ARTICLE pubs.acs.org/JPCC
Self-Assembly of DNA Segments on Graphene and Carbon Nanotube Arrays in Aqueous Solution: A Molecular Simulation Study Xiongce Zhao†,* Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States ABSTRACT: Molecular dynamics simulations were performed to study the interaction of double-stranded DNA segments with the surfaces of graphene and carbon nanotube arrays in aqueous solution. Several different kinds of self-assembly phenomena were observed. First, it is found that a DNA segment can ‘stand up’ on the carbon surfaces with its helix axis perpendicular to the surfaces of graphene or nanotube arrays to form a forestlike structure. Second, a DNA segment can also lie on the carbon surface with its axis parallel to the surface if both of its ends can form stable structure with the carbon surfaces. In the latter case, the ending basepairs of the DNA are broken due to severe deformations. Third, it is observed that short DNA segments can concatenate to each other to form a longer DNA when they are placed in the grooves of nanotube bundles. The self-assembly observed in this study usually happens in less than 50 ns. Exploration on the molecular details and self-assembly mechanism indicates the primary driving force is the π stacking interaction between the ending basepairs of DNA and the carbon rings. This study confirms the dominant role of hydrophobic π stacking in the interaction between nucleotides and carbon-based nanosurfaces in aqueous environment.
I. INTRODUCTION Recently, a great deal of research attention has been paid to the interaction of biomolecules and carbon-based nanoparticles such as nanotubes (CNT) and C60. The studies of these systems are motivated by a wide spectrum of applications ranging from drug discovery,15 to environmental protection,6,7 and to biomedical nanodevices.813 For example, as one of the representative carbonaceous nanoparticles, water-soluble C60 derivatives have been used as gene therapy agent14 and photosensitizer in disease treatments.15 Experimental efforts are also made to explore the impact of C60-based nanoparticles on living cells.16,17 The other popular nanoparticle, CNT, which has even more diverse properties, had also received much attention in a variety of biomedical applications.1820 Representative examples include using CNT as biosensors,21,22 as drug delivery agents,23 and as diagnosis tools.24 Obviously, developments of aforementioned applications are critically dependent on our understanding of the fundamental interaction of nanoparticles with biomolecules, much of which remain as open questions. Compared with applications of carbon-based nanoparticles in other fields such as materials sciences and electronics, one of the challenges in studying nanoparticlebiomolecule interaction comes from the existence of solvents in the system. Nearly all of the biomolecules require an aqueous environment to function properly. Therefore, the hydrophobicity of nanoparticles usually plays a significant role in their interaction with biomolecules, in turn impacts the properties of most nanobio systems. One typical example is the r 2011 American Chemical Society
DNACNT/C60 system. Experimental and theoretical investigations in the past few decades all suggest that the hydrophobic/ hydrophilic interactions between the DNA bases and the carbon rings in CNT/C60 in aqueous environment is one of the dominant forces in deciding the mechanism and structure of the system. This has been confirmed directly or indirectly in studies by different groups. For example, the simulation work by Gao and co-workers2527 indicates that single-stranded DNA (ssDNA) segments can readily adsorb from solution inside a singlewalled carbon nanotube of appropriate diameter due to the hydrophobic parts of DNA being attracted to the interior of the nanotube. This was subsequently observed experimentally by Okada and co-workers.28 Johnson and co-workers29 found that carbon nanotube can induce ssDNA to undergo a spontaneous conformational change that enables the hybrid to self-assemble via the ππ stacking interaction between ssDNA bases and the nanotube sidewall. This helps to elucidate the experimental observations by Zheng and co-workers30,31 that ssDNA can wrap onto carbon nanotubes. In a previous work, we investigated the interaction of DNA segments and CNT/C60 in aqueous solution via molecular dynamics simulations. We found that C60 or CNT can interact with nucleotide segments strongly, either through hydrophobic ππ stacking,32,33 or through hydrogen bonding33,34 when the Received: October 19, 2010 Revised: February 20, 2011 Published: March 18, 2011 6181
dx.doi.org/10.1021/jp110013r | J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C nanoparticles are functionalized. Under either circumstance, the carbon-based nanoparticles form very stable hybrids with the DNA segments, with typical binding energies in the range of 20 to 40 kcal/mol. Binding energies of such magnitudes are much stronger than that between nanoparticles themselves (i.e., the binding energy of two pristine C60 is about 7 kcal/mol in water34). The hydrophobic forces calculated in our studies are in qualitative agreement with recent results calculated using first principle methods. For example, Stepanian and co-workers35 found that the interaction energy in the complex formed by cytosine and a fragment of (10,0) single-walled carbon nanotube is in the range of 60 kJ/mol. And the first principle calculations by Antony and Grimme36 indicate that the noncovalent interactions of stacked nucleobase and graphene are in the range of 20 to 25 kcal/mol. All of these results underline the importance of hydrophobic interaction between the bases in the nucleotides and nanoparticles with carbon ring structure. Indeed, pristine C60 or CNT are almost nonsoluble in water due to their strong hydrophobicity. Dispersion of CNT or C60-based molecules into aqueous solution has to be realized by either adding in cosolvents or through functionalizing them with hydrophilic groups. In contrast, the backbone of a nucleotide segment is strongly hydrophilic. The primary hydrophobic parts on a DNA molecule are the