Hydrophobic Patterned Surfaces as Templates for DNA

Lide, D. The CRC Handbook of Chemistry and Physics, 74th ed.; CRC Press: Ann Arbor, MI. 1993; pp 6−144. There is no corresponding record for this re...
0 downloads 0 Views 150KB Size
Download PDF
Langmuir 2000, 16, 7223-7228

7223

Hydrophilic/Hydrophobic Patterned Surfaces as Templates for DNA Arrays S. D. Gillmor,† A. J. Thiel,‡ T. C. Strother,‡ L. M. Smith,‡ and M. G. Lagally*,† Department of Materials Science and Engineering and Department of Chemistry, University of WisconsinsMadison, Madison, Wisconsin 53706 Received July 29, 1999. In Final Form: May 11, 2000 Addressable DNA arrays are created by photopatterning self-assembled monolayers to form hydrophilic and hydrophobic regions on a gold surface. The hydrophilic regions aid to contain small volumes of different DNA solutions placed on them using an automated pin-tool loading strategy. The method allows for efficient attachment, manipulation, and hybridization of pure DNA strands on the surface.

Langmuir 2000.16:7223-7228. Downloaded from pubs.acs.org by UNIV OF EDINBURGH on 01/29/19. For personal use only.

I. Introduction The rapid development and implementation of DNA arrays on solid supports in recent years have been driven by their potential for the high-throughput analysis of DNA sequences, genetic variations, and gene expression. Methods to create DNA arrays on surfaces fall into two broad categories: (1) base-by-base attachment to build up different DNA strands at different sites in the array and (2) attachment of different complete strands at individual array sites. The former approach is exemplified by work from Affymetrix, where Foder et al.1 have developed a photolithographic technique to create DNA arrays through a series of exposures through masks to control the addition of bases to specific sites on the surface. This technique has been reported to have a repetitive yield of ∼0.95 per base pair attachment,2 considerably limiting the length of the oligonuclotide that can be synthesized with fidelity. Piezoelectric printing of each base followed by standard washing and deprotection for the subsequent deposition of another base onto the anchored DNA strand is a second example of base-by-base DNA synthesis on a surface. Piezoelectric printing has a reaction success rate of ∼0.99, similar to controlled pore glass (CPG) synthesis, a common traditional oligonucleotide synthesis technique. Methods of DNA strand synthesis on the surface do not allow for oligonucleotide purification after synthesis to eliminate incomplete strands or ones with an excess number of bases, considerably reducing the overall array reliability. Shalon and Brown have created arrays using the second method, attachment of complete strands of DNA to a surface.3 A pin tool is used to load the array mechanically. The oligonucleotides can be rigorously purified before surface attachment to allow for longer strands of high fidelity, something not possible with in situ synthesis. After CPG synthesis of DNA strands, a standard cartridge filter process is used to eliminate the failed sequences. The correct sequences will still have a tri-p-methylpphenyl group (trityl) attached, creating a highly hydrophobic tail that can be separated from the non-trityl species, which † ‡

Department of Materials Science and Engineering. Department of Chemistry.

(1) Fodor, S. P. A.; Read, J. L.; Pirrung, M. C.; Stryer, L.; Tsai, L. A.; Solas, D. Light-Directed, Spatially Addressable Parallel Chemical Synthesis. Science 1991, 251, 767-773. (2) Marshall, A.; Hodgson, J. DNA chips: An Array of Possibilities. Nat. Biotechnol. 1998, 16, 27-31. (3) Shalon D. Gene Expression Micro-Arrays: A New Tool for Genomic Research. Genome Methods 1996, 6 (7), 639-645.

