Megadalton-sized Dityrosine Aggregates of α-Synuclein Retain High Degrees of Structural Disorder and Internal Dynamics

Graphical abstract


Introduction
Lewy body (LB) aggregates of the pre-synaptic protein aSyn are pathological hallmarks of Parkinson's disease 1,2 and related synucleinopathies. 3 Whereas ageing and cellular oxidative stress are common denominators in synucle-inopathies, LB morphologies and aSyn aggregate/fibril structures differ in a source-of-origin and disease-specific manner, [4][5][6][7] suggesting that aggregation pathways, fibril strains and aggregate compositions are likely influenced by contextdependent factors that are altogether poorly understood. 8 Similarly, the toxicity and spreading of aSyn oligomers and fibrils varies between different brain regions and cell types, 4 which underscores the confounding compositional, functional and structural heterogeneity of aSyn aggregates in these disorders. 5 Mitochondria emerged as key organelles in mediating possible aggregation scenarios of aSyn, 9 especially in relation to reactive oxygen species (ROS) production, oxidative stress sensing, and as executioners of programmed cell death, i.e. apoptosis. 10 Recent findings substantiate this notion by revealing direct interactions between aSyn and mitochondrial membranes and lipids, 11,12 aSyn-induced organelle toxicity, 4,13,14 stressrelated re-localization of aSyn to mitochondria 15,16 and pathological co-localization with mitochondria (and other non-proteinaceous structures) in brainderived LB inclusions 17 and cellular models of LB formation and maturation. 4 In addition, the functions of many other Parkinson's disease proteins converge on mitochondria and are directly linked to mitochondrial biology. 18 Cytochrome c (cyt c) resides in the inner mitochondrial membrane (IMM) where it is bound to cardiolipin, an organelle-specific, negativelycharged phospholipid. 19 Under physiological cell conditions, cyt c transfers electrons between complex III and IV of the electron transport chain and oxidizes superoxide anions to molecular oxygen (O 2 ) as a ROS scavenger. Upon cumulating oxidative stress, cardiolipin oxidation frees cyt c from the IMM and the protein is actively released into the cytoplasm where it acts as an apoptotic messenger. 20 Cytoplasmic cyt c ultimately induces caspase activation and cell death, although mitochondrial outer membrane permeation (MOMP) and cyt c release do not strictly trigger apoptosis and several non-deleterious functions have been discovered recently. 20 Amongst them, cyt c acts as a peroxidase and oxidizes proteins and lipids in the presence of ROS. 21 In substrates such as aSyn, cyt c-mediated oxidation primarily targets exposed tyrosine residues and leads to the formation of dityrosine crosslinks that locally concatenate individual molecules into covalent, high molecular weight aggregates. 22,23 aSyn is a 140-residue protein that contains four tyrosines. One in its N-terminal part (Y39) and three at its C-terminus (Y125, Y133 and Y136). aSyn dityrosine adducts are found in Lewy body deposits of Parkinson's disease (PD) patients and in neuronal inclusions of PD mouse models 23,24 that also stain positive for cyt c. 25,26 Cyt cmediated dityrosine aggregation of aSyn has been reconstituted with isolated components in vitro, [26][27][28][29][30][31][32] however, no attempts have been made to study the structures of the resulting species at the atomic-resolution level. Here, we delineate the structural and dynamic properties of aSyn dityrosine aggregates by solution-state nuclear mag-netic resonance (NMR) spectroscopy and complementary methods including dynamic light scattering (DLS), circular dichroism (CD), atomic force (AFM) and negative-stain transmission electron microscopy (TEM), and mass spectrometry (MS). We show that despite their exceedingly large size, they retain individually crosslinked aSyn molecules in highly dynamic disordered conformations that resemble the biophysical properties of the monomeric protein. Importantly, we establish that residues of the amyloidogenic NAC region of aSyn are minimally perturbed in these assemblies and display no signs of b-aggregation, regardless of their close proximity and high local concentrations. We further demonstrate that cyt c-mediated aSyn aggregates form in macromolecularly crowded, cellular environments, where they exhibit structural and dynamic characteristics that are indistinguishable from aggregates reconstituted with purified components in vitro.

