Generation of Transducers for Fluorescence-Based Microarrays with

The present paper describes a novel generation of microchips suitable for fluorescence-based assays, such as cDNA, oligonucleotide, or protein microar...
0 downloads 0 Views 109KB Size
Anal. Chem. 2003, 75, 2571-2577

Generation of Transducers for Fluorescence-Based Microarrays with Enhanced Sensitivity and Their Application for Gene Expression Profiling Wolfgang Budach,* Dieter Neuscha 1 fer, Christoph Wanke, and Salah-Dine Chibout

Novartis Pharma AG Switzerland, 4002 Basel, Switzerland

The present paper describes a novel generation of microchips suitable for fluorescence-based assays, such as cDNA, oligonucleotide, or protein microarrays. The new transducers consist of a fully corrugated surface coated with a thin layer of Ta2O5 as a high refractive index material. Tuning of the incident excitation light beam to abnormal reflection geometry results in a confinement of the energy within the thin metal oxide layer. Consequently, strong evanescent fields are generated at the surface of these microchips and fluorophores located within the fields showed up to a 2 order of magnitude increase in fluorescence intensities relative to the epifluorescence signals. We have attributed this phenomenon as evanescent resonance (ER). Due to the surface architecture, propagation distances of the incident energy and fluorescence photons are in the micrometer range, thus preventing cross talk between adjacent regions. ER microchips offer a significant increase in fluorescence intensities in both “snapshot” fluorescence setups and commercial fluorescence scanners. The underlying principle of the novel chips is explained, and quantitative data on the fluorescence enhancement are provided. To demonstrate their potential, the novel chips are used to investigate the dependence of expression levels from metabolic genes in rat liver on drug treatment. In contrast to competitive hybridization, labeled samples were hybridized to individual ER microchips, and changes were observed by comparing with normalized data from different chips. Results obtained in gene expression profiling experiments with phenobarbital-treated rats are shown. The development of optical sensors for the detection of bioaffinity reactions has been of strong interest in the past decades. Besides cost reduction, many of these efforts were made to reduce the preparation steps involved, for example, by using label-free detection of analytes,1-3 or to lower the detection limits by using chemical amplification schemes.4-6 In addition, various * Corresponding author. E-mail: [email protected]. (1) Sauer, M.; Brecht, A.; Charisse´, K.; Maier, M.; Gerster, M.; Stemmler, I.; Gauglitz, G.; Bayer, E. Anal. Chem. 1999, 71, 2850-2867. (2) Kochergin, V.; Avrutsky, I.; Zhao, Y. Biosens. Bioelectron. 2000, 15, 283289. (3) Kunz, R. E.; Edlinger, J.; Sixt, P.; Gale, M. T. Sens. Actuators, A 1995, 4647, 482-486. (4) Hartley, D. P.; Klaassen, C. P. Drug Metab. Dispos. 2000, 5, 608-616. 10.1021/ac026390k CCC: $25.00 Published on Web 05/01/2003

© 2003 American Chemical Society

physical principles have been further developed such as total internal reflection, surface plasmon resonance, or resonance light scattering7-10 in order to increase the sensitivity and to lower the detection limits. However, although many of these new optical sensors have shown excellent performance for specific applications, the technical requirements/costs are high and the throughput is limited. At the same time, new biological methods such as gene and protein profiling based on microarrays have been developed by academical and industrial institutions.11,12 In contrast to abovementioned methods, many of these microarray approaches are based on multiplexed fluorescence assays on “passive” glass substrates and lack precision and sensitivity. For instance, in gene expression studies with commercial microarrays, about 40-50% of the genes are found to be below the detection limit; i.e., the sensitivity is not sufficient to provide information on low-abundant genes (so-called “absent” genes). The experimental errors impede differentiation between false positive/negative results as well as determination of significant changes for the expression levels of genes. Increasing the sample amount in order to increase the data quality is not always possible because of costs and limited availability of the sample material. However, the new methods used in genomics, proteomics, toxicogenomics, genetics, etc, are becoming extremely important for the pharmaceutical industry and probably soon also in other areas (such as medical care, diagnostics, point of care, and personalized medicine). The microarray concept requires low cost, high sensitivity, low cross talk, high reproducibility, robustness, and high degree of multiplexing (high density of capture elements) in order to miniaturize and reduce the sample requirement. These necessities have to be considered for the development of sensors. As a step toward fluorescence-based sensor systems with enhanced sensitivity, we have been reporting about planar (5) Zhang, D. Y.; Brandwein, M.; Hsuih, T. C. H.; Li, H. Gene 1998, 211, 277285. (6) Rijke, F.v.d.; Zijlmans, H.; Li, S.; Vail, T.; Raap, A. K.; Niedbala, R. S.; Tanke, H. Nat. Biotechnol. 2001, 19, 273-276. (7) Vo-Dinh, T.; Cullum, B. Fresenius J. Anal. Chem. 2000, 366, 540-551. (8) Pasternack, R. F.; Collings, P. J. Science 1995, 269, 935-939. (9) DeLisa, M. P.; Zhang, Z.; Shiloach, M.; Pilevar, S.; Davis, Ch. C.; Sikis, J. S.; Bentley, W. E. Anal. Chem. 2000, 72, 2895-2900. (10) Pipino, A. C. R. Phys. Rev. Lett. 1999, 83, 15, 3093-3096. (11) Chee, M. S.; Yang, R.; Hubbell, E.; Berno, A.; Huang, X. C.; Stern, D.; Winkler, J.; Lockhart, D. J.; Morris, M. S.; Fodor, S. P. A. Science 1996, 74, 610-614. (12) Schena, M.; Shalon, D.; Davis, R. W.; Brown, P. O. Science 1995, 270 (5235), 467-470.

