Everyone’s pouring money and effort into targeted protein degradation, and no wonder: it’s a completely new therapeutic mode that promises the ability to do something we’ve never been able to do before – that is, just make a particular protein disappear from inside the cell. Sure, we can (fairly often) make small-molecule inhibitors of some of these proteins, but that just blocks up their active sites, and most proteins have more going on than just that. And we can target such things back at the genomic level, but that gives the cells (and the tissues) time to come up with alternate pathways. Yanking a given protein right out of its whole interaction network in real time promises different effects, unattainable by other means, and that’s often what you see.
But there are great big swathes of this process that we don’t understand very well, and this new paper addresses one of them. These degrader molecules, be they bifunctional molecules with linking groups in the middle or smaller “molecular glues”, have to get into the cell and will presumably find their way to different regions in varying amounts. Intracellular phamacokinetics is (famously) a black box for us in most cases, so that’s about all we can say. Now, one of the things these degraders have going for them is that they are catalytically active. After they deliver their target protein to the embrace of the ubiquitinating enzymes (and thenceforth to the destroying maw of the proteasome) the degrader molecules escape destruction, not being targets themselves for the proteasome active sites and not being covalently attached to their own protein targets. They can then circulate around and do the whole trick again, which means that you might not need such large amount of them around. That’s a good thing, since some of these have challenges in their cell penetration and since (as just mentioned) we don’t really know where they’re going anyway.
There’s another factor at work, too: what is the overlap between abundance of the protein being targeted for degradation and the E3 ligase enzyme that’s being hauled over to work on it? The paper linked above looks at several types of bifunctional degraders targeting different proteins that are known to have widely varying subcellular localization. This was done by modifying the target proteins with either Halo-tag or dTag regions, which sets them up for degradation via either cereblon (CRBN) or VHL-mediated pathways, and using well-established localization signals to send them to various subcellular compartments (as well as control experiments with no such localization). And there are indeed many differences! A Halo-VHL degrader system performed well in all the compartments except the Golgi apparatus lumen. By contrast, the dTag-VHL system was noticeably less effective in the outer mitochondrial membrane, the Golgi, the peroxisome, and the lysosome, but held up fine in the nucleus, plain ol’ cytoplasm, and the endoplasmic reticulum. The dTag-CRBN system did those last three too and added back efficacy in the outer mitochondrial membrane, but again failed in the Golgi, perioxisomes, and lysosomes. A structurally different dTag-VHL system (dTagV-1) had the same subcellular compartment behavior overall, however it was more efficient than “dTag classic” in the ER and less efficient in the peroxisomes.
That tells us that we really can’t assume anything. The differences between the Halo-tag and dTag degraders would suggest that there are changes across these two systems in how the target proteins react with the degrader molecules and how they’re presented to the ubiquitinating complex. This ternary complex formation is (I think it’s fair to say) one of the biggest black boxes in the whole process. And the two different dTag-VHL degraders give you an example where two very similar systems that are recruiting the same ubiquitin ligase are mostly similar but can still differ in ways that you wouldn’t have been able to predict, either. And you can of course see differences across the choice of CRBN or VHL – the former, for example, seems to be better for nuclear-localized proteins in general, while VHL was better for targets in the ER overall. This work ties in with a 2020 report on differently localized solute carrier transporter proteins, showing that CRBN-mediated degradation varied across the different locations as well.
One problem here is that we don’t know how much CRBN and VHL are available in these different spots to start with. That data might explain most of these differences, but there are surely other factors that are kicking these numbers around as well. Note, for one thing, how a Hal-VHL system was able to work in some combinations that a dTag-VHL wasn’t able to: it’s hard to explain that by VHL abundance! And as the authors note, these results suggest that a recent review of the possibilities for protein degradation was probably too pessimistic, because that one downgraded the predicted chances for some of the membrane-enclosed compartments that worked in this latest paper. But overall, I’d say that we’re still very much in a “try it and see” mode in targeted protein degradation. I am skeptical about making actionable predictions under current conditions, because there are just too many variables to pick from that we know so little about. You choose a ligand for your target protein, a linker system to the E3 ligase ligand, and from different E3 ligases, and all of these choices have their own efficacies modified by pharmacokinetics, by ternary complex formation, by subcellular concentrations of all the components and who knows what else. Don’t talk yourself out of things, then, is my advice. . .
