Modeling the Binding Affinity of Peptides for Graphitic Surfaces

Apr 23, 2009 - Interactions between peptide sequences and graphitic surfaces—carbon nanotubes and graphite—are investigated using molecular dynami...
1 downloads 14 Views 4MB Size
Download PDF
8778

J. Phys. Chem. C 2009, 113, 8778–8785

Modeling the Binding Affinity of Peptides for Graphitic Surfaces. Influences of Aromatic Content and Interfacial Shape Susana M. Toma´sio and Tiffany R. Walsh* Department of Chemistry and Centre for Scientific Computing, UniVersity of Warwick, CoVentry CV4 7AL, U.K. ReceiVed: October 3, 2008; ReVised Manuscript ReceiVed: April 1, 2009

Interactions between peptide sequences and graphitic surfacesscarbon nanotubes and graphitesare investigated using molecular dynamics simulations with a polarizable force-field. Peptide sequences selected to have a strong affinity for carbon nanotubes [Nat. Mater. 2003, 2, 196.] are rich in tryptophan. We investigate the importance of the tryptophan residue for two of these sequences by mutating each tryptophan with either tyrosine or phenylalanine. We find that, in line with recent experimental observations, the original, tryptophancontaining sequences support relatively stronger binding to both nanotubes and graphite, compared with the mutants. We ascribe this behavior to the additional structural stability conferred by the indole group at the interface. We also explore the effect of interfacial curvature on the binding affinity. Our findings suggest that these nanotube-binding peptides have also been selected for interfacial shape. For the graphite surface our results indicate a compromise exists between maintaining strong ring-surface interactions and allowing nonaromatic groups to also approach the surface. Introduction The recognition between inorganic materials and peptides in nature is a remarkable phenomenon. The discovery that peptides can recognize different artificial inorganic materials1,2 via combinatorial biology approaches has opened new vistas for fabrication of nanostructures.3-6 These screening approaches have identified hundreds of peptide sequences that bind to a range of materials such as metals, oxides, and semiconductors.7-9 However, how and why a given peptide sequence binds to a particular material is currently not well understood. Recent work suggests that the binding affinity cannot always be interpreted as a mere sum of possible residue-surface interactions but points to a more complex interplay between sequence, conformation, and binding.10-14 A number of previous studies have used screening techniques to identify sequences that bind to graphitic surfaces such as carbon nanotubes15-18 (CNTs), fullerenes,19 and nanohorns.20 There is considerable interest in the applications arising from tailored nanotube-peptide interactions,21 ranging from dispersal in aqueous media22,18 to biosensor23 and drug-delivery24 applications. The role of aromatic groups in the nanotube-peptide interaction has been highlighted,17,25 with the importance of tryptophan in particular under recent scrutiny.22,26-28 Experimental measurements of pyrene binding to CNTs29 indicated that this interaction is quite strong due to π-stacking, and therefore, it is not surprising that aromatic residues also have an affinity to graphitic surfaces. Atomic force microscopy (AFM) measurements and optical absorption spectra indicated that increasing the electron density of the aromatic residue increases the ability to disperse single-walled CNTs.30 Direct AFM measurements reported by Li et al.26 revealed that poly tryptophan exhibited greater adhesion on nanotubes compared with poly lysine. Salzmann et al. measured the dispersion effect of polypeptides comprising both pure and mixed sequences of lysine/tryptophan and lysine/tyrosine, and reported the best performance for the tryptophan-containing molecules.28 Fur* Corresponding author. E-mail: [email protected].

thermore, Xie et al.22 reported that among the three aromatic amino acids (tryptophan, phenylalanine, tyrosine) tested, the sequence containing tryptophan supported the greatest affinity for CNTs while the tyrosine was found to be more selective for individual CNTs, and phenylalanine the aromatic residue with the lowest affinity for CNTs. Although there is considerable growth in the experimental study of peptide-inorganic recognition, there are not as many modeling studies in this area. It is essential to understand the nature of molecular recognition and the degree of affinity of a selected peptide to a given surface, so that it can be rationalized, predicted, and optimized for a range of applications;31 molecular simulation can serve as a complementary tool alongside other experimental characterization techniques in realizing this goal. A number of studies have appeared for modeling recognition at the peptide-metal interface, including gold32 and platinum.33,34 In particular, modeling studies of phage-display peptides with different levels of affinity to the Pt(100) surface33 showed good agreement with experimental data despite using only van der Waals (vdw) terms to describe the nonbonded interactions. Peptide-oxide surface molecular dynamics (MD) simulations have been reported for peptides selected to bind to copper oxide and zinc oxide surfaces.35 In our previous work,36 we performed MD simulations of single-walled CNTs interacting with “strongbinder” and “weak-binder” aptamers, as identified by Wang et al.,16 with our validated extension of the AMOEBAPRO force-field,37,38 which treats both the molecule and the CNT as polarizable. Our results confirmed experimental observations16 of relative binding affinity for the systems studied. With the exception of our previous work, all of these other studies used traditional force-fields that are based on a distributed charge model for the electrostatics and do not include a description of polarization effects. Taking into account the importance of the aromatic content, our aim is to explore whether mutations of the experimentally determined “strong binder” sequences, HWKHPWGAWDTL (denoted B1) and HWSAWWIRSNQS (denoted B3), can recover similar binding affinity to CNTs. To this end we ran

