Colloidal Silica Films for High-Capacity DNA Probe Arrays

Chem. Mater. 2001, 13, 4773-4782 4773 Colloidal Silica Films for High-Capacity DNA Probe Arrays M. Glazer,*,†,‡ J. Fidanza,§ G. McGall,§ and C. Frank‡ Departments of Chemical Engineering and Materials Science and Engineering, Stanford University, Stanford, California 94305, and Affymetrix, Santa Clara, California 95051 Received June 25, 2001 The use of arrays of immobilized DNA “probes” for high-throughput analysis of genomic samples is expanding rapidly. The detection sensitivity of these arrays depends on the quantity and density of immobilized probe molecules as well as on the thermodynamics and kinetics of nucleic acid hybridization. We have prepared and investigated substrates with a porous, “three-dimensional” surface layer as a means of increasing the surface area available for the synthesis or immobilization of oligonucleotide probes, thereby increasing the number of available probes and the amount of detectable bound target per unit area. Surfaces with pores 5 nm and larger were created by spin-coating colloidal suspensions of silica particles, followed by thermal curing. DNA arrays were synthesized on the resulting surfaces by photolithographic patterning, and the performance on the high-capacity substrates was compared to that on standard flat glass surfaces. The colloidal silica films created via this route show equivalent performance to flat glass substrates in terms of the efficiency of chemical synthesis and resolution of photolithographic patterning. DNA targets are able to penetrate the porous layers, and under saturating conditions, the quantity of bound target is proportional to the layer thickness. The result is an enhanced hybridization signal that is 20 times higher than flat glass for a colloidal particle layer that is 0.5 µm thick. The thermodynamic stability of probe/target duplexes in the matrix is the same as that for their counterparts on flat surfaces, although the colloidal silica films reach saturation more slowly than flat surfaces. Introduction sample using standard molecular biology protocols.5 The analyte sequences are fragmented and labeled with Recent initiatives to obtain complete genomic se- reporter molecules (such as fluorescent labels) for detec- quences for humans and other organisms have increased tion, and the mixture of labeled fragments is applied to the demand for techniques that enable the rapid, large- the array under controlled conditions, allowing binding scale analysis of DNA and RNA. Hybridization-based to surface probes that have complementary sequences sequence analysis using high-density polynucleotide (hybridization). The array is then imaged, typically with probe arrays is capable of efficiently accessing large a laser confocal scanning microscopy system, to locate amounts of genetic information from biological samples and quantify the binding of target sequences from the in a single assay procedure.1,2 These arrays, or “DNA sample to complementary sequences on the array. chips,” are essentially large sets of nucleic acid “probe” Application-specific software reconstructs and presents sequences that are immobilized in defined locations on the sequence data for further querying. the surface of a flat substrate, typically glass. Light- The ability to detect the binding of target molecules directed combinatorial synthesis has enabled the fab- from a nucleic acid sample to complementary probes on rication of oligonucleotide probe arrays with very high an array depends on a large number of factors, including resolution and information content on a commercial the quantity and density of immobilized probe molecules scale.3,4 In a typical application, DNA or RNA “target” on the substrate.6 A variety of methods have been used sequences of interest are isolated from a biological to increase the number of probe molecules available for binding and thereby increase the sensitivity of DNA * To whom correspondence should be addressed. E-mail: mglazer@ leland.stanford.edu. arrays. Much of the focus has been on organic surface † Department of Materials Science and Engineering, Stanford modification, including polymer gel layers,7-9 organic University. ‡ Department of Chemical Engineering, Stanford University. § Affymetrix. (4) Pease, A. C.; Solas, D.; Sullivan, E.; Cronin, M. T.; Holmes, C. (1) Lockhart, D. J.; Dong, H.; Byrne, M. C.; Follettie, M. T.; Gallo, P.; Fodor, S. P. A. PNAS 1994, 91, 5022-5026. M. V.; Chee, M. S.; Mittman,M.; Wang, C.; Kobayashi, M.; Horton, H.; (5) Schena, M.; Davis, R. W. In Microarray Biochip Technology; Brown, E. L. Nat. Biotechnol. 1996, 14, 1675-1680. Schena, M., Ed.; Eaton Publishing: Natwick, MA, 2000; p 1. (2) Lipshutz, R. J.; Morris, D.; Chee, M.; Hubbell, E.; Kozal, M. J.; (6) Steel, A.; Torres, M.; Hartwell, J.; Yu, Y.; Ting, N.; Hoke, G.; Shah, N.; Shen, N.; Yang, R.; Fodor, S. P. A. BioTechniques 1995, 19, Yang, H. In Microarray Biochip Technology; Schena, M., Ed.; Eaton 442-447. Publishing: Natwick, MA, 2000; pp 87-117. (3) Fodor, S. P. A.; Read, J. L.; Pirrung, M. C.; Stryer, L.; Lu, A. T.; (7) Heller, M. J.; Forester, A. H.; Tu, E. Electrophoresis 2000, 21, Solad, D. Science 1991, 251, 767-773. 157-164. 10.1021/cm010578n CCC: $20.00 © 2001 American Chemical Society Published on Web 11/30/2001 4774 Chem. Mater., Vol. 13, No. 12, 2001 Glazer et al. membranes,10,11 and dendrimeric linking molecules.12 characteristics. For example, dip and spin-coating tech- However, these approaches present certain drawbacks. niques for the deposition of colloidal films have been well For instance, the kinetics of diffusion of the target established.19 Furthermore, the use of rigid colloidal nucleic acid molecules in a polymer gel will be slower silica particles rather than molecular precursors (such than that in solution due to the higher viscosity of the as alkoxides) gives mechanical strength to the porous gel and also due to nonspecific interactions with the layers with minimal sintering. Additionally, the use of immobilized nucleic acid probes and potentially with the rigid precursors makes the deposition process more matrix itself.13 DNA dendrimers have also been utilized robust and less sensitive to the environment (humidity, to increase the number of sites available for binding temperature) and stoichiometry parameters that are target molecules, but the sensitivity gains with this critical