Impact of Electronic Polarization on Preformed, β-Strand Rich Homogenous and Heterogenous Amyloid Oligomers

Abstract

Amyloids are a subset of intrinsically disordered proteins (IDPs) that self-assemble into cross-β oligomers and fibrils. The structural plasticity of amyloids leads to sampling of metastable, low-molecular-weight oligomers that contribute to cytotoxicity. Of interest are amyloid-β (Aβ) and islet amyloid polypeptide (IAPP), which are involved in the pathology of Alzheimer’s disease and Type 2 Diabetes Mellitus, respectively. In addition to forming homogenous oligomers and fibrils, these species have been found to cross-aggregate in heterogeneous structures. Biophysical properties, including electronic effects, that are unique or conserved between homogenous and heterogenous amyloids oligomers are thus far unexplored. Here, we simulated homogenous and heterogenous amyloid oligomers of Aβ_16-22 and IAPP_20-29 fragments using the Drude oscillator model to investigate the impact of electronic polarization on the structural morphology and stability of preformed hexamers. Upon simulation of preformed, β-strand rich oligomers with Drude, structural rearrangement occurred causing some loss of β-strand structure in favor of random coil content for all oligomers. Homogenous Aβ_16-22 was the most stable system, deriving stability from low polarization in hydrophobic residues and through salt bridge formation. Changes in polarization were observed primarily for Aβ_16-22 residues in heterogenous cross-amyloid systems, displaying a decrease in charged residue dipole moments and an increase in hydrophobic sidechain dipole moments. This work is the first study utilizing the Drude-2019 force field with amyloid oligomers, providing insight into the impact of electronic effects on oligomer structure and highlighting the importance of different microenvironments on amyloid oligomer stability.

2. Introduction

Intrinsically disordered proteins (IDPs) are involved in a variety of complex biological processes including cellular growth, protein modification, signaling cascades, and gene expression despite their lack of defined secondary structure^1-5. Amyloids are an interesting subset of IDPs that self-assemble to form cross-β aggregate fibrils. Some amyloids can misfold to form complex, cytotoxic oligomeric structures, which are associated with diseases including Alzheimer’s disease (AD), Parkinson’s disease, and Type 2 Diabetes Mellitus (T2DM)^6-8. Amyloid aggregation is typically investigated via NMR spectroscopy⁹, X-ray crystallography³, Förster resonance energy transfer (FRET) spectroscopy^{3, 10}, and more, but these techniques are limited in resolution due to the heterogeneity of amyloid conformational ensembles and the rate at which they aggregate and precipitate. Exploring how amyloids misfold and organize into low-molecular-weight oligomeric structures at the atomistic level is necessary for understanding the aggregation pathway of amyloids and the associated impact on disease progression.

Two commonly studied and disease-related amyloids are amyloid-β (Aβ) and islet amyloid polypeptide (IAPP). Aβ is involved in the disease progression of AD whereas IAPP is co-secreted with insulin in the pancreas and has been associated with loss of pancreatic β-cell mass in T2DM. Typically, Aβ and IAPP exacerbate these disease states by self-assembling and forming homogenous oligomeric intermediates inducing cell death, exerting cytotoxicity via ion leakage and perturbing membrane integrity^11-13. These peptides then rearrange into a highly organized, β-sheet rich fibrillar morphologies that are less cytotoxic than low-molecular-weight oligomeric species^14-16. To further compound the complexity of amyloid folding mechanisms and associated cytotoxicity, IAPP can potentiate AD via cross-amyloid interactions with Aβ, or cross-seeding, forming heterogenous aggregates¹⁷.

