pH-induced molecular shedding drives the formation of amyloid fibril-derived oligomers

Significance

Oligomers formed en route to amyloid fibrils are thought to be the perpetrators of toxicity in many amyloid disorders. How amyloid fibrils contribute to disease, however, is less clear. Here, using β₂-micoglobulin (β₂m) as a model system, we show that the stability of amyloid fibrils is highly pH-dependent, with mild acidification enhancing the formation of fibril-derived nonnative oligomers that disrupt membranes and alter cellular function. Enhancing fibril stability by incubation with the molecular chaperone, hsp70, or by cross-linking, protects against fibril-induced membrane disruption and cellular dysfunction. The results highlight the importance of pH in determining fibril stability and suggest that uptake of fibrils into acidic cellular compartments may contribute to amyloid disease by pH-induced molecular shedding of toxic species.

Abstract

Amyloid disorders cause debilitating illnesses through the formation of toxic protein aggregates. The mechanisms of amyloid toxicity and the nature of species responsible for mediating cellular dysfunction remain unclear. Here, using β₂-microglobulin (β₂m) as a model system, we show that the disruption of membranes by amyloid fibrils is caused by the molecular shedding of membrane-active oligomers in a process that is dependent on pH. Using thioflavin T (ThT) fluorescence, NMR, EM and fluorescence correlation spectroscopy (FCS), we show that fibril disassembly at pH 6.4 results in the formation of nonnative spherical oligomers that disrupt synthetic membranes. By contrast, fibril dissociation at pH 7.4 results in the formation of nontoxic, native monomers. Chemical cross-linking or interaction with hsp70 increases the kinetic stability of fibrils and decreases their capacity to cause membrane disruption and cellular dysfunction. The results demonstrate how pH can modulate the deleterious effects of preformed amyloid aggregates and suggest why endocytic trafficking through acidic compartments may be a key factor in amyloid disease.

Amyloid assembly involves the formation of an ensemble of aggregate species that lead to the onset of pathological cascades and disease (1). The formation of amyloid is initiated by self-association of partially unfolded or intrinsically disordered precursors, which stimulates the production of an array of oligomeric assembly intermediates. These species, which precede the formation of mature amyloid fibrils, are commonly described as the agents of toxicity in amyloid disease (1). The ephemeral nature of oligomeric species, combined with their heterogeneity and copopulation with other species, provides immense challenges for characterizing amyloid assembly intermediates in structural detail and determining their role(s) in fibril assembly (1). It also highlights the difficulty in identifying which species among the aggregate ensemble contributes to cellular dysfunction and the onset of amyloid disease.

Although oligomers formed from many different proteins have been shown to be toxic (2), the role that fibrils play in amyloid disease is not well understood. Mounting evidence suggests that fibrils can be toxic (3, 4), or may contribute to disease by acting as reservoirs of toxic oligomers (5). Fibrils may also contribute to disease by catalyzing the formation of oligomers on their surface via secondary nucleation processes (6), or by disrupting protein homeostasis by consuming the capacity of the proteostatic network through the sequestration of molecular chaperones (7). Reduced fibril length can also contribute to disease by increasing the uptake of fibrils into intracellular compartments (8–10), or by enhancing transmission of aggregates between cells in prion-like processes (11). In addition, fibrils have been shown to interact with lipid bilayers, causing membrane deformations via interactions with fibril ends (9, 12, 13), or promoting the release of toxic oligomers by membrane-induced fibril disassembly (14). Together these data support a role of amyloid fibrils as direct or indirect agents of amyloid disease.

Because amyloid formation is under kinetic control, minor changes in the population of an amyloid precursor can lead to dramatic changes in the rate of fibril formation and in the repertoire of nonnative species formed (15, 16). Conversely, the kinetic stability of amyloid fibrils can be enhanced by alterations in the protein sequence (17), or by adding small molecules (18, 19) or molecular chaperones (20). Such enhancements in fibril stability can protect against toxicity, presumably by supressing the formation of oligomers. In addition, fibrils may encounter environments in vivo that are distinct from those which promote their formation; yet how localized cellular environments affect the dynamic equilibrium of amyloid fibrils remains unclear.

