UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Development of a regenerable glucose biosensor probe for bioprocess monitoring Phelps, Michael R. 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1993_fall_phelps_michael.pdf [ 5.44MB ]
JSON: 831-1.0058516.json
JSON-LD: 831-1.0058516-ld.json
RDF/XML (Pretty): 831-1.0058516-rdf.xml
RDF/JSON: 831-1.0058516-rdf.json
Turtle: 831-1.0058516-turtle.txt
N-Triples: 831-1.0058516-rdf-ntriples.txt
Original Record: 831-1.0058516-source.json
Full Text

Full Text

DEVELOPMENT OF A REGENERABLE GLUCOSE BIOSENSORPROBE FOR BIOPROCESS MONITORINGMICHAEL R. PHELPSB. Eng., The Royal Military College of Canada, 1990A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF CHEMICAL ENGINEERINGWe accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIAOctober 1993© Michael R. Phelps, 1993In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my v■ ,rittenpermission.(Signature) Department of Chemical EngineeringThe University of British ColumbiaVancouver, CanadaDate  September 28, 1993DE-6 (2188)ABSTRACTThe implementation and commercialization of enzyme-based biosensors for on-linebioprocess monitoring and control has been slowed by problems relating to the in situsterilizability of the probe and the stability of the enzyme component. A novel technologyis presented here which addresses both of these difficulties. The approach is based on thereversible immobilization of enzymes conjugated with the cellulose binding domain (CBD)of cellulases from Cellulomonas fimi. A regenerable biosensor probe is configured with acellulose matrix onto which the solubilized enzyme-CBD conjugate can be repeatedlyloaded (via the attachment of the CBD) and subsequently eluted by perfusing the cellulosematrix with the appropriate loading or eluting solution.The chemical conjugation of the enzyme glucose oxidase (GOx) with CBD byglutaraldehyde is described. The GOx-CBD conjugate retained the enzymatic activity ofthe glucose oxidase and the binding affinity of the CBD. The GOx-CBD conjugate wasused in an experimental glucose biosensor based on a platinum rotating disk electrodefitted with a cellulose immobilization matrix to demonstrate the feasibility of multiplecycles of loading and elution of the conjugate and to develop suitable protocols andreagents for the loading and elution procedures. A prototype glucose biosensor andreagent flow system were designed and built for use in fermentation monitoring. Acustom-designed membrane system consisting of a sterilizable, glucose-permeable outerNafion membrane for the sensor and a cellulose acetate coating on the indicating electrodewas developed for use in a microbial fermentation. The prototype glucose biosensor wasused successfully to monitor medium glucose concentration for 16.5 continuous hoursduring a 20 L fed-batch cultivation of E. coli in minimal medium. Michaelis-Mentenenzyme kinetics were used as an empirical model for the calibration of the experimentalbiosensor. The development of a computer-controlled prototype glucose biosensor andfermentation monitoring system is discussed.iiThese results are the first to demonstrate the concept, feasibility, and utility of aregenerable biosensor based on reversible immobilization of the enzyme using CBDtechnology and represent a significant step toward better instrumentation for fermentationmonitoring and control.iiiTABLE OF CONTENTSPageABSTRACT^ iiTABLE OF CONTENTS^ ivLIST OF TABLES viiLIST OF FIGURES^ viiiACKNOWLEDGEMENTS xINTRODUCTION^ 1CHAPTER 1 LITERATURE REVIEW^ 41.1 BIOSENSOR DEVELOPMENT 41.1.1 Introduction^ 41.1.2 Enhancing Long Term Stability^  81.1.3 Cellulose Binding Domain Technology  101.2 GLUCOSE MONITORING AND CONTROL^ 121.2.1 Optimization of Bioprocesses^  121.2.2 Strategies for Glucose Monitoring and Control^ 141.2.3 In situ Enzyme Electrode Probes^  18CHAPTER 2 BACKGROUND AND THEORY 222.1 INTRODUCTION^ 222.2 ENZYME ELECTRODE COMPONENTS^ 222.3 BIOCHEMICAL CONSIDERATIONS 242.4 ELECTROCHEMICAL CONSIDERATIONS^ 262.5 MODELLING CONSIDERATIONS^ 282.5.1 Assumptions and Approximations 302.5.2 Diffusion Equations^ 322.5.3 Enzyme Reaction-Rate Equations^ 33iv2.5.4 The General Modelling Approach^ 372.5.5 Multi-layer Modelling^ 392.6 PRINCIPLE OF OPERATION 41CHAPTER 3 SYNTHESIS AND CHARACTERIZATION OF THE GLUCOSEOXIDASE - [CELLULOSE BINDING DOMAIN] CONJUGATE^ 463.1 INTRODUCTION^ 463.2 MATERIALS AND METHODS^ 463.2.1 Enzymes and Chemicals 463.2.2 Enzyme-CBD Conjugations^ 473.2.3 Total Protein and Enzyme Activity Assays^ 483.2.4 Gel Electrophoresis and Immunoblotting 493.3 RESULTS AND DISCUSSION^ 49CHAPTER 4 DEMONSTRATION OF THE FEASIBILITY OF A REGENERABLEGLUCOSE BIOSENSOR^ 544.1 INTRODUCTION 544.2 MATERIALS AND METHODS^ 544.2.1 Instrumentation^ 544.2.2 Experimental Glucose Biosensor Preparation^ 554.3 RESULTS AND DISCUSSION^ 56CHAPTER 5 DESIGN AND CHARACTERIZATION OF THE EXPERIMENTALGLUCOSE BIOSENSOR PROTOTYPE^ 665.1 INTRODUCTION^ 665.2 MATERIALS AND METHODS^ 665.2.1 Prototype Design and Construction^ 665.2.2 Prototype Characterization Experiments 775.2.3 Glucose Monitoring During Fed-Batch Cultivation of E. coli^ 805.3 RESULTS AND DISCUSSION^ 815.3.1 Prototype Characterization 825.3.2 Glucose Monitoring During Fed-Batch Cultivation of E. coli^ 99CHAPTER 6 CONCLUSIONS^ 1116.1 CONCLUDING REMARKS^ 1116.2 FUTURE WORK^ 114NOMENCLATURE SUMMARY 117REFERENCES ^ 119APPENDIX SAMPLE STRIP CHART RECORD^ 130viLIST OF TABLESTable 1.1^Industrially important enzymatic assays requiring oxidase enzymes.^ 6Table 3.1^Specific activity of various samples of soluble GOx-CBD conjugate. ^ 53Table 4.1^Properties of the cellulose matrices used in the experimental glucosebiosensor system. ^ 59Table 4.2^Apparent enzyme kinetic data derived from the calibration data ofFigure 4.2^ 63Table 5.1^Characteristics of different membranes tested as potential outer membranesfor the glucose biosensor prototype.^ 84Table 5.2^Prototype sensor calibration constants for the first and second enzymeloadings^  105viiLIST OF FIGURESFigure 1.1^Schematic representation of the cellulase exoglucanase from C. fimi. ^ 10Figure 2.1^Schematic diagram of a glucose sensitive enzyme electrode. ^ 23Figure 2.2^Schematic diagram of a biosensor based on CBD-immobilized enzymes ^ 43Figure 2.3^Process flow diagram demonstrating the principle of operation of afermentation monitoring system using the renewable biosensor probe. ^ 43Figure 3.1^A. Gel electrophoresis (SDS-PAGE) of the GOx-CBD conjugate.B. Western blot of the gel in A. ^ 51Figure 4.1^Typical calibration data for the experimental glucose biosensor^ 58Figure 4.2^Calibration data for multiple cycles of loading and elution of the GOx-CBDconjugate^ 60Figure 4.3^Lineweaver-Burk plot of the calibration data in Figure 4.2.^ 61Figure 5.1^Construction of the Ingold CO2 probe^ 67Figure 5.2^Diagram of the internal electrode assembly. 71Figure 5.3^Stainless steel adapter for mounting the internal electrode unit.^ 73Figure 5.4^Schematic diagram of the biosensor prototype showing the reagent flowsystem and instrumentation ^ 75Figure 5.5^Stainless steel dummy electrode 78Figure 5.6^Comparison of the response of different cellulose acetate coated Ptelectrodes to hydrogen peroxide.^ 87Figure 5.7^Normalized calibration data for the prototype biosensor in PBS and Luriabroth (L-B) using different membranes. ^ 89Figure 5.8^Sensor equilibration time after insertion in Luria broth using differentmembranes.^ 92viiiFigure 5.9Figure 5.10Figure 5.11Figure 5.12Figure 5.13Figure 5.14Figure 5.15Figure 5.16Normalized calibration data for the prototype sensor in PBS and Luriabroth (L-B) using the Nafion membrane before and after autoclaving ^ . 94Effect of temperature on the sensor signal at steady-state^ 95Effect of medium pH on the sensor signal at steady-state.^ 95Effect of medium dissolved oxygen tension on the sensor signal at steady-state.^ 97Time-course of the fermenter variables during fed-batch cultivation of E.coli in a 20 L fermenter. ^  102Medium glucose concentration measured by the prototype glucose sensorand the Beckman off-line glucose analyzer during fed-batch cultivation ofE. coli in minimal medium (M-9) in a 20 L fermenter.^ 103Cross-correlation plots of the prototype glucose sensor output and theBeckman off-line glucose analyzer results^  107Ratio of the Beckman glucose analyzer results and the prototype glucosesensor output calculated at various points in time after fresh enzyme wasloaded.^  110ixACKNOWLEDGEMENTSI would like to thank my supervisor, Dr. Robin Turner, for his helpful advice andguidance throughout the course of my thesis work, and for his support and encouragementwhen it was needed. I would also like to thank the members of my supervisorycommittee, Dr. Doug Kilburn, Dr. John Hobbs and Dr. Ken Pinder, for their usefuldiscussions and examination of this thesis.I would also like to acknowledge the expert technical assistance of the technicians,students, and staff in the Biotechnology Laboratory and the Department of ChemicalEngineering at the University of British Columbia. I am especially grateful to Dr. EdgarOng and Dr. Andrew Wierzba for their technical assistance in purifying and characterizingthe GOx-CBD conjugate, to Gary Lesnicki for his help in the fermenter pilot plant, and toDiane Hasenwinkle and Eric Jervis for their assistance and advice regarding the cultivationof E. coli in minimal medium.Finally, I would like to acknowledge the financial support for this work providedby the Natural Sciences and Engineering Research Council of Canada (NSERC).xINTRODUCTIONThe increasing commercial importance of bioprocesses has stimulated research infermentation monitoring in order to optimize the performance of bioreactors. Thespecificity and selectivity provided by the biological component of biosensors offerenormous potential, in principle, for continuous, on-line analysis in complex fermentationmedia. Numerous examples of amperometric enzyme electrode biosensors have beendescribed in the literature (Brooks et al., 1991; Mascini and Palleschi, 1989). However,the development and implementation of these biosensors for bioprocess control has beenslowed by problems relating to the sterilizability and stability of enzyme-based probes inbioreactors. In this thesis, a novel enzyme immobilization technology is used in thedevelopment of a biosensor for bioprocess control which potentially addresses both ofthese shortcomings.The new technology is based on the reversible immobilization of enzymes viaconjugation with the cellulose binding domain (CBD) of cellulases from Cellulomonasfimi. These cellulases have a modular structure consisting of two or more structurallyseparate domains (Kilburn et al., 1992). The binding domain functions independently ofthe catalytic domain and can be chemically or genetically conjugated to other proteins(e.g. enzymes) which then bind strongly to cellulose. Under the appropriate solutionconditions, the binding can be disrupted and the conjugate protein eluted from thecellulose matrix. In this work, the CBD is chemically conjugated to glucose oxidase todevelop a regenerable glucose biosensor using reversibly immobilized enzyme.Briefly, the hardware for the regenerable biosensor system consists of a platinumindicating electrode, a porous cellulose matrix, and a protective dialysis membrane, allincorporated into a stainless steel probe for insertion into the bioreactor. The rate ofH202 evolution from the enzyme-catalyzed oxidation of glucose is measuredamperometrically at the platinum electrode. The cellulose matrix for immobilization of the1enzyme-CBD conjugate is incorporated into the enzyme chamber of the sensor body,sandwiched between the surface of the indicating electrode and the dialysis membrane.Inlet and outlet tubing in the probe body allow perfusion of the cellulose matrix with theenzyme-CBD conjugate solution and/or the elution buffer. The basic design is similar tothat presently employed in commercial CO2 probes (Ingold, 1990).After the steam sterilization of the bioreactor and the probe body, the sensor isloaded by perfusing the cellulose matrix with the enzyme-CBD conjugate solution,resulting in attachment of the enzyme via the CBD. The sensor is calibrated byinnoculation of the fermenter, and an internal calibration check can be performedperiodically during the fermentation. If the enzyme activity deteriorates to anunacceptable degree, the sensor can be regenerated without interrupting the fermentationby perfusing the cellulose matrix with elution buffer to remove the attached enzyme. Thesensor is then reloaded with enzyme-CBD conjugate as before, and recalibrated tocontinue monitoring the fermentation.The advantage of this design is that the complete process of diagnosis,regeneration, and recalibration could potentially be performed in situ and under computercontrol. The applications of the system can be expanded by conjugating the CBD to otherenzymes, such that the sensor hardware could be used for monitoring a variety of differentanalytes.The primary purpose of this work is to develop the technology for regenerablebiosensors based on enzyme-CBD conjugates. Specifically, the objectives of this thesisare:1. To synthesize and characterize a chemical conjugate of glucose oxidase andcellulose binding domain.2. To demonstrate the feasibility of the concept of a regenerable glucosebiosensor based on the GOx-CBD conjugate protein.23. To demonstrate the potential of the CBD technology for the developmentof on-line bioprocess sensors by designing, constructing, and testing an experimentalbiosensor prototype that can be used for glucose monitoring during a microbialfermentation.3CHAPTER 1 LITERATURE REVIEW1.1 BIOSENSOR DEVELOPMENT1.1.1 IntroductionThe Clark oxygen electrode (Clark, 1956) can be said to be the cornerstone ofmodern biosensor technology. The electrode detected dissolved oxygen polarographicallyand was used for monitoring blood dissolved oxygen levels in patients during surgery(Schultz, 1991). In 1962, Clark and Lyons first conceived of a glucose sensor based onthe Clark oxygen electrode (Clark and Lyons, 1962). Glucose oxidase was immobilized ina gel on the surface of the oxygen electrode, and the rate of consumption of oxygenduring the enzymatic oxidation of glucose could be related to the blood glucoseconcentration. Updike and Hicks coined the term "enzyme electrode" in 1967 anddeveloped the idea one step further (Updike and Hicks, 1967) by incorporating an oxygenelectrode without immobilized enzyme into the sensor system. A differential measurementcould then be performed to correct for variation of the background oxygen concentration.In 1969, Guilbault designed an enzyme electrode for the measurement of urea in bodyfluids using the enzyme urease (Guilbault and Montalvo, 1969). The ammonium ionproduced by the enzyme-catalyzed reaction of urea was detected potentiometrically usingan ion selective electrode (ISE). In 1970, Clark patented the idea of using a platinumelectrode for the amperometric detection of hydrogen peroxide produced in the oxidationof glucose by glucose oxidase (Clark, 1970). This development became the basis for thecommercially available laboratory glucose analyzer marketed by the Yellow SpringsInstrument Company (Yellow Springs, Ohio, U.S.A.).4Since the enzyme electrode was first conceived, a significant interest has developedin the field of biosensors because of the simplicity and selectivity of these sensors. Adramatic increase in biosensor interest began in the 1980's, evident by the publication of anew international journal Biosensors (Stoecker and Yacynych, 1990). The rapid growthof biotechnology in the last decade has now created a demand for more and better on-linesensors that can be interfaced with computers to control and optimize bioprocesses.Numerous enzyme electrode configurations have been published, and the amount ofresearch activity today is too great to be covered in a single review (Freitag, 1993).Present-day applications of biosensors can be found in industrial bioprocess monitoring,environmental monitoring, the food and drink industry, and clinical and in vivoapplications in medicine (Schultz, 1991; Reach and Wilson, 1992).The work in this thesis will concentrate on biosensors for fermentation monitoringand industrial bioprocess control. The focus will be on the development of the technologyfor a regenerable glucose biosensor using glucose oxidase (GOx) because of theimportance of glucose as the main carbon source and growth-limiting substrate inindustrial fermentations (Filippini et al., 1991; Huang et al., 1991; Freitag, 1993).However, it should be understood that the technology developed here could, in principle,be used in conjunction with numerous other oxidase enzymes to monitor other industriallyimportant analytes (see Table 1.1).The oxidation of glucose by glucose oxidase in the presence of oxygen results inthe production of gluconolactone and hydrogen peroxide. A number of differenttransducers, such as electrochemical detectors, optical detectors, and calorimeters, havebeen used to measure the rate of the enzyme-catalyzed reaction and provide a useableelectrical signal for further analysis. Calorimetry is a highly versatile technique as it can beapplied to virtually any enzymic reaction (Guilbault and Luong, 1989). However, thesensitivity and measuring range are relatively low, depending on the heat output of the5Table 1.1. Industrially important enzymatic assays requiring oxidase enzymes. Thegeneral oxidase enzyme-catalyzed reaction is of the form:Substrate(reduced form) + 02 Oxidase > Product(oxid ized form) + H202SUBSTRATE(Reduced Form)f3-D-GlucoseL-LactateEthyl AlcoholLactoseGlycerolCholesterolPyruvateUric AcidAcetaldehydeXanthineCholineL-GlutamateAcetylcholine*L-Glutamine*Maltose*Starch *Sucrose*ENZYMECATALYSTGlucose OxidaseL-Lactate OxidaseAlcohol OxidaseGalactose OxidaseGalactose OxidaseCholesterol OxidasePyruvate OxidaseUricaseAldehyde OxidaseXanthine OxidaseCholine OxidaseL-Glutamate OxidaseAcetylcholine EsteraseL-GlutaminaseGlucoamylaseAmyloglucosidaseInvertase + MutarotasePRODUCT(Oxidized Form) 8.-GluconolactonePyruvateAcetaldeyhdeGalactose DialdehydeGlyceraldehyde4-cholesten-3-oneAcetyl PhosphateAllantoinAcetateUrateBetainea-KetoglutarateCholineL-Glutamate13-D-Glucose13-D-Glucosef3-D-Glucose* Assays that require a multi-enzyme system.6given reaction, and difficulties have been encountered ensuring that the referencetemperature is constant (±0.01 °C) (Chaplin and Bucke, 1990). Huang et al. (1991) havecoupled the luminescent reaction of luminol and H202 in the presence of horseradishperoxidase with the glucose oxidase catalyzed oxidation of glucose and used an opticaldetector for signal transduction. The advantages of optical sensors are high sensitivity,stable calibration, and no requirement for a reference electrode. However, opticaldetectors are expensive and are not suitable for use as on-line, in situ, sensors due to theinterference from turbid, coloured media and the effect of background light.Electrochemical transducers are most commonly used for reasons of simplicity, highsensitivity, and low cost. Many examples have been cited in the literature (Kobos, 1980;Mascini and Palleschi, 1989; Hendry et al., 1990). Signal transduction can be based oneither the amperometric detection of oxygen consumption or hydrogen peroxideproduction, or the potentiometric determination of the local pH change due to theproduction of gluconic acid. The greatest drawback of the potentiometric method is thatthe quantification of pH change necessitates a weakly buffered measurement solution if asignificant change in pH is to be observed. In addition, the sensitivity of potentiometricsensors is governed by the Nernst equation, which dictates a logarithmic dependence ofthe signal on the hydrogen ion concentration. The biosensor developed in this thesisemploys the principle of amperometric detection of hydrogen peroxide production at aplatinum indicating electrode. The electrical current output from amperometric sensors isrelated to the rate of reaction by Faraday's law and the sensor response will, in theory, belinear. In addition, amperometric sensors are not as sensitive to changes in medium pH aspotentiometric sensors and can be used in buffered media.71.1.2 Enhancing Long Term StabilityDespite the number of biosensor research papers published each year in thescientific literature, relatively few biosensors are commercially available. Unfortunately,many practical problems remain which have prevented the widespread application ofbiosensors under real conditions and these have not been successfully addressed in theliterature to date. The development and commercialization of biosensors has been slowedby problems of instability and drift of the sensor signal, narrow measuring range for theanalyte, and long response times.The problems of long term stability can be attributed to changes in the enzymecomponent, such as inhibition or deactivation by components of the analyte medium. Driftof the sensor signal over time may be due to time-dependent changes in the sensorcalibration constants, which are caused by membrane fouling or electrode poisoning.Many of these issues can be alleviated to some extent by careful selection of the sensormembrane(s). Permselective membranes for amperometric biosensing have been reviewedin the literature (Wang, 1992). In addition, the sensor can be recalibrated periodically tocorrect for changes in the calibration constants. The denaturation of the enzyme over timeis irreversible, however, and although the lifetime of the enzyme may be prolonged insome cases, replacement of the enzyme when the activity has degraded to anunsatisfactory degree will eventually be necessary. The capability to replace the enzymecomponent of the sensor would not only extend the sensor operating lifetime but wouldallow for the substitution of other enzymes in order to change the analyte specificity of thesensor.A few sensor systems have been described in the literature with the capacity forenzyme replacement. Brooks et al. (1987/88) and Bradley and Schmid (1991) havedescribed the immobilization of the enzyme on graphite discs which could be replacedmanually. In the ideal case, the enzyme would be reversibly immobilized, such that the8enzyme could be exchanged in situ, without dismantling the sensor or interrupting thefermentation.Reversible enzyme immobilization techniques have been reported in the literature.Pieters and Bardeletti (1992) described the immobilization of enzymes to magnetic beads,which could then be manipulated using magnetic fields. The technique has been used inwaste-water treatment, affinity separation processes, cell sorting, immunoassays, and drugdelivery (Pieters and Bardeletti, 1992). Miyabayashi et al., (1989) reported apotentiometric enzyme electrode using chymotrypsin deposited on magnetic particles,which were then trapped in a magnetic field at the indicating electrode. The time to reachsteady state response was 30 minutes. The immobilization of glucose oxidase to magneticparticles was described by Pieters and Bardeletti (1992), and it may be possible to applythis technique to an amperometric glucose biosensor.Glucose oxidase has been reversibly immobilized in an enzyme reactor coupled toa flow injection analysis system using a chain of biospecific reactions based on the bindingof biotin-labelled (i.e., biotinylated) antibodies to an avidin coated matrix (de Alwis andWilson, 1989). Biotinylated antibodies and streptavidin-labelled reporter enzymes (suchas alkaline phosphatase or horseradish peroxidase) are commercially available, and thebiotin/streptavidin system has been exploited for immunoassays (Brillhart and Ngo, 1991).de Alwis and Wilson covalently attached avidin to a packed column, followed byattachment of biotinilated anti-glucose oxidase antibodies, which would then bind glucoseoxidase to the reactor column. The high affinity and strength of the avidin-biotin linkage(Kd = 1045 M4 ) was considered to be irreversible in this case, although the bindingcould be disrupted by 6 M guanidine, pH 1.5. The enzyme was eluted from the columnwithout affecting the avidin-biotin linkage by disrupting the antibody-enzyme bond using0.1 M phosphate buffer, pH 2.0, and fresh enzyme could then be attached by washing thecolumn with a solution of glucose oxidase. The enzyme could be loaded and elutedrepeatedly.91.1.3 Cellulose Binding Domain TechnologyThe glucose biosensor developed in this thesis uses a novel technique for thereversible immobilization of enzymes based on the cellulose binding domain (CBD) of thecellulases from C. fimi. These cellulases have been shown to consist of two or morestructurally distinct and independently functioning catalytic and binding domains (Gilkes etal., 1991). The cellulose binding domain of the cellulase exoglucanase, shown in Figure2.2, has been expressed in E. coli to obtain the isolated CBD polypeptide (CBDc ex) (Onget al., 1993). Proteins conjugated to CBD (by genetic or chemical methods) haveacquired the ability to bind reversibly to cellulose. The exact nature of the bindingmechanism has not been determined, however the binding to cellulose has been reportedto be virtually instantaneous. Previously published CBD binding studies indicated thatadsorption of the CBD to cellulose was complete within the shortest incubation timefeasible under the conditions of the experiment (i.e., 0.2 minutes) (Gilkes et al., 1992), butthe actual adsorption kinetics are likely much faster.Proline - ThreonineLinkerCatalyticDomainCOOH316 335 I 443Cellulose BindingDomainFigure 1.1:^Schematic representation of the cellulase exoglucanase from C. fimi. Thediagram shows the two structurally distinct and independent domains which can becleaved at the proline-threonine linker.10For example, a genetically engineered 