The design and testing of a silica sol–gel-based hybridization array

Journal of Non-Crystalline Solids 350 (2004) 39–45 www.elsevier.com/locate/jnoncrysol The design and testing of a silica sol–gel-based hybridization array J.R. Phinney a, J.F. Conroy a, B. Hosticka a, M.E. Power a, J.P. Ferrance b, J.P. Landers b, P.M. Norris a,* a Department of Mechanical and Aerospace Engineering, University of Virginia, 122 EngineerÕs Way, Charlottesville, VA 22904, USA b Department of Chemistry, University of Virginia, Charlottesville, VA 22904, USA Available online 6 November 2004 Abstract An approach for building DNA hybridization arrays has been developed, in which silica aquogel arrays are produced using micropiezoelectric printheads. When supercritically dried, the pads in these gel arrays have a footprint size of 0.4 ± 0.1 mm in diameter in which the porous silica has a density of 0.045 g cm
3, pore diameters ranging between a few nanometers and 0.5 lm, and internal surface areas >800 m2 g
1 (5 · 10
4 m2 per pad). This enormous internal surface area, combined with reasonable accessibility to internal binding sites, makes aquogel pads ideally suited for hybridization arrays. Silanized probe DNA was immobilized within the gel pads during gelation, and hybridization was carried out with fluorescently labeled target DNA. The array sites containing probe DNA and the control sites without probes are readily distinguishable using laser-assisted fluorescent scanning.  2004 Elsevier B.V. All rights reserved. PACS: 82.70.G; 87.15.K 1. Introduction Hybridization of single-stranded nucleic acids to their complementary sequences is commonly used in many areas of molecular biology, including gene identification, gene expression, and mutation analysis. Nucleicacid hybridization is a selective binding that is based on the base-pairing of nucleotides. Hybridization arrays are often used to sequence DNA and to study gene expression, as well as to discriminate single nucleotide polymorphisms [1,2]. Known sequences of singlestranded probe DNA are attached to specific locations within the array. One or more unknown target sequences with either fluorescent or radioactive labels are then washed over the array and allowed to hybridize with complementary or nearly complementary probe. By identifying the locations within the array where the * Corresponding author. Tel.: +1 434 924 6295; fax: +1 434 982 2037. E-mail address: [email protected] (P.M. Norris). 0022-3093/$ - see front matter  2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2004.08.228 targets have hybridized with probe, it is possible to determine the sequence(s) of the unknown target(s). The identification step can be experimentally difficult due to a low signal-to-noise ratio. In this study, we seek to develop a three-dimensional pad array that has the potential to achieve a higher concentration of hybridized nucleic acid per individual pad. With these increased hybridized nucleic acid concentrations, the detectable signal (fluorescence in our study) to be measured by a scanning system should also be increased at the pad sites where hybridization has occurred. The increased signal should eliminate the need for background noise removal. The use of piezoelectric printheads for direct solid free-forming of ceramics and sol–gels is a well-established technique. Xiang et al. demonstrated one of the earliest applications of piezoelectric printing [3]. They reported that the formulation of a stabilized ceramic ink and the adaptation of a drop-on-demand jet printer have exciting implications for the miniaturization of ceramic circuits and the fabrication of solid oxide fuel 40 J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 cells. Atkinson et al. reported another early application of inkjet printing using a sol as the ink, where the ink was used as decoration on ceramic articles [4]. The ink can either be directly printed on the ceramic or printed onto paper and applied as a gel transfer. MacCraith and McDonagh reported the interaction of analytes with entrapped analyte-sensitive fluorophores in microporous sol–gel-derived matrices as the analytes diffused through the matrix [5]. These authors also discussed sensor optimization made possible by the versatility of the physicochemical tailoring of the film. Silica sol–gels have many characteristics that make them ideal for use in DNA hybridization arrays. These characteristics include high internal surface areas and porosities, as well as production routes that do not require extreme temperatures and are thereby biocompatible, for example, that developed by Conroy et al. [6]. In addition, the surface of a silica sol–gel can be functionalized to increase the energy cost associated with non-specific adsorption of DNA. Since non-specific adsorption is a significant