Familial Alzheimer’s Disease Osaka Mutant (ΔE22) β-Barrels Suggest an Explanation for the Different Aβ_1–40/42 Preferred Conformational States Observed by Experiment

Abstract

An unusual ΔE693 mutation in the amyloid precursor protein (APP) producing a β-amyloid (Aβ) peptide lacking glutamic acid at position 22 (Glu22) was recently discovered, and dabbed the Osaka mutant (ΔE22). Previously, several point mutations in the Aβ peptide involving Glu22 substitutions were identified and implicated in the early onset of familial Alzheimer’s disease (FAD). Despite the absence of Glu22, the Osaka mutant is also associated with FAD, showing a recessive inheritance in families affected by the disease. To see whether this aggregation-prone Aβ mutant could directly relate to the Aβ ion channel-mediated mechanism as observed for the wild type (WT) Aβ peptide in AD pathology, we modeled Osaka mutant β-barrels in a lipid bilayer. Using molecular dynamics (MD) simulations, two conformer ΔE22 barrels with the U-shaped monomer conformation derived from NMR-based WT Aβ fibrils were simulated in explicit lipid environment. Here, we show that the ΔE22 barrels obtain the lipid-relaxed β-sheet channel topology, indistinguishable from the WT Aβ_1–42 barrels, as do the outer and pore dimensions of octadecameric (18-mer) ΔE22 barrels. Although the ΔE22 barrels lose the cationic binding site in the pore which is normally provided by the negatively charged Glu22 side-chains, the mutant pores gain a new cationic binding site by Glu11 at the lower bilayer leaflet, and exhibit ion fluctuations similar to the WT barrels. Of particular interest, this deletion mutant suggests that toxic WT Aβ_1–42 would preferentially adopt a less C-terminal turn similar to that observed for Aβ_17–42, and explains why the solid state NMR data for Aβ_1–40 point to a more C-terminal turn conformation. The observed ΔE22 barrels conformational preferences also suggest an explanation for the lower neurotoxicity in rat primary neurons as compared to WT Aβ_1–42.

INTRODUCTION

A rare mutation in the amyloid precursor protein (APP) with a deletion of glutamic acid, ΔE693, was first identified in patients of Japanese pedigree.¹ In clinical studies on propositus, it was discovered that the variant of APP is closely linked to pathogenesis of Alzheimer’s disease (AD) with symptoms similar to AD-type dementia. The ΔE693 mutation in APP produces a form of β-amyloid (Aβ) peptide that lacks a Glu22 residue, ΔE22, which is known as the Osaka mutant. Initial studies in vitro suggested that the mutant did not form fibrils, but presented subcellular oligomers in transfected cells.² Subsequent studies in vivo showed that transgenic mice exhibited age-dependent intraneuronal Aβ oligomerization without extracellular amyloid deposits.³

In contrast to earlier reports that the Osaka mutant (ΔE22) did not form fibrils,^1–3 recent studies, however, demonstrated that the mutant peptides derived from Aβ_1–42 and Aβ_1–40 have strong tendency to form fibrils faster than those of wild type (WT) Aβ peptides.^4–6 The analysis of secondary structure dynamics showed that elimination of Glu22 from both Aβ_1–42 and Aβ_1–40 (below referred to ΔE22-Aβ_1–42 and ΔE22-Aβ_1–40, respectively) substantially increase β-sheet formation propensities.⁴ Subsequent fibril morphology study demonstrated that both ΔE22-Aβ_1–42 and ΔE22-Aβ_1–40 form short protofibrillar and fibrillar structures with high conformational stability. Mixture of ΔE22-Aβ_1–40 and WT Aβ_1–40 also produced fibrils with morphology similar to that of fibril made of pure ΔE22-Aβ_1–40, indicating that ΔE22-Aβ_1–40 fibrils served as a seed for WT Aβ_1–40 fibril elongation.⁵ In rat primary neuron cultures, ΔE22-Aβ_1–40 was neurotoxic, while WT Aβ_1–40 was found to be nontoxic.⁶ In contrast, ΔE22-Aβ_1–42 was less toxic than WT Aβ_1–42.

The presence of Glu22 point substitutions in the Aβ peptide implicates an early onset of familial Alzheimer’s disease (FAD) and cerebral amyloid angiopathy (CAA).⁷ These point mutations include the E22Q, associated with Hereditary Cerebral Hemorrhage with Amyloidosis ‘Dutch type’ (HCHWA-D);⁸ the E22G, known as the Arctic mutation;^9,10 and the E22K, known as the Italian mutation.¹¹ The Arctic mutant increased bilayer disruption due to high hydrophobicity. The Italian mutant was observed to increase the rate of aggregation¹² as did the Dutch mutant which aggregated faster than WT Aβ.¹³ These mutants formed fibrils in solution morphologically similar to those formed by WT Aβ peptides and presented polymorphic aggregates on a lipid membrane.¹² Taken together, these results emphasized the importance of the point substitution at Aβ position 22 in FAD and CAA.

