Interplay of α-synuclein binding and conformational switching probed by single-molecule fluorescence

Abstract

We studied the coupled binding and folding of α-synuclein, an intrinsically disordered protein linked with Parkinson's disease. Using single-molecule fluorescence resonance energy transfer and correlation methods, we directly probed protein membrane association, structural distributions, and dynamics. Results revealed an intricate energy landscape on which binding of α-synuclein to amphiphilic small molecules or membrane-like partners modulates conformational transitions between a natively unfolded state and multiple α-helical structures. α-Synuclein conformation is not continuously tunable, but instead partitions into 2 main classes of folding landscape structural minima. The switch between a broken and an extended helical structure can be triggered by changing the concentration of binding partners or by varying the curvature of the binding surfaces presented by micelles or bilayers composed of the lipid-mimetic SDS. Single-molecule experiments with lipid vesicles of various composition showed that a low fraction of negatively charged lipids, similar to that found in biological membranes, was sufficient to drive α-synuclein binding and folding, resulting here in the induction of an extended helical structure. Overall, our results imply that the 2 folded structures are preencoded by the α-synuclein amino acid sequence, and are tunable by small-molecule supramolecular states and differing membrane properties, suggesting novel control elements for biological and amyloid regulation of α-synuclein.

Alpha-synuclein, a highly acidic 140-residue protein expressed at high levels in the human brain and enriched in presynaptic nerve termini, is a member of the growing class of intrinsically disordered proteins that adopt ordered structure upon interaction with cellular partners (1–5). This natively unfolded protein plays crucial roles in the pathogenesis of several neurodegenerative disorders including Parkinson's disease and Alzheimer's disease (6–9). Although several physiological functions have been proposed for the protein, including roles in the regulation of distinct pools of presynaptic vesicles (10, 11), maintenance of SNARE protein complexes (12), modulation of neural plasticity (13), control of dopamine neurotransmission (14), and ER-Golgi trafficking (15), its precise biological role remains unclear. Nevertheless, membrane interaction is generally believed to be a key modulator of α-synuclein function (16, 17).

Sequence analysis predicts α-synuclein interaction with lipid membranes through amphipathic α-helices encoded by 7 imperfect 11-residue repeats, approximately 4 of which are located in the highly basic N-terminal region of the protein, and 3 in the highly acidic and hydrophobic NAC region (non-Aβ component of Alzheimer's disease amyloid) (13, 18). Not surprisingly, the protein undergoes structural transitions upon binding to either brain-derived or synthetic acidic phospholipid vesicles, adopting α-helical conformations in the membrane-bound form (17–19). Similarly, α-synuclein assumes helical structures upon interaction with micellar SDS, a well-characterized phospholipid mimic widely applied in α-synuclein biophysical investigations (3, 20–25).

Several ensemble biophysical methods including NMR (20, 21, 26, 27), EPR (28–30), CD (21, 22, 31, 32) and fluorescence spectroscopy (33, 34) have provided valuable insights into the structural features of disordered and folded α-synuclein. In the context of membrane-binding, solution NMR studies performed in the presence of SDS spherical micelles demonstrated that micelle-bound α-synuclein adopts a broken α-helix structure, which consists of a pair of curved antiparallel helices connected by a well-ordered extended linker, and followed by a short extended region and a predominantly unstructured mobile tail (Fig. 1A) (20, 21, 26). EPR analyses of α-synuclein derivatives bound to small unilamellar lipid vesicles (SUVs) have provided evidence for an elongated helical structure devoid of significant tertiary packing (28), or suggested a broken helical structure more recently (29, 30). To relate these reported micelle- and vesicle-bound experimentally-observed structures, we previously carried out a detailed ensemble thermodynamic characterization of the SDS-induced folding of α-synuclein in broad ranges of SDS concentration, temperature, and pH, using CD spectroscopy, providing evidence for an inherently multistate α-synuclein folding (22). However, long-range structural and dynamics information critical for the unambiguous structural assignment of the protein thermodynamic states were unavailable from our ensemble study.

Fig. 1.

