J. Phys. Chem. B 2009, 113, 9447–9453
9447
Lipids Enhance Apolipoprotein C-II-Derived Amyloidogenic Peptide Oligomerization but Inhibit Fibril Formation Andrew Hung,† Michael D. W. Griffin,‡ Geoffrey J. Howlett,‡ and Irene Yarovsky*,† School of Applied Sciences, RMIT UniVersity, GPO Box 2476 V, Victoria 3001, Australia, and Bio21 Molecular Science and Biotechnology Institute, UniVersity of Melbourne, Melbourne, Victoria 3010, Australia ReceiVed: February 4, 2009; ReVised Manuscript ReceiVed: May 19, 2009
We investigated the effect of submicellar lipids on amyloid fibril formation. Thioflavin T fluorescence studies showed that submicellar levels of the short-chain phospholipids, dipentanoylphosphatidylcholine and dihexanoylphosphatidylcholine, strongly inhibited amyloid fibril formation by an 11-residue peptide derived from human apolipoprotein C-II (apoC-II60-70). In contrast, sedimentation equilibrium analysis of these peptide-lipid mixtures indicated the presence of soluble oligomeric complexes. To acquire insight into the atomic level influences of these lipids on the initial stages of aggregation of the peptide, we performed molecular dynamics (MD) simulations coupled with umbrella sampling to determine dimerization free energies of a number of β-stranded and random coil dimer complexes, both in the presence and absence of lipids. The simulations indicate that, in contrast to their inhibitory effects on fibril formation, short-chain phospholipids promote the formation and stabilization of dimers by enhancing intersubunit hydrophobic interactions. On the basis of these experimental and computational results, we propose that peptide-bound lipids can inhibit amyloid fibril formation by trapping of dimers and other oligomeric species in diverse nonfibril forming conformations, reducing their likelihood of acquiring subunit conformations prone to fibril nucleation and growth. In light of the demonstrated cytotoxicity of amyloid peptide oligomers, our results suggest that, by enhancing the stability of oligomeric peptide species, the presence of solvated lipids may contribute to the cytotoxicity of fibrillogenic proteins and peptides. 1. Introduction A number of debilitating human disorders including Alzheimer’s disease, type II diabetes, and the prion diseases1 are characterized by the extracellular deposition of amyloid plaques. These deposits are composed of protein fibrils and nonfibrillar components including lipids and lipid complexes. Proteinaceous fibrils are generally considered as self-assembled and misfolded proteins formed by intermolecular backbone-backbone hydrogen bonding in a cross-β pattern, with the H-bonds perpendicular to the long axis of the fibrils.2 Interest in the disease-related properties of amyloid fibrils arises from the findings that small oligomeric intermediates in the assembly pathways are toxic.3-5 Amyloid deposits are also associated with atherosclerotic plaques, where a number of proteins, including human apolipoprotein (apo) C-II (a 79-residue protein involved in lipid metabolism) have been identified.6,7 ApoC-II adopts a primarily R-helical structure in the presence of lipid mimetics.8-10 However, under lipid-depleted conditions, apoC-II readily forms homogeneous fibrils with a “twisted ribbon” morphology and all of the characteristics of amyloid fibrils.11 Hydrogen/deuterium exchange and proteolysis studies identified core regions within these fibrils postulated to drive fibril formation.12 Consistent with this hypothesis was the observation that a tryptic peptide derived from a core region, apoC-II56-76, and a related shorter peptide, apoC-II60-70, retained the ability of the full-length protein to form fibrils. The presence of phospholipids below and above the critical micelle concentration (CMC) has been demonstrated * Corresponding author. E-mail address:
[email protected]. Phone: +61 3 9925 2571. Fax: +61 3 9925 5290. † RMIT University. ‡ University of Melbourne.
