N-terminal Acetylation Stabilizes N-terminal Helicity in Lipid- and Micelle-bound α-Synuclein and Increases Its Affinity for Physiological Membranes*

Background: The functional effects of normal N-terminal acetylation of the Parkinson disease protein α-synuclein are unknown. Results: N-Acetylation stabilizes helical structure at the N terminus of membrane-bound forms of synuclein, including a novel partly helical state. Conclusion: Stabilization of helicity increases affinity for membranes similar to synaptic vesicles. Significance: In vivo N-acetylation of α-synuclein likely affects its physiological function and dysfunction. The Parkinson disease protein α-synuclein is N-terminally acetylated, but most in vitro studies have been performed using unacetylated α-synuclein. Binding to lipid membranes is considered key to the still poorly understood function of α-synuclein. We report the effects of N-terminal acetylation on α-synuclein binding to lipid vesicles of different composition and curvature and to micelles composed of the detergents β-octyl-glucoside (BOG) and SDS. In the presence of SDS, N-terminal acetylation results in a slightly increased helicity for the N-terminal ∼10 residues of the protein, likely due to the stabilization of N-terminal fraying through the formation of a helix cap motif. In the presence of BOG, a detergent used in previous isolations of helical oligomeric forms of α-synuclein, the N-terminally acetylated protein adopts a novel conformation in which the N-terminal ∼30 residues bind the detergent micelle in a partly helical conformation, whereas the remainder of the protein remains unbound and disordered. Binding of α-synuclein to lipid vesicles with high negative charge content is essentially unaffected by N-terminal acetylation irrespective of curvature, but binding to vesicles of lower negative charge content is increased, with stronger binding observed for vesicles with higher curvature. Thus, the naturally occurring N-terminally acetylated form of α-synuclein exhibits stabilized helicity at its N terminus and increased affinity for lipid vesicles similar to synaptic vesicles, a binding target of the protein in vivo. Furthermore, the novel BOG-bound state of N-terminally acetylated α-synuclein may serve as a model of partly helical membrane-bound intermediates with a role in α-synuclein function and dysfunction.

A significant limitation of in vitro structural and functional studies is the difficulty of capturing important aspects of the dynamic cellular environment of a living organism. Post-translational modifications, for example, both static and dynamic, can alter the chemical, structural, and functional properties of proteins in vivo. One such modification, which occurs with great frequency in eukaryotic organisms, is N-terminal amine acetylation. Up to 80% of mammalian and 50% of yeast proteins have been found to be N-terminally acetylated (1). Although the addition of an N-terminal acetyl group has been shown to have some important consequences in protein-protein interactions and cellular functionalities (2-7), the effects of N-terminal acetylation on protein structure have only rarely been investigated.
N-terminal acetylation of ␣-synuclein (aSyn), 2 a small (15 kDa) soluble presynaptic protein implicated in Parkinson disease (8,9), has recently received attention as a potential modulator of this protein's function and aggregation. aSyn is the major component of intra-neuronal proteinaceous aggregates known as Lewy bodies (10) that are a diagnostic hallmark of Parkinson disease, and five point mutations in the aSyn gene have been linked to familial Parkinson disease (11)(12)(13)(14)(15). The physiological function of aSyn remains to be conclusively established, but its direct interactions with lipids and synaptic vesicles in vitro and in vivo combined with reports connecting it with the regulation of synaptic vesicle exocytosis (16 -23) indicate that membrane interactions are important for aSyn function. Although aSyn has long been considered to belong to the class of intrinsically disordered proteins, recent studies have suggested that the functional form of aSyn in vivo is a native helical tetramer. All such studies involved the use of forms of aSyn that were N-terminally modified either by N-terminal acetylation (24,25) or by an N-terminal fusion-protein remnant (26). Because aSyn is known to be N-terminally acetylated in vivo (27), these studies suggested that this modification could profoundly alter the behavior of the protein.
Subsequent studies determined that N-terminal acetylation does not drastically alter the structure or behavior of aSyn in vitro or in vivo and that the protein is predominantly disordered in both the brain and other organs and tissues (28 -31). Nevertheless, N-terminal acetylation does influence the secondary structure of the protein (30,32,33). Because synuclein undergoes a disorder-to-helix transition upon binding to membranes and membrane mimetics (34,35), it might be expected that the effects of N-terminal acetylation on helical secondary structure could alter aSyn-membrane interactions. Interestingly, the nonionic detergent ␤-octyl-glucoside (BOG) figured prominently in two reports of the purification of a helical tetrameric form of N-terminally modified aSyn (25,26), suggesting an interplay between N-terminal acetylation of aSyn and its interactions with detergents, fatty acids, or lipids. Additionally, in yeast, knock-out of the N-acetyltransferase complex B gene responsible for N-terminal acetylation of aSyn results in a change from plasma membrane localization to diffuse cytosolic localization (36).
