Article pubs.acs.org/Biomac
Effect of Hydrophobic and Hydrophilic Surfaces on the Stability of Double-Stranded DNA Robert M. Elder,*,†,‡ Jim Pfaendtner,§ and Arthi Jayaraman*,‡,∥ †
U.S. Army Research Laboratory, Aberdeen Proving Ground, MD 21005, United States Department of Chemical and Biological Engineering, University of Colorado, Boulder, Colorado 80309, United States § Department of Chemical Engineering, University of Washington, Seattle, Washington 98195, United States ∥ Departments of Chemical and Biomolecular Engineering and Materials Science and Engineering, University of Delaware, Newark, Delaware 19716, United States ‡
S Supporting Information *
ABSTRACT: DNA hybridization is the foundation for numerous technologies like DNA origami and DNA sensing/microarrays. Using molecular simulations, enhancedsampling methods, and free-energy calculations, we show the effects of hydrophilic and hydrophobic surfaces on DNA hybridization. Hydrophilic surfaces compete with terminal bases’ H-bonds but stabilize central base stacking. Hydrophobic surfaces strengthen terminal H-bonds but destabilize central base stacking. Regardless of surface chemistry, for terminal bases, melting proceeds through breaking H-bonds, followed by unstacking from the neighboring base. For central bases in bulk or near hydrophobic surfaces, melting proceeds by disruption of H-bonds, followed by unstacking, whereas on hydrophilic surfaces, unstacking from one neighboring base precedes complete disruption of the H-bonds, followed by unstacking from the second neighboring base. Kinetic barriers to melting and hybridization show that the central bases melt rapidly near hydrophobic surfaces, which can accelerate conformational searching and thereby accelerate folding into the desired conformation.
■
INTRODUCTION Hybridization of single-stranded DNA (ssDNA) into doublestranded DNA (dsDNA) underpins many technologies and experimental techniques. DNA origami, for example, involves self-assembly of complex DNA-based nanostructures using the sequence-specific hybridization of DNA.1 While hybridization for many technologies occurs in solution, technologies involving origami-templated assembly require the delicate DNA nanostructures to be transferred onto a solid substrate without damage, and thus, there is motivation to study dsDNA stability on surfaces.2,3 To preserve a DNA nanostructure on a substrate, the specific interactions underlying hybridization (e.g., hydrogen bonds, base−base stacking) must overcome complex nonspecific interactions with the surface. For example, hydrophobic and hydrophilic surfaces alter the conformations of DNA hairpins, with hydrophobic surfaces stabilizing the hairpin by increasing folding rates.4 Despite a critical need to understand dsDNA stability near surfaces, interfacial hybridization has received little attention compared to hybridization in solution.5−7 While applications involving surface-grafted DNA, like the assembly of DNA-grafted nanoparticles and DNA microarrays,8−20 have received significant attention, the thermodynamics governing stability of grafted DNA is significantly different from that of free DNA near surfaces. Compared to solution, surfaces may favor or disfavor the © XXXX American Chemical Society
hybridized state depending on the DNA sequence and the surface properties (e.g., hydrophobicity, charge), hinting that many distinct DNA−surface interactions underlie the overall surface effect.21−25 To elucidate these competing interactions, we study a dsDNA tetramer, d(CGCG)2, adsorbed to model hydrophilic (oligoethylene glycol, OEG), and hydrophobic (oligomethylene, OMe) surfaces (Figure 1) using atomistic molecular dynamics (MD) simulations, enhanced-sampling methods, and free-energy calculations. We quantify the stability of Watson− Crick (WC) hydrogen bonds (H-bonds) and base−base stacking (stacking), the interactions that govern the stability of dsDNA,26,27 and find that the surfaces significantly affect the stability of these interactions. The effect depends on the location of a base within the dsDNA; bases are either terminal (i.e., at the blunt end of the DNA) or central (i.e., away from the termini). Relative to bulk solution, the polar, well-hydrated OEG surface weakens the WC H-bonds of terminal bases through competition, whereas the lack of competition near the relatively dry OMe surface stabilizes the terminal WC H-bonds. For the central bases, the effect of the surfaces is reversed, with Received: April 9, 2015 Revised: May 8, 2015
A
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules
CH3(CH2)14CH3), as shown in Figure 1. These surfaces are identical to those in our recent publications and are intended to provide model hydrophobic and hydrophilic interfacial environments.6,7 Additional details of surface construction and equilibration are available in SI Methods. We simulated three cases: one consisting of a self-complementary DNA double strand d(CGCG)2 in bulk solution (i.e., without a surface), and two cases consisting of the same dsDNA with one of the two model surfaces. The three systems were neutralized with Na+ counterions and solvated with TIP3P water molecules.28 As in previous studies,6,7 no additional salt was included. As a result, these simulations were conducted with zero ionic strength, which, due to electrostatic repulsion between DNA strands, will tend to decrease dsDNA stability compared to physiological conditions.29 Other limitations of using zero added salt are discussed in our previous studies.6,7 The AMBER f f10 combination of force fields was used to parametrize DNA, water molecules, and counterions.30−32 The oligomers comprising the surfaces were parametrized with the general AMBER force field.33,34 Additional details of system construction and equilibration prior to use in production simulations are available in SI Methods. The GROMACS 4.6.4 MD engine and the PLUMED plugin were used to perform all simulations. VMD was used for visualization.35−37 Production simulations were conducted in the constant temperature and constant volume ensemble (NVT). The benefits and limitations of using constant volume simulations rather than constant pressure simulations are discussed in our previous publication.6 All results are at 300 K. The remaining details of our simulation set up are presented in SI Methods. Enhanced-Sampling Methods and Free-Energy Calculations. Previous simulation studies of dsDNA stability have shown that enhanced-sampling methods are necessary to observe hybridization with a tractable computational cost. In one study with particular relevance to the present work, a combination of replica exchange and