Microfluidic-Based DNA Purification in a Two-Stage, Dual-Phase

In this report, we show that a novel capillary-based photopolymerized monolith offering unprecedented efficiency (∼80%) for DNA extraction from subm...
0 downloads 0 Views 474KB Size
Anal. Chem. 2007, 79, 6135-6142

Microfluidic-Based DNA Purification in a Two-Stage, Dual-Phase Microchip Containing a Reversed-Phase and a Photopolymerized Monolith Jian Wen,†,‡ Christelle Guillo,‡ Jerome P. Ferrance,‡ and James P. Landers*,†,‡,§

Department of Molecular Physiology and Biological Physics and Department of Chemistry and Department of Mechanical Engineering, University of Virginia, Charlottesville, Virginia 22904, and Department of Pathology, University of Virginia Health Science Center, Charlottesville, Virginia 22908

In this report, we show that a novel capillary-based photopolymerized monolith offering unprecedented efficiency (∼80%) for DNA extraction from submicroliter volumes of whole blood (Wen, J.; Guillo, C.; Ferrance, J. P.; Landers, J. P. Anal. Chem. 2006, 78, 1673-1681) can be translated to microfluidic devices. However, owing to the large mass of protein present in blood, both DNA binding capacity and extraction efficiency were significantly decreased when extraction of DNA was carried out directly from whole blood (38 ( 1%). To circumvent this, a novel two-stage microdevice was developed, consisting in a C18 reversed-phase column for protein capture (stage 1) in series with a monolithic column for DNA extraction (stage 2). The two-stage, dual-phase design improves the capability of the monolith for whole blood DNA extraction by ∼100-fold. From a 10-µL load of whole blood containing 350 ng of DNA, 99% (340 ( 10 ng) traverses the C18 phase while ∼70% (1020 ( 45 ug) of protein is retained. A total of 240 ( 2 ng of DNA was eluted from the secondstage monolith, resulting in an overall extraction efficiency of 69 ( 1%. This provided not only an improvement in extraction efficiency over other chip-based DNA extraction solid phases but also the highest extraction efficiency reported to-date for such sample volumes in a microfluidic device. As an added bonus, the two-stage, dual-phase microdevice allowed the 2-propanol wash step, typically required to remove proteins from the DNA extraction phase for successful PCR, to be completely eliminated, thus streamlining the process without affecting the PCR amplifiability of the extracted DNA.

are known to inhibit PCR amplification of DNA,1 a core methodology for almost all genetic analyses. Typically, 1 µL of a lysed human blood sample contains 35 ng of DNA amid ∼150 µg of protein.2 Purification of this DNA is most often performed using solid-phase extraction (SPE), which has largely replaced liquid extraction methods for purifying DNA from biological samples. However, the proteins in biological samples, particularly many of those in whole blood, bind avidly to the solid phase, thus limiting the DNA binding capacity and requiring a wash step to effectively remove protein PCR inhibitors from the solid phase. The development of solid-phase extraction for DNA purification has also allowed for miniaturization of the DNA extraction process to minimize sample handling, reduce contamination, and expedite analysis time.3,4 Silica phases are the most often utilized solid supports for miniaturized DNA SPE,5 but the proclivity of proteins to bind to the silica surfaces necessitates a wash step. Tian et al.6 reported DNA purification from white blood cells and whole blood on silica beads in a bed volume of only 0.5 µL, but noted decreased capacity and lower extraction efficiencies in comparison with the extraction of prepurified genomic DNA. This indicated that competition from proteins for binding sites was problematic, and as a result, only 0.2 µL of whole blood could be purified on the device. Breadmore et al.7 used a tetraethoxyorthosilicate-based sol-gel to immobilize the silica beads in a microdevice, but could also show effective extraction of DNA from only 0.2 µL of whole blood. Wu et al.8 utilized a tetramethyl orthosilicate (TMOS)-based sol-gel phase using poly(ethylene glycol) as the porogen to provide a large silica surface area, but again, only 0.2 µL of whole blood was loaded, with 60% of DNA recovered in ∼12 µL of elution buffer. A microdevice coupling on-chip DNA purification and

DNA purification and preconcentration is a requirement for most genetic analysis applications, primarily due to the complex nature of biological samples. Whole blood is a particularly complex mixture of nucleic acids, proteins, lipids, metabolites, and inorganic ions. Some of these species, e.g., hemoglobin and heparin,

(1) Wilson, I. G. Appl. Environ. Microbiol. 1997, 63, 3741-3751. (2) Wen, J.; Guillo, C.; Ferrance, J. P.; Lander, J. P. Submitted. (3) Bienvenue, J. M.; Duncalf, N.; Marchiarullo, D.; Ferrance, J. P.; Landers, J. P. J. Forensic Sci. 2006, 51, 266-273. (4) Legendre, L. A.; Bienvenue, J. M.; Roper, M. G.; Ferrance, J. P.; Landers, J. P. Anal. Chem. 2006, 78, 1444-1451. (5) Wolfe, K. A; Breadmore, M. C; Ferrance, J. P.; Power, M. E.; Conroy, J. F.; Norris, P. M.; Landers, J. P. Electrophoresis 2002, 23, 727-733. (6) Tian, H.; Huhmer, A. F.; Landers, J. P. Anal. Biochem. 2000, 283, 175191. (7) Breadmore, M. C.; Wolfe, K. A.; Arcibal, I. G.; Leung, W. K.; Dickson, D.; Giordano, B. C.; Power, M. E.; Ferrance, J. P.; Feldman, S. H.; Norris, P. M.; Landers, J. P. Anal. Chem. 2003, 75, 1880-1886. (8) Wu, Q.; Bienvenue, J. B.; Giordano, B. C.; Hassan, B. J.; Kwok, Y. C.; Norris, P. M.; Landers, J. P.; Ferrance, J. P. Anal. Chem. 2006, 78, 5704-5710.

