Development of a long-life capillary enzyme bioreactor for the determination of blood glucose

Talanta 71 (2007) 391–396

Development of a long-life capillary enzyme bioreactor for the determination of blood glucose Ja-an Annie Ho a,b,∗ , Li-chen Wu b , Nien-Chu Fan a , Ming-Shih Lee c,d , Hung-Yi Kuo b , Chung-Shi Yang b a Department of Chemistry, National Tsing Hua University, Hsinchu 300, Taiwan Department of Applied Chemistry, National Chi-Nan University, Puli, Nantou 545, Taiwan c Department of Medical Laboratory, Taichung Veterans General Hospital, Taichung 407, Taiwan d Department of Medical Technology, Chung Shan Medical University, Taichung 402, Taiwan b

Received 11 January 2006; received in revised form 11 April 2006; accepted 11 April 2006 Available online 5 June 2006

Abstract A long-life capillary enzyme bioreactor was developed that determines glucose concentrations with high sensitivity and better stability than previous systems. The bioreactor was constructed by immobilizing glucose oxidase (GOx) onto the inner surface of a 0.53 mm i.d. fused-silica capillary that was part of a continuous-flow system. In the presence of oxygen, GOx converts glucose to gluconic acid and hydrogen peroxide (H2 O2 ). Hydrogen peroxide detection was accomplished using an amperometric electrochemical detector. The integration of this capillary reactor into a flow-injection (FIA) system offered a larger surface-to-volume ratio, reduced band-broadening effects, and reduced reagent consumption compared to packed column in FIA or other settings. To obtain operational (at ambient temp) and storage (at 4 ◦ C) stability for 20 weeks, the glucose biosensing system was prepared using an optimal GOx concentration (200 mg/mL). This exhibited an FIA peak response of 7 min and a detection limit of 10 ␮M (S/N = 3) with excellent reproducibility (coefficient of variation, CV < 0.75%). It also had a linear working range from 101 to 104 ␮M. The enzyme activity in this proposed capillary enzyme reactor was well maintained for 20 weeks. Furthermore, 20 serum samples were analyzed using this system, and these correlated favorably (correlation coefficient, r2 = 0.935) with results for the same samples obtained using a routine clinical method. The resulting biosensing system exhibited characteristics that make it suitable for in vivo application. © 2006 Elsevier B.V. All rights reserved. Keywords: Continuous-flow system; Flow-injection analysis; FIA; Glucose; Blood glucose; Glucose oxidase

1. Introduction Human body needs to maintain blood glucose within a very narrow range of 70–110 mg/dL. People who have diabetes or increased fasting levels of glucose have elevated blood glucose levels because of an inability to use insulin properly. This is often referred to as insulin resistance. Statistics show that diabetes has reached epidemic levels in the U.S. because of increased incidence among older Americans, as well as more obesity in the population. About 2200 people are diagnosed with diabetes each day, but about one-third of the individuals who have diabetes are not aware of it until one of its life-threatening complications has developed. Diabetes results in long-term health consequences,

∗

Corresponding author. Tel.: +886 3571 5131x31286; fax: +886 3 571 1082. E-mail address: [email protected] (J.-a.A. Ho).

0039-9140/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2006.04.023

including cardiovascular disease, nephropathy, neuropathy, diabetic retinopathy and blindness. Recent research has indicated that hyperglycemia is common in critically ill patients, even in those without diabetes mellitus. It has been reported that aggressive glycemic control may reduce mortality in this population [1]. However, the relationship among mortality, the control of hyperglycemia, and the administration of exogenous insulin is still unclear. Therefore, it is very important to have a simpler, more-stable, and more-sensitive method that allows the monitoring of blood glucose in clinics and laboratories. The glucose sensor reported by Clark and Lyons in 1962 [2] has generally been recognized as the first biosensor. Since then, many types of sensors have been developed for medical diagnosis applications. The use of glucose oxidase (GOx)-based electrodes is a well-established method of detection for in vivo levels of circulating glucose [3–5]. In this approach, glucose is converted to gluconic acid and easily detectable hydrogen perox-

