Anal. Chem. 1981, 53, 51-53 51

with different slopes of the conductance curves. The thermal forces depend on the dielectric constant of the medium and the solvation depends on the attractive forces between solute ions and between these ions and the solvent molecules.

Also the phenol and its derivatives can be titrated with DPG, obtaining one or more end points depending on the substituent on the benzene ring.

In order to further our knowledge on this field of conduc- tometric titrations, we studied citric acid to determine if multiple end points could be obtained. The titration of this triprotic acid gave three end points corresponding to 1:1, 1:2, and 1:3 acid:base ratio. The complete linear conductance behavior beyond a theoretical 1:4 acidbase stoichiometry confirms the presence of only three equivalence points.

This result is very interesting because it gives some indi- cations regarding the possibility to determine quantitatively acids and phenols in binary or ternary mixtures in the cases in which such determinations should be impossible in aqueous solutions owing to the very close pK values. This is really true because the dissociation constants of citric acid in aqueous solution differ by about one unit of magnitude from each other (pKI = 3.14, pK2 = 4.77, and pK3 = 6.39 at 18 OC) (6).

From the above discussed titrations it is possible to de- termine the percentage recovery of the acid. These per-

centages are listed in Tables 1-111 in terms of millimoles of acid taken and recovered. The recoveries are reasonably good, with an error of approximately &(0.5-1.0)%.

ACKNOWLEDGMENT We thank L. Tassi for his experimental work and the

“Centro di Calcolo Elettronico” of the University of Modena for the computing support.

LITERATURE CITED (1) Marple, L. W.; Scheppers, G. J. Anal. Chem. 1968, 38, 553-558. (2) Chentml. M. K., Jr.; Kdthoff. I. M. J. Phys. Chem. 1978, 82,

9941000, and references therein. (3) Kdthoff, I. M.; Chantmi, M. K., Jr. Anel. Chem. 1978, 50,

1440-1446, and references therein. (4) Kratochvil, B. Anal. Chem. 1978, 50, 153R-161R. (5) “Handbook of Chemisby and physics”, 56th ed.; Weest, R. C., Ed.;

The Chemical Rubber CO.: Cleveland, OH, 1975-1976; p C-317. (6) Reference 5, p D-150. (7) Anderson, M. L.; Hammer. R. N. Anal. Chem. 1988, 40, 940-944. (8) Van Meurs, N.; Dahmen, E. A. M. F. Anal. Chlm. Acta 1958, 79,

64-73. (9) Lippmaa, E. T. J. Anal. Chem. USSR (Engl. Trans/.) 1955. 70,

157-182. (10) Bryant, P. J.; Wardrop, A. W. H. J. Chem. Soc. 1957, 895-906.

RECEIVED for review March 3,1980. Accepted September 11, 1980. This work has been supported by the National Research Council (C.N.R.) of Italy.

Enzyme Electrode for the Determination of Glucose

Esther Lobel”‘ and Judith Rishpon

Research Products Rehovot, Kiryat Weizmnn, P.O. Box 138, Rehovot, Israel

A new design of a glucose electrode based on the ampero- metric determlnatlon of hydrogen peroxide was investigated. An enzyme membrane is attached to a platinum net anode, whlch permits free diffusion of atmospherlc oxygen Into the membrane. As a result, oxygen is in a nonrate limiting con- centration near the enzyme, and higher concentrations of glucose can be directly determined. Variation of the oxygen tension of the test solution does not affect the measurements. The contact between the enzyme membrane and the anode has been improved by direct coating of the membrane with a thin layer of gold. The use of a negatively charged dialysis membrane In front of the enzyme membrane decreases the interference of species found in human serum. The interfer- ence of ascorbic acid, urk acid, MtiruMn, and giutathbne was investigated. The electrode was tested under flow conditions and lifetime of the gold-coated membrane was limited to 1-3 days.

The principle of measuring glucose by the electrochemical determination of hydrogen peroxide formed in the presence of glucose oxidase is the subject of a number of publications (1-3) and the basis of two commercial glucose analyzers (2, 4).

One is the Yellow Springs glucose analyzer (4 ) in which small samples of 25 pL are diluted by 1:14 prior to mea- surement; the second is the Biostator (2) in which blood is continuously withdrawn from the patient and then further

‘Present address: ville, IL 61832.

