Detection of side-chain proton resonances of fully protonated biosolids in nano-litre volumes by magic angle spinning solid-state NMR
We present a new solid-state NMR proton-detected three-dimensional experiment dedicated to the observation of protein proton side chain resonances in nano-liter volumes. The experiment takes advantage of very fast magic angle spinning and double quantum 13C–13C transfer to establish efficient (H)CCH correlations detected on side chain protons. Our approach is demonstrated on the HET-s prion domain in its functional amyloid fibrillar form, fully protonated, with a sample amount of less than 500 µg using a MAS frequency of 70 kHz. The majority of aliphatic and aromatic side chain protons (70%) are observable, in addition to Hα resonances, in a single experiment providing a complementary approach to the established proton-detected amide-based multidimensional solid-state NMR experiments for the study and resonance assignment of biosolid samples, in particular for aromatic side chain resonances.
KeywordsAmyloid fibrils Solid-state NMR Proton detection Very fast MAS Protein NMR
Solid-state magic angle spinning (MAS) NMR (SSNMR) is a broadly applicable biophysical technique for the structural investigations of insoluble and non-crystalline biological protein assemblies, ranging from membrane proteins (McDermott 2009; Tang et al. 2013; Baker and Baldus 2014; Andreas et al. 2015a, b, c; Brown and Ladizhansky 2015; Kaur et al. 2016), to bacterial filaments and capsids (Goldbourt 2013; Loquet et al. 2013; Yan et al. 2013; Linser 2017), and fibrillar aggregates (Meier and Bockmann 2015; Tycko 2016; Meier et al. 2017; Wel 2017). In particular, several three-dimensional (3D) structural models of amyloid proteins in their relevant fibrillar forms have been solved using MAS SSNMR methodology at atomic resolution (Meier et al. 2017; Wel 2017; Wasmer et al. 2008; Xiao et al. 2015; Colvin et al. 2016; Hoop et al. 2016; Tuttle et al. 2016; Walti et al. 2016; Qiang et al. 2017), revealing fundamental aspects of protein aggregation and propagation of the amyloid fold. A prerequisite for defining 3D structures by ssNMR is the collection of a large number of internuclear distance restraints derived from chemical shift assignments. This assignment step usually relies on the identification of 13C and 15N chemical shifts and is based on a well-established methodology, now in use for more than a decade, which combines 13C- and 15N-detected multidimensional SSNMR experiments (Sun et al. 1997; Hong 1999; McDermott et al. 2000; Detken et al. 2001; Pauli et al. 2001), such as 13C–13C, 15N–13C, 15N–13C–13C and 13C–15N–13C correlation spectra. Assignment of the protein backbone usually permits position-specific characterisation of secondary structure and can be used to probe the extent of conformational polymorphism and structural order. The detection of side chain 13C resonances, and their related connectivity to backbone resonances, is usually encoded in 13C–13C, 13C–13C–13C and 15N–13Cα–(or 13CO)–13CX experiments. The additional side chain assignments further facilitate the assignment process, help to distinguish amino acid type and, most importantly, provide access to the structural restraints mostly established between side-chain carbons that are necessary to perform structure determination and identify quaternary organisation or define intermolecular interactions.
