Skip to main content

Hard Thinking About Protein Degradation and Bifunctional Molecules

This paper (open access) is not going to be to everyone’s taste, but the people who are working in its area – and there’s an ever-increasing number of them – will want to read it closely. It’s a close look at the red-hot fields of targeted protein degradation and “molecular glues”, and for those who aren’t into those, the next paragraphs will give you some background. If you’re already into degradation – that didn’t quite come out right – skip ahead if you like:

So the idea behind targeted protein degradation (and now many related ideas besides) is the creation of a “bifunctional” molecule to hijack an existing cellular pathway. For TPD, you’re latching onto the protein ubiquitination system. Ubiquitin is a short protein tag that’s appended to lysine residues of other proteins, yet another post-translational modification, and Ub resides can be “stacked” onto each other into various sorts of oligomeric chains. If the right ones get added in the right quantity, it’s a signal that the protein thus tagged is ready for the cellular scrap heap, like a fluorescent orange waste disposal sticker. These Ub tags are applied by a set of enzymes (there’s a whole collection of them in cells) that are specialized for just that purpose.

Polyubiquitinated proteins are taken up by the proteasome, which as I’m fond of saying is a tubular structure resembling the planet-destroyer thingie in the old Star Trek episode “The Doomsday Machine“. Proteins enter one end of the proteasome and are ripped to shreds by a wall of protease enzymes facing the inside, and recycled amino acid rubble comes out the other end. So what if you could target specific proteins for this fate before their time?

That’s the plan. What you do is find a molecule that binds to some component of that ubiquitinating machinery (usually the part called the E3 ligase). There are many of these known, targeting different ligases, with (weirdly enough) the infamous thalidomide as one of the first to be proven for a particular E3. Then you find a molecule that binds your target protein – it can be an inhibitor, an allosteric site binder, whatever, just so long as it binds well. Now the so-obvious-it’s-dumb part: you take both of those molecules and build a linker between them to turn them into One Big Odd-Looking Molecule. One end of your new beast will grab onto the target protein, and the other end will grab onto the ubiquitination complex, and you will thus bring them into a close proximity that otherwise they would surely not otherwise enjoy.

The crazy thing is, it works: we have many examples now of such bifunctional degraders that manage to get into cells and cause drastic, sudden reductions in the amount of their targeted proteins as they get dragged, no doubt kicking and protesting, off to the proteasome shredder. What’s more, the bifunctional molecule survives the experience and goes back to degrade again, so you get true catalytic cycles out of the things. These have been shown to work in animal models, and several have now been dosed in human clinical trials – it’s a whole new way to attack targets of disease that we’ve never had before. We can block up active sites of proteins, but that leaves many of their other functions with other binding partners undisturbed. We can genetically silence expression of proteins, but that gives the cells time to come up with emergency backup plans in many cases. But reaching in during the middle of the cellular business day and zapping targeted proteins out of existence – that’s something new!

The idea doesn’t stop there. You can potentially bring *any* two proteins together if you have molecules targeting each partner. So you could haul a kinase over and have it phosphorylate the other, or a phosphatase to have it take any nearby phosphates off (this stuff has been demonstrated), and once you start thinking in that direction a whole list of ideas will occur to you. It’s not a small field now, but it’s going to get a *lot* bigger.

Molecular glues are another bizarre thing, with some similarities and some differences. In this case, you’re bringing two proteins together with a single small molecule – and some of them really are small – that binds to both of them simultaneously and may even arrange the two protein surfaces into a more favorable shape. I would not have thought that this was likely, or even possible, if you’d tried to sell me on it twenty years ago. But that’s what thalidomide itself does (when it’s by itself), and it’s also what how the auxin plant hormones work. I would have told you that this would be like trying to stick two cruise ships together by jamming a sailboat in between them, but proteins are a lot more complicated and interesting than your average cruise ship. People are currently searching all over the place for more of these things and trying to understand when they can work and how.

OK, then. By now there are a lot of examples of bifunctionals and several molecular glues, but the field has been, well, extremely empirical. Take those linkers that you have to build into make the bifunctionals. How long should they be, and what should they look like? Alkane chains, polyethylene glycol-like stuff, amides, sulfonamides, triazoles made via click chemistry? There are a lot of ways to do that, and they vary greatly in polarity, molecular weight, conformational flexibility, and more. People have tried all kinds of variations, and what we know for sure is that different linkers vary widely in their effectiveness once you’ve settled into a given system. But there are very few useful systematic rules; often enough you just make things and keep making them until something works, and when it works you’re not quite sure why.

