Brazilin Removes Toxic alpha-Synuclein and Seeding Competent Assemblies from Parkinson Brain by Altering Conformational Equilibrium

Alpha-synuclein (α-syn) fibrils, a major constituent of the neurotoxic Lewy Bodies in Parkinson’s disease, form via nucleation dependent polymerization and can replicate by a seeding mechanism. Brazilin, a small molecule derived from red cedarwood trees in Brazil, has been shown to inhibit the fibrillogenesis of amyloid-beta (Aβ) and α-syn, prompting our inquiry in its mechanism of action. Here we test the effects of Brazilin on both seeded and unseeded α-syn fibril formation and show that the natural polyphenol inhibits fibrillogenesis of α-syn by a unique mechanism that is distinct from other polyphenols and is also distinct from its effect on Aβ. Brazilin preserves the natively unfolded state of α-syn by stabilizing the compact conformation of the α-syn monomer over the aggregation-competent extended conformation. Molecular docking of Brazilin shows the molecule to interact both with unfolded α-syn monomers and with the cross-β sheet structure of α-syn fibrils. Brazilin eliminates seeding competence of α-syn assemblies from Parkinson’s disease patient brain tissue, and treatment of pre-formed fibril assemblies with Brazilin significantly reduces their toxicity in primary neurons. Our findings suggest that Brazilin has substantial potential as a neuroprotective and therapeutic agent for Parkinson’s Disease. Highlights - The natural polyphenol Brazilin binds to monomeric, oligomeric and fibrillar α-syn - Brazilin shifts the equilibrium away from aggregation-competent monomer conformations - Brazilin inactivates seeding-competent α-syn isolated from Parkinson patients’ brains - Brazilin detoxifies α-syn aggregation intermediates and stabilizes mature amyloid fibrils Graphical Abstract


