Optimization of High-Speed DNA Sequencing on Microfabricated

Anal. Chem. 1999, 71, 566-573

Optimization of High-Speed DNA Sequencing on Microfabricated Capillary Electrophoresis Channels Shaorong Liu,† Yining Shi, William W. Ja, and Richard A. Mathies*

Department of Chemistry, University of California, Berkeley, California 94720

DNA sequencing separations have been performed in microfabricated electrophoresis channels with the goal of determining whether high-quality sequencing is feasible with these microdevices. The separation matrix, separation temperature, channel length and depth, injector size, and injection parameters were optimized. DNA fragment sizing separations demonstrated that 50-µm-deep channels provide the best sensitivity for our detection configuration. One-color sequencing separations of singlestranded M13mp18 DNA on 3% linear polyacrylamide (LPA) were used to optimize the twin-T injector size, injection conditions, and temperature. The best one-color separations were observed with a 250-µm twin-T injector, an injection time of 60 s, and a temperature of 35 °C. The first 500 bases appeared in 9.2 min with a resolution of >0.5, and the separation extended to 700 bases. The best four-color sequencing separations were performed using 4% LPA, a temperature of 40 °C, and a 100-µm twin-T injector. These four-color runs were complete in only 20 min, could be automatically base-called using BaseFinder to over 600 bp after the primer, and were 99.4% accurate to 500 bp. These results significantly advance the quality of microchip-based electrophoretic sequencing and indicate the feasibility of performing highspeed genomic sequencing with microfabricated electrophoretic devices. The Human Genome Project has catalyzed interest in the development of high-speed, high-throughput DNA sequencing methods. DNA sequencing separations have traditionally been performed by slab gel electrophoresis.1 Subsequently it was demonstrated that capillary electrophoresis (CE) is an effective high-speed method for DNA fragment sizing and sequencing,2-6 † Current address: Molecular Dynamics Inc., 928 East Arques Ave., Sunnyvale, CA 94086. (1) Smith, L. M.; Sanders, J. Z.; Kaiser, R. J.; Hughes, P.; Dodd, C.; Connell, C. R.; Heiner, C.; Kent, S. B. H.; Hood, L. E. Nature 1986, 321, 674-679. (2) Cohen, A. S.; Najarian, D. R.; Paulus, A.; Guttman, A.; Smith, J. A.; Karger, B. L. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 9660-9663. (3) Kasper, T. J.; Melera, M.; Gozel, P.; Brownlee, R. G. J. Chromatogr. 1988, 458, 303-312. (4) Drossman, H.; Luckey, J. A.; Kostichka, A. J.; D’Cunha, J.; Smith, L. M. Anal. Chem. 1990, 62, 900-903. (5) Cohen, A. S.; Najarian, D. R.; Karger, B. L. J. Chromatogr. 1990, 516, 4960. (6) Swerdlow, H.; Wu, S.; Harke, H.; Dovichi, N. J. J. Chromatogr. 1990, 516, 61-67.

