In Vivo Brain Glucose Measurements: Differential Normal Pulse

Anal. Chem. 1996, 68, 4358-4364

In Vivo Brain Glucose Measurements: Differential Normal Pulse Voltammetry with Enzyme-Modified Carbon Fiber Microelectrodes Larissa I. Netchiporouk,†,‡ Nataliya F. Shram,†,‡ Nicole Jaffrezic-Renault,† Claude Martelet,*,† and Raymond Cespuglio§

Laboratory of Physicochemistry of Interfaces UMR CNRS 5621 IFoS, Ecole Centrale de Lyon, B.P. 163, 69131 Ecully, France, Sector of Bioelectronics, Kiev University, P.O. Box 152, Kiev-1 252001, Ukraine, and Department of Experimental Medicine U INSERM 52, Cl. Bernard University, 8 avenue Rockefeller, 69373 Lyon, France

The enzyme glucose oxidase was immobilized on the surface of carbon fiber microelectrodes (CFMEs) either by cross-linking in glutaraldehyde vapor or by enzyme entrapment in electropolymerized films of m-phenylenediamine or resorcinol. The cross-linked enzymatic layer was, in the given conditions, covered with an additional membrane of Nafion or cellulose acetate. The prepared glucose sensors were tested using differential normal pulse voltammetry (DNPV, in which the scan comprises successive double pulses (“prepulse and pulse”), the prepulses are of increasing amplitude, and the current measured is the differential of the current existing between each prepulse and pulse). With properly chosen DNPV parameters, the response to glucose presented a peak at a potential of about 1 V versus an Ag/AgCl reference, owing to the oxidation of enzymatically produced hydrogen peroxide. The calibration curves obtained (peak height/glucose concentration) were linear from 0.3-0.5 up to 1.5-6.5 mM and showed a sensitivity ranging from 1.4 up to 34.5 mA M-1 cm-2, depending on the sensor type. The DNPV response to glucose exhibited an essential insensitivity toward easily oxidizable interfering substances such as ascorbic acid and acetaminophen present at physiological concentrations. Peptides, the interfering species typical of the cerebral medium, were effectively retained by the above additional membranes. Concentration values of glucose in plasma and cerebrospinal fluid, determined in vitro from the DNPV peak height, agreed well with those measured by standard procedures. In the anesthetized rat, extracellular brain concentration of glucose was also monitored during administration of either insulin or glucagon. Under such pharmacological conditions, the changes observed in the peak height were in perfect agreement with the known effects induced by both substances. Considerable efforts have been devoted to the search for a reliable miniaturized glucose sensor. This is primarily due to the importance of glucose monitoring in diseases like diabetes mellitus1-3 but also to the still unsolved questions of fundamental research related to energy production and consumption.4 Indeed, †

Ecole Centrale de Lyon. Kiev University. § Cl. Bernard University. ‡

4358 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

despite the cardinal role played by glucose in cerebral metabolism, its concentration in the extracellular brain environment is still poorly documented. Recently, different studies have been reported on glucose detection in extracellular fluid5,6 and on the relationship with neuronal activity.7,8 All the assays were carried out using the enzymatic method, which remains the most widely employed. Glucose measured by this method undergoes enzymecatalyzed oxidation in accordance with the following reactions:

β-D-glucose + GOx/FAD f glucone-δ-lactone + GOx/FADH2 (1) GOx/FADH2 + O2 f GOx/FAD + H2O2

(2)

