Materials and Methods

Lab-scale system setup for pressure characterization

An experimental setup capable of operating TFF, rTFF, HPTFF and scTFF was built to characterize the different filtration systems at lab-scale (Figure 1A ). Pressure transmitters (PREPS-N-038, PendoTECH, Princeton, USA) were installed in the retentate loop at the inlet (\(\text{PT}_{R1}\)) and outlet of the filter (\(\text{PT}_{R2}\)). Further pressure transmitters were positioned in the filtrate loop at the inlet (\(\text{PT}_{F1}\)) and outlet (\(\text{PT}_{F2}\)) of the module. A PES hollow fiber-module with a total module length of 70 cm, an effective fiber length of 65 cm, an inner fiber diameter of 1 mm, a pore size of 0.2 μm and a membrane surface area of 0.15 m2 was used (S06-P20U-10-N, Repligen, Waltham, Massachusetts, USA). For the pressure tests, the hollow fiber module was modified by drilling holes and gluing additional pressure transmitters (\(\text{PT}_{A1-5}\)) into the filtrate side to monitor the filtrate pressures along the filter length. Considering the total length of the hollow fiber module, the additional pressure transmitters (\(\text{PT}_{A1-5}\)) were place at 5.5 cm, 20 cm, 35 cm, 50 cm and 64.5 cm measured from the start of the module. Flow sensors (LFSC-i10X-001, Levitronix GmbH, Zurich, Switzerland) were installed in the retentate loop (\(\text{FS}_{R}\)) and in the filtrate loop (\(\text{FS}_{F}\)). Three levitated centrifugal pumps (PuraLev i30SU, Levitronix GmbH, Zurich, Switzerland) were integrated into the experimental setup. Two pumps were integrated into the retentate loop, of which one was directed towards the filter inlet (\(\text{CD}_{R1}\)) and the other towards the filter outlet (\(\text{CD}_{R2}\)). The third centrifugal pump (\(\text{CD}_{F}\)) was inserted into the filtrate loop directed towards the filter inlet side. A peristaltic pump (\(\text{PP}_{H}\)) was inserted into the harvest stream. Data was recorded by connecting 3 process control units (LCO-i100, Levitronix GmbH, Zurich, Switzerland).

Large-scale system setup for pressure characterization

For large-scale pressure characterization, a similar experimental setup as described for lab-scale experiments was built to achieve TFF, HPTFF and scTFF operation (Figure 7AB ). A PES hollow fiber-module with a total module length of 78 cm, an effective fiber length of 68 cm, an inner fiber diameter of 1 mm, a pore size of 0.2 μm and a membrane surface of 7.15 m2 was used (X06-P20U-10-N, Repligen, Waltham, Massachusetts, USA). Pressure transmitters (PREPS-N-1-1, PendoTECH, Princeton, USA) were installed in the retentate loop (\(\text{PT}_{R1}\) and \(\text{PT}_{R2}\)) and in the filtrate loop (\(\text{PT}_{F1}\) and \(\text{PT}_{F2}\)). To measure pressures as closely as possible to the inlet and outlet of the hollow fibers, holes were drilled to place pressure transmitters (PREPS-N-038, PendoTECH, Princeton, USA) into the adapter piece of the filter module (\(\text{PT}_{RC1}\) and \(\text{PT}_{RC2}\)). Further pressure transmitters (PREPS-N-038, PendoTECH, Princeton, USA) on the filtrate side were attached along the filter length at 15.5 cm, 27.5 cm, 38.5 cm, 50 cm and 62.5 cm measured from the start of the module (\(\text{PT}_{A1-5}\)). Additionally, pressure transmitters at the same filter length but in the back were attached (\(\text{PT}_{AB1-5}\)). Flow sensors (LFSC-i35X, Levitronix GmbH, Zurich, Switzerland) were installed in the retentate (\(\text{FS}_{R}\)) and filtrate loops (\(\text{FS}_{F}\)). One levitated centrifugal pump (PuraLev-2000SU, Levitronix GmbH, Zurich, Switzerland) was inserted into the retentate loop (\(\text{CD}_{R1}\)) and a second levitated centrifugal pump (PuraLev-600SU, Levitronix GmbH, Zurich, Switzerland) was placed in the filtrate loop (\(\text{CD}_{F}\)). A peristaltic pump (\(\text{PP}_{H}\)) was inserted into the harvest stream. Data recording was achieved by coupling 6 process control units (5x LCO-i100 and 1x LCO-600, Levitronix GmbH, Zurich, Switzerland).

