|Home | About | Journals | Submit | Contact Us | Français|
Fluorescent dyes provide specific, sensitive, and multiplexed detection of nucleic acids. To maximize sensitivity, fluorescently labeled reaction products (e.g., cycle sequencing or primer extension products) must be purified away from residual dye-labeled precursors. Successful high-throughput analyses require that this purification be reliable, rapid, and amenable to automation. Common methods for purifying reaction products involve several steps and require processes that are not easily automated. Prolinx®, Inc. has developed RapXtract™ superparamagnetic separation technology, affording rapid and easy-to-perform methods that yield high-quality product and are easily automated. The technology uses superparamagnetic particles that specifically remove unincorporated dye-labeled precursors. These particles are efficiently pelleted in the presence of a magnetic field, making them ideal for purification because of the rapid separations that they allow. RapXtract-purified sequencing reactions yield data with good signal and high Phred quality scores, and they work with various sequencing dye chemistries, including BigDye and near-infrared fluorescence IRDyes. RapXtract technology can also be used to purify dye primer sequencing reactions, primer extension reactions for genotyping analysis, and nucleic acid labeling reactions for microarray hybridization. The ease of use and versatility of RapXtract technology makes it a good choice for manual or automated purification of fluorescently labeled nucleic acids.
Fluorescent dye labels have become essential tools for molecular biology and biotechnology research, enabling a variety of applications for probing complex biological systems. Recent advances leading to higher-intensity fluorescence signals and better detection technologies make possible highly sensitive detection of targets, enabling measurement of single molecules and analyses of samples in microarray formats.1, 2 Fluorescent dyes emitting at different wavelengths can be monitored simultaneously, enabling performance of multiplex analyses on a single sample.3 The utility of fluorescent dyes for labeling nucleic acids is well established,4 particularly for DNA sequencing,5– 8 genotyping,9– 11 nucleic acid microarrays,2, 12 and quantitative polymerase chain reaction (PCR).13, 14
Many applications using fluorescent dye-labeled nucleic acids are performed on instrumentation including liquid handling systems, capillary DNA sequencers, and microarray scanners. Automated systems allow researchers to reduce turnaround times from sample preparation to analysis, to minimize error caused by human manipulation, and to attain high sample throughput if needed. Rapid methods of sample preparation, particularly those amenable to automation, are required to meet the potential of these automated instruments.
Labeling of nucleic acids is generally accomplished by incorporating synthetic fluorescently labeled nucleotides using established enzymatic methods such as random-primed incorporation, PCR amplification, or 3′-end tailing.4, 15 For these reactions, fluorescent precursors are added in large excess to promote efficient incorporation. When labeling is complete, significant amounts of unincorporated precursors remain in the reaction mix. Removal of these dye-labeled precursors is necessary for accurate analysis of the labeled products.3
In fluorescent DNA sequencing, labeling occurs by incorporation of dye-labeled chain-terminating dideoxynucleotides (ddNTPs) that are provided in specific ratios to unlabeled deoxynucleotides (dNTPs). In these reactions, unincorporated dye terminators have high electrophoretic mobility and co-elute with sequence extension products, resulting in large peaks that can obscure as many as 20 bases.16, 17 The effect of excess unincorporated dye terminators on analysis of sequencing reactions is shown in Figure 1A and BB . . Similar complications arise from failure to remove unincorporated dye precursors in other applications using fluorescent dye-labeled nucleic acids.18, 19 The effect of using unpurified fluorescently labeled targets in microarray hybridization is demonstrated in Figure 1C and DD .
Whereas processes for labeling nucleic acids can be performed using automated platforms, methods commonly used to purify dye-labeling reactions are neither rapid nor easily automated. Purification methods involving precipitation by ethanol or isopropanol are well established,15 but they require long centrifugation steps and are labor intensive. Sample recovery and dye removal are also variable. Size-exclusion chromatography yields highly purified samples18, 20, 21 but requires centrifugation, which cannot be easily automated, or vacuum filtration, which is not compatible with all liquid-handling systems and may not provide optimal sample recovery. Filtration techniques using molecular-weight-cutoff membranes have also been used for sequencing reaction purification,21, 22 but with limitations similar to those of size exclusion.
