Abstract
Ligand-dependent protein degradation has emerged as a compelling strategy to pharmacologically control the protein content of cells. So far, only a limited number of E3 ligases have been found to support this process. Here, we use a chemical proteomic strategy to discover that DCAF16 – a poorly characterized substrate recognition component of CUL4-DDB1 E3 ubiquitin ligases – promotes nuclear-restricted protein degradation upon modification by cysteine-directed heterobifunctional electrophilic compounds.
Conventional small-molecule probes and drugs act by directly perturbing the functions of proteins (e.g., blocking enzyme catalysis or antagonizing receptor signaling). Many proteins, however, possess multiple functional domains, and therefore a compound that binds to only one of these domains may fail to fully inactivate the protein. An alternative emerging strategy uses chemical probes that direct proteins to the proteolytic degradation machinery of the cell, leading to the loss of protein expression1, 2. This targeted protein degradation approach leverages two types of small molecules – those that form tripartite complexes with specific E3 ubiquitin ligases and neosubstrate proteins (e.g., the IMiD class of therapeutics3, 4) and bifunctional compounds, often referred to as PROTACs (proteolysis targeting chimeras), which couple E3 ligase ligands to substrate ligands via a variably structured linker5-7. Targeted protein degradation has the potential to act in a catalytic manner8 that may lower the drug concentrations required to produce a pharmacological effect and operate sub-stoichiometrically to avoid antagonizing the natural functions of the participating E3 ligase. Despite these advantages, to date, only a handful of the 600+ human E3 ligases have been found to support ligand-mediated protein degradation5-7, 9-11. Importantly, these E3 ligases has been found to show distinct and restricted substrate specificities12, 13, a parameter that is not yet predictable or easily controlled, underscoring the need to discover additional ligandable E3 ligases with differentiated properties to realize the full scope of targeted protein degradation as a pharmacological strategy.
We have recently introduced chemical proteomic platforms for globally and site-specifically mapping the reactivity of electrophilic small molecules in native biological systems14-16. These experiments have identified a subset of fragment electrophiles, referred to hereafter as ‘scouts’15, that capture a remarkably broad fraction of the 100s-1000s of covalent small molecule-cysteine interactions discovered so far by chemical proteomics. We surmised herein that these scout fragments, when incorporated into heterotypic bifunctional compounds containing optimized protein ligands, may offer an expedited path to discover E3 ligases capable of supporting targeted protein degradation through covalent reactivity with cysteine residues. With this goal in mind, we fused three scout fragments – KB02, KB03, and KB05 – to the SLF ligand that binds tightly and selectively to FKBP12 (Fig. 1a), a cytosolic prolyl isomerase that has been frequently used to study ligand-induced protein degradation17, 18. We then stably expressed FLAG-tagged variants of FKBP12 (FLAG-FKBP12) or FKBP12 with a C-terminal nuclear localization sequence (FLAG-FKBP12_NLS) in HEK293T cells to provide cell models for evaluating cytosolic-and nuclear-localized E3-mediated degradation pathways, respectively (Fig. 1b and Supplementary Fig. 1a).
We first confirmed by anti-FLAG western blotting that both cytosolic and nuclear FKBP12 are degraded by lenalidomide-SLF (Supplementary Fig. 1b, c), a bifunctional molecule comprised of SLF coupled to lenalidomide, a ligand for the E3 ligase cereblon (CRBN)17, 18. We next evaluated the electrophilic scout fragment-SLF bifunctional compounds for effects on FKBP12 degradation in HEK293T cells. Under the treatment conditions (2 μM, 8 or 24 h), none of the compounds altered cytosolic FKBP12 (Fig. 1b). On the other hand, the KB02-SLF compound promoted a substantial reduction in nuclear FKBP12 (Fig. 1b) that was sustained across a 4-72 h time frame (Supplementary Fig. 2a). Cell imaging studies confirmed the selective loss of nuclear-localized FKBP12 in KB02-SLF-treated cells (Fig. 1c). These imaging studies, along with western blotting experiments (Supplementary Fig. 1a), pointed to a fraction of FKBP12 that remained cytosolically localized in FLAG-FKBP12_NLS-transfected cells and was consequently unaffected by KB02-SLF treatment (Fig. 1c).
