Anal. Chem. 1995,67, 1240-1244
Design and Optimization of a Selective Subcutaneously Implantable Glucose Electrode Based on LCWiredgg Glucose Oxidase Elisabeth Cs&egi,t David W. Schmidtke, and Adam Heller* Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712-1062
An implantable 0.29 mm 0.d. flexible wire electrode was designed for subcutaneous monitoring of glucose. The electrode was formed by sequentiallydepositingin a 0.09 mm deep shielded recess at the tip of a polyimideinsulated 0.25 mm gold wire a "wired" glucose oxidase (GOX) sensing layer, a mass transport limiting layer, and a nonfouling biocompatible layer. The glucose sensing layer was formed by cross-linking {poly[(1-vinylimidazoly1)osmium(4,4-dimethylbipyridine)2Cll}+'2+ (PVI13dme-Os) and GOX with poly(ethy1ene glycol) diglycidyl ether (PEG). The glucose mass transport restricting layer consisted of a poly(ester sulfonic acid) film (Eastman AQ 29D) and a copolymer of polyaziridine and poly(viny1 pyridine) partially quaternized with methylene carboxylate. The outer biocompatiblelayer was formed by photocross-linking tetraacrylated poly(ethy1ene oxide). The three layers contained no leachable components and had a total mass less than 2.2 pg (-50 ng of Os). When poised at +200 mV vs SCE and operated at 37 "C, the 5 x cm2 electrode had in vitro a sensitivity of 1-2.5 nA mM-'. The current increased with the glucose concentration up to 60 mM, and the 10-90%response time was 1min when the glucose concentrationwas abruptly raised from 5 to 10 mM. The sensitivity decreased by less than 4% over a test period of 1week, during which the electrode was operated continuously in a 10 mM glucose physiological buffer solution at 37 "C. The variation in the current when interferants were added at their physiological concentrations (0.1 mM ascorbate, 0.17 mM acetaminophen, 0.48mM urate) was 1SD of the assays (*2%), and the glucose electrooxidation current was nearly independent of oxygen partial pressure, even at low glucose concentrations. When an electrode was implanted subcutaneously in a rat, it retained its in vitro calibration and tracked the blood glucose concentration before, during, and after intraperitoneal glucose infusion.
-
Several research groups, including ours, are considering the continuous subcutaneous monitoring of glucose levels using amperometric sensors.1s2 Implanted glucose sensors must meet safety, reliability, performance, convenience, and ease of use * FAX, 1-512-471-8799.
Permanent address: University of Lund. Department of Analytical Chemism, P.O. Box 124, 5221 00 Lund, Sweden. (1) Reach, G.; Wilson, G. S. Anal. Chem. 1992,64, 381A-386A. ( 2 ) Wilson, G. S.; Zhang, Y.; Reach, G.; Moatti-Sirat, D.; Poitout, V.; Thevenot, D. R.; Lemonnier, F.; Klein J.-C. Clin. Chem. 1992,38/9,1613-1617. +
1240 Analytical Chemistry, Vol. 67, No. 7, April 7 , 7995
criteria. Concerning safety, the sensor must not have leachable components, the amount of foreign matter introduced in the body must be small enough to avoid hazard, and the sensor must not trigger immune reaction. Recently, several systems that have no leachable components, some of which may be of use in implanted sensors, have been d e ~ c r i b e d . ~With - ~ respect to reliability, the sensor must provide information sufticiently accurately for making valid clinical decisions during its specijied period of use; as a minimum, it must be able to trigger alarms in cases of hypo- or hyperglycemia or failure of the system. Concerning performance, the sensor must measure glucose concentrations through the clinically relevant 2-30 mM glucose concentration range, in the presence of all chemicals, including common electrooxidizable interferants like ascorbate, acetaminophen, and urate. Ideally, it must allow one-point calibration (recalibration by withdrawal and analysis of one sample of blood). This requires that the current associated with both electrooxidizable and electroreducible