Identification of Synaptosomal Proteins Binding to Monomeric and Oligomeric α-Synuclein

Monomeric α-synuclein (αSN) species are abundant in nerve terminals where they are hypothesized to play a physiological role related to synaptic vesicle turn-over. In Parkinson’s disease (PD) and dementia with Lewy body (DLB), αSN accumulates as aggregated soluble oligomers in terminals, axons and the somatodendritic compartment and insoluble filaments in Lewy inclusions and Lewy neurites. The autosomal dominant heritability associated to mutations in the αSN gene suggest a gain of function associated to aggregated αSN. We have conducted a proteomic screen to identify the αSN interactome in brain synaptosomes. Porcine brain synaptosomes were fractionated, solubilized in non-denaturing detergent and subjected to co-immunoprecipitation using purified recombinant human αSN monomers or oligomers as bait. The isolated αSN binding proteins were identified with LC-LTQ-orbitrap tandem mass spectrometry and quantified by peak area using Windows client application, Skyline Targeted Proteomic Environment. Data are available via ProteomeXchange with identifier PXD001462. To quantify the preferential binding an average fold increase was calculated by comparing binding to monomer and oligomer. We identified 10 proteins preferentially binding monomer, and 76 binding preferentially to oligomer and a group of 92 proteins not displaying any preferred conformation of αSN. The proteomic data were validated by immunoprecipitation in both human and porcine brain extracts using antibodies against monomer αSN interactors: Abl interactor 1, and myelin proteolipid protein, and oligomer interactors: glutamate decarboxylase 2, synapsin 1, glial fibrillary acidic protein, and VAMP-2. We demonstrate the existence of αSN conformation selective ligands and present lists of proteins, whose identity and functions will be useful for modeling normal and pathological αSN dependent processes.


Introduction
The neurodegenerative a-synucleinopathies are dominated by PD, DLB, and multiple system atrophy (MSA), which are characterized by aggregation and deposition of aSN [1,2]. aSN is a 14 kDa natively unfolded protein, and in vitro studies have shown that monomeric, soluble oligomeric, and insoluble fibrillar species of aSN exists in equilibrium [3][4][5]. aSN is normally located in the presynaptic nerveterminals in high concentrations [6]. The function of aSN in the nerve terminal remains ambiguous, but recent reports suggest a functional role as chaperone for the SNARE complex [7]. aSN is causally linked to autosomal dominant forms of PD where missense mutations in aSN (A30P, E46K, A53T, H50Q, and G51D) [8][9][10][11][12] and multiplications of the normal aSN coding reading frame cause autosomal dominant PD and DLB [13,14]. Moreover, GWAS studies have identified variation in the aSN gene as the strongest genetic risk factor for sporadic PD [15][16][17], and substitutions of A18T and A29S have been associated with sporadic PD [18].
The autosomal dominant disease associated mutations of aSN suggest a gain of toxic function and the soluble aSN oligomers are hypothesized to represent the culprit (Reviewed in [19,20]) based on biophysical, cellular and in vivo studies in models and man [21][22][23][24]. Evidence suggests aSN aggregation begins in nerve terminals and spread from synapses via axons to cell bodies forming Lewy neurites and Lewy bodies [25]. Hence, presynaptic proteins that interact with aggreated aSN species may represent the first neuronal partners in protein networks that are off-set by early aSN aggregates and recent data on prion-like spreading of aSN pathology in mouse suggest a role for transsynaptic spreading of aSN pathology [26,27].
Mass spectrometry based proteomic analyses are powerful tools to profile changes in proteins and identify protein-protein interactions in biological systems [28,29]. aSN has previously been the target for studies of its presence and post translational modifications [30][31][32]. aSN binding partners have been investigated in co-immunoprecipitation experiments in cell extracts, e.g. subjected to oxidative stress [33,34] and in brain extracts for Ser-129 phosphorylation dependent interactions and targets for insoluble aSN filaments [35,36].
To investigate the interactome disease-related oligomeric aSN in a preparation as close to its normal subcellular localization we used purified porcine brain synaptosomes. We conducted semi-quantitative comparison of the interactions for monomeric and oligomeric aSN to get an impression on the potential gain-of-function caused by the process of aggregation. The aSN oligomers used in this screen share structural similarities with oligomer species present in human dementia with Lewy body brain tissue and species developing in aSN transgenic mouse at the time of degeneration [21,22]. Synaptosomes have classically been used as the starting material for purifying synaptic vesicles because they are enriched in axonal nerve terminals, but also contains mitochondria, myelin and other brain cell structures [37]. Because sample complexity and the wide dynamic range of concentrations of analytes counteracts the ability of our MS based strategy to characterize the potential aSN interactome proteome, we reduced sample complexity by subfractionating the crude synaptosomal preparation into synaptosomal membranes, synaptosomal lysate, synaptic vesicles, and cytosol prior to non-denaturing detergent extraction [38]. The extract was incubated with purified aSN monomers or oligomers prior to co-immunoprecipitation (Co-IP) using anti-aSN antibodies. The interacting proteins were identified by label-free proteomic analysis on a LC-LTQ-orbitrap tandem mass spectrometer followed by a semi-quantitative analysis using Skyline Targeted Proteomic Environment [39]. Based on the quantitative analysis, we could group the ligands in three classes of interacting proteins, preferential monomeric (MP), preferential oligomeric (OP), and aSN interacting proteins with no detectable conformational preferences (NPB). Among the ligands were proteins involved in pathways genetically associated to PD and functionally linked to aSN dependent degeneration, e.g. VAMP-2. The identification of aSN targets and their semiquantitative binding preference for monomeric versus oligomeric aSN bring information that can substantiate current and and help build future hypotheses and models.

