by MS on an Agilent 1100 LC/MSD. The Library of Pharmacologically Active Compounds (LOPAC) collection and pure samples of CLC were obtained from Sigma. The anti-amyloid mAb WO1 (11) was obtained from high-density cell culture and used without purification. The amount of CLC in CLC-Aβ aggregates was determined by both HPLC and MS, in both cases by using a standard curve generated from analysis of a stock solution of CLC.

Microtiterplate Elongation and Screening Assays. The Aβ(1-40) microplate elongation assay has been described (12). For screening the compound library, plates were coated with 100 ng of Aβ (1-40) fibrils and washed. The assay was initiated by adding to each well, in order, 80 μl of assay buffer [PBSA (PBS containing 0.05% (wt/vol) sodium azide) plus 0.05% Tween 20], 10 μl of a 1 mM solution of test compound in 50% DMSO, and 10 μl of 100 nM biotinyl-Aβ in assay buffer. Plates were incubated 30 min at 37°C and then washed, and the signal was developed as described by using a streptavidin-europium-linked fluorescence assay (12, 13). Each microplate collection of compounds was assayed in triplicate, and results reported as percentage inhibition compared to control wells lacking test compounds.

Solution-Phase Aggregation Assays. Aβ(1-40) peptides were disaggregated just before use, as described (12, 14). Peptides (50 μM in fibril formation reactions, ≈30 μM in CLC aggregation reactions) were incubated in PBSA at 37°C. Reaction progress was followed by ThT fluorescence and/or quantitative HPLC on centrifugation (30 min, 315,000 × g) supernatants of reaction aliquots (12, 15). The C_r value is the concentration of Aβ in solution when this value stops changing with time. For amyloid fibril growth, this value represents a position of dynamic equilibrium and is the reciprocal of the growth equilibrium constant for the amyloid fibrils (15); for the CLC-induced aggregation, the value is a more qualitative measure of the stability of the Aβ-CLC aggregates. Samples of aggregates were stored at -80°C after snap-freezing of washed pellets. Reactions with CLC contained 10% DMSO from the CLC stock solution.

HX by MS. The set-up and protocols used for HX-MS experiments were as described (8, 16, 17). All aggregates and fibrils were collected by centrifugation, washed, and resuspended in either protonated or deuterated 2 mM Tris buffer. The sample solution and sample processing solvent containing water, acetonitrile, and formic acid were mixed on-line in a T tube before electrospray ionization. All of the HX data presented here were collected by using a Quattro II spectrometer (Micromass, Manchester, U.K.), and corrected for artifactual exchange of rapidly exchanging side chain and terminal protons. Data for CLC-Aβ aggregates were also corrected for artifactual forward and backward exchange of backbone amide protons, as described (17), by using exchange data from fully deuterated CLC-Aβ(1-40) aggregates.

Ab-Binding Assays. Aβ fibrils and CLC-Aβ aggregates were immobilized onto 96-well high-binding microtiter plate (Costar) by adding 50 μl of a 2 ng/μl solution of the fibrils/protofibrils in PBSA to each well and allowing the plate to dry by overnight incubation unsealed at 37°C. The plates were then washed twice with PBSA containing 0.05% Tween 20, blocked with 1% gelatin in PBSA, and used to determine binding to aggregates of the Ab WO1 or control mouse myeloma IgM (Calbiochem) as described (11).

Results

We screened a small library of 640 compounds, the LOPAC library (Sigma), by using a microplate-based Aβ elongation assay. In this assay, plastic wells containing immobilized fibrils are incubated with a biotin-tagged monomer in the presence or absence of potential modifiers (13, 18). This assay format has been successfully implemented to identify inhibitors of polyglutamine aggregate elongation from the LOPAC library (V.B. and R.W., unpublished data). When formatted for Aβ(1-40) elongation, however, we observed no significant inhibitors, with only a few LOPAC compounds at 100 μM exhibiting inhibition levels barely exceeding the 50% mark (Fig. 1). At the same time, however, we found a number of compounds with potent abilities to enhance deposition of the biotin-tagged Aβ(1-40) (Fig. 1). The most potent of these compounds is CLC (Fig. 1 Inset), which at 100 μM stimulates the deposition of approximately six times as much Aβ compared to the elongation reaction in the absence of added compound. Interestingly, there was no effect of CLC in a similar screen for polyglutamine aggregation modulators (V.B. and R.W., unpublished data).

We characterized the activity of CLC in the microplate assay in more detail. The time course of Aβ deposition in the presence of CLC yields a pattern similar to the reaction without added compound (Fig. 2a), suggesting a simple acceleration of the normal mechanism. However, when the assay is repeated on a microplate lacking Aβ fibrils, the dramatic deposition of biotinylated Aβ in the presence of CLC was maintained (Fig. 2a), suggesting a direct ability of CLC to interact with soluble Aβ and thereby produce rapid deposition. Dose-response analysis shows that this stimulation of the deposition of 10 nM Aβ occurs at concentrations of CLC in the 10- to 100-μM range (Fig. 2b).

