ANALYST, MAY 1991, VOL. 116 459 Studies on Enzyme Electrodes With Ferrocene and Carbon Paste Bound With Cellulose Triacetate S. K. Beh, G. J. Moody and J. D. R. Thomas* School of Chemistry and Applied Chemistry, University of Wales College of Cardiff, P.O. Box 912, Cardiff CFI 3TB, UK A ferrocene-based chemically modified electrode has been prepared from a mixture of carbon paste and ferrocene, bound with cellulose triacetate. Glucose oxidase immobilized onto nylon net placed over the chemically modified indicator electrode completed the assembly of a robust ferrocene-mediated enzyme electrode. This was housed in a three-electrode Stelte micro-cell modified for flow injection according t o previous studies, and further modified by introducing a viscose acetate exclusion membrane between the outermost nylon-enzyme mesh and the ferrocene-carbon paste layer.Glucose was determined ampero- metrically by monitoring the product of hydrogen peroxide enzymolysis at +I60 mV versus a silver-silver chloride reference electrode. The enzyme electrodes showed a detection range of 0.01-70 mmol dm-3 glucose and the lifetime of the chemically modified electrode exceeded 24 months with intermittent use. Interference from ascorbic acid was minimal, while the maximum useful range was extended t o 100 mmol dm-3 glucose by simply covering the electrode surface with an exclusion membrane. A simplex optimization procedure was employed in evaluating electrodes without the use of an exclusion membrane. Keywords: Chemically modified glucose enzyme electrode; ferrocene; flow injection; simplex optimization The use of electron-transfer mediators has significantly improved the scope and performance of amperometric probes. These mediators are redox couple agents of low relative molecular mass, which shuttle electrons from the redox centre of the enzyme catalyst to the surface of the indicator electrode.During the catalytic cycle, the mediator, M,,, reacts with the reduced enzyme, and then undergoes rapid charge transfer at the electrode surface as illustrated for adenine flavin dinucleotide (FAD): Glucose + FAD + gluconic acid + FADH2 (1) FADH2 + M,, -+ FAD2+ + Mred + 2H+ Mred -+ M,, + ne- (2) (3) Provided M,, does not react with oxygen, it substitutes for oxygen in the classical enzymic reaction [eqns. (1)-(3)], and the rate at which the reduced mediator, Mred, is produced can be measured amperometrically at a suitable electrode.A practical mediator needs to be of low relative molecular mass, easily adsorbed onto an electrode surface, reversible, fast reacting, regenerated at low potential, pH independent, stable in both oxidized and reduced forms, unreactive with oxygen and non-toxic. Among the most successful mediators are those based on ferrocene [bis(n-cyclopentadienyl)iron] and its derivatives,' all of which fulfil the stated criteria, and those with a standard electrode potential (Eo) ~ 1 6 0 mV versus a saturated calomel electrode. The first successful mediated enzyme electrode was based on 1,l-dimethylferrocene which was adsorbed onto a graphite electrode with the enzyme having been chemically immobi- lized using the carbodiimide route.' The upper linear detec- tion limit was 30 mmol dm-3 glucose, and response times were 60-90 s.A variety of oxidoreductases2 have since been used in association with the ferrocene-modified electrode. The method seems to be generally applicable3 and mediators other than ferrocene have been used but generally they do not have the versatility of ferrocene. Dimethylferrocene-mediated electrodes are the most developed and form the basis of a commercial glucose monitor.' Traditionally, graphite with immobilized glucose oxidase and coated with a ferrocene redox mediator has been used as a * To whom correspondence should be addressed. dip-type glucose sensor,' but for flow injection (FI), the modified electrode needs to be more robust.Covalent binding onto polymer film has been studied, but the system is relatively unstable.4 Another approach is the use of a carbon paste electrode, where a quinones-6 or dimethylferrocene7~~ mediator is mixed with the carbon-binder matrix to form the working electrode. In the present study the carbon powder and mediator are