interferants, as well as oxygen, be negligible relative to the signal, even at low glucose concentration, and that the function defining the relationship between the signal and the glucose concentration be known. Moreover, the signal to noise ratio must be adequate, and the response must be fast enough. Concerning convenience and ease of use, the sensor should be insertable with a simple device, such as a small needle, and its presence in the tissue must not be irritating. For one-point in vivo calibration and for accuracy, the absence of interferences is of essence. The often encountered interference by electrooxidizable species can be lessened by reducing the operating by enzymatically preoxidizing the interfera n t ~ or , ~by , ~overcoating the sensing zone with films that are less permeable to interferants.l0-I4 (3) Ohara, T.; Rajagopalan, R; Heller, A Anal. Chem. 1993,65, 3512-3517. (4) Ohara, T.; Rajagopalan. R; Heller, A Anal. Chem. 1994,66, 2451-2457. (5) Kaku, T.; Karan, H.; Okamoto, Y. Anal. Chem. 1994,66, 1231-1235. (6) Tatsuma, T.; Saito, K; Oyama, N. Anal. Chem. 1994,66, 1002-1006. (7) Marcinkeviciene, J.; Kulys, J. J. Biosens. Bioelectvon. 1993,8, 209-212. (8) Maidan, R; Heller, A Anal. Chem. 1992,64, 2889-2896. (9) Csoregi, E.;Quinn, P. C.; Schmidtke, W. D.; Lmdquist, S.-E.;Pishko, V. M.; Ye, L.; Katakis, I.; Hubbell, A. J.; Heller, A Anal. Chem. 1994,66, 31313138. (10) Palmisano, F.; Zambonin, P. G. Anal. Chem. 1993,65, 2690-2692. (11) Moussy, F.; Hanison. D. J.; O'Brien, D. W.; Rajotte, R V. Anal. Chem. 1993, 65, 2072-2077. (12) Chen, C.-Y.; Gotoh, M.; Makino, H.; Su. Y.-C.; Tamiya, E.; Karube, I. Anal. Chim. Acta 1992,265, 5-14. (13) Zhang, Y.; Hu, Y.; Wilson, G. S. Anal. Chem. 1994,66, 1183-1188. (14) Lowry, J. P.; McAteer, IC; El Atrash, S. S.; Duff, A; O'Neill, R D. Anal. Chem. 1994,66,1754-1761. 0003-2700/95/0367-1240$9.00/0 0 1995 American Chemical Society
Recently we reported a recessed, four-layered sensor made of a polymer-insulated gold wireg based on ''wired GOX, meeting most of the criteria listed above. Speciiically, the four layers of the sensor were a sensing layer, an insulating and mass transport limiting layer, an interference eliminating layer, and a nonfouling biocompatible layer. This sensor could be calibrated in vivo by using only one withdrawn blood sample, i.e., its one-point in vivo calibration was practical. However, the instability of the enzymes in the interference eliminating layer (peroxidase and lactate oxidase) limited the half-life of the sensor when operating in the presence of electrooxidizable interferants to -35 h at 37 "C. In this article we describe a novel structure based on a redox polymer similar to the one reported by Ohara et where the interference eliminating layer is no longer required and the number of enzymes in the structure is reduced to one: glucose oxidase. This redox polymer has a lower redox potential of 95 mV SCE, which allows the poising of the operating electrode at 200 mV vs SCE. At this potential, urate, ascorbate, and acetaminophen at their physiological concentrations are not electrooxidized at appreciable rates. The sensor also has a novel mass transport limiting layer between the wired enzyme sensing layer and the biocompatible outer layer. The three-layered electrode is simpler to make, is more stable, and, signiticantly, no longer requires the presence of oxygen for interference elimination. It has only 2.2 pg of material that might be lost in the body in a worst case accident, of which 0.9pg is a biocompatible polymer. EXPERIMENTAL SECTION Chemicals. Glucose oxidase from Aspergillus niger (GOX) (EC 1.1.3.4,159 units mg-I), L-ascorbic acid (Na salt) (AA), 4acetaminophen (AAP), and uric acid (Na salt) (UA) were purchased from Sigma (St. Louis, MO); Nafion (5% w/v in 90% lower aliphatic alcohol and 10% water) from Aldrich; Eastman AQ 29D (30%water solution) (EAQ) from Eastman Chemical Products Inc. (Kingsport, TN); polyfunctional aziridine (XAMA-7) (PAZ) from E.I.T. Inc. (Lakewilie, SC); and polyallylamine (PAL) and poly(ethy1ene glycol) diglycidyl ether 400 (PEG) from Polysciences (Warrington, PA). The glucose oxidase wiring polymer {poly[(1-vinylimidazoly1)osmium (4,4'-dimethylbipyridine)~C11}+/~+ (PVII$-dme-Os),where the subscript indicates that every 13th vinylimidazole mer is complexed to an Os center, was synthesized according to the previously published protoc01.