source of background fluorescence, such functionalization can decrease the background noise in the system. Biocompatible silica gels can be formed in a two-step gel formation process that yields silica gels having an almost entirely aqueous interstitial fluid, rather than the ethanol produced by traditional sol–gel methods [6,7]. The first step in the gel formation process is hydrolysis of an alkoxysilane such as tetraethoxysilane (TEOS) in a low pH aqueous solution. This hydrolysis yields various silicates that are sufficiently polar to dissolve in a largely aqueous solvent [6]. Conroy et al. recommend the use of an aqueous nitric acid solution to catalyze the hydrolysis [6]. Here, however, propyltrichlorosilane (PTS) replaced nitric acid as the catalyst, since a molar equivalent of HCl is released by hydrolysis of PTS in water. After the TEOS was hydrolyzed and the hydrolysis products dissolved in the aqueous solvent, the solution temperature was increased to 85 C and an open-pot distillation was performed for 1 h to remove hydrolysis reaction product ethanol [6]. The second step in the gel formation process involves an increase in the pH of the acidic sol using NaOH to catalyze gelation. Silica gels have a porous internal structure that results from the manner in which the gels form. Factors such as the presence of dispersants (e.g., polymers) in the gelation solution, temperature, electrolyte concentration, and type of alkoxide precursor can affect the structure of silica gels [6,8]. Prior to gelation, oxymetallates diffuse and condense in solution, leading to chain growth until a solid network spans the container yielding a gel in the shape of the container. The small size of the precursor and the speed with which gelation occurs prevent settling and leads to a relatively isotropic structure. After an initial spanning of the structure, gels can be allowed to Ôage.Õ An aging process involves the addition of material to the silica matrix by condensation of any unreacted species. The solid network of silica gels commonly has a porous fractal structure [9]. In this study, the solid network had pore sizes ranging from a few to a few hundred nanometers. It is within these pores that we intend to bind probe DNA. Assuming that the internal matrix contains sufficiently large pores, the target DNA can diffuse through the gel pads and hybridize with the probes bound within. 2. Experimental 2.1. Sol preparation A biocompatible sol, referred to as 2· PTS TEOS, was produced as follows [6]. To 50 ml of deionized water, 86 ll of proplytrichlorosilane (PTS, Aldrich Chemicals 98% grade) was added, mixed at 225 rpm, and heated to 60 C. An aliquot of 12.5 ml of tetraethoxysilane (TEOS, Aldrich Chemicals 98% grade) was then combined with the previous mixture and stirred for 10 min before increasing the temperature to 85 C for 1 h. The solution was then allowed to cool to room temperature and sealed for storage. Stirring continued during storage to prevent premature gelation. 2.2. Printing technique Inkjet printheads have been used to dispense biological materials in the past [10]. An Epson Stylus Color 580 printer based on the Ô4-color drop-on-demand Micro Piezo inkjet technologyÕ was used [11]. The Micro Piezo inkject technology is considered preferable to the more common thermal inkjet printheads because heat may increase the rate of gelation of the sol and thus potentially clog the printing nozzle. Also, the selected printhead is separate from the ink cartridge, thus making the printhead easier to remove and clean. Finally, the Epson Stylus Color 580 printhead is one of the least expensive means for producing drops with volumes as low as 6 picoliters without heating and causing premature gelation. The removable Epson Stylus Color 580 printhead consists of four separate color channels leading to a series of piezoelectric nozzles, each with an 20 lm diameter orifice. The nozzles are arranged in two rows with the ÔblackÕ channel leading to 48 nozzles and each of the three ÔcolorÕ channels leading to 45 nozzles [11]. Upon receipt, the printhead was removed from the printer and the passageways and nozzles were cleaned by attaching a 1 ml BD latex-free syringe to each inlet and flushing with ethanol. A dry syringe was used to flush out any remaining ethanol from the body of the printhead. Eppendorf microcentrifuge tubes (1.5 ml, Fisher Scientific) with the bottom shaved off were used J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 41 as a reservoir for the fluid to be dispensed. Glass cover slips, used as a deposition substrate for the gel pads, were activated by submerging in 1 M NaOH for 3 h at 60 C and dried. 