Spatial DNA Melting Analysis for Genotyping and Variant Scanning

geometry was drawn to scale in Adobe Illustrator CS (Adobe Systems, Inc., CA). ... FC5100-75, Graphtec, CA) where a small blade traced out the pat...
0 downloads 0 Views 694KB Size
Download PDF
Anal. Chem. 2009, 81, 2053–2058

Spatial DNA Melting Analysis for Genotyping and Variant Scanning Niel Crews,†,§ Carl T. Wittwer,‡ Jesse Montgomery,‡ Rob Pryor,‡ and Bruce Gale*,† Departments of Mechanical Engineering and Pathology, University of Utah, Salt Lake City, Utah 84112 A continuous-flow, temperature gradient microfluidic device was used to demonstrate spatial DNA melting analysis with the resolution and reproducibility necessary for clinical SNP scanning and genotyping of human genomic DNA. With a steady-state temperature gradient of 20-30 °C across a sample, melting curves were constructed from a single fluorescence data acquisition. This technique was used to scan for heterozygotes and to fully genotype single base changes using unlabeled probes. Signal-to-noise ratios of 150-300 were achieved. The thermal effects of sample flow were examined, and temperature control was aided by inclusion of an isothermal channel inlet and thermal relaxation times in the experimental protocol. Human single base variants examined by spatial DNA melting analysis included rs354439, HTR2A 102T > C, and three alleles that affect appropriate warfarin dosage (CYP2C9*2, CYP2C9*3, and VKORC1 1173C > T). Heterozygote scanning was demonstrated with rs354439, while the other PCR targets were genotyped using unlabeled probes with Tm differences of approximately 5 °C between genotypes. To validate the method, 12 blinded DNA samples were genotyped at the three warfarin-related sites by spatial DNA melting analysis with 100% accuracy. Single nucleotide polymorphisms (SNPs) can cause genetic disorders,1,2 influence phenotype expression,3,4 and affect cellular metabolic activity.5-7 DNA scanning refers to the detection of any * To whom correspondence should be addressed. Phone: (801) 585-5944. Fax: (801) 585-9826. E-mail: [email protected]. ‡ Department of Pathology. † Department of Mechanical Engineering. § Current address: Institute for Micromanufacturing, Louisiana Tech University, Ruston, Louisiana 71272. (1) Chowdari, K. V.; Mirnics, K.; Semwal, P.; Wood, J.; Lawrence, E.; Bhatia, T.; Deshpande, S. N.; Thelma, B. K.; Ferrell, R. E.; Middleton, F. A.; Devlin, B.; Levitt, P.; Lewis, D. A.; Nimgaonkar, V. L. Hum. Mol. Genet. 2002, 11, 1373–1380. (2) Quist, J. F.; Barr, C. L.; Schachar, R.; Roberts, W.; Malone, M.; Tannock, R.; Basile, V. S.; Beitchman, J.; Kennedy, J. L. Mol. Psychiatry 2000, 5, 537–541. (3) Uda, M.; Galanello, R.; Sanna, S.; Lettre, G.; Sankaran, V. G.; Chen, W. M.; Usala, G.; Busonero, F.; Maschio, A.; Albai, G.; Piras, M. G.; Sestu, N.; Lai, S.; Dei, M.; Mulas, A.; Crisponi, L.; Naitza, S.; Asunis, I.; Deiana, M.; Nagaraja, R.; Perseu, L.; Satta, S.; Cipollina, M. D.; Sollaino, C.; Moi, P.; Hirschhorn, J. N.; Orkin, S. H.; Abecasis, G. R.; Schlessinger, D.; Cao, A. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 1620–1625. (4) Haase, B.; Jude, R.; Brooks, S. A.; Leeb, T. Anim. Genet. 2008, 39, 306– 309. (5) Hill, C. E.; Duncan, A.; Wirth, D.; Nolte, F. S. Am. J. Clin. Pathol. 2006, 125, 584–591. 10.1021/ac801495w CCC: $40.75  2009 American Chemical Society Published on Web 02/19/2009

