In Vivo Cross-linking Reveals Principally Oligomeric Forms of α-Synuclein and β-Synuclein in Neurons and Non-neural Cells*

Background: αSyn is central to Parkinsonism, but its native state is unsettled. Results: A new, facile method for cross-linking αSyn in living cells, including neurons, reveals a major 60-kDa form consistent with a tetramer. Cell lysis destabilizes it, yielding mostly monomers. Conclusion: αSyn exists principally as a metastable tetramer in vivo. Significance: Models of native αSyn as an unfolded monomer should be reconsidered. Aggregation of α-synuclein (αSyn) in neurons produces the hallmark cytopathology of Parkinson disease and related synucleinopathies. Since its discovery, αSyn has been thought to exist normally in cells as an unfolded monomer. We recently reported that αSyn can instead exist in cells as a helically folded tetramer that resists aggregation and binds lipid vesicles more avidly than unfolded recombinant monomers (Bartels, T., Choi, J. G., and Selkoe, D. J. (2011) Nature 477, 107–110). However, a subsequent study again concluded that cellular αSyn is an unfolded monomer (Fauvet, B., Mbefo, M. K., Fares, M. B., Desobry, C., Michael, S., Ardah, M. T., Tsika, E., Coune, P., Prudent, M., Lion, N., Eliezer, D., Moore, D. J., Schneider, B., Aebischer, P., El-Agnaf, O. M., Masliah, E., and Lashuel, H. A. (2012) J. Biol. Chem. 287, 15345–15364). Here we describe a simple in vivo cross-linking method that reveals a major ∼60-kDa form of endogenous αSyn (monomer, 14.5 kDa) in intact cells and smaller amounts of ∼80- and ∼100-kDa forms with the same isoelectric point as the 60-kDa species. Controls indicate that the apparent 60-kDa tetramer exists normally and does not arise from pathological aggregation. The pattern of a major 60-kDa and minor 80- and 100-kDa species plus variable amounts of free monomers occurs endogenously in primary neurons and erythroid cells as well as neuroblastoma cells overexpressing αSyn. A similar pattern occurs for the homologue, β-synuclein, which does not undergo pathogenic aggregation. Cell lysis destabilizes the apparent 60-kDa tetramer, leaving mostly free monomers and some 80-kDa oligomer. However, lysis at high protein concentrations allows partial recovery of the 60-kDa tetramer. Together with our prior findings, these data suggest that endogenous αSyn exists principally as a 60-kDa tetramer in living cells but is lysis-sensitive, making the study of natural αSyn challenging outside of intact cells.

Since its discovery as the first causative gene product for Parkinson disease (3) and the major constituent of Lewy bodies (4), ␣-synuclein (␣Syn) 2 has been increasingly implicated as a key pathogenic protein in both sporadic and familial forms of the disorder. ␣Syn has long been thought to occur normally as a natively unfolded monomer of ϳ14.5 kDa (5). Accordingly, recombinant and cell-derived ␣Syn has often been studied using denaturing methods, as it was believed to be unstructured already. In contrast to these assumptions, our laboratory recently used a range of methods to observe that ␣Syn can be found in untransfected neuroblastoma cells and other cell lines and in human erythrocytes in apparent tetrameric assemblies that, upon purification under non-denaturing conditions, were found to contain substantial ␣-helical structure (1). An independent report at the same time described a protocol to recover bacterially expressed ␣Syn as a helically folded tetramer and provided an NMR structure supporting this model (6). A third laboratory has also reported evidence that bacterially expressed ␣Syn can form helically folded oligomers depending on the purification method (7), and a fourth has successfully reproduced our purification of folded, oligomeric ␣Syn from erythrocytes (8).
Nevertheless, these new findings have been controversial. For example, a recent paper reported experiments that suggested to the authors that ␣Syn existed almost exclusively as unfolded monomers in various cell lines or after purification from erythrocytes (2). Since we published our original findings (1), we have conducted extensive experiments to better delineate the assembly state of native ␣Syn in cells. We recognized that most of the methods we had used to identify native ␣Syn tetramers, including sedimentation equilibrium analytical ultracentrifugation, scanning transmission electron microscopy, and native gel electrophoresis (1), are not facile and thus are difficult for investigators to apply to the rapid analysis of the assembly state of ␣Syn in cells and tissue. The lack of an easy, consistent method to observe native ␣Syn oligomers in cells has made it challenging to obtain clarity about their existence and properties.
We have now developed a simple and reproducible method employing in vivo cross-linking that readily enables us to detect the apparent assembly state of ␣Syn in intact cells. Using this method and employing extensive controls, we show here that the major form of endogenous ␣Syn in several different cell types, including primary neurons, is an oligomer of ϳ60 kDa consistent with the size of a tetramer. The method also traps smaller amounts of ␣Syn species migrating at ϳ80 and ϳ100 kDa on SDS-PAGE that have the same isoelectric point as the 60-kDa putative tetramer and may thus be conformationally distinct homo-oligomers. Surprisingly, standard lysis of the cells followed by the same cross-linking protocol applied in vitro yields predominantly free monomers plus some of the 80-kDa oligomer, with marked destabilization of the 60-kDa apparent tetramer. However, if the lysis protocol is modified to maintain high protein concentrations, the 60-kDa tetramer is preserved in a concentration-dependent manner. These and additional findings herein are consistent with the existence of metastable oligomers that principally size as tetramers in intact, normal cells, in accord with the model proposed by Bartels et al. (1) and Wang et al. (6). Our findings have important implications for properly studying endogenous, native ␣Syn inside and outside of intact cells, and for modeling ␣Syn misfolding and pathogenic assembly in brain disease.
Primary neurons were cultured from E18 Sprague-Dawley rats (Charles River, Wilmington, MA). Rats were euthanized with CO 2 followed by cervical dislocation. Embryonic cortices were isolated and dissociated with trypsin and trituration. Cells were plated in DMEM supplemented with 5% FBS, 10 units/ml penicillin, 10 g/ml streptomycin, and 2 mM glutamine at 680 cells/mm 2 on BioCoat poly-D-lysine-coated culture dishes (BD Biosciences). After 4 h, medium was changed to Neurobasal medium supplemented with B-27, 2 mM GlutaMAX, and 100 g/ml gentamicin. Half of the medium was replaced every 4 days; on the first medium change, 5-fluoro-2Ј-deoxyuridine (Sigma) and uridine (Sigma) were added to concentrations of 100 and 500 g/l, respectively, to inhibit glial cell growth.

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
37°C with rotation unless stated otherwise. The reaction was quenched with the addition of 1 M Tris, pH 7.6, to 50 mM final concentration and incubated for 15 min at RT. After quenching, in vivo cross-linked samples were lysed by a 15 s sonication (Sonic Dismembrator model 300, Fisher, using a microtip at a setting of 40) then ultracentrifuged 30 min at 213,000 ϫ g unless stated otherwise. For in vitro cross-linking, the cells were first lysed by a 15 s sonication and then either cross-linked directly and ultracentrifuged (crude) or else ultracentrifuged before cross-linking just the supernatant (pure cytosol). To generate total protein lysates (that also contain membrane proteins), the detergent Triton X-100 (TX-100) was added after cell lysis at a final concentration of 1% (v/v) followed by 10 min of incubation on ice and centrifugation.
