Anal. Chem. 1996, 68, 3832-3839
Miniaturized Biosensors Employing Electropolymerized Permselective Films and Their Use for Creatinine Assays in Human Serum Marcel B. Maˇdaˇras¸ and Richard P. Buck*
Department of Chemistry, University of North Carolina, Chapel Hill, North Carolina 27599-3290
Miniaturized, disposable amperometric biosensors for determination of creatinine in human serum are described. The base electrodes are fabricated using microelectronics techniques, to build a multilayer film structure on a polyimide foil. By using a thin electropolymerized film of poly(1,3-diaminobenzene), the electrochemical interferences from ascorbate, urate, acetaminophen, and other oxidizable species are greatly diminished. The multienzyme system (creatininase, creatinase, sarcosine oxidase) is immobilized on top of the permselective layer using cross-linking of the proteins with glutaraldehyde. The electropolymerization conditions for obtaining almost ideal permselectivity of the inner layer are defined, as well as the optimal enzyme layer preparation. A composite polymeric outer membrane [Nafion + poly-(2-hydroxyethyl methacrylate)] is used for diffusion control and to protect the enzyme layer from fouling. The reagentless planar sensors for creatinine and creatine have fast response time (t95 ) 1 min), linear response up to 1.2 mM in batch-type and 2.0 mM in flow injection analysis and a detection limit of 10-20 µM. They are applied in a differential setup for creatinine assay in control and hospital human serum samples and are suitable for incorporation in a portable analyzer. Measurement of the creatinine levels in serum and determination of the renal clearance of creatinine are widely used for laboratory diagnosis of renal and muscular function. The assay of creatinine in human serum samples is a challenging analytical problem, due to the possible interferences and the fact that creatine is also present in the serum at approximately the same concentration levels as creatinine. The creatinine reference range in serum for adults is between 0.6 and 1.2 mg/dL (53-106 µM); men have slightly higher values than women.1 In clinical practice, there are two significant decision levels: values above 140 µM indicate the necessity of performing other tests for assessment of renal function such as the creatinine clearance test; values above 530 µM are almost invariably associated with severe renal impairment. Most creatinine determinations in the clinical laboratory and almost all the commercially available analyzers use spectrophotometric procedures (either end point or kinetic) based on the Jaffe´ reaction.2 The method monitors the absorbance of the (1) Whelton, A.; Watson, A. J.; Rock, R. C. In Tietz Textbook of Clinical Chemistry, 2nd ed.; Burtis, C. A., Ashwood, E. R., Eds.; W. B. Saunders & Co.: Philadelphia, PA, 1994, pp 1531-75.
3832 Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
addition complex resulting from the reaction between creatinine and picrate in alkaline solution. However, this reaction is not specific for creatinine because many substances could interfere in this assay. The long list of these so-called “non-creatinine chromogens” includes R-keto acids such as acetoacetate and pyruvate, bilirubin, dopamine, glucose, cephalosporines, and others.3 Although kinetic methods and different pretreatments of the sample were developed to address the interferences problem, the determination of the “true” creatinine concentration in serum remains a problem. Beside the well-established spectrophotometric procedure, separation methods such as HPLC4 and HPCE5 have been described for the creatinine determination. These methods are not well-suited for large-scale routine clinical analysis but are useful as reference techniques. Several enzymatic methods have been proposed,6 but they are time-consuming and rather costly and the elimination of interferences in the determination of creatinine in serum has still not been achieved. In comparison with the homogeneous enzymatic assays, biosensors promise reduction of the time and cost of clinical analyses. In recent years, a new class of portable clinical analyzers has been designed to provide rapid turnaround time with the precision and accuracy comparable to the benchtop analyzers. These new systems require integrated and miniaturized sensing elements. With this aim in mind, we have recently described7 a first generation of creatine and creatinine amperometric biosensors using microelectronics techniques for the fabrication of the base electrodes. These sensors had a relatively long response time and high sensor-to-sensor variability mainly due to the manual deposition of the inner permselective layer. Several other amperometric sensors for creatinine determination have been proposed.8-12 All of these biosensors are based on the enzyme catalytic sequence introduced by Tsuchida and Yoda:8 (2) Jaffe´, M. Z. Physiol. Chem. 1886, 10, 391-400. (3) Weber, J. A.; van Zanten, A. P. Clin. Chem. 1991, 37, 695-700. (4) (a) Ambrose, R. T.; Lauff, J. J.; Kasper, M. E. Clin. Chem. 1983, 29, 2569. (b) Diez, M. T.; Arin, M. J.; Resines, J. A. J. Liq. Chromatogr. 