Anal. Chem. 2001, 73, 2412-2420
Evaluation of Three-Dimensional Microchannel Glass Biochips for Multiplexed Nucleic Acid Fluorescence Hybridization Assays Vincent Benoit,† Adam Steel,‡ Matt Torres,‡ Yong-Yi Yu,‡ Hongjun Yang,‡ and Jonathan Cooper*,†
Gene Logic, Inc., 708 Quince Orchard Road, Gaithersburg, Maryland 20878, and Bioelectronics Research Centre, Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8LT, Scotland, UK
Three-dimensional, flow-through microchannel glass substrates have a potential for enhanced performance, including increased sensitivity and dynamic range, over traditional planar substrates used in medium-density microarray platforms. This paper presents a methodology for the implementation of multiplexed nucleic acid hybridization fluorescence assays on microchannel glass substrates. Fluorescence detection was achieved, in a first instance, using conventional low-magnification microscope objective lenses, as imaging optics whose depthof-field characteristics match the thickness of the microchannel glass chip. The optical properties of microchannel glass were shown, through experimental results and simulations, to be compatible with the quantitative detection of heterogeneous hybridization events taking place along the microchannel sidewalls, with detection limits for oligonucleotide targets in the low-attomole range. In the analytical sciences, microarray technology involves the miniaturization and pooling of biorecognition sites (antibodies, nucleic acids, etc.) on a single substrate, to allow heterogeneous bioassays to be performed in a highly parallel fashion, thereby dramatically increasing throughput.1-3 Over the past decade, numerous applications of life science research and pharmaceutical development, such as diagnostics and molecular medicine, have been shown to greatly benefit from being implemented in a microarray format.4-6 Most microarrays developed to date, whether used in polymorphism analysis,7,8 differential gene9-13 or protein14 †
University of Glasgow. Gene Logic, Inc. (1) O’Donnell-Malonay, M. J.; Little, D. P. Gen. Anal. 1996, 13, 151-157. (2) Wang, J. Nucleic Acids Res. 2000, 28 (16), 3011-3016. (3) Hoheisel, J. D. TIBTECH 1997, 15, 465-469. (4) De Benedetti, V. M. G.; Biglia, N.; Sismondi, P.; De Bortoli, M. Int. J. Biol. Markers 2000, 15, 1-9. (5) Lemieux, B.; Aharoni, A.; Schena, M. Mol. Breeding 1998, 4, 277-289. (6) Stratowa, C.; Wilgenbus, K. K. Curr. Opin. Mol. Therap. 1999, 1 (6), 671679. (7) Zhang, Y.; Coyne, M. Y.; Will, S. G.; Levenson, C. H.; Kawasaki, E. S. Nucleic Acids Res. 1991, 19 (14), 3929-3933. (8) Gunderson, K. L.; Huang, X. C.; Morris, M. S.; Lipschutz, R. J.; Lockhart, D. J.; Chee, M. S. Genome Res. 1998, 8, 1142-1153. (9) Bowtell, D. D. L. Nat. Genet. 1999, 21, 25-29. (10) Schena, M.; Shalon, D.; Heller, R.; Chai, A.; Brown, P. O.; Davis, R. W. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 10614-10619. (11) Granjeaud, S.; Bertucci, F.; Jordan, B. R. BioEssays 1999, 21, 781-790. ‡
2412 Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
expression, or sequencing by hybridization,15 are based on a twodimensional (2D) format, in which a suitable set of biorecognition elements are arrayed on the surface of a planar, unpenetrable substrate.16-20 Although this approach does offer high levels of multiplexing, the density of spots is ultimately limited by either the dispensing mechanism used in the arraying scheme or the amount of biological material that has to be present within each spot, to provide appropriate levels of sensitivity, dynamic range, and speed. In an effort to further increase the analytical performance of microarrays, a novel concept has been devised, based on the use of three-dimensional, uniformly porous substrates featuring a regular array of discrete, ordered, high aspect ratio (>50:1) microchannels.21 Biorecognition elements, henceforth referred to as probes, are immobilized on the sidewalls of the microchannels, and biorecognition events take place while the sample flows “through” the chip. In the current implementation presented in this paper, each spot typically encompasses a few hundred microchannels. Since the effective surface area available for probe immobilization is increased by expansion into the third dimension, a higher density of probes can be achieved as compared to a planar substrate with the same lateral dimensions. In addition, each microchannel acts as a miniature analytical chamber, in which mass-transfer kinetics are enhanced by spatial confinement. As a result, flow-through microchannel substrates provide significant advantages over planar substrates, such as (12) Lashkari, D. A.; DeRisi, J. L.; McCusker, J. H.; Namath, A. F.; Gentile, C.; Hwang, S. Y.; Brown, P. O.; Davis, R. W. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 13057-13062. (13) Duggan, D. J.; Bittner, M.; Chen, Y.; Meltzer, P.; Trent, J. Nat. Genet. 