Viral Vectors Explained | Cell Culture Technology

Back to Technology, Institution & Company Guides

TECHNOLOGY EXPLAINER

Viral Vectors
— how the container that carries a gene is made, and sorted

In practice, a gene therapy is a therapeutic gene packed into a viral shell. An AAV can carry only about 5 kb. And most of the capsids that come out of production are empty, which makes separating the empty ones from the full ones the most awkward separation problem in the field.

Built from primary sources: peer-reviewed papers (Signal Transduction and Targeted Therapy, Biomedicines, Molecular Therapy — Methods & Clinical Development) and FDA-approved prescribing information / Last updated September 2026

Abstract conceptual image of small polyhedra of identical shape loosely arranged against a dark background — **Conceptual image (AI-generated).** An impression of large numbers of identically shaped containers lined up. **It does not represent real virus particles, an electron micrograph or a product.**

What this article covers

What a viral vector is (the short version)
Four vectors compared — integration, capacity, use
The "5 kb" wall for AAV
A materials engineer's view (1): a capacity limit is a formulation problem
How they are made — plasmids and cells
How lentiviral vectors are made
Empty and full capsids — an impurity that looks identical
Our calculation: how much is discarded to make one dose
A materials engineer's view (2): separation is hard because there is almost no difference
How titre and quality are measured
Where the field stands, seen through approved products
Glossary / References / Claim-to-source audit

How claims are labelled in this article

Sourced = stated in published material (link given)
Our calculation = a figure this article derived, with the assumptions spelled out
Not yet confirmed = a plan or a target with no confirmed track record
Structural summaries and readings from a materials or process-design standpoint are marked separately as Commentary.

A note on medical content

This article explains manufacturing and materials technology. It does not assess the efficacy or safety of any treatment, and it is not medical advice. Approval status and product specifications are quoted only where they appear in material published by the regulator (the FDA).

1. What a viral vector is (the short version)

A virus is a structure specialised in one thing: getting its own genes inside a cell. A viral vector borrows only that property, with the contents swapped for a therapeutic gene.

What it is for: the review by Bulcha and colleagues frames gene therapy as the in vivo delivery of therapeutic genes to patients by vectors based on retroviruses, adenoviruses and adeno-associated virusesSourced. There is also an ex vivo route, in which cells are taken out of the body, given the gene and returnedSourced
The three families in use today: the same review states that the three major vector strategies at present are those based on adenoviruses, adeno-associated viruses and lentivirusesSourced
The main constraint: a viral shell is of fixed size, so there is an upper limit on the length of gene it can hold. For AAV the review says it is inevitably limited to a capacity of roughly 5 kbSourced

The single most important line in this article

The hard part about a viral vector is not making the virus. It is getting rid of the shells that have nothing inside them (empty capsids) afterwards. Wright's paper states that with transient transfection of HEK293 cells, empty capsids can account for 50% to 90% of all AAV particles generated in the cell cultureSourced. The product and the impurity have almost the same shape — that is the central problem of this field.

2. Four vectors compared — integration, capacity, use

Fig. 1 Conceptual diagram (vector drawing). Each entry follows Bulcha et al. (2021) [Reference 1]. Setting the four side by side is this article's own arrangement and implies no ranking. The "use" entries are our summary of the applications the paper lists.

Vector	Length of gene it can carry	Integration into the host genome	What the paper says
AAV	About 5 kb	No (persists in circular form)	"The wild-type genome is about 4.7 kb" and vectors are "inevitably limited to a capacity of roughly 5 kb". Recombination of the ITRs circularises the genome, and in non-dividing, terminally differentiated cells it can be detected more than ten years after administration
Adenovirus	4.5 / 6.5 / 10.5 / about 36 kb	No	The first generation replaced the E1A/E1B region, giving 4.5 kb; an E1/E3 double deletion gave 6.5 kb; the second generation 10.5 kb; and the third generation (helper-dependent, "gutless") about 36 kb
Lentivirus	Up to 9 kb	Yes	The HIV-1 genome is 9.7 kb. "Lentiviral vectors are integrating vectors that allow long-term transgene expression. They have a packaging capacity of up to 9 kb"
Gammaretrovirus	—	Yes	"Lentiviral vectors can be delivered into post-mitotic and quiescent cells, whereas other retroviral platforms such as gammaretroviral vectors require cell division to establish infection"

All Sourced (Bulcha et al. 2021 [Reference 1]). Lentiviruses are a genus within the retrovirus family: the paper treats gammaretroviruses as an example of a "simple retrovirus" and HIV-1 (a lentivirus) as an example of a "complex retrovirus".

