Fibroblast Cells: A Sensing Bioelement forGlucose Detection by Impedance Spectroscopy

Chaker Tlili,*,†,‡ Karine Reybier,† Alain Geloe1n,§ Laurence Ponsonnet,†Claude Martelet*,† Hafedh Ben Ouada,§ Michel Lagarde,§ and Nicole Jaffrezic-Renault*,†

Laboratoire d’Ingenierie et de Fonctionnalisation des Surfaces, UMR-CNRS 5621, ECL-Lyon, 36 Avenue Guy deCollongues, 69134 Ecully Cedex, France, Laboratoire U352 Biochimie et Pharmacologie de la mediation lipidique,INSERM-CNRS, 20 Avenue Albert Einstein, 69621 Villeurbanne Cedex, France, and Laboratoire de Physicochimie desInterfaces, Universite de Monastir, Avenue de l’Environnement, 5019 Monastir, Tunisie

Modifying the electrical properties of fibroblasts againstvarious glucose concentrations can serve as a basis for anew, original sensing device. The aim of the present studyis to test a new biosensor based on impedancemetrymeasurement using eukaryote cells. Fibroblast cells weregrown on a small optically transparent indium tin oxidesemiconductor electrode. Electrochemical impedancespectroscopy (EIS) was used to measure the effect ofD-glucose on the electrical properties of fibroblast cells.Further analyses of the EIS results were performed usingequivalent circuits in order to model the electrical flowthrough the interface. The linear calibration curve wasestablished in the range 0-14 mM. The specification ofthe biosensors was verified using cytochalasin B as aninhibitor agent of the glucose transporters. The nonreac-tivity to sugars other than glucose was demonstrated.Such a biosensor could be applied to a more fundamentalstudy of cell metabolism.

A global epidemic of diabetes is predicted for the first quarterof the 21st century. In the United States in 1998, ∼6.5% of thepopulation (over 16 million people) were diagnosed with type 2diabetes and a roughly equivalent number was estimated to beundiagnosed.1 Recent evidence reveals that the prevalence ofdiabetes in the United States increased by 33% from 1990 to 1998with a further 6% increase over the following year. These statisticsare not confined to the United States. The worldwide prevalenceof diabetes in persons g20 years old was 4% in 1995 and isexpected to reach 5.4% (over 300 million people) by 2025. Thismeans that diagnosing diabetes has become a major issue.

Diabetes is diagnosed exclusively by demonstrating a persis-tent increased glucose concentration in the blood. Continuousglucose sensing with a stable in vivo glucose sensor is expectedto lead to improved regulation of the glucose concentration andthus to a reduction of the number of complications in diabetes

mellitus patients. Since Clark and Lyons2 reported their enzymeelectrode for measuring glucose in 1962, an enormous amount ofliterature has been published concerning enzyme-based biosen-sors. Amperometric biosensors based on glucose oxidase are themost widely used biosensors. This enzyme catalyses the oxidationof â-D-glucose consuming molecular oxygen, producing glucono-lactone, gluconic acid, and hydrogen peroxide. Biosensor responseis generally based on the electrochemical oxidation of H2O2.3,4 Thiskind of biosensor is often used in short-term evaluation5,6 onlydue to an activity loss of the enzyme and the relative toxicity ofH2O2 for biological entities.

Several studies on glucose sensing by numerous investigatorshave aimed at developing an implantable glucose sensor,7-10 butto obtain reliable glucose measurements in vivo, the sensor musthave an outer coating in order to prevent contact between tissueand the enzymatic membrane,10,11 and its lifetime does not usuallyexceed 3 months. This is why the development of new biosensorsis of considerable interest, and microorganisms are good candi-dates. Some authors have proposed to replace the enzyme withfragments of yeast cell wall,12 since bacteria and yeasts arerecognized as organisms that metabolize all kinds of sugars verywell.13,14 However, to improve the selectivity of such biosensors,there must be use a second anti-interference enzymatic-layer.

The use of living cells as sensor elements provides theopportunity for high sensitivity in a broad range of biologicallyactive substances that affect the response of the cells. It has long

* Corresponding authors. E-mail: Claude.Martelet@ec-lyon.fr; chaker_tlili@yahoo.fr.

