A Wireless, Remote Query Glucose Biosensor Based on a pH

This paper describes a wireless, remote query glucose biosensor using a ribbonlike, mass-sensitive magnetoelastic sensor as the transducer. The glucos...
16 downloads 0 Views 295KB Size
Download PDF
Anal. Chem. 2004, 76, 4038-4043

A Wireless, Remote Query Glucose Biosensor Based on a pH-Sensitive Polymer Qingyun Cai,† Kefeng Zeng,† Chuanmin Ruan,† Tejal A. Desai,‡ and Craig A. Grimes*,†

Department of Electrical Engineering, and Department of Materials Science and Engineering, 217 Materials Research Laboratory, The Pennsylvania State University, University Park, Pennsylvania 16802, and Department of Biomedical Engineering, Boston University, 44 Cummington Street, Boston, Masschusetts 02215

This paper describes a wireless, remote query glucose biosensor using a ribbonlike, mass-sensitive magnetoelastic sensor as the transducer. The glucose biosensor is fabricated by first coating the magnetoelastic sensor with a pH-sensitive polymer and upon it a layer of glucose oxidase (GOx). The pH-responsive polymer swells or shrinks, thereby changing mass, respectively, in response to increasing or decreasing pH values. The GOx-catalyzed oxidation of glucose produces gluconic acid, inducing the pH-responsive polymer to shrink, which in turn decreases the polymer mass. In response to a time-varying magnetic field, a magnetoelastic sensor mechanically vibrates at a characteristic resonance frequency, the value of which inversely depends on sensor mass loading. As the magnetoelastic films are magnetostrictive, the vibrations launch magnetic flux that can be remotely detected using a pickup coil. Hence, changes in the resonance frequency of a passive magnetoelastic transducer are detected on a remote query basis, without the use of physical connections to the sensors.The sensitivity of the glucose biosensors decreases with increasing ionic strength; at physiological salt concentrations, 0.6 mmol/L of glucose can be measured. At glucose concentrations of 1-10 mmol/ L, the biosensor response is reversible and linear, with the detection limit of 0.6 mmol/L corresponding to an error in resonance frequency determination of 20 Hz. Since no physical connections between the sensor and the monitoring instrument are required, this sensor can potentially be applied to in vivo and in situ measurement of glucose concentrations. Diabetes mellitus results from the body’s abnormal response to or production of insulin, resulting in hyperglycemic glucose values. The effective method to reduce the risk of chronic complications in diabetic patients is to in vivo and continuously monitor blood glucose levels, instantly providing the patients with the information required to optimize insulin therapy and metabolic control. During the last 30 years, numerous efforts have focused on glucose sensor development.1-18 Of these studies, electro* Corresponding author: [email protected]. † The Pennsylvania State University. ‡ Boston University. (1) Koschinsky, T.; Heinemann, L. Diabetes Metab. Res. Rev. 2001, 17, 113123. (2) Gough, D. A.; Armour, J. C. Diabetes 1995, 44, 1005-1009.

