Packed bed chromatography is the workhorse of the majority of downstream purification processes used for the manufacture of biopharmaceutical therapeutics. This high dependence upon chromatography has lead to concerns being raised regarding the manufacturing costs and also the potential constraints on plant productivity imposed by packed bed processes, particularly in light of advances seen in upstream operations. This has not unsurprisingly generated a significant degree of discussion amongst the bioprocessing community on how best to deal with these challenges. Amongst the proposed strategies, is the adoption of what may be termed \alternative" bioseparation techniques, which may potentially offer higher processing capacities at a lower cost. In this study a Multi-Attribute Decision Making MADM) based framework was used to evaluate these bioseparation techniques being considered as potential alternatives to packed bed chromatography. This evaluation included consideration of a wide range of process characteristics, beyond just performance and cost related attributes, but also considering areas such as the ease of process development, operation and scalability. The use of this framework not only allows the most promising technologies to be identified, but manipulation of the non-deterministic outputs of this framework permitted indications to be obtained as to the most productive directions for further technology development. Using the indicators provided by this framework, experimental studies were carried out in order to study the performance of these most promising alternatives, when used as part of a whole downstream processing train. These studies yielded information upon the interactions between these different bioseparation technologies and allowed the impact on process productivity and process economy to be evaluated. The collective findings from this study reinforce the generally held opinion that none of these alternative bioseparation techniques can currently be considered the key to overcoming the challenges faced by downstream processing. Indeed evaluations of the techniques imply that the ubiquitous nature of packed bed chromatography is likely to remain for the foreseeable future. However the tools which have been developed allow for a rational explanation of what exact factors are the barriers to adoption for these techniques, and as a consequence provide guidance as to the areas in which future development is most required.

Bacteriophage are a diverse class of viruses that infect bacterial cells. As a result of over 60 years of molecular biology advances, bacteriophage today feature as candidates for vaccination, gene therapy, biomaterial and antibacterial purposes. Consequently, scientific, commercial and public awareness of bacteriophage is growing rapidly. There is now an increasing need for the establishment of strong biochemical engineering foundations to serve as a guide for future bacteriophage bioprocessing. It has been the purpose of this study to contribute towards this knowledge base, by understanding the properties of the filamentous bacteriophage M13. Ultimately, this work has aimed to allow for the more efficient assembly of a large-scale production process. By the application of well-understood small-scale predictive techniques, it has been found that bacteriophage M13 should not be severely damaged by hydrodynamic shear forces of the duration and magnitude imparted by fermentation, pumping or continuous centrifugation operations. Thus, it may well be possible to manufacture on the large-scale using existing large-scale equipment designs. Amongst bacteriophage, the reproduction strategy of M13 is unusual in that propagation occurs by the non-lethal extrusion of progeny through the cell wall of the E. coli host. Investigation of bacteriophage M13 propagation indicated that growth in a medium that increased host cell density concomitantly increased bacteriophage yield; a four-fold increase to 2 x 1012 pfu ml-1 was achieved. At the end of culture, concentrations of supernatant DNA and protein contaminants were found to vary amongst three E. coli strains studied. Post-fermentation, bacteriophage M13 can be precipitated from the cell-free process fluid by as little as 2 % (w/v) PEG 6 000 plus 25 mM magnesium sulphate, or by isoelectric precipitation. Purification factors in excess of 100 were achieved by PEG-salt precipitations with regards to the reductions in DNA and protein concentrations. Methods used in this study have increased the processing knowledge of bacteriophage M13 and have a broader applicability to both derivatives of M13 and other bacteriophage.