Immunoprecipitation, Gel Electrophoresis, and Immunoblotting-All materials were purchased from Invitrogen unless stated otherwise. Total protein concentrations were determined by BCA protein assay (Thermo Scientific) according to the manufacturer's directions. Samples were prepared for electrophoresis by dilution with the respective lysis buffer, the addition of 4ϫ NuPAGE LDS sample buffer, and boiling for 10 min. If not stated otherwise, 30 g of total protein were loaded per lane. Immunoprecipitation was carried out by incubating 3-5 g/l (protein concentration) cytosolic lysates with anti-FLAG M2 affinity gel beads (Sigma) at 4°C for 3 h followed by 3 brief washes with PBS/protease inhibitor (PI), and boiling in 1ϫ sample buffer. An equivalent to 300 -500% of the starting material was loaded. Samples were electrophoresed on NuPAGE 4 -12% Bis-Tris gels with NuPAGE MES-SDS running buffer and the SeeBlue Plus2 molecular weight marker. After electrophoresis, gels were electroblotted onto Immobilon-P 0.45-m or Immobilon-PSQ 0.2-m PVDF membranes (Millipore) for 1 h at 400 mA constant current at 4°C in 25 mM Tris, 192 mM glycine, 20% methanol transfer buffer. After transfer, membranes were incubated in 0.4% paraformaldehyde, PBS for 30 min at RT (11), rinsed twice with PBS, stained with 0.1% Ponceau S in 5% acetic acid, rinsed with water, and blocked in 5% nonfat milk in PBS containing 0.1% (v/v) Tween 20 (PBS-T) for either 30 min at RT or overnight at 4°C. After blocking, membranes were incubated in primary antibody in 5% milk in PBS-T with 0.02% sodium azide for either 1 h at RT or overnight at 4°C. Membranes were then washed 3 times for 10 min in 1% milk in PBS-T at RT and incubated in secondary antibody in 1% milk in PBS-T for 45 min at RT. Membranes were then washed 3 times for 10 min in PBS-T and developed with ECL Plus or ECL Prime (GE Healthcare-Amersham Biosciences) according to the manufacturer's instructions.
Sequential Protein Extraction-After treatment with DSG or else DMSO (vehicle) alone in PBS/PI without TX-100, cells were lysed by sonication and centrifuged at 435,000 ϫ g for 30 min, and the supernatant was collected (PBS cytosolic fraction). The pellet was resolubilized in PBS/PI with 1% TX-100 and again centrifuged at 435,000 ϫ g for 30 min. The resulting supernatant was collected (TX fraction: membrane proteins), and the pellet was boiled in 1ϫ NuPAGE sample buffer containing 2% lithium dodecyl sulfate (LDS fraction: LDS-soluble proteins).
Two-dimensional Gel Electrophoresis-All materials were purchased from Invitrogen unless stated otherwise. Samples were first electrophoresed under non-denaturing conditions on Novex pH 3-7 isoelectric focusing gels, fixed according to manufacturer's recommendations, Coomassie-stained (GelCode Blue, Thermo Scientific), and prepared for SDS-PAGE according to the manufacturer's directions. The isoelectric focusing lanes were then loaded onto NuPAGE Novex 4 -12% Bis-Tris gels with two-dimensional wells, electrophoresed, and blotted as above. Incompletely cross-linked samples were used for twodimensional gel analyses as this strongly reduced the background of WB signals and made the detection of cross-linking intermediates as well as endproducts possible. Incomplete in vivo cross-linking was achieved by incubating the cross-linking reaction at 4°C.
Preparation and Cross-linking of Recombinant ␣Syn-Recombinant ␣Syn (wt ␣Syn) was prepared, purified, and desalted as described (12). Pure recombinant ␣Syn was cross-linked in vitro at a concentration of 10 ng/l in PBS/PI using a range of DSG concentrations. As a control, pure egg lysozyme (MP Biomedicals) was cross-linked under identical conditions. For in vivo cross-linking, bacteria were pelleted, washed, cross-linked, quenched, and lysed analogously to mammalian cells using a range of DSG concentrations.

Cross-linking of Intact Cells Reveals Endogenous ␣Syn in Distinct Higher Molecular Weight Species with a Size
Range of 60 -100 kDa-To assess the physiological assembly states of ␣Syn in a natural cellular environment, i.e. in the cytoplasm of intact cells, we sought both the least invasive method and a readily cultured cell line with substantial endogenous expression of human ␣Syn. Primary neurons are an attractive system in which to study an abundant neuronal protein like ␣Syn, but variations in primary cultures and the fragility and limited quantity of these cells make the extensive biochemical optimization of a new protocol difficult. Immortalized cells are more available and easily handled, but even lines of neuronal origin such as SK-N-MC or M17D human neuroblastoma cells have low endogenous ␣Syn expression compared with primary neurons (Fig. 1A), making detection difficult and raising questions about the physiological relevance of potential findings. Common solutions for the detection problem such as overexpressing ␣Syn were avoided here whenever possible, as the natural steady-state condition of ␣Syn in such cells might be altered. Fresh erythrocytes, an abundant source of endogenous ␣Syn (13,14), express high levels of hemoglobin that can interfere with ␣Syn detection by immunoblotting due to similar gel migration. We reasoned that if erythrocytes are rich in ␣Syn, their progenitor cells may also be, and we thus identified the human erythroleukemia cell line HEL (15) as our initial system for development of a cross-linking protocol. HEL cells combine certain advantages of immortal cells and primary neurons; they are easy to culture and express abundant endogenous human ␣-synuclein ( Fig. 1A) but have very low levels of hemoglobin (not shown).
Regarding possible methods to assess ␣Syn assembly state in vivo, we considered a FRET/fluorescence lifetime imaging microscopy approach that has been used to detect interactions of ␣Syn molecules in intact cells (16). However, this requires exogenous expression of tagged ␣Syn, raising concerns about any aggregation induced by overexpressing and/or modifying the protein. Also, the method cannot define the size of oligomers exhibiting FRET. An alternative approach is to trap native assemblies through the in vivo cross-linking of endogenous, unlabeled ␣Syn with small, cell-permeant cross-linkers. Therefore, to define the assembly state of native ␣Syn in living HEL cells, we screened several membrane-permeable cross-linkers with differing spacer lengths. We initially tested the aminereactive DSS (spacer length 11.4 Å) used in our earlier study (1) and the additional amine-reactive cross-linkers DSG (7.7 Å) and DSP (12.0 Å). All 3 compounds were initially tried at a concentration of 1 mM per the manufacturer's recommendations, and the effectiveness of the cross-linking was analyzed by immunoblotting. The expected shift of a portion of cellular proteins toward the higher M r range, as visualized by Ponceau staining of PVDF blots (Fig. 1B, first panel), and the successful cross-linking of a portion of endogenous DJ-1 in its known dimeric form (17)(18)(19) (Fig. 1B, fourth panel) confirmed the general effectiveness of the three cross-linkers in the HEL cells. Negative controls to rule out nonspecific cross-linking of cellular monomeric proteins into artificial oligomers by our protocol included endogenous Ran and ␤-tubulin, each of which remained monomeric after applying the DSG protocol to the same HEL cells ( Fig. 1B; second and third panel). The ␣Syn mAb 15G7 (9) detected a predominant ϳ60-kDa species and a corresponding decrease in the levels of free monomer after in vivo (intact cell) application of each of the three cross-linkers compared with vehicle alone (Fig. 1B, fifth and sixth panels), consistent with our previous ␣Syn cross-linking results using DSS (1). The invariant observation of this endogenous ϳ60-kDa species by in vivo cross-linking throughout the current study is consistent with the existence of tetrameric ␣-synuclein in cells (1, 6); ␣Syn purified from human erythrocytes had previously FIGURE 1. HEL cells express high levels of endogenous ␣Syn that can be specifically cross-linked in vivo as a major ϳ60-kDa species and minor species that include ϳ80 and ϳ100 kDa. A, cytosols (post-20,000 ϫ g) from HEL cells, rat primary cortical neurons (DIV13), and M17D and SK-N-MC human neuroblastoma cells were immunoblotted with the ␣Syn mAb Syn1. Identical exposures of the same blot are shown; film was cut at the white line. B, HEL cells underwent in vivo cross-linking using the cell-permeant amine-reactive cross-linkers DSG, DSS, and DSP (each at 1 mM). Control cells (Ϫ) were treated with DMSO vehicle alone. Cytosols were generated by lysing cells via sonication for 15 s in PBS/PI and centrifuging at 20,000 ϫ g. As controls for cross-linking efficiency, blots were stained with Ponceau solution or a rabbit pAb (10) to the dimeric cytosolic protein DJ-1. As negative controls, the monomeric GTPase Ran as well as ␤-tubulin were detected using specific antibodies. ␣Syn was detected using rat mAb 15G7; short and long exposures of the same blot are shown. C, HEL cells were subjected to in vivo cross-linking with the cleavable cross-linker DSP (1 mM), and cytosols were prepared (post-20,000 ϫ g). For lanes labeled DSP/␤ME (␤-mercaptoethanol), the DSP cross-linker was cleaved after cross-linking by boiling in sample buffer with 5% (v/v) ␤-mercaptoethanol, whereas lanes labeled DSP were boiled in sample buffer alone. General cross-linking efficiency was visualized by Ponceau (Ponc.) staining and by blotting for DJ-1. D, HEL cells underwent in vivo cross-linking using the short spacer length cross-linker DFDNB (1 mM) versus DMSO alone (Ϫ), and 20,000 ϫ g cytosols were prepared. Ponceau staining and blotting for DJ-1 verified the general efficiency of the cross-linker.

