Atomistic-level study of the interactions between hIAPP protofibrils and membranes: Influence of pH and lipid composition

Abstract

The pathology of type 2 diabetes mellitus is associated with the aggregation of human islet amyloid polypeptide (hIAPP) and aggregation-mediated membrane disruption. The interactions of hIAPP aggregates with lipid membrane, as well as the effects of pH and lipid composition at the atomic level, remain elusive. Herein, using molecular dynamics simulations, we investigate the interactions of hIAPP protofibrillar oligomers with lipids, and the membrane perturbation that they induce, when they are partially inserted in an anionic dipalmitoyl-phosphatidylglycerol (DPPG) membrane or a mixed dipalmitoyl-phosphatidylcholine (DPPC)/DPPG (7:3) lipid bilayer under acidic/neutral pH conditions. We observed that the tilt angles and insertion depths of the hIAPP protofibril are strongly correlated with the pH and lipid composition. At neutral pH, the tilt angle and insertion depth of hIAPP protofibrils at a DPPG bilayer reach ~52° and ~1.62 nm with respect to the membrane surface, while they become ~77° and ~1.75 nm at a mixed DPPC/DPPG membrane. The calculated tilt angle of hIAPP at DPPG membrane is consistent with a recent chiral sum frequency generation spectroscopic study. The acidic pH induces a smaller tilt angle of ~40° and a shallower insertion depth (~1.24 nm) of hIAPP at the DPPG membrane surface, mainly due to protonation of His18 near the turn region. These differences mainly result from a combination of distinct electrostatic, van der Waals, hydrogen bonding and salt-bridge interactions between hIAPP and lipid bilayers. The hIAPP-membrane interaction energy analysis reveals that besides charged residues K1, R11 and H18, aromatic residues Phe15 and Phe23 also exhibit strong interactions with lipid bilayers, revealing the crucial role of aromatic residues in stabilizing the membrane-bound hIAPP protofibrils. hIAPP-membrane interactions disturb the lipid ordering and the local bilayer thickness around the peptides. Our results provide atomic-level information of membrane interaction of hIAPP protofibrils, revealing pH-dependent and membrane-modulated hIAPP aggregation at the early stage.

1. Introduction

The formation of fibrillar amyloid deposits is implicated in many human diseases, including Alzheimer’s disease, Parkinson’s disease and type 2 diabetes (T2D) mellitus [1,2]. These deposits are observed in over 90% of T2D patients, and their cytotoxicity is closely related to the membrane damage of insulin-producing β-cells [3,4]. The primary component of islet amyloid and actual fibril-forming molecule is human islet amyloid polypeptide (hIAPP, or amylin). hIAPP is a polypeptide co-secreted with insulin by pancreatic β-cells, and it can form β-sheet-rich aggregates of varied sizes and structures [3]. Increasing evidence suggests the early intermediates of fibrillation, such as oligomers and protofibrils, contribute to trigger pathological processes [3,5,6].

Monomeric hIAPP mainly adopts a random coil structure in aqueous solutions, with residues 8–19 displaying transient α-helical conformations [7–9]. Its aggregation follows a nucleation-dependent pathway as other amyloidogenic proteins, which can be dramatically accelerated in the presence of lipid membranes [10]. It was also reported that in the presence of model membranes, the α-helical transient intermediates were observed in the early stage of membrane-hIAPP interaction [11]. Intermediate oligomers can disrupt membrane integrity and permeability, lead to ionic homeostasis and signal disturbance, and thereby cause cell death [4,6,12–14]. Earlier electron paramagnetic resonance (EPR) spectroscopy and nuclear magnetic resonance (NMR) studies showed that a membrane-bound hIAPP monomer adopts pH- and membrane composition-dependent helical structure in one (residues 9–22) or two regions (residues 7–17 and 21–28) [15,16]. Upon enrichment of membrane-adsorbed peptides and subsequent formation of toxic oligomers, hIAPPs cooperatively transform into β-sheet aggregates [17–19]. Membrane-bound hIAPP oligomers not only promote amyloid formation but also induce membrane leakage [20–22]. However, it remains challenging to experimentally detect how hIAPP oligomers interact with membranes.