INTRODUCTION
Parkinson's Disease (PD), the second most common neurodegenerative disease following Alzheimer's Disease (AD), is characterized by the successive loss of dopaminergic neurons in the substantia nigra pars compacta along with pathological aggregation of the natively unfolded α-synuclein (α-syn) protein into Lewy Bodies [1][2][3] . In its native state, αsyn is a 140 residue intrinsically disordered protein that can adopt an α-helical structure when bound to membranes either as part of the protein complexes or, possibly, through selfassociation [4,5]. Monomers of α-syn contain three regions with distinct compositions of amino acids (a.a.): the amphipathic N-terminal region (a.a.  which can adopt alpha helical structure, the hydrophobic non-amyloid-β component (NAC) region (a.a 61-95) responsible for the formation of β-sheet structure, and an unstructured C-terminal region (a.a. 95-140) which includes 14 acidic a.a. [6,7] 4 Formation of α-syn fibrils is a complex process with many intermediates including oligomers, ring-like oligomers, and protofibrils [6,8]. The fibrillization process for amyloids is a nucleation dependent pathway [9,10], where post nucleation species of oligomers and protofibrils act as seeds for monomers, which accelerate the formation of insoluble fibrils [4,11]. It has been suggested that endocytosis of the misfolded protein may be the initial step in the replication of the misfolded protein structures by prion-like mechanisms [12][13][14].
Toxicity of α-syn in vivo is related to the oligomeric forms of α-syn [15] as well as purified protofibrils and fibrils [16]. Processes that generate seeds will cause α-syn to autocatalytically replicate via secondary nucleation, a general mechanism, which is observed in other misfolded protein diseases such as AD and prion diseases [4,12]. Pathological α-syn assemblies can replicate in vitro and in tissues [17,18]. Mutant strains of α-syn, A30P, E46K, and A53T, have natively indistinguishable monomer conformations, but are associated with early onset PD, likely due to accelerated spherical oligomer and fibril formation [9,19,20].
In solution, α-syn exists in equilibrium between extended and more compact conformations. The dynamic ensemble between these two conformational families can be shifted by environmental factors and ligand binding, [21] and pH of solution along with temperature, crowding agents, membranes, metal ions and interacting proteins have also been shown to modulate the conformational states of α-syn [22]. It is well documented that polyvalent metals increase the aggregation potential of α-syn and lead to increased fibril formation of distinct morphology and toxicity [23][24][25][26], and a shift to the more compact form of α-syn upon metal binding has been observed by ESI-IMS-MS [27][28][29], suggesting a role for the more compact conformations of α-syn as a precursor to amyloid formation [30,31], while a transient α-helical monomer may be a precursor to an alternative nucleation pathway [32,33].
An extensive number of small molecules have been identified to inhibit amyloidogenesis in AD and PD (reviewed in: [34]). Recently, studies on Lacmoid, Orcein and its derivatives, Congo red and EGCG, have explored alternative therapeutic strategies, such 5 as derailing amyloid formation, stabilizing mature fibrils and the depletion of oligomeric aggregation intermediates [4,21,[35][36][37][38][39]. Hasegawa and coworkers investigated the effect of 79 compounds belonging to 12 chemical classes on the assembly of α-syn into fibrils, and identified seven classes (polyphenols, phenothiazines, polyene macrolides, porphyrins, rifamycins, Congo red and its derivatives, and terpenoids) containing strong fibril inhibitory effects [40]. Much effort has been specifically focused around the study of polyphenols. This class of compounds is an important beneficial constituent in the human diet found in a wide range of fruits, vegetables, and beverages including tea and red wine [41]. Polyphenols such as resveratrol, curcumin, and methylene blue have been implicated as treatment in many neurodegenerative disorders, as well as inflammatory and common ophthalmic disorders [42][43][44]. It is suggested that phenol rings have a different mode of stacking compared to benzene rings, and that this difference allows phenol rings to non-covalently bond the amyloidogenic aromatic residues of monomers / oligomers, as well inhibiting efficient fibril assembly [45]. It has also been proposed that the presence of vicinal hydroxyl groups (in the manner of three > two > one -OH) hinders the progress of the fibril self-assembly process and/or destabilizes its amyloid structure and that antioxidants within the natural polyphenols scavenge reactive oxygen species as well as inhibit deposition of α-syn fibrils in the brain [41]. The presence of vicinal hydroxyl groups is a common feature of some of the most effective natural inhibitors of α-syn aggregation. In an assay of 169 compounds screened to inhibit α-syn fibril formation, all but one of the inhibitors was a catechol containing compound, including dopamine, L-dopa, norepinephrine, and epinephrine [46].
Brazilin, a natural compound extracted from the wood of Caesalpinia sappan, was recently shown to have an inhibitory effect on the fibrillogenesis and cytotoxicity of Aβ [47] and hIAPP [48]. Aβ aggregation is the crucial event in the pathogenesis of AD, but soluble Aβ oligomers are believed to be the main neurotoxic agents [47]. Oligomeric aggregation intermediates of α-syn, which can form membrane pores, are likewise toxic in neuronal models [49,50]. The structure of Brazilin is similar to other natural compounds shown to have 6 therapeutic effects on the aggregation of α-syn, namely the two vicinal hydroxyl groups and the two phenol rings (Figure 1a). The similarities of Aβ and α-syn aggregation combined with Brazilin's structural similarities to other α-syn fibrillogenesis inhibitors as well as the inhibitory effects Brazilin has on Aβ, suggest that Brazilin could be a key candidate as a potential neuroprotective and therapeutic agent for PD. Very recently, it was confirmed that Brazilin inhibits α-syn fibril formation in vitro [51]. Here, we analyzed in detail the mechanism how Brazilin disrupts α-syn self-assembly and inactivates seeding competent α-syn assemblies derived from PD brain homogenates. Our biochemical and molecular findings suggest Brazilin to be a potential therapeutic and neuroprotective agent for PD with a novel mechanism of action compared to other polyphenolic compounds.

Alpha synuclein expression
α-Syn was expressed in E. coli as described previously [35]. Bacteria were grown at 37° C in LB and ampicillin (100 μg/mL). α-Syn expression was induced with IPTG once the OD 600 reached 1.1. The bacteria were then grown overnight at 30° C. The bacterial suspension was spun for 30 minutes at 3166 ×g at 4°C. Two pellets of varying size were frozen at -80°C for 2 hours and then resuspended in 300 and 150 μL (large and small pellet respectively) of lysis buffer (10mM Tris HCl, 1mM EDTA, pH 8) containing Roche complete protease inhibitor. The solutions from the small and large pellets were then sonicated for 30 and 45 minutes respectively, boiled for 20 minutes, and spun down at 3185 ×g for 30 minutes at 4°C. The resulting α-syn pellet was collected and dissolved in 25 mM Tris, pH 7.7. The αsyn was purified using a FPLC Resource Q anion exchange column with a salt gradient from 0-1 M NaCl. The protein was dialyzed overnight in ammonium bicarbonate buffer (10 mM, pH 7.5), spun down at 3185 ×g for 30 minutes at 4°C, frozen and then lyophilized for storage. 7 Mutant K23Q α-syn for RT-QuIC assays was expressed in E. Coli and purified using an osmotic shock protocol as previously reported by Groveman et. al. (2018) [52].
ThT aggregation assays α-Syn (1 mg/mL) was dissolved in 10 mM NaOH and sonicated (VWR® Symphony ™ Ultrasonic Cleaner) for 20 minutes. The α-syn was then spun down at 90,900 ×g for 20 minutes at 4°C. The concentration of monomerized α-syn was then determined by OD280 using an extinction coefficient of 5960/M/cm. The α-syn was then aggregated in a nonbinding 96 well plate (Corning, #3651) at a concentration of 30 μM with intermittent shaking in aggregation buffer (100 mM NaP, pH 7.4, 10 mM NaCl, 0.1% NaN3, and 20 μM ThT). A 2 mm diameter glass bead was added to each well to accelerate the aggregation through stirring.
The plate was kept at 37° C and agitated by orbital shaking once every 1 minute for 5 seconds.
The ThT fluorescence was recorded with an excitation wavelength of 436 nm and an emission wavelength of 482 nm in a fluorescence plate reader (Infinite F200, Tecan).