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
been sized at 57.8 kDa by sedimentation equilibrium analytical ultracentrifugation (mass of the cellular N-acetylated monomer ϭ 14,502 daltons (1)). Longer exposure of the blot revealed additional ␣Syn immunoreactive species of ϳ80 and ϳ100 kDa in size (Fig. 1B, sixth panel) that were less abundant than the 60-kDa species. This pattern of three principal mid-M r ␣Syn bands accompanied by decreased levels of free monomers upon in vivo cross-linking was confirmed by blotting with several different ␣Syn antibodies ( Fig. 2A and see below). The specificity of the cross-linking was further supported by the fact that the three species (designated ␣S-60, ␣S-80, and ␣S-100 hereafter) reverted to monomer (␣S-14) upon ␤-mercaptoethanolinduced cleavage of the reducible cross-linker DSP (Fig. 1C, right panel), just as occurred with the known DJ-1 dimer in the same cells (Fig. 1C, middle panel). Interestingly, the short spacer length (3.0 Å), amine-reactive cross-linker DFDNB was unable to trap the ␣S-60, -80, and -100 species (Fig. 1D, right panel), although its general ability to cross-link some cellular proteins was shown by Ponceau-staining of the blot (Fig. 1D, left panel) and partial trapping of the endogenous DJ-1 dimer (Fig. 1D, middle panel). This negative result with DFDNB suggests a defined molecular structure (i.e. a particular spacing of the reactive amine groups) in the cross-linked ␣Syn assemblies, further supporting the specificity of the results. Because DSG appeared to be the most efficient of the three cross-linkers in these initial experiments as judged by the relative ␣Syn oligomer-to-free-monomer ratio (Fig. 1B, fifth and sixth panels), we used DSG in subsequent experiments.
We found that cross-linking the cells at the physiological temperature of 37°C was substantially more efficient than at room temperature (supplemental Fig. S1A; note the ratios of putative oligomers to free monomer). Applying a DSG concentration of 1 mM to the cells routinely allowed good visualization of the cross-linked endogenous ␣Syn species, and these were more pronounced at 2 mM with less free monomers (supplemental Fig. S1B , right panel). DSG at 5 mM resulted in a marked decrease or loss of all ␣Syn bands and mid-to-high M r smearing and/or gel-excluded ␣Syn immunoreactivity (supplemental Fig. S1B , right panel), indicating the occurrence of nonspecific over-cross-linking. In contrast, DJ-1 behaved differently, as a DSG concentration of 5 mM allowed ϳ90% of the protein to be trapped in its native dimeric state (supplemental Fig. S 1B , middle panel), with nonspecific high M r smears only appearing at concentrations Ͼ5 mM (not shown). However, the cross-linking behavior of ␣Syn was not unique, as we observed a general loss of focused protein bands and an increase in high M r smears and gel-excluded material on Ponceau-stained membranes when cross-linker concentrations were Ͼ1 mM (supplemental Fig. S1B , left panel). In further optimization experiments, cross-linking was most efficient when the volume of the crosslinking solution was 10 -20-fold higher than the volume of the cell pellet; a ratio of 4 ϫ 10 6 HEL cells in 200 l of 1 mM DSG in PBS ϩ PI mixture incubated at 37°C for 30 min consistently led to the detection of high amounts of ␣S-60, -80, and -100 and accordingly low levels of free monomer (see, e.g. supplemental Fig. S1C , right panel , second lane).
Lee and Kamitani (11) recently described the use of 0.4% paraformaldehyde treatment of blots to enhance the immunodetection of ␣Syn monomers. We found that such treatment of membranes after transfer was important to allow quantitatively meaningful results with cross-linking, as ␣Syn monomers, but not the cross-linked 60 -100-kDa species, were otherwise largely lost during the process of Western blot development (supplemental Fig. S1C). Without paraformaldehyde treatment of the blots, the relative level of apparent oligomers was overestimated due to the greater susceptibility of ␣Syn monomers to wash off the membrane during development (11). Thus, only when blots were paraformaldehyde-treated did the combined FIGURE 2. Multiple ␣Syn antibodies confirm the specificity of the ϳ60-, ϳ80-, and ϳ100-kDa species, all of which are enriched in the high speed cytosol fraction. A, HEL cells were treated in vivo with DMSO alone (Ϫ) or 1 mM DSG (ϩ), sonicated in PBS/PI, and spun at 435,000 ϫ g for 30 min. The resultant cytosols were blotted with ␣Syn antibodies 15G7, 2F12, C20, Syn1, 211, or LB509 and with DJ-1 antibody (10) as a control for cross-linking efficiency. Note that all six ␣Syn antibodies revealed lower levels of free monomer (14 kDa) after crosslinking. See "Results" for details. The asterisk marks a crossreactive band detected by the Syn1 antibody. B, shown are TX-100 total lysates and sequential extracts of HEL cells treated in vivo with DMSO (Ϫ) or 1 mM DSG (ϩ). Total lysates (left panel) were generated by bringing cells lysed by sonication in PBS/PI to 1% TX-100 (TX) followed by 10 min incubation on ice and ultracentrifugation (435,000 ϫ g, 30 min). For sequential extractions (right three panels), see "Experimental Procedures" and "Results." Enrichment of soluble proteins in the PBS cytosol was confirmed by strong signals for the cytosolic protein DJ-1 (monomer and dimer) and the absence of transferrin receptor (TfR). The TX-100 fractions showed the opposite pattern, consistent with enrichment of membrane proteins. mAb 15G7 revealed co-fractionation of all ␣Syn species with DJ-1 in the cytosol; only minor ␣Syn amounts were detected in the TX and LDS fractions. Identical exposures of the same blot are shown; film was cut at the white lines.
immunoreactivities of all ␣Syn species post-cross-linking approximate the amount of monomer alone pre-cross-linking. Approximating this "conservation of immunochemical matter" was a goal of the optimization of our cross-linking protocol and was achieved in virtually all figures herein.
In Vivo Cross-linking Reveals a Cytosolic 60-kDa Species as the Predominant ␣Syn Form in Intact Cells-To establish the specificity of the ␣Syn bands we detected and to better estimate their cellular levels, we examined a variety of anti-␣Syn antibodies on HEL cells treated with 1 mM DSG versus vehicle alone using the optimized protocol described in the previous section ( Fig. 2A). The well characterized antibodies Syn1, C20, and 15G7 as well as our newly generated mAb 2F12 confirmed that the 14-, 60-, 80-, and 100-kDa bands all contained ␣Syn. However, C20 showed relatively decreased sensitivity for the crosslinked species versus the monomer and stronger "smears" (although ␣S-60 was still the major species). 2F12 or 15G7 did not exhibit preferential reactivity for any of the species and thus were considered suitable to assess the relative levels of the different cellular ␣Syn species. Although both these antibodies led to similar results throughout this study, 15G7 was preferred, as it had been fully characterized before (9) and had slightly less background ( Fig. 2A). In addition, Syn1 reacted similarly to 15G7 (and 2F12), but it showed the known strongly cross-reactive, nonspecific band at ϳ50 kDa (20). Two other widely used ␣Syn antibodies, LB509 and 211, detected the decrease in free monomer levels upon cross-linking but failed to detect the cross-linked putative oligomers ( Fig. 2A), likely due to inaccessibility of their epitopes caused by this cross-linker.