hIAPP is a 37 amino acid peptide, its N-terminal 1–19 region involves membrane binding, the amyloidogenic 20–29 fragment governs fibril formation, and the C-terminal region involves peptide self-association [23]. Aromatic stacking interactions play an important role in driving the self-assembly of hIAPP [24–26]. Histidine 18 is responsible for the pH-dependence of the fibrillation and membrane interactions of hIAPP [27,28]. His18 modulates the HiAPP_1–19 orientation in anionic palmitoyloleoyl-phosphatidylglycerol (POPG) liposomes, and has a reduced toxicity via membrane disruption at pH 6.0 [27]. The kinetics of fibril formation and membrane damage of full-length hIAPP was observed to be much slower at pH 5.5 than at pH 7.4, which is closely related to His18 protonation state [28]. The aggregation of membrane-bound hIAPP is also affected by lipid composition, and can be significantly accelerated in the presence of negatively charged lipids, such as phosphatidylglycerol (PG) or phosphatidylserine (PS) [3,29,30]. These studies reflect the important role of electrostatic interaction in hIAPP-membrane interaction.

All-atom molecular dynamics (MD) simulations can provide detailed information of the interactions between amyloid proteins and membranes. Increasing number of research groups have examined the membrane interaction of monomeric/oligomeric amyloidosis-related proteins, including β-amyloid (Aβ), α-synuclein, hIAPP, and huntingtin, etc [31–44]. For example, Jang et al studied the structural properties of β-sheet-rich Aβ transmembrane channels and they observed in their MD simulations that preformed Aβ channels in lipid bilayers undergo a dynamic process comprised of subunit association, dissociation, and channel rearrangements [31]. Pannuzzo et al explored Aβ oligomerization in POPC membranes using MD simulations and molecular modeling, and reported that the insertion of Aβ into lipid membranes is related to the presence of high energy “frustrated helices” which could lead to the formation of highly dynamic ion channels [32]. Tsigelny et al. reported that α-synuclein can penetrate the membrane rapidly with a helix-coil transition, and the penetration of annular α-synuclein oligomers may form pore-like structure [33]. Zhao et al explored the dynamic properties of constructed hIAPP channels and they found that annular-like hIAPP channels with different sizes and topologies lose the initial continuous β-sheet network and break into oligomeric subunits [34]. They also investigated the property of double hIAPP channels in the DOPC bilayer using MD simulations and experiments, and reported the non-selective ion channel activity of hIAPP double channels [35]. Christensen et al examined the formation process of trimers and tetramers using a highly mobile membrane mimetic model and their simulations showed that initially membrane-bound α-helical hIAPPs can self-assemble into β-sheet structures [42]. Brown et al carried out MD simulations of disordered Aβ1–42 tetramers on zwitterionic lipid bilayers and they observed that the binding of Aβ tetramers on POPC bilayers resulted in a greater membrane perturbation than that on cholesterol-rich membranes [43]. Recently, Dong et al investigated systemically the adsorption dynamics, structural stability and membrane perturbation of protofibrillar Aβ trimers constructed using three different NMR-derived fibril structures, 2BEG, 2LMN and 2M4J. Their simulations showed that regardless of the morphologies and the initial orientations of the three different protofibrillar Aβ trimers, the N-terminal β-sheet of all trimers preferentially binds to the membrane surface [44].