An enzymatic process using shrimp alkaline phosphatase (SAP) and exonuclease I (exo) can also reduce interference with electrophoretic analyses by primers and unincorporated nucleotides. This process degrades primers (single-stranded DNA) and dephosphorylates nucleoside triphosphates, altering their electrophoretic mobility.23 Digestion with SAP/exo can also alter the mobility of dye-labeled nucleotides and thus has been used to reduce dye interference in some labeling reactions.19 However, this reaction does not actually remove dye-labeled nucleotides, and because dye-labeled products remain in solution, this method is not optimal for analysis by capillary electrophoresis.
Superparamagnetic particle-based dye removal methods provide simple and rapid magnetic separations obviating the need for vacuum filtration or centrifugation. However, most of these methods involve capturing sequencing extension products and purifying them through successive washes followed by a release step.24 These procedures are lengthy and generally require flammable solvents such as ethanol or specially modified DNA sequencing primers.
Prolinx, Inc. has developed the unique RapXtract superparamagnetic separation technology that affords methods that are rapid, yield high-quality products, and are amenable to automation. This technology is based on superparamagnetic particles that specifically remove dye-labeled precursors from solution, leaving behind purified extension products in the supernatant. This “reverse purification” process is simple and easy to perform: add unpurified fluorescently labeled nucleic acids, mix, and recover purified nucleic acids (Fig. 22 ). ). Mixing can be performed on a laboratory vortex mixer or by pipet agitation using manual or automated liquid-handling devices.
The RapXtract particles are useful for extracting a broad range of fluorescent dye-labeled precursors, and they have been used to purify reactions from different dye chemistries for DNA sequencing, primer extension for genotyping, and random-primed incorporation for microarray hybridization.
Dye terminator sequencing reactions using the BigDye version 1.0 (Applied Biosystems, Foster City, CA) or BigDye version 3.0 (Applied Biosystems) Ready Reaction Mixes were prepared at various dye dilutions. Dye dilution is indicated by the ratio of microliters of BigDye mix to microliters of total reaction volume, where full-strength BigDye reactions are denoted 8:20 (microliters BigDye:microliters total volume). Reactions were performed in 96-well plates in a PE9700 thermal cycler (Applied Biosystems): 96°C for 5 min, then 30 cycles of: 96°C, 20 s; 50°C, 20 s; 60°C, 4 min; then hold at 4°C. For diluted reactions, 5× Sequencing Reagent (Applied Biosystems) or Better Buffer (The Gel Company, San Francisco, CA) was used as indicated.
Sequencing reactions in 384-well plates were prepared as described above but were performed in a PCR Express thermal cycler with a 384-well block (Hybaid, Middlesex, UK). Reactions were prepared using M13 forward, M13 −47, and M13 −48 primers and pUC19 plasmid (New England Biolabs, Beverly, MA) or pGEM-Zf+ (Promega, Madison, WI) as DNA template. Noncommercial PCR amplicon template from Heterosigma akashiwo chloroplast DNA was kindly provided by R. A Cattolico and J. Veluppillai (University of Washington, Seattle, WA).
Dye terminator DNA sequencing reactions using near-infrared fluorescence dyes were prepared from IRDye700 or IRDye800 Termination Mixes (LI-COR, Lincoln, NE) according to the manufacturer’s instructions (2 μL IRDye:8.5 μL total volume). The reactions were primed with M13 forward primer and used noncommercial plasmid templates 3–7 kilobases in size. Templates were purified using the Plasmid Purification System (Qiagen, Valencia, CA) or the Quantum Prep Plasmid Purification Kit (Bio-Rad, Hercules, CA). Reactions were performed in 96-well plates in an iCycler thermal cycler (Bio-Rad): 95°C for 2 min, then 30 cycles of: 95°C, 30 s; 50°C, 30 s; 72°C, 45 s; then hold at 4°C.