␣Syn is generally believed to be predominantly cytosolic, with a small portion being membrane-associated (21,22). To further characterize the subcellular distribution of the apparent ␣Syn oligomers in intact cells, we performed sequential extractions of whole HEL cell lysates after in vivo cross-linking. We found the ␣S-60, -80, and -100 kDa species to be present almost exclusively in the PBS-soluble extract (cytosol), which was devoid of the transmembrane protein transferrin receptor (TfR) and enriched in the cytosolic protein DJ-1 (Fig. 2B). To rule out the possibility that the ␣Syn species might be associated with remaining (small) vesicles or membrane fragments in the PBS cytosol, we applied a very high ultracentrifugation force of 435,000 ϫ g. Comparison of this very high speed PBS cytosol with total protein extracts made by lysing the cells in PBS, 1% Triton X-100 (Fig. 2B, compare lanes 1 and 2 to lanes 3  and 4) showed no differences in the relative levels of the four principal ␣Syn bands, indicating that all species of ␣Syn detected were overwhelmingly cytosolic. In accord, membrane fractions prepared in the absence or presence of cross-linker (Fig. 2B, lanes 5 and 6) contained only small amounts of principally monomeric ␣Syn. Finally, the subsequent centrifugal fraction soluble in 2% LDS (Fig. 2B, lanes 7 and 8) contained small amounts of high M r ␣Syn near the top of the gel, which may correspond to large, LDS-soluble ␣Syn aggregates in the cells and/or nonspecifically cross-linked ␣Syn molecules. Taken together, all of the findings thus far suggest that a principal form of endogenous ␣Syn in intact HEL cells is a soluble, cytosolic species that migrates at the size of a tetramer (60 kDa) after in vivo cross-linking. In addition, smaller amounts of cyto-solic 80-and 100-kDa cross-linked species and free monomers (14 kDa) are detected.
60-, 80-, and 100-kDa-cross-linked ␣Syn Species Are Homo-oligomers-The results obtained so far are consistent with the recent recognition that ␣Syn can exist in part in homooligomeric, particularly tetrameric, forms (1,6,7). However, the very low abundance of bands that could be interpreted as oligomeric intermediates such as dimers or trimers upon crosslinking raised the alternative interpretation that the 60-kDa band could be a synuclein monomer-bound to an unknown ϳ45-50-kDa interactor or to multiple smaller interactors, i.e. a hetero-oligomer. We reasoned that such a prominent and consistent interactor that is cross-linked to ␣Syn in all of our in vivo cross-linking experiments in different cell types (see below) would likely be one that had already been described in the extensive literature on ␣Syn over the last 20 years. Although there is a long list of single reports of ␣Syn-interacting proteins, most of them have not been confirmed independently. Nevertheless, we examined a virtually complete list of published ␣Syn interactors for their possible presence in the mid-M r ␣Syn bands (especially the major ␣S-60 species) and were able to rule out all of these candidates for a variety of specific biochemical reasons (explained in supplemental Table S1 and its legend). Most importantly, for the cross-linked ␣S-60, -80, or -100 cytosolic species, we found no support for the presence of the most widely accepted interactors: Synphilin-1 (Ref. 23 To address this homo-oligomer question further, we searched for cross-linking conditions that led to some detection of putative dimer/trimer intermediates by Western blotting. We found that performing the in vivo cross-linking reactions at 4°C significantly reduced the efficiency of the cross-linking, as expected, and led to more apparent monomer, less apparent ␣S-60 tetramer, and more ␣S-30 and ␣S-40 bands that we interpreted as ␣Syn dimers and trimers, respectively (Fig. 3C, bottom left panel, and data not shown). This observation could be attributable to partial (incomplete) cross-linking, not necessarily to the existence of significant levels of endogenous dimers/ trimers of ␣Syn in intact cells. Importantly, we also routinely detected significant amounts of apparent ϳ30-kDa dimers by our standard 37°C cross-linking protocol when we overexpressed ␣Syn in the human neuroblastoma cell line, M17D, which has very low endogenous ␣Syn expression ( Fig. 3A; see also Fig. 6D, right panel). This accentuation of dimers was especially the case when C-terminally tagged ␣Syn was overexpressed (Fig. 3D). Together, these results suggest that the efficient oligomerization of ␣Syn fully into tetramers may be regulated by limiting cellular factors and may depend on the presence of an unmodified C terminus.
The conclusion that the cells contain homo-oligomers of ␣Syn was further supported by applying our in vivo DSG cross-linking to Escherichia coli (BL21 strain) expressing human ␣Syn. In the complete absence of eukaryotic proteins, we still observed the ␣S-60, -80, and -100 species in the bacterial cytosols post cross-linking (Fig. 3B, right panels; the left panel shows wild-type (wt) bacteria as a negative control). Given the identical electrophoretic positions of the cross-linked species in ␣Syn-expressing bacteria and normal eukaryotic cells, the theoretical possibility of a bacterial protein forming the same putative hetero-oligomers with ␣Syn as occurs in mammalian cells is extremely unlikely. Our observation of trapping of ␣Syn oligomers in the bacterial cytosol is consistent with the recent reports of the purification of human ␣Syn tetramers from bacteria under certain non-denaturing conditions (6,7). Of interest, the DSG trapping of the ␣S-60 tetramer at the expense of free monomer was less efficient in E. coli than in eukaryotic cells, and more ␣Syn was trapped at a dimer position (␣S-30); this was espe- , suggesting partial intramolecular modification of lysines in the monomer. A similar shift toward the lower pH range was observed for the ␣S-60, -80, and -100 species such that their isoelectric points aligned with the acid-shifted monomers. D, shown is the co-IP analysis of differently tagged ␣Syn molecules after co-expression and cross-linking. ␣Syn-mycHis and ␣Syn-FLAG 3 were either expressed separately in two different cell populations and mixed just before in vivo cross-linking (M) or co-expressed in the same cell population and then mixed with mock-transfected cells just before in vivo cross-linking (C). After the in vivo cross-linking, lysates were subjected to FLAG-IP. Starting materials (lysate) and anti-FLAG IPs (IP M2) were analyzed by WB using specific antibodies for the myc-epitope (A14), the FLAG-epitope (M2), and ␣Syn (15G7). A short exposure of the monomer bands in the anti-myc blot demonstrates the relative depletion of monomers in comparison to the mid-M r species shown in the long exposure of the blot. IP purity was confirmed by the absence of DJ-1 immunoreactivity (left panel) in the IP lanes. The asterisk marks IgG light chain bands. MARCH 1, 2013 • VOLUME 288 • NUMBER 9 JOURNAL OF BIOLOGICAL CHEMISTRY 6377 cially true when the bacterial expression was markedly enhanced by IPTG induction (Fig. 3B, right panel).

