Biosensors Based on Membrane-Bound Enzymes Immobilized in a 5

Gold electrodes were modified through chemisorption of 5-(octyldithio)-2-nitrobenzoic acid (ODTNB). ODTNB includes a long chain in a short-length thio...
0 downloads 8 Views 183KB Size
Download PDF
Anal. Chem. 2000, 72, 3784-3792

Biosensors Based on Membrane-Bound Enzymes Immobilized in a 5-(Octyldithio)-2-nitrobenzoic Acid Layer on Gold Electrodes M. Darder, E. Casero, F. Pariente, and E. Lorenzo*

Departamento de Quı´mica Analı´tica y Ana´ lisis Instrumental, Universidad Auto´ noma de Madrid, Canto Blanco 28049, Madrid, Spain

Gold electrodes were modified through chemisorption of 5-(octyldithio)-2-nitrobenzoic acid (ODTNB). ODTNB includes a long chain in a short-length thio acid, providing a heterogeneous-like alkanethiol layer after adsorption on gold electrodes. Membrane-bound enzymes, in particular D-fructose dehydrogenase (FDH), D-gluconate dehydrogenase (GADH), and L-lactic dehydrogenase (cytochrome b2) (Cyb2), were immobilized onto ODTNB-modified gold electrodes simply by adsorption. The short-length thio acid may provide electrostatic interactions with enzyme surface charges, while the alkanethiolate enables hydrophobic interaction with the largely lipophilic, membranebound enzymes. The immobilization of FDH, GADH, and Cyb2 onto ODTNB-modified gold surfaces has been studied with the quartz crystal microbalance (QCM). Spectrophotometric and electrochemical assays indicate that the immobilized enzyme retains its enzymatic activity after immobilization onto the ODTNB-modified gold surface. The amount of immobilized (and active) enzyme was estimated from QCM to be of the order of 2.5 × 10-125.3 × 10-12 mol‚cm-2. A fructose biosensor was developed, making use of a gold surface modified with ODTNB and fructose dehydrogenase, employing hydroxymethylferrocene as a mediator in solution. Calibration curves exhibited a linear relation between the biosensor response and the substrate concentration up to 0.7 mM. Statistical analysis gave an excellent linear correlation (r ) 0.9993) and a sensitivity of 6.1 mM-1 fructose. The biosensor shows a significant stable catalytic current for at least 25 days. One of the greatest obstacles to the success of biosensors is the inability to manufacture reproducible devices. Theoretical models of the classical geometry enzyme electrodes, where the enzyme is immobilized in a three-dimensional reaction matrix placed over a planar electrode, show that the response of the sensor is very sensitive to the thickness of the reaction matrix.1-6 (1) Staros, J. V.; Wright, R. W.; Swingle, D. M. Anal. Biochem. 1986, 156, 220-222. (2) Leypoldt, J. K.; Gough, D. A. Anal. Chem. 1984, 56, 2896-2904. (3) Martens, N.; Hall, E. A. H. Anal. Chem. 1994, 66, 2763-2770. (4) Schulmeister, T. Sel. Electrodes Rev. 1990, 12, 203-260. (5) Parker, J. W.; Schwartz, C. S. Biotech. Bioeneg. 1987, 30, 724-735.

3784 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

An alternative geometry enzyme electrode, with a response independent of the thickness of the reaction layer, was developed by Gooding and Hall.6 However, this geometry is not a generic solution, since it is too slow for some applications. An obvious conclusion of this would be that, in order to obtain better reproducibility and shorter response time, a two-dimensional reaction zone, where the recognition event is a surface reaction, would, perhaps, be a preferred solution. However, for any twodimensional reaction layer method to be successful, it also requires that the layer should be constructed in a highly reproducible manner, which suggests that self-assembled layers of alkanethiols could have great potential for enzyme electrodes. The bonding of alkanethiols to metal surfaces has been studied extensively7-9 and shown to produce stable monolayers with a high degree of order. Gold electrodes modified with alkanethiols have been used as anchors to immobilize biological molecules in order to construct both enzyme biosensors10-18 and immunosensors.19,20 The use of alkanethiols to construct enzyme electrodes often requires functionality at the terminal end of the molecule to allow the formation of a bond between the enzyme and the monolayer. In many of these approaches described in the literature, the alkane end terminates with an amine group and the biological molecule is immobilized using a cross-linker, usually, glutaraldehyde.21 However, the cross-linking with glutaraldehyde creates multilay(6) Gooding, J. J.; Hall, E. A. H. Electroanalysis 1996, 8, 407-413. Gooding, J. J.; Hall, E. A. H. J. Electroanal. Chem. 1996, 417, 25-33. (7) Ulman, A. An Introduction to Ultrathin Organic Films From LangmuirBlodgett to Self-Assembly; Academic Press: London, 1991. (8) Bain, C. D.; Whitesides, G. M. Science 1988, 240, 62-63. (9) Pan, W.; Durning, C. J.; Turro, N. J. Langmuirr 1996, 12, 4469-4473. (10) Kajiya, Y.; Okamoto, T.; Yoneyama, H. Chem. Lett. 1993, 2107-2110. (11) Willner, I.; Lion-Dagan, M.; Marx-Tibbon, S.; Katz, E. J. Am. Chem. Soc. 1995, 117, 6581-6592. (12) Imamura, M.; Haruyama, T.; Kobakate, E.; Ikariyama, Y.; Aizawa, M. Sens. Actuators 1995, B24-B25, 113-116. (13) Creager, S. E.; Olsen, K. Anal. Chim. Acta 1995, 307, 277-289. (14) Dong, X. D.; Lu, J.; Cha, C. Bioelectrochem. Bioenerg. 1995, 36, 73-76. (15) Wagner, P.; Hegner, M.; Kernen, P.; Zaugg, F.; Semenza, G. Biophys. J. 1996, 70, 2052-2066. (16) Willner, I.; Katz, E. Angew. Chem., Int. Ed. 2000, 39, 1180-1218. (17) Tarlov, M. J.; Bowden E. F. J. Am. Chem. Soc. 1991, 113, 1847-1849. (18) Prime, K. L.; Whitesides, G. M. J. Am. Chem. Soc. 1993, 115, 10714-10721. (19) Mrksich, M.; Whitesides, G. M. Annu. Rev. Biomol. Struct. 1996, 25, 5578. (20) Katz, E.; Willner, I. J. Electroanal. Chem. 1996, 418, 67-72. (21) Kajiya, Y.; Okamoto, T.; Yoneyama, H. Chem. Lett. 1993, 2107-2110. 10.1021/ac000276p CCC: $19.00