Here, using amyloid fibrils formed in vitro from β₂-microglobulin (β₂m) as a model system, we explore the effect of altering pH on fibril stability and in determining the nature of the soluble products that result from fibril disassembly. Using biochemical, biophysical and cellular assays, we show that β₂m fibril disassembly is highly sensitive to changes in pH, with a reduction in pH from 7.4 to 6.4 (similar to the difference in pHs encountered early during endocytic trafficking) resulting in dramatically different outcomes of disassembly. The results reveal how subtle changes in pH can alter the delicate kinetic balance of intermediates in fibrillation reactions and suggest that increasing fibril stability, and/or decreasing fibril uptake into acidic cellular compartments, may be a viable strategy to alleviate the progression of amyloid disease.

Results

pH-Dependent Fibril Dynamics.

To investigate how pH alters the stability and dynamic behavior of amyloid fibrils, β₂m monomers were assembled into amyloid fibrils by incubation at acidic pH (Methods). These fibrils have been analyzed structurally in detail and have been shown to possess the canonical parallel in-register β-sheet amyloid structure formed by many proteins and peptides, including β₂m fibrils isolated ex vivo (21, 22). In addition, β₂m fibrils formed under acidic conditions are structurally similar to those formed in vitro at near-neutral pH from the N-terminally truncated β₂m variant, ΔN6 (21). They also display similar effects (internalization into the endo-lysosomal pathway and the inhibition of cellular reduction of MTT) when incubated with cell lines and with primary cells relevant to the pathology of dialysis-related amyloidosis (DRA) (monocytes, chondrocytes, and osteoblasts) (9, 10, 23, 24). Moreover, fibrils formed from ΔN6, but seeded with wild-type β₂m fibrils formed at low pH, have been shown to inhibit osteoclast formation from primary human monocytes, rationalizing the osteoarticular pathology associated with DRA (23).

The kinetic stability of the β₂m fibrils was measured by diluting fibrils formed at acidic pH into buffer at pH 7.4 or 6.4 (at an identical ionic strength of 164 mM; Methods). pH values of 7.4 and 6.4 were chosen to mimic the environments encountered by fibrils in the extracellular milieu and in early endosomes, respectively. In previous experiments we have shown that β₂m fibrils are taken up into endosomes and lysosomes in a length-dependent manner, wherein they persist as protease-resistant material that retains a cross-β structure and results in the perturbation of lysosome function (10, 23, 24). pH-induced changes in fibril yield and conformation were monitored by ThT fluorescence and SDS-PAGE (Methods). Dilution into each buffer resulted in an initial increase in ThT fluorescence, presumably due to alterations in ThT binding sites in the fibril-bound state, followed by a slow phase in which the fluorescence intensity decreases (Fig. 1 A and B). Fitting to a double exponential revealed that the rate constants for the enhancement and decay of fluorescence are similar at both pH values (SI Appendix, Fig. S1 A–D). Most strikingly, however, the amplitude of the second phase of the reaction is larger at pH 7.4 than 6.4 (corresponding to 94% and 55% at each pH, respectively). At both pH values, SDS-PAGE analysis of soluble material revealed that the rate of loss of ThT fluorescence mirrors the rate at which soluble material is formed (Fig. 1 A and B). Analysis of the fibril length distributions using AFM 30 and 60 min after sample dilution at pH 7.4 or 6.4, respectively (at which times ∼25% soluble material is formed; Fig. 1 A and B), showed a reduction in fibril length upon incubation at both pH values (SI Appendix, Fig. S1 E and F). Similar effects were not observed upon diluting fibrils into growth buffer at pH 2.0 (SI Appendix, Fig. S1G). Moreover, mutation of each of the three His residues known to reside within the core of β₂m fibrils (21) significantly decreased the pH dependency of the loss of ThT fluorescence (SI Appendix, Fig. S2), suggesting that the protonation state of these His residues plays a role in determining the extent of pH-dependent fibril disassembly leading to the formation of soluble material.