Purification of Biosynthetic Human Insulin (BHI or Humulin), involves multimodal chromatographies as well as intermediate crystallizations. Size exclusion chromatography (SEC) is the final chromatographic step in BHI purification, which is conducted in acetic acid at high protein concentrations. The purpose of the SEC step is to remove the higher-molecular weight species proinsulin (molecular weight (MW) ≈ 10 kDa) and covalently linked insulin dimer (MW ≈ 11.6 kDa) from native insulin monomer (MW 5808 Da). Because covalent insulin dimers are recognised as a significant factor in immunological responses seen in diabetic patients (Darrington & Anderson 1995), limiting the amount of covalent in the final active pharmaceutical ingredient (API) is a key requirement for the development of the purification process. The presence of covalent dimers in insulin process streams results from chemical degradation of BHI either through intermolecular disulphide bond formation or from transamidation. Complete characterisation of the covalent dimeric species present in commercial SEC feed streams has not been performed. However, because specific purification technology is employed to remove disulphide bonded polymeric impurities upstream of the SEC step, dimeric impurities are thought to be predominantly transamidated species. Further, recent laboratory-scale SEC studies have indicated that there is a covalent dimer form present in the SEC feed streams that is difficult to separate from the native BHI molecule by SEC. The objectives of this study were to determine the types of insulin covalent dimer entering the BHI SEC step, understand the impact of the covalent insulin dimer type on SEC separation of dimer from the native insulin and to learn more about the SEC step itself i.e., investigations into the effect the SEC resin (G-50 Sephadex), may have had on the BHI structure. In order to achieve this, various techniques were explored to analyse BHI conformation, purity and degradation trends. Such techniques included Sodium dodecyl sulphate polyacrylamide gel electrophoresis SDS-PAGE, High Performance Liquid Chromatography (Reverse Phase and Size Exclusion Chromatography), Circular Dichorism and Surface Enhanced Laser Desorption Ionisation (SELDI). Breakdown studies on BHI were carried out using reducing agents DTT and GdnHCl and the effects of environmental conditions (pH, temperature and protein concentration), on native BHI as well as the reduced protein were investigated. Further to this, studies carried out at Eli Lilly and Company explored the occurrence of the fronting observed after SEC purification. Here the human proinsulin (HPI), high molecular weight polymer (HMWP) and BHI were thought to exist together and were separated from the main BHI peak, and then recycled in order to separate and recapture BHI in order to increase yield. Studies were conducted to understand the occurrence of the fronting better, as well as to see if factors such as presence of Zinc could alter the fronting and whether this could ultimately lead to steps being taken to modify the manufacturing process to make it more efficient and economically sound. Finally, the folding reaction where human proinsulin s-sulfonate (HPSS) is converted to HPI during BHI manufacture was investigated. During this step, the HPI purity and behaviour was assessed in order to determine whether changes could be made here in order to make the manufacturing process more efficient. Overall it was found that the species present during BHI analysis were BHI monomer and transamidated dimer. Upon reduction by denaturing agents such as DTT with GdnHCl, the BHI monomer was found to become reduced to form its constituent A and B chains. More of a dimer peak was observed also, which was a combination of transamidated dimer, but also BHI monomer being eluted sooner during HPLC due to its reversible formation of self-associated dimer as it flowed through the column. An additional small peak was also observed which was suspected to be an AA-AA dimer, but further experimentation is required to confirm its identity. Addition of Sephadex caused BHI monomer to be more stable, thus more difficult to reduce using denaturing agents such as DTT and GdnHCl to its constituent A and B Chains. The sephadex induced a shift in equilibrium between BHI and Chains A and B such that the BHI protein remained in a more folded state, thus making it harder for the DTT to access and reduce the disulphide bonds. With regards to environmental conditions, it was found that an increase of temperature lead to an increase of degradation/reduction of BHI monomer, resulting in an increase of all the other peaks. The general trend observed with regards to pH was that BHI monomer was most stable at pH 3 at all the temperatures tested, and pH 4 was the most unstable except at 40°C where an increase of pH appeared to encourage misfolding and avoid aggregation. Complete unfolding of BHI was only found to occur at pH 9 (where all the pH’s examined were 1.5, 2, 3, 3.5, 4, and 9). Concentration changes resulted in an expected increase in peak areas in SEC HPLC, and it did show that more BHI dimer was likely to form as a result of increased concentration. A number of theories were tested to understand the occurrence of fronting in the BHI elution peak from SEC. It was found to result from the BHI forming a self-associated dimer on the column, thus eluting earlier but then returning to its original monomeric state. This explained why BHI monomer was found in the fronting area along with HMWP and HPI, and therefore had to be recycled and repurified. It was also found that chelating zinc ions from the BHI sample did not have an effect on the equilibrium between monomer and dimer, or on the fronting observed on the large scale SEC HPLC. It was seen however that when the G-50 SEC column was saturated with ZnCl2 the BHI monomer peak shifted towards being eluted at a higher MW (towards the left), thus promotion of zinc induced BHI dimerisation. Although this resulted in increased fronting, this is more likely now to be due to the harmless self-associated dimers that are reversibly formed as opposed to non-reversible covalently formed dimers, so the need for the cut and recycle step to remove the covalent dimer, would no longer be needed. However it was also found that the HPI content was significantly increased in the Main Stream (MS) step (as illustrated in Figure 7.1) which decreased the final BHI purity, and so overall the saturation with zinc did not allow the recycle step to be avoided during manufacturing. Thus it was better to keep the initial fronting and avoid Zn induced dimerisation. Finally with regards to HPI purity during the manufacturing step where HPSS was converted to HPI, it was found that disulphide shuffling occurred with a corresponding increase of HMWP, but only for the first 40-50 hours, after which the polymer fraction decreased and HPI fraction increased until a plateau was reached. A variation in the time to reach the plateau is likely to be due to slight variations in added cysteine concentration. It is recommended that the cysteine concentration should be investigated in further experimentation in order to speed the reaction up so that the plateau could be reached faster. This decreased manufacturing time would result in a more time and cost efficient process.

