Recombinant Adeno-Associated Virus Production,the Beginning of the End of Uncertainty

In the early days of recombinant adeno-associated virus (rAAV) vector development, there was only one primary approach to produce these reagents. ¹ Owing to the complexity of rAAV production, rAAV vector manufacturing was more of an art than a science and mastered by a handful of laboratories with respect to reproducing quality and quantity. Over time, rAAV vectors have become a mainstay scientific research tool as well as an essential component of gene therapy drugs for curing human diseases. ²

As a result, a variety of methods for rAAV vector production have been developed to address the various needs of AAV manufacturing. Today, clinical rAAV vectors are either produced in Sf9 cells (baculovirus system), or mammalian cells such as 293 cells using herpes simplex virus, adenovirus (Ad), or triple transfection approaches. Although the quantity of rAAV production is no longer an obstacle (i.e., reaching amounts over 1E17 particles/batch), quality and effectiveness of vectors transduction made by different methods have been in continuous debate. Surprisingly, characterization of the exact contents in rAAV preps remain underdeveloped.

In this issue, Tran et al. present an extensive comparison of vectors made from 293 cells and insect cells, primarily from a genomic perspective. ³ In this study, conventional technology, such as ultracentrifuge density gradient and denaturing gel analysis, showed more heterogeneity with vectors made from insect cells. Oversized genomes (larger than unit length vector) appeared to be characteristics of full particles derived from insect producer cells purified by immunoaffinity method.

Consistent with traditional assays, modern technology, such as PacBio single molecule sequencing, revealed that vectors from insect cells again exhibited different profiles when the entire AAV genome population was mapped and sequenced. Such differences exist in both full and “empty” particle population. Notably, there was a higher degree of truncated and unresolved species in vectors from insect cells. The key cis element in rAAV vectors, namely, the inverted terminal repeats (ITRs), again generated more mutative forms in the vectors derived from insect cells. ⁴

While the authors avoided making a direct conclusion on what vector is a better fit for developing clinical therapeutics from this study, the reader is left with the obvious question of “what to use now” when faced with such choices. From this study, there is a clear difference between vectors produced in 293 cells versus Sf9 cells (at both the population and individual vector genome level). When looking for an explanation, it is likely that the production environment contributes to these discrepancies.

After all, AAV is a mammalian virus and 293 cells are considered to be a natural host for AAV production. Although previous studies suggest that insect cells are equally competent for rAAV production and that there are no significant differences in the final products when assayed using common molecular biology methods, ⁵ recent studies have illustrated composition and performance differences between vectors produced using these two systems. ⁶

From an AAV basic biology perspective, virion DNA compositions are complicated due to the strong secondary structure of the AAV ITR and its plenteous functions. ⁷ It remains unclear if the biological properties of rAAV vector particles are exactly the same as the wild-type virus since rAAV vectors are manufactured under conditions that are different from wild-type AAV propagation. ^8,9 To improve rAAV vector yield, cis and trans elements for replication and packaging have been tweaked and deviated significantly from wild-type AAV growth. ^10,11

Owing to the uncontrollable factors in vector production, rAAV vectors made by different laboratories have been difficult to compare directly using a few simple parameters (e.g., full to empty ratio and % packaging of incomplete genomes). This eventually led to the establishment of an AAV reference standard so vectors can be directly compared. ¹² As to vectors made by insect cells versus mammalian cells, the first distinction noted was the vast difference in potency of transduction (293 > Sf9).

Since this observation, various methods have been developed to enhance the correct capsid composition made in insect cells. Clearly, significant improvements have been reported concerning this aspect. Nevertheless, there still seems to be transduction behavior differences between vectors made by Sf9 cells compared with AAV 293 cells. Not surprisingly, significant variations between vectors made from 293 cells and insect cells are observed depending on the characterization methods. Notably, clinical vector doses from vectors derived from insect cells are significantly higher than those from mammalian cells. ¹³

However, in vivo transgene expression is not the only parameter to consider when using vectors in patients. Extraordinary heterogenous vectors have been reported to express high-level transgene protein in animal studies. ^14,15 Multiple groups have demonstrated that vectors way over the packaging capacity of AAV (i.e., up to 9kb from ITR to ITR) can still lead to transgene expression. Since AAV, similar to most DNA viruses, package by a “Head full” mechanism, it is unlikely any intact oversized genomes are packaged inside the capsid, but instead these vectors carry fragmented genomes.

Fragmented rAAV genomes appear to recombine with one another and yield some intact oversized genomes eventually. ¹⁶ Interestingly, the Tran et al. study clearly shows that vectors made from insect cells have features of those of fragmented AAV vectors, regardless of genome size and even those within the packaging capacity of AAV capsids. Many studies have demonstrated vectors derived from insect cells express high-level transgenes; however, it is unclear if there is any increase in mutant transgene genome derived from recombination among these fragmented AAV genomes. ⁵

One major feature of fragmented AAV vectors is the strong recombination potential. ^16,17 Such genomes exist in various forms: single stranded DNA, duplex DNA, or partial ssDNA. Nevertheless, all of them have a copy of the viral ITR at the 3′ end, the end reported to initiate AAV packaging. ¹⁸ The ITR itself can undergo site-specific integration in the presence of Rep proteins. Since vectors from insect cells are used clinically at much higher doses, it warrants future studies to monitor the effects of fragmented AAV genomes on host cells, especially integration potential.

Regardless, the mechanisms contributing the genome differences between vectors from insects and mammalian cells remain unclear. Generally, AAV genomes packaged into the capsid are processed through several key steps, including (1) AAV plasmid genome rescue, (2) vector replication, (3) vector packaging initiation, and (4) appropriate vector termination. ¹⁹ Notably, the rescued/replication of AAV DNA dictate what template molecules are available for AAV encapsidation.

Erroneous rescue/replication of AAV genomes, therefore, may give rise to numerous forms of abnormal genomes with ITRs. ²⁰ In contrast, erroneous packaging initiation will likely utilize nonstandard AAV packaging signal to initiate AAV encapsidation, which may result in viral particles containing genetic elements found in (1) the producer cell line (i.e., host cellular DNA), (2) viral/plasmid backbone, and/or (3) AAV/Ad helper sequences.

In contrast, the consequences of erroneous packaging due to premature termination resulting in non-unit length genomes may be the primary source of incomplete genomes and/or the generation of oversized genomes. As pointed out by the authors, the highly recombinogenic nature of Autographa californica multiple nucleopolyhedrovirus replication could be one potential source for generation of these abnormal genomes. In addition, AAV replication appears to be different in insect cells. An excessive amount of closed-end AAV genomes (unresolved AAV replication substrate) has been observed primarily in the insect cells. ^4,21

Despite the genome differences between rAAV derived from these two popular platforms, ³ vectors from both systems are commonly used for human gene therapy development. This study has clearly demonstrated what is different between vectors manufactured by these two platforms. Now, both basic science researcher and clinical research product developers should have a better scientific basis for choosing the appropriate vectors for their clinical efforts. More importantly, based on this study, vector manufacturers now have a better understanding of additional characterization needed for perfecting rAAV vector production and removing the uncertainty of these therapeutic products going forward.

Author Disclosure

The authors declare having potential competing financial interests. WX is a founder of Ivygen Corporation and Nikegen Inc and holds equity in the companies. RJS is a consultant for AskBio BioPharmaceutical, Inc., a wholly own company of Bayer AG.