© 2000 American Chemical Society Published on Web 07/20/2000

ers. Thus, a three-dimensional matrix is formed, and the advantages of a well-defined monolayer of immobilized enzyme are lost. Other studies have reported enzymes covalently bound directly to the self-assembling monolayer (SAM) without a cross-linker.22,23 With this approach, a two-dimensional reaction layer of enzyme at the electrode surface would be expected to be produced. In the same way, a good alternative consists of the immobilization of a membrane-bound enzyme in a lipophilic, self-assembled alkanethiolate layer on gold electrode, as it has been demonstrated previously for either fructose dehydrogenase (FDH)24 or Escherichia coli fumarate reductase.25 Alkanethiols, such as octadecyl mercaptan, chemisorb quite readily to gold, forming well-ordered, densely packed monolayers.26 This approach provides a membranelike environment for the enzyme, and the alkyl chains of the chemisorbed species effectively mimic the ion barrier properties of the organized hydrocarbon chains in lipid bilayers; thus, the enzyme microenvironment in this configuration is close to that from its native lipid layer. In addition, the hydrophobic monolayer can reduce access of polar electroactive species to the electrode surface and avoid potential interferences.25-28 In this paper, we describe a very simple method to immobilize membrane-bound enzymes in an alkanethiolate layer adsorbed on gold electrodes. We chose, as our model system, membranebound FDH, D-gluconate dehydrogenase (GADH), and L-lactic dehydrogenase (cytochrome b2) (Cyb2). GADH is a bacterial membrane-bound oxidoreductase that catalyzes the oxidation of D-gluconate (GlcA) to 2-keto-D-gluconate in vivo at cytoplasmic membrane surfaces to give electrons to ubiquinone in the membrane. The enzyme is composed of three subunits with molar masses of 66 000, 50 000, and 22 000 g, each of which contains covalently bound FAD, heme c and Fe-S clusters, respectively, allowing unidirectional electron flow through the enzyme from GlcA in solution to ubiquinone in the membrane. FDH is also a membrane-bound oxidoreductase containing PQQ and heme c as redox active sites, PQQ being the site for reacting with D-fructose to produce 5-ketofructose. Cyb2 is a flavohemoprotein which contains one riboflavin phosphate (flavin mononucleotide) group and one protoheme group per unit of 82 000 molecular weight. Cyb2 catalyzes the oxidation of L-lactate to pyruvate with subsequent transfer of electrons to cytochrome c or other artificial electron acceptors.29 In our system, the membrane-bound enzyme is immobilized onto a gold electrode modified with 5-(octyldithio)-2-nitrobenzoic acid (ODTNB) apparently in much the same way that these enzymes organize in lipid layers, improving both stability and reproducibility of the enzymatic electrode. On binding proteins to gold surfaces covered with SAMs of thio compounds, the best results were often obtained when heterogeneous alkanethiol layers (22) Willner, I.; Heleg-Shabtai, V.; Blonder, R.; Katz, E.; Tao, G.; Buckmann, A. F.; Heller, A. J. Am. Chem. Soc. 1996, 118, 10321-10322. (23) McRipley, M. A.; Linsenmeier, R. A. J. Electroanal. Chem. 1996, 414, 235246. (24) Kinnear, K. T.; Monbouquette, H. G. Anal. Chem. 1997, 69, 1771-1775. (25) Kinnear, K. T.; Monbouquette, H. G. Langmuir 1993, 9, 2255-2257. (26) Porter, M. D.; Bright, T. B.; Allara, D. L.; Chidsey, C. E. D. J. Am. Chem. Soc. 1987, 109, 3559-3568. (27) Creager, S. E.; Collard, D. M.; Fox, M. A. Langmuir 1990, 6, 16171620. (28) Chailapakul, O.; Crooks, R. M. Langmuir 1993, 9, 884-888. (29) Xia, Z.; Mathews, F. S. J. Mol. Biol. 1990, 212, 837-863.