2.3. Probe DNA In order to covalently link probe DNA to the silica sol–gel, silanization of the probe oligonucleotide, 5 0 AGT ACC GGG AGG CGA GCG ACC CAC TTG TTG ACT TGG-3 0 , was carried out. First, 20 ll of 5 0 aminated DNA was added to 10 ll of aldehyde-modified silane and allowed to react for 1 h at 20 C [12]. A 200 ll aliquot of 0.1% sodium dodecylsulfate (SDS) solution in deionized (DI) water was added to the reaction and put in a 10 000 MW cutoff centrifuge filter and spun at 7000 g for 10 min, then 200 ll of sodium borohydride buffer was added to the residue in the filter and allowed to react for 3 min before centrifuging again for 10 min at 7000 g [12]. Following the second filtration, 200 ll of 95 C DI water was added to the residue in the filter and allowed to sit for 5 min before centrifugation at 7000 g for 30 min. Finally, 100 ll of Tris-EDTA (TE) buffer was added to the residue in the filter and the entire volume transferred to 0.5 ml Eppendorf tubes that were stored at
20 C until used. 2.4. Printing the sol–gel pads As discussed above, the TEOS-based gel recipe of Conroy et al. requires the neutralization of an acidic prehydrolyzed sol to achieve gelation. Upon neutralization, gelation occurs within a few minutes. It is thus difficult to first neutralize the prehydrolyzed sol and then print within this short time period. Therefore, in our system, neutralization was used to form pad arrays by printing the various reagents from multiple nozzles onto the same location on a substrate [6]. Our printing system combined paper drive systems from several defunct printers as well as an old printhead carriage-drive setup. The control signals from a working printer were used to drive the amalgamation. Instead of driving paper, a belt made from a clear plastic film was used to translate an activated cover slip in one direction while the printhead was rastered perpendicularly above it to scan an x
y plane. Fig. 1(a) shows a schematic of the experimental setup while Fig. 1(b) shows a top view of the actual device. The carriage with one of the fluid reservoirs and the belt can be seen. The selected print pattern was a Microsoft Word document that included a single page of period characters (i.e., Ô. . .Õ) set to 18 point, Times New Roman font, with a color selected by eye to contain a majority of black followed by a small combination of magenta and yellow. The color was selected from a predefined palette within Microsoft Word and has red, green, Fig. 1. (a) Schematic of the device built to control drop placement; (b) top view of the actual device. and blue values of 153, 51, and 0, respectively. This print pattern thus used three different nozzles (i.e., black, magenta, and yellow) at once. The bottomless 1.5 ml Eppendorf tube reservoirs were secured to the inlets to the black, magenta, and yellow ink channels. The channels were then primed and the reservoirs filled as follows. The black channel was filled with the 2· PTS TEOS sol that was filtered using a 0.20 lm Nalgene syringe filter prior to loading into the reservoir. The magenta channel was filled with 1M NaOH that had also been filtered using a 0.20 lm syringe filter. The yellow channel was filled with 50 ml of sterilized water and 10 ml of a 0.2–0.4 pmol ml
1 silanized probe DNA. For the control gel pads, the yellow channel was left empty. Once the print command was submitted through the computer, the cover slip was placed on the backside of the belt with tweezers as soon as the printer started to feed in paper. As soon as the cover slip had cleared the nozzles on the printhead, the slip was removed from the belt with tweezers and placed on a platform inside a 100 mm polystyrene Petri dish that was partially filled with water. The Petri dish cover was then sealed to the base using Parafilm and the Petri dish was set aside to 42 J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 allow the pads to gel. The additional water in the dish was necessary to prevent the pads from drying out. Each of the pads had a probe DNA concentration between 6.1 · 10
8 and 1.2 · 10
7 pmol ml
1. 2.5. Hybridization Gel pads were aged for not less than 24 h prior to hybridization. The hybridization buffer comprised 20% formamide, 0.1% SDS, and 2· saline-sodium citrate (SSC). A 20 ll aliquot of SDS was added to 2 ml of SSC and 4 ml of formamide, followed by 14 ml of sterilized water. This formulation was used in order to decrease any non-specific binding of DNA target molecules. Next 10 ll of the fluorescein-labeled DNA targets (2 pmol/ll, sequence (5 0 to 3 0 ) fluoroscein-TCA TGG CCC TCC GCT CGC TGG GTG AAC AAC TGA ACC) was added to the 20 ml of buffer and again gently mixed. Two ml of the hybridization buffer was then added to five wells in two six-well, sterile, flat-bottomed cell culture plates. One cover slip was submerged with the gel pad side up in each of the filled wells. The cell culture plate covers were then fitted, and the culture plates were moved to a rotating apparatus (Thomas model #3623) that was covered and set to gently shake for 48 h at room temperature. This hybridization period was chosen to provide sufficient time for the target to diffuse through the gel pad. Following hybridization, each of the wells was drained of hybridization buffer and filled with 2 ml of a wash buffer solution to rinse the pads. The wash buffer solution was the hybridization buffer without DNA. The plates were returned to the rotating apparatus and rinsed for 1 h. The rinse was repeated twice. 