Moving PAT from Concept to Reality

Process analytical technology (PAT) is applied in the biopharmaceutical industry for analysis of raw materials, in-process monitoring, and final product analysis. PAT is not only enabling, but essential, to continuous bioprocessing. With sufficient advances, emerging PAT solutions should ultimately make real-time release testing possible.

In-line, at-line, and on-line

To understand the state of PAT technology development, it is necessary to understand the different types of PAT operations used in biologics manufacturing. In-line sensors are placed in a process vessel or process stream to conduct the analysis in-situ. On-line sensors are connected to process side streams and perform periodic automatic sampling, returning fluid back to the process streams after analysis is complete. At-line—or off-line—analyses involve collection of a sample from the process, with analysis performed away from the process. In-line and on-line sensors allow for continuous process measurement and control, while at-line measurements cause a delay between sampling and availability of the results, preventing their use for direct process control.

Still on the learning curve

The biopharmaceutical industry and regulatory agencies have recognized the value of PAT. Although many large multinational manufacturers have adopted and implemented various forms of PAT, many companies are still struggling to get started, according to Joe Makowiecki, enterprise solution architect with Cytiva.

“Despite the fact that the concept is supported by most scientists, implementation is limited at commercial manufacturing sites. Adoption of PAT is certainly not progressing as fast as was expected 16 years ago,” asserts Moheb M. Nasr, principal consultant with Nasr Pharma Regulatory Consulting. “While we [have seen] an increase in the interest in and value perception of PAT among our customers within the last two to five years, we believe that its full potential is not yet being used,” agrees Svea Grieb, product manager for PAT at Sartorius Stedim Biotech.

In many cases, Makowiecki adds, reliable, low cost, on-line/at-line analytical sensors needed for the measurement of ­critical product quality attributes and product-related species are not commercially available. An additional challenge is the development of sensor/detector/measuring technologies that allow for PAT in the single-use space.

For companies that have adopted PAT, these tools are applied for raw material analysis and monitoring/control of process performance, according to Arpan Mukherjee, technical committee coordinator for the National Institute for Innovation in Manufacturing Biopharmaceuticals (NIIMBL). Raw-material characterization is performed using handheld Raman/near infrared (NIR) scanners. In bioreactors, in-line sensors measure pH, temperature, dissolved oxygen (DO), and carbon dioxide; in-line optical sensors (Raman, NIR) are coupled with multivariate modeling to measure metabolites. In addition, a variety of on-line tools measure cell density (dielectric spectroscopy, fluorescence spectroscopy); protein aggregates (size-exclusion chromatography and ultra-high-performance liquid chromatography [UHPLC]); product concentration (HPLC); and protein charge variants (ion exchange chromatography). Numerous at-line analyses evaluate host cell proteins (HCPs), DNA, viable cell density, bioburden, aggregates, product concentration, and particulates.

One of the greatest difficulties with many current at-line methods is the extended length of time required to receive results—for instance, 28 days for most cell-based assays. The focus of research in these areas is the development of rapid methods that are sufficiently accurate and robust to allow adoption.

Upstream focus

Most PAT tools that find wide use today are designed for application in upstream processes. “This focus is in large part due to the fact that initial protein quality and effectiveness are established during cell culture or fermentation. Changes in the bioreactor conditions during protein production will impact all downstream processes. In addition, the largest number of sensor technologies available for use in biologics manufacturing are designed for upstream applications,” says Richard D. Braatz, Edwin R. Gilliland professor at the Massachusetts Institute of Technology (MIT). “Trace metals, glucose, lactate, pH, DO, amino acid content, glycosylation profiles, and HCPs must be measured to monitor the process and enable feedback/feedforward control to optimize yield (in-line and on-line only for control),” Mukherjee observes.

The challenge with downstream unit operations is partly the lack of obvious opportunities for using sensors, Braatz notes. In column chromatography, for instance, analysis cannot be performed until after the column, preventing any in-process monitoring. The diversity of downstream processes is also a consideration. “Not only are there many possible downstream unit operations, but also the critical quality attribute (CQA) most impacted by each step is diverse. The traditional methods for quality attribute measurement are also often product-specific,” says Makowiecki. As a result, the diversity of downstream processing design makes it difficult to develop a universal set of technologies that will cover all of downstream processing for all molecules and their associated quality metrics.