1992, 15, 1337-50. (5) (a) Shi, H.; Ma, Y.; Ma, Y. Anal. Chim. Acta 1995, 312, 79-83. (b) Miyake, M.; Shibukawa,A.; Nakagawa, T. J. High Resolut. Chromatogr. 1991, 14, 181-5. (6) (a) Jeppesen, M. T.; Hansen, E. H. Anal. Chim. Acta 1988, 214, 147-59. (b) Shmuzu, S.; Kim, J. M.; Yhamado, H. Clin. Chim. Acta 1989, 185, 24152. (c) Fossati, P.; Prencipe, L.; Berti, G. Clin. Chem. 1983, 29, 1494-6. (7) Madaras, M. B.; Popescu, I. C.; Ufer, S.; Buck, R. P. Anal. Chim. Acta 1996, 319, 335-45. (8) Tsuchida, T.; Yoda, K. Clin. Chem. 1983, 29, 51-5. (9) Yamato, H.; Ohwa, M.; Wernet, W. Anal. Chem. 1995, 67, 2776-80. (10) Nguyen, V. K.; Wolff, C. M.; Seris, J. L.; Schwing, J. P. Anal. Chem. 1991, 63, 611-4. S0003-2700(96)00239-9 CCC: $12.00
© 1996 American Chemical Society
CA
creatinine + H2O \ y z creatine Cl
creatine + H2O 98 sarcosine + urea
(1) (2)
SO
sarcosine + H2O + O2 98 glycine + HCHO + H2O2 (3)
where CA ) creatinine amidohydrolase, CI ) creatine amidinohydrolase, SO ) sarcosine oxidase. Tsuchida and Yoda8 and Nguyen and co-workers10 have shown results from application of their sensors to human serum measurements. Nguyen sensors were based on the amperometric response to O2, and because of the large size of the electrode, the simultaneous measurement of creatinine and creatine concentrations in serum was not possible. Tsuchida and Yoda have used their multienzyme membranes in conjunction with the probes of a YSI glucose analyzer (Model 23A). Recently, Nova Biomedical Co. (Waltham, MA) introduced the first commercial electrodebased creatinine sensor for its clinical benchtop analyzer (Nova 16). Their “membrane sandwich”-type sensor uses the same enzymatic scheme (eqs 1-3) for the creatinine assay and a classical H2O2-detecting electrode. In this paper we present results from the determination of creatinine with miniaturized disposable biosensors (“biochips”) using electropolymerized permselective films. Urban and coworkers recently described miniaturized biosensors for other analytes (glucose, lactate, glutamate, glutamine) using similar electropolymerized films,13 but no details about their permselectivity were provided. We describe a detailed study of the conditions for electropolymerization and present results from the determination of creatinine in hospital human serum samples in addition to the control serum assay. Measurements in real samples are more challenging to perform since each sample has a different matrix (interferences level). EXPERIMENTAL SECTION Reagents. Creatinine amidohydrolase (creatininase, EC 3.5.2.10., from Pseudomonas sp.), creatine amidinohydrolase (creatinase, EC 3.5.3.3., from Actinobacillus sp.), sarcosine oxidase (EC 1.5.3.1., from Bacillus sp.), creatinine hydrochloride, creatine hydrate, 1,3-diaminobenzene (1,3-DAB), 1,3-benzenediol (resorcinol), acetaminophen (APAP), L-ascorbic acid (AA), uric acid (UA), bovine serum albumin (BSA), and glutaraldehyde (GA) 25% aqueous solution were purchased from Sigma Chemical Co. (St. Louis, MO). Hydrogen peroxide, 30% aqueous solution, was from Mallinckrodt (Paris, KY). Dade Moni-Trol ES Level I Chemistry Control serum was from Baxter Diagnostics, Inc. (McGaw Park, IL). Nafion (perfluorosulfonic acid, 5% solution in a mixture of low molecular weight aliphatic alcohols and water), ferrocene monocarboxylic acid (Cp2FeCOOH), and hydrogen hexachloroplatinate(IV) hydrate (chloroplatinic acid) were obtained from Aldrich Chemical Co. (Milwaukee, WI). Poly(2-hydroxyethyl (11) Pfeiffer, D.; Setz, K.; Klimes, N.; Makower, A.;. Schulmeister, T.; Scheller, F. W. Biosensors: Fundamentals, Technologies and Applications; GBF Monographs, 17; VCH: New York, 1992, pp 11-8. (12) Schneider, J.; Gru ¨ ndig, B.; Renneberg, R.; Cammann, K.; Madaras, M. B.; Buck, R. P.; Vorlop, K.-D. Anal. Chim. Acta 1996, 325, 161-7. (13) (a) Urban, G.; Jobst, G.; Aschauer, E.; Tilado, O.; Svasek, P.; Varahram, M. Sens. Actuators B 1994, 18-19, 592-6. (b) Moser, I.; Jobst, G.; Aschauer, E.; Svasek, P.; Varahram, M.; Urban, G.; Zanin, U. A.; Tjoutrina, G. Y.; Zharikova, A. V.; Berezov, T. T. Biosens. Bioelectron. 1995, 10, 527-32.
