Parallel Electrophoretic Analysis of Segmented Samples On Chip for

Oct 15, 2010 - samples at a rate that matches analysis speed. To address ... further increases in throughput for higher numbers of samples by parallel...
19 downloads 0 Views 1MB Size
Anal. Chem. 2010, 82, 9261–9267

Parallel Electrophoretic Analysis of Segmented Samples On Chip for High-Throughput Determination of Enzyme Activities Jian Pei,† Jing Nie,† and Robert T. Kennedy*,†,‡ Departments of Chemistry and Pharmacology, University of Michigan, Ann Arbor, Michigan 48109-1055, United States Capillary electrophoresis (CE) on microfabricated structures has achieved impressive sample throughput by combining fast separation speed and parallel operations. One obstacle to further increasing throughput has been lack of methods for loading and injecting individual samples at a rate that matches analysis speed. To address this issue, we have developed a microfluidic device in which samples stored as nanoliter volume plugs segmented by a fluorocarbon oil are introduced sequentially to an array of three electrophoresis channels. A microfluidic interface consisting of patterned surface chemistry and geometric restriction was used to extract samples from each segmented flow channel and transfer to the respective electrophoresis channel for separation. Fluorescence detection was achieved by imaging the chip using a fluorescence microscope equipped with a chargecoupled device. Characterization of the system shows that injection volume is controlled by sample plug volume, flow rate during introduction, and voltage applied to the electrophoresis channel. The system was tested for a GTPase assay. Peak area ratios of enzyme product and internal standard had 6% relative standard deviations. Cross-contamination between peaks was 7%. Throughput of 120 samples in 10 min was achieved. Further development of the system may allow application to highthroughput applications such as drug screening. Micro total analysis systems (µTAS) have greatly evolved since the first demonstration of capillary electrophoresis (CE) separation on a microfabricated device in 1992.1 CE has remained the dominant separation technique for µTAS because it is easy to implement, consumes small samples, requires no external pumping, and enables rapid analysis of a broad range of analytes.2,3 An important potential advantage of CE-based µTAS is high throughput by integrating multiple electrophoresis systems on a single device that can be controlled in parallel. Microfluidic devices with * To whom correspondence should be addressed. Department of Chemistry, University of Michigan, 930 N. University Ave, Ann Arbor, MI 48109-1055. Phone: 734-615-4363. E-mail: [email protected]. † Department of Chemistry. ‡ Department of Pharmacology. (1) Manz, A.; Harrison, D. J.; Verpoorte, E. M. J.; Fettinger, J. C.; Paulus, A.; Ludi, H.; Widmer, H. M. J. Chromatogr., A 1992, 593, 253–258. (2) Dolnik, V.; Liu, S. R. J. Sep. Sci. 2005, 28, 1994–2009. (3) Wu, D. P.; Qin, J. H.; Lin, B. C. J. Chromatogr., A 2008, 1184, 542–559. 10.1021/ac101755y  2010 American Chemical Society Published on Web 10/15/2010

