On-line integration of PCR and cycle sequencing in capillaries: from human genomic DNA directly to called bases

Abstract

A fully integrated system has been developed for genetic analysis based on direct sequencing of polymerase chain reaction (PCR) products. The instrument is based on a serially connected fused-silica capillary assembly. The technique involves the use of microreactors for small-volume PCR and for dye-terminator cycle-sequencing reaction, purification of the sequencing fragments, and separation of the purified DNA ladder. Four modifications to the normal PCR protocol allow the elimination of post-reaction purification. The use of capillaries as reaction vessels significantly reduced the required reaction time. True reduction in reagent cost is achieved by a novel sample preparation procedure where nanoliter volumes of templates and sequencing reaction reagent are mixed using a micro- syringe pump. The remaining stock solution of sequencing reaction reagent can be reused without contamination. The performance of the whole system is demonstrated by one-step sequencing of a specific 257-bp region in human chromosome DNA. Base calling for the smaller fragments is limited only by the resolving power of the gel. The system is simple, reliable and fast. The entire process from PCR to DNA separation is completed in ∼4 h. Feasibilities for development of a fully automated sequencing system in the high-throughput format and future adaptation of this concept to a microchip are discussed.

INTRODUCTION

Since its inception, the Human Genome Project (HGP) (1,2) has called for developing technologies for cost-effective, high-speed and high-throughput DNA sequencing. Highly multiplexed capillary electrophoresis (CE), especially when combined with replaceable linear polymer solutions for DNA sequencing (3–5), has been demonstrated by many research groups (6–8). The attractive features of such a system are its speed, throughput, resolution, reliability and sensitivity compared with those of traditional slab gel electrophoresis instruments. Commercial multiplexed CE instruments (9–11) have accelerated the HGP significantly such that sequencing of the human genome is now essentially complete. However, the need for DNA sequencing has actually increased as we start to use the information in biological and medical research (12–28).

Currently, DNA sequencing efforts are primarily based on cycle-sequencing chemistry (29,30), which allows a linear increase in product amount with cycle number. This implies that one must start with an adequate concentration of templates (e.g. 100–500 ng/µl of dsDNA for ABI’s cycle-sequencing kit). Therefore, biological amplification by cloning into yeast or bacteria generally precedes the sequencing reaction. Additional steps include template preparation, template purification, reaction preparation, thermal cycling, sample purification, sample loading and electrophoretic separation. Since tens of microliters are still required for sample preparation, subsequent purification, and loading, CE has not lowered the reagent cost substantially. Also, the total analysis time has not improved due to the slow sample preparation tasks.

In polymerase chain reaction (PCR) (26), a specific target sequence on a large DNA molecule is amplified exponentially by an enzymatic reaction. This can serve as an alternative to biological amplification for producing adequate DNA templates. The ability to sequence DNA templates generated by the PCR has gained increasing attention (31–39). Clearly, an integrated sample preparation/DNA sequencing system that also allows for the direct sequencing of PCR products will be very attractive. Previously, we accomplished on-line integration of sequencing reaction, purification, injection and gel separation for sequencing of M13mp18 and pGEM using conventional fused-silica capillaries for CE (40,41). The integration strategy was based on the use of capillaries as a container for the reaction and as an on-line electrophoretic purification column. The integrated system provides significant reductions in the required volume and the total analysis time. If the system also covers template preparation, subsequent template purification will be a major bottleneck for the achievement of full integration. PCR makes use of two primers, dNTPs and DNA polymerase to produce multiple copies of a specific DNA sequence. When complete, most of the dNTPs and primers remain, and will interfere with cycle sequencing, which also makes use of primers and nucleotides. Therefore, the PCR product needs to be purified from the residual primers and unused dNTPs before it is compatible with the sequencing phase (42–46). PCR product clean-up requires spin column, sample transfer and centrifugation steps that are not readily amenable to automation or miniaturization.

It has been found that PCR product clean-up is not always necessary when sequencing with fluorescent dye-primers (34–37). This is usually carried out by setting up the PCR with limited amounts of primers and dNTPs, so that most of these are exhausted during amplification. Another way to carry out direct sequencing of PCR products without purification is to dilute the completed PCR mixture. This reduces the amount of primers and dNTPs carried into the sequencing reaction, as well as the amount of the original template. Sequencing unpurified PCR products with fluorescent dye-terminators is even more problematic than those with fluorescent dye-primers. As far as we know, direct sequencing of PCR products without clean-up using dye-terminators is limited to the study performed by Van Den Boom et al. (39), wherein fairly dilute PCR products [5% (v/v)] were used in the sequencing reaction. Dilution constitutes an additional step and is not broadly applicable since the large amount of PCR products that is required (prior to dilution) cannot always be obtained.

The work presented herein describes a system for direct sequencing of PCR products where the entire process of amplification of a specific target by PCR, introduction of reagents, cycle-sequencing reaction, purification of the sequencing ladder, injection and size separation is integrated in a single instrument using a serially connected multi-capillary assembly. Simple modifications to the normal PCR protocol allow use of the crude PCR product without purification for dye-terminator cycle sequencing. Novel sample manipulation using capillaries and microsyringe pumps is employed for true savings in reagents consumed. The performance of the whole system is demonstrated by the sequencing of a specific target fragment on the Y chromosome of human DNA. We also discuss the feasibilities and challenges for future development of fully automated and highly multiplexed instruments and implementation on a microchip for genomic sequencing.

MATERIALS AND METHODS

Reagents, buffers and separation matrix

Fuming hydrochloric acid (HCl), potassium chloride (KCl), magnesium chloride hexahydrate (MgCl₂·6H₂O) and tris (hydroxymethyl)aminomethane (Tris) were obtained from Fisher Scientific (Fair Lawn, NJ). Ten times PCR buffer (500 mM Tris–HCl, pH 8.3 with 30 mM MgCl₂ and 2.5 mg/ml bovine serum albumin, BSA), 10× BSA (2.5 mg/ml) and enzyme dilution buffer (10 mM Tris–HCl, pH 8.3, 2.5 mg/ml BSA) were purchased from Idaho Technology (Idaho Falls, ID). Nuclease-free water, dNTP mix (10 mM each), Taq polymerase (5 U/µl) and male human genomic DNA were purchased from Promega (Madison, WI). The primer set used to amplify the 257 bp target fragment of the Y chromosome of human genomic DNA is 5′-GGTTATCATAGCCCACTATACTTTG-3′ (forward) and 5′-ATCTTTATTCCCTTTGTCTTGCT-3′ (reverse). One of them was used as the primer for the cycle-sequencing reaction. The primer set was synthesized by the DNA facility at Iowa State University (Ames, IA). Poly(ethylene oxide) (PEO) and urea were received from Aldrich (Milwaukee, WI) and ICN Biomedicals (Aurora, OH), respectively. ABI PRISM™ dye-terminator cycle sequencing ready reaction kit with AmpliTaq DNA polymerase, FS (Perkin-Elmer, Foster City, CA) was used.