The ABI PRISM SNaPshot Multiplex Kit (Applied Biosystems) was used for obtaining single-base primer extension products. The template was a 791-bp PCR amplicon from bases 6371–7162 of lambda DNA. The three primers specific to sites in this amplicon were synthesized in-house: (1) a 20-base primer starting from base 6767 of lambda, (2) a 21-base primer starting from base 7162, and (3) an 42-base primer starting from base 6371 in lambda and containing a 21-base nonhybridizing tail. Reactions were cycled in a PE9700 thermal cycler (25 cycles of: 96°C, 10 s; 50°C, 5 s; 60°C, 30 s).
Reactions were prepared using a CyScribe First Strand Labeling Kit (Amersham Biosciences, Piscataway, NJ) and 1 μg Saccharomyces cerevisiae messenger RNA (mRNA) (BD Biosciences Clontech, Palo Alto, CA) or 2 μg mouse brain RNA (Ambion, Austin, TX). The mRNA was reverse-transcribed with incorporation of Cy3-labeled deoxyuridine triphosphate (dUTP) or Cy5-labeled dUTP, according to manufacturer’s specifications.
RNA was isolated from Haemophilus influenzae Rd KW20 grown aerobically or anaerobically in 125-mL sterile Erlenmeyer flasks with 25 mL of Brain Heart Infusion Medium (Acumedia Manufacturers, Inc., Baltimore, MD) supplemented with 10 mg/L each of beta-nicotinamide adenine dinucleotide (Sigma, St Louis, MO) and bovine hemin chloride (Sigma). Flasks were inoculated with 25–50 μL of overnight culture. The anaerobic flask was capped with a rubber septum, sealed with foil tape, and flushed with nitrogen while agitating for approximately 5 min. Aerobic and anaerobic flasks were incubated in a 37°C shaker at 275 rpm, until cultures reached late log-phase. Cells were harvested by centrifugation, and RNA was isolated using the Masterpure Kit (Epicentre Technologies, Madison, WI) and accompanying protocols. Purity of RNA was assessed qualitatively on ethidium bromide-stained agarose gels, and concentration was determined by measuring UV absorbance at 260 nm. The RNA was labeled using Alexa Fluor dyes of the ULYSIS Nucleic Acid Labeling Kits (Molecular Probes, Eugene, OR). Alexa Fluor dye 594 (aerobic isolate) or Alexa Fluor dye 660 (anaerobic isolate) were used for this experiment. The reaction was slightly modified from the manufacturer’s specifications: 6 μL of dye in 50% N,N-dimethylformamide/water was added to 10 μg RNA, the mixture was heated at 95°C for 5 min, and then the mixture was placed in an ice slurry bath to quench the reaction.
All labeling reactions and RapXtract particles were allowed to equilibrate at room temperature before purification.
Each sequencing reaction was added to an individual well of RapXtract particles (RapXtract II Kit) from which storage buffer had been removed. The resulting suspensions were mixed on a Genie 2 model G-560 vortex mixer (Scientific Industries, Bohemia, NY) that had been precalibrated, according to RapXtract Kit recommendations, to determine the optimal vortex speed. The time of mixing depended on dye dilution,25 and was 20 s for BigDye 1:10 (μL BigDye:μL total volume), 40 s for BigDye 2:10 or 4:10 (μL BigDye:μL total volume), and 50 s for IRDye700 and IRDye800 full-strength reactions. Sequencing reactions were separated from particles using the Prolinx 96-Well Bar Magnetic Separator, then diluted with 10 μL water for loading onto capillary sequencers or dried for loading on slab gel sequencers. Samples for capillary electrophoresis were loaded from an aqueous solution, which has been shown to be more convenient than other loading solutions (such as formamide) and to yield comparable signal strengths and sequence quality.25
Purification was performed using RapXtract 384 on the Quadra 3 automated assay workstation (Tomtec). Suspensions of particles with samples were mixed using a programmable retractable magnetic nest, according to the protocol developed for this instrument.26
Each reaction was added to combined particles from two wells of RapXtract II and mixed on a Genie 2 model G-560 vortex mixer for 40 s at a setting two steps higher than that recommended above for purification of sequencing reactions. Purified primer extension reactions were separated from particles using the Prolinx 96-Well Bar Magnetic Separator and dried for shipment.
