Anchoring of Amyloid-β onto Polyunsaturated Phospholipid Membranes

Abstract The interaction between the toxic amyloid-β and phospholipid membranes in the early stage of Alzheimer’s disease is complicated and depends on many factors. It was found that polyunsaturated fatty acids affect the incidence of Alzheimer’s disease. The number of double bonds in the phospholipid layer may play an important role in the molecular dynamic behavior of amyloid-β on cell membranes. In the present paper, the interactions between Aβ(25-35) and each of four phospholipids, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-stearoyl-2-arachidonoyl-sn-glycero-3-phosphocholine (SAPC), 1-stearoyl-2-docosahexaenooyl-sn-glycero-3-phosphocholine (SDPC), and 1,2-diarachidonoyl-sn-glycero-3-phosphocholine (DAPC), are investigated by using all-atom molecular dynamics simulation. It is interesting that, as the number of double bonds in the membrane increases, the peptide fragment prefers to stay in the surface region of the membrane rather than penetrates deeply into the membrane. With the increasing number of double bonds, the interaction between Aβ(25-35) and the membrane surface becomes stronger, especially for the interaction between the residue 28 (LYS28) in Aβ(25-35) and the phospholipids, anchoring Aβ(25-35) onto the membrane. The double bonds in phospholipid determine not only the adsorption of the peptide fragment Aβ(25-35) but also its conformation, which will influence further aggregation of Aβ in later stages. Communicated by Ramaswamy H. Sarma


Introduction
Since the discovery of Alzheimer's disease in 1906, the pathogenesis of the disease has remained unsolved, and there are no effective drugs to treat it (L. K. Huang et al., 2020). The transmembrane protein, amyloid-beta precursor protein (APP), is cleaved by b-secretase to release C-terminal fragment of the amyloid-b (CTFb), and the product of CTFb segment cleaved by c-secretase is amyloid-b (Ab) (Seubert et al., 1992;Shoji et al., 1992). Ab is regarded as a key protein with a major impact on the cause of the disease (Hardy & Higgins, 1992). The most abundant Abs in the brain are Ab40 with 40 amino acids and Ab42 with 42 amino acids which is more toxic (Roher et al., 1993;Yan & Wang, 2006).
Ab can self-aggregate into oligomers on the membrane surface and then develop into fibers (Cohen & Calkins, 1959;Hayden & Teplow, 2013), or aggregate with other molecules outside the cell (Strittmatter et al., 1993;Venegas et al., 2017). It affects the biological activity of neurons and causes cell inflammation, and ultimately leads to the death of nerve cells (Behl et al., 1994;Hensley et al., 1994;Shearman et al., 1994). Ab fibrils were previously believed to be the key to the disease, but recent studies have shown that oligomers are more toxic (Hayden & Teplow, 2013;Selkoe & Hardy, 2016). The oligomers adsorb on the membrane and then disrupt the membrane. For this reason, studying how the monomer molecule of Ab, the starting point of oligomers, interacts with the membrane becomes crucial.
The most abundant phospholipid in the brain are phosphatidylethanolamine (PE), and phosphatidylcholine (PC), but the proportion of phospholipids in the membrane of the brain changes with age, especially in the brain of Alzheimer's patients (S€ oderberg et al., 1991). In recent years, the impact of polyunsaturated fatty acids (PUFAs) has been studied, among which omega-6 and omega-3 are the most notable (Calon et al., 2004;Lukiw et al., 2005;Wall et al., 2010).
The content of docosahexaenoic acid (DHA, 22:6 n-3), one of the omega-3 fatty acids, is high in the frontal gray matter, while the content of arachidonic acid (ARA, 20:4 n-6), one of the omega-6 fatty acids, decreases in old age (Norris et al., 2015). In cancer tissues, it was observed that the content of DHA decreased, but the content of ARA increased (Z arate et al., 2017). PUFAs are also closely related to other diseases, such as lung cancer and Parkinson's disease (Kim et al., 2015;Tamtaji et al., 2019). Experiments have shown that DHA has a protective effect on microglia, reduces inflammation, and decreases the deposition of Ab42 and Ab40 in the brain (Hjorth et al., 2013;Lim et al., 2005;K. Wu et al., 2016). Interestingly, some experiments showed that ARA could reduce the deposition of Ab in the mice's brain, whereas the DHA cannot improve the mice's cognitive performance (Arendash et al., 2007;Hosono et al., 2015). Compared to untreated participants, the healthy old people with DHA supplementation show no differences in cognitive decline, and the supplement of DHA for Alzheimer's patients seems to have an inconspicuous benefit . All these suggest that the influence of PUFA on Alzheimer's disease is complicated. It is worthy to note how membranes with different double bonds affect the structure and conformation of the monomer of Ab.
Ab(25-35) is a type of short peptide fragment that naturally exists in human brain. Ab(25-35) retains the full-length cytotoxicity of Ab and can form a fiber structure rich in b-sheet (Serpell, 2000;Shanmugam & Polavarapu, 2004). Its interaction with membranes has been extensively studied. Its self-aggregation on the membrane surface leads to membrane perforation, promotes disruption of the blood-brain barrier, and can also promote neurotoxicity (Cuevas et al., 2019;Lin & Kagan, 2002;Pike et al., 1995). Experiments showed that Ab(25-35) could be combined with anionic membranes and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) zwitterionic membranes (Dante et al., 2002;Terzi et al., 1994). In addition to the aggregate form, Ab(25-35) can be directly inserted into the membrane as a monomer (Clementi et al., 2005). The insertion of Ab(25-35) into the membrane has an impact on cytotoxicity (Tsai et al., 2014). It is found that Ab(25-35) on the DMPC membrane had two states, in which the C-terminal inserts into the PC membrane or the C-terminal is far from the double layer (Ermilova & Lyubartsev, 2020;Smith & Klimov, 2018).

