The objective of this study was to understand the mechanism of lipid fouling in chromatography through the investigation of a hydrophobic interaction chromatography (HIC) operation. This was motivated by the need to understand this phenomenon during the manufacture of biological products such as vaccines. The systematic approach and novel analytical techniques employed create a unique platform to study fouling of other chromatographic adsorbents and process feed materials. HIC is employed as a primary capture step in the purification of yeast derived hepatitis R surface antigen (HBsAg), where the required cell disruption and detergent liberation steps release high levels of lipid content into the feed stream. From lipid- rich and lipid-depleted feedstocks, comparative analysis was able to quantify the deterioration in HIC performance (binding capacities, purities and recoveries) under successive cycles. Furthermore, a full mass balance on host lipids identified the highly hydrophobic triacylglyceride as the main foulant. Intra-particle distribution and progression of lipid fouling and its effects on material adsorption and diffusion were then examined under confocal laser scanning microscopy (CLSM). In addition, high- resolution scanning electron microscopy (SEM) images of the fouled bead (after 40 cycles) confirmed that a thick lipid layer was building up on the outer bead surface. Based on these findings, the mechanism of fouling was thought to be the rapid accumulation of lipid foulant at the rim of the bead, which was aggravated by the possible diffusion hindrance resulting from multi layer adsorption. Finally, pretreatments to reduce this mechanism of chromatography fouling were evaluated in terms of improvement on feed quality and HIC performance. Selective adsorbent polystyrene XAD-4 demonstrated promising lipid removal capabilities with satisfactory HBsAg VLP recoveries. The improved feed into the column resulted in a three-fold increase in product capacity, whilst the overall yield remained constant over 40 cycles.

In this thesis, a multiplexed micro-fabricated cell culture device is presented to probe the soluble micro-environment of stem cells. A large number of biological, physical and chemical variables determine the micro-environment of stem cells and affect therefore their fate. The soluble micro-environment is believed to play a pivotal role in controlling stem cell fate and its optimisation may therefore allow exploitation of applications such as regenerative medicine or drug discovery. However, novel tools are required to probe and optimise the soluble micro-environment. The ability to move small volumes of liquid and the minimal use of resources are important characteristics of microfluidic systems and thus, they are perfectly suited to study the micro-environment of stem cells. A microfluidic cell culture device must ful fil three important requirements to be of utility in stem cell bioprocess development. The first aspect addresses the need for adaptability and flexibility to implement changes in microfluidic designs to take account of the rapid progress in stem cell research. To this end, a packaging solution specifi cally designed for a microfluidic cell culture device has been developed. The packaging system has thus been complemented with a rapid fabrication method using a micro-milling machine to quickly fabricate disposable and easily reconfi gurable microfluidic chips. The packaging solution has a maximum burst pressure of approximately 7.5 bar and the fabrication method has a dimensional fidelity with less than 10% deviation from the nominal value. The second aspect focuses on the scalability and comparability of results and the feasibility of continuous culture of stem cells - both critical elements for the success of a microfluidic cell culture system for stem cell bioprocess development. A novel cell seeding method has been developed using a pipette to directly and carefully seed cells into a culture chamber within a microfluidic cell culture device. To prevent a wash-out of viable cells during continuous culture, a low hydrodynamic shear stress microfluidic chip has been developed. The cell culture device has been successfully tested using human embryonic stem cells (hESC) on feeder cells. The third aspect concerns the automated monitoring of stem cell bioprocesses in the cell culture device. A multiplexed micro fluidic bioreactor platform with time-lapse imaging has been developed to obtain data-rich experimental sets. The platform consisting of a cooled media reservoir and a pumping mechanism has been characterised and tested using mouse embryonic stem cells (mESC) as a proof of concept. The combination of these three aspects provides a basis towards a multiplexed microfabricated cell culture device, which allows data-rich experimentation and comparability with current benchscale culture vessels for stem cell expansion and di fferentiation.