where GOx (glucose oxidase) is the oxidoreductase enzyme and FAD is the oxidized form of the prosthetic group flavin adenine dinucleotide. The most common concept of glucose sensor consists of the amperometric oxidation of the hydrogen peroxide produced according to the two above reactions, GOx being immobilized at or near the electrode surface. The difficulties encountered when using these sensors can be classified into two categories. The first concerns the performance of the enzyme electrode, which depends on the enzyme immobilization method, the thickness and stability of the entrapping membrane, and the activity and stability of the entrapped enzyme. The second is related to the specificity of the measurement in the working conditions. The amperometric detection of glucose in the living brain suffers from the direct faradic interferences produced by ascorbic acid (AA), uric acid, and acetaminophen, species having an oxidation potential much lower than that of H2O2. Several methods to overcome this problem have been suggested. They can be summarized as follows: (1) Turner, A. P. F., Karube, I., Wilson, G. S., Eds. Biosensors: Fundamentals and Applications; Oxford University Press: Oxford, 1987. (2) Meyerhoff, M. E. Clin. Chem. 1990, 36, 1567-1572. (3) Schumann, W. S.; Schmidt, H. L. In Advances in Biosensors; Turner, A. P. F.; Ed.; Jai Press: London, 1992; Vol. 2, pp 79-131. (4) Erecin ˜ska, M.; Silver, I. A. J. Cerebral Blood Flow 1989, 9, 2-19. (5) Zilkha, E.; Koshy, A.; Obrenovitch, T. P.; Benetto, H. P.; Symon, L. Anal. Lett. 1994, 27 (3), 453-473. (6) Lowry, J. P.; McAteer, K.; ElAtrash, S. S.; Duff, A.; O’Neil, R. D. Anal. Chem. 1994, 66, 1754-1761. (7) Fellows, L. K.; Boutelle, M. G.; Fillenz, M. J. Neurochem. 1992, 59 (6), 2141-2147. (8) Silver, I. A.; Erecin ˜ska, M. J. Neurosci. 1994, 14 (8), 5068-5076. S0003-2700(96)00190-4 CCC: $12.00

© 1996 American Chemical Society

(1) Covering the enzymatic layer with an additional membrane able to exclude the interfering molecules on the basis of their charge9,10 or their size.11,12 This procedure, rather simple, has been used in many studies.13,14 An interesting variant of this approach is the immobilization of GOx in permselective electropolymerized films that allow the passage of H2O2 but prevent the interferents from reaching the electrode surface.15-17 Apart from limiting the interferences, additional membranes have been used as diffusional barriers for glucose, thus reducing the “oxygen deficit” (reaction 2). However, although up to four additional membranes have sometimes been employed,8 the in vivo performance of these amperometric sensors still remains poorly reliable. (2) Incorporation of an active component such as an enzyme, a redox reagent, or a ligand in the outer layer of the sensor in order to inactivate the incoming interferents.8,18 While this method has promoted some in vitro applications, its use in vivo has always been reported in combination with additional membranes.8 (3) Lowering the potential applied to the electrode (a) by incorporation of a mediator shuttling the electrons between the reduced form of the enzyme and the electrode,19-22 (b) by coimmobilization of GOx with the enzyme horseradish peroxidase (HRP), in order to detect H2O2 at a working potential of 0 mV versus SCE (standard calomel electrode) via a direct electron transport between HRP and the electrode surface,23 or (c) by optimization of the applied potential in order to decrease the current produced by oxidation of interfering species while leaving the current generated by glucose virtually unchanged.24 Unfortunately, the construction of a multicomponent system also generates the above-mentioned problems with regard to the sensor performance. In addition, for the enzymatic redox processes, the mediator must be competitive with the oxygen of the medium (reaction 2). Clearly, such a reliable mediator is not yet available. As for the applied potential, the lowering of which is reported to reduce interfering effects, it is rather characteristic of a given sensor design and of a particular interfering species (acetaminophen). (4) Use of a preoxidizing cell maintained at a potential great enough to destroy the interfering compounds.5 This method, (9) Moussy, F.; Jakeway, S.; Harrison, D. J.; Rajotte, R. V. Anal. Chem. 1994, 66, 3882-3888. (10) Wang, J.; Lin, S. M. Electroanalysis 1990, 2, 253-256. (11) Bindra, D. S.; Zhang, Y.; Wilson, G. S.; Sternberg, R.; The´venot, D. R.; Moatti, D.; Reach, G. Anal. Chem. 1991, 63, 1692-1696. (12) Christie, I. M.; Treloar, P. H.; Vadgama, P. Anal. Chim. Acta 1992, 269, 65-73. (13) Vaydya, R.; Wilkins, E. Electroanalysis 1994, 6, 677-682. (14) Zhang, Y.; Hu, Y.; Wilson, G. S.; Moatti-Sirat, D.; Poitout, V.; Reach, G. Anal. Chem. 1994, 66, 1183-1188. (15) Sasso, S. V.; Pierce, R. J.; Walla, R.; Yacynych, A. M. Anal. Chem. 1990, 62, 1111-1117. (16) Malitesta, C.; Palmisano, F.; Torsi, L.; Zambonin, P. G. Anal. Chem. 1990, 62, 2735-2740. (17) Rishpon, J.; Gottesfeld, S.; Campbell, C.; Davey, J.; Zawodzinski, T. A., Jr. Electroanalysis 1994, 6, 17-21. (18) Maidan, R.; Heller, A. J. Am. Chem. Soc. 1991, 113, 9003-9004. (19) Cass, A. E. G.; Davis, G.; Francis, G. D.; O’Hill, H. A.; Aston, W. J.; Higgins, I. J.; Plotkin, E. V.; Scott, L. D. L.; Turner, A. P. F. Anal. Chem. 1984, 56, 667-671. (20) Katakis, I.; Heller, A. Anal. Chem. 1992, 64, 1008-1013. (21) Ohara, T. J.; Rajigopalan, R.; Heller, A. Anal. Chem. 1993, 65, 3512-3517. (22) Karyakin, A. A.; Gitelmacher, O. V.; Karyakina, E. E. Anal. Lett. 1994, 27 (15), 2861-2869. (23) Cso¨regi, E.; Gorton, L.; Marko-Varga, G. Electroanalysis 1994, 6, 925933. (24) Johnson, K. W.; Bryan-Poole, N.; Mastrototaro, J. J. Electroanalysis 1994, 6, 321-326.