3. The "5 kb" wall for AAV

AAV is the most widely used vector for delivery inside the body, and its constraints are correspondingly clear.

What the paper says

A key consideration in AAV vector design is that the wild-type genome is about 4.7 kb. Vectors based on it are therefore inevitably limited to a capacity of roughly 5 kb. Every element needed for proper expression has to be shortened and minimised to fit inside this small capsid.Sourced

Both ends of the genome carry a 145-base ITR (inverted terminal repeat), which acts both as the origin of replication and as the packaging signalSourced.

What bites here is that the therapeutic gene is not the only thing that has to fit into those 5 kb. The ITRs, the promoter, the poly(A) signal — every part needed for expression goes into the same bag. Indeed, ZOLGENSMA, an approved AAV product, is described as a recombinant self-complementary AAV9 containing a transgene encoding the human SMN protein under the control of a cytomegalovirus enhancer/chicken β-actin hybrid promoterSourced.

4. A materials engineer's view (1): a capacity limit is a formulation problem

A materials engineer's view: what, and how much, goes into a bag of fixed size

In materials terms, the 5 kb ceiling is the same kind of constraint as having to keep a resin formulation adding up to 100%. Increase the main ingredient and something else has to give.

The main ingredient: the therapeutic gene itself
The additives: the promoter (when and where it is expressed), the poly(A) signal (stability) and the ITRs (essential for replication and packaging)
What cannot be trimmed: the ITRs are fixed at 145 bases each, twice overSourced

That ZOLGENSMA is self-complementary can be read the same waySourced. A self-complementary genome is packaged in double-stranded form, so the length that actually fits is smaller still (our commentary). That it was adopted anyway reads naturally as a property worth sacrificing length for.

The other important point is that if capacity runs short, changing vector is an option. A third-generation adenoviral vector can hold about 36 kbSourced. But on making them, Bulcha and colleagues note that the main challenge in HCAd production is reliably preventing contamination with helper virusSourced. Take the capacity and manufacturing gets harder. The same trade-off as in choosing a material (our commentary).

5. How they are made — plasmids and cells

AAV vectors are not made by growing a virus. Instead, the necessary parts are split across separate plasmids and delivered into cells together.

Fig. 2 Conceptual diagram (vector drawing). The make-up of the plasmids, the fact that HEK293 cells express E1a/E1b, that output "peaks 2 to 3 days after transfection", and the structure of the Sf9/baculovirus route all follow Bulcha et al. (2021) [Reference 1]. The drawing shows the logic of the process, not real equipment or layout.

(1) Triple transfection of HEK293 cells

Bulcha and colleagues state that preclinical and research-grade AAV is still produced mainly by the standard triple transfection method in HEK293 cellsSourced. The three plasmids are as followsSourced.

The cis plasmid: the transgene cassette of interest, flanked by ITRs
The trans plasmid: the rep and cap genes (the blueprints for replication and for the capsid)
The helper plasmid: E2a/E2b, E4 and VA RNA

Because HEK293 cells are transformed by adenovirus, they supply the essential E1a and E1b themselvesSourced. In other words, the cell itself provides one of the parts. Culture is semi-adherent, output peaks 2 to 3 days after transfection, and viability declines thereafterSourced.

(2) Packaging and producer cell lines

Stable cell lines that express the vector components constitutively have also been developedSourced. Bulcha and colleagues note, however, that unlike HEK293-based platforms, these stable lines do not express E1a/E1b, so they require infection with a replication-competent adenoviral helper or an Ad-AAV hybrid virusSourced. Convenience comes at the price of another source of impurity — the helper virus.

(3) Insect cells and baculovirus

The option that avoids mammalian cells altogether uses Sf9 cells derived from the fall armyworm together with the baculovirus expression vector (BEV) systemSourced. One BEV carries the vector genome flanked by ITRs, the other carries rep and cap, and both are used to infect Sf9 cellsSourced. Because the baculovirus itself supplies the helper functions, pairing it with an Sf9 line that stably expresses rep allows flexible, high-titre, large-scale production from a single-BEV systemSourced.

