3.3. Perfusion runs using XCellTM ATF-2 and inclined cell settler as cell retention devices
Based on the promising results of pseudoperfusion cultures obtained at spin tube scale, perfusion bioreactor cultivations were carried out. At first, a XCellTM ATF-2 was chosen as cell retention device, since it has been widely used not only for production of recombinant biopharmaceutical proteins, such as monoclonal antibodies (Clincke et al., 2013), but has also been reported in the recent years for virus (Coronel et al., 2019; Nikolay et al., 2018) and VLP production processes (Lavado-García et al., 2020; Alvim et al., 2019).
In our previous work on zika VLP production (Alvim et al., 2019), although a perfusion run could be stably operated for 30 days at approximately 30 million cells/mL, product retention inside the bioreactor due to membrane fouling was observed. However, considering a recent report (Nikolay et al., 2020b) that showed a polysulfone hollow fiber cartridge to be highly permeable for the yellow fever virus, we decided to use a cartridge of hollow fibers having the same characteristics as reported by these authors, with the aim of avoiding fouling.
Cell pool 1 was inoculated at 0.5 × 106 cells/mL and maximum viable cell density (VCD) was 28.8 × 106cells/mL on day 13, as measured offline by trypan blue exclusion in a cell counter. We also monitored VCD online using a capacitance-based biomass sensor. As shown in Figure 3A, the profiles of viable cell density determined using both methods were very similar, with the online biomass sensor showing a trend to be somewhat more sensitive. The use of this sensor can be very important in processes where the cell-specific perfusion rate is controlled, since the online biomass data can be used to automatically adjust the fresh medium feed rate (Dowd et al., 2003).
Cell viability remained above 90% until day 16 (Figure 3A), but then decreased because of nutrient depletion (Figure 3B). Until day 10, perfusion rate had been gradually increased along time in order to meet nutrient demands by the growing cell concentration. However, by the time viable cell density was approximately 25 million cells/mL, no increases in permeate withdrawal rate beyond 0.45 vvd were possible anymore (D_harvest, Figure 3B) due to progressive fouling of the membrane cartridge. Because the rate of permeate withdrawal was limiting, fresh culture medium addition rate (D_feed, Figure 3B) for 3 consecutive days exceeded the permeate and affected bioreactor working volume. Even so, from day 15 on glucose exhaustion in the bioreactor occurred, causing cell death. Another evidence that membrane fouling was progressively occurring was the fact that product retention inside the bioreactor started being observed from day 12 on, causing an increase in VLP concentration inside the bioreactor and a decrease in VLP recovery in the harvest (Figure 3C).
As discussed above, product retention had already been observed for virus and VLP production when typical polyethersulfone ATF cartridges with 0.2 μm pore size and 0.13 m2 filtration area were used (Alvim et al., 2019; Coronel et al., 2019; Lavado-García et al., 2020). In spite of using in the present work a different membrane cartridge, having the same characteristics (polysulfone membranes with 0.4 µm pore size and 0.005 m² filtration area) as the cartridge that Nikolay et al. (2020b) found to promote an efficient yellow fever virus recovery, product retention did occur again.
In their work, Nikolay et al. (2020b) tested the cartridge in a tangential-flow filtration experiment lasting just a few hours, whereby a bioreactor with YFV-infected BHK-21 cells set in a closed recirculation loop was just used to provide cell/virus suspension to challenge the membrane cartridge – no perfusion cultivation was carried out. Thus, considering the results obtained in the present work, it seems probable that in order to effectively avoid fouling in a long-term perfusion process, a membrane cartridge with the characteristics selected by Nikolay et al. (2020b) would need to have a much larger filtration area, such as for example the 0.13 m2 of the traditional cartridge sold by the ATF manufacturer for the ATF-2 system.
Thus, to overcome product retention and membrane fouling issues observed in ATF-based perfusion herein, large-area membrane cartridges of specific materials and porosities could possibly be a solution and should be tested. However, in view of the low production cost that is desirable for a vaccine to be accessible also for the population of low-income countries, we considered it more adequate to proceed investigations using an inclined lamella settler, which is a low-cost cell retention device that does not use membranes or any other consumables and is intellectual-property free. Although no previous reports of HEK293 perfusion processes using a settler was found in literature, this cell retention device has been shown to provide very high separation efficiencies with no product retention in previous works with other mammalian cell lines (Coronel et al., 2020a; Coronel et al., 2020b).
As shown in Figure 4, a perfusion run with a CS-10 inclined settler using cell pool 2 resulted in a maximum VCD of 41.2 × 10cells/mL on day 12, thus reaching the same viable cell concentration as obtained in pseudoperfusion on the same cultivation day (Figure 2B). Medium exchange rate was gradually increased during the run, with the aim of maintaining glucose levels at approximately 1 g/L (Figure 4B). In order to sustain VCDs above 25 million cells/mL, perfusion rates above 1.5 vvd had to be set, confirming how importantly membrane fouling in the ATF-based perfusion affected the necessary nutrient supply. Regarding the VLP product, as expected no product retention was observed, and VLP concentrations inside the bioreactor and in the perfusate harvest were of approximately 200 mg/L from day 10 on (Figure 4C) at dilution rates of up to 2 vvd (Figure 4B), thus reaching volumetric productivities of approximately 400 mg/L/d, which are higher than those obtained in the pseudoperfusion run (270 mg/L at 1 vvd, thus volumetric productivity of 270 mg/L/d). Based on these results, the use of the inclined cell settler combined to the double-sorted cell pool 2 holds a promising potential to establish an intensified perfusion process for yellow fever VLP production.