Here we investigate in greater detail the effects of N-terminal acetylation on the interactions of aSyn with lipid membranes and detergents. We observe that the conformation adopted by N-terminally acetylated aSyn (Ac-aSyn) in the presence of the common anionic detergent sodium dodecyl sulfate (SDS) is similar to that previously reported for the unmodified protein but exhibits a decrease in helix fraying at the protein N terminus. We also find that Ac-aSyn binds to nonionic BOG micelles via its N-terminal ϳ30 residues, which adopt a partly helical conformation in the bound state. This is the first direct observation of such a partly helical state of ␣-synuclein, although previous studies have indirectly detected the existence of such states. Binding to BOG is greatly reduced in the absence of N-terminal acetylation. Finally, we reconcile previously conflicting reports of the effects of N-terminal acetylation on aSyn binding to lipid vesicles by showing that this modification selectively enhances binding of aSyn to vesicles with a physiological proportion of negatively charged lipids.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-N-terminally acetylated and unmodified aSyn were produced as previously reported (30,35). Protein uniformly labeled with 15 N and 13 C, 15 N for heteronuclear NMR experiments was produced by the media swap method (37). Briefly, Escherichia coli BL21 (DE3) cells transformed with a plasmid encoding aSyn or plasmids encoding aSyn and the yeast N-acetyltransferase complex B (38) were grown in rich media at 37°C to an optical density of ϳ0.6. For production of unlabeled protein, cultures were induced with isopropyl 1-thio-␤-D-galactopyranoside and grown for 4 h at 37°C before harvesting by centrifugation. For production of 15 N-or 13 C, 15 N-labeled protein, cells were pelleted, resuspended in minimal media as a wash step, pelleted again, and then resuspended in minimal media containing 15 Nlabeled ammonium chloride or 15 N-labeled ammonium chlo-ride and 13 C-labeled glucose, respectively, then induced with isopropyl 1-thio-␤-D-galactopyranoside and grown for 3 h at 37°C. Purification of unmodified and N-terminally acetylated aSyn was the same and consisted of ammonium sulfate cuts, anion-exchange chromatography, and reversed-phase HPLC, as previously described. The purified protein was lyophilized and stored at Ϫ20°C. N-terminal acetylation was verified by mass spectrometry as well as by NMR for isotopically labeled protein.
Preparation of Protein and Lipid Samples-Lyophilized protein was dissolved in NMR buffer (100 mM NaCl, 10 mM Na 2 HPO 4 , 10% D 2 O, pH 6.8) and filtered using a 100-kDa cutoff centrifugal filter to remove large molecular weight aggregates. The protein was then mixed with NMR buffer and stock solutions of BOG (reagent grade; Amresco Inc., Solon, OH), deuterated BOG (Cambridge Isotope Laboratories, Tewksbury, MA), SDS (OmniPur, EMD Biosciences), or lipid vesicles to achieve the desired solute concentrations. All BOG and lipid vesicle stock solutions were freshly prepared, and all samples containing BOG or lipid vesicles were kept at 4°C. For spin label-containing samples, 5-doxyl-stearate (Aldrich) was dissolved in DMSO then mixed into BOG solutions before the addition of protein solutions. Lipid vesicles were prepared by drying mixtures of different lipids dissolved in chloroform (Avanti Polar Lipids, Alabaster, AL) under nitrogen and resuspending the lipid film in NMR buffer. For SUVs, the resuspended lipid mixture was immersed in a bath sonicator for 15 min in 3-min increments, then clarified by ultracentrifugation at 130,000 ϫ g for 2 h as modified from Bussell and Eliezer (39). The supernatant was used as SUV stock solution. For LUVs, the resuspended lipids were subjected to freeze-thawing 10 times, then passed through a hand-held extruder, first 21 times using filters with pore size of 400 nm and then 21 times using filters with pore size of 100 nm. The resultant solution was used as LUV stock solution.