13-glucosidase-CBDc ex fusion protein (Onget al., 1989; Ong et al., 1991) was shown to be as active as the native enzyme and retainedmore than 40% of the enzyme activity when bound to cellulose. In a cellulose columnperfused continuously with substrate, no activity loss was observed over 10 days ofoperation at 37 °C. In addition, binding to cellulose was stable for prolonged periods oftime at temperatures up to at least 70 °C, at ionic strengths from 10 mM to greater than 1M, and at pH values below 8. Binding could be reversed by distilled water, 1 M NaOH,or 8 M guanidinium HC1.The CBD technology has been used for the immobilization of enzymes or otherproteins, and as an affinity tag for the purification of recombinant proteins using cellulosecolumns (Kilburn et al., 1992). A genetically engineered Protein A-CBD conjugate hasbeen constructed which is stably bound to paper and dried. The paper strips can then beused to bind antibodies in immunoassays or other diagnostic tests. A conjugate proteinwas also constructed from CBD and IL-2 with a factor X protease cleavage site in themiddle. The conjugate protein can be purified in single step by affinity chromatography oncellulose, and then the IL-2 can be cleaved from the conjugate, either in solution or boundto cellulose. Other proposed applications of the CBD technology include the binding ofCBD-conjugated dyes or ink to cellulose-based textiles or paper, and the characterizationof cellulose fibre structure using fluorescently labeled CBD.In the application presented here, the CBD from C. fimi exoglucanase (CBDCex)is chemically conjugated to glucose oxidase using glutaraldehyde. An experimentalglucose biosensor is constructed by incorporating a cellulose matrix adjacent to thesurface of a platinum electrode, such that the GOx-CBD conjugate can be reversiblyimmobilized on the electrode. The enzyme is eluted from the cellulose by washing with asuitable elution buffer, and then reloaded using a fresh solution of GOx-CBD conjugate.The enzyme can be loaded and eluted repeatedly in this manner.111.2 GLUCOSE MONITORING AND CONTROL1.2.1 Optimization of BioprocessesThe optimization of the performance of bioprocesses depends on the ability tomonitor and control the parameters which describe the bioreactor environment. Atpresent, only a few parameters can be reliably monitored on-line (e.g. temperature, pH,dissolved oxygen tension, stir rate) without the use of highly sophisticated and costlyinstrumentation. The analysis of fermentation substrates, products, and metabolites isusually achieved by off-line methods (Brooks et al., 1987/88). However, optimal controlof a bioprocess requires that measurable parameters be determined as frequently aspossible, which in turn requires frequent sampling that increases the risk of contamination(Huang et al., 1991). Furthermore, off-line methods are usually too slow to be used in aclosed-loop control system, and it is often difficult to ensure that samples are notsignificantly degraded or changed during the sampling/analysis procedure. A sensorsystem based on an in situ probe which could provide continuous, real-time analysis wouldbe extremely valuable, particularly for high-density, fed-batch processes. Biosensors haveenormous potential (in principle) as in situ probes for the analysis of complex fermentationmedia due to the specificity and selectivity of the biological component of the sensor.The potential for the development of new feedback control strategies and adaptivecontrol techniques for the optimization of growth rate and protein production in fed-batchprocesses has motivated research and development of new, reliable, on-line sensors forfermentation monitoring. The economical commercial production of recombinant proteinsnecessitates the development of engineering strategies for the maximization of thevolumetric productivity of expressed products (Hardjito et al., 1993). To do so requiresthe optimization of gene expression in a high density cell culture. Biochemical engineershave become particularly interested in fed-batch growth techniques because of the12potential to separate the phases of cell growth and cloned-gene expression (Patkar andSeo, 1992) compared to batch or continuous culture. It can be shown that the productionof recombinant proteins can be optimized by maximizing the biomass yield on substrate toobtain a high cell density, and then maximizing cell specific productivity through high ratesof gene expression. High biomass yield on a given substrate can be obtained bycontrolling metabolism and minimizing the excretion of inhibitory metabolites through theregulation of substrate level (Smith and Bajpai, 1985). Due to the importance of glucoseas the main carbon and energy source for microbial growth in industrial fermentations, itwould be desirable to develop a glucose monitoring and control system in order to operatebioprocesses under optimum conditions.For instance, the specific growth rate of a hybridoma cell line (AFP-27) was foundto be directly related to the glucose concentration in continuous culture under glucose-limited conditions (Frame and Hu, 1991). However, in fermentations of the yeast S.cerevisiae, cell mass yield has been found to be higher at low glucose concentrations(Patkar and Seo, 1992). At high glucose levels, limitations of the respiratory capacity ofcertain yeasts results in ethanol production, even under aerobic conditions (Hardjito et al.,1993). Although ethanol produced by the fermentative metabolic pathway can be utilizedby cells when glucose is exhausted (diauxic growth), ethanol also inhibits growth andlowers the growth rate. In addition, the fermentation of glucose is inefficient, producingonly 2 moles of ATP per mole of glucose (Patkar and Seo, 1992). At low concentrationsof glucose in aerobic cultures, glucose is oxidized completely to CO2 using the respiratorypathway, which produces 16-18 moles of ATP per mole of glucose (Patkar and Seo,1992). Thus, the increase in cell mass yield at low glucose levels is due to the dominanceof the much more efficient respiratory metabolic pathway. Similarly, E. coli can be grownto high cell densities by maintaining low glucose concentrations. In a typical batch cultureof E. coli, glucose ranges from 20 g/1 to 0 g/1 over a period of 8 hours (Stamm et al.,1992). At high glucose levels, however, depletion of dissolved oxygen due to a high cell13respiration rate, particularly in high cell density cultures, causes a shift to fermentativemetabolism. The production of acetate during the fermentative metabolism of glucoselimits growth (Smith and Bajpai, 1985). Thus, a fed-batch culture is preferred over batchculture in order to maintain consistently low glucose levels and avoid the effects offermentative metabolism, thereby achieving high cell density.Once a high cell concentration has been obtained, the reactor environment can beadjusted to achieve a high protein yield and therefore high production rate. Demain lists alarge number of commercially important proteins (e.g. streptomycin, neomycin, penicillin,etc.) which are only expressed during idiophase. These proteins are not expressed duringgrowth because the enzymes responsible for their formation are repressed until thetropophase nears completion (Demain, 1972). In the case of some constitutive promoters,such as CYC 1, PGK, and GAPDH, maximum expression is observed during glucosedepletion (Hardjito et al., 1993). Patkar and Seo (1992) found that expression of theyeast SUC2 gene was derepressed at glucose levels below 2 g/L, and that there existed aninverse functional relationship between medium glucose concentration and the specificactivity of the invertase produced. Some inducible promoters require a high cell density inorder for the inducer to accumulate before the gene is derepressed (Demain, 1972). Or, ifthe inducer is added exogenously, a high cell density may be desired before induction inorder to maximize production rate. This may be particularly important for unstablerecombinant proteins subject to proteolysis, in which case a high production ratenecessitates the minimization of the run time before the product is harvested (Hardjito etal., 1993).1.2.2 Strategies for Glucose Monitoring and ControlWithout a glucose monitoring system to provide continuous, on-line, real-time,glucose analysis of the fermenter medium, efficient control of fed-batch cultures has been14difficult. A number of strategies have been employed to develop feeding schedules forfed-batch culture, such as empirically or mathematically derived models to predict feedingtime based on measurable parameters (Smith and Bajpai, 1985). As always, however,certain assumptions must be made in modelling, and an on-line glucose sensor wouldincrease the accuracy of the models by decreasing the number of assumptions necessary.Open-loop control schemes have been developed, using increases in dissolved oxygen ordecreases in CO2 in the reactor off-gases as an indicator of a decrease in the cellrespiration rate due to glucose exhaustion. However, cell respiration rate may alsodecrease if another substrate in the medium becomes limiting, such as ammonium in E.coli fermentations.Some indirect closed-loop strategies have also been investigated. Kole et al.,(1986) controlled glucose concentration during a cultivation of E. coli based on feedbackcontrol of ammonium concentration using an ammonia gas electrode. The ratio of thecellular consumption rates of ammonium and glucose was determined to be 13 molesglucose per mole ammonium, and a corresponding mixture was used to feed the fermenterbased on feedback from the ammonium controller. Ammonium and glucoseconcentrations were maintained at a constant level (15 mM and 8 g/L, respectively) andbiomass yield calculated on the basis of glucose or ammonium was increased relative tobatch growth. However, acetate production can still be growth inhibiting at a glucoseconcentration of 8 g/L. In addition, the ammonia electrode required a pH of 11.0 andcould not be used in situ, requiring a complex hardware system to withdraw and monitorsamples from the fermenter in a separate measurement vessel.Glucose concentration in samples of the medium can be measured directly usingoff-line enzymatic assays or laboratory glucose analyzers. Numerous enzymatic assays forglucose are commercially available, requiring from 5 to 45 minutes per assay (Sigma,1993). However, off-line enzymatic assays are not practical for controlling glucoseconcentration, as the assays are labour intensive and can require up to 1 hour per sample15for the complete procedure of sampling and assay (Patkar and Seo, 1992), thus precludingthe use of computerized feedback control systems. Other off-line methods, such as HPLCand gas chromatography, may provide faster assays but are capital intensive and require anexperienced operator for instrument set-up and optimization.Laboratory glucose analyzers, such as the YSI 2300 STAT Glucose and L-LactateAnalyzer (Yellow Springs Instruments, Inc., Yellow Springs, OH, U.S.A.) or theBeckman Glucose Analyzer 2 (Beckman Instruments Inc., Fullerton, CA, U.S.A.) are lessexpensive and can reduce the assay time to less than 90 seconds, but are difficult toimplement on-line. However, Hoist et al.(1988) modified a commercially availableglucose analyzer (Gambro AB, Lund, Sweden) for use as a glucose monitor during a 3 Lcultivation of E. coli. The system was completely automated and monitored by computer.Fermenter broth was drawn continuously at 3 mL/min and mixed with metabolic inhibitorto prevent cell-associated changes in glucose concentration during transport to theanalyzer. A response time of 6 minutes and a measuring range of 0-5 g/L were reported,and the measured value was updated every 90 seconds. However, the system iscomplicated, and the sterility of the sampling port and the possibility of influx of metabolicinhibitor into the fermenter were cited as potential problems. Moreover, the consumptionof medium by the sampling system could cause a significant decrease in the volume ofmedium during the course of the fermentation. Recently, Stamm et al. (1992) alsoreported a modification of the Yellow Springs Instruments YSI 2700 Glucose Analyzerfor on-line monitoring of glucose in batch fermentations of E. coli. The BIOPEM (Braun,Melsungen, Germany) sampling module was used, which is based on cross-flow membranefiltration of cells from the sample stream, and the filtrate flow rate was 0.5 mL/min. From4-50 samples could be measured per hour over a linear measuring range of 0.3-25 g/L,with a response time of 4 minutes.Numerous other automated sampling systems have been published (Mandenius etal., 1984; Romette, 1987; Bradley et al., 1991). In addition, the use of flow injection16analysis (FIA) in conjunction with sampling devices and off-line substrate analyzers hasseen rapid development for on-line fermentation monitoring in recent years (Kittsteiner-Eberle et al., 1989; Valero et al., 1990; Huang et al., 1991; Renneberg et al., 1991;Bradley et al., 1991). FIA is characterized by precisely controlled injection of samplesinto a carrier stream which flows to a detector. The advantages are low sampleconsumption, dilution of the analyte in the carrier stream (which can be adjusted byvarying the flow rate of the carrier), high frequency of analysis, and flexibility andadaptability of the monitoring system due to the typically modular construction. Some ofthe inherent disadvantages of off-line sample analysis that are often cited, however,include discontinuous measurement, depletion of the medium volume, and increased riskof fermenter contamination. The presence of dead volume in the sampling system alsoslows response time and allows for changes in the composition of the sample duringtransport to the detector. In addition, a high degree of automation and extra hardware(e.g. detector(s), pumps, valves, tubing, etc.) is required, which often necessitates atrained or experienced operator to optimize the system.The advantages and disadvantages of automated sampling systems in contrast within situ probes have been discussed in the literature (Ogbomo et al., 1990; Bradley et al.,1991; Filippini et al., 1991; Lildi et al., 1992; Cleland and Enfors, 1983). It is generallyagreed that in situ probes would be the more desirable approach, provided such devicescan be made to operate reliably under the conditions and constraints imposed by currentbioprocess designs. A sensor system based on an in situ biosensor probe has the potentialto provide continuous, real-time analysis of complex fermentation media due to thespecificity and selectivity of the biological component of the sensor.171.2.3 In situ Enzyme Electrode ProbesNumerous electro-enzymatic biosensor probes have been reviewed and described(Enfors and Molin, 1978; Turner et al., 1987; Mascini and Palleschi, 1989; Guilbault andLuong, 1989; Hendry et al., 1990; Bradley et al., 1991; Moody and Thomas, 1991), and awide variety of analytes could be monitored using different enzyme-based systems, asshown in Table 1.1. Unfortunately, the practical concerns of using an in situ biosensorprobe, such as in situ sterilizability, long-term stability, adequate measuring range, andmembrane fouling, have thus far prevented the widespread application andcommercialization of this approach. However, these difficulties are not viewed asinsurmountable, and some in situ enzyme electrode probes have been used to monitorglucose concentration for limited periods of time during cultivations of Baker's yeast(Bradley et al., 1988, 1989, 1991), Candida utilis (Enfors, 1981), and Escherichia coli(Cleland and Enfors, 1983, 1984b; Brooks et al., 1987/88).For fermentation applications, in situ enzyme electrode probes in general shouldhave the following characteristics:1. The measuring range of the sensor must be sufficient to cover the range ofanalyte concentration encountered in the fermenter, with adequate sensitivity over theworking range.2. The sensor response time must be sufficiently fast to monitor changes inthe analyte concentration and permit process control.3. The sensor must be insensitive to changes in the chemistry of the medium,such as dissolved oxygen tension, pH, and ionic strength.4. The sensor system must be resistant to electrochemical and enzymaticpoisons, inhibitors, interferents, etc. in the analyte medium.5.^The sensor system should allow for re-calibration during fermenteroperation. Periodic re-calibration is desirable in order to correct for changes in the18calibration constants of the sensor which may occur during a fermentation due to fouling,poisoning, deactivation, or other internal chemical changes in the probe (Kok and Hogan,1987/88).6. Sufficient long-term stability of the enzyme component of the sensor isnecessary in order to be useful in a typical fermentation. This may dictate that periodic re-calibration of the sensor is required if the enzyme activity diminishes. Alternatively, thismay require the capacity for replacement of the enzyme component when the activity hasdegraded to an unsatisfactory level. This regeneration of the probe must be possiblewithout interrupting the fermentation.7. In situ sterilization of the probe by conventional methods, such asautoclaving, must be possible. Steam sterilization is preferred by industry over othersterilization methods, such as ethanol, chloroform, radiation, etc:.The implementation and commercialization of in situ enzyme-based biosensors forbioprocess control has been particularly slowed by problems relating to the in situsterilizability of the probe and the stability of the enzyme component. These problemshave been addressed in a variety of ways by other investigators over the past 15 years.Enfors and Molin (1978), and Enfors and Nilsson (1979) reported an autoclavableenzyme electrode consisting of a potentiometric transducer in a stainless steel housingwith a dialysis membrane. The bioreactor and the probe hardware were steam sterilized insitu, after which a solution of enzyme was pumped into the enzyme chamber formed bythe transducer, the electrode housing, and the dialysis membrane. When the activity of theenzyme had diminished, the enzyme chamber could be perfused with fresh enzymesolution, however the sensor could not be recalibrated using internal standards pumpedinto the enzyme chamber or the enzyme would be washed out. Furthermore, it is difficultto concentrate the enzyme activity in a region near the surface of the indicating electrode,19leading to excessive H202 generation throughout the interior of the probe and potentialproblems due to H202 accumulation.Cleland and Enfors (1983) envisaged a probe consisting of a stainless steel housingand a dialysis membrane which could be autoclaved in situ. The enzyme would beimmobilized on an internal electrode which would then be inserted into the sterilizedhousing. A sterile barrier between the bioreactor medium and the electrode would bemaintained by the dialysis membrane. Cleland and Enfors (1984a, 1984b) also designed aninternally buffered enzyme electrode where buffer flowed continuously through theenzyme chamber. The buffer flow across the dialysis membrane provided continuousdialysis of the analyte and permitted variation of the sensor's working range by altering thebuffer flow rate. However, the enzyme could not be replaced without dismantling thehardware, which would presumably be difficult to automate.Bradley et al. (1988, 1989a, 1989b, 1990, 1991) and Brooks et al. (1987/88)reported an autoclavable enzyme electrode with provisions for internal calibration andrapid replacement of the enzyme. The stainless steel sensor housing and dialysismembrane were sterilized in situ, after which the internal electrode was inserted. Theinternal electrode consisted of a shaft with graphite discs for electrodes. The enzyme wasimmobilized on the graphite discs. A flow of buffer through the enzyme chamber could bemaintained if desired, and the sensor could be calibrated internally by pumping calibrationstandards into the enzyme chamber. Rapid enzyme replacement was performed whenrequired by removing the internal electrode and exchanging the graphite discs. However,it was necessary to carry out the process of enzyme replacement manually.Btihler and Ingold (1976) designed a mechanically retractable probe that could beisolated from the fermenter completely by a ball valve. The probe could be cleaned,recalibrated, modified, or replaced entirely, without compromising the fermentation, andthen re-sterilized with steam before re-introduction to the fermenter. Kok and Hogan(1987/88) described an in situ electrode calibrator, which isolated the probe tip from the20fermenter broth using two half-cylinders forced together around the probe pneumaticallyto form a small chamber. The external surface of the sensor membrane could be cleanedby washing or spraying with a jet of cleaning solution, and calibration solutions could beinjected into the chamber to recalibrate the sensor. The probe tip was then re-sterilizedwith steam before re-exposing to the fermenter medium.In the probe designs described above, the combined functions of enzymereplacement and recalibration of the sensor cannot be performed without operatorintervention. Thus, an operator must be standing by during a long fermentation run tomanually replace the enzyme component periodically and then recalibrate the sensor. Theregenerable biosensor probe developed in this thesis incorporates the essentialcharacteristics of the probes described above, but the hardware system is relatively simpleand the enzyme component can be replaced when necessary using computer-controlledpumps and valves. This enables automated, in situ replacement of the enzyme componentperiodically during a long fermentation, without interrupting the fermentation ordismantling the sensor. Furthermore, a single "generic" probe body can, in principle, beused to implement a variety of different electro-enzymatic sensors, based on differentenzymes and/or transducers. The design and principle of operation of the proposed sensorsystem is described in Chapter 2.21CHAPTER 2 BACKGROUND AND THEORY2.1. INTRODUCTIONIn general, the design and operation of enzyme-based biosensors capitalizes onbiochemical reactions involving the species of interest, (i.e., the analyte) which arecatalyzed by naturally occurring biological molecules called enzymes. Usually, the analyteis a substrate or co-substrate of the enzyme reaction. In the case of enzyme electrodes,the sensor operation is based on the electrochemical determination of reactant(s) orproduct(s) associated with the enzyme reaction. The advantage of incorporating theenzyme component in the biosensor is that the natural specificity of the enzyme for theanalyte is imparted to the sensor, thus the biosensor can be used for analyte determinationin typically complex biological solutions. In addition, enzymes are able to catalyzereactions faster than many non-biological catalysts in the temperature and pH rangenormally encountered in biological environments (Bailey and 011is, 1986). Thecombination of the chemical signal amplification due to enzyme catalysis and the highsensitivity inherent to electrochemical detectors results in a highly sensitive electro-enzymatic sensor.2.2. ENZYME ELECTRODE COMPONENTSA schematic diagram of a typical glucose sensitive enzyme electrode is shown inFigure 2.1. The approach used here is based on the amperometric detection of hydrogenperoxide produced as a by-product of the enzymatic oxidation of glucose by glucoseoxidase. A potentiostat is used to maintain the necessary bias potential for theelectrochemical oxidation of hydrogen peroxide at a noble metal indicating electrode, and2202GlucoseGluconicAcid1 0 2H202GOx1^p-D-Glucose + 0 2 + H 2 OIndicating^Immobilized^DialysisElectrode^Enzyme Membrane^Membrane22^H2 0 2^2H + 0 2 + 2e -• D-Gluconic Acid + H2 0 2E a + 0.7 V vs. SCEElectrodeCurrentInterferingSpeciese -Figure 2.1:^Schematic diagram of a glucose sensitive enzyme electrode. The basiccomponents of the sensor hardware and the principle biochemical and electrochemicalreactions are shown.23also converts the glucose-dependent electrode current into a useable output voltage. Theenzyme is immobilized in a gel or thin film, (the enzyme membrane) in close proximity tothe surface of the indicating electrode, which is typically platinum, gold, or carbon. In thiscase, the indicating electrode is platinum. One or more dialysis or covering membranesare normally used over the enzyme membrane. Glucose and oxygen diffuse into theenzyme layer and react with the enzyme to produce gluconic acid and hydrogen peroxide.Some of the hydrogen peroxide diffuses to the electrode where it is oxidizedelectrochemically, liberating oxygen, protons, and electrons. Thus, the current measuredat the indicating electrode is related to the concentration of the analyte in the bulksolution.The outer membrane of the sensor is very important, as it represents the interfacebetween the biosensor and the analyte medium. The purpose of the dialysis membrane isto allow the diffusion of glucose, oxygen and electrolytes into the enzyme layer whileexcluding potential interfering species which may be present in the analyte medium, suchas cells, proteins, enzyme inhibitors, or electrochemical interferents. The dialysismembrane also provides a significant mass transfer resistance which increases the linearityof response and the working range of the sensor, although at the expense of sensitivity.Selectively permeable membranes can be used to increase the mass transfer of oxygen withrespect to glucose, thereby minimizing the possibility of oxygen limitation of thebiosensor. A covering membrane may also be used to protect the physical integrity of theenzyme membrane.2.3. BIOCHEMICAL CONSIDERATIONSThe enzyme