To confirm the microtiter plate assay, we carried out solution-phase experiments by using unlabeled Aβ(1-40). Aβ incubated alone exhibits a thioflavin T (ThT) lag phase of several days before progressing to a rapid amyloid fibril growth phase (Fig. 3a). During the ThT lag phase, the amount of soluble, monomeric Aβ is reduced by approximately a factor of two, consistent with the formation of protofibrillar structures with low ThT response (2). In contrast, in the presence of 100 μM CLC, Aβ(1-40) is rapidly converted to a pelletable aggregate, so that within 1-2 h only 2-3 μM Aβ remains in solution after centrifugation (Fig. 3 a and b); when these aggregates are isolated by sedimentation, they are found to contain a ratio of 4 mol of CLC/mole of Aβ(1-40) (results of independent analyses against both HPLC and MS standard curves). The significance of this ratio is not clear. Interestingly, this deposition is not associated with any apparent ThT signal (Fig. 3a); the low ThT signal obtained at reaction initiation persists for days, then gradually increases over a period of 14 d to reach a level ≈20% of that observed for the same weight of mature amyloid fibrils (Fig. 3a).

EM suggests that the aggregates induced by CLC more resemble protofibrils than mature fibrils. Thus, in an incubation of Aβ(1-40) without CLC, aggregates collected at day 2 (within the ThT lag phase) are curvilinear, rough-edged protofibrils (Fig. 4a), as reported (1, 2, 7, 19-21). Aggregates collected from the same reaction at day 14 are typical amyloid fibrils (Fig. 4b). Aggregates isolated from the 2-d reaction in the presence of CLC exhibit large spheroids (not shown) that, when rigorously vortexed, yield EM images showing a dense lawn of structures (Fig. 4c) resembling normal protofibrils (Fig. 4a). Aggregates isolated from a 14-d reaction contain similar spheroidal clusters of protofibrils, accompanied by long fibril-like ribbons (Fig. 4d). This apparent partial conversion of protofibrils to fibrils after a 14-d incubation is consistent with the increased ThT signal observed for this reaction (Fig. 3a). Although these fibrillar structures are associated with the globular protofibril clusters on the EM grids, it is not possible to say whether the fibrils grow out of the protofibril clusters or associate with them postsynthetically.

Biochemical analysis of the CLC products is consistent with the structures observed in the EM. In one analysis, aggregates were collected at days 2 and 14 by centrifugation from Aβ(1-40) incubations done in the absence or presence of CLC. The concentrations of the aggregates were calibrated by HPLC analysis of dissolved aliquots, and equal weights of each immobilized onto microtiter plate wells. Neither the oligomers grown from Aβ alone (Fig. 4a), nor the CLC-Aβ aggregates collected after 2 d (Fig. 4c), were effective seeds for elongation by monomeric Aβ(1-40) (Fig. 2c). This inability to seed elongation is consistent with the known instability of Aβ protofibrils, which tend to dissociate when removed from high concentrations of Aβ (7). In contrast, mature Aβ amyloid fibrils (Fig. 4b) exhibit high elongation seeding ability, whereas the 14-d CLC-Aβ aggregates (Fig. 4d) give relatively low-seeding activity consistent with EM and ThT evidence of their being a mixture of fibrils and protofibrils.

The CLC-Aβ aggregates also are very similar to Aβ protofibrils in their HX behavior. Previously we showed that approximately one-half of the 39 backbone amide hydrogens of Aβ(1-40) are very stably protected in HX of mature amyloid fibrils (16, 17), whereas substantially fewer of these amide hydrogens are protected in Aβ(1-40) protofibrils (8). Fig. 5a shows that the degree of protection in the CLC-Aβ aggregates is exactly the same as that seen previously for authentic Aβ protofibrils isolated from normal fibril formation reactions (8). The coincidence between the kinetics and extents of protection provides further evidence that CLC-Aβ aggregates are very similar to the protofibrils normally observed during Aβ fibril assembly.

In our previous work on HX of protofibrils, it was only possible to correct the HX data for rapidly exchanging side chain protons (8). To obtain an absolute number for protection in protofibrils, it is also necessary to correct the data for artifactual forward and backward exchange into the backbone amide protons (17). We prepared CLC aggregates from fully deuterated Aβ(1-40) and determined their back exchange rate which, along with forward exchange data from protonated aggregates, was used to correct the experimental data (17). The fully corrected kinetic points are plotted in Fig. 5b, which also shows the fitted data for fully corrected amyloid fibril HX as well as the total number of exchangeable backbone amide protons. The data indicate that the CLC stabilized Aβ(1-40) protofibrils contain only ≈12 highly protected backbone amide protons after 2 d of exchange time, in contrast to ≈22 for Aβ(1-40) fibrils.