The 1× TBE buffer solution (89 mM Tris, 89 mM boric acid and 2 mM ethylenediaminetetraacetic acid) was prepared by dissolving the premix (Amresco, Solon, OH) in deionized water. 10 000× SYBR Gold as the intercalating dye was purchased from Molecular Probes (Eugene, OR). Low DNA Mass™ ladder was obtained from Invitrogen (Carlsbad, CA). The sieving matrix with 1× SYBR Gold was made by dissolving 0.5% (w/v) of M_r 8 000 000 PEO in 1× TBE buffer for the estimation of the PCR product amount as the preliminary experiments.

The electrolyte used in the capillary zone electrophoresis (CZE) separation was prepared by dissolving 10 mM Tris, 2 mM MgCl₂ and 3 mM KCl in deionized water. The pH of this electrolyte was adjusted to 8.0 with 0.1 M HCl. Then, 0.3% (w/v) of M_r 1 000 000 poly(vinyl pyrrolidone) (PVP) obtained from Polyscience (Warrington, PA) was added to the electrolyte for dynamic coating of the capillary. This solution was shaken for 30 s and left to stand for 1 h to remove bubbles.

The capillary gel electrophoresis (CGE) sieving matrix for the separation of the cycle-sequencing product was an entangled polymer solution made by dissolving 1.5% (w/v) of M_r 8 000 000 PEO and 1.4% (w/v) of M_r 600 000 PEO in 1× TBE buffer with 7 M urea. The solution was stirred vigorously overnight until all material was dissolved. The solution for coating the CGE capillary was prepared by dissolving 2% (w/v) of M_r 1 000 000 PVP in 1× TBE buffer.

Reaction protocols

PCR protocol. The improved buffer system used in this work is optimized for rapid temperature cycling with a hot-air thermocycler (Idaho Technology). PCR was performed with the following components: 50 mM Tris, pH 8.3, 475 µg/ml BSA, 3 mM MgCl₂, 100–200 µM each dNTP, 0.2–0.4 µM each primer, 9.8 ng/µl male human genomic DNA and 0.05 U/µl Taq polymerase.

The solution was made fresh daily and kept on ice. The temperature protocol for the PCR in the capillary was adjusted to the following: the sample mixture was heated to 95°C and held for 20 s to fully denature the template, followed by a series of 10 cycles at 57°C annealing temperature, followed by two series of 20 cycles at 56 and 50°C annealing temperatures, respectively (50 cycles in total). The denaturing and extension temperatures were fixed at 94 and 72°C, respectively. A step time of 10 s at the extension temperature was employed in each cycle. Each cycle had no holding time (0 s) at the denaturation and annealing temperatures. For off-line experiments, reaction was carried out in a plastic tube using an aluminum block thermocycler (Perkin-Elmer). The temperature cycling protocol was set as follows: initial holding, 2 min at 95°C; denaturation, 20 s at 95°C; annealing, 1 min at 56°C; extension, 1 min at 72°C, 35 cycles; final holding, 5 min at 72°C to assure that all the products were annealed to their double-stranded form.

Cycle-sequencing reaction protocol. Modifications to the standard cycle-sequencing reaction mixture developed for the ABI Model 9600 thermocycler were made to fit the small-volume reaction in the capillary. The reaction mixture typically used in this study consisted of 8 µl of terminator ready reaction mix, 1 µl of 10 µM primer, 1 µl of 25 mM MgCl₂, 2 µl of 10× BSA, and 7 µl of nuclease-free water (19 µl in total). This mixture was prepared in advance and a 1.672 µl aliquot was used for a series of reactions by adding 88 nl of PCR product.

The temperature protocol for the on-column cycle- sequencing reaction was adjusted to the following: the sample mixture was heated to 95°C and held for 20 s; 50 cycles were performed with denaturation at 95°C for 0 s, annealing at 53°C for 0 s, extension at 60°C for 10 s. For off-line experiments, reaction was carried out in a plastic tubes using the block thermocycler. The temperature protocol was as follows: the sample mixture was heated to 95°C and held for 1 min; 35 cycles were carried out with denaturation at 96°C for 10 s, annealing at 53°C for 5 s, and extension at 60°C for 1 min.

Instrumentation

Figure 1 shows the schematic diagram of the set-up. Different inner diameters of capillaries (C₁–C₇, 363 µm o.d.) (Polymicro Technologies, Phoenix, AZ) were used for the two reactions, the two separations and sample manipulation. The multiple capillaries were serially connected to execute the required functions. The entire system consisted of microthermocyclers (MTC₁ and MTC₂), a freeze/thaw (F/T) valve control system (47), microsyringe pumps (MSP₁ and MSP₂) (Kloehn, Las Vegas, NV), a CZE and a CGE electrophoretic system.

Figure 1.

Schematic diagram of instrumental set-up. AB, air blower; C₁ and C₄, solution delivery capillaries; C₂, reaction capillary for PCR; C₃, reaction capillary for cycle-sequencing reaction; C₅, CZE purification capillary; C₆, waste arm; C₇, CGE capillary; CL₁ and CL₂, convex lenses; F/T valve, freeze/thaw valve; LPF₁, 560 nm long pass filter; LPF₂, 630 nm long pass filter; M₁, M₂ and M₃, mirrors; MO₁, MO₂ and MO₃, microscope objectives; MSP₁ and MSP₂, microsyringe pumps; MT₁ and MT₂, microtees; MTC₁ and MTC₂, microthermocyclers; NF, 514 nm notch filter; PMT₁, PMT₂ and PMT₃, photomultiplier tubes; R₁, R₂ and R₃, buffer reservoirs; TEC, thermoelectric cooler; UCC, universal capillary connector. The inset shows the F/T valve control system using a TEC.

