Decoding the Roles of Amyloid-β (1–42)’s Key Oligomerization Domains toward Designing Epitope-Specific Aggregation Inhibitors

Fibrillar amyloid aggregates are the pathological hallmarks of multiple neurodegenerative diseases. The amyloid-β (1–42) protein, in particular, is a major component of senile plaques in the brains of patients with Alzheimer’s disease and a primary target for disease treatment. Determining the essential domains of amyloid-β (1–42) that facilitate its oligomerization is critical for the development of aggregation inhibitors as potential therapeutic agents. In this study, we identified three key hydrophobic sites (17LVF19, 32IGL34, and 41IA42) on amyloid-β (1–42) and investigated their involvement in the self-assembly process of the protein. Based on these findings, we designed candidate inhibitor peptides of amyloid-β (1–42) aggregation. Using the designed peptides, we characterized the roles of the three hydrophobic regions during amyloid-β (1–42) fibrillar aggregation and monitored the consequent effects on its aggregation property and structural conversion. Furthermore, we used an amyloid-β (1–42) double point mutant (I41N/A42N) to examine the interactions between the two C-terminal end residues with the two hydrophobic regions and their roles in amyloid self-assembly. Our results indicate that interchain interactions in the central hydrophobic region (17LVF19) of amyloid-β (1–42) are important for fibrillar aggregation, and its interaction with other domains is associated with the accessibility of the central hydrophobic region for initiating the oligomerization process. Our study provides mechanistic insights into the self-assembly of amyloid-β (1–42) and highlights key structural domains that facilitate this process. Our results can be further applied toward improving the rational design of candidate amyloid-β (1–42) aggregation inhibitors.


■ INTRODUCTION
Pathogenic amyloid aggregation of intrinsically disordered proteins (IDPs) is related to multiple neurodegenerative diseases. 1 The pathological hallmark of Alzheimer's disease (AD)�which is the most prevalent type of dementia�is the accumulation of amyloid-β (Aβ) and microtubule-associated protein (MAP) tau. 2,3 Fibrillar aggregation of Aβ leads to the formation of senile plaques in the brains of AD patients. 4 Numerous reports, including genome-wide association studies, imply that the appearance of amyloid aggregates is closely related to neurodegeneration. 5 Nevertheless, the exact pathophysiological role of the Aβ protein remains unclear.
The Aβ proteins are formed by enzymatic cleavage of the amyloid precursor protein (APP) and commonly have 40 or 42 residues (Aβ40 and Aβ42, respectively). 6 The longer Aβ42 tends to favor hydrophobic interaction-mediated self-assembly more so in comparison with Aβ40. 7 The Aβ proteins form various oligomeric species, and each species facilitates the subsequent assembly of several intermediates, eventually culminating in the formation of mature fibrils. 8 Although multiple studies have presented evidence that the Aβ protein promotes neuronal cell death and hyperphosphorylation of MAP tau, the molecular mechanism of amyloid aggregation is yet to be clearly elucidated owing to the highly heterogeneous nature of the pathological cascade. 9,10 To date, antibodies and small molecules that selectively bind to aggregates of Aβ fibrils have been introduced as potential therapeutic agents for AD. 11 In addition, a number of drug candidates target the Aβ-related pathway, which regulates the production or existence of Aβ and modulates the ratio of Aβ proteins. 12 Although many potential drug candidates against pathogenic amyloid aggregates have been proposed for AD treatment, their effectiveness remains debatable. Therefore, it is necessary to develop a more potent therapeutic strategy for AD based on a more in-depth understanding of the fibrillar amyloid aggregation process at a molecular level.
Recent studies have attempted to rationally design candidate drugs in order to improve existing therapeutic approaches or suppress the fibrillar amyloid aggregation process altogether 13−16 based on the accumulated knowledge of interlinked dynamics of IDPs, intermediate oligomers, and their pathogenesis to date. 17−20 For example, a pharmacophore in MAP tau for the binding of small molecules was shown to enhance the binding of potential therapeutic agents. 15 In addition, an epitope-specific method for antibody design has been reported, which allows development of antibodies to the desired binding epitopes of amyloidogenic IDPs. 16 Since the self-assembly property of Aβ can be modulated by interrupting or enhancing the contact of the hydrophobic core domains, it would be possible to target the hydrophobic core regions in order to prevent or reduce the capacity of Aβ for selfassembling. Based on such findings, as well as in silico and in vitro analyses, we have previously shown that rationally designed point mutants of Aβ42 can suppress fibrillar aggregation. 14 In particular, we identified two hydrophobic residues (Phe19 and Ile32) which are crucial for facilitating oligomerization of Aβ42. 14 The designed point mutants against those sites cross-interacted with Aβ42 in early-stage amyloid aggregation and interfered with the Aβ42 self-assembly. 14 In this study, we aimed to identify the target region responsible for disrupting fibrillar amyloid aggregation of Aβ42 and further delineate its role in this process. We employed a machine learning (ML) algorithm to predict the binding domain of prefibrillar Aβ42 with various fragments of Aβ42. 21 Next, we selected two Aβ42 fragments [ 16 KLVFFAE 22 , referred to as fragment a (F a ), and 25 GSNKGAIIGLM 35 , referred to as fragment b (F b ), respectively], which displayed self-assembly properties, while also containing one of the two crucial hydrophobic residues (Phe19 and Ile32). Moreover, we designed candidate inhibitor peptides against fibrillar amyloid aggregation of Aβ42 using the two selected fragments and applied a point mutation strategy to eliminate their selfassembly properties. We also confirmed the suppressive effects of the inhibitor candidates on amyloid cluster formation and the alleviation of Aβ42-dependent cytotoxicity in vitro using an SH-SY-5Y cell line. In addition, we examined the disassembly of preformed Aβ42 fibrils upon co-incubation with the inhibitor candidates. To understand the molecular mechanisms underlying the suppression of Aβ42 aggregation, we investigated the interference of inhibitor candidates in early-stage aggregation using an interdisciplinary biophysical approach. Hydrogen−deuterium exchange mass spectrometry (HDX-MS) and ion mobility mass spectrometry (IM-MS) provided evidence of the structural conversion of Aβ42 by designed inhibitor candidates. Combining ML algorithm predictions of peptide-inhibitor complex structures with comprehensive molecular dynamics (MD) simulations provided the molecular basis for peptide-inhibitor interactions. The in silico results and the electron transfer dissociation mass spectrometry (ETD-MS) analysis revealed the binding sites of the inhibitor candidates on Aβ42. We also performed small-angle X-ray scattering (SAXS)-based modeling of the Aβ42 ensembles to understand the solution-phase structural dynamics of Aβ42. Finally, in order to delineate the role of the Aβ42 C-terminal end in amyloid aggregation, we utilized a mutant isoform of Aβ42, I41N/A42N, in which the two C-terminal hydrophobic residues were substituted with polar amino acids.

