Anal. Chem. 2000, 72, 2514-2518
Microfabricated Electrophoresis Chips for Simultaneous Bioassays of Glucose, Uric Acid, Ascorbic Acid, and Acetaminophen Joseph Wang,* Madhu Prakash Chatrathi, Baomin Tian, and Ronen Polsky
Department of Chemistry and Biochemistry, New Mexico State University, Las Cruces, New Mexico 88003
A micromachined capillary electrophoresis chip is described for simultaneous measurements of glucose, ascorbic acid, acetaminophen, and uric acid. Fluid control is used to mix the sample and enzyme glucose oxidase (GOx). The enzymatic reaction, a catalyzed aerobic oxidation of glucose to gluconic acid and hydrogen peroxide, occurs along the separation channel. The enzymatically liberated neutral peroxide species is separated electrophoretically from the anionic uric and ascorbic acids in the separation/reaction channel. The three oxidizable species are detected at the downstream gold-coated thickfilm amperometric detector at different migration times. Glucose can be detected within less than 100 s, and detection of all electroactive constituents is carried out within 4 min. Measurements of glucose in the presence of acetaminophen, a neutral compound, are accomplished by comparing the responses in the presence and absence of GOx in the running buffer. The reproducibility of the on-chip glucose measurements is improved greatly by using uric acid as an internal standard. Factors influencing the performance, including the GOx concentration, field strength, and detection potential, are optimized. Such coupling of enzymatic assays with electrophoretic separations on a microchip platform holds great promise for rapid testing of metabolites (such as glucose or lactate), as well as for the introduction of high-speed clinical microanalyzers based on multichannel chips. An enormous amount of research has been devoted to the development of new diagnostic tools for the benefit of diabetic patients.1,2 As a result, new biosensors and bioassays have been introduced for monitoring glucose. Most often, diabetics selfmonitor their blood glucose using disposable amperometric biosensors connected to a hand-held electrochemical meter.3 Such biosensors are generally based on screen-printed electrode strips with immobilized glucose oxidase (GOx) and a proper electron mediator.4 While returning the chemical information rapidly, such devices have limits in versatility and scope, due to a minimal sample-handling capability (associated with their limited surface chemistry). For example, electrooxidizable constituents (e.g., (1) Wang, J. Anal. Chem. 1999, 71, 328R. (2) Henry, C. Anal. Chem. 1999, 71, 5009. (3) Kirk, J.; Rheney, C. C. J. Am. Pharm. Assoc. 1998, 38, 210. (4) Lewis, B. D. Clin. Chem. 1992, 38, 2093.
2514 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
ascorbic acid) in physiological fluids commonly interfere with the amperometric assay. A more versatile approach is to pretreat the sample and perform a separation. If neither the speed nor the size requirements are compromised, and if additional clinically relevant species could be measured simultaneously, such an approach should greatly enhance the power of glucose diagnostic devices, and of “point-of-care” clinical microanalyzers, in general. In this report, we describe a new chip-based protocol, based on the coupling of enzymatic bioassays and electrophoretic separations, for rapid measurements of metabolites such as glucose. Microfabricated fluidic devices integrating the sample preparation process and the measurement step onto microchip platforms are of considerable recent interest.5,6 Particular attention has been given to micromachined capillary electrophoresis (CE) chips, due to their remarkable separation and sample-handling capabilities.7 These devices can integrate the sample injection, separation, and pretreatment (mixing/derivatization, dilution, etc.) on a planar chip platform. Such miniaturization greatly reduces the time of CE separations (to periods approaching those of conventional biosensors) and results in a negligible reagent consumption. Microchip separation devices have previously been used for performing enzyme (galactosidase or acetylcholinesterase) inhibition assays8,9 but not in connection with measurements of glucose or other clinically important metabolites. A conventional CE system with on-column and precolumn enzymatic reactions and laser-induced fluorescence detection has been reported for measurements of glucose.10 The combination of immunoassays with microchip CE separations has also been reported.11 In the following sections, we will demonstrate the ability to quantitate simultaneously glucose, uric acid, and ascorbic acid on a biochip platform (Figure 1). In this glass microchip system, the sample and the enzyme GOx are mixed; the neutral glucose substrate and hydrogen peroxide are then separated by electrophoresis from the anionic urate and ascorbic species, which migrate at a slower rate (Figure 2). The three analytes are readily measured at the downstream thick-film electrochemical detector (5) Freemantle, M. Chem. Eng. News 1999, 77 (Feb 22), 27. (6) Kovacs, G. T. A.; Petersen, K.; Albin, M. Anal. Chem. 1996, 68, 407A. (7) Effenhuser, C.; Manz, A.; Widmer, H. Anal. Chem. 1993, 65, 2637. (8) Hadd, A. G.; Raymond, D. E.; Halliwell, J. W.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1997, 69, 3407. (9) Hadd, A. G.; Jacobson, S. C.; Ramsey, J. M. Anal. Chem. 1999, 71, 5206. (10) Jin, Z.; Chen, R.; Colon, L. A. Anal. Chem. 1997, 69, 1326. (11) Chiem, N. H.; Harrison, D. J. Clin. Chem. 1998, 44, 591. 10.1021/ac991489l CCC: $19.00
© 2000 American Chemical Society Published on Web 04/20/2000
Figure 1. Layout of the separation/reaction microchip for bioassays of glucose, ascorbic acid, uric acid, and acetaminophen. See text for details.
