This case study describes the implementation of process analytical technology on the cultivation process step of a whole-cell vaccine against whooping cough disease.
Online monitoring of product quality and continuous process optimization strategies are common in many industries, but only recently has the US FDA recognized the need for the pharmaceutical industry to come to the same level. To meet this need, the FDA launched the process analytical technology (PAT) guideline,1 which prescribes quality assurance to become part of the manufacturing process, referred to as Quality by Design (QbD), in contrast to quality by testing, which relies on the postproduction testing and release of the product. Under QbD, the observed processes need to be well-characterized, and manufacturers need to know the critical attributes of each process, each product and their interactions.2
The EMEA and the Japanese regulatory authorities have adopted PAT principles, which, through the International Conference for Harmonization (ICH), resulted in three new guidance documents (Q8, Q9 and Q10). These documents are now considered the world standard for this new regulatory concept (i.e., QbD with PAT) for product and process quality.3 For pharmaceutical small molecules with relatively simple production processes, the application of PAT is becoming increasingly common. Acceptance of PAT concepts has been much slower for biological products. The FDA has therefore invited biopharmaceutical companies to work with them for pilot submissions of biopharmaceutical PAT applications.4
(Aaron Graubart/Getty Image)
The Netherlands Vaccine Institute's (NVI) PaRel project is a good example of a successful PAT implementation in a bioprocessing environment. The project's aim was to implement PAT in the cultivation process step in making a whole-cell vaccine used to prevent whooping cough. This process step involves batch cultivation of the Bordetella pertussis bacterium — the most complicated and critical step in this production process and, as such, was chosen for the development of the tools, equipment, and knowledge necessary for full PAT application. Compared with a small-molecule drug, this vaccine is relatively undefined and complex, which makes it more difficult to ensure product quality during processing. The FDA has acknowledged that the approach chosen in the PaRel project can lead to approval of a PAT application for a complex biopharmaceutical product such as a whole-cell vaccine.
Understanding the cell-cultivation process requires an understanding of the interior physiology of the cell and how it responds to external changes, either because of process changes or disturbances. This understanding is important, especially in a batch-wise operated bioreactor, where continuously changing conditions may influence cellular physiology. To achieve this, a full genome DNA microarray analysis was developed for monitoring messenger RNA (mRNA) expression profiles of B. pertussis. By measuring gene expression levels during cultivation, the pathways or specific proteins affected by changes during processing can be identified. In this way, disturbances of specific process parameters can be assessed for impact on product quality. It is also possible to make determinations regarding reproducibility between batches and the optimal harvest point by examining the gene expression profile.
In a whole-cell vaccine, the outer membrane proteins are the targets presented to the immune system; therefore, the outer membrane protein composition is crucial for vaccine quality. Streefland et al. described a method for identifying genes that encode those outer membrane components known to induce a protective immune response.5 A highly conserved molecular switch, called the Bordetella virulence gene (Bvg) system, regulates the virulence genes of the Bordetella genus. This means that a single extracellular signal can induce the complete physiological switch between the virulent and the nonvirulent state.6 All genes involved are controlled by the same operon system. For some strains, these virulence genes have already been investigated,6,7 but not in terms of optimal outer membrane composition for vaccine manufacturing.
The dissolved oxygen concentration was one of the process parameters investigated in the PaRel project. Because B. pertussis is an obligate aerobic organism, the availability of oxygen during cultivation is essential. To investigate this, a series of cultivations were oxygen-limited for 90 min, and samples were taken immediately before and after limitation, as well as at the end of cultivation for microarray analysis. Oxygen limitation had a strong effect on gene expression levels of many genes immediately after the event, including virulence genes. At the end of cultivation, however, the oxygen-limited cultures could not be distinguished from the standard control cultivations. This indicates that oxygen limitation can have an effect on vaccine quality, but this effect is reversible. As long as oxygen limitation does not occur at or near the end of cultivation, it is not a crucial factor in manufacturing the vaccine.8
Using microarrays to monitor crucial genes enabled us to analyse the critical process attributes for the cultivation of B. pertussis. High expression of the 56 virulence marker genes is associated with high cellular virulence and thus good expected product quality. Low expression levels are associated with poor expected quality. Therefore, a weighted average of the expression levels of these virulence marker genes can be used to predict the quality of the bacterial suspension at the end of cultivation. This so-called product quality score can be used to objectively compare batches or samples of the same batch with each other.
