Discovery of Monovalent Direct Degraders of BRD4 that Act via the Recruitment of DCAF11

Targeted protein degradation (TPD) is a rapidly emerging therapeutic modality that utilizes the cellular protein homeostasis machinery to eliminate pathogenic proteins. As a general mechanistic principle, TPD exploits small molecules to induce proximity between the protein-of-interest (POI) and an E3 ligase substrate receptor, resulting in ubiquitination of the POI and its subsequent degradation via the 26S proteasome (1). Typically, these small-molecule degraders fall into two distinct classes, molecular glues and heterobifunctional degraders. Most molecular glues bind to the E3 ligase component and create a unique protein surface that is recognized by neosubstrates, leading to their degradation. This class of degrader is exemplified by numerous immunomodulatory drugs (IMiDs), which bind to the E3 ligase, cereblon, and recruit various neosubstrates for efficient elimination (2, 3). Beyond the IMiD family, other molecular glues have been disclosed, but their discovery has often relied on serendipity (4, 5), making prospective design to specific neosubstrates difficult. Heterobifunctional degraders bypass fortuitous discovery by taking advantage of linking known ligands to both the E3 ligase and the POI, thus inducing protein proximity (6, 7). This modality has greatly accelerated the field by enhancing the ability to design tool degraders of various pathogenic proteins and has led to the first clinical trials involving designed heterobifunctional degraders (1). However, designing drug-like heterobifunctional degraders is challenging due to their relatively high molecular weight and overall complexity, and their typically nondrug-like properties.

A key challenge with prospective design of drug-like degraders, particularly with molecular glues, has been redirecting an E3 to a pathogenic protein of choice. Specific structural motifs or amino acid sequences, known as degrons, have been identified in various neosubstrates which are critical for mediating protein–protein interactions with molecular glue-bound E3 ligases (8). However, applying established degrons to novel neosubstrates and/or discovering new degrons is an emerging field with few guidelines for success (8–11). Furthermore, both molecular glues and heterobifunctional degraders have so far been limited to using only a few E3 ligases (e.g., cereblon, VHL, etc.; ref. 1). An alternative strategy to approach targeted protein degradation is to design prospective degraders to bind to the POI itself, facilitating its degradation through induced interactions with an E3 ligase that is selected by the cellular machinery. These molecules mechanistically parallel molecular glues, in that they only have significant binding affinity for one of the two proteins within the ternary complex, with binding affinity designed to the POI instead of the E3 ligase. Phenotypic screens are particularly amenable for direct degrader discovery utilizing this approach. Binders to the POI can be designed in an E3 ligase agnostic manner and then exposed to the repertoire of cellular E3 ligases to potentially induce productive interactions (6, 12–14). However, like molecular glues, discovery of direct degraders are often rare events.

To explore the potential of discovering, developing, and characterizing degraders that target a specific protein of interest, we utilized our ultra-high throughput screening (uHTS) methodology which measures degradation of target proteins upon exposure to diverse E3 ligase agnostic chemical libraries. We chose BRD4 as a target due to its functional importance in controlling the expression of oncogenes and antiapoptotic proteins, as well as its proven degradation feasibility and availability of chemical starting points (15). BRD4 is a member of the bromodomain extraterminal (BET) family of proteins, consisting of BRD2, BRD3, BRD4, and BRDT, that all share homologous tandem bromodomain (BD) acetyl-lysine binding sites. This homology has made the development of selective inhibitors very challenging. Clinical development of pan-BET inhibitors has been hindered by dose-limiting toxicities; therefore, there is ongoing effort to design selective small-molecule modulators. The TPD modality has been utilized to design degraders that can selectively modulate specific protein isoforms (16–18). We were interested in implementing our monovalent direct degrader strategy to develop BRD4 isoform selective degraders. Utilizing reported co-crystal structures of inhibitors bound in the BD1 binding domain of BRD4 (12), vectors were identified on the small molecule that were amenable for designing bead-based libraries that were screened in our uHTS platform. The result of our screening and subsequent medicinal chemistry efforts led to the optimized compound, PLX-3618. In this paper, we describe the properties of BRD4 degradation that are induced by our novel direct degrader. PLX-3618 binds nonselectively to BET family proteins but induces selective degradation of BRD4 resulting in potent in vitro and in vivo antitumor activity that is superior to inhibitor-based strategies. Using CRISPR screens and mutational studies, we demonstrate that PLX-3618 mediates degradation of BRD4 through the recruitment of the E3 ligase substrate receptor, DCAF11.

See Supplementary Tables S3 and S4.

Cell lines were typically cultured in T-75 flasks and checked monthly for mycoplasma contamination using a PCR-based detection kit (Applied Biological Materials). Cells were kept at low passage numbers (<30) in humidified incubators at 37°C and 5% CO₂. Cell lines were obtained from ATCC (HEK-293T, CRL-3216; MV-4-11, CL-9591; LNCaP, CRL-1740; KG-1, CCL-246; Kasumi-1, CRL-2724; THP-1, TIB-202), DSMZ (Nomo-1, ACC 542; Molm-13, ACC 554; NB-4, ACC 207), Promega (HiBiT-BRD4-HEK293, CS3023269), and System Biosciences (HEK293-Cas9, CAS630A-1). Cell lines were authenticated prior to receipt and no additional authentication was performed. The following media were used: RPMI-1640 + 10% FBS (KG-1, LNCaP, THP-1); RPMI-1640 + 10% heat-inactivated FBS (Nomo-1, Molm-13, NB-4); RPMI-1640 + 20% FBS (Kasumi-1); IMDM + 10% FBS (MV-4-11); DMEM + 10% FBS (HEK-293T, HEK293-Cas9, HiBiT-BRD4-HEK293). All media contained 1% penicillin/streptomycin (Thermo Fisher, 15140122).

After compound treatment (see specific assay methods), cells were washed once with cold PBS and subsequently lysed using 1× RIPA lysis buffer (Thermo Fisher) and Complete EDTA-free Protease Inhibiter Cocktail (Roche). Cell lysates were clarified by centrifugation at maximum speed for 10 minutes at 4°C, transferred to a new microcentrifuge tube, and total protein quantified using the Pierce bicinchoninic acid (BCA) Protein Assay (Roche). Protein samples were diluted to 0.2 mg/mL using 0.1× sample buffer (Protein Simple) and separated using 66 to 440 kDa separation modules and Wes/Jess/Abby instruments according to manufacturer’s instructions. Proteins of interest were normalized to vinculin. Antibody concentrations are noted in the Supplementary Tables.

