Viral Vectors: The Major Types and How Each Delivers a Transgene

Choosing a viral vector is not a search for the “best” one. Four constraints decide it before preference enters the picture: how big your expression cassette is, whether the target cells divide, how long expression has to last, and whether the recipient already has antibodies against the capsid. Get any one of those wrong and no amount of optimisation rescues the experiment.

The most common and most expensive mistake is the first one. People budget the cargo limit against their coding sequence. The vector packages the whole construct — inverted terminal repeats or long terminal repeats, promoter, transgene, any tag or selection marker, polyadenylation signal, and regulatory elements. A 3.6 kb coding sequence that “fits comfortably” inside adeno-associated virus (AAV) frequently does not, once a strong promoter and a polyA are added. Size the cassette, not the gene.

The five vector classes and what each actually does

Published cargo limits vary between reviews because they depend on which generation of the vector you mean and how packaging efficiency is measured. The figures below are quoted from the sources listed at the end; where reviews disagree, both numbers are given rather than averaged.

Adeno-associated virus (AAV)

AAV is a small single-stranded DNA virus with a wild-type genome of about 4.7 kb, giving vectors roughly a 5 kb capacity. Delivery is receptor-mediated uptake followed by trafficking to the nucleus, where the vector genome circularises through recombination between its inverted terminal repeats and persists as a stable episome rather than integrating. That is why AAV gives durable expression — transgene product has been detected well beyond ten years — without the insertional-mutagenesis profile of a retroviral vector.

It is also why AAV suits post-mitotic tissue and is a poor choice for dividing cells: an episome is diluted out as cells divide. There are at least twelve natural serotypes across five clades and more than a thousand engineered variants. AAV2 binds heparan sulfate proteoglycans; engineered capsids such as AAV.PHP.B were evolved specifically to cross the blood–brain barrier.

The practical way to read serotype tropism is off the approved products, because each pairs a capsid with a route and a promoter:

• AAV2, subretinal — LUXTURNA (voretigene neparvovec), inherited retinal dystrophy.
• AAV5, intravenous, liver-directed — HEMGENIX (etranacogene dezaparvovec), a non-replicating recombinant AAV5 carrying Factor IX Padua under a liver-specific promoter (LP1).
• AAV9, systemic with CNS reach — ZOLGENSMA (onasemnogene abeparvovec), spinal muscular atrophy.
• AAVrh74, muscle — ELEVIDYS (delandistrogene moxeparvovec), Duchenne muscular dystrophy.

Note the HEMGENIX detail: targeting is not the capsid alone. A liver-specific promoter restricts expression even where the capsid distributes more broadly. If your specificity requirement is tight, the promoter is doing as much work as the serotype.

Lentiviral vectors

Lentiviral vectors are the workhorse where you need permanent modification of a cell you can grow. Their defining property is that the pre-integration complex is actively transported through an intact nuclear pore, so they transduce non-dividing cells — something gammaretroviruses cannot do.

Modern designs are split across plasmids so no single construct carries a replication-competent genome. Second-generation systems use three plasmids with the accessory genes vif, vpr, vpu and nef removed; third-generation systems use four and add the central polypurine tract and central termination sequence, producing the central DNA flap that raises transduction efficiency. The envelope is usually swapped for VSV-G, which broadens tropism to a wide range of cell types — and is the reason a lentiviral prep is not target-restricted in the way the parent virus was.

Self-inactivating (SIN) design is the safety feature that matters. A deletion in the U3 region of the 3’ LTR removes the TATA box and transcription-factor binding sites; during reverse transcription that deletion is copied into the 5’ LTR, so the integrated provirus cannot drive full-length viral transcription. Expression comes from an internal promoter you chose instead. If you are setting up the bench work, our lentiviral transduction protocol guide covers packaging, titering and MOI.

Gammaretroviral vectors

Murine leukaemia virus–based vectors integrate too, but they reach chromatin only when the nuclear envelope breaks down during mitosis. That makes them useless for quiescent targets and is the single clearest reason to prefer a lentiviral vector for primary cells.