methacrylate) (p-HEMA) with MW ) 300 000 was obtained from Scientific Polymer Products, Inc. (Ontario, NY), and tetra-nbutylammonium perchlorate (TBAP) was from GFS Chemicals (Columbus, OH). All other chemicals were analytical reagent grade. Double distilled water (Barnstead Nanopure II, Boston, MA) was used for all the solutions. Solutions. The electrolyte used for platinization of the gold base electrodes was a solution containing chloroplatinic acid (3%), (NH4)2HPO4 (11.2%), and Na2HPO4 (23.6%) prepared according to Blum and Hogaboom.14 The amperometric (batch-type and FIA) measurements were performed in isotonic phosphate buffer saline (PBS, pH 7.4) containing 0.053 M Na2HPO4, 0.015 M NaH2PO4, 0.051 M NaCl, 1.5 mM MgNa2EDTA, 1 mM NaH2PO2 (hypophosphite) as a protective agent for CI and CA, and 0.2 mM NaN3 as a preservative. The electropolymerization of the inner layers was performed in phosphate buffer (PB) 0.1M, pH 7. The creatinine and creatine solutions were freshly prepared every 3 days and stored at 4 °C when not in use. Apparatus and Procedures. Electrochemical measurements were performed using an EG&G potentiostat/galvanostat 273A (Princeton, NJ) controlled by a PC-AT computer via a GPIB (IEEE488) National Instruments (Austin, TX) card or a PINE RDE4 (Grove City, PA) bipotentiostat for the differential measurements. For the batch-type amperometric measurements, a two-electrode configuration was used with an Ag/AgCl electrode as reference, a miniature reference electrode with flexible barrel (Cypress Systems, Inc., Lawrence, KS) for the single disk electrodes, or a three-electrode configuration was used that incorporated counter and pseudoreference electrodes on the biochips. The electropolymerization of the inner layer was performed in a conventional three-electrode electrochemical cell (Bioanalytical Systems, West Lafayette, IN), using an Ag/AgCl reference electrode and a Pt wire counter electrode. Platinization of the Au electrodes, preparation of the Ag/AgCl pseudoreference electrodes on the chips and the flow injection experiments were realized using an EG&G potentiostat/galvanostat (Model 363). For the batch-type experiments, a simple Plexiglas cell (5 mL volume) was used and the solutions were stirred at constant speed. The noise due to the magnetic stirring was removed using an analog 0.7 Hz Butterworth-type low-pass filter. After the potential was applied (+ 0.6 V vs Ag/AgCl), the background current was allowed to stabilize. The required volume of investigated solution was injected in the cell and the current-time response of the biosensor recorded on a Yokogawa 3025 X-Y Recorder (Tokyo, Japan). For the flow injection analysis measurements, a Rainin Rabbit (Woburn, MA) peristaltic pump and a Rheodyne 7125 (Cotati, CA) seven-port injection valve with a 100 µL injection loop were used. The flow rate of the carrier buffer was set at 1.9 mL/ min, and the residence time of the sample in the FIA system was 20 s. The biochips were used in a wall jet-type configuration, with the biochip structure orthogonal to the carrier buffer flow. This single-channel setup does not require a flow cell because all three electrodes are in contact with the solution flowing from the end of a glass capillary tube. All the measurements were performed at room temperature (∼25 °C). Preparation of Microfabricated Biosensors. The fabrication of the film structure for the base electrodes comprises several (14) Blum, W.; Hogaboom, G. B. Principles of Electroplating and Electroforming; McGraw-Hill: New York, 1949; pp 383-4.
Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
3833
Figure 1. Schematic drawing of the miniaturized, planar-type, biochip: WE, working electrode (d ) 1.5 mm); CE, counter electrode; RE, reference electrode. All the dimensions are in millimeters.
steps, described in more detail elsewhere.15 Before the Pt deposition, the surface of the working electrode was electrochemically cleaned using a current density of 0.5 mA/cm2 for 30 s, with the polarity of the electrodes reversed. Then, Pt was galvanostatically deposited using 12.7 mA/cm2 for 6.5 min. After Pt electrodeposition, the inner membrane was prepared as described below, then the chips were individually cut from the larger wafer, and the electrical connections were made. The pseudoreference Ag/AgCl electrode on each biochip was prepared after the individual chips were cut from the wafer. First, the Ag deposition was performed from a 1% KAg(CN)2 aqueous solution, using a current density of 5 mA/cm2 for 10 min. Then, a portion of the electrodeposited Ag was converted to AgCl by inserting each chip in a 0.1 M HCl solution and anodizing using 5 mA/cm2 for 2.5 min. The schematic diagram of the new design biochip with all three electrodes incorporated on the chip is shown in Figure 1. These biochips were used for the serum measurements, while the majority of the work concerning the optimization of the individual layers was done using single-disk electrodes (dWE ) 1 mm). To provide the desired chemical sensitivity and selectivity, a structure consisting of a three-layer system, subsequently described, was built on the working electrode: Inner Membrane (IM). An inner permselective layer on the Pt electrode was used to diminish simultaneous oxidation of interferences such as ascorbate, urate, acetaminophen, and other electroactive molecules likely to be present in the serum. The inner layer was formed by electropolymerization from 0.1 M PB (pH 7) solutions containing 1,3-DAB and resorcinol to form the poly(1,3-DAB/resorcinol) copolymer film or 1,3-DAB alone for the poly(1,3-DAB) films (see Table 2). The electropolymerization was done by cyclic voltammetry between +0.2 and +0.8 V vs Ag/ AgCl, using different scan rates and electropolymerization times or by chronoamperometry at + 0.7 V vs Ag/AgCl. The oxidation of the monomer at the surface of the electrode produces an anodic peak (at + 0.5 V vs Ag/AgCl) that gradually decreases in amplitude in subsequent cycles, indicating that a nonconducting film forms at the electrode surface. Preliminary trials have used (15) Cosofret, V. V.; Erdosy, M.; Johnson, T. A.; Buck, R. P.; Ash, R. B.; Neuman, M.R. Anal. Chem. 1995, 67, 1647-53.
