Biosensors& Bioelectronics6 (199 1) 407-412

Internal supply of coenzyme to an amperometric glucose biosensor- based on a-chemically modified electrode Mikael Department

of Analytical

Skoog & Gillis Johansson*

Chemistry.

University

of Lund, PO Box 124, S-221 00 Lund, Sweden

(Received 24 April 1990: revised version received IO October 1990; accepted

11 October 1990)

Abstract: A biosensor for glucose using glucose dehydrogenase immobilized on a chemically modified graphite electrode was supplied with coenzyme, nicotinamide adenine dinucleotide (NAD+), through pores in the material. A graphite rod was hollowed out, leaving 0.3 mm at the end contacting the solution, filled with 10 mM NAD+ and pressurized. The response factor was 40% of that obtained when 2 mM NAD+ was mixed with the sample solution in a flow system. The coenzyme consumption was 11~1 h-l representing a 500-fold saving compared to supply through the bulk solution. The biosensor had a linear calibration curve from the detection limit, 1 PM, to 2 mM glucose and a repeatability of 0.3%. The graphite electrode was modified by adsorption of a bis-(benzophenoxazinyl)-terephthaloyl derivative in order to be able to oxidize NADH at 0 mV versus Ag/AgCI, 0.1 M KCI.

Keywords: nicotinamide adenine dinucleotide, FIA, enzyme electrode, NADH oxidation, phenoxazine mediator, immobilized enzyme, glucose dehydrogenase.

INTRODUCTION Biosensors containing dehydrogenases require normally either NAD+/NADH or NADP+/ NADPH as a coenzyme. The coenzyme is usually added to the sample or standard solutions, either batch-wise or in the carrier in a flow manifold. This is not completely satisfactory because of additional handling and dilution of the sample and a relatively large consumption of expensive coenzyme. Several alternative ways to provide coenzyme to the reaction site have been reported in the literature. *To whom all correspondence

should be addressed.

Bio~ensors & Bioelectronics 0956-5663/91/$03.50@

NAD+ can be attached to a polymer such as dextran so that it can be retained around an electrode by a dialysis membrane (Davies & Mosbach, 1974; Lee ef al., 1974; Coughlin et al., 1976; Riva et al., 1986; MAnsson & Mosbach, 1987; Buch-Rasmussen, 1989). The reduced coenzyme can be regenerated by a second enzymatically catalysed reaction, e.g. using alcohol dehydrogenase (Schulman et al., 1974). Regeneration in a biosensor is more complex than in a biotechnical process because it has to result in the formation or consumption of a detectable species in an equivalent amount. The regeneration can also be made electrochemically

by

407 1991 Elsevier Science Publishers

oxidation

of NADH

back

Ltd, England. Printed in Great Britain

to

Mikael Skoog, GillisJohansson

408

NAD+ at a platinum anode (Coughlin et al., 1976). Blaedel and Jenkins (1976) report on the electrochemical regeneration of a coenzyme bound in a cross-linked lactate dehydrogenase matrix. Another study reports on the covalent attachment of NAD+ to alcohol dehydrogenase, which was then bound to an electrode (Torstensson et al., 1980). The electrode material, graphite, was found to inactivate the complex very quickly, however. Similar observations were made with dextran-bound NAD+ in contact with a chemically modified graphite electrode (Buch-Rasmussen, 1989). This paper reports on an alternative way to supply coenzyme to the enzyme layer. A coenzyme solution is pressed through a porous electrode material from the interior of the biosensor. The enzyme layer thus receives a flux of substrate from the bulk solution and a flux of coenzyme from pores in the electrode.

EXPERIMENTAL Enzyme electrode A graphite rod (Ringsdorff-Werke, type RWO. od 3.1 mm), typically 25 mm long, was hollowed out with a 1.8 mm drill from one end, leaving about 1 mm material at the other. The end surface was polished with a line emery paper until about 0.3 mm material remained. A solution of poly(methyl methacrylate) in chloroform was painted on the outer walls of the tube to seal pores in the graphite and prevent leakage. The graphite rod was mounted in a Teflon holder with the circular end surface exposed to the solution in a flowthrough cell. A few drops of an ether solution of a phenoxabis-(benzophenoxazinyl)derivative, zine terephthaloyl chloride, BPT (Polasek et al.), was applied to the exposed electrode surface. The obtained surface coverage of mediator, as determined by cyclic voltammetry, was 5-10 nmol cm-*.

applying a drop of buffer solution, 16 ~1,containing 2 mg enzyme (16 U) on the surface (O-073 cm*) of the modified graphite. The excess was washed away after 10 min. The cross-linked electrodes were prepared as above but with the subsequent addition of a drop of 5% glutaraldehyde solution. The excess aldehyde was washed away after 30 s. The electrodes were stored in a refrigerator when not in use.

