MSU LIBRARIES “ RETURNING MATERIALS: Place in book drop to remove this checkout from your record. FINES will be charged if book is returned after the date stamped below. «‘- a . “5m PERFORMANCE OF BETA-GALACTOSIDASE FROM BACILLUS §TEAROTHERMOPHILUS IN HOLLOW FIBER REACTORS By Richard John Knob A THESIS Submitted to MIchigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Chemical Engineering 1988 ABSTRACT PERFORMANCE OF BETA-GALACTOSIDASE FROM BACILLUS STEAROTHERMOPHILUS IN HOLLOW FIBER REACTORS By Richard John Knob The performance of asymmetric polyamide hollow fiber reactors (HFRs) backflush loaded with a thermophilic enzyme, fi* ggggxgghgrmgphilgg fi-galactosidase, was investigated as a means of alleviating the problems of lactose removal in dairy products. The free-solution thermophilic enzyme kinetic parameters were determined to be K -3.69 mM, V -2.08 uMol mg.1 min-1, and K - 7.16 mM at 55°C and pH m max c 6.7. Under the operational conditions used, the enzyme was adsorbed and retained by the polyamide fiber. Reaction rate in the SFRs was independent of flow rate but proportional to enzyme loading. The performance of diffusion-limited SFRs (effectiveness factors of less than 0.2) was interpreted through a mathematical model dependent on the experimentally derived effective diffusivities of lactose and galactose 6 6 (0.86x10- and 1.85x10- cmz/sec, respectively) and the model-derived immobilized enzyme kinetics (Km-20.35 mM, Vmax-14.53 pMol mg.1 min-1, and Kc-0.765 mM). The model was used to scale the reactor performance to industrial scale systems. DEDICATION To Carol, Your inspiration has helped me overcome my fear of the future. Your warmth, compassion, and understanding has eternally captured my heart. The pages which follow contains our thesis ................................. iii ACKNOWLEDGMENTS Without the assistance and patience of Dr. Daina Briedis, my academic advisor and friend, a project of this depth could not have been completed. Thank you for your understanding and support. It was my fortune to gain the advise of another professional, Dr. Mark Worden, who was there when Dr. Briedis was unavailable. I would like to take this time to remember my friends at Michigan State University who had given me my everlasting memories of Michigan. They were Mike Bly, Jeff Bowles, Alan Powell, Steve Reiken, and the entire congenial staff of the Chemical Engineering Department. Services, finances, and materials were given through the generosity of the following contributors: Wiscosin Milk Marketing Board Dr. Mansel Griffiths, Hannah Research Institute W Dr. Patrick Oriel, Dept. of Microbiology and Public Health Dr. John Partridge, Dept. of Food Science and Human Nutrition Dr. Dennis Miller, Dept. of Chemical Engineering Without the financial support of the Department of Chemical Engineering, my experience at MSU would not have been possible. I am deeply grateful to all the above mentioned. iv TABLE OF CONTENTS List of Tables List of Figures Introduction Chapter 1: Background I. Lactose Digestion II. Current Methods of Hydrolyzing Lactose Free Enzyme Method Immobilized Enzyme Method III. Enzyme Selection IV. Hollow Fiber Reactors V. Immobilized Enzyme Kinetic Models VI. Experimental Program Chapter 2: Enzyme Kinetics and Production Optimization 1. Introduction II. Enzyme Kinetics Mathematics and Data Treatment Kinetic Parameter Verification III. Materials and Methods Enzyme Assays Endpoint Assay Method Continuous Assay Method Determination of Kinetic Parameters Batch Conversions ix ll 19 21 23 23 23 25 27 28 28 28 3O 31 33 IV. Chapter 3: I. II. III. Chapter 4: I. II. III. IV. Enzyme Production Optimization Results and Discussion Kinetic Parameters Batch Conversions Enzyme Production Optimization Additional Product Enzyme Retention and Stability Introduction Enzyme Immobilization Fiber Preparation of Loading and the Single Fiber Reactor System Experimental Method Enzyme Retention Enzyme Stability Results and Discussion Enzyme Retention Enzyme Stability Hollow Fiber Characterization Introduction Theory Diffusivities Fouling Materials and Methods Analytical Techniques and Data Treatment Permeability/Diffusivity Experiments Fouling Experiments Results and Discussion Permeability/Diffusivity Studies Fouling Studies vi 34 36 36 39 45 47 50 50 50 51 53 53 54 55 55 58 61 61 61 62 63 65 65 66 68 70 70 71 Chapter 5: Reactor Performance I. Introduction II. Theoretical Basis for Data Analysis Effectiveness Factor versus Modulus Bolus Flow III. Materials and Methods Reactor Operation Flow Rate Experiments Enzyme Loading Experiments Data Treatment Bolus Flow Experiments IV. Results and Discussion Flow Rate Experiments Enzyme Loading Experiments Bolus Flow Experiments Chapter 6: Immobilized Enzyme Kinetic Modelling I. Introduction‘ II. Theoretical Basis for Data Analysis III. Materials and Methods IV. Data Treatment and Results V. Discussion Intrinsic Immobilized Enzyme Kinetics Immobilized Enzyme Kinetic Model Scale-up Chapter 7: Conclusions and Recommendations I. Enzyme Enzyme Selection Enzyme Retention and Stability II. Reactor vii 77 77 78 78 so 81 81 84 84 84 86 88 88 88 96 97 97 97 100 100 111 111 111 112 113 113 113 114 115 Single Fiber Reactor Enzyme Loading III. Kinetic Model Scale-up Enzyme Distribution Reaction Species Diffusivities IV. Reactor Operation Method of Operation Substrate Solutions/ Fouling V. Concluding Summary Appendix: Computer Programs and Flowchart Program A Flowchart Computer Programs A-J List of References viii 115 115 116 116 117 117 118 118 118 119 120 121 139 LAWN U!“ N." LIST OF TABLES Approximate Composition of Milk Reaction Mixtures Used in the Determination of Kinetic Parameters Enzyme Production Protocol Fiber Cleaning Process Activity Retention of B* figggzgghgzmgphilug Beta- Galactosidase in Two Different Polyamide Hollow Fibers Total Protein Retention from Bi fitgargthegmgphilug Beta-Galactosidase Solution in Two Different Polyamide Hollow Fibers Results of the PAlO Fiber Permeability Study Reactor Enhancement - Bolus Flow Results and Specifications Non-Linear Regression Analysis Results Enzyme Kinetic Parameters ix 32 52 56 57 7O 96 102 102 NH 99 94> LIST OF FIGURES Electron Micrograph of Single Hollow Fiber. Single Fiber Reactor (SFR).- Shell material is boro- silicate glass, 22.0 cm effective active fiber, 0.8 cm o.d., tapered to 0.5 cm diameter x 1 cm length, end tubes with side tubes 0.5 cm diameter x 1 cm length; fittings illustrated on left were applied to both ends of the reactor: L- Male and Female Leur lock fittings; C- Silicone tube; HF- Hollow ultrafiltration fiber. The hollow fiber was retained by a plug of epoxy potting resin (Dow Chemical) between the Leur fittings and the hollow fiber. Single Fiber Reactor (SFR) System.- Res- Reservoir flask, P- Gear or peristaltic pump (Cole Palmer model #7520-25 unified drive; Micropump #81281 head or Cole Palmer #7520-14 head); f1, f2- flowmeters (Cole Palmer #FM044-40 and #FMlOZ-S); pl- lumen-side inlet port; p2- shell-side inlet port; p3- lumen outlet port; p4- shell- side outlet port; tl- 15 psig pressure transducer (Omega #PXl42-01565V); t2- 5 psi differential pressure transducer (Omega #PXl42-005D5V); 0- outlet sample port. Photomicrograph of PM10 Hollow Fiber, Longitudinal View. Lineweaver-Burk Plot Bacillus W Beta- Galactosidase. 54.50C. (No Inhibitor). Lineweaver-Burk Plot Bacillus Sgggrgghgzmgphilgg Beta- Galactosidase. 54.50C. (10 mM Inhibitor). Batch Data - Beta-Galactosidase (0.401625 units/ml). Reaction Mixture Contains Mn and Mg. Batch Data - Beta-Galactosidase (0.80325 units/ml). Reaction Mixture Contains Mn and Mg. Enzyme Stability in Batch Reactions with Lactose. Stir Rate- 50 rpm. Reaction Rate in Batch Reactions (0.401625 units Beta- Galactosidase/ m1). Reaction Mixture Contains Mn and Mg. Reaction Rate in Batch Reactions (0.80325 units Beta- Galactosidase/ m1). Reaction Mixture Contains Mn and Mg. Serial Bacillus gtegrgthermophilus Growth in Batch Cultures. All 5% Innoculations of Varying Cell Density. Enzyme Stability in Batch Reactions without Lactose. Stir Rate- 50 rpm. Permeability Determination Plot. Dual Closed-Loop Dialysis System for Membrane Permeability. Whey Permeate Fouling Study. Permeability Determination Plot. Dagano Cheese Whey Fouling Study. Permeability Determination Plot. 12 15 16 17 37 38 40 41 42 43 44 46 60 64 67 72 73 b9 UIUI U'IUI Skim Milk Fouling Study. Permeability Determination Plot. Extended Skim Milk Fouling Study. Permeability Determination Plot. Program A Flowchart. (Appendix) Graphical Illustration of the Dependence of the Normalized Overall Reaction Rate, R, on the Total Flow Rate in Homogeneous, Fh or Slug Flow, W, through a Straight Tube. Bubble-Generating System. Effect of Varying Flow Rate on Conversion in SFR. Thermophilic Enzyme. Effect of Varying Beta-Galactosidase Loading on Conversion with Time in SFR. Mesophilic Enzyme. Effect of Varying Beta-Galactosidase Loading on Conversion with Time in SFR. Thermophilic Enzyme. Comparison of SFR Reaction Rates with Predicted Rates for Free Enzyme. Effectiveness Factor vs Generalized Modulus at Different Mesophilic Enzyme Loadings. Reactor Flow Rate-6.5 ml/min. Effectiveness Factor vs Generalized Modulus at Different Thermophilic Enzyme Loadings. Reactor Flow Rate-6.5 ml/min. Determination of a Non-Linear Regression Analysis of a 48 Hour Reactor Performance Study [18.28 Units of Thermophilic Enzyme Loaded]. Verification of Corrected Model of a 48 Hour Reactor Performance Study [18.28 Units of Thermophilic Enzyme Loaded]. Effectiveness Factor vs Generalized Modulus at Different Thermophilic Enzyme Loadings with Corrected Model Prediction Curve. Data for the Determination of Time to Complete Various Conversions [at Optimum Thermophilic Enzyme Loading]. Estimation of the Time Needed to Complete Various Conversions [at Optimum Thermophilic Enzyme Loading]. Number of Cartridges to Achieve Conversion with Volume- 100 gal. and Time-10 hrs. at Optimum Thermophilic Enzyme Loading. Grams of Enzyme Required for Conversion with Volume- 100 gal. and Time-10 hrs. at Optimum Thermophilic ’Enzyme Loading. Units of Enzyme Required for Conversion with Volume- 100 gal. and Time-10 hrs. at Optimum Thermophilic Enzyme Loading. xi 74 76 82 87 89 90 91 93 94 95 101 104 105 106 107 108 109 110 INTRODUCTION The purpose of the work presented in this thesis has been to examine the use of enzymes immobilized on hollow fibers for the processing of dairy products to improve digestability and marketability and for the treatment of dairy waste streams. 0f the components of milk and other dairy streams, lactose, a disaccharide consisting of glucose and galactose monomers, most inhibits wider use of dairy products and easier disposal of dairy waste streams. Lactose poses the problem in the digestability and processing of dairy products. Lactose intolerance is a fairly common malady affecting approximately 10% of the adult population. Most other adult mammals also share the inability to digest lactose, thus limiting the use of dairy by-products for livestock and pet feeding. In dairy product processing and utilization, the presence of lactose presents problems arising from its low sweetness and low solubility compared with other sugars, including its constituent monosaccharides. Lactose has a low solubility, between 11.9 g/mg water (25°C), compared to galactose, the less soluble of the constituent monomers, which is almost twice as soluble as lactose. This low solubility prevents the development of concentrated milk and syrups containing high lactose concentrations. Its solubility is also sharply temperature dependent (Banks, Dalgleish, and Rook 1981); therefore, frozen dairy products often develop graininess due to ‘actose crystallization. Any dairy waste streams containing lactose (such as 2 whey permeate) also present a problem in waste disposal that requires the removal of lactose. In the case of whey permeate treatment, the converted stream may be reused as a sweetener in other dairy products such as yogurt, beverages, frozen deserts, syrups, and confections. Since milk is a food product, any processing method to produce low- lactose products must adhere to strict standards that maintain purity, shelf life, organoleptic characteristics, and nutritional standards. Standard chemical methods of disaccharide hydrolysis, e.g. acid and cationic resins, can only be used on fairly pure lactose solutions and alter the product unfavorably. The use of acid would precipitate casein and change other milk constituents. The enzyme fl-galactosidase, or lactase, selectively hydrolyzes the fi(1-4) glycosidic bond in lactose. If used free in solution in the dairy product to convert lactose, upon completion of the reaction the enzyme remains in the dairy product. Since the enzyme is a protein, it denatures and may give off flavors to the product. Therefore, an immobilization technique whereby the enzyme is separated from the dairy stream and is retained for reuse was investigated. Chapter 1 W I. Lactose Digestion In order to assist the dairy industry in dairy product processing and lactose intolerance problems, the objective of this study has been to develop a method to cost-effectively remove the milk sugar lactose, from milk and other dairy streams without affecting the final product's important nutritional and physical characteristics. This may be accomplished through an enzymatic hydrolysis. The enzyme called fl-d- galactosidase, or lactase, selectively hydrolyzes the 5(1-4) glycosidic bond in lactose to yield sweeter, i.e. glucose and galactose. A degree of lactose hydrolysis of 50-80% is considered sufficient in alleviating the problems associated with lactose in dairy products. Lactose is a disaccharide comprised of glucose and galactose monomers and is found in milk and dairy products. Like other disaccharides, lactose cannot be transported across the human intestinal membranes. In order for lactose to be digested, it must be hydrolyzed to its monosaccharide components by lactase. Lactose intolerance signifies a total intestinal lactase activity that is insufficient to hydrolyze usual amounts of dietary lactose. The undesirable effects of lactose intolerance, such as cramps, nausea, and diarrhea, are the result of fermented undigested lactose reaching the colon {Lampert 1975). Lactase deficiency develops in the first years of life and is inherited {Paige and Bayless 1981). This deficiency is particularly widespread among the inhabitants of the Third World as well as the Western World. Unfortunately, it is these populations that could A benefit most from the consumption of milk, an inexpensive, virtually complete food. ‘ The composition of milk determines its nutritive quality, its value as a raw material for making food products, and many of its physical and chemical properties. The approximate composition of milk is given in Table l. 1. Table 1.1: Approximate Compositipp pf Milk {Walstra and Jenness 1984) Comment W Water 3 Solids (not fat) Pat in dry matter Lactose Fat Protein Casein Mineral substances Organic acids COHNNUNVW Labiknb'pboi-alo'm 0 on ubmmwooo moub'm'uao'o'u *1 These values will be exceeded rarely, possibly in 1-2% of samples of separate milkings of individual cows. The principal components are those present in the largest concentrations; they are not necessarily the most important ones. For instance, the vitamins A, B, B2, C, D, and E are important in nutritive value. Enzymes in milk are important catalysts of deteriorative reactions, and some minor components contribute markedly to the taste of milk (Walstra and Jenness 1984}. The composition of cow's milk varies for a number of reasons including the breed and individuality of the cow, stage of lactation, environmental temperature, time of milking, interval between milkings, and nutrition of cows {Eddy 1939). The presence of lactose in dairy product processing and utilization creates problems due to the sugar's low sweetness and low solubility. 