~ Poly(viny1pyridine) was quaternized with bromoacetate to form PVPA according to the procedure of Kataki~,'~ as follows: poly(vinyl pyridine) (PVP,MW 40 000) was dissolved in DMFethylene glycol (1:l) and heated with bromoacetic acid at 70 "C for 18 h, using a molar ratio of pyridine to bromoacetic acid 5:2. The product was precipitated by slowly adding the reaction mixture to well-stirred acetone. The solid was redissolved in water and dialyzed for 48 h using a Spectropor 3200 MW cutoff dialysis membrane against deionized water. Elemental analysis and NMR show that 20% of the pyridines were quatemized. The outer nonfouling and biocompatible layer of tetraacrylated poly(ethy1ene oxide) (PEO) was formed by photocross-linking.gJ6 All chemicals were used as received. The solutions (if not otherwise specified) were prepared with deionized water. (15) Katakis, I. Development and Analysis of Operation of Enzyme Electrodes Based on Electrochemically 'Wired" Oxidoreductases. Ph.D. Thesis, The University of Texas at Austin, 1993. (16)Pathak, C.;Sawhney, A.; Hubbell, J.J. Am. Chem. SOC.1993,114,83118312.
Figure 1. Structure of the PVlj3-dme-Os polymer.
Procedures. Electrode Preparation. The electrodes were prepared as described earlier? with the following changes in procedure. A manual microinjector (Model NA-1 from Sutter Instrument Co., Novato, CA) was used for filling the recess under the microscope. With this microinjector, there was less than 10% difference in the current output of similarly prepared electrodes, n = 18. The following layers were sequentially deposited in the recess. The sensing layer was made by wiring GOX to the electrode through a redox hydrogel formed of the redox polyelectrolyte PVIu-dme-Os (Figure l), GOX, and PEG.4 The calculated composition of this polymer was C, 53.26;C1, 3.54;H, 6.38;N, 20.94;and Os,9.48. Elemental analysis (Oneida Research Services, Whitesboro, NY) showed C, 54.06;C1, 3.74;H, 5.92;N, 19.74;and Os, 9.52. The sensing films were made of solutions of 10 mg mL-l redox polyelectrolyte, 10 mg mL-l GOX (in HEPES 10 mM at pH 8.1),and freshly dissolved PEG (2.5mg mL-') mixed at 78:166wt % ratio. After coating, the electrodes were rinsed with water five timesg and cured at 45 "C for 15 min. The mass transport restricting layer was formed of one of the following: Nafion (0.5wt % in 95% ethanol); PAL (4.5mg mL-' in 100 mM HEPES, pH 7.0)with PAZ (30mg mL-') in 1:2volume ratio; WPA (25 mg mL-l) with PAZ (30mg mL-') in 1:2volume ratio; EAQ (5 wt % in water); or EAQ overcoated with PVPA (25 mg mL-1) and PAZ (30mg mL-l) in a volume ratio of 1:2.The PAZ solution was freshly prepared and used within 15 min of dilution. The EAQ films were cured at room temperature with 20 min intervals between coatings. Nafion, PAL+PAZ, and PVPAfPAZ layers were cured at room temperature for 8 h. The nonfouling biocompatible layer was formed of a sensitized 10 wt % aqueous tetraacrylated PEO solution by photo-cross-linking (45 s UV exposure) after each coating, as described!J6 Batches of five equally prepared electrodes were made. Unless otherwise speciiied, these were tested in physiological buffer (PBS) containing 10 mM glucose. The reported layer thicknesses were calculated by assuming that all introduced material was deposited on the electrode surface, even though part was actually deposited on the wall of the recess. In measuring the interferantcaused currents, glucose was dissolved first, and then the interferants were added in the sequence ascorbate-acetaminophen-urate. The current was measured, after it stabilized, after each addition. Instrumentation. In vitro measurements were carried out using batches of electrodes in a thermostated and stirred 10 mL Analytical Chemistry, Vol. 67, No. 7, April 1, 1995