to alter the kinetics of apoC-II fibril formation and morphology;13,14 submicellar, short-chain phospholipids accelerate fibrillization; while micellar phospholipids composed of fatty acid with chain lengths between 4 and 14 carbons inhibit fibril formation.14 In the present work, we combine experimental and computational techniques to study the effects of dihexanoylphosphatidylcholine (D6PC) and dipentanoylphosphatidylcholine (D5PC) on fibril formation and dimerization by the tryptic peptide apoCII60-70. This peptide corresponds to a core region of the fulllength apoC-II fibrils12 which retains the ability to form fibrils independently. Short peptide fibrils share the backbone structural properties of those derived from full-length proteins, and so understanding the mechanisms of peptide fibrillization and its inhibition by lipids serves as a foundation for understanding the lipid-dependent aggregation of larger proteins. We monitor the time course of fibril formation of the peptide under lipiddepleted and lipid-rich conditions using Thioflavin T fluorescence and apply ultracentrifuge sendimentation analysis to examine the nature (mass distribution) of the species obtained as a result of lipid-peptide interactions. We subsequently apply molecular dynamics simulation methods to obtain atomic-level details of the effects of lipids on the initial stages of peptide aggregation. 2. Materials and Methods Peptide Synthesis. ApoC-II60-70 (MSTYTGIFTDQ) was assembled fully manually by Bio21 Peptide Technologies (Vic, Australia) using Fmoc solid-phase synthesis. The purity (>95%) and identity of the peptide were confirmed by reversed-phase high-performance liquid chromatography (HPLC) and mass spectrometry using a Q-TOF LC mass spectrometer (Agilent
10.1021/jp901051n CCC: $40.75 2009 American Chemical Society Published on Web 06/19/2009
9448
J. Phys. Chem. B, Vol. 113, No. 28, 2009
Technologies, USA). Concentrated stocks of apoC-II60-70 were maintained in 5 M guanidine hydrochloride, 10 mM Tris · HCl, pH 8.0, to prevent aggregation of the peptides. The concentration of each stock solution was confirmed by UV absorbance at 280 nm (molar extinction at 280 nm in 5 M guanidine hydrochloride ) 1280 M-1 cm-1). Experimental Fibril Preparation and ThT Assays. In all cases, fibril formation was initiated by dilution of concentrated peptide stocks to 240 µM peptide in buffer, thereby reducing the guanidine hydrochloride concentration to levels where it exerts little effect on the peptide. Fibril formation was conducted at room temperature (22 °C). For experiments examining the effect of lipid, apoC-II60-70 was diluted with 100 mM sodium phosphate buffer, pH 7.4, in the presence or absence of the stated concentrations of D6PC. We note that the ThT-monitored peptide fibrillization time courses measured in the presence of D5PC under similar conditions, i.e., peptide and lipid concentrations, ratios, and temperature, yielded results similar to those presented in this work.15 Thioflavin T fluorescence assays were performed as follows. At fixed time points after initialization of fibril formation, 20 µL aliquots of the fibril suspension were removed and mixed thoroughly with solutions of thioflavin T (ThT) in a 96-well microtiter plate well (final concentration 10 µM ThT in 100 mM sodium phosphate buffer, pH 7.4, 250 µL final volume). Fluorescence intensities were measured using an fmax fluorescence plate reader with excitation and emission filters of 444 and 485 nm, respectively. All measurements were made in duplicate. Sedimentation Equilibrium Analytical Ultracentrifugation. Analytical ultracentrifugation experiments were performed in a Beckman XL-I analytical ultracentrifuge equipped with UV/ vis scanning optics and an An-60 Ti 4-hole rotor. ApoC-II60-70 (240 µM) was prepared in 100 mM sodium phosphate buffer, pH 7.4, in the presence of D5PC at concentrations of 960 µM, 1.44 mM, 2.88 mM, 4.32 mM, and 5.76 mM. Sample (120 µL) and reference (140 µL) solutions were loaded into 12 mm double-sector cells with quartz windows. Centrifugation was performed at 40 000 rpm until equilibrium distribution had been obtained in the sample contining 2.88 mM D5PC (∼16 h). The rotor speed was then increased to 50 000 rpm and maintained until the sample had again reached equilibrium distribution. Radial absorbance data were collected at each speed at a wavelength of 280 nm with radial increments set to 0.001 cm. Molecular Dynamics Simulation Parameters and Setups. Molecular dynamics simulations were performed under constant particle number, pressure, and temperature (NPT) conditions using GROMACS version 3.316,17 and the GROMOS96 forcefield18 with the 43a1 parameter set. Forcefield parameters for the D5PC lipid molecules were derived from those of Berger et al.6 Simulation trajectories were integrated using time steps of 2 fs. Particle mesh Ewald (PME) summation19 was employed for evaluation of long-range electrostatics, while a cutoff radius of 1.4 nm was specified for short-range nonbonded interactions. Covalent bond lengths were constrained via the LINCS algorithm.20 The simulation systems were maintained at 300 K and 1 bar using Berendsen temperature and pressure coupling.21 Analyses of MD trajectories were performed using the GROMACS suite of analysis tools. Visualization of molecular graphics was performed using VMD.22 We calculate the dimerization free energies (using the umbrella sampling method described below) for the apoC-II60-70 peptide dimers with four different initial conformations, two of which are extended and another two in coiled conformations. The extended conformations chosen are the antiparallel (ANTI)
Hung et al.
Figure 1. Effect of D6PC on fibrillization by apoC-II60-70. ApoCII60-70 at 240 µM was incubated in the absence (filled circles) and presence of D6PC at 240 (open circles), 480 (inverted filled triangles), 960 (open triangles), and 1440 µM (filled squares).