glucose oxidase 63-D-glucose:oxygen 1-oxido-reductase, EC to a class of enzymes called flavoprotein oxidases. Most studies are performedusing the enzyme from Aspergillus niger, although the enzyme from Penicillium24amagasakiense or Penicillium notatum has been used. Glucose oxidase from A. niger is adimer with two tightly bound FAD molecules per dimer. The tertiary structure of glucoseoxidase from A. niger was recently reported by Hecht et al. (1993). Molecular weightsbetween 155 and 186 kD have been reported for this enzyme by different sources, but avalue in the range of 155 kD is most common. The physical properties and the enzymekinetics of glucose oxidase have been studied extensively and can be found in standardtexts on enzymology (Bright and Porter, 1975; Dixon and Webb, 1964).The glucose oxidase catalyzed reaction of glucose is often presented in a singlestep:13-D -Glucose + 02 + H2O GOx > D - Gluconic Acid + 11202^(2.1)The reaction shown above is actually a net reaction which consists of two redox steps anda non-enzymic hydrolysis step. The electron transfer path is from the electron donor(glucose) to the coenzyme flavin adenine dinucleotide (FAD), and then to an electronacceptor. The first redox step involves the oxidation of glucose by glucose oxidase andthe reduction of the coenzyme FAD:(3- D - Glucose --> 5 - Gluconolactone + 2H + + 2e-^(2.2)FAD + 2H + + 2e - <----> FADH2^ (2.3)The second redox step involves the reduction of FADH2 and the transfer of electrons tothe electron acceptor:25FADH2 4----> FAD + 2H + + 2e -^ (2.4)02 + 2H+ + 2e - <^ > H202^ (2.5)In this case, oxygen plays the role of electron acceptor. The summation of the two redoxsteps yields:f3 -D -Glucose + 02 GOx > 5 Gluconolactone + H202^(2.6)The 5-gluconolactone undergoes a spontaneous hydrolysis reaction given by:5 - Gluconolactone + H2O <---> D - Gluconic Acid^(2.7)The sum of Equations 2.6 and 2.7 gives the overall net reaction given in Equation ELECTROCHEMICAL CONSIDERATIONSSome of the hydrogen peroxide produced by the enzymatic oxidation of glucose inthe enzyme layer diffuses to the surface of the indicating electrode and is oxidizedelectrochemically according to the reaction:H202 <---> 2H + + 02 + 2e -^(2.8)The standard electromotive force (emf) of the half-cell reaction in Equation 2.8 isE° = +0.695 V versus the standard hydrogen electrode (SHE) at pH 0. At pH 7, the26standard emf is shifted by approximately -0.410 V. At an electrode potential that isslightly more positive of the standard emf, the reaction becomes spontaneous, although itmay be kinetically limited. To ensure that the electron transfer at the anode occurs at themaximum rate, i.e., that the electrochemical reaction is diffusion limited rather thankinetically limited, an electrode bias potential of +0.94 V versus SHE is normally applied.Under diffusion limited conditions, the current measured at the indicating electrode will beproportional to the concentration of hydrogen peroxide.The saturated calomel electrode (SCE) or the silver/silver chloride (Ag/AgC1)electrode are typically used as reference electrodes as the SHE reference is ratherinconvenient. However, the standard emf of each these half-cell reactions is different.The emf of the SHE is taken to be 0 V by convention. The half-cell reactions are givenbelow.:2H+ + 2e- < > H2 (1 atm) E° = 0.000 V (2.9)Hg2C12 + 2C * > 2Hg + 2C1 - E° = +0.241 V vs. SHE (2.10)2AgC1 + 2e - <---> 2C1 - + 2Ag E° = +0.222 V vs. SHE (2.11)Thus, the normal operating electrode bias potential for most biosensor applications usingthe principle of operation above is approximately +0.7 V versus saturated calomel orAg/AgC1 reference electrodes. A comprehensive treatment of electrochemicalconsiderations can be found in standard texts on electrochemistry (Bard and Faulkner,1980; Sawyer and Roberts, 1974).272.5. MODELLING CONSIDERATIONSIn this thesis, all of the experimental results were empirical and no attempt atmodelling was undertaken. However, the study of modelling theory is worthwhile for aclearer understanding of the behaviour of electro-enzymatic biosensors and the processesthat determine biosensor performance. Due to the small physical dimensions and othergeometric considerations of enzyme electrodes, it is generally not possible to measureconcentrations and fluxes of substrates and products within the membrane layers of thesesensors. For this reason, mathematical modelling has become an important tool to analyzeand improve the performance of enzyme electrodes. Ideally, a successful mathematicalmodel should predict the measuring range, sensitivity, response time, and detection limit ofa given sensor configuration. The model can then be used to evaluate variations inparameters such as membrane thickness and immobilized enzyme activity (enzymeloading) with the aim of improving the analytical characteristics of the sensor. Trial anderror experiments to improve these sensor characteristics have produced what often seemlike conflicting results (Fortier et al., 1990; Foulds and Lowe, 1986). Hence, thecombination of modelling and experimental research can lead to a more directeddevelopment effort.Theoretical studies of immobilized enzyme electrodes are complicated by the non-linear nature of the enzyme kinetics and the mass transport processes. The overallmechanism is categorized as a diffusion-reaction problem, which is described by a set ofnon-linear, partial differential equations that can be solved analytically or numerically. Theoverall process may be diffusion controlled or kinetically controlled. Most publishedmodels are one layer models and consider only the immobilized enzyme layer, althoughreal sensors almost always employ one or more additional membrane layers which increasethe mass transport resistance. Consideration of the effects of covering membranes in28terms of diffusion retardation and modification of the effective permeabilities of substrateand co-substrate is presented in Section 2.5.5.A general model applicable to all sensor arrangements cannot normally beformulated due to the number of special cases that can exist. Furthermore, it is necessaryto weigh the advantages of a complex model with high integrity that accounts for many ofthe physical phenomena, against the convenience and simplicity of a more approximate yetqualitative model (Schulmeister, 1990). The intended application of the model must beheld in mind throughout its development. Once a basic model has been well developed,the features of more complex sensor systems can be incorporated into the model.The type and complexity of the model to be used also depends on whether theprinciple of measurement is transient or steady-state. When the transient response of thesensor to a step change is followed, the result is very rapid and very sensitive, and themodel can also be used to estimate the time to approach steady-state. However, themodelling is much more difficult because of the system of differential equations whichmust be solved. If the response is allowed to reach steady-state, then the time derivativesof the differential equations are zero and a system of algebraic equations results.The models presented here will be limited to monoenzyme electrodes, specificallythe glucose oxidase electrodes for glucose determination. The co-substrate is oxygen.Mediated enzyme electrodes and other modified enzyme electrodes are not considered.Amperometric electrodes are of primary interest, however the same reaction rate anddiffusion equations are employed to describe amperometric and potentiometric systemsbut different initial and boundary conditions for the solution of the resulting partialdifferential equations are used. For potentiometric electrode sensors, the concentrationsof substrate and product at the electrode surface are permitted to vary, but the flux of bothsubstrate and product must be zero. For amperometric sensors, the flux of theelectroactive species to be determined is non-zero at the electrode surface, due to theelectrochemical reaction which occurs (Janata, 1989; Leypoldt and Gough, 1984).29Similarly, the mode of operation of the amperometric sensor can be based on detection ofco-substrate (oxygen) or product (hydrogen peroxide); the observations and results arequalitatively similar (Leypoldt and Gough, 1984).2.5.1. Assumptions and ApproximationsThe following are typical assumptions and approximations made in the modellingliterature involving enzyme electrodes for glucose:1. At the boundaries of a given membrane, the Dirichlet condition applies.That is, the boundary values for all substances are equal to the concentration values on theopposite side of the interface (Schulmeister, 1990). This is generally true if the partitioncoefficient is equal to one. See assumption number 2.2. No partitioning of reactants and products occurs between the membraneand the solution. However, this may not always be true. For example, Maresse et al.(1987) found that the partition coefficient for potassium ferrocyanide was greater than onefor albumin-glutaraldehyde membranes.3. Reactions occurring on the external surface of the membrane areunimportant in comparison with the reactions occurring within the membrane. Thecontribution from the reactions at the external surface only becomes important in the limitwhere the reactions are extremely fast compared to the rate of diffusion (Schulmeister,1990).4. The bulk solution is assumed to be perfectly stirred and acts as an infinitereservoir of substrate if the volume is great enough (Schulmeister, 1990; Jochum andKowalski, 1982). Thus, the bulk solution concentration of substrate is assumed to beconstant.5. The concentration gradients within the enzyme layer can be assumed to belinear (Leypoldt and Gough, 1984). In actual fact, the concentration gradients are curvi-30linear, however, for the purposes of numerical analysis, the gradients may beapproximated as linear if a sufficiently small step size is used.6. The diffusion process can be reduced to one spatial dimension. Thus, aplanar electrode is approximated as semi-infinite and edge effects are ignored (Janata,1989). Cylindrosymmetrical electrodes can be reduced from two spatial dimensions (axialand radial) to one spatial dimension if the ratio of the diameter of the sensor surface to thethickness of the enzyme membrane is large (Schulmeister, 1990).7. No depletion layer exists at the membrane/solution boundary if the solutionis stirred rapidly. However, there exists a stagnant boundary layer even under stirring.This thin layer (10-3 - 10-1 mm depending on viscosity and stir rate) (Schulmeister, 1990)is often ignored since the diffusion in solution is rapid compared to diffusion in themembrane. Nevertheless, a shallow depletion layer can exist within the stagnant boundarylayer.8. The immobilized enzyme is uniformly distributed in a homogeneousmembrane. Thus the mass transport and kinetic parameters are constant at all points in themembrane. Membrane homogeneity is important to the model for simplicity and isassumed in all the published models. A heterogeneous membrane would be extremelycomplex to model because of the variation of parameters at every point in the membrane.However, the validity of the assumption of homogeneity of the immobilized enzymemembrane is an important issue with some uncertainty.9. The long-time processes of enzyme inactivation can be ignored(Schulmeister, 1990). This is a reliable assumption over the time period of a singlemeasurement and when enzyme loading is high.10.^The rates of protonation reactions are assumed instantaneous. This is asafe assumption on the time scale of the other processes in the overall mechanism (Janata,1989).3111.^The electrochemical double-layer which exists within the membrane at theelectrode surface is ignored. The electron transfer processes at the electrode surface arealso considered to be fast compared to the membrane processes.2.5.2. Diffusion EquationsThe enzyme electrode is often considered as a series of resistances (Janata, 1989):1) The resistance to charge transfer at the working or indicating electrode; 2) theresistance to mass transport; and 3) the resistance to charge transfer at the auxiliary orcounter electrode. Normally, the counter electrode area is much larger than the indicatingelectrode area, so the resistance to charge transfer at the counter electrode is negligible.Furthermore, the electron transfer reaction at the indicating electrode is considered to befast compared to the processes which occur in the membrane, so the resistance at theindicating electrode is generally neglected as well. The mass transport resistance can besubdivided into: 1) The external mass transport resistance, which occurs at thesolution/membrane interface; and 2) the internal mass transport resistance, which occurswithin the membrane (Cronenberg and van den Huevel, 1991, Hall, 1991). Furthermore,the process of mass transport occurs by the three mechanisms of diffusion, migration, andconvection, and can be described by the Nernst-Planck equation. For a species i:ji ^-D. dCi z.FU. cl(1) + C . (^1^1  dx^" dx^l'vx"(2.12)The first term is the diffusion term and is usually the only current limiting term.The second term is for migration and is considered negligible, if there exists an excess ofinert electrolyte, or constant, if the electric field in the membrane is constant at a givenpotential. The third term is the convection term and is also considered constant when the32electrode is rotating or vibrating, or it is assumed negligible in the membrane layer(Schulmeister, 1990).Mass transfer in the bulk solution is generally considered to be quite fast comparedto mass transfer in the membrane because of the higher diffusion coefficients in thesolution and because of stirring. Thus, only the internal mass transfer resistance is usuallyconsidered. Within the membrane, the Nernst-Planck equation reduces to Fick's law:aci (x, t)(x, t) = - Dm, i^ax (2.13)Normally, x = 0 at the electrode surface and x = L at the membrane/solution interface,where L is the thickness of the enzyme membrane.2.5.3. Enzyme Reaction-Rate EquationsLet the enzyme reaction rate be given by R(S,O) for substrate, S, and oxygen, 0.Most commonly, the published models assume Michaelis-Menten enzyme kinetics,described by the rate equation:V SV — maxKm + S(2.14)where V is the enzyme reaction rate, Vmax is the maximum reaction rate, and Km is theMichaelis constant for the enzyme (Bailey and 011is, 1986). In the special case thatoxygen is present in excess and the glucose concentration is low, the reaction can beconsidered first-order and the reaction rate can be modelled linearly (Bailey and 011is,331986). This assumption holds when the substrate concentration is much less than theMichaelis constant, that is, S° << K m. Therefore Equation 2.1 reduces to:V = kS = R(S)^ (2.15)where k — VmaxKmOne-substrate models have been successful in simulating the calibration curves fortwo-substrate enzyme electrodes when the co-substrate concentration is always close tothe saturation value, such that the substrate is the limiting reagent in the reaction.However, the response at lower co-substrate concentrations demonstrates unusualbehaviour that cannot be predicted by the one-substrate models (Leypoldt and Gough,1984). The characteristic, two-substrate rate expression used is the non-linear, Ping-Pongreaction rate equation (Schulmeister, 1990; Leypoldt and Gough, 1984; Ylilammi andLehtinen, 1988; Bailey and 011is, 1986):R(S, 0) — ^Vmax (S)(0) (0)(S + Ks) + KoSor,^R(S, 0) — Vmax1 + Ks + K0S^0(2.16)(2.17)The concentration of oxygen remains a parameter in the rate expression, and theoxygen concentration will be included in the glucose-dependent current, ig. The values ofKs and Ko for glucose oxidase are typically 0.11 M and 5x10 -4 M, depending on the datasource (Linek et al., 1980). In the case where the substrate concentration is low andoxygen is present in excess, then Ks/S>>K0/0 and Equation 2.1 becomes:34R(S) = Vmax S = kSK sif k — Vmax Ks(2.18)(2.19)The Ping-Pong rate expression reduces to the first-order reaction rate equation onceagain.If Michaelis-Menten kinetics are assumed for the enzyme reaction rate, then thetypical kinetic parameters which must be determined are the maximum rate, Vmax, and theMichaelis constant, Km, for the substrate. However, the kinetic parameters for theimmobilized enzyme are not necessarily the same as for the soluble enzyme. For thatmatter, the use of Michaelis-Menten kinetics for the immobilized enzyme may be invalid,since the Michaelis-Menten model is defined for a homogeneous and mobile (i.e., soluble)system. However, the Michaelis-Menten model fits the experimental data well and is usedfor its simplicity. In the case of an amperometric enzyme electrode, the sensor current isoften taken to be representative of the enzyme rate since the electron transfer at theelectrode is relatively fast and the electrochemical reaction can be ignored. The parameterV is replaced by I in the Michaelis-Menten equation (Equation 2.14) and the sensorcurrent and substrate concentration data are used to determine the enzyme kineticparameters. The Michaelis-Menten equation becomes:R(S) = I — 'max S ^(2.20)Km + SHowever, the kinetic parameters that are determined must be denoted as "apparent"kinetic parameters, 'max and Km, because any unresolved mass transfer resistances are35included in the apparent kinetic parameters (Leypoldt and Gough, 1984). I max comprisesterms that include the amount of enzyme activity immobilized in a given sensorconfiguration. The parameter Km includes the characteristic Michaelis constant of theenzyme as well as any unresolved mass transport resistances of the given sensorconfiguration (e.g. choice of covering membrane system, immobilization method, andsensor geometry).However, the kinetic parameters need not necessarily be known explicitly. Theapproach suggested by Leypoldt and Gough (1984) and supported by Schulmeister (1990)is to determine the mass transfer properties of each substrate without reaction using amembrane covered disk electrode. Then, measuring known substrate concentrations usinga given sensor and drawing a calibration curve, the unknown kinetic parameters can bedetermined by fitting an appropriate model to the calibration curve using a non-linearleast-squares fit. The parameters determined can be used repeatedly for the given sensor,as long as the calibration curve is unchanged. The method is useful for characterizingindividual electrodes, since the immobilized enzyme membranes cannot be reproducedwith very high precision and it becomes difficult to analyze their exact dynamics usinggeneral values for parameters.It should be noted that if the substrate concentration is high, that is, S/Km is large,then Michaelis-Menten kinetics predict that the reaction rate becomes zero-order:R = Vmax^(2.21)The reaction rate is independent of the substrate concentration and becomes afunction of the number of active sites of immobilized enzyme (Tran-Minh and Broun,1975). Thus, the enzyme electrode is not useful if the substrate concentration in theinterior of the enzyme membrane greatly exceeds K m because the sensor signal essentiallybecomes independent of bulk solution substrate concentration So in this regime.36acp (x , t)at- Dm , pa2cP (x , t) + kS (2.24)ax22.5.4. The General Modelling ApproachThe total rate of change of concentration of reactants and products in the enzymelayer is represented as the sum of the diffusion term and the reaction term:aci (x, 0 _ aJd , i (x, 0^ a2ci (x, t) ^ ± R(S,O) = Dm, i^± R(S,O)at^ax ax(2.22)The following system of equations results:acs (x,t)^,_,^a2cs (x,t)- Lim,s kSat ax2(2.23)aco(x,t)a2c0 (x,t)^ — D- _ n1,0^kSat 2(2.25)where S is for substrate (glucose), P is for product (hydrogen peroxide) and 0 is foroxygen. This system of equations can be solved using the appropriate initial and boundaryconditions. At time t = 0, the bulk solution concentration of substrate is S° and theproduct concentration, Po, is zero. In the membrane, substrate and productconcentrations are also zero. The membrane is assumed to be saturated with oxygen. Attime t > 0, the sensor is immersed in the analyte solution. Substrate molecules diffuse intothe enzyme layer and react with the enzyme. Product molecules diffuse throughout theenzyme layer. Since amperometric electrodes are operated with an applied potential onthe diffusion-limiting plateau, all of the electroactive species reaching the electrode are37reacted and the surface concentration is effectively zero. These conditions can beexpressed as follows:At the solution/membrane interface and for all time t:Cs(L,t) = S°^ (2.26)Cp(L,t) = 0 = P° (2.27)Co(L,t) = C *0 = 0°^ (2.28)For all x such that 0 <= x <= L and time t = 0:Cs(x,0) = 0^ (2.29)Cp(x,0) = 0 (2.30)Cp(x,0) = ape°^ (2.31)For an amperometric, product sensitive electrode, at all time t:Cp(0,t) = 0acs (0,t) _ 0axaco(0,0 — 0ax(2.32)(2.33)(2.34)The conditions for oxygen detection mode are qualitatively similar.The goal of the model is to determine the glucose dependent current, ig. In theproduct sensitive case, the number of ion equivalents, n p , reaching a unit area of theelectrode is:38dPn = -Dndx(2.35)The resulting anodic current is:acn (0,0ig (t) = z F A Dm,p ax^(2.36)where z is the number of charge equivalents per mole, F is the Faraday constant, and A isthe electrode area. The sensor current can be obtained by solving the system of partialdifferential equations for the product flux at the electrode surface and substituting intoEquation 2.36. The details of the numerical solution will not be presented here, but theresulting product flux is a function of the concentration of substrate in the bulk solution,So. Hence, the anodic current is also a function of bulk solution substrate concentration.2.5.5. Multi-layer ModellingLinear one-layer models are able to describe quite well the dynamic behaviour ofreal arrangements and are often sufficient, but multi-layer models are necessary forgreater accuracy (Schulmeister, 1990). A multi-layer model may be suitable in thefollowing situations:1. The bulk solution cannot be considered well-stirred. A two-compartmentmodel is often used in this case, where the bulk solution is considered to be onecompartment and the enzyme membrane layer is the second compartment (Jochum andKowalski, 1982).2. A front membrane, such as a dialysis membrane, is employed to stabilizethe enzyme membrane.393. The sensor uses selective layers which are permeable to the desiredsubstance(s) but impermeable to others. Selective layers can increase selectivity, enhancethe diffusion of oxygen over glucose, or provide protection for the metal surface of theelectrode or the enzyme (Schulmeister, 1987).4. The model includes the effects of the boundary layer at themembrane/solution interface and the internal electrolyte layer at the electrode surface.The usual system of reaction-diffusion equations and initial and boundaryconditions is used. In the model by Schulmeister (1987), a number of layers, 1, areimagined, where i = 1 to 1 and each layer i has width di. The mathematical problem is thatthe boundary conditions at the interface of each layer are unknown. Schulmeisterproposes a boundary function for substrate, S, and product, P, at each interface that isdefined as:^Si*(t) = Si(di,t) = Si+ 1 (0,0^(2.38)Pi* (t) = Pi(di,t) = Pi+1(0,t)^for i = I to 1 - 1 and t > 0^(2.39)Conservation of mass at each interface requires that:,^s i ) _ , (as il-us,it at^us,i+i^atD p i ( --all = D • ( aPi+1 ) at^p,i+1 at(2.40)(2.41)The conservation of mass equations are used to solve for the boundary function ateach i and t. Thus for each interval of time, the concentration profiles must beapproximated for all i successively, because the values of Si * (t) and Pi * (t) are needed for40the next time step. The multi-layer model was fitted to current-time data obtained from acommercial glucose analyzer with a three-layer membrane arrangement. The three-layermodel yielded very good fits and could be implemented on a personal computer(Schulmeister, 1987).2.6. PRINCIPLE OF OPERATIONFigure 2.2 is a schematic diagram of a biosensor based on CBD-immobilizedenzymes. The basic design consists of a platinum electrode, a porous cellulose matrix, anda protective dialysis membrane, all incorporated into a stainless steel probe for insertioninto the bioreactor. Inlet and outlet tubes in the probe body allow for perfusion of thecellulose matrix with the enzyme-CBD conjugate solution and the elution buffer. Theinternal electrode can be raised to permit complete perfusion of the cellulose matrix andthen lowered into contact with the immobilization matrix to facilitate substrate monitoring.The concept is similar to that employed in the commercial CO2 electrode fromIngold (Ingold, 1990). In this work, a prototype glucose biosensor is built from amodified Ingold CO2 probe. For operation as a glucose sensor, the probe requires aglucose-permeable outer membrane, a cellulose matrix, and a custom - built internalelectrode assembly with three electrodes for amperometric operation. The cellulosematrix for immobilization of the enzyme-CBD conjugate is enclosed in the enzymechamber formed by the probe body, the dialysis membrane, and the internal electrode unit.The internal electrode unit can be raised and lowered via a threaded shaft. Whennecessary, the internal electrode unit is raised and the enzyme is eluted by perfusing thecellulose matrix with a suitable elution buffer (e.g. distilled water, sodium hydroxide, orguanidine hydrochloride), followed by perfusion with soluble GOx-CBD conjugate toreplace the immobilized enzyme. For glucose monitoring, the internal electrode is loweredsuch that the cellulose matrix is sandwiched between the dialysis membrane and the41platinum indicating electrode. The enzyme chamber is filled with electrolyte whichequilibrates with the electrolyte of the external analyte solution.Figure 2.3 outlines the principle of operation of the sensor system for continuousfermentation monitoring. The entire probe, exclusive of the enzyme, is sterilized in situduring steam sterilization of the fermenter and contents. Following sterilization, theenzyme is loaded into the sensor by perfusion of the cellulose matrix with soluble GOx-CBD conjugate, resulting in immobilization of the enzyme (behind the sterile barrierprovided by the dialysis membrane) via attachment of the CBD. The sensor is calibratedduring the addition of glucose to the fermenter. Periodically during the fermentation, thesensor calibration can be checked by off-line analysis of samples of the broth, or bypumping internal calibration standards through the enzyme chamber.Bradley and Schmid (1991) have reported that the relationship between the sensorresponse to internal calibrant and to external glucose concentration was constant for agiven sensor configuration. Thus, response to internal calibration standards can bemeasured and compared to the external calibration response, and if the results of theinternal calibration check do not compare well then the sensor calibration constants maybe updated.If the internal calibration check indicates that the sensor performance hasdeteriorated to an unacceptable degree due to deactivation of the enzyme, then theenzyme can be eluted by perfusion of the cellulose matrix with elution buffer. The sensoris reloaded with fresh GOx-CBD conjugate as before, the sensor is re-calibrated using theinternal calibration standards, and fermentation monitoring continues. The sensorcalibration can be verified, if desired, by comparison of the sensor output with off-lineglucose analysis of a sample of the fermenter broth.42Porous Cellulose ^ Probe BodyMatrixCells, Protein,etc.