conditions for control of the hydrolysis and approach are lower than expected, suggesting that many condensation reactions that occur during the spin- of the binding sites may be inaccessible due to steric coating of alkoxides.19 Finally, the size of the precursors crowding.14 can be chosen to create the appropriate pore size to Porous inorganic substrates are receiving increasing accommodate different target molecules, with the poros- attention as potential supports for DNA arrays. Inor- ity being independent of the particle size (for relatively ganic materials are inert to most processes used for monodisperse particles). To further control and optimize array fabrication and testing, which is one of the reasons the morphology of these types of films, latex particles for the widespread use of flat glass substrates. Previ- can be incorporated during the film deposition process ously, oligonucleotides or DNA fragments have been and then removed at high temperatures, leaving behind immobilized on alumina filters,10 porous silicon,15 and voids of a controlled size.18 Pore sizes less than 30 nm “flow-through” chips consisting of glass capillaries.6 can be readily obtained, resulting in films with ex- However, due to optical scattering that would result tremely high surface area. from the large features and opaqueness (in the case of In this work, we present detailed characterization of silicon), these materials are not well-suited to high- colloidal silica films relevant to their use as substrates resolution photolithographic array synthesis. Further- for high-capacity DNA arrays and we discuss colloidal more, since confocal microscopy is extremely sensitive silica film deposition, chemical synthesis of arrays, to scattered light,16 scattering from large features in the optical imaging, and the hybridization of oligonucleotide film adversely affects the resolution of the imaging target molecules. The use of oligonucleotide targets in system. this study eliminates concerns about nonspecific binding Photolithographic patterning enables the synthesis of of background cell material present in biological samples DNA arrays with very small feature sizes, and thus so that fundamental differences in thermodynamic and extremely high information content can be obtained. kinetic behavior of probe/target binding in the matrix This process involves the selective removal of terminal- can be observed. protecting groups from growing oligonucleotide chains This study has been conducted simultaneously with in predefined regions of a glass support by controlled the evaluation of colloidal silica substrates in complex, exposure to light through photolithographic masks.3,4 high-density biological assays such as those used for Oligonucleotide building blocks with photolabile 5′- disease management and Human Gene Expression protecting groups are added to the growing probes in a profiling. Using colloidal silica films, we have demon- step-by-step process. For photolithography, the pore size strated an 8 to 10-fold improvement in signals and a of the features must be substantially smaller than the higher percentage of correct calls of the target sequence wavelength of light to avoid optical scattering. We have in an assay for the protease and reverse transcriptase recently reported preliminary results using colloidal genes of the HIV virus.20 The performance of Human silica films as substrates for photolithographically pat- Gene Expression profiling arrays using these substrates terned arrays.17,18 Deposition of colloidal silica as a route will be discussed in a separate report.21 for creating porous thin films has several favorable Experimental Procedures Film Deposition. Testing was performed on soda-lime (8) Proudnikov, D.; Timofeev, E.; Mirzabekov, A. Anal. Biochem. 1998, 259, 34. (Erie Scientific) and fused silica (U.S. Precision Glass) glass (9) Yershov, G.; Barsky, V.; Belgovskiy, A.; Kirillov, E.; Dreindlin, substrates. Test pieces were cut to either 2 × 3 × 0.027 in. or E.; Ivanov, I.; Parinov, S.; Guschin, D.; Drobishev, A.; Dubiley, S.; 5 × 5 × 0.027 in. All substrates were cleaned via a three-step Mirzabekov, A. PNAS 1996, 93, 4913-4918. process of soaking in Nanostrip (sulfuric acid and hydrogen (10) Englert, D. In Microarray Biochip Technology; Schena, M., Ed.; peroxide, Cyantek, 15 min), sodium hydroxide solution (10% Eaton Publishing: Natwick, MA, 2000; pp 220-240. (11) Matysiak, S.; Hauser, N. C.; Wurtz, S.; Hoheisel, J. D. by wt in water, 70 °C, 3 min), and hydrochloric acid (0.4% by Nucleosides Nucleotides 1999, 18, 1289. wt in water, 1 min). (12) Beier, M.; Hoheisel, J. D. Nucleic Acids Res. 1999, 27, 1970. Film deposition was accomplished via spin-coating of col- (13) Livshits, M. A.; Mirzabekov, A. Biophys. J. 1996, 71, 2795. loidal silica suspensions. The silica solutions used were Snow- (14) Wang, J.; Miang, M.; Nilsen, T. W.; Getts, R. C. J. Am. Chem. tex S50, 20L, and ZL (Nissan Chemicals), which were analyzed Soc. 1998, 120, 8281-8282. prior to deposition by dynamic light scattering (DLS) with laser (15) Lin, V. S.-Y.; Motesharei, K.; Dancil, K.-P. S.; Sailor, M. J.; Ghadiri, M. R. Science 1997, 278, 840. wavelength of 514.5 nm (Brookhaven Instruments) to observe (16) Schermer, M. J. In DNA Microarrays; Schena, M., Ed.; Oxford the actual size distribution of the particles. The measured sizes Press: New York, 1999; pp 17-42. of the S50, 20L, and ZL particles were 16 ( 5, 54 ( 13, and (17) Glazer, M.; Frank, C.; Vinci, R. P.; McGall, G.; Fidanza, J.; 65 ( 16 nm, respectively. Beecher, J. In Organic/Inorganic Hybrid Materials II; Klein, L, C., Francais, L. F., DeGuire, M. R., Mark, J. E., Eds.; Materials Research Society Proceedings: PA, 1999; pp 371-376. (19) Brinker, C. J.; Scherer, G. W. Sol-Gel Science; Academic (18) Glazer, M.; Frank, C.; Lussi, J.; Fidanza, J.; McGall, G. In Press: San Diego, 1990; pp 787-797. Organic/Inorganic Hybrid Materialss2000; Laine, R. M., Sanchez, C., (20) Fidanza, J.; Glazer, M.; Mutnick, D.; McGall, G.; Frank, C. Brinker, C. J., Giannelis, E., Eds.; Materials Research Society Proceed- Nucleosides Nucleotides 2001, 20, 533-538. ings, 2001; pp CC10.4.1-CC10.4.6 (online publication). (21) Fidanza, J.; Glazer, M.; Frank, C.; McGall, G. In preparation. High-Capacity DNA Probe Arrays Chem. Mater., Vol. 13, No. 12, 2001 4775 The suspensions were diluted to the desired weight concen- tration