Figure 2. Enzymatic and separation processes along the reaction/ separation channel of the CE biochip.
within less than 4 min, with glucose alone under 60 s. Comparison of the responses with and without GOx allows sequential measurements of the neutral glucose and acetaminophen species. The quantitation of uric acid and acetaminophen in biological fluids is also of clinical significance.12,13 The operating conditions and benefits of the new glucose biochip are discussed below. Such on-chip coupling of enzymatic assays, electrophoretic separations, and electrochemical detection represents a conceptually new approach for bioanalysis. While the concept is presented within the framework of rapid glucose testing, it should eventually lead to complete high-throughput clinical microanalyzers based on parallel channels and multiple runs on single-chip platforms. EXPERIMENTAL SECTION Apparatus. A homemade high-voltage power supply was used for controlling the separation and had a potential range between (12) Nakaminami, T.; Ito, S.; Kuwabata, S.; Yoneyama, H. Anal. Chem. 1999, 71, 4278. (13) Gilmartin, M.; Hart, J. P. Analyst 1994, 119, 2431.
0 and +4000 V. Safety Considerations: The high-voltage power supply should be handled with extreme care to avoid electrical shock. Amperometric detection was performed with a model 621 electrochemical analyzer (CH Instruments, Austin, TX) connected to a Pentium 166 MHz computer with 32 MB RAM. Chip Design. A schematic of the biochip is shown in Figure 1. The chip consisted of a four-way injection cross, followed by a 72 mm long, 50 µm wide reaction/separation channel. The chip, model MC-BF4-001, was fabricated by the Alberta Microelectronic Co. (AMC, Edmonton, Canada) using wet chemical etching and thermal bonding techniques. The original waste reservoir was cut off by AMC, leaving the channel outlet at the end side of the chip (see Figure 1), thus facilitating the end-column amperometric detection. The amperometric detector was located in the waste reservoir (at the channel outlet side) and consisted of an Ag/ AgCl wire reference electrode, a platinum-wire counter electrode, and a gold-modified screen-printed carbon working electrode. The screen-printed working electrode was placed opposite the channel outlet; the distance between the electrode surface and the channel outlet was controlled by a plastic screw and a thin-layer spacer (50 µm). A Plexiglas holder was fabricated for holding the separation chip and housing the detector and the reservoirs. A short pipet tip was inserted into each of the three holes on the glass chip for solution contact between the channel on the chip and the corresponding reservoir on the chip holder. Platinum wires, inserted into the individual reservoirs, served as contacts to the high-voltage power supply. Additional details of the integrated thick-film detector/CE chip microsystem were given recently.14 Reagents. Glucose, glucose oxidase (GOx), uric acid, ascorbic acid, and acetaminophen were obtained from Sigma. The gold atomic absorption standard solution (1000 mg/L) was purchased from Aldrich. All chemicals were used without further purification. A phosphate buffer (pH 7.4, 10 mM) solution served as the electrophoresis buffer. Stock solutions were prepared daily in the electrophoresis buffer and filtered with 0.45 µm filter (Gelman Acrodisc). Sample solutions were prepared by diluting the corresponding stock solutions with the phosphate buffer. Screen-Printed Electrodes. The screen-printed electrodes were fabricated with a semiautomatic printer (model TF 100, MPM, Franklin, MA). An Acheson ink Electrodag 440B (49AB90) (Acheson Colloids, Ontario, CA) was used for printing electrode strips. Details of the printing processes were described previously.14 The detector strip (0.4 in. × 1.333 in.) contained the carbon-band (line) working electrode and its silver contact printed on a ceramic substrate. The active working electrode area (0.30 mm × 2.50 mm) was defined by a layer of insulator. The carbon working electrode area was coated with a gold film, prepared by applying a pulse waveform (square-wave pulse potential between -0.2 and +0.75 V, vs Ag/AgCl, with a pulse width of 0.6 s) for 20 min in a solution containing 300 ppm of Au(III), 0.1 M NaCl, and 1.5% (w/v) HCl. Electrophoresis Procedure. The channels were treated before use by rinsing with 0.1 M sodium hydroxide and deionized water for 20 and 5 min, respectively. To perform the separation, the buffer reservoir was filled with the phosphate buffer solution, while the buffer-with-GOx reservoir was filled with phosphate (14) Wang, J.; Tian, B.; Sahlin, E. Anal. Chem. 1999, 71, 5436.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2515
