With the advent of high-throughput screening technologies, key constraints associated with identifying potential drug candidates are being removed - the bottleneck in the timely delivery of new drugs will inevitably shift toward process development. Conventionally, the pilot plant studies required for process design only start when there is confidence that the drug candidate will make it to market. This means that this development period is very limited, and the flowsheet that enters the manufacturing phase is often sub-optimal. Collaborative work at UCL has been developing new strategies that aim to change this paradigm. The use of scale-down experiments in the drug discovery phase, in conjunction with modelling techniques, has been shown to be capable of providing more robust process definition early on in development. Such methodologies allow the study of more process options and hence rapid identification of optimal conditions, and thus mitigate the risks associated with equipment capital investment. Further advantages are that the experiments require less material, and can be done prior to, as well as in parallel to, pilot plant development. Scale-down can also be useful for the analysis of existing processes, e.g. for validation and troubleshooting, especially where material is valuable and/or scarce. This thesis describes the design and development of a scale-down device for membrane filtration, with a focus on tangential flow microfiltration for primary clarification. The device uses a rotating disk suspended above a static membrane surface to generate surface shear, in order to mimic the global, hydrodynamic conditions found in commercial crossflow modules. Results are presented showing how filtration performance of the scale-down device (3.5xl0-4m2 membrane area) correlates well with pilot scale data (O.1m2) for a range of representative biological materials including yeast, bacterial and mammalian cell cultures. Methodologies for confident assessment of microfiltration performance are given, which are capable of dealing with different feedstocks, membrane types, and a range of operating strategies, using greatly reduced quantities of feed. The steps required to design and build the next generation of filtration scale-down device for rapid process development are also addressed, along with a discussion of the related business and regulatory issues that shape the industry.

Exposure to high shear during bioprocessing operations has been reported to be associated with losses in protein stability and lead to aggregation, rendering a product unfit for use. High shear forces occur in many process operations, therefore it is important to understand the influence of shear and to control its effects during bioprocessing. The purpose of this work was to assess the stability of human monoclonal antibodies of IgG4 isotype using a device designed to generate defined, quantifiable levels of shear in the presence of a solid-liquid interface. The device, based on a rotating disk, produced shear strain rates of up to 3.4x10^4 s^-1 and was designed to exclude air-liquid interfaces. Computational fluid dynamics (CFD) was used to study the fluid flow patterns within the devices and to determine the shear strain rate (s^-1) at a range of disk speeds. The structural integrity of the IgG4 after exposure to interfacial shear effects was studies by SDS-PAGE, IEF, dynamic light scattering, and peptide mapping by LC-MS. This analysis revealed that the main denaturation pathway of IgG4 exposed to these effects was the formation of large insoluble aggregates. Soluble aggregation, breakdown in primary structure, and chemical modifications were not detected. Factor associated with the solution conditions (pH, ionic strength, surfactant concentration, temperature), and the interface (surface roughness, and hydrophobicity) were studies for their effect on the rate of IgG4 monomer loss under high shear conditions. The dominant factors found to affect the rate of monomer loss under interfacial shear conditions were found to be pH and the nanometre-scale surface roughness associated with the solid-liquid interface. The addition of surfactant was found to have a significant stabilising effect at concentrations up to 0.02%(w/v). These studies highlight that the interaction of shear and protein-surface interactions are important factors to consider when designing stability studies for protein production.

Of the four human IgG antibody subclasses IgG1-IgG4, IgG4 is unusual in that it does not activate complement and exhibits atypical self-association including the formation of bispecific antibodies. Given the therapeutic importance of human monoclonal antibodies their solution properties are critical to understanding clinical function and response to manufacture and storage. Thus IgG4 was studied under a range of relevant bioprocess, storage and accelerated degradation conditions and analysed by synchrotron X-ray scattering, analytical ultracentrifugation, circular dichroism and other relevant biophysical techniques. The first study was carried out at a range of protein concentrations at pH 7.4. This indicated a small concentration dependence of the IgG4 solution structure. The averaged conformation of the Fab regions appear able to hinder complement C1q binding to the Fc region and the self-association of IgG4 through the Fc region. These results provide a starting point to understand antibody structural stability during their manufacture. Low pH conditions were also studied as these are typically used for both viral inactivation and elution from protein A affinity resins during IgG manufacturing bioprocesses. Time and concentration dependent conformational changes and aggregation events were studied for the IgG4 at pH 3. It was observed that at all protein concentrations studied, the monomer size decreases over the first 100 minutes at pH 3. It was found that the molecular conformation changes along with an intramolecular increase in the beta-sheet content of the protein. At the higher protein concentrations, these events preceded aggregation suggesting that they may have a role in the propensity of the IgG4 to aggregate. The IgG4 was also studied over a range of temperatures at pH 7.4 using classical accelerated degradation experiments. However, no significant loss of protein was observed over the time and conditions studied.

