Fluidic Preconcentrator Device for Capillary Electrophoresis of

comprehensive list of citations to this article, users are encouraged to perform a search inSciFinder. ... Chan Joo Lee , Jae Hwan Jung , and Tae ...
1 downloads 0 Views 114KB Size
Anal. Chem. 2003, 75, 5207-5212

Fluidic Preconcentrator Device for Capillary Electrophoresis of Proteins Juan Astorga-Wells† and Harold Swerdlow*,‡

Center for Genomics Research, Karolinska Institutet, SE-171 77 Stockholm, Sweden

A new preconcentration device was developed for analysis of proteins by capillary electrophoresis (CE). The microfluidic device uses an electric field to capture proteins that pass through the system. The capture zone is maintained in the flow stream by the interaction between hydrodynamic and electrical forces. The device consists of a flow channel made of PEEK tubing with two electrical junctions, each of which is covered with a conductive membrane. A syringe pump provides the flow stream and also allows the injection of up to 13.5 µL of a dilute sample. The system can be easily connected to a CE device postcapture for off-line preconcentration of proteins. For the proteins used in this study, preconcentration factors up to 40-fold can be achieved. CE detection limits for bovine carbonic anhydrase, r-lactalbumin and β-lactoglobulins A and B were in the nanomolar range using UV detection at 200 nm. Preconcentration is dependent on both time and initial protein concentration. We show the possibility of using an off-line fluidic preconcentrator device employing counterflow capillary electrophoresis with minimum sample manipulation, achieving detection limits similar to on-line approaches. The post-genomic era is demanding new sensitive and selective analytical tools for protein and peptide identification. Capillary electrophoresis (CE) has already proven to be an essential tool for large-scale DNA sequencing projects as a result of its high separation efficiency, speed, potential for parallelization, low sample and buffer consumption, and the possibility of automation. With these characteristics, CE also has the potential to become an important tool for proteomics and other protein-based research.1 Although significant advances have been made toward this goal, for example, high-resolution separation of complex protein mixtures from lysates2 and integration of CE with mass spectrometry for high-throughput protein identification,3 certain issues remain to be addressed. * Corresponding author. Phone: +44-1799 532 306. Fax: +44-1799 532 301. E-mail: [email protected]. † Current address: Department of Medical Biochemistry and Biophysics, Karolinska Institutet, SE-171 77 Stockholm, Sweden. ‡ Current address: Solexa Ltd., Chesterford Research Park, Little Chesterford, Essex, CB10 1XL, U.K.. (1) Dolnik, V.; Hutterer, K. M. Electrophoresis 2001, 22, 4163-4178. (2) Shen, Y.; Xiang, F.; Veenstra, T. D.; Fung, E. N.; Smith, R. D. Anal. Chem. 1999, 71, 5348-5353. (3) Jensen, P. K.; Pasa-Tolic, L.; Peden, K. K.; Martinovic, S.; Lipton, M. S.; Anderson, G. A.; Tolic, N.; Wong, K. K.; Smith, R. D. Electrophoresis 2000, 21, 1372-1380. 10.1021/ac0300892 CCC: $25.00 Published on Web 08/23/2003