Formation of α-syn Seeds for Seeded Aggregation Assays
α-syn seeds were generated by sonication of α-syn aggregates at different aggregation time points for 10 minutes in a water bath sonicator (VWR® Symphony ™ Ultrasonic Cleaner). 5% (m/m) α-syn seeds were then added to fresh α-syn monomers and aggregated with ThT as described above.

Atomic Force Microscopy
Aliquots of aggregated peptides (10 μL) were placed on clean, freshly cleaved grade V-1 mica (Cat#: 01792-AB, Structure Probe, Inc., USA). After 10 minutes, the solvent was wicked off by filter paper and the mica was washed 4 times with 20 μL of ultrapure (Milli-Q) water to remove 8 salts and buffer from the sample. Samples were dried overnight, and AFM images were acquired in tapping mode on a Veeco Dimension 3100 machine (Bruker) with Bruker FESP tips. Images were visualized using the Bruker Nanoscope software v1.5. overnight. Additional water washes were performed to de-stain the gel.

Circular Dichroism
The α-syn samples (50 μM) incubated under constant shaking at 37°C taken at various time points (24, 72, 168 h) were diluted 2:1 in aggregation buffer (100 mM NaP, pH 7.4, 10 mM NaCl, 0.1% NaN3, and 20 μM ThT). The 168 h incubation samples were sonicated for 2 minutes in a water bath sonicator (VWR® Symphony ™ Ultrasonic Cleaner). Circular dichroism (CD) spectra were recorded between 200 and 260 nm with a step size of 1 nm in a CD spectrometer (J-720, Jasco, Japan) [4]. The CD spectra of solutions without α-syn were subtracted as background from the CD signals with α-syn to isolate the α-syn-specific changes. All spectra were the average of 5 consecutive scans for each sample. software supplied with the mass spectrometer.

Brain Homogenate Preparation
All subjects provided consent to clinical assessment under UCSD IRB-approved protocol #080012, and had consented to their brains being obtained at autopsy. Clinical assessment and autopsy brain analysis were performed as outlined in [52]. Brain homogenates (BH; 10% w/v) were prepared by homogenizing the tissue in PBS using a Bead Beater (Biospec Products; 11079110z) for 1 min at maximum speed. The homogenate was then spun at 2000 ×g for 2 min at room temperature and the supernatant was transferred to a new tube and stored at − 80 °C for α-syn RT-QuIC analysis. For α-Syn RT-QuIC testing, brain homogenates were serially diluted in PBS.

Primary Neuronal Culture and Toxicity Analysis
Primary neuronal cultures were derived from brains of inbred FVB/N mice. Hippocampi of E17 mouse brains were dissected in HBSS (ThermoFisher Scientific) supplemented with 1% Lglutamine, 1% HEPES and 1% pen-strep. Cells were dissociated using 0.25% trypsin+0.04% benzonase, triturated mechanically and counted using a Neubauer haemocytometer. Cells were plated in DMEM supplemented with 10% horse serum (#26050-88, Invitrogen) at 10K/well to the inner 60 wells of poly-L-lysine-coated 96 well plates (Greiner, 655936). At 1 h post-plating, DMEM medium was aspirated and exchanged for Neurobasal medium (21103049, ThermoFisher Scientific) supplemented with 0.25% Glutamax (35050061, ThermoFisher Scientific) 2% Gibco B27 supplement (17504044, ThermoFisher Scientific) and incubated at 37°C (20% O2, 5% CO2). FVB/N neurons were maintained in culture for 11-12 days prior to a 72 h treatment. Images of live cells were taken on an IncuCyte S3 reader (Sartorius) with a 20x objective in phase contrast. 4 views were captured in each well every 4 h. 11 Neurite lengths were evaluated using the NeuroTrack module of the IncuCyte S3 software package (rev 2019A) using the following parameters: cell body cluster segmentation 0.7; cleanup 0; cluster filter 0; neurite sensitivity 0.25; neurite width 1 µm. Detected neurite masks are highlighted in pink in images. Neurite length data were normalized to the initial (0 h) value for each well and means ± standard deviation were calculated from quadruplicate sample wells. ANOVA statistical analysis (two-tailed, equal variances) was performed in GraphPad Prism.