Predictions of Complex Structures
To predict the complex structures of Aβ42 and its fragments or peptide inhibitor candidates, we employed the AlphaFold multimer (version 2.1.1). 21 Source code is available at https://github.com/ deepmind/alphafold. Structural images were created using UCSF Chimera (version 1.16). 22

ThT Fluorescence Assay
Thioflavin T (ThT) assays were used to track the amyloid aggregation of Aβ42 proteins in the presence or absence of the designed peptides. The concentration of Aβ42 was set at 2 and 5 μM. ThT was incubated at 37°C without agitation in 20 mM Tris−HCl buffer (pH 7.4) in SPL 96-well black microplates (SPL Life Science, Seoul, Korea) with sealing (EASYseal sealing film, Greiner-Bio-One). The designed peptide inhibitor candidates were applied in 1:1, 1:5, or 1:25 molar ratios. Fluorescence intensity was measured continuously at 1 h intervals, for up to 48 h using a Synergy H1 microplate reader (BioTek, Winooski, VT, USA) with excitation and emission wavelengths of 446 and 482 nm, respectively, from the top of the plate. The ThT assays for the I41N/A42N mutant were performed under the same experimental conditions, except for the Aβ42 I41N/ A42N concentration (10 μM).

Transmission Electron Microscopy
Transmission electron microscopy (TEM) images of the Aβ42 fibrils were obtained using negative staining protocols (Supporting Information). Briefly, incubated fibril samples (5 μL) were spotted on a Cu(II) grid for 3 min at 20°C and removed. The grids were washed twice with HPLC-grade water immediately after sample removal. Then, each sample was stained with 5 μL of uranyl acetate solution (0.5% w/v) for 1 min. The samples were dried for 4 h at 20°C after removal of the uranyl acetate solution. The TEM images were recorded using a JEM-F200 (TFEG; JEOL Ltd., Japan) fieldemission transmission electron microscope [National Center for Inter-University Research Facilities (NCIRF), Seoul National University, Seoul, Republic of Korea] at various magnifications (200 kV; 6000×, 15,000×, 30,000×, and 60,000× magnification).

Cell Viability Test
MTT assay was used to determine cell viability. In each well of a 96well plate, SH-SY-5Y cells (15,000 cells) were seeded and incubated for 24 h. Cells were treated with preincubated fibrils in DMEM for 48 h. The MTT solution (5 mg/mL) was added to the medium and incubated for 3 h at 37°C to form blue MTT-formazan products which were assessed by measuring absorbance at 540 nm using a microplate reader. To ensure reproducibility, each set was tested in triplicate, and the assay was repeated three times.

Disassembly of Preformed Fibrils
Aβ42 (10 μM) was incubated for 24 h at 37°C without agitation in 20 mM Tris−HCl buffer (pH 7.4). Fibrillar amyloid aggregation of Aβ42 proteins (10 μM) was monitored using a ThT fluorescence assay. The peptide inhibitor candidates were applied to the preformed fibrils of Aβ42 (diluted to a concentration of 2 μM), and the final peptide inhibitor candidate concentration was 50 μM. Except for the addition of the peptide inhibitor candidate, the control fibrils from the same preformed fibrils of Aβ42 were treated identically. All the samples were incubated at 37°C without agitation for 24 h. Disassembly of preformed fibrils was monitored by endpoint ThT fluorescence intensity and TEM image analysis at various magnifications (6000×, 15,000×, 30,000×, and 60,000× magnification). The density of the fibrillar aggregates was calculated using the area fraction covered by fibrils in the TEM images. TEM images were further analyzed using the ImageJ software to determine fibril density. Six images at 6000× magnification were processed using MATLAB (version R2020a). The adaptthresh and wiener2 functions were used to implement the adaptive thresholding and noise reduction steps, respectively.