NMR Experiments-NMR experiments were performed on 600-MHz Varian Inova, 600-MHz Bruker Avance, 800-MHz Bruker Avance, or 900-MHz Bruker Avance spectrometers equipped with cryogenic probes. Typical spectral widths for 1 H, 15 N-HSQC (heteronuclear single quantum coherence) experiments were 25 ppm in the 15 N dimension and 20 ppm in the 1 H dimension. Assignments for SDS-bound (40 mM) Ac-aSyn were transferred from SDS-bound unmodified aSyn assignments (39) with the aid of an HNCA experiment with spectral widths of 20, 28, and 30 ppm in the 1 H, 15 N, and 13 C dimensions, respectively, and 136 complex points in the 13 C dimension to achieve sufficient resolution to allow for unambiguous assignment transfer. Similarly, BOG-bound Ac-aSyn assignments were transferred from free-state Ac-aSyn assignments (30) using HNCA experiments with spectral widths of 12, 28, and 30 ppm in the 1 H, 15 N, and 13 C dimensions, respectively, and 176 complex points in the 13 C dimension at 100 mM BOG and with spectral widths of 12, 26, and 26 ppm in the 1 H, 15 N, and 13 C dimensions, respectively, and 160 complex points in the 13 C dimension at 300 mM BOG. For the latter HNCA, deuterated BOG was used in the sample to decrease detergentmediated relaxation and improve signal-to-noise. HNCA experiments provided ␣-carbon chemical shifts for secondary shift analysis. Binding to BOG and lipid vesicles for both Ac-aSyn and unmodified aSyn was followed via HSQC experiments on protein samples with increasing binding partner concentration. Proteins were dissolved in NMR buffer, and BOG or vesicle stock was added before collecting HSQC experiments. One sample was used for the complete titration of each protein with each binding partner. Concentrations of BOG and lipid vesicles were determined using the intensity of alkyl chain peaks in onedimensional proton NMR spectra. R 2 relaxation rates were measured through a series of 9 HSQC-based T 2 experiments with relaxation time delays of 10, 10, 10, 30, 30, 50, 70, 90, and 110 ms. The intensity decay of each cross-peak was fit to a single exponential decay function using internal NMRViewJ fitting module. Amide proton exchange experiments were carried out on SDS-bound samples of Ac-aSyn and unmodified aSyn at a pH of 8.4 following Ulmer et al. (40). HSQC-based experiments with and without a water-inversion Q3 pulse (5 ms) (41) applied 50 ms before the beginning of the HSQC pulse sequence were collected, and amide cross-peak intensity ratios were determined. The intensity ratio is proportional to the amide exchange rate, with protected regions retaining more signal intensity in the experiments with water inversion applied. All NMR experiments were conducted at 10°C, except those including SDS, which were conducted at 40°C; protein concentration was ϳ200 M for detergent experiments and 140 M for lipid binding experiments, whereas lipids were at 3 mM, BOG was at 100 or 300 mM, and SDS was at 40 mM. NMR data were processed with NMRpipe (42) and analyzed with NMRViewJ (43).
CD Spectroscopy-Far-UV circular dichroism experiments were performed on an AVIV Biomedical Model 410 CD spectrometer using bundled software. CD spectra from 200 to 260 nm, with a wavelength step of 1 nm and an averaging time of 1.5 s were obtained at 25°C on 100 M ␣-synuclein samples with varying concentrations of BOG dissolved in NMR buffer. The cell path length was 0.02 cm, and 4 scans for each sample were averaged, then the scan of buffer alone subtracted. The CD signal at wavelengths below 200 nm was highly variable due to increased scattering by the detergent. Titrations were fit using Equation 1 (44,45), where ⍜ 222 is the mean residue ellipticity at 222 nm, P t is the total protein concentration (set to 0.1 mM), B is the varied BOG concentration, K D is the equilibrium dissociation constant, ⍜ max is the mean residue ellipticity at full binding (set to Ϫ5060 deg-cm 2 /dmol), and ⍜ min is the mean residue ellipticity at no binding (set to Ϫ2343 deg-cm 2 /dmol). Fitting was performed using the nonlinear curve-fitting module in XMgrace and the curve_fit function in SciPy.
Lipid Binding Data Analysis-Lipid binding for unmodified and Ac-aSyn was measured using HSQC NMR experiments performed on samples of protein with and without lipid vesi-cles. The intensity ratio of cross-peaks in the lipid-free to lipidbound spectra was calculated on a residue-by-residue basis. Representative error estimates were obtained by analyzing the variability in two repeated measurements at one lipid concentration for each lipid composition. Intensity ratios were used to extract bound populations for different binding modes of aSyn, similarly to a previously published analysis (46). The population of all bound states was calculated as the ratio of the average intensity ratios of residues 3-9, which are expected to be bound in both fully and partly helical binding modes, to the average intensity ratio of residues 129 -137, which remain unbound even at high lipid concentrations, subtracted from 1. The population of the extended helix state was calculated as the ratio of the average intensity ratio of residues 65-80, which are in the second half of the lipid binding domain and have well resolved HSQC peaks, to the average intensity ratio of residues 129 -137, subtracted from 1. Apparent K D values were derived from fitting the bound populations of each state at several lipid concentrations to the following equation, derived from a simple bimolecular binding equilibrium (45), where the symbols are defined as above in Equation 1. Although the binding of synuclein to lipid vesicles cannot accurately be considered to be a bimolecular reaction, the derived apparent dissociation constants are useful as a vehicle for comparing the binding affinity of the protein to vesicles of different sizes and compositions. Fitting was performed using the nonlinear curve-fitting module in XMgrace and the curve_fit function in SciPy. Estimated errors of the intensity ratios were propagated to the final fit results.

