Carbon pastes are suitable electrode material for enzyme- based biosensors and have been used since the 1970s [i]. Unique to carbon pastes is the method of enzyme immobilisation where the enzyme, mediators and enzyme stabilisers can be mixed into the carbon composite.

However, the oil in the paste is not suitable for in vivo use [ii]. Replacing the oil with epoxy resin makes robust and reusable sensors with good biocompatibility and the same advantages as bulk-modified carbon paste sensors.

Various groups have tried to analyse and model the behaviour of the carbon composites as they found that the composite acts as microelectrodes arrays [iii, iv, v]. Due to enhanced diffusive flux they have a very weak dependence of faradaic current on bulk convection because the Nernst layer is small compared with the typical boundary layer thickness caused by natural convection.

Detection of H202 is of intrinsic interest in process monitoring in cosmetics, food, chemicals, fine chemicals and pharmaceutical manufacturing. It is also the end product of all oxidase enzyme reactions with dissolved dioxygen. There have been literature reports on the use of metallised carbon particle composite electrodes, bulk modified with various oxidase enzymes for the determination of glucose, lactate, glutamate, etc.

In all these devices, conductivity was maximised and the advantages obtained from more dilute dispersions were not apparent in the data.

Understanding the H202 detection mechanism is critical to

enhancing overall sensor performance of oxidase-based biosensors. The present invention results from an investigation into the factors affecting sensitivity (here defined as electrolytic current per unit concentration) of the H202 detection, in particular the effect of electrode modification with the oxidase enzymes and the co-factor flavin adenine dinucleotide (FAD) and related compounds. Durliat et al have reported the use of FAD as a mediator between glucose oxidase and platinum electrodes when present in the electrolyte solution [vi].

In vivo characterisation is important before implantable electrodes and sensors such as pace makers are used in the complex and hostile biological milieu. The two key factors are selectivity and biocompatibility. Electroanalysis in vivo presents a formidable analytical challenge due to the presence of unknown and time-varying concentrations of endogenous electroactive materials and surface-active compounds which can block the electrode reaction. Selectivity to the analyte can be improved by the use of permselective membranes [vii] or chemical modification (platinum group metals e. g. rhodium, iridium, ruthenium [viii] and palladium-gold mix [ix]) that can decrease the overvoltage for the oxidation or reduction of H202.

Blood plasma contains over 100 proteins. We have only considered those that are expected to compete effectively for the interface in the initial collision or contact process, i. e. those that are present at sufficient plasma concentration to be considered major constituents and can also adsorb to the surface. Albumin is ubiquitous, present in high concentrations and adsorbs to the surface. Fibrinogen is also present in relatively high concentrations but more importantly it adsorbs"irreversibly" [x]. Platelets then adhere to this proteinated surface and this is the first observable step in thrombosis on a foreign surface [xi].

Polycarboxybetaine (PCB), a phosphobetaine-based polymer, has useful anti-bioadherent properties with respect to mammalian cell adhesion and finds applications in contact lenses and vascular stents [xii].

The present invention is based on the finding that electrodes made from metallised carbon epoxy composites modified in a specific manner surprisingly show substantially increased sensitivity to the oxidation and reduction of H202 and other reactive oxygen species.

According to the present invention there is provided metallised carbon epoxy composites that have been modified with isoalloxazine, riboflavin, flavin mononuleotide and/or flavin adenine dinucleotide. It will be appreciated that all four of these modifiers have in common an isoalloxazine ring system in their structures. In each case, the reaction is believed to occur on the isoalloxazine molecule.

According to the present invention there is also provided a method for the modification of metallised carbon epoxy composites with isoalloxazine, riboflavin, flavin mononucleotide and/or flavin adenine dinucleotide. The modification can be performed using the techniques described in the experimental account that follows.

According to the present invention there is still further provided an electrode which comprises a metallised carbon epoxy composite modified with isoalloxazine, riboflavin, flavin mononucleotide and/or flavin adenine dinucleotide.

The modification may be either bulk or surface modification.

In the case of bulk modification the flavin adenine dinucleotide or other isoalloxazine derivative is added in the range 0. 1% (w/w) to 10% (w/w).