4038 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

chemical methods4-10 and optical methods11-20 have received considerable attention. Most electrochemical methods are based on the glucose oxidase (GOx)-catalyzed oxidation of glucose. The enzymatic oxidation of glucose produces gluconic acid and hydrogen peroxide with consumption of oxygen. The measurement signal is derived from the electrochemical oxidation of hydrogen peroxide4 or reduction of oxygen5 at a fixed electrode potential; the working electrodes are composed of noble metals, and in most cases, an Ag/AgCl reference electrode is used. Since hydrogen peroxide can deactivate GOx, catalase, which catalyzes hydrogen peroxide decomposition to oxygen and water, is generally employed with GOx in oxygen-based glucose sensors.5 A drawback to the practical use of electrochemical sensors is the interference caused by compounds that can be oxidized or reduced at the working potential. For example, compounds commonly encountered in clinical samples such as uric acid, acetaminophen, and ascorbic acid can be oxidized at the H2O2 potential interfering with detection. To eliminate electroactive interference, mediators are frequently used to lower the anodic potential of the H2O2-based sensors.6,7 Mediators are usually low (3) Zhao, W.; Song, Ch.; Pehrsson, P. E. J. Am. Chem. Soc. 2002, 124, 1241812419. (4) Retama, J. R.; Lopez-Ruiz, B.; Lopez-Cabarcos, E. Biomaterials 2003, 24, 2965-2973. (5) Baker, D. A.; Gough, D. A. Anal. Chem. 1996, 68, 1292-1297. (6) Groom, C. A.; Luong, J. H. T.; Thatipalmala, R. Anal. Biochem. 1995, 231, 393-399. (7) Chen, T.; Friedman, K. A.; Lei, I.; Heller, A. Anal. Chem. 2000, 72, 37573763. (8) Wilkins, E.; Atanasov, P. Med. Eng. Phys. 1996, 18, 273-288. (9) Jung, S.; Trimarchi, J. R.; Sanger, R. H.; Smith, P. J. S. Anal. Chem. 2001, 73, 3759-3767. (10) Luong, J. H. T.; Nguyen, A. L.; Guilbault, G. G. Adv. Biochem. Eng. Biotechnol. 1993, 50, 85-115. (11) Shiomi, Y.; Saisho, M.; Tsukagosi, K.; Shinkai, S. J. Chem. Soc., Perkin Trans. 1 1993, 2111-2117. (12) Yoon, J.-Y.; Czarnik, A. W. J. Am. Chem. Soc. 1992, 114, 5874-5875. (13) Cameron, B. D.; Gorde, H. W.; Satheesan, B.; Cote, G. L. Diabetes Technol. Ther. 1999, 1, 125-143. (14) Berger, A. J.; Koo, T. W.; Itzkan, I.; Horowitz, G.; Feld, M. S. Appl. Opt. 1999, 38, 2916-2926. (15) Zhang, L.; Small, G. W.; Arnold, M. A. Anal. Chem. 2002, 74, 4097-4108. (16) Shafer-Peltier, K. E.; Haynes, C. L.; Glucksberg, M. R.; Van Duyne, R. P. J. Am. Chem. Soc. 2003, 125, 588-593. (17) DiCesare, N.; Pinto, M. R.; Schanze, K. S.; Lakowicz, L. R. Langmuir 2002, 18, 7785-7787. (18) Ye, K. M.; Schultz, J. S. Anal. Chem. 2003, 75, 3451-3459. (19) Shafer-Peltier, K. E.; Haynes, C. L.; Glucksberg, M. R.; Van Duyne, R. P. J. Am. Chem. Soc. 2003, 125, 588-593. (20) Asher, S. A.; Alexeev, V. L.; Goponenko, A. V.; Sharma, A. C.; Lednev, I. K.; Wilcox, C. S.; Finegold, D. N. J. Am. Chem. Soc. 2003, 125, 3322-3329. 10.1021/ac0498516 CCC: $27.50

© 2004 American Chemical Society Published on Web 05/20/2004

molecular weight species that work as electron acceptors, replacing the role played by oxygen, shuttling electrons between the redox center of the enzyme and the working electrode. An alternative method to eliminate electroactive interference is to use a porous membrane to protect the electrode. In O2-based sensors, an oxygen-permeable hydrophobic membrane is included so that compounds of molecular weight greater than 100-200, such as uric acid and ascorbic acid, can be excluded. However, the membrane acts as a diffusion barrier and affects sensor performance, including parameters such as the detection limit, linear range, and response time.10 In addition, smaller molecules can still pass through the membrane, resulting in electrochemical poisoning. Inevitably sensor sensitivity decays with continuous use due to the electrochemical poisoning and electrode polarization21,22 Under physiological conditions, glucose molecules exchange between the blood plasma and interstitial fluid. Due to the obvious risks related to placing a sensor in vivo, for continuous monitoring of glucose concentration, glucose sensors are commonly placed in skin tissue where they are exposed to interstitial fluid. The glucose level in interstitial fluid is related to that in blood, but a time delay exists between blood and interstitial fluid glucose levels; the physiological lag time can vary from several seconds to tens of minutes.23,24 Noninvasive optical glucose sensors obtain information from the reflection characteristics of light incident upon the skin. Reflected light properties are altered by interactions of the light with the glucose molecules, e.g., laser polarimetry13 and vibrational spectroscopy14-16 or due to the indirect effect related to the interaction between glucose and an analytical reagent that induces changes in skin or humor physical properties such as fluorescence spectroscopy,17,18 Raman scattering,19 and diffraction spectroscopy.20 Although the noninvasive basis of optical detection makes it an ideal candidate for continuous glucose monitoring, biomolecules similar to glucose normally exist in a biological environment that can interfere with detection giving false positive results.13 We make note that recently nonenzyme glucose sensors have received attention25,26 that, compared with enzyme-based sensors, are more stable and with longer lifetimes but suffer from low selectivity. In this work, a wireless magnetoelastic glucose sensor was fabricated. In response to a time-varying magnetic field, the magnetoelastic ribbon longitudinally vibrates at a fundamental characteristic resonance frequency fr that inversely depends on the sensor length L, with width and thickness much smaller than length, Young’s modulus E, and density F:27,28

