677
Anal. Chem. 1991, 63,677-682
Amperometric Glucose Biosensors Based on Redox Polymer-Mediated Electron Transfer Paul D. Hale,*slLeonid I. Boguslavsky,l Toru Inagaki, Hiroko I. Karan? Hung Sui Lee, and Terje A. Skotheim Department of Applied Science, Division of Materials Science, Brookhaven National Laboratory, Upton, New York 11973
Yoshi Okamoto Department of Chemistry, Polytechnic University, Brooklyn, New York 11201
Electrical communlcatlon between the flavin adenlne dlnucleotide redox centers of glucose oxldase and a conventlonai carbon paste electrode has been achleved by using electron-transfer relay systems based on polyslloxanes. SIX materials for ampermetric blosensors are descrlbed In which ferrocene and dlmethylfertocene electron relays are covalently attached to Insoluble slloxane polymers. Sensors Contalnlng these polymeric relay systems and glucose oxldase respond rapidly to glucose, with steady-state current responses achieved In less than 10 s. The response to glucose under N, saturation shows apparent Mlchaells-Menten constants, KII*W, In the range 16-71 mM and llmltlng current densltles, jmrx,of 29-275 pA/cm2. The dependence of the sensor response on the nature of the siloxane polymer and the type of polymer-bound relay Is discussed.
INTRODUCTION Amperometric glucose electrodes based on glucose oxidase undergo several chemical or electrochemical steps which produce a measurable current that is related to the glucose concentration. In the initial step, glucose converts the oxidized flavin adenine dinucleotide (FAD) center of the enzyme into its reduced form ((FAD)H2). Because these redox centers are essentially electrically insulated within the enzyme molecule, direct electron transfer to the surface of a conventional electrode does not occur to any measurable degree. The most common method (1-4) of indirectly measuring the amount of reduced glucose oxidase, and hence the amount of glucose present, relies on the natural enzymatic reaction glucose
+ O2
glucose oxidase
gluconolactone
+ H20z
where oxygen is the electron acceptor for glucose oxidase. Oxygen is reduced by (FAD)Hz to hydrogen peroxide, which may then diffuse out of the enzyme and be detected electrochemically. The working potential of such a device is quite high ( H z 0 2is oxidized at approximately +0.7 V vs the saturated calomel electrode, or SCE), however, and the sensor is therefore highly sensitive to many common interfering electroactive species, such as uric acid and ascorbic acid; HzOz is also known to have a detrimental effect on glucose oxidase activity. Alternatively, one could use the electrode to measure the change in oxygen concentration that occurs during the above reaction. In both of these measuring schemes, this type Present address: Moltech Corporation, Chemistry Building, State University of New York at Stony Brook, Stony Brook, NY 11794.
*Permanent address: Division of Natural Science and Mathematics, Medgar Evers College, City University of New York, Brooklyn, NY 11225. 0003-2700/91/0363-0677$02,50/0
of sensor has the considerable disadvantage of being extremely sensitive to the ambient concentration of 02. In recent years, systems have been developed that use a nonphysiological redox couple to shuttle electrons between (FAD)H, and the electrode by the following mechanism: glucose
+ GO(FAD)
GO((FAD)HJ 2Mred
-
-
+ GO((FAD)H2) GO(FAD) + 2M,d + 2H+
gluconolactone
-
+ 2MOx 2MOx+ 2e-
(at t h e electrode)
In this scheme, GO(FAD) represents the oxidized form of glucose oxidase and GO((FAD)H,) refers to the reduced form. The mediating species Mox/Mredis assumed to be a oneelectron couple. Sensors based on derivatives of the ferrocene/ferricinium redox couple (5-8), on quinone derivatives (9-11), and on electrodes consisting of organic conducting salts such as TTF-TCNQ (tetrathiafulvalene-tetracyanoquinodimethane) (12-1 7) have recently been reported. In potential clinical applications, however, sensors based on electronshuttling redox couples suffer from an inherent drawback the soluble, or partially soluble, mediating species can diffuse away from the electrode surface into the bulk solution, which would preclude the use of these devices as implantable probes in clinical applications and restrict their use in long-term in situ measurements (e.g., fermentation monitoring). With this in mind, several research groups have been investigating systems where the mediating species is chemically bound in a manner that allows close contact between the FAD/(FAD)H, centers of the enzyme and the mediator yet prevents the latter from diffusing away from the electrode surface. For instance, Degani and Heller (18-20) have designed an electron-transfer relay system where the mediating species (ruthenium pentammine or ferrocene derivatives) are chemically