umbrella sampling of the same dsDNA tetramer in implicit solvent was required to observe rehybridization.38 However, since we used explicit solvent, the number of atoms in our systems is larger, which would require a prohibitive number of replicas in standard replica exchange. Therefore, we used the parallel tempering metadynamics in the welltempered ensemble (PTMetaD-WTE) method to enhance conformational sampling with a reasonable computational cost.39 Metadynamics and the PTMetaD-WTE protocol are described in detail in other publications,39−42 and we only briefly summarize the methods here; additional details are in SI Methods. Metadynamics (MetaD) is an enhanced-sampling method wherein a time-dependent potential energy bias is constructed to discourage the system from revisiting previous conformations. In parallel tempering (PT), a set of simulations with increasing temperature are run in parallel, and the simulations are periodically allowed to exchange temperatures subject to the Metropolis criterion. MetaD enhances sampling along specific collective variables (CVs) that describe a property of interest, while the high temperature simulations in PT enhance sampling of all degrees of freedom. The well-tempered ensemble (WTE) decreases the computational cost of PT simulations.40 Furthermore, while MetaD is a nonequilibrium method, the MetaD bias can be reweighted to reproduce the equilibrium Boltzmann distribution, allowing the calculation of the equilibrium distribution and free energy of any property.43 The combination of these methods (PTMetaD-WTE) is thus a powerful tool for efficient conformational sampling. To verify that we achieved sufficient sampling, we conducted two independent PTMetaD-WTE simulations, and we assess the convergence of our simulations in the SI Methods (Figures S1−S3). Choice of Collective Variable. To use the PTMetaD-WTE method, one or more CVs must be defined that describe the property of interest, which in this case is the stability of a dsDNA. We describe a few of the most relevant past simulation studies to explain our choice of CVs. Collective Variables Used in Literature. The distance between two DNA strands has been used successfully to describe DNA hybridization.38 While this is appropriate for simulations in bulk solution, it is
Figure 1. (a) Oligoethylene glycol (OEG) and oligomethylene (OMe) oligomers used to construct hydrophilic and hydrophobic surfaces. (b) Cytosine and guanine nucleobases comprising the d(CGCG)2 dsDNA tetramer. Atoms involved in Watson−Crick (WC) H-bonds, and the glycosidic nitrogen atoms used to define stacking, are labeled. (c) Simulation snapshots of the dsDNA tetramer at two different values of the WC H-bonds collective variable used during metadynamics simulations. For near-surface simulations, the center of mass of one ssDNA strand is constrained to be 10 Å from the top of the surface.
the hydrophobic OMe surface destabilizing H-bonds, and the hydrophilic OEG surface stabilizing H-bonds. The central bases are less influenced by hydration because they are partially shielded from the interfacial environment by neighboring bases. Instead, the stability of the central H-bonds is linked to stacking, which favors H-bonds by prealigning the bases. Relative to bulk solution, stacking of terminal and central bases is stabilized by the OEG surface and destabilized by the OMe surface. The polar, hydrated environment near the OEG surface disfavors exposure by unstacking of the relatively hydrophobic nucleobases. In contrast, the relatively dry environment near the OMe surface allows bases to unstack more easily, and the OMe surface further destabilizes stacking through direct hydrophobic interaction with the bases. We discuss how these effects on individual base−base interactions may impact overall dsDNA stability near surfaces. Finally, we identify the pathways by which base−base interactions are disrupted (i.e., melting). Base melting generally proceeds first through breaking H-bonds followed by unstacking from neighboring bases, indicating that prealignment by stacking with neighboring bases favors the formation of WC H-bonds. Surface chemistry can qualitatively change the melting pathway. Additionally, the surfaces affect the kinetic barriers along the melting pathways, suggesting that surfaces may be designed to influence the kinetics and stability of selfassembled DNA nanostructures.
■
METHODS
Simulation Setup. We constructed hydrophilic (oligoethylene glycol, OEG) and hydrophobic (oligomethylene, OMe) surfaces from oligomers of hydroxyl-terminated OEG (chemical formula H(OCH2CH2)5OH) and methyl-terminated OMe (chemical formula B
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules not clear how to appropriately correct for the differences in accessible volume and concentration near surfaces.44,45 Surfaces may favor hybridization by reducing the accessible volume, as suggested in recent studies,4,24 or may disfavor hybridization by concentrating the negative charge of the DNA.5 Furthermore, the interstrand distance may not sufficiently sample other important degrees of freedom, such as stacking, WC H-bonds, or DNA−surface interactions. In previous studies,46,47 stacking was described as the distance between the glycosidic nitrogen atoms of two bases (N1 and N9 in Figure 1b). Umbrella sampling of the internitrogen distance was conducted, and the sequence dependence of the stacking free energy was in qualitative agreement with experimental results, suggesting that this CV accurately describes stacking in ssDNA dimers. However, the use of this CV with more than two bases is ambiguous, because multiple internitrogen distances must be considered. Hagan et al. conducted MD simulations of a dsDNA trimer and identified CVs that describe the destabilization of a terminal base pair in bulk solution.48 These CVs were the distances and interaction energies between a terminal base and its nearest neighbors in the canonical DNA structure, which are related to stacking and WC Hbonds. While these CVs successfully describe unbinding of a single terminal base, it is unclear how to generalize these CVs to multiple bases away from the termini. Our Collective Variables. Building on previous studies, we developed two CVs to enhance sampling of two major contributions to dsDNA stability, WC H-bonds and stacking.26 Compared with previous work, the chief advantage of our CVs is that they are applicable to multiple bases. WC H-bonds were defined between a hydrogen atom in one strand and the corresponding heavy atom in the opposite strand (Figure 1b) and quantified using a switching function with the following mathematical form.
