Three Long-Range Distance Constraints and an Approach Towards a Model for the α-Synuclein-Fibril Fold

Abstract

Amyloid fibrils and plaques are the hallmark of neurodegenerative diseases. In Parkinson’s disease, plaques (Lewy bodies) consist predominantly of the α-synuclein (αS) protein. To understand aggregation, the molecular architecture of αS fibrils needs to be known. Here, we determine nm-distance constraints for the protein in the fibril by double electron–electron paramagnetic resonance (DEER) on doubly spin-labeled αS variants, diamagnetically diluted with wild-type αS to suppress intermolecular interactions. Intramolecular distances in three pairs (56/69, 56/90 and 69/90) are reported. An approach to derive a model for the fibril fold from sparse distance data assuming only parallel β-sheets is described. Using the distances obtained in this study as input, a model is obtained with three strands, comprising residues 56–90, in which the strands consist of 8–12 residues each. Limitations of the approach are discussed in detail, showing that the interpretation of the data does not yet yield an unambiguous structure model. Possible avenues to improve this situation are described.

Appendix

As described in the main text, an approach to interpret scarce distance data in terms of a simplified fibril architecture is useful to guide further experiments. The approach assumes an in-register configuration of the fibrils, parallel β-sheets (Fig. 1a, b) and considers the protein to be confined to a plane perpendicular to the fibril axis. The β-sheets of the fibril are perpendicular to that plane, so when viewed along the fibril axis as in Fig. 1b, c they appear as a set of parallels, separated by the inter-sheet distance (d). The distances measured between the residues 56, 69 and 90 define the relative positions of these residues as the corners of a triangle in the plane containing the protein. To perform an unbiased search for the location of the residues 56, 69 and 90 on such an idealized model fibril, an orientation is sought where the positions of each of the three residues is as close as possible to a particular one of these parallels by rotating the triangle with respect to the parallels.

Since, by DEER, the distance between the NO-groups of the pyrrolidine rings is measured, we need to relate the positions of the nitroxide groups to the positions of the C_β-atoms. The linker between the nitroxide and the C_β-atom comprises four single bonds, length approximately 0.5 nm, making the linker flexible. Programs to model the linker orientation [45–47] require atomic-level information on the structure, which for αS fibrils is not available. Therefore, we constructed a geometric model that considers sets of different linker orientations at the three positions 56, 69 and 90, thus providing an unbiased approach to eliminate the effect or at least derive worst-case scenarios for the influence of the orientation of the spin label linker on the structure.

First, we describe the procedure to derive a fibril model for the case where all linkers point in the same direction (parallel linkers) and below, we will discuss the effect of other linker orientations. Residue 56 is taken to be at the origin of the axis system used, and the triangle is rotated about this origin. Several orientations are found. Arrangements in which the residues are not sequential on successive parallels are excluded on the grounds that such an arrangement would result in too many turns of the protein given the number of intervening residues. That leaves four possible orientations. Two of these solutions involve same-strand locations for the pairs 56 and 69 or 69 and 90 and are therefore excluded (see above).

In the remaining two solutions the residues are at the positions marked by black squares in Fig. 1c. To generate a model for the fold of the protein given the positions of the residues 56, 69 and 90, the protein sequence is threaded through these points. The protein follows the direction of the β-sheet and subsequent residues within a strand are considered to be separated by 0.35 nm (l). Turns are placed at sequence positions such that they optimally complement opposing strands. Turns are considered to be made up of three residues, where residue i is the last residue in the preceding β-strand, and residue i + 4 is the first residue of the subsequent strand that is part of the opposite β-sheet. There are two ways of threading the protein through the three squares shown in Fig. 1c: starting at residue 56, the protein sequence can be threaded on strand I in the direction away from the position of 69, i.e., in the direction of the red arrow in Fig. 1c. The other possibility is to thread the protein from position 56 towards 69, i.e., in the direction shown by the red arrow in Fig. 8a. In the latter case, residues 70–74 would be needed to complement strand III. A turn involving residues 74–78 and a two-residue strand followed by a turn at 80–84 would allow to cover the distance to position 90, leaving a stretch of residues from 86–90 uncomplemented by an opposing β-sheet. This scenario, which is illustrated in Fig. 8a, seems unlikely, and is therefore discarded. Consequently, the model shown in Fig. 1c is the most likely fold amongst the solutions found. It consists of four zippered β-sheets, (strands II–V, Fig. 1c) requiring that intermediate sheets (e.g., strand III) would form ‘dry interfaces’ [1] at both sides of the sheet. The prediction of turns is somewhat arbitrary, but assuming similar lengths of β-strands, turns at residues 58–62 (residue i and residue i + 4), 70–74, and 85–89 are plausible. Next we discuss how different linker orientations could affect this model.