bound together with cellulose triacetate, and the enzyme is chemically immobilized with the use of a nylon mesh matrix. Experimental Reagents and Materials Glucose oxidase (E.C. 1.1.3.4,1.667 pkat mg-1, purified from Aspergillus niger) , p-benzoquinone, lysine, 25% glutaral- dehyde solution and P-D-( +)-glucose were all obtained from Sigma (Poole, Dorset, UK). Nylon net was obtained from Henry Simon (Stockport, Cheshire, UK), viscose acetate (Visking tubing, 0.32 mm thickness and of relative molecular mass > 150000) was obtained from Gallenkamp (Loughbor- ough, Leicestershire, UK), cellulose triacetate from Kodak (London, UK) and carbon powder from Goodfellow Metals (Cambridge, UK).The enzyme was stored desiccated in a freezer (-5 "C). All other reagents used were of the best analytical grade available and were used without further pre-treatment. Sodium dihydrogen orthophosphate buffer (0.1 mol dm-3, pH 4.5 when freshly prepared) was of pH 4 when used. This was adjusted to the appropriate higher pH values by spiking with 4 mol dm-3 sodium hydroxide. Glucose standards were prepared from fresh P-D-( + )-glu- cose (0.1 mol dm-3) in sodium dihydrogen orthophosphate buffer (0.1 mol dm-3, pH 7) which was also used in the FI carrier stream.Immobilization of Enzyme The chemical immobilization of glucose oxidase onto nylon net was carried out as previously described.9 Nylon net (95-150 pm mesh size, 1 x 1 cm) was treated with dimethyl sulphate (30 cm3) in a boiling-tube, and placed in a water-bath at 75 k 3 "C for exactly 5 min with constant swirling. The460 ANALYST, MAY 1991. VOL. 116 boiling-tube was immersed in ice to stop the reaction. After cooling, the membrane was washed twice (or more if necessary) with methanol (30 cm3) until the methanol washings became clear. The lysine spacer molecule was attached by immersing the membrane in 30 cm3 of 0.5 mol dm-3 lysine for 2 h at ambient temperature.9 After rinsing with 0.1 mol dm-3 sodium chloride the membrane was placed in a saturated solution of p-benzoquinone for 2 h at ambient temperature. Finally, in order to attach the enzyme, the membrane was dipped into a solution of glucose oxidase (50 mg) in 5 cm3 of phosphate buffer (100 mmol dm-3, pH 7) for 2 h at ambient temperature, or overnight at 4 "C.Electrode Fabrication The chemically modified electrode material was prepared by thoroughly mixing carbon powder and ferrocene, and was bound with 20% cellulose triacetate (1 + 2 + 1 m/m) in 1,2-dichIoroethane. The ferrocene, carbon powder and poly- mer mixture was then packed into the well of an electrode holder and smoothed over with a clean flat spatula. A small drop of the cellulose triacetate solution was then placed on the electrode surface to form a protective covering.The electrode was oven-dried for 24 h at SO "C and smoothed using very fine emery paper. Flow Injection Apparatus The FI system described previously9 was used to evaluate the ferrocene-type glucose oxidase electrode using glucose stan- dards. For this, the mediated glucose oxidase electrode was completed by placing the nylon net with immobilized glucose oxidase over the chemically modified carbon paste indicator electrode. This was then set up in a modified three-electrode Stelte micro-cell (Metrohm E A 1102) assembly.'* The elec- trode potential was controlled and the current was monitored by using a Metrohm (Herisau, Switzerland) VA-detector E611 potentiostat in conjunction with a Linear Model 500 y-t chart recorder.The carrier stream and sample propulsion were driven by a four-channel Watson-Marlow (Falmouth, Cornwall, UK) peristaltic pump, and an Omnifit (Atlantic Reach, NY, USA) sample injection valve was used. All connecting tubes were of either silicone rubber or poly- tetrafluoroethylene with a nominal i.d. of 1.27 mm. A pulse suppressor was fitted between the pump and the injection valve. The indicator electrode was set at +160 mV versus a silver-silver chloride reference electrode. The following scheme illustrates the reaction sequence:' Glucose + GOD,, + gluconolactone + GODred (4) GODred + 2Fecp2R+ + GOD,, + 2Fecp2R + 2H+ ( 5 ) 2Fecp2R - 2e- + 2Fecp2R+ ( 6 ) where GOD,, and GODred are the oxidized and reduced forms of glucose