Spending a week at the 7th TPD & Induced-Proximity meeting in Boston (see my blog summaries of day 1, day 2, day 3 & day 4) gave me the chance to hear about some great science and catch up with many friends from pharma & biotech to get a sense of where we, as a TPD community, have gotten to now that we’re nearly 10 years in to the latest gold rush of TPD-based drug discovery. It’s now much clearer which areas we’re getting rather good at (and there’s a lot of these) and which challenges are occupying the brightest minds in the field where I expect to see breakthroughs in the coming months and years.

Firstly, over 20 companies have now successfully progressed a plethora of TPD agents, both bifunctional PROTAC-like agents and smaller, glue-like drugs into the clinic. Initial signs of efficacy are encouraging with the first approval possible next year but, more importantly, no strong evidence of class-related issues or toxicities has apparently emerged. This is a really important observation for any new modality, especially a more complex one where multiple proteins must be engaged to deliver efficacy and multiple theoretical routes for things to go wrong could be envisaged. Now, clinical safety must always continue to be monitored of course, and it’s possible we may encounter target-based effects (as with all drugs) in some cases, but the lack of more general concerns gives us confidence to keep pushing degrader approaches into more challenging targets and more, non-oncology therapy areas. Sure, many approaches still rely on cereblon and there are outstanding unknowns regarding some aspects of cereblon-based toxicology given the long history in the area dating back to the 1950s but it’s positive to see the agencies and industry consortia are now engaged to help plot a useful path through these.

Whilst nearly a quarter of the ~90 TPD-focused biotechs have now successfully brought TPD agents onto clinical studies, it’s also great to see some of the pharma now translating their deep investments into vibrant pipelines. BMS (incl. Celgene) for example may have up to 10 TPD agents in clinical development – I’m sure we’ll see others catching up soon.

I’ve written recently about TPD target selection (here) and E3 ligase selection (here) but current TPD approaches have already enabled many, many targets with the scope set to broaden further. PROTACs require a target-binding ligand of course (more on that later) but when a suitable ligand is in hand, a large proportion of intracellular targets appear to be PROTAC-able with varying amounts of optimisation needed. Glues could access many targets without high affinity small molecule binding sites and break open huge swathes of classically undruggable deep blue ocean target space. Just focusing on CRBN, Monte Rosa and others have identified well in excess of 1000 targets with the “classical” G loop, beta-hairpin-like motif whilst Phoremost also used their loop-insertion, GlueSeeker approach to probe other targets which could be within reach of CRBN-based glues but which may lack these G loop-like regions, potentially expanding the CRBN target reach further.

So, CRBN may be able to degrade a wide range of substrates via glue (or PROTAC) mechanisms but it’s still likely that many targets will not work well or at all so other effectors are needed. There are increasing numbers of apparent glues, both degrading and non-degrading, being identified, often by directed proximity-based screening or phenotypic screens but the high hanging fruit here is true “any vs any” glue discovery. This may require large training sets to understand the tractable, global “glu-ome” and is an area many groups are now actively tackling. A good challenge for ML to address and certainly an area I expect more exciting disclosure in the near future.

Back with PROTACs, these still have, in many ways, the most tractable optimisation path but, as we get more and more interested in transcription factors and other less well-understood target exotica, ligand finding quickly becomes the main challenge. If you really care about a target, spending time on novel ligand ID is time well spent. Use DNA-encoded libraries (DEL), use fragment screening, use HTS, use virtual screening - or use combinations of all of these. That may be what you need. Simply running a DEL screen and expecting nM hits for intractable targets to fall in your lap every time is simply not realistic, some hard yards may be needed but finding and optimising ligands using a binding assay is med chem bread and butter of course so there’s a good chance of success if you’re patient and thoughtful.

We’ve also seen great strides made in the optimisation of PROTACs and glues. Brute force high throughput chemistry (using eg Direct-to-biology methods) allows many design hypotheses to be tested quickly in cellular degradation assays whilst the increasing availability of cryo EM and x-ray ternary complex structures now also gives more rational design options. With bifunctionals, our growing understanding of the relationship between linker design/conformation and degradation efficiency/pharmacokinetics/other druglike properties means we can focus optimisation in the right areas to shorten the overall process though it’s likely we’ll still encounter cases where small SAR changes have big effects so we need to be alert to these.