5 In the manufacturing of ice cream, a lactose content which is too high can, with inexpert handling and in long storage at fluctuating temperatures, cause a sandy texture in the ice cream due to lactose crystallization {Kessler 1981}. Crystallization of lactose in sweetened condensed milk cannot be avoided. The low sweetness of lactose, its slow absorption into the blood, and its undesirable crystallization properties disqualify its use as a good sweetener. Lactose-hydrolyzed (LH) milk and whey, as products of the hydrolysis process studied in this work, possess several altered chemical and physical properties of interest to the dairy manufacturer. Benefits include reduction of lactose content, prevention of lactose crystallization, increase in solubility and sweetness, and more readily fermentable sugars. But the most obvious benefit is the supply of low lactose dairy products for the lactase deficient or lactose intolerant individual. LH milk may be consumed directly or may be used to produce products such as concentrates, cheeses (cottage and cheddar types), yogurt, buttermilk, and dehydrated products. A frozen 3:1 concentrate is the preferred way of preserving milk with minimal flavor change. A LH concentrate reduces problems that occur due to the crystallization of lactose and improves the physical stability of concentrated milk during storage. Products that require fermentation of sugar include buttermilk, yogurt, and cottage- and cheddar-type cheese. The processing of these products is accelerated if the carbon source is primarily glucose and not lactose. This may be achieved using LH milk. Of even greater economic significance is the problem of the disposal of whey. Whey is the serum or watery part of milk that is separated from the coagulable part or curd. Because little more than half of the 32 billion pounds of whey produced annually in the 0.8. is 6 utilized (Coughlin and Charles 1980}, there appear to be significant environmental and economic incentives for converting this high biological oxygen demand wastewater by—product into usable products, thereby returning whey to the food chain. Modification of whey by lactose hydrolysis (or production of cheese from LH milk to produce LH whey) may enable whey to reenter the food chain in the form of whey beverages, syrups, alcohol, food additives, and confections. A lactose- hydrolyzed whey permeate has been used to develop a prototype snack-type soft drink, "Lactofruit", projected to be cheaper to produce and sell than other whey beverages because of the unique process employed, i.e. lactose hydrolysis by enzymatic electrocatalysis {Fresnel and Moore 1978}. Another possibility for utilization of lactose-treated whey is as an ingredient in novelty water ices on a stick. This is advantageous in that addition of acid whey permits the manufacture of less acid water ice; dental cavity preventatives, calcium and phosphate, present in the whey become constituents of the ice pop and could be beneficial in reducing dental cavities (Wagg, Friend, and Smith 1965}. Relatively few organisms can ferment lactose when whey is utilized as a fermentation substrate in wine production. LH whey does allow microbial species to ferment glucose, but effectively converting galactose into alcohol requires strict fermentation conditions {O'Leary et a1. 1977}. II. W The need for hydrolyzing lactose has led to the development of many hydrolysis methods, each having its advantages and disadvantages. One method of hydrolyzing lactose is the use of strongly acidic cation exchangers which operate at a temperature of 90°C {Poulsen 1984}. 7 Although very cost-effective, the nutritional value of the dairy product is reduced drastically because the integrity of the dairy proteins is changed. In order to retain product purity and maintain the nutritional value of dairy products, enzymatic lactose hydrolysis has been suggested as the preferred method. Enzymatic hydrolysis can be utilized in either a free enzyme or an immobilized enzyme method. A comparison of these methods will be necessary to evaluate their relative merits. Wm Most dairy plants use a free enzyme, i.e. fungal lactase from Apppggillpp piggg, in milk processing. If the fungal lactase is added to cold milk, the enzyme is heat-activated as the temperature rises to 63°C for pasteurization, and hydrolysis of the lactose occurs during the 30-60 minute holding period. The hydrolysis process occurs while growth of pathogenic microorganisms is minimized. Thus, dairy plants may batch pasteurize LH milk in one operation {Paige and Bayless 1981}. However, the use of a free enzyme method prevents the possibility of reusing the enzyme since the protein is retained in the product The major benefit of using the free enzyme method is the ease of its implementation. MW Methods of fl-galactosidase immobilization have involved one or a combination of the following three immobilization methods: (1) pppzppngnt - purified lactase from Escherichia £21; or yeast has been immobilized by entrapment in a triacetyl cellulose fiber; the enzyme was very stable and retained the initial activity after continuous operation for 30 days {Morisi, Pastore, and Viglia 1973}, (2) pdgpzppipn - the adsorption of fi-galactosidase on porous glass particles for the hydrolysis of acid in acid whey has been studied by Wierzbicki, Edwards, and Kosikowski {1974}, and (3) covalent bonding - this technique is not 8 used in food and dairy applications due to the health hazards which severely limit the type of reagents which can be covalently bonded to the enzyme. The primary factors affecting the cost of using an immobilized enzyme process are (l) enzyme cost, (2) immobilization cost, (3) system performance, (4) capital investment, and (5) cleanup costs. It should be emphasized that in order to judge a process to be economically feasible, all possible benefits from a processing plant must outweigh the cost of a plant. The benefits in immobilizing lactase for whey processing in a cheese plant illustrates this point. For example, a complete whey recovery and utilization program reduces cost for wastewater treatment, allows production of whey proteins that have high nutritional and functional qualities, and by treatment with immobilized lactase, permits production of a concentrated syrup from whey permeate {Roland 1980}. While any one of these processes alone would not be economically feasible, taken together, and with large volumes of products produced, the recovery becomes profitable {Swaisgood 1985}. There is always the undesirable potential for microbial contamination of a food product. In addition, reactor operating conditions (pH, temperature, and shear rates) may provide excellent conditions for microbial growth which may reduce reactor productivity. It is essential to control the microbial population by appropriate reactor design, by the choice of operating conditions, or by using a periodic sanitation procedure {Swaisgood 1985}. For example, a high shear environment discourages microbial attachment to a hollow fiber without inactivating the enzyme. The extended temperature and pH ranges in which hollow fibers can be operated allow the use of operating conditions which reduce the chance for microbial growth. 9 In any immobilization scheme, the possibility of recovering and recycling the enzyme's activity will take the highest priority in making the scheme economically feasible. The hollow fiber reactors used in this work offer advantages and solutions to many of the problems discussed in this section. They will be discussed in greater detail later in this chapter. III. M Whether an immobilization technique will be cost-effective or not is greatly influenced by the choice of the appropriate enzyme. The enzyme's specific activity, stability, and its ability to be recovered and reused after immobilization must be considered. An additional factor for dairy processing includes the enzyme's resistance to thermal inactivation, optimum pH range, and probable acceptability, i.e. can the enzyme be generally recognized as safe (GRAS) by the Food and Drug Administration. The properties of lactase will vary among these sources and between strains. The enzyme is required to remain stable and active up to 130°F to inhibit microbial growth in milk processing. Optimal enzyme activity should be exhibited near the pH of milk which ranges from 6.5 to 6.8 (Lampert 1975}, since pH adjustments may precipitate milk proteins and lower the integrity of the product. When the physical characteristics of milk are altered upon heating, the pH will decrease {Walstra and Jenness 1984). Therefore, the enzyme selected should be characterized by a broad pH range to accomodate the wider range of pH milk would exhibit upon heating. Some organisms will produce different types of lactases, extracellular and intracellular, which behave differently when removed from the cell. Extracellular enzymes may endure relatively severe 10 environments and are expected to be more stable than intracellular enzymes. This is partly due to the glycosylated form of the extracellular enzyme which helps protect it from the environment. Therefore, extracellular enzymes may be preferred over the intracellular form for immobilization. Although all living cells produce enzymes, one of three sources- plant, animal, or microbe - may be favored for a given application. Commercially available lactases have been isolated from fungi (mums: nixeLand A. was). bacteria (£129.11). and yeast (figggglgmxggg lappig and Si figpgilig). The activity of E; gpli lactase is drastically reduced in milk relative to buffered lactase solutions {Morisi, Pastore, and Viglia 1973}. Yeast lactases rapidly lose activity at temperatures above 40°C. Due to a low pH optimum for fungal lactases, they are generally more useful in hydrolyzing lactose in acid whey than in non-acidified milk products. In a previous Michigan State University study {Powell 1988}, fungal lactase from 5‘ 91115; was selected based upon its temperature stability, insensitivity to ions found in milk, and a broad pH range. It was necessary to use this enzyme at conditions which reduced its maximum activity by 50%. This enzyme exhibits strong product inhibition, poor retention in the immobilized form without the addition of stabilizers, and poor overall reactor effectiveness. These results using A. ppyng fl-galactosidase suggest the need for alternative thermostable lactases, such as those from nonpathogenic, noncommercial organisms, i.e. Bacillus ppgggpphpxmpphilpg and figrppppppggpg phgxmpphilpgp Beta-galactosidase from hp spppxpghgxppphilpp is quite stable at 55°C with its maximum activity between pH 5.6 and 7.0; its activity is enhanced by ions normally present in milk. The size of this enzyme is expected to be equivalent 11 to a molecular weight (MW) of 530,000 (MW of fl-galactosidase from Sp phgxnpphilpg) {Greenberg and Mahoney 1982}. This larger enzyme is likely to have improved retention in the immobilized form. The thermostability of the enzyme from 51 steagothermophilus and the broad pH range over which it is active makes this enzyme suitable for use in the dairy industry {Griffiths and Muir 1978}. For the above stated reasons, this enzyme has been selected for use in this project. Improving the production of this enzyme will be investigated in this study. It should be noted that since this enzyme has not been used in food processing as yet, it has not been tested by the FDA as GRAS. IV. 821W Hollow fiber membrane bioreactors (HFMB) have been used for immobilizing enzymes (Rony 1971; Waterland, Michaels, and Robertson 1974}, animal cells {Knazek et al. 1972}, plant cells {Shuler 1981}, and microbial cells {Kan and Shuler 1978; Vickroy, Blanch, and Wilke 1982}. The scale-up of hollow fiber reactors with fi-galactosidase for operation in dairy conditions has never been investigated. The lack of lactose conversion data up to the desired conversion (70%) makes scale-up difficult since the response of the reactor to various substrate and product concentrations are unknown. The potential applications of enzymatic catalysis in numerous diverse areas of chemical technology has brought about a need to develop and characterize new techniques for enzyme immobilization, the design of efficient immobilized reactors, the testing of thermophilic enzymes in membrane reactors, and the evaluation of reactor performance at realistic industrial conditions. The work described in this thesis addresses some of these needs. Before discussing hollow fiber reactors further, an understanding of hollow fibers themselves is necessary. A hollow fiber (Figure 1.1), 12 Region 3 (Spongy Layer) 1 (Lumen) egion Region 2 (Ultrathin Membrane) Ja. l.‘ .m . an . urtwuv. Electron Micrograph of Single Hollow Fiber. .Figure 1.1 13 or ultrafiltration fiber, consists of a thin semipermeable inner ultrafiltration membrane (region 2), approximately 0.5 microns thick, and an outer supporting macroporous spongy layer (region 3) which surrounds a cylindrical void space (lumen-region l). The nominal MW cutoff provided by the thin semipermeable skin may range from 3,000 to 500,000 daltons. The development of new synthetic hollow fibers has become attractive for reactor design since the fibers can be used at temperatures near 50°C so that microbial growth is retarded {Horton 1982}. Membranes can be make from polymers such as polysulfone, polypropylene, polyamide, cellulose acetate, and others; each membrane material may have a different compatibility with an enzyme. In the work of Korus and Olson {1977}, a-galactosidase from figpillpg gpppzpphgpmpphilpg was loaded into two types of hollow fibers. They observed 50% activity losses in acrylic copolymer membranes (MW cutoff 50,000) over seven days due to leakage from the fibers. Alpha- galactosidase losses from polysulfone membranes (MW cutoff 10,000) seemed to result from inactivation of the enzyme. Korus and Olson {1975} had found that yeast fl-galactosidase rapidly lost activity on polysulfone fibers. It was found in a previous study [Powell 1988} that fi-galactosidase was more compatible with a polyamide membrane than polysulfone. Therefore, polyamide membranes were used exclusively in this study. Since the enzyme is not chemically bound to the support, this permits reuse of the membranes. Enzymes may be physically immobilized in the spongy layer by either static loading {Waterland, Robertson, and Michaels 1975} or backflush loading {Breslau and Kilcullen 1978}. In these studies, hollow fiber reactor cartridges consisting of approximately 100 hollow fibers were 14 used. In static loading, the shell-side of the cartridge is filled with enzyme solution that diffuses into the fiber's spongy layer. By repeatedly filling and draining the shell-side with stock enzyme solution, the concentration of enzyme in the spongy layer approaches that of the stock solution. This procedure takes a longer period of time to accomplish the same enzyme loading than by backflush loading. In backflush loading, pressure is applied to the shell-side, and enzyme stock solution is backflushed through the fiber into the lumen (Figure 1.2), i.e. solution is forced through the fiber in a direction opposite to that used in ultrafiltration. The enzyme is trapped within the fiber wall because of the thin semipermeable skin on the lumen-side and the air-space maintained on the shell—side during reactor operation. This method achieves higher loadings over shorter periods of time. After loading by either of the above methods, the stock solution remaining on the shell-side is drained from the hollow fiber reactor. Subsequent cross-linking of the enzyme with reagents, such as glutaraldehyde, may be used to retain the enzyme {Breslau and Kilcullen 1978}. The enzyme's retention in the hollow fiber matrix (spongy layer) will vary with the enzyme's molecular weight, fiber pore size, and loading method. Enzyme stability can be enhanced by the support provided by the hollow fiber matrix. Additionally, the presence of substrate is known to enhance enzyme stability. The retention and stability of fi-galactosidase in hollow fiber reactors. Single fiber reactors (SFRs) were used throughout this project to budget the limited supply of enzyme available. The single fiber reactor (Figures 1.3 and 1.4) for this study operates as follows: the substrate passes across the thin skin to react with the immobilized enzyme in the spongy layer. The products diffuse back into the lumen because the air- space at the outer surface of the spongy layer prevents their escape and Figure l.2: 15 'Shéll side - HF Tube side (lumen of HF) Single Fiber Reactor (SFR).