The difference that shaped the whole field, though, is where the two integrate. Gammaretroviruses insert preferentially near transcription start sites, CpG islands and DNase I–hypersensitive sites — that is, next to promoters. Lentiviruses insert across the bodies of actively transcribed genes, so they are more likely to disrupt a gene than to switch one on.

The consequences were clinical. In the early SCID-X1 trials using LTR-driven gammaretroviral vectors, 4 of 9 patients in the French trial developed T-cell acute lymphoblastic leukaemia between 23 and 68 months after infusion, and 1 of 10 in the UK trial; LMO2 activation was the recurrent lesion, with CCND2 in one case. Two patients in an X-CGD trial developed myelodysplastic syndrome progressing to acute myeloid leukaemia, with MDS1/EVI1, PRDM16 and SETBP1 activation. The mechanism was the strong enhancer in the retroviral LTR acting on nearby proto-oncogenes. Removing that enhancer — SIN design with an internal cellular promoter — is the mitigation, and it is why LTR-driven vectors are no longer used clinically.

Lentiviral integration is safer, not safe: clonal dominance driven by aberrant HMGA2 splicing was seen in a lentiviral beta-thalassaemia trial. Both classes still integrate, and that is what triggers the regulatory obligations below.

Adenoviral vectors

Adenovirus carries a 26–45 kb linear double-stranded DNA genome and transduces quiescent and dividing cells with high efficiency. The DNA stays epichromosomal, so expression is transient — which makes it the right tool for vaccination, oncolytic and short-term local applications, and the wrong one for a chronic monogenic disease.

Capacity scales with how much of the viral genome you delete. First-generation vectors are quoted at up to 4.5 kb in one review and up to 8.2 kb in another; second-generation at up to 12 kb; helper-dependent (“gutless”) vectors at roughly 30–36 kb. If you need more than about 8 kb, helper-dependent adenovirus is one of the few options that exists at all.

The trade-off is immunogenicity. Adenovirus provokes strong innate responses to capsid protein and robust adaptive responses, and over 60% of adults carry neutralising antibodies. This is not an abstract risk: the 1999 death of Jesse Gelsinger followed an innate response to capsid protein that triggered a cytokine storm, and it reshaped gene therapy oversight. The approved adenoviral product, ADSTILADRIN (nadofaragene firadenovec), sidesteps systemic exposure entirely — it is a non-replicating adenovirus serotype 5 vector carrying interferon alfa-2b, instilled directly into the bladder.

HSV-1 vectors

Herpes simplex virus type 1 has a linear double-stranded DNA genome of about 152 kb, and that is the whole point: non-replicative genomic vectors carry more than 30 kb of cargo and amplicon vectors up to roughly 150 kb. Nothing else comes close. HSV-1 is natively neurotropic, with efficient retrograde axonal transport, and in sensory neurons the genome is maintained as an extrachromosomal episome with lytic genes epigenetically repressed — long-term persistence with no insertional mutagenesis.

Two HSV-1 products are approved: IMLYGIC (talimogene laherparepvec), a live attenuated HSV-1 expressing GM-CSF for melanoma, and VYJUVEK (beremagene geperpavec), a replication-defective HSV-1 expressing type VII collagen, applied topically for recessive dystrophic epidermolysis bullosa.

Integration sets your regulatory clock, not just your biology

This is the consequence people underestimate when they pick a vector for a programme with any translational ambition. FDA’s guidance on long term follow-up after administration of human gene therapy products ties observation duration directly to vector class:

• Fifteen years for integrating vectors — gammaretroviral, lentiviral, and transposon elements.
• Up to fifteen years for herpesvirus vectors (or oncolytics) capable of establishing latency.
• Up to fifteen years for microbial vectors known to establish persistent infection.
• Up to fifteen years for genome editing products.
• Up to five years for AAV vectors.

FDA also recommends testing for vector sequences by PCR in surrogate samples at intervals of no greater than six months for the first five years, then no greater than yearly for the next ten, until vector sequences are no longer detectable. Note that these are non-binding recommendations and a risk-based argument can modify them — but the default is a decade of difference in follow-up burden between an integrating vector and AAV. That belongs in the feasibility conversation, not at the IND stage. If your work is heading toward clinical manufacture, the GMP requirements for ATMPs diverge from standard GMP in ways worth understanding early.