© 2003 American Chemical Society

The use of small-bore capillaries minimizes band-broadening, providing efficient dissipation of Joule heat, and allowing the application of high electric fields. While high numbers of theoretical plates are attainable, the small size of the capillaries limits the sample volume that can be injected to 1-10 nL. Coupled with a short path length for optical detection, this leads to poor detection limits. DNA can be amplified by the polymerase chain reaction (PCR), reducing the impact of this issue; however, for proteins, CE analysis is usually limited to the micromolar range. In this context, different capillary geometries, novel optical designs, and sample preconcentration methods have been developed.4 In the last 10 years, a large effort has been made in the development of on- and off-line sample preconcentration methods, making CE more attractive for real bioanalytical applications (for a review of preconcentration methods, see ref 5). Using on-line approaches, for example, field-amplified sample stacking6 or largevolume sample stacking,7 a high degree of preconcentration can be achieved, but these methods suffer problems with the analysis of high-conductivity samples. Another useful approach is solidphase extraction (SPE) using immunoaffinity resins,8 reversedphase HPLC packing materials,9 and membrane-based preconcentration methods.10 SPE-CE has some drawbacks, because the separation efficiency is highly dependent on the volume of desorption solution used,10 and electroosmotic flow reversal can occur at low pH.11 To overcome problems with on-line SPE-CE, Bonneil and Waldron12 suggest disconnection of the separation capillary from the SPE device. This semi-off-line approach avoids the requirement for application of a low-pressure hydrodynamic flow during electrophoresis (due to the reversed electroosmotic flow) and the chromatographic effects seen during the SPE elution. Although on-line methods are more suitable for automation, their main disadvantage is that the preconcentration conditions (4) Albin, M.; Grossman, P. D.; Moring, S. E. Anal. Chem. 1993, 65, 489497A. (5) Osbourn, D.; Weiss, D.; Lunte, C. Electrophoresis 2000, 21, 2768-2779. (6) Mikkers, F. S.; Everaerts, F. M.; Verheggen, T. P. J. Chromatogr. 1979, 169, 11-20. (7) Chien, R. L.; Burgi, D. S. Anal. Chem. 1992, 64, 1046-1050. (8) Guzman N. A.; Trebilcok, M. A.; Advis, J. P. J. Liq. Chromatogr. 1991, 14, 997-1015. (9) Hoyt, A. M., Jr.; Beale, S. C.; Larmann, J. P., Jr.; Jorgenson, J. W. J. Microcolumn Sep. 1993, 5, 325-330. (10) Tomlinson, A. J.; Braddock, L. M.; Benson, R.; Oda, S.; Naylor, S. J. Chromatogr., B 1995, 669, 67-73. (11) Strausbauch, M. A.; Landers, P. J.; Wettstein, P. J. Anal. Chem. 1996, 68, 306-314. (12) Bonneil, E.; Waldron, K. J. Chromatogr., B 1999, 736, 273-287.

Analytical Chemistry, Vol. 75, No. 19, October 1, 2003 5207

must be compatible with the subsequent electrophoretic step, usually limiting the choice of CE running buffer and the quantity of the sample that can be pretreated. Off-line methods can offer increased flexibility, but their main problem is the lack of suitable techniques to manipulate submicroliter samples for proper injection into a CE capillary. To overcome these problems, we have designed a small-volume fluidic preconcentrator system, able to capture proteins in a narrow-channel flow stream, without using SPE materials or membranes, allowing injection onto a CE device with minimum sample manipulation.The preconcentration device uses a form of hydrodynamic counterflow (HC) to capture charged molecules in the flow stream.13 HC has been used to retard, halt, or reverse analytes’ electrokinetic migration.14-16 As a preconcentration method, HC has been coupled with isotachophoresis-capillary electrophoresis (ITP-CE).17-19 For ITP-CE preconcentration, HC was used in the following manner. After sample injection (sample diluted in the terminating electrolyte, TE), analytes were focused without application of HC. After focusing, HC was applied to push the isotachophoretic sample zones back toward the anode. Counterflow and voltage were switched off just before the sample reached the inlet of the capillary. The TE buffer was replaced with leading buffer, and the complete length of the capillary was used for a conventional CE run. Although this method provides high concentration factors, the obligatory use of a leading buffer for the electrophoresis run and the rigorous timing required (to avoid having the sample zones migrate out of the capillary) limits this method. In this report, a form of HC is used to concentrate samples in a small device consisting of two electrified junction zones. In the junction regions, a conductive and semipermeable membrane surrounds the fluidic channel, and a pair of electrodes is situated outside the fluidic channel. An electric field of appropriate strength and direction is applied between these zones and counter to the direction of a flow stream. Charged molecules (those attracted to the upstream electrode) will resist the hydrodynamic sweeping force and will be trapped in a sharp zone within the flow stream. After capture, the molecules can be injected into an electrophoresis capillary and separated by CE using a buffer different from the one employed in the preconcentration step. EXPERIMENTAL SECTION Reagents and Chemicals. All buffers and solutions were filtered through a 0.2-µm nylon syringe filter and degassed prior to use. Water used to prepare all solutions was obtained from a Nanopure water-purification system (Barnstead, Dubuque, IA). Tris-HCl (1 M), pH 8.0, was purchased from United States Biochemicals (Cleveland, OH) and diluted before use. Horse muscle myoglobin, bovine serum albumin, human hemoglobin (13) Park, S.-R.; Swerdlow, H. Anal. Chem. 2003, 75, 4467-4474. (14) Everaerts, F. M.; Vacı´k, J.; Verheggen, T. P.; Zuska, J. J. Chromatogr. 1970, 49, 262-268. (15) Culbertson, T.; Jorgenson, J. W. Anal. Chem. 1994, 66, 955-962. (16) Chankvetadze, B.; Burjanadze. N.; Bergenthal, D.; Blaschke, G. Electrophoresis 1999, 20, 2680-2685. (17) Reinhoud, N.; Tjaden U.; van der Greef, J. J. Chromatogr. 1993, 641, 155162. (18) Hjerte´n, S.; Liao, J.; Zhang, R. J. Chromatogr., A 1994, 676, 409-420. (19) Bergmann, J.; Jaehde, U.; Schunack, W. Electrophoresis 1998, 19, 305310.