Analytical Chemistry, Vol. 75, No. 11, June 1, 2003 2571

waveguides for multiplexed oligonucleotide assays.13 However, besides their high sensitivity, planar waveguides are critical with respect to robustness, reproducibility, and hardware requirements. In addition, planar waveguides show an increased background fluorescence and fluorescence cross talk due to optical near-field effects.14 In this paper, we present a novel approach using strong evanescent fields for an improved excitation of fluorophores, while avoiding the above-mentioned disadvantages. The novel transducer makes use of a uniformly corrugated surface coated with a high refractive index material (e.g., Ta2O5) to generate a localized evanescent at the illuminated chip surface. The phenomenon occurs under conditions for abnormal reflection,15,16 when the light incident on the chip has a specific angle with respect to the chip normal. We have attributed the formation of a strong evanescent field as evanescent resonance (ER). Since the surface of the ER chips is uniformly corrugated, the chips show virtually no fluorescence cross talk and are suitable for low-density to highdensity arrays without requiring any changes of the geometry. Sullivan and Hall17 used corrugated structures for increasing fluorescence signals. However, their sensor geometry required additional layers (LiF and metal), and the fluorescence enhancement was not correlated with the abnormal reflection phenomenon. ER transducers were optimized for almost perpendicular incidence of the excitation light and are, therefore, compatible with commercial instrumentation such as fluorescence scanners and microscopes. The amplification factor depends on the parameters of both the used chip and the excitation light and was found to be up to 2 orders of magnitude. The paper describes the principle of the novel evanescent resonator microchip and demonstrates the fluorescence enhancement based on gene expression profiling with real-world samples from pharmaceutical drug development. The novel ER microchip has the potential to improve the performance of microarray experiments and opens up new options for diagnostical tools. EXPERIMENTAL SECTION Chemicals. Dichlormethane (puriss. p.a., >99.8% (GC)), acetonitrile (for HPLC, >99.8% (GC)), o-xylene (puriss. p.a., > 99% (GC)), 2-propanol (Grade), and methanol (Grade) were used as solvents (Fluka). The silanization reagents (3-glycidoxypropyl)trimethoxysilane (purum, >97% (GC)), N-ethyldiisopropylamine (purum, >98% (GC)), and 2-amino-6-methylheptane (Grade) were purchased from Fluka. A mixture of 50% (v/v) Expresshyb hybridization buffer (Clontech) in formamide (>99%, Microselect, Fluka) containing 100 µg/mL sonicated salmon sperm DNA (Stratagene) was used as the hybridization buffer (HB). The wash buffer (WB) consisted of 50 mM NaCl (g 99.5%, Fluka), 20 mM sodium phosphate buffer pH 6.5 (>99%, Merck), 1 mM EDTA (Invitrogen), and 0.5% (w/ v) SDS (Fluka). (13) Budach, W.; Abel, A. P.; Bruno, A. E.; Neuschaefer, D. Anal. Chem. 1999, 71, 16, 3347-3355. (14) Marowsky, G.; et al. WO 99/40415, 1999. (15) Golubenko, G. A.; Svakhin, A. S.; Sychugov, V. A.; Tishchenko, A. V. Sov. J. Quantum Elecron. 1985, 15 (7), 886-887. (16) Avrutsky, I. A.; Sychugov, V. A. J. Mod. Opt. 1989, 36 (11), 1527-1539. (17) Sullivan, K. G.; Hall, D. G. J. Opt. Soc. Am. 1997, 14 (5), 1160-1166.