Teepak, Inc., 915 N. Michigan Avenue, Dan-

diluted and pumped to the glucose sensor. Dilution of the sample is mentioned in an additional electrode system based on an “enzyme spacer” (5). One possible reason for the need of sample dilution is the presence of limiting concentrations of oxygen in the test solution which affects the second-order reaction of glucose oxidase, especially when high glucose levels of diabetics are measured. It can be improved by an electrode design in which atmospheric oxygen is allowed to diffuse directly to the immobilized enzyme layer through a porous anode. The demonstration of this idea is the subject of the present paper.

The influence of blood components on the performance of the glucose electrode was examined. The anode was coated directly on top of the membrane and a charged dialysis membrane was mounted in front of it in order to improve the low current efficiency in blood.

EXPERIMENTAL SECTION Apparatus. The glucose measuring system consista of three

electrodes: the glucose electrode, a counterelectrode, and a reference electrode connected to a potentiostat (Elscint CHP-2). The glucose working electrode consists of a fine platinum net anode pressed lightly to the glucose oxidase membrane. The platinum net is of a commercial type. It is made of densely woven wires that are about 0.2 mm thick. It was cleaned in ethanol and nitric acid. The net was placed behind the enzyme membrane, and both were mounted in an Orion electrode body or in a self-manufactured electrode body. Electrical contact was made between the anode and the potentiostat. Alternatively a gold- coated enzyme membrane was connected to the potentiostat through the porous gold coating.

The platinum or gold anode is exposed to atmospheric oxygen through an opening that was made in the electrode shaft. Some of the enzyme membranes were separated from solution by a

0003-2700/81/0353-0051$01.00/0 0 1980 American Chemlcal Society

52 ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981

dialysis membrane. Enzyme membranes that were prepared from a skinned reverse osmosis type membrane were in direct contact with the measuring solution. The glucose electrode surface im- mersed in the sample was about 0.1 cm2.

The counterelectrode is a platinum plate of 1 cm2 area. The working electrode was biased by +750 mV relative to a saturated calomel reference electrode when the anode was platinum. In the presence of a gold anode, it was biased by +LO V.

For electrode evaluation under flow conditions, a flow system was assembled by using a peristaltic pump that was connected to a transparent plastic tube with a diameter of 0.6 cm and wall thickness of 0.2 cm. Into this tube, holes were cut for the three electrodes. The flow rate was 30 mL/min.

Procedure. Experiments were performed at room temperature (25-27 "C). Test solutions consisted of @-D-glUCOSe, phosphate buffer (pH 7.2), and 0.15 M NaC1. In some experiments, standard human serum (Dade) or human blood received from a clinical laboratory and containing 0.75% sodium citrate was used. So- lutions were stirred. The glucose concentration of the test so- lutions was varied by adding small increments of a saturated glucose solution. The concentration of glucose in blood was determined by Dextrostix test strips and the Eyetone reflectance colorimeter (Ames) prior to electrode measurements. The pre- cision of this system, according to the manufacturer, is f5 mg/dL in the range of 10-250 mg/dL blood glucose and A10 mg/dL in the range of 250-400 mg/dL. The precision of the glucose electrode was measured by repeating the electrode calibration between 0 and 12 mM glucose at least five times for the platinum and the gold anode electrodes. The standard deviation from mean at the various concentrations did not exceed 6%.

Membrane Preparation. GOD Membranes. Three types of GOD membranes were prepared a Millipore f i k r covered by a dialysis membrane (Spectrapor, Cole Palmer), RO skinned membranes, and a Nuclepore membrane (Nuclepore Corp.). The Millipore filter, Type VS, pore size of 0.025 pm was loaded with glucose oxidase (Sigma, Type 11) by pressing a solution of 50 mg/mL enzyme dissolved in phosphate buffer, pH 6.8,0.02 M, and containing 1 % glutaraldehyde, through the membrane in a pressure cell at 1 atm. The enzyme was allowed to cross-link in the membrane overnight at 4 "C. Enzyme activity was assayed by a titrimetric procedure as described in the Sigma product bulletin. Membranes of a known area of about 10 cm2 were cut into smaller units and placed in a glucose-containing glacial acetic acid buffer of pH 5.1 at 35 "C. The solution was aerated for 15 min with a sintered glass sparger. Aeration caused also some mixing. Excess 0.1 N sodium hydroxide was added to stop the reaction. The sodium hydroxide was back-titrated with 0.05 N hydrochloric acid by using phenolphthalein indicator.

For the gold-coated membranes, the enzyme was allowed to penetrate the uncoated side.

