Exposure to Long Chain Polyunsaturated Fatty Acids Triggers Rapid Multimerization of Synucleins*

Detergent-stable multimers of (cid:1) -synuclein have been found specifically in the brains of patients with Parkinson’s disease and other neurodegenerative diseases. Here we show that recombinant (cid:1) -synuclein forms multimers in vitro upon exposure to vesicles containing certain polyunsaturated fatty acid (PUFA) acyl groups, including arachidonoyl and docosahexaenoyl. This process occurs at physiological concentrations and much faster than in aqueous solution. PUFA-induced aggregation involves physical association with the vesicle surface via the large apolipoprotein-like lipid-binding domain that constitutes the majority of the protein. (cid:2) - and (cid:3) -synucleins, as well as the Parkinson’s disease-associated (cid:1) -synuclein variants A30P and A53T, show similar tendencies to multimerize in the presence of PUFAs. Multimerization does not require the presence of any tyrosine residues in the sequence. The membrane-based interaction of the synucleins with specific long chain polyunsaturated phospholipids may be relevant to the protein family’s physiological functions and may also contribute to the aggregation of (cid:1) -synuclein observed in neurodegenerative disease. The synucleins are small highly conserved proteins of uncer-tain function, abundant in the vertebrate nervous Their dominant structural feature is the presence of a long apo-lipoprotein-like domain capable of

The synucleins are small highly conserved proteins of uncertain function, abundant in the vertebrate nervous system (1). Their dominant structural feature is the presence of a long apolipoprotein-like domain capable of binding reversibly to specific phospholipids (2). Increased expression of ␣-, ␤-, and ␥-synucleins has been detected in breast, ovarian, and bladder cancers (3)(4)(5)(6) and has been mechanistically linked to increased potential for tumor growth (7) and metastasis (8). In the song control system of the avian forebrain, expression of ␣-synuclein has been correlated with periods of developmental (9, 10) and steroid-regulated (11) change in neuronal structure and function.
In a previous study of ␣-synuclein's interaction with phospholipids, we incidentally observed that exposure of recombinant ␣-synuclein to phosphatidylinositol stimulates the formation of stable apparent multimers that are maintained throughout SDS gel electrophoresis (2). Phosphatidylinositols play a critical role in vesicle cycling at presynaptic terminals, where ␣-synuclein is particularly enriched. Moreover, factors that stimulate multimerization of ␣-synuclein may have a direct role in promoting neurodegenerative disease. Therefore, we set out to characterize more thoroughly the basis for the enhancement of synuclein multimerization by specific phospholipids.

EXPERIMENTAL PROCEDURES
Cloning of Synucleins and Mutants-The human ␣-synuclein cDNA used in this study was the generous gift of M. Irizarry. Human ␤and ␥-synuclein cDNAs were cloned from an adult human brain cDNA library (CLONTECH). The canary ␣-synuclein cDNA was originally cloned from mRNA representing a portion of the canary telencephalon (9). For bacterial expression, these cDNAs were subcloned into pET28(a) (Novagen), which directs inducible expression under a T7 lac promoter. All mutant forms of human and canary ␣-synuclein were constructed by long polymerase chain reaction-based techniques using Pfu polymerase (Stratagene), 5Ј-phosphorylated primers, and subsequent re-circularization of polymerase chain reaction products by T4 ligase. The correct DNA sequences of all constructs were confirmed by DNA sequencing.
For SDS-polyacrylamide gel electrophoresis studies, protein concentrations were always 1 M. Lipids of interest and proteins were incubated at molar ratios of 5:1 ( Figs (2) by size exclusion chromatography on a calibrated Superose 6 FPLC column (Amersham Pharmacia Biotech) into 1-ml fractions, which were then mixed with Laemmli SDS-loading buffer and boiled for 5 min prior to SDS-polyacrylamide gel electrophoresis and transblotting. Immunoblots were probed with H3C monoclonal antibody (1:100,000) and horseradish peroxidase-anti-mouse secondary (1:3000; Amersham Pharmacia Biotech). Elution profiles of multilamellar vesicles, SUV, and unbound protein were monitored by light scattering (vesicles) and absorbance at 273 nm (protein) (2,44).

