Self-Assembly of Cyclo-diphenylalanine Peptides in Vacuum

May 30, 2014 - ABSTRACT: The diphenylalanine (FF) peptide self-assembles into a variety of nanostructures, including hollow nanotubes that form in aqu...
0 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/JPCB

Self-Assembly of Cyclo-diphenylalanine Peptides in Vacuum Joohyun Jeon and M. Scott Shell* Department of Chemical Engineering, University of California Santa Barbara, Santa Barbara, California 93106-5080, United States S Supporting Information *

ABSTRACT: The diphenylalanine (FF) peptide self-assembles into a variety of nanostructures, including hollow nanotubes that form in aqueous solution with an unusually high degree of hydrophilic surface area. In contrast, diphenylalanine can also be vapor-deposited in vacuum to produce rodlike assemblies that are extremely hydrophobic; in this process FF has been found to dehydrate and cyclize to cyclodiphenylalanine (cyclo-FF). An earlier study used all-atom molecular dynamics (MD) simulations to understand the early stages of the self-assembly of linear-FF peptides in solution. Here, we examine the self-assembly of cyclo-FF peptides in vacuum and compare it to these previous results to understand the differences underlying the two cases. Using all-atom replica exchange MD simulations, we consider systems of 50 cyclo-FF peptides and examine free energies along various structural association coordinates. We find that cyclo-FF peptides form ladder-like structures connected by double hydrogen bonds, and that multiple such ladders linearly align in a cooperative manner to form larger-scale, elongated assemblies. Unlike linear-FFs which mainly assemble through the interplay between hydrophobic and hydrophilic interactions, the assembly of cyclo-FFs in vacuum is primarily driven by electrostatic interactions along the backbone that induce alignment at long-range, followed by van der Waals interactions between side chains that become important for close-range packing. While both solution and vacuum phase driving forces result in ladder-like structures, the clustering of ladders is opposite: linear-FF peptide ladders form assemblies with side-chains buried inward, while cyclo-FF ladders point outward.

■

INTRODUCTION The diphenylalanine (FF) peptide has emerged as a rather remarkable self-assembling building block for bionano materials.1,2 Its self-assembly properties were originally discovered by Reches and Gazit3 in searching for a minimal model for the aggregation of the amyloid beta protein. Namely, they found that FF peptides can form nanotubes in aqueous solution at room temperature. In turn, these nanotubes offer a viable template for synthesizing inorganic nanomaterials such as silver nanowires. FF nanotubes have since been recognized to have exceptional properties for a biomaterial assembled from noncovalent interactions, with extraordinarily mechanical strength4 and stability to temperature and a variety of solvents.5,6 Moreover, it has been found that materials made from FF peptides under different conditions or with targeted mutations or chemical modifications can be fabricated easily and self-assemble into a variety of nanostructures, including nanowires,7 nanoplates8 and nanofibrils.9 Such materials have received recent interest in biomedical applications for drug delivery microtubes,10 hydrogels9 and biosensors.11 Quite interestingly, while many of the first studies of nanotube formation were performed in solution, AdlerAbramovich et al. found that FF peptides also self-assemble when deposited on a surface by physical vapor deposition, in this case forming large arrays of linearly aligned nanotubes called nanoforests.12 By varying the deposition time, the method was found to allow controllable fabrication of FF nanoforests of given thickness, offering great opportunity to produce interfaces of very high surface area due to decoration © 2014 American Chemical Society

with dense and homogeneous layers of FF nanotubes. It was later found that the nanoforests were not composed entirely of FF nanotubes of the type formed in solution, which accounted for only about 10−20% of the forest; the remaining 80−90% of the cylindrical assemblies instead appeared to be nanorods composed of a modified FF peptide.13 Namely, the linear lyophilized FF peptides were found to dehydrate during the vapor deposition process and form cyclo-FF peptides that possess a backbone ring with two aromatic side-chains (Figure 1). It is these cyclo-FF peptides that self-assemble into the nanorods found in the forests, and unlike the nanotubes, they are solid, lack hydrophilic channels, and are very hydrophobic. Specifically, the contact angle of surfaces functionalized with cyclo-FF nanorods is around 130°.13 Consistent with these

Figure 1. Dehydration of the diphenylalanine (FF) peptide. The linear backbone of FF cyclizes and the peptide becomes cyclo-FF. Received: February 11, 2014 Revised: May 29, 2014 Published: May 30, 2014 6644