Pre-existing immunity screens patients out before you dose

Neutralising antibodies against the capsid are an eligibility criterion, not a side note, and the thresholds are on the labels:

• ZOLGENSMA — clinical trial patients were required to have baseline anti-AAV9 antibody titers of ≤ 1:50 by ELISA. Safety and efficacy above 1:50 have not been evaluated; retesting may be performed if titers are reported above that.
• ELEVIDYS — select patients with anti-AAVrh74 total binding antibody titers < 1:400, as measured by an FDA-authorised test. Do not administer at titers ≥ 1:400.

Reported seroprevalence in humans is substantial: AAV2 is the most prevalent at roughly 30–72%, with AAV5, AAV8 and AAV9 in the 15–47% range. Over 60% of adults have neutralising antibodies to adenovirus. For HSV-1 products, 58% of IMLYGIC trial patients and 64% (14 of 22) of evaluated VYJUVEK patients were seropositive at baseline — local administration tolerates that in a way systemic dosing does not.

Route matters here. Enclosed compartments — subretinal space, bladder, intraparenchymal delivery — are far less exposed to circulating neutralising antibody than an intravenous infusion. Choosing a less prevalent serotype and choosing a compartmental route are two ways of solving the same problem.

You can often read the vector off the drug name

Nonproprietary names for gene therapy products are built from stems assigned by the USAN Council and the WHO INN programme, and the vector is encoded in the suffix. Across the current FDA list of approved cellular and gene therapy products the pattern is consistent:

• -parvovec — AAV: voretigene neparvovec (AAV2), etranacogene dezaparvovec (AAV5), onasemnogene abeparvovec (AAV9), delandistrogene moxeparvovec (AAVrh74), valoctocogene roxaparvovec, fidanacogene elaparvovec.
• -repvec — herpesvirus: talimogene laherparepvec, beremagene geperpavec.
• -denovec — adenovirus: nadofaragene firadenovec, zopapogene imadenovec.
• -leucel / -autotemcel — ex vivo modified leukocytes or autologous stem cells, where the vector is named in the label rather than the stem: tisagenlecleucel uses a lentiviral vector, axicabtagene ciloleucel a replication-incompetent retroviral vector.

This is a fast sanity check when you meet an unfamiliar product name, but confirm against the label — the stem tells you the viral family, not the serotype, generation or whether the vector integrates.

What this means at the bench

Vector choice is also a containment decision. Work with replication-defective viral vectors carrying recombinant nucleic acid falls under institutional oversight, and the review must be complete before the work starts — not before publication. Our guides on what an Institutional Biosafety Committee does and on transfection methods and the biosafety approval you need cover the approval path and how non-viral delivery compares when a vector is more machinery than the experiment requires.

A short decision path that reflects the constraints above: if the cassette exceeds about 8 kb, you are choosing between helper-dependent adenovirus and HSV-1. If it fits under 5 kb and the target is post-mitotic and you want years of expression, AAV. If you need permanent modification of dividing cells you can culture, a third-generation SIN lentiviral vector. If you need transient, high-level, local expression, adenovirus. Gammaretroviral vectors survive mainly in legacy CAR-T manufacturing, where the target cells are proliferating anyway.

Frequently asked questions

What is the actual size limit for an AAV transgene?

The wild-type AAV genome is about 4.7 kb and vector capacity is quoted at roughly 5 kb, but that figure covers the entire packaged genome — both inverted terminal repeats plus your promoter, transgene, any tag, and the polyA signal. Usable coding space is therefore construct-dependent and meaningfully smaller than 5 kb. Dual-vector systems that split a cassette across two AAVs and reconstitute it in the cell are the standard workaround for larger genes.

Why choose a lentiviral vector over a gammaretroviral one?