566 Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

and very high performance CE-based DNA sequencing has been reported.7-10 Although CE separations are faster and more efficient than slab gel separations, conventional CE allows analysis of only one sample at a time. Capillary array electrophoresis (CAE), using confocal scanning detection,11-14 provided a practical solution to these throughput challenges; a number of alternative CAE formats have been subsequently described.15-17 A 96-capillary, four-color confocal scanning CAE apparatus has now been developed and is being applied to DNA fragment sizing and genomic sequencing.18,19 However, to meet the ever growing demand for increased speed and sample throughput, even more advanced approaches to DNA sequencing and analysis are needed. Recently, microfabricated electrophoretic devices have been introduced that increase the speed and sample throughput of DNA separations by an order of magnitude.20-24 Microfabrication of electrophoretic separation channels using photolithographic technologies was first described in 1992 by Manz and co-workers.25 These devices were used to perform separations of fluorescent (7) Zhang, J. Z.; Fang, Y.; Hou, J. Y.; Ren, H. J.; Jiang, R.; Roos, P.; Dovichi, N. J. Anal. Chem. 1995, 67, 4589-4593. (8) Ju, J.; Kheterpal, I.; Scherer, J. R.; Ruan, C.; Fuller, C. W.; Glazer, A. N.; Mathies, R. A. Anal. Biochem. 1995, 231, 131-140. (9) Carrilho, E.; Ruiz-Martinez, M. C.; Berka, J.; Smirnov, I.; Goetzinger, W.; Miller, A. W.; Brady, D.; Karger, B. L. Anal. Chem. 1996, 68, 3305-3313. (10) Kim, Y.; Yeung, E. S. J. Chromatogr., A 1997, 781, 315-325. (11) Huang, X. C.; Quesada, M. A.; Mathies, R. A. Anal. Chem. 1992, 64, 967972. (12) Huang, X. C.; Quesada, M. A.; Mathies, R. A. Anal. Chem. 1992, 64, 21492154. (13) Mathies, R. A.; Huang, X. C. Nature (London) 1992, 359, 167-169. (14) Kheterpal, I.; Scherer, J. R.; Clark, S. M.; Radhakrishnan, A.; Ju, J.; Ginther, C. L.; Sensabaugh, G. F.; Mathies, R. Electrophoresis 1996, 17, 1852-1859. (15) Takahashi, S.; Murakami, K.; Anazawa, T.; Kambara, H. Anal. Chem. 1994, 66, 1021-1026. (16) Quesada, M. A.; Zhang, S. Electrophoresis 1996, 17, 1841-1851. (17) Ueno, K.; Yeung, E. S. Anal. Chem. 1994, 66, 1424-1431. (18) Mansfield, E. S.; Vainer, M.; Harris, D. W.; Gasparini, P.; Estivill, X.; Surrey, S.; Fortina, P. J. Chromatogr., A 1997, 781, 295-305. (19) Bashkin, J.; Marsh, M.; Barker, D.; Johnston, R. Appl. Theor. Electrophor. 1996, 6, 23-28. (20) Woolley, A. T.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 1134811352. (21) Woolley, A. T.; Mathies, R. A. Proc. Int. Soc. Opt. Eng.-SPIE 1995, 2386, 36-44. (22) Woolley, A. T.; Mathies, R. A. Anal. Chem. 1995, 67, 3676-3680. (23) Woolley, A. T.; Sensabaugh, G. F.; Mathies, R. A. Anal. Chem. 1997, 69, 2181-2186. (24) Simpson, P. C.; Roach, D.; Woolley, A. T.; Thorsen, T.; Johnston, R.; Sensabaugh, G. F.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 2256-2261. (25) Harrison, D. J.; Fluri, K.; Seiler, K.; Fan, Z.; Effenhauser, C. S.; Manz, A. Science 1993, 261, 895-897. 10.1021/ac980783v CCC: $18.00