Both Aβ and IAPP display biochemical similarities that play a pivotal role in amyloid aggregation. Structurally, Aβ and IAPP are short peptides that contain hydrophobic cores, contained within residues 16-22 of Aβ¹⁸ and residues 20-29 of IAPP^{19, 20} that are often used as experimental model systems for fibril growth. Homogenous fibril structures of these amyloids have been identified as organized steric zippers of monomeric β-strand repeats^{20, 21}. The cytotoxic oligomer intermediates form β-barrels, constructed from a ‘β-strand-β-strand’ motif for Aβ^{22, 23} and IAPP²⁴. Recent work has resolved the full-length and fragment fibril structures of Aβ²⁵ and IAPP²⁶, but there are limited data on the more cytotoxic low-molecular-weight oligomeric intermediates of these peptides. Molecular dynamics (MD) simulations have been utilized to explore the formation of the amyloid oligomeric states given studying these amyloid oligomers is experimentally challenging.

Simulating the structures of Aβ and IAPP and exploring their organization and formation of oligomer intermediates has been challenging due to force field inaccuracies, limitations on conformational sampling associated with limits in simulation lengths, and protein concentrations²⁷. Many MD simulation force fields were designed for structured proteins and have shown to have limited accuracy and inadequate reproduction of experimentally characterized conformational ensembles due to overstabilization or over-compacting of IDP systems². Employing MD simulations with force fields more robust for amyloid fragment simulation is necessary for accurately exploring amyloid aggregation pathways. Some success has been found utilizing Amber99SB^{28, 29}, OPLS-AA³⁰, GROMOS96 53A6³¹, and GROMOS96 54A7³², but lingering issues remain related to the overstabilization of certain types of secondary structures³³. The need to utilize more robust force fields that can more accurately simulate and model amyloids such as Aβ and IAPP is an essential next step for understanding disease pathology. Recent advancements in force field development that explicitly represent electronic polarization have allowed for more accurate representation of electrostatic effects, which may provide additional insight into the oligomeric organization of amyloids. Polarizable force fields may be more suitable for simulating these structures given their ability to explicitly represent dipole response as a function of local electric fields, an important feature when considering heterogeneous microenvironments that are found in oligomers and fibrils of Aβ and IAPP.

Most common force fields for biomolecular systems employ a fixed-charged electrostatic model, limiting their ability to accurately represent anisotropic charge distribution and electronic interactions^{34, 35}. Polarizable force fields overcome this limitation by explicitly representing electronic degrees of freedom in a simulation system³⁶. Two polarizable protein force fields that have recently emerged are AMOEBA³⁷ and the classical Drude oscillator force field^38-40. The Drude force field is based on the classical Drude oscillator model, in which auxiliary particles carrying negative charge are attached to non-hydrogen atoms, thereby modeling electronic degrees of freedom. The harmonic oscillations between the core atoms and their associated Drude particles model dipole response in the system. The Drude force field has been previously used to simulate Aβ unfolding to reveal cooperative effects in stabilizing a key α-helix in the central hydrophobic cluster⁴¹, and stabilizing quaternary interactions in amyloid fibrils⁴². Interestingly, previous simulations of amyloid fibrils with the Drude force field suggest that glycine electronic plasticity is particularly important in stabilizing complex fibril structures of multiple amyloid types⁴². To date, no studies on cytotoxic amyloid oligomer intermediates employing the Drude force field (or any polarizable force field) have been performed. Polarizable simulations may provide more detailed atomic insight into homogenous and heterogenous oligomer intermediates. This work seeks to explore the influence of electronic polarization on homogenous and heterogenous Aβ_16-22 and IAPP_20-29 hexamer structures from previous simulations to better characterize the properties of these species.