* To whom correspondence should be addressed. Phone: 434-243-8658. Fax: 434-243-8852. E-mail: [email protected]. † Department of Molecular Physiology and Biological Physics, University of Virginia. ‡ Department of Chemistry and Department of Mechanical Engineering, University of Virginia. § Department of Pathology, University of Virginia Health Science Center. 10.1021/ac0703698 CCC: $37.00 Published on Web 07/11/2007

© 2007 American Chemical Society

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007 6135

amplification was shown to successfully amplify extracted DNA from a slightly larger volume of blood (0.6 µL), but the capacity and extraction efficiency were not measured.4 The integrated device illustrated a problem with the wash step, in that residual 2-propanol from the wash step contaminated the early eluted fractions (which coincidentally contained much of the DNA) and, subsequently, inhibited the amplification. Thus, it is clear that the predominant presence of proteins in any complex biological sample (including blood) complicates the extraction of DNA, decreasing the binding capacity of the solid phase for DNA and necessitating the use of a wash step with a reagent (2-propanol) that is an inhibitor of the subsequent PCR-based DNA amplification. To minimize protein binding and avoid the use of chaotropic/ organic reagents in the DNA purification process, alternative solid phases have been developed. Witek et al.9 reported the preconcentration of DNA using a photoactivated polycarbonate microfluidic device with a capacity of 150 ( 30 ng of DNA from Escherichia coli whole cell lysate. An extensive wash step was required, but the device could be dried before the DNA was eluted to remove any inhibition due to the ethanol in the wash buffer. Larger elution volumes were required, however, and this work did not address DNA purification from human whole blood. Nakagawa et al.10 demonstrated the use of an amine-coated microchip to extract DNA from whole blood. This method avoided the use of PCR-inhibiting reagents, such as 2-propanol, but a large elution volume (>15 µL) and poor efficiency (27-40%) were reported as only 10 ng of DNA was recovered from the ∼44 ng of DNA expected to be present in 1.25 µL of blood extracted. The problematic nature of protein binding in DNA extraction methodologies is, perhaps, best encapsulated by the work of Cao et al.,11 who used a novel chitosan-coated multichannel microchip to demonstrate DNA purification in a totally aqueous system (i.e., avoiding the use of organic reagents). High extraction efficiency (75%) in small elution volumes and increased blood capacity (up to 1.5 µL) were observed due to low protein binding to the chitosan phase. While the volume capacity of 1.5 µL of blood represents a 7-fold improvement, larger capacity still is required for many clinical analyses.12,13 Approaches to improving the capacity of extraction phases for chaotrope-based DNA purification from blood will likely exploit one of two pathssincrease the surface area of the phase or decrease the protein binding to potential DNA binding sites. Increasing the extraction-phase surface area can be achieved by packing smaller diameter beads (e.g., 5 µm), resulting in high back pressure, or by etching pillars in the channels during fabrication,14 which increases fabrication cost. Larger functional surface areas for DNA binding can also be met by monolithic (9) Witek, M.; Llopis, S. D.; Wheatley, A.; McCarley, R. L.; Soper, S. A. Nucleic Acids Res. 2006, 34, e74/1-e74/9. (10) Nakagawa, T.; Tanaka, T.; Niwa, D.; Osaka, T.; Takeyama, H.; Matsunaga, T. J. Biotechnol. 2005, 116, 105-111. (11) Cao, W.; Easley, C. J.; Ferrance, J. P.; Landers, J. P. Anal. Chem. 2006, 78, 7222-7228. (12) Sauvaigo, S.; Barlet, V.; Guetfrari, N.; Innocenti, P.; Parmentier, F.; Bastard, C.; Seigneurin, J. M.; Chermann, J. C.; Teoule, R.; Marchand, J. J. Clin. Microbiol. 1993, 31, 1066-1074. (13) Westervelt, P.; Pollock, J.; Oldfather, K. M.; Walter, M. J.; Ma, M. K.; Williams, A.; DiPersio, J. F.; Ley, T. J. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 9468-9473. (14) Cady, N. C.; Stelick, S.; Batt, C. A. Biosens. Bioelectron. 2003, 19, 59-66.