Transmission Electron Microscopy
End-point samples from wt α-syn aggregation assays (5 µL) were loaded onto carbon-coated 300 mesh copper grids (Electron microscopy Sciences) that had been glow discharged for 40 seconds using an PELCO easiGLOW™ glow discharge unit (Ted Pella Inc., USA).
Samples were left to bind for 1 min, blotted dry, washed in water (3 x 30 µL), blotted, and then stained with 10 µL Nano-W (methylamine tungstate) stain (Nanoprobes) for 1 min. Images were acquired on a FEI Tecnai T10 electron microscope (FEI, Eindhoven, NL). RT QuIC samples were collected for TEM imaging after 16 h of incubation. To collect solutions, a pipet tip was used to vigorously scrape the well surfaces and pipet the solution. 2-8 wells were pooled for each reaction condition and the solutions briefly sonicated. Ultrathin carbon on holey carbon support film grids (400 mesh, Ted Pella) were briefly glow-discharged before being immersed into droplets of the fibril solutions for 30-60 min at room temperature. Grids were sequentially washed three times in MilliQ water before being negatively stained with Nano-W stain and wicked dry. Grids were imaged at 80 kV with a Hitachi H-7800 transmission electron microscope and an XR-81 camera (Advanced Microscopy Techniques, Woburn, MA).

Molecular Modelling
Setup of systems. 1) Unfolded α-syn monomers were generated by selecting the ten bestenergy structures from a brief minimization (7000 EMC steps) of a pool of 60000 randomly generated conformations using the ab initio peptide folding program RAFT [53] while weighting the residues psi/psi angle selection by secondary structure propensity. Repeating this process 100 times gave a population of 1000 viable random coil structures that were converted to allatom models with SCWRL4 [54]. One hundred structures were chosen from the pool at random and each placed in a simulation box 2 nm larger than the protein in every dimension for MD. BUDE [55,56] was used in surface-scanning mode to locate the best ten nonoverlapping Brazilin binding sites on an α-syn monomer as the start point for the 100 simulations of α-syn monomers bound to 10 Brazilin molecules. 2) A bespoke program was used to pack 64 of the monomer initial simulation boxes into the smallest cube giving a 46 nm cube. Performing this with and without Brazilin bound gave the starting points for the two large simulations to explore aggregation. 3) The fibril fragment containing 12 α-syn monomers was centered in a 15.67 nm cube. BUDE was used to define a 12 X 1.2 nm grid within the box and add Brazilin in random orientations to this grid. Brazilin molecules (222, not overlapping the protein) were selected at random.
Simulations: All molecular dynamics simulations were performed with GROMACS 2018 [57] under the following protocol: Hydrogen atoms were added to α-syn consistent with pH 7. The periodic-boundary boxes were filled with TIP3P waters and 0.15 M sodium chloride to neutrality. The system was parameterized with the amber99SB-ildn forcefield [58]. Brazilin was parameterized as a mixture of both tautomeric catechol mono-anions with the general Amber force field (GAFF) [59]. Short range electrostatic and van der Waals interactions were truncated at 1.4 nm and long-range electrostatics treated with the particle mesh Ewald method. An initial relaxation of 10000 steps of steepest descents energy minimization was performed. Subsequently, 0.2 ns of dynamics at 310 K was initialized while tethering the protein to its initial position, then the position restraints removed for production runs. The simulations were performed under periodic boundary conditions. The temperature was 13 maintained at 310 K using the v-rescale thermostat and the pressure at 1 bar with the Berendsen barostat. Twin temperature baths were used, one for the protein and the other for the water and ions. Bond constraints were applied to the water (SETTLE) and the protein (LINCS) to allow a 2 fs timestep for the leap-frog integrator.
Analysis: The simulation data were analyzed using the tools in the GROMACS suite and with bespoke programs written in-house.

