Coarse Grained Molecular Dynamics of Engineered Macromolecules

Jun 5, 2013 - Department of Chemical and Biochemical Engineering, Rutgers ... the NLB mucic acid head resulted in enhanced interaction with the propos...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/Biomac

Coarse Grained Molecular Dynamics of Engineered Macromolecules for the Inhibition of Oxidized Low-Density Lipoprotein Uptake by Macrophage Scavenger Receptors Michael D. Tomasini,† Kyle Zablocki,† Latrisha K. Petersen,† Prabhas V. Moghe,†,‡ and M. Silvina Tomassone*,‡ †

Department of Biomedical Engineering, Rutgers University, Piscataway, New Jersey 08854, United States Department of Chemical and Biochemical Engineering, Rutgers University, Piscataway, New Jersey 08854, United States

‡

ABSTRACT: Atherosclerosis is a condition resulting from the accumulation of oxidized low-density lipoproteins (oxLDLs) in arterial walls. Previously developed macromolecules consisting of alkyl chains and polyethylene glycol (PEG) on a mucic acid backbone, termed nanolipoblockers (NLBs) are hypothesized to mitigate the uptake of oxLDL by macrophage scavenger receptors. In this work, we developed a coarse grained model to characterize the interactions between NLBs with a segment of human scavenger receptor A (SR-A), a key receptor domain that regulates cholesterol uptake and foam cell conversion of macrophages, and studied NLB ability to block oxLDL uptake in PBMC macrophages. We focused on four different NLB configurations with variable molecular charge, charge location, and degree of NLB micellization. Kinetic studies showed that three of the four NLBs form micelles within 300 ns and of sizes comparable to literature results. In the presence of SR-A, micelle-forming NLBs interacted with the receptor primarily in an aggregated state rather than as single unimers. The model showed that incorporation of an anionic charge near the NLB mucic acid head resulted in enhanced interaction with the proposed binding pocket of SR-A compared to uncharged NLBs. By contrast, NLBs with an anionic charge located at the PEG tail showed no interaction increase as NLB aggregates were predominately observed to interact away from the oxLDL binding site. Additionally, using two different methods to assess the number of contacts that each NLB type formed with SR-A, we found that the rank order of contacts coincided with our experimental flow cytometry results evaluating the ability of the different NLBs to block the uptake of oxLDL.

■

ability to efficiently metabolize oxidized LDL.4 Excessive uptake of oxidized LDLs results in the formation of foam cells that become engorged, die, and subsequently form atherosclerotic plaques. These plaques grow in size and can occlude the arterial lumen, obstructing blood flow and ultimately resulting in an infarction. Kodama and co-workers purified bovine SR-A and determined the structure to be a trimer composed of five or six structural domains dependent on alternate RNA splicing: a cytoplasmic N-terminal domain, a membrane-spanning region, a spacer region, an α-helical coiled-coil domain, a collagen-like domain, and a cysteine-rich C-terminal domain present in SR-A type I but not SR-A type II receptors.5,6 Truncation studies implicated the α-helical coiled-coil domain in trimer formation and showed that the collagen-like domain mediates ligand recognition and binding.7 Site-directed mutagenesis uncovered a conserved cluster of positively charged amino acids at the distal end of the collagenous domain: Lys327, Lys334, Lys337,

INTRODUCTION Atherosclerosis is characterized by the buildup of lipid-rich plaques in arterial walls leading to cardiovascular disease, a major cause of death in the United States.1 The pathology of atherosclerosis can be broken into two parts: the accumulation of oxidized low-density lipoproteins (LDLs) in the vascular intima followed by an inflammatory signaling cascade resulting in plaque development, arterial blockage, and thrombosis. Atherosclerosis begins when circulating LDLs are transported into the intimal space and become trapped in the extracellular matrix where they are subject to oxidative modification. Oxidized LDLs cause damage to surrounding endothelial cells and initiate an inflammatory signaling cascade that results in the recruitment of monocytes to the site of injury.2 Monocytes differentiate into macrophages that express a range of different surface scavenger receptors (SRs) active in the nonspecific uptake of modified LDLs. Two primary SRs control the majority (75−90%) of the modified LDL uptake in macrophages: scavenger receptor type A (SR-A) and the type B scavenger receptor, CD36.3 Unlike receptors for native LDL, SRs are not down-regulated by increased cholesterol concentrations within the cell, and macrophages do not have the © 2013 American Chemical Society

Received: November 13, 2012 Revised: May 31, 2013 Published: June 5, 2013 2499

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

■

Article

MODEL AND SIMULATION DETAILS Simulation System. The simulation system for the functional interaction of differently structured NLBs with SRA included the collagenous domain of SR-A, 50 NLB molecules, adequate concentration of sodium and chloride ions for an electrically neutral system, and approximately 100 000 CG water particles. The NLB concentration was equivalent to 5.9 × 10−3 M, well above the experimental critical micelle concentration (∼1.3 × 10−6 M). The components of the simulation were described using the CG MARTINI force-field developed by Marrink et al. which has been shown to accurately reproduce structural and thermodynamic properties of various systems including lipid, protein, sugar, and polymer systems.22−25 In the MARTINI force-field, groups of atoms are lumped together into CG interaction sites (beads) which are parametrized to retain the underlying molecular detail of the corresponding atoms. Beads are generally composed of four heavy atoms (not hydrogen) with an effective diameter of 0.47 nm. When a finer level of detail is necessary, as in the case with ring-like structures, two to three heavy atoms with a diameter of 0.43 nm comprise an interaction site. Beads are grouped into four types: polar (P), nonpolar (N), apolar (C), and charged (Q) with further subtypes of each bead available to address the degree of polarity (1 = lower polarity and 5 = higher polarity) and the ability for hydrogen bonding (d = donor, a = acceptor, da = donor and acceptor, 0 = none). Nonbonded interactions between beads are treated with a shifted Lennard-Jones (12−6) potential energy function:

and Lys340 of bovine SR-A (corresponding to Arg325, Lys332, Lys335, and Lys338 in human SR-A) forming a positively charged groove that interacts electrostatically with anionic oxidized LDL.7,8 Further mutation studies by Andersson and Freeman9 proposed additional cationic residues (such as Arg317) nearer to the central and N-terminal regions of the SR-A collagenous domain that may also be involved in binding oxidized LDL. One approach hypothesized to counteract the formation of atherosclerotic plaques relies on inhibiting cellular uptake of oxidized LDL using agents selectively designed to bind SRA.10,11 Tian et al.12 developed a class of anionic, micelleforming molecules comprised of a mucic acid headgroup with attached aliphatic chains and a long polyethylene glycol (PEG) tail that were subsequently found to be effective in blocking oxidized LDL uptake in murine macrophages.13−17 Plourde et al.18 explored the structure−function relationship of these macromolecular inhibitors (referred to henceforth as nanolipoblockers (NLBs)) to determine the influence of charge and charge location upon the ability of NLBs to inhibit oxidized LDL uptake by SR-A. The authors used a combined experimental and computational (molecular dynamics (MD) docking techniques) approach and reported that NLBs containing hydrophobic-bound carboxylate groups were the most efficient inhibitors of oxidized LDL uptake. Although the authors did find agreement between experimental and computational results, the all-atom force field used for molecular docking limited the size of the computational system to the collagen-like domain of SR-A and one truncated NLB unimer, thus precluding insights into interactions of NLB assemblies. Experimental studies have shown that the assembly of NLBs is a likely precursor event that can influence the efficacy of SR-A binding and inhibition of oxidized LDL uptake, as demonstrated through micellization18,19 and kinetically assembled nanoparticles.20 This indicates that the unitary interactions of NLBs with SR-A may not be sufficient to fully capture the inhibitory functional behaviors of NLBs. In this work we developed a coarse-grained (CG) MD approach based on the MARTINI force-field of Marrink and co-workers.21,22 CG MD simulation is a model that allows for the simulation of larger complexes on closer to physiologic time-scales at the expense of neglecting the fine atomistic detail of the underlying system. In this work, we employ CG MD to probe large enough system sizes and time-scales to study the aggregation behavior of complete NLB molecules and the binding of NLB aggregates to the collagenous domain of SR-A, alleviating limitations of earlier structure−function studies. We assessed the micelle-forming capabilities of different NLBs and probed the specific interactions between NLBs and SR-A to characterize the mechanism by which preaggregates of NLBs may influence the inhibition of oxidized LDL uptake by SR-A model receptors. As a point of comparison for molecular simulation studies of the interaction of NLBs with SR-A, we performed flow cytometry to obtain the dose-dependent ability of differently structured NLBs to inhibit oxidized LDL uptake in peripheral blood mononuclear cells (PBMCs) derived macrophages, finding good agreement between the model and experimental results. Our findings suggesting the mechanism of interaction between NLBs and SR-A may be useful in guiding development of more effective inhibitors to oxidized LDL uptake by SR-A.

⎡⎛ σij ⎞12 ⎛ σij ⎞6 ⎤ ULJ(r ) = 4ϵij⎢⎜ ⎟ − ⎜ ⎟ ⎥ ⎝r ⎠⎦ ⎣⎝ r ⎠

where εij is the energy of interaction and σij is the distance of closest approach between particles i and j. Electrostatic interactions are treated with a screened Coulomb potential: qiqj Uel(r ) = 4πεrε0 where qi and qj are the charge on particles i and j, ε0 is the permittivity of free space, and εr is the relative dielectric constant. Bond and angle interactions are described by the following harmonic and cosine-harmonic potentials: Ubond =

1 k b(r − r0)2 2

Uangle =

1 kθ(cos(θ) − cos(θ0))2 2

with equilibrium bond distance r0 and angle θ0 and bond force constant kb and angle force constant kθ. Proper dihedral angles are treated with a periodic cosine function to prevent out-ofplane distortions of planar groups: Upd = kd[1 + cos(nψijkl − ψd)]

where kd is the force constant, n is the multiplicity, and ψd is the equilibrium proper dihedral angle. Improper dihedrals are implemented to maintain protein secondary structure23 and described by a harmonic potential: Uid = k id(ψijkl − ψid)2

with kid the force constant and ψid the equilibrium improper dihedral angle. 2500

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

Figure 1. Amino acid sequence of the collagen-like domain of SR-A used in simulations. The proposed binding pocket for oxidized LDL is underlined in bold with the positively charged residues highlighted in gray. Other cationic residues outside of the binding pocket are also highlighted in gray.

Figure 2. Chemical structure and corresponding CG representation of the four different NLBs simulated in this work. 0CM is a neutral NLB with no carboxylate group (top-left); 1CM has a carboxylate group near the aliphatic head of the molecule (top-right); 1CP has the carboxylate group moved to the PEG tail (bottom-left); PEG-COOH contains only a carboxylate headgroup and a PEG tail (bottom-right).

The pdb2cg.awk script from the MARTINI Web site28 was used to convert the atomistic protein into a CG MARTINI representation. The resultant CG structure of SR-A was energy minimized and equilibrated in explicit solvent. Parameterization of NLBs. Structures of the four different types of NLBs are given in Figure 2 along with their CG representation. A total of seven different types of interaction sites were needed to describe all NLBs: (i) Na accounted for the mucic acid backbone beads as they were equivalent to methylformate fragments (ii) C1 for aliphatic groups,22 (iii) SN0 for PEG residues,29 (iv) N0 for the terminal PEG groups representing a methoxyethane group, (v) SNa for the mucic acid to PEG linker region, (vi) P3 for the propyl formamide head groups of 0CM and 1CP, and (vii) Qa for the negatively charged carboxylate groups. Standard MARTINI parameters were used for the non-PEG portion of the molecules: bonded

Parameterization of SR-A. For simplicity, our simulations only considered the collagenous domain of SR-A given the importance of this region in mediating binding of oxidized LDL.7−9 Due to the CG nature of the MARTINI force-field, in addition to the nonbonded and bonded interactions between the protein particles, the force-field includes additional terms used to describe protein secondary structure. As the exact three-dimensional structure of the SR-A receptor is not known, we modeled the collagenous domain of SR-A as a collagen triple-helix using the parameters developed by Gautieri et al.26 Residues 272−341 that make up the collagenous domain of human SR-A are given in Figure 1 (here relabeled 1−70 in this work). A suitable starting structure of the protein was built by first mapping an atomistic representation of the collagen-like domain of SR-A onto an ideal collagen triple helix using the software THeBuScr (Triple-Helical collagen Building Script).27 2501