fr )

xEF 2L1

(1)

The magnetoelastic qualities of the sensor, high permeability and magnetization with low coercive force, enable efficient (21) Sternberg, F.; Meyerhoff, C.; Mennel, F. J.; Mayer, H.; Bischof, F.; Pfeiffer, E. F. Diabetologia 1996, 39, 609-612. (22) Moussy, F.; Harrison, D. J.; Rajotte, R. V. Int. J. Artif. Organs 1994, 17, 88-94. (23) Rebrin, K.; Steil, G. M.; Van Antwerp, W. P.; Mastrototaro, J. J. Am. J. Physiol.Endocrinol. 1999, 277, E561-E571. (24) Boyne, M. S.; Silver, D. M.; Kaplan, J.; Saudek, C. D. Diabetes 2003, 52, 2790-2794. (25) Shoji, E.; Freund, M. S. J. Am. Chem. Soc. 2001, 123, 3383-3384 (26) Park, S.; Chung, T. D.; Kim, H. Ch. Anal. Chem. 2003, 75, 3046-3049.

Figure 1. Schematic showing operation of the wireless magnetoelastic glucose sensor. The sensor is placed within a single coil that serves to both interrogate and detect the sensor. A magnetic field impulse is used to excite the sensor, while the sensor excitation ring-down is analyzed to determine resonance frequency.

conversion of the applied magnetic field energy into elastic energy, inducing the sensor to mechanically vibrate. Since the magnetoelastic ribbon is also magnetostrictive, the mechanical vibrations, in turn, generate a magnetic flux density that extends above the sensor and can be detected by a remotely placed pickup coil.29 No direct physical connections to the sensor are needed, nor is any internal power source (i.e., battery) needed for sensor operation. Figure 1 is a schematic illustration of the send/receive telemetry of magnetoelastic sensors. The inherent passive, wireless, and remote query nature of the magnetoelastic sensors offers an outstanding opportunity for in situ and in vivo monitoring. For example, a low-cost (approximately $0.001/sensor) magnetoelastic pH sensor for in situ monitoring of gastric pH has recently been described.30,31 Magnetoelastic sensors have been developed for detection of various physical, chemical, and biological properties such as mechanical stress,32 pressure,33,34 viscosity,28,35 vapor analysis,36,37 and microbial analysis.38 The small sensor size (common sensor lengths are a few mm to a few cm) and low material cost (the material is purchased in ribbon form at a cost of ∼$200/km) make it suitable for medical applications where disposable sensors are of the utmost utility. In this paper, a wireless glucose sensor was fabricated using a 4 cm × 1.3 cm × 28 µm magnetoelastic ribbon, which was first coated with a pH-sensitive polymer followed by a second coating containing glucose oxidase. The enzymatic oxidation of glucose results in a decreasing pH that is detected by the pH-sensitive polymer. In response to the pH decrease, the pH-sensitive polymer shrinks, reducing the mass load on the magnetoelastic transducer, (27) Landau, L. D.; Lifshitz, E. M. Theory of elasticity, 3 rd ed.; ButterworthHeinemann: New York, 1986; Chapter II. (28) Stoyanov, P. G.; Grimes, C. A. Sens. Actuators, A 2000, 80 (1), 8-14. (29) Oguey, H. J. Rev. Sci. Instrum. 1960, 31, 701-709. (30) Cai, Q. Y.; Grimes, C. A. Sens. Actuators, B 2000, 71 (1-2), 112-117. (31) Ruan, C. M.; Ong, K. G.; Mungle, C.; Paulose, M.; Nickl, N. J.; Grimes, C. A. Sens. Actuators, B 2003, 96 (1-2), 61-69. (32) Germano, R.; Ausanio, G.; Iannotti, V.; Lanotte, L.; Luponio, C. Sens. Actuators, A 2000, 81, 134-136. (33) Grimes, C. A.; Stoyanov, P. G.; Kouzoudis, D.; Ong, K. G. Rev. Sci. Instrum. 1999, 70, 4711-4714. (34) Karl, W. J.; Powell, A. L.; Watts, R.; Gibbs, M. R. J.; Whitehouse, C. R. Sens. Actuators, A 2000, 81, 137-141. (35) Loiselle, K.; Grimes, C. A. Rev. Sci. Instrum. 2000, 71, 1141-1446. (36) Cai, Q. Y.; Cammers-Goodwin, A.; Grimes, C. A. J. Environ. Monit. 2000, 2, 556-560. (37) Cai, Q. Y.; Jain, M. K.; Grimes, C. A. Sens. Actuators, B 2001, 77 (3), 614619. (38) Ruan, C. M.; Zeng, K. F.; Varghese, O. K.; Grimes, C. A. Anal. Chem. 2003, 75, 6494-6498.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4039