appealing as it may seem, can only be used in the dialysis technique, the main disadvantages of which are the low time resolution (from 2 to 10 min) and the relatively large size of the probes (minimum of about 200 µm).25 It should be mentioned that most of the constructed glucose biosensors are based on platinum electrodes5,6,8,9,11-14,16,17 or platinized carbonaceous materials26,27 and are generally used with constant-potential amperometry. However, most in vivo experiments now use carbon epoxy electrodes, carbon fiber electrodes, or carbon paste electrodes, together with voltammetric pulsed techniques for current detection.25 The main advantage of these methods is that the potential window and the pulse amplitude, duration, and frequency can be adjusted, so that better resolution of the target substrate current is obtained. In the present paper, we report on the detection of glucose in the rat brain by use of differential normal pulse voltammetry (DNPV) and enzyme-modified carbon fiber microelectrodes. The specificity as well as the limited size of the sensor prepared now offer the possibility to investigate the glucose concentrations directly in limited brain areas of the living animal. EXPERIMENTAL SECTION Chemicals and Solutions. The enzyme glucose oxidase (GOx), extracted from Aspergillus niger (EC 1.1.3.4, type X-S, activity 174 units mg-1), was obtained from Sigma. R-D(+)Glucose, L-ascorbic acid, cellulose acetate, and glucagon, extracted from a mixture of bovine and porcine pancreas and PBS (phosphatebuffered saline tablets), were obtained from Sigma. Glutaraldehyde (a 25% aqueous solution) was obtained from Merck, human insulin (Actrapid, 40 units mg-1) from Novo Nordisk Pharmaceutique S.A., and peptides (posthypophyse extracts) from Choay Laboratory. Resorcinol, m-phenylenediamine, and Nafion, obtained from Aldrich, were used as packaged. All other reagents used were of analytical purity grade. Bidistilled water was used to prepare the solutions. For in vitro experiments, a stock solution of glucose (1 M) was left to mutarotate overnight and then kept at +4 °C. PBS solution containing AA or acetaminophen (0.2 mM) was prepared just before use. When the interfering effects of peptides were studied, posthypophyse extracts were added to PBS in a 1:9 (v/v) ratio. For in vivo experiments, iso- or hypertonic (5% or 30%) glucose solutions, available from a pharmacy, were used. Glucagon was prepared as a 0.25 mg mL-1 solution in normal saline. Working Electrodes. These electrodes were prepared as previously described.28 Briefly, each of them was composed of a single carbon fiber (35 µm diameter, type AVCO, Lowell, MA), placed in a glass micropipet and emerging from its drawn tip. At the emergence point, the carbon fiber was sealed with epoxy resin and cut down at 500 µm. At the back, a contact was provided using a silver wire. Prior to any deposition of the enzymecontaining film, the electrodes were electrochemically pretreated in PBS by successive cycling (+2.4 V for 10 s; +1.8 V for 15 s; +1.3 V for 3 s). The potential applied consisted of triangular pulses delivered at 70 Hz. It has been previously shown that such (25) O’Neil, R. Analyst 1994, 119, 767-779. (26) Geise, R. J.; Yacynych, A. M. In Chemical Sensors and Microinstrumentation; Eddman, P. G., Wang, J., Eds.; ACS Symposium Series 403; ACS: Washington, DC, 1989; pp 65-77. (27) Wang, J.; Li, R.; Lin, M. Electroanalysis 1989, 1, 151-154. (28) Cespuglio, R.; Burlet, S.; Fabre, B.; Bidan, G. French Patent 94-1020-25-0894, 1994.

Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

4359

a pretreatment enhances the detection sensitivity and allows a reproducible smooth carbon surface to be obtained with a welldefined half-wave potential for hydrogen peroxide oxidation.29 To immobilize GOx onto the surface of the microelectrodes, two experimental procedures were employed. The first involved depositing a thick cross-linked enzymatic membrane. For this purpose, the microelectrodes were dipped into a 30% (v/v) solution of GOx in PBS, containing 10% glycerol. They were then dried for 10 min, exposed for 90 min to saturated vapors of glutaraldehyde, and dried again for 15 min. When used with an additional membrane, the microelectrodes were dip-coated with Nafion or cellulose acetate (the latter prepared as a 2%, 5%, or 10% solution in a 1:1 v/v mixture of acetone and cyclohexanone) and dried for at least 10 min before use. The second procedure involved trapping GOx in electropolymerized m-phenylenediamine or resorcinol films. The deposition procedure was similar to that described by Malitesta et al.16 and Geise et al.26 Briefly, a monomer deoxygenated solution (5 mM) was prepared in phosphate buffer (KH2PO4-NaOH, 20 mM, pH ) 7.4). One milliliter of this solution was added to a cell containing 1000 units of GOx and stirred for 1 min. The polymer film was then electrochemically deposited using the conventional three-electrode system composed of a platinum wire as the auxiliary electrode, and a potassium chloride saturated calomel electrode (SCE) as the reference electrode; a carbon fiber microelectrode as the working electrode. The potential applied to the working electrode was cycled continuously between 0 and +0.8 V at 50 mV s-1 for about 10 min (20 scans). Afterward, the electrodes were washed in PBS. All the prepared enzymemodified electrodes were stored at +4 °C when not used. Equipment and Voltammetric Parameters Used. The electrochemical deposition described above was carried out using a PRT 20-2X potentiostat (Tacussel, France) together with a pilot voltage generator (type GSTP, Tacussel). The level of current running in the working electrode circuit was controlled through a voltage drop by means of a resistor (R ) 100 kΩ, graph recorder Sefram, type TRP). The difference in the potential imposed between the reference and working electrodes was fed to the input “X” of the recorder. Electrochemical pretreatment of the microelectrodes and DNPV measurements was performed with a pulsed voltammetric unit (Biopulse, Solea, France) and a chart recorder. The DNPV method is now well-known and widely used.30 It involves applying successive double pulses (prepulse and pulse). The prepulse is increasing in amplitude. The current measured is the differential of the current existing between the ends of the pulse and the prepulse. The signal obtained is defined as an oxidation peak (Figure 1). For in vitro and in vivo experiments, a silver chlorinated wire (Ag/AgCl) was used as the reference electrode. The electrode system was successively immersed by means of microlifts in plastic cells containing glucose solutions (2 mL at different concentrations). Interfering chemicals were added according to the experimental protocol. The glucose concentrations were determined in plasma and CSF within aliquots of 150 µL. The height of the DNPV peak was measured from its top to the baseline, which was determined by the line crossing the two minima of the voltammetric signal (Figure 1b). (29) Netchiporouk, L. I.; Shul’ga, A. A.; Jaffrezic-Renault, N.; Martelet, C.; Olier, R.; Cespuglio, R. Anal. Chim. Acta 1995, 303, 275-283. (30) Suaud-Chagny, M. F.; Cespuglio, R.; Rivot, J. P.; Buda, M.; Gonon, F. J. Neurosci. Methods 1993, 48, 241-250.

