Author's Accepted Manuscript

Biofuel cells based on direct enzymeelectrode contacts using PQQ-dependent glucose dehydrogenase / bilirubin oxidase and modified carbon nanotube materials V. Scherbahn, M.T. Putze, B. Dietzel, T. Heinlein, J.J. Schneider, F. Lisdat

www.elsevier.com/locate/bios

PII: DOI: Reference:

S0956-5663(14)00360-1 http://dx.doi.org/10.1016/j.bios.2014.05.027 BIOS6789

To appear in:

Biosensors and Bioelectronics

Received date: 26 February 2014 Revised date: 6 May 2014 Accepted date: 10 May 2014 Cite this article as: V. Scherbahn, M.T. Putze, B. Dietzel, T. Heinlein, J.J. Schneider, F. Lisdat, Biofuel cells based on direct enzyme-electrode contacts using PQQ-dependent glucose dehydrogenase / bilirubin oxidase and modified carbon nanotube materials, Biosensors and Bioelectronics, http://dx.doi. org/10.1016/j.bios.2014.05.027 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Biofuel cells based on direct enzyme-electrode contacts using PQQ-dependent glucose dehydrogenase / bilirubin oxidase and modified carbon nanotube materials V. Scherbahn1, M. T. Putze1, B.Dietzel2, T. Heinlein3, J. J. Schneider3, F. Lisdat1 1 Biosystems Technology, Technical University of Applied Sciences, 15745 Wildau, Germany 2 Institute for Thin Film and Microsensoric Technology, 14513 Teltow, Germany 3 Technical University Darmstadt, Eduard-Zintl-Institute for Inorganic and Physical Chemistry, 64287 Darmstadt, Germany

Abstract Two types of carbon nanotube electrodes (1) buckypaper (BP) and (2) vertically aligned carbon nanotubes (vaCNT) have been used for elaboration of glucose/O2 enzymatic fuel cells exploiting direct electron transfer. For the anode pyrroloquinoline quinone dependent glucose dehydrogenase ((PQQ)GDH) has been immobilized on[poly(3-aminobenzoic acid-co-2methoxyaniline-5-sulfonic acid), PABMSA] - modified electrodes. For the cathode bilirubin oxidase (BOD) has been immobilized on PQQ - modified electrodes. PABMSA and PQQ act as promoter for enzyme bioelectrocatalysis. The voltammetric characterization of each electrode shows current densities in the range of 0.7-1.3 mA/cm2. The BP-based fuel cell exhibits maximal power density about 107 µW/cm2 (at 490 mV). The vaCNT-based fuel cell achieves a maximal power density of 122 µW/cm2 (at 540 mV). Even after three days and several runs of load a power density over 110 µW/cm2 is retained with the second system (10 mM glucose). Due to a better power exhibition and an enhanced stability of the vaCNT-based fuel cells they have been studied in human serum samples and a maximal power density of 41 µW/cm2 (390 mV) can be achieved.

Keywords Enzymatic fuel cell; PQQ-dependent glucose dehydrogenase; Bilirubin oxidase; Buckypaper; Vertically aligned carbon nanotubes

1   

1.

Introduction

During the last decade enzymatic biofuel cells (EBFCs) have becoming an interesting research topic particularly with respect to their sustainability and potential application as power supply for portable, implantable devices in medicine and biosensor systems [1]. Their stability, generated power and the cell voltage depends to a large extent on the choice of the enzymes for anode and cathode reaction and the electrode architecture used. For the cathode multicopper enzymes such as bilirubin oxidase (BOD) [2], laccases [3] and ascorbate oxidases [4] are suitable biocatalysts. For bioanodes, the application of different oxidizing enzymes allows to harvest the energy out of diverse biofuels e.g. glucose, fructose, cellobiose, alcohol or hydrogen. Thus, enzymes such as glucose oxidase, nicotinamide adenine dinucleotide and pyrroloquinoline

quinine

dependent

glucose

dehydrogenase

(PQQ)GDH,

fructose

dehydrogenase, cellobiose dehydrogenase and hydrogenases are suited for this purpose [5,6]. An efficient electrical communication between the enzyme and the electrode can be achieved via direct electron transfer (DET). It allows a current flow at potentials near the E° of the redox centre of the bound enzyme and avoids side reactions. Alternatively the addition of shuttle molecules results in a mediated electron transfer (MET) that may enhance the maximum rate of enzyme-electrode electron transfer, compared to DET. Here the enzymes do not need to contact the electrode surface directly [6], but the redox potential of the mediator influences the cell potential and problems with leakage can occur in case of soluble mediator or problems in enzyme accessibility for the substrate in case of polymer-bound mediators. The approach to develop a membrane-less EBFC requires a complete insensitivity toward oxygen during the anodic reaction of the fuel since oxygen is mostly used as electron acceptor on the cathode. Hence, the choice of a suitable enzyme and strategies for an effective competition with oxygen are important. For this purpose, (PQQ)GDH is an interesting enzyme because it can be produced in a recombinant way, it bears a high catalytic activity at physiological pH and is oxygen insensitive [7]. Several studies report different strategies for functional (PQQ)GDH - electrode contacts. Often this communication occurs via mediators e.g. ferrocene derivatives [8], PQQ [9], cytochrome c [10] or osmium containing polymers [11]. DET between the enzyme and the electrode surface has also been achieved byusing self assembling monolayers on gold with a covalent enzyme attachment [12], by carbon black modified carbon paste [13], by aniline derivatives modified carbon nanotubes [14], by activated buckypaper with covalently attached (PQQ)GDH [15], by attaching the enzyme on a carbon cryogel electrode [44] or by titan oxycarbide nanostructures [45]. 2   

