AAV vector manufacturing process design and scalability - Bending the trajectory to address vector-associated immunotoxicities

Main text

The pace and diversity of adeno-associated virus (AAV) vector based clinical trials over the last 5 years has been high, with definitive therapeutic benefits observed with the two products licensed by the US Food and Drug Administration, as well as several investigational products in pivotal trials demonstrating the paradigm-shifting potential of AAV based gene therapy for serious unmet medical needs. However, this excitement has been tempered by the emergence of serious adverse events (SAEs) in several clinical trials, many of which seem to be caused by innate and adaptive immune responses to the vector (https://www.fda.gov/media/151599/download).¹

Product critical quality attributes (CQAs) are by definition features that must be within an appropriate limit, range, or distribution to ensure product safety and efficacy. The latter are also inextricably linked to manufacturing processes—the product is the process—and efforts should focus on understanding the product features that have contributed to vector-related SAEs and how these can be influenced by manufacturing process design. The challenges of comparability studies to support process evolution during clinical development notwithstanding, establishing high capacity, cost-effective manufacturing processes in the absence of a comprehensive understanding of product CQAs is putting the cart before the horse. Defining all product safety CQAs, including those not identified in pre-clinical studies but that subsequently emerge from clinical experience, is a pre-requisite to finalizing a manufacturing process. Ease of scale-up and cost efficiency are important co-requisites that will ultimately be required for the industrialization of AAV gene therapy.

In contrast with recombinant therapeutic human proteins made in mammalian cell lines, the non-self-nature of AAV vectors with respect to the human immune system is more akin to vaccines designed to stimulate immune responses. The presence of viral antigens and pathogen-associated molecular patterns (PAMPs) contributed by AAV capsid amino acid sequences, residual viral DNA sequences, microbial epigenetic PAMPs in vector genomes,² and the use of microbial components (recombinant plasmids and viruses) for AAV manufacture together ensure that the in vivo administration of vectors will trigger innate and adaptive immune responses. Achieving safety and durable efficacy may require significant vector and manufacturing process design changes in response to clinical experience. Byrne and colleagues³ recently compared the immunological challenges using the systemic administration of AAV vectors with those encountered during the development of organ transplantation to support that improved immune management is key to mitigate immunotoxicity after administration—transplantation—of AAV vectors to human subjects. Building on this comparison, the establishment of tissue typing to maximize histocompatibility, i.e., minimize the immunological differences between transplanted tissue and recipient is another key requirement for successful organ transplantation. By analogy, identifying and decreasing potentially inflammatory vector attributes by manufacturing process design improvements is important to complement better immune management. One prudent objective in response to the emergence of immune response-related SAEs is to critically evaluate, and where necessary, modify upstream (cell culture-based vector generation) and downstream (vector purification) manufacturing processes to decrease product immunogenicity.

As has been recognized and highlighted by other investigators, there remains a need to establish AAV upstream processes that can meet projected vector production capacity as clinical programs progress and more products are commercialized.⁴ To highlight the manufacturing capacity gap and potential for improvement, a back-of-the-envelope calculation estimates that approximately 10¹⁹ AAV vectors have been manufactured in total for administration to human subjects in clinical trials and commercial dosing to date, corresponding with approximately 100 g recombinant AAV (rAAV), a quantity that for a typical monoclonal antibody can be manufactured from 20 L of cell culture. The current generation of rAAV production technologies using microbial helper components has supported remarkable progress, but the low volumetric production efficiency combined with product-associated immunotoxicities revealed by clinical experience argue that the acceleration of alternative AAV production platforms should be an important future direction. Engineered cell lines that can be induced to express the gene(s) required for product generation represent the industry standard for recombinant therapeutic proteins such as monoclonal antibodies. The challenges that have hindered the development of such cell lines for AAV vector production, including the complexity and cytostatic/cytotoxic nature of the multiple gene products required, are being overcome by advancing technologies, including more efficient and better choreographed induction of production gene expression. Importantly, the continued development of such producer cell lines should help to decrease the immunogenicity of the current generation of AAV vectors, e.g., by eliminating the need for microbial helper components, increasing vector genome packaging efficiency to decrease the fraction of empty capsids in the upstream harvest, and decreasing unmethylated CpG motifs in AAV vector genomes implicated in TLR9 innate signaling. Accelerating development of such producer cell lines could meet the twin aims of better process scalability and decreased product immunogenicity.

For vector purification, while standard bioprocess steps developed for protein purification have been successfully adapted for AAV vectors, one unique and challenging problem is the heterogeneity of AAV particles generated using current upstream platforms. There is variability in the nomenclature used to describe the diverse AAV particles formed. Table 1 proposes division of the AAV particle types into four categories and summarizes their associated immunological risks. The term “partials” is proposed to be reserved for AAV particles that contain a portion of the intended vector genome, which may be common for oversized vector genomes, that contribute to target cell transduction.⁵ Empty capsids are generally the most abundant particle type generated in cell culture production because of the inefficiency of vector genome packaging into pre-formed capsids and they are difficult to separate from the target vector. The most efficient method to remove empty capsids that is used by some sponsors is a gradient ultracentrifugation step. Column chromatography can also decrease empty capsids, but with a lower resolution and less robustly than ultracentrifugation. Some sponsors have chosen to accept the co-purification of empty capsids with the vector product because the steps for their removal are difficult to scale up and validate; such purified products may contain 50%–90% empty capsids, translating to a 2- to 10-fold higher capsid antigen dose for a given vector genome dose. Therefore, the decision of whether or not to remove empty capsids can represent a choice between greater process scalability versus lower product immunogenicity; which should prevail? The emergence of serious immunotoxicities after high-dose AAV vector administration, including complement activation that are likely capsid dose dependent, argues that total capsid dose is a critical quality attribute for safety at high vector doses and that empty capsids should be efficiently removed.

Table 1.

Proposed categorization of AAV particles generated during cell culture

	AAV particle description	Other names	Category	Immunological risk
1.	Packaged full-length vector genomes	“Fulls”	Product	Inherent product risk
2.	Packaged partial vector genomes	“Partials”	Partial product	Reduced transduction efficiency leads to higher dose requirement
3.	Packaged DNA impurities (nuclease-resistant cell and helper DNA)	Residual packaged DNA impurities	Product-related impurities	Microbial DNA PAMPs and ORFs lead to innate and adaptive immune responses
4.	Empty capsids	“Empties”	Product-related impurity	Lead to multi-fold increase in total viral capsid dose