N-terminal Acetylation Increases Helicity at the N Terminus of SDS-bound aSyn-
In the presence of spheroidal SDS micelles, the N-terminal domain of unmodified aSyn adopts a broken-helix conformation consisting of two stable antiparallel ␣-helices that are connected by a flexible non-helical linker between residues 37 and 45 and lie on the surface of the micelle with the hydrophobic face inserting slightly into the micelle surface (35,39,40,(47)(48)(49). The 1 H, 15 N-HSQC spectrum of Ac-aSyn in the presence of 40 mM SDS is similar to the corresponding spectrum of unmodified aSyn, with a few notable differences (Fig. 1A). A plot of a weighted average of the amide proton and nitrogen chemical shift changes reveals significant perturbations for the N-terminal 10 amino acids (Fig. 1B). Based on prior studies of phosphorylated aSyn (50 -52), these chemical shift perturbations extend farther up the peptide chain than would be expected for a small covalent modification such as acetylation. ␣-Carbon secondary chemical shifts in turn indicate an increase in the helicity of the N-terminal ϳ10 residues for SDS-bound Ac-aSyn ( Fig. 1C) when compared with the unmodified protein. Aside from these localized differences, the ␣-carbon secondary shifts closely resemble those of the unmodified protein, suggesting that the broken-helix confor-mation previously observed for the unmodified protein is retained upon N-terminal acetylation. To further evaluate the stability of helical structure at the N-terminal end of SDS micelle-bound Ac-aSyn, we performed amide proton exchange experiments in the presence of SDS at high pH (8.4) following Ulmer et al. (40). Amide proton exchange rates were approximated by the ratio of amide cross-peak intensity in HSQC spectra preceded by a water-selective inversion pulse to the intensity in spectra obtained without the inversion pulse. A lower relative intensity suggests faster amide proton exchange. A comparison of intensity ratios for Ac-aSyn and unmodified aSyn (Fig. 1D) shows that the pattern of amide exchange rates is quite similar throughout the protein, except for increased protection at the N terminus of the acetylated protein, extending to approximately residue 10. These data taken together suggest that N-terminal acetylation limits fraying at the N-terminal end of the first micelle-bound helix.
Ac-aSyn Binds BOG Micelles in a Novel Partly Helical Conformation-The addition of increasing concentrations of BOG to both acetylated and unmodified aSyn leads to an increase in helical secondary structure content as reported on by CD spectroscopy (Fig. 2), but the transition for Ac-aSyn occurs at lower concentrations than that for unmodified aSyn. To further compare these transitions for modified and unmodified protein, we followed chemical shift perturbations of amide cross-peaks upon the addition of increasing concentrations of BOG. Several well resolved resonances are observed to shift as well as broaden in the presence of increasing amounts of BOG (Fig. 3, A and B), suggesting intermediate-to-fast (on the NMR timescale) exchange between free and bound protein. The transition appeared largely complete at ϳ300 mM BOG for Ac-aSyn, whereas for the unmodified protein only the beginning of a transition was observed in the same concentration range. This is consistent with the CD data ( Fig. 2C), which when fit to a simple bimolecular binding equilibrium yield an apparent K D of 34 Ϯ 9 mM for Ac-aSyn and 563 Ϯ 61 mM for unmodified aSyn, indicating that N-terminal acetylation greatly enhances the interaction between aSyn and the nonionic BOG detergent. Fitting of the chemical shift perturbations was not attempted because the resonances do not begin to shift until BOG concentrations between 25 and 50 mM, likely because the interaction requires the formation of BOG micelles.