The tricky part is that a ternary complex has to form for all these ideas to work, and that takes you into pretty complex kinetic and thermodynamic territory. As you can picture, there are a lot of possibilities about what binds to what and in what order, and plenty of rate constants that can kick the outcomes all over the place. This new paper tackles these in a comprehensive way. One key, as it details, is to separate the ideas of cooperativity and binary affinity. Many bifunctionals have very high binary affinity (each end binds strongly to its partner) but are nothing special in cooperativity. And many molecular glues are at the other end of the spectrum: the glue molecule has terrible binding (one-on-one) to Protein A, and terrible binding to Protein B, but bring all three of them together and you get a nice tight complex. Counterintuitive, to say the least, but this “induced cooperativity” is what makes the whole thing work, as the protein structures shift around. And it’s what makes screening for these things rather labor-intensive, and it also shows you how my cruise ship analogy breaks down, because they aren’t quite so flexible.

If you overdo the binary affinity part (through very tight interactions or high concentrations), bifunctionals suffer from what’s been called a “hook effect” – the binary interactions get so favorable that the ternary complex doesn’t form properly, which is truly not what you want. Any consideration of thermodynamics will of course break down into entropy and enthalpy, and there’s plenty of that to go around here. The enthalpy terms (energies of the actual binding interactions) can be cancelled out by the entropic penalties of assembling a more ordered ternary complex, particularly one that involves flexible parts that might have suddenly had to get a lot less flexible. And there are all sorts of potential ternary complexes, some of which are more productive than others.

The middle of the paper goes into detail on all this, and the latter part talks about cellular context. Which frankly, is even more of a black box. You get into all sorts of fuzzy discussions about the “tone” of the various systems (how much capacity they have, how much range, how much disruption of them might be needed to achieve the desired effects, and so on). The best way forward here seems to be the development of better high-throughput cellular assays, although no one’s going to turn up their nose at any new fundamental insights, either.

So if you’re a degrader/glue type, have a look. You’ll run into some things you already knew, but you’ll find others that you hadn’t devoted enough thought to, as well. Very few of us have devoted enough!

27 comments on “Hard Thinking About Protein Degradation and Bifunctional Molecules”

  1. awfulcritic says:

    Oh my.

    My brain had a Bulls**t bingo anyeurism just reading the title : by “not to everyone’s taste” I guess you mean “here’s the PROTAC hype train reformulated for all you leansixsigma types”.

    This is unfortunate becase it really does look like this paper

  2. GT says:

    on the Covid vaxx topic, I have not heard much chat on when boosters (delta, african variants) will be rolled out. One would think by late summer given one would swap out modular cassettes of the RNA approach. Any news?

    1. Derek Lowe says:

      It looks like the current mRNA vaccines (at any rate) still provide good protection against the Delta variant. Beyond that, though, who knows?

      1. GT says:

        Yes, protection against serious illness, hospitalizations, death.
        But , anecdotally in my personal sphere breakthrough infection even after double dose.
        I believe there is and will be data to support above two sentences.
        So, a case could be made for a late-summer variant cocktail vaxx, including the unknown durability of original jabs (months, year, years, decade, decades, century, century plus).

        1. GT says:

          Not sure how to transfer without taking too much of my time.
          But I placed comment in the current blog post (on TPD) only to give you the new idea for a future blog topic on boosters – scientific rationale, time lines, costs, etc
          Thank you for what you do.

        2. David says:

          There is a study comparing effectiveness of the Biontech and Astra Zeneca vaccines between the Alpha and the Delta variant in symptomatic cases. There was no signifact difference found for full vaccinated patients. There might be a small reduction but the data is not good enough (there is not enough data) to tell:
          “With BNT162b2 there was a small reduction in effectiveness post dose 2 from 93.4% (95%CI: 90.4 to 95.5) with B.1.1.7 to 87.9% (95%CI:78.2 to 93.2) with B.1.617.2. Numbers vaccinated with 2 doses of ChAdOx1 were smaller and the overall 2 dose vaccine effectiveness was lower than with BNT162b2 however the difference in vaccine effectiveness between B.1.1.7 and B.1.617.2 was small and non-significant: 66.1% (95% CI: 54.0 to 75.0) and 59.8% (95%CI: 28.9 to 77.3) respectively.”

          1. David says:

            Seems like my URL was eaten or I forgot it. Here is the DOI: 10.1101/2021.05.22.21257658

            And the title:
            Effectiveness of COVID-19 vaccines against the B.1.617.2 variant

            It has not been peer-reviewed, though!