Small-Angle X-ray Scattering
Solution SAXS experiments were carried out at the 4C SAXS II beamline at the Pohang Accelerator Laboratory (Pohang, Republic of Korea). 23 All SAXS experiments used freshly prepared samples. Ensemble structures were modeled using EOM analysis from the ATSAS software package (version 3.0.3). 24 External pools were generated by REMD simulations and used when running a genetic algorithm to optimize ensembles (GAJOE). The ensemble size was set at 50 curves per ensemble, and repetitions were not permitted. The experimental details are described in the Supporting Information.

MD Simulations
MD simulations and umbrella sampling were performed using the GROMACS software package (version 2020.4). 25 To estimate the interchain contact probability between the Aβ42 and each peptide inhibitor candidate, a REMD simulation (33 replicas, 150 ns) was performed for each complex system using the CHARMM36m force field and the TIP3P solvent model. 26,27 The binding free energy was calculated using umbrella sampling combined with a pulling simulation. This was performed using the same force field and solvent model as described above. The 600 ps pulling simulation trajectories were divided into 30 windows, and the MD simulation was run for 10 ns in each window. External generation for EOM analysis was performed using the GROMACS software package (version 4.5.5) with the CHARMM36m force field and the General Born implicit solvation model, where five replicas were simulated for 20 ns. 25,26,28 The DSSP program was used to calculate the solventaccessible surface area (SASA) per residue as well as the total hydrophobic or hydrophilic surface area (gmx do_dssp module).

ESI-MS Combined with the HDX Technique
Freshly prepared Aβ42 and peptide inhibitor candidates were deuterated with D 2 O before quenching with formic acid. The final concentrations of Aβ42 and the peptide inhibitor candidates were 10 and 50 μM, respectively. The samples were immediately injected into a Synapt G2-Si HDMS quadrupole time-of-flight (Q-TOF) mass spectrometer (Waters, Manchester, UK) after the quenching agent was added. The backward-exchanged hydrogens were corrected with D 2 O-hydrated fully deuterated Aβ42.

Ion Mobility Mass Spectrometry
IM-MS was performed in the positive ion mode on a Synapt G2-Si HDMS Q-TOF mass spectrometer (Waters, UK, Manchester). The final concentration of the Aβ42 protein was 10 μM in 20 mM ammonium acetate (pH 6.8). The designed inhibitors were applied at a concentration of 50 μM, and the proteins were transferred to the gas phase using an electrospray ionization (ESI) source. The ESI parameters and detailed descriptions for calibrating the collision cross-section are given in the Supporting Information.

Electron Transfer Dissociation Mass Spectrometry
ETD was performed using a Synapt G2-Si HDMS Q-TOF mass spectrometer with ETD functionality (Waters, UK; Manchester). Aβ42 and each peptide inhibitor candidate were mixed at concentrations of 10 and 50 μM, respectively. The samples were introduced at a flow rate of 10 μL/min using an ESI source. To generate anions, the ETD reagent 1,4-dicyanobenzene was introduced into the glow-discharge source. Anions were refilled in the trap cell for 0.2 s at a time interval of 1 s. The ETD mode parameters were a 0.2 V trap wave height, 14 mL/min trap gas flow, and 475 V ETD RF trap. Peak assignment was performed manually by comparing it to a fragment list generated by the ProteinProspector web server (https:// prospector.ucsf.edu/prospector/mshome.htm). To compare the MD simulation and ETD-MS results, an interchain contact probability map of the complexes was projected onto the x-axis and overlaid on the ETD-MS results.

