Structural Organization of α-Synuclein Fibrils Studied by Site-directed Spin Labeling*

Despite its importance in Parkinson's disease, a detailed understanding of the structure and mechanism of α-synuclein fibril formation remains elusive. In this study, we used site-directed spin labeling and electron paramagnetic resonance spectroscopy to study the structural features of monomeric and fibrillar α-synuclein. Our results indicate that monomeric α-synuclein, in solution, has a highly dynamic structure, in agreement with the notion that α-synuclein is a natively unfolded protein. In contrast, fibrillar aggregates of α-synuclein exhibit a distinct domain organization. Our data identify a highly ordered and specifically folded central core region of ∼70 amino acids, whereas the N terminus is structurally more heterogeneous and the C terminus (∼40 amino acids) is completely unfolded. Interestingly, the central core region of α-synuclein exhibits several features reminiscent of those observed in the core region of fibrillar Alzheimer's amyloid β peptide, including an in-register parallel structure. Although the lengths of the respective core regions differ, fibrils from different amyloid proteins nevertheless appear to be able to take up highly similar, and possibly conserved, structures.

The deposition of amyloid fibrils has been linked to a variety of slow-onset degenerative diseases, such as Alzheimer's disease, Parkinson's disease (PD), 1 type II diabetes mellitus, and spongiform encephalopathies. In the case of PD, these amyloid fibrils are formed by ␣-synuclein (1), which is found in intracellular inclusions of dopaminergic neurons, called Lewy bodies. Several lines of evidence suggest that ␣-synuclein plays a causative role in PD. Linkage studies have shown that rare familial forms of early-onset PD, inherited as an autosomal dominant form, result from two independent missense mutations (A53T and A30P) within the ␣-synuclein gene (2,3).
Moreover, experimental fly (4) and other animal (5, 6) models have suggested an important role of ␣-synuclein in the etiology of PD.
␣-Synuclein is a 140-amino acid, thermally stable (7,8), cytoplasmic protein, found in presynaptic terminals of neuronal cells (9,10). Although the precise physiological role of this protein is not fully understood, it has been suggested that ␣-synuclein is involved in the modulation of neurotransmitter release (11). In aqueous solution, ␣-synuclein has been found to be highly dynamic and, therefore, has been classified as a natively unfolded protein (8). However, upon aggregation, some not yet well-defined regions of ␣-synuclein undergo a conformational change and take up a cross-␤-structure (12,13), in which individual ␤-strands run perpendicular to the fiber axis.
Originally, a 35-amino acid fragment of ␣-synuclein was isolated from brain tissue of Alzheimer patients (14). This fragment (amino acids 61-95), designated NAC (non-amyloid ␤ (A␤) component of Alzheimer's disease amyloid), comprises the hydrophobic core of the protein and has been shown to be important in fibril formation (15). More recent analysis suggests, however, that the core domain of ␣-synuclein might be even shorter. In the latter study, a 12-amino acid peptide (residues 71-82), located within the NAC region, was shown to be capable of undergoing self-aggregation (16). The importance of this hydrophobic stretch is further supported by its absence in ␤-synuclein, a homologue of ␣-synuclein (17), with strongly reduced propensity for fibril formation (12,18). In contrast, a third line of evidence, based on protease digestion studies (19), suggests that the core region of ␣-synuclein is longer. In this study, a 7-kDa fragment (comprising residues 31-109) was shown to be protected from proteinase K digestion. This region contains the putative 12-residue core domain, as well as the NAC region.