The use of small scale bioreactors that are mechanically and functionally similar to large scale reactors is highly desirable to accelerate bioprocess development because they enable well defined scale translations. In this study a 25 mL miniaturised stirred tank bioreactor (MSBR) has been characterised in terms of its power input, hydrodynamics and volumetric oxygen transfer coefficient (kLa) to assess its potential to grow high cell density (HCD) cultures using adequate scale-down criteria. Engineering characterisation results showed scale-down, based on matched specific power input (PG/V) and kLa was feasible from the 20 L and 75 L pilot scale stirred tank bioreactors used in this study. In addition, kLa in the MSBR was found to be highest among all three bioreactors suggesting that it was feasible to perform high cell density fermentation in the MSBR. Improvement of mechanical and operational design and optimisation of fed-batch operation resulted in high cell density establishment in the MSBR. Scale-down was performed from 20 L STR to the 25 mL MSBR at matched PG/V, matched kLa and based on dissolved oxygen (DOT>30%) using Fab’ producing E. coli W3110. Comparison of results from the three scaledown strategies shows that matched PG/V was the best strategy and 25 mL MSBR accurately scaled-down the 20 L fermentation performance in terms of growth, Fab’ production and harvest material characteristics at matched PG/V. Scale-down based on dissolved oxygen did not produce reproducible results. Successful scale-down at matched PG/V in the MSBR resulted in maximum cell density of OD600nm~ 114 and total Fab’ concentration of 940 μg/mL compared to OD600nm~118 and 990 μg/mL in 20 L STR. Furthermore, the use of the MSBR in conjunction with primary recovery scale-down tools to assess the harvest material of both reactors showed comparable extracellular viscosity, shear sensitivity and centrifugation performance at both scales when scale-down was performed at matched PG/V. The conjoint use of the MSBR with ultra scale-down centrifugation mimics can provide a cost-efficient manner in which to design and develop bioprocesses that account for good upstream performance as well as their manufacturability downstream. To assess the feasibility of the MSBR to be used for bioprocess development purpose an in-house variant strain of E. coli W3110 co-expressing Fab’ and Staphylococcus aureus nuclease was characterised in 25ml, 20 L and 75 L bioreactor for growth and productivity. Growth and productivity profiles in the pilot scale reactors were successfully scaled-down by 25 mL MSBR when Fab’ fermentation was performed at matched PG/V. Results highlighted in this thesis for successful scale-down high cell density cultivation, use of scale-down recovery tools to predict downstream processibility of the 25 mL harvest, and application of the MSBR for upstream process development of a new strain, suggest that the MSBR can be used for bioprocess development in parallel with pilot scale reactors.

Increasingly, high throughput studies, coupled with microscale techniques, are being employed as an integral element of bioprocess development to generate valuable knowledge from the early stages of the development train. This has led to high throughput bioprocess development as a time- and cost-efficient approach to developing bioprocesses. High throughput studies aim to identify influential process parameter,and the operating ranges leading to favorable process performance. These are refined as development progresses to establish eventually the process design space. At the early stages of bioprocess development a greater number of parameters with wider ranges are investigated and these lead to the assessment of an increased number of samples. This can frequently result in a bottleneck which is a persisting challenge in implementing high throughput bioprocess development. The goal of this thesis is to alleviate this bottleneck and it seeks to achieve it with three systematic methodologies (1) strategic assay selection; (2) strategic assay deployment; and (3) a hybrid experimental simplex algorithm for identifying process 'sweet spots'. The first two attack the bottleneck from the perspective of analytical methods and they aim to select methods which are fit for high throughput applications and then deploy them in a fashion that reduces the analytical burden and concurrently ensures the derivation of correct conclusions upon the completion of a study. The third methodology is an algorithmic approach for iteratively stepping through an experimental space to identify and define favorable process operating regions while reducing the expenditure of resources for investigating sub-optimal regions. These methodologies are described and then demonstrated by applying them to industrially relevant case studies. The obtained results are analyzed and used to support the argument that they have the potential to facilitate high throughput activities and consequently high throughput bioprocess development.