The observation and assignment of side chain proton resonances offers perhaps the richest potential source of intra and inter-molecular distance restraints due to their ubiquity and greater proximity to molecular interaction sites. Their larger intrinsic sensitivity (greater gyromagnetic ratio compared to 13C and 15N heteronuclei) and natural abundance make them very attractive NMR-active reporters. Yet, until recently, this information has remained largely untapped due to the technological challenges involved with the fast sample spinning that is necessary to diminish strong proton dipolar-couplings and observe sharp, well-resolved, NMR signals from biosolid samples. In lieu of ultra-fast spinning probes, which have only recently become commercially available, early attempts at detecting proton chemical shifts focussed on combining fast MAS spinning with partial deuteration, in order to reduce the proton density (Chevelkov et al. 2003, 2006; Paulson et al. 2003; Zhou et al. 2007; Huber et al. 2011; Knight et al. 2011; Linser et al. 2011; Sinnige et al. 2014). The principal approach to proton spin dilution includes perdeuteration of a biosolid sample (Chevelkov et al. 2003; Paulson et al. 2003; McDermott et al. 1992; Reif et al. 2001; Reif and Griffin 2003) with either subsequent back-exchange of the exchangeable protons or the incorporation of specifically protonated moieties. Nevertheless, such samples tend to be far from ideal in that they are expensive to prepare, may not be appropriate for all systems (i.e. achieving uniform back-exchange or of sufficient yield) and that each sample still only allows the observation of a subset of isolated proton species. Moreover, such samples are also markedly less practical for the acquisition of popular experimental mixing schemes such as proton-driven spin-diffusion (PDSD) (Bloembergen 1949) and diffusion-assisted rotational-resonance (DARR) (Takegoshi et al. 2001); wherein the proton bath to be exploited in magnetization transfer is no longer present for efficient spin diffusion to occur. Furthermore, partially deuterated samples have the disadvantage that proton-mediated long-range distance contacts cannot be detected on deuterated side-chains. There are of course many examples of productive and insightful studies that have employed proton-detected SSNMR for structural and dynamics investigations (Asami and Reif 2013; Andreas et al. 2015; Fricke et al. 2017). Specific pulse sequences have also been developed for proton-detection of specifically protonated samples for side chain carbon-detection (Kulminskaya et al. 2016), and non-deuterated samples for aliphatic proton (Vasa et al. 2016) and carbon (Xiang et al. 2016) observation. Indeed, the practicality of new generation ultra-fast spinning SSNMR probes (> 60 kHz) is already facilitating standard proton detection and structure determination in fully protonated samples (Andreas et al. 2015; Penzel et al. 2015; Stanek et al. 2016; Struppe et al. 2017; Xue et al. 2017). In this study, we propose a novel 3D SSNMR experiment to efficiently detect side chain proton resonances in biosolid samples, by establishing CCH correlations. The experiment starts from a 1H to 13C cross polarization (CP) step, evolution, followed by a double quantum generation and evolution, and then a CP-back to proton. The 3D (H)CCH experiment exploits an indirectly detected 13C double quantum polarization transfer and a 13C double quantum dimension. We take advantage of very fast MAS frequencies to establish the detection of aliphatic protons using a sample amount of approximately 500 µg. The resulting spectrum has no diagonal signals and provides correlations between backbone Cα nuclei to side chain carbons and their attached protons, as well as side chain—side chain 13C–13C correlations connected to their corresponding side chain protons. Side chain aromatic resonances, usually difficult to detect, are observable through various combinations, including Cγ–Cβ–Hβ, Cγ–Cδ–Hδ, Cδ–Cε–Hε and Cζ–Cε–Hε. Our experiment is demonstrated on a fully-protonated sample of HET-s prion-forming domain amyloid fibrils.
Materials and methods
Uniformly [13C,15N]-labelled samples of the HET-s prion domain (218–289) were prepared as previously described (Balguerie et al. 2003). Approximately 0.5 mg of hydrated protein was used to pack a 1 mm zirconia rotor for solid-state NMR experimentation.
(H)CCH pulse sequence composition
The building block of the (H)CCH experiment is a combination of 13C detected 13C DQ/13C SQ correlation and 1H detected 13C/1H CP-HSQC experiments. Instead of 13C detection in the 13C detected 13C DQ/13C SQ correlation experiment, data is acquired through 1H detection together with the additional magnetization transfer from 13C to 1H in the (H)CCH experiment. First, the 13C longitudinal magnetization is prepared from 1H thermal magnetization using the first 1H→13C cross polarization followed by a 90° 13C flip-pulse. The DQ excitation pulses convert the 13C longitudinal magnetization into 13C DQ coherence. Following which, the 13C DQ coherence is allowed to evolve during the t1 period under 1H decoupling and then back transferred to 13C longitudinal magnetization by the DQ reconversion pulses. We used the BABA-xy16 (Saalwachter et al. 2011) homonuclear recoupling sequence during the DQ excitation and reconversion periods, since BABA-xy16 has shown to be extremely robust towards experimental imperfections. While 13C magnetization is stored along the z-axis, the residual 1H magnetization and solvent signals are suppressed by homospoil pulsed field gradient (PFG) followed by homonuclear rotary resonance recoupling (HORROR) irradiation. The 13C 90° pulse excites the 13C SQ coherence that evolves during the t2 period under 1H decoupling. The 13C SQ coherence is then transferred to 1H that evolves during the t3 period under the 13C decoupling. The coherence pathway is selected by 8 step phase cycling.