10.1021/jp8087594 CCC: $40.75  2009 American Chemical Society Published on Web 04/23/2009

Binding Affinity of Peptides MD simulations of B1 and B3, and the corresponding mutants B1Y, B1F, B3Y, and B3F, where in each case each W was replaced by either Y or F, adsorbed on a carbon nanotube, to see if there was any change to the binding structures and binding affinity of the mutants. For example, the B1Y mutant corresponds with the sequence HYKHPYGAYDTL. Furthermore, experimental studies have indicated that peptide selection can distinguish between sequences that bind to CNTs and those that bind to graphite.16,18 To this end, we also ran similar simulations with these peptides adsorbed on a graphene sheet. Since the chemical composition of graphite and carbon nanotubes is the same, we considered these systems to be the ideal test-bed for investigating the influence of interfacial shape on the binding behavior of peptides. Our molecular simulations provide a detailed examination of peptide-CNT/peptide-graphite interactions for a system consisting of an (8,0) zigzag CNT segment/ graphene sheet and one of each of the six aptamers (each with 12 residues). Methods We performed MD simulations of systems comprising one surface (CNT or graphene sheet) and one of the peptide sequences B1, B3, B1Y, B1F, B3Y, B3F, and the control weakbinder NB1 (LPPSNASVADYS),16 using our own extensions of the AMOEBAPRO37,38 force-field. We have previously given details of the force-field extension for description of carbon nanotubes36 and graphite,39 and in each case good agreement is obtained when compared with results from electronic structure theory. Our previous results using the nanotube extension to the force-field also yielded excellent agreement with experimental observations.16 What makes AMOEBAPRO distinct from many other commonly used force-fields is (a) the description of electrostatics via the distributed multipole approximation,40,41 up to and including quadrupoles, and (b) inclusion of atombased polarizabilities to model induction effects. We believe that standard force-fields may not always yield a satisfactory description of the interaction between graphitic surfaces and the aromatic peptide groups, because in these force-fields electrostatic interactions are described by distributed charges. Because these idealized surfaces are modeled carrying little to no charge on each carbon atom, nonbonded interactions in this case reduce to a level of van der Waals (vdw) terms only. Use of distributed multipoles in our work enables an appropriate description of π-stacking between aromatic groups and the surface. Previous experimental work suggests a weak chargetransfer interaction between aromatic groups and carbon nanotubes,30 a phenomenon we can model with this polarizable forcefield. In the case of the peptide-CNT systems, we used a hydrogenterminated zigzag (8,0) nanotube with 976 carbon atoms such that the tube was at least four times the contour length of the peptide. In the case of the graphite surface interacting with the peptides, we modeled a graphene sheet comprising 2772 carbon atoms. All simulations were carried out in the canonical ensemble at room temperature. The Verlet42 algorithm was used to solve Newton’s equations of motion with an integration time step of 1.0 fs and a cutoff of 8 Å applied to all nonbonded interactions. The systems were equilibrated for 1 ns, followed by an additional production run of 1 ns. We checked that structural and energetic equilibration criteria were satisfied during our equilibration runs, such as measuring the fluctuations in the average distance between the tube surface and the aromatic ring center-of-mass, and the fluctuations in the average interaction energy between the tube and the surface; examples