Results
aSyn dityrosine crosslinking produces high molecular weight aggregates To obtain structural insights into aSyn dityrosine adducts, we reconstituted peroxidase reactions with equal amounts of recombinant, N-terminally acetylated aSyn (50 lM) and purified cyt c to which we added 50, 100, or 500 lM of hydrogen peroxide (H 2 O 2 ). We resolved the resulting mixtures by denaturing sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and visualized proteins by Coomassie staining (Figure 1(a)-(c)). In line with previous reports, 22,26,27 we detected multiple low molecular weight aggregates (LMWAs) at limiting H 2 O 2 concentrations, whereas high molecular weight aggregates (HMWAs) formed in the presence of excess peroxide (Figure 1(a)). Although we did not observe protein precipitation in any of these reaction mixtures, soluble HMWAs were retained in the loading slots and stack portions of SDS gels. At high H 2 O 2 concentrations, aSyn was quantitatively incorporated into HMWAs, whereas input levels of monomeric cyt c were largely preserved, as expected for an enzymatic reaction and in agreement with published data. 27 To evaluate the minimal stoichiometric requirements for cyt c in these reactions, we reconstituted oxidative aSyn aggregation with reduced amounts of peroxidase (Figure 1(b)). At aSyn (50 lM) to cyt c molar ratios of 100:1 and 50:1, LMWAs formed readily. At a 10:1 ratio of aSyn to cyt c, most of aSyn was incorporated into HMWAs. These results confirmed that aggregation of aSyn into covalent adducts in the presence of H 2 O 2 only required catalytic amounts cyt c. Next, we asked whether LMWAs constituted reaction intermediates en route to HMWA formation. We added increasing amounts of cyt c to preformed (c) Preformed low molecular weight aggregates (50 lM aSyn, 1 lM cyt c, 10 mM H 2 O 2 , marked with an asterisk) reacted with 2, 10 or 50 lM cyt c. All reactions were allowed to proceed for 30 min at 25°C. Loading slots and gel stacks are indicated with arrowheads. (d) Atomic force microscopy (AFM, top) and negative stain transmission electron microscopy (TEM, bottom) of purified high molecular weight aggregates (HMWAs) of aSyn. Control amyloid fibrils are shown on the right. (e) Biophysical characterization of monomeric aSyn (50 lM, black) and purified HMWAs (50 lM, red) by CD spectroscopy, DLS, SEC and fluorescence emission spectroscopy. (f) Coomassie-stained SDS-PAGE of aSyn low and high molecular weight aggregates (LMWAs and HMWAs, left) and corresponding Western blot (WB) analyses, with antibodies against Nand C-terminal aSyn epitopes, cyt c and dityrosine adducts. Dotted horizontal lines connect WB signals to corresponding Coomassie protein bands. Loading slots and gel stacks are indicated with arrowheads.
LMWAs and allowed reactions to pursue for 30 min (Figure 1(c)). Having resolved the respective mixtures by SDS-PAGE, we found that LMWAs efficiently converted into HMWAs and that final HMWA abundance correlated with cyt c input concentrations. Aiming to further characterize the structural details of cyt c/H 2 O 2 -mediated aSyn aggregates, we purified HMWAs and analyzed them by AFM and negative-stain TEM (Figure 1(d)). Both methods revealed amorphous globular structures of 30-50 nm diameters (10-15 nm heights), with different macroscopic features compared to canonical amyloid fibrils. Solution characterization by CD spectroscopy, DLS, sizeexclusion chromatography (SEC) and ultraviolet (UV) absorption measurements showed that HMWAs retained high degrees of structural disorder and uniform DLS distributions centered at 30 nm, roughly the diameter of assembled eukaryotic ribosomes, 33 compared to~6 nm for monomeric aSyn (Figure 1(e)). HMWAs eluted as single peaks in SEC void volumes, separated from non-incorporated cyt c and well set off from monomeric aSyn, and displayed fluorescence emission bands at 404 nm, which are characteristic for dityrosine adducts. 34 To determine the molecular compositions of aSyn aggregates, we probed LMWAs and HMWAs with different aSyn, cyt c and dityrosine-specific antibodies by Western blotting (Figure 1(f)). Antibodies against N-terminal (aa6-23), or C-terminal (aa120-125) aSyn epitopes recognized dimer-, trimer-and higher oligomericspecies in LMWAs that corresponded in size to Coomassie-stained protein bands, confirming that these adducts contained aSyn. The dityrosine antibody reacted with the same set of bands, validating the presence of intermolecular crosslinks. It showed no signal for monomeric aSyn, which suggested that intramolecular dityrosines did not form to a significant extent. By contrast, the cyt c antibody recognized different sets of low molecular weight bands that corresponded to monomeric cyt c and to cyt c oligomers with no corresponding Coomassie bands. These adducts contained no aSyn based on Western blotting results, which established that aSyn-aSyn dityrosine adducts constituted the primary species in LMWAs. Antibodies against C-terminal portions of aSyn, where Y125, Y133 and Y136 are located, and against dityrosines failed to recognize their epitopes in HMWAs, which suggested that crosslinking rendered these sites inaccessible for antibody binding. Antibodies against the N-terminus of aSyn and against cyt c produced HMWA signals, indicating that Nterminal aSyn residues remained accessible and that some co-aggregated cyt c was present. These results confirmed that intermolecular dityrosine crosslinks between individual aSyn molecules formed the basis of low and high molecular weight aggregates.