Fig. 1.

pH dependence of fibril disassembly. β₂m amyloid fibrils formed at pH 2.0 were diluted into pH 7.4 (A) or pH 6.4 (B) buffer and disassembly monitored by ThT fluorescence (solid lines show a fit to a double exponential). Soluble material released as fibril depolymerization proceeds was isolated using a filter-trap assay, resolved by SDS-PAGE (Inset), and quantified by densitometry of the gel (dashed line).

The Mechanism of Fibril Depolymerization Is pH-Dependent.

To determine the nature of soluble species released during fibril disassembly, fibril depolymerization was monitored at pH 6.4 and 7.4 using ¹H-¹⁵N NMR and the appearance of resonances was monitored in real time (Fig. 2 A and B and SI Appendix, Fig. S3). Most resonances that were detected have a chemical shift identical to those of the native protein, enabling the rate of production of native species to be determined at each pH (Fig. 2 C and D). The results revealed that at each pH native amide resonances increase in intensity at a similar rate across the sequence of β₂m (Fig. 2 C and D). However, the average rate of appearance of native resonances is dependent on the pH. The average per-residue time constant (τ) for the appearance of native β₂m resonances at pH 7.4 (105 ± 7 min) mirrors τ for the decay in ThT fluorescence (95 ± 5 min) (Fig. 2C), indicating that fibril disassembly leads to the formation of native β₂m monomers. At pH 6.4, by contrast, although the decrease in ThT fluorescence (99 ± 6 min) occurs at a similar rate to the appearance of soluble material (Fig. 1B), the appearance of native β₂m resonances measured using NMR is delayed significantly (171 ± 10 min; Fig. 2D). This observation suggests that fibril disassembly at pH 6.4 results in the formation of NMR-invisible species that precede the development of native and/or unfolded monomers. Indeed, weak resonances with chemical shifts distinct from those of the native monomer are detected early during fibril depolymerization at pH 6.4, but not at pH 7.4 (SI Appendix, Fig. S4), suggesting that the “NMR invisible” species are nonnative, but not unfolded. The absence of a detectable lag in the appearance of the native resonances at pH 6.4 suggests that the NMR invisible nonnative species could be off-pathway. Alternatively, an on-pathway model in which the exchange rate for oligomer formation does not exceed significantly the rate of monomer formation would also fit the data. More experimental data will be needed to resolve this issue. However, irrespective of the kinetic model for their formation, the results show that reducing the pH by a single unit alters the kinetics and/or kinetic mechanism of fibril disassembly such that nonnative species accumulate only at the lower pH.

Fig. 2.

The formation of native β₂m monomers during fibril depolymerization monitored by real-time NMR. (A) Select regions of ¹H-¹⁵N HSQC spectra collected at different times during depolymerization of ¹⁵N-labeled β₂m fibrils at pH 7.4 (purple) or pH 6.4 (green). Residue assignments are shown on the bottom panels. (B) The peak intensity of residue 19K versus time during fibril depolymerization at pH 7.4 (purple) or pH 6.4 (green). The data are fitted to single exponential functions (solid lines). Colored bars highlight the time points at which the spectra shown in A were acquired (framed in the same color). (C and D) Time constants (τ) of appearance of native resonances during fibril depolymerization at pH 7.4 (C) or pH 6.4 (D). Black dashed lines indicate the average under each condition. Red dashed lines depict τ of ThT fluorescence decay calculated in Fig. 1 A and B.

Oligomeric Species Accumulate During Fibril Disassembly at pH 6.4.

To characterize and compare the soluble species that form during fibril disassembly in more detail fluorescence correlation spectroscopy (FCS) was used. FCS exploits fluctuations in fluorescent signals as molecules diffuse through a confocal volume of known size (see SI Appendix, SI Methods for more detail). Given its ability to detect species that differ in size over several orders of magnitude, in complex ensembles and without purification, FCS is an ideal technique to identify and track the ensemble of species that form during amyloid assembly/disassembly over time.