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

interactions (Rb = 0.47, kb = 1250 kJ mol−1 nm−2), and a harmonic angle potential was used to keep the aliphatic chains linear (θ = 180°, kθ = 25 kJ mol−1). The bonded parameters for the PEG tails used the values of Lee et al.25 Simulation Details. Simulations were performed with the GROMACS software package version 4.5.5.30 In all simulations, in accordance with the MARTINI force-field, Lennard-Jones interactions were smoothly shifted to zero between a distance of 0.9 and 1.2 nm, and electrostatics were smoothly shifted to zero between 0 and 1.2 nm. The nonbonded neighbor list was updated every 10 steps with a neighbor list cutoff of 1.4 nm. All simulations were performed in an NPT ensemble with the system coupled to a Berendsen thermostat31 at 310 K with a coupling constant of τT = 1.0 ps. Berendsen pressure coupling31 was used to maintain the system at a pressure of 1.0 bar using a coupling constant of τP = 2.0 ps and a compressibility of 3 × 10−5 bar−1. The integration time step was 10 fs. The CG nature of the MARTINI force field yields a smoother energy landscape that has been shown to result in 4-fold faster dynamics.21 Our results report the effective time incorporating this scale-up factor. All analyses were performed using GROMACS tools, and visualizations were created using VMD.32 Initial simulations were performed with NLBs in the absence of SR-A to characterize the assembly of the different NLB types into micelles. Each system contained 25 molecules randomly placed in a simulation cell and solvated with 70 000 water beads (280 000 water molecules) and run for 800 ns. Subsequent simulations of the interaction of NLBs with SR-A contained 50 NLB molecules equilibrated for 400 ns to which was added equilibrated SR-A resulting in a total system size of approximately 20 × 20 × 35 nm3. As it has been shown that the α-helical coiled-coil domain and not the collagenous domain is responsible for trimer formation in SR-A,7 position restraints were applied to the backbone beads of the protein with a restraint force of 1000 kJ mol−1 nm−2 to prevent unraveling of the triple helix. The combined SR-A and NLB simulation was then allowed to equilibrate for 800 ns followed by 1 μs of simulation time.

■

packs for 15 min. Cells were transferred to polystyrene tubes (BD Falcon), centrifuged (1000 rpm) for 10 min at 4 °C, and resuspended in a solution of PBS containing 0.1% sodium azide, 0.5% bovine serum albumin, and 1% normal goat serum. Cells were centrifuged and resuspended twice before fixing with 1% paraformaldehyde. The samples were then analyzed using a Becton-Dickenson FACSCaliber flow cytometer to determine the extent of oxidized LDL uptake. After collecting 10 000 events per sample, data was processed using FlowJo software (Tree Star Inc.).

■

RESULTS Dose−response of NLBs with PBMC Derived Macrophages. The dose-dependent ability of different NLBs to inhibit the uptake of oxidized LDL in PBMC macrophages was examined by incubating cells with varying concentrations of NLBs and DiO-labeled oxidized LDL for 24 h under serum-free conditions (Figure 3). At lower doses (10−6 − 10−7 M) 1CM

Figure 3. Dose−response curves for the inhibition of oxidized LDL uptake by PBMC-derived macrophages following 24 h incubation with varying concentrations of NLBs and 5 μg/mL oxidized LDL in serumfree conditions. Error bars represent the bounds of duplicate experiments.

inhibited the uptake of oxidized LDL the most followed by 0CM and then 1CP. At higher doses (10−4−10−5 M), there was little difference between 1CM, 0CM, and 1CP, with all reaching roughly 100% inhibition of oxidized LDL uptake at a dosage of 10−4 M. Conversely, PEG-COOH showed no inhibition of oxidized LDL uptake, actually increasing the amount of oxidized LDL uptake for doses greater than 10−7 M. The responses for 1CM, 0CM, and 1CP were fit to a standard dose−response curve to calculate of the EC50 of the three NLBs (Table 1). The EC50 is the NLB concentration necessary to produce half of the maximal response and can be used to define a potency of each NLB. It was not possible to measure an EC50 for PEG-COOH since oxidized LDL inhibition decreased with increasing NLB concentration; however, it is clearly the poorest

MATERIALS AND METHODS

Cell Culture. PBMCs were isolated from human buffy coats (Blood Center of New Jersey) by centrifugation through Ficoll-Paque density gradient (GE Healthcare). PBMCs were plated into T-175 flasks (Corning), and mononuclear cells were selected via adherence to plastic after 24 h by washing thrice with phosphate buffered saline (PBS). Monocytes were cultured for 7 days in RPMI 1640 media (ATCC) supplemented with 10% fetal bovine serum, 1% penicillin/ streptomycin, and 50 ng/mL M-CSF (Peprotech) for differentiation into macrophages. Dose−Response Effects of NLBs to Oxidized LDL Uptake. The ability of NLBs to inhibit the uptake of oxidized LDL was studied by performing a dose−response assay using four differently structured NLBs. The different NLB molecules were named in accordance with Plourde et al.18 1CM contained a carboxylate group near the mucic acid head. 0CM was an uncharged NLB where the carboxylate group of 1CM was changed to type Na. 1CP had a carboxylate group attached to the PEG tail and a headgroup equivalent to that of 0CM. PEG-COOH consisted only of a carboxylate headgroup attached to a PEG tail with the aliphatic groups from other NLBs absent. The structure of each NLB type is given in Figure 2. PBMC macrophages were coincubated with 5 μg/mL of 3,3′-dioctadecyloxacarbocyannine (DiO) labeled oxidized LDL (Kalen Biomedical) and NLBs (10−7 M to 10−4 M) for 24 h in serum-free RPMI 1640 media. The adherent PBMCs were removed from culture dishes by treatment with 2 mM ethylenediaminetetraacetic acid (Invitrogen) and incubation on cold

Table 1. EC50 Values from the Dose-Response of 1CM, 0CM, and 1CPa NLB

EC50 (μM)

1CM 0CM 1CP

0.28 ± 0.04 0.54 ± 0.10 1.17 ± 0.34

a

Errors in the EC50 values represent the bounds of duplicate experiments. PEG-COOH was not included in the EC50 calculations. 2502

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

Figure 4. Time-lapse trajectories illustrating the self-assembly processes of NLBs. Each system began with 25 randomly distributed NLB unimers and proceeded for 800 ns. NLBs containing aliphatic chains (1CM, 0CM, and 1CP) assembled into micelles within 200−400 ns, whereas PEG-COOH remained as unimers. The particles colored red are the PEG tails, and those colored in cyan are the mucic acid and the aliphatic groups.

