Automation and control of an integrated continuous downstream bioprocess for the purification of monoclonal antibodies

4

Introduction
Improved patient access through cost reduction is motivating the implementation of process intensification. Beyond pure cost incentives, the desire for real-time drug release (1-3), improved product quality (2, 3), minimized waste and facility footprint (4, 5), increased manufacturing flexibility (5), and decreased environmental impact (6) drive the industry to adopt automated continuous processing (7).

 

Legacy mAb processes have been facilitated predominately through fed-batch bioreactor operations coupled with batch purification unit operations. In recent years, key integrated continuous bioprocessing (ICB) enabling technologies such as perfusion cell culture, multi column chromatography (MCC), and single-pass tangential flow filtration (SPTFF) have driven the industry towards more modern manufacturing capabilities. The earliest published ICB implementation with CHO cell perfusion cell culture was demonstrated with four column periodic counter-current chromatography (PCC) for continuous capture (8).

In the subsequent decade, a variety of end-to-end lab scale processes have been accomplished with some pilot scale demonstrations following suit (9-14).

 

In a traditional batch process, individual equipment is automated, performs its function, and passes its processed product to a hold tank. The subsequent unit operation is performed without much (if any) interaction or interconnectivity with the preceding or ensuing unit operations in the process sequence. When deploying an ICB approach however, each of these unit operations forms an interconnected and interdependent process. The interconnection challenges presented by ICB are numerous (15, 16). Some prevailing questions revolve around a) how can inconsistent outputs of a unit operation be handled, such as with bind and elute chromatography or during process disturbances b) should minimal holdup volume be preserved or should surge vessels be implemented and c) how can unit operations be managed to prevent product holds in unfavorable conditions and maintain continuous processing or divert as needed?

 

The downstream platform discussed in this work begins with periodic counter-current (PCC) Protein A capture, which can be implemented to account for the variability in product titer and impurity profile of a perfusion feed stream, as well as maximizing resin capacity by loading two columns in series (17, 18). For all other downstream unit operations, alternating (swing-system) approaches can be employed. This swing system approach facilitates a continuous flow, but otherwise operates in a similar way to current batch processes and should not present a new regulatory burden (19). Dual tank virus inactivation (VI) follows Protein A to inactivate enveloped viruses and acts as a product conditioning step for the subsequent polishing chromatography unit operations. Polishing chromatography follows VI; in this instance Mustang Q anion exchange chromatography in flowthrough mode removes HCP and DNA contaminants while bind and elute on Capto S ImpAct cation exchange resin reduces the aggregate population. Polishing chromatography leads into virus filtration and single-pass final formulation – inline concentration (ILC), inline diafiltration (ILDF) and sterile filtration.

Figure 1 provides a graphical overview of the order of operations in the described mAb purification process.

 

While there have been advancements and technological improvements to the unit operations themselves when operated in a continuous mode, a challenge of automated fluid management and integration between unit operations remains. Here we propose a novel modular universal valve approach to the automation and control of a downstream ICB platform process that has been successfully implemented at lab scale. Surge vessels and scales between unit operations enable uniform process outputs and drive fluid management through the valve orientation. Sufficiently mixed product is especially critical for bind and elute chromatography steps, as well as sampling and ensuring complete viral inactivation. The implementation of advanced, flexible control strategies structured around these surge vessels addresses fluid management challenges and facilitates continuous bioprocessing.

 

Control Strategy
To provide sufficient fluid management, consistent process performance, simplicity, and flexibility, a universal valving and pump approach associated with a surge vessel was developed. In principle, this approach addresses the main risks associated with this platform including mismatched flow rates of unit operations, startup, shutdown, steady state operation, and filter fouling. Utilizing weight-based control, a balance associated with each unit operation’s surge vessel is used to dictate both the valve configuration and pump speed to maintain steady state. Adjusting the weight and pump speed control setpoints based on the unit operation type and expected throughput then provides refined control over this deployment.

 

We have developed a universal hardware approach to manage the handoff of product between unit operations and facilitate continuous processing.

This universal hardware approach, consisting of a series of pumps and valves, manages flow for each unit operation based on process parameters including surge tank weight, pressure, conductivity, or pH. With this approach, valves are enumerated similarly across the unit operations and perform the same function, with the same control commands. Pumps are under a similar control scheme, and together with valves manage process fluid. An example of this strategy is shown below in Figure 2, in which each unit operation valve ending in 01 is always set to send fluid to the surge vessel, to waste, or quarantine. Valves ending in 02 provide the ability to maintain continuous operation by drawing from buffer instead of process fluid, which is particularly useful for operations such as chromatography, where pausing in process fluid or caustic solutions is to be avoided. Valves enumerated as 03 and 04 work in tandem to divert to a standby filter in the event of fouling, or divert to a hold tank if all consumables are inoperable. Likewise, valves ending in 05 and 06 manage the process output and allow the control strategy to send fluid to the subsequent unit operation or to waste if product is out of specification by pH or conductivity.