Dityrosine aggregates are disordered and dynamic
To derive high-resolution insights into the structural and dynamic properties of aSyn in HMWAs, we reconstituted aggregates with 15 N isotope-labeled, N-terminally acetylated aSyn and unlabeled cyt c in the presence of peroxide. We removed H 2 O 2 and free cyt c by SEC and pooled HMWA fractions for solution-state NMR experiments. Despite the exceedingly large size of these assemblies, usually beyond the scope of solution NMR measurements, 2D 1 H-15 N heteronuclear correlation spectra of HMWAs were of excellent quality and revealed the majority of aSyn signals at resonance positions corresponding to those of the monomeric protein ( Figure 2(a)). These NMR characteristics substantiated CD results about the disordered nature of aSyn in these aggregates (Figure 1(e)) and established that crosslinked aSyn molecules retained high degrees of internal dynamics. Upon closer inspection of NMR spectra, we noted pronounced chemical shift changes for N-terminal aSyn residues 1-10, including M1 and M5, that we and others had previously identified as characteristic for oxidized methionines (i.e. methionine sulfoxides), 35,36 confirming that NMR samples had indeed formed under oxidative conditions. Importantly, we also noted continuous stretches of severely line-broadened NMR signals for residues around Y39, Y125, Y133 and Y136, the four tyrosines of aSyn and possible sites of dityrosine crosslinks. To verify that detected NMR signals indeed originated from aSyn aggregates, we performed diffusion ordered spectroscopy (DOSY) measurements on HMWAs (Figure 2(b)). DOSY results indicated that HMWA diffusion was greatly reduced compared to monomeric aSyn and corresponded to a megadalton-sized~35 nm diameter particle, in good agreement with TEM and DLS results (Figure 1(d) and (e)).
To further explore aSyn dynamics in HMWAs, we analyzed signal intensity attenuations on a residueresolved basis and performed longitudinal (R 1 ) and transverse (R 2 ) relaxation experiments as well as heteronuclear Overhauser effect (hetNOE) measurements (Figure 2(c)). We determined intensity ratios (I/I 0 ) for well-resolved HMWA signals (I) in relation to monomeric aSyn (I 0 ) and found that site-selective line broadening centered at Y39 and C-terminal Y125, Y133 and Y136. In relation to these sites, attenuations weakened in a distance-dependent manner, with more expansive effects around Y39 compared to Y125, Y133 and Y136 (Figure 2(c), top). aSyn residues comprising the aggregation-prone non-amyloid component (NAC) region (aa60-95) displayed minimal line-broadening, which suggested that their overall structural and dynamic properties were retained in HMWAs. N-terminal aSyn residues 1-10 exhibited similarly small linebroadening effects, indicating equally preserved dynamics in HMWAs and probably explaining the observed antibody accessibility (Figure 1(f)). Quantitative R 1 and R 2 results supported these conclusions by establishing that fast picosecond to nanosecond amide backbone motions (R 1 ) were marginally affected, whereas R 2 profiles showed position-dependent attenuations consistent with restricted motions and/or conformational exchange on the microsecond to millisecond timescale (Figure 2(c), middle). hetNOE data confirmed that HMWAs displayed independent segmental motions characteristic for a disordered protein (Figure 2(c), bottom). The qualitative picture that emerged from these experiments stipulated that dityrosine crosslinks restricted backbone motions of concatenated aSyn molecules only in the vicinity of crosslinked residues, whereas remote sites retained high degrees of flexibility and conformational dynamics. On the macroscopic level, these results supported a model of loosely packed, macromolecular assemblies of disordered molecules held together by repeating, intermolecular side-chain crosslinks.
To independently verify the general nature of these dityrosine aggregate characteristics, we resorted to photoinduced crosslinking of unmodified proteins (PICUP) and reacted N-terminally acetylated aSyn with ammoniumpersulfate (APS) and Ruthenium (Ru 3+ ) under light as reported previously. 28 PICUP HMWAs displayed similar SDS-PAGE migration properties as cyt c/H 2 O 2 -mediated aggregates in that aSyn was quantitatively converted into high molecular weight species retained in loading slots and stacking gels (Figure 2(d)). Exploiting the homogenous nature of PICUP HMWAs, we analyzed aggregates by MS to delineate dityrosine positions and relative distributions ( Figure S1(a)). We identified crosslinks of Y39 peptides to aSyn fragments containing Y39, Y125, Y133 and Y136. Similarly, we detected intermolecular Y133-Y133 and Y136-136 crosslinks, but no intra-or inter-molecular adducts between other C-terminal tyrosines (Figures 2(d) and S1(a)). These results supported the central role of Y39 as a connectivity hub in aSyn PICUP aggregates, in line with previous findings. 28 To compare the structural features of PICUP HMWAs and cyt c/H 2 O 2 -mediated assemblies, we recorded NMR spectra on photo-induced aggregates of 15 N isotope-labeled aSyn (Figures 2(e) and S1(b)). Surprisingly, both types of HMWAs displayed virtually indistinguishable spectral features, especially in terms of site-selective line broadening. Different to cyt c/H 2 O 2 HMWAs, however, we did not observe N-terminal aSyn residues 1-10 in NMR spectra of PICUP aggregates. Our combined NMR data established that dityrosine HMWAs of aSyn exhibited similar structural and dynamic properties irrespective of whether they were generated via a protein-catalyzed peroxidase reaction or directly by photo-induced crosslinking.