LNCap cells were treated with 100 nmol/L PLX-3618 for 24 hours. Sample processing and TMT 16-plex isobaric labeling was conducted at Sanford Burnham. Cells were lysed in 8 mol/L urea, 50 mmol/L ammonium bicarbonate and Benzonase, and the lysate was centrifuged at 14,000 × g for 15 minutes to remove cellular debris. Supernatant protein concentration was determined using a BCA protein assay (Thermo Scientific). Disulfide bridges were reduced with 5 mmol/L tris(2-carboxyethyl)phosphine (TCEP) at 30°C for 60 minutes, and cysteines were subsequently alkylated with 15 mmol/L iodoacetamide in the dark at room temperature for 30 minutes. Urea was then diluted to 1 mol/L urea using 50 mmol/L ammonium bicarbonate, and proteins were subjected to overnight digestion with mass spec grade Trypsin/Lys-C mix (Promega). Following digestion, samples were acidified with formic acid and subsequently peptides were desalted using AssayMap C18 cartridges mounted on an AssayMap Bravo Platform (Agilent Technologies).

Peptide concentration was determined using a NanoDrop spectrophotometer (Thermo Scientific) and 20 micrograms of total peptide was used for TMT labeling. Briefly, dried peptides were reconstituted directly in 50% acetonitrile in 50 mmol/L HEPES (pH 8.5) containing one of the TMT tags from the TMT16plex reagent (Thermo Fisher). Peptide-TMT mixture was incubated for 1 hour at 25°C, and the reaction was stopped by addition of hydroxylamine to a final concentration of 0.2% and incubated for 15 minutes at 25°C before pooling the TMT-labeled samples. After pooling, the sample was dried using a SpeedVac system, resuspended in 0.1% formic acid and desalted using a C18 TopTip (PolyLC) according to the manufacturer’s recommendation, and finally the organic solvent was removed in a SpeedVac system.

Dried pooled samples were reconstituted in 20 mmol/L ammonium formate pH ∼10 for chromatography fractionation using a Waters Acquity BEH C18 column (2.1 × 15 cm, 1.7 µm pore size) mounted on an M-Class Ultra Performance Liquid Chromatography system (Waters). Peptides were then separated using a 35-minute gradient: 5% to 18% B in 3 minutes, 18% to 36% B in 20 minutes, 36% to 46% B in 2 minutes, 46% to 60% B in 5 minutes, and 60% to 70% B in 5 minutes (A = 20 mmol/L ammonium formate, pH 10; B = 100% acetonitrile). A total of 48 fractions were collected and pooled in a noncontiguous manner into 24 total fractions which were then dried to completeness in a SpeedVac concentrator prior to mass spectrometry analysis.

Dried peptide fractions were reconstituted with 2% ACN, 0.1% FA, and analyzed by LC-MS/MS using a Proxeon EASY nanoLC system (Thermo Fisher Scientific) coupled to an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific). Peptides were separated using an analytical C18 Aurora column (75 µm × 250 mm, 1.6 µm particles; IonOpticks) at a flow rate of 300 nL/minutes using a 75-minute gradient: 1% to 6% B in 1 minute, 6% to 23% B in 44 minutes, 23% to 34% B in 28 minutes, and 27% to 48% B in 2 minutes (A = FA 0.1%; B = 80% ACN: 0.1% FA). The mass spectrometer was operated in positive data-dependent acquisition mode. MS1 spectra were measured in the Orbitrap with a resolution of 60,000, at accumulation gain control (AGC) target of 4 ×10⁵ with maximum injection time of 50 ms, and within a mass range from 350 to 1,500 m/z. The instrument was set to run in top speed mode with 3-second cycles for the survey and the MS/MS scans. After a survey scan, tandem MS was performed on the most abundant precursors with charge state between +2 and +7 by isolating them in the quadrupole with an isolation window of 0.7 m/z. Precursors were fragmented with higher-energy collisional dissociation (HCD) with normalized collision energy of 35% and the resulting fragments were detected in the Orbitrap at 50,000 resolution, at AGC of 5 ×10⁵ and maximum injection time of 86 ms. The dynamic exclusion was set to 20 seconds with a 10 ppm mass tolerance around the precursor.

All mass spectra from were analyzed with MaxQuant software (v. 1.6.11) using the TMT 16-plex default settings. The search criteria were set as follows: full tryptic specificity was required (cleavage after lysine or arginine residues unless followed by proline), two missed cleavages were allowed, carbamidomethylation (C), TMTpro (K and peptide n-terminus) were set as fixed modification and oxidation (M) as a variable modification. The false identification rate was set to 1%.

Statistical analysis of the TMT data were carried out in R (version 3.5.1, 64-bit), using CRAN and Bioconductor packages such as data.table, tidyverse, limma, and MSstatsTMT. First, TMT reporter intensities were log₂-transformed and loess-normalized, using the normalizeCyclicLoess function from limma package, across all samples to account for systematic errors. Following normalization, all non-protein-group-specific peptide sequences and precursor isolation interference below 0.8 were removed from the list prior to statistical test. Finally, PSM’s with the same peptide sequence were rolled up by summing up TMT intensities within each channel and protein-level quantification and statistical testing for differential abundance were performed using MSstatsTMT bioconductor package.

NanoLuc-BRD4 plasmids (Promega) were transiently transfected into HEK293T cells at 10 ng DNA/well of six well plate, using Lipofectamine 3000 transfection reagent (Thermo Fisher) according to manufacturer’s instructions. Following overnight incubation at 37°C, 5% CO₂, cells were trypsinized, washed with PBS, and diluted into Opti-MEM (Gibco) at a density of 2 × 10⁵ cells/mL and plated into 384 well plates (Greiner 781080-20). A total of 50 µmol/L stock BRD4 tracer was diluted into Tracer Dilution Buffer (Promega) and added to cells at a final concentration of 0.5 µmol/L. Following a quick spin and mixing using a MultiDrop, compounds were dispensed using a TECAN D300e digital dispenser and incubated for 2 hours at 37°C, 5% CO₂. Plates were brought to RT and complete NanoBRET Nano-Glo reagent was added to wells (1:166 Nano-Glo Substrate + 1:500 Extracellular NanoLuc Inhibitor). Plates were mixed, incubated briefly, and then read at 450 nm (donor) and 610 nm (acceptor) wavelengths on a Clariostar (BMG) plate reader.

Immunoprecipitation (IP) of ubiquitinated proteins was carried out using the Signal Seeker Ubiquitin Detection Kit (Cytoskeleten Inc, BK161). In brief, 3 × 10⁶ LNCaP cells were plated in 10 cm plates and incubated overnight at 37°C. Cells were treated with 500 nmol/L bortezomib for 2 hours prior to the addition of DMSO (- control) or 10 nmol/L PLX-3618 and incubation was carried out for four additional hours. Cells were harvested and processed according to Signal Seeker Ubiquitination Detection Kit specifications. BRD4 levels were detected in the immunoprecipitated protein pool via capillary electrophoresis.