6136

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

materials.15 With the advantages of large surface area, controllable pore size, and high mass transfer from the porous structure, monoliths have been recently used for DNA isolation16 and preconcentration.17 A commercially available large-volume monolithic matrix for nucleic acid purification, the anion-exchange-based CIM monolithic disk (BIA Separations, Ljubljana, Sloveia), has been reported to have greatly increased binding capacity for plasmid and genomic DNA;16 however, the DNA is eluted under high-salt and high-pH conditions, both of which are problematic for the subsequent PCR.18,19 Our group recently reported on the use of a 3-(trimethoxysilyl)propyl methacrylate (TMSPM) monolith grafted with TMOS in a capillary format for DNA purification,17 showing significantly increased DNA binding capacities with purified genomic DNA over other silica phases. This method utilized UV-induced photopolymerization of a liquid precursor within a small region of the capillary,20 the volume of the monolith being similar to solid-phase volumes utilized in microfluidic DNA extraction devices. In addition, the UV polymerization method allowed accurate placement21 and reproducible fabrication of the monolith with no weirs or special chip fabrication procedures required, making this method ideal for transfer to an integrated microfluidic device where DNA extraction is only one step in the analysis process. Use of this phase with blood, however, was limited due to proclivity of proteins for binding to the matrix, thus occupying potential DNA binding sites. The work reported here shows successful transfer of the derivatized TMSPM monolith from a capillary to a microfluidic format with DNA binding capacities and extraction efficiencies determined using both prepurified genomic DNA and human blood. The decreased capacity observed in the capillary system for DNA extraction from microliter volumes of whole blood was overcome by the two-phase system capturing proteins from blood on the reverse phase (C18) strategically placed upstream of the TMSPM monolith selective for DNA extraction. MATERIALS AND METHODS Materials and Reagents. 3-(Trimethoxysilyl)propyl methacrylate (TMSPM, minimum 98%), tetramethyl orthosilicate (TMOS, 98%), λ-phage DNA, and PCR reagents (magnesium chloride, 10× PCR buffer, dNTP, Taq DNA polymerase) were obtained from Sigma-Aldrich (Milwaukee, WI). Toluene (99.9%), 2-propanol (HPLC grade), guanidine hydrochloride (GuHCl; electrophoresis grade), ethanol (95%), and tris(hydroxymethyl)aminomethane (Tris) were purchased from Fisher Scientific (Fairlawn, NJ). EDTA was obtained from American Research Products (Solon, OH). Photoinitiator Irgacure 1800 was generously donated by Ciba (Tarrytown, NY). Octadecyl (C18) silica beads (13-17-µm particle size, 300-Å-wide pores, carbon content 11%) were obtained from J.T. Baker (Phillipsburg, NJ). The PicoGreen dsDNA assay kit (15) Svec, F.; Fre´chet, J. M. J. Anal. Chem. 1992, 64, 820-822. (16) Bencˇina, M.; Podgornik, A.; Sˇ trancar, A. J. Sep. Sci. 2004, 27, 801-810. (17) Wen, J.; Guillo, C.; Ferrance, J. P.; Landers, J. P. Anal. Chem. 2006, 78, 1673-1681. (18) Hagihara, Y.; Aimoto, S.; Fink, A. L.; Goto, Y. J. Mol. Biol. 1993, 231, 180184. (19) Ling, L. L.; Keohavong, P.; Dias, C.; Thilly, W. G. PCR Methods Appl. 1991, 1, 63-69. (20) Kato, M.; Dulay, M. T.; Bennett, B. D.; Quirino, J. P.; Zare, R. N. J. Chromatogr., A 2001, 924, 187-195. (21) Morishima, K.; Bennett, B. D.; Dulay, M. T.; Quirino, J. P.; Zare, R. N. J. Sep. Sci. 2002, 25, 1226-1230.

was obtained from Molecular Probes (Eugene, OR) for DNA quantification, and BCA protein assay kit was obtained from Pierce (Rockford, IL) for protein quantification. Polyetheretherketone (PEEK) tubing of 360-µm o.d., ferrules, rings, and nut assemblies for chip-based microchip connections were purchased from Upchurch Scientific (Oak Harbor, WA). All solutions were prepared with Nanopure water (Barnstead/Thermolyne, Dubuque, IA). Microchip Fabrication. All glass microchips were fabricated as described previously22 through standard photolithography, wet etching, and thermal bonding (640 °C, 8 h). Borofloat glass slides (127 mm × 127 mm × 1.1 mm) were purchased from Telic (Valencia, CA), precoated with chrome and positive photoresist (AZ1500 resist, 5300 Å). Channels were etched using 40% hydrofluoric acid, and differential depths were achieved by etching in a stepwise fashion using HF-resistant dicing tape (Semiconductor Equipment Corp., Moorpark, CA). Two designs were investigated: a monolith DNA extraction chip (Figure 1A) and a twostage DNA extraction chip (Figure 1B). The monolith DNA extraction design contained channels (20 mm long, 200 µm initial width) that were etched 185 µm deep measured using an XP-1 stylus profilometer (Ambios Technology Inc., Santa Cruz, CA). The two-stage DNA extraction device was composed of a protein capture precolumn and a monolith DNA extraction channel. The protein capture precolumn design contained four parallel chambers (2-mm initial width), a loading arm, and an elution arm; all arm channels have an initial width of 500 µm. The protein capture precolumn chambers were etched 385 µm deep and contained a 30-µm-deep weir in each chamber to retain the beads. The precolumn and the DNA extraction channel were connected by a 10-mm-long, 20-µm-initial width channel etched 200 µm deep. Reservoirs were drilled into a cover plate using “triple ripple” diamond-tipped bits of 1.1- or 1.8-mm diameter (Crystalite Corp., Lewis Center, OH). Preparation of the Photopolymerized Monolith on a Microchip. The microchip channel was treated with 1 M NaOH for 20 min, ddH2O for 10 min, and 1.2 M HCl for another 20 min, then rinsed with 95% ethanol, and dried in an 80 °C oven for 1 h. The channel was then modified using 30% (v/v) TMSPM solution (in 95% ethanol), adjusted to pH 5 with acetic acid, for 12 h in the dark.17 The treated channel was subsequently rinsed with 95% ethanol and dried in the 80 °C oven for 1 h before use. The monomer solution preparation was similar to the one described previously.17 A precursor solution consisting of 85% (v/ v) TMSPM and 15% (v/v) 0.1 M HCl was stirred at room temperature in the dark for 1 min to afford a clear solution. The monomer solution was prepared with 24% (v/v) precursor solution, 76% (v/v) toluene, and 5% (w/v) photoinitiator Irgacure 1800 and stirred in the dark at room temperature for 30 s. The treated channels were then filled with the monomer solution and the reservoirs covered with a plastic sheet to prevent solvent evaporation. The microchip was sandwiched between tape-covered glass slides with a 1 mm × 4 mm window in the tape on the top glass providing the exposure area; the microchip was exposed to UV light for 5 min using a 100-W broad UV wavelength lamp (model RSM100W, Regent Lighting Corp., Burlington, NC) to initiate (22) Manz, A.; Fettinger, J. C.; Verpoorte, E.; Luedi, H.; Widmer, H. M.; Harrison, D. J. Trends Anal. Chem. 1991, 10, 144-149.