3834 Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
deoxygenation of the monomer(s) solution before and during electropolymerization using N2, but this approach produced gas bubbles on the surface of the electrode which could alter the uniformity of the electrodeposited layer. Therefore, deaeration was not used in our subsequent studies. Enzyme Layer (EL). The enzymatic layer was obtained by cross-linking the enzymes with GA in the presence of BSA. For a typical preparation, 3.8 mg of CI (12 units/mg), 1.2 mg of SO (45 units/mg), and 0.6 mg of BSA were dissolved in 60 µL of PBS (pH 7.4) in a small conical vial and mixed together. After 5 min, the mixture was separated in two equal portions, in one of the portions 0.29 mg of CA (79 units/mg) was added to prepare the enzyme layer for the creatinine sensors; the other half was used for the creatine sensors. The appropriate amounts (see Results and Discussion) of 1% glutaraldehyde aqueous solution (freshly prepared) were added to each vial to start the crosslinking process. Two microliters of the resulting mixtures was quickly deposited on the working electrodes of the creatinine or creatine sensors, respectively, using a microsyringe. The EL obtained was allowed to cross-link in air, at room temperature, for 1 h. The amount of immobilized enzymes was estimated from the activity of the lyophilized enzymes, as specified by the manufacturer. No measurement of the residual activity after the immobilization was performed. Outer Membrane (OM). To complete the preparation of the biosensors, an outer membrane was deposited above the enzymatic layer. Stock solutions of p-HEMA (5 and 10% by weight) were prepared by dissolving the p-HEMA beads in methanol (99.9%). The deposition solutions were prepared by mixing equal volumes of the 5% p-HEMA and Nafion solutions to yield a mixture with 2.5% p-HEMA/2.5% Nafion or by diluting the p-HEMA and Nafion stock solutions with methanol. The outer membrane was fabricated by placing 1-3 µL (depending on the area of the WE) of the chosen mixture on top of the enzymatic layer with a microsyringe. The solvent evaporation at room temperature produced a transparent and compact outer layer. After solvent evaporation and when not in use, these biosensors were stored at 4 °C in PBS. Serum Measurements. The serum measurements were performed in a differential setup; the recorded signal was the difference between the creatinine sensor current and the creatine sensor current, both sensors simultaneously present in the sample. First, the control serum was reconstituted from the lyophilized preparation according to the manufacturer recommendations. A 2 mL aliquot of PBS was placed in the cell containing the creatine and creatinine planar biosensors, and after the stabilization of the background current, 1 mL of the serum sample was added. After the current again reached a steady-state value, a standard addition procedure16 (using successive injections of a 10 mM standard creatinine solution) was used to evaluate the creatinine concentration in the serum. For the human serum samples (kept frozen until ready to use), an identical procedure was used. RESULTS AND DISCUSSION Optimization of the Permselective Layer. We studied the possibility of obtaining permselective layers on the Pt microfabricated sensors using electropolymerization, inspired by the work (16) Harris, D. C. Quantitative Chemical Analysis, 4th ed.; W. H. Freeman & Co.: New York, 1995; pp 140-2.
Table 1. Interference Rejection Comparison for a Single-Disk Creatinine Sensor, Having a Poly(1,3-DAB/resorcinol) Inner Membrane
substancea AA APAP H2O2 creatinine AA/creatinine (%) APAP/creatinine (%) AA/H2O2 (%) APAP/H2O2 (%)
conc (mM) 0.16 0.16 0.0145 0.16
bare electrode ibb (nA) day 0
IM coated icb (nA) day 0
% rejectionc on day 0 (no EL-OM)
67.0 114.0 10.5
1.0 0.1 9.3
98.5 99.91 11.4
IM-EL-OM i (nA) day 1 day 6 0.7 0.1 14.4 4.9 0.7
638.1 1085.7
10.7 1.1
0.6 1.3 5.6 14.6 4.1 8.9 10.7 23.2
a AA ) ascorbic acid; APAP ) acetaminophen. Electropolymerization conditions: 1.6 mM 1,3-DAB + 1.6 mM resorcinol, scan rate 2 mV/s, 12 cycles. b Without stirring. c % rejection, (ib - ic) × 100/ib.
