The widely used 0.2/0.22 µm polymer sterile filters were developed for small molecule and protein sterile filtration but are not well-suited for the production of large nonprotein biological therapeutics, resulting in significant yield loss and production cost increases. Here, we report on the development of membranes with isoporous sub-0.2 μm rectangular prism pores using silicon micromachining to produce microslit silicon nitride (MSN) membranes. The very high porosity (~33%) and ultrathin (200 nm) nature of the 0.2 µm MSN membranes results in a dramatically different structure than the traditional 0.2/0.22 µm polymer sterile filter, which yielded comparable performance properties (including gas and hydraulic permeance, maximum differential pressure tolerance, nanoparticle sieving/fouling behavior). The results from bacteria retention tests, conducted according to the guidance of regulatory agencies, demonstrated that the 0.2 µm MSN membranes can be effectively used as sterile filters. It is anticipated that the results and technologies presented in this study will find future utility in the production of non-protein biological therapeutics and in other biological and biomedical applications. K E Y W O R D Sbiological therapeutics, silicon nanomembranes, sterile filtration

Efficient downstream processing represents a significant challenge in the rapidly developing field of therapeutic viruses. While it is known that the terminal sterile filtration step can be a major cause of product loss, there is little known about the effect of host cell impurities (DNA and protein) on filtration performance. In this study, fractions of relatively pure Vero host cell protein and DNA were spiked into a highly pure preparation of vesicular stomatitis virus (VSV). Then, the resulting solutions were sterile filtered using two commercially available 0.22 µm rated microfiltration membranes. A combination of transmembrane pressure measurements, virus recovery measurements, and post-filtration microscopy images of the microfiltration membranes was used to evaluate the sterile filtration performance. It was found that increasing the amount of host cell protein from approximately 1 µg/mL (in the un-spiked VSV preparation) to 25 µg/mL resulted in a greater extent of membrane fouling, causing the VSV recovery to decrease from 89% to 65% in experiments conducted with the highly asymmetric Express PLUS PES membrane and to go as low as 48% in experiments conducted with the symmetric Durapore PVDF membrane. Similar effects were not seen when bovine serum albumin, a common model protein used in filtration studies, was spiked into the VSV preparation, which indicates that the sterile filtration performance is critically dependent on the complex composition of the mixture of host cell proteins rather than the presence of any protein. The results presented in this work provide important insights into the role of host cell impurities on the performance of sterile filtration processes for therapeutic viruses.

No abstract