Materials and Methods
Production and isolation of αSN monomer and oligomer Human recombinant aSN was produced as described previously [40]. In order to prepare monomer and oligomer, lyophilized aSN was dissolved in PBS (7.2 mM Na 2 HPO 4 , 2.8 mM NaH 2 PO 4 , 140 mM NaCl, pH 7.4) to a final concentration of 10 mg/ml (690mM) and incubated for 30 min on ice. The dissolved protein was centrifuged on a tabletop centrifuge at 14,800 rpm for 5 min at 4°C and afterwards loaded on a Superdex 200 10/30 GL column (GE Healthcare) and eluted with 0.5 ml/min PBS [41]. Oligomers were collected from 17-22 min, and monomer aSN was collected at 35 min ( Fig. 1) Protein concentrations of the fractions were determined by Bicinchoninic acid protein concentration assay (BCA) (Pierce).

Preparation of synaptosomal lysates from porcine brain
The synaptosomal fractions were prepared as described previously [42], with minor modifications. A schematic flowchart of the preparation of the synaptosomal fractions can be found in S1 Fig. After collecting fresh porcine brain tissues (Jensen's Slagtehus, DK), all steps were carried out on ice or at 4°C. The cortical grey matter region was dissected to avoid myelin-rich areas. The harvested tissue was homogenized in 3 x volume (w/v) ice-cold homogenization buffer (320 mM sucrose, 4 mM HEPES-NaOH, 2 mM EDTA, and Complete protease inhibitor mix (Roche), pH 7.4) using a loose-fitting glass-Teflon homogenizer (10 up-and-down strokes, 700 rpm). The homogenate was centrifuged for 10 min at 1000 G in a Sorvall RC 5C plus centrifuge. The resulting pellet (P1) was discarded, while the supernatant (S1) was collected and centrifuged for 15 min at 12 000 G. The supernatant (S2) was removed, and the pellet (P2) was washed by resuspension in 30 mL of homogenization buffer and recentrifuged for 15 min at 13 000 G to yield a supernatant, S2 0 , and a pellet, P2 0 . The latter pellet represents a crude synaptosomal fraction, which was subsequently resuspended in homogenization buffer, then homogenized in a glass-Teflon homogenizer. 10 x (v/v) ice-cold water containing protease inhibitors were added, and the whole suspension was immediately subjected to five up-and-down strokes at 1000 rpm. (hypotonic lysis). 1 M HEPES-NaOH buffer (pH 7.4) was immediately added to a final concentration of 10 mM HEPES followed by 30 min incubation on ice. The lysate was centrifuged for 20 min at 33 000 G to yield the lysate pellet (LP1-synaptosomal membranes) and a lysate supernatant (LS1-synaptosomal lysate). Subsequently, the supernatant was centrifuged for 2 hours at 260 000 G in Beckman Optima LE-80K ultracentrifuge. The supernatant was isolated (LS2-high speed cytosol) and pellet (LP2-synaptic vesicles) was resuspended in homogenization buffer. Complete protease inhibitor (Roche) and Triton X-100 to a final concentration of 0.5% were added to all samples.

Preparation of synaptosomal lysates from human brain
The human brain tissue from an 80 year old neurological intact male patient was from the South Australian Brain bank. The synaptosomal fractions were prepared as described for the porcine brain.
Co-immunoprecipitation of αSN binding proteins 2 mg purified aSN monomer or aSN oligomer was incubated with 1 mg/ml synaptosomal lysate, synaptosomal membranes, synaptic cytosol, or synaptic vesicles in PBS, 0.5% Triton X-100 in a total volume of 1 ml at 4°C in rotating tubes overnight. Affinity purified rabbit polyclonal aSN specific antibody (ASY-1) covalently bound to Sepharose beads 4B (GE Healthcare) was added in a 5 times molar excess (5 IgG: 1 aSN) to the samples followed by 2 hours incubation at 4°C in rotating tubes. The Sepharose beads were isolated and washed twice with PBS, 0.5% Triton X-100, and Co-IP proteins were eluted by incubation in non-reducing SDS loading buffer for 2 hours at room temperature. Three individual Co-IP were conducted in the four The misfolded state of the oligomers was demonstrated by dot blotting of 250 ng αSN monomer and oligomer in duplicates using the aggregate-specific antibody Fila-1 and the total loading of αSN on the filter by pan α-synuclein antibody (αSN BD). (C) Characterization of the porcine synaptosomal fractions used for pull-down experiments. 30µg from synaptosomal membranes (LP1), synaptosomal lysate (LS1), synaptic vesicles (LP2) and synaptosomal cytosol (LS2) were immunoblottet and presence of specific markers for synaptic vesicle synaptophysin, endoplasmic reticulum 78 kDa glucose-regulated protein (GRP78), mitochondria voltage-dependent anionic channel (VDAC), lysosomal LAMP-2a, and α-tubulin. (D) The stability of the isolated αSN monomer and oligomer in brain fractions during the co-immunoprecipitation protocol was studied by a subsequent gel filtration. Detergent extracts of porcine brain (LS2) was incubated with buffer as negative control (Lysate), monomer (Lysate + M) and oligomer (Lysate + O) for 24 h at 4oC, followed by gelfiltration. Samples from fractions eluting corresponding to O, M and a fraction eluting at 25 min between monomer and oligomer were resolved by SDS-PAGE and subjected to immunoblotting with monoclonal anti-α-synuclein antibody (αSN BD). The nature of αSN does not change either as demonstrated by the blot of recombinant monomeric αSN. Clearly the monomer migrates in the M fraction and the oligomer predominantly in the O fraction. The level of endogenous porcine αSN is low compared to exogenous recombinant αSN. described synaptosomal preparations using either monomeric, oligomeric aSN, or buffer as negative control, resulting in 36 samples to be analyzed by mass spectrometry.