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
To further assess the question of homoversus hetero-oligomers of ␣Syn in intact cells, we performed two-dimensional gel analyses on cross-linked ␣Syn species from incompletely (4°C) cross-linked Hel cells (Fig. 3C, large panels on the right; for comparison, representative one-dimensional gels are shown on the left). In the first dimension (x axis), proteins were separated by isoelectric point (isoelectric focusing), whereas the second dimension (y axis) was a standard SDS-PAGE separating proteins primarily by molecular weight. Immunoblot signals with identical isoelectric points on the x axis but different sizes on the y axis can be considered to be monomers and oligomers composed of the same protein. In non-cross-linked samples (Fig. 3C, large top right panel and short exposure below), the large majority of ␣Syn focused as a single spot at 14 kDa (y axis) and around its theoretical isoelectric point of pH 4.67 (x axis), as expected for the free monomer. The analysis of amine-crosslinked ␣Syn is complicated by the fact that the DSG modification of primary amines (lysines) in proteins shifts their isoelectric points to more acidic positions (a cysteine-based cross-linker, which would leave isoelectric points virtually unchanged, cannot be used with ␣Syn as it has no cysteines). Running the in vivo DSG-cross-linked samples on the two-dimensional gels yielded a 14-kDa spot at the expected pH ϳ 4.7 position of unmodified monomers plus some monomers shifted in the acidic direction, suggesting modification of lysine residues by the cross-linker and possibly some intramolecular cross-linking (Fig. 3C, large bottom right panel and short exposure below). For the ␣S-60, -80, and -100 in vivo cross-linked species, we observed a similar stepwise shift toward acidic pH as for the ␣S-14 monomer. Importantly, the ␣S-60, -80, and -100 species in the in vivo cross-linked sample shared their less acidic isoelectric points with the acid-shifted monomers and with faint ␣Syn species of ϳ30 and ϳ40 kDa representing probable dimers and trimers in these incompletely cross-linked samples. The observed array of fully and partially cross-linked 14-kDa and mid-M r species with overlapping isoelectric points as well as the more completely cross-linked higher M r species that are shifted in an acidic direction conforms to the results expected for homo-oligomeric proteins when one combines two-dimensional gels with amine-reactive cross-linking as here. Collectively, these two-dimensional gel results strongly suggest that the mid-M r ␣Syn species detected by the in vivo cross-linking method are homo-multimers of ␣Syn. The presence of a heterologous protein(s) bound to ␣Syn in the ␣S-60, -80, and -100 species is highly unlikely, as this would be expected to shift their isoelectric points to different positions than that of the free monomer.
To obtain further evidence for homo-oligomers, i.e. more than one ␣Syn molecule in the mid-M r bands, we next transiently co-expressed differentially tagged ␣Syn molecules (␣S-FLAG 3 ; ␣S-mycHis) either as pairs together in the same cells or, for negative control purposes, in separate cells that were later mixed together (Fig. 3D). At 48 h post-transfection, all cells were subjected to in vivo cross-linking, lysis, and an immunoprecipitation (IP) of their cytosols for the FLAG tag. The extent of any co-immunoprecipitation (co-IP) of the ␣Syn-mycHis proteins with the ␣S-FLAG 3 was detected by immunoblotting for the myc-tag (Fig. 3D). As we hypothesized, only in the case of co-expression in the same cells did we detect significant co-IP of ␣Syn-mycHis with ␣S-FLAG 3 in the putative oligomeric positions, indicating that the respective mid-M r bands indeed represent oligomers that form inside intact cells and consist of differentially tagged ␣Syn proteins. A similar result was obtained by co-expressing ␣S-FLAG 3 and ␣S-V5 (supplemental Fig. S2). Of interest, the in vivo cross-linking of these highly overexpressed and C-terminally tagged ␣Syn proteins led to more ␣Syn running as monomers that were decreased but not fully absent upon in vivo cross-linking and co-IP, suggesting the presence of some non-cross-linked ␣Syn oligomers in the lysates.
The 60-, 80-, and 100-kDa-cross-linked ␣Syn Species Are Homo-oligomers Distinct from the Products of "Diffusion-controlled" Cross-linking-A recent study reported the detection of ␣Syn predominantly at a dimeric position after the application of cross-linker to intact cells (2), a finding that is not necessarily inconsistent with the findings we present above. Nevertheless, these authors interpreted their observed cross-linking as diffusion-controlled rather than the trapping of actual oligomers existing in the cells under physiological conditions. Yet a diffusion-controlled (nonspecific) association that would lead to abundant, well defined species running principally at the tetramer position with very small amounts of dimers (Figs. 1 and 2) appears improbable. First, if ␣Syn existed solely as monomers in the cytosol (2), then a truly diffusion-controlled mechanism should not favor the interaction of ␣Syn with other ␣Syn molecules but rather would cross-link ␣Syn to numerous different molecules it may randomly contact, which potentially comprises many cytosolic proteins; this would lead to a diffuse smear detectable by WB, not discrete bands. Second, the trapping of ␣Syn at defined dimeric positions (2) and even more so at mostly tetrameric positions (this study) seems an unlikely result for a nonspecific process. To assess what to expect from the diffusion-controlled (nonspecific) cross-linking of unfolded monomeric ␣Syn, as Fauvet et al. (2) proposed, we subjected recombinant ␣Syn that had been conventionally purified under denaturing conditions (12) to in vitro cross-linking using a gradient of DSG concentrations up to 100 M; i.e. up to a crosslinker-to-␣Syn ratio much higher than the 1 mM DSG we used in the cytoplasm of intact cells (Fig. 4A, left panel). We detected a ladder of cross-linker-induced ␣Syn oligomers that appeared to follow a stochastic process and whose pattern was DSG concentration-dependent. Cross-linker concentrations Ͻ30 M trapped mostly monomers plus oligomer bands whose abundance decreased stepwise as their size increased, whereas 100 M led to the detection of less oligomer bands and a large amount of ␣Syn smears in the high M r range (Fig. 4A, left  panel). By systematically varying the ␣Syn and DSG concentrations, we found that 30 M DSG for a 10-ng/l solution of pure ␣Syn was the best condition with regard to optimal trapping of low-and mid-M r ␣Syn species. However, neither this purposefully optimized cross-linking ratio nor any other ratio we tried could generate an ␣Syn band pattern resembling that obtained with intact cell cross-linking, i.e. the prominent 60-kDa tetrameric band (this study) or even the abundant cellular dimers

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
seen by Fauvet et al. (2). Moreover, applying an identical DSG gradient to a solution of the known monomeric protein lysozyme (purified from hen eggs) produced a closely similar, ladder-like pattern of oligomers (Fig. 4A, right panel), consistent with the conclusion that the pattern we see with pure ␣Syn in vitro is indeed the result of diffusion-controlled (random) events leading to the nonspecific cross-linking of a monomeric protein. Strikingly, adding only a 10-fold excess of pure lysozyme to our solution of pure recombinant ␣Syn before cross-linking led to ␣Syn-immunoreactive bands at altered positions that constitute hetero-oligomers of ␣Syn and lysozyme (Fig. 4B, right lane). This result suggests that the unfolded ␣Syn monomer does not favor in vitro interactions with other ␣Syn molecules over interactions with other random proteins it may contact. Based on this result, unfolded ␣Syn monomers in the cellular context might be expected to be cross-linked to other random proteins rather than preferentially to other ␣Syn molecules at defined positions, as we invariantly observed throughout the current study. Moreover, our cross-linking in intact cells leads to an oligomer pattern that is not remotely in agreement with the results of nonspecific (diffusion-controlled) cross-linking of free monomers, as established in Fig. 4A, left panel. We directly compared the results of the nonspecific cross-linking of pure, unfolded ␣Syn monomers at optimal DSG concentration (30 M) to the cross-linking of endogenous ␣Syn in the cytosol of HEL cells (Fig. 4C). Obvious differences include (a) the virtual absence of intermediate dimers and trimers in the HEL cells, (b) the strong preference for the ϳ60-kDa ␣Syn tetramer in HEL cells, although this position is not favored with nonspecific cross-linking in vitro, (c) the cellular tetramer migrating slightly faster than the pure tetramer generated by cross-linking the E. coli-expressed ␣Syn, suggesting that the former species sizes as a tetramer, but its precise manner of cross-linking may be slightly different from that of the analogous non-specifically cross-linked tetramer, and (d) the presence of the aS-80 and aS-100 species in the HEL cells that do not overlap with any of the products of the nonspecific cross-linking of pure unfolded monomers. Based on our extensive findings on the intracellular ␣Syn oligomers presented so far (e.g. two-dimensional gels, co-IP data, comparisons to cross-linking of pure monomers, ruling out of potential candidate interactors), the presence of heterologous protein(s) bound to ␣S-80 and -100 species is highly unlikely. Rather, these in vivo cross-linked species may differ from the predominant tetrameric ␣S-60 as to (a) the number of ␣Syn subunits (e.g. they are tetramers, hexamers, and octamers) or (b) different native conformations of the same-sized oligomer (e.g. the tetramer) that are thus trapped slightly differently by the DSG (i.e. they are conformers of each other). Regarding the latter possibility, we treated an in vivo cross-linked sample with a harsh denaturant (5 M urea) but still detected the same ␣Syn population of 60-, 80-, and 100-kDa bands (Fig. 4D), suggesting that if these are conformers, the DSG cross-linking precludes their subsequent denaturation by 5 M urea.