Figure 1. Dye-filled monolith DNA extraction microdevice (A) and integrated protein capture (green)/DNA extraction (red) in two-stage, dual-phase microdevice (B). The channels are filled with dyes for better visualization of the protein and nucleic acid capture phase regions within the microchannel architecture. Arrow shows flow from stage 1 to stage 2. Chip dimension: 3 cm (length) × 2.5 cm (width).

polymerization. After polymerization, the microchip was mounted in a Plexiglas holder and ethanol was flushed through the channel using PEEK tubing and a syringe pump (KD Scientific, Holliston, MA) to remove excess monomer reagent and visualize the monolith. Once the monolith was formed, the surface was derivatized by treatment with 85% TMOS solution, as optimized in our previous work.17 Briefly, a mixture of TMOS and 0.1 M HCl (85%/ 15% v/v) was vortexed at room temperature for 1 min. The TMOS solution was then flushed through the monolith at room temperature at a flow rate of 4 µL/min (various derivatization times were investigated). The monolith was then washed for 15 min with 95% ethanol to remove the excess TMOS reagent at 4 µL/min. DNA Extraction on the Monolith. To investigate DNA extraction on the monolith, a prepurified human genomic DNA sample (18 ng/µL) was prepared in equilibration buffer (6 M GuHCl, pH 5.8). Before sample loading, the SPE monolithic column was rinsed with equilibration buffer for 10 min at a flow rate of 4 µL/min. The DNA sample was then delivered onto the column using a syringe at a constant flow rate (4 µL/min). For measurement of DNA capacity, a fraction was collected every 10 Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6137

µL and the amount of DNA that had passed through the monolith without binding was quantified in each fraction using a modified PicoGreen assay (Molecular Probes, Eugene, OR). One microliter of each fraction was diluted into 49 µL of TE solution (10 mM Tris, 1 mM EDTA, pH 8.0), mixed with another 50 µL of 0.5% PicoGreen solution in a 96-well PCR plate, and then incubated for 10 min at room temperature, protected from light. After incubation, sample fluorescence was measured using a Bio-Rad IQ 5 real-time PCR instrument (Hercules, CA). A standard calibration curve was used to quantify the extracted DNA. For DNA extraction, 15 (load sample concentration, 1.5 ng/µL), 50, or 100 ng (load sample concentration, 5 ng/µL) of genomic DNA was loaded onto the monolith; a wash step was subsequently performed using 20 µL of 80% (v/v) 2-propanol in water to remove unbound DNA and proteins/contaminants. DNA was then eluted off the monolith with 20 µL of TE buffer at a 4 µL/min flow rate, and the total amount of DNA eluted was determined using the PicroGreen assay. The blood capacity of the monolith was determined using human whole blood, which was chemically treated prior to extraction to release the DNA from the white blood cells. For different chip capacity investigations, proteinase K (10 µL of a 10 mg/mL solution prepared in ddH2O) was added to either 8 or 32 µL of whole blood; the solution was then diluted to 400 µL (0.02 or 0.08 µL of blood/µL of lysate) in cell lysis buffer (6 M GuHCl, 0.1% v/v Triton, pH 5.8) and vortexed for 1 min before incubation in a water bath at 56 °C for 10 min. After equilibration of the phase with the equilibration buffer (6 M GuHCl, pH 5.8) for 10 min at a flow rate of 4 µL/min, the 0.02 µL of blood/µL of lysate sample was loaded onto the monolithic phase with fractions collected every 5 µL; the DNA present in each fraction was quantified using the modified PicroGreen assay. DNA Extraction on the Integrated Protein Capture/DNA Extraction Microchip. For DNA extraction on the integrated microchip, the 4-mm-long DNA capture matrix was prepared in the monolithic channel and the surface was derivatized using TMOS solution, which was allowed to flow only in the monolithic channel. The precolumn section was then packed with C18 beads followed by equilibration buffer loading for 10 min at a flow rate of 4 µL/min to equilibrate both phases. Before DNA extraction from blood samples, the monolith was preconditioned by running two mock extractions, loading 100 ng of λ-phage DNA, following by eluting with 20 µL of elution buffer (10 mM Tris, 1 mM EDTA, pH 8). DNA extraction from a blood sample (0.08 µL of blood/µL of lysate) was performed in four steps. In the first step, 110 µL of blood sample was loaded onto the precolumn at 5 µL/min, and the waste was collected at the outlet before the DNA monolith. At this point, the outlet between the columns was closed and another 15 µL of blood sample was loaded onto the precolumn at 4 µL/min, with the column effluent now passing through the monolith. The loading of blood sample was then stopped, and equilibration buffer was loaded to move the remaining DNA from the C18 phase to the monolithic phase; the amount of protein and DNA present in each fraction collected from the monolith during the loading step was determined. The final step was elution of the DNA, by passing elution buffer only through the monolith at 4 µL/min. Twelve fractions of 1 µL were collected, with 0.5 µL of each fraction used for DNA quantification and the remaining 0.5 6138