Dityrosine aggregates form in complex environments
Next, we asked whether cyt c/H 2 O 2 HMWAs of aSyn formed in the presence of competing interactions with other macromolecules in crowded environments. We initially employed bovine serum albumin (BSA), which has been shown to transiently interact with N-terminal aSyn residues and to bind to Y39 via weak hydrophobic contacts, 16,37 and reconstituted cyt c/H 2 O 2 aggregation reactions with 15 N isotope-labeled aSyn and increasing amounts of unlabeled BSA (Figure 3 (a)). SDS-PAGE analysis revealed that stacking gel-retained HMWAs specifically formed in solutions of up to 10 mg/mL of BSA, which was the highest concentration we used to avoid overloading the gel. We reacted 15 N isotope-labeled aSyn with unlabeled cyt c and H 2 O 2 in the presence of a 20fold higher amount of BSA (200 mg/mL) and recorded in situ NMR experiments on the resulting mixture (Figures 3(b) and S2(a)). Similar to previous results, we obtained high quality NMR spectra that revealed the characteristic line-broadening profile of dityrosine-crosslinked aSyn, which confirmed that cyt c/H 2 O 2 aggregation of aSyn occurred under highly crowded in vitro conditions and regardless of protein-protein interactions that involved Y39. Following these observations, we extended our analysis to neuronal RCSN3 cell lysates. 38 We adjusted total lysate protein concentrations to 3 mg/mL and added equimolar amounts of aSyn and cyt c (50 lM) and 0.5-500 mM of H 2 O 2 . Western blotting revealed that aggregate formation only occurred at the highest peroxide concentration ( Figure 3(c), top), in contrast to previous experiments with isolated components ( Figure 1) and in BSA-crowded solutions (Figure 3 (a)). We hypothesized that peroxide-metabolizing enzymes may have cleared exogenously added H 2 O 2 . Indeed, when we measured remaining H 2 O 2 levels using a colorimetric assay, we found that only 1.8 ± 0.8% of input peroxide (10 mM) was present in these lysates (Figure 3(c)). To weaken endogenous peroxidase activities, we treated RCSN3 lysates with diethyl pyrocarbonate (DEPC), a chemical that inactivates enzymes by irreversible carbethoxylation of active site residues. 39 Because DEPC has a limited lifetime in aqueous solutions and decomposes readily, one of its advantages over enzyme inactivation by acid, SDS, or high temperature denaturation is the preservation of otherwise native lysate conditions. Accordingly, we adjusted the timing of lysate inactivation by DEPC ( Figure S2(b)) and added   aSyn, cyt c and H 2 O 2 as outlined before (Figure 3 (c), bottom). Under these conditions, 40 ± 0.5% of exogenously added peroxide was preserved and we detected aSyn aggregation at correspondingly lower H 2 O 2 concentrations. We repeated these reactions with 15 N isotope-labeled aSyn and recorded in situ NMR experiments in DEPCtreated RCSN3 lysates (Figures 3(d) and S2(c)). 2D NMR spectra exhibited compelling degrees of similarity with other HMWA samples. We clearly observed the spectral fingerprints of dityrosine crosslinks manifested by the conserved linebroadening of residues neighboring Y39, Y125, Y133 and Y136, and the characteristic chemical shift changes of N-terminal aSyn residues due to methionine oxidation (see Figure 2(a) for comparison). Together with results in BSA-crowded solutions, these findings demonstrated that cyt c/ H 2 O 2 -mediated aggregates of aSyn formed efficiently in complex physiological environments, especially under conditions of sustained ROS availability required for dityrosine crosslinking.

Dityrosine aggregates bind membranes, inhibit amyloid formation and are non-cytotoxic
Because N-terminal aSyn residues that mediate lipid-anchoring contacts 40 remained accessible in HMWAs, we asked whether aggregates retained the ability to interact with membranes. We added 15 N isotope-labeled monomeric and HMWA aSyn to small unilamellar vesicles (SUVs) that we reconstituted from pig-brain polar lipid extract. In line with previous results on such SUVs, 37 monomeric aSyn displayed continuous line-broadening of residues 1-100, which serves as a widely used indicator for aSyn membrane binding 41,42 (Figures 4(a) and S3(a)). Surprisingly, 2D NMR spectra of 15 N isotope-labeled HMWAs exhibited equally pronounced line-broadening in the presence of SUVs (Figure 4(b) and S3(b)). By contrast, however, remaining HMWA resonances included stretches of residues within the first 100 amino acids of aSyn, spanning V70 to Q79 for example, which collectively suggested that HMWA-SUV contacts were different than those of the monomeric protein.
Given that aSyn interacts with membranes in helical conformations, 43 we asked whether HMWAs adopted similar helical structures and recorded CD spectra of monomeric and HMWA aSyn in the presence of SDS micelles, a common surrogate for studying Syn's a-helical conversion upon membrane binding 44 (Figure 4(c)). We found that monomeric and HMWA aSyn displayed helical signatures in corresponding CD spectra, which suggested that HMWA-membrane interactions triggered some degrees of a-helical rearrangements. Finally, we sought to revisit HMWA effects on the aggregation behavior(s) of monomeric aSyn. Previous studies suggested that dityrosine adducts of aSyn impaired the formation of cross-b amyloid structures in a concentration-dependent manner. 31 In agreement with these findings, we observed the quantitative inhibition of thioflavin T (ThT)-positive aggregate formation in shaking assays when we added 10% HMWAs to monomeric aSyn (Figure 4  (d)). To investigate the mode of inhibition, we repeated these assays with 15 N isotope-labeled, monomeric aSyn and unlabeled HMWAs, and measured 1D 15 N-edited NMR spectra at different time-points of the reaction. In the absence of HMWAs, we observed the progressive attenuation of NMR signals as expected for the incorporation of monomeric aSyn into growing amyloid fibrils (Figure 4(e)). In the presence of HMWAs, NMR signal intensities of aSyn did not diminish substantially, suggesting that inhibition was mediated by the preservation of the monomeric protein state and not by the generation of ThT-negative, offpathway aggregates as has been observed for many other inhibitors. 45 To assess whether HMWAs were cytotoxic, we measured the metabolic activity of PC12 cells to which we added protofibrils generated from monomeric aSyn, purified HMWAs and aliquots of inhibited aggregation mixtures containing aSyn and HMWAs that we collected after 22 and 48 hours (Figure 4(f)). Addition of protofibrils decreased the conversion of 3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbro mide (MTT) into formazan in a concentrationdependent manner, in line with previous reports of cell metabolic activity 46 and characteristic of cytotoxic aggregates. 47 By contrast, neither purified HMWAs nor aSyn-HMWAs mixtures impaired cell viability and specimens displayed no signs of cellular stress (Figure 4(f)). These results established that exogenously added dityrosine aggregates of aSyn inflicted no adverse effects on the metabolic activity and viability of PC12 cells.