were used.23,30,31 On the basis of these results, we have selected ODTNB since this compound has the advantage of including a long chain in a short-length thio acid. Thus, after adsorption on gold electrodes, an heterogeneous-like alkanethiol layer is formed. Quartz crystal microbalance (QCM) experiments have been carried out in order to study the time evolution of the immobilization of FDH, GDH, and Cyb2 onto the ODTNB layer. We also describe the use of such enzyme electrodes for the amperometric determination of fructose and gluconate. Hydroxymethylferrocene was used as diffusional mediator between the active site of the immobilized enzyme and the electrode surface. EXPERIMENTAL SECTION Materials. The following enzyme preparations were purchased from Sigma Chemical Co: D-gluconate dehydrogenase (EC 1.1.99.3, from Pseudomonas spergilius), solution in 50% glycerol containing MgCl2, sodium gluconate, and 1% Triton X-100 and 79 units/mg of protein (4.0 mg of protein/mL); D-fructose dehydrogenase (EC 1.1.99.11, from Gluconobacter sp., grade III), lyophilized powder containing 56 units/mg of protein); L-lactic dehydrogenase (cytochrome b2) (type IV-SS; EC 1.1.2.3, from baker’s yeast), suspension in 3.2 M (NH4)2SO4 solution, pH 6.0, containing 9.7 mg of protein/mL and 1.7 units/mg of protein; glucose oxidase (GOx) (type II; EC 1.1.3.4. from Aspergillus niger) containing 30 units/mg of solid; peroxidase type I (HRP; EC1.11.1.7, from horseradish) containing 120 units/mg of solid. Solutions of GADH and Cyb2 were stored at 4 °C as received. For the other enzymes, the following stock solutions were prepared: FDH, 1.0 mg of the FDH lyophilized powder was dissolved in 50 µL of 0.1 M phosphate buffer (pH 4.5) and 50 µL of glycerol. GOx, 2.0 mg of the GOx lyophilized powder was dissolved in 60 µL of 0.1 M phosphate buffer (pH 6.0) and 60 µL of glycerol. HRP, 1.1 mg of the HRP lyophilized powder was dissolved in 250 µL of 0.1 M phosphate buffer (pH 7.0). All these stock solutions were aliquoted (10 µL) and stored at -30 °C; under these conditions the enzymatic activities remain stable for several weeks. ODTNB was purchased from Fluka and stored at 4 °C. It was, at least, 95% purity and used as received. Hydroxymethylferrocene (Aldrich 33,506-1) was used as received. Sodium gluconate (USP), D-(-)-fructose 99%, L-(+)-lactic acid lithium salt 97%, and D-(+)glucose 99.5% were purchased from Sigma Chemical Co. Hydrogen peroxide solutions were obtained by diluting a 30% solution from Carlo Erba. The Test Combination D-glucose/D-fructose assay kit was obtained from Boehringer Mannheim. Sodium phosphate (Sigma Chemical Co.) was used in the preparation of buffer solutions. Water was purified with a Millipore Milli-Q-System. Voltammetric Measurements. Cyclic voltammetric studies were performed with an Autolab/PGSTAT10 potentiostat from Eco-Chemie. The electrochemical experiments were carried out in three-compartment electrochemical cells with standard taper joints so that all three compartments could be hermetically sealed with Teflon adapters. Gold disk electrodes (2-mm diameter, 0.031cm2 geometric area, and 0.090-cm2 microscopic area) sealed in (30) Madoz, J.; Kuznetzov, B. A.; Medrano, F. J.; Garcia, J. L.; Fernandez, V. M. J. Am. Chem. Soc. 1997, 119, 1043-1051. (31) Spinke, J.; Liley, M.; Schmitt, F. J.; Guder, H. J.; Angermaier, L.; Knoll, W. J. Chem. Phys. 1993, 99, 7012-7019.

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

3785

soft glass were used as working electrodes. A large-area coiled platinum wire was employed as a counter electrode. All potentials are reported against a sodium-saturated calomel electrode (SSCE) without regard for the liquid junction, all solutions were deaerated with nitrogen gas before use, and the gas flow was kept over the solution during experiments. QCM Measurements. AT-cut quartz crystals (5 MHz) of 25mm diameter with Au electrodes deposited over a Ti adhesion layer (Maxtek Co.) were used for EQCM measurements. An asymmetric keyhole electrode arrangement was used, in which the circular electrode geometrical areas were 1.370 (front side) and 0.317 cm2 (backside). The electrode surfaces were overtone polished. Prior to use, the quartz crystals were cleaned by immersion in piranha solution, H2SO4/H2O2 (3:1). Caution: Piranha solution is extremely reactive!. They were subsequently rinsed with water and acetone and dried in air. The quartz crystal resonator was set in a probe (TPS-550, Maxtek) made of Teflon in which the oscillator circuit was included, and the quartz crystal was held vertically. The probe was connected to a cell by a homemade Teflon joint which was immersed in water-jacketed beaker thermostated at the assay temperature with a thermostatic bath (Digital Temperature Controller Haake F6). The frequency was measured with a plating monitor (PM-740, Maxtek Inc.) and simultaneously recorded by a personal computer. Procedures. Electrode Conditioning. Polycrystalline disk gold electrodes were polished with 1.0-µm diamond paste (Buehler), rinsed with water, and sonicated for 10 min in distilled water. The electrodes were activated by holding the potential at +2.0 V for 5 s in 0.1 M H2SO4 and then at -0.35 V for 10 s, followed by potential cycling from -0.35 to +1.5 V at 4 V/s for 1 min. Finally, the cyclic voltammogram characteristic of a clean polycrystalline gold electrode was recorded (from -0.2 to +1.5 V) at 100 mV/s and used to calculate the microscopic area by integration of the cathodic peak associated with the reduction of the gold oxide. The electrode was subsequently rinsed with water and ethanol and used immediately in the monolayer preparation. Adsorption of ODTNB and Enzyme Immobilization. The conditioned electrode was immersed for 3 h at room temperature in a 10 mM ODTNB solution in ethanol (EtOH). Afterward, the electrode was thoroughly rinsed with ethanol, water, and finally 0.1 M phosphate buffer, pH 7. Electrodes were employed immediately after preparation. For immobilization of the enzymes on polycrystalline ODTNBmodified gold electrodes, the following solutions were prepared daily and placed on the ODTNB-modified electrode surface for 1 h at 4 °C: (a) 10 µL of GADH stock solution mixed with 10 µL of 0.1 M phosphate buffer solution, pH 6.0; (b) 10 µL of FDH stock solution mixed with 10 µL of 0.1 M phosphate buffer solution, pH 4.5; (c) 10 µL of Cyb2 stock solution; (d) 20 µL of GOx stock solution; (e) 20 µL of HRP stock solution. In all cases, after the immobilization step, the electrodes were thoroughly rinsed with phosphate buffer and used immediately. To follow the enzyme immobilization kinetics by QCM experiments, an ODTNB-modified QCM Au resonator was immersed in a thermostated solution (10 mL of 0.1 M phosphate buffer, pH 7.0, 6.0, or 4.5 for FDH) and the frequency monitored as a function of time. After the temperature and frequency had stabilized, an aliquot of 50 (GADH, HRP, GOx), 20 (FDH), or 25 µL (Cyb2) of 3786