The in Vivo Interactions of Cellular ␣Syn Are Sensitive to Cell Lysis-Having examined ␣Syn species in intact cells by in vivo cross-linking, we next asked to what extent these species can also be cross-linked in cell lysates (in vitro). To this end we modified our in vivo cross-linking protocol in just one way; cell lysis occurred before rather than after cross-linking (Fig. 5B). All other parameters described earlier were left unchanged to  Plus2). B, a schematic is shown of in vivo and in vitro cross-linking protocols (see "Experimental Procedures" and "Results" for details). C, increasing DSG concentration does not overcome the inability to trap ␣Syn species other than ␣S-80 in vitro. 4 ϫ 10 6 HEL cells in 200 l PBS/PI were lysed by sonication, treated with 1, 1.5, 2, or 5 mM DSG and spun at 213,000 ϫ g. D, in vitro cross-linking in lysates of high protein concentration partially recovers the in vivo ␣Syn cross-linking pattern by stabilizing ␣S-60. Concentrated (15 g/l) lysates were generated by lysing HEL cells in small volumes of either PBS/PI (two panels on the left) or PBS/PI/1% TX-100 (two panels on the right) followed by centrifugation at 213,000 ϫ g; decreasing concentrations were generated by diluting this 15 g/l sample. After cross-linking at the indicated protein concentrations, all samples were normalized to 5 g/l, and equal volumes were loaded. Of note, oligomerization intermediates, the putative dimer ␣S-30, and the putative trimer ␣S-40 could be detected in some of the sample to variable degrees. enable meaningful comparison. The in vitro cross-linking was usually done on crude cell lysates (i.e. cross-linker was added immediately after cell lysis without centrifugation), but no major differences were observed if we first ultracentrifuged the lysates and cross-linked just the high speed cytosol. Coomassiestained SDS-PAGE gels of vehicle-only and DSG treatments performed in vivo and in vitro confirmed the general efficiency of both cross-linking approaches, as indicated by a similar shift of many proteins to the higher M r range (Fig. 5A, first panel). Importantly, several control proteins, DJ-1 (17)(18)(19), voltagedependent ion channel (28,29), GAPDH (30), Drp-1 (31), and HSP-70 (32), which are known to exist endogenously as oligomers in cells, could each be successfully cross-linked in their respective oligomeric states by both in vivo and in vitro crosslinking, although some modest protein-specific differences were observed (Fig. 5A, second to sixth panels). In striking contrast, cross-linking of ␣Syn in vivo versus in vitro revealed a major difference; the former trapped the now-familiar pattern of ␣S-60, -80, and -100 (and some very minor species) with low levels of monomers, whereas in vitro cross-linking led to the detection of mostly free monomers and a small amount of just the 80-kDa band (Fig. 5A, seventh panel). This observation is consistent with our finding that the efficient co-IP of differentially tagged ␣Syn species depended on cross-linking in vivo, whereas little interaction was detected without in vivo crosslinking (see Fig. 3D). Intriguingly, the ability to trap ␣S-80 by cross-linking in vitro was markedly decreased or abolished if the cell lysates were sonicated for 30 or 60 s (rather than our standard 15 s) before applying the DSG cross-linker (supplemental Fig. S3A , right panel), suggesting that the 80-kDa species is highly sensitive to the heat and/or shearing induced by sonication. DJ-1 analyzed simultaneously in the same cells behaved as expected; it was trapped in its endogenous dimeric state as efficiently by in vitro as by in vivo cross-linking, and sonication before in vitro cross-linking had no significant adverse effect (supplemental Fig. S3A, left panel). When we spiked recombinant human ␣Syn monomer into the HEL lysates before in vitro cross-linking, this pure, exogenous 14-kDa monomer specifically contributed to (i.e. augmented) the endogenous 80-kDa band, supportive of the interpretation that the endogenous ␣S-80 species is composed of just ␣Syn (supplemental Fig. S3B).
We consistently observed that the ␣S-60 species trapped by in vivo cross-linking was highly sensitive to cell lysis by various methods (e.g. liquid nitrogen with or without protease inhibitors; sonication) performed before in vitro cross-linking, whereas the ␣S-80 species persisted, albeit at substantially lower levels and accompanied by more free monomer than was observed in vivo (supplemental Fig. S3C). Raising the crosslinker concentration did not overcome the inability to recover ␣S-60 in vitro, i.e. we could detect only ␣S-80 and ␣S-14 by cross-linking after cell lysis (Fig. 5C). A possible explanation for this lability of the major in vivo species upon cell lysis in a standard buffer volume could be a dependence of the normal interactions of intracellular ␣Syn on high macromolecular concentrations. To test this hypothesis, we repeated the in vitro cross-linking approach after lysing the cells in small buffer volumes, i.e. at much higher protein concentrations, to better sim-ulate the intracellular macromolecular crowding that obtains during in vivo cross-linking. Through a stepwise increase in protein concentration above the conventional 1-6 g/l in which cells were routinely lysed, we observed a striking concentration-dependent recovery of ␣S-60 in the more highly concentrated lysates whether this was done in PBS cytosols or in TX-100/PBS total lysates (Fig. 5D). Under these conditions, ␣S-60 became more abundant than the ␣S-80 that was the sole mid-M r species recovered after in vitro cross-linking in conventional dilute buffer (Fig. 5D). Also, the pattern obtained by this in vitro cross-linking under protein crowding generally resembled that of in vivo cross-linking and looked nothing like the ladder of oligomers that pure ␣Syn formed under diffusioncontrolled (nonspecific) cross-linking in vitro (Fig. 4A). In sharp contrast to the results with ␣Syn, higher protein concentrations led to decreased efficiency in trapping DJ-1 in its dimeric state, probably due to the decreased cross-linker-toprotein ratio. Even at the highest protein concentrations we could test, the ␣Syn pattern upon in vitro cross-linking was still somewhat different from that of in vivo cross-linking in that the lowered level of free monomers routinely seen in vivo (e.g. Fig.  5A, far right panel) could not be achieved in vitro (Fig. 5D). Collectively, these findings support the hypothesis that cellular ␣Syn exists in proteinaceous solution in assemblies requiring specific intermolecular interactions, with high macromolecule concentrations favoring the occurrence or stabilization of those interactions.