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

peptides has also been studied in dichloromethane solution by Govindaraju21 and in 50% aqueous methanol solution by Joshi and Verma,22 and in that case it was also suggested that intermolecular hydrogen bonds form cyclo-FF molecular chains and that aromatic pi−pi interactions further stabilize the assembly of these chains into 3D nanostructures such as nanofibers and nanorods. Simulations offer useful ways to understand the molecular mechanisms underlying the self-assembly of FF peptides and the relative roles of different interaction forces in generating their unique structures and stabilities. However, limited studies appear to have been performed so far. A few studies of FF selfassembly in water used simple lattice23 or coarse-grained24,25 models, and the molecular structures found from these bear some agreement with the experimental FF nanotube crystal structure. This includes the existence of an interior central channel and the intermolecular association and packing of the aromatic side chains. However, it is still unclear whether such simple models can generate the unique, hierarchical nanotube molecular structure containing thousands of small hydrophilic pores inside the tube walls. On the other hand, Tamamis et al.26 examined the association of atomically detailed FF peptides using an implicit water model, and observed the transient formation of both open and ringlike FF peptides networks and network-containing structures with β-sheet characteristics. They suggested that the initial formation of FF aggregates could be promoted by both hydrophobic side-chain and backbone hydrogen-bonding interactions. More recently, Sayar et al.27 developed a new coarse-grained model for FF that aims to reproduce the trans- to cis- conformation change upon aggregation or in a cyclohexane and water biphasic system. They found that the conformational behavior of the FF is affected by the cyclohexane/water interaction potential, and the performance of the CG model depends on the delicate balance between the bonded and nonbonded interactions. We previously studied FF self-assembly using molecular simulations with atomically detailed and explicit solvent models, and found evidence for early mechanisms of oligomer self-assembly.28 Namely, our simulations showed that small FF oligomers leverage cooperative interactions between polar and hydrophobic forces: charged groups and hydrogen bonding facilitate the formation of peptide ladders with a single face that presents hydrophobic side-chains. These ladders can zip together and associate through hydrophobic and aromatic interactions to form larger-scale, amphiphilic structures. These simulations suggested that regions in between the nanoscopic hydrophilic channels in the FF nanotube structure might be the most relevant to the initial self-assembly pathway in solution. In the present study, we investigate the self-assembly of cyclo-FF peptides in vacuum to understand the formation of nanorods in the vapor deposition experimental procedure. A major question is by what mechanism do these peptides assemble when hydrophobic interactions are not present, and to what extent are there similarities with the FF nanotube assembly process in solution? We find that like FF in solution, cyclo-FF peptides form ladder-like structures; however, in this case the ladders are stabilized by double hydrogen bonds, and they assemble with each other and self-align through electrostatic interactions between backbone rings and van der Waals interactions between side chains, rather than hydrophobic forces. This alignment of the quasi-one-dimensional linear cyclo-FF peptide ladders is likely to be a key step for further assembly into larger nanostructures. To our knowledge, this is

results, it was found that the presence of water vapor during deposition increases the fraction of nanotubes versus nanorods in the nanoforest.14 From an applications perspective, this initial work by AdlerAbramovich et al. found that FF nanoforests could be used to coat carbon electrodes in model supercapacitors and significantly increase their performance, even exceeding similar strategies in which the electrodes were coated with carbon nanotubes. 12 This exceptional increase in double-layer capacitance was suggested to be due to the special atomic structure of the FF nanotubes as assessed by X-ray characterization.15 Specifically, an FF nanotube contains numerous ∼10 Å diameter hydrophilic channels within the walls of the tube itself and aligned with the tube axis. These nanoscale hydrophilic surfaces dramatically increase the “functional” surface area of carbon electrodes when decorated with FF nanoforests.16 While the hydrophobic cyclo-FF nanorods clearly do not contribute to ion transfer in the nanoforestdecorated carbon electrodes in these supercapacitors, their hydrophobicity has found great utility in other applications. For example, surfaces functionalized with nanoforests of FF nanorods can be used to pattern superhydrophobic domains for directing water channels in microfluidic devices12 and for creating superhydrophobic, self-cleaning surfaces.17 Because cyclo-FF peptide nanorods can be densely coated on many surfaces by physical vapor deposition, they appear to offer great possibilities in patterning 2D surfaces with superhydrophobic and biocompatible moieties. The crystal structure of bulk cyclo-FF crystals obtained from X-ray diffraction data18,19 shows that the peptides are linearly connected in ladder-like configurations by double hydrogen bonds along the z-direction (Figure 2). Cyclo-FF side-chain rings are tightly packed in both T-aligned and parallel forms which are common configurations of aromatic ring stacking.20 It is worthwhile to note that the self-assembly of cyclo-FF

Figure 2. Crystal structure of cyclo-FF nanowires. The unit cell is duplicated in the x and z directions to illustrate side chain packing and hydrogen bonds. (A) Aromatic side chains are closely packed by parallel and triangular stacking. (B) Cyclo-FF peptides are linearly connected by double hydrogen bonds. Side chains are omitted. 6645

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

the first simulation study of cyclo-diphenylalanine self-assembly in vacuum, and it may provide useful insights for understanding and/or designing these and related materials in the future.

