Large-scale production of adenoassociated virus (AAV) vectors at Takeda’s 1,000-L good manufacturing practice (GMP) gene therapy facility in Orth, Austria.

Adenoassociated virus (AAV) has emerged as the leading vector for gene therapy delivery. Compared with options such as lentivirus and adenovirus, AAV exhibits a strong safety profile because it has low pathogenicity and requires a helper virus to replicate. AAV is also capable of long-term gene expression, and it can infect both dividing and nondividing cells (1–5). Developers of advanced therapies have found such advantages to be quite attractive. As of January 2021, two gene therapy products have gained US Food and Drug Administration (FDA) approval: Luxturna (voretigene neparvovec) from Spark Therapeutics and Zolgensma (onasemnogene abeparvovec) from AveXis/Novartis. Both products use AAV vectors (serotypes 2 and 9, respectively), as do many candidate therapies moving through clinical studies. Cost-effective AAV manufacturing remains elusive, however. Yields and expression titers from upstream processes continue to be low — an acute problem for a modality that requires doses with high vector concentrations.

Early in 2022, I corresponded with Barbara Kraus (head of the Gene Therapy Process Development department at Takeda’s site in Orth, Austria) to explore best practices for scaling AAV production. Kraus emphasized the importance of identifying and monitoring critical quality attributes (CQAs), which in turn necessitates careful coordination of process and analytical-method development.

Kraus has over 20 years of experience in biotechnological project development. Over the past five years, her focus has moved increasingly toward gene therapy development. She has a PhD in biotechnology from the University of Natural Resources and Life Sciences (BOKU) in Vienna and an MBA in pharmaceutical management from the Danube University Krems, both in Austria.

Our Conversation
How would you describe Takeda’s typical scale-up process for AAV vector production? We use a proprietary production toolbox to generate our AAV-based gene therapies for clinical trials. It is based on a transient transfection platform, using a traditional three-plasmid approach and a proprietary suspension-adapted human embryonic kidney (HEK293) cell line cultured in chemically defined, protein- and serum-free media. We believe that this system gives us the highest flexibility in producing different AAV vector variants for use in animal models and preclinical studies to identify lead candidates for clinical trials. We have scaled up the entire toolbox from 500-L to 1,000-L scale. Currently, we are using a good manufacturing practice (GMP) facility with 1,000-L capacity to produce clinical trial material (CTM). The facility is equipped for completely single-use operations. Based on our current knowledge, we will use the same facility for commercial manufacturing.

When we find a new lead candidate for CTM production, we usually start with short process and analytical feasibility studies to determine whether our current process settings are applicable for the new vector type. Such studies are performed using high-throughput technologies or at small scales. Based on the results, we decide whether to proceed directly into pilot-scale production of preclinical material. If we determine that yields and/or quality of the drug substance might not meet specifications, we initiate a short development phase and then test those findings at pilot scales before moving into the final 500-L or 1,000-L scale for CTM production.

What aspects of AAV production are most difficult to perform at commercial scales? It is well known that certain parameters for some process steps cannot be scaled up linearly — e.g., stirrer speed in a bioreactor. Those are the most difficult parameters to scale. However, an experienced team that knows the underlying physical principles and calculations should not encounter any problems in scaling them.

Equipment for ultracentrifugation of adenoassociated virus (AAV) vectors at Takeda’s gene therapy production facility in Orth, Austria.

Early in process development (at micro and laboratory scales), a team should consider limitations with process parameters that might occur in the equipment that is to be used at the final scale — e.g., minimum and maximum flow rates in chromatography and tangential-flow filtration (TFF) systems. Accounting for such limitations when designing a process at the laboratory scale helps to minimize need for adjustments at full scale. Sometimes, quality-critical and time-dependent process steps are overlooked because development teams can underestimate how long an activity might take at a scale 100× larger than it would at lab scale.

Triple transfection might not be the most difficult step in the entire AAV manufacturing process, but it is one of the most important. If it is not defined well enough and some quality-critical process parameters remain unknown, then problems can arise during a transition to manufacturing scales. Such problems can manifest as low yields, among other things.

We have not yet observed that percentages of full capsids decrease with increasing scale. However, we have detected repeatedly that packaging efficiency (the proportion of full to partially filled and/or empty capsids) relates strongly to the transgene and AAV subtype used. Through ongoing research into that topic, we plan to understand that packaging mechanism more fully and thus to increase the proportion of full to partial/empty capsids to the highest possible value.

