Anal. Chem. 1993, 65,2072-2077
2072
Performance of Subcutaneously Implanted Needle-Type Glucose Sensors Employing a Novel Trilayer Coating Francis M O U S S YD., ~Jed ~ ~ Harrison;*+Darryl W. O’BrienJ and Ray V. Rajottet.8 Department of Chemistry, University of Alberta, Edmonton, AB, Canada T6G 2G2,and Surgical-Medical Research Institute and Department of Surgery and Medicine, University of Alberta, Edmonton, AB, Canada, T6G 2N8
A miniature, needle-type glucose sensor based on a new trilayer membrane configuration has been prepared and evaluated both in vitro and in vivo. The perfluorinated ionomer, Nafion, was used as a protective, biocompatible, outer coating, and poly( @phenylenediamine)as an inner coating to reduce interference by small, electroactive compounds. Glucose oxidase immobilized in a bovine serum albumin matrix was sandwiched between these coatings. The entire sensor assembly of Pt working electrode and Ag/AgCl reference electrode was 0.5 mm in diameter and could be inserted subcutaneously through an l&gauge needle. The sensor current closely followed the plasma glucose level during a glucose tolerance test in active dogs, with a delay of 3 min, corresponding to the known lag time for subcutaneous glucose levels. The sensor remained functional after 1 week of implantation, but failed after 2 weeks due to degradation of the reference electrode. In vitro tests in pH 7.4 buffer or whole blood show the sensors have good selectivity, sensitivity of about 25 nA/mM, precision of 2-5%, and a 90% response time of 33 s, Stabilization following polarization requires 10-30 min in vitro and 30-40 min in vivo.
INTRODUCTION The potential uses of a glucose sensor for the treatment of diabetes include continuous glucose monitoring, the development of an alarm device for detecting hypoglycemia and, ultimately, inclusion as a component of a closed-loop insulin delivery system. For these reasons there has been a continued effort to develop an implantable glucose sensor, since the first demonstrated use of a sensor for continuous monitoring of physiological glucose levels.1 Due to the potential hazards of intravascular glucose sensing most studies have focused on the development of needle-type glucose sensors for subcutaneous glucose monitoring.2d The glucose concentration in subcutaneous tissue has been shown to closely follow plasma glucose concentrations.4
* Author to whom correspondence should be addressed. t
Department of Chemistry.
t Surgical-Medical Research Institute. 1 Department of Surgery and Medicine.
(1)Albisser, A. M.; Leibel, B. S.; Ewart, T. G.; Davidovac, 2.; Botz, C. K.; Zingg, W.; Schipper, H.; and Gander, R. Diabetes 1974,3,397-404. (2)Yamasaki, Y.Med. J. Osaka Uniu. 1984,35, 25-34. (3)Veeo,G. D.;Reach, G.;Thevenot,D. InBiosensors: Fundamentals and Applzcatzons; Turner, A. P. F., Karube, I., Wilson, G. S., Eds.; Elsevier: New York, 1987, pp 390-408. (4)Fisher, U.;Ertle, R.; Abel, P.; Rebrin, K.; Brunstein, E.; Hahn von Dorsche, H.; Freyse, E. J. Diabetologia 1987,30,940-945. (5)Pickup, J. C.; Shaw, G. W.; Claremont, D. J. Diabetologica 1989, 32,213-217. 0003-2700/93/0365-2072$04.00/0
Several needle-type, enzymatic glucose sensors have been developed,” but poor biostability (biocompatibility) has limited their effectiveness when implanted.6J These sensors are based on the oxidation of glucose by glucose oxidase in the presence of oxygen to produce HzOz, which is then detected amperometrically via oxidation. For glucose sensors in a biological matrix the use of a semipermeable outer membrane is critical. The membrane must prevent degradation of the enzyme and electrode in the biological environment, reduce interferences, and provide biocompatibility. In addition it must decrease the rate of glucose mass transfer relative to oxygen in order to obtain a linear response. Several membranes have been suggested for this purpose, including polyurethane,2,3p69 cellulose and cellulose acetate,10-13and the perfluorinated ionomer Nd10n.l”~~Bindra et a1.8 and MoattiSirat et al.9 have reported excellent results for a needle-type sensor coated with polyurethane, which worked for at least 10 days after implantation