attached to the enzyme itself. In this scheme, however, the chemical modification of the enzyme can cause a measurable decrease in its activity. More recently, these studies have been extended to include systems where the mediating redox moieties are covalently attached to polymers such as polypyrrole (21), poly(viny1pyridine) (20,22-24), and polysiloxane (25-29). These systems serve to “electrically wire” the enzyme (20), facilitating a flow of electrons from the enzyme to the electrode. In the present paper, we report the design and response of amperometric glucose sensors based on ferrocene-containing siloxane polymer “wires”. The general structure of the polysiloxane relay system is shown in Figure 1. These materials are random block copolymers (i.e., the polymeric subunits are randomly distributed). We have previously found (25)that the highly flexible siloxane backbone, which has a very small energy barrier to rotation (30), allows the covalently bound ferrocene mediators into close contact with the FAD/(FAD)H, centers of glucose 0 1991 American Chemical Society
678
ANALYTICAL CHEMISTRY, VOL. 63, NO. 7, APRIL 1, 1991
800 h
N
i
A
E
3
600
v
x d 1-
VI
m: n --
P o I ymc r
x R -
I
35:O
2
H
II
1:l
2
H
Ill
I:?
2
H
IV
1 :7.5
2
H
V
1:2
9
H
VI
I:2
2
CH,
Flgure 1. Schematic diagram of the ferrocene-containing siloxane random block copolymers used as electron-transfer mediators in glucose oxidase modified carbon paste electrodes.
oxidase, so that electrical communication is achieved. In our preliminary studies, we found that the efficiency of this communication is sensitive t o changes in the polymer structure. As indicated in Figure 1, several important structural features can be adjusted in these materials. For example, the ratio of ferrocene-modified siloxane subunits to unsubstituted siloxane subunits (m:nratio) can be varied, as can the length ( x ) of the alkyl side chain onto which the ferrocene moiety is attached. In addition, the substituents (R)bound to the ferrocene relays can be changed in order to adjust both the redox potential of the mediators and the rate a t which they can reoxidize the reduced enzyme. In the present work, we examine in detail the dependence of glucose sensor response on these structural features of the polysiloxane electron relay systems.
EXPERIMENTAL SECTION Chemicals. The methylhydrosiloxane homopolymer and methylhydrosiloxane-dimethylsiloxane copolymers were obtained from Petrarch Systems (Bristol, PA). Ferrocene and vinylferrocene were obtained from Aldrich (Milwaukee, WI). 9Ferrocenyl-1-nonene was prepared by the following procedure. Ferrocene was reacted with 6-bromohexanoyl chloride in the presence of A1Cl3 to produce (6-bromohexanoy1)ferrocene.This produce was reduced with LiAlH, to (6-bromohexyl)ferrocene, which was treated with Mg to give a Grignard reagent. This Grignard reagent, (6-ferroceny1)magnesiumbromide, was reacted with 3-bromopropene to produce the final product. Graphite powder (Product No. 50870) and paraffin oil (Product No. 76235) were obtained from Fluka (Ronkonkoma, NY). Glucose oxidase (EC 1.1.3.4,type VII, 129 units/mg) was obtained from Sigma (St. Louis, MO). Glucose (Sigma) solutions were prepared by dissolving appropriate amounts in 0.1 M phosphate/O.l M KCl buffer (pH 7.0); the glucose was allowed to reach mutarotational equilibrium before use (ca. 24 h). All other chemicals were reagent grade and were used as received. Polymer Synthesis. The synthesis of the ferrocene- and dimethylferrocene-modifiedsiloxane polymers has been described previously (27). Briefly, the methyl(2-ferrocenylethy1)-and methyl[2-(dimethylferrocenyl)ethyl]siloxane polymers were prepared by the hydrosilylation of vinylferrocene or 1,1'-dimethyl-3-vinylferrocene with the methylhydrosiloxane homopolymer or the methylhydrosiloxane-dimethylsiloxane copolymers ( n : nratios of 1:1, 1:2, and 1:7.5; see Figure 1)in the presence of chloroplatinic acid as a catalyst. The methyl(9-ferrocenylnony1)siloxane-dimethylsiloxane (1:2) copolymer was prepared via hydrosilylation of 9-ferrocenyl-1-nonene with the methylhydrosiloxane-dimethylsiloxane (1:2) copolymer. The molecular weight range of the ferrocene- and dimethylferrocene-modified polymers is approximately 5000-10 000. Purification of the polymers was achieved by reprecipitation from chloroform solution, via dropwise addition into a large excess amount of aceto-
2
400
c
C
; (3 .-vV
200
I 0
0
200
Potential
400 600 (mV vs. SCE)
Flgure 2. Cyclic voitammograms for the ferrocene-modified polysibxane/ghmse oxldase/carbon paste electrodes (scan rate: 5 mV/s) in pH 7.0 phosphate buffer (with 0.1 M KCI) solution with no glucose present (dashed line) and in the presence of 0.1 M glucose (solid line). The electrode contained polymer VI as the electron relay system.