up to 384 electrophoresis channels have been reported4 with application to genetic analysis,5 immunoassay,6,7 and enzyme assays.8-11 When combined with rapid separations, a parallel system can generate impressive analysis rates. For example, a microfluidic device with 36 parallel electrophoretic channels could analyze 36 enzyme assay mixtures in 30 s.11 A limiting factor in further increases in throughput for higher numbers of samples by parallel, rapid separations is the time required to add new samples to the chip, which has typically involved emptying, cleaning, and reloading on chip sample reservoirs and sample transfer channels. Therefore, further improvement in sample handling is required to achieve better throughput and enable applications such as high-throughput screening (HTS). Several “world-to-chip” interfaces for continually introducing samples to chips have been reported.12-14 In these designs, samples are pumped into the device rather than loaded into a reservoir on the chip. Such schemes improve throughput but still require time to rinse between samples and a mechanical or electrical mechanism to change flow between discrete samples and rinse solutions. An alternative sample introduction method is to transport and feed serial samples entrained in a multiphase flow to the analyzer. Segmented flow (also called “continuous flow analysis”) techniques have been known for decades15,16 but were only recently adapted to microfluidic devices.17 This approach provides a convenient way to introduce samples to a µTAS with little cross-contamination and no requirement of valves or other switching mechanisms between samples. Analyzing samples encased in oil by electrophoresis requires a method of extracting (4) Emrich, C. A.; Tian, H. J.; Medintz, I. L.; Mathies, R. A. Anal. Chem. 2002, 74, 5076–5083. (5) Woolley, A. T.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 11348– 11352. (6) Cheng, S. B.; Skinner, C. D.; Taylor, J.; Attiya, S.; Lee, W. E.; Picelli, G.; Harrison, D. J. Anal. Chem. 2001, 73, 1472–1479. (7) Dishinger, J. F.; Kennedy, R. T. Anal. Chem. 2007, 79, 947–954. (8) Xu, H.; Ewing, A. G. Electrophoresis 2005, 26, 4711–4717. (9) Lin, S.; Fischl, A. S.; Bi, X.; Parce, W. Anal. Biochem. 2003, 314, 97–107. (10) Perrin, D.; Fremaux, C.; Scheer, A. J. Biomol. Screening 2006, 11, 359– 368. (11) Pei, J.; Dishinger, J. F.; Roman, D. L.; Rungwanitcha, C.; Neubig, R. R.; Kennedy, R. T. Anal. Chem. 2008, 80, 5225–5231. (12) Fang, Q. Anal. Bioanal. Chem. 2004, 378, 49–51. (13) He, Q. H.; Fang, Q.; Du, W. B.; Huang, Y. Z.; Fang, Z. L. Analyst 2005, 130, 1052–1058. (14) Chen, G.; Wang, J. Analyst 2004, 129, 507–511. (15) Furman, W. B. Continuous-Flow Analysis. Theory and Practice; Marcel Dekker, Inc.: New York, 1976. (16) Snyder, L. R. Anal. Chim. Acta 1980, 114, 3–18. (17) Linder, V.; Sia, S. K.; Whitesides, G. M. Anal. Chem. 2005, 77, 64–71.

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

9261

the samples and transferring them to the electrophoresis channel. Recently, several methods for achieving such extraction and injection have been described, but these methods have yet to be used in parallel for high-throughput assays.18-20 Further, such interfaces have been used for either demonstrations or twodimensional separations rather than analysis of arrays of distinct sample plugs. In this report, we describe an approach to high-throughput electrophoresis based on coupling segmented flow to electrophoresis channels in parallel on a microfluidic device. The method uses a novel method of extracting samples from segmented flow for serial electrophoresis analysis that is compatible with parallel electrophoresis systems on one microfluidic device. Using this method, we demonstrate that arrays of sample plugs stored in tubes can be pumped into parallel CE channels for continuous analysis resulting in throughput that is limited by the electrophoresis separation time. The entire analysis was automated with no need for voltage or mechanical switches between samples. The method is applied to analyzing GTPase reactions as a model enzyme assay system. We envision eventual utilization of this device for automated high-throughput analysis such as drug discovery. EXPERIMENTAL SECTION Materials. BODIPY FL GTP (BGTP) and BODIPY FL GDP (BGDP) were obtained from Molecular Probes, Inc. (Eugene, OR). GRo was expressed and purified as described previously and stored at -80 °C until used.11 All other chemicals were purchased from Sigma. All buffers were made using Milli-Q (Millipore, Bedford, MA) 18 MΩ deionized water and filtered using 0.2 µm surfactant-free cellulose acetate (SFCA) membrane filters (Nalgene Labware, Rochester, NY). BGTP hydrolysis was performed in Tris-glycine buffer (containing 25 mM Tris and 192 mM glycine, pH 8.3) supplemented with 1 mM ethylenediaminetetraacetic acid (EDTA) and 10 mM MgCl2 (TGEM buffer). Electrophoresis buffer was Trisglycine buffer supplemented with 5 mM MgSO4 (TGM buffer). Microfluidic Chip Fabrication. Microfluidic channels were fabricated on two separate 76 mm × 76 mm borosilicate (D263) glass substrates using standard photolithographic procedures. The substrates were 0.5 mm thick borosilicate glass precoated with a 530 nm layer of AZ1518 positive photoresist on top of another 120 nm layer of chrome from Telic Co., Santa Monica, CA. UV light exposure for 6 s at 26 W/cm2 through a custom-made photomask (Fineline Imaging, Colorado Springs, CO) was used to pattern the microfluidic channel designs onto substrates. The exposed photomask blanks were developed in AZ915 MIF developer (Clariant Corp., Summerville, NJ) and then treated with CEP-200 chrome etchant (Microchrome Technologies, Inc., San Jose, CA). Channels (80 µm deep) were formed in the substrate that would carry the segmented flow by etching the exposed glass in 50:50 (v/v) HCl/HF for 345 s. Access holes (360 µm diameter) were drilled onto this etched plate (18) Edgar, J. S.; Pabbati, C. P.; Lorenz, R. M.; He, M. Y.; Fiorini, G. S.; Chiu, D. T. Anal. Chem. 2006, 78, 6948–6954. (19) Roman, G. T.; Wang, M.; Shultz, K. N.; Jennings, C.; Kennedy, R. T. Anal. Chem. 2008, 80, 8231–8238. (20) Niu, X.; Zhang, B.; Marszalek, R. T.; Ces, O.; Edel, J. B.; Klug, D. R.; deMello, A. J. Chem. Commun. 2009, 7, 6159–6161.