System construction
The structure of the amyloid-b peptide (25-35) was gained from the Protein Data Bank (1YQT), and the sequence is GSNKGAIIGLM (D'Ursi et al., 2004). The C-terminal was embedded in the membrane, and the peptide was aligned along the z-axis. The N-terminal and C-terminal of the peptide were capped with acetylated-terminus and amidated-terminus, respectively, to avoid the electrostatic effect of the peptide. The two residues (Ile 31, Ile 32) resided at the membrane-water interface. All models were constructed by CHARMM-GUI (E. L. Wu et al., 2014).
The peptide was inserted into four membranes POPC, SAPC, SDPC, and DAPC, respectively. PC means phosphocholine, which is electroneutral and one of the most abundant lipids in the cell membrane. POPC bilayer is formed by 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine with one double bond, SAPC bilayer is formed by 1-stearoyl-2-arachidonoylsn-glycero-3-phosphocholine with four double bonds, SDPC bilayer is formed by 1-stearoyl-2-docosahex-aenoyl-sn-glycero-3-phosphocholine with six double bonds, and DAPC bilayer is formed by 1, 2-diarachidonoyl-sn-glycero-3-phosphatidylcholine with eight double bonds. The chemical structures of four lipids are illustrated in Figure S1 in Supplementary Material. The ratio between the upper leaflet and the lower leaflet is set to 50:52. The lower leaflet has slightly more lipids than the upper leaflet to make sure that the initial area of the upper and lower leaflets is approximately equal. One chloride ion was randomly placed in a periodic rectangular parallelepiped box, making the system electrically neutral and the box was infused with modified-TIP3P water (Jorgensen et al., 1983).