RESULTS
Acceleration of Multimerization by Exposure to PUFA-containing Surfaces-Recombinant human ␣-synuclein was purified from bacteria and exposed to vesicles prepared from a base of synthetic POPC, following conditions previously described (2). The protein/vesicle mixtures were then fractionated on SDS-polyacrylamide gels and analyzed by immunoblotting with the H3C monoclonal antibody (9). When the vesicles were supplemented with liver-derived PI (2) or brain-derived PIP 2 , ␣-synuclein migrated on SDS-polyacrylamide gels in two predominant forms (Fig. 1): a monomer (ϳ21 kDa) and a larger apparent dimer (ϳ45 kDa). Smaller amounts of apparent trimers and even larger multimers were also sometimes evident, especially with prolonged incubations, after which immunore-activity would appear in the stacking gel. Upon size exclusion chromatography of synuclein PIP 2 mixtures, both monomeric and multimeric forms of synuclein eluted with SUV fractions (Fig. 1), indicating a stable and specific physical association with membrane surfaces of high curvature (2). In parallel incubations using vesicles containing several other synthetic phospholipids, only monomeric synuclein eluted, as free protein with neutral phospholipids and with SUV fractions when acidic phospholipids were used (not shown; see Ref. 2).
We reasoned that the apparent selectivity of multimerization for PI and PIP 2 might be attributable to the lipid head group (phosphoinositol), the lipid acyl chains, or some combination of both. The PI and PIP 2 used in these studies were purified from animal sources and thus bear a heterogeneous mix of acyl groups (Table I). To test for sensitivity to acyl chain composition, we compared the response to PI derived from either bovine liver or soy. Liver PI, but not soy PI, induced a large amount of ␣-synuclein dimer and detectable trimer after 12 h ( Fig. 2A). Soy PI is abundant in polyunsaturated fatty acid (PUFA) tails of 18 carbons (linoleoyl and linolenoyl) but lacks the longer chain PUFAs characteristic of animal-derived phospholipids (Table I). Soy PI is also deficient in the intermediate length saturated acyl chain, stearoyl.
Since soy PI was less effective at inducing multimerization and was especially deficient in arachidonoyl and stearoyl (Table I), we tested the hypothesis that the presence of these acyl chains in the phospholipid vesicles is sufficient to induce multimerization. As Fig. 2B demonstrates, multimerization did not occur when synuclein was exposed to acidic vesicles containing only palmitoyl and oleoyl acyl chains (lanes labeled -). However, when synuclein was incubated with POPC vesicles supplemented with both stearoyl and arachidonoyl (by inclusion of 10% SAPC), robust multimerization occurred. Inclusion of stearoyl chains alone did not induce multimerization (Fig. 2B, SOPC), whereas arachidonoyl inclusion alone was sufficient (Fig. 2B, PAPC). We also tested another long chain PUFA acyl group with a different arrangement of unsaturated bonds, docosahexaenoyl (Fig. 2B,  SDPC). This acyl chain was as effective as arachidonoyl at inducing multimerization of ␣-synuclein, indicating a potentially general role for long chain polyunsaturated acyl groups in triggering ␣-synuclein multimer formation.
In a related experiment (Fig. 2C), we tested whether PUFAs alone (as free fatty acids) can cause the same effect as PUFAcontaining phospholipids. Incubation of ␣-synuclein with free AA at 10 -100 M (lanes 3-6) caused little if any acceleration of multimerization compared with that observed for ␣-synuclein in buffer with no AA (lanes 1 and 2). At 1 mM (above the critical  (Table I) and recombinant human ␣-synuclein was separated by size exclusion chromatography (see "Experimental Procedures"). Shown is an H3C immunoblot of consecutive fractions (lanes from left), with elution profiles of multilamellar vesicles (MLV), SUV, and unbound protein (free) indicated below. Markers on the left indicate the migration of prestained protein standards (in kilodaltons). Free monomeric ␣-synuclein migrates with a mass of 21 kDa (not shown), but incubation with PIP 2 results in a small retardation of the monomer's electrophoretic mobility and the formation of apparent ␣-synuclein dimers (ϳ45 kDa) and trimers (ϳ70 kDa). micellar concentration for AA), robust multimerization occurred (lanes 7 and 8). This concentration dependence suggests that PUFAs exert their effects on synuclein when organized as either a micellar (free fatty acid) or vesicular (phospholipid) surface.
