- Email: [email protected]
DNA Microarrays Based on Noncovalent Oligonucleotide Attachment and Hybridization in Two Dimensions
Analytical Biochemistry 292, 250 –256 (2001) doi:10.1006/abio.2001.5088, available online at http://www.idealibrary.com on
DNA Microarrays Based on Noncovalent Oligonucleotide Attachment and Hybridization in Two Dimensions Yuri Belosludtsev,* ,1 Bonnie Iverson,* Sergey Lemeshko,† Rick Eggers,* Rick Wiese,* Sandy Lee,* Tom Powdrill,* and Mike Hogan* ,† *Genometrix Inc., 3608 Research Forest Drive, Suite B7, The Woodlands, Texas 77381; and †Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030
Received December 7, 2000; published online April 17, 2001
Short oligonucleotide probes have been linked to a solid support by simple electrostatic adsorption onto a positively charged surface film. Attachment was obtained by microfluidic application of unmodified oligonucleotides in distilled water onto amino-silanized glass. It has been demonstrated that an extremely stable monolayer of oligonucleotide is obtained by this method, at a density of about 10 11 molecules/mm 2, which approaches the limit expected for a two-dimensional closest-packed array. Application of oligonucleotide by adsorption is followed by capping with acetic anhydride in the vapor phase, and then capping with succinic anhydride in solution to form a surface with weak negative charge. The capping method has been successfully employed for microarray fabrication and for the analysis of single nucleotide polymorphisms in the k-ras gene. The data reveal that, subsequent to capping, the adsorptive association of oligonucleotide to the surface yields a probe layer which is capable of single nucleotide base mismatch discrimination and high apparent binding affinity. © 2001 Academic Press Key Words: DNA microarrays; oligonucleotide attachment; two-dimensional hybridization.
The birth of DNA microarray technology and its ability to collect massive amounts of data in a short period of time has proven to be an important milestone in biotechnology (1, 2). Based upon the use of such twodimensional arrays of nucleic acid probes, large-scale genotype analysis and the analysis of gene expression at the mRNA level are becoming a reality. The key to these advances has been the development of immobilization chemistry for the spatially addressable attach1 To whom correspondence should be addressed. Fax: 281-4655002. E-mail: [email protected].
250
ment of DNA probes to the solid support, so as to form the desired microarray (3). In spite of the fact that adsorptive attachment of probes to fibrous nitrocellulose or nylon membranes has been used for many years as a standard hybridization technique (4), research in the microarray field has been concentrated on the development of covalent coupling of oligonucleotide or high-molecular-weight probes to planar surfaces. Such covalent coupling requires activation of the underlying planar surface with cross-linking reagents and/or modification of the DNA molecule with a reactive group (5–7). It has been shown previously that when a short DNA oligonucleotide is adsorbed to a surface by multiple constraining contacts, it becomes a poor hybridization probe due, presumably, to close proximity to the underlying surface, the attendant loss of configurational freedom, and the related loss of the capacity to form the double helix with a cognate target at the surface (8). On the other hand, if only a few adsorptive contacts are made between probe and the surface, it might be expected that the noncovalently attached DNA molecule becomes unstable and susceptible to rapid removal from the surface during hybridization or washing (9). Therefore, to date the use of oligonucleotides as probes for DNA microarray fabrication has been based upon covalent attachment strategies. Here we describe an extremely simple and reproducible method for the production of DNA arrays, employing adsorptive, noncovalent attachment of oligonucleotide probes to an amino-silanized glass surface. The method consists of two steps: deposition of unmodified oligonucleotide probes as a water solution onto an amino-silanized glass substrate, followed by drying and capping of “unused” amines, i.e., those amines not involved in direct association with adsorbed probe molecules. After drying and capping in that way, we show 0003-2697/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
251
FIG. 1. Chaperone probe binding. Solution state binding of chaperone probe to target facilitates opening of local secondary structure and allows capture probe hybridization.
that the attached oligonucleotides cannot be removed from the surface under standard hybridization and washing conditions, including high salt and high pH treatments. In spite of the fact that adsorbed oligonucleotide is presumably bound via multiple contacts to the surface and therefore may have lost configurational freedom required to form a perfect double helix with its cognate target, we show that the product of such adsorptive coupling, followed by judicious capping, displays specificity for duplex formation which is as high as that seen in a standard solution state hybridization reaction, or for surface hybridization to probes linked covalently to the surface at a single point.
acetic anhydride (Aldrich) in 3 ml of DMF 2 in a vacuum oven at 22 in. Hg for 1 h at 50°C with “printed” amino slides assembled in a rack. The acetylated amino-derivatized slides were then capped with succinic anhydride by dipping slides in a tank with 0.5 M succinic anhydride in DMF at room temperature for 1 h. Slides were cleaned by washing in acetone (three times), in distilled water (twice), and again in acetone (twice). As a quality control, 30 l of fluorescent-labeled oligonucleotide in hybridization buffer (see below) was deposited onto the slide surface: if no background was observed after 15 min, slides were considered ready for hybridization experiments. Amplicon Preparation
MATERIALS AND METHODS
Microarray Preparation Substrates used were silica slides (Erie) cleaned in an ultrasonic bath with detergent (2 min) followed by washing with distilled water (three times) and methanol (twice) and drying (30 min at 40°C). Slides were then silanized with 3-aminopropyltrimethoxysilane (Aldrich) or 3-glycidoxypropyltrimethoxysilane (Aldrich) in vapor phase by placing a petri dish with a mixture of 3 ml of corresponding silane and 3 ml of p-xylene (Aldrich) in a vacuum oven at 25 in. Hg overnight at 70 – 80°C with cleaned slides assembled in a rack. Oligonucleotides, purchased from Midland Certified Reagent Co. (Midland, TX) were deposited as 25 nl solutions (in 15 mM NaOH at 200 M for epoxy-silanized surface and in distilled water at 5 M for aminosilanized slides) upon the silanized surface using a Hamilton Micro Lab 2200 robot. After “printing” oligonucleotides, epoxy-derivatized slides were ready for processing (hybridization studies with target oligonucleotides or amplicon). After being dried (15 min at 40°C or overnight at room temperature) amino-derivatized slides were capped with acetic anhydride in vapor phase by placing a petri dish with a solution of 3 ml of
The 152-bp K-ras amplicons were obtained by the polymerase chain reaction (PCR). Wild-type amplicon (K-ras 1) was obtained by amplification of a commercial genomic DNA source (Sigma). K-ras 2 and K-ras 7 mutants were obtained by amplification of human genomic DNA from cell lines A549 and SW480, respectively. The PCR protocol was the following: one pre-PCR cycle 94°C for 12 min, 60°C for 1 min, and 72°C for 1 min; 35 PCR cycles 95°C for 1 min, 57°C for 1 min, 72°C for 1 min; hold cycle 72°C for 7 min, 4°C hold. PCR primers for k-ras amplicons were labeled with digoxigenin at their 5⬘ ends and had the following sequences: 5⬘-dig-actgaatataaacttgtggtagttggacct-3⬘ and 5⬘-dig-tcaaagaatggtcctgcacc-3⬘. A chaperone oligonucleotide, labeled with digoxigenin, was used in hybridization solutions to enhance hybridization signals by binding in the solution phase with the target immediately proximal to the target–probe hybridization site. This chaperone binding facilitates holding the target into a locally opened state, thereby minimizing undesired side effects of target secondary structure (10) (Fig. 1). In this particular case, in the absence of a chaperone, the PCR product forms a stable secondary structure which pre2
Abbreviation used: DMF, dimethyl formamide.