MD simulation
An all-atom MD simulation was performed by the Gromacs software (version 2020.02), with an acceleration of GPU (P all et al., 2020). A CHARMM36m force field was employed for all the MD simulations in the present work (J. Huang et al., 2017). Energy minimization before heating the system was carried out to avoid unreasonable contact. The minimization was performed by 5000 steps using the steepest descent algorithm and ended when the maximum force is smaller than 500 kJ mol À1 nm À1 . Then, the system was slowly heated to the target temperature 310 K with the 1000 kJ mol À1 nm À1 force imposed on the backbone of the peptide and membrane. Afterward, the system was stabilized for 200 ps in the NVT ensemble by using Berendsen (Berendsen et al., 1984). The MD simulation was conducted in the NPT ensemble to allow the system to change its volume as the real physiological environment. This process continued for 1 ns with the relaxation of the constraint on peptide and membrane. The production simulation was run 800 ns, which facilitated an integration time step of 2 fs by using the leap-frog algorithm. The temperature was sustained at 310 K using V-rescale (Bussi et al., 2007), and the pressure was controlled at 1 bar using semi-isotropic pressure with the Parrinello-Rahman barostat (Parrinello & Rahman, 1981). The electrostatic interaction was calculated with particle mesh Ewald (PME) method with a cutoff at 12 Å (Essmann et al., 1995). The Lennard-Jones potential interaction was cut at 12 Å with a switching function used between 10 Å to 12 Å. The LINCS algorithm was employed to constraint all bonds with H-atoms but not water, which is constrained by the settle algorithm (Hess et al., 1997;Miyamoto & Kollman, 1992). After 800 ns simulation, we clipped the last 400 ns of trajectory for analysis.
The parameter Z p was defined to be the average distance of phosphorus atoms of the upper leaflet from the center of mass of the bilayers. The average thickness of the membranes, order parameter S CD , and area per lipid of bilayers were calculated by using MEMBPLUGIN (Guix a-Gonz alez et al., 2014). For particularly, the order parameter was defined by with h z as the angle between the z-axis and the vector from C i to H i bonded to it. All snapshots were made by PyMol (Schr€ odinger, LLC, 2015). The secondary structure of the peptide with time was calculated by using the Define Secondary Structure of Proteins (DSSP) algorithm.

Anchoring of Ab(25-35) on the phospholipids
Simulations are performed to study the effect of double bonds in phospholipids on lipid-peptide binding. The final states of the systems after 800 ns of MD simulations are illustrated in Figure 1. One can see that the peptides in all cases can float on the membranes after the equilibrium is achieved. The hydrophobic C-terminals (residues 30-35) are located below the layer of hydrophilic heads of the upper leaflet. This occurs because of the hydrophobic interaction between these residues and the tail of phospholipids. The helix structure of the peptide in the POPC membranes makes its C-terminal hydrophobic and stabilizes in the hydrophobic region in the membrane. With increasing double bonds numbers in phospholipid, the N-terminal of the peptide gradually approaches the phosphatidylcholine. The hydrophilic N-terminal of the peptide can adopt two orientations, either vertical to the membrane or aligning to the surface due to the hydrophilicity of the head of the membranes and its outside surroundings. In the case of the DAPC, the chain of the peptide horizontally lays on the surface of the membrane.

