Sequence-Dependent DNA Immobilization: Specific versus

Langmuir 2004, 20, 3357-3361

3357

Sequence-Dependent DNA Immobilization: Specific versus Nonspecific Contributions Lauren K. Wolf, Yang Gao, and Rosina M. Georgiadis* Department of Chemistry, Boston University, 590 Commonwealth Avenue, Boston, Massachusetts 02215 Received November 11, 2003. In Final Form: January 23, 2004 We present results of the first systematic study on in situ sequence-dependent kinetics for short singlestrand oligonucleotide surface immobilization. By measuring film coverage for both thiolated and nonthiolated homo-oligomers as a function of adsorption time, we determine the relative contribution of specific thiol-surface and nonspecific DNA-surface interactions to the overall mechanism of DNA-thiol attachment to gold. We find that sequence-dependent nonspecific surface interactions play a significant role in DNA-thiol immobilization, influencing not only the kinetics but also the extent of oligomer adsorption. For example, sequences that initially form strong, rapid nonspecific contacts with the surface hinder long-time thiol adsorption (i.e., poly(dA)-thiol). In contrast, sequences with nucleotides that initially bind slowly and weakly to the surface (i.e., poly(dT)-thiol) do not obstruct further thiol adsorption, resulting in higher film coverage and Langmuir immobilization kinetics. This view of the DNA-thiol immobilization mechanism is further supported by sequence-dependent rinsing losses observed for thiolated DNA strands but not for analogous nonthiolated strands. Nonthiolated strands contact the surface strongly in a more horizontal orientation, whereas thiolated strands attain a more upright orientation that allows vertical DNA-DNA base-stacking. The results clearly illustrate the importance and interplay of competitive specific and nonspecific forces in forming DNA-thiol films. The specific coverage attained and the time dependence of the adsorption process depend on the prevailing sequence composition.

Introduction Because of the widespread use of biosensor technology, intense interest in the fundamentals of DNA-surface attachment is growing. A common method for fabricating nanoparticle and planar DNA sensors at gold surfaces relies on the assembly of prefabricated oligonucleotides via a covalent thiol linker. Unlike simple alkanethiol selfassembly, which involves headgroup (linker) covalent attachment and interchain van der Waals interactions, the assembly process for thiolated DNA films involves contributions from both specific linker-surface and nonspecific strand-surface interactions. Although not well-understood, nonspecific interactions participate significantly in the DNA-thiol immobilization process. Furthermore, we find that sequence-dependent nonspecific interactions affect not only the kinetics of the assembly process but also oligomer coverage and orientation within the film. Because the performance of a surface biosensor depends crucially on its probe density1 and availability for target binding,2 it is clear that the contributions of nonspecific binding to the overall DNA-thiol immobilization mechanism must be carefully investigated. Mirkin and co-workers recently studied nonspecific interactions of nucleobase and nucleoside binding to gold substrates with nanoparticle agglomeration and ex situ vacuum temperature-programmed desorption (TPD) experiments.3,4 These studies, as well as a competitive adsorption ex situ FTIR investigation5 of 5-mer homooligonucleotides, showed sequence dependence in the * To whom correspondence should be addressed. E-mail: [email protected]. (1) Peterson, A. W.; Heaton, R. J.; Georgiadis, R. M. The effect of surface probe density on DNA hybridization. Nucleic Acids Res. 2001, 29, 5163-5168. (2) Levicky, R.; Herne, T. M.; Tarlov, M. J.; Satija, S. K. Using selfassembly to control the structure of DNA monolayers on gold: A neutron reflectivity study. J. Am. Chem. Soc. 1998, 120, 9787-9792.