The arrayed CRISPR multiguide ubiquitin proteasome system (UPS) library (Synthego) was reverse transfected into Cas9-expressing HEK293 cells (SBI) at a final concentration of 0.15 µmol/L single-guide RNA (sgRNA)/well using Lipofectamine 3000 (0.16 μL/well) according to Synthego’s recommendations. Sixteen hours post-transfection, an equal volume of complete media was added to wells and plates were incubated until 56 hours post-transfection. At that time, 1 µmol/L PLX-3618 was added to wells using a TECAN D300e digital dispenser and incubated for the final 16 hours of the experiment (experiment total time was 72 hours from start of transfection). Cellular BRD4 levels were detected using immunofluorescence staining (protocol follows) and fluorescence signal was quantitated as described below. The BRD4 average cell intensity median was calculated for each plate and a robust Z score was determined using the equation 0.6745 × (X−μ1)/MAD, with values defined below:

• X = BRD4 average cell intensity

• μ1 = median of all BRD4 average cell intensity values on the plate

• MAD = median of the absolute deviation from the median of all the data points

DCAF11 sgRNA pool (Synthego) was transfected into HEK293 Cas9 cells (SBI) using a Neon Electroporation System (ThermoFisher). Genome editing was complete at 72 hours post-transfection and cells were expanded and collected to determine editing efficiency. Genomic DNA was made using QuickExtract DNA Extraction Solution 1.0 (Biosearch Technologies) and the edited DCAF11 genomic DNA fragment was PCR amplified using KAPA HiFi Hotstart Ready Mix 2× (Roche). The oligos used for PCR and Sanger sequencing were as follows:

Forward: CCT​TGG​TAG​TTG​GCA​TAG​GCT

Reverse: CTG​TCC​TGA​GCG​AAT​CCC​AT

Sequencing: CTT​GGT​AGT​TGG​CAT​AGG​CTA​AAG​AAA​AG.

Editing efficiency of DCAF11 knockout (KO) pools was determined to be 70% (% Indels, KO score of 63) using Synthego’s proprietary ICE tool. Knockout pools were also treated with 1 µmol/L PLX-3618 for 16 hours to demonstrate BRD4 degradation dependence on DCAF11. Following treatment, cell lysates were analyzed on a Simple Western Instrument (Bio-Techne), blotting for BRD4, DCAF11, and vinculin.

Corning 3764 plates (black well, clear bottom) were coated with poly-D-lysine (ThermoFisher) diluted to 50% with PBS (+Ca₂ +Mg) prior to reverse transfection of the Synthego sgRNA library. Following library transfection and PLX-3618 treatment, media was removed from cells and 2.5% formalin (10% Formalin, Electron Microscopy Sciences 15740) diluted in PBS (+Ca₂ +Mg) was added to cells. Plates were incubated at 37°C for 20 minutes for fixation followed by a PBS (+Ca₂ +Mg) wash. Methanol was then added to wells and plates were stored at −20°C overnight. The following day, PBS (−Ca₂ −Mg) was added to dilute the methanol in the wells, followed by two additional PBS washes. Blocking solution (1× fish gelation + 0.3% Triton X-100 in PBS (−Ca₂ −Mg) was added to wells and incubated at room temperature for 1 hour. Anti-BRD4 antibody was diluted in 1× fish gelatin + 0.1% Triton X-100 in PBS (−Ca₂ −Mg; Antibody Dilution Buffer), added to wells, and incubated overnight at 4°C. The following day, primary antibody was removed, and wells were washed three times in PBS (−Ca₂ −Mg) + 0.1% Tween-20, incubating 5 minutes on the final wash. Secondary antibody (anti-rabbit AF488, Southern Bio) and DAPI were diluted in Antibody Dilution Buffer, added to plate wells, and incubated for 2 hours at room temperature. Plates were washed three times as above followed by a final wash in 1× PBS (−Ca₂ −Mg). Plate imaging was performed on an ImageXress Pico system, followed by fluorescence intensity analysis calculations by Cell Reporter Xpress software.

HiBiT-BRD4 knock-in (KI) HEK293 cells (Promega CS3023269) were grown in DMEM + 10% FBS. Cells were plated at density of 2.67 × 10⁵ cells/mL into 384 well plates (Greiner 781080-20) and incubated overnight at 37°C, 5% CO₂. Cells were dosed with compounds using a TECAN D300e digital dispenser, plates spun briefly to mix, and incubated for 6 hours at 37°C, 5% CO₂. In combination studies using proteasome inhibitor (Bortezomib, 0.1 µmol/L) or neddylation inhibitor (MLN4924, 1 µmol/L), cells were first treated with the inhibitor for 1 hour followed by test compound dosing and incubation for an additional 6 hours. Post-incubation, plates were equilibrated to room temperature and equal volume of prepared Nano-Glo HiBiT Lytic Detection System (Promega; Lytic Buffer + 1:100 LgBiT + 1:50 HiBiT substrate) was added. Plates were incubated on an orbital plate shaker for 10 minutes and luciferase signal was read on a Clariostar plate reader (BMG).

HiBiT-BRD4 degradation assays using exogenously expressed BRD4 (BD1-BD2) WT and BD point mutants were performed as above except HEK293T cells were first transfected with pBit3.2N (TK-HiBiT) BRD4 constructs (10 ng DNA/well of six well plate) using Lipofectamine 3000. The next day, transfected cells were diluted in growth media to a density of 2 × 10⁵ cells/mL, replated into 384 well plates, and treated with compound for 6 hours.

pBit1.1C BRD4 (BRD4 (BD1-BD2)-LgBit, (WT or BD containing point mutants), pBit1.1C BRD2 [BRD2 (BD1-BD2)-LgBit], and pBit1.1C BRD3 [BRD3 (BD1-BD2)-LgBit] were transfected into HEK293T cells along with pBit2.1C DCAF11 (DCAF11-SmBit) and DDB1 using Lipofectamine 3000. Following overnight incubation at 37°C, 5% CO₂, transfected cells were diluted in Opti-Mem + 10% FBS and replated into 384 well plates at a density of 2 × 10⁵ cells/mL. Cells attached overnight and Nano-Glo Vivazine Live Cell Substrate (Promega) diluted in Opti-Mem and was added to cells to 1× concentration the following day. Plates were incubated for 1 hour to allow for Vivazine substrate accumulation and then dosed with compound. Luciferase measurements were taken at indicated timepoints.

