There are increasing numbers of reports of new TPD agents using tried and tested mechanisms (eg cereblon or VHL E3 ligases) but also a plethora of new degrading approaches which recruit other E3 ligases or "effector" proteins. How can you tell which have the potential to be developed towards the clinic and which others may struggle? It’s not an easy question but I’ll try to give some pointers here by way of a simple beginners guide based on 3 tests inspired by a useful article published recently by Ingo Hartung from Merck and colleagues. 

Many molecules, including bifunctional, PROTAC-like agents as well as more traditional small molecules (glues, monovalent degraders etc) will appear to lower levels of a given protein in a cellular environment – if you take the time to look for it. Some of these molecules work by well-defined, optimisable mechanisms and can give exquisitely selective and potent effects suitable for clinical development – other molecules work through shadier routes and may be doing many other things which will greatly complicate attempts to develop or optimise them into drugs. But, at an early stage, how do you tell the difference?

A few simple observations and controls will often quickly give a good sense of what is going on, especially with bifunctional, PROTAC-type agents. I don’t propose a long check list of studies with hard cutoffs as assessing the quality of a molecule is a judgement-based balance of many types of data. I also realise many exciting degrader discoveries may come from groups who may not have resource to complete an exhaustive dataset so instead I highlight certain, “sentinel” data which often suggests the overall quality of the agent in 3 main areas:

1 – Activity: Firstly, most degrader data should be reported as cellular DC50 (drug conc which gives 50% degradation of Protein of Interest [PoI]) and Dmax (maximum amount of protein degraded) to directly show a drop in cellular protein levels which can subsequently be correlated with whatever cellular functional efficacy endpoint is desired. There are now so many examples of highly potent degraders that for a new report (especially using well precedented E3 ligases), a good dataset should ideally expect to include one or more analogs which show DC50 <1uM and Dmax >80% (though for some challenging targets, Dmax values 50-80% may also be very valuable especially where full protein knockdown may not be required to give a functional benefit). 

That is not to say that compounds weaker than this are not useful, simply that the potential for additional, more complex mechanisms contributing to the degradation observed increases as potencies move from the nM range to the uM range. 

The cell type and assay used are also important. Though it may be necessary to use overexpressing cells for technical reasons of assay sensitivity in some initial cases, degradation should always be confirmed in cells with endogenous levels of PoI and ligase/effector (ideally in more than one type) with no overexpression to get a sense of what you may see in vivo or in people. If using Western blot or other antibody-based approaches to quantify protein, please always validate the specificity of the reagents using eg siRNA or CRISPR line controls (and don’t always trust the image with a single band in the reagent’s accompanying datasheet…).

Researchers should also take the time to generate full DC50 curves for key example molecules using at least 8-10 concentrations using half log dilutions. The dose response should be smooth – published examples using sketchy Western blots with little to no degradation at say 10nM & 100nM but then 90% degradation at 1uM don't give a useful sense of the dose response and may be indicative of the onset of an off-mechanism effect over a threshold drug concentration which is often not easy to optimise or develop. Using many concentrations over wide range also allows any Hook effect to be easily seen. Seeing a Hook effect is actually good mechanistic evidence of a ternary complex-mediated effect where you would expect to see one (though likely will not be seen with molecular glues of course). NB Many people seem to be worried about the Hook effect and what it might mean to in vivo dose responses and bell-shaped efficacy profiles. In reality, it usually just results in slower degradation (which can usually be modelled anyway) at supra-therapeutic doses and doesn't cause insurmountable issues.

As degradation is a kinetic process, time to Dmax can also be informative. Intracellular degradation via the proteasome is usually fast so good degradation should be seen within say 2-4h (NB some non-proteasomal mechanisms may be slower). If it requires 24h to see any degradation, this may still be genuine and indicative of an unoptimized molecule, but it may also suggest indirect effects, involving transcriptional suppression or feedback or even the onset of cellular toxicity or stress. When cells are unhappy, they may not fall over and die in the way you measure in a cytotox assay but they may partially downregulate translation as a survival mechanism. To minimise the risk of this, always use a cytotoxicity (or preferably cell health) assay and show that there is no frank toxicity in cells at concentrations at least 10 fold over the DC50.