Unlike point substitutions, the Osaka mutant eliminated Glu22 in patients of FAD. This recently described unusual variant of Aβ mutation prompted us to interrogate the biological properties of the mutant for its structure and function in the cell membrane. Here, we modeled octadecameric (18-mer) Aβ barrels of ΔE22-Aβ_1–42 (below we refer to “ΔE22” throughout the text) and WT Aβ_1–42 peptides in a lipid bilayer containing 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC). In our Aβ barrels, both mutant and WT peptides adopted the U-shaped motif of β-strand-turn-β-strand predicted by simulations¹⁴ and found in NMR experiments.^15,16 Overall, we observed that the ΔE22 barrels present similar morphologies and dimensions as observed for the WT Aβ_1–42 barrels, suggesting that they share features of monomer folding and aggregation into a toxic oligomer state. Due to the deletion of a charged residue, the hydrophobicity of the mutant is increased, and the increased hydrophobicity can cause faster kinetics of nucleation and membrane insertion leading to toxic channel formation. We suggest that the conformational similarity of the mutant with the WT Aβ peptide reflects functional similarity in the disease stage. As the WT Aβ peptide in the aggregation state is highly polymorphic,^17–19 the mutant would explore conformational space along the aggregation pathway of the highly polymorphic conformational states. Our simulations provide a possible conformational species of the barrels which is more highly populated by the Osaka mutant in atomic detail, to help drug discovery efforts targeting the mutant and WT Aβ peptides.

MATERIALS AND METHODS

Recruiting Aβ Monomer Conformations

Two Aβ monomer conformations with the β-strand-turn-β-strand motif (known as U-shaped or β-arcade²⁰ structures) were extracted from Aβ_1–42 fibrils, where the structure was defined by hydrogen/deuterium-exchange NMR data, side-chain packing constraints from pair-wise mutagenesis, ssNMR and EM (PDB code: 2BEG),¹⁵ and small Aβ_1–40 protofibrils (PDB codes: 2LMN and 2LMO),¹⁶ where the structure was based on the ssNMR model. In both structures, the N-terminal coordinates, residues 1–16 for the former and 1–8 for the latter structure, are missing due to disorder. We used the Aβ_1–16 coordinates, in the absence of Zn²⁺ (PDB code: 1ZE7),²¹ for the missing portions of the peptides. For each combination of the N-terminal structure with the U-shaped motifs, two Aβ_1–42 conformers were generated.^22–25 Conformer 1 has a turn at Ser26-Ile31, and conformer 2 at Asp23-Gly29. In the latter conformer, two C-terminal residues, Ile41 and Ala42 were added to create Aβ_1–42. For convenience, we divide both WT Aβ conformers into four domains: N-terminal chain (residues 1–16 and 1–8 for conformer 1 and 2, respectively), pore-lining strand (residues 17–25 and 9–22 for conformer 1 and 2, respectively), turn (residues 26–31 and 23–29 for conformer 1 and 2, respectively), and C-terminal strand (residues 32–42 and 30–42 for conformer 1 and 2, respectively).

Generating the Osaka Mutant, ΔE22, from the Aβ Conformers by the Deletion of Glu22

To create the mutant, we removed the Glu22 residue from the pore-lining β-strand. As a result, pore-lining residues 10–21 for both conformers flipped their side-chains, while the other domains kept intact. Two ΔE22 conformers (Figure 1A), each derived from the WT Aβ_1–42 conformers (Figure 1B) still retain the U-shaped structure with the β-strand-turn-β-strand motif. The 41-residue long mutant has the same turns as the corresponding WT peptides; Ser25-Ile30 for conformer 1 and Asp22-Gly28 for conformer 2. We similarly divide both mutant conformers into four domains: N-terminal chain (residues 1–15 and 1–7 for conformer 1 and 2, respectively), pore-lining strand (residues 16–24 and 8–21 for conformer 1 and 2, respectively), turn (residues 25–30 and 22–28 for conformer 1 and 2, respectively), and C-terminal strand (residues 31–41 and 29–41 for conformer 1 and 2, respectively).

Figure 1.

(A) Monomer conformations of the Osaka mutant (ΔE22) with different turns at Ser25-Ile30 (conformer 1) and Asp22-Gly28 (conformer 2), and (B) the wild type Aβ_1–42 peptides with different turns at Ser26-Ile31 (conformer 1) and Asp23-Gly29 (conformer 2). Pore-lining residues are marked with blue callouts, and especially the Glu22 residues in the wild type Aβ_1–42 peptides are marked with red callouts. In the peptide ribbon, hydrophobic, polar/Gly, positively charged, and negatively charged residues are colored white, green, blue, and red, respectively. Both termini of peptide are denoted as blue letter “N” for the N-terminus and red letter “C” for the C-terminus.