No description available.

A high throughput method for the study of normal flow microfiltration operations has been established using a custom designed 8-24 well filter plate (0.8 cm2) and a commercial 96-well Multiscreen filter plate (0.3 cm2). Integration of this new approach with a typical robotic platform has enabled automation of the experimental procedure. Membrane resistance data can be quantified using either filter plate. The accuracy of these measurements has helped to determine that plate position does not affect experimental results and applied pressure difference does not vary across either plate. Each of the two filter plate designs has been used to demonstrate that cell condition following fermentation, buffer type and media composition are all important factors influencing the specific cake resistance of E.coli TOP10 cells. The microscale method therefore allows parallel quantification of the impact of upstream process conditions on microfiltration performance. The custom filter plate, optimised for bioprocess studies, allows multiple membrane types to be evaluated on a single plate and the measurement of both permeate and retentate masses to ensure against cross-contamination or loss. Lower variation in specific cake resistance values is seen in the custom filter plate compared with the commercial filter plate. These automated microscale normal flow microfiltration techniques have also been combined with factorial experimentation to identify the key factors and interactions which influence the protein transmission and specific cake resistance during filtration of an E.coli and protein mixture. Results indicated pH and ionic strength were important factors. The pH and ionic strength interaction was further investigated using response surface methodology and a window of operation was generated showing the pH (5.5 ± 0.1) and ionic strength (153 ± 8 mM) values necessary to achieve a protein transmission above 95% and a specific cake resistance below 80 × 1012 m.kg-1. The custom microwell filter plate cake resistance and transmission data from the response surface models scaled up by a factor of 17 to conventional laboratory scale equipment, showing that the optimum conditions achieved in the microwell could be replicated at a larger scale. In addition to this, experiments at the laboratory scale confirmed the optimum indentified by the custom microwell filter plate. This demonstrated that the combination of experimental design and the custom microwell filter plate is capable of investigation, optimisation and scale-up of a complex separation process. Finally, the approaches established here have been expanded to a whole process sequence for the purification of plasmid DNA. A non-chromatographic process sequence, which might be used in industrial practice, involving 7 consecutive processes (4 filtration steps) has been run with 72 combinations of 8 different factors in parallel, collecting hundreds of scaleable data points. The key filtration challenges were identified as the lysis clarification and the removal of lipid removal agent (LRA). Once again the important interactions and trends were identified using microscale experimentation. A major discovery was the filter aid effect of the adsorbent; increasing LRA concentration showed a dramatic reduction in specific cake resistance. This trend was repeated in larger scale devices with different filter formats at high area (150X) and volumetric (2000X) scale-up factors. This shows that the microscale techniques developed in this thesis are capable of determining quantitative, scaleable data for early stage evaluation of whole microwell process sequences.

The use of Diels Alder cycloaddition chemistries to access a diverse range of useful cyclic structures is well established throughout literature. However, the value of the product may be enhanced further still by linking this reaction with subsequent (biocatalytic) steps to create novel, structurally demanding, optically pure compounds. This project investigates the linking of Diels Alder (DA) chemistry to the enzyme, transketolase (TK) as a model integration pathway of a chemical syntheses and a biological transformation. The two-step process aims to provide a framework to synthesise small structurally diverse compounds with high enantiomeric excess. The demand for optically pure compounds is becoming a necessity due to the adverse affects frequently introduced by racemic compounds and the cost implications of the material possessing often only 50% active compound. Recombinant wild type Eschericha coli transketolase (EC 2.2.1.1) (WT-TK) was overexpressed in E. coli for the biocatalytic step of this two step synthesis. A substrate walking approach whereby a range of sequentially linked cyclic aldehydes, were applied to wild type transketolase and potential activity detected. Transketolase mutants, previously constructed based on information derived from the structural position within the active site of the dimeric enzyme were subsequently screened for activity with the cycloadduct of the Diels Alder reaction as aldehyde acceptor substrate for TK. Following identification, selection, culturing and sequencing of variants indicating enhanced activity towards the novel, bulky, hydrophobic, cyclic aldehyde the enantioselectivity and absolute stereochemistry of the product were determined. Activity displayed by wild type transketolase indicated an 18,000 fold activity improvement. Michaelis- Menten kinetic parameters were subsequently determined to have an apparent Km of 69.9 mM, kcat of 17.5 s-1 and a vmax of 0.07 mM.min-1 and preliminary process compatibility including product and substrate inhibition issues were highlighted.