bases. Therefore, a DNA segment would bind with the carbon surface of these nanoparticles through π-stacking between the nucleobase and carbon surface when they are put into aqueous environment. Given the very strong (sometimes dominant) nature of such force, one natural question is how DNA segments would interact with smooth carbonaceous surfaces with aromatic structures, such as graphene, in solution. Previous studies suggest that long DNA strands can wrap on a single CNT to form a spindle-like structure, and such observation has been utilized in nanofabrications involving DNA and CNT.37,38 However, it is unclear how short DNA segments assemble on aromatic carbon surfaces. Obviously, short DNA segments (such as those with less than 20 bases) are unlikely to wrap a CNT due to the strong deformation energy required for the DNA to bend.33 It is plausible though, that the DNA would aggregate to form clusters or align on the carbon surface because of the DNADNA interaction. To our knowledge, there is no study in this area, especially on a molecular level. Given the complexity of the systems, it is challenging to probe them by experiment. An alternative approach is theoretical modeling, which has been prevalent in recent years in studying nanobiological systems. Molecular simulation can provide insight to questions exposed in experiments, as well as predict useful candidates for targeted applications. This study attempts to utilize molecular dynamics simulations to investigate the self-assembly properties of short DNA segments on two types of carbon surfaces, namely, the graphene surface and the surface of carbon nanotube arrays.
II. SIMULATION METHODS The double-stranded DNA investigated in this study consists of 8 or 12 basepairs, which is composed of two or three consecutive repeats of AGTC duplexes (d-poly(AGTC)2, d-poly(AGTC)3). The sequence and length of the DNA segments were chosen with several concerns. First, using repeats of AGTC enables one to investigate all 4 types of nucleotides. In addition, DNA with such sequence allows us to study DNA with 2 different types of ending basepairs, AT and GC. This is
ARTICLE
important because our previous studies3234 implied that interaction of GC or AT basepairs with a graphitic surface be different due to different pairing strengths in GC and AT. This study confirmed this, as will be discussed in details. Relatively short DNA segments were used in the simulation to keep the computation at a manageable level. Self-assembly of four to six DNA segments on two types of carbon surfaces were investigated, namely a graphite surface and a carbon nanotube array surface. The graphite was modeled as five stacked layers of graphene sheets, each graphene consisting of 3680 carbon atoms, with an area of 9.88 9.84 nm2 (xy plane). The distance between graphene layers is 3.35 Å. On the basis of a previous study, five layers of graphene sheets are sufficient to mimic a semi-infinite graphite surface.39 The carbon nanotube array consists of five to seven aligned (6,6) singlewalled carbon nanotubes. Each nanotube includes 44 units, with a length of 10.67 nm and a diameter of 8.02 Å. The surface formed by the CNT array is referred to as the xy plane, with the x axis being along the axes of CNTs. The distance between the axes of neighboring aligned CNTs is 11.42 Å. Four to six DNA segments are placed above the graphite or CNT array surfaces, with the axis of DNA helix parallel to the surface. The nearest initial distance between any pair of atoms on DNA and surface is about 10 Å. The DNA/CNT or DNA/graphite are wrapped in explicit water boxes. The solvated box had water buffer layers at least 2.0 nm thick between the solute surface and simulation box boundary in all three directions. Appropriate number of Naþ ions are added to the system to neutralize the negative charges from DNA. Total number of atoms for a typical system is about 120 000 to 160 000, depending on system details. Schematic description of the systems (water not shown) can be found in Figures 1 and 6. The DNA and ions were modeled by the AMBER 1999 force field.40 The carbon atoms in graphene and CNTs were modeled by the AMBER 1999 force field for generic aromatic carbon atoms. The details about the AMBER force fields can be found in the literature.40 The LJ interaction parameters between different atoms were calculated by the standard LorentzBerthelot combing rules, σij = (σi þ σj)/2 and εij = (εiεj)1/2. The cut off distance for LJ interactions was 1.0 nm with smooth shift, and atom-based pair-list with 1.1 nm were updated during the simulation. Water molecules were modeled by the TIP3P potential.41 Periodic boundary conditions are applied in all three directions. The particle-mesh Ewald method with a fourth order interpolation and direct space tolerance of 106 was applied to evaluate electrostatic interactions. Molecular dynamics (MD) simulations were performed within the constant pressure (1 bar) and constant temperature (300 K) ensemble.42 The NAMD43 software package was employed to integrate the equations of motion. The graphite and CNTs are fixed during all the simulations. Each simulation included 20 000 steps of energy minimization using a conjugate gradient algorithm, followed by gradual heating from 0 to 300 K in 3 ps, solvent equilibration for 5 ps with the DNA backbone atoms constrained, and equilibration of 100 ps without any constraints (except fixed carbon surfaces). Typical production simulations lasted up to 50 ns. Based on results from our previous studies,3234 simulations with such time scales are sufficient for the system to equilibrate and obtain stabilized molecular structures. For example, in one of our previous work,33 we simulated the interaction of a DNA with CNT in water solution by starting from two completely different initial configurations to monitor 6182
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
ARTICLE
Figure 1. Self-assembly of DNA segments on graphene layers. Snapshots were taken at (a) t = 0 ns, (b) t = 8 ns, (c) t = 16 ns, (d) t = 42 ns. Both top view and side view are shown for each snapshot. Color scheme: gray(C), red(O), blue(N), yellow(P), white(H).