The product quality score was used to determine a critical event in batch cultivation: the optimal harvest point.9 As biomass accumulates exponentially, nutrient concentrations drop exponentially (Figure 1), which can result in a range of unwanted effects, including cell lysis or the suppression of virulence genes.10 To investigate the optimal harvest point, four identically operated cultivations were each sampled at 11 time points for microarray analysis. A continuously changing gene expression pattern was expected because of the continuously changing extracellular environment associated with batch cultivation. However, gene expression proved to be relatively constant, which resulted in high product quality scores during most of the cultivation. Towards the end of cultivation, the product quality score dropped sharply (Figure 1). This drop coincided with the depletion of the nutrients lactate and glutamate; a determination that allowed the optimal harvest point to be accurately predicted. Therefore, measuring lactate and glutamate concentrations during cultivation demonstrated that the bacteria can be consistently harvested before nutrients become limiting, assuring the optimal composition of the bacterial outer membrane.
Figure 1: Batch profile analysis for the cultivation of Bordetella pertussis. Diagram (a) shows the increase in cell numbers for four cultivations and the decrease in average concentrations of the nutrients lactate and glutamate. Lines A to K in Diagram (a) indicate the points at which samples for microarray analysis were taken. Diagram (b) shows the relative product quality scores at the sample points A to K based on Diagram (a). It is clearly shown that the product quality score is consistently high for most of the process. Towards the end of the cultivation (as shown by points J and K), the product quality drops sharply. This decline corresponds with the depletion of important nutrients [Diagram (a)] in real time.
The definition of PAT states that it is a "system for designing, analysing 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".1 With the online measurement of nutrient concentrations and the supporting evidence that this correlates with virulence gene expression and subsequently outer membrane composition, the requirements for a PAT application are fulfilled, at least for the harvest-point determination. For other critical process attributes, several sensors are available to measure pH or dissolved oxygen, for example. These sensors, however, only monitor a single parameter, which means one could risk missing a critical parameter or be required to mount numerous probes on the bioreactor system for each measured parameter.
To monitor many parameters simultaneously, a near infrared (NIR) spectroscopic probe that measured a spectrum between 800 and 2500 nm was placed inside the bioreactor. The instrument's range, situated between visible light and far infrared, is sensitive to chemical changes (band vibration overtones) and physical changes (combination bands) such as optical density, viscosity, particle size and particle morphology. This range makes the technique highly suitable for monitoring bacterial cultivation processes, as was confirmed by initial pilot studies.11
To make use of the "timely measurements" (as stated in the PAT definition) to control the manufacturing process, the online process PAT data, including NIR, need to be readily available for process control models that feed back into the process. To allow this, the SIPAT software from NVI's project partner, Siemens (Belgium), was used. This software gathers all data that are measured online (i.e., pH, dissolved oxygen, temperature, NIR, controller outputs, gas flows, etc.) and stores them in a central database with aligned timestamps. The applied IT architecture is schematized in Figure 2.
Figure 2: The PAT IT infrastructure consists of two bioreactors, each monitored by a base station (Reactors 1 and 2). The process data are collected from Siemens (Belgium) PCS7, a process control system, as OPC (an industrial standard for data transfer between IT systems) tags by SIPAT, a software system that allows an online check if a new process is running within the tested process design space. Near infrared (NIR) measurements are collected from a Bruker (Germany) Matrix F Multiplex analyser running Bruker OPUS software. One collector station controls both measurement channels in parallel. Operational data are stored in a local database and then pushed to the central SIPAT database running on the SIPAT server. Quality predictions are made in real time by the integrated Simca Q calculation engine on the server. The bioreactors are controlled by a batch server running Siemens Batch CC, a batch engine system
The SIPAT database is accessible in real time for process models that can be run in the integrated Umetrics Simca (Sweden) software. Output from these models flows back into SIPAT and can be used by the bioreactor control system to make adjustments to the process and to enable closed loop controls. This software allows the use of any process sensor to monitor and control any process step. In this case, the development of the SIPAT software allowed the integration of NIR data with the other process data, making a true PAT application and a real-time release strategy possible.