In Figure 2A, a valving diagram for a dead-end filtration unit operation (i.e. depth filtration, sterile filtration, and/or virus filtration) is shown. In this example, the dead-end filtration unit operation has an idle filter train in parallel. An additional layer of control for filter-based operation relies on pressure data before and after each filter. If the first filter train reaches a high-pressure setpoint, valves switch to direct fluid to a secondary idle filter train. The secondary filter train is a necessary strategy to manage filter fouling and provide the ability to continue processing while an operator can replace a fouled filter. In the event both membranes are fouled, depending on setpoint values chosen, fluid can either be directed to waste or quarantine, or the pumps can be paused to allow the operator to replace fouled filters and resume processing.

 

Building on this framework, the valving strategy can be adapted for an array of unit operations including chromatography, virus inactivation (VI), and single-pass tangential flow filtration (SPTFF). In Figure 2B, a valving diagram of the chromatographic configuration is shown. In this configuration, the valving in positions 1, 2, and 6 are maintained from the previous example and retain the same functionality. Here the chromatography system, in this case the ÄKTA pcc, can be programmed locally in its native UNICORN environment and the valving can be used to manage fluid surrounding the system. Furthermore, due to its communication protocols and tag availability once the control system is connected, further control over parameters such as flow rate, pause/start commands, or specific variables such as loading challenge can be established. Similar approaches can be taken with more autonomous equipment referred to as “black box” equipment.

 

More sophisticated valving schemes and additional pumps may be needed for unit operations such as SPTFF. These pumps are enumerated 1 through 4 and maintain their specific functionality dependent on the location in the unit operation similarly to the valving. Additional “new” valving is required here and is associated with control that is specific to these particular unit operations. In Figure 2C of the ILC we see the addition of one of these new valves “7” which is always associated with the permeate of the ILC unit. Building upon this further, the ILDF and ILC unit operations can be performed within the same unit such as may be desired for final formulation.

 

Weight-based setpoints

A weight-based approach is one of many control strategies that can be implemented. Balances and their associated surge vessels offer benefits to process robustness. Surge vessels can act to dampen any process variability or noise, particularly when considering a bind and elute step where the elution peak is not homogenous. This normalization of product concentration and other product quality parameters minimizes variability entering the subsequent unit operation.

 

Four surge vessel weight-based setpoints are implemented as the framework for the valving and pump process control, Figure 3. These setpoints are defined as the Low, Startup, Reset, and High and can be set independently for the valving and pump control. These standalone setpoints allow the layering of pump control and valve control. When a process is initiated, the system is in a waiting state, with all pumps paused and valves directing to waste. As process fluid fills the surge vessels, the Startup setpoint is reached, and steady state operation begins at the default (steady state) pump flowrate. If there are any deviations from steady state, the valve control and pump control work together to maintain the process. If the fluid level drops below the Startup setpoint, the flowrate is reduced to prevent complete emptying of the surge tank. If the surge tank level continues to fall below the Low setpoint, the pump turns off and the unit operation is paused. Similarly, if a surge tank begins to overfill and the High setpoint weight is passed, the flowrate will speed up in efforts to drain the tank and the valving will redirect the fluid to waste to prevent the surge vessel from overflowing. Once the surge tank level drops below the Reset setpoint, the valving configuration and the pump speed return to default (steady state) values. For processes where the flow rate is a critical process parameter, the customizable nature of these setpoints allow the operator to simplify the pump control scheme to an on/off state by setting the pump flowrate to zero when setpoints are exceeded.

 

This universal, flexible valving and pump approach allows layering of control so that all unit operations are managed appropriately and effectively. This is particularly impactful in continuous operation, where the goal is often to manage all unit operations with a single control system. A process layout imbedded with modularity and consistent structure simplifies implementation. This underlying strategy is also valid across a wide range of process scales, limited only by hardware choices. We demonstrate effective control of processes between 6.5L to 81L of harvest material, processed through downstream unit operations.

