UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Toward an in-situ glucose biosensor for animal cell culture monitoring and control Fong, Fenton 1998

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

Item Metadata

Download

Media
831-ubc_1998-0437.pdf [ 6.81MB ]
Metadata
JSON: 831-1.0088469.json
JSON-LD: 831-1.0088469-ld.json
RDF/XML (Pretty): 831-1.0088469-rdf.xml
RDF/JSON: 831-1.0088469-rdf.json
Turtle: 831-1.0088469-turtle.txt
N-Triples: 831-1.0088469-rdf-ntriples.txt
Original Record: 831-1.0088469-source.json
Full Text
831-1.0088469-fulltext.txt
Citation
831-1.0088469.ris

Full Text

TOWARD AN IN-SITU GLUCOSE BIOSENSOR FOR ANIMAL C E L L CULTURE MONITORING AND CONTROL by FENTON FONG B.Sc, The University of Alberta, 1988  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF T H E REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF MICROBIOLOGY AND IMMUNOLOGY We accept this thesis as conforming to the required standard  T H E UNIVERSITY OF BRITISH COLUMBIA May 1998 © Fenton Fong, 1998  In presenting this thesis in partial fulfilment of the  requirements for an advanced  degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department  or by his  or her  representatives.  It  is  understood  that  copying or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department of  ffliirJtfvttfij  f  The University of British Columbia Vancouver, Canada  Da,a  DE-6  (2/88)  L  P,  /9  V  )n*M«dtf^,  ii  Abstract This thesis describes the development and characterization of an in-situ sterilizable, regenerable biosensor for monitoring and control of glucose concentration in bioreactorscale animal cell culture. Long-term stability and in-situ steam sterilization capability are obstacles preventing development of industrially acceptable biosensors. These difficulties are magnified in animal cell culture applications given their longer duration and high susceptibility to microbial contamination. A novel enzyme immobilization technology is presented which helps address these difficulties. The developmental starting point utilized an existing prototype originally designed for microbial culture applications. The prototype contains a cellulose matrix sandwiched between a platinum electrode and a dialysis membrane. Glucose oxidase enzyme, chemically conjugated with cellulose binding domain protein, is easily immobilized to and removed from the cellulose. A n improved conjugation protocol yielding conjugate with higher specific activity is described. Conjugate was characterized for binding characteristics to cellulose and lower detection limit during prototype use. Binding experiments were the first for characterizing a heterogeneous chemical fusion of these proteins. Re-design of the prototype membrane assembly and incorporation of continuous perfusion are described and were necessary modifications to produce a functional prototype for animal cell culture. Longer term sensor signal stability was characterized and enhanced. A first-order equation adequately described signal decay behavior during the first long-term experiment. The combination of hardware modifications resulted in ca. 2.5-fold increase in signal stability during controlled environment conditions. A subsequent cell culture experiment  Ill  revealed enhanced signal stability; although, undefined factors resulted in inconsistent decay patterns. A Chinese hamster ovary cell line was transfected with pNUT plasmid expressing H6E2FX protein to develop a model cell line for culture and glucose studies utilizing the prototype.  Positive subclones, after adaptation to serum-free medium over 7 weeks,  displayed undetectable levels of H6E2FX. Several possible explanations could account for unstable expression but no specific one was determined. Described hardware modifications yielded a prototype functional for longer periods of time and with enhanced signal stability. Stability experiments are the first for characterizing longer term prototype signal behavior. Results in this thesis contribute to existing data demonstrating the feasibility and potential of this biosensor technology for implementation in industrial bioprocesses.  Table of Contents Abstract Table of Contents List of Tables List of Figures Acknowledgment  ii iv vi vii ix  CHAPTER 1 LITERATURE REVIEW AND P R O P O S E D THESIS W O R K  1.1 Introduction 1 1.2 Literature Review 1 1.2.1 Biosensor Development 1 1.2.2 Glucose as a Carbon Source in Animal Cell Culture 5 1.2.3 Large-Scale Animal Cell Culture 8 1.2.4 Optimization of Large-Scale Animal Cell Culture and Nutrient Control 10 1.2.5 Benefits of Controlled Glucose Concentration 11 1.2.6 Current State-of-the-Art for Glucose Control in Animal Cell Culture 13 1.2.7 Challenges for Developing In-Situ Enzyme Sensors for Animal Cell Culture... 14 1.2.8 In-Situ Enzyme Sensors 16 1.2.9 Usefulness of Continuous Measurement and Future Potential of the Prototype 19 1.3 Features of the Prototype 21 1.3.1 Glucose Oxidase 21 1.3.2 Cellulose Binding Domain Technology 24 1.3.3 Operational Features 25 1.3.4 Prototype Performance in Microbial Culture 28 1.4 Thesis Overview and Thesis Objectives 29 CHAPTER 2 H A R D W A R E MODIFICATIONS A N D P R O T O T Y P E C H A R A C T E R I Z A T I O N  2.1 Introduction 2.1.1 Hardware Components and Principle of Operation 2.1.2 Motivation for Hardware Modifications 2.2 Materials and Methods 2.2.1 Membrane Cartridge Assembly 2.2.2 Continuous Perfusion 2.2.3 Conjugation of CBD and Glucose Oxidase 2.2.3.1 Total Protein and Enzyme Activity 2.2.3.2 Protein Gels and Western Blots 2.2.4 Lower Limit of Detection 2.2.5 Binding Capacity of Cellulose Matrix 2.2.5.1 C B D Binding to Cellulose 2.2.5.2 Conjugate Binding to Cellulose 2.2.5.3 Desorption over Time  32 32 40 47 47 52 52 54 55 56 56 56 57 59  2.3 Results and Discussion 2.3.1 Membrane Cartridge 2.3.2 Continuous Perfusion 2.3.3 Improved Conjugate Using a Hetero-Bifunctional Crosslinker 2.3.3.1 Comparison of Resultant Specific Activities 2.3.3.2 Gel Electrophoresis 2.3.4 Lower Limit of Detection 2.3.5 Binding Capacity of Cellulose Matrix 2.3.5.1 Desorption over Time 2.3.5.2 Saturation of Cellulose  59 59 60 61 61 66 69 71 72 74  CHAPTER 3 D E V E L O P M E N T O FA M O D E L C H O C E L L LINE  3.1 Introduction 3.2 Materials and Methods 3.2.1 Transfection 3.2.2 Gel Electrophoresis 3.2.3 Western Blots 3.2.4 Limiting Dilution, Dot Blot ELISA, Specific Protein Production Rate 3.2.5 Adaptation to Serum-free Medium 3.3 Results and Discussion CHAPTER 4 LONG-TERM  78 79 79 80 81 82 84 85  STABILITY  4.1 Introduction 93 4.2 Materials and Methods 93 4.2.1 Cell Line and Cultivation Conditions 93 4.2.2 Glucose Sensor Calibration and Data Analysis 95 4.2.3 Stability of Conjugate in Solution 95 4.3 Results and Discussion 96 4.3.1 First Long-Term Experiment with 2E11 Hybridoma Culture 96 4.3.2 Second Long-Term Experiment: Sensor Performance in a Controlled Glucose Environment 113 4.3.3 Third Long-Term Experiment with P9 Hybridoma Culture 117 CHAPTER 5 CONCLUSIONS  5.1 Concluding Remarks 5.2 Future Work  123 125  REFERENCES  130  vi  LIST OF TABLES Table 1.1  Summary of industrially important enzymatic assays requiring oxidase enzymes  28  Table 2.1  Summary and comparison of resultant specific activities from conjugations using glutaraldehyde and S M C C crosslinker 63  Table 2.2  Binding constants and No for CBDce and S M C C conjugate protein on 2 different cellulose types based on mass quantities 71  Table 4.1  Comparison of derived kinetic parameters using Michaelis-Menten and Hanes model analysis  X  99  Vll  LIST OF FIGURES Figure 1.1  Schematic representation of the Cellulose Binding Domain (CBD) protein. 24  Figure 2.1  Reprinted figure of Ingold C 0 2 probe  Figure 2.2  Cross-sectional schematic diagram of prototype sensor tip illustrating the configuration of the dialysis membrane, porous cellulose matrix, platinum electrode and inner perfusable space 37  Figure 2.3  Schematic diagram of the 3-electrode system  Figure 2.4  Schematic diagram illustrating the reagent flow system and instrumentations8  Figure 2.5  Structure of sulfosuccinimidyl 4-(N-maleimidomethyl) cyclohexane-1 -  36  37  carboxylate (SMCC) crosslinker  44  Figure 2.6  Original membrane cartridge configuration  50  Figure 2.7  Redesigned membrane cartridge configuration  51  Figure 2.8  SDS-PAGE and western blot of conjugate protein  68  Figure 2.9  Comparison of the lower limit of detection (LLD) in D M E M using both glutaraldehyde and SMCC conjugate protein 70 Figure 2.10 Binding Isotherm of C B D protein on Avicel and Whatman cellulose based on mass quantities 76 Figure 2.11 Binding Isotherm of S M C C conjugate protein on Avicel and Whatman cellulose based on mass quantities  77  Figure 3.1  Schematic diagram of pNUT vector harboring H6E2FX c D N A  78  Figure 3.2  SDS-PAGE of first 7 surviving colonies from transfection  86  Figure 3.3  Western blot of first 7 surviving colonies from transfection  86  Figure 3.4  Western blot of remaining 16 colonies showing two positives clones  87  Figure 4.1  Michaelis-Menten kinetics for three glucose sensor calibrations in the bioreactor during the first long-term experiment  107  Figure 4.2  Hanes plots for three glucose sensor calibrations in the bioreactor during the first long-term experiment 108  Vlll  Figure 4.3  Transformed sensor output compared against off-line values obtained from Beckman analyzer 109  Figure 4.4  Extraction of first-order decay constant for two periods of time between calibrations  110  Figure 4.5  Transformed sensor output corrected for decay with the extracted T value 111  Figure 4.6  Stability of G O X and conjugate in solution  Figure 4.7  Comparison of sensor signal decay rates in second long-term experiment and the first glutaraldehyde conjugate derived decay curves in Figure 4.4 116  Figure 4.8  Sensor output for first 24 hours of culture  120  Figure 4.9  Sensor output for 48-64 hours  121  Figure 4.10 Sensor output for 190-210 hours  112  122  ix  Acknowledgment Several people who have contributed to my work described in this thesis deserve gracious acknowledgment. I would first like to thank my immediate supervisors, Dr. Robin F.B. Turner and Dr. Douglas G. Kilburn for their continuous technical guidance and encouragement at times when it was needed. M y other committee members, Drs. Jamie Piret and R.A.J. Warren deserve acknowledgment for their support and guidance during committee meeting discussions relating to my work and thesis. Several members in the bioprocess monitoring group and cellulase group were helpful throughout the way. In particular: Kim Blouin, Jason Dowd and Jiirgen Koska for their mathematical and computer expertise as well as perspectives from their engineering backgrounds; Marta Guarna provided guidance with tissue culture work; Emily Kwan and Chris Sherwood assisted with western blots and protein gels. The "boys" in the pilot plant facility provided assistance during the fermenter experiments. As well, I would like to acknowledge Eric Jervis for his regular inputs and ideas, some of which led to new findings. A n acknowledgment should go out to all other members on "the third floor" and the N C E building past and present who are too numerous to mention. They all contributed to me in small but immeasurable ways both inside and outside the laboratory and made my time spent here very enjoyable . M y parents and family must be thanked for their support and faith as none of this would have been possible without them. It is to them that this thesis is dedicated. I thank Genentech and NSERC for their financial support which allowed me the opportunity to study. And finally I would like to give a special thanks to Carol for her unwavering support, enthusiasm and inspiration, all of which contributed greatly to my well being and helped me during the course of my thesis work. M y wish is to always continue reciprocating those same qualities back to you.  "Ancora imparo" (Still I am learning) -Michelangelo  1  CHAPTER 1 Literature Review and Proposed Thesis Work 1.1 Introduction This chapter is divided into 3 sections.  The first section provides a literature  review covering highlights of biosensor research from its beginnings over 40 years ago, with emphasis on glucose biosensors. The review also encompasses animal cell culture biotechnology as it relates to biosensor work in this thesis. In the second section, the operational features and advantages of the proposed prototype glucose sensor are presented.  The last section provides a brief overview of this thesis and a statement of  research objectives.  1.2 Literature Review 1.2.1 Biosensor  Development  A biosensor has been defined as a device which can detect or respond to environmental chemicals or analytes through specific interaction with a biological sensing component that is in close contact with a physical transducer (Wollenberger et al., 1993). Analyte recognition is usually performed by immobilized biocomponents such as whole cells, antibodies, protein receptors, organelles, enzymes or organic chemical ligands. The biosensors described in this review concentrate mainly on configurations employing enzymes but the reader should be aware that enzyme sensors represent only a fraction of the total biosensor field.  Biosensor research and development is often considered to have started with the invention of the Clark oxygen electrode (Clark, 1956) for polarographically measuring dissolved oxygen concentration in blood during surgery. Six years later Clark and Lyons (1962) utilized the same oxygen electrode, coupling it with glucose oxidase enzyme to create the first glucose biosensor and the first biosensor to employ a true biological component. In their configuration, the enzyme was immobilized in a gel on the surface of the oxygen electrode and the drop in localized dissolved oxygen due to the enzymatic action was related to the glucose concentration. The term "enzyme electrode" was coined in 1967 by Updike and Hikes when they reported an improvement upon the Clark & Lyons system from 5 years earlier by incorporating a second oxygen electrode with no enzyme in order to correct for normal variations in the background oxygen concentration (Updike & Hicks, 1967). The next major development was Clark's patent for the idea to use a platinum electrode transducer to amperometrically detect hydrogen peroxide generated from enzymatic oxidation of glucose in a sensor (Clark, 1970). The majority of enzyme-based glucose biosensor configurations (including the one in this thesis) utilize this principle of amperometric detection.  Furthermore, Clarks' patent became the basis for the first  commercially available laboratory glucose analyzer marketed by Yellow Springs Instrument Company (Yellow Springs, Ohio, USA). Amperometric biosensors providing off-line discontinuous sample analysis has been the most successful branch of biosensor research (Wollenger et al, 1993).  3 Urea was the next physiologically relevant analyte to be targeted for measurement using an enzyme electrode. Guilbault designed such an electrode using the enzyme urease (Guilbault & Montalvo, 1969). Ammonium ions generated from the enzyme-catalyzed reaction were detected potentiometrically using an ion selective electrode. In subsequent years, many enzymes specific for other industrial and medically important analytes were incorporated into enzyme electrode configurations.  However, a number of on-going  problems and challenges common to biosensors had become apparent by this time. These challenges prevented, and in many cases, continue to prevent commercial realization of the majority of proposed configurations. A large majority of proposed sensors suffered from problems relating to interference from other elements in the measured matrix, adequate signal sensitivity, calibration drift, long term signal stability, changes in response times, and hardware fouling. Investigators have attempted to address these issues in a number of ways over the years and with varying degrees of success. Examples of other analytes targeted by enzyme electrode biosensors include: 1) ethanol, using alcohol dehydrogenase. These authors report a configuration in which 2 identical working electrodes were incorporated with one used to correct for interfering compounds such as ascorbic acid.  They also report some interference from higher  alcohols which maybe present in fermentation's with bakers yeast (Kitagawa et al, 1991); 2) penicillin, using penicillinase immobilized at the tip of a glass pH electrode.  The  proposed sensor had a wide working range of concentrations but was also sensitive to interfering compounds and pH fluctuations (Meier et al., 1991); 3) glutamate, using glutamate oxidase and glutaminase co-immobilized onto aminopropyl glass beads.  The  4 configuration provided fast 2 minute response times and was reportedly stable for over 200 hours (Cattaneo et al,  1992); 4) lactate, using lactate oxidase and detection of  produced peroxide by chemiluminescence.  However, the standard curve was highly  variable and the signal stability was poor (Benthin et al, 1992). These examples represent a small fraction of the types of configurations that have been developed, the range of analytes that have been quantified and the types of problems encountered. The wide array of enzymes used and differences in catalytic mechanisms resulted in development of many different biosensor configurations.  This development required  expertise in a number of diverse disciplines, stimulating research in areas such as enzyme immobilization technologies (Bowers, 1986; Mosbach, 1976), membrane materials science (Wang, 1992; Byfield & Abuknesha, 1994), and transduction methods (Sethi, 1994; Griffiths & Hall, 1993). From these beginnings, progress and interest in biosensors has and continues to grow due to their simplicity, selectivity and specificity. The initiation of an international journal dedicated to the field {Biosensors and Bioelectronics, Elsevier Science Publishers Ltd.) in the early 1980's, reflects the rapid growth of the industry.  A number of  comprehensive books and symposiums (Turner, 1987; Hall, 1990; Cass, 1990; Twork & Yacynych, 1990; Eddman & Wang, 1992) provide wide coverage of this increasingly diversified field. Enzyme electrode biosensors have become useful and applicable in a number of areas other than clinical monitoring. The food and beverage manufacturing industry, environmental monitoring and agriculture, security and defense offer themselves as  5  potential markets. In more recent times, rapid development of the biotechnology industry has created a demand for new and better on-line sensors. The work in this thesis will concentrate on biosensors for fermentation monitoring and industrial bioprocess control. A number of reviews in the scientific literature provide summaries of the latest developments in biosensors relevant to biotechnology and bioprocess monitoring (Bradley etal, 1991;Mulchandani, 1995; Scheper etal, 1996). Glucose is one of the most important target analytes for detection with biosensors. The motivation for this arises from three reasons: 1) the paramount importance of glucose in physiological systems as a main carbon source 2) necessity for active management of its concentration in diabetic patients 3) glucose oxidase is an ideal enzyme for utilization for biosensor development due to its high specificity, high relative stability and low cost (Wilson & Turner, 1992). In industrial fermentation's, glucose is usually the growthlimiting substrate and as such requires accurate quantification (Fillipini et al., 1991). The vast majority of glucose biosensor configurations reported in the literature utilize glucose oxidase for selectivity. This enzyme is also the most widely employed enzyme as an analytical reagent (Raba & Mottola, 1995). 1.2.2 Glucose  as a Carbon  Source  in Animal  Cell  Culture  Mammalian cells utilize glucose as a primary carbon source for energy production and can produce lactate as a metabolic by-product (Miller et al., 1988; Sugiura, 1992; Kurokawa et al., 1993). The primary mode of uptake for most mammalian cells is facilitated diffusion through the plasma membrane via a family of glucose transporter proteins known as GLUT 1-5 and 7. The transporter exists in 6 isoforms each with  6 different kinetic properties with a K„, o f 1 to 2 raM (Wheeler, 1985; Renner et al,  1972).  The distribution o f the different isoforms is cell specific and the sole driving force for entry into cells is the glucose concentration gradient across the cell membrane. also be taken up by a high affinity ( K  m  less than 0.2 m M ) saturable N a  +  Glucose can  symport process  driven by the sodium gradient (Wright, 1993). The net flux o f glucose across a plasma membrane is influenced not only by the cell-specific transporter proteins but also by the actions o f several hormones and growth factors on the expression o f G L U T transporters. F o r instance, expression of G L U T 1 and 4 gene products is stimulated by insulin resulting in an increase in glucose uptake in muscle and fat cells (Klip et al., 1994). Glucose is oxidized through a series o f biochemical pathways intermediate by-products required for various cellular processes.  to generate  In addition, glucose  oxidation is coupled with A T P production, the universal currency o f free energy in living systems (Stryer, 1985).  These pathways are covered in greater detail in any standard  biochemistry textbook (Stryer; Lehninger, 1992) but a brief overview is useful here. U p o n entering a cell, glucose is oxidized to varying extents depending primarily on its concentration, the concentration o f other participants in the metabolic pathway and finally, the capacity o f the cell to utilize it completely. The first major metabolic pathway, glycolysis, takes place in the cytosol o f many cells. As with other metabolic pathways, a series o f intermediates are formed, some of which are channeled as precursors into other biosynthetic pathways. The intermediate steps in glycolysis include isomerization, transfer o f small groups such as a hydrogen or phosphate followed by ring splitting to yield 2 pyruvate molecules for each glucose molecule (Bailey & Ollis, 1986).  This process is  7  coupled with the formation of 2 ATP molecules.  The subsequent fate of pyruvate is  dependent upon its concentration at the mitochondrial membrane which in turn depends, in part, on the initial glucose concentration. High concentrations of pyruvate overload the mitochondrial capacity for further oxidation and pyruvate is then diverted to form lactate which in turn is excreted by the cell. Lactate is inhibitory to growth of animal cells partly because excretion leads to culture acidification and increased osmolarity. Lactate ion itself inhibits growth and antibody formation in hybridoma cells (Glacken et al., 1988; Ozturk et al., 1992). In addition, lactate formation markedly reduces the stoichiometric coefficient of ATP production since less pyruvate per glucose molecule enters the Krebs cycle. High concentrations of glucose thus lead to overflow metabolism and accumulation of inhibitory byproducts (Ljunggren and Haggstrom, 1994). During aerobic conditions in the mitochondrion, pyruvate which enters the next pathway, the Krebs cycle, becomes completely oxidized to C 0 and H 0 with the eventual 2  2  formation of 36 ATP for every glucose molecule. Complete oxidation of glucose under aerobic conditions leads to its most efficient usage as an energy source. The energetics of glucose oxidation under anaerobic conditions will not be treated here as this condition is generally avoided in cultures of recombinant animal cells. Glucose concentration also has profound effects on other aspects of cell culture and will be discussed in later sections of this chapter. Glutamine is also important as a major carbon source in cell culture and high concentrations result in formation of ammonia which is inhibitory to culture productivity at high concentrations because it affects compartmental pH.  Glutamine metabolism  g  proceeds through different pathways than glucose, but they converge with the formation of pyruvate. Decreasing both glucose and glutamine levels provokes the use of more efficient energy pathways and reduces the formation of inhibitory by-products (Ljunggren & Haggstrom, 1994). 1.2.3 Large-Scale  Animal  Cell  Culture  During the mid 1950's, impetus from medical research lead to the development of in-vitro culturing techniques for tissue from different warm blooded animals.  These  animals were chosen because their normal and pathological development are most similar to human. Tissue types cultured include chick embryo tissues because they could provide a diversity of cell types in primary culture and rodent tissues because they often lead to the production of continuous cell lines (Freshney, 1994). Such cell lines were transformed by external factors to become immortalized and able to propagate indefinitely. Human tumors also gave rise to continuous cell lines and resulted in the establishment of various cell lines such as the widely used HeLa cells (Gey et al, 1952). Subsequently, many other cell lines have since been established from epithelial tissues, connective tissue, blood and lymph of several animals including man, hamster, monkey, and mouse (Bailey & Ollis, 1986). Cell fusion techniques pioneered throughout the 1960's and 1970's led to the creation of hybridoma cells (Kolher & Milstein, 1975) for manufacturing monoclonal antibodies.  Such antibodies have proven to be revolutionary tools for biochemical  research, diagnosis and more recently have been approved for specific cases of immunotherapy.  This has provided the impetus for the large scale cultivation of  9  hybridoma cells.  In addition, cloning techniques and recombinant D N A technology  developed in the early 1970's (Boyer & Cohen, 1973) permitted recombinant expression of other proteins of choice in appropriate bacterial hosts on a large scale. With time however, it became clear that many other proteins of commercial or potential therapeutic value (y-interferon, t-PA, F X , EPO and others) were very complex in nature and could only be manufactured by animal cells such as Chinese hamster ovary (CHO) or baby hamster kidney cells (BHK).  These cells harbor the proper cellular mechanisms to  orchestrate post-translational modifications necessary for producing bioactive molecules. Modifications include proper folding of polypeptide chains, disulfide bridge formation and protein glycosylation (discussed later). As a result, substantial efforts to culture these types of cells on a large-scale for recombinant protein production have been in progress since the late 1970's. Large-scale recombinant animal cell culture technology has grown in sophistication and plays an increasingly important role in the biotechnology industry.  This trend is  expected to continue in the future because animal cells remain the only way to produce practical quantities of many newer and equally complex recombinant proteins such as therapeutic and diagnostic monoclonal antibodies, vaccines and hormones.  By 1995,  approximately half of the $5 billion dollar annual turnover of the biotechnology industry was based upon this technology and in many cases with bioreactor volumes of more than 1000 L (Cooney, 1995).  