with deionized water, dispensed onto the glass sub- strates using a syringe with a 0.45 µm filter, and then spun at 2500 rpm for 30 s. Drying typically occurred in 15-20 s, as could be observed by the color change of silicon substrates spun under the same conditions. The films were then heated at a rate of 10 °C min-1 and held at 350 °C for 4 h. This tem- perature was sufficient to provide stable films (see Results) while avoiding excessive sintering that could change the film morphology and cause shrinkage if elevated temperatures were used. Nitrogen Adsorption. To obtain samples of sufficient size for the measurement of surface area, pore size, and pore vol- ume, colloidal silica suspensions were dried in an oven at 80 °C for 12 h and then at 100 °C for 6 h, and samples of the dried materials were analyzed with a Beckman-Coulter SA-3100. Samples were outgassed at 300 °C for 60 min, and nitrogen was used as the adsorbate gas. The Brunauer, Emmett, and Teller (BET) method was used for calculating surface area, and the Barret, Joyner, and Halenda (BJH) method was used for calculating pore volume and size distri- bution.22 Film Characterization. Ellipsometric measurements were conducted on films deposited on silicon, which provides a higher contrast in the index of refraction than glass, with the native oxide surface layer on the silicon providing a substrate similar to the glass surface. Measurements were made with a Figure 1. Procedures for quantification of (a) surface hydroxyl Gaertner ellipsometer at a wavelength of 632.8 nm. Film site density (pmol cm-2) and (b) stepwise synthesis efficiency thickness measurements were made with a Dektak 3 surface (RSY). The cleaved products are quantified by HPLC; CE ) profilometer (Veeco). 2-cyanoethyl and Piv ) pivaloyl. Electron microscopy measurements were made on a Hitachi S4700 field emission scanning electron microscope. A thin midite coupling protocols.25 The amount of bound fluorescein layer of gold was sputtered onto the surface before imaging was analyzed by directly imaging the fluorescence with a con- the glass. focal system or by cleaving the fluorescein molecules from the Surface area of the deposited films is calculated by combin- surface and then quantifying the total amount of fluorescein ing results from ellipsometry and nitrogen adsorption. From molecules released using high-performance liquid chromatog- measurements of the index of refraction, the mass of the solid raphy (HPLC). phase (Ms) in the layer is calculated by eq 1, which is based Surface Fluorescence Analysis. Quantitative studies of on the Lorentz-Lorenz equation.23,24 the density and uniformity of chemical coupling on the substrates were conducted using “surface-staining” methods25 ( )( ) Nf2 - 1 Nsi2 + 2 in which a fluorescein phosphoramidite derivative is coupled Ms ) tFsi (in g cm-2 of lateral surface) to the free hydroxyl groups on the silane. Prior to staining, Nf2 + 2 Nsi2 - 1 the fluorescein phosphoramidite monomer was diluted with (1) the appropriate amount of a diluent monomer (T β-cyanoethyl phosphoramidite (DMT-T-CEP), Amersham Pharmacia) in Nf and Nsi are the indices of refraction of the film and pure acetonitrile to a total concentration of 50 mM. To deprotect SiO2 (1.46), respectively, t is the film thickness, and Fsi is the bound fluorescein molecules, the substrates were treated the density of pure SiO2 (2.2 g cm-3). The area factor (AF), in a 1:1 solution of ethylenediamine/ethanol for 1 h, rinsed which quantifies the multiple of surface area relative to a flat with deionized water, and dried with nitrogen. The substrates glass substrate for a given lateral area, is then calculated with were then scanned with a confocal laser scanning microscope,25 eq 2. with the signal obtained being a function of the number of available sites on the surface. The signal was calibrated with AF ) MsAN + 1 (2) solutions of fluorescein sodium salt (Sigma) in buffer to correlate values obtained with the actual surface density of fluorophores. AN is the specific surface area (m2 g-1) of the dried colloidal HPLC Analysis. A HPLC assay26-28 was used to measure material from nitrogen adsorption measurements, and the the hydroxyl site density available for synthesizing oligonucleo- additional term of unity is included to account for the original tides, the synthesis efficiency on the support, and the extent flat surface on which the layer was deposited. of adsorption or entrapment of the reagents within the porous Chemical Coupling Efficiency. Prior to synthesis of oligo- matrix (see Figure 1). For site density measurement, a nucleotide probes on the substrates, the surface was silanated cleavable linker (5′-phosphate-ON reagent, ChemGenes Corp.) with bis(2-hydroxyethyl)-3-aminopropyltriethoxysilane.17,18 Two was attached to the surface with phosphoramidite chemistry, complementary techniques were then used to study the cou- followed by a spacer molecule (C3 spacer phosphoramidite, pling of phosphoramidites to the resulting surface hydroxyl Glen Research) and a fluorescent tag prepared separately (5′- groups on the silane, which serve as initiation sites for oligo- carboxyfluorescein phosphoramidite27). The surface was then nucleotide synthesis. The hydroxyl groups were labeled with a fluorescein phosphoramidite using standard phosphora- (25) McGall, G. H.; Barone, A. D.; Diggelmann, M.; Fodor, S. P. A.; Gentalen, E.; Ngo, N. J. Am. Chem. Soc. 1997, 119, 5081-5090. (22) Allen, T. Particle Size Measurement, Surface Area and Pore (26) U.S. Patent 5,843,655. Size Determination, 5th ed.; Chapman & Hall: London, 1997; Chap- (27) Eur. Pat. Appl. EP 967217, 1999. ter 2. (28) McGall, G. H.; Barone, A. D.; Beecher, J. E.; Diggelman, M.; (23) Pettit, R.; Ashley, C.; Reed, S.; Brinker, J. In Sol-Gel Technol- Fodor, S. P. A.; Goldberg, M. J.; Ngo, N.; Rava, R. P. In Innovation ogy for Thin Films, Fibers, Preforms, Electronics, and Specialty Shapes; and perspectives in solid phase synthesis and combinatorial libraries, Klein, L., Ed.; Noyes Publications: NJ, 1988; p 85. 1998: peptides, proteins and nucleic acids: small molecule organic (24) Born, M.; Wolf, E. Principles of Optics, 6th ed.; Pergamon chemical diversity: collected papers, fifth international symposium, 5th Press: New York, 1980; pp 86-88. ed.; Epton, R., Ed.; Mayflower: London, 1999; pp 97-100. 