^4t,•its.' •Enzyme-CBDConjugateFigure 2.2^Schematic diagram of a biosensor based on CBD-immobilized enzymes.Load^Check Calibration CleanEnzyme-CBD^Run^ok?^ElectrodeSterilize Calibrate Yes EluteEnzyme-CBD Figure 2.3^Process flow diagram demonstrating the principle of operation of afermentation monitoring system using the renewable biosensor probe.43Before the cellulose matrix is reloaded with fresh enzyme-CBD conjugate, thesensor could also be perfused with electrode/membrane cleaning solutions and theelectrode could be cycled and anodized in order to regenerate the platinum surface ifnecessary. It would also be possible to clean the external surface of the membrane at thispoint, without breaching the sterility of the fermentation, by mechanically isolating theprobe tip from the fermentation broth and washing or spraying with a jet of cleaningsolution, as described by Kok and Hogan (1987/88). The probe tip would then be re-sterilized with steam before re-exposing to the fermenter medium. To continuefermentation monitoring, the sensor is reloaded with enzyme-CBD conjugate. Theseelectrode and membrane cleaning and regeneration processes may or may not benecessary, depending on the resistance of the system to fouling during the course of atypical fermentation.The proposed sensor system described here addresses the sterilizability andstability problems of electro-enzymatic sensors, but differs from sensor systems describedby other researchers in that the complete process of diagnosis and regeneration couldpotentially be performed under computer control and without interrupting thefermentation. This includes automated, in situ, replacement of the enzyme component.The proposed system allows for:1. conventional autoclaving of the electrode housing exclusive of the enzyme,2. periodic internal calibration verification,3. long term sensor operation by replacing the enzyme when necessary, insitu, and without interrupting the fermentation,4. recalibration following enzyme replacement using internal standards, and5.^the possibility of cleaning the membrane and the electrode during theenzyme replacement cycle, if necessary.44The prototype glucose biosensor described in this thesis has been developed withall of the above design characteristics in mind, but the development of the protocol forinternal calibration and the equipment for membrane and electrode cleaning were beyondthe scope of this project. The automation of the various protocols is discussed, howeverthe sensor functions described are performed manually in this work The chapters whichfollow describe the synthesis and characterization of the GOx-CBD conjugate protein, thedemonstration of the concept of a regenerable glucose biosensor using the GOx-CBDconjugate, and the construction and testing of an experimental glucose biosensorprototype which demonstrates the potential of this technology for on-line bioprocessmonitoring and control..45CHAPTER 3 SYNTHESIS AND CHARACTERIZATION OF THE GLUCOSEOXIDASE - [CELLULOSE BINDING DOMAIN] CONJUGATE 3.1. INTRODUCTIONThe first stage of the prototype development involved the conjugation of theenzyme glucose oxidase (GOx) with the cellulose binding domain from the cellulaseexoglucanase (CBDCex) from Cellulomonas fimi. In this work, a chemically conjugatedGOx-CBD protein was synthesized using the bi-functional cross-linking agentglutaraldehyde. The synthesis and characterization of the GOx-CBD conjugate isdescribed.3.2. MATERIALS AND METHODS3.2.1. Enzymes and ChemicalsGlucose oxidase (EC was Type X from A. niger (Sigma Chemical Co., StLouis, MO) and was used without further purification. The cellulose binding domain fromC. fimi exoglucanase (CBDCex) was harvested from recombinant E. coli and purifiedaccording to methods described elsewhere (Ong et al., 1993). CBDCex antiserum wasproduced in rabbits (Whittle et al., 1982). Grade 1 glutaraldehyde, 25% aqueous solution(Sigma Chemical Co.) was used as received for the GOx-CBD conjugation. Thesynthesis, purification, and storage of the GOx-CBD conjugate was carried out in 50 mMpotassium phosphate buffer, pH 7. Cellulose powder (Avicel, type PH101, FMCInternational Food & Pharmaceutical Products, Cork, Ireland), washed with distilled water46and phosphate buffer, was used for purification of the GOx-CBD conjugates. All otherchemicals were analytical grade and used as received.3.2.2. Enzyme-CBD ConjugationA 10 mg/mL solution of glucose oxidase in 50 mM phosphate buffer, pH 7, wasfirst activated with glutaraldehyde using a 50-fold excess of glutaraldehyde for each aminogroup (lysine residue or N-terminus) on the enzyme (Gibson and Woodward, 1992).40 p.L of glutaraldehyde (25% aqueous solution) was added per mL of GOx solution.After incubation overnight at 4 °C, the excess glutaraldehyde was removed by dialysisversus phosphate buffer using Spectra/Por 2 dialysis tubing, MWCO 12-14 kD (SpectrumMedical Industries Inc., Los Angeles, CA, U.S.A.). Dialysis was performed for 24 hoursversus 0.25 L of 50 mM phosphate buffer per mL of activated GOx solution. The dialysisbuffer was changed after 12 hours. CBDCex was added in a 1:1 molar ratio based onglucose oxidase and incubated overnight at 4 °C. 0.693 mL of 11.2 mg/mL CBDCexstock solution was added per mL of activated GOx solution. The CBD binds to theactivated GOx either via the single lysine residue present or via the N-terminus of theCBD polypeptide (Coutinho et al., 1992; Hansen and Mikkelsen, 1991) according to thenet reaction:GOx-NH2 + OHC-(CH2)3-CHO + H2N-CBD —> GOx-N=CH-(CH2)3-CH=N-CBD (3.1)Excess CBD was removed by buffer exchange with 50 mM phosphate buffer in anAmicon stirred cell using an Amicon PM30 (non-cellulose) membrane (Amicon CanadaLtd., Oakville, Ontario). Buffer exchange was performed until the absorbance (A280) ofthe filtrate versus phosphate buffer approached zero. The GOx-CBD conjugate was thenpurified in a single step by binding to Avicel, using at least 150 mg of Avicel per mL of47activated GOx solution. The conjugate was eluted from Avicel by washing twice with 0.1M NaCI in 50 mM phosphate buffer, followed by ten washes with 5 mM phosphate buffer,and then one wash with de-ionized, distilled water. The final wash was saved and madeup to 50 mM in potassium phosphate using 0.5 M phosphate buffer. The purified GOx-CBD conjugate was stored in 50 mM phosphate buffer at 4 °C. For longer term storage,the conjugate was stored bound to Avicel in 50 mM phosphate buffer at 4 °C and elutedwhen required. Batches of 1 mL and 10 mL volumes of GOx-CBD conjugate wereprepared using this protocol.3.2.3. Total Protein and Enzyme Activity AssaysTotal protein and enzyme activity assays were performed to determine the specificactivity (i.e., units of activity/mass of protein) of a conjugate sample. Total protein assayswere also performed on conjugate samples before and after binding the conjugate tocellulose, and the amount of protein bound was determined from the difference.Total protein was assayed using the Bio-Rad protein assay (Bio-Rad Laboratories,Richmond, CA, U.S.A.). Samples were assayed in triplicate on a 96 well plate using theMolecular Devices Vmax kinetic microplate reader (Menlo Park, California). The resultswere compared to a standard curve determined simultaneously with GOx standardsranging from 0 to 1200 pg/mL. For each assay, 160 [IL of sample and 16 of dyereagent concentrate were used, and the absorbance was measured at 595 nm after aminimum 15 minute incubation period.GOx activity was assayed using a modified Sigma GOx activity assay. Thiscolorimetric, kinetic assay is based on the rate of oxidation of o-dianisidine in the presenceof peroxidase and hydrogen peroxide produced from the oxidation of 13-D-glucose byglucose oxidase. The red colour resulting from the oxidation of o-dianisidine wasmeasured spectrophotometrically at 490 nm. Samples were assayed in triplicate on a 9648well plate using the kinetic microplate reader. For each assay, 240 p.L of dye-buffersolution (66 lig,/mL o-dianisidine in 50 mM phosphate buffer, pH 7. *Caution:o-dianisidine is a known carcinogen.), 10 f.IL of peroxidase solution (60 Purpurogallinunits/mL H20), and 10 1.1L of sample were added to each well. At time zero, 50 p.L ofglucose reagent (0.1 g glucose/mL H2O) was added to each well using a multi-channelpipette and mixed quickly. The increase in absorbance was measured over four minutesand the maximum rate was used for calculation. The results were compared to a standardcurve determined simultaneously with GOx standards ranging from 0 to 2.5 U/mLl. Theactivity assay was performed in 50 mM phosphate buffer, pH 7, at ambient temperature.3.2.4 Gel Electrophoresis and ImmunoblottingGel electrophoresis and Western blotting of the GOx-CBD conjugate wereperformed according to the protocols described by Harlow and Lane (1988). Ahomogeneous, 7.5% polyacrylamide gel was used, suitable for a molecular weight rangeof 45-200 Ica Bands containing CBDCex were identified by a Western blot using rabbitanti-CBDCex for the primary antibody (1:1000 dilution). The secondary antibody washorseradish peroxidase-labelled anti-rabbit antibody (1:10 000 dilution). The blot wasdeveloped using ECL Western blotting detection reagents (Amersham International plc,Amersham, UK) and exposed onto autoradiography film.3.3. RESULTS AND DISCUSSIONFigure 3.1 shows the results of gel electrophoresis (SDS-PAGE) of the GOx-CBDconjugate and Western blotting using primary antibody against CBDc ex• The GOx1 One unit is defined as the glucose oxidase activity which oxidizes 1 p.mole of (3-D- glucose perminute at pH 5.1 and 35°C.49control in lane 2 shows a band between 68 kD and 97 kD, due to the subdivision of GOx,Mr 155 kD, (Wilson and Turner, 1992) into two equal subunits. The CBDCex control,Mr 11 kD, (Ong et al., 1993) was not resolved on this gel but was carried to the anode inthe dye front. The raw GOx-CBD conjugate before purification is shown in lane 4.Uncoupled GOx was present in the raw GOx-CBD conjugate, but not after purification ofthe sample by binding to cellulose, as shown in lane 3. SDS-PAGE of a sample of purifiedGOx-CBD conjugate both before (lane 6) and after (lane 5) binding to cellulose does notdemonstrate any significant change due to binding to cellulose. The GOx-CBDconjugation product appears in bands at 200 kD and higher. The high molecular weight ofthe product is attributed to cross-linking of glucose oxidase by glutaraldehyde reagent inthe activation step of the conjugation protocol.The cross-linking of glucose oxidase may result in a GOx-CBD conjugation ratiogreater than unity. This feature of the conjugate protein could be advantageous in that ahigher enzymic activity could be loaded onto a fixed number of CBD binding sites on agiven cellulose matrix. On the other hand, the cross-linking of the glucose oxidase maycause a change in the conformation of the protein which could reduce the enzymic activityor change the enzyme specificity. It may be possible to tailor the GOx-CBD conjugationratio by modifying the conjugation protocol in order to increase, reduce, or eliminate thiscross-linking.Two experimental protocols were tested in an attempt to reduce cross-linking andensure a 1:1 conjugation ratio. In these protocols, the CBD was activated first, in eitherthe soluble form or bound to cellulose, followed by the addition of GOx to the activatedCBD, but an active conjugate was not obtained. Other conjugation protocols andimmobilization methods may still be investigated. For example, glucose oxidase has beenimmobilized using the carbohydrate moieties of the enzyme to prevent covalent bonding ofglutaraldehyde to lysine residues in the active site of the enzyme (Pieters and Bardeletti,1992)50A^ B1 2 3 4 5 6^1 2 3 4 5 6200 kD97 kD68 kD43 kDFigure 3.1: A. Gel electrophoresis (SDS-PAGE) of the GOx-CBD conjugate. A 7.5%polyacrylamide gel and Coomassie staining were used. Lane designations are: 1, sizemarkers; 2, GOx control; 3, raw conjugate after binding to cellulose; 4, raw conjugatebefore binding to cellulose; 5, purified conjugate after binding to cellulose; 6, purifiedconjugate before binding to cellulose. B. Western blot of the gel in A. Primary antibodywas rabbit - ocCBDc ex •51In the conjugation protocol, distilled water was used for eluting the purified GOx-CBD conjugate from Avicel. The use of 8 M guanidine HCl and 1 M NaOH for elution,as suggested by Kilburn et al. (1992), caused irreversible loss of enzyme activity. Usingdistilled water for elution, the soluble conjugation product was found to retain greaterthan 60% of the activity of the original, unconjugated enzyme. In addition, the GOx-CBDconjugate retained the binding affinity of the CBD for cellulose, and exhibited GOxactivity when bound to cellulose comparable to that of GOx immobilized by othertechniques (Harrison et al., 1988). The conjugate could be adsorbed and desorbed fromcellulose repeatedly, and could also be dehydrated for storage at room temperature, thenreconstituted with PBS, pH 7.4, without a significant loss of activity. Non-specificbinding of unconjugated GOx to Avicel or regenerated cellulose was found to beinsignificant. The specific GOx activities of various samples of soluble GOx-CBDconjugate are shown in Table 3.1 to illustrate the typical stability of the conjugate overtime. In glucose biosensor experiments using the GOx-CBD conjugate, it was found thatthe conjugate retained sufficient activity to be useful after up to 2 months in storage (whenstored bound to Avicel).Further characterization of the GOx-CBD conjugate will follow the optimizationof the conjugation protocol. A potential alternative to chemical conjugation would be todevelop an appropriate genetic construct that yields an active GOx-CBD fusion protein.Apropos of this, the gene encoding the glucose oxidase protein of Aspergillus niger hasbeen cloned and expressed in yeast (Frederick et al., 1990; DeBaetselier et al., 1991). Themain advantages of a genetically engineered conjugate would be greater uniformity andhomogeneity of the conjugate reagent, as well as considerable simplification of large-scaleconjugate production. A possible disadvantage, however, might be that the size and GOx-CBD ratio of the resulting conjugates could not be as easily tailored when compared tothe chemical conjugation method, which may limit the efficacy of the GOx-CBD reagent.52Further characterization and optimization of the chemically conjugated reagent shouldreveal the feasibility of this approach.Table 3.1: Specific activity of various samples of soluble GOx-CBD conjugate. Thespecific activity was calculated from the results of GOx activity and total protein assays.The relative activity was determined compared to unconjugated GOx. GOx-CBD Conjugate^3 weeks^Purified on Avicel;Batch #2 eluted by distilledwater after 1 weekin storageGOx-CBD Conjugate^5 weeks^UnpurifiedBatch #3aGOx-CBD Conjugate^6 weeks^UnpurifiedBatch #3bSpecificActivity(U/mg)RelativeActivity118 100%68 58%91 77%72 61%30 25%Time inSample^Storage^ConditionsGOx-CBD Conjugate^9 weeks^Purified on Avicel;Batch #3a eluted by distilledwater after 7 weeksin storageUnconjugated GOx^0^Freshly prepared in50 mM phosphatebuffer, pH 753CHAPTER 4DEMONSTRATION OF THE FEASIBILITY OF A REGENERABLEGLUCOSE BIOSENSOR4.1. INTRODUCTIONIn this chapter, the feasibility of a regenerable glucose biosensor based on thereversible immobilization of the enzyme using CBD technology is demonstrated. TheGOx-CBD conjugate was used in an experimental glucose biosensor constructed from aplatinum rotating disk electrode (RDE) with a cellulose matrix for enzyme immobilizationvia the CBD. Using glucose standards, the biosensor is calibrated repeatedly in PBSduring multiple cycles of loading and elution of the GOx-CBD conjugate in order tosimulate the periodic regeneration of a glucose biosensor during a fermentation.4.2. MATERIALS AND METHODS4.2.1. InstrumentationElectrochemical data were obtained using a Pine AFRDE4 bi-potentiostat and aPine AFMSRX analytical rotator (Pine Instrument Co., Grove City, PA, U.S.A.). A PineAFMDI1980 0.5 cm diameter platinum rotating disk electrode (RDE) was used for theexperimental indicating electrode. A platinum counter electrode was fabricated in-housefrom 1.0 mm diameter platinum wire (Aldrich, Milwaukee, WI, U.S.A.) bound in flintglass tubing (5 mm outside diameter) with epoxy (Chemgrip, Norton Co., Wayne, NJ,U.S.A.). A 25 mm X 25 mm, 52 mesh platinum wire gauze (Aldrich) was crimped ontothe platinum wire to increase the surface area of the counter electrode. The referenceelectrode was a saturated calomel electrode (SCE) (Fisher Scientific, Ottawa, Ontario,54Canada, No. 13-620-52). A Kipp & Zonen Model BD 112 strip chart recorder (Kipp &Zonen, Delft/Holland) was used for recording sensor output, and a Kipp & Zonen ModelBD91 XYY't recorder was used for recording cyclic voltammograms. The platinumworking electrode (RDE) was prepared by polishing with 0.05 pm alumina and rinsingwith distilled water. Immediately before each experiment, the working electrode wascycled in quiescent 0.5 M H2SO4 from -0.26 V to +1.2 V at 100 mV/s for 10 minutes,anodized at +1.8 V for 10 minutes, then cycled again until a stable cyclic voltammogramwas obtained.4.2.2. Experimental Glucose Biosensor PreparationThe rotating disk electrode was modified using a retainer such that differentcellulose matrices could be held in place over the platinum disk. The retainer was either aTeflon ring or a rubber ring (formed by cutting a transverse slice off the end of a piece ofTygon tubing) sized to fit tightly over the RDE. Three different cellulose-basedmembranes were used during the experimental biosensor trials: 1) Whatman No. 1qualitative filter paper (Whatman International Ltd., Maidstone, England), 2) Spectra/Por2 regenerated cellulose dialysis membrane, MWCO 12-14 kD (Spectrum MedicalIndustries Inc., Los Angeles, CA, U.S.A.), and 3) nitrocellulose protein transfer membranewith 0.45 pm pores (Schleicher & Schuell, Keene, NH, U.S.A.). The GOx-CBDconjugate was loaded by immersing the electrode assembly in a solution of GOx-CBDconjugate, followed by a single wash with 50 mM phosphate buffer.The RDE was rotated at 500 rpm and potentiostated at +0.7 V versus SCE in a250 mL beaker with 100 mL of electrolyte. The electrolyte used in all biosensorexperiments was phosphate buffered saline (PBS), pH 7.4, p = 0.2 M (0.1 M NaC1, 5 mMNaH2PO4, 30 mM Na2HPO4, preserved with 1 mM EDTA, and 5 mM sodiumbenzoate). The sensor was calibrated in PBS by recording the steady-state RDE current in55response to a series of aliquots of glucose standard (0.1 M glucose in PBS). The sensorresponse to decreasing concentrations of glucose was recorded by removing aliquots ofthe cell electrolyte and replacing with fresh, glucose-free PBS. Elution of the GOx-CBDconjugate was achieved by immersing the electrode assembly in the elution buffer,followed by a series of washes with 50 mM phosphate buffer only before reloading freshGOx-CBD conjugate. Three different elution buffers were used : 1) distilled, deionizedwater, 2) 1 M NaOH, and 3) 8 M guanidine hydrochloride (99+%(C1) from SigmaChemical Co., St. Louis, MO, U.S.A.), prepared in 50 mM phosphate buffer and filteredthrough Whatman qualitative filter paper. All chemicals were analytical grade and used asreceived.The experimental system was intended to simulate what will eventually be anautomated process. In the prototype system, the appropriate reagent solutions will beperfused through the cellulose matrix enclosed in the enzyme chamber of the probe via theinternal tubing of the Ingold CO2 probe body. In a future stage of the prototypedevelopment the system will be automated using computer controlled pumps and valvesfor the reagents.4.3. RESULTS AND DISCUSSIONAn experimental glucose biosensor was assembled using an RDE system in orderto investigate the feasibility of loading, eluting, and reloading the enzyme-CBD conjugateon a platinum electrode with an immobilized cellulose matrix. The RDE system wasadvantageous during the first stages of development because experimental parameterscould be easily manipulated. In addition, the rotation of the RDE created well-definedhydrodynamic conditions at the surface of the electrode which contributed to a stable,reproducible sensor signal, as well as reproducible conditions for loading/eluting the GOx-CBD conjugate. In the experimental system, this was essential for evaluating and56comparing different procedural modifications and conditions. In order to enable easyaccess to the cellulose matrix/enzyme membrane, no covering dialysis membrane was usedin the work reported here. Within the scope of the present feasibility investigation, theanalyte medium was limited to well-defined solutions of PBS and glucose, and the use of acovering membrane was not imperative. However, a suitable covering dialysis membranewill be required for monitoring glucose in fermentation broths, and a number of candidatematerials are currently being evaluated.Figure 4.1 shows a typical calibration curve obtained using the experimentalglucose biosensor. Relatively large electrode currents in response to glucose wereobtained, indicating that a substantial amount of enzyme had been immobilized (seebelow). The sensor was calibrated using increasing and decreasing concentrations ofglucose in order to demonstrate the reversibility of the sensor, a necessary characteristicfor fermentation monitoring. To investigate the possiblity of desorption of the enzyme-CBD conjugate during the experiment, the background signal was monitored using a bare,enzyme-free, platinum electrode, and the electrolyte in the electrochemical cell wasassayed for GOx activity following the experiment. No GOx activity could be detected,indicating that the enzyme was sufficiently strongly bound to the cellulose matrix.Different response time, signal magnitude, and durability was observed for eachcellulose matrix used (see Table 4.1). The response time was measured as the timerequired for the sensor signal to reach steady-state following the injection of an aliquot ofglucose standard into the electrolyte. The relative signal attenuation due to the cellulosematrix was determined by injecting an aliquot of H202 into the electrolyte and recordingthe magnitude of the sensor current for each cellulose matrix. The filter paper matrix wasfound to have the fastest response time (15-50 seconds) and the highest signal magnitudebut disintegrated over time when subjected to the shear stress encountered at high rpm onthe RDE. The regenerated cellulose matrix was more durable, but the response time wasmuch slower (2.5-5 minutes) and the signal was severely attenuated, indicative of the570.5 - ■ Increasing concentration+ Decreasing concentration^ Non-linear parametric fit2.0 -0.0 I^I^I0^5 10 15^20Glucose Concentration (mM)Figure 4.1: Typical calibration data for the experimental glucose biosensor. The GOx-CBD conjugate was bound to a nitrocellulose matrix on the Pt RDE. The RDE wasrotated at 500 rpm and potentiostated at +0.7 V versus SCE. Data points representincreasing and decreasing changes in glucose concentration. Solid line represents a non-linear parametric fit of the calibration data (concentration changes in the increasingdirection) using the Michaelis-Menten model for enzyme kinetics. Apparent kineticparameters were Imax = 4.4 p.A, Km = 21.3 mM.58substantially larger mass transport resistance of the regenerated cellulose when comparedto the filter paper. The nitrocellulose matrix with the 0.45 pm pore size was equally asdurable as the regenerated cellulose matrix when exposed to the 8 M guanidine elutionbuffer, but hardened and cracked when using 1 M NaOH as the elution buffer.Nevertheless, the response time when using the nitrocellulose matrix was 20-50 secondsand the signal attenuation was not as severe as that of the regenerated cellulose. For thebest combination of