The immunological approaches to cancer therapy that have come on in the last few years are some of the best news that field has had in a long time, and with any luck the revolution is just beginning. There’s a huge amount of research going on to profile various sorts of tumors for specific antigens, both to target therapies to them and to understand what these things mean clinically. The mutations that are constantly being generated in cancer cells should produce altered proteins on the cell surface, and these “neoantigens” would be natural candidates for immune targeting, since they shouldn’t exist in the normal cell population. But it’s tricky: that same mutation rate may also produce such a variety of neotantigens (in such a variety of cells) that specific targeting stops becoming feasible. (Even if it works, this is probably going to be personalized medicine for sure, with every patient having their own set of mutant proteins). As pointed out here, there are also two broad possibilities for such therapies – that once you start an immune response to the tumor cells that it picks up momentum on its own as they get destroyed, or that you’re going to have to be specific all the way, because otherwise the immune response is going to end up chasing after various proteins in a way that’s at the least ineffective and at worst harmful.
There’s a new paper out in Science that should advance the field. A large (and high-powered) team spanning the UK and US (and funded by Cancer Research UK and the Rosetrees Trust) has been profiling antigens in developing tumors (two varieties of lung cancer) and correlating that to clinical outcome. The quick answer is that in adenocarcinoma, when there’s been a broad clonal similarity in the antigens, the patients treated with the PD-1 antibody Keytruda (pembrolizumab) have had much better responses than when there’s a lot of subclonal variety. The same relationship is seen in melanoma samples and responses to the CTLA-4 antibody Yervoy (ipilimumab).
The paper notes that such tumors are under immune attack already in patients, but not to a useful degree. Thus the immunoenhancing antibody therapies – one of the problems is that these tumors secrete various immune-suppressing proteins, and taking out those signaling pathways allows an effective immune response to develop. It appears now that such an effective response is strongly correlated with the clonal neoantigen state.
That also helps to confirm that (as had been suspected) that if you’re raising responses to tumor cell antigens that you should go down as close to the main branch as possible, and pick antigens that are shared widely. Once you start sequencing for potential neoantigens, you could have a choice of hundreds (or thousands) of them, and it appears that you’d better choose wisely. But this paper also shows another area that needs some work, because the authors also looked at squamous cell carcinoma samples, and found that these cells tended not to display many neoantigens at all, which is the opposite problem. A target-poor environment is even worse than a target-rich one.
And this brings up something for us small-molecule people to think about: as the march of the immunotherapies continues, it’s possible that we’re going to have to modify our own targets. It could well be that the unmet medical need is going to concentrate more in the tumor types that are not good candidates for immune approaches (such as the squamous cell example above), and that for maximum clinical impact we should be deliberately looking for these areas. Thoughts?