Introduction
CPI Biologics, in collaboration with Cytiva, SCIEX, BiologIC Technologies, and Biopharm Services, have developed a bench-scale demonstration of continuous processing that operates from the perfusion bioreactor through the downstream processing steps, to finish with a formulated high-concentration mAb product.1 The potential for this development system to be adapted to GMP readiness has been analyzed and gaps identified. This report summarizes the gap analysis in the hope of provoking wider discussion in the industry, leading to further developments to capitalize on the potential of intensified manufacturing technologies and improve the availability of biotherapeutics to patients.
CPI Continuous Processing Lab
The continuous processing lab at CPI consists of a series of interconnected unit operations. These are, in order, perfusion bioreactor, alternating tangential flow filtration (ATF) for cell retention, Protein A capture operated via periodic counter-current chromatography, low pH virus inactivation, depth and sterile filtration, flow-through anion exchange chromatography (AEX), and bind and elute cation exchange chromatography (CEX) polishing steps (both operated in alternating mode without an interconnection of columns), virus filtration, inline diafiltration, and inline concentration.
Biologic Technologies has developed a novel mass balancer and mass router to manage the flow between unit operations. The mass router enables the switching of input and output feed streams for priming and cleaning of flow paths, as well as the diversion of material to waste or to quarantine material. The mass balancer enables a direct connection between unit operations with minimal holdup volume, resulting in the ideal scenario of near “one-piece flow”. Normally this is a challenge, as unit operations are driven by their own system pumps. Slight differences in the flow rates when directing flow from one pump to another can cause pressure build-up if the first pump is operating faster than the second. Alternatively, if the second pump is operating at a faster flow rate than the first, low pressure can induce cavitation. To overcome this, the mass balancer can operate “on the fly” to instantaneously add a buffer if the downstream pump is demanding a higher flow rate than the upstream pump. Where the downstream pump is operating more slowly than the upstream pump, a fraction of the flow can be diverted to waste to balance the volumetric throughput of the two pumps.
Residence Time Distribution
The mass balancer adds a degree of complexity to the process but also simplifies the challenge of monitoring product residence time distribution. Residence time distribution is not an issue in batch processing, where the product is processed through one operation and captured in a single tank before the next unit operation begins. The material in the hold tanks between unit operations is considered homogenous. However, in continuous processing, the product flows directly to the next unit operation before the previous one has processed all the material.
To accommodate this, a surge tank is commonly employed between unit operations. Where there is a constant flow into and out of the surge tank, there can be considerable residence distribution, and a small proportion of products may reside in the tank for surprisingly long periods. Therefore, residence time distribution is required for tracking “batches” and “lots” of consumables used in the process along with deviation tracking and quarantining of out-of-spec products that may have already moved downstream through the continuous operations.
It is then a major advantage to move to straight-through processing without the hold-up of a surge tank. Decreasing the residence time distribution can also enable process development through perturbation studies to define an operational design space, as is done for small molecule/oral-solid dose continuous processing.2 Additionally, developing the operational design space on the same system where the clinical materials are manufactured avoids the cost of multiple systems and the need for tech transfer and scale-up.
Bioburden Control
Perhaps the biggest opportunity in the current benchtop-scale continuous process is in developing and ensuring a functionally closed flow path that maintains low bioburden for the duration of a perfusion process. The challenges can be arranged into a series of sub-categories:
Tubing and connections
Currently, CytoFLEX™ tubing is used widely throughout the CPI laboratory. There are many pieces of tubing built into manifolds and then sanitized with NaOH once in place. Developing pre-sterilized single-use tubing manifolds would decrease bioburden risk and reduce the number of aseptic connections required - a big step toward a functionally closed process.
Additionally, the majority of tubing connections are made by Luer lock. To move to a closed process, these would have to be replaced with sterile connectors. Again, building single-use manifolds would significantly reduce the number of connections required.
