Control Strategies for Perfusion Cell Culture

Published on: 
BioPharm International, BioPharm International-11-01-2020, Volume 33, Issue 11
Pages: 16–19

PAT advances are enabling improved process understanding, process control, and error prevention.

Interest in continuous bioprocessing continues to rise as more initial successes in the area prove out the potential benefits. In a 2020 survey conducted by BioPlan Associates, more than half of the 200 companies polled indicated they were planning to actively or informally evaluate continuous upstream technologies in the next 12 months (1). Not surprisingly, the global market for continuous bioprocesses is estimated to be expanding at a compound annual growth rate of 23% from 2020 through 2030 to reach a value of $310 million (2).

Upstream continuous bioprocessing involves perfusion. One of the benefits of such processes is the ability to operate at steady-state conditions for extended periods of time, resulting in more consistent and higher product quality. Maintaining a steady state, however, requires effective process monitoring and control.

Support from regulatory authorities for continuous bioprocessing and encouragement and guidance for improving process control and consistency have enabled the industry to drive advances in process analytical technology (PAT) and eased the path for implementation of novel solutions into perfusion processes, according to Atul Mohindra, senior director of research and development for Lonza. “PAT tools are becoming more accepted within the biotech industry, and their implementation is leading to novel control strategies that previously could not be realized,” adds Thaddeus Webster, lead scientist, research and development with Lonza. As a result, he notes that improved process control, error prevention, and understanding have become more and more of a reality each day.

Monitoring challenges to overcome

Effective monitoring of perfusion processes must be achieved before they can be controlled, but real-time collection of valuable data is not always a simple task. The main challenges include monitoring the accurate concentration of cells, the main nutrients and metabolites, and the relationship between growth and metabolic data using typical in-line, real-time monitoring of parameters such as pH, temperature, and dissolved oxygen, according to Claudia Berdugo-Davis, director of process development at Catalent Biologics. “The ability to develop an effective control strategy is impacted by the lack of understanding of the critical process parameters in perfusion processes,” she says.

On a more granular level, on-line monitoring of cell density is needed to achieve the cell-specific perfusion rate, according to Andreas Castan, strategic technology partnership leader with Cytiva. In addition, he notes that control of the feed, harvest, and bleed rates is important, and all of the parameters that characterize the filtration process like the flow rate and pressure of the feed, and the filtrate rate and pressure are important to design a robust and reproducible filtration process.

Particular attention should be paid to metabolite monitoring and viable cell volume, Webster adds. “While daily adjustments can be made to cell volume, doing so could lead to an oversimplification of the cell culture demands because the cellular metabolism of the culture can change in a dynamic environment,” he explains. Daily measurements of metabolite levels and adjustment of perfusion rates based on these results can lead to similar issues.

Emphasis on control of product quality is equally important, according to Mohindra. “Traditionally during fed-batch production, product attributes are measured at the end of a batch as part of quality-control release testing. However, given that perfusion processes could be designed to operate for more than 60 days, it is crucial to be able to monitor additional parameters in order to ensure consistent product quality. The challenge is to measure the correct parameters, do so accurately, and ensure that timely feedback loops are put in place,” he comments.

At the macro scale, Avril Vermunt, strategic technology partnership leader at Cytiva, observes that accounting for variability greatly impacts the complexity of the required downstream equipment and the control strategy. “Implementing improved monitoring and control of the perfusion process stream helps reduce the burden on downstream process design,” she states.

Common PAT tools

A number of common sensor technologies are used for the monitoring of perfusion processes. In addition to well-established methods for monitoring pH, temperature, and dissolved oxygen, biomass capacitance sensors for the monitoring of cell growth and various spectroscopy tools for understanding the relationship between biomass and some key metabolites have been developed, according to Berdugo-Davis.

Recently, adds Webster, inline viable cell count/viable cell volume (VCC/VCV) monitoring using capacitance probes has begun to be used for perfusion rate/bleed rate adjustment. “These PAT tools allow near-continuous feedback of the perfusion rate, which should lead to more consistent processing during perfusion,” he says. In addition to the wide use of capacitance-based sensors for viable cell density and gravimetric control of the feed, harvest, and bleed flows, Castan also highlights disposable pressure sensors for the assessment of transmembrane pressure.

Measurement of key nutrients is also being used for the monitoring and control of perfusion processes, according to Mohindra. “Superficially, the control of glucose and certain amino acids in a perfusion culture can enable a given process to be better controlled in terms of culture performance and product quality,” he observes.

Inline analysis and control, Webster stresses, remove the operator from the equation because no sterile barrier needs to be broken to remove a sample for analysis and control. “This approach reduces process complexity from an operator standpoint while potentially removing the need for daily calibration/maintenance of offline equipment,” he asserts. In fact, with a toolkit of available sensors and control strategies, well-designed tangential flow filtration-based perfusion processes can easily be developed, according to Castan.

Measurable monitoring advances


Advances in PAT tools have occurred on several fronts. One important area is the transfer of established sensors and control strategies from the stainless-steel world into single-use operations, which Castan believes have been a great enabler for modern perfusion processes.

