Abstract
Neomorphic mutations and drugs can elicit unanticipated effects that require mechanistic understanding to inform clinical practice. Recurrent indel mutations in the Kelch domain of the KBTBD4 E3 ligase rewire epigenetic programs for stemness in medulloblastoma by recruiting LSD1-CoREST-HDAC1/2 complexes as neo-substrates for ubiquitination and degradation. UM171, an investigational drug for haematopoietic stem cell transplantation, was found to degrade LSD1-CoREST-HDAC1/2 complexes in a wild-type KBTBD4-dependent manner, suggesting a potential common mode of action. Here, we identify that these neomorphic interactions are mediated by the HDAC deacetylase domain. Cryo-EM studies of both wild-type and mutant KBTBD4 capture 2:1 and 2:2 KBTBD4-HDAC2 complexes, as well as a 2:1:1 KBTBD4-HDAC2-CoREST1 complex, at resolutions spanning 2.7 to 3.3 Å. The mutant and drug-induced complexes adopt similar structural assemblies requiring both Kelch domains in the KBTBD4 dimer for each HDAC2 interaction. UM171 is identified as a bona fide molecular glue binding across the ternary interface. Most strikingly, the indel mutation reshapes the same surface of KBTBD4 providing an example of a natural mimic of a molecular glue. Together, the structures provide mechanistic understanding of neomorphic KBTBD4, while structure-activity relationship (SAR) analysis of UM171 reveals analog S234984 as a more potent molecular glue for future studies.
Introduction
The rapid growth of genomic and chemogenomic data has unveiled unexpected complexity in the mechanisms of action of disease-causing mutations and small molecule drugs. As well as inducing loss or gain of function, mutations and small molecules can impart neomorphic effects that yield a different protein function1,2,3,4,5,6. Mutations in isocitrate dehydrogenases (IDH1/2) drive tumour formation by inducing the neomorphic production of oncometabolite 2-hydroxyglutarate7. Conversely, immunomodulatory drugs (IMiDs) drive the neomorphic destruction of zinc finger proteins by establishing molecular glue interactions with cereblon, the substrate receptor of a cullin4-RING E3 ligase complex (CRL4CRBN)8,9,10. Neo-substrates of this IMiD-bound complex include SALL4, responsible for the teratogenicity of thalidomide, as well as IKZF1 and IKZF3, which account for the anti-tumour effects of IMiDs lenalidomide and pomalidomide8,9,10. Defining the molecular basis of neomorphic function is therefore critical for uncovering mechanisms of pathogenicity, as well as for establishing therapeutic approaches.
The promise of targeting “undruggable” proteins has led to a surge of interest in targeted protein degradation that has yielded the discovery of other CRL4-dependent molecular glues, including degraders of RBM39 that utilise the substrate receptor DCAF1511 and cyclin K degraders that utilise a drug-induced complex of DDB1 and CDK1212,13. The targeted protein degradation paradigm was recently expanded by the discovery of neomorphic mutations in an E3 ligase14. Kelch repeat and BTB domain-containing protein 4 (KBTBD4) is the substrate receptor of a cullin3-dependent E3 ligase complex (CRL3KBTBD4). Recurrent in-frame insertions in KBTBD4 have been reported as putative driver mutations in group 3 and group 4 medulloblastomas15, as well as in pineal parenchymal tumours with intermediate differentiation (PPTID)16. These cancer hotspot mutants were found to elicit neomorphic ubiquitination that promoted the degradation of the CoREST transcriptional repressor complex, as well as the increased stemness of medulloblastoma cancer cells14. All mutations clustered across just six amino acids within the KBTBD4 Kelch domain. The indel mutation p.R313delinsPRR (herein R313PRR) was particularly prevalent across both brain tumour types and demonstrated a strong neomorphic effect14,15,16,17.