(∼1.3 × 10−6 M)12 except for PEG-COOH, which does not assemble into micelles. We validated the NLB parametrization scheme using comparison of simulations to experiment measurements on micallar size and the experimental diffusion coefficient of 1CM in water. The micelle radius was found by measuring the root-mean-square (RMS) distance from the terminal NLB tail beads to the micelle center of mass (COM). The computed micelle radii for NLBs containing 2000 and 5000 MW PEG tails are reported in Table 2. For the case of 2000 MW PEG, the micelle radii ranged from 4.15 to 5.45 nm. Additional self-assembly simulations of 1CM with 114 PEG repeats (5000 MW) resulted in micelles with radii 7.20 ± 1.62

inhibitor of oxidized LDL uptake, and as such can be defined as having the lowest potency of the four NLBs. Therefore, the rank order of NLB potency was 1CM > 0CM > 1CP > PEGCOOH. In the remainder of this work, we will attempt to correlate the NLB to SR-A interaction strength with the NLB potency under the assumption that increasing NLB potency is the result of enhanced interaction of the NLB with SR-A. NLB Micelle Formation. The micellization process of the NLBs was explored by performing simulations of NLB selfassembly using 25 molecules randomly placed in a water box. The NLB region containing the aliphatic groups is referred to as the head of the molecule, and the PEG portion as the tail. This is opposite of common micelle convention based on lipid molecules in which the hydrophilic portion is labeled as the headgroup and the hydrocarbon chains as the tails. As such, the NLB head region lies within the micelle interior, and the tails are in the periphery exposed to the aqueous solvent. The timelapse simulation of each NLB type is shown in Figure 4. 0CM, 1CM, and 1CP formed stable aggregates after 200−400 ns, while PEG-COOH did not aggregate and remained primarily as unimers. In all simulations, the NLB concentration was much greater than the experimental critical micelle concentration

Table 2. Cluster Analysis of the Micelle-Forming NLBs

number of clusters molecules per cluster max cluster size RM: 2000MW PEG (nm) RM: 5000MW PEG (nm) 2503

0CM

1CM

1CP

2.96 ± 0.20 8.50 ± 0.83 11 4.15 ± 0.62 -

4.99 ± 0.10 5.01 ± 0.12 10 5.13 ± 0.38 7.20 ± 1.62

4.83 ± 0.38 5.21 ± 0.47 11 5.45 ± 1.20 -

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

simulations were performed of four different types of NLB molecules in the presence of the collagenous portion of SR-A. For each NLB type, 50 molecules (5.9 × 10−3 M) were combined with an equilibrated trimeric SR-A molecule and tracked for a total of 1 μs. The final configurations for all NLBs interacting with SR-A are shown in Figure 6a−d. For micelle-

nm. The decreased PEG tail length of our simulations with relation to the NLBs used experimentally19,33 are not expected to markedly affect the binding modes of NLBs to SR-A. Tian et al.12 measured NLBs with both 2000 MW and 5000 MW PEG tails and found the overall micelle radii to be between 5 and 10 nm, which corroborates our results. As a further validation of our parametrization scheme, we measured the diffusion of a 1CM micellar aggregate and found D1CM = 7.72 × 10−7 cm2/s, which is fairly close to the experimental diffusion coefficient of 1CM (4.16 × 10−7 cm2/s) as measured by dynamic light scattering. The slight difference in values is likely the result of the simulations having been performed at body temperature, 310 K, whereas the experimental dynamic light scattering was performed at room temperature, 293 K. We quantified the aggregation behavior of the different NLBs using a cluster analysis in which two molecules were defined as belonging to the same cluster if the COM of their aliphatic beads (Type C1 in Figure 2) were within 2.5 nm of each other, the distance at which the radial distribution function exhibited a minimum. The self-assembly results presented in Table 2 give the largest and average cluster aggregation numbers for each micelle forming NLB. A complete probability plot of the micelle aggregation number would require several simulations of self-assembly for each NLB to obtain proper statistics. This calculation is computationally expensive and beyond the scope of this work; however, it is the focus of future work. 1CM formed stable aggregates of 5 micelles after approximately 300 ns as shown by the cluster analysis in Figure 5. Similar

Figure 6. Interaction following 1 μs of simulation of SR-A with each type of NLB (clockwise starting in the lower-left) 1CM, 0CM, 1CP, and PEG-COOH. In all snapshots, the orientation of SR-A is the same with the N-terminal region at the left and the C-terminal region on right. The protein is shown in green and yellow, the NLBs are in red and cyan.

forming NLBs (0CM, 1CM, 1CP), the interaction with SR-A was preferentially in the form of micellar aggregates rather than NLBs in a monomeric form. PEG-COOH, which did not form micelles, interacted as single unimers with the PEG tail wrapped around the protein. In Figure 6a 0CM interacts with one large micelle interacting with the N-terminal region of SRA, while a smaller micelle is present toward the C-terminal end. Also to note is that there were two small micelles observed in solution that did not appear to interact with the protein. 1CP aggregates interacted with SR-A near the C-terminal end with a small NLB aggregate around the central region of the protein. Near the N-terminal region, there existed only unimer interaction of 1CP with SR-A. Conversely, the 1CM simulation resulted in two large micelles interacting with both the Nterminal and C-terminal ends of SR-A. The observation that 1CM had enhanced interaction over 1CP and 0CM with the Cterminal end of the collagenous domain, which contains the proposed binding site for oxidized LDL, is consistent with our above results showing 1CM to be a more potent inhibitor of oxidized LDL uptake as well as previously published reports relating NLBs and oxidized LDL uptake.18,33 Finally PEGCOOH interacted as monomers somewhat nonspecifically along the protein, wrapping around SR-A with a number of PEG-COOH molecules remaining in solution (Figure 6d). Contacts between NLB Molecules and SR-A. Subsequent analysis focused on quantitatively assessing the extent of interaction between the four different NLBs and SR-A by