J. Phys. Chem. C, Vol. 113, No. 20, 2009 8779 of such data are shown as a function of time in Figure S1 in the Supporting Information. We also give (Table S1 in the Supporting Information) the fluctuations in such properties in the post-equilibration period. To attempt to identify many different binding configurations, several different initial geometries were used for each aptamer. We employed a range of strategies for devising initial configurations for our simulations. We first constructed peptide configurations by hand and placed these near the nanotube surface. These geometries ranged from simple extended backbone configurations roughly alligned with the long axis of the nanotube, through to configurations where the backbone torsion angles were set to create pseudohelical peptide geometries that wrapped the nanotube. Second, we ran simulations of the peptide without the presence of the nanotube, in effective continuum solvent conditions (Vide infra). We took configurations from these lone peptide runs and placed these close to the nanotube surface. Finally, we took peptide-nanotube geometries that yielded strong binding and subsequently performed a number of “on the fly” mutations in both the forward (W to Y/F) and reverse (Y/F to W) directions. All simulations were performed using the TINKER43 package. To account for solvation effects, we used a continuum effective solvent, since the inclusion of water at the atomistic level would increase the simulation time by several orders of magnitude. The implicit solvent employed was the ASP model,44,45 used in conjunction with a modified background dielectric. This approach was found to yield physically reasonable behavior when used previously.36 Furthermore, in Table S2 and Figure S2 in the Supporting Information, we provide a comparison of structural properties for an example strong-binding B1Wnanotube configuration in both the implicit solvent and explicit solvent cases. These data clearly show that this configuration has not changed significantly due to the presence of explicit water, and that trends exhibited in these properties (and in the fluctuations in these properties) has not changed appreciably. Our ongoing test simulations of the peptide-CNT interface using explicit solvation therefore indicate this approximation is a reasonable one. We attribute the apparent success of this approximation to a combination of effects. We suggest that a vital role is played by the well-known lack of solvent structuring at the nanotube-water interface.46 This is in contrast with hydrophilic surfaces such as the rutile TiO2(110) surface, where water structuring is regarded as substantial.47 We have found in such hydrophilic cases that binding configurations generated from implicit solvent simulations changed significantly upon translation into an explicit solvent environment.48 In the current system, we suspect that it is this lack of structured solvent at the interface, in partnership with the high hydrophobic content of the aptamers studied, that has led to minimal structural changes once the explicit solvation is accounted for. We used both energetic and structural metrics to analyze our simulation data. To quantify binding affinity, we calculated the normalized interaction energy (interaction energy between peptide and surface, divided by the number of atoms in the peptide); this enables us to make a more fair comparison between systems with differing numbers of atoms in the sequence. Clearly, a calculated free energy of binding would a preferred measure of binding affinity. However, we remain convinced that the degree of statistical sampling required to obtain meaningful free energies of binding puts such calculations out of reach for such large systems at present. We examined other two other measures of binding affinity, related to EN. Instead of assigning a weight of 1 to each atom in the peptide as we have done in calculating EN, we tried two other weighting

8780

J. Phys. Chem. C, Vol. 113, No. 20, 2009

Toma´sio and Walsh

Figure 1. Normalized interaction energies as a function of time for original sequences B1 and B3, a mutated sequence B3Y, and the weakbinder control NB1 interacting with the nanotube.

TABLE 1: Averaged Normalized Interaction Energies, EN (kJ mol- 1 atom- 1), and the Standard Error of the Mean (SEM) of EN, Taken for the Four Distinct Trajectories of Lowest Energy for each Aptamer, for the Peptide-Nanotube Interface aptamer

EN

SEM

B1 B3 B1Y B3F B3Y B1F NB1

-1.32 -1.29 -1.15 -1.14 -1.12 -1.07 -0.87

0.04 0.03 0.06 0.03 0.07 0.08 0.06

schemes: one where we assigned a weight of 1 to any peptide atom within 6 Å of the surface with a zero weighting for all peptide atoms further from the surface, and one where we assigned a weight of 1 for any peptide atom closer than 4 Å away from the surface, with an exponentially decaying weight for all peptide atoms more distant. These data are shown in Table S3 of the Supporting Information, suggesting that the broad ranking of aptamer binding (mutants consistently showing diminished binding, with NB1 showing the least binding) does not change regardless of which of the three weightings are used. We analyzed the structure at the interface by calculating the distance from the center of each aromatic ring to the surface, and the orientation of each aromatic ring (with respect to the surface normal) was also calculated. For the ring-surface distances, we present our data as histograms, averaged over the production run trajectories. Results Peptide-Nanotube Simulations. For each aptamer, we report normalized interaction energies, EN, averaged over the four distinct binding trajectories that yielded the best binding affinities. Profiles of EN from the “best” trajectory of B1, B3, B3Y and NB1 interacting with the CNT are shown in Figure 1. The best trajectories of B1Y, B1F and B3F are similar to B3Y, so are omitted from Figure 1 for clarity. Normalized interaction energies (averaged over the four best trajectories) for all cases are presented in Table 1, alongside the standard error of the mean (SEM), as defined by the standard deviation in the sample means. The breakdown in