Figure 2. Schematic of the sensor surface modification procedure for fabrication of glucose sensors.

which consequently results in an increase in resonance frequency of the magnetoelastic sensor. Owing to the specificity of the enzymatic reaction, the glucose sensor has a high degree of specificity. As needed, interfering analytes such as solution pH and salt concentration can be eliminated by cross-correlation with a reference sensor coated only with the pH-sensitive polymer. The described biosensor offers a potentially useful route for continuous in vivo or in situ glucose detection for diabetic patients by insertion of the sensor into, presumably, skin tissue while the detecting system can be packaged in a wristwatch-like form. Since no physical connections between the sensor and the detecting system are required, this device would be quite practical in use, as contrasted to electrochemical methods, and with a higher degree of specificity than optical glucose-level detection techniques. BACKGROUND The glucose sensor was fabricated by coating the magnetoelastic ribbon with pH-sensitive polymer, glucose oxidase, and catalase as shown in Figure 2. In the presence of glucose the sensor reactions are GOx

β-D-glucose + O2 + H2O 98β-D-gluconic acid + H2O2

where P-COOH denotes the pH-responsive polymer, which is electrically neutral in acidic solution. Polymer swelling is the result of interaction between the solvent (water) and polymer, with the extent of polymer swelling depending upon the polymer hydrophilicity and the degree of cross-linking. Linear polymers that are slightly cross-linked can experience significant swelling, while highly cross-linked polymers exhibit minimal swelling due to the covalent bonding of adjacent chains. Polymer hydrophilicity increases with increasing acrylic acid content. In alkaline solution, the polymer is electrically charged due to acrylic acid dissociation and additional swelling arises from the osmotic pressure exerted by the charged carboxylic group and its mobile counterions.39 Water molecules enter the polymer to lower the concentration of network ions until an equilibrium of osmotic pressure is achieved both inside and outside the polymer. The sensor transduction signal is derived from the polymer mass difference between its relatively shrunken state in acidic solution and relatively swollen state in alkaline solution. The addition of isooctyl acrylate can increase the polymer hydrophobicity, decreasing both the polymer solubility in water and its swelling ability in acidic solution.37 The shift in sensor resonance frequency ∆f as a function of mass can be expressed as28

catalase

H2O2 981/2O2 + H2O

∆f ) -

fr ∆m 2 M

(2)

