Cryo-EM structure of alpha-synuclein fibrils

1Parkinson’s disease (PD) is a neurodegenerative disorder characterized by the presence of Lewy bodies (LB) and Lewy neurites (LN). identified fibrils formed by the presynaptic protein alpha-synuclein (α-Syn, 140 residues, ~14 kD) as the main component of these human brain inclusions.

2Certain α-Syn fibril forms can seed LB-like and LN-like inclusions in cell culture and intra-neuronal aggregation of mouse α-Syn in vivo. In addition, abnormal α-Syn produces neuronal cell inclusions and axonal spheroids, as well as oligodendrocytic aggregates, known as glial cytoplasmic inclusions, found abundantly in Multiple System Atrophy (MSA), which makes α-Syn fibrils an important target for the development of diagnostic tools and therapeutic strategies for PD and related synucleinopathies.

3Despite α-Syn fibrils, other forms of α-Syn might also be involved in neurodegeneration, such as an oligomeric α-Syn intermediate, or the process of fibril aggregation itself. Fibrils of α-Syn show significant fibril strain polymorphism.

4Several factors point to α-Syn as an important player in the onset of PD: (i) six known point mutations in the α-Syn gene (SNCA) are associated with familial forms of synucleinopathies: A30P, E46K, H50Q, G51D, A53E, and A53T; (ii) animal models suggest a role of α-Syn in the etiology of PD, Dementia with Lewy Bodies (DLB), and MSA; (iii) individuals with duplications or triplications of the α-Syn gene exhibit overexpression of α-Syn and develop PD.

5Two related proteins, β-synuclein (β-Syn) and γ-synuclein (γ-Syn), with sequence homology to α-Syn, have been described. β-Syn and α-Syn share the greatest aminoacid sequence homology, with β-Syn lacking 12 amino acids (residues 71 to 82) within the non-amyloid component region (NAC; residues 61–95 in α-Syn). In synucleins, regions with the highest homologies are located in the structurally heterogeneous, amino-terminal half (residues 10–84 in α-Syn) composed of 5 to 6 imperfect repeats with the consensus sequence KTKEGV. In contrast, the carboxyl terminus is highly negatively charged and unstructured.

6A number of post-translational modifications have been described for α-Syn including phosphorylation, acetylation, ubiquitination, and C-terminal truncation. C-terminal truncation of α-Syn occurs normally in vivo, under physiological conditions and it has been shown to promote fibrillization. In turn, truncated forms of α-Syn play a role in inducing Lewy body formation, suggesting that truncation by proteolysis may be important in the pathological process.

7In vivo studies investigating α-Syn aggregation demonstrated that activation of the inflammasome and more specifically caspase-1, the enzymatic component of the inflammasome, leads to the production of an α-Syn fragment truncated at aspartic acid 121 (D121). This C-terminally-truncated α-Syn form (α-Syn(1-121)) aggregates more rapidly than full-length α-Syn (including disease-associated mutants), and its production is associated with cell toxicity. Furthermore, the use of VX-765, a pro-drug that produces a specific inhibitor of caspase-1 in vivo, improved survival of a neuronal cell model of PD, and reduced neurodegeneration in a transgenic mouse model of MSA, suggesting an important role of α-Syn(1-121) in cellular toxicity in both, cell cultures as well as a mouse model.

8To this date, high resolution structures of α-Syn fibrils are limited to the results of a micro-electron diffraction (microED) study of two small segments of the protein and a solid-state NMR structure obtained from ~5 nm diameter, single protofilaments, in addition to solid state NMR studies at the secondary structure level, and X-ray diffraction studies of shorter segments of α-Syn, or α-Syn bound to other molecules.

9Here, we report the atomic structure of α-Syn(1-121) fibrils determined by cryo-electron microscopy (cryo-EM). The structure allows conclusions about the organization of α-Syn fibrils at near-atomic resolution, suggest mechanisms for fibril formation and growth, and allows conclusions on fibril stability.

10Several preparations of recombinant human α-Syn fibril were screened by negative stain transmission electron microscopy (TEM; Figure 1—figure supplement 1). These included fibrils formed by full length α-Syn (Figure 1A), α-Syn phosphorylated at serine 129, N-terminally acetylated, and C-terminal truncated α-Syn comprised of residues 1–119 (α-Syn(1-119)), 1–121 (α-Syn(1-121)), or 1–122 (α-Syn(1-122)).