3. Methods

3.1. System construction

The starting coordinates for each system were extracted as the dominant cluster from RMSD clustering over the last 500 ns of previous simulations⁴³ with the united-atom GROMOS96 53A6 force field (Figures S1-S3). Briefly, six peptide fragments were initially separated by at least 1.5 nm, and aggregation was simulated over 2000 ns. A total of 12 systems were simulated (4 replicates each of Aβ_16-22, IAPP_20-29, and an equimolar mixture of Aβ_16-22 and IAPP_20-29). We converted these united-atom structures into all-atom coordinates using the internal coordinate builder in CHARMM⁴⁴, to be compatible with the CHARMM36m (C36m)⁴⁵ force field. The N- and C-termini of each peptide were capped with acetyl and amide groups, respectively, to avoid spurious end effects. Each system was solvated in a cubic box of CHARMM-modified TIP3P^46-48 water and 150 mM KCl. Details of each system, including the number and type of atoms, are given in Table S1. Systems were energy-minimized in CHARMM⁴⁴ for 500 steps of steepest descent minimization and 500 steps of adopted-basis Newton-Raphson (ABNR) minimization.

3.2. Molecular Dynamics Simulations

Following minimization, equilibration was performed for 1 ns in NAMD⁴⁹. Position restraints (5 kcal mol⁻¹ Å⁻²) were applied to all non-hydrogen protein atoms and an NPT ensemble was maintained by regulating the temperature at 298 K with the Langevin thermostat method^{50, 51} with a friction coefficient, γ, of 5 ps⁻¹. Pressure was set to 1 atm using the Langevin piston method⁵⁰; the oscillation period was 200 fs and decay time was 100 fs. Periodic boundary conditions were applied in all directions and the short-range van der Waals forces were smoothly switched from zero to 10 to 12 Å. The particle mesh Ewald (PME) method^{52, 53} was used to calculate the electrostatic interactions with a real-space cutoff of 12 Å and grid spacing of approximately 1 Å. Bonds to hydrogens atoms were held rigid using the SHAKE⁵⁴ algorithm, allowing for the integration time step of 2 fs in C36m simulations.

Following C36m equilibration, the systems were converted to the Drude-2019 polarizable model³⁹ in CHARMM by adding Drude oscillators and lone pairs to the equilibrated C36m coordinates. The TIP3P water model was converted to the polarizable SWM4-NDP⁵⁵. The Drude oscillators were relaxed using 1000 steps of steepest descent and 500 steps of ABNR energy minimization in CHARMM. During this minimization, all real atoms were restrained with a force constant of 10⁸ kcal mol⁻¹ Å⁻².

Equilibration of the polarizable systems was carried out for 1 ns in NAMD. The same restraints and constraints were applied, and the nonbonded settings were the same with the exception that the van der Waals potential, not force, was switched to zero from 10 to 12 Å, as is the convention for the Drude-2019 force field. Drude simulations employed a dual Langevin thermostat to regulate temperature, in which real atoms were coupled to a thermostat at 298 K with γ = 5 ps⁻¹ and Drude oscillators coupled to a relative thermostat at 1 K and γ = 20 ps⁻¹. Following equilibration, unrestrained production simulations were performed with an in-house version⁵⁶ of OpenMM^{57, 58} version 7.1, using NVIDIA P100 GPUs. Four replicate simulations of each system were performed for 1 μs each, for a total of 4 μs of total sampling for the Aβ_16-22, IAPP_20-29, and a heterogenous Aβ_16-22/IAPP_20-29 hexamers. The NPT ensemble was maintained using the same thermostat as during equilibration, however pressure was regulated at 1 atm using the Monte Carlo barostat in OpenMM with box scaling attempted every 25 integration steps.

3.3. Analysis

Secondary structure analysis was calculated according to the DSSP⁵⁹ method as implemented in GROMACS⁶⁰. Minimum distance calculations were performed as implemented in GROMACS. Standard CHARMM analysis tools were used to calculate radius of gyration (R_g), root-mean-square deviation (RMSD), dipole moments, and solvent-accessible surface area (SASA). Example CHARMM input scripts are available via the Open Science Framework at https://osf.io/mqcfd/.