Discussion
Recent years saw a wealth of high-resolution structural information about aSyn oligomers 48 and amyloid fibrils, 49 ranging from in vitro reconstituted assemblies to specimens directly derived from post-mortem brains. 50 Correlative imaging studies delivered complementary insights into the compositions and morphologies of Lewy bodies and their complex nature in terms of membrane and organelle inclusions, and the abundance of heterogeneous aggregate structures. 4,5,17 Parallel to these advancements, functional links between ageing, cellular oxidative stress, neurodegeneration and oxidative aSyn modifications have been explored for some time. 51 In this context, our results add primary insights into the structure and dynamics of non-canonical aSyn aggregates characterized by covalent dityrosine crosslinks between N-(Y39) and C-terminal (Y125, Y133, Y136) residues. We demonstrate that these aggregates efficiently form under oxidative conditions in the presence of catalytic amounts of a peroxidase enzyme (cytochrome c) (Figure 1) and that they constitute amorphous, locally concatenated assemblies that retain high degrees of structural disorder and internal dynamics (Figure 2(a)-(c)). Although we intentionally chose 'exhaustive' reaction conditions to produce homogenous endpoint products (i.e. HMWAs), we reason that the structural and dynamic properties that we elucidated are similarly displayed in mixtures of lower molecular-weight aggregates (i.e. LMWAs), which constitute reaction intermediates (Figure 1(c)). Having established that aSyn dityrosine aggregates also form in artificially crowded in vitro solutions and cell lysates (Figure 3) further strengthens the notion that crosslinking avidity and specificity are well preserved in complex physiological environments and that these aggregates may co-exist with monomeric, oligomeric or fibrillar forms of aSyn in cells. Furthermore, it seems plausible that mixed dityrosine adducts between different aSyn species can form, especially under conditions of sustained oxidative stress and considering the persistent reactivity of unmodified tyrosine residues. Such mixed crosslinks may modify non-covalent aSyn aggregates or stabilize existing aSyn fibrils. 22,52 In this regard, it is interesting to note that Y39, Y125, Y133 and Y136 in structures of reconstituted aSyn fibrils are not part of the ordered amyloid cores and do not contribute to protofilament packing, 49 which suggests that they may remain accessible for oxidative crosslinking. Indeed, results from dityrosine-aSyn EM immunogold co-labeling studies on PD patient-derived Lewy bodies confirm the presence of such species in physiological samples. 24 Co-aggregation with non-covalent aSyn fibrils may further explain certain aspects of morphological changes during Lewy body maturation, especially at later stages when fibrillar structures convert into extended spherical inclusions. 4 Intriguingly, this transition coincides with the recruitment of mitochondria into Lewy bodies, 4 which are abundantly found in mature deposits. 17 This puts aSyn, mitochondrial cytochrome c and the sites of ROS production in close proximity and it is straightforward to imagine how these encounters may drive or perpetuate the formation of dityrosine aggregates, as well as impede cellular functions such as proteasomal degradation, chaperone-mediated autophagy, 53 or illicit inflammatory activation of neighboring cells upon excretion. 54 In this regard, results pertaining to the non-cytoxicity of HMWAs when externally applied to PC12 cells ought to be considered with caution (Figure 4(f)). Because standard MTT assays do not provide information about aggregate internalization, we are unable to assess whether they cross the plasma membrane and reach the cytoplasm to cause intracellular effects. Alternatively, monomeric or oligomeric aSyn may directly engage with mitochondria and actively contribute to ROS production, cytochrome c release and, in turn, dityrosine aggregate formation. 55 It is important to stress, however, that the generation of aSyn dityrosine adducts does not strictly require a peroxidase enzyme (i.e. cytochrome c) and that they can form as reaction side-products of tyrosine nitration, by dopamine-mediated autoxidation, or in the presence of oxidized lipids or divalent metals. 51 Thus, cellular scenarios in which such aggregates are generated with or without the involvement of mitochondria are manifold.
Dityrosine crosslinks may further render involved residues inert to other post-translational modifications (PTMs) such as nitration 56 and phosphorylation. 57 All aSyn tyrosines undergo reversible phosphorylation 58 and PTM cross-talk between Y125 and S129 phosphorylation, as well as M127 oxidation is well established. 36,59 Crosslinks are likely to impair the accessibility of these sites and, thereby, modulate the enzymatic establishment or removal of individual PTMs. With regard to aSyn S129 phosphorylation, the pathological hallmark of Lewy body inclusions, 60 they may help to preserve this modification by restricting access of bulky phosphatase holoenzymes. 61 Intriguingly, mitochondrial recruitment and phospho-S129 accumulation coincide with Lewy body maturation. 4 Similarly, dityrosine adducts and aberrant PTM states may diminish aSyn interactions with other proteins, including the endocytosis GTPase Rab8 62 and the synaptic vesicle v-SNARE component synaptobrevin-2/VAMP2, 63 which may exert debilitating effects on the physiological functions of aSyn, probably exacerbated by the residual membrane-binding activity of these adducts (Figure 4(b)). They may further affect Cterminal proteolytic processing events 64 and the cellular clearance of aSyn. 51 Given that covalent dityrosine crosslinks represent irreversible protein modifications that cannot be undone by cellular enzymes, their impact on aSyn's biology is persistent.
In summary, our results established that oxidative aggregation of aSyn via dityrosine crosslinking imposes minor effects on N-terminal and central portions of the protein, with residues involved in membrane anchoring and of the amyloidogenic NAC region being largely unperturbed (Figure 2). The resulting degrees of conformational freedom appear to allow for structural rearrangements required for helical membrane-binding (Figure 4(b) and (c)) and to mediate the inhibition of NAC-dependent cross-b amyloid formation (Figure 4(d) and (e)), probably via capping of growing fibril ends. Because three of the four tyrosines of aSyn are located in the Cterminus of the protein, functional consequences are expected to be more drastic for activities mapped to this region, including the establishment and removal of PTMs and various protein-protein interactions. With regard to the involvement of cytochrome c, oxidative dityrosine crosslinking may scavenge the activity of this proapoptotic messenger and, in turn, delay the onset of programmed cell death, as has been suggested earlier. 26 By doing so, the outlined reaction scenarios may perpetuate detrimental conditions of oxidative stress and exacerbate cellular insults that contribute to PD pathology.