252
BELOSLUDTSEV ET AL.
vents it from hybridization to the surface capture probe. The chaperone oligonucleotide had the following sequence: 5⬘-taggcaagagtgccttgacgatac-3⬘-dig. Oligonucleotide Probes K-ras amplicons used in this study had different point mutations in codon 12, so specific complementary oligonucleotides were designed to serve as microarray capture probes. They had the following sequences: k1, 5⬘-gacctggtggcg-3⬘; k2, 5⬘-gacctagtggcg-3⬘; k3, 5⬘-gacttggtggcg-3⬘; k4, 5⬘-gacctcgtggcg-3⬘; k5, 5⬘-gacctgatggcg3⬘; k6, 5⬘-gacctgctggcg-3⬘; k7, 5⬘-gacctgttggcg-3⬘. Hybridization and Detection Amplicon analysis. Prehybridization solution, containing 150 mM sodium citrate (with respect to sodium ion concentration), 5⫻ Denhardt’s solution, pH 8.0, was applied to the array for at least 10 min. It was vacuumed off and hybridization solution (1 nM amplicon, 0.1 M chaperone, 150 mM sodium citrate in respect to sodium, 5⫻ Denhardt’s solution, pH 8.0) was applied to the array. In these studies, only amplicons complementary to capture probes k1, k2, and k7 have been used (i.e., K-ras 1, K-ras 2, K-ras 7). After 2 h of hybridization, the array was washed two times in 100 mM sodium citrate with respect to sodium, 10 min each, followed by a brief rinse in 1⫻ SSC. The digoxigenin-labeled amplicon was detected using anti-digoxigenin antibody linked to alkaline phosphatase (Boehringer Mannheim) at 1:1000 dilution in the blocking buffer from the ELF-97 mRNA In Situ Hybridization Kit (Molecular Probes), followed by washing in buffer A from the same kit and by application of ELF as described in the kit, which is a substrate for alkaline phosphatase (11). After being cleaved by alkaline phosphatase, ELF molecules precipitate and become fluorescent under UV excitation. The fluorescence intensities were detected with an Alpha Imager 2000 apparatus and processed using GeneView 1.0 (Genometrix, Inc.), Microsoft Excel 97, and Sigma Plot 3.0 software. Short synthetic targets. For chemically capped microarrays, oligonucleotide targets were hybridized as for amplicons. For uncapped microarrays, blocking was obtained with Denhardt’s solution. 5⫻ Denhardt’s solution was applied to the array in water, followed by a prehybridization solution, containing 150 mM sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5, for at least 10 min. For both methods of surface preparation, prehybridization solution was vacuumed off and hybridization solution ( 33P-labeled CGCCACCAGGTC, 150 mM sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5) was applied to the array. After 2 h of hybridization, the array was washed two times in 150 mM
FIG. 2. Chemistry of oligonucleotide attachment. (A) Covalent binding to an epoxy-silanized surface (prepared as in Ref. 7), (B) adsorptive attachment to an amino-silanized surface.
sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5, and the hybridization signal was detected with a Cyclone phosphorimager (Packard Instruments), as described above. Oligonucleotide Radiolabeling The 12-mer CGCCACCAGGTC was 33P labeled using polynucleotide kinase and standard protocol (4). RESULTS AND DISCUSSION
We have compared adsorptive vs covalent strategies for oligonucleotide attachment to a planar glass substrate. Covalent attachment has been obtained by reaction of a 5⬘-amino-modified oligonucleotide with an epoxysilanized surface (Fig. 2A), a well-known method for covalent attachment to surfaces (7), which yields a terminal secondary amine linkage. Adsorptive attachment has been obtained by application of an unmodified oligonucleotide in water to an amino-derivatized surface (Fig. 2B). For covalent attachment, amino-modified oligonucleotides were placed upon the epoxy-silanized surface by deposition of 25 nl of oligonucleotide at 200 M in 15 mM NaOH using a Hamilton Micro Lab 2200 robot. A high concentration of the applied oligonucleotide is required in this covalent coupling reaction to compensate for the slow reaction of primary amines with an
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
epoxide moiety in basic solution. Adsorptive attachment chemistry was obtained by depositing 25 nl of oligonucleotide at 5 M in distilled water. We have determined the saturability and stability of oligonucleotide adsorption to the amino-modified surface. Twenty-five-nanoliter spots of a radiolabeled 12mer (a synthetic target complimentary to the k1 12mer probe) were deposited at concentrations from 0.01 to 40 M. Arrays were formed, one each per well, in an 8 ⫻ 12-well microtiter format. The excess of deposited oligonucleotide was then washed off with repeated water washes and bound surface density was determined with a storage phosphorimager. This test revealed a saturating probe density of 3 ⫻ 10 11 molecules/mm 2 (Fig. 3A). Assuming that one 12-mer oligonucleotide occupies a surface area of about 600 Å 2 (that is, 10 Å width and 5 Å rise per base repeat), we calculate that the maximum number of oligonucleotide molecules needed to form a closely packed monolayer on the surface would be 2 ⫻ 10 11 molecules/mm 2. Thus, comparison of experimental and calculated density gives evidence that a densely packed monolayer of oligonucleotides is forming during the adsorption process. We have repeated a similar adsorption analysis with a 36-mer probe. Maximum saturable probe density with the 36-mer was found to be 0.6 ⫻ 10 11 molecules/mm 2. Comparison to the density calculated for a closest packed 36-mer monolayer (0.6 ⫻ 10 11 molecules/ mm 2) suggests that a densely packed probe film can be formed for probes as long as 36 bases. We have found that after drying to achieve a tightly adsorbed, closely packed monolayer of oligonucleotide on the surface, subsequent treatment with high salt concentrations (3 M NaCl or 1 M sodium phosphate) or high pH (up to 11.5) near room temperature did not displace the attached radiolabeled probe monolayer. The loss of radioactivity was less than 10% in both instances. Noticeable removal (more than 50%) of radiolabeled probe from the adsorbed monolayer was found only during washes at elevated temperatures (95°C) with high salt (1 M sodium phosphate). For a positively charged surface covered with a monolayer of adsorbed probe, there are two independent processes that would cause a solution state target nucleic acid to bind to the surface: adsorptive electrostatic attraction to the cationic surface and duplex formation with the attached probes. If there were a high density of amines on the surface, not involved in direct interaction with probe, electrostatic attraction between target and the surface could become a dominating force, thereby overwhelming the contribution of duplex formation. Poor base sequence selectivity would be seen in that instance. Thus, we have set out to discover ways to “cap” the excess of unused surface cation, while retaining a tight adsorptive interaction between probe and the surface.