Surface modification may be achieved by electropolymerisation from aqueous solutions containing other solvents (specifically methanol, ethanol, acetonitrile) either alone or with water. The isoalloxazine containing molecule should be present in solution in the concentration range 1 mM to 100 mM. Chemical oxidation may also be used to effect polymerisation.

Surface films may also be applied either by painting or spin-coating from similar solutions followed by either passive or convection-assisted evaporation of the solvent.

The electrodes of this invention are intended for use in the

detection and measurement of reactive oxygen species, such as superoxide (02-), perhydroxyl (HO2-), hydroxyl radical (OH), peroxide (H02) and, in particular, hydrogen peroxide (H202). One important application of H202 measurement is in oxidase-based biosensors, where H202 is produced by the enzyme reaction with dissolved oxygen. Other uses of the electrodes of this invention include as a component of fuel cells, including implantable fuel cells that are a power source for pace makers and other implanted devices.

The metal component of the metallised carbon epoxy composite may comprise any of the platinum group metals, namely ruthenium, osmium, rhodium, iridium, palladium and platinum ; rhodium is preferred. It may comprise combinations of two or more such metals or of one or more platinum group metals and other metals.

The present invention will now be further described by reference to the following example and the accompanying drawings:- Figure 1: Calibration of the glucose sensor in 0.1 M PBS, pH 7.4, with 3 mg/ml fibrinogen + 4% albumin and without the proteins at 0.4 V. 1 mM aliquots of glucose solution added every 2 mins from 1-20 mM. Experiments done at 37°C.

Figure 2: Calibrations at 0.4 V of a glucose sensor coated with polycarboxybetaine. 0.1 M PBS, pH 7.4 (a) with 4% albumin and (b) with 3mg/ml fibrinogen. 1 mM aliquots of glucose solution added every 2 mins (1-20 mM). After rinsing calibration repeated in PBS. Experiments done at 37°C.

Figure 3: Calibrations of Rh/C, glucose with and without PEI and lactate sensors in 0.1 M PBS, pH 7.4, using hydrogen peroxide as the analyte at OV. 1 mM aliquots of H202 solution added every 2 mins (1-20 mM). Cathodic currents were obtained for all the sensors but they have been shown as positive currents here.

Figure 4: Comparison of the steady state calibrations in 0.1 M PBS, pH 7.4, at 0.4 V. (a) Rh/C and RhC+FAD using H202 as the

analyte and (b) lactate sensors (with and without PEI) modified with 1% FAD using lactate as the analyte.

EXAMPLE Experimental Procedures Reagents Ultra F purity graphite (1 pm powder) was obtained from Alfa.

The epoxy EpoTek H77 was acquired from Promatech Ltd.

Poly (ethylenimine) (PEI, 50% solution in water), poly (vinyl chloride) (PVC), o-phenylenediamine (o-PD) and 5% Rhodium-on-carbon (Rh/C) were from Aldrich. Glucose oxidase (GOD, EC 1.1. 3.4, Type VII-S), lactate oxidase (LOD, EC 1.1. 3.2), D- (+)-Glucose, L- (+)-Lactic acid, DL-Dithiothreitol (DTT) and flavin adenine dinucleotide (FAD) were bought from Sigma.

Poly (carboxybetaine) (PCB) was a gift from Prof. A. W. Lloyd, School of Pharmacy & Biomolecular Sciences, University of Brighton. All reagents were used as received unless stated otherwise.

Solutions Stock solution of glucose (1 M) was prepared and left overnight in a refrigerator at 4°C to allow for mutarotation to occur. It was then stored frozen in vials and thawed before use when required. 1 M solutions of lactic acid were prepared and stored as above. Fresh solutions of approximately 1 M hydrogen peroxide were prepared on the day of use and assayed by iodide titration [xiii]. 0.1 M phosphate buffered saline (PBS), at pH 7.4 was used as the electrolyte in all the experiments. All the solutions were prepared in deionised water. Appropriate aliquots of these stock solutions were added to the buffer as required.

