Challenges in Predicting Molecular Structures - ACS Publications

Jun 30, 2017 - substantial progress in developing interatomic potentials for these interfaces and adapting advanced conformational sampling approaches...
71 downloads 10 Views 4MB Size
This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article pubs.acs.org/accounts

Pathways to Structure−Property Relationships of Peptide−Materials Interfaces: Challenges in Predicting Molecular Structures Tiffany R. Walsh* Institute for Frontier Materials, Deakin University, Geelong, VIC 3216, Australia CONSPECTUS: An in-depth appreciation of how to manipulate the molecularlevel recognition between peptides and aqueous materials interfaces, including nanoparticles, will advance technologies based on self-organized metamaterials for photonics and plasmonics, biosensing, catalysis, energy generation and harvesting, and nanomedicine. Exploitation of the materials-selective binding of biomolecules is pivotal to success in these areas and may be particularly key to producing new hierarchically structured biobased materials. These applications could be accomplished by realizing preferential adsorption of a given biomolecule onto one materials composition over another, one surface facet over another, or one crystalline polymorph over another. Deeper knowledge of the aqueous abiotic− biotic interface, to establish clear structure−property relationships in these systems, is needed to meet this goal. In particular, a thorough structural characterization of the surface-adsorbed peptides is essential for establishing these relationships but can often be challenging to accomplish via experimental approaches alone. In addition to myriad existing challenges associated with determining the detailed molecular structure of any molecule adsorbed at an aqueous interface, experimental characterization of materials-binding peptides brings new, complex challenges because many materials-binding peptides are thought to be intrinsically disordered. This means that these peptides are not amenable to experimental techniques that rely on the presence of well-defined secondary structure in the peptide when in the adsorbed state. To address this challenge, and in partnership with experiment, molecular simulations at the atomistic level can bring complementary and critical insights into the origins of this abiotic/biotic recognition and suggest routes for manipulating this phenomenon to realize new types of hybrid materials. For the reasons outlined above, molecular simulation approaches also face challenges in their successful application to model the biotic−abiotic interface, related to several factors. For instance, simulations require a plausible description of the chemistry and the physics of the interface, which comprises two very different states of matter, in the presence of liquid water. Also, it is essential that the conformational ensemble be comprehensively characterized under these conditions; this is especially challenging because intrinsically disordered peptides do not typically admit one single structure or set of structures. Moreover, a plausible structural model of the substrate is required, which may require a high level of detail, even for single-element materials such as Au surfaces or graphene. Developing and applying strategies to make credible predictions of the conformational ensemble of adsorbed peptides and using these to construct structure−property relationships of these interfaces have been the goals of our efforts. We have made substantial progress in developing interatomic potentials for these interfaces and adapting advanced conformational sampling approaches for these purposes. This Account summarizes our progress in the development and deployment of interfacial force fields and molecular simulation techniques for the purpose of elucidating these insights at biomolecule−materials interfaces, using examples from our laboratories ranging from noble-metal interfaces to graphitic substrates (including carbon nanotubes and graphene) and oxide materials (such as titania). In addition to the well-established application areas of plasmonic materials, biosensing, and the production of medical implant materials, we outline new directions for this field that have the potential to bring new advances in areas such as energy materials and regenerative medicine.



phage display to isolate the first peptide sequences that could recognize solid Ag surfaces4 and silica surfaces.5 These pioneering works gave rise to a new field of research. Peptide sequences have since been identified for a wide range of materials, as summarized by Care et al.6 It is noteworthy that some substrates can be oxidized under aqueous conditions, such as Ti7 and Al,8 which carries implications for the

BACKGROUND Since the first pioneering study published by Stanley Brown in 1997,1 there has been huge interest in the use of combinatorial libraries to identify peptide sequences with the ability to recognize and bind to different solid materials. This concept was subsequently and rapidly popularized by four landmark publications: Whaley et al.,2 who used phage-display libraries to isolate sequences with specific binding for semiconductor surfaces; Sarikaya and co-workers,3 who further probed Brown’s Au-binding sequences; and Naik and co-workers, who used © 2017 American Chemical Society

Received: February 2, 2017 Published: June 30, 2017 1617

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research

and protonation state of the peptide; and the appropriate interfacial solvent structure. The solid surface typically considered in many abiotic−biotic modeling studies is crystalline, but approximations to the amorphous surface have been advanced, e.g., for SiO2.13,14 While crystalline substrates are preferred for many molecular simulations, the use of crystalline targets in quantitative experimental binding analyses is rare. These measurements, such as with quartz crystal microbalance (QCM) observations or surface plasmon resonance (SPR) spectroscopy, hinge on the ability to coat the sensor with a thin layer of the target material. This coating may be amorphous or polycrystalline at best. Therefore, for MD simulations to connect with SPR and QCM data, it must be acknowledged that the target substrates are likely quite different. However, experimentally measured amino acid binding strengths for both crystalline and amorphous titania interfaces show agreement in binding trends,15,16 and comparisons with simulation data are also consistent.17 The question of nonideal substrates aside, adsorption onto different facets of the same crystalline material may be very different in the case of amino acids18,19 and peptides,19,20 as was considered by some of the first simulations of biocombinatorially selected Au-binding peptides.21 A recent example from our laboratories is facet-dependent peptide adsorption on the solid Au surface.20 We predicted the binding free energies of a dodecapeptide, AuBP1,22 in solution for three different Au facets and used knowledge of the facet prevalence on polycrystalline Au to obtain a weighted free energy average for direct comparison with QCM and SPR measurements. This strategy yielded excellent agreement with the experimental data, which would not have been possible if only the Au(111) interface had been considered. The validity of approximating the aqueous Au polycrystalline surface with aqueous Au(111) is another aspect. The Au(111) surface is thought to be the dominant facet at the aqueous interface of polycrystalline Au. For flat, planar Au surfaces (i.e., not on nanoparticle surfaces), the consensus view is that the Au(100) (1 × 1) surface reconstructs to Au(100) (5 × 1), yielding a lateral hexagonal arrangement of Au atoms in the surface plane similar to that on Au(111), as shown in Figure 1.23 This lateral similarity could explain why peptide adsorption on the aqueous Au(100) (5 × 1) surface was similar to that on Au(111),20 suggesting that the Au(111) surface is an appropriate model for comparison with experimental data obtained from polycrystalline Au QCM sensor surfaces. However, this argument is not necessarily true for Au nanoparticle surfaces because it is not currently known whether the {100} planes on the surface of a faceted Au nanoparticle under aqueous conditions are in the native or reconstructed form. This may impact peptide adsorption (aside from edge effects, etc.), as our simulations indicated that peptide adsorption at the native Au(100) surface is considerably weaker than that on Au(100) (5 × 1) or Au(111) .19,20 Overall, the task of modeling the experimentally relevant substrate can be complex, even for a material as well-studied as Au. The presence of structural defects on surfaces, which effectively introduces a degree of nonperiodicity to the model, and their impact on peptide adsorption merit deeper investigation. However, there is a lack of experimental data regarding the spatial distribution of defects on the surface. In our own work on modeling of oxygen-containing defects on carbon nanotubes (CNTs), we investigated the influence on