1999, 21, 10-14. (14) De Wildt, R. M. T.; Mundy, C. R.; Gorick, B. D.; Tomlinson, I. M. Nat. Biotechnol. 2000, 18, 989-994. (15) Drmanac, S.; Kita, D.; Labat, I.; Hauser, B.; Schmidt, C.; Burczak, J. D.; Drmanac, R. Nat. Biotechnol. 1998, 16, 54-58. (16) Lamture, J. B.; Beattie, K. L.; Burke, B. E.; Eggers, M. O.; Ehrlich, D. J.; Fowler, R.; Hollis, M. A.; Kosicki, B. B.; Reich, R. K.; Smith, S. R.; Varma, R. S.; Hogan, M. E. Nucleic Acids Res. 1994, 22 (11), 2121-2125. (17) Lipshutz, R. J.; Fodor, S. P. A.; Gingeras, T. R.; Lockhart, D. J. Nat. Genet. 1999, 21, 20-24. (18) Vo-Dinh, T.; Alarie, J. P.; Isola, N.; Landis, D.; Wintenberg, A. L.; Ericson, M. N. Anal. Chem. 1999, 71, 358-363. (19) Caillat, P.; David, D.; Belleville, M.; Clerc, F.; Massit, C.; Revol-Cavalier, F.; Peltie´, P.; Livache, T.; Bidan, G.; Roget, A.; Crapez, E. Sens. Actuators, B 1999, 61, 154-162. (20) Heller, M. J.; Forster, A. H.; Tu, E. Electrophoresis 2000, 21, 157-164. (21) Steel, A.; Torres, M.; Hartwell, J.; Yu, Y. Y.; Ting, N.; Hoke, G.; Yang, H. In Microarray Biochip Technology; Schena, M., Ed.; Eton: Natick, MA, 2000; pp 87-117. 10.1021/ac000946r CCC: $20.00
© 2001 American Chemical Society Published on Web 04/21/2001
increased sensitivity and dynamic range, higher throughput, and shorter assay times.21 However, this potential for enhanced performance only translates into an improved bioanalytical device if the substrate is compatible with an appropriate detection scheme. Fluorescence is widely used for the detection of heterogeneous nucleic acid hybridization, either through direct prompt intensity measurements17-20,22,23 or in more elaborate schemes such as fluorescence resonance energy transfer (FRET)24-26 and “molecular beacons”.27-29 When performed in a multiplexed format on conventional 2D arrays, most of these assays rely on confocal scanning microscopy,30,31 in which high numerical aperture (NA) optics are used to maximize the fluorescence excitation and collection efficiencies and to provide increased lateral and axial resolutions, thereby resulting in enhanced signal-to-noise characteristics. In the case of 3D microarrays, fluorescence must be excited and collected over a large axial distance commensurate with the chip thickness, to take full advantage of the inherently higher levels of sensitivity and dynamic range provided by the 3D format and to ensure linearity of the measurements. In this paper, we demonstrate that 3D microchannel glass (MCG) substrates32 are compatible with multiplexed fluorescence oligonucleotide hybridization assays. Appropriate levels of analytical performance were achieved through an epifluorescence imaging scheme, in which a low-magnification image of the array was formed by high depth-of-field optics onto the surface of a twodimensional photon detector. Although a larger depth of field necessitates tradeoffs, including reduced lateral resolution and lower NA, this is compensated by the remarkable waveguiding properties of MCG, as well as by the use of a highly sensitive photon detector, namely, a cooled CCD camera. Since the use of MCG chips of increased thickness provides a way of increasing the number of probe molecules immobilized within each spot without increasing the lateral dimensions of the spots, a range of chip thicknesses were considered in this paper, as a first step toward the optimization of the substrate geometry. Potential limitations to the benefits associated with an increase in chip thickness are discussed, in terms of the effectiveness of both the fluorescence detection scheme and the process of probe immobilization on the microchannel sidewalls. In this respect, the physical-chemical processes that control the formation of spots on microchannel substrates, i.e., the wetting properties of the substrates, as well as the main parameters involved in the arraying process (concentration of the probe solution and volume of the aliquots dispensed), have been investigated. (22) Guo, Z.; Guilfoyle, R. A.; Thiel, A. J.; Wang, R.; Smith, L. M. Nucleic Acids Res. 1994, 22 (24), 5456-5465. (23) Kurg, A.; Tonisson, N.; Georgiou, I.; Shumaker, J.; Tollett, J.; Metspalu, A. Genet. Test. 2000, 4 (1), 1-7. (24) Cardullo, R. A.; Agrawal, S.; Flores, C.; Zamecnik, P. C.; Wolf, D. E. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 8790-4. (25) Bernard, P. S.; Wittwer, C. T. Clin. Chem. 2000, 46 (2), 147-148. (26) Parkhurst, K. M.; Parkhurst, L. J. Biochemistry 1995, 34, 285-292. (27) Tyagi, S.; Bratu, D. P.; Kramer, F. R. Nat. Biotechnol. 1998, 16, 49-53. (28) Brown, L. J.; Cummins, J.; Hamilton, A.; Brown, T. Chem. Commun. 2000, 621-622. (29) Liu, X.; Tan, W. Anal. Chem. 1999, 71, 5054-5059. (30) Cheung, V.; Morley, M.; Aguilar, F.; Massimi, A.; Kucherlapti, R.; Childs, G. Nat. Genet. 1999, 21, 15-19. (31) Fodor, S. P. A.; Rava, R. P.; Huang, X. C.; Pease, A. C.; Holmes, C. P.; Adams, C. L. Nature 1993, 364, 555-556.