 

Platform Integration Requirements
The global and international nature of the biotechnology industry presents complications, with the availability and preferences of different regions for a variety of programmable logic controllers (PLCs) (20). Each PLC is sold with a specific set of communications protocols it is compatible with. If the other equipment in the lab is not able to send data over those communication protocols, middleware software is needed which contributes to additional cost and complexity. When these communications are implemented, tags associated with process data are required, yet could be limited by a particular software platform. These limitations could be due to constraints like cost, maximum tag count, or processing capability. More traditional 0 – 20 mA or 4 – 20 mA analog signals are generally compatible with any system when appropriate scaling of the signal is performed (21).

 

A supervisory control and data acquisition (SCADA) system or a distributed control system (DCS) can be used to control the equipment and execute commands in a specified manner. DCS’s are most often used in manufacturing environments where the need for flexibility is less critical. A SCADA platform was selected primarily for flexibility in terms of hardware components and communication protocols. The SCADA approach was also more cost effective for the scope of this project.

An Allen-Bradley PLC paired with an Ignition SCADA platform was used in this case. The PLC software contains the majority of the control strategy while the SCADA software contains the screen layout and functionality, as well as connectivity to devices for monitoring and historization. For some equipment using older communication protocols middleware software was used to convert one communication protocol to a compatible one. An AVEVA PI System (formerly OSIsoft PI System) is included in the network to record and historize the process data. This platform resides on a virtual machine, which can be duplicated and easily scaled out when additional hardware is available.

Modular remote I/O “pods”, that physically link hardware to the PLC, are deployed and can be moved with or without associated hardware around the suite, Figure 4. Choosing I/O deployments that are not hard-wired into the bioprocessing suite facilitates a variety of process sequences, or potentially other therapeutic modalities in the future. Sensors or pumps from a unit operation configuration can be relocated to another unit within the suite to avoid moving entire units. The software is also structured to allow the user to change the unit configuration by adjusting the available hardware on the pod and the control type. Unused valves or pumps can simply be plumbed out of the flow path and reinstated if the unit operation changes. This approach maximizes flexibility in the suite and can facilitate different process sequences and simplify hardware changes.

 

Once these nuances of the PLC selection, acceptable communication protocols, and the human-machine interface (HMI)/SCADA platform is decided, hardware can be selected and integrated into the control system.

 

Results
The full deployment of these solutions in the Westborough, Massachusetts Cytiva continuous downstream lab is depicted in Figure 5. The prototype valving manifold is situated in front of each unit operation on an independent valving rack. The HMI resides in the middle of the lab and displays each unit operation in its relative location throughout the lab. From this HMI, a lab user can control all the equipment present in the suite.

 

Four continuous mAb production processes were successfully completed in this downstream lab. The implementation of the described control scheme managed process fluid and maintained process flow throughout the duration of the runs, coordinating pump speeds to prevent tank drain and overflow, pausing pumps when filters hit pressure limits, and diverting out of spec fluid to waste. Table 1 summarizes run duration, productivity, and the critical quality attribute profile.

Conclusions
This flexible control strategy towards maintaining a continuous process successfully accommodates processes regardless of scale or duration. However, challenges exist around calculating mass balances, filter priming, startup, and shutdown. For small scale, short duration processes, closing a mass balance is particularly challenging due to large holdup volumes relative to the total processing volume. There are limitations as to how to determine the mass balance through only using the balances throughout the lab. This opens the opportunity to add in additional sensors, such as flow sensors, to accurately account for the processed fluid. Other functionalities, such as using tags for recording when a pump is on or off and the flow rate it is operating at, would also improve this platform. To start and complete the processes in a more controlled manner, startup and shutdown are also areas where additional phase creation can provide benefits.

 

Although there are many challenges related to full end-to-end automated control of a continuous bioprocess, this universal valve approach offers benefits and has been demonstrated to facilitate a platform process. Implementing a flexible universal valving approach allows the user the maximum adjustability for individual unit operations and facility configuration. In addition, this approach allows the interchangeability of process hardware and software between unit operations, with I/O pods, HMI units, and valving manifolds easily configurable to any type of operation. By structuring the software and hardware in this flexible manner and implementing similar control for all unit operations, a standardized approach can be applied, and interdependencies can be identified. Careful planning, equipment and software selection, and deployment of a suite as described can achieve the automation and control of an integrated continuous bioprocess.

Acknowledgements
The authors would like to recognize the special contributions of Harvey Branton and his group from the Centre for Process Innovation (CPI) in the UK.

 

Declarations
The authors declare no conflict of interest.

 

Figure 1. Downstream mAb purification unit operations.

 

Figure 2. Unit operation valving and pump process instrumentation diagrams.

 

Figure 3. Valving and pump control setpoint strategy.

 

Figure 4. Automation software architecture.

 

Figure 5. Cytiva Continuous Lab (Westborough, MA).

 

Table 1. Summarized process conditions and results of automated continuous processes.