are failed sequences. Further purification can include HPLC, ion exchange, size exclusion, or electrophoresis gel procedures to achieve highly pure DNA samples. Nanogen’s technique, electrically polarizing an array site so that the DNA sample is attracted to that site only, is another example of attachment with presynthesized DNA onto a surface. Finally, ink jet technology has also been used to print previously synthesized oligonucleotides onto an active surface.4 In these latter methods, while the attachment chemistry of the DNA strands to the surface is vital, it is the delivery system, such as the ink jet, the pin tool, or the photolithographic technique, that controls the precise placement of the oligonucleotides onto the array element. Blanchard and Hood proposed5 a technique in which the surface would assist in the placement, through the use of a polyfluorinated hydrophobic background and hydrophilic active areas on a silicon support. A patent on DNA microarrays, issued to ProtoGene, also demonstrates an application of chemical patterning to aid in placement with a glass support and fluoroalkylsilane chemistry.6 In this paper, we demonstrate the use of alkanethiol chemistry to create a patterned hydrophobic/hydrophilic array to confine minute volumes (100-200 nL) of DNA solutions. The alkanethiol system is a stable, wellcharacterized chemistry with a fast reaction rate that creates well-formed self-assembled monolayers and can be used to create a patterned surface. We create an array pattern on the surface with ultraviolet photolithography to control the placement of the different alkanethiols. Methyl-terminated alkanethiol molecules are attached to a gold film on a solid support as a hydrophobic background layer. DNA is attached to active hydrophilic regions, linked to hydrophilic alkanethiols, arranged in an array. We control the placement of the DNA onto the surface using a computer-programmable automated pin tool. The hydrophobic interaction with the aqueous DNA solutions prevents DNA attachment between the elements of the array and thus acts as a barrier for diffusion of DNA between the elements. Furthermore, any DNA that is placed on a hydrophobic region near a hydrophilic element (4) O′Donnell-Maloney, M. J.; Little, D. P.; Microfabrication and Array Technology for DNA Sequencing and Diagnostics. Genet. Anal.: Biomol. Eng. 1996, 13, 151. (5) Blanchard, A. P.; Kaiser, R. J.; Hood, L. E. High-density Oligonucleotide Arrays. Biosens. Bioelectron. 1996, 11 (6/7), 687-690. (6) Brennan, T. M. ProtoGene Laboratories, Inc., Mountain View, CA Patent number US 5474796, Dec 12, 1995.

10.1021/la991026a CCC: $19.00 © 2000 American Chemical Society Published on Web 08/05/2000

7224

Langmuir, Vol. 16, No. 18, 2000

diffuses into this region. With our strategy, we can attach completely synthesized and purified DNA strands to create high-purity arrays with high positional fidelity, excellent stability, and minimal cross talk between elements. II. Experimental Section In this section, we first describe aspects of the DNA attachment to surfaces that are general and not specific to the existence of a laterally defined pattern. We then describe the materials used in our procedure. Finally, we describe surface patterning and aspects of the attachment chemistry that are specific to the existence of a pattern. A. Uniform Surface. (1) Overview. A glass slide or silicon wafer is prepared by evaporating on it 50 Å of chromium followed by 1000 Å of gold. 11-Mercaptoundecanoic acid (MUA), a hydrophilic alkanethiol, is attached to the surface and orders into a self-assembled monolayer (SAM).7 Poly-L-lysine hydrobromide (PL) is attached to the MUA surface.8 The sample is then covered with a solution of sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SSMCC) followed by placement of 5′-thiol-modified DNA solutions onto the surface. The DNA reacts with the SSMCC, forms a bond, and becomes anchored to the surface.9 DNA attached onto the SSMCC areas is detected by exposing the surface to a solution of the fluorescently tagged complement of the DNA and by imaging the surface with a fluorescent imaging technique.10 (2) Surface Modification Chemistry. The gold surface is immersed in 1 mM ethanolic MUA for at least 18 h and is then covered with 1 mg/mL of PL of pH 8.0 for 30 min. An amine group of the PL reacts with a carboxylic acid group of the MUA to form an electrostatic bond.9,10 Subsequent reaction with SSMCC occurs for 30 min; during this time the remaining amine groups of the PL react with the carboxylic acid groups of the SSMCC to form a covalent bond. The surface is then ready for DNA attachment.9,10 (3) Contact Angle Measurements. To investigate the surface free energies of the several levels of the chemically modified surfaces, we measured the contact angles of water at room temperature with a model 100-00 contact angle goniometer (Rame´-Hart, Inc.). The contact angles are recorded immediately, after dispensing 10 µL droplets with a Gilson pipet onto the surface. The reported values for the alkanethiols are an average of 20 different measurements taken on three individually prepared samples; the MUA and MUA + PL data are for 14 measurements on each of two samples. The MUA + PL + SSMCC data are for 18 measurements each from three different samples. The results are discussed below. (4) DNA Attachment. A glass slide or silicon wafer is chemically modified for DNA attachment as described above. A purified 1 mM aqueous solution of 5′-thiolmodified DNA is placed onto the slide in 0.8 µL droplets and allowed to react for at least 6 h in a humid chamber (7) Nuzzo, R. G.; Allara, D. L. Adsorption of Bifunctional Organic Disulfides on Gold Surfaces. J. Am. Chem. Soc. 1983, 105, 4481-4483. (8) Jordan, C. E.; Frey, B. L.; Kornguth, S.; Corn, R. M. Characterization of Poly-L-Lysine Adsorption onto Alkanethiol-Modified Gold Surfaces with Polarization-Modulation Fourier Transform Infrared Spectroscopy and Surface Plasmon Resonance Measurements. Langmuir 1994, 10, 3642-3648. (9) Frutos, A. G.; Liu, Q.; Thiel, A. J.; Sanner, A. M. W.; Condon, A. E.; Smith, L. M.; Corn, R. M. Demonstration of a Word Design Strategy for DNA Computing on Surfaces. Nucleic Acids Res. 1997, 25 (23), 47474757. (10) Liu, Q.; Frutos, A. G.; Thiel, A. J.; Corn, R. M.; Smith, L. M. DNA Computing on Surfaces-Encoding Information at the Single Base Level. J. Comput. Biol. 1998, 5 (2), 269-278.