The adsorption of hIAPP aggregates onto anionic PG membranes was monitored by means of chiral sum frequency generation (SFG) spectroscopy, and the β-sheet aggregates were found to orient at 48° relative to the interface [45]. Inspired by the experimental study, Poojari et al. simulated the stability and orientation of membrane-embedded hIAPP trimers and tetramers [46]. Their 150-ns molecular dynamics simulations showed that these lipid bilayer-inserted hIAPP oligomers may cause water permeation and Na⁺ intrusion. Very recently, Zhang et al. examined the adsorption of hIAPP oligomers onto zwitterionic lipid bilayers by MD simulations [47]. They found that hIAPP has stronger interactions with mixed palmitoyloleoyl-phosphatidylcholine (POPC)/palmitoyloleoyl-phosphatidylethanolamine (POPE) lipids than with pure POPC lipids, without thinning or curving the bilayers during 200-ns simulations. Our previous studies examined the structure and dynamics of hIAPP monomer and dimer at anionic bilayers [48,49]. However, the interactions between hIAPP protofibrillar oligomers (protofibrils) and anionic membranes with different lipid compositions under different pH condition remain elusive. Here we investigated the conformational dynamics and insertion depth of hIAPP protofibrils in anionic dipalmitoyl-phosphatidylglycerol (DPPG) membrane by performing multiple independent 300-ns MD simulations. We also examined the influence of pH and lipid compositions to gain more insight into membrane interactions of hIAPP protofibrils.

2. Materials and methods

2.1. hIAPP protofibrils and lipid bilayers

The hIAPP protofibrillar pentamer was modeled on the basis of a previous solid-state NMR study [50]. The amino acid sequence of hIAPP is 1-KCNTATCATQRLANFLVHSSNNFGAILSSTNVGSNTY-37. Each hIAPP peptide chain in the protofibril is composed of a disordered N-terminal loop with the Cys2 and Cys7 forming a disulfide bond, two β-strand regions (Ala8-Val17 and Ser28–Tyr37), and a turn region of His18-Leu27 (Fig. 1(a)). The N-terminus was charged (NH₃⁺) and the C-terminus was amidated to mimic experimental conditions. At neutral pH, the side chains of Lys1 and Arg11 are positively charged; at acidic pH, the side chain of His18 is also positively charged.

Fig. 1.

The structures of a peptide chain in the hIAPP protofibril (a), a DPPG and a DPPC lipid molecule (b), and initial states of three simulated systems: hIAPP-DPPG(N), hIAPP-DPPG(A) and hIAPP-DPPC/DPPG(N) (c), where N and A in the parentheses denote neutral and acidic pH, respectively. Color codes: choline/glycerol group (black), phosphate group (red), ester group (green), and other carbon atoms (blue) in a lipid; residues Lys1 (red), Arg11 (orange), positively charged His18 (yellow), hydrophilic residues (cyan), and hydrophobic residues (gray) in the hIAPP protofibril; DPPG lipids (silver), DPPC lipids (green) in membrane bilayer, with phosphorus atoms in spheres. The atom index of acyl chain 1 (sn-1) in DPPG and DPPC lipids is also labeled.

The DPPG bilayer provides an ideal model to study the mechanism for the binding of hIAPP at anionic membrane and the membrane-mediated hIAPP aggregation, and it has been widely used in previous studies of hIAPP–membrane interaction [18,19,46]. To examine the influence of lipid composition, zwitterionic phosphatidylcholine (PC), the most abundant phospholipids in pancreatic islets [51], was introduced and mixed with DPPG lipids at a molar ratio of ~7:3 to mimic the membrane of pancreatic islet cells [52]. The chemical structures of a DPPG and a DPPC bilayer are shown in Fig. 1(b). A pure 2×64 DPPG bilayer and a mixed 2×64 DPPC/DPPG (90:38 in lipid number) bilayer were constructed based on a previous work [53], and they were well equilibrated in a 100-ns MD simulation at 323 K, above the gel-liquid crystal phase transition temperature of DPPG and DPPC lipid bilayers [54,55]. The molar ratio of peptide:lipid in our simulations is close to a previous experiment in which the ratio is 1:20 [28]. Na⁺ and Cl⁻ ions were added to neutralize the systems and a salt concentration of 0.1 M.

