AAV vector production: Troublesome host innate responses in another setting

Main text

Products based on recombinant adeno-associated virus (AAV) continue to advance, with multiple AAV gene therapies approved and additional AAV products reaching the licensure application stage in 2022. However, the limited manufacturing capacity of AAV vectors has posed challenges for expanding clinical and commercial product development needs. A key problem is the low volumetric productivity of AAV generation in cell culture, limiting the number of doses that can be manufactured from bioreactors, especially for high-dose indications such as those requiring systemic administration. In this issue of Molecular Therapy – Methods and Clinical Development, Chung and colleagues¹ report a study that uses time-series transcriptional analysis and Gene Ontology to characterize the cellular pathways in HEK293 production cells that are activated by transient transfection-based vector production. The study reveals potential critical bottlenecks in AAV vector production that, if addressed, could increase AAV production efficiency toward the much higher levels achieved with recombinant proteins such as monoclonal antibodies.

The AAV vector manufacturing productivity gap

To estimate the potential for increasing the volumetric productivity of current AAV production systems, it is informative to compare the manufacturing yields of AAV vectors and monoclonal antibodies. A general observation is that kilogram amounts of monoclonal antibodies, but only gram amounts of AAV vectors (10¹⁷ vg ∼1 gm), are produced from comparable manufacturing-scale bioreactor volumes and quantities of production cells. Table 1 provides a side-by-side comparison of per-cell and volumetric productivities for the generation of AAV vectors and monoclonal antibodies, with production yields expressed both in mass units and molecule copy numbers. The comparison illustrates the large relative deficit in the production efficiency of recombinant AAV (rAAV), emphasizing the potential to address AAV manufacturing capacity limitations by identifying and removing bottlenecks to the efficient biosynthesis of AAV vectors. Achieving production of 20–50 pg/cell/day, the current standard for monoclonal antibodies in CHO cells,² establishes a feasible level of recombinant product synthesis in mammalian cells that corresponds to 2–5 × 10⁶ vg/cell/day for rAAV. This rate of vector production maintained for even one day in HEK293 cells would generate ∼10¹⁹ vector genomes in a 1,000 L bioreactor. The magnitude of the rAAV production capacity gap emphasizes the importance of identifying the key bottlenecks in current vector generation processes.

Table 1.

The AAV production efficiency gap: Comparison of AAV and monoclonal antibodies

Biologic product	Per-cell productivity		Volumetric productivity (batch)
	Molecules/cell/day	pg/cell/day	Molecules/L	g/L
AAV vectors	1–2 × 105	1–2	0.1–1 × 1015	0.001–0.01
Monoclonal antibodies	800–2,000 × 105	20–50	400–3,200 × 1015	0.1–0.8

Inflammatory and antiviral responses may contribute to the productivity gap

The authors also offer a key insight toward specific strategies to address the AAV productivity gap. They describe RNA sequencing (RNA-seq) analysis on samples taken from production bioreactors at time intervals from the point of transfection of HEK293 cells through to vector harvest. This time-series transcriptional analysis revealed 1,850 differentially expressed genes (DEGs). They used a qualified scaled down 2 L bioreactor to accurately model a production-relevant 50 L scale and performed frequent sampling (0, 3, 6, 12, 24, 36, 48, and 72 h) to provide sufficiently high resolution to distinguish early and late differential gene expression patterns over the course of transfection and vector production. Gene Ontology analysis based on the 1,850 DEGs identified the following cellular responses: (1) regulation of MAPK activity, (2) upregulation of histone genes, and (3) pathways related to inflammation and innate responses to viruses. Upregulation of histones was deemed likely associated with the increased DNA and protein synthesis requirements of AAV capsid and genome synthesis and assembly. However, the changes in MAPK expression and inflammation and innate response pathways were deemed features of recognition of bacterial (plasmid) and viral nucleic acids by receptors in transfected HEK293 cells, i.e., an active response to pathogen-associated stimuli. Inflammatory and antiviral responses are part of the evolved immune response in living organisms that limit virus propagation. Therefore, a hypothesis that emerges from these findings is that the inflammatory and innate antiviral responses reduce vector production efficiency. The authors further describe different post-transfection time frames and cellular pathway features that they map to an early phase (3–6 h) consistent with inflammatory responses to non-viral transfection components such as plasmid DNA and the transfection reagent. Responses in the mid-late (6–12 h) and late (24 h and after) were typical of evolving innate antiviral responses concurrent with the nascent generation of AAV particles at increasing titers.

This rigorous analysis of differential gene expression caused by AAV production in HEK293 cells informs future strategies to improve vector production efficiency. One approach would be to genetically modify the HEK293 cell line to remove cornerstone components of inflammatory and innate response pathways so that inhibition of vector synthesis by these pathways is prevented. The findings also raise the question of to what degree similar pathogen pattern and antiviral responses occur in other production cell lines and if they reduce vector production efficiency in those systems. AAV vector production platforms that use recombinant helper viruses such as rBac and rHSV rather than plasmid DNA to introduce genes needed for AAV vector generation would be expected to similarly demonstrate innate response, and time-series transcriptional analyses performed on those systems using the approach described by Chung and colleagues would be informative. The recent report that innate pathogen pattern receptors are especially sensitive in human cells compared with other mammals³ poses the possibility that reduction of AAV vector productivity would be less in non-human mammalian cells such as CHO.

Unwanted innate responses in both vector production and clinical vector administration

Innate immune responses in human recipients have been recognized and characterized as a significant risk for AAV product safety and efficacy.⁴ Emerging research indicates that innate responses to AAV vectors are also a concern for production of AAV vectors. Common pathogen-associated molecular patterns, including those associated with AAV vectors and the microbial components used for making vectors, are now implicated in inducing innate pathways in human production cells that limit vector production yields. The findings reported by Chung and colleagues, specifically the HEK293 response pathways that they propose are induced by nascent AAV vectors during production, may provide useful insights and even a partial model for innate responses in human recipients of AAV vectors.