appropriate choice of structural model used in the simulations. Given the wealth of materials-binding peptide sequences reported to date, the bottleneck of this research is now concerned with acquiring and interpreting structure/function data to refine, optimize, and adapt these sequences to realize practical goals in peptide-mediated growth, assembly, and activation of functional nanomaterials in aqueous media. A chief obstacle hindering progress in elucidating these structure−function relationships lies in obtaining reliable structural data. The challenges in experimentally obtaining atomic-level details of the structure of any molecule adsorbed at an aqueous surface are numerous. While techniques such as NMR spectroscopy can readily resolve molecular structures in solution, this problem escalates to a challenging task once the molecule is adsorbed at an aqueous solid interface. Advances in techniques such as in situ atomic force microscopy,9 sumfrequency generation spectroscopy,10 and NMR spectroscopy11 are progressing, but specific challenges remain for materialsbinding peptides because typically these peptides are thought to be instrinsically disordered (and herein are termed intrinsically disordered peptides/proteins (IDPs)). Typically, IDPs12 cannot be adequately categorized by one single structure and are better described as a collection (ensemble) of structures. Consequently, materials-binding peptides may lack well-defined secondary structure. Moreover, experimental techniques that rely on the presence of secondary structure (e.g., circular dichroism spectroscopy), which are usually employed to characterize biomolecule adsorption at aqueous interfaces, can be uninformative when applied to materials-binding peptides in the adsorbed state. For this reason, advanced molecular simulation approaches have been pivotal to providing these much-needed structural details. However, such simulations also involve challenges; in summary, three key conditions must be adequately addressed for molecular simulations of the abiotic−biotic interface to be both credible and relevant to experimental findings: (1) the interatomic interactions (particularly across the solid−liquid interface) must be reasonably captured; (2) conformational sampling of the structural ensemble of the surface-adsorbed IDP must be appropriate; and (3) the atomic-scale structure of the solid surface must be physically reasonable. Validating the outputs and inputs of molecular simulation are essential; one of the serious caveats of molecular modeling in general is that it is always possible to generate data, regardless of whether these data are meaningful. For this reason, close connection with experiments is a necessary condition that should be met wherever possible. Furthermore, there always exists a compromise between system size (the number of atoms) and the quality of the simulation (the degree of sophistication of the implemented simulation approach).



CHALLENGE 1: STRUCTURAL MODELS Atomic-scale structural models of the interface that are appropriate to experimental conditions while also remaining computationally tractable are an essential element of any credible molecular dynamics (MD) simulation of the abiotic− biotic interface. By necessity, these structural models must be approximate. This does not imply that data from such MD simulations are not meaningful or useful, but instead requires that these data be interpreted appropriately. By “structural model” we refer to the composition and atomic-scale structure of the solid surface; the sequence, bond-topological structure, 1618

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research

Figure 3. Illustration of exemplar periodic simulation cells comprising (a) a slit-pore configuration containing a partially hydroxylated negatively charged TiO2(110) slab, Na+ (blue), Cl− (green), and the Ti-1 peptide (water not shown for clarity; data taken from ref 69) and (b) a 5 nm Au nanoparticle in an aqueous medium in the presence of several adsorbed Au-binding peptides (data taken from ref 27). The blue lines highlight the edges of the periodic simulation cell.

Figure 1. Differences in the lateral spatial distribution of atoms and their lateral interfacial water structuring in the surface planes of (a) Au(111), (b) the native Au(100) surface, and (c) the reconstructed Au(100) (5 × 1) surface. The colors used to indicate the atoms in (c) are not related to those in the right-hand legend. Reproduced with permission from ref 20. Copyright 2015 Royal Society of Chemistry.