in animals. This sensor was well protected against interferences, but exhibited low sensitivity and required about 4 h to stabilize after implantation and subsequent polarization. We have previously shown that Nafion is an excellent choice as an outer dialysis layer for glucose oxidase (GOx) based enzyme electrodes in whole blood in vitro15 and that Nafion exhibits a biocompatibility similar to that of medical-grade silastic tubing.14J7 Nafion coatings provide some, but not complete, protection from redox-active interferences in the sample matrix. To overcome this, our previous experiments utilized a differential measurement of the output of matched working electrodes, for which one of the pair was glucose insensitive.’5 This approach is inconvenient with the needletype electrode design, so as others have done, we used an inner coating to screen out small interfering chemicals. For this purpose, electrodeposition of a poly(o-phenylenediamine) (PPD) film as the first layer on the working electrode was adapted from the method of Malitesta et a1.18 In addition to characterization of the assembled Pt/PPD/GOx/Nafion electrode both in vitro and in vivo, we report on the flow rate (6)Schichiri, M.; Asakawa, N.; Yamasaki, Y.; Kawamori, R.; Abe, H. Diabetes Care 1986,9,298-301. (7)Schichiri, M.; Kawamori, R.; Goriya, Y.; et al. Diabetologia 1983, 24, 179-184. (8)Bindra, D. S.;Zhang, Y.; Wilson, G.; Sternberg, R.; Thevenot, D. R.; Moatti, D.; Reach, G. Anal. Chem. 1991,63,1692-1696. (9)Moatti-Sirat, D.; Capron, F.; Poitout, V.; Reach, G.; Bindra, D. S.; Zhang, Y.; Wilson, G. S.; Thevenot, D. R. Diabetologia 1992,35,224-230. (10) Clark, L.C.; Duggan, C. A. Diabetes Care 1982,5,174-180. (11)Thevenot, D. R. Diabetes Care 1982,5,184-189. (12)Ikeda, S.;Ito, K.; Kondo, T.; Ichikawa, T.; Yukawa, T.; Ichihashi, H. R o c . Chen. Sens. 1983,17,620-625. (13)Gough, D. A.; Lucisano, J. Y.; Tse, P. H. S. Anal. Chem. 1985,57, 2.151-2.157. -- - - --- . . (14)Turner,R. F.B.; Harrison, D. J.; Rajotte,R. V.;Baltes, H. P. Sens. Actuators 1990,B l , 561-564. (15)Harrison, D. J.;Turner, R. F. B.; Baltes, H. P. Anal. Chem. 1988, 60,2002-2007. (16)Fan, 2.; Harrison, D. J. Anal. Chem. 1992,64, 1304-1311. (17)Turner, R. F.B.; Harrison, D. J.;Rajotte, R. V. Biomaterials 1991, 12, 361-368. (18)Malitesta, C.; Palmisano, F.; Torsi, L.; Zambonin, P. G. Anal. Chem. 1990,62, 2735-2740. 0 1993 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 65, NO. 15, AUGUST 1, 1993 varnish for insulation
WORKING ELECTRODE: Coiled platinum wire coated with - poly(opheny1enediamine) film - GO/albumin/glutaraldehyde
REFERENCE ELECTRODE: Coiled Ag/AgCl wire entire sensor coated with Nafion
--
insulated copper wires
0.5 mm
Flgure 1. Schematic diagram of the sensor.
dependence of the sensor, which is dependent on the permeability of the PPD layer. Careful design of the various film thicknesses led to a flow rate independent signal. The resulting sensor has a higher sensitivity, shorter response time, and shorter stabilization time than previously reported for needle-type sens0rs.m EXPERIMENTAL SECTION Material. High-purity glucose oxidase (Aspergillus niger, Calbiochem, La Jolla, CA) bovine serum albumin (fraction V, 98-99% albumin, Sigma), and glutaraldehyde (25% aqueous solution, Sigma) were used as received. All other chemicals were reagent grade. Solutions were prepared from doubly distilled, deionized water. A pH 7.4 phosphate buffer solution (PBS) was prepared from phosphate salts ( p = 0.05 M) with sodium benzoate (5 mM) and ethylene diaminetetraacetic acid (1 mM) as preservatives and NaCl(O.1M) as electrolyte. Glucose (0.1 M) was added, allowed to mutarotate overnight at room temperature, and then stored at 4 OC. Solutions of interfering species in pH 7.4 buffer were prepared just before use, as was 5 mM o-phenylenediamine (oPD) (Aldrich) in an acetate buffer (pH 5.5). Nafion solutions (Solution Technology Inc., Mendenhall, PA) of 0.5 and 3 wt % were prepared by dilution with 1:l 2-propanol and water. Equipment. Amperometry was performed using a Pine RDE-4potentiostat(Pine Instrument Company,GroveCity,PA). A Pine MSR rotator and Pt rotating disk electrode (0.5-cm diameter), were also used. A three-neck, round-bottom flask served as the electrochemical cell. Stirring was provided with an air-driven magnetic stirrer. Data were recorded using a x-y strip chart recorder. Sensor Fabrication. A 10-cm-long(0.2-mm diameter) varnished copper wire was used as the supporting element of the needle-type sensor (Figure 1).Electrical contact with a platinum wire (0.1-mmdiameter) (Puratronic,Johnson Matthey)was made at one end by removal of 1mm of varnish. The platinum wire was coiled 10 times around the insulated copper wire. The exposed copper wire was then insulated using liquid varnish (Red GLPT insulating varnish, Cardinal, Edmonton). One millimeter from the coiled platinum wire, a silver wire (0.1-mm diameter) (Puratronic, Johnson Matthey) was coiled 15 times and was connected to another varnished copper wire (0.15" diameter). Silver chloride was formed on the Ag wire by anodizing galvanostaticallyat 0.04 mA for 30 min in stirred 0.1 M HCl and then rinsing with deionized water. The coiled platinum wire was anodized at 1.9 V and cycled between -0.26 and +1.1 V vs a saturated calomel electrode (SCE) in 0.5 M HzSOd. Poly(ophenylenediamine) film was grown electrochemically on the Pt from a fresh, deaerated, unstirred o-PD solution at 0.65 V vs SCE (18). About 1 pL of acetate-buffered (pH 5.5) solution of 19.5 mg/mL glucose oxidase, 73.2 mg/mL albumin (BSA), and 5 mg/mL gl~taraldehydel~ was deposited onto the platinum (19) Yao, T. Anal. Chim. Acta 1983,148, 27-33.
2073
electrode by passing the sensor through the drop formed in a V-shape wire previously dipped in the enzyme solution. The sensor was dried for 0.5 hat room temperature. The entire sensor (Pt and Ag/AgClwires)was then dip coated with successivelayers of Nafion solutions (0.5%, 3 % , 5 % ) to 1&15 pm thick (estimated from the response time, as discussed below). The sensor was stored dry at room temperature or in 0.05 M PBS at 4 "C. In Vitro Sensor Evaluation. The needle-type sensor was characterized in pH 7.4 phosphate buffer or heparinized canine blood at 0.7 V vs the incorporated Ag/AgCl reference electrode (19.2 0.2 mV vs SCE in PBS). The steady-state current was measured as a function of added glucose concentration. It was possible to obtain the same response from a given sensor by carefui adjustment of the stir rate and sensor positioning. In Vivo Sensor Evalution. The acute experiment was performed on healthy, nonanesthetized female dogs. An indwelling 20-gauge catheter was placed in a foreleg vein for glucose infusion and blood sampling. A sensor first tested in uitro was inserted througha skin fold of the neck using an 18-gaugecatheter. The catheter was removed, leaving the sensor under the skin, and the sensor was then biased at +0.7 V. After the sensor signal stabilized, a blood sample was taken. A bolus of glucose (0.5 g/kg of body weight) was then injected through an indwelling venous catheter and blood was taken at varying intervals. Silasticcatheters, 0.062 in. i.d. X 0.125 in. o.d., used for chronic implantation of the glucose sensor (4 cm long) and the venous catheter (70 cm long),were prepared as described by OBrien e t ~ 1 Briefly, . ~ a~ 2-cm diameter, double velour, Dacron phlange (Meadox, Oakland, NJ) was placed around each catheter near the external end and held in place with medical-grade silastic adhesive (Dow Corning, Midland, MI). A glucose sensor, first tested in uitro, was inserted into the 4-cm catheter and secured in place by injecting silicone adhesive. The sensitive tip of the sensor extended 4 mm from the catheter's end. The sensor and venous catheters were then sterilized with UV irradiation and ethylene oxide, respectively. Under halothane anesthesia the sensor was inserted up to the Dacron phlange through a small incision made in the neck, which was then sealed. The venous catheter, used for blood sampling, was inserted in the right, external jugular. The use of Dacron phlanges promotes tissue growth and prevents infection and removal of the catheters. The sensor was tested immediately after surgery (and stabilization) and then once a week, as described for the acute experiments.