■

METHODS We use replica exchange molecular dynamics (REMD) simulations29 to understand the association of 50 cyclo-FF peptides in vacuum. The REMD method facilitates the surmounting of energy barriers through a coupling of different simulation replicas over a range of temperatures. This is particularly important for the present case under vacuum conditions, in which the interactions between different peptides are extremely strong due to the absence of screening through a solvent. We employ the AMBER12 simulation package and the Generalized Amber Force Field (GAFF),30 building force field parameters for a cyclo-FF peptide with the Antechamber tool.31,32 Atom charges are reported in Table S1 in the Supporting Information. To perform simulations in vacuum, we use the igb=6 option in AMBER. No boundary conditions are applied, but all 50 cyclo-FF peptides are constrained to lie within a spherical region of radius 60 Å (27 mg/mL) using flatbottom harmonic restraints. REMD simulations are performed for 130 ns and the first 30 ns is discarded from the analysis as an equilibration time. Temperatures span from 300 K from 570 K, the latter of which roughly corresponds to the experimentally identified melting temperature of FF nanotubes.6 The average acceptance ratios between replicas are at least 27%. The convergence of the REMD simulations is tested in Figure S1 in the Supporting Information. Although Figure S1 shows that this system is not at equilibrium yet and evolves in time, the changes are relatively small for the last 50 ns. Still, it may be possible that further structural evolution may occur on longer time scales, beyond what is practical for the REMD runs here. We apply a k-means algorithm33,34 to cluster the structures of the peptides into common configurational motifs and to assess the populations thereof. We also perform umbrella-sampling replica exchange molecular dynamics (UREMD)35 to study in detail the interaction between two cyclo-FF tetramers. UREMD aids in determining a high accuracy potential of mean force (the free energy) along a given reaction coordinate, which in the present case corresponds to the distance between the centers of mass of two associating groups. UREMD involves a cascade of replicas that differ in an umbrella distance for the reaction coordinate; at selected replicas where particularly large free energy barriers are expected, a temperature cascade is built as well (Figure 3). Details of the approach are given in ref 35. To construct a tetramer group, we tether four cyclo-FF peptides together using flat-bottom harmonic constraints with a parabolic side that takes effect at the peptide−peptide distance of 3.5 Å (force constant 3.0 kcal/mol/Å2). Subsequently, two such groups are used as the associating species in the UREMD method. Specifically, an umbrella restraint with a force constant of 1.0 kcal/mol/Å2 is added between the center of mass of a cyclo-FF peptide from each of the tetramers. The restraint is also a flatbottom harmonic form and is wide enough to allow sampling a range of reaction coordinate spanning the previous and following umbrella distances. In the UREMD simulations, a total of 49 umbrella distances spanning from 2 to 60 Å are used; these are more closely spaced where energy barriers are expected to exist, that is, in the associated state. In addition, four temperature replica

Figure 3. A schematic of the two-dimensional replica cascade used in the UREMD simulation.

cascades are added at the umbrella distances 4, 8, 13, and 60 Å to facilitate sampling over energetic barriers; each temperature cascade contains 12 replicas at the same umbrella distance (Figure 3). The total number of replicas is 93. Postsimulation we use the weighted histogram analysis method (WHAM)36−38 to reweight data from all replicas and to obtain the potential of mean force along the reaction coordinate. In addition, simple energy minimization simulations are performed for stable conformations found from the UREMD simulation to analyze energetic contributions to single and double cyclo-FF tetramers. In these, 5000 steps of conjugate gradient minimization are performed after 10 000 steps of steepest descent minimization.

■

RESULTS Cyclo-FF Peptides Form Ladder-Like Structures Connected by Double Hydrogen Bonds. We first consider the results of the REMD simulation of 50 cyclo-FF peptides, taken from the lowest temperature replica at 300 K. At this condition, the peptides form strongly ordered structures, clustering into well-defined oligomeric species that become partially aligned. Figure 4 shows the clustered conformation with the highest percentage from the simulation (10%). The peptides mainly associate on the basis of double hydrogen bonds between two central backbone ringsthrough their backbone carbonyl and amine groupsand these doubly connected peptide oligomers form linear molecular chains that resemble “ladders.” This hydrogen bonding pattern closely resembles that of the cyclo-FF crystal structure (Figure 2), although the ladders themselves are not well aligned with respect to each other in this conformation. To be precise, we define a ladder as a set of cyclo-FF peptides that are connected linearly by double hydrogen bonds in the backbone rings, using a N−O cutoff distance of 3.5 Å and an angle criterion (H−N··· O angle 3) quite often, but these ladders seem to be transient with lifetimes of the order of a few ns. The maximum length of a cyclo-FF ladder observed in the simulation is 13 and the average length of all ladders for the final 100 ns is 2.8 ± 0.4. Cyclo-FF Ladders Self-Align As They Grow. For cycloFF ladders to ultimately assemble into a structure reminiscent of nanorods and of the X-ray structures, they must align into an 6646

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

question is whether or not association and growth into larger ladders leads to a concomitant increase in the degree of orientational alignment between distinct ladders. To measure the alignment, we define a ladder angle between each pair of cyclo-FF ladders that have lengths longer than two, where the angle is calculated from the vectors connecting the center of masses of the first and last peptides in each ladder. The ladder angle is calculated for all pairs of ladders in each snapshot and then averaged. The time-dependent angles at three different temperatures are shown in Figure 5. Interestingly, the ladder angle seems somewhat anticorrelated with the length. That is, the average ladder angle decreases when the average ladder length increases, as is particularly evident in the progression of the simulation from during the times 30−70 ns and from 120 to 130 ns. The average angle throughout the last 100 ns is 50 ± 8°. A more quantitative relationship between ladder length and angle is shown in Figure 6a, which presents a two-dimensional

Figure 4. Most probable clustered conformation (10%) from an REMD simulation of 50 cyclo-FF peptides in vacuum. Hydrogen bonds are shown as dotted red lines, and only doubly connected cycloFF peptides are shown. The system forms a number of “ladders” of continuously connected peptides through backbone hydrogen bonds. These ladders are colored differently for the sake of illustration. Side chains are omitted from this diagram, although the inset shows the general structure of the double hydrogen bond motif and the side chains.