to ensure that the interfacial solvent structuring can decay to bulk-solvent values within the limits of the periodic cell. In our recent work predicting the catalytic properties of Au nanoparticles, our largest system size incorporated a peptide-capped Au nanoparticle with a diameter of ∼5 nm into a cell of dimensions 100 Å × 100 Å × 100 Å, comprising ∼100 000 atoms,27 as depicted in Figure 3b. While standard MD simulations of systems with this number of atoms are not uncommon, standard MD simulations of these complex interfaces are not guaranteed to yield meaningful results. Instead, advanced sampling approaches are required. This system size illustrated in Figure 3b represents our upper limit in terms of carrying out simulations using advanced conformational sampling techniques, which required substantial access to high-performance computing facilities. Such large system sizes effectively limit both the degree of conformational sampling that can be achieved (even when advanced sampling is used) and the sophistication of the description of the interatomic interactions, since all three factors are mutually interdependent. In the future, closer integration between molecular simulation efforts and experimental approaches that can resolve the atomic-scale structure of the substrate, such as that reported by Bedford et al.,27 is desirable. Furthermore, advances in experimental techniques for resolving the spatial distribution of defect sites at the molecular level on a given substrate would enable modeling approaches to more credibly explore the structural consequences of surface nonideality.

peptide adsorption by considering two extremes of defect spatial distribution, either grouped together or dispersed (see Figure 2).24 Moreover, as a result of the use of three-

Figure 2. Representative snapshot taken from simulations reported in ref 24, corresponding to peptide adsorption on a nonidealized carbon nanotube surface where hydroxyl defects are grouped together.

dimensional periodic boundary conditions, solid surfaces in MD simulations in this field are usually represented as slabs of a finite thickness but modeled with effectively infinite lateral extent, resulting in a planar “slit-pore” type configuration of the interface (see Figure 3a). If the solid surface is modeled as a substrate on which the peptide adsorbs (i.e., if the process of materials nucleation is not relevant), the atoms of the surface may be partially or completely held fixed, e.g., because the use of high-temperature sampling protocols might otherwise melt the substrate. Our studies suggested that neglect of surface flexibility does not result in substantial differences in predicted binding strength compared with a flexible model.25 However, this infinite-planar simulation setup is not appropriate when comparing MD simulations with experimental characterization of peptide adsorption on nanomaterials, where the presence of nonperiodic features such as edges and vertices cannot be captured.26 The task of modeling peptide adsorption on nanomaterials in aqueous media is more challenging because this generally requires the use of a large periodic simulation cell to accommodate the nonperiodic substrate (e.g., nanoparticle) and a sufficient amount of solvent



CHALLENGE 2: FORCE FIELD DEVELOPMENT Interfacial interatomic potentials (herein called force fields) are essential for describing all relevant interactions across the abiotic−biotic interface and are closely connected with the structural model of the solid substrate. Validation of the biointerfacial force field is a priority prior to its use in production simulations. An unfortunate caveat is that anyone with access to a molecular simulation software package can naively conflate two force fields together (one for the substrate and one for the biomolecule) and subsequently use this unvalidated force field to generate molecular simulation output, the result of which may or may not be meaningful. Caution and careful appraisal are advisible when citing data from such sources. Validation and refinement of biointerfacial force fields remain an ongoing prospect. Counterpart experimental data required for this validation can be challenging to obtain and open to 1619

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research

found that bespoke cross-terms for specific interfacial interactions were necessary for our polarizable biointerfacial force fields for Au, Ag, and graphene. The harmonization process also must take the force field for water, ions, etc. into account. The most widely used biomolecule force fields have been developed for use with specific water models; for example, the CHARMM family40 of force fields is intended for use with the TIP3P water model.41 However, some mineral/water force fields42 have been developed in conjunction with nonbiological water models, such as SPC/fw,43 and the use of a water model different from that used to parametrize the biomolecule force field may result in unreliable results. That said, our own work comparing the conformational ensemble of tripeptides described using the CHARMM-27 force field, generated with TIP3P water model versus the SPC/fw water model, suggested that the SPC/fw−CHARMM hybrid performed acceptably in this instance.44 A similar degree of caution should be exercised to ensure that the interaction between aqueous ions and peptides can be captured appropriately. For instance, we reparametrized bespoke pairwise interactions in CHARMM22*40,45 to capture carboxylate−Ca2+ interactions in solution in order to ensure agreement with experimental data.46 Among the most extensively developed force fields for peptide−surface interactions, the substrates Au and graphite/ graphene/CNTs are strongly featured. In this instance, the inclusion of polarization effects may be desirable.47,48 We developed a peptide/CNT force field48 based on the polarizable AMOEBA force field49 using the Thole model of polarization50 and the distributed multipole approximation.51 Simulations using this force field24,52,53 were extremely demanding in terms of computational resources,35 thus limiting the degree to which the peptide conformations could be sampled and (by necessity) neglecting a molecular-level description of liquid water. To enable modeling of substantial system sizes (including solvent) and to allow for robust conformational sampling, polarization effects must be accommodated as economically as possible into the force field. Worthy of note in this area was the development of an economical polarizable biointerface force field for Au(111), GolP, advanced by Corni and co-workers,34 which was a pioneering contribution to the field. The parametrization of GolP was intended for use with the OPLS force field; we consequently modified GolP to harmonize this with the more suitable CHARMM family of force fields and extended its applicability to the Au(111), Au(100) (1 × 1), and Au(100) (5 × 1) interfaces, resulting in the GolP-CHARMM force field.23,36 We have applied a similar strategy for generating polarizable biointerface force fields for graphene and Ag(111)/ (100). Despite the dearth of amino acid binding data from experimental sources, the consistency in our approach in creating these force fields for Au, Ag, and graphene allowed us to compare adsorption on an equal footing across these three materials. We used metadynamics simulations54 to predict the adsorption free energies of amino acids at aqueous interfaces of Au(111), Ag(111),55 and graphene (0001).56 In Figure 5 we compared amino acid adsorption for Au/Ag, Au/graphene, and Ag/graphene. For example, these data indicate that very few amino acids have a thermodynamic preference to adsorb at aqueous Ag(111) over Au(111), which may reflect the difficulty of isolating peptide sequences that are strongly selective for Ag over Au. However, we note that peptide adsorption is not merely an additive sum of the binding affinities of the individual