© 1999 American Chemical Society Published on Web 12/22/1998

dyes26 and fluorescently labeled amino acids.27,28 It has also been shown that DNA restriction fragments,20,29 PCR products,20 and short oligonucleotides30 can be rapidly and effectively separated with CE chips. Woolley et al.31 demonstrated that PCR-based sample preparation can be functionally integrated into these microfabricated devices, and other approaches for on-chip DNA sample preparation were subsequently described.32 High-speed separations of DNA sequencing fragments suitable for diagnostic and screening applications (e200 bases) have also been achieved on CE chips,21,22 and recent results have extended the read length of single-color sequencing separations to 400 bases.33 Further improvements are needed in the resolution, sensitivity, and read length of separations on these microfabricated CE devices to make them practical for genomic sequencing applications. In this report, we explore the optimization of DNA sequencing separations on microfabricated CE channels. Compared to our earlier work,22 we have increased the separation channel length, optimized the twin-T injector size, increased the separation channel depth, and elevated the separation temperature to enhance the resolution, read length, and sensitivity of these separations. The best separations on 4-in.-diameter substrates are obtained using 3-4% linear polyacrylamide (LPA) as the sieving matrix, a temperature of 35-40 °C, 100-250-µm twin-T injectors, and 50µm-deep straight channels that have an effective separation distance of 6-7 cm. On these devices, one-color sequencing experiments exhibit single-base resolution to ∼500 bases in ∼10 min and four-color DNA sequencing experiments produce 500 bp of data in under 20 min that can be automatically base-called to 99.4% accuracy. EXPERIMENTAL SECTION Chip Fabrication. The microfabrication procedures have been described previously.24,34 All devices were made using 10-cmdiameter Borofloat glass wafers (Precision Glass & Optics, Santa Ana, CA). These wafers were pre-etched in concentrated HF for 15 s and cleaned before deposition of a 1500-Å amorphous silicon sacrificial layer in a plasma-enhanced chemical vapor deposition (PECVD) system (PEII-A, Technics West, San Jose, CA). Then, wafers were primed with hexamethyldisilazane (HMDS), spincoated with photoresist (Shipley 1818, Marlborough, MA) at 5500 rpm, and soft-baked at 90 °C for 20-30 min. A contact mask aligner (Quintel Corp. San Jose, CA) was used to expose the photoresist layer with the mask design, and the exposed photoresist was removed using a 1:1 mixture of Microposit developer (26) Jacobson, S. C.; Hergenroeder, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113. (27) Manz, A.; Harrison, D. J.; Verpoorte, E. M. J.; Fettinger, J. C.; Paulus, A.; Ludi, H.; Widmer, H. M. J. Chromatogr. 1992, 593, 253-258. (28) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 26372642. (29) Jacobson, S. C.; Hergenroeder, R.; Koutny, L. B.; Ramsey, J. M. Anal. Chem. 1994, 66, 1114-1118. (30) Effenhauser, C. S.; Paulus, A.; Manz, A.; Widmer, H. M. Anal. Chem. 1994, 66, 2949-2953. (31) Woolley, A. T.; Hadley, D.; Landre, P.; deMello, A. J.; Mathies, R. A.; Northrup, M. A. Anal. Chem. 1996, 68, 4081-4086. (32) Waters, L. C.; Jacobson, S. C.; Kroutuchinina, N.; Khandurina, J.; Foote, R. S.; Ramsey, J. M. Anal. Chem. 1998, 70, 158-162. (33) Schmalzing, D.; Adourian, A.; Koutny, L.; Ziaugara, L.; Matsudaira, P.; Ehrlich, D. Anal. Chem. 1998, 70, 2303-2310. (34) Simpson, P. C.; Woolley, A. T.; Mathies, R. A. Biomed. Microdevices 1998, 1, 7-26.

Figure 1. Mask design used in this work. From left to right, these channels have cross and 100-, 250-, and 500-µm twin-T injectors. The circle illustrates the size of the 10-cm-diameter Borofloat substrate. The distance from the injector to the anode was 7.5 cm for each channel.

