Interactions between fatty acids and α-synuclein

α-Synuclein (αS) is an amyloidogenic neuronal protein associated with several neurodegenerative disorders. Although unstructured in solution, αS forms α-helices in the presence of negatively charged lipid surfaces. Moreover, αS was shown to interact with FAs in a manner that promotes protein aggregation. Here, we investigate whether αS has specific FA binding site(s) similar to fatty acid binding proteins (FABPs), such as the intracellular FABPs. Our NMR experiments reveal that FA addition results in i) the simultaneous loss of αS signal in both 1H and 13C spectra and ii) the appearance of a very broad FA 13C-carboxyl signal. These data exclude high-affinity binding of FA molecules to specific αS sites, as in FABPs. One possible mode of binding was revealed by electron microscopy studies of oleic acid bilayers at pH 7.8; these high-molecular-weight FA aggregates possess a net negative surface charge because they contain FA anions, and they were easily disrupted to form smaller particles in the presence of αS, indicating a direct protein-lipid interaction. We conclude that αS is not likely to act as an intracellular FA carrier. Binding to negatively charged membranes, however, appears to be an intrinsic property of αS that is most likely related to its physiological role(s) in the cell.

a-Synuclein (aS), a 140 residue cytosolic protein that is water-soluble in its native state, represents in fibrillar form the most abundant protein component of the characteristic Lewy bodies, the pathological hallmark of Parkinson's disease. Despite the pathogenic features associated with misfolded aS, its physiological structure and function have not been resolved to date.
aS is ubiquitously found in the brain, constituting z1% of the total brain protein content (1). It is localized in part to presynaptic terminals (2), where it loosely associates with synaptic vesicles (3). Several very rare missense mutations of aS (A53T, A30P, and E46K) have been observed in families with autosomal dominant, early-onset forms of Parkinson's disease (4)(5)(6). Moreover, triplication of the aS gene was recently discovered in two different families with Parkinson's disease of rather typical onset (7,8). In the vast majority of cases, however, the wild-type form of the protein is implicated in the development of Lewy bodyrelated abnormalities (9).
Protein misfolding and aggregation are increasingly recognized as the basis for numerous previously idiopathic neurodegenerative diseases in the aged population (10). For example, several intriguing parallels have been reported between the intraneuronal Lewy body deposits of aS in Parkinson's disease and the extracellular amyloid plaques of amyloid b-protein (Ab) in Alzheimer's disease (11). Moreover, the non-Ab component (NAC) of amyloid plaque found in the brains of Alzheimer patients turned out to be identical to a 35 residue peptide segment within the aS sequence (12).
One approach toward understanding the role of neuronal proteins such as aS in the pathogenesis of human neurodegenerative diseases is to study their structural organization at the various stages of fibrillization (13) and inclusion formation. A better understanding of the molecular basis of amyloidogenesis could lead to rational therapies that suppress the formation of Lewy bodies and other neurotoxic inclusions. It has been proposed that the accumulation of partially or fully soluble oligomeric forms of aS (some of which have been referred to as protofibrils), rather than the occurrence of mature amyloid fibrils, may be the principal pathogenic event responsible for neuronal dysfunction (14). In the example of Alzheimer's disease, there is evidence that soluble Ab oligomers can interfere with synaptic efficacy in the absence of protofibrils or fibrils (15).
Although the pathological formation of b-pleated sheet structures in amyloid fibrils is well documented, the physiological conformation(s) and function(s) of aS are still unresolved. In dilute aqueous solution, aS appears to be largely unfolded (16), whereas interactions with phospholipid vesicles or detergent micelles appear to induce a highly helical conformational state of the lipid-bound protein (17,18). Various biophysical studies have reported a supposed preference of aS binding to i) unilamellar phospholipid vesicles, ii) smaller vesicles with highly curved surfaces, iii) negatively charged phospholipid surfaces, iv) lipids with long polyunsaturated acyl chains as tail groups, v) phospholipids with inositol as the head group, vi) phospholipid membranes containing lipid raft-like combinations of oleic acid (OA; 18:1 D 9 ) and PUFA chains, and vii) defect structures in the membrane-water interface (17,(19)(20)(21)(22)(23). Furthermore, one of these studies indicated a tendency of the aS molecules to oligomerize upon in vitro exposure to bilayer vesicles consisting of phospholipids with only polyunsaturated acyl chains (21). Moreover, the association of aS with biological membranes has been linked to protection of the protein from oxidation or nitration, leading to a reduction of pathological fibrillization (24). Most of these studies, however, did not explain the postulated lipid binding behavior of aS on a molecular basis.