Figure 5. Evolution of the cluster distribution for 1CM indicating that 1CM aggregated into micelles after approximately 300 ns with an average of five molecules per micelle.

trajectories and cluster formation were observed for 0CM and 1CP (data not shown). After 800 ns, 1CM established five clusters with an average of five molecules per cluster with the largest cluster containing 10 unimers. The cluster features at the end of the 800 ns simulation are elaborate. Resulting from the definition of a cluster, an NLB unimer in solution was counted as an individual cluster with 1 molecule per cluster. Therefore to obtain an accurate measurement of the cluster size, only the largest cluster was measured. The micelle radii of gyration were measured for each NLB type in their principal axes to provide for an estimation of micelle shape. 1CM and 0CM were somewhat ellipsoidal with principal axis ratios of (1: 1.16: 1.26) and (1: 1.14: 1.24) for 1CM and 0CM, respectively. 1CP was highly ellipsoidal with ratios of (1: 2.10: 2.16). Interaction of NLBs with SR-A. To probe the specific interactions of different NLB architectures with SR-A, 2504

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

Figure 7. Distance distribution of NLB molecules from the surface of SR-A. On the right are enlarged insets of the initial peak (top-right) and further from the protein surface (bottom-right).

Figure 8. Contact map of interactions between NLBs and SR-A. All panels show the collagenous domain of SR-A in a ball-and-stick with the residues from Figure 1 (charged binding pocket) represented as spheres. The protein beads are all colored according to the number of contacts that each bead makes with NLB molecules. A contact is defined as an NLB molecule within 2 nm of the protein bead (residue) that persists for greater than 75% of the simulation time. The atoms shown in blue interact with 0−3 NLB molecules, those colored in white interact with 4−7 NLB molecules, and those colored in red interact with greater than 7 NLBs for more than 75% of the simulation time.

quantifying the distance the NLB molecules resided from the protein surface. The distance distributions of each NLB type from any part of the collagenous domain of SR-A are plotted in Figure 7. The majority of NLBs had some portion of the molecule within 0.5 nm of SR-A as evidenced by the large initial peak in the distributions. Looking closer at the relative sizes of the peaks for each NLB (inset of Figure 7) shows the peak height to follow the trend 1CM > 0CM > 1CP > PEG-COOH.

Assuming that increased blocking of the SR results from a higher concentration of NLBs near the receptor surface, our simulation results agree with the experimental order of NLB potency. The distance distribution for 1CM is such that all molecules were within a distance of 4 nm from SR-A. This is in contrast to the other NLBs, which have secondary peaks in the density distribution in the range of 6 to 8 nm, signifying NLBs in solution away from the protein surface. This is consistent 2505

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

overall average number of persistent contacts for each NLB falls in the order 1CM > 0CM > 1CP > PEG-COOH, equivalent to the experimentally determined NLB potency order. Sustained interaction of 0CM with SR-A occurred primarily near the Nterminal region (away from the proposed binding pocket) and extended into the middle of the protein with minimal interaction at the C-terminal end. The only cationic residue of the SR-A binding pocket (as given in Figure 2) that 0CM showed a high interaction with was Arg54. 0CM had a low interaction with all C-terminal lysines (Lys61, Lys64, and Lys67). Compared to 0CM, 1CM had more persistent interactions with the middle of SR-A, including some strong interactions with Arg46, Arg54, and Lys61; moderate contacts with Lys64; and only a weak interaction with Lys67. Overall, 1CM had the most contacts with SR-A consistent with the binding energy simulations of Plourde et al.18 1CP showed minimal strong contacts with SR-A focused only in the Nterminal region. There was a moderate interaction with Arg46, but a low number of persistent interactions with the other cationic residues of the binding pocket. PEG-COOH exhibited the least contact with SR-A, with none of the regions having more than 7 persistent contacts and only weak interactions with all residues of the binding pocket. The minimal interaction of PEG-COOH with SR-A as a whole may explain why it was the least effective of the four NLBs at inhibiting oxidized LDL uptake (Figure 3). Interaction of NLBs with the Binding Pocket of SR-A. It is hypothesized that the residues important in oxidized LDL binding to human SR-A are Arg317, Arg325, Lys332, Lys335, and Lys338,7,9 which correspond to R46, R54, K61, K64, and K67 in our studies (Figure 1). To examine the electrostatic interactions between these residues and NLBs, the number of contacts between residues of the charged binding pocket and either the head (negatively charged terminal bead of 1CM and PEG-COOH) or the tail (negatively charged in 1CP) was

with the system snapshots of Figure 6, which show aggregates (or unimers in the case of PEG-COOH) in solution away from the protein for both 0CM and 1CP. The substantial number of NLBs residing away from the protein may indicate a lower affinity of 0CM and 1CP for SR-A than 1CM. In addition to the overall interaction of NLBs with the surface of SR-A, specific residues of the receptor with which the NLBs primarily interact were computed to determine whether different NLB types preferentially interact with the proposed LDL binding site on SR-A. Figure 8 shows a contact map in which SR-A is colored according to the number of persistent contacts each CG bead of SR-A makes with NLB molecules for each type. Here, a contact was defined by an NLB molecule within 2.0 nm of an SR-A bead and persisting for greater than 75% of the simulation time. The protein is in a ball-and-stick representation with the positively charged residues represented as spheres. Blue coloration indicates 0−3 persistent NLB contacts (low), the beads colored white represent 4−7 contacts (medium), and those colored red signify greater than 7 contacts (high) that last for at least 75% of the simulation time. The total number of protein beads falling into each category (low, medium, high) for each NLB is presented in Table 3. The Table 3. Number of SR-A CG Beads with Persistent Contacts at Three Different Levels (Low, Medium, High) for Each Type of NLB along with the Average Number of Persistent Contacts 0CM 1CM 1CP PEG-COOH