6. How lentiviral vectors are made

Lentiviral vectors are used where cells are engineered outside the body — in the manufacture of CAR-T cells, for example. To improve safety, the design principle of splitting the viral components across several plasmids has been pushed further with each generationSourced.

Fig. 3 Conceptual diagram (vector drawing). The contents of each plasmid, the self-inactivation mechanism, and the titre and safety points about HEK293T follow Bulcha et al. (2021) [Reference 1]. Drawing it as four parts is this article's own arrangement and does not show the real plasmid structures.

Generation	Change in design	Purpose
First (three plasmids)	env replaced with VSV-G and supplied from a separate plasmid; the transgene on a third plasmid	Pseudotyping with VSV-G lets the vector enter cells without depending on the native host receptor
Second	The accessory genes vif, vpr, vpu and nef removed	Take out genes involved in viral propagation and pathogenicity
Third (four plasmids)	tat and rev taken out of the packaging construct, with rev supplied from a separate plasmid; and the part of the 3' LTR containing the TATA box and related elements deleted	To "reduce the formation of unintended replication-competent proviruses" and to create a self-inactivating (SIN) vector packaging system

All Sourced (Bulcha et al. 2021 [Reference 1]). For clinical use, manufacture is by simultaneous transfection of the four plasmids, routinely into HEK293 or HEK293T cells.

The manufacturing weak point the paper identifies

While the four-plasmid transfection method is highly modular, batch-to-batch variability in viral titre remains a challengeSourced. The alternative is a stable producer cell line that functionally retains the lentiviral helper and packaging genes. In that case, because VSV-G and gag-pol are toxic to the cells, a tetracycline-inducible system is conventionally used to control their expressionSourced.

7. Empty and full capsids — an impurity that looks identical

This is the heart of the article.

What the paper says

An empty AAV capsid consists of essentially the same AAV capsid shell as the product, but with no nucleic acid molecule packaged inside.Sourced

Empty capsids are known to arise in high proportion in several AAV vector production systems, including transient transfection of HEK293 cells, and can account for 50% to 90% of all AAV particles generated in the cell culture.Sourced

A purification process that does not remove enough of this impurity can leave a final product with empty capsids in excess (say tenfold) over the intended vectorSourced.

Gimpel and colleagues call this proportion the content ratio: the share of full capsids among all capsids, and state that it can vary from below 1% to 30% in production cultures at harvestSourced. They add that it is not consistent between production lotsSourced.

Fig. 4 Conceptual diagram (vector drawing). The definition of an empty capsid and the 50 to 90% figure follow Wright (2014) [Reference 2]; the full-capsid share of under 1% to 30%, and the principles of density-gradient ultracentrifugation and anion exchange, follow Gimpel et al. (2021) [Reference 3]. The circles and curve are schematic and do not show the real shape, size or internal structure of the particles.

Why empty capsids are a problem

Wright names two concernsSourced.

An immunological concern: they can strengthen the adaptive immune response to viral capsid antigens
An efficiency concern: by competing for binding sites on target cells, they can lower the transduction efficiency of the intended vector, which pushes the required dose up

At the same time, the paper mentions reports that empty capsids can act as a decoy, using their immunological similarity to reduce neutralisation of AAV vectors by pre-existing antibodiesSourced. Less a uniform villain than a quantity to be controlled (our commentary).

How they are separated

Wright lists the following means of separationSourced.

Density-gradient ultracentrifugation: methods using caesium chloride or iodixanol are effective but present scalability challenges
Ion exchange chromatography: reported for AAV2, AAV6 and AAV5

Gimpel and colleagues explain the principle of anion exchange chromatography (AEC) as separating empty from full capsids on a positively charged stationary phase, based on the difference in isoelectric point that arises from the negatively charged genome insideSourced. The only thing that makes a difference is the charge on the nucleic acid inside.