5208

Analytical Chemistry, Vol. 75, No. 19, October 1, 2003

A, bovine carbonic anhydrase, R-lactalbumin, and β-lactoglobulins A and B were purchased from Sigma Chemical Co. (St. Louis, MO). Proteins were dissolved in 10 mM Tris-HCl, pH 8.0, 0.9 mM MgCl2, and 23 mM NaCl for analysis. For CE protein separations, 50 mM borate, pH 8.6, with additive (Catalogue No. P-2063, Sigma) was used. Capture Device Fabrication and Operation. An opening was made for the anode zone with a razor blade in one side of a piece of PEEK tubing (125-µm i.d., 512-µm o.d., Upchurch Scientific, Oak Harbor, WA) at a position 2.6 cm from the end. This upstream region (anode junction) was resealed with a piece of Nafion tubing (360-µm i.d.; 510-µm o.d.; nominal pore size, 10 Å; Permapure Inc., Toms River, NY). The downstream end of the PEEK tubing (cathode junction) was covered with another piece of the conductive membrane in such a way that the Nafion tubing protruded 6 mm from the end. The two electrical junction zones were placed into separate electrode chambers, made of 0.5 mL plastic microcentrifuge tubes (Eppendorf, Hamburg, Germany) and glued in place with epoxy. At the cathode, the plastic tube was glued in such a way that the Nafion tube protruded 1 mm from the tube wall. In this way, the effective length of the cathode junction extended from the end of the PEEK tubing to the tube wall. The electrode chambers were then filled with 100 mM Tris-HCl, pH 8.0. A syringe pump (model 33, Harvard Apparatus, Holliston, MA), equipped with a 500-µL gastight syringe (Hamilton Corp., Reno, NV), produced the flow stream in the PEEK channel. The capture device was connected to the syringe using a short length of Teflon tubing (256-µm i.d., 1.58-mm o.d., Upchurch), carefully drilled to ensure good connection with the PEEK tubing of the capture device. Flow rate was set between 0.2 and 1.0 µL/min according to the manufacturer’s setting on the pump, without further calibration or measurement. Proteins loaded into the syringe pump were injected continuously into the capture device. A high-voltage power supply (CZE1000PN30, Spellman, Plainview, NY) was used for both the electrophoresis and capturedevice operation. Depending on the flow rate, 300-1000 V was applied between the junction zones such that the electrode in the upstream chamber was positive relative to the downstream electrode. Release of captured proteins was accomplished simply by turning off the high-voltage power supply. The device was used in two modes, as explained in the following paragraphs. “CE Separation Mode” describes the offline coupling of the capture device to a CE system, and “System Characterization Mode” describes the on-line coupling of a UVabsorbance detector at the outlet of the device. The latter mode was only used for optimization procedures. Once optimum conditions of voltage and flow rate were obtained, the “CE separation mode” was used for all subsequent analyses. (A) CE separation mode (Figure 1A). The outlet of the device was left open during capture for subsequent injection onto the CE system. A few seconds before the injection of the captured proteins onto the CE system, the fused-silica capillary was manually inserted into the downstream cathode connector of the PEEK capture device. After this, the high voltage was turned off. With the CE cathode reservoir situated 10 cm below the capture device level, a hydrodynamic flow injection was performed for 10 s. After injection, the capillary inlet was removed from the capture