Membrane Coating A solution of o-PD dihydrochloride (5 mM) was prepared in

PBS (pH 7.4). The membrane was electropolymerised onto the electrode surface at 0.9 V vs. AglAgCl for 15 min in a stirred solution (henceforth referred to as o-PPD). Thereafter it was air-dried and left overnight to hydrate in buffer. A 12% solution of PVC was made up in tetrahydrofuran and the electrode was coated by dipping it in solution and air-drying.

Electrodes were also dip-coated with PCB, which was used as received.

All coated electrodes were hydrated overnight in PBS before use.

Apparatus Electrochemical measurements were carried out using a CV37 potentiostat (BAS) using a homemade AglAgCl reference electrode (to which all electrode potentials are referred) and a stainless steel needle as auxiliary electrode. The potentiostat was linked to a computer via an A/D converter for data acquisition and analysis.

Electrode Preparation Rh/C was mixed with the epoxy in a ratio of 5: 16 (carbon: epoxy) by weight. The optimal graphite proportion for electroanalytical applications is related to the volume fraction for the type of carbon and epoxy used. It is associated with the conductivity and minimal double layer capacitance of the composite [xiv, xv]. The oxidase enzyme was mixed with 2% poly (ethylenimine) (PEI), which acts as the enzyme stabiliser [xvi]. DL- Dithiothreitol (DTT) was also used as a stabiliser for lactate oxidase [xvii].

The carbon composite was mixed in with 1.5% (w/w) of the enzyme- stabiliser mixture. Electrode bodies consisting of wired nylon cannulae set in epoxy resin with a cavity for packing the composite were constructed.

The biocomposite was packed into the electrode cavity and cured for 3 weeks at 4°C. Plain carbon and Rh/C epoxy electrodes were prepared without the enzyme. Flavin adenine dinucleotide (FAD) modified electrodes were prepared by incorporating FAD (1% w/w) directly into the composite both with and without enzyme. Other methods by which these

modified electrodes could be prepared include painting FAD solution on to the surface, spin coating, electro-polymerisation and chemical polymerisation. Some of the sensors were coated with polyvinyl chloride (PVC), o-phenylenediamine (o-PPD) and polycarboxybetaine (PCB) membranes as described above.

Results and Discussion Effect of Stabilisers on the Enzyme Sensors Polyethylenimine (PEI) was used as an enzyme stabiliser for the glucose and lactate sensors. The sensitivity was increased as much as 60% in the presence of the PEI. The reaction velocity (Vmax), which is directly proportional to the amount of active enzyme present, was calculated using Lineweaver-Burke plots. In the case of the glucose sensor Vmax was increased from 1.61 to 4.32 x 104 mols-m~2. Lactate sensors were prepared with both PEI and DL-Dithiothreitol (DTT) as stabilisers. For the lactate biosensor without stabiliser Vmax = 4.18 x 10-7 mols~'m~2, with DTT it is increased to 1. 18x 10 mols'm-2 and with PEI it was a maximum of 3.35 x 10 mols'm-2. Thus both the stabilisers give increased sensitivity but the lactate sensor with PEI appears to give a higher and more stable current response as compared with DTT. This indicates that the stabiliser is effective in retaining enzyme activity.

Selectivitv During calibrations, ascorbic acid and uric acid were added in quantities similar to their maximum expected biological concentrations in plasma and acetaminophen in maximum likely therapeutic concentration [xvii]. Successive aliquots of 1 mM glucose or interferent (concentrations as in Table 1) were added and the current recorded. These calibrations were done on bare sensors and sensors membrane coated with o- phenylenediamine (o-PPD) and polyvinyl chloride (PVC) membranes were

examined to see if they improve selectivity.

Table 1 : Representative concentration values of the common interfering species found in human blood.

Interfering species Representative Concentration (mM) Acetaminophen (paracetamol) 0.13 Ascorbic Acid 0.125 Uric Acid 0.33 Two indices were used to characterise selectivity:- Relative Sensitivity = SInt/S. # 100% Equation 1 A Maximum Expected Error = Ibt x 100% Equation 2 A where S, nt is sensitivity of the sensor to the interferent, Sa is its sensitivity to the analyte, lint is the current response to maximum concentration of interferent as found in plasma and lais the current response to normal concentration of analyte in plasma.