EXPERIMENTAL SECTION Materials. Substrates and Reagents. The use of microchannel glass as a substrate for biochips has been described previously.21 The fabrication process for MCG is identical to that for nanochannel array glass, as described by Tonucci et al.32 A cylindrical boule of MCG (Galileo Electro-Optics Corp., Sturbridge, MA) was wafered to produce 0.5-, 1-, and 2-mm-thick chips with a channel diameter of 10 µm and an open fraction of 0.55. Care was taken to ensure that the chips were diced so that the front and back planes were perpendicular to the sidewalls of the microchannels. The surfaces of the chips were not polished after dicing. Standard float glass microscope slides (BDH, Poole, U.K.) were used as 2D substrates, for comparative purposes. (Mercaptopropyl)trimethoxysilane (MPTS), toluene (99.8%, HPLC grade), absolute ethanol, 20× sodium saline citrate (SSC) buffer (0.3 M sodium citrate and 3 M NaCl), and 20× saline sodium phosphate/EDTA (SSPE) buffer (0.2 M phosphate buffer, 2.98 M NaCl, and 0.02 M Na-EDTA) were from Sigma (St. Louis, MO). The heterobifunctional cross-linker sulfo-γ-maleimidobutyryloxysuccinimide (s-GMBS) was purchased from Pierce (Rockford, IL). The fluorescently labeled oligodeoxynucleotide (ODN) probe, P0, used in microarraying studies was an 18-mer of sequence “F”CCCAGGGAGACCAAAAGC“T”, where “T” denotes a 3′-aminohexyldeoxythymidylate modifier and “F” a 5′carboxyfluorescein phosphoramidite (6-FAM) label. Hybridization experiments involved the use of unlabeled probes and fluorescently labeled ODN targets. The sequences for the probes P1, P2, and P3 and the targets T1, T2, and T3 were “T”CCGCTGCCGCTGTCA, “T”CCCTGGTATGAGCCCATCTATC, “T”CACCAGGATGCTCACATTTAAGTT, “F”TGACAGCGGCAGCGG, “F”AAGATAGATGGGCTCATACCA, and “F”AAACTTAAATGTGAGCATCCTGGTG, respectively, with “T” and “F” as above. The complementary pairs P1-T1, P2-T2, and P3-T3 are representative of the genes C-myc, TNF-R, and IL2, respectively. All the ODNs were purchased as a lyophilized powder from Glen Research (Sterling, VA) and subsequently dissolved to appropriate concentrations in 1× SSC (probes) or 5× SSPE (targets). The blocking agent was an aqueous solution of 0.02% poly(vinylpyrrolidone) (PVP) and 0.02% Ficoll. Tetramethylrhodamine-5-maleimide (mal-TAMRA) was purchased from Molecular Probes (Eugene, OR). Instrumentation. Spotting of MCG and flat glass chips was performed using a Packard BioChip piezoelectric spotter (Packard Instrument Co., Meriden, CT). Fluorescence imaging was carried out using either an Eclipse E800 epifluorescence microscope (Nikon, Tokyo, Japan), fitted with an Orca C4742-95 color, 12bit CCD camera (Hamamatsu, Hamamatsu City, Japan), or a Microphot microscope (Nikon) with epifluorescence attachment, coupled to an SV10K monochrome, 16-bit CCD camera (PixelVision, Tigard, OR). The fluorescence filter sets, from Nikon, were DM510-B2A for FAM and DM580-G2A for TAMRA. Objective lenses, also from Nikon, were 10× and 4× Plan Apo, 2× and 1× Plan UW, and 4× Plan DL, with NA values of 0.45, 0.20, 0.06, 0.04, and 0.13, respectively. A specially designed flow cell21 was used for the convective pumping of sample and washing solutions through MCG chips. (32) Tonucci, R. J.; Justus, B. L.; Campillo, A. J.; Ford, C. E. Science 1992, 258, 783-785.
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2413
Figure 1. Epifluorescence imaging procedure for the investigation of the optical properties of microchannel glass chips, using a 2D fluorescent (TAMRA) spot, ∼260 µm in diameter, immobilized on the surface of a silanized flat glass piece. (A) In-focus imaging of the 2D spot; (B) imaging of the 2D spot through the chip; (C) imaging of the front surface of the chip, with the 2D spot underneath; (D) imaging of the 2D spot with a level of defocusing equal to the chip thickness. The dimensions of the 2D spot and the microchannels have been exagerated for clarity.