Polyether ether keton microtees (MT₁ and MT₂) with 29 nl dead volume (Upchurch Scientific, Oak Harbor, WA) were used to connect the capillaries. Microtee MT₁ was employed to join the 8 cm long reaction capillary for cycle-sequencing reaction (C₃, 150 µm i.d.), the 40 cm long buffer delivery capillary (C₄, 150 µm i.d.) and 30 cm long CZE capillary (C₅, 75 µm i.d., 27 cm effective length). Microtee MT₂ was used to connect the CZE capillary (C₅), 63 cm long CGE capillary (C₇, 75 µm i.d., 50 cm effective length) and 8 cm long waste arm (C₆, 250 µm i.d.). A much larger i.d. capillary was chosen as the waste arm because of its small flow resistance for rinsing and gel filling. Tight connection was confirmed by the stable current during CE separations.

Microthermocycler. The laboratory-built microthermocyclers MTC₁ and MTC₂ were used for PCR and cycle sequencing, respectively. Figure 2 shows the schematic diagram of MTC₁ in greater detail. It was constructed by holding two frames (7.0 × 3.5 × 0.3 cm) together with screws around the four corners. The reaction capillary for PCR (C₂, 150 µm i.d., 10 cm) was sandwiched between two pieces of thin brass sheet (7.0 cm × 3.5 cm × 76 µm) (Small Parts, Miami Lakes, FL) and a Kapton insulated flexible heater (7.5 cm × 2.5 cm × 0.25 mm, 10 W/in²) (Omega, Stamford, CT) with the inlet tip of the capillary ∼21 mm away from the edge of the brass sheets. Double-stick tape was used to bond the brass sheet and the flexible heater. Thermal conductive silver grease (Arctic Silver, Visalia, CA) was applied onto the capillary surface and between the brass sheets to ensure proper heat transfer. For a 150 µm i.d. capillary, 20 mm reaction length corresponded to ∼350 nl reaction volume. One could change to different inner diameters to accommodate different reaction volumes. In order to accurately measure and control the inner temperature of the reaction capillary, a miniature bare type K thermocouple (75 µm diameter, Omega) was inserted into a 150 µm i.d. × 363 µm o.d. capillary. The inner and outer diameters are identical to those of the reaction capillary so the total heat capacity should be similar. The thermocouple probe was placed in the middle of the thermocycler. The silver thermal grease was applied to the surface of the probe capillary as well for better heat conduction.

Figure 2.

Schematic diagram of laboratory-made microthermocycler MTC₁. Thin brass plates provide heat transfer from the heating tape to the reaction capillary. C₁, nuclease-free water delivery capillary; C₂, reaction capillary for PCR.

A proportional-integral-derivative (PID) temperature controller (CN77300, Omega) was used to control the temperature profile for both PCR and cycle sequencing. This mode of control varies the amount of heat applied to the sample in proportion to the difference between the actual and the set temperature. The advantages of PID control over traditional ON/OFF control is that the system can be carefully tuned to compensate for temperature overshoot and ringing. 4-W aquarium air pumps (Tetra Secondnature, Blacksburg, VA) were used to blow room air onto the heater via an air blower (AB) to cool the assembly down quickly during the ramp from denature to annealing steps. A single cooling system was used alternatively for MTC₁ and MTC₂ by simply changing the direction of AB. A laboratory-developed Labview program was used to control the microthermocycler. A standard RS-232 serial port was used to communicate with the temperature controller. The digital TTL-level parallel port output was fed to a solid-state relay (Omega) to control the ON/OFF of the air pumps.

Microfluidic control system. The fluidic control was performed using the programmable microsyringe pumps (MSP₁ and MSP₂), the F/T valve control system, the microtees MT₁ and MT₂, the reaction and liquid delivery capillaries C₁–C₄, and a universal capillary connector (Alltech, Deerfield, IL). The pumps MSP₁ and MSP₂, each equipped with a 25 µl microsyringe and a one-way valve, were used to accurately aspirate and dispense small volumes of solutions for sample manipulation.

The F/T valve system was built on the base of a Peltier thermoelectric cooler (TEC) (Melcor, Trenton, NJ) (see the inset in Fig. 1). When the DC power is activated, heat is removed from the capillary at the cold junction to the hot junction by the semiconductor and dissipated by the cool water flowing through the brass tube. The liquid in the capillary is then frozen, ‘closing’ the F/T valve. When the power is off, heat is transmitted from the hot junction to the cold junction. The liquid in the capillary is then thawed, ‘opening’ the F/T valve.

Detection system. Laser-induced fluorescence (LIF) detection was used in both CZE monitoring and DNA sequencing. A 15 mW Ar-ion laser at 514.5 nm (Uniphase, San Jose, CA) was used for excitation in both separation systems. The laser was divided into several beams by a prism. The strongest beam was used for CGE detection, while a weaker one was used for CZE detection. Uncoated plano-convex lenses (CL₁ and CL₂) (Edmund Scientific, Barrington, NJ) with 12 mm focal length were used to focus the laser beams to the capillary windows. Ten times microscope objectives (MO₁, MO₂ and MO₃) (Edmund Scientific) were used to collect the fluorescence perpendicular to the excitation laser. A 560 nm long-pass filter (LPF₁) was employed to block the scattered light from entering PMT₁ (R928, Hamamatsu, Bridgewater, NJ). A 630 nm long-pass filter (LPF₂), which was used to reduce the stray light entering PMT₂, was also used for ‘red-channel’ detection. A notch filter (NF) (Kaiser Optical System, Ann Arbor, MI) was deployed to prevent the stray light from going into PMT₃, and was also used for ‘blue-channel’ detection. All PMTs were terminated with 10 kΩ resistors before connecting to a 22-bit A/D interface (Model 55, IOtech, Cleveland, OH). A Pentium III 266 MHz computer was used to control the system and for data acquisition.

Operation protocols

The following procedures were sequentially performed for fully integrated DNA sequencing. Initially, capillaries C₁ and C₂ and the syringe pump MSP₁ were filled with nuclease-free water. Capillaries C₃, C₄, C₅ and C₆ and the syringe pump MSP₂ were filled with the CZE separation buffer by pushing the buffer from the open end of C₃. Capillaries C₆ and C₇ were flushed with 2% (w/v) PVP, followed by introduction of PEO gel from the open end of C₇. Then, CZE separation buffer was delivered from the open end of capillary C₃ into capillary C₅ again to push out the gel solution in C₆.