- Shell material is boro- silicate glass, 22.0 cm effective active fiber, 0.8 cm o.d., tapered to 0.5 cm diameter x 1 cm length, end tubes with side tubes 0.5 cm diameter x 1 cm length; fittings illustrated on left were applied to both ends of the reactor: L= Male and Female Leur lock fittings; C= Silicone tube; HF= Hollow ultra- filtration fiber. The hollow fiber was retained by a plug of epoxy potting resin (Dow Chemical) between the Leur fittings and the hollow fiber. l6 Bypass Res Figure l.3: Single Fiber Reactor (SFR) System.- Res= Reservoir flask, P= Gear or peristaltic pump (Cole Palmer .model #7520-25 unified drive; Micropump #81281 head or Cole Palmer #7520-14 head); fl, f2= flowmeters (Cole Palmer #FM044-40 and #FMlOZ-S); pl= lumen-side inlet port; p2= shell-side inlet port; p3= lumen outlet port; p4= shell-side outlet port; tl= 15 psig pressure transducer (Omega #PXl42-OlSGSV); t2= 5 psi differential pressure transducer (Omega #PXl42-OOSDSV ); 0= outlet sample port. 17 - FIBLR LUMEN ACTIVE MEMBRANE SURFACE . - OUTER SUPPORT STRUCTURE (SPONGE LAYER). Figure 1.4: Photomicrograph of PMlO Hollow Fiber, Longitudinal View. 18 because of the concentration driving force. Only small molecules, particularly substrate, enter the spongy layer since the nominal molecular weight cutoff for the membranes used in this study ranged from 3,000 to 30,000 daltons (MW of lactose is 342.3). Membranes are commonly utilized in the protein ultrafiltration of dairy products and in the concentration of dairy products by reverse osmosis (Delaney and Donnelly 1977}. The superior heat and chemical resistance of synthetic hollow fibers are the reasons for their growing acceptance within the dairy industry. Furthermore, the design of the hollow fiber and the high shear stresses in the lumen may reduce fouling from milk solids (Yan, Hill, and Amundson 1979}. Proteases {Howell and Velicangil 1980} may also be used in a cleaning procedure to degrade the fouling protein. A hollow fiber reactor scheme protects the immobilized enzyme from the microbes in the substrate-product stream and from the beneficial high shear environment or the harmful effects of protease cleaning solutions. Whether the enzyme can be retained in the fiber or not, would have to be investigated. In conclusion, the use of hollow fiber reactors for immobilizing enzymes was selected for further study in this project due to the following advantages (Chambers, Cohen, and Baricos 1976; Powell 1988}: 1) Quick and easy preparation without the necessity of chemically altering the enzyme. 2) Ability to retain and reuse the enzyme. 3) The possibility of relatively small changes in the kinetic properties of the enzyme since it is assumed to reside in the spongy layer in its free-solution state. 4) Prevention of microbial and antibody access to the enzyme. 5) Capability of retaining the important physical and chemical characteristics of the reaction solutions. l9 6) Large ratio of surface area to substrate volume. 7) Continuous operation at low pressure. WW The behavior of an immobilized enzyme reactor utilizing asymmetric hollow fibers has been studied using a theoretical model based on differential mass balances {Waterland, Michaels, and Robertson 1974}. This model considers axial laminar flow and radial diffusion in the lumen, radial diffusion across the hollow fiber membrane, and the radial diffusion across the hollow fiber membrane, and the radial diffusion and reaction in the spongy layer. It predicts conversions and effectiveness factors using steady-state assumptions. The steady-state assumptions of their study can be summarized as follows: 1) A Poiseuille-type radial velocity profile in the lumen. 2) A first order reaction rate expression. 3) No flux across the spongy layer-shell-side of reactor interface; flux of substrate across the ultrathin membrane is equal on either side of the membrane. 4) System assumed radially symmetric with an initial condition of of the solution entering the reactor having a concentration of C . 0 5) Substrate diffusivity in spongy layer same as its free-solution diffusivity. 6) A tenfold higher resistance across the hollow fiber membrane than in the lumen due to interfacial resistance. 7) No mass transfer mechanisms such as bulk flow across the membrane due to pressure gradients; no axial diffusion. 20 The analytical solution to the Waterland et a1. model was presented in terms of a Thiele modulus, the ratio of the maximum possible reaction rate to the maximum possible diffusion rate, V 2 2 max a ¢ - (1.1) K D m 3 and dimensionless length, 2 Z ' —a_x— (1'2) where V a x -—‘L— (1.3) I’1 is the Peclet number, D1 is free-solution substrate diffusivity, a is the inner radius of the fiber, a is the inner radius of the fiber, V0 is the maximum flow velocity, 2 is the axial coordinate, vmax is maximum enzyme reaction rate, Km is the Michaelis constant, and D3 is the spongy layer substrate diffusivity. The assumed first order kinetics predict rapidly increasing conversions at constant dimensionless length (2) as the Thiele modulus varies from 10'2 to 10. As 2 decreases, outlet conversions decrease and approach an asymptotic conversion more slowly as the Thiele modulus increases, i.e. the shift to a completely diffusion-controlled regime occurs at higher values of the Thiele modulus. As the reaction approaches zero order kinetics, the range of Thiele modulus values for transition from kinetic- to diffusion-control becomes narrower and shifts to higher values. While this model accurately predicts experimental conversions, the calculations involved are very extensive [Kim and Cooney 1976} and unnecessarily rigorous, i.e. variations in 21 some of the parameters yield negligible changes in predicted conversions. Webster and Shuler {1978} have treated a hollow fiber reactor as a CSTR rather than a plug type reactor in their model. This model eliminates the consideration of axial and radial concentration gradients in the lumen. Using the CSTR type operation; however, presents the problem that a product-inhibited enzyme operates at the lowest catalytic reaction rate in a CSTR. Other models have been presented (Lewis and Middleman 1974; Davis and Watson 1985} which are modifications of the Waterland, Michaels, and Robertson model. Inhibition complicates the kinetic expressions and may require that the concentration distribution of the product in the system be incorporated into the model. The models all consider evenly distributed catalytic activity in the spongy layer. The application of the above models to the data obtained from the experiments performed for this thesis is complicated by kinetics and the method of fi-galactosidase immobilization. The modelling and scale-up methods used in this thesis will be addressed in detail within Chapter 6. VI- W The experimental program is part of a continuing study designed to obtain lactose-hydrolyzing hollow fiber reactor data relevant to dairy applications: The objectives of the experiments are to test the enzyme's stability, retention within hollow fibers, enzyme production, enzyme kinetics in a single fiber reactor (SFR), reactor performance, and the hollow fiber's susceptibility to fouling. The data obtained on the enzyme kinetics and the diffusion of reaction species are necessary for modelling. Dairy products such as skim milk, sweet whey, and whey 22 permeate were investigated as the primary substrates for the hollow fiber reactor system. The following chapters describe the methods and results for experiments that determine: Chapter 2) Chapter 3) Chapter 4) Chapter 5) Chapter 6) Enzyme Kinetics and Production Optimization - the parameters needed for preliminary modelling and assay of the behavior of enzyme in free solution; development of a procedure to optimize the production of thermophilic fi-galactosidase. Enzyme Retention and Stability - the study of the retention of backflush-loaded enzyme in a SFR; comparison of the stability of the enzyme in a SFR to that in free-solution. Hollow Fiber Characterization - measurement of the diffusion of substrate and product for modelling; check of the possibility of fouling from skim milk, whey, and whey permeate. Reactor Performance - operation of single fiber reactors with different enzyme loadings and flow rates to assess performance; determination of the retention and recoverability of the enzyme under actual operating conditions; testing of bolus flow to enhance reactor performance. Immobilized Enzyme Kinetics Modelling - modelling of the immobilized enzyme kinetics and reactor scale-up. This experimental program continues the work of Powell (1988} with new emphasis on fouling, substrate and product diffusion, the use of a thermophilic enzyme, reactor performance enhancement, modelling, and scale-up. Chapter 2 Epgype Kinetips apd Erpdpction Optimizatipn 1-W The objectives of the experiments discussed in this chapter were to determine the kinetic parameters needed in preliminary modelling of reactor kinetics, to assay the behavior of the enzyme in free-solution, and to optimize the production of thermophilic fl-galactosidase from the two strains of hp spppxpphpgmpphilpp (HR16 and HR18) donated by the Hannah Research Institute (Ayr, Scotland). In the course of these experiments, the improvement of the kinetic assay and glucose analysis techniques was accomplished, the existence of an unexpected lactose hydrolysis product was revealed, and the stability and kinetic characteristics of the enzyme in free-solution were investigated. Kinetics of immobilized enzymes may vary from their free-solution kinetics. The intrinsic reaction kinetics are necessary in modelling the performance of a reactor. Therefore, the fip ppppxpphgxmpphilpp fi- galactosidase kinetic parameters were studied at the temperature and pH of the reactor in free-solution and in the immobilized state. The immobilized enzyme kinetics are presented in Chapter 6. II. W In biochemical processes, it is common to model enzyme kinetics by the general inhibition form of the Michaelis-Menten equation: ds Vpgx s V " F ' s(1+i/Ku) + Km(l+i/Kc) (2'1) 23 24 which contains the parameters vmax’ the maximal enzyme reaction rate, Km, the substrate concentration at which the reaction rate is half its maximal rate, and Kc and Ku’ competitive and an uncompetitive inhibition constants, respectively {Stryer 1975}. The quantity vmax is equal to keo which is the first order rate constant (k) times the enzyme concentration in free-solution (e0). Additionally, s is the substrate (lactose) concentration, V is the initial reaction rate, and i is the inhibitor (galactose) concentration. Galactose, a product of lactose hydrolysis, inhibits fi-galactosidase activity while the other product, glucose, does not. When lactose is hydrolyzed, galactose occupies the active site while glucose is unreactive to this site. Equation 2.1 adequately characterizes the behavior of most enzymes studied. The equation above describes mixed inhibition. This equation can be used to describe competitive inhibition, uncompetitive inhibition, or a mixture of both. A competitive inhibitor is usually similar in structure to the substrate, capable of reversible binding to the enzyme active site, and its presence increases the value of Kn but does not alter Vm An uncompetitive inhibitor does not bind in the active site ax' of an enzyme, but binds at some other region of the enzyme molecule {Boyer 1986}. Upon binding of the uncompetitive inhibitor, the Vmax of the enzyme decreases without changing the value of Km“ If ku approaches infinity, the Michaelis-Menten equation may be reduced to the following rate equation which describes competitive inhibition: v - vmpx s (2.2) s + xm(1+1/xc) 25 For uncompetitive inhibition, Kc approaches infinity, Ku is finite, and the (1+i/Kc) term of Equation 2.1 is unity. Griffiths and Muir {1978} have studied the activity of Bi ppppxpphpxmpphilpg B-galactosidase at various values of pH, temperature, and in the presence of various ions. The optimum temperature and pH ranges of enzyme activity in the purified and immobilized whole cell form were found to be 54-6506 and 5.8-6.7, respectively. In the study described in this thesis, an optimal reaction temperature of 55°C was chosen due to the temperature limitation of the hollow fiber material. The highest optimal pH (6.7) was chosen to approximate the pH of dairy products. Since enzyme activity was found to be enhanced by ions normally present in milk, magnesium and manganese were included in the reaction solution for the experiments discussed here. In the Griffiths and Muir paper, the Michaelis and inhibition constants for the enzyme at 65°C were determined to be 2.06 mM and 20 mM, respectively. These values were used as estimates in determining the kinetic parameters at 55°C for our studies. Before the use of the thermophilic enzyme can become economically feasible, the improvement of its production from the organism is necessary. Presently only small amounts can be obtained with the method used by Griffiths and Muir {1978}. Modifications to their procedure were performed to optimize enzyme production. These modifications included different growth mediums, serial batch growths, and selection between the strains provided (HR16 and HR18). Wm; Lineweaver-Burk plots of the inverse of the reaction rate versus the inverse of substrate concentration (l/V vs. l/s) data yield estimates of the kinetic parameters from a linear plot, i.e. the value 26 of the ordinate intercept is l/Vmax and the slope of the line is Km/Vmax as shown by the following linear relationship obtained by rearranging a simple Michaelis-Menten rate expression: 1 K 1 l __m_ ' v '3‘ + v (2‘3) max max <| where s is the substrate concentration. The linear relationship of l/V versus l/s data also allows estimation of the inhibition constants, Ku and Kc, when the effects of inhibition are experimentally tested. The difference between uninhibited and inhibited enzyme kinetics are evaluated when the parameters describing the intercept and slope are represented as ‘ follows: For uncompetitive inhibition, __1___1_(,_11_,1) V . V (2.4) max(apP) max and competitive inhibition, K K i v_m.