Sanger DNA-Sequencing Reactions Performed in a Solid-Phase

Department of Biological Sciences, Louisiana State University, Baton Rouge, Louisiana 70803-1804. A miniaturized, solid-phase nanoreactor was develope...
0 downloads 0 Views 116KB Size
Anal. Chem. 1998, 70, 4036-4043

Sanger DNA-Sequencing Reactions Performed in a Solid-Phase Nanoreactor Directly Coupled to Capillary Gel Electrophoresis Steven A. Soper,* Daryl C. Williams, Yichuan Xu, Suzanne J. Lassiter, Yuling Zhang, and Sean M. Ford

Department of Chemistry, Louisiana State University, Baton Rouge, Louisiana 70803-1804 Richard C. Bruch

Department of Biological Sciences, Louisiana State University, Baton Rouge, Louisiana 70803-1804

A miniaturized, solid-phase nanoreactor was developed to prepare Sanger DNA-sequencing ladders which was directly interfaced to a capillary gel electrophoresis system. A biotinylated fragment of the rat brain actin gene (1 kbp) was amplified by PCR and attached to the interior wall of an (aminoalkyl)silane-derivatized fused-silica capillary tube via a biotin/streptavidin/biotin linkage. Coverage of the capillary wall with the biotinylated DNA averaged 77 ( 10%. Stability of the anchored template under pressure (33 nL/s) and electroosmotic flows (11.3 nL/ s) were favorable, requiring rinsing for >150 h to reduce the surface coverage by only 50%. In addition, the immobilized template was stable toward temperatures required for preparing sequencing ladders, even under cycling conditions. Standard Sanger dideoxynucleotide termination performed in a large-volume (∼8 µL) solidphase reactor using the thermally stable polymerase enzymes Taq and Vent and the polymerases T7 and Bst with off-line slab gel electrophoresis and autoradiographic detection indicated that acceptable fragment generation was achieved only in the case of the thermally stable polymerases. Banding was not apparent for T7 and Bst since all reagents were inserted into the column in a single plug at the beginning of the reaction. A small volume reactor (volume ∼62 nL) was then used to perform DNA polymerase reactions and was coupled directly to a capillary gel column for separation. The capillary reactor was placed inside a thermocycler to control the temperature during chain extension and was directly connected to the gel column via zero dead volume fused-silica connectors. The complementary DNA fragments generated (C-track only) in the reactor were denatured using heat and directly injected onto the gel-filled capillary for size separation with detection accomplished using nearIR laser-induced fluorescence. Extension and single-base separation resolution of the C-track, which was directly injected onto the gel column, was estimated to be >450 bases from the primer annealing site with plate numbers ranging from 1 × 106 to 2 × 106/m.

complementary to the DNA template typically via Sanger chaintermination protocols; (3) separation of these fragments by a gel electrophoretic technique; (4) detection of the fragments; (5) data analysis (base calling). The acceleration and/or automation of these steps, which is one of the primary missions of the Human Genome Project, would greatly facilitate all types of genetic analysis, especially when large-scale sequencing projects are being tackled. In addition, reductions in the cost of obtaining such data is also a primary goal of the Human Genome Initiative. The use of laser-induced fluorescence (LIF) detection of dyelabeled primers or didexoynucleotides in place of autoradiographic imaging has allowed an enormous decrease in the time needed to evaluate sequencing data and has reduced the number of lanes required to perform base calling.1-3 Various LIF-based detection and base-calling schemes have been integrated into a number of commercially available DNA-sequencing instruments. In fact, many of these devices can essentially provide automated separation, detection, and reconstruction of a particular DNA sequence. For example, the ABI 377 DNA sequencer, which employs a slab gel format and LIF detection using a CCD camera with base calling accomplished via a four-color approach, can produce sequencing data on the order of 100 000 bases/day. Several groups have also made advances in data throughput by increasing the speed of DNA separations using capillary gel (CGE)4-8 and, recently, microchip electrophoresis.9-11 The principle advantages associated with these electrophoretic formats is that the surface-to-volume ratio is much larger than in conventional

DNA sequencing basically involves five steps; (1) preparation of the chromosomal or template DNA; (2) generation of fragments

(1) Ansorge, W.; Sproat, B.; Stegemann, J.; Schwager, C.; Zenke, M. Nucleic Acids Res. 1987, 15, 4593-4602. (2) Smith, L. M.; Sander, 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. (3) Prober, J. M.; Trainor, G. L.; Dam, R. J.; Hobbs, F. W.; Robertson, C. W.; Zagursky, R. J.; Cocuzza, A. J.; Jensen, M. A.; Baumeister, K. Science 1987, 238, 336-341. (4) Drossman, H.; Luckey, J. A.; Kostichka, A.; D′Cuhna, J.; Smith, L. M. Anal. Chem. 1990, 62, 900-903. (5) Dovichi, N. J. In Handbook of Capillary Electrophoresis; Landers, J. P., Ed.; CRC Press: Boca Raton, FL, 1994; pp 369-387. (6) Cohen, A. S.; Smisek, D. L.; Keohavong, P. TrAC, Trends Anal. Chem. 1993, 12, 195-202. (7) Monnig, C. A.; Kennedy, R. T. Anal. Chem. 1994, 66, 280R-314R. (8) St. Clair, R. L. Anal. Chem. 1996, 68, 569R-586R. (9) Effenhauser, C. S.; Manz, A.; Widmer, H. M. Anal. Chem. 1993, 65, 26372642. (10) Wolley, A. T.; Mathies, R. A. Anal. Chem. 1995, 67, 3676-3680. (11) Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1996, 68, 720-723.