11The diameters of the α-Syn fibrils produced varied from 5 nm to approximately 10 nm when studied by negative stain TEM. The fibrils formed by α-Syn(1-121) were straight, between 20 and 500 nm long and the only ones of consistent diameters of 10 nm (Figure 1B, Figure 1—figure supplement 1E). This fibrillar form α-Syn(1-121) has been described as an aggregation-prone species resulting from α-Syn truncation by caspase-1. The recombinantly produced α-Syn(1-121) used here showed a similarly aggressive aggregation profile.

12Preparations of α-Syn(1-121) fibrils were quick-frozen in the holes of fenestrated carbon coated cryo-electron microscopy (cryo-EM) grids, and imaged with a Titan Krios 300kV cryo-EM instrument, equipped with a Quantum-LS energy filter and a K2 Summit direct electron detector. Helical image processing of recorded cryo-EM movies produced a 3D reconstruction of the α-Syn(1-121) fibril at an overall resolution of 3.4 Å (Figure 1C and D, Figure 1—figure supplement 2, Figure 2, and Video 1).

13Our 3D map shows that fibrils are formed by two protofilaments, each of 5 nm in diameter (Figure 1). These lack C2 symmetry, but are related by an approximate 21 screw symmetry, akin to the symmetry exhibited by the paired helical filaments of tau and by amyloid-ß(1-42) filaments. α-Syn(1-121) fibrils are therefore polar, meaning that both protofibrils are aligned into the same direction. The position of a given ß-sheet in a protofilament is produced by the rotation of 179.5° of one sheet around its axis (helical twist), followed by a vertical translation of 2.45 Å (helical rise). This ß-sheet arrangement results in a spacing of 4.9 Å between α-Syn subunits in successive rungs of a single protofilament (Figure 1C and D). The quality of the EM map allowed an atomic model of the region between residues L38 and V95 to be built.

14Each α-Syn(1-121) molecule comprises eight in-register parallel β-strands (i.e. residues 42–46 (β1), 48–49 (β2), 52–57 (β3), 59–66 (β4), 69–72 (β5), 77–82 (β6), 89–92 (β7), and 94-(~102) (β8)), which are interrupted by glycine residues (i.e. G41 before β1, G47 between β1 and β2, G51 between β2 and β3, G67 and G68 between β4 and β5, G73 between β5 and β6, G84 and G86 between β6 and β7, and G93 between β7 and β8) or an arch (i.e. E57-K58 between β3 and β4) (Figure 1A,F and G). The β-strands β2-β7 wind around a hydrophobic intra-molecular core composed of only alanine and valine residues and one isoleucine (i.e. V48, V49, V52, A53, V55, V63, A69, V70, V71, V74, A76, V77, A78, I88, A89, A90, A91). Considering that these hydrophobic clusters are maintained along the fibril, they are likely to contribute to the stability of the protofilament. The hydrophobic core is surrounded by two hydrophilic regions (i.e. (i): Q79, T81, and (ii): T72, T75, T54, T59, and E61) both still within the core of the structure (Figure 3). While most of these side chains form so-called side chain hydrogen bond ladders, the second hydrophilic region comprising four threonine residues and a negatively charged glutamic acid side chain surrounds a tunnel filled with some ordered molecules of unknown nature, as evidenced by an additional density (Figure 1—figure supplement 3D). The less well defined β1 and β8 strands are attached to the core, while the first 37 N-terminal residues and the last ~20 C-terminal residues of α-Syn(1-121) are not visible in the 3D reconstruction (Figure 1E and Figure 1—figure supplement 2A), indicating a disordered structure in line with quenched hydrogen/deuterium exchange – solution-state NMR (H/D exchange NMR) and limited proteolysis, which showed these terminal segments to be unprotected in nature. Together with our results, this suggests that approximately 40 residues of both the N- and C-terminal ends of full-length human α-Syn are flexible, and surround the structured core of the fibril with a dense mesh of disordered tails, similar to the ‘fuzzy coat’ recently described in the cryo-EM tau structure.