The overall reaction becomes 1

GOx + catalase

β-D-glucose + /2O2 98β-D-gluconic acid The dissociation of gluconic acid produces H+, which diffuses to both the bulk solution and the pH-sensitive polymer resulting in polymer shrinking. The pH-sensitive polymer is a slightly crosslinked polyelectrolyte gel synthesized from acrylic acid and isooctyl acrylate monomers. Isooctyl acrylate was used to increase the hydrophobicity so that it is undissolvable in water. In addition, the branch alkyl group seems to enhance the ability of the polymer to swell. In response to pH change, acrylic acid dissociates according to the following equilibrium:

P-COOH (in acidic solution) T P-COO- + H+ (in alkaline solution) 4040

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

where M is the mass of the sensor and ∆m the mass change. Equation 2 shows that the sensitivity (∆f/∆m) is proportional to the frequency f and inversely proportional to the sensor mass. Smaller sensors have higher resonance frequencies, and lower masses, offering higher sensitivities. However, as the sensor size decreases, the detected signal amplitude decreases, making electronic signal processing issues more of a challenge per the lower signal-to-noise ratio. EXPERIMENTAL SECTION Materials. Acrylic acid, poly(ethylene glycol) diacrylate (PEGD), and isooctyl acrylate were purchased from Aldrich (Milwaukee, WI) and purified using an inhibitor remover disposable column (Aldrich) prior to use. 2,2′-Azobis(isobutyronitrile) (AIBN), glu(39) Khokhlov, A. R.; Starodubtzev, S. G.; Vasilevskaya, V. V. Adv. Polym. Sci. 1993, 109, 123-175.

Figure 3. Magnetoelastic sensor “reader” box. The sensor is placed within a vial containing the test solution, which is then placed inside the coil seen at the top of the box. The electronic analysis box is connected via a RS232 port to a computer for data display and storage.

cose, glutaric dialdehyde (50 wt % in water), dimethylaminopropyl3-ethylcarbodiimide (EDC), and N-hydroxysuccinimide (NHS) were purchased from Aldrich and used as received. Glucose oxidase (EC 1.1.3.4, 196 units/mg), catalase (EC 1.11.1.6, 2350 units/mg), and bovine serum albumin (BSA) from Sigma Co. were used as received. Bayhydrol 110, an anionic dispersion of an aliphatic polyester urethane resin in water/N-methyl-2-pyrrolidone solution (50% w/v) was purchased from Bayer Corp. (Pittsburgh, PA). Micro-Cleaning solution was purchased from International Products Co. (Burlington, NJ). Deionized and distilled water were used throughout the experiment. A 28-µm-thick ribbon of Metglas alloy 2826MB, alloy composition Fe40Ni38Mo4B18, was used as received from Honeywell Corp. The sensors, 4 cm × 1.3 cm rectangles, were laser-cut from the ribbon. The resonance frequency of an uncoated sensor in air is ∼54 kHz. Polymer Preparation. The pH-sensitive polymer is a copolymer synthesized by free radical copolymerization of acrylic acid and isooctylacrylate in dimethylformamide with an initial mole ratio of 4:1 at an overall monomer concentration of 4.5 mol/L. PEGD, 0.02 mol %, was added as cross-linker and 0.4 mol % AIBN was added as initiator. After 1-h bubbling of nitrogen to deoxygenate the solution, the temperature was raised to 65 °C to start the polymerization and maintained for 2 h to complete the polymerization under nitrogen atmosphere. The polymeric residue was washed with toluene repeatedly to remove any unreacted components. The resulting polymer was dried in a vacuum oven at 80°C under reduced pressure (10 Torr) overnight. The resulting polymer dissolves in ethanol and does not dissolve in water. Sensor Fabrication. The magnetoelastic sensors were ultrasonically cleaned in Micro-Cleaning solution, followed by a water and acetone rinse, and then dried in a stream of nitrogen. An ∼1µm-thick layer of Bayhydrol 110 was applied to the cleaned sensors by dip-coating in 40 wt % water/N-methyl-2-pyrrolidone solution. The thickness was estimated by weighing the sensor before and after coating using a microbalance (Sartorius Micro, Goettingen, Germany; resolution 0.1 µg), assuming a density of 1 g/cm3. The polyurethane-coated sensors were dried in air and then heated at 150 °C for 2 h to form a robust protective membrane, which offers a NH group for binding of the pH-sensitive polymer and protects the iron-rich magnetoelastic substrate from corrosion. The poly-