Design of Amyloid Aggregation Inhibitors
Recent crystallography studies have revealed that the hydrophobic domains of Aβ42 form a hydrophobic core in the fibril structure of Aβ42. 29 Prompted by previous findings, we designed novel point mutants of Aβ42 to suppress its selfassembly property. 14 Among the four selected residues (Phe19, Ile32, Leu34, and Ala42), our results showed that Phe19 and Ile32 were essential for fibrillar amyloid aggregation as amino acid substitutions in these positions effectively disrupted the process. 14 Therefore, we hypothesized that the region including one of the two crucial residues plays a key role in suppressing fibrillar amyloid aggregation of Aβ42. The previously reported mutant D1 (F19N/I32N) cross-interacts with wild-type Aβ42 during early-stage oligomerization and as such interferes with fibrillar amyloid aggregation of Aβ42 ( Figure 1a). 14 Herein, we aimed to identify the key domain which can effectively inhibit Aβ42 aggregation and to target that region by using smaller Aβ42 fragments to modulate the self-assembly function of Aβ42.
Our criteria were that smaller Aβ42 fragments must have self-assembly properties and include at least one of the essential residues (Phe19 or Ile32). To systematically search for binding domains, we compiled previously reported small fragments of Aβ42 (F a to F q , 17 fragments in total) and employed AlphaFold to predict the binding of each fragment to prefibrillar Aβ42 (Figures 1b and S1 and Table S1). 30−39 The Aβ42 tetramer is predicted to stack in parallel, and the fragment peptide KLVFFAE (F a ) is expected to bind to its cognate region ( 16 KLVFFAE 22 ) in a parallel conformation. The peptide fragment GSNKGAIIGLM (F b ) also associates with the central hydrophobic region of Aβ42. However, most of the N-terminal fragments of Aβ42 were expected to bind to the Cterminal end of Aβ42 ( Figure S1, F h −F q ). A short fragment, such as IIGLM, also bound to the C-terminal end of Aβ42 ( Figure S1, F g ). Several of the fragments we examined reportedly demonstrated self-assembly properties, and as such, they were expected to bind to the central hydrophobic region. 30,31 Therefore, we posited that targeting the central hydrophobic region using fragment-based inhibitor candidates could be effective in disrupting the amyloid aggregation process.
Based on the predictions from AlphaFold, we selected two short, aggregate-prone peptides, F a and F b , along with their point mutants, to target each domain of Aβ42 with their corresponding sequences while monitoring their effectiveness in inhibiting Aβ42 amyloid aggregation (Figure 1a). The peptides F a and F b both include the mutation target sites mentioned above and are also specific for the central and Cterminal regions, respectively. These two peptides have been reported to form insoluble amyloid bodies in vitro and are widely used as model systems in amyloid aggregation studies. 30,31 We monitored the cluster formation of the two fragments using a ThT fluorescence assay, whose fluorescence intensity increases upon binding to amyloid fibrils, and found that both fragments have self-assembly capabilities ( Figure S2). As the intermolecular contacts are mainly initiated by hydrophobic interactions, we expected that each fragment with self-assembly properties would still be able to crossinteract with Aβ42 via the corresponding domain.
We chose KLVNFAEDVGSNKGAINGLM (P ab ) as our first experimental inhibitor candidate peptide because it spans both the 16 KLVFFAE 22 (F a ) and 25 GSNKGAIIGLM 35 (F b ) regions and contains both F19N and I32N point mutations ( Figure  1a). The suppressive effect of the designed peptide inhibitor candidate P ab on Aβ42 amyloid aggregation was monitored by a ThT incubation assay and TEM image analysis after negative staining (Figure 1c,d). The Aβ42 sample showed an increase in the ThT fluorescence intensity as a function of time. The mixture of Aβ42 and P ab at equimolar concentrations displayed lower ThT fluorescence intensity than the Aβ42 sample at plateau intensity. In addition, the peptide inhibitor candidate P ab suppressed fibrillar amyloid aggregation in a concentrationdependent manner (Figure 1c). The TEM image analysis supports our ThT assay results, showing fewer amyloid fibrils in the mixture of Aβ42 and P ab (Figure 1d).

Effects of Fragments and Designed Peptides on Fibrillar Amyloid Aggregation of Aβ42
To understand the specific effects of different regions on fibrillar amyloid aggregation, we further examined how the aggregation of Aβ42 is affected by two smaller peptide inhibitor candidates, KLVNFAE (P a ) and GSNKGAINGLM (P b ), which span the central and C-terminal hydrophobic regions, respectively. While both P a and P ab suppressed Aβ42 aggregation to an extent, P ab was more effective in this regard than P a (Figures 1c and 2a,c). Interestingly, the inhibitor candidate P b did not suppress Aβ42 aggregation (Figure 2b,c). The TEM image analysis agreed with the ThT assay results, suggesting an inhibitory effect of P a on Aβ42 aggregation, while no considerable effect was observed for P b . Notably, both P ab and P a , which contain the F19N point mutation, effectively suppressed Aβ42 amyloid aggregation (Figures 1d and 2d). We further investigated the inhibitory effect of designed peptides on Aβ42 aggregation at higher Aβ42 concentration (25 μM) since distinct forms of the Aβ oligomer have been reported above the critical concentration (20−25 μM). 40,41 The results show that even though the Aβ42 concentration increased, the suppressive effect of the designed peptides persisted ( Figure  S6).
To monitor the cellular responses after addition of the designed inhibitor candidates, we performed the cytotoxicity test on the SH-SY-5Y cells using the modified thiazolyl blue tetrazolium bromide (MTT) (Figure 2e). We incubated the pure Aβ42 sample, as well as mixtures of Aβ42 and the designed peptides for 48 h at 37°C, and applied these to SH-SY-5Y cells. The results showed that the cytotoxicity of Aβ42 decreased significantly as the concentration of P ab increased. The inhibitor candidate P a also alleviated the cytotoxicity of Aβ42 in a concentration-dependent manner. In contrast, P b did not significantly diminish the cytotoxicity of Aβ42. These findings indicated that suppression of Aβ42 aggregation by the designed peptide inhibitors, P ab and P a , also reduced the toxicity of Aβ42 (Figures 1c and 2a,e).
Moreover, we assessed the disassembly of amyloid Aβ42 clusters using these three peptides. After a 24 h incubation period, we applied the designed peptides to the preformed Aβ42 fibrils and allowed these to act for a further 24 h at 37°C . The TEM image analysis revealed that the designed peptides affected the density of the fibrillar aggregates ( Figure  2f,g). The area fraction covered with fibrils in the TEM images was used to calculate the density of fibrillar aggregates. The density of fibrillar aggregates in the TEM images of the Aβ42 sample was 3.45%, and this decreased to 1.48, 2.82, and 3.01% upon addition of P ab , P a , and P b , respectively (Figure 2f). The endpoint ThT fluorescence intensity of each sample was monitored to quantify ThT-positive aggregates. Compared to the Aβ42 sample treated with incubation buffer (as a negative control), the ThT fluorescence intensity in each of the Aβ42inhibitor mixture samples treated with either P ab , P a , and P b decreased to 68, 81, and 92%, respectively (Figure 2h). The designed peptide P ab , which covered both the F a and F b regions in Aβ42 and included F19N and I32N point mutations, induced the disassembly of preformed amyloid aggregates of Aβ42. The P a designed peptide, which covers the F a region in Aβ42 and includes an F19N point mutation, seemed to have similar trends to P ab . However, P b , the sequence of which corresponds to that of F b and contains an I32N point mutation, did not show notable disassembly of the preformed amyloid aggregates of Aβ42. The MTT assay showed alleviated

