Advances in Molecular Modeling of Nanoparticle–Nucleic Acid

Oct 24, 2016 - With recent advances such as the use of graphics processing units (GPUs) for simulations, computational modeling has the potential...
0 downloads 0 Views 3MB Size
Download PDF
Review pubs.acs.org/bc

Advances in Molecular Modeling of Nanoparticle−Nucleic Acid Interfaces Jessica A. Nash, Albert L. Kwansa, James S. Peerless, Ho Shin Kim, and Yaroslava G. Yingling* Department of Materials Science and Engineering, North Carolina State University, Raleigh, North Carolina 27606, United States ABSTRACT: Nanoparticles (NPs) play increasingly important roles in nanotechnology and nanomedicine in which nanoparticle surface chemistry allows for control over interactions with other nanoparticles and biomolecules. In particular, for applications in drug and gene delivery, a fundamental understanding of the NP−nucleic acid interface allows for development of more efficient and effective nanoparticle carriers. Computational modeling can provide insights of processes occurring at the inorganic NP−nucleic interface in detail that is difficult to access by experimental methods. With recent advances such as the use of graphics processing units (GPUs) for simulations, computational modeling has the potential to give unprecedented insight into inorganic−biological interfaces via the examination of increasingly large and complex systems. In this Topical Review, we briefly review computational methods relevant to the interactions of inorganic NPs and nucleic acids and highlight recent insights obtained from various computational methods that were applied to studies of inorganic nanoparticle−nanoparticle and nanoparticle−nucleic acid interfaces. In this Topical Review, we first give an overview of computational methods and force fields relevant to the inorganic nanoparticle−nucleic acid interface and then provide notable examples of the application of such methods to gain insights into NP−NP and NP−nucleic acid interactions. The NP−nucleic acid interface covers a broad range of length and time scales and, due to the constrained length of a Topical Review, it is not possible to review all aspects of molecular modeling approaches to describe the interface. Here, we focus on results from classical techniques such as molecular dynamics and dissipative particle dynamics.

1. INTRODUCTION Engineered inorganic nanoparticles (NPs) have shown potential for use as gene delivery vehicles,1 contrast agents for imaging,2 or anticancer agents3 due to their unique optical or electronic properties. However, many obstacles remain to successfully integrate NPs into biological systems. Challenges include nonspecific distribution, and accumulation and localization of NPs at target sites.4 To optimize the use of NPs in these and future applications, a comprehensive understanding is needed of the effect of NP properties, such as the core material(s), core size, NP shape, and surface functionalization, on interactions with biomolecules such as nucleic acids or proteins. For example, the extent of interactions between NPs and nucleic acids may depend on nanoparticle design such as size, shape, surface chemistry, or ligand length and also on nucleic acid sequence and type. Computational techniques can provide rules for rational NP design and be used as powerful tools for fundamental investigations of systems in detail not accessible by experimental methods. Depending on the computational method utilized, researchers can study optical properties, electronic structure, time-dependent NP−biomolecule binding, or equilibrium conformations of functionalized nanoparticles and biomolecules. Robust and well-tested methods and force fields exist with which to study inorganic NPs and nucleic acids, and recent advances in computing algorithms and architecture, such as the use of graphics processing units (GPUs), allow for the simulation of increasingly large and complex systems. For instance, a single GPU-accelerated server can provide a comparable computational throughput to ∼50 CPU-only servers at a markedly reduced cost and power utilization.5 Consequently, simulation methods show tremendous potential for providing details of NP−nucleic acid complexation and NP design rules. © 2016 American Chemical Society

2. METHODS FOR MODELING THE NANOPARTICLE−NUCLEIC ACID INTERFACE A variety of computational methods have been developed to model the electronic structure and time-dependent behavior of molecular systems at different time scales and system resolutions (Figure 1). Each modeling technique has strengths and weaknesses; for example, methods that are able to capture the electronic behavior of molecules are applicable to only equilibrium structures or short time scales, while larger-scale methods lose electronic or atomic resolution. Thus, the phenomena of interest determine the choice of a computational method. In Figure 1, we summarize methods relevant to the NP−biomolecule interface with their various size scales and performance limits based on recently reported benchmarks.6−11 Density Functional Theory. Density functional theory (DFT) is most commonly applied to calculations of electronic structure, which can be used to study chemisorption, Special Issue: Interfacing Inorganic Nanoparticles with Biology Received: September 16, 2016 Revised: October 21, 2016 Published: October 24, 2016 3

