Small molecule sequestration of amyloid-β as a drug discovery strategy for Alzheimer’s disease

Disordered proteins are challenging therapeutic targets, and no drug is currently in clinical use that has been shown to modify the properties of their monomeric states. Here, we identify a small molecule capable of binding and sequestering the amyloid-β peptide (Aβ) in its monomeric, soluble state. Our analysis reveals that this compound interacts with Aβ and inhibits both the primary and secondary nucleation pathways in its aggregation process. We characterise this interaction using biophysical experiments and integrative structural ensemble determination methods. We thus observe that this small molecule has the remarkable effect of increasing the conformational entropy of monomeric Aβ while decreasing its hydrophobic surface area. We then show that this small molecule rescues a Caenorhabditis elegans model of Aβ-associated toxicity in a manner consistent with the mechanism of action identified from the in silico and in vitro studies. These results provide an illustration of the strategy of targeting the monomeric states of disordered proteins with small molecules to alter their behaviour for therapeutic purposes.


Introduction
Alzheimer's disease is a fatal neurodegenerative condition that affects over 50 million people worldwide, a number that is predicted to rise to 150 million by 2050 unless methods of prevention or treatment are found, with a cost to the global economy that exceeds one trillion dollars per year 1,2 . Despite over 25 years of intensive research and hundreds of clinical trials, there is still no drug capable of modifying the course of this disease 1,2 .
The aggregation of the amyloid-β peptide (Aβ) in brain tissue is one of the hallmarks of Alzheimer's disease [3][4][5][6][7] . This process involves at least three forms of Aβ: (i) a monomeric state, which is highly disordered, (ii) oligomeric aggregates, which are heterogeneous, transient and cytotoxic, and (iii) fibrillar structures, which are ordered and relatively inert, although they are capable of catalysing the formation of Aβ oligomers 8,9 . More generally, the aggregation of Aβ involves a complex non-linear network of inter-dependent microscopic processes, including: (1) primary nucleation, in which oligomers form from monomeric species, (2) elongation, in which oligomers and fibrils increase in size by monomer addition, (3) secondary nucleation, whereby the surfaces of fibrillar aggregates catalyse the formation of new oligomeric species, and (4) fragmentation, in which fibrils break into smaller pieces, increasing the total number of oligomers and fibrils capable of elongation 10,11 .
Aβ is produced by proteolysis from the transmembrane amyloid precursor protein, and its 42residue form (Aβ42) is the predominant species in deposits characteristically observed in the brains of patients with Alzheimer's disease 6,7,12 . Kinetic analysis shows that, once a critical concentration of Aβ42 fibrils has been formed, secondary nucleation overtakes primary nucleation in becoming the major source of Aβ42 oligomers, as fibril surfaces act as catalytic sites for their formation 8 . The fact that the oligomers appear to be the most toxic species formed during the aggregation process [13][14][15] , however, suggests that therapeutic strategies targeting Aβ aggregation should not primarily aim at inhibiting fibril formation per se, but rather doing so in a manner that specifically reduces the generation of oligomeric species 16 . Complex feedback mechanisms between the different microscopic steps in the aggregation reaction can lead to an increase in the concentration of oligomers even when the formation of fibrils is inhibited, and hence result in an increase in pathogenicity 16 .
Previous studies have suggested that effective strategies for inhibiting Aβ aggregation could be based on targeting fibril surfaces to supress the generation of oligomers, or on the reduction of the toxicity of the oligomers [17][18][19][20][21] . It is unclear, however, whether sequestering Aβ in its soluble state could be an effective drug discovery strategy against Alzheimer's disease. Stabilisation of monomeric Aβ into a β-hairpin conformation with large biomolecules has been previously demonstrated to inhibit aggregation, for example using an affibody protein 22 . However, whether such stabilisation of Aβ in its monomeric form can be achieved via small molecule binding in a drug-like manner is still under debate. While there is research indicating a stabilising effect of small molecules on the soluble state of Aβ, there are contradictory reports of their effects on its aggregation [23][24][25] . It should also be considered that such molecules may not be specific, as for example some appear to bind monomeric Aβ in a manner similar to low concentrations of sodium dodecyl sulphate (SDS) [23][24][25] . Furthermore, it has been proposed that the binding of these small molecules to monomeric Aβ may be mediated by colloidal particles formed by the small molecules 26 , although this observation has also been disputed 23,24,27 . The uncertainty of whether monomeric Aβ is a viable drug target is caused, in part, by a lack of understanding of the molecular properties of monomeric Aβ and how to stabilise this peptide with specific small molecules that have the potential to be developed as drugs.