strength of binding for these small units of DNA. Nevertheless, it has not been investigated how these sequence-dependent nonspecific interactions affect DNAthiol attachment. There is some evidence for significant sequence-dependent variation in the adsorption coverage of modified short-strand homo-oligomers6 and sequences containing homo-oligomeric spacer regions.7 Migration studies have also suggested sequence dependence in the orientation of 36- to 43-mer strands attached to nanoparticles.8 However, there has been no systematic investigation of the sequence-dependent kinetics or the competitive specific and nonspecific contributions that occur during DNA immobilization. Because surface plasmon resonance (SPR) spectroscopy detects the adsorption of molecules at surfaces based on a local change in refractive index, we can measure both specific and nonspecific contributions to the in situ DNA film formation without the need for labeling. In this study, we have used SPR to examine the binding of both thiolated (ssDNA-thiol) and nonthiolated (ssDNA) homo-oligo(3) Storhoff, J. J.; Elghanian, R.; Mirkin, C. A.; Letsinger, R. L. Sequence-dependent stability of DNA-modified gold nanoparticles. Langmuir 2002, 18, 6666-6670. (4) Demers, L. M.; Ostblom, M.; Zhang, H.; Jang, N. H.; Liedberg, B.; Mirkin, C. A. Thermal desorption behavior and binding properties of DNA bases and nucleosides on gold. J. Am. Chem. Soc. 2002, 124, 11248-11249. (5) Kimura-Suda, H.; Petrovykh, D. Y.; Tarlov, M. J.; Whitman, L. J. Base-dependent competitive adsorption of single-stranded DNA on gold. J. Am. Chem. Soc. 2003, 125, 9014-9015. (6) Steel, A. B.; Herne, T. M.; Tarlov, M. J. Electrochemical quantitation of DNA immobilized on gold. Anal. Chem. 1998, 70, 46704677. (7) Demers, L. M.; Mirkin, C. A.; Mucic, R. C.; Reynolds, R. A.; Letsinger, R. L.; Elghanian, R.; Viswanadham, G. A fluorescence-based method for determining the surface coverage and hybridization efficiency of thiol-capped oligonucleotides bound to gold thin films and nanoparticles. Anal. Chem. 2000, 72, 5535-5541. (8) Parak, W. J.; Pellegrino, T.; Micheel, C. M.; Gerion, D.; Williams, S. C.; Alivisatos, A. P. Conformation of oligonucleotides attached to gold nanocrystals probed by gel electrophoresis. Nano Lett. 2003, 3, 33-36.

10.1021/la036125+ CCC: $27.50 © 2004 American Chemical Society Published on Web 03/03/2004

3358

Langmuir, Vol. 20, No. 8, 2004

Wolf et al.

Table 1. Sequences of Homo-oligonucleotide Strands homo-oligonucleotide poly(dA)-thiol poly(dT)-thiol poly(dC)-thiol poly(dA) poly(dT) poly(dC)

sequence HS-(CH2)6-5′-(A)25-3′ HS-(CH2)6-5′-(T)25-3′ HS-(CH2)6-5′-(C)25-3′ 5′-(A)25-3′ 5′-(T)25-3′ 5′-(C)25-3′