4036 Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

S0003-2700(98)00288-1 CCC: $15.00

© 1998 American Chemical Society Published on Web 08/29/1998

slab gel electrophoresis (SGE), and as a result, higher electric field strengths can be used to drive the separation, decreasing the time required to effectively separate the DNA. Improvements in resolution are also encountered when separations are performed in capillaries versus slab gels. Reports have detailed a nearly 3-fold improvement in resolution and a 25-fold increase in the speed of the separation in CGE compared to SGE due primarily to the ability of CGE to more effectively dissipate Joule heat, allowing operation at high electric field strengths.4 However, in both capillary gel and microchip electrophoresis techniques, the amount of sample required for analysis ranges in the nanoliter regime, which could potentially offer significant savings in the use of costly reagents required for preparing the sequencing fragments. While advances in separation, detection, and base calling have been numerous, they have outpaced the development of methodologies for preparing DNA-sequencing fragments in a volume more commensurate with the microcolumn separation techniques used to fractionate these DNAs. Standard DNA-sequencing protocols generate products in the microliter range, therefore, not exploiting the low (nL) sample requirements for capillary and microchip separations. Low consumable costs for any sequencing protocol are of critical importance, especially if it is to be financially feasible to sequence large genomes, such as the human genome. For example, to sequence the human genome, 75 million sequencing reactions are required (assuming a modest 400-base average read length per electrophoretic run and 10× redundancy typically required for shotgun sequencing strategies) for complete coverage. At approximately $5.00 per reaction, the cost of reagents alone would exceed $375 000 000 using conventional sample preparation methodologies. Solid-phase DNA-sequencing methods possess several advantages which could potentially facilitate their integration into a miniaturized DNA-sequencing system. Solid-phase reactors have been described for a number of biochemical applications such as enzymatic cleavage,12,13 oligonucleotide purification,14 and DNAsequencing reactions.15,16 DNA sequencing using a solid support is advantageous in that it affords an easy method by which the products can be purified, leading to higher efficiency separations due to the removal of excess template, primers, and salts.14,17 Solidphase sequencing protocols have typically employed the use of magnetic beads as the support media with species attached to the support via a streptavidin-biotin linkage.18-24 The streptavidin-biotin anchor is typically accomplished by attaching the (12) Nashabeth, W.; El Rassi, Z. J. Chromatogr. 1992, 596, 251-264. (13) Amankwa, L. N.; Kuhr, W. G. Anal. Chem. 1992, 64, 1610-1613. (14) Tong, X.; Smith, L. M. Anal. Chem. 1992, 64, 2672-2677. (15) Hultman, T.; Stahl, S.; Moks, T.; Uhlen, M.Nucleosides Nucleotides 1988, 7, 629-638. (16) Ohara, R.; Ohara, O. DNA Res. 1995, 2, 123-128. (17) Tong, X.; Smith, L. M. Sequencing Mapping 1993, 4, 151-162. (18) Stahl, S.; Hultman, T.; Olsson, A.; Moks, T.; Uhlen, M. Nucleic Acids Res. 1988, 16, 3025-3038. (19) Hultman, T.; Stahl, S.; Hornes, E.; Uhlen, M. Nucleic Acids Res. 1989, 17, 4937-4946. (20) Uhlen, M. Nature 1989, 340, 733-734. (21) Hultman, T.; Bergh, S.; Moks, T.; Uhlen, M. BioTechniques 1991, 10, 8493. (22) Haung, S.-C.; Swerdlow, H.; Caldwell, K. D. Anal. Biochem. 1994, 222, 441449. (23) Rolfs, A.; Weber, I. BioTechniques 1994, 17, 782-787. (24) Hawkins, T. L.; McKernan, K. J.; Jacotot, L. B.; MacKenzie, J. B.; Richardson, P. M.; Lander, E. S. Science 1997, 276, 1887-1889.