As a further test of the role of surface association in PUFAdependent multimerization, we compared arachidonoyl-containing vesicles prepared with only neutral head groups (PC) to those supplemented with 10% phosphatidic acid (PA). The inclusion of acidic phospholipids greatly accelerates vesicle binding by ␣-synuclein (2); under the conditions used here, vesicle binding is complete at the zero time point of the assay. 2 Consistent with Fig.  2B, inclusion of PA in vesicles lacking PUFA acyl groups did not stimulate multimerization (Fig. 3, PA lanes). Inclusion of arachidonoyl-containing lipids in neutral vesicles did stimulate multimerization, but the effect was comparatively slow, developing between 12 and 24 h (Arach lanes). However, the combination of both PA and arachidonoyl in vesicles (PA ϩ Arach lanes) synergistically enhanced the rate of multimerization (Fig. 3, compare 12-and 24-h points, Arach versus PA ϩ Arach lanes).
Mapping of Interaction Domains-To determine whether the tendency to multimerize in the presence of PUFAs is a general property of synuclein family proteins, recombinant ␤and ␥-synucleins were also prepared and tested. These proteins were equivalent to ␣-synuclein in their tendency to multimerize following PUFA exposure (Fig. 4A). Equivalent PUFA-dependent multimerization also occurred with both human PDassociated variants of ␣-synuclein, A30P and A53T (Fig. 4B). ␣-, ␤-, and ␥-synucleins share nearly identical N-terminal apolipoprotein-like domains but diverge somewhat in their acidic C termini (1), suggesting that their common multimerization behavior may be mediated by the N-terminal domain. Consistent with this hypothesis, removal of the N-terminal domain from ␤-synuclein by clostripain cleavage (at residue Arg 86 ) abolished multimerization of the remaining C-terminal fragment as detected by the H3C antibody (Fig. 4A, ␤c lanes). Clostripain cleavage of the canary ␣-synuclein occurs at residue Arg 43 , leaving approximately half of the lipid-binding domain attached to the C-terminal fragment detected with H3C. This fragment multimerizes robustly (Fig. 4A, ␣c lanes), indicating that some but not the entire lipid-binding domain is both necessary and sufficient for PUFA-dependent multimerization.
We then tested whether various recombinant human ␣-synuclein deletion mutants, each lacking a part of the protein encoded by an individual exon, would form multimers when incubated with arachidonoyl containing vesicles. Since deletions of exon 6 or 7 disrupted binding to the H3C antibody epitope, identical immunoblots were probed with monoclonal antibodies SYN303 (to the N terminus of the protein (46)) and LB509 (to an epitope within exon 6 (47)). The results (Fig. 5) show that each of these single exon deletion constructs will form multimers in the presence of PUFAs. Together with Fig.  4, these results indicate that the capacity to support multimerization is present in at least two of the three exons comprising the lipid-binding domain (exons 3, 4, and 5 (44)). Curiously, the SYN303 antibody detected multimeric synuclein only very weakly, except in the case of the ⌬6 construct (Fig. 5). This result suggests that PUFA-bound multimers may exist in a conformation in which the N terminus (detected by SYN303) is masked somehow by residues in exon 6.
Recently, Souza et al. (48) demonstrated that the conserved tyrosines in ␣-synuclein could become nitrosylated and form protein-protein cross-links in the presence of peroxynitrite. Moreover, Giasson et al. (49) demonstrated that these modifications actually exist in the Lewy bodies of PD brain tissue. Since PUFAs are obvious targets and propagators of oxidative damage and lipid peroxidation, we wondered whether PUFAstimulated multimerization might be mediated primarily via these conserved tyrosines. However, when we replaced all four tyrosines of canary ␣-synuclein with phenylalanines, exposure  2. Exposure to PUFAs in vesicles or micelles is sufficient to stimulate synuclein multimerization. Recombinant ␣-synuclein was incubated with various phospholipid vesicles or free arachidonic acid, and the reactions were analyzed by H3C immunoblot. Size markers are as in Fig. 1A) ␣-Synuclein was incubated for 12 h with POPC vesicles containing 8% liver or soy PI (Table I) to PUFA-containing vesicles still caused multimerization (Fig.  6, Y4F). Furthermore, supplementation with antioxidant BHT (500 M) or chelator EDTA (5 mM) did not block multimer formation (data not shown). DISCUSSION Here we show that synuclein will rapidly form stable multimers when exposed to vesicles containing long chain polyunsaturated acyl groups. These multimers remain intact on SDS gels and also resist disruption by various chromatographic procedures, prolonged boiling, SDS concentrations above 8%, acidification, Folch lipid extraction (50), or the addition of urea (data not shown). When membrane binding is promoted by the inclusion of acidic phospholipids, multimerization is accelerated and occurs within hours or less, with physiological concentrations of protein and phospholipid. These observations have potential relevance to both the normal function and pathological behavior of synuclein proteins.
