Development of novel methodologies for the optimization of production processes of biopharmaceuticals

Abstract

Process Analytical Technology (PAT) is defined by the FDA as a "System for designing, analyzing and controlling manufacturing through timely measurements of critical quality and performance attributes of raw and in-process materials and processes, with the goal of ensuring final product quality". The goal of implementing PAT is defined therein as enhancing the understanding and the control of a production process. This broad definition encompasses testing of raw material for batch consistency as well as online sensors that provide feedback for the process control. In this context, rapid methods placed at-line, i.e. close to the process, can be considered to be PAT applications since they will both contribute to the understanding how individual steps impact on product quality and will accelerate and facilitate process development and optimization decisions. There are many tools available that enable process understanding for scientific, pharmaceutical development. These tools can provide effective and efficient means for acquiring information to facilitate process understanding and continuous improvement. From a physical, chemical and biological perspective, pharmaceutical products and processes are complex multi-factorial systems. Methodological experiments based on multivariate statistical principles provide useful means for the identification and study the effect and interaction of product and process variables. Traditional one-factor-at-time experiments cannot address these kinds of interactions. These tools enable the identification and evaluation of product and processes variables that may be critical to product quality and performance. Thanks to the PAT initiative, spectroscopic sensors systems have gained interest for bioprocess monitoring because they allow rapid and non-destructive monitoring of product quality attributes. The improvements in spectrometers, detectors and optics have led to interesting applications related to PAT. The main goal of the thesis is the development of novel methodologies based on spectroscopic techniques coupled with multivariate data analysis for the optimization of production process of biopharmaceuticals. One of the approaches described here combines the strengths of various spectroscopic techniques, such as Circular Dichroism (CD), Infrared (IR), Raman, Fluorescence and UV-Visible-NIR measurements, to provide a more comprehensive description of a substance, the so-called "fingerprint". This, in combination with Principal Component Analysis (PCA) may be use to establish and define quality, equivalence, and comparability of substances while also providing a means to monitor processes and provide relevant information about molecular changes in product. Moreover, it can highlight the relationships between different properties, for example that between structure and aggregation, and a better understanding of the nature of a product. The protein used in this study is a homo-dimeric Fc-fusion protein. In particular, for the first part of the project, focused on PCA of multispectroscopic data, were used ten batches of drug substance produced with the current process called "process C". This bulk material has a concentration of "not less than" 160 mg/ml. Within this set, the batches differ with respect to hydrolysates that were used as feed during the fermentation process. In addition to these difference, three additional batches form the new "process D", after some minor optimisations with respect to "process C", were analysed. In order to generate a wider diversity of samples, with aspects of deterioration, solutions were diluted with water (instead of using a buffer) and then stored at room temperature for several weeks before spectroscopic analysis. The major consequence of this treatment is a change in buffer/additive concentrations, a change in pH, and a deterioration of solutions through aggregation, structural and chemical decomposition. A comprehensive set of spectra of eleven batches thus were acquired, using a variety of techniques. Some twelve variants of five spectroscopies have been employed, covering the complete wavelength range from far-ultra violet to infrared and involving phenomena including absorption, fluorescence, Raman scattering, Rayleigh scattering and circular dichroism. Both concentrated stock and deliberately deteriorated dilute solutions were investigated. All of the techniques employed have yielded useful data of some forms in terms of identifying variance in the batches and are potentially complementary and cross-supporting. Each set of spectra were subjected to multivariate data analysis, primarily PCA, to highlight patterns and differences between batches. Such analysis highlighted an apparent connection between the spectra and the history of batches regarding production date and hydrolysate type used. In particular, a series of anomalous absorptions in the visible wavelength region, together with potentially related fluorescent species, were identified. These may derive from contaminants, post-translational modifications (PTM) dependent on production conditions. Another aim of the thesis was to assess the feasibility of obtaining quantitative data about degradation products of a therapeutic protein when employing Circular Dichroism and infrared spectroscopy in combination with multivariate data analysis, primarily Partial Least Squares (PLS) regression, and an extension of PLS, Orthogonal Partial Least Squares (O-PLS). This is a novel approach since the classical applications for CD and IR spectroscopy are the determination of secondary structure content of proteins. Also the use of multivariate statistical methods for the determination of secondary structure content is reported. Nevertheless, the present approach is to our knowledge the first one that seeks to exploit PLS in order to correlate CD and IR spectral data with quantitative data of common protein degradation forms. In order to generate a suitable calibration matrix, a set of samples containing pre-defined levels of aggregates, oxidized forms, and free Fc, was generated. In order to ensure non-correlation of the degradation levels within the calibration matrix, the target concentrations therein were chosen according to an approach described by Brereton (2000). All the samples generated were then analyzed separately for each of the three degradation forms employing dedicated chromatographic QC assays in order to obtain accurate degradant levels. Furthermore, both CD (near and far UV) and IR spectra were measured. Both the QC and the spectroscopic data form the basis for the generation of various PLS/O- PLS models, i.e. based respectively CD or IR spectra alone, as well as CD and IR data combined. The feasibility of employing PLS/O-PLS analysis to extract quantitative data for common protein degradation forms was successfully demonstrated for an Fc fusion protein. Both CD and IR spectra contained the relevant information, nevertheless, CD-based O-PLS models achieved a higher accuracy compared to that of IR-based models for predicting aggregate and oxidation levels, while the accuracy for free Fc levels could be equally well predicted. Combining CD and IR data improved the accuracy of the prediction for all degradation forms. In addition, we demonstrated that O-PLS models yielded to a better accuracy compared to that obtained with PLS models. The last part of the thesis is based on the "protein design" methodologies. Aim of the present thesis is to study the scaffold stability of contryphan-Vn, a small peptide isolated from the venom of Conus ventricosus formed by only 9 residues and characterized by the presence of a single disulfide bridge, after substitution of 4 of 9 amino acids of its sequence. Contryphans are bio-active peptides, isolated from the venom of marine snails of the genus Conus, which are characterized by the short length of the polypeptide chain and the high degree of unusual post-translational modifications. The cyclization of the polypeptide chain through a single disulphide bond, the presence of two conserved Pro residues and the epimerization of a Trp/Leu residue confer to Contryphans a stable and well defined structure in solution, conserved in all members of the family. The potential of Contryphans as scaffolds for the design of redox- active (macro)molecules was tested by engineering a copper binding site on two different variants of the natural peptide Contryphan-Vn, named Cupryphan and Arg-Cupryphan through the introduction of four His residues. The binding site was designed by computational modelling and the redesigned peptides were synthesized and characterized by optical, fluorescence, electron spin resonance and nuclear magnetic resonance spectroscopy. The novel peptides, named Cupryphan and Arg-Cupryphan bind Cu2+ ions with a 1:1 stoichiometry and a Kd = 1.3(± 0.2) x 10-7 M and 1.0(± 0.4) x 10-7 M, respectively. Other divalent metals (e.g. Zn2+ and Mg2+) are bound with much lower affinity. In addition, Cupryphans catalyze the dismutation of superoxide anions with an activity comparable to other non-peptidic superoxide dismutase mimicks. We tested the potential of conopeptides as scaffolds for the engineering of novel, metal based, biocatalysts starting from the simplest prototype of disulphide constrained conopeptides: the Contryphans. The results of the present work indicate that indeed this class of peptides could be successfully exploited to engineer novel, stable and redox active macromolecules.