Reductions in FKBP12_NLS were not observed with an analogue of KB02-SLF where the electrophilic alpha-chloroacetamide group was replaced with an unreactive propanamide (C-KB02-SLF; Fig. 1d, e), indicating that the mechanism of action of KB02-SLF involved covalent modification of one or more proteins. KB02-SLF, as well as analogues of this compound with different linker lengths (Supplementary Fig. 2b), promoted the loss of FKBP12_NLS across a concentration of ~0.5-5 μM, but showed varied reductions in activity at higher concentrations (Fig. 1f and Supplementary Fig. 2b). Such parabolic concentration-dependence is a feature of PROTACs, where binary complexes gain prevalence over ternary complexes at higher cellular concentrations of compound19. Consistent with this mode of action, we did not observed reductions in FKBP12_NLS in cells treated with the two separate components (KB02 and SLF) of the KB02-SLF bifunctional compound (Supplementary Fig. 2c). Pre-treatment with SLF did, on the other hand, block the KB02-SLF-induced loss of FKBP12_NLS (Supplementary Fig. 2d). We also found that KB02-SLF induced polyubiquitination of nuclear FKBP12_NLS (Fig. 2a and Supplementary Fig. 3), but not cytosolic FKBP12 (Fig. 2a) and KB02-SLF-mediated loss of FKBP12_NLS was blocked by the proteasome inhibitor MG132 (Fig. 2b) and neddylation inhibitor MLN492420 (Fig. 2c). Similar effects of KB02-SLF on FKBP12_NLS were observed in a second human cell line (MDA-MB-231) (Supplementary Fig. 4). These data, taken together, supported that KB02-SLF promoted the proteasomal degradation of nuclear-localized FKBP12 via the action of a Cullin-RING ubiquitin ligase (CRL).
Working under a model where KB02-SLF induced the degradation of FKBP12_NLS by covalently reacting with one or more CRL(s), we first attempted to identify the recruited CRL(s) using a chemical proteomic method termed isoTOP-ABPP (isotopic tandem orthogonal proteolysis-activity-based protein profiling), where ligand-engaged cysteines in cells are mapped by competitive blockade of reactivity with an iodoacetamide (IA)-alkyne probe21. However, we did not observe any cysteines on CRL(s) that were strongly engaged (> 75%) by KB02-SLF (10 μM, 2 h) in HEK293T cells by isoTOP-ABPP (Supplementary Table 1). We alternatively considered that KB02-SLF may show low stoichiometry engagement of a CRL, which would then function catalytically to drive FKBP12_NLS degradation. Such low fractional occupancy ligand-protein interactions are not easily mapped by competitive profiling methods like isoTOP-ABPP, as they would minimally perturb IA-alkyne-cysteine reactions. We therefore employed an alternative proteomic approach, in which FLAG-mediated affinity enrichment was used to identify proteins that associated with FKBP12_NLS in a KB02-SLF-dependent manner (Fig. 2d). This experimental set up identified two substrate receptor components of CRLs – DCAF16 and DTL – that were substantially enriched (> 5-fold) in HEK293T cells treated with KB02-SLF compared to DMSO-treated control cells (Fig. 2e, Supplementary Fig. 5a and Supplementary Table 2). Neither DCAF16 nor DTL were enriched by KB02-SLF in an additional control experiment performed with mock-transfected cells lacking FKBP12_NLS (Supplementary Fig. 5a, b and Supplementary Table 2). We also observed KB02-SLF-dependent enrichment of DCAF16 in MDA-MB-231 cells (Supplementary Fig. 4d, e and Supplementary Table 2).
Notably, both DCAF16 and DTL are predicted to be nuclear proteins (https://psort.hgc.jp/). DTL (also known as CDT2 and DCAF2) plays a key role in cell cycle control and DNA damage response22. DCAF16, on the other hand, remains poorly characterized and shares negligible homology with other DDB1-and CUL4-associated factor (DCAF) proteins that participate in CRLs23. We confirmed the nuclear localization of an HA-tagged DCAF16 in transfected HEK293T cells (Fig. 2f). Consistent with KB02-SLF engaging functional CRL complexes potentially containing DCAF16 and/or DTL, we also observed enrichment of additional CRL components – DDB1 and CUL4B – in KB02-SLF-treated cells (Fig. 2e and Supplementary Fig. 5a). We next used small hairpin RNA (shRNA)-mediated knockdown to find that reductions in DCAF16, but not DTL, substantially prevented KB02-SLF-mediated degradation of FKBP12_NLS (Supplementary Fig. 5c, d). shRNA-knockdown of DCAF16 also blocked KB02-SLF-induced polyubiquitination of FKBP12_NLS (Supplementary Fig. 5e). We further verified the involvement of DCAF16 in KB02-SLF-mediated degradation of FKBP12_NLS by CRISPR/Cas9 genetic knockout24. We used genomic sequencing and mass spectrometry-based proteomics to confirm the genetic disruption of DCAF16 in three independent clones (DCAF16-/-) compared to three wild type DCAF16 clones (DCAF16+/+) (Supplementary Fig. 6a-c and Supplementary Table 3). KB02-SLF supported the degradation of FKBP12_NLS in all three DCAF16+/+ clones, but not in any of the three DCAF16-/- clones (Fig. 3a and Supplementary Fig. 6d). Moreover, recombinant expression of HA-tagged DCAF16 restored KB02-SLF-induced FKBP12_NLS degradation in DCAF16-/- cells (Fig. 3b) and further enhanced the extent of KB02-SLF-induced FKBP12_NLS degradation in DCAF16+/+ cells (Supplementary Fig. 6e). We did not observe the degradation of HA-DCAF16 itself following treatment of cells with KB02-SLF (Fig. 3b and Supplementary Fig. 6e, f). These data, taken together, demonstrate that the degradation of FKBP12_NLS induced by KB02-SLF is mediated by DCAF16.