9262

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

with diamond-tipped drill bits (Kyocera, Ervine, CA). Eighteen micrometer deep separation channels were created on another glass substrate by etching in the same solution for 100 s. Both glass substrates were thoroughly cleaned in Piranha solution (3:1 v/v H2SO4/H2O2) for 30 min and then a 60 °C RCA solution (5:1:1 v/v/v H2O/NH4OH/H2O2) for 40 min. The cleaned substrates were pressed together when wet and bonded at 606 °C for 8 h in a Neytech Centurian Qex furnace (Pacific Combustion, Los Angeles, CA). Nanoport microfluidic reservoirs (Upchurch Scientific, Oak Harbor, WA) were attached to the bonded chip over the access holes after bonding. Additional 360 µm access holes were drilled from the side of the chip at the capillary docking guide using flat-tipped drill bits from Kyocera so that they connected with the segmentedflow channels.21 Water was perfused through the channels during drilling to prevent debris from clogging the channels. The chip was thoroughly sonicated after drilling to clean out all remaining debris. Fused-silica capillaries (360 µm o.d. by 150 µm i.d.; pretreated with octadecyltrichlorosilane (OTCS) to make the inner wall hydrophobic) were inserted into the access holes and glued in place using fast setting epoxy (Royal Adhesives & Sealants, Belleville, NJ) as the inlet for segmented flow. Figure 1 shows the completed device. Channel Wall Modification. Extraction of aqueous plugs from the segmented flow was achieved by pumping segmented flow past an interface that had a parallel flow of electrophoresis buffer flow in it. To stabilize the interface of flowing oil and buffer, the segmented flow channel was selectively modified and made hydrophobic with OTCS.19,22 To achieve selective modification, 0.5% OTCS solution in anhydrous hexadecane was pumped through the segmented flow channel at 10 µL/min while hexadecane was pumped at 2 µL/min as shown in Figure 1d. A 6-port valve (VICI Valco Instruments Co. Inc., Houston, TX) was used for switching the flow through the segmented flow channel between OTCS solution and hexadecane without interrupting the flow profile. Bubbles were flushed out and stable flow was maintained for 4 min before the flow was switched from hexadecane to OTCS solution for 7 min. The flow was switched back to hexadecane for another 14 min in order to flush out all residual OTCS. After derivatization, all channels were flushed three times with 200 µL of anhydrous hexane and then thoroughly dried under vacuum. Sample Preparation. GTPase assay for G protein R unit (GRo) was performed according to conditions reported before with some minor modifications.11 The hydrolysis reaction was initiated by mixing 2 µM BGTP with 100 nM GRo in the TGEM buffer, and the mixture was incubated at room temperature for some amount of time before it was quenched by adding 20 µM GTP. After quenching, up to 40 sample plugs of 2-18 nL each were loaded into a 150 µm i.d. by 360 µm o.d. Teflon tube (IDEX Health & Science, Oak Harbor, WA) by alternately aspirating sample and oil (Fluorinert FC-40 supplemented with 1% perfluoro-1-octanol) from a 96-well plate using a syringe pump (PHD 200, Harvard Apparatus, Holliston, MA). An xyz-micropositioner (XSlide, Velmex Inc., Bloomfield, NY) controlled by a Labview program written (21) Bings, N. H.; Wang, C.; Skinner, C. D.; Colyer, C. L.; Thibault, P.; Harrison, D. J. Anal. Chem. 1999, 71, 3292–3296. (22) Zhao, B.; Moore, J. S.; Beebe, D. J. Science 2001, 291, 1023–1026.