The following amounts of RapXtract particles were used to purify labeled cDNA or labeled RNA: Cy3-labeled cDNA, 600 μg; Cy5-labeled cDNA, 600 μg; Alexa Fluor-labeled RNA, 1200 μg. Unpurified labeled nucleic acids were added to superparamagnetic particles from the RapXtract Fluorescent Nucleic Acid Purification Kit and mixed for 30 s (Cy3, Cy5, Alexa Fluor 594) or 60 s (Alexa Fluor 660) on a Genie 2 model G-560 vortex mixer calibrated according kit recommendations. The labeled cDNA or RNA was removed from particles using a Prolinx Microcentrifuge Tube Magnetic Separator. For comparisons between purified and unpurified labeling reactions, two reactions of Cy3-labeled cDNA were combined and divided into two equal parts. One half was not purified, and the other was purified using RapXtract technology. Both preparations were used for microarray hybridization. Six reactions of Cy5-labeled cDNA were combined and then divided into six equal parts. One part was not purified, and one was purified using RapXtract technology. The remaining four were each purified using one of the following size exclusion systems, according to manufacturer specifications: Auto-Seq G-50 (Amersham Biosciences), Micro Bio-Spin 6 (Bio-Rad), Microcon YM-30 (Millipore, Bedford, MA), or Microcon YM-100 (Millipore). Purified and unpurified labeled cDNA and RNA were analyzed to assess the removal of unincorporated dye-labeled precursors and the yield of purified products, and to determine the amount of nucleic acid to use for hybridization to microarray slides. These analyses were performed electrophoretically using agarose gels for RNA or spectrophotometrically in an HP 845x UV-Visible Chemstation (Hewlett-Packard, Palo Alto, CA) for cDNA.
DNA sequencing analysis was performed either on an ABI PRISM 3100 Genetic Analyzer using POP6 polymer (Applied Biosystems) or on an ABI PRISM 3700 DNA Analyzer using POP5 polymer (Applied Biosystems). Instruments were loaded using injection parameters optimized for samples purified with RapXtract particles,25 as described in Figures 11 , , 33 , , 44 and 66 , , and run according to the manufacturer’s specifications. Quality scoring for BigDye samples was assessed using Phred basecaller software (Codon Code Corp., Dedham, MA).
IRDye700 and IRDye800 DNA sequencing reactions were purified using RapXtract particles or using the ethanol precipitation procedure recommended by the instrument manufacturer (LI-COR). The purified products were dried by vacuum centrifugation and resuspended in 4 μL of a modified stop/load solution (95% formamide, 10 mM EDTA, 30 mM Tris). Aliquots of 0.8 μL from each sample were loaded onto 41-cm gels prepared from Long Ranger gel solution (BioWhittaker Molecular Applications, Walkersville, MD) using RapidLoad 2.0 membrane combs (The Gel Company). Samples were analyzed on an NEN Global IR2 DNA Analyzer (LI-COR) according to the manufacturer’s specifications. Quality scoring and determination of number of bases (read length) were performed using e-SeqV2 sequencing software (LI-COR).
Vacuum-dried samples were sent to the University of Utah Core Genomic Facility (R. Scholl, Salt Lake City, UT) for analysis on an ABI PRISM 377 equipped with Genotyper version 2.0 software. Samples were resuspended in 3 μL of loading dye solution (83% deionized formamide, 8.3 mg/mL blue dextran, 4.2 mM EDTA, pH 8), and 1 μL was loaded per lane.