† UMR-CNRS 5621.‡ INSERM-CNRS.§ Universite de Monastir.

(1) Greenberg, R. A.; Sacks, D. B. Clin. Chim. Acta 2002, 315, 61.

(2) Clark, L. C.; Lyons, C. Ann. N. Y. Acad. Sci. 1962, 102, 29.(3) Cosnier, S.; Senillou, A.; Gratzel, M.; Comte, P.; Vlachopoulos, N.; Jaffrezic-

Renault, N.; Martelet, C. J. Electroanal. Chem. 1999, 469, 176.(4) Poyard, S.; Jaffrezic-Renault, N.; Martelet, C.; Labbe, P.; Besombes, J. L.;

Cosnier, S. Sens. Actuators, B 1996, 33, 44.(5) Ward, W. K.; Jansen, L. B.; Anderson, E.; Reach, G.; Klein, J. C.; Wilson, G.

S. Biosens. Bioelectron. 2002, 17, 181.(6) Wilkins, P.; Atanasov, B.; Muggenburg, A. Biosens. Bioelectron. 1995, 10,

485.(7) Schmidtke, D. W.; Heller, A. Anal. Chem. 1998, 70, 2149.(8) Pickup, J. C. Diabetes Cares 1999, 16, 535.(9) Kros, A.; Gerritsen, M.; Sprakel, V. S. I.; Sommerdijk, N. A. J. M.; Jansen,

J. A.; Nolte, R. J. M. Sens. Actuators, B 2001, 81, 68.(10) Galeska, I.; Chattopadhyay, D.; Moussy, F. Biomacromolecules 2000, 1,

202.(11) Quinn, C. A. P.; Connor, R. E.; Heller, A. Biomaterials 1997, 18, 1665.(12) Barlıkova, A.; Svorc, J.; Miertus, S. Anal. Chim. Acta 1991, 247, 83.(13) Rotariu, L.; Bala, C.; Magearu, V. Anal. Chim. Acta 2002, 458,

215-222.(14) Svitel, J.; Curilla, O.; Tkac , J. Biotechnol. Appl. Biochem. 1998, 27, 153.

Anal. Chem. 2003, 75, 3340-3344

3340 Analytical Chemistry, Vol. 75, No. 14, July 15, 2003 10.1021/ac0340861 CCC: $25.00 © 2003 American Chemical SocietyPublished on Web 06/03/2003

been recognized that the membrane of biological materials thatinclude cells exhibits dielectric properties. For example, whenculturing cells over electrode contacts, any change in the effectiveelectrode impedance permits an assay of cultured cell adhesion,spreading, and mortality.15-20 Impedance measurements rely onthe observation that living cells have excellent insulating propertiesat low frequencies. As cells grow or migrate to increase coverageover an electrode surface, the effective electrode impedance rises.Impedance measurements have been used to monitor the behaviorof an array of nonexcitable cell types including macrophages,16

endothelial cells,17 and fibroblasts.18

This paper describes a new fibroblast-based sensing devicefor the determination of glucose using impedance spectroscopy.Indeed, 3T3-L1 fibroblasts are able to metabolize glucose throughthe activation of specific glucose transporters (Glut 1 and Glut 4).

EXPERIMENTAL SECTIONMaterials. Glass slides coated with 110-nm-thick indium tin

oxide (ITO) films (R < 20 Ω/cm2) were purchased from MerckDisplay Technologies. Ultrapure water was used throughout thisstudy (Millipore-Q-Systems; R > 18 MΩ/cm), and all otherchemical reagent were purchased from Aldrich.

ITO Electrodes. The active area of these electrodes was 0.07cm2. Before cell culture, ITO electrodes were carefully cleanedby ultrasonication, twice in 2% Fluka cleaning (ref 61257) solutionfor 30 min and then twice in ultrapure water for 30 min. Eachsonication step was followed by 10 rinsing cycles in ultrapurewater. ITO surfaces were dried under a nitrogen flow. The ITOtreated with cleaner presents sufficient hydrophilicity for cellattachment (θeau ) 60°).19 The sterilization of the probes wasperformed using an autoclave at 137 °C and 2.4 bar for 30 min.