Figure 2. Sedimentation equilibrium data obtained for apoC-II60-70 in the presence of 2.88 mm D5PC at 40 000 rpm (gray filled circles) and 50 000 rpm (open circles). C represents the equilibrium distribution of peptide concentrations, while C0 represents the initial loading concentration of peptide. Data calculated for equilibrium distribution of a homogeneous solution of monomeric apoC-II60-70 at 40 000 rpm and 50 000 rpm are shown as dashed and dotted lines, respectively.
and parallel (PAR) β-stranded dimers (Figure 3A and B). The ANTI orientation was chosen on the basis of its structural stability, demonstrated by persistent interbackbone H-bonding between the peptides throughout the course of a ∼70 ns equilibrium MD simulation. We performed the unbiased simulation at 330 K to test the stability of the system. Such stability suggests that it is likely to be a low free-energy dimer species in solution. Additionally, previous studies on amylogenic peptides of similar length have shown that dimers can form backbone hydrogen bonds with a preferred antiparallel β-stranded organization.23 The PAR orientation is an unstable dimer, and equilibrium simulations revealed significant structural drift, with both monomers collapsing to random coil forms within ∼30 ns. The PAR orientation was chosen in the present work to determine the extent to which lipids are capable of enhancing the stability of a dimer structure known to be unstable in water. The additional two coiled conformations were generated by using GRAMM24 to dock identical random-coil monomers, obtained from ∼600 ns unbiased equilibrium MD simulations,15 in energetically favorable orientations quantified by surface complementarity. Calculating the dimerization free energies of the latter two dimer configurations enables characterization of the initial stage of oligomerization, which involves contact encounters between the peptides in conformations consistent
Lipids Effect on Amyloid Fibril Formation
Figure 3. Initial structures of the four dimer configurations: (A) ANTI, (B) PAR, (C) GRAMM1, and (D) GRAMM2. Peptide backbones are presented by red ribbons, while the aromatic side chains Y63 and F67 are represented by van der Waals’ spheres with the following color coding: red ) oxygen, white ) hydrogen, blue ) nitrogen, and cyan ) carbon.
with the monomeric state. The effects that lipids have on the energetics of “initial encounter” between two peptides can also be elucidated. These are labeled GRAMM1 and GRAMM2 and were chosen because they are of opposite relative orientations, thus enabling more conformational sampling (Figure 3C and D). The GRAMM methodology has been previously applied to generate initial “guess” geometries for free energy calculations of peptide dimerization.25 All simulations were performed in the presence of explicit SPC water.26 The initial simulation cell dimensions for ANTI and PAR are 60 × 60 × 60 Å3, while for GRAMM1 and 2, the dimensions are 50 × 50 × 70 Å3, with minor deviations of ∼2 Å in each dimension throughout the course of the NPT simulations. The relatively large cell dimensions chosen resulted in a separation between the image peptide molecules of >20 Å for all systems. This separation precludes Lennard-Jones and short-range Coulombic interactions as well as minimizes longrange Coulombic interactions between image peptides in adjacent cells; test calculations indicate no direct electrostatic interactions between the peptides at this separation (data not shown). Furthermore, indirect interactions (e.g., via solvent) are unlikely to have a significant impact on the computed dimerization free energy values. This is demonstrated by the lipidfree PMF profiles (Figure 4), all of which exhibited “plateauing” behavior between 1.5 nm (GRAMM1) and 2 nm (ANTI) separation between the same-cell peptides, indicating complete loss of interactions. Thus, since the separation between adjacent image peptides is more than 2 nm, it is likely that there are no significant nonphysical interactions between them. For each “window” of our umbrella sampling simulation, the six lipids were arranged in random orientations and positions relative to the peptides subject to the following constraint: the initial position of each lipid was placed in such a manner as to allow
J. Phys. Chem. B, Vol. 113, No. 28, 2009 9449 at least 10 Å separation between the pair of atoms, one each from the lipid and a peptide, which are in closest mutual proximity. For each window, the initial peptide structures were placed at the prescribed COM separation (with each monomer maintaining its initial conformation as described; i.e., ANTI, PAR, GRAMM1, and 2), lipids positioned as described above (for the lipid simulations), and solvated with the following numbers of SPC waters: ANTI (sol) 7043, (lip) 6919; PAR (sol) 7040, (lip) 6972; GRAMM1 (sol) 5680, (lip) 5554; and GRAMM2 (sol) 5699, (lip) 5558. Two Na+ ions were added to each system to ensure charge neutrality. For all four dimer systems, we acquired potentials of mean force (PMF) profiles both in pure water and in D5PC lipid solution. We employed umbrella sampling and WHAM27 to determine the PMF, or ∆G, as a function of separation distance between the centers of mass of two peptides, both in the absence and presence of six D5PC lipids. The latter systems correspond to peptides in the presence of a submicellar concentration of D5PC. In the present work, ∆G and PMF both refer to the free energy required to bring the two