The complexity of targeting monomeric Aβ is caused, in part, by the fact that Aβ is intrinsically disordered, as it lacks a well-defined structure and instead exists as a heterogeneous ensemble of conformationally distinct states 28 . The dynamic nature of disordered proteins, and the consequent absence of stable and persistent binding pockets, implies that they do not readily lend themselves to conventional mechanisms of drug-binding, such as the well-established lock-andkey paradigm, in which a drug fits snugly into a single, well-defined binding site. [29][30][31] As a result, targeting disordered proteins with small molecules has not been considered a promising drug discovery strategy, and there are no small molecules on the market directly targeting disordered regions despite their high prevalence in disease 2 . A deeper understanding of the possible mechanisms by which small molecules can modify the behaviour of disordered proteins may open new avenues for drug development, not only against Alzheimer's disease and other neurodegenerative disorders but also many other medical conditions involving disordered proteins, including type II diabetes, and certain forms of cancer and cardiovascular disease 28,29 .
10074-G5 has been previously identified to inhibit c-Myc-Max heterodimerization 33 specifically by binding and stabilizing the intrinsically disordered c-Myc monomer 34,35 . Here, we also observe that 10074-G5 binds monomeric Aβ42, a disordered peptide unrelated to c-Myc. As a result of this interaction, 10074-G5 significantly delays both primary and secondary nucleation pathways in Aβ42 aggregation. We characterise this interaction using biophysical experiments and integrative structural ensemble determination techniques, and observe that Aβ42 remains disordered in the bound form with decreased hydrophobicity. Notably, we also observe that the conformational entropy of Aβ42 increases upon interacting with 10074-G5, suggesting that exploiting this phenomenon may be a potential therapeutic strategy for disordered proteins. We further show that this molecule inhibits the pathogenesis associated with Aβ42 aggregation in a Caenorhabditis elegans model of Aβ42-mediated toxicity 36 in a manner consistent with the binding mechanism described in silico and characterised in vitro.

Selection of the system
We selected the compound 10074-G5 as model system to understand whether and how a small molecule inhibits the aggregation of Aβ by binding the monomeric form of this peptide. We used this molecule as it has been reported to bind the oncogenic disordered protein c-Myc in its monomeric form, and it contains a nitrobenzofurazan moiety, which has been previously shown to inhibit the aggregation of Aβ 37 .

Characterisation of the binding of 10074-G5 to monomeric Aβ42
We characterised the binding of 10074-G5 with monomeric Aβ42 using a multidisciplinary approach based on experiments and integrative structural ensemble determination. First, we carried out bio-layer interferometry (BLI, see Materials and Methods) measurements to characterise this interaction in real-time. We immobilised N-terminally biotinylated monomeric Aβ42 on the surface of super streptavidin sensor tips (Materials and Methods) and exposed them to varying concentrations of the small molecule (Figure 1b). We observed a concentrationdependent response, indicative of binding. By globally fitting the binding curves to simple one-step association and dissociation equations, such that the fit constrains all curves to share single association ( on ) and dissociation ( off ) rates, we determined on to be 1.5 x 10 3 ± 0.2 x 10 3 M -1 sec -1 and off to be 3.2 × 10 -2 ± 0.3 × 10 -2 sec -1 , corresponding to a binding dissociation constant ( D ) of 21 µM. This affinity value is comparable to other small molecule interactions with disordered proteins 34 .