response time, signal magnitude, and durability, the nitrocellulosematrix was used in conjunction with the 8 M guanidine elution buffer for the glucosebiosensor experiments using the modified RDE.Table 4.1: Properties of the cellulose matrices used in the experimental glucose biosensorsystem. The cellulose matrix was held on the Pt RDE by a retaining ring. For the glucoseresponse time, the cellulose matrix was loaded with GOx-CBD conjugate.Sensor Current in Response Response Time FollowingCellulose Matrix^to 50^of 0.03% H202^Glucose Standard Injectionin PBS (prA) ^Filter Paper^0.188^15-50 secRegenerated Cellulose^0.050 2.5-5 minNitrocellulose 0.110^20-50 secFigure 4.2 shows calibration curves obtained during five complete cycles ofenzyme loading, elution, and replacement. The GOx-CBD conjugate sample was dividedinto equal aliquots before the experiment, and a fresh aliquot was used for each loadingcycle. Total protein assays of the aliquots following the experiment showedapproximately 20% variation in the amount of protein loaded. The sensor was calibratedfollowing each GOx-CBD conjugate loading procedure (Load #1 - Load #5). After eachcalibration curve, the cellulose membrane was washed with the elution buffer and tested59—0— Load #1—•— Load #2—A— Load #3—N— Load #4—o— Load #5—A— ElutionCN0 20a)(I)5^10^150 20Glucose Concentration (mM)Figure 4.2: Calibration data for multiple cycles of loading and elution of the GOx-CBDconjugate. See Figure 4.1 for experimental conditions. The GOx-CBD conjugate samplewas divided into equal aliquots before the experiment, and a fresh aliquot was used foreach loading cycle. The elution buffer was 8 M guanidine in 50 mM phosphate buffer. Atypical glucose response curve following elution of the GOx-CBD conjugate is shown.608 ^7 -6 -7,---.1 5-C -LI 4 -(1)c) -60 3 _ca)u) -' 2 -1 -0 ^0.0I^'^I^'^I^'^I^'^1^i0.5^1.0^1.5^2.0^2.51/(Glucose Concentration) (mM 1 )0 — Load #1•— Load #2A- Load #3■ — Load #4a— Load #53.0Figure 4.3: Lineweaver-Burk plot of the calibration data in Figure 4.2.61for glucose response to confirm that the enzyme had been completely eluted. The elutionbuffer used was 8 M guanidine, as the NaOH elution buffer was found to degrade thenitrocellulose membrane and distilled water did not elute 100% of the attached enzyme,resulting in a residual, glucose-dependent current after elution. A typical calibration curvefollowing the elution step is included in Figure 4.2, demonstrating that the sensor does notrespond to glucose after elution of the enzyme with 8 M guanidine. The elution procedurewas found to be effective in clearing all active enzyme from the cellulose matrix, howeverthe possibility exists that some denatured enzyme may have remained attached. Noattempt was made to quantify residual denatured enzyme that remained bound to thecellulose matrix following the elution procedure. The absence of a clear, sequence-dependent trend in the calibration curves reported in Figure 4.2 suggests that very littleenzyme remained attached. The differences in the calibration curves following eachloading step are not unlike the typical unit-to-unit variations found between amperometric,enzyme-based biosensors fabricated by other methods, and are attributed to differences inthe amount of enzyme activity loaded. For a given experiment, the specific activity of theGOx-CBD conjugate sample and the protein loading data could be used to calculate theamount of enzyme activity loaded in each loading cycle. Results were typically in therange of 0.5-3 units.The shape of the calibration curves suggested that the Michaelis-Menten equationfor enzyme kinetics might be useful as an experimental model. In the case of anamperometric enzyme electrode, the sensor current can be taken to be representative ofthe enzyme rate, since the electron transfer at the electrode is relatively fast, and thesensor current and substrate concentration data are used to determine the kineticparameters of the Michaelis-Menten equation, as described in Chapter 2. The use ofMichaelis-Menten kinetics for the immobilized enzyme may be invalid, since the Michaelis-Menten model is defined for a homogeneous and mobile (i.e., soluble) system of single-substrate enzyme. Nevertheless, the Michaelis-Menten model is useful for its simplicity,62and was found to fit the experimental data very well, as shown in Figure 4.1. Thisobservation was supported by the linearity of a Lineweaver-Burk plot of the data from themultiple cycle experiment (Figure 4.3) and the correlation coefficients from linearregression analysis of the Lineweaver-Burk data (Table 4.2).Table 4.2: Apparent enzyme kinetic data derived from the calibration data of Figure 4.2.Linear Regression of the^Non-linear Parametric Fit of theLineweaver-Burk Plot (Figure 4.3) Calibration Data in Figure 4.2 Usingof the Calibration Data in Figure 4.2^the Michaelis-Menten EquationImax (4A) Km (mM) r2 Imax (IAA) Km (mM) Chi2Load #1 7.0 18.3 0.9999 6.4 15.7 0.0010Load #2 5.2 15.1 0.9999 5.3 15.2 0.0004Load #3 7.6 18.0 0.9998 5.9 12.8 0.0008Load #4 7.4 15.9 0.9999 6.0 11.8 0.0009Load #5 8.5 17.4 0.9998 6.3 11.2 0.0016Mean 7.14 16.9 6.0 13.3Standard 1.2 1.4 0.4 2.0DeviationCoefficientof Variation17% 8% 7% 15%The kinetic parameters determined for the immobilized enzyme are not necessarilythe same as for the soluble enzyme. The kinetic parameters determined for theimmobilized enzyme will include unresolved mass transport resistances due to theimmobilization matrix and the configuration of the sensor system (e.g. coveringmembrane, sensor geometry, etc.) and, hence, must be denoted as "apparent" kinetic63parameters. The maximum sensor current, I max, is analogous to the maximum reactionvelocity, Vmax, and is useful as an indicator of the amount of enzyme loaded or theamount of immobilized enzyme activity. The apparent Michaelis constant, Km, is intrinsicto the enzyme and also the sensor configuration used. Table 4.2 shows the apparentenzyme kinetic data determined from the Lineweaver-Burk plot in Figure 4.3 and from anon-linear parametric fit of the data in Figure 4.2 using the Michaelis-Menten equation.Kinetic data determined by the non-linear parametric fit is deemed to be more accurate, asthe slope (m ./Imax) of the Lineweaver-Burk plot is strongly determined by the substrateand current values which lie far from the origin, but which are the least accuratelymeasured (Bailey & 011is, 1986). Using these values of I max and Km, an accurateempirical model for the calibration curve can be obtained, which enables operation of thesensor over a working range much greater than the linear range normally reported inbiosensor literature.It is interesting to note that the coefficients of variation of the apparent kineticparameters for the five cycles of enzyme loading were roughly 15% or less (see Table4.2). Complete characterization of the GOx-CBD conjugate and the sensor system mayprove that one or both of these parameters will be consistent for a given conjugate sampleand sensor configuration. Using the analogy of a straight line, the sensor can be calibratedusing only a single point if one of the parameters of the line is known (i.e., the intercept orthe slope). Using the Michaelis-Menten function, the sensor system could potentially becalibrated using a single point if one of the values of Km or I max can be determinedreliably. Failing this, the sensor could possibly be calibrated using 2 or 3 points to doublecheck the accuracy of the calibration curve based on the expected values of K m' or Imax .In the proposed prototype biosensor, this calibration check could potentially be performedusing internal calibration standards which are pumped into the enzyme chamber of theprobe body. Bradley and Schmid (1991) have reported that the relationship between thesensor response to internal calibrant and to external glucose concentration was constant64for a given sensor configuration. The feasibility of these proposed calibration protocolscan be investigated following the complete characterization of the biosensor design.65CHAPTER 5 DESIGN AND CHARACTERIZATION OF THE EXPERIMENTAL GLUCOSE BIOSENSOR PROTOTYPE5.1. INTRODUCTIONIn this chapter, the development and characterization of an experimental glucosebiosensor prototype is described. The prototype is based on a modified Ingold CO2 probeand uses cellulose binding domain technology for the reversible immobilization of theenzyme. A cellulose matrix is incorporated into the "enzyme chamber" of the probe body,and the GOx-CBD conjugate described in Chapter 3 is loaded and eluted from thecellulose matrix by pumping the appropriate loading or eluting buffer through the enzymechamber via inlet and outlet tubes in the probe body. The construction of the prototypeand modifications of the Ingold CO2 probe are detailed. The prototype is characterized inpilot-scale experiments in the absence of cells, and a glucose-permeable outer membrane isdeveloped for use in a real microbial fermentation. Finally, the prototype glucose sensor isused in a 20 L fermenter to monitor glucose consumption during a fed-batch cultivation ofE. coli.5.2. MATERIALS AND METHODS5.2.1. Prototype Design and ConstructionA diagram of the stainless steel Ingold CO2 probe that was used as the basis forthe prototype glucose biosensor is shown in Figure 5.1 (Ingold, 1990). The Ingold probewas designed to withstand temperatures from 20 - 125 °C and pressures from 0 - 2 bars.In its original configuration, the CO2-permeable membrane was silicone rubber reinforced661 20 ml Syringes2 High-temp. coaxial cable3 Cable union nut4 Ring nut for retracting5 Plugs6 Feed tubes7 Weld-in socket8 Guide tube (probe body)9 Electrode shaft10 pH electrode11 Reference electrode12 CO2 electrolyte13 Membrane cartridge14 Calibration buffer15 Glass membrane16 Reinforced silicone membraneFigure 5.1: Construction of the Ingold CO2 probe. Reprinted from the Ingold productcatalogue with permission from Ingold Electrodes Inc..67with a stainless steel mesh and a nylon net, mounted on a sterilizable plastic membranecartridge with a stainless steel sleeve and a silicone rubber washer. The internal electrodewas a glass pH electrode. The pH electrode was attached to a threaded shaft and could beraised from, and lowered to, the silicone membrane by turning a knurled knob at the endof the shaft. The probe body contained inlet and outlet tubes used for injecting electrolyteand pH calibration standards into the electrolyte chamber and was equipped with thenecessary fittings for insertion into the 25 mm side-ports of Chemap fermenters.The basic hardware of the Ingold CO2 probe had many of the features of theproposed glucose biosensor hardware described in the schematic diagram of Figure 2.2,made available in a convenient, fermenter compatible package. Several modificationswere performed to convert the Ingold probe to the prototype glucose biosensor:1. A cellulose matrix was incorporated into the electrolyte chamber.2. The silicone membrane was replaced with a glucose-permeable membrane.3. The internal pH electrode was replaced with a custom designed electrodeassembly with three electrodes for amperometric operation.4. To fit the new internal electrode unit into the probe body, a customdesigned adapter was required to mount the electrode assembly onto the threaded shaftused for raising and lowering the internal pH electrode.5.^The syringes used for injecting electrolyte and pH electrode calibrant intothe CO2 probe were replaced with tubing, valves, and a peristaltic pump.Cellulose matrix:The characterization of different cellulose matrices is described in Chapter 4. Thecellulose matrix used in the experimental prototype was a disk of Whatman No. 1qualitative filter paper (Whatman International Ltd., Maidstone, England) cut to68approximately 7 mm in diameter. The cellulose matrix was sandwiched between theinternal electrode and the glucose-permeable outer membrane.Glucose-permeable outer membrane:The original silicone membrane of the CO2 probe was removed by separating thesteel sleeve from the plastic membrane cartridge and dissolving the silicone with SiliconeSealant Remover (Dow Corning Canada Inc., Mississauga, Ontario, Canada). The newglucose-permeable membrane was draped over the membrane cartridge and secured usingsteel wire and/or Silastic medical adhesive (Dow Corning Corporation, Medical Products,Midland, MI, U.S.A.). The steel sleeve was placed over the membrane and cartridge andsealed in place with Silastic adhesive.A number of different membranes were tested as glucose-permeable outermembranes. Candidate membranes were chosen on the basis of structure, strength,molecular weight cut-off, availability, and ease of use. Non-cellulose membranes werechosen where possible to prevent conjugate binding to surfaces other than the intendedcellulose matrix and thereby simplify the system of experimental variables:1. Spectra/Por 2 dialysis tubing from regenerated cellulose, MWCO12-14 kD, cut lengthwise to form a flat sheet (Spectrum Medical Industries, Inc., LosAngeles, CA, U.S.A.),2. Amicon PM10 (MWCO 10 kJ)) and PM30 (MWCO 30 kD) non-celluloseultrafiltration membranes with support backing (Amicon Canada Ltd., Oakville, Ontario,Canada),3. Filtron Omega 10 (MWCO 10 kD) polysulfone ultrafiltration membranewith support backing (Filtron Technology Corporation, Northborough, MA, U.S.A.), and4. A custom designed perfluorosulfonic acid (Nafion) membrane (AldrichChemical Company, Inc., Milwaukee, WI, U.S.A.) cast on a 0.2 )1111 Metricel GA-8(cellulose triacetate) membrane filter (Gelman Instrument Co., Ann Arbor, MI, U.S.A.).69One coat of 250 p.L of 0.5% Nafion (in 50/50 isopropyl alcohol in water) was solutioncast on a 25 mm diameter membrane filter with a 50 p,m stainless steel mesh for rigidsupport. The membrane filter and stainless steel mesh were secured onto the membranecartridge with steel wire before casting the Nafion membrane. The Nafion solution wasallowed to dry in air for at least 1 hour before the stainless steel sleeve was placed overthe membrane cartridge and sealed with Silastic adhesive. The stainless steel mesh facedthe interior of the membrane cartridge such that the Nafion coated membrane filter was onthe exterior and would be in direct contact with the fermentation broth. Using thisarrangement, a smooth Nafion coating could be cast on the outer surface which wouldlikely behave in a more well-defined manner in a stirred solution than the steel mesh. Thesteel mesh would also provide structural support for the membrane during the highpressure steam sterilization of the fermenter.Internal electrode assembly:The internal electrode assembly designed and built in-house is shown in Figure 5.2.The three-electrode unit had an outside diameter of 6 mm and consisted of a Pt indicatingelectrode, Pt counter electrode, and Ag/AgC1 reference electrode. The Pt indicatingelectrode was a 1.0 mm diameter platinum wire (Aldrich, Milwaukee, WI, U.S.A.) in aglass shroud made from 4 mm O.D. flint glass tubing. One end of the glass tubing waspartially closed by melting the glass over a Bunsen burner, such that the inside diameterwas just greater than 1 mm. The platinum wire was then bonded into the glass usingChemgrip epoxy (Norton Co., Wayne NJ, U.S.A.) to form a seal. The glass-shroudedplatinum wire was ground and sanded with successively finer grades of sandpaper,followed by polishing to a mirror finish with 0.3 pm and 0.05 p.m alumina, leavingexposed a circular platinum disk with a surface area of 0.785 mm2 .70Pt INDICATING Pt COUNTER AcI/AQC1 REFERENCEELECTRODE^ELECTRODE^ELECTRODEGLASS^EPDXY^GOLD^LEADSHROUD CRIMPS WIRESFigure 5.2: Diagram of the internal electrode assembly. The glass-shrouded platinumindicating electrode, platinum counter electrode, and silver-silver chloride referenceelectrode are shown. The outside diameter of the three-electrode unit was 6 mm.71The counter electrode and reference electrode were 1.0 mm diameter Pt and Agwire, respectively (Aldrich, Milwaukee, WI, U.S.A.), tightly coiled around the glassshrouded indicating electrode. The surface area of the counter electrode was muchgreater than the surface area of the indicating electrode to ensure that the area of thecounter electrode was not charge transfer limiting. The Ag/AgC1 reference electrode wasfabricated by anodizing the Ag wire in the presence of Cl', based on the method of Sawyerand Roberts (1974). The Ag wire was potentiostated at +0.2 V versus SCE for 8 hours,using a Pt wire counter electrode and an electrochemical cell containing 0.1 M KCI.Modem cable was used for the electrode leads, because the shielding of theindividual wires in the modem cable allowed the reference electrode lead to be shieldedfrom the other electrode leads. The electrode leads were attached to the electrode wiresusing gold crimps and soldered.Immediately before use, the bare Pt indicating electrode was cycled in quiescent0.5 M H2SO4 from -0.26 V to +1.2 V at 100 mV/s for 10 minutes, anodized at +1.8 Vfor 10 minutes, then cycled again until a stable cyclic voltammogram was obtained. Forsome experiments, described below, the indicating electrode was coated with celluloseacetate using a modification of the method described by Wang and Hutchins (1985). A7.5 piL drop of 2.5% cellulose acetate (BDH Ltd., Poole, England) in a 1:1 solution ofacetone and cyclohexanone (stirred for 12 hours) was applied to the indicating electrodeand allowed to dry overnight. The indicating electrode was cycled and anodizedimmediately before coating with cellulose acetate.Mounting adapter for the internal electrode unit:The stainless steel adapter used to mount the internal electrode unit to theelectrode shaft of the probe body is shown in Figure 5.3. The flange on the adapter was72Figure 5.3: Stainless steel adapter for mounting the internal electrode unit. The threadedend mounts onto the electrode shaft of the probe body; the three-electrode unit protrudesfrom the opposite end. The electrode lead wires were routed through the inside of theadapter and the electrode shaft to the opening at the top of the probe body. The flange onthe adapter prevented over-insertion of the electrode unit by stopping against a steelshoulder inside the probe body. Drawing not to scale.73designed to meet a shoulder inside the probe body to prevent the electrode shaft frombeing inserted too far into the probe body. An 0-ring was used between the adapter andthe inside wall of the probe body to maintain a sealed internal chamber. An insulatingwrap of black electrical tape was used between the electrodes and the adapter. Silasticadhesive was used to bond the internal electrode unit into the adapter and provide a liquidseal. The distance from the tip of the internal electrode unit to the flange on the adapterwas set at 35 mm, such that the indicating electrode would contact the outer membranewhen fully lowered. It was also necessary to bore the minimum inside diameter of theprobe body from 8.13 mm to 9.55 mm.Reagent flow system:A Gilson Minipuls 3 peristaltic pump (Gilson Medical Electronics, Middleton, WI,U.S.A.) was used to pump the appropriate reagent solutions through silicone tubing intothe probe body. A combination of three-way valves, shown in Figure 5.4, was used toselect the reagent to be pumped. The three reagent reservoirs contained 1) the internalelectrolyte and washing buffer, 2) glucose standard, for internal calibration, and 3) theelution buffer. The internal electrolyte and washing buffer was PBS (0.1 M NaCl, 5 mMNaH2PO4, and 30 mM Na2HPO4, pH 7.4, pt. = 0.2 M, preserved with 1 mM EDTA and5 mM sodium benzoate). The elution buffer was PBS with 8 M guanidine hydrochloride.At this stage of the prototype development, pump control and valve switching wasperformed manually.The protocols used for loading and eluting the GOx-CBD conjugate are listedbelow. Flow rate was calibrated against pump speed, and the volumes of the differentsegments of the flow system shown in Figure 5.4 were determined by measuring the timerequired for a fluid to pass through the different segments at a given flow rate. The timerequired to wash all unbound GOx-CBD conjugate out of the enzyme chamber after the74R1 R2 R3 R45.6 mLsFigure 5.4: Schematic diagram of the biosensor prototype showing the reagent flowsystem and instrumentation. RI, internal electrolyte and wash buffer reservoir (PBS); R2,internal calibrant reservoir (not used); R3, elution buffer reservoir (8 M guanidine in PBS);R4, waste reservoir; VI and V2, three-way valves; P, peristaltic pump; E, enzymechamber; PO, potentiostat; CR, chart recorder. The volume of the each segment of theflow lines and the enzyme chamber are shown. The volume of the enzyme chamberdepended on the position of the internal electrode unit (i.e., raised or lowered). Solid anddashed lines represent flow lines and electrical connections, respectively.75enzyme loading cycle was determined by collecting fractions of the effluent from the probewhile washing the enzyme chamber with PBS. The presence of protein in the fractionswas determined by measuring the absorbance at 280 nm versus fresh PBS. The timerequired to wash all traces of the guanidine elution buffer out of the enzyme chamber afterthe enzyme elution cycle was determined using the Pt indicating electrode as a detector forguanidine at the normal operating potential of +0.7 V versus Ag/AgCl. The enzymechamber was washed with PBS until the sensor signal returned to baseline.Enzyme loading protocol:1. Raise the internal electrode assembly.2. Pump GOx-CBD conjugate solution for 1.0 minute at 4.0 rpm (i.e., 2.2 mls/rnin)3. Pump PBS for 4.5 minutes at 4.0 rpm.4. Stop flow for 1.0 minute. At this point the bolus of GOx-CBD conjugate has filledthe enzyme chamber.5. Pump PBS for 7.5 minutes at 4.0 rpm to wash unbound conjugate from theenzyme chamber.6. Lower the internal electrode assembly.The GOx-CBD conjugate was pumped into the flow system by removing the inlet tubingfrom the PBS reservoir and inserting into a reservoir of conjugate. The soluble conjugatesolution prepared for use in the enzyme chamber of the prototype biosensor was made0.1 M in NaCI to approximate the composition of the PBS electrolyte and satisfy therequirements of the Ag/AgC1 reference electrode. Once a bolus of the GOx-CBDconjugate was loaded into the flow system, the tubing inlet was rinsed with distilled waterand re-inserted into the PBS reservoir. PBS was pumped through the flow system to pushthe bolus of GOx-CBD conjugate into the enzyme chamber. The total time required forthe loading protocol was 14.0 minutes.76Enzyme elution protocol:1. Raise the internal electrode assembly.2. Pump guanidine elution buffer for 2.5 minutes at 4.0 rpm.3. Pump PBS for 2.0 minutes at 4.0 rpm.4. Pump PBS for 5.5 minutes at 48.0 rpm (maximum pump rotation speed)5.