392

J.-a.A. Ho et al. / Talanta 71 (2007) 391–396

ide (H2 O2 ) by the enzyme glucose oxidase. The process requires oxygen as a cosubstrate. The produced H2 O2 is then measured using a charged platinum electrode surface. Hydrogen peroxide has become by far the most widely used method of signal transduction in enzyme biosensors, and a majority of all biosensors (65%) use hydrogen peroxide detection [6]. According to Wilson and Thevenot [7], the construction of a hydrogen peroxide sensor usually involves a platinum anode and a silver/silver chloride cathode. When the anode is poised at +0.6–0.7 V [8], the plateau of oxidation of peroxide is reached at the anode. The enzyme employed in the construction of hydrogen peroxideproducing biosensors frequently involves the immobilization of oxido-reductases to the surface of the sensor by glutaraldehyde cross-linking [8]. Other methods have been reported for the immobilization of enzyme, such as physical deposition onto solid supports, covalent binding [9], and entrapment within a polyer matrix [10]. In recent years, sol–gel technology has been widely used to entrap enzyme for different uses [11–14], because it retains better enzyme activity compared to the free enzyme. Matrixes are usually prepared under ambient conditions and exhibit tunable porosity, high thermal stability, and chemical inertness [12]. However, the silica sol–gel matrixes have some drawbacks, including fragility, complicated preparation procedures, and a tendency to be hydrolyzed at high acidity, which often results in the loss of enzyme stability and also limits their application and feasibility in the development of electrochemical sensors [15,16]. Many methods have been developed in an effort to find a noninvasive detection system for circulating glucose at in vivo levels, including ultrasound-assisted transdermal monitoring, electromagnetic-based sensor, and fluorescence-affinity hollowfiber sensors [17–27]. Other methods of glucose determination that have been reported include those based on a genetically engineered protein [28], on concanavalin A [29], and on a microcantilever [30]. In this work we develop a longer-life capillary enzyme bioreactor for the determination of glucose. The greatly improved activity and stability of this new enzyme bioreactor is facilitated by the direct attachment of GOx to the wall of a 530 ␮m i.d. fused-silica capillary. To the best of our knowledge this is the first demonstration of the capillary glucose oxidase bioreactor with improved enzyme stability and longer shelf life, which provides a higher surface-to-volume ratio, maximizing the interaction between glucose and GOx compared to bead-packed column. The results from blood sample analysis promised well for the use of this biosensing system in online blood glucose monitoring of critically ill patients before and after surgical operations. Reduced mortality can therefore be achieved by intensive glycemic control. 2. Materials and methods

sodium metaperiodate, sodium cyanborohydride and triethanolamine, glycine, and Trizma® Base tris[hydroxymethyl] aminomethane (Tris) were obtained from Sigma Chemicals Co. (St. Louis, MO). The fused-silica capillary (0.53 mm i.d.) was obtained from Alltech (Deerfield, IL). All other inorganic chemicals and organic solvents were of reagent grade or better and were purchased from Aldrich Chemical Co. (St. Louis, MO). The pre-analyzed blood plasma samples from patients were obtained fresh from the Veteran General Hospital—Taichung (VGHTC). The use of these samples in no way contradicts the Helsinki Declaration. De-ionized distilled water was obtained from a Milli-Q system (Milford, MA). 2.2. Methods 2.2.1. Capillary modification For a sensitive flow-injection analysis (FIA) enzyme reactor, we required a high-enzyme loading comparable to the dead volume of the bed. For such an enzyme assay, the immobilization support must be rigid and have a mild, very stable, covalent immobilization chemistry. Our group has previously demonstrated the successful immobilization of biomolecules such as antibody without loss of activity and decreased stability on the inner surface of capillary column precoated with a glycerylpropyl layer to minimize the adsorption of the analyte. In the current study the microcapillary enzyme reactor was modified based on previously described procedures [31–36]. Detailed modification procedures were as follows: Step 1: The 85 cm fused-silica capillary (0.53 mm i.d.) was treated with 1 M NaOH overnight. Step 2: 1 M HCl and distilled water were used to rinse the capillary, which was subsequently filled with 3glycidoxypropyltrimethoxysilane (GPTMS) and heated at 90 ◦ C for 2 h. Step 3: The capillary was rinsed and treated with 10 mM sulfuric acid at 90 ◦ C for 10 min to convert the residual epoxy groups to diols. Step 4: After washing with distilled water, diols were cleaved and oxidized to aldehydes with sodium metaperiodate containing potassium carbonate at room temperature for 2 h. Step 5: 190 ␮L of GOx (200 mg/mL) and sodium cyanborohydride (5 mg/mL) in 0.1 M phosphate buffer (pH 7.3) were passed slowly into the capillary and incubated overnight to reduce the Schiff base. Step 6: The capillary was rinsed with 0.2 M triethanolamine buffer (pH 8.2), 1 M NaCl, 0.1 M glycine/HCl buffer (pH 2.5), and Tris buffered saline (TBS), pH 7.0, sequentially. Finally, the capillary enzyme reactor, filled with TBS (pH 7.0), was then stored at 4 ◦ C until use. In this way glucose oxidase was covalently attached on the inner wall of capillary column.