Procedures. Preparation of MCG and Flat Glass Chips for Optical and Microarraying Studies. Covalent attachment of aminoterminated ODNs to MCG and flat glass chips was achieved as described previously.21 Briefly, the chips were cleaned by degreasing in organic solvents, surface activated by acidic treatment, and silanized by immersion in 2.5% MPTS in toluene followed by rinsing with EtOH and overnight baking at 80 °C. Prior to spotting, ODN probes were incubated for 1 h at room temperature with s-GMBS in 1× SSC, at a 10:1 cross-linker-to-probe ratio and at probe concentrations ranging from 5 to 71 µM. The volume dispensed for each spot ranged from 2 to 20 nL. Flat chips were also spotted after silanization with mal-TAMRA (2 nL/spot). Washing of noncovalently bound probes and blocking of subsequent nonspecific adsorption were achieved by flowing 10 mL of the PVP and Ficoll solution through the MCG chips, and over the flat chips, at 0.5 mL min-1. Imaging of Spotted Chips and Image Analysis. Epifluorescence and bright-field microscopy images of spotted chips were acquired after evaporation of the spotting buffer, prior to washing. Chips were also imaged after washing, immersed in aqueous solution, through the optical window of the flow cell. All the images of spotted MCG chips were acquired with the focus set on the front surface. Image analysis involved attributing each spot with both a disk, covering the entire surface of the spot, and a ring, surrounding the disk, outside the spot area. The net fluorescence signal for each spot, S, was determined by subtracting the average pixel intensity of the ring from that of the disk. The background noise for the whole array, N, was determined as the variance in the intensities of the rings. Optical Properties of MCG Chips. The optical properties of MCG chips relevant to the fluorescence imaging process, namely, transparency, waveguiding, and interchannel cross-talk, were investigated by means of the epifluorescence imaging procedure schematically presented in Figure 1. In this procedure, nonspotted MCG chips of various thicknesses were used in combination with a 2D fluorescent spot. The latter was a spot of mal-TAMRA immobilized on the surface of a silanized flat glass piece. The diameter of the 2D spot was ∼260 µm, covering an area that would encompass 370 individual microchannels on a MCG chip. All the measurements were carried out using a 4×, 0.13 NA microscope objective lens. 2414
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
The procedure depicted in Figure 1 was carried out manually, for each chip, during the data acquisition process, as follows. First, a focused image of the reference spot was acquired (image A). The MCG chip of interest was subsequently placed on top of the reference spot, in the optical path of the lens, and an image of the reference spot was acquired (image B). The sample was then moved away from the lens by a distance equal to the chip thickness, so that the top surface of the chip was now in the focal plane, and image C was acquired. Finally, the chip was removed and an image of the reference spot was acquired (image D), with a negative amount of defocusing equal to the chip thickness. Quantitative data were obtained by determining the signal, S, as the area under the peak in the radial intensity profile, after twopoint baselining of the curve. Simulation of the Optical Properties of MCG. Investigations into the waveguiding properties and the interchannel optical crosstalk characteristics of MCG were carried out using OptiCAD software (OptiCAD Corp., Santa Fe, NM). OptiCAD is an unconstrained ray-tracing program that allows the layout and analysis of three-dimensional imaging and nonimaging optical systems. For the purpose of the present studies, fluorescence emission from individual fluorescent labels was simulated by defining each fluorophore as a self-luminous, incoherent point source to which an isotropic emission distribution was associated. A 3D cutout section of a MCG chip, comprising 19 channels arranged in an hexagonal packing pattern, was defined as an OptiCAD object. The refractive index (RI) of the MCG material was set equal to 1.7. The RI of the immersion medium (internal volume of the microchannels, plus layers of fluid on each side of the chip) was set equal to either 1.0 (air) or 1.333 (aqueous solution). The bulk absorbance of the MCG substrate was assumed to be negligible. Volume scattering and autofluorescence of the material were not considered in the simulations. As a model to assess interchannel optical cross-talk, a single light source object simulating a fluorophore, as described above, was placed inside the central channel, at various depths inside the channel, and at a fixed distance from the sidewall. The radiant energy being emitted by the fluorophore and propagating within the MCG structure was mapped by recording the spatial distribution of energy impingent on a set of virtual radiometer film planes placed perpendicular to the channel axis, at different distances between the front and back faces of the chip. Typically, a minimum of 10 000 rays was traced. Hybridization Experiments. MCG chips, 0.5 mm thick, 10 µm channel diameter, were used in hybridization experiments. Each chip was spotted to feature 16 replicates of a 4 × 4 array, the latter comprising spots of the four probes P0, P1, P2, and P3, each in quadruplicate. Each spot was produced by the dispensing of a 5-nL aliquot of a 30 µM probe solution onto the substrate. The spacing between adjacent spots was 400 µm. Prior to use in flowthrough hybridization experiments, the chips were immersed in 0.2 mL of blocking agent for 10 min and then blow-dried and baked at 80 °C for 1 h. The hybridization sample consisted of 50 µL of a mixture of the targets T1, T2, and T3 at 2, 20, and 10 nM concentrations, respectively, in 5× SSPE buffer. The 50-µL sample volume was recirculated through the MCG chip under a constant flow rate of 0.5 mL min-1, for 3 h, at room temperature. Pre- and posthybrid-
Figure 2. Scanning electron microscope images of the front surface of a microchannel glass chip (Galileo Electro-Optics Corp.). A ∼10nm layer of Au-Pd was evaporated onto the substrate in order to prevent charging during SEM imaging.
ization washing steps were carried out by flowing 5× SSPE through the chip at 0.5 mL min-1 for 15 min. RESULTS AND DISCUSSION Characterization of Microchannel Glass. The morphology of MCG chips was characterized by scanning electron microscopy (SEM). Figure 2 shows SEM images obtained using a Hitachi S-800 scanning electron microscope. The regularity of the hexagonal packing, as well as the high uniformity in the shape and dimension of the microchannels, is clearly seen. The microchannel matrix is made of a clear potash lead glass containing 55% lead oxide and 35% silicon dioxide. The high lead oxide content endows the glass with a density of 4.44 g cm-3, providing mechanical strength to the structure. In terms of optical properties, this type of glass is characterized by a low optical absorption coefficient in the visible and a high RI of ∼1.7 at 500 nm. Factors Affecting Fluorescence Imaging in Microchannel Biochips. In microchannel biochips, biorecognition sites are located throughout the chip thickness, and thus fluorescence has to be detected over a correspondingly extended volume. This can be achieved by sacrificing lateral resolution for improved depth of field. When analysis of a microarray-based multiplexed bioassay is implemented as an imaging scheme, the sole requirement in terms of lateral resolution is an ability to distinguish between neighboring individual spots, to provide, for each spot, a relative fluorescence intensity value which is quantitatively correlated to the extent of biorecognition taking place within the spot of interest. In the case of microarrays manufactured by postsynthesis attachment of the probes, the performance of the spotting device eventually limits the minimum achievable spot diameter to a few tens of micrometers. These dimensions lie within the spatial resolution provided by even low-magnification objective lenses, thereby providing the possibility to trade lateral resolution for a greater depth of field. In MCG biochips, although fluorescence originates from inside individual microchannels, radiations are only partially confined within each channel, due to the high level of transparency of the glass matrix at the wavelengths of interest. As a result, interchannel optical cross-talk is expected to take place, which can lead to the fluorescence image of a spot to be significantly different from its physical shape. This has to be taken
Figure 3. Radial intensity profiles associated with the epifluorescence images of a 2D TAMRA spot, imaged at different levels of defocusing, with a 4×, 0.13 NA microscope objective lens. Each curve was obtained by averaging four adjacent pixel rows.