sij
rij n
() = 1−( ) 1−
Figure 2. Effect of surfaces on Watson−Crick hydrogen bonds. (a) Free-energy landscape of WC H-bonds and base-surface distance for terminal bases. Data is for OEG; data for OMe is in SI Results (Figure S4). Contours are spaced by 1 kcal/mol. The dashed line is the location of the free-energy profiles shown in panel b. (b) Free energy as a function of WC H-bonds for terminal bases in bulk solution and near (0.4 nm from) the OEG and OMe surfaces. Error bars are the standard error of the mean. (c) Similar to (a), but for the central bases. (d) Similar to (b), but for the central bases at 1.1 nm surface distance. The locations of the terminal and central bases are shown in the diagram at the top.
r0 rij m r0
(1)
The tuning parameters n, m, and r0 determine the shape of the switching function. We used n = 8, m = 12, and r0 = 2.5 Å, which have been used to describe H-bonds in peptides.41 The sum of the WC Hbonds in all bases was used as a CV during metadynamics. For each base, both base pairing partners in the opposite strand were included in the sum (e.g., the cytosines in strand 1 can pair with either guanine in strand 2). We prevented the strands from fully separating and exploring irrelevant conformations by using a half-harmonic wall at WC H-bonds = 1. We found that biasing the sum of the WC H-bonds was insufficient to sample stacking (see SI Methods), and consequently, we introduced a second CV to enhance sampling of stacking. To define stacking between two bases, we used the distance between the glycosidic nitrogen atoms of the bases (Figure 1b), which previous studies have found to be a qualitatively reasonable measure of stacking.46,47 This distance was quantified using eq 1 with tuning parameters n = 6, m = 12, and r0 = 6 Å. The value r0 = 6 Å is a heuristic that previous authors used to delineate stacked and unstacked conformations.46,47 The sum of all possible stacking pairs, both within and between DNA strands, was used as a CV during metadynamics. By biasing both the stacking CV and H-bonding CV, we simultaneously enhanced the sampling of both CVs during PTMetaD-WTE simulations.
individual bases rather than all bases. For simplicity, we only include H-bonds between canonical base pairing partners, so each base can form a maximum of 3 H-bonds. Base-surface distance is defined as the distance between the center-of-mass of the topmost heavy atoms in the surface and center-of-mass the heavy atoms in individual nucleobases. Free-energy landscapes are calculated individually for each base and averaged appropriately: within each system (bulk, OEG, and OMe), results for the four terminal bases are averaged together, and results for the four central bases are averaged together. Figure 2 shows the location of terminal and central bases. The error bars in Figure 2 are the standard error of the mean over the four bases and two trials. For the terminal bases, four free-energy minima are apparent in the free-energy landscape (OEG, Figure 2a; OMe, Figure S4). Considering base-surface distance, two of these minima are located near the surface (∼0.4 nm) while two of these minima are located far from the surface (∼1.4 nm). Considering the number of WC H-bonds, two of the minima are in a melted state (∼0 H-bonds) and two of the minima are in a hybridized state (∼3 H-bonds, i.e., the number of H-bonds in a GC base pair). Thus, it is apparent that WC H-bonds in terminal bases can be either melted or hybridized when the bases are close to or far from the surfaces. To examine the effect of the surfaces on the stability of Hbonds in terminal bases, we extract 1D free-energy profiles at a
■
RESULTS AND DISCUSSION Stability of Watson−Crick H-Bonds. To show the effect of the hydrophilic OEG and hydrophobic OMe surfaces on the stability of WC H-bonds, we construct 2D free-energy landscapes as a function of WC H-bonds and base-surface distance. These 2D landscapes are constructed using the reweighting procedure described previously and implemented in PLUMED2.36,43 WC H-bonds in Figure 2 are quantified similarly to the H-bonds CV used for metadynamics but for C
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules base-surface distance of 0.4 nm on the 2D landscapes for the OEG and OMe systems (Figure 2b). Since a landscape like that in Figure 2a cannot be constructed for the bulk system, a 1D free-energy profile for the bulk system is calculated directly using the reweighting procedure.43 We define the quantity ΔGH‑bond as the difference in free energy at the melted state (∼0 H-bonds) and the hybridized state (∼3 H-bonds), which is negative if the formation of H-bonds is thermodynamically favorable. In bulk solution ΔGH‑bond of the terminal bases is about +1.0 kcal/mol, indicating that terminal WC H-bonds are only metastable. This finding is reasonable since terminal base pairs are known to fray,49 and the overall contribution of WC H-bonds to dsDNA stability has been shown to be negligible or even unfavorable.26 Near OEG, ΔGH‑bond is about +2 kcal/mol, while near OMe ΔGH‑bond is about 0 kcal/mol. Therefore, relative to bulk solution, terminal WC H-bonds are more stable near the hydrophobic OMe surface and less stable near the hydrophilic OEG surface. For the central bases the landscapes (OEG, Figure 2c; OMe, Figure S4) show only two minima, corresponding to the hybridized state (∼3 H-bonds) and the melted state (∼0 Hbonds). To examine the effect of the surfaces on the stability of central WC H-bonds, we extract 1D free-energy profiles at 1.1 nm base-surface distance (Figure 2d), which intersects the deepest parts of the minima in the 2D landscapes. Because ΔGH‑bond is negative for all three systems, H-bonds in the central bases are stable regardless of the presence or chemistry of a surface. However, the magnitude of ΔGH‑bond is affected by the surfaces. In bulk solution, ΔGH‑bond for the central bases is about −1 kcal/mol, which is reasonably similar to the literature value of about 0 kcal/mol for GC base pairs.26 Near the OMe surface, ΔGH‑bond is about −0.5 kcal/mol, while on the OEG surface, ΔGH‑bond is about −3.5 kcal/mol. Therefore, relative to bulk solution, the OEG surface stabilizes central WC H-bonds, while the OMe surface destabilizes these H-bonds, which is opposite to the effect on the terminal H-bonds. Stability of Base−Base Stacking. To show the effect of the surfaces on stacking, we construct 2D free-energy landscapes as a function of stacking and base-surface distance using the reweighting method (for OEG see Figure 3a and Figure 3c; for OMe see Figure S5).43 Stacking in Figure 3 is quantified similarly to the stacking CV used for metadynamics but for individual bases rather than all bases. For simplicity, only canonical stacking is included: for terminal bases, stacking can occur with one neighboring base; while for central bases, stacking can occur with two neighboring bases. Other details of the construction of Figure 3 are identical to Figure 2. We note that cytosine bases, due to their smaller size, are expected to stack less stably than guanine bases.46,50 Nonetheless, we average the results for cytosine and guanine together because the effect of base identity is small compared to the effects of location in the dsDNA (i.e., terminal vs central) and the effects of the surfaces. For terminal bases, these landscapes have two free-energy minima (Figure 3a), one near the surface (∼0.4 nm) and one far from the surface (∼1.4 nm). Both minima correspond to one stacking interaction because terminal bases have one canonical stacking partner. Notably, the landscape does not contain any minima corresponding to fully unstacked conformations, indicating that stacking of terminal bases is stable regardless of proximity to a surface. To compare the bulk, OEG, and OMe systems, we extract 1D free-energy profiles (Figure 3b) from the 2D free-energy landscapes at 0.4 nm base-
Figure 3. Effect of surfaces on base−base stacking. (a) Free-energy landscape of base−base stacking and base-surface distance. Data is for OEG; data for OMe is in SI Results (Figure S5). Contours are spaced by 1 kcal/mol. Dashed line is the location of the 1D free-energy profiles shown in (b). (b) Free energy as a function of base−base stacking for terminal bases in bulk solution and near (0.4 nm) the OEG and OMe surfaces. Error bars are the standard error of the mean. (c) Similar to (a), but for the central bases. (d) Similar to (b), but for the central bases at 1.1 nm surface distance.