urethane-protected sensor was then coated with 50 µL of 23 g/L in ethanol pH-sensitive polymer solution, which contains 0.46 g/L EDC and 0.38 g/L NHS. The mole ratio of NHS and EDC to acrylic acid monomer is 2 and 1%, respectively. The polymercoated sensors were dried in air and then heated in a vacuum oven at 120 °C. Activated by EDC and NHS, the pH polymer was bonded on the polyurethane coating through amide bonding. The pH-sensitive polymer-coated sensor can be used as reference sensor as is or further coated to achieve the working glucose biosensor. To fabricate the working sensor, the sensor was first dipped in 0.02% EDC and NHS solution in water and then rinsed with water. Finally 0.5 mL of solution containing 5 g/L GOx, 0.5 g/L catalase, 12.5 g/L BSA, and 0.5% glutaric dialdehyde in water was prepared, and 60 µL of the solution was applied on the both sides of the pH polymer-coated sensor. The resulting sensor was dried in air at room temperature. After each coating step, the sensor was weighed using a microbalance and the coating mass determined. The typical coating weight (in mg/cm2) is, polyurethane ∼0.1, pH polymer 0.11, and biofilm 0.10. The enzyme amount (in active unit/cm2) is, GOx 5.4 and catalase 6.5. The sensor-coating steps are illustrated in Figure 2. Setup and pH Calibration. The system electronics used for determining sensor resonance frequency is shown in Figure 3; the sensors are inserted into the coil seen at the top of the box. Only one coil is needed for sensor excitation and detection. The sensors were placed into a vial containing 1.8 mL of test solution; the container volume is 10 mL. The electronics are capable of resonance frequency determination through frequency counting, damping ratio measurement, and spectrum analysis up to 450 kHz.40 Measurement of resonance frequency takes ∼5 ms. A computer, connected to the electronics through a RS232 port, coordinates system operation, data display, and data storage. Prior to use, the sensor was activated in 0.01 mol/L pH 7 PBS buffer solution for 20 min to wet the polymer. The sensor was then immersed in the test solution with various glucose concentrations. Sensor resonance frequency was recorded as a function of immersion time. After each measurement, the sensor was rinsed (40) Zeng, K. F.; Ong, K. G.; Mungle, C. M.; Grimes, C. A. Rev. Sci. Instrum. 2002, 73, 4375-4380.

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4041

Figure 5. Effect of salt concentration on the sensor sensitivity. Figure 4. Effect of glutaric dialdehyde concentration on sensor sensitivity and stability in 0.1mol/L NaCl over a period of 25 days. Sensor coating contains 0.3 mg of GOx, 0.03 mg of catalase, 0.75 mg of BSA, and variable amounts of glutaric dialdehyde as indicated in the legend.

and placed in 0.01 mol/L pH 7 PBS buffer solution for storage. All measurements were performed at room temperature (23 ( 1 °C). RESULTS AND DISCUSSION Sensor Stability. In addition to a high degree of specificity and accuracy, in vivo or in situ glucose sensors require excellent long-term stability. Glucose oxidase is a relative stable enzyme remaining intact for at least 3 months at room temperature and 6 months if stored in a refrigerator (∼1 °C). The key to achieving a stable glucose sensor is enzyme immobilization. As illustrated in Figure 2, the enzymes were chemically bonded on BSA through glutaric dialdehyde to form a biofilm layer that was bonded to the pH polymer through the amide bond. A higher degree of bonding results in greater sensor stability but at the cost of decreased sensitivity. For the pH-responsive polymer, the carboxylic groups provide bonding sites; with increasing EDC and NHS concentration, the pH sensitivity decreases without a corresponding increase in polymer stability. A satisfactory stability can be achieved at the molar ratio of 2 and 1% of EDC and NHS to acrylic acid monomer, respectively, with a pH sensitivity loss of ∼20%. Figure 4 shows the effect of glutaric dialdehyde concentration on sensor sensitivity and stability. To support the enzyme, a factor of 2.5 times BSA and a small excess of catalase was coimmobilized to catalyze the decomposition of H2O2 (reaction 2). The 10:1 mass ratio of GOx to catalase was used with an activity unit ratio of 4:5. The stability was measured by observing the sensitivity change over a period of 20 days. The sensitivity was calculated from the frequency change during the first 60 min of testing upon which a relative stable response was achieved. The results show that with increasing glutaric dialdehyde concentration the sensitivity decreases, with an increase in sensor stability that then saturates. Satisfactory results can be obtained by using 0.05% glutaric dialdehyde, equivalent to 0.3 mg on the sensor coating. Effect of Electrolyte. Increasing salt concentration results in the pH-sensitive polymer shrinking due to the effect of osmotic 4042 Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