streptavidin protein directly to the solid support and allowing a biotin-modified molecule to bind to the immobilized streptavidin. The biotin-streptavidin linkage is a suitable choice for sequencing applications due to the strength of the couple (Kd ) 10-15 M-1) and, as such, can potentially withstand the large and rapid temperature changes necessary to perform sequencing reactions. While sequencing using magnetic beads as the support medium allows for simple purification of the reaction products, the system is not well suited for producing ultrasmall volumes of material. A solid-phase strategy has been described by Ulhen and co-workers which involves the immobilization of the DNA template onto an agarose support via a streptavidin-biotin linkage.25 While very effective in producing small volumes of sequencing fragments, this procedure does not easily provide a means for a direct (online) connection to a microseparation medium as required for an autonomous sequencing system. The work presented herein describes the operational characteristics and use of a miniaturized solid-phase DNA polymerase nanoreactor which provides a significant reduction in the amounts of reagents and other consumables used in the preparation of Sanger, dideoxynucleotide-terminated DNA-sequencing fragments. In addition, the nanoreactor can be directly coupled to a capillary gel column for DNA separation, providing on-line analysis of sequencing fragments in an automated fashion on a nanoliter scale. The reactor consists of a conventional fused-silica capillary tube (V ) 62 nL), in which biotin is covalently attached to the wall with the subsequent addition of streptavidin to provide an anchoring point for a biotinylated DNA template prepared using the polymerase chain reaction (PCR). Long-term stability of the immobilized template under constant-flow conditions and temperature variations will be reported. In addition, the preparation of sequencing fragments directly in the nanoreactor will also be demonstrated, with reagents needed to prepare these fragments pumped in and out of the reactor using electrokinetic pumping. To provide the necessary temperature conditions for denaturation, annealing, and chain extension, the nanoreactor was placed in a thermocycler which consisted of an aluminum block sandwiched between a resistively heated coil and Peltier device. Due to the low thermal mass associated with the capillary tubes and the thin aluminum block allowing rapid thermal equilibration, fast temperature transitions could be effectively utilized, reducing the time necessary to prepare the DNA sequencing ladders.26-29 EXPERIMENTAL SECTION Preparation of Biotinylated DNA Using PCR Amplification. A 900-bp fragment of the actin gene was obtained by PCR from rat brain RNA and cloned into a carrier vector, pCRII (3.9 kb) (Invitrogen, San Diego, CA) using previously published procedures.30 The DNA/pCRII construct was then amplified by PCR in a Perkin-Elmer 2400 series thermocycler. The PCR (25) Stahl, S.; Hultman, T.; Olsson, A.; Moks, T.; Uhlen, M. Nucleic Acids Res. 1988, 16, 3025-3038. (26) Wittwer, C. T.; Garling, D. J.; Hillyard, D. R. Nucleic Acids Res. 1989, 17, 4353-4357. (27) Wittwer, C. T.; Fillmore, G. C.; Garling, D. J. Anal. Biochem. 1990, 186, 328-331. (28) Wittwer, C. T.; Garling, D. J. BioTechniques 1991, 10, 76-83. (29) Swerdlow, H.; Dew-Jager, K.; Gesteland, R. F. BioTechniques 1993, 15, 512519. (30) Bruch, R.; Medler, K. NeuroReport 1997, 7, 2941-2944.

Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

4037

mixture contained 1 µL of pCRII vector, 10 µL of 1× PCR buffer (20 mM Tris-HCl, pH 8.4, 50 mM KCl), 2 µL of dNTPs, 85 µL of double-distilled water (ddH2O), 1 µL of SP6 biotinylated forward primer (1 µM), and 1 µL of T7 reverse primer (1 µM). The terminal adenosine on the SP6 primer was labeled with a biotin molecule at the 6 position on the purine ring of the base, which allowed 32P labeling of the primer at the γ position of the triphosphate group attached to the 5′ position of the deoxyribose sugar. PCR was performed under “hot start” conditions, which entailed addition of 1 µL of the Thermus aquaticus (Taq) DNA polymerase (Gibco-BRL, Gaithersburg, MD) after the reaction temperature reached 80 °C, which ensured high fidelity during DNA amplification. Twenty-five PCR cycles were performed using the following program: (1) denature dsDNA at 94 °C for 45 s; (2) anneal primer at 55 °C for 30 s and; (3) extend primer at 72 °C for 90 s. The reaction products were separated by agarose gel electrophoresis (0.8% agrose, 1 µM EtBr, 25 V/cm) visualized under a UV lamp and excised from the gel matrix using a scalpel. The DNA was removed from the agarose using a gel removal centrifugation apparatus obtained from Amicon (Beverly, MA) and stored at -20 °C. Radioactive Labeling of Biotinylated Actin Fragment. The amplicon, when required, was radiolabeled at the 5′ terminus with [γ-32P]dATP (DuPont NEN, Wilmington, DE) before immobilization. DNA (2 pmol) was suspended in 29 µL of ddH2O and 5 µL of 10× kinase buffer. To the DNA-buffer solution, 15 µL of [γ-32P]dATP (3000 Ci/mmol, 10 mCi/mL) and 1 µL (10 units/µL) of the T4 polynucleotide kinase labeling enzyme (Gibco-BRL) were then added. The reaction mixture was incubated at 37 °C for 30 min. The action of the kinase enzyme was halted by the addition of 2 µL of 0.5 M EDTA and heating of the mixture to 65 °C for 2 min. Unincorporated dATP was removed by performing a selective precipitation of labeled DNA of >20 bp in length, which was accomplished by the addition of 25 µL of 7.5 M ammonium acetate and 150 µL of cold (4 °C) ethanol. The solution was allowed to sit 30 min at -20 °C and then centrifuged at 10 000 rpm for 20 min at -20 °C to form a DNA pellet. The supernatant liquid was decanted after centrifugation was complete. The efficiency of 32P labeling was determined by an assay using a DNA binding cellulose membrane. Labeling efficiency averaged 40%, which compared favorably to the 50% labeling efficiency reported by the product literature. Immobilization of DNA Template to Reactor Wall. The DNA was immobilized to the wall of a fused-silica capillary tube using a biotin-streptavidin-biotin system by following procedures similar to those outlined by Kuhr and co-workers, but with slight modifications to immobilize dsDNAs instead of trypsin proteins.13,31 Fused-silica capillary tubes (20-, 50-, or 100-µm i.d., 360-µm o.d.) were cut into 50-cm lengths and rinsed successively with 1N NaOH, ddH2O, and 1 N HCl for 10 min each using a vacuum. The capillary was purged with air (10 min) and oven-baked at 80 °C for 10 min to remove any residual water. The capillary was then filled with a 4% solution of (3-aminopropyl)triethoxysilane (APTS) in acetone, which was purchased from Sigma Chemical (St. Louis, MO), for 30 min, air-purged for 5 min, and finally incubated for 24 h at 45 °C. After incubation, the tube was filled with a bicarbonate solution (50 mM, pH 8.3) containing 5.0 mg/

