Engineering Guide · Vendor-Neutral

AAV vs Lentivirus Production: Manufacturing, Yield and Use-Case Guide

AAV vs lentivirus vector side-by-side comparison ~25 nm ssDNA · 4.7 kb IN VIVO · episomal 1E11–1E14 vg/mL AAV VS IN 100–200 nm RNA · 8 kb · integrating EX VIVO · CAR-T · HSC 1E7–1E10 TU/mL Lentivirus
Figure 1: AAV is a small non-enveloped icosahedral capsid delivering episomal ssDNA. Lentivirus is a larger enveloped particle delivering integrating RNA. Size, envelope and integration differences drive almost every downstream manufacturing decision.
Quick Verdict

Choose based on delivery route, not brand. AAV wins for in vivo gene therapy into non-dividing tissue with transgenes under 4.7 kb. Lentivirus wins for ex vivo modification of dividing cells (CAR-T, hematopoietic stem cells) and larger payloads up to 8 kb. AAV scales to 2000 L bioreactors. Lentivirus is capped by a 48 to 72 hour productive window and typically runs 50 to 500 L.

Key differences at a glance

Side-by-side comparison

Factor AAV Lentivirus
Particle size ~25 nm, non-enveloped 100 to 200 nm, enveloped
Genome ssDNA ssRNA (reverse-transcribed to DNA)
Packaging capacity ~4.7 kb ~8 kb
Integration into host genome Episomal (very low integration) Integrating (SIN-LV reduces oncogenic risk)
Dividing cell transduction Poor (episome diluted by division) Excellent (passed to daughters)
Non-dividing cell transduction Excellent (retina, CNS, muscle, liver) Good (VSV-G pseudotyped)
Typical crude titer 1E11 to 1E14 vg/mL (physical) 1E7 to 1E10 TU/mL (functional)
Bioreactor scale 200 to 2000 L stirred tank / SUB 50 to 500 L (VSV-G toxicity ceiling)
Storage / cold chain 2 to 8 °C for weeks; −80 °C for long term −80 °C mandatory; 20 to 50% titer loss per freeze-thaw
Delivery model In vivo (direct patient injection) Ex vivo (apheresis to transduce to reinfuse)

Values reflect typical published specifications for suspension HEK293 production. Vendor datasheets and your specific serotype or pseudotype take precedence. See our AAV production yield guide and lentiviral vector production guide for parameter-level detail.

AAV in detail

Adeno-associated virus is a small, non-pathogenic parvovirus with a 25 nm icosahedral capsid and a 4.7 kb single-stranded DNA genome. Recombinant AAV (rAAV) is produced by removing the wild-type Rep and Cap genes from the vector genome (keeping only the two ITRs flanking the transgene of interest) and supplying Rep and Cap in trans, along with adenoviral helper functions (E2A, E4, VA RNA). Thirteen natural serotypes and hundreds of engineered capsid variants let researchers target specific tissues: Spark Therapeutics uses AAV2 for the retina in Luxturna, Novartis uses AAV9 for CNS in Zolgensma, and Sarepta Therapeutics uses AAVrh74 for skeletal muscle in Elevidys.

How it works

Production is dominated by transient triple transfection of suspension HEK293 cells with plasmids encoding the ITR-transgene, RepCap, and adenoviral helper functions. Delivery uses Polyplus PEIpro, Thermo Fisher ExpiFectamine or an equivalent cationic reagent. AAV assembles inside the cell nucleus over 72 hours. Downstream harvest uses cell lysis (Triton, Tween, or freeze-thaw) then endonuclease digestion, clarification, affinity capture on an AAVX or AAV9-specific resin, and anion exchange or CsCl gradient to separate full from empty capsids. Alternative platforms include stable producer cell lines from Cytiva (ELEVECTA), Asimov (AAV Edge) and Lonza, plus the baculovirus/Sf9 system used by Roche for Elevidys manufacturing.

When AAV wins

AAV dominates when the therapeutic gene fits under 4.7 kb and the delivery target is a non-dividing tissue accessible by injection: retinal photoreceptors, spinal motor neurons, hepatocytes, cardiomyocytes, skeletal muscle. Because the episomal genome is not diluted by cell division, expression can persist for years in a quiescent tissue from a single dose. Serotype engineering gives fine control over tissue tropism, which is not really possible for lentivirus. The AAV Yield Calculator lets you estimate vg per batch from cell density and specific productivity.