Fibril depolymerization was measured using FCS by labeling the β₂m variant R3C with maleimide-Alexa488 (R3C488) (Methods). Residue 3 was chosen as the site to incorporate a fluorescent label as the N-terminal six residues of β₂m do not form part of the fibril core (21). R3C488 was incorporated into fibrils by seeded elongation in the presence of a 1,000-fold molar excess of wild-type β₂m monomers. Fibril depolymerization was then initiated by diluting these samples into pH 7.4 or 6.4 buffer. Autocorrelation curves (ACs) were acquired immediately and over the course of depolymerization (Methods and SI Appendix, Figs. S5 and S6). The curves were fitted using a maximum entropy method (MEM) (25) to extract the apparent hydrodynamic radius (R_H) of soluble species released from fibrils during depolymerization (see SI Appendix for details).

The FCS analysis revealed that soluble species released over the course of fibril depolymerization at pH 7.4 are dominated by a low molecular weight species with an apparent R_H of 1.9 nm (Fig. 3A, Left), equivalent to the apparent R_H determined for native β₂m monomers (SI Appendix, Fig. S7). By contrast, at pH 6.4, a second species with an apparent R_H of 20 nm is highly populated and persists up to 135 min after the initiation of disassembly (Fig. 3A, Right). Fitting the ACs to simpler one or two component functions confirmed that this larger species is only populated significantly during fibril disassembly at pH 6.4 (SI Appendix, Fig. S8 and Tables S1 and S2). Analysis of samples 30 min (pH 7.4) or 60 min (pH 6.4) after initiating depolymerization using negative stain TEM (Fig. 3 B and C) revealed the presence of spherical oligomers at pH 6.4, but not at pH 7.4. These species range in diameter from 30 to 60 nm, measured using cryo-EM (Fig. 3 D and E and SI Appendix, Fig. S9). Previous studies have shown that oligomers of this size do not accumulate during fibril formation of ΔN6 at pH 6.2 or from wild-type β₂m at pH 2.0, and thus their accumulation is a unique consequence of fibril disassembly at pH 6.4 (16, 26). Further structural characterization revealed that these species are not recognized by the antibody WO1 (which recognizes the cross-β fold of amyloid; ref. 27) (Fig. 4A and SI Appendix) and give rise to a far UV CD spectrum consistent with significant nonnative (more highly unfolded) structure (Fig. 4C). By contrast, CD analysis of the soluble species released at pH 7.4 is consistent with the formation of predominantly native species (Fig. 4B). The oligomers formed at pH 6.4 also expose greater hydrophobic surface area than those at pH 7.4, measured by the fluorescence of the hydrophobic dye, 8-anilinonaphthalene-1-sulphonate (ANS) (Fig. 4 D and E). Together, the results indicate that pH alters the energy landscape of fibril depolymerization such that nonnative, spherical oligomers only become highly, and persistently, populated at pH 6.4.

Fig. 3.

pH-dependent formation of oligomers during fibril disassembly. (A) Apparent R_H distributions of soluble species extracted from autocorrelation curves collected by FCS and analyzed by MEM at different times during β₂m fibril disassembly at pH 7.4 or pH 6.4. The apparent R_H values shown are the peak center of each ensemble. Negative stain TEM image of fibrils 30 min or 60 min after the initiation of fibril depolymerization at pH 7.4 (B) or pH 6.4 (C). (Scale bar, 150 nm). (D) Cryo-EM class averages of oligomers released from fibrils at pH 6.4, 5 min after initiating disassembly. Each box is 60 × 60 nm (see also SI Appendix, Fig. S9). (E) Histogram showing size distribution of the particles used to compute the class averages in D.

Fig. 4.