Materials and Methods
Proteins Unlabeled or uniform 15 N or 15 N/ 13 C isotope-labeled, Nterminally acetylated, human alpha-synuclein (aSyn) was expressed and purified under native conditions as previously described. 37 Recombinant proteins were resuspended in 20 mM phosphate buffer, 150 mM NaCl at pH 6.4 (NMR buffer). Protein concentrations were determined by UV/VIS absorption measurements at 280 nm (k) with an extinction coefficient (e) of 5600 M À1 cm À1 . Purified horse heart cyt c was obtained from Sigma, BSA was purchased from Roth. Stock solutions were prepared in NMR buffer by weighing in the respective amounts of protein powder and confirming absorbance at 410 nm for cyt c (e 106.1 mM À1 cm À1 ) and at 280 nm for BSA (e 43824 M À1 cm À1 ).  (Figure 1(d) and (e)), NMR measurements (Figure 2(a) and (e)) and testing of membrane binding and aggregation interference (Figure 4), HMWAs were purified by size exclusion chromatography on a Superdex-75 column to remove non-incorporated cyt c and excess hydrogen peroxide (H 2 O 2 ). Pooled HMWA fractions were either used immediately or flash frozen in liquid nitrogen and stored at À80°C. For NMR experiments, estimates of purified 15 N isotope-labeled HMWA concentrations were derived based on 1D 15 N-edited proton NMR spectra in comparison to reference samples of monomeric aSyn at known concentrations. Accordingly, samples were adjusted to 50 lM.

Polyacrylamide gel electrophoresis (PAGE)
Denaturing sodium dodecylsulfate (SDS) PAGE gels used in this study were prepared on Miniprotean III systems (BioRad) with a constant separating-gel pore size of 18% acrylamide (w/v). Proteins were visualized by gel staining with Coomassie Brilliant Blue (Sigma Aldrich).

Atomic force microscopy (AFM)
Sheet mica (Nanoworld) was glued to microscope glass slides and 20 lL of 25 lM HMWAs or aSyn amyloid fibrils were adsorbed for 10 min onto freshly cleaved mica (1 Â 1 cm), washed with filtered, deionized water (3 Â 30 lL) and dried overnight. Reference amyloid fibrils were obtained by shaking 100 lL of 50 lM aSyn in NMR buffer at 200 rpm, 37°C for 7 days. Fibrils were stored at 4°C and shortly vortexed before experiments. Amyloid fibril concentrations were considered equal to input amounts of aSyn. Dry AFM images were recorded on a Nanowizard II/Zeiss Axiovert setup (JPK) using intermittent contact mode and FEBS cantilevers (Veeco).

Negative-stain transmission electron microscopy (TEM)
Twenty microliters of 25 lM HMWAs and aSyn amyloid fibrils were adsorbed onto glow-discharged carbon-coated copper grids for 1 min. Excess liquid was removed with filter paper and grids were washed twice with H 2 O before staining with 2% (w/v) uranyl acetate for 15 s. TEM images were acquired on a Philips CM100 microscope.

