Status and future of recombinant adeno‐associated virus vector manufacturing

Abstract

Sixty years of adeno‐associated virus (AAV) research illustrates a trajectory marked by basic science exploration, iterative innovation, persistent challenges, a number of clinical setbacks, as well as commercial therapeutic triumphs. This continual evolution has led to recombinant AAV (rAAV) becoming a cornerstone of modern gene therapy. Significant advancements in molecular design, process development, and manufacturing have been made over the past three decades; these improvements are expected to significantly improve the safety, efficacy, and economics of rAAV gene therapies. Beyond rare disease, rAAV vectors have the potential to be used in prevalent conditions such as arthritis, heart failure, diabetes, Alzheimer's disease, and Parkinson's disease. Meeting the vector demands of these disease treatments will require continued innovations in rAAV manufacturing, including further improvements in process optimization and molecular engineering. In addition, the adoption of process intensification and automation strategies, pioneered in other biologics such as monoclonal antibody manufacturing, should prove pivotal in advancing the scale, robustness, and efficiency of rAAV production.

1. INTRODUCTION

Gene therapy is a revolutionary approach to medicine aimed at treating diseases by replacing, repressing, or editing faulty genes. The practice has been utilized to address a wide range of diseases, including but not limited to neurological disorders, cancer, viral infections, inherited blindness, cardiovascular disease, and metabolic disorders. ¹ Gene therapy can be implemented through two primary approaches: ex vivo and in vivo. In ex vivo applications, cells are taken from the patient, modified to incorporate the therapeutic gene, and then reintroduced to the patient. For in vivo gene therapy, the genetic therapeutic is directly delivered to the target tissue inside the patient's body. ¹ , ² Both ex vivo and in vivo have their unique strengths and limitations, depending on various factors such as the specific disease to be treated, the target cells, and the overall treatment strategy. Whereas ex vivo gene therapy approaches are typically employed for indications such as CAR‐T cell therapies and hematopoietic stem cell‐based treatments for sickle cell disease, in vivo gene therapies are generally reserved for delivery of genetic material to post‐mitotic or difficult‐to‐access tissues, including neurons, muscle, liver, and retina. ³

Introduction of therapeutic genes into target cells can be achieved by two delivery mechanisms: viral and non‐viral methods. Some examples of non‐viral delivery methods are naked plasmid delivery by electroporation, microinjection, or ultrasound, lipid nanoparticles (LNP), extracellular vesicles (EV), and cationic polymers. ⁴ , ⁵ Non‐viral gene delivery approaches offer advantages compared to viral delivery, such as lower immunogenicity and the potential for repeat dosing with reduced neutralizing antibody responses. However, non‐viral delivery methods generally suffer from lower transduction efficiency and limited ability to achieve precise tissue or cell‐type targeting in vivo. Viral methods exploit the natural ability of viruses to infect host cells to deliver genetic material. While both viral and non‐viral delivery methods have their unique advantages and challenges, viral delivery methods are generally considered superior in terms of efficiency, specificity, and ability to provide stable and long‐term gene expression. ²

The viral vectors used for gene therapy have been derived from a wide range of virus families, including adenoviruses, adeno‐associated viruses (AAV), herpes simplex viruses, retroviruses (including lentiviruses), alphaviruses, flaviviruses, measles viruses, rhabdoviruses, Newcastle disease virus, poxviruses, picornaviruses, reoviruses, and polyomaviruses. ⁶ , ⁷ Despite the diversity of viral platforms explored, adenovirus serotype 5 (Ad5), lentiviral, and recombinant adeno‐associated virus (rAAV) vectors have emerged as the most widely used platforms in clinical development due to their transduction efficiency, safety profile, and manufacturability. ² Adenoviral vectors are commonly used either as oncolytic agents, where their lytic replication and immunostimulatory properties are leveraged for cancer therapy, or as helper‐dependent (“gutless”) vectors that enable high‐level transient expression with a large payload capacity. ² , ⁷ Lentiviral vectors have become the preferred choice for ex vivo gene therapy, including CAR‐T cell and hematopoietic stem cell applications, due to their ability to stably integrate into the genome, allowing for durable modification of dividing cells. ² , ⁶ , ⁷ In contrast, rAAV vectors have become the dominant platform for in vivo gene therapy, particularly for targeting post‐mitotic tissues, owing to their favorable safety profile and capacity for long‐term transgene expression. ¹ , ² , ⁷ , ⁸

The use of rAAV in gene therapy is a significant advancement in the treatment of genetic disorders and other diseases. Some of the distinct features of rAAV are its relative stability, non‐pathogenic nature, durable episomal transgene expression, low likelihood of replication competence, and modifiable capsids. ¹ , ² , ⁸ Following cellular entry, the rAAV single‐stranded DNA transgenes undergo second‐strand synthesis and form stable, non‐integrating, circular DNA structures (episomes) within the nucleus. ⁶ The episomes are not lost during cell division and can facilitate long‐lasting therapeutic gene expression, but are diluted in actively dividing cells. Therefore, rAAV treatments are usually applied to static tissues, like muscle, liver, or the central nervous system. Its persistence as extrachromosomal elements minimizes the risk associated with random integration into the host genome, thereby reducing the potential for insertional mutagenesis and the activation of oncogenes. ⁸

Additionally, rAAV is highly amenable to transgene and capsid design. Capsid engineering, through rational design, directed evolution, and machine‐learning‐guided approaches, has produced novel AAV variants with enhanced tissue tropism, improved intracellular trafficking, reduced off‐target transduction, and partial evasion of pre‐existing neutralizing antibodies. ⁷ Optimized rAAV transgene designs can be achieved through molecular enhancements, such as cell type‐specific promoters, codon optimization, and sequence evaluation of packaged transgenes. Together, these attributes are enabling lower doses, reduced immunogenicity, and the continued establishment of rAAV as a major pillar of modern gene therapy development.

Manufacturing of rAAV is at an inflection point, driven by innovation, clinical demand, expanding target indications, and rapid advances in molecular design and bioprocess technology. Significant bioprocess innovations have driven substantial improvements in both upstream yield and downstream recovery, as well as purity in rAAV. The recent technological advancements are expected to lead to an increase in vector yields per manufacturing batch, making rAAV‐based gene therapies more accessible to larger patient populations. Recent advances in vector design, leading to improved delivery efficiency and lower rAAV doses per patient, together with process improvements and enhanced manufacturing scalability, are expected to increase therapeutic access and improve the affordability of rAAV‐based gene therapies. In this work, we review recent improvements in rAAV production processes and discuss future directions for further advancing manufacturing efficiency, scalability, and cost‐effectiveness.

2. BASICS OF rAAV VECTORS

Wild‐type adeno‐associated virus (wtAAV) (20–25 nm in diameter) is a non‐enveloped virus containing a single‐stranded DNA payload, and it was originally discovered as a contaminant in adenovirus preparations. ⁹ The genome of wtAAV is approximately 4.7 kb in size and contains two main open reading frames (ORFs), rep and cap, flanked by two 145 base inverted terminal repeats (ITRs). ¹⁰ The rep gene products are non‐structural proteins that facilitate viral replication and integration, while the cap gene products are structural proteins that form the protective capsid, enabling the virus to deliver its genetic material into the host cell. ¹¹ , ¹² The capsid of AAV has an icosahedral structure and contains 60 proteins in the approximate molar ratio of 1:1:10 (VP1:VP2:VP3). ¹³ , ¹⁴ The rep gene uses the p5 promoter to express Rep78 (78 kDa) and its splice variant Rep68 (68 kDa) protein and p19 promoter to express Rep 52 (52 kDa) and its splice variant Rep 40 (40 kDa) proteins. For the cap gene, the p40 promoter drives expression of the viral proteins VP1, VP2, and VP3 at 87, 72, 62 kDa respectively. ¹⁵ There are at least 13 different natural serotypes of wtAAV, which are characterized by variations of viral proteins and ITRs, and they have been isolated primarily from human and non‐human primates. ¹⁶ Since wtAAV is a Dependovirus within the Parvoviridae family, it cannot replicate independently and therefore relies on helper viruses, such as Adenovirus or HSV, for transcription and replication. ¹⁷

rAAV is a replication‐incompetent form of wtAAV in which the wild‐type protein‐coding sequences have been replaced with a therapeutic gene. ¹⁸ For traditional gene replacement, this is done by substituting the rep and cap genes in the wtAAV with a therapeutic transgene, a promoter and polyadenylation signal sequence between the ITRs to form the rAAV (Figure 1). Traditional rAAV vectors package as a single‐stranded genome (ssDNA), which must be converted into double‐stranded DNA (dsDNA) in the nucleus before transcription can begin. This conversion step, which depends on host‐cell DNA polymerases, can delay the onset of transgene expression. To bypass this bottleneck, self‐complementary AAV (scAAV) was developed by modifications of ITRs to generate double‐stranded genome, and this led to an increase in transduction efficiency in vitro and in vivo. ¹⁹ , ²⁰ However, this advantage comes at the cost of approximately halving the effective packaging capacity of the vector (≈2.3–2.5 kb vs. ≈4.7 kb). ²¹

FIGURE 1.

Wild type AAV (wtAAV), rAAV, and rAAV transduction.

3. CLINICAL USE OF rAAV

rAAV has been used in human gene therapy for approximately 30 years, evolving from early proof‐of‐concept studies in the early 1990s through periods of significant technical and immunological challenges to multiple landmark clinical successes. ⁸ Early clinical gene transfer experiments established the safety of rAAV but highlighted limitations, such as immune responses and variable efficacy. ² Continued advances in vector design, capsid engineering, and manufacturing ultimately enabled transformative therapies, underscoring the potential of rAAV as a durable platform for treating genetic, neurological, and other diseases. ²² The first human gene therapy trials using rAAV vectors were conducted in 100 cystic fibrosis (CF) patients enrolled in five separate trials of rAAV2‐CFTR (AAV2 encoding the CF transmembrane conductance regulator protein) administered via nasal, endobronchial, maxillary sinus, and aerosol delivery. ²³ Although these studies demonstrated safety, they failed to deliver therapeutic benefit, primarily because of limited airway transduction, epithelial turnover, and immunity that prevented re‐dosing. These early setbacks dampened enthusiasm but provided an important foundation for later use of rAAV in humans. ²³

The first clinical trial for hemophilia, using AAV vector expressing human coagulation factor IX (FIX) for treating Hemophilia B administered intramuscularly (IM), began in 1999. ²⁴ Initial trials for hemophilia B (employing AAV2 capsids) were pioneering for both liver‐ and muscle‐directed gene transfer, but resulted in blocking of transduction by pre‐existing neutralizing antibodies (NAbs), a dose‐dependent rise in liver enzymes caused by an immune response to the AAV2 capsid. ²⁴ Subsequent clinical studies initiated in 2010 utilized AAV8 capsid, which had shown superior liver transduction compared with AAV2, and reduced prevalence of pre‐existing NAbs in the human population. ²⁴ , ²⁵ In these studies, all subjects in the 2 × 10¹² vg/kg dose cohort achieved sustained, multi‐year expression of FIX in the range of 3%–7%, representing a clinically meaningful therapeutic level. This breakthrough result led to the first sustained therapeutic expression from liver gene transfer in human clinical trials. ²⁴ , ²⁵

Following the observed clinical trial success of Nathwani AAV8/FIX trial, ²⁵ there were several notable setbacks. While Glybera, the first commercial rAAV product, was authorized in the EU in 2012, it was taken off the market after 5 years due to the high price, reimbursement challenges, and few patients treated. ²⁶ Then, in 2015, Celladon reported that their lead product, Mydicar (AAV1‐expressing sarcoplasmic reticulum calcium ATPase (SERCA2a)), failed to meet its primary and secondary endpoints in the phase IIb CUPID2 trial. ²⁷ These failures highlighted critical challenges for rAAV delivery and underscored the need for more potent capsids, improved delivery strategies, and better patient selection to drive reduced costs and achieve positive commercial outcomes. Importantly, this experience helped recalibrate expectations for rAAV gene therapy and informed the subsequent shift toward capsid engineering and dose‐efficient vector design.