nucleotides at the bare gold-solution interface. Specifically, we have characterized and compared the sequencedependent binding of 25-mer poly(dA), poly(dT), and poly(dC) to a gold substrate. Poly(dG) was not investigated because its preference to form G-quartets (from Hoogsteen base-pairing)9 makes synthesis challenging. Its unique secondary structure would also make comparison to other sequences difficult. Experimental Section Materials. All oligonucleotides were purchased from Integrated DNA Technologies (IDT). We used 5′ thiol-modified, HPLCpurified 25-mers for all specific binding experiments, Table 1. These were received in disulfide form, protected by a mercaptohexanol group. This group was cleaved from the oligonucleotides by treatment with 0.04 M dithiothreitol (DTT) in 0.17 M phosphate buffer (pH 8.0) at room temperature for at least 16 h. The aqueous oligomer fraction of the cleaved products was separated and collected upon running through a size-exclusion NAP-10 column (Amersham Biosciences) equilibrated with 0.01 M sodium phosphate buffer (pH 6.8). The thiolated single-strand eluent was tested with UV-vis spectroscopy (Varian Cary 100 Bio) at 260 nm to establish concentration and converted into a 1 µM DNA/1 M KH2PO4 solution for immediate use. Nonthiolated, PAGE-purified 25-mer sequences were used as received in all nonspecific binding experiments, Table 1. KH2PO4 (Sigma-ACS Grade) was used to prepare 1 M solutions for all binding experiments in 18 MΩ cm distilled water. Refractive index increment values were determined for dA, dT, and dC nucleotides (2′-deoxyadenosine 5′-monophosphate, thymidine 5′-monophosphate, and 2′-deoxycytidine 5′-monophosphoric acid, respectively) obtained from Sigma. Formation of ssDNA Films. Gold films were evaporated by electron-beam directly onto SF-14 hemicylindrical prisms. The substrate prisms were cleaned with “piranha solution” (7:3 H2SO4/H2O2 (30% solution)) at >50 °C for 15 min, rinsed with copious amounts of distilled water, and dried under nitrogen prior to evaporation. Once at a base pressure of (8-9) × 10-7 Torr, the flat prism surfaces received ∼10 Å Cr (Alfa Aesar, 99.999%) at 0.3 Å/s, followed by ∼450 Å Au (Kurt J. Lesker Inc., 99.99%) at 0.9 Å/s. Prior to DNA film formation, the gold-coated prisms were cleaned with piranha solution at >50 °C for 5 min, rinsed with distilled water, and dried under nitrogen. This treatment always yielded a completely hydrophilic (wetted) gold surface. The fresh bare gold surfaces were immediately sealed via a Kalrez O-ring to an all-Teflon cell (∼3 mL capacity) and rinsed with water and 1 M KH2PO4. The ssDNA films were then prepared by injecting a 1 M KH2PO4 solution containing 1 µM DNA into the cell. Formation was monitored continuously (in situ) for at least 12 h at room temperature with surface plasmon resonance spectroscopy. Finally, the homo-oligonucleotide films were rinsed with copious amounts of 1 M KH2PO4 solution and water. The 1 µM DNA concentration used in these studies was convenient for measuring kinetics of formation. Saturation film coverage is not significantly affected by the use of oligonucleotide concentrations above 1 µM. Previous studies have shown that a 100-fold increase in 8-mer probe concentration above 1 µM does not increase the final film coverage.10 We and others find that (9) Bloomfield, V. A.; Crothers, D. M.; Tinoco, I. Nucleic Acids: Structures, Properties and Functions; University Science Books: Sausalito, 2000. (10) Steel, A. B.; Levicky, R. L.; Herne, T. M.; Tarlov, M. J. Immobilization of nucleic acids at solid surfaces: Effect of oligonucleotide length on layer assembly. Biophys. J. 2000, 79, 975-981.