Pulse sequence optimisation and acquisition
The experimental conditions for the (H)CCH experiment were optimized using 13C, 15N uniformly labelled l-alanine at 70 kHz MAS. 1H (13C) 90° pulses of 0.85 µs (0.7 µs) with RF field strengths of 354 (357) kHz were used. The 13C BABA pulses were optimized to 0.80 µs at 357 kHz and the duration of 16 τr (= 229 µs) was used for DQ excitation and reconversion to achieve the maximum signal. We tested different rf fields for BABA recoupling (from 156 to 333 kHz), and found only marginal differences at rf > 227 kHz. At 156 kHz, a 20% loss is observed. For CP transfers, DQ matching condition (n = 1) was used with RF strengths of 25 and 45 kHz on 1H and 13C, respectively, with ± 7 kHz ramp on 13C. The CP contact times were set to 2 ms for 1H→13C transfer and 0.4 ms for 13C→1H transfer. The low-power 1H TPPM and 13C WALTZ heteronuclear decouplings were employed with the RF field strength of 7 and 9 kHz, respectively. The residual 1H signals were suppressed using 40 ms homospoil PFG and 0.2 s HORROR irradiation with an RF strength of 36 kHz. The repetition delay was set to 1.3 s. Eight scans were accumulated for each indirect point with 64 increments for both t1 and t2 durations. Spectral widths were 52, 200 and 464 ppm for the 1H, 13C SQ (2.12 ms maximal evolution) and 13C DQ (0.91 ms maximal evolution) dimensions, respectively. The States-TPPI method was employed to discriminate the sign in the indirect dimensions. The total measurement time was 60 h.
The 2D CC PDSD experiment was acquired at 800 MHz using a Bruker AVANCE III spectrometer equipped with a 3.2 mm 1H/13C/15N MAS probe. Sample temperature was maintained at 277 K and the experiment was acquired with 48 scans (~ 20 h acquisition) at 11 kHz MAS with 2374 × 737 complex points and 295 and 122 ppm spectral widths, respectively. A mixing time of 50 ms was used and proton decoupling was achieved with SPINAL-64.
The 3D (H)CCH solid-state NMR experiment was carried out at 274 K with 70 kHz MAS using a JEOL RESONANCE Inc., JNM-ECA600II spectrometer, of 14.08 T field strength (599.67 MHz 1H Larmor frequency), equipped with a 1 mm double-resonance ultrafast MAS probe. Data were processed using NMRPipe (Delaglio et al. 1995), with 200 Hz Lorentzian line broadening in the directly detected 1H dimension and zero-filling (four-fold) in both indirect 13C dimensions. Spectra were analysed using the CCPNMR analysis software package (Vranken 2005).
Our approach is based on a three-dimensional (3D) experiment that correlates two neighbouring carbons, followed by detection on the attached proton. It is achieved by multiple polarization transfers enabling a (1H)–13C–13C–1H polarization pathway, the first 1H not being detected. In multi-dimensional NMR experiments, where multiple magnetization transfer steps are involved, the overall magnetization transfer efficiency must be sufficient to observe the resulting correlation peaks. Figure 1 shows the pulse program developed in this study.
In order to evaluate the experimental feasibility of the (H)CCH experiment, we optimised the efficiency of the magnetization transfer using a set of one-dimensional (1D) 1H-detected experiments. 1D experiments using either a (1H–13C)–1H (using CP-based HSQC polarization transfer steps) or (1H–13C–13C)–1H were performed on a 13C, 15N uniformly labelled sample of l-alanine to compare the filtering efficiencies. The second experiment uses a 13C double quantum (DQ) to 13C single quantum (SQ) polarization transfer step (here 13CDQ filtering) to correlate two 13C spins. DQ polarization transfer was done using the BABA sequence (Saalwachter et al. 2011). Approximately 40% of the magnetization observed from the single pulse experiments survived after the two-way 1H→13C→1H transfer for Hα, out of which ~ 50% remained after 13CDQ filtering (Supporting Information, Fig. S1). Therefore, 20% of the overall 1H signal was observed following the 1H→13CDQ→13CSQ→1H filtering. The benefits of double quantum transfer were recently employed by our group on small organic molecules (Zhang et al. 2017). When strictly considering polarization transfer efficiency, other recoupling schemes such as RFDR (radio frequency-driven recoupling) have also been shown to be highly efficient and practical under fast-MAS conditions (Nishiyama et al. 2014). We therefore tested the inclusion of different mixing schemes within a two-dimensional (2D) (H)CC experiment performed at 60 kHz MAS frequency. Optimization of the mixing period for the RFDR recoupling scheme leads to a maximal polarization transfer efficiency at ~ 10 ms (Supplementary Information, Fig. S2), showing a better efficiency compared to zero-quantum (ZQ) mixing in 13C–13C SQ/SQ correlations (PDSD and DARR) (Supplementary Information Fig. S3). RFDR and BABA recoupling schemes give comparable transfer efficiency under this experimental set up. While RFDR gives total correlations due to relayed transfers, BABA only provides correlations between bonded 13C pairs, giving a better selectivity compared to RFDR. This can be explained by the recoupled DQ Hamiltonians in the BABA sequence that do not commute each other, resulting in dipolar truncation. The filtering efficiency of the pulse sequence steps was also evaluated for the prion-forming domain of HET-s amyloid fibrils (residues 218–289). Although the two-way 1H→13C→1H filtering efficiency is unclear due to severe overlap of protein and water signals, the 13C DQ transfer efficiency was found to be similar to the l-alanine sample, and therefore sufficient for the (H)CCH experiments (data not shown).