Haematopoietic stem cell (HSC) transplants save the lives of cancer patients, but suffer from the limited number of progenitor cells in umbilical cord blood18. UM171 was discovered as a small molecule agonist of HSC expansion19 and shows promise as an ex-vivo drug treatment for HSC transplants in phase 2 clinical trials20,21. UM171 was found to induce degradation of the CoREST complex in a wild-type KBTBD4 and proteasome-dependent manner22,23,24, mirroring the neomorphic activity of mutant KBTBD414. Our understanding of these important neomorphic effects is limited by a lack of knowledge of the direct interactors of KBTBD4 and their molecular mechanisms. CoREST proteins function as a scaffold for assembly of a transcriptional repressor complex that contains CoREST1, 2 or 3, the lysine specific demethylase LSD1, histone deacetylase HDAC1 or 2 and various DNA-associated factors25,26. Here, we identify HDAC1/2 as the direct interaction partner of KBTBD4 through cellular, biophysical and structural investigations. Cryo-EM structures of wild-type and mutant KBTBD4-HDAC2 complexes confirm UM171 as a bona fide molecular glue and reveal structural mimicry between UM171 and mutant KBTBD4. These data reveal a shared mechanism of neo-substrate recruitment by an E3 cancer mutant and small molecule drug, as well as structural models to guide future drug development for cancer treatment.
Results
CoREST complex recruitment requires the full-length KBTBD4R313PRR dimer
BTB-Kelch family proteins are known to recruit their substrates through the Kelch domain. In agreement, the KBTBD4 Kelch domain harbours the cancer hotspot mutations that confer neo-substrate interaction. To test whether the Kelch domain alone was sufficient for KBTBD4 mutant binding to CoREST complex, we expressed FLAG-tagged KBTBD4R313PRR Kelch domain and full-length constructs and performed FLAG immunoprecipitations in HEK293 cells. Surprisingly, the Kelch domain failed to replicate the neo-substrate binding observed with full-length KBTBD4R313PRR, indicating that additional domains of KBTBD4 were necessary for binding (Fig. 1a).
Fig. 1: CoREST complex recruitment requires the full-length KBTBD4R313PRR dimer.
figure 1
a Representative immunoblots of lysates and anti-FLAG immunoprecipitates from HEK293 cells expressing FLAG-KBTBD4R313PRR full-length or Kelch domain protein. Immunoprecipitation was performed twice with similar results. Source data in this figure are provided in the Source Data file. b Size-exclusion chromatography showing oligomerization states of KBTBD4R313PRR and its monomeric variant (mono KBTBD4). c Representative immunoblots of lysates and anti-FLAG immunoprecipitates from HEK293 cells expressing FLAG-KBTBD4 variants as indicated. Immunoprecipitation was performed twice with similar results. Source data in this figure are provided in the Source Data file.
Full size image
BTB-Kelch proteins dimerize through the BTB domain to present two Kelch domains for substrate recruitment. Full-length KBTBD4R313PRR may therefore be necessary to establish bivalent substrate interactions involving both Kelch domains within the KBTBD4 dimer. To test this hypothesis, we engineered a monomeric form of the protein by introducing mutations predicted by previous work to disrupt the BTB dimer interface27, and confirmed its monomeric state using size-exclusion chromatography (Fig. 1b). FLAG immunoprecipitations were then performed comparing the monomeric FLAG-KBTBD4R313PRR construct to dimeric controls. Consistent with previous work14, the full-length KBTBD4R313PRR dimer was able to immunoprecipitate CoREST1, HDAC2 and LSD1 from the endogenous CoREST complex (Fig. 1c). By contrast, the monomeric KBTBD4R313PRR construct lost binding to all three proteins similarly to the KBTBD4WT dimer (Fig. 1c). Taken together, these results suggest that CoREST complex recruitment to KBTBD4 requires both a dimeric BTB domain and a Kelch domain harbouring a neomorphic mutation.
KBTBD4R313PRR is recruited to CoREST N-terminal domains
Our previous proteomics study identified all three CoREST isoforms binding to KBTBD4R313PRR, but not alternative LSD1 or HDAC1/2-associated scaffolding proteins, such as SIN3A or MTA128,29, suggesting that CoREST proteins may be important for KBTBD4 binding. Cryo-EM studies of the full-length LSD1-CoREST1-HDAC1 complex showed that the dynamic features of the complex limited the structure resolution and affected assay readouts25. We therefore reasoned that shorter CoREST1 constructs and subcomplexes may be desirable for the further characterization of the structural mechanisms of neomorphic KBTBD4.
