De novo aggregation of Alzheimer’s Aβ25-35 peptides in a lipid bilayer

A potential mechanism of cytotoxicity attributed to Alzheimer’s Aβ peptides postulates that their aggregation disrupts membrane structure causing uncontrollable permeation of Ca2+ ions. To gain molecular insights into these processes, we have performed all-atom explicit solvent replica exchange with solute tempering molecular dynamics simulations probing aggregation of the naturally occurring Aβ fragment Aβ25-35 within the DMPC lipid bilayer. To compare the impact produced on the bilayer by Aβ25-35 oligomers and monomers, we used as a control our previous simulations, which explored binding of Aβ25-35 monomers to the same bilayer. We found that compared to monomeric species aggregation results in much deeper insertion of Aβ25-35 peptides into the bilayer hydrophobic core causing more pronounced disruption in its structure. Aβ25-35 peptides aggregate by incorporating monomer-like structures with stable C-terminal helix. As a result the Aβ25-35 dimer features unusual helix head-to-tail topology supported by a parallel off-registry interface. Such topology affords further growth of an aggregate by recruiting additional peptides. Free energy landscape reveals that inserted dimers represent the dominant equilibrium state augmented by two metastable states associated with surface bound dimers and inserted monomers. Using the free energy landscape we propose the pathway of Aβ25-35 binding, aggregation, and insertion into the lipid bilayer.

. Random walk of replicas over temperatures in a representative REST trajectory. The distribution of replicas over temperatures at the trajectory start is color-coded according to the scale at the right. Han and Hansmann [1] have proposed a quantitative measure of replica mixing defined as where T is the REST temperature and tr is the total number of REST iterations spent at T by replica r. If the total number of replicas is R=8, then the optimum theoretical value of m(T) for any temperature is 1 -1/R 1/2 = 0.65. In Fig. S2 we display the measure m(T) after averaging over all REST trajectories. Although there are some deviations near the ends of temperature range due to boundary effects, m(T) approaches the theoretical value at most REST temperatures. Thus, this figure suggests nearly ideal replica mixing.

Conformational sampling in REST simulations:
Our REST trajectories have been initiated with an equal mix of peptides that are surface bound or inserted into the DMPC bilayer. In all, we have produced six REST trajectories, in which each of eight replicas was simulated for 60 ns. To monitor equilibration, we first considered the probability distributions P(Zm) of the z-position of A25-35 peptide center of mass Zm along the DMPC bilayer normal. These distributions were computed at 330K over the six batches of 10 ns each using all six trajectories. Six corresponding distributions P(Zm) are presented in Fig. S3a. The first four distributions collected over first 40 ns of sampling demonstrate shifting probabilities of surface bound and inserted states reflecting the process of equilibration of binding of A25-35 dimers to the bilayer. However, the two distributions collected over the last 20 ns of sampling are nearly identical suggesting settling of the system in equilibrium state. The distance R between the centers of mass of A25-35 peptides forming a dimer in a leaflet can be used as a second measure of aggregation equilibration. To this end, we plot in Fig. S3b the distance R, which is averaged over six REST trajectories and pairs of dimers, as a function of REST time. If interactions between A25-35 peptides are at equilibrium, R should reach a plateu as a function of REST simulation time. Fig. S3b shows that this is indeed the case after approximately 40 ns of sampling. Thus, we determine that the equilibration time in our REST simulations of A25-35 dimers is eq≈40 ns. Aβ25-35 secondary structure: Table S1 compares secondary structure in Aβ25-35 monomers [2] and dimers by listing the average probabilities to observe helical <H>, −turn <T>, and random coil <RC> conformations. It reveals a moderate increase in the helical fraction <H> coupled with a simultaneous decrease in the turn content <T>. Similar but more pronounced changes are observed in the R4 C-terminus, which features a stable helical structure (>0.50).

Aβ25-35 intrapeptide interactions:
Effect of aggregation on intrapeptide interactions was assessed via the difference contact map <ΔC(i,j)> = <C(i,j)> -<C(i,j)>M, where <C(i,j)> and <C(i,j)>M are the dimer and monomer [2] contact maps reporting the formation of contacts between amino acids i and j. The contacts most affected by aggregation (|<C(i,j)>| ≥ 0.1) are shown in Table S2. These interactions reflect stabilization of helical structure in A25-35 dimers. In fact, two helix contacts (Gly29-Leu32, Gly29-Gly33) became particularly stable as their probability of formation reaches 0.74 and 0.70, respectively. However, because only three contacts out of 55 topologically possible are noticeably affected by aggregation and two of them (Gly29-Leu32, Gly29-Gly33) are already stable in A25-35 monomers (<C(i,j)>M>0.4) [2], we surmise that aggregation does not cause a radical change in A25-35 tertiary structure.

Table S2
List of intrapeptide contacts affected by aggregation.

A25-35 interactions with the DMPC bilayer:
Differences in the binding mechanism between A25-35 dimers and monomers were explored using the contact map <Cl(i,k)>, which reports the formation of contacts between amino acids i and lipid groups k . Fig. S4 shows the contact maps for A25-35 dimers and monomers as well as the difference in the number of contacts with lipids per amino acid <Cl(i)>= <Cl(i)> -<Cl(i)>M, where <Cl(i)> is the number of contacts formed by amino acid i with all lipid groups and subscript M refers to monomer [2]. The figure shows that peptide aggregation enhances interactions of all A25-35 amino acids with the DMPC bilayer. Detailed analysis of amino acid -lipid interactions is given in the main text.