In an effort to determine which of these core regions potentially correspond to the cross-␤-structure and to decipher the overall structural features of ␣-synuclein fibrils, we used sitedirected spin labeling (SDSL), together with electron paramagnetic resonance (EPR) spectroscopy. Although SDSL had originally been developed to study the structural and conformational dynamics of soluble and membrane proteins (20), this method has recently been applied successfully to structural and dynamic studies of other amyloid proteins, such as A␤ (21) and transthyretin (22). SDSL is based upon the introduction of a cysteine-specific nitroxide spin label, which serves as a reporter molecule, into selected sites of the protein to generate the side chain R1 (Fig. 1). The structural environment of the spin label can then be monitored by EPR spectroscopy. The mobility information contained in the EPR spectra can be used to distinguish between loop, surface, or buried sites (23)(24)(25). In nitroxide scanning experiments, this information * This work was supported by the Hillblom Foundation (to R. L. and J. C.) and the Arnold and Mabel Beckman Foundation (to R. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  can furthermore be used to determine secondary structural elements (20). In addition, the distance between two R1 side chains can be estimated based on magnetic dipolar interactions (20).
Our findings show that ␣-synuclein, in its fibrillar state, has a distinct domain organization. We identified a core region of ϳ70 amino acids (residues 34 -101), packed in highly ordered, parallel fashion, with the same residues from different strands in exact register. Similar parallelism was also observed in the core regions of synthetic A␤ fibrils assembled in vitro using SDSL (21). The N terminus of fibrillar ␣-synuclein is structurally more heterogeneous, displaying a less ordered structural organization. The C-terminal region of the protein remains unfolded, even within fibrillar bundles. Taken together, these findings suggest that there may be a common mechanism underlying the formation of amyloid fibrils, at least in Alzheimer's disease and PD.

EXPERIMENTAL PROCEDURES
␣-Synuclein Mutagenesis-The wild-type human ␣-synuclein expression construct (pRK172) was kindly provided by Dr. M. Goedert. Wildtype ␣-synuclein sequence contains no cysteine residues. Single cysteine mutants for spin labeling were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA). The introduced mutations were verified by DNA sequencing.
␣-Synuclein Expression and Purification-Escherichia coli BL21(DE3)pLys-S cells were transformed with the pRK172 construct by heat shock. Expression of ␣-synuclein was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside at 25°C overnight. Bacterial cells were harvested by centrifugation at 4,000 ϫ g for 10 min and lysed in 500 mM NaCl, 100 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, and 1 mM phenylmethylsulfonyl fluoride. The cell lysate was then boiled for 10 min and subsequently centrifuged for 30 min at 13,000 ϫ g. The resulting supernatant was precipitated in acid at pH 3.5 and centrifuged for an additional 30 min at 15,000 ϫ g. The supernatant was dialyzed against 20 mM Tris-HCl, pH 8.0, 1 mM dithiothreitol, 1 mM EDTA, pH 8.0, and loaded onto a HiTrap ANX column (Amersham Biosciences) equilibrated in the same buffer. Proteins were eluted in a 0 -1 M NaCl gradient. ␣-Synuclein-containing fractions were pooled and applied to a HiTrap Q XL column (Amersham Biosciences) and were identified by SDS-PAGE. Protein purity was Ͼ95%, as judged by Coomassie Blue staining and electrospray ionization mass spectrometry. The concentration of purified proteins was determined using the Micro BCA protein assay kit (Pierce).
Fibrillar or monomeric ␣-synuclein was loaded into sealed glass capillaries for subsequent EPR analysis. EPR spectroscopy was performed using the Bruker EMX spectrometer (Bruker Instruments, Bil-lerica, MA), fitted with a loop-gap resonator (26) for room temperature measurements and with an ER 4119HS resonator for measurements taken at Ϫ40°C. X-band room temperature EPR spectra were obtained at 2 mW incident microwave power and at a field modulation of 1.5 Gauss at 100 kHz over a scan range of 150 Gauss. Spectra were accumulated over 20 scans. All first-derivative EPR spectra were normalized by double integration to represent the same number of spins.
To correct for small amounts of free spin label and/or nonspecific labeling, all spectra were baseline corrected by subtracting the spectrum of spin-labeled (Cys-less) wild-type ␣-synuclein, which yielded a highly characteristic sharp line shape of 1.6 Gauss. This subtraction amounted to ϳ5% of the signal (data not shown).