Fibroblast Cell Monolayer. 3T3-L1 fibroblasts were obtainedfrom American Type Culture Collection (Manassas, VA). The cellswere grown in Dulbecco’s modified Eagle’s medium (DMEM)supplemented with 10% newborn calf serum, 4 mM glutamine, 4nM insulin (Actrapid Human, Novo), 10 mM Hepes, and with 25µg of sodium ascorbate, 100 IU of penicillin, 100 µg of strepto-mycin, and 0.25 µg of amphotericin B per milliliter at 37 °C inwater-saturated atmosphere at 5% CO2 in air, in a Heraeusincubator (BB16). Probes were used 2-4 days after cell additiononce they had reached the confluent state. Under these conditions,cells formed a thin regular monolayer. For impedance measure-ments, the medium was replaced by DMEM to which increasingconcentrations of glucose were added.

Impedance Spectroscopy. Impedance spectroscopy wasrecorded using a Voltalab40 (Radiometer Analytical SA) imped-ance analyzer, which provides the perturbation sine wave voltagesignals, records the current response of the system, and calculatesthe complex impedance in polar coordinates |Z(ω)| and |æ(ω)|.The experimental setup is shown in Figure 1. Because of the

polarizable properties of ITO, a reference electrode was notrequired for Ubias ) 0 mV, but in the case of the Mott-Schottkyanalysis, where different bias potentials are necessary, the three-electrode setup is applied. We used sine waves between a smallITO electrode (0.07 cm2) and a large ITO counter electrode (0.385cm2) with an amplitude of (5mV (peak to peak) for frequenciesin the range 10-2-105 Hz at five fixed values of the frequencythat were equidistant on the logarithmic scale. The impedancedata are evaluated by fitting (nonlinear least squares) theparameters of appropriate equivalent circuits to the experimentaldata using the Equivcrt.Pas (EQU) (version 4.51) program writtenby Boukamp.25

To relate the measured complex impedance to the electricalbehavior of the system, the semiconductor, the cells, and theelectrolyte were represented by simple electrical elements inequivalent circuits. As demonstrated previously,26,27 each elementcontributes predominantly to different frequency regimes, whichallows separation of the effects of the different components of theelectrical interface.

However, the impedance spectra derived from equivalentcircuits composed of ideal elements, like capacitances or resis-tance, do not fit the measured spectra in a satisfactory way becauseof inhomogeneities of surface. To overcome this problem, it ispossible to replace the ideal elements by constant-phase elements(CPEs) CCPE,28,29 which account for the nonlinearities and the

(15) Giaever, I.; Keese, C. R. IEEE Trans. Bio-Med. Eng. 1986, 33, 242.(16) Mitra, P.; Keese, C. R.; Lawrence, D. A.; Giaever, I. Biotechniques 1990,

11, 504.(17) Xiao, C.; Lachance, B.; Sunahara, G.; Luong, J. H. T. Anal. Chem. 2002,

74, 1333.(18) Luong, J. H. T.; Habibi-Razaei, M.; Meghrous, J.; Xiao, C.; Male, K. B.;

Kamen, A. Anal. Chem. 2001, 73, 1844.(19) Ehret, R.; Baumann, W.; Brischwein, M.; Schwinde, A.; Stegbauer, K.; Wolf,

B. Med. Biol. Eng. Comput. 1998, 36, 365.(20) Ehret, R.; Baumann, W.; Brischwein, M.; Schwinde, A.; Stegbauer, K.; Wolf,

B. Biosens. Bioelectron. 1997, 12, 29.

(21) Kowolenko, M.; Keese, C. R.; Lawrence, D. A.; Giaever, I. J. Immunol.Methods 1990, 127, 71.

(22) Hillebrandt, H.; Abdelghani, A.; Abdelghani-Jacquin, C.; Aepfelbacher, M.;Sackmann, E. Appl. Phys. A 2001, 73, 539.