Figure 1. Schematic representation of the capture device-CE system. Two junction zones were made using a narrow-bore PEEK tube and covered with an ion-conductive membrane, as described in the text. The zones were placed in 1.5-mL plastic micro centrifuge tubes fitted with electrodes, and the tubes were filled with 100 mM Tris-HCl, pH 8.0. A syringe pump provided continuous injection of sample. Part A shows the CE separation mode. After sample capture, the inlet of the separation capillary was positioned in the outlet of the capture device, and the sample was injected hydrodynamically. The separation capillary was disconnected from the capture device and placed into the anode buffer reservoir, and the electrophoresis voltage was applied. Figure 1B shows the system characterization mode. Here, the outlet of the capture device was directly connected to a UV-absorbance detector through a short length of fused-silica capillary.

device and placed into the CE anode buffer chamber. The electrodes from the capture device were removed and placed into the CE buffer reservoirs, such that the anode was on the inlet side of the separation capillary. Last, high voltage was applied, and electrophoresis proceeded in a conventional manner. The CE separation was performed on a homemade system with UV detection. An untreated fused-silica capillary (100-µm i.d., 375µm o.d., 36-cm effective length, 46-cm total length, Polymicro Technologies, Phoenix, AZ) was used for all runs. Electrophoresis was carried out at 200 V/cm. An HPLC UV-absorbance detector (model 1783A, Applied Biosystems, Foster City, CA) was used at 200 nm to detect proteins on-column. An optical window was made in the separation capillary by removing the polyimide coating using a small flame. Analogue output from the detector was recorded on a chart recorder (Kipp and Zonen, Holland). (B) System Characterization Mode (Figure 1B). The downstream end of the device was connected directly to a short piece of fused-silica detection capillary (100-µm i.d., 360-µm o.d., Polymicro Technologies), which was glued in place with epoxy. A UV-absorbance detector (model 229, ISCO, Lincoln, Nebraska) operating at 254-nm wavelength was used to measure protein concentration at the outlet of the device. Analogue output was digitally converted by a PCMCIA-based A/D converter (CyberResearch, Branford, CT) and processed on a notebook-PC using Labtech Notebook (Laboratory Technologies Co., Wilmington, MA). RESULTS AND DISCUSSION To improve the detection limit of CE for protein analysis, a fluidic preconcentration device was constructed. The counterflow device, originally designed for use with DNA,13 employs electrokinetic force to move certain charged molecules against the direction of hydrodynamic flow, capturing those analytes attracted to the upstream electrode. The electrical junctions are made of Nafion tubing. This ion-selective polymer has good electrical

