Molecular conformation and dynamics of the Y145Stop variant of human prion protein in amyloid fibrils

SSNMR Spectra Identify a Highly Ordered, Compact Amyloid Core with a Predominantly β-Strand Conformation.

The fingerprint 2D ¹⁵N-¹³C′ (NCO), ¹⁵N-¹³C^α (NCA), and ¹³C-¹³C (CC) chemical-shift correlation spectra of uniformly ¹³C,¹⁵N-labeled huPrP23–144 amyloid fibrils at 0°C are shown in Fig. 1 and supporting information (SI) Fig. S1. Essentially identical spectra (i.e., |Δδ|_avg < 0.2 ppm) were obtained for two independently prepared fibril samples, indicating high reproducibility of the experiments. Remarkably, only ≈20% of the cross-peaks (i.e., 28 of a total of 125 or 126) expected in 2D NCO, NCA, or CC spectra (C′–C^α region) are observed under the present experimental conditions (note that huPrP23–144 used in this study contains 126 residues, including a 4-residue N-terminal extension). Given that cross-polarization (CP), which requires the presence of substantial through-space magnetic dipole–dipole couplings (29, 30), was used to transfer magnetization between ¹H, ¹³C, and ¹⁵N nuclei, this clearly indicates the presence of a compact (≈28 residues), relatively rigid “core” region of the amyloid. In contrast to this rigid core region, the remainder of the protein backbone in huPrP23–144 fibrils shows a high degree of conformational flexibility (see Variable Temperature NMR Studies of Conformational Dynamics in huPrP23–144 Amyloid). The existence of such rigid and dynamic segments has been previously reported for amyloid fibrils formed by other proteins, including α-synuclein (21, 22) and HET-s (23, 31).

Site-specific ¹³C and ¹⁵N NMR linewidth measurements further reaffirm that the amyloid core of huPrP23–144 fibrils is highly ordered on the atomic level. Indeed, as demonstrated by representative ¹³C and ¹⁵N 1D traces shown in Fig. 1, for most residues 2D NCO and NCA spectra display relatively narrow and symmetric resonances. Quantitative measurements for a set of 20 well resolved peaks in the 2D NCA spectrum (processed with no apodization) gave the average ¹⁵N and ¹³C widths at half-height of 0.98 ± 0.18 and 0.63 ± 0.13 ppm, respectively. These linewidths, comparable with those recently reported for amyloids formed by a synthetic fragment of transthyretin (27, 28) and HET-s protein (23), approach the limit of narrowness observed for peptides and globular proteins in a highly ordered microcrystalline state (32). Moreover, a 2D NCO spectrum recorded in the presence of ¹H and ¹⁵N spin decoupling during ¹³C detection shows that for residues distributed throughout the amyloid core, the ¹³C spectral resolution is to a large extent limited by the presence of ¹³C′-¹³C^α J-couplings (see Fig. S2). This further confirms high degree of order in this part of huPrP23–144.

To obtain the sequential ¹³C and ¹⁵N backbone and side-chain resonance assignments for the amyloid core region, a set of 2D and 3D chemical-shift correlation spectra was recorded at 0°C (see Materials and Methods). Representative strips from 3D ¹³C′-¹⁵N-¹³C^α (CONCA), ¹⁵N-¹³C^α-¹³CX (NCACX), and ¹⁵N-¹³C′-¹³CX (NCOCX) spectra (Fig. 2) show sequential backbone connectivity for residues A118–V122. The resonance assignments are listed in Table S1 and shown in the 2D spectra in Fig. 1 and Fig. S1. Importantly, no signals from residues 23–111 and 142–144 were observed in any of the experiments, clearly indicating that the amyloid core of huPrP23–144 maps to the region encompassing residues 112–141. Resonances for residues 112, 130, and 141 exhibited notably reduced intensities, and no cross-peaks were observed for residues 128–129 (except M129 ¹³C′, via a weak correlation to L130 ¹⁵N in the NCO spectrum). This points to the presence within the 112–141 core of regions characterized by varying degrees of conformational flexibility, with the short segment Y128–M129 being especially highly dynamic. Note that the identification of residues within the 112–141 core region is unambiguous. This unambiguousness of assignment, confirmed by multiple 2D and 3D NMR spectra, was facilitated by the fact that several amino acid types with unique ¹³C chemical-shift patterns (i.e., Ala, Ile, Val) are present only in this part of huPrP23–144.