Mass spectrometric peptide mapping
Samples were tryptic digested, and analyzed by LC-LTQ-Orbitrap mass spectrometry as previously described [43] but with minor modifications. Briefly, digested samples were acidified with 0.1 volume of 10% formic acid and cleaned using Ultra Microspin columns (Cat# SUM SS18V, The Nest Group, Southborough, MA, USA) following manufacturer's instructions. Samples were then dried to completion in a rotor speed vacuum and resuspended in 80 ml of 98% ddH 2 O, 2% acetonitrile, 0.1% formic acid for analysis. Liquid chromatography was performed using a Waters NanoAcquity UPLC (Waters Corporation, Milford, MA). Peptides were separated online with 75 mm i.d. × 20 cm home-packed fused silica columns (100 Å Magic C18AQ: Michrom Bioresources, Auburn, CA, USA) with a 125 min 5-90% acetonitrile/ water gradient containing 0.1% formic acid. All data were acquired on a LTQ-Orbitrap XL mass spectrometer (Thermo Corporation, San Jose, CA). Data Dependent Acquisition mode was employed for this instrument using a "Top 10" method and 125-minute acquisition time. The high resolution (mass accuracy: 5 ppm or less) MS data was acquired (60,000 at 400 m/z) followed by 10 ion trap MS/MS acquisitions. The AGC target value for the precursor scan was 1E6 counts and for the ion trap MS/MS scans to 3E4 counts at a minimum MS signal level of 2000 counts and an isolation window of 1.6 m/z. Charge state screening was considered for all MS/MS targets, utilizing only +2,+3, and +4 charge state isotope distribution. Dynamic exclusion was used with parameters set at 1 repeat count, 30 second repeat duration, 500 item exclusion list size, and an exclusion duration of 180 seconds. The mass spectrometry proteomics data have been deposited at the PRIDE partner repository (dataset identifier PXD001462) of the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) [44].
Semi-quantitative identification of proteins based on mass spectrometric peptide mapping RAW data files were converted to mgf and mzXML files using MS Convert (Proteowizard, v3.0.3951). The mgf were then searched using ProteinPilot v4.1 (AB SCIEX, Framingham, MA, USA) referencing a Sus Scrofa database containing all canonical sequences plus all manually reviewed isoform sequences retrieved from UniProt (July 2012; 26,685 protein entries) for peptide and protein identification. Default settings for peptide assignments and protein identifications were used (e.g., Identification as the Sample Type, Biological modifications as the ID Focus, and Unused ProtScore of 0.05 as the Detected Protein Threshold). In case that multiple proteins are identified by the software as equivalent "winners" (i.e., they share all or nearly all of the same peptides identified in the sample and are usually different isoforms in a protein family), the first one is arbitrarily chosen as the representative winner presented in the tables. The mzXML files were searched using X Tandem! Database searching as part of the TransProteomics Pipeline (Institute for Systems Biology) to create spectral libraries; static carbamidomethyl modifications to cysteine and variable oxidation of methionine, as well as allowance for two missed tryptic cleavages were considered. The resulting~pep.xml result files were grouped according to experimental condition in Skyline (v 1.3, MacCoss Lab, Department of Genome Sciences, University of Washington, Seattle, WA, USA) to create distinct spectral libraries. 36 spectral libraries were generated from the three biological replicates of Co-IP from the four different synaptosomal preparations (LP1, LS1, LP2, LS2) using the three different baits (PBS, monomer, and oligomer). All peptides from these libraries were added to the project to populate the peptide list, resulting in 11087 peptides derived from 2049 Sus scrofa proteins. The 36 RAW files from the experimental isolation conditions were then imported to create the MS1 Extracted Ion Chromatograms (XIC) for each peptide. From the spectral libraries, the identification retention time was indicated on these XICs where identification was made from the database search for that RAW file. This was done in Skyline by making connections between the RAW filename and the pep.xml filename in the spectral library. All peptide XICs were manually inspected and only peptides that had at least two positive identifications were retained. Additionally, peak assignments were manually corrected for all files based on these identification retention times, ensuring the correct MS1 signal was integrated. To further validate the specificity of these peak assignments, XICs were created in Skyline for all precursor isotopes, a predicted isotopic distribution of the peptide sequence was then calculated, and this predicted ratio was compared with the actual isotopic ratios for the MS1 signal of all peaks. From this, an 'idotp' dot product number, where 1.00 is a perfect match, was generated in Skyline. These idotp values were also considered for peak assignment. Identified proteins were pre-filtered by excluding proteins with higher peak areas in buffer controls than both monomer and oligomer samples. Peak areas of the remaining proteins were logarithmic transformed in order to reduce the variability between the biological replicates and normalized to averaged peak areas of buffer-control. The average fold increase was calculated as the ratio between peak areas for monomer and oligomers for monomer preferential binding proteins and as ratio between peak areas for oligomers and monomers for preferential oligomer binding proteins. In order for a protein to be considered as either monomer or oligomer binding Student's t-test was applied and conventional significance of 95% was accepted. The criteria of ! 2-fold increase in average fold increase has to be fulfilled in order to be considered as preferential binding protein, similar to the criteria set up in the phospho-peptide interaction study by McFarland et al. [35]. Based on the filtering aSN preferential monomer binding proteins are listed in Table 1, and preferential aSN binding proteins are listed in Table 2. Proteins with significant higher binding to aSN compared to buffer are listed as non-conformation specific aSN binding proteins in Table 3. The porcine proteome is not completely described and unknown proteins are listed with the predicted name derived from an Ensembl automatic analysis pipeline. A flowchart of how quantitative data was obtained from sample preparation to semiquantitative analysis including criteria for exclusions/inclusion is presented in S2 Fig., and semi-quantitative data can be found in S1 Table along with number of identified peptides, number of quantified peptides, and percent sequence coverage.