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

the corresponding enzyme stock solution was added to final concentrations of 0.14 µM GADH, 0.55 µM FDH, 0.15 µM Cyb2, 1.1 µM GOx, and 2.3 µM HRP. Spectrophotometric Determination of Enzyme Activity. Gluconate dehydrogenase, fructose dehydrogenase, L-lactic dehydrogenase, and horseradish peroxidase activity immobilized on an ODTNB-modified electrode was determined spectrophotometrically. Gluconate dehydrogenase was assayed as follows: 0.1 mL of 0.1 M potassium ferricyanide, 0.2 mL of 0.5 M sodium acetate buffer, pH 5.5, and distilled water up to 1 mL were mixed in a test tube. Subsequently, a GADH-ODTNB-modified electrode was immersed in this solution and the reaction was started by adding 0.1 mL of a 0.2 M D-gluconate solution. After 10 min at 37 °C, 0.5 mL of the ferric sulfate/Dupanol reagent (0.5 g of Fe2(SO4)3‚nH2O, 0.3 g of sodium lauryl sulfate, 9.5 mL of 85% phosphoric acid, and distilled water to 100 mL) was added. The mixture was allowed to stand at room temperature for 1 min. Water was added to a total volume of 10 mL, and the absorbance was read in an 1-cm-path length quartz cuvette at 660 nm32 in a Milton Roy Spectronic 3000 Array spectrophotometer. Fructose dehydrogenase activity was determined following the decolorization of 2,6-diclorophenol indophenol spectrophotometrically at 590 nm, at room temperature, according to the method reported by Yamada.33 As in the case of the gluconate dehydrogenase, instead of adding the enzyme to the reaction solution, the FDH-ODTNB-modified electrode was immersed in it. To determine the L-lactic dehydrogenase (cytochrome b2) activity, a Cyb2-ODTNB electrode was immersed in a spectrophotometric cuvette containing 1 mM ferricyanide, 10 mM L-lactate, and 100 mM phosphate buffer, pH 7.0, with 1 mM EDTA. At 30 °C, the variation of absorbance at 420 nm was recorded according to the method reported by Labeyrie.34 Horseradish peroxidase was assayed with TMB (0.5 mM) and H2O2 (50 µM) following the procedure described by us in a previous paper.35 Biosensor Description. The fructose biosensor consists of a gold electrode modified with an ODTNB layer, as described above, onto which subsequently a layer of FDH at a coverage of ∼2.5 × 10-12 mol/cm2, was deposited following the procedure described in the Experimental Section. The immobilized enzyme (FDH) oxidizes the D-fructose to 5-ketofructose in the presence of 0.1 mM hydroxymethylferrocene, which acts as soluble mediator in solution. RESULTS AND DISCUSSION Immobilization of Enzymes onto an ODTNB Layer. ODTNB is a bifunctional molecule that combines a long hydrophobic chain with a short-length hydrophilic bisulfide. In a previous study,36 we characterized the nature of the adsorbed layers of ODTNB on gold substrates and especially its concentration dependence. When gold electrodes were modified with low(32) Wood, W. A.; Fetting, R. A.; Hertlein, B. C. Methods Enzymol. 1962, 5, 287291. (33) Yamada, Y.; Aida, K.; Uemura, T. J. Biochem. 1967, 61, 636-646. (34) Labeyrie, F.; Baudras, A.; Lederer, F. Methods Enzymol. 1978, 53, 238256. (35) Darder, M.; Takada, K.; Pariente, F.; Lorenzo, E.; Abrun ˜a, H. D. Anal. Chem. 1999, 71, 5530-5537. (36) Darder, M.; Casero, E.; Dı´az, D.; Abrun ˜a, H. D.; Pariente, P.; Lorenzo, E. Langmuir, submitted.

Figure 1. Time dependence of the frequency changes of a ODTNB-modified QCM gold resonator in 0.1 M phosphate buffer solution at 4 °C, upon the following additions: (A) 50 µL of the GADH stock solution (29 µM), final concentration of GADH 0.14 µM; (B) 20 µL of the FDH stock solution (193 µM), final concentration of FDH 0.55 µM; (C) 25 µL of the Cyb2 stock solution (60 µM), final concentration of Cyb2 0.15 µM; and (D) 50 µL of the HRP stock solution (460 µM), final concentration of HRP 2.3 µM.