Instrumentation and methods for improved control (and  therefore increased productivity) of animal cell bioprocesses has thus become an important part of biotechnology research.  10 1.2.4 Optimization  of Large-Scale  Animal  Cell Culture  Through  Nutrient  Control  Several aspects of large-scale animal cell culture are currently being investigated in order to further optimize such processes: 1) developing methodologies for increasing cell density and increasing oxygenation efficiency to obtain cultures that are more productive on a per volume basis (Cooney, 1995), and per cell basis (Xie & Wang, 1997), 2) determining effective feeding strategies and nutrient compositions for optimal growth and production (Zeng & Deckwer, 1995; Xie & Wang, 1997; Kurokawa et al., 1994), and 3) development of biosensors and other instrumentation to measure and control various parameters of a cell culture (Kurokawa et al., 1994, White & Turner, 1995). These 3 issues are inter-related and the work in this thesis focuses on the latter relating to biosensor instrumentation development. The optimization of bioprocesses depends on the ability to  accurately  monitor and control parameters describing the  bioreactor  environment. At present, the analysis of fermentation products, substrates and metabolites is usually achieved using off-line methods (Brooks et al., 1987/88). The availability of online sensors to monitor and control such analytes as glucose would provide an essential tool to the researcher for elucidating the relationships between metabolic processes inside cells and how they are affected by nutrient composition and other environmental factors in the growth medium.  Knowledge of these relationships is vital for the design and  implementation of optimized feeding strategies leading to increased cell culture productivity. A reliable on-line glucose analysis system would undoubtedly improve the  11 efficiency of fermentation monitoring, allowing for development of new feedback strategies and adaptive control techniques which presently rely upon mathematical models and estimation techniques (Bradley et al., 1991). Improvements in the economics of mammalian cell production systems are likely to be achieved by better medium design and process control (Kurokawa et al,  1994).  Control of optimal levels of glucose and  glutamine in a bioreactor is of crucial importance for achieving high concentration and productivity of cells and the desired protein products (Glacken et al., 1986).  Carefully  controlled and reproducible cultivation of recombinant mammalian cells is important for the development and scale-up of specific production processes. 1.2.5 Benefits  of Controlled  Glucose  Concentration  Substantial literature has been published over the last 10 years relating to the effects of nutrient composition and control in animal cell culture. Much evidence exists to show that glucose has effects on many different aspects of a culture.  Glucose  concentration has been shown to directly affect the efficiency, homogeneity and reproducibility of recombinant protein glycosylation patterns. Glycosylation is the most extensive of post-translational modifications and affects protein secretion, antigenicity and in-vivo clearance time (Jenkins et al., 1996).  Proteins can be glycosylated in three  different ways: 1) through N-glycosidic bonds with the R-group of Asn residues having the concensus sequence Asn-X-Ser/Thr; 2) through O-glycosidic bonds with the R-group of Ser or Thr; 3) glycophosphatidylinositol (GPI) anchors to secure some membrane bound proteins (Jenkins et al., 1996).  12  Using chemostat culture, Hayter et al, (1993) have shown that the pattern of glycosylation in y-IFN remains constant over a wide range of physiological conditions, but is altered under conditions associated with low dilution rates (0.013 h' ) in glucose-limited 1  chemostat culture. In addition, low dilution rates lead to conditions similar to the latter stages of batch cultures of CHO 320 cells (also expressing y-IFN) where cell growth rate is greatly decreased and nutrients are depleted (Curling et al,  1990).  Under such  conditions, the proportion of full glycosylated y-IFN decreases by 15-20%. Hayter further states that y-IFN glycosylation patterns remain essentially constant under steady-state conditions but is influenced by transient changes to different steady state conditions or by glucose additions during steady-state.  Xie et al., (1997), found that glucose starvation  leads to deteriorated y-IFN glycosylation efficiency. These authors all suggest that finer control of glucose concentration can lead to more homogeneous and consistent product. Secondly, glucose concentration has been found to have direct effects on waste product and inhibitory by-product accumulation which in turn affect culture viability, culture productivity and longevity. This concept was introduced briefly in section 1.2.2. Maintaining glucose concentrations at lower levels (0.44 mM) in ST cells was shown by Hu et al, (1987) to lead to reduced lactate formation and a higher percentage of glucose being utilized efficiently through complete oxidation. Miller et al. (1988) found the product distribution from glucose metabolism is determined in large part by concentration. Below 0.5 mM, at least half of the glucose in rat hepatoma cell cultures was incorporated into nucleotides; but at 5 m M 90% of the glucose was converted to lactate. These authors attributed the increased productivity and cell mass to a reduction of inhibitory by-products  13 such as ammonia and lactate. Extension of culture productivity is especially important because the productivity of animal cell cultures is inherently low as compared to bacterial cultures and this is in part due to lower viable cell densities and longer production phases. The body of literature at present suggests that maintaining glucose concentrations at low levels will result in greater productivity, more efficient use of energy and reduction of inhibitory by-product accumulation. Kurokawa et al. (1994) have shown that both cell concentration and antibody production increased 2 fold when both glucose and glutamine were maintained at ca. 1 m M and ca. 0.7 mM, respectively (using on-line HPLC). Hayter et al. (1993) report that maintaining a glucose concentration which is sufficient to support the minimum growth rate, u„un (  m  this  c a s e  c a  - ^0 uM), would be most efficient in terms  of cell utilization of glucose. A number of groups have reported attempts to regulate glucose concentration in cultures at ca. 1 m M or lower using different strategies. Hayter et al. (1991) was able to maintain glucose between 0.15-0.35 m M using chemostat cultures as did Haggstrom et al. (1994) for glucose concentrations <1 mM. Kurokawa et al. (1994) utilized an on-line H P L C system to maintain glucose between 1-10 mM. 1.2.6 Current  State-of-the-Art  for Nutrient  Control  in Animal  Cell  Culture  Numerous configurations of amperometric enzyme electrode biosensors for glucose and other analytes have been proposed in the literature but further development of practically useful and industrially acceptable sensors has been slowed because none have been able to adequately address the on-going challenges of long term signal stability and in-situ sterilizability (Pons, 1993). Less literature exists documenting glucose control in animal cell culture than for microbial culture due to the relative newness and smaller  14 market for this type of bioprocess. To the author's knowledge, there are no reports of true in-situ probe configurations developed for use in bioreactor-scale cultures of animal cells. Instead, the majority of reported glucose sensor systems utilized for closed-loop feedback control of animal cell cultures consist of a conventional enzyme-based bench top analyzer which is external to the bioreactor culture and has been linked via an automated sampling system and complex plumbing.  Drawn samples, which may undergo  pretreatment, are delivered to the analyzer through a carrier stream. A feeding pump and concentrated nutrient source is also linked with the bioreactor. The output of the analyzer is fed into the computer at predefined intervals and decisions are made based on the output as to whether or not to feed the cells by automatic activation of the feeding pump. These systems provide ex-situ "at-line" measurement, are typically configured as flow injection analysis (FIA) systems, and were first described by Ruzicka & Hansen (1975). These systems are complicated and at best difficult to implement, maintain and prevent from some form of failure.  Problems typically involve contamination, complexity,  hardware breakdown, sampling time, sample pretreatment.  Nevertheless, successful  experimental applications of these systems have been reported (White et al, 1995; Male et al, 1997; Ozturkera/., 1997). 1.2.7 Challenges  for Developing  In-Situ Enzyme  Sensors  for Animal  Cell  Culture  The fundamental challenge of developing a practically useful biosensor is to develop a stable configuration that is sufficiently sensitive to the analyte of interest and relatively insensitive to all other "external" factors. External factors are numerous and  15 include fouling and interference of the sensor component(s) from other analytes and molecules in the matrix, changes in the mass transport characteristics of the analyte of interest across the sensor interface, changes in the activity of an inherently fragile biological component of the biosensor and changes in environmental parameters such as pH, temperature and dissolved oxygen. Such external factors have a direct effect on the biosensor signal and result in signal instability and calibration drift over time. Accounting for and minimizing the effects of external factors is indeed challenging and has been the principle reason for the unavailability of on-line sensors for industrial applications.  The challenge of providing and maintaining long-term signal stability is  heightened in animal cell culture applications because of the need for much longer duration of biosensor usage and the much slower rate of glucose consumption. A number of issues specific to application of biosensor technology for large-scale animal cell culture processes need to be considered. 1. Protein containing matrices.  Protein is a major potential source of membrane  fouling and clogging because it readily adsorbs to many types of surfaces especially after longer term exposure. As a result, protein may significantly alter the mass transport characteristics of a sensor membrane and in turn lead to calibration drift. 2. Lower working range of glucose concentrations.  The glucose concentrations in  cell culture are in the range of 25 m M and usually less. For glucose optimized cultures, the concentration likely needs to be maintained at levels below 0.5 m M . A sensor must have adequate sensitivity to accurately measure such low  16  concentrations The typical commercial analyzer has a lower detection limit of ca. 0.8-3 m M which is 2-6X above the limit for effective glucose control. 3. A high susceptibility to bacterial contamination and hence the requirement for an absolute sterile barrier at the sensor interface. The sensor membrane is likely the most fragile barrier separating culture volume from the external environment. Preservation and maintenance of membrane integrity is an absolute necessity to prevent contamination of a culture. 4. Long term signal stability. This is perhaps the most difficult challenge and lasting obstacle to developing a practically useful in-situ glucose biosensor. 1.2.8 In-Situ Enzyme  Sensors  It is generally agreed that the best sensor configuration is an on-line, in-situ probe that provides real time continuous measurement of a culture parameter. Technologies for instrumentation to measure parameters such as temperature, pH and D O in this fashion already exist. Availability of an on-line sensor for glucose would prove extremely useful as a tool to achieve, maintain and study high cell density animal cell culture.  Rapid  assessment of metabolite concentrations is critical in more sophisticated mammalian cell bioreactors such as fed-batch and high density perfusion systems (Ozturk, 1994). Off-line methods may be too slow and/or labor intensive to be used in closed-loop control especially for high cell density systems where high frequency of sampling is required. It can also be difficult to ensure that samples are not significantly degraded/changed during the time required for sampling/analysis.  In addition, sampling increases the risk of  contamination (Huang et al., 1991). There is as yet no report of a functional glucose  17 sensor that truly measures on-line and in real time in a continuous fashion for useful periods of time. The longest experiments reported in the literature an in-situ enzyme sensor for pilot-scale fermenter use are 13 hours (Cleland & Enfors, 1983), 24 hours (Cleland & Enfors, 1983) and 9 hours (Brooks et al., 1987). There are several requirements and characteristics of an in-situ enzyme electrode probe, some of which were mentioned in the previous section because they are particularly relevant for animal cell culture applications. Other requirements for an in-situ probe are: 1. The response time of the sensor must be fast enough to effectively monitor changes in the analyte concentration in real time and hence enable feedback control. Response time is a critical issue in specialized situations such as high cell density cultures where the analyte concentration may be changing very quickly. Present technology is only beginning to permit high cell density animal cell culture and as such, closed-loop control of glucose concentration using an in-situ electrode has not yet been investigated. 2. The sensor must be insensitive to changes in typical culture parameters such as temperature, pH and dissolved oxygen.  This requirement is not normally an  important issue as current technology exists to effectively control these parameters in animal cell culture. 3. The sensor must be resistant to electrochemical poisons, enzymatic poisons, fouling agents and inhibitors which maybe present or are produced over time in a culture.  18 4.  The sensor must allow for re-calibration capability in order to periodically correct for changes in the calibration constants as a result of #3.  5. The sensor must have the capability to be steam sterilized with the bioreactor.  No in-situ on-line glucose sensor is commercially available that can fulfill all of the above stated requirements.  A small number of research groups have approached the  problem of developing an in-situ sensor for industrial use over the last 17 years, but all have been unsuccessful in fulfilling one or more of the above requirements. Enfors and Molin (1979) designed an in-situ sterilizable stainless steel probe housing which contained a dialysis membrane and an inner perfusable space for introduction of enzyme reagent. The shortcoming of this configuration was the difficulty in localizing enzyme to the electrode tip and the inability to internally calibrate the sensor once it had been loaded with enzyme. Introduction of glucose calibrant into the chamber would wash out the enzyme solution. Five years later, Cleland and Enfors (1984) developed a similar in-situ sterilizable probe body in which buffer could be continuously perfused through the inner space of the sensor.  The continuous perfusion feature was novel because it prevented the  accumulation of potential fouling species in the sensor chamber and also extended the linear range of the sensors measuring capability. However in their configuration, enzyme was loaded on an internal electrode and the electrode was placed in manually. Presumably this configuration was not designed to be amenable for automation.  19  Brooks et al. (1987) & Bradley et al. (1991) developed a similar configuration in which a stainless steel in-situ sterilizable probe body could be inserted through a side port of a bioreactor.  Their configuration allowed for internal calibration as the enzyme  component was immobilized onto graphite discs and placed inside the probe body behind a dialysis membrane.  In addition, their configuration allowed for internal buffer flow if  desired. The drawback to their system was the apparent difficulty in replacing the graphite disks. Presumably it was not amenable to automation. The systems described above require hands-on intervention in which a skilled operator must be on hand to deal with the regeneration of the enzyme component and the calibration of the resultant regenerated biosensor. The prototype described in this thesis is amenable to automation of these steps. 1.2.9 Usefulness  of Continuous  Measurement  and Future  Potential  of  the  Prototype  At present, the average glucose uptake rate of a conventional fed-batch animal cell culture is not high enough to require continuous sampling/off-line measurement or even sampling at more frequent than ca. 6 hours in order to maintain glucose concentration within close range of a relatively high set point (ca. 10 mM). Instead, infrequent off-line sampling coupled with open-loop control based upon models and equations to predict glucose consumption rates and hence glucose concentration into the short term future of a culture have been demonstrated to be useful for such situations.  Sampling at frequencies  of once or twice a day provide sufficient information for model prediction to enable reasonably fine control of a culture at a high set point glucose concentration. However,  20 margins of error exist in both the calculated glucose uptake rate and the off-line glucose measurement.  As a result, there is error associated with the forcasted direction of the  uptake rate and the near future glucose concentration.  The margin of error increases  when higher glucose uptake rates prevail and/or the time between sampling increases. Tolerance for this margin of error decreases as the set point is lowered. At set point levels close to zero, it is critically important to ensure that the real glucose concentration never falls to zero and so the frequency of sampling must necessarily increase to ensure that glucose is not exhausted. At some limiting set point, it will be necessary to sample and measure continuously. The prototype envisioned in this work would be most useful as a tool during high cell density animal cell culture when glucose uptake rates are much higher and when the set point glucose concentration is low. In such a situation, effective continuous real-time measurement is required and such capability would provide invaluable information describing the metabolic state of the cells. Additionally, attainment of effective continuous monitoring under such dynamic conditions would possibly enable closed-loop feedback control of nutrient composition to maintain high cell density and viability.  Present  technology using open-loop control models for fed-batch culture or the use of perfusion cultures is only beginning to allow such high cell densities to be reached. Applicability of open-loop methods of control is limited because the models they rely on contain certain assumptions and are therefore inherently error prone.  Reports of glucose monitoring  and/or control under high cell density and high glucose uptake rate conditions have not yet taken place yet and is due in part to the unavailability of appropriate biosensor  21 instrumentation. The potential of the prototype envisioned in this work is to fulfill this specific instrumentation need for a highly specialized process that, into the new millennium, will play an ever increasingly important part of biotechnology.  1.3 Features of the Prototype 1.3.1 Glucose  Oxidase  Most glucose sensors are amperometric and utilize glucose oxidase enzyme immobilized at the face of a platinum electrode. The prototype presented in this work relies on the same principle.  These sensors detect electrons generated  from  electrochemical oxidation of hydrogen peroxide, a product of enzymatic oxidation of glucose. The quantity of electrons or current is related to the amount of glucose in the solution being measured. The major source of glucose oxidase is from Aspergillus niger sp. preparations. The enzyme is a 1135 amino acid glycoprotein homodimer with each subunit of the dimer having a tightly bound F A D group and each subunit coded for by the same gene. The molecular weight of the active holoenzyme has been reported to be between 155 and 186 kD depending on the source, with 155 kD being the most common. There is substantial literature available presenting detailed comprehensive descriptions of the mechanism of glucose oxidase enzyme (Wilson & Turner, 1992). A brief introduction is appropriate for this work.  22 The enzymatic oxidation of the P anomer of D-glucose by glucose oxidase (GOX) is often described by the following overall net reaction equation: P-D-Glucose + O2 + H2O  D-Gluconic Acid + H2O2  (l.l)  This equation is actually the sum of 2 redox reactions and a spontaneous non-enzymatic hydrolysis step. The first redox step involves the oxidation of glucose to a lactone with the generation of 2 electrons and concomitant reduction of the F A D prosthetic group of G O X to its reduced form: P-D-Glucose FAD + 2 H  +  > 8-Gluconolactone + 2 H  +  + 2e"  (12)  + 2e" <—-> FADH2  (1.3)  The second redox reaction involves the transfer of electrons from F A D H to molecular 2  oxygen in order to regenerate the flavin co-factor back to the oxidized form.  The  corresponding reduction of O2 leads to the formation of hydrogen peroxide: FADH  0  2  2  <  + 2H  > FAD + 2 H  +  + 2e" <  + 2e"  +  > H 02 2  (14)  (1.5)  The final step in the breakdown of glucose is the spontaneous hydrolysis of the lactone to the acid form: 8-Gluconolactone + H2O <  > D - Gluconic Acid  Equations 1.2-1.6 sum to yield the first overall reaction (Eqn. 1.1).  (16)  23 Some of the hydrogen peroxide generated by the enzymatic action of G O X (Eqn. 1.5) diffixses to the surface of the platinum working electrode and is electrochemically oxidized according to the following equation:  H2O2 <  >2 H  +  ( i 7)  + O2 + 2e"  The standard potential for this half cell reaction is E° = +0.695 V versus the standard hydrogen electrode potential (SHE) at pH 0. At pH 7.0, the standard potential is shifted by -0.410 V . The applied potential used for biosensor applications is usually slightly more positive than the standard potential to ensure that the reaction is not kinetically limited. An applied potential of +0.94 V is normally used.  Under such conditions, the  electrochemical reaction is limited only diffusionally which is desirable. A saturated calomel electrode (SCE) or silver/silver chloride (Ag/AgCl) reference electrode is normally used for biosensor applications because the SHE is rather cumbersome and inconvenient. The standard potentials for these electrode half reactions differ from each other and relative to the SHE they are given below:  2H  +  + 2e" <  >H  H g C l 2 + 2e" < 2  2AgCl + 2e" <  2  ( i atm)  > 2Hg + 2C1"  > 2C1" + 2Ag  E° = 0.000 V  (1.8)  E° =+0.241 V v s . SHE  (1.9)  E° =+0.222 V vs. SHE  (1.10)  With all this taken into account, the normal electrode bias potential for many biosensor applications is approximately 0.7 V versus the saturated calomel or silver/silver chloride reference  electrode.  A more detailed and comprehensive  presentation  of the  24 electrochemical theory can be found in standard texts of electrochemistry (Bard & Faulkner, 1980; Sawyer & Roberts, 1974). 1.3.2 Cellulose  Binding  Domain  Technology  A novel immobilization method is used in the sensor enabling simple localization, removal, and replacement of the G O X enzyme component in the sensor.  The method  utilizes a cellulose binding domain (CBD) protein which binds strongly to and can be efficiently removed from cellulose. The gene coding for this protein has been cloned and expressed in E. coli and the resultant C B D polypeptide isolated. The binding properties of the polypeptide to cellulose have been studied extensively (Gilkes et al,  1991, 1992;  Creagh etal, 1996; Jervis etal, 1997).  Proline - Threonine Linker  NH,  COOH 316 335 Catalytic Domain  443  Cellulose Binding Domain  F i g u r e 1.1: Schematic diagram of the structure of cellulase exoglucanase from C. fimi. This diagram shows the two distinct and independent domains which are linked by the proline-threonine linker.  The binding protein comes from a family of cellulase enzymes originating from Cellulomonas fimi. These enzymes are modular in structure and contain distinct catalytic domains that are separate from the binding domain. Using molecular biology techniques,  25 the coding region for the binding domain has been removed from the catalytic domain and genetically linked with a gene encoding for a second protein to form a chimera or fusion protein. It has been found that the activities of the binding domain and its fusion partner are preserved even when bound to cellulose.  CBD's have been fused with several  different proteins such as 1) protein A (Ramirez et al, 1995), 2) P-glucosidase (Ong et al, 1991), and 3) IL-2 (Ong et al, 1995) in order to immobilize them onto cellulose. C B D ' s have proven to be effective and powerful affinity tags for affinity purification of fusion proteins using cellulose-based columns (Kilburn et al, 1992). In this work, C B D protein is chemically linked with glucose oxidase protein to form a conjugate molecule which can be immobilized to a cellulose matrix in the prototype sensor.  The  chemical linking  utilizes sulfosuccinimidyl 4-(N-maleimidomethyl)  cyclohexane-l-carboxylate (SMCC), a hetero-bifunctional crosslinker. 1.3.3 Operational  Features  The hardware for the regenerable biosensor system consists of a platinum indicating electrode, a porous cellulose matrix, and a modified permselective dialysis membrane which is all incorporated into a stainless steel probe body that inserts into the sideport of a bioreactor (Figures 2.1-2.4).  Detailed descriptions of the prototype are  presented in Chapter 2. The porous cellulose matrix is sandwiched between the tip of the platinum indicating electrode and the dialysis membrane.  The probe body contains an  inner perfusable space, linked by inlet and outlet ports, enabling enzyme-CBD conjugate in an appropriate electrolyte loading buffer to be introduced into the enzyme chamber and bound to the sandwiched cellulose.  