The bare sensors gave a constant response to glucose with a sensitivity of 17.8 nA/mM. All the interferents were detectable as summarised in Table 2. However, uric acid gave the maximum expected error over glucose closely followed by acetaminophen. When the sensor is coated with o-PPD, there is a greatly enhanced response to ascorbic acid, a 10% increase in response to uric acid while the sensitivity to acetaminophen is reduced. The sensitivity to glucose is also decreased to 6.65 nA/mM. Table 2 gives the relative sensitivity and the maximum expected error that can be caused by the interferents, which were

calculated from Equations 1 and 2, respectively.

Table 2: Comparison of the response of a glucose sensor to interferents when bare and coated with o-PPD in 0.1 M PBS, pH 7.4 at 0.4 V. Normal concentration of glucose taken as 5 mM and interferent concentration from Table 1.

Interferent Relative Sensitivity as Maximum Expected Compared to Glucose Percentage Error in Plasma Bare o-PPD Bare o-PPD Acetaminophen 2000% 1400% 50% 40% Ascorbic Acid 900% 3800% 20% 90% Uric Acid 1000% 1100% 65% 70% The effect of interferents on the lactate sensor with DTT was also tested with the results summarised in Table 3. The bare sensor is again very sensitive to the interferents but on coating it with PVC the respective sensitivity is greatly reduced (by 80% to acetaminophen, 70% to ascorbic acid and 65% to uric acid). However the PVC coating also reduced the sensitivity to lactate by 88%.

Table 3: Comparison of the response of a lactate sensor to interferents when bare and coated with PVC at 0.4 V. Remaining conditions same as Table 2.

Interferent Relative Sensitivity as Maximum Expected Compared to Lactate Percentage Error in Plasma Bare PVC Bare PVC Acetaminophen 3800% 800% 1000% 200% Ascorbic Acid 1600% 500% 400% 125% Uric Acid 1300% 450% 800% 300% The relative sensitivity calculated in Table 2 and Table 3 appears to be very high, i. e. the sensor is very sensitive to the interferents.

This is a major problem with oxidase-based sensors (even those commercially available). However, the actual amount of interferent present in vivo is a fraction of that used in the calculation as we are calculating maximum error in the worst case. Since the normal concentrations of glucose (5 mM) and lactate (0.5 mM) in blood have been used, a larger error is found in the lactate sensor as compared with the glucose sensor.

Comparing the data obtained for both o-PPD and PVC it appears that o-PPD increases the sensitivity of the sensor towards ascorbic acid by a factor of four. This may be due to accumulation of anionic ascorbate in the cationic o-PPD matrix. This is consistent with the observation that sensitivity to uric acid was also increased by 10%.

However, sensitivity to acetaminophen was reduced by 30%. PVC reduces the effect of all the interferents by greater amounts (Acetaminophen by 80%, Ascorbic acid by 70% and uric acid by 65%). Neither membrane was

totally effective in excluding interferents and o-PPD actually displayed an affinity towards these species that cannot be explained, especially ascorbic acid, contrary to much of previously published data [xviii, xix]. The ineffectiveness might be due to incomplete coverage of the sensor with the polymer as found by O'Neill on platinum electrodes [xx].

A difficulty in comparing the present data with previous results is the number of arbitrary indices used to characterise interference. One exception is the study by Lowry and O'Neill [xx] that reported the relative sensitivity for glucose over ascorbic acid. Fitness for purpose can more easily be assessed where errors are calculated using Equations 1 and 2.

In terms of these indices, they found a relative sensitivity of 7835% for ascorbic acid and about 340% for uric acid with respect to glucose. The maximum error ascorbic acid this would cause is 195% and uric acid is 22%.

The effects of operating potential. on interference and selectivity for the bare glucose sensor are summarised in Table 4. At all the potentials the sensor responds to glucose but the sensitivity to interferents is greater than that to glucose. The pattern is somewhat complex e. g. , at 0 V there is high selectivity in the presence of acetaminophen and ascorbic acid but effect of uric acid is reduced at 0.4 V.