JACS Au
pubs.acs.org/jacsau Article cytotoxicity of preformed Aβ42 fibrils after inhibitor-induced disassembly, and the results were in good agreement with our fibril disassembly assays (Figure 2i).

Identification of the Three Key Domains of Aβ42
To understand the different roles of the three regions (central hydrophobic region, 17 LVF 19 , Φ 1 ; C-terminal hydrophobic region, 32 IGL 34 , Φ 2 ; and C-terminal end, 41 IA 42 , Φ 3 ) ( Figure  S7), we investigated the solution-phase dynamics of Aβ42 using SAXS experiments. We performed an ensemble optimization method (EOM) analysis since the Kratky plot showed that Aβ42 has a highly unstructured conformation (Figure 3a), and the theoretical analysis was well fitted to the experimental curve (R g,theo = 22.40 ± 3.19 Å and R g, exp = 22.81 ± 1.25 Å, respectively) ( Figure 3b). Fifty ensemble structures of Aβ42 were obtained from EOM analysis ( Figure 3c) and subjected to calculations for the intramolecular contacts of each structure. The resulting intramolecular distances were plotted as a contact probability map of Aβ42 to better understand the transient interactions in the monomeric state of Aβ42 (Figure 3d). We observed that the Φ 2 Aβ42 site interacts with both Φ 1 and Φ 3 on the same molecule. The interaction between Φ 2 and Φ 3 should be avoided for the successful association of Φ 1 with Φ 2 . The latter interaction shields the Φ 1 domain by positioning it inward in the Aβ42 structure. Based on these observations, we speculate that the structural fluctuations of Aβ42 originate from intramolecular hydrophobic interactions in the C-terminal regions. We separated the modeled ensembles into three groups according to the radius of gyration (compact conformers were assigned to group 1, intermediates to group 2, and extended conformers to group 3) to deconvolute the structural basis of the Aβ42 fluctuations ( Figure S8). We assessed the exposure of each domain (5-residue windows) based on SASA (Figure 3e). These results suggest that Φ 2 is mostly located in the interior sites of the protein; thus, Φ 1 is likely exposed to solvent molecules and other Aβ42 peptides. The total solventaccessible hydrophobic surface area decreased in proportion to the radius of the Aβ42 gyration (Figure 3f). The sum of the solvent-accessible hydrophilic surface area also decreased in group 1, which suggests that compact conformations of Aβ42 lead to the shielding of both hydrophobic and hydrophilic surface areas (Figure 3f,g). On the other hand, the intermediate conformers of Aβ42 did not induce shielding of the hydrophilic surface area, which indicates that the structural fluctuations of Aβ42 conformers are mainly driven by intramolecular contacts of the hydrophobic regions ( Figure  3g). Investigation of the solution-phase dynamics of Aβ42 revealed the interactions between its three key domains (Φ 1 , Φ 2 , and Φ 3 ). These observations showcase that the deviation in the solvent-accessible area between groups 2 and 3 mainly arises from two intramolecular hydrophobic interactions: between Φ 1 and Φ 2 or between Φ 2 and Φ 3 (Figure 3e−g). We previously reported that self-assembly of Aβ42 initiates with the interchain hydrophobic contacts of Φ 1 and Φ 2 . 14 Intramolecular hydrophobic interaction between Φ 1 and Φ 2 leads to shielding of both interchain interaction sites. However, when Φ 2 interacts with Φ 3 , the Φ 1 domain is consequently exposed. Based on these findings, we hypothesized that Φ 1 is shielded by Φ 2 , thereby delaying fibrillar amyloid aggregation. As such, it may be possible that the shielding effect of Φ 2 could be weakened when the domain interacts with Φ 3 .