Brazilin Does not Delay Spontaneous α-syn Aggregation but Inhibits Conformational Change of α-syn to β-Sheet
To examine the inhibitory effects of Brazilin on the aggregation of α-syn, aggregation was monitored using a thioflavin-T (ThT) fluorescence assay. ThT increases in fluorescence upon binding to β-sheet rich amyloid-like structures [36,60]. These data could indicate that (a) Brazilin prevented α-syn incorporation into amyloid structures, (b) acted during the later stages of α-syn aggregation, permitting the formation of an early aggregate species with reduced ThT binding or (c) that it competitively inhibited ThT binding without affecting fibril formation [61]. In the latter case, it would be expected that αsyn aggregates would retain their β-sheet structure in the presence of the small molecule. We 14 measured the secondary structure of α-syn in the presence and absence of Brazilin using circular dichroism (CD) over the course of one week to distinguish between these hypotheses. CD spectra of α-syn at t = 0 h and 24 h are consistent with a lack of secondary structure, as has been previously observed [4,35,49,62] (Figure 1d-e). Upon protein aggregation, CD spectra of untreated samples show a minimum at 218 nm while α-syn treated with Brazilin fully (10x) or partially (1x) retains the unstructured state (Figure 1f-g), which demonstrate that Brazilin inhibits the formation of β-sheet structures of α-syn in vitro. ThT fluorescence recorded from CD samples matches the inhibition of β-sheet formation (Figure 1h), excluding that ThT inhibition merely reflects competitive binding of the small molecule. ThT and CD data would be consistent with a mechanism, in which Brazilin suppresses a large percentage of amyloid formation at the primary nucleation stage, but does not substantially affect the lag phase, because overall kinetics are dominated by secondary nucleation processes.

Aggregates
To test whether the reduction in ThT signal reflected a reduction in α-syn fibril formation, we imaged the samples analyzed in Figure 1d  In order to quantify aggregation, the same samples used for CD and AFM were centrifuged at 100,000 ×g and total (T), soluble (S) and pellet (P) fractions were analyzed b y SDS PAGE and Coomassie blue stain (Figure 3). Initially all α-syn was soluble and ran as a 14 kD monomer band. At t = 72 h, non-treated α-syn had almost completely aggregated, while α-syn treated with 1x and 10x Brazilin stayed soluble. Tenfold excess of Brazilin quantitatively prevented α-syn aggregation even after one-week incubation (Figure 3d). Taken together, the data from ThT fluorescence, CD, AFM, TEM and ultracentrifugation indicate that Brazilin prevents the formation of α-syn amyloid fibrils and keeps the protein in an unfolded state, which is most likely the monomer. Interestingly, a small amount of SDS-resistant α-syn dimer appeared in Brazilin treated samples, which may reflect the amorphous aggregates observed in AFM and TEM.

Brazilin Binds Preferentially to the Compact Conformation of α-syn
We analyzed α-syn by native ESI-IM-MS to determine how Brazilin prevented aggregation. The native mass spectrum of the sample treated with 2% DMSO only (same amount of DMSO in the Brazilin-treated sample) showed predominantly monomer species of α-syn (Figure 4a). Based on the bimodal charge state distribution observed in the spectrum, two major conformational families spanning from 15+ to 4+ were detected, reflecting extended and compact conformations of the protein, respectively ( Figure S1) [21,30].
As shown in Figure 4a, Brazilin bound preferentially to the low charge states of α-syn (5+ to 7+) with up to eight Brazilin ligands associated with a single molecule at the 1:5 (mol/mol) equivalent of α-syn and Brazilin. Notably, no brazilin bound to the more highly charged states representative of the extended form of the protein. Such a binding "selectivity" for the more compact lower charge states was also previously found for EGCG, whereas dopamine showed the exact opposite behavior by binding preferentially to the more extended highly charge states of α-syn [21]. It is possible that multiple molecules of Brazilin bind to each

Brazilin Does not Prevent the Recruitment of α-syn Monomers into Amyloid Fibrils
Preformed fibril seeds obviate nucleation [63], allowing us to analyze to effect of Brazilin on fibril growth. Solutions of α-syn (30 μM) were prepared with 5% (w/v) seed, formed by sonication of α-syn fibrils, and 3, 15, 30, 60, and 300 μM concentrations of Brazilin to test whether the small molecule prevented the recruitment of α-syn monomers to preformed fibrils (Figure 5a, b). Similar to de novo aggregation, Brazilin reduced the amplitude of ThT fluorescence in a concentration dependent manner. However, after normalizing ThT amplitudes to the final ThT amplitude, it was clear that Brazilin concentration has no effect on the seeded aggregation of α-syn. We calculated apparent growth rates from the initial slopes of ThT kinetics (Figure 5c, d). These growth rates were similar across all concentrations of small molecule when normalized for ThT amplitudes. We quantified total (T), soluble (S) and aggregated (P) α-syn from the end point of the seeding assay by ultracentrifugation and SDS-PAGE (Figure 5e). Their relative fractions remained unchanged by Brazilin, confirming that the small molecule did not directly inhibit the incorporation of α-syn monomer into growing fibrils.