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry

fields employed. Some all-atom MD force fields, such as AMBER,23 CHARMM,24−26 and OPLS-AA,27 have been carefully parametrized (and are continuously refined) for biologically relevant molecules such as carbohydrates, lipids, nucleic acids, and proteins. Force fields that describe the inorganic−biomolecule interface are a subject of active research and are much more difficult to develop because parameters will change as a function of nanoparticle inorganic core material, surface and ligand chemistry, and solvent medium. Current force-field parameters have been based upon, for example, experimental densities and surface free energies of metals28 and DFT-calculated and experimental interaction energies between gold and aminoacid-like model compounds.29 Coarse-Grained Molecular Dynamics. Though recent advancements in computing allow for simulation of increasingly large all-atom systems, it is often desirable to simulate largerscale behavior. Coarse-grained MD (CGMD) involves the grouping of atoms into beads, thereby reducing the resolution or detail of the molecular representation and allowing for larger system sizes, longer time scales, and larger time-steps. MARTINI is a popular coarse-grained force field in which four heavy atoms are grouped into one bead, although occasionally other grouping schemes are used.30,31 MARTINI was originally developed to model lipid systems, but it has been extended for proteins32,33 and DNA.34 CGMD permits longer time-steps to be used due to the decreased molecular resolution involved. While all-atom MD typically involves time-steps of 1−4 fs, CGMD force fields such as MARTINI can employ time steps of 10−40 fs, which greatly improves the practical time scale of the simulation. MARTINI force-field parameters have been derived from all-atom structural distributions and experimental data such as densities, diffusivities, and partitioning free energies. However, one should be careful in translating time-dependent properties such as diffusivity because the allatom to coarse-grained mapping and smoother CG potential energy surface may not permit direct comparison to all-atom and experimental quantities.31 Some challenges with CG models can include the absence of potentially important structural details and interactions (e.g., hydrogen bonds), difficulties with modeling nonliquid phases (e.g., gas and solid phases), and the exclusion of long-range electrostatic interactions.31 Choice of Solvent Model. When modeling NP−nucleic acid interfaces, choosing an appropriate solvent model is an important factor. Explicit solvent models, in which solvent particles are included in the simulation, have been most widely used, and nonpolarizable fixed-charge models have been the most prevalent.35 Explicit solvent can afford the prediction of important solvent effects such as solute−solvent interactions. However, due to the increased number of simulated particles, considerable computational resources are required to simulate systems containing explicit solvent.36 Moreover, careful parametrization is required to represent solvents and solute− solvent interactions accurately, and the choice of the force field for solvent becomes imperative.37 Even for water alone, there are many different all-atom models that have been developed, including the most frequently used OPC,38 SPC,39 SPC/E,40 TIP3P,41 TIP4P,42 and TIP5P models,43 all of which are able to represent different properties of water with different degrees of accuracy. In explicit solvent simulations, the choice of force field parameters for salts is also important because nucleic acids are charged molecules. For example, incorrect parametrization

Figure 1. Performance limit (simulation time per day) vs system size scale for various molecular-modeling methods. The highest resolution methods in the lower left corner describe systems with the most detail, but these systems also have the lowest performance limit and system size. As size scale increases, the systems are described with less detail, and the performance limit increases due to the increased timesteps (Δt) that are possible when describing a system with less detail. The bottom of the image shows snapshots depicting the application of various computational methods to the NP−nucleic acid interface. DFT image rendered with data from ref 12, published by the PCCP Owner Societies.

physisorption, or optical properties of inorganic nanoparticles.13 DFT involves an approximate solution to the Schrodinger equation to calculate ground-state energy as an electron density functional (i.e., energy expressed as a function of the probability density of electron positions).14 For example, DFT has been used to gain insights into how amino acids interact with gold and silver15 or platinum nanocrystals.16 DFT has also been successfully used with nuclear magnetic resonance (NMR) spectroscopy to identify chemical reaction mechanisms during the functionalization of silicon nanoparticles17 and amino acid adsorption mechanisms onto gold.18 However, as shown in Figure 1, system sizes for DFT are relatively small (up to a couple thousand atoms). Furthermore, DFT is traditionally unable to describe timedependent behavior and has difficulty describing van der Waals interactions.19 All-Atom Molecular Dynamics. For phenomena that require larger size scales or time-dependent binding behavior, methods such as molecular dynamics (MD) are usually employed (Figure 1). Please note that all subsequent use of “traditional MD” or “all-atom MD” refers specifically to allatom nonreactive MD. Traditional MD neglects the electronic nature of molecules and instead models atoms as spheres connected by spring-like bonds, which move according to Newton’s equations of motion. All-atom MD simulations are now able to describe the behavior of biomolecular systems that are up to hundreds of millions of atoms in size8,20 or for maximum time scales on the order of microseconds.21,22 The success of these studies depends on the quality of the force 4

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry of monovalent ions has led to incorrect results or conclusions in simulation studies based on artifacts.44 Implicit solvent models for all atom simulations, in which the solvent is treated as a continuum dielectric with the average properties of real solvent, can significantly reduce computational expense due to the reduced number of simulated particles and their associated degrees of freedom; furthermore, the absence of viscosity increases the rate at which the solute explores its conformational space and reduces the real-world time needed to capture such dynamic processes.45−47 However, implicit solvent models cannot capture, e.g., viscosity effects48 and specific solute−solvent interaction mechanisms.49 In the case of explicit solvent models for CGMD and DPD, multiple solvent molecules are grouped into beads (e.g., four water molecules in a MARTINI CG bead).30,31,50 CG implicit solvent models have also been developed, e.g., to simulate lipid bilayers.51 Coarse-grained solvent models in particular require very careful interpretation of predicted dynamic behaviors.31,52 Dissipative Particle Dynamics. For studies of large-scale dynamic phenomena, such as self-assembly or phase separation, it is typical to use mesoscale methods such as dissipative particle dynamics (DPD). DPD is a coarse-grained molecular modeling technique in which the net force acting on a soft sphere is composed of conservative, dissipative, and random components.53 Because DPD is based on a momentumconserving thermostat and a soft repulsive potential, relevant hydrodynamic effects can be captured, and very large system sizes comparable to those of experiments are possible. DPD has been used, for example, to predict the formation of gold clusters or nanoparticles in the presence of thiol-terminated ligands and solvent.54 Atomistic details are lost in DPD simulations, given that multiple atoms, entire molecules, or even molecular clusters may be represented as single beads in a similar fashion to CGMD. Moreover, due to the simplifications of Newtonian relationships present in DPD, the calculation of quantitative physical properties is nontrivial. Hence, DPD is more commonly used for comparative studies and investigation into morphological phenomena and phase behavior.

Figure 2. Nanoparticle ligand morphologies and PMF calculations. (A) Schematics of striped, Janus, and patchy NP ligand morphologies. (B) PMF calculations are used to quantify the interaction energies between nanoparticles as a function of distance and can be obtained from computational modeling. This graph shows contributions from electrostatic, van der Waals, and hydrophobic (indicated as “phobic” above) interactions to the total interaction energy between two NPs. Snapshots on the bottom show representative snapshots of particles at various interaction distances. Panel B is adapted with permission from 58 (copyright 2013, American Chemical Society).