into account in both the chip arraying and the image analysis methods, to allow for unbiased quantitative data treatment of fluorescence images. Imaging Properties of Microscope Objective Lenses in Three Dimensions. By design, the ability of a microscope objective lens to produce a sufficiently sharp image is restricted to a short distance on each side of the focal plane, in the object space, referred to as the depth of field of the lens. In a nonconfocal epifluorescence imaging scheme, images formed on the detector include contributions from fluorophores and scattering sites located within a whole range of optical planes. The weight of the contribution from each optical plane is determined by the imaging properties of the lens in that plane. If the object is thicker than, or lies out of, the depth of field of the lens, significant blurring of the image will occur. Figure 3 illustrates how the imaging properties of a microscope objective lens are affected by the position of the object (in this case, a 2D fluorescent spot) relative to the focal plane. As the level of defocusing increases, the edges of the spot become blurred, resulting in the overall diameter of the spot image to increase. The overall intensity of the spot image also decreases. In the case of the 4×, 0.13 NA lens used in the experiment of Figure 3, both the increase in diameter and the decrease in intensity were found to be roughly linear with the level of defocusing. At a defocusing distance of 0.5 mm, the diameter and brightness of the spot image were 1.15 and 0.75 times that of the focused image, respectively. Optical Properties of MCG Chips. Multiple information can be inferred from the procedure represented in Figure 1. Comparison of images A and B indicates how the imaging properties of the lens are altered by the presence of the chip in the optical path; this characterizes the level of transparency of the chip. Comparison of images A and C reflects the waveguiding efficiency of the chip, combined with the imaging properties of the lens for a defocusing distance equal to the chip thickness, while comparison of images C and D gives direct insight into the waveguiding properties of the chip. Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2415
Figure 4. (a-d) Epifluorescence images A-D associated with the procedure described in Figure 1, for a 0.5-mm-thick microchannel glass chip. (e) Image C, for the same chip, acquired with a higher magnification lens. The 2D fluorescent spot is ∼260 µm in diameter.
Figure 6. Characteristic ratios S(C)/S(A) (circles), S(C)/S(D) (squares), and S(B)/S(A) (diamonds), as a function of the microchannel glass chip thickness. Average and 95% confidence intervals for three independent measurements. See text for details. Figure 5. Radial intensity profiles associated with the images of Figure 4a-d. Each curve was obtained by averaging four adjacent pixel rows.
Panels a-d of Figure 4 show images A-D, as described in Figure 1, for a 0.5-mm-thick, 10-µm-channel diameter MCG chip. In all the images, the 2D spot is clearly seen, with various levels of sharpness. Variations in the levels of signal and background with chip thickness are also evidenced. As seen in Figure 4e, imaging at higher magnification showed that individual microchannels appear brighter than the interstitial surface (i.e., the surface of the glass matrix, located on the front face of the chip, perpendicular to the sidewalls of the channels). The fact that the 2D spot is not uniform, as seen in Figure 4a and discussed later, provides a means to check that the extent of radiation scrambling through the MCG chip is limited. Figure 5 shows the radial intensity profiles of the images shown in Figure 4. Each curve was obtained by averaging the four median columns of pixels in each spot image. Figure 5 also illustrates the relative levels of signal and background associated with images A-D. Curve A is representative of the focused image of the 2D spot and shows the nonuniformity of the latter. Curve D is associated with the image of the 2D spot at an amount of negative defocusing equal to the chip thickness and shows the associated blurring of the spot, as discussed above. Curve B is representative of the image of the 2D spot as seen through the chip and shows an attenuation in the net fluorescence signal and an increase in 2416 Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
the level of background, both caused by the MCG structure. Finally, curve C is associated with the image of the 2D spot with the focus adjusted on the top surface of the chip. As compared to curve B, curve C is characterized by similar levels of signal and background, which demonstrates the compatibility between the chip thickness and the imaging properties of the lens. Figure 6 shows the effect of the chip thickness on the characteristic ratios S(C)/S(A), S(C)/S(D), and S(B)/S(A), where S is the net fluorescence signal and A-D are as defined in Figure 1. As expected, the fraction of net fluorescence signal collected by adjusting the focus on the front surface of the chip, characterized by the ratio S(C)/S(A), decreases as the chip thickness increases. This results from a combination of two factors: attenuation of the fluorescence signal caused by the presence of the substrate in the optical path, and loss in the excitation and collection efficiencies of the lens induced by the defocusing. The contribution from the latter factor is evidenced by comparison with S(B)/S(A), in which no defocusing is involved. Importantly, S(C)/S(A) and S(B)/S(A) are not statistically different in the case of a 0.5-mm-thick MCG chip, suggesting that no significant attenuation of the signal results from defocusing itself. Signal attenuation by the substrate is due, for the major part, to reflection and scattering of light on the front surface of the chip and, thus, is not proportional to the chip thickness. The ratio S(C)/S(D) characterizes the waveguiding efficiency of the chip, since it represents the relative intensity of fluorescence that is detected
Figure 7. Bright-field (a) and epifluorescence (c) microscopy images, and associated radial intensity profiles (b), of spots formed by the dispensing of 5-nL aliquots from a 30 µM solution of fluorescently labeled probe in 1× SSC, onto a flat glass slide (left) and a microchannel glass chip (right). The images were acquired using a 4×, 0.2 NA lens. Each profile curve was obtained by averaging four adjacent pixel rows.