4776 Chem. Mater., Vol. 13, No. 12, 2001 Glazer et al. Figure 2. Scanning electron microscope (SEM) image of colloidal silica particles deposited on glass. diced into 1 cm2 pieces, weighed, placed in a glass vial, and Table 1. Parameters of Dried Colloidal Silica Measured treated with 1:1 (by vol) ethylenediamine/water for 4 h at 50 by Nitrogen Adsorption °C to cleave the linker and release 3′pC3-fluorescein5′. The mean BET mean pore pore resulting solution was diluted, co-injected with an internal particle surface area diameter volume standard, and analyzed by HPLC. The internal standard, sample size (nm) (m2 g-1) (nm) (mL g-1) 3′pC C -fluorescein5′, was prepared separately on an ABI 3 3 synthesizer and quantified independently by UV-Vis spectra S50 16 ( 5 104 5.3 ( 0.2 0.19 on a Varian Cary 3E spectrophotometer (Varian). 20L 54 ( 13 49 14.2 ( 0.7 0.22 HPLC analyses were performed on a Beckman System Gold ZL 65 ( 16 35 22.5 ( 2.2 0.21 employing an ion-exchange column and fluorescence detection at 520 nm. Elution was performed with a linear gradient of mismatches at base 10 (A substituted for T) and two-base 0.4 M NaClO4 in 20 mM Tris pH 8 (or other suitable buffer mismatches with substitutions at both base 10 and base 12 system), at a flow rate of 1 mL min-1. If fluorescent molecules (A substituted for G).29 Discrimination was evaluated by have remained adsorbed or trapped on the surface non- comparing the relative signal intensity (after background covalently, they will appear first in the chromatogram, fol- correction) in the mismatch regions with the signal in the lowed by fluorescein that has coupled to the C3 spacers and perfect-match region. finally the 3′pC3C3-fluorescein5′ internal standard. Integration of HPLC peak areas was used to quantify the total cleaved Results fluorescein and thereby the total site density. The surface site density per unit area was determined by dividing the total Suspension and Film Characterization. Figure 2 sites by the lateral surface area available for synthesis shows SEM images of the S50, 20L, and ZL silica (calculated from the weight of the glass sample). particles deposited on a glass surface (after 350 °C The efficiency of oligonucleotide synthesis was determined treatment). The particles pack randomly with no notice- by examining the yield of a short homopolymer, such as able short- or long-range order. The broad distribution hexathymidylate. The cleavable linker was coupled to the surface, followed by a fluorescein phosphoramidite and six observed by light scattering is also reflected in the nucleoside phosphoramidites (see Figure 1). The homopolymer distribution of particle sizes in the deposited layer. The was then cleaved from the support, mixed with an internal implications of the broad particle size distribution are standard, and analyzed by HPLC. The relative synthesis yield considered in the Discussion. The index of refraction of (RSY) on the surface was calculated by dividing the integrated the colloidal silica films was typically 1.3 before the area of the 6-mer peak by the total area of all products cleaved 350 °C thermal treatment and did not increase following from the surface. The RSY is indicative of the efficiency of the step-by-step base-coupling reactions on the solid support. the treatment, indicating that the curing process did Hybridization to Oligonucleotide Targets. 20-mer probes not cause significant densification of the films. There- were synthesized in a checkerboard pattern on Affymetrix fore, the properties of the dried material should be com- photolithographic synthesizers using phosphoramidite chem- parable to the structure following thermal treatment. istry, followed by removal of protecting groups in 1:1 (by vol) The pore size, surface area, and pore volume for bulk- ethylenediamine/ethanol solutions for a minimum of 4 h.25 A dried colloidal silica are summarized in Table 1. Pores typical probe sequence used for analysis was (from surface) in this system are the void spaces between the solid 3′-GACTTGCCATCGTAGAACTG-5′. The arrays were hybrid- ized to solutions of 100 nM 5′-fluorescein-labeled oligonucle- particles. The smaller particle sizes have correspond- otide target in 0.1 M MES (2-[N-morpholino]ethanesulfonic ingly smaller pores and higher surface area, whereas acid (Sigma)) buffer (0.89 M NaCl, 0.03 M NaOH). A target the total pore volume is nearly constant. The pores are concentration of 100 nM was used, as this concentration has considerably smaller than the pores in standard con- previously been shown by Forman et al.29 to saturate the trolled pore glass (CPG) supports for DNA synthesis, surface probes in a similar system, resulting in the highest which are typically 50-100 nm.30 adsorbed target density. The films were scanned at several time points in 25 °C buffer (36 h) and then transferred to buffer Site Density and Coupling Yield. The surface-site- maintained at 45 °C (50 h) with an environmental incubator density procedure was used to evaluate the efficiency shaker (New Brunswick Scientific). Prior to scanning, the of phosphoramidite chemistry on the porous films. samples were extensively washed with fresh buffer to remove Analysis of the HPLC chromatograms for the cleaved nonspecifically bound target. products showed peaks corresponding only to the ex- To evaluate discrimination, probes were synthesized with pected 3′pC3-fluorescein5′ and internal standard. The C3 the perfect-match sequences, as well as with single-base spacer serves as a control for incomplete coupling or removal of reagents from the porous matrix. If the (29) Forman, J. E.; Walton, I. D.; Stern, D.; Rava, R. P.; Trulson, M. O. In Molecular Modeling of Nucleic Acids; Leonitis, N. B., SantaLucia, J., Jr., Eds.; ACS Symposium Series 682; American (30) Adams, S. P.; Kavka, K. S.; Wykes, E. J.; Holder, S. B.; Gallupi, Chemical Society: Washington, DC, 1998; pp 206-228. G. R. J. Am. Chem. Soc. 1983, 105, 661. High-Capacity DNA Probe Arrays Chem. Mater., Vol. 13, No. 12, 2001 4777 Figure 4. Comparison of the density of accessible hydroxyl sites to the surface area of ZL colloidal silica films of varying Figure 3. Comparison of the density of accessible surface thickness. The measured site density increases in direct hydroxyl sites to the surface area of the porous films. Reagents proportion to the surface area as the film thickness increases. are able to fully access the 20L and ZL surfaces, but complete coupling is not achieved