concentrate (Shipley) and H2O. Developed wafers were then hardbaked at 120 °C for 10-15 min, and the exposed amorphous silicon was removed using a CF4 plasma in the PECVD reactor. Wafers were chemically etched with concentrated HF at room temperature (etch rate 7 µm/min) to produce channels with depths from 10 to 50 µm. The remaining photoresist was stripped using 3:1 concentrated sulfuric acid and 30% hydrogen peroxide, and the amorphous silicon was removed with a CF4 plasma etch. Access holes were drilled into the etched wafers with a 0.75-mmdiameter diamond drill bit (Crystalite, Westerville, OH). A finished CE chip was prepared by thermally bonding an etched and drilled plate to a flat wafer of the same size in a programmable vacuum furnace (Centurion VPM, J. M. Ney, Yucaipa, CA). The design and layout of the chip used in this work is presented in Figure 1. Channels were masked to 30-µm width; the final etched channel width depends on the etch depth, ranging from 70 to 130 µm for channels from 20 to 50 µm in depth, respectively. Channel Derivatization. Channel surfaces were coated with linear polyacrylamide with minor modification of the Hjerten procedure.35 First, channels were washed with 1 M NaOH for ∼45 min and rinsed with water, and then a solution of 0.4% (v/v) of [γ-(methacryloxy)propyl]trimethoxysilane (Sigma, St. Louis, MO) and 0.2% acetic acid in acetonitrile was drawn through the channels for ∼1 h using vacuum. The channels were rinsed with acetonitrile and filled with a degassed 4% (w/v) acrylamide solution containing 0.01% (w/v) ammonium persulfate and 0.01% (v/v) N,N,N′,N′tetramethylethylenediamine (TEMED). This solution was allowed to polymerize in the channel for ∼5 min, and then the channel was flushed with water and dried by drawing air through the channel with vacuum. Separation Matrix Preparation. LPA was prepared following the procedure of Carrilho et al.9 A 10-mL solution of 6% (w/v) (35) Hjerten, S. J. Chromatogr. 1985, 347, 191-198.

Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

567

acrylamide at 0 °C was purged with high-purity helium for ∼1 h. Ammonium persulfate (10 µL of 10% w/v) and 10 µL of 10% (v/v) TEMED were added to the acrylamide solution to initiate the polymerization. The polymerization was allowed to proceed for ∼24 h in the ice bath. Urea, 10× Tris-Taps (0.50 M Tris, 0.50 M Taps, and 20 mM EDTA) solution, and deionized water were then added to the polymerized (6%) LPA solution to produce a final LPA separation matrix containing (as needed) 3 or 4% LPA, 7 M urea, and 1× Tris-Taps. For fragment sizing a 0.75% (w/v) hydroxyethylcellulose (HEC) sieving matrix in 1× TAE (40 mM Tris, 40 mM acetate, 1 mM EDTA, pH 8.2) was used.20 Sequencing Sample Preparation. Sequencing extension reactions were produced using dideoxy sequencing chemistry with cyanine-donor energy-transfer dye-labeled primers.36 Singlestranded M13mp18 DNA and other reagents used to perform the sequencing reactions were obtained from Amersham Life Science (Cleveland, OH). The DNA sequencing fragments used in the onecolor experiments were made using 0.4 pmol of the C10R110 primer (see ref 36 for nomenclature) and 0.2 µg of single-stranded M13mp18 DNA template, and they were terminated using ddTTP. Cycle sequencing reactions were performed according to protocols provided with the Amersham Life Science reagent kit (20 cycles of 30 s at 95 °C for denaturing, 15 s at 45 °C for annealing, and 45 s at 70 °C for extension). The reaction products were precipitated, washed with ethanol, and resuspended in a mixture of 2 µL of deionized formamide and 1 µL of deionized water. The sample used in the four-color separation experiments was generated using the same amount of primer and DNA template for each of the four reactions. The primers used for the reactions terminated by ddCTP, ddTTP, ddGTP, and ddATP were C10R110, C10G, C10T, and C10R, respectively. The sequencing reactions for the fourcolor sequencing samples were carried out using the same protocols employed for the one-color sequencing samples. The four reaction products were precipitated, washed with ethanol, and resuspended in a mixture of 2 µL of deionized formamide and 1 µL of deionized water. Electrophoresis Methods. The separation and cross channels were filled with separation matrix by pumping the LPA solution through the anode access hole with a syringe until the separation and cross channels were completely filled. Pipet tips were then inserted into the cathode, waste, and anode access holes and filled with the same separation matrix. The sample access hole was rinsed with deionized water, and then 1 µL of the sequencing sample was pipetted into the sample hole. A sample reservoir was then formed by inserting an empty pipet tip into the sample hole; the solution level rose by ∼1-2 mm above the chip surface as the pipet tip was pressed into the hole. Electrodes were then inserted into all the reservoirs and the chip was aligned in the confocal fluorescence detector. The sample was loaded using a pinched injection procedure which consists of applying relative voltages of 150 V for 60 s to the waste reservoir while keeping all other reservoirs at 0 V. Electrophoretic separation was carried out at 150 V/cm; back-bias voltages of 90 and 120 V were applied to the sample and waste reservoirs, respectively, to clear excess sample from the injection region. The microfabricated chip was mounted on a hollow aluminum plate; the temperature of the plate was thermostated by flowing water from a regulated bath through (36) Hung, S.; Mathies, R. A.; Glazer, A. N. Anal. Biochem. 1997, 252, 78-88.