Oxidative biocatalysts have a vast industrial and biotechnological potential in areas such as fine chemical and antibiotic synthesis. They offer an environmentally compatible and sustainable route to catalysis, often simpler and more specific than chemical alternatives. However, the routine use of biocatalysts in biopharmaceutical manufacture has been hindered by biocatalyst complexity and the experimental burden necessary for implementation. This thesis aims to investigate, using automated microscale technologies, how oxidative biocatalytic bioprocesses can be designed and developed at a reduced cost and timeframe compared to conventional laboratory scale experimentation. A robotic platform was used with 96-Deep square well microtiter plates to develop an effective bioprocess for investigating cyclohexanone monooxygenase (CHMO). E. coli cultivations for CHMO production, bioconversion, liquid-liquid metabolite extraction and analytic techniques were conducted using the developed microscale automated approach. Each step allowed rapid and reproducible collection of quantitative kinetic data over multiples runs achieving ‘walk away operation’. Whole bioprocess evaluation was achieved, whereby linking multiple unit operations enabled rapid assessment of process interactions. Factors influencing CHMO activity and bioconversion yields were investigated along with alternative bioconversion substrates. From identified limitations of the CHMO system an optimised process was developed where the processing time was almost halved and CHMO activity increased 5-fold. Two novel self-sufficient cytochrome P450 systems, P450SU1 and P450SU2 were investigated using an automated approach where factors limiting bioconversion were identified. Implementation of the required improvements resulted in a 5-fold improvement in enzymatic expression and 5-fold and 1.5-fold increase in product formation from cytochrome P450SU1 and P450SU2, respectively. A matched oxygen transfer coefficient approach was used for predictive scale-up. The optimised microscale CHMO and P450 processes were scaled to 75 L and 7.5 L bioreactor scale, respectively. Growth and bioconversion kinetics were found to be identical between scales for the CHMO system whereas differences were observed for the P450 systems. Results described in this thesis have demonstrated the benefits of microscale automated methodologies for the creation, investigation and predictive scale-up of oxidative biocatalytic bioprocesses. The established strategies evaluated in this work contribute to meeting the current demand to decrease developmental costs and timelines.

Ion exchange chromatography is a ubiquitous unit operation within bioprocessing. There are a large number of ion exchange resins, however there is limited time and material available for a full assessment. The work reported is primarily concerned with the microfluidic scale down of process chromatography. Microfluidic technologies offer the potential to investigate events occurring within the bed (such as adsorption profiles) that would be more difficult to observe in larger systems. In addition, the devices may be used for the investigation of adsorbent type, buffer type, pH, flowrate etc. by employing parallel architecture to increase throughput. To facilitate the objectives described above, the fluidic handling and detection system was characterised. For the packing of microfluidic chips to be relevant for comparison, it was necessary to analyse the quality of the packing. Plate calculations combined with 3 dimensional imaging using confocal microscopy were used. According to the plate calculations, the microfluidic column was poorly packed, however the 3 dimensional imaging allowed a full assessment of the packing contained within the microfluidic column. Both frontal and elution chromatography results were generated using the microfluidic chip. Useful breakthrough curves using lysozyme were produced which compared well with both laboratory scale curves and previously published data. The general rate model was then used to understand whether the microfluidic column's breakthrough curves could simulate laboratory scale curves. The results produced showed that a new model was required to further develop the simulation. Separations were achieved using the egg white protein system containing 3 proteins. Microfluidic column separations were achieved using both step and linear gradients. These separations compared favourably with laboratory scale separations, although the use of linear gradients would require further work, as they proved challenging to reproduce at the microfluidic scale. However, the separation generated was similar to the laboratory scale.

