I have always had a liking for the technique of having target proteins assemble their own inhibitors. This goes under several names: target-guided synthesis or protein-templated reactions more generally, and in situ click chemistry when the triazole/alkyne reaction is used as the assembly method. But the idea is the same in each case. You bring in two molecules, each of which are functionalized with weakly reactive groups, and if they come into proximity due to their binding conformations in the protein, they will have a chance to react much faster than they would in solution. You could in theory sort through a number of combinations simultaneously, and only the ones that lined up properly in the active site would produce anything. And if they line up that way in the active site, the molecule produced is quite likely to be an inhibitor. The triazole click reaction is quite useful for this work, because the azide and alkyne are pretty inert by themselves, and the background rate of reaction (without proximity-induced acceleration) is very low.
This has been demonstrated in a number of systems over the years (here’s a recent review on the area). I don’t mind mentioning that I’ve done some demonstrating myself (nontriazole work that is unpublished, although I’ve given a couple of talks about the work). No one is still sure how generalizable a technique it is, but when it works, it’s like having magic powers – the affinities of the final products can go up as far as the products of the two precursors’ separate affinities (perhaps even more if the newly formed linker contributes to the binding as well?), which means that you can start with two fifty-micromolar fragments and end up with a nanomolar inhibitor (perhaps).
You normally do these experiments with purified protein in vitro, and to be honest, that’s one disadvantage: the more protein you shove in there, the better the chances you have of making enough product to detect if some combination actually works. Target-guided synthesis is not an efficient way to crank out a pile of inhibitor; you’re only going to make a little of it, which is why you need a background rate that makes essentially none. LC/MS detection is the norm, because of its sensitivity, but you can still end up using a pretty fair amount of protein to explore the combinations with confidence. Product inhibition is presumably a problem, too, depending on off-rates: if you do make a great inhibitor, it can sit in the protein’s binding site and preventing any more from being made.
More recently, there have been reports of doing this sort of thing inside living cells. That’s quite a step, since you’re no longer availing yourself of artificially high protein or reactant levels. That latter paper, though, shows the formation of a triazole-containing carbonic anhydrase inhibitor, previously shown to form by target-guided synthesis in vitro, by using red blood cells instead. You’d have to think that this has the best chance of working when you’re targeting a protein that’s present in high concentrations naturally, and this certainly fits. But you know what else fits? Ribosomes. Boy, are there are a lot of ribosomes.
That’s where this paper came from a couple of years ago, from a team at Temple who showed that the antibiotic solithromycin, which already has a click-derived triazole in its structure, could be produced inside live bacteria from its two click precursors. The mode of interaction between solithromycin and the bacterial 70S ribosomal subunit is known from X-ray crystallography; the triazole side chain is making binding contacts.
Now the same group reports an extension of this work. They’re using a strain of S. aureus that’s resistant to most macrolide antibiotics, but is still susceptible to ketolides like solithromycin, apparently because of those extra binding interactions with the side chain: the azide-functionalized macrolide is very weakly active, while solithromycin itself, after the triazole side chain has formed, is >100x more potent. (As controls in both these papers, the alkyne fragments were shown to be almost totally inactive themselves, and the solithromycin formation was shut down by adding saturating amounts of another ribosome-binding compound for competition). In this new paper, eleven new alkyne-bearing side chain percursors were assayed, and the formation of the new products correlated with the potency of the products. And these could actually be detected in a 96-well MIC (minimum inhibitory concentration) assay, whose trends were confirmed by LC/MS analysis.
That’s something that you don’t see in this field – detection through the activity of the products. They’re typically formed in such small amounts (and you typically have so much protein in there) that a functional readout isn’t feasible. But in the living bacteria, things are lined up right for it to work: you have a lot of ribosomal protein, and its inhibition is a big problem for the cells. This opens up the possibility of doing target-guided antibiotic screens on a larger scale, with direct readouts on potency. False negatives could come from using reacting partners that aren’t very cell penetrant, whereas false positives could show up if a reactant is inhibitory on its own. But there should be plenty of room for a good assay window, you’d think.
The authors even suggest that this could be done more broadly than just targeting this known ribosomal binding site, and I’d very much like to see that idea tried out. It’ll be a foot race between the intrinsic amounts of the target proteins and the consequences of their inhibition – that is, is there enough protein to form products, and does its inhibition give you enough of a readout? But I think it’s certainly worth a try. My first suggestion would be to piggyback on known antibiotic structures, but run the click experiments in resistant strains to see if you can overcome that resistance with the new side chains you’re forming, but there are a lot of other possibilities. Pretty much anything that speeds up antibiotic drug discovery or opens up new areas in it will get my vote!