Bioactive proteins have emerged as the next generation of biopharmaceuticals. While downstream purification remains a bottleneck in manufacturing, as conventional techniques are expensive, crystallisation is increasingly viewed as a viable greener purification step. However, proteins are notoriously difficult to crystallise due to their complex macromolecular structure with patchy surfaces. The work presented in this thesis aims to enhance crystallisation of bioactive proteins, such as human insulin, by molecular templating the solution properties, e.g., altering the pH, salt type and concentration or introduction of novel soft-templates. To demonstrate the feasibility of crystallisation to replace conventional downstream purification techniques, a novel concept for a continuous tubular bulk crystalliser will be presented. Combining both aspects, this study will showcase the feasibility to achieve a controlled and continuous crystallisation at scales towards downstream purification.
One of the most crucial parameters in protein crystallisation is the pH as it modulates the net surface charge and therefore dictates the strength of the electrostatic interactions. Studying crystallisation of insulin at different pH (6.0 to 6.7) and supersaturation ratios (4 to 21) revealed that with increasing pH, the solubility increases by 5-fold whereas nucleation was accelerated by up to 8 times. This opposing effect has not been reported anywhere else. Investigating the protein-protein interactions revealed that, for the investigated pH range, the protein-protein interactions are pH independent. It is consequently suggested that the solvent-protein and the ion-protein interactions and the resulting water structuring are crucial in insulin crystallisation.
Stemming from the crucial role of the solvent-protein and ion-protein interaction, amino acids (0.01 to 0.1 M) as novel soft-templates to control nucleation at low supersaturation (S=18) are introduced. It is demonstrated that both L-arginine and L-leucine promote the nucleation of human insulin even at low supersaturation, whereas L-glycine does not promote nucleation. As the amino acids are dissolved, it is hypothesised that it is the intermolecular interactions between the protein’s residues and the amino acid’s residue in combination with changing the protein-solvent and protein-ion interactions, that results in an enhancement of nucleation. By studying the solubility of insulin, it was found that the enhancement in nucleation is not due to a change in the thermodynamic equilibrium between the crystalline and bulk-liquid phase. Lastly, it is demonstrated that L-arginine and L-lysine accelerate insulin crystallisation at high supersaturation (S=180) by more than 3 times at scales of 10 mL and beyond. This demonstrates the feasibility of the soft-templating concept to control protein crystallisation at scales towards downstream purification.
To demonstrate the feasibility and operability of bulk crystallisation as a tool for downstream purification, a continuous crystallisation platform consisting of a tubular slug flow crystalliser is presented. Thereby, the effect of both physical mixing and equipment surface chemistry on the design and operability was investigated, using lysozyme as a model protein and under laminar flow conditions (Reynolds number » 1). Firstly, it was found that proteins adsorb nonspecific on glass surfaces independent of their surface functional groups, e.g., OH or CH3, resulting in an undesired acceleration of the heterogeneous nucleation of up to 11.5 times and unstable slug flow. Recovering the surface functional groups by applying surface-specific cleaning protocols comprising of NaOH or liquid detergent, resulted in a stable slug flow and controlled nucleation. Secondly, a two-step-mixing approach, consisting of an intermediate mixing step, was developed to overcome poor mixing and prevent what seems to be amorphous precipitation in laminar flow conditions.
Investigating the obtained crystals with respect to their shape, average size and size distribution is crucial in designing and scaling-up bulk crystallisation processes. Therefore, a MATLAB® based automated image analysis routine will be presented which enables the derivation of crystal size distributions from images. This image-based technique requires only a small amount of crystal slurry, e.g. 10-20 μL, and therefore enables the characterisation of crystal sizes during the initial scale-up from small scale screening to continuous bulk crystallisation processes. By analysing more than 5000 crystals it was found that the crystal sizes follow a log-normal distribution. As image-based derivation of crystal sizes are traditionally executed manually with ImageJ®, the performance of the in MATLAB® developed image analysis routine was compared to ImageJ®. Although both software return similar size distributions, the average deviation in x10,2, x50,2 and x90,2 is less than 5 μm which is the microscope’s resolution, because of the automation, fast execution, and flexibility of MATLAB®, in combination with tailoring image-specific operations, MATLAB® is superior to ImageJ®.
Overall, this study showcases the control of crystallisation of biopharmaceutical proteins with molecular templates. Moreover, it is demonstrated that the concept of molecular templating enables the control of crystallisation at scales beyond millilitres. Coupled with the presented continuous tubular slug flow crystalliser, this study highlights and demonstrates the feasibility of achieving a continuous and controlled crystallisation at scale, which ultimately has the potential to replace conventional downstream purification techniques.