through the chip with the focus on the front surface of the chip, as compared to that detected directly. For the range of chip thicknesses studied here, S(C)/S(D) is less than unity, which indicates that the signal originating from the 2D spot is actually attenuated by the presence of the chip in the optical path. However, as the chip thickness increases, the waveguiding ability of the chip progressively compensates for the defocusing-induced loss of detection efficiency of the objective lens. Generation of ODN Probe Spots on MCG Chips. Figure 7a shows bright-field microscopy images of spots formed by the dispensing of identical volumes of a probe solution onto flat glass and MCG. Figure 7c shows epifluorescence images of the same spots. In the case of the flat glass substrates, both the brightfield image and the fluorescence image show a poor level of uniformity, including the presence of buffer crystals having formed during evaporation of the spotting solution, and bright outer fluorescent rings. A number of publications concerned with the arraying of DNA probes onto planar, impenetrable substrates have reported such inhomogeneities30,33,34 and ascribed the effect to the transport of material from the center to the edges of the deposited droplets upon evaporation. By comparison, the spots on the MCG substrate appear to be much more uniform. In particular, no outer rings can be seen in the fluorescence image. In addition, the spots formed on MCG exhibit a smaller footprint, with a diameter of ∼40% that measured on flat glass. This illustrates the particular wetting properties of the microchannel substrates. When a droplet of liquid, released from the capillary tip of the piezoelectric spotter, impinges onto the surface of a MCG chip, the liquid undergoes lateral and axial redistribution, through the combined effects of spreading over a certain area of the front surface of the substrate and drawing into the microchannels by (33) Graves, D. J.; Su, H. J.; McKenzie, S. E.; Surrey, S.; Fortina, P. Anal. Chem. 1998, 70, 5085-5092. (34) Yoon, S. H.; Choi, J. G.; Lee, S. Y. J. Microbiol. Biotechnol. 2000, 10 (1), 21-26.
capillary action. Since the latter takes place on a short time scale with respect to evaporation, probe and buffer molecules present in the dispensed liquid can be assumed to be uniformly distributed between all of the microchannels encompassed by the dispensed volume. Any redistribution of probe molecules occurring upon subsequent evaporation of the solvent is confined within the volume of each microchannel, rather than taking place over the whole spot area as in the case of flat substrates. As the wetting properties of MCG substrates result from a combination of surface properties (hydrophobicity) and capillary action, it can be expected that only part of the total dispensed volume will be drawn into the microchannels, the remaining fraction forming a convex meniscus on the surface of the chip. From the knowledge of the lateral dimensions of a spot generated by the delivery of a given volume of solution, the fractions of the dispensed volume that either fills the microchannels or forms a meniscus at the surface of the substrate can be estimated. The results of a calculation of the ratios Vdr/Vch, where Vdr is the volume of the aliquot of probe solution dispensed onto the microchannel substrate and Vch the internal microchannel volume encompassed by the resulting spot, are shown in Table 1, for MCG chips between 0.5 and 2 mm in thickness, spotted with aliquots of probe solution between 2 and 20 nL in volume. The internal microchannel volume encompassed by a spot was estimated from the footprint of the spot, determined by bright-field imaging, and the morphology of the microchannel substrate. The calculated ratios Vdr/Vch are seen to encompass a wide range of values above and below unity. The latter point arises due to the possibility of the dispensed volume being less than the total microchannel volume over the spot area. All the chips exhibit a clear trend, with the values for Vdr/Vch increasing as the dispensed volume increases. In the case of the 1- and 2-mm-thick chips, the dispensed volume is not enough to fill the entire volume of the microchannels, suggesting that probe attachment in the Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2417
Table 1. Calculated Ratios Vdr/Vcha dispensed volume (nl) chip thickness (mm)
1
2
3
5
10
20
0.5 1 2
0.54 nab na
0.70 na na
0.84 na na
1.03 0.75 0.53
1.08 0.94 0.64
1.26 0.99 0.82
aV dr is the volume of an aliquot from a 30 µM solution of oligodeoxynucleotide probe P0 in 1× SSC buffer, dispensed onto a microchannel glass substrate of given thickness, and Vch is the effective microchannel volume encompassed by the resulting spot. The microchannel volume encompassed by a spot was estimated from the footprint of the spot, determined by bright-field imaging, and the morphology of the microchannel substrate. b na, not available.
