This research service has spent a large portion of time writing about novel small molecule efforts. The most exciting of these efforts aim to reach previously undruggable proteins, with RevMed’s daraxonrasib serving as the paradigmatic example. Some other efforts focus on targets that, while currently druggable, may be hampered by side-effect issues. Relay’s lead candidate is a good instance of this: side-effects of current PIK3CA inhibitors can be severe enough that they end up playing second fiddle to other metastatic breast cancer options.
Fibrolamellar Carcinoma (FLC) is an example of a disease whose driving protein is in the undruggable column. The vast majority of cases are caused by a fusion of the DNAJB1 and PRKACA genes. Both genes are located on chromosome 19, and a 400kb deletion causes them to fuse together. This DNAJ-PKAc fusion protein is not in principle undruggable: PRKACA codes for a portion of protein kinase A, and kinase inhibitors are frequently used to fight cancer. Sadly, protein kinase A is so essential for cellular functioning that hitting its wild-type form is an unacceptable side-effect.1
Surgery is the standard treatment for FLC, but there’s frustratingly little beyond that. Chemotherapy is fairly ineffective, as is standard immunotherapy (FLC tumors tend to have a low mutational burden, be mismatch repair proficient, and have low levels of PDL1 expression. In other words, the exact opposite profile of a tumor that would respond well to Keytruda). 5-year survival rates are around ~40%, a sobering statistic when patients are typically diagnosed as adolescents or young adults.
Fortunately, there are more novel therapeutic options that could show promise in treating FLC:
1) Degraders (whether a PROTAC or molecular glue) that target the fusion protein directly. The good news with FLC is the fusion protein is the cancer cause and contributes to the cancer’s continued growth.
2) Degraders that target other proteins implicated in FLC growth.
3) Cancer vaccines.
I won’t spend time on (3), but there’s a paper by a team out of Johns Hopkins on building a cancer vaccine to target the DNAJ-PKAc fusion. Unlike what Moderna’s working on for melanoma, such a vaccine would not need to be customized for each patient’s tumor, and so could be administered soon after diagnosis.
Degraders as a class are clearly of interest in FLC because, in principle, they offer a way to get around the selectivity issues posed by traditional kinase inhibitors. Targeting the kinase active site leads to undesirable off-target effects, so going via an alternative route (whether a non-functional binding pocket or a degron) should be a way to circumvent these issues. The most straightforward application of degraders here is to figure out a way to uniquely target the fusion DNAJ-PKAc over wild-type PRKACA. The Fibrolamellar Cancer Foundation is funding research on this. My assumption, given the Principal Investigator’s background, is she was able to use chemoproteomics to find a potentially compelling novel binding site.
Some of the other current research/clinical work on FLC degraders, whether directly hitting the fusion protein or not, is also very interesting. This work elucidates benefits of PROTACs/molecular glues that aren’t necessarily leveraged in the more well-known degrader clinical trial candidates, and so are instructive for FLC and degraders more broadly.
The first example of this is Dialectic Therapeutics’ DT2216, in clinical trials for FLC (in combination with irinotecan) and platinum resistant ovarian cancer (in combination with paclitaxel). The FLC trial is currently suspended due to unacceptable toxicity, but nonetheless has some important takeaways (and will hopefully be resumed). The idea for this trial came out of work done by Dr. Sanford Simon at Rockefeller University. Dr. Simon screened FLC tumor lines against >5000 compounds, looking for evidence of some ability to kill off the cancer cells. A few compounds showed promise: irinotecan, a topoisomerase I inhibitor, and some inhibitors of Bcl-xL. Irinotecan was a surprising find, as topoisomerase I inhibitors tend to be helpful in cases where cancer is fast growing; FLC is on the opposite end of that spectrum. Bcl-xL is a protein that helps prevent cell death, and so is one you don’t want to be especially active in cancers. Unfortunately, inhibiting Bcl-xL leads to a rate of cell death that’s far too high in megakaryocytes (which produce platelets). This in turn results in thrombocytopenia, a low blood platelet count that can lead to prolonged bleeding.
It is here that DT2216 gets really interesting. Recall that PROTACs must do two things to enable protein ubiquitination and subsequent degradation:
1) Bind to the protein of interest
2) Bind to a portion of an E3 ligase (today this is most often VHL or cereblon)
From there, the POI is tagged for degradation and broken down by the proteasome. PROTACs don’t work if they can’t do both of those steps. Without the necessary E3 ligase, there’s no way to degrade the protein.