interpretation. Experimentally observed adsorption free energies of amino acids and short peptides are a highly preferred contribution to the process of fitting and validating such force fields; however, few such data are available.15,16,28,29 Alternatively, while numerous experimental binding energies are available for larger peptides, the utility of these in the fitting and validation process, even for homopeptides, requires careful inference to isolate residue-specific binding information. This difficulty in extracting residue-level binding propensities from peptide binding can be attributed to the lack of conformational control inherent to their IDP character. This results in a complex interplay between a peptide sequence, its conformational ensemble, and its concomitant adsorption strength at the interface. For instance, for Au there are very few amino acid adsorption data from experimental sources. Worthy of note is the binding free energy reported for phenylalanine adsorbed at the aqueous Au(111) interface;28 using GolP-CHARMM our predicted value30 of −20.6 ± 0.5 kJ mol−1 compared favorably to the experimentally determined range of −18 to −37 kJ mol−1. Beyond this, quantitative binding data have been reported only for larger peptides. Excellent examples of a careful fitting and validation procedure can be found in the work of Latour and co-workers,31 who have made extensive use of host−guest peptide sequences, along with SPR experiments and molecular simulation, to isolate residue-level binding propensities, albeit at self-assembled monolayer surfaces and not at aqueous interfaces of hard materials. There are different philosophies regarding the process of parametrizing biointerfacial force fields, e.g., fitting to experimental data32−34 or, as was done in our laboratories, fitting to data from first-principles calculations.23,35−37 For the latter, advances in the developments in density functionals that can reasonably capture the nonbonded interactions between molecules and surfaces, such as the vDW-DF family of functionals,38 were a critical factor that enabled this parametrization strategy. We used vdW-DF calculations on small molecules adsorbed at the Au(111) surface as part of our fitting data set; the comparison between the experimental and calculated values was excellent (Figure 4). In addition,

Figure 4. Comparison of Au(111)-binding energies of small molecules obtained from experiment and vdW-DF calculations. Adapted from ref 36. Copyright 2013 American Chemical Society.

challenges exist for combining the force field for the biomolecule with that of minerals, where the partial atomic charges for similar types of atoms (e.g., oxygen) can be very different but (usually) should not be altered. Procedures to harmonize these descriptions via the generation of bespoke cross-term potentials have proved successful.39 Similarly, we 1620

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Accounts of Chemical Research



Article

CHALLENGE 3: CONFORMATIONAL SAMPLING

The main consequence of the IDP character of many materialsbinding peptides is that these molecules cannot be readily categorized as possessing a distinct “structure”. While the archetypal energy landscapes57 associated with materialsbinding peptides remain to be comprehensively established, indications point to these landscapes as complex. For this reason, merely running microsecond-long standard MD simulations may not provide sufficient sampling,30 and targeted conformational sampling strategies may be necessary. Furthermore, given the lack of well-defined structural traits in many materials-binding peptides, sampling approaches that do not a priori assume any particular structural traits are preferable. To address this, we adapted the replica exchange with solute tempering molecular dynamics (REST-MD) simulation approach58,59 and have applied this strategy to a wide range of materials-binding peptides. In general, replica-exchange approaches seek to accelerate conformational sampling by surmounting barriers on the potential energy landscape; they do so without requiring a priori knowledge of the relevant reaction coordinates on the potential energy landscape. The basic premise of the replica-exchange approach is the use of several replicas (chemically identical copies of the system), each of which is run in an MD simulation in synchrony with the others. In principle, the key outcome of a REST-MD simulation is the Boltzmann-weighted ensemble of conformations of the peptide. In temperature-based REMD, each replica is run at a slightly different thermal temperature. At fixed time intervals during the simulation, a swap between neighboring replicas is attempted, the success of which is determined by applying the Metropolis criterion. By this swapping process, conformations from the “hotter” replicas can ultimately appear in the baseline replica trajectory. REMD simulations in principle can be applied to the aqueous peptide−materials interface, but in practice, the number of replicas required can escalate to impractical levels if liquid water is modeled explicitly. While the evolution of implicit solvent models is progressing, it is still a commonly held view that a description of explicit solvation is required to capture physically reasonable properties of these biointerfaces. This REMD impasse can be addressed using REST-MD, which scales selected parts of the Hamiltonian, not the thermal temperature; the result is a substantial reduction in the number of replicas needed to ensure reasonable acceptance probabilities.58 In addition to REST-MD simulations, numerous alternative computational strategies are available. Such approaches include (but are not limited to) umbrella sampling (US),60 accelerated molecular dynamics (aMD),61 and the temperature interval with global exchange of replicas (TIGER) algorithm and subsequent variants).62−64 Steered MD, TIGER, US, and metadynamics-based54 approaches have found use in the application to materials-adsorbed peptides.20,31,65−69 In summary, simulations using advanced conformational sampling strategies should ideally become the norm for these biointerfaces, not the exception. From an experimental perspective, advances in techniques devised specifically for characterizing the structure(s) of IDP systems and their translation to biointerfacial systems would be the ideal complement to these simulation efforts.