A two-dimensional (2D) 13C–13C projection for the full 3D (H)CCH spectrum of the HET-s prion-forming domain (PFD) (218–289) is presented in Fig. 2. Preliminary evaluation of the spectrum quickly identified four distinct groups of resonance correlations; three of which arose via through-bond 13C–13C magnetization transfers for aliphatic, Ser/Thr aliphatic and aromatic correlations (Fig. 2a, b and c, respectively). Additionally, in contrast to the expected selective transfers, the fourth group of correlations corresponded to both single-quantum (SQ) and double-quantum (DQ) carbonyl excitations. These non-selective peaks presumably arise from longer range H-to-C (for DQ CO shifts) and C-to-H (for SQ CO shifts) magnetisation transfers when no proton was directly bound; a fact further evidenced from the observation that these non-selective peaks corresponded to both aliphatic and amide proton chemical shifts. Due to the combination of severe peak overlap in this region and the achieved 1H resolution (~ 500 Hz, 1.2 ppm FWHH), correlations involving carbonyl transfers were not, in general, assigned in this work. However, it can be hypothesised that with greater 1H resolution or with the use of specific labelling techniques, these correlations could give rise to useful sequential amide proton assignments. With respect to spectral dispersion, the inclusion of both double and single quantum 13C dimensions was advantageous in simplifying the identification of coupled spin-system carbons, as is the case for other INADEQUATE-type experiments. However, in the case of the 3D (H)CCH experiment, this advantage is expanded upon with the addition of a third, heteronuclear, dimension which further reduces the spectral complexity for a given plane. Spectral overlap was observed for the side-chain protons of Ile, Leu, and Val residues, however resonance assignment was otherwise largely unambiguous for protons of HET-s amyloid fibrils (see Table 1).
Statistics of the assignment percentage performed with the 3D 1H-detected 13CDQ/13CSQ experiment on the HET-s prion domain amyloid fibrils
Percentage assignment (%)
The observation of selective C–C–H transfers (excluding non-selective carbonyl transfers) was again advantageous by concomitantly acting to reduce spectral complexity and effectively restricting the observation of directly bound H–C chemical shift correlations to the direct 1H and indirect 13CSQ dimensions, respectively. In this respect, it was then possible to essentially ‘dial-in’ proton resonances by navigating in the spectrum to the summed chemical shift for their expected bound and adjacent carbons in the 13CDQ dimension and that of their directly bound carbon in the 13CSQ dimension. This is shown in both Fig. 3 and Supplementary Fig. S4 for examples of Ser/Thr and Ile Hα and Hβ assignment, respectively.
Thorough analysis of the (H)CCH spectrum allowed for proton resonance assignment for both backbone and side-chain protons of the HET-s rigid core (residues 226–241 and 262–277) (Table 1), based on 13C/15N resonance assignment performed by the Meier group (BMRB entry 11,064 (Melckebeke et al. 2010)). We found that the large majority (31 out of 32 residues) of the amyloid core of HET-s PFD can be identified based on their Hα–Cα correlations. Concerning the side chain moieties, we identified 85% of the expected 13C–1H side chain correlations included in the HET-s PFD, validating the efficiency of the approach.