concentration solutions of ODTNB (less than 70 µM), a welldefined monolayer with a surface coverage of 4.2 × 10-10 mol cm-2 was obtained. This type of layer, formed by adsorption of single molecules, is in contrast with those obtained when gold electrodes were modified with high-concentration solutions of ODTNB (2.0-10.0 mM). In this case, ODTNB appears to form multilayer equivalent aggregates likely in the form of micelles. The structure of such a type of aggregates could mimic the structure of a native bilipid layer and could provide an environment conducive to the immobilization of enzymes with large hydrophobic domains, such as membrane-bound enzymes, with retention of activity. To ascertain this, we have modified gold electrodes by adsorption of 10 mM ODTNB and the immobilization of membrane-bound enzymes, in particular, GADH, FDH, and Cyb2, was studied. To carry out these studies, we have employed the QCM technique. This technique allows the measurement of mass changes at surfaces through changes in the resonant frequency of the quartz crystal. Mass changes associated with the immobilization process of membrane-bound enzymes (GADH, FDH, Cyb2) were compared with those obtained when cytosolic enzymes, such as GOx or HRP, were used. The immobilization process was carried out, as described in the procedures, in a homemade cell coupled to the QCM probe containing the ODTNB-modified Au quartz resonator in contact with 10 mL of buffer solution. The assembly was thermostated at 4 °C. After the temperature and frequency had stabilized, an aliquot of the stock solution of the enzyme to be studied was added to the buffer solution. As can be seen in the Figure 1, upon addition of GADH (A), FDH (B), and Cyb2 (C), a rapid decrease in the frequency is

Table 1. First-Order Rate Constant (k), Frequency Change (∆Fmax), Mass of Enzyme (m) Deposited on a ODTNB-Modified QCM Gold Resonator, and Surface Coverages (Γ) for the Immobilization Process of GDH, FDH, and Cyb2 k (min-1) ∆Fmax (Hz) m (ng‚cm-2) Γ (× 1012 mol‚cm-2) GADH FDH Cyb2

0.07 1.05 0.13

41.0 19.5 33.4

999 473 810

5.3 2.5 3.7

observed during the first 15 min; afterward, the frequency decreased more slowly until a steady state was reached. Assuming that the immobilization process is kinetically controlled, the data were fit to a first-order kinetics equation:

∆F ) -∆Fmax(1 - e-kt)

(1)

where ∆F is the frequency change (in hertz), ∆Fmax is the frequency change between the initial and the steady-state frequencies, and k is the first-order rate constant (min-1). The values of ∆Fmax and k obtained from the data fit, for GADH, FDH, and Cyb2, are summarized in Table 1. Assuming that the frequency decrease is only due to the change in mass arising from the immobilization of the enzyme, one can calculate the amount and the surface coverage of the immobilized enzyme layer by using the Sauerbrey equation:37

∆m ) -Cf∆F Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

(2) 3787

where ∆m is the mass change (ng‚cm-2) and Cf (17.7 ng‚Hz-1‚ cm-2) is a proportionality constant for the 5.0-MHz crystals used in this study. As can be seen in Table 1, values for the surface coverage from 2.5 × 10-12 to 5.3 × 10-12 mol‚cm-2 were obtained for the different enzymes assayed. In the case of GADH, the surface coverage value obtained is comparable to that reported by Ikeda et al.38 for GADH directly adsorbed on gold electrodes. As mentioned above, similar QCM experiments were employed to study the immobilization of cytosolic enzymes, in particular GOx (1.1 µM) and HRP (2.3 µM), on ODTNB-modified electrodes. In this case, the frequency-time course plots were very different from those obtained with the membrane-bound enzymes. As can be seen in Figure 1D, upon addition of HRP the frequency remains practically constant. Similar behavior was observed upon addition of GOx (figure not shown). These results clearly suggest that these enzymes are not immobilized on the ODTNB layer. To serve as a comparison to the QCM measurements, spectrophotometric measurements to determine the amount of active enzyme immobilized on the ODTNB-modified gold electrode were also carried out. For this purpose, the procedures described in the Experimental Section were followed. According to the results obtained from these experiments, only electrodes modified with GADH or FDH reveal a measurable enzymatic activity. In addition, surface coverages of active enzyme calculated by spectrophotometric methods were found to be about 4.5 × 10-12 and 2.7 × 10-12 mol‚cm-2, for GADH and FDH, respectively. These values agree well with those calculated by the QCM. Thus, the QCM technique allows one not only to ascertain whether enzyme immobilization takes place but also to determine the amount of enzyme immobilized. The spectrophotometric results confirm that the immobilization of GADH or FDH onto ODTNB-modified gold electrodes takes place, giving rise to catalytically active layers. In addition, as one would expect based on the results obtained by QCM, no enzymatic activity was observed for HRP, which confirms the absence of this enzyme in the ODTNB layer. In the case of Cyb2, the absence of enzymatic activity obtained in the spectrophotometric measurements for Cyb2-ODTNB electrodes may be due to the low specific activity of this enzyme (as will be discussed below). The concordance between the results obtained by QCM and spectrophotometry allows us to ensure that the decrease in the frequency observed upon the addition of membrane-bound enzymes is mainly due to the incorporation of membrane-bound enzymes into the ODTNB layer. The formation of the ODTNB layer may have the additional advantage of effectively blocking all, not just anionic, electroactive polar species from nonspecific interaction with the electrode surface, which could be very useful to avoid possible interferences with regard to biosensor applications. QCM experiments were also performed to study the direct adsorption of these membrane-bound enzymes onto a bare gold quartz crystal resonator in the same conditions described above for ODTNB-modified electrodes. In general, a decrease in frequency was observed when bare gold quartz crystal resonators were placed in contact with all membrane-bound enzymes studied. This fact indicates that hydrophobic membrane-bound enzymes (37) Sauerbrey, G. Z. Phys. 1959, 1555, 206-222. (38) Ikeda, T.; Miyaoka, S.; Miki, K. J. Electroanal. Chem. 1993, 352, 267-278.