Construction of Aβ Barrels in the Lipid Bilayer

To construct the β-barrel structure, ΔE22 was inclined ~37° relative to the pore axis²⁶ and then an 18-fold rotational symmetry operation was performed with respect to the pore axis creating an 18-mer ΔE22 barrel (Figure S1A and S1B of the Supporting Information). The WT Aβ_1–42 barrel was also constructed in an analogous manner (see Figure S1C and S1D of the Supporting Information for comparison). We modeled the Aβ barrels with β-sheet structure by mimicking naturally-occurring β-barrels observed in transmembrane proteins that are found frequently in the outer membranes of bacteria, mitochondria, and chloroplasts. The β-barrel motif is a large β-sheet composed of an even number of β-strands. Some known structures of β-barrel membrane proteins have β-strands ranging in size from 8 to 22.²⁷ The 18-mer Aβ barrel has 18 β-strands enclosing the solvated pore. This number is also in the range of the number of β-strands for natural β-barrels ranging from 8 to 22, which can form a β-barrel motif. Our previous simulations for Aβ channels indicate that different numbers of Aβ monomers could produce channels with different outer and pore dimensions.^{22–26,28–32} We found that Aβ channels obtained a preferred size range of 16–24 β-strands lining the pores.^30,32 This range was also found to hold for other toxic β-sheet channels; K3 (a fragment of β₂-microglobulin) channels with 24 β-strands³³ and protegrin-1 (PG-1) channels with 16–20 β-strands.^34,35 In this work, the outer/pore diameters of the WT 18-mer Aβ barrels are in good agreement with the experimental AFM ranges.^23,25,36 The AFM experiments provide images of channels with a wide variety of sizes and shapes, but simulated Aβ barrels are limited to cover all ranges of channel sizes that are imaged by AFM.

To construct two layers of DOPC lipid in a unit cell, simple van der Waals (vdW) spheres representing lipid headgroups are placed in two parallel planes separated by the expected headgroup-to-headgroup DOPC lipid bilayer thickness.^37,38 These planes can be regarded as membrane surfaces. Dynamics is performed on the spheres, constrained to their respective planes, with the embedded barrel structure held rigid. This planar harmonic constraint ensures that the vdW spheres are randomly distributed onto the planes and well packed around the Aβ barrel. The lipid bilayer is then constructed with headgroups at the positions of the vdW spheres. The DOPC lipid molecules were randomly selected from a library of preequilibrated liquid crystalline state lipids. After replacement of the vdW spheres with lipid molecules, a series of minimizations is performed to remove overlaps of the alkane chains and gradually relax the system. For DOPC, the cross-section area per lipid is 72.4 Ų and headgroup distance across the bilayer is 36.7 Å at 30 °C.³⁹ With a choice for the number of lipid molecules, the optimal value of lateral cell dimensions can be determined. The DOPC bilayer containing 420 lipids constitutes the unit cell with TIP3P water, added at both sides with lipid/water ratio of ~1/110. Updated CHARMM⁴⁰ all-atom additive force field for lipids (C36)⁴¹ and the modified TIP3P water model⁴² were used to construct the set of starting points and to relax the systems to a production-ready stage. The system contains Mg²⁺, K⁺, Ca²⁺, and Zn²⁺ at the same concentration of 25 mM to satisfy a total cation concentration near 100 mM. The bilayer system containing an Aβ barrel, lipids, salts, and water has almost 210,000 atoms.

MD simulations employed the zwitterionic DOPC bilayer. All Aβ barrels were pre-assembled and simulated in the lipid bilayer. It has been known that anionic bilayers with negatively charged surfaces facilitate the interactions with Aβ peptides.⁴³ However, once the peptides are inserted into the membrane core and subsequently assembled to form a channel, the hydrophobic interactions between lipid-facing residues in the channel and lipid tails should be an important factor in stabilizing the channel conformation. In our previous simulations, we have employed anionic lipid bilayer composed of DOPS/POPE in order to complement the electrophysiological recoding studies.^22–25 The DOPS/POPE phospholipid combinations can form stable lipid bilayers, and allow conductance measurements in PLB experiments. However, we observed that there are no significant differences in the critical results of subunits formation in the channel conformation in zwitterionic and anionic bilayers.