Fig. 8

Models of the αS fibril illustrating alternating threading (a) and the effect of different linker orientations (b–d). Colored arrows indicate successive β sheets, and dark blue, bent lines, turns. Italic numbers residues at the start and the end of the turn (see text). a Alternative way of threading the protein for the positions shown in Fig. 1c. For details see text. b Effect of linker orientation: Black squares, all linkers parallel. Pink and orange dots give the relative positions of residues 69 and 90 for the two other linker orientations that are compatible with the fold shown in Fig. 1c. The similarity of the solutions reveal that this fold is consistent with cases in which all linkers are parallel (black squares, shown also in Fig. 1c), and in which one or both linkers at 69 and 90 are almost parallel to the strand but on the same side of the respective strand, as the linker at residue 56 (for details see text). The threading model is shown for the pink positions. Hashed turns indicate turn positions for the model shown in Fig. 1c of the main text. With respect to the model shown in Fig. 1c, strand IV is shifted by 2 residues relative to strand II, and turn positions differ by one (turn between strand II and III and two residues (turn between strand IV and V). c Fold for linker orientations with the linker at residue 56 opposite to those at residues 69 and 90 for two different linker orientations (green and purple dots). The purple dot and the square close to the green dot labeled 69 are the alternative positions of residue 69. The black square between strands IV and V shows the position of residue 90 in the model shown in Fig. 1c of the main text. d Same positions as for the model in c with alternative threading of the protein (see text)

1.1 The Influence of Different Spin Label Linker Conformations

To asses the effect of the flexible linker joining the nitroxide group to the protein backbone, a geometric model was developed which mimics different orientations of the nitroxide groups. In this model, the linker is considered as having a fixed length of 0.5 nm and it can have different orientations with respect to the triangle formed by the positions of residues 56, 69 and 90 (black squares in Fig. 1c). Three orientations are considered per residue, each 120° apart to account for the difference in position of the protein backbone and the nitroxide group. Since only relative orientations are of interest, there are nine possible combinations of linker orientations for the residues 56, 69 and 90. Besides the possibility that all linkers are parallel, five other arrangements are found. In Fig. 8b three combinations of linker orientations are shown: The black squares for parallel linker orientations, and the orange and pink dots, for which the linker direction at residue 69 has different orientations. The three-strand model prevails for residues 56 to 90, showing that the model shown in Fig. 1c is robust, but the strands are shifted parallel to each other, juxtaposing other residues than in Fig. 1c at opposite strands. So, even if the linkers at residue 69 and 90 are oriented nearly parallel to the strand (orange dots, Fig. 8b), the corresponding model of the fold is similar to the one shown in Fig. 1c, with the exception that strand IV is shifted by two residue positions with respect to strand II, and turn positions shift by maximally three residues.

If, however, the linkers at residues 69 and 90 point in a direction opposite to that of residue 56, i.e., if the linker at residue 56 points to the left of the strand in Fig. 1c, linkers at residues 69 and 90 point to the right side of their respective strand, or vice versa a different solution results (Fig. 8c, d). This solution is possible for two sets of orientations of linkers (green and purple dots, Fig. 8). The two ways of threading the protein in this case are shown in Fig. 8c, d. The main difference is that there is only one turn between residues 69 and 90. A single turn between residue 69 and 90 was suggested in [10–12, 14], however, these models predict two turns between residues 56 and 69, which is not compatible with the model shown in Figs. 1c and 8. Also, some studies [9] position residue 90 outside the β-sheet core, indicating that it could be in a turn region. We consider the folds shown in Fig. 8c and d less likely, because the overall extent of the fibrillar region seems smaller than expected from the outside dimensions of the protofibrils [10], but we note that we cannot fully exclude it.