The Electronic Properties of DNA Bases.
Mingsheng Xu,Robert G. Endres,Yasithiko Arakawa
DOI: https://doi.org/10.1002/smll.200600732
IF: 13.3
2007-01-01
Small
Abstract:Base level: The electronic properties of the four DNA nucleosides are studied by ultrahigh-vacuum scanning tunneling techniques (see image), and the results are compared with first-principles density functional calculations of the methylated bases. Base-specific electronic signatures are observed, such as molecular energy levels and currents, which shed light on charge injection into and migration along DNA molecules and electronic DNA sequencing. DNA, the blueprint of life, has recently attracted much interest for its potential application in DNA-based nanoelectronics1–3 and due to the search for alternative DNA sequencing methods.4 Prior to realizing such applications, the electronic properties of the DNA bases need to be fully investigated. Scanning tunneling microscopy (STM) studies showed that DNA molecules or its bases, namely, guanine (G), adenine (A), cytosine (C), and thymine (T), self-assemble into various structures on different substrates.5–19 Recently, Ohshiro et al.20 reported STM images in solution that showed facilitated electron tunneling between a DNA base and a complementary base-modified STM tip. Through theoretical simulation, Zwolak et al.21 showed that the four DNA bases G, A, C, and T in single-stranded (ss) DNA support currents of different magnitudes in the direction perpendicular to the backbone, and argued that electronic DNA sequencing may be possible. The findings of Zwolak et al. were recently contrasted by Zikic et al.,22 who reported a different order of current magnitudes through the four bases using density functional theory (DFT) calculations. The same authors also found that the current is very sensitive to the geometrical conformation of the DNA bases and their orientation to the electrodes. However, very few experimental current–voltage (I–V) characteristics of individual bases are available for comparison. Here, we report the electronic properties of the four nucleosides (DNA bases with ribose), that is, guanosine, adenosine, cytidine, and thymidine, as well as of 40-base-pair-long, double-stranded (ds) poly(GC) and poly(AT) DNA, by exploiting ultrahigh-vacuum (UHV) STM at room temperature.23 For comparison we also performed first-principles DFT calculations of the four methylated bases (ribose replaced by methyl group to speed up simulations). We find that the four DNA bases (nucleosides and methylated bases) show base-specific electronic signatures, although the electronic differences among the four bases are small. The base-specific signatures not only provide clues for understanding charge injection into and migration along DNA molecules but may also shed light on the issue of electronic DNA sequencing. We performed control experiments to study the effect of the K2HPO4+Tris–ethylenediaminetetraacetic acid (TE) buffer solution, which was used to dissolve the nucleosides and the dsDNA molecules. As shown in Figure 1, careful examination of the clean and the buffer-treated Au(111) surfaces suggests that our protocol is suitable for examining nucleosides and DNA molecules.23 For the nucleosides on Au(111) surfaces, we observed well-ordered patterns of adenosine (Figure 2 b) and cytidine (Figure 2 c) with periodic repeats of about 6.2 and 6.8 nm, respectively (see profiles in insets), but not of guanosine (Figure 2 a) and thymidine (Figure 2 d). The latter adsorbed as isolated molecules (or small clusters). For all four nucleosides we observed isolated molecules (or small clusters) as well as uniform-appearing films (see Figure S1 in the Supporting Information as an example). As the formation of such structures on various substrates was described previously,5–8 these images are not discussed further. Representative normalized conductances of metallic clean and buffer-treated Au(111), as well as of insulating thymidine and poly(GC) dsDNA on Au(111). UHV STM images (100×100 nm2) of the four nucleosides on Au(111) surfaces acquired at preset tunneling conditions of 2.0 V and 30 pA. a) Guanosine, where spots may indicate single or very few molecules; b) adenosine, which shows a well-ordered pattern with a periodic repeat of about 6.2 nm (see the profile along the line in the inset); c) cytidine, which shows a well-ordered pattern with a periodic repeat of about 6.8 nm (see the profile along the line in the inset), and d) thymidine. Figure 3 a (bottom/left axes) and Figure 3 b show representative I–V curves that were obtained from uniform-appearing films and the corresponding normalized conductances of the four nucleosides, respectively. Figure 3 a (top/right axes) shows the corresponding current ratios (IX/IT) between the current of nucleoside X (A, G, or C) and the current of nucleoside T at different biases. Figure 3 c shows the histogram of the current at bias −2.44 V, obtained from 20 different sample locations (including uniform-appearing films, isolated clusters, and well-ordered patterns if available) for each nucleoside. For negative bias voltages below the bandgap threshold, nucleoside T exhibits the smallest current among the four nucleosides (Figure 3 a and c), as theoretically predicted.21, 22 However, the current ratios are generally