the stabilizing times required. It was found that they converged to the same final configurations in less than 10 ns. Compared with that study, the DNA segments employed here is shorter. So the time required for the system to stabilize is expected to be shorter or in the same order of magnitude. Nonetheless, in this
study we continue each simulation for additional 1020 ns after stable self-assembled structure is observed, to ascertain that the system is converged. The integration time step chosen was two fs. The structural configurations were saved every 1 ps for subsequent analysis. 6183
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
ARTICLE
Figure 2. Evolution of angles between the DNA axes and the graphene surface normal as a function of simulation time. Blue, DNA 1; black, DNA 2; green, DNA 3; red, DNA 4.
Visualizations and analysis were performed using the VMD44 software packages.
III. RESULTS AND DISCUSSION A. Self-Assembly of DNA Segments on Graphite. We begin our study by simulating the self-assembly of double-stranded DNA segments on graphene surface. Four d-poly(AGTC)3 segments are placed on the surface at the beginning of simulation, with their axes parallel to the graphite surface. The four DNA segments are aligned, as shown in Figure 1, and labeled as 14, respectively. The initial distance between the axes of DNA 1 and DNA 3 along the y direction (which is perpendicular to the DNA helix axis in Figure 1) is 3.6 nm. And the distance between the center of DNA 1 and 2 along the x direction (which is parallel to the DNA axis in Figure 1) is 6.2 nm. Interestingly, it was observed that the DNA segment can adjust their geometric orientations rapidly and form two types of distinct self-assembled structures on the graphite surface. First, DNA segments 1, 3, and 4 were able to rotate and stabilize in a ‘stand-up’ state. The axes of the double helices turned gradually from a parallel to a perpendicular geometry within 30 ns. This results in a forestlike structure on the surface by the rotated DNA segments. Given the relatively large size of the DNA molecules and the existence of solvent, such a self-assembly process (within tens of ns) is surprisingly fast. In the second type of self-assembled structure observed, a DNA segment (i.e., segment 2) was lying on the surface and keeping its original orientation. Such structure is also very stable since no perturbation was observed up to 50 ns of simulation time. The two types of assembly process are illustrated by the snapshots shown in Figure 1 and the trajectories shown in Figure 2. In Figure 1, the snapshots of the simulation were taken at t = 0, 8, 16, and 42 ns, respectively. It is seen that DNA 4, 1, and 3 fully rotated from parallel to perpendicular geometry at about t = 10, 20, and 30 ns, respectively. This is reflected in the plots of the angles between their axes and the surface (red, blue, and green lines in Figure 2). In contrast, DNA 2 never rotated and its axis kept an angle of about zero degree to the graphite surface (Figure 2, black line). After examining the interacting details of each DNA segment with the graphite surface, we found that the hydrophobic ππ stacking is the dominant force behind the first type of self-assembly.
Figure 3. (a) Evolution of the π-stacking parameters between one ending basepairs (A1T24) of DNA segment 3 and graphene surface, (b) binding energy between DNA 3 and the surface, (c) the final stabilized structure of DNA 3.