£3129_L_‘7.m_(K +1) (2.5) max(app) max c where Km and V have been determined in assays without added inhibitor. Conversely, K m(app) and V are determined in assays maX(aPP) with excess inhibitor. These equations were based on the Equation 2.3 relationship. The estimates from graphical straight-line approximations 27 are improved when statistical methods are used to estimate intercepts and slopes. The Wilkinson statistical data treatment was used to obtain more accurate estimates of the kinetic parameters from the experimental data. Wilkinson {1961} developed a method which involves linear regression analysis and weighting the data according to the nature of its accuracy. Residuals are obtained by subtracting the experimental and the calculated rates of reaction from the linear relationship. Outliers, or residuals, which are greater than two experimental standard deviations (suspect of being erroneous) are identified and removed from the data set. In this study, the estimation of enzyme kinetic parameters was analyzed using the Wilkinson method in a computer program called WILMAN 4 {Brooks and Suelter 1986}. W Confirmation of the accuracy of the kinetic parameters calculated as described above required that they predict batch conversions over time. The predictions of competitively inhibited reaction rate data were obtained from the integrated form of Equation 2.2: l K s s v {(so-s)(l-—l-(m) + xm(1+—KQ) In (19)) (2.6) max C C t- Initially, the use of Equation 2.6 and the kinetic parameters yielded a poor prediction of experimental batch conversions. Because of a suspicion of possible enzyme instability, a study of the stability of the enzyme in these batch conversions was performed so that enzyme decay could be included in the prediction. The enzyme decay was determined for each data point in the batch conversion versus time plot to accurately reflect the prediction of Equation 2.6. 28 III. MUN—11m Enzyme and the hp ppppxpphegppphilpp strains were donated by Dr. Mansel Griffiths of the Hannah Research Institute (Ayr, Scotland). Lactose (cat. #L3625), galactose (cat. #60625), and glucose (cat. #65000) were obtained from Sigma Chemical Company. W Assays for enzyme activity may be performed by a continuous method or by an endpoint method if the rate of reaction is linear over time. An endpoint assay was used since lactose and its hydrolysis products cannot be assayed directly during the enzymatic reaction. A continuous assay method was used when o-nitrophenyl fi-galactopyranoside (ONPG) was used as the substrate; the product of this reaction is o-nitrophenol (ONP) which is directly measured spectrophotometrically. All enzyme assays were conducted in 0.05 M potassium phosphate buffer (pH-6.7) which also contained 0.5 mM MgSO4 and 0.1 mM MnCl The 2. buffer composition approximated the ionic content of milk. Magnesium has been identified as an activation ion for the enzymatic hydrolysis of lactose. Manganese was not tested for this property. W The accuracy of endpoint assays for reaction rate constant determination requires that a linear relationship hold between product appearance and time over the duration of the experiment. This linear relationship was verified experimentally for up to 10 minutes for enzyme concentrations up to 0.218 mg/ml at various lactose concentrations with and without galactose. The endpoint method involves the initialization of the reaction with the addition of the enzyme to a reaction mixture containing a known concentration of substrate. The reaction is allowed to proceed for a 29 set period of time and is stopped by heat or by lowering the pH. In these experiments, the reaction was stopped in initial experiments by placing the reaction vessel in a water bath at 90-10000. Since a heat denaturing technique was found to give only crude estimates of kinetic parameters, a pH lowering method was used. An equal volume of 1.4% perchloric acid was added to the reaction mixture to instantaneously stop the reaction. The acid itself was suspected of hydrolyzing lactose, but this did not prove to be the case. After the reaction was stopped, glucose was analyzed to determine the amount of lactose converted in order to calculate units of enzyme activity. A unit is defined as the liberation of one micromole of product (ONP or glucose) per minute at 55°C and pH-6.7 under the conditions of the assay. When acid was used to stop the reaction, glucose determinations were performed by a phosphatase-glucose oxidase (PGO) method (Sigma Diagnostics, St. Louis, MO, procedure #510) in which the resulting solution of 0.5 ml analyte and 5.0 ml of a PGO enzyme/color reagent solution was assayed at 450 nm with a UV/VIS spectrophotometer (Perkin Elmer, Lambda 3A). The accuracy of a PGO analysis for glucose is only acceptable when the analyte solution's optical density (OD) is less than 0.8. The response of this analysis becomes nonlinear for OD. > 0.8; therefore, it became necessary to dilute samples to ensure accurate analyses. 0 This analysis also detects the presence of galactose, but at a lower level of sensitivity. In samples which contained glucose and galactose, the samples were corrected for the concentration of galactose by subtracting the response of the estimated galactose concentration. Galactose concentrations were estimated to be equivalent to the preliminary uncorrected glucose concentration. This technique was 30 verified by several HPLC analysis (described below) - PGO analysis comparisons. Preliminary enzyme assays used heat to stop the reaction and obtain a rough estimate of kinetic parameters. Glucose and galactose were analyzed using a Waters HPLC system (Waters 600 solvent delivery system; Waters 490 detector; Waters 410 differential refractometer) with a Bio- gel Sugar-Pak I column under the following operating conditions: water flow rate - 0.5 ml/min. refractive index cell temp.- 35°C column temperature - 90°C elution time for lactose - 7.41 min. glucose - 8.84 min. galactose - 9.81 min. The use of this method was limited by the ionic strength and pH of the sample. A guard column was installed to protect against the ions present in the buffered samples. This technique could not be used when the samples analyzed were highly acidic because it would be too damaging to the guard and Sugar-Pak I column. HPLC analysis revealed the presence of other carbohydrates in the product distribution. These were not studied further in this work. WW An easily performed continuous assay method, which used ONPG and ONP obtained from Sigma Chemical Co., was used when a relative enzyme activity between samples was investigated. The activity of £- galactosidase in samples were determined by following its initial reaction rate in 6.25 mM ONPG at 55°C and pH-6.7. Magnesium and manganese were present in the phosphate buffer as described previously. Assay mixtures consisted of 0.625 ml of 12.5 mM ONPG stock solution in 31 buffer, 0.025 to 0.625 ml of sample solution, and buffer to bring the total volume to 1.25 ml in the flow cuvette. Hot water from a water bath was pumped through the jacketed cuvettes to maintain the reaction temperature at 55°C when measuring absorbances. The temperature was monitored with a thermocouple. The reaction was initiated by adding the enzyme sample to the solution. Units of activity as defined previously were determined by monitoring the time to achieve an absorbance change of 0.1 at 420 nm. Lower absorbance changes were used for assays with extremely slow reaction rates. Linearity was checked periodically. The rates were calculated from these absorbances relative to a standard solution of product which consisted of 0.4 mM GNP in buffer. WWW Endpoint enzyme assays on reaction mixtures which ranged from 0.5 to 20 mM lactose were performed to approximate Km. Lactose concentrations used bracketed the expected Km estimated from literature (Griffiths and Muir 1978). An HPLC analysis was used to determine product (glucose) concentrations. Results from these initial experiments led to the refined method in determining the kinetic parameters. In the refined method, test tubes containing the reaction mixtures (Table 2.1) were placed in a constant temperature shaker water bath (New Brunswick Model G76D) at 55°C at least 10 minutes prior to initializing the reaction. After enzyme was pipetted into each tube, the reaction was initiated by vortexing each tube and then returning each to the water bath. The kinetic parameter experiments were conducted with enzyme concentrations of 0.059 and 0.118 mg/ml. Here, the reaction was stopped after the appropriate incubation time by adding an equal amount of 1.4% Table 2 . 1: 32 Reaction Mixtures Used ip the Dgtermination of Kinetic Parameters Total volume - 2.0 ml NOTE: For each case below, at least two experiments were performed with enzyme concentrations of 0.059 and 0.118 mg/ml. Case #2 With Inhibitor Case #1 Without Inhibitor Tube # (Lactose),mM 1 l. 2 2. 3 5 4 8 5 16 6 24 7 32 8 40 25 50 .00 .00 .00 .00 .00 .00 Tube # II III IV VI VII VIII (Lactose),mM 2.50 4.00 8.00 16.00 32.00 48.00 64.00 80.00 (Galactose),mM 10 10 10 10 10 10 10 10 33 perchloric acid to the reaction mixture to lower the pH to 1.2. Following inactivation, glucose concentrations were determined by the spectrophotometric (PGO) method described earlier. The maximum required dilution to ensure OD > 0.8 was 4:1. Since enzyme kinetics will differ for different enzyme sources, the kinetic parameters were determined for the one source of enzyme used in the reactor study. The kinetic parameters in all of these experiments were analyzed using the WILMAN 4 program. To calculate specific activities, protein concentrations were determined by the Lowry method (Cooper 1977). Win: The experimentally determined enzyme kinetic parameters were checked by their ability to predict batch reaction curves. Each batch run was initiated with the addition of a buffered-enzyme solution to a buffered-lactose solution to result in a 20 ml solution with 146.7 mM lactose. The reaction vessels containing lactose and buffer were held in a constant temperature water bath at least 10 minutes prior to adding the enzyme. The average specific activity of the enzyme stock solution was 3.2 units/mg. Enough enzyme stock solution was added to each vessel to yield lactase concentrations of 0.402 and 0.803 units/ml. These values were chosen to conserve the enzyme for future experiments. Immediately following addition of the enzyme, each vessel was shaken at 50 rpm and a 1 m1 sample was withdrawn with an Eppendorf pipettor to serve as a blank control. In each experiment, samples were withdrawn at 10, 20, 30, 45, 60, 90, 120, 150, and 240 minutes. The PGO method was used to determine glucose concentrations. The enzyme decay suspected in the above experiments led to parallel experiments which evaluated the effect of stirring on enzyme activity decay. Relative enzyme activity between samples was followed by the 34 continuous assay (PGO) method described previously. The enzyme activity decay results were incorporated into the model which predicted the batch results. This will be discussed later in this chapter. W The optimization of B, stearotpermophiipg fi-galactosidase production involved the selection of an appropriate growth medium, selection between two strains (HRI6 and HRIB), the evaluation of two cell-lysing techniques (addition of lysozyme of toluene/acetone), and the optimization of batch reaction sequencing. The optimum broths were chosen from L-broth, Penassay broth (PAB), and a modified PAB. The PAB broth was modified by the addition of lactose to stimulate the production of the enzyme from the genetic level. Modified PAB contained 1.5g yeast extract, 3.5g NaCl, 3.68g K2 HPOA, 1.32g KH2 P04, 0.1g lactose, and 8.0g nutrient broth per liter. Each strain of an organism may differ in its production of an enzyme due to the strain's genetics and environment. Therefore, HRI6 and HRI8 were evaluated for enzyme production in the growth media discussed above. The HRI8 strain was found to produce greater amounts of fi-galactosidase. The cell-lysing agents added to the cell lyse the cell membrane and possibly some enzymes. The latter possibility was checked. The first step in the optimization of enzyme growth was to determine which enzyme strain (HRI6 and HRI8 stored in -70°C glycerol) would produce the greatest amount of fi-galactosidase in the chosen medium, L—broth, PAB, or modified PAB. The strains were grown in 20 ml of each medium at a temperature of 55°C and stirring maintained by a shaker water bath. They were checked for cell density (growth) in each 35 broth after 2 hours by observing the broth's absorbance at 600nm in a Klettsummerson photoelectric calorimeter. Additional cultures were grown under the same conditions to determine an effective cell-lysing procedure. The cells were isolated using a Sorval Superspeed RC2 centrifuge at 9,000 rpm (15 mins.), and the cells were lysed with toluene/acetone and lysozyme to determine which agent would limit the obtainable enzyme activity. The cells were lysed with each method for 30 minutes. The supernatants were assayed for enzyme activity on a Gilford Model 240 spectrophotometer using an ONPG endpoint assay method. This method was similar to the endpoint method described earlier except the reaction mixture contained 6.25 mM ONPG and the reaction was stopped with the addition of 0.5 M sodium bicarbonate. Once the optimum strain and lysing method were determined, the method for large-scale batch growth was investigated. The age of the organism at inoculation was shown to be important in optimizing its growth without degeneration in progressively larger batches. An inoculation schedule was determined by checking the cell density and enzyme activity of each 20 ml of modified PAB serially inoculated with the organism at various ages. From this information and the procedure of Griffiths and Muir (1978), optimization and purification of the enzyme yield were possible. The purification of the enzyme solution involved the use of an ion-exchange column filled with DEAE-sepharose CL-6B (Pharmacia), a gel filtration column filled with Bio-gel A-0.5m, and ammonium sulfate for salt fractionation. protein levels were determined on a Gilford Model 240 spectrophotometer at 280nm. The optimal enzyme production procedure for a four liter batch operation is reported in the next section. 