α-Synuclein coupled folding and binding monitored using single-molecule FRET. (A) Representations of α-synuclein primary and SDS micelle-bound tertiary structures [1XQ8; (26)], with the N-terminal, NAC and C-terminal regions color-coded as blue, purple and red, respectively, and the dye-labeling positions indicated by green and red spheres. A mixture of 2 isomers (Cys8Donor-Cys84Acceptor and Cys8Acceptor-Cys84Donor) from the dual-cysteine labeling was used in the single-molecule experiments—for simplicity, only 1 of the 2 isomers is depicted in the figures. (B) A summary of the smFRET data on the isothermal titration of labeled α-synuclein (≈100 pM) against SDS in the presence of 20 μM WT protein. The 5 protein conformations observed are identified. (C) The population distributions as modulated by SDS concentration, determined from nonlinear least-squares fitting of individual smFRET histograms to Gaussian functions. See SI Appendix for additional details.

To better understand the complexities of α-synuclein folding, we turned to single-molecule experiments, which avoid loss of information due to ensemble averaging (35–40) and thus benefit studies of protein folding landscapes (41–43) by permitting multiple structural distributions and coexisting populations to be resolved and examined in a more straightforward and model-independent manner. Here, we use single-molecule fluorescence resonance energy transfer (smFRET) as a long-range distance ruler (44) to provide unequivocal evidence for the structural interplay between the broken and extended α-helix structures observed for the protein, induced by binding to SDS and phospholipid SUVs. In addition, correlation analysis of FRET fluctuations [FCS-FRET (45)] was carried out to quantify the structural dynamics of the different α-synuclein conformations.

Results and Discussion

smFRET Maps a Multistate Landscape for SDS-Induced α-Synuclein Folding.

smFRET was used to probe the coupled folding and binding of α-synuclein in the presence of SDS, a negatively charged lipid mimetic that is well-characterized, and widely-used for studying α-synuclein structure and function (3, 20–25). SDS provides a simple model for biological membranes, and enables the precise tuning of the properties of binding partners over a broad range of concentrations, allowing exploration of interactions with both small-molecule amphiphiles at low concentrations and membrane-like surfaces at higher concentrations. For FRET detection of α-synuclein conformational changes, the protein was dual-labeled at residue positions 7 and 84 with donor (Alexa Fluor 488) and acceptor (Alexa Fluor 594) fluorescent dyes (Fig. 1A), resulting in a mixture of Donor-Acceptor- and Acceptor-Donor-labeled proteins at the two positions. On the basis of the NMR structure of SDS spherical micelle-bound α-synuclein (26), these labeling positions flank the helical regions of the protein, and are ideal for FRET detection of global structural transitions involving long-range distance changes (20–70 Å) (36), like that between the broken and extended helical protein states (20–22, 26, 28). The smFRET data reported here are in good agreement with our previous ensemble data on wild-type α-synuclein (22), showing that dye-labeling does not introduce significant perturbations of protein properties. In the smFRET experiments, donor and acceptor fluorescence bursts were simultaneously recorded, and individual labeled protein molecules freely diffused through a small confocal detection volume (see Fig. S1) (44). FRET efficiencies (E_FRET) for individual molecules were calculated from donor and acceptor bursts, and E_FRET histograms were constructed (SI Appendix).

Isothermal protein-SDS titrations monitored by smFRET were performed using SDS concentrations that span both monomeric and micellar conditions. A 3-dimensional map of raw FRET histogram data collected in the presence of background unlabeled protein and presented as a function of SDS concentration (Fig. 1B) provides a model-free protein phase diagram. Simple visual examination identifies a total of 5 distinct protein conformations, distinguished as peaks with specific E_FRET, and resolved independent of thermodynamic models and data fitting, highlighting the strength of smFRET measurements in charting complex folding landscapes (41–43). Below the SDS critical micelle concentration (CMC), which is ≈0.9 mM in the buffer conditions used in the present study (22, 24), and where the lipid mimetic is predominantly monomeric, 3 nonzero FRET peaks were detected, consistent with the natively unfolded protein (U conformation) specifically binding monomeric SDS molecules to form 2 differently liganded conformations (I and F). Above CMC, where SDS is largely micellar, 2 additional nonzero FRET peaks were detected (I_m and F_m). The designations for the different conformations observed are the same as those we used in ref. 22, and are based on the measured relative helical contents (U < I/I_m < F/F_m) and the type of binding partners, i.e., small molecules like SDS monomers (I and F) vs. micelles or membranes (I_m and F_m).