oxidase, respectively, and R represents substituents in the Fecp2 ring system for Fecp2R+ and Fecp2R. Results and Discussion Optimization of the FI System for Glucose Determination Optimization of reaction p H Each enzyme electrode was optimized by varying the pH between 5 and 9 through increments of 0.5 pH unit.The resulting peak height versus pH plots reached a plateau at pH 6.8-7.2. All further work was therefore carried out at pH 7. immobilization technique and the nature of the support material, and for immobilized glucose oxidase an optimum pH of 7.0 is normal. Effect of temperature The effect of temperature on glucose sensing was studied by slowly raising the temperature from 5 to 75 "C over a period of 2 h. The sample solution and carrier stream were kept at the same temperature in a water-bath for each run.The glucose signal rose from 154 nA at S "C to a maximum of 280 nA at 38-42 "C; thereafter the signal decreased (to 230 nA at 75 "C), presumably due to denaturation of the enzyme. A repeat run on the same electrode gave a similar profile, but with a reduced glucose signal for temperatures ranging from 5 to 75 "C; the signal was reduced by 27 nA at 5 "C, 52 nA at 35- 42 "C and 41 nA at 75 "C. Optimization of flow conditions The sample volume and carrier solution flow-rate were optimized for electrodes without the use of viscose acetate membranes, and on 100 mmol dm-3 glucose standards, using an additional modified simplex optimization procedure,l3 this being an adaptation of the modified simplex optimization algorithm.14+15 For this, the control parameters of flow-rate and sample volume were examined with respect to the response criteria of peak height and run time, i.e., the time taken from sample injection to the attainment of the maximum current signal and a return to the baseline.The bias of the optimization in previous studies13-15 was maximization of the peak height with minimization of the run time. Different degrees of importance can be placed on the various response criteria according to the requirement expected of the enzyme electrode. The criteria of convergence for the simplex occur when the standard deviation (SD) of the signal responses of its vertices is less than five times the signal fluctuation (noise) of the system. The signal fluctuation can be determined by taking the SD of the signals of ten runs; this is necessary to prevent degeneracy of the simplex caused by the signal fluctuation.It is important to note that with the search method of optimiza- tion the data need to be verified, as the computation of each search point, and its direction, is dependent on the previous point. Therefore, a minimum of three runs was carried out for each subsequent cycle, the data being accepted only when there were three points with a percentage SD:signal ratio within the corresponding ratio of the original ten runs. If large signals are required, regardless of the run time, then greater weighting is placed on signal size rather than on run time. At the other extreme a system may be required that can 'O 60 L L z F Other workers have reported a broad pH range of 4.0-7.0 with a maximum response at about pH 5.5 for solubilized glucose oxidase.11J2 However, the optimum pH range is a direct result of the micro-environment of the enzyme and is related to the Fig.1 Simplex optimization of :nzyme electrode without viscose acetate membrane. Open squareF represent the experimental condi- tions; closed squares represent the normalized response of peak height and experimental run time, ie., the TRFANALYST, MAY 1991, VOL. 116 461 Table 1 Study of the interference of organic acids on the glucose sensor with and without a viscose acetate exclusion membrane Separate injection Mixed injection AilnA AilnA + 160mV +600 mV + 160 mV +600 mV Organic Without With Without Without With Without acid membrane membrane membrane membrane membrane membrane Ascorbic 0.05 0.03 3.00 1.05 1.03 4.00 Gluconic 0.00 0.00 0.00 1 .OO 1.00 1 .oo Lactic 0.00 0.00 0.00 1 .oo 1 .oo 1.00 Citric 0.00 0.00 0.00 1.00 1 .oo 1.00 Acetic 0.00 0.00 0.00 1.00 1 .00 1.00 4000 3000 P 2 2000 1000 0 4 3 P a 2 a 0, 1 - 1 0 0.1 0.2 [ G lucose]/mol dm -3 -4.0 -3.0 -2.0 -1.0 Log([glucosel/mol dm-3) Fig.2 ( a ) Glucose