monomers from their dimeric, associated form (in this form, the dimer’s free energy is defined as 0 kcal/mol) to some separation d (nm). PMF profiles for four different initial dimer conformations (i.e., the dimer configuration at ∆G ) 0 kcal/ mol) were acquired to ensure the statistical significance of the results. To acquire a PMF profile for each system studied, a series of simulations (windows) are performed in which the separations between the COM of the peptides were restrained in each window by Hookean functions with force constants of 20 kcal/mol. Adjacent windows are separated by 0.5 Å. Fortyfive windows were applied for calculating the ANTI and PAR profiles, while thirty windows were applied for calculating the GRAMM1 and 2 profiles. Each window was simulated for 3 ns except for the ANTI with lipids system, simulated for 6 ns per window. WHAM27 was subsequently applied to the final 2 ns (final 5 ns, for the latter system) of the umbrella sampling trajectories to remove the biasing potentials and obtain the unbiased PMF profiles. 3. Results and Discussion Fibrillization was monitored by the induction of thioflavin T (ThT) fluorescence (Figure 1). The ThT fluorescence profile for apoC-II60-70 alone has a sigmoidal shape characteristic of an initial lag phase followed by rapid fibril growth. Increasing molar ratios of D6PC from 1:1 to 1:6 caused progressive inhibition of fibrillization with complete inhibition at the submicellar concentration of 1.44 mM over the time course studied (48 h). Inhibition of fibril formation by apoC-II60-70 was also observed with D5PC, using a similar range of molar ratios.15 Sedimentation equilibrium analytical ultracentrifugation experiments were conducted to ascertain the effect of D5PC on the solution properties of the apoC-II60-70 peptide. Figure 2 shows sedimentation equilibrium data, presented as a “log C/C0 vs r2” analysis, obtained for apoC-II60-70 in the presence of 2.88 mM D5PC (peptide:lipid ratio of 1:12) at 40 000 rpm (gray filled circles) and 50 000 rpm (open circles). The dashed (40 000 rpm) and dotted (50 000 rpm) straight lines indicate the predicted behaviors for a homogeneous solution of monomeric peptide at the respective rotor speeds. Curvature in the experimental data at higher radial positions, with data tending toward higher gradients than those calculated (expected) for monomeric peptide, indicates significant oligomeric complex formation by the peptide. These data suggest that short chain phopspholipids act to stabilize peptide complexes and that the formation of soluble peptide:lipid complexes may be one mechanism of the
9450
J. Phys. Chem. B, Vol. 113, No. 28, 2009
Hung et al.
Figure 4. PMF profiles for the dimerization of (A) ANTI, (B) PAR, (C) GRAMM1, and (D) GRAMM2 complexes, in D5PC lipid-free (black line) and lipid-rich (gray line) environments. Simulations were performed using Gromacs.15 Graphics produced using VMD.20 Insets show typical system configurations at indicated separations. Aromatic residues represented in CPK format, peptide backbones as ribbons, and lipids as thin lines.
observed inhibition of fibril formation. At D5PC concentrations of 1.44 mM (lipid:peptide ratio of 6:1) or below, equilibrium distributions were not reached due to loss of material from the solution column during the course of the experiment (∼24 h), precluding analysis of these data sets. This is consistent with our previous analysis of fibril formation by the peptide under these conditions and indicates time-dependent aggregation of the peptide to form fibrils.15 At D5PC concentrations of 4.32 mM (peptide:lipid ratio of 1:18) or above, equilibrium distribution of the solution was also hindered by a slow time-dependent loss of material. This was an interesting observation since, under these conditions, fibril formation is known to be completely inhibited over extended time periods. Furthermore, no amorphous precipitation of peptide has been observed in this lipid environment, suggesting that time-dependent formation of highorder, soluble, peptide-lipid complexes may account for the loss of material from the solution column. In these cases, the oligomers may be too large to contribute to the equilibrium distribution at the high rotor speeds used. These experimental studies demonstrate the importance of lipids in inhibiting fibrillization by apoC-II (and perhaps related) peptides and indicate the formation of stable, soluble, oligomeric peptide-lipid complexes which do not proceed to fibril elongation. Circular dichroism (CD) spectra for apoC-II60-70 alone indicate that the petide is predominantly unstructured in solution. Spectra acquired in the presence of submicellar D6PC (data not shown) do not differ significantly from that of peptide alone, suggesting that short-chain phospholipid does not induce gross secondary structural changes in the peptide. Accordingly, to understand the atomic mechanisms of peptide aggregation and the influence of lipids on this process, we applied computer simulations to examine the effects of solvated lipids on the association between peptide monomers. Potentials of mean force (PMF) profiles detailing the dimerization free energies of all four peptide dimer configurations