3788 Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

Figure 2. Time dependence of the frequency changes of a QCM gold resonator in 0.1 M phosphate buffer solution at 4 °C, upon the addition of 50 µL of the GADH stock solution (29 µM), final concentration of GADH 0.14 µM. The solid line represents simple (A) or double (B) exponential fit to the data. The vertical dashed line separates the two intervals in case (B).

adsorb on metal electrodes as has been previously described by other authors.38 However, the shape of the resulting frequency changes as a function of time, as can be seen in Figure 2 for direct adsorption of 0.14 µM GADH, is different from that depicted in Figure 1A, where the immobilization of GADH was carried out onto a ODTNB layer. Assuming that the immobilization process onto the bare electrode is kinetically controlled as well, we attempted to fit data from Figure 2 to eq 1 but were unable to obtain an acceptable fit to a simple exponential (Figure 2A). However, excellent results were obtained by a double-exponential fit, as shown in Figure 2B (solid line). The vertical dashed line indicates the two separate regions. From the fits, values of 0.25 and 0.03 min-1 were obtained for the first and second processes, respectively. Although largely speculative on our part, we believe that the initial process corresponds to a crude adsorption of the enzyme molecules and the second process could be due to a rearrangement of the initial monolayer involving molecules of solvent and components included in the GADH commercial preparation, such as glycerol and Triton X-100. The surface coverage of GADH deposited on the bare gold quartz crystal resonator surface was estimated from the overall decrease in frequency at 50 min, the time at which the second process of immobilization reaches a steady state. Assuming that the major decrease in the frequency is only due to the immobilized GADH and using eq 2, the coverage was found to be 7.4 × 10-12 mol‚cm-2. However, the spectrophotometric assay carried out in

order to determine the amount of active GADH immobilized on the bare gold did not give any appreciable enzymatic activity. Similar results were obtained for Cyb2, suggesting that immobilization of these two enzymes on bare gold gives rise to either layers of denatured protein unable for catalysis or layers of enzyme with very low activity. These results are in clear contrast with those obtained for layers of FDH directly adsorbed on bare gold electrodes. In this case, catalytic activity was observed, although with significant differences when compared to that obtained at FDH-ODTNB-modified electrodes, as will be discussed below. Voltammetric Response of Enzymatic Electrodes. Electrical communication/connection between redox proteins and the electrode surface provides a general means of enhancing the activity of the redox-active biocatalyst.39 Direct contact between protein’s redox center and the electrode interface is, however, generally ineffective due to the insulation of the active site by the protein matrix. Various methodologies have been employed to enhance electrical contact between redox proteins and electrodes. One of the more common approaches involves modification of the electrode surface with promoter molecules that bind and/or align the protein to a configuration that facilitates electron transfer.40,41 To ascertain if this was the case for the enzymeODTNB-modified gold electrodes, their activity toward substrate concentration was examined. However, cyclic voltammograms conducted in 0.1 M phosphate buffer solution for GADHODTNB-, FDH-ODTNB-, or Cyb2-ODTNB-modified gold electrodes in the absence and in the presence of the corresponding substrates (D-gluconate, fructose, and L-lactate, respectively) did not exhibit catalytic peak currents, suggesting that the ODTNB had little, if any, effectiveness as a promoter. On the other hand, from QCM experiments, we have established that the amount of adsorbed enzyme is in the range of (2.55.3) × 10-12 mol‚cm-2. This amount of adsorbed enzyme may be too small to produce a surface redox wave clearly distinguishable from the base current obtained with bare electrodes, as has been previously mentioned by other authors.38 In general, it is found that when promoters are used the heterogeneous charge-transfer rate is typically much lower than it is when redox mediators are used.42 Thus, an alternative approach, based on the use of redox mediators, was pursued. In the flavocytochrome enzymes (GADH, Cyb2) and quinocytochrome (FDH) enzymes, the domains containing FAD, FMN, or PQQ are the sites to accept electrons from the corresponding substrates, these electrons reach the heme redox centers through the enzyme molecule, and the external redox mediator acts as an electron acceptor from the heme group. In solution, electrontransfer mediators operate by a diffusional route facilitating electrical contact between the enzyme’s redox center and the electrode surface. In the case of GADH, FDH, and Cyb2 several redox mediators have been reported as efficient artificial electron