calibration plots for electrodes with (0) and without (A) viscose acetate exclusion membranes. ( b ) Log-lo glucose calibration plots for electrodes with (0) and without (A7 viscose acetate exclusion membranes Table 2 Response parameters of the two enzyme electrodes at + 160 mV versus silver-silver chloride reference electrodes. F-factor obtained when comparing the two electrode systems = 26.67 at p = 0.01, denominator = 8, numerator = 1 with critical value = 8.29, shows that the two electrode systems gave significantly different signals Linear Aifor detection 1 mmol range/ dm-3 mmol glucose1 Response Washout Electrode dm--3 nA SD*/nA timels time/s Without 0.10-70 22.0 0.791 15 30 With 0.10-100 20.0 0.354 25 45 * n = 5 .handle a large throughput of sample, thus placing greater emphasis on minimizing the run time. Operating conditions of sample volume and flow-rate for the systems in this study were optimized with equal emphasis being placed on the two response criteria, namely peak height and run time, which together make up the total response function (TRF) (Fig. 1). The open squares represent the experimental conditions and the closed squares represent the normalized response of peak height and experimental run time, i.e., the TRF.The initial search starts from iterations 1 , 2 and 3, and the optimum range occurs at iterations 16, 17 and 18 as shown in Fig. 1. The optimum range was the same for different search starting points. As a result of the optimization, the carrier stream flow-rate adopted was 2.3 cm3 min-1 and the corresponding sample size was 0.5 cm3, based on 100 mmol dm-3 glucose standards so as to realise large currents (= 2000 nA) without recourse to the use of instrumental noise adjustments. Sensor Selectivity The possible interference of ascorbic, gluconic, lactic, citric and acetic acids, particularly important in the analysis of food, was assessed by using two different techniques. In the first, 1 mmol dm-3 of interferent was injected directly into the system and in the other a mixture of substrate and interferent, both at a concentration of 1 mmol dm-3, was injected into the sensing system.The signal was normalized with respect to the signal obtained for 1 mmol dm-3 glucose which was taken as zero (Table 1). Ascorbic acid interfered very slightly but this interference was further reduced by covering the indicator electrode with a viscose acetate membrane. However, this acid causes severe interference even at +300 mV with the analogous glucose electrode, described by Wang et al. ,8 where the enzyme is physically, rather than chemically, immobilized. Electrode Calibration The electrodes were calibrated by FI with glucose standards over the range 0.01-200 mmol dm-3 using the optimized conditions.The lowest detectable concentration of glucose was 0.1 mmol dm-3, while the calibration was linear to 70 mmol dm-3 glucose (Fig. 2). Sensitivity analysis was carried out to determine the linear portion of the calibration plot because at higher concentra- tions the tailing-off effect of the signal from the system was caused by a physical phenomenon rather than by random error. The non-linear portion of the calibration plot cannot therefore be rejected on the basis that the points on this part of the graph are simply outliers. The sensitivity analysis was performed by determining the coefficient of regression (6) of the calibration plot, wherein each point is the mean signal for three replicate determina- tions. The point representing the highest concentration is rejected and the value of 1-2 obtained from the remaining points is compared with the previous value. If the value of r2 does not approach unity, the remaining final point of the calibration plot is again rejected and the rz of the calibration determined rejecting the point of highest concentration each time until r2 approaches unity. This procedure is carried out because the ‘outliers’ are not caused by noise (or random462 ANALYST, MAY 1991, VOL.116 error), but are the result of phenomena concerning the electrode itself. After optimization, the inclusion of the viscose acetate membrane over the electrode surface extended the linear detection range to between 0.01 and 100 mmol dm-3 (Fig. 