Sampling
Taking samples in a continuous process can be challenging for many reasons. Most obviously, a downstream purification may operate for 20 days or more, which means samples may need to be collected at inconvenient times, including through the night. Additionally, there is an increased risk of sampling continuous processes due to the greater number of cycles and the necessity to take more samples.
A possible solution is sterile automated sampling. However, the commercially available systems add significant cost and are designed to operate at the pilot scale and above. The minimum volume such systems typically sample is 10 ml, which may be better suited to scaled-up operation than to bench-scale purification.
Systems
Systems involved in bench-scale continuous processing do not typically contain single-use flow paths. The ÄKTA pcc™ chromatography system used in the CPI lab has a “hard-wired” flow path comprising polyetheretherketone (PEEK) tubing, piston pumps, and rotary valves. These may not be considered traditional or ideal GMP-ready solutions. However, there remains the possibility of using the system in a GMP scenario by dedicating the flow path to a single drug product. The flow path can be sanitized with 1N NaOH to minimize bioburden between processes with the same product. Switching to another product requires a complete replacement of the product-contact flow path, including valves, pumps, tubing, and sensors. With this caveat that the system would have to be re-plumbed for a new product, the ÄKTA pcc system could be conceived of as GMP-ready.
Columns
Chromatographic media are not widely available in pre-sterilized column format. It is more usual for packed columns to be sanitized with NaOH at a molarity that the medium will withstand. For most columns and for some modern Protein A resins such as MabSelect PrismA™ resin, this is 1N NaOH. Sanitizing in place with NaOH is not as desirable, as it can be difficult to destroy microorganism spores that can be highly resistant to NaOH. Additionally, during sanitization, the downstream flow path may be exposed to bioburden. For these reasons, pre-sterilization of columns is an active area of research.3 However, for batch processes, the risks of using chromatography with sanitization (rather than pre-sterilization) have been well managed over the years, with some suppliers moving to highly controlled manufacturing environments.
Microbial testing
Despite efforts to create a completely closed process and minimize microbial contamination, testing is still required. Endotoxin testing is probably the least problematic of such release tests. Limulus amoebocyte lysate (LAL)-based methods are well-established and relatively fast. Alternative approaches are also under development.
Bioburden is more of an issue. Microorganism testing as described in USP chapter 61 is more prolonged.4 The main element of the chapter is a plate count assay that requires up to 14-day incubation times on agar plates. This is a major pain point for lot release. Additionally, it poses another problem, as contamination detection takes such a long time it offers no opportunity to correct a process if bioburden is beginning to appear.
Similarly to bioburden testing, viral contamination testing currently involves cell tissue culture with protracted incubation times. Thus, rapid detection methods are an active area of interest, and the lead alternative involves next-generation sequencing (NGS). While this approach is faster, it involves a significant amount of data analysis and poses substantial problems with respect to process validation for GMP.5 It remains to be seen how regulatory authorities will accept such data. Overall, microbial testing represents the main analytical issue to continuous manufacturing and releasing lots in a timely manner.
Cleanroom requirements
The proposed improvements to the flow path may enable the development of a fully closed manufacturing process, which in turn would eliminate the risk of adventitious agents entering the manufacturing environment. A potential positive impact of such a closed process would be the reduced requirements for a cleanroom classification, which would greatly reduce cost and facility complexity. Fully closed processes could potentially be operated in open ‘ballrooms’. Additionally, there would be no need to physically segregate the pre- and post-viral removing steps (e.g., the operations before and after virus filtration). However, fully defining the requirements and validation for a closed system remains a challenge.6
The location of the individual unit operations in an enclosure or secondary container could essentially create a secondary environment to house the closed flow path. Controlling that secondary environment would further reduce the risk of contamination and may provide the confidence to locate a continuous bioproduction capability outside a traditional cleanroom environment.