The miniaturized sample manipulation steps from the aspiration of the PCR mixture to on-line cycle sequencing reaction are schematically shown in Figure 3. First, 616 nl of PCR mixture (actual reaction volume) in a tube (T₁) was aspirated into capillary C₂ by pump MSP₁ (Step 1), followed by the aspiration of 369 nl of nuclease-free water in a tube (T₂) so that the PCR mixture plug moved toward the temperature-controlled region in microthermocycler MTC₁ (Step 2). Then, thermal cycling for PCR was started with the inlet tip of C₂ immersed in water in T₂ to avoid evaporation. After PCR, 633 nl of the solution in capillary C₂ was drained (Step 3). Then, 88 nl of the PCR product solution, which included the region with the highest concentration, was dispensed into 1.672 µl of the cycle-sequencing reaction mixture in a narrow tube (T₃) (Fischer Scientific, Fairlawn, NJ) (Step 4). After dispensing, tube T₃ was transferred from MTC₁ and MTC₂ and the inlet tip of capillary C₃ was immersed in T₃. The reaction solution was left for 5 min at ambient temperature to allow complete mixing by diffusion before its introduction into capillary C₃. 352 nl of the cycle-sequencing solution in T₃ was aspirated into capillary C₃ by pump MSP₂ (Step 5), followed by the aspiration of 492 nl of the CZE separation buffer in a tube (T₄) for cycle sequencing at microthermocycler MTC₂ (Step 6). During Steps 5 and 6, the F/T valve was closed so that the CZE separation buffer in capillary C₅ could not move toward microtee MT₁ to affect the aspiration volumes. Finally, thermal cycling for cycle-sequencing reaction was started. The inlet tip of C₃ was immersed in CZE buffer in T₄ during the reaction. The F/T valve was ‘closed’ even while the cycle-sequencing reaction was being conducted in order that the reaction solution in C₃ can sit at the appropriate temperature-controlled region without any influence from hydrodynamic flow caused by the small difference in height between the ends of capillaries C₃ and C₆.

Figure 3.

Schematic representation of sample preparation steps. 1, aspiration of 616 nl PCR mixture in T₁; 2, aspiration of 369 nl nuclease-free water in T₂; 3, drainage of 633 nl solution in C₂; 4, dispensing of 88 nl PCR product into 1.672 µl cycle-sequencing reaction mixture in T₃; 5, aspiration of 352 nl reaction solution in T₃; 6, aspiration of 492 nl CZE buffer in T₄. C₂ and C₃, reaction capillaries for PCR and cycle-sequencing reaction; MTC₁ and MTC₂, microthermocyclers; T₁, T₂, T₃ and T₄, tubes containing PCR mixture, nuclease-free water, cycle-sequencing reaction mixture and CZE buffer, respectively. Steps 1–4 were performed using pump MSP₁ with C₂ while Steps 5 and 6 were carried out using pump MSP₂ with C₃. Tube T₃ was transferred from MTC₁ to MTC₂ after Step 4. The inlet tip of C₂ was immersed in nuclease-free water in T₂ during PCR while the inlet tip of C₃ was immersed in CZE buffer in T₄ during the cycle-sequencing reaction in order to avoid evaporation.

After the cycle-sequencing reaction, the F/T valve was ‘opened’ and CZE separation was immediately started to get rid of the excess dye-terminators and salts by applying 13 kV to reservoirs R₁ and R₂, which contained CZE separation buffer (see Fig. 1). Purification was completed in ∼15 min. When the DNA fragments were detected at the detection window, an appropriate delay was imposed to allow for the fragments to move to the microtee MT₂. Electrokinetic injection of the DNA into the CGE capillary was initiated by switching the positive height-voltage electrode from reservoir R₂ to R₃. After injection at 20 kV for 5 min, the CZE capillary was flushed with 1× TBE buffer with 7 M urea to remove the uneluted excess dye-terminators and salts. Reservoir R₂ was replaced by 1× TBE buffer with 7 M urea for sequencing separation at 11 kV applied to reservoirs R₂ and R₃. Two-channel electropherograms were analyzed by Grams32 software (Thermo Galactic, Salem, NH) for peak picking. Base calling was manually conducted based on the two-color ratio in the blue and red channels (48).

Recycling the capillaries is an important issue in on-line DNA sequencing. In the manual mode in this work, regeneration of the whole system was performed after CGE separation was complete. The CZE capillary, including the reaction section and the microtees, was regenerated by simply flushing with 100 µl of the CZE separation buffer. The CGE capillary was regenerated by flushing with ∼300 µl of deionized water, 50 µl of 2% (w/v) PVP and 15 µl of PEO gel, sequentially. The reaction capillary for PCR was replaced between runs.

RESULTS AND DISCUSSION

Elimination of PCR product clean-up step

First, we prepared PCR products using a heating block with 200 µM of each dNTP and 0.4 µM of each primer. We tried using the PCR product without purification directly in a cycle-sequencing reaction. A small PCR product volume [∼10% (v/v)] was incorporated in the reaction solution to lower the carryover primer and dNTP amounts to as little as possible. It is preferable that the PCR product is quantified prior to the sequencing reaction since the amount used is very important for optimum results. Usually, quantification is carried out by measuring the absorbance at 260 nm. We roughly estimated the product amount by comparing the peak area of the product with that of a 200 bp DNA standard by SYBR Gold staining in CGE (Fig. 4). The manufacturer recommends that 30–90 ng of PCR product is employed at the concentration of 10–30 ng/µl in 20 µl of final sequencing reaction volume. The optimal template amount depends on how large the fragment size is. Since the amplified DNA fragment here is fairly small (257 bp), smaller amounts of products compared with the recommended one should be applicable. We therefore used 1 µl of the PCR product, which gave 30.6 ng of product in 20 µl of sequencing mixture. The template volume [5% (v/v)] in the sequencing reaction mixture is smaller than the normal volume [15–30% (v/v)], resulting in lower amounts of excess primers and dNTPs.

Figure 4.

Electropherogram of a mixture of PCR products obtained by a heating block instrument and DNA standards. Separation matrix, 0.5% PEO (M_r 8 000 000) in 1× TBE buffer containing 1× SYBR Gold. The PCR product was injected into the DNA standards solution prior to loading into the separation capillary (DNA standard:PCR product was 1000:88, v/v). Product amount was estimated by comparison with the peak area of the 200 bp DNA fragment (0.46 ng/µl).