Figure 1. Details of the completed microfluidic device. (a) Schematic showing how the chip functions (not drawn to scale). Segmented sample flows enter the chip as indicated by the shaded arrows. Electrophoresis buffer flow inlets are indicated by the solid arrows. Sample droplet is extracted into the electrophoresis buffer flow while oil from the segmented flow emerges at the oil waste. Buffer waste is grounded and a negative injection/separation voltage is applied to common waste. (b) A close-up photograph of the interface for extracting water samples from the segmented flow. The sample inlet capillary forms a dead volume-free connection with the microfluidic channel. Segmented flow and buffer flow enter the chip in the directions indicated. Extraction occurs at the interface and flow passes the sample past the electrophoresis channel. (c) Cross section of the interface (not drawn to scale). Both channels are 80 µm deep while the gap between them is only 5-10 µm deep. (d) Channel modification procedure. The clear part shows the OTCS hexadecane solution flow (at 10 µL/min) while the black region indicates the hexadecane flow (at 2 µL/min from each direction) during patterned surface derivatization. Arrows indicate direction of flow. This pattern results in only the segmented flow channel being derivatized.

with deionized water and electrophoresis buffer for 5 min each. Sample cartridges were connected to the chip inlet capillaries using 360 i.d. Teflon connectors. To start sample analysis, both buffer and segmented-flow were initiated with the cathode reservoir (common waste in Figure 1a) electrically connected to a negative high voltage (-HV) power supply (CZE1000R, Spellman High Voltage Electronics, Hauppauge, NY), and buffer waste were reservoirs grounded. Voltage applied was controlled by a personal computer equipped with a multifunction board (ATMIIO-16, National Instruments, Austin, TX) using software written in-house. Fluorescence Detection and Data Analysis. Fluorescence detection of all separation channels was accomplished by collecting fluorescence images using an inverted epi-fluorescence microscope (IX71, Olympus America, Inc., Melville, NY) as described before.11 Briefly, light from a 300 W Xe arc lamp (LBLS/30, Sutter Instrument Company, Novato, CA) was passed through a FITC filter cube (Semrock, Rochester, NY) before being focused on the chip detection region (Figure 1a) with a 10× objective lens with 0.40 NA (Olympus America Inc., Melville, NY). Fluorescent emission was collected with the same objective and detected using an electron-multiplying CCD camera (C9100-13, Hamamatsu Photonic Systems, Bridgewater, NJ). Images were collected at 10 Hz, stored, and analyzed with Slidebook software (Intelligent Imaging Innovations, Inc., Denver, CO). Fluorescence intensity corresponding to each separation channel was extracted from the images to produce electropherograms that were analyzed using software written in-house.24

in-house (National Instruments, Austin, TX) was used to maneuver the inlet of the Teflon tubing from well-to-well. Samples in the 96-well plate were covered with a layer of oil, as described previously,23 so that oil was aspirated into the tube during transfer from one sample well to the next. An aspiration rate of 1 µL/min was used to form the sample array or “cartridge”. Sample plugs were measured under a fluorescence microscope when they passed through the field of view, and their sizes were estimated to have a variance of 7%. Sample tubing diameter was used a reference for measuring plug size. Mircrofluidic Chip Operation. Before use, chips were conditioned by perfusing 0.1 M NaOH in 50:50 (v/v) H2O/ methanol through separation channels for 10 min. Care was taken so that the solution did not enter the hydrophobic channels. This treatment was followed by conditioning channels