Figure 6. Time-dependent sensor response profile as a function of glucose concentration. The sensor coating contains 0.3 mg of GOx, 0.03 mg of catalase, 0.75 mg of BSA, and 0.3 mg of glutaric dialdehyde.

pressure, which reduces the swelling ability of the polymer and consequently the pH sensitivity. The sensor sensitivity to glucose decreases with increasing salt concentration,30 as shown in Figure 5. The sensitivity in 0.1 mol/L NaCl solution is ∼60% of that in 0.01 mol/L NaCl solution. However, the response is more sensitive to the salt concentration change at low concentrations than at high concentrations. A concentration change of 0.01mol/L NaCl to 0.02 mol/L NaCl results in a 12% change in sensor sensitivity. However, a concentration change of 0.1 to 0.11 mol/L NaCl results in a 4.5% change in sensor sensitivity. This result is favorable to biological glucose detection since the physiological salt concentration is approximately equivalent to 0.15 mol/L NaCl. In such a salt concentration, small variations from the baseline have little effect on the sensitivity of the sensor. In addition, the effect is only dependent upon salt concentration, not the salt species. Glucose Detection. Under normal physiological conditions, serum glucose concentration is 3.8-6.1 mmol/L. Diabetic urine appears when blood glucose is generally higher than 9 mmol/L, the renal glucose threshold. Therefore, we calibrated our sensor to glucose concentrations ranging from 1 to 15 mmol/L while to imitate physiological conditions, 0.15 mol/L NaCl was added. The time-dependent response profiles of an illustrative glucose sensor are shown in Figure 6. The calibration curve for an illustrative

Figure 7. Frequency shift to glucose concentration calibration curve.

sensor is shown in Figure 7, with the sensitivity calculated from the frequency shift achieved during the first 60 min. The response is linear from 1 to 10 mmol/L glucose, becoming sublinear at higher concentrations. Defining the detection limit as the concentration that can be detected at 3 times the noise level, the detection limit is 0.6 mmol/L glucose within 1-10 mmol/L, given a system noise level of 20 Hz, which corresponds to the precision at which the resonance frequency can be determined.

CONCLUSIONS A novel wireless glucose biosensor is described, based on a mass-sensitive magnetoelastic sensor coated with a pH-sensitive polymer upon which is coated a bienzymatic system containing GOx and catalase. The enzymatic oxidation of glucose results in a decreasing pH that is sensed by the pH-responsive polymer and results in polymer shrinking. The polymer shrinking decreases the mass loading on the sensor, and as a result, the sensor resonance frequency increases. The response of the biosensor to changing glucose concentrations is reversible and linear from 1 to 10 mmol/L. Sensitivity decreases with increasing ionic strength; in 0.15 mol/L NaCl, the sensitivity is 55% of that in 0.001 mol/L NaCl. However, at physiological salt concentration, ∼0.15 mol/ L, the glucose sensitivity of 0.1 mol/L is high enough for glucose detection in diabetic patients. Since no physical connections between the sensor and the monitoring electronics are required, and the sensor is entirely passive without requiring an internal power supply, this sensor platform offers an exciting opportunity for developing a useful in vivo and in situ glucose measurement technology. ACKNOWLEDGMENT Support of this work by the National Institutes of Health under Contract 1-R21570-01 is gratefully acknowledged. Received for review January 26, 2004. Accepted April 15, 2004. AC0498516

Analytical Chemistry, Vol. 76, No. 14, July 15, 2004

4043