253
FIG. 3. (A) Titration of the amino-silanized surface with a 12-mer oligonucleotide. (Filled circles) oligonucleotides deposited and then washed. (Open circles) oligonucleotides deposited on the surface and quantified without washing. (B) Binding curves for 33P-labeled k-ras 1 synthetic 12-mer hybridized to k-ras probes that were attached to an amino-silanized surface via adsorption. Chemical capping (circles and triangles) or adsorptive capping with Denhardt’s solution (squares) was employed prior to hybridization. Hybridization buffer was (150 mM of sodium bicarbonate) at pH 9.5 (circles and squares) or pH 7.5 (triangles). A perfectly matched k-ras 1 target to k-ras 1 probe pairing is shown in filled symbols. A representative single base mismatch (k-ras 1 target to k-ras 3 probe) is shown in open symbols. T is the number of target molecules bound/mm 2 and P is the number of probe molecules/mm 2. The following binding function for homogeneous and noninteracting sites was used for fitting of the binding curves: Y ⫽ n*k[T]/(1 ⫹ k[T]), where Y ⫽ T/P is the bound target– probe ratio, [T] is free target concentration in solution, k is the affinity constant, and the value of Y is at saturation. The parameters k and n for the binding curves are shown in Table 1.
Covalent Capping In initial experiments, using noncovalent capping by prehybridization at pH 7.5 with 5⫻ Denhardt’s solution, amplicon interaction with residual positive charge gave rise to poor specificity and high background. Therefore, covalent capping reactions were investigated as an alternative. We used acetic anhydride
254
BELOSLUDTSEV ET AL.
Adsorptive Capping
TABLE 1
Binding Parameters for Synthetic 12-mer k-ras Target (See Fig. 3) Conditions k1 target:k1 probe, adsorptive capping, pH 9.5 k1 target:k1 probe, chemical capping, pH 8.0 k1 target:k1 probe, chemical capping, pH 9.5
n, (T/P) max
k/10 ⫺6 M dissociation constant
0.8
1.0
0.5
1.1
0.5
1.8
(at 0.5 M in DMF) for acetylation of amino-silanized surfaces after probe coupling. Capping reactions were carried out at room temperature for 1 h. Due to incomplete covalent capping, the above reactions yielded a slightly positive surface which still produced some nonspecific target binding at pH 7.5. Therefore, we decided to make neutral, slightly negatively, and highly negatively charged surfaces to determine the optimal covalent capping protocol yielding specificity of hybridization. A neutral surface was obtained by capping with acetic anhydride in the vapor phase at 50°C and 22 in. of Hg in a vacuum oven. Capping reactions were carried out 1 h and overnight. It was found that the overnight reaction significantly reduced electrostatic attraction and improved specificity. However, the 1-h reaction product retained a slight positive charge and gave poor specificity for amplicon targets. To make a slightly negatively charged surface, we used arrays that had been first capped with acetic anhydride, followed by a second capping with 0.5 M succinic anhydride in DMF at room temperature for 1 h. We found that such “double capped” arrays gave very good amplicon target binding specificity. A highly negatively charged surface was prepared by a one-step reaction with succinic anhydride, as described above. However, this surface was found to be too negatively charged to support measurable hybridization. Therefore, to achieve the best amplicon target binding specificity, we used the two-step capping procedure with acetic anhydride in the vapor phase at 22 in. Hg for 1 h at 50°C, followed by 0.5 M succinic anhydride in DMF at room temperature for 1 h. In the first acetylation step, an almost quantitative transformation of free amines into neutral amides was achieved (background or nonspecific binding of target amplicon to the surface was decreased almost to 0). By the second acylation step (with succinic anhydride) the remaining amines were transformed into their acylated derivatives. This introduction of carboxyl-groups imparted a slight negative charge and yielded surfaces capable of high amplicon target binding specificity.
To minimize the free energy of nonspecific target association with the surface, while retaining the capacity to adsorb oligonucleotide probe, we have tested the utility of a traditional blocking polymer mix: 1 mg/ml each of Ficoll, polyvinyl pyrrolidone, and bovine serum albumin (5⫻ Denhardt’s solution) as a coating to effect a reduction of electrostatic interaction between unused surface amines and solution state target. Additional fine-tuning of target interaction was obtained by an increase of solution state pH, reducing the surface charge as the pK of surface bound amine groups is approached. Thus, we have found that a prehybridization step at 150 mM sodium carbonate buffer at pH 9.5 and 5⫻ Denhardt’s solution, followed by hybridization in the same buffer produces a surface capable of high selectivity and high-affinity nucleic acid hybridization. We have compared these two methods of capping by performing analytical hybridization of a radiolabeled synthetic 12-mer k-ras gene target to a microarray containing its Watson–Crick complement (k1, Materials and Methods) and a related probe which differed by a single base change (k3). As seen in Fig. 3B (middle curves), the covalent method of capping has given rise to a binding isotherm with an apparent dissociation coefficient near to 1 ⫻ 10 ⫺8 M, reaching saturation around 0.5 targets bound per probe equivalent. These data are not significantly affected by pH in the range from 7.0 to 9.5. For comparison, at pH 9.5, and in the presence of Denhardt’s solution as the blocking agent, the measured dissociation constant is also seen to be about 1 ⫻ 10 ⫺8 M (Fig. 3B, upper curve). This result suggests that, as expected, the covalent capping process had not altered the nucleic acid probe in any way that could be detected via hybridization. At the highest applied k-ras target concentration, the limiting capacity to bind a synthetic 12-mer target appears to saturate at a slightly higher value for noncovalent blocking with Denhardt’s (0.8 targets bound per probe equivalent, compared to about 0.5), but considering the standard error of the data, this apparent difference is not significant. That secondary consideration aside, the data suggest that the two methods of excess surface charge neutralization are seen to be nearly equivalent in terms of target binding affinity and binding capacity per unit area. For both methods of capping, the capacity to form duplex with a short synthetic target is characterized by single nucleotide mismatch discrimination (Fig. 3B, lower curves). We have used the above adsorptive probe attachment chemistry (with covalent capping) to perform single base polymorphism analysis on the 152-bp PCR product and have compared the results thus obtained with arrays prepared using covalent attachment of
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
FIG. 4. (A) Single nucleotide polymorphism analysis with electrostatically attached probes. (B) Single nucleotide polymorphism analysis with covalently attached probes.
