A colleague mentioned to me the other day that Sunesis Pharmaceuticals had let many of its remaining research staff go back during the summer – they’re battening down to try to get their main clinical candidate through for leukemia and ovarian cancer. That’s a common phase of life for a small company trying to go it alone. Clinical trials are expensive, and so are scientists, and sometimes a company finds that it can’t afford both at the same time. Amylin, to pick one example, went through so many cycles of that (starting in the mid-1990s) that I completely lost count.
The Sunesis news struck me, though, because if you go back a few years in the literature, they’re all over the place. The company was aggressively investigating (and promoting) a technique called “tethering” as a platform for drug discovery. Back around 2003, they were all over the journals with it.
Tethering was one of those neat ideas which seems to have been a lot of work to reduce to practice. It’s a variation, in its way, of another one of those techniques called Dynamic Combinatorial Chemistry. In DCC, you take a good-sized collection of compounds which can form reversible bonds with each other. Thiols (R-SH) have been used a lot, since they can form disulfides (R-SS-R), which can easily come apart and re-form with other thiols. In the presence of some target or template, such as the binding site of a protein, the idea is that any disulfide combination that manages to bind well will get enhanced in the final mixture, since it spends more time out of the swim of potential reactants. Comparing the product distribution with and without the target protein can point you to a potential lead structure to optimize. (You can also turn it around and make synthetic receptors (PDF) for molecules that you're interested in).
The idea behind tethering was, at least in one of its main variations, to introduce an extra thiol group into a target protein somewhere close to its active site. Then this mutant protein would be screening against a library of small molecules with thiol groups of their own, with the idea that if there was a binding site near that thiol that it would be found by preferential disulfide formation between it and some member of the screening library. Then came the second step. Normal, unmutated protein would be exposed to a mix of that preferred thiol and a library of other potential thiol coupling partners, in an attempt to find another preferred extension into the binding cavity. So this was basically a way to do DCC, but giving it a leg up by trying to make sure that there was a good amount of at least one thing that could bind to some relevant part of the target.
That tells you that standard from-the-ground-up DCC must have some difficulties, since if it worked as well as its concept you wouldn’t need to put your thumb on the scales like that. But I was never sure how well tethering worked, either. The company published numerous examples of it, but I don’t know if any of these compounds ever got anywhere (and indeed, I’m not at all sure that their current clinical candidate was discovered by this technique).
There are several places where things could break down. Making a mutant protein introduces some uncertainty, for starters. That SH group might not change things, or it might change them just enough so that the binding site you find doesn’t quite exist when you switch to the wild type. And any binding site you find in the first round isn’t necessarily a productive one – the original protein SH group was targeted to try to dangle out over the right part of the protein, but there are no guarantees about that. Past that, even if you get through the second round and find some new disulfide hits (no sure thing), they are, well. . .they’re disulfides. And those are poor bets for drugs.
That’s where the real weak point of DCC is in general, to my mind. Using reversible reactions gives you compounds with too much potential to fall apart, so the first thing you have to do is replace those bonds with something sturdier – and that’s not always easy, or even possible. There are very, very few clean substitutions available in the chemical world. Nothing’s quite like a nitrile except a nitrile, and there’s only one thing shaped exactly like a t-butyl group: another t-butyl. Likewise, the only thing that’s guaranteed to look and act like a disulfide is a disulfide. A two or three carbon chain replacement is the logical place to start, but that might be synthetically tricky, or (even more often) might turn out to be a completely different sort of compound once you’ve made it.
In the end, I think tethering turned out to be an excellent means to get some very interesting papers published in some good journals. (The publications have continued to this day). But beyond that, I’m not so sure. I’d be glad to hear from any ex-Sunesis people with other views. . .