Busra
Demir‡
abc,
Hashem
Mohammad‡
d,
M. P.
Anantram
c and
Ersin Emre
Oren
*ab
aDepartment of Materials Science & Nanotechnology Engineering, TOBB University of Economics and Technology, Ankara, Turkey. E-mail: eeoren@gmail.com
bBionanodesign Laboratory, Department of Biomedical Engineering, TOBB University of Economics and Technology, Ankara, Turkey
cDepartment of Electrical and Computer Engineering, University of Washington, 98195 Seattle, WA, USA
dDepartment of Electrical Engineering, Kuwait University, P. O. Box 5969, Safat 13060, Kuwait
First published on 19th May 2023
DNA's charge transfer and self-assembly characteristics have made it a hallmark of molecular electronics for the past two decades. A fast and efficient charge transfer mechanism with programmable properties using DNA nanostructures is required for DNA-based nanoelectronic applications and devices. The ability to integrate DNA with inorganic substrates becomes critical in this process. Such integrations may affect the conformation of DNA, altering its charge transport properties. Thus, using molecular dynamics simulations and first-principles calculations in conjunction with Green's function approach, we explore the impact of the Au (111) substrate on the conformation of DNA and analyze its effect on the charge transport. Our results indicate that DNA sequence, leading to its molecular conformation on the Au substrate, is critical to engineer charge transport properties. We demonstrate that DNA fluctuates on a gold substrate, sampling various distinct conformations over time. The energy levels, spatial locations of molecular orbitals and the DNA/Au contact atoms can differ between these distinct conformations. Depending on the sequence, at the HOMO, the charge transmission differs up to 60 times between the top ten conformations. We demonstrate that the relative positions of the nucleobases are critical in determining the conformations and the coupling between orbitals. We anticipate that these results can be extended to other inorganic surfaces and pave the way for understanding DNA–inorganic interface interactions for future DNA-based electronic device applications.
Controllable conductivity, mechanical integrity, and structural stability are among the critical elements that molecular nanodevices should possess. DNA, like other biomolecules, fluctuates and interacts with its surrounding environment. Several studies have been conducted to understand the relationship between conformational changes and the charge transmission along DNA molecules. Woiczikowski et al.21 reported that the effect of structural fluctuations can be different based on the DNA's sequence and the base pair dynamics are important in determining the charge transport properties. Artés et al.22 demonstrated that the conductance of DNA molecules increases by nearly one order of magnitude when their conformation is changed from the B-form to the A-form. Bruot et al.23 outlined that conductivity is significantly responsive to mechanical stretching and has a weak dependence on the DNA's length. Saientan Bag et al.14 noted that the pulling direction plays a significant role in changing the conformation of DNA and therefore the conductance.
DNA is subjected to conformational changes when it is in contact with a substrate. Since DNA would be in interaction with a substrate in DNA-based nanodevices, the conformational changes due to substrate interactions would be a key element in determining the electronic properties of DNA-based nanodevices. Although great efforts have been made to understand the effect of structural fluctuations on the electronic properties as mentioned above, the effect of structural changes due to an external substrate on the electronic properties remains an open subject. Depending on the sequence, length, and environmental conditions, the interactions between biomolecules and inorganic substrates can change. This affects the number of interacting atoms and creates a variety of interfaces. This should in turn impact both the charge couplings between DNA and substrate atoms and between the bases. Therefore, it is important to explore the variation in conductance values depending on the conformation on the substrate.
In this study, we examine the electronic properties of DNA when in contact with an Au substrate. The Au substrate is chosen due to its chemical inertness and biocompatibility, making it a great candidate for bio-electronic and bio-medical applications. We focus on three different DNA sequences where GC base pairs are separated by AT base pairs. Then, we conceptualize a setup where DNA molecules lay horizontally on a substrate and make electrical contact between the central AT base pairs and the underlying gold atoms. Our motivation for horizontal DNA molecules comes from DNA origami tiles which can self-assemble on a surface and form long, repeating DNA wires,24–27 which could be used in a circuit.