A more unique element of the (H)CCH experiment is the ability to cleanly observe and assign aromatic side-chain protons. In the here-presented case of the HET-s (218–289) construct, only single tyrosine (Y), tryptophan (W) and phenylalanine (F) residues are present and the observable correlations were clear and tractable (although non-stereospecific). Figure 4 shows the observed correlations for Y281, wherein Hβ, Hδ and Hε protons were assignable, as were their related carbon nuclei. Peak correlations were however weaker than the principal aliphatic correlations, and so a lower contour threshold was required for analysis of aromatic correlations in contrast to that presented in Fig. 2. In fact, for Y281, 100% of the expected 1H side chain assignments were possible (including Hη) and a near-complete assignment was possible for F286 (all resonances except C/Hδ). Fewer assignments were possible in the case of W287, where a larger number of potential carbon-couplings with similar chemical shifts caused greater ambiguity. Tryptophan assignment could however prove less challenging in other systems exploring a greater conformational space (exclusively β-strand in HET-s) or with the use of selective glycerol- or glucose-based labelling schemes (Hong 1999a, b; Castellani et al. 2002). The clear observation of these aromatic protons should be of particular interest for future SSNMR studies that aim to study protein structure because the identification of long-range restraints internal to the hydrophobic interiors and interfaces of proteins, and that are typically occupied by aromatic residues (Moreira et al. 2013), can prove highly significant structural reporters.
From the extent of the HET-s PFD proton resonances that could be assigned, clear agreement can be seen with other published data (Wasmer et al. 2008; Siemer et al. 2006). Firstly, the observation of proton resonances for residues within the hydrophobic core of HET-s PFD (Fig. 5) serves to reinforce the description of the rigid β-solenoid fold; in contrast to the resonances of residues previously characterized as flexible (Siemer et al. 2006a, b) which were not observed. Figure 5 shows example side chain proton assignments for residues internal to the HET-s hydrophobic core of the first β-solenoid structural repeat. Similar to the aromatic proton assignment, the clean observation of protons within hydrophobic folds should prove useful for identification of long range 13C–13C restraints, 1H-detected, in future structural studies with the incorporation of longer mixing times for homonuclear 13C–13C recoupling.
Moreover, as Hα proton chemical shifts are sensitive reporters on protein secondary structure, we were also able to probe the conformational behaviour of HET-s (Fig. 6). Secondary Hα chemical shifts (deviation from random coil) well describe the twin four-β-strand structural elements as well as the flexibility of the linker residues in-between. This also highlights another potential use of the (H)CCH experiment, as the Hα-based secondary structure determination presented here could improve or even substitute the standard secondary chemical shift calculation (Luca et al. 2001) (based on Cα and Cβ chemical shifts); further contributing to structural studies of proteins which fall short of calculating 3D models.
The presented experiment represents an efficient SSNMR approach for the structural investigation of biosolids, as illustrated here on amyloid fibrils of the HET-s PFD. We believe the (H)CCH experiment to be an important addition to the growing portfolio of SSNMR experiments available for the identification and assignment of aliphatic and aromatic side chain proton resonances, especially in the case of fully-protonated samples. The increased sensitivity from proton detection allows for experimental acquisition in sensible time frames and the logical correlations between single and double quantum 13C chemical shifts also allows for a rapid and efficient means of 1H resonance assignment. It offers a powerful alternative to similar hCCH experiments proposed by the Reif and coworkers on perdeuterated samples (Agarwal and Reif 2008; Asami et al. 2012). The experiment has the disadvantage of relying on two highly time intensive indirect dimensions (13C SQ and DQ), restricting the use to long acquisition times to completely digitalize the 13C signal. The resulting poor 13C resolution in indirect dimensions still remains an obstacle towards performing a stand-alone resonance assignment based solely on this particular experiment, thus requiring additional 13C–13C correlation assignments. Performed at 70 kHz spinning frequency, it can be combined with previous methods (Asami and Reif 2013; Andreas et al. 2015; Fricke et al. 2017; Barbet-Massin et al. 2014) based on 1H amide-detected experiments to provide a full assignment of backbone and side chain atoms. The use of selectively deuterated samples would also further increase the spectral resolution (Xue et al. 2017) to reach high molecular weight complexes in a single 3D experiment using ultra-fast MAS. In this respect, the spectrum should also prove extremely useful for the future generation of structural restraints based on proton–proton proximities important to the hydrophobic cores and interactions common to protein assemblies. Our laboratory is currently exploiting this approach in combination with more complex isotopic labelling schemes to investigate the 3D atomic structures of large macromolecular assemblies, even at higher MAS frequencies (> 100 kHz).