3. PROPERTIES OF NANOPARTICLES AND THE NANOPARTICLE LIGAND CORONA The surface chemistry of inorganic nanoparticles can be controlled through attachment of organic ligands to the nanoparticle. Changes in surface chemistry may be achieved through ligand identity (e.g., negative charge and hydrophobicity) and control over the surface morphology of ligands (Figure 2; Janus and striped NPs, etc.). Understanding the properties of functionalized nanoparticles prior to combination with biomolecules is crucial for the successful optimization of the NP−nucleic acid interfaces. Before modeling complex interactions with biomaterials, one should fundamentally know (1) the behavior of the ligand corona grafted to the NP surface and (2) the particle−particle interactions within a medium. In the case of (1), this may be a phase separation between ligands on the surface of the NP, whereas in (2), the ability of ligands to stabilize NP suspensions within a specific matrix may be critical to performance.55 Although DFT methods have been extensively employed toward understanding of the crystal structure of the NP core,56,57 these core properties are not critical to nanoscale behavior at the biointerface and thus are not discussed at length. Modeling Ligand Behavior. Since the pioneering work of Ghorai and Glotzer, which first used CGMD to describe the

phase behavior of alkanethiol ligands on a gold NP surface,59 computational models have been influential in predicting and describing NP corona behavior. Many recent MD studies have focused on descriptions of multispecies ligand phase separations to produce various ligand morphologies, including striped and Janus NPs (Figure 2),60−63 revealing ligand composition and length as crucial design parameters. For instance, in systems with mixed ligand lengths, both atomistic and DPD simulations have shown the formation of striped domains to be entropically controlled; longer ligands increase their free volume (and conformational entropy) by being surrounded by short NP ligands, leading to smaller striped domains where the ratio of NP ligand lengths is large.61 Overall, computational investigations have been able to reveal and explain experimentally relevant ligand behavior in diblock copolymerand mixed-polymer-coated NPs utilizing CGMD64 and efficient field theory65 approaches. Modeling Interparticle Interactions. In the case of simulations in which multiple nanoparticles interact with biomolecules, knowledge of interparticle interactions and solubility are important. The distance-dependent interaction energy profile between two NPs can be used to predict selfassembly and phase behavior. Moreover, a detailed explanation 5

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry

Figure 3. Results from computational studies of DNA−NP interactions. (A) Illustration of all-atom vs coarse-grain models for nanoparticles and DNA. (B) Results from CGMD study of NPs binding to DNA. Depending on NP size, modes of interaction are adsorption, wrapping, or collection. Modified and reproduced with permission from ref 80 (copyright 2007, American Chemical Society). (C) Snapshots of an all-atom simulation of 1.5 nm gold nanoparticles binding to double-stranded DNA and RNA in 0.1 M NaCl. The NP charge is indicated below the snapshot. NPs of high charge are able to induce wrapping of DNA but do not induce wrapping of RNA. Modified and reproduced with permission from ref 84 (copyright 2015, American Chemical Society). (D) All-atom molecular dynamics provides insight into the NP−nucleic acid interface with a high level of detail. This plot shows a comparison of the conformation of DNA when wrapped around a charged NP to that of DNA in a crystal structure. Modified and reproduced with permission from ref 84.

4. THE NUCLEIC ACID−NP INTERFACE NP−nucleic acid conjugates are created either through surface functionalization of NPs with nucleic acids or through noncovalent binding of nucleic acids with ligands of functionalized nanoparticles. In the first class of material, hybridization with complementary nucleic acid molecules allows for assembly of complex periodic structures, while in the second class, NPs can be designed to wrap and cause compaction of nucleic acids. Here, we focus on the second class of material in which nucleic acids interact with NPs through noncovalent interactions. Nucleic acids, such as DNA and RNA, are biological molecules consisting of a negatively charged backbone and bases capable of base pairing with one another. In applications such as drug or gene delivery, cationically functionalized inorganic NPs bind to nucleic acids through electrostatic interactions and are capable of changing the nucleic acid structure. Nanoparticles have been shown experimentally to induce conformational changes of nucleic acids, such as bending and wrapping,78 or denaturation.79 Methods that are most appropriate for studying the interface at this level are CGMD and all-atom MD due to their ability to probe timedependent binding behavior and the final structure of the nucleic acid complex. Studying large-scale nucleic acid−nanoparticle binding using computational methods presents a challenge in system size. Biological DNA may often be thousands to millions of base pairs in length, but until recently, practical all-atom simulation was limited to systems which were less than 1 000 000 atoms in size, including solvent. Thus, to observe NP−nucleic acid binding and conformational changes, coarse-grained methods

of the components of this interaction energy profile allows researchers to predictably manipulate and modify NP−NP interactions via their environment. Given the complex nature of NP functionalized groups, it is often found that generalized approaches such as the Derjaguin−Landau−Verwey−Overbeek (DLVO) theory fail to fully describe NP−NP interaction energies.58,66,67 Hence, many studies have calculated interaction energies as a function of distance in silico by investigating systems containing two NPs. A common method for determining interaction energies between two bodies as a function of distance is through calculation of the potential of mean force (PMF). There are many different methods to compute PMF through simulation, and a robust generalized review of these methods68 has been performed. Recent studies have employed CGMD techniques69−71 as well as steered MD72,73 to calculate PMF between two NPs on a variety of ligand compositions and matrix materials. Figure 2B shows a PMF curve generated from Monte Carlo simulation of two charged nanoparticles at varying distances. These calculations showed that hydrophobic interactions between ligands are the driving force causing aggregation (Figure 2B).58 Aside from PMF calculations, additional thermodynamic relations have been derived in conjunction with simulation methods to elucidate the nature of NP−NP interactions for various functionalization parameters.58,74 Beyond two-NP systems, the prediction of aggregation and self-assembly behavior of many NPs in a medium is often desired but require the use of CGMD70,75,76 or DPD77 approaches, as the high number of atoms necessary makes all-atom approaches prohibitively slow. 6