In Vivo Cross-linking of Primary Neurons Expressing ␣and ␤-Synuclein Yields Closely Similar Putative Oligomers as Those in HEL Cells-To assess the relevance of the above findings on endogenous ␣Syn in HEL cells to endogenous ␣Syn in neurons, the target cells in human synucleinopathies, we analyzed primary rat neurons as well as the human neuroblastoma cell line, M17D. In agreement with earlier studies (33), rat primary neurons needed to be cultured for 7 days or longer to achieve substantial ␣Syn expression, which then was in the range of HEL cells (Fig. 1A). We routinely plated primary cortical rat neurons at high density (ϳ12 ϫ 10 6 cells per 10-cm dish) and harvested them on day 13 in vitro (DIV13). Similar to the HEL cells, neurons treated with 1 mM DSG in vivo underwent the trapping of a major ␣S-60 and minor ␣S-80 and ␣S-100 species and a much lower level of free monomers in the cytosol (Fig. 6A, fifth panel). Just as in the HEL cells, the in vitro treatment of neuronal lysates with DSG led to the detection of only the ␣S-80 species and more free monomer (Fig. 6A, fifth panel). Immunoblotting for synaptobrevin-2 (Sbr-2) in a TX-100 total lysate confirmed the neuronal nature of the cells (Fig. 6A, first panel, first lane), whereas the absence of any Sbr-2 immunoreactivity in all the neuronal cytosol samples showed that the latter were devoid of membranes, as expected. The detection of DJ-1 dimers after in vivo and in vitro DSG cross-linking (Fig. 6A, third panel) once again confirmed the general effectiveness of both cross-linking protocols. We also probed the neurons for the known monomeric protein Parkin and observed only the monomer after either in vivo or in vitro cross-linking (Fig. 6A, second panel), again underscoring the lack of any evidence that the ␣Syn oligomers we detect are artificially induced by the cross-linking method. As in HEL cells, neuronal ␣Syn was similarly abundant MARCH 1, 2013 • VOLUME 288 • NUMBER 9

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
in cytosol and TX-100 total lysates (Fig. 6B, second panel, compare first and second lanes). Sbr-2 was again exclusively found in total lysate, not cytosol (Fig. 6B, first panel). After in vivo crosslinking and lysis of the neurons in the presence of TX-100, we found a closely similar pattern of the ␣Syn mid-M r species (Fig.   6B, second panel, third lane) as we saw in the cytosol (Fig. 6A,  fifth panel). Here, all ␣Syn species could be cleanly detected by mAb Syn1, whose cross-reactive ϳ50-kDa band was virtually absent in these DIV13 neurons. Blotting the in vivo cross-linked neuronal samples for Sbr-2 revealed its monomer and trace FIGURE 6. Cross-linking of endogenous ␣and ␤-synuclein in primary neurons and endogenous and transfected ␣Syn in M17D cells validates the findings in HEL cells. A, shown is an immunoblot analysis of cytosols (213,000g) after DSG cross-linking of rat neurons and M17D cells. Primary cortical rat neurons (DIV13) were cross-linked in vivo or in vitro just as with HEL cells. pAb to Sbr-2 (left panel) detected a strong signal in a total protein lysate (to.) but not in cytosols. Blotting for DJ-1 confirmed the equal efficiency of in vivo and in vitro cross-linking. pAb EP1537Y was used to probe for ␤Syn, and mAb 2F12 was used for ␣Syn. B, shown is an immunoblot analysis of 1% TX-100 total lysates (213,000 ϫ g) after in vivo DSG cross-linking of primary cortical rat neurons (DIV13). For comparison, a cytosol (cy.) is shown. Unlike Sbr-2, which was absent in cytosol, ␣Syn was equally present in cytosolic (cy.) and TX-100 total lysates (total). A weak ϳ35-kDa band (arrowhead) was detected for Sbr-2 after cross-linking. The asterisks marks the nonspecific band detected by the antibody (see also Fig. 2B). C, rat neurons (DIV13) were cross-linked with increasing concentrations of DSG (0 -5 mM) and immediately lysed in 2% LDS sample buffer. The asterisk marks a presumably nonspecific band detected by the antibody in rat neuronal lysates; g.e., gel-excluded. D, Wt ␣Syn and the NAC-domain mutant ⌬71-82 were transiently expressed in M17D cells for 48 h followed by in vivo cross-linking with DSG (right panels in each pair) or just DMSO (left panels in each pair). As a control, vector only was transfected (mock). DJ-1 was a control for cross-linking efficiency and equal loading (left panel). Identical exposures of the same blot are shown; film was cut at white lines.

␣-Synuclein Occurs as Soluble Oligomers in Living Cells
amounts of a 35-kDa band (arrowhead) that did not co-migrate with any of the cross-linked ␣Syn species (Fig. 6, A and B, first  panels). This result rules against the possibility that a recently proposed interaction between ␣Syn and Sbr-2 (25) is responsible for our detection of the mid-M r ␣Syn species; moreover, the latter occurs in the cytosol (Fig. 6A).
The human neuronal cell line M17D had low endogenous ␣Syn expression levels (Fig. 1A), but the findings from HEL cells and primary neurons, ␣S-60, -80, and -100 after in vivo crosslinking and just ␣S-80 after in vitro cross-linking, were again observed (Fig. 6A, sixth panel).
An advantage of the primary neurons over M17D and HEL cells was their significant endogenous expression of ␤Syn, enabling us to probe for any apparent oligomeric species of this partial homologue. In agreement with published data (33), we found ␤Syn monomer running slightly higher than ␣Syn monomer on SDS-PAGE despite the fact that ␤Syn is 6 residues shorter (134 versus 140 amino acids). This slightly higher gel migration as well as our inability to immunodetect any ␤Synreactive bands in lysates of M17D cells transiently overexpressing abundant ␣Syn (not shown) confirmed that our ␤Syn antibody does not detect ␣Syn. Using this specific antibody, a gel pattern strikingly reminiscent of that of ␣Syn was seen for endogenous ␤Syn in primary neurons undergoing DSG crosslinking; this is, a principal ϳ60 kDa ␤Syn species with an associated decrease in monomer levels upon in vivo cross-linking and the sensitivity of this 60-kDa species to cell lysis when in vitro cross-linking was performed (Fig. 6A, fourth panel). In contrast to HEL and M17D cells, only a very weak 80-kDa ␣Syn band was detected in neurons (especially upon in vitro crosslinking), and the analogous ␤Syn band was barely detectable (Fig. 6A, compare the three right-most panels), suggesting that the putative oligomer patterns can vary subtly as a function of cell type and synuclein homolog. Because it is known that ␤Syn does not share the aggregation propensity of ␣Syn or play a pathogenic role in Parkinson disease (4,34,35), our closely similar in vivo and in vitro cross-linking data on ␤Syn and ␣Syn suggest that the ␣Syn oligomers we detect are neither the result of a pathological aggregation process nor induced by the crosslinker but rather occur normally in neurons.
In an attempt to better assess the relative abundance of the in vivo cross-linked ␣Syn species in primary neurons, we tested increasing concentrations of DSG from 0 to 5 mM (applied at our standard ratio of ϳ4 ϫ 10 6 cells per 200 l of cross-linking solution). Also, in order not to miss any of the possible cellular ␣Syn species in this analysis, we lysed the cells after the in vivo cross-linking by boiling them directly in 2% LDS sample buffer, thus generating a total protein extract potentially also containing more insoluble species. We observed that increasing DSG concentrations led to a dose-dependent depletion of monomers but also to increasing amounts of gel-excluded ␣Syn immunoreactivity at higher DSG concentrations (Fig. 6C, right panel). This was in agreement with Ponceau staining of the blots, which showed a DSG dose-dependent increase in gel-excluded proteins in general (not shown) as well as with results obtained for DJ-1 (Fig. 6C, left panel). The highest degree of detection of the mid-M r ␣Syn species (accompanied by low levels of free monomers) in neurons occurred at concentrations of 1-2 mM DSG (Fig. 6C, right panel), closely similar to what had been seen in HEL cells (compare with supplemental Fig. S1C). Thus, the data indicate that the 60-kDa putative tetramer is the prevalent endogenous ␣Syn species in primary cortical neurons.
To provide further evidence that the endogenous ␣Syn oligomers we detect are not due to pathogenic ␣Syn aggregation, we transiently expressed in M17D cells either wt human ␣Syn or an artificial ␣Syn mutant ⌬71-82 that lacks a 12-amino acid stretch in the NAC domain shown to be required for ␣Syn aggregation in vitro (36) and performed an in vivo cross-linking analysis. For both wt and ⌬71-82 ␣Syn, DSG treatment resulted in lower levels of the monomer and the usual 60 (major)-, 80-, and 100-kDa mid-M r species (Fig. 6D, middle and  right panels). This lack of difference in the in vivo pattern of ⌬71-82 versus wt ␣Syn, like the analogous pattern seen for the (non-pathogenic) endogenous ␤Syn, supports the physiological (not pathological) nature of the putative oligomeric ␣Syn forms we invariably detect upon in vivo cross-linking of endogenous ␣Syn.