2572

Analytical Chemistry, Vol. 75, No. 11, June 1, 2003

RNA Isolation and Labeling. Total RNA from homogenized tissue was isolated using either the SV Total RNA Isolation System (Promega) or TRIzol Reagent (Invitrogen) and checked for integrity by agarose gel electrophoresis. A total of 5-10 µg of total RNA was denatured in the presence of 0.5 mg of oligo dT12-18 primer (Invitrogen) for 10 min at 70 °C and immedialtely cooled on ice. First strand cDNA synthesis was carried out for 2 h at 42 °C in the presence of 40 units of AMV reverse transcriptase in the respective buffer (Finnzymes), dNTP mix (0.2 mM dTTP, 0.5 mM dA/G/CTP), and 0.1 mM cyanine 5-dUTP (Perkin-Elmer) in a total volume of 20 µL. After incubation, samples were denatured at 100 °C for 5 min and immediately cooled on ice. Remaining RNA was digested by incubation for 15 min at 37 °C in the presence of 1 µg of RNase A (Sigma). Finally, labeled firststrand cDNA was purified using Microcon PCR columns (Millipore), dried down in a Speed-vac, and resuspended in HB. Chip Silanization Protocol. The evanescent resonator chips were mounted into chip racks, placed into a glass beaker, covered completely with CH2Cl2, and then sonicated for 15 min at room temperature. The solvent was decanted, and the chips were sonicated twice for 15 min in fresh 2-propanol and then transferred directly into a solution containing 2% v/v (3-glycidoxypropyl)trimethoxysilane and 0.2% v/v N-ethyldiisopropylamine in 500 mL of o-xylene (freshly prepared at 75 °C). The reaction vessel was closed with a watch glass and the reaction mixture stirred at a temperature of 75 °C. After 7 h, the chips were removed from the reaction mixture, immersed in fresh CH3CN, and sonicated for 15 min (3 times). Subsequently, the chips were incubated with 2% (v/v) 2-amino-6-methylheptane in acetonitrile (overnight, at room temperature). Finally the chips were sonicated each twice in acetonitrile and methanol and dried in a vacuum. Arraying. A GMS 418 was used for the production of the microarrays. The dimensions of the printed area was 1 cm2. Sets of operon 70-mer oligonucleotides (Qiagen) were used for the preparation of the chips. The number of printed capture elements and intrachip replicates (4, 10, or 30) varied depending on the experiments; please see in the respective sections. Printing was carried out at 26 °C, and humidity was adjusted to 70-80% relative humidity. Evanescent Resonator Microchips. The transducers (dimensions 18 × 18 mm2) were manufactured by Unaxis AG, Liechtenstein, according to specifications described elsewhere.18 Glass substrates were nanostructured in order to create a fully corrugated surface (grating period 360 nm). Subsequently, the corrugated surface was coated with Ta2O5 as dielectric material (150-nm thickness, refractive index material n ) 2.1). Fluorescence Snapshot Setup. The experimental setup is shown schematically in Figure 1. A HeNe laser (632.8 nm, 5 mW, Melles Griot, Zevenaar, The Netherlands), mounted onto a goniometer (Newport MM3000 controller, goniometer UBG120, Newport Corp., CA), was used as excitation light source. The transducer lies virtually in the center of rotation of the goniometer, allowing adjustment of the incident light without changing the position of the light beam on the chip. The laser beam was expanded by a Newport beam expander 30× (Newport). To achieve the resonance condition, the coupling angle was optimized by means of the goniometer (minimum for (18) Neuscha¨fer, D.; et al. Biosens. Bioelectron., in preparation.

Figure 1. “Snapshot” CCD setup used for demonstration of the fluorescence enhancement of evanescent resonator microchips (scheme). The ER chips used in the present study showed abnormal reflection (resonance) at 2° with respect to the transducer normal. By changing between normal and 2 °C incidence, off-resonance (epifluorescence) and resonance were compared.