It is well established that NMR chemical shifts are highly sensitive reporters of protein secondary structure, both in solution and the solid phase. Here, the ¹³C and ¹⁵N shifts were used to predict the backbone conformation of the huPrP23–144 amyloid core region. The ¹³C′, ¹³C^α, and ¹³C^β secondary shifts, Δδ (i.e., the differences between the experimental chemical shifts and the corresponding random-coil values), are plotted as a function of the residue number in Fig. 3. The inspection of these plots reveals that the ¹³C′ secondary shifts for all but two residues are negative, with an average value of Δδ_avg = −1.9 ± 1.6 ppm. Furthermore, most ¹³C^α and ¹³C^β secondary shifts are negative (Δδ_avg = −0.6 ± 1.5 ppm) and positive (Δδ_avg = 1.5 ± 1.4 ppm), respectively. This immediately indicates that the majority of residues in the huPrP23–144 amyloid core region adopt a β-strand conformation. To further probe this issue, a semiquantitative conformational analysis was performed by using the program PSSI (33). The normalized probabilities of different secondary structure types (i.e., α-helix, β-strand, and coil) and the secondary structure classification are shown in Fig. 3. Altogether, this analysis clearly indicates that most residues between 112 and 122, and 130 and 139 are in a β-strand conformation, whereas unordered structure is predicted for residues 140–141 at the edge of the amyloid core, as well as residues 123–129. The latter region contains four glycines and the highly flexible Y128–M129 segment.

Variable Temperature NMR Studies of Conformational Dynamics in huPrP23–144 Amyloid.

As noted above, most of the NMR signals, including those corresponding to the entire N-terminal segment 23–111, were not observed in 0°C experiments with cross-polarization. This suggests that large segments of huPrP23–144 in the amyloid state remain highly dynamic, attenuating nuclear dipole–dipole couplings required for efficient cross-polarization. To confirm the presence of this segmental dynamics, we performed systematic SSNMR studies as a function of temperature between −30 and 30°C. Fig. S3 shows a series of 1D ¹³C MAS spectra, where the initial ¹³C magnetization was generated by either ¹H-¹³C CP (i.e., the conditions that “filter out” flexible parts of the protein) or direct polarization (DP) (i.e., a single 90° ¹³C pulse, which excites with equal efficiency nuclei in both rigid and flexible regions). These experiments may be summarized as follows. First, whereas the CP spectra between −20 and 30°C show only modest temperature-dependent differences, the −30°C spectrum is markedly different, displaying substantially broader signals and an ≈2-fold increase in the total integrated intensity as compared with the spectrum recorded at −20°C. Second, the linewidths in the DP series of spectra progressively decrease as the temperature increases; a particularly striking example is the region corresponding to side chains of His, Phe, Trp, and Tyr (≈100–140 ppm). Third, for all CP/DP pairs recorded between −20 and 30°C, significant differences in peak widths and intensities (i.e., apart from a uniform intensity scaling of the entire spectrum) are observed at each temperature. Only for the −30°C CP/DP pair is the CP spectrum an essentially 2-fold enhanced version of the DP spectrum, as would be expected in the rigid-lattice limit (29). Altogether, these variable-temperature studies confirm the presence of nonuniform backbone and side-chain dynamics of huPrP23–144 molecules in the amyloid fibrils. Although the amyloid core remains highly ordered at all temperatures probed, the entire region between the N terminus and residue 112 exhibits significant conformational dynamics. This segmental flexibility is attenuated at low temperatures, becoming effectively “frozen out” at approximately −30°C.

To gain higher-resolution insight into temperature-dependent changes in the segmental flexibility of huPrP23–144 fibrils, we recorded 2D NCA and CC spectra at −30°C, 0°C, and 20°C (Fig. 4 and Fig. S4). Although the 0°C and 20°C spectra are similar to each other, major changes in peak intensities and widths are evident at −30°C. In addition to the amyloid core signals, numerous additional peaks appear at −30°C in all spectral regions associated with backbone and side-chain ¹⁵N-¹³C and ¹³C-¹³C correlations. The appearance of these additional resonances indicates that at −30°C parts of the protein outside the amyloid core region become sufficiently rigid to enable efficient dipolar magnetization transfers between ¹H, ¹³C, and ¹⁵N nuclei, in complete agreement with the 1D ¹³C spectra discussed above. In fact, quantitative analysis of the 2D NCA spectra reveals that most residues in huPrP23–144 become rigid at −30°C. Using the integrated cross-peak volumes, it can be estimated that ≈83–120 total residues and ≈35–47 glycines contribute to the −30°C spectrum. Given the approximate nature of such an analysis, these estimates are in reasonable agreement with the expectation that 125 residues in huPrP23–144, including 40 glycines, should contribute to the spectrum in the rigid-lattice limit. Finally, it is notable that, in contrast to the amyloid core resonances, additional cross-peaks that appear at −30°C exhibit substantial inhomogeneous linebroadening. This suggests that the flexible N-terminal region of huPrP23–144 in amyloid fibrils likely exists as an ensemble of largely disordered states characterized by a random-coil-like conformation. This conclusion is further supported by characteristic random-coil chemical shifts observed for all Gly ¹³C^α and Thr ¹³C^β resonances that can be uniquely identified in the 1D ¹³C DP spectrum recorded at 30°C (Fig. S3I).