NanoLuc-BRD4 [pNLF1N-BRD4 (BD1-BD2)] WT and bromodomain point mutants were each cotransfected with HaloTag-CRBN (Promega) and DDB1 (Promega) into HEK293T cells and incubated overnight at 37°C, 5% CO₂. Transfected cells were resuspended in Opti-Mem + 4% FBS at a density of 2 × 10⁵ cells/mL and split into two pools. To one pool, DMSO was added to 0.1% while HaloTag NanoBRET 618 Ligand was added to the other pool to 100 nmol/L (final concentration). Cells were allowed to attach overnight (37°C, 5% CO₂). Nano-Glo Vivazine Live Cell Substrate (Promega) diluted in Opti-Mem was added to cells to 1× concentration the following morning. After 1 hour incubation, cells were then dosed with dBET1 at indicated concentrations and plates were read on a Clariostar (BMG) plate reader at both donor emission (460 nm) and acceptor emission (618 nm) wavelengths at the indicated timepoints.

Five million HEK293T cells/10 cm dish were plated in complete media (minus Pen Strep). The next day, cells were transfected with BRD4 (BD1-BD2)-FLAG and HA-DCAF11 (5 µg plasmid DNA each) using Lipofectamine 3000. Twenty four hours post transfection, cells were dosed first with 0.1 µmol/L bortezomib, followed by 1 µmol/L PLX-3618 cotreatment on selected plates for an additional 4 hours. Cells were collected and lysed with 0.2% NP-40 lysis buffer (25 mmol/L Tris-HCl pH 7.4, 150 mmol/L NaCl, 10% glycerol, 0.2% NP-40) supplemented with Halt Protease Inhibitor Cocktail (Pierce). Cells were vortexed, placed on ice 15 minutes, then passed through a 1 mL syringe with a 23G needle 10 to 15 times. After centrifugation at 16,000 g for 10 minutes at 4°C, the supernatant was collected and added to anti-DYKDDDDK magnetic agarose beads (Pierce) and incubated for 2 hours at 4°C. Post-incubation, beads were washed four times with lysis buffer and 2× Lammeli sample buffer was added (minus reducing agent) to the beads. The beads were heated to 70°C for 10 minutes to elute the BRD4-FLAG and its interacting proteins. The eluate was transferred to a new tube to which β-mercaptoethanol was added. Samples were boiled at 95°C and loaded on 4% to 20% Mini Protean TGX gels (Bio-Rad) for Western blot analysis using the Tris-glycine buffer system (Bio-Rad). Twenty-five percent of immunoprecipitated material and 1% input were analyzed. Gels were transferred using a Genie Blotter (Idea Scientific) to 0.2 μmol/L nitrocellulose (Bio-Rad). Following incubation with primary and secondary antibodies and standard washes with TBS-T (0.2% Tween-20), membranes were imaged on a Li-Cor Odyssey CLx Imager.

Codon optimized human DCAF11 (41–546; UniProt ID Q8TEB1) and DDB1 (1–1,140; UniProt ID Q16531) were synthesized by GeneArt (ThermoFisher Scientific Inc.) and cloned into a proprietary vector. A tag composed of 8× His-MBP-PreScission-Avi-TEV cleavage site was added to the N-terminus of the DCAF11 sequence. The proteins were coexpressed in mammalian HEK293 cells and in vivo biotinylated. Cells were frozen at −80°C after harvesting.

The DCAF11-DDB1 complex was purified using multiple steps of IMAC and SEC. The cell pellet was resuspended in lysis buffer (20 mmol/L HEPES pH 7.4, 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 1 mmol/L TCEP), with added protease inhibitor (Complete, EDTA-free Protease Inhibitor, Merck, 04693159001) and 2.5 U/mL benzonase nuclease (Merck, E1014), and lysed by homogenization at 2.4 kPsi at 4°C. The cleared lysate was loaded onto a His-Trap column equilibrated in IMAC buffer A (25 mmol/L Tris pH 8.2, 500 mmol/L NaCl, 1 mmol/L TCEP, 10 mmol/L imidazole) and eluted with a linear gradient of IMAC buffer B (25 mmol/L Tris pH 8.2, 500 NaCl, 1 mmol/L TCEP, 500 mmol/L imidazole). Soluble aggregates were separated with SEC in 20 mmol/L Tris pH 8, 200 mmol/L NaCl, 1 mmol/L TCEP. The N-terminal 8× His-MBP-PreScission tag was cleaved by incubating the soluble DCAF11-DDB1 complex with His-tagged HRV 3C protease (Merck, SAE0045) in 1:1 w/w ratio O/N at 4°C. 3C protease and the cleaved tag were separated from the Avi-tagged protein complex using a TALON column equilibrated in SEC buffer. Cleaved avi-DCAF11-DDB1 was collected in the flow-through and polished with a final SEC step (Superdex S200) in which a small excess of DDB1 was isolated from the protein complex. The proteins were stored at −80°C in SEC buffer with added 5% glycerol (v/v). The yield of pure biotinylated avi-DCAF11-DDB1 complex ranged between 1 and 2 mg/L.

Human BRD4 (44–460) and human BRD4 (333–460; UniProt ID O60885) were tagged at the N-terminus with an 8His-TEV cleavage site. Codon optimized constructs were synthesized by GenScript and cloned in the pET-28a (+) vector. The expression was performed in E. coli BL21 (DE3; Bioké, C250H) in TB medium and induced at OD₆₀₀ 0.8 with 0.5 mmol/L IPTG. The expression was carried out at 16°C for 18 hours at 250 RPM. Cells were frozen at −80°C after harvesting.

Both human BRD4 (44–460) and BRD4 (333–460) were purified in a two-step purification based on IMAC and SEC. The cells were resuspended in lysis buffer (50 mmol/L HEPES pH 7.5, 500 mmol/L NaCl, 2 mmol/L TCEP, 20 mmol/L imidazole), protein inhibitor (cOmplete, EDTA-free Protease Inhibitor, Merck, 04693159001), 2.5 U/mL of benzonase nuclease (Merck, E1014), and lysed through homogenization at 30 kPsi at 4°C. The lysate was centrifuged, and the supernatant was run through a His-Trap column equilibrated in IMAC Buffer A (50 mmol/L HEPES pH 7.5, 500 mmol/L NaCl, 2 mmol/L TCEP, 20 mmol/L imidazole). The proteins were eluted with a linear gradient of IMAC Buffer B (50 mmol/L HEPES pH 7.5, 500 mmol/L NaCl, 2 mmol/L TCEP, 1 mol/L imidazole). The buffers used for the purification of BRD4 (44–460) contained 10% glycerol (v/v). The SEC step (Superdex S75) in 50 mmol/L HEPES pH 7.5, 500 mmol/L NaCl, 2 mmol/L TCEP, recovered around 10 mg per liter of culture.