mL NHS-LC-biotin (Sigma) for 4 h at room temperature. Following biotinylation, the reactor was exposed to a 4.0 mg/mL solution of streptavidin (Sigma) in 50 mM sodium phosphate buffer (pH 7.3). The streptavidin solution was allowed to incubate for several hours at 4 °C followed by the removal of any free streptavidin by rinsing the capillary tube with ddH2O and stored in a refrigerator until required for use. The column, when required for use, was cut to the desired length and filled with a 1 µM solution of the biotinylated template and incubated at 4 °C for ∼30 min. In some cases, biotinylated single-stranded (ss) DNAs were immobilized to the wall of the nanoreactor by quickly heating the doublestranded (ds) PCR product to 95 °C for several minutes followed by rapid cooling (5 °C) and immediately inserting the solution into the reactor. Excess template was removed by rinsing with ddH2O and then the capillary was inserted into the appropriate apparatus for evaluation and/or sequencing experiments. Scintillation measurements of the 32P-labeled template in the reactor were performed by submerging the reactor into a scintillation liquid (3 mM p-terphenyl in toluene). Prior to inserting the reactor into the scintillation cocktail, the reactor was capped on both ends with quartz capillary caps obtained from Polymicro. Measurements were performed on an LS6000IC series scintillation counter (Beckman Instruments, Fullerton, CA). DNA Thermocycler. The nanoreactor was inserted into an aluminum block (9 cm × 0.75 cm) which possessed guide holes that were ∼400 µm in diameter (see Figure 1A) and could accommodate up to 12 capillaries. A Peltier plate/heat sink, which was attached directly to the aluminum block, was used to cool the temperature of the block while heat was generated via resistive heating of a 100-W filament pad which was also attached to the aluminum block. The temperature was monitored using a thermocouple and an in-house-constructed active feedback circuit was used to control the temperature to within (1 °C. The heating/cooling rate of the thermocycler was determined to be ∼10 °C/min. During preparation of the sequencing fragments, the reactor was flanked on each side by two 20-cm lengths of fused-silica tubing which were linked to the reactor via zero dead volume glass capillary connectors (MicroQuartz, Phoenix, AZ). The flanking capillaries were coated with a 2% linear polyacrylamide in order to minimize the electroosmotic flow, which allowed the reagents to be delivered into the reaction vessel from the cathodic end of the system.32 Following primer extension, the gel-filled separation column was attached directly to the reactor using the connector with the removal of one of the flanking LPAcoated tubes (see Figure 1A). In Figure 1B is shown an optical micrograph of the interface between a 20-µm-i.d. capillary nanoreactor and a 75-µm-i.d. gel-filled capillary column. The sequencing fragments generated in the nanoreactor were then injected into the gel column for separation by applying a negative potential at the cathodic end of the LPA-coated flanking capillary (see Figure 2A). Solid-Phase Nanoscale Sequencing Reactions. The extension of the immobilized ssDNA was performed directly inside the nanoreactor. The reactor was filled with a 1 × 10-6 M solution of IRD800-labeled T7 primer (Li-COR, Lincoln, NE) and the temperature raised to 95 °C followed by cooling to 25 °C to allow annealing of dye primer to the surface-immobilized template. The

(31) Amankwa, L. N.; Kuhr, W. G. Anal. Chem. 1993, 65, 2693-2697.

(32) Herjten, S. J. Chromatogr. 1983, 270, 1-6.

4038 Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

Figure 2. Normalized surface coverage of immobilized biotinylated ssDNAs to the wall of the capillary nanoreactor as a function of gravity flow (A) and electropumping (B). The capillary nanoreactor used in these experiments was 4 cm in length and 100-µm i.d. The buffer (TBE) flow was allowed to run continuously with the capillary reactor removed from the flow system at fixed time intervals and inserted into the scintillation vial for determining the activity of the reactor (amount of immobilized DNA). The data plotted are the average of five runs with the error bars representing the standard deviation in the measurement.

Figure 1. Schematic diagram of CE system with nanoreactor and capillary gel column (A). The Al block consisted of 12 guide holes to allow accommodation of up to 12 capillaries possessing a 365-µm o.d. The nanoreactor was connected to the flanking LPA capillary, and the gel column using zero dead volume unions. The flanking capillary was coated with a 2% linear polyacrylamide to reduce the magnitude of the electroosmotic flow, was 20 cm in length, and possessed an inner diameter of 75 µm. The gel column was a 3% T/3% C cross-linked polyacrylamide gel with 7 M urea as the denaturant. The inner diameter of this column was 75 µm with a total length of 70 cm, 40 cm to the detection window. (B) An optical micrograph of the union between a 20-µm-i.d. nanoreactor and the 75-µm-i.d. capillary gel column. One end of each capillary forming the junction was cleaved and then polished on 3- and 0.3-µm polishing paper in water prior to insertion into the union.