Similar results were obtained for the lactate sensor. These differences presumably reflect the physicochemical properties of the interferents and might be exploited to minimise the effect of a dominant interferent in a particular application. However, in general, using 0 V and lower as the operating potential reduced the effect of interferents. This is a direct effect of using rhodium-modified carbon which enables a wider operating potential range compared with plain carbon or electrodes fashioned from bulk platinum group metals or of gold.

Table 4: Comparison of different operating potentials for a bare glucose sensor in the presence of acetaminophen, ascorbic acid and uric acid.

Remaining conditions similar to Table 2.

Interferent Percentage Sensitivity Maximum Expected as Compared to Glucose Percentage Error in Plasma 0. 4V 0V-0. 1 V 0. 4V 0V-0. 1 V Acetaminophen 2000% 400% 950% 50% 10% 25% Ascorbic Acid 900% 650% 2300% 20% 15% 55% Uric Acid 1000% 1100% 1400% 65% 70% 90% Biocompatibility Biocompatibility was studied by examining the effect of plasma proteins albumin (4%) and fibrinogen (3 mg/ml). The protein concentrations used are those found in human blood plasma.

Polycarboxybetaine (PCB) a novel biocompatible coating (Figure 4 shows the monomer) was tested along with the o-PPD and PVC membranes. All experiments in this section were done at 37°C.

A bare glucose sensor was calibrated in PBS containing 3mg/ml fibrinogen and 4% albumin using glucose as the analyte. The sensor was rinsed with deionised water and the calibration repeated in PBS (Figure 1). There is about 35% loss in sensitivity to glucose in the protein solution. This could be due to: (i) the adsorption of protein on to the sensor surface (ii) portioning of glucose in the protein or (iii) the diffusion coefficient of glucose in solution is reduced. On rinsing the sensors and re- calibrating, the sensitivity was regained. When the sensors were coated with PCB and calibrated in PBS with 4% albumin, the calibration was reproducible. Similarly, when the coated sensor was calibrated in 3mg/ml solution of fibrinogen, then rinsed and recalibrated in buffer the results were

reproducible (Figure 2). Thus it appears that the protein adsorbs onto the bare sensor surface and either blocks the active sites on the electrode surface or creates an additional diffusion barrier between the sensor and the analyte.

To further test the biocompatibility of the PCB membrane, the effect of exposure to tissue sample (bovine caudal intervertebral disc) on the coated sensor surface were studied. When a bare lactate sensor was exposed to the disc there was about 13% loss in sensitivity to lactate after exposure for 3-4 hours. Comparing this with the above results obtained for the bare sensor in fibrinogen and albumin solutions, it becomes clear that the above protocol used is a valid method of testing biocompatibility. When the experiment was repeated with a glucose sensor coated with PCB the calibrations before and after exposure were not significantly altered.

Detection of Hydrogen Peroxide on Enzyme Sensors The glucose and lactate sensors were calibrated with H202.

Glucose oxidase and lactate oxidase modified electrodes showed increased sensitivity towards the H202 by about 60% as compared with the Rh/C electrode as seen in Figure 3. When the enzyme stabiliser PEI was included in the composite, it further enhanced the sensitivity to 80%. PEI modified electrodes (Rh/C+PEI) without enzyme showed no change in its sensitivity to H202 (data not shown) indicating that the current increase in enzyme-modified sensors is due to the presence of more active enzyme.

Table 5 summarises the results of all the types of electrodes investigated in this publication with respect to their response to H202.

Table 5: Summary of results for all types of sensors investigated with respect to their sensitivity to H202.

Type of Sensitivity to Linear Range Limit of Electrode H202 (mM) Detection (mM) (nA/mMmm2) Bare Carbon 0. 5 0-20 0. 5 Rh/C 102.5 0-20 0.3 Rh/C+GOD 171.04 0-4 0.2 64.65 5-20 Rh/C+LOD 118.4 0-20 0.7 Rh/C+PEI 105 0-20 1.4 Rh/C+GOD+PEI 320.66 0-20 0.2 Rh/C+FAD 347.42 0-20 0.1 Role of C-enzyme Flavin Adenine Dinucleotide (FAD) The increased sensitivity of the enzyme-modified electrode to H202 is due to some component present in the biocomposite. Since the co-enzyme flavin adenine dinucleotide (FAD) is the redox centre for the enzyme and FAD has been shown to react with both H202 and superoxide [xxi], the possible role of FAD as an electrocatalyst was examined. FAD modified electrodes were constructed and the calibrations compared with the Rh/C and enzyme modified electrodes.