Compounds were serially diluted in DMSO and then to 4× concentrations in TR-FRET buffer (20 mmol/L HEPES pH 7.5, 150 mmol/L NaCl, 1 mmol/L TCEP, 0.05% P20, 0.01% BSA). Five microliters of 4× compound dilutions were dispensed into wells of a black proxiplate (PerkinElmer). To the compound wells, 5 µL of 400 nmol/L His-tagged BRD4 (44–460) and 5 µL of 10 nmol/L biotinylated Avi-tagged DCAF11-DDB1 were added to initiate ternary complex formation. After 1 hour of incubation at 20°C, 5 µL of a TR-FRET detection pair consisting of 0.625 nmol/L Eu-Streptavidin (PerkinElmer/Cisbio) and 2 ng/µL of anti-6His-d6 monoclonal antibody (PerkinElmer/Cisbio) was added to each well, and the mixture was incubated for an additional 2.5 hours at 20°C. Prior to the TR-FRET signal measurement, 2 µL of 2 mol/L KF was added for signal enhancement. The fluorescent intensities from the ternary complex mixtures were measured using a BMG Clariostar with an excitation filter at 320 nm and emission filters at 615 and 665 nm. The TR-FRET signal was determined by the ratio of intensities between 665 and 615 nm. To visualize the compound-dependent ternary complex formation, the TR-FRET signals were normalized to DMSO control and plotted against the compound concentrations in GraphPad Prism.

All SPR experiments were executed at 15°C on a Biacore T200. BRD4 (44–460) and BRD4 (333–460) proteins were minimally biotinylated via primary amines using NHS-LC-LC-biotin (Thermofisher scientific). About 1 μmol/L of biotinylated BRD4 (44–460) and BRD4 (333–460) were injected at 5 µL/minutes flow rate into flow cell 2 (FC2) and 3 (FC3) of a SA-chip (Cytiva) in buffer containing 20 mmol/L HEPES pH 7.0, 150 mmol/L NaCl, 1 mmol/L TCEP, and 0.005% (v/v) P20 tween. The residual biotin binding sites on all FCs were blocked by injection of 50 μmol/L PEG-Biotin for 300 seconds at 5 µL/minutes (Thermofisher scientific).

Using single-cycle kinetics injection on the Biacore T200, compounds were injected at five different concentrations with a flow rate of 30 µL/minutes with association and dissociation times of 90 and 900 seconds, respectively. The buffer used for the compound titration was 20 mmol/L HEPES pH 7.0, 150 mmol/L NaCl, 1 mmol/L TCEP, 1% (v/v) DMSO, and 0.005% (v/v) Tween 20. SPR sensorgrams were analyzed by Biacore Evaluation using double referencing methods. After DMSO solvent correction, the resulting sensorgram was fit to 1:1 kinetic model using Biacore Insight Evaluation software to determine k_on, k_off, and K_D of the interactions.

The cell panel proliferation screen was performed at Crown Bioscience as part of their OmniScreen platform. Cells were treated with increasing concentrations of PLX-3618 for 72 hours, after which cell viability was monitored using Cell-Titer Glo reagent. Acute myeloid leukemia (AML) focused cell proliferation assays were performed in 96 well plates (Corning), starting with 500 to 2,000 cells per well. Compound was added using a Tecan digital dispenser and a day 0 CellTiter Glo reading was measured for reference. After a 72-hour incubation period, cell viability was measured using Cell-Titer Glo according to manufacturer’s specifications. Values were plotted relative to day 0 and negative controls (DMSO), and resulting curves were fitted using a four-parameter nonlinear model to determine IC₅₀ values. Apoptosis was monitored in replicate plates using Caspase-Glo 3/7 reagent (Promega) and normalized to CellTiter Glo values. Data were plotted using GraphPad Prism as for the proliferation assays.

Experiments were conducted at ReachBio Research Labs. In brief, serial dilutions of test compounds were assessed for their effect on clonogenic properties of human megakaryocyte progenitors in a semi-solid, collagen-based matrix for 14 days. Following the treatment period, cultures were fixed using methanol/acetone and stained using an anti-human CD41 antibody and an alkaline phosphate detection system according to manufacturers’ instructions. The colonies were assessed microscopically, scored by trained personnel, and values obtained used to generate dose-response curves. Resulting IC₅₀s were divided by each compound’s respective antiproliferative IC₅₀ for MV-4-11 tumor cells. Values >1 signify greater potency toward the tumor cells versus megakaryocyte progenitors.

All animal studies were carried out in accordance with the guidelines established by the Institutional Animal Care and Use Committee at Explora BioLabs (ACUP# EB17-010-074). NOD/SCID female mice (4-week-old) were purchased from Charles River Laboratories and implanted with 1 × 10⁷ MV-4-11 tumor cells in 100 µL serum-free media into their right flank for tumor development. When the mean tumor size reached approximately 200 mm³, the mice were randomized and size-matched into vehicle and treatment groups (eight animals/group). Tumor size was measured in length and width with a caliper twice a week. The tumor volume was calculated by the formula L × W × W/2 according to NCI standards. Body weights were collected prior to study start and twice a week during the study. PLX-3618 was formulated in 40% Captisol in ultrapure water and vortexed until a clear solution was formed. Statistical analysis of difference in tumor volume among the groups were conducted on the data obtained at the last day of treatment and subsequently evaluated using the one-way ANOVA, no matching and corrected for multiple comparisons using Dunnett’s t test (equal variance assumed). All data were analyzed using GraphPad Prism, where P < 0.05 was considered as statistically significant.

Raw data for proteomic studies were generated at Sanford Burnham Prebys Medical Discovery Institute (MassIVE MSV000094973). Raw data for SPR and TR-FRET studies were generated at ZoBio. Processed data for these studies and raw data for all others are available from the corresponding author upon request.

In an effort to discover direct degraders of BRD4, we conducted an uHTS using our proprietary picowell platform, which combines target-specific combinatorial bead-based libraries with a cell-based phenotypic readout of POI degradation (19). HEK293T cells were exposed to a BRD4 ligand focused library for 24 hours, after which BRD4 protein levels were assessed via immunofluorescent imaging. Consistent with the rarity of discovering monovalent degraders, our screening efforts yielded a low percentage (<1%) of robust hits. The hits were confirmed via orthogonal assays (e.g., Western blot) and then further optimized for potency, selectivity, and ADME properties, ultimately resulting in the drug-like degrader, PLX-3618 (Fig. 1A; ref. (14), compound P-4-d). PLX-3618 elicits potent (DC₅₀ = 12.2 nmol/L; Fig. 1B) and rapid (Supplementary Fig. S1) degradation of BRD4 in endogenously HiBiT-tagged BRD4 HEK293T cells. The core BRD4-binding element of PLX-3618 is derived from the pan-BET inhibitor JQ1 (12), and like JQ1, PLX-3618 exhibited selective bromodomain (BD) binding to the extraterminal (BET) family of proteins (Fig. 1C), with near equal potency to BD1 and BD2 of BRD4 (Supplementary Fig. S2). However, despite similar binding affinity to BRD4, BRD2, and BRD3, PLX-3618 was a selective degrader of only BRD4, as exemplified by both Western blot and proteomic analyses (Fig. 1D and E; respectively). This selective degradation activity was in stark contrast to the JQ1 and cereblon-based PROTAC, dBET1, which induced nonselective degradation of all three BET family members (Fig. 1D).