2 – Controls The right controls are also very instructive for bifunctional agents. Always make sure the PoI binder alone and the E3 ligase (or “effector” protein) binder are characterised to see if they induce any degradation alone – sometimes these molecules also degrade  (which may be interesting to understand) confounding interpretation of data using the designed degrader. Other controls including bifunctional analogs with inactivated ligase/effector binder or PoI binder (eg a stereoisomer or with key binding interaction removed) and competition of designed degrader with free PoI ligand and E3 ligase ligand are also important and should abrogate activity. Addition of proteasome inhibitors, NEDDylation inhibitors or other mechanistic probes at appropriate active, non-toxic concentrations (often 1-10uM) can also infer involvement of the proteasome and cullin-based ligases etc respectively. If these agents fail to block the observed apparent degradation, there may be more going on than meets the eye. It's only a couple of extra Westerns to generate this dataset so there's really no excuse for not.doing this.

3 – Selectivity Finally, some indication of selectivity of degradation is really instructive to help interpret cellular functional data. Expression proteomics (TMT or other, say at 6h using 10x conc. over DC50) is increasingly accessible and should be used to give a sense of specificity of effect. Not all unoptimised molecules need show the exquisite selectivity of degrading just a single protein out of ~9000 quantified and a few off-targets degraded may be expected but if dozens to hundreds of proteins are seen going down (or up), it suggests a mechanism which may be more difficult to understand and optimise. If your degrader requires 24h to show degradation and you need to run proteomics at this time point, you may get a much more complex pattern with direct effects as well as feedback and transcriptional effects greatly complicating analysis – another reason why fast degraders often more easily show compelling profiles (see above). Proteomics is not infallible of course as many proteins are not sufficiently detectable but it gives a good sense of overall effect. If proteomics is not accessible to you, please try to assess changes in levels of as many other proteins as you easily can. Reports of "potent degradation" of a target (especially kinases) are of little use unless there is a clearer idea of what else may be degraded (or indeed just inhibited - degraders based on PoI inhibitors are still themselves inhibitors so functional data should always be interpreted with this in mind).

The guide: Overall, my simple degrader quality checklist relies on assessing only 3 areas in order to identify degraders with a higher chance of being developed as therapeutics working through a well-defined mechanism:

1. Activity: A smooth DC50 curve in endogenous, non-overexpressing cells with DC50 <1uM and Dmax>80% measured at 4h or less, preferably with some evidence of a Hook effect and no cytotoxicity within at least 10 fold of the DC50.
2. Controls: All controls (PoI & E3 binder alone; inactivated analogs; addition of proteasome & NEDDylation inhibitors or other) have been run and shown to abrogate cellular degradation completely.
3. Selectivity: Initial characterisation of selectivity at the proteome level shows only a modest number of other proteins affected at 10x DC50 without more general changes of unrelated proteins

There are many, many other factors to be taken into account of course including a range of physicochemical factors, SAR observations, more detailed mechanistic characterisation etc as nicely summarised in Ingo’s paper - my outline is just a beginners guide to help highlight the more robust data sets and areas of the data which may need to be understood in more detail. Note these guides are absolutely not “rules” but should be used to help you form an opinion if the agents are working cleanly through their proposed mechanism and to guide further questions and experiments.

Once you have some level of confidence that the fundamental degradation is sound, other factors need to be considered as you assess potential application to therapeutic discovery: correlating levels of protein knockdown to functional cellular phenotypes and in vivo responses, depth and duration of knockdown needed, safety, stability & metabolism etc etc but those are topics for another day. Ingo’s review touches on a few of these and also gives some valuable information on assessing covalent probes – well worth a read.

I hope my quick guide is useful for TPD researchers (and journal manuscript reviewers…) and that the studies highlighted are not too onerous but instead are a simple way for everyone to generate a simple baseline set of data to either highlight that your molecules are well behaved or else may require a little more thought...