surface distance, and we directly calculate a similar 1D freeenergy profile for the bulk system using the reweighting method.43 We define ΔGstack as the free-energy change from the unstacked state (stacking ≈ 0) to the stacked state (stacking ≈ 1), which is negative if stacking is favorable. In bulk solution, ΔGstack for terminal bases is about −2 kcal/mol, which is similar to past results that the stacking free energy in CG and GC ssDNA dimers is about −3 kcal/mol.46,47 Additionally, because stacking is the primary stabilizer of base−base interactions in dsDNA,26 our value of ΔGstack may be compared to overall base pairing free energy, which, similar to our result, is about −1.5 to −2.0 kcal/mol.26,29,48 We note that stacking stability depends strongly on solution conditions (e.g., temperature, salt concentration, DNA concentration) and nature of the DNA molecule in use (e.g., nucleobase vs nucleoside vs nucleotide, monomers vs oligomers).51 Overall, our value of the stacking free energy for terminal bases in bulk solution is in reasonable agreement with past results. Both the presence and chemistry of the surfaces affect ΔGstack, with values of about −2.5 kcal on OEG and about −1 kcal/mol on OMe (Figure 3b). Therefore, compared to bulk solution, stacking of the terminal bases is more stable near the hydrophilic OEG surface and less stable near the hydrophobic OMe surface. In addition to the 1D profiles at 0.4 nm surface distance (Figure 3b), we extracted profiles at a surface distance of 1.4 nm (SI Results, Figure S6), and ΔGstack shows the same trend as at 0.4 nm. Thus, the surfaces influence the stability of stacking even when bases are not in direct contact with the surface. The landscapes for the central bases (Figure 3c) show two minima, one global and one local. The global minimum corresponds to canonical stacking with both neighboring bases. By contrast, in the local minimum, central bases are near the D
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules
Figure 4. Effect of surfaces on base melting pathways. Free-energy landscape of WC H-bonds and base−base stacking for terminal (top: a−c) and central (bottom: e−g) bases in bulk (left: a, e), near OEG (middle: b, f), and near OMe (right: c, g). To create the landscapes for the near-surface systems, we constructed 3D free-energy landscapes with base-surface distance as the third CV and extract a 2D slice at 0.4 nm base-surface distance. Contours are spaced by 1 kcal/mol. The black line is the minimum free energy pathway (MFEP) between the hybridized and melted states, indicated by asterisks. Diagrams of the melting pathways are shown for terminal (d) and central (h) bases. The green solid arrows in the pathway denote unstacking and red dashed arrows denote H-bond disruption.
surface (∼0.4 nm) and have highly disrupted stacking interactions (stacking ≈ 0.3). The local minimum occurs on the OEG and OMe surfaces, suggesting that interactions between surfaces and nucleobases can destabilize stacking, which we discuss in the next section. To quantify the effect of the surfaces on stacking stability, we extract 1D free-energy profiles from the 2D free-energy landscapes at 1.1 nm surface distance, and construct a similar 1D free-energy profile for the bulk system using the reweighting method.43 These free-energy profiles (Figure 3d) show that the surfaces affect stacking in the central bases in the same way as the terminal bases. Overall, regardless of location within the dsDNA, stacking is more stable near the hydrophilic OEG surface and less stable near the hydrophobic OMe surface, even when bases are not in direct contact with the surface. These long-range effects on stacking suggest that surfaces induce long-range structural changes in dsDNA, which might occur, for example, directly through DNA−surface interactions or indirectly through surfaceinduced changes to the properties of water. We address these interactions in the following section. Interactions Affecting Stability of H-Bonds and Stacking. The greater stability of terminal H-bonds near OMe can be attributed to the effects of hydrophobic surfaces on water. The lower density of water,7 and the greater propensity for density fluctuations,52 allows adsorbing molecules to “dry” the hydrophobic interface. The dielectric constant is lowered in the dry region,53 strengthening electrostatic interactions such as H-bonding. The scarcity of water also reduces the number of competing H-bonding partners (Figure S7), thus, strengthening WC H-bonds. Conversely, the lower stability of terminal H-bonds near the hydrophilic OEG surface is due to higher water density near the surface than in bulk water,7 increasing competition by water molecules with WC H-bonds (Figure S7). Additionally, −OH groups in the OEG surface form H-bonds with DNA that may compete with WC H-bonds.7
The effect of the surfaces on WC H-bonds in central bases is opposite to their effect on the terminal H-bonds: OMe destabilizes central WC H-bonds, while OEG stabilizes these H-bonds. This reversal suggests that different mechanisms are dominant for the central WC H-bonds. While the mechanisms acting on the terminal bases (e.g., solvation) presumably affect the central bases, these mechanisms apparently have less influence (Figure S7), probably because the central bases are shielded from water and the near-surface environment by neighboring bases. Instead, stacking plays a greater role in the stability of the central WC H-bonds: stable stacking aligns bases in conformations that stabilize WC H-bonds. Thus, because the OEG surface stabilizes stacking for central bases, it also stabilizes the central WC H-bonds; the OMe surface has the opposite effect. The polar, well-hydrated OEG surface stabilizes stacking by disfavoring exposure (i.e., unstacking) of the relatively hydrophobic bases. Stacking is stabilized for bases not in direct contact with OEG, even outside the ∼1 nm dense hydration layer above the surface.7 This long-range stabilization is explained by the cooperative nature of stacking:54 stacking for bases near the OEG surface promotes stacking for neighboring bases farther from the surface. Although OEG provides interactions that would tend to destabilize stacking (Figure S8), such as van der Waals interactions and H-bonds,7 these interactions are apparently overwhelmed by the above stabilizing mechanisms. In contrast, the nonpolar, poorly hydrated environment near OMe destabilizes stacking. Furthermore, the OMe surface provides hydrophobic interactions that directly compete with stacking (Figure S8). Base Pair Melting Pathways. Next we characterize the effect of surfaces on the melting pathways for individual bases. Because WC H-bonds and stacking are the two main contributions to dsDNA stability,26 and because CVs related to H-bonds and stacking have been used to describe the melting pathway of a terminal base,48 we use these two CVs to E