A groundbreaking success was achieved through rAAV mediated delivery of the gene for retinal pigment epithelium‐specific 65 kDa protein (RPE65) in patients suffering from retinitis pigmentosa. Three patients were administered AAV2.hRPE65v2 in the initial safety studies and each showed an improvement in measures of retinal function as indicated by subjective tests of visual acuity. ²⁸ Additional studies were conducted to assess the safety and efficacy of rAAV2 carrying the RPE65 gene in 15 patients at four dose levels and two different injection strategies. ²⁹ The results showed that no systemic toxicity was detected and visual function improved in all patients to varying degrees. ²⁹ Long‐term follow‐up confirmed sustained expression and an acceptable safety profile, ultimately leading to the regulatory approval of voretigene neparvovec (Luxturna) in 2017 as the first FDA‐approved in vivo gene therapy. The FDA approval firmly established rAAV as a clinically viable platform for treating inherited retinal diseases.

The first‐in‐human trial of AAV9‐mediated SMN1 gene replacement conducted by AveXis (later acquired by Novartis) represents one of the most transformative milestones in rAAV gene therapy. Spinal muscular atrophy (SMA) type 1 is a severe, often fatal neuromuscular disorder caused by loss of the SMN1 gene, leading to progressive motor neuron degeneration. ³⁰ AveXis developed AVXS‐101, an intravenously delivered AAV9 vector encoding a functional SMN1 transgene, leveraging the unique ability of AAV9 to cross the blood–brain barrier and efficiently transduce motor neurons following systemic administration. The results demonstrated that in patients with SMA1, a single intravenous infusion of rAAV containing DNA coding for SMN resulted in longer survival, superior achievement of motor milestones, and better motor function than in historical cohorts. ³⁰ Many patients achieved previously unattainable motor milestones such as sitting unassisted and, in some cases, standing or walking. The durability of SMN expression and sustained clinical benefit observed in long‐term follow‐up provided compelling evidence that systemic AAV gene therapy could safely and effectively treat a devastating neurodegenerative disease, ultimately leading to FDA approval of onasemnogene abeparvovec (Zolgensma). A summary of the history of rAAV, from discovery to first‐in‐human use, first FDA approval, and high‐yield processes, is shown in Figure 2.

FIGURE 2.

History of rAAV from discovery to first in human use to first FDA approval to high yield processes.

Gene therapy using rAAV has continued to see increased application for the treatment of multiple transformative and fatal conditions, with 240 clinical trials covering major disease areas such as ocular, neuromuscular, neurological, cardiovascular, metabolic, hematological, and cancer. ¹⁶ , ³¹ , ³² The approved rAAV products, as well as the serotype, the indication, and the dose, are summarized in Table 1. ³³ , ³⁴ While these commercial approvals have been meaningful for the field, two rAAV products have been withdrawn from the market, largely due to decreased demand impacted by the high cost of the drugs. A major component of these higher prices is the use of older rAAV manufacturing technologies that lead to a limited number of patients that can be treated per batch.

TABLE 1.

Approved products used in rAAV gene therapy.

Year	Product	Indication	Production method	Serotype/target	Dose
2012	Glybera a (uniQure)	Lipoprotein lipase deficiency	Infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors	AAV1/skeletal muscle	1 × 1012 vg/kg
2017	Luxturna (Spark therapeutics)	Retinal dystrophy (RPE 65)	Transient transfection of HEK 293	AAV2/eye	1.5 × 1011 vg/eye
2019	Zolgensma (Novartis)	Spinal Muscular Atrophy (SMA)	Transient transfection of HEK 293	AAV9/CNS	1.1 × 1014 vg/kg
2022	Upstaza b (PTC therapeutics)	Aromatic L‐Aino acid decarboxylase deficiency (AADC)	Transient transfection of HEK 293	AAV2/brain	1.8 × 1011 vg
2022	Hemgenix (CSL Behring)	Hemophilia B	Infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors	AAV5/liver	2 × 1013 vg/kg
2023	Elevidys (Sarepta)	Duchenne muscular dystrophy (DMD)	Transient transfection of HEK 293	AAVrh74/Skeletal muscle	1.3 × 1014 vg/kg
2023	Roctavian (BioMarin)	Hemophilia A	Infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors	AAV5/liver	6 × 1013 vg/kg
2024	Beqvez a (Pfizer)	Hemophilia B	Transient transfection of HEK 293	AAVRh74var/liver	5 × 1011 vg/kg
2024	Kebilidi b (PTC Therapeutics)	Aromatic L amino acid decarboxylase (AADC) deficiency	Transient transfection of HEK 293	AAV2/brain	1.8 × 1011 vg
2025	Itvisma (Novartis)	Spinal muscular atrophy (SMA)	Transient transfection of HEK 293	AAV9/CNS	1.2 × 1014 vg
2026	Otarmeni (Regeneron)	Sensorineural hearing loss (>90 dB HL)	Transient transfection of HEK 293	AAV1/inner hair cells	7.2 × 1012 vg

^a Product was taken off the market.

^b Same product approved in different regions (EMA 2022, FDA 2024).

4. HISTORY OF rAAV MANUFACTURING AND PRODUCTION PLATFORMS

4.1. Production process for rAAV

Cloning of rAAV sequence into bacterial plasmid and rescue of rAAV into human cells was performed in 1982 by Samulski et al. ³⁵ and the expression of foreign genes in mammalian cells was performed by Hermonat and Muzyczka. ³⁶ The production process originally developed for rAAV used a plasmid containing therapeutic genes flanked by ITRs, rep and cap genes from wtAAV or a second plasmid and an adenovirus for helper functions. ³⁷ After about four decades of process development, the four main methods currently being used for scalable rAAV production are transient plasmid transfection of HEK293 cells, infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors, infection of HEK293 (or BHK21) cells with recombinant herpes simplex virus (rHSV), and the use of stable producer cell lines (Figure 3, Table 2). ³⁸

FIGURE 3.

The four main production processes: transient plasmid transfection of HEK293 cells, infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors, the use of stable producer cell lines, and infection of HEK293 (or BHK21) cells with recombinant herpes simplex virus (rHSV).

TABLE 2.

Summary of the production processes for rAAV production.

Production method	Transient plasmid transfection of HEK293 cells	Infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors	Stable producer cell lines	Infection of HEK293 (or BHK21) cells with recombinant herpes simplex virus (rHSV)
Cell line	HEK293 or HEK293T (suspension or adherent)	Sf9	HeLa, HEK293, A549	HEK293 or BHK
Cell line origin	Human	Insect	Human	BHK (Hamster)
Transgene	Plasmid	Baculovirus expression vector (BEV)	Integrated in cell line	rHSV
Helper	Plasmid	Baculovirus expression vector (BEV)	Integrated in cell line	rHSVs (transgene and RepCap)
RepCap	Plasmid	Baculovirus expression vector (BEV)	Integrated in cell line	rHSV
Method of production	Transfection	Infection	Induction	Infection
Yield (vg/L)	+++	+++	++	+++
Production system	Suspension, adherent	Suspension	Suspension	Suspension
Advantages	Cell are human origin, simplicity, flexibility, fast to produce	High yield, efficient large scale production	Cell are human origin, efficient large scale production, low cost	High yield, efficient large‐scale production
Disadvantages	High cost, low scalability	Non‐mammalian origin, rBV stability, ratio of VPs being different	Time consuming, limited flexibility	Risk of rHSV, contamination, rHSV stability, production of high‐titer rHSV
Host cell	HEK 293 (T) 1–3 plasmids	Sf9	HEK293, A549, HeLa	HEK293 or BHK
Input	Different combinations for 3, 2 or 1 plasmid	Different combinations for 1 Bac and 2 Bac system	With or without wtAd5 cell lines	rHSVs
Approved products (FDA/EU)	Yes	Yes	No	No

4.2. Adherent versus suspension systems

The upstream production process is performed with either adherent or suspension cells. In adherent cultivation systems, the cells are attached to a matrix, typically by using roller bottles or cell stacks for small scale productions. ³⁹ , ⁴⁰ Large scale adherent bioreactors used include the iCELLis fixed‐bed technology with iCELLis 500 (culture area up to 500 m²) and Univercells Scale‐X (culture up to 600m²). ⁴¹ Suspension cell culture involves the use of the production cells in an agitated vessel where the cells are free‐floating, and this is performed in shake flasks for small‐scale productions (typically <2 L), in rocking bioreactors for intermediate scale (2–25 L), or stirred tank bioreactors for large scale manufacturing (>50 L). Adherent cell culture production of rAAV has typically resulted in higher cell‐specific productivities of >100,000 vg/cell compared to suspension cell cultures with productivities >50,000 vg/cell. However, it is challenging to generate the amount of vector required for clinical or commercial use in adherent cell culture due to the large surface area required in adherent bioreactors. ⁴⁰ , ⁴¹ Suspension‐based bioreactors are more scalable compared to adherent systems, and rAAV production has been scaled up to 500, 1000, and 2000 L in single‐use bioreactors. ⁴⁰ , ⁴² , ⁴³

Adaptation of adherent cells to suspension systems represents a significant advancement in the production of rAAV vectors, enabling the capability to scale up to higher bioreactor volumes. ⁴⁴ , ⁴⁵ Greiger et al. ⁴⁶ adapted an adherent HEK293 to animal‐component free suspension conditions in shake flasks and rocking bioreactors to allow for rapid and scalable rAAV production. The triple transfection method was applied to the suspension HEK293 cell line and a titer greater than 1 × 10¹⁴ vg/L, corresponding to 1 × 10⁵ vg/cell. ⁴⁷ The rAAV vectors generated by suspension cells were demonstrated to be equivalent and may have higher potency compared to vectors generated by adherent cells. ⁴⁷ High‐purity vectors were generated with AAV serotypes 1–6, 8, 9 after optimization of process variables such as media and transfection reagent, and process conditions. ⁴⁶ These advancements in the ability to produce rAAV using suspension systems provide opportunities to implement additional advancements, such as the introduction of fed‐batch cultures and perfusion systems, that could be used to increase the cell density at transfection and vector yield per liter of culture.