36 IV. W K ete During the course of this project, the source of enzyme varied. Enzyme was produced at MSU from the HR18 strain of Bi speagpthermophilus with the hope that enough enzyme could be obtained to complete the project. This was not the case. Since the amount of enzyme produced allowed only the completion of the retention and stability study (Chapter 3), another source of enzyme was used. Dr. Mansel Griffiths of the Hannah Research Institute provided enough intracellular and extracellular enzyme from the organism to investigate the reactor performance. Lineweaver-Burk plots for the intracellular enzyme were found to show good linearity both with (Figure 2.1) and without (Figure 2.2) inhibition. The kinetic parameters for the extracellular enzyme were not determined because the reactor performance was not studied using this enzyme in this thesis. Statistical analysis of the results by the WILMAN 4 program yielded mean parameter values and standard deviations as follows: For no inhibition, Km - 3.69 mM (S.D.-0.40) vmax - 2.08 units/mg (S.D.-0.06) With 10 mM galactose, K - 0. 4 . .- . m(app) 8 mM (S D 0 82) V - . 9 . .- . max(app) 2 0 units/mg (S D 0 16) 37 ‘1), Hal-I“ a. nu nu .9.1 9v cu nu . Ll. nu m. a... / nu. . ,W N... m. he . b w T :11: 1 Cl 1.0L. mmmcsm N.Eu a 02..» 30: m mxpmmimzam “Tu 5253318 Benson q T). a . . w 1 . . .l. lob 0.0 PM CL. Pm Pa ._.O n ._l11_4 ._.N ._.a. ._.m fm Nb a\ roofiomm 00383320: A35 raamamm.m 0. A20 Harmamdomuw 38 ) m 02.) 23: 1.. mxmmmEqum mu . I §_E:mos.m 3826a _. nu U a; .. a» S O I. t nu an... an .111 nu. . 7m “WI ....w.. C . up. w. 1.. / .. OI - a — q _ . _ . — . l0.» 0.0 0.» 0.5. 0.0 0.0 a .0 a\ rooaomm oozooscdzo: A373 mmocsm m.mu rmsmmmm 23: N mxuém 0 mamcsm m.u” d — d to no mm .5 mo mo no mo u: - u — q - d — a 9:00.86 Oozoozsdmo: A35 xmmnamoz mmflm a: mono: mmmnnaozm Ao.mowmm czmnm mmamummdmnnomaammm\smv. xmmnflmoa zmxacxm nosnmazm 2: man go. 45 Verification of the kinetic parameters permits their use in reactor kinetic modelling;however, the kinetics of enzyme in free-solution man change when the enzyme is immobilized. This was the case in this work. Chapter 6 discusses the determination of the immobilized enzyme kinetics. o u t 0 at 0 Of the two strains supplied by the Hannah Research Institute, HRIS was found to produce the largest amount of enzyme in a modified PAB. The HRI6 and HRIS strains grown in L-broth, and PAB and HRI6 grown in modified PAB exhibited a lower amount of enzyme produced. It was also found that the use of lysozyme to lyse the cells was less destructive to the enzyme than using toluene/acetone. The optimization of enzyme growth was determined by observing the relative Bi pgppzpphgzppphilpp growths in serial batch cultures (Figure 2.8). All of the serially inoculated batches reached an optimum growth between 6 and 7 hours regardless of the cell density of each inoculum. It is at this time that an optimum enzyme growth in larger batch volumes can be initiated without proceeding to the death phase of the organism which decreases the enzyme yield. As a result of these experiments, an experimental protocol (Table 2.2) was developed which yielded the optimal conditions for enzyme production. A sequence of three batch growths (6-7 hours), each near its optimum production, was necessary to obtain approximately 2100 units of unpurified enzyme activity from a total batch volume of 4 liters. The purification of the crude enzyme solution was carried out according to the methods described by Griffiths and Muir (1978). After purification (gel filtration and ion-exchange), the resulting 110 m1 of enzyme solution had only 1140 units of activity which provided enough enzyme for less than 20 experiments. The resulting solution had a 46 mooa a ) 1 p s / a \We/ K n u j ( 1 D won. 0 4 _ m. e. e . D .1 l. mg +1. .309 :63 5:03 A0 N.» 33.00030 .mw xlx .300. $03 .303. A0 q :3.00I.o£.::oo$ C 4 I .300. #03 5:06» A0 a ambulumuasoouv I 5300. ‘33 «Snow A0 N rabolamuisoomv . I .3300. :63 lqo 0 3626. “$9.30 M u _ u _ — u - -|l d i|_ c a N u a m m u m o _.0 ad .336 Arocav mmccxm N.mn mmwmm. wmnmdzcm mnmmxonsmxaouxmdcm meet”: a: mean: nczacxmm. >Ld ma szoncumiosm om 53131.6 no: c333: 47 protein concentration of 1.79 mg/ml. Although the first trial of this protocol was quite successful, its second trial was not as successful. It was necessary to repeat the serial batch culture study. The time and effort expended and the cost of enzyme preparation and purification must be drastically improved in order to make the thermophilic enzyme feasible for large-scale industrial use. W The HPLC analysis of the reaction products from the enzymatic hydrolysis of lactose revealed the existence of an additional product besides the peaks for lactose, glucose, and galactose. This may be explained by the following hypothesis: when lactose is hydrolyzed by 8- galactosidase, galactose occupies the enzyme's active site. The galactose-enzyme complex can react galactose with another lactose molecule to form a trisaccharide. This mechanism has been suggested by Professor Prakesh Dey (MSU-private communication). The presence of a trisaccharide could explain the peak observed before the lactose peak in an HPLC analysis. The rate at which this phenomena occurs was observed to be small. Its occurrence was dependent on having a high concentration of the substrate present, so as the amount of lactose decreased, the rate of formation of the trisaccharide decreased. In an industrial setting where the 50-70% conversions of lactose are desired, the contribution of this phenomenon would be negligible. At the highest observed rate of trisaccharide formation, the resulting error for the calculated rate of lactose conversion was found to be very small. A further investigation of this additional product was not undertaken for the purposes of this study. 48 Table 2.2 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) ll) 12) 13) 14) e oduct o o Inoculate 20 ml of modified Penassay broth with a single colony from -70°C glycerol medium of Bapiilug stearptheppophilus (Hannah Research Institute, Scotland. HR18 strain). d ed ssa roth mP B 8.0g Nutrient broth 1.5g yeast extract 1 l l l l l i 1 1 3.5g NaCl Dissolve in 900 ml distilled 3.68g KZHPO4 H20 and autoclave. 1.32g KH2P04 ---Add 100 ml 18 lactose solution which was filter sterilized - expected pH-7.22. Grow at 55°C (aerated - stirred @ 100 rpm). After 6-7 hours, transfer 10 ml of culture to 200 ml mPAB. Grow at 55°C for 6-7 hours (aerated). Transfer 80-100 ml portions of new culture to two 2 liter batches of mPAB. Grow at 55°C for 6-7 hours (aerated) - follow growth optimum at 600 nm - for example see Figure 2.8. At growth optimum, centrifuge cultured broth at 10,000 rpm - 12,000 rpm for 15 mins. - discard supernatant. Resuspend pellets in 100 m1 of 4 mg/ml solution of lysozyme in 50 mM potassium phosphate buffer, pH-6. 7. Incubate at 37°C for 30 mins. to lyse cells. Centrifuge at 10,000-12,000 rpm for 15 mins.-discard pellet. Filter supernatant through 0.2 micron filter (filter sterilized). Bring resulting cold (0-4OC) solution to 70% saturation with ammonium sulfate. Centrifuge at 13,000-15,000 rpm for 10-15 mins. - discard supernatant. Resuspend pellet in 10-15 ml of 50 mM potassium phosphate buffer, pH-6.7. 49 121119.211 (COM) 15) Dialyze sample in buffer for 48 hours, changing 1,000 ml stock every 16 hours. 16) Apply sample to buffer equilibrated slurry of DEAE-sepharose CL-6B (bed volume 10 m1) and mix "sample-slurry" mixture for at least 30 mins. 17) Prepare an ion-exchange column with a 0-2 M KCl gradient (it was found that fi-galactosidase elutes between 0.64 and 0.56 M KCl). 18) Collect fractions with high specific activity. 19) Prepare a (2.5 x 50 cm)) Bio-gel A-0.5m column. 20) Apply 10 ml of sample to the gel filtration column and collect high specific activity fractions (if sample volume is too large - reduce volume by dialyzing over polyethylene glycol crystals). 21) Filter sterilize final solution to reduce microbial growth. Chapter 3 e ete a d tab t I. ngroductipp The interaction of an enzyme with its immobilization support is an important factor in the effectiveness of a hollow fiber membrane bioreactor design. The enzyme's activity and/or stability may be enhanced or reduced by such an interaction. The objectives of the experiments described in this chapter were to determine 1) if the enzyme can be retained and recovered from the hollow fiber matrix and 2) how long the enzyme remains active within the matrix. W Backflush loading was selected as the method for the immobilization of enzyme in SFRs. Based on the work presented by Breslau and Kilcullen (1978), Powell (1988), and Jones, Yang, and White (1988), this method offers the advantage of being a rapid method of loading and gives the ability to achieve higher enzyme concentrations than static loading. With this immobilization technique the recovery of enzyme and reuse of the support is also possible. The polymeric structure and large molecular weight of B*_§§p§;pphgxmpphiip§ fi-galactosidase was expected to improve its retention in polyamide fibers without the need of adding binding or stabilizing agents to increase the steric hindrance effect as was done in previous work in this research program {Powell 1988}. It was also suggested during the previous study that use of a gear pump for enzyme loading may denature the enzyme due to pump cavitation. A syringe pump was used in this study to prevent any enzyme loss due to the pumping technique. 50 51 The selection of the hollow fiber material was determined in previous experiments by Powell. These experiments showed an incompatibility of B-galactosidase from 5* pgyng with polysulfone hollow fibers, while the enzyme was relatively unaffected by polyamide. The retention experiments described in this chapter compared polyamide fibers, PA10 and PA30, with nominal molecular weight cutoffs of 10,000 and 30,000, respectively. The fibers were evaluated for retention of protein and intracellular (removed from within the pi ppppppphppppphiipp cells) enzyme activity, recoverability of enzyme, and enzyme inactivation over time. Enzyme inactivation was compared between continuously stirred batch reactors and the enzyme immobilized in SFRs. The extracellular form of this enzyme concentrated from its growth medium was also included in the immobilization stability study. The recoverability of enzyme and enzyme leakage was investigated during the operation of the SFR (Chapter 5). WW MW Single fiber reactors (Lo et a1. 1978) were prepared using PA10 and PA30 hollow fibers; a SFR is shown in Figure 1.2. All fibers were donated by Romicon, Inc. For the construction of a SFR, a hollow fiber is placed through the fittings, tubings, and the glass tube shown in Figure 1.2. Epoxy potting resin (Dow Chemical) was placed inside the cup of a female Leur lock fitting prior to joining it with the male fitting. By joining the two fittings, the epoxy filled the spaces between the hollow fiber and the inside wall of the fittings. This bonding method, which results in the separation of the shell- and lumen-sides of the SFR, was found to be stronger than any previous construction method used on this project at MSU. 52 The SFRs were cleaned and sanitized of any bacteria and extraneous proteins before the enzyme was loaded onto the fiber. The cleaning procedure used is described in Table 3.1: Table 3.1: e ess Time Beam; 30 mins. 0.7% phosphoric acid solution 15 mins. distilled water 30 mins. 0.5% potassium hydroxide solution 15 mins. distilled water 20-30 mins. 200 ppm sodium hypochlorite 15 mins. distilled water These reagents were pumped through the fiber using a Cole Palmer pump drive (model #7553-30) with a model #7514-20 pump head. Rinses included recycling through the lumen and ultrafiltering through the spongy layer. After cleaning, the SFRs were tested for leaks by pressurizing the lumen to 15 psig with air from a syringe. Enzyme loading and unloading was accomplished using a Sage Instruments syringe pump (model 341A). Just before the enzyme immobilization process was performed, the hollow fiber was equilibrated with 0.05 M potassium phosphate buffer, pH-6.7. The reactor system shown in Figure 1.3 consisted of the SFR, feed reservoir, flowmeter, pump, shaker water bath, and 1/8” diameter insulated tygon tubing to carry liquids. The hollow fiber reactor was operated using a Cole Palmer pump which pumped the reservoir fluid through the flowmeter, the SFR, and back to the reservoir. The valving around the SFR permitted the reactor fluid to flow in several modes: (1) through the shell-side or the lumen, (2) by ultrafiltration - from lumen to the shell-side, (3) by backflushing - from shell-side to the lumen, or (4) in a combination of the modes above. 53 II. ta e d W The experiments described in this section were conducted to determine the retention of enzyme immobilized in the SFR by backflush loading. Before loading the enzyme, the SFR was cleaned, sanitized, and flushed with buffer as described earlier. The shell-side of the SFR was filled with the enzyme solution, and the shell-side loop then was closed so all the solution would be forced into the hollow fiber lumen by the backflush loading technique. A back pressure of approximately 17 psig on the shell-side was achieved using a syringe pump for all loadings. Loadings of 16 to 30 units of enzyme were accomplished. The solution forced into the lumen, the backflush effluent, was saved for protein and enzyme analyses. The shell-side and fiber surface was rinsed with 30 m1 of buffer, and the rinses were saved for analysis. To test the leakage of enzyme into the lumen and the activity of immobilized enzyme, the lumen was filled with a buffered lactose solution, and the SFR was stored at 0-4°C overnight. No enzyme leakage was observed during this operation. The enzyme was then removed from the fiber wall by ultrafiltering the enzyme from the fiber with a solution of buffer with lactose added for stabilization. Four 8 m1 ultrafiltrate samples were collected for analysis of enzyme activity and protein content. Each enzyme loading was followed by the same cleaning, sanitizing, and flushing procedure. The concentration of the enzyme solution varied between a one- to three- fold dilution of the enzyme stock solution (6.4 units/mg and 1.8 mg/ml protein, initially). These experiments were performed on SFRs of either PA10 or PA30 fibers. The amount of enzyme 54 loaded was equal to the enzyme activity backflushed into the fiber less that detected in the backflush effluent and removed by rinsing of the shell-side. All enzyme and protein analyses used were as described in Chapter 2. e tab t Enzyme stabilities in the immobilized and in the free-solution form were compared. For free-solution results, enzyme stability was investigated in batch reactions gighpp; lppgppg by the same method as described in the previous chapter. The same units of enzyme were used as in the SFR study. This study proved an effective means of comparing the stability of the enzyme in free-solution to its stability in the SFR. Intracellular and extracellular forms of B-galactosidase were separately loaded into two similar PA10 SFRs. Twenty-seven units of the enzyme were loaded in each SFR after the fibers were conditioned by the cleaning-sanitization-flushing method described previously. Enzyme activity was checked upon loading and throughout a period of 30 days by periodically operating each reactor with 50 ml of a 146.7 mM buffered- lactose solution for one hour. The SFR system was operated with a 9 ml/min feed flow rate using a peristaltic pump (ISCO). The SFR was held to 54.5+/-0.5°C by submersion in a water bath. The conversions were determined using the PGO enzyme method and were used as an indication of enzyme activity. In operating the system as a reactor, enzyme loading was assumed to equal the enzyme activity backflushed into the fiber less that detected in the backflush effluent which was negligible. To again confirm that a negligible amount of enzyme leaked into the substrate solution during SFR operation, half of each reaction sample was incubated in the water 55 bath until the next sampling time. Lack of significant additional conversion confirmed the absence of leakage. III. C a i uss on W The comparison of the enzyme "activity balances" between PA10 and PA30 fibers are reported in Table 3.2. Both fibers exhibited the same percentage of enzyme activity retained in the fiber (approx. 