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules
influence the rate of hybridization. The overall process of hybridization can be described as a two-step process: a slow step, bimolecular nucleation of a few consecutive base pairs, followed by a fast step, zipping of the remaining base pairs.58 The rate of nucleation is dominated by diffusion and geometric factors,58 which are independent of base-pairing kinetics, but the rate of zipping should be influenced by base-pairing kinetics. Zipping involves multiple intermediate structures, as observed recently for a DNA hairpin,60 and base-pairing kinetics will influence the transition rates between these structures. Additionally, DNA melting should be influenced by base-pairing kinetics, since melting occurs through base-pair breathing and the nucleation of melted “bubbles”.26 Thus, it seems likely that the overall kinetics of hybridization will be affected by base-pairing kinetics. The heights of base-pairing barriers depend on the direction of travel along the MFEP, so we separately calculate the barriers to hybridization, ΔG‡h, and to melting, ΔG‡m (Table 1).
characterize the disruption of base−base interactions (i.e., base melting). To this end, we construct free-energy landscapes as a function of WC H-bonds and stacking for terminal and central bases in bulk, near OEG, and near OMe (Figure 4). We identify the pathway of destabilization of base−base interactions by calculating the minimum free energy pathway (MFEP) between the hybridized and melted states using the finite-temperature string (FTS) method.55 Details of the FTS calculations are available in SI Methods. The landscapes for the terminal bases (Figure 4, top) exhibit three minima separated by kinetic barriers. These minima correspond to the fully hybridized state (H-bonds ≈ 3 and stacking ≈ 1), a partially melted state with broken H-bonds (Hbonds ≈ 0 and stacking ≈ 1), and the fully melted state (Hbonds ≈ 0 and stacking ≈ 0). We emphasize that we are using the term “melting” to mean the disruption of H-bonds and stacking for individual bases rather than the dehybridization of two DNA strands; an individual base could be in the “partially melted” state in fully melted ssDNA. For bulk, OEG, and OMe, the MFEPs (i.e., black dotted lines in Figure 4) show that melting of terminal bases proceeds in a sequential fashion by first breaking the H-bonds, after which unstacking from the neighboring base can occur (see diagram in Figure 4d). This result confirms that stacking with the neighboring base prealigns terminal bases to form WC H-bonds, and, conversely, that destacking of terminal bases obstructs reformation of the WC H-bonds. This finding is in accord with Hagan et al.,48 who found that a terminal base in bulk solution must unstack to prevent easy reformation of the WC H-bonds. Similarly, Wong and Pettitt found that fraying and unstacking of the terminal bases in bulk allows the DNA strands to peel apart.56 Here our new results demonstrate that the pathway for terminal base melting is unaffected by the surfaces. The melting pathways for central bases (Figure 4, bottom) differ from those of the terminal bases. Starting from the fully hybridized state (H-bonds ≈ 3 and stacking ≈ 2), the MFEP proceeds to a transition region (H-bonds ≈ 1.5 and stacking ≈ 1), which shows that melting of central bases initially occurs through partial disruption of H-bonds and unstacking from one neighboring base (see diagram in Figure 4h). For the bulk and OMe systems, H-bonds are disrupted first and unstacking occurs second, which is similar to the sequential pathway followed by the terminal bases. In contrast, for the OEG system, unstacking occurs first and H-bonds are disrupted second, likely because central WC H-bonds are more stable near OEG. Thus, surfaces can qualitatively affect the pathway for melting. From the transition region, melting proceeds first through complete disruption of the H-bonds, followed by unstacking from the second neighboring base. Similar to the terminal bases, the pathways for the central bases indicate that prealignment by stacking with neighboring bases favors the formation of WC H-bonds. Furthermore, progressive unstacking and disruption of H-bonds in neighboring bases allows central bases to melt. Kinetic Barriers to Base Pairing. To examine the kinetic barriers to base pair hybridization and melting, we extract the free energy along the MFEPs (SI, Figure S9). We focus on the largest barrier along the MFEPs because, as the ratedetermining step in the framework of transition-state theory,57 it has the greatest impact on base-pairing kinetics. We note that the hybridization of two ssDNA strands into dsDNA exhibits non-Arrhenius behavior58,59 and cannot be described by a single kinetic barrier, but barriers to base-pairing are likely to
Table 1. Kinetic Barriers to Melting, ΔG‡m, and Hybridization, ΔG‡h, for Terminal and Central Basesa terminal bases
a
central bases
system
ΔG‡m
ΔG‡h
ΔG‡m
ΔG‡h
bulk OEG OMe
2.5 ± 0.3 3.1 ± 0.4 4.0 ± 0.3
3.3 ± 0.2 4.9 ± 0.4 3.5 ± 0.4
6.0 ± 0.3 6.0 ± 0.6 4.3 ± 0.4
3.2 ± 0.4 3.0 ± 0.6 3.3 ± 0.5
Units are kcal/mol. Error bars are the standard error of the mean.