top part of the channels will be compromised. In the case of the 0.5-mm-thick chip, there is an excess liquid for dispensed volumes larger than 5 nL; these conditions are favorable, in terms of probe attachment, for coating the whole microchannel wall. A dispensed volume of 5 nL closely matches the volume of the microchannels, while lower volumes result in partially filled microchannels. Optical Cross-Talk in MCG Chips. The fluorescence images of spots on MCG, shown in Figure 7c, feature an outer ring of lower intensity. In effect, the fluorescence image of a spot on MCG appears wider than the actual physical dimensions of the spot, as evidenced by the different shapes of the radial intensity profiles of the bright-field and epifluorescence images, shown in Figure 7b. This phenomenon is caused by the out-of-focus imaging properties of the lens (Figure 3), combined with possible interchannel optical cross-talk. Nevertheless, the spot-widening effect associated with the epifluorescence imaging of a 500-µm-thick MCG chip with a 4× objective lens (Figure 7b) only amounts to a ∼80% increase in spot diameter. As a result, equivalent spots have a smaller fluorescence footprint on MCG than on flat glass, suggesting the possibility of arraying MCG at a higher density than flat glass with identical volumes of probe solution. Figure 8 illustrates the phenomenon of interchannel optical cross-talk in MCG, based on the results of a simulation in which a single model fluorophore was placed inside the central channel of a model MCG chip, at a distance of 10 µm from the front surface. From the simulation, it is apparent that most of the emitted fluorescence radiation emerges from the opening of the central channel, where the emitting fluorophore is located, while only a small fraction of fluorescence radiation is seen to have crossed the solid glass wall before emerging from the opening of neighboring channels. Simulated fluorescence emission patterns in which the fluorophore was placed at increasing distances (50450 µm) from the front surface featured an even less significant level of cross-talk than that represented in Figure 8. Figure 8 also shows that no significant amount of fluorescence radiation emerges from the interstitial surface, in the absence of light scattering on the front surface of the chip. A close analysis of the ray propagation pattern in OptiCAD confirmed that, although a number of rays travel inside the solid glass part of the structure, these rays undergo total internal reflection at the front surface of the substrate. This suggests the possibility of minimizing the amount of surface scattering, hence the level of background, by polishing the front surface of MCG chips to an optical flat. 2418 Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
Figure 8. Simulated 3D fluorescence intensity profiles on the front surface of a microchannel glass chip, illustrating the interchannel cross-talk of fluorescence radiations emitted by a single fluorophore, located inside the central channel, at a distance z ) 10 µm from the front surface of the chip. The units used in the labeling of the x and y axes are distances, in micrometers, along two orthogonal directions on the front surface of the model microchannel glass chip.
Coverage Density of ODN Probes on the Microchannel Sidewalls. The coverage density of probes within each spot significantly affects mixed-phase hybridization efficiencies. In particular, it is crucial to alleviate steric crowding that can hinder duplex formation.35 For instance, Steel et al.36 experimentally found the hybridization efficiency for a 25-mer ODN target to a 25-mer ODN probe to exhibit a maximum for a probe coverage density of 66 fmol mm-2, which is ∼50% lower than the density of a closely packed layer of duplexes laying flat on the surface. From above, MCG chips with a thickness of 0.5 mm appear to be well suited in terms of fluorescence detection and ODN spot generation, and this thickness was therefore considered further in this study. For a given dispensed volume, the coverage density of probes that can be achieved within a spot is controlled by the concentration of the probe solution, provided that a sufficient number of surface anchoring sites are available. The total number of thiol anchoring sites available within the effective area of a spot, calculated from the silane coverage density (determined as ∼2 × 1011 molecules per mm2, from XPS measurements, data not shown) and the spot footprint area, was less than the amount of probe present in the dispensed aliquot, for all concentrations. The attachment chemistry used in this work has an overall efficiency of covalent immobilization of 60%.21 In these conditions, the average coverage density of probes covalently attached on the sidewalls of the microchannels, resulting from the dispensing of a 5-nL aliquot from a 30 µM probe solution, was estimated as 46.4 fmol mm-2. This value lies below the limit coverage for steric hindrance of duplex formation.36 (35) Southern, E.; Mir, K.; Shchepinov, M. Nat. Genet. 1999, 21, 5-9. (36) Steel, A. B.; Herne, T. M.; Tarlov, M. J. Anal. Chem. 1998, 70, 4670-4677.
Figure 10. Epifluorescence image of a three-dimensional microchannel glass oligoarray, showing a pattern of fluorescently labeled oligodeoxynucleotide targets hybridized to spots of complementary probes. The image shows 12 replicates of a 4 × 4 array, the latter comprising four different probe sequences P0, P1, P2, and P3, each spotted in quadruplicate as indicated in the bottom left corner of the image. Each spot of the oligoarray comprises an estimated 90 fmol of probe molecules. The sample was 50 µL of a mixture of the targets T1, T2, and T3 at 2, 20, and 10 nM concentrations, respectively, in 5× SSPE buffer. Hybridization was carried out by iteratively circulating the sample through the microchannel glass oligoarray, under a constant flow rate of 0.5 mL min-1, at room temperature, for 3 h. Pre- and posthybridization washing steps consisted in flowing 5× SSPE through the chip at 0.5 mL min-1 for 15 min. A 4×, 0.20 NA microscope objective lens was used for imaging.
Figure 9. Net fluorescence intensities of spots generated by the dispensing of 5-nL aliquots from a fluorescently labeled probe solution, onto a 0.5-mm-thick microchannel glass chip, as a function of the concentration of the probe solution. Each curve is associated with images taken with the same objective lens, of given magnification factor and NA. (a) Spots imaged prior to washing the chip; (b) spots imaged after washing the chip with 10 mL of blocking solution, circulated at 500 µL min-1 through the flow cell. Key: circles, 1×, 0.04 NA lens; triangles, 2×, 0.06 NA lens; squares, 4×, 0.20 NA lens; diamonds, 10×, 0.45 NA lens.