the only way to significantly increase the maximum probe coverage during oligonucleotide immobilization is to employ higher ionic strength or multivalent cation deposition solutions.1,11 Probe oligomers reach a threshold saturation coverage at which electrostatic repulsion between strands prevents further surface attachment. Our selection of 1 M KH2PO4 (pH ∼ 4.5) immobilization solution should not have any significant effect on the overall findings of this study. We and others have shown that the coverage and kinetics of DNA adsorption are most affected by ionic strength and cation valency rather than pH.1,11 However, our experimental solution pH might be expected to affect cytidine (pKa ) 4.2) containing sequences such as poly(dC). We performed control experiments in 1 M NaCl-Tris/EDTA (pH ∼ 7.7) solution for poly(dT)-thiol and poly(dC)-thiol and found no significant pH dependence. Within our reported sample-to-sample variation, the relative kinetics and coverages for these sequences were unaffected by the presence of buffer and neutral pH. SPR Measurements. The SPR theory, apparatus, and methods of data analysis have been described in detail elsewhere.12-14 Briefly, p-polarized light from a visible (632.8 nm) He-Ne laser is focused through the back of the SF-14 prism onto the gold surface. The entire system, including the flow cell, is mounted to a motorized θ/2θ goniometer. The reflectivity of light is collected as a function of incident angle by a photodiode and fit to a multilayer Fresnel optical model. Changes in cell temperature are measured by a built-in poly(tetrafluoroethylene) (PTFE)-coated Chromel-Alumel thermocouple and accounted for by the fitting algorithms. The multilayer Fresnel model, when used to fit SPR spectra of adsorbed molecular layers on gold, yields film properties dlayer (thickness) and layer (dielectric constant). For very thin films, a unique dlayer-layer pair can be obtained by specialized methods such as two-color SPR.15 Accurate coverages, however, can be calculated regardless of the specific dlayer-layer pair through careful selection of a reasonable film dielectric constant. For this reason, we present our experimental results for DNA submonolayers in terms of coverage (molecules/cm2). We fit all data using a dielectric constant for the adsorbed layer, layer ) 2.0, and determined a corresponding “optical thickness”, rather than a physical thickness, for DNA films from the Fresnel model. Coverage was then calculated from this pair of parameters. In using layer ) 2.0, no assumptions are made about the orientation of the DNA molecules on the gold surface. This value is a good choice because it falls in a general range of dense DNA films formed under 1 M salt conditions that we and others1,7,10 have studied from 25% volume of a close-packed upright monolayer (layer ) 1.9, dlayer ∼ 73 Å) to 75% of a monolayer in which the strands are lying down (layer ) 2.3, dlayer ∼ 15 Å). Coverages calculated within this range have an associated error of, at most, 8%. Another factor to be considered during data analysis, particularly during data fitting, is the electronic perturbation at the gold surface that occurs during molecular adsorption. Allara and co-workers observed such electronic variations while monitoring alkanethiol adsorption on gold by ellipsometry.16 The perturbations in these studies were accounted for through the introduction of an additional thin interfacial layer in the optical fitting model. (11) Petrovykh, D. Y.; Kimura-Suda, H.; Whitman, L. J.; Tarlov, M. J. Quantitative analysis and characterization of DNA immobilized on gold. J. Am. Chem. Soc. 2003, 125, 5219-5226. (12) Peterlinz, K. A.; Georgiadis, R. In situ kinetics of self-assembly by surface plasmon resonance spectroscopy. Langmuir 1996, 12, 47314740. (13) Peterlinz, K. A.; Georgiadis, R. M. Observation of hybridization and dehybridization of thiol-tethered DNA using two-color surface plasmon resonance spectroscopy. J. Am. Chem. Soc. 1997, 119, 34013402. (14) Georgiadis, R.; Peterlinz, K. P.; Peterson, A. W. Quantitative measurements and modeling of kinetics in nucleic acid monolayer films using SPR spectroscopy. J. Am. Chem. Soc. 2000, 122, 3166-3173. (15) Peterlinz, K. A.; Georgiadis, R. Two-color approach for determination of thickness and dielectric constant of thin films using surface plasmon resonance spectroscopy. 1996, 130, 260-266. (16) Shi, J.; Hong, B.; Parikh, A. N.; Collins, R. W.; Allara, D. L. Optical Characterization of Electronic-Transitions Arising from the Au/S Interface of Self-Assembled N-Alkanethiolate Monolayers. Chem. Phys. Lett. 1995, 246, 90-94.