Inter-spin distances can be quantified based upon magnetic dipolar interactions (20,(27)(28)(29)(30). Here, distances between the same R1 labels within ␣-synuclein strands were determined using a Pake patternbased simulation software generously provided by Drs. Hubbell and Altenbach (31), which has proven successful in measuring inter-spin distances within fibrils of A␤ (21). Using this software, distances between R1-labeled derivatives of ␣-synuclein were measured by comparing the EPR spectra of fully labeled protein with those co-mixed with an equal molar ratio of unlabeled wild-type protein (spin diluted), as described previously (21,31). Any residual coupling in the spin-diluted spectra were removed by subtracting the spin-coupled spectrum using the built-in option in the program.
␣-Synuclein Fibril Assembly-Purified spin labeled proteins (4 mg/ ml) were incubated in 10 mM HEPES buffer, pH 7.4, 100 mM NaCl, 0.1% NaN 3 at 37°C for 7-10 days with stirring. Fibrils were isolated by centrifugation and washed twice with 10 mM HEPES buffer, pH 7.4. Purified, unlabeled wild-type ␣-synuclein was used for spin-dilution experiments. Spin-labeled derivatives of ␣-synuclein were mixed with unlabeled wild-type protein in a 1:1 molar ratio at 4 mg/ml of total protein concentrations.

Structural Changes upon Transition from Monomeric to
Fibrillar ␣-Synuclein-To identify the structural organization of soluble and fibrillar ␣-synuclein, single spin labels were introduced at selected sites throughout the protein, and their EPR spectra were recorded in solution and in fibrils. As shown in Fig. 2, the EPR spectra for all ␣-synuclein derivatives in solution (monomeric; black traces) are very similar and are characterized by sharp and narrowly spaced lines. Such spectral features arise from the rapid motion of the R1 label and are typical of highly dynamic and unfolded structures, in close  2. A, normalized EPR spectra of soluble (black trace) and fibrillar (green trace) ␣-synuclein obtained at room temperature. R1 was introduced at selected sites in the C terminus, as indicated in the figure. The scan width is 150 Gauss. B, EPR spectra of selected ␣-synuclein derivatives taken at room temperature, in the soluble state (black trace) and in the fibrillar state (green trace). To show the spectral features of 68R1 and 88R1, the respective spectra were magnified as indicated. All spectra were normalized by double integration to the same number of spins and were recorded at 150 Gauss scan width. agreement with the notion of ␣-synuclein as a "natively unfolded" protein (8).
Next, by growing fibrils from spin-labeled ␣-synuclein derivatives, we sought to determine how this unfolded structure is altered upon fibril formation. Fibril morphology for all sites was verified by electron microscopy (data not shown) and was found to be very similar to fibrils taken from wild-type ␣-synuclein (15,32,33). Fig. 2A shows the effect of fibril formation on the EPR spectra of ␣-synuclein derivatives containing spin labels at selected sites in the C terminus (103R1, 106R1, 107R1, 124R1, 127R1, 130R1, 131R1, and 136R1; green trace). As was the case in solution, these sites gave rise to very sharp and narrowly spaced lines, even after fibril formation. Thus, we found little evidence of any ordering in the C-terminal region of ␣-synuclein aggregates and believe it is likely that the C-terminal region remains largely unfolded, even after fibril formation (also see discussion of Fig. 3 below).
To test whether structural changes might occur in the Nterminal 100 amino acids of ␣-synuclein, we recorded the room temperature EPR spectra of fibrillar ␣-synuclein labeled at positions 5R1, 26R1, 68R1, and 88R1, respectively (Fig. 2B,  green and red traces). Residues 5R1 and 26R1 showed some decrease in signal amplitude (Fig. 2B, green trace), whereas residues 68R1 and 88R1, located in the NAC region of the protein, showed a marked decrease in amplitude (Fig. 2B, green  trace). To clearly visualize details of the line shape, spectra of fibrillar 68R1 and 88R1 were scaled at 10-and 5-fold, respectively (Fig. 2B, red trace). In addition to a strong reduction of signal intensity, we also observed a pronounced spectral broadening, characterized by the overall width of the spectrum exceeding 100 Gauss (all spectra were accumulated in a magnetic scan range of 150 Gauss). These spectral features are indicative of very strong spin-spin interactions and are reminiscent of what was observed for fibrillar A␤ peptide (21), suggesting that some regions of ␣-synuclein, as in A␤, might also be arranged in parallel fashion.