      2. J Curwen says:

        Saunders Lab has now shown ADE in vitro and partially (what they call not) in vivo.
        They used the wildtype virus but it is well known that different serotypes (as in Dengue) are the real stuff. Since we now already approach more or less different serotypes with Delta and Delta Plus, I wonder what will be the outcome of these new variants in combination with the vaccine?

    2. sgcox says:

      For me the whole idea of regularly boosters because the level of circulating antibodies drops with time looks a bit iffy:
      Immune system naturally restricts expansion of B-cells if antigen is no longer present. Otherwise we would all become big balloons by keeping making no longer useful antibodies from previous infections.
      I remember it was discussed here before and the estimate was like Blue Whale in size.
      As long as the immune memory maintained, any future reinfection should be supressed quickly.
      For many serious illnesses we are inoculated only once and only in some cases booster is used years later.
      Flu is of course a different story due to its peculiarity of antigens shuffling.
      This is just a hunch but I think we will not really need yearly boosters for Sars-Cov-2

  3. Tim says:

    Um, did anyone else first keep reading it as “molecular glutes”, and wonder about the metaphor?

  4. sgcox says:

    With molecular glues, I am still struggling a bit to understand how tri-molecular complex is formed in real life if affinity of “glue” to either protein is weak. Complexes are not formed by 3 molecules suddenly bumping into each other. At least one binary intermediate should have a decent life time to give a chance for a third party to cement it. Surely some binary Kd must be well under uM for it to happen at physiological concentrations.

    1. GK Adam says:

      Glues do show detectable affinity, 10s to 100s nM depending on case, to one partner and it is the thus formed biomolecular complex that presents a new interaction surface suitable for the neosubstrate. This is well documented now across all known examples, such as lenalidomide, CC-885, CR8 and so on

      1. sgcox says:

        In case of CRBN ligands, yes.
        I had indisulam in mind here.
        according to ITC on Fig.1, its affinity to individual partners are 50 uM at best.

    2. sgcox says:

      More importantly, how can it happen in open systems.
      Yes, if you mix 3 components which have low binary affinity but very cooperative avidity in Eppendorf an let it sit for a while in the fridge, nice complexes will eventually form according to the laws of Thermodynamics. But in vivo…

  5. h3k27ac says:

    I prefer the biologics MOA to a degrader/degradomer. Very difficult to degrade proteins that form insoluble aggregates with small molecules.

  6. sgcox says:

    I understand the current obsession with PROTAC but how can any serious discussion about molecular glues not to mention drugs which already helped millions of people ?
    Like Tacrolimus, Ciclosporin, Rapamycin, etc. 🙂

    1. Sam says:

      Exactly, many people like to bitch about Stuart Schreiber, but they should remember his seminal work with FK506, Rapamycin and Cyclosporin in the 90s showing their effects as “Chemical inducers of dimerization”

  7. CB says:

    “The idea doesn’t stop there. You can potentially bring *any* two proteins together if you have molecules targeting each partner” A nice example was already shown in 1995: two different protein binding domains linked together with an optimized spacer length inspired by molecular modelling. Such synthetic molecules entered the clinic:

  8. Paul says:

    I love how I spent the first 10 years of my career being publicly humiliated for daring to suggest making a molecule outside property space and then PROTAC comes around and it’s all cool because “catalytic”.

    1. John Wayne says:

      Sounds like we worked at the same sort of place. It was better to be wrong with the majority than do something different that might work.

    2. odonovdh says:

      Yep, it’s a nice side-benefit of the PROTAC hype machine that most companies are beginning to look more seriously at how to create drugs beyond the rule-of-five.

      1. tommysdad says:

        That started long before PROTACs; just look at venetoclax. Bottom line is simple: if your molecule has properties sufficient for the intended route of administration, you have the data to counter any “groupthink”. But medicinal chemists have been very good at chasing potency at the expense of good drug-like properties for a long, long time. The point is to think like a drug-hunter from the very early stages of the project, and not just like a chemist.

  9. He PROTAC says:

    For those who are interested in the topic, Dana Farber has been having an excellent biweekly Webinar series since about Sep 2020.

    1. sgcox says:

      Second to that.
      Dana-Farber keeps recording of most presentations but unfortunately not all.
      I would really like to re-watch excellent Nicolas Thomä talk.

  10. bifictional says:

    Also, the kinase recruited to protein has also been demonstrated!

  11. Kaleberg says:

    Thanks for a great article. I thought I had just read something about this, but it was about bispecific antibodies, not bifunctional molecules. I’m guessing there’s a big bi-bandwagon out there.

    1. Jerry Liu says:

      Could you give a link to this bispecific antibody paper? thanks

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.

This site uses Akismet to reduce spam. Learn how your comment data is processed.