We then investigated the binding of 10074-G5 and monomeric Aβ42 at the ensemble-averaged, single residue level. To do so, we performed 2D H N-BEST CON nuclear magnetic resonance This experiment, which relies on heteronuclear direct detection with minimal perturbation of proton polarization, provides a valuable tool to study solvent exposed systems in which amide protons experience fast hydrogen exchange 38 . In particular, the signals of amide nitrogen atoms become attenuated when their directly bound protons are in fast exchange with the solvent. After testing this on a well characterized protein (ubiquitin), we performed this experiment on Aβ42.
In the absence of 10074-G5, we observed that several hydrophobic residues, particularly Cterminal residues are protected from solvent exchange (Leu17, Leu34, Val36, Ile41, Ala42, see I/I0 values as shown in Figure 1c,d). Notably, in the presence of 10074-G5, we observed the quenching of several residues across the sequence of the monomeric Aβ42 peptide that were not quenched in the absence of 10074-G5 (Figure 1c,d), suggesting that some residues have increased solvent exposure in the presence of the small molecule. This change in the solvent exchange profile suggests that 10074-G5 interacts with monomeric Aβ42 in a manner that increases the solubility of at least some of the conformations within the monomeric structural ensemble 29,30 .
To obtain further insight into the thermodynamic properties of this interaction, we quantified the heat changes upon 10074-G5 binding to Aβ40 using isothermal titration calorimetry methods ( Figure S2). In these experiments, we used Aβ40 instead of Aβ42 because of the higher solubility of Aβ40; we have, however, shown that 10074-G5 has similar effects on the aggregation of Aβ40 as on that of Aβ42 ( Figure S3a). The observation of minimal heat changes ( Figure S2) suggests that the interaction of 10074-G5 with monomeric Ab is likely to be entropic, as found for the interactions of another small-molecule with a disordered peptide 40 .
To obtain a structural description of how 10074-G5 affects the disordered structural ensemble of Aβ42, we employed metadynamic metainference, an integrative structural ensemble determination approach 41 (Figure S6a) 44 . As expected, we observed instead no difference between the average number of hydrogen bonds formed by water molecules in the bulk between the apo and holo simulations ( Figure S6b). To determine whether or not this binding could be characterised by the release of water molecules upon association, we calculated the average number of water molecules in the hydration shell and show that this value is similar with and without association ( Figure S6c).
To structurally characterise the unbound and bound conformations of Aβ42 and assess convergence, we performed a clustering analysis (Figures S7, S8). We observed that while 10074-G5 binds the extended form in a non-specific manner, all other structural clusters show localisation of the compound within well-defined pockets of Aβ42 for specific conformations (Figure 2c and S7a) involving hydrophobic, hydrophilic, charged, and polar residues ( Figure   S8). Interestingly, we also observed that the conformational entropy of Aβ42 is increased in the bound form, exhibiting the so-called "entropic expansion" mechanism which may contribute favourably to the binding free energy 29 ( Figure S7c). In stark contrast to the "lock and key" binding mechanism, this observation suggests that the identification of small molecules which increase the conformational entropy of the disordered proteins may be a promising therapeutic strategy.
To probe the energetic contributions to this interaction on an ensemble-averaged, residuespecific level, we analysed Lennard-Jones and Coulomb contributions between 10074-G5 and each residue. We observe strong Lennard-Jones interactions, particularly between aromatic residues (Figure 2e) Tyr10, Phe19 and Phe 20 and 10074-G5. The strongest Coulomb interactions occur at charged residues Lys16 and Lys28 (Figure 2e).