Sequence-Dependent DNA Immobilization This layer and its optical properties described the region encompassing the newly formed gold-thiol bond. In our studies, electronic perturbations associated with DNA surface adsorption manifested in slight changes to the reflected intensity and shape profile at the SP resonance position. We accounted for such intensity changes by allowing the imaginary dielectric constant of the entire gold layer (Au,imag) to vary in data fitting. This parameter, which is normally fixed during data analysis and describes the damping of the surface plasmon,17 is affected by chemical bond formation between the molecule and gold. Our data treatment, which introduces an additional parameter (Au,imag) to the fit, is optically equivalent to the method used by Allara and co-workers and results in excellent agreement between the optical model and experimental data. Film formation experiments were repeated multiple times on numerous gold surfaces, for which the gold optical properties were not precisely the same from sample to sample. Nevertheless, in all cases, Au,imag increased by 0.13 ( 0.04 for covalently bound ssDNA-thiol films and by 0.05 ( 0.01 for all nonspecifically bound ssDNA films during immobilization. The stronger the attachment to gold, the more the plasmon was damped and the further Au,imag increased. Adjusting Au,imag during data analysis results in a more reliable fit to the experimental data and yields, at most, a 5% difference in calculated coverage. Refractive Index Increment Determination. The refractive index increment (dn/dc) of each DNA adsorbate must also be established in order to derive accurate film coverage, Γ (molecules/cm2), given by

Γ)

-1

(dn dc )

(nlayer - nsolution)dlayer

where nlayer is the refractive index of the DNA layer (for transparent media, nlayer ) (layer)1/2, so here, nlayer ) (2.0)1/2), nsolution is the known refractive index of the immobilization solution, and dlayer is the optical thickness of the DNA layer determined from a multilayer Fresnel fit. Values of refractive index increment (RII) for oligonucleotides are not available in the literature; measurement would require costly mM concentrations of synthetic oligomers. As such, we determined RII values for individual dA, dT, and dC nucleotides in water and 1 M KH2PO4 (see the Supporting Information for details). The RII values obtained in 1 M KH2PO4 were applied to our coverage calculations with the assumption that the value per base is additive. If the RII values are not additive due to the effects of nucleotidenucleotide bonds present within the oligomers studied, the relative coverage calculations remain valid. Nucleotide-nucleotide bonds are present in all of the sequences and should affect each in the same way. Similarly, the small hexanethiol group present on our thiolated oligonucleotides may have a slight effect on the molecular polarizabilities and, thus, the RII values of these sequences. However, the hexanethiol functionality is common to all of our thiolated sequences and the relative resulting coverages in this study should not be affected. The RII values obtained for dA, dT, and dC in water were 0.217 ( 0.001, 0.174 ( 0.001, and 0.225 ( 0.002, respectively. The RII values obtained for dA, dT, and dC in 1 M KH2PO4 were 0.206 ( 0.001, 0.151 ( 0.001, and 0.218 ( 0.004, respectively. The differences in these values indicate differences in the polarizabilities of the nucleotides and, by extension, the homooligonucleotides.

Results and Discussion Figure 1 displays clear sequence dependence in the immobilization of poly(dA)-thiol, poly(dT)-thiol, and poly(dC)-thiol films. The kinetic rates of adsorption for each of these strands display reproducibility on different gold samples over the course of months. Poly(dT)-thiol always binds rapidly to the surface, while poly(dA)-thiol never reaches saturation coverage and continues to adsorb slowly during the time of the experiment (>12 h). The adsorption (17) Raether, H. Surface Plasmons; Springer: Berlin, 1988.

Langmuir, Vol. 20, No. 8, 2004 3359

Figure 1. Immobilization of poly(dA)-thiol (open circles), poly(dT)-thiol (black circles), and poly(dC)-thiol (gray circles) onto a bare gold surface. All immobilization solutions were 1 M KH2PO4 containing 1 µM thiolated homo-oligomer. Each trace is the average of two experiments on different piranha-cleaned gold surfaces. Sample-to-sample kinetic rates of adsorption were reproducible for each homo-oligomer, and the average sampleto-sample deviation in coverage was 9%. The inset shows the kinetic behavior of each oligomer during the first 15 min of film formation. Initial adsorption rates of each sequence (