Brazilin Inactivates Synthetic and Brain-derived α-syn Seeds
We then tested whether Brazilin could affect the activity of α-syn seeds themselves. α-Syn fibrils (30 μM monomer) were formed at 37 °C in phosphate buffer in a fluorescence plate reader and Brazilin was added after 65 h incubation (Figure 6a). We observed a gradual concentration dependent decrease in ThT fluorescence (t½ ~ 3h) upon addition of Brazilin as compared to the control sample, which may indicate a structural remodeling of the fibrils.
The end products of Brazilin treatment were then sonicated in a water bath for 15 minutes to generate seed, which were added to fresh α-syn monomer (Figure 6b). Treatment with Brazilin reduced seeding capacity in a concentration-dependent manner, indicated by the increase in lag phase when compared to untreated seeds, with 10x Brazilin having aggregation kinetics similar to that of untreated α-syn. In order to ensure that the effect was not due to residual Brazilin present in the seed solution from seed generation, we compared seeding kinetics of untreated fibrils in the presence and absence of Brazilin (15 µM, equivalent to the final concentration for seeds treated with 300 µM Brazilin), but found no effect on fibril growth rate, when normalizing for the reduction in ThT amplitude caused by the compound ( Figure S2).
However, TEM revealed that amyloid fibrils of α-syn were still present after Brazilin treatment ( Figure 6c) and quantification of aggregated α-syn via ultracentrifugation revealed 18 no solubilization of α-syn fibrils by the Brazilin treatment, but rather the presence of fibril clusters (Supplementary Figure 3, Figure 6d). Rather, we found pelleted α-syn in insoluble high molecular weight aggregates that did not enter the SDS-gel after Brazilin treatment. This suggests that Brazilin, rather than dissolving amyloid fibrils, promoted the formation of large fibrillar assemblies, which were more stable and less seeding competent than untreated α-syn fibrils. Similar SDS-resistant high molecular weight aggregates were observed at high Brazilin concentrations in the previous seeding experiment (Figure 5e), supporting the interpretation that Brazilin coats and stabilizes large α-syn fibrils.
Finally, we tested the capability of Brazilin to inactivate α-syn from Parkinson's disease brain homogenates. Seeds of α-syn can be amplified in vitro and detected via ThT fluorescence in real-time quaking-induced conversion assays (RT-QuIC; Figure 7a) [17,52].
Here, the seed concentration provided to the reaction can be estimated to be roughly 10 8 -fold lower than the seed concentration used above in Figure 5 based on comparisons of end-point dilution RT-QuIC titrations of Parkinson's disease brain and known quantities of synthetic αsyn fibrils [52]. Under these conditions, Brazilin treatment prevented the induction of ThT positive aggregates. Control brain homogenates from control corticobasal degeneration (CBD) patients did not induce ThT fluorescence. Imaging by TEM confirmed that Brazilin treatment prevented the formation of fibrillar α-syn aggregates in RT-QuIC (Figure 7b). Conversely, RT-QuIC reactions seeded with CBD brain homogenates did not show α-syn fibrils ( Figure S3).
Brazilin was also tested in an RT-QuIC reaction using a dilution series of PD brain 3089 at a sub-stoichiometric concentration relative to α-syn monomers, and was found to completely inhibit the seeding of α-syn relative to positive controls, demonstrating potent inhibition of fibril formation ( Figure S4a, b). The potent inhibition of seeding under RT-QuIC condition seemingly contradicts the observation that Brazilin does not directly inhibit fibril elongation ( Figure 5). However, with the much lower initial seed concentration in the RT-QuIC reactions, we assume that fibril elongation alone, i.e., without the exponential seed amplification provided by secondary nucleation via lateral nucleation of fibril formation, would not be detectable. Thus, these data are consistent with the notion that Brazilin markedly reduced the activity of Parkinson's-associated α-syn seeds by reducing secondary nucleation on fibril surfaces, even if, as suggested by data in Figure 5, primary fibril elongation was unimpeded.

