Catechol-Containing Compounds are a Broad Class of Protein Aggregation Inhibitors: Redox State is a Key Determinant of the Inhibitory Activities

Background: A range of neurodegenerative and related aging diseases, such as Alzheimer’s disease, Parkinson’s disease, and type 2 diabetes, are linked to toxic protein aggregation. Yet the mechanisms of protein aggregation inhibition by small molecule inhibitors remain poorly understood, in part because most protein targets of aggregation assembly are partially unfolded or intrinsically disordered, which hinders detailed structural characterization of protein-inhibitor complexes and structural-based mechanistic elucidation. Methods: Herein we employed a small molecule screening approach to identify inhibitors against three prototype amyloidogenic proteins in neurodegeneration and related proteinopathies: amylin, Ab and tau. We further systematically investigated selected class of inhibitors under aerobic and anaerobic conditions to uncover a key determinant of the inhibitory activities. Results: One remarkable class of inhibitors identied from all three parallel screenings against different amyloidogenic proteins was catechol-containing compounds and redox-related quinones/anthraquinones. Further mechanistic studies determined that the redox state of the broad class of catechol-containing compounds is a key determinant of the amyloid inhibitor activities. Conclusion: Our small molecule library screening platform was able to identify a broad class of amyloid inhibitors. Redox was found to be a key factor not only regulating the inhibitory activities but also involving the mechanism of inhibition. The molecular insights we gained not only explain why a large number of catechol-containing natural compounds, often enriched in healthy diet, have anti-neurodegeneration and anti-aging activities, but also could guide the rational design of therapeutic or nutraceutical strategies to target a broad range of neurodegenerative and related aging diseases.

involving the mechanism of inhibition. The molecular insights we gained not only explain why a large number of catechol-containing natural compounds, often enriched in healthy diet, have antineurodegeneration and anti-aging activities, but also could guide the rational design of therapeutic or nutraceutical strategies to target a broad range of neurodegenerative and related aging diseases. respectively. 300 mesh formvar-carbon-coated copper grids and uranyl acetate replacement solution (UAR) were purchased from Electron Microscopy Sciences (Hatfeild, PA).
Lyophilized amylin powder (0.5 mg) was initially dissolved in 100% HFIP at a nal concentration of 1-2 mM. The additional lyophilizing step was employed to eliminate traces of organic solvents, which have been shown to affect amylin aggregation. Aliquots were either lyophilized again prior to use in cell-based assays or dissolved directly into DPBS, 10 mM phosphate buffer pH 7.4 or 20 mM Tris-HCl pH 7.4 for all amylin amyloid-related in vitro assays. All remaining 1-2 mM stocks in 100% DMSO were stored at -80 °C until later use. The lyophilized powder from all compounds and ThT were dissolved in DMSO (10 mM) and distilled water or relevant buffer (1-4 mM). These stocks were stored at -20 °C until later use.
Residual DMSO in the nal samples used for all in vitro assays ranged from 0-9.5%. We determined that these DMSO concentrations had negligible effects on amylin amyloid aggregation as re ected by ThT uorescence, TEM, PICUP assay, and inhibitor-induced amylin amyloid remodeling assays.
UV-Vis absorption spectra were collected on a Varian Cary 50 UV-Vis spectrophotometer. Equal concentration of catechol, norepinephrine, and rosmarinic acid was used for each compound. Aerobic and anaerobic conditions referred in each case as exposure to air or incubation in an anaerobic chamber (described below) at the speci ed duration. (c) Screening against Aβ42 peptide In contrast to the ThT uorescence assays for amylin and 2N4R tau, non-continuous, two-point measurements were taken to represent starting and ending (at plateau) uorescent signals for Aβ42. Unaggregated solutions of Aβ42 were typically prepared by dissolving lyophilized Aβ42 powder with 100% HFIP. Subsequent peptide quanti cation of this solution was estimated based on a standard micro plate BCA assay. Initial stock solution preparation of Aβ42 for the ThT assay were prepared by evaporating HFIP treated stocks followed by a modi ed two-step aqueous dissolution process as previously described: Herein, HFIP evaporated stocks of Aβ42 were dissolved in 60 mM NaOH. Next, these solutions were further diluted in 1X DPBS such that the nal solution consisted of approximately 2.4 mM NaOH (i.e. 4% 60 mM NaOH, 96% 1X DPBS). Immediately after the addition of 1X DPBS, 6.48 µL aliquots of Aβ42 and 0.52 µL aliquots of compound or buffer treated controls were distributed to the ThT reading plates. All plates were sealed to prevent evaporation and allowed to incubate at 37 °C until the estimated plateau of aggregation (24 hours). Next, the reading plates were centrifuged prior to receiving 14 µL of a 45 µM ThT solution prepared in 50 mM Glycine-NaOH, 8.6 pH. Finally, all plates were read at excitation 444 nm/emission 490 in order to estimate plateau phase amyloid aggregation as indicated by ThT uorescence. Note, the nal concentration of Aβ42 prior to being titrated with ThT, was approximately 12 µM, at a 3:1 molar ratio of compound to Aβ42.

