In 2010, a research team led by Steven Rosenberg at the National Cancer Institute treated a 39-year-old woman whose colon cancer had spread to her lungs and liver. She received an infusion of her own immune cells, re-engineered to carry a CAR — a chimeric antigen receptor, a synthetic receptor stitched together so that a T-cell, the immune system’s central response cell, will latch onto one chosen molecule on a cell’s surface and respond. The chosen molecule here was ERBB2, more commonly called HER2, a protein found at high levels on many solid tumors.
Within about fifteen minutes she developed difficulty breathing. Her oxygen levels fell; she progressed to respiratory failure and cardiac arrest, and she died five days later. A surge of inflammatory signaling molecules appeared in her blood within hours. When the investigators reconstructed what had happened, reported by Morgan and colleagues in Molecular Therapy, they reached a conclusion that has shaped the field ever since: the engineered cells had found HER2 — not on the tumor alone, but at a low level on the normal cells lining the lungs — and had acted there.
The target was real on the cancer. It was also present, faintly, on a tissue no patient can lose. That is the landmark example of what is now called on-target, off-tumor toxicity, and it is the single clearest argument for why validating a target for a cell therapy is a different exercise from anything that came before it.
