Today's renewed interest in perfusion culture is due to an increased awareness of its advantages, some general improvement in equipment reliability, and a broadening of operational skills in the biomanufacturing industry. Some misperceptions persist, however, according to a 2011 review by Eric Langer (1). Our view here of the history of perfusion and fed-batch processes includes some discussion of technological process improvements and challenges that the bioprocess industry faces.

A team of authors at Serono in Switzerland wrote in 2003:

The major advantage of the perfusion mode is high cell number and high productivity in a relatively small-size bioreactor as compared with batch/fed-batch. In order to sustain high cell number and productivity, there are needs to feed medium during the cell propagation phase and the production phase. In contrast to batch and fed-batch processes, where there is no metabolites removal, in continuous processes medium is perfused at dilution rates exceeding the cellular growth rate. For this, a good separation device is needed to retain cells in the bioreactor. (2)

Many cell retention devices perform well, to a greater or lesser degree, at small scale, including gravity-based cell settlers, spin filters, centrifuges, cross-flow filters, alternating tangential-flow filters, vortex-flow filters, acoustic settlers (sonoperfusion), and hydrocyclones. All are described well in the 2003 paper mentioned above. But only a few types are reliable at larger scales and scalable enough for bioindustrial use.

Perfusion performed at CMC Biologics (Click to enlarge)

Here I compare the ATF System from Refine Technology with spin filters, cell settlers, and centrifuges. I am not including other technologies here because of scalability limitations and a lack of proven market acceptance.

Perfusion's Early Potential

The advantages of using perfusion for enhancing production of cell-derived products were realized in the late 1980s and early 1990s. In those early days of the modern biotechnology industry, production cell lines were not fully developed, and their product expression was very small — from a few micrograms to a few hundred milligrams per liter in batch or fed-batch. Attainable cell concentrations were only a few million per milliliter.

Spin Filters: Perfusion offered a way to derive more product from such low producers. It was well known that perfusion could increase cell concentration by as much as an order of magnitude (3). The spin filter was the most common perfusion device used; it was the best cell-separating device available at the time, supported by reputable equipment manufacturers.

Spin filters remain in use at a few sites but have been largely phased out, largely because of their limited scale-up potential and unreliability: When a bioreactor's volume scales up by the cube of its radius, the surface area of its spin filter screen scales by the square of its radius. An internal spin filter can take up a significant portion of production space within a vessel, and once its screen fouls, the run is terminated. An external production spin filter may solve this shortcoming, but it has drawbacks related to cost, maintenance and sterilization difficulties.

A more important factor behind the lackluster acceptance of perfusion in those early years was the rapid evolution of cell biology. New, more productive expression systems and improved media development permitted large increases in culture productivity; product concentrations were increasing from several hundred milligrams to about a gram per liter. Production needs could, therefore, be achieved with the well-understood fermentation technologies, batch and fed-batch. Scale up was accomplished simply by moving to bigger vessels.

The success of batch and, more important, fed-batch, not only inhibited the wider acceptance of spin filters, but also of other evolving cell-separation technologies. The difficulties associated with spin-filter operations and the undeveloped state of new perfusion technologies stigmatized the process. The dominance of fed-batch continued well into the next decade.

However, despite the dominance of fed-batch as an industry standard, perfusion continued to be championed. Perfusion offered an excellent solution for production with unstable proteins that could not remain in the toxic environment of an ever-deteriorating fed-batch culture. With perfusion, such products could be removed rapidly from a vessel and stored appropriately to preserve their stability. Many people chose perfusion to bypass constraints of space and cost factors. Furthermore, as culture productivity increased, and although it greatly benefited fed-batch processes, perfusion promised even greater output from a continuous culture.

So the use of perfusion never died; in fact, as the use of spin-filters declined, other cell separation devices slowly emerged. Those were based on filtration, gravity settling, and centrifugation. Continued development of numerous products that held out the promise of commercialization provided the driving force to experiment with new culture technologies. Occasionally a perfusion process, primarily one based on using spin-filters, cell settling, and centrifugation, was scaled to commercial production.

High-Concentrations Are a Game-Changer: From the early 2000s and particularly in the past few years another critical transition in biopharmaceutical manufacturing occurred. Further advancements in development of cell lines, expression systems, and media formulations resulted in an impressive ability to grow cells to very high concentrations and achieve product concentrations previously inconceivable. Using fed-batch as a reference, in the mid 1990s attainable cell concentrations were about 5 × 10⁶ cells/mL, with record product concentrations of 1–2 g/L; today those are greater than 15 × 10⁶ cells/mL, with product concentrations of up to 10 g/L. Although those concentrations are still not typical, they indicate where the field is heading. Those results are amplified by the use of perfusion, through which substantially higher cell concentrations and product output can be achieved (4,5).