2). The inclusion of the membrane over the electrode surface probably promotes this by limiting mass transfer to the electrode surface.This effect can be seen by comparing the time of response of the electrode system without the exclusion membrane ( ~ 1 5 s) and that with the exclusion membrane in place (=20 s); respective washout times were -30 and =45 s (Table 2). These times compare favourably with those of the analogous glucose electrode described by Wang et al. ,8 whose linear range, however, is inferior, namely, 0.5-8.0 mmol dm-3. Membrane Lifetimes Membrane lifetimes and storage stability are significant factors with regard to a wider practical role for biosensors. In order to determine membrane lifetime with respect to substrate, a 1 mmol dm-3 glucose solution was continuously pumped (2.3 cm3 min-1) over the immobilized glucose oxidase electrode.At daily intervals the electrode was washed and re-calibrated. The study showed that the membrane could withstand at least 24 h of glucose flow before any loss of enzyme activity was detected. After 7 d the signal had fallen by It was also noted that membranes stored at 4 "C in buffer gave electrodes that responded well to glucose; after frequent intermittent use (1 h per week) over 4 months the signal was 70% of that for a new electrode. 25%. Electrode Lifetimes The carbon-ferrocene chemically modified electrodes are highly stable with lifetimes of >2 years with intermittent use. The immobilized enzyme membrane can be changed when there is a signal loss due to enzyme deterioration. To achieve this long lifetime the electrodes are stored in a cool, dry and dark place when not in use.This indicates that the electrode lifetime is independent of the immobilized enzyme, the long life being attributed to the carbon paste being bound with a porous polymer matrix rather than the Nujol used by other workers. 16 Conclusion Cellulose triacetate is sufficiently porous to permit electrical contact between the reaction substrate and the electrode material, and the pores are small enough to prevent the electrode modifying material from leaching away. This is indicated by the long lifetimes of the electrodes used under tortuous FI conditions. The maximum linear limit can be extended, and interference from ascorbic acid minimized, by placing a viscose acetate membrane over the electrode surface, as the mass transfer to the electrode surface is thus reduced. The authors thank the Trustees of the Analytical Chemistry Trust Fund of the Royal Society of Chemistry for the award of an SAC Research Studentship and the Committee of Vice Chancellors and Principals for a concurrent Overseas Research Scheme Studentship (both to S.K. B.). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 References Cass, A . E. G . , Davis, G . , Francis, G. D . , Hill, H. A . O., Aston. W. J . , Higgins, I. J., Plotkin, E. V . , Scott, L. D . I . , and Turner, A . P. F., Anal. Chem., 1984,56, 667. D'Costa, E. J . , Higgins, 1. J . , andTurner, A. P. F., Biosensors, 1986, 2, 71. Mattews, D. R . , Holman, R. R., Brown, E., Steemson, J., Watson, A., Hughes, S . , and Scott, D., Lancet, 1987. 1, 778. Foulds, N. C., and Lowe, C. R., Anal. Chem.. 1988,60, 2473. Ikeda, T., Hamada, H., Miki, K . , and Senda, M., Agric. Biol. Chem., 1985.49, 541. Senda, M., Ikeda, T., Miki, K., and Hasa, H . , Anal. Sci., 1986, 2, 501. Dicks, J. M., Aston, W. J.. Davis, G . , and Turner, A . P. F., Anal. Chim. Acta, 1986, 182, 103. Wang, J . , Wu, L.-H., Lu. Z . , Li. R.. and Sanchez, J . , Anal. Chim. Acta, 1990, 228, 257. Beh, S. K . , Moody, G. J . , andThomas, J. D. R., Analyst, 1989, 114, 1421. Moody, G. J., Sanghera, G. S., and Thomas, J . D. R . , Analyst, 1986, 111, 605. Bright, H. J . , and Appleby, M. J . , J . Biol. Chem., 1969, 244. 3625. Weibel, M. K . , and Bright, H. J., J . Bid. Chem., 1971, 246, 2734. Beh, S. K . , Moody, G. J., and Thomas, J. D . R . , Anal. Proc., 1990, 27,82. Beh, S. K . , Moody, G. J . , and Thomas, J . D. R., Anal. Proc., 1989, 26, 290. Beh, S. K., Moody, G. J . , and Thomas, J . D . R . , Analyst, 1989, 114, 29. Gunasingham, H., and Tan, C.-H.. Analysr, 1990, 115, 35. Paper 0102581 J Received June 11 th, 1990 Accepted January Ist, 1991