4.3. Transient transfection of HEK 293 cells

Transient transfection with plasmid DNA remains the most widely used method for producing rAAV in suspension or adherent cultures, although the field has increasingly shifted toward the use of suspension systems in single‐use bioreactors. This method has been applied to five of the nine approved products for rAAV production (Table 1). The popularity of this method for rAAV production is due to the relatively rapid development speed and the simplicity of designing and introducing plasmids with transfection reagents. ¹³ , ¹⁶ This is particularly useful when speed to the clinic is critical in development timelines, where potential rAAV candidate products can be quickly produced and tested in animal models and in early‐phase clinical trials. However, the cost of goods, particularly plasmids, and the challenge of scaling the transfection process to higher volumes (>2000 L for suspension HEK293) have been drawbacks of this platform. ³⁸ , ³⁹

The transfection method for rAAV production requires the efficient delivery of one, two, or three plasmids (or plasmid DNA alternatives) to HEK293 cells. Plasmid DNA (pDNA) required for rAAV production is traditionally manufactured in Escherichia coli using microbial fermentation followed by downstream purification. ⁴⁸ As an alternative to fermentation‐based methods, novel, in vitro produced, synthetic DNA manufacturing technology has been developed to produce DNA constructs without using bacterial hosts. ⁴⁹ , ⁵⁰ Transfection efficiency is a notable source of manufacturing variability, as effective production of rAAV requires cotransfection of the rep/cap genes, virus helper genes, and the GOI transgene. ⁵¹ Early work in the field used calcium phosphate to form pDNA complexes, ¹⁷ but calcium phosphate has been largely replaced with more robust transfection reagents, such as polyethyleneimine (PEI), Mirus Bio AAViator, and FectoVIR. ⁵² While these reagents have improved the percentage of transfected cells, there is typically a subset of cells in the culture that are not fully transfected, leading to lower productivity and the production of empty particles.

4.4. Infection of Spodoptera frugiperda (Sf9) with baculovirus expression vectors

In the baculovirus/Sf9 (Bac/Sf9) manufacturing system, rAAV is produced in insect cells (fall armyworm, Spodoptera frugiperda). The first approved rAAV product (UniQure's Glybera) and three out of the nine approved products were manufactured using the Bac/Sf9 system (Table 1). The ITR‐flanked GOI, rep, and cap genes are delivered to the Sf9 cells, usually in suspension and serum‐free medium, using engineered baculovirus expression vectors (BEVs), where the baculovirus provides the needed helper functions. The volume of inoculum required is small due to the low multiplicity of infection (MOI) of the BEVs needed for rAAV production. ³⁸ Some of the advantages of this system include high vector yields, scalability, improved full‐to‐empty capsid ratios, ⁵³ and the lower manufacturing costs due to the use of baculovirus infectious insect cells (BIICs). The disadvantages of this system include the molecular optimization needed to achieve the correct AAV viral protein ratios for specific capsids. If the viral capsid proteins are not properly tuned to achieve VP1:VP2:VP3 at a ratio of 1:1:10, the resulting rAAV produced by this method may have lower potency. Another drawback of using BEVs is the genetic instability created by the removal of parts of the genome during replication, which can result in mutant rAAV vector. ¹⁶ Also, the Bac/Sf9 system generates rAAV vectors which contain unique impurities not found in rAAV produced from mammalian cells, including baculovirus proteins, baculovirus DNA, insect cell‐specific post‐translational modifications, and glycosylation patterns specific to insect cells. ⁵⁴ , ⁵⁵ While insect cells support post‐translational modification of viral proteins, the structures and complexities of these modifications differ from those produced in mammalian systems. ⁵³ , ⁵⁶ Evidence in the literature regarding differences in the potency of rAAV produced in insect cells versus mammalian systems is mixed; Giles et al. ⁵⁶ report differences in potency, whereas Lu et al. ⁵³ observe no significant differences in potency or clinical outcomes. Specialized characterization of rAAV for baculovirus‐related impurities such as residual DNA, proteins, infectious baculovirus particles also needs to be performed. ⁵⁴ However, baculovirus DNA has a low likelihood of transcriptional activity in human cells and therefore results in a lower risk of residual host‐cell DNA (hcDNA) in the final drug substance compared with mammalian production systems. ⁵⁵ In addition, rAAV vectors only contain a very low level of sporadic glycosylation, such that changes in the glycosylation patterns are likely not detrimental. Lastly, since baculovirus is a labile enveloped virus, it is generally straightforward to implement viral removal steps to clear it from the final rAAV product. ⁵⁷

4.5. Stable producer cell lines

A “producer cell line” refers to a cell line with some or all of the genetic elements required for assembling a rAAV vector that are stably integrated into the genome, including the “rep” and “cap” genes for viral replication and packaging, as well as the GOI flanked by ITRs. These cells often utilize helper functions from adenovirus to facilitate efficient rAAV production, either by transient infection with an Ad helper virus or by having the Ad helper genes stably integrated within the cell line. If adenovirus is used to provide helper functions, the removal of residual adenovirus contaminants in the final product must be demonstrated. When generating stable producer cell lines, a drug resistance gene is often included in the construct to select for cells that have successfully integrated the desired genes. Common cell lines like HEK293, A549, and HeLa are often used to generate producer cell lines, with specific subclones like C12, H44, and B50 sometimes being preferred. ⁵⁴

A stable producer cell line (PCL) derived from HeLa for generating rAAV vectors was first described by Clark et al., ⁵⁸ where the cells were integrated with AAV rep‐cap genes, which could produce rAAV upon infection with Adenovirus 5. The authors achieved this by creating neomycin‐resistant HeLa PCL containing both the AAV rep‐cap genes and rAAV vector that produced infectious rAAV particles when infected with adenovirus. ⁵⁸ The production process with HeLa PCL was used to manufacture rAAV for clinical use and was demonstrated to be scalable. ⁵⁹ , ⁶⁰ More recently, Ultragenyx developed a novel AAV perfusion‐enhanced expression process (APEX) to achieve higher cell densities and rAAV titers approaching 1 × 10¹⁵ vg/L across multiple serotypes. ⁶¹

CVEC (acquired by Cytiva) developed the ELEVECTA™ producer cell line by stably integrating the Ad helper, rep, cap, and GOI genes into the host cell's genome with an inducible system. Two ELEVECTA™ alpha cell lines, modified clonal HEK293 or a cell line derived from human amniocytes (CAP™) with helper and rep genes already integrated, are used to generate the producer cells after stable integration of capsid and GOI genes. ⁶² Stable cell lines have been adapted to suspension cultures for scalable rAAV production at 2000 L scale. ³⁸ , ⁵⁴ Some advantages of a stable cell line include scalable, efficient production without the need for plasmids. However, generating a stable producer cell line is time‐consuming, and achieving stable expression can be challenging. Since a new cell line must be developed for every product, rAAV drug developers must include additional time for cloning, evaluation, Master Cell Bank (MCB) manufacturing, and MCB release testing. ¹⁶ , ³⁸

4.6. Infection of HEK293 (or BHK21) cells with recombinant herpes simplex virus

This method involves the use of replication‐defective recombinant herpes simplex virus (rHSV) to infect HEK293 or baby hamster kidney cells (BHK21). Two distinct rHSVs, one carrying the GOI flanked by ITR and the other carrying rep/cap genes, are used to infect HEK293 or BHK21 cells with rHSV, providing helper functions for rAAV production. The shuttle vectors, rHSV are typically produced in V27 cells (Vero cells stably transformed with HSV‐1 UL54 gene encoding ICP27). ³⁸ The advantages of using this method include the high yield that can be obtained and the ability to scale efficiently. ⁶³ The drawbacks of this approach include the risk of rHSV contamination and additional purification required to remove associated contaminants from the final rAAV products. ⁵¹ , ⁵⁴ The production of high‐titer, infectious rHSV stocks can be challenging because of several features inherent to HSV biology: (1) the production efficiency and the product safety profile are usually inversely correlated, and (2) HSV particles are highly sensitive to production and processing conditions (e.g., temperature, shear, solvents, and detergents) and can easily be inactivated during manipulation. ⁶³ Despite the difficulties, the rHSV system has been used to produce a number of clinical products. ⁶⁴ A summary of the four main production methods is shown in Table 2.

4.7. Novel production methods

Beyond the four major rAAV production platforms, several novel and emerging systems have been developed to improve yield, scalability, and process control. One such innovation is the tetracycline‐enabled self‐silencing adenovirus (TESSA) system, which employs a modified, self‐regulating adenovirus to provide helper functions without generating contaminating adenovirus particles. ¹⁶ By inserting a tetracycline repressor binding site into the adenoviral major late promoter, the helper virus self‐inhibits its replication while tightly activating rep expression during AAV genome replication and packaging. ⁶⁵ This approach has been shown to increase rAAV yields by up to 30‐fold compared with helper‐free plasmid systems and has been successfully scaled to 50 and 200 L bioreactors, achieving titers exceeding 7 × 10¹⁴ vg/L and drug substance yields greater than 1 × 10¹⁷ genome copies from a 200 L batch. ⁶⁶ Another infection‐based strategy leverages HSV‐1, whose highly processive polymerase supports efficient rAAV genome replication and packaging. ⁶⁷ By placing HSV‐1 polymerase under inducible control, researchers demonstrated significant increases in encapsidated rAAV titers and full particle content without a change in total capsids. ⁶⁷ The HSV system has also been adapted for rAAV production in CHO cells, yielding rAAV with infectivity, transduction efficiency, and biodistribution comparable to those of HEK293‐derived vectors. ⁶⁸ Saccharomyces cerevisiae has been shown to support AAV capsid formation and genome encapsidation, although yeast‐derived rAAV exhibits reduced infectivity and altered capsid composition. ⁶⁹ , ⁷⁰ In addition, microalgae may serve as an alternative expression host for rAAV production. ⁷¹ More recently, plant‐based expression systems have demonstrated the ability to assemble functional rAAV particles in plant tissue, achieving titers of 10¹² to 10¹⁴ vg per plant following downstream purification. ⁷² Together, these platforms highlight innovative directions for future rAAV manufacturing beyond conventional systems.

5. CURRENT STATUS OF rAAV MANUFACTURING

The production process for rAAV has seen significant advancements over the last few years, including improved vector design, cell line optimization, transfection reagents, and additives for enhanced yield, scalable production at larger bioreactor volumes, and innovations in vector purification and recovery. The most common production process, transient transfection in suspension HEK293 cells, is illustrated in Figure 4.

FIGURE 4.

Process flow diagram for the most common approach used for rAAV production, triple transfection with suspension HEK293 cells. In the case of producer cell lines, no transfection is performed. *Ultracentrifugation can be performed with either CsCl or iodixanol.