In terms of biocathode, BOD is a favorable enzyme for oxygen reduction. It has the advantage of being stable at neutral pH, possesses a high activity and its reduction process starts at potentials ~0,5 V vs. Ag/AgCl [16]. DET between BOD and electrodes has been achieved at modified carbon- [35,36], gold- [34,38] and CNT-electrodes [33]. Oxygen supply during the reduction reaction has been recognized as a limiting factor for the power output, thus the development of gas-diffusion cathodes based on carbon-black modified carbon toray paper has been shown to avoid this problem [18,19]. Despite the constant improvement of the EBFCs, their power output and lifetime are still not optimal for direct and long-time applications. Despite this, first implantable EBFCs have been reported [15,20]. Beside the choice of the biocatalysts, suitable electrode materials with a high surface area are important for improving the EBFC performance. Due to their excellent electrocatalytic activity and an immense surface area several types of CNTarchitectures have been used for developing bioelectrodes such as immobilized CNTs on carbon [33,37] or gold surfaces[17], vertically aligned CNTs [21,22] or buckypaper [23]. In order to increase the bio-compatibility of the rather hydrophobic CNT-surface different pretreatments can be applied: sonication- [17], acid- [24], plasma- [25], electrochemical or chemical pretreatment [26]. Because of their dual advantages of productive binding of enzymes in an active form and allowing for electron transport towards the electrode different conducting

and

biocompatible

polymers

such

as

modified

polyaniline- [27],

polythiophene [28,29] and polypyrrole-derivatives [30] can be combined with CNTs. The aim of the present study is to develop glucose/O2EBFCs based on two different CNT-architectures: (1) buckypaper (BP) and (2) vertically aligned carbon nanotubes (vaCNT). In order to promote the direct contact between the enzyme and the electrode surface both the anode and the cathode have been separately evaluated. At the anodic side (PQQ)GDH is immobilized at polymer modified CNT-electrodes. At the cathodic side BOD has been used as biocatalyst. The performance of the developed EBFCs has been characterized in artificial and physiological (human serum) solutions.

2. Experimental 2.1 Materials Buckypaper (BP) have been obtained from Buckeye Composites (USA). Human serum samples containing 3-4 mM glucose are received from LIMETEC Biotechnologies GmbH, Germany as a kind gift. Pyrroloquinoline quinine (PQQ) is purchased from Wako Pure 3   

Chemical Industries. Soluble GDH (Acinetobacter calcoaceticus) is provided as apo-enzyme by Roche Diagnostics GmbH as a kind gift. The enzyme is recombinantly expressed in E. coli. Poly(3-aminobenzoic acid-co-2-methoxyaniline-5-sulfonic acid) - PABMSA has been synthesized by chemical oxidative polymerization according to the procedure described herein [46]. N-hydroxysuccinimide (NHS), 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), D-glucose and bilirubon oxidase from Myrothecium verrucaria(BOD) are purchased from Sigma-Aldrich Chemie GmbH (Germany). 2-(N-morpholino)ethansulfonic acid (MES) isfrom ApliChem GmbH (Germany). Calcium chloride (CaCl2) and citric acid are received from Carl Roth GmbH + Co. KG (Germany). For all aqueous solutions 18 MΩ deionized water (Eschborn, Germany) is used. The vertically aligned vaCNT@Si-electrodes (vaCNT@Si) have been prepared by a water assisted chemical vapour deposition (CVD) method [47,48]. A silicon substrate (Sb doped, , 0.01-0.02 Ω/cm, from Silicon Materials), size 5 mm x 15 mm is covered with a mesh containing cavities of 370 µm x 370 µm in size. Deposition of 11.6 nm Al and 1.4 nm Fe by e-beam evaporation followed and the mask is removed. CVD synthesis has been started by heating the substrate to 850 °C in a gas mixture of argon and hydrogen (40 % hydrogen) in a quartz tube with inner diameter 85 mm. The patterned structure of vaCNTs is obtained by introducing a flow of ethine (200 sccm) as carbon source for 15 min. Microscopic and spectroscopic investigations are performed using SEM (XL 30 FEG, Philips), TEM (CM20, Philips) and Raman (LabRamHR8000, Horiba) - see Fig. 1.

2.2.Enzyme electrode preparation (PQQ)GDHis reconstituted by dissolving 2 mg/ml (20 µM) of apoGDH in 5 mM MES + 1 mM CaCl2, pH 6.5. Next PQQ has been added with a molar ratio of 1 (PQQ/apoGDH). The solution has been incubated for 3 h at room temperature in the dark. The resulting enzyme solution is stored at 4 °C before use. 1 mg BOD has been dissolved in 1ml citrate phosphate buffer (100mM CiP, pH 7) and stored as 30 µl aliquots at -20 °C before use (concentration amounts to 20 µM). For the electrode preparation the BP material has been cut into rectangular pieces. The approximate surface of the electrode immobilized with enzyme is 0.05-0.11 cm2. Anodes: The BP- und vaCNT@Si-electrodes have been first incubated with PABMSA solution of different concentrations (0.1, 1, 2, 3 or 5 mg/ml) in 5 mM MES + 1 mM CaCl2, pH 6.5 for one hour. After this the electrodes are washed three times with the same buffer. For (PQQ)GDH adsorption the electrodes have been placed in enzyme solution for 1 h. In order to 4   

fix the enzyme covalently, the PABMSA-modified electrodes are placed in an EDC/NHSsolution (100 mM/ 25 mM) in 5 mM MES + 1 mM CaCl2, pH 6.5 for 15-20 min and after that washed 3x with the same buffer before enzyme incubation. The (PQQ)GDH/PABMSAelectrodes are stored in 5 mM MES + 1 mM CaCl2, pH 6.5 at 4 °C. Cathodes: The BP- and vaCNT@Si-electrodes have been incubated with a PQQ solution (1 mM or 2.73 mM) in 100 mM citrate phosphate buffer (CiP), pH 7 for 1 h. Then the electrodes have been washed three times with the same buffer. For BOD adsorption on the surface the electrodes are placed in enzyme solution (10 µM) for 1 h. In order to fix the enzyme covalently, the PQQ-modified electrodes have been placed in EDC/NHS-solution (100 mM/25 mM) in 100 mM CiP for 15-20 min and washed three times with the same buffer before enzyme incubation. The BOD/PQQ-electrodes are stored in 100 mM CiP, pH 7, 4 °C.

2.3. Electrochemical measurements The voltammetric experiments are performed using the potentiostat PGSTAT 12 (Metrohm-Autolab, Netherlands). A three electrode 1ml home-made chemical cell is used with a Pt wire as counter electrode, an Ag/AgCl (1 M KCl) electrode as reference electrode and the modified BP- and vaCNT@Si-electrodes as working electrodes. For studying the anodes buffer solutions (5 mM MES + 1 mM CaCl2 pH 6.5 and 100 mM CiP + 1 mM CaCl2 pH 7) in absence and in presence of glucose have been used as electrolyte. For cathode characterization 100 mM CiP buffer, pH 7 (Ar- and air-saturated) has been used. For all bioelectrocatalytic measurements a scan rate of 10 mV/s is applied. The biofuel cells are characterized by performing galvanodynamic measurements using the potentiostat Reference 600 (Gamry Instruments, USA). Since no difference in power output could be seen at scan rates 2 nA/s and 3 nA/s, the scan rate of 3 nA/s has been applied resulting in a test period of about 2 h for each measurement. For the serum measurements 1 ml serum have been placed in a home-made electrochemical cell and then the prepared electrodes have been inserted and characterized accordingly.