Validation of protein targets by immunoblotting
Antibodies targeting the proteins selected for validation were tested by immunoblotting on all four subcellular fractions and the fraction with the highest signal was used as source for the validating co-IP (S3 Fig.). Proteins were resolved on 10-16% gradient SDS-PAGE under reducing conditions followed by immunoblotting, as described previously [45]. Quantification of band intensity on western blot was conducted in Image J (Research Services Branch, http://rsb.info.nih. gov/ij/index.hmtl). For the anti-aSN immunoblot in Fig. 1C-D, the filter was fixed in 0.4% PFA for 30 min at room temperature to reduce dissociation from the PVDF membrane, and boiled in PBS for 5 min prior to blocking followed by conventional immunoblotting procedure [46].

Results
aSN oligomers can be formed in vitroin various ways e.g. oxidatively modified by dopamine [50]. We isolate spontaneously assembled, non-covalently modified aSN oligomers by gel filtration [41], which display the FILA-1 epitope that is shared with oligomers present in human dementia with Lewy body brain tissue and aSN transgenic mouse [21][22][23]. Such oligomers are structured but readily depolymerized upon denaturation [51]. Fig. 1A demonstrates the oligomers elute between 17 and 22 min and this pooled fraction is designated O. The monomers elute between 32 min and 37 min and the fraction collected at 35 min, designated M, was used for the further investigations. The presence of aggregate specific epitopes on oligomers, but not monomers, are demonstrated by their binding of the antibody Fila-1 (Fig. 1B), which bind an epitope that is shared with mature fibrils and occur in pathological aSN aggregates in brain [21][22][23].
aSN is predominantly a presynaptic protein so we investigated fractions of brain synaptosomes rather than crude brain homogenates in search for potential binding partners for exogenous monomeric and oligomeric aSN. The non-denaturing detergent Triton X-100 (0.5%) was Co-immunoprecipitations of proteins from detergent extracts of porcine synaptosomal fractions using recombinant αSN monomer, αSN oligomer, or buffer as bait, were conducted, and the interacting proteins were identified by LC-LTQ orbitrap tandem mass spectrometry. The identified proteins are specified by name, Uniprot/swiss-Prot entry, and the synaptosomal fraction in which they were identified. Quantitative analyses were based on Peak Area Intensity of peptides occurring in at least two of three individual immunoprecipitations. The average fold increase is calculated as the ratio between the peak areas for monomer and oligomers for monomer preferential binding proteins. The criteria of ! 2 fold increase compared to oligomer had to be fulfilled in order to be considered a preferential αSN monomer binding protein. In order for a protein to be considered as either monomer or oligomer binding Student's t-test was applied to the logarithmic transformed peak areas and significance < 0.05 was accepted. * indicates the monomer preferential proteins that were validated by western blotting in Fig. 3.   Co-immunoprecipitations of proteins from detergent extracts of the synaptosomal fractions from porcine brain using recombinant αSN monomer, αSN oligomer, or buffer as bait, were conducted, and the interacting proteins were identified by LC-LTQ orbitrap MS/MS. The identified proteins are specified by name, Uniprot/swiss-Prot entry, and the synaptosomal fraction in which they were identified. Quantitative analyses were based on Peak Area Intensity of peptides occurring in at least two of three individual immunoprecipitations. The average fold increase is calculated as the ratio between the peak areas for monomer and oligomers for oligomer preferential binding proteins. The criteria of ! 2 fold increase compared to monomer had to be fulfilled in order to be considered a preferential αSN oligomer binding protein. In order for a protein to be considered as either monomer or oligomer binding Student's t-test was applied to the logarithmic transformed peak areas and significance < 0.05 was accepted. * indicates the oligomer preferential proteins that were validated by western blotting in Fig. 3.
doi:10.1371/journal.pone.0116473.t002 used to extract the four subfractions, synaptosomal membranes (LP1), soluble synaptosomal lysate (LS1), and its subfractions synaptic vesicles (LP2) and synaptosomal cytosol (LS2) (S1 Fig.). The sub-fractionation was conducted to reduce the complexity of the samples in order to favor the downstream MS based identification of proteins co-purifying with aSN or Co-immunoprecipitations of proteins from detergent extracts of the synaptosomal fractions from porcine brain using recombinant αSN monomer, αSN oligomer, or buffer as bait, were conducted, and the interacting proteins were identified by LC-LTQ orbitrap MS/MS. The identified proteins are specified by name, Uniprot/swiss-Prot entry, primary accession number, and the synaptosomal fraction in which they were identified. Quantitative analyses were based on Peak Area Intensity on peptides occurring in at least two of three individual immunoprecipitations. The presented as αSN binding proteins without any apparent conformational preferences are those where the ratio between peak areas for monomer and oligomers were < 2. doi:10.1371/journal.pone.0116473.t003 Identification of Synaptosomal Proteins Binding to α-Synuclein buffer control. The composition of the sub-fractions was analyzed by immunoblotting for proteins known to be present in specific compartment. Synaptophysin, a marker for synaptic vesicles, was enriched in LP2, but could also be detected in LS1 and LP1 (Fig. 1C). The content of endoplasmic reticulum, was analyzed by the presence of 78 kDa glucose-regulated protein (GRP78), and GRP78 was also enriched in LP2, and detectable in LS1, LP1, and LS2. The mitochondrial content was visualized by voltage-dependent anionic channel (VDAC) and found exclusively in the LP1 fraction. Lamp2a, a marker for lysosomes, was enriched in LP2. Furthermore, all the synaptosomal fractions contained a-tubulin. First, we analyzed whether the oligomers and monomers were stable during the procedure of co-IP by comparing their gel filtration behavior before and after incubation for 24 h at 4 o C in the detergent containing synaptosomal cytosol (2mg/ml recombinant monomer or oligomer aSN in 1 mg/ml extract) as determined by anti-aSN immunoblotting of fractions corresponding to O, and M (Fig. 1D). A fraction (I) eluting at 25 min, which is expected to contain low amounts of exogenous aSN (Fig. 1A), was also analyzed in order to investigate a dissociation of the oligomer into smaller species. The endogenous aSN level in the LS2 fraction was low, compared to samples supplemented 2mg recombinant monomer or oligomer aSN, but can be detected in the lysate sample without recombinant aSN primarily eluting as monomer (Fig. 1D). The exogenous monomer eluted in the M fraction after incubation for 24 h with small amounts of high molecular weight material present in fraction I. The oligomers also retained their elution in the high molecular weight fraction O, but with some dissociation into complexes eluting in fraction I and with little detectable material in the fraction M, likely corresponding to endogenous porcine aSN. This shows that the purified oligomers and monomers largely retain their molecular sizes during the incubation in detergent extract of brain lysates for 24 h at 4°C.