The above procedure was also used to prepare glucose oxidase membranes from RO-930 reverse osmosis membranes (DDS). The porous layer of the membrane pointed upward in the pressure cell. Nuclepore filters (Nuclepore Corp.) of 0.1 pm pore size were immersed in the GOD solution for 6 h in order to allow the diffusion of the enzyme into the filter prior to cross-linking.

Gold-Coated Membranes. Thin gold coatings were produced by two methods: heat evaporation of gold under vacuum and spattering by a spattering device for the scanning electron mi- croscope. The coatings were performed on the Millipore filters, and their thickness, by the first method (estimated from the weight of the gold used), was a few hundred angstroms. For the second method, the thickness was not determined, but the time of coating could be controlled (2.5 min).

Charged Membranes. Dialysis membranes were negatively charged by means of a triazinyl dye (Brilliant Orange, ICI). Two types of dialysis membranes were used: very dense ones with a molecular weight cutoff of lo00 and more open ones of 6oo(t8OOO. The charging procedure was as follows: the dialysis membrane was agitated in an aqueous solution of 10% dye and 3% sodium carbonate at pH 10.5 for 2 h at room temperature and then thoroughly rinsed.

RESULTS Linear calibration curves up to 15 mM glucose were ob-

tained when oxygen was allowed to diffuse into the system

" h , . 3 h

I I t I " I t < l I I I r t 0

0 2 4 6 8 10 12 14 16 18 20 GLUCOSE CONCENTRATION, mM

Figure 1. Electrode current as a function of glucose concentration: (curve 1) anode open to oxygen: (curve 2) anode closed to oxygen: (curve 3) anode closed and nitrogen bubbling in test solution.

r

0 4 S 12 16 20 GLUCOSE CONCENTRATION, mM

Flgure 2. Effect of enzyme membrane on electrode slope: (curve 1) RO 930 membrane: (cuve 2) Nudepore membrane; (curve 3) Mtllipore membrane. The test solutions contain glucose in saline.

and only up to 5 mM when the oxygen diffusion was prevented by covering the anode with polyethylene film. If, in addition, nitrogen was bubbled into the test solution, there was no response of the electrode to changes in glucose concentration (Figure 1). The electrode system was a Millipore filter with immobilized GOD, covered by a dialysis membrane. The platinum anode was positioned between the enzyme mem- brane and the polyethylene film, contacting both.

Another group of experiments which emphasize the im- portance of free diffusion of oxygen to the glucose oxidase showed that, for the same membrane type, there is almost no difference in the slope response whether the test solution is saturated with nitrogen, air, or a 10% oxygen-nitrogen mix- ture. "he current increased linearly over a concentration range of 0-17 mM glucose and the slopes ranged over 0.49-0.52 pA/mM.

An RO-930 type membrane with free access of oxygen to the glucose oxidase gave also satisfactory results (Figure 2). Nuclepore on the other hand did not produce a linear slope response. A Millipore membrane without Wig combined with a dialysis membrane gave a linear response only up to 7 mM glucose.

Slope response curves for the Milliporedialysis and the RO type electrode were measured as a function of time. Initial slopes were about 0.25 pA/mM in buffer and 0.1 pA/mM in blood or serum.

Among 15 glucose electrodes of RO type tested, the period of proper functioning was between 5 and 15 days out of which the electrodes were stored and tested 30 h in whole blood or serum. For five Millipore type electrodes, the lifetime varied

ANALYTICAL CHEMISTRY, VOL. 53, NO. 1, JANUARY 1981 53

measurement without further dilution of sample and may lead to an implantable electrode. Glucose can be determined over a range of 15 mM. Two of the membranes described meet the requirements for linear electrode performance. These requirements were analyzed in an article by Racine and Mindt for a lactate electrode (6).

One of the requirements to achieve linearity of the current as a function of external substrate concentration is to ensure that the internal substrate concentration near the enzyme is very smaU as compared with the K, of the enzyme. Practially, this can be achieved by a membrane with a low permeability to glucose. Therefore, it was necessary to add a dialysis membrane in front of a Millipore filter or to use a skinned reverse osmosis membrane with the skin pointing toward the test solution. The Nuclepore filter, which is very thin (20 pm) is too highly permeable to glucose, hence the nonlinear slope.

The decrease in slope response in serum may be caused by a low current efficiency due to the presence of reducing substances in blood which may cause side reactions with the hydrogen peroxide. An alternative possibility for the low current efficiency being due to an inappropriate contact be- tween platinum and membrane was not validated. The gold coating alone did not solve the problem of low current effi- ciency.