Figure 5. Comparison of predicted adsorption free energies for amino acids at the (a) Au(111) and Ag(111), (b) graphene and Au(111), and (c) graphene and Ag(111) aqueous interfaces. Data were taken from refs 55 and 56. Residues for which (a) Ag binding ≥ Au binding, (b) graphene binding ≥ Au binding, and (c) graphene binding ≥ Ag binding are highlighted.

residues in the peptide sequence and that the position and environment of each residue in the peptide may modulate its individual binding propensity.30 That said, amino acid binding propensities provide a valuable benchmark against which such modulations can be observed and can be used to help interpret the adsorption properties of peptides. In summary, thorough validation processes for interfacial force fields are essential in this field. From a modeling perspective, a greater focus on the application of advanced sampling techniques, particularly those based on metadynamics, to the prediction and evaluation of peptide−surface adsorption free energies is needed, particularly for larger peptides, to enable connections with previous experiments. To complement this, advances in experiments are also needed to enable quantitative binding evaluations for smaller molecules, such as tripeptides, which are amenable to more accurate predictions via molecular simulation. We also remark here that the inference of binding constants from experimental observations typically involves similar challenges with assumptions, e.g., regarding structural models of the substrate, etc. 1621

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research



OUTLOOK The maturity of many aspects of simulation approaches used to investigate aqueous peptide−materials interfaces has progressed over the past decade. This has enabled the transition from investigation of fundamental interfacial structure(s) to interpretation and prediction of their properties, particularly for those based on Au and graphene. Translation of these achievements into biointerfaces made with earth-abundant substrates is necessary to expand the range of possible applications. Complex oxide materials, magnetic materials, and multimetallic materials (e.g., FeNi) all have promising potential, particularly for the bio-based production of cheaper nanocatalysts. The lack of mature, validated structural models and force fields for these more complex materials hinders this potential, and these developments should be a priority for future focus. The production and dispersion of nanoparticles with nonspherical shapes in aqueous media without recourse to covalent functionalization is another application that holds great promise for catalysis and sensing. Fundamental concepts on the use of peptides as shape-control agents for realizing aqueous nanoparticle growth are evolving, but significant developments are still needed. Another key direction for peptide−materials interfaces lies in the nanomedicine space, particularly in the more complex realm of protein−nanoparticle interfaces. For example, exploitation of simulation approaches, as a complement to experimental efforts, to elucidate a deeper comprehension of the structure (and therefore properties) of the protein corona would provide foundational contributions to this field. In this respect, the development and use of nonatomistic, coarsegrained models that can recover the relevant chemistry and physics of these biointerfaces is a necessary but challenging prospect70 for realizing this goal. While significant strides have been made in this area, more development is needed. For example, the development and use of bespoke hardware devoted to solely to MD simulations offers the ability to perform simulations with substantially enhanced time scales pushing beyond the millisecond regime. Moreover, the contentrich datasets produced by molecular simulation, in combination with quantitative experimental data, have enormous potential for exploitation via data-mining approaches. Finally, advancements in molecular simulation approaches depend on the development of new, more sophisticated experimental techniques. Collaboration with experimental teams and the cutting-edge development of experimental approaches are indispensable for enabling progress in molecular simulation approaches to help in meeting these future challenges.



Materials at the University of Oxford, she joined the faculty of the University of Warwick in the Department of Chemistry. In 2012 she returned to Australia to the Institute for Frontier Materials at Deakin University, where she is now a Professor of Bio/Nanotechnology. Her research interests focus on computational modeling of the interface between biomolecules and synthetic materials using molecular dynamics simulations.



ACKNOWLEDGMENTS Support from the EPSRC (Grant EP/I001514/1), the AFOSR (Grant FA9550-12-1-0226), and veski is gratefully acknowledged. The author is grateful for access to supercomputing resources provided by the University of Warwick, U.K.; the ARCHER National Supercomputing Service, U.K.; the Victorian Life Sciences Computation Initiative (VLSCI), Australia; the National Computing Infrastructure (NCI), Australia; and the Pawsey Centre, Australia. The author thanks past and present team members for their valuable contributions to this research area.



REFERENCES

(1) Brown, S. Metal-Recognition by Repeating Polypeptides. Nat. Biotechnol. 1997, 15, 269−272. (2) Whaley, S. R.; English, D. S.; Hu, E. L.; Barbara, P. F.; Belcher, A. M. Selection of Peptides with Semiconductor Binding Specificity for Directed Nanocrystal Assembly. Nature 2000, 405, 665−668. (3) Brown, S.; Sarikaya, M.; Johnson, E. A Genetic Analysis of Crystal Growth. J. Mol. Biol. 2000, 299, 725−735. (4) Naik, R. R.; Stringer, S. J.; Agarwal, G.; Jones, S. E.; Stone, M. O. Biomimetic Synthesis and Patterning of Silver Nanoparticles. Nat. Mater. 2002, 1, 169−172. (5) Naik, R. R.; Brott, L. L.; Clarson, S. J.; Stone, M. O. SilicaPrecipitating Peptides Isolated from a Combinatorial Phage Display Peptide Library. J. Nanosci. Nanotechnol. 2002, 2, 95−100. (6) Care, A.; Bergquist, P. L.; Sunna, A. Solid-Binding Peptides: Smart Tools for Nanobiotechnology. Trends Biotechnol. 2015, 33, 259−268. (7) Sano, K. I.; Shiba, K. A Hexapeptide Motif That Electrostatically Binds to the Surface of Titanium. J. Am. Chem. Soc. 2003, 125, 14234− 14235. (8) Adams, B. L.; Finch, A. S.; Hurley, M. M.; Sarkes, D. A.; StratisCullum, D. N. Genetically Engineered Peptides for Inorganics: Study of an Unconstrained Bacterial Display Technology and Bulk Aluminum Alloy. Adv. Mater. 2013, 25, 4585−4591. (9) Chen, C. L.; Zuckermann, R. N.; DeYoreo, J. J. Surface-Directed Assembly of Sequence Defined Synthetic Polymers into Networks of Hexagonally Patterned Nanoribbons with Controlled Functionalities. ACS Nano 2016, 10, 5314−5320. (10) Ding, B.; Jasensky, J.; Li, Y. X.; Chen, Z. Engineering and Characterization of Peptides and Proteins at Surfaces and Interfaces: A Case Study in Surface-Sensitive Vibrational Spectroscopy. Acc. Chem. Res. 2016, 49, 1149−1157. (11) Mirau, P. A.; Naik, R. R.; Gehring, P. Structure of Peptides on Metal Oxide Surfaces Probed by Nmr. J. Am. Chem. Soc. 2011, 133, 18243−18248. (12) Habchi, J.; Tompa, P.; Longhi, S.; Uversky, V. N. Introducing Protein Intrinsic Disorder. Chem. Rev. 2014, 114, 6561−6588. (13) Butenuth, A.; Moras, G.; Schneider, J.; Koleini, M.; Koppen, S.; Meissner, R.; Wright, L. B.; Walsh, T. R.; Colombi Ciacchi, L. Ab Initio Derived Force-Field Parameters for Molecular Dynamics Simulations of Deprotonated Amorphous-SiO2/Water Interfaces. Phys. Status Solidi B 2012, 249, 292−305. (14) Emami, F. S.; Puddu, V.; Berry, R. J.; Varshney, V.; Patwardhan, S. V.; Perry, C. C.; Heinz, H. Force Field and a Surface Model Database for Silica to Simulate Interfacial Properties in Atomic Resolution. Chem. Mater. 2014, 26, 2647−2658.