Next-generation therapies are creating additional opportunities for PAT adoption, though. In personalized medicines such as autologous cell therapies, for instance, the nature of the treatment demands a process that is flexible and can dynamically adjust to wide variations in starting material, according to Grieb. PAT can account for the variations and peculiarities of the cells from different patients in an automated fashion, enabling a high process consistency irrespective of the starting material.

PAT can also help overcome the analytical challenges presented by viral vectors for novel vaccines and gene therapies. These products are not well-characterized molecules like monoclonal antibodies, but a complex of various proteins, DNA/RNA, and in some cases lipid membranes. “This complexity makes it hard to identify and understand the factors influencing the product CQAs. Hence, these processes benefit from a stricter control strategy, where high levels of automation and implementation of PAT and advanced data analytics play a key role,” Grieb asserts.

Essential for continuous manufacturing

Process intensification/continuous bioprocessing is a hot topic in the biopharma industry at the moment because it enables increased productivity of single-use facilities while decreasing the footprint. It also brings incentives to invest in PAT, and advances in sensor technology/connectivity will make continuous processing possible, according Makowiecki.

Intensified processes are much more complex than conventional fed-batch processes and therefore require tighter monitoring and control, adds Grieb. “PAT and automation do not only provide this, but also reduce the complexity for the operator,” she says. In Nasr’s view, PAT is critical for achieving the enhanced controls required to ensure quality throughout the entire batch during operation of continuous processes. “Regulators expect process monitoring and controls based on the entire batch data and the ability to revise controls to compensate for raw material variability or the use of different batches of raw materials,” he observes. In addition to in-line process control, there is also a need for advanced in-line PAT for continuous/hybrid batch manufacturing to enable real-time batch release, according to Mukherjee.

Many drivers for additional development

One of the biggest advantages to using PAT and biopharmaceutical manufacturing is the increased process understanding that is gained, which leads to more consistent product quality, according to Ruben Carbonell, chief technology officer for NIIMBL. It also allows for increased productivity, connectivity and automation, says Makowiecki. Reducing the need for manual sampling lowers the risk of operator error and contamination, while the timely identification and correction of process irregularities will help minimize the risk of lost batches, Grieb notes.

Furthermore, asserts Grieb, a well-characterized and monitored process together with scalable hardware can significantly reduce the cost and efforts of process scale-up and scale-down. Overall, therefore, PAT contributes to acceleration of process development timelines for reduced time to market, concludes Carbonell.

PAT should also ultimately make real-time product release possible. “PAT is bringing quality control testing closer to the manufacturing floor for on-time/real-time release and development of predictive models to enable active process control and reduce batch rejection rates,” Makowiecki asserts. “Real-time release is important for reducing the time to market. Instead of sitting on a shelf in storage for up to several months, medicines can be put in the hands of patients very quickly if real-time release is possible,” Braatz states.

Speeding drug development is crucial given that many drugs in clinical trials fail to reach the market. Drug companies would like to start manufacturing process development later in the overall development cycle—preferably during Phase III of clinical trials—in order to focus their investments on candidates that have the greatest likelihood of obtaining regulatory approval. By reducing process development timelines, PAT is making it possible for companies to take this approach, which saves time, money, and effort, according to Braatz.

On a similar note, advanced PAT solutions often serve as better methods for advanced process development. They allow for monitoring of processes and product attributes that can provide the most optimal candidates to take forward in terms of quality attributes and productivity, according to Stacy L. Springs, executive director of the Biomanufacturing Program at MIT.

Braatz adds that PAT provides increased production flexibility. “There is a lot of uncertainty about the market size when drugs are in development; the actual demand will depend on whether a drug reaches the market first before other competitors and how accurate predictions of the user base are. The more information that a company has about its processes, the better decisions it can make, such as about whether to scale up in single-use or stainless-steel equipment,” he explains.

But barriers to adoption remain

While PAT implementation affords many benefits, there also exist many hurdles that must be overcome before it becomes widely adopted across the biopharmaceutical industry. Cost is certainly an issue. So is the lack of experience and experts with training in advanced analytics and modeling, according to Nasr.

There are regulatory challenges as well, says Grieb. “Some concepts of modern automation technologies and sensor technologies are not yet covered by regulatory guidelines, particularly those used for multivariate data analysis, which takes all available data and integrates them into a fingerprint. The adoption of such batch-fingerprinting concepts must be considered by the regulatory bodies.” The same questions arise for multi-analyte sensors that are based on computational models, which is the case for spectroscopy, for example, she adds. The risk of delay to a filing application can result in the lack of a business case, according to Carbonell. The complex regulatory framework around the world also creates risk aversion.