conductance, and its pore size of only ∼10 Å20,21 will prevent migration of large molecules into the electrode chambers. Additionally, Nafion is negatively charged, restricting passage through the membrane to small cations. Small- to medium-sized proteins, like those used in our experiments, would not easily pass across the membrane, since they have an average diameter in the range of tens of angstroms and they are negatively charged at the buffer pH used. Experimental data obtained using a similar setup show that positively charged peptides also do not easily pass through this membrane.22 Electrical Properties and Theoretical Behavior. Under different voltages and flow rates, the capture device shows complex electrical behavior. In the absence of hydrodynamic flow, the system develops high electrical resistance, decreasing the current to almost zero, independent of the magnitude of the applied voltage. Since the electrodes are in electrical contact with the fluidic channel by a cation selective barrier, only cations are responsible for the charge transport between the channel and the electrode buffer chambers. Thus, a disturbance in the overall resistance should be related to a reduced outflux or influx of cations between said zones. A complete description of all the ionic interactions between the various compartments in the device is very complex. In conventional gel electrophoresis, zones of altered salt concentration are formed at the interfaces between the gel and electrolyte solutions. Such zones are formed by differences in the transference numbers between the gel solution and the liquid in the buffer chambers.23 The situation becomes even more complex when one of the media is selective to either cations or anions. Polarization, boundary effects, electroosmosis, and current-transport limits can lead to disturbances in the electric field as a result of changes in (20) Bookman, P.; Nicholson J. Developments in Ionic Polymers; Wilson, A., Prosser, H., Eds.; Elsevier Applied Science Publishers: London. 1986; Vol 2, 269-283. (21) Koter, S. J. Membr. Sci. 2000, 166, 127-135. (22) Astorga-Wells, J.; Jo ¨rnvall, H.; Bergman, T. Anal. Chem., in press. (23) Spencer, M. Electrophoresis 1983, 4, 36-41.

Analytical Chemistry, Vol. 75, No. 19, October 1, 2003

5209

Figure 2. Capture-device current versus flow rate. Current was measured after 15 min of operation at flow rates between 0 and 5 µL/min using a sample solution of 10 mM Tris-HCl, pH 8.0, and 300 V (187 V/cm) across the device.

the transference numbers for permeable ions, as described by Rubinstein et al.24 Since the Nafion membrane blocks the circulation of anions, these transference-number effects are likely to lead to high electrical resistance under zero-flow conditions in our device. Thus, from the experimental data shown in Figure 2 and the cation-selective properties of the Nafion membrane, a theoretical electrical behavior model can be proposed. With respect to the cathode buffer gap (downstream), anions cannot enter the flow channel from the cathode buffer chamber; however, they are being pulled within the peek channel toward the upstream junction. This imbalance produces a cationic depletion zone in the channel near the downstream gap (Figure 3A). Macroscopic electroneutrality forces the cations to behave in a similar fashion. Thus, there is an ion-depleted zone downstream, which leads to the development of high electrical resistance. At the other end of the capture device (anode, upstream junction), anions are moving upstream, but they cannot be removed to the anode through the Nafion membrane, and thus, they accumulate. A zone containing a high concentration of both ions forms close to the anode membrane. Being a highconductivity, low-resistance zone, this latter region has little effect on the overall electrical resistance of the device. In summary, the absence of hydrodynamic flow causes an ion concentrated zone at the anode junction (upstream) and a high electrical resistance depletion zone at the cathode junction (downstream). At 187 V/cm and flow rates 0 and 1.5 µL/min (current independent of the flow rate). Black and white spheres represent anions and cations, respectively. As explained in detail in the text, the ion-depletion zone formed by the cation-selective properties of Nafion likely creates a stacking effect and is also responsible for the low current seen at low flow rates (Figure 2).

applied on an anion have opposite direction. Certain anions situated at the depletion zone (downstream) will be subject to a hydrodynamic force of lower magnitude than the electrical force, therefore pushing these ions upstream toward the ion concentrated zone (lower electric field area). As they move upstream, the electrical force decreases. At some point, the magnitude of the electrical force will become lower than the hydrodynamic force, and they will therefore be pushed downstream where the process is repeated. By this mechanism, slow anions can be stacked or captured in a narrow band in the flow stream. A similar phenomenon is observed for ITP-CE preconcentration using hydrodynamic counterflow.17-19 Ions with mobilities between those of the leading and terminating ions are concentrated at the boundary between said zones as a result of differences in the electric field. In the present study, the leading buffer would correspond to the ion concentrated zone and the terminating buffer to the depletion zone. Further experiments must be performed in order to confirm if proteins are captured by a similar phenomenon in our system. Sample Preconcentration. An initial configuration was chosen in which negatively charged proteins would be attracted toward the upstream electrode (anode). The selected proteins were dissolved in 10 mM Tris-HCl, pH 8.0, to ensure that their net charge was negative. The use of colored proteins, such as horse muscle myoglobin and human hemoglobin A, permitted the visualization of protein