Figure 6. Two dimensional free energy surface (F = −kBT ln Prob(l, θ)) with respect to average cyclo-FF ladder length (l) and the average angle between cyclo-FF ladders (θ) at different temperatures. The length is defined as the number of peptides composing a ladder, while the angle is defined as that between two ladders. Values of length and angle for all ladders in each snapshot are averaged to create this representation. Around 460 K, the well-aligned state involving long ladders begins to vanish. F has units of kcal/mol

free energy surface as a function of these two variables. While admittedly not wholly distinct, it is possible to identify two states in this free energy representation. The most probable state (A, 98%) is a dispersed, unaligned ensemble in which the peptides form only small oligomers, that is, ladders with a small average length (30). On the other hand, a second state (B, 2%) has a much smaller population but corresponds to ordered configurations in which ladders have large lengths (≥3.5) and small alignment angles (≤30). The modest correlation between ladder length and alignment in the free energy surface suggests that ladders may self-align as they grow during early stages of self-assembly. Moreover, the existence of a low-population between regions A and B, and hence an associated free energy barrier, may be a signature of cooperative effects in the simultaneously alignment

Figure 5. Average cyclo-FF ladder lengths and angles versus time during the REMD simulation. Each point represents an average over 10 ps. There are significant fluctuations in both quantities that are often anticorrelated. Fluctuations increase at 352 K, while at 512 K, most peptides are dissociated.

ordered 3D nanostructure. In many of the conformations observed in the REMD simulations, multiple ladders are present but adopt mutual orientations that are not parallel. One 6647

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

surface as the temperature is varied, as shown in Figure 6. As expected, the population of state B shrinks as the temperature increases, nearly disappearing at 512 K (Figure 6d). It is interesting to note that this temperature is slightly lower than 575 K, the melting temperature of bulk cyclo-FF crystals.7,40 Figure S2 in the Supporting Information shows that several other metrics reveal significant changes in the system in the range 450−500 K, including the number of hydrogen bonds and the average radius of gyration of all 50 peptides. Of course, a fair comparison of the simulation to experiment is difficult due to the length and time-scale limitations of the calculations. Interestingly, one might expect that state B should be increasingly prominent at lower temperatures. However, Figure 8 shows that the percent of B is highest near 80 °C (350−400

and growth of ladders. Importantly, conformations in state B (Figure 7) show an alignment that would be conducive to

Figure 8. Temperature-dependence of the ensemble fraction of state B. Each point is averaged from four 25 ns time-blocks during the final 100 ns of the simulation. The percent of state B is highest near 350−400 K and monotonically decreases after 400 K.

K, 5%) and, while it vanishes near 512 K as expected, it also shrinks at the lowest temperatures. Given the strength and directionality of the peptide−peptide interactions in vacuum, a significant kinetic barrier between state A and state B may exist that makes conformational rearrangements difficult near room temperature. This would be consistent with the existence of a free energy barrier separating these two states. Perhaps coincidentally, the experiments showing the formation of cyclo-FF nanoforests on surfaces was performed at 80 °C, the same temperature as the maximum-B state found here.12 How Do Cyclo-FF Ladders Grow and Self-Align? The free energy surfaces shown in Figure 6 do not specifically delineate the driving forces that induce growth and selfalignment of cyclo-FF ladders. To better understand this, we consider the simpler process of the association of two tetramers. We first determine the potential of mean force and average energy for the association of two cyclo-FF tetramers using the UREMD technique. The reaction coordinate is chosen as the center of mass distance between two internal peptides that are located in the center of each cyclo-FF ladder (tetramer). Figure 9 shows that as soon as two peptides from each tetramer are within interaction range (corresponding to 26 Å for this particular reaction coordinate), they rapidly form a double ladder of length four. As before, this structure corresponds to a state in which the backbone rings are in close contact between the individual ladders, with the aromatic side chains pointing outward. The stability of this initial structure is already rather large (∼15 kcal/mol), but with closer approach the free energy continues to decrease until the most stable structure is reached around 4 Å separation distance with a stability of almost 45 kcal/mol. During this process, the

Figure 7. (A) One clustered conformation (population 0.5%) from the REMD simulation showing alignment of multiple, long peptide ladders. Hydrogen bonds are shown as dotted red lines and ladders are colored separately for illustration. Cyclo-FF peptides not participating in ladders are colored green. Side chains are omitted for clarity. (B) Top-view of orange, purple, and magenta ladders. Side-chains are shown explicitly.

forming nanowires, and perhaps ultimately, nanorods. While the simulations here are limited in scale due to computational expense, it is likely that with larger numbers of peptides and longer times, a shift to even more structured nanowire motifs might emerge. Using the aggregate simulation data from the entire REMD replica cascade, we also examine the behavior of this free energy 6648

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

Figure 10. Average van der Waals energy (Evdw), electrostatic energy (Eelec) and total energy (Etot) with respect to the average distance between ladders. Each energy term and distance are averaged at each replica from the same UREMD simulation used to generate Figure 9. The energies shown here reflect total energies due to both intra- and interladder interactions, and are shifted so as to approach zero at large distances.

