Emergence of Alternative Structures in Amyloid Beta 1-42 Monomeric Landscape by N-terminal Hexapeptide Amyloid Inhibitors
Alzheimer’s disease (AD) is characterized by deposition of amyloid beta (Aβ) peptides into senile plaques in the brain. While most familial mutations are associated with early-onset AD, recent studies report the AD-protective nature of two genetic human Aβ variants, i.e. A2T and A2V, in the heterozygous state. The mixture of A2V Aβ1-6 (Aβ6) hexapeptide and WT Aβ1–42 (Αβ42) is also found neuroprotective. Motivated by these findings, in this study we investigate the effects of WT, A2V, and A2T Aβ6 hexapeptide binding on the monomeric WT Aβ42 landscape. For this purpose, we have performed extensive atomistic Replica Exchange Molecular Dynamics simulations, elucidating preferential binding of Aβ42 with the A2V and A2T hexapeptides compared to WT Aβ6. A notable reorganization of the Aβ42 landscape is revealed due to hexapeptide association, as manifested by lowering of transient interactions between the central and C-terminal hydrophobic patches. Concurrently, Aβ6-bound Aβ42 monomer exhibits alternative structural features that are strongly dependent on the hexapeptide sequence. For example, a central helix is more frequently populated within the A2T-bound monomer, while A2V-bound Aβ42 is often enhanced in overall disorder. Taken together, the present simulations offer novel molecular insights onto the effect of the N-terminal hexapeptide binding on the Aβ42 monomer structure, which might help in explaining their reported amyloid inhibition properties.
Alzheimer’s disease (AD), presently accounting for 60–70% of the 47.5 million dementia cases worldwide1, is symptomatic of progressive and irreversible loss of cognitive brain functions including memory, language skills, and spatiotemporal perception. AD is neuro-pathologically characterized by the aggregation of amyloid beta (Aβ) peptides into extracellular senile plaques, formation of intracellular neurofibrillary tangles, and subsequent neu- ronal degeneration2, 3. The causative Aβ peptide exists in varying lengths, Aβ40 and Aβ42 being the most abundant isoforms. Of these, Aβ42 is more aggregation-prone and toxic in nature4. Solid-state (ss) NMR studies reveal that both Aβ40 and Aβ42 fibrils are composed of intermolecular stack of β-strands, termed as cross-β conformation, formed via self-complementary side-chain interfaces – so-called dry steric zipper5. In both Aβ40 and Aβ42 fibrils, residues around the central hydrophobic cluster (CHC) and the hydrophobic C-terminal region (CTR) form two parallel, in-register β-strands separated by a hydrophilic turn region6, 7. Structural polymorphism has been reported within Aβ fibrils, revealing alternative β-sheet arrange- ments, such as a triple β-motif8 or a double horse-shoe like structure composed of five β-strands9. Aβ aggregation is initiated by a rate-limiting primary nucleation process, in which monomers self-associate into oligomeric ‘paranuclei’, followed by fibril assembly10. Oligomers can also be formed via a fibril-catalyzed sec- ondary nucleation pathway from the monomeric peptides11. Soluble Aβ oligomers are reportedly more toxic than the fibrils and monomers, and are considered as pathological target for the treatment of AD12, 13. The neurotox- icity of those oligomers is correlated to their β-sheet content14 that increases with the order of polymerization15.