terminal amine-modified probes to an epoxysilanecoated surface (Fig. 2). Results obtained using amplicon targets derived from the k-ras gene show that specificity achieved using adsorptive probe binding followed by chemical capping (Fig. 4A) is very much comparable in affinity and specificity to the best that could be obtained with covalently attached probes on a neutral epoxysilane surface (Fig. 4B). As seen in both instances, the relatively stable G–T mismatch (K-ras 1 target amplicon with k3 probe) generates signals that are diminished by a factor of 10 relative to perfectly matched 12-bp pairing (K-ras 1 target with k1 probe). Discrimination for all other single base mismatches is measured to be significantly greater than a factor of 25. CONCLUSIONS
There are several practical advantages to the current method of microarray formation by adsorptive probe attachment. First, the affinity and selectivity of duplex formation are at least as good as the best that
255
can be obtained by traditional methods of probe attachment. Second, the adsorptive method requires the use of inexpensive unmodified probes, at an applied concentration that is at least five times lower than that required for the currently available covalent attachment chemistries. As a result, the probe cost of manufacture is reduced at least 10-fold. Finally, because the probe is printed in distilled water (which denatures nucleic acid secondary structure) and because binding is adsorptive, which tends to maximize surface contacts, probes affixed to a surface in this way may tend to show smaller artifacts due to intramolecular folding interactions. The mechanism of a surface-directed nucleic acid condensation has been suggested in previous literature (12, 13), but little was known about the structure of the resulting nucleic acid layer or its capacity to bind target nucleic acid. Here, it has been shown that subsequent to an appropriate capping reaction, target binding to such probes displays a high level of sequence selectivity and high overall binding affinity, thereby confirming that a base-paired duplex is being formed. Yet, the high stability of the probe interaction with the surface suggests that probes bound to the surface in this way must be constrained by multiple contacts. Thus it can be asked how a tight (nearly irreversibly) adsorbed probe layer might be capable of tight, sequence-selective duplex formation with a high-molecular-weight nucleic acid target, especially in light of previous literature (8), which had suggested that such adsorbed probes would be unavailable to bind a cognate target due to such limitations. As a partial resolution of that dilemma, we suggest that duplex can still form with such probes because, on a weakly cationic aminosilane surface, probes are not bound rigidly to the underlying surface, but are instead “trapped” upon the planar surface in a potential well, unable to diffuse off it, but relatively free to engage in rotation and other reorienting motions upon a surface that are necessary to form a double helix with a cognate DNA or RNA target. As such, the two-dimensional probe layer has properties that approximate those of liquid crystal, with the capacity for nearly each probe to form duplex within the liquid crystal layer. Alternatively, it may be formally possible that structures approximating a ribbon-like, nonhelical duplex may be forming at the surface. Such structures have been reported in different contexts, such as hybridization at air–water interfaces (14). Work is in progress to resolve these potential configurations. ACKNOWLEDGMENT This work was funded in part by a Program Project grant from the National Cancer Institute.
256
BELOSLUDTSEV ET AL.
REFERENCES 1. Collins, F. S. (1999) Microarrays and macroconsequences. Nat. Genet. Suppl. 21, 2. 2. Lander, E. S. (1999) Array of hope. Nat. Genet. Suppl. 21, 3– 4. 3. Ramsay, G. (1998) DNA chips: State-of-the art. Nat. Biotechnol. 16, 40 – 44. 4. Ausubel, F. M., et al. (1997) Current Protocols in Molecular Biology Units 2.9 and 3.10. Wiley, New York. 5. Lee, G. U., Chrisey, L. A., and Colton, R. J. (1994) Direct measurement of the forces between complimentary strands of DNA. Science 266, 771–773. 6. Guo, Z., Guilfoyle, R. A., Thiel, A. J., Wang, R., and Smith, L. M. (1994) Direct fluorescence analysis of genetic polymorphisms by hybridization with oligonucleotide arrays on glass supports. Nucleic Acids Res. 22, 5456 –5465. 7. Eggers, M., et al. (1994) A microchip for quantitative detection of molecules utilizing luminescent and radioisotope reporter groups. BioTechniques 17, 516 –524. 8. Duggan, D. J., Bittner, M., Chen, Y., Meltzer, P., and Trent, J. M.
9.
10.
11.
12.
13.
14.
(1999) Expression profiling using cDNA microarrays. Nat. Genet. Suppl. 21, 10 –14. Chrisey, L. A., Lee, G. U., and O’Ferrall, C. E. (1996) Covalent attachment of synthetic DNA to self-assembled monolayer films. Nucleic Acids Res. 24, 3031–3039. Maldonado-Rodriguez, R., Espinosa-Lara, M., Calixto-Suarez, A., Beattie, W. G., and Beattie, K. L. (1999) Hybridization of glasstethered oligonucleotide probes to target strands preannealed with labeled auxiliary oligonucleotides. Mol. Biotechnol. 11, 1–12. Haugland, R. P. (1996) Handbook of Fluorescent Probes and Research Chemicals, 6th ed., pp. 117–120. Molecular Probes, Inc., Eugene, OR. Fang, Y., and Hoh, J. H. (1998) Surface-directed DNA condensation in the absence of soluble multivalent cations. Nucleic Acids Res. 26, 588 –593. Balladur, V., Theretz, A., and Mandrand, B. (1997) Determination of the main forces driving DNA oligonucleotide adsorption onto aminated silica wafers. J. Colloid Interface Sci. 194, 408 – 418. Ebara, Y., Mizutani, K., and Okahata, Y. (2000) DNA hybridization at the air–water interface. Langmuir 16, 2416 –2418.