568

Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

the interior of the plate. The chip and thermal plate assembly was mounted on an x-y-z translation stage for precise adjustment of the detection system focus and alignment with the channels. Laser Confocal Fluorescence Detection System. The detection systems used for the one-color20 and for the four-color8 sequencing separations have been described previously. The onecolor system employed a 32× NA 0.4 objective imaged to a 400µm pinhole. This optical system defined a depth of focus of ∼50 µm at the sample. In the four-color detection system, the 488-nm line from an argon ion laser was focused within the channel using a 20× NA 0.5 microscope objective. Fluorescence was collected by the objective lens and passed through a series of dichroic filters to divide the fluorescent signal into four spectral regions (510540, 540-570, 568-592, and >590 nm). The fluorescent signal in the first three spectral regions was filtered by a band-pass filter; the fourth region was filtered with a 590-nm long-pass and a 660nm short-pass filter; the emitted light in all four spectral channels was focused through 100-µm confocal pinholes with a 100-mmfocal length lens to spatially filter the fluorescence before photomultiplier detection. The depth of focus defined by the image of this pinhole on the channel was ∼30 µm. Signal from the photomultipliers was sampled at 10 Hz with a 16-bit ADC board (NB-MIO-16XL-18, National Instruments, Austin, TX) controlled by a program written in LabView. Base-Calling Procedure. Raw DNA sequencing data traces were reduced and base-called using the program, BaseFinder, provided by Lloyd Smith.37 BaseFinder uses a set of modules, each with user-input parameters, to define a script that can be saved and used on multiple runs. The data were treated first by baseline correction and then reduced by performing a multicomponent matrix transformation to correct for spectral cross-talk.37-39 The script used to analyze the chip data reported here included the following: primer peak deletion, baseline subtraction, spectral separation to remove cross-talk, two rounds of successive noise filtering and deconvolution, a final noise filtering, histogram equalization, mobility shift correction, and base-calling.40 BaseFinder allows the final sequence to be saved in a variety of formats, including the SCF format,41 which is directly usable by the basecalling program, Phred.42 (37) Giddings, M. C.; Severin, J. S.; Westphall, M.; Wu, J. Z., Smith, L. M. Genome Res. 1998, 8, 644-665. (38) Yin, Z. B.; Severin, J.; Giddings, M. C.; Huang, W. A.; Westphall, M. S.; Smith, L. M. Electrophoresis 1996, 17, 1143-1150. (39) Giddings, M. C.; Brumley, R. L.; Haker, M.; Smith, L. M. Nucleic Acids Res. 1993, 21, 4530-4540. (40) The script steps included the following: primer peak deletion; baseline subtraction (calculated over regions of 75 s or 750 data points); spectral separation to remove cross-talk due to spectral overlap of the four dye primers (BaseFinder allows formulation of the matrix by simply selecting four isolated peaks, one for each base); two rounds of successive noise filtering (first round, window width M ) 10, Gaussian width σ ) 2; second round, M ) 10, σ ) 1.8) and deconvolution (four iterations with the default spread function and R ) 0.25); a final noise filtering (M ) 5, σ ) 1); preliminary base-calling to allow for sequence-adjusted normalization of the four channels (default values except for maximum iterations, 3); normalization of the signal intensities; histogram equalization (to cut off high, spurious peaks); mobility shift correction; and base-calling. The final base-calling module used parameters that evenly weighted base spacing, peak widths, and peak heights. A maximum of 10 iterations was allowed, and beginning and end width tolerances were set at 1.70 and 1.35, respectively. Spacing tolerance was set at 1.45. (41) Dear, S.; Staden, R. DNA Sequence 1992, 3, 107-110. (42) Ewing, B.; Hillier, L.; Wendl, M. C.; Green, P. Genome Res. 1998, 8, 175194.