26 The ease of binding to and elution off of cellulose using C B D technology coupled with an inner perfusable space confers unique operational advantages to the sensor prototype. Firstly, in-situ steam sterilization of enzyme-based biosensors in a bioreactor has traditionally not been possible due to the thermal sensitivity of the enzyme component. In the present case, the probe hardware is sterilized in-situ but without the enzyme. After sterilization, enzyme conjugate is loaded into the sensor through the inner perfusable space using a peristaltic pump. Secondly, when during the course of longer term use the activity of the enzyme has deteriorated to unacceptable levels, the signal can be regenerated online in a relatively simple fashion. Regeneration involves removing the "spent" conjugate using an appropriate elution buffer via the inner perfusable space followed by replacement with fresh conjugate. The net result is an extension of the operational lifetime of the sensor without having to breach the sterility of the bioreactor or interrupt the culture in progress. These features represent significant improvement in the functionality of such sensors. Another advantage of this system is the potential for automation of the enzyme-conjugate loading and regeneration system. The implementing of automation was beyond the scope of this thesis. The mechanism of glucose oxidase enzyme and its exploitation for usage with C B D technology in the present amperometric sensor configuration can be expanded to include a wide range of different analytes, many of which are physiologically relevant. A n entire family of oxidoreductase enzymes exist which catalyze the breakdown of different substrates with the formation of peroxide. As a result, it should be possible in principle to  27 construct a "generic" sensor whose analyte specificity is dictated in part by the choice of enzyme-CBD conjugate (Table 1.1).  T a b l e 1.1. Industrially important enzymatic assays requiring oxidase enzymes. The general oxidase enzyme-catalyzed reaction is of the form: Oxid&sc Substrate(reduced form) + 0  2  > Product(oxidized form) + H 0 2  2  SUBSTRATE  ENZYME  PRODUCT  (Reduced Form)  CATALYST  (Oxidized Form)  p-D-Glucose  Glucose Oxidase  D-Gluconolactone  L-Lactate  L-Lactate Oxidase  Pyruvate  Ethyl Alcohol  Alcohol Oxidase  Acetaldeyhde  Lactose  Galactose Oxidase  Galactose Dialdehyde  Glycerol  Galactose Oxidase  Glyceraldehyde  Cholesterol  Cholesterol Oxidase  4-cholesten-3 -one  Pyruvate  Pyruvate Oxidase  Acetyl Phosphate  Uric Acid  Uricase  Allantoin  Acetaldehyde  Aldehyde Oxidase  Acetate  Xanthine  Xanthine Oxidase  Urate  Choline  Choline Oxidase  Betaine  L-Glutamate  L-Glutamate Oxidase  a-Ketoglutarate  Acetylcholine*  Acetylcholine Esterase  Choline  L-Glutamine*  L-Glutaminase  L-Glutamate  Maltose*  Glucoamylase  p-D-Glucose  Starch*  Amyloglucosidase  p-D-Glucose  Sucrose*  Invertase + Mutarotase  p-D-Glucose  * Assays that require a multi-enzyme system.  29  1.3.4 Prototype  Performance  in Microbial  Culture  The prototype which was originally designed for use in microbial cultures performed satisfactorily in that application.  It was used to monitor the glucose  concentration for 16.5 consecutive hours in a low-density E. coli culture. During this time period, the enzyme component was regenerated once on-line and re-calibrated. A crosscorrelation plot of the prototype sensor output to off-line Beckman values showed reasonable correlation. The prototype was also employed in a 20 hour high cell density E. coli culture. The prototype was tested for its potential for closed-loop feedback control. During a 3 hour period of the culture, the glucose concentration was maintained manually at a low and constant level based on the output of the sensor. Cellular growth rate during this same time period increased 30%. Increased growth rate suggested that closed-loop glucose control based on the sensor signal is feasible and has a measurable effect on the course of the fermentation. The output reading of the prototype was compared to off-line values and it was found that the sensor values were consistently higher than those from the off-line analyzer (Phelps, et al, 1994, 1995).  1.4 Thesis Overview and Research Objectives It was anticipated that problems and challenges would arise during steps to implement the original prototype sensor for animal cell culture use. The overall objective of this work was to identify and solve these problems and then to utilize the modified sensor in animal cell culture with the aim of characterizing and improving long-term signal stability. The results of long-term signal stability investigations are documented in 3 major  30 experiments in chapter 4. In pursuit of the overall objective, numerous problems with the prototype became apparent especially after the first long-term experiment in chapter 4. These problems led to modifications and characterization experiments which became important objectives of this work and are described in chapter 2. The reader should keep in mind that the contents of chapter 2 were actually motivated by results from the first long-term experiment in chapter 4. The results in chapter 2 enabled execution of the final 2 experiments in chapter 4.  In conjunction with efforts to develop and characterize the prototype sensor for longer-term usage, an attempt was made to develop a model animal cell line expressing a recombinant Factor X protein derivative and this work is described in chapter 3. The objective was to utilize the developed cell line for glucose monitoring studies using the prototype sensor. The specific objectives of this research were to: a. Modify existing prototype hardware in order to meet the requirements of animal cell culture.  Specifically, to redesign the membrane assembly for increased  mechanical strength and to incorporate a continuous perfusion system to discourage accumulation of gas bubbles in the sensor chamber. b. Develop an improved chemical conjugation protocol to produce conjugate protein with increased specific activity as compared to an existing protocol and to characterize this conjugate in terms of lower limit of detection in the sensor and binding characteristics to cellulose. c. Develop a model cell line (using CHO cells) for glucose monitoring studies.  31 d. Investigate the longer term stability and behavior of the sensor signal during use in bioreactor scale animal cell culture and characterize the signal stability.  A great deal of difficulty was encountered with developing a prototype that was functional for the longer term. However, in the end, the prototype was successfully modified such that the specific objectives could be fulfilled. The results in the following chapters reveal the challenges faced with developing an in-situ enzyme electrode and as well, demonstrate the potential for the prototype to provide continuous on-line real time measurement of a culture variable which at present cannot be measured reliably in such a fashion.  32  CHAPTER 2 Hardware Modifications and Prototype Characterization 2.1 Introduction This chapter begins with a description of the hardware components and operation of the prototype sensor system in section 2.1.1. elsewhere (Phelps, 1993).  Detailed descriptions are outlined  Section 2.1.2 discusses the motivation for hardware  modifications and characterization experiments. The remainder of this chapter (sections 2.3, 2.4) describes the hardware modifications and presents results from prototype characterization experiments. 2.1.1 Hardware  Components  and Principle  of  Operation  The basic hardware is composed of a probe body, a porous cellulose matrix, a glucose permeable outer membrane, an internal electrode assembly, the reagent flow system, and finally an electrochemical detection and data recording system. Probe Body: The probe body consists of a modified stainless steel Ingold C O 2 electrode (Ingold Electronics specification's manual, 1990) in which some of the components were redesigned to accommodate the needs of an in-situ glucose biosensor for microbial cell culture (Phelps, 1993). The body has an inner perfusable chamber that is accessible via two ports for inflow and outflow of liquid (Figure 2.1, 2.2). This perfusion system is used to introduce enzyme-CBD  cex  conjugate protein as well as to remove it with the appropriate  elution buffer when necessary.  Regulation and control of flow through the perfusable  space is carried out using silicon tubing, valves and a peristaltic pump (Figure 2.4).  33 Porous Cellulose Matrix: The porous cellulose matrix consists of a double layer of Whatman #1 filter paper (Whatman International Ltd., Maidstone, England) cut to approximately 7 mm diameter. This material was tested along with several other cellulose-based materials and chosen as the most suitable in terms of the combination of porosity, response time, resultant signal sensitivity, and ease of handling (Phelps et al, 1993). The matrix is sandwiched between the platinum electrode and a permeable outer membrane. The placement of the cellulose matrix is illustrated in Figure 2.2. Internal Electrode Assembly: The internal electrode assembly was designed and fabricated in-house and consists of a 1-mm diameter platinum wire (Aldrich, Milwaukee, WI, USA) encased in 4-mm O.D. flint glass tubing. The exposed circular platinum disk had a surface area of approximately 0.785 mm . 2  A 1-mm diameter platinum counter electrode and 1-mm diameter silver-  silver chloride reference electrode was tightly coiled around the glass shrouded working electrode. The Ag-AgCl reference electrode was manufactured in-house by anodizing the A g wire in the presence of CI' ions, based on the method of Sawyer and Roberts (1974) and is illustrated in Figure 2.3. The entire electrode assembly is housed in a stainless steel casing that is inserted into the probe body and lowered onto the cellulose as shown in Figure 2.1. On frequent occasions the bare platinum indicating electrode was cycled in quiescent 0.5 M H S 0 between -0.26 V and +1.2 V at 100 mV/s for 10 minutes followed 2  4  by anodizing at +1.8 V for 10 minutes.  The cycle was repeated until a stable  34 voltammogram was obtained. This was done to electrochemically clean the bare platinum, after which, the surface was coated with cellulose acetate solution based on modification of a method by Wang and Hutchins (1985, 1992) and described elsewhere (Phelps, 1993). A stainless steel dummy electrode plug was also designed and fabricated to replace the real electrode assembly during steam sterilization of the probe body with the bioreactor.  After sterilization, the dummy electrode was replaced with the internal  electrode assembly followed by enzyme loading. Glucose Permeable Outer Membrane: The membrane material consists of 0.2 um cellulose triacetate sterile filtration membrane (Gelman Instrument Company, Ann Arbor, M I , U S A ) impregnated with NAFION™ polymer (Aldrich Chemicals, Milwaukee, WI, U S A ) as illustrated in Figure 2.2. Nafion polymer has been used in many different membrane preparations because it confers a negative charge to the membrane such that it is better able to reject negatively charged species from crossing the membrane and fouling the electrode or enzyme or both (Wang, 1992; Scouten etal, 1995). Ascorbic and uric acid, both of which are negatively charged, are known to foul platinum electrode surfaces (Wang, 1992). Nafion polymer also decreases membrane pore size resulting in increased mass transport resistance and extension of the linear dynamic range. The purpose of the dialysis membrane is to allow the diffusion of glucose, oxygen and electrolytes into the enzyme layer while excluding potential interfering species such as cells, proteins, enzyme inhibitors and electrochemical interferents which may be present in the analyte medium. Equally important, the dialysis membrane must also act as a sterile  35 barrier to prevent the passage of contaminating species such as bacteria in the external environment from entering into the culture volume.  A consequence of its role as a  permselective membrane is that it is the most delicate and fragile barrier separating the culture in a fermenter from the external environment. The integrity of the membrane must therefore be maintained absolutely at all times. The membrane also creates significant mass transport resistance which increases the linearity of response and extends the working range of the sensor. The mass transport characteristics must also remain stable as instability will contribute to drift in the sensor calibration. Reagent Flow System: This system (Figure 2.4) consists of reservoir bottles connected together and linked with the perfusable inner space of the sensor body via a combination of three way valves and lengths of silicon tubing. The reservoir bottles contained: R l : PBS buffer electrolyte (0.1 M NaCl, 5 m M N a H P 0 , and 30 m M N a H P 0 , p H 7.2, preserved with 2  4  2  4  1 m M E D T A and 5 m M sodium benzoate); R2: 8 M guanidine hydrochloride for elution; R3: waste collection. A Gilson Minipuls 3 peristaltic pump (Gilson Medical Electronics, Middleton, WI, U S A ) was used to transport fluids to and from the reservoir bottles and through the inner perfusable space. In the present case, all reagent flow protocols were controlled manually, but the system has the potential to be automated using electronically activated and computer controlled pumps and valves. Electrochemical Detection and Data Recording System: A Pine AFRDE4 bi-potentiostat (Pine Instrument Co., Grove City, PA, USA) was used to generate the +0.7 V bias potential necessary to oxidize glucose concentration  36 dependent hydrogen peroxide and collect/amplify the resulting current.  The current  output was recorded on a Kipp & Zonen Model BD112 strip chart recorder (Kipp & Zonen, Delft/Holland) and a Kipp & Zonen Model BD91 X Y Y T recorder was used for recording cyclic voltammograms.  Measurement  Calibration  1 2 0 m l Syringes 2 H i g h - l e m p . coaxial cable 3 Cable onion nut 4 Ring n u t for retracting 5  Plugs  6  Feed tubes  7 W e l d - i n socket  •  8 Guide tube (probe body) 9  Electrode shaft  10 p H e l e c t r o d e 11 R e f e r e n c e e l e c t r o d e 12 C 0  2  electrolyte  13 M e m b r a n e c a r t r i d g e 14 C a l i b r a t i o n b u f f e r 15 G l a s s m e m b r a n e 16 R e i n f o r c e d s i l i c o n e m e m b r a n e  F i g u r e 2.1 Construction of the Ingold C 0 2 probe. Reprinted from the Ingold product catalogue with permission from Ingold Electrodes Inc.  37  On-Line Regenerable Glucose Biosensor  Porous Cellulose Matrix  Probe Body  * <8  Dialysis Membrane  [ Em>m«^80R«aa«nt  {  Inlet Conduit ln«ul»tor  Glucose + O,  Electrode  Cells, Protein, etc. «  Insulator  Iii y Efizynw-CBD Buwrt  ^  Outlet Conduit  Enzyme-CBD Conjugate — Bioreactor Wall -  F i g u r e 2 . 2 Schematic cross-sectional diagram of sensor tip illustrating the inner perfusable space, dialysis membrane and porous cellulose matrix. (Adapted and reprinted from Phelps, 1993 with permission)  Pt INDICATING ELECTRODE  Pt COUNTER ELECTRODE  A g / A g C I REFERENCE ELECTRODE  GLASS SHROUD  EPOXY  GOLD CRIMPS  LEAD WIRES  F i g u r e 2 . 3 Diagram of the internal electrode assembly. The glass-shrouded platinum working electrode, platinum counter electrode and silver-silver chloride reference electrode are shown. (Adapted and reprinted from Phelps, 1993 with permission)  38  CR  0.5 mLs  r  7.2 mLs  5.6 mLs  V2  V l ^ 2.2 mLs v ^ ^ ' V  2.2 mLs  2.6 mLs  2.4-2.5 mLs Rl  R2  R3  R4  F i g u r e 2 . 4 Schematic diagram illustrating the reagent flow system and instrumentation. R1, internal electrolyte and wash buffer reservoir (PBS); R2, internal calibrant reservoir (not used); R3, elution buffer reservoir (8M guanidine in PBS); R4, waste reservoir; V1 and V2, three way valves; P, peristaltic pump; E, enzyme chamber; PO, potentiostat; CR, chart recorder. The volume of the enzyme chamber depended on the position of the internal electrode unit (ie., raised or lowered). Solid and dashed lines represent flow lines and electrical, respectively. (Adapted and reprinted from Phelps, 1993 with permission)  39 Operation Protocol: The protocol for loading and eluting conjugate protein using the reagent flow system is described in greater detail elsewhere (Phelps, 1993). Briefly, the protocol for loading consists of: 1. Raising the internal electrode assembly. 2. Pumping G O X - C B D  cex  conjugate solution for 1 minute at pump setting of 4 rpm  (2.2 ml/min). 3. Pumping PBS for 5 minutes at 4 rpm to move the bolus of conjugate into the enzyme chamber of the sensor. 4. Stop flow for 5 minutes to allow binding of conjugate to the cellulose matrix. 5. Pumping PBS for 7.5 minutes at 4 rpm to wash unbound conjugate from the enzyme chamber. 6. Lower the internal electrode assembly. Conjugate was introduced into the flow system by disconnecting the inlet line from the PBS reservoir bottle and reconnecting with a reservoir of conjugate.  After intake of a  bolus of conjugate by pumping for 1 minute, the line was reconnected with the PBS reservoir. Total time required for loading was 14 minutes. The elution protocol consists of: 1. Raising the internal electrode assembly. 2. Pumping 8 M Guanidine H C L elution buffer for 2.5 minutes at 4 rpm. 3. Pump PBS for 2 minutes at 4 rpm. 4. Pump PBS for 5.5 minutes at 48 rpm (maximum pump speed).  40 5. Lower internal electrode assembly or load new enzyme. Total time required for elution was 10 minutes. 2.1.2  Motivation  for  Hardware  Modifications  and  Characterization  Experiments  Results from the first long-term experiment with the prototype in chapter 4 and efforts to set up subsequent long-term experiments revealed a number of problems with the hardware. Problems related to membrane cartridge integrity, the inability of the probe to respond to glucose when enzyme was loaded and finally, typical loss of responsiveness after 24-48 hours.  These issues were addressed in order to develop a prototype that  would be functional for the required duration of an animal cell culture process. Glucose Permeable Outer Membrane Cartridge Assembly: The original design (Figure 2.6) was not durable enough to withstand harsh steam sterilization conditions. The original steam sterilization protocol of the bioreactor with the sensor incorporated an external pressure equalization line to directly link the perfusable space of the sensor with the internal bioreactor chamber in order to relieve any pressure differential which may be created. It was found that the pressure equalization line did not adequately alleviate a 1 bar pressure differential created during 40 minute steam sterilization cycles and caused the delicate sensor membrane to balloon outwards.  After  aborting and dismantling contaminated bioreactor experiments, the membrane component of the sensor was always found to be wrinkled and stretched. membrane  This deformation of the  material could possibly have resulted in microscopic rupturing and  compromised integrity. Bacterial contamination was determined most often to be due to a  41 disruption in the sterile barrier normally offered by the dialysis membrane or leakage due to inadequate strength of sealing between the membrane cartridge and the probe body. The definitive test implicating the membrane assembly was to steam sterilize the bioreactor containing bacterial cell broth and the sensor prototype inserted through the side port. The system was then left untouched for several days to ensure sterility. Afterwards, a small volume of viable E. coli bacterial culture was manually injected into the inner perfusable space of the sensor to directly challenge the sensor membrane assembly integrity. Inevitably the broth media in the bioreactor rapidly became contaminated with E. coli and so it was postulated that stretching and ballooning had placed unnatural stress on the membrane itself and/or the sealing system around it, possibly leading to its rupture. The membrane cartridge assembly was redesigned with the goal of increasing its mechanical strength such that it would be able to withstand steam sterilization conditions. Continuous Perfusion: On numerous occasions of biosensor usage in benchtop or bioreactor experiments, the prototype lost sensitivity to glucose after 24-48 hours of use. Examination of possible explanations indicated that whenever the inner perfusable space of the sensor was flushed after it had lost sensing capability, a series of air bubbles would emerge in the outlet flow tube indicating that gas bubbles had formed inside the chamber over the preceding time period. In addition, when the sensor was dismantled after long periods in the bioreactor, a water-line mark was usually seen on the inside walls of the sensor chamber. This indicated that the chamber was not completely filled with liquid but instead contained gas bubble(s).  42 A system to continuously perfuse the inner space of the sensor with de-oxygenated buffer was developed to counteract gas bubble accumulation leading to sensor malfunction. Buffer flow through an enzyme chamber of an enzyme electrode has been reported in the literature by Enfors (1984).  The author developed an in-situ enzyme probe  configuration for glucose which incorporated continuous perfusion of the enzyme chamber space with buffer solution. The intended purpose of perfusion was to dilute glucose at the electrode surface thereby extending the linear dynamic range of the sensor. A secondary effect of the buffer flow was to mitigate sensor sensitivity to normal fluctuations in microbial culture D O since a constant and continual supply of oxygen was provided by the buffer flow. Improved Conjugate Using a Hetero-Bifunctional Crosslinker: Several characteristics of the original glutaraldehyde conjugate were found to be undesirable. Firstly, initial batch syntheses of glutaraldehyde conjugate lead to wide batchto-batch variations in resultant specific activity. Secondly, the sensor signal sensitivity afforded by the glutaraldehyde conjugate at lower glucose concentrations (ca.<5 mM) was usually not adequate. Higher probe sensitivity would be needed for animal cell cultures grown at low glucose concentrations. Thirdly, results from a first long-term hybridoma culture (chapter 4.3.1) showed that the longer term sensor signal stability afforded by this conjugate between calibrations was poor and impractical for a potentially useful sensor. It was postulated that part of the explanation for the above observations was the choice of crosslinker and manner of crosslinking.  The original crosslinking protocol  utilized glutaraldehyde. This crosslinker is homo-bifunctional, having two aldehyde ends  43 that are reactive towards primary amine groups such as the e-amino group of lysine residues and the amino terminus of polypeptide chains. There are approximately 30 lysine residues on the surface of glucose oxidase (Rasmol Molecular visualization program, Sayle, 1995). CBDce has a single surface exposed lysine residue (Jervis et al, 1997) and, X  comprised as a single polypeptide chain, also contains a single amino terminal group (Ong et al., 1993). The availability of a large number of potential crosslinking sites on the surface of G O X coupled with the homo-bifiinctionality of glutaraldehyde led to a high degree of both inter and intramolecular glutaraldehyde crosslinking between the two proteins which was difficult to control and reproduce.  The net result of the coupling  reaction is the formation of heterogeneous populations of conjugate polymers all differing in size, ratio of C B D  c e x  to G O X composition, structural stability and ultimately, resultant  specific activity. Additionally, the Schiff base bond created between the amine group of lysine and the carbonyl group of glutaraldehyde is inherently unstable and reversible over time unless it is reduced further with reducing reagents such as sodium borohydride (Fersht, 1985; Hansen & Mikkelson, 1991). As a result, it is possible that some parts of the conjugate polymers are slowly unlinking and released from the cellulose matrix, thus contributing to the instability of resultant conjugate protein molecule structure and ultimately, the biosensor signal. Glutaraldehyde is a widely used and effective reagent for many protein crosslinking applications but the stability of the Schiff base bond may become a consideration for applications requiring immobilization for longer term usage. FernandezLafuente et al. (1995) immobilized Penicillin G acylase on glyoxyl agarose gels using  44  glutaraldehyde and investigated the stability of acylase activity in these derivatives. They found an additional 10-20% decrease in residual activity of non-reduced derivatives after 20 hours compared to borohydride reduced derivatives after the same time period. The authors suggest that reduction of aldehyde-amide bonds is absolutely necessary in order to stabilize and terminate a crosslinking reaction and that the usual omittance of this reduction step in protein crosslinking can lead to results that are difficult to explain. Difficulty in reproducing glutaraldehyde crosslinking results have also been attributed to the instability of glutaraldehyde preparations; the crosslinker, unless freshly purified, consists of polymers which can form as a result of complex chemistry (Scouten et al., 1995).  F i g u r e 2.5 Structure of sulfosuccinimidyl 4-(N-maleimidomethyl) cyclohexane-1carboxylate (SMCC) crosslinker.  The modified protocol utilizes a sulfosuccinimidyl 4-(N-maleimidomethyl) cyclohexane-1-carboxylate (SMCC) hetero-bifunctional crosslinker, M W = 436.37 (Pierce Chemicals, Rockford, IL, USA).  This crosslinker (Figure 2.5) contains 2 chemically  reactive ends, each reacting with different side groups of constituent amino acid residues. The N-hydroxy succinimide (NHS) ester end reacts with primary amines and the N terminus of polypeptide chains. The maleimide end reacts with free sulfhydryl groups of  45 cysteine residues. The modified protocol was designed to target 2 free amino groups in CBD e C  X  protein (Ong et al, 1993) and free sulfhydryl groups to be introduced onto surface  lysine residues in G O X using Traut's reagent, 2- iminothiolane (Pierce). S M C C crosslinker has been used successfully to conjugate G O X with rabbit IgG for various enzyme assay reagents (Yoshitake et al., 1979) and other proteins as well. Imagawa et al. (1982) used SMCC to create antibody-horse radish peroxidase (HRP) conjugates for E L I S A applications. In this case, conjugation results were compared with results obtained using glutaraldehyde. The authors reported that the HRP activity retained was ca. 10% higher and heterogeneity of resultant conjugate molecules was lower for conjugates produced with S M C C versus glutaraldehyde. Lower Limit of Detection: As mentioned previously, the activity of conjugate loaded into the sensor would need to be increased for measurement at lower glucose concentrations and this provided part of the impetus to devise a new conjugation protocol with S M C C crosslinker. The lower limit of detection using SMCC conjugate was determined and compared with the original glutaraldehyde conjugate. The lower limit of detection of an analytical instrument is generally defined as that measured value corresponding to a signal that is three standard deviations above the mean of the background (Skoog, 1985). Binding Isotherms: Solution-depletion binding isotherm experiments were done in order to determine if non-binding, failure to bind or unstable binding to the cellulose matrix was taking place and hence an explanation for inability of the sensor to respond to glucose during many  46 trials and experiments.  Whatman paper cellulose was determined previously to be the  most suitable sorbent material for use in the sensor (Phelps, 1994). However, nothing was known about the affinity or binding capacity of conjugate for the Whatman cellulose matrix and the implications for biosensor performance. Unanswered questions include: Is the cellulose matrix being saturated each time a loading is made? required for saturation?  