The transcription factor androgen receptor (AR) is a key driver of prostate cancer, the most frequent cancer among European men. Most available AR inhibitors compete with endogenous ligands by blocking the ligand-binding pocket, which is an effective treatment in many patients. However, escape mutations inevitably lead to constitutive AR reactivation causing the therapy-resistant, lethal form known as castration-resistant prostate cancer. Consequently, novel therapeutics with alternative mechanisms of action are highly sought after. To target therapy-resistant AR mutants, we aim to degrade AR by targeting it to the ubiquitinproteasome pathway using heterobifunctional degrader molecules. Among others, consensus-designed tetratricopeptide repeat proteins (CTPRs) were deployed as artificial modular scaffolds, onto which AR- and E3 ligase-binding moieties were grafted to induce AR ubiquitination and degradation. As the degraders bind AR through protein-protein interactions, we can effectively target sites distant from the ligand-binding pocket that are not affected by mutations of the latter.

To challenge the grafting capacity of CTPRs, 17 highly diverse, unstructured peptides were introduced into the inter-repeat loop of CTPR2 protein (comprising two repeats). We found that even long loops of over 50 amino acids could be accommodated and all designs were thermostable. While the effect on protein stability was dependent on the length of the inserted loop, the relationship between solubility and loop length was considerably more complex.

Three different approaches were explored in the design of heterobifunctional AR degraders, each targeting one of the three AR domains. In the first approach, we aimed to bind the AR ligand-binding domain by grafting peptides derived from AR coactivators onto the terminal helices of the CTPR using a rational design strategy. We observed that, in contrast to previous findings, helical peptide grafting is not straightforward for all peptides and can be much more complex than simply the transfer of key interaction residues, not only for retaining binding function but also for obtaining well-behaved soluble proteins. Computational saturation mutagenesis can potentially be used to screen for optimal interaction residues that do not abolish protein stability or solubility.

The second approach targeted AR’s DNA-binding domain (DBD) by utilising a short DNA sequence corresponding to the AR response element (ARE). ARE was covalently attached to a CTPR containing an E3 ligase KEAP1-binding peptide from the protein NRF2. Ternary complex formation between AR DBD, CTPR-NRF2-ARE and the E3 ligase KEAP1 was demonstrated using biophysical techniques, and the results of preliminary experiments assessing endogenous AR levels in prostate cancer cells were encouraging.

The third approach aimed to target the unstructured N-terminal domain of AR by exploring various known binders identified from the literature. The most promising results were obtained for the BRD4-BD1 domain onto which the NRF2 peptide was grafted to bind to KEAP1. The designed BRD4-NRF2 pulled down AR in cells, but unexpectedly revealed low binding affinity using biophysical methods.

The identification of suitable AR binders was challenging, and it is unknown which E3 ligases can degrade AR, considering their subcellular locations and specific geometries required for successful degradation. Therefore, two assays were developed to identify suitable degrons and E3 ligases, respectively. In the first assay, 16 degrons grafted on a CTPR scaffold were fused to AR, transfected into prostate cancer cells, and levels of AR-fusion proteins were measured to assess whether the degrons induce AR degradation. In the second assay, 10 chimeric E3 ligases were designed to degrade GFP-tagged AR. Comparative results obtained for another cancer target-GFP fusion showed that the AR-GFP fusion protein is more resistant to degradation. The results of these two assays not only provide new insights into future strategies for induced AR degradation but also constitute transferable toolkits for characterising novel targets.