Figure 9. Potential of mean force and average energy with respect to the distance between the centers of mass (COM) of two peptides (cyan and pink peptides that are located in the center of each cyclo-FF ladder) in two cyclo-FF tetramers (green and orange). Peptides within each ladder are mildly constrained to remain associated as a ladder as described in the text. At far distances, the tetramers are noninteracting and each bends so as to form a small double ladder of length two. At close distances, the tetramers from one large double ladder, becoming increasingly stable until the free energy minimum near 4 Å. In the bottom panel, the total energy reflects intra- and interladder interactions, and is shifted so as to approach zero at large distances.

between two cyclo-FF tetramers. These hydrogen bonds linearly connect one cyclo-FF peptide in one tetramer to a neighboring cyclo-FF peptide in the other. There is a second large drop of Eelec from 26 to22 Å and at this stage, two more hydrogen bonds are formed, resulting in each tetramer elongating as a double ladder of length four. From 22 to7 Å, there are several electrostatic energy barriers (involving both intra- and intertetramer interactions) due to the hydrogen-bond attachment and detachment during configurational rearrangement. Finally, the steepest drop of Eelec occurs between 6 and 4 Å as the cyclo-FF double ladder finds the optimal alignment that minimizes electrostatic interactions between all carbonyl and amine groups in eight cyclo-FF peptides. On the other hand, the van der Waals energies show a more gradual evolution during the association of two ladders, with less pronounced fluctuations. Evdw decreases slowly from 30 to 22 Å as the side chains from each tetramer begin to interact. From 22 to4 Å as the cyclo-FF ladder structures rearrange, there are additional 10 kcal/mol fluctuations in Evdw. As a second means of investigating the various energetic contributions, we examine energy-minimized configurations of both double and single ladders. From the UREMD simulation we pick four distinct, top-clustered double ladder conformations from the replica located near energy minima (4 and 12 Å). We also create single-ladder conformations by extracting one ladder from each of these double ladder structures. The energy-minimized configurations are shown in Figure 11. The contribution of the potential energy to the formation of a double ladder is estimated from the minimized energies as ΔU = Udouble − 2Usingle. We find overall that ΔU = −64.6 ± 2.1 kcal/mol, and that the van der Waals component of this energy is −40.3 ± 7.1 kcal/mol while the electrostatic part is −25.8 ± 3.5 kcal/mol. Unlike the results in Figure 10 the electrostatic contribution is smaller than that of van der Waals interactions, but it is still significant. The difference is almost certainly due to the fact that the energy-minimized configurations neglect fluctuations; thermalized structures are likely to have less efficient molecular packing that has a strong effect on the shortranged van der Waals energies. Another difference stems from the fact that the dissociated state corresponding to Figure 10 does not involve completely elongated ladders but more

peptides in the double ladder optimize their interactions mainly by shifting register along backbone. Indeed, there appear to be a significant number of locally stable double ladder structures throughout the association reaction coordinate. It seems telling that there are noticeable free energy differences along the reaction coordinate even for structures that appear quite similar. For example, the locally stable doubleladder conformations near 4 and 26 Å separation distance have a root-mean-square distance (RMSD) of 2.6 Å and yet differ by almost 30 kcal/mol. This suggests that there are strong and directional interactions between cyclo-FF backbone rings that require very specific alignments to optimize the free energy. There are also a few free energy barriers along the reaction coordinate, typically around 5 kcal/mol, which may suggest kinetic constraints for conformational rearrangements. Of course, these barriers depend on the particular choice of reaction coordinate; the one used here is not likely to be the precise kinetically relevant one, but qualitatively the analysis remains informative. The energy of hydrogen bond formation in vacuum is on the order of 6−7 kcal/mol,41 which may at least partially suggest an origin of these barriers. Interestingly, Figure 9 also shows that the average potential energies of the different ladder configurations are quite varied. Indeed, extremely large potential energy barriers of up to ∼40 kcal/ mol exist along this particular reaction coordinate. To understand the energetic contributions to the free energy, Figure 10 shows the thermally averaged van der Waals energy (Evdw), electrostatic energy (Eelec) and total energy (Etot) as a function of ladder separation distance. We find that the electrostatic interactions play an important role in cyclo-FF ladder growth and the alignment of the double ladder conformation. Upon approach from far distances, the ladders quickly align at 30 Å due to steering by a sharp decrease in the electrostatic interactions. That is, there is a steep drop of Eelec from 32 to26 Å due to the formation of two hydrogen bonds 6649

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

rings. These ladders then begin to associate on the basis of two factors, one of which involves long-ranged backbone−backbone interactions as in the formation of double ladders, but a second is that ladders can attach through favorable van der Waals interactions between side chains. As the structures grow beyond the size of the systems studied here, it is likely that the packing of side chains among ladders becomes increasingly important and drives the configurations into more ordered and regular structures. Linear-FF Peptides in Water and Cyclo-FF Peptides in Vacuum. These results show the interplay between electrostatic and van der Waals interactions that allow cyclo-FF peptides to align in vacuum, form ladder structures, and ultimately assemble into larger nanostructures. By comparison, our previous study of linear-FF peptides in water28 showed the relevance of the interplay between hydrophobic and hydrophilic (electrostatic) interactions to alignment, ladder-formation, and assembly. While both systems form ladders at a small scale, the distinct balance of interactions leads to an opposite effect when multiple ladders associate at a slightly larger scale: linear-FF ladders in solution pack side chains toward the interior of multiple-ladder structures, while cyclo-FF ladders in vacuum align along the backbone and point side-chains outward. To compare and quantify these behaviors, we calculate atom−atom contact probabilities from the simulations using a cutoff length of 6 Å. Figure 13 shows that the interactions