The degradation mechanism was explored using specific inhibitors of the UPS. Treatment of cells with the potent proteasome inhibitor, bortezomib, blocked PLX-3618-induced degradation of BRD4, which confirmed that cellular degradation was mediated via the UPS (Fig. 1B). Additionally, the complete blockage of degradation activity by the addition of the neddylation inhibitor, MLN4924, indicated that ubiquitination of BRD4 was mediated via an E3 ligase complex belonging to the cullin-RING family (Fig. 1B; ref. 13). To further explore the mechanism of degradation, cells treated with 1 μmol/L PLX-3618 for 6 hours ± 0.1 μmol/L bortezomib were lysed and subjected to ubiquitin IP. When the IP samples were blotted for BRD4, PLX-3618 treatment with bortezomib clearly led to an enrichment of BRD4 within the total pool of ubiquitinated proteins (Fig. 1F), demonstrating that PLX-3618 induces its ubiquitination. Collectively, these data demonstrate that PLX-3618, a nonselective binder to the BET family members, selectively degrades BRD4 via the proteasome.

Early characterization of the BRD4 degradation mechanism induced by PLX-3618 indicated the involvement of the 26S proteasome and more specifically, a cullin-RING ligase (CRL) complex (Fig. 1B). However, since the screen was conducted in cells where the whole repertoire of E3 ligases is available to potentially facilitate compound-induced interactions, the specific E3 ligase complex that mediated ubiquitination and subsequent degradation of BRD4 was unknown. To identify the responsible ligase complex, a focused CRISPR screen targeting 1,159 genes involved in the UPS pathway was conducted. Single-gene KO cells were incubated with 1 μmol/L PLX-3618 for 24 hours and subsequently fixed and stained for BRD4 protein, resulting in the identification of 27 specific gene KO events that had significantly impaired BRD4 degradation activity (Fig. 2A; Supplementary Table S1). Contained within the hits were genes encoding subunits of a specific CRL complex including the substrate receptor, DCAF11, as well as CUL4B, UBE2G1, NEDD8, and RBX1. (Fig. 2A; Supplementary Table S1). Proteins known to regulate CRL function, such as the signalosome proteins CAND1 and COPS5 and subunits of the 26S proteasome, were also identified (Fig. 2A; Supplementary Table S1; refs. 13, 20).

To confirm the involvement of DCAF11 in PLX-3618-induced degradation of BRD4, DCAF11-KO HEK293 cells were generated via CRISPR and used to monitor BRD4 degradation upon exposure to PLX-3618. DCAF11 KO pools treated with PLX-3618 had significantly reduced degradation of BRD4 compared to the parental line, confirming the role of DCAF11 (Fig. 2B). Studies to determine if PLX-3618 induced a protein–protein interaction between BRD4 and DCAF11 were pursued in both biochemical and cell-based contexts. Cellular IP experiments, using a truncated protein construct containing the first and second bromodomains of BRD4 (BD1-BD2-FLAG) and HA-tagged DCAF11, confirmed that PLX-3618 mediated a functional interaction between BRD4 and the DCAF11 CRL (Fig. 2C). Further, TR-FRET assays using a tandem bromodomain construct of BRD4 (BD1-BD2) and a purified DCAF11:DDB1 protein complex showed clear induction of protein–protein interactions in the presence of increasing concentrations of PLX-3618 (Fig. 2D). In contrast, the bromodomain based inhibitor, JQ1, was not able to significantly induce a similar interaction (Fig. 2D). Finally, cell-based NanoBRET assays were utilized to interrogate the selectivity of interaction between DCAF11 and BRD4, and the highly related proteins BRD2 and BRD3. Consistent with the BRD4-selective degradation observed (Fig. 1F), PLX-3618 induced a ternary complex with DCAF11 and BRD4, but not with BRD2 or BRD3 (Fig. 2E), suggesting that selective degradation arises from selective complex formation. Together, our studies have identified and confirmed the involvement of DCAF11 in the PLX-3618-mediated degradation mechanism of BRD4.

BRD4 contains two tandem bromodomains (BD1 and BD2) and binding studies demonstrate that PLX-3618 binds to both bromodomains with nearly equipotent affinity, similar to what is reported for JQ1 (Fig. 1C; Supplementary Fig. S2; ref. 12). To assess binding stoichiometry, SPR experiments were conducted using PLX-3618 with either the isolated BD2 domain of BRD4 or the tandem bromodomain protein construct, BRD4 (BD1-BD2). Data generated using isolated BD2 protein clearly demonstrated that PLX-3618 bound with a 1:1 stoichiometry (Fig. 3A). In contrast, when BRD4 (BD1-BD2) was used, the resulting sensorgram closely matched theoretical response levels that are consistent with a 1:2 protein:ligand stoichiometry (Fig. 3B), suggesting that two molecules of PLX-3618 are binding to BRD4 (BD1-BD2). The degradation mechanism of PLX-3618 acts through the recruitment of DCAF11 to BRD4, so we also tested the ability of PLX-3618 to bind directly to DCAF11. PLX-3618 did not exhibit appreciable binding to DCAF11 in SPR-binding assays (>5 μmol/L), differentiating it from heterobifunctional degraders that possess high affinity to both the protein-of-interest and the E3 ligase.