heterozygous variants within a DNA segment,8 whereas genotyping involves the identification of homozygous and heterozygous variants at a specific locus. The polymerase chain reaction (PCR) is commonly used to amplify the region flanking the variant, which can subsequently be examined by probe hybridization,9,10 allelespecific extension,11 temperature gradient or capillary electrophoresis,12-14 or DNA melting analysis.15 Techniques for DNA melting analysis can be classified as either solid phase or liquid phase. Solid phase analysis involves the immobilization of DNA or hybridization probes. Dodge et al.16 examined the melting temperature of immobilized oligonucleotides (34-bp) using a molecular beacon fluorescence technique. The heating protocol consisted of a 1 °C/s temperature ramp for 5 s, followed by a temperature plateau for 5 s. The fluorescence was subsequently measured. Six data points between 40 and 65 °C were obtained using this heat-hold-image/heat-hold-image technique. Piunno et al.17 demonstrated the melting analysis of immobilized fluoresceinlabeled oligonucleotide duplexes (20-bp). Fluorescence was measured continuously while the system was heated at a rate of approximately 0.005 °C/s over a 60 °C temperature range. Stimpson et al.18 and Khomyakova et al.19 incorporated pinspotting techniques to perform simultaneous melting analysis of (6) Chen, J. S.; Lipska, B. K.; Halim, N.; Ma, Q. D.; Matsumoto, M.; Melhem, S.; Kolachana, B. S.; Hyde, T. M.; Herman, M. M.; Apud, J.; Egan, M. F.; Kleinman, J. E.; Weinberger, D. R. Am. J. Hum. Genet. 2004, 75, 807– 821. (7) Palles, C.; Johnson, N.; Coupland, B.; Taylor, C.; Carvajal, J.; Holly, J.; Fentiman, I. S.; Silva, I. D.; Ashworth, A.; Peto, J.; Fletcher, O. Hum. Mol. Genet. 2008, 17, 1457–1464. (8) Gundry, C. N.; Vandersteen, J. G.; Reed, G. H.; Pryor, R. J.; Chen, J.; Wittwer, C. T. Clin. Chem. 2003, 49, 396–406. (9) Kajiyama, T.; Miyahara, Y.; Kricka, L. J.; Wilding, P.; Graves, D. J.; Surrey, S.; Fortina, P. Genome Res. 2003, 13, 467–475. (10) Trau, D.; Lee, T. M. H.; Lao, A. I. K.; Lenigk, R.; Hsing, I. M.; Ip, N. Y.; Carles, M. C.; Sucher, N. J. Anal. Chem. 2002, 74, 3168–3173. (11) Russom, A.; Ahmadian, A.; Andersson, H.; Nilsson, P.; Stemme, G. Electrophoresis 2003, 24, 158–161. (12) Zhu, L.; Lee, H. K.; Lin, B. C.; Yeung, E. S. Electrophoresis 2001, 22, 3683– 3687. (13) Zhang, H. D.; Zhou, J.; Xu, Z. R.; Song, J.; Dai, J.; Fang, J.; Fang, Z. L. Lab Chip 2007, 7, 1162–1170. (14) Salimullah, M.; Hamano, K.; Tachibana, M.; Inoue, K.; Nishigaki, K. Cell. Mol. Biol. Lett. 2005, 10, 237–245. (15) Reed, G. H.; Kent, J. O.; Wittwer, C. T. Pharmacogenomics 2007, 8, 597– 608. (16) Dodge, A.; Turcatti, G.; Lawrence, I.; De Rooij, N. F.; Verpoorte, E. Anal. Chem. 2004, 76, 1778–1787. (17) Piunno, P. A. E.; Watterson, J. H.; Kotoris, C. C.; Krull, U. J. Anal. Chim. Acta 2005, 534, 53–61. (18) Stimpson, D. I.; Hoijer, J. V.; Hsieh, W. T.; Jou, C.; Gordon, J.; Theriault, T.; Gamble, R.; Baldeschwieler, J. D. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 6379–6383. (19) Khomyakova, E. B.; Dreval, E. V.; Tran-Dang, M.; Potier, M. C.; Soussaline, F. P. Cell. Mol. Biol. 2004, 50, 217–224.

Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

2053

an array of immobilized oligonucleotides. Fiche et al.20 demonstrated a label-free melting analysis by using a surface plasmon resonance (SPR) imaging technique. Data was acquired while the immobilized oligonucleotides (15-bp) were heated at a rate of 0.033 °C/s. While solid phase analysis allows for the study of highly parallel reactions within a single microfluidic chamber,18 surface modification protocols are required to prepare the devices for use. Liquid phase DNA melting analysis techniques can eliminate this processing step but allow for only a single analysis per well. Either fluorescently labeled oligonucleotide probes,21 unlabeled probes in the presence of a saturating DNA dye,22 or amplicon detection by the saturating dye alone23-25 can be used for liquid phase DNA melting analysis. Germer and Higuchi,26 and later Papp et al.,27 demonstrated allele-specific PCR of human genomic template and its subsequent melting analysis by using a fluorescent dye (SYBR Green I) rather than covalently labeled oligonucleotides.21 The fluorescence of the dye was monitored continuously during temperature ramps of 0.017 and 0.003 °C/s, respectively. Current commercial systems operate at temperature ramp rates from 0.009 to 0.1 °C/s. Slow heating ensures that the samples (∼10 µL) maintain thermal equilibrium during the continuous fluorescence acquisition. These commercial instruments can perform DNA melting analysis with acquisition times ranging from 2 to 90 min and signal-to-noise ratios ranging from 60 to 4000.28,29 Liquid phase DNA melting analysis by temperature ramping has been achieved in microfluidic devices. SNP genotyping of PCR-amplified DNA with a saturating dye was demonstrated by Sundberg et al.,30 who fabricated a microfluidic channel with 10 nL wells using a glass/polymer xurographic31 technique. Sample heating was achieved with a Peltier element at a rate of 0.1 °C/s while a fluorescence microscope acquired the melting data. The authors were able to obtain signal-to-noise ratios between 60 and 170 with a sample volume of 10 nL. Neuzil et al. demonstrated DNA melting by end-point analysis after a real-time PCR protocol.32 The instrument consisted of a small array of microelectromechanical system (MEMS) heaters, over which a glass coverslip was placed. Fluid droplets located on the coverslip were thermally cycled from beneath by means of the MEMS heaters and monitored from above using fluorescence microscopy. After (20) Fiche, J. B.; Fuchs, J.; Buhot, A.; Calemczuk, R.; Livache, T. Anal. Chem. 2008, 80, 1049–1057. (21) Howell, W. M.; Jobs, M.; Brookes, A. J. Genome Res. 2002, 12, 1401– 1407. (22) Zhou, L. M.; Wang, L.; Palais, R.; Pryor, R.; Wittwer, C. T. Clin. Chem. 2005, 51, 1770–1777. (23) Zhou, L.; Myers, A. N.; Vandersteen, J. G.; Wang, L.; Wittwer, C. T. Clin. Chem. 2004, 50, 1328–1335. (24) Wittwer, C. T.; Reed, G. H.; Gundry, C. N.; Vandersteen, J. G.; Pryor, R. J. Clin. Chem. 2003, 49, 853–860. (25) Liew, M.; Pryor, R.; Palais, R.; Meadows, C.; Erali, M.; Lyon, E.; Wittwer, C. Clin. Chem. 2004, 50, 1156–1164. (26) Germer, S. r.; Higuchi, R. Genome Res. 1999, 9, 72–78. (27) Papp, A. C.; Pinsonneault, J. K.; Cooke, G.; Sadee, W. BioTechniques 2003, 34, 1068–1072. (28) Herrmann, M. G.; Durtschi, J. D.; Bromley, L. K.; Wittwer, C. T.; Voelkerding, K. V. Clin. Chem. 2006, 52, 494–503. (29) Herrmann, M. G.; Durtschi, J. D.; Wittwer, C. T.; Voelkerding, K. V. Clin. Chem. 2007, 53, 1544–1548. (30) Sundberg, S.; Wittwer, C.; Greer, J.; Pryor, R.; Elenitoba-Johnson, O.; Gale, B. Biomed. Microdevices 2007, 9, 159–166. (31) Bartholomeusz, D. A.; Boutte, R. W.; Andrade, J. D. J. Microelectromech. Syst. 2005, 14, 1364–1374. (32) Neuzil, P.; Pipper, J.; Hsieh, T. M. Mol. Biosyst. 2006, 2, 292–298.