nitrile at room temperature. This reprecipitation was repeated 2-3 times to ensure that no low molecular weight species (which could act as freely diffusing electron-transfer mediators) were present. Thin-layer chromatography and high-performance liquid chromatography showed that no oligomeric materials were present in the purified materials. Electrode Construction. The modified carbon paste for the sensors was made by thoroughly mixing 50 mg of graphite powder with a measured amount of the ferrocene-containingpolymer (the latter was first dissolved in chloroform); in the present work, the molar amount of the ferrocene moiety was the same for all electrodes (36 pmol of ferrocene per gram of graphite powder). After evaporation of the solvent, 5 mg of glucose oxidase and 10 p L of paraffin oil were added, and the resulting mixture was blended into a paste. We have previously shown (25,29),as have others (31),that glucose oxidase retains its activity when incorporated into a carbon paste matrix. The paste was packed into a 1.0-mL plastic syringe (7.0-mm outer diameter; 1.8-mm inner diameter) that had previously been partially f i e d with unmodified carbon paste, leaving approximately a 2-mm-deepwell at the base of the syringe. The electrodes were polished by rubbing gently on a piece of weighing paper, which produced a flat shiny surface with an area of approximately 0.025 cm2. Electrical contact was achieved by inserting a silver wire into the top of the carbon paste. Equipment. Cyclic voltammetry and constant potential measurements were performed by using a Princeton Applied Research potentiostat (Model 173), a Universal programmer (Model 175), and a Western Graphtec X-Y-T recorder (Model WX2400). All experiments were carried out in a conventional electrochemical cell containing pH 7.0 phosphate (0.1 M) buffer with 0.1 M KCl at 23 (h2) "C. Except when testing for oxygen interference, all experimental solutions were thoroughly deoxygenated by bubbling N2 through the solution for at least 10 min; in the constant potential experiments, a gentle flow of N2 was also used to facilitate stirring. Glucose samples injected into the cell were also thoroughly deoxygenated. In addition to the modified carbon paste working electrode, a saturated calomel reference electrode (SCE) and a platinum wire auxiliary electrode were employed. In the constant potential experiments, the background current was allowed to decay to a constant value before samples of a stock glucose solution were added to the buffer solution. A constant background current was attained approximately 5-10 min after application of the potential.
RESULTS AND DISCUSSION Cyclic Voltammetric Characterization of Enzyme/ Polymer-Modified Electrodes. Figure 2 shows typical voltammetric results for a carbon paste electrode containing
ANALYTICAL CHEMISTRY, VOL. 63, NO. 7, APRIL 1, 1991
100
150
--
4
679
- T
75 -.
50
-.
25
-. 0
0 0.2
0
10
20
30
Time (min)
Flgure 3. Typical response of the ferrocene-modified polysiloxane/ glucose oxidase/carbon paste electrode to addition of glucose at E = M O O mV (vs SCE). The electrode contained polymer VI as the electron relay system. Indicated are the amounts (in mL) of 0.1 M glucose added to the test solution (initial solution volume: 10 mL).