For stable producer cell line, the production process is very similar apart from the transfection step. The process typically utilizes HEK293 suspension cells, and many groups have developed their own version of this cell line ²⁸ , ⁴⁴ , ⁷³ and there are commercially available versions, such as the VPC cells. ⁷⁴ , ⁷⁵ The process begins with thawing a cell bank and passaging it through seed train expansion before going into the production vessels, typically at the 50, 200, 500, 1000 or 2000 L scale. ⁴³ , ⁷³ , ⁷⁶ For the transient transfection process, once the cells reach the required density, they are transfected with plasmid(s) (GOI, RepCap, Ad helper) with the help of a transfection reagent. The culture is then monitored for several days, and at harvest, cells are lysed with a detergent and treated with nuclease prior to clarification. The clarified lysate is taken through a TFF or directly loaded on the appropriate affinity capture column, and the eluted product is further polished through CsCl ultracentrifugation, iodixanol gradient, or AEX (anion exchange) purification. The full capsid product from the polishing step is buffer exchanged into the final formulation and taken through viral and sterile filtration steps to generate the drug product. ⁴⁴ , ⁷³ , ⁷⁶

5.1. Molecular improvements

For decades it was known that the production of rAAV in HEK293 cells requires the following genes: E1A, E1B, E2A, E4ORF6, VA RNA from Adenovirus (or similar HSV genes) for transcription and replication. ¹⁷ In HEK293 triple transfection, E1A and E1B are provided by HEK293 cells, and E2a, E4ORF6, and VA RNA are supplied in trans as Adhelper plasmid. However, newer work has further interrogated and advanced our understanding of AAV and the adenovirus genes needed for rAAV generation. One region that is included in various Adhelper plasmid iterations is the E2a promoter region, which overlaps with the adenoviral L4 transcription unit, encoding the 100K (L4‐100K) protein and regulatory proteins L4‐22K and L4‐33k. ⁷⁷ The adenovirus L4‐22k protein controls viral gene expression at the post‐translational level and regulates the accumulation of the L4‐33k protein, another critical viral regulator. ⁷⁸ Adsero et al. ⁷⁷ applied molecular techniques to demonstrate that adenoviral 22k protein is essential for rAAV in HEK293 cells and the 33k protein synergistically increases rAAV yield. In the production of rAAV using packaging cell lines, Su et al. ⁷⁹ demonstrated that adenovirus 22/33k protein is required for episomal amplification of integrated rep/cap genes. These insights have enabled the removal of excessive adenoviral sequences from helper plasmids, resulting in smaller Adhelper plasmids and ~2× improvement in vector genome productivity across multiple capsid serotypes, genome designs, and transfection platform, highlighting how continued basic research in our understanding of AAV can dramatically impact rAAV manufacturing. ⁷⁷ , ⁸⁰

Although HEK293 cells with E1A gene represent one of the best platforms for stable producer cell lines, constitutive expression of E1A gene, a rep transcription activator, leads to the expression of cytotoxic rep proteins. ⁸¹ Jalši et al. ⁸² addressed this limitation by developing an inducible HEK293‐based rAAV producer cell line in which Rep40 and Rep68 expression was placed under the control of two inducible promoters. ⁸² Vector yield (vg/L) and productivity (vg/cell/day) can be improved through cell line engineering by characterizing the transcriptomics of different production cell lines to identify potential gene targets for cell engineering. The mRNA expression profile of three HEK293 cell lines was evaluated by Pistek et al. to identify genes that correlate with increased productivity of rAAV. ⁸³ Some of the most significant findings show that genes related to endoplasmic reticulum, sterol metabolism, and biosynthesis have an impact on rAAV yield, and these genes could be targeted for improvements in vector yield and productivity. ⁸³

5.2. Better transgene design

Transgene design encompasses improvements to the internal vector genome, including ITRs, promoters, enhancers, and introns. The ITRs provide critical functions such as genome packaging, replication during vector production, and episomal stability within host cells following transduction. ³² The modification of ITRs in the D‐regions was previously shown to improve AAV replication compared to the wild type ITRs. ⁸⁴ , ⁸⁵ , ⁸⁶ The wild type ITRs contain CpG motifs (cytosine followed by guanine), and this can send signals to toll‐like receptor TLR‐9 and induce an inflammatory response. ¹⁶ Replacing the wild‐type ITR with CpG‐free ITR did not affect vector genome encapsidation, transgene expression, or potency but resulted in a reduced rAAV titer. ⁸⁷ Efforts in rAAV vector engineering evaluated reducing the CpG motif to minimize innate immune activation and improve transgene persistence. ¹⁶ , ⁸⁸ Faust et al., ⁸⁸ demonstrated that CpG‐depleted AAVrh32.33 vectors exhibited persistent transgene expression, evaded immune responses, and minimized infiltration of effector immune cells. It is important to ensure that improvements in transgene design do not compromise rAAV productivity, process robustness, or manufacturability.

Promoters play a critical role in protein expression, cell and tissue specificity, and duration of expression in rAAV production. ⁸⁹ The cytomegalovirus (CMV), chicken beta‐action (CBA), cytomegalovirus immediate enhancer/chicken β actin/rabbit β‐globin (CAG), and human elongation factor‐1α‐subunit (EF‐1α) have been commonly used to provide robust, long‐term expression in all cell types. ¹⁶ , ⁸⁹ , ⁹⁰ However, the use of strong ubiquitous promoters can lead to adverse effects resulting from protein overexpression, often exceeding physiological levels by more than 100‐fold. Cell or tissue‐specific promoters can precisely regulate localized gene expression and therefore minimize off‐target effects and improve efficacy and safety. ⁹⁰ Some examples of cell or tissue‐specific promoters are Synapsin (hSyn), ⁹¹ targeting central nervous system neurons, thyroxine‐binding globulin (TBG), ⁹² targeting the liver, and muscle creatine kinase family (MCK), targeting the skeletal muscle. ⁹³ Recently, Wang et al. ⁹⁴ used deep learning models trained on single‐cell chromatin accessibility data from human retinas to generate synthetic AAV promoters targeting individual retinal cell types. Histological analysis of transduced retinas showed >90% cell‐type specificity for retinal ganglion or horizontal cells, demonstrating the potential of deep learning‐designed de novo AAV promoters to precisely target specific cell types and broadly advance promoter development for gene therapies. ⁹⁴ Examples of technology providers include Synpromics (now part of AskBio), which develops data‐driven, tissue‐specific synthetic promoter elements, and Form Bio, which offers integrated computational platforms to support gene therapy vector design, sequence optimization, and analytical workflows.

The regulatory elements within the rAAV expression cassette work together to control where, when, and how strongly a transgene is expressed, and small changes in these elements can dramatically alter vector performance. Promoter activity can be further tuned by enhancers and introns, which boost transcription and mRNA processing. ⁹⁵ The woodchuck hepatitis virus post‐transcriptional regulatory element (WPRE) is often used in rAAV cassettes to improve viral titer and transduction efficiency by stabilizing mRNA and regulating transcriptional termination. ⁹⁰ Although the WPRE can boost expression and prevent long‐term silencing in combination with several promoters, the presence of an intron could mitigate its effectiveness in boosting transgene expression levels. ⁹⁵ When introns are positioned proximal to the promoter or within a 5′ untranslated exon, they can increase transcriptional efficiency, facilitate recruitment and processivity of RNA polymerase II, and promote efficient splicing, nuclear export, and translation of the mRNA. ⁹⁵ , ⁹⁶ At the 3′ end of the transgene cassette, the choice of polyadenylation signal (e.g., SV40, bGH, or synthetic polyA) influences mRNA stability, cleavage efficiency, and transcriptional read‐through, and is now routinely evaluated during preclinical optimization. ⁹⁷ The regulatory RNA motifs, such as miRNA target sites (to suppress expression in off‐target tissues or specific immune cell subsets), riboswitches, or other RNA stability elements, can be used to sharpen tissue specificity. ⁹⁸ Together, these regulatory elements—ITRs, promoters, enhancers, introns, polyA signals, and RNA motifs—form a modular toolbox that allows rAAV genomes to be finely engineered for efficient, cell‐selective, and durable transgene expression. ⁹⁵ , ⁹⁶ , ⁹⁷ , ⁹⁸

5.3. Novel serotypes

Novel rAAV serotypes are engineered capsid variants that extend beyond the classical 13 natural serotypes and are designed to enhance tissue tropism and transduction efficiency while overcoming key limitations of natural AAVs, including pre‐existing immunity, suboptimal transduction efficiency, and limited tissue specificity that constrain their therapeutic potential. ⁹⁹

Beyond altering tissue tropism and immune evasion, specific capsid protein mutations have been shown to enhance rAAV production by improving capsid assembly efficiency, genome packaging, and the ratio of full to empty particles. ¹⁰⁰ The three approaches typically used to develop novel rAAV capsids are rational design, directed evolution, and machine learning or artificial intelligence.

In rational design, structural, and functional insights are utilized to modify capsids for enhanced performance. Zhong et al. ¹⁰¹ used rational capsid design by performing site‐directed mutagenesis of surface‐exposed tyrosine residues, and the authors reported nearly a 30‐fold improvement in transduction efficiency at a one‐log lower vector dose. The authors demonstrated that the increased transduction efficiency of tyrosine‐mutant vectors is attributed to a reduction in capsid ubiquitination and enhanced intracellular trafficking to the nucleus. ¹⁰¹ In another study, a domain‐swapping strategy was employed to generate 27 chimeric capsid genes, which contained exchanged domains between AAV2 and AAV8. ¹⁰² The study demonstrated that the more efficient liver transduction achieved by AAV8 was closely related to the components of its interstrand Loop IV domain, particularly the subloops 1 and 4. ¹⁰² Rational design has been employed in inserting cell‐penetrating peptides into the viral capsid, resulting in enhanced CNS transduction after systemic delivery. ¹⁰³ The authors demonstrated that this variant exhibits enhanced transcytosis at the blood–brain barrier (BBB) and can be used to deliver antitumor payloads in a mouse model of glioblastoma. ¹⁰³

Directed evolution utilizes selective pressure to isolate a capsid variant with desired properties, such as enhanced transduction of a target tissue, reduced liver uptake, increased rAAV titer, or resistance to neutralizing antibodies. ¹⁶ Large libraries of capsid variants are typically generated by error‐prone PCR, DNA shuffling, or peptide insertion into exposed capsid loops, and then subjected to iterative rounds of selection in vivo or in relevant primary cells to enrich variants that best meet the performance criterion. ⁹⁹ A directed evolution approach using random point mutagenesis to generate capsid libraries and selecting AAV2 variants with enhanced properties through high‐throughput screening was employed to isolate mutants that evade antibody neutralization. This approach resulted in a variant with a 96‐fold increase in antibody resistance compared to wild‐type AAV2. ¹⁰⁴ Koerber et al. ¹⁰⁵ used directed evolution involving DNA shuffling of the cap genes of numerous parent AAV serotypes (AAV1, 2, 4–6, 8, and 9) to generate chimeras with broad diversity for cell tropism and neutralizing antibody resistance. A directed evolution platform developed by Voyager Therapeutics, TRACER™ (Tropism Redirection of AAV by Cell‐type‐specific Expression of RNA), was utilized to create a novel capsid with enhanced CNS‐tropic properties and improved BBB‐crossing potential, and demonstrated the importance of utilizing expression based screen methods to identify capsids with improved transduction properties. ¹⁰⁶ Importantly, targeting of the transferrin receptor has now been demonstrated by several groups to enhance modified rAAV crossing of the BBB. ¹⁰⁷

Machine learning (ML) or artificial intelligence in capsid engineering relies on the computational identification of patterns in a dataset that can be harnessed to design and optimize rAAV capsids. ⁹⁹ Zhu et al. ¹⁰⁸ used a ML method for designing AAV peptide insertion libraries that achieve fivefold higher packaging fitness than the standard NNK library with negligible reduction in diversity. The authors demonstrate that ML yields approximately 10‐fold more successful variants than the NNK library after selection for infection of human brain tissue, resulting in a promising glial‐specific variant. Deep learning was used to design highly diverse adeno‐associated virus 2 (AAV2) capsid protein variants that remain viable for packaging DNA payload. ¹⁰⁹ Focusing on a 28‐amino acid segment, even when trained on limited data, deep neural network models accurately predict capsid viability across diverse variants. ¹⁰⁹ In summary, these novel rAAV serotypes yield vectors with enhanced transduction efficiency, reduced dosing requirements, and improved resistance to neutralizing antibodies, thereby overcoming key limitations of natural AAV serotypes.