␣-Synuclein Fibrils Exhibit a Core Region of Parallel and
In-register Structure-To further confirm the presence of spinspin interactions and to probe the extent of a parallel, ordered structure, we recorded the EPR spectra of the R1-labeled ␣-synuclein fibrils in their frozen state (Fig. 3).
Two factors can significantly contribute to changes in amplitude and line shape, namely, mobility and spin-spin interactions between nearby labels because of dipolar coupling and spin exchange (20). In the frozen state, EPR spectra are no longer influenced by local structure and mobility; therefore, major changes in the EPR line shapes can be attributed solely to spin-spin interactions.
EPR spectra of sites ϳ34R1 through ϳ101R1 show clear spin-spin interactions, as reflected in the marked drop in signal intensity and a spectral line broadening beyond 100 Gauss (Fig. 3, green trace). A lesser effect was seen at the N-terminal regions of fibrillar ␣-synuclein, and little or no effect was observed at C-terminal sites (103R1, 106R1, 107R1, 109R1, 124R1, 127R1, 131R1, and 136R1), which exhibited relatively large amplitudes. These EPR spectra are indicative of the absence of strong spin-spin interactions, in agreement with the data shown in Fig. 2A, indicating that this region is not ordered.
To further confirm and quantify the effect of spin-spin interactions, we performed spin-dilution experiments, wherein R1labeled and wild-type ␣-synuclein were co-mixed and allowed to form fibrils over time (see "Experimental Procedures"). Such experiments were conducted based upon the following rationale: if wild-type ␣-synuclein and its R1-labeled derivatives were to co-mix interchangeably to form fibrils, the potential of one spin label located on a given ␣-synuclein protein coming in close proximity to another spin label located on a neighboring protein would be expected to decrease as a function of increasing concentration of unlabeled wild-type protein. Thus, spinspin interactions would be reduced upon co-mixing of the R1labeled and wild-type ␣-synuclein. As a result, the EPR spectral line shapes would become sharper and narrower. Dilution experiments were performed, and once again, fibril formation was verified by electron microscopy (data not shown). As shown in Fig. 3, these wild-type dilutions resulted in highly similar EPR spectra for all of the sites (black traces). A comparison of the spectra from R1-labeled (green traces) and wildtype-diluted fibrils (black traces) shows that the largest spectral changes occurred in sites between ϳ34R1 and ϳ101R1. Within this region, spin dilution caused a significant loss of spin-spin interaction, as can be seen from the increase in signal intensity and the concomitant decrease in spectral breadth. In contrast, the EPR spectral changes in the N-and C-terminal regions of ␣-synuclein fibrils are much smaller. In addition to its importance in demonstrating spin-spin interactions, the co-mixing of wild-type and R1-labeled proteins indicates that the R1-labeled and wild-type proteins are able to adopt similar structures within the fibril. Thus, as has already been observed for Alzheimer's A␤ (21), the introduction of R1 is tolerated remarkably well in amyloid fibrils.