The biopharmaceutical industry, which relates to human health, witnessed very fast development in the last few years to match the biodrug market demand. Manufacture of high value therapeutics usually requires the use of at least 2 or 3 chromatographic steps, which contributes to the significant cost of downstream processes. Therefore, chromatographic process optimization is an essential part of bioproduct manufacture development. In chromatographic separation, the compressible agarose-base matrices, which are most widely, used as column packing material. Over the past years there has been a steady move toward the adoption of more rigid, porous particles in order to combine ease of manufacture with increased levels of productivity. The latter is still constrained by the onset of compression where the level of wall support becomes incapable of withstanding flow-induced particle drag. In this study it investigates how, by the installation of cylindrical column inserts, it is possible to enhance the level of wall support to improve the column hydrodynamic performance. Experiments were conducted to examine the effect of the position of the insert in the column, and also of the insert dimensions on the critical velocity at which the onset of compression occurs. It was found that when installed at the bottom of the column, single inserts in different dimensions can provide 5% to 15% critical velocity increment, and inserts combination can provide up to a 20% increase in critical velocity without significantly affecting column hydrodynamics (less than 10%), as measured by the level of axial dispersion. A solid mechanics model was established to simulate the pressure drop, flowrate, and packing material compression properties in a chromatography column. Based on the Biot’s theory, which describes consolidation of porous materials, the model can relate the pressure drop to compression in chromatographic process. Darcy’s law is also applied, and combines with the Konezy-Carman equation for permeability. Comsol Multiphysics software, which can solve physical phenomenon by the finite element method, was employed for the model established. The inverse method in Matlab is used for parameter determination. With this simulation, the Young’s modulus, and bed voidage fraction were specified in one experiment condition. The determined parameters were then input to the model to simulate the flowrate, pressure drop, and bed displacement. The simulation results fit the experimental result quite well. The average error between experimental and simulation data was 3% for linear velocity and 4% for pressure drop. The simulation could also predict the superficial critical velocity for the same column packed but with inserts in different dimensions. Besides the experimental and simulation study on hydrodynamics in chromatography column having inserts, the effects of inserts applied on protein separation were also considered. The column inserts allowed higher operational flowrates, and process duration, which is positive to process productivity. However, it had a negative effect on the resolution, and caused larger elution volumes, especially when more than one insert present. In a case study of column inserts affecting on productivity, the productivity could increase 18% by two inserts setup in a 1 L column. These inserts lead 20% critical velocity increment, and 10% plate number decrement.

The production of therapeutic proteins with complex post-translational modifications is typically performed with suspension adapted mammalian cells, in stirred tank bioreactors (STRs). In this environment cells are exposed to hydrodynamic forces derived from both agitation and aeration. An improved understanding of these forces and the resulting cellular response is therefore critical. This thesis has characterised the hydrodynamic conditions within a STR using computational fluid dynamics and investigated the lethal and sub-lethal effects of this environment on CHO cells during the production of a monoclonal antibody. The hydrodynamic forces varied within the vessel, with the maximum forces being present in the impeller region. Energy of dissipation rates obtained from single and multiphase models were found to be higher than those previously reported. Having gained understanding of the hydrodynamic forces within the bioreactor, the environment was changed in a systematic manner. To create a bubble and bubble free environment within the STR direct gas sparged and silicone membrane aerated systems (SMAS) were used. When cells were subjected to the stress of gas-liquid interfaces they entered apoptosis at an earlier stage of cell culture, had decreased F-actin intensity and a modified cell morphology. This had implications on cell strength and impurity release during primary recovery. The application of fed-batch mode and mild hypothermic conditions were shown to be beneficial; delaying the onset of lethal and sub-lethal effects, enabling higher cell densities to be obtained, prolonging the culture duration and increasing product titre. The vital role of shear protectant agents at gas-liquid interfaces was highlighted; in absence of Pluronic F-68 cell death occurred within 24 hours of inoculation, but growth proceeded when using the SMAS. The glycosylation profile was not significantly affected by the STR environment, with harvest point being shown to have a greater impact on the relative abundance of different glycoforms.