Lentivirus in detail

Lentivirus is a genus of retroviruses that includes HIV-1, from which most third-generation lentiviral vectors are derived. The vector particle is 100 to 200 nm with a lipid envelope displaying an envelope glycoprotein (usually VSV-G for broad tropism, but measles H/F, baboon endogenous, or cocal envelope are also used). Inside the envelope, a conical capsid holds two copies of about 8 kb single-stranded RNA. On entry, reverse transcriptase converts the RNA to double-stranded DNA and integrase inserts it into the host chromosome, giving permanent expression that survives cell division. The three-plasmid third-generation SIN system splits gag-pol, rev and env onto separate plasmids and deletes the U3 enhancer in the LTR, lowering the risk of insertional oncogenesis. Vendors and CDMOs in this space include Oxford Biomedica, Lonza, WuXi Advanced Therapies, MilliporeSigma and Yposkesi.

How it works

Suspension HEK293T cells are transfected with four plasmids: transfer (containing the transgene between LTRs), gag-pol (structural proteins), rev (RNA export) and envelope (VSV-G by default). PEI delivers the plasmid cocktail at 1 to 2 µg total DNA per million cells. Lentivirus buds continuously from the plasma membrane, so harvest is from the supernatant rather than cell lysate. The productive window is 48 to 72 hours before VSV-G cytotoxicity limits further production. Downstream uses tangential flow filtration for concentration, benzonase to remove residual DNA, mixed-mode or anion exchange chromatography, and sterile filtration. Cold chain is unforgiving: lentivirus loses 20 to 50% functional titer per freeze-thaw, so fill-finish and shipping are engineered around single-use aliquots stored at −80 °C. See our detailed lentiviral vector manufacturing guide for a 50 L bioreactor walk-through.

When lentivirus wins

Lentivirus dominates ex vivo cell therapy. When you extract T cells or CD34+ hematopoietic stem cells from a patient, transduce them in a closed-system bag or bioreactor with lentiviral vector, and reinfuse, the modified genes need to persist through the many divisions those cells undergo after re-engraftment. That is only possible with an integrating vector. All approved CAR-T products (Kymriah, Yescarta, Tecartus, Breyanzi, Abecma, Carvykti) and the approved lentiviral HSC therapies (Zynteglo, Skysona, Lyfgenia) rely on this integration property. The 8 kb payload also matters for constructs like bicistronic CARs with a safety switch, or gene addition of dystrophin fragments where AAV's 4.7 kb ceiling cannot fit the required cargo.

Pros and cons

AAV

Advantages

  • Non-pathogenic, non-integrating, low insertional mutagenesis risk.
  • Scales cleanly to 2000 L stirred-tank on a robust suspension HEK293 platform.
  • Thirteen or more natural serotypes and hundreds of engineered capsids for tissue-specific targeting.
  • Direct in vivo delivery: no apheresis, no ex vivo cell processing, no reinfusion.

Disadvantages

  • 4.7 kb packaging limit rules out large genes (dystrophin, CFTR, factor VIII require truncation or dual-vector tricks).
  • Pre-existing neutralising antibodies in 30 to 60% of adults exclude patients or require immune suppression.
  • High per-dose vector requirement (up to 1E14 vg/kg systemic) drives GMP manufacturing cost.
  • Empty capsid removal is a downstream burden (10 to 70% of physical particles are empty in crude harvest).

Lentivirus

Advantages

  • Permanent integration gives durable expression in dividing cells (T, B, HSC).
  • ~8 kb payload accommodates bicistronic CARs, multi-cassette editors, larger transgenes.
  • Broad tropism via VSV-G pseudotyping; alternative envelopes for cell-selective targeting.
  • Strong clinical precedent: six approved CAR-T products and three HSC therapies as of 2026.

Disadvantages

  • Insertional oncogenesis risk remains, even with third-generation SIN vectors.
  • 48 to 72 hour VSV-G cytotoxicity ceiling caps productive window and complicates scale-up.
  • Unforgiving cold chain (−80 °C mandatory; 20 to 50% titer loss per freeze-thaw).
  • Requires ex vivo cell processing infrastructure, apheresis, and closed-system transduction.

Which should you choose?

The dominant constraint is almost always delivery route and cell biology, not manufacturing preference. These four scenarios cover most real programmes.

In vivo, non-dividing tissue, gene under 4.7 kb

Retinal dystrophy, spinal motor neuron disease, monogenic liver disease, cardiomyopathy. Direct injection into a stable tissue with a small transgene is exactly what AAV was engineered for.

Choose AAV

Ex vivo modification of dividing cells (CAR-T, HSC)

B-cell lymphoma, sickle cell disease, beta-thalassemia, cerebral adrenoleukodystrophy, iPSC engineering. Cells will divide many times post-transduction and need heritable expression.

Choose Lentivirus

Transgene over 5 kb or multi-cassette construct

Full-length dystrophin, CFTR, factor VIII, epigenetic editors with dual promoter cassettes. AAV cannot package without truncation or dual-vector reconstitution.