The small molecule 10074-G5 sequesters monomeric Aβ42 and inhibits its aggregation
We measured the kinetics of Aβ42 aggregation at a concentration of 1 µM in the presence and absence of increasing concentrations of 10074-G5. Measurements were performed by means of a fluorescent assay based on the amyloid-specific dye thioflavin-T (ThT), which reports on the overall fibril mass formed during the aggregation process 5,8,16,[45][46][47] . We found that 10074-G5 has a significant effect on Aβ42 aggregation (Figure 3a,b). Specifically, the data show that the final value of the ThT fluorescence, which corresponds to the end point of the aggregation reaction, is dependent on the concentration of the compound (Figure 3a). The observation of a significant decrease in the final ThT intensity could be due to several non-mutually exclusive possibilities including: 1) interference of the ThT signal by 10074-G5, 2) formation of soluble off-pathway aggregates, 3) sequestration of Aβ42 during the aggregation process 18 .
Given the fact that 10074-G5 is a coloured compound, we sought to investigate whether the decrease in the fluorescence intensity of ThT was exclusively due to an interference of 10074-G5 with the dye, or also due to a decrease in the mass of the fibrils formed during the aggregation process. To this end, we performed a ThT-independent dot-blot assay in which we explicitly measured the quantity of soluble Aβ42 over time in the presence and absence of 10074-G5 using the W0-2 antibody, which binds to Aβ (Figure 3c-e). The solubility was determined by measuring the amount of Aβ42 that did not sediment after 1 h of ultracentrifugation at 100,000 rpm. We observed that in the presence of a 20-fold excess of 10074-G5, approximately 40% of the total amount of Aβ42 remained in a soluble form (Figure 3d,e). These experiments indicate that not all Aβ42 monomers are incorporated in ThT-binding fibrils at the end of the aggregation process, and, thus, that the presence of 10074-G5 sequesters Aβ42 in its soluble form.  48 . This experiment reports on early aggregation events that may be invisible to ThT, which is specific for cross-β sheet content as early aggregates such as oligomers or multimers may lack β-sheet structure. 49 We observed, however that in the presence of an equimolar concentration of 10074-G5, the fluorescence intensity remains constant over time (Figure 3g), thereby suggesting that Aβ42 does not self-associate in the presence of 10074-G5.
To further demonstrate that 10074-G5 alters the kinetics of aggregation, we performed 3-D morphological analyses of fibrils using high resolution and phase-controlled 50 atomic force microscopy (AFM) on the time scale of the aggregation process (Figures 3b and S9). Singlemolecule statistical analysis of the aggregates in the 3-D maps shows that Aβ42, both in the absence and presence of 10074-G5, forms non-mature aggregates with average cross-sectional diameters of approximately 2-3 nm, and mature fibrillar aggregates with average diameters of approximately 5-6 nm, as previously observed 51,52 . It has been previously shown that fibrillar species with diameters less than 6 nm lack a cross-b sheet structure fully stabilised by a tight network of intermolecular hydrogen bonding, as compared to mature fibrillar aggregates 51 .
Notably, we observed that at the same time point of aggregation, the fibrillar aggregates formed in the presence of 10074-G5 had smaller cross-sectional diameters than those formed in its absence, with a significantly higher abundance of non-mature species with respect to mature fibrillar species. These data suggest that the process of fibril formation and maturation of crossb sheet structure in the presence of this compound is considerably slower than in its absence 52,53 (Figures 3b and S9b).

10047-G5 does not chemically modify Aβ42
To determine whether or not the binding of 10074-G5 to Aβ42 is covalent or induces other chemical modifications, we performed mass spectrometry on Aβ42 incubated in the presence and absence of 10074-G5. Samples were incubated overnight at 37 °C and then spun down using an ultracentrifuge (Materials and Methods). The supernatant and resuspended pellet of the aggregation reactions were analysed by matrix assisted laser desorption/ionization (MALDI) mass spectrometry ( Figure S10). No mass increase was observed following the incubation with 10074-G5, indicating that its presence does not result in detectable covalent chemical modifications to Aβ42.

