This may be an oversimplification, and I could be wrong, but the way I understand it is this:
Relapse events are often triggered by an alternative mutation in the cancer. So let's say your skin cell becomes cancerous because gene 1111 mutates. You have, let's say, a million other genes that dictate your skin cell, and maybe like a few dozen mutations that could be cancerous (some mutations are harmless). So the T cell targets that gene 1111 mutation, but as a response, either that same mutated cell mutates again, or a healthy cell mutates. Now it's gene 1232 that mutates, but your T cells are modified to only recognize that gene 1111 mutation, so the new gene 1232 mutation sneaks by without being killed by the T cell.
There are a lot of intuitions I have about why you can't just do it again, and ultimately I think it depends. Maybe gene 1232 doesn't change a protein that T cells can target without killing your other cells. Maybe your skin cell has dozens of mutations but 99% of them aren't cancerous, and we don't yet have the right signal/noise reduction in sequencing / the right AI to determine the mutation that is causing the cancer. Here we need to keep in mind that these proteins are very small and very hard to determine the structure of. Cancer cells can be very hard to differentiate from normal cells in a way that you can target them, so a new mutation is a whole new puzzle to solve.
If I am wrong please correct me - my degree is in chemistry / applied math and not genetics/biology and this info comes from domain knowledge I pick up at my job (SDE for a cancer company) + books I've read.
So let's say you're targetting cd19 (the protein). You're saying that the target might stop emitting cd19. Okay, but then why not target cd21 or another target that the cancer is emitting, and run CAR-T again?
Sometimes the cancer doesn't emit any marker proteins, or maybe the ones it does are difficult to detect/create antibodies to bind to.
(I'm in the same situation as 2 above, I'm a software engineer with domain knowledge from working in a cancer research lab at a university; I may be entirely wrong.)
You need to choose a marker that you can "afford" to use for destruction - which either doesn't exist at all in healthy cells, or, more likely, it exists but losing all your healthy cells with the same marker won't kill you (but is still likely to harm you, just in ways that can be treated). Such markers are scarce, and not guaranteed to exist; if we find one for this cancer (like CD19) that works, but if it fails, then it's exceedingly likely that targeting "cd21 or another target" will simply kill you because it will also destroy some vital cells in your liver or veins or brain.
Relapse events are often triggered by an alternative mutation in the cancer. So let's say your skin cell becomes cancerous because gene 1111 mutates. You have, let's say, a million other genes that dictate your skin cell, and maybe like a few dozen mutations that could be cancerous (some mutations are harmless). So the T cell targets that gene 1111 mutation, but as a response, either that same mutated cell mutates again, or a healthy cell mutates. Now it's gene 1232 that mutates, but your T cells are modified to only recognize that gene 1111 mutation, so the new gene 1232 mutation sneaks by without being killed by the T cell.
There are a lot of intuitions I have about why you can't just do it again, and ultimately I think it depends. Maybe gene 1232 doesn't change a protein that T cells can target without killing your other cells. Maybe your skin cell has dozens of mutations but 99% of them aren't cancerous, and we don't yet have the right signal/noise reduction in sequencing / the right AI to determine the mutation that is causing the cancer. Here we need to keep in mind that these proteins are very small and very hard to determine the structure of. Cancer cells can be very hard to differentiate from normal cells in a way that you can target them, so a new mutation is a whole new puzzle to solve.
If I am wrong please correct me - my degree is in chemistry / applied math and not genetics/biology and this info comes from domain knowledge I pick up at my job (SDE for a cancer company) + books I've read.