Figure 4 (a) shows the steady state calibration of the FAD electrodes using H202 as the analyte. An increased sensitivity of about 70% in Rh/C+FAD electrode as compared with the Rh/C electrode was observed over the range tested (0.1-20 mM H202).

With glucose sensors an increase of about 50% was detected in the sensitivity to glucose, especially in the low concentration range below

3 mM. Vmax was increased from 4.39 x 10-7 mols~m~2 to 1.76 x 10-6 mols-'m-2 and KM is increased from 1.87 mM to 4.19 mM. Lactate sensors were prepared and calibrated in a similar way as the glucose sensors, both with and without FAD. The calibrations are shown in Figure 4 (b) using lactate as the analyte. Besides the increased sensitivity, an extension in the linear range was observed. Vmax was again increased from 9.43 x 10-2 mols'm2 to 1.27 x 10-1 mols~'m~2. Also the limit of detection for the glucose and lactate sensors with FAD was decreased.

It has previously been shown by others that flavin mononucleotide [xxii] and FAD [xxiii] can regenerate apparently denatured aged enzyme suggesting that denaturation of the enzyme is to some extent accounted for by loss of co-factor. These data support this hypothesis and suggest that FAD can also catalyse the co-substrate reaction. Bulk modification with FAD thus provides a general way of finessing linear range and sensitivity of oxidase-based sensors.

Mechanism of the FAD Redox Reaction Previous investigations of H202 on carbon electrodes [xxiv] show that the first electron transfer to hydrogen peroxide is the rate- determining step of the reaction. In the reduction of 02 on carbon the one electron reduction to 02-is the slow step of the reduction. This is consistent with the current-voltage curves obtained for these electrodes (data not shown).

At physiological pH FAD forms a free radical (semiquinone) by one electron reduction. Chi and Dong [xxv] have demonstrated that FAD in solution (at-0.47 V) and in films (at-0.37 V) can catalyse electrochemical reduction of 02 to H202. Though their proposed mechanism shows a two electron reduction of the flavin film, it is well known that the one electron semiquinone product can react with the superoxide intermediate to form a hydroperoxide [xxi]. 4a-Flavin hydroperoxides (i. e. the addition product of FAD*+ 02'-) are intermediates

in both flavoprotein oxidase and mono-oxygenase chemistry [xxvi]. It is therefore likely that such radical-radical coupling at essentially diffusion- limited rates can account for the enhanced sensitivity of the FAD-modified electrodes towards 02/H202 reaction.

Microelectrode Array like Behaviour of the Electrode One mole of H202 is generated per mole of glucose oxidised by the enzyme. The H202 formed at the end of the enzyme reaction is freely exchanged with the stirred solution. Thus it is important to establish the fraction of H202, which is"captured"by the electrode since high efficiency would reduce the dependence on [02 (aq) ] thus allowing these devices to be used over a wider range of oxygen concentrations.

Calculating the ratio of the glucose sensitivity to H202 sensitivity for the enzyme biosensor gives the"capture efficiency"of that biosensor. For the glucose sensor it was found to be typically 0.30 0.04.

Thus 30% of the H202 produced by the enzyme reaction is detected at the electrode surface demonstrating high diffusional fluxes towards the electrode.

Whilst the steady state calibrations were being done the stirrer was switched on and no significant shift was observed in the baseline current. Therefore the sensors do not show any significant stirring artefact. O'Hare et al have reported similar results when they compared the effect of stirring on a glassy carbon electrode with the graphite composite electrode [xvii]. Slow rate cyclic voltammograms of the sensors showed sigmoidal current voltage curves which are characteristic of microelectrodes. These observations are confirmed by their study.

The present invention thus enables the production of carbon composite electrodes that are easy to make, biocompatible, robust, have a short response time (25-35 s) and a working life of about 12 months. This makes them highly suitable for in vivo applications especially when coated with PCB. The membrane provides an exceptional barrier to prevent

protein adsorption and fouling of the electrode surface.