Structural analysis of soluble material formed during fibril disassembly. (A) Dot blots of insoluble and soluble material (filtrate) generated 30 min or 60 min after the initiation of fibril disassembly at pH 7.4 or pH 6.4, respectively. Samples were immunoblotted with anti-fibril (W01; ref. 27) or anti-β₂m antibodies (SI Appendix). Control blots are of β₂m fibrils at pH 2.0. Mean residue ellipticity of the filtrate isolated as stated above at pH 7.4 (B) or pH 6.4 (C) measured using far-UV CD. Spectra of native β₂m (blue) were acquired at the same protein concentration (SI Appendix). Difference spectra are shown in dashed lines. ANS fluorescence emission spectra of the filtrate isolated as stated above at pH 7.4 (D) or pH 6.4 (E) and of native β₂m at equivalent concentrations. Curves are normalized to the maximum amplitude of the spectrum of acid unfolded β₂m at pH 2.0 (gray dashed line) acquired at the same protein concentration. In C–E, native β₂m monomer controls are shown in blue.

Fibril-Released Oligomers Cause Membrane Disruption.

The ability of oligomers produced during fibril disassembly at pH 6.4 to disrupt membranes was next examined by an in vitro dye release assay using large unilamellar vesicles (LUVs) containing 12 mol% bismonoacylphosphate (BMP). BMP is enriched in endosomal/lysosomal membranes and has been shown to sensitize LUVs to disruption by β₂m amyloid fibrils (12). The disruption of BMP-containing membranes by β₂m fibrils is also pH dependent, with a large increase in the extent of membrane disruption observed when the pH is decreased from 7.4 to 6.4 (12). To determine the extent of membrane disruption caused by soluble species released during fibril depolymerization, β₂m fibrils were allowed to depolymerize for 60 min at pH 6.4. Filtration was then used to separate soluble species from fibrillar material, so that the effect of each fraction on membrane integrity could be assessed individually. These experiments showed that the soluble material released from fibrils accounts for ∼70% of the membrane disruption activity of the total (fibril + soluble) sample (Fig. 5A). Determining the dye release efficiency per µM protein showed that the soluble material is ∼5 times more membrane active than the same monomer-equivalent concentration of β₂m fibrils (Fig. 5B). As native β₂m monomers cause minimal dye release (Fig. 5 A and B), nonnative oligomers that accumulate during fibril disassembly at pH 6.4 must be responsible for the majority of fibril-mediated membrane disruption.

Fig. 5.

Fibril-derived oligomers cause membrane disruption and cellular dysfunction. (A) Membrane disruption of total protein (fibrils + soluble material) versus soluble material (filtrate) isolated 60 min after the initiation of fibril disassembly at pH 6.4 measured using a carboxyfluorescein dye release assay (SI Appendix). The dye release of the equivalent concentration (to total protein sample) of native monomers was also measured (native β₂m). (B) Normalized dye release efficiency per µM of protein (determined using SDS-PAGE densitometry). The fibril concentration was calculated as the amount of nonfiltrated material present 60 min after the onset of fibril disassembly. (C) Kinetic stability of β₂m fibrils (12 µM monomer equivalent concentration) at pH 6.4 measured using ThT fluorescence in the presence or absence (black dashed) of 0.25 molar (orange) or 3 molar (green) equivalents of hsp70. A control in the presence of a threefold molar excess of BSA (BSA) is also shown (gray). (D) Membrane disruption by β₂m fibrils 30 min after the initiation of fibril disassembly as measured by dye release in the presence of 0.25 or 3 molar equivalents of hsp70. (E) The reduction of MTT in the presence of 1.2 µM β₂m amyloid fibrils with or without the addition of 0.25 or 3 molar equivalents of hsp70. MTT reduction is expressed as a percentage in comparison with cells incubated with hsp70 in the absence of fibrils. (F) SH-SY5Y cells were incubated with buffer, 1.2 μM (monomer equivalent concentration) native β₂m monomers or fibril samples after preincubation in the presence (light gray) or absence (dark gray) of a threefold molar excess of hsp70. The lysosomal degradation of endocytosed Alexa647-labeled ovalbumin was quantified by flow cytometry (10). *P < 0.05.