0−3 contacts

4−7 contacts

8+ contacts

average #

66 63 125 280

237 140 250 110

87 187 15 0

6.63 8.14 5.57 3.75

Figure 9. Average number of contacts within 0.8 nm (top) for the NLB head (a) or tail (b) and the average number of contacts within 2.0 nm for the head (c) and tail (d) beads. In 1CM and PEG-COOH, the head beads are negatively charged, whereas in 1CP the tail bead is negatively charged. 0CM is neutral at both the head and tail and carries no charge. The legends refer to the one-letter amino acid code as given in Figure 1. Error bars represent the difference between the first and second halves of the simulation. 2506

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

magnitude of the negative charge (charge = 0 for 0CM and −1 for 1CM, 1CP, and PEG-COOH) location of the charge (near the macromolecule head in the case of 1CM or at the tail in the case of 1CP) or the ability of the NLBs to form micelles (PEG-COOH lacks the aliphatic groups of the other NLBs and thus did not form micelles). Of the NLBs that formed micelles (0CM, 1CM, 1CP), the micelles generally formed within ∼300 ns and had a micelle diameter between 8 and 11 nm, consistent with the results of Tian et al.12 One of the new insights from the modeling efforts in this work is that amphiphilic NLBs preferentially interacted with SR-A in the form of micelles rather than single monomers, indicating that NLBs primarily interact with SR-A in an aggregated state. 0CM and 1CP formed micelles that interacted principally with the N-terminal and center portions of SR-A (as indicated by the contact map in Figure 8) while also having micelles in solution away from the protein surface. The dose−response experiments showed that 1CM was most effective at inhibiting uptake of oxidized LDL by PBMC macrophages followed by 1CP, 0CM, and finally PEG-COOH, which produced minimal uptake inhibition. These results are consistent with the experiments of Plourde et al.18 and, subsequently, Iverson et al.34 for oxidized LDL uptake in THP-1 macrophages. The presence of noninteracting 0CM micelles in solution may be indicative of weaker binding between 0CM and SR-A and a possible explanation of why 0CM does not inhibit oxidized LDL uptake as potently as 1CM. 1CM associated with SR-A by way of two large micelles near both the C-terminal and N-terminal ends with no aggregates observed in solution, suggesting that 1CM micelles interact more favorably with SR-A than either 0CM or 1CP micelles. Alternatively, PEG-COOH did not form micelles and interacted with SR-A as monomers with a number of molecules remaining in solution. Ideally, a direct comparison of the magnitude of the experimental EC50 values with the thermodynamic free energy of NLB binding to SR-A would provide the most useful information regarding the energetic contributions to binding, thus allowing for better prediction of optimal NLB structures. The energetics of binding is best described by the change in the Gibbs free energy ΔG = ΔH − TΔS, where ΔH is the change in enthalpy, T is the thermodynamic temperature, and ΔS is the variation in the entropy. From the simulations of NLB interaction with SR-A, we can compute the enthalpic term from the potential energy of the system and the change in volume at constant pressure; however, the enthalpic (ΔS) term is nontrivial to compute requiring several simulations of a fluctuating number of NLB aggregates and varying distances from SR-A. Furthermore, as the system is composed of several chain-like molecules that aggregate to form large assemblies, we expect the entropic term to be a substantial portion of the Gibbs free energy and cannot be neglected. In this work we therefore make correlations between the rank order of experimental EC50 values with the order of structural factors such as the number of contacts formed between the NLBs and SR-A and the distribution of those contacts. Exploring the distance distributions that NLBs reside from SR-A indicated that 1CM and 0CM were located nearest to the surface of the receptor followed by 1CP and PEG-COOH. One mechanism by which NLBs could inhibit uptake of oxidized LDL is to occlude the entire binding region preventing receptor contact with oxidized LDL. A higher concentration of NLBs near the surface of SR-A implies better steric blocking of the receptor to disallow interaction of oxidized LDL with SR-A.

calculated for all four NLBs (neither the head nor tail is charged in 0CM). A contact was defined with a distance cutoff of 0.8 nm between any two beads corresponding to a distance slightly less than two of the smallest CG particles (0.43 nm) in contact with one another.34 Figure 9 displays the average number of contacts at 0.8 nm for the head and tail beads (9a and 9b, respectively) and the average number of contacts at 2.0 nm for the head and tail beads (9c and 9d, respectively). The error bars represent the difference between the first and second halves of the simulation. The number of close contacts ( 0CM > 1CP > PEG-COOH, which aligns with the experimental results of NLB potency. The same NLB ordering scheme was found when the number of persistent contacts, defined to be contacts that are present for greater than 75% of the simulation, was calculated. Finally, the specific contacts between the cationic residues of the proposed oxidized LDL binding pocket and the charged portions of NLBs indicated that the addition of a negative charge on the PEG tails of NLBs did not act to enhance interaction with the residues of the SR-A binding pocket. However, an anionic charge at the NLB head increased interaction with the positively charged residues of the binding pocket, confirming the experimental finding that 1CM is the best inhibitor of oxidized LDL uptake as it is most strongly associated with SR-A.