In further support of ternary complex formation involving a DCAF16-CRL, we found that HA-DCAF16 and DDB1 co-immunoprecipitated with FLAG-FKBP12_NLS in the presence of KB02-SLF (Fig. 3c) or its linker analogues (Supplementary Fig. 7a). In the course of performing these co-immunoprecipitation experiments, we noticed that KB02-SLF-treated cells showed a higher molecular weight (HMW) form of HA-DCAF16, consistent with covalent modification of this protein by KB02-SLF (Fig. 3c and Supplementary Fig. 7a-f). Only a modest fraction of HA-DCAF16 was converted to this HMW in the presence of increasing concentrations of KB02-SLF (Fig. 3c and Supplementary Fig. 7b). Notably, however, co-immunoprecipitations with FKBP12_NLS exclusively pulled down the HMW, but not LMW form of DCAF16 (Fig. 3c and Supplementary Fig. 7), supporting a ternary complex model where KB02-SLF is covalently and non-covalently bound to DCAF16 and FKBP12_NLS, respectively. Also consistent with this model, DCAF16 was not co-immunoprecipitated from cells treated with the non-electrophilic control compound C-KB02-SLF (Supplementary Fig. 7c) and minimally enriched from cells treated with the other two electrophilic bifunctional compounds (KB03-SLF and KB05-SLF) that did not support FKBP12_NLS degradation (Supplementary Fig. 7d, e).
DCAF16 is a 216 aa protein that is highly conserved across mammals (e.g., human and rabbit DCAF16 share 97% identity), but absent in rodents. The DCAF16 protein has eight cysteine residues, including a cluster of four cysteines between amino acids (aa) 173-179 (Fig. 3d). We next attempted to map KB02-PEG0-SLF-reactive cysteine(s) in DCAF16 by treating HA-DCAF16-transfected HEK293T cells with DMSO or KB02-PEG0-SLF (10 μM, 2 h), affinity purifying HA-DCAF16, and subjecting the protein to trypsin digestion and LC-MS/MS analysis (Fig. 3d). We searched the MS1 (parent ion) profiles for the m/z values of unmodified and KB02-PEG0-SLF-modified DCAF16 tryptic peptides and identified a KB02-PEG0-SLF-modified form of the tryptic peptide (aa 168-184) containing C173 and C177-179 (Supplementary Fig. 8a). In contrast, we did not identify KB02-PEG0-SLF-modified forms for tryptic peptides containing C100/C103 (aa 97-106) or C119 (aa 107-133) (Supplementary Fig. 8b, c). We did not detect the tryptic peptide containing C58, likely because it is a very small in length (four aa). Tandem (MS/MS) analysis of the KB02-PEG0-SLF-modified DCAF16 peptide indicated that the most likely modified residue(s) was C177, C178, and/or C179 (Fig. 3e). Our efforts to evaluate the contribution of C177-179 and other cysteine residues to KB02-SLF-induced degradation of FKBP12_NLS by site-directed mutagenesis were hindered, in part, by the dramatic reduction in expression of cysteine-to-serine mutants for multiple residues (e.g., C177, C179) (Supplementary Fig. 9a). These mutagenesis studies did reveal that the C58S-, C173S-, and C178S-DCAF16 mutants were expressed at near-wild type levels and supported KB02-SLF-induced degradation of FKBP12_NLS (Supplementary Fig. 9a, b), indicating that C58, C173, and C178 are unlikely to constitute primary sites of engagement for KB02-SLF. We additionally analyzed whether cysteine-to-serine mutants of DCAF16 could form a ternary complex with KB02-SLF and FKBP12_NLS in cells pre-treated with the proteasome inhibitor MG132, which we found to normalize DCAF mutant protein expression levels (Supplementary Fig. 9c). Among the five mutants tested, C58S-, C173S- and C178S-DCAF16, but not C177S- and C179S-DCAF16, co-immunoprecipitated with FKBP12_NLS in KB02-SLF-treated cells (Supplementary Fig. 9c). While our data, taken in total, support C177 and/or C179 in DCAF16 as possible sites of modification responsible for KB02-SLF-induced degradation of FKBP12_NLS, we should qualify this interpretation by noting the potential for improper folding of the C177S- and C179S-DCAF16 mutants as a confounding factor.