For proteomic analysis the brain extracts were incubated with recombinant human aSN in its monomeric, and oligomeric state, or buffer-control followed by pull-down using the polyclonal aSN antibody ASY-1 coupled to sepharose beads. Co-IP proteins were identified by tryptic peptide mapping by LC-MS/MS. Three identical experiments were conducted using the four different synaptosomal preparations and each with monomeric, oligomeric aSN, or buffer control as bait. The 36 individual co-IPs were analyzed by LC-LTQ orbitrap tandem mass spectrometry, which returned 11087 identifying peptides corresponding to 2049 proteins by searching of a Sus scrofa database. The criterion to be fulfilled for proteins to be considered as potential binding partners was identification in at least two of the three biological samples. To compare if binding of a protein is preferential to aSN monomer or aSN oligomer, semiquantitative data was obtained as the peak area. The peak area of individual peptides varied considerably between runs as exemplified by the peak area from ELEAVCQDVLSLLDNYLIK peptide of 14-3-3 protein gamma that ranged from 397,866 to 1,415,206 in buffer controls, from 509,836 to 724,293 in monomer samples and from 1,439,298 to 9,995,353 in oligomer samples. In order to normalize the individual runs, a logarithmic transformation of the MS peak area output was conducted before comparing the runs. This resulted for the ELEAVCQDVLSLLDNYLIK peptide in control values from 5.6 to 6.15, monomer values from 5.71 to 5.86 and oligomer values from 6.16 to 7.0 thus demonstrating a significant difference (p-value 0.03) between oligomer and monomer or buffer control when analyzed by student's t-test. A graphical illustration of the sample preparation, analysis and semi-quantitative analysis and the criteria used can be found in S2 Fig. A total of 178 proteins fulfilled the criteria of being aSN-binding proteins that are present significantly higher in aSN fractions than in buffer controls. 87% of the aSN-binding proteins occurred uniquely among one of the fractions. Tubulin beta-4 and synapsin 1 were present in three fractions and only reticulocalbin 2 was present in all four fractions.
The aSN binding proteins were filtered for their preferential occurrence in oligomer or monomer samples based on the calculated average fold increase between the samples or those being evenly distributed in these fractions. This resulted in lists of preferential monomerbinding proteins (MP; Table 1), preferential oligomer-binding proteins (OP, Table 2) and those not displaying preference between the aSN forms (NPB; Table 3). The tables also display the predicted subcellular localization of the proteins based on Uniprot KB database.
Ten preferential monomer-binding proteins (Table 1) were identified and their predicted subcellular localizations are dominated by mitochondria and cytosol ( Fig. 2A). It should be kept in mind that the detergent extraction removes all membranes and thus binding of e.g. intra-mitochondrial proteins must be interpreted with caution.
The preferential oligomer-binding proteins ( Table 2) comprised a large group of 76 that displayed a complex subcellular distribution although it was dominated by cytosolic and cytoskeletal proteins that may be immediately accessible to cytosolic aSN albeit a large group of mitochondrial proteins were also present (Fig. 2B). It was reassuring to identify the protein p25a/tubulin polymerization promoting protein in this group because it previously has been identified as a preferential binder to aggregated aSN species [48].
The non-preferential group was the largest with 92 proteins (Table 3). Its subcellular distribution was complex and slightly more evenly distributed between the categories than the preferential oligomer-binding protein (Fig. 2C).
The increased binding of proteins to oligomers and monomers are displayed by the foldincrease of binding to the corresponding aSN form (oligomer vs. monomer) based on the average of all their identified peptides (Tables 1 and 2). These values varied greatly from 2 to 802 and thus call for caution when interpreting the increase for proteins that are based on few peptides. To corroborate their qualitative nature, we performed additional Co-IP experiments followed by immunoblotting for selected proteins. Proteins for validation were selected based on availability of well-characterized antibodies raised against human epitopes with at least 95% identity to the porcine protein. As representatives for the preferential monomer-binding proteins, Myelin proteolipid protein (mPLP) and Abl interactor 1 (Abl1) were chosen (Fig. 3A, B) and for preferential oligomer-binding proteins, Glial fibrillary acidic protein (GFAP), Glutamate decarboxylase 2 (GAD2), and Synapsin 1 (Syn1) were chosen (Fig. 3C, D). The data confirm the aSN conformation preference identified by the MS-based technique. The quantitative analyses of three individual pull-down experiments were conducted and resulted in approximately 4-6 times increased binding to oligomers vs. monomers for the oligomer preferential proteins. The results for the monomer binding proteins also clearly demonstrated the preference for monomers although the absolute value for the monomer preference of Abl 1 is unclear because the oligomer binding was as low as the buffer control thus making estimation of ratios pointless.