buffer containing 75 U/mL glucose oxidase (Figure 1). The sample reservoir was filled with mixtures containing glucose, uric acid, ascorbic acid, and/or acetaminophen. The detection/waste reservoir was filled with the phosphate buffer solution. Other reservoirs were filled with 200 µL of the corresponding solutions, thus maintaining equal hydrostatic levels. The initial filling of the injection channel (between the separation channel and the sample reservoir) with the sample solution was achieved by applying a potential of +1500 V for 20 s to the sample reservoir with the detection reservoir grounded and other reservoirs floating. The actual assays were performed by loading the sample plug into the separation/reaction channel, by applying +1500 V to the sample reservoir for 2 s (with the detection reservoir grounded and other reservoirs floating). Subsequently, for simultaneous measurements of glucose, uric acid, and ascorbic acid, the separation voltage was applied to the buffer-with-GOx reservoir. Mixing of the glucose substrate (in the sample plug) with the enzyme (in the running buffer) started at the intersection and proceeded primarily down in the separation channel. The neutral hydrogen peroxide species (produced in the separation/reaction channel) and the uric and ascorbic acids were separated in the separation/reaction channel (Figure 2), and the three oxidizable species were detected amperometrically at different migration times. Alternately, measurements of glucose and acetaminophen were carried out by comparing the responses with and without the enzyme. For this purpose, a total signal was measured with the running buffer containing GOx, while the acetaminophen signal alone was recorded by applying the separation voltage to the buffer reservoir (containing no GOx). The current difference was used for quantifying the glucose concentration in the sample mixture. Electrochemical Detection. The electropherograms were recorded with a time resolution of 0.1 s while the detection potential was applied (usually +0.9 V vs Ag/AgCl wire). Sample injections were performed after stabilization of the baseline. No software filtration of the signal was used. All bioassays were carried out at room temperature. RESULTS AND DISCUSSION The ultimate goal of this effort is to develop miniaturized clinical analyzers, by coupling enzymatic assays and electrophoretic separations on a chip platform. While such an attractive combination is demonstrated below for the rapid testing of glucose, parallel operations (using multichannel networks) should lead to the simultaneous measurements of other clinically relevant metabolites (e.g., lactate, cholesterol, and creatinine). In the present single-channel chip manifold (Figure 1), the enzyme (GOx)/running buffer and sample solutions are mixed at the channel intersection and in the separation channel using electrokinetic flow. The enzymatic reaction occurs along the separation/ reaction channel while the enzyme (in the running buffer) and glucose (in the sample plug) diffuse downstream: GOx
glucose + oxygen 98 hydrogen peroxide + gluconic acid (1)
The enzymatically liberated neutral peroxide species is electro2516 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
Figure 3. Electropherograms for 1 × 10-3 M glucose (a) solutions containing increasing levels of ascorbic acid (b) and uric acid (c): 2 × 10-4 M (A); 5 × 10-4 M (B); 7 × 10-4 M (C). Conditions: running buffer, a phosphate buffer (10 mM, pH 7.4) solution, containing 75 U/mL GOx; separation potential, +1500 V; injection potential, +1500 V; injection time, 2 s; detection potential, +0.9 V.