To further characterize interactions of Ac-aSyn with BOG, we obtained additional NMR data in the presence of 100 mM BOG, a concentration at which much of the protein (Ͼ70%) is expected to be bound based on the CD and NMR titrations, but binding-associated resonance broadening is not yet too severe. 1 H, 15 N-HSQC spectra under these conditions exhibit a number of resonance shifts when compared with that of the protein alone (Fig. 4A). A plot of the amide chemical shift differences shows that significant changes extend from the N terminus of the protein to approximately residue 30 (Fig. 4B), suggesting that the N-terminal ϳ30 residues of Ac-aSyn interact with BOG micelles.  FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6 ␣-Carbon secondary shifts for Ac-aSyn in the presence of 100 mM BOG (Fig. 4C) show significant positive deviations for the first 30 residues, whereas the remainder of the protein exhibits secondary shifts very similar to those previously determined for free Ac-aSyn (30). The amplitude of the N-terminal secondary shifts decreases from just above 2 ppm to just below 0.5 ppm. In  the presence of 300 mM BOG, where binding is nearly saturated, the secondary shifts increase slightly (Fig. 4C), approaching 3 ppm for the N-terminal portion of the protein, consistent with highly helical structure. The decrease in secondary shifts toward the C-terminal portion of the BOG-interacting region, however, indicates that the helical structure becomes frayed and likely quite labile. Overall, the helical content suggested by the ␣-carbon secondary shifts is reasonably consistent with that inferred from the CD data for the bound state (ϳ15% helicity, corresponding to 21 fully helical residues (53)). 15 N R 2 relaxation measurements for Ac-aSyn in the presence of 100 mM BOG (Fig. 5A) show that the N-terminal ϳ30 residues of the protein exhibit higher R 2 relaxation rates vis à vis the remainder of the protein, suggesting that these N-terminal residues experience a slower effective tumbling rate, probably as a result of binding to BOG detergent micelles. To ascertain more directly whether this or other regions of the protein are in close contact with BOG micelles, we doped the micelles (100 mM BOG) with 4 mM 5-doxyl-stearic acid, a fatty acid conjugated with a small paramagnetic moiety. Unpaired electrons increase the relaxation rates of NMR-active nuclei in a distance-dependent fashion through a process termed paramagnetic relaxation enhancement, resulting in attenuation of the NMR signals arising from nuclei that spend time in the proximity of the spin-label. Fig. 5B shows that the N-terminal 25-30 residues of Ac-aSyn exhibit significantly attenuated resonances in the presence of BOG micelles doped with spin label compared with undoped micelles, indicating that these residues likely associate with the micelles. The strong paramagnetic relaxation enhancement effect around residue 40 has been observed previously and likely arises from nonspecific interaction between the paramagnetic moiety and the aromatic residue tyrosine 39 (48).

Effect of N-Acetylation on aSyn-Membrane Binding
N-terminal Acetylation Increases aSyn Binding to Lipid Vesicles of Moderate Charge-Unmodified aSyn has been shown to bind lipid vesicles with a specific preference for highly curved   FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6 vesicles as well as those with a high proportion of negatively charged lipid headgroups (34, 54 -58). Prior studies of the effects of N-terminal acetylation on membrane binding by aSyn reported conflicting results, with one indicating an absence of any effect (30), whereas the other indicated a significant increase in binding (33). The first study employed large unilamellar vesicles with very high negative charge content, whereas the second study utilized small unilamellar vesicles with moderate negative charge content, but neither study investigated the influence of differing curvatures or charge content. To determine whether these factors play a role in the seemingly contradictory results obtained in prior studies, we examined the binding of Ac-aSyn and unmodified aSyn to lipid vesicles of different curvature (ϳ40-nm diameter SUVs produced by sonication versus ϳ120-nm diameter LUVs produced by extrusion) and charge (15% versus 50% DOPS) using NMR. Binding was assayed in a residue-specific manner by determining the intensity ratio of equivalent resonances in lipid-containing versus lipid-free samples. As previously demonstrated (44,46,59), binding leads to attenuation of individual peak intensities, as any individual residue that becomes liposome-bound in turn becomes NMR-invisible due to the large size of the resultant protein-vesicle complex.

Effect of N-Acetylation on aSyn-Membrane Binding
We selected protein and lipid concentrations (140 M protein, 3 mM lipids, 22:1 lipid:protein ratio) that resulted in partial binding of the N-terminal region of aSyn for all cases examined, permitting a residue-by-residue comparison of the extent of binding (Fig. 6). At the higher proportion of negatively charged lipids (50% DOPS), Ac-aSyn binds approximately to the same extent as unmodified aSyn to either LUVs or SUVs, with both forms exhibiting stronger binding to SUVs as expected (Fig. 6,  A and B). At the lower proportion of negatively charged lipids (15% DOPS), Ac-aSyn binds more tightly than unmodified aSyn to both LUVs and SUVs (Fig. 6, C and D). In comparison to the more highly charged vesicles, unmodified aSyn binds less tightly to both SUVs and LUVs with less negative charge, whereas the level of Ac-aSyn binding is similar to that observed for vesicles of higher negative charge content.