How much time is  What affinity does the conjugate have for the cellulose?  Is  conjugate protein desorbing from the cellulose over time in the sensor? These questions could begin to be addressed using the Langmuir adsorption isotherm equation (Adamson, 1990) as a mathematical model for describing the observed adsorption behavior.  r=^5Z  ( 3 1 )  1+ K F  v  1  This adsorption isotherm and the underlying theoretical context represent the most convenient method for analyzing experimental data. The T term represents the fraction of the maximal number of available sites on a sorbent surface occupied by a ligand, in this case, bound conjugate protein. N is the maximal number of binding sites on the surface 0  of the sorbent,  is the equilibrium association constant of adsorption, and F is the  concentration of free ligand or conjugate protein. The units of N and K depend on the Q  units of T and F.  a  For protein adsorption, free ligand, F, is commonly expressed as  umole/ml and bound protein, T, as umole/mg sorbent. In this case, the units of N become 0  umoles/mg and K becomes ml/umol. Units for T and F can also be expressed in terms of a  mass (ug ligand/mg sorbent and ug ligand/ml respectively) instead of molar quantities. In  47 this case, the units for N become p:g conjugate/mg cellulose and the units for K» become a  inverse density, D ' or ml/mg. 1  2.2 Materials and Methods 2.2.1 Membrane  Cartridge  Assembly  The original configuration was prepared as follows: a 25 mm diameter 0.2 urn cellulose triacetate membrane material was wetted with 50/50 d H 0 : isopropyl alcohol 2  and then draped over a 50 urn stainless steel wire mesh screen support which in turn was draped over a hard plastic rim as shown in Figure 2.6. The membrane was then fastened in place over the plastic rim by tying a length of stainless steel wire around the base of the plastic rim. Nation was impregnated into the fastened membrane by using a micropipet to add one coat of 250 | i l 0.5 % Nation (diluted from a 5% stock using 50/50 dH 0/isopropyl alcohol). After at least 1 hour air drying time, a thin coat of silastic 2  adhesive (Dow Corning Canada Inc., Mississauga, Ontario, Canada) was spread around the base of the plastic rim and over the tied down stainless steel wire. A brass collar that fit over the edge of the hard plastic rim was fitted in place to seal all the components together. In addition to the ballooning effect caused by the pressure differential, draping the delicate membrane material over the plastic rim placed additional unnatural stress on the membrane especially at the folded edge and this contributed to the possibility of membrane rupture during the course of usage. The membrane cartridge was redesigned such that the cellulose triacetate membrane remained flat and was sandwiched between two stainless steel wire mesh  48 screens instead of the original single screen as shown in Figure 2.7. The result was a 3layer configuration.  In addition, the membrane material was impregnated with Nation  before assembling the cartridge as follows: cellulose triacetate circles (22 mm diameter) were cut from sheets and wetted with 50/50 dH 0/isopropyl alcohol. Excess wetting 2  solution was removed by touching the edge of the membrane to absorbent tissue paper using forceps. The membrane was then placed over top of a flat silicon O-ring (20 mm I D . , 2 mm thickness) on a flat surface. A second identical O-ring was placed over top of the membrane. The O-rings were used to suspend the membrane surface and enclose the area to be impregnated by Nafion. Nation solution (0.5% diluted from a 5% stock using 50/50 dH 0/isopropyl alcohol) was cast onto wetted membranes in two applications of 2  100 uL with 30 minutes air drying time between applications. A micropipet was used to dispense the Nafion solution and care was taken to disperse the solution as evenly as possible over the entire surface. The membrane was allowed to air dry overnight after the second application followed by thermal curing in a dry heat oven at 120°C for 60 minutes. In addition to the Nafion-impregnated membrane material, 2 stainless steel wire mesh screens and 150 |im thick silicon-sheet O-rings were cut to 22 mm O.D. dimensions. The O-rings were cut to 19 mm I.D. Four O-rings in total were used to separate and seal each of the three layers. Layers were assembled in succession by first coating the outside edge of a layer with silastic diluted ca. 5-fold (with toluene to form a viscous yet fluid consistency) using a micropipet. After coating with silastic, layers were built on top of each other with one O-ring alternating in between each layer. The 3-layer assembly shown in Figure 2.7 was allowed to set overnight by sandwiching between two Teflon sheets and  49  compressing by placing a heavy object on top. After setting, the circular edge of the sealed layers was trimmed using a razor blade to remove excess silastic. A second coat of diluted silastic was added along the circular edge to ensure uniform sealing between all 3 layers followed by setting overnight. The end result was a well sealed membrane system which was supported on both sides by a wire mesh and hence able to protect the membrane from ballooning outwards when subjected to a pressure differential. An in-house manufactured stainless steel collar shown in Figure 2.7 was designed to fit over the original hard plastic rim to provide a flat surface for the assembled membrane layers to sit on. Assembly of the probe for use involved placing the membrane sandwich flat over top of the stainless steel collar and compressing the membrane into place with an in-house designed stainless steel cap (Figure 2.7) that screwed onto the tip of the probe body.  50  Stainless steel threaded screw cap  Brass collar  Draped membrane Stainless steel wire Hard plastic rim  Stainless steel probe body tip  F i g u r e 2.6 Original membrane cartridge configuration  Stainless steel mesh  51  Re-designed stainless steel threaded screw cap  Stainless steel mesh Silicon O-rings f==  Cellulose triacetate membrane  Stainless steel collar  Hard plastic rim  Stainless steel probe body tip  F i g u r e 2.7 Re-designed membrane cartridge illustrating the membrane sandwich and stainless steel collar  52 2.2.2  Continuous  Perfusion  The peristaltic pump used to introduce and remove conjugate protein from the sensor chamber was also used for continuous perfusion. Flow rates at a given setting (2.1, 1.5, 1.0, 0.7, 0.5 ml/min) were maintained for 48 hours during benchtop experiments. A beaker containing 100 ml of PBS solution and 10 m M in glucose was monitored continuously. A flow rate of 0.70 ml/minute was found to be the minimum necessary for preventing gas bubble accumulation. The reagent flow system in Figure 2.4 was modified slightly by incorporating larger (2 L ) reagent bottles as reservoirs for both phosphate buffer and waste collection. Use of the minimum flow rate required refilling the PBS reservoir and draining the waste reservoir every 48 hours.  Two-litre aliquots of PBS  buffer were autoclaved, cooled rapidly and transferred into the reagent reservoir bottle for use. 2.2.3  Conjugation  of CBD and Glucose  Oxidase  Cellulose binding domain from Cellulomonas fimi exoglucanase ( C B D , M W = cex  11 kD) was obtained from recombinant expression in E. coli. Cultivation, harvesting and purification of C B D  c e x  is described elsewhere (Ong et al, 1993). The following protocol  is optimized from a suggested protocol provided by the crosslinker supplier. From a 15 mg/ml stock of C B D  c e x  solution in d H 0 , 14.2 mg was added with 14.1 mg (25x molar 2  excess) of SMCC crosslinker, and total volume was made up to 3 ml with potassium phosphate buffer pH 7.5 (KP-7).  The reaction was stirred for 60 minutes at room  temperature followed by gel filtration using a 10 ml Sephadex G-25 column (Pharmacia Biotech, Baie d'Urfe, Quebec, Canada) to remove excess crosslinker.  The reaction  53 mixture was loaded onto the column and column void volume was collected by adding and eluting 4 ml KP-7. Simultaneous with the activation of C B D , G O X was modified with Traut's cex  reagent in a second vessel. The modification time was critical and 40 minutes resulted in the highest specific activity of final conjugate preparations (data not shown). Glucose oxidase (Type X-S from Aspergillus niger, 118-225 units/mg protein from Sigma Chemicals) was modified by dissolving 100 mg and 13.3 mg of Traut's reagent (25x molar excess) using T E A buffer (50 m M triethanolamine, 150 m M sodium chloride and 1 m M E D T A , pH 8) to a total volume of 3 ml. The reaction was stirred at room temperature for 40 minutes followed by gel filtration using a second 10 ml Sephadex G-25 column to remove excess Traut's reagent. The reaction was loaded onto the column and the column void volume was collected by adding and eluting 4 ml KP-7. Activated C B D  c e x  and modified G O X were combined (total volume ~ 8 ml) and  rotated at 4°C overnight to effect conjugation.  Crude conjugate solution was then  dialyzed using an Amicon stir cell (Amicon Canada Ltd., Oakville, Ontario) with a 30kd M W C O non-cellulose membrane by passing through 6 x 50 ml KP-7 buffer at 4°C to remove excess unconjugated C B D . The resultant conjugate was diluted to 50 ml with cex  KP-7 and partitioned into 5 x 10 ml aliquots in 15 ml conical centrifuge tubes. Into each tube, 0.75 gm washed Avicel cellulose powder (type PH101, F M C International Food and Pharmaceutical Products, Cork, Ireland) was added, and crude conjugate was purified in one step by adsorption to Avicel by rotating overnight at 4°C.  54 Working volumes of conjugate were obtained by eluting aliquots of conjugate protein as follows: stored Avicel bound conjugate was centrifiiged at 3000 rpm for 5 minutes and supernatant containing non-bound protein was decanted.  Avicel bound  conjugate was washed several times with 2 x 15 ml KP-7 + 0.1M NaCl followed by 8 x 15 ml washes with 5 m M KP-7. Between washes, Avicel samples were centrifiiged for 5 minutes at 3000 rpm and supernatant was discarded.  The final elution proceeded by  incubation and rotation of washed Avicel with 10 ml d H 0 at 4 °C overnight. Eluted 2  conjugate was removed as supernatant after final centrifiigation.  Conjugate was  replenished to 0.1 M NaCl and 50 mM KP-7 to make the working solution. 2.2.3.1  Total Protein and Enzyme  Activity  Conjugate protein was assayed for total protein and glucose oxidase activity to determine specific activity. Activity was measured using a modified Sigma G O X activity assay method (Sigma Chemicals). Briefly, the assay is a colorimetric and kinetic method based upon spectrophotometric determination of the rate of production of oxidized odianisidine reagent (Sigma) at 490 nM. Oxidized o-dianisidine (red) arises from the reduced form (colorless) by electron transfer to  H2O2  and this reaction is catalyzed by  peroxidase which is present in non-limiting concentration.  The rate of o-dianisidine  oxidation is proportional to the rate of H 0 production by GOX. Unconjugated G O X in 2  2  the range of 0-2.5 units/ml (based on activity specifications of the supplier) was used for the standard curve in the assay. Ten-uL samples were combined in 96 well microtiter plates with 10 ul peroxidase solution (Sigma chemicals 60 purpurogallin units/ml in d H 0 ) 2  and 240 ul o-dianisidine (66 ag/ml in 50mM phosphate buffer). At time zero, 50 uL of  55 10% glucose was quickly added to triplicate sample wells with a multichannel micropipet. The microtiter plate was then placed in the plate reader, mixed automatically and monitored for rate of increase in  A490  over 4 minutes.  Absorbance measurements for  activity and total protein assays were carried out on a Molecular Devices Vmax kinetic microtiter plate reader (Menlo Park, California) at room temperature. Total protein was measured using the BioRad assay (BioRad Laboratories, Missisauga, Ontario, Canada). Samples were assayed in triplicate on a 96-well microtiter plate with G O X standards spanning 0-21 (xg/ml concentration. One-hundred ul of sample was combined with 10 ul of Coomassie blue dye concentrate and incubated for 15 minutes at room temperature.  Endpoint absorbance was measured at 595 n M . Results from the  BioRad method for some samples were also compared against two other methods, absorbance at 280 n M and the Bicinchoninic acid (BCA) method (Pierce). 2.2.3.2 Protein  Gels and Western  Blots  Conjugate protein was analyzed by SDS-PAGE gel electrophoresis to visualize any physical differences between conjugate preparations using the two different crosslinkers. Conjugate samples were run in denaturing, non-reducing 7.5% gels with SDS using standard protocols (Harris and Angal, 1989). For western blots, hybridized membranes were probed with rabbit anti-CBD  cex  primary antibody diluted 10,000X with PBS/TWEEN. Secondary antibody was goat antirabbit IgG H R P conjugate (Gibco) diluted 7000X with PBS/TWEEN. Hybridized blots were developed with the Amersham chemiluminescent reagent system (Amersham International P L C , Amersham, UK).  56 2.2.4 Lower  Limit of  Detection  Experiments were carried out on the bench top using a 150 ml beaker and an air driven magnetic stirrer (HP89055A, Hewlett-Packard Canada Ltd., Missisauga, Ontario, Canada). For each experiment, 100 ml of glucose-free D M E M media was used and the sensor was dipped to the 70 ml mark on the beaker. Six calibration curves were made by sequential additions of 0.1 M p-D-Glucose to media at 37°C. Each of the 6 calibrations was carried out with a new loading of enzyme conjugate so that an average response value for a typical loading could be obtained. The standard deviation (SD) of the background noise was determined by recording the sensor output in standard D M E M containing zero glucose and recording the magnitude of peaks and troughs in the baseline signal over a period of time. More than 125 peak/trough values were obtained over 20 minutes of measurement and used to calculate the SD. 2.2.5 Binding  Capacity  of Cellulose  Matrix  Whatman cellulose was prepared by pre-cutting sheets into small pieces and then soaking in PBS to make up a 4 mg/ml stock solution. The cellulose mixture was vortexed frequently to effect break up of the fibers into a homogeneous slurry. Avicel cellulose sorbent (FMC chemicals, Cork, Ireland, 40-75 u M presieved particle size) was prepared as a 4 mg/ml stock solution in PBS. 2.2.5. 1 CBD Binding to Cellulose All binding experiments were carried out in 1.5 ml Eppendorf tubes. Each C B D  c e x  ligand concentration in the isotherm was made in triplicate by combining 1 mg Avicel or Whatman cellulose, an appropriate volume of C B D c e protein ligand (from a 20 m M stock X  57 in 50 m M KP-7), d H 0 and 50 ul of 1 M KP-7 to 3 Eppendorf tubes. The final solution 2  in the Eppendorf tubes contained 1 ml of  CBDcex  cellulose sorbent. Eleven concentration points of  protein ligand in 50 m M KP-7 and 1 mg CBDcex  ligand were used and spanned 0-  10 u M . A 4th Eppendorf tube with no cellulose sorbent and was included with each triplicate to represent the starting free ligand concentration at each point. Tubes were rotated at 2 rpm for 2 hours at 37 °C followed by centrifugation at 15,000 rpm for 10 minutes to pellet cellulose sorbent in the triplicate tubes. samples  and corresponding sorbent-free  concentration point) were measured for  A 8o 2  sample (4  th  Supernatant from triplicate eppendorf tube from each  to determine the remaining free concentration  and the starting free concentration respectively.  Absorbance from the  sorbent-free  samples also provided the standard curve to convert absorbance values to mass quantities of C B D . Bound C B D was determined by difference between starting free concentration and remaining free concentration. Results were fitted with the Langmuir model. 2.2.5.2  Conjugate  Binding to  Cellulose  Avicel cellulose was used in the first binding isotherm. Each experiment required twelve Eppendorf tubes and all tubes were precoated with B S A by rotating overnight at 4°C with 500 ul of a 4 mg/ml solution in PBS followed by aspiration to remove excess volume. Fifty-ul (0.2 mg) aliquots of Avicel cellulose sorbent stock solution in PBS were transferred into each of the 12 Eppendorf tubes. Conjugate protein ligand (2 mg/ml stock solution) was aliquoted into 12 Eppendorf tubes to concentrations spanning 0-120 ug/ml ligand in 2 fold serial dilution's starting at the highest concentration. The total volume in each Eppendorf tube was 500 ul.  The highest concentration was chosen to be  58 approximately 10X the saturation capacity of CBDcex alone {ca. 40 ug/mg Avicel, Ong et al, 1993) in order to ensure that saturation would be reached. The 2-fold increment was chosen to ensure sufficient spread in the data points and to have enough concentrations in the lower range so that a significant initial depletion would take place. This initial slope gives an indication of the affinity of the protein for the sorbent. For each isotherm experiment, two sets of 12 Eppendorf tubes were used. Both sets were identical except that the second set, containing no sorbent, represented starting free protein concentration for each point on the isotherm curve and provided a standard curve for conversion of protein mass from assayed G O X activity. Binding (in the first set of tubes) was initiated by incubating both sets of Eppendorf tubes at 37° C overnight and rotating at 2 rpm for mixing and attainment of binding equilibrium. After binding, the first set of Eppendorf tubes were microcentrifuged at 13,500 rpm for 10 minutes to pellet the Avicel sorbent. Ten-ul samples of supernatant from all Eppendorf tubes in both sets were assayed for activity to determine unbound or remaining free  conjugate  protein  concentration (the first set) and the starting free protein concentration (the second set). The assay utilized the modified Sigma method as described in 2.2.3.1. Residual activity in the second set (containing no cellulose sorbent) was correlated with the known concentration of conjugate added, yielding a linear standard curve. From this standard curve, the concentrations of residual free conjugate in the first sorbent containing set were determined, enabling calculation of free (second set) and bound (second set minus the first set of eppendorf tubes) protein concentrations. Plotted results of the free protein versus the bound protein served as the basis for analysis using the Langmuir model.  59 This same experiment was repeated using #1 grade Whatman cellulose paper sorbent (Maidstone, England). One-hundred ul (0.4 mg) aliquots of Whatman cellulose stock were transferred into each of 12 Eppendorf tubes for binding experiments.  The  same range of starting conjugate protein ligand concentrations were used as for Avicel experiments. 2.2.5.3 Desorption  over  Time  The possibility of conjugate protein desorbing from the cellulose after adsorption was investigated.  Supernatant in Eppendorf tubes containing Avicel or Whatman  saturated with conjugate was removed followed by washing of the cellulose. Cellulose was washed 3 X using 500 ul PBS buffer each time and microcentrifuging at 13,000 rpm for 5 minutes between washes. After the third wash, cellulose was resuspended in 500 ul PBS and rotated overnight at 37°C along with the second set of Eppendorf tubes used for calculation of staring free concentrations.  Afterwards, cellulose containing Eppendorf  tubes were microcentrifuged and triplicate samples of the supernatant from each tube were assayed for desorption by measuring glucose oxidase activity. The assay for desorption was carried out again after repeating the washing and overnight incubation process a second time.  2.3 Results and Discussion 2.3.1 Membrane  Cartridge  The integrity of the new membrane cartridge configuration was tested by carrying out an E. coli challenge experiment. Following steam sterilization of the bioreactor with  60 the sensor, E. coli injected into the inner perfusable space of the sensor did not contaminate the bioreactor contents after 48 hours of incubation. The incidence and frequency of bacterial contamination decreased drastically with the re-designed membrane configuration. This provided further evidence that the original membrane cartridge design was the predominant factor leading to culture contamination.  2.3.2 Continuous  Perfusion  The presence of gas bubbles in the sensor chamber was unexpected and their source remains unclear. Nevertheless, incorporation of a continuous perfusion protocol proved effective in removing the problem. A possible explanation for bubble formation is dissolved gases in either the culture medium and/or the PBS perfusion buffer coming out of solution. There are several potential driving forces which would result in such a process. Firstly, a temperature difference exists between PBS buffer in the R l reservoir bottle (ca. room temperature), the interior of the sensor chamber and the culture medium (37 °C). Given the dependence of gas solubility in solution on temperature, there may be sufficient driving force to cause gases to come out of PBS solution and form bubbles in the sensor chamber.  Secondly, the concentration of dissolved C O 2 is high in culture  medium as it is used in conjunction with bicarbonate to buffer the medium. The p H in the micro-environment around the electrode surface may be lowered as a result of the production of gluconic acid from enzymatic oxidation of glucose (Eqn. 1.1, 1.6).  This  localized drop in p H would definitely shift the equilibrium of CC>2-bicarbonate balance such that C 0 would accumulate and perhaps come out of solution. Either of the above 2  61 scenarios would result in a steady accumulation of dissolved gas in the sensor chamber until it came out of solution. At this point, bubbles could collect by rising to the highest point or adsorb to the sides of the chamber or become trapped in the cellulose matrix. A l l scenarios could result in alteration of the mass transport characteristics of either glucose diffusion into the enzyme layer or hydrogen peroxide diffusion to the electrode surface. Detailed investigation of the mechanism(s) for gas bubble formation would require analysis of gas bubble composition however this was beyond the scope of the present work. The flow rate for continuous perfusion was determined based on the minimum required to remove/prevent bubble formation and presumably, the slower the flow rate, the less impact there would be on signal sensitivity (recall from 2.1.2 that Enfors et al, used perfusion to extend linear dynamic range). Reduced signal sensitivity did not create a problem for glucose monitoring at low concentrations using S M C C conjugate in the prototype. 2.3.3 Improved  Conjugate  Using  a Hetero-Bifunctional  Crosslinker  The new conjugate was characterized in terms of resultant specific activity and size distribution on SDS-PAGE.  These results are compared to those obtained from the  previous glutaraldehyde conjugation method. 2.3.3.1 Comparison  of Resultant  Specific  Activities  Table 3.1 shows specific activity data compiled from several preparations of conjugate using both crosslinkers. Statistical analysis of the resultant specific activity data (last column) in Table 2.1 indicates significant differences between the two types of  62 conjugate.  A one tailed t-test (Zar, 1984) comparing the mean activity of S M C C  conjugate with the mean activity of original glutaraldehyde conjugate indicates with greater than 99% confidence, at least 20% higher average specific activity results with S M C C conjugate (/o.oi, (i), B  =  -2.65, t - -6.03). The data suggest that the crosslinking  process using S M C C was probably less disruptive on G O X structure as compared to using glutaraldehyde and this is reflected in the preservation of a higher percentage of native enzyme activity in the final conjugate. The method of crosslinking presumably has a large effect on the enzyme activity and this has been described by Wingard (1983) who showed that the thickness and glutaraldehyde content of crosslinked membranes can have a significant effect on the response (activity) of an enzyme sensor configuration.  63 T a b l e 2.1 Compilation of Resultant Specific Activities From Several Conjugation Batches Batch  crosslinker used  starting GOX activity (units/mg protein)  total protein using BioRad assay (ug/ml)  assayed activity (units/ml)  specific activity (units/mg protein)  % starting activity retained (normalized to batch 10)  1  glutaraldehyde  118  28.5  1.57  55.2  0.34  2  glutaraldehyde  118  44.9  3.31  74  0.46  3  glutaraldehyde  118  26.6  2.02  76  0.45  4  glutaraldehyde  118  30  1.95  65  0.40  5  glutaraldehyde  118  44  2.31  52  0.32  6  glutaraldehyde  185  37  3.3  89  0.35  7  glutaraldehyde  185  92.3  10  108.5  0.43  8  SMCC  185  92  13.9  150  0.59  9  SMCC  185  2.3  0.42  182  0.72  10  SMCC  185  38.5  9.7  251  1  11  SMCC  185  70  13.8  197  0.78  12  SMCC  185  80  16.4  205  0.82  13  SMCC  118  300  31.35  104  0.65  14  SMCC  118  250  23  92  0.57  15  SMCC  225  122  17  139  0.46  64 Initial observation of the data in Table 2.1 indicates that a surprisingly small amount (if any) of starting enzyme activity appears to be lost with SMCC conjugate (data in bold) as a result of crosslinking.  In some of the cases, specific activity actually  increased. This result appears puzzling but is an artifact arising from the method of total protein assay used. The BioRad method yields total protein values for conjugate protein that are consistently 2-5 fold lower than other methods such as B C A or A go which were 2  used for comparison (data not shown). The discrepancy in results between methods is not surprising and is due to the differential dependence of each method on protein structure and amino acid composition. The A o method relies on the presence of aromatic amino 28  acids while the B C A method detects peptide nitrogen's and is rather insensitive to amino acid composition (Harris, 1989). The BioRad method used in this work relies on the binding of Coomassie blue dye to basic amino acids such as arginine's and to aromatic amino acids.  Other workers have determined that C B D  c e x  protein is not accurately  quantifiable by the BioRad method but instead results in assay results far below expected values (Doheny & Johannson, personal communication).  