AUTHOR INFORMATION

Corresponding Author

*E-mail: tiff[email protected]. Phone: +61 352 273 116. ORCID

Tiffany R. Walsh: 0000-0002-0233-9484 Notes

The author declares no competing financial interest. Biography Tiffany R. Walsh graduated with a B.Sci(Hons) from the University of Melbourne and then earned her Ph.D. in theoretical chemistry from the University of Cambridge, U.K., as a Cambridge Commonwealth Trust Scholar. Following a Glasstone Fellowship in the Department of 1622

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research

Using 12−6 and 9−6 Lennard-Jones Potentials. J. Phys. Chem. C 2008, 112, 17281−17290. (34) Iori, F.; Di Felice, R.; Molinari, E.; Corni, S. GolP: An Atomistic Force-Field to Describe the Interaction of Proteins with Au(111) Surfaces in Water. J. Comput. Chem. 2009, 30, 1465−1476. (35) Hughes, Z. E.; Tomasio, S. M.; Walsh, T. R. Efficient Simulations of the Aqueous Bio-Interface of Graphitic Nanostructures with a Polarisable Model. Nanoscale 2014, 6, 5438−5448. (36) Wright, L. B.; Rodger, P. M.; Corni, S.; Walsh, T. R. GolPCHARMM: First-Principles Based Force Fields for the Interaction of Proteins with Au(111) and Au(100). J. Chem. Theory Comput. 2013, 9, 1616−1630. (37) Hughes, Z. E.; Wright, L. B.; Walsh, T. R. Biomolecular Adsorption at Aqueous Silver Interfaces: First-Principles Calculations, Polarizable Force-Field Simulations, and Comparisons with Gold. Langmuir 2013, 29, 13217−13229. (38) Berland, K.; Cooper, V. R.; Lee, K.; Schroder, E.; Thonhauser, T.; Hyldgaard, P.; Lundqvist, B. I. Van Der Waals Forces in Density Functional Theory: A Review of the vdW-DF Method. Rep. Prog. Phys. 2015, 78, 066501. (39) Freeman, C. L.; Harding, J. H.; Cooke, D. J.; Elliott, J. A.; Lardge, J. S.; Duffy, D. M. New Forcefields for Modeling Biomineralization Processes. J. Phys. Chem. C 2007, 111, 11943− 11951. (40) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. All-Atom Empirical Potential for Molecular Modeling and Dynamics Studies of Proteins. J. Phys. Chem. B 1998, 102, 3586−3616. (41) Neria, E.; Fischer, S.; Karplus, M. Simulation of Activation Free Energies in Molecular Systems. J. Chem. Phys. 1996, 105, 1902−1921. (42) Raiteri, P.; Demichelis, R.; Gale, J. D. Thermodynamically Consistent Force Field for Molecular Dynamics Simulations of Alkaline-Earth Carbonates and Their Aqueous Speciation. J. Phys. Chem. C 2015, 119, 24447−24458. (43) Wu, Y.; Tepper, H. L.; Voth, G. A. Flexible Simple Point-Charge Water Model with Improved Liquid-State Properties. J. Chem. Phys. 2006, 124, 024503. (44) Desmond, J. L.; Rodger, P. M.; Walsh, T. R. Testing the InterOperability of the Charmm and SPC/fw Force Fields for Conformational Sampling. Mol. Simul. 2014, 40, 912−921. (45) Piana, S.; Lindorff-Larsen, K.; Shaw, D. E. How Robust Are Protein Folding Simulations with Respect to Force Field Parameterization? Biophys. J. 2011, 100, L47−L49. (46) Church, A. T.; Hughes, Z. E.; Walsh, T. R. Improving the Description of Interactions between Ca2+ and Protein Carboxylate Groups, Including Gamma-Carboxyglutamic Acid: Revised CHARMM22*Parameters. RSC Adv. 2015, 5, 67820−67828. (47) Lu, D. Y.; Li, Y.; Ravaioli, U.; Schulten, K. Empirical Nanotube Model for Biological Applications. J. Phys. Chem. B 2005, 109, 11461− 11467. (48) Tomasio, S. D.; Walsh, T. R. Atomistic Modelling of the Interaction between Peptides and Carbon Nanotubes. Mol. Phys. 2007, 105, 221−229. (49) Ren, P. Y.; Ponder, J. W. Polarizable Atomic Multipole Water Model for Molecular Mechanics Simulation. J. Phys. Chem. B 2003, 107, 5933−5947. (50) Thole, B. T. Molecular Polarizabilities Calculated with a Modified Dipole Interaction. Chem. Phys. 1981, 59, 341−350. (51) Stone, A. J.; Alderton, M. Distributed Multipole Analysis Methods and Applications. Mol. Phys. 1985, 56, 1047−1064. (52) Akdim, B.; Pachter, R.; Kim, S. S.; Naik, R. R.; Walsh, T. R.; Trohalaki, S.; Hong, G. Y.; Kuang, Z. F.; Farmer, B. L. Electronic Properties of a Graphene Device with Peptide Adsorption: Insight from Simulation. ACS Appl. Mater. Interfaces 2013, 5, 7470−7477.