aSN has been demonstrated as a chaperone for Vesicle-associated membrane protein 2 (VAMP-2), also named synaptobrevin-2, in nerve terminals [7] so it was a surprise that we did not detect VAMP-2 in our proteomic screen. However, inspection of the potential tryptic peptides generated from VAMP-2 demonstrates that only 2 recognizable peptides can be formed. Hence, such peptides may have been missed in the MS-based peptide search. Fig. 3E demonstrates a co-IP analysis of the VAMP-2 binding to monomer and oligomer aSN. This clearly demonstrates that oligomerization aSN enhances the interaction. In order to ensure that the interactions were not artifacts due to species differences between human aSN and porcine brain synaptosomal preparations, we performed validative Co-IPs using human brain synaptosomal preparations, and were able to confirm the finding from porcine brain synaptosomal extracts (Fig. 3G-I).  Table 1), oligomers (OP, N = 76, Table 2) and proteins not displaying any preferences (NPB, N = 92, Table 3). They were grouped based on their subcellular localization as described by their principal localization in the Uniprot database to demonstrate the aggregation state of αSN have potential for significantly changing its cellular targets.   F) and human (G-I) brain extracts were analyzed by immunoblotting using antibodies against antigens selected among the monomer and oligomer binding proteins. Monomer binding antigens were myelin proteolipid protein (mPLP) and Abl interactor 1 (Abl1) and oligomer binding proteins were glial fibrillary acidic protein (GFAP), glutamate decarboxylase 2 (GAD2), and synapsin 1 (Syn1). VAMP-2 was tested because it has been reported to bind αSN, although it was not detected in our proteomic analysis. One representative of three experiments is presented for porcine αSN binding proteins (A, C, E), and the quantification of the three experiments is presented in panels B, D, F. The quantification of bands was performed after subtracting the non-specific signal in the buffer control from the specific bands in monomer and oligomer samples. Bars represent mean ratio between monomer and oligomer AE S.D. of the three replicates. The values for binding to monomer and oligomer were compared by Student's t-test and the resulting p-values are listed above the bars. * Indicates that the band intensity from oligomer did not differ significantly from background making quantifications impracticable. In order to ensure that the interaction were not due to species differences between human and porcine proteins we conducted validations in

Discussion
The functionality of aSN in neurons is related to its state of aggregation. Native monomeric species are hypothesized to conduct physiological roles in nerve terminals, where aSN is present at high concentrations. In brain tissue affected by PD and DLB, aSN accumulates as soluble oligomeric or protofibrillar species in terminals, axons and the somatodendritic compartment and filamentous in Lewy inclusions and Lewy neurites. Moreover, aSN is also present in astrocytes in PD [52] and oligodendrocytes in MSA [53]. Hence, at least three factors have to be considered for possible molecular pathogenic pathways initiated by aSN in disease (Fig. 4). First, a gain of function by novel structures on aggregated oligomeric species. Second, novel functions of aSN species, being monomeric and oligomeric, concentrated at abnormal sites like axons and the cell body or in astrocytes and oligodendrocytes, where they will encounter new interaction partners. Third, potential loss of function by the native monomer population if reduced due to aggregation or abnormal sorting.
We conducted a proteomic screen in sub-fractions from porcine synaptosomes to identify potential interacting networks for aSN that could be affected in a-synucleinopathies. The semi-quantitative proteomic screen was conducted using exogenous human aSN as bait in its monomeric and pathological oligomeric state. Many types of in vitro formed aSN oligomers have been reported, some forming spontaneously [41], some being stabilized by organic solvents [54], and chemical modifications like dopamine [50] and some being off-versus onpathway in the aggregation process [55]. We choose to use the protocol for spontaneously formed oligomers, which we have characterized extensively by surface hydrogen/deuterium exchange combined with mass spectrometry and FILA-1 binding [36,41,51]. These oligomers share surface epitopes for the conformational specific antibody FILA-1 with aSN oligomers present in pathological human and mouse brain tissue [21][22][23]. It was imperative for the experiment that the monomer and oligomer aSN retained their gross molecular structure during the 24h period of the incubations in order to perform a pull-down analysis of ligands for the two conformations. This is especially critical because the non-modified oligomers we use display some dissociation into monomers [51]. Comfortingly, we could demonstrate that the majority of the exogenously added monomer and oligomer retained their sizes as determined by gel filtration of the detergent brain extracts after the 24 h incubation. The elution of monomeric input aSN as monomers after the incubation with brain extracts indicates most remains unbound or so loosely associated to targets that the complex dissociates during the gel filtration. The large and heterogeneous size of oligomers based on their gel filtration profile does not allow judgment of whether they have bound ligands or not by the subsequent gel filtration analysis.
