Process Analytical Technologies in Biopharmaceutical Process Development

Biopharmaceuticals already represent a broad field of new and innovative medicines that promote a longer and healthier life, either by preventing and inhibiting the progression of diseases, or by treating diseases such as hepatitis, diabetes, multiple sclerosis, AIDS, stroke, rheumatoid arthritis, breast cancer, leukemia, and other rare diseases. Currently hundreds of millions of patients benefit from these medicines and it is expected that this number will increase even further in the future. However, process development for biopharmaceuticals presents a number of significant constraints, primarily as a result of the high complexity of these molecules with respect to their structure and biological activity. Also, biopharmaceuticals are synthesized by living cells with inherent variability, further enhanced by sensitivity to the manufacturing environment. Due to the challenges associated with the production of biotech therapeutics and the aim to protect consumers, global regulatory agencies have imposed stringent quality guidelines. Over time, these expectations have evolved in the form of the Quality by Design (QbD) framework that encloses the scientific understanding of the whole product life cycle including the manufacturing process, which together with risk-based approaches can build quality into the process (http://www.ich.org/). Process Analytical Technologies (PAT) represents a key tool in the QbD paradigm, as it is used to control the whole manufacturing process, based on monitoring and analysis of the critical attributes of the raw materials and the process. PAT is therefore used to design, measure, analyze and control the whole manufacturing process, thus leading towards a high quality and regulatory compliant product. Furthermore, PAT may be used for speeding up the development process towards an optimized and more economic manufacturing process, as will be demonstrated in this special issue. The main focus of this special issue is on PAT applications in biopharmaceutical process development, and includes some PAT examples from related fields as well. We have been fortunate to receive contributions from the leaders in this exciting field. The scope of the issue covers key concepts, enabling technologies as well as applications. The first paper from Rathore and Kapoor1 presents a review of PAT applications for monitoring commonly used downstream biotech unit operations. The review covers recent advancements (last five years) and discusses analytical tools that have been used in these approaches with reference to their speed of analysis, robustness and sensitivity. The next mini-review2 is from Mandenius and Gustavsson and discusses how soft sensors can enable PAT implementation. The paper provides an excellent overview of the existing soft sensor alternatives and discusses how these can be configured to meet typical industrial needs. We expect this mini-review to inspire further implementation of soft sensors in the industry. The next set of papers3-9 discusses applications of the various analytical tools towards creating PAT applications for upstream unit operations such as microbial fermentation and mammalian cell culture. The contribution from Ashton et al.3 investigates the applicability of UV resonance Raman (UVRR) spectroscopy as a probe for residual cellular DNA and RNA in mammalian cell culture medium. Variations in DNA and RNA UVRR spectral profiles of medium-cellular footprint samples were identified and related to time of harvest and increased cell lysis that is associated with a loss in cell viability. Svendsen et al.4 present a case study where a lactic fermentation process is monitored by different techniques (BRIX, NIR and Fluorescence Spectroscopy) providing different data structures (zero-, first- and second-order measurements structures). Multivariate data analysis (PCA and PARAFAC) has then been applied on the first- and second-order datasets. The paper presents and clarifies the advantages and variation in the structure of the different data structures. Lopes et al.5 explore the possibility of in-situ monitoring of plasmid production in E. coli cultures using a near infrared (NIR) fiber optic probe for predicting bioprocess critical variables such as the concentrations of biomass, plasmid, carbon sources (glucose and glycerol) and acetate. The paper demonstrates that NIR spectroscopy combined with PLS modeling provides a fast, inexpensive and contamination-free technique to accurately monitor plasmid bioprocesses in real time, independently of the medium composition, cultivation strategy and the E. coli strain used. Wechselberger et al.6 attempt to estimate a novel physiological variable, bio-density, using the combination of dielectric spectroscopy and soft sensors in recombinant E. coli bioprocesses. Applying this concept, batch/ fed-batch and induction phase physiological changes were successfully