In this PhD project a 50µL volume oscillating jet driven microfluidic microbioreactor was developed. This system features dissolved oIn this PhD project a 50µL volume oscillating jet driven microfluidic microbioreactor was developed. This system features dissolved oxygen (DO) contIn this PhD project a 50µL volume oscillating jet driven microfluidic microbioreactor was developed. This system features dissolved oxygen (DO) control and has fulfilled several design criteria and performance goals, which were demonstrated with fermentations conducted with Saccharomyces cerevisiae in YPD media. The device’s principle of operation is novel and newly demonstrated. The oscillating jet system generates active mixing without on-chip moving parts. The system is the only in its class to demonstrate DO control with yeast, in this case the industrially and scientifically important eukaryote S. cerevisiae. This is a new and novel result. The device has the smallest volume of any actively mixed microbioreactor, and does not suffer from evaporation of water from media. Reproducibility of DO monitoring and control has been demonstrated. This was done under the standard conditions for fermentation of S. cerevisiae, and has been extended up to 30 hours in duration, and to cell densities measured by optical density (OD) of over 20cm-1. DO control was maintained up to cell densities of ~ OD 10cm-1, which corresponds to a biomass concentration of 4.7 g-dcw/L. The system was demonstrated to be robust and reliable in mechanical and fluidic function. There are indicators that the DO control system has an effect on cell metabolism, an important consideration for microbioreactors as they are ultimately intended for use as high-throughput systems for process development, metabolic engineering, and synthetic biology. The volumetric mass transfer coefficient of the system is 174hr-1, which is high for its class. This is the key parameter for determining bioreactor oxygen transfer performance and scale-up suitability. Monitoring of cell growth via OD is less consistent than desired however. This may be due to the system’s limitation in maintaining S. cerevisiae cells in suspension at higher densities. This may also limit the “real-world” oxygen transfer performancerol and has fulfilled several design criteria and performance goals, which were demonstrated with fermentations conducted with Saccharomyces cerevisiae in YPD media. The device’s principle of operation is novel and newly demonstrated. The oscillating jet system generates active mixing without on-chip moving parts. The system is the only in its class to demonstrate DO control with yeast, in this case the industrially and scientifically important eukaryote S. cerevisiae. This is a new and novel result. The device has the smallest volume of any actively mixed microbioreactor, and does not suffer from evaporation of water from media. Reproducibility of DO monitoring and control has been demonstrated. This was done under the standard conditions for fermentation of S. cerevisiae, and has been extended up to 30 hours in duration, and to cell densities measured by optical density (OD) of over 20cm-1. DO control was maintained up to cell densities of ~ OD 10cm-1, which corresponds to a biomass concentration of 4.7 g-dcw/L. The system was demonstrated to be robust and reliable in mechanical and fluidic function. There are indicators that the DO control system has an effect on cell metabolism, an important consideration for microbioreactors as they are ultimately intended for use as high-throughput systems for process development, metabolic engineering, and synthetic biology. The volumetric mass transfer coefficient of the system is 174hr-1, which is high for its class. This is the key parameter for determining bioreactor oxygen transfer performance and scale-up suitability. Monitoring of cell growth via OD is less consistent than desired however. This may be due to the system’s limitation in maintaining S. cerevisiae cells in suspension at higher densities. This may also limit the “real-world” oxygen transfer performancexygen (DO) control and has fulfilled several design criteria and performance goals, which were demonstrated with fermentations conducted with Saccharomyces cerevisiae in YPD media. The device’s principle of operation is novel and newly demonstrated. The oscillating jet system generates active mixing without on-chip moving parts. The system is the only in its class to demonstrate DO control with yeast, in this case the industrially and scientifically important eukaryote S. cerevisiae. This is a new and novel result. The device has the smallest volume of any actively mixed microbioreactor, and does not suffer from evaporation of water from media. Reproducibility of DO monitoring and control has been demonstrated. This was done under the standard conditions for fermentation of S. cerevisiae, and has been extended up to 30 hours in duration, and to cell densities measured by optical density (OD) of over 20cm-1. DO control was maintained up to cell densities of ~ OD 10cm-1, which corresponds to a biomass concentration of 4.7 g-dcw/L. The system was demonstrated to be robust and reliable in mechanical and fluidic function. There are indicators that the DO control system has an effect on cell metabolism, an important consideration for microbioreactors as they are ultimately intended for use as high-throughput systems for process development, metabolic engineering, and synthetic biology. The volumetric mass transfer coefficient of the system is 174hr-1, which is high for its class. This is the key parameter for determining bioreactor oxygen transfer performance and scale-up suitability. Monitoring of cell growth via OD is less consistent than desired however. This may be due to the system’s limitation in maintaining S. cerevisiae cells in suspension at higher densities. This may also limit the “real-world” oxygen transfer performance.