Production Runs

We generated at least 10 different initial configurations for each Aβ barrel for the relaxation process in order to obtain the best initial configuration toward a starting point. In the pre-equilibrium stages, a series of minimizations was performed for the initial configurations to remove overlaps of the alkane chains in the lipids and to gradually relax the solvents around the harmonically restrained peptides. The initial configurations were gradually relaxed through dynamic cycles with electrostatic cutoffs (12 Å). The harmonic restraints were gradually diminished with the full Ewald electrostatics calculation and constant temperature (Nosé–Hoover) thermostat/barostat at 303 K. For t < 30 ns, our simulation employed the NPAT (constant number of atoms, pressure, surface area, and temperature) ensemble with a constant normal pressure applied in the direction perpendicular to the membrane. After t = 30 ns, the simulations employed the NPT ensemble. Production runs of 100 ns for the starting points with the NAMD code⁴⁴ were performed on a Biowulf cluster at the NIH. Averages were taken after 30 ns, discarding initial transients. Analysis was performed with the CHARMM programming package.⁴⁰

RESULTS

The Osaka Mutant (ΔE22) Adopts the β-Barrel Topology in the Lipid Bilayer

We performed 100 ns all-atoms molecular dynamics (MD) simulations on Aβ barrels, constructed using the Osaka mutant with complete elimination of Glu22 and WT Aβ_1–42 peptides, embedded in a DOPC bilayer. Both ΔE22 and WT Aβ barrels comprising two different U-shaped conformers were initially pre-assembled as an annular shape. During the course of the simulations, no immediate peptide dissociation in the barrels was observed, but the peptides were gradually relaxed in the lipid bilayer (Figure S2A, B of the Supporting Information). Previous Aβ channel simulations showed that amyloid ion channels composed of truncated (Aβ_17–42 (p3), Aβ_9–42 (N9), and p3-F19P)^26,28–32 and full-length (L-Aβ_1–42, D-Aβ_1–42, F19P, and F20C)22–25 Aβ peptides/mutants are heterogeneous compared to functional gated ion channels. In our current simulations, the mutant also presents heterogeneity in barrel conformations as observed in the WT barrels (Figure 2). Non-homogeneous peptide interactions with the surrounding environments, as evident by the independent peptide fluctuations, induce the heterogeneity in the mutant and WT barrel conformations (Figure S2C-F of the Supporting Information). In Table 1, the calculated outer dimensions and pore sizes are summarized for the 18-mer mutant and WT Aβ barrels. The outer/pore diameters are ~7.91/~2.1 and ~7.62/~1.52 nm for the conformer 1 and 2 ΔE22 barrels, and ~7.87/~1.98 and ~7.97/~2.18 nm for the conformer 1 and 2 Aβ_1–42 barrels, respectively. Although the conformer 2 ΔE22 barrel presents slightly reduced outer diameter and significantly reduced pore size, those dimensions for the conformer 1 ΔE22 barrel are in the range of the WT barrel.

Figure 2.

Averaged pore structures calculated with HOLE⁵⁰ embedded in the average barrel conformations during the simulations for the (A) conformer 1 and (B) conformer 2 Osaka mutant (ΔE22) barrels, and the (C) conformer 1 and (D) conformer 2 Aβ_1–42 barrels. In the barrel structures with the ribbon representation, hydrophobic, polar/Gly, positively charged, and negatively charged residues are colored white, green, blue, and red, respectively. For the pore structures in the surface representation, red denotes pore diameter of d < 1.4 nm, green denotes pore diameter in the range, 1.4 nm = d = 2.0 nm, and blue denotes pore diameter of d > 2.0 nm.

Table 1.

The calculated outer and pore diameters, d_o and d_p, with a description of the pore status, % of β-sheet secondary structure in the pore/C-terminal strands, and “straightness” of the strands by the β-strand order parameter using $S_{\beta }=(1/N_v){\sum }_{k=1}^{N_v}((3{cos}^2{\theta }_{\alpha }-1)/2)$, where θ_α is the angle between the positional vectors connecting two C_α atoms and N_v is the total number of vector pairs for the residues in the pore/C-terminal strands for the Osaka mutant (ΔE22) and wild type Aβ_1–42 barrels.

Aβ barrels	do (nm)	dp (nm)*	Pore status	% of β-sheet pore/C-term	Sβ pore/C-term
Conformer 1 Osaka mutant (ΔE22)	~7.91	~2.10	Opened	41.6/2.4	0.58/0.45
Conformer 2 Osaka mutant (ΔE22)	~7.62	~1.52	Reduced	46.6/5.1	0.57/0.43
Conformer 1 Aβ1–42 (wild type)	~7.87	~1.98	Opened	46.0/3.8	0.56/0.45
Conformer 2 Aβ1–42 (wild type)	~7.97	~2.18	Opened	50.4/5.8	0.57/0.45

^*Pore diameters are averaged along the pore axis within the cutoffs defined by the height of pore.