Molecular Mechanisms Underlying Suppressed Amyloid Aggregation of Aβ42
We performed MS-based structural analyses to further investigate the molecular mechanisms underlying the suppression of the fibrillar amyloid aggregation process. The deuteration level of proteins during HDX reflects the accessibility of protein regions to solvent molecules. A higher HDX propensity indicated more frequent solvent exposure with an extended conformation. We employed HDX-MS analysis to monitor the structural conversion of Aβ42 in the presence of the designed peptides ( Figure 4a). As a reference, the increment in the deuteration level of inhibitor-free Aβ42 (incubated for 1 s in water prior to acidic quenching) during the MS data acquisition for 90 s was 1.07 ± 0.21%. All three designed inhibitor candidates facilitate Aβ42 deuteration, with P b causing the greatest increase in Aβ42 deuteration level (1.72 ± 0.01, 1.63 ± 0.21, and 1.94 ± 0.38% for P ab , P a , and P b , respectively). Therefore, it is anticipated that the designed peptides, in general, induce Aβ42 to be extended by peptide− protein interactions.
Next, we attempted to dissect the interaction modes of the Aβ42 complexes with different peptides. The structural dynamics of Aβ42 and Aβ42-inhibitor complexes were monitored in the gas phase using IM-MS by determining the collision cross-sections (CCSs) of each complex ion ( Figure  4b). According to the IM spectra, the complex ions (+4 charge state) showed increased CCS values compared to free Aβ42 ions. The complex ions of Aβ42 with P a or P b showed two conformers (compact and extended conformers), while the Aβ42 complex ion with P ab did not show extended conformers (Figure 4b). It should be noted that the compact conformer of the Aβ42−inhibitor complex ion is still extended compared to the intact Aβ42 ion's conformer (Figure 4a,b). The Aβ42-P a complex ions preferred a compact conformation over the extended conformation. In contrast, the P b -bound complex ions displayed an increase in the population of extended conformers. Taken together, the Aβ42-P ab complex ions presented only with a compact conformer, while Aβ42 complexes with the other designed peptides showed both compact and extended conformers. In addition, the Aβ42 complex ions with P a and P b showed extended conformers to a significant extent. The Aβ42-P b complex ions displayed a preferential tendency for extended conformers in contrast to the Aβ42-P a complex. The degree of extension of Aβ42 may play an important role in suppressing the fibrillar amyloid aggregation of Aβ42.
To further investigate the atomistic details of the observed structural changes in Aβ42 and various amyloid complexes, we utilized AlphaFold to predict complex structures between the Aβ42 monomer and each of the three designed peptides. The major binding site was the central hydrophobic region of Aβ42 ( Figure S11). The complex structures were fed into replicaexchange MD (REMD) simulations as initial structures. The REMD simulation sets consisted of 33 different temperatures ranging from 300 to 400 K with equal exchange probabilities. For each designed peptide, we obtained five R g distributions of Aβ42 and five interpeptide contact probability maps from independent REMD simulations and weight-averaged them based on the binding free energy obtained by umbrella sampling (Table S2). The R g distributions of Aβ42 when the protein coexists with the designed peptide P b showed an extra population with an extended conformation centered at 2.4 nm (Figure 4c). The HDX-MS results and R g distribution data indicate that P b induces Aβ42 to be extended more than the other two designed peptides (Figures 4a,c). All designed peptides bound to Aβ42 without a specific interaction site (Figure 4d). We observed a preference for peptide inhibitors associating with the central hydrophobic region along with the C-terminal hydrophobic region of Aβ42. P ab , which spans both the two hydrophobic regions of Aβ42, interacted with Φ 1 and Φ 2 of Aβ42 as expected. P a and P b , whose sequences correspond to the Φ 1 and Φ 2 regions on Aβ42, respectively, also associated with Φ 1 . It is noteworthy that both P a and P b interacted with the Aβ42 Φ 3 domain. Although P a and P b were designed to bind only to their respective Φ 1 and Φ 2 cognate sites on Aβ42, we observed that these inhibitors could also concomitantly interact with the Φ 1 and Φ 3 domains of Aβ42. Because P a preferred contact with the Φ 1 domain of Aβ42 over the Φ 3 region, P a showed effective suppression of Aβ42 aggregation ( Figure S13). On the other hand, P b still contacts the Φ 1 domain most frequently, but contact probabilities do not differ significantly among the three hydrophobic domains ( Figure S13). P ab , which suppressed Aβ42 aggregation the most effectively, also made significant contacts with the Φ 2 domain of Aβ42, but P ab 's binding to the Φ 1 and Φ 2 domains occurs concomitantly not independently ( Figure S14). The (e) Comparison of the residue-specific interpeptide contact probabilities obtained from MD simulations and bindings between Aβ42 and inhibitor candidates, which were monitored by using ETD-MS fragmentation of the Aβ42-inhibitor complex. The product ions are indicated in gray, and the MD simulated contact probability is represented in green, blue, and purple (Aβ42-P ab , Aβ42-P a , and Aβ42-P b , respectively). The error bars represent the standard deviation from three independent experiments.