Using the above pull-down protocol 178 proteins were enriched in the aSN containing samples compared to the buffer control according to our specified criteria. These proteins were selected from a total of 11087 MS spectra corresponding to 2049 proteins identified by searching a Sus scrofa database containing all canonical sequences plus all manually reviewed isoform sequences retrieved from UniProt (July, 2012). This number of spectra is comparable to the number of spectra identified in a previous proteomic screen in mouse brain synaptosomes for ligands of the C-terminal 40 residues of aSN [35]. We identified only 41 significantly enriched proteins in the soluble synaptosomal lysate compared to 121 in its two resulting subfractions, synaptic vesicles and synaptosomal lysate. This demonstrates the power of reducing the human brain extracts. One representative of two experiments is presented for each validated protein. The validation for both porcine and human of mPLP, Abl1, Syn1 and VAMP-2 was conducted in the LP2 fraction enriched in synaptic vesicle and the validation to GFAP and GAD2 in the LS1 fraction of synaptosomal lysate. doi:10.1371/journal.pone.0116473.g003 Identification of Synaptosomal Proteins Binding to α-Synuclein complexity in targeted proteomic searches but also highlights a limitation in deciphering complex interacting networks spanning cytosolic constituents and organelles. Of the 2049 identified proteins, 178 proteins had significant increased binding to aSN compared to background. The 178 proteins could be grouped in three categories using > 2 fold-increase as indicative of preferential binding. Preferential monomer binding proteins (MPB) are at risk of loosing their function when aggregation occurs. Preferential oligomer binding protein (OPB) may initiate novel conformation-dependent interactions and No-preference binding (NPB) are proteins that can form novel interactions if the cellular/subcellular context of aSN is changed. To validate the preferential binding to monomer, oligomer and buffer control by immunoblotting, we selected mPLP and Abl1 among the 10 MPB, and GFAP, GAD2 and Syn1 among the 76 OPB and obtained for all selected a positive validation of their predicted target preference. Three amino acids differs between human and porcine aSN, so in order to ensure that the identified interactions were not artifacts due to these differences, the interactions were all positively validated using human synaptosomal preparations. This supports the concept of groups of cellular proteins preferentially binding either monomeric or oligomeric aSN and thus representing interaction networks at risk of being offset in a-synucleinopathies. The larger number of identified interaction partners of OBP compare to MBP could be due to the unfolded nature of monomer aSN compared to oligomer that display distinct structural feature as demonstrated by antigen presentation [21] and deuterium exchange mass spectrometric analysis [51]. During disease αSN undergo aggregation and this lead to novel conformation-dependent interactions (red), which represents a gain of function. In addition, αSN species (monomeric and oligomeric) are concentrated at abnormal sites, like axons and the cell body, or in astrocytes and oligodendrocytes, which give rise to novel interactions because new partners are introduced (green). Finally, an abnormal sorting and aggregation leads to a loss of, or reduced normal αSN concentration in nerve terminals where critical monomer specific interactions will be compromised thus representing a loss of function. Identification of Synaptosomal Proteins Binding to α-Synuclein aSN is considered a predominantly cytosolic protein and cytosolic proteins also represent the largest group with 51 of 178 aSN binding proteins. However, it is evident that aSN species holds potential for targeting proteins in organelles like mitochondria, endoplasmic reticulum, vesicles, plasma membrane and cytoskeleton. As cautionary note, one must be aware that the use of detergents dissolve natural membrane barriers and may give rise to false positive interactions. With that mentioned, mitochondria have previously been reported to interact with and take up aSN [56,57] and cellular aSN expression increases mitochondrial calcium uptake [58] and reduces mitochondrial membrane potential [59]. Our data support that aSN can modulate mitochondrial functionality as we confirm binding to the previously found mitochondrial chaperone mortalin [34] and identify novel targets e.g. in the outer mitochondrial membrane like voltage-dependent anion-selective channel protein 1 and translocase of outer mitochondrial membrane but also components involved in ATP production like ATP synthase subunits alpha and beta. Interestingly, ATP synthase subunit alpha is increased in aSN expressing SH-SY5Y cells [60].
Autophagy has attracted significant interest in PD research because of its ability to degrade intracellular proteins in general and misfolded proteins in particular [61] and because mutations in lysosomal glycocerebrosidase, causing Gaucher's disease, increase the risk of developing PD [62]. Moreover, dysfunctional autophagic flux results in increased cellular excretion of aSN species, which have been sampled from the cytosol by quality control autophagy [63]. Such species may hold increased potential for spreading misfolded aSN to other cells compared to the normal constitutively excreted aSN. Tightly regulated vesicle sorting and acidification of vesicular content is critical for the autophagic flux and proper enzymatic degradation of its cargo. We find aSN hold potential for disturbing these processes because the all three aSN binding fractions contain subunits associated with acidifying proton pumps (V-type proton ATPase subunit D, D1 and F), along with vacuolar sorting proteins (vesicle-associated membrane protein-associated B; vacuolar protein sorting-associated protein 53).
aSN aggregates have been demonstrated in synaptic terminals of brain tissue affected by Lewy body dementia [25]. In the terminals oligomers may inhibit normal synaptic vesicle function because aSN oligomers display increased binding of synaptic vesicle proteins amphiphysin, synapsin 1, synaptic vesicle glycoprotein 2a and VAMP-2 (also known as synaptobrevin 2). aSN has been demonstrated to function as a critical chaperone for VAMP-2 function [7] and this function is hypothesized to rely on a tethering mediated by vesicle binding to the N-terminus and VAMP-2 binding to the C-terminus of aSN [64]. Oligomerization of aSN will enhance VAMP-2 binding and likely inhibit the ordered tethering mediated by native aSN. We did not identify VAMP-2 by our proteomic investigation, but could demonstrate its preferential binding to oligomers by co-IP. This highlights the limitations of proteomic studies where negative results have to be interpreted with care.