IBM Thomas J. Watson Research Center, Yorktown Heights, NY, 10598, USA. Correspondence and requests for materials should be addressed to P.D. (email: [email protected]) Possible mechanism for Aβ oligomer toxicity includes direct association with membrane, resulting in membrane disruption and/or pore formation16.The monomer to oligomer transition is thought to be initiated by transient misfolding of the monomer to an aggregation-competent, hairpin-like conformation17–22. This structure is topologically similar to the peptide mon- omer conformation in fibrils, in which CHC and CTR are in direct contact, and residues 22–29 form a turn-like structure19, 23, 24 (hereafter referred as the turn region, see Fig. 1a). Designed Aβ sequences that fold into β-hairpin monomer conformation reportedly stabilize toxic, β-sheet rich oligomers and inhibit amyloid fibril formation20. However, translating the structural knowledge of the ordered amyloid fibrils to the Aβ monomer and pre-fibrillar, soluble oligomers remains non-trivial using conventional experiments, as those species are highly metasta- ble, structurally heterogeneous, and disordered in nature25, 26. Aβ peptide belongs to the class of Intrinsically Disordered Peptides (IDP)27 that lack unique 3-dimentional structure, and instead exist as a dynamic ensemble of interconverting conformations. Molecular simulations at various resolutions28–33 have served as a useful means, particularly when combined with experiments, to characterize structure and binding of the disordered peptides. The Aβ N-terminal region (NTR), that is primarily disordered in fibril structures6, 7, has long been relatively neglected in terms of its role in downstream events of AD.
However, a number of familial mutations leading to aberrant monomer misfolding, aggregation, and related toxicity, have been identified at the NTR, such as the English (H6R) and the Tottori (D7N)34. Interestingly, a familial A2V mutation, while causing demen- tia in homozygous carriers, has been reported to provide protection in heterozygous carriers35. In a seminal whole-genome sequencing study on an Icelandic population, the rare A2T mutation was found to be AD protec- tive and also offered protection from age-related cognitive decline in non-AD patients36. Experimental37–41 and simulation studies42–44 have revealed that, both A2V and A2T mutations differentially affect the Aβ monomer and oligomer structures, peptide aggregation kinetics, and associated toxicity. The protective nature of the WT/ A2V(T) Aβ cross-interactions thus provides a natural path towards design of mechanism-based AD therapeutics. Along this line, Di Fede et al.45 recently showed that Aβ1-6A2V in D-form directly interacts with WT Aβ40/42 and prevents amyloid fibril formation in vivo. Further testing of the hexapeptide tagged with the TAT sequence [Aβ1- 6A2VTAT(D)] in a mouse model revealed alteration of oligomer size distribution, aggregation inhibition, and cerebral amyloid deposition46. However, little is known about the molecular mechanism underlying the inhibitory effect of those short NTR peptide fragments.
Given the importance of monomer misfolding in the amyloidogenesis pathway, it can be expected that bind- ing of these Aβ-derived short inhibitor peptides results in remodeling of the full-length Aβ monomer landscape, thereby interfering with the subsequent aggregation. In this study, we take the first steps towards quantifying the effects of the binding of three different Aβ NTR hexapeptide (Aβ6) variants on the Aβ42 monomeric landscape. The hexapeptide variants studied are WT, A2V, and A2T Aβ6. For this purpose, we have performed large-scale, atomistically-detailed Replica Exchange Molecular Dynamics (REMD) simulations of the full-length peptide and hexapeptide variant in explicit water. In recent years, REMD has been extensively used as a tool to suc- cessfully sample conformational ensembles of amyloid proteins43, 44, 47–49. Our simulations reveal that these NTR-derivatives cause a substantial and sequence-dependent reconfiguration of the Aβ42 landscape. Taken together, this simulation study illustrates the potential of the NTR-based hexapeptides toward mechanism-based therapeutic design and calls for further extensive structural investigation, both computational and experimental.