RESULTS AND DISCUSSION Chip Overview. Safety Considerations. Piranha solution (70% concentrated H2SO4, 30% H2O2) and hydrofluoric acid (HF) are very dangerous and should be handled with extreme caution. Gloves and eye protection are required. Both solutions are prepared and used in fume hood to prevent inhalation. Piranha solution should be made in open glass containers and cooled down before disposal. HF should be kept cool in polystyrene container and disposed in an appropriate manner. In this work, preformed sample arrays stored as discrete plugs segmented by an immiscible oil phase in a tube were pumped into a microfluidic chip (Figure 1) where plugs were sequentially sampled and analyzed by electrophoresis without mechanical or electrical valves. A capillary pretreated with OTCS formed a zero dead volume connection with the microfluidic channel, as shown in Figure 1b, and served to transfer samples to the chip without coalescence of adjacent plugs. Aqueous sample plugs were extracted from the segmented flow for electrophoretic injection using a microfabricated interface where the oil stream was in contact with a parallel flow of electrophoresis buffer. The interface was created by isotropically etching two channels which were originally 150 µm apart in the photomask. After well-controlled etching, the two channels connected with each other via a gap about 5-10 µm high (cross section shown in Figure 1c). In addition, the segmented flow channel was selectively rendered hydrophobic by derivatizing with OTCS as shown in Figure 1d. The combination of geometric restriction and surface patterning

(23) Chabert, M.; Dorfman, K. D.; de Cremoux, P.; Roeraade, J.; Viovy, J. L. Anal. Chem. 2006, 78, 7722–7728.

(24) Shackman, J. G.; Watson, C. J.; Kennedy, R. T. J. Chromatogr., A 2004, 1040, 273–282.

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

9263

Figure 2. Typical electropherograms showing cross contamination. Two sample plugs containing 65.4 nM rhodamine 110 internal standard and 2 µM BGTP in TGEM followed by three blank plugs were extracted and injected. Each plug was about 9 nL. Immiscible phase was Fluorinert FC-40 supplemented with 1% perfluoro-1octanol. Both segmented flow and buffer flow rates were 4 µL/min, and separation voltage was 6 kV.

was found to facilitate formation of a stable water/oil interface or “virtual wall” similar to previous reports.19,25 As aqueous plugs flowed past the interface, they crossed the virtual wall and merged into the hydrophilic (underivatized) region of the interface so that the electrophoresis buffer flow carried the plug content past the inlet of the electrophoresis channel allowing a portion of the individual plugs to be sampled and separated. The chief advantage of this interface over a similar, previously reported design for extraction and electrophoretic injection of sample plugs19 is ease of fabrication, especially for parallel channels. In previous work, the geometric restriction was created by manually aligning two microfluidic networks of different depths under microscope. Although the critical alignment was feasible for a single electrophoresis system, it was impractical for multiple systems. The present interface had much larger tolerances for alignment so that the two wafers could be arranged manually under the naked eye even with parallel systems on the chip. This was because the geometric restriction was microfabricated (Figure 1c), and the only alignment required was to lay the electrophoresis channel over the buffer channel as shown in Figure 1b. A direct benefit of fabrication multiple units on a single chip was proportionally increased analysis throughput. In principle, other interfaces18,20 would also be compatible with parallel separations. Cross-Contamination. As a goal of this system was to use it for rapid serial analysis, we evaluated the potential for crosscontamination between sample plugs. To evaluate cross contamination, a series of two sample plugs containing 65 nM rhodamine 110 and 2 µM BGTP in TGEM buffer followed by three blank plugs, each about 9 nL, were serially analyzed on chip (Figure 2). Cross contamination averaged 7 ± 2% (n ) 3 replicates on 3 chips). The biggest source of cross contamination was an aqueous layer that formed in the segmented flow channel just before the interface. This aqueous layer was present because a small region of the channel was not derivatized by the surface patterning due to inevitable leakage of the solvent during surface patterning as illustrated in Figure 1d. To minimize the volume of this layer, we (25) Aota, A.; Hibara, A.; Kitamori, T. Anal. Chem. 2007, 79, 3919–3924.