The hIAPP protofibril was pre-inserted partially into the membranes by the program INFLATEGRO [56], with the backbone perpendicular to the bilayer surface. The turn region of hIAPP protofibril was located at the half depth of the upper leaflet so that hIAPP may choose to insert deeper if it favors the lipid environment or move backward to the membrane surface if it favors the water solution (Fig. 1(c)). This initial state allows the N-terminal residues to be exposed to the water environment and core hydrophobic residues 20–27 to be located in the lipid phase, as suggested in previous studies [57,58]. The systems of hIAPP protofibril with a DPPG bilayer at neutral pH, hIAPP protofibril with DPPG bilayer at acidic pH, and hIAPP protofibril with mixed DPPC/DPPG (7:3) bilayer at neutral pH, are labeled as hIAPP-DPPG(N), hIAPP-DPPG(A) and hIAPP-DPPC/DPPG(N), respectively. Figure 1 displays the structures of a peptide chain in hIAPP protofibril, a DPPG/DPPC lipid molecule, and the initial states of three systems. The average z-coordinate of the phosphorus atoms in the upper leaflet is defined as z=0 (corresponding to the membrane surface), and the z-coordinate of the bilayer center is at about −2 nm. The β-strand vector (Cα-Cα vector from T9 to V17 residues) in each peptide chain is used to define the protofibril orientation, and the initial tilt angle relative to the membrane surface is ~90° (Fig. 1(c)).

2.2. Simulation details

Atomistic MD simulations were performed in the isothermal-isobaric (NPT) ensemble using the GROMACS 4.5.3 software package [59]. The peptides and lipids were respectively described with the GROMOS [60] and modified Berger force fields [61] with an adaption of Tieleman’s lipid parameters. The systems were solvated with SPC water molecules. The integration time step is 2 fs. The pressure is kept at 1 bar using the semi-isotropic Parrinello-Rahman’s method [62,63] with a coupling constant of 1.0 ps and a compressibility of 4.5×10–5 bar⁻¹. The temperature is maintained at 323 K using Nose-Hoover’s method [64,65] with a coupling constant of 0.1 ps. Van der Waals (vdW) interaction is described with a space cutoff of 1.4 nm. Long-range electrostatic interaction is described by the Particle Mesh Ewald (PME) method [66] with a recommended cutoff of 1.2 nm, especially for charged lipids [67]. For each system, four independent 300-ns MD runs were carried out with different initial velocities.

2.3. Analyses

MD trajectories were analyzed using in-house-developed codes and the GROMACS toolkits. The z-position of a residue is defined as the displacement of the centroid of each residue relative to the average z-coordinate of the phosphorus atoms (z=0) in the upper bilayer. An atomic contact is defined when two non-hydrogen atoms are located within 0.54 nm. The peptide-lipid interaction energy U is calculated using the GROMACS tools g_ener and mdrun-rerun (using the formula U_inter = U(peptide+lipid) - U(peptide) - U(lipid)). The number of hydrogen bonds (H-bonds) is calculated using GROMACS tool g_hbond. An H-bond is considered as formed only if the distance between donor D and acceptor A is less than 0.35 nm and the angle of D-H… A is greater than 150°. A salt bridge is considered to be formed if the minimum distance between the nitrogen atom of sidechain NH₃⁺ group and the phosphorus atom of lipid PO₄⁻ group is less than 0.4 nm. The order parameter S_CD of lipid tail is calculated using the formula S_CD = 0.5<3cos²θ−1>, where θ is the angle between the bilayer normal direction and the C-H bond vector (in simulations) or the C-D bond vector (in experiments). The angular brackets denote a time and ensemble average [68]. This parameter reveals the orientation and the (dis)order degree of lipid acyl chains, correlated with the perturbation of hIAPP on the lipid tail. The larger the S_CD is, the more ordered the lipid acyl chains are. The thickness of lipid bilayer is estimated by the average of the phosphorus-to-phosphorus distance [69]. The local bilayer thickness is estimated by the average z-position of the phosphorus atoms in the upper leaflet minus that in the lower leaflet within a certain cutoff from hIAPP peptides.