Results
Simulating Aβ42 monomer in presence of WT, A2V, or A2T Aβ6 hexapeptide variant. The amino acid sequence of the Aβ42 monomer and the three Aβ6 NTR-derivatives studied here are shown in Fig. 1a. For brevity, we will refer these three Aβ42 + Aβ6 systems as WT-bound (or + WT), A2V-bound (or + A2V), and A2T- bound (or + A2T) henceforth. The simulated system comprising of the solvated full-length peptide monomer and hexapeptide variant is represented in Fig. 1b. To quantify the effect of Aβ6 binding on Aβ42 landscape, we com- pared the simulations of Aβ42 in presence of Aβ6 with earlier reported simulations of the free Aβ42 monomer42. To ensure that the choice of the initial peptide conformation did not bias the results, we computed the root-mean-square deviation (RMSD) of the Aβ42 backbone from the initial peptide structure as a function of sim- ulation time. Figure 1c depicts the backbone RMSD evolution of the 308.4 K replica as an example. The backbone RMSD reaches a value of 1.00 nm around 20 ns, then quickly equilibrates to ~1.20 nm by 50 ns, and fluctuates around this value (with a standard deviation of 0.12 nm) for the rest of the simulation. Based on this result, we consider that the REMD trajectories reached equilibration by around 50 ns, and the 60–200 ns portion of the trajectories was regarded as the production run. The production ensemble consists of structures extracted every 50 ps from the twelve replicas (276.0–308.4 K), resulting into 33,600 structures. We further confirmed sampling convergence by comparing the residue-wise turn propensity of Aβ42 over the time intervals of 60–130 ns and 60–200 ns (Fig. 1d, also see Table S1). The overall turn propensity shows a mean value of ~50% for the hexapeptide-bound Aβ42 monomer, and is almost indistinguishable over the two time-intervals. Equilibration and convergence of the simulations were also tested in terms of additional structural properties (see Supplementary Information, Fig. S1, and Table S1). These results suggest that the production ensemble is not biased towards the initial conformation and has reasonably converged to a quasi-equilibrium state.
The radius of gyration (Rg) distribution of the Aβ42 monomer, as estimated from the production ensemble (see Supplementary Fig. S2), reveals a mean value around 1.05 nm for the free Aβ42 and is unchanged upon hexa- peptide binding. This value agrees well with the value of the hydrodynamic radius (0.9 nm) reported in fluores- cence correlation spectroscopy measurements50 and Rg values reported in earlier REMD simulation studies47, 51. The estimated Rg value indicates a collapsed structure, as Rg scales as N1/3 (where N = 42)52, 53. We have also quantified the extent of disorder within the production ensemble by performing an RMSD-based clustering (see Supplementary Information). The results, as shown in Supplementary Fig. S2, imply that the top 50 clusters cumulatively represent ~90% of the total population for all systems. Thus, the Aβ42 peptide remains intrinsically disordered in presence of the hexapeptide variant, to the same extent as it is in its free state. Hexapeptide interaction modifies Aβ42 secondary structure profile. Figure 2 shows the ensemble-averaged, residue-wise population of the secondary structure elements within the free Aβ42 ensemble. The standard error values were obtained from standard deviations estimated by dividing the simulation data into four 35 ns long, non-overlapping blocks between 60–200 ns. The calculated standard error values are negligible, indicating statistical significance of the values reported (see Model and Methods in Supplementary Information). A notable feature of the secondary structure profile of free Aβ42, is a >20% β-strand propensity around the CHC (residues 17–21) and the CTR (residues 30–41). The residue-wise secondary structure distributions of hexapeptide-bound Aβ42 can be found in Supplementary Fig. S3. The overall disordered nature of Aβ42 in all sys- tems is illustrated, as coil and turn population together account for 70–75% of secondary structure. Hexapeptide binding in general results in considerable lowering of the CHC and CTR β-strand propensity (at least by 10% in majority of those residues, see Fig. 2b–d), when compared to free Aβ42. The β-strand tendency reduction is most prominent in the CHC residues. In contrast, the preCHC region (residues 11–16) exhibits enhanced β-strand tendency in all three hexapeptide-bound Aβ42 systems. β-strand formation is also noticed around residues 6 and 29 in WT-bound Aβ42 and residues 28–30 in A2V-bound Aβ42. The first 17 residues appear more helix-rich in the WT-bound and A2T-bound peptide (Fig. 2b,d). On the other hand, the helicity becomes almost negligible due to A2V hexapeptide binding.
The short peptide mainly populates coil structure (≥60%) in all three variant forms (Supplementary Fig. S4). Interestingly, the A2V hexapeptide shows notable (~15%) propensity to form β-strand around residues 2 and 3, when compared to the two other variants.