Figure 2. Effect of channel depth on sensitivity. (top) Separations of ΦX174 HaeIII (20 ng/µL) in 1× TAE performed with a 250-µm twin-T injector on a 7-cm-long separation channel having the indicated etch depths. (bottom) Relative fluorescence intensity of the 603-, 872-, and 1353-bp bands as a function of channel depth. The 0.75% HEC sieving matrix contained 400 nM thiazole orange in 1× TAE.

RESULTS AND DISCUSSION Figure 1 presents the channel layout used for the optimization of the separation parameters. This chip contains four separate electrophoresis channels that have a cross and a 100-, a 250-, and a 500-µm twin-T injector. Each channel provides a total separation column length of ∼7.5 cm (from the edge of the anode reservoir to the injector), and the effective separation length is 6.5-7 cm depending on the position of the detector. Figure 2 illustrates the effect of increased channel depth on the signal strength. Fragment sizing separations of ΦX174 HaeIII DNA were performed on a channel with a 250-µm twin-T injector and 0.75% HEC as the sieving matrix. The signal increased significantly as the channel depth was changed from 7 to ∼21 µm, but increasing the depth to 55 µm resulted in only a modest additional improvement. Further increases in the channel depth beyond 55 µm made it easier to load the sample matrix, but these very deep channels exhibit increased Joule heating. One would expect the signal to increase until the channel depth was on the order of the depth of focus of the objective. For the 20× NA 0.50 objective and 100-µm pinhole used in the four-color detection system, we calculate a depth of focus of ∼30 µm while for the

Figure 3. Effect of injector size on the resolution and signal strength of DNA sequencing separations. The resolution in the indicated size ranges was determined from isolated doublets that are one base apart and is presented as an average (( one standard deviation) for each region. Electropherograms were generated with cross and 100-, 250-, and 500-µm twin-T injectors which are expected to produce nominal injection plugs of 130, 230, 380, and 630 µm. Separations were carried out using 3% LPA containing 7 M urea and 1× Tris-Taps. The one-color DNA sequencing sample was prepared with 0.25 µg of M13mp18DNA template. The extended fragments were ethanol precipitated and suspended in 3 µL of 70% formamide and 30% water for injection. Voltages for sample loading and separation are described in the Experimental Section.

32× NA 0.4 objective used in the one-color system we calculate a depth of focus of ∼50 µm. The channels used subsequently in this work were ∼50 µm deep to provide good signal strength as well as facile alignment and matrix filling. Figure 3 presents average band areas and resolution values at various positions in the separation of M13mp18 T-extension fragments on channels having a 7-cm length and twin-T injectors with an offset of 0, 100, 250, and 500 µm. The resolution was calculated for all pairs of T-extension products separated by one base using the equation

R ) (2 ln 2)1/2 (t2 - t1)/(hw1 + hw2)

(1)

Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

569

Figure 4. Effect of injection time on the signal strength of DNA sequencing separations using a 250-µm twin-T injector.