This thesis documents the progress made in utilising the alkane hydroxylase complex of P.putida GPo1 expressed in E.coli as a whole-cell biocatalyst for the oxidation of n-dodecane to 1-dodecanol. The process is of considerable interest due to the difficulty in performing the reaction using conventional chemistry and the large global market for fatty alcohols. The first results chapter compares the fermentative bio-oxidations using E.coli pGEc47ΔJ on n-octane and n-dodecane in a stirred tank reactor. The first reported conversion of n-dodecane in-vivo using this enzyme system in a recombinant host is reported. A number of bottlenecks were identified in this chapter; in particular, (i) poor induction of the alkS expression system with ndodecane, which controls the expression of the alk enzymes (ii) a suspected mass transport limitation (iii) substantial over-oxidation of the desired 1- dodecanol product to dodecanoic acid. The second results chapter firstly describes the development of a microwell platform in order to characterise the system more efficiently. Phase mixing limitations and organic phase spillage/evaporation were overcome in order to develop the microwell platform for the fermentative bio-oxidation which is the first reported microwell scale-down which matches the volumetric and specific rates achieved in a bioreactor for a substrate of such low solubility. Secondly, the microwell platform was used with design of experiments (DoE) methodology to rapidly and systematically characterise the overoxidation issue and identify appropriate solutions. Using this approach, substrate solubility was identified as the most critical factor affecting the tendency for overoxidation; the use of cosolvents to improve n-dodecane solubility in the aqueous phase was found to improve the 1-dodecanol yields and reduce dodecanoic acid yields. Oxygen availability and carbon source availability also proved important factors in the extent of overoxidation. Despite the improvements made the problem was only partially overcome and it was decided, based on the results, that biological engineering of the strain was necessary to remove the downstream aldehyde dehydrogenase alkH which was likely to be exacerbating overoxidation. The process of designing and constructing 3 new plasmids is described in results chapters four and five. These plasmids were designed with the aim of identifying the role of various alk proteins and ultimately identifying ways of improving substrate access to the enzyme and reducing overoxidation. It was found as a result of this work that overoxidation was reduced by removal of alkH but that the alkane-1-monooxygenase alkB was still capable of direct overoxidation to the dodecanoic acid even in the absence of alkH. More significantly, the function of an outer membrane protein of unknown function was also confirmed by this work. It was found to be essential for conversion of n-dodecane in-vivo but was also found to be toxic to the host organism when overexpressed. Finally, it was found that the alkane-1-monooxygenase enzyme system was also capable of C14 and C16 alkane oxidation; this has not previously been reported in literature in-vivo.