Experimental procedure for pressure characterization

Pressure characterization studies were performed with water as a medium. To characterize the TFF system, only pump \(\text{CD}_{R1}\) was operational. To realize the HPTFF and scTFF systems, pump\(\text{CD}_{R1}\) in the retentate loop and pump \(\text{CD}_{F}\) in the filtrate loop were active simultaneously. For the rTFF systems, pump\(\text{CD}_{R1}\) was used for the first phase and pump\(\text{CD}_{R2}\) was used for the second phase with reversed crossflow. For all lab-scale setups, crossflow was kept at 650 mL/min, except for ramping experiments where the crossflow was ramped from 0 – 1500 mL/min. For crossflow ramping in HPTFF, a delta pressure PI control (P-term: 2 rpm/Δmbar; I-term: 5 rpm/Δmbar*sec) was used with a setpoint of 0 mbar, meaning that the pump speed of pump (\(\text{CD}_{F}\)) was controlled such that pressure \(\text{PT}_{F1}\) matched\(\text{PT}_{R1}\). For scTFF characterization, the co-current filtrate flow rate was set at 870 mL/min (scTFF phase 1) and 1890 mL/min (scTFF phase 2) to achieve a delta pressure control of ± 10 mbar. To characterize large-scale TFF operation, the crossflow was ramped from 0 – 45 L/min. For HPTFF pressure characterization, additional filtrate flow was applied by a delta pressure PI control (P-term: 2 rpm/Δmbar; I-term: 5 rpm/Δmbar*sec) such that \(\text{PT}_{F1}\) was kept 6 mbar higher than \(\text{PT}_{R1}\). Filtrate flow ramping to assess scTFF operation was achieved by maintaining a crossflow of 14.5 L/min and ramping the co-current filtrate flow from 0 – 70 L/min. To compare filtration conditions between filtration module scales, shear rates provided by the manufacturers lookup tool were considered (Repligen, Waltham, Massachusetts, USA).

Perfusion culture process

A proprietary CHO-K1 cell line producing a bispecific mAb was expanded in an incubator (Multitron, Infors HT, Bottmingen, Switzerland) for 21 days using a proprietary chemically defined perfusion platform medium and an on-demand proprietary feed (Merck KGaA, Corsier-sur-Vevey, Switzerland). Perfusion bioreactors (Labfors 5 Cell, Infors HT, Bottmingen, Switzerland) were inoculated at a seeding density of 0.6 x 106 viable cells/mL. Culture conditions were maintained at 36.5 °C with a dissolved oxygen setpoint at 50% (VisiFerm DO Arc, Hamilton, Bonaduz, Switzerland). The pH was controlled at 7.07 ± 0.17 (EasyFerm Plus Arc probe Hamilton, Bonaduz, Switzerland) by sparging CO2 and using a 1.1 M Na2CO3 solution. Bioreactors were operated at 2 L working volume. After an initial growth phase, an online capacitance probe (Incyte Arc, Hamilton, Bonaduz, Switzerland) was used to keep the viable cell volume (VCV) constant at 12%. Perfusion started on day 0 and was kept constant at 1.3 reactor volumes per day until the end of the process.

Cell retention devices in perfusion cell culture

Bioreactor harvests were gravimetrically controlled to maintain the bioreactor weight constant using the same hollow fiber module as describes in section 2.1. Cell retention devices were either operated as TFF (Figure 1B ), rTFF (Figure 1C ), HPTFF or scTFF systems (Figure 1D ). In all setups, the crossflow was generated by levitated centrifugal pumps (PuraLev i30SU, Levitronix GmbH, Zurich, Switzerland) to ensure comparability of sieving studies. Pump speeds in the retentate loop were set to 3500 rpm, initially corresponding to a crossflow velocity of 650 mL/min. For TFF, HPTFF and scTFF operations, the retentate loop pump (\(\text{CD}_{R1}\)) was stopped every 3 minutes for 3 seconds to release accumulated air bubbles within the centrifugal pump head. For HPTFF operations, a delta pressure PI control (P-term: 2 rpm/Δmbar; I-term: 5 rpm/Δmbar*sec) between pressure \(\text{PT}_{R1}\)and pressure \(\text{PT}_{F1}\) was used with a setpoint of 0 mbar to control the speed of the filtrate pump (\(\text{CD}_{F}\)). As such, filtrate pressures were matched with retentate pressures along the filter length. scTFF operation was achieved by delta pressure control of -5 mbar during scTFF phase 1 and +5 mbar during scTFF phase 2 between pressure \(\text{PT}_{R1}\) and pressure \(\text{PT}_{F1}\). rTFF operation consisted of two phases. In the first phase, the first retentate pump (\(\text{CD}_{R1}\)) was operational and the second retentate pump (\(\text{CD}_{R2}\)) was stopped. During the second phase, pump (\(\text{CD}_{R2}\)) was operational and the pump (\(\text{CD}_{R1}\)) was stopped. Phase times were 20 sec.

Reference analytics

Cell density, viability, cell diameter, and pH were measured using a BioProfile FLEX2 (Nova Biomedical, Waltham, USA). Bioreactors were automatically sampled by the FLEX2 On-Line Autosampler (Nova Biomedical, Waltham, USA) and samples were fractionated using a Teledyne Cetac ASX-7200 (Teledyne CETAC Technologies, Omaha, Nebraska, USA). Viable cell volume was calculated as follows (Metze et al., 2020):