To further quantify the binding of modified and unmodified aSyn, we performed titrations of both proteins with each of the four abovementioned vesicle types (Fig. 7). The average peak intensity ratios for different regions of aSyn for each condition were used to calculate the bound fraction of protein (Table 1). Averages taken over the C-terminal region of the lipid binding domain reflect the fraction of protein in which that entire domain is vesicle-bound and helical (labeled "Extended helix"). Averages taken over only the most N-terminal region of the protein (labeled "All bound states"), reflect the fraction in which any portion of the protein is vesicle-bound, which may include both fully and partly helical states (46,59). The bound populations were fit to a simple bimolecular binding equilibrium to extract apparent dissociation constants (Fig. 8). The results ( Table 1) clearly demonstrate that there is little change in apparent affinity upon N-terminal acetylation for highly charged SUVs and LUVs but an ϳ3-fold increase in affinity (decrease in K D ) for moderately charged SUVs and LUVs. Somewhat surprisingly, the observed -fold increase in affinity is approximately the same for both LUVs and SUVs (2.9 versus 3.3 for all bound states and 2.3 versus 2.0 for the extended helix state).

DISCUSSION
Although the precise physiological roles of aSyn in neurons or other cell types remain unclear, it is generally accepted that membrane binding plays a requisite role in at least some functions of this protein. Potential target membranes include synaptic vesicles, mitochondrial membranes, and the nuclear envelope, and potential activities include (but are not limited to) regulating different aspects of vesicle trafficking and fusion, participating in mitochondrial homeostasis, and regulating fatty acid transport. All of these potential functions may also involve membrane remodeling by aSyn (60 -63), a process that could involve transitions between different membrane-  FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6 bound forms of the protein (64 -67). In addition to playing a critical role in aSyn function, membrane binding can facilitate aSyn aggregation, providing a potential in vivo context for initiating the aggregation cascade that ultimately leads to the formation and deposition of aSyn amyloid fibrils in Parkinson disease (46, 59, 64, 68 -70). Ultimately, a detailed understanding of synuclein-membrane interactions is likely to prove critical to elucidating both the physiology and the pathology associated with this protein.

Effect of N-Acetylation on aSyn-Membrane Binding
Endogenous aSyn in living organisms is N-terminally acetylated (27). We and others have shown that the primary consequence of N-terminal acetylation in the free state of aSyn is to increase the helicity of the N-terminal ϳ10 residues (30,32,33), an effect that likely results from the known ability of an N-acetyl group to act as a helix cap (71,72), which would stabilize the transiently helical structure formed at the N terminus. We (35,39) and others (33,73) have postulated that transient helical character at the very N-terminal region may be important in The population of all bound states was calculated as the ratio of the average intensity ratio of residues 3-9, which are expected to be bound in both fully and partly helical binding modes, to the average intensity ratio of residues 129 -137, which remain unbound even at high lipid concentrations, subtracted from 1. b The population of the extended helix state was calculated as the ratio of the average intensity ratio of residues 65-80, which are in the second half of the lipid binding domain and have well resolved HSQC peaks, to the average intensity ratio of residues 129 -137, subtracted from 1. c Apparent dissociation constants were derived from fitting the bound populations of each state at several lipid concentrations to Equation 2, derived from a simple bimolecular binding equilibrium. See "Experimental Procedures" for a further description. mediating membrane binding and concomitant folding of aSyn, and the increased helicity engendered by N-terminal capping due to N-terminal acetylation could, therefore, influence synuclein-membrane interactions. Here we show that indeed N-terminal acetylation significantly affects the binding of aSyn to membranes and detergents, with different effects in different contexts.
Early structural studies of the membrane interactions of unacetylated recombinant aSyn using the anionic detergent SDS as a membrane mimetic revealed that the protein adopts a broken-helix conformation consisting of two stable antiparallel helices comprising residues 3-37 and 45-94 separated by a non-helical linker (39,40,47,49). Our results show that this broken-helix conformation is not dramatically altered by N-terminal acetylation, although we do not directly assess the relative orientation of the two helices. However, the presence of an N-terminal acetyl group results in higher helicity and increased stability at the N terminus of the first helix in the SDS micelle-bound state. This result suggests that this helix is somewhat frayed at its N terminus in the absence of acetylation and that this fraying is reduced or eliminated upon N-terminal acetylation. The helical secondary structure of aSyn in the SDS micelle-bound state is thought to be highly similar to that in its vesicle-bound states, outside of the linker region, making it likely that N-terminal acetylation also prevents fraying at the N terminus of aSyn in its vesicle-bound states. Finally, because the broken-helix state of aSyn has been posited to be important in the function of the protein (64 -67, 74, 75), alterations in the stability of either of the two helices could significantly influence the protein's role in vivo.