DNA Microarrays Based on Noncovalent Oligonucleotide Attachment and Hybridization in Two Dimensions Yuri Belosludtsev,* ,1 Bonnie Iverson,* Sergey Lemeshko,† Rick Eggers,* Rick Wiese,* Sandy Lee,* Tom Powdrill,* and Mike Hogan* ,† *Genometrix Inc., 3608 Research Forest Drive, Suite B7, The Woodlands, Texas 77381; and †Baylor College of Medicine, One Baylor Plaza, Houston, Texas 77030
Received December 7, 2000; published online April 17, 2001
Short oligonucleotide probes have been linked to a solid support by simple electrostatic adsorption onto a positively charged surface film. Attachment was obtained by microfluidic application of unmodified oligonucleotides in distilled water onto amino-silanized glass. It has been demonstrated that an extremely stable monolayer of oligonucleotide is obtained by this method, at a density of about 10 11 molecules/mm 2, which approaches the limit expected for a two-dimensional closest-packed array. Application of oligonucleotide by adsorption is followed by capping with acetic anhydride in the vapor phase, and then capping with succinic anhydride in solution to form a surface with weak negative charge. The capping method has been successfully employed for microarray fabrication and for the analysis of single nucleotide polymorphisms in the k-ras gene. The data reveal that, subsequent to capping, the adsorptive association of oligonucleotide to the surface yields a probe layer which is capable of single nucleotide base mismatch discrimination and high apparent binding affinity. © 2001 Academic Press Key Words: DNA microarrays; oligonucleotide attachment; two-dimensional hybridization.
The birth of DNA microarray technology and its ability to collect massive amounts of data in a short period of time has proven to be an important milestone in biotechnology (1, 2). Based upon the use of such twodimensional arrays of nucleic acid probes, large-scale genotype analysis and the analysis of gene expression at the mRNA level are becoming a reality. The key to these advances has been the development of immobilization chemistry for the spatially addressable attach1 To whom correspondence should be addressed. Fax: 281-4655002. E-mail: [email protected].
250
ment of DNA probes to the solid support, so as to form the desired microarray (3). In spite of the fact that adsorptive attachment of probes to fibrous nitrocellulose or nylon membranes has been used for many years as a standard hybridization technique (4), research in the microarray field has been concentrated on the development of covalent coupling of oligonucleotide or high-molecular-weight probes to planar surfaces. Such covalent coupling requires activation of the underlying planar surface with cross-linking reagents and/or modification of the DNA molecule with a reactive group (5–7). It has been shown previously that when a short DNA oligonucleotide is adsorbed to a surface by multiple constraining contacts, it becomes a poor hybridization probe due, presumably, to close proximity to the underlying surface, the attendant loss of configurational freedom, and the related loss of the capacity to form the double helix with a cognate target at the surface (8). On the other hand, if only a few adsorptive contacts are made between probe and the surface, it might be expected that the noncovalently attached DNA molecule becomes unstable and susceptible to rapid removal from the surface during hybridization or washing (9). Therefore, to date the use of oligonucleotides as probes for DNA microarray fabrication has been based upon covalent attachment strategies. Here we describe an extremely simple and reproducible method for the production of DNA arrays, employing adsorptive, noncovalent attachment of oligonucleotide probes to an amino-silanized glass surface. The method consists of two steps: deposition of unmodified oligonucleotide probes as a water solution onto an amino-silanized glass substrate, followed by drying and capping of “unused” amines, i.e., those amines not involved in direct association with adsorbed probe molecules. After drying and capping in that way, we show 0003-2697/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
251
FIG. 1. Chaperone probe binding. Solution state binding of chaperone probe to target facilitates opening of local secondary structure and allows capture probe hybridization.
that the attached oligonucleotides cannot be removed from the surface under standard hybridization and washing conditions, including high salt and high pH treatments. In spite of the fact that adsorbed oligonucleotide is presumably bound via multiple contacts to the surface and therefore may have lost configurational freedom required to form a perfect double helix with its cognate target, we show that the product of such adsorptive coupling, followed by judicious capping, displays specificity for duplex formation which is as high as that seen in a standard solution state hybridization reaction, or for surface hybridization to probes linked covalently to the surface at a single point.
acetic anhydride (Aldrich) in 3 ml of DMF 2 in a vacuum oven at 22 in. Hg for 1 h at 50°C with “printed” amino slides assembled in a rack. The acetylated amino-derivatized slides were then capped with succinic anhydride by dipping slides in a tank with 0.5 M succinic anhydride in DMF at room temperature for 1 h. Slides were cleaned by washing in acetone (three times), in distilled water (twice), and again in acetone (twice). As a quality control, 30 l of fluorescent-labeled oligonucleotide in hybridization buffer (see below) was deposited onto the slide surface: if no background was observed after 15 min, slides were considered ready for hybridization experiments. Amplicon Preparation
MATERIALS AND METHODS
Microarray Preparation Substrates used were silica slides (Erie) cleaned in an ultrasonic bath with detergent (2 min) followed by washing with distilled water (three times) and methanol (twice) and drying (30 min at 40°C). Slides were then silanized with 3-aminopropyltrimethoxysilane (Aldrich) or 3-glycidoxypropyltrimethoxysilane (Aldrich) in vapor phase by placing a petri dish with a mixture of 3 ml of corresponding silane and 3 ml of p-xylene (Aldrich) in a vacuum oven at 25 in. Hg overnight at 70 – 80°C with cleaned slides assembled in a rack. Oligonucleotides, purchased from Midland Certified Reagent Co. (Midland, TX) were deposited as 25 nl solutions (in 15 mM NaOH at 200 M for epoxy-silanized surface and in distilled water at 5 M for aminosilanized slides) upon the silanized surface using a Hamilton Micro Lab 2200 robot. After “printing” oligonucleotides, epoxy-derivatized slides were ready for processing (hybridization studies with target oligonucleotides or amplicon). After being dried (15 min at 40°C or overnight at room temperature) amino-derivatized slides were capped with acetic anhydride in vapor phase by placing a petri dish with a solution of 3 ml of
The 152-bp K-ras amplicons were obtained by the polymerase chain reaction (PCR). Wild-type amplicon (K-ras 1) was obtained by amplification of a commercial genomic DNA source (Sigma). K-ras 2 and K-ras 7 mutants were obtained by amplification of human genomic DNA from cell lines A549 and SW480, respectively. The PCR protocol was the following: one pre-PCR cycle 94°C for 12 min, 60°C for 1 min, and 72°C for 1 min; 35 PCR cycles 95°C for 1 min, 57°C for 1 min, 72°C for 1 min; hold cycle 72°C for 7 min, 4°C hold. PCR primers for k-ras amplicons were labeled with digoxigenin at their 5⬘ ends and had the following sequences: 5⬘-dig-actgaatataaacttgtggtagttggacct-3⬘ and 5⬘-dig-tcaaagaatggtcctgcacc-3⬘. A chaperone oligonucleotide, labeled with digoxigenin, was used in hybridization solutions to enhance hybridization signals by binding in the solution phase with the target immediately proximal to the target–probe hybridization site. This chaperone binding facilitates holding the target into a locally opened state, thereby minimizing undesired side effects of target secondary structure (10) (Fig. 1). In this particular case, in the absence of a chaperone, the PCR product forms a stable secondary structure which pre2
Abbreviation used: DMF, dimethyl formamide.