Circular dichroism (CD) spectroscopy
CD measurements were performed on a Jasco J-810 spectropolarimeter at 25°C with samples at 10 lM. Far-UV CD spectra were collected in NMR buffer using a 0.1 cm pathlength cuvette. Five scans were averaged and blank (buffer) scans were subtracted. CD experiments with SDS-micelle ( Figure 4(c)) were performed in the same manner. Micelles were prepared by dissolving 10 mM SDS in NMR buffer as reported in. 44 Dynamic light scattering (DLS) DLS measurements were acquired on a Zetasizer Nano ZS (Malvern Instruments) operating at a laser wavelength of 633 nm equipped with a Peltier temperature controller set to 25°C. Data were collected on 50 lM protein samples in NMR buffer in 3 Â 3 mm cuvettes. Using the Malvern DTS software, mean hydrodynamic diameters were calculated from three replicates in the volume-weighted mode.

Nuclear magnetic resonance (NMR) spectroscopy
NMR experiments were performed on Bruker Avance 600 or 750 MHz spectrometers, equipped with cryogenically-cooled triple resonance 1 H{ 13 C/ 15 N} TCI probes and z axis selfshielded gradient coils. Data acquisition, processing, and spectral analyses were carried out in TOPSPIN 2.1, iNMR 3.6.3 and CCPNmr 2.1.5. 2D 1 H-15 N correlation spectra of monomeric uniform 15 N nitrogen labeled aSyn and HMWAs were acquired using the SOFAST-HMQC pulse sequence on 50 lM samples dissolved in NMR buffer at 10°C, with 1 K and 256 complex points for sweep-widths (SW) of 16 ppm and 28 ppm in 1 H and 15 N dimensions, respectively. Spectra were recorded with 32 scans and interscan delays of 60 ms. Processing was done by zero-filling to 4 times the number of complex points, followed by a sine modulated window function multiplication, Fourier transformation (FT) and baseline correction in both dimensions. Diffusion ordered spectroscopy (DOSY) was performed on 50 lM monomeric aSyn and homogeneous HMWA samples in NMR buffer. Diffusion coefficients were obtained by fitting intensity curves from bipolar pulse-pair longitudinal-eddy-current delay (BPPLED) experiments with an array of linearly shifted gradient power. A series of 28 1D NMR spectra were collected as a function of gradient amplitude. Intensities were measured by integrating signals from 1D proton ( 1 H) NMR spectra between 0.75 ppm and 3.2 ppm. Monomeric aSyn and its hydrodynamic radius/diameter (Rh) obtained by DLS was used as the internal reference to determine Rh for HMWAs. 15 N amidebackbone relaxation data were acquired using standard pulse sequences provided in the Bruker Topspin library. Spectra used for longitudinal relaxation T 1 (1/R 1 ) analysis were collected using the following delay times (in ms) : 12, 52, 102, 152, 202, 302, 402, 602, 902, 2002, 5002. T 2 data (1/R 2 ) were measured using a pulse sequence employing a CPMG pulse train with the following delays (in ms): 6,10,18,26,34,42,82,122,162,202,242, 322 and 462. Duplicate spectra were collected for several time points to estimate uncertainty. To determine R 1 and R 2 relaxation rates resonance signal intensities were extracted and fit as a function of the relaxation delay time in CCPNmr. To obtain steady-state hetero-nuclear (het) NOE values, we calculated ratios of peak intensities in paired spectra collected with and without an initial 4 s period of proton saturation during the 5 s recycling delay. For relaxation experiments, the concentrations of aSyn and HMWA aggregates were 50 lM. Resolution settings and processing were the same as for 2D SOFAST-HMQC experiments. 8, 16 and 32 transients were used for T1, T2 and hetNOE experiments, respectively. To determine residue-resolved NMR signal intensity ratios, only cross-peaks of well-resolved aSyn residues were used. Absolute HMWA signal intensities (I) were divided by corresponding signals of monomeric aSyn (I 0 ) and I/I 0 ratios were plotted against the protein sequence.

Mass spectrometry (MS)
PICUP HMWAs (20 lg) were digested (protein concentration of 0.6 mg/mL in a 20 mM HEPES, 20 mM NaCl, 5 mM MgCl 2 buffer) using the endoprotease Glu-C. HMWAs were first reduced by adding 5 mM DTT, with incubation for 30 min at 60°C, followed by alkylation with 15 mM iodoacetamide, with incubation in the dark for 15 min at 20°C. Glu-C was then added at a protease-to-protein ratio of 1:13 (w/w) and digestion allowed to proceed for 18 h at 37°C. The resulting digest was acidified to pH 2 with trifluoroacetic acid and desalted using C18 StageTips before analysis on the mass spectrometer. Peptides were loaded (at a flow rate of 0.5 ll min À1 ) directly on a spray analytical column (75 lm inner diameter, 8 lm opening, 250 mm length; New Ojectives, Woburn, MA) packed with C18 material (ReproSil-Pur C18-AQ 3 lm) using an air pressure pump (Proxeon Biosystems). Two gradients (at a constant flow rate of 0.3 ll min À1 ) were used for chromatographic separation (A: 0.1% formic acid in water, B: 0. and analyzed using a "high/high" acquisition strategy, detecting peptides and at high resolution in the Orbitrap and analyzing fragmentation products also in the Orbitrap. Precursor scan (MS) spectra were acquired in the data-dependent mode, detecting in the Orbitrap at 100 000 resolution. The eight most intense triply charged or higher precursors for each acquisition cycle, were isolated with a 2 Th m/z window and fragmented in the ion trap with collision-induced dissociation (CID) at a normalized collision energy of 35. Subsequent product (MS2) fragmentation spectra were then recorded in the Orbitrap at 7500 resolution. A dynamic exclusion window (with single repeat count) of 60 s was applied and automatic gain control targets were set to 1x 10 6 (precursor scan) and 1 Â 10 5 (product scan). Raw files for crosslinking searches were processed using MaxQuant software (v. 1.2.2.5) using default parameters, except for the setting "Top MS/MS peaks per 100 Da", which was set to 100. Peak lists were searched against the sequence of aSyn using Xi software (v. 1.7.5.1). The following search parameters were applied: MS accuracy, 6 ppm; MS/MS accuracy, 20 ppm; missing mono-isotopic peaks, 2; variable modifications, carbamidomethylation (Cys) and oxidation (Met); enzyme, V8, with a maximum of four missed cleavages allowed; crosslinker modification mass (Tyr-Tyr), À2.015650 Da. Matched crosslinks were mapped to the primary protein structure using xiNET.