2.6. Detection The fluorescence detection system is shown in Fig. 2. A beam from an argon ion laser (Reliant 150 m; LaserPhysics) is directed through a 505 DRLP filter dichroic mirror (Omega Optical) functioning as a beamsplitter. The beam is sent through a 16· Melles Griot microscope objective that focuses the beam on the gel pad to be scanned. If the gel pad has fluorescein-labeled target DNA, then it is excited by the laser and light reflects back through the objective and through the beamsplitter to a Melles Griot mirror. The reflected beam is then sent through a bandpass filter (Omega Optical 530DF30) and into a photomultiplier tube, PMT (H5784-01; Hamamatsu). The signal from the PMT is recorded by the data collection system using Labview software [12]. A special platform was designed to maintain a constant velocity of the cover slip through the fluorescent probing field. A plexiglass platform with a metal rod and a hole in the center ensured that the beam would hit only the samples on the cover slip and not the plexiglass platform, which would reflect enough light to saturate the PMT. This platform was then driven at a constant velocity of 0.15 mm s
1 by a syringe pump (KD Scientific model 200). For each scan, the PMT was set to maximum gain. After a pad was located, scanning was stopped, reset, and the platform was set to translate the pad through the beam and data collection started. Each scan was recorded separately. 3. Results Light microscopic observations indicate gel pads were produced with a diameter of 0.4 ± 0.1 mm when printed as described above. In the scan depicted in Fig. 3, four gel pads are clearly evident with widths of (moving from left to right) 1.56, 1.04, 1.07, and 1.39 mm (population mean 1.5 ± 0.4 mm). The spacing between the four pads, when measured from peak-to-peak between nearest neighbors, is 2.29, 2.28, and 3.35 mm (population mean 3.6 ± 0.9 mm). The control pads in Fig. 4 show corresponding pad widths of 0.9, 1.48, 1.14, and 1.2 mm (population mean 1.4 ± 0.7 mm). The peak-to-peak spacing of these pads between nearest neighbors, working from left to right, is 3.55, 3.15, and 4.28 mm (population mean 3.4 ± 0.9 mm). The fluorescence measurements cannot be used to measure absolute spot size, because the fluorescence sig- Fig. 2. Diagram of the fluorescence detection system used to detect hybridization in the gel pads [12]. J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 3.5 Gel Pads Fluorescence (rfu) 3 2.5 2 1.5 1 0.5 0 0 5 10 Displacement (mm) 15 Fluorescence (rfu) Fig. 3. Fluorescence scan of 2· PTS TEOS pads with probe DNA present (taken at a displacement rate of 0.15 mm s
1 with the PMT set to maximum gain). 2 1.8 1.6 1.4 Gel Pads 1.2 1 0.8 0.6 0.4 0.2 0 0 5 10 Displacement (mm) 15 Fig. 4. Fluorescence scan of 2· PTS TEOS pads without probe DNA present (taken with the PMT set to maximum gain and displacement rate of 0.15 mm s
1). nal is detected as a function of time as the gel pad is pushed through the laser at a constant rate and then converted to pad displacement. The fluorescence is measured from the location of first detection to where it can no longer be distinguished from background noise. Thus any fluorescent signal that continues after the pad has passed out of the beam and the beam size at that location would play roles in the apparent size shown and would also decrease the ultimate resolution. The fluorescent signal does confirm the light microscopic evidence that the gel pads were consistently evenly spaced by the inkjet printer apparatus and that DNA hybridization has occurred. A comparison of the fluorescent scans depicted in Figs. 3 and 4 demonstrate a significant difference in fluorescence between pads made with probe DNA and those pads made without any DNA. The conditions of hybridization for both sets of pads were the same in that target DNA was allowed to diffuse into the pads and then non-specifically bound target was washed off under reasonably stringent conditions. Fig. 3 is a representative scan of gel pads containing probe DNA. From left to right the magnitudes are 2.866, 1.418, 1.453, and 2.058 relative fluorescent units (rfu). 