glucose oxidase and the methyl[@-(1’,3‘-dimethylferroceny1)ethyl]siloxane-dimethylsiloxane (1:2) copolymer (polymer VI) as the electron relay system. With no glucose present, the voltammogram displays very low anodic currents; that due to the oxidation of the polymer-bound dimethylferrocene moieties is nearly too small to be seen. Upon addition of glucose, however, the voltammetry changes dramatically, with a large increase in the oxidation current and no increase in the reduction current. The fact that the reduction current does not increase along with the oxidation current is indicative of the enzymedependent catalytic reduction of the ferricinium ion produced a t oxidizing potential values. Upon comparison of the voltammograms with and without glucose present, it is apparent that the ferrocene-containing siloxane polymer can act as an efficient electron-transfer relay system between the FAD/(FAD)H, centers of glucose oxidase and a carbon paste electrode. Electrodes containing glucose oxidase and the other polymeric mediators in Figure 1 display similar voltammetric characteristics upon addition of glucose. Voltammograms made with carbon paste electrodes containing only glucose oxidase with no polymeric relay system do not display this catalytic behavior. Constant Potential Response to Glucose. A typical glucose response trace (current density vs time) is shown in Figure 3 for a carbon paste electrode containing glucose oxidase and polymer VI as the electron relay system. This trace clearly shows the rapid response and good sensitivity of the sensor to clinically relevant glucose concentrations. The time required to reach 95% of the steady-state current was typically less than 10 s after addition of the glucose sample. The response times of sensors containing the ferrocene-modified polymers (I-V) were similar to that shown in Figure 3, although the magnitude of the response was dependent on several characteristics of the polymeric mediator. This dependence is discussed in more detail below. Dependence of Glucose Response on Polymer Subunit Ratio. As mentioned above, siloxane polymers are known to be extremely flexible (30).This flexibility will, of course, be sensitive to the amount of side-chain substitution present along the polymer backbone. For instance, in the homopolymer used in these studies (polymer I), the presence of a ferrocenylethyl moiety bound to each silicon subunit should provide an additional degree of steric hindrance, and thus a barrier to rotation about the siloxane backbone, in comparison with the copolymers, which have ferrocene relays attached to
100 200 300 400 Applied Potential (mV vs. SCE)
Flgure 4. Steady-state response to 31.5 mM glucose of the ferrocene-modified polysiloxane/glucose oxidase/carbon paste electrodes at several applied potentials. The polymeric relay systems are indicated next to each curve. Each curve is the mean result for four electrodes. 125
4
0
5 10 15 20 25 30 Glucose Concentration (mM)
35
Flgure 5. Glucose calibration curves for the ferrocene-modified polysiloxane/glucoseoxidaselcarbon paste electrodes at E = +300 mV (vs SCE). The polymeric relay systems are indicated next to each curve. Each curve is the mean result for four electrodes.
only a fraction of the Si atoms. Because these siloxane polymers are insoluble in water, their flexibility is an important factor in their ability to facilitate electron transfer from the reduced enzyme. As has been demonstrated in our laboratory and elsewhere (32),relays contained within more rigid redox polymers, such as poly(vinylferrocene), cannot achieve close contact with the enzyme’s redox centers and thus cannot serve as effective electron-transfer mediators. The importance of this feature can be seen quite clearly by comparing the mediating ability of homopolymer I with that of copolymers 11-IV, as shown in Figures 4 and 5. As expected, the dependence of the glucose response on the applied potential (Figure 4) is similar (maximal responses a t potential values 2300 mV vs. SCE) for sensors containing polymers I-IV (because each polymer contains unsubstituted ferrocene as the mediator), but the magnitude of this response is highly dependent on the polymer structure. By, in effect, systematically replacing a fraction of the ferrocenylethyl groups with methyl groups, it is possible to increase the flexibility of the polymer as well as the average spacing between the redox sites. From the results in Figure 4, and from the glucose calibration curves shown in Figure 5, it is apparent that these factors play a key role in the interaction between the polymeric relay system and the FAD redox centers in glucose oxidase. These results show an enhanced sensitivity to glucose for the electrodes containing polymers I1 and I11 as electron relay systems; these polymers have intermediate values for m:n. On the other hand, when the average spacing between the electron relays becomes very large, as in the case of polymer IV (m:n = 1:7.5), the increased polymer flexibility does not
680
ANALYTICAL CHEMISTRY, VOL. 63, NO. 7, APRIL 1, 1991 100
4
1
150
1
0
Glucose Concentration (mM)
Figure 6. Effect of increased alkyl chain length on glucose sensor response: glucose callbration curves for !he ferrocenemodlfled polysiloxane/glucoseoxidase/carbon paste electrodes at E = +300 mV (vs SCE). The polymeric relay systems are indicated next to each curve. Each curve is the mean result for four electrodes.