Although synuclein's normal function is not understood, an interaction with phospholipids is likely to be central for several reasons (1,51). Biochemical fractionation of brain homogenates shows that synuclein partitions into both soluble and membranous compartments (9). In vitro, synuclein binds reversibly to vesicles containing acidic phospholipids (2). The domain involved in this interaction with vesicles spans roughly twothirds of the molecule (44) and, upon binding, adopts an amphipathic secondary structure virtually identical to that of the class A 2 family of lipid-binding helices in apolipoproteins (2,52). This lipid-binding domain is conserved in all species and in all members of the synuclein family (1) and is also the domain that mediates PUFA-induced multimerization (Fig. 4). Thus, the large complexes induced by exposure to PUFAs could represent a discrete functional intermediate, perhaps coupling local PUFA concentrations to some specific intracellular response.
Given the stability of the multimers and their specific requirement for long chain PUFAs, it is possible that PUFAs themselves are a component of these larger "soluble" synuclein complexes. If so, synuclein binding may influence the turnover or local organization of PUFAs in cellular membranes. Such an effect could have a number of biological consequences, given the demonstrated roles for PUFAs in processes including ion channel inhibition (53), synaptic vesicle recycling (54), and phospholipase activation (55). Consistent with this hypothesis, both ␣and ␤-synucleins have been found to inhibit the PIP 2dependent activation of phospholipase D2 (56), and genetic disruptions of synuclein expression have apparent effects on synaptic vesicle pools in both cultured neurons (57) and the striatum of knockout mice (58).
It is also possible that the synuclein multimerization observed here is related to the protein's behavior in diseases that involve synuclein aggregation. Much attention has been focused on the role of synuclein in neurodegenerative diseases involving Lewy bodies (12)(13)(14)(15)(16)(17)(18)(19)(20), since synuclein appears to be the major protein component of these inclusions. Attempts to model this process have shown that ␣-synuclein on its own will eventually form fibrillar aggregates in aqueous solution (37-41, 48, 59 -62), apparently proceeding through a nucleation process that begins with the formation of smaller "protofibrils" (39). The multimers we observe could serve an equivalent function in the nucleation of fibrils in vivo. However, synuclein multimers do not typically accumulate in healthy tissue (15,17,24,25,42,43,63), despite the presence of arachidonoyl and docosahexaenoyl groups in the healthy brain (64) and despite our observation that PUFA-induced multimerization is more rapid than aqueous multimerization by several orders of magnitude (as assessed by comparing the concentrations and incubations times used in various published studies). It may be that enzymatic processes in the cell act to uncouple or degrade PUFA-induced synuclein multimers. Some forms of synuclein may be degraded via a proteasomal pathway, for example, although the evidence for this is still equivocal (65)(66)(67)(68) and the signals that might regulate this process are not yet defined. Thus, it is conceivable that alterations in the production or removal of PUFA-dependent multimers could have an early role in the nucleation of pathological aggregates. Subsequent oxidative modifications may accelerate and stabilize the formation of large fibrillar aggregates, such as are found in Lewy bodies in Parkinson's disease.
FIG. 6. Tyrosine residues are not required for PUFA-induced multimerization. Recombinant wild-type canary synuclein (AS) and canary mutant Y4F, devoid of tyrosines, were incubated with 80% POPC, 10% POPA, 10% PAPC vesicles and probed with HATCAN mAb, raised against residues 108 -120 of canary ␣-synuclein. Size markers are as in Fig. 1. gested that oxidative mechanisms are important in forming or stabilizing large synuclein aggregates in vitro (48,69) and in tissue (49). The specific requirement for PUFAs observed here is consistent with an oxidative mechanism, since the multiple unsaturated bonds readily participate in oxidative reactions and can produce reactive species (e.g. lipid hydroperoxides), which can propagate damage to membranes and proteins (70). One set of studies suggested that oxidative/nitrative conditions can lead to formation of dityrosine cross-links in ␣-synuclein (48), but this mechanism does not appear to be relevant to the PUFA-dependent pathway described here, since site substitution of all four tyrosines in ␣-synuclein did not abrogate multimerization. Lysine residues, which are abundant in synuclein, are also especially susceptible to oxidation (71). We did not detect any obvious effect of several pro-and antioxidants when added to the multimerization reaction, however. Additional studies are warranted to characterize the structure of the multimers and to define more precisely the nature of the chemical reaction leading to the formation of these structures.