Taking DNA 3 as an example, we found that one of its ending base pairs (A1T24) interacts with the graphene carbon rings strongly during the self-assembly process to form very stable π stacking structure. Two parameters are monitored to characterize the π stacking interaction. The first one is the relative angles (γ) between the contacting ending basepair and the graphite surface, and the second one is the distance (d) between the basepair plane and the surface. These two parameters for A1T24 of DNA 3 are plotted as a function of simulation time, as shown in part a of Figure 3. It is found that very stable π interaction was established as early as at t = 30 ns. At the beginning, γ was fluctuating rather randomly (blue line in part a of Figure 3). Correspondingly, the value of d was fluctuating between 0.4 to 3.0 nm. At about 30 ns, the values of γ and d suddenly drop to about zero degrees and 3.4 Å respectively, which represent standard π stacking features between two ring structures. The two parameters stay at these two values in the remaining time of the simulation, with negligible fluctuations, implying that the π interaction between the basepair and the surface is highly stable. We also calculated the binding energy of the DNA and the surface. Here the binding energy is defined as the potential energy between the bound DNA and the surface. Such energetic information can be readily calculated from the postsimulation analysis of the trajectories collected. The binding energy is approximated by averaging the DNAgraphene interaction potential energies computed from frames collected once the system of interest reaches stable state. An example for the binding energy of DNA 3 and graphite surface is given in part b of Figure 3. The interaction between the DNA and surface starts from about 10 kcal/mol. There is a slight increase during t = 010 ns. The interaction is 6184
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
ARTICLE
Figure 4. π-interaction between one ending basepair, A1T24, of DNA 4 and graphene surface.
essentially zero between 10 and 20 ns, indicating the DNA is not interacting with the surface during this period. At about t = 30 ns, there is a dramatic drop in the binding energy, corresponding to the rotation of the DNA and establishment of stable π stacking structure with the surface. The energy stabilizes at about 53 kcal/mol afterward. The corresponding final structure of the DNA and graphite is shown in part c of Figure 3. It is found that the formation of stable π stacking structure for DNA and graphite surface happens in a very short time. One example is shown in Figure 4, for DNA segment 4. It can be seen that stable π structure is formed in as short as 3 ns. Some lag time may exist between the formation of stable π stacking and the establishment of a fully assembled DNA geometry on the graphene surface. For example, for DNA 4 the initial π stacking was formed by t = 3 ns, whereas the segment’s final orientation was stabilized until t = 12 (red line in Figure 2). Likewise for DNA 3, the π interaction was initiated around t = 18 ns and fully established by t = 28 ns, whereas the final DNA-surface structure stabilized at about t = 30 ns (green line in Figure 2 and the binding energy curve in Figure 3). Theoretically, both ends of the DNA would have an opportunity to interact with the hydrophobic graphene surface. Then, one may expect to see the DNA to bend into a horseshoe shape for both ends to contact with the surface. However, for the short DNA segments investigated in this study, bending them into horseshoe shape would require significant energetic penalties. Therefore, to keep its conformation in standard B-format, the DNA segments ‘prefer’ to a slow rotation until they fully ‘stand up’, with their axes perpendicular to the surface. The rotation is driven by the strong π interaction and antibending force but resisted by the energy required to rearrange the solvent molecules around the segment. The fact that the process is able to finish in a relatively short time, suggesting that the π stacking between the basepairs and graphite surface is dominant. On the basis of such observations, it is highly likely that a sufficiently long DNA segment, which is more flexible for bending, would assemble itself on a graphene surface with both its ends attached to the surface while its middle part floating in the solvent due to the hydrophilic nature of the DNA backbones. The terminal basepairs of DNA segments modeled in this study are either AT or GC. Interestingly, it is found that for the first type of self-assembly, the AT end of the DNA is more likely to initiate a π interaction with the carbon surface than the GC end. For example, the assembled DNA segments 1, 3, and 4 shown in part d of Figure 1 all have their AT ends in contact with the graphite surface while their GC ends extend into solution.
Figure 5. π-interaction between two ending bases, C12 and A1, of DNA 2 and graphene surface. (a) and (b) describe the π stacking parameters and (c) and (d) are the binding energy and the stabilized molecular structure of DNA 2, respectively.
This can be explained by the difference in the pairing strength of AT and GC. Given that the beginning orientation of the ending baseplanes are perpendicular to the surface, the ending AT or GC planes has to rotate 90° to have a face-on contact with the graphene to initiate a π stacking. The most significant deformation observed is the breaking of ending basepairs. After the breaking, ending bases are flexible to rotate and stack effectively with the surface. The AT basepair only contains two hydrogen bonds, whereas the GC basepair has three of them. Therefore, it is relatively easier for the ending AT basepair on the DNA to deform and/or open. The phenomena observed in this study is consistent with the implication by our previous studies33,34 that the interaction between DNA and graphene surfaces are dominated by the π stacking force between the baseplanes of ending nucleotides (AT or GC), which are the only exposed hydrophobic surfaces on the DNA, and the carbon rings in graphene. Molecular feature of the second type of self-assembly of DNA 2 on graphene surface is illustrated by Figure 5. Details of the stabilized structure indicates that it is also driven by π stacking force. It is found that two nucleobases on each end of the DNA segment were able to form stable π-stacking with the graphene, as illustrated by the evolution of γ and d shown in parts a and b of Figure 5. A snapshot of stabilized structure is given in part d of Figure 5, featuring the stacking between the end bases and the carbon rings in graphene. Different from the ones observed for DNA 1, 3, and 4, the π stacking of each end of DNA 2 on the graphite surface only involves one base. That is, the ending basepairs of DNA 2 were broken during the binding in order for 6185
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
ARTICLE
Figure 6. Assembly of dsDNA segments on single-walled carbon nanotube surface. (a) Snapshot at t = 0 ns, (b) snapshot at t = 28 ns.