That these unique aggregates also share some structural features with mature fibrils is suggested by the ability of the CLC-Aβ aggregates to bind, with ≈10-fold reduced affinity, to the amyloid-specific Ab WO1 (11), as shown in Fig. 3c. This Ab has been shown to recognize amyloid fibrils from many different proteins, while not binding to the monomeric precursor proteins. It has not previously been studied with protofibrils, partially because of difficulties maintaining protofibril integrity during binding assays. The results suggest that the amyloid epitope detected by the Ab is already present in the CLC-Aβ aggregate and hence in protofibrils.

To further probe the relationship between the conformation of Aβ in fibrils and protofibrils, we turned to scanning proline mutagenesis analysis, which we used previously to provide insights into the Aβ conformation in mature amyloid fibrils (15). Proline point mutants of Aβ(1-40) were treated with CLC and the aggregation reactions followed until the amount of Aβ in solution after centrifugation was unchanged (Fig. 3b). These values, which represent the degree to which proline at each sequence position destabilizes the aggregate (15), are plotted in the bar graph in Fig. 6a. The corresponding values for amyloid fibril formation by the same Aβ(1-40) mutants, most of which were reported (15), are shown in Fig. 6b.

The comparison shows both similarities and differences in the sensitivities of the two aggregates to proline substitution. In both aggregates, replacement of residues in the N-terminal region is negligibly destabilizing, whereas proline substitution at many other positions in the 15-40 segment, especially the hydrophobic clusters at positions 16-21 and 31-36, are destabilizing. Illustrative of this, a double Pro mutant at positions 19 and 32 produces the most destabilization of any mutant for both mature fibrils and CLC aggregates. Fig. 6 also shows some significant differences between fibrils and CLC protofibrils. In a striking example, a tetra-substituted mutant peptide (designated P4 in the figure), with prolines at positions 9, 23, 30, and 37, makes amyloid fibrils of the same stability as WT Aβ(1-40), but produces protofibrils that are significantly destabilized compared to WT. The source of this destabilization is apparent in closer examination of the single point mutant data, which shows that prolines at positions 30 and 37 significantly destabilize CLC aggregates but not fibrils. Conversely, Pro replacement at residues 24-28 more strongly affects amyloid fibril formation than CLC aggregate formation. The implications of these results will be treated in Discussion.

It should be pointed out that, although WT Aβ(1-40) exhibits very similar C_r values for both fibril (0.8 μM) and protofibril (1.6 μM) formation, the stability of CLC protofibrils is augmented by the binding energy of the CLC component. We assume that it is for this reason that the destabilizing effects of proline replacement are significantly dampened in the CLC-Aβ aggregates compared to amyloid fibrils (Fig. 6b). The overall resemblance of the profiles for fibrils and protofibrils in Fig. 6 strongly suggests that the overriding structural bases for proline destabilization are similar for the two aggregated states of Aβ.

Discussion

The results described here show that the aggregated structures formed from Aβ(1-40) in the presence of CLC resemble the protofibrillar structures normally observed during the formation of mature amyloid fibrils from this peptide. The results also suggest that the CLC-Aβ aggregate and, hence, Aβ protofibrils are similar to mature fibrils in some structural aspects and different in others.

Previous studies demonstrated the ability of HX to reproducibly distinguish among a variety of Aβ assemblies (8). The close match in both exchange kinetics and final amplitude between protofibrils and CLC-Aβ aggregates strongly suggests a structural relationship. Segmental HX protection factors, determined by using pepsin digestion in-line with MS (I.K., M.C., K.D.C., and R.W., unpublished data), may eventually provide more detailed comparisons of protofibrils and CLC-Aβ aggregates. Until such data are available, it must be kept in mind that the two nonfibrillar aggregates compared here may differ in which residue amide protons are involved in H-bonded structure, even though net H-bonded structure appears to be identical.

CLC aggregates exhibit fine structure highly reminiscent of a particular type of protofibril often observed by EM (2, 19, 20) and atomic force microscopy (1, 7, 21). Normal Aβ protofibrils incubated in the absence of excess Aβ monomer tend to dissociate (7); whereas CLC stabilizes the Aβ aggregates, minimizing dissociation, these aggregates are incapable of elongating via monomer addition (Fig. 2c), supporting a resemblance to normal protofibrils. Another resemblance between the CLC-Aβ aggregates and normal protofibrils is their low ThT response. Fig. 3a shows that the pelletable Aβ that accumulates both in the early stages of spontaneous amyloid growth in the absence of CLC, and the CLC-Aβ aggregates, give similar, low ThT responses. These observations are in qualitative agreement with previous ThT measurements of Aβ protofibrils isolated in the lag phase of normal Aβ amyloid growth (2).