After sequencing reaction using a heating block, the sequencing ladder was separated from the excess dye-terminators and salts by CZE and the purified sequencing ladder was injected on-line into the CGE capillary for sequencing separation. As expected, the sequencing results showed ambiguities such as peak broadenings, peak asymmetries and peak overlaps (Fig. 5A). Ambiguous peaks were noticeable especially in the 173–177 and 200–206 bp regions. When 2 µl of the PCR product, which corresponded to 10% (v/v) of the reaction volume, was employed, the ambiguity was even more remarkable (Fig. 5B). Excessive template will often cause trailing peaks as seen in Figure 5A and B. However, when hydrolytic enzymes exonuclease I and shrimp alkaline phosphatase, which can degrade only the remaining primers and dNTPs (44,45), were deployed to treat the PCR product, a higher quality of electropherogram than those in Figure 5A and B was obtained, even when a larger volume of template [20% (v/v)] was used (Fig. 5C). The smaller the PCR product volume, the more one can reduce the excess primers and dNTPs in the sequencing reaction solution. However, the template amount in the sequencing reaction mixture is also reduced in proportion to the decrease in PCR product volume. Too little template for sequencing results in low, noisy signals with few well defined peaks. Moreover, if the PCR product volume is too small [<5% (v/v)] compared with the sequencing reaction volume, liquid handling will be more difficult. For example, an injection of 1% (v/v) PCR product into a 20 µl final reaction volume corresponds to 200 nl dispensing of PCR product by micropipetting. So, we simply lowered the primers and the dNTP concentrations two times (0.2 µM each primer and 100 µM each of the four dNTPs) in PCR and used a small volume of the PCR product [5% (v/v)] to minimize carryover of unreacted primers and dNTPs. Although the PCR product yield (22.4 ng/µl) was reduced due to the smaller amounts of initial primers and dNTPs, an electropherogram, where all the peaks were well defined except for the compressed peaks region (∼40 min), was obtained. Figure 5D shows that the amounts of leftover reaction components were reduced to levels that do not significantly interfere with dye-terminator cycle sequencing.

Figure 5.

Effect of carryover primers and dNTPs on the DNA sequencing quality. Forward primers were used in the cycle-sequencing reactions. Primer and dNTP concentrations used in PCR were 0.4 µM each and 200 µM each, respectively, except that 0.2 µM of each primer and 100 µM of each dNTP were used for experiment (D). The template volumes in cycle-sequencing reactions for the experiments were 5, 10, 20 and 5% (v/v), respectively. The PCR products used in experiment (C) were treated with exonuclease I and shrimp alkaline phosphatase at 37°C for 15 min, followed by heating at 80°C for 15 min and then cooled down to 4°C. The cycle-sequencing products were cleaned up by the CZE purification system except for experiment (C).

We compared the amounts of critical components at each stage from PCR to cycle-sequencing reaction with respect to the sequencing results in Figure 5A, B and D in Table 1. The concentration ratio of dNTPs to dye-labeled terminators in the reaction mixture must be well defined to yield optimal sequencing results. Carryover dNTPs will change the ratio, resulting in failure of the well balanced extension/termination chemistry. As shown in Table 1, most of the dNTPs were not exhausted by PCR, regardless of their initial concentrations. Therefore, the initial dNTP concentrations are critical to minimize the amount of carryover dNTPs. It appears that excess dNTPs do not significantly degrade the sequencing results if they are used at concentrations less than ∼20 µM in the sequencing reaction.

Table 1. Approximate quantification of critical components at each stage in the experiments in Figure 5.

		Figure 5A	Figure 5B	Figure 5D
Before PCR	Primer conc., each (nM)	400	400	200
	Total dNTP conc. (µM)	800	800	400
After PCR	Obtained product conc. (ng/µl)a	30.6	30.6	22.4
	Remaining primer conc., each (nM)b	216.6	216.6	65.8
	Remaining dNTP conc. (µM)b	714.5	714.5	337.4
Cycle-sequencing reaction	Template volume (% v/v)	5	10	5
	Template amount (ng)c	30.6	61.2	22.4
	Unwanted primer conc. (nM)d	10.3	21.6	3.3
	Reaction primer conc. (nM)d	510.3	521.6	503.3
	Unwanted dNTP conc. (µM)	35.7	71.4	16.8

^aThe product concentrations are estimated by comparison with the DNA standard (200 bp) (see Fig. 4).

^bThe concentrations of the remaining dNTPs and primers are calculated from the obtained PCR product concentrations^a on the assumption that the molar mass of a double-stranded DNA fragment is (number of bp) × (649 g/mol/bp).

^cFor a final reaction volume of 20 µl.

^dThe carryover reverse primer is the unwanted primer when the forward primer is used as the reaction primer and vice versa.

With dye-terminator chemistry, the extension products from the residual primers, i.e. the carryover reverse primer in the experiments in Figure 5, will also be labeled. This will result in the presence of a second sequencing ladder (43). The concentrations of the reverse primer in the three sequencing reactions were 10.3, 21.6 and 3.3 nM, respectively. It is hard to think that such small differences in the carryover reverse primer concentration can produce the large variations in the amount of the second sequencing ladder that affect the sequencing quality. Moreover, it should be noted that those concentrations were much lower than that of the forward primer (∼500 nM), which was 3-fold more than the recommended amount (160 nM) in order to compete with any carryover reverse primer during cycle sequencing. In control experiments, overlapping sequences based on the multiple sequencing ladders were not seen when the PCR product used in Figure 5D was subjected to a commercial four-colors DNA sequencing (Fig. 6).

Figure 6.

The sequencing separation and detection of the labeled ladders were performed by ABI PRISM 377 DNA sequencer using 4.5% polyacrylamide as the separation matrix. BigDye terminator cycle sequencing ready reaction kit (ABI) was used with a commercial heating block instrument for the reaction. The same PCR product as used in Figure 5D was employed. The amounts in the 20 µl of reaction solution are 13.0 ng/5 pmol/3.8 × 10^–2 pmol (PCR product/forward primer/carryover reverse primer) while 22.4 ng/10 pmol/6.6 × 10^–2 pmol (PCR product/forward primer/carryover reverse primer) in 20 µl for Figure 5D. The numbers above the peaks do not indicate the actual base numbers.