EN for each of the four trajectories for each aptamer are given in Table S4 of the Supporting Information. While neither mutation diminishes the binding affinity to a level comparable with NB1 (the weak-binding “control” peptide), both B3 and B1 show a reduction in binding affinity upon mutation of tryptophan for either phenylalanine or tyrosine. The gap in normalized energies between the wildtype and mutants is small, but taking into account that each peptide sequence has on average 200 atoms, it is still a considerable difference in the binding energy. Further, the differences in EN between the original and mutant sequences are greater than the SEM associated with each case. The differences among the binding affinities of the tyrosine mutants and phenylalanine mutants is smaller again. In terms of the best’ trajectory for each aptamer (see Table S4, Supporting Information) the tyrosine mutants show slightly greater binding than the phenylalanine mutants. The effects here are very subtle; the ordering among the phenylalanine and tyrosine mutants may change if these same simulations were conducted in explicit water. We propose that the origin of these energetic differences lie in the differences in binding geometry. In Figure 2 we show the distribution profile of the center-of-mass to CNT surface distances for H, W, Y, and F for the best trajectories of B1, B1Y and B1F. These data clearly show that all tryptophans (Figure 2a) are on average closer to the nanotube surface compared with the corresponding tyrosine and phenylalanine groups (Figure 2, parts b and c). This trend was found for all the configurations studied and also for the other peptide sequences B3, B3Y and B3F (Figure S3 in Supporting Information). Our best trajectories of B1 and B3 yield binding geometries where all the aromatic residues (tryptophans and histidines) bind to the CNT, whereas in the case of the mutant peptides, only two or three aromatic residues adopt binding configurations reasonably close to the surface. In general, the spatial proximity of the histidine in the first position (H1) was found to be more diffuse (as evidenced by broad peaks in Figure 2, even for B1), as it is at the extremity of the peptide chain. In Figure 3, an orientational analysis of the ring-tilt angle with respect to the surface normal reveals the origin of these differences in ring-surface distance. Here, we show example profiles of the orientation of tryptophan in position 2 (W2) for both B3 and B1, contrasted with the same profiles for B1Y and B3Y. For both B1 and B3 (Figure 3a), W2 remains steadily in a predominantly flat orientation (such that the rings lie tangential to the surface plane), while the mutants exhibit less orientational stability. Figure 3b highlights the differences in the origin of this instability; for B3F, our example trajectory supports an orientational profile of F2 where rapid jumps between flat orientations are noted, indicating ring-flipping events, while for B3Y, our example trajectory supports a profile for Y2 where the ring gently oscillates around an almost perpendicular orientation. In these examples we are not suggesting that F mutations must always yield ring-flipping events or that Y mutants must always show oscillatory behavior; we are merely using these examples to illustrate how ring orientation behavior for the mutants show differences compared with the original sequences, and how these differences will have consequences for the binding affinity. In the first case, ring flip events will necessarily place the ring center-of-mass further from the surface compared with a ring that does not flip. In the second case, a ring oriented perpendicularly to the surface will again not be able to approach the surface as closely. In both cases this will serve to decrease the tube-ring binding of these groups.

Binding Affinity of Peptides

J. Phys. Chem. C, Vol. 113, No. 20, 2009 8781

Figure 3. Profile of ring-tilt angle (relative to surface normal) at the peptide-nanotube interface as a function of trajectory for (a) W2 in both original sequences B1 and B3 and (b) Y2 and F2 in B3Y and B3F, respectively.

Figure 2. Distribution profile of the distance from the nanotube surface to the ring center of mass for all the aromatic residues H, W, Y and F for the peptide sequences (a) B1, (b) B1Y, and (c) B1F.

We found that all tryptophans in B1 and B3, and the nonterminal histidine in B1 (H4), always adopted a flat orientation at the interface. This observation is emphatically underscored by the average ring-tilt angle and fluctuation in this angle (calculated as the root-mean-square deviation of the tilt angle), presented as a bar chart in Figure 4 for B1 (Figure 4a), and the mutated sequences B1Y and B1F (Figure 4, parts b and c). The remaining data for B3, B3Y, and B3F can be found in Figure S4 in the Supporting Information. An average angle of around 0° indicates the ring is oriented flat on the surface, whereas a tilt of 90° indicates a perpendicular orientation. Large fluctuations indicate that the side chain of the aromatic residue is flexible and therefore that the aromatic ring is not so strongly bound, such that the ring can adopt several orientations relative to the surface. Figure 4a demonstrates that

tryptophans in B1 spend most of the time flat on the tube surface. The corresponding fluctuation is very small for all rings but H1, suggesting that once tryptophans bind to the CNT surface they remain stable with no significant changes in orientation. In contrast, in each of the mutated sequences, B1Y and B1F (Figure 4, parts b and c), at least one of the mutated aromatic residues (Y and F, respectively) exhibits a large average tilt with a greater corresponding fluctuation. These data provide evidence that in general, the orientation of the rings in the tyrosine and phenylalanine side groups is less rigid than is seen for the indole ring in tryptophan when interacting with the surface. The same phenomenon was evidenced for the other configurations studied and also for the other peptide sequences B3, B3Y, and B3F (Figure S4, Supporting Information), reinforcing the notion that tyrosine and phenylalanine are bound less tightly at the interface compared to tryptophan. In the case of H4 in B1, a small average tilt and a small fluctuation indicates that this residue is also flat and stable on the nanotube surface. However, for the mutated sequences B1Y and B1F, the corresponding H4 average tilt is around 90°, with a greater fluctuation. In each case, one of the nearby aromatic residues is also orientationally unstable (Y6 in Figure 4b and F2 in Figure 4c), suggesting possible structural effects that give rise to nonlocal modulation in the binding of these “key” residues. In keeping with our distance analysis, a general trend was again found for H1; in all the configurations studied, H1 yielded a large fluctuation in the tilt angle. Peptide-Graphite Simulations. The EN, as calculated as the average over the four best trajectories for each aptamer, and the corresponding SEM for the peptide-graphite interface are presented in Table 2. The breakdown in EN for each of the four trajectories for each aptamer are given in Table S5 of the Supporting Information. As found for the peptide-nanotube interface, these data show B1 and B3 to have greater binding affinity to the sheet compared with the mutants; however, in this case, the spread of interaction energies (as shown in Table S5, Supporting Information), especially for the original sequences, is wider than was seen for the peptide-nanotube case. Therefore, there is not as clear a distinction in the interaction energies between the original sequences and the mutant

