Decoding CRBN Conformation Dance: Implications for PROTACs

Most people in the TPD field are likely familiar with the concept of CRBN's open and closed forms, as highlighted in a ground-breaking publication by Gabriel Lander on CRBN dynamics. Using cryo-EM reveals that the binding of molecular glue induces a change in conformation from open to closed. This closed conformation can then bind to the neosubstrate, which is ubiquitinated and degredated by the UPS system.

Unfortunately, there is no publication regarding PROTACs, and I think a lot of us are wondering if the dynamics of CRBN can be translated to PROTACs modalities too. Recently, I found a publication from 2018 that describes how CRBN is degraded by VHL-CRBN heterodimerizing PROTACs. This publication provides an important clue on how the open-closed dynamics can be significant for PROTACs as well.

Open-Close Dance of CRBN domains

But first, before going into that publication, here's a quick review of the mechanism of open-closed dynamics of CRBN induced by molecular glues. You can skip this section if you are familiar with the concept, but for those new to the TPD field, this will be a good introduction for further discussion.

CRBN is a protein that can be divided into four parts, as shown in Figure 1:

• N-terminal disordered region (NT Belt): ~1-64 residues
• N-terminal Lon protease-like domain (Lon domain): ~65-186 residues
• Helical bundle (HB), responsible for binding to DDB1 protein: ~187-318 residues
• C-terminal thalidomide-binding domain (TBD) that contains the tri-Trp imid binding pocket: ~319-442 residues (TB). Inside TBD, we can also define the sensory loop (SL): 341-361 residues and region responsible for binding ligands (RB): ~380-400 residues

In the Apo form, which is also called the open form, the N-terminal disordered region and sensory loop are disordered, and the Lon domain and TBD are not interacting with each other - Figure 2 on left.

In the first step, imid binds to the tri-Trp imid binding pocket in the TBD domain, which is sufficient by itself to bind the drug without the Lon domain. When the ligand is bound, the sensory loop rearranges from a disordered to a beta-hairpin motif, causing a very fast change into the closed conformation of CRBN. The last step of this Lon and TBD domain dance is the rearrangement of the N-terminal belt, which additionally stabilizes the closed conformation. You can see how this belt tightly interacts in the cartoon representation of the close conformation in Figure 3.

One of the most important take-home messages is that only the closed conformation of CRBN can bind the neosubstrate, and the open-closed dynamics equilibrium should be considered more important than just the binary Kd to CRBN. Based on this, all factors that can stabilize the closed conformation should increase the degradation of the neosubstrate, as demonstrated in the publication for mezigdomide.

The effect of NT belt stabilization on the active closed conformation was studied using the CRBN∆NTD construct with truncated 1-63 residues. They found that deletion of the N-terminal residues changed the ratio of closed to open CRBN conformation from 20% to 2% in the case of pomalidomide.

Heterodimerizing PROTAC

Now, going back to the PROTAC publication from 2018, in this paper, they studied the degradation of CRBN and VHL using a heterodimerizing degrader with both VHL and CRBN warheads (Figure 4). They observed only CRBN degradation using this bifunctional degrader. More importantly for us, they found that the N-terminally truncated CRBN construct ∆1-81 residues is not degraded by VHL-CRBN heterodimerizing PROTACs.

Now, if we combine these observations from both papers and connect the dots, we can see that the same N-terminal belt plays a crucial role in effective degradation for both molecular glues and PROTACs. Based on this information, we can confidently assume that the closed conformation is necessary for effective degradation of PROTACs in this case and likely can be translated to others

Going Further

The reasoning behind the presence of CRBN dynamics

The first and most obvious reason is energy. In the case of degradation, only proteins that possess defects should be targeted. Due to the high cost of their resynthesis, this process needs to be precisely controlled by the selectivity of the UPS system .

Based on publications from Christina Woo which explore the natural the natural substrates of CRBN, we observe that CRBN is responsible for degrading proteins that possess a terminal glutarimide ring, likely formed through internal protein chain cleavage. The presence of a glutarimide ring signals the degradation of defective proteins, which can then bind to the tri-Trp pocket of CRBN. An additional mechanism for discriminating between open and closed conformations further distinguishes amino acids beyond the glutarimide in the sequence, ensuring that only correct degrons capable of stabilizing CRBN in the closed conformation are degraded.