Figure 11. Energy-minimized structures of a representative single ladder (A) and double ladder (B).

globular tetramers since there are no explicit restraints to guarantee this. From the energy-minimized configurations, it is apparent that van der Waals interactions can be significant in stabilizing a double ladder composed from two single ladders. The REMD oligomer simulations described earlier also suggest a strong role for van der Waals interactions, particularly between the sidechains that extend out from single and double ladders. Namely, we observe structures in which a number of short cyclo-FF ladders are linearly aligned and closely packed such that their side chains fit together like puzzle pieces, as illustrated in Figures 7B and 12. These illustrations show that cyclo-FF

Figure 13. Atom−atom contact maps of linear-FF and cyclo-FF peptides. Contacts between every heavy atom (cutoff distance = 6 Å) are calculated, but only nitrogen and oxygen atoms in the peptide backbone and one carbon atom in each aromatic side-chain are shown in the axis labels. The contact maps show relative contact probabilities: all contacts for each atom−atom pair are divided by the total number of contacts.

Figure 12. (A) Linearly aligned and closely packed ladders seen in the REMD simulation (5%). For clarity, each ladder is colored differently and side-chains are omitted. (B) Top-view with side-chains shown explicitly to illustrate their packing.

ladders can effectively cluster in vacuum without hydrophobic interactions. Unfortunately, we only observe cyclo-FF ladders that are somewhat irregularly assembled in our REMD simulation when compared to the crystal structure in Figure 2A that shows very specific triangular and parallel aromatic ring packing with interlocking side-chains. Supporting Information, Figure S3 reveals a broader distribution of the orientations of aromatic rings in the simulations, relative to that of the crystal structure. The structural differences may be due to limitations in simulation time scale and system size, or to the use of a classical force field.42 We leave an investigation of more sophisticated force fields to a future study. Taken together, the results from the simulations in this study suggest that the large-scale structure of cyclo-FF aggregates may follow a semihierarchical mechanism: multiple peptides become linearly aligned via double hydrogen bonds between backbone

between nitrogen and oxygen atoms in the peptide backbone are similarly strong for both systems of linear-FF and cyclo-FF peptides. This is clearly the case since both linear- and cyclo-FF ladders involve backbone hydrogen bonds along their lengths. In contrast, the interactions between side chains are much more prominent for the case of linear-FF peptides due to hydrophobic interactions. When water molecules are present, the aromatic side chains of these peptides form a hydrophobic cluster, with the hydrophilic backbones along the exterior (Supporting Information, Figure S4B). We proposed that such clusters of ladders might gather to form the honeycomb-like arrangement of small hydrophilic pores that is seen experimentally (Figure S4A). In vacuum, the electrostatic interactions between backbones become the most significant driving forces and lead to outward-facing aromatic side chains. 6650

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

■

This basic propensity may ultimately lead to larger rodlike structures that bury polar backbone moieties while exposing large surface areas due to aromatic groups. Such behaviors might explain why cyclo-FF peptide nanostructures are very compact and hydrophobic, whereas linear-FF nanotubes are less compact and hydrophobic.

CONCLUSIONS We used replica exchange simulations to study the early selfassembly of the cyclo-diphenylalanine peptide in vacuum. This system has been found in experiment to form elongated nanorod structures that are quite hydrophobic and that have a variety of emerging technological applications. Our modestscale simulations reveal that cyclo-FF peptides form long “ladder-like” structures during early stages of self-assembly, largely due to the formation of double hydrogen bonds between the (cyclized) backbones of peptide pairs. As these ladders grow, pairs, triplets, and other oligomers of them can then associate to form higher levels of structure. Importantly, the association and mutual alignment of multiple ladders seems highly correlated with their growth (i.e., increase in length) and may be cooperative. A free energy analysis shows the emergence of a distinct stable state of associated, long-length ladders at low temperatures that disappears at higher temperatures near the melting temperature of bulk cyclo-FF crystals. Finally, we show that electrostatic, van der Waals, and side-chain packing interactions contribute significantly to the formation of ordered structures and are likely to direct assembly in a hierarchical way. These insights may provide molecular interpretations of the early assembly physics driving the formation of cyclo-FF nanorods in experiments. ASSOCIATED CONTENT

S Supporting Information *

Temperature dependence of the radius of gyration and the number of hydrogen bonds for cyclo-FF peptides; probability distributions of cyclo-FF ladder length, ladder angle, intrapeptide dihedral angle, and interpeptide angle; crystal structure of linear-FF peptide nanotubes and hexamer of linear-FF peptides; and force field parameters of cyclo-FF. This material is available free of charge via the Internet at http://pubs.acs.org.