phoretically separated from the anionic uric and ascorbic acids in the separation/reaction channel (Figure 2), and the three oxidizable species are detected at the downstream working electrode at different migration times. Figure 3 displays typical electropherograms obtained for 1 × 10-3 M glucose solutions containing increasing levels of ascorbic acid and uric acid (2 × 10-4-7 × 10-4 M, B-D). These electropherograms indicate convenient and rapid separation and detection of all three compounds, with a total time of around 4 min using a separation potential of +1500 V. Glucose alone is actually detected within less than 100 s. Under these conditions, the microsystem offers a nearly complete isolation from the high separation potential, as indicated from the flat baseline and low noise level. The increasing levels of ascorbic and uric acids do not affect the glucose response. (Note that these levels greatly exceed the physiological levels of ascorbate and urate and can affect the accuracy of amperometric glucose biosensors.15) Note also that the current peaks of ascorbic and uric acids are proportional to their concentrations. Detailed calibration experiments will be presented in the following sections. Acetaminophen, a common neutral interference,15 is carried solely by the electroosmotic flow and cannot be resolved from the neutral glucose/peroxide species. Selective measurements of glucose in the presence of acetaminophen can be performed by comparing the responses in the presence and absence of GOx in the separation buffer [Figure 4 (A, B, D vs C)]. Well-defined peaks with identical migration times (∼100 s) are observed for glucose (A) and acetaminophen (D) using the GOx-containing buffer; a mixture of the two compounds thus yielded the expected additive response (B). As expected, glucose was not detected in the absence of GOx (not shown); accordingly, only an acetaminophen contribution was observed for the glucose/acetaminophen mixture (C). The glucose signal (in such mixtures) can be readily obtained by the difference in the responses with and without the enzyme (B - C). Notice that this current difference is identical to the current of glucose alone in the presence of GOx (A). Such a (15) Wang, J.; Liu, J., Chen, L.; Lu, F. Anal. Chem. 1994, 66, 3600.
Figure 4. Measurements of glucose and acetaminophen: portions of electropherograms recorded in the presence (A, B, D) and absence (C) of GOx in the running buffer. Sample solutions: (A) 1 × 10-3 M glucose; (B) 1 × 10-3 M glucose and 1 × 10-4 M acetaminophen; (C, D) 1 × 10-4 M acetaminophen. Other conditions were the same as those for Figure 3.
Figure 6. Influence of glucose oxidase concentration (in the running buffer). Other conditions were the same as those for Figure 3.
Figure 7. Reproducibility of the current response for 1 × 10-3 M glucose (a) and 5 × 10-4 M uric acid (b) and of the glucose-to-uric acid peak ratio (c). Asterisks denote the use of deliberately shorter sample injection times. Other conditions were the same as those for Figure 3.
Figure 5. Effect of the separation voltage. The sample mixture contained 1 × 10-3 M glucose, 6 × 10-4 M ascorbic acid, and 4 × 10-4 M uric acid. Separation voltages: (a) +1000, (b) +1500, (c) +2000, (d) +2500, and (e) +3000 V. Other conditions were the same as those for Figure 3.
strategy is applicable for addressing the interference of all other coexisting neutral electroactive species and for assays of other clinically relevant substrates. Yet, discrimination among individual neutral species is not possible. Figure 5 depicts the influence of the separation potential upon the separation efficiency and overall chip performance. As expected, increasing the separation potential from 1000 to 3000 V (in 500 V increments, a-e) decreases dramatically the migration time for glucose from 135 to 45 s. The corresponding times for ascorbic and uric acids are reduced from 330 and 360 to 105 and 115 s, respectively. The separation efficiency, represented by the plate number, decreases from 6000 to 1100 (for ascorbic acid) and from 5200 to 1050 (for uric acid) upon raising the separation potential from 1000 to 2500 V. The broader glucose/peroxide peak, associated with the enzymatic reaction, is characterized by smaller plate numbers ranging from 100 (at 2500 V) to 800 (at 1000 V). The relatively low separation efficiency is attributed to the floated injection (used for facilitating the introduction of the anionic species). The separation potential has a negligible effect upon the background noise level. However, a larger initial baseline slope
is observed for potentials ranging from 2000 and 3000 V, indicating an incomplete isolation at high separation potentials. Most subsequent work employed a potential of 1500 V. This separation potential offers convenient glucose measurements within less than 100 s. An even faster glucose detection (on a time scale of 1 min) would be expected upon improving the separation efficiency and the injection protocol and/or providing a connection to a precolumn enzymatic reaction. Note that the response time of common amperometric glucose biosensors is 30-45 s.3 Figure 6 examines the effect of the GOx level in the reagent solution upon the response to the 1 × 10-3 M glucose substrate. The current increases rapidly upon raising the GOx concentration between 0 and 50 U/mL, then increase more slowly, and finally starts to level off above 125 U/mL. All subsequent work employed 75 U/mL GOx. Enzyme levels higher than 100 U/mL resulted in increased background noise and adsorption onto the channel walls. Hydrodynamic voltammograms (i.e., plots of current response versus the applied potential) were used for selecting the detector potential. The gold-coated carbon detector displayed a defined wave-shaped voltammogram for glucose, with the current starting at +0.50 V and leveling off above +0.80 V (not shown). Most work employed a detection potential of +0.90 V, in view of the high background noise above +1.0 V. The bare carbon surface required higher potentials for the peroxide detection, with the current starting at +0.80 V and a plateau above +1.20 V. A problem associated with electrokinetic injections (particularly floated ones) is the run-to-run variations in the injection time. Such Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2517
Figure 8. Calibration data for glucose (A), ascorbic acid (B), and uric acid (C). Portions of the electropherograms corresponding to successive 1 × 10-4 M concentration increments (a-e). Also shown (insets) are the resulting calibration plots. Other conditions were the same as those for Figure 3.