4. Results and Discussion

Given the involvement of Aβ and IAPP in AD and T2DM, extensive computational work has been performed studying amyloid oligomers, comprised of both full-length peptides^61-64 and fragments^{43, 65, 66}, with additive force fields. The fixed point-charge approximation in these force fields fails to account for inevitable changes in local electric fields over the simulation trajectory⁴⁰. As such, characterization of biomolecules using additive force fields may be insufficient to accurately describe structural changes and the evolution of associated heterogenous microenvironments that may depend on electronic polarization. Oligomers of Aβ_16-22 and IAPP_20-29 have yet to be characterized in terms of these electronic effects. This work seeks to assess the structures adopted by amyloid oligomer fragments utilizing the Drude polarizable model and probe the role of polarization on oligomer compactness and structural changes.

4.1. Polarizable Force Field Favors Random Coil Structures in Preformed Amyloid Oligomers

It has been experimentally determined using methodologies such as transmission electron microscopy, that Aβ oligomers form structures that are rich in antiparallel β-strands⁶⁷ and can form β-barrels⁶⁸. IAPP oligomers, in contrast, are thought to be more disordered⁶⁹, with residues 23-27 adopting transient β-strand structure on pathway to aggregation and fibril formation¹⁹. Very little experimental structural information is available for heterogenous Aβ/IAPP oligomers; however, previous computational work has predicted the formation of pore-like cross-aggregate oligomers⁴³. High β-sheet content in the hydrophobic amyloid fragments is often reported in computational work with both full-length^{61, 68} and fragment oligomer systems^{27, 43, 65}; however, commonly used force fields such as GROMOS96 53A6 in SPC water⁷⁰ and CHARMM22⁷¹ in TIP3P water⁴⁶ are thought to over-stabilize β-sheets and α-helices, respectively, when simulating an Aβ₄₀ monomer³³. However, it should be noted that while GROMOS96 53A6 tends to over-stabilize β-sheets, it was superior in replicating NMR data for an Aβ₄₀ monomer in a force field comparison study⁷². The recently released Drude-2019 force field was optimized to address issues regarding under-stabilization of β-hairpins previously observed in the Drude-2013 force field³⁹. However, the relative stability of secondary structures using this force field has not been explored in the context of amyloid quaternary structures, which are expected to be highly dynamic and heterogeneous.

The starting structures used in this work were rich in β-sheets, with replicates of Aβ_16-22 and heterogeneous systems adopting pore-like structures (Figures S1-S3). At the outset of simulation, the oligomers contained ~50% β-sheet and ~50% coil, whereas IAPP_20-29 was on average ~60% coil and ~40% β-sheet (Table 1). Over the course of the 1-μs simulations with the Drude force field, the β-strand content of the oligomers decreased, ultimately favoring random coil structures (Figure 1, Figures S4-S9, Table 1). β-strand structure was preserved primarily in hydrophobic regions of homogenous systems (Figure S10). In Aβ_16-22 systems, residues Val18, Phe19, Phe20, and Ala21 retained β-strand as the dominant structure (Figure S10). Similarly, β-strand was dominant for residues Phe23, Gly24, Ala25, and Ile26 in IAPP_20-29 systems (Figure S10). As such, our polarizable simulations of Aβ_16-22 and IAPP_20-29 oligomers resulted in the peptides behaving more like intrinsically disordered peptides, favoring disorder, and sampling transient secondary structure. We hypothesize that this change in structure and modest rearrangement of structures is a result of adjusting for the now-present electronic effects as provided by the Drude force field. Interestingly, experimental evidence suggests that these oligomers, particularly Aβ_16-22, should adopt higher β-strand structure than we observed here⁶⁸. In the case of Aβ simulations and experimental results, full-length oligomerization is predicted to enhance formation of β-barrels, with residues 10-21 forming β-sheets at high probabilities⁶⁸. One possible explanation for this observation is a difference in force fields; that is, whereas the GROMOS96 53A6 simulations led to ordered structures rich in β-strands, the disordering observed in the Drude simulations suggests that this starting point was disfavored when modeled with explicit polarization. Another possibility is that β-strand structure is somewhat under-sampled when using the Drude-2019 force field, although there is no demonstration of such instability in ordered proteins against which the force field was validated³⁹. Future studies of monomeric IDPs and other amyloid proteins should be pursued to determine whether β-strand content is consistent with experimental observations.