FIA manifold A flow injection analysis (FIA) manifold was set up as shown in Fig. 1. The system contained a peristaltic pump (Gilson Minipuls 2) a pneumatic injector (Cheminert SVA 8031) and a wall-jet three-electrode electrochemical cell made at this laboratory (Appelqvist et al., 1985). A separate syringe pump pressed coenzyme solution (10 mM NAD+ in water) through the modified graphite electrode. The pressure was monitored with a transducer of the strain-gauge type and the flow rate was adjusted to give an overpressure of 300 kPa (3 bar) unless otherwise stated. All measurements were made at an applied voltage of 0 mV versus Ag/Agcl, 0.1 M KCl. The cell was connected to a potentiostat covering the range from 2 nA to 2 ,uA (LKB 2050) and the signal was registered by a recorder. Samples, 50~1 throughout, were injected into the water channel. The merging channel contained a 0.25 M phosphate buffer at pH 6.7. The flow rate was 0.5 ml min-’ in each channel. Teflon tubings, id 0.5 mm, were used for all connections.

Pressure meter .

NAD+

Water Ref Buffer

Enzyme immobilization Glucose dehydrogenase, GDH, EC 1.1.1.47, from Bacillus megaterium (Merck 13732)was immobilized either by adsorption or by cross-linking. The electrodes with adsorbed enzyme were made by

Fig. 1. Schematic diagram of the flow injection manifold used for test of the biosensors. NAD+-solution. 10 mM, is supplied by a syringe pump into a hollow graphite electrode mounted in ajlow-through electrochemical cell. A loOp1 knitted tubing improves mixing and pulse-damping.

409

Supply of coenzyme to a glucose biosensor

RESULTS AND DISCUSSION Reactions The biosensor combines the enzymatic reactions with the reaction sequence of the modified electrode. P-o-glucose

+ NAD+,

S-gluconolactone S-gluconolactone

GDH

W

+ NADH + H+ + Hz0 -

gluconic acid NADH + BPT+ __+ BPTH -

(1)

BPT+ + H+ + 2e-

1

400

Overpre%m/kPa

(2) NAD+ + BPTH

300

100

(3) (4)

The enzymatic reaction (1) produces NADH at the electrode surface. Reaction (2) has a low spontaneous reaction rate but it is speeded up by lactonase present in the GDH preparation (Hanazato et al., 1988). The reaction removes the formed gluconolactone in a thermodynamically favoured hydrolysis step so that the back-reaction of (1) can be neglected. Some of the reduced coenzyme will form a complex with the mediator, BPTeNADH, which then decomposes to NADf and the reduced mediator BPTH (Gorton et al., 1984, 1985, in press; Polasek er al.). The net result of the mediator-coenzyme reactions is given by reaction (3). Detection takes place through the re-oxidation of the mediator, reaction (4). BPT was selected for this study since it adsorbs more strongly on graphite than other phenoxazines, thus increasing the expected life-time of the electrode. The reaction mechanism of BPT has not been studied in sufficient detail and the reactions (3) and (4) are therefore tentative and partly based on studies of other phenoxazines (Gorton et al., 1984.1985). The oxidation plateau is flat from - 150 mV and upwards (Polasek et al.). An oxidation potential of 0 mV was selected since the background current is low and the electrochemical interferences are few at this potential. Glucose sensor properties A coenzyme solution is pumped to the hollow electrode using a pulse-free syringe pump. Transport through the final distance (nominally03 mm) to the enzyme layer takes place through the pores of the spectrographic graphite. A higher pressure

Fig. 2. Response of a biosensor with adsorbed enzyme versus the overpressure of the NAD+-solution. The samples were 100,~ glucose and the set-up was the same as in Fig. I.

within the electrode results in a more rapid transport and therefore a higher concentration of NAD+ in the enzyme layer. Figure 2 shows the pressure dependence of the current obtained when standards of 100~~ glucose were injected. The response is good even when no external pressure is applied. The signal increases with pressure but not as fast as the expected increase in flow rate. The response factor at an overpressure of 3 bar corresponds to 40% of that obtained when the sample carrier contained 2 mM NAD+. An increase of the coenzyme concentration from 10 to 20 mM resulted in a 40% increase in the response factor. The comparison was made without external pressure. The response factor in normal FIA with a similar mediator reaches a maximum at 1.5 mM NAD+ (Appelqvist et al., 1985). High NAD+ concentrations will inhibit the enzyme. The supply of coenzyme is thus suboptimal compared to a supply through the bulk solution. The explanation is probably that a flow of coenzyme solution will counteract the diffusional influx of glucose to the enzyme layer and also increase the fraction of NADH which escapes the immobilized mediator. The distribution of coenzyme may be inhomogeneous over the electrode surface, especially at the circumference where the distance to the internal solution is more than the nominal 0.3 mm measured at the centre. It is known from previous work that the graphite is inhomogeneous even within the same rod and a local excess of NAD+ at pores near the centre