43 An example of a scan done on control gel pads without probe DNA, but subjected to hybridization with fluorescein-labeled DNA and washing similarly to those pads with the probe DNA is shown in Fig. 4. There are four distinguishable gel pads present; however, the amount of fluorescence is significantly lower than in those pads with probe DNA. This fluorescence is most likely the result of non-specific adsorption of fluorescently labeled target. This may be due, in part, to washing conditions that were not stringent enough. From left to right, the fluorescent magnitudes are 0.774, 0.879, 0.781, and 0.808 rfu. An analysis of variance (ANOVA) was performed to determine whether there was a significant difference in the amount of fluorescence in the experimental and control groups of gel pads. The F ratio associated with the data is 20.9, which is much greater than the critical value for this test (4.009877). The null hypothesis that there is no difference between the mean values of the two sample populations with greater than 95% confidence in our results can be rejected [13]. To narrow this further, the P value specific to these data is 2.67 · 10
5, providing 99.99733% confidence that there is a difference between the mean values of the two populations. Hence the target DNA did not degrade and was hybridized to its complement probe DNA within the gel pads [13]. Experiments carried out with longer hybridization times resulted in the pads washing off the cover slip (data not shown). In order to calculate the pad volumes and thus enable calculation of the probe concentration, the drop diameters were measured from micrographs for both the individually printed pad components and the total pads. These diameters were used to calculate the pad volumes by assuming the pads to be hemispherical. The mean volume for the components was 5.8 ± 1.3 nl of silica sol plus 200 ± 40 pl of 1 M NaOH catalyst for the gel pads without probe DNA, and for the pads with probe DNA, these values plus the addition of 11 ± 5 pl of a solution containing 10 ll of 0.2–0.4 pmol/ll probe DNA with 50 ll of sterilized water. From these drop-size calculations, the volumetric ratio of silica sol:cat:probe solution was determined to be approximately 524.3:18.4:1. This volumetric ratio was used to make larger gels of the same composition that could then be used for structural characterization. Also, on the basis of this information, it appears that there were between 1.1 · 10
3 and 6.6 · 10
3 fmol of probe in each pad in the array. Bulk gels were produced with a volumetric ratio of silica sol:cat:probe of 524.3:18.4:1. These gels were then supercritically dried to produce aerogels so that the microstructure could be examined using a gas adsorption-desorption isotherm technique [7]. The results of the tests using a Micromeritics ASAP 2010 accelerated surface area and porosimetry system showed the Brunauer–Emmett–Teller (BET) measured internal surface area to be 832 m2 g
1. The average pore size 44 J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 Fig. 5. Electron micrograph taken of the surface of a supercritically dried 2· PTS TEOS gel taken at 25 kV and 5000· magnification. Note the 1 lm scale bar in the lower right corner. was reported to be close to 10 nm, although pores with sizes on the order of 0.5 lm in diameter are clearly present in the micrograph shown in Fig. 5. This discrepancy is not surprising, as nitrogen sorption porosimetry is only capable of detecting pores between 1.7 and 300 nm in diameter. The larger pores observed in the micrograph confirm the fluorescent observation that diffusion of the small oligonucleotide target is feasible. 4. Discussion DNA hybridization within silica gel pads was demonstrated. The results of the fluorescent scanning confirm that the probe DNA remained within the gel pad and was not denatured during either the printing process or sol–gel formation and at least a certain population of the probe DNA was also accessible to its complementary DNA for hybridization. Additionally, Fig. 3 shows that the scans performed on the 2· PTS TEOS gel pads produced readily identifiable gel pads with fluorescence most intense near the center of the gel pads and decreasing towards the edges. This behavior suggests that the probe DNA hybridizes to its complementary DNA strand inside of the gel pads instead of just on the surface and that the pores of the gel are large enough to allow the fluorescently labeled target DNA to enter the matrix and hybridize with the probes within. Indeed, if the hybridization takes place on a three-dimensional substrate, then the strongest signal should be produced where the gel pad is thickest. If hybridization were only to take place on the surface, then the signal should be similar across the pad. This new technique compares well with other methods currently in use and under development, as shown in Table 1. The probe concentration, while