Certain information technology questions remain, too, Grieb notes. A comprehensive automation strategy for an entire bioprocess, and potentially an entire production site, requires connectivity of all components and a centralized control unit. “That would require data sharing and access that implies safety risks. We experience reluctance among our customers to adopt new technologies such as cloud computing and wireless communication of PAT components,” she explains.

On a more basic level, Carbonell adds that there is often lack of confidence in testing robustness and integrity testing. In addition, lack of thorough process understanding can lead to ambiguity regarding the attributes that should be measured. This barrier is readily evident with the lack of integration into the existing process equipment installed base, automation platforms, and control strategies, adds Makowiecki. He also comments that the ability to identify the right opportunities to implement PAT without causing commercialization delays or rework, the need to design in PAT solutions through process development stages, and to really validate them during scale-up prevents adoption.

For continuous bioprocessing in particular, the lack of experience with PAT at commercial sites that have typically performed batch processing combined with the perceived business and regulatory risks of implementing new technologies are hindering PAT implementation, according to Nasr. “The lack of availability and reliability of cost-effective in-line/on-line analytical sensors needed for the measurement of CQAs and product-related species is one of the gaps in continuous processing, but with many companies evaluating continuous processing, there is also synergy to explore PAT as an enabler,” adds Makowiecki. Sartorius also thinks intensified/continuous processing will boost novel PAT solutions, because as companies establish new manufacturing pipelines with unique requirements, they can justify the costs and efforts of going through the approval for commercial manufacturing.

Technology gaps to be addressed

Overall, there needs to be a standard path to implementation for PAT, with technology available that fits each application and does so across all manufacturing scales and a practical approach to connectivity, according to Makowiecki. A higher degree of automation for upstream bioprocesses and standardization of the process steps would lead to improved batch-to-batch consistency, and in turn, product quality, Grieb agrees.

Grieb also notes that, while most PAT implementation has targeted upstream processes, there are plenty of examples where PAT could significantly improve downstream processes as well. Examples are automated venting of filters and protein quantification during column loading. “We expect to see more downstream targeted PAT implementation within the next years,” she says. In particular, Springs points to a need for low-cost, robust, downstream PAT solutions that are non-invasive and do not require advanced interfaces with equipment, which can impact sterility.

There are also specific sensor technologies that need improvement or have yet to be developed, according to Kelvin Lee, NIIMBL’s director. Examples include high-resolution, specific, robust PAT tools with a low limit of detection and in-line, reusable sensors with high frequency measurements that do not require calibration over the course of the batch. The ability to characterize proteins in a bioreactor on-line via a reliable, reproducible, rugged, high-precision, and high-throughput liquid chromatography–mass spectrometry (LC–MS) instrument at a cost of $300,000 rather than $1.3 million would be a game-changer, asserts Braatz.

Although a difficult problem to solve, sensors that rapidly measure viral and microbial contamination are needed the most, Braatz states. The long (28-day) time to receive results from cell-culture-based sterility tests could delay product release when needed, especially for vaccines or autologous cell therapies like chimeric antigen receptor (CAR) T. That is why application of rapid methods and next-generation sequencing for sterility testing is creating so much excitement, adds Springs. As importantly, if contamination is caught sooner using in-process rapid testing for virus or microbes, contamination can be stopped before spreading downstream in the process, saving both money and time.

Emerging tech is exciting

Many PAT technologies are under development in both academia and industry. Braatz and collaborators at MIT and Biogen conducted an 18-month study to identify different technologies in use and in development for a wide range of CQAs and classified them according to the timeframe in which they were likely to be adopted (1).

Some of the promising emerging PAT solutions include slanted nano-arrays and spectral analysis with partial least squares for protein aggregates; microfluidic technology for analysis of post-translational modifications and sequence variants; nanotechnology-based analysis of charge heterogeneity that does not require protein purification; aptamer-based biosensors for N-glycosylation evaluation; next-generation sequencing (NGS) for viral and microbial contamination quantification; surface plasmon resonance, amplified luminescent proximity homogeneous assay, and fluorescence resonance energy transfer for target and Fc binding analysis; liquid chromatography coupled with ultraviolet, fluorescence, or MS detectors for evaluation of process-related impurities; and variable pathlength spectroscopy solutions for protein concentration measurement in real time (1).