Gillmor et al.

at room temperature, during which time the alkene of the SSMCC pyrrole group reacts with the thiol-modified DNA to tether the DNA strand to the surface at the 5′ terminus. The surface is rinsed with water and soaked for 1 h in a solution of 300 mM NaCl, 20 mM sodium phosphate, 2 mM EDTA, and 6.9 mM sodium dodecyl sulfate (2X SSPE/ 0.2% SDS) at room temperature, after which the sample is ready for hybridization.9,10 (5) Hybridization. Hybridization to the attached DNA probe molecules is accomplished by exposure to a 2 µM solution of 5′-fluorescein-labeled target oligonucleotides in 2X SSPE/0.2% SDS buffer. A 30 µL drop of the complement solution is placed onto the sample and then spread over the entire surface by placing a clean cover slip on top of the sample. Hybridization adsorption is allowed to proceed for 30 min, after which the sample is rinsed with water and is immersed in a solution of 2X SSPE/0.2% SDS buffer without the 5′-fluorescein-labeled complements for 10 min.9,10 (6) Surface Fluorescence Measurements. We use fluorescence to detect hybridization, with a Molecular Dynamics FluorImager 575. The sample is placed face down on top of a glass scanner tray, with a 30 µL droplet of buffer between the sample and the glass tray, and then scanned at 530 nm. A signal is measured on the slide at each level of chemistry to get background and noise figures. An image of the surface is obtained.9,10 (7) Subsequent Hybridizations. After the fluorescence signal is recorded, the sample is placed in 8.3 M urea solution at 37 °C for 20 min to remove the fluorescent complement and is washed thoroughly with water. The sample is then ready for a second hybridization for the detection of another target sequence.9,10 B. Materials. The chemicals, 11-mercaptoundecanoic acid (MUA) (Aldrich), n-octadecyl mercaptan (ODM) (Aldrich), poly-L-lysine hydrobromide (PL) (Sigma), sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SSMCC) (Pierce), urea (Biorad), and triethanolamine hydrochloride (TEA) (Sigma), were all used as received. Gold substrates were prepared by vapor deposition of 50 Å chromium followed by 1000 Å of gold onto a microscope slide cover (Fisher No. 2, 18 × 18 mm) or silicon surfaces (3 in. diameter, n-doped wafer, 1-10 Ω cm, (100), Wafer World). Millipore filtered water was used for all aqueous solutions and for rinsing. All oligonucleotides were synthesized on an ABI DNA synthesizer at the University of Wisconsin Biotechnology Center. Glen Research 5′-thiol modifier C-6, and ABI 6-FAM were used for 5′-thiolmodified and 5′-fluorescein-modified oligonucleotides, respectively. Before use, the oligonucleotides were purified by reverse-phase binary gradient elution HPLC (Shimadzu SCL-6A). However, if the HPLC column or the injecting needle has not been sufficiently cleaned, the next DNA sample that is run through the HPLC may have some slight contamination.11 We can see this contamination if we inject a sample of water through the HLPC directly after a run with a DNA sample. We see a peak that corresponds to the previous DNA sample. Through better design of the HLPC and sufficient cleaning, we can eliminate DNA solution contamination as a source of error. The 5′-thiol DNA solutions used on the surface attachment reactions had a DNA concentration of 1 mM in a pH 7.0, 100 mM TEA buffer unless otherwise noted. All DNA concentrations were verified using a HP8452A UV-vis spectrophotometer. DNA hybridization and rinsing employed a “2X SSPE/0.2% SDS” buffer that consisted of 300 (11) Qinghua Liu Unpublished data.

