Bardoxolone conjugation enables targeted protein degradation of BRD4

Targeted protein degradation (TPD) has emerged as a powerful tool in drug discovery for the perturbation of protein levels using heterobifunctional small molecules. E3 ligase recruiters remain central to this process yet relatively few have been identified relative to the ~ 600 predicted human E3 ligases. While, initial recruiters have utilized non-covalent chemistry for protein binding, very recently covalent engagement to novel E3’s has proven fruitful in TPD application. Herein we demonstrate efficient proteasome-mediated degradation of BRD4 by a bifunctional small molecule linking the KEAP1-Nrf2 activator bardoxolone to a BRD4 inhibitor JQ1.

www.nature.com/scientificreports/ amine 2 and the resulting amide (see 3) converted into TFA salt 4 which was used without further purification 33,34 . Amide bond formation between 4 and CDDO then delivered the bifunctional degrader CDDO-JQ1. We tested CDDO-JQ1 in the 231MFP human breast cancer cell line and observed dose-responsive degradation of BRD4; at higher concentrations loss of BRD4 degradation was observed, presumably due to the "hook" effect, a phenomenon not unexpected with a highly-reversible, covalently-binding bifunctional molecule (Fig. 2B). Interestingly, we also observed loss of KEAP1 at higher concentrations in cells treated with CDDO-JQ1, a finding that has been reported previously by treating cells with electrophilic stressors (Fig. 2B) 35 . Notably extensive degradation of BRD4 was observed in the 100-200 nM range without any optimization of linker length or composition 36 .
We further demonstrated that the CDDO-JQ1 mediated degradation of BRD4 was attenuated by pre-treatment with the proteasome inhibitor bortezomib as well as the E1 activating enzyme inhibitor MLN7243 (Fig. 2C). Given that KEAP1 belongs to the CUL3 family of E3 ligases that require NEDDylation for activity, we further demonstrated that the BRD4 degradation conferred by CDDO-JQ1 was also significantly attenuated by the NEDD8 activating enzyme inhibitor MLN4924 (Fig. 2D).
To show that the observed degradation by CDDO-JQ1 was not due to hydrophobic tagging of BRD4 leading to local protein unfolding and subsequent ubiquitination and degradation [38][39][40][41] , we synthesized two negative control compounds, both having significantly altered electrophilicity and resultant covalent protein reactivity, but crucially having similar physicochemical properties (Fig. 3A). First, hydrogenation of 5 (H 2 , Pd/C) generated ketone 6 as a mixture of diastereomers and enol tautomers. Coupling of this material with 4 then generated H 2 -CDDO-JQ1. Importantly, H 2 -CDDO-JQ1 did not induce BRD4 degradation in comparison to CDDO-JQ1 (Fig. 3B). Similarly, we prepared 3-oxo-oleanolic acid-JQ1 (3-OOA-JQ1), which possesses no potentially reactive alkenes of any type by coupling 7 42 with 4 and found it also did not degrade BRD4 as compared to CDDO-JQ1.
Finally, we prepared the des-cyano variant of CDDO-JQ1 by coupling 8 43 and 4 (Fig. 4A), and found that this compound (de-CN-CDDO-JQ1) also does not degrade BRD4 in 231 MFP cells indicating the criticality of the entire α-cyanoenone motif to this process (Fig. 4B). Interestingly, this chemical modification had also previously reduced the activity of bardoxolone methyl ~ 1,000 fold as an anti-inflammatory agent 30 . In general, all three of the changes made to the CDDO portion of CDDO-JQ1 (Figs. 3 and 4) were known to greatly reduce Nrf2 activation in the medicinal chemistry campaigns involving bardoxolone methyl as an anti-inflammatory drug 28-30 . conclusion In summary, the first bifunctional protein degrader (PROTAC) based on the known KEAP1 ligand bardoxolone (CDDO) is reported. While robust, proteasome-dependent degradation of BRD4 was observed, a number of mechanistic questions remain: firstly, CDDO is not known to bind the Kelch domain of the KEAP1/Cul3 complex, the key area for recognition of Nrf2 and where degrader 44 , and multiple small molecule inhibitors presumably bind [45][46][47][48] . While a crystal structure of CDDO bound to Cys-151 of the BTB domain of KEAP1 has been solved, various reports have also posited that multiple cysteines are targeted by this class of compounds during the Nrf2 activation process in cells [49][50][51][52][53] . This raises the question how (or if) a CDDO-based PROTAC can mechanistically induce neosubstrate degradation in a KEAP1-dependent manner. Secondly, while bardoxolone is thought to activate Nrf2 through the targeting of reactive cysteines on KEAP1, it also interacts with additional pharmacological targets, including IKKb which modulates NF-kB signaling 54 , mTOR 55 , and others 55,56 . It should be noted that the direct detection of all proteome-wide targets of CDDO by pulldown studies has proven www.nature.com/scientificreports/ challenging given the highly reversible nature of its cysteine interactions 30,50,53,[54][55][56] . Thus, we cannot rule out at this moment that the degradation observed here may be due to one (or more) other cullin-family E3 ligases that are targeted by CDDO-JQ1. Emerging evidence suggests that clinical resistance to PROTACs can occur through rewiring of the cellular E3 ligase machinery 57,58 , thus highlighting the critical need for more and diverse E3 ligase recruiters. Our combined results reported herein strongly implicate E3 ligase involvement and covalent cysteine reactivity in the mechanism of BRD4 degradation by CDDO-JQ1. Future chemoproteomic and genetic studies to map the proteome-wide targets of CDDO-JQ1 will be revealing in further understanding the mechanism of degradation reported herein. Nevertheless, the degradation toolkit has now been expanded to include CDDO as an easily prepared recruiter, and reversible, covalent E3 engagement as a promising concept for future TPD applications.

