Here’s a nice protein degradation paper (and its earlier BioRxiv version) that suggests a new assay that could turn out to be quite useful. In a lot of these experiments, protein levels are quantified (more or less) by Western blots. That’s fairly labor intensive and imperfectly reproducible, and I think that a lot of people get the feeling that hey, it’s 2022 now, isn’t there a better way to follow the levels of a specific protein in the cell. There certainly are other techniques, but some of these rely on engineering or labeling the protein of interest, and that’s not a lot of fun, either (not to mention the constant suspicion that you might have altered the native protein’s function in some way that you’re missing).
This new one is a TR-FRET assay, which is certainly a technique that’s proven itself over the years in too many high throughput screening campaigns to count. Like every assay, it has its idiosyncracies, but those are pretty well worked out by now, and this current idea avoids some of the ones that you hit when you’re directly screening a bit library of compounds in vitro (chief among them weirdo fluorescent emitters/quenchers that kick your data around to random places). Here’s how it works: so you’ve got a protein of interest – in the paper, they demonstrate the technique with BRD4, which is a popular platform for this sort of thing. As with all protein degradation ideas, you do need a compound that binds to that protein, and for BRD4 you have the longtime favorite JQ1 that came out of the Bradner lab and spread out to research groups everywhere. The engineering that comes in here is the production of a labeled JQ1 – the authors use a fluorescein conjugate and validated that it does still bind to BRD4 the way it’s supposed to. That’s the acceptor of the TR-FRET system, what’s going to glow when there is indeed resonance energy transfer.
You also need a good antibody to the protein of interest, and there are indeed such for BRD4. The paper uses “nano-secondaries”, which are small single-domain antibodies (from our friends the alpacas) that bind to that anti-BRD4 antibody (thus the “secondary”, where you have an antibody binding to an antibody). The use of the alpaca nanobodies cuts down on the size and complexity of the whole system and makes the assay more feasible. Those secondary antibodies are in turn engineered with a TR-FRET donor (CoraFluor-1), so now you have your total TR-FRET system: if the anti-BRD4 antibody is bound to its target, those nanosecondaries bound to it are along for the ride. If the FITC-labled JQ1 binds in turn to BRD4, you get a strong FRET readout as the fluorescein and the CoraFluor come into proximity. All is well. That means that you can take cell lysate that has this system glowing away in it and start screening compounds to find other ligands for BRD4, should you so desire. You start out with a strong FRET signal, and look for things that decrease it. I’m sure that there are going to be some false positives – there always are in FRET assays – but you’ll also find things that displace the labled JQ1 from its binding site, killing the FRET activity in the process.
But you can run the assay in a different way to use it to quantify protein levels. Take your cells, and treat them any way you want – such as dropping a BRD4 degrader into them, or some new compound that you hope turns out to be one. You wait the appropriate amount of time (you’re going to want to take several time points, most likely) and lyse the cells, then drop that whole antibody/secondary nanobody/labled JQ1 mix into them. The resulting FRET readout will be proportional to the amount of BRD4 that’s floating around in the lysate: voila, protein quantitation.
The limitations of the assay are that you need a good antibody for your protein of interest – selective of course, and one that doesn’t interfere with the sites on the protein that you’re hoping a degrader molecule will bind to. You need a known binder, too (like JQ1 for BRD4) as a reagent, and you need to be able to make a fluorescent-conjugate version of it that still binds to the protein and doesn’t get up to any other funny business. To be sure, that binder probably doesn’t have to be selective for your protein (the selective antibody with the other half of the FRET setup should take care of that issue). Not every system will bring the time and the place and the loved one all together (you get Robert Browning today instead of A. E. Houseman), but when you have them, this looks like a good system to try. On the plus side, many of the other assays that try to do this need some extensive cell engineering, which of course has to be re-done and re-validated every time you think about looking at another cell type, while this one just drops into whatever cell line you choose.
Now one of the things about the bifunctional protein degrader field is that it has been leaning very heavily on two enzymes: cereblon and VHL. These are the E3 ligases that bring in the ubiquitinating complexes that do the deed to the protein target, slapping ubiquitin side chains on it and thus causing it to be dragged off to the proteasome for destruction. Those two are the E3 ligases for which we have by far the most well-characterized small molecules for binding: all sorts of maleimide derivatives for cereblon, and various hydroxyproline-benzyl-thiazole structures for VHL. Other E3 ligases have been tried out as various ligands for them are discovered (and it’s for sure that not all of these are public yet), but the huge bulk of the work in the field is still with those two. But it’s a wild frontier right now, because picking which E3 ligase might do the job for your targeted protein (and which degrader molecule might do the best job of recruiting it) are still stubbornly empirical questions that we’re having to answer by relentless experimentation. So new assays are very much welcome.
This paper goes on to attack that issue by making a bifunctional degrader out of JQ1 on one end and a molecule called celastrol on the other. Celastrol is an interesting critter. It’s a triperpene derivative, related to the steroid family, with a very reactive quinone methide group in its structure. So it roams around pouncing on nucleophiles (or letting nucleophiles pounce on it, depending on your point of view), particularly exposed and/or reactive cysteine residues on proteins. It does all kinds of stuff in cell assays and animal models because of this, and you can basically find papers that will illustrate it doing anything you feel like seeing – antibacterial, antiinflammatory, anticancer, antidiabetic, antioxidant, insecticide, you name it. It’s probably a decent floor polish, although I’m not sure I’d recommend its use as a dessert topping. But there are a lot of E3 ligases that have exposed Cys groups on them, which various people have been trying to target with reactive molecules to make covalently linked bifunctional degraders. That’s another wild frontier, as you can well imagine, because there’s covalent-covalent and reversible covalent, for starters, and it might be that compounds of the first category don’t survive the passage of their linked protein into the maw of the proteasome (and therefore doesn’t partake of the multicycle catalytic nature of other bifunctional degraders). Or maybe they do! This is still an open question. This has notably been tried out with the reactive triterpenoid bardoxolone by the Nomura group.
And when you make this celastrol/JQ1 degrader, it does indeed lower BRD4 levels, as illustrated by the assay outlined above. The group did the appropriate control experiments, such as looking for an increase in ubiquitinated BRD4, demonstrating rescue with inhibitors of proteasome function, and so on. The degradation profile seems to be different from the bardoxalone-derived degraders, which probably reflects differences in the various E3 ligases that are being roped in. Just which E3 ligases those are is another question, naturally.
The assay described by this work really does look useful, and I expect we’ll see it adopted shortly in other work. The celastrol degraders too, most likely, for people who want to target some particular protein and who don’t care what ubiquitinating complex is doing the work. I regard celastrol as a compound of easy virtue and might worry a bit about what else its bifunctional is doing in the cell, but it remains to be seen if that’s a problem or not!