410

Mkael Skoog Gillis Johansson

may inhibit the enzyme. Graphite is also known NAD+/NADH caralytically to decompose (Wallace & Coughlin, 1978; Torstensson etal., 1980) but the rate should be low enough to be insignificant in the present application. The flow rate of coenzyme solution through the electrode was 11~1 h-‘, corresponding to a consumption of 0.11 ,umol NAD+ h-‘. This can be compared with a consumption of 60,umol NAD+ h-’ when the coenzyme is mixed into the carrier (0.5 ml min-‘, 2 mM NAD+). The consumption of coenzyme is thus 0.2% of that required in a normal flow injection procedure. The biosensor was used for glucose measurements in a flow-injection manifold containing a knitted tubing of 100,ul to damp flow pulsations and increase the mixing efficiency between the sample and the buffer solution. The response of the biosensor itself was very fast, but the throughput was limited to 60 samples h-’ because of the dispersion caused by the knitted tubing. The repeatability for injections of 1 mM glucose was 0.3% (n = 6, cross-linked enzyme). Figure 3 shows calibration curves for the biosensors with adsorbed and with cross-linked GDH. Both curves are linear from the detection limit, 1 ,UM, to 2 mM glucose. The range can be extended upwards by decreasing the injection volume as usual in FIA. The linearity was retained also for ageing biosensors. The response factor of the biosensor is thus sufticiently high for practical purposes, except in extreme low-level applications. It varies with pressure and it is therefore necessary to keep control of this variable.

800

1

P

4w-

Concentrotion/mM

Fig. 3. Calibration curvesfor a biosensor with (0) adsorbed and fl) cross-linked enzyme. The overpressure of the NAD+-solution was 300 kPa (3 bar).

Fig. 4. Response ratio of the biosensors versus age. The ordinate shows the ratio between the currents obtained with a 100 pcIMglucose sample and a 100 pu NADH sample. The curves relate to electrodes with adsorbed (0) and crosslinked (8) enzyme.

Figure 4 shows the time stability for the two electrodes. The ordinate is the ratio of the response for a 100 FM glucose injection to that for a 100 FM NADH injection. It can be seen that the relative response for the electrode with the adsorbed enzyme decreases, presumably because of desorption of enzyme. The relative response for the biosensor with cross-linked enzyme increases, however. Several effects may contribute to the observed increase in response with time. The cross-linked enzyme layer may swell with time giving rise to an increasing permeability for glucose. A swelling may also increase the availability of NAD+ at the catalytic sites. The peak current obtained with a standard glucose injection was a few percent lower after seven days than on the first day for the cross-linked electrode. This might indicate some deactivation of mediator, which was compensated for in Fig. 4 by plotting the ratio of the responses. NADH injections produce a unidirectional flux of NADH towards the electrode surface whereas a glucose injection causes a bidirectional flux of NADH from the enzyme layer, i.e. towards the electrode surface and towards the NADH-free bulk. No attempts were made to determine the fraction of NADH captured by the mediator. The glucose response should be half of the NADH response if the fluxes had been equal and the reaction with the mediator, reaction (3) had been instantaneous. Furthermore, the enzyme can only produce NADH from the/?-D anomer, which accounts for 63.5% of the total amount of glucose

Supply of coenzyme to a glucose biosensor

411

in a mutarotated solution. The enzyme layer is not sufficiently thick to give a diffusion limited glucose conversion. All these factors combine to give an electrode response of about 10% of that for NADH (see Fig. 4). The response is thus surprisingly high considering that the enzyme layer is very thin and that the coenzyme supply is suboptimal. The noise level of the hollow electrode was about twice as high as that of a solid electrode supplied with coenzyme from the carrier. This is because of the additional contact area between solution and graphite inside the electrode. A buffered coenzyme solution was found to give higher noise and background current than an unbuffered one. The reason is that narrow channels deep in the electrode material become sufficiently conducting to transmit contributions to the background current. The coenzyme was therefore dissolved in pure water during this study, in order to keep down the conductivity.