JACS Au
pubs.acs.org/jacsau Article transient binding of designed peptides into the Φ 3 of Aβ42 may disrupt intramolecular contact between Φ 2 and Φ 3 ( Figure S15). We experimentally examined the probable binding sites of Aβ42 peptide inhibitors by ETD analysis. We used 1,4dicyanobenzene to generate anions, and the chemical reaction between the Aβ42 cation and radical reagent anion led to peptide backbone cleavage. The ETD-MS results revealed that the peptide inhibitors were located around multiple regions of Aβ42 (Figure 4e). Structural analysis of Aβ42 and inhibitor complexes suggests that P a and P b could interact with Aβ42 in two different hydrophobic regions (Φ 1 and Φ 3 , respectively). Although we found that P a and P b could each separately occupy the Φ 1 and Φ 3 Aβ42 sites at the same time, the binding of the designed peptides to different regions may lead to structural conversion of Aβ42 to different extents. Taken together, interaction of the designed peptides with the specific hydrophobic region of Aβ42 leads to conformational changes and suppresses protein aggregation.
In summary, the designed peptides interacted with Aβ42 in multiple hydrophobic domains. The P ab inhibitor associated with the Φ 1 and Φ 2 regions of Aβ42, whereas the other peptides interacted with Φ 1 and Φ 3 . Since binding of either P ab or P a leads to the extension of Aβ42 to a similar degree and conveys an inhibitory effect on fibrillar amyloid aggregation, we are prompted to conclude that the Φ 1 domain on Aβ42 is likely the main target for suppression of amyloid aggregation (Figures 1c, 2a, and 4). These results also demonstrate that binding of inhibitor candidates to the C-terminus of Aβ42 results in an extension of the Aβ42 conformation and cannot effectively disrupt fibrillar amyloid aggregation (Figures 2b and   4). The AlphaFold-predicted binding sites of F a and F b in prefibrillar Aβ42 are also the central hydrophobic regions of the protein. This implies a potential competition for interactions near the central hydrophobic region, where they may also bind adjacent Aβ42 molecules in the amyloid fibrils, inducing Aβ42 fibril disassembly (Figure 1b). 15