of Yacynych and co-workers on classical macroelectrodes17 a,18,19 and microelectrodes.17b The procedure used for testing the permselectivity of different electropolymerized films consisted in measuring the steady-state response (current) of the sensor to 0.16 mM AA and 0.16 mM APAP. These values were selected to be at the upper limit of their physiological levels.20 For comparison, the sensor response to 0.0145 mM H2O2 and/or 0.16 mM analyte (for the complete biosensors) was recorded. The H2O2 level was selected to provide a signal of a magnitude comparable with the expected signal from a low to normal physiological level of the analyte, therefore mimicking the most unfavorable conditions for the operation of the biosensor. For the purpose of estimating the efficiency of different electropolymerized films, a combined signal from AA + APAP less than 5% of the analyte signal was considered a desirable level of rejection, between 5 and 10% acceptable, and above 10% unacceptable; sensors showing poor rejection were discarded. Yacynych used a copolymer of 1,3-DAB and resorcinol as the preferred film for blocking interferences for long periods of time (months) from the surface of carbon or partially platinized carbon electrodes in a FIA setup.17a Other researchers21 concluded that the same film formed on Pt surface is not satisfactory in rejecting acetaminophen 2-3 days after preparation. We wanted to explore further this possibility, and single disk planar sensors having an electropolymerized poly(1,3-DAB/resorcinol) inner layer were studied for this purpose. The results from a typical sensor are presented in Table 1. It is obvious that while the IM-coated electrode has a very good rejection for both AA and APAP (on day 0), the combined effect of the two interferences amounts to 5.6 (on day 1) and 13.0% (on day 6) of the analyte signal for the complete structure. In a recent review,22 Emr and Yacynych pointed out that the electropolymerized films can have significantly different charac(17) (a) Geise, R. J.; Adams, J. M.; Barone, N. J.; Yacynych, A. M. Biosen. Bioelectron. 1991, 6, 151-60. (b) Reynolds, E. R.; Geise, R. J.; Yacynych, A. M. In Biosensors & Chemical Sensors: optimizing performance through polymeric materials; Edelman, P. G., Wang, J., Eds.; ACS Symposium Series 487; American Chemical Society: Washington, DC, 1992; pp 186-200. (18) Geise, R. J.; Rao, S. Y.; Yacynych, A. M. Anal. Chim. Acta 1993, 281, 46773. (19) Manowitz, P.; Stoecker, P. W.; Yacynych, A. M. Biosens. Bioelectron. 1995, 10, 359-70. (20) (a) Cso¨regi, E.; Schmidtke, D. W.; Heller, A. Anal. Chem. 1995, 67, 12404. (b) Abdel-Hamid, I.; Atanasov, P.; Wilkins, E. Anal. Chim. Acta 1995, 313, 45-54. (21) Zhang, Y.; Hu, Y.; Wilson, G. S.; Moatti-Sirat, D.; Poitout, V.; Reach, G. Anal. Chem. 1994, 66, 1183-8. (22) Emr, S. A.; Yacynych, A. M. Electroanalysis 1995, 7, 913-23.
Table 2. Optimization of the Electropolymerization Conditions: Comparison of 1,3-DAB vs 1,3-DAB + Resorcinol, Time of Electropolymerization, and Electrochemical Technique iAA/iH2O2 (%) iAPAP/iH2O2 (%) conditions for sensora electropolymb day 1 day 2 day 3 day 1 day 2 day 3 1 2 3 4 5 6 7d 8e
CA, 5 min CA, 10 min CA, 20 min CV (5; 6)c CV (2; 6) CV (2; 60) CV (2; 6) CV (2; 54)
13.3 10.8 9.9 6.8 5.7 4.5 3.0 3.9
7.9 6.2 6.6 5.3 3.7 3.8 3.5 5.0
7.8 6.0 7.1 3.7 3.2 3.8 3.5 3.8
106.8 8.7 5.9 4.3 1.2 0.0 2.0 0.5
104.0 31.7 21.3 18.1 3.8 0.5 10.4 4.0
114.6 58.8 82.4 32.5 8.2 0.7 12.2 10.6
a Deposition solutions were 10 mM 1,3-DAB in 100 mM PB (pH 7), except as noted. b CA, chronoamperometry; CV, cyclic voltammetry. c First number in the parentheses represents the scan rate (in mV/s); the second is the number of cycles. d Deposition solution, 5 mM 1,3DAB + 5 mM resorcinol. e Deposition solution, 1.5 mM 1,3-DAB + 1.6 mM resorcinol.
teristics when formed on different electrode materials and in different electropolymerization conditions. Our results obtained on single-disk Pt electrodes (summarized in Table 2) confirmed that the permselectivity behavior of the films is greatly influenced by the electropolymerization conditions. The copolymer films from 1,3-DAB and resorcinol performed less well than the poly(1,3-DAB) alone obtained under similar conditions (compare sensors 5 and 7) especially in terms of acetaminophen rejection. Cyclic voltammetry was preferred over chronoamperometry for the formation of the film. A slower scan rate favors formation of a less permeable film (compare sensors 4 and 5). All the films became gradually more permeable for acetaminophen, whereas the normalized response to ascorbic acid was relatively constant (∼3-5% for the best rejecting films). This behavior and the different dynamic response for AA (gradual increase of the signal) and APAP (almost instantaneously reaching a steady-state value) indicated a different transport mechanism for the two main interferences in the electropolymerized films. Clearly, the best permselectivity was obtained by using a long time (10 h, sensor 6) cyclic voltammetry deposition. However, the miniaturized sensors described here are meant to be disposable and batch fabricated; therefore, a compromise between the preparation time of the inner layer and its permselective properties should be found. For our purpose, an electropolymerization time of 2-3 h (12-18 cycles) was found to provide acceptable membranes to be used in the miniaturized disposable sensors. Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
3835
Figure 2. Cyclic voltammetry of ferrocenecarboxylic acid at two single-disk electrodes: (a) poly(1,3-DAB) on the Pt electrode; (b) bare Pt electrode. Conditions: 1 mM Cp2FeCOOH in 0.1 M TBAP/MeCN, scan rate 100 mV/s.