As a result, the C B D  c e x  component of conjugate molecules is likely being underestimated by the BioRad assay. It is possible that the inability to accurately quantify the C B D  c e x  component of the conjugate  molecule is a principle reason for the lower assay results using the BioRad method, which in turn, would artificially inflate calculated values of specific activity. It could be argued that the BioRad assay, yielding smaller values of total protein and hence resulting in artificially inflated specific activity values, is not the best method to determine specific activity values. However, total protein data from the SMCC batches  65  and 2 of 7 glutaraldehyde batches in Table 2.1 (all of which were produced in this body of work) utilized the same BioRad protein assay method that was used for 5 of the 7 glutaraldehyde batches in Table 2.1 originating from previous work (Phelps, 1992). As a result, the same error associated with using the BioRad method is actually present in the total protein data for all batches in Table 2.1. The relative differences in resultant specific activity between the two types of conjugation methods are thus comparable. If the BCA and A280 methods had been used to determine total protein for all of the batches in Table 2.1, all batches would have higher resulting total protein values. This would reduce the calculated specific activity values for all batches but retain the relative differences between glutaraldehyde and SMCC conjugate. The new conjugation method was developed with the objective of decreasing the degree of intra and intermolecular crosslinking so that batch preparations would have higher resultant specific activity (Table 2.1), be less heterogeneous in terms of size homogeneity (next section) and conjugate composition.  The composition was not  investigated in detail but given the hetero-bifiinctional nature of the crosslinker, both intramolecular crosslinking within an individual protein and intermolecular crosslinking between two identical proteins (CBDcex-CBDcex or GOX-GOX) was prevented. The only crosslinks that could be formed are those between different proteins. As a result, the composition of conjugate molecules was greatly affected and likely in a manner which would reduce composition heterogeneity since fewer possible crosslink combinations may be formed between the two proteins.  66 The S M C C crosslinker used in the new conjugation method also formed more stable bonds than glutaraldehyde crosslinker. The NHS-ester end of the crosslinker forms a covalent amide bond linkage with primary amines while the maleimide end forms stable thioether linkages with sulfhydryl groups which are not cleaved under physiological conditions (Mattson et al., 1993).  Increased bond stability would increase conjugate  molecule structure stability which in turn would contribute to longer term sensor signal stability since the likelihood of immobilized conjugate molecules falling apart is decreased. 2.3.3.2  Gel  Electrophoresis  Figure 2.8 shows the results of gel electrophoresis (SDS-PAGE) and a western blot of the G O X - C B D c e x conjugates prepared using the two different crosslinkers. The C B D , M W 11 kd, (Ong et al., 1993) control runs close to the bottom in lane 2. The cex  glucose oxidase control in lane 3 shows a band between 97.4 and 68 kd, due to the dissociation of GOX, M W 155 kd, (Wilson and Turner, 1992) into 2 equal subunits under denaturing conditions. Avicel purified samples of both conjugates appear in lane 4 and 5 at ~ 180 kd and higher. The higher molecular weight of the conjugate is attributed to the crosslinking process and it is apparent that definite differences exist between the two types of conjugate proteins in terms of distributions in conjugate sizes.  The glutaraldehyde  conjugate in Gel A , lane 4 runs to approximately 200 kd and higher whereas S M C C conjugate in Gel A, lane 5 is more homogeneous in size, forming a tighter band at slightly less than 200 kd.  Increased homogeneity may be the result of decreased intra and  intermolecular crosslinking using SMCC crosslinker. It is also possible that the bands visualized in Figure 2.8A,B lanes 4 and 5 represent only a fraction of the conjugate protein  67 loaded onto the gel; that some of the conjugate polymers were too large to be fractionated in a 7.5% gel. As a result, a definitive increase in conjugate homogeneity due to S M C C crosslinking cannot be concluded. However, the use of S M C C crosslinker may lead to a population of concatameric conjugate molecules which differ in molecular weight by the size of glucose oxidase molecules. Such concatamers may arise from a reduction in the number of types of allowable crosslinks as a result of using SMCC versus glutaraldehyde. The two types of conjugate protein in lanes 4 and 5 of Figure 4.8 A also appear to differ in degrees of dissociation to glucose oxidase subunits under denaturing conditions. The differences may be attributed to the stability of crosslink bonds. Under the stress of the heat denaturation step (to coat protein samples with SDS), the Schiff base crosslink bonds formed with glutaraldehyde crosslinker are likely to be broken more readily than the bonds formed using S M C C crosslinker which as previously mentioned, are very stable. The net result would be a larger proportion of conjugate molecules that have dissociated to the glucose oxidase subunits when glutaraldehyde crosslinker is used and this is especially apparent in the western blot in Figure 2.8B. In summary, decreasing the amount of intra and intermolecular crosslinking as well as increasing the stability of the formed crosslink bonds using S M C C crosslinker resulted in a number of beneficial effects on conjugate protein including higher amount of resultant specific activity and increased structural stability.  68  F i g u r e 2 . 8 A SDS-PAGE gel electrophoresis of two different GOX-CBD conjugates. Gel is stained with Coomassie brilliant blue. All lanes were loaded with 10 ug protein. Lane: 1. Protein size standards; 2. Purified C B D c e control; 3. Glucose oxidase control; 4. Avicel-purified glutaraldehyde conjugate; 5. Avicel-purified SMCC conjugate. 2 . 8 B Western blot of same samples in gel A, using r a b b i t - a C B D c e x primary antibody for detection. X  69 2.3.4 Lower  Limit  of  Detection  The results o f sensor sensitivity experiments at l o w glucose concentrations are shown in Figure 2.9. This figure illustrates the magnitude o f sensor responses to additions o f small quantities o f glucose to D M E M solution. The L L D o f the prototype sensor using S M C C conjugate is significantly lower than with glutaraldehyde conjugate. This was the result o f higher sensitivity o f the sensor which in turn is a reflection o f the higher specific activity o f the enzyme conjugate. The L L D is in the range o f 10-50 u M glucose which is well in the range o f that necessary for monitoring and or controlling the glucose concentration in a glucose limited fed-batch culture (Hayter et al, 1992; Kurokawa et al, 1994). The L L D is also more than an order o f magnitude lower than that for a typical commercial bench top analyzer such as the Beckman analyzer (Fullerton, C A ) which can measure with reasonable accuracy to ca. 0.6 m M . The L L D in Figure 2.9 is at a level which is at least 5-fold lower than the lowest concentrations controlled at in the literature (1 m M ) using an H P L C for detection (Kurokawa et al, 1994).  70  Glucose Concentration (uM)  F i g u r e 2 . 9 Comparison of resultant sensor signal sensitivity using glutaraldehyde conjugate (-0-) and SMCC conjugate (--•--) protein. The dotted line near the origin defines the boundary of the inset figure showing the best-fit line of the data and an improved lower limit of detection (LLD) as a result of using SMCC conjugate. The 3o line represents three standard deviations of the noise above the average background sensor signal and is the minimum measurable value. Experiments were carried out in DMEM at 37°C on the benchtop.  71  2.3.5 Binding  Capacity  of Cellulose  Matrix  Isotherm data in Figures 2.10-2.11 was fitted to the Langmuir equation by nonlinear curve fitting using a Levenberg-Marquardt curve-fitting algorithm (Microcal Software Inc., M A . ) .  The resulting kinetic constants N and K and 95% confidence 0  a  intervals are summarized in Table 2.2. The standard deviation (s) of the regression is also included in the legends for Figures 2.10-2.11.  T a b l e 2.2 Binding isotherm parameters based on mass concentration of ligand Ka (mg/ml)-  No (ug /mg)  0.094 ±0.01  37.08 ±1.9  CBD ex on Whatman cellulose  0.11 ±0.008  33.10 ±1.4  S M C C conjugate on Avicel cellulose  0.11 ±0.02  22.78 ±1.2  S M C C conjugate on Whatman cellulose  0.10 ±0.006  14.52 ±0.3  1  CBD  c e x  on Avicel cellulose  C  Conjugate protein has an affinity value, K , for both Avicel and Whatman cellulose a  which is not significantly different from that of C B D  c e x  alone. This result likely indicates  that the crosslinking process with SMCC has not grossly disrupted C B D cellulose binding properties. CBDcex  c e x  structure or its  When comparing N between the conjugate protein and Q  using either types of cellulose in Table 3.1, values appear to be roughly 2-fold  smaller for the conjugate which is intuitively expected given the much greater size of conjugate molecules compared to  CBDcex-  Large and complex conjugate molecules would  likely present steric hindrances which would directly affect cellulose binding capacity. The  72 ca. 50% difference in N of Whatman cellulose as compared to Avicel for either conjugate a  protein or C B D  c e x  in Table 2.2 is likely due to differences in structure between the two  types of cellulose. Highly crystalline or highly ordered cellulose fibers such as B M C C and Valonia (from V. ventricosa) are known to have higher binding capacities for C B D  c e x  (Jervis, unpublished results). Regenerated cellulose preparations such as Whatman have more amorphous, inaccessible and non-crystalline regions as compared to Avicel cellulose; hence the lower N values. 0  Another comparison of the binding properties of C B D  c e x  versus  conjugate  molecules to different cellulose surfaces could be made based on molar quantities of conjugate protein if the molecular weight of the conjugate was known. However, the heterogeneous nature of conjugate molecules in terms of structure, composition and populations (recall from section 2.3.3.2 that conjugate in SDS-PAGE gels runs at -180 kd and higher) does not enable determination of the molecular weight or even an approximate average value. 2. 3.5.1 Desorption  over Time  Desorption experiments were carried out using Eppendorf tubes containing Avicel and Whatman cellulose sorbent saturated with adsorbed conjugate protein.  The  experiments revealed a small percentage of non-binding conjugate molecules for both Avicel and Whatman cellulose. Assay after the first desorption incubation revealed a small amount of detectable enzyme activity corresponding to ca. 3% percentage of bound mass in Whatman cellulose and ca. 5% of bound mass in Avicel cellulose. Assay for desorption  73 after the second overnight incubation showed insignificant levels of enzyme activity being released. The small portion of non-binding conjugate protein that is present appears to slowly desorb over time. This finding is significant and has implications for long-term biosensor signal stability. If conjugate protein is desorbing slowly over time off of the cellulose, this undoubtedly contributes to decreases in long-term signal stability. In the present case, only a small percentage of bound protein mass (3-5%) is desorbed and only during the first 15 hours.  Further desorption appears to stop or at least become  undetectable. This non-binding or desorbing population, when accounted for, has the net effect of shifting the binding isotherm curves in Figure 2.10-2.11 rightwards away from the ordinate (to reflect the corrected higher value of F) and downwards toward the abscissa (to reflect the corrected lower value of T). The data in Figures 2.10-2.11 was corrected for the 3-5% non-binding or desorbing population and thus reflects the described curve shift. Several possibilities exist to explain the presence of a small non-binding and/or desorbing population of conjugate molecules. population of C B D  c e x  The conjugation may affect a small  molecules such that they will have altered or lost their ability to bind  cellulose thereby leading to its desorption.  CBD  c e x  is known to bind strongly and  essentially irreversibly to B M C C cellulose (Creagh et al, 1996). The cellulose matrix being porous, may physically trap some of the large conjugate molecules as opposed to binding or anchoring them onto a cellulose surface. Over time, it is conceivable that some of these trapped conjugate molecules will become untrapped and be released into solution.  74 A small amount of unconjugated G O X could also be present in the mixture which associates non-specifically with the cellulose. It is also possible that some conjugate molecules are falling apart due to unstable crosslinking.  This explanation does not  however seem likely given the stable bonds formed by the NHS ester and maleimide ends of the crosslinker. Such a scenario is more likely with glutaraldehyde conjugate and has been discussed in 2.1.2. 2.3.5.2  Saturation  of  Cellulose  Experiments were done to determine if a longer incubation period would result in higher bound activity. Pre-cut Whatman cellulose sorbent circles designated for use in the sensor were incubated overnight at 4 °C with conjugate protein (100 ug/mi) and then placed inside the sensor chamber for use. The inner chamber was flushed for 7.5 minutes as described in the loading protocol followed by lowering the electrode. Calibration of the probe did not result in higher signal sensitivity as compared to the normal exposure time of 5 minutes for the conjugate during loading. This result indicates that the exposure time used is sufficient to attain saturation and binding equilibrium. Other work with C B D  c e x  indicates that binding is fast and complete within 0.2 minutes of exposure to Avicel (Gilkes et al., 1992).  The concentration of working solutions of conjugate was kept  sufficiently high (>100 ug/ml) in order to ensure saturation concentration conditions. The binding isotherm data are the first of their kind for characterizing the binding behavior of a heterogeneous chemical fusion of C B D  c e x  with another protein. Although  the Langmuir model assumes a homogeneous ligand species, the isotherm data appears to fit the equation quite well.  The binding characteristics of CBDcex and whole cellulase  75  enzymes have been measured (Ong, 1993; Gilkes et al, 1992; Jervis et al., 1996; Creagh et al, 1997) along with an A B G - C B D genetic fusion (Ong et al., 1991) but as yet no published data exists to describe the binding kinetics of chemical fusions with C B D . cex  These initial results from the adsorption isotherm experiments may prove to be important because chemical fusion is arguably, in some cases, the only practical way to link two different proteins together such that both partners retain their biological activity. Undoubtedly, there will continue to be increasing numbers of biotechnological applications for protein fusions and, as such, ability to describe the binding behavior of chemical fusions based on the Langmuir model is significant.  76  Figure 2 . 1 0 Langmuir  Binding Isotherm of C B D Isotherm  yields  A/ =38.1 0  cex  to 1 ) Avicel sorbent ( • ) , fitting with the  ug/mg,  Ka=0.094  (mg/ml)' , 1  s=1.7; 2 )  Whatman cellulose sorbent ( • ) , A/ =33.1 ug/mg, Ka=0.056 (mg/ml)" , s=1.2. 1  0  77  0  5  10 15 20 25 30 35 40 45 Free Ligand Concentration (ug/ml)  50  55  F i g u r e 2.11 Binding isotherm of 1) SMCC conjugate protein to Avicel cellulose sorbent ( • ) ,  /V =22.8 ug/mg, K.=0.11 (mg/ml)- , s=0.7; 2) SMCC conjugate 1  0  protein to Whatman cellulose ( O ) , /V =14.5 (ug/mg)" , Ka=0.10 (mg/ml)" , s=0.2. 1  0  1  78  C H A P T E R  3  Development of a Model CHO Cell Line 3.1 Introduction This chapter describes efforts to develop a model animal cell line for growth and glucose monitoring in a bioreactor using the sensor prototype. Chinese Hamster Ovary (CHO) cells, DHFR" (Crl-9096, ATCC), were transfected with a pNUT construct (Palmiter et al, 1987) expressing a His6-tagged derivative of the human clotting enzyme Factor X (E2FX), herein referred to as H6E2FX (Guarna et al, 1996). C H O cells were chosen as the host because of their abilities to grow in suspension which would enable relatively easy large scale growth in a bioreactor and also because they are able to perform complex post-translational modifications required to produce the biologically active protein.  SV40 early promoter  „  P  K  nl  I  ,  mMT1 Promoter  Mutant DHFR cDNA  H6E2FX cDNA HBV3'  hGH 3'  Figure 3.1 Schematic diagram of pNUT vector harboring H6E2FX cDNA with expression regulated by the metallothionien promoter. Also shown is the mutant DHFR gene. Reprinted in part with permission from C.H. Fann and M. Guarna, 1996.  79  3.2 Materials and Methods 3.2.1  Transfection  All cell culture reagents were from Life Technologies (Gaithersburg, M D , U S A ) unless otherwise noted.  H6E2FX-pNUT plasmid D N A was obtained from miniprep  purification of an available E. coli stock using the QIAGEN™ mini prep system (Qiagen Industries, CA). Purified plasmid D N A was quantified by nucleic acid absorbance at 260 nm, aliquoted into 20 ug portions in Eppendorf tubes, precipitated with ethanol and stored at -20°C until use.  For each transfection, 20 ug aliquots of precipitated D N A were  dissolved in 1 ml H B S buffer, pH 6.95 (HEPES buffered saline: 137 m M NaCl, 5 m M KC1, 0.7 m M N a H P 0 , 6 m M dextrose, 21 m M HEPES) and 0.125 M CaCl . 2  4  2  Transfections were carried out in 6-well plates with CHO cells at l x l 0 cells/well growing s  in complete medium containing D M E M , 5% Newborn Calf Serum, 0.1 m M non-essential amino acids and hypoxanthine-thymidine supplement (0.1 and 0.016 m M respectively). Cell counts were done by Trypan Blue dye exclusion using a Fuchs Rosenthal Hemocytometer. Cells were incubated in 6 well plates with complete medium (5 ml/well) at 37 °C overnight to permit attachment.  Transfections were started by aseptically  transferring dissolved 20 ug D N A aliquots dropwise into wells containing attached cells and complete medium using a sterile pasteur pipette. Plates were swirled during dropwise addition to effect mixing. Attached cells were exposed to the foreign D N A solution for 3.5 hours followed by aspiration to remove all liquid supernatant.  The cells were  replenished with complete medium (5 ml/well) and allowed to recover from the  80 transfection with further incubation for 15 hours. Control transfections were performed with the pNUT vector without an insert and HBS buffer with no D N A . After the recovery period, culture medium was exchanged with selection medium to begin the selection process for positive transfectants. Selection medium was comprised of D M E M , 5% dialyzed serum, non-essential amino acids and 0.05 m M methotrexate, a 4amino analog of folic acid. Selection medium was changed every 3 to 4 days and selection pressure was maintained for over 6 weeks. At the end of this period approximately 30 colonies were isolated. Surviving colonies were transferred to smaller wells in 24 well plates as follows: selection medium in the 6 well plates was aspirated away and sterile stainless steel cloning rings lined with sterile vacuum grease on the bottom edge were carefully placed over marked colonies. Into each ring, two drops of trypsin solution was added and left for 5 minutes. A single drop of serum was then added for trypsin neutralization. Cells were detached and mixed into serum solution by aspiration with a pasteur pipette. The cells were transferred to a 24 well plate and supplemented with 1 ml complete non-selective medium. After a 7 day period, 23 colonies were established in the 24 well plates and the medium was exchanged with low protein Ultroser G (UG) medium (complete medium with 1% v/v U G replacing 5% NCS). Cells were grown with 1 ml U G medium for 72 hours to obtain E2H6FX protein samples for analysis by SDS-PAGE and western blot. 3.2.2  Gel  Electrophoresis  One-ml of U G samples were prepared for Coomassie blue SDS-PAGE by microcentrifuging to remove debris and then 20 uL were boiled for 2 minutes with 5 uL  81  5X loading buffer containing SDS, bromophenol blue and glycerol. The positive control was 1 ug E2H6FX protein and the negative control was fresh U G protein free medium. Samples and standards were loaded onto 10% denaturing, non-reducing SDS-PAGE gels and run using a BioRad protein gel apparatus (BioRad Laboratories, Hercules, C A , USA) at 150 V for 60 minutes. After electrophoresis, gels were stained with Coomassie brilliant blue R-250 dye for 30 minutes followed by destain and preservation. 3.2.3  Western  Blots  Samples identical to those used in Coomassie stained gels were prepared as follows for western blots. Ten-ul samples consisting of 1 ul U G sample, 2 ul 5x running buffer and 7 |il d H 0 were loaded into each well. The positive control consisted of 1 ng 2  E 2 H 6 F X protein, negative control was 10 ul U G medium and 10 ul prestained markers were loaded. Samples were loaded onto SDS-PAGE and run as for the Coomassie blue gel.  After electrophoresis, the gel was placed on a custom cut P V D F membrane  (Millipore Corporation, Bedford, M A , USA), pretreated by soaking in methanol for 15 seconds followed by soaking in I X CAPS buffer (3-[cyclohexylamino]-l-propanesulfonic acid (Sigma), 10% methanol, 90% water).  The P A G E gel-PVDF membrane were  sandwiched between layers of Whatman paper absorbent and the entire sandwich was placed in a plastic holder and subjected to an electric field at maximum voltage, limiting current at 0.5 A for 20 minutes to effect transfer of protein bands to P V D F membrane. After transfer, the membrane was prepared for antibody detection by preblocking for one hour with gentle shaking in 30 ml preblock buffer (3% B S A in P B S / T W E E N 20). This was followed by several washes with PBS/TWEEN and then addition of 20 ml  82 primary antibody anti F X (diluted 7000X to 20 ml with P B S / T W E E N , 0.5% BSA) and incubated for 60 minutes on a shaker. The primary Ab was decanted followed by several washes with P B S / T W E E N and addition of secondary Ab, goat anti-rabbit IgG H R P conjugate (Amersham) diluted 10,000x in 15 ml PBS/TWEEN).  Secondary Ab was  incubated on a shaker for 60 minutes followed by several washings. Antibody hybridized membranes were developed using the E C L chemiluminescent development system (Amersham). 3.2.4  Limiting  Dilution,  Dot  Blot  ELISA  and  Specific  Protein  Production  Rate  Two positive clones identified by western blots were subcloned by limiting dilution to obtain pure cultures (subclones) derived from a single cell as follows: positive clones in 24 well plates were scaled up in T-flask cultures. Confluent cultures were trypsinized, counted and then serially diluted into 10 ml portions with concentrations of 1, 5, 10, 20 and 50 cells/100 ul. Into each well of a 96 well plate, 100 ul of a particular dilution was transferred.  In total, 5 plates were used with each plate having 96 replicates of a  particular dilution. Diluted cells in each of the wells were incubated with 0.5 ml selection medium to permit attachment. A dissection microscope was used to identify and select wells containing single cells. After 21 days, 127 subclones were obtained. The medium in each of these wells was exchanged with U G medium and incubated for 72 hours. Dot blot E L I S A ' s on the harvested U G samples from all subclones were done to confirm H6E2FX expression and estimate specific production rate.  83 Dot blots were carried out according to manufacturers instructions (Gibco) using P V D F membrane material and the same antibody reagents as described above for western blots.  Dot blot intensities on the developed autoradiograghs were scanned by  densitometry (Molecular Dynamics, CA.). The 8 subclones with the darkest spot intensities were chosen and analyzed for residual glucose using the Beckman analyzer to calculate consumed glucose. The ratio of glucose consumed to spot intensity was then calculated. Specific protein production rates of the 8 subclones was determined by calculating the ratio of protein produced to the log mean change in cell number over a time period as follows: 2 x 10 cells/well from each of the 8 subclones were cultured in 6-well plates for s  24 hours with 5 ml/well of serum containing medium to allow attachment. The medium was then exchanged for 3 ml/well of U G medium and cells were cultured for 72 hours. Culture medium was harvested and attached cells were trypsinized and counted. The log mean change in cell numbers over the 72 hour period was calculated. Harvested culture medium was assayed by E L I S A on a 96 well plate using the same antibody reagents as for western blot. The E L I S A protocol consisted of coating each of the 96 wells with 1 ug primary antibody with coating buffer (15 m M Na2CC>3, 34.9 m M NaHCC«3) containing 1 ug antibody/100 ul overnight at 4°C. Coating buffer was removed and wells were washed 2x with PBS/TWEEN. Wells were then preblocked with 100 ul of 1% B S A for 1 hour at 37°C followed by removal and wash. Standards of E2H6FX protein antigen spanning 0-1.0 ng/well were prepared using 5% spent U G medium (spent U G medium diluted in PBS/TWEEN) for dilution.  Culture medium  84 samples containing protein antigen from the 8 clones were diluted 20-fold with 5% spent U G medium. Both standards and sample antigen were prepared to make up 1 ml and 100 were transferred to triplicate wells. Protein antigen was allowed to bind for 1 hour at 37°C followed by removal and 2 X wash. Secondary antibody was diluted 2000 fold with PB S/TWEEN/B SA and 100 ul was added to each well and incubated at 37°C for 1 hour. Secondary antibody was removed followed by washing 3X with P B S / T W E E N and 2 X with PBS. One hundred-ul T M B Horseradish Peroxidase substrate (Kirkegaard and Perry Laboratories Inc., Gaithersburg, M D , USA) was added to each well and incubated at 37°C for 30 minutes followed by addition of 100 ul 1 M phosphoric acid to halt the reaction and final reading at 450 nm. The mass of E2H6FX protein antigen in samples was derived from the standard curve allowing calculation of the specific protein production rate for each subclone. 3.2.5 Adaptation  to Serum-free  Medium  The 3 subclone clones with the lowest ratio of glucose consumed to spot intensity were chosen for adaptation to serum-free medium. Serum containing cultures of these three clones were passaged into 2.5%, 1%, 0.5% and then 0% serum containing medium for 4 days at each step in T-flasks. The total adaptation time was 7 weeks. During the adaptation period, two sets of T-flasks were used for each of the 3 subclones. The sets were identical except that one set utilized medium with methotrexate for selection pressure. It was anticipated that adaptation with selection pressure may be too stressful for the cells.  85  3.3 Results and Discussion Coomassie blue stained SDS-PAGE gel and corresponding western blot for the first 7 surviving clones is shown in Figure 3.2 and 3.3. None of the seven clones displayed positive expression. Western blots of the remaining 16 clones is shown in Figure 3.4A,B and they show two positive colonies out of a total of 23. The H6E2FX positive control band in lane 1 in both gels run at the approximate expected size of 66 kD and the 2 positive clones show uniform bands. The 2 positive colonies were subjected to limiting dilution for 4 weeks. One hundred and twenty seven subclones were isolated and assayed for expression by dot blot ELISA. A range of spot intensities in the dot blot results were seen and more than half of all the subclones displayed little to no spot intensity. This result reflects the heterogeneous expression levels of individual cells in a mixed population before limiting dilution. Culture supernatant from 8 subclones with the darkest spots were analyzed for residual glucose using the Beckman analyzer. A dark spot intensity does not necessarily preclude higher expression level than a second lighter spot if the culture glucose for both subclones differed substantially.  