Table 1.

Average secondary structure percentages for Aβ_16-22, IAPP_20-29, and Aβ_16-22/IAPP_(20-29, at simulation start, and over the last 500 ns of simulation duration. Averages taken across four replicates per system.

	Simulation Start (t = 0)			Last 500 ns
System	Coil (%)	β-Sheet (%)	Helix (%)	Coil (%)	β-Sheet (%)	Helix (%)
Aβ 16-22	43 ± 3	57 ± 3	0 ± 0	55 ± 7	44 ± 6	1 ± 2
IAPP20-29	53 ± 7	47 ± 7	0 ± 0	72 ± 9	28 ± 9	0.01 ± 0.03
Aβ 16-22/ IAPP20-29	46 ± 6	54 ± 6	0 ± 0	60 ± 3	40 ± 3	0.1 ± 0.1

Figure 1.

Average secondary structure percentage over time compared with the average β-strand and random coil percentages over the last 500 ns of simulation with GROMOS96 53A6 force field. Secondary structure percentage is shown for A) Aβ_16-22, B) IAPP_20-29, and C) Aβ_16-22/IAPP_20-29. Percentages for α-helix (black), β-strand (red), and coil (blue) were averaged over all replicates. The standard deviation across replicates is shown as shading for each time series. The average α-helical content over the last 500 ns of simulation with GROMOS96 53A6 is not shown, as it is near 0.

β-strand content was observed across all residues in the heterogeneous Aβ_16-22/IAPP_20-29 oligomers (Figure S10). Hydrophobic residues in the central positions of each peptide were again the most enriched in these structures. It is notable that in the heterogeneous oligomers, β-strand content in Aβ_16-22 was slightly lower than in the homogeneous Aβ_16-22 oligomers, and β-strand content in IAPP_20-29 was comparable in both systems. The stability of β-strands in the heterogeneous oligomers suggests that the Drude-2019 force field does not inherently underestimate the stability of β-strands, but rather is additionally influenced by microenvironments.

In addition to increased coil content, low levels of α-helical structure were sampled in at least one replicate of each system (Figures 1-2 and Figures S4-S9). This behavior was particularly prevalent in peptides that partially dissociated from the core oligomeric structure. Replicates 3 and 4 of IAPP_20-29 in both homogenous and heterogeneous systems adopted transient helical structure; helices in homogenous IAPP_20-29 were observed in residues Phe23, Gly24, and Ala25 (Figure 2A), whereas in heterogenous systems, helices were comprised of IAPP_20-29 residues Ile26, Leu27, and Ser28 (Figure 2B). These helices were transient, with the most stable and long-lasting helices being sampled in replicate 3 of the heterogeneous Aβ_16-22/IAPP_20-29 system, persisting for ~100 ns (Figures S8 and S9). In contrast, Aβ_16-22 replicate 4 adopted a stable helix comprised of residues Val18, Phe19, Phe20, and Ala21, which persisted over the last ~250 ns of simulation (Figures 1, 2C, S4, S7). Interestingly, α-helical structures were not sampled in measurable quantities over the course of simulations with GROMOS96 53A6⁴³, though this force field is known to under-stabilize such structures. The low levels of α-helical structure in all systems simulated with Drude-2019 suggest that the polarizable force field models at least some of these peptides as sampling structures closer to a helix-coil equilibrium.

Figure 2.