primer solution also contained 0.01 unit of a DNA polymerase (Taq, Vent, Bst or T7), the four deoxynucleotides, and a single dideoxynucleotide (ddCTP for CGE or ddGTP for SGE). To build a sufficient population of extension fragments to aid in detection for the CGE experiments, the temperature of the reactor was cycled 10 times with each cycle consisting of 92 °C for 30 s, 55 °C for 30 s, and 72 °C for 90 s. For the SGE experiments, only one temperature cycle was used in order to get a reasonable comparison of the extension efficiency of the thermally stable polymerases to Bst and T7. The matrix used for separation in the capillary format was a 3% T/3% C polyacrylamide gel containing 7 M urea as the denaturant. The field strength was set to 200 V/cm during electrophoresis and the running buffer consisted of 1× TBE (pH 8.6). The capillary gel column possessed a total length of 70 cm, with an effective length of 40 cm (injection to detection). LIF/CGE System. The DNA fragments were detected using a custom-built near-IR laser-induced fluorescence apparatus which

was similar in design to that repored earlier by our group.33 Basically, the system consisted of a GaAlAs diode laser producing 20 mW of laser light at 780 nm which was focused onto the capillary using a singlet lens to a spot size of ∼10 µm (1/e2). The fluorescence was collected with a 60× microscope objective (NA ) 0.85) and imaged onto a spatial filter with a slit width set to 0.6 mm which produced a sampling volume in the gel column of ∼0.78 pL. After spatial filtering, the emission was spectrally filtered with an eight-cavity band-pass filter (CWL ) 820 nm; HBW ) 30 nm; Omega Optical, Brattleboro, VT) and focused onto the face of the photodetector using a 20× microscope objective. The photodetector was a single-photon avalanche diode (SPAD; EG&G Optoelectronics, Vadrieulle, Canada) that was passively quenched and interfaced to a PC which contained a counting board (CT101, Cyber Research, Bradford, CT) with the data acquisition software written using LabTech Notebook7 (Cyber Research). The mass limit of detection for this diode-based system has been reported to be in the 10-60-zmol range for near-IR dye-labeled amino acids separated by capillary zone electrophoresis.33 RESULTS AND DISCUSSION The efficiency of DNA immobilization onto the capillary wall was initially studied using a large-diameter nanoreactor (i.d. ) 100 µm; L ) 5 cm) which generated sufficient product volumes to allow off-line analysis using scintillation counting. The degree of surface coverage was determined by incorporating a 32P label into the template, which contained the biotin linkage at the 5′ terminus which was bound to the wall-immobilized strepavidin species. The amount of activity generated from the reactor was then measured by submerging the entire reactor into a vial (33) Legendre, B. L.; Moberg, D. L.; Williams, D. C.; Soper, S. A. J. Chromatogr., A 1997, 779, 185-194.

Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

4039

Table 1. Calculation of Immobilization Efficiency of Biotinylated Double-Stranded DNA to Biotin/ Streptavidin-Modified Capillary Nanoreactora reactor surface area (m2) total volume (L) streptavidin cross-sectional area (A2) number of binding sites moles of immoblized DNA template specific activity of 32 P (Ci/molecule)b conversion factor for Ci to dpm (dpm/Ci) expected scintillation intensity (dpm) measured scintillation intensity (cpm) immobilization efficiency (surface coverage)

1.57 × 10-5 314 × 10-9 2000 7.85 × 1011 1.04 × 10-12 7.23 × 10-20 2.22 × 1012 125 997 90 115 71.5%

a The reactor used for scintillation measurements was 4 cm in length and possessed an inner diameter of 100 µm. b The specific activity was corrected for the half-life and age of the 32P label.

containing a scintillation cocktail and monitoring the activity using the scintillation counter. The theoretical scintillation intensity for complete monolayer coverage (see Table 1 for summary of results) was determined by calculating the surface area available for immobilization inside the reactor and assuming that the degree of coverage was restricted by the area occupied by the streptavidin protein (2000 Å2).13,31 The coupling ratio of biotinylated DNA to the streptavidin anchor was assumed to be 1:1 due to steric interactions that would result when multiple 1-kbp templates would bind to a single immobilization site. Based on these assumptions, the efficiency of template immobilization to the interior reactor wall was determined to be 77% ((10%). From the physical dimensions of this reactor (100 µm i.d. × 4-cm length), the average amount of DNA immobilized in this system was ∼1 × 10-12 mol, yielding an effective DNA concentration of 3.3 × 10-6 M in this 314-nL volume. The results depicted in Figure 2B show the effects of longterm rinsing using electrokinetic pumping (11.3 nL/s) with 1× TBE buffer solution (pH 8.6) through the reactor on the immobilized DNA. The stability of the template anchored to the capillary wall under electrokinetic flow was favorable, requiring >150 h of continuous rinsing to reduce the surface coverage by only 50%. Removal of the immobilized DNA from the reactor wall was slightly accelerated over the rate calculated using the known biotin-strepavidin dissociation constant (Kd ) 10-15 M-1) and this volumetric flow rate. After 250 h of continuous rinsing, the surface coverage was reduced to 38% whereas, on the basis of the known dissociation constant for the biotin-streptavidin couple and this volumetric flow rate, one would predict a surface coverage of 84% after this same period of time. To determine whether the accelerated surface loss was due to the influence of the applied potential across the reactor during electropumping, experiments using gravity-driven flow were also performed. The loss rate using gravity flow is shown in Figure 2A. These results showed an increased depletion rate versus the theoretical prediction but had a smaller deviation from the predicted loss compared to electrokinetic pumping. At this volumetric flow rate, the predicted surface coverage would be 38% after 250 h of rinsing and the observed value was 16%. The higher loss rate encountered in both experiments may be explained by the alkaline nature of the running buffer (1× TBE pH 8.6), which could expedite denaturation of the streptavidin protein or loss of biotin from the wall of the capillary nanoreactor due to hydrolysis of the siloxane bonds. 4040 Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