Comparison of A25-35 inserted dimers and monomers:
Our simulations of A25-35 dimers suggest that aggregation does not qualitatively change the secondary and tertiary structure of A25-35 monomers.
To provide a direct evidence, we plot in Fig. S6 the residue-specific distribution of helical structure in inserted A25-35 ID dimers (Fig. 5) and inserted monomers I sampled in our previous REST simulations [2]. In both A25-35 conformational ensembles the helical structure is localized in the C-terminus (<H(R4)>=0.60 in ID and 0.40 in I). Consistent with Fig. 2 this figure also indicates that aggregation stabilizes helical structure in A25-35. In addition, Table S3 presents the list of stable intrapeptide contacts in these two A25-35 species. It follows from the table that all five stable ID intrapeptide contacts are also present as stable interactions in I. Fig. S6 and Table S3 also include the data for the inserted monomers IM sampled in the dimer simulations (Fig. 5). As in the two ensembles discussed above IM features a stable helix in R4 (<H(R4)>=0.66) and four out of five stable ID contacts appear among stable intrapeptide IM interactions. It is also worth noting that IM and I ensembles share three most stable contacts in the same descending order of stability. Thus, Fig. S6 and Table S3 demonstrate that A25-35 dimers utilize monomer-like peptide conformations, i.e., A25-35 monomers inserted into the DMPC bilayer are aggregation-ready.

Figure S6
Helical propensities <H(i)> for amino acids i in A25-35 dimers ID (in black), inserted monomers I from the previous study [2] (in red), and inserted monomers IM from the current study (in blue). Vertical bars show sampling errors. Regions R3-R4 are colored according to Fig. 1a.

Effect of insertion depth on helical propensity:
The helical propensity <H(Zm)> in A25-35 peptide as a function of the position of its center of mass along the bilayer normal Zm is presented in Fig. S7. It shows that in the peptides forming dimers <H(Zm)> steadily increases with the depth of their insertion in the bilayer. In contrast, the helical fraction remains largely unchanged, when A25-35 monomer binds to the DMPC bilayer. The likely reason for the differing outcomes is an increase in the hydrophobic moment of ID dimer composed of two head-to-tail helices compared to a stand-alone monomer. Indeed, using the hydrophobic scale of Wimley et al [3], we found that the hydrophobic moments of ID and inserted monomer IM are 5.0 and 4.1 Å kcal/mol, respectively. Increase in the helix fraction with the insertion depth has also been observed by Garcia and coworkers for WALP-16 peptide [4].

Possible impacts of bilayer composition and post-translational modifications on A25-35 aggregation:
It is important to discuss potential impact of bilayer composition on A25-35 aggregation. The aggregation mechanism summarized in Figs. 5 and 6 is likely to remain valid as long as binding to the bilayer induces helical structure in the A25-35 C-terminus. Although the precise details and the extent of this secondary structure transition depend on the bilayer composition, it was observed upon binding of A peptides to diverse bilayers, including the zwitterionic (DMPC, current work and [2]), anionic (DMPS [5]), or cationic (DMPC + lipopeptides [6]) bilayers and the cholesterol-enriched DMPC bilayer [7]. Therefore, we hypothesize that the formation of ID dimer may be robust against changes in the bilayer composition. Interestingly, our preliminary data on A25-35 monomer binding to the equimolar ternary bilayer composed of DMPC, PSM, and cholesterol support this conclusion by showing that binding to this bilayer promotes helical fraction approximately to the same extent as the pure DMPC bilayer [2]. It is important to note that the arguments presented above apply to A25-35 dimers, whereas larger oligomers may respond differently to changes in bilayer composition.
It is interesting to discuss the question whether A25-35 dimers promote permeation of Ca 2+ ions. We have studied binding of A10-40 monomers to the DMPC bilayer coincubated with 150 mM of Ca 2+ ions [8]. That work showed that Ca 2+ enhances binding affinity of anionic A10-40 by disrupting the intrapeptide salt bridge and by making zwitterionic DMPC partially cationic due to strong coordination of Ca 2+ with phosphate groups. In addition, anionic amino acids in A10-40 attract Ca 2+ ions bringing them with the peptide upon its shallow insertion into the bilayer. None of these factors are applicable to A25-35, which is cationic and does not have intrapeptide salt-bridges. Therefore, it is conceivable that Ca 2+ ions will decrease A25-35 affinity to the bilayer consistent with the experimental findings of Tatulian and coworkers [9]. Hence, we tentatively conclude that A25-35 dimer may not strongly affect Ca 2+ permeation and larger aggregates embedded in the bilayer are needed to change this outcome.
Importantly, the latter suggestion agrees well with the recent experiments showing that Ca 2+ permeating pores are built of up to eight A25-35 oligomers each involving from four to eight peptides [9] or formed by a single -barrel oligomer composed of six A25-35 peptides [10].
Finally, it is of interest to evaluate a possible effect of Ser26 phosphorylation on A25-35. Previous studies have shown that this post-translational modification stabilizes A1-40 oligomers and increases their cytotoxicity [11]. We are not aware of phosphorylation studies of A25-35, but our current findings allow us to speculate on its plausible implications. Phosphorylation of Ser26 will introduce a negative charge in the peptide N-terminus. Since Ser26 is bound to the DMPC headgroups, we do not expect that phosphorylation will directly affect the location of this residue within the bilayer. Also, because Ser26 is not part of ID dimer aggregation interface, phosphorylation is unlikely to directly impact the ID species. However, phosphorylated Ser26 may form a salt-bridge with Lys28 as it was recently shown in replica exchange simulations of A21-30 fragment [12]. If so, phosphorylation will indirectly compromise Lys28 binding to DMPC phosphate groups and, in turn, reduce A25-35 propensity to insert into the bilayer thus shifting its conformational ensemble away from the inserted dimers. This hypothesis will be tested in our future simulations.