Novel AAV capsids must meet manufacturing requirements, that is, upstream volumetric productivity and efficient downstream recovery. Although several capsid engineering approaches have improved tissue tropism, potency, and immune evasion, manufacturability remains a critical, often under‐optimized design constraint. Modifications to AAV capsids can lead to imbalances in viral protein sequence homogeneity, stoichiometry, and functional transduction units, thereby introducing new challenges. ¹¹⁰ Capsid mutations can unintentionally disrupt assembly by affecting overlapping AAP sequences; therefore, restoring AAP compatibility has been shown to rescue vector yield. ¹¹¹ Some recent work by Ohba et al. showed that modulating capsid protein expression timing and stoichiometry, without changing capsid sequence, significantly increased rAAV yield and reduced empty capsid formation across multiple serotypes. ¹¹² Together, these studies establish capsid assembly as a central rate‐limiting step in rAAV manufacturing and highlight capsid sequence and expression control as powerful, underutilized levers for yield improvement.

5.4. Transfection improvements

Optimization of the transfection step explores parameters such as cell density at transfection, total pDNA/cell, transfection reagent/DNA, and plasmid ratios of transgene, packaging, and Adhelper plasmid. Initial attempts to evaluate these parameters used one‐factor‐at‐a time (OFAT) approach and achieved clarified lysate titers greater than 1 × 10¹⁴ vg/L. ⁴⁶ By using Design of Experiments (DoE), where the parameters (cell density at transfection, total DNA/cell, transfection reagent/DNA, and plasmid ratios of transgene, packaging, and Adhelper plasmid) could be simultaneously varied, Zhao et al., achieved clarified lysate titers of 3 × 10¹⁴ vg/L. ¹¹³ The choice of DoE method and design is crucial for avoiding misleading results; therefore, Tzimou et al. ¹¹⁴ compared four approaches: rotatable central composite design (RCCD), Box–Behnken design (BBD), face‐centered central composite design (FCCD), and mixture design (MD) in the triple transfection process. The results reveal that blocking is essential to reduce variability from uncontrolled random effects, and MD coupled with FCCD outperformed other approaches, resulting in a reported volumetric productivity that is >100‐fold ¹¹⁴ higher than the control condition.

Early scale‐up work in the field utilized three common transient transfection reagents: DNA coprecipitation with calcium phosphate, polyethyleneimine (PEI) coprecipitation, and cationic lipids. ¹¹⁵ Work on new and improved transfection reagents has been an active area of development, with meaningful impact for rAAV manufacturing. For example, FECTOVIR was formulated as an alternative to PEI and has been reported to enhance physical titers by approximately 3‐fold. ¹¹⁶ , ¹¹⁷ By using FECTOVIR in the design of experiments in small‐scale bioreactors and optimizing parameters such as pH, complexation time, viable cell density at transfection, transfection reagent to DNA ratio, the optimized conditions resulted in clarified lysate titers of ~1 × 10¹⁵ vg/L. ¹¹⁷ These results show up to a log increase in rAAV yield (1 × 10¹⁴ to 1 × 10¹⁵ vg/L), giving the potential to treat 10‐fold more patients per batch.

5.5. rAAV production enhancers

Another area of active research that has substantial impact for rAAV production is the derivation and screening of small molecule enhancers. The addition of small bioactive molecules such as Nocodazole (an anti‐mitotic agent) and M344 (a selective histone deacetylase inhibitor) has been shown to significantly increase recombinant AAV production in HEK293 cells. ¹¹⁸ When Nocodazole was added to HEK293 suspension cells that were transiently transfected to produce rAAV, the cells were arrested in the G2/M phase of the cell cycle, leading to an increase in average cell volume, a decrease in viable cell density, and a significantly higher final crude genome vector titer (more than double) compared to untreated cultures. ¹¹⁸ One company, Virica, is focused on developing enhancers for vaccine and vector production and has assembled a proprietary collection of small molecules that have been demonstrated to enhance the production of viruses by transiently and efficiently dampening cellular antiviral defenses. ¹¹⁹ In another study, Ascend Gene and Cell Therapies developed a miniaturized suspension adapted high‐throughput screening strategy: ATLAS (Arrayed Targeted Library for AAV Screening) to screen small molecules. Optimization studies show translatable, reproducible, and comparable AAV9 yields from 96 well to 125 mL shake flask format, and screening of small molecules was performed using a curated compound library of over 700 small molecules. ¹²⁰ Targets identified include epigenetic modulators, DNA damage response, GPCR and transmembrane transporters, cell cycle modulators, anti‐infection, and metabolic targets. ¹²⁰ Importantly, Mirus Bio (now Millipore) recently developed an enhancer, RevIT AAV enhancer that produces 2–3× higher genome titers in suspension HEK293 cells when used with other transfection reagents or their transfection reagent, TransIT‐VirusGen. ¹²¹

Beyond chemical enhancers, process intensification strategies such as fed‐batch and perfusion culture have increasingly been explored to enhance rAAV productivity by increasing viable cell density at the time of transfection (for transient systems) or induction (for stable or infection‐based systems). Perfusion‐based processes enable maintenance of high cell densities while controlling nutrient availability and waste metabolite accumulation, thereby creating favorable conditions for rAAV production. ¹²² When combined with optimized transfection conditions and molecular enhancers, high‐density perfusion cultures have the potential to substantially increase volumetric vector yield without proportionally increasing bioreactor volume.

5.6. Downstream improvements

In the downstream processing of rAAV, the initial step to recover the viral vector typically involves cell lysis using either non‐ionic detergents or mechanical disruption methods, which breaks open the cells to release the rAAV particles into the supernatant for further purification. ³⁸ , ¹²³ The total viral particles generated (vp) consist of full capsids containing the complete vector genome (vg), partially filled capsids with incomplete genomes (partials), empty capsids lacking genetic material (empties), and capsids containing packaged DNA impurities. ¹²⁴ An example of a purification process is shown in Figure 4, where clarified lysate is passed through an affinity resin column and the rAAV capsids bind to the ligand, allowing for selective isolation of capsids from contaminants. The dynamic binding capacity of AAVX resins is reportedly 10¹³ to 10¹⁴ viral particles/mL of resin, and that of AVB Sepharose is greater than 10¹² viral particles/mL of resin (Table 3).

TABLE 3.

Affinity resins used for rAAV production.

Affinity resin	Supplier	AAV serotype specificity	Binding capacity	Affinity ligand	Residence time
POROS™ CaptureSelect™ AAVX	Thermofisher	AAV1, 2, 3, 4, 5, 6, 7, 8, 9, AAVrh10, synthetic serotypes	1013 to 1014 vp/mL resin	VHH	≥3 min
POROS™ CaptureSelect™ AAV9	Thermofisher	AAV9	>1014 vp/mL resin	VHH	≥3 min
POROS™ CaptureSelect™ AAV8	Thermofisher	AAV8	>1013 vp/mL resin	VHH	≥3 min
AVB Sepharose HP	Cytiva	AAV1, 2, 3, 4, 5, 6, 8, AAVrh10	>1012 vp/mL resin	VHH	≥3 min
AVIPureR AAV2	Repligen	AAV2	>2 × 1013 vp/mL resin	Peptide	≥30 s
AVIPureR AAV5	Repligen	AAV5	>2 × 1013 vp/mL resin	Peptide	≥30 s
AVIPureR AAV8	Repligen	AAV8	>1 × 1015 vp/mL resin	Peptide	≥30 s
AVIPureR AAV9	Repligen	AAV9	>2 × 1014 vp/mL resin	Peptide	≥30 s
AAVidity	LigaTrap	AAV1, 2, 3, 4, 5, 6, 7, 8, 9, AAVrh10, synthetic serotypes	>5 × 1013 vp/mL resin	Peptide	3 min

Note: VHH is antigen binding fragment of heavy chain antibody.

Step recovery from the affinity steps is usually in the 50%–100% range, depending on the resin type, serotype, and the elution conditions. ¹²⁵ A problem with most of the available affinity resins is the need to elute in strongly acidic conditions (pH ≤3), which can reduce rAAV transduction activity and shorten the resin's lifespan. ¹²⁵ Newer resins (e.g., peptide–ligand based) are being developed for mild elution conditions (~pH 6.0) and can be manufactured at a lower cost compared to antibody‐based ligand (VHH). ¹²⁶ These advances in the development of new affinity resins with milder elution conditions, longer lifespan, and lower production cost could have a significant impact on the recovery, product quality, and commercial viability of rAAV products.

After affinity chromatography, a polishing step is required to enrich full capsids by exploiting small charge and density differences between full and empty capsids. Full capsids have slight charge differences compared to empty capsids; therefore, cation/anion exchange chromatography can be used to separate them. ¹²⁶ Some common cation/anion exchange resins used are Capto Q, Poros HQ, Mustang S/Q and CIMultus QA. ¹²⁷ Packed beds, membranes, and monoliths can be used as stationary phases for cation/anion exchange, and the mobile phase of solutions with a wide range of binding ionic strengths and pH can be employed. For elution, a shallow linear gradient with increasing salt concentration is typically used to enrich full particles, but product‐specific optimization is needed for improving the recovery and percentage of full capsids. ¹²⁷ The product‐specific optimizations may include testing different salt concentrations of NaCl and MgCl₂, and the addition of salt modifiers such as ammonium acetate, magnesium sulfate, sodium acetate, or different quaternary ammonium salts. ¹²⁷ , ¹²⁸ , ¹²⁹ After product‐specific optimization of linear gradients, transitioning from a linear gradient to a step gradient is often necessary to achieve better scalability and consistent production.

Gradient ultracentrifugation has been used for decades for the enrichment of full rAAV particles. Since full rAAV capsids are denser than empty capsids the differences in density can be utilized in ultracentrifugation gradients to separate full particles from empty and partially packaged particles, as the denser full capsids sediment faster and collect at a lower position in the gradient compared to the lighter empty or partial capsids. The reagents most commonly used to form the gradients are either cesium chloride (CsCl) or iodixanol. ¹²³ , ¹³⁰ Both CsCl and iodixanol ultracentrifugation have been used for the manufacturing of commercial AAV products, but each has limitations. Iodixanol is a sugar contrast agent used in clinical settings, so it is considered a safer reagent, but has limited scalability and less clear separation of rAAV particles within the gradient. CsCl is a naturally occurring mineral salt, but is not currently available as a cGMP‐grade reagent. CsCl has been evaluated as a cancer therapy in clinical trials with no reported safety issues, ¹³¹ but there have been reported adverse events when CsCl is repeatedly self‐administered at high doses as an alternative therapy by cancer patients. ¹³² The benefits of CsCl ultracentrifugation include self‐forming gradients, high loading capacity, and visible banding. ¹³³ The full and empty capsids of rAAV products are typically collected from gradients using a needle and rely on a high degree of operator training for consistent band pulling. Even though ultracentrifugation has been used for commercial rAAV products, it is labor‐intensive and not convenient for large‐scale cGMP environment. ¹²³ Continuous ultracentrifugation is an attractive approach for the large‐scale separation of rAAV capsids, which eliminates the labor‐intensive nature of the process and enhances consistency in cGMP manufacturing, but the systems are not currently available with single‐use product contact consumables. ¹³⁴ The differences between the polishing methods are summarized in Table 4.