4360 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

A

B

Figure 1. (A) Principles of differential normal pulse voltammetry (DNPV). The current is sampled just at the end of the prepulse (Pr in A) and the pulse (P in B). The difference (differential) of these two elements is recorded. The advantage of DNPV versus other methods is that, between each double pulse (Pr + P), the electrode is held at the initial potential (Pi). In this way, the background noise is decreased. Abbreviations: T, period of each double pulse, about 400 ms; t1, t2, prepulse and pulse durations; ∆v, pulse amplitude and prepulse increment. (B) Differential normal pulse voltammograms recorded in solutions with different glucose contents (0-5 mM). The working GOx-modified electrode was covered with an additional membrane of Nafion. Abbreviation: h, peak height measured.

To verify the validity of DNPV measurements, we also checked plasma and CSF glucose contents by using two commercially available glucose analyzers: Beckman glucose analyzer 2 (electrochemical principle, glucose oxidase-H2O2) and Ciba Corning CL 7200 glucose analyser (colorimetric principle, glucose oxidase + chromogen). Surgery. Male Sprague-Dawley rats weighing 280-360 g were implanted under chloral hydrate anesthesia (400 mg kg-1, via the intraperitoneal route, i.p., 10% solution in normal saline) with the same electrode assembly used in vitro. A working microelectrode was inserted into the first 500 µm of the right cortex (3 mm above λ and 2 mm lateral31) through a window opened in the bone and the dura. During this procedure, special care was taken to avoid blood vessel damage. The platinum and Ag/AgCl wires were placed in the posterior cortex in contact with the dura. Afterward, all the electrodes were connected to the instrumental setup and recordings immediately started. (31) Paxinos, G.; Watson, P. The Rat Brain in stereotaxic coordinates, 2nd ed.; Academic Press: Tokyo, 1986.

Insulin, glucose, glucagon, normal saline, and additional doses of anesthetics were administered via the i.p. route. For glucose determination, blood and cerebrospinal fluid (CSF) were sampled from anesthetized rats following standard procedures. The experimental protocol as well as the surgical procedure were in compliance with the rules of the French Agriculture Ministry (Decree No. 87.848). RESULTS AND DISCUSSION Glucose Voltammetric Measurements. DNPV parameters usually employed for neurotransmitter measurements30,32 are not adequate for the voltammetric measurement of enzymatically produced hydrogen peroxide. This might be due to a difference in the conditions of diffusion of the oxidized species concerned. A DNPV peak appearance is the result of a competition between the rise in the amount of oxidized species due to an increase in potential and the depletion of diffusion layer in these species. In the case of neurotransmitters, the diffusion is directed toward the electrode surface, while for hydrogen peroxide, produced in the vicinity of the enzymatic layer, it has an opposite direction. Therefore, to obtain a well-resolved DNPV peak, it was necessary to lower the magnitude of the measurement pulse and/or increase the duration of the measurement pulse and prepulse (as compared with 40-50 mV and 50-60 ms values respectively used for neurotransmitters detection; Figure 1b). Experimentally, the following suitable DNPV parameters have been established: 1030 mV for pulse magnitude and 60-100 ms for pulse and prepulse duration. When using differently covered microelectrodes, several combinations of these parameters were possible to obtain a wellresolved H2O2 oxidation peak. Furthermore, in order to facilitate a comparison between different electrodes, a magnitude of 20 mV and a duration of 80 ms for pulse and prepulse were used in our experiments. An initial scan potential of +400 mV versus Ag/ AgCl reference was chosen in order to be above the oxidation peaks of most of the electrooxidized interferents (neurotransmitters, acetaminophen, ascorbic acid). In vivo and in vitro, in order to obtain a well-resolved peak, the DNPV scan was extended up to +1350 mV versus Ag/AgCl reference. We did not observe any influence of the scan rate on the peak shape for rates up to 20 mV s-1. Finally, to prevent the depletion of hydrogen peroxide produced around the working electrode, the interval existing between successive pulses was determined to be at least 0.3 s. After the above check of the DNPV parameters, our measurements were currently performed every 2 min at a scan rate of 10 V s-1 (a step of 4 mV per interval of 0.4 s). In these conditions, the CFME modified with a cross-linked GOx membrane allowed a stable voltammetric response to be obtained in glucose solutions. Usually, the sensitivity of our DNPV measurements was about 34.5 mA M-1 cm-2 and linear up to 1.5 M. To improve this limit of detection, most probably related to the “oxygen deficit” (see reaction 2), we applied additional membranes in order to decrease the glucose:oxygen ratio near the microelectrode surface. Cellulose acetate (CA) and Nafion were used for this purpose. CA membranes were deposited from 10%, 5%, and 2% CA solutions. However, a considerable loss of sensitivity (up to 15 times) was observed for the biosensors coated with additional CA (5% and 10%) membranes; they were thus discarded for our experimental protocol. Adequate results were obtained when CA (2%) and (32) Cespuglio, R.; Faradji, H.; Riou, F.; Buda, M.; Gonon, F.; Pujol, J. F.; Jouvet, M. Brain Res. 1981, 223, 287-311.