on the S50 surface due to poor fluid flow in the small pores. Table 2. Synthesis Yield on Colloidal Silica Filmsa particle relative synthesis yield type (flat glass ) 1) S50 0.46 ( 0.44 20L 1.07 ( 0.09 ZL 1.11 ( 0.10 a Normalized to flat glass [) RSYfilm/RSYflat]. fluorescein reagents were not completely washed out of the matrix during coupling, a separate peak, corre- sponding to the fluorescein phosphate monomer, would elute prior to that for the 3′pC3-fluorescein5′. The fact that all of the fluorescein that eluted from the surface was covalently attached to C3 spacers indicated that coupling was efficient in the porous film. Figure 5. Surface fluorescence on a flat glass surface and a Figure 3 shows the quantification of the hydroxyl site 0.5 µm thick ZL silica film that have been silanated and the density experiments for each of the particle types in hydroxyl sites labeled with fluorescein (see Experimental comparison to the actual surface area of the film. Layers Section). Fluorescence intensity is in linear proportion to were deposited from 20 wt % suspensions, resulting in fluorescein/diluent ratio up to a ratio of approximately 0.05; 0.3 µm films for all three particle types. Measured site it reaches a maximum at 0.2 and decreases further at higher density on the flat glass was 120 ( 10 pmol cm-2. For concentration. the 20L and ZL films, the increase in site density agrees with that predicted by the increase in surface area On the basis of these results, the ZL particles were within experimental uncertainty. For the S50 films, the chosen for further experiments. In addition to facilitat- site density measured by the HPLC procedure is lower ing efficient coupling, the pores in the ZL films, which than the expected value based on surface area, which are the largest of the three particle types tested, should is presumably due to incomplete access of synthesis aid the penetration of large target molecules and reagents to surface sites within the film. fluorescent protein conjugates such as streptavidin- The site density results are further supported by phycoerythrin (SAPE). SAPE is a commonly used label- Table 2, which shows the relative hexathymidylate ing agent in GeneChip assays1 and has dimensions as synthesis yield for each of the particle substrates large as 11 nm.31 Figure 4 shows that the site density normalized to the yield of the flat glass control () on the ZL films is proportional to the surface area (and RSYfilm/RSYflat). For the 20L and ZL substrates, the the film thickness). For the 0.5 µm film, an increase of yield of full-length 6-mer probes was equal to that of 25-fold over flat glass is obtained. Accessibility of surface the flat control samples within experimental error, synthesis sites is independent of film thickness over this indicating that coupling was equally efficient on these range, indicating efficient penetration of synthesis films as on flat glass. On the S50 substrate, synthesis reagents throughout the layer. yield was lower and there was a high degree of vari- Surface Fluorescence. To examine the sensitivity ability, which is consistent with the data from the site of the system to fluorescence quenching, flat and ZL density measurement and indicates that there is poor surfaces were stained with fluorescein in a range of fluid flow through the small pores in the S50 film during concentrations. Figure 5 shows that the signal on both synthesis. Optimization of synthesis (e.g., using higher surfaces is linear with fluorescein/diluent ratio up to reagent concentrations and/or longer coupling and 0.05, which corresponds to 1 pmol cm-2 of fluorescein. washing times) may improve the efficiency of chemical coupling on the S50 substrates but was not pursued (31) Glazer, A. N.; Stryer, L. Methods Enzymol. 1990, 184, 188- further. 194. 4778 Chem. Mater., Vol. 13, No. 12, 2001 Glazer et al. Figure 6. Ratio of binding of oligonucleotide targets (to perfect-match surface-bound probe sequences) on ZL colloidal silica films to fused silica glass. Samples were incubated in 100 nM oligonucleotide target at 25 °C and then at 45 °C. All time points have reached equilibrium within 12 h at 25 °C, except for the 0.4 and 0.5 µm surfaces. Figure 7. Discrimination ratio on porous glass films after 24 Quenching effects become noticeable at a ratio of 0.1, h at 45 °C under saturating conditions (100 nM target). Under the signal reaches a maximum at 0.2, and the signal these conditions, the discrimination on the porous glass films is equivalent to the flat glass control at each thickness and decreases on both films at higher concentrations. On the for both one- and two-base mismatches. basis of results from Forman et al.29 with a similar system, the maximum density of bound fluorophore for ratio of the hybridization signal on the films to the 20-mer probes at 25 °C should be approximately 1 pmol signal on the flat glass does not change with additional cm-2, which is in the linear range for both surfaces. hybridization time. To ensure that the target was not Optical Scattering. A single front-side “print” with depleted, fresh solution was replenished after each scan. the porous layer in contact with the photolithographic Hybridization reactions for gene expression assays are mask was used to determine if the multiple silica/air commonly performed at 45 °C for 16-24 h1, which interfaces in the layer cause significant scattering of corresponds to the first 45 °C scan in Figure 6. The final light during photolithographic patterning or fluorescent time point at 45 °C was taken to ensure that the films scanning. This is a close mimic of the actual procedures had reached equilibrium. The hybridization signal used for manufacturing GeneChip arrays. Colloidal multiple increases in proportion to the film thickness silica layers 0.5, 1, and 2 µm thick were deposited by and reaches ∼80% of the multiple of the ratio of the successive spin-coating depositions and stained with hydroxyl site density to flat glass (from Figure 4). fluorescein dye.17 For all the coatings tested, the transi- Background and Discrimination. High specificity tion “edge” between the background (nonphotolyzed) of discrimination for target molecules as well as low and stained (photolyzed) regions was as sharp as flat background signals are critical aspects of the functional glass. The 20-30 nm pore size is sufficiently smaller performance of DNA