When assembled in a barrel, the U-shaped peptides form two layered annular β-sheets. The pore-lining strands form a β-sheet with intermolecular backbone hydrogen bonds (H-bonds), tightly encompassing a solvated pore, as evident by the secondary structure analysis (Figure 3). However, the outer β-sheet is not so preserved because of the longer H-bonding pair distance between the C-terminal strands due to the larger curvature. In contrast to the pore-lining strands with high β-sheet content, the C-terminal strands have low β-sheet content, suggesting that the evolution of secondary structure is rather dynamic. For the same conformers, the ΔE22 barrels exhibit slightly lower percentage of the β-sheet secondary structure for the membrane embedded pore and the C-terminal strands than the Aβ_1–42 barrels (Table 1). However, these differences are very subtle. In addition, both mutant and WT barrels have similar β-strand order parameters for the membrane embedded portions, suggesting analogous barrel conformations in the lipid bilayer.

Figure 3.

The description of secondary structure by STRIDE⁵¹ averaged during the simulations for the (A) conformer 1 and (B) conformer 2 Osaka mutant (ΔE22) barrels, and the (C) conformer 1 and (D) conformer 2 Aβ_1–42 barrels.

Interactions of the Osaka Mutant (ΔE22) in the Lipid Environment

To observe how the deletion of Glu22 affects peptide interactions in such complex lipid environments, we calculated the interaction energy for each peptide with the lipids, water, and other peptides, and then averaged over the number of peptides in the barrel, yielding per peptide interaction energy (Figure S3 of the Supporting Information). For comparison, in the calculation only membrane embedded portions of the peptides, i.e. residues 17–41 and 9–41 for the conformer 1 and 2 ΔE22, respectively, and residues 17–42 and 9–42 for the conformer 1 and 2 Aβ_1–42 peptides, respectively, were considered. These portions of the peptides represent the intrinsic U-shape obtained from the experimental coordinates.^15,16 Figure 4 shows averaged peptide interaction energy obtained from the per peptide interaction energy averaged over time. In the comparison of the same conformers between mutant and WT barrels no apparent energy difference from the mutation in the peptide/lipid interaction can be observed, since the lipid interaction energy of the mutant is similar to that of the WT peptide. This suggests that the mutation site is located at the solvated pore, far from lipids, and each conformer mutant has the lipid-contacting C-terminal strand sequentially identical to the corresponding conformer WT peptide (Figure 1). However, energy differences from the mutation in the peptide/water and peptide/peptide interactions are immediately apparent. Comparison of corresponding conformers of mutant and WT Aβ_1–42 barrels indicates that each conformer ΔE22 barrel exhibits relatively weaker peptide/water (mainly pore water) and relatively stronger peptide/peptide interactions. The energy contributions of the mutation in the peptide/water interaction are ~−44.7 and ~−119.1 kcal/mol per peptide for the conformer 1 and 2 ΔE22 barrels, respectively. In the peptide/peptide interaction, they are ~83.8 and ~82.5 kcal/mol per peptide for the conformer 1 and 2 ΔE22 barrels, respectively. The negative sign in the energy contribution denotes “loss” of attraction, while the positive sign corresponds to “gain” of attraction. The conformer 2 ΔE22 loses attraction more than the conformer 1 ΔE22 in the peptide/water interaction, suggesting that the conformer 2 ΔE22 barrel slightly reduces the size of the water pore (Table 1). However, both conformer mutants similarly gain attraction in the peptide/peptide interaction, since the repulsive force between the negatively charged Glu22 side-chains in the water pore disappears from the peptide/peptide interaction upon deletion of the residue.

Figure 4.

Averaged peptide interaction energy for the conformer 1 and 2 Osaka mutant (ΔE22) barrels (denoted as D1 and D2, respectively), and the conformer 1 and 2 wild type Aβ_1–42 barrels (denoted as W1 and W2, respectively).

Ion Activities in the Osaka Mutant (ΔE22) Pore

In the pores of WT Aβ channel/barrel, the negatively charged Glu22 side-chains can attract cations into the pore, forming a cationic ring.^26,28,29 However, this is not the case for the ΔE22 barrel in the absence of the Glu22 residue. To locate the ionic binding sites in the mutant barrel, we calculated three-dimensional (3D) density maps of ions around Aβ barrels (Figure 5). Highly populated ionic binding sites are enclosed by 3D meshes with the colors for Mg²⁺ (green), K⁺ (red), Ca²⁺ (blue), Zn²⁺ (cyan), and Cl⁻ (gray). As expected, in the ΔE22 barrel pores the cationic ring at the vestige of Glu22 binding site is invisible. In contrast to the mutant barrels, Aβ_1–42 barrels present a cationic binding site at a cluster of Glu22 side-chains. In addition to the cationic binding site, the WT pores provide an anionic binding site at a cluster of Lys16 side-chains. This is not the case for the mutant barrels, since the Lys16 side-chains are located behind the solvated pore (Figure 1). However, in the pores of ΔE22 barrels, cations strongly interact with the Glu11 side-chains at the lower bilayer leaflet, suggesting that the side-chains serve as an emerging cationic binding site.

Figure 5.