252
BELOSLUDTSEV ET AL.
vents it from hybridization to the surface capture probe. The chaperone oligonucleotide had the following sequence: 5⬘-taggcaagagtgccttgacgatac-3⬘-dig. Oligonucleotide Probes K-ras amplicons used in this study had different point mutations in codon 12, so specific complementary oligonucleotides were designed to serve as microarray capture probes. They had the following sequences: k1, 5⬘-gacctggtggcg-3⬘; k2, 5⬘-gacctagtggcg-3⬘; k3, 5⬘-gacttggtggcg-3⬘; k4, 5⬘-gacctcgtggcg-3⬘; k5, 5⬘-gacctgatggcg3⬘; k6, 5⬘-gacctgctggcg-3⬘; k7, 5⬘-gacctgttggcg-3⬘. Hybridization and Detection Amplicon analysis. Prehybridization solution, containing 150 mM sodium citrate (with respect to sodium ion concentration), 5⫻ Denhardt’s solution, pH 8.0, was applied to the array for at least 10 min. It was vacuumed off and hybridization solution (1 nM amplicon, 0.1 M chaperone, 150 mM sodium citrate in respect to sodium, 5⫻ Denhardt’s solution, pH 8.0) was applied to the array. In these studies, only amplicons complementary to capture probes k1, k2, and k7 have been used (i.e., K-ras 1, K-ras 2, K-ras 7). After 2 h of hybridization, the array was washed two times in 100 mM sodium citrate with respect to sodium, 10 min each, followed by a brief rinse in 1⫻ SSC. The digoxigenin-labeled amplicon was detected using anti-digoxigenin antibody linked to alkaline phosphatase (Boehringer Mannheim) at 1:1000 dilution in the blocking buffer from the ELF-97 mRNA In Situ Hybridization Kit (Molecular Probes), followed by washing in buffer A from the same kit and by application of ELF as described in the kit, which is a substrate for alkaline phosphatase (11). After being cleaved by alkaline phosphatase, ELF molecules precipitate and become fluorescent under UV excitation. The fluorescence intensities were detected with an Alpha Imager 2000 apparatus and processed using GeneView 1.0 (Genometrix, Inc.), Microsoft Excel 97, and Sigma Plot 3.0 software. Short synthetic targets. For chemically capped microarrays, oligonucleotide targets were hybridized as for amplicons. For uncapped microarrays, blocking was obtained with Denhardt’s solution. 5⫻ Denhardt’s solution was applied to the array in water, followed by a prehybridization solution, containing 150 mM sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5, for at least 10 min. For both methods of surface preparation, prehybridization solution was vacuumed off and hybridization solution ( 33P-labeled CGCCACCAGGTC, 150 mM sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5) was applied to the array. After 2 h of hybridization, the array was washed two times in 150 mM
FIG. 2. Chemistry of oligonucleotide attachment. (A) Covalent binding to an epoxy-silanized surface (prepared as in Ref. 7), (B) adsorptive attachment to an amino-silanized surface.
sodium bicarbonate, 5⫻ Denhardt’s solution, pH 9.5, and the hybridization signal was detected with a Cyclone phosphorimager (Packard Instruments), as described above. Oligonucleotide Radiolabeling The 12-mer CGCCACCAGGTC was 33P labeled using polynucleotide kinase and standard protocol (4). RESULTS AND DISCUSSION
We have compared adsorptive vs covalent strategies for oligonucleotide attachment to a planar glass substrate. Covalent attachment has been obtained by reaction of a 5⬘-amino-modified oligonucleotide with an epoxysilanized surface (Fig. 2A), a well-known method for covalent attachment to surfaces (7), which yields a terminal secondary amine linkage. Adsorptive attachment has been obtained by application of an unmodified oligonucleotide in water to an amino-derivatized surface (Fig. 2B). For covalent attachment, amino-modified oligonucleotides were placed upon the epoxy-silanized surface by deposition of 25 nl of oligonucleotide at 200 M in 15 mM NaOH using a Hamilton Micro Lab 2200 robot. A high concentration of the applied oligonucleotide is required in this covalent coupling reaction to compensate for the slow reaction of primary amines with an
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
epoxide moiety in basic solution. Adsorptive attachment chemistry was obtained by depositing 25 nl of oligonucleotide at 5 M in distilled water. We have determined the saturability and stability of oligonucleotide adsorption to the amino-modified surface. Twenty-five-nanoliter spots of a radiolabeled 12mer (a synthetic target complimentary to the k1 12mer probe) were deposited at concentrations from 0.01 to 40 M. Arrays were formed, one each per well, in an 8 ⫻ 12-well microtiter format. The excess of deposited oligonucleotide was then washed off with repeated water washes and bound surface density was determined with a storage phosphorimager. This test revealed a saturating probe density of 3 ⫻ 10 11 molecules/mm 2 (Fig. 3A). Assuming that one 12-mer oligonucleotide occupies a surface area of about 600 Å 2 (that is, 10 Å width and 5 Å rise per base repeat), we calculate that the maximum number of oligonucleotide molecules needed to form a closely packed monolayer on the surface would be 2 ⫻ 10 11 molecules/mm 2. Thus, comparison of experimental and calculated density gives evidence that a densely packed monolayer of oligonucleotides is forming during the adsorption process. We have repeated a similar adsorption analysis with a 36-mer probe. Maximum saturable probe density with the 36-mer was found to be 0.6 ⫻ 10 11 molecules/mm 2. Comparison to the density calculated for a closest packed 36-mer monolayer (0.6 ⫻ 10 11 molecules/ mm 2) suggests that a densely packed probe film can be formed for probes as long as 36 bases. We have found that after drying to achieve a tightly adsorbed, closely packed monolayer of oligonucleotide on the surface, subsequent treatment with high salt concentrations (3 M NaCl or 1 M sodium phosphate) or high pH (up to 11.5) near room temperature did not displace the attached radiolabeled probe monolayer. The loss of radioactivity was less than 10% in both instances. Noticeable removal (more than 50%) of radiolabeled probe from the adsorbed monolayer was found only during washes at elevated temperatures (95°C) with high salt (1 M sodium phosphate). For a positively charged surface covered with a monolayer of adsorbed probe, there are two independent processes that would cause a solution state target nucleic acid to bind to the surface: adsorptive electrostatic attraction to the cationic surface and duplex formation with the attached probes. If there were a high density of amines on the surface, not involved in direct interaction with probe, electrostatic attraction between target and the surface could become a dominating force, thereby overwhelming the contribution of duplex formation. Poor base sequence selectivity would be seen in that instance. Thus, we have set out to discover ways to “cap” the excess of unused surface cation, while retaining a tight adsorptive interaction between probe and the surface.