To determine population distributions without the use of thermodynamic models (22, 46, 47), individual smFRET histograms were analyzed and fitted to Gaussian functions. The calculated fractional populations for the 5 conformations observed are displayed as a function of SDS concentration (Fig. 1C). The ease with which population distributions were determined without using complicated models or performing SDS titrations as a function of additional variables (e.g., temperature or pH) demonstrates another strength of single-molecule techniques in revealing structural energetic information in complex systems.

In our previous thermodynamic characterization of the SDS-induced folding of α-synuclein via ensemble far-UV CD measurements and protein phase diagram analyses (22), we detected 3 types of protein conformational states exhibiting different degrees of helicity (i.e., U, I-type and F-type states). Using the long-range conformational information from the current smFRET measurements, we now directly observe the coexistence and transitions between these states, and can definitively correlate the previously observed changes in helical content with striking changes in long-range protein structure. To probe and understand the different protein structures in more detail, we next analyzed the different FRET peaks and performed correlation experiments.

Structure and Dynamics of α-Synuclein Conformational States Probed Using smFRET and FCS-FRET.

Global structures of the different α-synuclein conformations detected by smFRET were characterized using E_FRET as a ruler for measuring interdye distances (see SI Appendix) (44). Because of the FRET distance dependence, protein species characterized by high E_FRET have short interdye distances, and vice versa. Representative smFRET histograms acquired in solution conditions favoring a predominant population of U, I, F or I_m, and F_m are shown in Figs. 2A and 3A, respectively. In addition, long-range conformational dynamics were obtained using correlation analysis to probe FRET fluctuations on the nanosecond to millisecond timescales (SI Appendix) (45). FCS-FRET data for U, I, F and I_m are shown in Fig. 2C and summarized in Fig. S2. Because a single FRET distance alone is not sufficient to make structural assignments, here we use our smFRET and FCS-FRET data in combination with our previous far-UV CD ensemble data (22), the NMR structure of SDS spherical micelle-bound protein (26), and the elongated helical α-synuclein structure observed by EPR (28) to structurally assign each of the observed protein conformations.

Fig. 2.

Structure and dynamics of α-synuclein conformations. (A) smFRET histograms recorded under conditions favoring predominant population of the specified protein species (i.e., for U, I, F and I_m conformations, 0, 0.25, 1.2, and 10 mM SDS and 20, 0, 20, and 20 μM unlabeled protein were used, respectively). Thin lines represent Gaussian fits to data; bold lines are the fitted peaks for the indicated individual conformations. (B) Diagram representation of suggested structures for the different conformations. (C) Correlation analyses showing rapid (sub-microsecond), intermediate (low-microsecond), and slow (> 50 μs) FRET fluctuations (yellow symbols). Green and red symbols represent pseudo cross-correlation of donor signals and cross-correlation between donor and acceptor signals, respectively. See SI Appendix for additional details.

Fig. 3.

α-Synuclein folding induced by interaction with lipid membrane mimics. α-Synuclein in the presence of very high concentration of SDS (i.e., 450 mM) exhibits low FRET efficiency (E_FRET), consistent with an extended helical F_m structure bound to cylindrical micelles (A). The addition of the cosurfactant hexanol transforms spherical SDS micelles ([SDS] = 50 mM) into a stack of flat bilayers, and induces a protein conformational switch from the high-E_FRET spherical micelle-bound I_m form (red curve) to the low-E_FRET bilayer-bound F_m species (blue curve; overlaid histogram) (B). Thin lines represent Gaussian fits to data; bold lines are the fitted peaks for the indicated conformations. See SI Appendix for additional details.

The protein state U that exists in dilute buffer conditions is known to be intrinsically disordered, exhibiting a “random coil” CD spectrum (3, 22). The U FRET histograms obtained in 0–0.1 mM SDS show a single nonzero peak, with E_FRET of ≈0.4, which is consistent with the dimensions of an unfolded α-synuclein (interdye distance ≈ 50 Å) (Fig. 2 A and B, Table S1, and SI Appendix for distance measurements and limitations). Moreover, FRET-correlation analysis revealed sub-μs fluctuations (Fig. 2C and Fig. S2), indicative of rapid interconversion between numerous conformations in a disordered ensemble, consistent with the study in ref. 27, recently observed (48) for an intrinsically unstructured yeast prion.