9264

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

decreased its length to 300 µm as shown in Figure 1b. In addition, we found that by supplementing the FC-40 with 1% (v/v) perfluoro1-octanol the oil phase would expand and displace part of the aqueous layer when pumping segmented flow through, thus further decreasing its volume. After the first blank plug, signals from the other two plugs did not drop significantly, which indicated either poor washout of this region or possibly another source cross contamination. Thus, in cases where the 7% crosscontamination is too high, a single wash plug could be inserted between each sample plug at the cost of decreasing throughput by half. Further measures to reduce cross contamination could be considered if necessary. For example, because the source of cross contamination, the aqueous layer, has a constant volume, lower cross contamination should be possible if larger sample plugs are used. One alternative that was considered was to change the flow pattern during surface modification so that the oil flow was in the opposite direction to that shown in Figure 1d; however, this resulted in worse separation efficiency as discussed in the Supporting Information. Sample Injection. Sample injection control is essential for speed and quality of CE operation and performance. Plug size, flow rate, and voltage were expected to affect injection and separation performance and, thus, studied in detail. Plugs generated from solution containing 65 nM rhodamine 110 and 2 µM BGTP in TGEM buffer were used for all evaluations. Increasing sample plug length should increase the time the plug is passing past the inlet of the electrophoresis channel. Therefore, it was expected that increasing plug size would generate larger peaks but also increase bandwidth if injection was the dominant source of band broadening. To evaluate this effect, groups of sample plugs of different sizes were generated and then analyzed on chip (Figure 3a). Peak width at half peak height was used to evaluate peak broadening because determining actual separation efficiency (i.e., plate number) was complicated by the difficulty of determining the injection time (and, hence, migration time) for each sample plug. (Determining actual migration time would require detecting each plug as it entered the electrophoresis capillary in addition to detecting the peaks after separation because no trigger is used to initiate injection.) As shown in Figure 3b, peak height increased with plug size. Above 5 nL, the peak width also increased. These results show that plug volume can be varied to control injection volume and yield the typical trade-off for sensitivity versus separation efficiency. Increasing the flow rate of electrophoresis buffer is expected to cause the extracted plugs to pass the column inlet more quickly and, therefore, decrease injection volume and improve efficiency. To evaluate this effect, 9 nL plugs were pumped into the system while varying the electrophoresis buffer flow rate from 2 to 5 µL/ min. The segmented flow was matched to the electrophoresis buffer so that neither of the flow rates would be rate limiting for plug transfer. (In all cases, video analysis showed that the entire plug was extracted and transferred to the electrophoresis buffer stream.) As seen in Figure 4a, increasing flow rate directly resulted in higher injection frequency and, thus, higher throughput. Peak intensity decreased with increasing flow rate but separation efficiency was not significantly improved (Figure 4b). Thus, with this system a trade-off between throughput and sensitivity can be made by varying sample and buffer flow rates.

Figure 3. Effect of sample plug volume on CE separation. (a) Five groups of sample plugs of different sizes were analyzed on chip. Each group had four plugs of the same size as indicated in the figure. Each plug contained 65.4 nM rhodamine 110 internal standard and 2 µM BGTP in TGEM buffer. Flow rate was 4 µL/min, and separation voltage was 7 kV. (b) Rhodamine 110 peak heights and peak widths at half peak height were plotted against plug sizes. Error bars are (1 standard deviation (SD).

Because the applied potential was kept constant during chip operation, the same voltage was used for injection and separation. Therefore, we expect a trade-off in that higher voltages will give larger peaks, faster separations, and higher efficiency but the latter is countered by extra column band broadening from injection. To evaluate this effect, voltage was varied from 4 to 7 kV for 9 nL plugs pumped into the chip at 4 µL/min (Figure 5a). As shown in Figure 5b, peak height increased with voltage while the zone width decreased slightly but then became stable. This latter effect probably reflects the band broadening of over injection countering the effect of voltage on width. Extension to higher voltages was not possible because of arcing. Even if arcing was prevented, it seems unlikely that substantially higher fields could be used without Joule heating because at 7 kV the electric field in the separation channel was fairly high at 1.8 kV/cm. These results show that although injections are performed passively with this system, i.e., there is no active control of valves or potentials, injection volumes can be adjusted through a variety of parameters. The passive control has some advantage in simplifying the system, especially for parallel systems where the plugs may not always enter the channel at the same time on all channels. It may be possible to use active control, for example, lowering the voltage during injection and then increasing for separation; however, this would require capability to detect the presence of a plug at the channel inlet to trigger the voltage changes. In terms of throughput on a single channel, these results suggest that higher sample flow rates and higher voltages will