Figure 4. Image of the downstream cathode junction during capturedevice operation. A solution of 0.5 mg/mL horse muscle myoglobin, 0.25 mg/mL human hemoglobin A, and 0.4 mg/mL bovine serum albumin in 10 mM Tris-HCl, pH 8.0, was injected onto the system and captured at a flow rate of 0.4 µL/min and 300 V (187 V/cm) across the device. A few seconds after application of the voltage, a concentrated protein band was observed. Note that, at this pH, the selected proteins are attracted toward the upstream anode junction; that junction is not visible in the picture.

behavior during system operation. With a steady flow of proteins through the device, a few seconds after the voltage was turned on, a sharp band of protein was observed at the outlet (cathode side) of the PEEK tubing (see Figure 4). By increasing the applied voltage, or decreasing the flow rate, the protein band could be shifted back toward the anode junction. The minimum applied voltage necessary to capture the proteins was found to be proportional to the flow rate (data not shown). Similar results were obtained with DNA using the capture device.13 To avoid excessive Joule heating, the conditions were chosen to be 187 V/cm at 0.3 µL/min. Initial experiments, including capture device optimization, used the “system characterization mode” shown in Figure 1B. Bovine serum albumin, horse muscle myoglobin, and human hemoglobin A in 10 mM Tris-HCl, pH 8.0, either individually or mixed together, could be captured and concentrated in the system during a period of at least 45 min (total volume of sample about 13.5 µL). After turning off the voltage, a single peak from these three proteins was observed in the downstream UV detector (Figure 5). Recovery was typically >90%, and depending on the capture time, the peak height of the released protein band was between 10 and 40 times higher than the original concentration. Peak height of the released proteins was both capture-time- and proteinconcentration-dependent, with linear responses in both cases (data not shown). Interfacing a home-built CE system to the capture device was accomplished by connecting the outlet of the system to the separation capillary inlet (“CE separation mode” and Figure 1A). After a 10-s hydrodynamic injection, the capillary was disconnected from the capture device and placed into the electrophoresis running buffer, and high voltage was applied. Leaving the syringe pump on during injection did not interfere with the separation. Sample volume injected onto the CE system did not vary significantly from a standard 10-s hydrodynamic injection (see below). Overall reproducibility of the analysis was also good; peak-

Figure 5. Typical capture-device current and capture/release profiles using the system characterization mode. A fused-silica capillary was attached to the end of the device, and a UV-absorbance detector (254 nm) monitored protein as it left the device. A solution of 0.5 mg/mL horse muscle myoglobin, 0.25 mg/mL human hemoglobin A, and 0.4 mg/mL bovine serum albumin in 10 mM Tris-HCl, pH 8.0, was injected continuously at 0.3 µL/min and 167 V/cm into the device, giving a steady level of protein passing through the detector. Approximately 110 s after the application of voltage (at 450 s), a decrease in the signal indicates that protein is being captured by the device. Shortly after turning off the voltage supply (at 640 s), a single peak of concentrated proteins was observed at the detector.

height reproducibility after capture, injection, and CE separation was 4.4% RSD (n ) 10). A mixture of 40 µg/mL bovine carbonic anhydrase, R-lactalbumin, and β-lactoglobulin A and B was preconcentrated for varying times, injected onto the CE system, and separated (Figure 6A). Additionally, using a fixed capture time (18 min), varying concentrations of the same protein solution were analyzed (Figure 6B). The experiments show that peak heights were both capturetime- and protein-concentration-dependent, with excellent linearity (r2 ) 0.99, in both cases). As in the system characterization mode experiments, preconcentration factors between 10- and 40-fold could be achieved. Limits of detection (S/N ratio ) 3) were 1.5 µg/mL for β-lactoglobulin A and B (80 nM) and 0.8 µg/mL for bovine carbonic anhydrase and R-lactalbumin (26 and 56 nM, respectively). An example of the separation and preconcentration obtained is shown in Figure 7B and compared to a standard hydrodynamic injection (Figure 7A). Since the height of the “injection” peak obtained from the sample buffer (asterisk in Figure 7) was similar in Figure 7A,B (in both, 10 s of hydrodynamic flow injection was used), we can conclude that the volume of sample injected in the experiments was equivalent. However, the amount of protein injected was at least 40 times higher using the capture conditions compared to the conventional injection. The concentration factor can likely be improved simply by increasing the volume of sample injected into the device, through an increase in either flow rate or injection time. However, an increase in the flow rate necessitates an increase in the capture electric field, leading to an increase in both current and Joule heating. We have observed that current values higher than 200 µA lead to unstable capture because of bubble formation. Therefore, it is preferable to increase injection time in order to increase the preconcentration factor (Figure 6A). There is no obvious plateau in the peak-height curve of Figure 6A; therefore, Analytical Chemistry, Vol. 75, No. 19, October 1, 2003