Downstream bioprocessing and especially chromatographic steps, commonly used for the purification of multicomponent systems, are significant cost drivers in the production of therapeutic proteins. Lately, there has been an increased interest in the development of systematic methods where operating conditions are defined and chromatographic trains are selected. Several models have been developed previously, where chromatographic trains were selected under the assumption of 100% recovery of the desired product. Removing this assumption gives the opportunity not only to select chromatographic trains but also determine the timeline in which the product is selected. Initially, a mixed integer non-linear (MINLP) programming mathematical model was developed to tackle that problem and was tested using three illustrative examples. Later on, this model was linearised by applying piecewise linear approximation techniques and computational efficiency was improved. Next, an alternative MILP model was developed by discretising the recovery levels of the product and computational efficiency improved even by 100-fold. Finally, the equilibrium dispersive model was used in a simple 4-protein mixture and the MINLP model was validated. This research represents a significant step towards efficient downstream process operation and synthesis

Heterologous protein expression in E. coli can lead to the formation of dense insoluble aggregates named inclusion bodies (IB). The refolding of protein derived from IB is often characterised by low yields of active product. Process optimisation is often achieved empirically and requires significant resource and time efforts. Microscale experimentation may provide a valuable alternative by enabling representative process studies to be conducted early on in process development, using minimal quantities of product, parallel experimentation and automated liquid handling procedures. An automated robotic platform has been used to develop a dilution refold microscale process-screening tool with a set of hierarchical assays to rapidly determine optimal refolding conditions. The hierarchical orthogonal assays enable the simplest, cheapest and most generic high-throughput assays to first screen for a smaller subset of potentially high-yielding conditions. Absorbance can be used as an initial filter to measure particulate formation and fluorescence boundaries can then be used to select the conditions with the most native-like tertiary structure. The subset can then be analysed for native protein yield by slower, more expensive or protein specific assays, thus saving resources whilst maximising information output, alleviating the analytical bottleneck. This approach has been demonstrated in this work using lysozyme, with fluorescence boundaries to select 30% of highest yielding samples, and also with DHFR. An automated whole bioprocess sequence comprising fermentation, cell harvest and lysis, inclusion body harvest, denaturation and refolding has been developed at the microscale to study the effect of fermentation conditions on inclusion body yield and quality. The approach has been applied to dihydrofolate reductase (DHFR) and insulin, allowing a more thorough understanding of the effect of fermentation feeding, media and induction strategies on protein refolding yield and purity. This approach allowed yields of active insulin of increased from 10% to 68%. The results obtained from this approach have been compared to larger scales of operation, illustrating the challenges of scale-up. The process sequence, integrated with rapid analytical assays, provides a powerful tool for understanding the interaction between fermentation conditions and downstream processing yields, allowing a whole process approach to optimisation.

Lyophilization is a common method used for long term stability of pharmaceutical and biopharmaceutical products that are unstable in the liquid state for a substantial period of time. Currently, formulation and cycle development are often determined empirically. Although this approach is gradually changing as scientific publications reveal more about the nature of protein stability, nevertheless the lack of material during early stage development prevents large screening investigations to identify optimum formulations. The use of high throughput methods coupled with factorially designed experiments enables a far more efficient and wider screening and optimisation of viable formulations for development. This thesis explores the use of micro titre plates for formulation development with emphasis on formulations for lyophilization. This is coupled with design of experiment methods to provide a powerful engineering tool for the formulation scientist. While much has been done to model freeze drying cycles and optimize cycle parameters, current models are generic and require system specific data which can be hard to collect. By applying design of experiment principles, a system specific model was developed to allow the optimisation of cycle development to identify key parameters and produce a product that would meet critical quality attributes. Such a platform would lend itself well to quality by design and its application in lyophilization development.

No description available.

Escherichia coli is used to produce a wide variety of therapeutic proteins, such as Fab fragments. Many such proteins are expressed and retained within the periplasmic region of the cell, and after fermentation the proteins must be recovered for downstream processing using steps such as homogenisation. The aim of this study was to investigate a new way to modify E. coli cells that would increase the levels of release of specific periplasmic proteins of interest into the cell culture broth, in order to allow these proteins to be recovered without the need for lengthy fermentations or the usual downstream processing steps. This thesis describes the construction of plasmid vectors designed to express antisense RNA that would inhibit the translation of cell membrane components including murein lipoprotein (Lpp), the most abundant protein in E. coli and a major component of the outer cell membrane. This was conducted using a strong promoter that could be induced during the cells’ exponential growth, to ensure the expression of the antisense RNA as the cells were growing so that subsequent generations of the cells would have lower levels of the lipoprotein within their cell membrane, and as a result leak a greater proportion of an overexpressed protein, alpha-amylase, into the cell culture broth. The system was also tested with a pair of Fab fragments which were induced with a distinct promoter so that the antisense RNA could be generated as the cells grow, while the Fab production would begin after the exponential growth phase has completed. These strains were tested from shakeflask levels up to 20 L fermentations (with up to 9 L working volumes). The periplasmic proteins examined showed increased levels of release in strains with the antisense plasmids, increasing the levels of release from as low as 10 - 20% in the unmodified strains up to almost 50% in some cases. This was achieved without compromising the growth rates of the cells which make them suitable for industrial fermentations. In summary, this antisense system, can be used to increase release levels of periplasmic proteins although further refinements and research into the use of antisense RNA would be required for the system to achieve full release of periplasmic proteins.