the transmitted light indicates abnormal reflection/evanescent resonance). The resonance condition was adjusted prior to each experiment. For the experiments described below, chips with a resonance angle of Θres ) 2° were used. The resonance angle depends on the refractive indices of substrate, superstrate, and metal oxide, thickness of the Ta2O5 layer, period of the incoupling grating (360 nm), and wavelength of the HeNe laser (632.8 nm). An Astrocam CCD slow-scan camera was used as photodetector (LSR Astrocam, Cambridge, U.K.). The camera head carried a Kodak KAF1600 chip with 16-bit dynamic range operated at a temperature of -30 °C (Peltier cooling). A Nikon Noct 58 mm/ 1.2 lens was used as the collection optics (Nikon Corp., Tokyo, Japan). The aperture was adjusted to 4, to increase depth of focus. The polarization of the excitation light was parallel to the grating grooves. Fluorescence was separated from excitation light by means of Omega interference filters (Omega 680RDF40, Omega, Brattleboro, VT) centered at the maximum of the Cy5 emission at 680 nm with 40-nm half-width value. The exposure time was 10 s/image for all measurements, and a shutter automatically blocked the laser beam in the dead time between the image acquisitions to reduce photobleaching. Fluorescence Scanner. For the experiments, a commercially available, unmodified Genetic Micros System GMS 418 fluorescence scanner was used. Typically, the laser power was adjusted to 100%, and the gain was adjusted to 60%. Fluidics. A GeneTAC station from Genomic Solutions, Inc. was used for the hybridization of the microarrays. Since the system was originally designed for 1 × 3 in. slides, we have constructed smaller flow cells that were compatible with the 18 × 18 mm2 evanescent resonator format used for the presented experiments. The microarrays were washed prior to each hybridization. The process steps prewash and hybridization/postwash are described below. (1) Prewash. Three cycles with WB, each 10-s flow at 75 °C and an additional 10-s wash with HB at 23 °C was used. (2) Hybridization/Postwash. The temperature was ramped to 42 °C, and a 30-µL sample in HB was introduced into the flow chambers into the liquid phase. The temperature was then increased to 75 °C (10 min) to denature any hybridized species.

Figure 2. Coupling/decoupling of a bundle of photons into/from the corrugated layer (scheme). The excitation energy is confined in the metal oxide resonator and generates strong evanescent fields at its surface.

Subsequently, the temperature was adjusted to 42 °C for 16 h. After hybridization, the chips were washed with three cycles each 12-s flow of WB at 42 °C and an additional cycle at 23 °C. The chips were removed from the system, flushed with deionized water, and dried in a Nitrogen flow. Image Analysis. The fluorescence images were analyzed by means of Imagene 4.1 (Biodiscovery, Inc.). RESULTS Chip Geometry and Evanescent Resonator Principle. The transducer consists of a glass substrate with a uniformly corrugated surface covered with a ∼150-nm-thick Ta2O5 layer as a high refractive index material. The period of the grating was 360 nm. The resonance angle Θres is defined by

sin Θres ) ne - mλ/Λ where m is the diffraction order (m ) (1), λ is the wavelength of the excitation light, and Λ is the grating period. ne corresponds to the effective index of refraction and is a function of wavelength and polarization of incident light, thickness of the layer, and refractive indices of the layers. Details on the structure and its preparation, as well as physical characterization of the used chips, are published elsewhere.18 The novel approach takes advantage of a phenomenon attributed as abnormal reflection. As illustrated schematically in Figure 2, a bundle of photons incident on the chip under resonance conditions couples into the corrugated metal oxide surface at the site of incidence and creates locally an evanescent field. Due to the high loss factors of the structure (2000/cm), decoupling of light takes place within the micrometer range. Depending on the optical setup used, the excitation beam has typically at least several micrometer diameters and the coupling/ uncoupling phenomenon takes place simultaneously in the illuminated region of the transducer. Therefore, the abnormal reflection is the result of a constructive interference of all reflected beams R and beams D- diffracted into the cover medium. In contrast, the light intensity in transmission direction is almost completely extinguished by destructive interference (beams T and D+). Analytical Chemistry, Vol. 75, No. 11, June 1, 2003

2573

Figure 3. Transmission through an ER chip as a function of the incident angle of TE polarized laser light (helium-neon laser, 633 nm), Resonance occurs at (2° resulting in “W”-type characteristics.

Figure 4. Fluorescence of an evanescent resonator microarray after hybridization of Cy5-labeled RNA sample from rat liver. Left: normal incidence of the excitation light (off-resonance, corresponds to epifluorescence); the fluorescence signals are almost not visible. Right: angle of incident excitation light tuned to evanescent resonance (=2° with respect to surface normal); the fluorescence signals were found to be ∼2 orders of magnitude increased. The gene expression pattern is visible under resonance conditions.

