How Are CAR-T Cells Actually Engineered? Inside the Vector and Gene-Editing Process

What This Article Covers

CAR-T therapy is often described in terms of its timeline — blood drawn, cells shipped, weeks of waiting, an infusion. The more remarkable story is what happens at the molecular level during that wait: a patient's own T cells are given a synthetic gene that did not exist in nature, delivered using tools borrowed from viruses that normally cause disease. This article goes inside the cell — what a chimeric antigen receptor actually is, how its gene is physically inserted into a T cell's DNA, and how gene-editing tools like CRISPR are being used to build "off-the-shelf" CAR-T that skips the need for a patient's own cells.

What a CAR Actually Is

A chimeric antigen receptor doesn't exist in nature — it's an engineered protein stitched together from parts that normally belong to different molecules ("chimeric" is the tell). According to the National Cancer Institute, the external part of a CAR is built from fragments of lab-made antibodies that let it latch onto a specific antigen, such as CD19 on B-cell leukemias, on the surface of a cancer cell. That targeting fragment is called a single-chain variable fragment, or scFv, connected through a hinge and transmembrane domain that anchor the receptor in the T cell's membrane. Inside the cell, signaling and costimulatory domains fire off the instructions that make the T cell attack and multiply once the scFv binds its target.

The choice of costimulatory domain shapes how the resulting cells behave for months or years afterward. The two domains used in every FDA-approved CAR-T product are CD28 and 4-1BB (CD137). Research comparing them describes CD28-based CARs as producing a fast, explosive attack with robust cytokine release, while 4-1BB-based CARs mount a slower response but persist far longer — sometimes detectable in patients for months or years — because they lean on durable, oxidative metabolism rather than a short burst of glycolysis. Both designs also carry a CD3-zeta domain, which delivers the core activation signal through structures called immune-tyrosine-based activation motifs. This isn't academic: Yescarta and Tecartus use CD28 domains, while Kymriah, Breyanzi, Abecma, and Carvykti all use 4-1BB.

How the Gene Gets Into the T Cell

Designing a CAR on paper is step one. The harder problem is getting the instructions for building it — one continuous gene — physically into the DNA of millions of a patient's T cells, so every daughter cell keeps making the receptor. The dominant solution borrows one of biology's most efficient delivery machines: viruses, stripped of their disease-causing genes and repurposed as inert shuttles.

Two viral vector families are used in FDA-approved CAR-T manufacturing: gamma-retroviral and lentiviral vectors. Both are made by replacing a virus's genetic payload with the CAR gene, then using packaging cells to assemble particles that "transduce" a T cell and splice that gene into its genome. Gamma-retroviruses can only integrate into actively dividing cells, and tend to insert near gene promoters in a way that raised cancer-risk concerns in earlier gene-therapy trials. Lentiviral vectors, derived from HIV but rendered replication-incompetent, don't require cell division and prefer to land inside actively transcribed genes rather than near their promoter switches — a safer integration pattern. Both remain in clinical use today: Yescarta and Tecartus rely on gamma-retroviral vectors, while Kymriah and Breyanzi use lentiviral ones. Transduction happens in a bioreactor after a patient's T cells are collected by apheresis and activated; the virus delivers the CAR gene into the cells' chromosomes, and the modified cells are then expanded to hundreds of millions before infusion.

Beyond Viruses: Non-Viral Engineering

Viral vectors are expensive and slow to manufacture at scale, which has pushed researchers toward non-viral alternatives. One of the most studied is the piggyBac transposon system, a "cut and paste" DNA element that, with the help of an enzyme called transposase, inserts a CAR gene into a T cell's genome with no viral machinery at all. Published methods for enzymatically produced piggyBac vectors report efficient non-viral generation of CD19-specific CAR T cells, and the approach is now being paired with gene-editing tools in experimental allogeneic platforms. Other strategies use mRNA electroporation — briefly opening a T cell's membrane with an electrical pulse to deliver CAR-encoding mRNA directly, producing CAR expression without permanently altering the cell's DNA, useful for shorter-acting or safety-focused applications.

The Rise of Gene-Editing for Off-the-Shelf CAR-T

Every approved CAR-T product is autologous — built from the patient's own T cells, which is why manufacturing takes weeks and why not every patient is healthy enough to donate cells that expand well. Gene-editing tools including CRISPR/Cas9, TALENs, and newer base and prime editing are now being used to engineer donor T cells that could, in principle, be manufactured in advance and used in any patient — an "allogeneic," or off-the-shelf, CAR-T product.

The core problem with someone else's T cells is that they carry their own T-cell receptor and HLA markers, which can trigger graft-versus-host disease or cause rejection by the recipient's immune system. Gene editing disables the genes responsible: knocking out TRAC or TRBC removes the native T-cell receptor to prevent graft-versus-host disease, while knocking out B2M — and in some programs CIITA — removes HLA molecules to blunt rejection. Some programs also knock out CD52, letting engineered cells survive an antibody-based conditioning regimen that would otherwise wipe them out. Base and prime editing are considered theoretically safer than CRISPR or TALENs because they can alter DNA sequence without creating a double-strand break.

The promise is real: an off-the-shelf product made in bulk from healthy donor cells could, in theory, reach a clinic in days rather than weeks, at lower cost and without excluding patients too depleted by prior chemotherapy to donate cells. Early trials are underway — for example CB-011, a CRISPR-edited, HLA-modified allogeneic CAR-T in multiple myeloma. But the caveats matter: no genome-edited allogeneic CAR-T has yet been FDA-approved, and the field's most sobering lesson came from the UCART19 trial, where simultaneous TRAC and CD52 editing produced an unexpected chromosomal translocation in every manufactured batch — a reminder that cutting DNA at multiple sites carries real, measurable genotoxicity risk still under study.

What's Still Experimental

Viral-vector-based autologous CAR-T is proven technology, underpinning seven FDA-approved products treating blood cancers since the first approval in 2017. Non-viral transposon and mRNA-based engineering are further along in the lab than the clinic, with most data still preclinical or early-phase. Gene-edited allogeneic CAR-T is furthest from routine use: candidates sit in Phase 1 trials, with proof-of-principle activity in some patients but open questions about durability, editing-related genotoxicity, and long-term safety.

Bottom Line

Underneath every CAR-T infusion is a feat of molecular construction: a synthetic receptor gene, built from antibody fragments and native signaling machinery, spliced into a living T cell's genome using tools adapted from viruses or, increasingly, enzymes that need no virus at all. That process is mature and proven for today's approved, patient-specific therapies. The next frontier — gene-edited, off-the-shelf CAR-T built from donor cells — carries genuine promise to make this therapy faster, cheaper, and more widely available, but it remains investigational, with real safety questions still being worked out rather than settled.