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry

stranded DNA and RNA, with system sizes and simulation times that were previously unfeasible with traditional CPU computing. The use of all-atom systems allowed for examination of the bending mechanisms of DNA and RNA, where structural changes could be evaluated by base pair and compared to experimentally determined crystal structures of DNA wrapping around the biological histone proteins (Figure 3D). These simulations showed that some NPs are able to induce smooth DNA bending similar to histone octamer. In addition to NP charge, simulations revealed that shape of the NP ligand corona influenced the ability of NPs to bend DNA, with more spherical nanoparticles of higher charge showing a greater degree of bending for DNA. Our simulations of cationic nanoparticles with DNA and RNA were the first to examine differences in interactions of inorganic nanoparticles with double-stranded RNA and DNA. Although double stranded DNA and RNA are both linear and highly charged polyelectrolytes, they show distinct differences in behavior. For example, in response to multivalent cations, DNA readily condenses, while RNA does not.86 A dual computational and experimental study proposed this was due to differences in binding of ions to the nucleic acid strands; for DNA, ions bind to the outside of the helix, allowing for condensation to occur, while for RNA, the ions bind internally, preventing condensation.87 Similar to the behavior of RNA with multivalent ions, even high-charge NPs were not able to bend RNA but readily caused DNA bending. Our simulations showed this is because long NP ligands bind to the inside of RNA’s major groove and inhibit RNA from bending.84 Recent atomistic MD simulations of single-stranded DNA and RNA binding to cationic ligand functionalized NPs have shown that the influence of NPs on the structure of single-stranded DNA and RNA depends on nucleic acid sequence, with polypyrimidines showing greater response to NP binding than polypurines.85

have been traditionally employed to study the NP−nucleic acid interface. In recent years, however, computer hardware and software advances have allowed for the simulation of progressively large NP−nucleic acid systems. Coarse-Grained Simulations of DNA and Cationic Nanoparticles. For coarse-grained simulations of the nanoparticle−nucleic acid interface, nucleic acids are typically represented as a generic polyelectrolyte chain and nanoparticles are represented as spheres with uniform charge, allowing for simulation of large systems at the cost of atomic detail (Figure 3A). An early coarse-grained MD study by Zinchenko et al., which examined the binding of cationic nanoparticles with long strands of DNA with both CGMD and experimental methods, identified three modes of interactions of NPs with DNA depending on the relative size of the NP, which are (1) adsorption of DNA to NP surfaces, (2) wrapping of DNA around NPs, and (3) collection of NPs on DNA (Figure 3B).80 CGMD has also been used to examine the effect of NP charge on nucleic acid binding. CGMD simulations have shown that the extent of DNA wrapping around cationic nanospheres depends on overall NP charge. NP charge was shown to influence the quality of DNA wrapping around the nanoparticles, where very high NP charges caused random orientation of the DNA chain on the surface of the NP, in contrast to lower charges, which resulted in ordered wrapping.81 Nonuniform charge distribution on nanoparticles may also occur in the case of Janus nanoparticles. CGMD has been used to examine the effect of charge distribution on the nanoparticle through simulation of polyelectrolyte binding to Janus particles in response to changes in nanoparticle size and surface charge density.82 The ligand chemistry at the nanoparticle surface may fundamentally change the interaction of the polyelectrolyte chain with the NP because ligand identity can make the nanoparticle considerably “softer”. In CGMD, the effect of nanoparticle functionalization has been examined through simulation of a spherical polyelectrolyte brush with a charged linear polyelectrolyte, which can be considered analogs to a functionalized nanoparticle and to DNA or RNA, respectively. Simulations revealed differences in polyelectrolyte ordering on the surface of the nanoparticle depending on grafting density, with high grafting densities showing unordered arrangements of the polyelectrolyte.83 Atomistic Modeling of the NP−Nucleic Acid Interface. The simulations described above mostly considered NPs as spherical beads with uniform charge or the nucleic acids as generic polyelectrolytes. However, details such as nanoparticle ligand identity or specific conformational changes of DNA and RNA may influence the behavior of the NP−nucleic acid complex. To observe the role of ligand corona, it is desirable to model chemically explicit ligands using all-atom methods, while the use of all-atom models of DNA and RNA allows for the evaluation of DNA- and RNA-bending mechanisms. Our group has performed multiple studies using atomistic molecular dynamics simulations of the binding of functionalized gold NPs to nucleic acid, revealing the key role ligands play in nucleic acid binding.79,84,85 For instance, Railsback et al. found that high concentrations of functionalized NPs with low charge could induce DNA strand separation in a study that combined results from atomistic MD, UV−vis spectroscopy, and gel electrophoresis.79 Recently, Nash et al. have reported all-atom GPU-enabled MD simulation of variable charge gold NPs binding to double-