We acknowledge financial support from the European Research Council (ERC) under the European Unions Horizon 2020 research and innovation programme (ERC-2015-StG GA no. 639020 to A.L.), IdEx Bordeaux (Chaire d’Installation to B.H., ANR-10-IDEX-03-02), the ANR (ANR-14-CE09-0020-01 to A.L., ANR-13-PDOC-0017-01 to B.H. and ANR-17-CE11-0035 to S.J.S) and the CNRS (IR-RMN FR3050). This work has benefited from the facilities and expertise of the Biophysical and Structural Chemistry Platform (BPCS, UMS3033).
- Andreas LB, Stanek J, Le Marchand T, Bertarello A, Cala-De Paepe D, Lalli D, Krejcikova M, Doyen C, Oster C, Knott B, Wegner S, Engelke F, Felli IC, Pierattelli R, Dixon NE, Emsley L, Herrmann T, Pintacuda G (2015c) Protein residue linking in a single spectrum for magic-angle spinning NMR assignment. J Biomol NMR 62:253–261CrossRefGoogle Scholar
- Barbet-Massin E, Pell AJ, Retel JS, Andreas LB, Jaudzems K, Franks WT, Nieuwkoop AJ, Hiller M, Higman V, Guerry P, Bertarello A, Knight MJ, Felletti M, Le Marchand T, Kotelovica S, Akopjana I, Tars K, Stoppini M, Bellotti V, Bolognesi M, Ricagno S, Chou JJ, Griffin RG, Oschkinat H, Lesage A, Emsley L, Herrmann T, Pintacuda G (2014) Rapid proton-detected NMR assignment for proteins with fast magic angle spinning. J Am Chem Soc 136:12489–12497CrossRefGoogle Scholar
- Chevelkov V, van Rossum BJ, Castellani F, Rehbein K, Diehl A, Hohwy M, Steuernagel S, Engelke F, Oschkinat H, Reif B (2003) 1H detection in MAS solid-state NMR spectroscopy of biomacromolecules employing pulsed field gradients for residual solvent suppression. J Am Chem Soc 125:7788–7789CrossRefGoogle Scholar
- Knight MJ, Webber AL, Pell AJ, Guerry P, Barbet-Massin E, Bertini I, Felli IC, Gonnelli L, Pierattelli R, Emsley L, Lesage A, Herrmann T, Pintacuda G (2011) Fast resonance assignment and fold determination of human superoxide dismutase by high-resolution proton-detected solid-state MAS NMR spectroscopy. Angew Chem Int Ed Engl 50:11697–11701CrossRefGoogle Scholar
- Stanek J, Andreas LB, Jaudzems K, Cala D, Lalli D, Bertarello A, Schubeis T, Akopjana I, Kotelovica S, Tars K, Pica A, Leone S, Picone D, Xu ZQ, Dixon NE, Martinez D, Berbon M, El Mammeri N, Noubhani A, Saupe S, Habenstein B, Loquet A, Pintacuda G (2016) NMR spectroscopic assignment of backbone and side-chain protons in fully protonated proteins: microcrystals, sedimented assemblies, and amyloid fibrils. Angew Chem Int Ed Engl 55:15504–15509CrossRefGoogle Scholar
- Struppe J, Quinn CM, Lu M, Wang M, Hou G, Lu X, Kraus J, Andreas LB, Stanek J, Lalli D, Lesage A, Pintacuda G, Maas W, Gronenborn AM, Polenova T (2017) Expanding the horizons for structural analysis of fully protonated protein assemblies by NMR spectroscopy at MAS frequencies above 100 kHz. Solid State Nucl Magn Reson 87:117–125CrossRefGoogle Scholar
- Tuttle MD, Comellas G, Nieuwkoop AJ, Covell DJ, Berthold DA, Kloepper KD, Courtney JM, Kim JK, Barclay AM, Kendall A, Wan W, Stubbs G, Schwieters CD, Lee VM, George JM, Rienstra CM (2016) Solid-state NMR structure of a pathogenic fibril of full-length human alpha-synuclein. Nat Struct Mol Biol 23:409–415CrossRefGoogle Scholar