Amyloid formation under physiological conditions proceeds via a native-like folding intermediate

Abstract

Although most proteins can assemble into amyloid-like fibrils in vitro under extreme conditions, how proteins form amyloid fibrils in vivo remains unresolved. Identifying rare aggregation-prone species under physiologically relevant conditions and defining their structural properties is therefore an important challenge. By solving the folding mechanism of the naturally amyloidogenic protein beta-2-microglobulin at pH 7.0 and 37 degrees C and correlating the concentrations of different species with the rate of fibril elongation, we identify a specific folding intermediate, containing a non-native trans-proline isomer, as the direct precursor of fibril elongation. Structural analysis using NMR shows that this species is highly native-like but contains perturbation of the edge strands that normally protect beta-sandwich proteins from self-association. The results demonstrate that aggregation pathways can involve self-assembly of highly native-like folding intermediates, and have implications for the prevention of this, and other, amyloid disorders.

Figure 1. Kinetic elucidation of the folding mechanism of wild-type β₂m and P32G.

(a,b) Kinetics of folding (a; top to bottom, 0.2–1.6 M GdmHCl in 0.1-M increments) and unfolding (b; top to bottom, 1.4–4.0 M in 0.2-M increments) of wild-type β₂m. (c) Folding mechanism of wild-type β₂m and associated microscopic rate constants, in the absence of GdmHCl (pH 7.0, 37 °C), determined by global fitting of all experimental data. (d,e) Kinetic folding and unfolding traces of P32G (traces colored as in a,b). (f) Folding mechanism of P32G and associated microscopic rate constants, determined as in c.

Figure 2. Amyloid-fibril formation from wild-type β₂m and P32G at pH 7.0 and 37 °C.

(a,b) Time course of the fibril formation of wild-type β₂m (a) in 0 M (black) and 0.6 M (red) sodium sulfate, and of P32G (b) at increasing concentrations of sodium sulfate from 0 M (top, black) to 0.6 M (bottom, red) in 0.1-M increments. Error bars show s.d. over three replicate samples. (c,d) Negative-stain EM images of fibrils formed from wild-type β₂m (c) and P32G (d) after 7 d of incubation at pH 7.0, 37 °C in 0 M sodium sulfate. Black bar represents 100 nm. (e) Free-energy diagram for P32G β₂m in the absence (black) and presence (red) of 0.6 M sodium sulfate. A pre-exponential term of 4.8 × 10⁸ s⁻¹ was used to calculate the energy of the transition states (TS). (f) Plot showing equilibrium concentrations of unfolded (blue), intermediate (red) and native state (black) of P32G at different sodium sulfate concentrations. (g) Correlation between initial rate of fibril formation and equilibrium concentration of I_T for P32G in different concentrations of sodium sulfate (open circles) and for wild-type β₂m in 0 M sodium sulfate (red circle). Inset, concentrations of unfolded (blue), intermediate (red) and native states (black) of P32G at equilibrium versus initial rate of fibril formation. Error bars indicate standard deviations resulting from fitting the experimental data.

Figure 3. Identification of conformational-exchange processes in P32G.

(a,b) 2D ¹H-¹⁵N HSQC spectra of wild-type β₂m (a) and P32G (b) at pH 7.0 and 37 °C. Insets show that line broadening of several resonances in P32G is not observed in the spectrum of the wild-type protein. (c,d) Example regions of spectra obtained from wild-type β₂m (c) or P32G (d) at 37 °C and at 4 °C in the absence or presence of 0.4 M sodium sulfate. Chemical exchange between two species (N_T and I_T), showing very similar chemical shifts, results in peak splitting at low temperatures, not observed in the wild-type protein. Consistent with the proposed mechanism (Fig. 1f), addition of sodium sulfate reduces the concentration of I_T (Fig. 2f), resulting in the appearance of a single residual resonance, further allowing the identification of peaks corresponding to either N_T or I_T.

Figure 4. Analysis of ¹H-¹⁵N HSQC spectra of wild-type β₂m and P32G at pH 7.0, 37 °C.

(a) Chemical shift differences (Δδ) for all clearly detectable peaks in spectra of the two proteins. Resonances of P32G show chemical shifts within <0.1 p.p.m. of their wild-type counterparts. Horizontal black bars A–G indicate positions of native β-strands labeled in c. (b) Peak intensity of resonances of P32G relative to their wild-type counterparts. Resonances arising from residues in the N-terminal region, native β-strands A and D and the BC and FG loops show less than 20% of the intensity of the equivalent peaks in the spectrum of wild-type β₂m, whereas only minor differences in intensity are observed for all other resonances. (c) Exchange broadening of NMR resonances in P32G, indicating residues with relatively large structural differences between the structure of the native state (N_T) and the intermediate (I_T). These residues are mapped onto the structure of wild-type β₂m. Green, resonances with intensities <20% of the equivalent resonance in wild-type β₂m; gray, those that could not be analyzed, as the corresponding resonances are not assigned; blue, all other resonances; yellow, the single disulfide bond; red sphere, Pro32. The figure was prepared using PDB entry 1JNJ and PyMOL.

Figure 5. Schematic free-energy landscape of wild-type β₂m.

The free-energy surface is represented as a function of two reaction coordinates, Φ₁ and Φ₂, indicating the increase of native intramolecular interactions during folding and non-native intermolecular interactions during aggregation, respectively. Local minima represent the ensembles of unfolded states (U_C and U_T), the intermediate ensembles (I_C and I_T) and the native state (N). The thermodynamically most stable structure, the β₂m monomer within an amyloid fibril, is indicated (F). I_T directly links the folding (black arrows) and aggregation (red arrow) pathways. In vivo, the energy landscape will be ‘rougher’ than depicted schematically here, as various micro- and macrostates are populated on both pathways.