result in an increase in the measured catalytic current. Since the total molar amount of the mediating species is kept constant in each electrode (the polymers with larger m:n ratios have fewer redox sites per mole, so more material is necessary), then a change in the glucose response at constant potential can be attributed to either a change in the interaction between the enzyme and the polymer-bound relays or to changes in the interactions between the relays themselves. In the case of polymer IV, the decrease in sensor response could be due to the lower relay density, which results in a less efficient electron transfer between adjacent relays or to an electrical insulation of the relays by the relatively larger amount of siloxane polymer, which prevents close contact between the relays and the FAD/(FAD)H, centers of the enzyme. From the results shown in Figures 4 and 5 , it is clear that a systematic adjustment of the siloxane polymer's m:n ratio can provide an ideal compromise between polymer flexibility and relay density, and an optimal sensor response. Dependence of Glucose Response on Length of Alkyl Side Chain. The average spacing between the polymer-bound relays can also be adjusted by changing the length ( x in Figure 1)of the alkyl chain onto which they are attached. The effect of increasing the alkyl side chain length is apparent in Figure 6, which shows glucose response curves for sensors containing polymers I11 (x = 2) and V ( x = 9) as the electron relay systems. The maximal current densities measured with sensors containing polymer V are approximately 2 times greater than those measured with sensors containing polymer 111. It is clear from these curves, and from the Michaelis-Menten analysis described below, that polymer V is a more efficient electron relay system than polymer 111. Again, the longer alkyl chain may facilitate a more intimate interaction between the ferrocene moieties and the FAD/ (FAD)H, centers of glucose oxidase or better electrical communication between the relays themselves. Dependence of Glucose Response on Ferrocene Substitution. I t is well-known from studies of monomeric ferrocene mediators (5) that modification of the cyclopentadienyl ring can cause substantial changes in the redox potential of the mediator, as well as changes in the rate of electron transfer between the reduced form of the enzyme and the oxidized ferrocene. For example, Cass et al. (5) determined that 1,l'-dimethylferrocene reoxidized reduced glucose oxidase at a rate that was nearly 3 times that of the unsubstituted ferrocene. We have found that the improved mediating ability of dimethylferrocene is maintained after the species is covalently attached to a siloxane polymer. This is clear from the results shown in Figure 7 , which shows glucose response curves
T
A
"I
5 10 15 20 25 30 35 Glucose Concentration (mM)
Flgure 7. Effect of ferrocene substitution on glucose sensor response: glucose calibration curves for ferrocene-modified (I1 I) and dimethylferrocene-modifled (VI) polysiloxane/glucose oxidase/carbon paste electrodes at E = +300 mV (vs SCE). Each curve is the mean result for four electrodes.
100
n
0.0
0.5 1.0 1.5 2.0 2.5 Current Density/Concentration (pA cm-*/mM)
Flgure 8. Typical electrochemical Eadie-Hofstee plot for the ferroc-
ene-modified polysiloxane/glucose oxidase/carbon paste electrodes at E = +300 mV (vs SCE). The electrodes contained polymer 111 as the electron relay system. for sensors containing polymers I11 (ferrocene) and VI (dimethylferrocene). Because these polymers have identical subunit ratios (m:n= 1:2) and alkyl side chain lengths (x = 2), the greater response of sensors containing polymer VI can be attributed to the fact that dimethylferrocene is more efficient than ferrocene a t reoxidizing the reduced enzyme molecules. Apparent Michaelis-Menten Constants. The linear response range of the sensors can be estimated from a Michaelis-Menten analysis of the glucose calibration curves in Figures 5-7. The apparent Michaelis-Menten constant KMaPP can be determined from the electrochemical Eadie-Hofstee form of the Michaelis-Menten equation (23, 33)
where j is the steady-state current density, j,, is the maximum current density measured under conditions of enzyme saturation, and C is the glucose concentration. As shown in Figure 8 for glucose sensors containing polymer 111,a plot of j vs j / C (an electrochemicalEadie-Hofstee plot (33))produces a straight line and provides both KMaPP(slope) and j,,, (y intercept). The results of this Michaelis-Menten analysis are summarized in Table I. It is clear from these KMaPPvalues, and from the results in Figures 5-7, that the response of the sensors begins to deviate from linearity even a t glucose concentrations below 10 mM (the response to glucose is expected to be strictly linear for concentrations