2054

Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

sample amplification, melting analysis was performed by slowly heating the droplets at a rate of 0.01 °C/s through a 30 °C temperature range. DNA melting analysis is traditionally a time-based process. A recent technique, spatial DNA melting analysis, monitors fluorescence as a function of position instead of time. This is achieved by creating a characteristic gradient within the DNA sample and imaging the resulting fluorescence. While temporal melting analysis uses a single photodetector over time, spatial melting data is collected only once with a detector array. This was achieved within a chemical gradient by Liedl and Simmel33 for doublestranded oligonucleotides ranging in size from 11-17 bp. Spatial DNA melting analysis in a temperature gradient was first demonstrated by Mao et al.34 using a glass substrate containing multiple microchannels that were each heated to a unique temperature. Prior to heating the channels, they were filled with a concentrated solution containing a DNA dye and double-stranded oligonucleotides (30-bp). The authors found that a single base pair mismatch near the center of the oligonucleotides produced a detectable change in the DNA melting curve. A similar demonstration was also performed by Baaske et al.,35 who used an infrared laser to generate a radial temperature gradient of 80 °C within a small (300 µm diameter × 20 µm deep) volume. The thermal denaturing of a concentrated sample containing a dyelabeled DNA hairpin (33-bp) was measured. These three prior studies (Liedl, Mao, Baske) were similar in that the DNA targets were all uniquely designed DNA sequences of very small size that were manufactured and purified for testing. Spatial DNA melting analysis for SNP scanning and genotyping of human genomic DNA templates has not been demonstrated prior to this current work. Recently, Crews et al.36 demonstrated simultaneous amplification and spatial melting analysis starting from a human genomic DNA template within a one-dimensional temperature gradient. The denaturing of the PCR product during each cycle was observed by imaging the continuous-flow temperature gradient PCR device, and the spatial melting curves of representative DNA samples were differentiated. However, the PCR products analyzed were very different in sequence, GC content, and size. Signal resolution was adequate to differentiate these very different samples but not to identify single base genotypes or scan for heterozygous variants. This article focuses on the experimental performance of a microfluidic device for spatial DNA melting analysis, developed to positively answer the question, “Can spatial DNA melting analysis achieve the resolution and reproducibility necessary for SNP scanning and genotyping from human genomic DNA?” Representative DNA targets were used to demonstrate heterozygote scanning within PCR amplicons and to completely genotype SNPs with unlabeled probes. Finally, a blinded study of 12 DNA samples for three alleles that contribute to appropriate warfarin dosing37 (CYP2C9*2, CYP2C9*3, and VKORC1 1173C > T) was performed. (33) Liedl, T.; Simmel, F. C. Anal. Chem. 2007, 79, 5212–5216. (34) Mao, H. B.; Holden, M. A.; You, M.; Cremer, P. S. Anal. Chem. 2002, 74, 5071–5075. (35) Baaske, P.; Duhr, S.; Braun, D. Appl. Phys. Lett. 2007, 91, 133901. (36) Crews, N.; Wittwer, C. T.; Palais, R.; Gale, B. Lab Chip 2008, 8, 919–924. (37) Sconce, E. A.; Khan, T. I.; Wynne, H. A.; Avery, P.; Monkhouse, L.; King, B. P.; Wood, P.; Kesteven, P.; Daly, A. K.; Kamali, F. Blood 2005, 106, 2329–2333.

Figure 1. (a) An image of a single-cycle thermal gradient device containing two channels used for spatial DNA melting analysis. (b) A composite diagram of the device and representative temperature gradient data obtained from an IR camera.

MATERIALS AND METHODS Device Fabrication. The microfluidic devices in this study were made by rapid prototyping using patterned polymer films sandwiched between glass substrates.31 The inexpensive, disposable devices were fabricated according to the following protocol: A two-dimensional outline of the channel geometry was drawn to scale in Adobe Illustrator CS (Adobe Systems, Inc., CA). The design was then exported to a cutting plotter (model no. FC510075, Graphtec, CA) where a small blade traced out the pattern on a double-sided adhesive polyimide film (KPTD-1, Kaptontape.com, CA) of uniform thickness (approximately 100 µm, including adhesive). Tweezers were then used to “weed” the thin film by removing the interior of the cut channel design. The patterned film was then manually aligned and applied to a precleaned glass microscope slide (Fisherfinest, Fisher Scientific, NH). A second microscope slide was prepared by drilling holes to mate with the inlet and outlet regions of the channel design. The holes were drilled with a variable speed Dremel Rotary Tool (no. 395, Dremel, WI), press (no. 212, Dremel, WI), and assorted diamond-tipped bits. A 2 mm thick layer of cured PDMS (polydimethylsiloxane) was then bonded to the top surface of the glass. This was done by activating the PDMS and glass surfaces with an oxygen plasma (PlasmaLab 80 Plus, Oxford Instruments, Oxfordshire, U.K.) for 30 s (at 73 W, 23.4 sccm O2) and pressing them together. A 1.5 mm diameter hole was then cored through the PDMS over each of the holes in the glass. The drilled and PDMS-coated glass slide was then affixed to the remaining exposed surface of the patterned adhesive. The resulting serpentine microfluidic channel possessed a roughly rectangular cross-section comprised of glass on the top and bottom and the film/adhesive layers on the sides. PEEK tubing (1531, Upchurch Scientific, WA) was inserted snugly through the cored holes in the PDMS to create a fluidic interface between the device and an external syringe. A representative device is shown in Figure 1a. For these devices, the lengths of the two parallel counterflow sections in each channel are 1 cm, having a width of 1 mm and a minimum wall thickness of 1 mm. Gradient Heating. A steady-state temperature gradient was induced along the microchannel, using the gradient heating