10074-G5 inhibits all microscopic steps of Aβ42 aggregation
In order to better understand the mechanism of inhibition of Aβ42 aggregation by 10074-G5, we performed a kinetic analysis on the ThT aggregation traces. Figure 4a shows the ThT kinetic curves normalized relative to the reaction end points. From the normalized data, we observe that 10074-G5 slows down the aggregation reaction in a concentration-dependent manner, consistent with the AFM and dot blot results, confirming a delay in the aggregation process of Aβ42 ( Figure S9). We then used a chemical kinetics approach 54 to determine whether the inhibition data could be explained by a monomer sequestration model, in which 10074-G5 inhibits Ab42 aggregation by binding monomeric Ab42 and, in this manner, reduces the concentration of monomers available for each microscopic step of aggregation (see SI). Specifically, we first fitted the measured aggregation kinetics in the absence of 10074-G5 to a kinetic model of Ab42 aggregation (see SI, Eq. S10) 54 to estimate the values of the unperturbed rates for primary nucleation, elongation, and secondary nucleation. We then formulated a master equation model for inhibited aggregation kinetics in the presence of 10074-G5 (Eq. S11). We derived explicit integrated rate laws describing inhibited kinetics (Eqs. S11-14 and Figure S11), which we used to fit the experimental ThT data in the presence of 10074-G5. For this analysis, we implemented the unperturbed rate constants for aggregation, leaving the value ofas the only fitting parameter. We performed a global fit; all ThT profiles at increasing concentrations of 10074-G5 were not fit individually, but rather using the same choice of -, with the dependence on the concentration of 10074-G5 being captured in the integrated rate law through Eq. S14. The result of this global fit is shown in Figure 4a and yields an affinity value of -= 40 µM, consistent with other affinities reported for small molecule binders of disordered proteins 34 . It is interesting to note that this estimated affinity is considerably weaker than small molecule binders of A key prediction from the monomer sequestration model is that a monomer-interacting compound should interfere with all three microscopic steps of aggregation. In fact, we find that the presence of an inhibitor that binds monomers quickly compared to the overall aggregation rate does not affect the topology of the reaction network. As a result, the inhibited kinetics can be interpreted in terms of effective rates of aggregation that depend on the concentration of inhibitor (Eq. S14). In Figure 4d, we show the values of the effective rates of aggregate proliferation through primary (λ) and secondary (κ) nucleation pathways as a function of the concentration of 10074-G5 predicted by this model (see Eq. S10 for a definition of λ and κ). The monomer sequestration model also predicts that the effective rate of elongation should be reduced (Figure S13b), although to a lesser extent than the nucleation pathways, which have a stronger monomer concentration dependence. To test this prediction, we performed seeded aggregation experiments in the presence of preformed Aβ42 fibrils to obtain independent measurements of the effective elongation rate as a function of 10074-G5 concentration. We observed that 10074-G5 indeed decreases the effective rate of fibril elongation (Figure S13a), consistent with the monomer sequestration mechanism.
We also sought to understand whether 10074-G5 binds monomeric Aβ40 with a comparable affinity to that of Aβ42. To address this question, we applied the monomer sequestration model as used in Figure 4a, to fit the inhibitory effects of 10074-G5 on aggregation of 10 µM of Aβ40 ( Figure S3a). From this analysis, we extracted an affinity constant of 10074-G5 for Aβ40 of 10 µM, a similar value to that obtained for Aβ42 ( Figure S3b). Given the increased toxicity of Aβ42 as compared to Aβ40, we anticipate the optimisation of small molecules more specific for monomeric Aβ42 over Aβ40.

Characterisation of the binding of 10074-G5 to stabilised Aβ40 oligomers
Next, we probed whether 10074-G5 alters the behaviour of oligomeric species of Aβ. Although it is extremely challenging to determine whether 10074-G5 modifies the oligomeric intermediates of Aβ42 formed on-pathway to aggregation, which are transient, heterogenous species, it is possible to carry out this analysis more readily on oligomers of Aβ40 stabilised using Zn 2+ 56 . Thus, we next considered whether or not 10074-G5 can alter the behaviour of these stabilised, pre-formed oligomeric species. We incubated pre-formed oligomers in the presence of 10074-G5, centrifuged the samples, and measured the quantities of Aβ40 in the pellet and in the supernatant by using SDS-polyacrylamide gel electrophoresis (SDS-PAGE, Figure S14a).