5. CONCLUSIONS AND FUTURE DIRECTIONS Here, we have highlighted the various computational techniques and their application to the investigation of NP− nucleic acid interfaces. The interactions of ligand-functionalized NPs and biomolecules encompass a range of length and time scales, with behavior ranging from adsorption to aggregation of multiple NPs, biomolecules, or both. Thus, multiple computational methods must be used to capture all of the details of the inorganic NP−nucleic acid interface. Each technique outlined here is appropriate for studying different NP−biomolecule phenomena and has associated strengths and limitations. This Topical Review shows that computational methods have already provided important insight into the adsorption of ligands onto NP surfaces, ligand behavior, NP aggregation, and NP−nucleic acid binding. For NP−nucleic acid interactions, simulations have determined some NP design rules that can be used to optimize NP binding to biomolecules. In future research, we expect that simulation methods will play an even more significant role in the study of the nucleic acid−nanoparticle interface as computing methods and architectures continue to improve. Recent advances in implementation of GPU-based computing architecture has expanded the length and time scales accessible for simulations. Of the methods outlined here, all-atom MD has benefitted the most from the use of GPUs, and we expect that in the future, all-atom simulations will play an increasingly important role in the description of the NP−biomolecule interface. The success 7

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry

(12) Zhang, X., Sun, C. Q., and Hirao, H. (2013) Guanine binding to gold nanoparticles through nonbonding interactions. Phys. Chem. Chem. Phys. 15, 19284−19292. (13) Zhong, J., Tang, X. Q., Tang, J., Su, J. C., and Pei, Y. (2015) Density functional theory studies on structure, ligand exchange, and optical properties of ligand-protected gold nanoclusters: Thiolate versus selenolate. J. Phys. Chem. C 119, 9205−9214. (14) Sholl, D., and Steckel, J. A. (2009) Density functional theory: A practical introduction; John Wiley & Sons: Hoboken, NJ. (15) Pakiari, A. H., and Jamshidi, Z. (2007) Interaction of amino acids with gold and silver clusters. J. Phys. Chem. A 111, 4391−4396. (16) Ramakrishnan, S. K., Martin, M., Cloitre, T., Firlej, L., and Gergely, C. (2015) Design rules for metal binding biomolecules: Understanding of amino acid adsorption on platinum crystallographic facets from density functional calculations. Phys. Chem. Chem. Phys. 17, 4193−4198. (17) Lee, D., Kaushik, M., Coustel, R., Chenavier, Y., Chanal, M., Bardet, M., Dubois, L., Okuno, H., Rochat, N., Duclairoir, F., et al. (2015) Solid-state NMR and DFT combined for the surface study of functionalized silicon nanoparticles. Chem. - Eur. J. 21, 16047−16058. (18) Karki, I., Wang, H., Geise, N. R., Wilson, B. W., Lewis, J. P., and Gullion, T. (2015) Tripeptides on gold nanoparticles: Structural differences between two reverse sequences as determined by solidstate NMR and DFT calculations. J. Phys. Chem. B 119, 11998−12006. (19) Klimes, J., and Michaelides, A. (2012) Perspective: Advances and challenges in treating van der waals dispersion forces in density functional theory. J. Chem. Phys. 137, 120901. (20) Perilla, J. R., Goh, B. C., Cassidy, C. K., Liu, B., Bernardi, R. C., Rudack, T., Yu, H., Wu, Z., and Schulten, K. (2015) Molecular dynamics simulations of large macromolecular complexes. Curr. Opin. Struct. Biol. 31, 64−74. (21) Pasi, M., Maddocks, J. H., and Lavery, R. (2015) Analyzing ion distributions around DNA: Sequence-dependence of potassium ion distributions from microsecond molecular dynamics. Nucleic Acids Res. 43, 2412−2423. (22) Salomon-Ferrer, R., Gotz, A. W., Poole, D., Le Grand, S., and Walker, R. C. (2013) Routine microsecond molecular dynamics simulations with AMBER on GPUs. 2. Explicit solvent particle mesh Ewald. J. Chem. Theory Comput. 9, 3878−3888. (23) Cornell, W. D., Cieplak, P., Bayly, C. I., Gould, I. R., Merz, K. M., Ferguson, D. M., Spellmeyer, D. C., Fox, T., Caldwell, J. W., and Kollman, P. A. (1995) A second generation force field for the simulation of proteins, nucleic acids, and organic molecules. J. Am. Chem. Soc. 117, 5179−5197. (24) MacKerell, A. D., Bashford, D., Bellott, M., Dunbrack, R. L., Evanseck, J. D., Field, M. J., Fischer, S., Gao, J., Guo, H., Ha, S., et al. (1998) All-atom empirical potential for molecular modeling and dynamics studies of proteins. J. Phys. Chem. B 102, 3586−3616. (25) Foloppe, N., and MacKerell, A. D., Jr. (2000) All-atom empirical force field for nucleic acids: I. Parameter optimization based on small molecule and condensed phase macromolecular target data. J. Comput. Chem. 21, 86−104. (26) Schlenkrich, M., Brickmann, J., MacKerell, A. D., and Karplus, M. (1996) An empirical potential energy function for phospholipids: Criteria for parameter optimization and applications. In Biological Membranes: A Molecular Perspective from Computation and Experiment (Merz, K. M., and Roux, B., Eds.); Birkhäuser: Boston, MA; pp 31−81. (27) Jorgensen, W. L., Maxwell, D. S., and TiradoRives, J. (1996) Development and testing of the OPLS all-atom force field on conformational energetics and properties of organic liquids. J. Am. Chem. Soc. 118, 11225−11236. (28) Heinz, H., Vaia, R. A., Farmer, B. L., and Naik, R. R. (2008) Accurate simulation of surfaces and interfaces of face-centered cubic metals using 12−6 and 9−6 Lennard-Jones potentials. J. Phys. Chem. C 112, 17281−17290. (29) Iori, F., Di Felice, R., Molinari, E., and Corni, S. (2009) GolP: An atomistic force-field to describe the interaction of proteins with Au(111) surfaces in water. J. Comput. Chem. 30, 1465−1476.

of such studies will still depend on careful parametrization of force fields for all-atom MD and CGMD. For instance, the description of metallic nanoparticles may benefit from treatment with polarizable force fields or reactive force-field potentials. Multiscale and multiresolution models may also play important roles in describing the inorganic−biomolecule interface.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS

■

REFERENCES

The authors wish to thank the National Science Foundation grants CMMI-1150682 and CBET-1403871 for the support. J.A.N. was supported by the National Science Foundation Graduate Research Fellowship under grant no. DGE-0946818.