User Experience
The challenges for users of continuous systems go beyond the standard 21 CFR part 11 requirements: “Persons who use open systems to create, modify, maintain, or transmit electronic records shall employ procedures and controls designed to ensure the authenticity, integrity, and, as appropriate, the confidentiality of electronic records…”.7 This entails different login credentials and software abilities for different operators, supervisors, and process engineers that are standard for manufacturing processes. However, a continuous process poses new challenges that require further development of the software. Programming systems to operate in unison for days or weeks is more challenging. It also requires development at both the unit operation and process levels. Thus, the manufacturing equipment may be best viewed holistically as a single system. For operators to set up and perform processes, enhanced programmability may be a necessity to minimize errors and ensure compatibility of the unit operation outputs throughout the process. Additionally, a user-friendly graphical interface with real-time data monitoring will be required for operators to monitor the process and see if any part of it is starting to head out of specification.
Data Management
One of the advantages of continuous processing is the ability to reduce the size of equipment and consumables through rapid cycling. From a data perspective, rapid cycling can be a double-edged sword. On the positive side, having more process data enables earlier identification of when the system starts to deviate from control so that early intervention can be deployed to keep all products in specification.
On the negative side, operating for long periods creates a data-management problem. Sensors, for example, can take measurements with millisecond frequency. This would generate a vast amount of data after multiple weeks of continuous processing, which then must be stored, managed, and presented in the batch report. Reporting by exception — focusing on occasions where the actual data differ from the prediction — would be the most likely path forward. Luckily, the large amounts of data generated would enable a more complete statistical analysis. This is a key area where digital twins of the process could be highly impactful.
Another challenge with data management is tying data together from different systems and sources. Here it will be important to employ a data historian and be able to timestamp and reconnect data that is generated later through analytics and offline sampling. This challenge is exaggerated for continuous processing due to the longer processing times and increased amount of data generated.
Verification and Validation
System and process validation will also become more challenging, as it will have to encompass the duration of the process. Despite operation at a steady state, the full duration of a process would likely have to be validated to ensure there was no significant change over time, including genetic drift of the production cells. As enticing as the prospect is of being able to modulate operation times based on demand, it is possible that operation would be allowed only for the validated period.
Lot Size Selection
With continuous manufacturing, the extended period of operation leads to a risk-based decision about how to deal with lot release. One scenario is to pool all the products from a multi-day campaign as a single lot. This poses more risk, as a deviation or an out-of-specification assay may precipitate loss of the whole batch. As an alternative, to minimize risk, it may be desirable to have many small batches so that if the process goes out of specification, only those batches directly affected will be lost — although this comes with increased cost and resource use. The challenge is that lot release is onerous and can include up to 40 different measurements of product identity and quality. Thus, a risk-based decision needs to be made on the lot size. To move to smaller lot sizes there is a desire, if not a necessity, to implement process analytical technologies and move toward real-time lot release. If we look at learnings from continuous processing in the small-molecule drug world, we can see that micro batches and real-time release go hand-in-hand to intensify the process.
Real-Time Release Testing (RTRT)
A major challenge for continuous manufacturing is the implementation of analytics, not just for process monitoring and process control, but also for lot release testing. Real-time release testing (RTRT) is necessary to fully intensify a process. RTRT is defined as “an ability to evaluate and ensure the safety, efficacy, and quality of the final drug substance and/ or drug product based on in-process data with reduced end-product testing”.8 To demonstrate that the process is consistently operating within the validated space, controlling, and monitoring at the process level, there are three remaining enabling challenges.
The first is identifying sensors that can directly monitor key critical quality attributes and process performance attributes in line. Such direct process measurement is part of the Process Analytical Technology (PAT) system, which applies timely analysis to process control to ensure final product quality.
Second, if direct measurement is not possible, there is the option of soft sensors that measure these attributes indirectly by combining output from perhaps multiple sensors and using a software model to compute the value of the attribute. Whatever the approach is, the sensors will have to operate for the duration of the process. This could involve switch-out and replacement or, preferably, sensors that can remain in calibration throughout the process. It may be necessary to employ strategies for hardware health monitoring and error detection, along with fault-mitigation strategies. This could entail predictive maintenance and strategies to prolong hardware health.