(15) Roddick-Lanzilotta, A. D.; McQuillan, A. J. An in Situ Infrared Spectroscopic Study of Glutamic Acid and of Aspartic Acid Adsorbed on TiO2: Implications for the Biocompatibility of Titanium. J. Colloid Interface Sci. 2000, 227, 48−54. (16) Roddick-Lanzilotta, A. D.; Connor, P. A.; McQuillan, A. J. An in Situ Infrared Spectroscopic Study of the Adsorption of Lysine to TiO2 from an Aqueous Solution. Langmuir 1998, 14, 6479−6484. (17) Sultan, A. M.; Hughes, Z. E.; Walsh, T. R. Binding Affinities of Amino Acid Analogues at the Charged Aqueous Titania Interface: Implications for Titania-Binding Peptides. Langmuir 2014, 30, 13321− 13329. (18) Wright, L. B.; Walsh, T. R. Facet Selectivity of Binding on Quartz Surfaces: Free Energy Calculations of Amino-Acid Analogue Adsorption. J. Phys. Chem. C 2012, 116, 2933−2945. (19) Hughes, Z. E.; Kochandra, R.; Walsh, T. R. Facet-Specific Adsorption of Tri-Peptides at Aqueous Au Interfaces: Reconciling Experiment and Simulation. Langmuir 2017, 33, 3742−3754. (20) Wright, L. B.; Palafox-Hernandez, J. P.; Rodger, P. M.; Corni, S.; Walsh, T. R. Facet Selectivity in Gold Binding Peptides: Exploiting Interfacial Water Structure. Chem. Sci. 2015, 6, 5204−5214. (21) Braun, R.; Sarikaya, M.; Schulten, K. Genetically Engineered Gold-Binding Polypeptides: Structure Prediction and Molecular Dynamics. J. Biomater. Sci., Polym. Ed. 2002, 13, 747−757. (22) Hnilova, M.; Oren, E. E.; Seker, U. O. S.; Wilson, B. R.; Collino, S.; Evans, J. S.; Tamerler, C.; Sarikaya, M. Effect of Molecular Conformations on the Adsorption Behavior of Gold-Binding Peptides. Langmuir 2008, 24, 12440−12445. (23) Wright, L. B.; Rodger, P. M.; Walsh, T. R.; Corni, S. FirstPrinciples-Based Force Field for the Interaction of Proteins with Au(100)(5 × 1): An Extension of GolP-CHARMM. J. Phys. Chem. C 2013, 117, 24292−24306. (24) Walsh, T. R.; Tomasio, S. M. Investigation of the Influence of Surface Defects on Peptide Adsorption onto Carbon Nanotubes. Mol. BioSyst. 2010, 6, 1707−1718. (25) Wright, L. B.; Freeman, C. L.; Walsh, T. R. Benzene Adsorption at the Aqueous (011) Alpha-Quartz Interface: Is Surface Flexibility Important? Mol. Simul. 2013, 39, 1093−1102. (26) Hughes, Z. E.; Walsh, T. R. Non-Covalent Adsorption of Amino Acid Analogues on Noble-Metal Nanoparticles: Influence of Edges and Vertices. Phys. Chem. Chem. Phys. 2016, 18, 17525−17533. (27) Bedford, N. M.; Hughes, Z. E.; Tang, Z. H.; Li, Y.; Briggs, B. D.; Ren, Y.; Swihart, M. T.; Petkov, V.; Naik, R. R.; Knecht, M. R.; Walsh, T. R. Sequence-Dependent Structure/Function Relationships of Catalytic Peptide-Enabled Gold Nanoparticles Generated under Ambient Synthetic Conditions. J. Am. Chem. Soc. 2016, 138, 540−548. (28) Li, H. Q.; Chen, A. C.; Roscoe, S. G.; Lipkowski, J. Electrochemical and FTIR Studies of L-Phenylalanine Adsorption at the Au(111) Electrode. J. Electroanal. Chem. 2001, 500, 299−310. (29) Cohavi, O.; Reichmann, D.; Abramovich, R.; Tesler, A. B.; Bellapadrona, G.; Kokh, D. B.; Wade, R. C.; Vaskevich, A.; Rubinstein, I.; Schreiber, G. A Quantitative, Real-Time Assessment of Binding of Peptides and Proteins to Gold Surfaces. Chem. - Eur. J. 2011, 17, 1327−1336. (30) Tang, Z. H.; Palafox-Hernandez, J. P.; Law, W. C.; Hughes, Z. E.; Swihart, M. T.; Prasad, P. N.; Knecht, M. R.; Walsh, T. R. Biomolecular Recognition Principles for Bionanocombinatorics: An Integrated Approach to Elucidate Enthalpic and Entropic Factors. ACS Nano 2013, 7, 9632−9646. (31) Snyder, J. A.; Abramyan, T.; Yancey, J. A.; Thyparambil, A. A.; Wei, Y.; Stuart, S. J.; Latour, R. A. Development of a Tuned Interfacial Force Field Parameter Set for the Simulation of Protein Adsorption to Silica Glass. Biointerphases 2012, 7, 56. (32) Heinz, H.; Lin, T. J.; Kishore Mishra, R.; Emami, F. S. Thermodynamically Consistent Force Fields for the Assembly of Inorganic, Organic, and Biological Nanostructures: The Interface Force Field. Langmuir 2013, 29, 1754−1765. (33) Heinz, H.; Vaia, R. A.; Farmer, B. L.; Naik, R. R. Accurate Simulation of Surfaces and Interfaces of Face-Centered Cubic Metals 1623