Choose Lentivirus

Single systemic dose, non-integrating preferred

Haemophilia B (Hemgenix), Duchenne (Elevidys), spinal muscular atrophy (Zolgensma). Regulator preference for episomal expression in monogenic in vivo indications is strong.

Choose AAV

Real-world use cases

Approved products and late-stage programmes have converged on one platform or the other for reasons that reveal the underlying decision logic.

In vivo · CNS · AAV9
Zolgensma (spinal muscular atrophy)

Novartis dosed at 1.1E14 vg/kg IV to cross the blood-brain barrier via AAV9 and transduce spinal motor neurons. Single dose, episomal, no ex vivo processing possible in newborns. AAV was the only viable route.

Ex vivo · CAR-T · Lentivirus
Kymriah (B-ALL, DLBCL)

Novartis apheresed patient T cells, transduced with a lentiviral vector encoding anti-CD19 CAR under EF1α, expanded, and reinfused. Integration was required to survive the many divisions during expansion and post-infusion.

In vivo · Liver · AAV5
Hemgenix (haemophilia B)

CSL Behring uses AAV5 to deliver factor IX to hepatocytes at 2E13 vg/kg. Liver is a stable episome-friendly tissue and factor IX cDNA is 1.4 kb, well under the 4.7 kb ceiling. A single systemic dose replaces lifelong infusions.

Ex vivo · HSC · Lentivirus
Zynteglo (beta-thalassemia)

bluebird bio harvests CD34+ HSCs, transduces with lentivirus encoding a modified beta-globin gene, and reinfuses after myeloablation. Integration is essential for stable expression across the patient's lifetime of red-cell turnover.

Estimating AAV yield for your batch?

The AAV Yield Calculator converts cell density and specific productivity into vector genomes per batch across suspension bioreactor scales from 50 mL to 2000 L. Use it to sanity-check budget estimates before locking a manufacturing plan.

Open the AAV Yield Calculator

Cost and lifecycle considerations

Batch cost and per-dose cost tell different stories

An AAV batch is expensive because every dose is large. A lentiviral batch treats many patients because every dose is small. Compare on cost-per-treated-patient, not cost-per-batch, when evaluating which platform fits your programme economics.

A 200 L GMP AAV batch at a CDMO like Charles River or AGC Biologics is typically around US$2 million all-in, dominated by GMP plasmid DNA ($50k to $150k per gram, 2 to 4 g per batch), transfection reagent, media, and downstream affinity resin. Plasmid alone can account for 30 to 40% of variable cost.

A comparable 200 L lentiviral batch runs around US$1.5 million because plasmid quantities are smaller (four plasmids at roughly 0.5 g each) but per-dose vector requirements are also smaller. A CAR-T dose uses only microgram quantities of lentivirus, so a single 200 L batch can supply hundreds of patients. Independent cost-of-goods analyses estimate lentiviral vector as roughly $5,000 per CAR-T dose, versus $50,000 to $500,000 of AAV per systemic dose in indications like SMA or DMD.

Cost component AAV (200 L GMP batch) Lentivirus (200 L GMP batch)
Media + reagents~$400k~$300k
GMP plasmid DNA$200k to $400k$100k to $200k
Downstream (chromatography, empty removal)~$700k~$500k
QC + release testing~$400k~$400k
Total per batch (all-in, CDMO)~$2.0M~$1.5M
Doses per batch (typical)1 to 10 systemic; 100+ intravitreal200 to 1000 CAR-T

Vendor landscape

Major CDMOs and platform vendors in each camp, with one-line positioning notes.