8782

J. Phys. Chem. C, Vol. 113, No. 20, 2009

Toma´sio and Walsh TABLE 2: Normalized Interaction Energies, EN (kJ mol-1 atom-1), and the Standard Error of the Mean (SEM) of EN, Averaged over the Four Distinct Trajectories of Lowest Energy for Each Aptamer, for the Peptide-Graphite Interface aptamer

EN

SEM

B1 B3 B3F B1Y B3Y B1F NB1

-2.01 -1.93 -1.88 -1.87 -1.84 -1.83 -1.61

0.18 0.22 0.07 0.08 0.07 0.11 0.03

mutants. In Figure 5, we present the distribution profile of the distances of the ring center-of-mass to the graphene sheet for the best trajectories of B1, B1Y, and B1F (see Figure S5, Supporting Information, for B3 and variants). In contrast to the case for the tube, all rings in the B1 sequence, including H1, have maintained very close contact with the surface, yielding

Figure 4. Average and fluctuation of the ring-tilt angle (relative to surface normal) at the peptide-nanotube interface for all aromatic residues in (a) the original sequence B1 and the mutated sequences (b) B1Y and (c) B1F. A value of approximately 0° indicates a flat ring orientation, parallel with the surface.

sequences at the peptide-graphite interface. Again, the mutations for either phenylalanine or tyrosine do not push the binding affinity into the lower range supported by the NB1 weak-binder “control” peptide. However, a diminishing of the binding affinity was again generally noted for the mutant sequences. The differences in EN among the mutants is again very small. The “best” trajectory for each aptamer (Table S5, Supporting Information) shows a contrast with the peptide-nanotube case, in that the B3F mutant binds slightly better than B3Y, with the opposite being true for the B1 mutants. Again, solvation effects may change this ordering. The ring to surface distance analysis shows some similarity with the nanotube data, in the sense that the original sequences again maximize the ring-surface contact, compared with the

Figure 5. Distribution profile of the distance from the graphite surface to the ring center of mass for all the aromatic residues H, W, Y, and F for the peptide sequences (a) B1, (b) B1Y, and (c) B1F.

Binding Affinity of Peptides

J. Phys. Chem. C, Vol. 113, No. 20, 2009 8783 floppysin agreement with the fact that these residues do not make close contact with the surface (Figure 5b). Discussion

Figure 6. Average and fluctuation of the ring-tilt angle (relative to the surface normal) at the peptide-graphite interface for all aromatic residues in (a) the original sequence B1 and the mutated sequences (b) B1Y and (c) B1F. A value of approximately 0° indicates a flat ring orientation, parallel with the surface.

distribution peaks that are markedly less broad than noted for the nanotube interface. B1F also has increased the number of rings in contact with the surface in going from the CNT to graphite, with only H1 not binding to the surface. B1Y also shows unusual behavior; in this case the number of rings in very close contact with the surface has decreased in moving from the CNT to graphite (Vide infra), from three to two. Instead, this system maintains two further rings with moderate surface contact. Again, the corresponding orientational analysis highlights the increased structural stability of the original sequences at the graphite surface. Figure 6 shows the average ring-tilt and corresponding fluctuation in each tilt angle for B1, B1Y, and B1F (see Figure S6, Supporting Information, for B3 and variants). These data are consistent with our distance profiles, showing all rings in B1 with quite flat orientations, while for B1F, only the H1 ring exhibited pronounced deviation from a flat orientation on average. For B1Y (Figure 6b), these data show the end-point regions of the chain are anchored to the surface via the flat orientations of H1 and Y9, while the orientations of the central rings in H4, Y2, and Y6 are significantly more