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624

Article

Accounts of Chemical Research (53) Tomasio, S. M.; Walsh, T. R. Modeling the Binding Affinity of Peptides for Graphitic Surfaces. Influences of Aromatic Content and Interfacial Shape. J. Phys. Chem. C 2009, 113, 8778−8785. (54) Barducci, A.; Bussi, G.; Parrinello, M. Well-Tempered Metadynamics: A Smoothly Converging and Tunable Free-Energy Method. Phys. Rev. Lett. 2008, 100, 020603. (55) Palafox-Hernandez, J. P.; Tang, Z. H.; Hughes, Z. E.; Li, Y.; Swihart, M. T.; Prasad, P. N.; Walsh, T. R.; Knecht, M. R. Comparative Study of Materials-Binding Peptide Interactions with Gold and Silver Surfaces and Nanostructures: A Thermodynamic Basis for Biological Selectivity of Inorganic Materials. Chem. Mater. 2014, 26, 4960−4969. (56) Hughes, Z. E.; Walsh, T. R. What Makes a Good GrapheneBinding Peptide? Adsorption of Amino Acids and Peptides at Aqueous Graphene Interfaces. J. Mater. Chem. B 2015, 3, 3211−3221. (57) Wales, D. J.; Miller, M. A.; Walsh, T. R. Archetypal Energy Landscapes. Nature 1998, 394, 758−760. (58) Terakawa, T.; Kameda, T.; Takada, S. On Easy Implementation of a Variant of the Replica Exchange with Solute Tempering in Gromacs. J. Comput. Chem. 2011, 32, 1228−1234. (59) Wright, L. B.; Walsh, T. R. Efficient Conformational Sampling of Peptides Adsorbed onto Inorganic Surfaces: Insights from a Quartz Binding Peptide. Phys. Chem. Chem. Phys. 2013, 15, 4715−4726. (60) Virnau, P.; Muller, M. Calculation of Free Energy through Successive Umbrella Sampling. J. Chem. Phys. 2004, 120, 10925− 10930. (61) Doshi, U.; Hamelberg, D. Towards Fast, Rigorous and Efficient Conformational Sampling of Biomolecules: Advances in Accelerated Molecular Dynamics. Biochim. Biophys. Acta, Gen. Subj. 2015, 1850, 878−888. (62) Li, X.; O’Brien, C. P.; Collier, G.; Vellore, N. A.; Wang, F.; Latour, R. A.; Bruce, D. A.; Stuart, S. J. An Improved Replica-Exchange Sampling Method: Temperature Intervals with Global Energy Reassignment. J. Chem. Phys. 2007, 127, 164116. (63) Li, X.; Latour, R. A.; Stuart, S. J. TIGER2: An Improved Algorithm for Temperature Intervals with Global Exchange of Replicas. J. Chem. Phys. 2009, 130, 174106. (64) Li, X. F.; Latour, R. A. The Temperature Intervals with Global Exchange of Replicas Empirical Accelerated Sampling Method: Parameter Sensitivity and Extension to a Complex Molecular System. J. Comput. Chem. 2011, 32, 1091−1100. (65) Deighan, M.; Pfaendtner, J. Exhaustively Sampling Peptide Adsorption with Metadynamics. Langmuir 2013, 29, 7999−8009. (66) Mijajlovic, M.; Penna, M. J.; Biggs, M. J. Free Energy of Adsorption for a Peptide at a Liquid/Solid Interface Via Nonequilibrium Molecular Dynamics. Langmuir 2013, 29, 2919−2926. (67) Schneider, J.; Colombi Ciacchi, L. Specific Material Recognition by Small Peptides Mediated by the Interfacial Solvent Structure. J. Am. Chem. Soc. 2012, 134, 2407−2413. (68) Baio, J. E.; Zane, A.; Jaeger, V.; Roehrich, A. M.; Lutz, H.; Pfaendtner, J.; Drobny, G. P.; Weidner, T. Diatom Mimics: Directing the Formation of Biosilica Nanoparticles by Controlled Folding of Lysine-Leucine Peptides. J. Am. Chem. Soc. 2014, 136, 15134−15137. (69) Sultan, A. M.; Westcott, Z. C.; Hughes, Z. E.; PalafoxHernandez, J. P.; Giesa, T.; Puddu, V.; Buehler, M. J.; Perry, C. C.; Walsh, T. R. Aqueous Peptide-TiO2 Interfaces: Isoenergetic Binding Via Either Entropically or Enthalpically Driven Mechanisms. ACS Appl. Mater. Interfaces 2016, 8, 18620−18630. (70) Rutter, G. O.; Brown, A. H.; Quigley, D.; Walsh, T. R.; Allen, M. P. Testing the Transferability of a Coarse-Grained Model to Intrinsically Disordered Proteins. Phys. Chem. Chem. Phys. 2015, 17, 31741−31749.

1624

DOI: 10.1021/acs.accounts.7b00065 Acc. Chem. Res. 2017, 50, 1617−1624