TABLE 4.

Comparison of cation/anion exchange, cesium chloride, and iodixanol ultracentrifugation.

Polishing steps	Cation/anion exchange	Cesium chloride ultracentrifugation	Iodixanol ultracentrifugation
Separation approach	Charge differences	Density differences	Density differences
Resins/chemicals	Capto Q, Poros HQ, Mustang S/Q and CIMultus QA 127	Cesium chloride	Iodixanol
Scalability	Very scalable	Scalable	Limited scalability
Percent full capsids	Moderate (process/capsid dependent) (~80%) 129	Gold standard (>90%) 130	Moderate (~80%) 130
Step recovery	Moderate to good (process/capsid dependent) (~70%–80%) 129	High (>90%) 130	High (>90%) 130
Impurity clearance	Very good. Effective removal of HCP and hcDNA	Good, but residual cesium chloride needs to be removed	Moderate, but residual iodixanol needs to be removed
Cost	Moderate	High	High
Process time	Short to moderate (~2–6 h per cycle), depending on the process. 133	Long (>12–24 h per spin) 130	Moderate (~ 2 h per spin) 130
Automation potential	High	Moderate	Moderate

5.7. Analytical improvements

Robust analytical characterization is essential to rAAV manufacturing, underpinning process development, lot release, and regulatory submissions. The optimization of upstream and downstream processes for the production of rAAV with consistent quality depends on the ability to rapidly characterize critical quality attributes, including virus titer, capsid content, aggregation, potency, purity, and safety for rAAV‐mediated gene therapy products. The suite of analytical assays used for rAAV has continued to be refined over the past few decades, especially in the emergence of digital PCR, analytical ultracentrifugation (AUC), mass spectroscopy, and next generation sequencing. Early work in the field relied on older technologies for rAAV titering, such as dot blot assays with radioactively labeled probes, before moving to qPCR using primer/probes (usually targeting the ITRs to ensure consistent design across products); however, these methods were less accurate than ddPCR and often led to overestimation of the actual titer. ¹³⁵ , ¹³⁶ Genome titer now is commonly measured using digital PCR (dPCR) or digital droplet PCR (ddPCR), but ddPCR has emerged as the preferred method. ¹³⁷ , ¹³⁸ Techniques such as AUC, mass photometry, charge detection mass spectrometry (CDMS), and capillary isoelectric focusing (cIEF) are increasingly used as characterization tools to resolve full, intermediate, and empty populations. ¹³⁷ , ¹³⁸ AUC is an improvement over previous methods such as negative‐stained transmission electron microscopy and has now become the gold standard assay since it can accurately measure not only empty and full particles, but also partially packaged and oversized particles.

CDMS is an emerging technology and has the potential to rapidly analyze both the packaged transgene subtypes as well as the capsid makeup. ¹³⁹ Mass spectroscopy is an improvement over capsid ELISAs for vector protein identity, since the ELISA methods can be misleading due to cross‐reactivity, and mass spectroscopy can not only accurately define the capsid identity but can also be utilized for detection of PTMs and protein impurities. ¹³⁸ Next generation sequencing is used to confirm transgene sequence integrity, assess inverted terminal repeat (ITR) fidelity, detect mutations or rearrangements, and identify partial or heterogeneous genome packaging events that may not be captured by other methods. ¹⁴⁰ , ¹⁴¹ Improving existing rAAV analytical methods and developing new methods to shorten turnaround times and increase throughput are essential for supporting efficient process development and manufacturing, as well as developing a more robust understanding of the CQAs that define rAAV product success criteria. ¹³⁷ , ¹³⁸

5.8. Process characterization

As these programs move toward commercialization and regulatory approval, process characterization becomes a critical step, shifting the focus from early process development toward generating a deep understanding of how process parameters impact critical quality attributes (CQAs). Following quality by design (QbD) principles, developers need to execute comprehensive process characterization strategies aligned with FDA/EMA guidance (ICH Q8, Q9, Q10). ¹⁴² , ¹⁴³ , ¹⁴⁴ This process begins with defining the quality target product profile (QTPP) and identifying CQAs, followed by parameter risk assessment (PRA) to evaluate the impact of process parameters on product quality using historical data and subject‐matter expertise. Depending on available information, failure mode and effects analysis (FMEA) is performed either right after the PRA or after process characterization studies. Subsequently, scale‐down models (SDMs) are developed that represent the process at scale to evaluate process performance and product quality. These models are used to perform process characterization studies, typically through design of experiments (DoE) and complementary approaches, to assess the impact of high‐risk parameters on process performance and CQAs. The outcome of this work results in the establishment of a process control strategy that defines parameter ranges, targets, and criticality, serving as the foundation for process performance qualification (PPQ) campaigns. A‐gene, ¹⁴⁵ an industry framework example, and a recent paper by Kowshik and Singh ¹⁴⁶ provide a great overview of process characterization for rAAV production.

5.9. Impact of regulations on manufacturing

Regulatory requirements significantly influence rAAV manufacturing by shaping process development, quality control, scalability, and compliance strategies throughout the product lifecycle. The FDA is offering expedited pathways—such as RMAT (Regenerative Medicine Advanced Therapy) ¹⁴⁷ and Fast Track ¹⁴⁸ —which, while speeding up approval, require intensive, continuous interaction with regulators and robust data to support preliminary evidence. Recently, the FDA's “Plausible Mechanism” framework, ¹⁴⁹ introduced via draft guidance in February 2026, is a regulatory approach designed to accelerate approval of highly personalized, or “bespoke,” therapies, particularly for ultra‐rare, serious genetic conditions. It addresses scenarios where traditional randomized clinical trials are not feasible due to extremely small patient numbers. The U.S. FDA is currently piloting Real Time Clinical Trials (RTCT) ¹⁵⁰ in 2026, aimed at accelerating drug approvals by allowing regulators to see data signals directly from clinical research platforms, thereby reducing the time to approval. These changes in speeding up clinical trials demand that manufacturing processes deliver a consistently characterized product earlier in development than traditional approval timelines. Sponsors need to pursue a more rapid CMC development program to accommodate the accelerated clinical trial pathways.

Despite the need for a rapid CMC development, manufacturing of rAAV still requires strict compliance with current Good Manufacturing Practices (cGMP) enforced by agencies like the FDA (US) and EMA (Europe) to ensure safety, purity, and potency. ¹⁴² , ¹⁴³ , ¹⁴⁴ Regulatory expectations are increasingly driving how process development and rAAV manufacturing strategies are designed, pushing organizations toward greater consistency, scalability, and control. Agencies now emphasize robust characterization of CQAs, process understanding, and data integrity across the entire lifecycle, from early development through commercial production. This has led to a shift toward platform‐based approaches, ¹⁵¹ implementation of QbD principles, and earlier adoption of scalable, GMP‐compatible processes. The platform approach enables the use of data from previous clinical trials to support new drug applications and streamline development activities. Other opportunities the regulators have offered to support CMC development include. CMC development and readiness program (CDRP), advanced manufacturing technologies (AMT) designation, and CMC flexibilities developing human cell and gene therapies. ¹⁵² , ¹⁵³ , ¹⁵⁴ Additionally, regulators expect more rigorous control of impurities such as empty capsids and host cell proteins, as well as comprehensive analytical methods to ensure product quality and comparability. As a result, companies are aligning their manufacturing strategies with regulatory guidance earlier in development to reduce risk, streamline approvals, and support long‐term product consistency.

5.10. Upcoming approvals

Beyond the currently approved products (Table 1), additional rAAV candidate products are under investigation in preclinical and clinical development spanning numerous therapeutic areas, including neuromuscular diseases, hemophilia, cardiovascular‐associated disorders, lysosomal storage diseases, mucopolysaccharidoses, primary central nervous system disorders, and ocular diseases. ¹⁵⁵ A summary of the upcoming approvals from the FDA and EMA in 2026, including the therapeutic developers, indications, and expected decision dates, is presented in Table 5.

TABLE 5.

2026 AAV pipeline of approvals.

Therapy name and developer	Therapy type (indication)	Expected decision date
REGENXBIO a RGX‐121	AAV therapeutic for the treatment of boys with MPS II, designed to deliver the iduronate‐2‐sulfatase (IDS) gene to the central nervous system (CNS).	Plans to resubmit 156 , 157
Ultragenyx DTX401	AAV gene therapy for the treatment of glycogen storage disease type Ia, a serious inherited disease that results in the inability to regulate blood sugar.	August 23, 2026 157 , 158
Sangamo Isaralgagene Civaparvovec	A liver‐tropic rAAV 2/6 vector carrying the cDNA of human alpha‐galactosidase A (GLA) for Fabry disease. Isaralgagene civaparvovec aims to deliver a functional copy of the GLA gene to the liver, enabling liver cells to produce functional α‐Gal A.	Rolling BLA in progress. 156 , 159
Nanoscope Therapeutics MCO‐010	AAV gene therapy with multi‐characteristic opsin (MCO) gene to treat retinitis pigmentosa. MCO‐010 is an AAV2‐based vector with the MCO gene expression cassette. It genetically modifies bipolar retinal cells to produce a light‐sensitive protein, allowing them to utilize the eye's remaining visual pathways and compensate for the death of the original light‐sensing cells. 160	FDA BLA to be submission possible in H1 2026 156 , 159
Ultragenyx UX111	AAV9 gene therapy to deliver SGSH gene to cells of the CNS and peripheral organs for the potential treatment of Sanfilippo syndrome (MPS IIIA).	September 19, 2026 159
RegenxBio RGX‐202	RGX‐202 is designed to use the AAV8 vector to deliver a transgene to muscle cells that encodes a novel microdystrophin that includes the functional elements of naturally occurring dystrophin including the C‐terminal (CT) domain.	FDA BLA to be submission possible in H2 2026 159
Ocugen OCU400	AAV gene therapy for retinitis pigmentosa using nuclear hormone receptor NR2E3 to the retina and retinal pigment epithelium (RPE). AAV5‐hNR2E3 gene therapy platform offers a promising approach to preserve and potentially restore retinal structure and function.	EMA MAA submission possible in 2026 156

^a FDA rejects RGX‐121 but plans to resubmit for approval with longer‐term data. ¹⁵⁷

6. CHANGING LANDSCAPE OF rAAV

In recent years, the commercial approvals and clear efficacy observed in some of the clinical trials targeting rare monogenic disorders has driven a resurgence in efforts to address broader and more prevalent disease populations using rAAV‐mediated gene therapy. Examples of large patient‐population diseases being investigated using rAAV gene therapy include knee osteoarthritis, Alzheimer's disease and mild cognitive impairment, heart failure, Parkinson's disease, and diabetes (Table 6, ¹⁵⁸ , ¹⁶¹ , ¹⁶² , ¹⁶³ , ¹⁶⁴ , ¹⁶⁵ ). In osteoarthritis, intra‐articular delivery of rAAV encoding interleukin‐1 receptor antagonist (IL‐1Ra) has shown encouraging results. ¹⁵⁸ A Phase 1 clinical trial in nine patients demonstrated sustained elevation of IL‐1Ra in synovial fluid for more than 1 year, accompanied by improvements in pain and function, supporting the safety and feasibility of this approach. ¹⁵⁸ In parallel, within the central nervous system, rAAV‐mediated delivery of brain‐derived neurotrophic factor (BDNF) is being evaluated in early Alzheimer's disease and mild cognitive impairment, building on preclinical evidence that BDNF supports neuronal survival and synaptic function. ¹⁶¹ The open‐label Phase I study is being conducted in 12 participants to assess the safety, tolerability, and preliminary efficacy of AAV2‐BDNF. ¹⁶¹

TABLE 6.