Calcium dysregulation is potentially involved in sporadic PD because treatment with calcium antagonists in epidemiological studies has demonstrated a protective effect [65]. We find that the type 4 plasma membrane Ca 2+ ATPase displays increased binding of oligomeric aSN, whereas binding to type 2 plasma membrane Ca 2+ ATPase displays no conformational preference. Additionally, we have in another pull-down study demonstrated increased binding of aSN oligomers to ER-associated Ca 2+ ATPase and this interaction functionally activates the pump (Betzer, Jensen, Manuscript in preparation). Hence, aggregation and redistribution of aSN from its normal presynaptic localization to larger neuronal compartments hold the potential of dysregulating these important Ca 2+ pumps, which are critical for restoring normal cytosolic calcium levels after locally increased Ca 2+ transients involved in signaling. Likewise, dysregulation of excitatory amino acid transporter 1 by aSN oligomeric may affect the removal of glutamate from the synaptic cleft [66] and thus cause hyperactive glutamate signaling and potentially excitotoxicity.
Dysregulation of protein translation has recently been associated to PD by the demonstration of mutations in the translation initiation factor EIF4G1 in familial PD [67] and the identification of specific R1205H variant as a PD risk factor [68]. We did not identify EIF4G1 or its directly interacting partners, but demonstrate several proteins associated to the subsequent steps in protein translation, elongation factors 1 alpha and beta, 40S ribosomal proteins S3, S3a, S5-like protein isoform 2, 60S proteins L6, L12, L22, L32 and acidic ribosomal protein 3-like isoform 2. Among these interactions are both aggregate and non-aggregated dependent interactions. Association of dysregulated protein synthesis in PD was thus initially highlighted by the EIF4G1 mutations [67], but our findings suggest that PD causing defects in protein synthesis can be triggered by dislocalized or misfolded aSN at other levels in this biosynthetic pathway.
Potential targets in signaling cascades were also identified. The cAMP and Protein kinase A (PKA) signaling system may be affected at more levels by interactions with catalytic subunit beta, regulatory subunit type 2, as well as scaffolding A-kinase anchor protein 5. Such interactions may be involved in the aSN dependent attenuation of cAMP dependent TH gene expression in cells [69]. Additionally, we identified Protein kinase C epsilon, calcium/calmodulindependent protein kinase type 2, alpha subunit and creatine kinase thus linking aSN sensitive targets to phospholipase C and its substrate diacylglycerol, calcium regulation and bioenergetics.
Unfolded protein stress in an aSN transgenic mouse, was associated to accumulation of misfolded aSN species in the endoplasmic reticulum of brain tissue and such accumulation was also demonstrated in humans brains affected by a-synucleinopathies [22]. Our demonstration of protein disulfide-isomerase A6 binding to aSN could provide a mechanistic link between accumulated aSN aggregates and interference with protein disulfide bridge formation, folding and ER stress.
The strong preference for sirtuin 2 binding to oligomeric aSN may indicate direct proteinprotein interactions as involved in the role of protein deacetylase sirtuin 2 in nigrostriatal damage in toxic and aSN transgenic PD models [70,71].

Conclusion
We identify a range of aSN interacting brain proteins among which several displays monomer and aggregate specific preferences. The identified ligands validate proteins previously reported to interact with aSN (including actin [33], tubulins [72,73], p25a [48], and mortalin [34]), but also highlights novel proteins from pathways already hypothesized to be involved in the molecular pathogenesis of a-synucleinopathies like calcium regulation, mitochondrial homeostasis, protein translation, kinase signaling and vesicular and lysosomal function. Hence, the data may both corroborate current hypotheses by expanding their involved targets and contribute to the generation of novel hypotheses.
Supporting Information S1 Fig. Fractionation process for generating the synaptosomal fractions for the interaction analysis. Porcine and human brain tissue was homogenized and subjected to differential centrifugation as described in Materials and Methods. Enclosed in the dashed box is the four synaptosomal preparations; synaptosomal membranes (LP1), synaptic lysate (LS1), synaptic vesicles (LP2), and synaptic cytosol (LS2), which were used in the co-immunoprecipitations and subsequent identifications of aSN binding proteins upon solubilized in 0.5% Triton X-100. (TIF) S2 Fig. Schematic illustration of the workflow and decision making used in the semi-quantitative mass spectrometric analysis. Brain tissue is prepared and fractioned resulting in four synaptosomal fractions (LP1, LS1, LP2, and LS2) before being mixed with either aSN monomer, oligomer, or buffer control yielding 12 samples. By conducting the investigation on three biological replicates this gives a total of 36 samples. The 36 samples were processed by co-IP, tryptic digestion, fractionation and analysis by LC-LTQ orbitrap tandem mass spectrometry. Peptides identified in at least two of the 36 samples were analyzed by integrating the peak area intensity (PAI) using Skyline to obtain semi-quantitative information. The semi-quantitative data was logarithmic transformed to reduce variation between biological replicates. All the quantitative data were manually inspected for peptides with higher buffer values compared to monomer or oligomer values and these peptides were excluded. To determine conformation preferences among identified ligand interactions a ratio was calculated between monomer signal and oligomer signal for each of the three biological replicates. The significance was tested by Students t-test and significant average fold increase of 2 or more designates conformation preferential binding. 30mg of each of the fractions synaptosomal membranes LP1, synaptosomal lysate (LS1), synaptic vesicles (LP2) and synaptosomal cytosol (LS2) were immunoblottet, and analyzed for the presence of myelin Proteolipid protein (mPLP), Abl interactor 1 (Abl1), Glial fibrillary acidic protein (GFAP), Glutamic acid decarboxylase 2 (GAD2), Synapsin 1 (Syn1), and Vesicle associated membrane protein 2 (VAMP2). (TIF) S1 Table. Semi-quantitative data from mass spectrometric analysis. semi-quantitative analysis of protein binding to alpha-synuclein, subsectioned into three conditions, monomer binding proteins, oligomer binding proteins, and no difference in binding monomer and oligomer including number of identified peptides, number of quantified peptides, and percent sequence coverage. (XLSX)