Rational design of molecular glues - the next frontier
By Jesse Chen
Jesse Chen is an entrepreneur-in-residence at RA Capital. In this role he works closely with RA’s Venture Team to evaluate various drug discovery platforms and initiate novel therapeutics programs.
October 12, 2021
Targeted protein degradation (TPD) has advanced beyond intriguing chemistry to promising therapy over the past five years. With at least 15 programs in the clinic by the end of 2021 and more than 400 preclinical assets in earlier-stage pipelines, TPD is expanding the landscape of drug discovery.
As more established TPD companies continue to produce promising data in the clinic, newcomers have been greeted with sustained enthusiasm from the investment community. While TPD is “thriving in the spotlight”, it’s worth learning about the field’s historical advances and imagining its future possibilities.
Here I would like to revisit a few key intellectual and scientific breakthroughs and focus on molecular glues as an exciting new frontier in drug discovery.
Rewire the degradation pathway? “Yes,” says the virus
Proteins are constantly made and destroyed in cells to maintain a delicate biological balance, carefully regulating each protein’s functions. Cells remove many proteins by tagging them with a smaller protein called ubiquitin, which becomes a signal to direct the tagged protein for degradation in the proteasome, a cellular “recycling center.”
For years, the accepted notion was that as proteins aged, they would become unfolded and damaged and require recycling via ubiquitin tagging. We also assumed that was the proteasome’s sole job. But all that changed in the early 1980s.
Peter Howley, then a researcher at the National Cancer Institute, was intrigued by E6, a protein encoded by human papillomavirus. E6 did not fall into any known viral protein categories. As Howley and colleagues discovered, E6 protein is an adaptor that induces a novel interaction between two host proteins: one that they named E6-associated protein (E6-AP) and p53, a protein with many key regulatory functions, including viral surveillance.
E6-AP turned out to be a ubiquitin ligase, a protein that tags (“ligates”) other proteins with ubiquitin to facilitate their removal. The virus was outsmarting its human host by hijacking the cell’s own machinery, redirecting p53 straight to the recycler. E6 was inducing proximity between a ubiquitin ligase and the cell’s sentry, removing it before it could sound an alarm.
The significance of Howley’s work was two-fold: First, it uncovered a novel ubiquitin ligase (actually, a class of them) that is not directly involved in recycling aging proteins. In fact, E6-AP, among other ubiquitin ligases, and the proteasome play a variety of roles including signal transduction and cell cycle regulation. Second, the whole machinery can be rewired to degrade new proteins, the core premise of TPD.
In 1998, Howley and his colleagues proposed that proteins such as E6 can be fused to new protein adaptors to target other proteins they would like to destroy using the proteasome, which is the earliest documented concept of TPD. Although this elegant idea did not attract much attention back then, in 1999, two researchers at Proteinex, John Kenton and Steven Roberts, further expanded this concept.
Kenton and Roberts proposed that in principle, small molecules could be designed to mimic the actions of viral E6 protein. In the patent filing, they illustrated their concept by a dumbbell-shaped molecule: one end to grab a ubiquitin ligase and the other end to engage the target protein. But this TPD approach suffered from the lack of a suitable molecule that can grab the ubiquitin ligase.
Not for long, though. In 1998, Ray Deshaies, a prominent ubiquitin biologist from Caltech, met Craig Crews, a chemist from Yale, at a research conference. The pair had similar ideas about degrading proteins. Even better, they also knew how to grab the ubiquitin ligase.
In their landmark 2001 paper, Deshaies and Crews demonstrated for the first time the degradation of a protein with an engineered ubiquitin ligase ligand. Since then, especially over the past ten years, additional high-affinity ubiquitin ligase ligands have been discovered. This has further fueled advances in the “heterobifunctional” molecule approach, often called proteasome-targeted chimeras (PROTACs).
A great advantage of PROTACs is their ability to target proteins that until recently were considered undruggable because of their lack of an active enzymatic site to be inhibited by a small molecule.Several PROTAC compounds have entered clinical development and produced encouraging early readouts, led by Arvinas and Kymera.
Birth of “molecular glues”
While the E6 and E6-AP story unfolded in the early 1990s, another group of scientists was intrigued by a small molecule called auxin (indole-3-acetic acid). Auxin is found in all plants and regulates plant growth. Most researchers at that time thought auxin would behave like other plant hormones and interact with an unidentified cellular receptor, triggering downstream signaling pathways.
They were wrong. A surprising finding from Ning Zheng and others in 2007 showed that auxin is a small molecule adaptor involved in TPD. In Ning’s words, “By filling in a hydrophobic cavity at the protein interface, auxin enhances the TIR1–substrate interactions by acting as a ‘molecular glue’.”