a complicated function of the bias voltage in the range from −3.0 to +3.0 V (see Figure 3 a). For instance, at negative bias the current decreases in the order G, A, C, and T, while at positive bias the current decreases in the order C, A, T, and G. Electronic signatures of DNA bases. a) Representative I–V curves (bottom/left axes) acquired at 2.3 V and 30 pA, as well as a scatter plot of the corresponding current ratios of X (X=A, G, or C) relative to T (top/right axes) at fixed biases; b) normalized conductances calculated from I–V curves in (a); c) histogram of currents of nucleosides at fixed bias based on 20 measurements, and d) histogram of nucleoside HOMO energies based on 20 measurements. The tunneling current depends exponentially on the tip–sample separation. To minimize the influence of the separation between the tip and the sample, we calculated normalized conductances from the I–V curves, which reflect the electronic density of states (DOS) of the bases on the surface (Figure 3 b). Based on the normalized conductances, we can assign the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies (see Experimental Section), shown in a histogram in Figure 3 d. The HOMO energies fluctuate with sample location of the nucleosides, which can influence the charge-injection barrier. Our observation of the variability of the HOMO energies is supported by our first-principles study, discussed below, and recent DFT calculations by Zikic et al.22 The averages and standard deviations of our measured HOMO energies are −1.39±0.07 (G), −1.56±0.06 (A), −1.69±0.06 (C), and −1.74±0.08 eV (T) (relative to the Fermi energy EF of Au(111)). Despite the variability, the relative order of our four HOMO energies is consistent with measured ionization energies,5, 24 for example, −7.926 (G), −8.503 (A), −8.968 (C), and −9.516 eV (T).5 To support our experimental findings, we performed first-principles calculations of the electronic structure of single methylated bases on a Au(111) slab. For each base two initial structures (differing only in the lateral orientation of the base on the slab) were relaxed to the nearest local minimum on the energy landscape. Binding distances between DNA bases and the top Au layer ranged from 2.6 to 3.1 Å; the angle between the normal vectors of DNA bases and Au(111) ranged between 4 and 22°, which indicates that bases adsorb almost parallel to the surface (see inset in Figure 4 c). Figure 4 a–d shows the DOS projected onto the four methylated bases. In each panel, the two upper curves show the results from the two simulations, where the solid curve is the simulation with the lower total energy and the dashed curve is the simulation with the higher total energy; ΔE denotes their energy difference. In Figure 4 a, c, and d the large value of ΔE reflects the fact that in these simulations the bases lost a hydrogen atom from their methyl group, which indicates unfavorable chemical adsorption. The difference of the DOS between simulations of the same base with comparable total energy (see Figure 4 b) may explain the fluctuations of HOMO energies observed in our experiments. For comparison, we also calculated the DOS of unbound bases by separating the bases by 6 Å from the surface (lower curve in each panel). In this case, the HOMO energies can clearly be assigned to −4.86 (G), −5.13 (A), −5.21 (C), and −5.40 eV (T). Note that adsorption generally leads to an energetic downward shift of the molecular energy levels, and that the DOS results of C and T are smeared out and show very few features. The similarity between C and T, as well as the order of the HOMO energies (best visible for unbound bases), are in line with our experimental findings. DOS of methylated bases on a Au(111) slab (see inset in (c) as an example), obtained with DFT code SIESTA: a) guanine, b) adenine, c) cytosine, and d) thymine. For clarity, only the projection onto the p orbitals of elements C, N, and O is shown. The relative order of the HOMO energies of the four DNA bases suggests that the hole-injection barrier from the electrode to the bases is the lowest for base G and the highest for base T. This finding is consistent with the HOMO energies of dsDNA, measured in the transverse direction (see Figures S2 and S3 of the Supporting Information). The average HOMO energies of the poly(GC) and poly(AT) dsDNA are −1.25±0.12 and −1.62±0.07 eV (relative to the Fermi level of Au(111)), respectively, obtained from ten different sample locations. This finding indicates that it is easier to inject a hole from the electrode to a G–C base pair than to an A–T base pair, and that an A–T base pair constitutes a charge-migration barrier of ≈0.37 eV for hole migration along the DNA helix.25 However, note that charge transport also depends strongly on the conformation-dependent electronic coupling between adjacent base pairs, as well as on other factors.2, 26 Electronic sequencing of DNA has recently attracted a lot of interest. Suggested approaches include STM/scanning tunneling spectroscopy (STS),21 and also nano-optics27 and solid-state nanopores.28, 29 The prerequisite for electronic sequencing of whole genomes is the identification of robust electronic signatures of the four bases, in particular in the context of ssDNA. To estimate the feasibility of sequencing, we used the