Binding of SDS monomers to U induces formation of the I conformation (22), characterized by high E_FRET (≈0.8) and short interdye distance (≈34 Å) (Fig. 2 A and B, Table S1). I displays fast (≪1 μs), intermediate (≈1 μs) and slow (≫1 μs) conformational fluctuations (Fig. 2C and Fig. S2). In comparison, the spherical micelle-bound I_m, shown to exhibit a broken α-helix structure (20, 21, 26), shows similarly high E_FRET (≈0.85) and short interdye distance (≈37 Å) (Fig. 2 A and B and Table S1). However, in contrast with I dynamics, I_m exhibits only intermediate-timescale conformational fluctuations (Fig. 2C and Fig. S2). The slow “breathing” fluctuations of I in the 5–100 μs time-range correlate well with a slightly lower helicity (22) relative to I_m.

F, which also binds SDS monomers (22), exhibits low E_FRET (≈0.1) and long interdye distance (≈67 Å) (Fig. 2 A and B and Table S1). Given that all histograms have a peak at zero E_FRET (i.e., “zero peak”; from molecules with photobleached, missing or otherwise nonfluorescent acceptor dye), we performed a control smFRET experiment with rapidly alternating donor and acceptor laser excitation (49) to validate the existence of the low E_FRET F species. Our data unequivocally demonstrate the presence of an F FRET-population close to but distinct from the zero peak (Fig. S3). No measurable FRET fluctuations were observed for F (Fig. 2C), although this may merely reflect the overall lack of energy transfer and hence lack of donor/acceptor signal fluctuations. Similarly, the elongated cylindrical micelle-bound F_m conformation shows low E_FRET (≈0.1) and long interdye distance (≈70 Å) (Fig. 3A and Table S1).

For more straightforward smFRET measurements, experiments were performed in solution conditions where α-synuclein does not readily aggregate, as shown in ref. 22. However, even with negligible protein–protein interactions, α-synuclein concentration still affects what is observed by smFRET due mainly to competitive protein binding with the monomer and micelle forms of SDS, and the very dilute FRET-labeled samples (100 pM) used to ensure single-molecule experimental conditions. In the absence of background protein, for example, because the monomer to micelle SDS transition is not infinitely cooperative, protein binding to SDS micelles was observed well below the CMC (Fig. S4 A and B). The addition of 20 μM background unlabeled protein extends the range of SDS concentrations where interaction of labeled α-synuclein to monomeric SDS can be observed, free of micelle binding (Fig. 1B). We note that the FRET characteristics of the populated conformations are unchanged with (Fig. 1B, 20 μM [Protein] and Fig. S5A and B, variable [protein]) or without (Fig. S4 A and B) background unlabeled protein. Additionally, we note that a major contributor of the peak broadening observed in the histograms is shot noise in addition to other experimental contributions that are difficult to fully quantify (44), including a potential contribution from differential dye properties in the mixture of labeling isomers (D-A and A-D); hence, we instead use FCS-FRET measurements to evaluate the presence of additional conformational distributions (see SI Appendix for general discussion of peak broadening and comparisons of histogram data with shot noise simulations). Thus, although we have denoted the observed peaks as conformations, consistent with the thermodynamics of the system (22), we note that each of these “conformations” represents an ensemble, borne out by our FCS-FRET fluctuation data and previous data from time-resolved FRET and tertiary contact dynamics measurements (33, 50).

On the basis of the smFRET and FCS-FRET protein-SDS titration data described above, and our previous ensemble data (22), the small-molecule-interacting I and spherical micelle-bound I_m conformations are both structurally and thermodynamically alike. That is, similar to I_m (26), I exhibits a broken α-helix structure, although these 2 I-type forms exhibit different dynamic signatures. Likewise, the remarkable similarities between the properties of monomer SDS-bound F and cylindrical micelle-bound F_m protein species suggest that these F-type conformations also have very similar structures. Specifically, the measured E_FRET values and the high helical contents of the F-type conformations (22) show that this protein ensemble exhibits an elongated helix structure. Furthermore, because the I-type and F-type structures exist independent of the presence of stabilizing hydrophilic-hydrophobic interfaces, we conclude that these structures are properties intrinsic to α-synuclein, encoded by the protein's amino acid sequence, thermodynamically linked and in equilibrium with each other in the presence and absence of stabilizing surfaces, with the population distributions modulated by the protein's environment.

