Developments and challenges in degradation-targeting chimeras

Back in 2015, I blogged about the buzz in the chemical biology world over targeted protein degradation: the use of chimeric molecules that bind both your protein of interest (POI) and something capable of degrading that protein, and bring them into chemically induced proximity to cause degradation to occur. While the concept has been around for years [Sakamoto 2001], it garnered a new level of attention thanks to the discovery of the mechanism of action of thalidomide, whose phthalimide chemical group turns out to bind the E3 ubiquitin ligase cereblon (CRBN).

In the years since then, this area of research has exploded. It turns out there are loads of different proteins, complexes, and pathways that target things for degradation, and that are operative in various cellular compartments, and for which ligands either existed or are now being discovered. This week I sat down to try to update my understanding of the literature on this topic, and identified no fewer than nine different systems that have been reported — and I am by no means certain this list is complete:

Above I have summarized these studies in terms of the protein machinery responsible for degradation, the cellular compartment(s) in which it is active, a description of the chimeric molecule system needed to induce degradation, and references. I only included systems that rely solely on the exogenous addition of a single chimeric molecule — this list does not even touch on the numerous systems that rely on transfection of cells to express a special construct, such as the “Trim-Away” system using TRIM21 [Clift 2017], or the auxin-inducible degron (AID) system [Holland 2012].

My reason for interest in all this is, of course, the hope that some such system could be leveraged to degrade prion protein (PrP) in the brain. We have excellent evidence that lowering PrP will be therapeutically beneficial in prion disease, but, like many of the targets people have worked on in these “degrader” studies cited above, just binding the protein might not be enough — ideally, the protein needs to not be present.

With that framing and motivation, where does the recent progress in development of “degrader” systems put us?

Whereas the first reported “degrader” systems were for cytosolic proteins, it is great news that we are finally seeing more diversity of cellular compartments: three systems cited above are operative in the ER or extracellular space, including one I blogged about previously [Gustafson 2015] as well as the recently reported LYTAC [Banik 2019] and ENDTAC [Nalawansha 2019] systems. It is worth discussing these two new systems in a bit more detail.

The LYTAC system [Banik 2019], excitingly, was shown to work not only on secreted proteins, but also on membrane-bound proteins such as EGFR. It relies on hijacking an endogenous system for degradation of glycosylated proteins. CI-M6PR’s native function is to bind mannose-6-phosphate (M6P), which is a glycan mark that cells use to target proteins for degradation. Prior work in delivery of enzyme replacement therapies had established that conjugation of a cargo to M6P can allow internalization by this receptor, but it hadn’t been used for targeted degradation before. Certain cargo internalized by CI-M6PR, such as PD-L1 (CD274) can often return to the cell surface through chaperone-mediated recycling, but they showed that this system actually overrides such recycling and can achieve up to 50% degradation of PD-L1. The so-called LYTAC (lysosome-targeting chimera) molecules are antibodies to a target protein, covalently attached to a 20- or 90-mer of M6P. Thus, the resulting molecule must be quite large, well over 100 kDa.

In contrast, the ENDTAC system [Nalawansha 2019] relies on a (relatively) small molecule. They used the recently developed small molecule CXCR7 ligands VUF11403 and VUF11207 [Wijtmans 2012], and added a linker and then a chloroalkane moiety. Chloroalkane is a ligand for HaloTag7, and they expressed a GFP-HaloTag7 fusion protein as an artificial proof-of-concept system. No endogenous proteins were targeted in this paper, but they demonstrated clear uptake and reduction of the secreted GFP fusion protein at concentrations of ENDTAC as low as 500 nM. With the CXCR7 ligand, linker, and POI ligand, the overall ENDTAC molecule is not tiny, but is far smaller than the antibody or peptide-based systems listed up top. Update 2020-01-29: the ENDTAC paper has now been retracted, as apparently the original experiments did not replicate.

All of these studies are exciting as proofs of principle, and I am excited to see the arsenal of degradation systems continue to grow. That said, there are several reasons why applying such a system to PrP is still a relatively distant dream.

1. Expression. In order to degrade your protein of interest in the disease-relevant tissue, the degradation-mediating protein or complex has to be expressed and active in that tissue. This can’t be taken for granted. The androgen receptor AR, used in the endoplasmic reticulum-based targeting system mentioned above [Gustafson 2015] is expresed only in the testis. IGF2R and ACKR3, the degraders used in the LYTAC and ENDTAC systems reespectively, both have relatively low brain expression.
2. Delivery. Because we need to reach the brain (the whole brain) to treat prion disease, drug delivery may be our biggest challenge. Many of the systems listed above rely on antibodies, peptides, and/or sugar polymers, pushing their overall size into the kilodaltons or even hundreds of kilodaltons. That means these exact systems, at least as currently described, might not be applicable to brain diseases. While there are of course those who argue that antibodies do cross the blood brain barrier in some appreciable quantity, they appear not do so in sufficient abundance to treat prion disease [White 2003]. And I have yet to see any good data regarding the distribution across a large brain of antibodies or proteins delivered intrathecally, into the cerebrospinal fluid. Therefore, I expect that any PrP-targeting degrader system would have to be a small molecule in order to be useful. Which brings us to the final two problems.
3. Need for POI binder. Finding small molecule ligands of PrP is really hard! No validated, monovalent binders currently exist, and as our research scientist Andrew Reidenbach presented at Prion2019, our NMR fragment screening efforts against PrP at Broad have met with a remarkably low hit rate. We are not giving up the search for PrP binders, but it’s worth acknowledging that it does appear to be a hard problem.
4. How small is small. Meanwhile, it is an open question how “small” a degradation-targeting chimera could ever be. Take for example the phthalimide-based PROTAC systems [Winter 2015, Lu 2015, Nabet 2018], which have perhaps gotten the most attention of any system here so far. This is kind of a best-case scenario, because phthalimide itself is tiny — just 274 Da, but by the time you add in a linker, you are up to 604 Da, and then you still need your POI binder. Thus, for example, the PROTAC dTAG-48 to degrade FKBP12 ends up being 1,208 Da [Nabet 2018]. That’s four times the size of most FDA-approved drugs for the central nervous system (CNS), which cluster around 300 Da, and twice the size of the largest such drug (about 600 Da). It is an open question whether anyone will manage to make a degradation-targeting chimera molecule small enough to get into the brain, and for that matter, we don’t even yet have any evidence as to whether the pharmacokinetics of these chimeric molecules can be successfully tuned even for robust delivery to peripheral tissues of interest.

Given all these limitations, a degradation-targeting chimera is not our fastest bet for developing a drug to lower PrP — thank heavens antisense oligonucleotides appear poised to get there much, much faster. Still, this is an exciting area of research and I intend to continue to keep an eye on this and think about how they could work for our disease.