For terminal bases, ΔG‡m is larger on OMe than in bulk or on OEG, while ΔG‡h is larger on OEG than in bulk or on OMe. These results suggest that hydrophobic surfaces cause terminal base pairs to form more rapidly and to melt more slowly relative to hydrophilic surfaces. These findings will aid the interpretation of experiments on DNA self-assembly. For example, these results are consistent with findings on hairpin hybridization kinetics, where it was found that a hydrophobic surface yielded faster folding and slower unfolding than a hydrophilic surface.4 It is reasonable that hybridization and melting rates of terminal base pairs would significantly influence the overall kinetics of hairpin folding, because terminal base pairs act as nucleation points for hybridization or melting.56,60 For central bases, ΔG‡m is lower on OMe than in bulk or on OEG, while ΔG‡h is similar for all three systems. These results suggest that central bases melt more rapidly near OMe than in bulk or near OEG, whereas central bases hybridize at a similar rate regardless of the presence or chemistry of the surfaces. The OMe surface reduces the barrier to melting likely by destabilizing stacking through direct hydrophobic interactions, which is similar to recent findings that attraction to a gold nanoparticle accelerates DNA melting.61 However, our finding of faster melting on the hydrophobic OMe surface is at odds with findings that ssDNA hairpins melt more slowly on a hydrophobic surface than on a hydrophilic surface.4 As an explanation, we speculate that the slower base melting we observe on the hydrophilic OEG surface kinetically traps the DNA in misfolded structures, as suggested previously.60 Conversely, faster base melting on the hydrophobic OMe surface allows more rapid exploration of conformational space, increasing the rate of folding into the correct hairpin structure. F
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules
■
CONCLUSION We have elucidated the effects of hydrophilic and hydrophobic surfaces on the stability of individual base−base interactions (i.e., Watson Crick H-bonds, stacking), from which we can infer the effects on overall duplex stability. Hybridized short dsDNA, containing a significant fraction of terminal bases, will be stabilized by hydrophobic surfaces and destabilized by hydrophilic surfaces. In contrast, hybridized long dsDNA, containing a large fraction of central bases, will be destabilized by hydrophobic surfaces and stabilized by hydrophilic surfaces as central WC H-bonds and base stacking are destabilized by the hydrophobic surface. The combined effect will depend on the relative importance of terminal and central bases to hybridization and melting, for example, the stability of terminal bases will be more important for hairpins, where the terminal base pair serves as a nucleation point for unzipping.60 The concentration of free DNA may also play a role, for example, high concentrations of free DNA may enhance the effect of the surfaces on DNA stability or drive hybridization to completion regardless of surface chemistry.62 This study serves as a step toward further studies that elucidate these effects of DNA concentration and crowding on DNA−surface interactions. Surface effects likely also depend on whether the dsDNA is free in solution, as in this study, or is grafted to the surface, as in other experimental and simulation studies.8−19 For example, on gold surfaces directly adsorbed (free) dsDNA is less stable than end-grafted dsDNA,25 indicating an interplay between surface interactions, grafting, and stability. One limitation to the above inferences on overall dsDNA stability is that the 4-base-pair oligomer studied here is much shorter than typical dsDNA. As a consequence, the central base pairs in our short dsDNA oligomer are subject to end-effects that arise from the instability of the terminal base pairs. Endeffects can be reduced by using at least a two-base-pair cap or, preferably, a full helical turn (10 base pairs),63,64 but only a onebase-pair cap is present in our oligomer. As such, the behavior of the central bases in this study may not be truly representative of the behavior of central bases within longer dsDNA, which, due to the cooperativity of stacking and WC H-bonds, will likely have more stable base−base interactions and larger barriers to melting. Despite this limitation, we do find differing results for terminal and central bases, even in our short dsDNA oligomer, which suggests that the central bases in this oligomer have at least some of the character of central bases in a longer dsDNA polymer. We have identified the pathways by which base melting takes place on hydrophilic and hydrophobic surfaces. For terminal bases, melting proceeds first through breaking H-bonds and then by unstacking from the neighboring base, regardless of the presence or chemistry of the surfaces. For the central bases in bulk or near the hydrophobic surface, melting proceeds in a stepwise fashion, alternating between disruptions of H-bonds followed by unstacking from neighboring bases. The pathway differs on the hydrophilic surface, where unstacking from one neighboring base precedes complete disruption of the H-bonds, followed by unstacking from the second neighboring base. The surfaces also affect the kinetic barriers to base melting and hybridization. In particular, our results suggest that central bases melt more rapidly near hydrophobic surfaces, which may accelerate conformational searching and thereby increase the rate at which misfolded DNA structures (e.g., origami) reach the correctly folded conformation. These findings may aid the
design of surfaces that stabilize self-assembled DNA nanostructures.
■
ASSOCIATED CONTENT
S Supporting Information *
Methods: additional details of our equilibration and simulation protocols, our use of the PTMetaD-WTE method, the convergence of the PTMetaD-WTE simulations, and the implementation of the finite-temperature string method. Results: additional details of the stability of WC H-bonds and stacking, quantification of the interactions that affect H-bonds and stacking, and the free energy along the minimum free energy pathways. References: 65−77. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.biomac.5b00469.