A

B

Figure 2. (A) Dependence of the DNPV peak height on glucose concentration for microelectrodes modified with (1) cross-linked GOx layer covered with CA (a) or Nafion (b); (2) electropolymerized GOx/ PPD (c) or GOx/polyresorcinol (d) films. (B) Initial region of the calibration curve for a sensor covered with CA membrane. This type of sensor was mostly used for in vivo measurements.

Nafion membranes were used (Figure 2A, curves a, b; Table 1). Reproducibility of the response of biosensors prepared according to the same procedure was within 20%, in spite of the dip-coating method used. The effect of oxygen on the response of the electrodes covered with CA additional membranes was studied while bubbling the calibration glucose solutions with O2. An extension of the initial dynamic range of the sensor response up to glucose concentration of 15-20 mM was obtained, while the initial slope of the calibration curve remained unchanged. Taking into account that pH and ionic strength of the physiological media are quasiconstant, and since their influence on similar glucose sensors responses was extensively investigated earlier,33-35 such a dependence was not studied here. To further improve the performance of our biosensor, we trapped GOx in m-phenylenediamine (PPD) and resorcinol films (33) Shul’ga, A. A.; Sandrovsky, A. K.; Strikha V. I.; Soldatkin, A. P.; Starodub, N. F.; El’skaya, A. V. Sens. Actuators B 1992, 10, 41-46. (34) Shul’ga, A. A.; Strikha V. I.; Soldatkin, A. P.; El’skaya, A. V.; Maupas, H.; Martelet, C.; Clechet, P. Anal. Chim. Acta 1993, 278, 233-236. (35) Soldatkin, A. P.; El’skaya, A. V.; Shul’ga, A. A.; Netchiporouk, L. I.; Nyamsi Hendji A. M.; Jaffrezic-Renault, N.; Martelet, C. Anal. Chim. Acta 1993, 283, 695-701.

Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

4361

Table 1. Biosensor Characteristics in the DNPV Measurement Mode

dynamic range, mM correlation coefficient sensivity, mA M-1 cm-2 selectivity to peptidesa life spanb no. of electrodes tested

GOx

GOx +CA (2%)

GOx +Nafion

GOx/PPD

GOx/ polyresorcinol

0.5-1.5 0.99 34.5 ( 2.8 0 >5 days 15

0.5-3 0.99 8.2 ( 0.6 0.93 ( 0.03 >7 days >50

0.3-6.5 0.99 2.7 ( 0.2 0.78 ( 0.04 >5 days 23

0.5-5 0.99 2.2 ( 0.1 0.71 ( 0.02 12-16 h 34

0.5-3.5 0.98 1.4 ( 0.1 0.50 ( 0.02 10-12 h 27

a Ratio of the DNPV peak produced by the enzymatic oxidation of glucose in the “clean” solution or in the presence of peptides. b Determined before a 20% loss of sensitivity, observed when using the sensor for at least 6 h/day.