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

µL used for PCR amplification of a 139-bp gelsolin fragment. The PCR mixture for analysis of human genomic DNA amplifications in conventional PCR contained 50 mM KCl, 10 mM Tris-HCl, pH 9.0, at room temperature, 3 mM MgCl2, 0.4 µM concentration of each primer for gelsolin (MWG Biotech, High Point, NC), 100 µM dNTP, and 0.3 units/µL Taq polymerase. Thermocycling conditions were as follows: 94 °C for 30 s, 60 °C for 30 s, and 72 °C for 30 s (30 cycles) with a 3-min preincubation at 95 °C and a final extension of 1 min at 72 °C. PCR products were analyzed using microchip gel electrophoresis performed in an Agilent Bioanalyzer 2100 (Palo Alto, CA) using the commercially available DNA 1000 kit. RESULTS AND DISCUSSION Development of a UV-initiated monolith presented the possibility for generating more reproducible DNA purification microdevices as a result of the ability to accurately control the characteristics and the placement of the immobilized phase within the microchannel architecture.20 However, the performance of this monolith for DNA extraction in a microfluidic platform had to be investigated to determine the capacity, efficiency, and application of this method for microfluidics to real biological crude samples, such as whole blood. Binding Capacity and DNA Extraction Efficiency of the Monolith. A DNA extraction method that provides reasonable DNA binding capacity is essential for most clinical applications involving genetic analysis of nucleic acids from complex samples, e.g., whole blood. Since we had shown that the DNA binding site density of a derivatized TMSPM monolith surface was closely related to the density of the derivatized TMOS layer,17 the effect of TMOS derivatization time on the monolith DNA binding capacity was investigated. To determine the DNA binding capacity of the monoliths established in the microdevice shown in the inset of Figure 2, a human genomic DNA (hgDNA) sample (18 ng/ µL) was continuously loaded through each monolith and fractions (10 µL) were collected from the channel outlet to generate breakthrough curves. Figure 2 shows the DNA breakthrough curves obtained with a 4-mm-long monolith in a straight channel chip (200-µm bottom width, 185 µm deep) using three different derivatization timess25, 45, and 75 min. The capacity of each device was determined from the first derivative of the breakthrough curve, while the binding capacity of the monolith itself was calculated based on the volume of the monolith in the device. The shortest monolith derivatization time investigated (25 min) was observed to not only generate the lowest chip-based DNA binding capacity (3.9 ( 0.6 µg/µL monolith, n ) 3) but was also associated with a higher standard deviation on different chips compared to the 45- and 75-min derivatization timessthis was possibly the result of an incomplete condensation reaction. The 45-min derivatization time showed a slightly higher binding capacity with less variation between monoliths. The longest derivatization time of 75 min provided a monolith with a binding capacity of 6.0 ( 0.4 µg/µL monolith (n ) 3), superior to any other microfluidic DNA extraction phase reported to date. It is worthy of note that the curve for the 75-min derivatization time had a significantly shallower profile, which is consistent with a previous report on the CIM monolithic disk,16 and was attributed to the longer relaxation time needed for large DNA molecules to adapt their shape when passing through the narrow channels of

Table 1. Extraction Efficiency on Monolithic Microdevicesa after Preconditioning run-to-run (intrachip) DNA extracted (ng)

extraction efficiency (%)

DNA extracted (ng)

extraction efficiency (%)

15 50 100

12.44 ( 0.48 43.76 ( 1.31 86.95 ( 4.30

83 ( 3 88 ( 3 87 ( 4

12.53 ( 0.77 43.23 ( 0.75 83.92 ( 0.91

84 ( 5 87 ( 3 84 ( 1

whole blood (µL) 0.1 0.2 0.3

1.81 ( 0.29 2.99 ( 0.62 10.84 ( 2.57

41 ( 7 34 ( 7 40 ( 2

hgDNA (ng)