We finally investigated whether covalent modification of DCAF16 could support the degradation of another nuclear protein. For these studies, we selected BRD4 as the nuclear protein, as it has a potent and selective ligand JQ1 that has been successfully coupled to other E3 ligands to promote degradation7, 17, 25, 26. We found that a KB02-JQ1 bifunctional compound (Fig. 3f) promoted, in a concentration-dependent manner, the degradation of BRD4 in HEK293T cells (Fig. 3g), and this effect was blocked by MG132 or MLN4924 (Fig. 3h). BRD4 was not degraded in cells treated with either the two separate components (KB02 and JQ1) of KB02- JQ1 or the FKBP12-directed bifunctional compound KB02-SLF (Supplementary Fig. 10a). We noted that much higher concentrations of KB02-JQ1 (20-40 μM) were required to degrade BRD4 compared to the degradation of FKBP12_NLS by KB02-SLF (0.5-2 μM). We suspect that this apparent difference in potency may reflect differential cellular uptake of the two KB02 bifunctional compounds, as KB02-JQ1 also showed a rightward shift in cytotoxicity (IC50 > 50 μM) compared to KB02-SLF and the parent electrophilic fragment KB02 (IC50 values of 14-23 μM) (Supplementary Fig. 10b). We then assessed the proteome-wide effect of KB02-JQ1 treatment in HEK293T cells and found that the KB02-JQ1-induced degradation of BRD4 was selective, with only one additional protein – ACAT1 – showing similar reductions in KB02-JQ1-treated cells (Fig. 3i and Supplementary Table 4). Interestingly, ACAT1 harbors a highly KB02-reactive cysteine (C126)14, 15, which suggests that the direct engagement by KB02- containing compounds like KB02-JQ1 could lead to degradation of this protein. Supporting a functional role for DCAF16 in KB02-JQ1-induced BRD4 degradation, we found that the HMW form of DCAF16 co-immunoprecipitated with BRD4 in a KB02-JQ1-dependent manner (Fig. 3j) and that BRD4 degradation was substantially blocked in DCAF16-/- cells (Fig. 3k).
In summary, we have used a chemical proteomic strategy to discover that the E3 ligase subunit DCAF16 supports targeted protein degradation when engaged by electrophilic PROTACs. DCAF16 may offer advantages as a new addition to the emerging subset of E3 ligases that promote ligand-induced protein degradation. First, our data suggest that DCAF16 exclusively promotes the degradation of nuclear proteins, which may offer a way to improve the specificity of ligand-induced protein degradation by, for instance, avoiding cytosolic proteins that engage a bifunctional compound. Second, DCAF16 appears capable of supporting ligand-induced protein degradation at very low fractional engagement, which we attribute, at least in part, to covalent interaction with electrophilic PROTACs. By converting a sub-stoichiometric portion of DCAF16 to a ‘neo-E3 ligase’, electrophilic PROTACs may support the degradation of target proteins while minimally perturbing endogenous substrates of DCAF16, which could presumably be degraded by the still predominantly unmodified fraction of this E3 ligase in cells. Our findings also underscore the value of broadly reactive scout fragments as tools to discover ligandable and functional sites on poorly characterized proteins like DCAF16. We should note, however, that scout fragments, even when active as electrophilic PROTACs at sub-μM concentrations, still modify several other proteins in cells, and the optimization of ligands for DCAF16-dependent protein degradation therefore represents an important future objective. Here, structural information on DCAF16, in particular, adducted to covalent PROTACs, would be valuable, as our chemical proteomic data point to a site(s) of modification that contains several clustered cysteines. We finally wonder whether this feature indicates the potential for endogenous electrophiles and/or oxidative stress to modify DCAF16 and shape its substrate scope in vivo.
Author contributions
X.Z. and B.F.C. conceived of the research and wrote the paper. X.Z. developed methods, performed experiments, and analyzed data. X.Z. and M.M.D. analyzed chemical proteomic data. X.Z. designed and synthesized KB02-SLF, KB02-PEG0-SLF, KB02-PEG4-SLF, KB03-SLF, KB05-SLF and C-KB02-SLF. V.M.C. designed and synthesized KB02-JQ1. T.G.W. designed and synthesized lenalidomide-SLF. V.M.C. characterized all the compounds.
Competing Financial Interests
Dr. Cravatt is a founder and scientific advisor to Vividion Therapeutics, a biotechnology company interested in developing small-molecule therapeutics.
Acknowledgements
This work was supported by the NIH (CA087660 (B.F.C.), CA231991 (B.F.C.), CA212467 (M.M.D.), CA211526 (V.M.C.)) and the Damon-Runyon Cancer Research Foundation (X.Z. DRG-2341-18).