The third is fully exploiting feedback process control, a key feature of PAT. It is unclear how regulators would view “on-the-fly” process modifications and how these would fit with a fixed and defined process.
Despite the recent advances in inline analytical technologies, it is not possible to eliminate at-line analytical methods. Such methods will require automation to provide timely data outputs. For such methods, sample handling is an important factor to ensure representative sampling and preferably a minimum of sample processing.
Facilities
The current automated integrated process equipment at CPI is spaced out over 7 m × 1.3 m with access to both sides of the equipment. There is a further 4 m2 for control and data analysis hardware and monitors, which would be located outside the process area.
The CPI lab is designed to handle the output of a 50L perfusion bioreactor. In a 30-day manufacturing scenario, the total media usage would be 2800 L, and the total volume of buffer is 4000 L, divided across 10 buffers. Depending on the frequency of make-up required, the total equipment space for media and buffers is 6-7 m2, with media and buffer held outside the process area, adding a further 4 m2.
It is usual for the equipment to be placed in a much larger area to facilitate the movement of materials and personnel; often ten times the equipment area. Automated, operator-free processing with occasional changeover of small consumable items and removal of drug substances and wastes can reduce the net process area to values closer to those of clinical facilities (typically 1/5th the space) or cell therapy facilities (1/5th -1/6th the space)9,10 – resulting in a 50 m2 process area and prep/ hold area of 25-50 m2 each for media and buffers.
Additional infrastructure for utilities, warehouse, HVAC, waste management, QC laboratory, and offices (QA, logistics, engineering) would add an estimated 1000 m2, but these spaces can be shared with and supply additional automated process units.
Further design work is needed to minimize the footprint of the automated system while retaining access to solution inputs, waste outputs, consumable changeover, product changeover, and maintenance. An assessment of the closed nature of the system flow path is also needed, including media supply, buffer supply, and sampling to guide the cleanroom classification required for the process areas.
Such a process would be analogous to the use of automated process units for cell therapy, where one therapy is manufactured by one machine, and multiple therapies are manufactured in one room, each within a closed system and with process times ranging from days to weeks depending on the expansion/perfusion time required.11
Opportunities for Small-Scale Continuous Processing
Smart process development
At present, process development for continuous processes is performed in batch mode with single operations at a time to generate a fundamental understanding of each unit operation in order. However, the mAb process is relatively well established and the unit operations and order are essentially fixed, which presents another opportunity: To perform process development directly using the continuous system. Process knowledge combined with continuous experiments could enable a rapid definition of not just each unit’s operational parameters, but also of the complete operation space and how changes upstream impact the process.
Process intensification for other modalities
The process intensification path for mAbs is becoming well understood, if not yet widely implemented. For other modalities that have emerged in the last decade, the manufacturing platforms are not as clear or fully defined. There will be opportunities to bring process intensification to these platforms to help reduce cost and make difficult-to-manufacture therapeutics accessible to more people.
The production of low-dose therapeutics, such as vaccines or gene therapies, would extend the opportunity of such small-scale manufacturing systems,11 particularly for in-country manufacturing.
Miniaturization for portable/regional manufacturing
This could most easily be described as developing the current lab into a small single-box solution. With the mass balancer and mass router used currently in the lab, we can envisage how microfluidics could be miniaturized into a single-box solution combining all the unit operations into an automated mAb-producing system in a box.
In addition to a reduced physical volume, future systems should also offer flexibility and reconfigurability to enable multiple therapeutic products to be created from the same common hardware. This should be – to the greatest extent possible – via common hardware with software-modified parameters.
While this could result in suboptimal biomanufacturing solutions for each therapy, the overall ability to rapidly reconfigure a regional biomanufacturing solution to address a local threat enables faster and more agile regional responses.
This has utility in the developed world and especially in parts of the developing world that lack biomanufacturing infrastructure.
Data management
Continuous processing offers the potential to capture orders of magnitude more process data, during development and during production. How can this data be organized, stored, and managed to get the most insights on critical process parameters, in real-time and as part of ongoing continuous improvement strategies or development of new products?