both DNA ends to attach to the surface. In this example, bases C12 and A1 on the two ends are in contact with the graphite surface, while their pairing bases G13 and T24 are dangling in solution. The hybrid structure formed therein corresponds to a binding energy of about 66 kcal/mol (part c of Figure 5). Such a binding energy is much stronger than the typical hydrogen bonds between the two ending basepairs, A1T24 or C12G13. As shown in part a of Figure 5, the binding of C12 with the surface started at about t = 21 ns. The other ending base, A1, begins to form a π bond with the surface at a much earlier time, about t = 2 ns (part b of Figure 5). Again, this is consistent with the fact that AT basepairs are easier to break than GC basepairs due to its fewer hydrogen bonds. The stacked structures are highly stable, indicated by the steady values of γ and d once the interaction was established. We also simulated similar systems with more DNA segments, including six d-poly(AGTC)3 instead of four. Qualitatively similar self-assembly phenomenon was observed. Typically it took less than 30 ns for some of the DNA segments to ‘stand up’ on the graphite surface, whereas the other segments lying flat on the surface with the ending basepairs disrupted and in contact with the carbon surface with a face-on pattern. Our simulations indicate that the probability of ‘stand-up’ versus ‘lie-flat’ for a particular DNA segment may depend on its starting configuration. If the DNA segment is tilted at the beginning, they are more likely to stand up within the simulated ns time scale. If they are lying on the surface initially, then both types of final configurations could be observed, with the ‘standup’ structure more prevalent. However, we did not observe a selfassembly mechanism other than these two types up to 50 ns of simulation. In real applications, one might be able to control the initial configurations of the DNA segments by applying electric field to the surface. For example, if the ‘lie-flat’ configurations are preferred, a positive charge can be applied on the surface to attract the negatively charged DNA backbones. B. Self-Assembly of DNA Segments on CNT Arrays. We have performed simulations to study the self-assembly of doublestranded DNA segments on surfaces composed by CNT arrays.
Figure 7. Evolution of angles between the DNA axes and the CNT surface normal as a function of simulation time. Green, DNA 1; black, DNA 2; red, DNA 3; orange, DNA 4; brown, DNA 5; blue, DNA 6.
Six short DNA segments, each consisting of 8 basepairs (d-poly (AGTC)2), are initially placed in two rows and three columns on the CNT surface, with their helix axes parallel to the CNT axis. One example of system setup is shown in part a of Figure 6. The space between the nearest atoms on two adjacent DNA segments is about 10 Å along the x direction (which is parallel to the CNT axes), and about 15 Å along y direction. Compared with the DNA/graphene systems, slightly shorter DNA segments are used for the DNA/CNT systems, allowing us to accommodate the system with more DNA segments in a computationally manageable simulation box. This is important for studying the interaction/association between DNA segments in addition to that between DNA and CNTs. In a previous work, we have investigated the interaction of longer DNA and CNT.33 Two distinct types of self-assembly features, which are similar to those found in DNAgraphite systems, are observed for DNA on CNT surface. DNA segments can either ‘stand up’ to a perpendicular orientation in respect to the CNT surface, or stay flat. Comparison of the starting and stabilized snapshots are given in Figure 6. In this simulation, DNA segments 1, 4, 5, and 6 were able to rotate to an upward position within 30 ns, with one of their ending basepairs in face-on contact with the CNT surface. DNA 2 and 3 remain a lying-flat geometry, like that 6186
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
Figure 8. Two different types of interaction between a DNA segment and surface of CNT arrays.