Expression and Purification of huPrP23–144.

The plasmid encoding huPrP23–144 with N-terminal linker containing His₆-tag and a thrombin cleavage site is described in ref. 49. Uniformly ¹³C,¹⁵N-labeled huPrP23–144 was expressed in Escherichia coli by using a minimal medium with ¹³C-glucose (3 g/liter) and ¹⁵NH₄Cl (1 g/liter) (Cambridge Isotope Laboratories) as the sole carbon and nitrogen sources, respectively, and purified by using a nickel-nitrilotriacetic acid agarose resin (19). The His₆-tag was cleaved by using biotinylated thrombin (Novagen), the thrombin was sequestered by using streptavidin-agarose beads, and the residual His₆-tag was removed by dialysis against ultrapure water. The purified 126-residue protein consisted of the huPrP23–144 sequence with an N-terminal Gly-Ser-Asp-Pro extension. The protein was stored as a lyophilized powder and had a final purity of >95% as determined by SDS/PAGE.

Preparation of huPrP23–144 Amyloid Fibrils.

Lyophilized huPrP23–144 was dissolved in ultrapure water at a concentration of 400 μM, and fibrilization was initiated as described previously (14, 19) by addition of potassium phosphate buffer (pH 6.4) to a final concentration of 50 mM. Fibrils were allowed to form at 25°C under the conditions of gentle rotation at 8 rpm. Before NMR analysis, amyloid fibril samples were characterized by atomic force microscopy (Fig. S5; see refs. 14 and 20 for experimental details). Fibrils (≈12 mg of ¹³C,¹⁵N protein) were pelleted by centrifugation, washed several times with 50 mM potassium phosphate buffer (pH 6.4), and packed by using low-speed centrifugation into a 3.2-mm, limited-speed, 36-μl Varian SSNMR rotor. The rotor was sealed by using custom-made spacers to prevent sample dehydration during experiments.

SSNMR Spectroscopy.

Spectra were recorded by using a three-channel, 500-MHz Varian spectrometer equipped with a 3.2-mm BioMAS probe in the ¹H–¹³C–¹⁵N configuration. The sample temperature was controlled by a stream of compressed air delivered to the sample via a variable-temperature (VT) stack. All temperature values listed in the article consistently refer to the temperature of the VT gas at the sample; under the experimental conditions used, the average sample temperatures were 5 ± 2°C higher than the VT gas temperature, as determined by lead nitrate calibration.

The pulse schemes used in this work are analogous to those used in previous SSNMR studies of proteins (32). The following 2D and 3D experiments were used to establish sequential resonance assignments of the huPrP23–144 amyloid core region at 0°C: 2D ¹³C-¹³C (CC), 2D ¹⁵N-¹³C^α (NCA), 2D ¹⁵N-¹³C′ (NCO), 2D ¹⁵N-(¹³C^α)-¹³CX (N(CA)CX), 2D ¹⁵N-(¹³C′)-¹³CX (N(CO)CX), 3D ¹³C′-¹⁵N-¹³C^α (CONCA), 3D ¹⁵N-¹³C^α-¹³CX (NCACX), and 3D ¹⁵N-¹³C′-¹³CX (NCOCX). Where applicable, the ¹⁵N-¹³C^α/¹³C′ correlations were established by using adiabatic SPECIFIC CP (30) (τ_mix,NCA/τ_mix,NCO = 3.5/6.0 ms), and ¹³C-¹³C magnetization transfer was achieved by using dipolar assisted rotational resonance (50) (τ_mix = 10 ms for 2D CC and 15 ms otherwise). Two-pulse phase modulated proton decoupling (51) at a field strength of ≈70 kHz was used during all chemical-shift evolution periods. Data were processed by using NMRPipe (52).