Aggregation Assays in Anaerobic Chamber
All reagents including buffers, protein and compounds were made anaerobic by ushing with nitrogen prior to being placed within an anaerobic chamber (glove box). H 2 gas mixed with inert gas is circulated through metal catalyst to remove O 2 gas in the chamber. Oxygen level was maintained between 15-50 ppm within the anaerobic chamber. Typical aggregating conditions within the chamber were maintained at room temperature (25-30 °C). All anaerobic chamber-based ThT uorescence assays were discontinuous, two point measurements (starting point, ending or plateau point).

ThT Fluorescence-Based Amyloid Remodeling Assays
Similarly with regular ThT uorescence amyloid formation assays, remodeling assays extended the monitoring of the uorescence signals continuously after an inhibitor or a control compound was spiked into the aggregation system (amyloidogenic protein, ThT, buffer, heparin the case of tau protein). The volume of spiked compound was made to 5% of total sample volume in each well such that the baseline uorescence signal was minimally changed.
Transmission Electron Microscopy (TEM) Analysis TEM images were obtained by a JEOL 1400 microscope operating at 120 kV. Samples consisting of 30 µM amylin (20 mM Tris-HCl, 2% DMSO, pH 7.4) in the presence of drug or vehicle control were incubated for ≥ 48 hours at 37 °C with agitation. Prior to imaging, 2-5 µL of sample were blotted on a 200 mesh formvar-carbon coated grid for 5 minutes and then stained with uranyl acetate (1%). Both sample and stain solutions were wicked dry (sample dried before addition of stain) by lter paper. Qualitative assessments of the amount of brils or oligomers observed were made by taking representative images following a careful survey of each grid (> 15-20 locations on each grid were surveyed).

Gel-Based Amyloid Remodeling Assay
Vehicle control or speci ed compounds were spiked into freshly dissolved amylin samples (containing amylin and buffer). Thereafter amylin aggregation was allowed to proceed for 3 days. Final amylin concentration was 15 µM that included 45 µM of compound (drug:amylin molar ratio was 3:1). After 3 days, these samples were vacuum dried and re-dissolved in 6.5 M urea containing 15 mM Tris and 1X SDS Laemmeli sample buffer, boiled at 95 °C for 5-10 minutes and subjected to SDS-PAGE followed by Western blot analysis with anti-amylin primary antibody (T-4157, 1:5000, Peninsula Laboratories, San Carlos, CA). All gel-based amyloid remodeling assays were repeated at least twice.

Photo-Induced Crosslinking of Unmodi ed Proteins (PICUP) Assay
Amylin aliquots from a master mix in 10 mM phosphate buffer, pH 7.4 were added separately to 0.6 mL eppendorf tubes containing small molecule inhibitors or DMSO vehicle loaded controls. Crosslinking for each tube was subsequently initiated by adding tris(bipyridyl)Ru(II) complex (Ruby) and ammonium persulfate (APS) (Typical amylin:Rubpy:APS ratios were xed at 1:2:20, respectively, at a nal volume of 15-20 µL), followed by exposure to visible light, emitted from a 150-Watt incandescent light bulb, from a distance of 5 cm and for a duration of 5 seconds. The reaction was quenched by addition of 1X SDS sample buffer. PICUP results were visualized by SDS-PAGE (16% acrylamide gels containing 6 M urea), followed by silver staining. The binary solvent system was composed of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). A 10-minute gradient was used for the analysis with the following conditions: initial and hold for 30 s at 1%B, a linear gradient to 90%B at 8 min, hold at 90%B to 8.5 min and return to initial condition at 9 min. Sample volumes for MS analysis were between 2-4 µl and injected onto a Waters UPLC BEH C18 (1.7 µm, 2.1 mm x 50 mm) held at 35 °C for MS analysis. Source conditions for the mass spectrometer were: capillary 2.8 kV, source temperature 125 °C, sample cone 30V, source offset 80V, desolvation gas 500 L/Hr, desolvation temperature 400 °C, cone gas 50 L/Hr, and nebulizer gas 6 par. The m/z scan range was 50-1800 for MS analysis and 50-600 for MS/MS analysis and a collision energy ramp of 15-40 eV was used for MS/MS analysis.