(1) Xu, Z. P., Zeng, Q. H., Lu, G. Q., and Yu, A. B. (2006) Inorganic nanoparticles as carriers for efficient cellular delivery. Chem. Eng. Sci. 61, 1027−1040. (2) Nam, J., Won, N., Bang, J., Jin, H., Park, J., Jung, S., Park, Y., Kim, S., and Jung, S. (2013) Surface engineering of inorganic nanoparticles for imaging and therapy. Adv. Drug Delivery Rev. 65, 622−648. (3) Sun, T. M., Zhang, Y. S., Pang, B., Hyun, D. C., Yang, M. X., and Xia, Y. N. (2014) Engineered nanoparticles for drug delivery in cancer therapy. Angew. Chem., Int. Ed. 53, 12320−12364. (4) Blanco, E., Shen, H., and Ferrari, M. (2015) Principles of nanoparticle design for overcoming biological barriers to drug delivery. Nat. Biotechnol. 33, 941−951. (5) Gao, Y. X., Iqbal, S., Zhang, P., and Qiu, M. K. (2015) Performance and power analysis of high-density multi-GPGPU architectures: A preliminary case study. In Proceedings of the 2015 IEEE 17th International Conference on High Performance Computing and Communications, 2015 IEEE 7th International Symposium on Cyberspace Safety and Security, and 2015 IEEE 12th International Conference on Embedded Software and Systems, New York, August 24−26, 2015; IEEE: New York; pp 66−71. (6) de Jong, D. H., Baoukina, S., Ingolfsson, H. I., and Marrink, S. J. (2016) Martini straight: Boosting performance using a shorter cutoff and GPUs. Comput. Phys. Commun. 199, 1−7. (7) De Palma, P., Valentini, P., and Napolitano, M. (2006) Dissipative particle dynamics simulation of a colloidal micropump. Phys. Fluids 18, 027103. (8) Phillips, J. C., Sun, Y., Jain, N., Bohm, E. J., and Kalé, L. V. (2014) Mapping to irregular torus topologies and other techniques for petascale biomolecular simulation. SC Conf Proc. 2014, 81−91. (9) Hakala, S., Havu, V., Enkovaara, J., and Nieminen, R. (2013) Parallel electronic structure calculations using multiple graphics processing units (GPUs), In Applied Parallel and Scientific Computing; 11th International Conference, PARA 2012, Helsinki, Finland, June 10−13, 2012, revised selected papers (Manninen, P., and Ö ster, P., Eds.); Springer: Berlin, Germany; pp 63−76. (10) Senftle, T. P., Hong, S., Islam, M. M., Kylasa, S. B., Zheng, Y., Shin, Y. K., Junkermeier, C., Engel-Herbert, R., Janik, M. J., Aktulga, H. M., et al. (2016) The ReaxFF reactive force-field: Development, applications and future directions. Npj Computational Materials 2, 15011. (11) Le Grand, S., Gotz, A. W., and Walker, R. C. (2013) SPFP: Speed without compromise-a mixed precision model for GPU accelerated molecular dynamics simulations. Comput. Phys. Commun. 184, 374−380. 8