As a result of this transducer geometry, the light incident on the transducer under resonance condition (Θ ) Θres) is locally confined into the thin corrugated layer of high refractive index material. This results in strong electromagnetic fields within the metal oxide layer and strong evanescent fields at the surface of the chip. The effect has been attributed as evanescent resonance. The evanescent fields increase the fluorescence intensity of chromophores attached to the surface, since the effective field strength is increased ∼100-fold by the confinement of the available excitation energy. The uniformly corrugated ER transducers have broad resonance curves (full width at half-maximum, fwhm, ∼2°) centered close to normal incidence and are, therefore, easily adaptable to, for example, fluorescence scanners and microscopes. Figure 3 shows the relative transmitted energy as a function of the angle of incidence. The ER chip shows minimum transmission at (2° (resonance) for transversal electric (TE) polarized HeNe laser light. Due to the broad fwhm value, both resonances overlap resulting in “W”-type characteristics. Fluorescence “Snapshot” Images of Oligonucleotide Microarrays. For the experiments, a setup equipped with a chargecoupled device (CCD) camera as fluorescence detector and an expanded helium-neon laser beam as excitation source, as described in Figure 1, was used. The angle of the light incident on the chip with respect to the chip normal was adjustable by means of a goniometer in order to “switch” between conventional epifluorescence illumination and ER. The main advantages of the CCD setup are that the expanded laser beam (Gauss profile) provides parallel excitation light, the angle of incidence can be adjusted, and microarray images can be obtained within a few seconds. ER chips with approximately 60-70-mer oligonucleotide capture elements (each with 30 replicates) representing rat genes capture elements were prepared. The size of the arrays was ∼1 cm2. Cy5-labeled total RNA samples from rat liver were hybridized to the chip, and an image was measured under conventional epifluorescence conditions (normal incidence of the excitation light). The obtained fluorescence image is shown on the left side of Figure 4. Subsequently, the angle of incidence was tuned to the evanescent resonance position and a second image was taken (angle of incidence 2° with respect to surface normal). The second image is shown on the right part of Figure 4.

Both images were scaled to the same intensity range in order to allow visual comparison. The images obtained under resonance conditions were significantly more intense as compared to the images obtained under epifluorescence (off-resonance) conditions. Signal variations in the fluorescence image can be explained by slight imperfections of the expanded laser beam (corners of the image) and by the spotting procedure. Under conventional conditions (off-resonance, epifluorescence), the microarray pattern is almost not visible. The mean fluorescence value of the off-resonance images (two replicates) was found to be six counts, and the mean standard deviation of the background was found to four counts (governed by camera electronics). Under resonance conditions, the mean value of all spots was increased to 170 counts. In contrast, fluorescence background and background standard deviation were only slightly increased (50% range). The data obtained with the CCD setup are summarized in Table 1 (upper part). The exact factor for the fluorescence enhancement depends strongly on the accurate quantification of the weak intensities obtained under epifluorescence conditions. To calculate a factor (ratio) for the fluorescence enhancement, spots that were 3-fold above standard deviation under offresonance conditions and the corresponding spots from the resonance images were used. The mean ratio of the fluorescence intensities resonance/off-resonance of chips measured in the CCD setup was found to be 52. In addition, reference glass chips processed under identical conditions produced fluorescence signals comparable to ER chips under off-resonance geometry (almost not visible, not shown). Further results on the evanescent resonance with above-mentioned CCD setup have been published elsewhere.18 Under the chosen experimental conditions (short exposure, read noise > shot noise), the ratios’ net signals/background standard deviations changed from ∼1 for the off-resonance images to ∼30 for the resonance images, which corresponds to a significant improvement of the signal quality. Longer exposure times will increase the fluorescence signals proportionally; however, the image quality will not automatically further be improved, since the shot noise contribution will increase proportional to the exposure time (shot noise > read noise). In addition, in the case of larger signals (shot noise > read noise), it is expected that the signal standard deviation will increase, especially in the case of

2574 Analytical Chemistry, Vol. 75, No. 11, June 1, 2003

Table 1. Mean Fluorescence Intensities, Mean Background Values, and Mean Background Standard Deviations of Identical Processed Microarrays Based on Glass and Evanescent Resonator Chips Obtained with Custom CCD Setup (Figure 1) and a Commercial Scanner (GMS 417) CCD Setup resonance

mean (counts) background (counts) standard deviation (counts)

off-resonance

ER chip 1

ER chip 2

mean

ER chip 1

ER chip 2

mean

150.4 43 5.1

189.4 43 5.3

170 43 5.2

5.7 29 4.3

5.9 28 4.1

6 28 4.2

Fluorescence Scanner resonance

mean (counts) background (counts) standard deviation (counts)

glass

ER chip 1

ER chip 2

mean

glass 1

glass 2

mean

6054 67 81.7

7882 65 79.7

6968 66 80.7

924