CONCLUSIONS A 500-fold saving of coenzyme can be achieved by an internal supply of reagent compared to a supply in the carrier of a FIA system. The coenzyme is added where it is needed, i.e. in the enzyme layer, rather than in the bulk of the solution. There is no need for additional enzymes or reagents for regeneration and a commercially available coenzyme can be used without synthetic modifications. Further development may result in self-consistent biosensors without need for

external pumps and supply lines. The loss in sensitivity was 60% compared to supply with the carrier, but this should be of little importance in most applications.

ACKNOWLEDGEMENT This work was supported by grants Swedish Natural Research Council.

from the

REFERENCES Appelqvist R, Marko-Varga. G., Gorton, L., Torstensson, A. & Johansson. G. (1985). Enzymatic determin-

ation of glucose in a flow system by catalytic oxidation of the nicotinamide coenzyme at a modified electrode. Anal. Chim. Acta, 169,237-47. Blaedel, W. J. & Jenkins. R. A. (1976). Study of a reagentless lactate electrode. Anal. Chem., 48, 1240-7. Buch-Rasmussen, T. (1989). Development of biosensors based on dehydrogenase enzymes and electrochemical detection, Thesis. University of Lund. Coughlin. R. W., Aizawa. M. & Charles. M. (1976). Preparation and properties of soluble-insoluble nicotinamide coenzymes. Biotechnol. Bioeng., 18, 199-208. Davies, P. & Mosbach, K. (1974). The application of immobilized NAD+ in an enzyme electrode and model enzyme reactors. Biochim. Biophys. Acta, 370, 329-38. Gorton. L., Torstensson, A.. Jaegfeldt, H. & Johansson, G. (1984). Electrocatalytic oxidation of reduced nicotinamide coenzymes by graphite electrodes modified with an adsorbed phenoxazinium salt, Meldola Blue. J. Electroanal. Chem.. 161, 10320. Gorton. L., Johansson, G. & Torstensson, A. (1985). A kinetic study of the reaction between dihydronicotinamide adenine dinucleotide (NADH) and an electrode modified by adsorption of 1,2benzophenoxazine-7-one. J. Electroanal. Chem., 196, 81-92. Gorton, L., Persson, B.. Polasek. M. & Johansson. G. Chemically modified electrodes for the electrocatalytic oxidation ofNADH. ElectroFinn Analysis, ConJ Rep.. Plenum, New York. in press. Hanazato. Y., Inatomi, K-1.. Nakako, M.. Shiono, S. & Maeda, M. (1988). Glucose-sensitive field-effect transistor with a membrane containing coimmobilized gluconolactonase and glucose ox&se. Anal. Chim. Acta, 212, 49-59. Lee. C. Y., Lappi. D. A.. Wermuth, B., Everse, J. & Kaplan, N. 0. (1974). 8-(6-Aminohexyl)-aminoadenine nucleotide derivatives for affinity chromatography. Arch. Biochem. Biophys., 163, 561-9. Mansson, M. 0. & Mosbach. K. (1987). Immobilized pyridine nucleotide coenzymes. In Pyridine Nucleotide Coenzymes, ed. D. Dolphin, R. Poulson & 0. Avramovic. vol. 2B. Wiley, New York, pp. 2 17-73. Polasek, M.. Gorton. L., Appelqvist, R., Marko-Varga. G. & Johansson, G. A glucose sensor based on crosslinked glucose dehydrogenase on a graphite electrode modified with a phenoxazine derivative. Anal. Chim.. to be published. Riva. S.. Carrea, G., Veronese, F. M. & Biickman, A F. (1986). Effect of coupling site and nature of the polymer on the coenzymatic properties of watersoluble macromolecular NAD derivatives with selected dehydrogenase enzymes. Enzyme Microb. Technol.. 9, 556-60. Schulmann, M. P.,Gupta, N. K.,Omachi.A.. Hoffman,

412 G. & Marshall, W. E. (1974). A nicotinamideadenine dinucleotide assay utilizing liver alcohol dehydrogenase. Anal. Biochem., 60, 302-l 1. Torstensson, A., Johansson. G., Mansson, M. 0.. Larsson, P. 0. & Mosbach, K. (1980). Electrochemical regeneration of NAD+ covalently bound

Mikael Skoog, GillisJohansson to liver alcohol dehydrogenase. Anal. Len., 13, 837-49. Wallace, T. C. & Coughlin, R. W. (1978). Rate of catalytic oxidation and decomposition of NADH by graphite in aqueous solutions. Biotechnol, Bioeng.. 20, 403-20.