3. Results and discussion In the present work we design protein electrodes based on DET for the application in biofuel cells by using two types of CNT-architectures and (PQQ)GDH as glucose converting enzyme and BOD as O2-reducing enzyme. We use here (a) buckypaper and (b) vertically aligned CNT (Fig. 1) as interface for the immobilization of redox enzymes both for anodes and cathodes. The vaCNT arrays are grown by a water assisted chemical vapor process which 5   

ensures a high quality of mostly double walled CNTs. Their structural and spectroscopic characterization gives a main diameter of 6-10 nm containing only a minor amount of multi walled CNTs with more graphitic shells. The ID/IG of ratio is 0.7 as obtained by Raman spectroscopy (see Fig. 1). In order to improve the biocompatibility of the CNT-structure two types of organic interlayers have been introduced: (1) an aniline-type polymer (PABMSA, Scheme 1a) for the anodes and (2) a 3-ring aromatic compound: pyrroloquinoline quinone (Scheme 1b) for the cathodes. Furthermore, parameters such as glucose sensitivity, the influence of the polymer concentration and different buffer systems have been investigated in order to elucidate suitable conditions for a maximal output of enzymatic biofuel cell (EBFC) operation. Finally, the stability of the EBFCs has been investigated in buffer and in human serum. 3.1. Buckypaper-based anode In order to couple PQQ-GDH to the electrode we have used buckypaper (BP) – a material of high surface area and composed of multiwalled carbon nanotubes. In addition it has been shown to interact productively with several redox enzymes [23,32]. Thus, it is first tried to adsorb (PQQ)GDH onto an untreated BP-electrode. In the presence of glucose a small catalytic current can be detected by linear sweep voltammetry (LSV). At 0.1 V vs. Ag/AgCl a current density of 3-4 µA/cm2 (10 mM glucose) is obtained. These measurements show that a direct interaction of the enzyme with the CNT-material is feasible, but efficiency of bioelectrocatalysis is very low. In order to improve the surface properties of the BP for immobilization and DET an aniline-based polymer film (PABMSA, see Scheme 1a) has been adsorbed on the electrode before enzyme fixation. This idea is based on previous studies showing that sulfonated polyaniline films can improve the interaction with (PQQ)GDH [39]. LSV and cyclic voltammetry (CV) are performed in order to study whether it is possible to achieve enhanced catalysis in presence of glucose by adsorbed (PQQ)GDH on the PBMSA-modified BP-electrode. In Fig. 2a cyclic voltammograms of PABMSA/BPelectrodes without and with adsorbed (PQQ)GDH in absence and in presence of glucose are depicted. It is evident that no glucose conversion occurs at PABMSA/BP-electrodes. When (PQQ)GDH is adsorbed on the PABMSA/BP-electrodes an oxidation current in the presence of glucose starts from about -0.1 V vs. Ag/AgCl and thus indicates efficient sugar conversion. Since no additional mediator is added a DET from the sugar reduced enzyme to the polymermodified electrode at the enzymes redox potential can be concluded. The polymer

6   

modification of the CNT surface obviously increases the amount of productively immobilized enzyme on the electrode. Furthermore, current densities about 400 µA/cm2 at 0.1 V vs. Ag/AgCl can be achieved which are reasonable high (for the calculation the geometrical area in contact with the solution has been used). Thus, it is evident that the PABMSA-film improves the biocompatibility and allows a much higher enzyme activity after its immobilization in comparison with unmodified BP-electrodes. In a next step we investigate whether different PABMSA concentrations during the BP modification will influence the bioelectrocatalysis of the immobilized enzyme in the presence of substrate. Fig. 2b depicts current densities at 0.1 V vs. Ag/AgCl achieved at (PQQ)GDH/PABMSA/BP-electrodes prepared with different PABMSA-concentrations in dependence on different glucose concentrations. Obviously the polymer concentration influences the efficiency of the glucose conversion and applying 5 mg/ml of PABMSA for the CNT modification maximal current can be obtained for a given glucose concentration. It is also studied how the buffer composition influences the current output. Ca2+-ions are a significant parameter for the stabilization of the enzyme and thus are always added to the buffer. Two buffer systems with different buffer capacity have been tested with respect to the performance of the anode:

100 mM citrate-phosphate (CiP) and 5 mM 2-(N-

morpholino)ethanesulfonic acid buffer (MES). Fig. 2c illustrates the response in presence of different glucose concentrations in MES and CiP buffer solutions at 0.1 V vs. Ag/AgCl. It is evident that using CiP higher current densities at higher glucose concentrations can be detected (current densities of ~0,75 mA/cm2 in presence of 10 mM glucose). The higher ionic strength and higher buffer capacity of the CiP buffer seems to be beneficial for an efficient glucose conversion at the (PQQ)GDH/PABMSA/BP-electrode. It can be stated that mainly the high ionic strength is responsible for the higher current output since application of a 50 mM MES buffer also results in higher current values compared to the measurements in 5 mM MES buffer. Obviously the enzyme electrode interaction is improved and the system can follow the catalytic ability of the enzyme at higher glucose concentrations probably also due to the higher buffer capacity leading to less fluctuation in local pH near the electrode. Besides the ability to increase active amount of the enzyme on the electrode, the PABMSA film bears carboxylic acid groups within its structure. In order to increase the stability (PQQ)GDH can be bound covalently by the EDC/NHS chemistry on the polymer film. By means of LSV a rather similar behavior was found with slightly lower maximal current densities. This approach is also used in the EBFC measurements. 7   