If a part of the protein binds to CRBN, it needs to be able to induce CRBN closure and increase residence time, which should increase the likelihood of progressing through all stages leading to ubiquitination. You can imagine that if a protein not targeted by CRBN binds to it and fails to induce closure, CRBN will have a shorter residence time. This causes quick dissociation from CRBN and saves the protein of interest from degradation.

Additionally, these open-close dynamics can contribute to greater substrate flexibility of CRBN. An example is seen in a presentation by Monte Rosa Therapeutics, where they demonstrated that CRBN can adopt a partially open conformation in the case of their NEK7 molecular glue ( link to the Webinar)

Mechanism behind the CRBN dynamics

First and most obviously, these processes can create a ternary complex with a long half-life. When a PROTAC or molecular glue binds to the open conformation or only to constructs with the TBD domain, you can observe binary interactions. However, compared to the closed conformation, these interactions should be weaker and have lower residence time.

After binding ligand to CRBN, it goes through at least three rearrangement steps (Figure 1) to transform from the open to closed conformation. To dissociate from CRBN, the entire process needs to reverse. When a protein needs to change conformation and adjust to the ligand, it increases the residence time, which in most cases leads to an increasing Kd. You can see how ligand is buried in protein in side view of CRBN in close conformation with bounded Pomalidomide on Figure 5. Moreover, in the closed conformation, we can observe additional interactions between the sensory loop (pink) and ligand, which also contribute to the stability of the binary complex in the closed state.

The longer CRBN remains in the closed conformation, or in a more AI-friendly way, the higher the probability that CRBN will be in the closed conformation necessary for degrading the neosubstrate the higher probability of creating a TCF with neosubstrate.

Moreover, high residence time is important to increase the probability of subsequent events after TCF creation, such as the assembly of the whole ubiquitination machinery in the right place and time, leading to ubiquitination. By increasing residence time, we can reduce the impact of CRBN dissociation from the complex during this cascade of events. Simplified point of view these mechanism is that increased occupancy can be seen as reducing the three-body problem to a binary interaction.

Additionally, the closed conformation of CRBN could create a more localized surface that can facilitate ordered protein-protein interactions with the neosubstrate compared to the open one.

Practical implication

Armed with this knowledge, we should also consider the important key takeaways from these papers and how they can be implemented in the degrader design process and principles. Below is a list of the most important ones in my opinion:

• Using only the TBD construct may result in binary interaction, but it might not translate to the degradation results observed in biology due to the conversion from the open to closed conformation being a limiting step. It would be recommended to use the full-length construct or some modification recently published in the literature that includes functional open-close dynamics in the system.
• It would be preferable and desirable to use methods that provide kinetic interaction data, such as surface plasmon resonance (SPR), rather than methods that only provide Kd values, like fluorescence polarization (FP), to determine your system.
• You could compare results from biophysics and biology between wild-type CRBN and truncated constructs with a ∆1-64 deletion to understand the role of the N-terminal disordered belt as a proxy for the impact of open-close dynamics on the degradation mechanism in your studied system.
• Molecular modelling should be performed on binary and ternary complexes with the closed conformation as a template.
• AI models can include the probability of the CRBN open/close ratio as a metric from calculations, such as molecular dynamics (MD), for example.

Additional Notes

• For more insights into the dynamics, you can refer to the original publication here and the Dana-Farber TPD seminar video, which I highly encourage you to watch if you haven't already.
• The N-terminal disordered part, often dismissed as not crucial, can have a significant impact on protein dynamics and structure of drug action.
• The partially open conformation of CRBN with NEK7 degrader close to the closed conformation. We can imagine some deviation from fully closed CRBN, which would work as long as the ligand can stabilize the conformation of CRBN in one position with a sufficiently long half-life.
• Fortunately, the publication used flexible peg and alkyl as a linker, which could help us understand the effects of N-terminal belt deletion on degradation. In future posts, I will delve deeper into the topic of stabilizing open and closed conformations by molecular glue and PROTACs, providing further explanations.
• Additionally, comparing CRBN's role to other ligases responsible for homeostasis control, such as VHL or those involved in creating large multicomplex structures, could lead to hypotheses about E3 ligases. However, this is a topic for another post.
• One of the most important questions to ask is whether this mechanism of regulating E3 activity is applicable to other E3 ligases or if there is a bias specific to CRBN.