As a result, it was necessary to correlate glucose  consumed with spot intensity. The specific H6E2FX production rate of the 8 subclones as determined by E L I S A was in the range of 600-900 ng protein/10 cells/day. 6  86  1 2 3 4 5 6 7 8 9 200kD 116kD 97 kD 66.2 kD  ^  10  fttitil  ^ir^iiiiMii'iiiiTt  1  45kD 31 kD 21.5 kD  F i g u r e 3.1 SDS-PAGE gel electrophoresis of Ultroser G (UG) medium supernatant from 7 colonies of CHO DHFR" cells transfected with a pNUTH6E2FX plasmid construct. Gel is stained with Coomassie brilliant blue. Lane 1. Protein size markers (10 ug); 2. UG medium negative control; 3. H6E2FX positive control (5 ug); 4-10. UG supernatant from surviving colonies.  1 2 3 4 5 6 7 8 9  10  200kD  F i g u r e 3.2 Western blot of same samples in Gel A using rabbit anti-FX primary antibody for detection.  87  A  1  2  3  4  5  6  7  8  9  B  1  2  3  4  5  6  7  8  9  F i g u r e 3 . 3 A and B Western blot of Ultroser G (UG) medium supernatant from 16 colonies of CHO DHFR cells transfected with a pNUT-H6E2FX plasmid construct. Lane 1 in both gels are 1 ng H6E2FX protein positive control; 2. In both gels is UG media negative control; Lane 8 in gel A and lane 5 in gel B are positive transfectants. The spotty background is due to use of paper towel absorbent instead of fine tissue for blot drying membrane prior to chemiluminescent development. -  88  The three subclones with the lowest ratio of glucose consumed to spot intensity were adapted to serum-free medium (SFM) for 7 weeks. After this period, none of the 3 subclones produced measurable amounts of H6E2FX protein as judged by E L I S A assay on 96-well plates.  The loss of expression indicated that the cell specific protein  production rate had dropped to undetectable levels due to some form of instability as a result of adaptation to SFM. Instability of heterologous protein expression in animal cells has been documented and characterized in the literature by a number of investigators.  There appear to be  different molecular mechanisms that result in observed instabilities. The mechanisms are cell line dependent. In addition, not all transfected cell lines display unstable expression. A priori, instability must arise as a result of one or both of two possible processes: decreases in the coordinated expression of the gene(s) of interest or decreases in gene copy number of integrated DNA. In mammalian cells, resistance to the folate analogue methotrexate arises from amplification of the dihydrofolate reductase gene (Kaufman et al., 1981).  During  amplification, flanking regions to either side of the enzyme will also be amplified. Thus by linking a gene of interest (in this case H6E2FX) to the original D H F R cDNA, H6E2FX will also be overexpressed in cells. This phenomenon has been exploited for the overproduction of heterologous proteins in animal cells. Weidle et al., (1988) reported that in three C H O cell lines excreting human t-PA, recombinant protein production takes place in a stable manner when selective pressure with M T X is maintained. In the absence of M T X ,  89  both the number of integrated expression constructs and the size of the amplified regions decreases. The mechanisms of increased stability in the presence of selection pressure is not well understood and some of the data in the literature conflicts. Numberg et al, (1978) reported that stably amplified DHFR genes in a MTX-resistant C H O cell line were localized in homogeneously staining regions (HSR) on a single long marker chromosome. These regions stained as well defined bands on host chromosomes.  Amplified  heterologous genes have been shown to be unstable when the selective pressure is removed (Pallavicini et al, 1990; Weidle et al, 1988). The size and distribution of the H S R regions decreases dramatically when the cells are propagated in the absence of selective pressure (Pallavicini et al, 1990). Kaufman et al. (1979) reported that unstably transfected D H F R genes are integrated in a manner such that they do not form HSR's but instead are associated with small paired chromosomal elements called "double-minute chromosomes".  These chromosomes are acentromeric and do not participate in equal  segregation during mitosis. Thus, the DHFR genes are largely extra chromosomal and can become unequally distributed in daughter cells during mitosis leading to loss from the cell population in subsequent generations.  In such cell lines, approximately 50% of the  amplified D H F R genes are lost from a cell population in as few as 20 cell doublings in MTX-free medium. Declining expression levels of monoclonal antibodies in hybridomas and transfectomas have also been documented.  General decline in antibody secretion  levels over long-term cultures have been observed and were attributed to loss of heavychain expression (Bae etal, 1995; Couture & Heath, 1995).  90  In some cases even in the presence of selection pressure, the D H F R levels and hence heterologous protein expression levels have been observed to decline over time and without detection of double minute chromosomes, (Morrison et al, 1997; Cossons et al., 1991). In other cell lines, the stable amplification of D H F R and high copy number of D H F R genes persists for up to 40 cell doublings even in the absence of M T X selection pressure.  These conflicting results indicate that resistance to M T X may arise by  mechanisms other than DHFR amplification and so integration stability is not necessarily dependent on M T X selection pressure. In fact, declines in the overall secretion rates of interferon-y in CHO cells with M T X after 60 days have been reported and attributed to complex factors other than changes in the copy number of D H F R genes (Cossons et al., 1991). In the present case, the rapid decline of H6E2FX expression to undetectable levels in S F M took place over the course of 50 days or rough equivalent of, in cell population doublings(average doubling time of recombinant animal cells is ca. 19 hours). This rapid and unexpected loss of expression in serum-free medium occurred with or without M T X selection pressure. Unstable H6E2FX expression was not M T X dependent so a number of alternative explanations may account for the observed instability. The stress of adaptation to S F M may have contributed to or induced a change in gene copy number and/or expression levels. The 3 subclones experienced substantial drops in viability (>90% to 60%) during the first 20 days or so of adaptation before recovery back to near the original levels.  91 The subclones were derived in M T X containing medium and it has been reported that progeny cells from a M T X selection process tend to become progressively heterogeneous with respect to gene copy number (Kaufman et al, 1981). It is possible that during S F M adaptation, populations of cells with decreased gene copy number (low or non-producers) could outgrow high gene copy number cells.  Growth rates of  transfected cells are often substantially lower than their wildtype counterparts.  Morrison  et al. (1997) reported a bimodal distribution of CHO cells expressing wildtype GP63 surface antigen protein. The culture was characterized by a progressive increase in nonproducer population after 51 days culture and accompanied by a decline in the expressing population. This group reported a 12-20% higher growth rate advantage for the nonproducers that resulted in 83% loss in expressing population after 138 days.  The  adaptation period for the CHO cells in the present case may have resulted in the enrichment of low or non-producers in the population of cells. Another possibility is that the 3 subclones used for adaptation to S F M (based on highest expression levels) were chosen on an incorrect basis.  Selection of the highest  producing subclones in serum containing medium for adaptation to S F M does not necessarily assure continuation of the same expression level in SFM, (Glacken, personal communication). Inclusion of more subclones for adaptation process, including those with lower specific protein production rates, might have increased the likelihood of obtaining a stable cell line. Finally, the apparent loss of expression might be explained by a lack of inducer for the metallothionein promoter in pNUT.  This promoter requires 20-100 um zinc for  92 activation and so perhaps the zinc concentration in S F M was not high enough.  To  investigate this further, the SFM adaptation process was carried out a second time with the three chosen subclones (cultured in serum containing conditions), this time using S F M supplemented with 100 um zinc sulfate. However, there was no detectable expression as determined by ELISA. The 3 SFM-adapted subclones which had lost expression during the first adaptation experiment were also cultured further in the zinc supplemented S F M for 5 weeks. None of these subclones exhibited detectable expression of H6E2FX. In conclusion, the mechanism(s) responsible for the observed unstable expression was not determined. One or more of the above described mechanisms could have taken place. A cell line producing H6E2FX was not essential to the main thrust of this thesis. As a result, further work to establish a stable cell line was not undertaken. Hybridoma cell cultures (Chapter 4) were used in place of the failed model cell line.  93  C H A P T E R  4  Long-Term Signal Stability 4.1 Introduction This chapter presents results from three major experiments designed to characterize the performance and long-term signal stability of the prototype during continuous use under animal cell culture conditions. Most of the work in chapter 2 was motivated by results from the first major experiment in this chapter (4.3.1). The main objective of the first experiment was to characterize and evaluate the performance of the unmodified prototype during long-term usage in cell culture.  The characterization  included an evaluation of signal instability. This signal stability was compared with the long-term stability of native G O X , soluble conjugate and immobilized conjugate in solution. The second experiment attempted to evaluate some of the steps taken towards addressing problems elucidated in the first experiment. Part of the evaluation included effects on long-term signal stability and determination of an average decay constant in culture conditions containing no cells.  The third experiment utilized the modified  prototype to investigate long-term signal stability in a second hybridoma culture.  4.2 Materials and Methods 4.2.1  Cell Line and Cultivation  Conditions  All reagents were obtained from Life Technologies unless otherwise noted. The 2E11 (BRC) and P9 (Terry Fox Laboratories) hybridoma cell lines were used for experiments 1 and 3, respectively. Cells were cultured in glucose-free D M E M (sterile  94  filtered into the bioreactor through a 0.2 u M high volume sterile filtration membrane from Gelman Sciences, MI) with 5% NCS. The culture was fed glucose in accordance with protocols appropriate to each experiment. Cell counts were done as described previously in Chapter 3 using the Trypan Blue dye exclusion method. Off-line glucose measurements were made using a benchtop Beckman analyzer (Beckman Instruments, Inc., Fullerton, C A , USA). All three experiments were carried out in a steam sterilizable 2.5 L Chemap Type S G bioreactor (Chemap A G , Switzerland) with a 2 L working volume.  The  temperature (37 °C) and stir rate (50 rpm) were controlled using a Chemap 3000 Series base unit and FZ3000 control unit. The pH was maintained at 7.2 and D O in a range between 75-100% air saturation by manually controlled aeration of the headspace with 5% C02/balance air in the first experiment. In the same manner, D O was maintained at 100% in the second experiment. In the third experiment, the DO was maintained at 100% air saturation using pure oxygen and computer control (Applikon). Culture parameters were monitored for p H with an Ingold Infit 764-50 p H electrode and dissolved oxygen with an Ingold sterilizable electrode.  Parameter outputs were datalogged using the Genesis  control series software (Iconics, Foxborough, M A , USA). The bioreactor was inoculated at starting cell concentrations of 2 x 10 cells/ml 5  from  500 ml spinner flask cultures (Bellco Laboratories) in hybridoma culture  experiments. Medium was supplemented with l g / L Pluronic F-68 (Sigma Chemicals). Hybridoma cell cultures were maintained as a sequence batch in which 90% of the fermenter volume was drained away at the end of an entire growth cycle (ca. 5 days). This was followed by volume replacement with an equivalent volume of fresh sterile  95 medium containing no glucose. The remaining 10% volume in the bioreactor served as a starting inoculum to initiate the next sequence of the batch culture. The sequence batch technique was used as a simple method for extending the duration of a culture thus permitting a first evaluation of prototype performance over the longer term.  Both  hybridoma cultures were fed manually by pulse additions of glucose from a 1M stock in a feed bottle linked with silicon tubing and a peristaltic pump. 4.2.2  Glucose  Sensor  Calibration  and Data  Analysis  The prototype sensor was used to monitor the glucose concentration. Set-up and operation of the sensor prototype was carried out as described in chapter 2. Calibrations were made by making a series of known glucose additions and recording the sensor output. The background or baseline signal was measured at several points during cultures by recording the sensor signal in the absence of loaded enzyme conjugate (usually after elution of a spent loading). The background was subtracted from the calibration data points and the plotted results displayed Michaelis-Menten type kinetics as shown in Figure 4.1. Non-linear curve fitting (Microcal Inc.) was used to derive the kinetic parameters, K'  m  and Vmax from the data.  The kinetic parameters were used to transform the raw  sensor output in units of nA into the glucose concentration readings in m M between calibrations and these values were compared to off-line assay results. 4.2.3  Stability  of Conjugate  in  Solution  Solutions of native G O X (100 ug/ml), soluble glutaraldehyde conjugate (100 ug/ml), soluble S M C C conjugate (100 ug/ml), Avicel-immobilized glutaraldehyde conjugate (20 ul diluted 10X), and Avicel-immobilized S M C C conjugate (20 ul diluted  96 10X) in D M E M with 2 m M glucose were prepared.  The mass amount of Avicel-  immobilized conjugate used could not be quantified because the cellulose particles interfered with the absorbance readings during the BioRad assay. Knowledge of the exact amount of conjugate immobilized, however, was not necessary for the experiment. The decay in G O X activity of these samples over time at 37 °C was determined using the previously described Sigma assay. Three 200 ul aliquots of each different type of sample was incubated in 96-well plates. At pre-determined time intervals, the microtiter plate was removed from the 37°C incubator and 10-ul samples were withdrawn for assay.  The  assay results were compiled and for each different sample, assayed activities were normalized to the activity at time zero.  4.3 Results and Discussion 4.3.1  First Long-Term  Experiment  with 2E11 Hybridoma  Culture  This culture was maintained for over 580 hours through 5 sequence batches. Glutaraldehyde crosslinked enzyme conjugate was loaded into the sensor, eluted and reloaded 8 times and the sensor was calibrated 19 times. The primary objective of this experiment was to become familiar with the challenges and problems that lay ahead for successful implementation of this sensor technology for large-scale animal cell culture applications.  Prior to this experiment, the longest period of continuous use for this  prototype had been 16.5 hours in microbial culture (Phelps, 1995). The Specific objectives of this experiment were to evaluate: 1. effectiveness of the signal regeneration system using the inner perfusable space of the prototype, sensor response time, effects of longer term exposure to culture conditions  97 on hardware components including the membrane and platinum electrode, signal sensitivity to glucose and oxygen; 2. evaluate the behavior of the sensor signal over the longer term in a real culture with respect to behavior of the background and also stability.  The sensor signal was calibrated and re-calibrated multiple times in order to observe the stability/repeatability of kinetic parameters. Each loading was calibrated and re-calibrated at least once. On 4 different occasions throughout the course of the culture, regeneration of the sensor with fresh enzyme conjugate did not result in a signal that was responsive to glucose additions. Instead, the sensor signal remained at a constant baseline value. The reasons for this were not apparent and may have been due to inefficiencies of conjugate binding to the cellulose matrix as a result of entrapped gas bubbles (as described in chapter 2). Nevertheless, on occasions of failed signal regeneration, the signal was successfully restored after elution and reloading a second time. Except for brief periods of unresponsiveness (again, possibly due to entrapped bubbles), the prototype was sensitive to glucose for the duration of the culture and the measured background current remained stable indicating that factors such as protein fouling of the membrane and/or electrochemical system including the platinum surface had not adversely affected sensor performance.  The response time of the sensor was not  significantly changed with increased exposure, remaining essentially constant between 1015 minutes, further supporting the inference that detrimental fouling of the membrane or electrode did not occur. Finally, no significant change in the background signal was  98  observed for the entire course of the culture, which varied only slightly around an average 4 nA value. This indicated that electrochemically interfering species were not fouling the platinum electrode appreciably and that the mass transport characteristics of the membrane were not grossly perturbed. The sensitivity of the sensor to glucose was adequate for this culture experiment (Figure 4.1) but higher sensitivity is required for reliable measurement at concentrations below 1 mM. The sensor signal was found to be highly sensitive to fluctuations in D O as expected based on earlier work by Phelps (1994) and others (Cleland & Enfors, 1983, 1984a,b; Cass et al, 1984; Claremont et al, 1986). The performance of amperometric biosensors utilizing oxidoreductases is known to have a dependence on D O since oxygen is a co-substrate of the enzymatic reaction. Efforts have been made to remove oxygen dependence by replacement of oxygen with chemical mediators to act as terminal electron acceptors, (Cass et al, 1984) or by introducing continuous oxygenated buffer flow into the enzyme chamber, (Cleland et al, 1983, 1984). No efforts were made to reduce or remove oxygen dependency as this was beyond the scope of the project. A continuous buffer flow was incorporated but was motivated for a different reason as described in chapter 2. Stability of the long term signal behavior of the signal was investigated over the course of the eighth and final loading of enzyme conjugate (530-580 hours) because the D O was controlled most stringently at 90% air saturation during this period. The loaded enzyme conjugate in the sensor was calibrated 3 times (Figure 4.1A-C) and are designated A - C in the order that they were done.  The pattern of data points exhibit Michaelis-  99  Menten kinetic behavior; this model has been used successfully to describe previous calibrations of this prototype in microbial culture, (Phelps et al, 1993, 1994). It can be seen that  Vma  X  progressively decreases with each calibration in Figure 4.1. This decrease  directly reflects the decay of enzyme activity over time. The data was also plotted and fitted using the Hanes model equation in Figure 4.2. The two functions resulted in slightly different derived values for the kinetic parameters because each function emphasizes different aspects of the data. The data from both analyses are summarized in Table 4.1.  Table 4.1 Comparison of extracted kinetic parameters using Michaelis-Menten and Hanes model analysis.  Michaelis-Menten plot  Hanes plot  K ' (mM)  A 0.82  B 0.89  C 2.71  A 0.84  B 1.43  C 2.3  V  30.97  17.31  11.25  31.15  18.72  18.76  m  m a x  (nA)  Kinetic parameters derived from the Hanes plot provided the best fit to the data and were used transform the raw sensor output values (in nA) into the glucose concentration signal (in mM) using the following equation:  [S]  (/-/)/ \  (4.1)  o / max  This equation is essentially the Michaelis-Menten equation rearranged to solve for [S] in m M with  V  and  Vmax  replaced by  I  and  Imax  in nA while  background current which is subtracted from the raw output.  I  0  is the baseline or  100 The results of the transformed on-line sensor signal are shown in Figure 4.3 superimposed with the off-line glucose analysis (using the Beckman analyzer).  The  transformed signal falls off rapidly after a calibration and does not follow the Beckman values (true glucose concentration) very closely. The shape of the transformed signal between calibrations suggested a smooth and consistent pattern of decay which was found to fit a first-order decay equation. The ratio of the sensor output to the Beckman value at each time point was calculated and plotted in Figure 4.4. This plot illustrates the decay rate independent of the changing glucose concentration environment. The three sets of data points clustered at y=l correspond to the calibration points while the remainder fall in a decay pattern that is consistent with an assumption of first-order kinetics. A n ideal enzyme sensor which exhibited zero decay and followed glucose concentration as well as the off-line method would always give a ratio value of 1 in such a plot. Curve fitting of the two regions of decay in Figure 4.4 yields two similar decay constants (3.27 and 3.55 hr' ) as expected since the data correspond to a single loading of enzyme conjugate. 1  The extracted decay constant values were used to correct the transformed sensor signal between calibrations in Figure 4.3 to give the results shown in Figure 4.5. The corrected on-line sensor output follows the off-line Beckman readings more closely, as expected since the off-line values themselves were used to derive the decay constant. However, results of this analysis are significant because they indicate that the pattern of decay is uniform and predictable using a relatively simple mathematical transformation. This suggests that, if a reasonably precise value for the decay constant can be determined (either on-line or by adequate off-line characterization), then real-time correction of the  101  sensor may be practical. Such on-line correction should at least reduce the cumulative error in the sensor signal due to loss of enzymatic activity, thus potentially extending the time periods between necessary calibrations. Towards this end, one objective of the second major experiment was to investigate the possibility of extracting a consistent decay constant of the sensor signal under controlled glucose concentration conditions. The decay behavior of conjugate protein and native G O X in solution is shown in Figure 4.6 and can be compared to the signal decay due to conjugate in the sensor in Figure 4.4. Data in Figure 4.6 shows that a pattern of first order decay takes place during an initial period (ca. 15 hours) as was observed with the conjugate immobilized in the sensor in Figure 4.4 although the rate of decay is slower. Beyond this initial period, a higher order process becomes apparent in Figure 4.6 and is characterized by a more stable plateau-like region followed by a more rapid decay. This experiment was also designed to provide a first glimpse of differences, if any, in conjugate stability in a soluble and immobilized state as a result of using SMCC crosslinker versus glutaraldehyde crosslinker. ; The stability curves in Figure 4.6 suggest that no significant differences in stability between S M C C conjugate and glutaraldehyde conjugate can be observed under these conditions. It was anticipated that SMCC crosslinking would result in conjugate with higher structural stability than glutaraldehyde crosslinking (recall from chapter 2 that glutaraldehyde forms Schiff bases). This result, were it true, would not manifest itself in Figure 4.6 because results in this experiment reflect only the catalytic stability (elaborated on in the next section as thermal deactivation) of the conjugate and not the structural stability. In contrast, the sensor signal reflects both the structural and thermal stability of  102 conjugate protein. Any immobilized conjugate in the sensor that is released as a result of structural instability over time directly affects the sensor signal. Glutaraldehyde conjugate, being less structurally stable than SMCC conjugate, breaks apart (ie. Crosslinks are broken and unconjugated enzyme drifts away) over time in the soluble conjugate experiments. However, these unconjugated portions are still detected because the assayed samples contain all of the unconjugated portions. The observed signal decay between calibrations in Figure 4.4 is drastic (>75% activity loss after 15 hours), especially in comparison with Figure 4.6, and represents another example of the major limitation and obstacle which has rendered enzyme based biosensors impractical for industrial use: signal instability and drift during long-term continuous use. Several possible explanations may account for the observed instability: 1) thermal deactivation, 2) accumulation of peroxide, 3) breakdown of immobilized glutaraldehyde conjugate and finally, 4) exposure to inhibitory agents in the growth medium. Published literature concerning the stability and mechanisms for loss of enzyme catalytic activity are based on the context of need for enzyme activity stabilization during industrial applications.  