Membrane Curvature Can Modulate α-Synuclein Global Structure.

To explain the observed conformational switch from I_m to F_m, we hypothesized that low-curvature surfaces in cylindrical SDS micelles formed in high SDS concentrations bind the protein and promote an elongated helical F_m conformation. To test this hypothesis, we next probed the effects of interaction with another low-curvature SDS surface in the form of a stack of flat SDS bilayers, transformed from a spherical micelle solution using a cosurfactant (hexanol) that counterbalances the curvature induced by SDS molecules (51). The corresponding FRET histogram (Fig. 3B) is characterized by a nonzero peak with E_FRET of ≈0.2, which corresponds to ≈61 Å interdye distance if I_m-like or F_m-like dye properties are assumed (Table S1). Therefore, free from curvature-induced structural strain (26), the F-type elongated helix structure is the preferred conformation of α-synuclein bound to SDS surfaces. Hence, changes in membrane curvature serve to effect conformational switching between I_m and F_m. Our observation of an extended helical F_m structure for α-synuclein bound to low-curvature SDS surfaces is consistent with the recent EPR results published in ref. 52, wherein double electron-electron resonance was used to probe the conformation of dual spin-labeled α-synuclein variants bound to cylindrical SDS micelles, and other low-curvature membrane structures, providing evidence for an extended helix structure for surface-bound protein. Interestingly, the authors point out indications in their data for the presence of additional α-synuclein conformations, consistent with our previous ensemble results (22). The interplay between different α-synuclein conformations is now unequivocally demonstrated by our current single-molecule data.

Overall, we have shown that in the absence of interacting surfaces, the α-synuclein structure in solution is an equilibrium between the 3 conformations U, I and F (Figs. 1 B and C and 2 A and C). Ligand binding to U induces I and/or F, with the I → F and F → I transitions controlled by ligand properties and concentration. Similarly, surface-bound α-synuclein may adopt I_m and/or F_m conformations (Fig. 3 A and B). Together, our results suggest that this I-type ↔ F-type conformational switching is a general property of α-synuclein (see Fig. 5), applicable not only to the SDS model system but also to other ligand and membrane systems.

Fig. 5.

Model for α-synuclein folding and conformational switching induced by binding to small-molecule ligands and stabilizing surfaces.

smFRET Detects an Extended Structure for Lipid Vesicle-Bound α-Synuclein.

Subsequently, the structural propensities of α-synuclein bound to phospholipid membranes (Fig. 4A) were investigated. Shown in Fig. 4 B–D, respectively, are FRET histograms of the protein bound to POPS SUVs of different sizes (Fig. 4B) and POPA/POPC SUVs of different compositions (Fig. 4 C and D), i.e., using a high fraction (50%) of the negatively charged lipid POPA or using a lower fraction (≈4%), similar to that found in biological membranes (18). In all of the experimental conditions used, α-synuclein mainly exists in low-E_FRET F_m conformations. For the protein bound to POPS SUVs, the F_m structure is the only detectable conformation, exhibiting low E_FRET and sharp FRET peaks (Fig. 4B). In the presence of POPA/POPC SUVs, the obtained FRET histograms (Fig. 4 C and D) are consistent with a low-E_FRET F_m species in equilibrium with the intrinsically disordered higher-E_FRET U state, i.e., an equilibrium between bound and unbound forms.

Fig. 4.

α-Synuclein folding induced by interaction with phospholipid small unilamellar vesicles (SUV). (A) Illustration of an α-synuclein molecule bound to a phospholipid SUV. Binding of α-synuclein to lipid SUVs composed of 100% POPS (B), and molar ratios of 1:1 (C) and 1:25 (D) POPA:POPC drives folding of the protein into low-E_FRET structures. The total lipid concentration used was kept constant at 250 μM. Data using ≈30-, ≈50-, ≈100- and ≈400-nm POPS SUVs are presented in B as blue, green, red, and black lines, respectively. For the POPA/POPC SUV data shown in C and D, the thin lines represent Gaussian fits to data, and the bold lines are the fitted peaks for the indicated conformations. See SI Appendix for additional details.