Examples of application rAAV in non‐rare diseases.

Disease area	rAAV trial/product	Notes
Osteoarthritis (knee OA)	IL‐1Ra gene was delivered by a self‐complementary (sc)–recombinant adeno‐associated virus (rAAV) serotype 2.5 (sc‐rAAV2.5IL‐1Ra). 158	Phase 1 clinical trial to test the safety of intra‐articular delivery of an adeno‐associated virus expressing interleukin‐1 receptor antagonist (IL‐1Ra), an endogenous inhibitor of the proinflammatory cytokine IL‐1. 158
Alzheimer's disease/mild cognitive impairment	Brain‐derived neurotrophic factor (BDNF) in AAV2. AAV2‐BDNF. 161	First‐in‐human AAV2–BDNF trial in patients with early Alzheimer's disease or Mild Cognitive Impairment (MCI), testing whether continuous BDNF expression can slow neurodegeneration. 161
Heart failure	AAV1/SERCA2a cardiac gene therapy carrying cardiac isoform of the sarcoplasmic reticulum calcium ATPase. 162	AAV‐mediated phase 2a trial assessing the safety and feasibility of delivering (AAV1/SERCA2a) to adult chronic heart failure patients implanted with a left ventricular assist device. 162 , 163
Parkinson's disease (PD)	Glial cell line‐derived neurotrophic factor (GDNF)–(AAV2‐GDNF) bilateral intra‐putaminal infusion in participants with advanced Parkinson's disease (PD). 164	Phase 1b study investigating AskBio's bilateral putaminal AAV2‐GDNF gene therapy in participants with mild or moderate PD. Preliminary assessments of motor function suggest that putaminal infusions with up to 1800 μL of 3.3 × 1015 vg/L AAV2‐GDNF may provide a beneficial disease‐modifying effect in the form of stabilization or restoration in the mild and moderate cohorts. The results support further investigation of AAV2‐GDNF in an ongoing phase 2 randomized, double‐blind, surgical‐controlled study (regenerated; NCT06285643). 164
Diabetes	Human GLP‐1 PGTx (RJVA‐001) transgene construct, consisting of human insulin promoter and GLP‐1 sequences, is packaged into (AAV)‐9 vector. 165	RJVA‐001 can be safely delivered endoscopically in a large animal model (pigs), demonstrating a single dose gene therapy for diabetes. Active GLP‐1 protein was five‐fold greater in RJVA‐001‐treated pigs compared to untreated controls (D, p < 0.02). RJVA‐001 first‐in‐human trial in type 2 diabetes is being planned. 165

In cardiovascular disease, early studies with AAV1‐SERCA2a showed safety but limited efficacy, highlighting challenges with delivery and transduction in the heart. ¹⁶² More recent trials using chimeric cardiotropic AAV (AB‐1002) have demonstrated favorable safety profiles and preliminary functional improvements, informing Phase 2 studies. ¹⁶³ Similarly, in Parkinson's disease, intraputaminal delivery of glial cell line‐derived neurotrophic factor (AAV2‐GDNF) has been well tolerated and associated with clinical stabilization or improvement in early‐stage trials. ¹⁶⁴ Fractyl Health's pancreatic delivery of an rAAV encoding GLP‐1 has demonstrated safe, localized gene expression in large animal models, supporting the feasibility of single‐dose gene therapy for type 2 diabetes. ¹⁶⁵ As these indications progress into Phase 2/3 trials, rAAV manufacturing transitions into a broader therapeutic modality targeting prevalent chronic diseases; a higher total vector demand will be required at manufacturing scale. For example, a disease with 10 million patients with a modest dose of 1 × 10¹⁴ vg would require 1 × 10²¹ vg. With upstream productivity of 1 × 10¹⁵ vg/L at a 50% downstream recovery would require 2 million liters of bioreactor volume. The successful deployment of rAAV in large‐population diseases will require leveraging advances in transgene design and capsid design to enable lower vector doses, as well as an improved, scalable manufacturing process to enable more doses per batch.

7. CHALLENGES IN rAAV MANUFACTURING

Manufacturing rAAV vectors for large patient populations requires highly scalable, high‐yield upstream processes, combined with efficient, robust downstream purification strategies, to meet the substantially increased vector demand. Development and manufacturing of rAAV are more complex than other biologics, like monoclonal antibodies (mAbs). This complexity is due to several factors, such as the technical maturity of mAbs compared to rAAV therapies, the complex biology of rAAV, and the unique nature of the analytical methods, which can lead to lab‐to‐lab variability and inaccuracies. ¹⁶⁶ The manufacturing yields of monoclonal antibodies have been compared with rAAV, and the results indicate that monoclonal antibodies produce at ~20–50 pg/cell/day whereas rAAV is produced at 1–2 pg/cell/day. ¹⁶⁷ Given that the productivity per cell is at least a log lower for rAAV than mAbs and the cell density achieved in a mAb process is another log higher, this makes rAAV at least two logs lower in productivity compared to mAbs. The key differences between rAAV and monoclonal antibody (mAb) manufacturing are summarized in Table 7. Notably, mAb production typically involves longer culture durations than rAAV processes; therefore, the final volumetric titers achieved for rAAV are at least three orders of magnitude lower than those routinely achieved for mAbs. ¹⁶⁷ The magnitude of the difference between mAb and rAAV productivity highlights the need to identify bottlenecks in rAAV production and the opportunities to continue improving rAAV manufacturing.

TABLE 7.

Comparison of rAAV and mAbs.

	Recombinant adeno‐associated virus (rAAV) (triple transfection in HEK293)	Monoclonal antibody (mAbs)
Cell line	HEK293	CHO
Duration of production per batch	~3 days	~2 weeks
Primary recovery	Cell lysis to remove the product	No cell lysis
Purification	2 step purification with an affinity step	2–3 step purification with an affinity step
pg/cell/day a	1–2 a	20–50 a
Molecules/cell/day a	1–2 × 105a	800–2000 × 105a
Volumetric productivity (g/L) a	0.001–0.01 a	4–10 a

^a Numbers obtained from Wright. ¹⁶⁷

Systems biology and cell‐engineering approaches that have been successfully applied in CHO‐based mAb processes could also be utilized for rAAV production in HEK293 cells to gain a deeper understanding of viral production mechanisms and enhance rAAV productivity. One possibility for the relatively lower productivity of rAAV compared to mAbs is the induction of inflammatory and antiviral responses in HEK293 cells during the triple transfection process. By using gene ontology analysis during triple transfection of HEK293, Chung et al. identified the regulation of MAPK activity, upregulation of histone genes, pathways related to inflammation, and innate response to viruses during rAAV production. ¹⁶⁷ , ¹⁶⁸ Research suggests that removing inflammatory and antiviral factors in HEK293 cells and temporally expressing the adenoviral helper elements and replication proteins at controlled ratios could provide optimal rAAV production. ¹⁵ , ¹⁶⁸ , ¹⁶⁹

One critical complication for the successful generation of rAAV vectors is that the protein capsid shell is formed within the cell prior to the packaging of the DNA transgene, which leads to a mixture of fully packaged rAAV particles, along with empty and partially packaged rAAV. In a triple transfection rAAV production process, full capsids might constitute only 10%–50% of total capsids. ¹¹⁷ This indicates that the majority of the capsids are either empty or partially packaged, and therefore improvements are needed to increase the percentage of full capsids. The heterogeneity of the starting upstream material increases the downstream processing burden. In a study using conformation‐specific antibody staining and flow cytometry to determine the percentage of cells producing rAAV, the authors identified that only 5%–10% of the HEK293 cells appear to produce measurable levels of assembled rAAV capsids despite an approximate 60% transfection efficiency. ¹⁷⁰ These results suggest that yield and product quality in the HEK293 transient transfection process could be improved by increasing the proportion of cells that are productively making rAAV.

As rAAV manufacturing scales toward indications with million‐patient populations, scalability and process robustness become dominant challenges. Unlike mAb manufacturing, where scale‐up (and increasingly scale‐out) strategies are highly standardized, rAAV processes often face constraints related to plasmid supply, transfection efficiency at large volumes, shear sensitivity, and variability in viral assembly. Achieving scalable production while maintaining consistent product quality will require process intensification approaches, including high‐density cultures enabled by fed‐batch or perfusion strategies, as well as modular, single‐use bioreactor systems capable of reliable operation at the multi‐thousand‐liter scale. In parallel, implementation of process analytical technologies (PAT) will be essential for risk mitigation and process control as rAAV manufacturing scales.

8. FUTURE DEVELOPMENTS IN rAAV MANUFACTURING

8.1. Automation and PAT

Despite significant advances in vector design, process development, and scale‐up, rAAV manufacturing remains largely manual and labor‐intensive, particularly in upstream production and downstream purification. This reliance on manual operations contributes to high manufacturing costs, operator‐to‐operator variability, increased risk of contamination and batch failure, and limited scalability, challenges that become more pronounced as rAAV therapies expand to larger patient populations requiring higher production volumes and consistent product quality. Transitioning toward greater automation, including automated liquid handling for transfection, automated operation of large‐scale single‐use bioreactors (>1000 L), programmable downstream equipment, and in‐line process analytics, will improve process reproducibility, increase throughput, and increase process robustness while reducing manual labor and human error.

Biopharma 4.0 refers to the application of Industry 4.0 principles—digitalization, automation, advanced analytics, and connectivity—to biopharmaceutical development and manufacturing. It aims to integrate cyber‐physical systems and smart factories to yield automated, end‐to‐end bioprocesses for biopharmaceutical production, optimizing efficiency by minimizing human intervention and associated uncertainty. ¹⁷¹ Recently, Iglesias et al. ¹⁷² proposed that soft sensing, combined with predictive modeling, is an important strategy for optimizing the upstream process of rAAV production by monitoring critical process variables in real‐time. ¹⁷² Soft sensors, also known as virtual sensors, are computational models that estimate critical process variables using easily measured online data, combined with mathematical or machine learning models. In biopharmaceutical manufacturing, many important attributes, such as viable cell concentration, specific growth rate, nutrient consumption, metabolite accumulation, or product titer, are difficult, time‐consuming, or expensive to measure directly in real‐time. ¹⁷³ In downstream processing, soft sensors can determine column loading capacity, breakthrough behavior, or impurity clearance by combining pressure, UV, conductivity, and flow data, supporting consistent chromatography performance. ¹⁷³ Soft sensors have helped improve production rates and product quality in monoclonal antibody production, and they are expected to support automation in rAAV manufacturing. ¹⁷² , ¹⁷³ Automation will be a key enabler for future rAAV manufacturing, reducing human error and supporting the scale and reliability required for the next generation of gene therapies.