15Two β-sheets (one from each protofilament) interact at the fibril core via a hydrophobic steric zipper-geometry comprised of β-strand β3 (i.e. residues G51-A56). As a consequence, two α-Syn molecules per fibril layer are stacked along the fibril axis (Figure 2B and C). The side chains of residues A53 and V55 form the inter-molecular surface contributing to the interface between the two protofilaments, which is further stabilized by a surface-exposed salt bridge between E57 and H50 that might be sensitive to pH, as an unprotected histidine has a pK of ~6.2 (Figure 1—figure supplement 3H). The same structure with a steric zipper topology was found in micro-crystals of the peptide comprising residues G47-A56. Interestingly, the β-strand β6 that is sandwiched between β-strands β2/β3 and β7 is also aligned with a neighboring molecule but shifted by one monomer along the fibril axis, as shown in Figure 1G and Figure 2—figure supplement 1. Thus, hetero and homo steric zippers are both present in the 3D structure. Of these, the homo steric zipper at the inter-molecular interface has an extensive and well-packed β-strand interface, forming a very densely packed fibril. This stacking generates an asymmetric fibril with two distinct ends. Furthermore, the hydrophobic core of the fibril is composed of β-strands that interact with each other in a half-stacked zipper topology, contrasting with the hydrophilic core comprised of β-strands β4 and β5, which are non-stacked (Figure 1G and Figure 2—figure supplement 1). The latter confirms previous results from site-directed spin labeling experiments, which show that the region including residues 62–67 at the beginning of the NAC region, has a pronounced lack of stacking interactions.

16The outer surface of the ordered region of the fibrils is mostly hydrophilic, with a few exceptions (i.e. L38, V40, V82, A85, A90, F94, V95) (Figure 3A). The side chain of V66 should probably not be classified as surface exposed because of its interaction with β-strand β8 (Figure 1—figure supplement 2A). If we ignore the influence of the non-polar alanine residues due to the small size of their side chains, the surface of the fibrils has two highly hydrophobic regions formed by residues L38 and V40, and by residues F94 and V95. Other interesting properties of the surface are the salt bridge formed by the side chains of E46 and K80 (Figure 1—figure supplement 3G) and the rather highly positive clustering of K43, K45, K58, H50 that requests the binding of a counter-ion, as it is supported by an observed density (Figure 1—figure supplement 3C).

17Six familial mutations in α-Syn are known to be associated with PD and other synucleinopathies (i.e. A30P, E46K, H50Q, G51D, A53E, and A53T). Of these, all but A30P are located in the heart of the core of the fibril structure presented here (Figure 1A and E). E46 forms a salt bridge with K80 (Figure 1—figure supplement 3G). The mutation of the glutamic acid E46 to a positively charged lysine in an E46K mutant would thus induce a charge repulsion between β-strands β1 and β6, likely destabilizing this α-Syn fibril structure. The familial PD/DLB-causing mutation E46K was found to enhance phosphorylation in mice, and its toxic effect was increased by the triple-K mutation (E35K, E46K, E61K) in neuronal cells.

18Previous high-resolution structures of α-Syn only included small peptides or single protofilaments. Our 3D map suggests structural contributions of some familial mutations to fibril stability, since H50, G51 and A53 are all involved in the inter-molecular contact between the two β-sheets from adjacent protofilaments at the core of the here studied α-Syn(1-121) fibrils. Mutation of the positively charged histidine 50 into a polar, uncharged glutamine in the H50Q mutant would likely interfere with the salt bridge established between residues E57 and H50 (Figure 1—figure supplement 3H). Adding to the absent side-chain of glycine 51 a negatively charged aspartic acid in mutant G51D, or transforming the small side-chain of alanine A53 into a larger threonine in mutant A53T, would likely disrupt the steric zipper interaction between the two protofibrils, whereby the A53T mutation would in addition change the highly hydrophobic surface at the zipper to partly hydrophilic one. In our α-Syn(1-121) fibril structure, A53 is part of a hydrophobic pocket that defines the interaction of protofilaments and likely contributes to fibril stability as the hydrophobic interactions exist along the fibril axis. Mutations at the core of this α-Syn fibril would compromise the formation of the structure presented here. This suggests that a different fibril structure (i.e. fibril strain) could be formed from α-Syn containing the above discussed familial PD mutations.

19Several features of our structure, such as non-functional hydrophobic surface patches (Figure 3), a hydrophilic tunnel (Figure 1—figure supplement 3D), and a positively charged side chain arrangement like the one comprised of residues K43, K45, K58, H50 (Figure 1—figure supplement 3C) are not found in functional amyloid structures such as that of HET-s. However, similar structural characteristics have been previously observed for pathological tau filaments obtained from Alzheimer’s disease brains where (i): lysine and tyrosine residues play a similarly stabilizing role in the interface region of two protofilaments of the straight filaments (SF), and (ii): the area in the center of the protofilaments is dominated by hydrophilic residues. It is plausible that these structural features might arise because folding to form the amyloid fibril structure is dictated by the need to bury the maximum number of hydrophobic side-chains as efficiently as possible, as is also the case for the Aβ(1-42) amyloid fibrils.