5211

Figure 6. Protein peak height versus capture time (A) and protein concentration (B), CE separation mode. In panel A, a solution of 40 µg/mL of each protein was captured for 15, 25, 35, and 45 min at 0.3 µL/min and 167 V/cm, injected onto the CE capillary for 10 s, as described in the text, and analyzed. In panel B, protein solutions of 10, 20, 30, and 40 µg/mL were captured for 18 min under the same conditions, injected onto the CE system for 10 s, and analyzed. Regression coefficients of r2 > 0.99 were obtained for the linear fits shown for all four plots.

we believe that we can preconcentrate samples for much longer than was done in this report, likely even for a few hours. Since proteins are captured into a sharp band, it is expected that aggregation or precipitation of proteins would limit the ultimate concentration factors that are achievable. CONCLUSION This work has demonstrated the feasibility of using a novel fluidic preconcentrator device for capillary electrophoresis of proteins. Using the polarity configuration described here, the device can electrokinetically capture negatively charged proteins from a flow stream. These proteins can then be injected onto a CE system and separated. The two-step procedure used here provides a degree of independence between sample enrichment and CE separation and increases the flexibility in the choice of CE running buffer. Detection limits similar to those obtained using on-line preconcentration methods can be achieved with the present method.1,18,19 In addition, the microfluidic device is easily assembled in about 30 min and has a lifetime of several months. Joule heating due to current in the system limits the capturedevice operation. Future efforts will be focused on microfabrication using silica or polymeric substrates. Small-diameter channels will allow preconcentration of higher conductivity samples. Further experiments are needed to validate the proposed theoretical behavior of the microfluidic system. In addition to the presented application, the device can be modified for use in sample preparation of peptides and proteins for analysis by MALDI-MS.22 5212 Analytical Chemistry, Vol. 75, No. 19, October 1, 2003

Figure 7. Electropherograms obtained from model proteins separated on the CE system using a conventional gravity hydrodynamic flow injection method (A) and the capture device (B). For both experiments, 3 µg/mL of each protein was dissolved in 10 mM TrisHCl, pH 8.0: (1) bovine carbonic anhydrase, (2) R-lactalbumin, (3) β-lactoglobulin A, (4) β-Lactoglobulin B, and (*) Tris buffer peak. For the hydrodynamic injection (A), the sample vial was elevated 10 cm above the cathode vial for 10 s. Capture-device preconcentration (B) was performed for 30 min at 0.3 µL/min and 187 V/cm. Hydrodynamic injection at 10 cm for 10 s was performed as described in the text.

Finally, the possibility to capture proteins in a sharp zone within a flow stream, without the use of solid supports or chemical bonding, suggests a new principle to manipulate molecules in micrometer-size structures, which is compatible with large-scale production. ACKNOWLEDGMENT This work was supported by grants from Pharmacia & Upjohn Corp. (N344010) to the Center for Genomics Research, Karolinska Institutet and from the Swedish Research Council (projects 03X3532, B5101-879/2001, and K5104-20005891). J.A.W. is grateful to Hans Jo¨rnvall and Tomas Bergman, Department of Medical Biochemistry and Biophysics at Karolinska Institutet for additional help and discussions.

Received for review March 6, 2003. Accepted June27, 2003. AC0300892