(23) Giaever, I.; Keese, C. R. Nature 1993, 366, 591.(24) Lee, J. H.; Lee, S. J.; Khang, G.; Lee, H. B. J. Biomater. Sci. Polym. Ed.

1999, 10, 283.(25) Boukamp, A. B. Equivalent circuit users manual. Report CT88/265/128.

Department of Chemical Technology, University of Twente, Twente, TheNetherlands, 1989.

(26) Stelzle, M.; Weissmuller, G.; Sackmann, E. J. Phys. Chem. 1993, 97,2974.

(27) Gritsch, S.; Nollert, P.; Jahnig, F.; Sackmann, E. Langmuir 1998, 14, 3118.(28) MacDonald, J. R. Impedance spectroscopy, Emphasing solid materials; John

Wiley and Son: New York, 1987.(29) Omanovic, S.; Metikos-Hukovis, M. Thin Solid Films 1995, 266, 31.

Figure 1. Schematic of the experimental setup. The impedanceanalyzer is controlled by a computer and generates the perturbationac voltage, which is applied between the two ITO electrodes. Thecurrent response of the system is measured, and the analyzercalculates the complex impedance Z.

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003 3341

frequency dependence of the elements by expressing the electricalimpedance in terms of a simple power law.

For ideal elements, the frequency exponent is R ) 1 and K )1/C for a capacitance and R ) 0 and K ) R for a resistance,respectively. The deviation of the exponent R from the ideal valuesis attributed to the inhomogeneities of the analyzed layer. By usingof the CPEs, the quality of the fitting of the experimental data byequivalent circuit models can be dramatically improved.

Medium for Experiments. Serum has the disadvantage ofbeing made up of numerous components whose compositionfluctuates, which renders its use difficult for reproducible experi-ments. For adherent cells, the intensity of adhesion is a veryimportant test factor, which is largely dependent upon themedium. Experiments concerning the influence of serum com-ponents have already been performed with an interdigitatedelectrode structure.20 In this study, we have chosen to performimpedance analysis in DMEM, to improve reproducibility ofmeasurements.

RESULTS AND DISCUSSIONEffect of Biological Cells on Interface Impedance. Biologi-

cal cells are very poor conductors at low frequencies (at leastbelow 10 kHz), and therefore, force electrical currents to bypassthem. As cells grow adherently on the ITO electrode, this reducesthe available electrode area, and the interface impedance isincreased as it is inversely proportional to the total electrode areareached by the current.30 The gap between the cell membraneand the substratum limits the influence of the cell membranecapacity on the interface impedance of the electrodes.

Impedance Analysis of the Electrodes/Culture MediumSystem without Cells. The Bode diagram of a bare semiconduc-tor ITO electrode, immersed in culture medium DMEM (corre-sponding to an electrolyte/semiconductor interface), is shown inFigure 2a. The module Z showed a plateau region at 105-103 Hzand increased linearly to 10-2 Hz. The phase shift æ decrease from105 to101 shows a constant in the vicinity of -90° from 10 to 10-2

Hz. The linear increase of |Z(ω)| from 103 to 10-2 Hz indicates asingle capacitance corresponding to the semiconductor/electrolyteinterface. The impedance spectrum of this system can thereforebe analyzed using an equivalent circuit that consists of a resistorand a CPE (replacing the capacitor) in series. The resistor R0 ismainly due to the conductivity of the bulk solution and the wireconnection whereas the CCPE represents the dielectric propertiesof the electrode/electrolyte and additional qualitative informationon the surface morphology of the ITO electrode. Applying a fittingprocedure to the Bode diagram of Figure 1a yielded the value ofR0 ) 33.8 ( 0.7 Ω cm2, CCPE ) 24.0 ( 0.3 µF cm-2, and R ) 0.929( 0.005.