Figure 3. Reactivation of the capillary nanoreactor after long-term rinsing with TBE buffer using gravity flow. The arrow indicates the point at which the nanoreactor was reactivated. The reactor was incubated with streptavidin followed by the addition of biotinylated DNA to reactivate the capillary nanoreactor. See Figure 2 for experimental details.

The potential applied to the reactor may also influence the stability of the biotin-strepavidin anchor under electropumping conditions since these materials are close to the wall where the greatest potential drop occurs. Figure 3 shows the efficiency of reactivation of the nanoreactor by reimmobilization of biotinylated 32P-labeled DNA. After extensive electrokinetic rinsing, streptavidin was added to the nanoreactor followed by addition of biotinylated [32P]DNA template. After incubation at 4 °C, the percent coverage of reimmobilized DNA was found to be nearly restored to its initial value. Addition of biotinylated [32P]DNA without first replenishing the streptavidin layer resulted in negligible increases in reactor surface coverage by the DNA. These results indicate that the loss of immobilized DNA is due primarily to the loss of anchored streptavidin and not the loss of biotin attached directly to the wall of the capillary tube. The results also show that the nanoreactor can be reactivated for subsequent rounds of sequencing by simply adding fresh streptavidin followed by the addition of the target biotinylated DNA without the need for replacement of the entire capillary reactor tube. We also found that we could effectively remove all of the immobilized streptavidin and/or DNA by filling the reactor with a 98% formamide solution and heating the reactor to 80 °C. After this treatment, streptavidin could be reimmobilized to the wall of the nanoreactor, restoring its activity. To determine the conditions necessary to denature wallimmobilized dsDNAs after primer extension, we next focused on the use of heat and/or strong base to denature the doublestranded constructs. In these experiments, the PCR product was not subjected to heat denaturation prior to wall immobilization. Figure 4 shows the result of raising the reactor temperature to 95 °C for 5 min, followed by cooling the reactor to 5 °C for 3 min, and finally rinsing (10 nL/s) the reactor. A 43% loss of scintillation intensity was seen from the 32P-labeled template, which we attributed to the removal of the complementary strand. Since both strands are labeled with 32P, the loss of one stand would be expected to yield a 50% reduction in scintillation intensity if the complement was completely lost after denaturation, consistent with our data. We also found that subsequent heating, cooling,

Figure 4. Relative activity of the capillary nanoreactor as a function of denaturing cycles. The denaturing cycle consisted of raising the reactor temperature to 95 °C for 2 min followed by a short rinsing step using gravity flow (15 nL/s) to remove the complement. See Figure 2 for experimental details. In this case, the immobilized DNA was double-stranded and consisted of the original PCR product.

and rinsing cycles produced only minimal reductions in scintillation signal. We also attempted to denature the template using 0.1 N NaOH in the absence of heat but observed that after two rinses an 81% reduction in scintillation signal occurred. The large loss of signal intensity after NaOH rinsing indicated a loss of both the complement and the tethered ssDNA using this denaturing step. Attempts to reactivate the reactor by adding streptavidin and DNA after NaOH treatment were unsuccessful, indicating a loss of the covalently anchored biotin to the capillary wall. An evaluation of several different enzymes for use in the nanoreaction vessel was then performed using the modified T7 and Bst enzymes as well as two thermally stable polymerase enzymes, Taq and Vent. The Taq and Vent sequencing protocols were most amenable to the nanoreactor format since it allowed the addition of all necessary reagents as a single plug into the reactor, while the T7 and Bst protocols were designed to operate in a static reaction vessel requiring the addition of selected reagents at different time intervals during the polymerase reaction. To be adaptable to the nanoreactor format, all reagents required for chain extension must be added into the reactor simultaneously. The stated protocols for the Bst and T7 enzymes suggested the addition of the polymerase enzyme, dNTPs, and ddNTPs at different intervals during the reaction. In the modified protocol, enzymes, dNTPs, and ddNTPs were added concurrently. In all cases, the polymerase reactions were performed at 37, 65, 72, and 72 °C for T7, Bst, Taq, and Vent, respectively, using only a single temperature cycle. Reactors with 100-µm i.d. and lengths of 100 cm (total volume 7.85 µL) were used in this experiment in order to provide a manageable amount of material for off-line analysis using slab gel electrophoresis with autoradiographic detection. The results of the sequencing reactions performed in these largescale reactors are shown in Figure 5 for the G-track only. As can be seen, the absence of banding was observed in the case of T7 (lane 3) and Bst (lane 4) polymerase enzymes using these reaction conditions. In the case of the Taq and Vent polymerase enzymes (lanes 1 and 2, respectively), banding was clearly evident with the pattern matching that of the control (lane 5). The absence of bands in the case of T7 and Bst enzymes most likely resulted from