8. Our calculation: how much is discarded to make one dose

First, the published figures

Titre range: 10¹⁰ to 10¹¹ vg/mL in the crude culture at harvest, and 10¹³ to 10¹⁴ vg/mL in the purified productSourced
The starting point for purification: as a scale able to support a phase 1 trial, Wright takes the recovery of 10¹⁵ highly purified vector particles, which he puts at about 10 mg (about 6.5 mg of capsid protein plus about 3.5 mg of vector DNA)Sourced
What it is mixed with: about 4 g of non-vector protein (3 g from about 10¹⁰ HEK293 cells plus about 1 g from fetal bovine serum) and about 350 mg of non-vector nucleic acid (320 mg from HEK293 cells plus 30 mg of production plasmid DNA)Sourced
The removal required: to bring impurities to 2% or less of the product, protein has to be reduced selectively by about 10⁵-fold and nucleic acid by about 10⁴-foldSourced
Doses of approved products: ZOLGENSMA is 1.1 × 10¹⁴ vg/kg at a nominal concentration of 2.0 × 10¹³ vg/mLSourced. For LUXTURNA the diluted dose is 1.5 × 10¹¹ vg in 0.3 mLSourced

Our calculation

Converting several of the figures aboveOur calculation:

Contents of one vial: 2.0 × 10¹³ vg/mL × 5.5 mL = 1.1 × 10¹⁴ vg
Dose for a body weight of 8 kg (assumption: 8 kg): 1.1 × 10¹⁴ × 8 = 8.8 × 10¹⁴ vg, which is almost 10¹⁵ particles
Eye versus whole body: 8.8 × 10¹⁴ ÷ 1.5 × 10¹¹ = about 5,900 times. Even with the same AAV, the amount needed varies by three to four orders of magnitude with the route of administration
The mass involved (assumption: applying Wright's conversion of 10¹⁵ vg to about 10 mg as it stands): 8.8 × 10¹⁴ vg is about 8.8 mg. The active ingredient in one dose comes to less than 10 mg
Mass ratio to what it is mixed with: 4 g ÷ 6.5 mg = about 615 times. Purification starts from a liquid holding more than 600 times as much protein as the capsid protein being sought
Culture volume needed (assumptions: 10¹¹ vg/mL at harvest, 100% recovery): 10¹⁵ ÷ 10¹¹ = 10⁴ mL = about 10 L. Change the assumption to 10¹⁰ vg/mL and it becomes 100 L
Allowing for empty capsids: if only 10% of capsids are full, then ten times as many capsids are being made to obtain the same number of full ones

Assumptions and limits: the 8 kg body weight is an assumption set by this article; real doses are determined per patient. Wright's mass conversion is a model example given in that paper and varies with serotype and genome length. The culture volume is a lower bound resting on the unrealistic assumption of 100% recovery; a real process loses material at every clarification and purification stage. These figures also say nothing about therapeutic meaning or safety.

Fig. 5 Drawn with our calculation included (vector drawing). The 10 mg, 4 g, and the 10⁵-fold and 10⁴-fold reductions are model figures Wright (2014) [Reference 2] gives to explain the level of purification required. The about 615 times is calculated by this article and is not a published value. Real values vary with manufacturing conditions.

9. A materials engineer's view (2): separation is hard because there is almost no difference

A materials engineer's view: in effect, only two property differences are available

An ordinary separation step works from a difference in physical properties between product and impurity: size, density, charge, hydrophobicity, solubility. Empty and full capsids, however, are shells of the same protein and the same size. Wright states that removal of empty capsids is difficult because they are structurally similar to the intended vector productSourced.

The usable differences are, in effect, only the two that come from having DNA inside.

Density: heavier by the mass of the nucleic acid, hence density-gradient ultracentrifugation with caesium chloride or iodixanolSourced
Charge: nucleic acid is negatively charged, hence anion exchange chromatography using the difference in isoelectric pointSourced

And here is where process design gets difficult. Density-gradient ultracentrifugation works well but is explicitly noted as hard to scale upSourced. Chromatography scales readily, but then the separation rests on a small difference in isoelectric point. The method that works does not scale; the method that scales lacks resolution — a pattern often seen in materials separation (our commentary).

Wright further notes that a process unable to separate empty capsids would not be expected to separate other product-related impurities either, such as AAV particles containing DNA fragments from the host cell or the helperSourced. If one separation axis fails, everything that sits on that axis stays in. Separation steps are not independent: they work, or fail to work, in bundles defined by the property difference they exploit (our commentary).

A materials engineer's view: process-related or product-related changes what you can do about it

Wright points out that residual host-cell DNA exists in two formsSourced.

Nuclease-sensitive, process-related impurity: carried along non-specifically with the vector
Nuclease-resistant, product-related impurity: packaged inside the AAV particles

The first can be removed by enzyme treatment or washing; the second is inside the shell, so an enzyme applied from outside cannot reach it. Wright states that minimising these different forms of residual host-cell DNA requires separate manufacturing process optimisation strategiesSourced.