^Lower the internal electrode assembly.Separate tubing inlets were used for the guanidine elution buffer and the PBS wash buffer,controlled by a three-way valve. The total time required for the elution protocol was 10.0minutes.Dummy electrode:In addition to the modifications described above, a stainless steel plug, or "dummyelectrode", shown in Figure 5.5, was fabricated so that the probe body could be steamsterilized in a fermenter with the internal electrode assembly removed. The internalelectrode assembly as constructed could not be autoclaved or the Chemgrip epoxy wouldbecome brittle and crack. A dummy electrode was required as a plug when the internalelectrode assembly was removed, in case failure of the probe membrane during steamsterilization released superheated steam through the electrode shaft.5.2.2. Prototype Characterization ExperimentsPrototype characterization experiments were performed in the absence of cells in a150 mL beaker and a Chemap Type SG 3.5 L fermenter (Chemap AG, Switzerland).Experiments in the 150 rnL beaker were performed using 100 mL of PBS for electrolyteand the probe was inserted to the 70 mL mark on the 150 mL beaker for everyexperiment. The electrolyte was stirred by a magnetic stir bar and an air-driven magnetic77Figure 5.5: Stainless steel dummy electrode. The dummy electrode was used to replacethe three-electrode unit during the steam sterilization process. Drawing not to scale.78stirring module from a Hewlett-Packard spectrophotometer (HP 89055A, Hewlett-Packard Canada Ltd., Mississauga, Ontario, Canada). A regulator was used to control airflow, and the air pressure was adjusted to 9 psig for each experiment. The indicatingelectrode was potentiostated at +0.7 V versus Ag/AgCI using a Pine AFRDE4 bi-potentiostat (Pine Instrument Co., Grove City, PA, U.S.A.) and the sensor output wasrecorded on a Kipp & Zonen Model BD 112 strip chart recorder (Kipp & Zonen,Delft/Holland). The probe was calibrated using aliquots of 0.1 M glucose standardprepared in PBS.To test the performance of a given sensor configuration in complex medium, thesensor was first calibrated in PBS. Residual glucose in the membrane and the enzymechamber was removed by pumping PBS internally and washing the probe externally infresh PBS until a stable baseline was obtained. The electrolyte in the 150 mL beaker wasthen replaced with fresh Luria Broth (10g/L tryptone (E. Merck, Darmstadt, Germany), 8g/L yeast extract (Merck), and 5 g/L NaCI, pH 7.2) and recalibrated. Antifoam C (SigmaChemical Co.) was added at a concentration of 1.6 mL/L to investigate the effect on thesensor signal.Experiments to characterize the effect of medium dissolved oxygen tension, pH,temperature, stir rate, and air flow rate were performed in the 3.5 L fermenter using theChemap Type 3000 base unit and Type FZ3000 control unit. The fermenter had three 25mm side-ports and the standard blade stirrer was installed. An Ingold InFit 764-50sterilizable pH electrode (Ingold Electrodes Inc., Wilmington, MA, U.S.A.) and an Ingoldsterilizable 02 electrode (No. 401814-06, 25 mm 0.D.) were used in two of the fermenterside-ports. The third side-port was used for the glucose biosensor prototype. Thebiosensor system used the same potentiostat, chart recorder, and reagent flow system asdescribed above. The fermenter was filled with 1.75 L of PBS and aliquots of 1 Mglucose standard in PBS were used for calibration. For the pH sensitivity experiments, the79fermenter was filled with 1.75 L of unbuffered saline (0.1 M NaCI) and the pH wasadjusted using HCl or NaOH.5.2.3. Glucose Monitoring During Fed-Batch Cultivation of E. coli:Organisms:A strain of E. coli JM101/pTUgEO7K3 was used for the cultivation. Theorganism contained the plasmid for the production of CBDCex and was stored in 10%DMSO at -70 °C. The plasmid consisted of the tac promoter and the leader sequence ofC. fimi exoglucanase (Cex), followed by the structural gene for CBDc ex . The resistancemarker was kanamycin and the inducer was IPTG (E. Ong, unpublished results). Theinducer was not added in this experiment.Media:Minimal medium M-9 was prepared with the following composition (g/L):Na2HPO4, 11.76; KH2PO4, 5.88; NaCI, 0.5; NH4C1, 1.0; MgSO4, 0.49; CaC12, 0.01;thiamine, 1.685; kanamycin, 0.025; and the following trace metals (mg/L):Al2(SO4)3 .7H20, 0.040; CoC12.6H20, 0.032; CuSO4.5H20, 0.008; H3B03, 0.004;MnC12•4H20, 0.080; NiC12 , 6H20, 0.004; Na2Mo04-2H20, 0.020; ZnSO4.7H20, 0.020.The starting glucose concentration was 2.40 g/L. Medium for the inoculum was preparedas follows (g/L): Na2HPO4, 6.00; KH2PO4, 3.00; NaC1, 0.5; NH4CI, 1.0; MgSO4, 0.49;CaCl2, 0.01; thiamine, 3.37; kanamycin, 0.05; glucose, 2.80 g/L.Cultivation:The cultivation was performed in a 20 L Chemap Type SG fermenter with thestandard blade stirrer and three 25 mm side-ports. The fermenter working volume was 8L. A 500 mL inoculum was prepared in shake flask culture. One side-port was used forthe Ingold InFit 764-50 pH electrode. A second side-port was used for an optical density80monitor (Cerex MAX Cellmass Sensor Probe, Cerex Corporation, Ijamsville, MD,U.S.A.). The third side-port was used for the glucose biosensor prototype. A 19 mmdiameter Ingold sterilizable 02 probe (No. 40180-03) was inserted through a port in thefermenter head plate. For the purpose of monitoring glucose concentration with theglucose biosensor prototype, it was unnecessary to use aseptic techniques during thefermentation.The Chemap 3000 Series base unit and controller were used. Initially, the air flowrate was set at 6 L/min and stir rate was controlled between 50 and 700 rpm in order tomaintain the dissolved oxygen setpoint at 95% of air saturation. As the fermentationreached higher cell density, it was necessary to increase the air flow rate to 7.5 L/min anddecrease the dissolved oxygen setpoint to 80%. Temperature was controlled at 37 °C.Medium pH was uncontrolled.Analyses:Samples were withdrawn from the fermenter at different intervals using thesampling/harvesting valve. Glucose concentration in each sample was analyzed using theBeckman Glucose Analyzer 2 (Beckman Instruments Inc., Fullerton, CA, U.S.A.) aftercentrifuging the sample at 14,000 rpm for 2 minutes. The absorbance of each sample at600 nm (versus distilled water) was also measured using a Varian DMS 200 UV-VISspectrophotometer (Varian Pty. Limited, Mulgrave, Victoria, Australia). In addition, theGenesis Control Series software package (Iconics, Foxborough, MA, U.S.A.) was used tolog temperature, pH, dissolved oxygen tension, stir rate, and optical density from the on-line sensors every five minutes during the course of the fermentation.5.3. RESULTS AND DISCUSSIONIn all experiments reported here, the operation of the pump and valves and theraising and lowering of the internal electrode assembly was performed manually in order to81simulate the automated process. It was beyond the scope of this thesis to produce a fullyautomated prototype; however, in subsequent development phases of the prototype anautomated version of the system described here can be realized by interfacing the pumpwith a personal computer. In addition, the valves can be actuated by computer controlledelectric or pneumatic actuators, and the operation of raising and lowering the internalelectrode assembly, which is accomplished by turning a knurled knob on a threaded shaft,can be driven by a computer-controlled stepper motor. The sensor output can bemonitored by a personal computer rather than a chart recorder, such that the data can beused in computer algorithms for sensor calibration, self-diagnosis, and regeneration, aswell as feedback control algorithms for bioprocess control.5.3.1. Prototype Characterization:In general, the performance of the experimental prototype fulfilled expectationsand was consistent with the results of the preliminary experimental studies described inChapter 4. The response of the prototype could be calibrated with respect to glucoseconcentration up to at least 23 mM (the maximum concentration tested) in mediumwithout cells. The sensor signal was relatively stable and noise-free, demonstrating lessthan 5% variation per hour and signal noise less than 1% of the sensor current. The GOx-CBD conjugate could be loaded and eluted successfully using the inlet and outlet tubing ofthe modified Ingold probe body and the loading and elution protocols described above.After elution of the enzyme, the sensor response to glucose was less than or equal to thebaseline signal, confirming that the enzyme had indeed been eluted.The Whatman qualitative filter paper, characterized in Chapter 4, proved to be asatisfactory cellulose matrix for the biosensor prototype. The low mass transfer resistanceof the filter paper compared to the other cellulose matrices tested was advantageous interms of fast sensor response time and low signal attenuation due to the matrix. The filter82paper also had a high porosity and cellulose surface area for binding the GOx-CBDconjugate, and was anticipated to be more easily perfused with the reagent solutions thanthe regenerated cellulose dialysis membrane or the nitrocellulose protein transfermembrane. Structural stability of the filter paper was not a problem, as in the RDEexperiments, due to the absence of shear stress from stirring or rotation. Cotton battenwas also tested as a potential cellulose matrix, but the relatively uniform thickness andincompressibility of the filter paper was preferred for experimental reproducibility.Characteristics of the glucose-permeable outer membrane:Table 5.1 lists the characteristics of a number of different membranes tested aspossible glucose-permeable outer membranes for the prototype. Four properties weredetermined to be essential for the outer membrane:1. The membrane must be autoclavable.2. The membrane must be sufficiently permeable to glucose and oxygen sothat the sensor response time is fast enough to follow changes in the glucose concentrationin the medium. For example, a high cell-density cultivation of E. coli with an opticaldensity of 40 (measured at 600 nm versus distilled water) can consume 2.5 g of glucosefrom 1 L of medium in approximately 10 minutes (D. Hasenwinkle and E. Jervis,unpublished results). A sensor response time of less than five minutes was consideredadequate for the present work, although faster response times would certainly beadvantageous.3. The membrane must be impermeable to electroactive species which willcontribute to a high background signal, as well as to medium components which willinhibit or denature the enzyme or poison the platinum surface of the indicating electrodeover the course of a fermentation run.4. The membrane itself must be resistant to fouling by protein or microbialadsorption (for example) over the course of a fermentation run.83Table 5.1: Characteristics of different membranes tested as potential outer membranes forthe glucose biosensor prototype.Dialysis MWCO Steam Performance ResponseMembrane (kD) sterilizable? rating in complexmediumtime(mM)None 5Spectra/Por 2 12-14 Yes 12PM30 30 No Poor 6PM10 10 No Fair 25Omega 10 10 No 00 1Nafion Yes Fair 10Nafion(autoclaved) 2 Yes Fair 3Nafion (autoclaved)/ - Yes Excellent 5Cellulose Acetate2,31 The sensor did not respond to glucose in the concentration range from 0-23 mM withthis membrane.2 The membrane cartridge with the Nafion membrane was submerged in PBS andautoclaved before testing.3 This notation refers to the combination of a Nafion membrane on the membranecartridge (the glucose-permeable outer membrane of the sensor) and a cellulose acetatecoating on the surface of the indicating electrode.No off-the-shelf membrane tested could satisfy all of the required conditions listedabove. A custom membrane was designed using a solution of perfluorosulfonic acid(Nafion) cast on a 0.2 um cellulose triacetate membrane filter with a 50 um stainless steelmesh. Nafion has been used with good results as a dialysis membrane material on G0x/Ptelectrodes for the determination of glucose in whole blood (Harrison et al., 1988). Thepermselectivity of Nafion is due to the rejection of anionic species by the negatively-84charged perfluorinated ionomer membrane, as well as the specific morphology of themembrane. The sterilizability of the Nafion membrane is discussed below. The 0.2 pmmembrane filter acted as a support for casting the Nafion membrane, and also ensured thata sterile barrier was maintained in case of failure of the Nafion coating. The stainless steelmesh provided rigid support.A cellulose acetate coating for the indicating electrode was also developed, to beused in conjunction with a glucose-permeable outer membrane. It was anticipated that thebackground signal in complex media would be high, due to the presence of electroactivespecies in the media that can be oxidized at a potential of +0.7 V. Furthermore, theindicating electrode is susceptible to poisoning by components of the analyte medium.This is particularly true in biological media, which contain proteins and other components(e.g., ascorbate ion) that may adsorb to the Pt surface, thereby attenuating theelectrocatalytic activity of the electrode surface. Dilution of the analyte using an internalbuffer flow through the enzyme chamber can decrease the background signal and slow theprocess of electrode fouling, but the internal buffer flow also causes a decrease in sensorsensitivity. Another approach is to coat the indicating electrode surface with an inert film,such as cellulose acetate, that rejects interfering electroactive species and prevents surfaceadsorption. Cellulose acetate films coated on the surface of platinum electrochemicaldetectors in liquid chromatography systems have been shown to prevent electrode foulingfrom protein adsorption and eliminate some electroactive interferences, while at the sametime permitting rapid diffusion of H202 (Sittampalam and Wilson, 1983; Wang andHutchins, 1985). Detector response time was not significantly different with the additionof the cellulose acetate coating (Wang and Hutchins, 1985). Cellulose acetate films havealso been shown to protect the Pt electrode component of glucose sensors for medicalapplications (Yamasaki, 1984). The most significant drawback of the coating is theattenuation of the electrode signal resulting in a loss of sensitivity (Sittampalam andWilson, 1983; Wang and Hutchins, 1985; Kuhn et al., 1989) However, controlled base85hydrolysis of the cellulose acetate film is known to increase the porosity of the film byremoving acetate functional groups, and has been shown to decrease the electrode signalattenuation as a result of the cellulose acetate film (Wang and Hutchins, 1985). Anadditional advantage of the cellulose acetate coating is that the cellulose polymer will bindthe GOx-CBD conjugate, thereby increasing the amount of enzyme activity that can beloaded. It was also found that the conjugate could be eluted from cellulose acetate usingthe 8 M guanidine elution buffer and that the coating was not degraded by the guanidine.A 2.5% solution of cellulose acetate in acetone and cyclohexanone (1:1) was usedfor coating the indicating electrode in some of the prototype biosensor experiments withthe Nafion membrane.' Experiments were performed to investigate the signal attenuationdue to the cellulose acetate coating. Identical Pt electrodes were coated with differentconcentrations of cellulose acetate and the electrode response in PBS was compared usingaliquots of hydrogen peroxide. Base hydrolysis of some electrodes was performed in 0.07M KOH. The results revealed that the least signal attenuation was obtained using a 2.5%cellulose acetate coating, without performing base hydrolysis (see Figure 5.6). The use oflower concentrations of cellulose acetate or longer periods of base hydrolysis were nottested as the membrane might become too fragile. The Pt electrode with the 2.5%cellulose acetate coating was also compared to a bare Pt electrode in Luria broth (data notshown). The electrode with the cellulose acetate coating showed a lower electrodecurrent in response to aliquots of hydrogen peroxide than the bare Pt electrode, but thebackground current and signal-to-noise ratio of the cellulose acetate coated electrode wasalso lower. Both electrodes demonstrated signal attenuation in Luria broth compared toPBS.1 The use of the custom-designed Nation membrane (described above) as the glucose-permeable outermembrane of the sensor in conjunction with a cellulose acetate coating on the surface of the indicatingelectrode is denoted as "Nafion/cellulose acetate".86—I— 2.5%; no hydrolysis—A-- 5.0%; no hydrolysis—4,— 5.0%; 15 min hydrolysis—• 5.0%; 30 min hydrolysis—+— 5.0%; 45 min hydrolysis01—8♦,----1/•----------1----------A4.--------^--+_____-------_-: • '^I^I^I^I^I• I^r0.000 0.001 0.002 0.003 0.004 0.005 0.006% Hydrogen PeroxideFigure 5.6: Comparison of the response of different cellulose acetate coated Pt electrodesto hydrogen peroxide. Different concentrations of cellulose acetate in a 1:1 solution ofacetone and cyclohexanone were used. Base hydrolysis was performed in 0.07 M KOH.Each electrode was potentiostated at +0.7 V vs SCE in PBS and the electrode current wasreported relative to a bare Pt electrode.87Performance in complex media:By far, the most demanding conditions in which the candidate membranes will beused are the conditions of a real fermenter run, due to the presence of cells, cellularmetabolites, proteins, and other medium components. Not only must the membrane beautoclavable and enable a fast sensor response time, but the membrane must contribute tostable sensor operation over the course of the entire fermentation. The potential of eachmembrane to perform satisfactorily in the fermentation environment was evaluated in150 mL beaker experiments using Luria broth as a complex medium for testing. Theprototype sensor with a given membrane was first calibrated in PBS and the response timeto the addition of aliquots of glucose was noted (Table 5.1). The electrolyte was thenimmediately changed to Luria broth and the sensor was recalibrated in order to assess theperformance of the given membrane in complex medium. The time between obtaining thetwo calibration curves was minimized to lessen the possible effect of enzyme deactivationon the sensitivity of the second calibration curve. The addition of Antifoam C to the Luriabroth was found to have no effect on the sensor signal at steady state.In order to compare the results of a series of experiments using differentmembranes on the same graph, the sensor calibration curve established in PBS for eachsensor configuration was normalized with respect to the maximum sensor current (i.e., at23 mM glucose) to give a maximum dimensionless sensor current of 1. The calibrationcurve in Luria broth from a given experiment was then expressed relative to the calibrationcurve in PBS by normalization with the same value, since the only experimental variablechanged was the electrolyte/medium.It can be seen from the results in Figure 5.7 that, in all cases, significant signalattenuation and sensitivity loss was observed in complex medium compared to definedmedium (i.e., PBS). This phenomenon has been reported for other enzyme electrodes, as88—0— PM30(PBS)-A- PM10(PBS)—o Nafion(PBS)—o— Nafion/CelluloseAcetate(PBS)—m— PM30(L-B)-A- PM10(L-B)Nafion(L-B)Nafion/CelluloseAcetate(L-B)0^5^10^15^20^25Glucose Concentration (mM)Figure 5.7: Normalized calibration data for the prototype biosensor in PBS and Luriabroth (L-B) using different membranes. For each membrane, the prototype was firstcalibrated in PBS and then recalibrated immediately afterwards in Luria broth. The Nafionmembrane used had been autoclaved previously.89well as ion-selective electrodes, mass spectrometers, and gas chromatographs, (Clelandand Enfors, 1983; Merten et al., 1986; Locher et al., 1992). Although not completelyunderstood, the observations are attributed to effects of the analyte matrix. A number ofpossible explanations have been forwarded, however the phenomenon may be due to morethan one chemical effect, or a symbiosis of several different effects:1. It is the activity and not the concentration of the analyte that is measuredby chemical sensors. The equality of concentration and activity is true only at infinitedilution. The presence of additional molecular species, as in fermentation media, mayreduce the activity of the analyte (Merten et al, 1986).2. Low-molecular weight molecules may complex with macromolecules inbiological media, thereby reducing the chemical activity with respect to concentration(Merten et al, 1986), the mass transfer through the sensor membrane, or the affinity of theenzyme for the substrate.3. Proteins, lipids, and other hydrophobic components of biological mediamay occupy a non-aqueous compartment of the solution which is not accessible to theanalyte. Thus the volume occupied by the analyte is lower than the total volume (Mertenet al., 1986). This does not necessarily explain the observed signal attenuation and loss ofsensitivity, but may explain the poor correlation in analytical results obtained by differenttechniques (Bradley et al., 1989a).4. Changes in the composition of the liquid phase may change the propertiesof mass transfer through membranes significantly (Locher et al., 1992). For example, thepartitioning coefficient of the membrane/solution interface may differ depending on thesolution. Protein or microbial adsorption to the membrane may increase the mass transferresistance of the membrane.5. In the case of enzyme electrodes, components of the medium may inhibit ordenature the enzyme or poison the electrode, causing a decrease in the measured sensorsignal or a loss of sensitivity. For example, competitive inhibition of glucose oxidase by90D-glucal (a substrate analog) and/or halide ions (CI - , Br, 1-) has been reported by Rogersand Brandt (1971). (However, published results by Cleland and Enfors (1983) for aglucose oxidase electrode have not shown any effect by chloride ion at concentrations upto 0.26 M NaCl. The immobilization of the enzyme was thought to prevent the inhibitoryeffect of Cl" ion.) In addition, other components of the medium, such as ascorbate, areknown to adsorb to platinum surfaces, forming a monolayer which blocks electrochemicalreactions at the surface, thereby attenuating the sensor signal and eventually poisoning theelectrode.6. In the case of oxidase enzyme-based systems, the analyte matrix may affectthe activity or concentration of dissolved oxygen, which is required as the electronacceptor for the reduced form of the enzyme cofactor FADH2. The solubility of oxygen,for example, is reduced by high concentrations of ionic species.Figure 5.8 shows the equilibration time of the prototype sensor when inserted intoLuria broth, before the addition of glucose. All of the membranes tested, except theNafion/cellulose acetate combination, demonstrated an initial current response peak whichthen decayed to a stable background. Although the mechanism for this behaviour is notclear, the initial peak in response was attributed to the oxidation of electroactive species inthe medium, and the subsequent decay of the peak was thought to be a result of membranefouling and/or electrode poisoning by adsorption of proteins or other species in themedium. The absence of the initial response peak in the case of the Nafion/celluloseacetate combination membrane may have been due to the protective cellulose acetatecoating on the indicating electrode and the permselective properties of the Nafionmembrane. The Nafion/cellulose acetate combination membrane also established thelowest background signal compared to the other membranes investigated, indicating a highrejection of interfering species, as expected.910 250.20 -00" PM30PM10   Nafion Nafion/CelluloseAcetate-^•^....................1^i^15^10^15 20 25^30Time After Immersing Sensor in Medium (min)Figure 5.8: Sensor equilibration time after insertion in Luria broth using differentmembranes. Equilibration time was the time for the sensor signal to reach a stable baselinebefore the addition of glucose.92These characteristics, in addition to the relatively fast sensor response time andlow signal attenuation and sensitivity loss in Luria broth (refer to Figure 5.7), pointed tothe Nafion/cellulose acetate membrane system as the best choice (among the membranesinvestigated here) for use in the biosensor prototype during glucose monitoring of a realfermentation. The sterilizability of the Nafion membrane was established by autoclavingthe membrane cartridge while submerged in PBS. After autoclaving the membrane andrecalibrating the sensor in PBS, the sensor response time was found to have decreasedfrom 10 minutes to 3 minutes. This may have been due to the dissolution of a smallfraction of the Nafion coating during the high temperature process of autoclaving (Mooreand Martin, 1986), resulting in a decrease in mass transfer resistance in the membrane anda faster sensor response time. Most importantly, however, Figure 5.9 demonstrates thatthe performance of the autoclaved Nafion membrane in complex medium was notsignificantly affected.Effect of temperature and pH:Using the Chemap fermenter control unit and the 3.5 L fermenter, a number offermenter operating parameters could be varied to investigate the effects on the sensorsignal at steady state. The effect of temperature was investigated over the range of normaloperating temperatures for industrial fermentations (see Figure 5.10). A directrelationship between the sensor signal and temperature was expected due to temperature-dependent increases in the reaction rate constant and the diffusion coefficient, as predictedby Arrhenius' law and the Stokes-Einstein equation, respectively. However, the decreasein medium dissolved oxygen concentration at higher temperatures (as a result of loweredoxygen solubility) may also have reduced the sensor signal due to limitation of the enzymekinetics. At temperatures greater than 40 °C, the enzyme glucose oxidase is reported tobe unstable (Nakamura et al., 1976). Fortier et al. (1990) have investigated the effect oftemperatures greater than 40 °C for glucose oxidase immobilized in polypyrrole on a Pt93—0— Nafion(PBS)-A- Autoclaved Nafion(PBS)■ Nafion(L-B)—A-- Autoclaved Nafion(L-B)1.0 -I^I^I^i5 10 15 20Glucose Concentration (mM)250.8 -4E'08 0.6 -ca)(.0a). 11- 0.4te5z0.2 -0.00Figure 5.9: Normalized calibration data for the prototype sensor in PBS and Luria broth(L-B) using the Nafion membrane before and after autoclaving. The prototype was firstcalibrated in PBS and then recalibrated immediately afterwards in Luria broth using apristine Nafion membrane. The experiment was then repeated after the membranecartridge had been removed, autoclaved while submerged in PBS in a media bottle, andreplaced.9420 25 30 35 400.0200.015 -0.010 -2 0.005 -a)0)0.00015■■4574 980.025 ^5 0.020 -y).. 