1241
three-electrode electrochemical cell at 37 "C, with the wired GOX as working, SCE as reference, and platinum wire as auxiliary electrodes. The electrolyte was 20 mM phosphate buffer containing 0.15 M NaCl at pH 7.15. The working electrode was poised at +200 mV vs SCE. The temperature was controlled with an isothermal circulator (Fisher Scientific, Pittsburgh, PA). The currents were measured using an Ensman E1 400 bipotentiostat (Ensman Instrumentation, Bloomington, IN), a Biometra EP 30 (Biometra GmbH, Gottingen, Germany) bipotentiostat, or a BAS 4B (BioAnalytical Systems Inc. West Lafayette, IN) monopotentiostat. The potentiostats were connected to a six-channel strip charter recorder (Kipp and Zonen type BD101, Bohemia, NY). In vivo measurements were camed out using 400 g male Sprague-Dawley rats fasted overnight. The rats were anesthetized with an intraperitoneal injection of sodium pentobarbital (65 mg/ kg of rat wt) prior to the experiment. The right jugular vein was then catheterized using a 0.0375 in. 0.d. medical grade silicone tubing (Baxter, McGraw Park, IL). A 3-way stopcock (Baxter) was then connected to the catheter to allow frequent blood sampling. Next, the animal was systemically heparinized by administering heparinized saline (100 units/kg of rat wt) through this catheter line. Clotting of the line was prevented by administering smaller doses of heparinized saline (1.75 units/mL) after each blood withdrawal. An electrode was then inserted subcutaneously in the rat's back using a 22 gauge Per-Q-Cath introducer (Gesco International, San Antonio, Tx). The Ag/AgCl surface skin reference electrode (Microelectrodes Inc., Londonderry, NH) was attached to a shaved portion of the rat's abdomen previously coated with a conductive gel. The implanted electrode was poised at f 2 0 0 mV vs Ag/AgCl using an EG&G Princeton Applied Research Model 400 EC bipotentiostat (Princeton, NJ), and glucose signals were recorded using a data logger (Rustrak Ranger, East Greenwich, N).The rat's body temperature was maintained at 37 "C by using a homeostatic blanket (Harvard Apparatus, South Natick, MA). The current was allowed to reach a constant baseline before venous blood sampling was started. Blood samples were collected in tubes spiked with sodium fluoride and then analyzed with a Model 23A glucose analyzer (YSI Inc., Yellow Springs, OH). Glucose was infused for about 1 h using a syringe pump (Harvard Apparatus) at a rate of 25 mg of glucose/ minekg of rat wt. The in vivo experiments were approved by the University of Texas Institutional Animal Use and Care Committee. RESULTS AND DISCUSSION
The new design improved the performance of our earlier described subcutaneous 0.29 mm 0.d. wire sensor by (a) making the sensor's interference rejection independent of the presence or absence of oxygen; (b) extending the response through the clinically relevant 2-30 mM range; (c) extending the life of the sensor in continuous use at 37 "C to 1 week by removing the previously used least stable interference eliminating layer that consisted of cross-linked peroxidase and lactate oxidase and by reducing mass transport through an improved flux controlling layer; (d) reducing the residual interference by 02 at low glucose concentrations; and (e) improving safety by reducing the total amount of electrode material that can be lost in a worst case accident to 2.2 pg, of which 0.9 pg was a biocompatible film. The depth of the 0.25 mm i.d., 0.02 mm wide polyimide sheathshielded recess, at the tip of the wire electrode, was reduced from 125 to 90 pm, as the -80 pm thick interference eliminating layer 1242 Analytical Chemistry, Vol. 67, No. 7, April 1, 1995