The nonionic detergent BOG is frequently used in purification and reconstitution of membrane proteins for x-ray crystallography (76,77). In the case of aSyn, low concentrations (0.1%, ϳ3 mM) of BOG were used in purification protocols that reportedly yielded a helical oligomeric form of N-terminally modified recombinant protein (25,26). Our results indicate that in the presence of micellar concentrations of BOG, aSyn adopts a unique conformation in which the first ϳ30 residues of the protein are associated with the micelle in a partly helical state, whereas the remainder of the protein remains unbound and highly dynamic, much as in the free state of the protein. At present, it is not clear why BOG binding induces this novel conformation, but the smaller expected size of BOG micelles compared with previously used SDS micelles, based on an estimated hydrocarbon chain length of 11.6 Å for BOG compared with 16.8 Å for SDS (78), may play a role.
Partly helical states of aSyn have been inferred to exist in several previous studies and have been postulated to play a role in membrane-induced aSyn aggregation. An ESR study of vesicle-bound aSyn indicated the presence of conformations in which residues C-terminal to position 69 are detached from the membrane, whereas residues N-terminal to position 19 are attached (79). A subsequent NMR study of aSyn in the presence of dilute SUVs proposed a so-called "SL1" lipid binding mode in which only the N-terminal ϳ25 residues are bound to the vesicles (46,59) and suggested that such a state could facilitate aggregation. Studies of the effects of intermediate concentrations of fluorinated alcohols on synuclein structure also implicated a partly helical (in the N-terminal region) state in membrane-induced aSyn aggregation (80). The BOG-bound state characterized here shows some structural similarity to each of these previously described states, with helical structure confined to the N terminus of the protein. Our estimate of the number of fully helical residues in the BOG-bound state (ϳ21) also corresponds well to that calculated for the partly helical trifluoroethanol-induced aSyn intermediate (ϳ24) as well as to the number of membrane-bound residues in the SL1 vesicle binding mode (ϳ25). Unlike in the above studies, however, the BOG-bound state of aSyn gives rise to easily detectable NMR signals (due to the much smaller size of BOG micelles compared with SUVs) and can be studied at NMR-compatible concentrations (unlike the trifluoroethanol-induced state). Thus, the BOG-bound state represents an opportunity for further high resolution study of an elusive state of aSyn that may be particularly relevant to aSyn aggregation and pathology.
The best validated interaction partners for aSyn in vivo are synaptic vesicles, which are typically ϳ 40 nm in diameter and contain ϳ15% phosphatidylserine headgroup lipids (81), which may, however, be asymmetrically distributed between the inner and outer membrane leaflets. Intriguingly, we find that whereas N-terminal acetylation does not alter the affinity of aSyn for synthetic lipid vesicles containing a higher proportion (50 -100%) of anionic lipids, irrespective of curvature (Fig. 6, A and B, and as previously reported by Fauvet et al. (30)), tighter binding of Ac-aSyn was observed for vesicles with a more moderate, synaptic vesicle-like, anionic lipid content (Fig. 6, C and D). Thus, in vivo, N-terminal acetylation should result in a higher population of aSyn bound to its likely physiological target, the synaptic vesicle surface.
A prior report of Ac-aSyn binding to SUVs with a physiological anionic content similar to that used here (33) also observed a significant increase in affinity relative to the unacetylated protein consistent with our own results (Fig. 6D). Interestingly, this report noted a more distinct increase in binding by the first 12 residues followed by a transition extending to residue ϳ25 to the flatter profile typically observed for the unacetylated protein. Our own data do not exhibit such a distinct transition nor the transitions noted in earlier studies (46,59) of the unacetylated protein.
Instead, we observe a relatively monotonic decrease in affinity moving away from the N terminus. The reasons for this are not entirely clear, but variations in vesicle preparation protocols and size distributions may play a role. In any case, it is intriguing that the stepped profile reported by Maltsev et al. (33) bears considerable resemblance to the BOG-bound state we describe above, lending some credence to the idea that this BOG-bound state of the protein may indeed reflect a conformation populated in the presence of membranes.