The results indicate that these pre-formed oligomers did not dissociate in the presence of 10074-G5. Furthermore, 10074-G5 was found not to alter the turbidity of solutions in which they were present as measured by their absorbance profiles (Figure S14b) suggesting that 10074-G5 does not cause such species to change detectably in size. Lastly, dot blots of pre-formed oligomeric samples in the presence and absence of the compound using the OC-antibody, which binds to βsheets 57 , show that the oligomers maintain their characteristic conformations (Figure S14c). Due

13
to the coloured nature of 10074-G5, it was neither possible to characterise the oligomers in the presence of the compound with dynamic light scattering nor analytical ultracentrifugation measurements. Taken together, these data suggest that 10074-G5 does not disaggregate the preformed oligomeric species or cause them to undergo further assembly. Nevertheless, it remains possible that this compound affects the evolution of oligomer populations formed during the aggregation reaction, potentially inhibiting their conversion into fibril-competent species.

10074-G5 inhibits Aβ42 aggregation in a C. elegans model of Alzheimer's disease
To determine if 10074-G5 can inhibit the formation of Aβ42 aggregates in vivo, we tested its effects using a C. elegans model of Aβ42-related toxicity (GMC101, Figures 5, S15, S16), in which age-progressive paralysis was induced by overexpression of Aβ42 in the body wall muscle cells 36 . The N2 wild-type strain 58 was used as a control.
10074-G5 was administered to worms from larval stage L4, and then continuously throughout their lifespan (see Materials and Methods and Figure 5a). First, we probed the quantity of the aggregates in the animals by means of an amyloid specific fluorescence probe, NIAD-4 20 (Figure 5b,c). The results show that the administration of 10074-G5 resulted in a markedly lower aggregate load. We also monitored a number of phenotypic readouts of worm health, including body bends per minute, the extent of the bending motion, the speed of movement, and also the rate of paralysis. We found that 10074-G5 improved all of these characteristic behavioural parameters in worms expressing Aβ42 in a dose-dependent manner when compared to the untreated worms (Figure 5d,e).  (Figures 3 and 4). Furthermore, these results suggest that the combination of in silico, in vitro, and in vivo drug discovery methods holds promise for the identification of novel small molecules which inhibit the toxic behaviour of disease-related disordered proteins.

Conclusions
We have characterised the binding of the small molecule 10074-G5 to monomeric Ab42 using a combination of experimental approaches and integrative structural ensemble determination methods. The real-time, dose-dependent responses that we have observed in the BLI experiments demonstrate that 10074-G5 binds Aβ42 in its monomeric form (Figure 1b). The NMR experiments and the metadynamic metainference simulations have illustrated that this binding is distinct from most small molecule interactions with structured proteins. In particular, we have observed that 10074-G5 does not bind to a single binding site, but rather binds transiently to many different sites (Figures 2, S7). We have also found 10074-G5 induces Aβ42 to adopt many different conformations, keeping it disordered and more solvent-exposed ( Lastly, we show that 10074-G5 is highly effective at reducing the Aβ42 aggregate load and its associated toxicity in a C. elegans model of Alzheimer's disease.
Collectively, these results indicate the importance of developing a more detailed understanding of the interactions between disordered proteins and small molecules, which in turn could lead to the development of new therapeutic approaches for the many diseases in which such disordered proteins are involved.

Data Availability Statement
The data and code that support the findings of this study are available from the corresponding author upon request.

Metadynamic metainference simulations.