Impedance Spectra of Fibroblast Cell Monolayer Grownon ITO Electrode. The influence of cells on the impedance signalis clearly shown in Figure 2b, which presents the Bode diagramwith fibroblast cell monolayer 3 days after confluence. The

presentation of the data in the form of a Bode impedance plotclearly reveals the presence of three separate frequency-dependentsegments. In the high-frequency region, the absolute impedancecurve is almost quasi-independent of frequency, when the phaseangle value approaches 0°. This is a typical response for a resistivebehavior, which corresponds to a resistance of the DMEMelectrolyte between the working electrode and the referenceelectrode and to the existence of a cell monolayer on the smallelectrode ITO. There is not a great difference between a bareelectrode and an electrode bearing a cell layer, since the junctionsbetween the cell are very weak.31 In the intermediate frequencies,a linear relationship can be observed between the absoluteimpedance and the frequency with a slope of -0.8 and a phaseangle of ∼-80° with a minimum at 10 Hz. This frequency rangecorresponds to a capacitive behavior of the electrode/electrolyteinterface. In the lower frequency region, there is a change in theslope of the absolute impedance and the phase angle presentsanother minimum after 4 × 10-2 Hz, indicating an increase in theresistive behavior of the electrode, relative to a capacitive behavior.

The impedance spectrum is interpreted in terms of theequivalent circuit shown in the inset of Figure 2b, where R0 )

(30) Schwan, H. P. In Determination of biological impedance in physical techniquesin research, Electrophysiological Methods, part B; Nesluk, W. L., Ed.; AcademicPress: New York, 1963; Vol. 6, p 323.

(31) Wegener, J.; Sieber, M.; Galla. H, J. J. Biochem. Biophys. Methods 1996,32, 151.

ZCPE ) Kω-R (1)

Figure 2. (a) Impedance spectrum of bare ITO electrode in contactwith DMEM electrolyte. The squares represents the measured data,while the solid line is the theoretical spectrum of the equivalent circuitinside the figure. (b) Impedance spectrum of ITO electrode coveredwith a confluent monolayer of fibroblast cells (3T3-L1).

3342 Analytical Chemistry, Vol. 75, No. 14, July 15, 2003

34.4 ( 0.7 Ω cm2, CCPE ) 68 ( 1 µF cm-2, and R ) 0.871 ( 0.004,but R ) 6.5 ( 0.3 kΩ cm2 and C ) 45 ( 2 µF cm-2 are the resistorand the capacity of the layer of the adsorbed proteins32 on theoxide surface and Rc ) 5.425 ( 0.720 Ω cm2 and Cc ) 0.76 ( 0.24µF cm-2 corresponds to the cell layer.

Effect of Glucose. Figure 3 shows the effect of glucoseaddition on the electrical properties of the fibroblast cell monolayerin terms of impedance spectra. In this case, the Nyquist diagramrepresentation was preferred since such a representation allowsa more sensitive analytical treatment. Analyses have been per-formed in a wide range of glucose concentrations from 1 to 23mM, which is wider than the physiological range to test theresponsiveness of our biosensor. Each glucose addition induceda change of the Nyquist diagram, which comes from a changeboth in the resistance and in the capacity of the cell monolayer.This variation is correlated to the change of the dielectricproperties σm and εm of cellular membrane under the effect ofD-glucose.33 As no impedance variation was recorded after glucoseadditions in the case of a bare ITO electrode, the variationsrecorded are due to an “interaction” between the cell monolayerand glucose molecules. This interaction could correspond to theglucose uptake through Glut 1 and Glut 4 transporters presenton the plasma membrane of the fibroblasts.

Figure 4 shows that, at glucose concentrations below 14 mM,a typical calibration graph of the absolute impedance response at4 kHz of fibroblast/ITO electrode versus glucose concentrationis produced. But when the concentration goes over 14 mM, thesystem reaches a saturation, as can be clearly seen from the figure.To correlate this variation of impedance to fibroblast behavior,the fitting procedure with the previous equivalent circuit, definedwithout glucose, is applied to impedance variation as a functionof glucose concentration. Values of R0, CCPE, R, and C, remainingconstant in the equivalent circuit, values of Rc and Cc componentsvary as a function of glucose concentration as shown in Figure 5.The error bars include the experimental error on impedancemeasurements and the discrepancy of experimental values related

to the model equivalent circuit. The general evolution of Rc andCc shows that they translate the fibroblast behavior, and asaturation effect is observed for both components for concentra-tions higher than 14 mM, which corresponds to the saturationeffect on the experimental curve (cf. Figure 4).