A series of experiments using bromodomain-selective inhibitors and point mutants was conducted to elucidate the domain dependency for PLX-3618-mediated degradation of BRD4. First, the selectivity of the tool inhibitors GSK778 (BD1 selective) and GSK046 (BD2 selective) was verified (21). In isolated bromodomain cellular target engagement assays, PLX-3618 bound to both BD1 and BD2 with similar potency (IC₅₀ = 10 and 30 nmol/L; respectively), whereas the binding selectivity of GSK778 to BD1 (BD1 IC₅₀ = 320 nmol/L; BD2 > 10 μmol/L) and GSK046 to BD2 (BD2 IC₅₀ = 320 nmol/L; BD1 > 10 μmol/L) was confirmed (Fig. 3C). These BD-selective compounds were utilized to compete for PLX-3618 binding to either bromodomain and surprisingly, both inhibitors were effective at blocking PLX-3618-mediated degradation of BRD4 (Fig. 3D). To rule out nonspecific binding activity of the tool compounds, BD1-BD2-NanoLuc protein constructs were engineered harboring bromodomain point mutations known to abrogate acetylated-lysine based binding of inhibitors (22) for use in degradation assays. PLX-3618 induced degradation of BRD4 with the introduction of a BD1 or BD2 point mutation was drastically diminished compared to wild type (WT), recapitulating the effects observed with selective competition (Fig. 3E). In contrast to PLX-3618, dBET1 activity in both BD1 and BD2 mutant constructs was similar to wild type (Fig. 3E). Both PLX-3618 and dBET1 degradation activity was completely abolished when both bromodomains were mutated, indicating productive binding is required for degradation (Fig. 3E). The bromodomain mutants were then utilized to assess ternary complex formation between BRD4 and DCAF11. Consistent with the degradation results, none of the bromodomain mutant constructs were able to support productive ternary complex formation with PLX-3618 (Fig. 3F). In contrast, both single bromodomain mutants supported dBET1-mediated ternary complex formation with cereblon, indicating that binding to only one bromodomain was necessary for heterobifunctional induced degradation (Fig. 3G). Taken together, the data suggest that both bromodomains of BRD4 need to be occupied by PLX-3618 for effective degradation by DCAF11.

It is well established that BRD4 can contribute to tumorigenesis due to its function in regulating chromatin stability and the expression of potent oncogenic drivers, such as c-MYC, at super enhancer regions (23–26). Several reports have shown that inhibition or degradation of BRD4 can potently inhibit proliferation of various tumor cell lines in vitro (25, 27–33). The antiproliferative activity of PLX-3618 was profiled across 110 different solid and hematopoietic tumor cell lines to identify tumor types that were sensitive to PLX-3618-mediated degradation of BRD4. Incubation for 72 hours with PLX-3618 elicited potent and differential antitumor activity in vitro (Fig. 4A). Cell lines grouped according to tumor type illustrated the differential sensitivity to BRD4 degradation, with hematopoietic, breast, and prostate lines exhibiting the lowest IC₅₀ values (Fig. 4A, top). Interestingly, this selective tumor sensitivity profile was similar to that reported for a BD2-selective inhibitor (34, 35). In addition, many cell lines had viability readings below the day 0 reading, indicating cell death (Fig. 4A, bottom). Cell lines that displayed both lower than average IC₅₀ values as well as cell death were deemed sensitive. Consistent with previous studies focused on BRD4 inhibition/degradation, PLX-3618 showed the most pronounced activity in hematopoietic cell lines, which are often driven by translocations, amplification, and overexpression of c-MYC (Fig. 4A, dashed box; refs. 27, 28, 30, 35–37). Apoptosis was confirmed in the AML model, MV-4-11, where PLX-3618 induced strong induction of caspase activity at and above degradation DC₉₀ concentrations (Fig. 4B). We compared PLX-3618 activity to a clinical-stage BRD4 inhibitor, CPI-0610 (38, 39), and found the antiproliferative and proapoptotic activity of PLX-3618 to be significantly more potent (Fig. 4B), despite the reported binding affinity of CPI-0610 [39 nmol/L (39)] being similar to PLX-3618. This finding is consistent with earlier studies comparing the BRD4 degrader, dBET1, to JQ1 (27). The observation of increased potency over a leading BRD4 inhibitor was expanded to include a small panel of AML lines. In all lines tested, PLX-3618 was significantly more potent than CPI-0610 (Fig. 4C; Supplementary Table S2), suggesting that degradation of BRD4, as opposed to inhibition, may have a more significant effect on tumor cell growth.

Due to the observed in vitro potency of PLX-3618 in hematopoietic cell lines, antitumor activity of PLX-3618 was explored in vivo using the AML MV-4-11 tumor model. Mouse pharmacokinetic (PK) studies demonstrated that PLX-3618 had moderately high clearance rates and low oral bioavailability (Cl = 39 mL/minutes/kg; oral bioavailability <5%), so intraperitoneal dosing regimens were pursued. NOD-SCID mice, bearing subcutaneous MV-4-11 tumors, were exposed to an acute IP dose of PLX-3618 at 5 mg/kg to assess the pharmacodynamic (PD) response. A single dose of PLX-3618 resulted in complete degradation of BRD4 in MV-4-11 tumors, and >50% loss of BRD4 protein was sustained for greater than 24 hours (Fig. 5A, gray bars). PK/PD relationships showed that substantial degradation of BRD4 was maintained even after plasma levels of PLX-3618 had dropped due to systemic clearance (Fig. 5A). This observation can be explained by fast degradation kinetics and slow BRD4 protein resynthesis rates [T_1/2 ∼ 18 hours (33)].

Given the rapid and complete degradation of BRD4 upon administration of PLX-3618, compound effect on tumor growth inhibition in vivo was tested and compared to the pan-BETi, CPI-0610. Animals treated with either 5 or 10 mg/kg QD PLX-3618 experienced significant tumor regression, without dose-limiting toxicity (Fig. 5B–D). A dose-dependent increase in tumor growth inhibition was observed with greater tumor regression detected upon administration of 10 mg/kg PLX-3618; seven out of eight animals at this dose group had no measurable tumor by day 14 (Fig. 5C). In comparison, BRD4 inhibition using CPI-0610 at 60 mg/kg resulted only in tumor growth inhibition with no regression, similar to earlier studies describing CPI-0610 activity (Fig. 5C; ref. 39). Attempts to elicit tumor regression with CPI-0610 by increasing the dose to 120 mg/kg resulted in partial tumor regression, but overall was not tolerated with four out of eight mice needing to be euthanized due to severe body weight loss (Fig. 5C and D).

PLX-3618 was compared to CPI-0610 in human megakaryocyte progenitor colony formation assays to assess potential thrombocytopenia risks, which is a known liability in developing pan-BET inhibitors (34). CPI-0610 potently inhibited the proliferation of megakaryocyte progenitor cells at concentrations below its antiproliferative IC₅₀ for MV-4-11 cells in vitro. In contrast, PLX-3618 inhibited proliferation of megakaryocyte progenitors only at concentrations above its IC₅₀ for MV-4-11 cells (Supplementary Fig. S3), suggesting that antitumor efficacy may be achieved without significantly causing hematologic toxicity. Collectively, our in vivo studies demonstrate that PLX-3618-mediated BRD4 degradation via DCAF11 can elicit robust tumor regression in an AML tumor model with superior efficacy and tolerability to a pan-BET inhibitor.