where t is the migration time of the nth peak and hw is the full width at half-maximum of the nth peak. The signal intensity was quantitated by the relative band area of these same doublets. The resolution was the highest for the cross injector (130-µm nominal plug size) and decreased as the plug size was increased. The cross injector gave the lowest signal intensity. The 100- and 250-µm offset injectors resulted in somewhat reduced resolution and a significant increase in signal strength. The 500-µm twin-T injector produced much poorer resolution. On the basis of these observations, we employed 100- or 250-µm twin-T injectors in our DNA sequencing experiments, depending on the resolution requirements. The separations were significantly improved when the samples were loaded by electrophoresing through the cross channels to the twin-T injector. During injection, this electrophoresis of the DNA fragments through the cross channels provides a partial cleanup and differential concentration of the sequencing sample. Little change in concentration will occur at the sample/gel interface for small DNA fragments and inorganic ions because 570

Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

their electrophoretic mobilities are similar in free solution and in LPA. On the other hand, a considerable increase in the steadystate concentration will occur at the sample/gel interface for the larger fragments because of their reduced mobility in the gel. It was therefore necessary to determine the injection time that allows the larger DNA fragments to reach a high steady-state concentration in the intersection. Figure 4 illustrates how the experimental results depend on injection time. Although the resolution is not sensitive to injection time, the signal strength increases with increased injection time and then plateaus. With a short injection time (for example, 15 s), the signal from the long fragments is lower than that from the short ones due to their mobility differences. At an injection time of ∼60 s, the majority of the fragments have reached a steady-state concentration in the injector, as evidenced by the more uniform and increased intensities. A very long injection time risks introducing larger molecules such as the DNA template and polymerase into the injector as well as depletion of the smaller DNA fragments in the reservoir. By using an injection time of 60 s, we were able to

Figure 5. Unprocessed data from an optimized one-color separation of an M13mp18 T-sequencing extension reaction on a 7.5-cm-long separation channel. Separation was performed with the chip thermostated at 35 °C, an injection time of 1 min, a 250-µm twin-T injector, and a channel depth of 50 µm. Separation was performed on 3% LPA at 150 V/cm, and the detector was placed 7 cm from the injector.

exclude the slower moving template and enzyme contaminants from the injector and achieve improved separations. Elevated temperature has been reported to improve the resolution and to reduce compressions in DNA sequencing separations.7,43,44 Elevating the temperature of the entire chip to 34-44 °C resulted in improved resolution and reduced compressions; further increases in temperature tended to degrade the separations, presumably due to enhanced diffusive broadening of the bands. It should be noted that the separation speed increased dramatically with increased temperature. At 20 °C, ∼600-bp fragments were detected in 12 min, while at 65 °C, these fragments were detected in 0.5) extends to ∼500 bases. This is significantly better than the 200 base reads achieved earlier on chips by Woolley and Mathies in 1995.22 This improvement is (43) Ruiz-Martinez, M. C.; Berka, J.; Belenkii, A.; Foret, F.; Miller, A. M.; Karger, B. L. Anal. Chem. 1993, 65, 2851-2858. (44) Kleparnik, K.; Foret, F.; Berka, J.; Goetzinger, W.; Miller, A. W.; Karger, B. L. Electrophoresis 1996, 17, 1860-1866. (45) Wang, Y.; Wallin, J. M.; Ju, J.; Sensabaugh, G. F.; Mathies, R. A. Electrophoresis 1996, 17, 1485-1490. (46) Wang, Y.; Hung, S. H.; Linn, J. F.; Steiner, G.; Glazer, A. N.; Sidransky, D.; Mathies, R. A. Electrophoresis 1997, 18, 1742-1749. (47) Schmalzing, D.; Koutny, L.; Adourian, A.; Belgrader, P.; Matsudaira, P.; Ehrlich, D. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 10273-10278.