0.015 -'50 0.010 -C-)cna)c 0.005 -tn0.0003 5^6Temperature (° C)Figure 5.10: Effect of temperature on the sensor signal at steady-state. Temperature wascontrolled using the Chemap fermenter control unit. Fermenter parameters were:medium, PBS; glucose concentration, 14 mM; pH, 7.4; stir rate, 150 rpm; air flow rate, 0L/min; dissolved oxygen tension, >98% of air saturation. The Nafion/cellulose acetatemembrane system was used.pHFigure 5.11: Effect of medium pH on the sensor signal at steady-state. Medium pH wasadjusted manually using HC1 and NaOH. Fermenter parameters were: medium, 0.1 MNaCI in distilled, deionized water; glucose concentration, 14 mM; temperature, 37 °C; stirrate, 150 rpm; air flow rate, 0 L/min; dissolved oxygen tension, >98% of air saturation.The Nafion/cellulose acetate membrane system was used.95electrode and reported adecrease in sensor current, due to temperature-induceddenaturation of the enzyme. However, the effect was found to be reversible up to atemperature of 50 °C.The effect of medium pH on the steady-state sensor signal is shown in Figure 5.11,which demonstrates a maximum at pH 5. The pH sensitivity of the sensor is largely due tothe pH dependence of the enzyme activity itself. Fortier et al. (1990) have reportedsimilar results, with a maximum observed at pH 6. Studies of the reaction rate of solubleglucose oxidase from pH 3-8 have found an optimum pH of 5.5-5.6 when oxygen is theelectron acceptor for the reduced form of the FAD (Wilson and Turner, 1992). Glucoseoxidase has been reported to be stable over the range from pH 3-10 (Bright and Porter,1975). However, in this study the sensor signal was seriously reduced at pH values below3.5, and at pH values greater than 8 the binding of the CBD might be disrupted (Kilburn etal., 1992).Effect of medium dissolved oxygen tension, stir rate, and air flow rate:The effect of medium dissolved oxygen tension was investigated in the 3.5 Lfermenter by sparging the fermenter simultaneously with nitrogen and air and adjusting theflow rates of the two gases to control the medium dissolved oxygen at various levels. Adirect relationship was observed between dissolved oxygen level and the steady-statesensor signal (see Figure 5.12). The requirement for oxygen as the electron acceptor toturn over the reduced form of the flavin group of glucose oxidase during the oxidation ofglucose has been discussed in Chapter 2. At high dissolved oxygen concentrations, wherethe enzyme kinetics are glucose limited, variations in the dissolved oxygen level are notcritical. The effect of medium dissolved oxygen on the sensor output becomes importantin fermentations where the dissolved oxygen level experiences large fluctuations. FromFigure 5.12, it can be seen that a constant dissolved oxygen concentration would have tobe maintained in order to eliminate the dissolved oxygen dependence of the sensor output.960.0200.015 -0.000 I^I^I^1^10^20 40 60 80 100p02 (% air saturation)Figure 5.12: Effect of medium dissolved oxygen tension on the sensor signal at steady-state. The fermenter was sparged simultaneously with nitrogen and air and the flow ratesof the two gases were adjusted to control the medium dissolved oxygen at different levels.Fermenter parameters were: medium, PBS; glucose concentration, 14 mM; pH, 7.4;temperature, 37 °C; stir rate, 150 rpm. The Nafion/cellulose acetate membrane systemwas used.97Dissolved oxygen control during a fermentation is easily accomplished usingcurrently available technology for feedback control of aeration and stir rate. The effect ofvariation of stir rate and air flow rate on the steady-state sensor signal was investigated inthe 3.5 L fermenter. The sensor prototype was found to be insensitive to stir rate over therange from 300 to 500 rpm. A 3% change in the steady-state signal was observed overthe range from 25 to 200 rpm, indicating that the sensor signal was limited to a smallextent by external mass transfer resistance in this range. At zero stir rate, the sensor signalbegan to increase, which was attributed to H202 accumulation in the enzyme chamber.Experiments with aeration demonstrated a 1-2% decrease in the steady-state signal withthe initialization of air flow, but no further change was observed over the range of air flowrates from 3 to 7 L/min. It should be noted that these observations were recorded for theprototype sensor using the Nafion/cellulose acetate membrane system. The results aredependent on the sensor configuration and the mass transfer properties of the membranesystem, however, and as part of the process of optimization of the prototype, the sensorresponse should be re-characterized following each design change.Cleland and Enfors (1983) reported an enzyme electrode for which the relationshipbetween sensor signal and medium dissolved oxygen level could be mathematicallydescribed by the linear relationship:I = Ice * DOT + Io (5.1)where ke is an electrode constant independent of glucose and 10 is a glucose dependentconstant. Thus, software compensation could potentially be used to correct the sensoroutput for the effect of variations in dissolved oxygen during the course of thefermentation. The linear model presented by Cleland and Enfors is not appropriate for thedata presented in Figure 5.12, but an appropriate empirical model could presumably bedetermined for the sensor described in this work. However, accurate characterization ofthe sensor response under varying conditions is essential for this approach.98As the medium dissolved oxygen was exhausted, a significant decrease in thesensor response was observed and the signal approached zero (data not shown). This mayhave been due to depletion of dissolved oxygen in the enzyme chamber by enzymaticconsumption and/or mass transfer into the fermenter medium through the sensormembrane. Glucose oxidase electrodes which depend on oxygen as the electron acceptorcannot be used in anaerobic environments unless oxygen is provided by some internalsource within the probe body (e.g., an oxygenated buffer flow stream), or substituteelectron acceptors, such as ferrocene derivatives (Cass et al., 1984), are used in place ofoxygen to mediate the electron transfer from the enzyme to the electrode. A number ofapproaches have been reported to address the performance of oxidase enzyme electrodesin anaerobic or oxygen-limited media. Unfortunately, the complete consideration of thesemethods is beyond the scope of this discussion, and the reader is referred to the availableliterature (Romette et al., 1979; Cleland and Enfors, 1983; Rishpon et al., 1990; Stoeckerand Yacynych, 1990; Sansen et al., 1992).5.3.2. Glucose Monitoring During Fed-Batch Cultivation of E. coliCultivation of E. coli was performed in a 20 L fermenter for the purpose ofmonitoring glucose concentration with the glucose biosensor prototype. The probe bodywas sterilized in situ but the internal electrode assembly was removed to preventdegradation of the Chemgrip epoxy used to bond the indicating electrode into the glassshroud. If this had occurred, liquid leakage into the glass shroud around the indicatingelectrode would have caused undesirable variations in the electrode current due to theexposure of electrochemically active internal materials contacting the Pt electrode. Otherhigh-temperature adhesives must be investigated to replace the Chemgrip epoxy in theinternal electrode assembly if the electrode is to remain inserted in the probe body duringsterilization (although this may not be essential).99The Nafion/cellulose acetate membrane system was used, but it was discoveredthat the Nafion coated membrane filter stretched during sterilization in the fermenter. Thiswas likely due to the pressure difference which existed across the membrane (i.e., betweenthe interior of the fermenter and the enzyme chamber of the probe) during the sterilizationcycle in the fermenter, thus it may be necessary to design a more rigid support for themembrane to prevent stretching. Alternatively, a pressure equalization manifold could beengineered which would connect the enzyme chamber to the interior of the fermenter (orsome other appropriately pressurized vessel) during the sterilization cycle to equalize thepressure on either side of the membrane. For the purpose of this experiment, however, themembrane cartridge was replaced with a spare and the probe was re-inserted into thefermenter.The GOx-CBD conjugate was loaded using the enzyme loading protocol describedabove, and the sensor was calibrated before innoculation by adding a known amount ofglucose to the medium in a series of aliquots. The sensor calibration curve wasdetermined by comparing the steady-state sensor signal after each aliquot to the calculatedglucose concentration in the fermenter. The Michaelis-Menten equation was fitted to thesensor calibration curve and used as a conversion function to calculate the mediumglucose concentration from the measured sensor signal during the fermentation. Thesensor response time was five minutes or less.After 8 hours, the enzyme was eluted and reloaded in situ. The internal electrodeunit was lowered into contact with the cellulose matrix after elution of the enzyme tomeasure the background signal. During this phase (i.e., the second enzyme loading of theprototype) the background signal was used as the baseline for recalibration of the sensorafter reloading fresh GOx-CBD conjugate. Ideally, the sensor would be recalibrated atthis point using internal calibration standards pumped into the enzyme chamber, however,the protocol for internal calibration has not yet been developed. The sensor wasrecalibrated by adding four aliquots of glucose to the fermenter, and the steady-state100sensor signal was compared to the results of off-line glucose analysis of medium samples(taken once the sensor signal had reached steady-state). This method is not ideal, as theinaccuracies of the off-line glucose analyzer are incorporated into the calibration curve. Inaddition, the medium glucose concentration is changing during the calibration due tocellular metabolism. However, the sensor response time in this experiment was relativelyfast, and it was found that steady-state sensor signals could be obtained within asufficiently short period of time to obtain a useful calibration.The time course of the fermenter variables monitored during the fermentation isshown in Figure 5.13. The fermentation run was carried out for a total of 16.5consecutive hours. Furthermore, the experiment was not terminated due to failure ordeterioration of the probe. The longest experiment reported in the literature (to thisauthor's knowledge) involving glucose monitoring during a fermentation with an in situenzyme electrode probe is 12 hours (Cleland and Enfors, 1983). Most of the experimentalresults reported in the literature ranged from 36 minutes to 5 hours of operation (Bradleyet al., 1988,1989,1991; Cleland and Enfors, 1983, 1984). The longevity of thisexperiment is attributed to the stability of the biosensor prototype provided by theNafion/cellulose acetate membrane system and the capacity for in situ enzymereplacement.Figure 5.14 shows the output from the prototype sensor and the results of the off-line glucose analyses over the course of the experiment. These results demonstrate theeffect of the analyte matrix on the sensor response and the importance of sensorcalibration under appropriate conditions. From Figure 5.14, it is obvious that the sensoroutput correlated more closely with the results of the off-line analyses after recalibrationof the sensor in the fermenter broth with cells, compared to the initial calibration of thesensor in fresh medium without cells. After reloading the enzyme and recalibrating thesensor, the profile of the sensor output followed the profile of the off-line analyses withsubstantially greater fidelity than the preceding phase, correctly indicating the exhaustion101^.10^-8Temperature. ........^..... . ..... . . . . . .. . • . .... . ..pH ^ Optical Density- 7...1:7...................•^ •........0a)?-a)100.....^ .......................^Setpoint changeDissolved Oxygen ^ Stir RateI,^I^i..^I^.^1^,^ 06^8 10^12 14 16Time (hrs)48060- 40- 20800 -Figure 5.13: Time-course of the fermenter variables during fed-batch cultivation of E. coliin a 20 L fermenter. The fermenter was inoculated at time zero. Fermenter variables wererecorded every 5 minutes using the Genesis Control Series software package and on-linesensors. The dissolved oxygen setpoint was changed from 95% of air saturation to 80%as the cells reached higher density .102First Enzyme Loading^Second Enzyme LoadingRecalibrationiT00ca)0 2a)00Enzyme Elutionand ReloadingC.500Time After Innoculation (hrs)• Beckman Analyzer^ Sensor Output^ Corrected Sensor OutputFigure 5.14: Medium glucose concentration measured by the prototype glucose sensorand the Beckman off-line glucose analyzer during fed-batch cultivation of E. coil inminimal medium (M-9) in a 20 L fermenter. The sensor current was converted to glucoseconcentration using the Michaelis-Menten type conversion function in Equation 51. Thecorrected and uncorrected sensor calibration constants for the first and second enzymeloadings are shown in Table 5.2103of the medium glucose and accurately following the infusion of glucose (and withoutsignificant delay. The improved correlation was expected, since the inaccuracies of theoff-line analyzer and the unresolved effects of the analyte matrix were included in thesensor calibration constants after recalibration. Innoculation of the fermenter followingthe initial calibration in fresh medium changed the composition of the sample matrix,therefore calibration of the sensor would have best been performed after innoculation andafter measurement of the background (i.e., prior to loading the enzyme). Ideally, on-linecalibration could be performed without disturbing the fermentation by using a series ofinternal calibration standards in a scheme similar to that proposed by Bradley and Schmid(1992). Alternatively, the sensor could be calibrated by adding glucose to the medium inaliquots after innoculation and determining the substrate concentration after each aliquotby calculation or by using the off-line glucose analyzer (in the same manner as calibrationwas performed following the second enzyme loading in this experiment).The glucose concentrations determined by the prototype sensor were consistentlylower than the results obtained from the off-line glucose analyzer. This behaviour mayhave been due to effects of the analyte matrix and/or systematic differences between thetwo analytical methods. These observations are consistent with similar comparisonspublished in the literature (Merten et al., 1986; Bradley et al., 1989a; Locher et al., 1992).An empirical model was formulated for the sensor calibration curve which was used toexperiment with corrections to the sensor calibration constants in an attempt to fit thesensor output more closely to the off-line glucose analyzer results. The Michaelis-Mentenfunction used for the conversion of the measured sensor current (.1.A) to glucoseconcentration (g/L) was obtained by manipulation of Equation 2.20. An additionalparameter, Io, was included to represent the value of the sensor baseline current which isnormally subtracted from the measured sensor current before conversion, giving:104S — (I - lo ) Km Imax^(I To)^ (5.2)where S is the glucose concentration (g/L), I is the measured sensor current (pA), I o is thesensor baseline (iiA), Imax is the maximum sensor current (piA), and Km is the apparentMichaelis constant (g/L). Using numerical analysis and several initial values for theparameters Io, Imax, and Km, the corrected sensor output shown in Figure 5.14 wasdetermined. The corrected and uncorrected sensor calibration constants for the first andsecond enzyme loadings are shown in Table 5.2.Table 5.2: Prototype sensor calibration constants for the first and second enzymeloadings.Load #1 Load #2Parameter Uncorrected Corrected Uncorrected CorrectedIo (p.A) 0.014 0.0068 0.0068 0.0068Imax GA) 0.047 0.040 0.0127 0.0135Km (mM) 8.6 9.99 6.44 10.10Correlation 0.9797 0.9710 0.9856 0.9917CoefficientThe results were found to be reasonable. The calibration constants for the firstenzyme loading could be corrected significantly by adjusting the baseline to account forthe change in the analyte matrix after innoculation. The value used for the sensor baselinewas the background signal determined during the fermentation (measured by lowering theinternal electrode unit into contact with the cellulose matrix after elution of the enzymeand recording the sensor current). This value was taken to be relatively constantthroughout the experiment, making the assumption that the membrane system effectively105rejected interfering species and resisted fouling during the course of the fermentation. Theapparent Michaelis constant was found to be nearly identical for the first and secondenzyme loading, which is consistent with the results of Chapter 4 for multiple cycles ofenzyme loading and elution using the modified rotating disk electrode. The values ofImax, which can be taken to be representative of the amount of enzyme loaded, were notchanged significantly by the correction procedure.It can be seen from Figure 5.14 that, after applying the corrected sensor calibrationconstants, the sensor output matches the results profile from the off-line analyses muchmore accurately. The transient fluctuations in the sensor output observed at the beginningof the experiment are not normally observed in the medium glucose concentration and arepresumed to be a result of some initial instability in the local environment of the probe(e.g., due to entrapped bubbles), although the actual cause in this case could not bedetermined. In any case, the perturbation was temporary and did not recur. Once theperturbation subsided, the correlation between the corrected sensor output and the off-lineresults was excellent.Although the correlation coefficients reported in Table 5.2 do not show anysignificant change, the cross-correlation plots in Figure 5.15 demonstrate that therelationship between the sensor output and the off-line results was shifted closer to the lineof direct proportionality after correction of the calibration constants. Part B of Figure5.15 also demonstrates that the sensor output and the off-line results correlated when theglucose concentration varied in a non-sequential manner. The cross-correlation plots inFigure 5.15 are consistent with results reported in the literature for the comparison ofdifferent analytical methods for glucose analysis (Bradley et al., 1989a).The approach used above assumes that the off-line glucose analyzer was preciseand accurate and that the sensor output was in error. According to Locher et al. (1992),the accuracy of the measured value of a single sensor can normally be validated bycomparison with alternative measurement methods. Unfortunately, different measuring106^ Sensor Output• Corrected Sensor Output^ Direct ProportionalityA^2.5 ^.....,.';_;-2.0 -=a_50 1.5 —L)2 1.0 -a)u)ow 0.5 -oo=^- 0- 0.0 ^0.0I^I^I^10.5 1.0 1.5 2.0Beckman Glucose Analyzer (g/L)2.5B2.5 ^-J^_--en.;_,""2.0 -no_^-50 1 - 5 —C52 1.0 —a)u)^-m 0.5 -0o^-=-6- 0.0 ^0.0^ Sensor Output• Corrected Sensor Output^ Direct Proportionality2.5I^i^I^.^I^.^I0.5 1.0 1.5 2.0Beckman Glucose Analyzer (g/L)Figure 5.15: Cross-correlation plots of the prototype glucose sensor output and theBeckman off-line glucose analyzer results using the data from Figure 5.14. A. Firstenzyme loading. Data recorded during the period of transient fluctuations in the sensorsignal at the outset of the experiment was not included. B. Second enzyme loading. Oneextreme outlier (which was measured at the instant of glucose feeding) was removed.107methods often will produce different results and the most accurate analytical methodcannot easily be determined. This is especially true in the case of biological systems,which are frequently more difficult to measure accurately than simple physical or chemicalsystems. Comparison of the off-line glucose analyzer results immediately afterinnoculation (2.17 g/L) with the glucose sensor output (1.98 g/L) and the calculatedmedium glucose concentration (2.40 g/L, based on the volume of medium and the amountof glucose added) reveals a significant discrepancy in both methods. The choice of themost accurate and/or reliable analytical method must often be based on experience. In thisexperiment, the results from the Beckman glucose analyzer were used as the standard forcomparison for reasons of practicality, availability, and relative ease of use.To determine the useable lifetime of the enzyme component of the sensor, periodicinternal calibration checks could be performed to determine at what point the activity ofthe enzyme had deteriorated to an unsatisfactory degree. In this experiment, the resultsfrom off-line glucose analysis of medium samples were used as a reference in an attempt toidentify drift in the prototype sensor output which could be attributed to enzymedeactivation over time. This approach makes the assumption that the results from the off-line glucose analyzer were stable and reliable over time. With respect to this, care wastaken to recalibrate the glucose analyzer before analyzing each medium sample. It wasexpected that if some systematic discrepancy existed between the off-line glucose analyzerand the glucose sensor output, the error would be consistent over time unless someprocess of membrane or electrode fouling or enzyme deactivation modified the resultsfrom one (or both) of the sensor(s). Figure 5.16 shows that the ratio of the off-lineglucose analyzer results and the uncorrected glucose sensor output was relatively constantfor a period of up to 6 hours after the initial loading of the enzyme at the outset of theexperiment. After this period, the sensor output began to drift significantly in the case ofthe first enzyme loading. This may have been an indication of a decay in enzyme activity,although in the case of the second enzyme loading, the experiment was terminated 6.5108hours after loading the fresh enzyme, at which time the sensor signal had not yet shownany discernible signs of drift or decay. This suggests that continuous glucose monitoringcould possibly be performed for at least 6 hours before elution and replacement of theenzyme would be necessary. However, further characterization is required to verify theseresults.1099 -0 8-c 7 -a)J) 6-a)u) 5-.0 04 -?.-9 3 _03▪ 2_a)▪ 0m—0— First Enzyme Loading—A— Second Enzyme Loading^^^[313/13 ♦• AA—A♦1^2^3^4^5^6^7^8^9Time After Loading (hrs)Figure 5.16: Ratio of the Beckman glucose analyzer results and the prototype glucosesensor output calculated at various points in time after fresh enzyme was loaded. Theuncorrected sensor data from Figure 5.14 was used for calculation.110CHAPTER 6CONCLUSION6.1 CONCLUDING REMARKSThe results presented in this thesis are the first to demonstrate the concept,feasibility, and utility of a regenerable biosensor based on reversible immobilization of theenzyme using CBD technology. The design and construction of a prototype glucosebiosensor based on this technology and the use of the prototype in a real microbialfermentation represent a significant step toward a practical, industrially acceptable probedesign, and also toward better instrumentation for fermentation monitoring and control.The following objectives have been achieved:1. Glucose oxidase and the cellulose binding domain have been successfullyconjugated. The GOx-CBD conjugate was synthesized chemically using a glutaraldehydelinkage and retained the activity of glucose oxidase and the binding affinity of the cellulosebinding domain.2. The concept of repeatedly loading and eluting the GOx-CBD conjugatefrom a cellulose matrix on a platinum electrode was shown, demonstrating the feasibilityof a regenerable biosensor based on reversible immobilization of the enzyme using CBDtechnology. The loading and elution protocols were defined and a number of cellulose-based materials were evaluated as potential immobilization matrices.3. A prototype glucose biosensor and reagent flow system were designed andbuilt with the capacity for loading and elution of the enzyme-CBD conjugate in order toregenerate the sensor during a fermentation. Enzyme loading and elution protocols weredeveloped for in situ enzyme replacement using the reagent flow system. The prototypeglucose biosensor was used successfully to monitor medium glucose concentration during111a fed-batch cultivation of E. coli in a 20 L fermenter, demonstrating the potential of theproposed biosensor system for on-line bioprocess monitoring and control.In addition, a custom-designed membrane system suitable for use in microbialfermentations was developed for the biosensor prototype. The membrane systemconsisted of a sterilizable, glucose-permeable, outer Nafion membrane and a celluloseacetate coating on the platinum indicating electrode. This membrane system was shown toeffectively reject interfering species and minimize the sensor background signal whilemaintaining high sensitivity and fast sensor response time. The prototype sensor with thismembrane system was used for 16.5 continuous hours in a microbial fermentation withoutfailure or discernible deterioration.An empirical model for the sensor calibration curve was developed based on theMichaelis-Menten equation for enzyme kinetics. The sensor calibration curve could becharacterized in terms of the apparent kinetic parameters Km and Imax and the sensorbaseline current. The model was used successfully during fermentation monitoring as aconversion function to transform the measured sensor current into the correspondingmedium glucose concentration.A sensor system was proposed which addressed the sterilizability and stabilityproblems of enzyme-based biosensors. The proposed sensor system could potentiallyperform the complete process of diagnosis, regeneration, and recalibration under computercontrol without interrupting the fermentation, including the automated, in situ,replacement of the enzyme. The prototype glucose biosensor described in this thesis hasbeen developed with all of the necessary design characteristics in mind, however, it wasbeyond the scope of this project to completely develop all of the features of the proposedsystem. Furthermore, the system functions described could be executed manually usingthe present version of the prototype, and the modifications required to automate thesystem have been discussed.112The development of more advanced and sophisticated fermentation processes hasgiven rise to an immediate need for improved instrumentation. In particular, fourimportant motivations for bioprocess monitoring have been outlined by Locher et al.