Miniaturized sample preparation for DNA sequencing

In-capillary template preparation. Temperature equilibration will always be achieved faster if the sample volume is small, if the container wall is thin, or if the surface-to-volume ratio of the sample exposed to the container wall is high. Therefore, the use of microbore tubes is more attractive for rapid temperature cycling than the use of conventional polypropylene tubes. A number of research groups have applied microbore tubes or microfabricated chambers as reaction vessels for PCR or cycle-sequencing reaction in order to realize a faster amplification (49–55). In the present study, cooling from the denaturation temperature (94°C) to the annealing temperature (56°C) takes ∼15 s, corresponding to a cooling rate of 2.5°C/s. The heating rate is ∼3°C/s. The required time for a single temperature cycle is ∼1 min. An even faster approach to the target temperature would be achieved if thinner brass sheets, which are used to sandwich the capillaries, were used and if more efficient cooling was employed with a well tuned PID temperature controller.

An aliquot of PCR product (88 nl, 0.5 cm long) in the middle of the product solution plug (616 nl, 3.5 cm long) was injected into 1 µl of DNA standard marker for quantification. The total reaction time was significantly reduced and 2–3-fold larger product amounts were obtained by applying short step times rather than long step times. However, the product yield (0.5 ng/µl) was much smaller than that by a heating block instrument (22.4 ng/µl) when the same cycle (35) and step temperatures were applied for the reactions (Fig. 7A and B). The following reasons may explain the difference in product yields. First, adsorption/inactivation of the DNA polymerase on the capillary surface degraded efficient amplification (56). It is noteworthy that the inner diameter of the capillary we used is much smaller than that of the Idaho Technology glass capillary (0.56 mm). It is essential that the sample includes BSA to lower adsorption. Secondly, diffusion of the amplified fragments and other PCR reagents axially during the reaction is inevitable in the present (open) system. This leakage is even more serious with convection due to the difference in temperature between the temperature-controlled and the ambient regions.

Figure 7.

Optimization of PCR product yield. Conditions: (A) 35 cycles with denaturing at 95°C for 20 s, annealing at 56°C for 1 min, and extension at 72°C for 1 min; (B) 35 cycles with denaturing at 95°C for 0 s, annealing at 56°C for 0 s, and extension at 72°C for 10 s; (C) 35 cycles as in (B) with capillary pre-treatment; (D) 50 cycles as in (B) with capillary pre-treatment; (E) 50 cycles touchdown PCR with capillary pre-treatment. The microthermalcycler was used for (B–E) while a heating block instrument was used for (A). More details are given in the text.

An alternative to compensate for the smaller product amounts for subsequent cycle-sequencing reaction is to employ a larger product volume. However, a large volume of crude PCR product in the sequencing reaction solution degrades the sequencing quality as discussed previously. Therefore, improving the product yield is essential.

When we manually aspirated large amounts of PCR cocktail (∼50 µl) into the reaction capillary using a disposable plastic syringe, we found out unexpectedly that the product yield could be greatly improved. We conclude that the enzyme adsorption sites could be effectively saturated by flushing the capillary with PCR reagents in advance. Interestingly, excess enzyme in the sample was not helpful to improving the final reaction yield, nor was excess BSA. Capillary pre-treatment with large amounts of PCR mixture is contrary to a cost-effective sample preparation strategy. Therefore, we simply flushed a fresh reaction capillary with 1 µl PCR solution that did not include genomic DNA or dNTPs by repeating aspirating and dispensing five times using the programmable syringe pump. The pre-conditioned capillary provided a 3-fold increase in product yield (Fig. 7C).

We employed a large cycle number (50 cycles) to increase the product amount further. Although such large cycle numbers (>40) are not normally used because they often produce unwanted fragments, 50 thermal cycles with the pre-treated capillary gave 5.1 ng/µl of products (Fig. 7D). It should be noted that the time that the enzymes are exposed to the high temperature (94°C) is much shorter than that in conventional temperature protocols using a heating block.

The annealing temperature is very important for specific amplification. Too low an annealing temperature at early cycles often causes false priming problems, which in turn lower the final yield of the target DNA fragments significantly. However, it is useful for increasing the product yield to lower the annealing temperature stepwise. Only the target DNA is amplified at the early cycles with higher annealing temperatures. The use of lower annealing temperature at the later cycles allows for more effective priming. We combined this technique, referred to as ‘touchdown’ PCR (57), with a pre-treated capillary and a large cycle number (50 cycles) and obtained 13.1 ng/µl of products (Fig. 7E). Although the amount obtained is still lower than that obtained by PCR with a heating block instrument (22.4 ng/µl), the product was amenable to subsequent sequencing since the concentrations of the remaining primers (6.0 nM) and dNTPs (18.1 µM) were comparable with those for products prepared by a heating block (see Table 1). Moreover, minor products that could lead to multiple sequencing (58) were not observed (Fig. 8). The reaction was completed in ∼50 min, which was less than half of the 35 cycles with a heating-block instrument (∼2 h). The capillary end was simply immersed in nuclease-free water in a tube during the reactions with no seals for most of our studies. Occasionally, the product was not obtained because the solution did not stay at the same position in the capillary due to hydrodynamic flow. This problem can be readily overcome by careful adjustment of the liquid levels or by sealing the end of the capillary with a universal capillary connector.

Figure 8.

Electropherogram of a mixture of PCR products obtained by the microthermal cycler and DNA standards. Conditions were as in Figure 4.

Manipulation of the sequencing reaction mixture. The aforementioned template volume [5% (v/v)] in the sequencing reaction solution must be adhered to since the product was not purified. The syringe pump used was amenable to nanoliter reagent dispensing. Only 88 nl of PCR product aliquot in the middle of the sample plug (616 µl, actual PCR volume) was injected into 1.67 µl of sequencing reaction reagents (see Step 4 in Fig. 3), which is about 10 times lower than the sample preparation volume (20 µl) of the manufacturer’s instruction and the previous work (39–41,47,55). Although a larger reaction mixture (19 µl) was prepared in advance, the remaining stock mixture (∼17 µl) that contains all components for the sequencing reaction except for PCR product as a template could be repetitively used without contamination, resulting in true cost savings. If the resolution of the syringe pump (∼0.5 nl/step with a 25 µl microsyringe) is taken into account, it should be possible to further reduce the volume of sequencing reaction mixture; e.g. a dispensing of 40 nl PCR product, which demands approximately 80 steps for the syringe pump, requires 0.76 µl of sequencing reaction mixture. After injection, the mixture was left for 5 min at ambient temperature to allow complete mixing by diffusion. Since the cycle-sequencing reaction mixture is relatively robust, this delay did not influence the sequencing reaction efficiency. We did not examine the minimum time required for proper mixing. However, the time is negligible compared with those for the other steps such as PCR (∼50 min), cycle-sequencing reaction (∼50 min) and gel separation (∼70 min).

