Skip to main content

Chemical Biology

If You Want to Degrade, You Have to Get In the Door

Back for more, and welcome to 2020! Frankly, that sounds like a year out of a science fiction story, but I will admit that it’s not the first year about which I’ve had that thought, either. Let’s get right into the drug discovery with a new paper from a team at UNC in the very fashionable field of targeted protein degradation.

As those who have worked in this area will have come to appreciate, there are a lot of moving parts in a successful degrader. Your bifunctional molecule has to bind the protein target and also bind an E3 ligase that will be dragged in to ubiquitinate it, of course. And it has to bring these proteins together in a productive manner – if the ligase can’t reach a lysine residue on the target protein, or if any ubiquitination that does happen isn’t quite right to set off targeting to the proteasome, then you’re not going to get far. And as the authors of the current paper note, if your bifunctional molecule doesn’t get into the cells very well to start with, you’re hosed right from the get-go. OK, that’s not quite the way that they phrase it, but that’s the case. “As a result,” they say, “degrader development often involves synthesizing and testing multiple iterations of compounds without a clear understanding of what exactly needs to be improved” and that’s no tall tale, either.

They’re proposing the use of a chloroalkane penetration assay (CAPA) to shed some light on the first step. That one is an offshoot of Halotag technology: what you do is set up a cell line that expresses a Halotag-GFP fusion protein that’s expressed on the outer side of the mitochondrial membrane. Then if you expose the cells to a chloroalkane-tagged (ct) molecule, the Halotag component of that protein will react with them on exposure, since that’s what it does. After that dose and a suitable waiting period, you then expose the cells to a ct-dye molecule whose cellular permeability has already been established. It goes in and labels whatever Halotag protein active sites didn’t pick up your earlier investigational molecule, and you measure the fluorescence of that species, sorting things out via flow cytometry and using the GFP readout as a control to make sure that protein levels didn’t do anything funny along the way. The more fluorescent the cells are at the dye wavelength, the worse the permeability of the original target molecule must have been.

What was done here was labeling of a well-known BRD4 degrader (MZ1) with a chloroalkane tag, a process that was helped along by the known crystal structure of MZ1 with its two protein partners. That degrader has a known BRD4 ligand (JQ1) and the standard ligand for the VHL ubiquitin ligase, and helpfully, the team also made a series of chloroalkane-tagged species (just the VHL ligand side, just the JQ1 side, and so on. Some of these had the ct tag attached pretty directly, and some had it spaced with a 3-unit polyethylene glycol chain. And finally, all of these compounds were also tested in the Caco-2 in vitro permeability assay, which I would say is simultaneously widely used and widely regarded with suspicion as a reliable indicator of real-world cell permeability. It does have something going for it, though, in that permeability in one direction is assumed to be largely passive diffusion, while permeability in the other is a mixture of passive diffusion and active transport.

So how do all these compounds look? Well, the most striking result is shown in the abstract and reproduced above. ct-JQ1 is quite permeable and does a fine job labeling the Halotag protein inside the cells. But the ct-MZ1 degrader is over 100,000 time less permeable. Even taking ct-JQ1 and adding the above-mentioned PEG spacer to its structure decreased its permeability by >16,000-fold. So yeah, if you were looking for some more empirical proof that cell permeability is a precious thing and easily lost, here you are. It’s worth noting, though, that the ct-VHL species was also pretty permeable (although certainly not as readily as JQ1), and adding PEG spacers to it only knocked it down by a few fold. In general, the various labeled components showed the sort of broad dependence on molecular weight and polar surface area that one would expect for cell permeability, but those JQ1 numbers show that you can’t take anything for granted – the effects may go in the general direction you expect, but can definitely be nonlinear.

Meanwhile, in the Caco-2 assay, JQ1 and its chloroalkane version showed passive diffusion (since this one doesn’t need the ct tag to run, a comparison of plain JQ1 and ct-JQ1 was possible), but neither the parent nor the chloroalkane-tagged versions of the VHL ligand nor the MZ1 degrader itself showed any. Interestingly, all six compounds showed some permeability when run in the basolateral-to-apical direction, the one that includes active transport.