In this study, we analyzed in vitro the interactions of purified aS with several different monounsaturated and polyunsaturated fatty acids using high-resolution NMR spectroscopy and electron microscopy (EM), as previous studies (25)(26)(27) had indicated mutual effects that were dependent on the aS and FA concentrations, both in vitro and in vivo. Initially, in vitro measurements of binding interactions between aS and unesterified FAs that were based on the Lipidex assay (25) had led to discussion about a possible relationship to intracellular fatty acid binding proteins (FABPs), a family of cytosolic lipid carriers with a very characteristic b-barrel structure (28). Recent titration microcalorimetry data, on the other hand, did not reveal any specific binding of single FA molecules, either saturated or monounsaturated, to aS (29), as typically observed for FABPs. Moreover, other experiments had demonstrated that, unlike FABPs, the levels of soluble aS oligomers in the cytoplasm are directly affected by FAs (26): PUFAs enhanced the formation of aS oligomers in living mesencephalic neuronal cells, whereas saturated FAs inhibited their formation; both effects were replicated in vitro using purified recombinant aS. As a consequence, we set out to compare aS with FABPs, because the interaction of aS with FAs is apparently of major physiological significance for the role of aS inside the cell. For example, an association between aS oligomer accumulation and altered FA composition was found in the brains of patients with certain synucleinopathies (27). More precisely, specific long-chain PUFA levels were increased in the soluble brain fraction of patients with either Parkinson's disease or a related disorder, Lewy body dementia, compared with normal agematched subjects. Similarly, altered FA compositions were detected in cells that either overexpress or entirely lack aS, in contrast with the normal cells (27). Just recently, a disruption of FA uptake and trafficking was observed in vivo in astrocytes from aS gene-ablated mice (30).
With protocols designed to detect specific interactions between a particular protein and individual FA molecules by the use of NMR, we explored whether aS has high-affinity FA binding sites, similar to those revealed for FABPs and serum albumin via the same method (28,31). Instead, however, the NMR data indicated the formation of high-molecular-weight FA-aS complexes. Subsequently, EM was used to visualize the morphology of these biomolecular aggregates. Rat intestinal (I-)FABP was prepared as described previously (32), with the exception that a Hi-Load Superdex G-75 column (60 cm 3 1.6 cm), equilibrated with 50 mM sodium phosphate buffer, 300 mM NaCl, and 0.05% sodium azide at pH 7.8, was used for size exclusion.

MATERIALS AND METHODS
Purified human aS, which was kindly provided by Ronit Sharon (Brigham and Women's Hospital, Boston, MA), had been prepared as described previously (16,33). Gas chromatographic analysis showed that the protein was free of endogenous FA. For molecular weight determinations, aS protein samples (2 mg) were applied on precast SDS gels (Tris-glycine, 8-16% acrylamide gradient) from Bio-Rad and subsequently subjected to Western blot analysis using LB509 (Chemicon, Temecula, CA) as anti-aS antibody. The relative band migrations were analyzed with a standard BenchMark prestained protein ladder (Invitrogen, Carlsbad, CA) using AlphaEase 5.5 software (Alpha Innotech Corp., San Leandro, CA).
Before each NMR and EM experiment, aS samples were prepared fresh from the lyophilized material that was filtered through Microcon concentrators with a 100 kDa cutoff (Millipore, Billerica, MA) to remove protein aggregates. The FAs were added as sodium or potassium salts from concentrated stock solutions (34).