Cycle-sequencing reaction. The same capillary pre-treatment as carried out for PCR may be helpful for achieving higher sequencing reaction efficiency. We eliminated the process because sufficient amounts of sequencing products for sensitive detection were obtained without pre-treatment. The reaction mixture was introduced into the reaction capillary, followed by aspiration of the CZE buffer to allow the mixture to be introduced to the temperature-controlled region. The injected sample plug length for PCR was 3.5 cm, while the length for the sequencing reaction was limited to 2 cm to preserve the resolution for the subsequent CZE separation. The short cycle times such as 0 s for denaturation and annealing and 10 s for extension are amenable to the sequencing reaction as well. Fifty thermal cycles were completed in ∼50 min.

Purification, gel separation and detection

We chose dye-terminator chemistry as opposed to dye-primer chemistry since only one reaction chamber is needed for each DNA sample. However, it is essential to remove the excess dye-terminators from the sequencing product mixture prior to the gel separation. Figure 9A clearly illustrates that incorporated dye-terminators interfere with base calling if they are not removed. The largest dye-terminator peak (marked 1) was seen at 37.5 min although this peak did not affect base calling. Other large dye-terminator peaks appeared around 40.0– 41.7 min and some of them partially overlapped with unresolved small DNA fragments (<70 bp). Relatively smaller dye-terminator peaks were seen around 86–88, 104–106 and 250–254 bp in the well resolved sequencing region and caused unreadable peaks and incorrectly called bases. Previously, we succeeded in completely separating dye-labeled M13mp18 from excess dye-terminators and injecting only the purified product using 10 mM Tris–HCl buffer (pH 8.2) with 1.5 mM MgCl₂, 2 mM KCl and 0.3% (w/v) PVP as the CZE separation electrolyte (41). The sequencing products in the present study were not completely separated from the dye-terminators when the same separation conditions as those in the previous study (41) were used. The lower resolution is principally caused by 4-fold larger sample volume introduced into the reaction capillary. The previous study (41) indicates that lower pH and higher salt concentrations in the separation electrolyte can provide better resolution at the expense of reaction efficiency and effective injection. Although we used slightly lower pH (8.0) and higher concentrations of salts (2 mM MgCl₂ and 3 mM KCl), we still could not completely separate the labeled DNA fragments from the dye-terminators as depicted in Figure 10A. However, we could effectively remove the dye peaks in the well resolved sequencing region (greater than ∼70 bp) with no compromise in detection sensitivity by carefully adjusting the injection time with fast switching (∼10 s) of the electrodes prior to injection (Fig. 9B). Furthermore, it can be seen that the dye-terminators that do not overlap with the unresolved DNA fragments (peak 1) are effectively removed as well.

Figure 9.

(Previous page and above) Effect of incorporation of dye-terminators. (A) CGE separation of unpurified cycle-sequencing product (thin line) and dye-terminators only (solid line). (B) CGE separation of purified cycle-sequencing products. Separation matrix, 1.5% PEO (M_r 8 000 000) and 1.4% PEO (M_r 600 000) in 1× TBE with 7 M urea. The lower panels in (A) and (B) represent the well resolved region in greater detail. The peak marked (1) indicates the largest incorporation. The samples were directly injected into a single gel-filled capillary in experiments of (A) while the sample was purified by CZE followed by on-line injection in experiment (B).

Figure 10.

Separation of dye-labeled DNA products and residual dye-terminators (A) and on-line product injection (B). The thin line in (A) represents CZE separation of the product and the residual dye-terminators while the solid line indicates electrophoretic migration of the dye-terminators only. The interval between a and b in (B) represents the time delay before injection. The positive electrode was switched from R₂ to R₃ between b and c. On-line injection was performed between c and d. Separation electrolyte, 10 mM Tris–HCl buffer (pH 8.0) containing 2 mM MgCl₂, 3 mM KCl and 0.3% (w/v) PVP. More details are given in the text.

The dips before and after injection in Figure 10B are artifacts due to the interruption of data acquisition while switching the electrodes in this experiment. The results showed that 5 min was the optimum injection time under the conditions employed here, such as the specific applied voltage and the conductivities of the CZE separation buffer and the PEO gel. The two-channel sequencing data in Figure 11 demonstrate high signal-to-noise ratios and adequate resolution for base calling from 74 to 257 bp in 70 min with no errors. For diagnostic tests based on direct sequencing of a target DNA region, e.g. mutation detection, read lengths of <500 bp are adequate (31–33). For these short DNA fragments, the two-color detection scheme with PEO gel has been proven to allow correct base calling (41).

The large A peak (256 bp) occasionally masked out the following T peak (257 bp). The last A peak in Figure 11 indicates that synthesis of the DNA ladder produced an extra nucleotide. The artifact is called terminal extendase activity, which often occurs in in vitro PCR (59,60). In this work, sequence detection for the smaller fragments (less than ∼70 bp) was limited by the resolving power of the gel. However, the unresolved sequence (less than ∼70 bp) in Figure 11 was readily revealed by including this region in a companion sequencing from the opposite direction with the reverse primer (data not shown). The entire process from the capillary pre-treatment to the gel separation was accomplished in ∼4 h with minimum operator intervention. The essential features of the present system are summarized in comparison with the previous works (49–55) (Table 2).

Figure 11.

CGE separation of on-line purified cycle-sequencing product. Separation matrix, 1.5% PEO (M_r 8 000 000) and 1.4% PEO (M_r 600 000) in 1× TBE with 7 M urea.

Table 2. Comparison of system characteristics.