DT2216 acts by binding to Bcl-xL and VHL. VHL isn’t expressed at high levels in platelets, and so the drug should be able to get around the dose-limiting thrombocytopenia risk that typically comes with Bcl-xL inhibition. This speaks to a key hope for PROTACs/MGDs in the future: today people like cereblon and VHL because they’re, for the most part, widely expressed across tissues. Degrader companies predominantly pitch the modality as a way to hit conventionally undruggable targets, and much less as a tool to get increasingly precise with where in the body the drug is active. Monte Rosa’s MRT-2359 is intriguing because the GSPT1 protein doesn’t have a conventional binding pocket, not because the therapeutic is somehow leveraging an E3 ligase only expressed in prostate cancer cells. As our knowledge of E3 ligases expands, however, the potential for tissue selectivity will become an increasingly important aspect of degraders. Kymera may be presently noteworthy because KT-621 is able to target a transcription factor, but this the future the company is eventually building toward, as discussed in its S1 filing:
“Leveraging the knowledge generated by our E3 Ligase Whole-Body Atlas, we are building a proprietary toolbox of differentiated E3 ligase binders for the development of next-generation targets and disease-specific degraders. We are focused on building an expanded library of E3 ligases and novel ligands with differentiated expression profiles in order to selectively pair them with targets in specific tissues, cell types and subcellular compartments. We believe that this approach to degrader design will lead to more selective and potent target degradation in disease contexts, while avoiding target degradation in tissues associated with known toxicities, which we believe will lead to a substantially improved overall safety profile for our degrader molecules. For example, as shown in the figure below, using the E3 Ligase Whole-Body Atlas we identified a novel bone marrow sparing E3 ligase, which has been successfully liganded and used to degrade a target protein.”
The second degrader in the FLC world (albeit preclinically) that’s worth discussing is the bioPROTAC discussed in this paper. BioPROTACs operate a little differently than the small molecule equivalent, although the principle is the same. Rather than two small molecules connected by a linker, they consist of a fusion protein: one half of the fusion binds to the protein of interest, and the other half is a portion of an E3 ligase. BioPROTACs are very early (there aren’t any that have entered clinical trials), but they’re primarily compelling because they expand the library of E3 ligases we can use. Small molecule PROTACs predominantly leverage VHL and Cereblon because we know how to bind to them. Many E3 ligases are not traditionally druggable, so figuring out how to hit them with a small molecule is not an insignificant hurdle! With BioPROTACs that’s not an issue. A portion of the E3 ligase is already part of the fusion protein, so there’s no need to bind to an endogenous one once the fusion has entered the cell.
The second benefit of BioPROTACs is you don’t have to spend time worrying about ternary complex formation. With traditional PROTACs, one has to optimize to some degree for binding affinity to the protein of interest/E3 ligase and then to a great degree on the strength of ternary complex formation between the PROTAC, protein of interest, and E3 ligase. With bioPROTACs, one only has to worry about binding affinity between the first half of the fusion protein and the protein of interest: the E3 ligase is already part of the fusion, so the odds of ubiquitination (as long as binding affinity is sufficient) are much higher.
In the case of the bioPROTAC for FLC, researchers fused mRNA coding for a peptide kinase inhibitor with SPOP, an E3 ligase active in the nucleus. Like Dialectic’s candidate, this construct seemed to have a problem, as it also binds to wild-type PRKACA, which should make things unworkable! There’s a crucial structural difference, however, between wild-type PRKACA and DNAJB1::PRKACA. Recall that the role of an E3 ligase is to facilitate ubiquitination of the protein of interest, which serves as a flashing signal for the proteasome to come along and degrade it. Typically, these ubiquitins are added to lysines on a protein. It so happens that the DNAJB1 fragment contains a number of accessible lysines, which in turn makes the DNAJB1::PRKACA fusion a better candidate for degradation. This is a pretty fascinating application of degraders, whereby selectivity is achieved based on how compelling a protein’s structure is to the proteasome.
There’s a natural question that arises with both DT2216 and the preclinical bioPROTAC above. While degradation may be selective in both cases, inhibition is not. This should cause problems: Bcl-xL inhibitors cause thrombocytopenia because they inhibit Bcl-xL; traditional kinase inhibitors haven’t worked in FLC because they inhibit wild-type PRKACA. Consequently, what good is a therapeutic candidate that degrades selectively but still inhibits indiscriminately?
This brings us back to a key degrader advantage that companies today spend plenty of time harping on: they act catalytically, and so go degrade one protein, then another, and then another. In other words, one degrader molecule can degrade multiple copies of a protein. This differs from traditional inhibitors: if the molecule dissociates from one mutant EGFR to go inhibit another mutant EGFR, that first protein is now back to being active (the scientific term for this is degraders have a ‘sub-stoichiometric’ mode of action). The upshot of degraders’ sub-stoichiometric capabilities is dosing can be lower than traditional drugs. As a result, while there is presumably still inhibition of Bcl-xL in megakaryocytes (in the case of DT2216) and inhibition of wild-type PRKACA (in the case of the bioPROTAC), the inhibition is at a low enough level that it shouldn’t cause major issues.
Acknowledgements: Thanks to Mason Grant for bringing my attention to FLC and the unmet need there.
Blueprint Medicines (acquired by Sanofi for >9bn) was working on a candidate for FLC but I assume was not able to sufficient avoid wild-type PRKACA.