RCSN3 cell lysates, DEPC treatment and H 2 O 2 quantification
RCSN3 cells (rat cortical substantia nigra neurons) were grown in DMEM HAM F12 medium (PAA) supplemented with 12.5% fetal calf serum (FCS) and 1% penicillin/streptomycin to confluency in 175 cm 2 flasks, in a 5% CO 2 humidified environment at 37°C. Following medium removal and washing with PBS, 100 lL of chemical lysis buffer (50 mM Tris-HCl pH 7.4, 0.1% Triton X, 0.25% Na-deoxycholate, 150 mM NaCl, 1 mM NaF, no reducing agents) was added to each flask. Cells were detached with a plastic scraper and collected lysates were centrifuged at 16,000g for 30 min. Total protein concentrations of soluble fractions were determined with a Bradford assay (BioRad) and adjusted to 3 mg/mL with NMR buffer. For inactivation with DEPC (Sigma) concentrated stock solutions were prepared in 100% ethanol. DEPC was added to RCSN3 lysates to final concentrations of 50 mM and incubated for 22 h at 25°C. aSyn (50 lM) and cyt c (50 lM) were mixed with native and DEPC-inactivated RCSN3 lysate aliquots and 0.5, 10, 50 and 500 mM of H 2 O 2 (input concentrations). To determine amounts of remaining H 2 O 2 in native and DEPC-inactivated cell lysates, a colorimetric peroxide detection kit was employed (AssayDesigns, Stressgen).

Small unilamellar vesicles (SUVs)
SUVs were prepared from commercial pig-brain polar lipid extract (Avanti) as reported in Theillet et al. 37 The lipid powder was dissolved in NMR buffer at a concentration of 16 mg/mL (aprox. 20 mM assuming an average lipid mass of 800 Da) and vortexed at room temperature for 30 min. The solution was frozen and thawed 5 times on dry-ice and a 37°C water-bath, sonicated for 20 min at 4°C using 30% sonicator output power (Brandelin Sonoplus) and centrifuged at 16,800g for 10 min to remove remnant insoluble material. The resulting average vesicle size was 60 nm as determined by DLS. For the binding experiments we added 50 lM of 15 N-isotopically enriched aSyn or HMWA to SUV solutions (final concentration~16 mM after aSyn or HMWA addition). NMR spectra were recorded at 30°C.

Thioflavin T (ThT) aggregation assay
Kinetic aggregation assays were performed with 100 lL aliquots of 50 lM monomeric aSyn (in 100 mM Na-phosphate, 10 mM NaCl, pH 7.2), or monomeric aSyn in the presence of 5 lM HMWAs (10%), and 20 lM Thioflavin T (ThT) in low-binding 96-well plates (Corning), shaken at 200 rpm with one 2 mm glass bead per well, at 37°C. Samples were shaken continuously for 2 h before ThT fluorescence emission was recorded at 485 nm (excitation at 440 nm) on a plate reader (InfinitE M200). Samples were set up in triplicates. Data points in Figure 4(d) correspond to the mean value, error bars denote standard errors of the mean (SEM).

MTT metabolic assay
PC12 (rat pheochromocytoma, American Type Culture Collection) cells were cultured in DMEM medium (Gibco BRL) supplemented with 5% FBS, 10% horse serum and 3 mM glutamine. Cells were plated at a density of 10 3 cells per well in 96-well plates coated with polylysine in 90 lL of fresh medium in a 5% CO 2 humidified environment at 37°C. After 24 h, aliquots of aSyn protofibrils, pure HMWAs or aSyn-HMWA aggregation mixtures were added at concentrations of 0.03, 0.06, 0.13, 0.25, 0.5 and 1 lM, in three replicates. Cells were incubated for 3 days at 37°C. Cytotoxicity was measured using an MTT assay kit (Promega) by determining Formazan absorbance at 590 nm on a Tecan Safire automated plate reader. Absorbance values obtained from samples containing aggregated or fibrillar species and from untreated cells were averaged and used to calculate cell viability. Viability was expressed in percentage compared to untreated cells (100%). Error bars indicate SEM values.