Cy3-labeled cDNA (500 ng) was added to a hybridization solution of 390 mM sodium chloride, 39 mM sodium citrate, 10 μg poly dA, and 0.2% (w/v) sodium dodecyl sulfate. The cDNA was then hybridized to a Yeast Collage array (Operon, Alameda, CA). Hybridization was performed at 67°C for 12 h and results were visualized on a ScanArray Lite scanner (Packard BioScience, Meriden, CT) at a laser output of 80%, photomultiplier tube (PMT) gain of 80%, and spatial resolution of 50 μm.
Purified Alexa Fluor-labeled RNA (5 μg each preparation) was hybridized in digoxigenin (DIG) hybridization solution (Roche Molecular Biochemicals, Indianapolis, IN) to an H. influenzae glass slide microarray designed by S. Stolyar. This DNA microarray contained PCR amplicons, average size 390 bp, from 188 H. influenzae genes. Each amplicon was spotted in four replicates on a single slide. The microarray was manufactured at the Institute for Systems Biology Microarray Facility (Seattle, WA). Hybridization was performed at 42°C for 15 h, and results were visualized in a ScanArray 5000 (Packard BioScience) at a laser output of 100% (Alex Fluor 594) or 95% (Alexa Fluor 660), PMT gain of 70% (Alexa Fluor 594) or 63% (Alexa Fluor 660) and spatial resolution of 10 μm, with technical assistance from A. Golden of the Institute for Systems Biology Microarray Facility.
The performance of RapXtract technology for removal of unincorporated dye terminators for DNA sequencing reactions is demonstrated in Figures 33 –6 . The sequence shown in Figure 33 was prepared from BigDye version 3 sequencing chemistry (Applied Biosystems) in a 384-well microplate format and purified using an automated purification protocol.26 The purified sequences represented in Figure 44 were prepared from BigDye version 1.0 dye terminators (Applied Biosystems) in a 384-well format and purified with RapXtract 384 using the same automated purification protocol.26 These reactions were prepared from two different formulations of 1:10 (μL BigDye:μL total volume) reactions, one using 5× Sequencing Buffer (Applied Biosystems) as the diluent and one using Better Buffer (The Gel Company). Each formulation yields good-quality sequence, although the average signal strength was somewhat stronger in the sequences obtained with the Better Buffer diluent. DNA sequence quality in this study was assessed using the quality (Q) scoring algorithm provided by the Phred basecaller.27 Q scores are a means of quantifying the accuracy of sequence data; the Q20 score for a particular sequence represents the number of peaks in that sequence for which the accuracy of basecalling is 99% or greater.
Figure 55 shows results from purification of a sequencing reaction prepared using the IRDye700 Termination Mix (LI-COR) and purified manually using the RapXtract II Dye Terminator Removal Kit. The chromatographs in Figures 33 and 55 show very low background, and the sequences are discernible past 700 bases (Fig. 33 ) ) or past 960 bases (Fig. 55 ).
Results from purification of IRDye700- and IRDye800- (LI-COR) labeled DNA sequencing reactions using RapXtract particles and ethanol precipitation are compared in Table 11 . . Samples purified using the RapXtract II kit consistently displayed longer read lengths and higher quality than those purified by ethanol precipitation.
The DNA sequences represented in Figure 66 , , prepared from BigDye version 1.0 dye terminators and purified with the RapXtract II Kit, are shown to demonstrate two examples of the relationship between sequence quality and signal strength. In Figure 6A6A , , sequencing reactions were prepared using a dilution ratio of 3:10 (μL BigDye:μL total volume) and used various templates of varying purity, resulting in variable signal strength and quality of sequence. In Figure 6B6B , , sequencing reactions were prepared using a dilution ratio of 2:10 (μL BigDye:μL total volume). The template was from commercial preparations, but the samples were loaded into the ABI PRISM 3100 (Applied Biosystems) capillaries using instrument default injection parameters, not the parameters recommended for samples purified with RapXtract technology.25 The sequencing reactions performed on templates of variable purity resulted in samples with variable signal strength and sequence quality. The nonoptimized injection results in sequences with reduced signal strength but consistently high sequence quality.