Templates for DNA Arrays

mM NaCl, 20 mM sodium phosphate, 2 mM EDTA, and 6.9 mM sodium dodecyl sulfate.9,10 C. Patterned-Surface Formation and DNA Attachment to Arrays. (1) Overview. Alkanethiol monolayers self-assemble to form robust, highly ordered films on a gold surface.7 The use of ω-functionalized alkanethiols such as MUA, whether commercially available or specifically synthesized, allows the surface reactivity to be tailored via the choice of terminal group. The DNA attachment takes advantage of both the ordered monolayer characteristic of alkanethiols and the selective reactivity of the terminal group. The selective removal of alkanethiols from the gold substrate can be achieved by exposure to UV light through a mask. After the gold surface has been chemically modified, the sample is exposed to UV light to cleave the gold-sulfur bond and selectively remove regions of the alkanethiol. Immersion in a solution of n-octyldecyl mercaptan (ODM), a methyl-terminated alkanethiol, makes the regions that had been exposed hydrophobic. The subsequent DNA attachment takes advantage of the hydrophilic/hydrophobic behavior of the patterned surface. (2) Formation of the Patterned Substrate. In Figure 1, the process of patterning a surface with alkanethiols is illustrated schematically. After a slide has been chemically modified with the attachment of MUA and PL (Figure 1a), the surface is exposed to 400 W ultraviolet light (xenon-mercury arc lamp, Oriel) for 1 h through a quartz mask (Figure 1b). UV light creates ozone that oxidizes the gold-sulfur bond.12 Rinsing the surface with ethanol for 30 s completely removes the alkanethiol in the exposed regions (Figure 1c) and leaves islands of hydrophilic MUA-PL regions and a background of bare gold. The slide is immersed for 5 min in a 1 mM ethanolic solution of ODM that attaches to the bare gold regions to create a hydrophobic matrix filling the space between the hydrophilic regions, and therefore a patterned surface (Figure 1d). The sample is then covered with SSMCC, which attaches to the amine group on the MUA-PL islands only, followed by the addition of DNA to the surface. The hybridization process is the same for a patterned surface as for an unpatterned one. When this patterned surface is imaged, the fluorescent signal reveals the DNA attachment and hybridization on the hydrophilic regions. This difference in behavior of the ODM and the MUAPL areas allows the selective adsorption of DNA to the array sites only. There are measurable differences in the surface, as characterized by the difference in contact angles as MUA, PL, SSMCC, and ODM are added to the surface. The contact angle measurements highlight the large difference in hydrophobicity between the MUA + PL + SSMCC areas, 41.5 ( 1.54°, and the ODM background, 91.8 ( 1.75° (Table 1). It is this barrier at the interface between the hydrophilic MUA + PL + SSMCC array elements and the hydrophobic background that defines areas where the DNA solutions can adsorb. An aqueous solution will wet the SSMCC surface while an aqueous solution will be repelled from the ODM surface. (3) Loading DNA onto an Array. To place DNA into the elements of our array, we use an automated mechanical loading system with three independent motorized axes and a PC-compatible computer to coordinate their motion. An AT6400 control board from Compumotor Division, Parker Hannifin Corp., translates the Turbo-Pascal 7.0 (12) . Tarlov, M. J.; Burgess, D. R. F.; Gillian, G. J. Am. Chem. Soc. 1993, 115, 5305-5306. Huang, J.; Dahlgren, D. A.; Hemminger, J. C. Photopatterning of Self-Assembled Alkanethiolate Monolayers on Gold: A Simple Monolayer Photoresist Utilizing Aqueous Chemistry Langmuir 1994, 10, 626-628.