Three-dimensional density map of Mg²⁺ (green mesh), K⁺ (red mesh), Ca²⁺ (blue mesh), Zn²⁺ (cyan mesh), and Cl⁻ (gray mesh) for the (A) conformer 1 and (B) conformer 2 Osaka mutant (ΔE22) barrels, and the (C) conformer 1 and (D) conformer 2 Aβ_1–42 barrels. Averaged channel structure is shown as the ribbon and transparent surface representations in gray. Density map indicates populated interaction sites for the ions, each with the same probability of 0.01 for cations and the probability of 0.03 for Cl⁻.

To obtain the probability distribution for ions across the bilayer, the 3D density maps of ions are projected onto a plane (Figure 6). The highly populated ion binding sites are reflected in peaks in the probability distribution curves. In the simulations, we observed that the ions mainly bind to the charged side-chains of the peptide with a strong electrostatic attraction. In particular, the cations also interact with a phosphate group in the lipid head and the C-terminus of peptide, reflecting that peaks at z = ~±2.0 nm represent those cations interacting with lipid head groups. For the Aβ_1–42 barrels, peaks at z = ~1.0 and ~−1.2 nm for the conformer 1, and at z = ~1.8 and ~−0.2 nm for the conformer 2 correspond to the Glu22 cationic and Lys16 anionic binding sites in the pore, respectively. However, for both conformer ΔE22 barrels, those peaks are absent from the probability distributions for ions in the pore. Instead, peaks at z = ~−2.9 and ~−1.4 nm emerge at the lower bilayer leaflet for the conformer 1 and 2 ΔE22 barrels, respectively, representing a highly populated cation binding site at a cluster of Glu11 side-chains.

Figure 6.

Probability distribution functions for Mg²⁺ (green line), K⁺ (red line), Ca²⁺ (blue line), Zn²⁺ (cyan line), Cl⁻ (gray line), and water (black dashed line) as a function of the distance along the pore center axis for the (A) conformer 1 and (B) conformer 2 Osaka mutant (ΔE22) barrels, and the (C) conformer 1 and (D) conformer 2 Aβ_1–42 barrels.

The locations of the peak in the 2D ion distribution curves are qualitatively consistent with the charged side-chain locations across the bilayer. To locate the charged groups in the Aβ barrels, the probability distributions for few selected charged groups in the peptides are calculated (Figure 7). The peaks in the distribution curves reflect the highly populated locations of the charged groups across the bilayer. During the simulations, we observed that four charged side-chains, Asp1, Glu3, Arg5, and Asp7 in the N-terminus of the peptides are mainly located in the bulky water area below the lower bilayer leaflet. Although these residues attract ions, they do not participate in the pore formation. In the ΔE22 barrels, the Glu11 side-chains are located at z = ~−2.9 and ~−1.4 nm for the conformer 1 and 2 barrels, respectively, attracting cations. In the Aβ_1–42 barrels, they are located at z = ~−2.0 and ~−1.5 nm for the conformer 1 and 2 barrels, respectively. It can be seen that Glu11 side-chain locations are highly correlated with the peaks in Figure 6 for the highly populated cationic binding sites across the bilayer. In the mutant barrels, the Glu11 side-chain points toward the water pore, while in the WT barrels it is directed away from the pore (see the peptide topology in Figure 1). This indicates that in the mutant barrels the Glu11 side-chains circularly cluster to form a negatively charged ring with cations screening, serving as a cationic binding site. The circular clustering by the Glu11 side-chains at the lower bilayer leaflet further encompasses the solvated pore, causing an extension of the pore. Thus, the lengths of both conformer mutant pores are relatively longer than the corresponding conformer WT pores (Figure 2). We observed that several mutant monomers participate in the β-sheet formation (Figure 3). At the anionic biding site, the positively charged Lys16 side-chains can attract Cl⁻ in the pore. In the mutant barrels, the Lys16 side-chains are located at z = ~−1.0 and ~0.5 nm for the conformer 1 and 2 barrels, respectively. In the WT barrels, they are located at z = ~−1.5 and ~−0.2 nm for the conformer 1 and 2 barrels, respectively. While WT pores provide the anionic binding site, it is absent from the mutant pores since the Lys16 side-chains are not present in the water pore. Thus, for the mutant barrels, the associated peak representing the anionic binding site in Figure 6 is missing, since the Lys16 side-chains are exposed in the hydrophobic core between the β-sheets. In contrast, in the WT barrel, the Lys16 side-chains point toward the water pore, forming the anionic binding site.

Figure 7.

Probability distribution functions for selected charged residues, Asp1 (red), Glu3 (green), Arg5 (yellow), Asp7 (blue), Glu11 (pink), Lys16 (cyan), and Glu22 (gray), and for the phosphate group of lipid head, PO₄ (dark red), as a function of the distance along the pore center axis for the (A) conformer 1 and (B) conformer 2 Osaka mutant (ΔE22) barrels, and the (C) conformer 1 and (D) conformer 2 Aβ_1–42 barrels.