Figure 4. Effect of segmented flow rate on CE separation. (a) Sample plugs of the same size and content were analyzed on chip but with varied flow rates as indicated. Three plugs were injected at each flow rate. Each plug was about 9 nL and contained 65.4 nM rhodamine 110 internal standard and 2 µM BGTP in TGEM buffer. Separation voltage was 7 kV. (b) Rhodamine 110 peak heights and peak widths at half peak height as a function of electrophoresis buffer and segmented flow rates. Error bars are (1 standard deviation (SD).

improve throughput. These effects also give rise to narrower peaks. The main limitation to faster flow rates is that eventually too small of a fraction of the plug is injected and the resulting signal is to low to be detected. Higher voltages are limited by Joule heating or arcing. Parallel Analysis of Segmented Samples. Our goal was to achieve high sample throughput that might be attractive to HTS applications by analyzing segmented samples in parallel. For testing the potential of this system, we used a CE assay for G protein GTPase activity.26 In this assay, the fluorescent GTP analog BGTP is used as a surrogate enzyme substrate. GTPase produces BGDP from the BGTP so that electrophoretic analysis of the reaction mixture allows enzyme activity to be determined from the amount of BGDP formed within a fixed reaction time. We have previously shown that this assay can be used to detect inhibitors of the interaction between regulator of G protein signaling protein (RGS) and G proteins, suggesting the potential for use in drug discovery for this novel target.26,27 In an initial test of the system, we evaluated injection and separation stability within a single unit and reproducibility across three parallel units. We first tested stability of a single unit by serially extracting and analyzing 40 plugs containing 65 nM (26) Jameson, E. E.; Roof, R. A.; Whorton, M. R.; Hosberg, H. I.; Sunahara, R. K.; Neubig, R. R.; Kennedy, R. T. J. Biol. Chem. 2005, 280, 7712–7719. (27) Neubig, R. R.; Siderovski, D. Nat. Rev. 2002, 1, 187–197.

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

9265

Figure 5. Effect of separation voltage on CE separation of segmented samples. (a) Sample plugs of the same size and content were analyzed on chip with different separation voltage applied. Four plugs were injected at each voltage. Each plug was about 9 nL and contained 65.4 nM rhodamine 110 internal standard and 2 µM BGTP in TGEM buffer. Flow rate was 4 µL/min. (b) Effect of voltage on rhodamine 110 peak height and peak width at half peak height. Error bars are (1 standard deviation (SD).

rhodamine 110 and 2 µM BGTP in TGEM buffer. The RSDs of BGDP peak area and the ratio of BGDP to rhodamine 110 peak area were 15% and 6%, respectively. Reproducibility across three parallel units was also evaluated. Sample plugs of the same size (9 nL) and content (65 nM rhodamine 110 and 2 µM BGTP) were analyzed at flow rate of 5 µL/min and voltage of 7 kV. Absolute peak area had high variation across three channels. Relative standard deviation of rhodamine 110 peak area was as high as 29% and BGDP 30%. Relative BGDP peak area to rhodamine 110 was more suitable for quantification. The RSD of the average ratio of BGDP to rhodamine 110 for three units was 2%. To demonstrate high-throughput analysis of enzyme reactions, 100 nM GRo was incubated with 2 µM BGTP for 0, 15, and 30 min at room temperature. Three cartridges were generated each containing an array of 40 plugs, 9 nL each, for each of the three quenched reaction mixtures. Content of all three cartridges was pumped out at 4.5 µL/min to the chip for separation at 7 kV. Figure 6 shows typical parallel electropherograms (selected from 15 replicates of this experiment). Each separation took 15 s, allowing 120 samples to be analyzed within 10 min. Different amounts of BGDP formed were easily detected due to the different reaction times as shown in Figure 6. Peak areas for BGDP relative to rhodamine 110 from the three units were 0.31 ± 0.03 (n ) 38), 0.70 ± 0.03 (n ) 38), and 1.14 ± 0.06 (n ) 36) for the 0, 15, and 30 min reaction times, respectively. 9266

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

Figure 6. Parallel analysis of segmented samples. (a) 100 nM GRo was incubated with 2 µM BGTP in TGEM buffer at room temperature for 0, 15, and 30 min. An array of 40 sample plugs from each reaction mixture was analyzed in parallel on the chip shown in Figure 1. Samples were pumped into the chip at 4.5 µL/min, and 7 kV was applied for separation. Inset shows expanded view of traces. (b) BGDP peak area ratio to rhodamine 110 plotted against injection number showing variation of quantitative measurement throughout analysis.