Langmuir, Vol. 16, No. 18, 2000 7225

Figure 1. The process of patterning a surface. (a) A solid support coated with gold is placed into a solution of 11mercaptaoundecanoic acid (MUA) for at least 18 h. The surface is subsequently covered with poly-L-lysine (PL), which binds to the MUA, creating a hydrophilic surface. (b) When the MUAPL surface is exposed to UV light through a quartz mask, the gold-sulfur bond between the MUA and the gold surface is oxidized. (c) Those regions where the UV light was blocked by the mask are preserved. Rinsing the surface completely removes the MUA-PL from the gold surface in the exposed areas. (d) n-Octyldecyl mercaptan (ODM) fills the bare gold regions and transforms the exposed areas into hydrophobic regions. (e) The surface is then reacted with sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SSMCC) and finally with 5′-thiol-modified DNA. The SSMCC reacts only with the MUAPL regions of the surface, not to the ODM areas, and the 5′thiol-modified DNA only attaches to the SSMCC sites. The ODM regions are hydrophobic while the MUA-PL-SSMCC areas are hydrophilic. The DNA solutions are confined to the specific hydrophilic regions. Table 1. Contact Angles of Water on a Gold Surface That Has Been Chemically Modifieda surface

ΘH2O

(1) 11-mercaptoundecanoic acid (MUA) (2) MUA + poly-L-lysine (PL) (3) MUA + PL + SSMCC (4) n-octadecyl mercaptan (ODM)

58.1 ( 3.54° 34.3 ( 2.44° 41.5 ( 1.54° 91.8 ( 1.75°

a (1) A bare gold surface exposed to a solution of 11-mercaptoundecanoic acid (MUA) for 18 h, and after rinsing with water and ethanol. (2) Further exposure of the surface to a solution of polyL-lysine (PL) for 30 min. (3) Further exposure of the surface to sulfosuccinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate (SSMCC) for 30 min. (4) A bare gold sample exposed to a solution of n-octadecyl mercaptan (ODM) for 18 h.

program13 into the motion commands for the axes (x, DCI, 305 mm; y, 5000 series Daedal Positioning Systems and (13) . For a copy of the Turbo Pascal code, contact the authors.

7226

Langmuir, Vol. 16, No. 18, 2000

Controls, 102 mm; z, 5000 series Daedal Positioning Systems and Controls, 56 mm). To transfer the DNA solution onto the sample, the robot manipulates a 0.5 mm mechanical pencil tip and places it in a microtiter plate well and ∼0.5 µL of solution is transferred into the tip. After depositing ∼100-200 nL of solution through capillary action by touching the tip onto a hydrophilic portion on the sample, the robot rinses the tip for 10 s in a water bath, dries it in a nitrogen gas (N2(g)) stream for 10 s, and then touches the tip into another DNA solution on a microtiter plate to repeat the process. Finding an absolute home, with a magnetic marker without relying on the mechanical motion of the robot and before placement of a DNA solution onto the patterned surface, allows the robot to perform precise, repeatable motions. The smallest step size within the robot’s capability is specified to be ∼0.2 µm,14 but absolute precision is approximately ∼50 µm, likely due to backlash in the mechanical motion. The positioning capabilities of the robot used in these investigations limited the array density and size that we were able to load. (4) Hybridization. The process used for hybridization on a patterned surface is similar to hybridization on an unpatterned surface, as described above. The surface is flooded with a solution of complementary fluorescently labeled DNA strands and is then covered with a clean cover slip. The hydrophobic regions repel the solution and do not allow it to spread homogeneously over the surface. After a reaction time of 30 min, the slide is immersed in 2X SSPE/0.2% SDS at 37 °C for 5 min, and then the step is repeated with fresh SSPE/0.2% SDS at 37 °C, which removes any nonspecifically bound fluorescent DNA solution. (5) Surface Fluorescence Measurements. A patterned surface is treated exactly the same as an unpatterned surface in fluorescence measurements. The resolution is limited by the CCD camera pixel size, and as a consequence, the highest resolution of the FluorImager is 100 µm × 100 µm. While our smallest feature in the present data is much larger than this limit, we do see some pixelation in the images. III. Results The premise of using a hydrophilic/hydrophobic patterned surface to create arrays is that the chemical patterning aids to confine small quantities of the DNA solutions to prescribed areas, from which they do not bleed to contaminate other areas. The success of this premise depends on the degree to which a particular DNA solution prefers to go to and stay on the specific element onto which it is placed. This preference depends on surface free energy differences between the hydrophilic and hydrophobic regions. As we decrease the array element size, the magnitude of the difference in the surface free energies of the hydrophilic and hydrophobic regions becomes increasingly important. Using contact angles, we can determine the difference in interfacial free energies between the surface/air interface and the surface/water interface from Young’s equation:15 (14) . Buxton, E. C. Automation in Large Scale DNA Sequencing. Ph.D. Dissertation, University of WisconsinsMadison 1996, pp 97100. (15) . Brody, J. H.; Rose, R. M.; Wulff, J. The Structure and Properties of Materials: Thermodynamics of Structure; John Wiley and Sons: New York, 1964; p 55.