arrays. Background due to non- than the wavelength of light such that optical scattering specific adsorption on the silica films was equivalent to does not cause loss of edge resolution during photolitho- flat glass within the uncertainty of the measurement. graphic synthesis or optical scanning. Background due to the substrate itself increases slightly Hybridization to Oligonucleotide Targets. Prior with film thickness, which may be due to fluorescent to hybridization experiments, the confocal microscope impurities in the film. The net result is that at 45 °C, system was calibrated with solutions of oligonucleotide the backgrounds on the 0.2 and 0.5 µm films are 1.4 target to correlate the signal to the density of bound and 2.1 times the background of the flat glass, respec- target on the flat substrates.29 Figure 6 shows the tively. By contrast, bound target reaches multiples of 9 hybridization signal ratio for the perfect-match probes and 19 times the flat glass for the same films, as on the 0.2, 0.4, and 0.5 µm ZL films relative to flat glass illustrated in Figure 6. at successive time points. Forman et al. have shown that To test discrimination, probes were synthesized with oligonucleotide probes on flat substrates reach equilib- single-base mismatches at base 10 and two-base mis- rium with a 10 nM target solution in approximately 1 matches at both base 10 and base 12. Figure 7 shows h,29 and therefore the 100 nM target used in this study that discrimination for oligonucleotide targets is equiva- should cause saturation in equal or less time. The bound lent on the flat glass and porous films after 24 h at 45 target density on flat glass reaches a maximum of 0.8 °C, which is the temperature typically used to destabi- ( 0.2 pmol cm-2 at 25 °C for the perfect-match probes, lize nonspecifically bound targets in gene expression which agrees well with the measured density for similar assays. Therefore, the ratio of the hybridization signal arrays by Forman et al. and is in the linear range of on the silica films to flat glass is independent of the fluorescence for both flat and ZL silica films (see Surface binding constant of the probe/target duplex. Since low Fluorescence). target densities, such as 0.3 ( 0.1 pmol cm-2 at 45 °C The 0.4 and 0.5 µm ZL films did not reach equilibrium for two-base mismatches on the flat glass, and high at the first 25 °C time point, but all films appeared to densities (0.8 ( 0.2 pmol cm-2 at 25 °C for the perfect- be at equilibrium at the successive time points, as the match probes) are enhanced by the same factor on the High-Capacity DNA Probe Arrays Chem. Mater., Vol. 13, No. 12, 2001 4779 Table 3. Comparison of Measurements on 0.5 µm ZL Films (measured ratio to flat glass)/ measurement (surface area ratio to flat glass) (%) surface area 100 [OH] site density 84 ( 10 surface fluorescencea 76 ( 5 hybridizationb 70 ( 6 a Average of measurements with fluorescein/diluent ratios from 0.001 to 0.05. b Perfect-match probes. silica films relative to flat glass, there is further indica- tion that quenching is not more significant on the silica films. If quenching was a factor on the silica films in the concentration range observed, then the multiple at high concentration would be less than the multiple at low concentration, where there is greater distance Figure 8. Properties of dried ZL silica solution by nitrogen between fluorophores. adsorption. The majority of the area is contained in pores that It should be noted that the 100 nM target concentra- are ∼23 nm in diameter. tion is much higher than typically present in genomic assays, which typically range from 1 to 100 pM, and was nitrogen molecule is used for determination of surface chosen to observe probe/target binding under saturating area, this measurement serves as a reference point for conditions. A much higher discrimination ratio for the the true adsorption capacity of the film. Synthesis and matched and mismatched probes is routinely achieved hybridization require the use of larger molecules which under practical conditions and hybridization protocols, may not be able to access all sites on the surface, as as discussed by Fidanza et al.,21 which presents further will be discussed in the following sections. evidence that discrimination on the colloidal silica films Before proceeding with an analysis of proximity is equivalent to flat glass for gene transcripts of average effects, it is first necessary to consider the film geometry. length approximately 50 bases1 over a range of concen- Figure 8 shows the incremental and cumulative surface trations. area vs pore diameter for the ZL films, with the pores Stability. The thickness of the films was tested before being the void spaces between the solid particles. The and after the hybridization experiments. Within the average pore size is 23 nm (see Table 1), and corre- uncertainty of the measurement, no difference in the spondingly the majority of the surface area is in these thickness of the layer was observed. Additionally, the pores (50% of the surface area is within one standard testing times were much longer than the standard deviation and 80% is within two standard deviations). conditions used for gene expression assays. Therefore, to further analyze proximity effects, it is necessary to develop a model of the area distribution Discussion within these pores. Colloidal silica films created from large colloidal Geometric Model of the Porous System. As is precursors, such as Snowtex 20L and ZL, provide good standard in the analysis of porous substrates, a simpli- substrates for high-capacity DNA arrays. Hydroxyl site fied model of a typical pore must be developed, which density on the colloidal silica films is proportional to we have done by consideration of the properties of the the surface area, and synthesis yield is equivalent to porous film. First, the index of refraction of the ZL films flat glass. Furthermore, the fact that the probes “func- is typically 1.3, which corresponds to films that are 68% tion” in the same manner in the porous matrix as on dense, close to the theoretical packing density for same- flat glass, as shown by the equivalent discrimination, sized spheres of 74%.32 For the ZL suspension, the serves as further evidence that chemical synthesis and particle size (Ds) is 65 ( 16 nm and the pore size (Dp) is hybridization proceed efficiently in the porous matrix. 