Both ΔE22 and Aβ_1–42 barrels present the solvated pore, wide enough for the ion conductance. To observe ion fluctuation across the pore, we calculated the change in the total charge in the pore as a function of the simulation time. In the calculation, a pore height cutoff along the pore axis is |z| < 1.5 nm was used (Figure 8). A larger pore height cutoff, |z| < 1.8 nm, was also considered (Figure S4 of the Supporting Information). The pore height cutoff with |z| < 2.0 nm ensures the charge fluctuations including only a contribution from the ions in the middle of the pore. For both conformer mutant barrels, we observed that charge fluctuations in the total charges are similar to those for the corresponding conformer WT barrels. The charge fluctuations with the larger pore height cutoff also show a similar pattern, suggesting that the ΔE22 barrel is ion permeable in the membrane.

Figure 8.

Change in total charge in the pore as a function of the simulation time for the conformer 1 and 2 Osaka mutant (ΔE22) barrels (1st and 2nd rows), and the conformer 1 and 2 Aβ_1–42 barrels (3rd and 4th rows). The pore height with cutoff along the pore axis, −1.5 < z < 1.5 nm was used.

DISCUSSION

We performed explicit molecular dynamics (MD) simulations of octadecamer (18-mer) Osaka mutant (ΔE22) and wild type (WT) Aβ_1–42 barrels in a DOPC bilayer. The monomer mutant conformation was derived from the Aβ_1–42 peptide via a deletion of Glu22. Two U-shaped mutant conformers, with a turn at Ser25-Ile30 for conformer 1 and a turn at Asp22-Gly28 for conformer 2, inherited the same turn conformations from the Aβ_1–42 conformers with slightly different turns, Ser26-Ile31 in conformer 1 and Asp23-Gly29 in conformer 2. However, the pore-lining residues invert their side-chain orientation due to the deletion of the residue. Here we show that the ΔE22 mutant forms a β-barrel-like channel in the lipid bilayer, with the membrane embedded conformation indistinguishable from the WT Aβ_1–42 barrel. For the conformer 1 ΔE22 barrel, we obtained outer and pore diameters, ~7.91 and ~2.1 nm, respectively, indicating that the overall dimensions are consistent with the Aβ_1–42 barrels in this study. Further, these dimensions are also in the range of the outer diameter, ~7.8 – 8.3 nm, and pore diameter, ~1.8 – 2.2 nm, for the same 18-mer Aβ_1–42 barrels in the anionic bilayer composed of DOPS/POPE.^22–25 However, for the conformer 2 ΔE22 barrel, we obtained outer and pore diameters, ~7.62 and ~1.52 nm, respectively, with the overall dimensions slightly reduced as compared to the Aβ_1–42 barrels. The decrease in the pore diameter resulted from the inverted Lys16 side-chains that unfavorably reside in the central hydrophobic core between the β-sheets (z = ~0.5 nm). As a result, the unfavorable force exerted on the backbones of pore-lining residues slightly relocates them toward the pore axis. This causes reduction in the pore diameter and hence in the outer diameter of the barrel as well. In contrast, the conformer 1 ΔE22 barrel could preserve the overall dimensions consistent with the WT barrels, since the Lys16 side-chains are located very close to the amphipathic interface of the lipid bilayer at the lower bilayer leaflet (z = ~−1.0 nm), stretching to interact with the C-termini or phosphate groups of lipids. The heterogeneity in the dimension of ΔE22 barrels suggests that ΔE22 may be less toxic than WT Aβ_1–42, since the mutant barrels with smaller pores would be populated. Recently, it was shown that ΔE22 exhibits less neurotoxicity than WT Aβ_1–42 in rat primary neurons.⁶