■

REFERENCES

(1) Yan, X.; Zhu, P.; Li, J. Self-Assembly and Application of Diphenylalanine-Based Nanostructures. Chem. Soc. Rev. 2010, 39, 1877−1890. (2) Rosenman, G.; Beker, P.; Koren, I.; Yevnin, M.; Bank-Srour, B.; Mishina, E.; Semin, S. Bioinspired Peptide Nanotubes: Deposition Technology, Basic Physics and Nanotechnology Applications. J. Pept. Sci. 2011, 17, 75−87. (3) Reches, M.; Gazit, E. Casting Metal Nanowires within Discrete Self-Assembled Peptide Nanotubes. Science 2003, 300, 625−627. (4) Kol, N.; Adler-Abramovich, L.; Barlam, D.; Shneck, R. Z.; Gazit, E.; Rousso, I. Self-Assembled Peptide Nanotubes Are Uniquely Rigid Bioinspired Supramolecular Structures. Nano Lett. 2005, 5, 1343− 1346. (5) Sedman, V. L.; Adler-Abramovich, L.; Allen, S.; Gazit, E.; Tendler, S. J. B. Direct Observation of the Release of Phenylalanine from Diphenylalanine Nanotubes. J. Am. Chem. Soc. 2006, 128, 6903− 6908. (6) Adler-Abramovich, L.; Reches, M.; Sedman, V. L.; Allen, S.; Tendler, S. J. B.; Gazit, E. Thermal and Chemical Stability of Diphenylalanine Peptide Nanotubes: Implications for Nanotechnological Applications. Langmuir 2006, 22, 1313−1320. (7) Amdursky, N.; Beker, P.; Koren, I.; Bank-Srour, B.; Mishina, E.; Semin, S.; Rasing, T.; Rosenberg, Y.; Barkay, Z.; Gazit, E.; et al. Structural Transition in Peptide Nanotubes. Biomacromolecules 2011, 12, 1349−1354. (8) Govindaraju, T.; Pandeeswar, M.; Jayaramulu, K.; Jaipuria, G.; Atreya, H. S. Spontaneous Self-Assembly of Designed Cyclic Dipeptide (Phg-Phg) into Two-Dimensional Nano- and Mesosheets. Supramol. Chem. 2011, 23, 487−492. (9) Mahler, A.; Reches, M.; Rechter, M.; Cohen, S.; Gazit, E. Rigid, Self-Assembled Hydrogel Composed of a Modified Aromatic Dipeptide. Adv. Mater. 2006, 18, 1365−1370. (10) Silva, R. F.; Araújo, D. R.; Silva, E. R.; Ando, R. A.; Alves, W. A. L-Diphenylalanine Microtubes as a Potential Drug-Delivery System: Characterization, Release Kinetics, and Cytotoxicity. Langmuir 2013, 29, 10205−10212. (11) Yemini, M.; Reches, M.; Gazit, E.; Rishpon, J. Peptide Nanotube-Modified Electrodes for Enzyme−Biosensor Applications. Anal. Chem. 2005, 77, 5155−5159. (12) Adler-Abramovich, L.; Aronov, D.; Beker, P.; Yevnin, M.; Stempler, S.; Buzhansky, L.; Rosenman, G.; Gazit, E. Self-Assembled Arrays of Peptide Nanotubes by Vapour Deposition. Nat. Nanotechnol. 2009, 4, 849−854. (13) Beker, P.; Rosenman, G. Bioinspired Nanostructural Peptide Materials for Supercapacitor Electrodes. J. Mater. Res. 2010, 25, 1661− 1666. (14) Bank-Srour, B.; Becker, P.; Krasovitsky, L.; Gladkikh, A.; Rosenberg, Y.; Barkay, Z.; Rosenman, G. Physical Vapor Deposition of Peptide Nanostructures. Polym. J. 2013, 45, 494−503. (15) Gorbitz, C. H. The Structure of Nanotubes Formed by Diphenylalanine, the Core Recognition Motif of Alzheimer’s betaAmyloid Polypeptide. Chem. Commun. 2006, 2332. (16) Beker, P.; Koren, I.; Amdursky, N.; Gazit, E.; Rosenman, G. Bioinspired Peptide Nanotubes as Supercapacitor Electrodes. J. Mater. Sci. 2010, 45, 6374−6378. (17) Lee, J. S.; Ryu, J.; Park, C. B. Bio-Inspired Fabrication of Superhydrophobic Surfaces through Peptide Self-Assembly. Soft Matter 2009, 5, 2717−2720. (18) Gdaniec, M.; Liberek, B. Structure of Cyclo(-L-Phenylalanyl-LPhenylalanyl-). Acta Crystallogr. C 1986, 42, 1343−1345. (19) Lee, J. S.; Yoon, I.; Kim, J.; Ihee, H.; Kim, B.; Park, C. B. SelfAssembly of Semiconducting Photoluminescent Peptide Nanowires in the Vapor Phase. Angew. Chem., Int. Ed. 2011, 50, 1164−1167. (20) Sinnokrot, M. O.; Sherrill, C. D. High-Accuracy Quantum Mechanical Studies of π−π Interactions in Benzene Dimers. J. Phys. Chem. A 2006, 110, 10656−10668.

■

■

Article

AUTHOR INFORMATION

Corresponding Author

*Tel: +1 805 893 4346. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors greatly appreciate the support of the Alfred P. Sloan Foundation.