Development of spectroscopic technologies for inline analysis is another. Many PAT tools already in use in bioprocessing today leverage spectroscopy, with Raman spectroscopy currently being considered state of the art in fed batch and presumably following soon in perfusion, according to Webster. “The ability to measure multiple process parameters via Raman should enable near-continuous perfusion and bleed rate adjustments that currently cannot be achieved using daily adjustments,” he remarks. Berdugo-Davis observes that the industry is currently looking expanding the scope even further to include online high-pressure liquid chromatography, near-infrared spectroscopy, and mass spectrometry. “These tools may help to gain knowledge from better understanding of the properties of cell functionalities in dynamic environments such as perfusion systems,” she says.

Yet other developments in PAT have been geared to reducing the need for offline sampling, according to Webster. “While offline analysis for viable cell concentration and metabolite analysis is critical to understanding the state of the bioreactor culture, it requires skilled operators to aseptically remove samples, calibrate equipment, perform analyses, and upload results. Replacing offline analysis with inline technologies reduces the potential for errors associated with offline analysis,” he comments.

For parameters that cannot be measured accurately via an inline probe or require processing prior to analysis (e.g., product quality attributes), Webster does point out that specific analytical methods have been developed with atline analysis in mind to take advantage of automated sampling systems that can pull on-demand samples for analysis. “These approaches aim to remove the need for operators to make manual interventions in the process while still collecting information about process changes,” he explains.

Another development is the rapid advances in data analytics capable of processing the large quantities of data generated by today’s PAT tools, according to Berdugo-Davis. “Such advances are needed in order to be able to quickly understand the relationships between multiple process parameters and apply that knowledge to consistently achieve optimum process performance,” she says.

Aided by improved sampling techniques

A sampling event is always a risk in the aseptic operation of a bioreactor. Therefore, says Castan, disposable aseptic sampling assemblies are widely used. Collaborative efforts are directed at the development of automated tools for sampling of high-throughput systems and bench-scale bioreactors, according to Berdugo-Davis. “Auto sampler systems integrated with PAT and control loops can help to elucidate critical process parameters for more effective control strategy definition,” she states.

Indeed, given the duration of perfusion processes, the improved robustness and reliability of aseptic sampling technologies have been crucial to ensuring “batch” success, Mohindra asserts. That includes any sampling that must occur in subsequent purification steps integrated with the perfusion process.

In particular, Webster points to advances in fluidic technologies as important for newer sampling technologies. For instance, he notes that Lonza’s MAST automated sampling system is capable of sending highly viscous samples greater than 80 feet to their final destination for analysis. “This ability enables atline sampling and analysis of several process parameters that have historically needed to be sampled and analyzed offline,” he says. When connected with automated sample prep systems, Webster adds that operator manipulation of samples can be eliminated, leading to improved robustness and analytical throughput.

Automation is an enabler

The implementation of PAT sensor strategies is the first step towards a robust and well-controlled process, according to Castan. “The next level,” he adds, “is the implementation of automation, of leveraging the sensors for feed-back or feed-forward control.” In this case, Castan notes it is a great advantage to have the perfusion skid with its automation connected as one unit to the bioreactor. Cytiva has introduced such a system that integrates with its perfusion bioreactors that provides automated filter switching, liquid management, and cell bleed.

Automation and PAT are, observes Vermunt, also evolving the industry toward an ideal future state where rather than using single-dimension ranges for multiple parameters to define operating spaces, operating spaces are fluid and dynamic and able to account for the interplay between multiple parameters and product attributes.

Issues still being tackled

Despite all of the advances to date that are enabling enhanced monitoring and control of perfusion processes, there are plenty of opportunities for further improvement. Vermunt would like to see non-specialized sensing technology that is accurate enough to control the direct contributors to yield, productivity, and/or quality without the need for development specific to each cell line and process.

Improvements in at-scale PAT solutions, particularly with respect to stability of inline sensors exposed to high cell densities, typical bioreactor disturbances, and extended culture durations are also needed, according to Berdugo-Davis. For example, Mohindra points to inline pH sensor technology, which still remains prone to drifting, thus potentially still requiring operators to extract samples and take offline measurements.

More sophisticated technologies for the measurement of traditional parameters such as dissolved oxygen, pH, temperature, glucose, and amino acids that provide more information about exactly what is happening within a cell are also needed, according to Mohindra, and will be a step change for the industry.

Vendor compatibility also still remains a challenge with respect to process automation, Mohindra adds. Aggregation of all perfusion-related data such as online viable cell density readings, flow rate measurements, and pressures into one repository together with other data from the bioreactor and on cell-culture performance is another issue for Castan. He believes that integration of the perfusion skid with the bioreactor is a significant step toward achieving that vision.


1. E.S. Langer, BioPlan 17th Annual Report and Survey of Biopharmaceutical Manufacturing Capacity and Production, (BioPlan Associates, April 2020).
2. Roots Analysis Research and Consulting Group, Continuous Manufacturing Market Research Report (Small Molecules and Biologics), 2020–2030, Report (March 2020).

About the Author

Cynthia A. Challener, PhD, is a contributing editor to BioPharm International.

Article Details

BioPharm International
Vol. 33, No. 11
November 2020
Pages: 16–19


When referring to this article, please cite it as C. Challener, “Control Strategies for Perfusion Cell Culture," BioPharm International, 33 (11) 2020.