The two most widely used viral vectors in gene therapy — adeno-associated virus (AAV) and lentivirus (LV) — are often discussed as competing platforms. That framing is misleading. AAV and lentivirus are not interchangeable tools competing for the same applications. They are fundamentally different biological systems with different mechanisms of action, different payload capacities, different tropism profiles, and different manufacturing requirements. Choosing between them is not a preference — it is a scientific and clinical decision that should be driven by the biology of the target disease, the tissue to be treated, and the durability required.

This article explains the defining biological and technical differences between AAV and lentiviral vectors, describes the disease contexts in which each performs best, and outlines the key considerations that should guide platform selection.

How AAV Works

AAV is a small, non-enveloped DNA virus belonging to the parvovirus family. In its recombinant form, the viral genome — approximately 4.7 kilobases of single-stranded DNA — is packaged inside a 25-nanometer icosahedral protein capsid. After systemic or local administration, the capsid binds to cell-surface receptors and is internalized. Once inside the cell, the vector travels to the nucleus, where the single-stranded DNA genome is converted to a double-stranded form and primarily exists as an episome — a circular DNA element that is not integrated into the host chromosome.

Because AAV DNA persists primarily as episomal DNA rather than integrating, it is not replicated when cells divide. In non-dividing or slowly dividing cells — such as neurons, liver hepatocytes, cardiomyocytes, and photoreceptors — episomal DNA is stable over years, enabling long-term gene expression from a single administration. In rapidly dividing cells, the episomal genome is progressively diluted as cells divide, which limits AAV’s utility in proliferative tissues.

How Lentivirus Works

Lentivirus is an enveloped RNA virus derived from HIV-1, engineered for safety by splitting its packaging components across multiple constructs and deleting replication-competent sequences. After entry into target cells, the single-stranded RNA genome is reverse-transcribed into double-stranded DNA by the viral reverse transcriptase. That DNA then integrates into the host cell genome — permanently and heritably — at sites distributed across the chromosome.

Integration is the defining feature of lentiviral vectors. It means that the therapeutic gene becomes a permanent part of the cell’s genetic material, is replicated with every cell division, and is passed to all daughter cells. This makes lentivirus the natural choice for applications targeting dividing cells, particularly hematopoietic stem cells and T cells, where the therapeutic benefit must be maintained through cell expansion and turnover.

The lentiviral genome is larger than AAV — typically accommodating payloads of 8 to 10 kilobases — which provides considerably more design flexibility for complex genetic constructs.

Payload Capacity: A Defining Constraint

The payload size difference between the two platforms is one of the most important practical distinctions. AAV’s approximately 4.7-kilobase limit constrains which diseases it can address. Many therapeutically relevant genes fit within this limit — the factor IX gene used in hemophilia B therapy, the RPE65 gene corrected in Luxturna, the SMN1 gene restored in Zolgensma — but some clinically important genes do not. The dystrophin gene associated with Duchenne muscular dystrophy, for example, is one of the largest genes in the human genome and cannot be packaged in a standard AAV capsid in full-length form. Developers working on Duchenne have had to engineer shortened versions of the protein (micro-dystrophins) that fit within the packaging limit — an approach with its own efficacy and regulatory implications.

Lentivirus, with its larger payload capacity, can accommodate more complex genetic constructs, including multiple regulatory elements, full-length gene sequences, and the chimeric antigen receptor constructs used in CAR-T cell manufacturing. The CAR construct — which encodes a fusion protein combining an extracellular antigen-binding domain, a transmembrane domain, and intracellular signaling domains — is delivered to T cells by lentiviral transduction in all currently approved autologous CAR-T products.

Integration vs. Episomal Persistence

The question of whether the therapeutic gene should integrate or persist episomally is not merely a technical preference — it carries clinical and safety implications.

Episomal persistence, the predominant mode for AAV, avoids the risk of insertional mutagenesis — the disruption or aberrant activation of host genes at the site of integration. Early gene therapy programs using gamma-retroviral vectors, which integrate somewhat preferentially near gene promoters, were associated with insertional activation of proto-oncogenes in several patients. Lentiviral vectors integrate more randomly across the genome and with a lower preference for promoter regions compared to gamma-retroviruses, but insertional mutagenesis remains a theoretical and monitored safety concern for any integrating vector.