65%), but the PA30 fiber had results with the most variability. For example, the percent loss standard deviation for the PA30 fiber (14%) is an indication of the greater variability compared to the PA10 fiber (7%). Specific activities were calculated from the results of the protein and activity analysis presented in Tables 3.2 and 3.3. The specific activity of the enzyme in the PA30 fiber allows filtration of extraneous proteins to a greater degree, thus providing a purer form of the . immobilized enzyme and improving the reactor performance. This does not appear to have occurred in the PA10 fiber reactors. However, because consistent enzyme loadings could not be attained with the PA30 fibers, probably due to the selective filtering of the extraneous proteins, the PA10 fibers were chosen for further experimentation. The lower percent loss of protein for each fiber (Table 3.3) compared to the percent loss of enzyme activity (Table 3.2) implied a lower degree of denaturation than inactivation, since the degree of experimental error was similar in each case. The greater variability of percent loss results for the PA10 fiber cold be due to differences in the size of the proteins in each solution loaded. The PA30 fiber was not as selective for various sizes of proteins and was unaffected by variations in the loading solution. 56 Iabip 3,2: Activity Retention of Bi fipppppphgpppphilps Beta-Galactosidase in Two Different Polyamide Hollow Fibers Retention Fiber Type 13mm 32112 283.9 % Specific activity1 100.7 (S.D.-19.3) 123.8 (S.D.- 8.2) % Activity retained in fiber2 68.9 (S.D.- 5.5) 61.5 (S.D.-14.2) % Activity leakage upon loading3 0.6 (S.D.- 0.5) 4.5 (S.D.- 2.0) a Loss of enzyme activitya 24.9 (S.D.- 7.0) 30.8 (s.o.-13.9) Average amount enzyme retained in fibers (units) 22.02 18.94 1 Activity of enzyme per mg protein in ultrafiltrate as % of stock solution unit activity. Activity of enzyme in ultrafiltrate as % of expected maximum activity loading. 3 Activity of enzyme in backflush effluent as % of expected maximum activity loading. Activity of enzyme in ultrafiltrate as % of the amount of enzyme activity calculated by activity balance: cuvu _ ] x 100 COVO (cbVb + csVs) % Loss - [ l - c- concentration u-ultrafiltration sample V- volume o-loading solution b-backflush loading effluent sample s-shell-side flush sample 57 Iabig 3.3: Total Protein Retention from Bi Stearothermophilus Beta-galactosidase Solution in Two Different Polyamide Hollow Fibers Retention Fiber Type Iafermasien £819 £839 % Protein retained in fiber1 60.7 (S.D.-12.8) 49.1 (S.D.- 8.7) % Protein leakage upon loading2 7.9 (S.D.- 6.4) 22.7 (S.D.- 2.7) % Loss of protein3 19.1 (s.p.-1s.5) 18.4 (S.D.-11.2) Average amount of protein retained in fibers (mg.) 2.67 2.72 l Ultrafiltrate protein as % of expected maximum protein loading. 2 3 Ultrafiltrate protein as % of the amount of protein calculated by mass balance: Backflush effluent protein as % of expected maximum protein loading. c V % Loss - [ 1 - u “ 1 x 100 coVo - (cbVb + csvs) c- concentration u-ultrafiltration sample V- volume o-loading solution b-backflush loading effluent sample s-shell-side flush sample 58 The retention study did not use the reactor operation temperature (54.500) but was conducted at 25°C. In SFR operation, it was found that the higher temperature and possibly the use of a different enzyme stock source may have caused the enzyme to adsorb on the hollow fiber matrix. Only about 30% of the enzyme could be recovered by ultrafiltration during the experiments run at 54.50C. The retention of enzyme within the same SFR was improved with each additional loading. The first loading on a new SFR provided retention results quite different for results from subsequent loadings. This was thought to be due to the pre-conditioning of the fiber by the first loading. Some adsorption of the protein from a previous run may have accounted for the increasing retentions. Walling The enzyme stability of B_L spppppphgpppphilpg fi-galactosidase in batch reactions without lactose is presented in Figure 3.1 as the percentage of the initial enzyme activity over time. Figure 3.1 shows the enzyme decay over time. When Figure 3.1 is compared to the same study with the presence of lactose (Figure 2.5), it was concluded that the stabilizing effect of lactose cannot prevent the enzymatic decay due to the increased destructive forces upon stirring (50 rpm). It was observed that higher concentrations of enzyme were not as prone to these forces than were lower concentrations. Immobilized enzymes might be stabilized by the combination of the low interfacial (shear) forces and the stabilizing effect of lactose. The extracellular and intracellular forms of the enzyme were evaluated for differences in stability. The intracellular form showed a 30-day half life in storage on the hollow fiber, while the extracellular form retained 68% of its initial activity after 30 days in the hollow fiber. The extracellular enzyme also showed better recovery from the 59 hollow fiber; however, the extracellular form is more difficult to obtain in adequate amounts from cultures. Both the intracellular and the extracellular forms seemed to prefer storage in a buffered-lactose olution. Although the intracellular form was adequate in its stability, improvements could be make in hollow fiber membrane bioreactor operation if improved enzyme techniques could be found for the extracellular enzyme. 60 Percent of Initial Enzyme Activity 8 I MIN oboumm c::m\3_ I oboamnm c3553. 0 maccwm w.un d - 1 — id a mo 15. J we. moo . moo j3o A33c3mv mz~kam mdmuuzmak a: mean: zmmnnaosm zanzocn ransomm. mama xmnmu as «us. Chapter 4 o ow Fi e C a c e at I. o u o In order to develop a mathematical model for the immobilized enzyme reactor (Chapter 5), it was necessary to obtain values for the diffusivities of reactant (lactose) and products of hydrolysis (glucose, galactose) in the hollow fiber membrane. This was determined by measuring permeabilities of the species through the hollow fiber wall. Fouling of hollow fiber membranes was also studied in the context of possible decreased membrane permeabilities due to fouling. This chapter describes the experiments and analyses performed to obtain the necessary data. II. Ibsen For mass transfer from the tube-side to the shell-side of a hollow fiber membrane, the overall membrane permeability resistance, R0, is broken down into component membrane and fluid film resistances in series: Ro - Rm.+ RLS + RLT (4-1) or in terms of mass transfer coefficients, K: _1_ L 1 1 + + K0 Pm KLs Kit (4.2) where Pm is the membrane (spongy layer) permeability, LS designates the shell-side liquid film, and LT is the tube-side liquid film {Smith et 61 62 a1. 1968}. In the experiments described in this chapter, these resistances contributed to the overall membrane permeability since the shell- and the tube-side flow rates were not large enough to negate the effects of the film resistances. Diffusivities The necessary model parameters, the diffusion coefficients, may be calculated from the overall permeability (KO) as follows. Using the wet membrane thickness as the diffusion path length (0.0445 cm), the effective diffusivities are calculated as Deff- Ko * wet membrane thickness. Since glucose and galactose have the same molecular weight, the diffusivity of galactose may be assumed to be the same as for glucose. . Most permeability studies for hollow fiber membranes have calculated the overall permeability by following the concentration changes between the tube-(lumen) and shell-side of the hollow fiber in a dialysis mode of operation {Colton et a1. 1971; Farrell and Babb 1973; Kim and Chang 1983). In this mode, the substrate flows through the tube-side, diffuses across the fiber wall, and is collected in a "blank" solution flowing on the shell-side. Material balances between the tube- side and the shell-side were used to calculate the overall membrane permeability for the transported species. In this work, the overall membrane permeability (KO) of lactose and galactose for a hollow fiber membrane was calculated with the material balance equation used by Park, Kim, and Chang {1985): m-(V + V ) 3 1n [V 5 g 9—1-—E°i(v +V)t (43) t (sto ' ssa) vt Vs t s 63 where m - stVt+sSVs and Vt and Vs are the volumes of solute solution in the tube- and shell-side, respectively; st and sS are the concentrations of solute in the tube- and shell-side; sto and 330 are the initial concentrations of solute in the tube- and shell-side; A is the mass transfer area of the hollow fiber membrane - 10.45 cm2;; and t is the time of sample. Since these are all known or measured quantities, the terms -ln [(m-(Vt+VS)cS)/(Vt(cto-cso))] versus time may be plotted (example-see Figure 4.1). The slope of the plot is equal to KoA(Vt+Vs)/(vtvs) as shown by Equation 4.3. The overall membrane permeabilities and diffusivities of lactose and galactose are then calculated directly. The permeability studies in this project were conducted without the presence of the immobilized enzyme and with the same flow rates used in the PA10 SFR performance studies. To reduce experimental error, possible non-equilibrium transport effects and the hydrostatic pressure differences between the shell- and the tube-side of the SFR were minimized. Eoplipg Fouling has been known to be a problem in the ultrafiltration of skim or whole milk through hollow fiber membranes (Harper 1980} and may cause serious deterioration of HFMB performance. The physical characteristics of the "fouling” material, i.e. pH, proteins, lipids, calcium salts, and phosphate interactions, will greatly affect the degree of fouling possible {Muller and Harper 1979}. The protein- protein and protein-lipid interactions in milk coupled with the properties of the hollow fiber materials may cause fouling. Therefore, the possibility of fouling with skim milk, Dagano cheese whey, and whey permeate in the polyamide fiber was studied in the context of its effect 64 Negative Log Expression 0.00 0.00.. 0.00.. 00.9... 0.00.. 0.00.. 0.._ 0.. 0.. 0... 0.:1 0; MI 0... 0.. 0.004 0.00.. 00?. 0.00.. X meBEo «amaze 0.001. mmocwm a.zu - — d ‘00 moo emsammcmdanx omamwaaamamoz amen. 2.8 .23m A33cfiomv « a~00 000 65 on membrane permeability. Fouling was defined in these experiments as a decrease in the mass transfer (permeability) of lactose across the fiber. In the experiments, the "fouling" solution (i.e. skim milk, whey, whey permeate) was changed every two hours to maintain a high driving force for fouling throughout the experimental run. The permeability values calculated for fouling solutions should not be considered as the true permeability of lactose since possible non-equilibrium transport effects were not controlled for each fouling solution, and the non-isothermal exchange of the "fouling" solution reestablished a new steady state operation with every change of solution. The purpose of these studies was only to determine changes in observed permeabilities from one solution to the next. During this study, a fiber was also intentionally fouled by ultrafiltration to determine if the fiber was susceptible to fouling under extreme operating conditions. III. W I: eat In the determination of the permeability/diffusivity of lactose from the tube-side to the shell-side, sucrose was used on the shell-side to reduce the possible effects of osmotic pressure. The presence of sucrose in the experimental sample rendered the use of HPLC analysis of the sample difficult since sucrose could not be separated from the lactose in a Sugar-Pak I column. Therefore, a Park-Johnson reducing sugar analysis {Cooper 1973} was used for lactose instead. This method for determining reducing sugars is based on the reduction of ferricyanide ions in alkaline solution by a reducing sugar (lactose). However, a high sucrose concentration (146.7 mM) on the shell-side 66 caused the Park-Johnson method to be inaccurate due to the need for high dilutions and the precipitation of the buffer ions during the analysis. Whenever possible, the use of a buffer was avoided. In the determination of the permeability/diffusivity of glucose, galactose was used on the shell-side to eliminate the possible non- equilibrium transport effects. A PGO method was used to analyze for glucose concentration as described in Chapter 2. The fouling studies were performed by following the concentration of lactose in the dialysate (shell-side) solution. A shell-side sucrose solution was not required because fouling could be quantified as a relative decrease in the mass transfer of lactose over time and not as the decrease in the true permeability. This made it possible to determine the concentration of lactose by HPLC analysis. b fu v t erime These experiments were designed to evaluate the permeability of substrate and product through the hollow fiber wall. The SFR was used in a dialysis mode in which the test species was circulated through the tube-side, and a ”blank" solution was circulated through the shell-side to collect the diffusing species. The dialysis configuration is illustrated in Figure 4.2 which shows a counter-current flow pattern through the reactor. The equipment used was the same as described in Chapter 3. Pulse dampeners were placed upstream of the reactor to reduce pulsatile flow. As described above, it was found in preliminary experiments that the potassium phosphate buffer caused precipitation problems in the Park-Johnson and PGO analyses. Since buffer was not essential in these experiments because enzyme was not present, all sugars were dissolved in doubly distilled filtered water only. One PA10 SFR was used throughout each series of experiments. Prior 67 Outlet Pressure Indicator Rotametsr V 3‘ Pump Temperature Control Bath Reservoir Reservoir Tube Side LOOP I lShell Sidel < LOOP 2 5122M Inlet Pressure Indicator Pi. P‘I SFR Figure 4.2: Dual Closed-Loop Dialysis System for Membrane Permeability. 68 to starting each experiment, the SFR was cleaned, sanitized, and equilibrated in filtered distilled water using the cleaning procedure described earlier. After the water bath temperature had achieved 54.5+/-0.5°C, the shell- and tube-side solutions were pumped through bypass lines in each independent flow circuit until equilibrium in temperature and flow rate was reached (approx. 