TIR1 is a plant ubiquitin ligase. Only in the presence of auxin does TIR1 bind its target protein and tag it with ubiquitin for its destruction. Ning Zheng had discovered a new type of small molecule, completely different from PROTACs, that enables TPD by binding in the protein-protein interface and inducing proximity between two proteins.
It turns out auxin is far from the only such “molecular glue.” Synthetic compounds including thalidomide and its analogs have been found to act by the same mechanism. One thalidomide analog, Revlimid, is now one of the best-selling cancer drugs in the world, yielding over $12B of revenue in 2020. These cancer therapies trigger degradation of a range of transcription factors involved in cancer cell growth and proliferation. It’s increasingly clear that molecular glues such as auxin and thalidomide are not oddities. They represent a fundamentally different type of interaction between small molecules and proteins.
Hallmarks of molecular glues
By comparing molecular glues with PROTACs we can quickly point to three hallmarks of a molecular glue:
1) Simple. These molecules are incredibly small and elegant, usually below 400Da. Unlike PROTACs, molecular glues do not contain two warheads that separately engage a ubiquitin ligase and a target protein. In fact, no direct binding of molecular glues to the target protein is required. This essentially expands the druggable target space to the entire proteome, regardless of whether the target features a small molecule binding pocket. Whereas PROTACs expanded the druggable target space to include undruggable proteins, molecular glues expanded it to include unligandable proteins.
2) Cooperative. Typically, a molecular glue has some weak affinity to one of the two proteins, most often to the ubiquitin ligase. Almost no binding affinity is observed between these two proteins in the absence of the molecular glue. But when the three parties come together (the ligase, the target protein, and the molecular glue), there is an enormous gain of affinity, sometimes >1000 fold. The complex is often stable enough to be purified by column chromatography and observed by X-ray crystallography.
3) Catalytic. Compared with a traditional small molecule drug that inhibits protein function by binding a pocket on its surface, a molecular glue does not need to be present all of the time to exert its biological function. It only needs to be at the interface of the ligase and the target long enough for ubiquitin tagging to take place, so fewer molecular glue molecules than target proteins are required for target ubiquitination.
(Some researchers in the field use the term “molecular glue” more broadly to include large molecules that induce protein-protein binding. Here I am proposing these three hallmarks in the TPD context to reflect the original intention of Ning Zheng, who first coined the term in the context of protein degradation.)
Discovering new molecular glues
More molecular glues have proven to be hidden in plain sight. There are many approved drugs and biologically active small molecules whose mechanisms are poorly understood. It’s quite plausible they could also act as molecular glues. Researchers are increasingly applying advanced technologies such as proteomics to detect protein level changes, hoping to catch more molecular glues in action.
As more molecular glues are being discovered serendipitously, people are also taking a so-called phenotypic approach. For example, more thalidomide-like molecules are being made and tested for novel biological functions and degradation of new targets that could have roles in diseases.
Solving the puzzle
Our most exciting opportunity is to rationally discover novel molecular glues that can induce proximity between any target and a ubiquitin ligase of choice. This is like solving a 3D jigsaw puzzle starting with only one piece. The task we are facing is to find a second piece (a ubiquitin ligase, of which we humans have >600) that fits not-quite-perfectly with the first one, but well enough so that we can put the two pieces together, look at the space between them, and grab the third one, a much smaller piece, to complete the puzzle.
The three-body problem in physics is notoriously difficult to solve. Rational molecular glue discovery is like solving the three-body problem in biology. On one hand, it totally changes the paradigm in drug discovery. For the past 50 years, we have been focusing on binary interaction between a target protein and a small molecule to create inhibitors or the warheads for PROTACs. We don’t need such binary interactions anymore. All we need is a protein surface, augmented by a molecular glue, that can specifically form new protein-protein interactions. The opportunity is tremendous.
On the other hand, the complexity of the task is vast. New computational and experimental tools are needed to reduce that complexity—and many are being developed. For example, establishing a preferred feature for binding to a ubiquitin ligase is a great starting point to narrow down chemical space; several research groups and companies have used this to identify thalidomide-like molecules. In addition, searching for weak protein-protein interactions can help us to find the second piece of the puzzle, by providing a hint for which proteins might be amenable to molecular glue enhancement.
With advances in new computational tools, especially machine learning algorithms that can rapidly score and match protein surface features, and our increasing ability to interrogate the vast chemical space using DNA-encoded compound libraries and virtual screens, rational molecular discovery is well within reach and is becoming the new frontier for discovering powerful medicines.
This is a frontier that the team at RA Capital is actively exploring with our partners. We look forward to sharing our progress in the coming months.