Gillmor et al.

γLV cos Θ ) γSV - γSL

(1)

where γLV, γSV, and γSL are the specific interfacial free energies for the water (solution)/air, surface/air, and surface/water interfaces at 25 °C, respectively. The value for γLV is 71.97 dyn/cm;16 using the measured contact angles, the difference of (γSV - γSL) is 53.9 dyn/cm for the hydrophilic MUA + PL + SSMCC surface and -2.25 dyn/ cm for the ODM surface. As already shown, an aqueous solution on ODM will have a tendency to bead up while an aqueous solution on an SSMCC surface will wet. We exploit these differences with our patterned surface. For a feature size of 500 µm, the hydrophilic and hydrophobic areas can be seen with the unaided eye in a humid environment when the surface is cooled. The larger the differences in (γSV - γSL) for the two array surfaces (hydrophilic/hydrophobic), the greater the driving force for the solution to remain on the hydrophilic region. We have fabricated and successfully loaded a 7 × 7 array of 500 µm × 500 µm squares. Figure 2 shows that DNA attaches on the energetically favorable regions. We see a well-formed array with discrete array sites. Loading the entire array with the same DNA sequence as in Figure 2 illustrates the effectiveness of the hydrophobic matrix and hydrophilic sites in containing DNA to regular sites without bleeding. The DNA sequence is 31 base pairs long, synthesized via CPG, and subsequently purified. For comparison, a 31 base pair long strand synthesized in situ would have a 20% fidelity. The image in Figure 2 is plotted using arbitrary fluorescence units. A value of 0 is a scan without a sample and a signal of 10 000 is the saturation limit of the detector. A scan of a sample with a MUA surface has an average signal of 157 ( 11 units, a MUA + PL surface has an average signal of 158 ( 13, and a MUA + PL + SSMCC surface has an average signal of 157 ( 12, indicating that for all purposes these surfaces are identical. For a patterned surface, there is no discernible difference in the fluorescence reading between the hydrophilic areas and the hydrophobic background even though the pattern is easily detectable by washing the sample with water or by cooling the surface in a humid environment. A scan of a patterned sample shows an average signal of 155 ( 13. The samples with single-stranded DNA without fluorophores attached to the patterned surface produce a signal of 158 ( 11 units. The intensity for a sample with the fluorescent complement is usually over 1000 ( 150. In Figure 2, this background signal of 155 has been subtracted. The error in this background, (13, is, of course, maintained in the signal that we show but is negligible compared to the magnitude of the signal and to the other sources of error described below. DNA solutions placed on the array elements remain confined to those sites and do not spread into adjacent array elements. A proof-of-principle experiment is illustrated in Figure 3. Two different DNA sequences, A and B (Figure 3e), have been placed on adjacent 500 µm squares in the arrangement shown in Figure 3a. After hybridization with both complementary sequences, all array sites on which DNA sequences had initially been placed fluoresce (Figure 3b). When the surface is hybridized with only one of the complementary sequences, signal is detected only from the corresponding location of that sequence in the array (Figure 3c,d). Figure 3 shows that we occasionally see a signal for elements in which we did not deliberately place any DNA (16) . Lide, D. The CRC Handbook of Chemistry and Physics, 74th ed.; CRC Press: Ann Arbor, MI. 1993; pp 6-144.