How can such concerns be minimized during scale-up? Two important considerations can facilitate scale-up, the most important of which is to identify process parameters that will have critical impacts on product quantity and/or quality. Depending on the experience level of a process development team, that task can be quite challenging. For a typical upstream process, critical attributes can include seeding density, pH, partial oxygen pressure (pO₂), agitation rate, and maximum cell density before culture split.

During scale-up, quality-critical parameters must be kept within specified ranges such that the quality and quantity of vectors produced at full scale are comparable to those obtained at pilot scale, which is the scale used to produce material for animal studies.

Thus, a second consideration comes into play: development of analytical methods and corresponding process-control strategies. For process development and scale-up to succeed, analytical methods must be able to measure quality-critical parameters with high precision and reliability. The interaction between process development and analytics should function smoothly, whereby the most important analytical methods preferably will have a turnaround time of only one or two days.

What keeps the scaling of AAV production processes interesting for me is the fact that vectors with different capsids and transgenes can have unique production requirements. The setpoints for parameters in different process steps need to be adjusted accordingly to achieve the requisite quantity and quality of an AAV drug substance.

What expression systems are available for AAV production, and how do those systems influence process scalability? Four main platforms are used in the gene therapy industry:

• transient transfection systems using HEK293 cells (both adherent- and suspension-culture formats are available).
• stable packaging/producer cell lines (e.g., HeLa cells).
• herpes simplex virus (HSV) systems.
• baculovirus expression vector systems (BEVS) using Spodoptera frugiperda (Sf9) cells.

These platforms all are used by gene therapy developers and contract development and manufacturing organizations (CDMOs) alike. Each system requires a high degree of specialization so that setpoints for critical parameters are known and well controlled. To some extent, each platform also requires a tailored analytical package that enables users to determine and control relevant process-related impurities. Raw-material variability is another important factor that must be controlled. For instance, risks come with variability across lots of fetal bovine serum (FBS), which sometimes is used during production of adherent cell lines.

Each platform has its own advantages and disadvantages. General considerations include requirements for producing and storing helper viruses and for demonstrating genetic stability of cell lines and viruses. But the most important parameter for choosing a platform is volumetric productivity. All the above systems demonstrate comparable vector titers, usually ranging from 5 × 10¹³ to 2.4 × 10¹⁴ vector genomes per liter (vg/L).

I consider the main advantages of the HEK293–plasmid transfection platform to be its flexibility for different AAV vector serotypes and its superior capacity for in vitro infectivity and in vivo efficacy. Its main disadvantage is its dependence on plasmids, which can be quite expensive. Expenditure on plasmids represents the highest proportion of the cost of goods in the production of AAV vectors. Usually, that percentage is in the double-digit range for all three plasmids.

Sometimes that limitation can be overcome by in-house production of the corresponding plasmids. That requires profound know-how of the production process and the corresponding analytical panel for in-process control and release testing, both for the Escherichia coli cell banks and for the plasmids themselves. Particularly difficult is the fact that process units such as alkaline lysis do not scale easily. Manufacturing plasmids according to GMP also comes with its own requirements: appropriate production rooms with qualified cleanroom concepts, production and storage of E. coli cell banks under GMP conditions, qualified equipment and personnel, and so forth. However, a cost–benefit analysis might show that it makes sense to produce plasmids in house despite their high costs and long delivery times.

How do different cell culture formats influence process scalability? So far as I know, all of the AAV production platforms I mentioned that depend on suspension cell culture (e.g., Sf9 and HEK293 cells) have been brought to scales up to 2,000 L in single-use stirred-tanked reactors.

Production processes that are based on anchorage-dependent cell lines can follow one of two paths. To increase cell-culture surface area, multiple cell culture systems can be run in parallel — a process known as scaling out. Available systems include CellFactory vessels (Thermo Fisher Scientific) and CellSTACK culture chambers (Corning). In the past few years, novel single-use, fixed-bed bioreactors (e.g., iCELLis systems from Pall) have become commercially available to ease scale-up of adherent cell-culture systems. Nevertheless, companies might stick to a scale-out approach if they have products in late life-cycle phases or if they want to prevent complications that can come with switching to suspension cell cultures or fixed-bed reactors due to little experience and lack of know-how in relation to a new cell culture system. Those can include low yields, uncontrollable critical process parameters, and reductions in product quality.

How much attention should developers give to scale-up during early (pre)clinical phases, and what considerations are most important at that stage? Takeda’s gene therapy process development team works closely with the therapy developer or research team that has designed the vector (capsid and transgene) being produced. Close collaboration helps to align timelines and demands. We on the chemistry, manufacturing, and controls (CMC) team also are involved in selecting a lead candidate, taking into account knowledge gained during preparation of preclinical material.