studied, with and without lipids, are shown in Figure 4A-D. All of the profiles indicate that the presence of D5PC lipids enhances the association free energy of dimers by 0.5-4 kcal/ mol, compared to those in the absence of lipids. Thus, more free energy is required to fully separate any given dimer when D5PC is present. We also considered the influence of lipids at different intermonomer separations. At short separation ranges, lipids do not appear to change the interaction between monomers in a consistent manner (when comparing the different dimer structures). For example, both water and lipid system profiles for ANTI and GRAMM2 have nearly identical PMF at close separations (Figure 4A, D), while the water system profiles are somewhat higher than those with lipids present for PAR and GRAMM1 at close separations (Figure 4B, C). However, above a certain separation (different for each system; the shortest being ANTI at ∼0.8 nm), the ∆G values for the lipid systems are always higher than those of the water systems, i.e., the “crossover” point. This corresponds to the separation at which lipids begin to exert their stabilizing influence. Prior to this point (i.e., at interpeptide separations below the crossover point), the main contributions to the PMF are from direct interpeptide interactions, whereas contributions to the PMF arise mainly from lipid-mediated interactions beyond the lipid/water PMF crossover points, for the lipid-present simulations. At these “intermediate” separations, lipids enhance dimer stability by promoting hydrophobic interactions between the monomers. Furthermore, we note that stabilization is greatest (by 3-4 kcal/mol) for dimer conformations whose initial structures involve aromatic contacts between F67 of one peptide and either F67 or Y63 of the other, namely ANTI (F67-Y63 pair), PAR (F67-F67), and GRAMM1 (F67-Y63) (see Figure 3). By contrast, stabilization is lowest for GRAMM2 (∼0.5 kcal/mol), in which the sole initial aromatic pairing is between two Y63 residues (Figure 3D). Thus, lipid stabilization is most effective when there are pre-existing aromatic contacts involving the highly hydrophobic phenyla-
Lipids Effect on Amyloid Fibril Formation
J. Phys. Chem. B, Vol. 113, No. 28, 2009 9451
Figure 5. Minimum contact distances for intermonomer Y63/F67 interactions with respect to COM separation for the lipid-free (black) and lipid-rich (gray) ANTI simulations.
lanine, whereas lipid-induced dimer stabilization is less effective when the sole aromatic contact is between two tyrosine residues, which has partial hydrophilic character. This suggests that lipids enhance dimer stability by increasing the hydrophobic interactions between the peptides, particularly those involving F67. We also note that the extent of dimer stabilization by lipids correlates with the intrinsic stability of the dimer conformation in the absence of lipids. To illustrate this, we consider first the relative stabilities of the dimers. An approximate ranking of the four systems by free energies at the minimum may be possible if we assume that, for large interpeptide separations (>2 nm), there exists a state of noninteracting, random-coil peptides (assumed to be similar to the present separatedmonomer states) which may serve as the ∆G ) 0 reference state for all four orientations. The free energies taken at the minima (well-depths) of the four systems produce the following ranking in decreasing order of stability. For the lipid-free systems: ANTI > PAR ≈ GRAMM1 > GRAMM2. For the lipidrich systems: ANTI > GRAMM1 g PAR > GRAMM2. The trend correlates with the extent of phenylalanine-mediated hydrophobic interactions between the two peptides (discussed above). It is also the same trend as in the extent of lipid stabilization measured as the change in ∆Gdimer in the presence of lipids. Thus, lipids impart the greatest additional stability to those conformations which are already the most stable in the absence of lipids, namely, those involving interpeptide F67 interactions. This may have implications for understanding the inhibitory action of lipids on the apoC-II60-70 peptide. For example, it can be suggested that the most stable dimer structures are not necessarily the most prone to fibrillogenesis (since lipids are able to encourage their formation and yet prevent fibril formation). Subsequently, it can be suggested that fibrillogenesis necessitates significant conformational transformations of the initial, “most stable” dimers involving transient disruption of aromatic interactions, a process which could be kinetically slowed by lipid binding. We illustrate the effects of lipids on interpeptide aromatic interactions by considering the aromatic ring contacts for the ANTI water-only and lipid-present systems. Figure 5 shows the minimum contact distances dmin (with respect to COM separation distance) between Y63 and F67 of one monomer with those of the other. Lower dmin indicates greater aromatic side chain interactions. Below ∼1 nm interpeptide separation, the persistence of interpeptide aromatic contacts is similar for both lipidrich and lipid-free systems (evidenced by similar values of dmin). However, between 1 and 2 nm separation, the aromatic contacts between the two monomers are more significant (i.e., generally
Figure 6. (A) Water radial distribution functions (RDF) with respect to peptides for the pure water system at far (black filled line) and at energy minimum (gray filled line) separations and for the lipid system at far (black dotted line) and at energy minimum (gray dotted line) separations. (B) Difference RDFs for the water (black) and lipid-present (gray) systems, indicating that water density depletion with respect to dimer formation is greater for the lipid system.