Ref.	Process	Reaction vessel	Reaction volume	Heating/cooling sources	Heating/cooling rate or cycling rate	Times for processesa	Results achieved
Present work	PCR-sample preparation for CSR-CSR-purification-CGE	Fuse-silica capillary tubing; 150 µm i.d. × 363 µm o.d.; 10 cm long for PCR and 8 cm long for CSR	616 nl for PCR and 352 nl for CSR	Thin-film heater/compressed air blower	∼3°C/s (heating) 2.5°C/s (cooling) ∼1 min/cycle for both PCR and CSR	All in 4 h for PCR (50 min)-sample preparation for CSR (5 min)-CSR (50 min)-purification (∼15 min)-CGE (∼70 min)	(i) 257 bp fragment of human genomic DNA is amplified using 50 cycles with the comparable yield with that obtained with the commercial thermocycler. (ii) Sequencing ladders amplified with 50 cycles are on-line purified, injected and separated.
49	PCR-purification-CGE or CSR-purification-CGE	Thin-wall Teflon tubing; 558 µm i.d. × 863 µm o.d.; 11 cm long	27 µl	Idaho Technology’s air thermal cycler	∼20 s/cycle for PCR ∼44 s/cycle for CSR	All in 20 min for PCR (8 min)-purification (4 min)-CE (8 min); all in 90 min for CSR (∼25 min)-purification (4 min)-CE (1 h)	(i) 303 bp fragment of M13mp18 DNA amplified with 25 cycles using TAMRA-labeled primers is on-line injected, separated and detected. (ii) Sequencing fragments of M13mp18 ssDNA amplified with 30 cycles using TAMRA-labeled primers are separated (∼600 nt).
50	PCR	1 mm i.d. × 1.58 mm o.d. PTFE tubing	1 µl	Three aluminum cubes (12.7 mm)	46 s/cycle	PCR (23 min)	500 bp fragment of genomic λ-DNA amplified with 30 cycles is obtained with 78% amplification efficiency.
51	CSR-CGE	Fused-silica capillary tubing; 20, 50 or 100 µm i.d. × 360 µm o.d.	62 nl for 20 cm of 20 µm i.d. capillary	9 cm × 0.75 cm aluminum block with 400 µm diameter guide holes/Peltier plate	∼10°C/min for heating and cooling	N/A	C-track sequencing fragment of biotinylated PCR product (1 kb) amplified with 10 cycles using dye-primers are on-line-injected and separated (∼450 bases from the primer annealing site).
52	PCR-offline SGE	Borosilicate glass capillary tube; 390 µm i.d. × 840 µm o.d.; 55 cm long	5 µl	Thin-film semiconductor ITO (3000 Å)/fan	44°C/s (heating) 15°C/s (cooling)	PCR (∼20 min)-offline SGE (N/A)	777 bp fragment of human genomic DNA amplified with 35 cycles is obtained with a lower yield than that with Idaho Technology’s air thermal cycler.
53	PCR-CGE	Microfabricated chamber (42 µm deep) in a glass sandwich structure	280 nl	1 cm2 thin-film heater/fan with nitrogen gas	∼10°C/s for heating and cooling 30 s/cycle	PCR (15 min)-CGE (N/A)	231 bp PCR product of human genomic DNA amplified with 30 cycles from single DNA molecule is on-line-injected and detected.
54	PCR-CGE	Microfabricated well using two glass plates	5–7 µl	Dual Peltier elements	2°C/s (heating) 3–4°C/s (cooling) ∼1 min/cycle	All in ∼30 min for PCR (∼25 min)-CGE (N/A)	199 bp fragment of λ-phage amplified with 25 cycles was on-line injected and detected.
55	PCR-offline SGE or CSR-offline SGE	Microchamber made from rectangular borosilicate glass stock; 500 µm × 5.0 mm o.d.; 13 mm long	28 µl	Infrared radiation from a tungsten lamp/solenoid gated compressed air blower	10°C/s (heating) 20°C/s (cooling) 40 s/cycle	PCR (∼20 min)-offline SGE (N/A)	(i) 216 bp fragment of T-cell receptor β-chain amplified with 30 cycles is obtained with the yield <30% of that obtained with the commercial thermocycler.
							(ii) Sequencing fragments provide sequence out to 430 bp.

^aSome minor processes are involved between each process. CSR, cycle sequencing reaction; SGE, slab gel electrophoresis.

CONCLUSIONS

Direct sequencing of PCR products automatically with small reagent volumes (1.672 µl cycle-sequencing reagent for sample preparation) has been realized by using a serially connected multicapillary assembly. The instrument allows one to carry out the various processes of sub-microliter PCR in a capillary tube, nanoliter template dispensing for preparation of the sequencing reaction solution, in-capillary sequencing reaction, purification of the sequencing products, injection, gel separation and detection. Four modifications to the usual PCR protocol allowed the elimination of the purification of residual primers and dNTPs. Lower amounts of primers and dNTPs were used. A large number of thermal cycles (50 cycles) were employed with ‘touchdown’ PCR to deplete the unreacted primers and dNTPs. Minimum volumes of PCR products [5% (v/v)] were transferred for sequencing to minimize carryover primers and dNTPs. Lastly, 3-fold more forward primer than the recommended amount was used to compete with the carryover reverse primer and vice versa. The elimination of a purification step led to simplification of the instrument set-up, shortening of the total analysis time, and reduction in the amount of PCR reagents used. The extremely short temperature steps applied to both PCR and cycle-sequencing reactions in capillaries were facilitated by a simple microthermocycler. The microfluidic control by using a micro-syringe pump enabled the automatic nanoliter template dispensing with high precision, and provided true savings in reagent cost. All the steps in the protocol were completed in ∼4 h with minimal intervention.

We envision modifying the present system further to meet the needs of high-speed, high-throughput and totally automated operation. In this study, some processes such as the replacement of the CZE buffer in capillaries C₃, C₅ and C₆ with 1× TBE with 7 M urea prior to the gel separation, the transfer of the tube that contained the sequencing reaction solution (Steps 4–5 in Fig. 3), and capillary reconditioning for next template were carried out manually since they could readily be implemented by robotics. Several groups have already developed automatic regeneration systems for multiple-capillary assemblies (47,49,61). Furthermore, the dimensions of the micro-thermal cycler allow for using multiple capillaries (16 capillaries with 4.5 mm spacing). Therefore, future incorporation of robotic arms, highly multiplexed capillary format or microfabrication (62–65) and automatic capillary reconditioning devices into the present scheme will bring to fruition fully automated high-throughput instruments for diagnostic tests based on PCR product direct sequencing.