■
AUTHOR INFORMATION
Corresponding Authors
*E-mail:
[email protected]. Phone: (410) 306 1926. *E-mail:
[email protected]. Phone: (302) 831 8682. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS R.M.E. was supported by an appointment to the Postgraduate Research Participation Program at the U.S. Army Research Laboratory administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the U.S. Department of Energy and USARL. J.P. acknowledges financial support from NSF award CBET-1264459. R.M.E and A.J. acknowledge financial support from NSF DMR Biomaterials Grant DMR-1206894 and supercomputing time through XSEDE (TG-MCB100140). This work used the Janus supercomputer, which is supported by the NSF (CNS0821794) and University of Colorado−Boulder. A.J. thanks J.J. de Pablo for useful scientific comments.
■
REFERENCES
(1) Rothemund, P. W. Nature 2006, 440, 297−302. (2) Sun, X.; Hyeon, K. S.; Zhang, C.; Ribbe, A. E.; Mao, C. J. Am. Chem. Soc. 2009, 131, 13248−13249. (3) Saccà, B.; Niemeyer, C. M. Angew. Chem., Int. Ed. 2012, 51, 58− 66. (4) Kastantin, M.; Schwartz, D. K. Small 2013, 9, 933−941. (5) Gong, P.; Levicky, R. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 5301−6. (6) Elder, R. M.; Jayaraman, A. J. Chem. Phys. 2014, 140, 155103. (7) Elder, R. M.; Jayaraman, A. Soft Matter 2013, 9, 11521−11533. (8) Seifpour, A.; Dahl, S. R.; Lin, B.; Jayaraman, A. Mol. Sim. 2013, 39, 741−753. (9) Stoughton, R. B. Annu. Rev. Biochem. 2005, 74, 53−82. (10) Park, S. Y.; Lee, J. S.; Georganopoulou, D.; Mirkin, C. A.; Schatz, G. C. J. Phys. Chem. B 2006, 110, 12673−12681. (11) Park, S. Y.; Lytton-Jean, K. R. A.; Lee, B.; Weigand, S.; Schatz, G. C.; Mirkin, C. A. Nature 2008, 451, 553−556. (12) Noh, H.; Hung, A. M.; Cha, J. N. Small 2011, 7, 3021−3025. (13) Striolo, A. Small 2007, 3, 628−635. (14) Vainrub, A.; Pettitt, B. M. J. Phys. Chem. B 2011, 115, 13300− 13303. (15) Ambia-Garrido, J.; Vainrub, A.; Pettitt, B. M. Comput. Phys. Commun. 2010, 181, 2001−2007. (16) Wang, Z.; Zeng, X.; Deng, Y.; He, N.; Wang, Q.; Huang, J. J. Nanosci. Nanotechnol. 2011, 11, 8457−8468. (17) Ozel, A. B.; Srivannavit, O.; Rouillard, J.-M.; Gulari, E. Biotechnol. Prog. 2012, 28, 556−566. G
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
Article
Biomacromolecules
(54) Alhambra, C.; Luque, F. J.; Gago, F.; Orozco, M. J. Phys. Chem. B 1997, 101, 3846−3853. (55) Vanden-Eijnden, E.; Venturoli, M. J. Chem. Phys. 2009, 130, 194103. (56) Wong, K.-Y.; Pettitt, B. M. Biophys. J. 2008, 95, 5618−5626. (57) Hammes, G. G. Physical Chemistry for the Biological Sciences; Wiley-Interscience: Hoboken, NJ, 2007. (58) Hinckley, D. M.; Lequieu, J. P.; de Pablo, J. J. J. Chem. Phys. 2014, 141, 035102. (59) Ouldridge, T. E.; Šulc, P.; Romano, F.; Doye, J. P. K.; Louis, A. A. Nucleic Acids Res. 2013, 41, 8886−8895. (60) Bowman, G. R.; Huang, X.; Yao, Y.; Sun, J.; Carlsson, G.; Guibas, L. J.; Pande, V. S. J. Am. Chem. Soc. 2008, 130, 9676−9678. (61) Sedighi, A.; Li, P. C. H.; Pekcevik, I. C.; Gates, B. D. ACS Nano 2014, 8, 6765−6777. (62) Lei, Q.-l.; Ren, C.-l.; Su, X.-h.; Ma, Y.-q. Sci. Rep. 2015, 5, 9217. (63) Pasi, M.; Maddocks, J. H.; Beveridge, D.; Bishop, T. C.; Case, D. A.; Cheatham, T.; Dans, P. D.; Jayaram, B.; Lankas, F.; Laughton, C.; Mitchell, J.; Osman, R.; Orozco, M.; Pérez, A.; Petkevičiutė ̅ , D.; Spackova, N.; Sponer, J.; Zakrzewska, K.; Lavery, R. Nucleic Acids Res. 2014, 42, 12272−12283. (64) Pasi, M.; Maddocks, J. H.; Lavery, R. Nucleic Acids Res. 2015, 43, 2412−2423. (65) Barducci, A.; Bonomi, M.; Parrinello, M. Biophys. J. 2010, 98, L44−L46. (66) Barducci, A.; Bussi, G.; Parrinello, M. Phys. Rev. Lett. 2008, 100, 020603. (67) Bussi, G.; Donadio, D.; Parrinello, M. J. Chem. Phys. 2007, 126, 014101. (68) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K. M.; Roberts, B.; Hayik, S.; Roitberg, A.; Seabra, G.; Swails, J.; Götz, A. W.; Kolossváry, I.; Wong, K. F.; Paesani, F.; Vanicek, J.; Wolf, R. M.; Liu, J.; Wu, X.; Brozell, S. R.; Steinbrecher, T.; Gohlke, H.; Cai, Q.; Ye, X.; Wang, J.; Hsieh, M.-J.; Cui, G.; Roe, D. R.; Mathews, D. H.; Seetin, M. G.; Salomon-Ferrer, R.; Sagui, C.; Babin, V.; Luchko, T.; Gusarov, S.; Kovalenko, A.; Kollman, P. A. AMBER 13; University of California: San Francisco, 2012. (69) Elder, R. M.; Jayaraman, A. Biophys. J. 2012, 102, 2331−2338. (70) Hao, M.-H. J. Chem. Theory Comput. 2006, 2, 863−872. (71) Hess, B. J. Chem. Theory Comput. 2008, 4, 116−122. (72) Jakalian, A.; Jack, D. B.; Bayly, C. I. J. Comput. Chem. 2002, 23, 1623−1641. (73) 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, 1781−1802. (74) Sousa da Silva, A. W.; Vranken, W. F. BMC Res. Notes 2012, 5, 1−8. (75) Szatyłowicz, H.; Sadlej-Sosnowska, N. J. Chem. Inf. Model. 2010, 50, 2151−2161. (76) Thomas, J. M.; David, A. C., Modeling unusual nucleic acid structures. Molecular Modeling of Nucleic Acids; American Chemical Society: Washington, DC, 1997; Vol. 682, pp 379−393. (77) Ulman, A. Chem. Rev. 1996, 96, 1533−1554.