2.1. Reagents and materials

2.3. Flow-injection analysis system

d-(+)-Glucose, glucose oxidase (glucose: oxygen oxidoreductase E.C. 1.1.3.4, from Aspergillus niger, 181.6 U/mg), 3-glycidoxy propyltrimethoxysilane, potassium carbonate,

The flow-injection analysis system (schematic diagram shown at Fig. 1) consists of a Hewlett Packard 1050 HPLC pump (Agilent, Foster City, CA) at the inlet of the capillary glu-

J.-a.A. Ho et al. / Talanta 71 (2007) 391–396

393

Fig. 1. A schematic diagram of enzymatic flow injection analytical system for blood glucose. WE: working electrode; RE: reference electrode; AE: counter electrode.

cose biosensing system that maintains a flow rate of 0.3 mL/min and a Rheodyne (Model 7725) injector with a 20 ␮L sample loop (Rainin, Emeryville, CA) for injection of the samples. Commercially available polyetheretherketone (PEEK) tubing (0.020 in. i.d.) and standard fingertight fittings were purchased from Upchurch Scientific Inc. (Oak Harbor, WA). The carrier used was 0.1 M potassium phosphate buffer (pH 7.2) containing 0.1 M NaCl that was vacuum-filtered before use. A HW-2000 Chromatography workstation, used for data collection, was purchased from Great Tide Instrument Co. (Taipei, Taiwan). Final signal integration was performed using the HW2000 Chromatography workstation system running on an Intel Celeron 2.20 GHz computer. All electrochemical measurements were performed on an electrochemical analyzer (model CL-4C amperometric detector) obtained from BAS (West Lafayette, IN). A BAS model CC-5E electrochemical flow cell was employed in these measurements. The conventional threeelectrode system was made up of a dual platinum electrode for thin-layer cross-flow cell (model MF-1012, BAS) as working electrode (3 mm in diameter), Ag|AgCl as reference (model MW-2078, BAS), and steel wire as counter electrode.

amperometric detector, and the current output was also stored on HW-2000. The height of a given FIA peak reflects the number of moles injected onto the capillary enzyme reactor. At a given analyte concentration, the peak height varies with sample volume, which is determined by the volume of the injection loop. The calibration curves for each assay were expressed in terms of the injected molar content in order to determine the linear dynamic range of the capillary FIA system for glucose. 2.5. Real-sample analysis Fresh human serum (20 ␮L) was diluted with 180 ␮L of 0.1 M potassium phosphate buffer (pH 7.2) containing 0.1 M NaCl and subsequently injected via the valve. The results obtained using this capillary-based method were compared to those obtained with a clinically used glucose analyzer (Hitachi 7170 automated analyzer). 3. Results and discussion 3.1. Optimization of parameters and characterization of the biosensing system

2.4. Experimental procedures For the determination of glucose, 20 ␮L of glucose standards were injected into the carrier stream and subjected to flowinjection analysis using the constructed microcapillary glucose oxidase enzyme reactor as described above, where the catalytic reaction involves glucose oxidation to produce hydrogen peroxide. The amperometric signal produced by hydrogen peroxide was measured by the amperometric detector in the system, which applied a potential to the electrochemical cell and monitored the resulting electrochemical reaction. The dual platinum working electrode was initially prepared by polishing for 3 min with 0.05 ␮m diamond polish on a polishing disk. After polishing, the dual electrode surface was rinsed with distilled water and maintained at 0.350 V versus Ag|AgCl for hydrogen peroxide measurements. The electrochemical oxidation of hydrogen peroxide at the dual 3 mm platinum electrode was measured with an