It is interesting to note the implications for WT Aβ from the Osaka mutant barrels. The coordinates for conformer 1 Aβ_1–42 were directly extracted from the NMR-based Aβ_1–42 fibrils (PDB code: 2BEG).¹⁵ For the conformer 2 Aβ_1–42, we obtained the coordinates from the Aβ_1–40 protofibrils (PDB codes: 2LMN and 2LMO),¹⁶ and then added two C-terminal residues, Ile41 and Ala42, generating Aβ_1–42. Although both conformers are the same U-shaped peptide with the β-strand-turn-β-strand motif, they can be distinguished by their different turns (Figure 1). This suggests that the turn at Asp23-Gly29 of the conformer 2 should belong to an intrinsic turn of Aβ_1–40. A similar turn at Val24-Ala30 for Aβ_1–40 was recently identified from a structural model of Aβ_1–40 fibrils also using comprehensive ssNMR techniques.⁴⁵ Thus, the conformer 2 Aβ_1–42 adopting the Aβ_1–40 turn topology is likely to be a relatively less populated conformation than the conformer 1 Aβ_1–42, suggesting an explanation for the ssNMR observation for this turn for the Aβ_1–40 peptide¹⁶ rather than the conformation of the Aβ_1–42.¹⁵ If the WT Aβ_1–40 were to adopt the turn conformation of Aβ_1–42, then the C-terminal strand will be too short to create a stable U-shaped conformation. Thus, while the less C-terminal turn conformation of Aβ_1–42 is more stable, the resulting shorter strand for the 1–40 sequence shifts the equilibrium toward the more C-terminal turn conformation. Recently, it was demonstrated that the C-terminal domain of Aβ_1–42 showed a distinct conformational dynamics from that of Aβ_1–40, which suggests that the Val36-Cly37 turn is the sine qua non of Aβ_1–42.46 For the Osaka mutant the conformer with the less C-terminal turn is still more populated, since the shortened sequence with the Glu22 deletion affects the N-terminal strand, rather than the C-terminal, thus retaining an equilibrium favoring the Aβ_1–42 turn.

The negatively charged Glu22 is implicated as a key site for point mutation, since amino acid substitutions at position 22 including Dutch (E22Q),⁸ Arctic (E22G),^9,10 and Italian (E22K)¹¹ mutants. While the Dutch mutation is closely associated with cerebral amyloid angiopathy (CAA),⁷ other mutants with a familial Alzheimer’s disease (FAD) linked point substitution at Glu22 are toxic species,¹³ suggesting that Glu22 plays a significant role in the pathogenesis of AD. However, in spite of the complete elimination of Glu22, rather than an amino acid substitution, Aβ peptide lacking Glu22 is still linked to FAD. Previously, we evaluated that in the solvated pore the charged side-chain of Glu22 plays an important role in conducting ions in the amyloid ion channel models. Our evaluation was supported by the observation that a circular cluster formed by the negatively charged Glu22 side-chains attracts cations into the pore and serves as a cationic binding site.^26,28,29 The larger the ions population at the binding site, the higher the probability for ions to conduct through the water pore. However, in the absence of Glu22, the ΔE22 barrels still attract cations into the pore and show large charge fluctuations in the pore similar to the WT barrels. Although the ΔE22 barrels lose the negatively charged Glu22 side-chains at the upper bilayer leaflet, they gain the Glu11 side-chains at the lower bilayer leaflet. Thus, the emerging Glu11 side-chains attract cations into the pore and contribute to pore elongating with the β-sheet formation. To evaluate the biological role of a charged side-chain in the pore, we suggest that a circular cluster of charged side-chains can be formed with a help of ion screening. This minimizes the charge repulsions between the ions, with the circular assembly of side-chains strongly sustaining the backbone β-sheet formed by the pore-lining residues. With well-established pore, which is wide enough, ions can move freely through the water pore.

Even though the Osaka mutant that we have modeled also appears to adopt a toxic channel conformation consisting of a barrel organization of the U-shape motif, this does not necessarily imply that such conformational species are always the preferred conformational states. As we have already emphasized the amyloid landscape is highly heterogeneous^17–19 and different conformations may be populated, including in membrane-permeated channels, suggesting that highly polymorphic conformations of Aβ channel could evolve from different seed formations.^47–49 It is a challenge to infer all possible highly populated states for different mutants, under different conditions and membrane compositions.

To conclude, our MD simulations provide a membrane-bound conformation of the Osaka mutant barrel in atomic-level detail, illustrating that the multimeric β-barrel-like channel can be indistinguishable from the WT Aβ_1–42 barrel. The U-shaped peptide with the β-strand-turn-β-strand motif supports the Osaka mutant barrel, suggesting the universality of the Aβ motif in aggregation. It has been known that due to the loss of charge, the mutant has higher hydrophobicity resulting in faster oligomerization and fibril formation.^4–6 We speculate that high production of mutant oligomers can lead to toxic channel formation followed by oligomers insertion into the cell membrane in FAD. The mutant conformational dynamics along with the membrane insertion and channel formation have not been well elucidated. The conformational space of Aβ monomer, oligomer, and membrane-embedded channel states is highly polymorphic,^17–19 with the mutant sharing these free energy landscapes, however with varied conformational preferences. Of particular interest, this deletion mutant suggests an explanation why the solid state NMR data for Aβ_1–40¹⁶ presented a more C-terminal turn conformation versus Aβ_1–42,¹⁵ and argues that the more toxic Aβ_1–42 species would preferentially populate the less C-terminal turn. Those two extra residues at the C-terminus shift the free energy landscape toward the more stable Lührs et al.¹⁵ conformation. Aβ_1–42 lacking Glu22 still retain these conformational preferences, since the deletion affects the conformation of the N-terminal, rather than the C-terminal strand.