10 mins.). The diffusion of a 146.7 mM solution of lactose or glucose was in the direction from the tube-side to the shell-side. The shell-side initially contained an _equimolar amount of sucrose or galactose for testing with lactose or glucose, respectively. Once equilibrium was achieved, the solutions were allowed to pass through the SFR. The hydrostatic pressure difference between each side of the reactor was monitored (see Figure 1.4) and minimized by creating back-pressure on the lower pressure side. This was done to avoid any convective flows due to a transmembrane pressure gradient. Samples were taken over a 4 hour period and analyzed. The experiments were repeated at a variety of flow rates. The shell-side flow rates ranged from 3.6 to 18.2 ml/min while the tube-side flow rates were fixed at values between 6.0 and 12.6 ml/min. The resulting diffusivities for lactose and galactose (glucose) were averaged for the optimal flow rate used in the SFR performance studies. Fo men Initially, it was thought that gross, macroscopic fouling along the lumen of the SFR could be observed. Experiments were run in which the "fouling" solution was recirculated at 9 ml/min through the SFR, and the pressure drop down the fiber was measured to determine if fouling was causing an increase in resistance to flow. Several fibers were dissected and inspected microscopically for evidence of fouling. Because of the lack of evidence of macroscopic fouling, it was decided 69 that fouling was best studied by following the permeability of lactose from the tube-side to the shell-side of the SFR. The general apparatus and procedure for these experiments were the same as the permeability/diffusivity experiments with the following modifications: 1) The shell-side solution contained the reaction mixture buffer to maintain the pH of the tube-side "fouling" solution (skim milk, Dagano cheese whey, or whey permeate). 2) The tube-side "fouling" solution was analyzed for its lactose concentration. All solutions were corrected for any lactose deficiency to maintain a 146.7 mM concentration of lactose in each. 3) The shell- and tube-side flow rates were fixed at 18.1 and 9.6 ml/min, respectively. 4) In some experiments the periodic replenishment of the tube-side solution caused non-isothermal operation. This led to erroneous results and was corrected by equilibration the fresh solution at the operating temperature before exchange. 5) The effect of doubling the time of the cleaning procedure was was investigated. 6) Except for the control case experiments, samples were taken over an 8 hour period. 7) A 4 hour control case experiment was run with 146.7 mM lactose in buffer (tube-side) against buffer (shell-side) for each SFR used. The same SFR was used for each series of experiment with the same "fouling" solution. Since our results indicated that no significant fouling occurred, to determine if fouling was possible in extreme conditions, an additional experiment was performed by Mr. Sean McGee, an undergraduate 70 independent study student. A control case and a fouling study using skim milk were performed on a SFR according to the normal procedure. Afterwards, skim milk was ultrafiltered through the SFR to force fouling of the hollow fiber. A permeability study was then repeated and compared to the control experiments. IV. s s 0 Fe ea fus v ud The permeability and diffusivity values for lactose and galactose through a PA10 hollow fiber are shown in Table 4.1. The SFR from this study was used as representative of the characteristics of other PA10 fibers, although the characteristics of other PA10 fibers may vary slightly. The large standard deviations in these values (Table 4.1) are due to experimental error. Experimental error may be attributed to hydrostatic pressure differences between the tube- and shell-side and the possible changing properties of the PA10 fiber from one experiment to the next. A probable reason for errors in the lactose results may be inherent in the difficulties experienced in the Park-Johnson analytical technique discussed earlier. Table 4.1: KW Solute Species We: 1.8.9.032 Mums). Overall Membrane Permeability (cm/min)x103 1.15 (s.p.-o.47) 2.49 (s.p.-0.59) Effective Diffusivity Deff (cm2/sec)x106 0.86 (s.p.-0.35) 1.85 (s.p.-0.44) Deff (lactose)/Deff (galactose) - 0.46318 71 Was In these studies, fouling was characterized by a decrease in the permeability of lactose through the hollow fiber wall. Fouling would have been characterized by a decrease in the slope of the negative log expression (Equation 4.3) versus time plot. During the fouling studies, the slope was observed to change after the first two hours of some experiments. This was attributed to an initial unsteady-state period discussed earlier and not to changes in permeability. Whey permeate contains very little protein and was found not to cause fouling. This is illustrated by comparing the slope of the control case (with a "clean" lactose solution) to the slope exhibited after two hours of operation for whey permeate (Figure 4.3). Dagano cheese whey was tested for fouling as shown in Figure 4.4. It was suspected that a non-isothermal exchange of whey disturbed the steady state conditions and drastically reduced the transport of lactose. After switching to an isothemal exchange of whey, the whey did not exhibit signs of fouling. The transport of lactose was greater than observed for the control case. This may have been caused by a combination of.the experimental errors discussed above. From Figure 4.5, it is difficult to determine if skim milk had fouled the fiber. The control case and one of the fouling experiments had demonstrated equal permeabilities of lactose resulting in the conclusion that fouling did not occur. It was also found that another experiment showed a lower permeability to indicate that fouling might have occurred. It was suspected that the cleaning procedure was not effective between each experiment. After doubling the cleaning time, the same reduced permeability was observed. To eliminate any doubt, a hollow fiber was intentionally ”fouled" with skim milk by ultrafiltration to force protein against the lumen .72 9.5 n .0. 0.00.. S S e . m. cm Pom- 9 . .. o L e Poet .w t n a m. is. N O “macaw e.wu N 23183230. 98330 x 203183930. nxozoaoo x 025.2 name . 8o 2.5 2.5 So d3m A33c$mv ssmk emwammam «osmium macaw. emxammcadmflk omflmxauamaaos when. 000 73 0L0L .9 23183330. 983300 a - w 53:23a— nxozoaoo . _. n .. x 633030. 883260 u .m . x 00326— oowo u S e 0.1—NI I rl p .. w x n w E I A I nus o Pom- L 1 I e I .W . u LL .. e 0 e e e e g 0.00.. e o a» N . a 1 X a x .. 0.00:1 . a . J . _ . _ 1 0 ._00 ~00 000 .5.00 000 .230 A33cfiomv mamewm e.eu cmomzo nzmmmm zsmk mocdmzo macak. emsammcmdmnk cmnmsamamfimoa anon. 74 . 0 $032.30. 08.40000 . N 2001003230. 008.3030 n .. x 2001003230. 00830000 .m 0...0.. x .. zoal.0030_.3\0000.0 0.00050 . S + $039.30. 0830000 . me. L 4 0033. 0000 n? X 9.»: Fr. . a e Q g .. O - L C e 0.00.. . .W . . LL .. O I m. o o... - e . . . . .. u a . . N n m u w 0 0.2L . _ . _ . _ . _. 0 .00 ~00 000 000 n.0010 e.mu .230 A3358. mx.a z..x m0:..:c mecak. 0013000...a< c0~013.=mn.0= v.0". 000 75 wall; the permeability results were compared to those of an unfouled fiber. The control case and the "fouled" hollow fiber exhibited similar permeability values of lactose (Figure 4.6) to definitely indicate that fouling did not occur. The experiments discussed above indicate that polyamide fibers are resistant to fouling for all dairy feed streams tested. This is probably due to the relatively high shear rates in the fiber lumen and demonstrates a clear advantage for the use of hollow fibers in dairy stream processing. 76 .. N 0300.00 30:02 300: J a. 100.00 30:02:00.. u n x 0002.0. 0000 . olOl I I S I S nY.N e .. F D. . x I E .. I g 0.001 x nu II. . u e .. u x {N .. LL . m. 0.9.... e 1 x N I . - d d o.oo+...._...._l. 0 d 1 d. I 1 d mo . 00 .00 N00 .430 “3.3:”.0mv n.0c10 0.0” mxnmaama mx.a z..x «0:..30 mncak. 0013000...fi< cmnmxa.:mfi.0= 0.0”. 4 - .Nwo. 1 a d 000 Chapter 5 low The primary focus for this thesis work was to evaluate the performance of h, ggggxgthggmgphilgg B-galactosidase in a hollow fiber reactor. The experiments described in this chapter investigated the effects of enzyme loading and flow rate on SFR performance, compared the reactor performances of the thermophilic immobilized enzyme to the mesophilic immobilized enzyme (A. ggyzgg fi-galactosidase), and studied the use of bolus flow to enhance reactor performance. The retention of thermophilic enzyme under actual reactor conditions was also investigated. The flow rates and enzyme loadings examined were equivalent to the range of values used in the study of a mesophilic enzyme reactor performance (Powell 1988} so that a direct comparison between the two enzymes could be make. As in Powell's study, the flow regimes selected were near those recommended by Romicon, Inc., to prevent fiber fouling during ultrafiltration. In order to avoid fouling due to cross-membrane pressure differences (ultrafiltration) and to increase the single-pass residence time in a reactor, lower flow rates (laminar flow) than recommended were used. An average flow rate of 6.5 ml/min was used in most experiments described in this chapter. The comparison of reactor performance using the two enzymes was made by observing differences between the effectiveness factor versus generalized modulus {Moo-Young and Kobayashi 1972) values for modelling diffusion and reaction in porous media. 77 78 Studies to evaluate the enhancement of thermophilic enzyme lactose hydrolysis by using bolus, or air-segmented flow, instead of conventional homogeneous flow in the lumen were designed to improve the rate of reaction by increasing substrate transfer rates into the spongy layer. Bolus flow was found not to enhance reactor performance. The theoretical development of effectiveness factor and generalized modulus expressions used in the data analysis and the basis for bolus flow experiments are described in the following section. The experimental apparatus and results are discussed in the Materials and Methods section. The effect of intraparticle diffusion in artificial membranes on the kinetic behavior of immobilized enzymes has been modelled by Moo- Young and Kobayashi {1972}. Their model is based on immobilized enzymes which obey Michaelis-Menten kinetics. The effectiveness factor used in their model is defined by Equation 5.1. The physical interpretation of a generalized modulus is the ratio of reaction rate to diffusion rate. E- w t (51) Total rate at bulk substrate conc. only ' In their derivation of the effectiveness factor and generalized modulus, Moo-Young and Kobayashi assumed the enzyme to be evenly distributed within the reactor with negligible changes in substrate and product diffusivities and enzyme kinetics due to the immobilization of enzyme. In this study, free-solution enzyme kinetics were used 79 initially in calculating effectiveness factors. the true effectiveness factors were determined in work described in Chapter 6 where the immobilized enzyme kinetics were investigated, and intrinsic enzyme kinetics could be used in calculating effectiveness factors. Moo—Young and Kobayashi {1972} expanded the form of a generalized modulus, m, {Bischoff 1965} for use with competitively inhibited enzyme reactions: m__..h_*_]__ (5.2) 51 + 32 J12(1.0) where, 1 13 +19 12(1.0)- 2 * [.92 - 191 In 47341 (5.3) 52 1 ._Vm_ h'zoss*L (5.4) .E.EQ. _Zmax Vm - Vc - Vc (5'5) with 131 - a1 (1 - w 02) (5.6) £2 - l - 6 al 02 (5 7) and K p 01 - 'gm— ; w - -;—'+ 6 (5-3) s D 02 - x , 6 -—DL (5.9) c P In these definitions, Ds and Dp represent diffusivities of the substrate (lactose) and product (glucose or galactose) within the fiber. Ds’Dp' and their ratio (6) are reported in Table 4.1. Lumen substrate 80 concentration (3) and the initial substrate concentration (30 - 146.7mM) 'were used to calculate the product concentration, p - so - s. The kinetic parameters vmax’ Kc’ and Km have been defined and quantified in Chapter 2. Equation 5.2 yields a value of the modulus for a flat sheet geometry; therefore, modulus values were corrected by defining a new characteristic length (L) as the ratio of catalyst volume to lumen surface area [Froment and Bischoff 1979}: (5.10) This adjustment is most accurate for first-order reactions at m>>l. For the sake of comparison only among different regimes of reactor operation, the effectiveness factor-generalized modulus values calculated for this chapter were acceptable. They will be re-examined in Chapter 6. We! Bolus flow is a form of two-phase flow in tubes in which the flow of one phase is separated by segments, or slugs of the second. Bolus flow occurs in the human cardiovascular system where blood cells separate plasma segments in capillary flow; the plasma between erythrocytes does not move in simple laminar flow, but exhibits an eddy- like motion which circulates the fluid between the cells (Middleman 1972}. This is thought to enhance the radial mass transfer in blood. The benefits of this phenomenon in enhancement of mass transfer in hollow fiber reactors may be simulated by segmenting liquid flow in the fiber lumen with sections of air flow. 81 In previous work on lactose hydrolysis in hollow fiber reactors {Powell 1988}, the reactor's lactose-hydrolyzing reaction was found to be limited by the diffusion of substrate into the enzyme region as indicated by high modulus values. The enhanced radial transport caused by bolus flow was thought to be a good mechanism with which to improve the rate of the hydrolysis reaction. In the work of Horvath, Solomon, and Engasser {1973), the benefits of (slug) flow over homogeneous flow for an enzyme mounted on the inside walls of a tubular reactor is illustrated in Figure 5.2. It was found that bolus flow did increase the rate of reaction for their enzyme reactor over what was exhibited under homogeneous flow conditions. Bolus flow was characterized by its slug length and the frequency of the air and liquid section flow. Horvath et al. showed that radial mass transfer increases rapidly with decreasing slug length and increasing slug frequency. These parameters were varied in the investigation for this thesis. 111. mm The immobilized enzyme SFR was used to investigate the effects of l) varying the flow rate in the lumen, 2) varying the amount of enzyme loaded in the spongy layer, and 3) bolus flow on SFR performance. The general method for assembling and operating the reactor system was similar for all experiments described in this chapter. The method is described in the following section. Wen The cleaned, sanitized, flushed, and buffer-equilibrated system described in Chapter 3 was drained of buffer before operation. The system was loaded with enzyme (see Chapter 3) and the reservoir was filled with 150 ml of 146.7 mM lactose solution in potassium phosphate buffer (pH-6.7) with Mn and Mg added to simulate a dairy feed stream. 82 0.6 *- O.