First, we model DNA molecules on top of the Au (111) surface via molecular dynamics (MD) simulations which have been proven to be an effective tool to investigate the DNA–gold interactions.28,29 Our calculations agree with prior work which finds several different conformations due to interactions with the gold surface. Then, we cluster the resulting conformations and use the top ten groups for the quantum mechanical calculations. We employ density functional theory (DFT) calculations to study the electronic properties of each selected DNA conformation. We present that the energy and spatial distributions of the molecular orbitals vary with sequence and molecular conformation. Next, we use Green's function method to perform charge transport calculations by considering the alterations in contact atoms caused by conformational changes. Our results demonstrate that the transmission of the HOMO energy varies by up to 60 times between different conformations.
This paper is organized as follows: first, we examine the structural changes of DNA lying on top of the Au (111) substrate. Subsequently, we present a comprehensive analysis of differences in conformations of each DNA sequence. This is followed by an in-depth investigation of the energy levels and molecular orbitals of each conformation. Then, we report the charge transport calculations through DNA aiming to elucidate the effect of the sequence and conformation on its electrical properties.
To reduce the computational time required for the MD simulations, we place the DNA structures 3.5 Å away from the surface. Then we solvate the whole system with 0.15 M KCl. We use the TIP3P forcefield30 to model water molecules and CHARMM3631 and INTERFACE32 force fields for the DNA and gold substrate, respectively. It is known that when a gold surface is in contact with water, surfactants, and biopolymers, an attractive polarization takes place at the interface33 due to highly mobile electrons. Therefore, while modeling the interactions between DNA and gold surfaces, it is important to include polarization effects on the surfaces. We use the INTERFACE force field, which has been generated using experimental material properties like density and surface energy as target data to fit the Lennard Jones (LJ) atomic radii and well depth, respectively.32 Consequently, in our simulations we only include van der Waals forces, and we neglect further Coulomb interactions between the DNA molecules and the gold substrate. We simulate the DNA–Au (111) systems for 50 ns in the NPT ensemble using the NAMD34 program. During the simulations, we keep the location of the Au atoms fixed and neglect the reconstruction of the surface for simplicity. We provide a comprehensive description of the simulation procedure in the Method section in the ESI.†
We first assess the global conformational changes of DNA upon placement on the surface using the conventional measure of the root-mean-square deviation (RMSD). The RMSD increases during the first 20 ns of the trajectory before reaching a plateau for C3T3C3 and G3A3G3 while it takes almost 40 ns for C3A3C3 to reach a plateau as shown in Fig. 2. We observe that the final RMSD values converged to ∼5 Å indicating that significant structural changes occurred for all three DNA sequences considered. Thereafter, we focus on the Root Mean Square Fluctuation (RMSF) for each individual atom to reveal the regions of the DNA structures that are most flexible (Fig. S2, ESI†). RMSF plots show that the DNA molecules exhibit significant mobility in the terminal region of the structure in all cases. On the other hand, the central AT region for each case stays stable during the 50 ns simulation. Therefore, we anticipate that with an increase in the simulation time, the terminal regions will continue to fluctuate, causing more conformational changes to the molecule, while the central AT-rich region will remain stable for an extended period. Since our focus is on the electronic properties of DNA and understanding the effects of conformational changes, rather than identifying the most thermodynamically stable conformation of DNA on a gold substrate, we restrict the simulation time to 50 ns. However, we acknowledge that the conformation of such short DNA sequences may exhibit greater conformational changes when the simulation time is extended.
