How small-molecule drugs fit into binding pockets in their targets is one of the central questions of medicinal chemistry. A new paper from a group at Oxford gives a good example of how varied that process can be – it’s looking at a number of drugs that have been shown to interfere (to some degree) with infection by the Ebola virus. Ebola has been the subject of a number of screens, which you can be sure were run by people paying very careful attention to their lab technique and protective equipment. A surprising number of approved drugs have shown some activity – no rockin’ inhibitors at the nanomolar level, unfortunately, but definitely a higher hit rate than one might have expected from an antiviral screen.
The compounds that hit certainly didn’t seem to have much in common. Among them are ibuprofen, the estrogen receptor modulator toremifene, the muscarinic antagonist benzatropine, the old calcium channel blocker bepridil, and the serotonin reuptake inhibitors Paxil (paroxetine) and Zoloft (sertraline). That’s a pretty good spread of mechanisms and structures, and no, the two SSRIs don’t seem to lead anywhere mechanistically. So what’s going on?
It turns out that all of these drugs bind to the Ebolavirus glycoprotein (EBOV GP), a key surface protein on the viral particles that is responsible for its attachment to cells and entry through the cell membrane. Not only do all these compounds bind to the same protein, they actually bind to the same pocket, albeit by forming different interactions in each case. At right is an overlay of the six structures, and you can see that they carve out a common volume, for the most part. You have to be careful with that sort of thinking, because most compounds this size can be jiggered so that they look to be carving out such a space, but in this case, though, these are structures based on X-ray data of the EPOV GP/ligand complexes, collected at the heavy-duty DIAMOND synchrotron source in the UK. There are several interactions that are suggested by the data, among them that hydrophobic subpocket near the V66/A101 residues (where a phenyl ring of bepridil fits perfectly). The X-ray structures also show some movements in the protein itself on the binding of different inhibitors, although none that seem directly related to the small molecule interactions.
But sertraline doesn’t hit either of the subpockets; it’s all out in the middle. Two molecule of benzatropine fit into the protein at the same time (one of those is shown in green in the figure above), and paroxetine (which is in grey) doesn’t seem to fill all of the common volume very well at all. Interestingly, all the interactions seen are hydrophobic ones – there’s not a hydrogen bond in sight. And none of the compounds exploit all the interactions seen across the whole set, which strongly suggests that there should be hybrid structures that do a better job.
All of these compounds are way up in the micromolar range in their binding constants, and they all show destabilization of the EBOV GP in thermal shift assays. Those of you familiar with fragment-based drug discovery will recognize the process – low affinity, multiple molecules per site, different compounds scattered around the binding pocket, and so on. These compounds are larger than traditional fragments, but the the region where they’re binding is larger than many traditional binding sites, so you end up in the same mode.
If you’re willing to go out into these weaker interactions, you’d probably find that an awful lot of drug molecules hit an awful lot of things at (say) 100 micromolar. It’s just that most of these binding events have little or no functional relevance. And viruses have very few moving parts to them, which often makes drug discovery difficult. In the Ebola case, though, this protein is a critical, unique, Swiss-army-knife thing whose functions cannot be messed around with. Bepridil, for example, provides 100% protection from Ebola infection in a mouse model and sertraline hits 70%, which is not too bad at all for micromolar phenotypic screening hits. Even more suggestively, the antiviral efficacy correlates quite well with the thermal shift data, a strong indication that messing with the protein’s conformation is indeed the mechanism of action.
You’d have to think that an optimized EBOV GP binder could be very effective indeed, and this paper points the way to assembling such compounds. You’ve got the protein target, you’ve got a strongly relevant biophysical assay, you’ve got X-ray data from a list of compounds in the binding site, and you’ve got an animal model – that’s about as much as a med-chem team can ask for down here on Earth, and I hope that something comes of it.