The interference rejection properties of the poly(1,3-DAB) film are presented in another form in Figure 2, which shows the cyclic voltammograms of 1 mM ferrocenecarboxylic acid in 0.1 M TBAP/MeCN at two planar Pt disk electrodes. While the redox behavior of ferrocenecarboxylic acid at the bare electrode is present (E°′ ) 0.67 V vs Ag/AgCl, ∆Ep ) 80 mV), there is no faradaic current flowing at the covered electrode, clearly showing that the ferrocenecarboxylic acid cannot permeate through the film to the electrode surface. Very important is the fact that while the film is efficient in rejecting relatively small molecules such as ascorbate, acetaminophen, and urate, it is almost completely transparent to H2O2 (see Table 1). We should mention that in some cases the IM was subjected to high temperature (100 °C for 15 min) during the process of making the electrical connections to the electrodes on the chip. This treatment did not influence the permselectivity properties of the poly(1,3-DAB) films, as observed for different sensor batches. Another important observation for the operation of the biosensors employing polymer films is that the selectivity of the inner layer is dependent on the extent of exposure to working conditions. Biosensors stored in PBS at 4 °C for up to 2 weeks without testing, entirely preserved their permselective properties, whereas similar sensors tested under operational conditions became gradually permeable to interferences, especially acetaminophen. This is a key observation, because it indicates the possibility of storage of the sensors in wet conditions for a period of time before usage. The lifetime of the biosensors depends to a great extent, however, on the operational stability of the enzyme layer. As shown in Figure 3, the sensitivity of a sensor having the EL on top of the IM decreased slightly upon storage in PBS at room temperature for 24 h, but significantly when the sensor was stored dry at room temperature for a few days. Similarly prepared sensors stored all the time at 4 °C in PBS when not in use did not show any significant change in sensitivity in the same time frame. From this experiment it was concluded that storage of the sensors in PBS at 4 °C is required in order to prolong their useful operational lifetime. The use of a protective OM is desirable, as discussed below. Enzyme Layer Considerations. The fabrication of the sensitive layer of the biosensors was performed by cross-linking 3836 Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
Figure 3. Dependence of a creatine sensor (IM-EL) response on the conditions of storage: 9, day 1; b, day 2; 2, day 8; 1, day 13. Storage conditions: days 0-7 PBS, 4 °C; days 7-8 PBS, room temperature; days 8-13 dry, room temperature.
Figure 4. Variability of the sensitivity of creatine sensors with the cross-linking factor f. For all sensors [GLT] ) 0.27%, (CI + SO)/BSA ) 2. Measurements at 25 °C in PBS (pH 7.4).
the enzymes with a bifunctional reagent (glutaraldehyde) in the presence of an inert protein, BSA. Details about the optimization of the EL with respect to the absolute activities and the activity ratios of the active enzymes were previously provided.7 During the preparation of the EL mixtures for different batches of biosensors, we recognized that another important parameter should be considered: the amount of cross-linking reagent in the layer. We define here this parameter, the cross-linking factor, as the ratio of the amounts of glutaraldehyde (GLT) and the total protein TP (active enzymes + BSA) in the mixture, expressed as weight percentage, f ) % GLT/% TP. Several creatine sensors were prepared with different amounts of enzymes in the layer while the ratio of activities were kept constant, ACI/ASO ) 1 as previously determined. The amount of GLT used in the deposition solution for each sensor was also constant. The variability of the sensitivity of these sensors with the cross-linking factor is presented in Figure 4. By increasing f, the sensitivity is lowered because the EL becomes less permeable. The diffusion of the analyte and the products of enzymatic reactions in the EL is restricted23 and/or the excessive cross-linking reduces the number
Table 3. Study of the Dependence of the Creatine Sensor Characteristics on the Outer Membrane Material and Thickness linear rangea (mM creatine) sensor
casting solution composn
V (µL)
day 1
day 2
1 2 3 4 5 6 7 8
no OM 5% Nafion 0.5% Nafion 10% p-HEMA 5% p-HEMA 1% p-HEMA 2.5% p-HEMA + 2.5% Nafion 2.5% p-HEMA + 2.5% Nafion
3 3 3 3 1 2 × 0.5 1
0.2 0.45 0.3 1.5b 1.2b 0.6 1.2 0.5
0.2 0.4 0.3
a
0.5 1.0 0.4
response time (t95) (s)
day 4
day 1
day 2
0.5 0.9 0.4
10 60 60 100 100 30 80 30
10 60 90 120 120 30 60 30
sensitivity (µA mM-1 cm-2)
day 4
day 1
day 2
day 4
30 60 30
64 6.2 44 6.0 16.4 46.5 16.2 22.9
58 5.8 28 5.5 15.4 52 17.3 22.9
55 15.6 23.6
R g 0.999. b At least (not studied for higher concentrations).