8.2. Process intensification and continuous processing

Process intensification involves making a manufacturing process faster, more efficient, and more productive, while maintaining or improving product quality and safety. Using monoclonal antibody production in CHO cells, Muller et al. ¹⁷⁴ used elements of process intensification, such as optimized clone selection, N − 1 perfusion, an intensified N stage, and continuous downstream purification to offer synergistic gains with yields up to 10‐fold and shortening production runs by 30%. ¹⁷⁴ For rAAV production, Zhang et al. ¹⁷⁵ used high cell density stirred tank perfusion culture with rAAV1 and rAAV9 vectors carrying eGFP cargo, with triple‐plasmid transfection performed in suspension HEK293 cells at a high viable cell density of 50 million cells/mL in culture, then maintained at ≥30 million cells/mL throughout production. The intensified process achieved rAAV9 production levels per cell comparable to those observed in reference shake flask cultures at 1 million cells/mL. ¹⁷⁵ This approach provided a 17‐fold intensification compared to transfection at 3 million viable cells/mL, a typical production density at scale, highlighting the benefit of significantly increased volumetric productivity at very high cell density. ¹⁷⁵ In another study, alternating tangential flow (ATF) and tangential flow depth filtration (TFDF) techniques were implemented using a high‐cell‐density perfusion process to produce AAV8 using mammalian HEK293 cells and transient transfection. ¹⁷⁶ The results demonstrated that rAAV titers could be improved by up to almost fourfold in TFDF and surpassed those of ATF and perfusion‐mocking experiments carried out in shake flasks. ¹⁷⁶

For clarification studies, single‐pass tangential flow filtration (SPTFF) was implemented at bench scale by feeding the clarified cell lysate (CCL) into cellulose membranes. ¹⁷⁷ As an initial proof‐of‐concept, an integrated process was developed at pilot scale, linking clarification, SPTFF, and affinity chromatography, and the integrated process offered an 81% reduction in total operating time (due to the reduced volume of load material for the affinity column after preconcentration by SPTFF), 36% improvement in affinity resin utilization (due to the higher AAV concentration in the column load), and an estimated 10% reduction in raw material costs. ¹⁷⁷ These improvements resulted in an 8.5‐fold increase in overall productivity compared to an equivalent batch process, underscoring the potential for SPTFF to intensify large‐scale AAV downstream processing. ¹⁷⁷

Affinity chromatography process intensification was achieved using a continuous multicolumn chromatography (MCC) approach for AAV8 affinity resin purification with the Resolute BioSMB PD platform, which involves dividing a single chromatography bed into smaller, connected columns. ¹⁷⁸ The continuous sequential multicolumn chromatography (S‐MCC) process developed resulted in higher AAV recovery compared to the batch multicolumn chromatography bed (B‐MCC), with a recovery of 76% for the highest load concentration evaluated, and a productivity of 1.8 × 10¹⁷ vphr⁻¹ mL⁻¹, 1.5 times higher than that of the B‐MCC. ¹⁷⁸ Additionally, for the same volume of processed feed, the S‐MCC halves the buffer consumption compared to the B‐MCC. ¹⁷⁸ The authors showed that S‐MCC significantly improves overall process efficiency, offering higher recovery and productivity while reducing resource consumption. ¹⁷⁸ Recently, the BioRMB™ column‐free purification platform, previously applied to monoclonal antibody manufacturing, was successfully used for rAAV affinity capture, achieving recoveries that meet or exceed those of traditional column‐based processes without compromising product quality. ¹⁷⁹

With the rapid increase in upstream AAV titers, some exceeding 1 × 10¹⁵ vg/L at the bioreactor scale, traditional polishing steps based on linear gradient anion‐exchange chromatography (AEX) are increasingly approaching their scalability limits. ¹⁸⁰ This challenge is particularly pronounced for processes employing monoliths and membranes, which are often favored over packed‐bed resins in AAV purification due to the elimination of pore‐diffusion limitations and improved mass transfer. ¹⁸⁰ As upstream productivity continues to rise, next‐generation downstream processes are required that can accommodate significantly higher loadings (≥1 × 10¹⁵ vp per cycle) while maintaining acceptable processing times, robustness, and cGMP suitability. Recently, Thakur et al. ¹⁸⁰ demonstrated an intensified AEX strategy by optimizing loading conditions and implementing UV‐based automated control, achieving greater than 10‐fold increases in material loaded per cycle while simultaneously reducing processing time by approximately 10‐fold. ¹⁸⁰ This approach is particularly well‐suited for rapid‐cycling and continuous chromatography operations, providing a viable downstream solution to support emerging high‐titer, intensified rAAV manufacturing processes. ¹⁸⁰

8.3. Lessons from monoclonal antibody manufacturing to enable rAAV production

Therapeutic monoclonal antibody manufacturing has evolved significantly since the mid‐1980s through advances in molecular design and bioprocess technology, enabling faster development, lower costs, and reliable supply to millions of patients. ¹⁸¹ Key innovations, including fed‐batch and perfusion cultures, improved media and cell line engineering, downstream purification advances, and process intensification, drove dramatic titer increases from ~0.1 to >10 g/L and supported large‐scale bioreactor operations (from 500 to 25,000 L). ¹⁸² , ¹⁸³ These coordinated developments transformed mAbs into one of the most scalable and widely adopted therapeutic modalities in modern medicine. A similar transformation is now underway in rAAV manufacturing, where applying lessons from mAb production could substantially increase vector yields, reduce the cost of goods, and enable expansion into larger patient populations. Stable producer cell lines could help unlock scalable AAV manufacturing beyond 5000 L by reducing reliance on transient transfection and improving batch‐to‐batch consistency. ¹¹⁸

8.4. Are we approaching the upper limit of rAAV productivity per cell?

In HEK293‐based systems (including transient transfection or producer cell lines) and Sf9 systems, the amount of rAAV produced per cell is limited by intrinsic cellular constraints on viral genome replication, capsid assembly, and intracellular trafficking. ⁴³ , ⁸¹ , ¹⁷⁵ Across the literature, cell‐specific productivities typically range from 1 × 10⁴ to 1 × 10⁵ vector genomes (vg) per cell, with the upper end achieved only under highly optimized conditions involving efficient transfection, balanced rep/cap expression, and carefully timed harvest. ⁴³ , ¹⁷⁵ The cell‐specific yields of wtAAV is on the order of 1 × 10⁵ to 1 × 10⁶ viral genomes (vg) per cell. ¹⁶⁹ With the recent improvements in productivity, titers above 5 × 10¹⁵ vg/L can be achieved for some vectors at a typical 3 million cells/mL at transfection in HEK293 transfection systems, resulting in a cell‐specific productivity of 1.6 × 10⁶ vg/cell. A key open question is whether current production platforms can meaningfully exceed the intrinsic limits of the cell‐specific productivity, or whether further engineering advances could enable rAAV yields to approach 10⁷ vector genomes per cell? Achieving a step‐change in rAAV productivity beyond current production levels will require simultaneously addressing three interconnected dimensions: molecular optimization of the viral and auxiliary elements, modulation of the host cell machinery (such as cell cycle regulation, DNA replication, capsid assembly, anti‐viral response) and its relationship to viral replication, and integration of upstream process intensification with downstream purification advances, rather than incremental optimization of any single parameter in isolation.

9. CONCLUSION

Since AAV was discovered in 1965 as a contaminant of adenovirus preparations, it has undergone a remarkable transformation from a virological curiosity to one of the most successful gene therapy platforms. Foundational studies in the 1970s and 1980s established AAV as a non‐pathogenic virus dependent on helper functions, and subsequent cloning and sequencing of the AAV genome enabled the development of rAAV vectors in which viral coding sequences were replaced with therapeutic transgenes. The first human trials, conducted in the early 1990s, included studies in cystic fibrosis and hemophilia, demonstrating an encouraging safety profile but revealing significant challenges, such as limited efficacy, immune responses, and constraints with repeat dosing. Over the following decades, advances in capsid biology, vector genome design, and manufacturing led to major clinical breakthroughs, including durable liver‐directed expression in hemophilia, restoration of vision in inherited retinal diseases culminating in the approval of Luxturna, and life‐saving systemic gene delivery with AAV9‐SMN1 (Zolgensma) for spinal muscular atrophy. At the same time, setbacks such as immune‐mediated loss of expression and failures in large‐organ indications underscored the biological and translational limits of the platform. Together, six decades of AAV research illustrate a trajectory marked by iterative innovation, clinical triumphs, and persistent challenges, positioning rAAV as a mature yet still evolving cornerstone of modern gene therapy.

Over the past three decades, the manufacturing process for rAAV has evolved from small‐scale, research‐driven methods to increasingly robust and scalable biopharmaceutical processes. HEK293 triple‐transfection platforms have evolved from small‐scale adherent cultures to suspension processes, incorporating improved plasmid design, transfection reagents, and process intensification strategies to boost productivity and consistency. Producer cell line–based systems have benefited from advances in stable cell line engineering and inducible expression technologies, improving process reproducibility, but they remain less flexible than transient transfection systems. Similarly, the baculovirus–insect cell system has evolved from early multi‐baculovirus approaches to more stable and efficient designs, such as dual‐ or single‐baculovirus platforms, which enable higher titers, better product quality, and compatibility with large‐scale bioreactor operations. Collectively, these iterative improvements have transformed all four platforms into more robust manufacturing solutions, with transient transfection and baculovirus systems emerging as the most clinically and commercially mature. Advances in upstream processing, such as optimized transfection, media additives, and process intensification, have increased volumetric productivity, while downstream innovations have improved product recovery and product quality. The adoption of high‐affinity, serotype‐agnostic capture resins and optimized anion exchange chromatography has improved recovery, purity, scalability, and cGMP compatibility, thereby significantly increasing the number of patients treated per manufacturing batch. Novel production systems, including CHO cells, yeast, algae, and plant‐based platforms, are under active investigation to further address scalability and cost, though most remain at early stages.

Beyond rare diseases, rAAV gene therapy is increasingly transitioning toward applications in larger patient populations, such as knee osteoarthritis, Alzheimer's disease/mild cognitive impairment, heart failure, Parkinson's disease, and diabetes, reflecting both clinical maturation and manufacturing advances. The expansion of rAAV gene therapy into large patient populations will strain an already burdened field that has experienced historical limitations in dose, scalability, and efficiency. Realizing the future of rAAV‐mediated therapies will necessitate continued efforts in molecular design, process improvement, and the adoption and adaptation of process‐intensification and automation strategies pioneered in the manufacturing of other biologics, enabling rAAV production to achieve the scale, robustness, and affordability needed for broader impact.

AUTHOR CONTRIBUTIONS

Frank Agbogbo: Conceptualization; writing – review and editing; formal analysis; writing – original draft; data curation; visualization. David Dismuke: Conceptualization; methodology; writing – review and editing; visualization.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.