apparatus and method previously described.38 Improvements to the apparatus included a clamping mechanism derived from microscope stage clips (no. 24V0240, Ward’s Natural Science, Rochester, NY) and a closed loop temperature controller that allows the temperatures of the heating and cooling strips to be set (rather than the heater voltage, as in the previous work). Because of improved thermal contact with the clamps, thermal interface pads36 were no longer necessary. Infrared (IR) thermometry was used to characterize the temperature gradient,38 and the IR data were used to approximate the maximum and minimum channel temperatures as a function of the set temperatures of the heating and cooling strips. An example temperature gradient across the substrate and its relationship to channel geometry are shown in Figure 1b. Data Acquisition and Analysis. The filtered LED light source and the EMCCD camera used for fluorescence acquisition have been previously detailed.36 A 30 s exposure of ten 3 s accumulations was used in this study to further improve the signal without saturating the detector array. A single fluorescence image of spatial DNA melting was obtained for each sample. The images (1000 × 1000 pixels) were mapped to a near linear temperature distribution within the channel region and analyzed using MATLAB and LabVIEW algorithms that were previously described.36 Sample Preparation. For heterozygote scanning based on amplicon melting curve analysis, a 78-bp PCR product bracketing a common polymorphism (102T > C) within exon 1 of HTR2A8 was analyzed. For SNP genotyping with unlabeled probes, four primer/probe sets were used. The selected targets included a 93bp segment encompassing a common polymorphism (rs354439) used in forensics,39 122-bp and 134-bp segments encompassing the *2 and the *3 alleles (respectively) of the CYP2C9,37 and a 190-bp segment encompassing position 1173 of the VKORC1 gene.37 The PCR mixture contained 5 ng/µL of human genomic DNA, 200 µM of each deoxynucleotide triphosphate (dNTP), 0.04 U/µL of KlenTaq1 polymerase (AB Peptides, MO), 0.064 ng/µL of AntiTaq Monoclonal Antibody (eENZYME LLC, Gaithersburg, MD), 3 mM MgCl2, 1X LCGreen Plus (Idaho Technology, UT), and 250 ng/mL bovine serum albumin (BSA) in a 50 mM Tris (pH 8.3) buffer. Target-specific oligonucleotide primers and probes were also included in the mixture. For the HTR2A target, 0.5 µM of both primers were used with no probe. For the rs354439 target, 0.1 µM of the forward primer, 0.5 µM of the reverse primer, and 0.5 µM of the unlabeled probe were used. For the three other targets, 0.1 µM of the forward primer, 1.0 µM of the reverse primer, and 1.0 µM of the unlabeled probe were used. Human genomic DNA of each genotype was used as a template to characterize the melting behavior of each of the five target PCR sequences, after which a blinded study was performed for the three warfarin-related targets. All samples were amplified in a commercial real-time PCR instrument (LightCycler 1.5, Roche, IN) before being analyzed on the spatial melting microdevice. The amplification protocol consisted of an initial denaturation at 95 °C for 5 s, followed by (38) Crews, N.; Witwer, C.; Gale, B. Biomed. Microdevices 2008, 10, 187–195. (39) Sanchez, J. J.; Phillips, C.; Børsting, C.; Balogh, K.; Bogus, M.; Fondevila, M.; Harrison, C. D.; Musgrave-Brown, E.; Salas, A.; Syndercombe-Court, D.; Schneider, P. M.; Carracedo, A.; Morling, N. Electrophoresis 2006, 27, 1713–1724.

Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

2055

45 cycles of 95 °C for 0 s, 60 °C for 0 s, and 75 °C for 0 s. A temperature ramp rate of 2 °C/s was set for the transition between 60 and 75 °C. Maximum temperature ramping was used for all other transitions. All amplified genotypes were also analyzed on the HR-1 instrument (Idaho Technology, Salt Lake City, UT) over an acquisition temperature range of 50-90 °C with a 0.5 °C/s ramp rate. Results from this commercial instrument were compared with data from the spatial DNA melting analysis device. Blinded Study. After instrument characterization, a validation study was performed using the three warfarin-related PCR targets. Twelve DNA samples including at least one example of each genotype for each target were randomized and blinded. The 36 genotyping results were compared to HR-1 analysis and confirmed by sequencing as necessary. Spatial Melting Protocol. The temperature range of interest for each DNA target was first obtained using standard instrumentation (HR-1). The temperature controller setpoint temperatures (Tcool, Thot) used in the spatial DNA melting analysis of the PCR samples were HTR2A (60 °C, 105 °C); rs354439 (40 °C, 100 °C); CYP2C9*2 (45 °C, 115 °C); CYP2C9*3 (45 °C, 115 °C); VKORC1 (45 °C, 115 °C). Once the heating apparatus achieved the setpoint temperatures, an additional 5 min was allowed for the microfluidic device itself to equilibrate. Up to 20 DNA samples were analyzed in series within the same channel, with no cleaning step in between. The protocol for the sample injection and the fluorescence acquisition was as follows: A single 10 µL sample was aspirated into a syringe (Hamilton, NV) and manually injected into the channel with a flow rate on the order of 100 µL/min. When the sample advanced to just fill the channel, the syringe was placed into a pump (KDS120, KD Scientific, MA) and a continuous volumetric flow rate of 1.5 µL/min was initiated. After 60 s of continuous sample flow, the dye in the sample was excited with the LEDs and the resulting fluorescence was measured. The sample was then manually pumped out of the channel (∼100 µL/min), and a subsequent sample was loaded, injected, and imaged in the same way. In the event of significant bubbles or particulate contamination in the microfluidic channel during melting acquisition, the affected sample was retested. For the representative samples of known genotype, each DNA type was analyzed at least twice. In the blinded study, known genotypes were serially melted both before and after the unknown samples. Spatial melting curves of the known samples were used as a reference to genotype the unknown samples. RESULTS AND DISCUSSION Data Collection and Analysis. The temperature ranges within the microfluidic channels were approximately 67-90 °C (HTR2A), 53-82 °C (rs354439), and 59-94 °C (CYP2C9*2, *3 and VKORC1). Each sample was injected across these temperature gradients and imaged (see Figure 2). The abrupt fluorescence transitions within the microchannels (∼81-83 °C, in Figure 2) indicate the melting and rehybridization of the PCR product. When the flow is stopped, melting and rehybridization curves are identical (data not shown). Under flow conditions (1.5 µL/min), characteristic differences exist between the respective melting and 2056

Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

Figure 2. A fluorescence image of the spatial melting of a 78 bp HTR2A PCR product from genomic DNA. The superimposed dotted lines indicate the sidewalls of the microfluidic channel. Flow through the channel is from top to bottom. The temperature distribution is such that the fluid is heating as it moves left to right in the upper channel section and cooling as it moves right to left in the lower section. The melting of the amplicon is visible in the upper section and its hybridization in the lower. The temperature controller setpoints were 60 and 105 °C for this fluorescence acquisition.

Figure 3. Comparison of DNA melting on the spatial melting device and the HR-1 using (a) a HTR2A amplicon for heterozygote scanning, and (b) a rs354439 unlabeled probe (low temperature transition) and amplicon (high temperature transition) for genotyping.

rehybridization curves (Figure 2) that cannot be attributed to a flow-induced temperature shift.40 This “hysteresis” demonstrates the contrast between the molecular kinetics of the two processes. DNA melting is predominantly a function of temperature only, such that there is no observable change in the spatial melt curve with changing flow rate. Although rehybridization is likewise temperature-dependent, it is also a diffusion and concentrationdriven process. Therefore, the spatial rehybridization curve is a function of the flow rate, while the melting curve is not. The spatial DNA melting behavior in the sample heating region of the microchannels (see Figure 2) was qualitatively similar to the temporal melting data obtained on the HR-1 instrument (see Figure 3). Figure 3a compares HTR2A melting curves amplified from wild type and heterozygous 102T > C DNA. While both products melt at approximately the same temperature, the heterozygous sample is characterized by a broader transition and shallower slope. While heterozygotes can be recognized in this way,8,25,41 the genotyping of homozygotes is often difficult without the use of unlabeled probes in conjunction with asymmetric PCR.23,41 In this technique, unlabeled oligonucleotide probes (40) Crews, N.; Ameel, T.; Wittwer, C. T.; Gale, B. Lab Chip 2008, 8, 1922– 1929. (41) Liew, M.; Seipp, M.; Durtschi, J.; Margraf, R. L.; Dames, S.; Erali, M.; Voelkerding, K.; Wittwer, C. Am. J. Clin. Pathol. 2007, 127, 341–348.