Examination of the Role of Three Key Domains in Aβ42 Self-Assembly
Multiple lines of experimental evidence have indicated that Φ 1 is the main target for inhibition of fibrillar amyloid aggregation. Based on the SAXS experiment, we hypothesized that Φ 1 would be shielded by Φ 2 , thereby delaying fibrillar amyloid aggregation, whereas this shielding effect would be weakened when the Φ 2 domain interacted with Φ 3 . To test our hypothesis, we used a double point mutant of Aβ42 (I41N/ A42N) in order to disrupt the association between the Φ 2 and Φ 3 regions on Aβ42. It is commonly accepted that the two hydrophobic residues at the C-terminal end of Aβ42 enhance the initiation of the fibrillar amyloid aggregation process compared to Aβ40. 7 The ThT fluorescence assay was performed to compare the fibrillation kinetics between Aβ42 and the double point mutant isoform (Aβ42 I41N/A42N) in the same peptide concentration ( Figure S17). We observed that the point mutations caused a significant kinetical delay in fibrillar amyloid aggregation (t 1/2,Aβ42 = 11.37 ± 0.82 h and t 1/2,Aβ42I41N/A42N = 22.07 ± 1.69 h).
Moreover, we revealed that the hydrophobic interactiondriven binding of the designed peptides (P a and P b ) was separated into multiple sites, including Φ 1 and Φ 3 ( Figure  4d,e). To investigate the effects of focusing the binding of designed peptides into Φ 1 and eliminating competition for this   (Aβ42 I41N/A42N) and three designed inhibitor candidates. P ab and P a still had an inhibitory effect on fibrillar amyloid aggregation in a concentrationdependent manner (Figure 5a,b). In addition, P b , which failed to abolish the fibrillar amyloid aggregation of wild-type Aβ42, did suppress the process effectively in the mutant isoform ( Figure 5c). The TEM image analysis results were in close agreement with those of the ThT incubation assay (Figure 5d). Interestingly, all three inhibitor candidates showed inhibitory effects to a similar degree, in contrast to wild-type Aβ42 amyloid aggregation (Figures 1 and 2). We confirmed that the conformations of the complex ion (Aβ42 I41N/A42N-P b ) merged into a single distribution in the gas phase of the IM spectra ( Figure S18). These findings indicate that focusing the binding of designed peptides to Φ 1 suppresses fibrillar amyloid aggregation, highlighting Φ 1 as the primary target for suppressing Aβ42 aggregation. Our previous study, which reported the structural difference between Aβ40 and Aβ42, also suggests that absence of the C-terminal end enhances the interaction between the central hydrophobic and the Cterminal hydrophobic regions, thereby increasing the population of compact conformers of Aβ40 compared to Aβ42. 42 Since the central hydrophobic region has been reported to be the key region for initiating Aβ40 aggregation as well, the limited role of the C-terminal end of Aβ42 on fibrillar amyloid aggregation was observed in comparison with that of both Aβ42 I41N/A42N and Aβ40. 43,44 ■ CONCLUSIONS In this work, we investigated the roles of the central ( 17 LVF 19 , Φ 1 ) and C-terminal hydrophobic ( 32 IGL 34 , Φ 2 ) regions, as well as the C-terminal end ( 41 IA 42 , Φ 3 ) of Aβ42 in the fibrillar amyloid aggregation process, using Aβ42 fragments. Taken together, our computational and experimental data suggest that: (1) the Φ 1 domain drives Aβ42 amyloid aggregation and that (2) the Φ 2 Aβ42 site can interact with either Φ 1 or Φ 3 . If Φ 3 is absent or is replaced by hydrophilic residues, then Φ 2 constantly (yet transiently) interacts with Φ 1 , which disrupts early-stage oligomerization. When Φ 3 is involved in hydrophobic interactions, this preferentially binds to Φ 2 , in turn increasing the likelihood of the Φ 1 domain being more exposed to solvent molecules. This exposure can lead to intermolecular hydrophobic interactions, which drive the oligomerization of Aβ42.
Our HDX-MS experiments revealed that nonspecific binding of designed peptides to Aβ42 induces the conversion of the Aβ42 structure to a more extended conformation, which is related to a lack of intramolecular hydrophobic interactions. Based on structural analyses of Aβ42 complexes using IM-MS, REMD simulations, and ETD-MS, we also observed that binding of the designed peptides to the Φ 1 domain of Aβ42 suppresses fibrillar Aβ42 aggregation. This compelling evidence suggests that the designed peptides' association with the Φ 1 domain of Aβ42 can disrupt intermolecular interactions, thereby interfering with Aβ42 cluster formation. Furthermore, we identified intramolecular interactions of Aβ42 using SAXS and MD simulations. The structural analysis results indicated that intramolecular hydrophobic interactions of the Φ 2 Aβ42 site determine the monomeric Aβ42 conformation and accessibility to the Φ 1 domain of the protein. To assess the consequent implications on Aβ42 aggregation, we monitored the effects of substituting the Cterminal end with a polar amino acid using the Aβ42 I41N/ A42N mutant and focusing the intramolecular interactions of Φ 2 to the Φ 1 domain. Our findings revealed that the aggregation kinetics of the double point mutant were delayed compared to that of wild-type Aβ42.
Because the Aβ42 I41N/A42N mutant and Aβ40 share similar features during the kinetical delay of fibrillar amyloid aggregation, we suggest that the probable role of the Aβ42 Cterminal end is to interfere with the interaction between Φ 1 and Φ 2 . We also observed the binding of the designed inhibitor candidates to Aβ42 by combining the MD simulations with MS-based analysis. Multiple lines of evidence indicate that disruption of interpeptide interactions between the Φ 1 domains suppresses the self-assembly properties of Aβ42. The Φ 2 Aβ42 site is not involved in catalyzing the formation of early-stage oligomers but, conversely, slows down the fibrillar amyloid aggregation process of Aβ42. We previously reported that the double point mutation I32N/L34N of Aβ42 promotes the self-assembly of the protein. 14 Based on our findings, we designed potential inhibitors of Aβ42 amyloid aggregation. Our hypothesis was simple: if the inhibitor reduces the likelihood of the Φ 1 domain being exposed, then oligomerization (and fibrillation) may be hampered. The key hydrophobic residues (F19 and I32) of the Aβ42 fragments were replaced by asparagine, which disrupted the alignment of the different Aβ42 chains. The designed peptides, P ab and P a , both of which spanned the Φ 1 domain, could suppress fibrillar amyloid aggregation while alleviating the cytotoxicity of Aβ42 and promoting the disassembly of preformed fibrils. In contrast, another designed peptide, P b , which lacks the Φ 1 domain, failed to disrupt the self-assembly properties of Aβ42.
A higher preference for interacting with the Φ 1 domain leads to the effective suppression of fibrillar amyloid aggregation in the context of drug discovery. Our findings herein prompt us to suggest that the seven-residue peptide, P a , could be a viable candidate as a therapeutic agent that may be further tested in preclinical models. Furthermore, this peptide inhibitor could also be potentially utilized in drug development as an adjuvant to existing therapeutic strategies, such as antibodies and molecular receptors, in order to increase biocompatibility, improve delivery techniques, and enhance the target molecule's binding affinity for the Φ 1 domain of Aβ42. Although binding of P a to the Φ 1 domain of Aβ42 suppresses protein aggregation, it may interact with Aβ42 in multiple domains competitively. According to our findings, competitive binding reduces the efficacy of P a ; thus, combining it with other approaches that increase selectivity to the Φ 1 domain of Aβ42 would be helpful to overcome its limitations.
Our proposed target domain may additionally be used in other types of drug design, including small-molecule inhibitors. Overall, our approaches for investigating inter-and intramolecular interactions and binding with multiple IDP binding sites may be utilized in the development of alternative candidate drugs against several amyloid-deposition neuropathology.
■ ASSOCIATED CONTENT * sı Supporting Information Additional experimental details, materials, and methods, including photographs of experimental setup (PDF)