observed in DNAgraphite studies. The assembly process is depicted by the angle between the DNA axes and the CNT surface normal as a function of simulation time, as shown in Figure 7. It is seen that the angles for DNA 1, 4, 5, and 6 turned quickly from zero to about 90° relative to the CNT array plane, whereas those for DNA 2 and 3 are fluctuating around zero throughout the simulation. Compared with the self-assembly of DNA segments on graphene surface, it took much shorter time for the DNA segments to stabilize on CNT surface. For example, most of the DNA segments reach stable states within 10 ns on the CNT surface, as shown by the plots in Figure 7. This is probably due to the relatively shorter DNA segments used. Compared with the DNA segments with 12 basepairs, the shorter segments are more spherical in geometric shape, which enable them to adjust their orientation in shorter time scales in solvent. The dominant driving force behind the self-assembly of DNA segments on CNT surface is still the hydrophobic π stacking interaction. Like in the DNAgraphene systems, the ending basepairs of DNA segments are able to interact with the CNT surface with a face-on pattern. Snapshots showing such a contacting mechanism are shown in Figure 8. For DNA segments that ‘stand up’, one of its ending basepairs readily ride on the sidewall of one CNT, with the tips of the backbones extruding into the interstitial grooves of the CNT bundle (part a of Figure 8). For DNA that lies flat on the surface, one of its end basepairs breaks up to stack on the CNT walls (part b of Figure 8). Those two hybrid structures correspond to interaction energies of 47 and 51 kcal/mol, respectively. Similarly, we recorded the two geometric parameters, d and γ, to monitor the formation of π stacking between DNA and CNT surface. One example for DNA 5 is shown in Figure 9. The angle between the base plane of one of the DNA ending basepairs, A1T16, and the CNT surface, as well as the distance between them, are plotted as a function of simulation time. We note that here the CNT surface is defined as the xy plane spanned by the CNT wall edges. It can be seen from Figure 9 that stable π stacking was formed rapidly once the DNA are brought to interact with the CNT surface. Typical time required for the π stacking to stabilize is less than 5 ns. During this period, the angle between the plane of interacting base and CNT surface goes from 90° to about zero. Correspondingly, the distance between the contacting planes drops from >14 Å to about 3.4 Å. Slightly different from that observed in DNAgraphite interaction, the fluctuations in γ and d here are more intensive (comparing Figure 9 with Figures 35). This is probably due to the surface corrugations from the CNT interstitial grooves,
ARTICLE
Figure 9. π-interaction between the A1T16 end of DNA 5 and CNT array surface.
which reduces the stability of the π stacking interaction with DNA bases and is reflected by the fluctuations in the two parameters monitored. Like that seen in DNAgraphite simulations, we found that the AT-basepaired ends of the d-poly(AGTC)2 are more likely to interact with the CNT surface than those of GC basepairs. For the example in Figure 6, all four stand-up DNA segments have their AT ends in contact with the CNT surface and their GC ends extending into solution. One interesting phenomenon observed in the DNACNT simulation was that short DNA segments can be spontaneously catenated with each other to form a longer segment. One such example is shown in the snapshots given in Figure 10. In this simulation, the starting configuration was exactly the same as that in part a of Figure6. After several nanoseconds, three of the DNA segments (4, 5, and 6) were able to stand up on the CNT surface, through the mechanism discussed above. The other three DNA segments (1, 2, and 3), which are aligned in a row initially, associate to each other via a head-to-tail pattern (part b of Figure 10). The A1T16 basepair on DNA 1 connects to the C8G9 basepair on DNA 2, and A1T16 of DNA 2 connects to C8G9 of DNA 3. A stable, long DNA segments consisting of 24 basepairs is formed within 5 ns. This process is quantitatively described by the relative geometry between the interacting basepairs from the adjacent DNA, namely, A1T16 of DNA 1 with C8G9 of DNA 2, and A1T16 of DNA 2 with C8G9 of DNA 3. Two geometric parameters, the relative angle and the distance between the interacting basepairs, are used to illustrate the association process, as shown in Figure 11. It is seen that in less than 5 ns the angles between these interacting basepair planes are stabilized to about zero, while their center of mass distances stabilized around 5 Å, which are typical relative geometry for neighboring basepairs in a standard B-DNA double helix. The structure of the stabilized 24basepair DNA segment is presented in the snapshots in Figure 12. We notice that the assembled long DNA segment kept a roughly parallel orientation to the CNT surface. The simulation was continued for 18 ns in this case, no rotation or break-apart was observed for the assembled long DNA. Formation of long DNA segments by short ones was only observed for DNACNT systems. One possible reason is that the existence of interstitial grooves on the CNT surface facilitated the self-assembly of adjacent DNA to form a longer one. When several short DNA segments are located in an interstitial groove with their axes parallel to the groove, their movements are 6187
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C
ARTICLE
Figure 10. Spontaneous formation of a long DNA by three short DNA segments on CNT array surface. (a) t = 0 ns, (b) t = 18 ns.
Figure 12. Spontaneous formation of a long DNA by three short DNA segments on CNT surface, final stabilized structure.
Figure 11. Spontaneous formation of a long DNA by three short DNA segments on CNT array surface. Interaction between A1T16 of DNA 1 and C8G9 of DNA 2 are described by the blue and red lines; that between A1T16 of DNA 2 and C8G9 of DNA 3 by the green and orange lines.
