(M1130-08-43) Final Sterile Filtration of Lentivirus

Purpose: Sterility of a biologic drug substance is often achieved by its filtration with a sterilizing grade membrane rated at ≤0.2 μm [1, 2]. When applied to lentiviral vectors (90-120 nm in diameter [3]), the sterile filtration requirement is viewed as problematic due to its similar size as the pores of a 0.2 μm membrane [4] and its fragile lipid envelope [5, 6]. Promoting this perception are low yields in lentivirus sterile filtration, which impose increased bioreactor volume and larger downstream sizing. This study’s purpose is to examine whether feed characteristics or membrane selection is more important in the final sterile filtration of lentivirus

Methods: This study included three trials, each beginning with fresh lentivirus harvested from suspension grown HEK293 cells. After DNase addition, each harvest was clarified using cellulose depth filter media followed by 0.2 µm membrane filtration. Bind and elute anion exchange chromatography provided additional purification. Purified lentivirus was concentrated by tangential flow filtration using a 100 kDa membrane. A subsequent diafiltration, using a buffer of 50 mM HEPES, 20 mM magnesium chloride, and 140 mM sucrose at pH 7.3, produced the sterile filtration feed. For the sterile filtrations, one prototype, one polyvinylidene fluoride (PVDF), and five polyethersulfone (PES) membranes having a nominal pore size of 0.2 μm were evaluated. To measure the membranes’ capacity (volumetric loading at 690 mbar transmembrane pressure) and virus transmission, they were screened using 13 mm (0.7 cm²) and 25 mm (3.5 cm²) sized holders. The smaller 13 mm holder ensured that membrane capacity was measured, and the 25 mm holder ensured sufficient filtrate for analytics. Membranes were tested in parallel at a filtrate flux of 150 L·m^-2·h^-1 using a multichannel syringe pump. Membrane inlet pressure during filtration was recorded using a pressure transducer located before the membrane holder. Lentivirus transmission was calculated using titers measured by a live cell infectivity assay and p24 ELISA. Particles sizes were measured by dynamic light scattering.

Results: The three diafiltered lentivirus pools had variable turbidity and titer. Turbidities of 70, 140, and 310 NTU and titers of 3.1×10⁸, 2.3×10⁸, and 2.6×10⁸ TU/mL were observed, respectively. Having an almost milky appearance, turbidity of the diafiltered lentivirus was plainly visible. The variable turbidity in a narrow range of titers proved beneficial as it allowed for an examination of how filterability changes with turbidity. As shown in Figure 1, increasing feed turbidity causes membrane capacity to decline sharply. The prototype membrane had the highest capacity across the observed turbidities. At 70 and 140 NTU, PES-3 was the second top performer. At 310 NTU, the capacities for all membranes studied was less than 20 L·m^-2. The capacity drop-off due to turbidity has profound implications for process scale-up, as low capacity may lead to significant losses to capsule void volumes when increasing relative to membrane surface area. Lastly, while the PVDF membrane appears to have performed poorly, this might be attributable to poor wettability of the PVDF membrane giving high initial inlet pressures. Prewetting this filter before filtration may improve capacity. A representative set of particle sizes measured by dynamic light scattering for the prototype membrane are shown in Figure 2. As shown in Figure 2a, the TFF feed has one large peak at 50-200 nm, where lentivirus is expected. The sterile filtration feed, shown in Figure 2b, also has this peak at 50-200 nm and an additional peak in the 700-3000 nm range. The presence of larger particles suggests the observed turbidity increase is from the formation of aggregates/precipitates of an unknown and possibly heterogenous composition. Being larger than 200 nm, retention of the aggregates/precipitates by the membranes should be unsurprising (Figure 2c). The poor membrane capacity observed may be explained by retention of these aggregates/precipitates. The average virus transmission across the three trials as calculated by infectivity assay and p24 ELISA are shown in Figure 3. The virus transmission for both the infectivity assay and p24 ELISA agreed, and, except for PES-3, a virus transmission between 80% and 100% was measured. Also, turbidity was found to not strongly affect lentivirus transmission calculated from titers measured by either assay.

Conclusion: The filterability of lentivirus using 0.2 µm membranes was studied using similar titer feedstocks with variable turbidity. While many of the membranes studied had 80-100% virus transmission, capacity decreased sharply with higher feed turbidity. This precipitous decline in capacity impacted even the prototype membrane, which stood out having significantly higher capacity than selected peer membranes. From these data, filter selection does matter with lower turbidity feeds, but as turbidity increases its significance in filterability becomes much greater. At some point, the feed turbidity high enough that it must be mitigated in order to further improve sterile filter capacity. This makes identification and mitigation of lentivirus pool turbidity an essential part of continued improvement in lentivirus downstream process development.

References: [1] U.S. Department of Health and Human Services Food and Drug Administration, "Guidance for Industry: Sterile Druge Products Produced by Aseptic Processing," September 2004.
[2] European Medicines Agency, "Guideline on the sterilisation of the medicinal product, active substance, excipient and primary container," 6 March 2019.
[3] S. Pettit, C. Glover, J. Hughes, J. Madsen and P. Jaluria, "Downstream Manufacturing of Gene Therapy Vectors," in Key Considerations in Gene Therapy Manufacturing for Commercialization, The Cell Culture Dish, Inc., 2018.
[4] R. I. Pinador, M. Kaabouch, R. B. Aim and J. I. Calvo, "Non-Destructive Characterization of Industrial Membrane Cartridges by Using Liquid–Liquid Displacement Porosimetry (LLDP)," Membranes, vol. 10, no. 12:369, 2020.
[5] C. J. Burrell, C. R. Howard and F. A. Murphy, "Virus Replication," in Fenner and White’s Medical Virology: Fifth Edition, London, UK, Elsevier, Inc., 2017.
[6] D. Escors and K. Breckpot, "Lentiviral Vectors in Gene Therapy: Their Current Status and Future Potential," Archivum Immunologiae et Therapiae Experimentalis, vol. 58, pp. 107-119, 2010.

Acknowledgments: Conflicts of Interest
Membranes PES-1 and PES-2 are Pall’s Supor EKV and Supor ECV membranes, respectively.

Figure 1. Sterile filtration feed turbidity vs. capacity.



Figure 2. Examples of observed particle size intensities of TFF feed (A), sterile filtration feed (B), and prototype membrane filtrate (C) measured using DLS performed on the Stunner (Unchained Labs, Gent, Belgium). Note that y-axis are different for each sample.



Figure 3. Virus transmission from TU and p24 averaged for three independent trials (a) and averaged across all membranes for the range of turbidities (b). Error bars represent standard deviations.