DISCUSSION
Here we systematically searched for endogenous assemblies of ␣Syn by applying cell-permeant cross-linkers to living cells. Using HEL cells that express high endogenous levels of ␣Syn as a new model system initially, we conducted extensive experiments to establish an optimized protocol based on the amine cross-linker DSG. We report that ␣Syn exists in intact cells, including HEL and primary cortical neurons, principally as mid-M r species of ϳ60, 80, and 100 kDa that can be detected by in vivo cross-linking, accompanied by low levels of free monomers. A 60-kDa ␣Syn species was the most abundant of the mid-M r species and was the major form among all ␣Syn-immunoreactive species detected in living cells. These new data are consistent with recent reports that ␣Syn can exist as a physiological tetramer in cells (1,6), as the combined molecular mass of four ␣Syn monomers (14,502 daltons each) adds up to 58 kDa. Like the solely monomeric ␣Syn observed without cross-linking, the mid-M r ␣Syn species detected after in vivo cross-linking are recovered almost exclusively in the high speed cytosol during subcellular fractionation. The nearly equal isoelectric points of these DSG-cross-linked assemblies with that of the DSG-modified monomers in two-dimensional gels suggest that all are composed of only ␣Syn. Furthermore, the addition of pure monomeric ␣Syn to cytosols before in vitro cross-linking selectively augmented (i.e. contributed to) the endogenous 80-kDa ␣Syn species. These data suggest that the mid-M r species of ␣Syn we detect in vivo and in vitro represent homo-oligomers rather than monomers bound to one or more distinct heterologous proteins. Indeed, we provide multiple criteria that rule strongly against the ␣S-60, -80, or -100 species being hetero-oligomers of ␣Syn monomers bound to a known ␣Syn-associated protein, to wit: exclusion of candidates as per supplemental Table 1; detection of oligomerization intermediates (dimers, trimers) under certain conditions and occurrence of molecular mass shifts by adding epitope tags to ␣Syn; co-IP of differentially tagged ␣Syn molecules in a 60-kDa oligomer; detection of the oligomeric species when ␣Syn is expressed in bacteria). On the basis of all cur-rently available data, we hypothesize that the ϳ60, 80, and 100 species observed in vivo are either composed of increasing numbers of monomers (4/5/6 or 4/6/8), or the ␣S-80 and 100 species represent conformers of the major 60-kDa cellular form. Parsimony suggests that the latter is likely to be a tetramer when our new data are integrated with the multiple lines of independent evidence (e.g. sedimentation equilibrium analytical ultracentrifugation; scanning transmission electron microscopy analysis; circular dichroism; NMR structure) that ␣Syn can exist in part as a helically folded tetramer (1, 6 -8).
We incorporated several key controls to support the physiological relevance of our cross-linking approach. First, in vivo cross-linking of endogenous ␣Syn by DSG (and DSS) occurred under the same conditions that allowed the observation of well established endogenous oligomeric proteins such as DJ-1, voltage-dependent ion channel, or Drp-1 in the same cells, whereas monomeric proteins such as Ran or Parkin were only detected at monomeric positions. Second, the ␣S-60, -80, and -100 species were all reduced quantitatively to monomers when the cleavable cross-linker DSP was applied to intact cells and then cleaved. Third, trapping of these species could not be achieved with certain other cross-linkers, e.g. the short spacer length DFDNB, suggesting that specific atomic spacing is required for successful stabilization of the species and that our observations are not just caused by non-specific cross-linking. Fourth, by using in vivo (intact cell) cross-linking as our standard method, by performing ultracentrifugation at extremely high speeds, by using various different lysis methods, and by demonstrating the absence of the synaptic vesicle protein synaptobrevin-2 in our neuronal cytosols, we could essentially rule out the possibilities that either endogenous membrane vesicles or membrane changes induced by cell lysis were responsible for the ␣Syn species we observed. Fifth, the pattern of oligomers obtained by non-specifically cross-linking pure, unfolded ␣Syn monomers in vitro is entirely distinct from the endogenous cellular oligomers we describe. Sixth, cellular co-expression of differentially tagged ␣Syn molecules and in vivo cross-linking leads to the co-IP of the two different monomers in the tetramers. Seventh, the confirmation of our basic findings by studying ␤-synuclein or else ␣Syn lacking an intact NAC domain (⌬71-82) has special interest, as it argues strongly against pathogenic (Parkinson disease-related) aggregation events being responsible for the detection of the mid-M r ␣Syn after cross-linking, thus underlining the physiological relevance of our findings.
The central purpose of the current study, as was the case for our earlier report (Bartels et al. (1), was to attempt to detect and characterize the physiological state of endogenous ␣Syn in cells, especially neurons. Such basic understanding is required to interpret the widely discussed concepts of "misfolding" and "pathological aggregation" of ␣Syn in human brain diseases. Although three recent papers strongly suggested the existence of oligomeric, folded ␣Syn in nature (1,6,7), the development of a facile and efficient in vivo cross-linking protocol we describe and validate here should now allow many laboratories to readily detect the oligomeric state of both endogenous and overexpressed ␣Syn in cellular models. We expect the method to be a useful tool for many further studies analyzing the effects of various cell stressors and of Parkinson disease-causing ␣Syn mutations, allowing dynamic studies of their effects on the native cellular state of ␣Syn. However, quantitative conclusions about the absolute abundance of the various mid-M r species versus monomeric ␣Syn are limited by the efficiency of in vivo cross-linking. The protocol we describe allows the detection of apparent endogenous oligomers and residual free monomers in living cells while largely avoiding ␣Syn-immunoreactive high M r smears, which we consider to arise from nonspecific crosslinking occurring only at high DSG concentrations.
Although the current study purposely focused on the development and exploitation of in vivo and in vitro cross-linking methods, the results can and should be interpreted in the context of our prior use of numerous other analytical methods (␣Syn purification from cells and concentration in the purified state, sedimentation equilibrium analytical ultracentrifugation, scanning transmission electron microscopy, CD (1)) that collectively support the hypothesis that the principal endogenous form of ␣Syn in cells is a folded tetramer of ϳ58 kDa. This model, which now requires more widespread confirmation, is also consistent with emerging data from other laboratories (6 -8, 16).
Very importantly, the striking sensitivity of the major ϳ60-kDa putative tetramer to cell lysis provides a strong caveat for the widespread study of cellular ␣Syn by conventional biochemical methods. This behavior of endogenous ␣Syn was distinct from that of endogenous DJ-1, voltage-dependent ion channel, GAPDH, DRP-1, and HSP-70, all of which showed their expected oligomeric species in the very same cells regardless of whether crosslinking was done in vivo or in vitro (see Fig. 5A). This unusual lability of cellular ␣Syn oligomers once cells are broken open is in agreement with an earlier study that detected intracellular ␣Syn oligomers (size not assessable) by FRET/fluorescence lifetime imaging microscopy measurements in intact cells but not by other methods after cell lysis (16). The lack of in vitro stability is unfortunate for the ␣Syn research community but must be taken into account in designing future studies of this key protein in Parkinson and Alzheimer diseases and other human synucleinopathies. Methods performed in intact cells (including new ways of performing FRET or cross-linking) and careful attention to the assembly state and folding of ␣Syn after its purification from brain tissue will now be necessary. We speculate that an unknown bioorganic molecule (perhaps a small lipid) normally stabilizes the endogenous tetramer (and perhaps higher oligomers) in intact cells, but this moiety is diluted and lost upon cell lysis except when very high macromolecular conditions are preserved during lysis (see Fig. 5D) or else concentration of the purified cellular protein is undertaken (1). The identification of such a factor(s) represents in our view an important new experimental goal, as the factor could help regulate ␣Syn assembly and stability in both health and disease.