NMR experiments
A Bruker (Rheinstetten, Germany) AVANCE 500 MHz spectrometer, equipped with a 5 mm inverse triple-resonance probe that has XYZ-gradient capability, was used to carry out all NMR experiments. All NMR spectra were collected at 258C in a phase-sensitive mode, implementing time-proportional phase incrementation for quadrature detection. The 1 H and 13 C chemical shift values were referenced to external 2,2-dimethyl-2-silapentane-5-sulfonate (Cambridge Isotope Laboratories). In the one-dimensional (1D) and two-dimensional (2D) 1 H experiments, the water signal (H 2 O/D 2 O ratio of 95:5) was suppressed by selective presaturation during the relaxation delay (1.3 s), with the carrier placed in the center of the spectrum on the water resonance. 1D 1 H data (128 scans, 4 k time domain size) were collected with 7,507.5 Hz (15 ppm) spectral width. For shorter data collection during the titration experiments, the 2D 1 H/ 1 Htotal correlation spectroscopy (TOCSY; spinlock time of 75 ms) and 2D 1 H/ 1 H-NOESY (mixing time of 150 ms) spectra were acquired with 7,507.5 Hz spectral width in both time domains, but only eight scans and 2,048 3 256 data points. Regular 2D data sets, on the other hand, were collected with 64 scans and 2,048 3 512 data points. For the 1D 13 C experiments of aS (30,000 scans, 16 k time domain size), a relaxation delay of 2 s and a spectral width of 27,777.8 Hz (221 ppm) was applied. The 1D 13 C spectra of I-FABP were collected with 6 k scans, 4 k time domain size, 1 s relaxation delay, and a spectral width of 28,985.508 Hz (230 ppm). All NMR spectra were acquired, processed, and analyzed using the XWINNMR 2.6 software package (Bruker).

EM experiments
Samples of 200 ml final volume were prepared in 10 mM Tris buffer (125 mM NaCl, pH 7.8) containing i) 1 mM OA, ii) 1 mM aS, or iii) 1 mM OA plus 1 mM aS. The samples were either mixed gently or vortexed for 90 s at maximal power. To visualize the aggregate structures by EM, a modification of the negative staining drop technique (35) was used. Aliquots of 5 ml were incubated for 10 s on carbon-and Formvar-coated, 200 mesh copper grids that had been glow-discharged to promote sample spreading (36). The grids were then rinsed with 10 drops of Millipore-filtered (0.2 mm) water to reduce the salt concentration. After removal of the excess fluid by blotting, the samples were stained for 10 s with 1% sodium phosphotungstate at pH 7.5, blotted again, and airdried. All images were recorded under low-dose conditions on SO163 film in a CM12 transmission electron microscope (Philips Electron Optics, Eindhoven, The Netherlands).

NMR data
To probe the interactions of FAs with aS at a molecular level, we performed high-resolution NMR experiments at 11.7 T. Homonuclear 1D and 2D spectra of aS in aqueous buffer solution (pH 5.5) were first obtained under conditions typical for NMR-based structural analysis of FABPs (37)(38)(39). In this acidic environment, FAs bind to FABPs with high affinity while the exchange of the amide protons is suppressed, thus allowing a better detection of the highly relevant amide proton resonances by NMR. The 2D TOCSY experiments are designed to show short-and long-range 1 H couplings through chemical bonds; they form the basis for making rigorous assignments of protein spectra. Both the 1D and 2D NMR spectra of aS without FAs showed a rather low signal dispersion and hence extensive spectral overlap ( Fig. 1), as is typical for unfolded proteins. Moreover, the 1 H resonances of most aS residues displayed standard chemical shift values (in ppm), as usually observed in "random coil" structures. Hence, a quantitative analysis of the protein secondary structure according to Wishart, Sykes, and Richards (40) yielded no significant a-helix or b-sheet percentages based on these 2D NMR spectra. This result is consistent with other NMR data of aS that were derived from a heteronuclear 2D experiment ( 1 H/ 15 N-HSQC) (18) but strikingly different from data obtained for FABP family members, which exhibit extensive secondary and tertiary structure under similar experimental conditions (38,39).
Because of this apparent lack of tertiary structure in watersoluble aS, we did not proceed with the analysis of NOESY spectra (data not shown), which directly reflect atomic distances between 1 H nuclei, to derive the solution-state structure of the protein. Instead, we focused on the interactions between aS and different FA species.
We initiated our studies of FA binding to aS with the PUFA linolenic acid (LA; 18:3 D 9,12,15 ), because PUFAs have been shown to induce aS oligomerization both in vitro and in vivo (21,26). The addition of equimolar or higher amounts of LA to a 1 mM aS solution at pH 5.5 produced no chemical shift changes in either the 1D or 2D NMR spectra, which would be indicative of conformational rearrangements within the protein structure. The so-called "fingerprint region" in the 2D 1 H/ 1 H-TOCSY spectrum, which exhibits protein signals between the backbone amide and the corresponding C a protons, is particularly sensitive for detecting even small conformational changes. However, the superposed fingerprint regions of two 1 H/ 1 H-TOCSY spectra collected i) without LA and ii) with a 10fold excess of LA (i.e., final FA concentration 5 10 mM) showed no additional or shifted signals in the presence of FAs (Fig. 2). Hence, these NMR spectra in solution show that even a large molar excess of FAs causes no discernible changes in the tertiary structure of water-soluble aS. Moreover, the possibility that the protein binds FA molecules in discrete high-affinity binding sites, which should produce at least some specific alterations of the local chemical environment inside the binding pocket (as seen with FABPs), is highly unlikely because of the apparent lack of chemical shift changes.