lower dmin) when D5PC is present. This suggests that lipids enhance interpeptide hydrophobic interactions, allowing the aromatic side chains to make contact with each other at a larger interpeptide separation compared to that in a lipid-free solvent environment. The enhancement of (indirect) interaction between the peptides at larger separation in the presence of lipids may be due to the more rapid formation of a larger hydrophobic cluster present in the lipid system (i.e., lipid tails and aromatic side chains) compared to a smaller cluster formed in the water-peptide system (i.e., aromatic side chains alone). At larger separations (>2 nm), there are no direct contacts between the monomers, yet lipids appear to enhance the attraction between the dimers by forming indirect, “bridging” interactions between them (see insets in Figure 4A-D, which show lipids in intermonomer spaces). The nature of this bridging interaction is hydrophobic, and the aliphatic tails of the lipids bound to each of the (separated) monomers tend to interact favorably with those of the other monomer, thus lending a driving force to dimer formation in addition to that intrinsic to the peptide. We have investigated the behavior of water with respect to interpeptide separation for both pure water and lipid-present systems by calculating the radial distribution functions (RDF, or g(r)) of all waters within the simulation cell with respect to the peptide dimer. Figure 6A shows the RDF plots for the ANTI dimer in pure water and in the presence of D5PC. For each of the plots, we overlay the water RDF calculated with respect to the “separated” peptides (i.e., at center-of-mass interpeptide separation of 2.4 nm) and with respect to the “dimerized” peptides (i.e., at the minimum energy separation of ∼0.2 nm). As expected, g(r) for the water system is overall greater than that for the D5PC system, indicating a lower level of water solvation due to lipid binding for the latter. However, examination of the differences in g(r) for the peptides at a large
9452
J. Phys. Chem. B, Vol. 113, No. 28, 2009
Figure 7. Histograms of minimum atomic contact distance between selected residues and phosphate of the D5PC lipids, calculated over all umbrella sampling windows for the ANTI (lipid present) simulations.
separation and when they are in close contact (dimerized) gives an indication of the extent to which water density changes in the vicinity of the peptide upon dimer formation. Effectively, this enables an approximation of the rate of desolvation of the peptides as they form a dimer complex. Under both solvent environments (Figure 6A), we find that dimerization results in an overall lowering of water density in the vicinity of the peptides, indicative of the depletion of peptide solvation waters upon interpeptide contact. However, the reduction in water density upon dimer formation is more significant in the presence of lipids, as can be ascertained by comparison of the plots in Figure 6A. It can be seen that the differences between g(r) at two peptide separations for the lipid system are larger than the differences between separated and dimerized g(r) for the lipidfree system. This is confirmed in Figure 6B, which shows the difference in water g(r), given by ∆g(r) ) g(r)dimerized g(r)separated under pure water (black line) and D5PC-present (gray line) environments. The D5PC system exhibits more negative ∆g(r) between ∼0.35 and 2 nm, indicating more rapid desolvation of the peptides upon dimer formation. The greater stability of the dimer in the presence of lipids (as indicated by the PMF profiles, discussed above) may partly be due to the higher peptide desolvation rate upon dimerization, which imparts greater favorable solvent entropic contributions to the dimerization free energy compared to that of the lipid-depleted system. We have furthermore examined the minimum contact distance between the phosphate headgroups of D5PC and all of the residues in the peptide, including the aromatics, tyrosine, and phenylalanine. Aromatic residue interactions with the phosphate groups are of special interest, since it has been well established that aromatic amino acids have a strong preference for binding to lipid bilayers at the interfacial region. Of particular interest is Y63, whose amphipathic character imparts a preference for binding to lipid bilayer interfaces due to its capability for H-bonding to phosphates, as well as favorable hydrophobic interactions with the lipid tail region of the bilayer. Figure 7 shows a plot of histograms of the minimum contact distance between selected residues and the phosphate headgroups, taken over all of the windows of our umbrella sampling simulation for the ANTI (with lipids) configuration. We note that examination of the minimum distances between phosphates and the residues as a function of intermolecular separation yields no obvious trend; this is to be expected, as lipids were found to interact strongly with both peptides at all interpeptide separations studied. Examination of the histograms indicate that the Hbonding-capable residues, Y, T, and S, are capable of strong,