For AAV, the trade-off is that episomal DNA is diluted by cell division. This is acceptable — and even desirable from a safety standpoint — in post-mitotic tissues. For liver-directed therapies in pediatric patients, however, the fact that hepatocytes divide as the liver grows means that AAV-based therapy administered in childhood may lose efficacy as the patient grows, potentially requiring retreatment. Re-dosing with AAV is complicated by the immune response to the capsid generated after the first administration, a challenge that does not apply in the same way to ex vivo lentiviral applications.

For lentiviral applications, stable integration means that a single manufacturing and transduction event produces a permanent correction in the cell and all its progeny — which is precisely the goal in HSC gene therapy and CAR-T manufacturing. The cells are modified ex vivo, expanded, and reinfused. Once the modified cells are engrafted or infused, no additional vector exposure occurs.

Tropism: Where Each Vector Goes

AAV and lentivirus differ fundamentally in how they are used in the clinic. AAV is primarily an in vivo vector: it is administered directly to the patient — intravenously, intramuscularly, intravitreally, intrathecally, or via other routes — and must reach and transduce the correct target tissue in the body. Tissue targeting depends on the capsid serotype, the route of administration, and the presence or absence of specific cell surface receptors in the target tissue.

Lentivirus is primarily an ex vivo vector: patient cells are collected, transduced outside the body in a controlled manufacturing environment, and reinfused. This means that lentivirus does not need to navigate the in vivo environment, evade immune clearance, or find its target tissue — the target cells are presented to the vector directly in a controlled setting. The consequence is that lentivirus does not need to be engineered for tissue tropism in the way AAV serotypes are selected for it.

Manufacturing Complexity

AAV and lentiviral vectors have substantially different manufacturing profiles. Both use transient transfection of HEK293 cells as a common production approach, and both require multi-step downstream purification. However, lentiviral vectors are enveloped — they are surrounded by a lipid bilayer derived from the producer cell membrane — which makes them considerably more fragile than the non-enveloped AAV capsid. Lentiviral vectors are sensitive to freeze-thaw cycles, shear stress, and temperature. They have shorter shelf lives and require more controlled handling conditions.

AAV, by contrast, is non-enveloped and physically robust. It tolerates freeze-thaw cycles, can be formulated for long-term storage, and has a more favorable stability profile for distribution and patient administration. This physical robustness has contributed to AAV’s commercial viability as an in vivo therapeutic.

Which Platform for Which Disease?

The practical guidance that follows from these differences is relatively clear. AAV is the platform of choice for monogenic diseases affecting post-mitotic tissues — liver, eye, CNS, muscle — where the payload fits within the 4.7-kilobase limit and where long-term episomal persistence provides durable correction. Approved programs in hemophilia A and B, Leber congenital amaurosis, spinal muscular atrophy, and various lysosomal storage disorders reflect this profile.

Lentivirus is the platform of choice for applications requiring stable gene integration in dividing cells — particularly ex vivo modification of hematopoietic stem cells for inherited blood disorders and immunodeficiencies, and T cell transduction for CAR-T manufacturing. Every approved autologous CAR-T product to date uses lentiviral transduction. Approved programs in beta-thalassemia, sickle cell disease, and ADA-SCID use lentiviral HSC transduction.

The boundaries between these categories are not absolute. Hybrid strategies, including dual-vector AAV approaches for large genes and in vivo lentiviral delivery for some indications, are active areas of development. But for most programs, the biology of the target disease, the tissue being treated, the proliferative status of the target cells, and the payload requirements together make the platform choice substantially more constrained than the framing of “AAV or lentivirus” implies.

The choice of vector platform is one of the most durable decisions made in a gene therapy program’s early life. Making it on the basis of biological fit rather than platform familiarity is the starting point for a development strategy that the science can sustain.

AAV vs. lentiviral vector platform comparison across ten decision dimensions. The ‘Platform Selection Implication’ column reflects the practical consequence of each biological and technical difference. Platform selection should be driven by biological fit across all dimensions simultaneously, not by any single attribute.