Figure 5. Slab gel electropherogram of sequencing fragments (Gtrack only) produced in large-scale nanoreactors using solid-phase sequencing and various polymerase enzymes. The gel was a 5% T/3% C cross-linked polyacrylamide gel. The extension products were labeled with a 32P label and the bands visualized using autoradiography after vertical slab gel electrophoresis. Lanes (1) Taq polymerase; (2) Vent polymerase; (3) T7 polymerase; (4) Bst polymerase enzyme; (5) Taq polymerase enzyme with the sequencing reaction performed in a standard microtube using the prescribed protocol for this polymerase enzyme. In all cases, only a single temperature cycle was used for the reactions.

the low population of terminated fragments produced during polymerization due to significant alterations in the reaction protocol. While the Taq and Vent polymeases were selected for this series of experiments, it would be expected that any thermally stable polymerase could be used in this application, since these enzymes allow the addition of all necessary reagents to the reaction in a single step. In all subsequent sequencing experiments, the Taq polymerase enzyme was employed. Integration of the nanoreactor directly to a capillary gel column for sequencing requires careful consideration to zone broadening due to extracolumn effects because of the stringent requirements on separation efficiency in sequencing applications in order to achieve single-base resolution. In the present case, the finite volume of the reactor (injection volume, Vinj, for a 20-µm-i.d. × 20 cm column, Vinj ) 62 nL) and the dead volume associated with the connector can potentially result in the reduction in the plate numbers. To assess the contributions of these two variances to the total variance, experiments were carried out to evaluate the relative contribution from the connector (σ2con) and reactor volume Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

4041

(σ2inj). This extra column variance (σ2xc) can be calculated from the following expression, assuming that the zone variance from the finite injection volume could be calculated from l2/12,34 where l is the length of the injection plug;

σ2xc )

[

]

µ2gel r4nr 2 l nr + σ2con 12µ2fs r4sc

(1)

where µgel is the electrophoretic mobility of the oligonucleotide in the gel column, µfs is the free solution mobility of the oligonucleotide (4.0 × 10-4 cm2/V‚s),35 rnr is the radius of the nanoreactor, rsc is the radius of the separation capillary, and lnr is the length of the nanoreactor. The term in the brackets represents the zone variance arising from the finite injection volume (σ2inj) and assumes that complete radial diffusion occurs when injecting into a column with a larger radius. As can be seen from this expression, to minimize σ2inj, it is necessary to use a highly cross-linked gel or high % T linear gel (small µgel) and also a small-diameter nanoreactor and larger diameter separation column. It should also be noticed from eq 1 that the zone variance from σ2inj will be highly dependent upon the number of bases comprising the oligonucleotide, since the longer oligonucleotides have a smaller µgel. In the present case, µgel < µfs and rnr < rsc, and therefore, zone compression should result when the contents of the nanoreactor are injected directly into the gel column. It should also be noted that since the separation column used 7 M urea as the denaturant, it is expected that zone compression should also result at the head of the column due to a conductivity difference between the column and the solution contained within the nanoreactor. The zone variance from the connector was determined by loading the nanoreactor with dye-labeled primer and interfacing it via the glass connector to a gel-filled column followed by electrokinetically injecting the dye primer onto the gel column. The electropherogram generated using this arrangement was compared to one generated after performing a direct injection of near-IR dye primer onto the CGE column under similar electrokinetic injection conditions. In both cases, the injection volume was kept constant, which was accomplished, in the case of the connector experiment, by removing the separation capillary and placing it in 1× TBE once the appropriate injection volume had been reached. The plate numbers (data not shown) were found to be 2.49 × 105 plates for direct injection and 2.45 × 105 plates when injection occurred across the zero dead volume connector. Since the only additional contribution to the zone variance was that arising from the connector between these two cases, the difference in the total zone variance calculated from the electropherograms yielded σ2con, which was determined to be 1.5 × 10-4 cm2. Even though the connector is stated as possessing zero dead volume, the loss in efficiency in this case is most likely due to the inability to polish the capillary ends correctly, producing a void volume at the interface. The nanoreactor used in our sequencing experiments was selected to have a length of 20 cm (lnr) and radius of 10 µm (rnr), (34) Sternberg, J. C. Adv. Chromatgr. 1966, 2, 205-211. (35) Grossman, P. D.; Colburn, J. Capillary Electrophoresis: Theory and Practice; Academic Press: New York, 1992.