Wild-type CHC-CTR interaction is reduced due to Aβ6 binding. Next, we compare the intramolecular contact map of the bound Aβ42 with that of the free peptide, to reveal the effect of short peptide binding on the Aβ42 tertiary structure (see Fig. 3 and Supplementary Fig. S5). It should be noted that all long-range (|i-j| > 8) con- tacts observed are transiently populated (with a ~12% average probability of formation) within the unbound Aβ42 ensemble, in accordance with its intrinsically disordered nature. Noticeable long-range interactions are: (i) a set of anti-parallel, hairpin interactions between the CHC and CTR residues, (ii) a second, less-extensive set of hairpin interactions within CTR residues 30–41, and (iii) contacts between two termini. The observed β-strand procliv- ity of the CHC and CTR residues, the dominant (>70%) turn tendency of residues 24–28, and the CHC-CTR tertiary interactions, together suggest transient population of a CHC-CTR β-hairpin monomer. Such transiently formed, extended hairpin-like monomer structures are believed to be aggregation-prone in nature, therefore triggering the formation of soluble, toxic oligomers and subsequent polymerization20, 23, 24. Interestingly, the Aβ42 tertiary contact map reveals significant changes due to hexapeptide binding. Only the contacts showing a probability difference value higher than 0.06 were considered for this analysis. First of all, hexapeptide binding induces a marked ≥20% reduction in the CHC-CTR interaction (Fig. 3). At the same time, the D23-K28 salt-bridge frequency decreases from 13% in free Aβ42 to 8–9% in the bound peptide. These find- ings together suggest that hexapeptide binding reduces the transient folding into an aggregation-prone β-hairpin structure, of which CHC-CTR contacts and D23-K28 salt-bridge are key features.
Additionally, interac- tions of extreme NTR with rest of the protein generally decrease in the bound Aβ42. On the other hand, enhanced interaction between the preCHC residues 12–16 and the CTR residues 30–35 is observed, following the order: A2T-bound (12%)
Disordered structures. The A2V-bound S1 and S2 states are consistent with an overall disordered structure (Fig. 6a and Supplementary Fig. S7), and together account for about 28% of the total population. Those Aβ42 structures mainly sample turn or coil conformation. The hexapeptide extensively engages with the CHC, 22–29 turn, or CTR residues, which directly competes with the CHC-CTR hairpin interactions needed to nucleate Aβ aggregation. V2 from the short peptide is frequently involved in these interactions. Central helix structures. The preCHC/CHC helix is primarily seen in the A2T-bound S1, A2T-bound S3, and in WT-bound S2 states (Fig. 6b and Supplementary Fig. S7). This helical feature is most dominant in A2T-bound Aβ42, accounting for about 34% of the total ensemble (Table 1). Additional features, such as β-strand at the Aβ42 NTR or within the short peptide, are also seen (see Supplementary Fig. S7). In these structures, the hexapeptide
residues directly interact with the CTR hairpin, forming a β conformation. In some structures, the NTR, preCHC, and/or CHC residues adopt helical form. Within this sub-population, the hexapeptide directly contacts with the CTR hairpin or the CHC residues, again disrupting the CHC-CTR intramolecular interactions. F4 and R5 from Aβ6 dominate the intermolecular binding. NTR-CTR β-sheet structures. The WT-bound S5 and the A2V-bound S4 states are characterized by an NTR-CTR β-sheet (Fig. 6d, Supplementary Fig. S7d). These sub-populations account for about 16.5% of the WT-bound and 23% of the A2V-bound ensemble (Table 1). The NTR of Aβ42 is mainly found to be in contact with the hexapep- tide, whereas F4 from Aβ6 governs the quaternary association (Table 1).