Distribution of Ab(25-35)
The location of the peptide on membranes can characterize the adsorption of the peptide to the membrane and the interaction between the peptide and the phospholipids. Mass density profiles of Ab(25-35) and PO 4 are plotted in Figure 2 to show the distribution of the peptide around the membrane. For the membranes with fewer double bonds, the peptides reside deeper in the hydrophobic tails. All peptides are distributed in the range from 0.5 nm to 3.5 nm, which means the peptide is not fully located within the bilayers. The distance of the maximum population of peptides from the midplane of the membrane is near or less than 1.5 nm for POPC, SAPC, and SDPC, respectively, but it is larger than 1.5 nm for DAPC. It means that the peptide on DAPC is closer to the region of the hydrophilic heads. Moreover, for the DAPC bilayers, Ab(25-35) has the highest and the narrowest density maxima, which is 45 kgÁm À3 , larger than 42 kgÁm À3 for POPC, 33 kgÁm À3 for SAPC, and 37 kgÁm À3 for SDPC, indicating that the peptide is more concentrated. The peptide in POPC bilayers also shows a relatively high-density maximum because of its compact helix structure as shown in Figure 1.
To examine the locations of the individual amino acids of Ab(25-35) in detail, we calculate the probability of distribution P (i, z) of amino acid i from the center of mass of the membrane along in the Z direction. The probability distribution of amino acids and the average position Z p of phosphorous atoms of the upper leaflet from the center of mass of the bilayers are given in Figure 3. In most cases, the residues (25-28) are above the average position Z p of phosphorus atoms, while the residues (29-35) are below Z p . LYS28 is always near the Z p , indicating the strong interaction applied on it to fix its location. One can see that the residue 33 (GLY33) fluctuates and is closer to Z p than its neighboring residues because it is more flexible. As shown in Figure 3(A) for the POPC system, the distribution of residues (30-35) is more concentrated compared to the other cases. The more concentrated distribution of residues (30-35) around POPC is due to the helix structure of the peptide. For the SAPC system in Figure 3(B), although the distances of residues (30-35) from the midplane are also shorter than Z p , the distributions of residues (30-35) are closer to Z p than those in the POPC system. For the SDPC system, the differences in the distribution of residues (30-35) are not obvious compared to the SAPC system, but from the Figure 3(C), it can be observed that the residues SER26 and ASN27 are more remote from the membrane than those in the system of SAPC. It coincides with the final states of the two systems in Figure 1. With more double bonds in DAPC, the N-terminal of the peptide leans on the membrane, and the residues (30-35) are near and above Z p (1.82 nm) (Figure 3(D)).
Furthermore, the distributions of the center of mass (COM) of residues (30-35) away from the COM of phosphate groups are calculated to examine the depth of intersection of Ab into the membrane (Figure 4). The residues in POPC have the lowest distance, changing from À1.25 to 0 nm, which means that the C-terminal of the peptide (residues 30-35) deeply penetrates the membranes. However, the distribution of the residues in SAPC or SDPC changes from À1.125 to 0.125 nm. The residues are farther from DPAC. These results verify that the double bonds in phospholipids lead to depletion of hydrophobic residues of Ab from the bilayer.
From the results above, it is found that the insertion depth of Ab(25-35) gets less as more double bonds in lipids. the location of the C-terminal of Ab(25-35) in POPC membranes in Figure 3(A) is in good agreement with Ab(25-35) monomer in DMPC (Smith & Klimov, 2018). Conversely, with more double bonds, the Ab(25-35) monomer gets closer to the membrane-water interface. This intriguing result is not consistent with some previous results in which the Ab(25-35) inserts deeper into omega-3 membranes (Ermilova & Lyubartsev, 2020;Vitiello et al., 2013). But those experiments or simulations account for oligomers, not a monomer. Experiments about the monomer have demonstrated that Ab(25-35) has two states, C-terminal inserted in membranes or C-terminal far away from the hydrophobic interior of membranes (Dante et al., 2002;Dies et al., 2014), and these two flexible states can mutually convert easily (Smith & Klimov, 2018). The double bond can have a dramatic effect on reducing the interfacial energy with water, which means more double bonds escalate more hydrophilicity. The increased hydrophilicity of membranes gives the hydrophobic C-terminus of Ab(25-35) immersed in the membrane a tendency to adjust its conformation to gain a more stable position. So Ab(25-35) floats on the membrane-water interface in DAPC.

Effects of Ab on the structure of membrane
The order parameters S CD for C-H bonds of phospholipids are calculated based on Equation (1) to explore the influence of the peptide on the structure of phospholipids ( Figure 5). For POPC, SAPC, and SDPC, the magnitude of S CD for C-H bonds of phospholipids within a distance of 5 Å from the peptide is slightly smaller than those at a distance larger than 5 Å. It implies that Ab induced interference when it approaches and inserts into these phospholipids. However, there is no obvious difference in the two S CD s of DAPC for two different regions.
The area per head group and the thickness of the membrane can be used to characterize the structure of the membrane. It is also found that there is a little increase in the area per head group of phospholipids in the upper leaflet as Ab disrupts the arrangement of the phospholipids (see Table S1). However, the thickness of the membrane almost remains unchanged whether Ab is present   was small. Figure 5 shows that the insertion of monomers has limited change on membrane order parameters, which are in good agreement with previous simulations (Poojari et al., 2013;Smith & Klimov, 2018). In particular, Figure 5(D) demonstrates that double bonds increase the flexibility of lipids to adjust the structure in response to the insertion of Ab(25-35), similar to the results reported by Cordom ı et al. (Cordom ı & Perez, 2007). The fluidity of the bilayer enhanced by excess double bonds counteracts the influence of insertion of Ab. The striking property indicates its protective role for membranes in Alzheimer's disease.

