CAR-T vector choice: lentivirus vs gamma-retrovirus

Lentivirus and gamma-retrovirus differ in whether the target cell must be dividing

The difference lies in how the gene reaches the nucleus. Gamma-retrovirus integrates its gene at the moment of cell division, when the nuclear envelope breaks down, so it can enter only cells that are proliferating. Lentivirus has a mechanism that carries its gene into a nucleus whose envelope stays intact, so it also enters resting cells.

The preferred integration sites differ as well. Gamma-retrovirus tends to integrate near transcription start sites and CpG islands. Lentivirus tends to integrate inside the bodies of actively transcribed genes. The representative approved products are Yescarta and Tecartus on the gamma-retroviral side, and Kymriah, Breyanzi, Abecma and Carvykti on the lentiviral side.

In CAR-T manufacturing, though, T cells are usually stimulated with beads carrying anti-CD3/CD28 antibodies and expanded before the vector goes in (see CAR-T manufacturing). The target is a dividing cell, so a gamma-retroviral process works, and Yescarta is exactly that. The gap shows up when you want gentler stimulation so differentiation does not advance too far, or when you want to keep memory T cells that are close to resting. There is a view that lentivirus's flexibility pays off in those designs.

The production methods also differ in a typical way. Gamma-retroviral vector has long been made in stable producer cell lines that keep releasing vector, and PG13 (derived from mouse NIH3T3 cells) is a representative packaging cell line. Lentiviral vector is widely made by transient transfection of plasmids each time. Why lentivirus is hard to turn into a cell line is covered in lentivirus production: transient transfection or stable producer line (in Japanese).

How far you plan to expand the target cells and which vector you choose cannot be decided separately.

Insertional mutagenesis depends on design, not on the virus family

Where the vector lands in the genome changes how large the risk is of switching on an oncogene by mistake. Gamma-retrovirus tends to land near transcription start sites. In early gene therapy trials for X-linked severe combined immunodeficiency, such insertions activated a proto-oncogene and leukemia followed (see insertional mutagenesis risk in gene therapy (in Japanese)). That lesson changed vector design a great deal.

Lentivirus tends to land inside the bodies of transcribed genes, and its concentration near start sites is not as strong as with gamma-retrovirus. Even so, the family name does not make it safe. Three things decide the risk. The first is whether the vector is a self-inactivating (SIN) design, in which the long terminal repeat sequence is deleted so that the integrated vector does not transcribe itself. The second is the strength of the internal promoter. The third is the number of copies that enter each cell.

FDA's 2024 warning says that T cell malignancies, including CAR-positive ones, can appear as early as a few weeks after infusion, and some cases were fatal. FDA did not withdraw or restrict these products, and its position is that the benefit outweighs the risk. The warning applied across both virus families, so when you compare vectors, read the design documents instead of deciding by family name.

The risk is set by the combination of SIN design, internal promoter strength, copy number and the nature of the target cell, not by the virus family alone. In a document, look for these four before the family name. How copy number is measured and specified is covered in measuring VCN for CAR-T (in Japanese).

Compare vectors on the design documents: whether it is SIN, which promoter it uses, and what the copy-number specification is.

Choosing a vector supplier: ask what the timeline assumes

Vector manufacturing is a heavy process, and it is a common point of supply and cost bottlenecks (see lentivirus manufacturing process (in Japanese)). CDMOs differ in their strengths, and the published numbers do not share the same premises.

Lonza says it supports both transient and stable packaging cell lines. AGC Biologics advertises about 9 months to reach GMP with its ProntoLVV platform. In April 2026, Oxford Biomedica (OXB) announced a lentiviral contract service that cuts development to GMP manufacturing, usually put at 12 to 18 months, to as little as 9 months. It is limited to eligible programs (see the news item (in Japanese)). All of these are company claims, and the point where the clock starts may not be the same.

What to ask about is the premise behind each number. Is it for a program that goes onto an existing platform, or one that starts from vector design? What scale and method are involved? Is there a track record with genes such as CARs that can burden the producer cell? Titer numbers cannot be lined up as they are, because each company measures with its own cells and its own method (see lentivirus titer measurement (in Japanese)).

Decide whether scale or start-up time matters more, then compare the numbers only after you know what each one assumes.

Transduction reagents and release tests belong in the vector comparison

When a vector is added to T cells, a reagent that brings vector and cells close together helps the transfer. Takara Bio's RetroNectin, a recombinant fibronectin fragment, has long been used for gene transfer into hematopoietic stem cells and lymphocytes. Such reagents are raw materials that enter the cell product's process. Check their grade and quality documentation just as you do for the vector.

Before release you confirm two things: that no replication-competent virus is present (RCR, and RCL for lentivirus), and that the vector copy number per cell is not excessive. The test targets and the days they take are covered in replication-competent virus testing (in Japanese). Commercial kits for RCL detection are also available, such as Bio-Rad's digital PCR kit (see the news item (in Japanese)).

Put the transduction reagent and the RCL and copy-number tests into the same comparison as the vector.

Non-viral does not mean safer

Taking the virus-making step out of the process eases vector sourcing and testing. One way is to deliver a transposon such as Sleeping Beauty or piggyBac (a sequence that moves within the genome) into cells by electroporation (see electroporation for cell therapy (in Japanese)).

But ease of manufacture and safety are separate questions. In a trial of CD19 CAR-T cells made with piggyBac, 2 of 10 patients developed CAR-positive T-cell lymphoma, and 1 died. According to the report, the first lymphoma had a high transgene copy number, and no insertion into a typical proto-oncogene was found (Micklethwaite et al., Blood 2021). Copy number and the nature of integration are points to check for viral and non-viral methods alike.

In the in vivo direction, where CAR-T cells are made inside the patient, the vector is not a raw material for a cell product. It is the medicine itself, given directly to the patient (see in vivo CAR-T (in Japanese)).

Do not assume non-viral means safe; check the integration behavior and the copy number in the data for every method.

Seven questions to put to a vector candidate or a CDMO proposal

Ask these when you have vector candidates or a CDMO proposal in front of you.

1. How far is the design meant to expand the target T cells? Do you want to keep cells close to resting, such as memory T cells?
2. Is the vector self-inactivating (SIN)? Which internal promoter does it use, and what supports its strength?
3. What are the copy-number specification and the measurement method? What does the replication-competent virus test cover, and how many days until results?
4. What is the supplier's method (transient or stable producer), its scale, and its time to GMP? What does that number assume?
5. Is the titer a functional titer? In which cells and by which method was it measured?
6. What grade is the transduction reagent, and what quality documentation exists for it?
7. If you change vectors during development, is there a comparability plan?

To see vector choice across gene therapy as a whole, go to choosing a gene therapy vector (in Japanese). For supplier checks, go to selecting a viral vector CDMO (in Japanese). You can find suppliers in the vector manufacturing CDMO directory, transduction reagents under lentivirus transfection reagents, and equipment under electroporators.