CAR T Cells Eliminate Previously Untreatable Solid Cancers in Mice

The tumour cells looked clean. Under the microscope, they registered negative for CD70, the molecule that Sophie Hanina’s engineered immune cells were programmed to hunt. Standard detection methods said they were clear. Conventional CAR T therapy, the living drug that has transformed blood cancer treatment, agreed. It killed the positive cells and left. The tumours grew back.

Hanina, a research associate scientist at Columbia University’s Initiative in Cell Engineering and Therapy, had a hunch the tests were wrong. Not wrong exactly, but insufficiently sensitive. Those supposedly negative cells, she suspected, weren’t truly negative at all.

She was right, and the implications reach across some of the most intractable cancers in medicine. Working with Michel Sadelain, who pioneered the CAR T therapies now used routinely against blood cancers, Hanina developed new detection methods fine enough to reveal a whisper of CD70 on cells that standard assays couldn’t register. The molecule was there. It was just at levels no existing technology had been designed to see. When she then deployed a new class of immune cell therapy sensitive enough to recognise those vanishingly faint signals, the results in mice were stark: kidney cancer, ovarian cancer and pancreatic cancer – three of the hardest tumours to treat – were completely eliminated. The work appears today in Science.

That last detail matters enormously. Complete elimination.

CAR T cells have rewritten the prognosis for B cell leukaemias and lymphomas since their introduction. They work partly because blood cancer cells are blanketed in CD19, a surface protein that acts like a neon sign for the engineered immune cells. Solid tumours are a different problem. They are biologically diverse, their cells expressing targets unevenly, some brightly, many dimly, others apparently not at all. “You can’t cure somebody if you just eliminate a small fraction or even 90% of their tumour,” Sadelain says. “You have to get down to the very last cell.” That 10% remnant – or the 50% that never expressed the target in the first place – is enough to rebuild a tumour and kill the patient.

CD70 has been on researchers’ radar for years as a potential target in solid cancers. It appears on between 70 and 80 per cent of clear cell kidney cancers, a similar proportion of ovarian cancers, and about a quarter of pancreatic tumours. Crucially, it is mostly absent from healthy adult tissues, turning up only on subsets of activated immune cells. That selectivity makes it attractive – a target with relatively few opportunities for friendly fire. The problem was always its patchiness within tumours themselves: some cells wearing it prominently, many appearing to have none at all. Phase I clinical trials of CD70-targeted CAR T cells in kidney cancer patients have reported objective response rates of roughly 6 to 21 per cent. Promising enough to continue, not good enough to celebrate.

Hanina’s insight was to question whether the patchiness was genuine. Using confocal microscopy and a technique called stimulated emission depletion imaging – which works in the far-red spectral channel to cut through cellular autofluorescence – she could distinguish genuine CD70 absence from expression too faint for conventional methods to register. What she found was a spectrum. The vast majority of supposedly negative tumour cells turned out to express CD70 after all, just at levels perhaps a hundred-fold lower than their brightly positive neighbours. The heterogeneity was an artefact of measurement. The tumour, in a real sense, was uniformly positive.

The reason for that dimness turned out to be epigenetic. An enzyme called EZH2 was chemically modifying the histone proteins wrapped around the CD70 gene in the low-expressing cells, placing a repressive mark – H3K27me3 – that clamped gene activity without switching it off entirely. The gene stayed on, just at a murmur. Blocking EZH2 with an inhibitor caused CD70 expression to come roaring back in those cells, confirming the mechanism. And because the silencing was epigenetic rather than genetic, the chromatin structure around the CD70 gene remained accessible – a fingerprint that could predict low-level expression even before treatment, potentially guiding patient selection in future trials.

To exploit this discovery, Hanina turned to HIT cells – HLA-independent T cell receptors – a more sensitive CAR variant developed in Sadelain’s lab. Where conventional CAR T cells borrow their antigen-recognition module from antibodies, HIT cells substitute that module into the architecture of a natural T cell receptor, co-opting the downstream signalling machinery the immune system evolved to detect even tiny quantities of foreign material. “They have the sensitivity of a natural T cell and can detect cancer cells that have only a vanishingly small number of target molecules,” Hanina says. They are, she adds, “the next generation of CAR T cells.”

In mouse models bearing patient-derived tumours grown orthotopically – in the actual organ, kidney or ovary or pancreas, rather than under the skin where xenografts are more convenient to establish and measure – CD70-targeted HIT cells cleared the tumours completely. Conventional CD70-targeted CAR T cells, tested in parallel, achieved partial responses or outright failure. The HIT cells also showed no excess toxicity against normal human tissues in humanised mouse models, accumulating in tumours without depleting bone marrow cell populations beyond what standard CD19-CAR T therapy already does. “Studies suggest that the escape of undetected cancer cells is the key impediment to therapeutic success with conventional CAR T therapy,” Hanina says. The HIT cells found those cells.

Hanina and Sadelain are now planning clinical trials of CD70-targeted HIT cells in ovarian cancer patients at Columbia University Irving Medical Center. More than 20 other solid tumour types express CD70 heterogeneously to some degree, among them glioblastoma and pancreatic ductal adenocarcinoma, two cancers with survival statistics that have barely shifted in decades. Whether the epigenetic silencing pattern discovered in kidney, ovarian and pancreatic tumours extends across that broader landscape remains to be established. But the principle Hanina’s work demonstrates may matter as much as the specific target: that what looks like absent expression in solid tumours is sometimes just expression below the threshold of what we’ve been able to see. “We hope our CD70-directed HIT cells help us find a way to eradicate the entire tumour,” she says. The question now is whether there are other stealth targets lurking in those supposedly negative cells, waiting for a sensitive enough eye to find them.

Study link: https://www.science.org/doi/10.1126/science.adv7378