effectively limited to translations along the groove direction. Therefore, the adjacent DNA segments have a better chance to concatenate to each other by a head-to-tail fashion than on a smooth graphene surface. For example, it is noticed that the assembled long segment lies right in the CNT groove (part b of Figure 10 and part b of 12), compared to their original locations (part a of Figure 10). This implies that the surface corrugation of the carbon surface would influence the self-assembly mechanism of DNA segments. We point out that the association between adjacent DNA segments is purely driven by physical forces. For the force fields employed in our simulations, no covalent bond breaking/formation was considered. Therefore, the end-to-end catenation between two neighboring DNA segments is dominated by the physical interaction between the two ending nucleotides. Nevertheless, such interaction is very strong and should be in the same order of magnitude compared with DNACNT interaction. We also performed additional simulations using a longer DNA segment, d-poly(AGTC)3, as well as using a larger CNT surface. The results are qualitatively similar to the ones shown, except
that the self-assembly of longer DNA segments took substantially longer time (∼30 ns) than the short ones. Simulations were also performed with different initial DNA orientations. It was found that most of the DNA segments would end up with a perpendicular structure if the simulation starts with DNA segment being tilted by 30° rather than lying flat on the surface, which is also consistent with that found in DNAgraphene simulations. In addition, under such an initial configurational setup, no DNADNA association was observed.
IV. CONCLUSIONS In summary, we found from molecular simulations that short DNA segments consisting of up to 12 basepairs can selfassemble on graphene or CNT array surfaces to form stable hybrid structures. The self-assembly occurs in a very short time, usually in less than 50 ns. Two types of assembly patterns were observed for DNA on graphene surface. First, DNA can rotate from a parallel to a perpendicular orientation on the surface to form a forestlike structure. Second, DNA can attach flat on the surface. The driving force behind both patterns are the π stacking interaction between the hydrophobic DNA basepairs and graphitic carbon rings. In particular, the ending basepairs of the DNA were broken apart when it binds to the surface in the second type of self-assembly. Interestingly, the AT basepairs are more likely to interact with carbon surfaces than the GC basepairs if DNA lies flat on the surface initially. 6188
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189
The Journal of Physical Chemistry C Short DNA segments were found to be able to assemble on the surface of CNT arrays. The self-assembly patterns were similar to those seen in DNAgraphene. However, the hybrid systems formed by DNA and CNT surface is less intensive than those from DNA and graphene, indicated by the slightly more noisy fluctuations in the π stacking parameters. Using of shorter DNA segments leads to an even faster self-assembly process. In addition, it was found that short DNA segments were able to assemble into longer ones on the CNT surface, probably facilitated by the surface corrugations on the CNT arrays. The results from this study again emphasize the importance of π stacking forces between nucleotides and surfaces with carbon rings in aqueous solution. Indeed, such forces play a significant role in any systems involving biomolecules, often dominating the conformation of biopolymers like DNA segments in aqueous environment. For biosystems that contain any solid surface with carbon ring structure, the hydrophobic π stacking interaction is inevitably one of the dominant driving force that may result in dramatic biopolymer-surface interfacial properties such as selfassembly. In closing, we note that further studies will be undertaken on the similar systems in the following aspects. First, the energetic information reported in this work is a simple binding energy cal culated from interaction potential energies. More meaningful energetic information is the binding free energy. One could calculate the free energy of the self-assembly observed by computing the potential of mean force associated with the binding process. However, such an approach requires the identification of a set of conformational coordinates that well define the process. In addition, solvent effect, which is often indirectly related to the binding free energy between the macromolecules in aqueous solution, is another important factor for such systems worthy of separate investigations.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. Present Addresses †
National Institutes of Health, Bethesda, MD 20892
’ ACKNOWLEDGMENT The author thanks Peter T. Cummings for many helpful discussions. This research was conducted at the Center for Nano phase Materials Sciences, which is sponsored at Oak Ridge National Laboratory by the Division of Scientific User Facilities, U.S. Department of Energy. This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. ’ REFERENCES (1) Da Ros, T.; Prato, M. Chem. Commun. 1999, 663. (2) Prato, M.; Kostarelos, K.; Bianco, A. Acc. Chem. Res. 2008, 41, 60. (3) Bianco, A.; Kostarelos, K.; Partidos, C. D.; Prato, M. Chem. Commun. 2005, 571–577. (4) Bianco, A.; Kostarelos, K.; Prato, M. Curr. Opin. Chem. Biol. 2005, 9, 674. (5) Bakry, R.; Vallant, R. M.; Najam-Ul-Haq, M.; Rainer, M.; Szabo, Z.; Huck, C. W.; Bonn, G. K. Int. J. Nanomedicine 2007, 2, 639.