A number of investigators have modeled the deactivation of  enzyme activities as a function of thermal effects (thermal deactivation). The deactivation kinetics of soluble and immobilized P-glucosidase over a temperature range were studied by Aguado et al. (1995) and found to behave in a two step process with each step obeying first-order kinetics. These authors also noted that immobilization of the enzyme resulted in increased stability. The kinetic equations fit their experimental results with a mean  103 deviation of less than 10%, however they did not discuss the actual mechanism by which thermal energy affected P-glucosidase activity. Nikolova et al. (1997), using differential scanning calorimetry (DSC) data coupled with catalytic activity measurements reported that P-l,4-glycanases may be irreversibly deactivated in one of two possible ways: 1) through a low-temperature route (<50°C) characterized by first-order kinetics; or 2) through a high temperature route characterized by an initial reversible step followed by an irreversible step. D S C data provides a quantitative description of the gradual inactivation that takes place below 50°C. Collectively, this data suggest that thermal deactivation may play a role in explaining the observed signal instabilities in this experiment. The deactivation kinetics of G O X have been documented in the literature as well. The enzyme is susceptible to inactivation by hydrogen peroxide, a product of its own catalytic reaction.  Inactivation involves oxidation of some methionine residues, which  may reside close to the active site, to form methionine sulfoxide (Kleppe, 1966). The author observed that the reduced form of the enzyme is inactivated much more rapidly than the oxidized form. Krishnaswamy & Kittrell, (1978) were one of the first to describe the deactivation kinetics o f G O X upon exposure to peroxide using a simple first-order equation.  Their calculated kj values increased from 0.024-0.11 hr  -1  when G O X was  exposed to peroxide concentrations from 0-10 mM. However, the upper range is not relevant as peroxide concentrations would never accumulate to such levels in a sensor. The kd values are not directly comparable with the present system because the literature experiments were carried out with soluble enzyme (versus immobilized as in the sensor) at pH 5.5 and the configurations of the enzyme reactor are different.  Nevertheless, the  104 results indicate a direct relationship between peroxide concentration and deactivation rate. These authors also demonstrated that dissolved oxygen affects the deactivation kinetics although it was unclear what role oxygen played.  Either oxygen itself was directly  inhibitory or the catalytic activity of the enzyme was increased with higher DO. The stability of immobilized GOX is affected by peroxide more so during continuous use than in storage tests and removal of peroxide from the micro-environment would decrease the rate of deactivation (Greenfield et al., 1975). Collectively, these observations suggest that if peroxide accumulation takes place inside the sensor chamber of the prototype described here, it could have a large impact on the GOX signal stability. Accumulation is highly unlikely as it would manifest itself as an obvious upward drift in the electrode current over time during constant glucose concentration conditions. On the other hand, GOX used in biosensor applications by other groups reveal contrary results. When GOX is utilized in cylindric, needle-type amperometric electrodes designed for in-vivo patient use, they are not detectably affected by peroxide when measuring in the range of glucose concentrations found in a physiological system, ca. 4-9 mM (Woedtke et al., 1994). An external glucose concentration as high as 30 mM (upper linear range of their configuration) would result in a peroxide concentration of ca. 0.18 mM in the configuration used in these particular studies and this did not appear to damage the GOX.  The contradiction in results may be attributed to comparing the effects of  peroxide between immobilized and soluble enzyme. By the same token, deactivation rates  105  of enzymes between in-vivo and in-vitro environments are difficult to compare (Towe et al, 1996). The deactivation kinetics of G O X by peroxide has also been investigated by Malikkides & Weiland (1982). These workers concluded that there is great difficulty in determining and monitoring all of the variables that contributed to enzyme deactivation. This conclusion led Towe et al (1996) to state that a simple first-order equation was adequate for fitting experimental results of deactivation rates for an implantable glucose biosensor configuration. These authors reported derived decay constants in the range of 0.198 to 1.3 per day but they are not directly comparable to those derived in this work because of differences in sensor geometry and the dependence of catalytic half-life on the immobilization method. A third possible explanation for the observed instability in Figure 4.4 is the nature of the glutaraldehyde crosslinked enzyme conjugate.  The possibility of glutaraldehyde  crosslinked conjugate structures destabilizing and breaking apart during longer term use in the sensor prototype was described briefly in the previous section and in chapter 2. This possibility provided part of the motivation to redesign the conjugation protocol as described in chapter 2 Another possible source of the observed instability in Figure 4.4 arises from the membrane component of the sensor hardware.  A low level contamination became  apparent half way through the long-term culture. The infection was controllable using antibiotics, and at the time of the experiment, the membrane was not suspected to have been the source of contamination. In retrospect, however, it is possible that a microscopic  106  rupture of the membrane assembly was the source of contamination and lead to exposure of G O X enzyme to inhibitory agents in the medium, thus contributing significantly to the observed instability in Figure 4.4.  107  Glucose Concentration (mM)  F i g u r e 4.1 Plot of sensor output versus glucose concentration for three different calibrations of the same enzyme loading. 4.1 A Calibration #16; 4.1 B Calibration #17; 4.1 C Calibration #18.  108  1  2 3 4 5 Glucose Concentration (mM)  F i g u r e 4.2 Hanes plots of the data in Fig. 4.1.  109  Figure 4.3  Transformed sensor output compared against off-line values  obtained from Beckman analyzer. (•): off-line transformed sensor output (not corrected for decay).  Beckman  values;  (-0-):  110  2.00 1.75 J o) c  1.50H  J  1.25-)  CD  =  1.00-  o) 0.75H |  0.50 H  §  0.25  CO  \ x = 3.27h-i  H  0.00530  x = 3.55h-i  •—E3EI 540  550 560 Culture Time (hours)  570  580  F i g u r e 4.4 Extraction of first-order decay constant, x, for two periods of time between calibrations.  Ill  530  540  550  560  570  580  Culture Time (hours)  F i g u r e 4.5 Transformed sensor output corrected for decay using the extracted x value. ( • ) : off-line Beckman values; (-0-): transformed sensor output corrected for decay.  112  T  T  Native GOX  l— —I— —I—'—I 1  80  1  100 120 140 160  i— —i—•—i—•—r 1  >  Soluble Conjugate  <  "D <D N "ro E  " i -  i— —i— —r 1  0  C  20  1  40 60 80 100 120 140 160 ~~i— —i— 1  Immobilized Conjugate  —I—i—|—i—|—.—|—i—|—i—|—r—  20  40 60 80 100 120 140 160 Time (Hours)  F i g u r e 4 . 6 Stability of GOX and conjugate in solution. All samples were prepared in DMEM with 2 mM glucose and incubated at 37°C. 4 . 6 A soluble, native GOX; 4 . 6 B soluble conjugate: glutaraldehyde ( — • — ) and SMCC ( — • — ) compared with soluble GOX (•); 4 . 6 C Avicel-immobilized conjugate in solution: glutaraldehyde ( - 0 - ) and SMCC (-•-) compared with soluble GOX (•). Each data point is the average of triplicate sample readings. Error bars are the standard deviation of triplicates.  113  4.3.2  Second Glucose  Long-Term  Experiment:  Sensor  Performance  in a  Controlled  Environment  The results of the first experiment led to modifications to the membrane design, the fermenter hardware configuration, redesign of the conjugation protocol, and to the incorporation of continuous perfusion. This work was necessary to develop a functional prototype for long-term use in the second and third experiment. The primary objective of this second long term experiment was to determine the effects of the combined modifications on the sensor signal and secondly, to determine if a reasonably consistent value of the decay constant under controlled conditions could be obtained. As previously mentioned, prior knowledge of the decay constant would reduce cumulative drift in the signal over time between calibrations and lessen the dependence on off-line values. The third objective was to determine if long-term stability would be improved upon and, to what extent from the first experiment, as a result of incorporation of all the above described modifications. This experiment consisted of several bioreactor trials with the sensor. The sensor was loaded with fresh S M C C enzyme conjugate and the resulting signal output in sterile D M E M was recorded over at least 15 hours in 3 trials at 1 m M glucose and 2 trials at 5 m M glucose. Different glucose concentrations were used to determine if the decay rate was concentration dependent. Replicate trials enabled evaluation of the reproducibility of decay patterns between different loadings of SMCC enzyme conjugate. Figure 4.7A compares the sensor signal decay pattern using glutaraldehyde crosslinked conjugate protein from the first experiment with the new S M C C conjugate. A  114 very substantial increase in longer term stability was achieved.  Between 75-85% of  starting activity, was retained, even after 40 hours of continuous use. This contrasts with the original glutaraldehyde conjugate which displayed >80 % loss in activity after only ca. 15 hours. The close overlap of decay patterns for all 7 enzyme loadings in Figure 4.7 reflect the consistency and reproducibility of decay behavior from one loading to the next. This result indicates that the average decay constant (x) for different loadings may be consistent for at least 24 hours, (Figure 4.7B). As a result, the decay-corrected sensor output signal can be relied upon for a longer length of time (compared to results in the first long-term experiment) before recalibration is required. A number of issues become apparent when comparing the decay rates between the improved S M C C conjugate and the original glutaraldehyde conjugate in Figure 4.7A. The asymptotic values to which the signal appears to approach with time is different for the two conjugates. Theoretically, one would expect that this asymptotic value would be the background signal of the sensor in the absence of loaded enzyme. The data in Figure 4.7 A indicate that the S M C C conjugate signal falls off rapidly, mimicking a first-order kinetic process, but then plateaus at some value near 0.75 after ca. 30 hours. In contrast, the glutaraldehyde conjugate approaches a value much closer to zero or the baseline after only ca. 18 hours. The biphasic pattern for SMCC conjugate signal suggests that a higherorder mechanism may be involved. This pattern of decay was also observed over a similar time period in the soluble stability data in Figure 4.6.  A similar plot of enzyme  denaturation kinetics for both a and B-amylase was also found (Ray et al, 1994) to be  115  distinctly biphasic, with an initial rapid phase, apparently obeying first-order kinetics, followed by a slower phase. S M C C signal data in Figure 4.7A is expanded in Figure 4.7B to illustrate in greater detail, the decay behavior over the first 24 hr period. The deactivation rate during this time period follows first-order kinetics well so the x values from S M C C conjugate can be compared with x values from glutaraldehyde conjugate which span approximately the same 24 hr period. In Figure 4.7B, a first-order function representing an average value of all the S M C C conjugate signal decay curves was modeled.  This curve indicates an  average x value of 9 hr' and is roughly 2.5X larger than the x values for glutaraldehyde 1  conjugate. The observed 3-fold increase in sensor signal stability evident in Figure 4.7A appears at first to conflict with the data shown in Figure 4.6, which showed no significant differences in catalytic stability between glutaraldehyde and S M C C conjugate.  Recall,  however, that the data in Figure 4.6 could not reveal differences between the two types of conjugates in terms of structural stability, because of the sampling procedure. The signal stability depicted in Figure 4. 7 does include structural information, since only enzyme that remain immobilized within the cellulose matrix can contribute to the signal. Given the insignificant differences in catalytic stability, coupled with the large differences in sensor signal stability arising from the two conjugate types, it is evident that the increased stability observed here is most likely the result of increased structural stability of the S M C C conjugate.  Furthermore, these results indicate that structural stability plays a  significant role in contributing to sensor signal stability.  116  1  1  Bsaii"""  I  I  1  I  1  ,,B  •••••••••••••••» .B«.ii.ii. . . . . B  B  B  B  x =  3  Q.<  T=  o  I  n  0  "D CD N  I  1  5  1  1  1  1  1  B  r -  .  r a  . . B  B  - 9 h-1  - 3 . 4 h-1 1  1  1  1  1  >  1  '  1  1  1  r-  10 15 20 25 30 35 40 45 50 55  "co E i_  o  z  0.80 0.75  H T  0  2  1  i  4  1  —i— —i— —i— —i— —i—•—i—i—i—i—i—i—r 1  6  1  1  1  8 10 12 14 16 18 20 22 24 Time (hours)  F i g u r e 4 . 7 A Sensor signal decay curves of several different loadings of SMCC conjugate in the sensor. The longer term decay was investigated under controlled conditions with no cells and different concentrations of glucose. The SMCC data is compared with the original glutaraldehyde conjugate data from Experiment 1 in chapter 4. 4 . 6 B Data from Figure 4.6A concentrating on the first 24 hours showing that an average decay constant obtained after this time period is significantly lower than with the original glutaraldehyde conjugate. ( ) 1 mM glucose; (•) 5 mM glucose; ( ) glutaraldehyde conjugate.  117 4.3.3  Third Long-Term  Experiment  With P9 Hybridoma  Culture  Extraction of an improved decay constant in the second experiment led to the third bioreactor culture experiment with P9 hybridoma cells. The objective of this experiment was to utilize the modified prototype in a bioreactor culture and determine if observed improvement in signal stability in the second experiment could be reproduced in a culture containing cells. This culture experiment lasted ca. 200 hours as 2 sequence batches. Three successful loadings of SMCC conjugate in the sensor were obtained.  The  transformed sensor output results are compared with off-line values (Figures 4.8A-4.10A). Signal decay rates derived from a plot of the ratio values of the sensor output to the offline value is shown in Figures 4.8B-4.10B. The raw current output of the sensor for each of the three loadings is also shown in the scale on the right axis of Figure 4.8A-4.10A. The similarities in the range of sensor currents on the right axis for each of the three loadings is indicative of the reproducibility of conjugate enzyme loadings. On 4 occasions during this experiment, regeneration of the sensor signal by re-loading fresh conjugate did not result in a glucose responsive - signal. In addition, 18 hours after the final loading ;  (Figure 4.10), a hardware problem associated with the automatic temperature control of the culture took place. The temperature regulator drifted from the 37°C setpoint and very quickly rose up to 100°C, effectively destroying the culture in progress. In the final analysis of data from this experiment, observed decay behavior was unexpectedly and strikingly different for all three loadings. The signal in the first loading (Figure 4.8A) displayed remarkable stability and appeared not to require correction at all for at least 22 hours. The sensor signal in this loading also correctly tracked the addition  118 of glucose aliquots after 8 hours. After ca. 22 hours, the sensor signal began to drop off very rapidly to the background level.  The second loading, in contrast with the first,  exhibited a first-order decay rate similar to that seen in the first experiment with glutaraldehyde conjugate (Figure 4.9). The third loading appeared to follow first-order decay as well but did not begin to decay as expected until after ca. 7 hours after loading (Figure 4.10).  The pattern displayed in the third loading has been observed in other  stability studies of immobilized enzymes (Towe et al., 1996). The explanations for inconsistent sensor signal behavior in this last experiment with hybridoma culture are not clear and indicate on-going and/or undetermined problems with the sensor.  One possible explanation is that having cells present extremely  complicates the dynamics of the measured matrix.  This complication may render it  difficult to measure reproducibly in such a matrix, unlike in the second experiment without cells.  A second possible explanation is that the population of conjugate molecules  immobilized to the cellulose matrix in a loading may have differential degrees of stability. Such a differential could conceivably arise as a result of the heterogeneous nature of chemical crosslinking. Another possibility for inconsistent sensor behavior relates to the efficiency of conjugate exposure to the cellulose matrix. The internal geometry of the sensor chamber may not be optimal in terms of solution mixing and flushing. In particular, the platinum electrode is raised during loading and elution but the raised distance is only ca. 3 mm. The close proximity of the electrode to the cellulose surface may affect the fluid dynamics during loading and elution protocols such that these protocols are not completed properly every time.  A final possibility is that minute gas bubbles are still  119 interfering with the sensor signal cannot be discounted as it was not proven that the problem was completely overcome by the incorporation of continuous perfusion. In any case, the results in this experiment indicate that a degree of increased stability has been attained as compared to the decay rates observed in the first hybridoma culture.  A number of unknown factors appeared to mask the observed increase in  stability. These factors may have been transient in nature given that the inconsistent signal behavior had not been observed before.  Collectively, the modifications described in  chapter 2 yielded a prototype that was functional for the long-term and displayed increased stability during animal cell culture use. Additional enzyme loadings in this third culture experiment, had they been possible, may have provided more insight into the apparent inconsistencies in signal behavior and provided further convincing evidence of increased stability.  120  4.0  • i • i • i • i • i • i • i • i • i • i • i •  3.5-1  130  -I 25 co CD  a) C/>  o o 3 O  ^3  i.o-l  - 20  §  - 15  g  - 10  5-  0.5 0.0  * i 0 2  1  i 4  1  i  1  i 6  i 8  1  i i i i i i 10 12 14 16 18 20  1  1  1  1  1  1  22  24  B  I  o o -•—< i_  1  i.o(5  i  1  0.8-  i  i • i • i <i  I  1  I  o—c>  OQ  0.6-  o  (j) 0.4rz  CD  CO  0.2-  -4—«  0.0  o o  CO  a:  O I  0  2  4  6  1  I  8  1  I  10  1  I  12  1  I  1  14  I  16  18  20  22  24  Time (Hours)  F i g u r e 4 . 8 A Transformed sensor output (•) for first 24 hours corresponding to the first SMCC enzyme conjugate loading and compared to off-line Beckman values (•). The equivalent raw current output of the sensor for this loading is shown in the right hand side scale; 4 . 8 B Ratio of sensor to off-line to extract signal decay pattern. The horizontal axis represents elapsed time following calibration of loading.  121  I  1  I  1  I  I  1  I  1  1  I  1  I  20 15  CD  V)  o  10 o •6" 5  5  0  i  2  0 10 12 14 16 18 20 22 24  • i • i • i • i • i • i •  4  6  8  i • i  i • i  x=  co  0.0  i • i • i • i • i • i  0  2  4  6  1  1  i • i • i  -5.18 h-""  i • i  1  i • i • i •  8 10 12 14 16 18 20 22 24  Time (Hours)  F i g u r e 4 . 9 A Transformed sensor output (•) for 48-64 hours corresponding to the 4th SMCC enzyme conjugate loading and compared to off-line Beckman values (•). The equivalent raw current output of the sensor for this loading is shown in the right hand side scale; 4 . 9 B Ratio of sensor to off-line to extract signal decay pattern. The horizontal axis represents elapsed time following calibration of loading.  122  0)  CD (fi  O  —\  O rz  o3 > 0  2  0  i 2  4  1  6  8  10  12 14 16  18 20 22  24  i • i • i • i • i • i • i • i • i • i 4 6 8 10 12 14 16 18 20 22  24  Time (Hours)  F i g u r e 4.1 OA Transformed sensor output (•) for 190-210 hours corresponding to the 6th SMCC enzyme conjugate loading and compared to off-line Beckman values (•). The equivalent raw current output of the sensor for this loading is shown in the right hand side scale; 4 . 1 0 B Ratio of sensor to off-line to extract signal decay pattern. The horizontal axis represents elapsed time following calibration of loading.  123  CHAPTER 5 Conclusions 5.1 Concluding Remarks Lack of long-term signal stability has been the main obstacle preventing the commercialization of on-line enzyme electrode biosensors.  Efforts to resolve this fact  provided impetus for the work presented in this thesis. Most of the work was focused on improving and characterizing an existing prototype in-situ glucose biosensor such that it would be functional and stable for longer term use in large scale animal cell culture applications. Towards this end, a number of objectives were achieved: 1. A n improved conjugation method using SMCC crosslinker to chemically link CBD ex protein with G O X was developed. C  The method resulted in higher  starting specific activity, increased sensor sensitivity and stability during use in the prototype sensor. 2. A number of hardware modifications were made including redesign of the membrane cartridge assembly and incorporation of continuous perfusion, both of which were necessary to yield a probe design that would be functional for the longer term. 3. A n investigation into the longer term signal stability was made.  Thermal  deactivation of the conjugate (catalytic instability) played a small role whereas structural instability of the conjugate played a larger role as the source of signal instability (as opposed to membrane or electrode fouling). The pattern of this instability was found to exhibit first-order decay kinetics and this behavior has  124  been observed in the literature. Long-term stability was significantly improved upon as a result of improvement in the conjugation protocol and the combination of hardware modifications. 4. The binding characteristics of a heterogeneous chemical fusion of C B D - G O X protein were determined for the first time and Langmuir adsorption theory was useful for analyzing the data. The prototype was developed to a stage where its utility for animal cell culture applications could be adequately investigated for the first time.  The prototype was  characterized during several long-term experiments under conditions of continuous operation, once for more than 550 hours. This is the longest reported duration of use for an in-situ enzyme electrode. The S M C C conjugation protocol was designed to reduce the amount of intra and intermolecular crosslinking between C B D and GOX. The protocol resulted in conjugate with a starting specific activity that provided sufficient sensitivity for sensor detection of glucose concentrations below 1 m M and increased structural stability. The redesigned membrane component was shown to effectively reject known interfering species and minimize sensor background signal while still providing adequate sensitivity, adequate response time and an absolutely aseptic barrier.  These effects  contributed to increased sensor signal stability as well. An unexpected problem of gas bubble accumulation inside the sensor chamber was effectively overcome by incorporation of a continuously perfusing buffer flow. continuous flow contributed to the observed enhancement of signal stability.  This  125 Longer term signal stability was investigated in detail for the first time using the Michaelis-Menten equation as an empirical model. The model was useful for transforming the raw sensor output into a value directly comparable with off-line or real glucose values. This direct comparison enabled evaluation of sensor signal stability. The finding of a firstorder decay pattern is significant and suggests that, if a reasonably precise value for the decay constant can be determined, then real-time correction of the sensor may be practical. This manner of on-line correction should at least reduce the cumulative error in the sensor signal due to loss of enzymatic activity, thus potentially extending the time periods between necessary calibrations.  The importance of long-term stability and a  summary of efforts to develop a stable prototype were presented and received with interest at the annual American Chemical Society (ACS) meeting (Fong et al, 1997).  5.2 Future Work The prototype developed in this work can be further optimized in a number of different areas. The importance of long-term stability and understanding the mechanisms responsible for the observed instabilities have been thoroughly discussed. A number of positive steps have been taken which enhanced long-term stability. Further investigations should lead to a more accurate idea of the signal decay characteristics.  In particular, the  inconsistent behavior of the sensor signal in the last experiment in chapter 4 (Figures 4.74.9) indicates that a number of unknown factor(s) still remain which influence signal behavior. Determination of these factor(s) and development of strategies to counteract them could further enhance signal stability.  Secondly, the apparent plateau effect  126  displayed by the sensor signal in the second experiment in chapter 4 (Figure 4.6) suggests a biphasic decay process in which a second slower process affecting signal stability may be taking place but does not predominate during the early periods of a loading. Accurate knowledge of the decay kinetics and the factors which influence them may enable further improvement in stability and are also necessary for the realization of a computer controlled system that could provide on-line continuous interpretation and correction of the sensor signal for decay. Continuous perfusion, incorporated to discourage gas bubble formation in the sensor chamber, solved the problem of loss of sensor responsiveness to glucose after 2448 hour periods (chapter 2). However, consistent regeneration of sensor responsiveness by reloading fresh conjugate enzyme was an on-going difficulty and requires further analysis (recall from the last experiment in chapter 4 that four attempts to regenerate the signal were unsuccessful).  The possibility that gas bubble accumulation in the sensor  chamber has not been completely eradicated cannot be discounted without further investigation. It is possible that minute gas bubbles trapped in the cellulose may interfere with the adsorption of conjugate and deactivation when the bolus is introduced into the sensor chamber. A first step to investigating this possibility would involve determining the composition of gas bubble(s) that tend to collect in the sensor chamber. This information might reveal the source of the bubbles and enable development of a more effective strategy to remove the problem completely.  Alternatively, occasional difficulty in  regenerating a glucose responsive signal may arise from issues relating to the internal geometry of the sensor chamber.  