(W0930-02-09) Recombinant Adeno-Associated Virus (rAAV) Production in Spodoptera Frugiperda (Sf9) Cells: Viral Cathepsin Mediated Capsid Cleavage and Mitigation Strategies

Purpose: Currently, the two most common methods of recombinant adeno-associated virus (rAAV) production are plasmid transfection of human embryonic kidney 293 (HEK293 cells) or infection of Spodoptera frugiperda (Sf9) insect cells with recombinant baculovirus (rBV). Our group focuses on rAAV production in Sf9 cells, and we have used this method to produce naturally occurring and novel rAAV capsid serotypes. While many capsids are produced without issue in the Sf9 system (such as AAV2, AAV9, and STAC-BBB), some AAV serotypes are prone to additional viral protein (VP) banding observed by SDS-PAGE or Western blotting (such as AAV8 and some novel capsids). Peptide sequencing confirmed that the additional bands are cleavage fragments from the VPs. Viral cathepsin (VCATH) is a known cathepsin L-like proteinase expressed by baculovirus as a late viral gene. Galibert et al. demonstrated that some naturally occurring serotypes are sensitive to VCATH-mediated VP cleavage during manufacturing in the Sf9 system, and such degradation can be prevented by eliminating or inhibiting VCATH (PMID: 30440025). We tested the addition of a known VCATH proteinase inhibitor (E64) and glycosylation inhibitor (tunicamycin) in Sf9 AAV productions with two affected serotypes (capsid A and B).

Methods: For capsid A and B, we performed multiple small-scale Sf9 productions with 50µM of E64 (n=7 or 4) or 10µg/mL of tunicamycin (n=2 or 3) vs. standard productions (n=10 or 7). We analyzed the AAV for cleavage products, yield, capsid ratio, and transgene potency. Our results showed that tunicamycin did not eliminate the cleavage products and instead caused additional aberrant capsid protein staining by SDS-PAGE for both capsids. E64 addition successfully reduced or eliminated the VP cleavage products. However, E64 addition negatively impacts AAV yield for capsid A, while capsid B slightly improved (58% and 1.08% of the av. abs. yield). Adding E64 improved the VP1 content for both capsids and more closely aligned the VP ratio to the ideal 1:1:10 (1:5:14 to 1:2:11 for capsid A and 1:5:34 to 1:2:14 for capsid B). Finally, we assayed the activity of a zinc finger repressor transgene from productions with E64 or tunicamycin vs. control productions. Adding either inhibitor improved the transgene expression and target repression for both capsids; however, capsid A substantially improved.

Results: Adding either inhibitor improved the transgene expression and target repression for both capsids; however, capsid A substantially improved. For capsid A, average transgene expression improved 4x with tunicamycin vs. 13x with E64, corresponding to 300x vs. 459x repression of the native gene target, respectively. For capsid B, average transgene expression improved 1x with tunicamycin vs. 2x with E64, corresponding to 6x vs. 2x repression. There is no consensus recognition sequence for VCATH, so capsid cleavage cannot be easily predicted before manufacturing or preemptively prevented by capsid engineering.

Conclusion: There is no consensus recognition sequence for VCATH, so capsid cleavage cannot be easily predicted before manufacturing or preemptively prevented by capsid engineering. However, our results demonstrate that E64 successfully inhibits VCATH-mediated cleavage of VP proteins with largely beneficial effects on AAV production when VCATH is present in the rBV backbone. Additionally, this work supports the adoption of VCATH null rBV to produce rAAV.

Acknowledgements: All authors are employees of Sangamo Therapeutics.