20The artificial, highly toxic, but not synucleinopathy-related mutant E57K is interesting to mention in the context of the 3D structure presented, because E57 is also at the inter-molecular interface (Figure 2). The presence of a positive lysine side chain at this position in the E57K mutant would significantly interfere with the formation of the interface and even the amyloid fibril. Indeed, this mutant was designed in a successful structure-based attempt to interfere with amyloid fibril formation (at least under some conditions). Furthermore, both in a lentivirus-rat system as well as in a transgenic mouse model, the E57K mutant formed a significant amount of oligomers and was highly toxic, resulting in a large decay of TH-sensitive neurons in the substantia nigra of rats and a motor phenotype reminiscent of PD in mice. Thus, the artificial mutant E57K can be regarded as a ‘familial PD-like’ mutation both from the in vivo and from the structure/mechanism-based point of view.

21Full-length α-Syn subunits in a fibril studied by NMR (, PDB 2N0A) were found to be in a roughly similar secondary structure arrangement as in the here reported structure of α-Syn(1-121) (Figure 4A), even though the primary structure and the side-chain interactions of our here reported structure are very different from the NMR structure. Most importantly, the fibrils used for the NMR study were only approximately 5 nm wide, which corresponds to the diameter of a single protofilament. The larger diameter of our fibrils, 10 nm, results from the interaction between two protofilaments, which allowed us to hypothesize on the nature of α-Syn(1-121) protofilament interactions. Fibrils of 5 to 10 nm in diameter found in substantia nigra samples from the brain of PD patients,, cingulate cortex of patients with DLB, cerebral cortex of PD patients, and in-vitro aggregated samples. had already suggested that the 10 nm filaments are the result of the interaction between 5 nm protofilaments.

22An important difference between our here reported structure and the NMR structure reported by is the orientation of residue A53. The mutation A53T is associated with early onset PD. In our structure, residue A53 faces the interface between the two protofibrils and thereby likely contributes to fibril stability. In contrast, reported in their NMR structure A53 to point towards the hydrophobic core of the one observed individual protofilament, which may explain the lack of 10 nm fibrils in their sample. However, it is also noted here that the NMR study by showed a significant disagreement among the ten lowest-energy NMR structures for residues 51–67 [Figure 3d in ], indicating a lower confidence for those residues in the NMR structure. Our here reported cryo-EM map has the side-chains for those residues pointing into the opposite direction as reported in the structure.

23Our structure includes a serine residue at position 87 (Figure 1—figure supplement 3E), which is one of the several phosphorylation sites in α-Syn, in addition to Y125, S129, Y133 and Y135. S87 is the only phosphorylation site located within the NAC region. The previous solid-state NMR structure of α-Syn placed the side chain of this residue towards the inside of the protofilament core, leading to the assumption that phosphorylation of S87 might be the only modification occurring at a region not accessible in the fibrillar state. However, in our cryo-EM structure, S87 faces the outside of the fibril and hence remains accessible for disease-associated modification in α-Syn fibrils.

24We also observed the arrangement of G47 and A78 described by, which was proposed to favor the interaction between residues E46 and K80 and allow them to form a stable salt bridge between two consecutive α-Syn monomers (Figure 1—figure supplement 3G). The conservation of the geometry adopted by these residues confirms their role in facilitating backbone-backbone interactions. In addition, our structure also confirms that residues A69 and G93 (and likely G68) help to stabilize the distal loop in a protofilament (Figure 1—figure supplement 3F).

25A microED structure obtained from crystals produced from a 10-residue peptide simulating the core of α-Syn fibrils (PreNAC, from 47 to 56; Figure 4B) and including a threonine instead of an alanine at position 53 (i.e. A53T), also proposed that residue 53 forms the hydrophobic core within a protofilament. In addition, the microED model suggested that the interaction between adjacent protofilaments would occur through residues 68 to 78 (referred to as NACore). However, their short peptides did not include most residues responsible for the α-Syn monomer topology that we observed. Instead, our cryo-EM structure reveals that the PreNAC is responsible for the interaction between protofilaments, and places the NACore at the very center (i.e.the core) of a single protofilament.