Although the membrane cartridge assembly was re-  127  designed such that changes in the internal geometry from the original configuration would be minimal (Figure 2.6, 2.7), it is possible that the existing geometry is not ideal in terms of maximizing exposure of conjugate bolus to the cellulose matrix. Fluid dynamic studies could provide answers to the issue of chamber geometry. Experimental work to further increase sensor sensitivity would be advantageous. In the present work, sensitivity was enhanced by improving the conjugation protocol. Other possibilities exist for additional improvement.  Sensitivity maybe enhanced by  investigating alternative cellulose matrices such as Valonia from Valonia ventricosa which is known to have a high capacity for CBD. Other issues such as porosity, diffusion characteristics and ease of handling would have to be investigated along with the binding capacity of the particular cellulose type. Higher sensitivity may also be attained by increasing the surface area of the platinum working electrode.  A number of  electrochemical considerations relating to the dynamics of electron transfer rates between the enzyme layer and the electrode surface as a result of increased surface area would also need to be considered. Further into the future, development of automated and computer-controlled methods to operate the sensor system would be very useful. At the present stage of development, a great deal of manual labour was required to generate, regenerate, calibrate, interpret and diagnose the sensor signal. The system was designed at the outset such that all of these procedures could potentially be carried out automatically.  As  mentioned previously, accurate knowledge of decay kinetics is necessary for implementation of computer control to continuously interpret and correct the signal for  128 decay.  Attainment o f computer control for this purpose alone would represent a  significant step in reducing the amount o f manual, specialized and highly-skilled labor necessary to operate the sensor. The feasibility and utility o f developing a genetically engineered fusion protein to replace chemical conjugation could also be investigated.  The advantages o f a genetic  fusion would be better reproducibility o f activity and homogeneous populations enzyme-CBD molecules.  of  The recombinant form o f G O X , expressed in yeast, has been  shown to have comparable specific activity, a wider optimal p H range and higher thermal stability than the native enzyme (Frederick et al,  1990; D e Baetselier et al,  1991). These  desirable characteristics may be imparted to a C B D - G O X genetic fusion. Development o f the fusion such that both proteins retain their biological activities may however be extremely challenging given the > 10-fold difference in size between the 2 proteins and the complexity o f G O X multi-subunit tertiary structure. The technology described in this work to develop a regenerable enzyme-based insitu sensor using C B D protein could be further extended to other analytes o f interest as mentioned previously (Table 1.1).  A great deal of work may be required to develop a  chemical or genetic conjugate with the particular oxidase enzyme and the same issues relating to stability, enzyme specificity, enzyme purity, specific activity and cost would need to be considered. The realization o f a reliable and complete on-line system for glucose analysis has enormous potential for glucose monitoring in bioprocesses and the evolution o f new fermentation control strategies. The availability o f such a tool for batch processes would  129  allow for glucose monitoring to evaluate not only consumption rates, but also to determine batch to batch variations, reproducibility and quality control standards. This information would enable determination of optimal feeding strategies for animal cells in culture and ultimately, the implementation of such strategies. In fed-batch processes, an on-line glucose monitoring system could be used to determine medium glucose exhaustion at which time feeding could be automatically triggered. Ultimately, the system could be utilized to its full potential as a closed-loop feedback control system to maintain glucose at an optimal set point concentration during high cell density (high glucose uptake rate) fedbatch culture. Evidence from the literature suggests that the optimal set point concentration is quite low (<1 mM). Maintenance of a high uptake rate at a low setpoint would require a system that is able to provide continuous real time measurement. Furthermore, current open-loop control methods coupled with discontinuous off-line sampling to predict glucose concentration in the near future would likely be unable to maintain glucose concentration effectively under such dynamic conditions.  130  References 1.  Adamson, A.W.„ "Physical Chemistry of Surfaces", Fifth edition, John Wiley & Sons, Inc. pp. 595, 1990.  2.  Aguado, J.; Romero, M.D.; Rodriguez, L . ; Calles, J.A.„ "Thermal Deactivation of Free  and  Immobilized B-Glucosidase From  Penicillium  funiculosum",  Biotechnology Progress, Vol. 11, pp. 104-106, 1995. 3.  Assouline, Z.; Graham, R.; Miller, R.C. Jr.; Warren, R.A.J.; Kilburn, D.G.„ "Processing of Fusion Proteins With Immobilized Factor Xa", Biotechnology Progress, Vol. 11, pp. 45-49, 1995.  4.  Bai, S.W.; Hong, H.J.; Lee, G.M.„ "Stability of Transfectomas Producing Chimeric Ab Against the Pre-S2 Surface Antigen of Hepatitis B Virus During Longterm Culture", Biotechnology and Bioengineering, Vol. 47, pp. 243-251, 1995.  5.  Bailey, J.E.; Ollis, D.F.„ "Biochemical Engineering Fundamentals", Second Edition, 1986.  6.  Bard, A.J.; Faulkner,  L.R.„ "Electrochemical Methods: Fundamentals and  Applications", John Wiley & Sons: New York, 1980. 7.  Benthin, S.; Nielsen, J.A.; Villadsen, J.„ "Flow Injection Analysis of Micromolar Concentrations of Glucose and Lactate in Fermentation Media", Analytica Chimica Acta, Vol. 261, pp. 145-153, 1992.  8.  Bowers, L.D.„ "Applications of Immobilized Biocatalysts in Chemical Analysis", Analytical Chemistry, Vol. 58, No. 4, pp. 513a-530a, 1986.  9.  Bradley, J.; Stocklein, W.; Schmid, R.D.„ "Biochemistry Based Analysis Systems For Bioprocess Monitoring and Control", Process Control and Quality, Vol.1, pp. 157-183, 1991.  10. Brooks,  S.L.; Ashby, R.E.; Turner,  A.F.P.;  Calder,  M.R.; Clarke, D.J.„  "Development of an On-Line Glucose Sensor for Fermentation Monitoring", Biosensors Vol. 3, pp. 45-56, 1987. 11. Byfield, M.P.; Abuknesha, R.A.„ "Biochemical Aspects of Biosensors", Biosensors and Bioelectronics, Vol. 9, pp. 373-400, 1994.  131 12. Cass, A.E.G.,, "Biosensors: A pratical Approach", IRL Press, 1990. 13. 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.F.P.,,  "Ferrocene-Mediated Enzyme  Electrode for Amperometric Determination of Glucose", Analytical Chemistry, Vol. 56, pp. 667-671, 1984. 14. Cattaneo, M . V . ; Luong, J.H.T.; Mercille, S.„ "A Chemiluminescence Fiber Optic Biosensor System for the Determination of Glutamine in Mammalian Cell Cultures", Biosensors and Bioelectronics, No. 7, pp. 329-334, 1992. 15. Claremont, D.J.; Penton, C ; Pickup, J.C.„ "Potentially Implantable, Ferrocene Mediated Glucose Sensor", Journal of Biomedical Engineering, V o l . 8, pp. 272274, 1986. 16. Clark L.C.„ "Monitor and Control of Blood and Tissue Oxygen Tensions", Transactions of the American Society of Artificial Internal Organs, Vol. 102, pp. 29-45, 1956. 17. Clark, L.C.„ "Membrane Polarographic Electrode System and Method With Electrochemical Compensation", US Patent 3 539 455, 1970. 18. Clark, L . C . ; Jr., Lyons, C.„ "Electrode Systems for Continuous Monitoring in Cardiovascular Surgery", Annals of the New York Academy of Science, Vol. 102, pp. 29-45, 1962. 19. Cleland, N . ; Enfors, S.O.„ "Control of Glucose Fed-Batch Cultivations of E. coli by Means of an Oxygen Stabilized Enzyme Electrode", European Journal of Applied Microbiology and Biotechnology, Vol. 18, pp. 141-147, 1983. 20. Cleland, N . ; Enfors, S.O.„ "Externally Buffered Enzyme Electrode for Determination of Glucose", Analytical Chemistry, Vol. 56, pp. 1880-1884, 1984a. 21. Cleland, N . ; Enfors, S.O.„ "Monitoring Glucose Consumption in an Escherichia coli Cultivation With an Enzyme Electrode", Analytica Chimica. Acta, V o l . 163, pp. 281-285, 1984b. 22. Cohen, S.; Chang, A.; Boyer, H ; Helling, R.„ "Construction of Biologically Functional Bacterial Plasmids In-Vitro", Proceedings of the National Academy of Sciences, Vol. 70, pp. 3240-3244, 1973.  132 23. Cooney, C.L.„ "Are We Preparedfor Animal Cell Technology in the 21st Century?", Cytotechnology, Vol. 18, pp. 3-8, 1995. 24. Cossons, N . H . ; Hayter, P . M . ; Tuite, M.F.; Jenkins, N.„ "Stability of Amplified DNA in Chinese Hamster Ovary Cells", from proceedings of the 10th annual E S A C T meeting concerning,, "Production of Biologicals from Animal Cells in Culture", pp. 309-315, 1990. 25.  Couture, M . L . ; Heath, C.A.„ "Relationship Between Loss of Heavy Chains and the Appearance of Non-producing Hybridomas", Biotechnology and Bioengineering, Vol. 47, pp. 270-275, 1995.  26. Curling, E . M . A . ; Hayter, P.M.; Baines, A.J.; Bull, A.T.; Gull, K . ; Strange, P.G.; Jenkins, N.„ "Recombinant human Interferon-/: Differences in Glycosylation and Proteolytic Processing Lead to Heterogeneity in Batch Culture", Biochemical Journal, Vol. 272, pp.333-337, 1990. 27. 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. niger Glucose Oxidase: Scale-Up, Purification and Characterization of the Recombinant Enzyme", Biotechnology, Vol. 9, pp. 559-561, 1991. 28. Enfors, S.O.„ "Oxygen Stabilized Enzyme Electrode for D-Glucose Analysis in Fermentation Broths", Enzyme Microbial Technology, Vol. 3, pp. 29-32, 1981. 29. Fernandez-Lafuente, R.; Rosell, C M . ; Rodriguez, V . ; Guisan, J.M.,, "Strategies for Enzyme Stabilization by Intramolecular Crosslinking With Bifunctional Reagents", Enzyme and Microbial Technology, Vol. 17, pp. 517-523, 1995. 30. Fersht, A . Enzyme Structure and Mechanism, Second Edition pp.69-75, 1985. 31. Fillipini, C ; Sonnleitner, B . ; Feichter, A ; Bradley, J.; Schmid, R.„ "On-Line Determinetion of Glucose in Biotechnological Processes: Comparison Between FIA and an In-Situ Enzyme Electrode", Journal of Biotechnology, Vol. 18, pp. 153-160, 1991. 32. Fong, F.; Turner, R.F.B.; Kilburn, D.G.„ "Characterization of an Experimental Glucose Sensor for Mammalian Cell Culture", Proceedings of the 1997 spring ACS meeting, April 13-17, 1997.  133  33. Frederick, K.R.; Tung, J.; Emerick, R.S.; Masiarz, F.R.; Chamberlain, S.H.; Vasavada, A . ; Rosenberg, S.„ "Glucose Oxidase From Aspergillus niger: Cloning, Gene Sequence and Secretion From Saccharomyces cerevisiae and Kinetic Analysis of a Yeast Derived Enzyme", Journal of Biological Chemistry, Vol. 265, No. 7, pp. 3793-3802, 1990. 34. Frederick, K.R.; Tung, J.; Emerick, R.S.; Masiarz, F.R.; Chamberlain, S.C.; Vasavada, A.; Rosenberg, S., "Glucose Oxidase From Aspergillus niger", Journal of Biological Chemistry, Vol. 265, No. 7, pp. 3793-3802, 1990. 35. Freshney, R.I., "Culture of Animal Cells", Second Edition, 1987. 36. Gey, G O . ; Coffman, W.D.; Kubicek, M.T., "Tissue Culture Studies of the Proliferative Capacity of Cervical Carcinoma and Normal Epithelium", Cancer Research, Vol. 12, pp. 364-365, 1952. 37. Gibson, T.D.; Woodward, J.R., "Protein Stabilization in Biosensor Systems", Chapter 5, Biosensors & Chemical Sensors, ACS, 1992. 38. Gilkes, N.R.; Jervis, E.; Henrissat, B.; Tekant, B.; Miller Jr, R.C.; Warren, R.A.J.; Kilburn, D.G., "The Adsorption of a Bacterial Cellulase and Its Two Isolated Domains to Crystalline Cellulose", Journal of Biological Chemistry, Vol. 267, No. 10., pp. 6743-6749, 1992. 39. Glacken, M . W . ; Adema, E.; Sinskey, A.J., "Mathematical Descriptions of Hybridoma Culture Kinetics I, Initial Metabolic Rates", Biotechnology and Bioengineering, Vol. 32, pp. 491-506, 1988. 40. Greenfield, P.F.; Kitrell, JR.; Laurence, R.L., "Inactivation of Immobilized Glucose Oxidase by Hydrogen Peroxide", Analytical Biochemistry, V o l . 65, pp. 109-124, 1975. 41. Griffiths, D.; Hall, G , "Biosensors -What Real Progress Has Been Made?", Trends in Biotechnology, Vol. 11, No. 4, pp. 122-130, 1993. 42. Guarna, M . M . ; Cote, H.C.F.; Amandoron, E.A.; MacGillivray R.T.A.; Warren, R.A.J.; Kilburn, D.G., "Engineering Factor X Fusions for Expression in Pichia pastoris", Annals of the New York Academy of Sciences Vol. 799, pp. 397-400, 1996.  134 43. Guilbault, G.G.; Montalvo, J., "A Urea-Specific Enzyme Electrode", Journal of the American Chemical Society, Vol. 91, pp. 2164, 1969. 44. Hall, E . A . H . , , "Biosensor^, Open University Press, 1990. 45. Hansen, E . H . ; Mikkelson, H.S., "Enzyme Immobilization by the Glutaraldehyde Procedure. An Investigation of the Effects of Reducing the Schiff Bases Generated, as Based on Studying the Immobilization of Glucose Oxidase to Silanized Controlled Pore Glass", Analytical Letters, Vol. 24, No. 8, pp. 1419-1430, 1991. 46. Harris, E . L . V . ; Angal, S., "Protein Purification Methods; a Practical Approach", IRL Press at Oxford University Press, pp. 19-39, 1989. 47. Hayter, P . M . ; Curling, E. M . A . ; Gould, M . L . ; Baines, A.J.; Jenkins, N . ; Salmon, I.; Strange, P . G ; Bull, A.T., "The Effect of the Dilution Rate on CHO Cell Physiology and Recombinant Interferon-y Production in Glucose Limited Chemostat Culture", Biotechnology and Bioengineering, Vol. 42, pp. 1077-1085, 1993. 48. Hayter, P . M . ; Curling, E . M . A . ; Gould, M . L . ; Baines, A.J.; Jenkins, N . ; Salmon, I.; Strange, P.G.; Tong, J.M.; Bull, A.T., "Glucose-Limited Chemostat Culture of Chinese Hamster Ovary Cells Producing Recombinant Human Interferon-/', Biotechnology and Bioengineering, Vol. 39, pp. 327-335, 1992. 49. Hu, W.S.; Dodge, T.C.; Frame, K . K . ; Himes, V . B . , "Effect of Glucose on the Cultivation of Mammalian Cells", Development of Biological Standards, Vol. 66, pp. 279-290, 1987. 50. Huang, Y . L . ; L i , S.Y.; Dremel, B . A . A . ; Biletewski, U . ; Schmid, R.D., "On-Line Determination of Glucose Concentration Throughout Animal Cell Cultures Based on Chemiluminescent Detection of Hydrogen Peroxide Coupled with Flow-Injection Analysis", Journal of Biotechnology, Vol. 18, pp. 161-172, 1991.  51. Imagawa, M . ; Yoshitaki, S.; Hamaguchi, Y . ; Ishikawa, E.; Niitsu, Y . ; Urushizaki, I.; Kanazawa, R.; Tachibana, S.; Nakazawa, N . ; Ogawa, H . , "Characteristics and Evaluation of Antibody-Horseradish Peroxidase Conjugates Prepared by Using a  135 Maleimide Compound,  Glutaraldehyde and Periodate", Journal of Applied  Biochemistry, Vol. 4, pp. 41-57, 1982. 52. Jenkins, N . ; Parekh, R.B.; James, D.C., "Getting the Glycosylation Right: Implications for the Biotechnology Industry , Review. Nature Biotechnology, Vol. 11  14, pp. 975-981, 1996. 53. Jervis, J. J.; Haynes, C.A.; Kilburn, D.G., "Surface Diffusion of Cellulases and Their Isolated Binding Domains on Cellulose", Journal of Biological Chemistry, Vol. 272, No. 38, pp. 24016-24023, 1997. 54. Kaufman, R. J.; Schimke, R.T., "Amplification and Loss of Dihydrofolate Reductase Genes in a Chinese Hamster Ovary Cell Line", Molecular and Cell Biology, Vol. 1, pp. 1069-1076, 1981. 55. Kaufman, R.J.; Brown, P.C.; Schimke, R.T., "Amplified Dihydrofolate Reductase Genes in Unstably Amplified Methotrexate-Resistant Cells are Associated With Double-Minute Chromosomes", Proceedings of the National Academy of Sciences, Vol. 76, No. 11. Pp. 5669-5673, 1979. 56. Kitigawa, Y . ; Kitabatake, K.; Suda, M . ; Muramatsu, H.; Ataka, T.; Mori, A ; Tamiya, E.; Karube, I., "Amperometric Detection of Alcohol in Beer Using a Flow Cell and Immobilized Alcohol Dehydrogenase", Analytical Chemistry, No. 63, pp. 23912393, 1991. 57. Kleppe, K . , "The Effect of Hydrogen Peroxide on Glucose Oxidase From Aspergillus niger", Biochemistry, Vol. 5, pp. 139-143, 1966. 58. Klip, A . ; Tsakiridis, T.; Marette, A . ; Ortiz, P.A., "Regulation of Expression of Glucose Transporters By Glucose: A Review of Studies In Vivo and In Cell Cultures", FASEB Journal, Vol. 8, pp.43-53, 1994. 59. Kohler, G.; Milstein, C , "Continuous Cultures of Fused Cells Secreting Antibody of Predefined Specificity", Nature, Vol. 256, pp. 495-497, 1975. 60. Kreichbaum, M . ; Heilmann, H.J.; Wientjes, F . J ; Hahn, M . ; Jany, K . D . ; Gassen, H.G.; Sharif, F.; Alaeddinoglu, G , "Cloning and DNA Sequence Analysis of the Glucose Oxidase Gene From Aspergillus niger", FEBS Letters, Vol. 255, pp. 63-66, 1989.  136  61. Krishnaswamy, S.; Kittrell, J.R., "Deactivation Studies of Immobilized Glucose Oxidase", Biotechnology and Bioengineering, Vol. 20, pp. 821-835, 1978. 62. Kurokawa, H . ; Ogawa, T.; Kamihira, M . ; Park, Y.S.; Iijima, S.; Kobayashi, T., "Kinetic Study of Hybridoma Metabolism and Antibody Production in Continuous Culture Using Serum Free Medium", Journal  of  Fermentation  and  Bioengineering, Vol. 76, No. 2, 128-133, 1993. 63. Kurokawa, H . ; Park, Y.S.; Iijima, S.; Kobayashi, T., "Growth Characteristics in FedBatch Culture of Hybridoma Cells With Control of Glucose and Glutamine Concentrations", Biotechnology and Bioengineering, Vol.44, pp. 95-103, 1994. 64. Lehninger, A . L . , "Principles of Biochemistry", Worth Publisher, Inc. New York, 1993. 65. Levering, P.R.; van Heijst, J.A.M.; Sunnen, C.M.G., "Physiology of Myeloma Cells Grown in Glucose-Limited Chemostat Cultures", Cytotechnology, V o l . 9, pp. 125130, 1992. 66. Ljunggren, J.; Haggstrom, L., "Catabolic Control of Hybridoma Cells by Glucose and Glutamine Limited Fed-Batch Cultures", Biotechnology and Bioengineering, Vol. 44, pp. 808-818, 1994. 67. Male, K . B . ; Gartu, P.O.; Kamen, A . A . ; Luong, J.H.T., "On-Line Monitoring of Glucose in Mammalian Cell Culture Using a Flow Injection Analysis (FIA) Mediated Biosensor", Biotechnology and Bioengineering, Vol. 55, No. 3, pp. 497504, 1997. 68. Malikkides, C O . ; Weiland, R.H., "On the Mechanism of Immobilized Glucose Oxidase Deactivation by Hydrogen Peroxide", Biotechnology and Bioengineering, Vol. 24, pp. 2419-2439, 1982. 69. Mattson, G ; Conklin, S.; Desai, G ; Neilander, M . D . Savage; Morgensen, S., "A Practical Approach to Crosslinking", Molecular Biology Reports, Vol. 17, pp. 167183, 1993. 70. Meier, H . ; Kumaran, S.; Danna, A . M . ; Tran, M . C . , "Rapid Measurement of Penicillin Contained in Complex Media Using Enzyme-Loaded Glass Electrodes", Analytica Chimica Acta, Vol. 249, pp. 405-411, 1991.  137 71. Miller, W . M . ; Wilke, C.R.; Blanch, H.W., "Transient Responses of Hybridoma Cells to Nutrient Additions in Continuous Culture: 1. Glucose Pulse and Step Changes", Biotechnology and Bioengineering ,Vol. 33, pp. 477-486, 1989. 72. Mosbach, K . Immobilized Enzymes, from Methods in Enzymology, Volume X L I V , 1976. 73. Moussy, F.; Jakeway, S.; Harrison, J.D.; Rajotte, R . V . , "In Vitro and In Vivo Performance and Lifetime of Per flourinated lonomer Coated Sensors After High Temperature Curing", Analytical Chemistry, Vol. 66, pp. 3882-3888, 1994. 74. Mulchandani, A . ; Bassi, A . S., "Principles and Applications of Biosensors For Bioprocess Monitoring and Control', Critical Reviews in Biotechnology, Vol. 15, No. 2, pp. 105-124, 1995. 75. Nikolova, P.V.; Creagh, A.L.; Duff, S.J.B.; Haynes, C.A., "Thermostability and Irreversible Activity Loss of Exoglucanase/Xylanase Cex From Cellulomonas fimi", Biochemistry, Vol. 36, pp. 1381-1388, 1997. 76. Numberg, J.H.; Kaufman, R.J.; Schimke, R.T.; Urlaub, G.; Chasin, L . A . , "Amplified Dihydrofolate Reductase Genes are Localized to a Homogeneously Staining Region of a Single Chromosome in a Methotrexate-Resistant Chinese Hamster Ovary Cell Line", Proceedings of the National Academy of Sciences, Vol. 75, pp 5553-5556, 1978. 77. Ong, E . ; Alimonti, J.B.; Greenwood, J.M.; Miller, R.C. Jr.; Warren, R.A.J.; Kilburn, D.G., "Purification of Human Interleukin-2 Using the Cellulose-Binding Domain of aProkaryotic Cellulase", Bioseperation, Vol.5, pp. 95-104, 1995. 78. Ong, E.; Gilkes, N.R.; Miller, R.C.; Warren, R.A.J.; Kilburn, D.G., "The Cellulose Binding Domain (CBDcex) of an Exoglucanase From Cellulomonas fimi: Production in Escherichia coli and Characterization of the Polypeptide", Biotechnology and Bioengineering, Vol. 42, pp. 401-409, 1993. 79. Ong, E . ; Gilkes, N.R.; Miller, R.C.; Warren, R.A.J.; Kilburn, D.G., "Enzyme Immobilization Using a Cellulose Binding Domain: Properties of a B-Glucosidase Fusion Protein", Enzyme Microbial Technology, Vol. 13, pp. 59-65, 1991.  138 80. Ozturk, S.S.; Riley, M.R.; Palsson, B.O., "Effects of Ammonia and Lactate on Hybridoma Growth, Metabolism, and Antibody Production", Biotechnology and Bioengineering, Vol. 39, pp 418-431, 1992. 81. Ozturk, S.S.; Thrift, J.C.; Blackie, J.D.; Naveh, D., "Real-Time Monitoring and Control of Glucose and Lactate Concentrations in a Mammalian Cell Perfusion Reactor", Biotechnology and Bioengineering, Vol. 53, No. 4, pp. 372-378, 1997. 82. Palmiter, R.D.; Behringer, R.R.; Quaife, C.J.; Maxell, F.; Maxwell, I.H.; Brinster, R.L., "Cell Lineage Ablation in Transgenic Mice by Cell-Specific Expression of a Toxic Gene", Cell, Vol. 50, pp. 435-443, 1987. 83. Phelps, M . R., M . App. Sc. Thesis, 1994. 84. Phelps, M . R . ; Hobbs, J.B.; Kilburn, D.G.; Turner, R.F.B., "Technology For a Regenerable Biosensor Based on Enzyme-Cellulose Binding Domain Conjugates", Biotechnology Progress, Vol. 10, pp. 433-440, 1994. 85. Phelps, M.R.; Hobbs, J.B.; Kilburn, D.K.; Turner, R.F.B., "An Autoclavable Glucose Biosensor for Microbial Fermentation Monitoring and Control', Biotechnology and Bioengineering, Vol. 46, No.6, pp. 514-524, 1995. 86. Pons, M . N . , "Biosensors for Fermentation Control', Current Opinion in Biotechnology, Vol. 4, pp. 183-187, 1993. 87. Raba, J.; Mottola, H.A., "Glucose Oxidase As An Analytical Reagent", Critical Reviews in Analytical Chemistry, Vol. 25, No. 1, pp. 1-42, 1995. 88. Ramirez, C ; Fung, J.; Miller, R.C. Jr.; Warren, R.A.J.; Kilburn, D . G . , "A Bifunctional Affinity Linker to Couple Antibodies to Cellulose", Bio/Technology, Vol. 11, pp. 1570-1573, 1993. 89. Ray, R.R.; Jana, S.C.; Nanda, G , "Biochemical Approaches of Increasing Thermostability of B-Amylase From Bacillus megaterium B ", FEBS Letters, V o l . 6  356, pp. 30-32, 1994. 90. Ruzicka, J.; Hansen, E.N., "Flow Injection Analysis Part I: A New Concept of Fast Continuous Flow Analysis", Analytica Chimica Acta, Vol. 78, pp. 145, 1975. 91. Sawyer, D.T.; Roberts, J.L. Jr., "Experimental Electrochemistry for Chemists", John Wiley and Sons: New York, pp. 39-41, 1974.  139 92. Sayle, R. Rasmol Molecular Modelling Program, G L A X O research and development greenford, Middlesex, UK., 1995. http://www.umass.edu/microbio/rasmol/index.html 93. Scheper, T.H.; Hilmer, J.M.; Lammers, F.; Muller, C ; Reinecke, M , "Biosensors in Bioprocess Monitoring", Review Journal of Chromatography A, Vol. 725, pp. 312, 1996. 94. Scouten, W.H.; Luong, J.H.T.; Brown, S., "Enzyme or Protein Immobilization Techniques For Applications In Biosensor Design", Trends in Biotechnology, Vol. 13, pp. 178-185, 1995. 95. Sethi, R.S., "Transducer Aspects of Biosensors", Biosensors and Bioelectronics, Vol. 9, pp. 243-264, 1994. 96. Sigma Catalogue, Sigma Chemical Co. St. Loius, M O , USA, pp. 522-524, 1998. 97. Simonsen,  C.C.; Levinson, A . D . , "Isolation and Expression of An Altered  Dihydrofolate Reductase cDNA", Proceedings of the National Academy of Sciences, Vol. 80, pp. 2495-2499, 1983. 98. Skoog, D. A., "Principles of Instrumental Analysis", Third Edition. Saunders College Publishing, pp. 23-24, 1985. 99. Stryer, L . , "Biochemistry", Second Edition. W . H . Freeman & Co., pp. 240-241, 1981. 100. Sugiura, T., "Effects of Glucose on the Production of Recombinant Protein C in Mammalian Cell Culture", Biotechnology and Bioengineering, Vol. 39, pp. 953959, 1992. 101. Towe, B.C.; Guilbeau, E.J.; Coburn, J.B., "In Vivo and In Vitro Deactivation Rates of PTFE-Coupled Glucose Oxidase", Biotechnology and Bioengineering, V o l . 11, No. 8, pp. 791-798, 1996. 102. Turner, A.F.P.;  Karube,  I.; Wilson, G.S., "Biosensors: Fundamentals and  Applications", First Edition, Oxford University Press, 1987. 103. Twork, J.V.; Yacynych, A . M . Sensors in Bioprocess Control, Marcel Dekker, Inc., 1990. 104. Updike, S.J.; Hicks, G.P., "The Enzyme Electrode", Nature, Vol. 210, pp. 986, 1967.  140  105. Wang, J., "Permselective Coatings for Amperometric Biosensing", Biosensors and chemical sensors: optimizing performance through polymeric materials; Eddman, P.G.; Wang, J., Eds.; A C S : New Mexico State University,; Symposium series 487, pp. 125-132, 1992. 106. Wang, I ; Honda, H , Lenas; P., Wantanabe, H ; Kobayashi, T., "Effective tPA Production by BHK Cells in Nutrients Controlled Culture Using an On-Line HPLC Measuring System", Journal of Fermentation and Bioengineering, Vol. 80, No. 1, pp. 107-110, 1995. 107. Wang, J.; Hutchins, L . D . , "Thin-Layer Electrochemical Detector With a Glassy Carbon Electrode Coated With a Base-Hydrolyzed Cellulosic Film", Analytical Chemistry, Vol. 57, pp. 1536-1541, 1985. 108. White, S.F.; Turner, A.P.F.; Biltewski, U ; Bradley, J.; Schmid, R . D . , "On-Line Monitoring of Glucose, Glutamate and Glutamine During Mammalian Cell Cultivations", Biosensor and Bioelectronics, Vol. 10, pp. 543-551, 1995. 109. Wilson, R.; Turner, A.P.F., "Glucose Oxidase: An Ideal Enzyme", Biosensors and Bioelectronics, Vol. 7, pp. 165-185, 1992. 110. Wingard, L . B . ; Cantin, L . A.; Castner, J.F., "Effect of Enzyme-Matrix Composition on Potentiometric Response to Glucose Using Glucose Oxidase Immobilized on Platinum", Biochimica et Biophysica Acta, Vol. 748, pp. 21-27, 1983. 111. Woedtke, T. von; Fischer, U . ; Abel, P., "Glucose Oxidase Electrodes: Effects of Hydrogen Peroxide Activity?", Biosensors and Bioelectronics, V o l . 9, pp. 65-71, 1994. 112. Wollenberger, U . ; Schubert, F.; Pfieffer, D.; Scheller, F.W., "Enhancing Biosensor Performance Using Multienzyme Systems", Review, Trends in Biotechnology, V o l . 11, No. 6, pp. 255-262, 1993. 113. Xie, L . ; Nyberg, G ; Gu, X . ; L i , H ; Mollborn, F.; Wang, D.C., "Gamma-Interferon Production and Quality in Stoichiometric Fed-Batch Cultures of Chinese Hamster Ovary (CHO)  Cells Under Serum-Free Conditions", Biotechnology  Bioengineering, Vol. 56, No. 5, pp. 577-582, 1997.  and  141 114. Xie, L.; Wang, D.I.C., "Integrated Approaches To the Design of Media and Feeding Strategies For Fed-Batch Cultures of Animal Cells", Trends in Biotechnology, V o l . 15, No. 3, pp. 109-113, 1997. 115. Yang, S.T.; Marchio, J.L.; Yen, J.W., "A Dynamic Light Scattering Study of 3Galactosidase: Environmental Effects on Protein Conformation and Enzyme Activity", Biotechnology Progress, Vol. 10, pp. 525-531, 1994. 116. Yoshitake, S.; Yamada, Y . ; Ishikawa, E.; Masseyeff, R., "Conjugation of Glucose OxidasefromAspergillus niger and Rabbit Antibodies Using N-Hydroxysuccinimide Ester of  N-(4-carboxycyclohexylmethyl)-Maleimide",  European  Journal  of  Biochemistry, Vol. 101, pp. 395-399, 1979. 117. Zar, J.H., "Biostatistical Analysis", Second Edition pp. 122-132, 1984. 118. Zeng, A.P.; Deckwer, W.D., "Mathematical Modelling and Analysis of Glucose and Glu famine Utilization and Regulation in Cultures of Continuous Mammalian Cells", Biotechnology and Bioengineering, Vol. 47, pp. 334-346, 1995.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0088469/manifest

Comment

Related Items