In recent years, virus filtration has been introduced as a robust, size-based operation capable of providing the viral clearance needed for production of therapeutic proteins. Several different companies make virus filtration membranes for bioprocessing applications. Although these membranes all provide at least 3-logs of viral clearance, the membranes have very different capacity and filtrate flux due to large differences in clean membrane permeability and protein fouling characteristics. The objective of this study was to develop a fundamental understanding of the mechanisms governing protein fouling and capacity of virus filtration membranes and the relationship between these performance characteristics and the underlying membrane structure.

Experiments were conducted using commercially available Ultipor DV20 (Pall Corp.) and Viresolve 180 (Millipore Corp.) virus filters, both made from polyvinylidene fluoride. The Viresolve 180 is a composite structure with the pore size increasing from less than 35 nm at the upper skin to micron size in the substructure. In contrast, the DV20 has a fairly uniform pore size throughout the structure, resulting in a hydraulic permeability that is more than 25 times smaller than that of the Viresolve 180. Bovine serum albumin (BSA) was used as a model protein. Data were obtained in a stirred ultrafiltration cell at both constant pressure and constant filtrate flux. The effects of fouling on the membrane were examined from both buffer permeability and dextran sieving measurements obtained with the clean and fouled membranes.

The Viresolve 180 membrane displayed a very sharp flux decline during constant pressure operation, with large differences in performance depending upon the orientation of the membrane (skin-side facing up or down). The dominant contribution to the flux decline in the skin-up orientation was concentration polarization, with membrane fouling being most important with the skin-side down. In contrast, the rate of flux decline was quite slow with the DV20 membrane even over 24 hr, which was directly related to the very low flux arising from the low hydraulic permeability of these membranes. The flux decline with the DV20 membrane was due almost entirely to fouling, with no measurable effect of concentration polarization. The volumetric capacity (L/m²) of the DV20 membrane was an order of magnitude larger than that of the Viresolve 180 for constant pressure operation, although the DV20 required much longer filtration times. In addition, the volumetric capacity of the DV20 was nearly independent of the bulk protein concentration, while the volumetric capacity of the Viresolve 180 membrane decreased almost linearly with increasing protein concentration. The very different behavior of these membranes is discussed in terms of the differences in underlying pore structures. These results provide important insights into the design and operation of virus filtration membranes.