The results of this work are consistent with experimental findings that 1CM is the most potent (of the four NLBs tested here) at inhibiting oxidized LDL uptake by SR-A. To further investigate this phenomenon, we computed the number of contacts between NLB heads (which are charged in 1CM and PEG-COOH) or the tails (which are charged in 1CP) with the residues implicated in oxidized LDL binding. In general, 1CP showed minimal enhancement of interactions with any of the binding pocket residues over the other NLBs. This could be due to the micelle structure that 1CP forms. With the negative charges placed on the PEG tails of 1CP, they extend radially outward from the micelle interior and are thus unlikely to cluster in one location. Additionally, it was found that 1CP formed micelles that were much more ellipsoidal in shape that either 0CM or 1CM, further altering the interaction of 1CP with SR-A. To enhance association of NLBs with the cationic residues of the SR-A binding pocket, it proved better to place a negative charge near the NLB head. Both 1CM and PEGCOOH had enhanced interaction of their head groups compared to NLBs with neutral head groups. This is in agreement with the work of Plourde et al.18 who found 1CM and PEG-COOH to be the energetically best binders to SR-A. However, experimentally, 1CM is a strong inhibitor of oxidized LDL uptake, whereas PEG-COOH did not deviate from the basal condition. This may be due to the fact that 1CM binds as micelles while PEG-COOH binds as monomers. Computing the distance distribution of NLBs from the surface of SR-A showed a secondary peak for PEG-COOH at approximately 8 nm from the surface of SR-A (Figure 7 inset), whereas no molecules of 1CM were found more than 4 nm from the receptor surface, suggesting that SR-A may not be sufficiently sterically blocked by PEG-COOH molecules. Oxidized LDL molecules are 22−28 nm in diameter,35 and it is possible that an incoming LDL molecule could form sufficient interactions with SR-A to displace a handful of PEG-COOH monomers. Alternatively, 1CM existing as a large micelle around SR-A could be sufficient to sterically disallow oxidized LDL to near the receptor and form any type of interaction.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (732) 445-2972. Fax: (732) 445-2581. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This study was partially supported by NIH grants (R01 HL107913, R21 HL093753) to Prabhas Moghe. Silvina Tomassone and Michael Tomasini were supported by NSF IGERT Nanopharmaceutical Engineering and Science Grant #0504497.

■

REFERENCES

(1) Roger, V. L.; Go, A. S.; Lloyd-Jones, D. M.; Benjamin, E. J.; Berry, J. D.; Borden, W. B.; Bravata, D. M.; Dai, S.; Ford, E. S.; Fox, C. S.; Fullerton, H. J.; Gillespie, C.; Hailpern, S. M.; Heit, J. A.; Howard, V. J.; Kissela, B. M.; Kittner, S. J.; Lackland, D. T.; Lichtman, J. H.; Lisabeth, L. D.; Makuc, D. M.; Marcus, G. M.; Marelli, A.; Matchar, D. B.; Moy, C. S.; Mozaffarian, D.; Mussolino, M. E.; Nichol, G.; Paynter, N. P.; Soliman, E. Z.; Sorlie, P. D.; Sotoodehnia, N.; Turan, T. N.; Virani, S. S.; Wong, N. D.; Woo, D.; Turner, M. B.; Association, A. H.; Subcommittee, S. S. Heart disease and stroke statistics–2012 update: A report from the American Heart Association. Circulation 2012, 125, e2−e220. (2) Li, A. C.; Glass, C. K. The macrophage foam cell as a target for therapeutic intervention. Nat. Med. 2002, 8, 1235−1242. (3) Kunjathoor, V. V.; Febbraio, M.; Podrez, E. A.; Moore, K. J.; Andersson, L.; Koehn, S.; Rhee, J. S.; Silverstein, R.; Hoff, H. F.; Freeman, M. W. Scavenger receptors class A-I/II and CD36 are the principal receptors responsible for the uptake of modified low density lipoprotein leading to lipid loading in macrophages. J. Biol. Chem. 2002, 277, 49982−49988. (4) Brown, M. S.; Goldstein, J. L. Lipoprotein metabolism in the macrophage: Implications for cholesterol deposition in atherosclerosis. Annu. Rev. Biochem. 1983, 52, 223−261. (5) Kodama, T.; Freeman, M.; Rohrer, L.; Zabrecky, J.; Matsudaira, P.; Krieger, M. Type I macrophage scavenger receptor contains αhelical and collagen-like coiled coils. Nature 1990, 343, 531−535. (6) Matsumoto, A.; Naito, M.; Itakura, H.; Ikemoto, S.; Asaoka, H.; Hayakawa, I.; Kanamori, H.; Aburatani, H.; Takaku, F.; Suzuki, H. Human macrophage scavenger receptors: Primary structure, expression, and localization in atherosclerotic lesions. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 9133−9137. (7) Doi, T.; Higashino, K.; Kurihara, Y.; Wada, Y.; Miyazaki, T.; Nakamura, H.; Uesugi, S.; Imanishi, T.; Kawabe, Y.; Itakura, H.

■

CONCLUSIONS The ability of engineered macromolecules to block the uptake of oxidized LDLs by macrophage scavenger receptors was investigated using experimental in vitro techniques in addition to CG MD simulations of the specific interactions between NLBs and SR-A. Previous studies of structure−function relations in NLBs used atomistic models that were only capable of assessing interaction between SR-A and a single truncated NLB. In this work, a CG model was employed to explore the dynamical assembly of NLBs and the interaction of NLB micellar aggregates with SR-A. In the simulations, 0CM, 1CM, and 1CP NLBs formed micelles within 300 ns and consisted of 10 to 11 NLB molecules with a micelle radius of 4.2−5.5 nm, in agreement with experiential observations, while PEG-COOH did not form micelles. As such, the interaction studies between NLBs and SR-A showed that 1CM, 1CP, and 0CM NLBs interacted with SR-A as micellar aggregates, while PEG-COOH associated with the protein predominately as single monomers. The interaction of both 0CM and 1CP was primarily with the N-terminal end and middle of the receptor, while 1CM interacted with both the N-terminal region and Cterminal region (containing the proposed oxidized LDL binding pocket) of SR-A. As an indirect assessment of binding affinity, the distribution of contacts that each type of NLB made 2508