RapXtract particles can be used for purification of fluorescent dye-labeled nucleotides from other types of reactions as well. Figure 77 shows results from purifying SNaPshot (Applied Biosystems) primer extension reactions using the RapXtract II Kit. The products of these reactions are labeled with the same dye-labeled nucleotides that are used in the BigDye sequencing reactions, but in these reactions the dye-labeled nucleotides are more concentrated than in DNA sequencing reactions and there are no unlabeled dNTPs in the reaction mixture. Purification using RapXtract particles results in good signal above background for the reaction products, as indicated in Figure 77 , , indicating that dye-labeled precursors have been removed from the sample.
Fluorescently labeled nucleic acids used for microarray analysis can be successfully purified using RapXtract particles. Figures 1D1D and 88 show microarray hybridization experiments using dye-labeled nucleic acid targets that were purified using RapXtract technology. The DNA microarray in Figure 1D1D was hybridized with cDNA labeled with CyDyes (First Strand cDNA Labeling Kit, Amersham Biosciences) and that in Figure 88 was hybridized with RNA labeled with Alexa Fluors (ULYSIS system, Molecular Probes). The results show that RapXtract technology has been successfully used for purification of both classes of dyes.
Unpurified fluorescent cDNA preparations do not show significant hybridization to a microarray (Fig. 1C1C ), ), whereas fluorescent cDNA that has been purified with RapXtract particles hybridizes well with good reproducibility of replicate spots (Fig. 1D1D ). ). The microarray in Figure 88 was hybridized with a mixture of two preparations of RNA, one labeled with Alexa Fluor 594 and one labeled with Alexa Fluor 660. Both RNA preparations were purified with RapXtract particles, and the resulting hybridization demonstrated good signal strength and low background fluorescence, indicating that unincorporated dyes had been removed.
Removal of excess dye from a nucleic acid preparation can be measured using spectrophotometry. In Figure 99 , , labeled cDNA preparations that were purified by RapXtract particles or by various size exclusion systems are compared with unpurified labeled cDNA. Each of the purification methods resulted in significantly reduced absorbance at 650 nm (the emission maximum of Cy5 dye), indicating that excess unincorporated dye-labeled nucleotides were removed. The yield of labeled cDNA, as determined by measuring absorbance at 260 nm, was significantly better for the sample purified using RapXtract particles than for samples purified using the various size exclusion systems (Fig. 99 ).
This study presents results using RapXtract superparamagnetic separation technology to purify reactions labeled using different fluorescent dye chemistries. We have successfully applied RapXtract technology for purification of nucleic acids labeled with energy transfer dyes for DNA sequencing or primer extensions, near-infrared dyes for DNA sequencing, and Alexa Fluor dyes for microarray hybridization.
Purification of DNA sequencing reactions using RapXtract technology results in good-quality data. The samples in Figure 44 purified with the RapXtract 384 Kit exhibit high Q20 scores with good consistency across the 384-well plate. These reactions are 1:10 (μL BigDye:μL total volume) dye dilutions, indicating that high-quality results can be obtained for highly diluted BigDye reactions.
Purification of IRDye-labeled DNA sequencing reactions using the RapXtract II kit also resulted in high-quality data. The representative experiment shown in Table 11 demonstrates that there can be a lot of variability in quality from one template to another. Purification using RapXtract particles was consistently shown to be more reliable than purification by ethanol precipitation with respect to both read length and Q20 score. The RapXtract protocol is easier to perform and takes significantly less time to complete than ethanol precipitation. DNA sequences purified with RapXtract particles can be analyzed on a range of automated DNA sequencing instruments with similar high-quality results (Table 22 ).
A method that can reproducibly yield high-quality sequences in a 384-well microplate format is significant because adaptation of other purification methods to the 384-well format has proven challenging.28 Superparamagnetic particle technology is well suited for automated systems, and the results in Figure 44 demonstrate that the RapXtract 384 Kit is an effective product for increasing throughput and reproducibility of DNA sequencing analysis using an automated purification process. The results shown here were performed on a Tomtec Quadra 3 instrument (Tomtec, Hamden, CT); other liquid handling platforms that are known to be compatible with purification using RapXtract Kits are given in Table 33 .