26Our 3D structure allows us to hypothesize a mechanism for fibril elongation (fibril growth). Because two different stacking modes are present (i.e. the half-stack at the intermolecular interface and the stacking of β-strand β6), the two ends of the fibrils are distinct, suggesting an end-dependent growth of the fibrils, as documented and also suggested for other amyloids. One end of the fibril includes a hydrophobic cleft formed between β-strands β2/ β3 on one side and β7 on the other side (residues V49, V52, A88, I89), providing a hydrophobic entry point for the next incoming molecule, with the matching segment consisting of 5 hydrophobic residues (V74-V82, Figure 5). This suggests that the initial binding event of fibril elongation might be a hydrophobic interaction involving residues V74-V82. This peptide segment is the central part of the NAC region and strong experimental evidence suggests that it is critical for fibril formation. In addition, it has been shown that β-synuclein, which lacks residues V74 to V82, is incapable of forming fibrils.

27It is intriguing to speculate that a small molecule binding into this hydrophobic cleft could be a potent fibril elongation inhibitor or tracer, with the potential to be applied in PD and other synucleinopathies. Finally, the inter-molecular stacking may also play a role in fibril elongation, since the zipper interaction is of hydrophobic nature. Furthermore, it is likely that fibril growth alternates between the two protofilament structures at the level of monomer addition. Failure thereof may result in the growth of a single protofilament with little stability, yielding a dynamic on- and off-binding of monomers and larger oligomers, which has been observed for other amyloid fibril systems.

28In conclusion, we present the structure of recombinant α-Syn(1-121) fibrils determined at a resolution of 3.4 Å by cryo-EM. Our structure encompasses nearly the complete protein (residues 38 to 95), and includes the NAC region (residues 61 to 95) of α-Syn. We determined that various residues associated with familial forms of PD and other synucleinopathies are located in the interacting region between two protofilaments, suggesting their involvement in fibril formation and stabilization. The cryo-EM structure presented here reveals how two protofilaments interact to form a fibril, and how the NAC region contributes to protofilament formation and stability. Our structure also presents novel insights into how several PD-relevant mutations of α-Syn would compromise the structure of this fibril, suggesting that in the case of certain familial forms of PD, a different structure of α-Syn than this fibril strain might be involved. Our findings on protofilament interaction and our hypothesis on the mechanism of fibril elongation invite for the design of molecules for diagnostics or treatment of synucleinopathies.

29Recombinant full-length α-Syn was expressed from the pRT21 expression vector in BL21(DE3) competent Escherichia coli (E. coli). For N-terminal acetylation of α-Syn, cells were pre-transfected by pNatB vector coding for the N-terminal acetylase complex (plasmid kindly provided by Daniel Mulvihill, School of Biosciences, University of Kent, Canterbury, UK). C-terminally truncated forms of α-Syn(1-119), α-Syn(1-121), and α-Syn(1-122) were expressed in BL21-DE3-pLysS competent E. coli. Purification of α-Syn strains was performed by periplasmic lysis, ion exchange chromatography, ammonium sulfate precipitation, and gel filtration chromatography as previously described. Polo like kinase 2 (PLK2) was expressed in BL21-DE3-pLysS competent E. coli, isolated via its His-tag and immediately used to phosphorylate purified α-Syn. This was followed by standard ion exchange and gel filtration chromatography to separate phosphorylated from non-phosphorylated α-Syn. Endotoxins were removed from all α-Syn strains by Detoxi-Gel Endotoxin Removing Gel (Thermo Scientific) usually in one run or until endotoxin levels were below detection level. The sequence of the expressed α-Syn strains was verified by tryptic digestion followed by MALDI mass spectrometry (MS) or HPLC/ESI tandem MS for total mass was performed. Purity and monodispersity was determined by Coomassie blue or Silver staining of the SDS PAGE gel and analytical ultracentrifugation and the concentration was determined by the bicinchoninic acid (BCA) assay (Thermo Scientific) with bovine serum albumin as a standard. Dialyzed and lyophilized α-Syn(1-121) was prepared by dialyzing the purified protein in a 2 kD Slide-A-Lyzer unit (Thermo Scientific, for max. 3 ml) against HPLC-water (VWR). 500 µg protein aliquots were pipetted into 1.5 ml tubes, frozen on dry ice, and lyophilized for 2 hr using an Eppendorf concentrator (Eppendorf). Lyophilized samples were stored at −80°C until use.

30Fibrils were prepared by dissolving dialyzed and lyophilized, recombinant α-Syn protein at 5 mg/mL in incubation buffer (DPBS, Gibco; 2.66 mM KCL, 1.47 mM KH2PO4, 137.93 mM NaCl, 8.06 mM Na2HPO4-7H2O pH 7.0–7.3). Reactions of 200 µL per tube were incubated at 37°C with constant agitation (1,000 rpm) in an orbital mixer (Eppendorf). Reactions were stopped after 5 days, sonicated (5 min in a Branson 2510 water bath), aliquoted, and stored at −80°C until use. The presence of amyloid fibrils was confirmed by thioflavin T fluorimetry and high molecular weight assemblies were visualized by gel electrophoresis.