3.2. vaCNT-based anode Previous studies reporting about vertically aligned CNTs (vaCNT) as suitable electrode surfaces for electrochemical applications [21,22] give the background for our studies. vaCNTs have been prepared in an array format with individual spots on n-doped silicon (see 2.1). Analogously to experiments with BP, we first adsorb the (PQQ)GDH onto untreated vaCNTelectrodes. In the presence of 10mM glucose current densities in the range of 5 µA/cm2 can be detected at 0.1 V vs. Ag/AgCl by means of LSV with such an electrode (5 mM MES pH 6.5, 1 mM CaCl2). These experiments show that the enzyme interacts directly with the vaCNTs, but the current output is insufficient for an application in an EBFC. In order to establish an efficient anode system based on vaCNTs the surface is modified in a way, which has been found optimal during the development of the BP- based anode, namely a PABMSA concentration of 5 mg/ml for electrode modification. Thus, vaCNT@Sielectrodes have been incubated in the copolymer solution and then (PQQ)GDH has been immobilized by adsorption. Subsequently the bioelectrocatalytic behaviour is analysed. Fig. 2d depicts CVs of a (PQQ)GDH/PABMSA/vaCNT@Si-electrode in absence and in presence of glucose using the MES buffer system. It is evident that a catalytic current in the range of about 590 µA/cm2 can be obtained. This value is about 1.5 times higher in comparison to the BP-electrode at similar potentials. Communication between the enzyme and the electrode occurs here in an analogous way via DET. Evaluating the glucose sensitivity using CiP buffer by means of LSV a defined dependence on the enzyme substrate concentration and higher current values can be found. The results are presented in Fig. 2e. Maximum current densities in the range of about 1.3 ± 0.18 mA/cm2 (at 0.1V vs. Ag/AgCl, n = 3) can be generated in presence of 6 mM glucose (the geometrical area is used for the calculation as sum of the individual CNTsquares).

3.3 Buckypaper-based cathode In order to construct BOD-cathodes three approaches have been followed: (1) adsorption of BOD on untreated buckypaper (BP), (2) adsorption of PQQ on BP before BOD adsorption and (3) adsorption of PQQ on BP before covalent fixation of BOD via EDC/NHSchemistry. The idea of applying PQQ as CNT modifier is based on earlier observations that PQQ can work as promoter for BOD bioelectrocatalysis [16]. The PQQ modification of the electrode has been performed with two different PQQ concentrations: 1 mM and 2.73 mM. 8   

The results of the bioelectrocatalytic reduction of O2 are depicted in Fig. 3a. Reduction currents can be detected in air-saturated 100 mM CiP buffer (quiescent solution, pH 7) by means of LSV starting at around +0.5 V vs. Ag/AgCl which agrees with the behavior of immobilized BOD on other carbon surfaces [2,16,17,33] indicating that the T1 center of the enzyme is in contact with the electrode. The approach of using PQQ as interface and a covalent attachment of the BOD shows the highest catalytic currents - about 1 mA/cm2 at 0.1 V vs. Ag/AgCl in an unstirred solution in comparison to BOD adsorbed on PQQ-modified or untreated BP (evaluation of the current change with respect to the result in Ar-saturated buffer). PQQ works here not as a mediator since it is not reduced when the electrode starts to transfer electrons via BOD towards oxygen (at +0.5 V). It enhances the productive enzymeelectrode interaction and thus serves as a promoter here. Moreover, applying BP as electrode surface no or rather weak diffusion limitation can be observed in the voltammetric curves. This phenomenon may be explained by the filter-like structure and the hydrophobicity of the buckypaper, which allows O2 not only to diffuse from the solution but also from air along the electrode down to the BOD-modified side.

3.4. vaCNT-based cathode Our second cathode system based on BOD has been established by applying vaCNT@Si-electrodes modified with PQQ as interface for the covalent fixation of the enzyme. Fig. 3b depicts the current behaviour of a BOD/PQQ/vaCNT@Si-electrode in Arand in air-saturated solution by means of LSV. It is obvious that bioelectrocatalytic oxygen reduction takes place and starts at potential of about + 0.5 V vs. Ag/AgCl. The shape of the catalytic current indicates a high catalytic activity of the immobilized enzyme and a diffusion limitation of the electrode process. This phenomenon can be explained by the measurement set up: During the measurements the spots with the vaCNTs are completely covered by buffer, thus oxygen can only be provided from solution. However, the higher reduction current found in the steepest part of the curve (in comparison to BP electrode) indicates that the modified vaCNTs host a significant access amount of active BOD. The steady-state catalytic current at +0.1 V vs. Ag/AgCl shows current densities of about 550 µA/cm2.

3.5. (PQQ)GDH | BOD – biofuel cell based on buckypaper In the current study we have shown that BP can be successfully applied to construct efficient enzyme electrodes. Consequently, an EBFC based on BP has been assembled applying following preparation conditions: (PQQ)GDH attached on PABMSA/BP as anode 9   

and BOD covalently bound to a PQQ/BP-electrode as cathode. One of the most important parameter for a fuel cell is the power output which has been studied under optimal conditions (100 mM CiP pH 7, 10 mM glucose, air saturated) by galvanodynamic sweep measurements. Such an experiment lasts for about 2 h providing real data of the performance under load. The behaviour of this EBFC is depicted in Fig. 4a. Here the power density and the cell voltage in dependence on the current density are given. The power density curve shows a maximum of about 100 µW/cm2 at a rather high cell potential of 490mV and a current density of 203 µA/cm2. Furthermore, the reproducibility of the electrode preparation can be shown by testing 3 biofuel cells which achieve in average: 104 ± 14 µW/cm2 (n = 3). The developed system exhibits an almost two times higher power density than the EBFCs from our previous studies also applying BOD and (PQQ)GDH as biocatalysts but based on a different CNT material and modification [9,14]. After three successive applications of each fuel cell (for about 3 x 2 h each) a power density decrease up to 30% has been found, whereas after four days a decrease even down to 13% can be observed. This behavior indicates a limitation caused probably by the instability of the adsorbed (PQQ)GDH at the anode. In order to improve the behavior the enzyme is covalently attached via EDC/NHS on the PABMSA/BP-electrode (PABMSA contains carboxylic groups, Scheme 1a). Using these electrodes in a fuel cell set up about 65% (70 ± 14 µW/cm2, n = 3) of the original power density (107 ± 5 µW/cm2, n = 3) can be retained after three successive measurements of each fuel cell. Comparing this result to the fuel cells with adsorbed (PQQ)GDH it can be concluded that the covalent coupling of the enzyme on the PABMSA/BP-electrode positively influences the stability of the BP-fuel cells.