Purification of a DNA sequencing reaction can result in some loss of sequencing extension products and, consequently, in lowered overall signal strength. However, removal of unincorporated dye-labeled nucleotides and other contaminants generally improves the overall resolution and quality of the DNA sequence despite the reduction in signal. The two experiments shown in Figure 66 are intended to demonstrate that poor signal strength does not always correlate with poor sequence quality. In Figure 6A6A , , the quality of sequence is poorer in samples that exhibit lower signal; in Figure 6B6B , , sequence quality remains high even when signal is low. The samples in Figure 6A6A were prepared from noncommercial template preparations of varying template homogeneity and sequence composition. These sequences show considerable signal strength variability, which correlates with variable Q20 scores. However, the magnitude of Q20 variability is considerably less than that of signal intensity. The samples in Figure 6B6B were prepared from high-purity commercial template but have reduced signal intensity because they were injected into the capillaries using instrument default injection parameters rather than those optimized for use with RapXtract Kits. Sequencing reaction samples are loaded into capillaries using a controlled-voltage electric field for a specified length of time. The conditions optimal for injection of samples purified using RapXtract technology vary depending on the type of instrument, and generally vary from the injection conditions recommended by the instrument manufacturer.25 Samples injected using default parameters exhibit low signal strength, and yet their quality is comparable to that of samples injected under optimal conditions (Fig. 44 ). ). The Q20 score is influenced by several factors, including concentration and purity of template and primer, affinity and specificity of primer to its binding site, uniformity of temperature during thermal cycling, and nucleotide sequence of primer and template. Highly efficient purification can yield good-quality sequence even when signal is low.
The need to remove unincorporated dye-labeled dideoxynucleotide triphosphate groups from DNA sequencing reactions is in part due to their labile nature. The triphosphates are susceptible to hydrolysis if exposed to heat (such as during thermal cycling or denaturation) or to certain chemicals. During a cycle sequencing reaction, the nucleotides and dye-labeled dideoxynucleotide triphosphate terminators undergo hydrolysis of the phosphodiester bonds. This hydrolysis results in species with high negative charge-to-mass ratios, including dye-labeled nucleoside monophosphates and diphosphates (ddNMPs and ddNDPs). Such species are particularly problematic for capillary DNA sequencing, where they cause intense signals that interfere with analysis of the DNA sequence as far as 250 bases into the sequence trace (Fig. 1A1A ). ). Preliminary studies in which DNA sequencing reactions were treated with phosphatase yielded large dye peaks that migrated more than 200 bases into sequence reads on capillary instruments, and were interpreted to be these hydrolysis products (Spicer DS, personal communication, September 1999). Moreover, because of their high negative charge-to-mass ratio, the hydrolysis products compete with sequencing extension products for electrokinetic injection into the capillary,20 further reducing the overall signal strength. Any labeling method using a thermal cycling reaction to incorporate dye-labeled nucleotides will likely generate similar hydrolysis products, regardless of the dye chemistry. Purification of dye hydrolysis products is necessary prior to any analysis in which such dye hydrolysis products can interfere with interpretation of results.
The results described above show that RapXtract technology has been used successfully to purify DNA sequencing reactions under a variety of conditions and from different classes of dye labels, including fluorescent energy transfer dyes and near-infrared fluorescent dyes. RapXtract Kits have also been used to purify DNA sequencing reactions made using BigDye dye primer kits and DYEnamic ET Terminator Kits (Amersham Biosciences), as shown in Table 44 .
