(E-172) Design of Direct 26S Proteasome Recruiters with Language Model-Derived Guide Peptides

Introduction:: The human proteome consists of around 20,000 proteins, many of which can acquire or present pathogenic consequences in disease states. A subset of these are considered “undruggable” by standard pharmacological techniques using small molecules, due to their unstable structural conformations, localizations, and/or post-translational modifications. Targeted protein degradation (TPD) represents a powerful modality to eliminate pathogenic proteins via the ubiquitin-proteasomal system.1 However, small molecule-based TPD platforms, such as PROTACs and molecular glues, rely on accessible binding sites on structured protein targets. We previously generated peptide-E3 ubiquitin ligase fusions to bind and degrade target proteins to achieve a programmable system which is reliant on the ubiquitin-proteasomal pathway, an indirect, often inefficient, process to achieve degradation of pathogenic proteins.2–4 A recent study discovered ligands that can directly recruit the 26S proteasome to target proteins, one of which is a linear peptide not requiring chemical stabilization.5 In this study, we adopt this ligand to create a genetically-encoded architecture that fuses our our computationally-generated guide peptides to this ligand for direct proteasomal degradation of pathogenic proteins. We demonstrate proteasomal degradation of diverse proteins using various peptide design algorithms. Overall, our method represents a modular, genetically-encoded system for TPD.

Materials and Methods:: Protein design algorithms based upon protein language models and AlphaFold were used to generate specific target-binding peptides from the target sequence or structure. Generated peptides were experimentally cloned into pcDNA plasmid vectors upstream of proteasome recruiting-ligand sequences. Post-transfection of these plasmids into target-expressing fluorescent reporter cell lines at equimolar ratios, cells were collected and analyzed via flow cytometry for target degradation or phenotypic output. To quantitatively evaluate the functional β-catenin degradation, we use a TOPFlash assay, consisting of a GFP reporter that serves as a reliable readout of β-catenin/TCF-dependent activity.


Results, Conclusions, and Discussions:: Results and Discussion
Our direct proteasomal degradation construct consists of a computationally-derived peptide linked to a ligand that recruits the PSMD2 subunit of the 26S proteasome. We achieve statistically significant direct proteasomal degradation with a variation of linker lengths between the PSMD2 ligand and the target peptide. Using our previously computationally-derived potent binder of β-catenin (termed E9), we demonstrate statistically significant degradation of transiently transfected β-catenin compared to a non-targeting PSMD2 ligand control in HEK293 cells (Figure 2A). We confirm the proteasomal mechanism of action our constructs by demonstrating lack of degradation when incubating cells with MG132, a known 26S proteasome inhibitor (Figure 2A). Furthermore, we show a substantial decrease in the tumorigenic transcriptional response via TOPFlash relative to a non-targeting control, when treating colorectal cancer cells (DLD1) with our E9-based proteasomal recruiters (Figure 2B).


Conclusion
In this work, we aim to create a direct, CRISPR-analogous degradation system, thus generating biologics that can directly and programmably degrade dysregulated pathological proteins. To do this, we create a modular genetically-encoded direct proteasomal degradation by fusing computationally-generated target peptides to a linear ligand that recruits 26S proteasome. Our experiment results show statistically significant degradation of the target protein, followed by the modulation of the transcriptional activity. Current work entails exploring a more comprehensive range of targets and diverse strategies for intracellular delivery. By attaching a cell-penetrating peptide (CPP) upstream of the peptide-proteasome recruiter construct, we can enable direct delivery of our compact architecture to diverse cell types of interest, without the requirement of a external delivery vehicle, such as lipid nanoparticles (LNPs) or adeno-associated viruses (AAVs). Overall, our unique, genetically-encoded degradation architecture represents a modular, programmable system for TPD, enabling broad access to the human proteome.

Acknowledgements (Optional): :

References (Optional): : 1. Behan, F. M. et al. Prioritization of cancer therapeutic targets using CRISPR–Cas9 screens. Nature 568, 511–516 (2019).

2. Palepu, K. et al. Design of Peptide-Based Protein Degraders via Contrastive Deep Learning. bioRxiv 2022.05.23.493169 (2022) doi:10.1101/2022.05.23.493169.

3. Chatterjee, P. et al. Targeted intracellular degradation of SARS-CoV-2 via computationally optimized peptide fusions. Communications Biology 3, 1–8 (2020).

4. Brixi, G. et al. Design of Peptide-Guided Protein Degraders with Structure-Agnostic Language Models. (2023) doi:10.21203/rs.3.rs-1694183/v1.

5. Bashore, C. et al. Targeted degradation via direct 26S proteasome recruitment. Nat. Chem. Biol. 19, 55–63 (2022).