Hung et al. persistent binding (H-bonding) to phosphates. This is manifested through the presence of a tall, narrow peak at ∼0.18 nm, indicating frequent “tight” contacts which form over short distances with low fluctuations. Y and T exhibit the strongest interactions, with S and the other polar residues exhibiting similar peaks in this region (T and Q, data not shown). However, Y exhibits a second, broader peak at ∼ 0.3 nm; all other polar residues exhibit their respective second peaks at greater distances, between 0.35 and 0.4 nm (illustrated in Figure 7 for S and T). This peak is indicative of nonspecific steric contacts. The shorter distance of this peak, for Y, is likely to be due to the use of united-atom representations for the aliphatic carbons of the nonaromatic residues. Taking this into account, the interaction between Y and phosphates is overall similar to that of T, with stronger nonspecific interactions but weaker specific interactions. All of the polar residues form tighter contacts than the nonpolar residues; the latter do not exhibit sharp, narrow peaks ∼0.18 nm, but do exhibit broader peaks at greater separations. For F, the steric interaction occurs at ∼0.28 nm (similar to Y), while for other nonpolar residues (illustrated for I in Figure 7), the peak is at ∼0.35 nm. This difference may again be attributed to the use of united-atoms for aliphatic carbons but explicit aromatic hydrogens for F. However, the nonspecific peak for F is narrower than that for other nonpolar residues. For example, I65, while exhibiting a peak at ∼0.35 nm, has a greater dispersion of long-distance contacts, indicating that phosphates do not make as much persistent contacts with it as compared to F67. In summary, Y interacts with phosphates to a similar extent as T and binds more strongly than other polar residues like S and Q. For the nonpolar residues, F forms more persistent nonspecific contacts with phosphates compared to other hydrophobic residues. Overall, we find that aromatics interact favorably with the charged phosphate groups. Our molecular simulation results revealed that dimerization free energy profiles for all four dimer configurations studied consistently indicate the enhancement of interpeptide association under lipid-rich conditions, consistent with the mechanism suggested by sedimentation ultracentrifugation analysis, namely, that solvated lipids may serve to kinetically trap dimers and higher-order oligomers by raising their free energy barriers of dissociation, including oligomers which are in nonfibrillogenic forms. The kinetic trapping of nonfibrillogenic dimers/oligomers renders them less likely to dissociate and reform in conformations more prone to subsequent fibril nucleation and growth. In light of the demonstrated cytotoxicity of amyloid peptide oligomers, our results suggest that, by enhancing the stability of oligomeric peptide species, the presence of solvated lipids may contribute to the cytotoxicity of fibrillogenic proteins and peptides 4. Conclusion In summary, we have investigated the effect of submicellar lipids on amyloid fibril formation by apoC-II60-70. ThT fluorescence studies showed that submicellar levels of the shortchain phospholipids, D5 and D6PC, strongly inhibited amyloid fibril formation by the peptide. In contrast, sedimentation equilibrium analysis of these peptide-lipid mixtures, as well as molecular dynamics simulations of dimer formation, indicated the presence and stability of soluble oligomeric complexes. This suggests that one mechanism by which lipids inhibit fibrillization is by increased stabilization of nonfibrillogenic oligomers. We note that the relatively short time scale trajectories do not allow significant conformational conversion from the initial structures. However, our simulations enabled the determination
Lipids Effect on Amyloid Fibril Formation of the influence of lipids on stabilization of a number of distinct, “pre-selected” dimer conformations. Our results are therefore focused on the effects of lipids on the initial association propensity between two peptides, prior to dimerization-induced conformational transition. By taking this approach, we were able to acquire an insight into the mechanism by which lipids stabilize dimers by comparing the extent of stabilization of the different dimer structures. It is of considerable interest to note that, contrary to their effects on the peptide apoC-II60-70, submicellar PC lipids enhance the fibrillization of the 79-residue full-length protein rather than inhibit it. The underlying mechanism for this fibrillization enhancement is unclear. However, based on the results reported here, we advance the following hypothesis. For the full-length protein, lipids in solution interact and bind to regions rich in aromatic side chains (including 60-70). As in the case of peptides, this results in enhanced interprotein association, especially promoting interprotein contact at the lipid-bound regions. However, this would also enhance the rate of interactions between the nonlipid-bound, fibrillogenic segments from the different monomers since the proteins are brought together more rapidly in the presence of lipids. This would subsequently lead to enhanced fibril growth compared to a lipid-free environment. The results from our present work also demonstrate the potential impact of lipids on amyloid forming pathways and the accumulation of toxic oligomeric intermediates. It has been postulated that the capacity for certain proteins to form insoluble fibrils is an evolutionary protective mechanism against the formation of persistent, toxic misfolded protein aggregates,28 which have been shown to exhibit greater cytotoxicity than mature amyloid fibrils.3-5 Given the possible stabilization of nonfibril-favoring peptide oligomers by lipids suggested by our current results, and the known positive correlation between the presence of small oligomeric species and cytotoxicity,28 it is possible that the presence of free (nonbilayer, nonmicellar) lipids may in fact promote the cellular toxicity of fibrillogenic peptides, despite their capability to inhibit fibrillization. The enhancement of cytotoxicity resulting from inhibition of fibril formation must therefore be taken into consideration when devising therapies against misfolding-related disorders, as has been recognized for, for example, the Aβ peptide.29 Acknowledgment. We thank the Australian and Victorian Partnerships for Advanced Computing (APAC and VPAC) for computational resources, the latter for funds through the eResearch Grants scheme, and our colleagues at RMIT University (Sue Legge, Nevena Todorova, and Akin Budi). In addition, we thank the anonymous reviewers for their insightful comments and helpful suggestions.