Fig. 2 RMSD change within 50 ns of MD simulation together with the RMSD histogram which shows RMSD mean and variation for each sequence. |
Next, we construct pairwise RMSD plots to monitor the conformational changes with time for both (i) the whole DNA structure and (ii) the central AT region. Fig. S3 (ESI†) illustrates the temporal evolution of the molecule's conformation obtained by comparing the RMSD change between the conformations saved every 1 ns. The right column of Fig. S3 (ESI†) shows that the central AT-region is stable and displays minimal conformational variability. We also observe the stability of this region in the RMSF plots as previously mentioned. Conversely, the terminal regions cause the whole structure to exhibit clusters of distinct conformations (see the left column of Fig. S3, ESI†). Therefore, we use a RMSD-based clustering algorithm35 and categorize the conformations of the whole DNA with a cutoff value of 1.5 Å RMSD. Then, we select the top ten clustered groups for further analysis. We consider the conformations that do not fit any of the top ten groups as un-clustered. Table S1 (ESI†) presents the number of conformations and their percentages for each cluster.
It is important to quantify the nature of the interaction between the DNA strand and the substrate as this results in the various transmission/conductance features which are also previously reported in the literature.15,36,37 From each of the ten different clusters, we choose the conformation that has the minimum RMSD difference from all others in the same cluster—we call this the representative structure. Because we only take a snapshot from MD simulations, depending on where we are in the conformational space, we may end up with a conformation that is slightly away from the local energy minimum. Therefore, we apply energy minimization to all thirty representative structures before DFT calculations. Fig. S4 (ESI†) illustrates all these structures. To identify the conformational differences, we initially calculate the percentage of contact atoms, which were defined as the DNA atoms within 5 Å of the Au (111) surface. Our observations indicate that, in all sequences, at least one nucleobase from each AAA extension contacts the Au (111) substrate. In the C3T3C3 sequence, two of the Adenines are fully adsorbed for most of the representative structures, while in C3A3C3 and G3A3G3 it decreases to one. Furthermore, we do not observe 100% interaction with the gold substrate in any nucleobase other than AAA extensions. We also note in the complementary strand lacking the AAA extension region, the nucleobases close to the GC-AT boundaries interact with the substrate for each cluster. Although these interactions depend on the initial placement of the DNA molecules on the substrate, our results indicate that the relative positions of nucleobases result in different conformations on the surface.
Subsequently, we compare the hydrogen bonds between the base pairs for each conformation (Fig. S5, ESI†). Notably, while the representative structures displayed various percentages of contact atoms, we see comparable numbers of hydrogen bonds across all representative structures. This indicates that the terminal AAA extensions provide base pair integrity for the dsDNA part of the sequences during the simulation time.
Next, we change our attention towards investigating the puckering angle since A- and B-form DNAs exhibit distinct conductance values. Fig. S6 (ESI†) demonstrates the sugar puckering angle for all representative structures, thereby shedding light on the potential DNA form changes. Interestingly, we note that most of the Ts display a pucker angle close to that of the A-form, rather than the B-form, while the rest of the structure stays close to the B-form sugar puckering.
In the next section, we study the electronic properties of each representative structure, which reveals that conformational changes induced by substrate interactions play an important role in determining the transmission.
First, we examine the effect of conformational changes on the energy levels of the molecular orbitals (MO). Fig. S7 (ESI†) illustrates the first 11 orbitals from both occupied (HOMO−10 to HOMO) and unoccupied (LUMO to LUMO+10) states. We note the difference between the highest and lowest HOMO levels is 155 meV for C3A3C3, 115 meV for C3T3C3, and 153 meV for G3A3G3. We find that C3A3C3 exhibits the highest HOMO energy difference between the representative structures. Furthermore, we also observe that C3A3C3 displays less variation in the energy of the HOMO levels in comparison to the other two sequences.
Furthermore, we observe that HOMO levels are generally higher for C3T3C3 and G3A3G3: 70% of the C3T3C3 and G3A3G3 conformations have a HOMO level above −5 eV, while it is only 20% for C3A3C3. Moreover, although there are shifts in the orbital energy levels, the band gaps of all representative structures lie within the 3.8–4.2 eV range (Fig. S8, ESI†). This clearly indicates that while the content of each sequence is the same, energy levels differ due to conformational changes.