1D 1 H-NMR spectra obtained with different FA/aS molar ratios up to 10:1 (Fig. 3A) revealed a stepwise decrease of the protein signal intensities with increasing fatty acid concentrations. The corresponding intensity changes upon addition of LA are displayed graphically for two selected aS amide proton peaks in Fig. 3B. aS solutions were also titrated with the PUFA docosahexaenoic acid (DHA; 22:6 D 4,7,10,13,16,19 ) at pH 6.0 and with monounsaturated OA at pH 7.7 to cover a wider range of experimental conditions; the spectral results (data not shown), however, were similar to those found for LA. This general decrease in the protein signal intensities with increasing FA ratios implies a reduction of the monomeric aS concentration in solution, even though no concomitant protein precipitate was observed. Moreover, no narrow FA signals, which would indicate a population of single FA molecules bound with high affinity to soluble aS, appeared with increasing FA concentrations. Instead, the results rather suggest that the presence of FAs caused aS to form high-molecular-weight complexes with the lipid, thereby reducing its molecular tumbling rate to the extent that the 1 H signals disappeared from the NMR spectrum as a result of excessive line-broadening.
Western blot analysis of the LA-aS mixture at 10:1 ratio, which was obtained from the corresponding NMR sam-ple, revealed the presence of various oligomeric aS forms, whereas the corresponding aS samples lacking FAs displayed only the monomeric protein form (Fig. 4). These data, as well as the fact that the decrease of monomeric  aS concentration in the NMR solutions is induced entirely by the addition of FA, suggest that the resulting high-molecular-weight complexes consist not solely of protein but also include lipid. Moreover, our NMR spectra do not support the existence of soluble aS dimers, whose molecular mass (28 kDa) is still far enough below the size limit at which excessive line-broadening would render signal detection impossible. Hence, assuming that the highly concentrated NMR sample we used consisted solely of high-molecular-weight FA-aS aggregates, the fairly large amount of aS dimer species observed on the Western blot may be the result of a reversible decomposition into smaller oligomer fragments as a result of FA loss on the gel or during gel preparation.
To define more precisely the physical state of the FAs in the aqueous buffer solutions used here, we next used 13 C-NMR and EM. 13 C-NMR spectroscopy with 13 C-carboxyllabeled FAs is a very sensitive and specific method for the detection of FAs bound to individual protein binding sites, as demonstrated previously in studies of intracellular FABPs and serum albumin (41,42). We now performed 13 C-NMR binding studies with nonlabeled rat I-FABP for direct comparison with aS. As shown in Fig. 5, the carbonyl resonances of I-FABP at natural 13 C abundance are near the noise level of the 1D 13 C-NMR spectra. With the addition of small amounts of 1-13 C-labeled OA at pH 7.4, a narrow signal (z30 Hz half-width) appears at 183.8 ppm. This peak represents a single OA molecule bound to I-FABP (41), because i) the concentration of unbound FA mono-mers in aqueous solution is too low to be detected by NMR and ii) high-molecular-weight FA complexes exhibit broader line shapes as a result of reduced molecular tumbling rates. The addition of OA above a FA/protein ratio of 1:1 resulted in the appearance of a second, z4-fold broader peak at 182.8 ppm, whereas the protein signal intensities remained the same. This new peak apparently derives from the excess FA molecules that could not bind to I-FABP above an equimolar ratio but rather formed highmolecular-weight lipid aggregates instead. The intensity of the signal at 183.8 ppm, which represents the FA-FABP complex, was not affected by the further increase of FA concentration beyond the 1:1 molar ratio.