Failed injections were occasionally observed, probably due to errors during plug extraction which resulted in unresolved peaks. Differences in quality of electropherograms were found for different channels. (See inset, Figure 6.) This effect could be ascribed to differences in derivatization quality which resulted in differences in speed of plug extraction at the interface. If analysis of such plug arrays could be extended at the same pace, this system would achieve analysis of 720 samples/h or 17 280 samples/day. This throughput would enable several interesting applications. The most immediate application would be high-throughput screening for drug discovery. Assays for enzyme inhibitors and binding interactions have been developed on the basis of electrophoresis that would be compatible with this system. Other potential applications where large numbers of assays are necessary include single cell analysis and coupling to

multiple separation columns for high-throughput two-dimensional separations. Achieving such throughput would require some improvements. Instability of the interface has prevented long trials so far. Preliminary studies suggest that the surface coating begins to lose its effectiveness after 3 to 4 h of operation. This effect may be due to loss of the surface modification caused by the basic electrophoresis buffer (pH of 8.3). (Chip performance was regained by rederivatizing the surface, allowing multiple uses of the device.) We believe that improved derivatization procedures would enable longer term operation and better RSDs.28 Other surface patterning methods such as photocatalytic patterning22,29 might also be superior even though it requires special equipment such as a mask aligner. Even higher throughput could be achieved with more channels. The current system would be compatible with aligning and fabricating up to 96 channels on the device (more with larger substrates). A limiting factor is the time and manual labor involved in interfacing sample cartridges to the chip. CONCLUSION We developed a microfabricated world-to-chip interface which coupled streams of segmented nanoliter samples to on chip CE (28) Darhuber, A. A.; Troian, S. M.; Miller, S. M.; Wagner, S. J. Appl. Phys. 2000, 87, 7768–7775. (29) Dulcey, C. S.; Georger, J. H.; Krauthamer, V.; Stenger, D. A.; Fare, T. L.; Calvert, J. M. Science 1991, 252, 551–554. (30) Shestopalov, I.; Tice, J. D.; Ismagilov, R. F. Lab Chip 2004, 4, 316–321. (31) Song, H.; Chen, D. L.; Ismagilov, R. F. Angew. Chem., Int. Ed. 2006, 45, 7336–7356. (32) Song, H.; Ismagilov, R. F. J. Am. Chem. Soc. 2003, 125, 14613–14619. (33) He, M. Y.; Sun, C. H.; Chiu, D. T. Anal. Chem. 2004, 76, 1222–1227. (34) Huang, K. D.; Yang, R. J. Electrophoresis 2008, 29, 4862–4870.

separation. This interface allows nanoliter plugs to be serially pumped into the system for electrophoretic analysis of discrete samples at 15 s intervals. The fabrication method allowed multiples of such interfaces to be integrated onto a single device which further increased the throughput to 120 samples/10 min for a model enzyme assay mixture. The passive injection system facilitated automation by eliminating the need for valves. The high throughput and low sample consumption suggest that this approach could be used as part of a HTS system; however, further work is required before use in screening such as simplified procedures for manipulating multiple cartridges and increasing the lifetime of the derivatized chip. In this work, all the samples were processed off-chip and then loaded into the cartridge. An interesting potential direction for this approach is to process the samples on chip using the methods of plug manipulation such as reagent mixing and incubation,30,31 dilution,32 and concentration.33,34

ACKNOWLEDGMENT This work was supported by NSF CHE-0514638. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review July 2, 2010. Accepted September 27, 2010. AC101755Y

Analytical Chemistry, Vol. 82, No. 22, November 15, 2010

9267