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509

Biomacromolecules

Article

Conformation and hydrodynamics. J. Phys. Chem. B 2009, 113, 13186−13194. (26) Gautieri, A.; Russo, A.; Vesentini, S.; Redaelli, A.; Buehler, M. J. Coarse-grained model of collagen molecules using an extended MARTINI force field. J. Chem. Theory Comput. 2010, 6, 1210−1218. (27) Rainey, J. K.; Goh, M. C. An interactive triple-helical collagen builder. Bioinformatics 2004, 20, 2458−2459. (28) www.md.chem.rug.nl/cgmartini. (29) Lee, H.; Pastor, R. W. Coarse-grained model for PEGylated lipids: Effect of PEGylation on the size and shape of self-assembled structures. J. Phys. Chem. B 2011, 115, 7830−7837. (30) Hess, B.; Kutzner, C.; van der Spoel, D.; Lindahl, E. GROMACS 4: Algorithms for highly efficient, load-balanced, and scalable molecular simulation. J. Chem. Theory Comput. 2008, 4, 435−447. (31) Berendsen, H. J. C.; Postima, J. P. M.; van Gusteren, W. F.; Dinola, A.; Haak, J. R. Molecular dynamics with coupling to an external bath. J. Chem. Phys. 1984, 81, 3684−3690. (32) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular dynamics. J. Mol. Graph. 1996, 14 (33−8), 27−8. (33) Iverson, N. M.; Sparks, S. M.; Demirdirek, B.; Uhrich, K. E.; Moghe, P. V. Controllable inhibition of cellular uptake of oxidized lowdensity lipoprotein: Structure-function relationships for nanoscale amphiphilic polymers. Acta Biomater. 2010, 6, 3081−3091. (34) Hall, B. A.; Sansom, M. P. S. Coarse-grained MD simulations and protein−protein interactions: The cohesin−dockerin system. J. Chem. Theory Comput. 2009, 5, 2465−2471. (35) Gotto, A. M., Jr,; Pownall, H. J.; Havel, R. J. Introduction to the plasma lipoproteins. Methods Enzymol. 1986, 128, 3−41.

Charged collagen structure mediates the recognition of negatively charged macromolecules by macrophage scavenger receptors. J. Biol. Chem. 1993, 268, 2126−2133. (8) Leyva, F. J.; Pershouse, M. A.; Holian, A. Modified low density lipoproteins binding requires a lysine cluster region in the murine macrophage scavenger receptor class A type II. Mol. Biol. Rep. 2010, 37, 2847−2852. (9) Andersson, L.; Freeman, M. W. Functional changes in scavenger receptor binding conformation are induced by charge mutants spanning the entire collagen domain. J. Biol. Chem. 1998, 273, 19592−19601. (10) Yoshiizumi, K.; Nakajima, F.; Dobashi, R.; Nishimura, N.; Ikeda, S. 2,4-Bis(octadecanoylamino)benzenesulfonic acid sodium salt as a novel scavenger receptor inhibitor with low molecular weight. Bioorg. Med. Chem. Lett. 2004, 14, 2791−2795. (11) Broz, P.; Benito, S. M.; Saw, C.; Burger, P.; Heider, H.; Pfisterer, M.; Marsch, S.; Meier, W.; Hunziker, P. Cell targeting by a generic receptor-targeted polymer nanocontainer platform. J. Controlled Release 2005, 102, 475−488. (12) Tian, L.; Yam, L.; Zhou, N.; Tat, H.; Uhrich, K. E. Amphiphilic scorpion-like macromolecules: Design, synthesis, and characterization. Macromolecules 2004, 37, 538−543. (13) Chnari, E.; Nikitczuk, J. S.; Wang, J.; Uhrich, K. E.; Moghe, P. V. Engineered polymeric nanoparticles for receptor-targeted blockage of oxidized low density lipoprotein uptake and atherogenesis in macrophages. Biomacromolecules 2006, 7, 1796−1805. (14) Wang, J.; Plourde, N. M.; Iverson, N.; Moghe, P. V.; Uhrich, K. E. Nanoscale amphiphilic macromolecules as lipoprotein inhibitors: The role of charge and architecture. Int. J. Nanomed. 2007, 2, 697− 705. (15) Harmon, A. M.; Uhrich, K. E. In vitro evaluation of amphiphilic macromolecular nanocarriers for systemic drug delivery. J. Bioact. Compat. Polym. 2009, 24, 185−197. (16) Iverson, N.; Plourde, N. M.; Wang, J.; Nackman, G. B.; Moghe, P. V. Convergence of nanotechnology and cardiovascular medicine: Progress and emerging prospects. BioDrugs 2008, 22, 1−10. (17) Lewis, D. R.; Kamisoglu, K.; York, A. W.; Moghe, P. V. Polymerbased therapeutics: Nanoassemblies and nanoparticles for management of atherosclerosis. Wiley Interdiscip. Rev. Nanomed. Nanobiotechnol. 2011, 3, 400−420. (18) Plourde, N. M.; Kortagere, S.; Welsh, W.; Moghe, P. V. Structure-activity relations of nanolipoblockers with the atherogenic domain of human macrophage scavenger receptor A. Biomacromolecules 2009, 10, 1381−1391. (19) Iverson, N. M.; Plourde, N. M.; Sparks, S. M.; Wang, J.; Patel, E. N.; Shah, P. S.; Lewis, D. R.; Zablocki, K. R.; Nackman, G. B.; Uhrich, K. E.; Moghe, P. V. Dual use of amphiphilic macromolecules as cholesterol efflux triggers and inhibitors of macrophage atheroinflammation. Biomaterials 2011, 32, 8319−8327. (20) York, A. W.; Zablocki, K. R.; Lewis, D. R.; Gu, L.; Uhrich, K. E.; Prud’homme, R. K.; Moghe, P. V. Kinetically assembled nanoparticles of bioactive macromolecules exhibit enhanced stability and celltargeted biological efficacy. Adv. Mater. 2012, 24, 733−739. (21) Marrink, S. J.; de Vries, A. H.; Mark, A. E. Coarse grained model for semiquantitative lipid simulations. J. Phys. Chem. B 2004, 108, 750−760. (22) Marrink, S. J.; Risselada, H. J.; Yefimov, S.; Tieleman, D. P. The MARTINI force field: Coarse grained model for biomolecular simulations. J. Phys. Chem. B 2007, 111, 7812−7824. (23) Monticelli, L.; Kandasamy, S. K.; Periole, X.; Larson, R. G.; Tieleman, D. P.; Marrink, S. J. The MARTINI coarse-grained force field: Extension to proteins. J. Chem. Theory Comput. 2008, 4, 819− 834. (24) Lopez, C. A.; Rzepiela, A. J.; de Vries, A. H.; Dijkhuizen, L.; Hünenberger, P. H; Marrink, S. J. MARTINI coarse-grained force field: Extension to carbohydrates. J. Chem. Theory Comput. 2009, 5, 3195−3210. (25) Lee, H.; de Vries, A. H.; Marrink, S. J.; Pastor, R. W. A coarsegrained model for polyethylene oxide and polyethylene glycol: 2509

dx.doi.org/10.1021/bm301764x | Biomacromolecules 2013, 14, 2499−2509