(18) Schmitt, T. J.; Knotts, T. A. J. Chem. Phys. 2011, 134, 205105. (19) Schmitt, T. J.; Rogers, J. B.; Knotts, T. A. J. Chem. Phys. 2013, 138, 035102. (20) Welling, R. C.; Knotts, T. A. J. Chem. Phys. 2015, 142, 015102. (21) Cruz, F. J. A. L.; de Pablo, J. J.; Mota, J. P. B. RSC Adv. 2014, 4, 1310−1321. (22) Cruz, F. J. A. L.; de Pablo, J. J.; Mota, J. P. B. J. Chem. Phys. 2014, 140, 225103. (23) Peterson, A. W.; Wolf, L. K.; Georgiadis, R. M. J. Am. Chem. Soc. 2002, 124, 14601−14607. (24) Watkins, H. M.; Vallée-Bélisle, A.; Ricci, F.; Makarov, D. E.; Plaxco, K. W. J. Am. Chem. Soc. 2012, 134, 2120−2126. (25) Schreiner, S. M.; Hatch, A. L.; Shudy, D. F.; Howard, D. R.; Howell, C.; Zhao, J.; Koelsch, P.; Zharnikov, M.; Petrovykh, D. Y.; Opdahl, A. Anal. Chem. 2011, 83, 4288−4295. (26) Yakovchuk, P.; Protozanova, E.; Frank-Kamenetskii, M. D. Nucleic Acids Res. 2006, 34, 564−574. (27) Kool, E. T. Annu. Rev. Biophys. Biomol. Struct. 2001, 30, 1−22. (28) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. J. Chem. Phys. 1983, 79, 926−935. (29) Bloomfield, V. A.; Crothers, D. M.; Tinoco, I. J. Nucleic Acids: Structures, Properties, and Functions; University Science Books: Sausilito, CA, 2000. (30) Pérez, A.; Marchán, I.; Svozil, D.; Sponer, J.; Cheatham, I.; Thomas, E.; Laughton, C. A.; Orozco, M. Biophys. J. 2007, 92, 3817− 3829. (31) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179−5197. (32) Joung, I. S.; Cheatham, T. E., III J. Phys. Chem. B 2008, 112, 9020−9041. (33) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. J. Comput. Chem. 2004, 25, 1157−1174. (34) Wang, J.; Wang, W.; Kollman, P. A.; Case, D. A. J. Mol. Graphics Model. 2006, 25, 247−260. (35) Hess, B.; Kutzner, C.; Van Der Spoel, D.; Lindahl, E. J. Chem. Theory Comput. 2008, 4, 435−447. (36) Tribello, G. A.; Bonomi, M.; Branduardi, D.; Camilloni, C.; Bussi, G. Comput. Phys. Commun. 2014, 185, 604−613. (37) Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graphics 1996, 14, 33−38. (38) Kannan, S.; Zacharias, M. Phys. Chem. Chem. Phys. 2009, 11, 10589−10595. (39) Deighan, M.; Bonomi, M.; Pfaendtner, J. J. Chem. Theory Comput. 2012, 8, 2189−2192. (40) Bonomi, M.; Parrinello, M. Phys. Rev. Lett. 2010, 104, 190601. (41) Deighan, M.; Pfaendtner, J. Langmuir 2013, 29, 7999−8009. (42) Laio, A.; Parrinello, M. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 12562−12566. (43) Bonomi, M.; Barducci, A.; Parrinello, M. J. Comput. Chem. 2009, 30, 1615−1621. (44) Boresch, S.; Karplus, M. J. Chem. Phys. 1996, 105, 5145−5154. (45) Gilson, M. K.; Given, J. A.; Bush, B. L.; McCammon, J. A. Biophys. J. 1997, 72, 1047−1069. (46) Murata, K.; Sugita, Y.; Okamoto, Y. Chem. Phys. Lett. 2004, 385, 1−7. (47) Norberg, J.; Nilsson, L. Biophys. J. 1995, 69, 2277−2285. (48) Hagan, M. F.; Dinner, A. R.; Chandler, D.; Chakraborty, A. K. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 13922−13927. (49) Pohorille, A.; Ross, W. S.; Tinoco, I., Jr. Int. J. Supercomput. Appl. 1990, 4, 81−96. (50) Norberg, J.; Nilsson, L. J. Am. Chem. Soc. 1995, 117, 10832− 10840. (51) Friedman, R. A.; Honig, B. Biophys. J. 1995, 69, 1528−1535. (52) Jamadagni, S. N.; Godawat, R.; Dordick, J. S.; Garde, S. J. Phys. Chem. B 2008, 113, 4093−4101. (53) Uematsu, M.; Franck, E. U. J. Chem. Phys. Ref. Data 1980, 9, 1291−1306. H
DOI: 10.1021/acs.biomac.5b00469 Biomacromolecules XXXX, XXX, XXX−XXX
