The revelation last June that several sartans, a group of angiotensin II receptor blocker (ARB) high-blood pressure medications, were contained via their manufacturing process with potentially carcinogenic nitrosamines NDMA and NMEA highlights the need for accurate and real-time analyses during pharmaceutical manufacturing. While this bioprocess monitoring would identify impurities during synthesis or purification, it could also provide a means of optimizing reaction conditions at the same time.
But any change in monitoring must also allow for high-throughput, and avoid intensive man-hours and disruption to the manufacturing process, both of which could introduce inefficiencies. This delicate balancing act of improving quality while maintaining efficiency is the job of process analytic technology (PAT).
The prospects for PAT to make an impact looked promising in 2004, with the promulgation of the FDA’s guidance on the topic. Soon, regulators were pushing quality by design (QbD) strategies as a tangible goal of bioprocess monitoring. Yet the biotech industry quickly learned that adapting technologies designed for food or semiconductor industries to the manufacture of therapeutic proteins would not be a smooth undertaking.
Despite the technological, economic, and inertial barriers to widespread implementation that have been encountered, PAT has proliferated. According to a 2017 report by research aggregator ReportLinker, global markets for PAT reached $410 million in 2017 and is expected to reach $540 million by 2022, an annual growth rate of 5.5%. How much of that growth, or interest in PAT itself, is a consequence of regulatory pressures and how much would have occurred as part of the normal desire to understand and improve a process is anyone’s guess.
Bioprocess sensors
The backbone of any bioprocess monitoring platform is the deployment of sensors able to provide real-time in situ snapshots of conditions inside a bioreactor. However, sensor issues related to performance, robustness, and suitability to sterile manufacturing environments were known from the early days of biotech. Even when sensors operate reliably in situ, they require calibration, which complicates their deployment in bioprocessing in ways not seen in consumer-product or even food manufacturing.
Single-use sensor manufacturing opens up novel opportunities to exploit the benefits of a use-and-toss monitoring model, but manufacturers have not yet reached a balance point between cost, performance, and suitability, even for monitoring process characteristics as familiar as pH. The knock on single-use sensors has been their lack of physical and performance robustness. Developers have had difficulty bringing standard glass electrochemical sensors, which are normally reusable, into single-use processes. Disposable sensing also introduces the dilemma of how to deal with the sensors after use: Discarding glass-electrode probes is frivolous, while reuse flies against the spirit of single-use bioprocessing.
The alternatives are less-robust, underperforming sensors specifically designed for single use. But these represent a step backward in terms of performance and, ultimately, gaining process understanding. Several companies offer single-use pH sensors, for example, Hamilton, whose OneFerm single-use pH sensors claim “the high-accuracy and performance of a traditional glass pH electrode, even after gamma irradiation and dry storage.” Mettler-Toledo has a line of single-use pH and dissolved oxygen sensors with “a unified design that allows safe and easy installation in a standard 1” Eldon James port disk.”
And earlier in 2019, Broadley-James announced a collaboration with Pall to commercialize single-use pH sensors. Broadley-James’s sensors leverage the older glass-electrode design, but with improvements that make the sensors more suitable to bioprocessing. For example, calibrated buffer storage eliminates previous constraints of having to calibrate the sensor before sterilization, which affects measurement accuracy.
HPLC: Too much, too little
The lesson from the current state of bioprocess sensing, particularly of the single-use variety, is that an analytic modality suitable for one industry, or one setting, doesn’t necessarily work in another industry or setting. This has been true for HPLC, a method used extensively in biopharmaceutical discovery and development, but which is encountering all sorts of problems in PAT settings.
HPLC, particularly when paired with mass spectrometry, represents the ultimate in biological analytics. LC is a natural fit for PAT, since it can monitor everything inside a bioreactor. Yet this technology, which is ubiquitous in the life sciences, has been difficult to adapt to the semi-real-time requirements of bioprocess PAT. Instrument systems employed during discovery and early characterization studies are complex, require experts to operate, and have run times that are too long for rapid responses.
That is why, despite its extensive democratization in laboratories, LC/MS remains underutilized on the production floor.
As Jeff Mazzeo, PhD, vice president of global marketing at Waters, explains, “Our biopharmaceutical customers use Q-TOF mass spectrometers for deep protein characterization, but they also do a lot of routine monitoring on those instruments.”