Restricting Molecular Shedding Decreases Membrane Damage and Fibril-Mediated Cellular Dysfunction.

To determine whether fibril-induced membrane damage can be inhibited by enhancing the kinetic stability of β₂m amyloid fibrils, the molecular chaperone hsp70 was added. Previous studies have shown that the association of hsp70 (or its constitutively expressed homolog, hsc70) with amyloid aggregates protects against amyloid toxicity (28–30). Recombinant hsp70-1A (29) was incubated with β₂m amyloid fibrils at pH 6.4 in a 1:0.25 or 1:3 molar ratio (monomer equivalent β₂m:chaperone) and the effect of the chaperone on fibril stability was measured using ThT fluorescence. These experiments showed that hsp70 enhances the kinetic stability of β₂m fibrils such that no depolymerization is observed up to 3 h after dilution to pH 6.4 (Fig. 5C). To determine whether the addition of hsp70 also affects the ability of β₂m fibrils to mediate membrane disruption, LUVs containing 12 mol% BMP were incubated with fibrils in the presence of hsp70 (Fig. 5D). These experiments showed that membrane disruption is decreased significantly in the presence of both 0.25 and 3 molar equivalents of hsp70, in comparison with controls lacking hsp70. The residual membrane disruption activity is presumably attributed to the activity of the fibrils themselves (Fig. 5 A and B; ref. 13). Chemically cross-linking fibrils to enhance their stability also resulted in a significant decrease in membrane disruption (SI Appendix, Fig. S10 A–C).

Finally, to determine whether hsp70 is able to protect cells against fibril-induced cellular dysfunction, SH-SY5Y cells were incubated with either 0.3 µM or 3 µM hsp70 for 6 h before adding 1.2 µM (monomer equivalent concentration) of β₂m fibrils. Previous studies have shown that β₂m fibrils inhibit MTT reduction without causing cell death (10). Instead, fibrils disrupt lysosomal function as evidenced by the inhibition of the degradation of endocytosed ovalbumin (10). Here, hsp70 was shown to rescue cells from fibril-mediated inhibition of MTT reduction when preincubated with 3 µM hsp70 before the addition of β₂m fibrils (Fig. 5E). Similarly, the reduction in the capacity of cells to degrade ovalbumin taken up by endocytosis in the presence of β₂m fibrils (10) was restored in hsp70-treated cells (Fig. 5F). Thus, an excess of hsp70 is able to protect against the deleterious effects of β₂m fibrils measured by these two assays. Decreasing the extent of fibril depolymerization by chemical cross-linking also partially rescued the inhibition of MTT reduction by fibrils (SI Appendix, Fig. S10D). Together, these results suggest that the pH-dependent shedding of oligomers from fibrils contributes to β₂m fibril-mediated membrane disruption in vitro and to cellular dysfunction, firmly ascribing the formation of these previously uncharacterized fibril-derived oligomers as facilitators of amyloid-induced cellular degeneration.

Discussion

Although amyloid fibrils have been shown to exhibit dynamic behavior in solution (5), the importance of such dynamic excursions in the context of biological activity has remained unclear. Here, by studying amyloid fibril dynamics in solution conditions distinct from those which promote aggregation and, most importantly, which mimic the cellular environment encountered early during endocytosis, we show how subtle differences in pH can dramatically influence the repertoire of species that form over the course of fibril disassembly in vitro. The importance of histidine residues in controlling the fate of β₂m fibril disassembly suggests that other amyloid assemblies may also display pH-dependent depolymerization pathways, especially for fibrils that are formed extracellularly and trafficked via the endocytic pathway. As shown here using β₂m fibrils formed in vitro as a model, a change in the pH by just one unit (pH 7.4 to 6.4) dramatically enhances the potential of fibrils to act as vehicles of cellular dysfunction by enhancing the accumulation of nonnative oligomers able to disrupt membranes in vitro and to contribute to cellular dysfunction.