which were electropolymerized at the CFME surface. A disadvantage noticed for these biosensors was that the DNPV peak for glucose did not appear just after the film preparation. It was obtained only after 2 days of storage in PBS or after DNPV potential sweeping (from one to a few hours for the GOx/ polyresorcinol and GOx/PPD films, respectively). However, on account of a comparatively high sensitivity (assuming that GOx was immobilized in a thin insulating layer) and an initially extended dynamic range of the sensors (see Figure 2 and Table 1), they were retained in order to test their selectivity toward interfering substances. Influence of Interfering Substances on the Glucose Voltammetric Signal. This aspect is illustrated in Figure 3A, where DNPV peaks are recorded with a GOx-modified microelectrode either in a pure 2 mM solution of glucose or in the same solution supplemented with 0.2 mM ascorbic acid. The presence of an interferent like ascorbic acid contributed to a DNPV baseline shift, while both the shape and the height of the signals remained unchanged. The same observations were obtained with acetaminophen at physiological concentration (0.2 mM). This insensitivity for the interferents having an oxidation potential much more negative than that of H2O2 can be attributed to the mode of measurement employed. The measured signal is the first derivative of the polarographic curve, assuming that the DNPV parameters chosen are correct for the extraction of capacitive currents. Meanwhile, particular attention has to be paid in order to discriminate between species which are oxidizable close to the hydrogen peroxide potential. In the case of in vivo brain measurements, those are peptides present in cerebral extracellular fluid in large concentrations. To evaluate their influence on the response of the microbiosensors to glucose (2 mM for the GOxmodified electrodes with and without additional membranes; 5 mM for GOx/PPD and GOx/polyresorcinol electrodes), measurements were performed either in a pure solution of glucose or added with posthypophyseal extract. The peptides contained in this extract exhibit an oxidation potential at about +700 mV versus Ag/AgCl reference and also produce a noticeable increase in the DNPV baseline. An additional Nafion membrane considerably restricted the contribution of the peptides to the baseline, although the response due to glucose was decreased (see Table 1). Only the electrodes coated with enzymatic membranes were adequate to overcome such an interference. In such a situation, the addition of a CA membrane left the signals related either to peptides or to hydrogen peroxide well separated (Figure 3B). Glucose DNPV Measurements in Vivo versus Standard Methods of Analysis. Table 1 summarizes the data related to the biosensors operating in the DNPV measurement mode. Taking into account the fact that the glucose concentration 4362 Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

A

B

Figure 3. Appearance of the DNPV peaks corresponding to a 2 mM concentration of glucose in PBS as compared with those recorded in the presence of ascorbic acid (A) or peptides (B) (case of GOx/ PPDfilms).

reported in cerebral extracellular fluid was between 0.47 ( 0.18 and 2.4 ( 0.1 mM (refs 7 and 8, respectively), it appeared that the sensitivity and dynamic range of the biosensors excluding those of the first type (with GOx membrane only) were sufficient to monitor the level of glucose in brain. However, the electropolymerized membranes lost their selectivity and/or sensitivity properties during the DNPV measurements. Under microscope

Figure 4. Level of glucose concentration and changes therein recorded in the living brain by the DNPV method while insulin and glucose (A) or glucagon (B) were successively injected. The working GOx-modified microelectrodes were covered with an additional membrane of (A) cellulose acetate (fitting equation of the calibration curve y [mM] ) 0.17x [nA] - 0.5) and (B) Nafion (y [mM] ) 0.51x [nA] - 0.3).