Many software tools are available. BioSolve Process software, used for the project discussed here, offers an overview of the process and its top-level inputs (recipes, workflow, operations, costs, labor, energy, etc.) and outputs (product throughput, facility metrics, cost of goods, capital intensity, etc). As a part of the project, this top-level architecture is being integrated with normal process development recipes, distributed control system (DCS) operational inputs, and process outputs, with the aim of enabling technology transfer from process development (PD) to clinical to commercial scale-down manufacturing.
As the second UK government-funded project at CPI is set to complete, plans are being made to extend the activities in the lab. Currently, the goal of a future project will be smart process development, perhaps with a focus that goes beyond mAb modalities to viral vectors and vaccine production.
References
- American Pharmaceutical Review. Integrating Continuous Technologies for Rapid Delivery of Cost-Effective Biotherapeutics to Patients. 2022 https://www. americanpharmaceuticalreview.com/featured-articles/589065-integrating-continuous-technologies-for-rapid-delivery-of-cost-effective-biotherapeutics-to-patients/
- Vanhoorne V, Vervaet C. Recent progress in continuous manufacturing of oral solid dosage forms. Int J Pharm. 2020;579:119194.
- Varner C, Patil R, Godawat R, Warikoo V, Konstantinov K, Brower KP. Gamma irradiating chromatography columns enable bioburden-free integrated continuous biomanufacturing. Biotechnol J. 2021;16(4):e2000298. doi:10.1002/biot.202000298
- USP 61 Microbiological examination of nonsterile products: Microbial enumeration tests, https://www.usp.org/sites/default/files/usp/document/harmonization/gen-method/ q05b_pf_ira_34_6_2008.pdf
- BioPhorum Group. Rapid detection of bacteria and viruses: justification, regulation, requirements, and technologies - how can industry achieve broad adoption? Published October 2019. https://www.biophorum.com/wp-content/uploads/bp_downloads/TRM-rapid-detection-of-bacteria-and-viruses-how-can-industry-achieve-broad-adoption[1]October-2019-1.pdf
- Jagschies G, Lindskog E, Łącki K, Galliher P, eds. Biopharmaceutical Processing: Development, Design, and Implementation of Manufacturing Processes. Elsevier. 2017.
- US Food and Drug Administration. Code of Federal Regulations Title 21, part 11: Electronic Records, Electronic Signatures.
- (ICH Q8[R2]) Pharmaceutical development - Scientific guideline https://www.ema. europa.eu/en/ich-q8-r2-pharmaceutical-development-scientific-guideline
- Kappeler SR, Nicholson G, Bohnenkamp HR, Bethke U. A plug-and-produce GMP plant for cell and gene therapy — part 1: a case study in modular facility design and deployment. Bioprocess International. 2022 (Sep 28).
- Ran T, Eichmüller SB, Schmidt P, Schlander M. Cost of decentralized CAR T-cell production in an academic nonprofit setting. Int J Cancer. 2020;147(12):3438-3445.
- Hamidi A, Yallop C, Drugmand J-C, Reniers A, Dubois S, Castillo J. A small footprint, integrated, and automated platform for viral production. GEN - Genetic Engineering and Biotechnology News. 2020 (Feb 1). https://www.genengnews.com/resources/tutorial/a-small-footprint-integrated-and-automated-platform-for-viral-production/
Author Details
Simon Hawdon,1 Sean Ruane,1 Liam Nattrass,1 Allan Watkinson,2 Matt McEwan,3 Nick Rollings,4 Rob Noel,5 Martin Glenz,6 John Welsh,6 Ruth de la Fuente Sanz,6 and Mark Schofield6
1 CPI2 Labcorp3 AMAT4 BiologIC5 Biopharm Services6 Cytiva
Publication Details
This article appeared in American Pharmaceutical Review:
Vol. 27, No. 2March 2024Pages: 8-13
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!