To generate the structural ensembles, we employed an integrative approach that incorporates NMR chemical shift data into molecular dynamics simulations. To this end, we used metadynamic metainference, which compensates for the inaccuracies of the force field, accounts for errors in experimental data, and enhances sampling. 41,42 All-atom metadynamic metainference simulations 41  Production runs were executed in the NPT ensemble at 278 K using the Parrinello-Rahman barostat 68 . A time step of 2 fs was used together with LINCS constraints on all bonds 69 . The van der Waals interactions were cut off at 1.2 nm, and the particle-mesh Ewald method was used for electrostatic interactions 70 . Bound simulations were performed as described above, using the starting structures obtained from the NVT equilibration at 600 K. The 10074-G5 molecule was added to a corner of the box and the system re-solvated with 11734 water molecules. The system was then minimized and equilibrated using the procedures described above. Preliminary parameters for 10074-G5 were taken from the CGenFF software 71,72 , and those with any penalty were explicitly re-parameterised using the Force Field Toolkit 73 and Gaussian 09 74 (see SI and Dot-blot assay. Blotting was performed using the Ab42 sequence-specific antibody (W0-2, MABN10, Millipore, Burlington, USA). Samples were removed from a solution containing 2 µM Ab42 in the presence and absence of three-and ten-fold equivalence of 10074-G5. To ensure only the monomer was placed on the blots, samples were spun down using an ultracentrifuge at 100,000 rpm for 1 h at 25 °C using a TLA100 rotor. 2 µL of the supernatant were pipetted onto a nitrocellulose membrane (0.2 µM; Whatman). After drying, the membrane was blocked with 5% (w/v) bovine serum albumin (BSA) in phosphate-buffered saline (8 mM Na2HPO4, 15 mM 20 KH2PO4, 137 mM NaCl, 3 mM KCl, pH 7.4, PBS) overnight at 5 °C, followed by three 15 min washes with PBS at room temperature. The membrane was then immunised with a 1/1000 dilution of WO-2 anti-Ab antibody in PBS with 5% BSA overnight at 5 °C, followed by three 15 min washes with PBS at room temperature. The membrane was then incubated for 2 h at room temperature in PBS supplemented with 0.05% tween and an anti-mouse-Alexa Fluor 594 secondary antibody conjugate (ThermoFisher Scientific, Waltham, USA) at room temperature, and then washed three times with PBS supplemented with 0.05% tween. Fluorescence detection was performed using Typhoon Trio Imager (GE Healthcare, Chicago, IL, USA). Blots were quantified using ImageJ. Data were fit to a competitive binding equilibrium model between free monomers and fibrils (Eq. S13). In this model, monomers are either free, aggregated (i.e. part of a fibril), or bound to 10074-G5; the binding of the compound to the monomer is described by a single binding free energy. The binding of monomers to fibril ends is stronger compared to the binding of monomers to the inhibitor. The concentration of free monomer in equilibrium with amyloid fibrils (critical concentration) measured in our experiments is 89:;:8<= = 93 nM, consistent with other reports. 79 The equilibrium concentration of unreacted soluble monomer after ultracentrifugation measured at varying inhibitor concentration is fit to Eq. S13 withas a fitting parameter. This procedure yields -= 7 ± 1 µM, as shown in Figure 3e. Aliquots of NGM media containing FUDR (75 µM) were autoclaved, poured, seeded with 350 µL OP50 culture, and grown overnight. After incubating for up to 3 days at room temperature, 2.2 mL aliquots of 10074-G5 dissolved in water at different concentrations were spotted atop the NGM plates. The plates were then placed in a sterile laminar flow hood at room temperature to dry. For the final experiments, worms were transferred onto the 10074-G5-seeded plates directly at larval stage L4 and they were exposed to 10075-G5 for the whole duration of the experiment.