Our research and CMC teams agree that good transduction efficiency of a vector and high expression of a transgene can lower dose requirements significantly. That can diminish the overall demands upon a product and increase its manufacturability.

What else could improve AAV scale-up? The biopharmaceutical industry needs a mixture of new technologies and improved understanding. There is need for excellently educated personnel who have a high level of biological understanding in combination with profound process knowledge; such qualities are key prerequisites for working in biotechnology. And the more experience that a team has gained developing new processes, optimizing existing ones, and troubleshooting GMP operations, then the faster and more effectively it will complete new projects.

As I mentioned, it is critically important that a team working at laboratory scales to optimize parameter settings for a new viral-vector product knows the limitations of its large-scale equipment. And I want to emphasize that process development teams must never forget the importance of know-how in the area of analytical method development. Perfect coordination of process and analytical development makes a team effective.

In recent years, another knowledge component has been added to the needs of process development teams: expertise in process modeling. That involves digitally mapping and simulating different activities to determine optimal setpoints for process parameters. Expertise in this area can help a team not only to plan experiments more effectively, but also to find solutions to questions that cannot be answered experimentally, especially if the investigated issue occurs only at full scale.

What stages of manufacturing-scale AAV production can be optimized most? When and how would you begin to make such adjustments? When a new project enters process development, it is important to balance speed, output, and quality. Some process adjustments will need to be made; otherwise, you might be unable to produce enough material for preclinical or clinical studies, or product quality might not meet specifications. During these first mandatory optimizations, if time permits, further adjustments can be made to increase yield.

The farther that a program moves toward product launch, the less that regulatory authorities will welcome process adjustments. Nevertheless, it might still be necessary to carry out further optimizations in phase 3 clinical studies. Around the time of process performance qualification, a process clearly should be optimized, and no further changes should be made.

What other advice do you have for scientists who are seeking to maximize the scalability of their AAV processes — and ultimately, their yields? Just referring to our AAV process toolbox here at Takeda, and thinking only about yield improvement, my first thought, of course, goes to upstream production. It is always tempting and interesting to fine-tune setpoints or to try adding media. However, downstream activities also must be considered. Depending on the vector construct, very low recovery rates might be observed for one or the other purification step. Steps exhibiting the greatest loss of capsids are the low-hanging fruits with the highest probability of yield improvement. Those should be optimized first.

References
1 Clément N, Grieger JC. Manufacturing of Recombinant Adeno-Associated Viral Vectors for Clinical Trials. Molec. Ther. Meth. Clin. Dev. 3(16002) 2016; https://doi.org/10.1038/mtm.2016.2.

2 Srivastava A, et al. Manufacturing Challenges and Rational Formulation Development for AAV Viral Vectors. J. Pharm. Sci. 110(7) 2021: 2609–2624; https://doi.org/10.1016/j.xphs.2021.03.024.

3 Rodrigues GA, et al. Pharmaceutical Development of AAV-Based Gene Therapy Products for the Eye. Pharm. Res. 36(2) 2019: 1–20; https://doi.org/10.1007/s11095-018-2554-7.

4 Castle MJ. Adeno-Associated Virus Vectors: Design and Delivery. Humana Press: Totowa, NJ, 2019.

5 Mével M, et al. Chemical Modification of the Adeno-Associated Virus Capsid to Improve Gene Delivery. Chem. Sci. 11(4) 2020: 1122–1131; https://doi.org/10.1039/C9SC04189C.

Suggestions for Further Reading
Barlow JF, et al. Insights on Successful Gene Therapy Manufacturing and Commercialization. Cell Culture Dish 14 December 2020; https://cellculturedish.com/insights-on-successful-gene-therapy-manufacturing-and-commercialization.

Mertens O. AAV Vector Production: State of the Art Developments and Remaining Challenges. Cell Gene Ther. Insights 2(5) 2016: 521–551; https://doi.org/10.18609/cgti.2016.067.

Zhao Z, Anselmo AC, Mitragotri S. Viral Vector–Based Gene Therapies in the Clinic. Bioeng. Transl. Med. 7(1) 2022: e10258; https://dx.doi.org/10.1002/btm2.10258.

Brian Gazaille is associate editor of BioProcess International, part of Informa Connect, PO Box 70, Dexter, OR 97431; brian.gazaille@informa.com; 1-212-600-3594.