(523c) Process Development Considerations for Ultrafiltration/Diafiltration of Viral Vectors Used in Gene Therapy: Theory and Practice

The biopharmaceutical industry has made significant progress in effectively treating diseases using viral gene therapies. However, a lack of standardized upstream and downstream processes poses significant challenges to developing these therapies both economically and at a commercial scale.

A critical part of purification in the downstream process (DSP) addressed within this work is the ultrafiltration/diafiltration (UF/DF) steps, which can be optimized through the proper selection of membrane molecular weight cutoff and operating mode. Properly designed tangential flow filtration (TFF) operations can provide concentration, buffer exchange, and impurity reduction through UF/DF in the downstream manufacturing of viral vectors. Viral vectors can be retained while also removing impurities like endonucleases using a 100 or 300 kDa membrane cut-off.

Most gene therapy manufacturing processes are based on adeno-associated virus (AAV) vectors.

AAV feed streams can only be filtered in a stable way with 100 or 300 kDa membranes up to a certain permeate flux for a given concentration; then the system pressure becomes unstable and rises rapidly. This is called a critical flux application. To prevent this pressure instability, the process can be run at a much lower controlled flux -historically set by a conservative rule of thumb. Another option is to start at a higher flux, but allow it to drop naturally by holding the transmembrane pressure (TMP) constant.

This work considers these alternative control strategies, and the underlying theory of critical flux operation, to allow for optimization of performance and scaling. Performance in lab trials was evaluated based on critical flux and UF/DF process simulations for clarified producing and non-producing adeno-associated virus (AAV2) cell cultures. Simulated process runs were conducted using a 10X concentration step followed by a 5-diavolume diafiltration step. Permeate control model flux level was set using a conservative rule-of-thumb approach (50% of the critical permeate flux). TMP control had a flux (or processing time) advantage over permeate control, while permeate control (of flux) had an impurity clearance advantage. Both control modes had similar yields (>85% recovery) of AAV2. Initial process development and optimization of parameters was performed with laboratory-scale 50 cm² cassettes using 50-200 ml volume. Lab-scale parameters were then utilized to scale-up to larger area devices, including cassettes and spiral-wound membrane capsules, demonstrating a successful 300-fold scale-up of the operation.

These experimental results show agreement with the underlying theory of critical flux operations. Ideas for future work to optimize conditions and select control mode, based on the underlying theory, will be presented as well. These findings will provide a framework for developing efficient and stable TFF processes for viral gene therapy DSP applications.