4042 Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

yielding a total volume of 62 nL. From previous measurements, we found that the mobility of a 150-mer in a 3% T/3% C polyacrylamide gel with a near-IR primer was 6.1 × 10-5cm2/V‚ s.36 Using these numbers as well as the free solution mobility of an oligonucleotide with >20 bases yielded a value of 3.4 × 10-3 cm2 for σ2nr. The total variance (σ2tot) for this 150-mer near-IR labeled single-stranded oligonucleotide using direct injection onto a cross-linked gel column was found to be 1.3 × 10-3 cm2 (Htot ) 2.5 × 10-5 cm, where H is the height equivalent of a theoretical plate) and the plate number was 2.0 × 106.36 Adding the zone variance contributions from the connector and the injection volume produced by the nanoreactor would give an estimate of the efficiency of this injection system, which yielded a value of 4.8 × 10-3 cm2 (Htot ) 8.6 × 10-5 cm) and a plate number of 5.8 × 105. Therefore, one could expect an approximate 65% reduction in separation efficiency using the injection system as configured above compared to direct injection. In Figure 6 is shown the capillary electropherogram for C-terminated fragments generated in the nanoreactor with subsequent separation in a 3% T/3% C cross-linked polyacrylamide gel column (i.d. ) 75 µm). From this electropherogram, we estimated an effective read length of ∼450 bases, which was determined by peak counting and then multiplying by 4 to a point in the electropherogram where resolution permitted single-base identification (R > 0.7). However, it should be noted that since only the C-track was analyzed, the stated read length will most likely be reduced when the system is transferred to a single-lane, four-base protocol. Calculation of the plate number for the nearIR dye-labeled 160-mer depicted in the figure inset yielded a value of 2.05 × 106 plates compared to 4.51 × 106 plates for direct injection (data not shown), in fair agreement with the efficiency loss calculated above. An interesting feature of this injection format is the fact that biases due to electrokinetic injection are absent, since in the present case we are injecting onto the gel column a fixed volume, which is defined by the physical dimensions of the capillary nanoreactor. Therefore, any differences in peak amplitude between the early- and latter-migrating components arises from the efficiency of the polymerase enzyme during chain extension. CONCLUSIONS We have demonstrated the ability to effectively produce sequencing ladders on a nanoliter scale, a volume that exploits the small sample requirements associated with the microcolumn separation techniques. The volume used in the present system (62 nL) represents an approximate 300-fold reduction in sample size typically used in Sanger chain-termination protocols. The net result is a significant reduction in the amount of consumables required for generation of sequencing ladders. A cost analysis for large-scale DNA-sequencing applications using standard protocols and the DNA nanoreactor system indicated that reagent costs could run as high as $12 750 per day when performing 2500 sequencing reactions (for an average read length of 400 bases per electrophoretic run, this corresponds to 1 × 106 bases of raw sequencing data per day) compared to only $40.65 per day when using the nanoreactor described in this work possessing a volume of 62 nL. Another attractive feature of this system is the ability (36) Williams, D. C.; Soper, S. A. Anal. Chem. 1995, 67, 4358-4365.

Figure 6. Capillary gel electropherograms of C-terminated Sanger sequencing fragments produced in the nanoreactor possessing a volume of ∼62 nL. The fluorescence of the labeling dye was excited with 4.5 mW of laser power at 780 nm. The column length, from injector to detector, was 40 cm with a total length of 70 cm. The electrophoresis was carried out using a field strength of 200 V/cm. The graph inset shows an expanded view with the 160-mer with the plate numbers calculated for this peak. In the figure, only 5000 s of data is shown.

to perform solid-phase sequencing in ultrasmall volumes with its incumbent advantages, such as removal of excess primer, salts, and dNTPs prior to separation, improving banding in the electropherogram. In addition, since the anchor used in the present system is stable toward typical sequencing conditions, one can subject the immobilized DNA template to multiple sequencing rounds making it amenable to such techniques as primer walking and cycle sequencing. Our preliminary data have indicated that we can subject the immobilized template to four to five rounds of sequencing before a significant reduction in signal is seen. However, for a four-base, single-lane format, this methodology will require the use of dye-labeled terminators (dideoxynucleotides) and not dye-labeled primers as used here. The use of dye-labeled primers would require four nanoreactors with subsequent pooling of the reaction products prior to separation. Work in our laboratory is currently focused on developing near-IR dyelabeled terminators for use in the nanoreactor system. Another advantage of the present system is its ability to operate in an automated fashion. Recently several research groups have discussed automated systems for the preparation of sequencing ladders using HPLC hardware and the associated pumps.37,38 In these examples, the volumes required for sequence analysis were on the same order as those typically used in slab gel applications and, also, the ancillary equipment can be somewhat prohibitive. In our system, ultrasmall volumes are utilized and, in addition, no mechanical pumps are required, with fluid pumping accomplished by simple electropumping. Therefore, by using (37) Tan, H.; Yeung, E. S. Anal. Chem. 1997, 69, 664-674. (38) Swerdlow, H.; Jones, B. J.; Wittwer, C. J. Anal. Chem. 1997, 69, 848-855.

electropumping to fill the nanoreactor with the appropriate amount of sequencing reagents (primer, sequenase, ddNTP, dNTPs, buffers, and salts) required for DNA polymerization, the reactor inlet can be inserted into a microtiter well containing these reagents and an electric field applied to fill the reactor. In this way, only a few nanoliters of materials are consumed, and since the template is not added to the mix, the remaining contents of the microtiter well can be used for subsequent sequencing rounds without having to discard them after one use. Another advantage associated with this system is the fact that biases due to electrokinetic injection are absent. As such, alterations in the dNTP/ddNTP ratios during DNA polymerization are not required and standard mixes included in the sequencing kit can be utilized. In addition, since we are immobilizing the target DNAs onto the wall of the capillary, long DNAs (>10 kbp) could be immobilized allowing one to implement primer walking strategies in this format, simplifying sequence reconstruction. ACKNOWLEDGMENT Financial support of this work by the NIH (Grant HG01499) is greatly appreciated. The authors also thank Dr. Huigen Dong for helpful discussions during the course of this work. Assistance with the construction of the thermal cycler by Don Patterson is also deeply appreciated.

Received for review March 12, 1998. Accepted July 28, 1998. AC980288Z

Analytical Chemistry, Vol. 70, No. 19, October 1, 1998

4043