As pointed out in our previous work,36 we emphasize that our force-field is not intended to reproduce absolute binding energies, but is designed to capture trends in binding. Our forcefield showed an overestimate of binding energies, while yielding correct trends36 compared with first-principles calculations,49 and we therefore suggest that it is likely our binding affinities are an overestimate of the observed binding affinities, while still exhibiting realistic trends in binding affinity upon mutation. We note that according to the experiments of Wang et al.,16 all sequences, including NB1, are observed to show at least some degree of binding to carbon nanotubes. The multipole-based description of the electrostatic interactions and inclusion of the polarizability are considered important in the capturing changes in binding energy as a function of the ring orientation relative to the surface; our previous work demonstrates that interactions based on van der Waals interactions (comparable in quality with a typical “off the shelf” force-field) alone do not compare well with first-principles calculations (most pronounced for both indole and imidazolium rings), whereas our force-field yields a satisfactory performance.36 We therefore believe the expense in employing our force-field is justified for this study, where we believe the structural and energetic characteristics of the interfacial aromatic rings are a key indicator of binding affinity. However, for other peptide-nanotube systems, where aromatic rings are not a critical feature of the binding behavior, “offthe-shelf” force-fields may provide an entirely sufficient description. We have also considered how our findings change with respect to minor modifications of our force-field. While it is not sensible to change multipole moments on atomic sites without recalculating the distributed multipoles on the entire molecular fragment under consideration,50 it is worthwhile to probe how binding changes upon changing polarizability of the carbon atoms in the nanotube. Our test calculations, employing a tube carbon polarizability either lower or higher (2.0 and 5.0 Å3 respectively) than our original value (3.5 Å3) reveals the trends in EN to remain unchangedssee Table S6 in the Supporting Information. Comparison of the results obtained for aptamers interacting with the nanotube and the graphite sheet suggest that B1 and B3 are strong binders for both surfaces, with the tyrosine and phenylalanine mutations yielding slightly less binding affinity. This finding is in general agreement with the ranking in binding affinities reported by experimental studies on similar systems,22,28 underscoring the importance of tryptophan in binding to carbon nanotubes.26,27 In our previous work we concluded that the aromatic content of the sequences is the key factor for the strong binding onto the nanotube. While tyrosine and phenylalanine are also aromatic, these residues are smaller than tryptophan. With two fused rings and a long aspect ratio, the indole group in tryptophan is able to support configurations where all the atoms interact closely with the surface, giving tryptophan a structural stability at the interface. We propose that it is this inherent structural stability that gives rise to the concomitant energetic stability of the tryptophan-containing peptide sequences. Tyrosine and phenylalanine may bind quite strongly on these graphitic surfaces on an atom-for-atom basis, but lack the same degree of stabilization as the indole group. This conclusion is not changed by interfacial shape effects; however, the modes of binding at the graphite interface appear to be different to those seen for the nanotube interface.

8784

J. Phys. Chem. C, Vol. 113, No. 20, 2009

Figure 7. Two snapshots taken from two different strong-binding trajectories (with different starting configurations) of the B1Y peptide interacting with the CNT. The backbone and the tyrosine residues are highlighted.

As in our previous work, our trajectories of B1, B3, and their mutants show that strong binders support different spatial arrangements of the aromatic groups on the nanotube/graphite surface, giving further evidence to support our earlier proposal that peptide affinity can be attributed, in part, to the notion that “strong-binder” aptamers support a variety of strong binding configurations.36 An example is illustrated in Figure 7, showing two different strong-binding configurations taken from different trajectories of the B1Y peptide interacting with the nanotube, emphasizing the different possible strong-binding conformations of the peptide backbone. In the previous section, it was noted that the best trajectory (strongest binding) of the B1Y mutant on the graphite surface actually supported fewer rings in close, flat contact on the surface compared with our nanotube simulations. Inspection of the normalized energies of B1Y versus B1F on the graphite surface (Table S5, Supporting Information) reveals however that the two are very close in binding affinity, despite the fact that B1F maintains a relatively greater number of rings close to the surface. In fact, in some of our trajectories for B1Y, we did find configurations that allowed more rings to approach the surface; however, in these cases, the binding affinity was weaker. We rationalize this apparent discrepancy by examining the remaining (nonaromatic) residues in the chain. In the case of the weaker-bound B1Y configuration that supported more ringsurface interaction (denoted as our second best trajectory), some nonaromatic residues in the peptide were pointing away from the surface and therefore did not contribute to the interaction, diminishing the binding affinity. We illustrate this in Figure 8a, clearly showing the buckling of the peptide backbone in the regions of the chain between the ring positions. In the case of our “best” B1Y trajectory, some of the rings (Y6, Y9) are maintained a medium distance away from the surface, and the peptide backbone is not so buckled, allowing the remaining

Toma´sio and Walsh

Figure 8. (a) Snapshot from the “second-best” trajectory for B1Y on the graphite surface, illustrating the degree of backbone buckling at the interface. (b) Number of peptide atoms within 6 Å of the surface as a function of time for the B1 peptide, for both the peptide-nanotube and peptide-graphite interfaces.

