I’ve been busy over the summer so my blog posting has suffered but now the weather is changing, I’ll do my best to get back to it!

When TPD exploded onto the drug discovery scene nearly 10 years ago it was accompanied with the usual hyperbole which often accompany shiny new things, namely that soon no drug target would remain undruggable as TPD would soon be degrading all those highly validated but hard to drug targets with the resultant tsunami of exciting clinical candidates. So, now we know more, how’s that been going?

Asher Mullard has just put out a great piece in Nature Reviews Drug Discovery which gathers the views of many leaders across the TPD space (even me) on how they see the TPD-able target landscape evolving and it’s well worth a read.

My summary of the article: good progress in novel targets in some areas and still great potential to do things that inhibitors, antibodies etc can’t, but now need to bring together all the learnings to really pick off the high hanging fruit.

The main flavours of TPD, certainly those that have produced the more advanced candidates, are bifunctional degraders (PROTACs etc) and monofunctionals (often called glues). While glues and PROTACs can be thought of on a mechanistic continuum relying on (co-operative) ternary complexes to induce ubiquitylation, their respective target spaces have certainly been different.

The prototypical glues are the IMiDs (lenalidomide, pomalidomide etc), discovered many years ago, which use the E3 ligase cereblon to degrade a family of structurally-related targets (primarily IKZF1/3). All of these targets are certainly undruggable when it comes to any definition involving “can you get a small molecule inhibitor to bind to it?” but they’re really very tractable to cereblon-based glues. Based on known IMiD substrates, advanced degraders targeting the related targets GSPT1, IZKF2, CK1a have also now been found which all contain the key G-loop degron in a zinc finger. The glue bioinformaticians at Monte Rosa have now shared their assessment of just how far you can persuade cereblon to reach out into other neosubstrates to degrade them and the answer seems to be – pretty far. They propose over 1600 “glueable” targets which have suitable structural regions (often based on b-hairpin or other G-loops) and we’ve already seen cereblon glues to novel targets like NEK7, VAV1 & WIZ1 starting to progress into development. So glues seem to be able to tackle a set of targets which may be really quite novel and a little orthogonal to those where we can find good small molecule inhibitors but finding glues can still require a lot of screening and a little luck. Many groups are now training machine learning on this tough problem which is a sensible thing to do - as long as they have enough useful training data to give the ML something to digest (which is not always the case).

Let’s now turn to PROTACs – these can in principle be designed more rationally. It all makes sense on paper: use your favourite assay to find a binder to a well validated, and possibly undruggable target (there are many assays from DNA-encoded libraries to biophysical assays to fragment screens) then simply append an E3 ligase binder (probably recruiting cereblon or VHL) via a suitable linker and you’ll have a degrader to take to the clinic, probably after some amount of chemistry optimisation.

The first wave of PROTACs however didn’t really push the undruggable envelope: AR, ER, BTK, IRAK4 etc. Whilst maybe not considered highly novel targets, they all have sound rationales why a degrader will outperform an inhibitor and indeed many of these projects continue to generate very meaningful clinical efficacy and safety data and are likely to reach approvals in the coming years. Let’s remember though that, in most cases, these degrader drugs will need to compete with the respective target inhibitors in the clinic, all of which are dosed orally. As prevailing wisdom in many companies may say that you need an oral drug to compete with an existing oral standard of care, this has forced TPD drug developers to learn how to make their PROTACs orally available. And they’ve been rather good at doing that, consigning Lipinski’s Rule of 5 to a distant memory in the process. Making oral PROTACs is not always easy but we know the right design areas to play in as Kymera and Arvinas have published on. However, there will be many targets where making orally available PROTACs will be impossible because the necessary target-binding ligands may be just too big or polar, but is that really so much of a problem?

There are increasing reports of injectable PROTACs which have dosing regimens which look a lot like injectable biologics (ie q2w, q3w or even less frequent), and we know that is a clinically very acceptable way to go, especially given the often shockingly low patient compliance with once daily oral therapies. Foghorn & Arvinas have shown preclinical in vivo target knockdown lasting at least two weeks after a single injectable dose and GSK also showed knockdown for up to 4-6 weeks after one dose of a PROTAC.

So back to target novelty. There will undoubtedly be some “undruggable” targets where you can find small target-binding ligands suitable to give oral PROTACs (STAT6 may be a good example now entering the clinic) but many high value targets will not yield suitable high affinity, low molecular weight ligands. They may however yield to larger binders (eg think of the BCLxL or MCL1 binders or other monsters) but, importantly, these can still be incorporated to make potent PROTACs. Sure, they’re unlikely to be orally available but if you can use some neat formulations to get a once every two week injectable regimen, you’ll likely have a great candidate, especially if your degrader of an undruggable target offers transformational efficacy or opens up a patient segment with no good standard of care.

So I suggest, if there are targets which have such a strong validation or genetic rationale that you can be confident that you’re likely to get excellent efficacy with a clinical degrader, you’ll just need to use whatever smart screening & med chem approaches (fragment growing/linking, cyclic peptides, other affinity approaches) you can to get a good ligand binding at say 1uM or better. It may have mw>1000 or have a structure that looks less than beautiful but make a PROTAC, let the catalytic mode of action work its magic and you’ll likely to have a cell active degrader with DC50 <10nM with all the other advantages of degrader approaches. 

How big can your PROTAC degraders get before they simply can’t get into cells any more? Well, pretty big it seems. Alessio Ciulli and colleagues have been playing with some rather exotic degraders which are trifunctional or even tetrafunctional. They show some very interesting cellular efficacy down to 1nM but hang on, some of these are compounds with mw>2000. Surely these beasts must have poor cell penetration? Well, enough gets in to drive nM effects so that’s good enough for me. 

If you take this approach on your favourite undruggable you might also end up with a mw 2000 PROTAC but, optimise the pharmacokinetics a little then use an injectable formulation and you may just have an exciting drug candidate. Ok, so the drug will need to get to the site of action in vivo you say, but turns out that big PROTACs seem to be rather good at that as well.

As I concluded Asher’s article saying: “If you want to talk about novel targets, I wouldn't count out PROTACs yet.” 

Indeed, we now have all the pieces in place to really mount a degrader-based assault on these pesky undruggable targets, and TPD approaches remain uniquely positioned to do this.