was no longer needed. With the recess shortened, adhesion of material to the polyimide wall of the recess during deposition of the layers was reduced. At 90 pm recess length, there still was an -40 pm deep unfilled "sleeve" protecting the top biocompatible layer. The shallower recess was formed by anodically dissolving the gold wire in a cyanide solution under galvanostatic (8 mA/ wire) control, passing 1C/electrode. The etching procedure was otherwise similar to the one described earlier.g When the sensing layer was overcoated with the mass transport limiting and the biocompatible layers, the current at 10 mM glucose concentration and at 25 "C increased with the thickness of the sensing layer up to 12 pm, and then decreased (8 pm, 42 nA; 12 pm, 62 nA; 15 pm, 55 nA; 19 pm, 20 nA). A 15 pm thickness, where the ratio of the glucose electrooxidation to interferant electrooxidation currents was high, was preferred. The 15 pm thick layer was formed by successive applications (20) of microdroplets of the premixed GOX, PV113-dme-Os, and PEG solutions. Curing at 45 "C for 15 min did not deactivate the GOX, yet it shortened the time needed for sensor preparation. The electrodes were poised at 200 mV(SCE), well above the reversible potential of PV113-dme-Os, which was at 95 mV(SCE). Mass transport limiting layers consisting of Nation, EAQ, PALfPAZ, and PVPA+PAZ were investigated. Nafion could not be reproducibly deposited on the sensing layer, because the alcohol of the supplied emulsion evaporated rapidly and the material deposited mainly on the wall of the recess. The electrodes with PAL+PAZ or PVPA+PAZ overcoatings lost less than 10%of their sensitivity when continuously operated at 37 'C in 10 mM glucose for 72 h, but their sensitivities were poor, about 0.5 nA mM-'. Though the EAQ layers were easy to deposit, the half-life of the continuously operating electrodes made only with an EAQ mass transport limiting layer was only 35 h at 37 "C. Investigation of the glucose mass transport restriction by an EAQ film showed that the apparent Michaelis constants (Km)of electrodes made with a 15 pm thick sensing layer increased when thicker EAQ films were formed by applying 3, 4, 5, or 6 microdroplets resulting, respectively, in 11,15,20, and 24pm thick EAQ films. The corresponding K, values were 18,33,52, and 49 mM glucose. The fact that K,,, no longer increased when the layer's thickness was increased from 20 to 24 pm suggests that transport was not restricted as much by the EAQ film itself as it was by the interacting EAQ polyanion/PVIIj-dme-Os polycation system, forming a heavily electrostatically cross-linked interface. The polycationic PVPA-PAZ film on the EAQ layer also formed such an interface. A mass transport limiting layer made of 13 pm thick EAQ and a 1.5 pm thick PVPA-PAZ (ratio 1:2) film was chosen for the design. The structure, shown in Figure 2, was completed by overcoating the mass transport limiting layer with the nonfouling, biocompatible, photo-cross-linked, -20 pm thick tetraacrylated PEO layer.16 Thus, the combined thickness of all layers was -50 pm. The sensitivity of the optimized electrode was 1.0-2.5 nA mM-', and its dynamic range was 0-60 mM glucose. In batches of electrodes that were made identically, some (such as the electrode of Figure 3) showed nearly linear response. In others, the sensitivity increased with glucose concentrations, possibly because glucose dynamically reduced the tight ion pairing of the polyanionic-polycationic interface that limited its transport. The difference was attributed to different intermixing of the EAQ polyanionic with the neighboring polycationic layers, changing
T
I
0.29 mm
Sensing layer glucose oxidase 110 ng PVII3-dme-Os 552 ng Peg 42 ng
I
Mass transport limiting layer EAQ PVPA PA2
560 ng 26 ng 62 ng
I
T 0.25 mm
I
PEO/TA
885ng
Figure 2. Schematic drawing showing the dimensions and constituents of the layers of the sensor.
100 aC .-
80
t
I e 4
\
60