Brazilin Removes α-Syn Toxicity in Primary Neurons
We then tested whether Brazilin could reduce the toxicity of preexisting α-syn assemblies using a novel method, which monitors neurite length of primary mouse hippocampal neurons in response to amyloid aggregates [64]. Fibrils were treated with Brazilin for 24 h at 1x, or 2x molar ratios (Figure 8a) and then added to neurons at 1 µM monomer equivalent concentration. Live neurons were imaged every 6 h and their neurite lengths were quantified relative to their pre-treatment state (Figure 8a, b). After 72 h incubation, α-syn treatment significantly reduced neurite lengths, whereas neurite lengths with Brazilin treated fibrils were not significantly different from untreated controls. Brazilin itself and monomeric α-syn did not reduce neurite lengths compared to buffer controls ( Figure S5).
In total, these data indicate that Brazilin treatment potently inactivates seedingcompetent and neurotoxic α-syn assemblies. They furthermore suggest that, unlike polyphenols such as EGCG, but similar to Orcein and its derivatives, it does so by promoting the formation of large insoluble aggregates over smaller seeding competent species. This in contrast to Brazilin's effect on Aβ42, where it was reported that Brazilin dissolved and remodeled amyloid fibrils into granular aggregates [47]. 20 Firstly, we explored the binding of Brazilin to α-syn monomers using molecular dynamics. The peptide structure-prediction software, RAFT [53] was used without extensive minimization to produce 1000 viable random coil structures giving an ensemble of unstructured α-syn monomers. One hundred of these were chosen at random and each simulated for 100 ns in explicit water both with and without Brazilin present, giving a total of 20 µs of trajectory data. The simulations with Brazilin were set up by docking 10 ligand molecules to the monomer chain using BUDE [55,56] Quantitative analysis of the simulation data shows a rapid equilibration such that on average about 80% of Brazilin remains bound to the monomeric α-syn ( Figure S7a). The presence of Brazilin has only modest effects on the evolution of the average radius of gyration of the monomer population ( Figure S8) and has a slight suppressing effect on the α-helical content of the monomers ( Figure S9). From these data we can deduce that Brazilin binds to the monomers without greatly influencing the structural ensemble adopted by the unfolded monomers.