Conformations of Ab(25-35)
From the snapshots of the final states in Figure 1, different structures of Ab can be observed. The compositions of coil, bend/turn, and helix of the peptide in the four lipids are shown in Table 1. The composition of secondary structures of Ab(25-35) in the POPC system is different from the others with 51% of coil, 17% of bend/turn, and 30% of helix. The compositions are similar in the systems of SAPC and SDPC, with 72% of coil and 26% of bend/turn in SAPC, and 75% of coil and 23% of bend/ turn in SDPC. Helix structure of Ab(25-35) is only found in the POPC system. Remarkably, there is more bend/turn structure found in the DAPC system with 39% of bend/turn.
The backbone of the Ab(25-35) is bent on the surface of the membrane, so it is interesting to calculate the bend/turn content along the residual sequence ( Figure 6). Compared to those in the SAPC, SDPC, the peptide in POPC forms the helix structure along the chain, so its bend/turn structure reduces. However, in DAPC, when the peptide approaches the hydrophilic phosphorus plane, there are more bend/turn structures in residues (30-34), which increases the bend/turn structure to 70% on average. This suggests that as the peptide contacts the membrane, the peptide takes more turn/bend structure to adjust its structure to interact with phospholipid heads. The change of structure of Ab(25-35) has also been observed in omega-3 membranes, which shows Ab(25-35)s tend to be in random structures in the omega-3 membrane in the early stage of aggregation (Emendato et al., 2016).
Furthermore, the change of Ab(25-35) monomer's conformation implies a protective effect of PUFA on the membrane. Recent experiments showed that the b-sheet structures of Ab(25-35) in membranes can lead to the permeation of Ca 2þ  cations (Kandel et al., 2017(Kandel et al., , 2019. However, Ab(25-35) were found to aggregate with helix structures in membranes in low concentrations, but it transforms into b-sheets in high concentrations (Smith et al., 2019;Tang et al., 2016). The rupture of the helical structure of Ab(25-35) in PUFA membranes may retard the formation of Ab(25-35) into b-sheets.

Membrane-peptide interaction
The electrostatic energy and van der Waals interaction energy between the peptide and the POPC, SAPC, SDPC, and DAPC membrane are calculated to study how the membrane affects the conformation of the peptide. In Figure 7, one can see that the electrostatic interactions between the peptide and the membrane are much stronger than their van der Waals interaction. For the peptide in POPC, the van der Waals interaction between the peptide and the POPC lipids is strong because of the hydrophobic interaction between lipid tails and the C-terminal of the peptide. For SAPC, LYS28 provides the most contribution in van der Waals interaction, and the contributions from other residues are negligible. For SDPC, the van der Waals interactions between residues 30-35 and the membrane are stronger. The peptide has the strongest interaction with DAPC. In all cases, LYS28 has a dominant contribution to the electrostatic interaction. The strong interaction of LYS28 with the PC membranes has also been verified by other studies (Kandel et al., 2017;Smith & Klimov, 2018;Terzi et al., 1994). Additionally, the increment of energy with more double bonds can partially be attributed to the proximity of the SER26 and ASN27 residues to the phosphorus atoms, which is consistent with the Ab(25-35) oligomer in omega-3 membranes (Ermilova & Lyubartsev, 2020). Cuco et al. (2016) proposed three steps of Ab(25-35) aggregation: adsorption, nucleation, and permeation. They suggested that the interaction between the Ab(25-35) and polar region of the lipid monolayer may affect its nucleation process since the adsorption of the Ab(25-35) monomer on the DAPC membrane is stronger than on the POPC membrane. For comparison purposes, the case of the peptide with the N-terminal initially inserted in membrane is also studied, and the results are presented in Supplementary Material. It is found that the role of the double bond is similar to that in the case with the C-terminal initially inserted. In addition, the N-terminal of Ab(25-35) remains inclined and is located on the surface of membranes, even if the N-terminal is initially inserted. Moreover, the interaction between LYS28 and the membrane dominates as LYS28 does in the case with the Cterminal initially inserted.

Conclusions
The MD simulation is carried out to study the equilibrium state of Ab(25-35) on phospholipid membranes with various double bonds. As the number of double bonds increases, Ab(25-35) prefers to stay on the surface of the membrane rather than penetrate deep into the film. The position of Ab(25-35) has a small effect on the structure of the membranes, only slightly varying the thickness of the membrane and the orderliness of the membrane around the peptide. However, the conformation of the Ab(25-35) peptide is affected by the various membranes, which is related to why the unsaturated phospholipid retard the aggregation of Ab. With increasing double bonds in phospholipid, the interaction between the membrane and Ab(25-35) was enhanced, and the adsorption becomes stronger. The electrostatic attraction between LYS28 and the membrane plays a key role.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by Key-Area Research and Development Program of Guangdong Province (2019B010940001).