3.6. (PQQ)GDH | BOD – biofuel cell based on vaCNT Our second EBFC system based on vaCNT@Si-electrodes has been constructed applying

the

same

preparation

conditions

as

the

BP-EBFC:

(PQQ)GDH/PABMSA/vaCNT@Si-electrodes as anode and BOD/PQQ/vaCNT@Si-electrode as cathode. In order to provide improved stability both BOD and PQQ-GDH are covalently attached on the modified electrodes. The performance of the vaCNT@Si-based EBFC is shown in Fig. 4b. The power density curve achieves its maximum of about 130 µW/cm2 at a cell potential of about 560 mV. It has to be mentioned here that the cell voltage is decreasing rather slowly by increasing the current flow through the system. In addition higher current densities can be obtained with this electrode combination. This supports the idea that a high

10   

enzymatic activity can be achieved within the modified CNT architecture. Further arguments in this direction can be collected analyzing the stability of the EBFC (see below). Compared to our previous studies, the power performance can be clearly enhanced (23 µW/cm2 [9], 65 µW/cm2 [14]). Moreover, the achieved power density of the developed fuel cells exceeds the results of reported fuel cells using GOD/Lac and GOD/BOD (43 µW/cm2 [50]), FDH/BOD on CNT and KetjenBlack electrodes (50 µW/cm2 [40]) or CDH and BOD on nanoporous gold (40 µA/cm2 [51]) and reaches the same range with 131 µW/cm2 using NAD-dependent GDH and Lac immobilized on SWCNT [3]. However, the power density is lower compared to EBFCs with FDH/BOD using Au-NP and carbon paper achieving 0.66 mW/cm2 or a GOD/Lac-fuel cell based on enzyme-CNT-composite discs with 1.3 mW/cm2 [41]. The performance of the EBFC has also been analyzed by repeated measurements (about 2 h each). After 3 successive measurements of the same EBFCs they show only a small decrease (18%) of the original maximum power density (122 ± 8 µW/cm2, n = 3) at 540 ± 50 mV and a rather constant value of current density of about 230 µA/cm2 (n = 3). The OCP is about 690 mV. This can be seen as a first hint for an optimized enzyme environment within the vertically aligned CNT architecture and provides the basis to study the performance within the period of several days. The key parameters of these experiments are shown in Tab. 1a. As one can see there are some fluctuations in the performance which increase with the storage time of the electrodes (overnight in the refrigerator). Even after 3 days more than 110 µW/cm2 as maximum power density can be retained. This is clearly a progress compared to the immobilization of (PQQ)GDH on top of disordered MWCNTs where already after the first day a significant loss of activity has been found [9]. Furthermore, a stable cell potential can be provided within several days of application. Another aspect of this work relates to the application of the EBFC in real biological fluids in order to evaluate the performance under more realistic conditions. Because the Si/vaCNT based EBFCs exhibit better cell parameters they are used for these studies in human serum samples. The results including power density and the cell potential are presented in Tab. 1b. A maximum power density of 41 ± 7 µW/cm2 (n = 3) is achieved at a cell potential of 390 ± 70mV (n = 3). These values are lower than biofuel cells operating in glucose containing buffer (Fig. 4b, Tab. 1a). The rather complex matrix influences obviously the enzyme-electrode contact since both the maximum current density and the potential at maximum power are smaller. But, this performance is found to be rather reproducible using three similarly prepared fuel cells. Another reason for the diminished performance might be 11   

the lower glucose concentration in serum. Repeated experiments in serum with added glucose clearly show that glucose sensitivity is maintained. Taking a medium glucose level of 3-4 mM, the addition of 5 mM glucose results in a rather small increase in power (Tab. 1b). The experiments demonstrate that the glucose concentration in serum is only a minor point with respect to the reduced performance found here in comparison to an artificial buffer solution. Nevertheless, even in a biological fluid reasonable power densities can be achieved compared to previous studies in pure buffer [9,14,42].Compared to other studies with EBFCs being investigated in human sera, no significant improvement concerning the power output can be obtained, but a rather good stability has been found for repeated measurements (no loss of maximum power within 3 repeated measurements).

4. Summary In the present study we have developed membraneless and mediatorless, glucose EBFCs based on (PQQ)GDH and BOD as biocatalysts and two different CNT architectures: buckypaper (BP) and vertical aligned CNTs (vaCNT). The separate evaluation of the anodes and cathodes shows a higher suitability of vaCNT as electrode interface for the enzymes compared to BP. For both systems a polyaniline based interlayer has been introduced for (PQQ)GDH coupling and a PQQ film for BOD fixation, thus allowing efficient electrochemical communication between the biocatalyst and the CNT surface. Maximal current densities about 1.3 ± 0.18 mA/cm2 (at 0.1 V vs. Ag/AgCl) for the (PQQ)GDH/PABMSA/vaCNT@Si-anode and about 550 µA/cm2 (at 0.1 V vs. Ag/AgCl) for the BOD/PQQ/vaCNT@Si-cathode can be achieved. For buckypaper as electrode material these values are 0.75 mA/cm2 for the anode ((PQQ)GDH/PAPMSA/BP) and 1 mA/cm2 for the cathode (BOD/PQQ/BP). Applying the developed electrodes in a fuel cell maximum power density of about 107 ± 5 µW/cm2 is detected for the bucky paper based system in (10 mM glucose). Covalent coupling of the enzyme to the polymer layer improves the stability of the system. The second biofuel cell is based on vaCNTs applying the same enzymes and modification steps. Here a power density of 122 ± 8 µW/cm2 (at 540 ± 50 mV) in presence of 10mM glucose can be generated, which is slightly higher than the BP based cell. Even on the third day of application of the cell a power density of more than 110 µW/cm2 is retained. This EBFC has also been applied in human serum samples. The results show a decreased power

12   

density of about 41 ± 7 µW/cm2. The diminished glucose serum concentration is only to a minor extent responsible for the lower performance in serum compared to buffer solutions.

Acknowledgements Financial support by the BMBF Germany is kindly acknowledged (project 03IS2201I). The authors want to thank Roche Diagnostics (Penzberg, Germany) and mainly Dr. Meier and Dr. von der Eltz for cooperation on supplying us with the GDH enzyme. We also thank the Karl und Marie Schaak-Stiftung, Frankfurt/Main, for generous financial support.