observed in real time and a correlation with the specific product body titer was observed. Lee et al.7 use various kinds of chemometric tools to understand the spectral dynamics of the dielectric spectroscopy over the extended period of a CHO cell culture, covering different cell growth phases. The authors concluded that the best prediction performance from the dielectric spectroscopy could be realized by using the regression model based on locally-weighted partial least squares (LWPLS), compensating the spectral dynamics observed during the different cell growth phases inherently in the given estimation model. The contribution from Striedner et al.8 involves evaluation of Partial Least Squares regression and Radial Basis Function Artificial Neural Network based models to predict gene dosage, product titer and solubility of the target protein expressed in E. coli. The paper from Yoon et al.9 examine the effects of raw material variability in regulating the intracellular metabolic pathways of CHO cells. Plant-derived supplements of wheat hydrolysates were employed as a model system of raw materials having high compositional variability. The intracellular metabolic fluxes of antibody-producing GS-CHO cells supplemented with different lots of wheat hydrolysates were quantified using the flux balance analysis (FBA) technique. The paper shows how metabolic flux analysis coupled with multivariate data analytics unravels the complicated effects of raw material variability on the cellular metabolism, and hence improves the overall productivity of an antibody-producing CHO cell culture. The role of multivariate data analysis and mathematical modeling in elucidating the complex interactions that exist in most biological systems is being appreciated more and more. Glassey et al.10 investigate the effect of operating conditions upon cell metabolism and the glycosylation profile of monoclonal antibody produced using hybridoma cell culture. Multi-way PCA analysis of on-line process data and amino acid concentration profiles reflecting the cell metabolism indicated significant dependence on the operating conditions, particularly DO and pH. Delvigne et al.11 investigate yeast based processes by using on-line flow cytometry in combination with a fluorescent transcriptional reporter (GFP) and viability fluorescence tags (propidium iodide, PI) for different bioreactor operating conditions. The authors point out the importance of understanding the segregation mechanisms for the applied fluorescent reporters, to judge whether simple mathematical tools may be applied or if more sophisticated computational tools are needed for the quantification of the microbial population segregation. Heins et al.12 explore the use of a two-compartment, scale-down setup, consisting of two interconnected stirred tank reactors, which was used in combination with mathematical modeling to mimic large-scale continuous cultivations. The first reactor represents the feeding zone with high glucose concentration and low oxygen, whereas the second one represents the remaining reactor volume. Using this setup, the concentration profiles of biomass and glucose could be successfully validated experimentally. The last paper of the issue by Mendhe et al.13 presents a case study on implementation of PAT in downstream processing. A variety of PAT based and pseudo-PAT based approaches are evaluated for pooling of a real commercial process chromatography column. The approaches examined include pseudo-PAT approaches such as ultraviolet (UV) absorbance, % peak height, retention time, and relative retention time. These approaches do not involve direct measurement of product quality attributes, but are operationally simpler to implement. In addition, a feed-forward control based model and a rapid HPLC method have also been explored. A comparison of the pros and cons of the various approaches that have been examined is also presented. In summary, the issue presents the spectrum of research that is being undertaken under the umbrella of PAT. We think that this space is likely to stay vibrant for the coming years as it presents researchers with a variety of avenues for further work. Newer analytical tools that offer rapid analysis of the quality attributes of the product and particularly integration of the monitoring data with smart control schemes need to be demonstrated. There is also a lot of scope and need for process modeling as well as in-depth economic analysis that takes a holistic view of the entire process and not just a single unit operation. We hope that researchers working in this area enjoy this issue, and will use it as a source of inspiration for their own PAT development activities. Lead Guest Editor Anurag S. Rathore, Indian Institute of Technology, India Guest Editors Krist V. Gernaey, Technical University of Denmark (DTU), Denmark Cecília R.C. Calado, Catholic University of Portugal, Portugal Seongkyu Yoon, University of Massachusetts, USA