In contrast, the 1D carbon NMR spectra of aS (Fig. 6), obtained with increasing amounts of 1-13 C-labeled OA, exhibited the following differences compared with I-FABP. On this type of gel, monomeric aS (arrow) typically migrates with an apparent molecular mass of z17 kDa, which is slightly above its true mass of 14 kDa. aS that was not exposed to any FA (left lane) appeared solely in the monomeric form. On the other hand, a LA-aS mixture at a ratio of 10:1 (right lane), which was generated in the NMR titration study depicted in Fig. 3A, clearly shows both monomeric and oligomeric protein forms. Note that the detected amount of aS is overall higher in the presence of FAs, even though identical protein quantities (2 mg per lane) had been applied; this seems to be attributable to FA-induced changes in the protein conformation, which enable improved recognition by the antibody (see Fig. 4   First, the carbon signals of the nonlabeled protein showed progressively decreasing signal intensities. This intensity decrease, which is analogous to the above-described 1 H-NMR results, was observed not only for the carbonyl resonances displayed in Fig. 6 but also for all other carbon resonances of aS (data not shown). A second significant difference in these 13 C spectra was that the carbon signal originating from the OA 13 C-carboxyl group turned out to be considerably broader (z470 Hz half-width) than the FA peaks that appeared in the OA titration of I-FABP, indicating a yet higher aggregate weight of the FA complex forms, probably because of the incorporation of aS molecules. This broad signal emerged at z183 ppm, next to a narrow peak at 183.9 ppm that derives from the side chain carboxyl groups of the 18 glutamate residues present in the aS sequence (Table 1).
Hence, these NMR data obtained upon the addition of OA to aS differ markedly from previous results (28,31) when OA was added to FABP or albumin at comparable protein and FA concentrations (i.e., 1-2 mM protein in the presence of equimolar amounts or stoichiometric excess of ligand), as demonstrated here with I-FABP. First, the protein signals of FABP (and albumin) show no intensity changes with the addition of FA; that is, these proteins remain in solution without aggregation or oligomerization. In the case of aS, on the other hand, the protein signals diminished progressively with increasing FA concentration, both in the 1 H and 13 C spectra. Second, the 1 H signals of residues located in the high-affinity binding site(s) of FABP family members usually display significant chemical shift changes in the presence of FAs owing to the immediate vicinity of the bound ligand (37). In the 1 H spectra of aS, however, no soluble protein state other than the monomeric apo form is observed. Finally, the OA carboxyl group yields a single narrow peak when bound to FABP (or a cluster of narrow peaks representing several discrete binding sites when bound to albumin), because the segmental motions of the FAs within the binding site(s) are fast compared with the aggregated FA state (43); the intensities of these narrow peaks increase until all protein sites are filled (1 site in I-FABP, 2 sites in liver FABP, and z10 sites in albumin), whereas an additional, broader OA carboxyl peak appears only after saturation of all highaffinity binding sites (34,44). Yet, the significantly broader OA carboxyl peak observed in the aS spectra only provides evidence for a high-molecular-weight aggregate state of the FAs. This aggregate state cannot be micellar, because micelles i) usually occur only above pH 9 and ii) generally display narrow line widths in the 13 C-NMR spectra (45), in accordance with a loosely structured aggregate of z8.4 kDa calculated mass. Rather, the FA aggregates observed with aS must be considerably larger to cause the extreme linebroadening effects observed in the NMR spectra. Moreover, narrow FA signals, as would be expected for a water-soluble protein with high-affinity FA binding site(s), did not emerge in the 13 C spectra of aS upon the addition of equimolar OA amounts, although under these conditions monomeric aS apparently remains in solution in sufficient quantities (Fig. 3A) to generally permit the binding of single FA molecules from the surrounding medium. These molecules would be derived from the FA pool, either as free FAs that exist in nanomolar to micromolar concentrations in aqueous solution (46) or as FAs bound in high-molecular-weight aggregates, such as acid-soap bilayers when the pH is near neutrality (45).

EM data
In an attempt to visualize and structurally characterize the FA aggregate forms that we have postulated from the . At the same time, the continuous increase of the extremely broad peak (z470 Hz half-width) at z183 ppm with increasing FA/aS ratios can be attributed to the appearance of highmolecular-weight lipid aggregates that seem to be considerably larger than those formed in the presence of I-FABP (see Fig. 5). The seven imperfectly repeated hexamer motifs are underlined. Acidic and basic residues are indicated in italic and boldface type, respectively. indirect NMR evidence, we used negative stain EM. To this end, we examined samples prepared under identical conditions as the above-described 13 C-NMR experiments.