3. Results and discussion

3.1. Orientation and insertion depth of a membrane-bound hIAPP protofibril

Membrane-bound hIAPP protofibrils remain inserted in the lipid bilayer, but reorient from their initial insertion direction and exhibit a stable conformational behavior. The time evolution of the root-mean-square deviation (RMSD) of Cα-atoms relative to the initial structure of hIAPP protofibril in Figs 2(a–c) shows that the Cα-RMSD in each MD run are smaller than 0.4 nm, and the average Cα-RMSD value for each of the three systems is about 0.2 nm, indicating that the hIAPP protofibril is highly stable in the presence of anionic membranes under both neutral and acidic pH conditions.

Fig. 2.

Time evolution of Cα-RMSD with respect to the initial structure of hIAPP protofibril (a-c) and the tilt angle of hIAPP protofibril relative to the bilayer surface (d-f) for different systems. Four independent MD runs for each system are denoted in different colors, and the average value is in black. The tilt angle is defined as the protofibril orientation relative to the membrane surface (see Fig. 1(c)).

The time evolution of the tilt angle of the hIAPP protofibril relative to the membrane surface in Figures 2(d) shows that the hIAPP in hIAPP-DPPG(N) system has a relaxation time of ~50 ns at the beginning, then adjusts the orientation slowly for tens of nanoseconds and finally reaches an average angle of 52°. This tilt angle is close to hIAPP orientation of 48° in DPPG monolayers, which was predicted by chiral SFG spectroscopy and quantum chemistry calculation [45]. However, at acidic pH, hIAPP reaches a smaller tilt angle of 40°, and one MD run with a long relaxation time of ~200 ns is observed (Fig. 2e). In the case of a mixed DPPC/DPPG membrane at neutral pH, the average tilt angle of hIAPP protofibril is 77° (Fig. 2f)), not far away from the initial tilt angle of 90°.

To quantify the peptide orientation relative to the membrane surface, we plot in Figs. 3(a–c) the time evolution of z-position of individual residues in a representative MD run. The low z-positions of N-terminal residues with respect to those of C-terminus residues reflect that the hIAPP protofibril leans to the N-terminal side in the DPPG bilayer at neutral pH. The z-position of each individual residue in hIAPP protofibril is quite stable during the whole 300-ns simulation time, with the turn region at 18–27 partially buried below the bilayer surface. At acidic pH, the turn region has a larger z-position in DPPG bilayer than that in hIAPP-DPPG(N) system, indicating that the insertion depth of the hIAPP protofibril at acidic pH is smaller than that at neutral pH. Clearly, the positively charged His18 at acidic pH prevents the deep membrane insertion. This change in orientation is consistent with a previous NMR experimental study showing that the hIAPP_1–19 peptides tend to orient at the micelle surface at pH 6.0 [29]. The larger z-position in the N-terminal region at the very beginning of the simulation also reflects a longer relaxation time. As for the hIAPP-DPPC/DPPG(N) system, the enlarged blue region and the higher N-terminal z-position indicate a deeper insertion and a larger tilt angle with respect to those of the hIAPP-DPPG(N) system.

Fig. 3.

Time evolution of residue-based z-position (a-c) and number of contacts between each individual residue of hIAPP protofibrils and lipid membranes (d-f) in a representative MD run for each hIAPP-membrane system.

The time-dependent number of contacts between each individual residue of the hIAPP protofibril and lipid membranes in a representative MD run is given in Figs. 3(d–f). At neutral pH, the N-terminal residues Lys1, Arg11 and Phe15 and residues in the turn region of hIAPP protofibril, especially residues Asn21, Asn22 and Phe23, have a large number of contacts with the DPPG bilayer. In the hIAPP-DPPG(A) system, the contact number increased in the N-terminal region and reduced in the turn region, corresponding to a larger tilt angle and a shallower insertion depth. His18 displays a remarkably stronger interplay with DPPG, while the contacts between Arg11 and lipids get smaller. As for the hIAPP-DPPC/DPPG(N) system, the turn region 18–27 has more contacts with membranes, among which Phe23 contributes the most. Residues Arg11 and Phe15 also have impressive contacts with lipids, while Lys1 has no contact with membrane due to the large tilt angle.