References [1]

Neto, S., Forti, J., Andrade, A. De. (2010). An overview of enzymatic biofuel cells. Electrocatalysis, (April), 87–94.

[2]

Brocato, S., Lau, C., Atanassov, P. (2012). Mechanistic study of direct electron transfer in bilirubin oxidase. Electrochim. Acta, 61, 44–49.

[3]

Karaśkiewicz, M., Nazaruk, E., Żelechowska, K., Biernat, J. F., Rogalski, J., Bilewicz, R. (2012). Fully enzymatic mediatorless fuel cell with efficient naphthylated carbon nanotube–laccase composite cathodes. Electrochem. Commun., 20, 124–127.

[4]

Falk, M., Blum, Z., Shleev, S. (2012). Direct electron transfer based enzymatic fuel cells. Electrochim. Acta, 82(0), 191–202.

[5]

Osman, M. H., Shah, A. A., Walsh, F. C. (2011). Recent progress and continuing challenges in bio-fuel cells . Part I : Enzymatic cells. Biosens. Bioelectron., 26(7), 3087–3102.

[6]

Barton, S. C. (2010). Enzyme catalysis in biological fuel cells. In Handbook of Fuel Cells. John Wiley Sons, Ltd.

[7]

Durand, F., Stines-Chaumeil, C., Flexer, V., André, I., Mano, N. (2010). Designing a highly active soluble PQQ-glucose dehydrogenase for efficient glucose biosensors and biofuel cells. Biochem. Biophys. Res. Commun., 402(4), 750–4.

[8]

Razumiene, J., Meškys, R. (2000). 4-Ferrocenylphenol as an electron transfer mediator in PQQ-dependent alcohol and glucose dehydrogenase-catalyzed reactions. Electrochem. Commun., 2, 307–311.

[9]

Tanne, C., Göbel, G., Lisdat, F. (2010). Development of a ( PQQ ) -GDH-anode based on MWCNT-modified gold and its application in a glucose / O 2 -biofuel cell. Biosens. Bioelectron., 26(2), 530–535. 13 

 

[10] Wettstein, C., Möhwald, H., Lisdat, F. (2012). Coupling of pyrroloquinoline quinone dependent glucose dehydrogenase to ( cytochrome c / DNA ) -multilayer systems on electrodes. Bioelectrochemistry, 88, 97–102. [11] Laurinavicius, V., Razumiene, J., Kurtinaitiene, B., Lapenaite, I., Bachmatova, I., Marcinkeviciene, L., Ramanavicius, A. (2002). Bioelectrochemical application of some PQQ-dependent enzymes. Bioelectrochemistry, 55, 29–32. [12] Sang, Y. K., Park, J., Lee, D., Kim, H. (2012). Electrochemical glucose biosensor by electrostatic binding of PQQ-glucose dehydrogenase onto self-assembled monolayers on gold. J. App. Electrochem., 383–390. [13] Razumiene, J., Vilkanauskyte, A., Gureviciene, V., Barkauskas, J., Meskys, R., Laurinavicius, V. (2006). Direct electron transfer between PQQ dependent glucose dehydrogenases and carbon electrodes: An approach for electrochemical biosensors. Electrochim. Acta, 51(24), 5150–5156. [14] Schubart, I. W., Göbel, G., Lisdat, F. (2012). A pyrroloquinolinequinone-dependent glucose dehydrogenase ( PQQ-GDH ) -electrode with direct electron transfer based on polyaniline modified carbon nanotubes for biofuel cell application. Electrocimica Acta, 82, 224–232. [15] Halamkova, L., Bocharova, V., Szczupak, A., Alfonta, L., Katz, E. (2012). Implanted Biofuel Cell Operating in a Living Snail. J.Am. Chem. Soc., 134, 8–11. [16] Göbel, G., Lisdat, F. (2008). Organic interlayers for oxygen reducing electrodes based on bilirubin oxidase and MWCNT modified gold. Electrochem. Commun., 10(11), 1691–1694. [17] Schubert, K., Goebel, G., Lisdat, F. (2009). Bilirubin oxidase bound to multi-walled carbon nanotube-modified gold. Electrochim. Acta, 54, 3033–3038. [18] Gupta, G., Lau, C., Rajendran, V., Colon, F., Branch, B., Ivnitski, D., Atanassov, P. (2011). Direct electron transfer catalyzed by bilirubin oxidase for air breathing gasdiffusion electrodes. Electrochem. Commun., 13(3), 247–249. [19] Miyake, T., Haneda, K., Yoshino, S., Nishizawa, M. (2013). Flexible layered biofuel cells. Biosens. Bioelectron., 40(1), 45–49. [20] Cinquin, P., Gondran, C., Giroud, F., Mazabrard, S., Pellissier, A., Porcu, P., Cosnier, S. (2010). A Glucose BioFuel Cell Implanted in Rats. PLoS ONE, 5(5), 1–7. [21] Javier, F., García-Céspedes, J., Muñoz, F. X., Bertrán, E. (2008). Vertically aligned carbon nanotube based electrodes : Fabrication , characterisation and prospects. Electrochem. Commun., 10, 1242–1245. [22] Liu, X., Baronian, K. H. R., Downard, A. J. (2009). Direct growth of vertically aligned carbon nanotubes on a planar carbon substrate by thermal chemical vapour deposition. Carbon, 47(2), 500–506.