23 ( 2 nm. The ratio Dp/Ds ) 0.35 is similar to the Fluorescence quenching affects both flat and ZL silica relative size of the octahedral interstitial site in a matrix films at the same concentration of surface-bound fluo- of close-packed, same-sized spheres, where Dp/Ds ) 0.41. rophores, and therefore the signal amplification ob- Therefore, this matrix and interstitial site can serve as tained with the ZL silica films is independent of the a model for the pore geometry, and to obtain a better bound target concentration over the concentration range estimate of the actual spacing in the pore, we have examined. developed a “finite-elements” analysis of the pore space Although both site density and hybridization enhance- based on this system. ment on the silica films are proportional to surface area, Figure 9a shows a “top view” schematic of the we observed, by comparison of Figures 4 and 6, that in octahedral pore site (with radius ) Rp) in a close-packed comparison to flat glass, the enhancement multiple for system of same-sized spheres. To calculate the actual hybridization < hydroxyl site density < surface area. spacing of adjacent surfaces in the pore, we examine a To understand these results, we must assess the im- section of sphere S0, shown as the shaded region. Figure pacts of proximity effects such as steric hindrance and 9b shows that this section, surface ABF, covers 1/16th electrostatic repulsion on the performance of the films. of the surface area of sphere S0 and is representative of The comparison of the multiples obtained from the various techniques provides a starting point to assess (32) Kingery, W. D.; Bowen, H. K.; Uhlmann, D. R. Introduction to these factors, as shown in Table 3. Since the small Ceramics, 2nd ed.; Wiley & Sons: New York, 1976; p 57. 4780 Chem. Mater., Vol. 13, No. 12, 2001 Glazer et al. Figure 9. (a) Geometric model of a pore (radius ) Rp) between ZL silica particles. The measured pore size (∼23 nm) is close to the predicted pore size for the octahedral pore in a close-packed system (∼27 nm). (b) For any point C (x1, y1, z1) on surface ABF, d1 is the shortest distance (“clearance”) to point D on an adjacent sphere. The distribution of surface area vs clearance can be determined by calculating the minimum of the distances to sphere S1 (at {2R, 0, 0}) and S2 (not shown, “above” pore at {R, R, 1.414R}) for each point on ABF. In the finite-elements approach, each point covers an incremental surface area R(dθ)(dφ). the surface inside the void space. For a given point C (at {x1, y1, z1}) on surface ABF, the distance d1 to the closest point D on adjacent sphere S1 can be calculated by drawing a line from C to the origin of sphere S1 (point E). As a line extending from the radius is perpendicular to the tangent at that point, the segment CD is the shortest distance to the adjacent particle. The same calculation can be done for sphere S2 (not shown, “above” pore at {R, R, 1.414R}). Equations 3a-3f have been used to determine the distance distribution. x ) R cos φ cos θ (3a) y ) R cos φ sin θ (3b) z ) R sin φ (3c) Figure 10. Distribution of surface area within 23 nm pores. Points on the surface of a given particle are much closer to a point on an adjacent surface than implied by the “average” d1 ) R(x(x1 - 2)2 + y12 + z12 - 1) (3d) pore size. The silane layer, which is 1-2 nm thick, causes ∼15% of the surface, located in the crevices of the large pores, to be inaccessible for probe synthesis. During hybridization, x d2 ) R( (x1 - 1)2 + (y1 - 1)2 + (z1 - x2)2 - 1) (3e) target molecules are not able to access probe sites in the crevices, excluding an additional 15% of the surface. d ) min(d1,d2) (3f) Site Density and Coupling Yield. Prior to coupling of oligonucleotide probes, the surface is coated with a d is defined as the “clearance” or the shortest distance silane coating, the thickness of which has been deter- to a point on an adjacent particle. θ and φ are shown in mined to be 1-2 nm.33 The density of sites measured Figure 9b. Surface ABF is divided into elements with by the HPLC C3-fluorescein procedure was 84% of the surface area R(dθ)(dφ), and the closest point to an surface area increase (see Table 3). On the basis of adjacent particle for each element is determined. By Figure 10, this result implies that the effect of the silane increasing θ from 0 to 45° and φ from 0 to 90°, we can layer is to occlude from probe synthesis approximately determine the distribution of surface area vs clearance, the first 1.5 nm of clearance, which agrees well with as shown in Figure 10. It is evident from the figure that the measured thickness of the silane. the actual distances between points on adjacent surfaces is much less than implied by the “average” pore size (33) Forman, J., unpublished data. Silane thickness measured by distribution. This distribution provides the context for atomic force microscopy “scratch test” using Digital Instruments assessing the factors in Table 3. nanoscope. High-Capacity DNA Probe Arrays Chem. Mater., Vol. 13, No. 12, 2001 4781 Surface Fluorescence. Figure 5 shows that neither clearance is excluded by the presence of the silane the flat surface nor the porous films are susceptible to coating, we find that these hindrances would block up quenching at surface concentrations below approxi- to ∼40 and ∼80% of the sites, respectively. Therefore, mately 1 pmol cm-2, which is greater than the measured based on the data in Table 3, where it is cited that bound density of adsorbed target for the assays in this study. target reaches ∼70% of the available surface area, we Although the measured surface fluorescence increase is conclude that the limiting factor is the duplex width slightly lower than the available hydroxyl site density rather than the length. increase, the two measurements agree within experi- For case 2, sites spaced more closely than the radius mental uncertainty (see Table 3), and quenching effects of gyration of the diffusing molecule would be inacces- within the pores will not be considered further. sible. For a single-stranded target molecule diffusing in Hybridization to Oligonucleotide Targets. Two solution, the radius of gyration Rg of the random coil cases of proximity effects, such as steric and electrostatic can be estimated from eq 5.35 repulsion, are possible: (1) effects due to lateral spacing of probes on the surface of a particle and (2) effects