The growing demand for quick and effective methods of producing large amounts of plasmid DNA for human therapy and vaccination has increased the practical challenges associated with process optimisation to improve supercoiled plasmid DNA yields obtained through current methods. The supercoiled isoform of DNA is the preferred form for use in gene therapy and vaccination as this isoform is known to produce higher levels of in vitro and in vivo transgene expression than other forms of plasmid DNA. This study was designed to investigate whether different strategies can be implemented early on in a process to improve supercoiled plasmid DNA yields obtained upstream, with the view to aid and/or ease further downstream stages. The main theme investigated is the influence of the host strain on supercoiled plasmid DNA production. Seventeen strains of Escherichia coli and three different plasmids were investigated at shake flask scale, before two strains were selected for scale up to 7L fermentation scale. The results obtained indicated that the host strain plasmid combination can heavily influence both the quantity and quality of plasmid DNA obtained and this behaviour cannot simply be determined by looking at the host strain genotype. Fermentation runs on the two strains selected for scale up (BL21 DE3 gWiz and HB101 gWiz) demonstrated that these two strains scale up very well, maintaining high specific pDNA yields (1.5mg/L/OD for BL21 DE3 gWiz) and high SC-DNA yields (98% for HB101 gWiz). Temperature amplification studies using strains harbouring pUC18 have shown that although most strain-plasmid combinations yielded more plasmid at a higher temperature of 40°C, the extent of this increase is highly influenced by the host strain. Indeed in some cases, such as for the strains ABLE K, W3110, W1485, a higher plasmid yield was obtained at 37°C. However, as the growth rates of these cultures were not measured, the extent of the accumulation of plasmid DNA due to the effects of the growth rate and/or temperature during the exponential phase of growth is unknown at this time. Similarities to what has been reported as temperature induced runaway plasmid replication have been observed in this study, although no experiments were conducted to confirm whether these observations were indeed as result of runaway replication as defined in the literature. Potential alternative strategies investigated included implementing anaerobiosis to test if these conditions can improve supercoiled plasmid DNA production at fermentation scale, and whether a ‘Quiescent cell expression system’ (a state where chromosomal replication and expression is temporarily shut down but residual proteins remain metabolically active) can be implemented to improve plasmid DNA yields by redirecting resources away from biomass production. The results suggest that under the conditions set in this study, these strategies do not increase plasmid DNA production or the percentage of supercoiled plasmid obtained. In conclusion, the results from this investigation have demonstrated that a highly effective and influential strategy for improving the quality and quantity of plasmid DNA obtained is the initial choice of the host strainplasmid combination. Further improvements can then be obtained through the application of other reported fermentation strategies.

The scale-up and manufacture of therapies based on intact whole cells presents a major challenge for development scientists and engineers due to the stress-reactive nature of these cells. The administered cells may be characterized in terms of their membrane integrity, size, surface markers and eventually their biopotency. The challenge is to process the cells at various scales and in ways which maintain these cell properties. Also during formulation the presence of cytokines produced by cells is a critical factor. This study presents an approach to allow the rapid characterization of human cell lines in terms of their resistance to hydrodynamic stress. An ultra scale-down method has been developed which allows investigation with small quantities of cells commonly available at the early discovery stage. The study describes controlled flow through a capillary device where cells are exposed to several defined hydrodynamic stresses. A Design of Experiments approach was used to understand the combined effect of process parameters such as flow rate, length of capillary and number of passes. This was followed by an additional set of detailed ultra scale-down experiments where other critical quality attributes such as cell size, surface phenotype, biopotency and cytokine release were measured. Computational fluid dynamics was used to describe the capillary entry region. The cells were then characterized in terms of a critical stress below which there is no significant damage to their integrity. The results were used to predict successfully a capillary design where no damage would occur at a specified high flow rate; for example as required for cell dispensing or vialling operations. Equally, the extent of loss of cell integrity was also successfully predicted in a capillary flow system designed to yield high levels of break up.