Discussion
In light of the AD-protective nature of familial A2V and A2T mutations in heterozygous carriers, therapeutic design based on the molecular basis of WT/A2V(A2T) Aβ cross-interaction appears promising. In vivo and in vitro studies have further confirmed the striking ability of the A2V and A2T Aβ variants to impede Aβ nuclea- tion, neurotoxicity, and aggregation37, 38, 41, 57, 58 by directly interacting with WT Aβ. Remarkably, this inhibitory effect was found to be retained, even when WT Aβ42 was co-incubated with a short 1–6 WT or A2V Aβ vari- ant, the effect being more prominent for the A2V variant35, 45, 46. Recently, short A2T NTR fragments of varying lengths have also been reported to inhibit fibrillization and rescue from toxicity59. A molecular characterization of the NTR hexapeptide binding with the full-length peptide is therefore crucial towards understanding of the aggregation-inhibitive and neuroprotective properties of those short peptides, which is useful for introducing new AD therapy. To our knowledge, the present simulation study reports for the first time the striking effect of the Aβ NTR hexapeptide binding on the WT Aβ42 monomeric landscape. Our results reveal remarkable differences between the WT, A2V, and A2T hexapeptide variant in terms of Aβ42 binding, and resulting structural changes.
While the overall disordered nature of Aβ42 persists, the transient CHC-CTR β-hairpin interactions that are implicated in Aβ aggregation and toxicity18–20 is strikingly reduced upon hexapeptide binding. Consequently, short-lived alternative populations emerge on the monomeric conformational landscape. The good agreement between the conformational landscape of the free and hexapeptide-bound monomer indicates involvement of conformational selection60 in the binding process. However, present simulations do not allow to quantify if, and to what extent, conformational selection governs binding. In the bound state, Aβ6 frequently populates a ‘fuzzy’ cloud of possible orientations around a transiently populated Aβ42 structure (Fig. 6 and Supplementary Fig. S7). In those fuzzy complexes61, the loss in overall binding enthalpy is somewhat compensated by a lower entropic loss. As a result of such dynamic interaction, long range electrostatic and transient physical contacts may increase61, leading to greater number of initial contacts and an increased capture radius62. Being an IDP, Aβ42 does not provide a unique binding groove to the hexapeptide, leading to promiscuous binding63. However, the Aβ42 NTR is often found engaged with the hexapeptide, which is desired, given the emerging pivotal role of NTR in Aβ structure, oligomerization/aggregation, associated toxicity, as well as interactions with anti-amyloid molecules64–66. A more flexible NTR is often associated with higher toxicity, as Aβ42 NTR is found more flexible than that of Aβ40 in monomer67 and dimer simulations68.
Four distinct structural features are consistently noticed within the emerging alternative Aβ42 populations due to hexapeptide binding: (i) an overall unstructured population, (ii) a preCHC/CHC helix, (iii) a CTR β-hairpin, and (iv) an NTR-CTR β-sheet. The relative propensity of hexapeptide-bound Aβ42 to visit these alternative sub-populations is strongly dependent on the Aβ6 sequence. For example, a central helix or a CTR β-hairpin is more frequently populated within the A2T-bound monomer, while A2V-bound Aβ42 is often disordered, or forms an intramolecular β-sheet involving both termini. The emergence of a preCHC/CHC helix in the A2T or WT hexapeptide-bound Aβ42 (Fig. 6b and Supplementary Fig. S7) is of interest, since helix stabilization in this region has been reported to reduce fibril formation69 and counteract toxic oligomer formation70. Mutation such as V18A/F19A/F20A or addition of phospho-L-serine in the CHC region increases helical content, and subsequently reduces aggregation69. Therefore, enhanced helical Aβ42 population resulting from WT or A2T hexapeptide binding may impede Aβ42 aggregation, which is consistent with experimental findings35, 45. A2V hexapeptide binding favors an overall unstructured population of monomeric Aβ42. Since Aβ mono- mer misfolding is implicated in formation of β-sheet rich oligomers20, the simulation results are in line with the experimental observation that binding of A2V hexapeptide inhibits β-structure formation in full-length Aβ45. Redirecting IDPs such as Aβ and α-synuclein towards stable, unstructured non-toxic oligomer formation has been associated with the aggregation and toxicity inhibiting effect of resveratrol71 and epigallocatechin gallate (EGCG) molecules72. Taken together, enhanced disorder in A2V Aβ6-bound Aβ42 might prevent toxic oligomer formation.