 

References and notes

1. Holzberg TR, et al. Sensors for biomanufacturing process development: facilitating the shift from batch to continuous manufacturing. Curr Opin Chem Eng. 2018;22(October):115-127. doi:10.1016/j.coche.2018.09.008.
2. Jiang M, et al. Opportunities and challenges of real-time release testing in biopharmaceutical manufacturing. Biotechnol Bioeng. 2017;114(11):2445-2456. doi:10.1002/bit.26383.
3. Wasalathanthri DP, et al. Technology Outlook for Real Time Quality Attribute and Process Parameter Monitoring in Biopharmaceutical Development – A Review. Biotechnol Bioeng. 2020; 1-17. doi:10.1002/bit.27461.
4. Xu J, et al. Biomanufacturing evolution from conventional to intensified processes for productivity improvement: a case study. MAbs. 2020;12(1). doi:10.1080/19 420862.2020.1770669.
5. Hong MS, et al. Challenges and opportunities in biopharmaceutical manufacturing control. Comput Chem Eng. 2018;110:106-114. doi:10.1016/j.compchemeng.2017.12.007.
6. Stankiewicz AI, Moulijn JA. Process intensification: transforming chemical engineering. Chem Eng Prog. 2000;96:22–33.
7. Fisher AC , et al. The current scientific and regulatory landscape in advancing integrated continuous biopharmaceutical manufacturing. Trends in Biotechnology. 2019;37:253–267. doi:10.1016/j.tibtech.2018.08.008.
8. Warikoo V, et al. Integrated continuous production of recombinant therapeutic proteins. Biotechnology and Bioengineering. 2012;109:3018–3029. doi:10.1002/bit.24584.
9. Brower M, et al. Monoclonal antibody continuous processing enabled by single use. Continuous Processing in Pharmaceutical Manufacturing 2014;255–296. doi:10.1002/9783527673681.ch11.
10. Coffman J, et al. A common framework for integrated and continuous biomanufacturing. Biotechnology and Bioengineering. 2021;118:1735–1749. doi:10.1002/bit.27690.
11. Coffman J, et al. The design basis for the integrated and continuous biomanufacturing framework. Biotechnology Bioengineering, 2021;118:3323–3333. doi:10.1002/bit.27697
12. Utturkar A, et al. A Direct Approach for Process Development Using Single Column Experiments Results in Predictable Streamlined Multi-Column Chromatography Bioprocesses. Biotechnol J. 2019;14(4). doi:10.1002/biot.201800243.
13. Schwarz H, et al. Integrated continuous biomanufacturing on pilot scale for acid-sensitive monoclonal antibodies. Biotechnol. Bioeng. 2022;119(8):2152-2166. doi:10.1002/bit.28120.
14. Coolbaugh M J, et al. Pilot-scale demonstration of an end-to-end integrated and continuous biomanufacturing process. Biotechnology and Bioengineering, 2021;118:3287–3301. doi:10.1002/bit.27670.
15. Kumar A, et al. Why Is Batch Processing Still Dominating the Biologics Landscape? Towards an Integrated Continuous Bioprocessing ternative Processes. 2020;8(12):1641. doi:10.3390/pr8121641.
16. Zydney AL. Continuous downstream processing for high value biological products: A Review. Biotechnol Bioeng. 2015;113(3):465-475. doi:10.1002/bit.25695.
17. Gomis-Fons J, et al. Optimization study on periodic counter-current chromatography integrated in a monoclonal antibody downstream process, Journal of Chromatography A. 2020;1621. doi:10.1016/j.chroma.2020.461055.
18. Hans Blom and Helena Skoglar, “The use of dynamic control in periodic counter current chromatography” in “Integrated Continuous Biomanufacturing II”, Chetan Goudar, Amgen Inc. Suzanne Farid, University College London Christopher Hwang, Genzyme-Sanofi Karol Lacki, Novo Nordisk Eds, ECI Symposium Series, (2015). http://dc.engconfintl.org/biomanufact_ii/83
19. Thakur G, et al. Control of surge tanks for continuous manufacturing of monoclonal antibodies. Biotechnol Bioeng. 2021;118(5):1913-1931. doi:10.1002/bit.27706.
20. Manser B, Bisschops M. Automation and Control in Continuous Bioprocesses Pall Biotech Technical Regulatory Topic. 2021
21. Dhanabalan G, Tamil Selvi S. Design of parallel conversion multichannel analog to digital converter for scan time reduction of programmable logic controller using FPGA, Computer Standards & Interfaces. 2015;39:12-21. doi:10.1016/j.csi.2014.12.002.