The studies presented herein describe the development and subsequent characterization of the small-molecule degrader of BRD4, PLX-3618, which is an example of a new class of targeted protein “direct” degraders. Such molecules are designed to bind directly to a protein of interest and induce its degradation via the proteasome, without the inclusion of a high-affinity E3 ligase binder. PLX-3618 illustrates this concept in that the central JQ1-based core afforded strong binding affinity to BET family proteins, inducing ubiquitination and selective degradation of BRD4 with nanomolar potency. Co-incubation with proteasome or neddylation inhibitors rescued degradation, clearly demonstrating that PLX-3618-induced degradation of BRD4 was mediated via the ubiquitin proteasome pathway, and more specifically via a cullin-RING ligase complex. As PLX-3618 was designed in an E3-ligase agnostic manner, a CRISPR screen focused on UPS-related genes was conducted to identify proteins required for the degradation mechanism. Hits from the screen yielded a well-defined CRL complex containing the E3 ligase substrate receptor, DCAF11. Subsequent cellular and biochemical experiments confirmed that PLX-3618-induced an interaction between BRD4 and DCAF11. While relatively little is known about DCAF11 biology, several reports have linked it to regulating various epigenetic and cell cycle functions (40–42). Perhaps the activity of DCAF11 on chromatin-associated proteins could explain it being co-opted into the BRD4 degradation mechanism induced by PLX-3618.

Recent reports have described small molecules containing covalent warheads that can react with DCAF11. Some of these have been incorporated into heterobifunctional degraders of BRD4 (43–45). However, unlike these covalent molecules, PLX-3618 does not contain obvious electrophilic functionality, and we do not observe any direct interaction with DCAF11 when tested by SPR. Therefore, PLX-3618 induces BRD4:DCAF11 interactions via a unique mechanism. During the preparation of this manuscript, work was published utilizing one of our patented compounds that demonstrated the recruitment of DCAF11 for selective degradation of BRD4 (14, 46). In their work, the authors invoke a 1:1 binding stoichiometry, with the molecule bridging the two bromodomains contained within BRD4. This model is reminiscent of prior studies focused on bivalent BET inhibitors (47–49). In our work, a combination of SPR, competition, and mutational studies clearly demonstrated that PLX-3618 binds to BRD4 in a 1:2 protein-to-ligand stoichiometry, and that binding to both bromodomains is required for efficient interaction with, and degradation by, DCAF11. Taken together, our data suggest that two molecules of PLX-3618 are required to bind to BRD4 for full degradation activity mediated by DCAF11. It is possible that the presentation of two pyrazolopyrimidine moieties is required to create an interface that DCAF11 recognizes, or that the two bound molecules of PLX-3618 cooperate to induce a BRD4 conformation which leads to the specific recognition by DCAF11. The exact mechanism of ternary complex formation is still under investigation.

Given the potency and selectivity of BRD4 degradation, PLX-3618 was advanced to in vitro and in vivo studies to monitor inhibition of tumor cell growth. In vitro data demonstrated potent and differential antitumor activity, with pronounced activity in hematopoietic cell lines. In vivo tumor growth inhibition studies using subcutaneous AML tumors derived from MV-4-11 cells demonstrated robust and well-tolerated tumor regression with IP administration of PLX-3618. Degradation of BRD4 by PLX-3618 led to superior tumor growth inhibition relative to the clinical stage pan-BET inhibitor, CPI-0610, consistent with earlier studies comparing BRD4 degradation versus inhibition (27). BRD4 inhibition has been shown to inhibit the expression of super-enhancer associated oncogenic drivers of AML (25, 28). Perhaps the complete depletion of BRD4 protein via targeted protein degradation leads to sustained downregulation of these oncogenic drivers, thus explaining the observed enhanced activity (50). The sustained activity can also be explained in part due to the unique exposure response relationship of small molecule degraders. PK/PD studies demonstrate that despite relatively rapid clearance of PLX-3618 resulting in plasma drug levels that are below the measured DC₅₀, the kinetics of degradation coupled with slower protein resynthesis rates sustain BRD4 depletion. In these studies, the BRD4 protein only returns to 50% levels at about 24 hours post-dose. This unique profile may also be in part responsible for the differences in tolerability when compared to CPI-0610, a pan-BET inhibitor. Unlike protein degraders, reversible inhibitors generally require consistent target coverage to achieve their effect. The intolerability observed with CPI-0610 at the higher dose could reflect nonselective target coverage of BRD4, BRD2, and BRD3. This suggests that the unique PK/PD profile which is possible from protein degradation could be an attractive option for minimizing off-target liabilities.

PLX-3618 represents a novel class of direct degraders, in which the recruitment of the DCAF11 CRL promotes the efficient and selective degradation of BRD4. The discovery of BRD4 direct degraders took advantage of our E3 ligase-agnostic ultra-high throughput cell-based screening methodology where ligands designed to bind to the target protein were exposed to the whole cellular repertoire of E3 ligases. This enabled the discovery and development of potent and selective direct degraders without being restricted to a particular E3 ligase. This E3 ligase-agnostic approach led to the discovery of the E3 ligase DCAF11, which we demonstrate is amenable to selective targeted protein degradation, thus expanding the E3 ligase toolbox for TPD development beyond cereblon and VHL. The insights gained from this discovery can be applied to the development of the next generation of targeted protein direct degraders, focused on selectively eliminating disease-causing proteins.

G.S. Parker reports personal fees from Plexium outside the submitted work. G. Blanco reports personal fees from Plexium outside the submitted work. A. Jamborcic reports personal fees from Plexium during the conduct of the study. A. Dearie reports personal fees from Plexium outside the submitted work. G. Leriche reports a patent for WO2022/221786 pending. P.A. Thompson reports personal fees from Plexium outside the submitted work. No disclosures were reported by the other authors.

G.S. Parker: Conceptualization, formal analysis, supervision, methodology, writing–original draft. J.I. Toth: Conceptualization, formal analysis, supervision, investigation, methodology, writing-review and editing. S. Fish: Investigation. G. Blanco: Investigation. T. Kampert: Investigation. X. Li: Formal analysis, supervision, investigation, methodology. L. Yang: Investigation. C.R. Stumpf: Supervision, investigation, writing–review and editing. K. Steadman: Supervision, investigation. A. Jamborcic: Investigation. S. Chien: Investigation. E. Daniele: Investigation. A. Dearie: Investigation. G. Leriche: Conceptualization, supervision, project administration. S. Bailey: Conceptualization, formal analysis, supervision, writing–review and editing. P.A. Thompson: Conceptualization, formal analysis, supervision, project administration, writing–review and editing.

We thank the entire Plexium team for their helpful scientific discussions. We thank Kenneth Chng, Erika Green, Michael Hocker, Elliot Imler, and Yi Zhang for the development of the screening platform. We acknowledge the members of ZoBio for their assistance executing surface plasmon resonance and TR-FRET experiments. We acknowledge Alex Campos from Sanford Burnham Prebys Medical Discovery Institute for his proteomic analyses. We also thank Kevin Freeman-Cook for many helpful discussions regarding this manuscript.