nearly the same as that of the other inkjet method [10], is significantly less than that used in the MAGIChip polyacrylamide gel pads [14] and more than that used by Affymetrix [15]. At the same time, the gel pads used are smaller than those of the other inkjet method but larger than the in-situ formation and polyacrylamide gel pad approaches [10,14,16]. With respect to hybridization protocols, the method discussed here used an intermediate concentration of fluorescently labeled target DNA, while only Affymetrix uses less [10,14,16]. The design parameters for the developed system may be altered to test different pad sizes as well as probe and target concentrations, hybridization and washing times, and temperatures. Any of the other methods, including those under development, require many steps to produce the desired array. To produce the GeneChips array, Affymetrix requires up to one hundred different masks, and each masking step requires washing to remove deprotection byproducts. This washing is then followed by the addition of a single activated and protected nucleic acid, which is followed by additional washes to remove excess reagent [16]. The MAGIChip gel pads require photopolymerization, multiple washings, and incubation in twophase systems with yet more washings [14]. One of the benefits of the sol–gel method described herein is that it eliminates washing and complicated and contaminating post-production procedures. The pads are spotted using three channels of an inkjet printhead. This patterning is done in a manner similar in complexity to that used by Allain et al. [10]. Although only a single probe was used in these experiments, additional probes can be added to adjacent pads by adapting the DNA print nozzle with a multichannel control. The silica gel pads provide an increase in the density of hybridization sites. This should provide increased fluorescence per unit area. There is over 5000 times more surface area available for probe binding in a threedimensional pad occupying a 350 lm diameter footprint Table 1 Comparison of four prominent array types Method Pad shape Pad size Number of probes per pad (fmol) Target concentration (pM) Affymetrix GeneChip [16] Inkjet printing of DNA onto glass MAGIChip [14] polyacrylamide gel Inkjet printing of 2· PTS TEOS gel pads onto glass Square Drop Rectangular prism Drop 20 · 20 lm2 420 lm diam. 100 · 100 · 20 lm3 353 lm diam. 1.83–3.32 · 10
8 0.95–1.02 · 10
3 30–300 1.12–6.55 · 10
3 1–100 5 · 10
5 1 · 10
6 1 · 10
3 J.R. Phinney et al. / Journal of Non-Crystalline Solids 350 (2004) 39–45 than there is in a two-dimensional array location. If even a fraction of this gel pad surface area can be used to detect hybridization, it could then lead to dramatic improvements in the fluorescent signal. Further experiments optimizing probe and target concentrations and hybridization times should enhance the value of this technique. We believe this is the first time DNA has been incorporated into a silica sol–gel for the purpose of hybridization and the applications of this work can extend from hybridization arrays to miniaturized pads to immobilize DNA for Ôlab-on-a-chipÕ hybridizations. Acknowledgment The authors gratefully acknowledge NIH grant #R21HG02024-01A1 for funding this research project. References [1] Z. Strezoska, T. Paunesku, D. Radosavljevic, I. Labat, R. Drmanac, R. Crkvenjakov, Proc. Natl. Acad. Sci. USA 88 (1991) 10089. 45 [2] D.J. Lockhart, H. Dong, M.C. Byrne, M.T. Follettie, M.V. Gallo, M.S. Chee, M. Mittmann, C. Wang, M. Kobayashi, H. Horton, E.L. Brown, Nature Biotechnol. 14 (1996) 1675. [3] Q.F. Xiang, J.R.G. Evans, M.J. Edirisinghe, P.F. Blazdell, J. Eng. Manufac. 211 (1997) 211. [4] A. Atkinson, J. Doorbar, A. Hudd, D.L. Segal, P.J. White, J. Sol–Gel. Sci. Technol. 8 (1997) 1093. [5] B.D. MacCraith, C. McDonagh, J. Fluorescence 12 (2002) 333. [6] J.F.T. Conroy, M.E. Power, J. Martin, B. Earp, B. Hosticka, C. Daitch, P.M. Norris, J. Sol–Gel. Sci. Technol. 18 (2000) 269. [7] C.J. Brinker, G.W. Scherer, Sol–Gel Science: The Physics and Chemistry of Sol–Gel Processing, Academic, 1990. [8] L.L Hench, J.K. West, Chem. Rev. 90 (1990) 33. [9] J. Fricke, T. Tillotson, Thin Solid Films 297 (1997) 212. [10] L.R. Allain, M. Askari, D.L. Stokes, T. Vo-Dinh, Fresnius J. Anal. Chem. 371 (2001) 146. [11] Epson America, Epson Stylus Color 580 Printer Specifications, 2000. [12] J. Ferrance, personal communication, 2002. [13] J.S. Milton, J.C. Arnold, Introduction to Probability and Statistics: Principles and Applications for Engineering and the Computer Sciences, 3rd Ed., McGraw-Hill, 1995. [14] D. Proudnikov, E. Timofeev, A. Mirzabekov, Anal. Biochem. 259 (1998) 34. [15] S. Trask, personal communication, 2002. [16] M.C. Pirrung, J.L. Read, S.P.A. Fodor, L. Stryer, U.S. Patent #6,261,776, 2001.