Microfluidic devices and contact-free PAT technologies are among the most recent advances but are still mostly in the laboratory proof-of-concept stages, according to Carbonell. “Microfluidics-based analyzers require a small sample volume and provide higher resolution than existing PAT technologies,” he explains. He points to three additional microfluidics technologies—particle analyzers for on-line protein aggregation detection and monitoring, devices for on-line or at-line detection of adventitious agents, capillary electrophoresis and ­capillary isoelectric focusing-based microfluidics devices for real-time protein analytics—and MS-integrated microfluidics.

“These technologies are currently being evaluated by the biopharma industry and are in different phases of development (alpha and beta testing). They have the potential of increasing process/product knowledge, early detection of contaminants, and reducing the need for comprehensive end-point quality testing,” Carbonell says.

Spectroscopic methods (both Raman and NIR) are currently deployed but are not yet widely adopted, according to Makowiecki. Sartorius Stedim Biotech also believes that spectroscopic techniques will become more abundant in both upstream and downstream bioprocessing, due to its capability of label-free, online measurements of several analytes, cell properties, and product quality attributes. “Spectroscopy has the potential to replace offline measurements during the bioprocess. We envision the use of a combination of different spectroscopic techniques—such as NIR, Raman, and UV-Vis—to be required for this,” says Grieb.

Soft sensing (transforming existing signals and leveraging advanced process understanding to infer/gain visibility to an attribute without direct measurement) is making it possible to monitor/predict performance through use of surrogates, according to Makowiecki. “As a result, the industry doesn’t have to wait for sensor maturity for CQAs to begin their PAT journey,” he says.

Commercial impacts

The application of sophisticated PAT tools in combination with multivariate data analytics is starting to have a high impact on commercial processing, according to Grieb. “Measurements are moving forward in the process to the point of controllability. Using process fingerprints, the state of the process can be assessed at any time. Furthermore, through real-time univariate and multivariate process monitoring, data can be used for simulation and modeling of process design and control and ultimately lead to prescriptive analytics for product quality,” she explains. In the near future, Sartorius Stedim Biotech also expects wider spread adoption of analytics in GMP that are already available, such as spectroscopy for metabolite control and bio-capacitance for viable biomass.

Braatz is looking forward when cost of LC–MS systems is reduced sufficiently to allow widespread adoption for real-time monitoring of bioprocesses. He notes that few people in the 1980s would have expected powerful, easy-to-use, small-sized LC–MS systems to be hundreds of thousands of dollars, so development is on the right trajectory. NGS technologies are also very promising. “Next-generation sequencing could be very helpful as it is applied to more advanced and complex therapies, such as genome-edited products,” Springs asserts.

Springs adds that some of the technologies being developed to measure the sterility of autologous cell therapies should be applicable to traditional biologics. “We may see more innovation in the cell-therapy space first,” she says.

Overall, Braatz believes that PAT development has reached a point where there have been sufficient advances that have resulted in real improvements in analytical capabilities that the concept of real-time-release testing can be realistically considered. “We are getting to a very exciting place now,” he says.

Collaboration is needed

Increased adoption of PAT will be driven by regulatory clarity and support, industry collaboration, advancements in connectivity and technology, the introduction of PAT-focused platforms and services by vendors, and further implementation of continuous processing, according to Makowiecki.

Emerging technologies are promising, but appropriate software and hardware development is necessary for integration into existing bioprocesses and for automation, especially for continuous/semi-continuous processes, notes Lee. “The coupling of these technologies with new automation systems is likely to dominate the future of biopharmaceutical manufacturing. A key challenge is the need to de-risk the adoption and implementation of these technologies in a highly regulated environment. We believe that public-private partnerships that enable multiple stakeholders to innovate collaboratively provide the opportunity to de-risk such advances,” he states.

“It is essential to reduce the technical risk of emerging PAT solutions, especially as the industry moves toward real-time feedback control and scales out the manufacturing of autologous cell therapy products,” according to Springs. “By working together, we can find the best solutions that allow medicines to get to the patients faster,” she concludes.

1. M. Jiang, et. al., Biotechnology and Bioengineering, 114 (11) 2445–2456 (2017).