14   

[23] Ahmadalinezhad, A., Wu, G., Chen, A. (2011). Mediator-free electrochemical biosensor based on buckypaper with enhanced stability and sensitivity for glucose detection. Biosens. Bioelectron., 30(1), 287–293. [24] Khabazian, S., Sanjabi, S. (2011). The effect of multi-walled carbon nanotube pretreatments on the electrodeposition of Ni–MWCNTs coatings. Appl. Surf. Sci., 257(13), 5850–5856. [25] Lee, S., Chang, Y.-P. (2009). Plasma treatment effects on surface morphology and field emission characteristics of carbon nanotubes. J. Mater. Sci., 20, 851–857. [26] Musameh, M., Lawrence, N. S., Wang, J. (2005). Electrochemical activation of carbon nanotubes. Electrochem. Commun., 7, 14–18. [27] Li, X., Du, Z., Zhang, C., Li, H., Zou, W. (2013). Preparation of polyaniline grafted multiwalled carbon nanotubes and conductive application in polyetherimide. Polym. Adv. Technol. Technologies, 24(January 2012), 151–156. [28] Liu, C., Huang, C., Wu, S., Han, J. (2010). Nanometer-thick patterned conductive films prepared through the self-synthesis of polythiophene derivatives. Polym. Int., 59(June 2009), 517–522. [29] Hsieh, G., Li, F. M., Beecher, P., Nathan, A., Wu, Y., Ong, S., Milne, W. I. (2009). High performance nanocomposite thin film transistors with bilayer carbon nanotubepolythiophene active channel by ink-jet printing High performance nanocomposite thin film transistors with bilayer carbon nanotube-polythiophene active channel by ink-jet pr. J. Appl. Phys., 123706(106). [30] Li, D., Wen, Y., He, H., Xu, J., Liu, M., Yue, R. (2012). Polypyrrole – Multiwalled Carbon Nanotubes Composites as Immobilizing Matrices of Ascorbate Oxidase for the Facile Fabrication of an Amperometric Vitamin C Biosensor. J. Appl. Polym. Sci., 126, 882–893. [31] Hussain, S. T., Abbas, F., Kausar, A., Khan, M. R. (2012). New polyaniline/polypyrrole/ polythiophene and functionalized multiwalled carbon nanotube-based nanocomposites: Layer-by-layer in situ polymerization. High Perform. Polym. (25), 70–78. doi:10.1177/0954008312456048 [32] Hussein, L., Urban, G., Kruger, M. (2011). Fabrication and characterization of buckypaper-based nanostructured electrodes as a novel material for biofuel cell applications. Phys. Chem. Chem. Phys., 13(13), 5831–5839. [33] Ivnitski, D., Artyushkova, K., Atanassov, P. (2008). Surface characterization and direct electrochemistry of redox copper centers of bilirubin oxidase from fungi Myrothecium verrucaria. Bioelectrochemistry, 74(1), 101–110. [34] Christenson, A., Shleev, S., Mano, N., Heller, A., Gorton, L. (2006). Redox potentials of the blue copper sites of bilirubin oxidases. Biochim. Biophys. Acta, 1757(12), 1634– 41.

15   

[35] Suraniti, E., Abintou, M., Durand, F., Mano, N. (2012). Heat and drying time modulate the O2 reduction current of modified glassy carbon electrodes with bilirubin oxidases. Bioelectrochemistry, 88, 65–9. [36] Nogala, W., Celebanska, A., Szot, K., Wittstock, G., Opallo, M. (2010). Bioelectrocatalytic mediatorless dioxygen reduction at carbon ceramic electrodes modified with bilirubin oxidase. Electrochim. Acta, 55(20), 5719–5724. [37] Kim, J., Parkey, J. (2009). Development of a biofuel cell using glucose-oxidase-and bilirubin-oxidase-based electrodes. J. Solid State Electrochem., 13, 1043–1050. [38] Ramírez, P., Mano, N., Andreu, R., Ruzgas, T., Heller, A., Gorton, L., Shleev, S. (2008). Direct electron transfer from graphite and functionalized gold electrodes to T1 and T2/T3 copper centers of bilirubin oxidase. Biochim. Biophys. Acta, 1777(10), 1364– 9. [39] Göbel, G., Schubart, I. W., Scherbahn, V., Lisdat, F. (2011). Direct electron transfer of PQQ-glucose dehydrogenase at modified carbon nanotubes electrodes. Electrochem. Commun., 13(11), 1240–1243. [40] Filip, J., Šefčovičová, J., Gemeiner, P., Tkac, J. (2013). Electrochemistry of bilirubin oxidase and its use in preparation of a low cost enzymatic biofuel cell based on a renewable composite binder chitosan. Electrochim. Acta, 87(0), 366–374. [41] Zebda, A., Gondran, C., Goff, A. Le, Cinquin, P. (2011). Mediatorless high-power glucose biofuel cells based on compressed carbon nanotube-enzyme electrodes. Nature Commun., 2, 370–376. [42] Coman, V., Ludwig, R., Harreither, W., Haltrich, D., Gorton, L., Ruzgas, T., Shleev, S. (2010). A Direct Electron Transfer-Based Glucose/Oxygen Biofuel Cell Operating in Human Serum. Fuel Cells, 10(1), 9–16 [43] Lee, J. Y., Shin, H. Y., Kang, S. W., Park, C., Kim, S. W. (2011). Application of an enzyme-based biofuel cell containing a bioelectrode modified with deoxyribonucleic acid-wrapped single-walled carbon nanotubes to serum. Enzyme Microb. Technol., 48(1), 80–84. [44] Flexer, V., Durand, F., Tsujimura, S., Mano, N. (2011).Efficient Direct Electron Transfer of PQQ-glucose Dehydrogenase on Carbon Cryogel Electrodes at Neutral pH. Anal. Chem., 83, 5721–5727. [45] Sarauli, D., Riedel, M., Wettstein, C., Hahn, R., Stiba, K., Wollenberger, U., Lisdat, F. (2012). Semimetallic TiO2 nanotubes: new interfaces for bioelectrochemical enzymatic catalysis. J. Mat.Chem., 22(11), 4615–4618. [46] Sarauli, D., Xu, C., Dietzel, B., Schulz, B., Lisdat, F. (2013). Differently substituted sulfonated polyanilines: The role of polymer compositions in electron transfer with pyrroloquinoline quinone-dependent glucose dehydrogenase. Acta Biomaterialia, 9(9), 8290–8298.