253
FIG. 3. (A) Titration of the amino-silanized surface with a 12-mer oligonucleotide. (Filled circles) oligonucleotides deposited and then washed. (Open circles) oligonucleotides deposited on the surface and quantified without washing. (B) Binding curves for 33P-labeled k-ras 1 synthetic 12-mer hybridized to k-ras probes that were attached to an amino-silanized surface via adsorption. Chemical capping (circles and triangles) or adsorptive capping with Denhardt’s solution (squares) was employed prior to hybridization. Hybridization buffer was (150 mM of sodium bicarbonate) at pH 9.5 (circles and squares) or pH 7.5 (triangles). A perfectly matched k-ras 1 target to k-ras 1 probe pairing is shown in filled symbols. A representative single base mismatch (k-ras 1 target to k-ras 3 probe) is shown in open symbols. T is the number of target molecules bound/mm 2 and P is the number of probe molecules/mm 2. The following binding function for homogeneous and noninteracting sites was used for fitting of the binding curves: Y ⫽ n*k[T]/(1 ⫹ k[T]), where Y ⫽ T/P is the bound target– probe ratio, [T] is free target concentration in solution, k is the affinity constant, and the value of Y is at saturation. The parameters k and n for the binding curves are shown in Table 1.
Covalent Capping In initial experiments, using noncovalent capping by prehybridization at pH 7.5 with 5⫻ Denhardt’s solution, amplicon interaction with residual positive charge gave rise to poor specificity and high background. Therefore, covalent capping reactions were investigated as an alternative. We used acetic anhydride
254
BELOSLUDTSEV ET AL.
Adsorptive Capping
TABLE 1
Binding Parameters for Synthetic 12-mer k-ras Target (See Fig. 3) Conditions k1 target:k1 probe, adsorptive capping, pH 9.5 k1 target:k1 probe, chemical capping, pH 8.0 k1 target:k1 probe, chemical capping, pH 9.5
n, (T/P) max
k/10 ⫺6 M dissociation constant
0.8
1.0
0.5
1.1
0.5
1.8
(at 0.5 M in DMF) for acetylation of amino-silanized surfaces after probe coupling. Capping reactions were carried out at room temperature for 1 h. Due to incomplete covalent capping, the above reactions yielded a slightly positive surface which still produced some nonspecific target binding at pH 7.5. Therefore, we decided to make neutral, slightly negatively, and highly negatively charged surfaces to determine the optimal covalent capping protocol yielding specificity of hybridization. A neutral surface was obtained by capping with acetic anhydride in the vapor phase at 50°C and 22 in. of Hg in a vacuum oven. Capping reactions were carried out 1 h and overnight. It was found that the overnight reaction significantly reduced electrostatic attraction and improved specificity. However, the 1-h reaction product retained a slight positive charge and gave poor specificity for amplicon targets. To make a slightly negatively charged surface, we used arrays that had been first capped with acetic anhydride, followed by a second capping with 0.5 M succinic anhydride in DMF at room temperature for 1 h. We found that such “double capped” arrays gave very good amplicon target binding specificity. A highly negatively charged surface was prepared by a one-step reaction with succinic anhydride, as described above. However, this surface was found to be too negatively charged to support measurable hybridization. Therefore, to achieve the best amplicon target binding specificity, we used the two-step capping procedure with acetic anhydride in the vapor phase at 22 in. Hg for 1 h at 50°C, followed by 0.5 M succinic anhydride in DMF at room temperature for 1 h. In the first acetylation step, an almost quantitative transformation of free amines into neutral amides was achieved (background or nonspecific binding of target amplicon to the surface was decreased almost to 0). By the second acylation step (with succinic anhydride) the remaining amines were transformed into their acylated derivatives. This introduction of carboxyl-groups imparted a slight negative charge and yielded surfaces capable of high amplicon target binding specificity.
To minimize the free energy of nonspecific target association with the surface, while retaining the capacity to adsorb oligonucleotide probe, we have tested the utility of a traditional blocking polymer mix: 1 mg/ml each of Ficoll, polyvinyl pyrrolidone, and bovine serum albumin (5⫻ Denhardt’s solution) as a coating to effect a reduction of electrostatic interaction between unused surface amines and solution state target. Additional fine-tuning of target interaction was obtained by an increase of solution state pH, reducing the surface charge as the pK of surface bound amine groups is approached. Thus, we have found that a prehybridization step at 150 mM sodium carbonate buffer at pH 9.5 and 5⫻ Denhardt’s solution, followed by hybridization in the same buffer produces a surface capable of high selectivity and high-affinity nucleic acid hybridization. We have compared these two methods of capping by performing analytical hybridization of a radiolabeled synthetic 12-mer k-ras gene target to a microarray containing its Watson–Crick complement (k1, Materials and Methods) and a related probe which differed by a single base change (k3). As seen in Fig. 3B (middle curves), the covalent method of capping has given rise to a binding isotherm with an apparent dissociation coefficient near to 1 ⫻ 10 ⫺8 M, reaching saturation around 0.5 targets bound per probe equivalent. These data are not significantly affected by pH in the range from 7.0 to 9.5. For comparison, at pH 9.5, and in the presence of Denhardt’s solution as the blocking agent, the measured dissociation constant is also seen to be about 1 ⫻ 10 ⫺8 M (Fig. 3B, upper curve). This result suggests that, as expected, the covalent capping process had not altered the nucleic acid probe in any way that could be detected via hybridization. At the highest applied k-ras target concentration, the limiting capacity to bind a synthetic 12-mer target appears to saturate at a slightly higher value for noncovalent blocking with Denhardt’s (0.8 targets bound per probe equivalent, compared to about 0.5), but considering the standard error of the data, this apparent difference is not significant. That secondary consideration aside, the data suggest that the two methods of excess surface charge neutralization are seen to be nearly equivalent in terms of target binding affinity and binding capacity per unit area. For both methods of capping, the capacity to form duplex with a short synthetic target is characterized by single nucleotide mismatch discrimination (Fig. 3B, lower curves). We have used the above adsorptive probe attachment chemistry (with covalent capping) to perform single base polymorphism analysis on the 152-bp PCR product and have compared the results thus obtained with arrays prepared using covalent attachment of
DNA MICROARRAY PRODUCTION USING OLIGONUCLEOTIDE ATTACHMENT
FIG. 4. (A) Single nucleotide polymorphism analysis with electrostatically attached probes. (B) Single nucleotide polymorphism analysis with covalently attached probes.