base-specific distributions of the HOMO energies and currents, and calculated the probability of predicting the correct base from the present electronic measurements (see Supporting Information). The obtained probabilities ranged from 0.5 to 0.9, which is considerably larger than guessing (≈0.25) but much smaller than the consensus accuracy of 0.9999 of the Sanger sequencer.30 This result demonstrates that single-molecule measurements are inherently noisy, most likely caused by variations in adsorption geometry and the tip–sample contact at room temperature. These effects are hard to control, especially in solution experiments. In the case of electronic sequencing by STM/STS in a vacuum, the proper sample preparation of ssDNA with regularly aligned bases may be achieved through DNA combing and stretching.31 In conclusion, by combining UHV STM with a first-principles calculation we demonstrated that DNA bases show base-specific electronic signatures, for example, molecular energy levels and currents. Our results aid the understanding of charge injection from electrodes to bases or base pairs and charge migration along DNA molecules. However, the distributions of the base-specific electronic signatures are broad and overlap each other, most likely caused by variations of the adsorption geometry, of the molecule–substrate contact, and of the location of the tip above the molecules. These effects need to be controlled to make electronic sequencing possible. STM and STS measurements: The four nucleosides (adenosine: A, guanosine: G, cytidine: C, and thymidine: T) were purchased from Sigma–Aldrich. Nucleoside solutions were prepared by dissolving nucleoside powder (0.02 mg) in 1 M K2HPO4+TE buffer solution (5 mL, pH 7.0). Immobilization of the four nucleosides on Au(111) substrate was performed by immersing the substrate in the nucleoside solutions for about 15 min. For the immobilization of the 40-base-pair-long poly(GC) and poly(AT) dsDNA on Au(111), the DNA molecules were dissolved in 1 M K2HPO4+TE buffer solution (pH 7.0) after hybridization from ssDNA.23 The Au substrate was immersed in about 1.5 μM dsDNA solution for 10 min. STM and STS were performed by using an etched tungsten tip with a bias voltage applied to the sample at room temperature.23, 32 We used the same tip for measuring all four nucleosides and another tip for the two dsDNA molecules but we cleaned the tips frequently. All the presented I–V curves are averages over five individual I–V curves obtained from the same location, to reduce noise. Furthermore, each I–V data point of an individual curve is an average of 32 repeated measurements collected by the STM control electronics. From the average I–V curves we calculated normalized conductances, defined as (dI/dV)/((I/V)2+c2)1/2. Introduction of the small constant c avoids dividing by zero.33 We chose c to be equal to 2.0 % of the maximum current of the I–V curve throughout this work. Based on normalized conductances, we assigned the HOMO and LUMO energies according to the following criterion: the first peak or shoulder larger than 2.0 to the left (right) of the bandgap was assumed to be the putative HOMO (LUMO). First-principles study: The DOS of the methylated bases on Au(111) was obtained with the SIESTA package.34, 35 SIESTA is based on ab initio DFT, and uses the generalized gradient approximation as well as pseudopotentials for core electrons and a localized atomic-orbital-like basis set for high efficiency. We used a double-zeta basis set with additional polarization orbitals. The basis set of Au includes 4f electrons explicitly, while 5d and 6s electrons are included in the pseudopotential. The Au surface was modeled with a periodically repeated three-layer slab of 4×5(4) Au atoms in each layer for a purine (pyrimidine) calculation using the experimental lattice constant 4.08 Å.36 The atoms of the base and top surface layers were relaxed until forces on atoms were below 0.04 eV Å−1. For each base, the structural relaxation was started from two slightly different initial structures where the base is oriented parallel to the surface at a distance of 2 Å but with small differences in the lateral base position. To check whether three surface layers including one mobile layer are sufficient we used the relaxed output structure of a cytosine calculation, added a fourth layer of Au at the bottom of the slab, and restarted the relaxation calculation while allowing the top two surface layers to be mobile. After restarting the calculation, the initial forces on the atoms of the DNA base were below 0.1 eV Å−1. The relaxed structure and DOS were essentially indistinguishable from before. Supporting Information: The different self-assembled structures of adenosine on Au(111) by STM and the corresponding STS, the STS results obtained in the transverse direction of the poly(GC) and poly(AT) dsDNA on Au(111), atomic force microscopy images of the poly(GC) and poly(AT) on Au(111), and the estimated sequencing accuracy are available online. This work was supported in part by Special Coordination Funds for Promoting Science and Technology Supporting information for this article is available on the WWW under http://www.wiley-vch.de/contents/jc_2296/2007/z600732_s.pdf or from the author. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.