What happens when you expose a reactive, covalent-bond-forming compound to cell extracts (or to living cells)? The answer is complicated. You might expect the compound to go around grabbing every reactive group it sees, shotgunning across the proteins it encounters and labeling without discrimination. But it turns out that you only see that with the nastiest compounds, the sorts of things that fume when you open the reagent bottle. Even things that look fairly reactive actually pick their targets more carefully than you’d think (here’s a post with a number of links to discussions on this).
There are several ways you can tune this effect. If you have a weakly active electrophile in your probe (such as an epoxide), then it’s probably only going to react when it fits just right next to a suitable nucleophile, which means that the affinity of the compound itself is going to mean more than the reactivity of the warhead. If you don’t have many binding features in your molecule and not much of a warhead, it’s probably not going to do much at all. That’s been the question behind the idea of reactive covalent-linking fragment-sized compounds, then: what’s the balance between the one-size-fits-many (to some degree of binding) behavior of fragments and the reactivity of the covalent linking group?
There have been a few examples exploring this area in the literature, and a new paper from Ben Cravatt’s group at Scripps has more to add. (Here’s a post at Practical Fragments on the paper as well, and an article from the San Diego Union Tribune). This work takes a library of 50 reasonably small electrophilic probes (mostly chloroacetamides and substituted acrylamides) and runs them across the human proteome. Careful mass spec analysis showed that these molecules labeled 758 Cys residues in 637 proteins (out of 2248 that they could quantify), and a closer look at just which residues and just which proteins showed some interesting features. Good old druggable proteins with reactive Cys residues in their active sites accounted for some of the hits, as you would have expected. But there were also plenty of unusual hits – transcription factors, scaffolding proteins, and others that don’t usually pick up on small molecule compounds, much less fragment-sized ones. Overall, only 14% of the proteins that reacted were found in the DrugBank database.
The reactivity was concentration-dependent, and even some of the more wide-ranging compounds showed selectivity if you dialed down their concentrations in the assay. And there were wide variations among the proteins when it came to which electrophiles they liked – looking at the heat map, I’d have to say that it would have been impossible to predict which of the fifty probes would show activity towards any given protein, because they’re all over the place. I get the impression that 150 probes or more would probably have given a similar result – there’s really no such thing as a Cys-reactive compound that’s just going to go in and pick them all up. (Update: well, maybe iodoacetamide, which was used as the positive control. See the comments section). You can predict whether a given protein has Cys residues that are more likely to be reactive, but not what they’re going to react with.
And there’s apparently no such thing as a single thiol that’s going to pick up all the reactive fragments, either. Interestingly, the hit rates of the fragments did not really correlate with their reactivity towards glutathione, which is a quick-and-dirty assay often used to assess reactive compounds. It may well be time to get rid of that one; there are just too many other variables, it seems, for it to tell you anything really useful.
The group also tried the experiment with 20 of the probes on intact cells. This mostly recapitulated the lysate data, but with some real outliers in both directions (which is a phenomenon that’s been seen before in protein-labeling experiments). One interesting effect was noted with caspase 8 – there were several fragments that labeled it in both experiments, but none of them seemed to be inhibitors of the enzyme when they were assayed, even at high concentrations. It turned out that they were labeling the (inactive) procaspase form of the protein, and the paper shows some further SAR development of one of the fragments into what seems like a very useful chemical probe.
Beyond chemical probes, there have been many reports (two recent ones) of covalent Cys targeting as a possible mode for drug action, and this sort of study would seem to lead right into some useful approaches for that as well. If your interests overlap at all with the topic of covalent drug mechanisms, chemical probes, cysteine reactivity, or new target discovery, the whole paper (which is a big one, once you get into the Supplementary Information) is well worth a close look.