Figure 2. Typical breakthrough curves obtained from purified genomic DNA loading on monolithic microdevices derivatized with TMOS for varying times (25, 45, and 75 min). Sample (18 ng of DNA/ µL in 6 M GuHCl, pH 5.8) was loaded at 4 µL/min, and fractions were collected every 10 µL. Inset: straight channel microchip utilized for monolith optimization; chip dimension, 3 cm × 2.5 cm (length width); channel dimensions (depth × bottom width), 185 × 200 µm; monolithic phase length, 4 mm. Inset figure shows a chip containing two parallel channels with a monolith formed in one of channels. Inset table shows monolith binding capacities for each derivatization time on different chips (n ) 3).

the monolith. Decreased pore size, owing to the TMOS-derivatized layer, was also indicated by the substantially increased back pressure after the 75-min derivatization. Generated under these conditions, the monolith was shown to be much less robust and more physically unstable under the high running pressure required; thus, a derivatization time of 45 min, which offered lower back pressure but sufficient binding capacity, was selected for further experiments. Having defined the reaction conditions for creating the monolith within the microdevice, the extraction efficiency of the device was determined by loading samples containing various masses of DNA (15, 50, and 100 ng). The first extraction on any new monolith suffered from low extraction efficiency, which was consistent with previously reported results from Wu et al.8 on monolithic TMOS-based sol-gels, probably owing to surface preconditioning and irreversible binding of DNA to sites in deadend flow paths. Overall, about 30-40 ng of DNA was consistently lost during the initial extractions on each newly formed monolith. This problem was remedied with chip preconditioning using a large mass (100 ng) of λ-phage DNA to block dead-end binding sites, yielding successful and reproducible extractions. Any nonhuman source of DNA presumably could suffice (e.g., salmon sperm DNA) with the only prerequisite being that it does not interfere with subsequent PCR amplifications of the human genomic DNA target(s). Table 1 shows the efficiency and reproducibility of consecutive hgDNA extractions (n ) 3) on five different monolithic devices after the surface preconditioning. For the 15-ng DNA samples, 12.44 ( 0.48 ng could be recovered, corresponding to an extraction efficiency of 83 ( 3%, with deviceto-device performance associated with an RSD of less than 5%. For larger masses of DNA, the overall average efficiency for the microchip extraction was roughly 87% (n ) 9), with a run-to-run (intrachip) RSD of less than 4% (n ) 3); interchip (chip-to-chip)

chip-to-chip (interchip)

a Chip dimension (top width ×depth ×bottom width), 570 × 185 × 200 µm. Monolithic phase length, 4 mm. Phase volume, 0.4 µL. TMOS derivatization time, 45 min. Loading hgDNA, 1.5 or 5 ng/µL. Flow rate, 4 µL/min. All hgDNA samples are dissolved in equilibration buffer (6 M GuHCl, pH 5.8).

Figure 3. DNA breakthrough curve obtained on a monolithic column with human whole blood as the sample (n ) 3). Channel and load conditions as in Figure 2; TMOS derivatization time, 45 min.

extraction efficiency was roughly 85% with a RSD of less than 3% (n ) 5). Importantly, all extracted DNA was observed to be eluted in just 3 µL of solution according to the elution profiles (data not shown), a substantial decrease from 12 to 100 µL in elution volume relative to other silica-based microdevices.6-11,14-17 The performance of the monolith was further investigated with the extraction of DNA from, perhaps, the most challenging of biological samplesshuman whole blood. To determine the DNA binding capacity, a lysed whole blood sample was loaded on a monolith that was not preconditioned and DNA in the collected fractions quantified to generate the breakthrough curve shown in Figure 3sbreakthrough occurred after 0.8 µL of blood had been loaded and plateaued at ∼1 µL. This indicated that, despite the large capacity observed with prepurified hgDNA (Figure 2), a DNA binding capacity of ∼28 ng (calculated from the DNA in 0.8 µL of blood) was achieved. The decreased DNA capacity with whole blood most likely involved loss of functional sites for DNA binding to the large mass of proteins (and other lysate components) present in the lysed blood; the extraction efficiency for the blood sample was ∼35-40%. It became clear that, in order to maximally exploit the extraction capabilities of the derivatized TMSPM monolith, interference from protein binding to the Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6139

Figure 4. Integrated protein capture/DNA extraction in two-stage, dual-phase microdevice for DNA extraction from whole blood. Chip dimensions as in Figure 1. Reversed-phase (C18) packed bed length, 10 mm; monolith dimensions as in Figure 2. Arrows from reservoirs 1 to 2 show flow-through stage 1. Arrows from reservoirs 2 to 3 show loading through stage 2. DNA elution was performed in the reverse direction from reservoir 3 to 2. Masses and percentages of protein and DNA at each stage were also incorporated to show the mass balance through the system.