of active redox centers of the immobilized enzyme(s). In this particular experiment, smaller values for f were not tested because the mixture became too viscous (the values for TP were 5.7, 10.3, and 19.1%, respectively, with decreasing f), but the data suggest that the sensitivity could be further improved by lowering f. Keeping the cross-linking factor constant (f ) 0.0085), another study was performed to note the influence of the total protein concentration in the EL on the sensitivity of the biosensors. We have considered using less protein to obtain thinner layers. By decreasing the TP concentration the biosensor sensitivities decreased ∼40% in going from 8.4 to 4.5% TP. Unfortunately, one set of sensors with even lower TP could not be measured. Too little GLT was used for their preparation, and the actual f value was only 0.0043. The cross-linking was not sufficient for entrapment and the enzymes were washed out from the sensor surface! This example illustrates the importance of properly selecting the cross-linking factor. From these two studies, we established that values of TP of ∼9-10% and f in the range 0.008-0.014 will provide optimal sensitivity for our creatine and creatinine sensors. The EL showed good adhesion to the polyimide-IM structure, because no peel-off of the EL was observed upon wet storage (even in the absence of the OM). This adhesion could in part be explained by the formation of covalent bonds between glutaraldehyde and the residual (unreacted) amino groups at the surface of the poly(1,3-DAB) layer. This explanation is supported by the recent results of Si and co-workers,24 who indicated that the electrochemical copolymerization of aniline and 1,3-DAB produces a functionalized copolymer with NH2 groups that can be applied for the covalent immobilization of biological materials. Outer Membrane. The outer membrane has several important functions: (1) to protect the EL from fouling with proteins and other large molecules from blood or serum, (2) to provide a diffusion barrier for the analyte, therefore increasing the linearity of the device beyond the apparent K′M of the immobilized enzyme system, and (3) to render the sensor less sensitive to changes in the pH and temperature of the sample. Ideally, the OM should be thin and have controlled porosity to create reproducible diffusion paths for the analyte and therefore increased sensor-tosensor reproducibility. An example of such membranes with a regular pore structure are the polycarbonate track-etched membranes Nuclepore (Costar Co., Cambridge, MA). This type of preformed membranes is difficult to confine at the surface of a (23) Mutlu, M.; Mutlu, S. Biotechnol. Tech. 1995, 9, 277-82. (24) Si, S. H.; Xu, Y. J.; Nie, L. H.; Yao, S. Z. Electrochim. Acta 1995, 40, 271521.
planar structure, and a different approach was used for our biochips. Various solvent-dissolved polymeric materials were dispensed with a microsyringe on top of the EL, and the resulting biosensors were tested for linearity, sensitivity, and response time. The goal was to obtain an OM with good adhesion to the polyimide substrate and ability to provide a linear range up to 1 mM analyte (creatinine or creatine) with a fast response time (t95 e 1 min). Results obtained for single-disk creatine sensors using different OM compositions are presented in Table 3. Sensors 1-5 were fabricated using the same EL mixture and had a poly(1,3-DAB) IM, while sensors 6-8 were fabricated using a similar EL composition and had no IM. The characteristics observed for all the sensors should be almost entirely dependent on the OM, since the IM does not significantly influence the sensitivity or the response time of the biosensor. The sensors having a Nafion OM had short linear response range. It is possible that the reduced linear range and sensitivity (sensor 2) are a result of deactivation of some of the enzyme material by the Nafion. This assumption is supported by the fact that similar sensors having a p-HEMA OM had larger linear range but also higher sensitivity (compare sensors 2 and 5). The membranes cast from 5 or 10% p-HEMA are too thick, and the response times of the respective sensors (4 and 5) are increased. Using more dilute p-HEMA solution (1%) and smaller volumes for casting, the performance of the sensors in terms of response time and sensitivity is improved, but the linearity is degraded, as expected (sensor 6). It is well-known that since its introduction as a permselective film,25 Nafion was extensively used26-28 in the discrimination against interferences. Our idea was to cast the OM from a mixture of p-HEMA and Nafion, in order to exploit the diffusion-limiting capability of p-HEMA and to minimize the damaging effect of the Nafion. When different polymers are used for solvent-casting of composite membranes, they should be soluble in the same solvent in order to obtain homogeneous films. Fortunately, this is the case for methanol-soluble p-HEMA and Nafion. (25) Gerhardt, G. A.; Oke, A. F.; Nagy, G.; Moghaddam, B.; Adams, R. N. Brain Res. 1984, 290, 390-5. (26) Kristensen, E. W.; Kuhr, W. G.; Wightman, R. M. Anal. Chem. 1987, 59, 1752-7. (27) Harrison, D. J.; Turner, R. F. B.; Baltes, H. P. Anal. Chem. 1988, 60, 20027. (28) Navera, E. N.; Suzuki, M.; Tamiya, E.; Takeuchi, T.; Karube, I. Electroanalysis 1993, 5, 17-22.
Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
3837
Figure 6. Response of a creatine biochip tested in a FIA experiment: Conditions: carrier buffer PBS pH 7.4; flow rate 1.9 mL/min; 100 µL injection loop size. Figure 5. Behavior of a miniaturized creatine biochip in a batchtype experiment. (top) Effect of interferences (a ) 0.3 mM UA; b ) 0.16 mM APAP; c ) 0.16 mM AA; d ) 0.16 mM creatinine; e ) 0.16 mM creatine). (bottom) Response of the biochip to successive additions of a standard creatine solution (final concentrations 0.16, 0.32, 0.48, 0.63, 0.91, 1.18, and 1.43 mM creatine).