Experimental section
General synthetic methods. Unless otherwise noted, all reactions were performed in flame-dried glassware under a positive pressure of nitrogen or argon. Air-and moisture-sensitive liquids were transferred via syringe. Dry dichloromethane (CH 2 Cl 2 ) and N,N-dimethylformamide (DMF) were obtained by passing these previously degassed solvents through activated alumina columns. Bardoxolone methyl (CDDO-Me) was purchased from Sigma-Aldrich. Oleanolic acid was purchased from Acros Organics. (+)-JQ1 was purchased from Enovation Chemicals. All reagents were used as received from commercial sources, unless stated otherwise. Reactions were monitored by thin layer chromatography (TLC) on TLC silica gel 60 F 254 glass plates (EMD Millipore) and visualized by UV irradiation and staining with p-anisaldehyde, phosphomolybdic acid, or Ninhydrin. Volatile solvents were removed under reduced pressure using a rotary evaporator. Flash column chromatography was performed using Silicycle F60 silica gel (60 Å, 230-400 mesh, 40-63 μm). Proton nuclear magnetic resonance ( 1 H NMR) and carbon nuclear magnetic resonance ( 13 C NMR) spectra were recorded on Bruker AV-600 and AV-700 spectrometers operating at 600 and 700 MHz for 1 H NMR, and 151 and 176 MHz for 13 C NMR (Supplementary Information). Chemical shifts are reported in parts per million (ppm) with respect to the residual solvent signal CDCl 3 ( 1 H NMR: δ 7.26; 13 C NMR: δ 77.16), CD 2 Cl 2 ( 1 H NMR: δ 5.32; 13 C NMR: δ 53.84). Peak multiplicities are reported as follows: s = singlet, d = doublet, t = triplet, dd = doublet of doublets, tt = triplet of triplets, m = multiplet, br = broad signal. IR spectra were recorded on a Bruker Vertex80 FTIR spectrometer. High-resolution mass spectra (HRMS) were obtained by the QB3/chemistry mass spectrometry facility at the University of California, Berkeley using a Thermo LTQ-FT mass spectrometer with electrospray ionization (ESI) technique.
Cell culture. 231MFP cells were obtained from Prof. B. Cravatt (Scripps) and were generated from explanted tumor xenografts of MDA-MB-231 cells as previously described 59 . They were cultured in L15 medium containing 10% (v/v) fetal bovine serum (FBS), maintained at 37 °C with 0% CO 2 .
Cell-based degrader assays and western blotting. For assaying degrader activity, cells were seeded (500,000 for 231MFP cells) into 6 cm tissue culture plates (Corning) in 2.0-2.5 mL of media and allowed to adhere overnight. The following morning, media was replaced with complete media containing the desired concentration of CDDO-JQ1 (or related CDDO-based degrader) diluted from a 1,000 × stock in DMSO. For rescue studies, the cells were pre-treated with proteasome inhibitor (bortezomib, 1 μM), E1 inhibitor (MLN7243, 1 μM) or NED-Dylation inhibitor (MLN4924, 1 μM) for 30 min. Cells were subsequently treated with vehicle DMSO or degraders for 12 h. To harvest cells, media was aspirated and cells were washed with 500 μL PBS then 100 µL Radioimmunoprecipitation assay buffer (RIPA buffer) was added to each well and incubated 5 min on ice before scraping and transferring to Eppendorf tubes. The collected cells were vortexed vigorously in the lysis buffer and allowed to sit on ice for 5 additional min before cellular debris was pelleted by spinning at maximum speed for 10 min at 4 °C. Supernatant was transferred to new tubes and total protein was normalized by Pierce BCA Protein Assay. Samples were denatured by addition of 4 × Laemmli's Loading dye and 30 µg of protein was loaded onto 4-20% TGX Precast gels (BioRad). After gel electrophoresis, proteins were transferred to a nitrocellulose membrane using semi-dry transfer on a Trans-Blot Turbo (BioRad) over 7 min. The membrane was then incubated for 1 h in 5% bovine serum albumin (BSA) in tris-buffered saline containing Tween 20 (TBST) before incubation with the corresponding primary antibody overnight at 4 °C. The membranes were washed in TBST before a 1 h room temperature incubation with secondary antibodies. After a final set of washes, blots were imaged on a LiCor CLX imager and band intensities were quantified using ImageJ software (Supplementary Information).