residues to also approach the surface; giving rise to a competitive binding affinity (Table S5, Supporting Information). This behavior was not noted in any of the peptide-nanotube simulations. These data lead us to propose that the nanotube binder sequences have not only been selected to bind strongly to a graphitic surface, but also to bind best at a curVed graphitic surface. As an example, we counted the number of peptide atoms within 6 Å of the surface in the case of our “best” B1 simulations on both the nanotube and graphite; these data are shown for these trajectories in Figure 8b, clearly showing that more peptide atoms are closer to the surface at the peptide-nanotube interface. It may be plausible that the best graphite-binding peptides will contain residues in particular sequences that confer close contact at the surface for most atoms in the chain. Conclusions We have investigated the binding affinity of a range of peptide sequences adsorbed at two graphitic surfaces, graphite and a carbon nanotube, using molecular dynamics simulations. We considered two tryptophan-rich strong-binding sequences, B1 and B3, plus mutations for each of these sequences where all tryptophans were mutated with either tyrosine or phenylalanine. We found that none of the mutants could surpass the original sequences in terms of binding affinity, in agreement with recent experimental observations. We rationalize these data on the basis of the relatively greater structural stability of the indole ring at the interface. While we found that the original sequences showed strongest binding regardless of interfacial shape, our data suggest that aromatic groups may be more abundant in peptides selected for optimal binding to curved graphitic surfaces rather than flat surfaces. Acknowledgment. The authors gratefully acknowledge the computing facilities of the Centre for Scientific Computing, University of Warwick. S.M.T. also thanks the University of Warwick for a graduate student scholarship. This project was

Binding Affinity of Peptides partially resourced by the EPSRC Materials Modelling Consortium Grant (GR/S80127/01) Modelling the Biological Interface with Materials. Supporting Information Available: Measures of system equilibration, comparison of properties calculated using implicit solvent and explicit solvent simulations, normalized binding energies calculated using alternative weighting schemes, averaged normalized interaction energies calculated for each of the four best trajectories for each aptamer for both the peptidenanotube and peptide-graphite interfaces, ring-surface distance distributions and the average ring-tilt and fluctuation for the B3, B3Y, and B3F sequences for both the peptide-nanotube and peptide-graphite interfaces, averaged normalized interaction energies calculated using different values of the tube atom polarizability. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Brown, S. Nat. Biotechnol. 1997, 15, 269. (2) Whaley, S. R.; English, D. S.; Hu, E. L.; Barbara, P. R.; Belcher, A. M. Nature (London) 2000, 405, 665. (3) Dai, H. X.; Choe, W. S.; Thai, C. K.; Sarikaya, M.; Traxler, B. A.; Baneyx, F.; Schwartz, D. T. J. Am. Chem. Soc. 2005, 127, 15637. (4) Zin, M. T.; Munro, A. M.; Gungormus, M.; Wong, N. Y.; Ma, H.; Tamerler, C.; Ginger, D. S.; Sarikaya, M.; Jen, A. K. Y. J. Mater. Chem. 2007, 17, 866. (5) Nam, K. T.; Kim, D. W.; Yoo, P. J.; Chang, C. Y.; Meethong, N.; Hammond, P. T.; Chiang, Y. M.; Belcher, A. M. Science 2006, 312, 885. (6) Sano, K. I.; Yoshii, S.; Yamashita, I.; Shiba, K. Nano Lett. 2007, 7, 3200. (7) Sarikaya, M.; Tamerler, C.; Jen, A. K. Y.; Schulten, K.; Baneyx, F. Nat. Mater. 2003, 2, 577. (8) Willett, R. L.; Baldwin, K. W.; West, K. W.; Pfeiffer, L. N. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 7817. (9) Baneyx, F.; Schwartz, D. T. Curr. Opin. Biotech. 2007, 18, 312. (10) Goede, K.; Busch, P.; Grundmann, M. Nano Lett. 2004, 4, 2115. (11) Peelle, B. R.; Krauland, E. M.; Wittrup, K. D.; Belcher, A. M. Langmuir 2005, 21, 6929. (12) Seker, U. O. S.; Wilson, B.; Dincer, S.; Kim, I. W.; Oren, E. E.; Evans, J. S.; Tamerler, C.; Sarikaya, M. Langmuir 2007, 23, 7895. (13) Oren, E. E.; Tamerler, C.; Sahin, D.; Hnilova, M.; Seker, U. O. S.; Sarikaya, M.; Samudrala, R. Bioinformatics 2007, 23, 2816. (14) Choe, W. S.; Sastry, M. S. R.; Thai, C. K.; Dai, H.; Schwartz, D. T.; Baneyx, F. Langmuir 2007, 23, 11347. (15) Dieckmann, G. R.; Dalton, A. B.; Johnson, P. A.; Razal, J.; Chen, J.; Giordano, G. M.; Munoz, E.; Musselman, I. H.; Baughman, R.; Draper, R. K. J. Am. Chem. Soc. 2003, 125, 1770. (16) Wang, S.; Humphreys, E. S.; Chung, S.-Y.; Delduco, D. F.; Lustig, S. R.; Wang, H.; Parker, K. N.; Rizzo, N. W.; Subramoney, S.; Chiang, Y.-M.; Jagota, A. Nat. Mater. 2003, 2, 196. (17) Su, Z.; Leung, T.; Honek, J. F. J. Phys. Chem. B. 2006, 23263. (18) Brown, S.; Jespersen, T. S.; Nygård, J. Small 2008, 4, 416.