Bifunctionals were once thought to have very undruglike properties however, once you have a potent degrading PROTAC, optimal ways to deliver it in vivo and in the clinic are now becoming better understood. For many CRBN PROTACs, oral availability is now routine (see my thoughts here). Oral delivery of VHL PROTACs in certainly not yet routine though I suspect some groups are quietly making good progress here…so keep watching. There will of course be some highly desirable targets where the size or complexity of the target-binding ligand makes oral delivery (CRBN, VHL or anything else) impractical so in these cases (if the glue route isn’t open to you), non-oral, long-acting injectable formulations can still enable dosing regimens which will look a lot like clinically-successful biologics and so will be acceptable to patients. With the increasing willingness to use injectable GLP1 agonists and the availability of new autoinjector technologies etc it will be interesting to see the pace with which use of non-oral delivery routes continue to grow. I predict that some of the most exciting efficacy from degrading drugs will come from non-oral dosed agents.

CNS penetration also, whilst likely an additional order or magnitude more difficult than oral absorption for larger drugs is now slowly being untangled. Combinations of using (or avoiding) transporters along with careful conformational control is now making this area more attractive and I expect to see this grow further – there are clearly a number of highly attractive targets in this space to motivate innovation here.

Turning to refining the selectivity of degraders, proteome-wide selectivity of degradation (routinely assessed by proteomics on >8000 proteins) is also now a largely solvable problem, though in some cases you may need more med chem to reduce some pesky off-targets. Remember though, the selectivity of your degrader may be different in different cell types or when different metabolites are produced. Now, this is exactly the same as with any typical small molecule drug so a key consideration with TPD agents is deciding how much you need to know and how many “known unknowns” you’re comfortable with as you move to the clinic. Don’t forget, small molecule inhibitors usually have significantly more gaps in our knowledge of selectivity/mechanism but a clean GLP safety profile usually acts to bypass many of these before the clinic.

The field has largely moved away from the concept of tissue-restrictive E3 ligases (tried it – too hard – move on). I still hope we find something here, and many are still persisting as the reward could be high, but using degrader-antibody conjugates (DACs) is likely a better way to achieve higher selectivity if degrading your target broadly across all tissues is unlikely to end well. As well as targeting tumor-enriched markers, going after specific immune cell populations or others holds great promise as a way of getting cell or tissue-specific degradation. TPD agents, with their often very high potency, make ideal payloads for antibodies where, outside a few ultra-nasty toxins, non-catalytic small molecules just don’t give the bang for the buck needed in the ADC space. 

Induced-proximity can also take a leaf out of the ADC playbook when using bifunctionals (eg RIPTACs) which selectively elicit their effects in cells highly expressing tumor markers and other targets: intracellular ADCs without the antibody part - enabled by proximity.

Everything I’ve spoken about so far relates mainly to intracellular, proteasomal degradation and indeed the target scope there is large but efforts are also gathering pace to go after intramembranous or secreted proteins as well as aggregated proteins and other potentially pathogenic species or organelles. The learning curve on these approaches is a few years behind their proteasomal cousins (and some areas of the literature still have a Wild West feel to them) but many groups are now doing good work in this area (see eg here). Whether using small molecules, biologics or RNA-encoded delivery of intracellular proteins, there are many approaches being assessed, all of which could expand the target scope and risk profile of any TPD portfolio. We’ll need to see which ones of these show clinical success but it’s great to see induced-proximity concepts using many classes of agents beyond the core small molecules.

So, as a company, whether you’re an investor, a startup, biotech or big pharma, is the TPD field a place you should be investing in? Er, yes. The multiple benefits of TPD approaches over other modalities have been demonstrated extensively so you really can’t afford not to be in the space. How you invest will be up to you. For fast access to the clinic, there is still a lot of mileage in mining out the CRBN space with your favourite targets for glues and PROTACs. If any nagging doubts about using drugs with a family tree which includes thalidomide keep you awake at night, VHL-based agents can be pretty good as well (and other approaches start to come onto the radar if you are happy to embrace protein-based agents) – certainly don’t consider these alternatives to be the poor relations to CRBN – in time they may even be better. 

There is still much room for innovation outside of these areas – new E3s may yet prove their worth and solving the global glue discovery problem is the glittering prize here which could yield degraders or other useful, non-degrading pharmacology – but combining now well-validated concepts into DACs, RIPTACs and a long list of other “Shiny-new-TACs” will also give agents which will mediate hitherto inaccessible pharmacology. 

The only limitation is imagination.

The future is bright, the future is proximal.