The endo-lysosomal pathway is emerging as a common route by which extracellular and intracellular amyloid aggregates can elicit mechanisms of toxicity (31, 32). For β₂m fibrils, cellular dysfunction is associated with internalization via endocytosis, with fibril uptake being enhanced for fibrils of short average length (<400 nm), such as those produced early during fibril formation or via fibril fragmentation (9). Long-term residence of protease-resistant fibrils within lysosomes leads to perturbation of the ability of these organelles to degrade endocytosed proteins (10, 24). Moreover, recent investigations have shown that synthetic membranes containing BMP, a lipid enriched in endosomes and lysosomes (12), have enhanced susceptibility to β₂m fibril-mediated membrane disruption compared with synthetic membranes lacking BMP (12). Together, decreased kinetic stability of fibrils under mildly acidic conditions, combined with the specific lipid composition of endosomes/lysosomes, creates a cellular environment that is uniquely susceptible to β₂m amyloid fibril-mediated cellular damage.

The results presented here also demonstrate that increasing the stability of amyloid fibrils by chemical cross-linking or the addition of hsp70 leads to decreased membrane disruption and a reduction in lysosome dysfunction. Hsp70 has been shown to inhibit aggregation and toxicity of a variety of amyloidogenic systems via distinct mechanisms (28–30). Here we show that hsp70 is able to protect cells against the deleterious effects of β₂m fibril-mediated cell dysfunction to a greater extent than chemical cross-linking. This observation suggests that hsp70 inhibits fibril-mediated cellular dysfunction by preventing molecular shedding, and potentially through additional mechanisms. Nevertheless, the results suggest that the pH-induced formation of fibril-derived oligomers is a significant source of β₂m fibril-mediated cellular dysfunction. pH-induced molecular shedding may also contribute to fibril-mediated toxicity of other amyloid aggregates that undergo endocytic trafficking. For example, the stabilization of fibrils by small molecules has been shown to reduce Aβ toxicity by accelerating assembly toward β-rich fibrils (18). Other studies have shown that small molecules can alleviate Aβ fibril-mediated toxicity (19), presumably by stabilizing fibrils and preventing the shedding of toxic species. Further studies using other amyloid systems are now needed to determine in more detail the role of the endo-lysosomal pathway in modulating fibril/aggregate stability and hence in contributing to the protection or exacerbation of disease.

By contrast with protein folding reactions, which, in most cases, result in the formation of a unique structural entity (the native state), amyloid formation is a complex kinetic process that can give rise to fibrils and aggregates with distinct structural, kinetic and/or thermodynamic properties (8, 33). For example, structurally distinct fibrils of α-synuclein, and those of Aβ, display different biological activities (8, 33). Depending on their structure, some fibril strains will be less kinetically stable than others and, thereby, could potentially be more dangerous biologically than their similar, but more stable counterparts. In addition, as shown here for β₂m, subtle changes in solution conditions can change the repertoire of oligomers originating from a single fibril strain by altering the energy landscape of fibril disassembly. Rather than being the innocuous end products of assembly, fibrils may thus be a dangerous source of toxic oligomers uniquely formed by molecular shedding in a mechanism that is dependent on the pH. A better understanding of the link between fibril dynamics, pH and toxicity, such as presented here for β₂m fibrils, may provide new avenues for the development of therapeutic strategies against amyloid disease.

Materials and Methods

Protein Expression, Purification, and Labeling.

Wild-type β₂m, ¹⁵N-labeled β₂m and β₂m R3C were expressed and purified as described (9, 16). The R3C variant was covalently labeled with maleimide-Alexa488 (Invitrogen) according to manufacturer’s instructions. Labeled monomer was stored at −80 °C in 25 mM sodium phosphate buffer, pH 7.2. The human hsp70-1A gene was kindly provided by Chris Dobson (University of Cambridge) and was expressed and purified as described (29).

Fibril Formation and Kinetic Stability Assay.

