Engineering immune system cells to do what we’d like them to do is one of the big areas of medical research these days, and this new paper could be a real advance in the area. A team out of UCSF with several collaborators reports on a new way around one of the big problems in this area – these various immune system cells are great and powerful, but they can be too great and powerful if you can’t aim them precisely.
T cells, NK cells and so on are there to hunt down and kill other cells if they detect that they’re diseased, and it would be a great thing if they could be persuaded to recognize cancer as a disease. But so many tumors fly under that radar, sending out constant “friend not foe” biochemical signals. In the last few years, though, we’ve seen some dramatic results from CAR-T (chimeric antigen receptor T cell) therapy, in which a sample of a patient’s own bone marrow is removed and the stem cells are engineered to produce T cells with synthetic receptors on their surface that recognize specific proteins found on leukemia cells. You then infuse the altered cells and turn them loose: now the patient is equipped with T cells that suddenly recognize, bind to, and kill the leukemia cells. The first dramatic success in this field damn near killed the recipient of the treatment, when his new T cells cleared out such a mass of malignant B cells so quickly that his kidneys nearly shut down, among other problems. This procedure (with variations) has been approved by the FDA as a therapy for some types of leukemia – it’s grueling indeed, but at the same time there are people walking around today because of it who were headed for certain death. (Update: clarified the procedure above where I’d bungled it some).
As you’ll appreciate, you need to have a very specific protein antigen signal to set these things loose on. There’s the problem, because it’s been difficult to find such things for many (most) tumors, especially solid ones, and the last thing you want is a bunch of revved-up T cells wandering around eating the wrong parts of your body. Not to put it too delicately. There are tumor types that present some promising antigens, but the problem is that these same surface proteins show up in other tissues as well.
What this latest paper does is an ingenious two-step verification process. The T cells are engineered with a synthetic version of the Notch receptor, set to recognize some particular surface protein that’s peculiar to the general tissue that you’re aiming at. You wire that Notch system up so that its activation then causes a new specific chimeric antigen receptor to be expressed on the cell surface, this one targeted towards one of those tumor antigens. The idea is that this way the CAR will only show itself in the presence of the right tissues, and not just hit the general circulation. If some of the cells with their new CARs do wander off, they lose the expression phenotype once their Notch receptor is no long being activated, and revert to something harmless anyway.
This is demonstrated in models of glioblastoma, which of course is a famously untreatable cancer in most cases with a tragically rapid course. Glioblastomas often express a particular surface protein called epidermal growth factor receptor splice variant III (EGFRvIII), but unfortunately aiming cell therapies at that one hasn’t really been successful: some of the tumor cells escape, and then grow back after the others have been killed off (which is generally the way that oncology treatments fail). Meanwhile, there are other surface proteins available, such as ephrin type A receptor 2 (EphA2) and interleukin 13 receptor α2 (IL13Rα2) that are associated with several types of cancer, but are unfortunately in that set that are expressed in other normal tissues as well. People are trying CAR-T cells aimed at them, but it’ll probably be an uphill fight.
So this group tried cells that had Notch receptors set for either EGFRvIII or another common brain protein, myelin oligodendrocyte glycoprotein (MOG), with the activation of those receptors set to then produce chimeric antigen receptors targeting EphA2 or IL13Rα2. The hope was that these fully activated cells would then only be produced in brain tissue. As the authors note, “Neither the priming nor killing antigens need to be perfect” – it’s the intersection of the two that gives you specificity that otherwise might be impossible to reach. If they each have a different specificity problem, you can get around those defects and use the best features of each one. And that seems to be what happened here: in mouse xenograft models of glioblastoma, these two-stage cells were able to completely clear the tumor tissue without apparent harm to other tissue types. It’s quite dramatic, and when this is pictured by fluorescence imaging, you can actually see the cells lighting up as the Notch/CAR combination kicks in – here’s a download (open access) of a short movie showing this happening.
There seems to be another benefit to this technique. When people have tried surface antigens like EphA2 or IL13Rα2, one problem has been “antigen exhaustion”. Leaving the T cells constantly flipped on for such things apparently triggers an eventual downregulation, making them much less active. But the intermittent only-when-Notch-goes-off activation of those receptors seems to keep the T cells in a much more lively state. The combination of this effect and the specificity trick inherent in the two-step priming process could be very potent indeed.
So these are excellent mouse xenograft results, but as someone who’s been on a lot of oncology projects over the years, there’s a limit to which I can get excited just by mouse xenograft results. What’s even better about this paper is the proof of concept: the “SynNotch” protocol is a very good idea, and so far it looks like it’s working the way its inventors hoped. I wish them rapid progress towards human trials, which is where all such therapies have to be proven in the end, and I’ll be watching with great interest along the way.