observation (Nachet Instruments, magnification ×400), a gradual alteration of the polymer membrane was noticed. As yet, in our hands no significant improvement for membrane stability could be obtained. Therefore, only sensors coated with additional membranes of CA and Nafion were used for in vivo measurements. When inserted into the cortex, the sensors needed a delay ranging from 20 to 40 min before stable and reproducible DNPV peaks could be obtained. The basal glucose level was monitored during at least 2 h before injection of insulin or glucagon. Figure 4A shows the typical changes occurring in the DNPV peak height after administration of insulin. The average time necessary for a 50% decrease in the peak height is about 30 min (n ) 10). Then, by administrating successive boluses of glucose, the effect of the drug could be reversed. Since the animals received a high dose of insulin, injections of isotonic solution of glucose yielded only a short duration increase in the DNPV signal. An injection of 2 mL of hypertonic solution of glucose allowed the base line recovery for at least 1 h. The administration of glucagon caused first an increase in the signal height, corresponding to a change in glucose concentration of about 0.5 mM. The average time of the signal rise was about 40 min (n ) 6). Then, a second injection of glucagon at the same dose provoked a rapid decrease in the signal height, followed by a slow recovery (Figure 4B). This fall in the glucose level after an increased dose of glucagon is ascribed to a release of insulin triggered by glucagon.36 In control experiments, after administration of normal saline only (NaCl 0.9%), the DNPV signal was continuously monitored in the cortex during 5 h at least. This compound did not change the baseline level nor the peak heights. Only a slight and transient alteration in the DNPV peak was observed during the 5-10 min (36) Jacot, E.; Assal, J. Ph. Re´ gulateurs de la glyce´ mie; Pharmacologie Edition; Frison-Roche: Paris, Slatkine, Gene`ve, 1989; Vol. 33, pp 481-494.

Figure 5. Bar graphs presenting values of glucose concentration in blood plasma (A-C) and CSF (cerebrospinal fluid) (A-C) determined by photocolorimetric method, the standard electrochemical glucose-analyzer and the microelectrodes operating in the DNPV measurement mode. Values Cplasma and CCSF are calculated on the basis of measurements carried out with diluted samples (1:2). The average values with standard errors (n ) 6) are presented.

period immediately following the anesthetic supply. This aspect is totally negligible. At this experimental step, the specificity of the DNPV signal toward glucose is thus clearly established. The electrodes were calibrated before and after implantation. A sensitivity loss (up to 20%) occurred after continuous use in vivo for at least 8 h. This aspect was taken into account when the baseline glucose level in the extracellular compartment of the brain was estimated. According to our measurements, this baseline level appears to be at a concentration of 1.5 ( 0.3 mM (n ) 21). Finally, experiments were also conducted to establish a correlation between glucose concentrations measured with either the DNPV technique or other standard procedures used in laboratories or in clinical practice. For this purpose, plasma and CSF samples were analyzed using our microbiosensors and two commercial glucose analyzers (Beckman and Ciba Corning). Taking into account the presumed content of glucose in the physiological liquids, plasma samples were diluted three times (1:2 v/v) with PBS and analyzed using only the microelectrodes coated with an additional Nafion membrane. The same dilution was used for CSF, and the measurements were conducted using the microelectrodes coated with an additional membrane of CA. The results obtained are summarized in Figure 5. Glucose concentrations were established throughout six assays and given as a mean ( SEM. It should be noticed that the glucose concentrations determined using the microelectrodes were, for each sample, in good agreement with those established by the standard procedures. CONCLUSIONS Differential normal pulse voltammetry, performed with the use of modified carbon fiber microelectrodes, is a reliable method for direct and specific glucose monitoring in the living brain. CFMEs possess good biocompatibility, are mini-invasive, and meet well the requirements imposed by the use of a pulsed technique. DNPV offers a better selectivity toward enzymatically produced hydrogen peroxide, as compared with conventional amperometry. Interfering species having an oxidation potential close to that of Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

4363

H2O2 (peptides) can be eliminated effectively. There is also good agreement between the glucose concentrations determined in vitro either by standard procedures or by the DNPV technique. Approaches related to the study of brain metabolism, with unanesthetized and freely moving animals, can now be seriously envisaged. ACKNOWLEDGMENT We thank Prof. J. Pichot and Dr. A. Geloen for their kind help in glucose determination by standard procedures. This work has

4364

Analytical Chemistry, Vol. 68, No. 24, December 15, 1996

been supported by French Foreign Ministry, CNRS, and INSERM U52. The framework of Franco-Ukranian scientific collaboration is also gratefully acknowledged. Finally, we thank C. Limoge for improving the English text. Received for review September 23, 1996.X

February

27,

1996.

Accepted

AC960190P X

Abstract published in Advance ACS Abstracts, November 1, 1996.