acceptors.43,44 In the present work, among the redox mediators assayed, benzoquinone, naphthoquinone, and hydroxymethylferrocene, the best results were obtained when hydroxymethylferrocene was used. We have reported36 on the blocking of the redox response of hydroxymethylferrocene, at gold electrode surfaces modified with adsorbed ODTNB as well as adsorbed decyl mercaptan (DM). This material was selected in order to assess the blocking effects of an alkyl chain thiol and to compare it against ODTNB. For gold electrodes modified with DM, the cyclic voltammetric response for hydroxymethylferrocene exhibited peak currents that were significantly diminished when compared to the behavior exhibited at ODTNB-modified gold electrodes, which was virtually identical to that observed with bare gold electrodes. This would indicate that the DM monolayer is much more tightly packed than the ODTNB layer. This would be the anticipated result since in DM the alkyl chains are close together so that chain/chain interactions are strong and dominant. In ODTNB, the alkyl chains are significantly further apart due to the presence of the aromatic (nitrobenzoic acid) substituent so that chain/chain interactions are significantly reduced. The combination of a long alkyl chain with a short aromatic substituent provides an environment conducive to the immobilization of enzymes with retention of activity and allowing for easy transport (permeation) of redoxactive probes used as mediators. The cyclic voltammetric responses (at slow sweep rate) of the enzyme-ODTNB-modified gold electrodes with hydroxymethylferrocene in solution and in the presence and absence of the corresponding substrate were used to assess the catalytic activity of these enzyme electrodes. Figure 3 (curve a) depicts the cyclic voltammetric response from -0.1 to +0.4 V at 3 mV/s for (A) a GADH-ODTNB-, (B) a FDH-ODTNB-, and (C) a Cyb2ODTNB-modified gold electrode in contact with a 0.1 M phosphate buffer solution containing 0.1 mM hydroxymethylferrocene in the absence of substrate. The well-behaved redox response of the hydroxymethylferrocene in aqueous media is readily apparent, as one would expect for a reversible electrochemical response in the absence of diffusional impediments. Upon addition of the corresponding substrate (to a final concentration of 10 mM), there was an enhancement of the anodic peak current, and in addition, no current was observed in the return (cathodic) wave for electrodes modified with GADH (Figure 3A, curve b) and FDH ((Figure 3B, curve b). This behavior is consistent with a strong electrocatalytic effect. It should be noted that in the case of electrodes modified with Cyb2 a small catalytic current was observed (Figure 3C, curve b), which is probably due to the low specific activity of the commercial preparation available. This result may explain the absence of enzymatic activity obtained by the spectrophotometric measurements, although we have reported above a decrease in frequency, from QCM, which corresponds to the incorporation of 3.7 × 10-12 mol‚cm-2 Cyb2 during the immobilization of Cyb2 onto the ODTNB layer.

(39) Bartlett, P. N.; Tebbut, P.; Whitaker, R. G. Prog. React. Kinet. 1991, 16, 55-60. (40) (a) Armstrong, F. A.; Hill, H. A. O.; Walton, N. J. Acc. Chem. Res. 1988, 21, 407-413. (b) Frew, J. K.; Hill, H. A. O. Eur. J. Biochem. 1988, 172, 261-269. (41) Ruzgas, T.; Cso ¨regi, E.; Emmeus, J.; Gorton, L.; Marco-Varga, G. Anal. Chim. Acta 1996, 330, 123-138. (42) Paddock, P. M.; Bowden, E. F. J. Electroanal. Chem. 1989, 260, 487-494.

(43) (a) Ikeda, T.; Matsushita, F.; Senda, M. Agric. Biol. Chem. 1990, 54, 29192924. (b) Ikeda, T.; Matsushita, F.; Senda, M. Biosens. Bioelectron. 1991, 6, 299-304. (44) (a) Khan, G. F.; Shinohara, H.; Ikariyama, Y.; Aizawa, M. J. Electroanal. Chem. 1991, 315, 263-273. (b) Khan, G. F.; Kobakate, E.; Shinohara, H.; Ikariyama, Y.; Aizawa, M. Anal. Chem. 1992, 64, 1254-1258. (c) Begum, A.; Kobakate, E.; Suzawa, T.; Ikariyama, Y.; Aizawa, M. Anal. Chim. Acta 1993, 280, 31-36.

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

3789

Figure 3. Curves a: Cyclic voltammetric response at 3 mV‚s-1 in the potential range from -0.1 to +0.4 V for GADH-ODTNB (A), FDHODTNB (B), and Cyb2-ODTNB (C) gold modified electrodes, in 0.1 M phosphate buffer solution containing 0.1 mM hydroxymethylferrocene. Curves b: upon addition of 10 mM gluconate (A), 10 mM fructose (B), and 10 mM lactate (C).

Figure 4. Catalytic currents vs fructose concentration for a FDHODTNB-modified gold electrode, in 0.1 mM phosphate buffer (pH 4.5) using 0.1 mM hydroxymethylferrocene as mediator. The inset shows the linear range up to 0.7 mM.

To confirm the role of the enzyme in the catalytic response to substrate, ODTNB-modified electrodes without immobilized enzymes were immersed in 0.1 M phosphate buffer containing 0.1 mM hydroxymethylferrocene. As one would expect, upon addition of the corresponding substrates no catalytic waves were obtained. Figure 4 depicts a typical calibration curve obtained with a FDH-ODTNB-modified electrode poised at a potential of 0.3 V vs SSCE where oxidation of hydroxymethyl ferrocene by the electrode is assured. To normalize the response to the different electrodes employed, the currents are plotted as (ia - id)/id, where ia is the anodic current under catalytic conditions (i.e., in the presence of substrate; fructose) and id is the current due to the mediator in the absence of substrate. As can be seen, initially the catalytic current increased linearly with fructose concentration and 3790

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

then leveled off, suggesting a saturation response. A similar behavior was also observed for GADH-ODTNB electrodes upon addition of increasing amounts of gluconate. For FDH-ODTNB electrodes, calibration curves exhibit a linear relationship between the biosensor response and the fructose concentration up to 0.7 mM. Statistical analysis gave an excellent linear correlation (r ) 0.9993). The sensitivity, or the slope, of the plot in Figure 4 was 6.06 mM-1. The lowest fructose concentration measured was 0.02 mM, and the relative standard deviation for five successive measurements of 10 mM fructose was 2.5%. The electrocatalytic activity of electrodes modified with ODTNB and glucose oxidase (as described in the Experimental Section) toward the oxidation of glucose was also tested. When the ODTNB-GOx-modified gold electrode was placed in a 0.1 M phosphate buffer solution containing 0.1 mM hydroxymethylferrocene, no electrocatalytic response was observed upon addition of 3 mM of glucose. A similar behavior was observed for HRPODTNB-modified gold electrodes in the same conditions and using hydrogen peroxide as substrate. These results are in accord with those obtained from QCM experiments and also confirm the absence of either GOx or HRP in the ODTNB layer. As mentioned above, bare gold electrodes modified by direct adsorption of FDH exhibited a significant catalytic activity toward the oxidation of fructose. Therefore, additional studies in order to compare this catalytic activity to that obtained at FDH-ODTNB gold-modified electrodes were carried out. For this purpose, a gold electrode was modified by placing it in a solution containing 10 µL of 0.1 M phosphate buffer solution, pH 4.5, and 10 µL of FDH stock solution for 1 h (at 4 °C). After being rinsed with phosphate buffer, the modified electrode was placed in an electrochemical cell containing 0.1 M phosphate buffer solution, pH 4.5, and 0.1 mM hydroxymethylferrocene. The catalytic activity of such modified electrodes was evidenced upon addition of 10 mM fructose. A further analysis of the catalytic waves obtained at different concentrations of fructose reveals that the catalytic responses were lower and less reproducible when compared to