Biocatalysis has emerged as a powerful tool for the synthesis of high value optically pure compounds. With advances in synthetic biology, it is now possible to design de novo non-native pathways to perform non-natural chiral bioconversions. However these systems are difficult to assemble and operate productively, severely hampering their industrial application. The purpose of this study was to develop a microscale toolbox for the rapid design and evaluation of synthetic pathways, in order to increase their operational productivities and speed-up their process development. The first aim of this work was to establish a microscale platform to accelerate the evaluation of different variants of transketolase (TK) and transaminase (TAm), in order to design and construct a de novo pathway for the one-pot synthesis of chiral amino alcohols. The second aim was to develop a microscale methodology to rapidly establish the complete kinetic models of the selected TKs and TAms, which would allow efficient operation of the one-pot synthesis. The third aim was to scale-up the production of the biocatalyst to pilot plant, while controlling and maintaining the desired level of expression of each enzyme. Finally the fourth aim of this project was to scale-up and simulate the complete one-pot syntheses to preparative scale, while predicting and applying the best reaction strategies and reactor configurations. The experimental microscale toolbox was based on 96 microwell plates with automation capacities, where the one-pot syntheses of the diastereoisomers (2S,3S)-2 aminopentane-1,3-diol (APD) and (2S,3R)-2-amino-1,3,4-butanetriol (ABT) were designed and performed with final product yields of 90% and 87% mol/mol respectively. For the synthesis of ABT and APD, the wild type E. coli TK and the engineered D469E TK were identified as the best candidates respectively, and both enzymes were paired with the TAm from Chromobacterium violaceum. A microscale methodology for kinetic model establishment was developed based on programmable non linear methods. The TAm step was found to be the bottleneck of the multi-step syntheses, due to the high a Michaelis constant of intermediate substrate erythrulose for the synthesis of ABT, and the low catalytic constants for the synthesis of APD. Also the amino donor substrate was discovered to be toxic for the TAm, as well as causing side reactions, thus affecting the overall performance of the de novo pathway. The production of the E. coli whole cells containing the de novo pathway were successfully scaled-up to pilot plant without losing catalytic activity. By manipulating the fermentation temperature and induction time of TAm, it was found the desired level of expression of each enzyme could be achieved. Finally, the complete one-pot syntheses were simulated using the previously established microscale kinetic models, which were found to be predictive of preparative scale bioconversions. A reactor with fed-batch addition of the amino donor was predicted as the best operating strategy in each case. Using this strategy, the one-pot syntheses allowed up to a 6-fold increase in product yield (% mol/mol), while using concentrations one order of magnitude higher than previously published preparative scale data. As a conclusion, this work is the first of its kind to develop such a microscale modelling toolbox, which is designed to exploit the synthetic potential of engineered and recombinant enzymes, in order to design, simulate and optimize de novo engineered pathways. This makes the results of this work an original contribution for the process development of synthetic pathways.

Enzymes from extreme environments possess highly desirable traits of activity and stability under process conditions. One such example is L-aminoacylase (E.C. Number 3.5.1.14) from the thermophilic archaeon Thermococcus litoralis (TliACY), which catalyzes the hydrolysis of the amide bond between the nitrogen and the carbonyl group of an N-protected L-amino acid. As this reaction is enantiomerspecific, L-aminoacylase is often used to resolve racemic mixtures in the preparation of chiral intermediates. Using protein engineering techniques, the capabilities of such biocatalysts can be extended. This thesis seeks to compare the ability of libraries created by error-prone PCR and structure-guided mutagenesis methods to identify residues governing substrate specificity. Libraries were constructed to screen for variants which showed a shift in substrate specificity towards aliphatic amino acids, whilst maintaining the preferred benzoyl protecting group. An existing fluorescent screen for proteases was adapted for the high-throughput screening of mutant L-aminoacylase libraries, and was demonstrated to be capable of quantitatively detecting millimolar quantities of product. From an error-prone PCR library over the dimerization region of the enzyme, 10000 variants were screened against a variety of N-benzoylated amino acid substrates. Sequence alignment and homology model construction allowed a 3-D model of TliACY to be built from its closest neighbours in the PDB. Phylogenetic comparison methods were used to identify regions and residues of significance, which were then examined using sitedirected saturation mutagenesis. Purification and characterisation of selected variants of TliACY examined two mutants in detail: S4C (S100T / M106K) which exhibited a 300% improvement in catalytic efficiency over wild-type on the Nbenzoyl valine substrate, and S3B (F251K) which showed a shift in substrate preference against aliphatic amino acid substrates.