First, samples containing 1 mM OA (pH 7.8) were investigated without any protein present to characterize their general appearance. The images revealed large, electron-lucent spherical objects up to 160 nm in diameter (the dark halo around these particles is indicative of stain buildup) as well as smaller tapered objects of z70 3 35 nm (Fig. 7A, B). It can be concluded that these heterogeneously sized objects all represent FA aggregates, because i) the micrographs showed characteristics of lipids, which typically exclude stain (47), and ii) samples containing higher OA concentrations displayed increased numbers of such objects (data not shown). The particle size and shape in OA preparations that had been gently mixed (Fig. 7A) were very similar to those in preparations vortexed vigorously for 90 s (Fig. 7B). At the selected conditions (1 mM OA, pH 7.8, and 258C), FAs are predicted to consist predominantly of a lamellar FA-soap phase (45), as discussed in detail below. Thus, the large particles in our EM images are most likely multilamellar structures. Their bilayer structure, however, was not visualized, and the membrane thickness could not be determined, because the phosphotungstate stain did not penetrate the negatively charged lipid surfaces to enter the cores of the particles. In the presence of aS, these lipid particles appeared slightly larger and more electron-dense as long as no additional energy input was applied (C). Upon vortexing, however, an extensive disruption of the lipid particles was observed (D). Original images were taken at 70,000-fold magnification; white bar in A 5 140 nm. The schematic drawing in the upper right corner of B represents a FA-soap bilayer structure (45), in which approximately half of the carboxylate head groups are uncharged (i.e., protonated) and the other half are negatively charged (i.e., deprotonated), as indicated by circles filled in white and black, respectively.
If we assume that these structures consist entirely of FA molecules without any water inside, the calculated aggregate weight of the spherical and tapered particles (z2 3 10 6 and 62 3 10 3 kDa, respectively) is much too large to yield observable NMR signals. And even if the observed structures were monolamellar vesicles consisting of only a single bilayer (Fig. 7B, inset), the aggregate weights of the two different particle forms (313 3 10 3 and 28 3 10 3 kDa, respectively, not including the enclosed water molecules) would still cause excessive line-broadening in the NMR spectra because of very slow tumbling rates.
When aS was added to OA at a 1:1 molar ratio (i.e., a large excess of protein molecules relative to the number of lipid particles), spherical particles up to 230 nm in diameter were present after gentle mixing (Fig. 7C). These particles were more electron-dense and larger in diameter than those in Fig. 7A. We assume that the overall positively charged N-terminal domain of aS (Table 1) attaches to the negatively charged surface of an OA aggregate. Hence, the increased electron-density of the OA-aS complex is at least in part attributable to aS located on the particle surface. For comparison, 1 mM aS with no FAs present, prepared under identical negative staining conditions, contained only thin, slightly electron-lucent strands surrounded by denser stain (data not shown).
Additional evidence for an interaction between aS and the FA aggregate structures is illustrated in Fig. 7D. The large FA particles, which were observed at an OA/aS ratio of 1:1 after gentle mixing (Fig. 7C), were disrupted by the vortexing procedure, leading exclusively to smaller particles (z30-70 nm in diameter; Fig. 7D). This particle disruption provides strong evidence that the presence of aS destabilizes the OA aggregates, because vigorous vortexing did not disrupt the protein-free lipid particles (compare Fig. 7B versus 7A).
Similar observations were reported by two recent studies in which soluble monomeric aS caused the disruption of membranes composed of phospholipid bilayers (20,48). Destabilization of the OA particles by aS might furthermore explain why the OA-aS aggregates in Fig. 7C appear larger than the protein-free OA aggregates in Fig. 7A, as a flattening of aS-laden particles that adhere to the carbon surface could possibly produce an increase in diameter accompanied by a (nondetectable) reduction in height.
It must be noted that our EM study cannot necessarily be extrapolated to the acidic pH conditions used in some of our NMR experiments. Below neutral pH, the acid-soap bilayer begins to accommodate the uncharged FAs until it eventually undergoes a phase change to an unstructured oil phase. The NMR results showing a decrease of the soluble protein concentration at pH 5.5 and 6.0 upon the addition of LA and DHA, respectively, therefore could reflect either the binding of aS to remnant acid-soap structures or a different mode of binding than that observed above neutral pH.