With the PAT initiative also came the concept of process design space. ICH defines "design space" as "the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality".3,12 This means that product quality is no longer assured at specific process settings, but rather that a range of settings should be explored. One of the best ways to explore the process design space is by using design of experiments (DoE).13 In this way, several crucial process parameters are tested at a range of settings using a minimal number of experiments. The main benefit of this approach is that it allows one to investigate the interaction between crucial parameters at different settings. This approach was used to explore the process design space of the cultivation of B. pertussis by executing a series of designed experiments in which several critical process parameters were varied simultaneously. SIPAT was used to collect all of the experiment's variables.2
Based on these experiments, a process model that describes the design space for the cultivation of B. pertussis was constructed for the tested ranges of the critical parameters. This process model can be executed online using the SIPAT software, which allows an online check if a new process is running within the tested process design space. Newly gathered data can be added to the process model so that it becomes more accurate with time, allowing the process to be optimized with time. Ultimately, this model can be validated so that it assures product quality online, allowing real-time product release.
The author says...
The principles of PAT can be applied to biopharmaceutical products, not just small-molecule drugs. Biopharmaceutical processes, particularly the cultivation process step, are intrinsically more complex than processes used to make chemical drugs. By applying a sound scientific approach for development of process understanding and by using the appropriate sensors and monitoring techniques, a biopharmaceutical process can become as stable and capable as any world-class manufacturing process.
The ultimate benefits of PAT application are mitigation of risks (failures) during manufacturing, flexibility in process optimization, and safer and more pure products. The effort to understand the process will be rewarded when scale-up, process changes, or even medium optimizations are more easily implemented throughout the product's manufacturing lifetime. Furthermore, online monitoring of product quality and subsequent adjustment to the optimal trajectory will lead to fewer failed runs and a more consistent product of high quality. Real-time quality assurance will lead to shorter cycle times and fewer stockpiles. This case study shows that monitoring cell cultivation, combined with process understanding condensed in chemometrical models, allows the optimal harvest point to be determined, and thereby allows the monitoring of product quality throughout the batch, opening the possibility for real-time product release. This increased process understanding leads to shorter production times and faster detection of batch failures, reduces cost and increases productivity.
These benefits have encouraged the FDA to make PAT application the mandatory backbone of future pharmaceutical process development and manufacturing. This case study demonstrates that it can even be applied to a complex and undefined product such as a whole-cell vaccine, which means it should also be feasible for any pharmaceutical or biopharmaceutical product currently on the market.
The PAT initiative can, therefore, help to increase the safety and efficacy of medicines while reducing the time-to-market for new products and the operational costs of manufacturing. This situation is good news for regulatory inspectors, manufacturers and patients.
Kjell François, PhD is senior PAT consultant for Siemens NV, Curie Square 30, 1070 Brussels, Belgium. Kjell.Francois@siemens.com
Mathieu Streefland, PhD is project leader of the PaRel project at The Netherlands Vaccine Institute (NVI).
Rebecca Vangenechten is a consultant for business and project development life sciences USA at Siemens NV.
Leo Hammendorp is director of business and project development for global life sciences at Siemens NV.
1. FDA, Guidance for Industry: PAT — A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance (Rockville, MD, 2004). www.fda.gov/cder/guidance/6419fnl.pdf.
2. M. Streefland, "From Process Understanding to Process Control: Application of PAT on the Cultivation of Bordetella pertussis for a Whole Cell Vaccine," (Doctoral thesis, Wageningen University, The Netherlands, 2009).
3. ICH, Q8 Pharmaceutical Development (Geneva, Switzerland, 2005) www.ich.org/LOB/media/MEDIA1707.pdf .
4. FDA, "Submission of Quality Information for Biotechnology Products in the Office of Biotechnology Products; Notice of Pilot Program," Federal Register73, 37972–37974 (Rockville, MD, 2008).
5. M. Streefland et al., Vaccine, 25(16), 2994–3000 (2007).
6. C.A. Cummings et al., J. Bacteriol., 188(5), 1775–1785 (2006).
7. D. Hot et al., Mol. Genet. Genomics, 269(4), 475–486 (2003).
8. M. Streefland et al., Biotechnol. Bioeng., 102(1), 161–167 (2009).
9. M. Streefland et al., Biotechnol. Bioeng., 103(5), 900–908 (2009).
10. M.M. Nakamura et al., Infect. Immun., 74(10), 5537–5548 (2006).
11. E.N.M. van Sprang et al., Qual. Engin., 19(4), 373–384 (2007).
12. ICH, Q8 Pharmaceutical Development, Q9 Quality Risk Management, and Q10 Pharmaceutical Quality System. www.ich.org/cache/compo/363-272-1.html.
13. D.C. Montgomery, J. Roy. Stat. Soc., 48(2), 159–177 (2001).