changes commonly result in relative standard deviations of ∼10%. A deliberate addition of internal standards was suggested for addressing these variations.16 In the present work, we present an approach in which one of the analytes (uric acid) is used as a “built-in” internal standard. As indicated from Figure 7, such use of uric acid greatly improves the reproducibility of repetitive glucose measurements in connection with measurements of the glucose/urate peak ratio. Both the glucose and uric acid peaks yielded relative standard deviations (RSDs) of 10.6 and 10.5%, respectively; an RSD of 4.2% was estimated for the glucose/urate peak ratio. Note that deliberately shorter injection times were used in three of the measurements of Figure 7. Smaller RSDs would be expected upon using the same injection times. The concentration dependence is examined in Figure 8, which shows electrophoretic peaks for increasing levels of glucose (A), ascorbic acid (B), and uric acid (C) in 1 × 10-4 M steps (a-e). Highly linear calibration plots are observed for both ascorbic and uric acids (see insets), with slopes of 21.6 and 21.8 nA/mM (and correlation coefficients r of 0.998 and 0.999, respectively). In (16) Dose, E. V.; Guiochon, G. A. Anal. Chem. 1991, 63, 1154. (17) Bright, H. J.; Gibson, Q. J. Biol. Chem. 1967, 242, 994.
2518 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
contrast, and as expected for biocatalytic reactions, the response for the glucose substrate displays a curvature at concentrations higher than 7 × 10-4 M. A highly linear response is observed for the initial linear portion (slope 17.3 nA/mM; r ) 0.999). Greater deviations from linearity (above 1 × 10-4 M glucose) have been reported for conventional CE/fluorescence bioassays of glucose.10 While conventional biosensors rely on mass-transport-limiting membranes for extending the linear range, biochips should rely on simple dilution for imparting wider dynamic ranges in connection with biocatalytic assays. Sample dilution could thus be readily integrated (with an additional channel to control dilution). The on-chip Km value, estimated from the corresponding Lineweaver-Burk plot, 6 × 10-3 M, is lower than the value (2.6 × 10-2 M) reported for GOx in solution.17 Discrepancies between on-chip and literature Km values were reported for other enzymes.8 Detection limits of 6 × 10-6 M glucose and 5 × 10-6 M ascorbic and uric acids were estimated on the basis of the signal-to-noise characteristics (S/N ) 3) of an electropherogram for a mixture containing 4 × 10-5 M glucose and 2 × 10-5 M ascorbic and uric acids (not shown). Such micromolar detection limits permit convenient on-chip sample dilution as desired, for example, for extending the linear range. Even lower detection limits would be expected upon using an on-chip precolumn enzymatic reaction (instead of the on-column one) in a manner analogous to that used for improvements in conventional CE systems.10 In conclusion, we have described a conceptually new approach for performing miniaturized bioassays of glucose using separation microchips. Electroosmotic flow has been used for mixing the sample with the enzyme glucose oxidase (GOx) and for separating the neutral hydrogen from the anionic ascorbate and urate species. The versatility resulting from such coupling of on-chip enzymatic assays, electrophoretic separations, and amperometric detection offers great promise for decentralized testing of glucose. Additional sample manipulations (e.g., cleanup by filtration or extraction) may be added to the chip platform, as needed to address the complexity of biological fluids. Work is in progress toward the reduction of the assay time, extension of the new approach to additional metabolite/enzyme systems, miniaturization of the potentiostatic circuitry and power supply, and design of multichannel separation chips for the simultaneous measurements of multiple analytes. Such new CE microchips could compete with traditional benchtop analyzers, as well as with conventional biosensors, in terms of performance, speed, sample volume, and size. ACKNOWLEDGMENT This project was supported by the National Institutes of Health (NIH Grant RO1 RR14173-0).
Received for review December 29, 1999. Accepted March 8, 2000. AC991489L