Statistical Analysis
All data are presented as the mean ± S.E.M and the differences were analyzed with a one-way analysis of variance followed by Holm-Sidak's multiple comparisons (amylin kinetics) or unpaired Student's t test. These tests were implemented within GraphPad Prism software (version 6.0). p values < 0.05 were considered signi cant.

Results
One of the salient ndings from the screening is that catechols and redox-related quinones/anthraquinones represent a broad class of amyloid inhibitors. As shown in Fig. 1A, these molecules made up a substantial portion of the identi ed strong inhibitors, which were de ned as exhibiting greater than three standard deviation units below the ThT RFU observed for buffer treated individual amyloidogenic protein controls (dotted line). Fully 13 out of 41 strong inhibitors (32%) for amylin, 11 out of 22 (50%) for tau (2N4R isoform), and 14 out of 29 (48%) for Aβ were catechols or quinone/anthraquinones (highlighted as red dots). In the screens against amylin amyloid, out of 22 catechols and quinones/anthraquinones from the NIHCC library, 21 of them exhibited signi cant amyloid inhibitory activities (Table S1) with isoproterenol being the exception, with only a weak inhibitory effect. The majority of these catechols and quinones/anthraquinones (16 out of 22 drugs) were further validated by an orthogonal biochemical assay, Photo-Induced Cross-linking of Unmodi ed Proteins (PICUP), which identi ed cross-linked oligomers (Figs. S1B and S3A), or by a biophysical method, transmission electron microscopy (TEM; Fig. 1C, Fig. S1C and Table S1).
Similar with amylin screening results, 17 out of 22 catechols and quinones/anthraquinones showed signi cant activities against tau 2N4R amyloid (Table S2), with a majority of them further validated by a ThT uorescence-based tau amyloid remodeling assay, accomplished by spiking testing compounds into pre-formed tau amyloids (Fig. S3B). Fully 18 out of 22 catechols and quinones/anthraquinones exhibited signi cant inhibition against Aβ amyloid, with nine of them validated previously by assays including TEM and atomic force microscopy (AFM) ( Table S3). Chemical structures of catechols and quinones/anthraquinones of the top hits are shown (Fig. S2). Numerous hits such as idarubicin (Compound #318), daunorubicin (Compound #321), and rifapentine (Compound #601), showed strong inhibitory effects to all three amyloidogenic proteins (boxed red dots in Fig. 1A), whereas a few hits displayed preferential inhibition, such as rutin (Compound #548; Tables S1-S3), which preferentially inhibited amylin amyloid formation.
To test the hypothesis whether the catechol functional group alone inhibits amyloid formation, we tested catechol in secondary assays and compared it with its control analog phenol. Catechol demonstrated moderate, yet signi cant activities in amylin amyloid inhibition in both ThT uorescence assays and TEM analyses, whereas phenol showed no such inhibitory effects in both assays (Figs. 1B and 1C). Another class of chemical structures enriched in our screens was the anthraquinones/quinones (redoxrelated to catechols) and tetracyclines. Anthraquinones were previously observed to inhibit tau aggregation (Pickhardt et al, 2005) and quinones were reported to inhibit insulin oligomerization as well as bril formation (Gong et al, 2014). With respect to tetracycline, our screen revealed that several variants were active in amyloid inhibition. This class of compounds was reported to also inhibit Aβ and Based on the fact that catechol-containing compounds and multiple anthraquinone/quinone compounds (redox related to catechols) exhibited strong anti-amyloid activities, we hypothesized that catechol autoxidation may be part of the general mechanism that signi cantly enhances the anti-amyloid activities of the catechol-containing compounds. To test this hypothesis, we compared the anti-amyloid activities of a collection of oxidized (or aged -exposed to the air for 48 hours) with non-oxidized (nonaged or freshly prepared) catechol-containing compounds. In virtually all cases, aged samples exhibited signi cantly greater activities than their identically prepared non-aged counterparts ( Fig. 2A). Oxidationinduced activity enhancement was not observed with phenol nor a structurally similar but non-catechol amyloid inhibitor, morin. These combined results strongly suggested a catechol-dependent enhancement speci city, which was recapitulated under stringently de ned aerobic/anaerobic conditions using an anaerobic chamber. Aerobic, but not anaerobic conditions, signi cantly enhanced anti-amyloid activities of several catecholamines and other catechol-containing inhibitors (Fig. 2D). The kinetic pro les of ThT uorescence-based amylin amyloid inhibition showed signi cantly stronger inhibition with aged RA versus non-aged RA, with an even more dramatic inhibition activity enhancement was observed with aged norepinephrine (Figs. 2B & 2C). Consistently, enhanced inhibition by norepinephrine occurred only under aerobic conditions (Fig. 2E), with marked reduction in bril formation (Fig. 2F).
Multiple small molecule amyloid inhibitors, many of which are catechol-containing polyphenols, perturb or "remodel" unaggregated and/or pre-aggregated amyloid species into denaturant-resistant aggregates that displayed broad-range molecular weights; characterized as "smear-type" distributions on SDS-PAGE gels (Ehrnhoefer et  Using RA and norepinephrine as two representative cases, treatment with a reducing reagent such as cysteine nearly eliminated their amyloid remodeling activities in a dose dependent manner (Figs. 3A &   3B). Similar effect was observed with glutathione and cystamine as well, and importantly, with other catechol-containing compounds including epigallocatechin gallate (EGCG) and dopamine, but not negative controls phenol and a non-catechol amyloid inhibitor, curcumin (Fig. 3C).
The chemical changes that occur during autoxidation were investigated by both UV-Vis and liquid chromatography-mass spectrometry (LC-MS) approaches. UV-Vis time course spectra con rmed that chemical changes were only detectable under aerobic conditions, and occurred coincidently with their enhanced anti-amyloid activities ( Fig. 4A; highlighted by green asterisks). Such chemical changes were re ected in broad UV absorption spectra changes, particularly in the region of 300-350 nm (Fig. 4A). In an effort to identify the oxidized chemical species that contribute to the enhanced amyloid inhibition, we performed LC-MS analyses of aged norepinephrine. New species/peaks with increasing ion intensities over time (elution was collected at 0-96 hours) were detected in LC at the elution time between 0.85 min -1.0 min (Fig. 4B) (Figs. 2E & 2F). These data agree well with studies that showed catecholamine oxidation products were effective anti-amyloidogenic agents against α-synuclein (Li et al, 2004) and tau (Soeda et al, 2015). Collectively, our data demonstrate that autoxidation is a general pathway enhancing the anti-amyloid activities of catechol-containing compounds (Fig. 4C). In supporting this covalent inhibition model, we demonstrated the covalent conjugate adduct by high resolution mass spectrometry in our speci c investigation on baicalein (Velander et al, 2016). Redox state is therefore a key factor that modulates the activities of a large number of catechol-containing amyloid inhibitors.