We have also successfully used the RapXtract II Kit to purify single nucleotide primer extension products using SNaPshot primer extension chemistry. As with DNA sequencing reactions, primer extension reactions contain dye terminators, but they do not contain unlabeled dNTPs. As a result, only one nucleotide is added to each primer during the reaction. Because the concentration of dye-labeled nucleotides is high, and because the reaction products are small and produce peaks that have mobilities close to those of hydrolyzed dye-labeled nucleotide species, these nucleotide species can cause significant interference with the analysis of results. Purification using RapXtract particles results in very low interference of unincorporated dye-labeled nucleotides (Fig. 77 ). ). RapXtract purification can be performed at room temperature and is complete in approximately 1 minute. The RapXtract method is simpler and more rapid for reaction purification than the commonly employed SAP/exo method, which requires 75 minutes and involves incubation at high temperatures.19 Moreover, RapXtract particles remove the unincorporated dye-labeled nucleotides from the samples, eliminating their potential to interfere with electrophoretic analysis. The SAP/exo method degrades dye-labeled nucleotides but does not remove them.
Nucleic acid microarrays are generally visualized by incorporating fluorescent dyes into the target nucleic acid preparation. The results in Figure 88 and Table 44 show that RapXtract technology can also be used for purification of these preparations. The microarray platform is highly sensitive, and purification of unincorporated dye-labeled precursors is critical to ensure data accuracy. Unpurified fluorescently labeled nucleic acid preparations exhibit very high fluorescence background and nonspecific hybridization to arrayed nucleic acid probes.29 Such conditions yield faint or undetectable signals (Fig. 1C1C ). ). Fluorescently labeled nucleic acids that have been purified with RapXtract particles exhibit good hybridization properties and low background fluorescence (Figs. 1D1D , , 88 ); ); these characteristics are critical for accurate analysis of hybridization intensities. The DNA microarrays in these figures were hybridized with cDNA (Fig. 1D1D ) ) or total RNA (Fig. 88 ) ) and were visualized using two different classes of dyes: CyDyes in Figure 1D1D and Alexa Fluors in Figure 88 . . The results show that RapXtract technology can be used for purification in both cases. The labeling methods for these two dye systems differ markedly from one another. Fluorescent cDNA (Fig. 1D1D ) ) was labeled by incorporation of fluorescent nucleotides into a nascent cDNA chain, whereas RNA (Fig. 88 ) ) was labeled directly using a nonenzymatic system in which the fluorescent dye is linked to a platinum complex that labels guanine residues of nucleic acids.30
Purification with RapXtract particles consistently produces high yields of labeled nucleic acids, as demonstrated in Figure 99 , , providing more target nucleic acid for visualization of hybridization. Purification using RapXtract particles is more rapid than purification by alcohol precipitation or size exclusion, enabling higher sample throughput than other methods.
A list of all dye labeling systems evaluated so far that are compatible with purification by RapXtract particles is shown in Table 44 . . Because of the broad range of fluorescent dye chemistries that have been successfully extracted using RapXtract particles, it is likely that other fluorescent dye chemistries will also prove compatible with this purification technology.
Fluorescent dyes are used for many applications in molecular biology research, and most of these applications require purification of unincorporated dye-labeled precursors from nucleic acid samples after labeling. RapXtract technology uses functionalized superparamagnetic particles to purify unincorporated dye-labeled precursors. These particles can be used with many different fluorescent dye chemistries and provide a process that is rapid, produces consistently high-quality results, and is compatible with automation. RapXtract technology provides a valuable tool for facilitating high-throughput analyses of nucleic acids using fluorescent dye labels.
We thank Rose Ann Cattolico, Eileen Savoy, and Jean Veluppillai for technical assistance with H. akashiwo DNA; Rebecca Scholl for GeneScan analyses; and Alex Picone and A. Golden for technical assistance with RNA preparation and microarray processing. We are also grateful to Leslie Linkkila for critical reviews, Mark Stolowitz for many helpful discussions, and Fred Hilerio for assistance with editing.
Work on H. influenzae arrays was supported by discretionary funds of L. Hood, Institute for Systems Biology, Seattle, WA. There are no conflicts of interest for the authors of this article, and no permissions or database submissions are required.