Figure 5. Elution of DNA from the C18 phase with continuous loading of a whole blood lysis solution (solid triangles) or loading of a lysis solution containing 10 µL of blood (dashed open squares). Reversed-phase chamber dimensions as in Figure 3. The flow rate was 5 µL/min, and fractions were collected every 5 µL.

monolith had to be preferably eliminated or, at the very least, substantially reduced. Consequently, a method for on-chip removal of proteins from blood was investigated for implementation upstream of the monolith. Two-Stage, Dual-Phase Microchip Containing Both a C18 Reverse and TMOS Monolith Phase. Considering that the most abundant proteins in blood have reasonable hydrophobicity (e.g., hemoglobin), the use of an octadecyl (C18) reversed-phase precolumn was investigated in another study2 as a phase that should effectively capture hydrophobic proteins with minimal binding of the DNA. A four-parallel channel C18 phase chip was shown to effectively capture protein from up to 10 µL of lysed human whole blood. With optimized selectivity in this proteomic chip, it could extract 70% of the proteins from 10 µL of blood while dispensing >97% of the nucleic acids. We leveraged this in the current work where the goal was to remove a major fraction of the protein prior to binding the nucleic acid components to the monolith. Consequently, the two phases (C18 and TMOS monolith) were combined on a single microdevice, with the C18 6140

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

reversed-phase precolumn used for protein capture (stage 1) in series with the monolithic column for DNA extraction (stage 2) (Figure 4). Figure 4 shows the microchip channels filled with the C18 and monolith phases. In order for the DNA to reach the TMOS monolith phase for capture, it had to traverse the C18 phase; thus, binding of DNA to the C18 phase under the chaotropic conditions associated with the lysis buffer had to be investigated. Figure 5 shows the DNA breakthrough curve generated by continuous loading of blood lysate (solid triangles) and by loading 10 µL of blood onto the C18 phase and then changing the loaded solution to an equilibration buffer in the form of 6 M GuHCl (buffer switch, dashed open squares). The DNA exited the C18 reverse phase in a postponed manner, suggesting that the C18 phase acts as a preconcentration step for DNA, as it binds to the reverse phase but is then replaced by stronger binding proteins. It was postulated that, if the load solution was “switched” from sample (lysed blood) to 6 M GuHCl equilibration buffer after a given volume of sample had been loaded, proteins would remain bound to the reverse phase while the DNA exited stage 1 (C18 phase) readied for transport to the monolith. This allows the majority of the loaded DNA to be recovered from the C18 column before the bulk of the proteins, eliminating the competition between the protein and DNA for the binding sites on the monolith. With the “buffer switch” approach, significant selectivity was achieved as shown in Figure 5. The net result was that >97% of the DNA (340 ( 10 ng) exited stage 1 available for selective stage 2 TMOS monolith capture of DNA (Figure 4). It should be noted that the DNA profiles in Figure 5 were optimized for DNA extraction from blood; however, the extent of DNA recovery will differ from other sample types depending on the protein concentration and content in each sample. Consequently, the buffer switch point will need to be reoptimized for different sample sources. DNA Extraction Using a Two-Stage, Dual-Phase Microchip. It has also been noted that some hydrophilic proteins pass directly through the C18 column without binding, and these might interact with the monolith and occupy binding sites before the

Figure 6. Extraction profiles showing device-to-device reproducibility for DNA extractions from whole blood using the two-stage, dual-phase microdevice. Left insets show the well packed C18 phase device, which provides reproducible results, and the poorly packed device, which significantly decreases the extraction capacity of the monolith. Right inset shows PCR profiles from amplification of a 139-bp gelsolin fragment directly from the eluted DNA with no wash step. DNA was eluted in 10 mM Tris, 1 mM EDTA, pH 8.0, at 4 µL/min with fractions collected every 1 µL.

DNA eluted from the C18 phase. Consequently, the monolithic column was, disconnected from the C18 column during sample loading (by creating the reservoir between the two stages); reconnection occurred when switching the load from lysed blood sample to equilibration buffer. Fractions collected at the outlet of the TMOS-derivatized monolith were quantified for proteins and DNA, and these are reported in Figure 4. From the difference in protein mass detected before and after the monolith, it was determined that tens of micrograms of proteins still bound to the monolith, resulting in ∼30% of the DNA recovered from stage 1 passing through the monolith without binding. After the DNA was bound to the monolith, it was eluted using a low ionic strength TE buffer, delivered through reservoir 3 (Figure 4), allowing it to pass only through the monolith. Figure 6 provides four DNA extraction profiles from four different twostage chips, each having extracted 10-µL blood samples. The 240 ( 2 ng DNA recovered from 10 µL of blood, loaded in a 125-µL total volume of lysis solution, represented an extraction efficiency of 69 ( 1% with the two-stage device (Figure 4). This compares favorably with the 50-65% extraction efficiency for whole blood observed with other chip-based solid phases and represents recovery of a significantly larger mass of DNA than other microchip solid phases reported to-date. Moreover, fractions 2 and 3 contained more than 50% of the extracted DNA (130.1 ( 2.0 ng) in only 2 µL of eluent with 88.4 ( 1.1% (211.3 ( 1.5 ng) eluting in just 5 µL; this represents a >20-fold increase in the DNA concentrationsa characteristic of significant importance for downstream processing of DNA in integrated microfluidic devices.23 With respect to the reproducible nature of the two-stage process, (23) Easley, C. J.; Karlinsey, J. M.; Bienvenue, J. M.; Legendre, L. A.; Roper, M. G.; Feldman, S. H.; Hughes, M. A.; Hewlett, E. L.; Merkel, T. J.; Ferrance, J. P.; Landers, J. P. Proc. Natl. Acad. Sci. U. S. A. 2006, 103, 19272-19277.