A series of experiments was performed to establish the conditions for maximum peak height. The applied voltage (0.300–0.500 V), sample injection volume (5–20 ␮L), enzyme concentration (100–200 mg/mL), and flow rate (0.1–0.5 mL/min) were investigated. To evaluate the effect of the voltage on the sensitivity of the biosensing system, different voltages were applied to the system. The voltages varied from 0.300 to 0.500 V. The maximum sensitivity was achieved at an applied voltage of 0.500 V versus Ag|AgCl using model glucose standards. However, this high voltage suffered from a low signal-to-noise ratio problem, and there was severe interference from the medium during serum sample analysis (Fig. 2). Oxidation of hydrogen peroxide at higher voltage (at a platinum electrode) is prone to interference from many other electroactive substances, such as ascorbic acid and uric acid; however, oxidation signals obtained

394

J.-a.A. Ho et al. / Talanta 71 (2007) 391–396

Fig. 2. The effect of applied voltage on real-sample analysis.

Fig. 3. Reproducibility of the signals generated by five replicates of a glucose standard (concentration = 10 mM).

tion for the glucose standards. at lower potential minimize these interference effects. Therefore, a voltage of 0.350 V was selected, and the system became more tolerant of interference. Although results obtained from Hitachi 7170 autoanalyzer and proposed FIA system correlated well, differences were still observed between these two sets of data. The accuracy of the proposed system should be satisfactory for screening purpose. The injected-sample volume was varied by changing sample loop size relative to the injection valve. The peak height increased with increasing injected-sample size. A sample volume of 20 ␮L was selected as a suitable volume. The enzyme concentration for the immobilization on the inner wall of capillary was varied from 100 to 200 mg/mL. The maximum peak height was obtained with GOx at 200 mg/mL. Due to the relatively high cost of glucose oxidase, enzyme concentration higher than 200 mg/mL was not considered in this study. The length of bioreactor (25–100 cm) was also an investigated parameters. The longer the bioreactor, the higher signal output could be obtained. However, broaden peaks were often found when length of bioreactor was longer than 85 cm. The flow rate was a very important parameter of the proposed system because the slower flow allowed sample glucose to react with immobilized enzyme more completely, and therefore higher signal output could be collected; however, the slow flow rate often resulted in peak broadening and limited sample throughput. After considering all of these factors, a flow rate of 0.3 mL/min was chosen for acceptable peak height and sample throughput.

H 2 O2

+0.350−0.650V vs Ag/AgCl

−→

O2 + 2H+ + 2e−

The amperometric signal was obtained upon the injection of variable glucose concentrations into the electrode cell under flow conditions, no signal was found in the absence of glucose. The linear response was observed between 10−2 mM (10 ␮M) and 10 mM of glucose standard solution with the regression equation of amperometric signal = 49.879 mM + 2.21 (R2 = 1.000). The sensing system was used for an average of 8 h a day, and the capillary enzyme reactor remained stable for at least 120 days at 25 ◦ C in operation condition and for the remainder of the time at 4 ◦ C in storage conditions. The repeatability and reproducibility of the proposed system was examined by injecting a 10 mM glucose standard. The uniformity of the FIA-amperometric peaks generated by five replicate injections is shown in Fig. 3. The largest value for the coefficient of variation (CV) for five replicate measurements was 0.75%, indicating that the reproducibility of this enzyme-based capillary glucose sensing system is acceptable. The proposed method was also validated for its accuracy by employing interday studies for 5 days where a glucose standard was measured, also at a concentration of 10 mM. As shown in Fig. 4, the coefficient of variation based on the peak height was found to be 3.05%. Additionally, though Pt electrode is prone to fouling by protein components in physiological fluids, no serious interference was found when our model serum solution (BSA spiked in glucose standard solution at final concentration of 7 mg/mL) was tested.

3.2. Assay performance The present enzyme-based capillary glucose biosensing system was based on glucose oxidase. Glucose + O2 ↔ H2 O2 + gluconicacid In the reaction sequence shown above, glucose oxidase catalyses the oxidation of ␤-d-glucose to gluconic acid, using oxygen as the electron acceptor. Since the gluconic acid level cannot be measured by the change in pH [8], the oxidation of hydrogen peroxide was measured by means of a charged platinum-based working electrode surface [3]. Based on earlier research [37–40], potassium phosphate buffered saline (pH 7.2) with a similar nature to physiological fluids was selected as the medium solu-

Fig. 4. Interday deviation for the enzyme-based capillary biosensing system. Glucose standard concentration: 10 mM.