Incorporating FAD into the composite of enzyme-modified electrodes enhances the sensitivity of the biosensor to the respective analyte, decreases the detection limit and also allows the manipulation of the linear range. These effects can be entirely explained by the catalysis of H202 oxidation or reduction at the electrode. Therefore incorporating FAD into the biosensor design would be a general approach to help enhance the performance of all oxidase-based sensors, especially when expensive enzymes with low activity are used due to high specificity.

References [i] Q. Yang, P. Atanasov and E. Wilkins, Biosensors and Bioelectronics, 1999, 14,203.

[ii] D. E. Ormonde and R. D. O'Neill, J. Electroanal. Chem., 1989,261, 463.

[iii] D. E. Tallman and S. L. Petersen, Electroanalysis, 1990,2, 499.

[iv] D. E. Weisshaar, D. E. Tallman and J. L. Anderson, Anal. Chem., 1981,53, 1809.

[v] N. Sleszynski, J. Osteryoung and M. Carter, Anal. Chem., 1984,56, 130.

[vi] H. Durliat, M. B. Barrau and M. Comtat, Bioelectrochem. & Bioenergetics, 1988,19, 413.

[vii] S. A. Emr and A. M. Yacynych, Electroanalysis, 1995,7, No. 10, 913.

[viii] A. J. Appleby, Modem Aspects Electrochem., (Eds. B. E. Conway and J. O'M. Bockris), Plenum, New York, 1974,9, pp. 369-478.

[ix] F. Céspedes, E. Martinez-Fabregas and S. Alegret, Anal. Chim.

Acta, 1993, 284,21.

[x] J. D. Andrade and V. Hlady, Annals Of The New York Academy Of Sciences, 1987,516, 158.

[xi] S. W. Kim and R. G. Lee, Applied Chemistry at Protein Interfaces, Chairman: R. E. Baier, Advances in Chemistry Series 145,1975, pp. 218- 229.

[xii] A. C. Hunter, E. Ampatizdou, S. Meikle, C. Steward and A. W. Lloyd, Int. J. Artificial Organs, 1999,22, No. 6,411.

[xiii] C. Watson, Official and standardised methods of analysis, RSC, Cambridge, 1994,166.

[xiv] J. Navarro-Laboulais, J. Trijueque, F. Vicente, H. Scholl, J.

Electroanal. Chem., 1994,379, 159.

[xv] J. Navarro-Laboulais, J. Trijueque, J. J. Garcia-Jareno, F. Vicente, H. Scholl, J. Electroanal. Chem., 1995,399, 115.

[xvi] J. Wang, L. Chen and J. Liu, Electroanalysis, 1997,9, 298.

[xvii] Q. Yang, P. Atanasov and E. Wilkins, Biosensors and Bioelectronics, 1999,14, 203.

[xviii] J. Wang, L. Chen, J. Liu and F. Lu, Electroanalysis, 1996,8, No. 12, 1127.

[xix] J. P. Lowry, M. Miele, R. D. O'Neill, M. G. Boutelle and M. Fillenz, J.

Neurosci. Methods, 1998,79, 65.

[xx] J. P. Lowry and R. D. O'Neill, Electroanal., 1994,6, 369.

[xxi] D. T. Sawyer, Oxygen Chemistry, Oxford University Press, Oxford, 1991, pp. 181-183.

[xxii] D. O'Hare, K. Parker and C. P. Winlove, Bioelectrochem. & Bioenergetics, 1990,23, 203.

[xxiii] A. Szucs, G. D. Hitchens and J. O'M. Bockris, Bioelectrochem. & Bioenergetics, 1989,21, 133.

[xiv] M. R. Tarasevich and E. l. Khrushcheva, Comprehensive Treatise of Electrochemistry, 1983,7, 295.

[xxv] Q. Chi and S. Dong, J. Electroanal. Chem., 1994,369, 169.

[xxvi] C. Walsh, Acc. Chem. Res. , 1980,13, 148.

[xxvii] D. O'Hare, J. V. Macpherson, A. Willows, Electrochemical Communications, 2002,4, 245.