due Rg ) (Lp/3)1/2 (5) to clearance between adjacent particles. Proximity Effects Due to Lateral Spacing of L ) Nbo, and for single-stranded DNA, bo has been Probes. On the basis of the good agreement between calculated as 4.3 Å.35 p is the persistence length, which the surface area of the films and the density of hydroxyl approaches the intrinsic persistence length pi for single- sites, it can be assumed that the lateral spacing between stranded DNA of 8-13 Å for the high ionic strength probes (and therefore between adsorbed target mol- solutions used in this study.35 With the use of p ) 13 Å ecules) on the silica particles is approximately the same and N ) 20, Rg was estimated at a maximum of 19 Å. as their spacing on a flat surface. The closest spacing This length is similar to the width of the duplex, so it of adjacent target molecules is estimated by assuming is difficult to assess which is the ultimate cause of a rectangular array of sites (eq 4), where σ is the density limitation, and further studies with targets of much of bound target (mol cm-2) and NA is Avogadro’s different probe lengths would be necessary. Addition- number. ally, long targets may be able to temporarily “unwind” from the random coil formation they take in free 1 1 solution and diffuse through openings smaller than the s) ) ) (σNA)1/2 ((1 × 10-12)(6.02 × 1023))1/2 radius of gyration. Fidanza et al.20 reports that the ability to uniquely identify target sequences was not 1.29 × 10-6 cm ) 12.9 nm (4) affected when fragmented biological targets with aver- For double-stranded DNA, L ) Nbo, where L is the age length of approximately 50 bases1 were hybridized length of the duplex, N the number of bases, and bo is to the porous matrix. In summary, the amount of bound 3.4 Å.34,35 For a 20-mer duplex, L ) 6.8 nm, which target on the porous layers approaches the maximum represents the maximum distance from the attachment possible for this type of surface morphology, given the to the surface that the duplex can extend, and therefore fundamental limitations of the duplex width and the probe/target duplexes should not exert significant steric size of the diffusing target molecules. or electrostatic influences on neighboring duplexes. Proximity Effects Due to Clearance. The ex- Conclusion tended duplex could be long enough to span small pores A porous inorganic system has been characterized and and crevices in large pores. As shown in Figure 8, small evaluated for use with small molecule assays such as pores of 10 nm and less comprise only ∼5% of the DNA arrays. Oligonucleotide synthesis and DNA hy- surface area in the film. It is evident that these small bridization have been carried out in pore sizes on the pores do not impede the penetration of the target or that order of the length of a probe/target duplex. The porous the target is able to find alternative routes, as the films show equivalent performance to flat glass in terms amount of bound target increases in proportion to the of efficiency of chemical coupling, resolution of photo- film thickness (see Figure 6). The influence of the lithographic patterning, optical scanning, and discrimi- smallest pores on hybridization kinetics has not been nation for target molecules but yield greatly increased evaluated, and it may prove that proximity effects in hybridization signals. Hybridization signals reach ∼70% the small openings have a profound effect on the of the maximum possible intensity based on the surface hybridization rate. area of the porous films, and we have presented a model Proximity effects within the crevices of the large pores that attributes the 30% hindrance to the portion of the were also considered. Two cases are possible: (1) initial film that is inaccessible to synthesis reagents and/or target penetrates the crevice, forms probe/target duplex, target molecules. and prevents penetration of further target or (2) target Colloidal silica substrates take longer to reach satu- never penetrates the crevice. For case 1, hindrance could ration than flat glass. Films that are 0.2 µm thick be possible from a minimum length of the duplex width reached equilibrium with the target solution in less than (∼2 nm for double-stranded DNA in solution34) to a 12 h at room temperature, whereas thicker films (0.4 maximum of the duplex length (6.8 nm). On the basis and 0.5 µm) reached equilibrium in 12-36 h. The of Figure 10 and assuming that the first 1.5 nm of difference in time constants is most likely due to a combination of limitations, including that mass transfer (34) Stryer, L. Biochemistry, 4th ed.; W. H. Freeman and Com- from solution may not be rapid enough to meet the pany: New York, 1998; p 791. (35) Tinland, B.; Pluen, A.; Sturm, J.; Weill, G. Macromolecules increased demand for target of the high-capacity surface 1997, 30, 5763-5765. and that the diffusivity of the target may be reduced 4782 Chem. Mater., Vol. 13, No. 12, 2001 Glazer et al. in the porous matrix. Aspects of the hybridization These properties make colloidal silica substrates prom- behavior such as kinetics and diffusion are currently ising candidates for future use in high-capacity DNA being investigated in greater depth. arrays. In comparison to other techniques for creating high- density surfaces for biomolecule array applications, Acknowledgment. We acknowledge the assistance colloidal silica films offer several advantages. For and support of Larry Bailey, John Bravman, Alice Gast, example, the use of silica as the base material makes Albert Lee, Fabian Pease, and Lubert Stryer of Stanford these films easy to integrate into array-fabrication University. We also thank Paul Bury, Paul Ciccolella, processes on glass or oxidized metal surfaces. Addition- Steve Fodor, Jon Forman, Bill Lyon, Dan Mutnick, ally, since DNA probes on the particle films thermo- Nineveh Parker, Michael Savage, Audrey Suseno, and dynamically behave the same as on flat glass, much of Mark Trulson of Affymetrix. Richard Vinci of Lehigh the knowledge gained from the extensive research University was extremely helpful with advice and conducted on standard flat glass substrates can be technical assistance, and Mike Alden of Beckman- readily translated to these films, without the neces- Coulter assisted with nitrogen adsorption measurements. sity of considering other types of target/matrix inter- actions that could occur with other matrix materials. CM010578N