J.-a.A. Ho et al. / Talanta 71 (2007) 391–396

Fig. 5. Linear regression of the correlation between glucose results measured by the proposed capillary glucose biosensing system and reference clinical measurements of blood samples obtained from 20 patients. X-axis: results of clinical measurements; Y-axis: results from the capillary glucose biosensing system. Each point represents the mean of three measurements.

3.3. Application The comparison study in measuring blood glucose of patients with proposed system and clinically used glucose analyzer (Hitachi 7170) is very important in validating a new analytical methodology towards its clinical applications. In this study, glucose was determined in human serum from adults by using the proposed system. Prior to analysis the sample serum was diluted with 0.1 M potassium phosphate buffer containing 0.1 M NaCl (pH 7.2) and centrifuged at 2000 × g at room temp for 5 min. Twenty microliters was taken directly from the supernatant and injected into the system. Linear regression of the correlation between blood glucose results measured by the enzyme-based capillary glucose biosensing system and a clinically used glucose analyzer (Hitachi 7170) is shown in Fig. 5. The largest value of the CV for five replicate injections was 6.05%. The best-fit regression line of the average of enzyme-based capillary glucose sensing system versus that determined using the clinical Hitachi 7170 automated analyzer indicated a strong correlation between the two data sets (r2 = 0.935). This shows that the proposed system is comparable to the Hitachi 7170 automated analyzer presently used in the Medical Laboratory Department at Taichung Veterans General Hospital. 4. Conclusions In summary, we have demonstrated the successful attachment of glucose oxidase to the inner wall of a fused-silica capillary while retaining enzymatic activity for more than 120 days. Our study has shown that an enzyme-based capillary glucose biosensing system was developed based on flow-injection analysis with amperometric detection. Operational and storage stability for greater than 4 months permitted the measurement of more than 300 samples. The glucose biosensing system prepared using the optimal GOx concentration (200 mg/mL) exhibited a FIAamperometric current response at 7 min. The sample throughput was about 9/h, and the reagent consumption was reduced. This

395

flow-injection type of sensing system holds promise for the determination of glucose content in clinical samples as well as in fruit juice. The proposed glucose sensing system was found to be responsive to glucose over a wide range of concentrations and has the following characteristics: a detection limit of 10 ␮M (based on the signal-to-noise characteristics, S/N = 3), linearity up to 104 ␮M, and reproducibility of under 0.75% coefficient of variation. These results clearly show the usefulness of this platform for direct detection of glucose and clinical diagnosis without complicated sample preparation and labeling. This proposed biosensing system has demonstrated its feasibility as a means of determining blood glucose in serum. The integration of this capillary system into a flow-injection system offers the advantages of a large surface-to-volume ratio, laminar flow, and reduced band-broadening effects compared to previous packed-column systems. These all help to increase the sensitivity and reproducibility of this capillary glucose measurement system. The current focus of our group is to use an enzyme-immobilization technique on microfluidic enzyme chips to improve the sample throughput. Future efforts will include attempts to incorporate an insulin biosensor into the present sensing system that will allow the parallel measurement of glucose and insulin. This will simplify studies on whether blood glucose level or the amount of insulin administered is associated with reduced mortality in critically ill patients. Acknowledgements This work was supported by the National Science Council in Taiwan, ROC, under grant NSC 91-2113-M-260-011 and the Taichung Veterans General Hospital-National Chi-Nan University Joint Research Program under grant VGHCN 92-72-02. References [1] S.J. Finney, C. Zekveld, A. Elia, T.W. Evans, JAMA 290 (2003) 2041. [2] L.C. Clark Jr., C. Lyons, Ann. N Y Acad. Sci. 102 (1962) 29. [3] G.G. Guilbault, G.J. Lubrano, Anal. Chim. Acta 64 (1973) 439. [4] Y. Zhang, G.S. Wilson, Anal. Chim. Acta 281 (1993) 513. [5] Y. Hu, G.S. Wilson, J. Neurochem. 68 (1997) 1745. [6] J. Davis, D.H. Vaughan, M.F. Cardosi, Enzyme Microbiol. Technol. 17 (1995) 1030. [7] G.S. Wilson, D.R. Thevenot, in: A.E.G. Cass (Ed.), Biosensor, A Practical Approach, IRL Press, Oxford, UK, 1990, pp. 1–17. [8] J. Woodward, in: A. Mulchandani, K.R. Rogers (Eds.), Enzyme and Microbial Biosensor, Techniques and Protocols, Humana Press, New Jersey, USA, 1998, pp. 67–79. [9] K.T. Lee, C.C. Akoh, Food Rev. Int. 14 (1998) 17. [10] D. Parra-Diaz, D.P. Brower, M.B. Medina, G.J. Piazza, Biotechnol. Appl. Biochem. 18 (1993) 359. [11] A. Hsu, T.A. Foglia, S. Shen, Biotechnol. Appl. Biochem. 31 (2000) 179. [12] J. Yu, H. Ju, Anal. Chem. 74 (2002) 3579. [13] Th. Noguer, D. Szydlowska, J.-L. Marty, M. Trojanowicz, Polish J. Chem. 78 (2004) 1679. [14] Y. Tang, E.C. Tehan, Z. Tao, F.V. Bright, Anal. Chem. 75 (2003) 2407. [15] Q. Chen, G.L. Kenausis, A. Heller, J. Am. Chem. Soc. 120 (1998) 4582. [16] O. Lev, M. Narvaez, E. Dominguez, I. Kataki, J. Electroanal. Chem. 425 (1997) 1.