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry

simulations with implicit solvent. J. Chem. Theory Comput. 11, 260− 275. (52) Rzepiela, A. J., Louhivuori, M., Peter, C., and Marrink, S. J. (2011) Hybrid simulations: Combining atomistic and coarse-grained force fields using virtual sites. Phys. Chem. Chem. Phys. 13, 10437− 10448. (53) Groot, R. D., and Warren, P. B. (1997) Dissipative particle dynamics: Bridging the gap between atomistic and mesoscopic simulation. J. Chem. Phys. 107, 4423−4435. (54) Juan, S. C. C., Hua, C. Y., Chen, C. L., Sun, X. Q., and Xi, H. T. (2005) Dissipative particle dynamics simulation of a gold nanoparticle system. Mol. Simul. 31, 277−282. (55) Boles, M. A., Ling, D., Hyeon, T., and Talapin, D. V. (2016) The surface science of nanocrystals. Nat. Mater. 15, 141−153. (56) Lopez-Acevedo, O., Akola, J., Whetten, R. L., Gronbeck, H., and Hakkinen, H. (2009) Structure and bonding in the ubiquitous icosahedral metallic gold cluster Au-144(SR)(60). J. Phys. Chem. C 113, 5035−5038. (57) Li, G., Abroshan, H., Chen, Y. X., Jin, R. C., and Kim, H. J. (2015) Experimental and mechanistic understanding of aldehyde hydrogenation using Au-25 nanoclusters with Lewis acids: Unique sites for catalytic reactions. J. Am. Chem. Soc. 137, 14295−14304. (58) Van Lehn, R. C., and Alexander-Katz, A. (2013) Ligandmediated short-range attraction drives aggregation of charged mono layer-protected gold nanoparticles. Langmuir 29, 8788−8798. (59) Ghorai, P. K., and Glotzer, S. C. (2007) Molecular dynamics simulation study of self-assembled monolayers of alkanethiol surfactants on spherical gold nanoparticles. J. Phys. Chem. C 111, 15857−15862. (60) Singh, C., Ghorai, P. K., Horsch, M. A., Jackson, A. M., Larson, R. G., Stellacci, F., and Glotzer, S. C. (2007) Entropy-mediated patterning of surfactant-coated nanoparticles and surfaces. Phys. Rev. Lett. 99.10.1103/PhysRevLett.99.226106 (61) Ghorai, P. K., and Glotzer, S. C. (2010) Atomistic simulation study of striped phase separation in mixed-ligand self-assembled monolayer coated nanoparticles. J. Phys. Chem. C 114, 19182−19187. (62) Van Lehn, R. C., and Alexander-Katz, A. (2013) Structure of mixed-monolayer-protected nanoparticles in aqueous salt solution from atomistic molecular dynamics simulations. J. Phys. Chem. C 117, 20104−20115. (63) Velachi, V., Bhandary, D., Singh, J. K., and Cordeiro, M. (2016) Striped gold nanoparticles: New insights from molecular dynamics simulations. J. Chem. Phys. 144, 244710. (64) Dong, J., Li, J., and Zhou, J. (2014) Interfacial and phase transfer behaviors of polymer brush grafted amphiphilic nanoparticles: A computer simulation study. Langmuir 30, 5599−5608. (65) Koski, J., Chao, H., and Riggleman, R. A. (2015) Predicting the structure and interfacial activity of diblock brush, mixed brush, and Janus-grafted nanoparticles. Chem. Commun. 51, 5440−5443. (66) Tay, K., and Bresme, F. (2005) Computer simulations of two dimensional gold nanoparticle arrays: The influence of core geometry. Mol. Simul. 31, 515−526. (67) Guerrero-Garcia, G. I., Gonzalez-Mozuelos, P., and de la Cruz, M. O. (2011) Potential of mean force between identical charged nanoparticles immersed in a size-asymmetric monovalent electrolyte. J. Chem. Phys. 135, 135−164705. (68) Trzesniak, D., Kunz, A. P. E., and van Gunsteren, W. F. (2007) A comparison of methods to compute the potential of mean force. ChemPhysChem 8, 162−169. (69) Jabes, B. S., Yadav, H. O. S., Kumar, S. K., and Chakravarty, C. (2014) Fluctuation-driven anisotropy in effective pair interactions between nanoparticles: Thiolated gold nanoparticles in ethane. J. Chem. Phys. 141, 154904. (70) de Moura, A. F., Bernardino, K., Dalmaschio, C. J., Leite, E. R., and Kotov, N. A. (2015) Thermodynamic insights into the selfassembly of capped nanoparticles using molecular dynamic simulations. Phys. Chem. Chem. Phys. 17, 3820−3831. (71) Estridge, C. E., and Jayaraman, A. (2015) Effect of homopolymer matrix on diblock copolymer grafted nanoparticle

(30) Marrink, S. J., de Vries, A. H., and Mark, A. E. (2004) Coarse grained model for semiquantitative lipid simulations. J. Phys. Chem. B 108, 750−760. (31) Marrink, S. J., Risselada, H. J., Yefimov, S., Tieleman, D. P., and de Vries, A. H. (2007) The MARTINI force field: Coarse grained model for biomolecular simulations. J. Phys. Chem. B 111, 7812−7824. (32) de Jong, D. H., Singh, G., Bennett, W. F. D., Arnarez, C., Wassenaar, T. A., Schafer, L. V., Periole, X., Tieleman, D. P., and Marrink, S. J. (2013) Improved parameters for the Martini coarsegrained protein force field. J. Chem. Theory Comput. 9, 687−697. (33) Monticelli, L., Kandasamy, S. K., Periole, X., Larson, R. G., Tieleman, D. P., and Marrink, S. J. (2008) The MARTINI coarsegrained force field: Extension to proteins. J. Chem. Theory Comput. 4, 819−834. (34) Uusitalo, J. J., Ingolfsson, H. I., Akhshi, P., Tieleman, D. P., and Marrink, S. J. (2015) Martini coarse-grained force field: Extension to DNA. J. Chem. Theory Comput. 11, 3932−3945. (35) Cramer, C. J. (2013) Essentials of Computational Chemistry: Theories and Models; John Wiley & Sons: Hoboken, NJ. (36) Cheatham, T. E., and Young, M. A. (2000) Molecular dynamics simulation of nucleic acids: Successes, limitations, and promise. Biopolymers 56, 232−256. (37) Levy, R. M., and Gallicchio, E. (1998) Computer simulations with explicit solvent: Recent progress in the thermodynamic decomposition of free energies and in modeling electrostatic effects. Annu. Rev. Phys. Chem. 49, 531−567. (38) Izadi, S., Anandakrishnan, R., and Onufriev, A. V. (2014) Building water models: A different approach. J. Phys. Chem. Lett. 5, 3863−3871. (39) Berendsen, H. J. C., Postma, J. P. M., van Gunsteren, W. F., and Hermans, J. (1981) Interaction models for water in relation to protein hydration. In Intermolecular Forces (Pullman, B., Ed.); D. Reidel Publishing Company: Dordrecht, Holland; pp 331−342. (40) Berendsen, H. J. C., Grigera, J. R., and Straatsma, T. P. (1987) The missing term in effective pair potentials. J. Phys. Chem. 91, 6269− 6271. (41) Jorgensen, W. L., Chandrasekhar, J., Madura, J. D., Impey, R. W., and Klein, M. L. (1983) Comparison of simple potential functions for simulating liquid water. J. Chem. Phys. 79, 926−935. (42) Jorgensen, W. L., and Madura, J. D. (1985) Temperature and size dependence for Monte Carlo simulations of TIP4P water. Mol. Phys. 56, 1381−1392. (43) Mahoney, M. W., and Jorgensen, W. L. (2000) A five-site model for liquid water and the reproduction of the density anomaly by rigid, nonpolarizable potential functions. J. Chem. Phys. 112, 8910. (44) Joung, I. S., and Cheatham, T. E. (2008) Determination of alkali and halide monovalent ion parameters for use in explicitly solvated biomolecular simulations. J. Phys. Chem. B 112, 9020−9041. (45) Roux, B., and Simonson, T. (1999) Implicit solvent models. Biophys. Chem. 78, 1−20. (46) Bashford, D., and Case, D. A. (2000) Generalized Born models of macromolecular solvation effects. Annu. Rev. Phys. Chem. 51, 129− 152. (47) Anandakrishnan, R., Drozdetski, A., Walker, R. C., and Onufriev, A. V. (2015) Speed of conformational change: Comparing explicit and implicit solvent molecular dynamics simulations. Biophys. J. 108, 1153−1164. (48) Feig, M., and Brooks, C. L. (2004) Recent advances in the development and application of implicit solvent models in biomolecule simulations. Curr. Opin. Struct. Biol. 14, 217−224. (49) Kleinjung, J., and Fraternali, F. (2014) Design and application of implicit solvent models in biomolecular simulations. Curr. Opin. Struct. Biol. 25, 126−134. (50) Hadley, K. R., and McCabe, C. (2012) Coarse-grained molecular models of water: A review. Mol. Simul. 38, 671−681. (51) Arnarez, C., Uusitalo, J. J., Masman, M. F., Ingolfsson, H. I., de Jong, D. H., Melo, M. N., Periole, X., de Vries, A. H., and Marrink, S. J. (2015) Dry Martini, a coarse-grained force field for lipid membrane 9