A combination of four substances found in blood produced a similar or somewhat larger reduction in electrode current than blood or serum. The lower response to glucose in the presence of blood could also be related to modification of diffusional properties of the membranes. Addition of a dense and charged dialysis membrane eliminated part of the problem possibly by a rejection mechanism.

A possible reason for the limited lifetime of the gold-coated electrode is the peeling off of the coating, especially when the membrane swells in solution. The coating technique must be improved in order to obtain electrodes with a longer lifetime.

ACKNOWLEDGMENT The authors thank 0. Kedem, Research Products Rehovot,

for many helpful suggestions.

LITERATURE CITED (1) Guiibautt, G. 0.; Lubrano, 0. J. Anal. Chim. Acta 1072, 60, 254. (2) Clarke, W. L.; Santiago, J. V. ArtM. Organs, 1077, 1 , No. 2. (3) Thevenot, D. R.; Stemberg. R.; Coulet. P. R.; Laurent, J.; @ W o n ,

(4) Chua, K. S.; Tan, I. K. Clin. Chem. ( Wnston-Salem, N .C . ) 1078, 24,

(5) Martiny, S. C.; Jensen, 0. J. "Ion and Enzyme Electrodes in Blolosy and Medicine"; Kessler, M., Clark, L. C., Jr., Lubbers, D. W., Silver, I. A., Simon, W., Ed.; Unlverslty Park Press: Baltimore, MD, 1078; pp 198-199,

D. C. Anel. Chem. 1970, 51. 96-100.

150-152.

(6) Racine, P.; Mindt, W. Experientla Suppl. 1971, No. 16, 525.

Table I. The Ratio between Slope Response in Serum To That in Buffer for Glucose Electrodes with Platinum Net and with Gold Coating

no. membranea serum/buffer slope ratio

1 Millipore' + Dialysis' 0.42 2 Millipore2 + Dialysis' 0.40 3 Millipore* + Dialysis2 0.37 4 Millipore2 + Dialysis', charged 0.47 5 Millipore2 + Dialysis2, charged 0.75

anode. Millipore2, Millipore GOD membrane, gold coated. Dialysis', molecular cutoff 6000-8000. Dialysis*, molecular cutoff 1000.

a Millipore', Millipore GOD membrane, platinum net

between 3 and 8 days, out of which the maximal exposure time to blood or serum was 3 h. The blood seems to decrease electrode lifetime. The enzyme membrane was replaced when the initial calibration slope was reduced by more than 50%.

A comparison of the ratio of slope response in serum to that in buffer for electrodes with a platinum net anode and elec- trodes with gold coatings is shown in Table I.

The slope ratio did improve when a negatively charged dense dialysis membrane was added to the system.

In the following experiments, four solutes were added to the buffer solution, f i i t separate and then combined. Ascorbic acid added to phosphate buffer at the physiological range of 0.5-1.5 mg/100 mL did not decrease the electrode slope as compared with a solution of glucose in buffer. The same is true for uric acid a t a concentration of 2.0-7.8 mg/100 mL. Glutathione, however, a t a concentration of 21 mg/100 mL decreased the electrode slope of a platinum net electrode to 62% as compared to glucose in buffer. Bilirubin at a con- centration of 1.4 mg/100 mL in the presence of 30 mg/mL albumin (in order to improve its solubility in buffer) also caused a current decrease.

A mixture of 1.5 mg/dL ascorbic acid, 10 mg/dL uric acid, 21 mg/dL glutathione and 1.4 mg/dL bilirubin caused a 70% decrease in the slope response of a Millipore-open dialysis- platinum net electrode as compared to electrode currents in buffer. This result is an average of four experiments.

Measurements with a glucose electrode under flow condi- tions as a function of time resulted in a linear current increase over a concentration range of 0-14 mM glucose. The electrode slope, however, decreased from 0.94 pA/mM glucose on the first day to 0.43 pA/mM on the third day. The electrode response time was less than 1 min, which is needed for the solution to be transported from the container to the electrode surface. The lifetime of a number of additional electrodes tested varied between 1 and 3 days.

DISCUSSION One can conclude that in principle the electrode design

described in this work is advantageous for direct glucose

RECEIVED for review July 21,1980. Accepted October 15,1980. This research was supported by a grant from the National Counsel for Research and Devlopment, Israel, and the GSF Munchen, Germany.