396

J.-a.A. Ho et al. / Talanta 71 (2007) 391–396

[17] M. Gourzi, A. Rouane, R. Guelaz, M. Nadi, F. Jaspard, J. Med. Eng. Technol. 27 (2003) 276. [18] J. Kost, S. Mitragotri, R. Gabbay, M. Pishko, R. Langer, Nature 6 (2000) 347. [19] R. Ballerstadt, J.S. Schultz, Anal. Chem. 72 (2000) 4185. [20] M. Gerritsen, J.A. Jansen, A. Kros, R.J.M. Nolte, J.A. Lutterman, Invest. Surg. 11 (1998) 163. [21] R. Ballerstadt, J.S. Schultz, Anal. Chim. Acta 345 (1997) 203. [22] E. Wilkins, P. Atanasov, B.A. Muggenburg, Biosens. Bioelectron. 10 (1995) 485. [23] D.L. Meadows, J.S. Schultz, Anal. Chim. Acta 280 (1993) 21. [24] S. Mansouri, J.S. Schultz, Biotechnology 2 (1984) 885. [25] P. Abel, A. Muller, U. Fischer, Biomed. Biochim. Acta 3 (1984) 577. [26] J.C. Pickup, D.R. Thevenot, European achievements in sensor research dedicated to in vivo monitoring, in Advances in biosensors, supplement 1, JAI Press, London, UK, 1993, pp. 273–288.

[27] S. Jung, J.R. Trimarchi, R.H. Sanger, P.J.S. Smith, Anal. Chem. 73 (2001) 3759. [28] L. Tolosa, I. Gryczynski, L.R. Eichhorn, J.D. Dattelbaum, F.N. Castellano, G. Rao, J.R. Lakowicz, Anal. Biochem. 267 (1999) 114. [29] R.J. Russell, M.V. Pishko, Anal. Chem. 71 (1999) 3126. [30] J. Pei, F. Tian, T.B. Thundat, Anal. Chem. 76 (2004) 292. [31] P. Larsson, Methods Enzymol. 104 (1984) 212. [32] M. de Frutos, S.K. Paliwal, F.E. Regnier, Anal. Chem. 65 (1993) 2159. [33] M.Y. Lee, R.A. Durst, J. Agric. Food Chem. 44 (1996) 4032. [34] J.A. Ho, R.A. Durst, Anal. Chim. Acta 414 (2000) 61. [35] F.E. Regnier, R. Noel, J. Chromatogr. Sci. 14 (1976) 316. [36] L.C. Wu, C.M. Cheng, Anal. Biochem. 346 (2005) 234. [37] J. Wang, M. Chatrathl, A. Iba˜nez, Analyst 126 (2001) 1203. [38] M.G. Booutelle, L.K. Fellows, C. Cook, Anal. Chem. 64 (1992) 1790. [39] T. Hoshi, J. Anzai, T. Osa, Anal. Chem. 67 (1995) 770. [40] S. Hrapovic, J.H.T. Luong, Anal. Chem. 75 (2003) 3308.