What about that ct-tagged MZ1 as an actual degrader? Well, Western blot analysis showed that although it’s wildly less permeable than the parent compound (as mentioned), it is just as effective in degrading BRD4 protein. So if you were looking for some more empirical proof that such degraders are indeed catalytic and that you don’t need much getting into the cells in order to be effective, here you are again! You can look on that result two ways – the sanguine approach would be to say hey, I don’t care much about permeability, do I, because it turns out that I hardly need any. But that would be a bit too laid-back, because your permeability can, in fact, be a flat zero and it’s kind of difficult to overcome that. You may not need or want to finely quantify these things, but knowing that you’re getting through the cell membrane (to any degree) versus bouncing off of it is important.

My own guess is that cell permeability is overall a lesser variable than productive ternary complex formation, etc., but individual cases can certainly bite you if you take them for granted. And we don’t understand that ubiquitination machinery very well, to be sure, but another thing that this paper illustrates is that we don’t understand something as basic as crossing a cell membrane all that well, either – not at the level of detail we’d like, anyway. That, sadly, has been the case for decades. What the world needs (well, our part of it) is a completely label-free assay to determine this sort of thing in living cells, but that’s another longstanding problem. There are a number of useful partial answers, but nothing general. . .yet.


12 comments on “If You Want to Degrade, You Have to Get In the Door”

  1. Barry says:

    That chloro-alkane appendage is not a small perturbation on the properties of the molecule tagged. Proportionately, it is a much bigger deal than e.g. farnesylating RAS, which is known to change its spatial distribution (and activity) dramatically.

  2. Mazdak says:

    Re “What the world needs (well, our part of it) is a completely label-free assay to determine this sort of thing in living cells, but that’s another longstanding problem.”

    Should be amenable to direct visualization by cryo-EM, electron tomography in particular:

    1. Derek Lowe says:

      That paper is in the “to blog about” queue as well, as it happens!

      1. Ben Engel says:

        Cool! Thanks for considering blogging about our paper on ER-associated proteasome clusters. Another paper, which coincidently came out alongside ours in the same journal, makes a related discovery: IRE1 clusters in the ER membrane during stress. I think there is a lot more to understand about the spatial compartmentalization of the ER membrane for different quality control and secretory pathway functions.

  3. PAMPA fan says:

    I think the permeability results are a great example of the nebulous, quantified but not understood, passive permeability penalty that crops up as you get far beyond 500 Da. The two molecules in the abstract there have roughly the same lipophilicity (calculated 4.8 versus 4.4), but dramatically differ in molecular weight (1225 vs 605). This paper (which I am shamelessly an author on) attempts to quantify the effect of size on passive permeability while controlling for lipophilicity. The plot in that paper’s abstract would suggest about a 5 log unit difference in passive permeability between something of ~600 MW and ~1200 MW, exactly what they find. ct-MZ1 is larger than cyclosporin. This class of molecules are going to be hard to medchem from a physiochemical property standpoint, but as you point out their catalytic MOA gives you much more wiggle room than a traditional, stoichiometric inhibitor.

  4. Wavefunction says:

    Also shows you how hard it would be to computationally predict permeability for beyond Ro5 compounds (it’s already pretty hard for standard druglike molecules).

  5. NotThatLipinsky says:

    Can PROTAC technology be used to degrade membrane-bound target proteins (e.g., GPCRs, ion channels, transporters, etc.)? Or only those target proteins that are NOT membrane-bound (e.g., select cytosolic enzymes, etc.)?

    1. Derek Lowe says:

      I’d say that’s an open question. You might think that the protein involved would need to be able to be dragged off to the proteasome in order for this technique to work, but on the other hand, ubiquitination of a membrane-bound species might cause it to be internalized for just that purpose. . .

    2. Anon says:

      Yes, it can – Of course, whether transporters, for example, can be internalized and degraded as efficiently / through similar pathways as RTKs is unknown, but there are at least examples of membrane protein-targeting PROTACs.

      1. Degrade Everything says:

        You could also use an ENDTAC to internalise and the trans-membrane protein and then degradation occurs.

  6. RBWoodward says:

    Zero membrane permeability and, don’t forget, zero oral bioavailability. But, hey, they’re “catalytic degraders”…so we don’t need much target engagement. Good Lord, we’re starting to spin fairy tales here. Drugging the undruggable with the undruglike. Who’s idea was that anyway? We’ll look back on these projects and cringe 5 years from now. RB

  7. fake says:

    b 0 omer

Comments are closed.