J. Phys. Chem. C, Vol. 113, No. 20, 2009 8785 (19) Morita, Y.; Oshugi, T.; Isawa, Y.; Tamiya, E. J. Mol. Catal. B: Enzym. 2004, 28, 185. (20) Kase, D.; Kulp III, J. L.; Yudasaka, M.; Evans, J. S.; Iijima, S.; Shiba, K. Langmuir 2004, 20, 8393. (21) Prato, M.; Kostarelos, K.; Bianco, A. Acc. Chem. Res. 2008, 41, 60. (22) H.Xie; Becraft, E. J.; Baughman, R. H.; Dalton, A. B.; Dieckmann, G. R. J. Peptide Sci. 2008, 14, 139. (23) Allen, B. L.; Kichambare, P. D.; Star, A. AdV. Mater. 2007, 19, 1439. (24) Dhar, S.; Lin, Z.; Thomale, J.; Dai, H. J.; Lippard, S. J. J. Am. Chem. Soc. 2008, 130, 11467. (25) Zorbas, V.; Smith, A. L.; Xie, H.; Ortiz-Acevedo, A.; Dalton, A. B.; Dieckmann, G. R.; Draper, R. K.; Baughman, R. H.; Musselman, I. H. J. Am. Chem. Soc. 2005, 127, 12323. (26) Li, X.; Chen, W.; Zhan, Q.; Dai, L.; Sowards, L.; Pender, M.; Naik, R. R. J. Phys. Chem. B. 2006, 110, 12621. (27) Su, Z.; Mui, K.; Daub, E.; Leung, T.; Honek, J. J. Phys. Chem. B. 2007, 111, 14411. (28) Salzmann, C. G.; Ward, M. A. H.; Jacobs, R. M. J.; Tobias, G.; Green, M. L. H. J. Phys. Chem. C 2007, 111, 18520. (29) Chen, R. J.; Zang, Y.; Wang, D.; Dai, H. J. Am. Chem. Soc. 2001, 123, 3838. (30) Poenitzsch, V. Z.; Winters, D. C.; Xie, H.; Dieckmann, G. R.; Dalton, A. B.; Musselman, I. H. J. Am. Chem. Soc. 2007, 129, 14724. (31) Evans, J. S.; Samudrala, R.; Walsh, T. R.; Oren, E. E.; Tamerler, C. MRS Bull. 2008, 33, 514. (32) Braun, R.; Sarikaya, M.; Schulten, K. J. Biomater. Sci., Polym. Ed. 2002, 13, 747. (33) Oren, E. E.; Tamerler, C.; Sarikaya, M. Nano Lett. 2005, 5, 415. (34) Kantarci, N.; Tamerler, C.; Sarikaya, M.; Haliloglu, T.; Doruker, P. Polymer 2005, 46, 4307. (35) Thai, C. K.; Dai, H. X.; Sastry, M. S. R.; Sarikaya, M.; Schwartz, D. T.; Baneyx, F. Biotechnol. Bioeng. 2004, 87, 129. (36) Toma´sio, S. D.; Walsh, T. R. Mol. Phys. 2007, 105, 221. (37) Ren, P.; Ponder, J. W. J. Comput. Chem. 2002, 23, 1497. (38) Ren, P.; Ponder, J. W. J. Phys. Chem. B. 2003, 107, 5933. (39) Walsh, T. R. Mol. Phys. 2008, 106, 1613. (40) Stone, A. J. Chem. Phys. Lett. 1981, 83, 233. (41) Stone, A. J.; Alderton, M. Mol. Phys. 1985, 56, 1047. (42) Verlet, L. Phys. ReV. 1967, 159, 98. (43) Ponder, J. W.; Ren, P.; Pappu, R. V.; Hart, R. K.; Hodgson, M. E.; Cistola, D. P.; Kundrot, C. E.; Richards, F. M. TINKERssoftware tools for molecular design, Version 4.2; Washington University School of Medicine: St. Louis, MO, 2004. (44) Eisenberg, D.; McLachlan, A. D. Nature (London) 1986, 319, 199. (45) Wesson, L.; Eisenberg, D. Protein Sci. 1992, 1, 227. (46) Walther, J. H.; Jaffe, R.; Halicioglu, T.; Koumoutsakos, P. J. Phys. Chem. B. 2001, 105, 9980. (47) Skelton, A. A.; Walsh, T. R. Mol. Simul. 2007, 33, 379. (48) Skelton, A. A.; Liang, T.; Walsh, T. R., Interplay of sequence, structure and binding at the peptide-inorganic interface as mediated by water. Manuscript in preparation. (49) Walsh, T. R. Phys. Chem. Chem. Phys. 2005, 7, 443. (50) Stone, A. J., The Theory of Intermolecular Forces; OUP, Clarendon Press: Oxford, U.K., 1997.

JP8087594