For a more quantitative evaluation of the spin-spin interactions, we determined the distances between spin labels, as described under "Experimental Procedures." Distances obtained by SDSL are very accurate, as reported previously in a comparison of distances obtained by x-ray crystallographic methods with those obtained by SDSL in T4 lysozyme (28). It is important to note, however, that SDSL measures distances between the nitroxide side chains of the spin labels rather than between the protein backbone atoms. A summary of distance distributions of filamentous ␣-synuclein using SDSL is shown in Fig. 4A, wherein the percentage of distance distributions is plotted against the corresponding residue number. Red columns represent distance distributions beyond 20 Å, yellow columns indicate distance distributions between 15 and 20 Å, and black columns are indicative of distance distributions of Ͻ15 Å. As depicted in this figure, the distance contributions from residues ϳ34 to ϳ101 are Ͻ15 Å (typically 8 -11 Å) (black columns), with negligible amounts of distance contributions from longer distances (red columns), which is indicative of a highly specific, in-register arrangement of ␣-synuclein amyloid fibrils. Interestingly, the degree of parallelism includes but extends beyond residues 71-82 and the amyloidogenic NAC region. The C-terminal residues of the protein primarily exhibit distance distributions of Ͼ20 Å (red columns), which is consistent with the data presented in Figs. 2 and 3, showing that the C-terminal region of ␣-synuclein is unstructured. A much more heterogeneous distance distribution is observed in the N-terminal region of ␣-synuclein. Structural heterogeneity may arise from the existence of multiple conformations as observed in other amyloid fibrils such as A␤ (21) studied by EPR and by protease digestion studies of the A␤ peptide (34). DISCUSSION The principal objective of this study was to identify the overall structural organization of ␣-synuclein fibrils. Using SDSL and EPR analysis, we observed a distinctive domain organization in ␣-synuclein fibrils. We found a core region of ϳ70 residues that is flanked by shorter N-and C-terminal domains of ϳ30 and 40 amino acids, respectively (Fig. 4B). Interestingly, the core region is characterized by strong spinspin interactions arising from the close proximity of R1 side chains (mostly, 8 -11 Å distances between nitroxide groups) of the same residues from different ␣-synuclein molecules, thus indicating the formation of highly ordered packing interactions that result in a homogeneous, parallel, and in-register structure. The core region of ␣-synuclein, identified by SDSL and EPR, contains the NAC domain, as well as the hydrophobic 12-residue peptide 71 VTGVTAVAQKTV 82 , which was found to be necessary and sufficient for fibril formation (16). However, our data demonstrate that the core region of ␣-synuclein fibrils extends beyond these regions to a significant extent, a finding that is further supported by a proteinase K-resistant fragment (residues 31-109) identified recently by Miake et al. (19).
It is important to note that a parallel arrangement of amyloid fibrils is not limited to ␣-synuclein alone. In fact, a number of studies on Alzheimer's A␤ peptide by SDSL (21) and solidstate NMR (35) have revealed that this peptide is also packed in a parallel and highly specific manner. Studies involving many different amyloid fibril-forming proteins have revealed an overwhelming number of similarities in the mechanism(s) of fibril formation. For example, amyloid fibrillogenesis consistently appears to be a nucleation-dependent process (36) that ultimately results in a cross-␤ structure, where individual strands run perpendicular to the fiber axis (37). Moreover, all amyloid fibrils exhibit specific tinctorial properties, i.e. they bind Congo red and Thioflavin S (38). Common structural features of amyloids are furthermore supported by conformational antibodies that appear to recognize a common amyloid fibril fold that seems to be shared by a number of different amyloid fibrils (39). This overall structural similarity might even extend to prefibrillar structures, considering that another conformationally specific antibody recognizes protofibrillar aggregates from various amyloidogenic proteins, including ␣-synuclein (40). Thus, in light of these similarities, it is possible that parallel arrangement is a general feature of the core regions of amyloid fibrils.
In contrast to this possibly conserved and highly organized structural arrangement in the core region of ␣-synuclein fibrils, we find little evidence of any ordered structure in the C terminus of ␣-synuclein fibrils. It is therefore unlikely that the C terminus contributes to the stability of fibrils. In fact, it is conceivable that the high negative charge density of the C terminus could cause significant electrostatic repulsions, a feature that may serve to prevent a further extension of the parallel-arranged core region beyond residue 101. A fibril-de- stabilizing role of the C terminus is further supported by kinetic analysis that has shown a significant enhancement in fibril formation upon removal of the C-terminal regions (12,41). Although it is unclear if removal of this destabilizing region contributes to enhanced fibril formation in vivo, it is interesting to note that partially truncated ␣-synuclein at the C terminus has been isolated from Lewy bodies of patients suffering from dementia with Lewy bodies (42).