Put in materials-process language: "dirt on the surface" and "foreign matter taken up inside" are different problems. The first can be cleaned up downstream; the second is settled at the moment of formation and can only be reduced by upstream conditions. It is the same logic behind Wright setting out two routes to fewer empty capsids: (1) optimising the conditions under which they form in culture (raising packaging efficiency and lowering the empty-to-full ratio) and (2) optimising the purification processSourced (the grouping is ours).

10. How titre and quality are measured

The amount of vector cannot be captured in a single number. There are at least three kinds of "titre"Sourced.

What is measured	Typical method	What Gimpel et al. state
Genome titre	ddPCR, qPCR	ddPCR has a reproducibility of 2 to 10% and takes 1 to 2 hours. It needs sample preparation to remove DNA outside the capsid and to denature protein
Capsid titre	ELISA, BLI	ELISA has a reproducibility of 10 to 20% and takes 2 to 5 hours
Content ratio (share of full capsids)	AEC, AUC, CDMS	AEC has a reproducibility of under 1% to 4%, takes 30 minutes and needs no purification. AUC has a reproducibility of 2%, takes 6 hours and needs purification
Infectious titre	Cell-based assays	"Quantification takes 1 to 3 days and variability spans orders of magnitude. It is therefore not practical for timely determination of product titre"
Aggregation	AUC and others	A single AUC experiment can assess both the presence of aggregates and the content ratio

All Sourced (Gimpel et al. 2021 [Reference 3]). The paper also notes that the content ratio has recently been added to regulatory product specifications for virus-based gene therapiesSourced.

On high-resolution methods, Gimpel and colleagues state that AUC remains the only practical method able to quantify partially filled capsids, but requires large amounts of sampleSourced. The fast method has low resolution; the high-resolution method is slow and eats sample — the same trade-off again (our commentary).

Conceptual image of a small plain glass vial on a dark surface holding a small volume of clear, very pale liquid — **Fig. 6 Conceptual image (AI-generated).** An impression of the fact that the final product is a very small volume of clear liquid. **It does not represent a real product, container, label or contents.**

11. Where the field stands, seen through approved products

The specifications of three products are set out below, as far as the FDA prescribing information states them. Every quotation is there to show what kind of thing is made and how it is handled, not how it works clinically.

Product	Vector	What the label states	Handling
ZOLGENSMA (onasemnogene abeparvovec)	AAV9 (self-complementary)	"A recombinant self-complementary AAV9 containing a transgene encoding the human SMN protein under the control of a cytomegalovirus enhancer/chicken β-actin hybrid promoter". Nominal concentration 2.0 × 10¹³ vg/mL, dose 1.1 × 10¹⁴ vg/kg. Supplied as a kit of 2 to 14 vials	Shipped frozen at −60 °C or below. Moved to 2 to 8 °C immediately on receipt, where it is stable for 14 days. Must not be refrozen
LUXTURNA (voretigene neparvovec-rzyl)	AAV2	"A live, non-replicating AAV serotype 2 that has been genetically modified to express the human RPE65 gene". The vial is 5 × 10¹² vg/mL, diluted tenfold to a dose of 1.5 × 10¹¹ vg in 0.3 mL. "May contain residual amounts of HEK293 cell DNA and protein, and trace amounts of fetal bovine serum"	Diluted before administration
KYMRIAH (tisagenlecleucel)	Lentiviral vector	"Autologous T cells genetically modified using a lentiviral vector to encode an anti-CD19 chimeric antigen receptor (CAR)". Prepared from peripheral blood mononuclear cells obtained by standard leukapheresis, activated with anti-CD3/CD28 antibody beads, expanded in culture and cryopreserved	Shipped as a frozen suspension in a patient-specific infusion bag. Must pass sterility testing before release

All Sourced (FDA-approved prescribing information [References 4, 5 and 6]). Indications and clinical results are not covered in this article.

What one sentence on the LUXTURNA label tells you

The very fact that a label states "may contain residual amounts of HEK293 cell DNA and protein, and trace amounts of fetal bovine serum" captures the character of this fieldSourced. What comes from the production cells cannot be taken to zero. So instead you measure how far it has been reduced and control it as a specification — the 10⁵-fold and 10⁴-fold requirements in Section 8 are the other side of that sentence (our commentary).