To highlight the specificity of the cells toward glucose and toreinforce the hypothesis of an impedancemetric signal correlatedto glucose uptake by fibroblasts, the same experiments have beenperformed in the presence of D-mannitol instead of glucose, asweetener not metabolized by the cells. As shown in the Nyquistdiagram presented in Figure 6, no variation was recorded forD-mannitol additions in the range from 1 to 15 mM. Consequently,the changes in capacity and resistance of the cell monolayerobserved in the presence of glucose (Figure 3) result from theincorporation of glucose in the fibroblast monolayer itself and notfrom indirect changes induced by the increased concentrationsof glucose such as, for example, increased osmotic pressure.

Inhibition of Glucose Uptake. To strengthen our hypothesis,we inhibited the glucose transporters of the 3T3-L1 fibroblasts.This was achieved by using cytochalasin B, which inhibits Glut

(32) Hiromoto, S.; Noda, K.; Hanawa, T. Corros. Sci. 2002, 44, 955.(33) Caduff, A.; Hayashi, Y.; Livshits, L.; Feldman, Y. 2nd International Conference

on Broadband Dielectric Spectroscopy and Its Applications, Leipzig,Germany, September 2-6, 2002.

Figure 3. Nyquist diagram of the effect of D-glucose on confluentmonolayer of fibroblast cells.

Figure 4. Calibration plot of the glucose biosensor. The signal islinear from 0 to 14 mM glucose concentration and reaches saturationfor higher concentration. Error bars were determined from threeexperimental results on the same electrode.

Figure 5. Variation of the Rc (9) and Cc (b) versus D-glucoseconcentration with calculated error bars according to the equivalentcircuit used to interpret the experimental variations of Z modulusversus glucose concentration.

Analytical Chemistry, Vol. 75, No. 14, July 15, 2003 3343

1- and Glut 4-mediated glucose uptake.34 The cell monolayer wasexposed for 10 min to 10-4 M cytochalasin B. Then cytochalasinB was removed and replaced by a DMEM solution to whichincreasing quantities of glucose were added. Figure 7 presentsthe Nyquist diagram after inhibition by cytochalasin B and glucoseadditions (10 and 20 mM). After glucose additions, the curvesare superimposed, which indicates that the resistance and thecapacitance of the cell monolayer are not modified and conse-quently that glucose is no larger incorporated into the cells. Theseresults demonstrate that the changes in the resistance and in thecapacity of the cell monolayer recorded in Figure 2 directly resultfrom glucose uptake and metabolization of glucose by the cells.

CONCLUSIONSThe present study shows for the first time the feasibility of

glucose sensing by fibroblasts over the physiological range ofglucose concentrations, using impedancemetry analysis. Indeed,

the addition of a carbohydrate not metabolized by cells did notgenerate any change in the impedance parameters of the sensor.Furthermore, inhibiting glucose uptake by cytochalasin B, aspecific inhibitor of glucose transporters, completely abolishedthe signals relating to the modification of glucose concentration.Further studies, over a wider frequency range, are under way inorder to improve the intensity of the performance of the sensingdevice. Subsequently, this system could furnish fresh insights fora better understanding of glucose metabolism in cells.

ACKNOWLEDGMENTThe authors acknowledge financial support from MIRA Re-

search Program and valuable discussion with Professor BelcourtAlain, Vice-President and Director Centre Europeen D’Etude duDiabete.

Received for review January 29, 2003. Accepted April 25,2003.

AC0340861(34) Karnieli, E.; Zaenowski, M. J.; Hissin, P. J.; Simpson, I. A.; Salans; L. B.;

Cushman, S. W. J. Biol. Chem. 1981, 256, 4772.

Figure 6. Nyquist diagram of the effect of D-mannitol on confluentmonolayer of fibroblast cells.

Figure 7. Nyquist diagram of the inhibition of glucose uptake bycytochalasin B on confluent monolayer of fibroblast cells.

3344 Analytical Chemistry, Vol. 75, No. 14, July 15, 2003