$ *- Ill .0 .e l homogeneous Now 0.2 - F; or W [ml/min] Figure 5.2: Graphical Illustration of the Dependence of the Normalized Overall Reaction Rate, R, on the Total Flow Rate in Homogeneous, Fh’ or Slug Flow, w, through a Straight Tube. 83 The reactor and fluid reservoir were held at 54.5+/-0.5°C in a shaker water bath (New Brunswick Model 6760). To assure constant temperature in the reactor, all external tubing was insulated. The substrate stream was preheated by passing it through a 50 cm by 5mm (o.d.) glass coil immersed in the bath immediately upstream of the reactor. After all ports to the reactor were closed to isolate the reactor, lactose solution was circulated through the recycle and bypass loops for approximately 15 minutes. The reservoir was emptied by opening the sample port and was refilled with approximately 100 ml of fresh lactose solution. After steady-state was achieved (about 15 mins.), the substrate solution was allowed to flow into a graduated cylinder, and the rate of filling was measured to determine flow rate. Except in the experiments in which the effect of flow rate on reactor performance was specifically examined, all experiments were conducted with a flow rate of 6.5 ml/min. Reactor operation was started by opening the lumen-side ports of the SFR and closing the bypass loop which allowed the solution to be recycled. Three milliliter samples were collected during operation by simultaneously closing the recycle loop and opening the sample port. Sample volumes were recorded. Each sample was then divided into two equal portions which were dispensed into disposable polypropylene centrifuge tubes. In order to test for enzyme leakage into the lumen, one portion of the sample was added to an equal volume of 1.4% perchloric acid to inactivate any enzyme that might have leaked into the lumen. The other portion was incubated in a water bath at the reactor temperature and was deactivated by acid at the next sampling-time. By comparing glucose concentrations in the two portions of each sample, enzyme leakage into the lumen could be detected. Since no significant 8.4 differences between glucose concentrations in the two portions of any sample were detected, negligible enzyme leakage was confirmed. At the end of each experiment, the fluid remaining in the system was drained into a graduated cylinder. All but one ml of the fluid was forced from the system by blowing it out with air. This total residual volume was then recorded. Total initial volume in the system and volumes during sampling intervals were determined by adding sample volumes to the residual volume. Elgw Rage Experimen§§ To evaluate the effects of substrate solution flow rate on reactor performance, product concentrations over time were compared at flow rates of 3.25, 6.5, and 13.0 ml/min in a reactor loaded with 27.5 units of thermophilic enzyme. Samples were collected at 0, 30, 60, 90, 120, 150, 180, and 240 minutes. W A series of experiments was developed to examine the effects of varying the amount of enzyme loaded on lactose conversions. Sample were collected under the same schedule as discussed in the previous section. Thermophilic enzyme loadings of 12.10, 25.48, 30.51, 37.71, and 51.02 units (or equivalently enzyme concentrations in the spongy layer of 25.2, 53.1, 63.6, 78.6, and 106.3 units/m1 based on annular volume) were used at a flow rate of 6.5 ml/min. These loadings corresponded to mesophilic enzyme loadings of 0.8, 1.7, 2.5, and 3.4 mg/ml {Powell 1988). A comparison between the mesophilic and thermophilic enzyme SFR performance results was made and will be discussed later. Da£a_Irsa£msn£ In the experiments described above, data were initially acquired in the form of glucose concentration versus time. Since the volume of solution in the system varied among experiments and changed with each 85 sample collection, comparisons among experiments required transforming sampling time into an mean residence time in the SFR. Mean residence time (r1) for each sampling interval was calculated according to the following equation: V t ' C _ _L_ 1 H where VLr the lumen volume (0.216 cm3), Q - volumetric flow rate of substrate solution, and vsi - volume of fluid in the system during the time interval (t1 -t1 _ 1). Product concentrations were plotted against T 1. A comparison between the mesophilic and thermophilic enzyme SFR performance results was also made using their effectiveness factors and generalized moduli values. The effectiveness factors for the thermophilic enzyme reactor were calculated by dividing the experimentally observed rate of reaction by the predicted rate of reaction at the substrate and product concentrations in the lumen. The predicted reaction rates were generated from the free-solution kinetic parameters discussed in Chapter 2 and the competitive inhibition rate equation (2.2). To enable the comparison of catalytic performance between the two enzyme types, a FORTRAN program (Appendix - Program A) was written to calculate the generalized modulus from diffusivity (as described in Chapter 4) and kinetic constants. The program logic is presented in the flowsheet in Figure 5.1 (Appendix). Equations 5.1 through 5.10 were 86 used to calculate effectiveness factors and generalized moduli based on free-solution kinetics. ‘ WHERE The bolus flow experiments were conducted at the optimum thermophilic enzyme loading of 63.6 units/cm3. All pretreatments were the same as described in the reaction operation section (homogeneous flow) except for the following modifications to the SFR system: the feed stream was pumped from the reservoir through a pulse dampener into a heat exchanger tube at 55°C. From this tube, the solution flowed into the bubble-generating system shown in Figure 5.3. This bubble- generating system created alternate segments of air and liquid by the controlled injection of air into the flowing feed stream. The segmented flow then passed into a viewing tube where the characteristics of the bolus flow, such as slug length and frequency, could be measured. From the viewing tube, the stream entered the SFR and then returned to the reservoir where the air segments were allowed to separate and liquid sampling was performed. Air flow rates were determined using a soap- bubble air flowmeter. The characteristic air segment lengths and the air and liquid flow rates were measured for each experiment. The largest possible experimental range of air segment lengths and frequencies were used as shown in Table 5.1 in the next section. Each experiment was run for 1.5 hours after achieving a steady- state condition (approx. 20 mins.) with either homogeneous or bolus flow. With a total reservoir of 51 ml of 146.7 mM buffered-lactose solution and constant liquid flow rate of 10.7 m./min, the resulting bolus flow conversions were compared to the homogeneous flow results. The same single fiber was used in each experiment to maintain geometric and functional consistency. The desired improvement in lactose conversion achieved by varying 87 ‘ from i from gas tank with pressure regulator (low motor to viewing into C-«— 0.0-0- - reactor MICROVALVE THICK-WALLED TEFLON TUBE STAINLESS STEEL NOZZLE - TEE INJECTION TUBE VIEWING TUBE Figure 5.3: Bubble-Generating System. C —'l (I) O 23 '0 u 88 the characteristic air segments lengths and frequencies was not observed; bolus flow did not cause enough of an enhancement in mass transfer to increase lactose transport to the fiber wall. A possible alternative reactor performance enhancement method is discussed in Chapter 7. IV. cu on Witness: Varying flow rates yield small differences in conversion with time (Figure 5.4). Lack of variation indicates that tube-side diffusion does not constitute a significant mass transfer resistance in the SFR over the range of flow rates examined. Since several of the models described in Chapter 1 consider tube-side resistance (Waterland, Michaels, and Robertson 1974; Kim and Cooney 1976}, our experiments suggest that consideration of radial concentration gradients on the tube-side may not always be necessary. This was also confirmed by the bolus flow experiments. Since a dramatic optimum flow rate was not observed, a flow rate of 6.5 ml/min was selected to allow a comparison between this study and the results from a mesophilic enzyme reactor performance study {Powell 1988). Wm The rate of glucose production increased at greater mesophilic and thermophilic enzyme loadings in the SFR as shown in Figures 5.5 and 5.6. An optimum enzyme loading was determined for each study as the loading which produced the smallest incremental change in conversion per unit of enzyme loaded. This was done in order to conserve enzyme. The optimum mesophilic and thermophilic enzyme loadings were determined to be 6.4 and 20.6 mg/ml, respectively. The lower attainable purity of 89 . . x 30: u 9N0 3.\3.: W no: x 39.. u 0.0 33:5 .. m ... N 302 ..l. .90 3.\3.: u ( 1 m 3.... n MU .. x 0 . Lrnu I n AMI I e 1 n C .. M... . . .. C 0.. .. e ... . S x m 1 N m A}... aux G . x A I 0.... _ . 7 . . _ . . 0.0 0... 0b 0.u 0.0 0.0 :00: $003080 .230 A33. m.mcxm m.0" mfifimnw 0m <01k.=m m.0z xmflm 0: n0=<0xm.0z .: max. 43013003...n m=~kam. 0.0 90 - H 7.4 mg/ml a H 6.4 mg/ml ‘ x—x 2.9 rug/ml .. o—e 1.0 mg/ml «n—I- 0.8 rug/ml 15'0“ e—e 0.4 mg/ml A .. IE ////' E '1 V , c a o :3 d 5 10.0- / 5 .. 0 d c o o . . o d 3 o 2 .. (D {SJD-i [I ”- _ ‘ / // 4 ,1 " / . .l I/ '1 d _7. 000 f I' T "rrI'I‘T'I'I'I T r 0.0 0.1 0.2 0.3 0.4 0.5 0.5 0.7 0.8 0.9 1.0 Mean Residence Time (min) Figure 5.5: Effect of Varying Beta-Galactosidase Loading on Conversion with Time in SFR. Mesophilic Enzyme. 91 00 x mm.» c3.~0\3. \M) N 00.. c3.~0\3_ m L a. 00.0 0320\3. {\ x <00 0320\3. n u . .090 03.n0\3. . :0. N01 t nu r t nn 6 I. C e u n .. o I C 8.. . e U 8 AV n» m I. . G Q d — I ‘i d _ CL -. q - .11 .l W 0.0 0... 0.» 0.0 0;. 0.0 0.0 oh {003 20050300 4.30 A3.3v v.0:10 m.mu mmwmnn 0* <0x<.3m wmnm-mm.mnaom.ammm romq.3c 0: ooz<0sm.03 3.0: 4.30 .3 mm». 43013003...n m3~ * ca) (6.1) where Q(l) and Q(2) are the curve-fitting parameters and CC represents the micromoles of glucose produced. Once the model had been verified as a best fit, the design of a large-scale hydrolysis system was performed. The SFR was modelled as a batch reactor in which the feed stream was recycled through the hollow fiber reactor until a desired conversion was achieved. This was justified because of the short residence times of the lactose solution within the SFR. The rate of lactose conversion in the SFR was modelled as follows: 99 Lactose V d[8] [3] feed vma conversion - —‘ - (- -E)[ __x ] (6.2) rate [e] Acart dt [5] + K mc(l+i/K ) where the intrinsic reaction kinetics in the right-hand parentheses are multiplied by the effectiveness factor, E, to describe the overall kinetics of the reactor. Enzyme concentration loaded per unit surface area of hollow fiber cartridges is represented by the term [e] (mg/ftz), [s] is the bulk lactose concentration (same as 3 used in other chapters), i-so-s, and Acart is the total cartridge surface area required to achieve a given conversion. Total surface area was used as a basis for scale-up since most membrane manufacturers sell cartridges based on total surface area requirements. Equation 6.2 may be rearranged and integrated in the following form: 3 ds [e] A t f [s 1 - ---951§ f dt (6.3) so feed 0 [( E)[[s] +mfix mc(1+1/x )1] Integration of the left-hand side of the expression was carried out from so-lh6.7 mM to the final desired concentration dictated by the desired end use of the product, e.g. s-73.4 mM for 50% conversion or s-36.7 mM for 75‘ conversion. The right-hand side of the expression, called the design time, has units of min*mg/ml. Depending on what in the process is fixed and/or known, the remaining design variable(s) may be calculated directly from the integrated results. 100 All predicted effectiveness factors were approximated by the following Moo-Young and Kobayashi expression to ensure negligible deviations from the actual E: _1_+_T.ésh_(.m)_ m m E - 1 + '61 (6.4) where m is the generalized modulus and fil was previously defined by Equation 5.6. Equation 6.4 represents an effectiveness factor for intermediate ranges of 81. III. als An experiment was performed with 18.28 units of intracellular thermophilic enzyme loaded in the SFR. Data were collected and analyzed for a 48 hour period. An extended time period experiment was necessary so that an accurate determination of the intrinsic kinetics could be accomplished. A suboptimal enzyme loading was used because of a. shortage of enzyme. The same materials and methods described in Chapter 5 were used to ensure identical conditions. Samples were taken hourly during the following time intervals: 0, 2-4, 16-21, and 42-48 hours. They were promptly analyzed by the PGO method described in Chapter 2 and analyzed as described in the next section. IV. W The data from the extended reactor time experiment were plotted as time versus micromoles of glucose produced (Figure 6.1). The data were curve-fit as described in Section II of this chapter. The curve-fit expression and its parameters are given in Table 6.1. 101 88 < u m3 .. mxunwfiv .. so + mg . mm n 8698 m» u 0.08 uooo: \sz \a. m - \ nu \\ .m \ m NOOOI \\ ( \ \ . e \\ m . \x F. 88.. \k....\.. \\\ lw\-\u\\\\\\ AguanN\\. _ a _ . _ . .fix . ._ q . . o Boo 88 «.80 38 mooo 88 303323 9. genome twoacoma V000 «mocwm m.mn cmflmxaaamnaoz om m zozurazmmx xmmxmmmao: >3m0osam INITIALIZE VARIABLES (N = Number of Data Sets) CALCUIATE GENERALIZED monams (Equations 5.2 - 5.10) 000 00000 00000 000 50 121 Program A PROGRAM THIELE.FOR Determine inhibitted enzyme reaction generalized modulus from kinetic and diffusion parameters. Generate effectiveness factor versus generalized modulus results. DIMENSION PI(100),EFF(100),EO(100) OPEN (60,FILE=’P£RF.DAT.’,STATUSz’OLD’) OPEN (61,FILE=’THRJK1.DAT.’,STATUS=’NEW’) OPEN (62,FILE=’THRJK2.DAT.’,STATUS=’NEW’) OPEN (63,FILE=’THRJK3.DAT.’,STATUS=’NEW’) OPEN (64,FILE=’THRJK4.DAT.’,STATUS=’NEW’) OPEN (65,?1LE-‘THRJK5.DAT.',STATUS=’NEW’) OPEN (66,FILE=’THRJK6.DAT.’,STATUS=’NEW’) INITIALIZE VARIABLES M=7 EKM=3.6884 EKENZ=2.0837 EKI=7.1647 808146.67 VOL=O.4801 SA=7.7132 DS=0.8554E-O6 ZETA=0.46318 READ DATA (3 of DATA SETS WHICH INCLUDES EFFECT- IVENESS FACTOR, GALACTOSE CONC., AND ENZYME (AMOUNT LOADED. READ (60,*) N DO so 1:1,N READ (60,*) EFF(I),PI(I),EO(I) CONTINUE DO 1oo,1=1,N CALCULATE GENERALIZED MODULUS VMsEKENZ * E0(I)/VOL . SI=SO-PI(I) VP=EKENZ*SI/(SI+EKM*(1+PI(I)/EKI)) EL=VOLISA HaELsSQRT(VM/(120*DS*SI)) AisEKM/SI A2=SIIEKI w=(PI(I)/SI)+ZETA Bi=A1*(1+(W*A2)) BZsi-(ZETAiAinAZ) E12=1/(82**2)*(82-(BiiLOG((Bi+BZ)/Bi))) DV=SORT(E12) TM=(HI(Bi+BZ))*(1/DV) 122 Program A (cont.) C WRITE EFFECTIVENESS FACTOR/GENERALIZED MODULUS C RESULTS TO FILES. C IF (I.LT.M+1) THEN WRITE (61,!) TM,EFF(I) ELSEIF (I.LT.2*M+1) THEN WRITE (62,!) TM,EFF(I) ELSEIF (I.LT.3*M+1) THEN WRITE (63,*) TM,EFF(I) ELSEIF (I.LT.4*M+1) THEN WRITE (64,*) TM,EFF(I) ELSEIF (I.LT.5*M+1) THEN WRITE (65,») TM,EFF(I) ELSEIF (I.LT.6*M+1) THEN WRITE (66,!) TM,EFF(I) ENDIF 100 CONTINUE CLOSE (60) CLOSE (61) CLOSE (62) CLOSE (63) CLOSE (64) CLOSE (65) CLOSE (66) STOP END 000 0000 0000 000 00000 50 25 123 Program 8 PROGRAM MODVER.FOR Generate an effectiveness factor versus generalized modulus prediction model based on free-solution kinetic parameters. OPEN (61,FILE=’MODVER.DAT.’,STATUS=’NEW’) INITIALIZE VARIABLES EKM=3.6884 EKENZ=2.0837 EKI=7.1647 SO=146.67 VOL=0.4801 SA=7.7132 DS=O.8554E-O6 ZETA=0.46318 GENERATE GENERALIZED MODULUS AND EFFECTIVENESS FACTOR PREDICTION VALUES. DO 25 184,18,2 DO 50 J=0,70,10 E031 PI=J CALCULATE GENERALIZED MODULUS VMsEKENZ * EOIVOL SIcSO-PI(I) VPsEKENZ*SI/(SI+EKM*(1+PI(I)/EKI)) EL-VOL/SA HaEL»SQRT(VM/(120*DS*SI)) A1=EKMISI A2=SIIEKI ws(PI(I)/SI)+ZETA Bi-Ai*(1+(W*A2)) BZsi-(ZETA!A1*A2) EIZsi/(82**2)*(BZ-(BiiLOG((BI+BZ)/Bi))) DV=SQRT