Although clearly distinct structures, Aβ mature fibrils exhibit a number of similarities to protofibrils and CLC-Aβ aggregates, including a fundamental fibrillar character in the EM, a shared epitope for the Ab WO1, and common possession of a subset of unusually stable H-bonds. Both mature amyloid fibrils and CLC protofibrils appear to be assembled from very similar conformations of the peptide, as revealed by scanning proline mutagenesis, which can reveal sites of restrictive peptide backbone geometry within fibrils (15, 22).

We found a substantial overlap in the destabilization profiles for the amyloid fibrils and CLC/protofibril forms in the scanning proline mutagenesis analysis (Fig. 6). This suggests that the structure providing HX protection to protofibrils lies in H-bonded β-sheet involving the same hydrophobic patches 16-21 and 31-36 involved in fibril structure. Significantly, in both fibrils and protofibrils, the N-terminal 14 residues and the C-terminal 3-4 residues seem uninvolved in the kind of rigid structure disrupted by proline substitution (Fig. 6). The involvement of residues in the 21-30 segment in rigid structure, however, differs between these two aggregated states. The data suggests that the CLC protofibril may house Aβ in more of a simple extended hairpin structure with a long, relatively flexible loop, involving residues 22-29, spanning the two β-structure segments. The major structural difference in the Aβ sequence when it adopts amyloid fibril structure may thus be a better definition of structure in the 22-29 segment. This is supported by pepsin-HX analysis of the CLC-Aβ protofibrils (I.K., M.C., K.D.C., and R.W., unpublished data).

Although these experiments give us a much clearer picture of protofibril structure, they cannot address the question of the relevance of protofibrils in the amyloid fibril assembly mechanism. The fact that CLC-Aβ aggregates partially transform into an amyloid-like structure on prolonged incubation, coupled with the physical proximity of spheroidal CLC-Aβ aggregates in EM grids (Fig. 4d), is consistent with a precursor-product relationship between protofibrils and fibrils. Details of HX studies on isolated protofibrils previously led to a similar suggestion (8). However, it remains possible that incubated CLC-Aβ aggregates could produce fibrils by dissociation and reassembly of monomeric Aβ, whereas the physical proximity on grids, as shown in Fig. 4d, could be an artifact of aggregate stickiness. It is also possible that the fundamental characteristics of the Aβ sequence, which displays β-extended chain structure propensity in regions 17-21 and 31-36 even as a soluble monomer (23), might influence the structures of both aggregates independently. Recent data, showing that under some conditions Sup35 fibril formation involves a small critical nucleus and growth by monomer addition (24), suggests that more work needs to be done to establish the role of protofibrils in amyloid assembly.

The ability of CLC to accelerate protofibril formation and stabilize protofibril structure was discovered during a random screen for modulators of Aβ fibril elongation. Other random screening efforts have unveiled compounds that stimulate formation of alternatively aggregated forms of Aβ and hence retard fibril formation (25-27). Although there are no strong structural similarities, these compounds share with CLC the combination of a basic nitrogen moiety with substituted aromatic or heteroaromatic groups. It is possible that aromatic interactions between these compounds and Phe residues 19 and 20 may contribute to the ability of compounds like CLC to stabilize Aβ protofibril structure. It has been speculated that phenylalanine interactions may promote amyloid formation in at least some amyloidogenic sequences (28, 29). In this regard, it may be relevant that the stabilization of protofibril structure is also a property of the “Arctic” mutant of Aβ (30), which involves a Glu → Gly mutation at position 22 of Aβ, directly adjacent to the Phe-Phe pair at positions 19 and 20. Interestingly, treatment of Aβ peptides with organic solvents can also induce formation of nonfibrillar aggregates (31). The relationship between any of these alternatively aggregated structures and the CLC-Aβ aggregates described here is not clear.

As shown here, compounds like CLC can prove useful in facilitating detailed studies of protofibril structure. Although counterintuitive, it remains conceivable that compounds that stabilize protofibril structure may also have therapeutic value. Although recent evidence suggests that oligomers and protofibrils may be the toxic entities in at least some amyloid diseases (4, 5), it is possible that a drug that facilitates sequestration of monomeric Aβ into stabilized protofibrils may provide therapeutic advantage, protecting cells while giving the body time to remove or inactivate these peptides.

Note Added in Proof. Results similar to those reported here for the 1-40 version of Aβ have now also been obtained by using the 1-42 version of the peptide, which is considered to be the more toxic form.