■

ABBREVIATIONS FF, diphenylalanine; cyclo-FF, cyclo-diphenylalanine; MD, molecular dynamics; REMD, replica-exchange molecular dynamics; UREMD, umbrella sampling replica-exchange molecular dynamics; WHAM, weighted histogram analysis method 6651

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652

The Journal of Physical Chemistry B

Article

(21) Govindaraju, T. Spontaneous Self-Assembly of Aromatic Cyclic Dipeptide into Fibre Bundles with High Thermal Stability and Propensity for Gelation. Supramol. Chem. 2011, 23, 759−767. (22) Joshi, K. B.; Verma, S. Participation of Aromatic Side Chains in Diketopiperazine Ensembles. Tetrahedron Lett. 2008, 49, 4231−4234. (23) Song, Y.; Challa, S. R.; Medforth, C. J.; Qiu, Y.; Watt, R. K.; Pena, D.; Miller, J. E.; Swol, F.; van Shelnutt, J. A. Synthesis of Peptide-Nanotube Platinum-Nanoparticle Composites. Chem. Commun. 2004, 1044. (24) Guo, C.; Luo, Y.; Zhou, R.; Wei, G. Probing the Self-Assembly Mechanism of Diphenylalanine-Based Peptide Nanovesicles and Nanotubes. ACS Nano 2012, 6, 3907−3918. (25) Guo, C.; Luo, Y.; Zhou, R.; Wei, G. Triphenylalanine Peptides Self-Assemble into Nanospheres and Nanorods That Are Different from the Nanovesicles and Nanotubes Formed by Diphenylalanine Peptides. Nanoscale 2014, 6, 2800. (26) Tamamis, P.; Adler-Abramovich, L.; Reches, M.; Marshall, K.; Sikorski, P.; Serpell, L.; Gazit, E.; Archontis, G. Self-Assembly of Phenylalanine Oligopeptides: Insights from Experiments and Simulations. Biophys. J. 2009, 96, 5020−5029. (27) Dalgicdir, C.; Sensoy, O.; Peter, C.; Sayar, M. A Transferable Coarse-Grained Model for Diphenylalanine: How To Represent an Environment Driven Conformational Transition. J. Chem. Phys. 2013, 139, 234115. (28) Jeon, J.; Mills, C. E.; Shell, M. S. Molecular Insights into Diphenylalanine Nanotube Assembly: All-Atom Simulations of Oligomerization. J. Phys. Chem. B 2013, 117, 3935−3943. (29) Sugita, Y.; Okamoto, Y. Replica-Exchange Molecular Dynamics Method for Protein Folding. Chem. Phys. Lett. 1999, 314, 141−151. (30) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. Development and Testing of a General Amber Force Field. J. Comput. Chem. 2004, 25, 1157−1174. (31) Case, D. A.; Darden, T. A.; Cheatham III, T. E.; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M.; et al. AMBER 12, 2012. (32) Pearlman, D. A.; Case, D. A.; Caldwell, J. W.; Ross, W. S.; Cheatham, T. E.; DeBolt, S.; Ferguson, D.; Seibel, G.; Kollman, P. AMBER, a Package of Computer Programs for Applying Molecular Mechanics, Normal Mode Analysis, Molecular Dynamics and Free Energy Calculations to Simulate the Structural and Energetic Properties of Molecules. Comput. Phys. Commun. 1995, 91, 1−41. (33) Lloyd, S. Least Squares Quantization in PCM. IEEE Trans. Inf. Theory 1982, 28, 129−137. (34) Dill, K. A.; Ozkan, S. B.; Shell, M. S.; Weikl, T. R. The Protein Folding Problem. Annu. Rev. Biophys. 2008, 37, 289−316. (35) Gee, J.; Shell, M. S. Two-Dimensional Replica Exchange Approach for Peptide−Peptide Interactions. J. Chem. Phys. 2011, 134, 064112. (36) Ferrenberg, A. M.; Swendsen, R. H. Optimized Monte Carlo Data Analysis. Phys. Rev. Lett. 1989, 63, 1195−1198. (37) Kumar, S.; Rosenberg, J. M.; Bouzida, D.; Swendsen, R. H.; Kollman, P. A. THE Weighted Histogram Analysis Method for Free Energy Calculations on Biomolecules. I. The Method. J. Comput. Chem. 1992, 13, 1011−1021. (38) Kumar, S.; Rosenberg, J. M.; Bouzida, D.; Swendsen, R. H.; Kollman, P. A. Multidimensional Free Energy Calculations Using the Weighted Histogram Analysis Method. J. Comput. Chem. 1995, 16, 1339−1350. (39) Baker, E. N.; Hubbard, R. E. Hydrogen Bonding in Globular Proteins. Prog. Biophys. Mol. Biol. 1984, 44, 97−179. (40) Jaworska, M.; Jeziorna, A.; Drabik, E.; Potrzebowski, M. J. Solid State NMR Study of Thermal Processes in Nanoassemblies Formed by Dipeptides. J. Phys. Chem. C 2012, 116, 12330−12338. (41) Ben-Tal, N.; Sitkoff, D.; Topol, I. A.; Yang, A.-S.; Burt, S. K.; Honig, B. Free Energy of Amide Hydrogen Bond Formation in Vacuum, in Water, and in Liquid Alkane Solution. J. Phys. Chem. B 1997, 101, 450−457.

(42) Gervasio, F. L.; Chelli, R.; Procacci, P.; Schettino, V. The Nature of Intermolecular Interactions between Aromatic Amino Acid Residues. Proteins Struct. Funct. Bioinf. 2002, 48, 117−125.

6652

dx.doi.org/10.1021/jp501503x | J. Phys. Chem. B 2014, 118, 6644−6652