ARTICLE
(6) Lam, C. W.; James, J. T.; McCluskey, R.; Arepalli, S.; Hunter, R. L. Crit. Rev. Toxicol. 2006, 36, 189. (7) Zhu, S. Q.; Oberdorster, E.; Haasch, M. L. Marine Environ. Res. 2006, 62, S5. (8) Daniel, S.; Rao, T. P.; Rao, K. S.; Rani, S. U.; Naidu, G. R. K.; Lee, H. Y.; Kawai, T. Sens. Actuators, B 2007, 122, 672. (9) Trojanowicz, M. Trac-Trends Anal. Chem. 2006, 25, 480. (10) Katz, E.; Willner, I. ChemPhysChem 2004, 5, 1085. (11) Yogeswaran, U.; Chen, S. M. Anal. Lett. 2008, 41, 210. (12) Cui, D. X. J. Nanosci. Nanotechnol. 2007, 7, 1298. (13) Balasubramanian, K.; Burghard, M. Anal. Bioanal. Chem. 2006, 385, 452. (14) Sitharaman, B.; Zakharian, T. Y.; Saraf, A.; Misra, P.; Ashcroft, J.; Pan, S.; Pham, Q. P.; Mikos, A. G.; Wilson, L. J.; Engler, D. A. Mol. Pharmaceutics 2008, 5, 567. (15) Kasermann, F.; Kempf, C. Rev. Med. Virol. 1998, 8, 143. (16) Kamat, J. P.; Devasagayam, T. P. A.; Priyadarsini, K. I.; Mohan, H.; Mittal, J. P. Chemico-Biological Interactions 2000, 114, 145. (17) Kamat, J. P.; Devasagayam, T. P. A.; Priyadarsini, K. I.; Mohan, H. Toxicology 2000, 155, 55. (18) Sinha, N.; Yeow, J. T.W. IEEE Trans. Nanobioscience 2005, 4, 180. (19) White, A. A.; Best, S. M.; Kinloch, I. A. Int. J. Appl. Ceram. Technol. 2007, 4, 1. (20) Foldvari, M.; Bagonluri, M. Nanomedicine-Nanotechnology Biology and Medicine 2008, 4, 173. (21) Wang, J.; Liu, G.; Jan, M. R. J. Am. Chem. Soc. 2004, 126, 3010. (22) Xu, Y.; Jiang, Y.; Cai, H.; He, P. G.; Fang, Y. Z. Anal. Chim. Acta 2004, 516, 19. (23) Martin, C. R.; Kohli, P. Nature Rev.: Drug Discovery 2003, 2, 29. (24) Peng, G.; Tisch, U.; Haick, H. Nano Lett. 2009, 9, 1362. (25) Gao, H.; Kong, Y.; Cui, D.; Ozkan, C. S. Nano Lett. 2003, 3, 471. (26) Gao, H.; Kong, Y. Annu. Rev. Mater. Res. 2004, 34, 123. (27) Pei, Q. X.; Lim, C. G.; Cheng, Y.; Gao, H. J. J. Chem. Phys. 2008, 129, 8. (28) Okada, T.; Kaneko, T.; Hatakeyama, R.; Tohji, K. Chem. Phys. Lett. 2006, 417, 288. (29) Johnson, R. R.; Johnson, A. T. C.; Klein, M. L. Nano Lett. 2008, 8, 69. (30) Zheng, M.; Jagota, A.; Semke, E. D.; Diner, B. A.; McLean, R. S.; Lustig, S. R.; Richardson, R. E.; Tassi, N. G. Nat. Mater. 2003, 2, 338. (31) Zheng, M.; Jagota, A.; Strano, M. S.; Santos, A. P.; Barone, P.; Chou, S. G.; Diner, B. A.; Dresselhaus, M. S.; McLean, R. S.; Onoa, G. B.; et al. Science 2003, 302, 1545. (32) Zhao, X. C.; Striolo, A.; Cummings, P. T. Biophys. J. 2005, 89, 3856. (33) Zhao, X. C.; Johnson, J. K. J. Am. Chem. Soc. 2007, 129, 10438. (34) Zhao, X. C. J. Phys. Chem. C 2008, 112, 8898. (35) Stepanian, S. G.; Karachevtsev, M. V.; Glamazda, A. Y.; Karachevtsev, V. A.; Adamowicz, L. Chem. Phys. Lett. 2008, 459, 153. (36) Antony, J.; Grimme, S. Phys. Chem. Chem. Phys. 2008, 10, 2722. (37) Wang, Z. D.; Lu, Y. J. Mater. Chem. 2009, 19, 1788. (38) Becerril, H. A.; Woolley, A. T. Chem. Soc. Rev. 2009, 38, 329. (39) Zhao, X. C.; Johnson, J. K. Mol. Simul. 2005, 31, 1. (40) Case, D. A.; Darden, T. A.; Cheatham, T. E.; III, Simmering, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K. M.; Wang, B.; Pearlman, D. A.; et al. , Amber 8.0, 2004. (41) Jorgensen, W. L. J. Am. Chem. Soc. 1981, 103, 335. (42) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Nola, A. D.; Haak, J. R. J. Chem. Phys. 1984, 81, 3684. (43) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kale, L.; Schulten, K. J. Comput. Chem. 2005, 26, 1781. (44) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 1.
6189
dx.doi.org/10.1021/jp110013r |J. Phys. Chem. C 2011, 115, 6181–6189