It is widely accepted that transiently formed β-strand/sheet-rich monomeric structures can promote aggre- gation and toxic oligomer formation in amyloidogenic proteins73–77, such as huntingtin exon1, α-synuclein and IAPP. Previous studies have also indicated a positive correlation between hydrophobic solvent exposure and aggregation propensity (and related toxicity)14. The hydrophobic solvent accessible surface area is consistently estimated to be smaller in the alternative Aβ42 structures with the C-terminal hairpin (2174 ± 62 Å2) or the NTR-CTR β-sheet (2375 ± 67 Å2), when compared to structures with a CHC-CTR β-hairpin (2567 ± 76 Å2). Thus, enhanced population of these atypical β-strand/sheet-rich conformations may help in lowering Aβ aggrega- tion and toxicity. Commensurate to this speculation, a monomer with a C-terminal hairpin was more frequently seen in simulations of the protective A2T Aβ variant compared to the wild-type variant42. On the other hand, a double β-hairpin was often visited by the causative A2V Aβ monomer. Redirecting Aβ monomers to atypical conformations appears a more feasible and attractive strategy for treating AD, compared to directly targeting Aβ production, as Aβ is essential in modulating synaptic activity, neuronal viability, and has shown potential antioxidative functions78, 79. The alternative conformations stabi- lized on the hexapeptide-bound Aβ landscape might be off the aggregation pathway, as indicated by the earlier experiments69–72. However, it has been proposed that the α-helical epitope, while present in a sub-population of monomeric Aβ, may be absent in oligomers and higher order aggregates. In fact, solution NMR in presence of helix promoting agents such as HFIP or SDS suggests helix formation only in Aβ monomers80. Thus, the A2T hexapeptide likely recognizes an epitope that is only present in a monomeric Aβ42 sub-population, resulting into a higher preference for monomer. On the other hand, the unstructured peptide conformation is found in a broad variety of Aβ species including oligomers and is related to lower toxicity71, 72. Therefore, targeting the unstruc- tured epitope, as observed in the A2V hexapeptide binding, seems a more rational approach for AD therapeutic design. Our simulations reveal that the A2V short peptide binding stabilizes the unstructured monomer, subse- quently sequestering the central and/or C-terminal hydrophobic regions and targeting the central turn region, thus eliminating the essential structural features needed for toxic Aβ oligomer and fibril formation. The unique population of disordered Aβ42, together with stronger binding affinity of the A2V hexapeptide, may explain its more pronounced effect on inhibiting β-sheet rich oligomer formation and aggregation of the full-length peptide, with respect to what was seen with the WT hexapeptide45.
A number of phenolic compounds, such as EGCG72, Congo Red dye81 and resveratrol71, have been previously identified to prevent Aβ aggregation and toxicity. Short peptide fragments derived from natural Aβ sequence have also been extensively studied as potential inhibitors of AD. The hydrophobic CHC derivatives82, while being effective fibrillation inhibitors, are reportedly less efficient in preventing oligomerization. On the other hand, CTR fragments83, 84 are more successful in mediating both oligomerization and fibrillogenesis through diverse mech- anisms. Recent experiments establish the amyloid inhibition properties of the NTR Aβ fragment that is mainly hydrophilic in nature. The present study consistently illuminates the sequence-dependent, differential ability of the NTR hexapeptides to reconfigure the Aβ folding landscape away from disease-implicated structures. It will be intriguing to study the effect of length and sequence variation of the short peptide on the Aβ42 A2ti-1 folding landscape, which will be addressed in future. In summary, our intriguing findings unravel the key structural features of the full-length Aβ monomer binding with N-terminal fragment derivatives, which offers novel molecular insights onto their amyloid inhibition properties and can further guide rational design of mechanism-based therapeutics for devastating protein aggregation diseases.