Presence of helical structure in amyloid peptides during simulation with the Drude-2019 force field. Helices adopted by A) homogenous IAPP_20-29, B) heterogeneous IAPP_20-29, and C) homogenous Aβ_16-22. Residues adopting helical configurations are shown as ball-and-stick, with carbons colored teal in IAPP_20-29, and as light purple for AB_16-22. Ball-and-stick nitrogen and oxygen atoms are colored blue and red, respectively. Peptides are shown as cartoon. N- and C-termini are shown as spheres, and colored dark blue and dark red, respectively. C) Hydrogen bond stabilizing α-helix between Leu17 and Ala21 backbone is shown with distance shown in Å.

4.2. Stability of Homogenous Amyloid Oligomer Structures

We sought to compare the simulated homogenous oligomers in terms of structural variation observed over time. To probe structural stability, we calculated radius of gyration (R_g) and RMSD over the course of the trajectory (Table S2, Figures S11-S14). Aβ_16-22 oligomers were the most compact of all systems, with an average R_g of 10.7 ± 0.3 Å, and deviated the least from its starting structures, with an average RMSD of 1.3 ± 0.1 Å (Table S2). Hydrophobic packing is a critical component in amyloid oligomerization, and burial of hydrophobic residues should result in a depolarization of their sidechains as a function of being occluded from the aqueous solvent. Moreover, the structures of the different oligomers may result in aggregate-specific shifts in dipole moments. To investigate the association between packing and electronic polarization, sidechain dipole distributions were calculated for hydrophobic residues in both homogenous systems. Of the residues that are present in both Aβ_16-22 and IAPP_20-29 (Leu, Phe, and Ala), no change in sidechain dipole moment was observed between Leu residues (Figure 3A, Table 2). Sidechain dipole moment for Ala21 in Aβ_16-22 was marginally decreased compared to Ala25 in IAPP_20-29 in terms of distribution (Figure 3D), although on average, there was no change (Table 2). The greatest differences in sidechain dipole moment were observed for Phe residues (Figure 3B, 3C, Table 2). On average, the sidechain dipole moment for Phe23 of homogenous IAPP_20-29 (0.9 ± 0.2 D) was elevated compared to homogenous Aβ_16-22 residues Phe19 (0.8 ± 0.1 D) and Phe20 (0.7 ± 0.1 D). The increased polarization of IAPP_20-29 Phe23 relative to Aβ Phe19 and Phe20 may influence aromatic contacts; homogenous Aβ, on average, had closer contacts between Phe residues (7.0 ± 0.3 Å) than IAPP_20-29 (8.0 ± 3.0 Å), as well as closer center-of-mass (COM) distances, on average, between Phe residues (13.0 ± 0.4 Å vs. 17.0 ± 5.0 Å) (Figure S15). Interestingly, Phe23 in homogenous IAPP_20-29 was more solvent-exposed (194 ± 4 Ų solvent-accessible surface area) than Phe residues in homogenous Aβ_16-22 (187 ± 4 Ų and 178 ± 3 Ų) (Table S2 and S3, respectively). Together, these data suggest that increased solvent accessibility is consistent with higher sidechain dipole moments in Phe, and reflects the plasticity of the electronic structure of this residue. Phe residues in Aβ_16-22 may be driven inward to a greater extent since there are two of them, which is subsequently reflected in their depolarization as a function of solvent occlusion.

Figure 3.

Sidechain dipole moment distributions in Aβ_16-22 and IAPP_20-29 systems and salt bridge formation. A-D) Sidechain dipole moment distributions for common residues in Aβ_16-22 and IAPP_20-29, A) Leu17 and Leu27, B) Phe19 and Phe23, C) Phe20 and Phe24, and D) Ala21 and Ala25. E-F) Observed salt bridge between Glu22 and Lys16 of Aβ_16-22 system. Peptides shown as cartoon, with N- and C- termini shown as spheres, colored blue and red, respectively. Glu22 and Lys16 are shown as ball-and-stick, with oxygens and nitrogen atoms colored red and blue, respectively. The distance measurement is shown in Å.