What remains difficult in manufacturing

Productivity: Gimpel and colleagues note that the virus concentration in a sample spans several orders of magnitude, from 10¹⁰ to 10¹¹ vg/mL at harvest to 10¹³ to 10¹⁴ vg/mL after purificationSourced. Doses are of the order of 10¹⁴, while the starting concentration is in the 10¹⁰ range
Lot-to-lot variability: the share of full capsids is not consistent between production lotsSourced. For lentiviruses too, batch-to-batch variability in titre remains a challengeSourced
The cost of measurement itself: Gimpel and colleagues note that the sample volumes analysis requires can result in significant loss of productSourced
Cost: no figure is given here, because manufacturing cost could not be confirmed in a primary source within the scope of this article.

The article in summary

An AAV can carry only about 5 kb, and the ITRs, promoter and poly(A) signal go into the same bagSourced
If capacity is what you need, adenovirus (about 36 kb); if you are engineering cells outside the body, lentivirus (up to 9 kb, integrating) — that is how the roles divideSourced
The standard way to make AAV is to put three plasmids into HEK293 cells. Insect cells with baculovirus are another routeSourced
Between 50 and 90% of the capsids made can be empty — an impurity hard to tell from the productSourced
Only two property differences are available for separation: density and charge. Ultracentrifugation works well but is explicitly noted as hard to scaleSourced
Against about 10 mg of product there is about 4 g of non-vector protein — a mass ratio of about 615 timesOur calculation

12. Glossary

Vector: The container that carries a gene into a cell. One that uses a viral shell is a viral vector.

AAV: Adeno-associated virus. Its genome is about 4.7 kb and vector capacity about 5 kb.

Capsid: The protein shell that encloses a viral genome.

Empty capsid: A capsid with no genome inside, treated as a product-related impurity.

ITR: Inverted terminal repeat. The 145-base sequence at each end of the AAV genome, needed for replication and packaging.

Plasmid: Circular DNA delivered into cells. In vector manufacture the parts are split across several of them.

Transfection: Introducing DNA and the like into cells. For AAV the standard is three plasmids delivered together.

HEK293 cells: A human embryonic kidney cell line, transformed by adenovirus and expressing E1a and E1b.

Sf9 cells: Insect cells from the fall armyworm, used with baculovirus to produce AAV.

Lentivirus: A genus within the retrovirus family, able to deliver genes into cells that are not dividing.

VSV-G: The glycoprotein of vesicular stomatitis virus. Pseudotyping with it removes dependence on the native receptor.

Self-inactivating (SIN): A design deleting part of the 3' LTR so that transcription is unlikely after integration.

vg (vector genome): The unit counting particles that contain a genome, used to state titre.

Content ratio: The share of full capsids among all capsids. One of the quality attributes.

ddPCR: Digital droplet PCR, used to quantify genome titre.

AUC: Analytical ultracentrifugation, which assesses aggregation and content ratio from sedimentation behaviour.

AEC: Anion exchange chromatography, which separates empty from full using the difference in isoelectric point.

Iodixanol: One of the media used for density-gradient ultracentrifugation, alongside caesium chloride.

13. References (primary sources)

Bulcha, J.T., Wang, Y., Ma, H., Tai, P.W.L., Gao, G. "Viral vector platforms within the gene therapy landscape", Signal Transduction and Targeted Therapy 6, 53 (2021). doi:10.1038/s41392-021-00487-6 — pmc.ncbi.nlm.nih.gov
Wright, J.F. "Product-Related Impurities in Clinical-Grade Recombinant AAV Vectors: Characterization and Risk Assessment", Biomedicines 2(1), 80–97 (2014). doi:10.3390/biomedicines2010080 — pmc.ncbi.nlm.nih.gov
Gimpel, A.L. et al. "Analytical methods for process and product characterization of recombinant adeno-associated virus-based gene therapies", Molecular Therapy — Methods & Clinical Development 20, 740–754 (2021). doi:10.1016/j.omtm.2021.02.010 — pmc.ncbi.nlm.nih.gov
FDA "ZOLGENSMA (onasemnogene abeparvovec-xioi)" prescribing information (revised June 2026) — fda.gov
FDA "LUXTURNA (voretigene neparvovec-rzyl)" prescribing information (revised May 2022) — fda.gov
FDA "KYMRIAH (tisagenlecleucel)" prescribing information (revised June 2025) — fda.gov