DISCUSSION
The synucleins are highly conserved neuronal proteins (9); the amino acid sequence of human aS is shown in Table 1. The N-terminal region of aS (residues 1-60) contains most of the basic residues (i.e., 11 of 15 lysines, without any arginines present), with a recurring 11 residue sequence that includes seven imperfectly repeated hexamer (KTKEGV) motifs. This region is very similar in all three members of the synuclein family (a-, b-, and gsynucleins). The central part of the aS sequence (residues 61-95) comprises the highly hydrophobic NAC domain, which by itself can readily form amyloid fibrils under physiological conditions (3); partial deletion of the NAC region in b-synuclein made this protein nonamyloidogenic. Finally, the less conserved C-terminal region of aS (residues 96-140) is characterized by a high content of acidic residues (two-thirds of all aspartate and glutamate residues occur in the last one-third of the protein sequence).
Although purified aS exists in a nearly completely unfolded state in aqueous solution, most structural studies of aS suggest a predominantly helical secondary structure in the presence of lipid membranes (17,18,49,50). The circular dichroism study of Davidson and coworkers (17) demonstrated that the helicity of water-soluble monomeric aS increases drastically from a negligible amount (z3%) up to 82% in the presence of unilamellar phospholipid vesicles. Moreover, the protein preferentially associates with small unilamellar vesicles that contain acidic phospholipids (net negative charge), whereas no binding was observed to vesicles with a net neutral charge. Based on these data, structural models have been developed that feature helices in which the positively charged lysine side chains are generally located at the water-lipid interface (17,49,50). Recently, NMR data have shown that the Nterminus of aS folds in the presence of SDS micelles into two approximately equally long helices (49)(50)(51). Only this net positively charged domain spanning residues 1-102 is involved in binding to the lipid surface, whereas the Cterminal end with net negative charge remains unaffected and freely mobile in solution.
In this study, we raised the question about the type of interaction that occurs between aS and FAs. Based on the close similarity in the molecular weights of FABPs and aS, and because of some correspondence in the amino acid sequences, it has previously been suggested that aS might be a member of the intracellular FABP family (25). However, there are several notable differences. FABPs i) have a welldefined tertiary structure with a high b-sheet content in solution; therefore, they ii) exhibit a high 1 H-NMR signal dispersion both in the absence and presence of FAs and iii) remain monomeric and structurally unchanged in solution in the presence of excess FAs. They also bind maximally one to two FA molecules per protein in specific binding sites and thus reveal narrow carboxyl signal line width(s) for bound FAs. As none of these features apply to purified aS, we conclude that aS is not a member of the FABP family. Moreover, the binding properties of aS do not resemble those of other known high-affinity FA binding proteins (31), including the highly a-helical serum albumin.
Nevertheless, our results support the general hypothesis that aS preferentially binds to negatively charged surfaces. At millimolar concentrations and over a range of several pH values centered at neutral pH, long-chain FAs are well above their solubility limit and tend to aggregate (45,52). At room temperature, OA forms either an oil phase at low pH or a lamellar acid-soap bilayer near neutrality. As we did not observe any phase separation at pH 5.5 or 6.0, the physical state of the FAs in those experiments is not clear. However, an interaction of aS with the FA phase was obvious from the changes in aS signal intensity that were observed. Hence, even at pH 5.5, the FAs apparently produced an environment suitable for the formation of aS-FA aggregates. In the case of the experiments performed at pH . 7, however, the FA aggregates definitely have a lamellar bilayer structure. These acid-soap bilayers (Fig. 7B, inset), which are similar to phospholipid bilayers in their molecular organization, are composed of both unionized (uncharged) and ionized (negatively charged) FAs, thus creating a net negative charge at the bilayer surface that could serve as a docking site for aS molecules. A recent study by Chen and Szostak (53) has shown that at pH 8.5 OA tends to aggregate, first forming small micelles, then medium sized intermediates with an average hydrodynamic radius of 45 nm, and finally, after a period of several hours, bilayered vesicles of 130 nm radius. Binding of more than one aS molecule to such aggregate structures could account for the previously reported in vitro appearance of aS oligomeric forms in the presence of high FA concentrations (21,26). It should be noted, however, that aS oligomerization has also been observed at lower FA concentrations near the solubility limit (26), at which no lipid aggregate structures could be detected by light scattering (Ronit Sharon, personal communication). Apparently, even a few FA molecules may provide the necessary lipid nucleus that is required for protein oligomerization to occur.