Table 2. Average sidechain dipole moments for common residues in homogenous Aβ_16-22 and homogenous IAPP_20-29.

Averages taken for residues in each peptide, then across replicates.

Aβ16-22 Residue	IAPP20-29 Residue	Aβ16-22 Average Sidechain Dipole Moment (D)	IAPP20-29 Average Sidechain Dipole Moment (D)	Difference
Leu17	Leu27	0.7 ± 0.1	0.7 ± 0.1	0.0
Phe19	Phe23	0.8 ± 0.1	0.9 ± 0.2	0.1
Phe20	Phe23	0.7 ± 0.1	0.9 ± 0.2	0.2
Ala21	Ala25	0.7 ± 0.04	0.7 ± 0.05	0.0

Aβ_16-22 contains charged residues Lys16 and Glu22 flanking the hydrophobic core. We hypothesized that the formation of salt bridges may influence the increased relative stability seen in homogenous Aβ_16-22 systems. In replicates 2-4, Aβ_16-22 Lys16 and Glu22 residues engaged in salt bridges in ~30% of frames across the trajectory (Table S5). Salt bridge formation was elevated in replicate 1, with charged residues engaging in salt bridges in 56% of frames (Figure 3E and 3F). Replicate 1 of homogeneous Aβ_16-22 was the most stable in terms of RMSD and R_g, although these differences may not be significant (Table S6). Furthermore, replicate 1 of homogenous Aβ_16-22 had the highest percentage of β-strand structure over the course of simulation, and was subject to less variation in secondary structure (Table S6). These observations indicate that salt bridge formation in homogenous Aβ_16-22 play an important role in modulating oligomer stability. Many nonpolarizable force fields overestimate the stability of salt bridges, and it has previously been proposed that implicitly polarized or explicitly polarizable force fields may be more suitable models for such interactions⁷³. Our results here suggest that modeling salt bridges correctly has implications for amyloid oligomer stability and should be investigated more thoroughly in the future.

4.3. Impact of Dipole Moments on Homogenous and Heterogenous Amyloid Oligomers

Heterogenous Aβ_16-22/IAPP_20-29 oligomers differed from their homogenous counterparts in terms of stability and compactness. The heterogeneous Aβ_16-22/IAPP_20-29 oligomers maintained R_g values that were intermediate between homogeneous Aβ_16-22 and IAPP_20-29 oligomers (Table S2), reflecting the combination of their different sizes (Aβ_16-22 being smaller than IAPP_20-29). Calculations of R_g over time also revealed that heterogeneous Aβ_16-22/IAPP_20-29 systems were less susceptible to having a peptide dissociate from the core oligomer structure than were the homogenous systems (Figure S11). To assess how differences in microenvironments may affect stability, we compared sidechain dipole moment distributions btabetween residues in the respective systems. Engagement in the heterogeneous Aβ_16-22/IAPP_20-29 oligomers led to a depolarization of Aβ_16-22 charged residues Lys16 and Glu22 and increased polarization of hydrophobic sidechains of Leu17, Val18, and Phe20 (Figure 4A, Figure S16). The depolarization of Lys16 and Glu22 are reflected in the lack of salt bridge formation compared to homogenous Aβ_16-22 (Table S4). Previous simulations of mature amyloid fibrils with Drude indicate that sidechain dipole moments for charged residues increase when engaged in salt bridges⁴², and the results obtained here suggest this phenomenon is true in low-molecular-weight oligomers and therefore may be a persistent feature along the amyloid aggregation pathway.

Figure 4.