J. Phys. Chem. B, Vol. 113, No. 28, 2009 9453 References and Notes (1) Bucciantini, M.; Giannoni, E.; Chiti, F.; Baroni, F.; Formigli, L.; Zurdo, J.; Taddei, N.; Ramponi, G.; Dobson, C. M.; Stefani, M. Nat. Biotechnol. 2002, 416, 507. (2) Nelson, R.; Eisenberg, D. Curr. Opin. Struct. Biol. 2006, 16, 260. (3) Kirkitadze, M. D.; Bitan, G.; Teplow, D. B. J. Neurosci. Res. 2002, 69, 567. (4) Walsh, D. M.; Klyubin, I.; Fadeeva, J. V.; Rowan, M. J.; Selkoe, D. J. Biochem. Soc. Trans. 2002, 30, 552. (5) Kayed, R.; Head, E.; Thompson, J. L.; McIntire, T. M.; Milton, S. C.; Cotman, C. W.; Glabe, C. G. Science 2003, 300, 486. (6) Berger, O.; Edholm, O.; Jahnig, F. Biophys. J. 1997, 72, 2002. (7) Stewart, C. R.; Haw, A.; Lopez, R.; McDonald, T. O.; Callaghan, J. M.; McConville, M. J.; Moore, K. J.; Howlett, G. J.; O’Brien, K. D. J. Lipid Res. 2007, 48, 2162. (8) MacRaild, C. A.; Hatters, D. M.; Howlett, G. J.; Gooley, P. R. Biochemistry (Mosc). 2001, 40, 5414. (9) Zdunek, J.; Martinez, G. V.; Schleucher, J.; Lycksell, P. O.; Yin, Y.; Nilsson, S.; Shen, Y.; Olivecrona, G.; Wijmenga, S. Biochemistry (Mosc). 2003, 42, 1872. (10) MacRaild, C. A.; Howlett, G. J.; Gooley, P. R. Biochemistry (Mosc). 2004, 43, 8084. (11) Hatters, D. M.; Howlett, G. J. Eur. Biophys. J. 2002, 31, 2. (12) Griffin, M. D. W.; Mok, M. L. Y.; Wilson, L. M.; Pham, C. L. L.; Waddington, L. J.; Perugini, M. A.; Howlett, G. J. J. Mol. Biol. 2008, 375, 240. (13) Hatters, D. M.; Lawrence, L. J.; Howlett, G. J. FEBS Lett. 2001, 494, 220. (14) Griffin, M. D.; Mok, M. L.; Wilson, L. M.; Pham, C. L.; Waddington, L. J.; Perugini, M. A.; Howlett, G. J. J. Mol. Biol. 2008, 375, 240. (15) Hung, A.; Griffin, M. D.; Howlett, G. J.; Yarovsky, I. Eur. Biophys. J. 2008, 38, 99. (16) Van der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H. J. C. J. Comput. Chem. 2005, 26, 1701. (17) Lindahl, E.; Hess, B.; van der Spoel, D. J. Mol. Model. 2001, 7, 306. (18) Gunsteren, W. F. v.; Kruger, P.; Billeter, S. R.; Mark, A. E.; Eising, A. A.; Scott, W. R. P.; Huneberger, P. H.; Tironi, I. G. Biomolecular Simulation: The GROMOS96 Manual and User Guide; Biomos & Hochschulverlag AG an der ETH Zurich: Groningen & Zurich, 1996. (19) Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089. (20) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. J. Comput. Chem. 1997, 18, 1463. (21) Berendsen, H. J. C.; Postma, J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684. (22) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33. (23) Hwang, W.; Zhang, S. G.; Kamm, R. D.; Karplus, M. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12916. (24) Katchalskikatzir, E.; Shariv, I.; Eisenstein, M.; Friesem, A. A.; Aflalo, C.; Vakser, I. A. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 2195. (25) Tarus, B.; Straub, J. E.; Thirumalai, D. J. Am. Chem. Soc. 2006, 128, 16159. (26) Berendsen, H. J. C.; Postma, J. P. M.; Gunsteren, W. F. v.; Hermans, J. Intermolecular Forces; Pullman, B., Ed.; Reidel: Dordrecht, 1981; p 331. (27) Roux, B. Comput. Phys. Commun. 1995, 91, 275. (28) Cohen, E.; Bieschke, J.; Perciavalle, R. M.; Kelly, J. W.; Dillin, A. Science 2006, 313, 1604. (29) Bagriantsev, S.; Liebman, S. BMC Biol. 2006, 4.
JP901051N