terminal amine-modified probes to an epoxysilanecoated surface (Fig. 2). Results obtained using amplicon targets derived from the k-ras gene show that specificity achieved using adsorptive probe binding followed by chemical capping (Fig. 4A) is very much comparable in affinity and specificity to the best that could be obtained with covalently attached probes on a neutral epoxysilane surface (Fig. 4B). As seen in both instances, the relatively stable G–T mismatch (K-ras 1 target amplicon with k3 probe) generates signals that are diminished by a factor of 10 relative to perfectly matched 12-bp pairing (K-ras 1 target with k1 probe). Discrimination for all other single base mismatches is measured to be significantly greater than a factor of 25. CONCLUSIONS
There are several practical advantages to the current method of microarray formation by adsorptive probe attachment. First, the affinity and selectivity of duplex formation are at least as good as the best that
255
can be obtained by traditional methods of probe attachment. Second, the adsorptive method requires the use of inexpensive unmodified probes, at an applied concentration that is at least five times lower than that required for the currently available covalent attachment chemistries. As a result, the probe cost of manufacture is reduced at least 10-fold. Finally, because the probe is printed in distilled water (which denatures nucleic acid secondary structure) and because binding is adsorptive, which tends to maximize surface contacts, probes affixed to a surface in this way may tend to show smaller artifacts due to intramolecular folding interactions. The mechanism of a surface-directed nucleic acid condensation has been suggested in previous literature (12, 13), but little was known about the structure of the resulting nucleic acid layer or its capacity to bind target nucleic acid. Here, it has been shown that subsequent to an appropriate capping reaction, target binding to such probes displays a high level of sequence selectivity and high overall binding affinity, thereby confirming that a base-paired duplex is being formed. Yet, the high stability of the probe interaction with the surface suggests that probes bound to the surface in this way must be constrained by multiple contacts. Thus it can be asked how a tight (nearly irreversibly) adsorbed probe layer might be capable of tight, sequence-selective duplex formation with a high-molecular-weight nucleic acid target, especially in light of previous literature (8), which had suggested that such adsorbed probes would be unavailable to bind a cognate target due to such limitations. As a partial resolution of that dilemma, we suggest that duplex can still form with such probes because, on a weakly cationic aminosilane surface, probes are not bound rigidly to the underlying surface, but are instead “trapped” upon the planar surface in a potential well, unable to diffuse off it, but relatively free to engage in rotation and other reorienting motions upon a surface that are necessary to form a double helix with a cognate DNA or RNA target. As such, the two-dimensional probe layer has properties that approximate those of liquid crystal, with the capacity for nearly each probe to form duplex within the liquid crystal layer. Alternatively, it may be formally possible that structures approximating a ribbon-like, nonhelical duplex may be forming at the surface. Such structures have been reported in different contexts, such as hybridization at air–water interfaces (14). Work is in progress to resolve these potential configurations. ACKNOWLEDGMENT This work was funded in part by a Program Project grant from the National Cancer Institute.
256
BELOSLUDTSEV ET AL.
REFERENCES 1. Collins, F. S. (1999) Microarrays and macroconsequences. Nat. Genet. Suppl. 21, 2. 2. Lander, E. S. (1999) Array of hope. Nat. Genet. Suppl. 21, 3– 4. 3. Ramsay, G. (1998) DNA chips: State-of-the art. Nat. Biotechnol. 16, 40 – 44. 4. Ausubel, F. M., et al. (1997) Current Protocols in Molecular Biology Units 2.9 and 3.10. Wiley, New York. 5. Lee, G. U., Chrisey, L. A., and Colton, R. J. (1994) Direct measurement of the forces between complimentary strands of DNA. Science 266, 771–773. 6. Guo, Z., Guilfoyle, R. A., Thiel, A. J., Wang, R., and Smith, L. M. (1994) Direct fluorescence analysis of genetic polymorphisms by hybridization with oligonucleotide arrays on glass supports. Nucleic Acids Res. 22, 5456 –5465. 7. Eggers, M., et al. (1994) A microchip for quantitative detection of molecules utilizing luminescent and radioisotope reporter groups. BioTechniques 17, 516 –524. 8. Duggan, D. J., Bittner, M., Chen, Y., Meltzer, P., and Trent, J. M.
9.
10.
11.
12.
13.
14.
(1999) Expression profiling using cDNA microarrays. Nat. Genet. Suppl. 21, 10 –14. Chrisey, L. A., Lee, G. U., and O’Ferrall, C. E. (1996) Covalent attachment of synthetic DNA to self-assembled monolayer films. Nucleic Acids Res. 24, 3031–3039. Maldonado-Rodriguez, R., Espinosa-Lara, M., Calixto-Suarez, A., Beattie, W. G., and Beattie, K. L. (1999) Hybridization of glasstethered oligonucleotide probes to target strands preannealed with labeled auxiliary oligonucleotides. Mol. Biotechnol. 11, 1–12. Haugland, R. P. (1996) Handbook of Fluorescent Probes and Research Chemicals, 6th ed., pp. 117–120. Molecular Probes, Inc., Eugene, OR. Fang, Y., and Hoh, J. H. (1998) Surface-directed DNA condensation in the absence of soluble multivalent cations. Nucleic Acids Res. 26, 588 –593. Balladur, V., Theretz, A., and Mandrand, B. (1997) Determination of the main forces driving DNA oligonucleotide adsorption onto aminated silica wafers. J. Colloid Interface Sci. 194, 408 – 418. Ebara, Y., Mizutani, K., and Okahata, Y. (2000) DNA hybridization at the air–water interface. Langmuir 16, 2416 –2418.