Waters saw the need for an LC/MS system for routine analysis that is accessible, easily deployed, and that could theoretically return process control to the production floor and reduce pressure on core LC/MS facilities staffed by PhD scientists. “Routine monitoring doesn’t require their expertise, but it takes up instrument time,” Mazzeo adds.
For these analyses, Waters introduced BioAccord, taglined as “the first SmartMS-enabled biopharmaceutical LC/MS.” BioAccord is optimized for intact protein, released glycan, and peptide monitoring applications.
“BioAccord is for when you know what you’re looking for, but you want to use LC/MS to monitor it,” Mazzeo says. “It was designed to decentralize routine monitoring that supports formulation and bioprocessing and put it into the hands of scientists who do that work.”
Part of Waters’ long-range plan for BioAccord is to adapt it for at-line, and possibly inline PAT. “We think we can get BioAccord close to biomanufacturing, at least at-line,” says Mazzeo. In addition to its regulatory compliance features, BioAccord is self-calibrating and self-diagnosing, further reducing the need for trained users. Waters is also working on automating the sample acquisition and preparation steps, which will greatly reduce sampling time and improve consistency.
“Long-term, LC/MS will be the best way to do PAT because it’s the best tool we have for getting at a protein’s quality,” he adds.
Despite the trend towards instruments that anybody can use, PhD analytical chemists need not worry about their jobs, at least not yet anyway. “You still need their expertise when you need to fully understand a previously uncharacterized molecule,” Mazzeo explains. “But once you’ve established what you need to be looking for, the analyses become routine.”
Replacing immune-based analysis
Part of the allure of LC/MS in process analytics is its ability to replace wet chemical and immunoassay methods for quantifying and characterizing protein products and impurities.
Bioprocessors monitor host-cell proteins (HCPs) through enzyme-linked immunosorbent assay (ELISA), an old technique that lacks the specificity and coverage to identify and quantify individual HCPs. Analysts must first identify HCPs, characterize them, and raise antibodies to them for use in the ELISA.
Lisa Sapp, biopharma segment market manager at Agilent Technologies, notes that “any difference in culture conditions, for example nutrient supply, pH, temperature, etc., can change the composition and quantity of the HCPs. Because of this high variability, bioprocessors must check HCP levels after each adjustment and purification step. In addition, some HCPs may not have sufficient affinity to the antigen presented, so the HCP might not be detectable at all by ELISA.”
HPLC seems like a logical replacement for immunoassays, since LC systems are already optimized for protein purification. Both ELISA and mass spectrometry-based methods can be automated to increase walk-away time and sample throughput. And although the upfront cost of the LC/MS system is high, reagent costs are low because no antibodies are involved.
However, on LC, HCPs often coelute with more highly abundant drug products or derivatives. Additionally, since regulations mandate that impurities be measured at each production step, companies who find acceptable levels of HCPs at one juncture do not receive a pass for subsequent steps.
Agilent’s approach to HCPs is based on superior separation of HCP peptides, and the broad dynamic range of their LC/Q-TOF system. In a recent Agilent study, investigators used the AssayMAP Bravo platform for automated sample preparation, which included protein digestion, desalting, and on-cartridge high-pH reversed-phase fractionation to clean up and purify the samples. AssayMAP Bravo uses micro chromatography cartridges and task-centric automation protocols for reproducibility, scalability, flexibility, and ease-of-use in sample preparation.
For the assays, samples are then subjected to high-resolution LC/Q-TOF for analysis. The Agilent 6545XT AdvanceBio Q-TOF system provides iterative MS/MS, which improves identification of low-abundance precursors. Iterative LC-MS/MS analyses allow for precursors that were previously selected for MS/MS fragmentation to be automatically excluded on a rolling basis, with customizable mass error tolerance and retention time exclusion tolerance.
Modern manufacturing relies on analytics to provide real-time control and understanding of complex production systems, but biomanufacturing is decades behind foods, fuels, and other highly-regulated industries in this critical area. Process analytic tools exist, but their implementation for at-line or in-line monitoring has been painfully slow.
Liquid chromatography stands out as a natural modality for PAT, particularly for its integration with mass spectrometry. Vendors are eager to work with drug sponsors on HPLC-based PAT but it takes two to tango, and the dance becomes even less likely in an era of fanatical risk aversion.