Identification of a distinct domain organization in ␣-synuclein fibrils, together with the finding of strong parallelism in the core region, constitutes a first step toward understanding the mechanism of fibrillogenesis and places constraints on possible structural models. Additional structural constraints can be obtained from simple geometrical considerations. For example, we can eliminate the possibility of parallel, fully extended strands containing 70 amino acids. Given the 3.5-Å axial distance between adjacent amino acids in an extended ␤-strand, the parallel polypeptide core region of ␣-synuclein would be ϳ250 Å (25 nm) in length; however, this length exceeds the narrowest fiber dimension (6 nm) reported for ␣-synuclein fibrils (12). Thus, several turn and bend regions must exist to account for such geometrical dimensions (21,43).
The existence of hairpins consisting of anti-parallel ␤-strands has been proposed for ␣-synuclein (13,44) and A␤ fibrils (45,46), based upon Fourier Transform Infrared spectroscopy analysis of amyloid fibrils and those of soluble proteins containing such antiparallel strands. A general model for a two-stranded hairpin with anti-parallel ␤-strands is shown in Fig. 5A, where the interstrand distance is ϳ5 Å, the intersheet distance is estimated at ϳ10 Å, and the ␤-strands run perpendicular to the fiber axis (37). To observe parallelism, one might envision a parallel arrangement of individual hairpins in which the same residues will be at ϳ10 Å distance from each other (Fig. 5A, red marks). Such a structure could be taken up entirely by shorter peptides, such as NAC or the A␤ peptide. However, for the much longer core region of ␣-synuclein, this arrangement would not be likely because additional strands (probably at least four) and bend regions are required to accommodate ϳ70 amino acids under the constraints of the fiber dimensions (see Fig. 5B). The resulting distances between the same residues in different molecules, repeated along the fibril axis (Ն25 Å), would be outside the detectable range using the current approach and clearly too long to account for the strong spin-spin interaction observed from the EPR data presented herein. Nevertheless, it is still possible to account for parallelism if the sheets are arranged in parallel fashion, as depicted in Fig. 5C. Although we cannot exclude such a model based on the EPR data alone, this model could not readily explain the growth and propagation properties of amyloid fibrils. As an example, the contact surface for growth in the fibril direction would be limited to only a single ␤-strand region, as schematically illustrated in Fig. 5D. Furthermore, if the parallel sheet packing is stabilizing, one might expect greater variations in the number of stacked sheets; in turn, this could result in highly irregular fibril thickness, a feature not generally observed in fibrillar assemblies.
Problems of this nature would not be encountered with the model shown in Fig. 5E. This model is based on the existence of parallel strands that are interrupted by a number of regions that cause bending of the entire sheet. The model shown in Fig.  5E represents one of several possible organizations to which such bent regions could be achieved. One positive feature of this model is that the fibril propagation can be easily explained, i.e. both fibril ends would expose an extensive surface of hydrogen bond donors and acceptors that could act as templates for incoming molecules (see Fig. 5F). Such "template-guided folding" would further ensure that precise conformation and fibril type would be maintained. Additional support for such a model is derived from a recent solid-state NMR analysis of A␤ 1-40 peptide (47), which showed that individual strands, rather than sheets of hairpins, are arranged in parallel.
Through future studies using SDSL and EPR, we hope to prove the validity of the model in Fig. 5E (and variations thereof, possibly including some ␤-helical structures) and to identify the location(s) of the bend or turn regions. Similarly, this approach should also enable a more detailed structural analysis of prefibrillar intermediates and their membranebound forms.