OK, chemists and biologists: name a small molecule that binds really, really tightly to its protein partner, without forming an outright covalent bond. Everyone in the room is probably going to shout “biotin“, and for good reason. It has perhaps the most famous binding constant of all, down into the femtomolar range with avidin protein. The interaction with the related streptavidin from bacteria is only a bit weaker. That’s roughly a million times more potent that most small-molecule drugs hit their targets, and this next-thing-to-irreversible behavior has of course been borrowed and used countless times for setting up assays and chemical biology experiments since then.
But how does it do that? Biotin is a rather small molecule, only 240 MW, so it has insanely high ligand efficiency in whatever way you care to measure it. It’s hard to imagine doing better. This new paper from Darryl McConnell at Boehringer Ingleheim examines that binding in detail, and it has some interesting lessons. Chief among them is that many of the key interactions are not things that we are used to exploiting in the normal practice of medicinal chemistry. There’s a urea group in the molecule, and a cyclic sulfide, and out on the side chain there’s a carboxylic acid. But a lot of the binding energy comes from the other C-H hydrogen atoms in the structure, which are things that we tend to ignore in favor of concentrating on heteroatoms. There are 18 interactions apparent in the X-ray cocrystal structure, but ten of those involve hydrogen atoms, and only two of them are classic hydrogen bonds (the NH parts of the urea). The other 8 are CH-to-pi and CH-to-O interactions.
There are lessons to be learned even among the more obvious interactions, though. The urea group fits beautifully into an “oxyanion hole” region, and McConnell emphasizes that this not only lets you form some great hydrogen bonds (those two donors mentioned, and three acceptor bonds onto the carbonyl), but it also reduced the penalty that you have to pay for pushing water molecules out of that region. About 30% of biotin’s binding energy is right there. The paper goes on to predict that the number of such oxyanion holes in known protein structures has been underestimated, because the position of the H atoms in that part of the protein is crucial and we generally don’t have hard data at that level of detail.
But it’s those CH-pi interactions that the paper emphasizes even more, since they’re even more seriously underappreciated. They start with a real advantage, being inducible dipoles instead of permanent ones. A permanent dipole is going to attract waters of solvation, and you’re going to pay the price when you shed those to fit into the protein binding site. But an induced dipole only comes to life once it’s in the binding site itself – no waters to shed. Many chemists will find it weird, but it’s the aliphatic side chain leading out to the COOH that accounts for a good part of biotin’s binding. There are several C-H interactions from what appear to be completely inert and invisible methylene groups, reaching three different tryptophan residues on the protein. Something to keep in mind if you find that your protein target has some reachable Trps, for sure. The CH-O interactions mentioned earlier have in fact been taken advantage of in several kinase inhibitors, as the paper notes, but there are still probably some missed opportunities in that area, too.
The carboxylate group itself also illustrates a desolvation principle. It’s not buried down in the binding site in a death-or-glory attempt to pick up polar interactions. Instead, it’s out on the edge of it where it can still interact with a couple of water molecules, reducing the desolvation penalty that’s so prominent with this charged group.
Now, 18 binding site interactions for a molecule this size is pretty impressive, and that’s how you get down to those crazy affinity numbers. We’re not going to be able to just stroll in and take advantage of all of these things at once and crank out the femtomolar binders at will. But we don’t need to! Femtomolar affinity, frankly, is too much for a lot of drug targets. I’d be happy to settle for driving a bunch of stuff down from micromolar to nanomolar – my needs are simple. You’ve probably heard the one about “great artists steal”, as attributed to Picasso (Steve Jobs spread that one around a lot). But Picasso himself probably never said it. The real quote is more subtle, and it’s from T. S. Eliot, in his 1920 essay on Phillip Massinger: “Immature poets imitate; mature poets steal; bad poets deface what they take, and good poets make it into something better, or at least something different. . .” That’s the spirit. Let’s steal from biotin, and make something different.
