Introduction

Hollow fiber-based tangential flow filtration has emerged as one of the most preferred cell retention technologies for mammalian perfusion processes with applications in main stage perfusion bioreactors and also in N-1 bioreactors improving existing fed-batch production units (Bielser et al., 2018; Coffman, Brower, Connell‐Crowley, et al., 2021; Wolf et al., 2020). Despite significant advancements in pump technology, such as low-shear diaphragm pumps or levitated centrifugal pumps (Blaschczok et al., 2013; Clincke et al., 2013; Kelly et al., 2014; Wang et al., 2017a), to improve culture viability and thereby reduce the load of fouling-provoking particles on the filter membrane, filter clogging and product retention remain major challenges on the way to robust manufacturing processes.
Tangential flow filtration (TFF) in unidirectional crossflow mode (frequently driven by a levitated centrifugal pump) and alternating tangential flow (ATF, driven by a diaphragm pump) are the most commonly reported systems in perfusion processes (Fisher et al., 2019; MacDonald et al., 2022; Matanguihan & Wu, 2022). Most studies revealed that ATF showed superior product sieving compared to TFF at lab-scale and suggested ATF as a more suitable technology for long-term perfusion operation (Clincke et al., 2013; Karst et al., 2016; Wang et al., 2017a). However, ATF systems driven by diaphragm pumps were associated with operational instability at manufacturing scale (Coffman, Brower, Connell-Crowley, et al., 2021; Pavlik, 2017, 2019; Shevitz, 2018). Furthermore, multiple parallel ATF system were required to operate 2000 L perfusion bioreactors, requiring considerably more floor space compared to similar TFF systems (Coffman, Brower, Connell‐Crowley, et al., 2021; Romann et al., 2023). Therefore, TFF systems were claimed to be the preferred choice within the industry at large scale due to smaller facility footprint and higher robustness, whereas ATF systems showed improved product sieving and reduced development time for pilot-scale operations (Coffman, Brower, Connell-Crowley, et al., 2021).
In TFF, concentration polarization and fouling can both affect product retention (Belfort et al., 1994; Chew et al., 2020; Field, 2010; Taddei et al., 1990; van Reis & Zydney, 2007). Compared to industrial TFF applications that have short operating times and high filtrate flux (Redkar & Davis, 1993; Tanaka et al., 1997; Weinberger & Kulozik, 2021b), TFF systems used as cell retention devices in perfusion processes must be operational for as much as several months (without cleaning) and are operated at comparably very low filtrate fluxes of around 2 L/m2/h (Radoniqi et al., 2018). Due to the low filtrate fluxes and high axial pressure drops, a reverse flow of filtrate back into the fiber lumen occurs at the filter exit. This so-called Starling recirculation (Starling, 1896) was modelled and shown to be significantly larger than the actual harvest rate during typical perfusion processes (Radoniqi et al., 2018). As a consequence, only slightly more than 50% of the actual hollow fiber membrane surface area is used for filtration (Radoniqi et al., 2018).
More recently, alternating TFF systems have been described either with one levitated centrifugal pump and valves to switch crossflow direction (Weinberger & Kulozik, 2021a), or with two alternating, inversely positioned centrifugal pumps in the retentate loop called reverse TFF (rTFF) (Pappenreiter et al., 2023; Weinberger & Kulozik, 2022). Both setups showed reduced product retention compared to unidirectional TFF systems. While each individual phase of alternating crossflow filtration (ATF, rTFF or alternating crossflow by valve switching) can be compared to the situation in a TFF system, the distinguishing factor lies in the alternating direction of the crossflow and therefore the change in the location of the filter inlet. When working with cell lines prone to aggregation, switching the filter inlet can prevent fiber blocking (Weinberger & Kulozik, 2021a; Zydney, 2016). Improved product sieving with alternating crossflow systems compared to TFF was further attributed to the short period of zero net flow between phases resulting in a very low TMP across the entire filter length, possibly leading to deposit layer relaxation (Weinberger & Kulozik, 2021a, 2021b). It was also suggested that the Starling recirculation, which switches between the two ends of the hollow fiber, could remove deposited material and thereby minimize fouling (Karst et al., 2016; Radoniqi et al., 2018). Additionally, alternating crossflow filtration makes use of the entire filter length, harnessing the full membrane surface of the module (Radoniqi et al., 2018). The increased performance of alternating TFF is likely due to a combination of these factors.
Although there are several advantages associated with alternating crossflow filtration, it is important to note that the backflush of filtrate at the filter exit must be counterbalanced by an increased filtrate flux near the filter inlet to maintain the same overall level of filtration. This causes an increase in the drag forces that push particles into the membrane (Ripperger & Altmann, 2002), which can lead to a denser deposit and greater particle penetration into the membrane, both detectable as an increase of irreversible fouling resistance (Weinberger & Kulozik, 2022). Sundar et al. (2023) demonstrated that the greatest fouling in ATF systems occurs at the ends of the hollow fiber module, i.e., where the filtrate flux is greatest. This might explain studies observing significant product sieving losses despite using properly sized ATF modules (Kim et al., 2016; Wang et al., 2017b).
These phenomena provide notable constraints in hollow fiber designs and determining optimal operating conditions for perfusion systems. Crossflow velocities, for example, must be kept low to decrease the pressure drop along the filter length and thereby reduce the Starling recirculation. However, reducing crossflow increases the residence time of the cells within the recirculation loop, risking oxygen depletion (Walther et al., 2019). In addition, low crossflow leads to greater concentration polarization, i.e., greater accumulation of cells at the membrane surface. The increasing axial pressure drop with increasing length of the hollow fiber modules favors the use of relatively short filtration modules, requiring multiple filters in parallel to meet the needs for larger filtration area. Increasing the inner diameter of the hollow fibers would also reduce the pressure drop along the fiber length, but at a cost of much greater hold-up volume within the module.
Although all these strategies to improve filtration performance try to reduce the impact of the pressure drop and Starling flow, none of them solve the fundamental problem that the local TMP is coupled to the magnitude of the crossflow (which determines the axial pressure drop) and the filter characteristics. In order to achieve a nearly uniform TMP throughout the module, a similar pressure drop must be generated on the filtrate side of the module as on the retentate side. This concept for the biopharmaceutical industry was originally called High Performance TFF (HPTFF) and was successfully demonstrated to control concentration polarization along the filter length to enable high resolution protein-protein separations in downstream operations (van Reis, 1993; van Reis et al., 1997) . It has also improved purification of viral vectors using ultrafiltration (Grzenia et al., 2008). The concept is further known in the dairy industry for microfiltration (Merin & Daufin, 1990; Sandblom, 1978; Vadi & Rizvi, 2001). The ability to control the filtrate flux and Starling flow independently of the crossflow and length of the filtration module offers a promising tool to overcome challenges in current TFF and ATF systems in perfusion processes. To the authors knowledge, HPTFF to alleviate product retention in perfusion processes has not been evaluated in the literature.
The aim of this study was to develop a co-current filtration flow system for perfusion processes based on single-use low-shear centrifugal pumps in combination with pressure sensors with the ultimate goal to reduce product retention. This so-called HPTFF system was characterized for a wide range of operating conditions by measuring the TMP along the length of the filter module. Steady-state perfusion cell culture runs demonstrated superior performance during HPTFF compared to standard TFF. Further, a new concept of stepping co-current TFF (scTFF) was introduced, allowing us to generate alternating Starling recirculation within a unidirectional TFF system. Subsequently, scalability of HPTFF and scTFF were successfully evaluated by recording pressure profiles along a modified large-scale hollow fiber module.