Of course, we cannot completely rule out low-affinity FA binding to aS monomers or oligomers in solution if this interaction is so weak that it causes no concomitant conformational changes in the protein. The strong propensity of aS to adopt a helical structure in the presence of negatively charged lipid surfaces such as acidic phospholipid bilayers or detergent micelles (17,19), however, makes this possibility seem very unlikely. In addition, although water-soluble aS dimers or trimers were not observed under the conditions used in our NMR experiments, we cannot exclude the possibility, based on the data presented here, that smaller, biologically active FA-aS complexes could exist in less concentrated solutions. Nevertheless, our data support the formation of high-molecularweight FA-aS complexes with slowed molecular tumbling rates that result in extensive line-broadening of both the protein and FA signals.
The broadened OA carboxyl signal in the presence of aS further indicates that this particular protein stimulates the formation of yet larger FA aggregate structures compared with those found in the presence of I-FABP. This latter effect was also observed in the EM micrographs of unperturbed OA-aS samples. Upon mechanical agitation in the presence of a high aS concentration, however, the large FA aggregates were broken down into smaller par-ticles. Presumably, the partially immersed helical protein molecules, in particular when aS oligomers are formed, cause a disruption of the lipid surface similar to membranolytic peptides such as melittin. The exact composition(s) of these FA-aS aggregates and the mechanism(s) of their formation will have to be clarified in future studies.
The natural function of aS thus appears to be related to its tendency to adopt a predominantly helical conformation in the presence of (preferably negatively charged) lipid surfaces, such as the biological membranes of various subcellular compartments. This interaction is apparently based on two physical criteria typically found in biological membranes: i) a net negative lipid surface charge that presumably interacts with the positively charged N-terminal domain of aS, and ii) a hydrophobic, fluid lipid core consisting, for example, of polyunsaturated acyl chains that better allow the protein to immerse into the membrane layer. Other studies have already shown that aS prefers to bind to PUFAs or to PUFA-esterified lipids (21). As the fatty acyl chains of PUFAs are more disordered compared to saturated FAs (54), they should allow a better penetration of the aS side chains below a lipidcomposed surface.
How do these in vitro results relate to the brain, where the steady-state concentration of free (unbound) FAs is very low? As in other tissues, most FAs are bound either to cell membranes or to intracellular lipid binding proteins such as FABPs, thus keeping the concentration of unbound FAs in the submicromolar range, well below their solubility limit. The documented presence of FA aggregates within cells is extremely rare and has been associated with severe pathology in peroxisomal disorders, such as X-linked adrenoleukodystrophy (55). Therefore, acid-soap bilayers, as formed in our in vitro studies, are not expected to be present in the cytosol of brain cells to act as catalysts for the adsorption of aS. Rather, aS appears to be attracted to negatively charged membrane surfaces (17,19).
Nevertheless, FAs could still play a role in the natural function of aS in the brain. Unesterified FAs enter the brain continuously as albumin circulates through brain capillaries and releases FAs (56). In addition, FAs are released inside brain cells by deesterification from the sn-2 position of phospholipids, as catalyzed by phospholipase A 2 (PLA 2 ) (57). FAs released by PLA 2 are mainly PUFAs, whereas the FAs delivered by albumin reflect the wide variety of dietary FAs. Two PUFAs that are especially abundant in the brain, arachidonic acid and DHA, are thus released continuously from membrane phospholipids by PLA 2 activity. The PUFAs so generated remain localized mainly in the membrane, where they are z50% ionized, hence increasing the net negative charge of the membrane surface (58). Therefore, their presence could initiate or enhance the binding of aS to the membrane.
Despite the fact that the physiological function of aS is still unknown, all available data suggest that the active protein form is represented by structured aS monomers or oligomers, which are adsorbed to lipid surfaces. If the altered FA metabolism in the presence of aS is in fact a result of the previously reported reduction in FA uptake (29,30), it may be concluded that the membrane-attached aS molecules affect the trafficking of FAs across the membrane, possibly by altering the equilibrium distribution of FAs in the leaflets of the lipid bilayer. This could occur (speculatively) either by diminishing the usually fast rates of FA flip-flop across the bilayer (59) or by influencing other membrane-bound proteins that allegedly act as FA transporters across the membrane (60). Conversely, if the reported lower FA uptake is induced by changes in FA turnover, it can be hypothesized that aS affects the corresponding enzymes that regulate FA storage and/or utilization inside the cell. In any case, the mechanisms underlying these effects need to be elucidated for a better understanding of the role that aS plays in the healthy brain.