Figure 6. Resolution as a function of base position for the M13 T ladder separation presented in Figure 5. Resolution was calculated for isolated doublets that are one base apart using eq 1.

a result of using longer (7.0 vs 3.5 cm) and deeper (50 vs 8 µm) separation channels, the optimized twin-T injector, the LPA separation matrix, and elevated temperature. Recently Schmalzing et al.33 presented single-color sequencing separations that extended to 400 bases (R g 0.5) in 14 min on 11.5-cm-long, 40-µm-deep microfabricated channels, and they developed a theoretical model for the performance of microchannel sequencing separations. On the basis of this work, they concluded that sequencing to more than 400 bases on microfabricated devices much shorter than 10 cm in length is unlikely. Our one-color results are significantly better than this prediction, Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

571

Figure 7. Analyzed four-color M13 DNA sequencing traces from a CE chip. Separation was performed on a 7-cm-long channel with a 100-µm twin-T injector using 4% LPA as the separation medium at 40 °C. Separation was performed with a voltage of 160 V/cm, and the detector was 6.5 cm from the injector. Only 0.2 µg of DNA template was employed per reaction, and 1 µL of the final reaction solution (33%) was loaded on the chip. See Experimental Section for other conditions. This run was complete in under 20 min.

producing ∼500 bases (R g 0.5) in 10 min on 7-cm channels. There are several differences in experimental conditions that may explain these results. First, the sequencing separation quality is dependent on the preparation of the linear polyacrylamide. The LPA gels used here were prepared based on a recently optimized protocol,9 where the polymerization is performed in water with helium purging at 0 °C. This might have produced a higher molecular weight LPA that has better selectivity, especially for larger fragments.9,48 Second, the reduced separation voltage (150 vs 200 V/cm) used in this work is expected to be more effective for sieving larger fragments and will reduce the voltage-dependent component of the diffusion broadening.49 Third, our use of energytransfer fluorescent primers coupled with confocal detection is expected to improve the sequencing results because enhanced (48) Sunada, T. M.; Blanch, H. W. Electrophoresis 1997, 18, 2243-2254. (49) Luckey, J. A.; Norris, T. B.; Smith, L. M. J. Phys. Chem. 1993, 97, 30673075.

572 Analytical Chemistry, Vol. 71, No. 3, February 1, 1999

sensitivity allows us to avoid overloading-induced loss of resolution. Although the DNA concentrations were not stated by Schmalzing et al.,33 the DNA concentrations and amounts were more than 10-fold higher in our earlier DNA sequencing experiments.22 Some combination of these factors may contribute to the improved performance obtained here. A four-color separation of an M13 sequencing sample on a microfabricated channel is presented in Figure 7. This separation was performed on a 6.5-cm-long channel filled with 4% LPA and run at 40 °C. With a separation field of 160 V/cm, the entire 600bp separation is complete in 20 min. The matrix concentration was increased to 4% LPA and a smaller 100-µm twin-T injector was employed to ensure enhanced resolution of the four-color traces for better base-calling. In addition, the temperature and field strength were increased to reduce the separation time. The data in Figure 7 were analyzed using BaseFinder,37 as described in the Experimental Section. The script used in the analysis of

Figure 8. Display of the cumulative base-calling errors and percent accuracy as a function of position for the chip sequencing data in Figure 7.

the chip data included the following: primer peak deletion; baseline subtraction; spectral separation; two rounds of successive noise filtering and deconvolution to counteract zone-broadening effects; a final noise filtering; normalization of the signal intensities; histogram equalization to remove excessively high peaks; mobility shift correction; and base-calling. Cumulative accuracy and error count are presented in Figure 8. The cumulative errors are 0 at 450 bp, 3 at 500 bp, and 28 at 600 bp. The percent accuracy is 100% at 450 bp, 99.4% at 500 bp, and 95.3% at 600 bp. These results are dramatically better than the only other published four-color sequencing on microchannels which only extended to 150 bases.22 CONCLUSIONS The electrophoretic channel design and methods have been optimized to permit high-quality DNA sequencing on 4-in.-diameter microfabricated substrates. Separations of sequencing fragments on straight 6.5-7.0-cm-long channels resolve 500 bases in