14. Claim-to-source audit

Claim in the text	Basis	Label
That gene therapy uses vectors based on retroviruses, adenoviruses and AAV, and that an ex vivo route exists. That the three major families at present are adenovirus, AAV and lentivirus. That the AAV genome is about 4.7 kb, the ITR 145 bases, and vector capacity limited to roughly 5 kb. That adenoviral vector capacity by generation is 4.5, 6.5, 10.5 and about 36 kb; that the third generation is called "gutless" or helper-dependent; and that the main manufacturing challenge is reliably excluding helper virus. That the HIV-1 genome is 9.7 kb, that lentiviral vectors carry up to 9 kb and integrate, that they can be delivered into post-mitotic and quiescent cells while gammaretroviruses require cell division. That lentiviral DNA integrates non-randomly, favouring transcriptionally active sites. That the AAV genome circularises through ITR recombination and can be detected for more than ten years in non-dividing, terminally differentiated cells, and that AAV is regarded as the least immunogenic of the current vectors. The contents of the cis, trans and helper plasmids in triple transfection, that HEK293 cells express E1a and E1b, and that output peaks in 2 to 3 days. That stable cell lines lack E1a/E1b and require helper virus infection. The structure of the Sf9/baculovirus route and high-titre large-scale production from a single-BEV system with a rep-expressing Sf9 line. The first to third generation design changes for lentiviral vectors (VSV-G, removal of accessory genes, separation of tat and rev, self-inactivation by 3' LTR deletion), manufacture with four plasmids, the use of HEK293 and HEK293T and the SV40 large T antigen concern, batch-to-batch variability, and stable producer lines using a tetracycline-inducible system	Bulcha et al. (2021)[Reference 1] https://pmc.ncbi.nlm.nih.gov/articles/PMC7868676/	Sourced
The definition of an empty capsid. That empty capsids can account for 50 to 90% of all AAV particles in production systems including transient transfection of HEK293 cells; that a process that does not remove them can leave them in excess (say tenfold) over the intended vector; and that lot-to-lot variation is large. That empty capsids can strengthen the adaptive immune response to capsid antigens and can lower transduction efficiency by competing for binding sites, while also being reported to act as a decoy reducing neutralisation by pre-existing antibodies. That their structural similarity makes removal difficult, and that a process unable to separate empty capsids would not be expected to separate particles containing host-cell or helper-derived DNA either. That density-gradient ultracentrifugation with caesium chloride or iodixanol is effective but presents scalability challenges, and that ion exchange separation has been reported for AAV2, AAV6 and AAV5. The two routes to fewer empty capsids (optimising culture conditions and optimising purification). That residual host-cell DNA exists both as a nuclease-sensitive process-related impurity and as a nuclease-resistant product-related impurity, each requiring its own optimisation strategy. That 10¹⁵ highly purified vectors correspond to about 10 mg (6.5 mg capsid protein plus 3.5 mg vector DNA); that non-vector protein amounts to about 4 g (3 g from about 10¹⁰ HEK293 cells plus about 1 g from fetal bovine serum) and non-vector nucleic acid to about 350 mg (320 mg from cells plus 30 mg plasmid DNA); and that bringing impurities to 2% or less of the product requires selective reduction of about 10⁵-fold for protein and about 10⁴-fold for nucleic acid	Wright (2014)[Reference 2] https://pmc.ncbi.nlm.nih.gov/articles/PMC5423478/	Sourced
That virus concentration runs from 10¹⁰ to 10¹¹ vg/mL at harvest to 10¹³ to 10¹⁴ vg/mL after purification. The definition of the content ratio, that it varies from under 1% to 30% at harvest and is not consistent between lots, and that it has recently been added to regulatory product specifications. That anion exchange separates empty from full using the difference in isoelectric point arising from the negatively charged genome inside. The reproducibility, time, sample requirements and need for purification of AEC, AUC, ddPCR, ELISA and CDMS. That infectious titre takes 1 to 3 days to quantify with variability spanning orders of magnitude, making it impractical for timely characterisation. That AUC remains the only practical method able to quantify partially filled capsids and requires large amounts of sample. That the sample volumes needed for analysis can result in significant loss of product	Gimpel et al. (2021)[Reference 3] https://pmc.ncbi.nlm.nih.gov/articles/PMC7940698/	Sourced