Molecular Modeling of Brazilin Binding
Next, we looked at the influence of Brazilin on the aggregation behavior of the monomers both with and without Brazilin present. Bespoke software was used to pack 64 of the orthorhombic boxes from the starting structures of the 100 monomer simulations into a cube with the smallest volume. The resulting cubic box was 46 nm on each side and after resoaking with model water and salt, contained nearly 10 million atoms. The size of these systems (64 α-syn monomers with and without 640 Brazilin molecules, ~1 mM and ~10 mM 21 respectively) means that sampling was limited to 200 ns for both simulations. Nonetheless, these calculations produced some interesting indications of behavior. As the simulations evolve a certain amount of aggregation of α-syn is observed. Intriguingly, when Brazilin is present there is a shift towards small aggregates of two or three monomers compared with αsyn alone where more monomers are present ( Figure S10). Like the individual monomer simulations, the percentage of Brazilin molecules bound to the 64 chains in this simulation is around 80%, corresponding to 8 molecules per α-syn monomer ( Figure S7a, b). Movies 2a-c illustrate the initial starting structure in the box and the final structures with and without Brazilin bound. These images correspond to the aggregate-size data shown in Figure S10.
Finally, to investigate the affinity of Brazilin to a fibril structure, we simulated a fibril fragment (PDB code 6a6b) with and without Brazilin. In this case we loaded the simulation box with 222 Brazilin molecules (~100 mM) to ensure extensive surface sampling to locate the best Brazilin binding regions on the fibril fragment. Since the α-syn monomers are truncated in this cryoEM structure, missing the first 36 and last 41 residues, these simulations are not directly comparable with the full-length monomer simulations described above. The fibril fragments are stable during the simulation as expected, with Brazilin enhancing rigidity in the structure ( Figure S11). Inspection showed that Brazilin suppresses the motions of the N and C terminal regions, while the fibril core structure remains rigid in both simulations. The relative affinity of Brazilin for different parts of the structure is illustrated in Figures 9c, 9d and S12, showing that binding to the exterior of the fibril is favored over the faces where new α-syn monomers would bind to extend the fibril. Unsurprisingly, the fibril structure occludes some of the binding sites along an α-syn strand that are occupied by Brazilin in the monomer simulations ( Figure S12).
It has been shown that Brazilin can efficiently inhibit amyloid fibrillogenesis, inactivate mature fibrils and reduce cytotoxicity of both Aβ and human islet amyloid polypeptide (hIAPP), and α-syn [47,48,51]. On the surface, its mechanism seems to mirror EGCG, in that Brazilin inhibits fibril formation and is able to reduce the ThT fluorescence and toxicity of preformed amyloid fibrils [47,51,64]. However, our analysis of Brazilin's effect on the de novo aggregation of α-syn, as well as its ability to remodel mature α-syn fibrils, reveals that its effect on α-syn is mechanistically distinct from EGCG and from previous reports of Brazilin on Aβ and hIAPP.
CD data demonstrate that Brazilin is able to maintain α-syn as an unfolded monomer ( Figure 1). Prior research has shown that unfolded α-syn exists as a conformational ensemble around two forms, one extended and one partially compact, and that the partially compact form is a precursor to amyloid formation [30,31]. From native IM-MS data we conclude that Brazilin is binding specifically to the compact conformation of α-syn between charge states 5+-7+ ( Figure 4). Interestingly, it has been shown that EGCG, and several other potent inhibitors of α-syn aggregation also bound specifically to the compact conformation, while dopamine binds specifically to the extended conformation of α-syn [21,30,65] , which may suggest a common specific mechanism of polyphenolic inhibitors on α-syn monomers.
While it is important not to over-interpret the limited sampling available from the molecular dynamics simulations, both single-monomer trajectory data and simulating 64 monomers in a box yielded a binding stoichiometry around 8 Brazilin molecules per monomer that was similar to that observed in ESI-MS experiments ( Figure S6). Correspondingly, Brazilin binding occurs throughout the α-syn chain, but affinity decreased near the C-terminus ( Figure   S12). Simulations suggest that the small molecule has little effect on the average secondary structure of the protein apart from a small reduction in alpha helicity and a corresponding small increase in coil and turn conformations ( Figure S8), corresponding to experimental data from CD experiments. The observation of more dimeric and trimeric aggregates in the simulation of 64 α-syn molecules at the expense of monomers when Brazilin is present ( Figure S9) is suggestive of stabilization of α-syn aggregates, again consistent with experimental data collected here (Figure 3d).
Simulation of the fibril fragment shows that Brazilin prefers to bind to the exterior surface of the fibril rather than to the two faces where recruitment of α-syn molecules would occur to extend the fibril length (Figures 9 and S12). In brief, the modelling results are consistent overall with a mechanism of amyloidosis suppression whereby Brazilin binds to αsyn monomers and small oligomers tightly enough to inhibit the formation of seeds that initiate fibrilization.
However, our data suggest that the effect of Brazilin on preformed amyloid fibrils is unlike other polyphenols like EGCG. Brazilin inactivates seeding competent assemblies, both in vitro and when treating assemblies in the brain of PD patients (Figures 7 and 8). While reduction in ThT fluorescence and the absence of single fibrils in AFM and EM images would suggest fibril disassembly [47,51,64], increased insoluble high-molecular weight aggregates were observed by SDS-PAGE in the pellet fraction of ultracentrifugation experiments ( Figure   3). The remodeled fibrils have a similar morphology to that of untreated α-syn; but appear in larger clusters (Figure 6c), which could explain their increased SDS-resistance and their decreased seeding capacity [4]. While Liu et al. did not observe fibril clusters under their experimental conditions, Brazilin slightly altered the morphology preformed α-syn fibrils at 2:1 molar ratio, resulting in wider fibrils, which may be precursors of the fibril clusters we found at 10:1 molar ratio ( Figure 6) [51].
Our results suggest that Brazilin inactivates seeding competent fibrils not by disassembly, but by promoting their assembly into large, inert aggregates, similar to the natural phenoxazine Orcein and derived compounds ( Figure 10) [4,52]. Correspondingly, Brazilin did not directly inhibit fibril elongation when fibrils are present at large concentrations ( Figure S2). Likely, formation of large aggregate clusters limits autocatalytic replication of αsyn by secondary nucleation on fibril surfaces [4]. Secondary nucleation dominates the α-syn assembly mechanism under physiological conditions [66], and which can autocatalytically replicate. These conditions are recapitulated by the RT-QuIC assay, in which Brazilin proves to be a very potent inhibitor of the replication of misfolded α-syn assemblies.
Correspondingly, sequestering smaller amyloid species into larger aggregates removes α-syn toxicity [4,52]. This is in agreement with the reduced toxicity of α-syn fibrils after Brazilin treatment in neuronal model cells [51] which has been confirmed in primary neurons in our study (Figure 8). Our molecular modelling suggests binding of the Brazilin anions to hydrophobic pockets on the fibril surface, specifically in the vicinity of the positively charged residues K43 and K45 (Figure 9), which was also reported by Liu et al., that may promote lateral assembly of fibrils ( Figure 10).
While our study cannot directly address the therapeutic potential of Brazilin, its inactivation of brain-derived α-syn assemblies suggests that it may be active in vivo. Previous studies on the pharmacokinetics of Brazilin found it to be highly stable in vivo. After a single oral dose or intravenous injection of 100 mg/kg in rats, the highest Brazilin concentrations in plasma were 82 μg/mL with half-lives of 4.5 or 6.2 h, respectively [67,68]. Brazilin has also been shown capable of crossing the blood brain barrier (BBB); after an intravenous injection of 50 mg/kg to rats Brazilin could be detected in the brain with an AUC of 340 ± 30 ng h/mL 25 and a Cmax of 254 ± 15 ng/mL [69]. This implies that Brazilin could cross the BBB in humans and target α-syn in a patient's brain, making it an appropriate CNS small molecule [47]. Based on its protective activity in vitro and its promising pharmacokinetic properties, Brazilin may be a strong candidate as a neuroprotective and therapeutic agent in PD.