If you go to the med-chem or pharmacology literature databases and type "Aurora kinase", you'd better stand back. A geyser of publications will come spraying out, most of them having to do with Aurora A and/or Aurora B as possible targets for cancer therapy. These enzymes are involved in different phases of cell division, among other things, and a lot of evidence points to them as key players in several cancer lines. There are a number of inhibitors compounds known for them as well, in various stages of development, some of which are selective and some of which hit both to different degrees. Attempts to unravel all the functions of the kinases through these compounds, and through various loss/gain of function mutations in cells, have been. . .well, "complex" is a judicious term to use. The functions of the two enzymes may well be tied to each other, so getting a clear look has been hard.
There's a new paper that illustrates just why it hasn't been easy. This one looks at an AstraZeneca compound, ZM447439, which inhibits both Aurora A and Aurora B in enzyme assays, but in cells seems to be the closest match to a clear knockout of B. The authors started with a well-known cancer cell line (HCT-116), and picked out mutants that had acquired resistance to the drug. They turned out, indeed, to have mutated forms of Aurora B in them, and when they introduced those mutant forms into other cells, they also became able to grow in the presence of ZM447439. That's about as good a test of mechanism as you're going to get in the oncology field, and as the commentary on the paper says, "Even had the authors stopped at this point, it would have been an important contribution."
But they kept on digging, and good for them - perhaps they were (rightly) suspicious that everything was working out a bit too neatly. They then chose two other Aurora inhibitors, VX-680 (which hits both forms) and MLN8054, which is known to be selective for Aurora A. When the cells with mutant forms of Aurora B were exposed to the VX compound, they grew anyway - which makes sense from the Aurora B side of things, since they could well have mutated the efficacy away from this compound, in the same way they got away from the AstraZeneca one. But VX-680 definitely seems to hit Aurora A, too - so is that pathway not doing anything at all for efficacy?
Well, when they treated the Aurora B mutant lines with the Aurora-A-selective MLNM compound, they died off, implying that Aurora A inhibition can do the job all by itself, so there's a pretty blatant contradiction here. The authors advance the two hypotheses that have to be looked at: either Aurora A is a good target and the VX compound isn't doing as much against it as everyone thought, or Aurora A inhibition is largely useless (at least in HCT-116 cells!), and the MLNM compound has another target that no one's realized yet. (It's important to realize that this situation could vary from tumor to tumor - here's a suggestion that Aurora A might be the way to go for pancreatic cancer, for example).
And there's another, rather troubling take-home lesson, having to do with the alacrity with which these cells mutated away from sensitivity to the Aurora inhibitors. As the authors put it:
"The rather surprising picture emerging from our studies and from previous studies on Abl and other tyrosine kinases is that the kinase scaffold is very tolerant of mutations in the hinge loop that lines the ATP-binding site. A discouraging consequence of this fact is that these mutations are likely to affect a wide range of ATP-competitive inhibitors—even ones from distinct chemical classes—as most ATP competitors are sensitive to the active site's architecture, to which the mutated residues contribute considerably."
Put simply, the kinases we're targeting have more room to maneuver than we do as medicinal chemists. They can mutate quite a bit and still function, shedding the key binding motifs that our drugs are targeting along the way. We're going to have to work a lot harder to come up with effective combinations.