31Cryo-EM grids were prepared using a Vitrobot Mark IV (ThermoFisher Scientific) with 95% humidity at 4°C. Amyloid fibrils (3 µL aliquots) were applied onto glow-discharged, 300 mesh, copper Quantifoil grids. After blotting, grids were plunge frozen in liquid ethane cooled by liquid nitrogen. Samples were imaged on a Titan Krios (ThermoFisher Scientific) transmission electron microscope, operated at 300 kV and equipped with a Gatan Quantum-LS imaging energy filter (GIF, 20 eV energy loss window; Gatan Inc.). Images were acquired on a K2 Summit electron counting direct detection camera (Gatan Inc.) in dose fractionation mode (50 frames) using the Serial EM software at a magnification of 165,000× (physical pixel size 0.831 Å) and a total dose of ~69 electrons per square angstrom (e-/Å2) for each micrograph. Micrographs were drift-corrected and dose-weighted using MotionCor2 through the Focus interface. Additional data collection parameters are detailed in Table 1.

32Helical reconstruction was carried out with the RELION 2.1 software, using methods described in. Filaments were manually selected using the helix picker in RELION 2.1. Filament segments were extracted using a box size of 280 pixels (233 Å) and an inter-box distance of 28 pixels. A total of 18,860 segments were extracted from 792 fibrils manually picked from 118 micrographs (Table 1). 2D classification was carried out with a regularization value of T = 10, and 2D class averages with a clear separation of β-strands were selected for further data processing. Power spectra of 2D class averages show the layer line at 1/ (4.9 Å) with peak intensities on both sides of the meridian (Bessel order n = 1). This is the result of an approximate 21 screw symmetry between α-Syn subunits on the two protofilaments (Figure 1—figure supplement 2). Segments assigned to the best 2D classes were used for 3D classification using a regularization value of T = 8 and with optimization of the helical twist and rise. For both 3D classification and refinement, a helical_z_percentage parameter of 10% was used, which defines the size of the central part of the intermediate asymmetrical reconstruction that is used to apply real-space helical symmetry. An initial reconstruction was calculated using a cylinder generated via the helix toolbox in RELION 2.1 as initial model. This reconstruction was low-pass filtered to 60 Å and employed as the initial model for a 3D classification with a single class (K = 1) and T = 20, an approach that allowed the successful reconstruction of amyloid filaments. The handedness of the reconstruction was determined by comparison with atomic force microscopy images, which showed left-coiled surface patterns for the fibrils.

33Refinement was carried out by the auto-refine procedure with optimization of helical twist and rise. This resulted in a structure with overall resolution of 3.8 Å. Post-processing with a soft-edge mask and an estimated map sharpening B-factor of −82.6 Å gave a map with a resolution of 3.4 Å (by the FSC 0.143 criterion). An estimation of local resolution was obtained using RELION 2.1 and a local-resolution-filtered map was calculated for model building and refinement.

34A model of the α-Syn(1-121) fibril was built into the Relion local resolution-filtered map using COOT, with the PDB ID 2N0A as an initial model for the early interpretation of the map. The structure helped to determine the directionality of the protein chain and facilitated the assignment of densities in the map to specific residues. However, due to the large differences between the NMR structure and our EM map, major rebuilding was necessary. The high quality of the EM map allowed us to unambiguously build residues 38–95. A comparison was also carried out between our structure and X-ray structures of α-Syn fragments 69–77 (PDB ID 4RIK), 68–78 (PDB ID 4RIL) and 47–56 (PDB ID 4ZNN; with the mutation A53T).

35The structure (10 monomers, 5 on each protofilament) was refined against the RELION local resolution-filtered map with PHENIX real space refine. Rotamer, Ramachandran restraints, and ‘NCS’ constraints were imposed, and two B-factors per residue were used during refinement. For validation, we randomized the coordinates (with a mean shift of 0.3 Å) and refined (using the same settings) against one of the refinement half-maps (half-map 1). We then calculated the FSC between that model (after refinement against half-map 1) and half-map 1, as well as the FSC between the same model and half-map 2 (against which it was not refined). The lack of large discrepancies between both FSC curves indicates no overfitting took place.