It is important that the linear range of the sensors can be controlled by the thickness of the OM. Comparing sensors 7 and 8, for the same type of OM, it can be seen that the linear range doubled when two aliquots of 0.5 µL were successively deposited on top of the EL (with drying in between) compared with a 1 µL single deposition. In this later case the solution spreads over a larger area; the resulting membrane is thinner and provides less diffusional restraint. The thickness and the structure of the outer layer can thus be somewhat controlled by the choice of the polymer matrix, concentration, and volume of the casting solution. On the basis of these observations we used the composite mixture (2.5% p-HEMA/2.5% Nafion) to produce the OM for our creatine and creatinine biosensors. The optimized membrane (sensor 7) has the required characteristics in terms of linearity and response time. At the same time, the sensitivity of these sensors is 10 times higher than for those fabricated using cellulose acetate (IM) and polyurethane (OM) in the past.7 Analytical Characteristics. The permselectivity characteristics of an optimized creatine biochip are presented in Figure 5. There was little response to the main interferences found in serum, and the creatine sensor did not respond to creatinine. The response time is relatively fast (t95 < 60 s), and the linear range extends up to 0.9-1.2 mM in a batch-type experiment, with a detection limit of 20 µM creatine (S/N ) 3). The fabricated sensors were also tested in a flow injection setup and their behavior is shown in Figure 6. The linear range of the response is further extended (up to at least 2 mM), since by virtue of dilution of the injected sample, the EL functions below the saturation threshold at even higher analyte concentrations. The creatine biochip had a detection limit of 10 µM creatine (S/N ) 3) and good reproducibility (coefficient of variation of 1.6% for six repetitive injections at the 0.2 mM level). The sample throughput was 30-35 samples/h. Similar characteristics were obtained with the creatinine biochips. The reproducibility of equally made electrodes, estimated as the coefficient of variation for the linear range slope, was 10-20%. The reproducibility was limited by the OM because thickness control is difficult to achieve by casting. Although the long-term operational stability 3838 Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
Table 4. Creatinine Assay Using the Miniaturized Biochips in Control Human Seruma run
A
B
corr coeff R (n ) 4)
[creatinine]sample (µM)
[creatinine]serum (µM)
1 2 3 4 5
21.61 25.73 25.92 27.75 26.62
641.9 633.1 653.6 679.8 691.8
0.999 96 0.999 90 0.999 92 0.999 89 0.999 89
33.7 40.6 39.6 40.8 38.5
101.1 121.8 118.8 122.4 115.5
a Linear regression for the standard addition: Y ) A + BX. Measured: [creatinine]serum ) [creatinine]av ( CL ) 116 ( 11 (µM) ) 1.31 ( 0.12 (mg/dL); assigned: [creatinine]serum ) 1.26 ( 0.05 (mg/ dL). Confidence limit CL ) s(t/n0.5), where s is the standard deviation, t is the Student variable for (n - 1, 95%), and n is the number of measurements (runs).
was not an issue for these disposable sensors, they have shown practically no change in response over a period of 30 days with periodic testing and storage in PBS at 4 °C. Serum Assay of Creatinine. A human control serum (Dade Moni-Trol Level I) was used and the results are presented in Table 4. The correlation between the measured creatinine level and the assigned value was very good. The measured value was only 4% higher than the assigned one. The assigned value was calculated as an average of 23 different clinical analyzers mean values provided by the manufacturer. The differential measurement assumes that the sensitivity of both sensors to creatine and interferences is very similar, so that by subtracting the current output values of the two sensors, the resulting signal will be proportional only to the creatinine level in the sample. The novelty of our approach is that the interferences problem is addressed twice: the inner membranes are physically blocking the interferences, and the currents due to the interferences are similarly small (if not zero) and will cancel out by subtraction. The creatinine levels in several human serum samples obtained from hospitalized patients were assayed. The correlation between the results provided by the biochips and the values obtained with the classical spectrophotometric method (Jaffe´), using a Hitachi 717 analyzer, gave a linear regression Y ) 0.07 + 1.04X (R ) 0.957, n ) 5) for samples in the normal physiological range. Unfortunately, due to the limited volume of serum available in each sample, statistical measurements of reproducibility could not be performed. Samples with high levels of creatinine (and presumably creatine) gave readings 5-10% lower than the assigned values. These lower than expected values were obtained due to the observed higher sensitivity to creatine of the creatine
sensor in comparison with the creatinine sensor. The subtracted current is therefore smaller than in the ideal “matched” pair, and as a result, the creatinine reading is lower, especially for samples with high creatine content. This fact points out one of the weaknesses of this approach: the necessity of having creatine and creatinine sensors with close sensitivities for creatine. New methods for reproducible deposition of the EL and OM or correction using appropriate software in the case of commercial applications can solve this problem. The method and the procedures described here can be extended to other pairs of analytes such as glutamine/glutamate, acetylcholine/choline, and sucrose/glucose, which require the use of multienzyme systems. The results obtained are encouraging for the perspective of using miniaturized disposable biosensors in diagnostic, portable instruments for monitoring of important metabolites in biological fluids, with little or no dilution of the sample. The electropolymerized films provide good rejection of unwanted endogenous and exogenous interferences, and the fabricated biosensors have higher sensitivity and shorter response
time than the previously used structures.7 The linear range of these devices is suitable for measurements in serum, both in normal- and high-level creatinine samples, and they could provide a quick screening test for renal function evaluation. ACKNOWLEDGMENT This research was supported by NSF Engineering Research Center Grant CDR-8622201. We thank Stefan Ufer and Bruce R. Ash (Biomedical Microsensors Laboratory, North Carolina State University) for the microfabrication of the base electrodes and Dr. John Hammond (Clinical Laboratory, University of North Carolina Hospitals) for providing the serum samples. We also acknowledge useful discussions with Sayed Marzouk and Dr. Ionel C. Popescu. Received for review March 11, 1996. Accepted August 7, 1996.X AC960239R X
Abstract published in Advance ACS Abstracts, September 15, 1996.
Analytical Chemistry, Vol. 68, No. 21, November 1, 1996
3839