16   

[47] R. Joshi. O. Yilmazoglu, J.J. Schneider, D. Pavlidis (2010).Patterned growth of ultra long carbon nanotubes. Properties and systematic investigation into their growth process. J. Mater. Chem. 20, 1717-1721. [48] R. K. Joshi, J. J. Schneider (2012).Assembly of one dimensional nanostructures into functional 2D and 3D architectures. Synthesis, arrangement and functionality. Chem. Soc. Rev. 41 (15), 5285-5312. [49] Ammam, M., Fransaer, J. (2010). Micro-biofuel cell powered by glucose/O2 based on electro-deposition of enzyme, conducting polymer and redox mediators: Preparation, characterization and performance in human serum. Biosens. Bioelectron., 25(6), 1474– 1480. [50] Rengaraj, S., Kavanagh, P., & Leech, D. (2011). A comparison of redox polymer and enzyme co-immobilization on carbon electrodes to provide membrane-less glucose/O2 enzymatic fuel cells with improved power output and stability. Biosensors & Bioelectronics, 30(1), 294–9. [51] Wang, X., Falk, M., Ortiz, R., Matsumura, H., Bobacka, J., Ludwig, R., Shleev, S. (2012). Mediatorless sugar/oxygen enzymatic fuel cells based on gold nanoparticlemodified electrodes. Biosensors and Bioelectronics, 31(1), 219–225.                                 17   

Tab. 1: Key parameters obtained from vaCNT-based enzymatic biofuel cells based on a (PQQ)GDH/PABMSA/vaCNT-anode and a BOD/PQQ/vaCNT-cathode (a) within three days. (EBFCs (n = 3), 10 mM glucose in 100 mM CiP + 1 mM CaCl2, pH 7 quiescent, air-saturated solution, galvanodynamic measurements at a scan rate of 3nA/s) and (b) in human serum samples. (EBFCs (n = 3), successive addition of 5 mM and 10 mM glucose, air-saturated, quiescent solution, galvanodynamic measurements with a scan rate 3 nA/s). (a) day max. power density [µW/cm2] potential at power maximum [mV] open circuit potential (OCP) [mV] current density at power maximum [µA/cm2]

1

2

3

122 ± 8

142 ± 48

113 ± 26

540 ± 50

519 ± 35

447 ± 21

707 ± 68

707 ± 38

690 ± 14

228 ± 5

273 ± 74

254 ± 71

human serum + 10 mM glucose

(b)

solution

human serum

human serum + 5 mM glucose

max. power density [µW/cm2]

41 ± 7

50 ± 5

51 ± 5

391 ± 70

418 ± 20

391 ± 23

660 ± 43

650 ± 24

620 ± 20

potential at power maximum [mV] open circuit potential (OCP) [mV]

18   

 

‐ ‐ ‐

We have deveeloped membbrane-less annd mediator-less, glucosse enzymaticc fuel cell W baased on (PQQ)GDH andd BOD as bioocatalysts Tw wo differentt carbon nan notube archittectures havee been used aas interface: buckypaperr annd verticallyy aligned carbbon nanotubbes Thhe applicatio on of the devveloped fuell cells allowss to generatee electrical power using gllucose as bio ofuel

Scheme 1: chemicaal structure of (a) pooly(3-aminobbenzoic acid-co-2-methhoxyaniline-5sulfonic acid) a – PABMSA and (b b) pyrroloquuinoline quin none - PQQ.

19

Captions  

Fig. 1: (a) photographical image of two vaCNT/Si chips (size 5 mm x 1.5 mm); (b) scanning electron micrograph (SEM) of the block-patterned CNT structures; (c) transmission electron micrograph shows double-walled CNTs with a diameter of 6-10 nm.; (d) Raman spectra of the block-patterned carbon nanotubes  Fig. 2: Cyclic voltammograms of (a) buckypaper/PABMSA-electrodes: 1) 0 mM glucose, 2) 10 mM glucose, 3) with adsorbed (PQQ)GDH and 0mM glucose, 4) with adsorbed (PQQ)GDH and 10 mM glucose. (b) Voltammetric current responses of (PQQ)GDH/PABMSA/buckypaper-electrodes in presence of different glucose concentration at 0.1 V vs. Ag/AgCl, 1 M KCl. Current density changes in dependence on different PABMSA-concentrations used for preparation of the electrodes: 1) 0,2 mg/ml; 2) 1 mg/ml; 3) 2 mg/ml; 4) 3 mg/ml; 5) 5 mg/ml. Buffer conditions: 5 mM MES + 1 mM CaCl2 pH 6.5, scan rate 10 mV/s. (c) current density changes in dependence on different buffer conditions during the measurement (using a PABMSA-concentration of 5 mg/ml for preparation of the (PQQ)GDH electrodes, n = 3): 1) 5 mM MES + 1 mM CaCl2 pH 6.5 and 2) 100 mM CiP + CaCl2 pH 7, scan rate 10 mV/s. (d) Cyclic voltammograms of (PQQ)GDH/ PABMSA/vaCNT/Si-electrodes: 1) 0 mM glucose, 2) 10 mM glucose, PABMSAconcentration during preparation 5 mg/ml, 5 mM MES+1 mM CaCl2 pH 6.5. Scan rate 10 mV/s. (e) Dependence of the glucose concentration on catalytic current. Buffer conditions: 100 mM CiP + 1 mM CaCl2 pH 6.5. Scan rate 10 mV/s.  Fig. 3: Linear sweep voltammogram of (a) 1) BOD/buckypaper-electrode, BOD adsorbed, Ar-saturated buffer; 2) BOD/buckypaper, BOD adsorbed, air-saturated buffer; 3) BOD/PQQ/buckypaper, [PQQ conc for adsorption]= 1 mM, BOD adsorbed, air-saturated buffer; 4) BOD/PQQ/buckypaper, [PQQ conc. for adsorption] = 1 mM , BOD covalently attached, air-saturated buffer; 5) BOD/PQQ/buckypaper, [PQQ conc. for adsorption] = 2,73 mM , BOD covalently attached, air-saturated buffer and (b) of BOD/PQQ/vaCNT/Sielectrode ([PQQ conc. during adsorption] = 2.73 mM) in 1) Ar-saturated buffer and 2) in airsaturated buffer. Buffer conditions: 100 mM CiP pH 7, quiescent solution. Scan rate 10 mV/s. Fig. 4: Performance curves of the established biofuel cell based on (a) buckypaper (BP) with a (PQQ)GDH/PABMSA/BP-anode and a BOD/PQQ/BP-cathode obtained from galvanodynamic measurements at a scan rate of 3 nA/s: Cell potential and power density in dependence on current density and (b) fuel cell based on vertically aligned CNTs with a (PQQ)GDH/PABMSA/vaCNT/Si-anode and a BOD/PQQ/vaCNT/Si-cathode obtained from galvanodynamic measurements at a scan rate of 3 nA/s: Cell potential and power density in dependence on current density. Measuring conditions: quiescent, air-saturated 100 mM CiP + 1 mM CaCl2, pH 7, 10 mM glucose.

20   

Fig. 1

21

Fig. 2

22   

Fig. 3

23   

Fig. 4

           

24