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10

Review

Bioconjugate Chemistry conformation and potential of mean force: A molecular simulation study. J. Polym. Sci., Part B: Polym. Phys. 53, 76−88. (72) Alsharif, S. A., Chen, L. Y., Tlahuice-Flores, A., Whetten, R. L., and Yacaman, M. J. (2014) Interaction between functionalized gold nanoparticles in physiological saline. Phys. Chem. Chem. Phys. 16, 3909−3913. (73) Villarreal, O. D., Chen, L. Y., Whetten, R. L., and Yacaman, M. J. (2015) Ligand-modulated interactions between charged monolayerprotected Au-144(SR)(60) gold nanoparticles in physiological saline. Phys. Chem. Chem. Phys. 17, 3680−3688. (74) Bauer, G., Lange, A., Gribova, N., Holm, C., and Gross, J. (2015) Effective potentials between gold nano crystals - functional dependence on temperature. Mol. Simul. 41, 1153−1158. (75) Lin, J. Q., Zhang, H. W., Chen, Z., Zheng, Y. G., Zhang, Z. Q., and Ye, H. F. (2011) Simulation study of aggregations of monolayerprotected gold nanoparticles in solvents. J. Phys. Chem. C 115, 18991− 18998. (76) Estridge, C. E., and Jayaraman, A. (2014) Assembly of diblock copolymer functionalized spherical nanoparticles as a function of copolymer composition. J. Chem. Phys. 140, 144905. (77) Go, E. M., Lee, T. K., Min, S. H., Jeon, W. C., Kim, B. S., Yeom, M. S., and Kwak, S. K. (2016) Theoretical study on enhancement of sensing capability of plasmonic dimer Au nanoparticles with amphiphilic polymer brushes. J. Phys. Chem. C 120, 11068−11077. (78) Goodman, C. M., Chari, N. S., Han, G., Hong, R., Ghosh, P., and Rotello, V. M. (2006) DNA-binding by functionalized gold nanoparticles: Mechanism and structural requirements. Chem. Biol. Drug Des. 67, 297−304. (79) Railsback, J. G., Singh, A., Pearce, R. C., McKnight, T. E., Collazo, R., Sitar, Z., Yingling, Y. G., and Melechko, A. V. (2012) Weakly charged cationic nanoparticles induce DNA bending and strand separation. Adv. Mater. 24, 4261. (80) Zinchenko, A. A., Sakaue, T., Araki, S., Yoshikawa, K., and Baigl, D. (2007) Single-chain compaction of long duplex DNA by cationic nanoparticles: Modes of interaction and comparison with chromatin. J. Phys. Chem. B 111, 3019−3031. (81) Cao, Q. Q., Zuo, C. C., Ma, Y. H., Li, L. J., and Zhang, Z. (2011) Interaction of double-stranded DNA with a nanosphere: A coarsegrained molecular dynamics simulation study. Soft Matter 7, 506−514. (82) de Carvalho, S. J., Metzler, R., and Cherstvy, A. G. (2014) Critical adsorption of polyelectrolytes onto charged janus nanospheres. Phys. Chem. Chem. Phys. 16, 15539−15550. (83) Cao, Q. Q., and Bachmann, M. (2013) Polyelectrolyte adsorption on an oppositely charged spherical polyelectrolyte brush. Soft Matter 9, 5087−5098. (84) Nash, J. A., Singh, A., Li, N. K., and Yingling, Y. G. (2015) Characterization of nucleic acid compaction with histone-mimic nanoparticles through all-atom molecular dynamics. ACS Nano 9, 12374−12382. (85) Nash, J. A., Tucker, T. L., Therriault, W., and Yingling, Y. (2016) Binding of single stranded nucleic acids to cationic ligand functionalized gold nanoparticles. Biointerphases, accepted. (86) Li, L., Pabit, S. A., Meisburger, S. P., and Pollack, L. (2011) Double-stranded RNA resists condensation. Phys. Rev. Lett. 106.10.1103/PhysRevLett.106.108101 (87) Tolokh, I. S., Pabit, S. A., Katz, A. M., Chen, Y. J., Drozdetski, A., Baker, N., Pollack, L., and Onufriev, A. V. (2014) Why doublestranded RNA resists condensation. Nucleic Acids Res. 42, 10823− 10831.

10

DOI: 10.1021/acs.bioconjchem.6b00534 Bioconjugate Chem. 2017, 28, 3−10