To ensure that the presence of 10074-G5 did not affect the OP50 E. coli consumed by the C. elegans, we performed a growth assay of the E. coli directly from the NGM plates in the presence of 10074-G5 or DMSO after one day of incubation at 24 °C (Figure S15). E. coli from the NGM plates were added to 4 mL of LB media and diluted to an optical density of 0.25. Then, 3 mL of this starter culture was added to 40 mL of sterile LB media, which was incubated at 37 °C and shaking at 180 RPM. Optical density measurements were collected every 30 minutes and the experiment was performed in duplicate.
All C. elegans populations were cultured at 20 °C and developmentally synchronized from a 4 h egg-lay. At 64-72 h post egg-lay (time zero) individuals were transferred to FUDR plates, cultured at 24°C to stimulate aggregation, and body movements were assessed over the times indicated. At different ages, the animals were washed off the agar plates with M9 buffer and spread over an OP50 un-seeded 9 cm plate. The swimming worms were visualized by using a high-performance imaging lens and a machine vision camera, after which their movements were recorded at a high number of frames per second (fps) for 30 s or 1 min 80,81 . Body bends were then quantified using a tracking algorithm 81,82 . Briefly, after an initial background subtraction, a second (nonadaptive) thresholding procedure was performed and worms were identified and labelled. The eccentricity, a measure of the ratio of the minor and major ellipse axes, of each tracked worm was then used to estimate the worm bending as a function of time 81,82 . The total health was calculated by summing the mobility, speed, bend measure, and viability of the worms 81,82 . Total health values were normalized using the values of the control worms. At least 150 animals were examined per condition, unless stated otherwise. All experiments were carried out in triplicate and the data from one representative experiment are shown in Figure 5. Control experiments to test the effects of 10074-G5 on the movement of N2 wild type C. elegans are shown in Figure S16. Two-tailed Student's t tests (unpaired) were used to calculate P values.
Statistical analysis was performed using the GraphPad Prism 6 software.
To stain the aggregates within the C. elegans, live transgenic animals were incubated with 1 µM NIAD-4 (0.1% DMSO in M9 buffer) for 4 h at room temperature 20 . After staining, animals were allowed to recover on NGM plates for about 24 h to allow destaining via normal metabolism.
Stained animals were mounted on 2% agarose pads containing 40 mM NaN3 as an anesthetic on glass microscope slides for imaging. Images were captured with a Zeiss Axio Observer D1 fluorescence microscope (Carl Zeiss Microscopy GmbH, Jena, Germany) with a 20 × objective and a 49004 ET-CY3/TRITC filter (Chroma Technology Corp, Bellows Falls, USA).
Fluorescence intensity was calculated using ImageJ software (National Institutes of Health) and then normalized as the corrected total fluorescence 20,83 . Only the head region was considered because of the high background signal in the intestinal regions. At least 25 animals were examined per condition, unless stated otherwise. All experiments were carried out in triplicate and the data from one representative experiment are shown in Figure 5. Two-tailed Student's t tests (unpaired) were used to calculate P values. Statistical analysis was performed using the GraphPad Prism 6 software. Quantification of the relative / ? intensities from (c) shows that the peptide amide groups are more exposed to solvent in the presence of 10074-G5. Arrows highlight regions along the sequence in which signals are detectable in the absence of the compound, but not in its presence, thus suggesting that 10074-G5 increases the solvent exposure of specific regions of Aβ42.   Global fit of normalized ThT kinetic curves to a monomer sequestration model (Eq. S11), in which 10074-G5 affects the aggregation by binding free monomers. Error bars represent ± one standard deviation. Measurements were taken in quintuplicate. The theoretical curves are obtained using Eq. S10 with unperturbed kinetic obtained from (b) leavingas the only global fitting parameter (Eq. S14). The global fit yields -= 40 µM. (b) Global fit to Eq. S10 of ThT kinetic traces of the aggregation reaction for increasing concentrations of Aβ42 (1, 1.  In all panels, error bars represent ± standard error of the mean (SEM). The symbols *** and **** indicate P < 0.001 and 0.0001, respectively by two-tailed Student's t-test. 10074-G5 shows minimal movement effects on wild type C. elegans (Figure S16).