The observation of the F-type conformation as the dominant ordered state for the vesicle-bound protein (using POPS and POPA/POPC SUVs) provides direct corroboration of an elongated helical structure for α-synuclein bound to POPS/POPC SUVs proposed based on EPR data (28), and is also consistent with recent EPR results for lipid bilayer-bound α-synuclein (52). Other recently published EPR results, however, suggested that α-synuclein bound to POPS SUVs (30) and POPG SUVs (29) can adopt the I_m broken α-helix structure (20, 21, 26). There are apparent disagreements between these results, i.e., if one ignores the differences in experimental conditions such as membrane composition and temperature [e.g., low temperatures can cause protein cold denaturation and induce F_m to I_m transition (22)], and, more importantly, assumes that α-synuclein only adopts a single vesicle-bound conformation. Our view, however, on the basis of the single-molecule data presented here is that the elongated helix (F-type) and broken helix (I-type) structures are both physiologically relevant and are inherent properties of α-synuclein, in the presence and absence of interacting amphipathic interfaces (Fig. 5). A survey of several reported far-UV CD data on vesicle-induced folding of α-synuclein (21, 31, 32) in the context of the expected spectral signatures of the I-type and F-type conformations (22) provides further evidence that both broken and elongated helical conformations are indeed populated in lipid vesicles, with the dominant structure determined by variables such as lipid composition, membrane size and shape, and solution conditions like temperature, pH and ionic strength. Conformational switching between the two structures in vivo could significantly affect protein function and dysfunction.

Another important aspect of α-synuclein biology is the relatively weak association of the protein with synaptic membranes, i.e., in vivo, a major fraction of the protein exists in soluble form (13). We showed here that by binding to small-molecule ligands, α-synuclein conformational switching also occurs in its soluble form. Therefore, we expect that α-synuclein conformational control is not a monopoly of lipid membranes and vesicles, but a role probably shared with other nonmembrane cellular partners.

Herein, we have directly observed how intermolecular interactions can modulate a complex equilibrium between several folded α-synuclein species. Our observations support a model where the α-synuclein sequence loosely encodes the ability to adopt two competing helical structures. Conformational partitioning can be controlled by binding to stabilizing surfaces and small-molecule ligands. Thus, regulation of α-synuclein conformational switching by both membranes and different assembly states of small molecules could very well be relevant to the protein's cellular function and aggregation. Moreover, because different conformational states of the protein may be linked to differential aggregation propensities (22), the supramolecular chemistry of antiamyloid compounds should be viewed as an important tunable element in their design. The ability to construct direct single-molecule dynamic structural maps of complex folding landscapes should prove to be of broad utility for understanding and modulating the roles of other intrinsically disordered proteins in biology and disease.

Materials and Methods

Protein expression and purification of WT α-synuclein and its variants were performed using published protocols (1, 22). For smFRET and FCS-FRET measurements, the G7,84C mutant of α-synuclein was used, dye-labeled with Alexa Fluor 488 (donor) and Alexa Fluor 594 (acceptor). smFRET and FCS-FRET experiments were performed using instrument setups described in detail in SI Appendix and Figs. S1 and S2. Determination of interdye distances was achieved by measuring relative fluorescence quantum yields of donor- and acceptor-labeled proteins, γ correction factors, and Förster distances as a function of protein conformation, as described in SI Appendix. FCS-FRET correlation analyses were carried out as described by Torres and Levitus (45).

smFRET SDS titration measurements were performed in 0.2 M NaCl, 10 mM sodium acetate, 10 mM NaH₂PO₄ and 10 mM glycine, pH 7.50 ± 0.05 at room temperature, in the presence and absence of 20 μM WT α-synuclein, using ≈100 pM dual-labeled protein. FCS-FRET experiments were carried out using 10 nM labeled protein in the same buffer conditions. See SI Appendix for more experimental details.