Figure 4. Derivative probe melting curves for genotyping the three warfarin-related targets. The spatial melting chip was used to analyze two DNA samples of each genotype for each SNP.

anneal directly over the variation site and, being typically less than 40 bases in size, have significant genotype-specific melting characteristics. Figure 3b compares DNA melting curves for the three genotypes of rs354439. Each curve has two melting regions, one corresponding to the lower melting temperature of the probe (36-mer) and the other to the higher melting temperature of the amplicon (93-bp). While only the homozygous and heterozygous samples can be differentiated by their amplicon melting curves, the corresponding probe melting curves are quite distinct for each of the three genotypes, with a temperature span of more than 2 deg separating the melting peaks of the wild-type and variant alleles. Thus, heteroduplex scanning is shown in Figure 3a and specific probe genotyping in Figure 3b and Figure 4. Derivative plots of the probe melting region for the three genotypes of each warfarin-related target are shown in Figure 4. For all three targets, the genotype-specific curves are distinct, with both the wild-type and the homozygous variants exhibiting prominent single peaks at different temperatures and the heterozygous genotype as a double peak of lower amplitude. The oligonucleotide probes for the CYP2C9*2 and VKORC1 targets were designed to be fully complementary to the wild-type sequence at the location of interest, and the probe for the CYP2C9*3 target was designed as a perfect match to the variant. For this reason, the probe on the wild-type CYP2C9*3 target melts at a lower temperature than on the homozygous variant allele. Spatial melting analysis was used to genotype the blinded DNA samples. The genotypes obtained by this technique were 100% concordant to their previously established genotypes. The 12 blinded DNA samples included 1 CYP2C9*2 homozygote, 3 CYP2C9*2 heterozygotes, 1 CYP2C9*3 homozygote, 3 CYP2C9*3 heterozygotes, 1 VKORC1 homozygote, and 6 VKORC1 heterozygotes. Data Quality. A comparison of the curve sets in Figure 3 indicates a disparity between the melting temperatures obtained from the two analysis platforms. Since the HR-1 is considered the gold standard of DNA melting analysis, this disparity reflects inaccurate calibration of the spatial melting device. The assumption of a linear temperature gradient introduces error, since the actual gradient along the microchannel can vary by up to ±1 °C/ mm.38 Likewise, the calculation of the channel temperatures from the surface temperature distribution can introduce additional inaccuracies. However, these errors are only observed when comparing melting curves obtained from different instruments or channels. By evaluating the melting of multiple samples tested

serially in the same channel, all data have the same bias. These errors then become unimportant when comparisons between data sets are made. This is illustrated in Figure 3b, where the high thermal precision between samples melted successively within the same microfluidic channel allows for clear differentiation of the three genotypes. The spatial resolution of DNA melting is determined by the CCD detector array, the optical detection path, and the temperature gradient in the spatial melting device. The spatial pixel resolution of the CCD detector used in these experiments was 63 pixels/mm, resulting in temperature resolutions of 0.04 °C/ pixel for HTR2A, 0.05 °C/pixel for rs254439, and 0.06 °C/pixel for CYP2C9*2, *3, and VKORC1. Considering the warfarin targets, where the probe signals for the three genotypes span approximately 10 °C (Figure 4), this spatial resolution equates to about 180 data points over the melting of each probe. Were a finer resolution desired, the temperature controller could be set to a narrower temperature range so that the melting region of interest would span a greater portion of the channel and the CCD array. This “zoom” feature potentially increases melting resolution without adding time to the acquisition or complexity to the instrumentation. Different commercial DNA melting instruments have varying precision, with a signal-to-noise ratio between 60 and 4000.28,29 The melting curves obtained from this spatial DNA melting device have a signal-to-noise ratio between 150 and 300. Images containing small bubbles or particulates are correlated with a lower signalto-noise ratio. At the current depth of the microchannel (100 µm), each pixel of the spatial image corresponds to 25 pL of sample mixture. Although the depth of the features could be increased to improve the signal-to-noise ratio, this would lessen the operational advantages that are inherent to microscale systems. Operational Features. Precise spatial DNA melting requires a consistent temperature distribution in the microchannel during each set of experiments. To ensure such uniformity, an isothermal channel inlet was included in the microchannel geometry. The sample, initially at room temperature, enters through the inlet and travels through the device. The first arm of the microfluidic channel is designed to lie along an isotherm, allowing the sample temperature to reach the first temperature of the subsequent thermal ramp. After traveling the specified length, the channel turns 90°. From this point, the sample heats as it crosses the chip, then cools on its return prior to exit from the channel. Without equilibration before the ramp, a temperature lag can Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