Templates for DNA Arrays

Figure 2. A DNA array viewed in fluorescence. (a) Forty nine array sites at which 31-base-long DNA strands have attached. Each array site is 500 µm × 500 µm and is separated by 1 mm center to center. 0.5 mM 5′-thiol-modified DNA in solution reacts with the hydrophilic array elements. Complementary strands, 5′-fluorescein-labeled oligonucleotides, hybridize with the single-stranded DNA in the array elements, allowing them to be fluorescently imaged. (b) The intensity of the fluorescence is color coded, in arbitrary units. The average value for the array elements is 876 ( 80. A filter has been applied of the data to remove noise from stray dust particles.

(Figure 3b) or the particular solution A or B. This signal is never higher than 20% of the minimum signal for a correct site. There are two possible reasons for the existence of these stray signals. One, inaccuracies in the loading procedure may have caused solutions to run to adjacent elements. All elements with adventitious signal lie next to ones in which we have deliberately placed DNA solution. Note that the hydrophobic regions between the hydrophilic elements are always free of signal. Our basic premise is thus sound. This first reason, however, cannot be the complete cause, because if it were, all elements showing a signal in Figure 3b should show a signal in either panel c or d of Figure 3. They do not. The second possibility is that the DNA solutions we loaded in the array are not entirely pure. If A contained some B, then one would observe an A signal in the sites nominally containing B. If B contained some A, then the reverse would be true. If either were true, we should consistently observe the error on all corresponding sites in panels c and d of Figure 3. We do not. The fact that all adventitious elements lie next to a deliberately loaded element suggests that some combination of the above causes must apply. We cannot, at this stage, determine which and thus conclude that our present experiments have a maximum uncertainty of 20% of the minimum real signal and as little as 8-10% of the average signal.

Langmuir, Vol. 16, No. 18, 2000 7227

Figure 3. Demonstration of selectivity of hybridization and isolation to array elements. (a) Two different 5′-thiol-modified DNA solutions (A and B) are placed in different elements of the array. (b) The array elements in which the DNA strands have formed duplexes with the fluorescein-tagged complements become visible in the fluorescent image. Complements of both A and B have both been added. (c) The fluorescent image of elements loaded with DNA sequence A is recorded, after the complements are removed from surface (b) with a urea wash, and the complement of A is added. (d) The fluorescent image of elements loaded with DNA sequence B is obtained, after washing off the complement of A and the complement of B is added. (f) The individual elements of (b), (c), and (d) have average values of 1251 ( 276. The figure shows that each array element can be individually loaded with a specific DNA solution and that the solution stays confined in the array site. A filter has been applied of the data to remove noise from stray dust particles.

As the feature size shrinks, the density of DNA sequences represented on a surface in an addressed-array fashion grows. We have been able to create arrays of 50 µm squares 100 µm apart center to center. At present, the limiting factor of this method is not the feature size but the loading mechanism; for example, 10 µm pixels spaced 20 µm apart present a formidable loading challenge. Repeatable motion in a highly precise automated tool and the ability to dispense volumes under a nanoliter in a controlled fashion are required. Also, while our present fluorescence imaging technique has a spatial resolution of 100 µm, higher-resolution fluorescence detection methods are available. In summary, the use of functionalized alkanethiols patterned on a gold surface to create hydrophobic and hydrophilic regions provides an efficient method to create small high-quality DNA arrays. Surface patterning is a highly flexible technique requiring one mask to create

7228

Langmuir, Vol. 16, No. 18, 2000

the array and simple modifications in the computercontrolled automated pin tool to place DNA solution in the array in a variety of different arrangements. The achievable array density is presently limited by the quality of the mechanical loading tool, but arrays with pixel sizes of the order of 50 µm are easily foreseeable, making a 1000 element array in 1 cm2 an achievable target. The application of such patterned surfaces is not limited to

Gillmor et al.

DNA arrays but can be modified to create a surface with a variety of terminal groups separated by an inert background for surface chemistry experimentation. Acknowledgment. This research was supported by NSF and DARPA. We thank J. Brockman, A. Frutos, J. Goomey, Q. Liu, and L. Wang for assistance. LA991026A