Malignant pleural mesothelioma (MPM) is a devastating malignancy of the pleural lining. It presents insidiously and at the time of presentation the disease is often diffusely spread throughout the chest cavity. There are few effective therapies available, the average survival is 9-12 months from diagnosis and 5-year survival rates are only 2%. Novel therapies are desperately needed. There is increasing interest in combined gene and cell therapy approaches and for a malignant disease this is particularly appealing. Mesenchymal stem cells (MSCs) are known to home to tumours and are readily transduced with viral vectors making them ideal cells for delivering targeted therapy. TNF-related apoptosis inducing ligand (TRAIL) is an exciting anti-cancer therapy as it selectively causes apoptosis in cancer cells without affecting healthy cells. This makes the combination of MSCs carrying TRAIL (MSCTRAIL) a viable prospect for the targeted treatment of MPM. Lentiviral vectors expressing TRAIL were used to transduce human bone marrow-derived MSCs (MSCTRAIL) which induced apoptosis and cell death in multiple human MPM cell lines. MPM cells were transduced with a luciferase-expressing lentiviral vector (MPMLuc) and were used to establish a murine model of MPM allowing me to track tumour growth in vivo using bioluminescent imaging. Systemic delivery of MSCTRAIL to MPM tumours resulted in a significant reduction in tumour growth but topical delivery was not effective. Using dual bioluminescent and fluorescent imaging I showed that MSCs home to tumours when delivered both systemically and topically but that they engraft in greater numbers following systemic delivery. Combining chemotherapy with MSCTRAIL showed promising effects in vitro but was not effective in reducing tumour burden in vivo. In summary, human bone marrow derived MSCs were shown to localise to areas of MPM and when modified to express TRAIL and delivered systemically they significantly reduced tumour burden.

An automated microscale bioprocessing platform for membrane filtration processes was established to identify key process issues early and aid the rapid design of robust and scaleable filtration processes. To demonstrate the utility of this platform, it was used to investigate the impact of upstream operations on microfiltration performance. The primary recovery of humanised antibody Fab’ fragments from Escherichia coli (supplied courtesy of UCB Celltech) were used as a case study to evaluate the microfiltration methodologies and devices created in this work. Initially, the methodology associated with the microscale dead-end filtration device previously created and investigated by Jackson et al. (2006) has been improved by reducing the required volume by 50% (~500 \mu L). This improved method demonstrated reproducibility and sensitivity to changes in feed preparation. The method was then applied in the study of the influence of various cell disruption operations on subsequent solid-liquid separation and hence, Fab’ product recovery. Results showed that the heat extracted cells showed better dead-end microfiltration performance in terms of permeate flux and specific cake resistance. In contrast, the cell suspensions prepared by homogenisation and sonication showed more efficient product release but with lower product purity and poorer microfiltration performance. Having established the various microscale methods, the linked sequence was automated on the deck of the Tecan™ robotic platform and used to illustrate how different conditions during thermo-chemical extraction impacted on the optimal performance of the linked unit operations of product release by extraction and subsequent recovery by microfiltration. The microscale approach was then extended for crossflow operations. A microscale crossflow filtration device was designed to enable integration also within the Tecan™ platform for automated processing. The device has an effective membrane area of 0.001 m2, which is a hundred-fold smaller than the larger scale Pellicon-2™ membrane module used for scale translation studies, and has two independent membrane channels for parallel analysis. The device was first characterised by determining the normalised water permeability (NWP) of a Poly(vinylidene fluoride) membrane and compared this with the NWP of the membrane by dead-end filtration. NWP is an inherent membrane property and as expected, the NWP values derived from crossflow filtration experiments match the values derived from dead-end filtration to within 90%. For scale translation studies, two types of feeds were used: a model feed, which is resuspended active dry yeast and Bovine Serum Albumen in phosphate buffer, and the antibody fragment expressing E. coli strain. Results showed, that at matched optimal shear rates and transmembrane pressure, the percentage differences between microscale and large scale values were up to ± 25% for the permeate flux, ± 10% for Fab’ and total protein yields. These scale-up predictions were achieved with a ten-fold reduction in feed material requirement for crossflow operation. Overall, the results illustrate the power of microscale techniques to identify and enable the understanding of key process performance attributes in a bioprocess sequence. The broader implications derived from using these microscale membrane devices, further applications and recommendations for future research are also discussed.