2057

produce inconsistencies in the corresponding melting/rehybridization curves. The calculated thermal entry length for a flow rate of 1.5 µL/min is less than 1 mm, as long as substrate cooling is ignored. During testing, however, the isothermal inlet portion of the channel needed to be at least 5 mm long for serial melting curve consistency (data not shown). This finding indicates that sample injection causes significant temperature perturbations in the vicinity of the device inlet.40 A temperature drift of about 0.1 °C, approximately 2 pixels, is predicted within the gradient heating and cooling sections of the microfluidic channel when sample flow is 1.5 µL/min.40 However, the sample injection protocol included a manual filling step with much higher flow rates. Video acquisition of DNA fluorescence during and after rapid manual filling of the channel revealed a thermal relaxation time of approximately 45 s, after which the fluorescence distribution remained virtually unchanged. Therefore, a 60 s thermal equilibration time was included in the protocol between the sample injection and the data acquisition. CONCLUSION This project has successfully demonstrated high-resolution SNP scanning and genotyping by spatial DNA melting analysis. Human genomic DNA samples were analyzed for clinically relevant polymorphisms using a continuous-flow temperature gradient microfluidic device. The success of this study indicates that (1) high-resolution spatial DNA melting can be achieved using complex genomic DNA. While artificial templates and conditions are conducive to proof-of-concept testing, the use of real samples is a better indicator of the viability of an analysis technique.20 While spatial melting was demonstrated by Liedl and Simmel,33 Mao et al.,34 and Baaske et al.,35 each used highly purified, concentrated, artificial oligonucleotide duplexes of small length (17, 30, and 33-bp, respectively). Prior to this current work, no known spatial melting studies have been performed with samples of clinical genomic DNA. (2) Variant scanning and SNP genotyping can be achieved by analyzing individual images, as an alternative to time-based data acquisition. This essentially uncouples the achievable resolution from the time of experiment. For temporal DNA melting instruments, higher resolution requires a slowing of the temperature ramp rate, potentially extending the analysis time to an hour or more.28,29 For spatial melting analysis, however, the time of the experiment remains unchanged, while higher resolution is achieved by altering the steady-state temperature gradient and upgrading the optical system. (3) Thermal effects within continuous-flow microfluidics can be stabilized such that high-resolution spatial DNA melting analysis of serial samples can be achieved. With the exception of our previous work,36 continuous-flow spatial DNA melting analysis has not been previously reported. Rather, gradients were generated in stationary samples and subsequently imaged.33-35 Although flow-induced thermal effects in temperature gradient microfluidics have been characterized,40 the physical application of this knowledge to a spatial DNA melting system has not previously been achieved.

2058

Analytical Chemistry, Vol. 81, No. 6, March 15, 2009

(4) High-resolution DNA melting analysis of PCR products amplified from human genomic DNA can be achieved at higher heating rates than prior reports.28,29 Under the given experimental conditions (geometry, temperature gradient, flow rate, etc.), the continuously flowing sample experienced an average temperature ramp rate of approximately 1 °C/s, during both melting (heating) and rehybridization (cooling). Were the channel geometry modified, this ramp rate would increase in direct proportion to the reduction in the cross-sectional area (e.g., 10 times smaller equates to 10 °C/s). For “macrofluidic” systems, increasing the heating rate risks temperature variations within the sample, potentially obscuring details of the melting curve. However, for the microfluidic spatial gradient system described here, the flow-induced temperature variations in both the device substrate and the flowing sample remain unchanged, as long as the cross-sectional area of the microfluidic channel is scaled evenly with its hydraulic diameter.40 (5) Both liquid phase DNA melting and its subsequent rehybridization can be examined simultaneously. This is achieved without labeling the DNA samples and without chemically modifying the microchannel. Although this simplification comes at the cost of reduced parallel processing, this technique is conducive to high serial processing, making it particularly suited to applications in personalized medicine or field diagnostics. This work has shown that spatial DNA melting analysis using microfluidics with steady-state temperature and flow conditions is a viable technique for scanning and genotyping of SNPs within genomic DNA samples. Melting data with a signal-to-noise ratio of up to 300 and a temperature resolution as fine as 0.04 °C/pixel was obtained. Both heterozygote scanning based on amplicon melting and unlabeled probe genotyping were demonstrated with correct genotyping of 12 blinded samples at three loci relevant to warfarin dose estimation. Integration of fluorescence imaging and continuous-flow microfluidics has a promising future. In these ongoing studies, however, channel feature geometry and flowinduced thermal effects should not be ignored. ACKNOWLEDGMENT Authors credit funding by the State of Utah Center of Excellence program and the NSF IGERT Program. Authors thank Josh Vandersteen for the design of the HTR2A primers and probe, Scott Sundberg and Oluwole Elenitoba-Johnson for their development of the fluorescence acquisition system, and other members of the Wittwer and Gale research laboratories whose efforts have advanced this work. SUPPORTING INFORMATION AVAILABLE Listing of the primer/probe sets used in this study. This material is available free of charge via the Internet at http://pubs. acs.org. Received for review July 17, 2008. Accepted January 29, 2009. AC801495W