CAR-T stands for chimeric antigen receptor T-cell therapy. In plain terms: doctors take a patient's own T cells and engineer them with a new receptor — a "chimeric" one — that can recognize a specific tumor antigen. The T cell is then turned into a living drug, trained to seek out and destroy cancer. CAR-T is older and more intricate than many people realize. This guide walks through where it came from, what it is made of, and how it actually kills a tumor.
A Brief History: Four Generations of Engineering
The first CAR-T concept paper appeared in 1993. Even in that early work the core idea was already present: a single-chain antibody fragment (scFv) fused to the signaling tail of CD3ζ. It established two facts that still define CAR-T design today — that an antibody-style binding domain could be grafted onto a T-cell signaling unit, and that CD3ζ could signal on its own.
Why a single chain?
Before scFv, researchers tried inserting two separate genes (the heavy-chain and light-chain variable regions) into T cells. Without efficient electroporation, the retroviral step killed most cells. Fusing VH and VL into one gene — borrowed from how bacteria make single-chain antibodies — dramatically improved transfection efficiency. That insight gave us the scFv.
What followed was a steady refinement of the signal a CAR delivers:
- First generation — the CAR carried only CD3ζ, delivering the "first signal" (antigen recognition). It could make T cells recognize tumor cells, but it did not keep them alive and active long enough to work well in patients.
- Second generation — a costimulatory signal was added alongside CD3ζ (either CD28 or 4-1BB). This second signal is what transformed CAR-T from a lab curiosity into the FDA-approved drugs we have today, and it is what makes CAR-T effective against blood cancers.
- Third generation — designers simply chained two costimulatory domains together (CD28 + 4-1BB). The results were disappointing: stacking signals did not improve efficacy and instead risked T-cell exhaustion and off-target toxicity, so this approach was largely abandoned.
- Fourth generation ("armored" CAR-T) — rather than adding more costimulation, designers added an immune-modulatory signal to fight the immunosuppressive tumor microenvironment of solid tumors — most often by engineering the cell to secrete cytokines such as IL-12, IL-15 or IL-7, and sometimes by adding chemokine receptors like CCR4.
The lesson of the third generation is the central tension of CAR-T: the real challenge is balancing efficacy and safety. Better results come from precise tuning, not from brute-force stacking of signaling pathways.
Anatomy of a CAR: What It Is Made Of
The single-chain antibody (scFv)
The scFv is built from the variable regions of an antibody's heavy chain (VH) and light chain (VL), joined by a flexible peptide linker. It is the part that recognizes an antigen on the surface of a cell. That surface-only recognition is the first built-in limitation of CAR-T:
- ① It must see a surface antigen. CAR-T cannot target the tumor-specific neoantigens that lie inside the cell. It relies on tumor-associated antigens (TAAs), which means healthy tissues expressing the same marker can be caught in the crossfire (the "on-target, off-tumor" problem).
- In blood cancers the targets are clean — CD19 and BCMA. In solid tumors no target is yet fully satisfactory. Candidates such as HER2, MAGE-A4, CEA and PSMA each bring problems: on-target/off-tumor toxicity, antigen escape, poor tumor infiltration, and an immunosuppressive microenvironment.
There is a second, subtler flaw. Unlike a natural antibody, the synthetic scFv lacks a constant region, so its VH and VL can stick to each other through hydrophobic interactions and form oligomers (especially dimers). When the scFv oligomerizes, it can switch on downstream signaling on its own — at a low level that does not require tumor binding — pushing the T cell toward an exhausted state. Researchers are addressing this by re-engineering the CAR (for example, using just a single VH domain, which can still recognize the target without the dimerization problem).
CD3ζ: the signaling engine
CD3ζ mimics the signaling role of the natural T-cell receptor (TCR)/CD3 complex: it tells the T cell it has found its target. On its own, however, CD3ζ signals relatively weakly and cannot sustain T-cell proliferation and activation by itself. The fix was already mentioned above — adding CD28 or 4-1BB as a costimulatory signal.
CD28 vs 4-1BB — same job, different temperament
CD28 activates the NF-κB pathway: it drives T cells to pump out large amounts of IL-2 and delivers strong, short-lived cytotoxicity. Use it when you want a potent, fast hit. 4-1BB activates the NFAT pathway: it secretes fewer cytokines but sustains a strong memory phenotype. Use it when you want durable, long-lasting activation.
How CAR-T Kills a Tumor: Three Mechanisms
A CAR-T cell does not destroy a tumor through a single trick. It uses at least three routes:
- ① Antigen-positive tumor cells — where the target antigen is expressed, the CAR-T cell releases perforin and granzyme, inducing tumor-cell lysis.
- ② Antigen-negative tumor cells — even cells that have lost the target can be killed through death-receptor pathways (e.g., Fas–FasL), inducing apoptosis.
- ③ The tumor microenvironment — CAR-T cells release cytokines such as IFN-γ that reshape the surrounding stroma and amplify CAR-T activity, extending the attack beyond the cells the receptor directly binds.
Source: adapted from a popular-science explainer by 曲别针生命科学 (Paperclip Life Sciences), originally published July 12, 2025.