Sidechain dipole moment comparison and intermolecular interactions. A) Average sidechain dipole moments per residue in (dark blue) homogenous Aβ_16-22 and (green) heterogeneous Aβ_16-22. B) Glu22 shows more frequent contacts with hydrophobic residues in homogeneous systems. Image shows Glu22 interacting with Leu17, Val18, and Phe20 of Aβ_16-22, and Ile26 of IAPP_20-29. Aβ coils colored lavender, and β-sheets colored as dark purple. IAPP_20-29 coil colored as light blue. Residues in the vicinity of Glu22 are shown as ball-and-stick, with carbons colored dark purple (Aβ) and light blue (IAPP). Oxygens in ball-and-stick are colored red. N- and C- termini are shown as spheres and colored blue and red, respectively. C) Average sidechain dipole moments per residue in (dark blue) homogeneous IAPP_20-29 and (green) heterogeneous IAPP_20-29. A and C) Averages are taken across all peptides and replicates over the entire trajectory.

The polarization of hydrophobic Aβ_16-22 residues Leu17, Val18, and Phe20 in heterogeneous systems may be influenced the introduction of polar residues from IAPP_20-29 to the microenvironment. These shifts in hydrophobic polarization may stabilize interesting configurations; distance matrices show that on average, Glu22 formed closer contacts with hydrophobic residue sidechains in heterogenous systems (Table S7). In replicate 4, Glu22 interacts with hydrophobic residues Leu17, Val18, and Phe20 of Aβ, and Ile26 of IAPP at the interface of solvent and the hydrophobic core (Figure 4B). Interestingly, Phe19 and Ala21 were resistant to shifts in polarization. In heterogeneous systems, Phe19 was less exposed to solvent (116 ± 5 Ų) than Phe20 (129 ± 6 Ų), suggesting this residue experienced more hydrophobic packing, and thus less influenced by polar residues (Table S3).

Few differences were observed between IAPP_20-29 dipole moments in homogenous and heterogeneous systems, with the exception of Gly24 (Figures 4C, S17, and S18). Gly24 dipole moment differences between heterogeneous and homogenous systems did not follow a clear trend, suggesting that this residue is highly sensitive to its microenvironment (Figure S18). There are no strong correlations between Gly24 and SASA in heterogeneous systems (Figure S19). The lack of correlation to SASA suggests the residue microenvironment in heterogeneous systems has more influence over Gly24 dipole moments than solvent exposure. Our previous simulations of amyloid fibrils with the Drude force field suggested that glycine residues exhibit electronic plasticity in response to microenvironment changes to promote the stabilization of a given structure⁴². The relatively increased stability in heterogeneous oligomers with respect to IAPP_20-29 may be attributed to glycine adjusting its electronic structure to favor stability.

5. Conclusions

This work sought to determine the impact of electronic polarization on homogenous and heterogenous amyloid oligomers. Atomistic detail of the structural morphologies of these metastable oligomers is mostly unknown, and computer simulations are sometimes limited in accuracy as a result of force field inaccuracies that lead to imbalance of different secondary structure elements. Utilizing the Drude-2019 force field, we observed that structural rearrangement occurred, most notably a modest decline in β-strand structure as compared to starting structures. The most stable system simulated was Aβ_16-22, whose salt-bridge formation modulated its relatively lower R_g and RMSD compared to other systems. Additionally, residues Phe19 and Phe20 in Aβ_16-22 had lower sidechain dipole moments and were less solvent accessible than Phe23 in IAPP_20-29, which contributed to stronger hydrophobic packing and compactness. In heterogeneous systems, Aβ_16-22 experienced a depolarization of charged residues Lys16 and Glu22, and a polarization of hydrophobic residues Leu17, Val18, and Phe20. These changes can be attributed to lack of salt bridge formation in heterogeneous systems and the introduction of polar residues from IAPP_20-29. Very little difference was observed between homogenous and heterogeneous IAPP_20-29 in terms of polarization, except for Gly24, whose electronic plasticity may work to stabilize heterogenous microenvironments. This work highlights the impact of electronic polarization on the stability and properties of amyloid oligomers, with special focus on the role of microenvironments in maintaining oligomer stability.