Two reasons. Lentiviral pre-integration complexes are actively transported through the nuclear pore, so they transduce non-dividing cells, while gammaretroviruses need the nuclear envelope to break down at mitosis. And lentiviruses integrate into the bodies of transcribed genes rather than near transcription start sites, so they are more likely to disrupt a gene than to activate a proto-oncogene — the failure mode behind the leukaemias in the early LTR-driven SCID-X1 and X-CGD trials.

Do AAV vectors integrate into the genome?

Recombinant AAV vector genomes predominantly circularise through recombination between their inverted terminal repeats and persist as episomes rather than integrating. That is why FDA recommends up to five years of long term follow-up for AAV, against fifteen for vectors with integrating activity. It also means expression is diluted in dividing cell populations, so AAV suits post-mitotic tissue.

Which viral vector carries the largest transgene?

HSV-1. Its genome is about 152 kb; non-replicative genomic HSV-1 vectors accommodate more than 30 kb of cargo and amplicon vectors up to roughly 150 kb. Helper-dependent (“gutless”) adenoviral vectors are next at approximately 30–36 kb. Both lentiviral and gammaretroviral vectors are capped under about 8 kb, and AAV under about 5 kb.

Why does adenovirus give only transient expression?

Adenoviral DNA remains epichromosomal and is not incorporated into the host genome, so it is lost as cells divide and is cleared by immune responses to the strongly immunogenic capsid. That makes adenovirus well suited to vaccines, oncolytics and short-term local delivery, and unsuited to chronic correction of a monogenic disease.

How do I know if a patient population will have neutralising antibodies?

Assume a substantial fraction will. AAV2 seroprevalence is reported at roughly 30–72% and AAV5, AAV8 and AAV9 at 15–47%; more than 60% of adults have neutralising antibodies to adenovirus. Approved AAV products set explicit eligibility thresholds — anti-AAV9 ≤ 1:50 for ZOLGENSMA, anti-AAVrh74 < 1:400 for ELEVIDYS — but the assay and cut-off are product-specific, so read the label rather than transferring a threshold between programmes.

Does vector choice change how long a trial must follow patients?

Yes, and by a decade. FDA recommends fifteen years of long term follow-up for integrating vectors including gammaretroviral and lentiviral vectors and transposons, up to fifteen for latency-capable herpesvirus vectors and for genome editing products, and up to five years for AAV. These are non-binding recommendations open to a risk-based argument, but they should inform vector selection at the programme-design stage rather than surface during IND review.

References

• Bulcha JT, Wang Y, Ma H, Tai PWL, Gao G. Viral vector platforms within the gene therapy landscape. Signal Transduction and Targeted Therapy, 2021.
• Trivedi PD, Byrne BJ, Corti M. Evolving Horizons: Adenovirus Vectors’ Timeless Influence on Cancer, Gene Therapy and Vaccines. Viruses, 2023.
• Yudaeva A, Kostyusheva A, Kachanov A, et al. Clinical and Translational Landscape of Viral Gene Therapies. Cells, 2024.
• Yew CHT, et al. Integrase deficient lentiviral vector: prospects for safe clinical applications. PeerJ, 2022;10:e13704.
• Borrenberghs D, et al. Nuclear entry and integration site selection of retroviral vectors. Nucleic Acids Research, 2018.
• Wu C, Dunbar CE. Stem cell gene therapy: the risks of insertional mutagenesis and approaches to minimize genotoxicity. Frontiers of Medicine, 2011.
• Le Hars M, Joussain C, Jégu T, Epstein AL. HSV-1-derived vectors for gene delivery. Gene Therapy, 2024.
• Li X, et al. Pre-existing antibodies to candidate gene therapy vectors (adeno-associated vector serotypes). PLoS ONE, 2019.
• US FDA. Long Term Follow-Up After Administration of Human Gene Therapy Products: Guidance for Industry, January 2020.
• US FDA. Approved Cellular and Gene Therapy Products.
• DailyMed prescribing information: ZOLGENSMA, ELEVIDYS, HEMGENIX, LUXTURNA, ADSTILADRIN, IMLYGIC, VYJUVEK, KYMRIAH, YESCARTA.
• American Medical Association. United States Adopted Names approved stems.