The mechanisms by which N-terminal acetylation increases synuclein membrane affinity are not entirely clear but likely involve alterations to both the intrinsic helicity and electrostatic charge of the protein. It is clear that N-terminal acetylation enhances the helicity of aSyn at its N terminus, and because membrane-bound aSyn becomes highly helical, it is reasonable that N-terminal acetylation would increase the affinity of the protein for membranes. Indeed, we and others have proposed FEBRUARY 7, 2014 • VOLUME 289 • NUMBER 6

Effect of N-Acetylation on aSyn-Membrane Binding
previously that helicity at the N terminus of synuclein is likely a key mediator of early lipid interactions (35,73). On the other hand, N-terminal acetylation eliminates the primary amine group of aSyn and thereby decreases the positive charge of its N-terminal lipid binding domain by 1 at physiological pH. Importantly, merely decreasing the charge of the N terminus (by increasing the pH) does not recapitulate the effects of N-terminal acetylation on detergent binding (Fig. 9), confirming that this change alone is not sufficient to account for the effects of N-terminal acetylation. Nevertheless, elimination of the charged primary amine group would be expected to decrease electrostatic attraction to negatively charged lipid headgroups and may explain why there is no net increase in affinity for highly charged membranes, as decreased electrostatic attraction may compensate for any increase engendered by greater helicity. At lower negative charge density, however, electrostatics likely play a lesser role in the overall affinity of the protein for membranes, and increased helicity through N-terminal acetylation leads to a net increase in affinity.
Unmodified aSyn is known to bind more tightly to highly curved membranes, and our data demonstrate that this holds true for Ac-aSyn as well. The basis for curvature sensing by aSyn may include defects in highly curved membranes that have been proposed to provide preferred binding sites for amphipathic helices, including those adopted by aSyn in its membrane-bound conformation (54,57,63,82). At the same time, the charge distribution of the amphipathic N-terminal helix of aSyn may also play a key role in curvature sensing (54,58). N-terminal acetylation alters both helicity and charge at the N terminus of aSyn and could, therefore, potentially affect aSyn curvature sensing. Our data, however, suggest that the effect of N-terminal acetylation on aSyn membrane affinity is similar for both SUVs and LUVs. Nevertheless, it has been shown that curvature selectivity by aSyn cannot be comprehensively characterized through measurements of affinity alone (54), and further studies will be required to evaluate more directly the curvature selectivity of Ac-aSyn. In addition, it is now generally accepted that aSyn can influence membrane properties, specifically curvature, in addition to sensing them (57,58,60,62,63,66,75,83), and the influence of N-terminal acetylation on this activity of the protein remains to be explored.
Our results suggest that contributions from helix formation and electrostatic attraction are delicately balanced in aSynmembrane binding, and different membrane compositions and/or protein modifications may alter this balance. Such an interplay between electrostatic attraction, hydrophobic partitioning, and coil-to-helix transitions in peptide-lipid interactions has been recognized before (84). The stabilization of amphipathic helical structure at the N-terminal end of aSyn by N-terminal acetylation likely contributes to the increased affinity of the protein for membranes with a physiological negative charge density, including synaptic vesicles in neurons.

CONCLUSIONS
Because the natural form of aSyn in vivo is the N-terminally acetylated form, it is important to consider the effect of this post-translational modification on the structure and function of the protein. In its free state Ac-aSyn exhibits increased helicity at its very N terminus but is otherwise similar to the unacetylated protein. It is now clear, however, that N-terminal acetylation influences the interactions of aSyn with membranes and detergents. Fraying at the N terminus of micelle-bound and probably membrane-bound aSyn is eliminated in Ac-aSyn, stabilizing bound states of the protein. Increased N-terminal helicity also contributes to an increased affinity of Ac-aSyn for vesicles with more moderate, physiological, anionic lipid content, possibly by facilitating binding to membrane defects that may be present in such vesicles. Finally, increased helicity at the N terminus of Ac-aSyn appears to favor states such as those we observed in the presence of the nonionic detergent BOG and that may also exist in the presence of vesicles, in which the  (A and B), unmodified aSyn with SDS at pH 6.8 is in black, Ac-aSyn with SDS at pH 6.8 is in red, and unmodified aSyn with SDS at pH 8.4 is in green. Arrows highlight peak shifts in the SDS-bound spectrum upon acetylation, which are not observed upon raising the pH for the unmodified protein. In the bottom four panels (C-F), unmodified aSyn (C and E) with 100 mM BOG at pH 6.8 (black) is compared with unmodified aSyn with 100 mM BOG at pH 8.4 (green), whereas Ac-aSyn (D and F) in buffer at pH 6.8 (blue) is compared with Ac-aSyn with 100 mM BOG at pH 6.8 (red). Arrows highlight peak shifts indicative of Ac-aSyn binding to BOG micelles, which are not observed upon raising the pH for the unmodified protein. Note that some peaks disappear in the high pH spectra due to increased amide proton exchange.
N-terminal 25-30 residues of the protein are partly helical while the remainder of the protein remains disordered. The potential roles of such states both in pathological aggregation and in the physiological functions of aSyn remain to be more clearly delineated.