That ZOLGENSMA is a recombinant self-complementary AAV9 encoding the human SMN protein under a CMV enhancer/chicken β-actin hybrid promoter. Its nominal concentration of 2.0 × 10¹³ vg/mL, dose of 1.1 × 10¹⁴ vg/kg, extractable volumes of 5.5 mL or 8.3 mL, and supply as a kit of 2 to 14 vials. That it is shipped frozen at −60 °C or below, is stable for 14 days once moved to 2 to 8 °C, and must not be refrozen	FDA ZOLGENSMA prescribing information[Reference 4] https://www.fda.gov/media/126109/download?attachment	Sourced
That LUXTURNA is a live, non-replicating AAV serotype 2 modified to express the human RPE65 gene. That the vial is 5 × 10¹² vg/mL and the dose after tenfold dilution is 1.5 × 10¹¹ vg in 0.3 mL. That it may contain residual HEK293 cell DNA and protein and trace amounts of fetal bovine serum	FDA LUXTURNA prescribing information[Reference 5] https://www.fda.gov/media/109906/download?attachment	Sourced
That KYMRIAH is autologous T cells carrying an anti-CD19 CAR delivered by a lentiviral vector; that it is prepared from peripheral blood mononuclear cells obtained by leukapheresis, activated with anti-CD3/CD28 antibody beads, expanded in culture and cryopreserved; and that it must pass sterility testing before release	FDA KYMRIAH prescribing information[Reference 6] https://www.fda.gov/media/107296/download?attachment	Sourced
Putting one vial at 1.1 × 10¹⁴ vg and the dose at a body weight of 8 kg at 8.8 × 10¹⁴ vg. Putting the ratio to LUXTURNA at about 5,900 times. Converting 8.8 × 10¹⁴ vg to about 8.8 mg. The 4 g ÷ 6.5 mg = about 615 times. Putting the culture volume needed for 10¹⁵ vg at about 10 L (at 10¹¹ vg/mL) or 100 L (at 10¹⁰ vg/mL). The statement that a 10% full-capsid share means ten times as many capsids are being made	Our calculation. The 8 kg body weight, the application of Wright's mass conversion and the assumption of 100% recovery are all assumptions set by this article. Real values vary with serotype, genome length and production route. Nothing here addresses therapeutic meaning or safety	Our calculation
The manufacturing cost of an AAV vector	No figure is given, because no published primary source could be confirmed within the scope of this article	Commentary
Differences in productivity between serotypes, and the process conditions or yields of particular companies	Not stated, because the primary sources consulted do not give process conditions for individual companies	Commentary
The indications, clinical results and safety assessment of each product	This article explains manufacturing and materials technology and does not assess therapeutic efficacy or safety. For approved products it quotes only statements on manufacture, specification and handling that appear on the label	Commentary
How far the reduction of empty capsids and scale-up will progress in future	At the time of writing (September 2026) this article could not confirm any primary source indicating a future end point	Not yet confirmed
Likening the 5 kb limit to keeping a formulation adding up to 100%; the reading that a self-complementary genome leaves less usable length; framing capacity against manufacturing difficulty as a trade-off; the summary that only density and charge are available as property differences; the reading that "the method that works does not scale and the method that scales lacks resolution"; likening process-related and product-related impurities to "dirt on the surface" and "foreign matter taken up inside"; the reading that empty capsids are a quantity to control rather than a uniform villain; and reading the level of purification required from the residual-component sentence on the LUXTURNA label	Our summary and commentary based on published content. Not views expressed by the authors of the papers or by the regulator	Commentary
That Figs. 1, 2, 3, 4 and 5 are explanatory drawings rather than real structures, equipment or observations, and that the hero image and Fig. 6 are AI-generated images	Our note	Commentary

Last updated 23 September 2026. Sources are limited to primary material (peer-reviewed papers and FDA-approved prescribing information). Because the article includes structural summaries and readings from a process-design standpoint, those are marked as Commentary and kept separate from sourced fact. Manufacturing costs, process conditions and yields at individual companies, and differences in productivity between serotypes are not stated here because they could not be confirmed in published primary sources. This article explains manufacturing technology; it does not assess the efficacy or safety of any treatment, and it is not medical advice. All figures are explanatory concept graphics. Figs. 1 to 5 are vector drawings; the hero image and Fig. 6 are AI-generated images, and none of them is a photograph of a real particle, piece of equipment or product.