Later, we analyze the spatial distribution of MO along the molecules. Fig. S9 (ESI†) represents the first 4 occupied MOs from each representative structure. For all cases, occupied orbitals are either located on the rightmost or leftmost guanines. However, depending on the conformation, the localization of orbitals changes; for instance, while in C3A3C3 C1 (Cluster 1), the HOMO and HOMO−4 are delocalized on the same side, in C2 they are on the opposite guanines. As a result, it is reasonable to expect sample-to-sample variations in the transmission values.
To calculate the transmission, we consider a conceptual scenario where the DNA on the Au (111) substrate makes a contact from the top of the DNA molecule as in STM and/or AFM measurements (Fig. 3a). We specifically place the top contact on the backbone atoms of the central AT base pair to ensure that the most stable portion of the structure is probed. We chose this configuration based on the MD simulation results which reveal that the central region is the most stable portion of the structure. Here, the top contact can be thought of as a gold tip (Fig. 3a), different interacting molecules, or covalently bonded functionalization groups. As for the bottom contact, we select the contact atoms that we previously defined in Fig. S5 (ESI†). It should be noted that as the conformation changes, so does the number of contact atoms (Fig. S5 (ESI†), Fig. 4).
Fig. 4 Bar plots showing the relation between time points, the number of contacts and transmission at the HOMO and HOMO+0.5 for the C3T3C3, C3A3C3 and G3A3G3 sequences. |
As mentioned earlier, we do not include explicit Au atoms in the DFT calculations due to computational limitations. Therefore, in modeling the charge transport properties along the DNA molecules, we need open boundary conditions resulting from the top and the bottom contact. To model charge transport including these open boundary conditions, we employ Green's function technique, which includes self-energy terms corresponding to each contact. Details about the model and calculation parameters are given in the Methods section of the ESI.†
Fig. 3b demonstrates the transmission plots for the ten conformations corresponding to each sequence. We shift the energy axis to make the HOMO energy of all representative structures to be at E = 0. First, we notice that for a given sequence, the transmission can vary by order of magnitude at the HOMO. The conformations which have the highest and the lowest transmission, as well as the variance between them, depend on the sequence. For instance, at the HOMO, while the transmission difference between C4 and C6 of C3T3C3 reaches up to 23 times, between C4 and C10 of C3A3C3, it is 60 times and between C6 and C8 of G3A3G3, it is 24 times. Inside the band gap, these differences reach up to 3 orders of magnitude due to the broadening of energy levels near the HOMO/LUMO edges, which increases the density of states inside the bandgap. Furthermore, we observe that the highest transmission does not always correspond to C1, which is the most observed conformation in the simulations.
Next, we analyze the relation between each representative structure and the number of contacts, transmission at the HOMO and the band gap by plotting the results based on time intervals that the representative structures correspond to (Fig. 4). The transmission at the band gap is selected from 0.5 eV away from the HOMO. The bar plots given in Fig. 4 indicate that the number of contacts increases as the simulation progresses in time, and this influences the transmission at the band gap. On the other hand, the transmission at the HOMO does not demonstrate one to one correlation with the number of contacts. We think that this is related to the spatial distribution of the HOMOs (Fig. S9, ESI†), i.e., the HOMO transmission values are highly dominated by the HOMO only, but the transmission values at the band gap are the collective effect of the other molecular orbitals as well.
All the results indicate that changing the position of base pairs affects (1) adsorption onto a gold substrate, (2) conformation, (3) the number of contact atoms, and (4) energy levels and spatial locations of the molecular orbitals. The cumulative effects of all these change the transmission of DNA molecules. While designing a DNA nanowire that is going to be used with a gold substrate, it is important to consider all these effects. Our results also indicate that it is challenging to predict transmission values just with sequence selection, since there will be different conformations.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2cp05009a |
‡ These authors contributed equally. |
This journal is © the Owner Societies 2023 |