Figures 4(a–c) represent the final snapshots of the hIAPP protofibril at the membrane interface for each of the three systems. The membrane-bound hIAPP exhibits that different conformational behaviors accompany different pH and lipid composition. Arg11 always anchors at the bilayer surface and Lys1 assists to adjust the oligomer orientation, different from the binding behavior of hIAPP monomer in which two basic residues Lys1 and Arg11 anchor the peptide to charged lipid headgroups by strong electrostatic interactions [48,49]. Figures 4(d–f) display the average z-position of each residue of hIAPP protofibirl at bilayer surface and the water density profile along the membrane normal. In the presence of the DPPG bilayer, Phe19 has the lowest z-position of −1.62 nm among all the residues at neutral pH, while this value becomes −1.24 nm at acidic pH. In addition, the N-terminal residues in the hIAPP-DPPG(A) system are located closer to phosphorus atoms in the upper leaflet. These results reveal that the hIAPP protofibirl at acidic pH has a small insertion depth and a preference to interact with DPPG bilayer surface. At neutral pH, Ser20 has the lowest z-position of −1.75 nm in the mixed DPPC/DPPG membrane and is not yet able to cross the bilayer center. The N-terminal residues 1–5 are ~1 nm higher and residues 18–27 in the turn region insert more deeply into the membrane than those in pure DPPG bilayer. Arg11 is no longer the closest residue to bilayer surface. These data reveal that the reduced net charge of mixed DPPC/DPPG membrane weaken the restriction of hIAPP to the bilayer surface. The water density profile shows that no water molecules enter the hydrophobic tail region of the lower leaflets (Fig. 4(e–f)). The hIAPP trimer was reported to lose the strand-turn-strand motif and induce the diffusion of Na+ deep into the membrane in the neutralizing ion only simulation [46]. However, when additional salt was added, the peptide motif was kept and no ions diffused into the hydrophobic region of bilayer, with water flow inside the membrane being much reduced [46]. The salt concentration in our simulated systems is 0.1 M, and few water molecules are observed to penetrate into headgroup region of the upper leaflet.

Fig. 4.

(a-c) Snapshots of hIAPP protofibril at membrane interface in a representative MD run of each system. The hydrophilic, hydrophobic, Lys1, Arg11, and positively charged His18 residues are colored in cyan, gray, red, orange, and yellow, respectively; DPPG and DPPC lipids are denoted in silver and green, with phosphorus atoms in spheres. (d-f) Average z-position of each residue of hIAPP protofibirl at bilayer surface and water density profile along the membrane normal for hIAPP-DPPG(N) (d), hIAPP-DPPG(A) (e) and hIAPP-DPPC/DPPG(N) (f) systems. Average z-positions of phosphorus atoms in upper and lower leaflets are denoted by green dashed line.

3.2. Electrostatic, van der Waals, hydrogen-bonding and salt-bridge interactions between hIAPP protofibrils and membranes