AAV production vendors and CDMOs

Lentivirus production vendors and CDMOs

Frequently asked questions

What is the difference between AAV and lentivirus for gene therapy?
AAV is a small (~25 nm) non-enveloped icosahedral virus with a ~4.7 kb single-stranded DNA payload that remains largely episomal in the host nucleus, giving stable expression in non-dividing cells without integration. Lentivirus is a larger (100 to 200 nm) enveloped retrovirus with an ~8 kb RNA payload that reverse-transcribes and integrates into the host genome, giving permanent expression that persists through cell division. AAV is used for in vivo gene therapy delivered directly to patients. Lentivirus is used for ex vivo modification of patient cells (CAR-T, hematopoietic stem cells) that are then re-infused.
Which produces higher yields, AAV or lentivirus?
Volumetrically AAV produces more particles: optimised transient triple transfection in HEK293 typically reaches 1E11 to 1E14 vg/mL crude harvest, and stable producer lines can push to 1E15 vg/L pre-purification. Lentiviral yields are lower on a functional basis, at 1E7 to 1E9 transducing units per mL crude in suspension HEK293T, with best-in-class optimised processes reaching 1E10 TU/mL. However the two units are not directly comparable: AAV titers count physical vector genomes while lentivirus is usually reported in functional transducing units, and lentiviral processes are constrained by a 48 to 72 hour productive window before VSV-G envelope cytotoxicity limits production.
What is the packaging capacity of AAV versus lentivirus?
AAV packages approximately 4.7 kb of single-stranded DNA. Anything larger than 5.0 kb sees dramatic loss of packaging efficiency and truncated genomes. Lentivirus packages approximately 8 kb of RNA payload, so it can carry roughly twice the transgene size. This matters when delivering large genes (dystrophin, CFTR, factor VIII) or multi-cassette constructs (bicistronic CAR + safety switch, epigenetic editors).
Does lentivirus integrate into the genome and does AAV?
Lentivirus integrates its DNA into the host cell genome using integrase, giving permanent expression that is inherited by daughter cells during division. Third-generation self-inactivating (SIN) lentiviral vectors have deletions in the U3 region of the 3' LTR that reduce the risk of insertional oncogenesis but do not eliminate it. AAV remains predominantly episomal: the ssDNA is converted to dsDNA and persists as circular episomes in the host nucleus, which is diluted out by cell division but stable in non-dividing tissues (liver, retina, muscle, CNS). Wild-type AAV can integrate at the AAVS1 site on chromosome 19, but recombinant AAV lacking Rep integrates only at very low frequency.
Which is safer for clinical use, AAV or lentivirus?
AAV is non-pathogenic in humans and has an excellent safety record for in vivo administration, with the main risks being immune response (pre-existing neutralising antibodies limit dosing in seropositive patients) and hepatotoxicity at very high doses. Lentivirus carries integration risk because insertion near proto-oncogenes can drive clonal expansion, though third-generation SIN vectors and clinical experience across thousands of CAR-T and hematopoietic stem-cell gene therapy patients show this risk is manageable and lower than the earlier gamma-retroviral vectors it replaced. In practice both are considered safe enough for their respective indications when handled to standard biosafety practices: AAV for in vivo, lentivirus for ex vivo where cells can be monitored for integration site distribution.
How much does AAV manufacturing cost compared to lentivirus?
A GMP 200 L suspension batch of AAV drug substance runs around $2 million at a CDMO. A comparable lentiviral batch is around $1.5 million. Per dose the picture flips because ex vivo lentiviral products (CAR-T, HSC gene therapy) need only microgram quantities per patient. The lentiviral vector component of an approved CAR-T can be around $5,000 per dose, while a systemic AAV dose (Zolgensma at 1.1E14 vg/kg) consumes a large fraction of a batch and pushes vector cost to $50,000 to $500,000 per dose. Both are dominated by downstream processing: empty capsid removal for AAV and cold-chain plus small-scale filling for lentivirus.
Which approved gene therapies use AAV, and which use lentivirus?
AAV-based approvals include Luxturna (voretigene neparvovec, AAV2, retinal dystrophy), Zolgensma (onasemnogene abeparvovec, AAV9, spinal muscular atrophy), Elevidys (delandistrogene moxeparvovec, AAVrh74, Duchenne), Hemgenix (etranacogene dezaparvovec, AAV5, haemophilia B), and Roctavian (valoctocogene roxaparvovec, AAV5, haemophilia A). Lentiviral approvals cover Kymriah (tisagenlecleucel, CAR-T for B-ALL and DLBCL), Yescarta (axicabtagene ciloleucel, CAR-T), Breyanzi (lisocabtagene maraleucel, CAR-T), Abecma (idecabtagene vicleucel, BCMA CAR-T), Zynteglo (betibeglogene autotemcel, beta-thalassemia), Skysona (elivaldogene autotemcel, cerebral adrenoleukodystrophy) and Lyfgenia (lovotibeglogene autotemcel, sickle cell disease).
When should I choose AAV over lentivirus for a gene therapy programme?
Choose AAV when: (1) delivery is to a non-dividing tissue you can inject directly (retina, CNS, liver, skeletal muscle, cardiac muscle); (2) the therapeutic transgene fits within 4.7 kb; (3) durable expression from a stable episome is acceptable and patients tolerate the immunogenic profile of the target serotype. Choose lentivirus when: (1) delivery is ex vivo into dividing cells (T cells, hematopoietic stem cells, iPSC-derived progenitors) that must retain expression through many divisions; (2) the transgene is larger than 4.7 kb or you need a multi-cassette construct; (3) permanent integration is desirable for the indication; (4) manufacturing infrastructure and clinical protocol are already built around apheresis-then-transduce-then-reinfuse workflows.

Resources and references