Fibril samples were prepared by dissolving lyophilised β₂m monomers in 10 mM sodium phosphate buffer, pH 2.0 containing 50 mM NaCl. The sample was filtered (0.22 µM Millipore syringe filter) and the protein concentration adjusted to 120 µM (monomer equivalent concentration). Fibril formation was seeded using 0.1% (wt/wt) of preformed β₂m fibrils as described (9). Samples were then incubated quiescently at room temperature for 48 h, which resulted in fibrils of average length ∼1.2 µm measured using AFM, consistent with previous results (9, 12). These fibrils were used for all subsequent experiments.

To monitor the stability of β₂m fibrils upon changes in pH, β₂m fibrils (120 µM monomer equivalent concentration) were diluted fourfold into 50 mM sodium phosphate buffer, pH 7.4 containing 35 mM NaCl, or 50 mM sodium phosphate buffer, pH 6.4 containing 85 mM NaCl (ionic strength 164 mM), previously equilibrated at 25 °C and supplemented with 10 µM ThT. ThT fluorescence was measured using a BMG Labtech Optima plate reader with excitation and emission band passes of 440 ± 5 nm and 480 ± 5 nm for up to 360 min. ThT fluorescence curves were fitted to a double exponential function. Errors represent the SD of a minimum of 20 replicates. Soluble material was also measured as a function of time by SDS-PAGE following filtration to separate fibrils from soluble material (SI Appendix).

NMR.

For NMR experiments, depolymerization of ¹⁵N-labeled β₂m fibrils was initiated by dilution to a final (monomer equivalent) concentration of 150 μM into the pH 7.4 or pH 6.4 buffers described above. Each buffer was supplemented with 10% (vol/vol) D₂O and 50 µM ThT (the latter was included to enable direct comparison with the rates of fibril depolymerization, measured at the same time using the same stock of fibrils, by ThT fluorescence). The appearance of NMR-observable species was then monitored over time by acquiring SOFAST ¹H-¹⁵N HSQC spectra (34) at 25 °C using a Varian Inova spectrometer (Agilent) at a ¹H frequency of 750 MHz, equipped with a cryogenic probe. The dead time of the experiment was 3–7 min, and the acquisition time of individual spectra varied between 5 and 10 min (four and eight scans per incremental delay, respectively). At least 20 sequential ¹H-¹⁵N HSQC spectra were acquired with a final acquisition (of 100% native monomer) performed 24 h after the initial dilution. The experiment was repeated several times, increasing the acquisition time to optimize the signal-to-noise ratio while maintaining time resolution. The average of two such experiments is shown in Fig. 2. The error bars represent the propagated error of the fits to single exponentials for each residue over the two experiments.

FCS.

Details of the instrument set-up, data collection, and analysis can be found in SI Appendix.

Hsp70 Binding and Cell Assays.

The kinetic stability of β₂m amyloid fibrils in the presence of hsp70 was monitored by diluting fibrils formed at pH 2.0–12 µM (monomer equivalent concentration) in pH 6.4 buffer to which 0, 0.25, or 3 molar equivalents of recombinant hsp70-1A or 3 molar equivalents of BSA were added. Depolymerization was then monitored at 25 °C by ThT fluorescence as outlined above. Data shown are the average from a minimum of three independent replicates.

MTT and lysosome proteolysis assays were performed on SH-SY5Y cells as described (10). Cells were cultured in the presence of 0, 0.3, or 3.0 μM hsp70 for 6 h before the addition of 1.2 µM (monomer equivalent concentration) of fibrils that had been fragmented for 48 h as described (9). Native β₂m monomer, buffer-only, hsp70-only, or 0.1% (wt/vol) sodium azide were added as controls. For the MTT assay, cells were incubated for a further 24 h before the addition of 20 µL of 5 mg⋅ml⁻¹ MTT. After further incubation for 60 min, the media was removed and the resulting formazan crystals were solubilized in 100% dimethyl sulfoxide. Reduction of MTT was quantified by measuring the absorbance at 570 nm and subtracting background light scattering at 650 nm. The error bars represent the SD from three independent experiments, each containing five replicates. Details of the lysosome proteolysis assay can be found in SI Appendix.