To identify the critical interactions that stabilize the orientation and structure of hIAPP protofibril in membranes, we plot in Figs. 5(a–c) the average interaction energy between each residue of hIAPP protofibril and lipid membranes (per lipid), decomposed into van der Waals (vdW) and electrostatic terms. Figure 5(a) shows that at neutral pH, residues Lys1 and Arg11 have the strongest interaction with DPPG bilayers, dominated by electrostatic components. Residues 18–27 in the turn region as well as Phe15 has a strong vdW interactions with the lipids, among which Asn21, Asn22 and Phe23 contribute the most. As Arg11 is the anchor residue, the orientation of the hIAPP protofibril results from the combination of electrostatic attraction between Lys1 and the bilayer surface and the vdW interaction between the turn region and the lipid bilayer. At acidic pH, residues Lys1, Arg11 and His18 interplay most strongly with membranes (Fig. 5b). The electrostatic attraction of Lys1/His18 with bilayers leads to a smaller tilt angle of hIAPP protofibril. This significantly enhanced attraction between His18 and the lipid bilayers reveals its key role in hIAPP-lipid interaction at acidic pH, consistent with the observation that His18 is critical in the modulation of pH-dependent hIAPP_1–19 orientation in the anionic membrane. In the case of mixed DPPC/DPPG bilayer under neutral pH, the electrostatic terms of residues Lys1 with lipids are significantly reduced due to the replacement of anionic DPPG lipids by zwitterionic DPPC lipids. The turn region 18–27 has an increased interaction energy, among which Phe23 interacts strongest with the bilayer mixture. The hIAPP-membrane interaction energy analysis reveals that besides charged residues K1, R11 and H18, aromatic residues Phe15 and Phe23 also display strong interactions with lipid bilayers. This finding is consistent with a previous study of hIAPP_12–18 aggregation in DMPC vesicles, showing the high affinity of Phe with lipid bilayer [24].

Fig. 5.

(a-c) The interaction energy (kcal/mol) between each individual residue of the hIAPP protofibril and lipid membranes (per lipid) for each of the three systems, decomposed into van der Waals (vdW) and electrostatic terms. (d-f) Number of H-bonds between each residue of hIAPP protofibril (per chain) and different lipid headgroups (glycerol, phosphate, and ester groups) for each simulated system. Data are averaged over the last 60 ns of four MD runs for each system.

The number of H-bonds between each residue of the hIAPP protofibril (per chain) and different lipid headgroups in Figs. 5(d–f) shows that the protofibril mostly forms H-bonds with the phosphate groups, and less with the ester groups, and least with the glycerol groups. Arg11 has the highest number of H-bonds with lipid headgroups in the three systems, which facilitate its role as the anchor point at the membrane surface. In the hIAPP-DPPG(N) system, the H-bonds are formed mostly by charged residues Lys1 and Arg11, and polar residues Asn21 and Asn22 with lipid headgroups. The number of H-bonds between His18 and the lipid heads is dramatically increased in the hIAPP-DPPG(A) system, and more DPPG ester groups are involved in H-bond formation. In the case of a mixed DPPC/DPPG bilayer at neutral pH, the total number of H-bonds between hIAPP and lipid heads is greatly reduced due to the weakened electrostatic attraction. As a result, hIAPP inserts deeper into the bilayer to maximize its interaction with the lipid bilayer. This resembles the NMR and fluorescence observation of an antimicrobial peptide MSI-78, in which the peptide inserts more deeply in PC lipid bilayers than in PG lipid bilayers [70].

The number of salt bridges formed between Lysl, Argil and His18 residues of hIAPP protofibril and phosphate groups of lipid bilayers is listed in Table 1. It shows that Argil has the most salt bridges with membranes. The Arg11-lipid salt bridges help to stabilize Arg11 at the bilayer surface. Lys1 and Arg11 form less salt bridges with DPPG lipids at acidic pH. The number of salt bridges between Lys1 and mixed DPPC/DPPG bilayer gets much smaller. The hIAPP-membrane interaction is enhanced by the additional force resulting from the formation of Lys1-PO₄ and Arg11-PO₄ salt bridges [47].

Table 1.

Number of salt bridges formed between Lys1, Arg11 and His18 residues and phosphate groups of lipid bilayers.