RESULTS AND DISCUSSION A needle-type geometry was chosen for the sensor to minimize the biological response to the implant and to facilitate the implantation procedure. When assembled, the sensor was 0.5 mm in diameter, comparable to the size of a 25-gauge needle, and could be inserted under the skin through an 18-gaugeneedle. As illustrated in Figure 1,coiled Pt and Ag/AgCl wires were wrapped around a Cu wire and served as the working and reference electrodes, respectively. A coiled reference electrode geometry was employed by Bindra et aZ.8 We have in addition coiled the working electrode to increase sensitivity by increasing the electrode's area. In a previous design using a Pt wire working electrode we found extensive cracking of the enzyme and Nafion layers often occurred.14~21The cracking pattern suggested this was due to hoop stress arising from the cylindrical geometry and the different swelling properties of the immobilized enzyme and Nafion layers. In the present design the immobilized glucose oxidase lies mainly between the coils, which significantly reduces cracking of the Ndion outer membrane. The underlying layer of poly(o-phenylenediamine) also reduced (20) OBrien, D. W.;Semple, H. A.; Molnar, G. D.; Tam, Y.; Coutta, R. T.;Rajotte, R. V.; Bayens-Simmonds, J. J.Pharmacol. Methods 1991, 25, 157-170. (21) Turner, R.F.B. Towards an Implantable Glucose Sensor. Ph.D. Thesis, University of Alberta, 1990.
ANALYTICAL CHEMISTRY, VOL. 65, NO. 15, AUGUST 1, 1993
2074
-
F
5.6 mM glucose pH 7.4 0.7 V vs SCE
10 min
0.5 mM Uric acid
c C
0.2 mM
E
Acetominophen
L
3
0
tt ii
6.1 mM
J
-
I
I
I
I
I
I
concna (mM)
ascorbic acid uric acid L-cysteine urea fructose acetaminophen
0.11 0.48 0.10 4.30 0.40 0.17
no PPD (n = 2)
5-min PPD
16-min PPD (n = 2)
5
0 5 i 3 poisoning 0 0 57 i 29
0 0 poisoning 0
39
( n = 5)
39
a Interferent is added at the maximum physiologicalconcentration in blood. b Error is expressed as the apparent percent increase in glucose concentration when the interferentis added to 6.6 mM glucoee, pH 7.4 buffer. The number of sensors tested is shown in parenthesis.
PtlPPDIGOxlNafion I
interferent
Ascorbate
J
Pt/GOx/Nafion I
mM Uric acid
5-
Table I. Sensor Response to Various Interferences % error re1 to 6.6 mM glucoseb
I
0.6
-
20 mM glucose ptl 7.4
Pt/GOx/Nafion n
the tendency of the coatings to crack, perhaps by enhancing adhesion. After months of dry storage or weeks of storage in pH 7.4 phosphate buffer at 4 O C , no decrease in sensitivity was observed and the response was still linear. However, after several wetting and drying cycles, the sensitivity increased and the linearity decreased. Optical microscopy showed cracks in the Nafion layer had developed, so repeated drying and wetting cycles were avoided. Permeationof Sensorswith PPDUndercoating. While the outer Ndion coating limits access of larger molecules and anions22 a significant background is still observed in blood, due to imperfect exclusion of anions and permeation by small neutral and cationic species.'e Malitesta et al. recently reported good selectivity using low-permeability PPD films with GOx coimmobilized during the PPD electrodeposition.18 Our initial attempts to prepare PPD films with coimmobilized GOx gave sensorswith very low sensitivity and poor selectivity, and we did not pursue this further. However, we found that electrodeposition of PPD alone, followed by a layer of immobilized GOx in bovine serum albumin and an outer coating of Nafion, gave sensors with improved selectivity over those without the PPD underlayer. The effect of several common interfering species present in biological samples was evaluated in phosphate buffer solution. Needle-type sensors with different PPD film thicknesses were prepared by electrodeposition for 5 and 15 min. Figure 2 illustrates the difference in response to the addition of ascorbate, uric acid, and acetaminophen at a Pt/ GOx/Nafion and a Pt/15 min PPD/GOx/Ndion electrode in a 5.6 mM glucose solution. The improved selectivity with PPD present is apparent, but it is clear that the commonly used drug acetaminophen still permeates the membranes. A more quantitative measure of selectivity was obtained by measuring the change in the sensor response when compounds were introduced a t their maximum physiological concentration. In Table I, the percentage increase in sensor current when the interferent was added to a 5.6 mM glucose solution is expressed as the apparent error. This is the maximum error that should be observed in blood for the compounds studied. With a GOx/Ndion coating all species studied interfered. Electrodepositionof PPD for 15min prior to coating with GOx and Nafion gave complete protection (22) Eiienberg, A., Yeager, H.L. Eds.Perfluorinated Zonomer Membranes; ACS Symposium Series 180; American Chemical Society: Washington, DC, 1982.