Figure 5. Catalytic currents vs time for a FDH-ODTNB-modified gold electrode (A) and a bare gold electrode modified with FDH by direct adsorption (B) in the presence of 10 mM fructose. Other conditions as in Figure 4.

FDH-ODTNB-modified electrodes. As an example, the relative standard deviation for five successive measurements of 10 mM fructose using FDH directly immobilized on the gold surface was 20%. This is in clear contrast to the 2.5% obtained for the relative standard deviation for five successive measurements of 10 mM fructose when FDH is immobilized on to a ODTNB-modified gold electrode. In addition, the catalytic response obtained with FDHODTNB electrodes was more stable than that obtained with FDH gold electrodes. Figure 5A shows the catalytic response versus time for FDH-ODTNB-modified gold electrodes (A) and for FDHmodified gold electrodes (B). As can be observed, the immobilized enzyme system, using the ODTNB, shows a significant stable catalytic current for at least 25 days, in spite of repeated rinsing and cyclic voltammetry studies run to steady state. However, when the enzyme (FDH) was immobilized directly on bare gold electrodes, the response was variable from day to day and tended to decay, as depicted in Figure 5B. An additional advantage of the ODTNB enzyme immobilization system is that the ODTNB layer coating the gold electrode surface limits the access of oxidable ascorbic acid to the gold surface. Indeed, over a range of 0-0.1 mM ascorbic acid, gold electrodes modified with ODTNB showed a ∼60% reduction in oxidative current response to ascorbic acid as compared to a bare gold electrode (see Figure 6). Thus, ascorbic acid at concentrations of 4% of fructose level, which is close to those found in citrus juice, caused errors in fructose measurements of only 2%. This value is very similar to that obtained by Kinnear et al.24 for a fructose biosensor based on fructose dehydrogenase immobilized in a membrane mimetic layer on gold and really smaller than the value of 80% reported for a sensor based on FDH immobilized on a carbon paste electrode.43b Finally, the application of the proposed FDH biosensor was focused on the analysis of fructose-containing beverages, like apple and orange juice, in the presence of ascorbic acid. As is the case

Figure 6. Cyclic voltammetry response of a bare (A) and a ODTNB-FDH (B) gold electrode to ascorbic acid 0.01 mM (curve b) and 0.3 mM (curve c) in 0.1 M phosphate buffer solution, pH 4.5. Scan rate, 3 mV‚s-1. Curve a: absence of ascorbic acid.

in many instrumental methods, analyses were based on a calibration curve obtained from known concentrations of fructose. Once a given electrode was calibrated, it could be employed in multiple determinations. As shown earlier, the response due to ascorbic acid at ODTNB-modified gold electrodes is greatly suppressed. Given this and the low concentration of ascorbic acid relative to fructose, any response due to the ascorbic acid oxidation will be considered negligible. The samples (5 µL of apple juice or 10 µL of orange juice) were diluted in 5 mL of 0.1 M phosphate buffer, pH 4.5, containing 0.1 mM hydroxymethylferrocene. The results presented are the average of four determinations and were compared to those obtained with the enzymatic spectrophotometric assay kit. For orange and apple juice, the biosensor yielded an average fructose concentration of 139 and 299 mM, respectively. Values that are in accord with the assay kit values of 138 and 302 mM (n ) 4, where n equals the number of assays) were obtained for orange and apple juices, respectively. The relative standard deviation for the orange juice sample was 3.4% and for the apple juice 1.9%. CONCLUSIONS An effective method of immobilizing membrane-bound enzymes in an ODTNB layer on gold electrodes has been described. The amount of active enzyme, in particular, GADH, FDH, and Cyb2, immobilized has been estimated from QCM measurements giving values from 2.5 × 10-12 to 5.3 × 10-12 mol‚cm-2. Electrochemical and spectrophotometric assays indicate that the immobilized enzyme retains its enzymatic activity after immobilization. A fructose biosensor was developed making use of a gold surface modified with ODTNB and FDH employing hydroxymethylferrocene as mediator in solution. Ascorbic acid, at concentrations similar to those found in citrus juice, caused errors in fructose determinations of only 2%. In fact, the average measureAnalytical Chemistry, Vol. 72, No. 16, August 15, 2000

3791

ment of fructose content of either orange or apple juice conducted with the biosensor was very similar to the value determined with the enzyme assay kit.

to´noma de Madrid (Grant 07M/0016/1999). M.D. and E.C. also acknowledge support by Fellowships from the Comunidad Auto´noma de Madrid.

ACKNOWLEDGMENT This work was supported by the CICyT of Spain through Grants PB97-0037 and PB98-0082, and by the Comunidad Au-

Received for review March 9, 2000. Accepted May 31, 2000.

3792

Analytical Chemistry, Vol. 72, No. 16, August 15, 2000

AC000276P