System	Lys1	Arg11	His18
hIAPP-DPPG(N)	5.0	6.0	—
hIAPP-DPPG(A)	2.8	3.2	1.1
hIAPP-DPPC/DPPG(N)	1.0	4.1	—

3.3. Perturbation of the membrane-bound hIAPP protofbirl on the local ordering of lipid bilayers

The toxicity of hIAPP oligomers is related to hIAPP-membrane interactions [12,17]. To examine the influence of membrane-bound hIAPP protofibril on the bilayers, we calculated the lipid tail order parameter S_CD of acyl chain 1 (sn-1) and the local membrane thickness (Fig. 6). Figure 6(a) shows that in the presence of the DPPG bilayer, the hIAPP protofibril disturbs the membrane integrity and the impact is more obvious at acidic pH than at neutral pH. This difference is attributed to the stronger peptide-lipid electrostatic and H-bonding interactions at acidic pH. Similar perturbation effect of hIAPP on lipid order was revealed by NMR spectra at both low and neutral pH [28]. As for the hIAPP-DPPC/DPPG(N) system, the disturbance of the protofibril on the lipid integrity is much weak. The local membrane thickness within different cutoffs from peptides in Fig. 6(b) shows that hIAPP protofibril thins the neighboring lipid bilayer. Similar change of bilayer thickness was reported previously, in which the DPPG bilayer is thinned by 1–2 nm in the vicinity of hIAPP, and the bilayer farther away maintain a thickness of ~4.2 nm as without hIAPP aggregates [46]. The hIAPP protofibril has a similar impact on DPPG bilayer thickness at neutral and acidic pH, and induces a slightly thinner thickness in the mixed DPPC/DPPG bilayer with respect to the pure mixed membrane. With the increase of membrane-bound peptide concentration, the hIAPP-lipid interaction is expected to cause larger membrane disruption.

Fig. 6.

Influence of membrane-bound hIAPP protofibril on the lipid tail order parameter S_CD of acyl chain 1 (sn-1) (a) and the local membrane thickness (b). In the S_CD calculation, only the lipids within a minimum distance of 1.0 nm from hIAPP peptides are considered. The S_CD of a neat DPPG or a mixed DPPC/DPPG (7:3) lipid bilayer is also given, averaged over the last 20 ns data of a respective 100-ns MD run. The average thicknesses of a neat DPPG lipid bilayer and a mixed DPPC/DPPG (7:3) lipid bilayer are 42.6±0.3 Å (green dashed line) and 41.1±0.4 Å (cyan dashed line) respectively, consistent with previous reports [46,71].

4. Conclusions

In this study, we have investigated the conformational behavior and lipid interaction of hIAPP protofibril at neutral pH in the anionic DPPG lipid membrane by carrying out multiple MD simulations. We have also simulated hIAPP at acidic pH and in the presence of a mixed DPPC/DPPG (7:3) lipid membrane to examine the influence of pH and lipid compositions. We found that the hIAPP protofibril reaches a tilt angle of 52° and an insertion depth of −1.62 nm relative to the bilayer surface. Arg11 anchors hIAPP protofibril to anionic lipid headgroups, and the peptide orientation results from a combination of Lys1-lipid electrostatic attraction and vdW interaction between the turn region and lipid bilayer. In addition to charged residues K1, R11 and H18, aromatic residues Phe15 and Phe23 also display strong interactions with lipid bilayers, indicating the important role of aromatic residues in stabilizing the membrane-bound hIAPP protofibrils. The H-bonds and salt bridges formed between hIAPP protofibril and membrane provide additional forces to stabilize peptide-lipid interaction. The acidic pH induces a smaller tilt angle of 40° and a shallower depth of −1.24 nm, in which the protonated His18 plays a key role in the modulation of hIAPP-membrane interaction. The presence of zwitterionic DPPC lipids induces a larger tilt angle of 77° and a deeper depth of −1.75 nm, due to the reduced electrostatic restriction for hIAPP to the bilayer surface. The hIAPP-membrane interaction leads to the disturbance of lipid integrity and a thinned bilayer thickness near hIAPP. Our results provide atomic-level information of membrane-hIAPP-protofibril interactions, which may help illuminate pH-dependent and membrane lipid composition-modulated hIAPP aggregation.