4
0.4 -
U
w
C
(3
0.2 -
L
Pt/PPD/GOx/Nafion
4
8
Time (min.) Figure 3. Sensor current in 20 mM glucose, pH 7.4 buffer, as the stirrer is turned off and on for (a) a Pt/Ox/Nafion electrode and (b) a Pt/ 15-min PPD/GOx/Nafion electrode.
against uric and ascorbic acids and urea. Sensors coated with PPD by a 5-min oxidation were less selective, but were still substantially protected from interference. However, PPD was permeated by acetaminophen. In addition, L-cysteine poisoned the sensors, producing a continuous, slow decline in sensor output. This effect could be reversed by rinsing and testing in fresh solution. Despite this, it is clear that the PPD film significantly enhances sensor selectivity. We have previously shown that Nafion must be present as an outer coating,16 so ita role is not supplanted by the PPD film. Hydrodynamic Flow Dependence. The effect of flow rate on sensor response is a critical factor, and it is desirable to obtain mass-transport rate independence. Sensors coated with GOx and Ndion show the expected decrease in current with decreasing mass-transport rate. However, sensors with an underlying PPD film prepared by 15-min electrolysis showed an increase in current as mass-transport rates decreased. Figure 3 shows this for a needle-type sensor in a stirred, 20 mM glucose solution. A similar effect was seen for Pt/PPD/GOx sensors that lacked the Ndion overcoat and for coimmobilized GOx in PPD films prepared by 15-min electrolysis. The dependence of the needle-type sensor current on masstransport rate in a 20 mM glucose solution was evaluated for sensors with PPD undercoatings prepared by 5 and 15 min of electrolysis. Table I1 shows the change in current caused when the magnetic stirrer was turned off, expressed as the percentage change of the current relative to a stirred solution. Tables I and I1show that a 5-min PPD film gives an excellent
ANALYTICAL CHEMISTRY, VOL. 65, NO. 15, AUGUST 1, 1993
Table 11. Hydrodynamic Effect on Sensor Output % change with no stirringb film coating GOx/N&ion 5-min PPD/GOx/Nafion 15-minPPD/GOx/Ndion
-20 4 (n = 3) 5 f 1 (n = 5) 42 f 25 (n = 3)
Table 111. In Vitro Characteristics of Needle-Type Glucose Sensors. background currentb(nA) sensitivity (nA/mM) response timeC(8)
Films coated on Pt, needle-typesensors. PPD, poly(o-phenylene diamine);GOx, glucose oxidase. b Percent change in sensor current when stirring is stopped in 20 mM glucose, pH 7.4, buffer. The number of sensors studied is shown in parenthesis.
2075
*
20 7 (n = 19) 25 f 10 (n = 19) 33 f 13 (n = 6)
@
3
* Omin. PPD 500
2min. PPD 5min. PPD 15min. PPD
--a
z E
CI
C
e
1
//$
400 300
c)
L
a
3
1 - tI
E 2
0
a Five-minute PPD/GOx/Nafion coating on needle-type Pt electrodes in pH 7.4 buffer. The number of sensors studied is shown in parenthesis. Measured after stabilization of the background following polarization (10-30 min). Rise to 90%responsefor a glucose concentration step from 0 to 5.6 mM.
C
! 3
1
200
0 100
0
I/0
10
20
30
40
Square root RPM Flgurr 4. Levich plot of current vs (rotation (rpm)I/2 of naked and ppD.coated Pt rotating disk electrodes in 20 mM H202, pH 7.4. PPD was deposited by electrolysis for 2, 5, and 15 min. compromise between protection from interferences and stir rate independence. This will of course be more significant for a sensor implanted intravenously, since there will be minimal flow in subcutaneous sites. The sensor flow rate dependence described above was studied in glucose solutions that did not contain the enzyme reaction product HzOz. This means there is a flux of HzO2 into the solution resulting from the concentration gradient. To understand the competition between transport of HZOZ back into solution vs through the PPD we examined the masstransport behavior of PPD-coated electrodes toward H202 in a 20 mM H 2 0 ~solution, using the controlled hydrodynamics of the rotating disk electrode (RDE). A Pt RDE was coated with PPD alone by electrodeposition for 2, 5, and 15 min. Figure 4 shows a Levich plot of the anodic current at 0.7 V vs SCE as a function of (rotation rate)l/2for a naked electrode and different coating thicknesses. The data show that the film’s permeability is a function of the electrodepositiontime, decreasingwith increasing film thickness. Further, the weak rotation rate dependence for even the shortest film deposition period shows PPD is poorly permeable to HzO2, despite the rapid (