(1992):1. Data collection during a bioprocess enables visualization of changes inbioprocess parameters in order to discuss observed phenomena.2. Storage of collected data enables documentation of a series of experimentsfor later comparison.3. The collected data can be used to evaluate the accuracy of mathematicalmodels for research.4. The output from bioprocess monitoring can be used for automatic processcontrol by expert systems.Thus, there is a need for monitoring techniques which can provide precise, highquality measurements of important bioprocess parameters, and the prototype describedhere has been shown to be useful for glucose monitoring in experimental systems, even atits present stage of development. The realization of a reliable, on-line system for glucoseanalysis has enormous potential for glucose monitoring in bioprocesses and the evolutionof new fermentation control strategies. In batch culture, the prototype sensor systemcould be used in experimental fermentations to monitor glucose consumption profiles forthe purpose of research, documentation, or model analysis. In fed-batch culture, thesensor could be used to replace open-loop glucose control systems with closed, feedbackcontrol loops which indicate medium glucose exhaustion and trigger glucose infusion.Alternatively, the full potential of the glucose sensor system could be realized usingfeedback control to maintain the medium glucose concentration at a set value. The rate ofglucose infusion could be matched to the rate of glucose consumption in order to avoidfluctuation of the medium glucose concentration and maintain a constant glucose level.Thus, the prototype sensor system could be used in pilot-scale fermentations for the113investigation and optimization of the growth rate and product production ofindustrial/commercial fermentations, based on glucose control.The technology for a regenerable enzyme-based biosensor for glucose usingglucose oxidase conjugated to CBD has been developed. However, the approach is notrestricted to the glucose sensor application presented here. In principle, it should bestraightforward to conjugate other oxidase-type enzymes to CBD, such that the samesensor hardware could be used to measure different analytes in different fermentations, orat different stages of the same fermentation, depending on which enzyme-CBD conjugatesolution is perfused through the cellulose matrix. A multiple enzyme sensor for analytesrequiring a multi-enzyme system (see Table 1.1) may also be realized by incorporating twoor more different enzyme-CBD conjugates in the same perfusate.6.2 FUTURE WORKMany possibilities exist for the optimization of the present version of theprototype. The importance of accurate calibration of the sensor has been discussed withrespect to fermentation monitoring and further development of the procedure for sensorcalibration during a fermentation is essential. It would be helpful to study the effects ofthe analyte matrix on the calibration constants of the sensor. In addition, internalcalibration protocols that can be used to perform periodic calibration checks during thecourse of the fermentation, such as the scheme proposed by Bradley and Schmid (1992),remain to be investigated. In addition, further work is required toward the completion ofa sterilization protocol, such as the design of a pressurization manifold for the interior ofthe sensor. The loading and elution protocols can also be optimized to increase theefficiency of these procedures.The protocols that have been developed can be automated following acquisition ofthe necessary equipment. The sensor system could then be interfaced with a personal114computer which would monitor the sensor output and use the data in computer algorithmsfor sensor calibration, self-diagnosis, and regeneration. A computer-controlled glucoseinfusion pump could also be added to the system and control algorithms could bedeveloped for feedback control of glucose concentration during bioprocesses, asenvisioned at the outset of this project. The equipment for incorporating electrode andmembrane cleaning into the regeneration cycle also remains to be developed, although theexperimental results thus far have not indicated that these procedures would be necessaryon a frequent basis for this system.In addition, there are possibilities for enhancing the sensor performance.Increasing the amount of enzyme activity loaded for a given sensor configuration would beanalogous to increasing Imax and would result in a higher sensor current and greatersensitivity. This could be accomplished by using a concentrated GOx-CBD conjugatesample for the loading protocol or a cellulose matrix with a higher binding capacity. Othermethods to increase the sensor signal include using a larger indicating electrode areaand/or a higher electrode bias potential. It would also be possible to amplify the sensorsignal, as the signal noise observed was very low (less than 1% of the measured sensorcurrent). Any noise that is recorded is of sufficiently high frequency that it could easily befiltered with a low-pass filter.The use of continuous internal buffer flow through the enzyme chamber has beenreported by Cleland and Enfors (1984a), Bradley and Schmid (1991), and Brooks et al.(1987/88) and should be investigated for the sensor prototype. There are numerousadvantages to this technique. Dilution of the analyte in the enzyme chamber can extendthe range of concentrations that can be measured by the sensor. The operatingconcentration range can be varied to suit the sensor application by varying the dilutionrate. A loss of sensitivity may result, as some of the influx of glucose and the hydrogenperoxide produced at the enzyme are carried away. However, harmful reaction products,enzyme inhibitors, electrochemical poisons, and other interferents are also washed away,115and a constant chemical environment is maintained within the enzyme chamber. Inaddition, an oxygenated buffer could potentially be used as an oxygen source for theenzyme during fermentation monitoring in anaerobic or low dissolved oxygenenvironments.The hardware configuration of the prototype sensor may also be advantageous forthe implementation and study of electro-enzymatic sensors mediated by alternate electronacceptors, such as ferrocene derivatives (Cass et al., 1984). Substitution of electronacceptors other than oxygen for the electron transfer between glucose oxidase and theindicating electrode can enable sensor operation at a lower electrode bias potential and inanaerobic or low dissolved oxygen media. Experiments could be performed to test thisconcept using solubilized mediators in the internal buffer of the enzyme chamber.Furthermore, experimental work on the optimization and further characterizationof the glucose oxidase-CBD conjugates should continue. The feasibility and utility ofdeveloping a genetically engineered fusion protein to replace chemical conjugation shouldalso be investigated. It would be ideal if a stable, consistent source of conjugate, such as alyophilized powder, could ultimately be obtained. Finally, the application of the CBD-immobilization approach to other oxidase enzyme systems should be explored in order todevelop a range of sensor systems for different analytes which are based on the samegeneric sensor hardware and can be used effectively for the improvement of bioprocessmonitoring and control.116NOMENCLATURE SUMMARYA^Electrode areaci^Concentration of species iCo *^Bulk solution oxygen concentrationDi^Diffusion coefficient for species iDm i^Membrane diffusion coefficient for species iEa^Applied potentialStandard half-cell potential or electromotive forceF^Faraday constantI^Sensor currentIo^Sensor baseline currentImax^Maximum sensor currentJi^Mass flux of species iMass flux of species i due to diffusionKd^Dissociation constantKm^Michaelis constantKm^Apparent Michaelis constantKs, IC0^Michaelis constant for substrate and oxygenL^Enzyme membrane thicknessMWCO^Molecular weight cut-offR^Reaction rateS, P, 0^Concentration of substrate, product, and oxygenSo, Po, 0°^Bulk solution concentration of substrate, product, and oxygen117Si*(t), Pi*(t) Boundary function for concentration of substrate or product at theinterface between membrane layer i and i+1 as a function of timeIli^Absolute mobility of species iV^Enzyme reaction velocityVmax^Maximum enzyme reaction velocitydi^Width of membrane layer ii g^Glucose-dependent sensor currentk^Constant of proportionality (k = Vmax/Km)1^number of membrane layersn P^Number of ion equivalents of productr2^Correlation coefficient for linear regressiont^Timex^X-axis positionz, zi^Number of charge equivalents per mole of species iao^Equilibrium partition coefficient for oxygend(1)/dx^Potential gradientil^Solution ionic strengthvx^Solution velocity in the x-direction118REFERENCESBailey, J.E.; 011is, D.F. Biochemical Engineering Fundamentals, 2nd ed.; McGraw-Hill:New York, 1986.Bard, A.J.; Faulkner, L.R. Electrochemical Methods: Fundamentals and Applications;John Wiley & Sons: New York, 1980.Benthin, S.; Nielsen, J.; Villadsen, J. Anomeric Specificity of Glucose Uptake System inLactococcus cremoris, Escherichia coli, And Saccharomyces cervisiae: Mechanism,Kinetics, And Implications. Biotechnol. Bioeng. 1992, 40, 137-146.Bradley, J.; Anderson, P.A.; Dear, A.M.; Ashby, R.E.; Turner, A.P.F. Glucose Biosensorsfor the Study and Control of Baker's Compressed Yeast Production. In ComputerApplications in Fermentation Technology: Modelling and Control ofBiotechnological Processes; Fish, N.M., Fox, R.I., Thornhill, N.F., Eds.; ElsevierApplied Science: New York, 1988; pp 47-51.Bradley, J.; Kidd, A.J.; Anderson, P.A.; Dear, A.M.; Ashby, R.E.; Turner, A.P.F. RapidDetermination of the Glucose Content of Molasses Using a Biosensor. Analyst 1989a,114, 375-379.Bradley, J.; Turner, A.P.F.; Schmid, R.D. An In situ Fermenter Probe for Baker's YeastPropagation Monitoring. In GBF Monographs. Biosensors: Application in Medicine,Environmental Protection and Process Control; Schmid, R.D., Scheller, R., Eds.;VCH: New York, 1989b; Vol. 13, pp 85-88.Bradley, J.; Schmid, R.D. Biosensors For In situ Fermentation Monitoring of GlucoseConcentration. Proc. Biosensors 90, May 1990, Singapore, p 255.Bradley, J.; Schmid, R.D. Optimisation of a Biosensor For In situ FermentationMonitoring of Glucose Concentration. Biosensors & Bioelectronics 1991, 6, 669-674.Bradley, J.; Stticklein, W.; Schmid, R.D. Biochemistry Based Analysis Systems forBioprocess Monitoring and Control. Process Contr. Qual. 1991, 1, 157-183.119Bright, H.J.; Porter, D.J.T. Flavoprotein Oxidases. In The Enzymes, 3rd ed.; P.D. Boyer,Ed.; Academic Press: New York, 1975; Vol. 3, pp 421-506.Brillhart, K.L.; Ngo, T.T. Simple And Rapid Procedure For "Sandwich" EnzymeImmunoAssay Development Using Biotin Labeled Microwells (AvidPlate-Biotin).Anal. Lett. 1991, 24(12), 2157-2169.Brooks, S.L.; Ashby, R.E.; Turner, A.P.F.; Calder, M.R.; Clarke, D.J. Development of anOn-Line Glucose Sensor For Fermentation Monitoring. Biosensors 1987/88, 3, 45-56.Brooks, S.L.; Higgins, I.J.; Newman, J.D.; Turner A.P.F. Biosensors For Process Control.Enzyme Microbiol. Technol. 1991, 13, 946-955.Baler, H.; Ingold, W. Measuring pH and Oxygen in Fermenters. Process Biochem.1976, 11(3), 19-24.Cass, A.E.G.; Davis, G.; Francis, G.D.; Hill, H.A.O.; Aston, W.J.; Higgins, I.J.; Plotkin,E.V.; Scott, L.D.L.; Turner, A.P.F. Ferrocene-Meditated Enzyme Electrode ForAmperometric Determination of Glucose. Anal. Chem. 1984, 56, 667-671.Chaplin, M.F.; Bucke, C. Enzyme Technology; Cambridge University Press: Cambridge,1990; pp 197-219.Clark, L.C. Monitor And Control of Blood And Tissue Oxygen Tensions. Trans. Amer.Soc. Artif. Intern. Organs. 1956, 2, 41.Clark, L.C.; Lyons, C. Electrode Systems For Continuous Monitoring in CardiovascularSurgery. Ann. N.Y. Acad. Sci. 1962, 102, 29-45.Clark, L.C. Membrane Polarographic Electrode System And Method WithElectrochemical Compensation. US Patent 3 539 455, 1970.Cleland, N.; Enfors, S.O. Control of Glucose Fed-Batch Cultivations of E. colt by Meansof an Oxygen Stabilized Enzyme Electrode. Eur. J. Appl. Microbiol. Biotechnol.1983, 18, 141-147.Cleland, N.; Enfors, S.O. Externally Buffered Enzyme Electrode For Determination ofGlucose. Anal. Chem. 1984a, 56, 1880-1884.120Cleland, N.; Enfors, S.O. Monitoring Glucose Consumption in an Escherichia coliCultivation With an Enzyme Electrode. Anal. Chico. Acta. 1984b, 163, 281-285.Coutinho, J.B.; Gilkes, N.R.; Warren R.A.J.; Kilburn, D.G.; Miller, R.C. Jr. The Bindingof Cellulomonas fimi Endoglucanase C (CenC) to Cellulose and Sephadex is Mediatedby The N-Terminal Repeats. Molec. MicrobioL 1992, 6, 1243-1252.Cronenberg, C.C.H.; van den Huevel, J.C. Determination of Glucose DiffusionCoefficients in Biofilms With Micro-Electrodes. Biosensors & Bioelectronics 1991, 6,255-262.de Alwis, U; Wilson, G.S. Strategies For The Reversible Immobilization of Enzymes byUse of Biotin-Bound Anti-Enzyme Antibodies. Talanta 1989, 36, 249-253.De Baetselier, A.; Vasavada, A.; Dohet, P.; Ha-Thi, V.; De Beukelaer, M.; Erpicum, T.;De Clerck, L.; Hanotier, J.; Rosenberg, S. Fermentation of a Yeast Producing A. nigerGlucose Oxidase: Scale-Up, Purification and Characterization of The RecombinantEnzyme. Bio/Technology 1991, 9, 559-561.Demain, A.L. Cellular And Environmental Factors Affecting The Synthesis And Excretionof Metabolites. J. Appl. Chem. Biotechnol. 1972, 22, 345-362.Dixon, M.; Webb, E.C. Enzymes; Academic Press: New York, 1964.Enfors, S.O. Oxygen-Stabilized Enzyme Electrode For D-Glucose Analysis inFermentation Broths. Enzyme MicrobioL Technol. 1981, 3, 29-32.Enfors, S.0.; Molin, N. Enzyme Electrodes for Fermentation Control. Process Biochem.1978, 13, 9-11.Enfors, S.O.; Nilsson, H. Design And Response Characteristics of an Enzyme ElectrodeFor Measurement of Penicillin in Fermentation Broth. Enzyme MicrobioL Technol.1979, 1, 260-264.Filippini, C.; Sonnleitner, B.; Fiechter. A.; Bradley, J.; Schmid, R. On-Line Determinationof Glucose in Biotechnological Processes: Comparison Between FIA And an In situEnzyme Electrode. J. Biotechnol. 1991, 18,153-160.121Fortier, G.; Brassard, E.; Belanger, D. Optimization of a Polypyrrole Glucose OxidaseBiosensor. Biosensors & Bioelectronics 1990, 5, 473-490.Foulds, N.C.; Lowe, C.R. Enzyme Entrapment in Electrically Conducting Polymers. J.Chem. Soc. Faraday Trans. 1986, 82, 1259-1264.Frame, K.K.; Hu, W.-S. Kinetic Study of Hybridoma Cell Growth in Continuous Culture:II. Behavior of Producers And Comparison to Nonproducers. Biotech. Bioeng. 1991,38, 1020-1028.Frederick, K.R.; Tung, J.; Emerick, R.S.; Masiarz, F.R.; Chamberlain, S.H.; Vasavada, A.;Rosenberg, S.; Chakraborty, S.; Schopter, L.M.; Massey, V. Glucose Oxidase FromAspergillus niger: Cloning, Gene Sequence, Secretion From Saccharomycescerevisiae and Kinetic Analysis of a Yeast-Derived Enzyme. J. of Biol. Chem. 1990,265, 3793-3802.Freitag, R. Applied Biosensors. Curr. Opin. Biotech.: Anal. Biotech. 1993, 4, 75-79.Gibson, T.D.; Woodward, J.R. Protein Stabilization in Biosensor Systems. In Biosensors& Chemical Sensors: Optimizing Performance Through Polymeric Materials;Edelman, P.G., Wang, J., Eds.; ACS: Washington, DC, 1992; pp 40-55.Gilkes, N.R.; Henrissat, B.; Kilburn, D.G.; Miller, R.C. Jr.; Warren, R.A.J. Domains inMicrobial 13-1,4-Glycanases: Sequence Conservation, Function, And Enzyme Families.Microbiol. Rev. 1991, 55, 303-315.Gilkes, N.R.; Jervis, E.; Henrissat, B.; Tekant, B.; Miller, R.C. Jr.; Warren, R.A.J.;Kilburn D.G. The Adsorption of a Bacterial Cellulase And Its Two Isolated Domains toCrystalline Cellulose. J. Biol. Chem. 1992, 267(10), 6743-6749.Guilbault, G.G.; Lubrano, G.L. An Enzyme Electrode For The AmperometricDetermination of Glucose. Anal. Chim. Acta 1973, 64, 439-455.Guilbault, G.G.; Luong, J.H. Biosensors: Current Status And Future Possibilities. Sel.Elec. Rev. 1989, 11, 3-16.122Guilbault, G.G.; Montalvo, J. A Urea-Specific Enzyme Electrode. J. Am. Chem. Soc., 91,1969, 2164.Hall, E.A.H. Biosensors; Prentice Hall: Englewood Cliffs, New Jersey, 1991; pp 216-218.Hansen, E.H.; Mikkelsen, H.S. Enzyme-Immobilization by The GlutardialdehydeProcedure. An Investigation of The Effects of Reducing The Schiff-Bases Generated,as Based on Studying The Immobilization of Glucose Oxidase to Silanized ControlledPore Glass. Anal. Lett. 1991, 24(8), 1419-1430.Hardjito, L.; Greenfield, P.F.; Lee, P.L. Recombinant Protein Production Via Fed-BatchCulture of The Yeast Saccharomyces cerevisiae. Enzyme Microb. Technol. 1993, 15,120-126.Harlow, E.; Lane, D. Antibodies: A Laboratory Reference Manual; Cold Spring HarborLaboratory: Cold Spring Harbor, New York, 1988; Ch. 12.Harrison, D.J.; Turner, R.F.B.; Baltes, H.P. Characterization of Perfluorosulfonic AcidPolymer Coated Enzyme Electrodes And a Miniaturized Integrated Potentiostat ForGlucose Analysis in Whole Blood. Anal. Chem. 1988, 60, 2002-2007.Hendry, S.P.; Higgins, I.J.; Bannister, J.V. Amperometric Biosensors. J. Biotech. 1990,15, 229-238.Hecht, H.J.; Schomburg, D.; Kalisz, H.; Schmid, R.D. The 3D Structure of GlucoseOxidase From Asperigillus niger. Implications For The Use of GOD as a BiosensorEnzyme. Biosensors & Bioelectronics 1993, 8, 197-203.Hoist, 0.; HAkanson, H.; Miyabayashi, A.; Mattiasson, B. Monitoring of Glucose inFermentation Processes Using a Commercial Glucose Analyser. Appl. Microbiol.Biotechnol. 1988, 28, 32-36.Huang, Y.L.; Li, S.Y.; Dremel, B.A.A.; Bilitewski, U.; Schmid, R.D. On-LineDetermination of Glucose Concentration Throughout Animal Cell Cultures Based onChemiluminescent Detection of Hydrogen Peroxide Coupled With Flow-InjectionAnalysis. J. Biotechnol. 1991, 18, 161-172.123Ingold, W. CO2 Measuring System Product Brochure, Ingold Electrodes Inc.,Wilmington, MA, c. 1990.Janata, J. Principles of Chemical Sensors; Plenum Press: New York, 1989; pp 175-195.Jochum, P.; Kowalski, B.R. A Coupled Two-Compartment Model For ImmobilizedEnzyme Electrodes. Anal. Chico. Acta. 1982, 144, 25-38.Kilburn, D.G.; Turner, R.F.B.; Coutinho, J.B.; Din, N.; Gilkes, N.R.; Greenwood, J.M.;Hobbs, J.B.; Miller, R.C. Jr.; Ong, E.; Phelps, M.R.; Ramirez, C.; Warren, R.A.J.Cellulose Binding Domains: Applications in Biotechnology. Proc. Cellucon 92, InPress.Kittsteiner-Eberle, R.; Ogbomo, I.; Schmidt, H.-L. Biosensing Devices For The Semi-Automated Control of Dehydrogenase Substrates in Fermentations. Biosensors 1989,4, 75-85.Kobos, R.K. Potentiometric Enzyme Methods. In ISE's in Analytical Chemistry; Freiser,H., Ed.; Plenum: New York, 1980; Vol. 2, p. 1 .Kok, R.; Hogan, P. The Development of an In situ Fermentation Electrode Calibrator.Biosensors 1987/88, 3, 89-100.Kole, M.M.; Ward, D.; Gerson, D.F. Simultaneous Control of Ammonium And GlucoseConcentrations in Escherichia coli Fermentations. J. Ferment. Technol. 1986, 64(3),233-238.Leypoldt, J.K.; Gough, D.A. Model of a Two-substrate Enzyme Electrode For Glucose.Anal. Chem. 1984, 56, 2896-2904.Linek, V.; Beneg, P.; Sinkule, J.; Holecek, 0.; Maly, V. Oxidation of D-Glucose in ThePresence of Glucose Oxidase And Catalase. Biotech. Bioeng. 1980, 22, 2515-2527.Locher, G.; Sonnleitner, B.; Fiechter, A. On-Line Measurement in Biotechnology:Exploitation, Objectives and Benefits. J. Biotech. 1992, 25, 55-73.124Liidi, H.; Garn, M.B.; Haemmerli, S.D.; Manz, A.; Widmer, H.M. Flow Injection AnalysisAnd In-Line Biosensors For Bioprocess Control: A Comparison. J. Biotech. 1992, 25,75-80.Mandenius, C.F.; Danielsson, B.; Mattiasson, B. Evaluation of a Dialysis Probe ForContinuous Sampling in Fermentors And in Complex Media. Anal. Chico. Acta 1984,163, 135-141.Maresse, C.A.; Miyawaki, 0.; Wingard, L.B., Jr. Simultaneous ElectrochemicalDetermination of Diffusion And Partition Coefficients of Potassium Ferrocyanide ForAlbumin-Glutaraldehyde Membranes. Anal. Chem. 1987, 59, 248-252.Mascini, M.; Palleschi, G. Design And Applications of Enzyme Electrode Probes. Sel.Elec. Rev. 1989, I I , 191-264.Merten,^O.-W.;^Palfi,^G.E.;^Steiner,^J.^On-Line^Determination^ofBiochemical/Physiological Parameters in The Fermentation of Animal Cells in aContinuous or Discontinuous Mode. In Advances in Biotechnological Processes;Mizrahi, A. Ed.; Alan R. Liss: New York, 1986; Vol. 6, pp 125-152.Miyabayashi, A.; Reslow, M.; Adlercreutz, P.; Mattiasson, B. A Potentiometric EnzymeElectrode For Monitoring in Organic Solvents. Anal. Chin,. Acta. 1989, 219, 27-36.Moody, G.J.; Thomas, J.D.R Amperometric Biosensors: A Brief Appraisal of Principlesand Applications. Sel. Elec. Rev. 1991, 13, 113-124.Moore, R.B.; Martin, C.R. Procedure For Preparing Solution-Cast PerfluorosulfonateIonomer Films And Membranes. Anal. Chem. 1986, 58, 2569-2570.Nakamura, S.; Hayashi, S.; ;Koga, K. Effect of Periodate on The Structure AndProperties of Glucose Oxidase. Biochein. Biophys. Acta 1976, 445, 294-308.O'Neill, G.; Goh, S.H.; Warren, R.A.J.; Kilburn, D.G.; Miller, R.C. Jr. Structure of TheGene Encoding The Exoglucanase of Cellunionas fimi. Gene 1986, 44, 325-330.Ogbomo, I.; Prinzing, U.; Schmidt, H.L. Prerequisites For The Control of MicrobialProcesses by Flow Injection Analysis. J. Biotechnol. 1990, 14(1), 63-70.125Ong, E.; Gilkes, N.R.; Miller, R.C. Jr.; Warren, R.A.J.; Kilburn, D.G. EnzymeImmobilization Using a Cellulose-Binding Domain: Properties of a j3-GlucosidaseFusion Protein. Enzyme Microb. Technol. 1991, 13, 59-65.Ong, E.; Gilkes, N.R.; Miller, R.C. Jr.; Warren, R.A.J.; Kilburn, D.G. The Cellulose-Binding Domain (CBDCex) of an Exoglucanase From Cellulomonas fimi: Productionin Escherichia coli And Characterization of the Polypeptide. Biotech. Bioeng. 1993,42, 401-409.Ong, E.; Gilkes, N.R.; Warren, R.A.J.; Miller, R.C. Jr.; Kilburn, D.G. EnzymeImmobilization Using The Cellulose-Binding Domain of a Cellulomonas fimiExoglucanase. Bio/Technology 1989, 7, 604-607.Patkar, A.; Seo, J.-H. Fermentation Kinetics of Recombinant Yeast in Batch And Fed-Batch Cultures. Biotech. Bioeng. 1992, 40, 103-109.Pieters, B.R.; Bardeletti, G. Enzyme Immobilization on a Low-Cost Magnetic Support:Kinetic Studies on Immobilized And Coimmobilized Glucose Oxidase AndGlucoamylase. Enzyme Microb. Technol. 1992, 14, 361-370.Reach, G.; Wilson, G.S. Can Continuous Glucose Monitoring be Used For The Treatmentof Diabetes? Anal. Chem. 1992, 64, 381-386.Renneberg, R.; Trott-Kriegeskorte, G.; Lietz, M.; Jager, V.; Pawlowa, M.; Kaiser, G.;Wollenberger, U.; Schubert, F.; Wagner, R.; Schmid, R.D.; Scheller, F.W. EnzymeSensor-FIA-System For On-Line Monitoring of Glucose, Lactate And Glutamine inAnimal Cell Cultures. J. Biotech. 1991, 21, 173-186.Rishpon, J.; Shabtai, Y.; Rosen, I.; Zibenberg, Y.; Tor, R.; Freeman, A. In situ GlucoseMonitoring in Fermentation Broth by "Sandwiched" Glucose-Oxidase Electrode(SGE). Biotech. Bioeng., 1990, 35, 103-107.Rogers, M.J.; Brandt, K.G. Multiple Inhibition Analysis of Aspergillus niger GlucoseOxidase by D-Glucal And Halide Ions. Biochem. 1971, 10, 4636-4641.126Romette, J.-L.; Froment, B.; Thomas, D. Glucose-Oxidase Electrode. Measurements ofGlucose In Samples Exhibiting High Variability In Oxygen Content. Clin. Chico. Acta1979, 95, 249-253.Romette, J.-L. Mammalian Cell Culture Process Control: Sampling And Sensing. BGFMonographs 1987, 10, 81-86.Sansen, W.; Suls, J.; Lambrechts, M.; Claes, A.; Jacobs, P.; Kuypers, M.H. AnaerobicOperation of a Glucose Sensor by Use of Pulse Techniques. Biosensors 1992, 156-162.Sawyer, D.T.; Roberts, J.L. Jr. Experimental Electrochemistry for Chemists; John Wileyand Sons: New York, 1974; pp 39-41.Schulmeister, T. Mathematical Treatment of Concentration Profiles And Anodic Currentof Amperometric Multilayer Enzyme Electrodes. Anal. Chico. Acta 1987, 198, 223-229.Schulmeister, T. Mathematical Modelling of the Dynamic Behaviour of AmperometricEnzyme Electrodes. Sel. Elec. Rev. 1990, 12, 203-260.Schultz, J.S. Sophisticated Descendants of The Canary in The Coal Mine Are Based onMolecular Components of Plants And Animals Bound to Microscopic Electrodes orOptical Fibers. Scientific American 1991, 264, 64-69.Sigma catalogue, 1993, Sigma Chemical Co., St. Louis, MO, U.S.A., pp 2104-2106.Sittampalam, G.; Wilson, G.S. Surface-Modified Electrochemical Detector For LiquidChromatography. Anal. Chem. 1983, 55, 1608-1610.Smith, M.; Bajpai, R. Fed-Batch Control of Escherichia coli Fermentation to High CellDensity. Proc. 15th Ann. Biochem. Eng. Symp. 1985, 15, 34-44.Stamm, W.W.; Seidler, W.; Kahlert, W.; Kappel, W. On-Line Monitoring of Glucose inBatch Fermentations of Escherichia coli With a Commercially Available BiosensorAnalyzer - Fermentation Control System. Proc. 2nd World Cong. Biosensors, May1992, Geneva, p 506.127Stoecker, P.W.; Yacynych, A.M. Chemically Modified Electrodes as Biosensors. Sel.Elec. Rev. 1990, 12, 137-160.Tran-Minh, C.; Broun, G. Construction And Study of Electrodes Using Cross-LinkedEnzymes. Anal. Chem. 1975, 47, 1359-1364.Turner, A.P.F., Karube, I., Wilson, G.S. Eds.; Biosensors: Fundamentals andApplications. Oxford University Press: Oxford, 1987.Updike, S.J.; Hicks, G.P. The Enzyme Electrode. Nature 1967, 214, 986-988.Valero, F; Lafuente, J.; Poch, M.; Sola, C.; Araujo, A.N.; Lima, J.L.F.C. On-LineFermentation Monitoring Using Flow Injection Analysis. Biotech. Bioeng. 1990, 36,647-651.Wang, J.; Hutchins, L.D. Thin-Layer Electrochemical Detector With a Glassy CarbonElectrode Coated With a Base-Hydrolyzed Cellulosic Film. Anal. Chem. 1985, 57,1536-1541.Wang, J. Permselective Coatings For Amperometric Biosensing. In Biosensors andChemical Sensors: Optimizing Performance Through Polymeric Materials; Eddman,P.G., Wang, J., Eds.; ACS: New Mexico State University, 1992; Symposium Series487, pp 125-132.Whittle, D.J.; Kilburn, D.G.; Warren, R.A.J.; Miller, R.C. Jr. Molecular Cloning of aCellumonas fimi Cellulase Gene in Escherichia coli. Gene 1982, 17, 139-145.Wilson, R.; Turner, A.P.F. Glucose Oxidase: An Ideal Enzyme. Biosensors &Bioelectronics 1992, 7, 165-185.Wirth, M.; Li, S.Y.; Schumacher, L.; Lehmann, J. Screening For And Fermentation ofHigh Producer Cell Clones From Recombinant BHK Cells. In Advances in AnimalCell Biology and Technology for Bioprocesses; Spier, R.E., Griffiths, J.B., Stephenne,J., Gooy, P.J., Eds.; Butterworth: Sevenoaks, Kent, England, 1989; pp 44-51.Yamasaki, Y. The Development of a Needle-Type Glucose Sensor For Wearable ArtificialEndocrine Pancreas. Medical Journal of Osaka University 1984, 35, 25-34.128Ylilammi, M.; Lehtinen, L. Numerical Analysis of a Theoretical One-DimensionalAmperometric Enzyme Sensor. Med. & Biol. Eng. & Comput. 1988, 26, 81-87.129APPENDIX SAMPLE STRIP CHART RECORDRecorded during first calibration of the biosensor prototype for glucose monitoringin a fermentation of E. coilConditions: Prototype with Nafion/cellulose acetate membrane systemFirst enzyme loadingFresh minimal medium (M-9)No cells130


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items