the performance seems to be predictable based on how well the C18 phase is packed. The upper three elution profiles of Figure 6 were obtained from three different two-stage microchips, all of which had a well-packed C18 bed based on visual inspections the upper inset photo shows an example. The three curves have essentially the same profile with similar fraction-to-fraction quantitative recovery of DNA. On the other hand, the lower inset photo of the chip shows abnormalities in the packed C18 phase, and this bears itself out in an extraction profile that is clearly associated with lower extraction efficiency. Future work aims to address this packing issue in more detail. The power of the two-stage DNA extraction system is illustrated in Figure 6 (upper right inset data), which shows the electropherograms generated by microchipbased electrophoresis of PCR products generated from amplification of the fractions targeted at a 139-bp fragment of the human gelsolin gene. The protein product of this gene plays an important role in the “gel” to “sol” transformation in cell motility, through the severing and capping of actin filaments, thereby regulating filament lengthsthese filaments are involved in cell structure, motility, apoptosis, and cancer.24,25 A positive PCR control reaction included 3 ng of purified human genomic DNA; a negative control reaction included 1 µL of TE buffer, which had been passed through the monolith bed without an extraction performed. The Figure 6 inset shows the 139-bp product is amplified in the first six fractions (0.5 µL of each) eluted from the two-stage system during a whole blood extraction. Successful amplification of the gelsolin fragment from the extracted DNA represents unequivocal (24) Mielnicki, L. M.; Ying, A .M.; Head, K. L.; Asch, H. L.; Asch, B. B. Exp. Cell Res. 1999, 249, 161-176. (25) Ohtsu, M.; Sakai, N.; Fujita, H.; Kashiwagi, M.; Gasa, S.; Shimizu, S.; Eguchi, Y.; Tsujimoto, Y.; Sakiyama, Y.; Kobayashi, K.; Kuzumaki, N. EMBO J. 1997, 16, 4650-4656.

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6141

proof that PCR inhibitors (including proteins, such as hemaglobin) have been successfully removed using the C18 precolumn. Note that we not only see clear-cut amplification of the 139-bp gelsolin fragment, but with the exception of fraction 1, all fractions contained PCR-amplifiable DNA. This stems from an important benefit of the two-stage extraction system, specifically, that the 2-propanol wash step (usually required to remove proteins from the solid phase) is not needed here. This simplifies the process and accelerates the purification process. While there is still some carryover of residual guanidine, a known PCR inhibitor, in the first fraction, this is minor compared to the carryover of 2-propanol, which can significantly contaminate up to half of the eluted fractions containing DNA.3 It should be noted that PCR amplification was also evaluated using 4 µL of the blood lysate sample loaded onto stage 1, and using 1 µL of the DNA/protein mixture collected after stage 1; both of these contained ∼3 ng of genomic DNA. The PCR results indicated that both of these samples failed to amplify due to a PCR inhibitor (such as the hemoglobin) or the high concentration of guanidine present (data not shown). The issue of residual guanidine can be addressed in future work by decreasing the monolith channel dead volume. The two-stage microdevice can be reused by removing the C18 beads in stage 1 with ethanol and flushing the TMOS-grafted monolith in stage 2 with 1 M NaOH overnight at a low flow rate (5 µL/min). CONCLUSIONS The need for high DNA binding capacity is of the utmost importance with many clinical applications that rely on whole blood as a source of genomic DNA. Examples include the detection of endogenous nucleic acid sequences associated with rare cells (disease markers in cancer cells) or the detection of infectious agents (exogenous DNA associated with bacteria)s both requiring the purification of DNA from samples exceeding the nanoliter volume scale.12,13 Previous microdevices exploiting chaotropic methods for DNA isolation have only demonstrated the processing of human whole blood samples in the 0.1-1.5-µL range with DNA extraction efficiency reaching to 50-60%. Inher-

6142

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

ent in most of these microchip methods is a protein wash that serves to ensure optimal purity of the DNA but increases the number of process steps involved in the process. This work reported here demonstrates a high-capacity, highextraction-efficiency, surface-derivatized monolith for DNA extraction, the reproducible fabrication of which was shown by high run-to-run and device-to-device reproducibility for genomic DNA extractions. The novel two-stage, dual-phase microdevice design is key for the ability to handle up to 10 µL of whole blood, which is facilitated by the capture of proteins by the in-line C18 phase; this represents an increase of at least 50-fold compared to that from most other DNA purification microdevices. Successful amplification of a fragment of a cancer marker gene, gelsolin, following purification of the DNA from human whole blood, illustrated the effectiveness of the method. The ability to achieve this without a protein wash step accelerates the analytical process, reduces the number of analytical steps, and eliminates any potential sample contamination that may occur from switching syringes or tubing. With the majority of the extracted DNA released from the monolith in as little as 2 µL, this system is ideal for the concentration and purification of DNA from whole blood for downstream applications on integrated microfluidic devices. ACKNOWLEDGMENT The authors acknowledge Dr. Weidong Cao and Prof. Michael Roper (Chemistry, Florida State University) for helpful and stimulating discussions. The authors also thank Ciba for their generous donation of Irgacure 1800 and Agilent Technologies, Inc. for the DNA 1000 kits used in this work. Funding for this work was provided by NHGRI through NIH Grant R01 HG002613.

Received for review February 22, 2007. Accepted May 25, 2007. AC0703698