Quantitative Fluorescence Detection in MCG Chips. A 0.5mm-thick MCG chip was spotted with 5-nL aliquots of 5, 12, 30, and 71 µM solutions of probe P0. Figure 9 shows the measured net fluorescence intensities of the spots, imaged with lenses of various NA, before (Figure 9a) and after (Figure 9b) washing the chip to remove noncovalently attached probe molecules. As shown in Figure 9a, the net fluorescence intensity measured using the 1×, 2×, and 4× lenses is proportional to the concentration of the probe solution, hence to the number of fluorescent labels present within the spot, over the range of concentrations studied. The plot for the 10×, 0.45 NA lens, however, shows saturation of the
fluorescence signal at high probe concentrations (71 µM). This can be explained by a higher density of probes in the lower part of the channels, beyond the depth of field of the higher magnification lens. A higher density of probes in the lower part of the channels could result from sedimentation of probe molecules within the high-concentration solution contained inside the microchannels, over the time frame of the evaporation process. Note that independent experiments showed self-quenching of the FAM labels not to be significant in the range of concentrations used here (data not presented). Direct assessment of the axial distribution of probes on the sidewalls of the microchannels by confocal optical sectioning microscopy could not be implemented satisfactorily, due to the fact that it was technically impossible to scan the confocal volume element across the array of microchannels at distances larger than 20 µm from the front surface. The lead glass matrix has a high RI (∼1.7) and no index matching fluid is available that could be used to fill the microchannels and thereby nullify RI inhomogeneities within the MCG structure. Figure 9b shows a plot of the same parameters as in Figure 9a, but after the chip was washed to remove noncovalently bound probe molecules. The plots for the 1×, 2×, and 4× lenses are linear for concentrations of the probe solution up to 30 µM (correlation coefficient r > 0.998). This indicates that, in these conditions, the epifluorescence imaging procedure allows for the quantitative determination of fluorophores bound to the microAnalytical Chemistry, Vol. 73, No. 11, June 1, 2001
2419
channel sidewalls. By extension, this is applicable to labeled targets hybridized to surface-tethered probes. The attenuation in signal observed for the highest concentration (71 µM) can be explained by a reduced binding efficiency, resulting from the limited number of available anchoring sites with respect to the number of probe molecules and, ultimately, steric crowding of probes on the available surface area. MCG Chip-Based Oligonucleotide Fluorescence Hybridization Assay. Chips prepared for hybridization experiments were 0.5-mm-thick MCG substrates, arrayed at 400-µm pitch with 5-nL spots of 30 µM probe solutions. Each spot was 140 µm in diameter and comprised an estimated 90 fmol of probe molecules. A total of 5.8 pmol of each of the four probe sequences was thus available for hybridization over the whole array, which represents a ∼6fold excess with respect to the most concentrated target present in the sample. Figure 10 shows the epifluorescence image of an area of the 256-spot chip, acquired with the 4× objective lens, after 3 h of hybridization with the target mixture. For each sequence, the coefficient of variation in signal to background noise, S/N, was calculated as ∼10%, reflecting nonuniformities in the flow pattern within the flow-through cell. The low levels of S/N observed for the P1-T1 duplex were attributed to a low degree of labeling of T1, by comparison to the two other targets, as evidenced from experiments in which the targets were present at equimolar levels (not shown). The detection limit, defined as the analyte concentration at which S/N equals 3, was determined in complementary experiments as 31 amol of 65-mer, FAM-labeled ODN target per spot, equivalent to a target concentration of 40 pM in the 50-µL sample volume. While 1.5 orders of magnitude in dynamic range could be achieved within a single image, a dynamic range of 3 orders of magnitude is possible by varying the integration time of the camera up to 10 s. These analytical figures of merit are obviously significantly affected by the nature of the fluorescent labels used, as well as by the spectral characteristics of both the fluorescence filters and the CCD detector. A FAM label was chosen here
2420
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
because its spectral characteristics matched the spectral detection efficiency of the CCD camera used. We expect the use of longer wavelength dyes such as TAMRA or Cy5, along with a CCD camera with enhanced detection efficiency in the red, to provide lower levels of background noise through reduced levels of scattering, hence improved performance. CONCLUSION The suitability of microchannel glass as a substrate for multiplexed, heterogeneous nucleic acid fluorescence hybridization assays has been demonstrated in this paper. In particular, the optical properties of MCG have been shown to be compatible with the quantitative detection of fluorophores located along the microchannel sidewalls, provided that imaging optics are used whose depth-of-field characteristics match the thickness of the MCG chip. Improved performance of the MCG biochip platform as an analytical device for hybridization assays, in the simple implementation of the detection scheme presented in this paper, requires further work to reduce the relatively high levels of background noise, caused by autofluorescence of the glass matrix and light scattering on the front surface of the substrate, which affect both the detectability and dynamic range of the assay. Accordingly, detection schemes that provide enhanced selectivity at the molecular level, including FRET-based methodologies such as molecular beacons, are being considered. Chemiluminescence is also envisaged as a potentially lower background scheme. The expected resulting increase in performance should make the MCG biochip platform a powerful tool for gene expression analysis and genetic diagnostics. Acknowledgment is made to Gene Logic Inc. for funding V.B. and to BBSR, EPSRC, and SHEFC for infrastructural support.
Received for review August 10, 2000. Accepted January 23, 2001. AC000946R
