The pileup in immuno-oncology is already the stuff of legend: it’s difficult to even count the number of therapies, combinations, and clinical trials that are underway or in development. And that’s for good reason, of course – the promise here is huge, the field is wide open, and there are vast tracts of things that we know little or nothing about and can only discover through experiment. Right in front of us, in real time, what may be the greatest change in cancer treatment since the advent of chemotherapy is taking shape.
But if you’re going to turn the immune system loose on tumor cells, you want to be extremely sure that it’s only tumor cells that you’re attacking. We have enough autoimmune diseases in the human population already to be sure that we don’t want to invent any new ones. A misfiring immune response can make a person miserable for the rest of their life, or (if things go really wrong) can end that life in a matter of minutes. No, this is a mighty weapon and has to be handled with great care.
What you’d want ideally is a list of tumor-specific antigens, targets that you can turn the T-cells and antibodies loose on without fear for the rest of the body. But people have been searching for that list for many years, and it’s not an easy thing to come up with. The progress in forms of leukemia has been driven by the identification of such markers, but for the vast majority of other tumor types we don’t have much to work with. This new paper, though, perhaps holds out some clues. It’s from a team in Montreal, and they report some unexpected sources of leads. As it is now, people have been sequencing tumor cells and looking for mutations that are associated with said tumor and are different enough to confer some specificity. But that’s been quite hard – almost all the candidates discovered this way turn out not to perform through the various steps of the immune response cascade.
They’re unexpected because the tumor-specific antigens they’re finding are disproportionately from completely noncoding regions of the genome. They would thus be missed by many of the more targeted approaches, such as anything that depends on looking for regularly expressed genes. The key, this paper says, is to not focus so much on mutated tumor-specific antigens (mTSAs) so much as aeTSAs, the aberrantly expressed ones. There is of course a lot more noncoding DNA than there are coding regions, and it can potentially be expressed (perhaps especially under the conditions found in tumor cells?)
Now, commonly available protein datasets (such as what you’d use for mass spec proteomics) tend not to contain even the “regular” TSAs, much less these things. So anyone working in this field has to start off with a big proteomics effort, but this group went out of their way to do as much alignment-free work as possible to increase their chances of finding things outside the usual regions. They used medullary thymic epithelial cells, which have a huge transcriptome already, as the controls: chopping the RNA-seq reads into 33-base pieces and getting rid of everything that showed up in those normal cells leaves a residue of potential TSAs, coding regions or not. They also got rid of sequences that were too close to what had been reported before, or things that showed up on a check of other normal tissue types.
After checking two mouse tumor lines and seven human ones, that leaves about 40 potential TSAs (17 murine and 23 human), and 90% of those are from noncoding regions. Interestingly, a number of them are also from endogenous retroelements (EREs) in the DNA sequences. The cDNAs of these have already been noted as a potential contributor to autoimmune disease, and there are clearly some deep connections between EREs and immune function. A lot of retroviruses have left their calling cards in our genomes over evolutionary time, and the big question is whether to recognize these things as signs of foreign viral invasion or whether to treat them as “self”. You can imagine that going too far in either direction could be bad news – raising the cutoff too much would allow for unchecked viral infections, but lowering it too much would set off responses to all sorts of ancient nonthreatening stuff (and probably more bystander proteins and cDNAs besides).
Immunization experiments in mice were promising, with activity depending (as it should) on the number of responsive T-cells. One of the appealing things about these antigens is that they have a greater chance of not being “private” ones, specific to just one particular cell line. A TSA that shows up in a variety of tumor clones but still not in normal tissue would be ideal, since the big problem in all sorts of cancer therapy is wiping out some, or most, but not all of the cells in a heterogeneous tumor.
Identifying and validating these things is still going to be labor-intensive (although the paper has some suggestions on how to make the workflow less onerous). But the results could be worth it – there are so many tumor types for which we have no real antigen candidates at all. These would be candidates for vaccines, modified T-cells, and whatever else we can envision, and if they’re lurking out there in the genomic wilderness, it’s time we rounded them up.