Discussion
A variety of covalent mechanisms can readily explain the observed effects of autoxidation for catecholmediated remodeling and the corresponding strong stability of the observed remodeled aggregates. Free radical cycling occurring during catechol autoxidation could directly cycle through nearby interacting amyloid proteins that subsequently lead to protein-protein and/or protein-compound adducts (Meng et al, 2009). Alternatively, remodeled aggregates may also form through covalent interactions between electrophilic o-quinone oxidized byproducts of catechol parent compounds and amyloid protein side chain amines (Zhu et  Nevertheless, caution has been exercised in our work to validate with multiple orthogonal in vitro and cellbased assays. ThT uorescence assays for example, one of the most commonly used assays for screening amyloid inhibitors, could lead to false positives with certain hydroxy avones (Noor et al, 2012). Orthogonal secondary assays, such as TEM analysis, PICUP, and cell-based assays, allowed us to follow up on the hits generated from ThT uorescence-based primary screenings, minimizing such false positives.
Multiple structural studies suggest that there are different conformers and/or molecular polymorphism in multiple amyloidogenic proteins including amylin, Aβ, tau, α-synuclein, and TDP-43 ( . These experimental observations suggest that the accumulation and assembly of various protein aggregates in many protein misfolding diseases is not totally driven by the amino acid sequences of aggregation-prone molecules; instead, they may be governed by the precise cellular and pathological environment of aggregation conditions. It will be interesting to identify different conformers, key environmental factors, protein modi cations, and correlate them with their cytotoxicity. Such structure and function classi cation will greatly facilitate the identi cation of speci c pharmacophores for structure-based inhibitor design in the future.

Conclusions
In summary, we applied a molecular screening platform and identi ed catechol-containing compounds as a broad class of protein aggregation inhibitors against three prototypes of amyloidogenic proteins. We further discovered that the redox state is a key determinant of this broad class of aggregation inhibitor activities as part of a covalent conjugation mechanism. These molecular mechanistic insights not only explain why a large number of catechol-containing natural compounds, often enriched in healthy diet, have anti-neurodegeneration and anti-aging activities, but also could guide the rational design of therapeutic or nutraceutical strategies to target a broad range of increasingly prevalent neurodegenerative and related proteinopathies aging diseases.