The N and C-terminal regions of CoREST1 are known to bind to HDAC1/2 and LSD1, respectively. To map the minimal region required for KBTBD4R313PRR interaction, we cloned four fragments of CoREST1 designed to explore each terminus with different truncations (Fig. 2a). These HA-tagged constructs were co-transfected into HEK293 cells with KBTBD4R313PRR for HA co-immunoprecipitation. Strikingly, we observed KBTBD4R313PRR binding to the N-terminal CoREST1 fragments NT1 and NT2, but not to the C-terminal fragments CT1 and CT2 (Fig. 2b). The conformational integrity of the N and C-terminal fragments was confirmed by their binding to endogenous HDAC2 and LSD1, respectively (Fig. 2b). These domain mapping experiments demonstrate that the CoREST1 N-terminal region that assembles with HDAC1/2 is required for its recruitment to KBTBD4R313PRR.
Fig. 2: KBTBD4R313PRR is recruited to CoREST N-terminal domains.
figure 2
a Domain organization of CoREST1 indicating the construct designs used for interaction site mapping. b Representative immunoblots of lysates and anti-HA immunoprecipitates from HEK293 cells expressing HA-CoREST1 fragments, as well as FLAG-KBTBD4R313PRR as indicated. Immunoprecipitation was performed at least twice with similar results and source data are provided in the Source Data file. c For microscale thermophoresis (MST) experiments, unlabelled CoREST1NT2-HDAC2 or LSD1 was titrated at indicated concentrations into 40 nM fluorescently-labelled KBTBD4R313PRR. CoREST1NT2-HDAC2 bound to KBTBD4R313PRR with a dissociation constant (KD) of 0.75 µM. All MST experiments were performed in triplicates. Data points and error bars represent the means and standard deviations. Source data are provided in the Source Data file.
Full size image
We selected the smaller CoREST1NT2 fragment (a.a. 71-259) for in vitro binding studies as it faithfully spanned the ELM2 and SANT1 domains needed for HDAC1/2 interaction. A previous structure of the homologous MTA1-HDAC1 complex revealed that HDAC1 was enveloped by a flexible arm in the ELM2 domain that would likely be disordered on its own29. We, therefore, co-expressed the CoREST1NT2 fragment with full-length HDAC1 or 2 for recombinant protein production in Expi293FTM cells. The CoREST1NT2-HDAC2 complex showed a higher yield and was purified for affinity measurements using microscale thermophoresis (MST). The resulting CoREST1NT2-HDAC2 complex bound strongly to full-length KBTBD4R313PRR with a dissociation constant (KD) value of 0.75 µM (Fig. 2c), consistent with the cellular domain mapping. In contrast, LSD1 alone showed no binding (Fig. 2c), suggesting its co-precipitation with KBTBD4R313PRR in cells was mediated by the CoREST1-HDAC2 subcomplex and not through its direct interaction.
CoREST1NT2-HDAC2 is a common binding entity for KBTBD4R313PRR and KBTBD4WT-UM171
We hypothesised that the small molecule drug UM171 would recruit wild-type KBTBD4 to CoREST complexes with a similar mode of action to the R313PRR mutation. We, therefore, purified the full-length wild-type and mutant KBTBD4 proteins and tested their binding to the CoREST1NT2-HDAC2 complex in the presence or absence of UM171. We first explored crosslinking of E3-neo-substrate complexes using bis-sulfosuccinimidyl suberate (BS3), which has lysine-reactive hydroxysuccinimide ester groups separated by a spacer arm length of 11.4 Å. Crosslinked and control samples were analyzed on SDS-PAGE to observe mass shifts indicative of complex formation (Fig. 3a). The crosslinked KBTBD4R313PRR control sample exhibited a clear shift to a ~250 kD E3-neo-substrate complex, validating the assay conditions. Encouragingly, an identical mass shift was observed for the mixture of UM171, KBTBD4WT and CoREST1NT2-HDAC2 (Fig. 3a), but not when UM171 was absent, showing that the same E3-neo-substrate complex was induced by the small molecule.
Fig. 3: CoREST1NT2-HDAC2 is the common binding entity for KBTBD4R313PRR and KBTBD4WT-UM171.
figure 3
a For crosslinking experiments, 2.5 μM of indicated proteins were mixed and incubated with 4 mM BS3 crosslinker. The reactions were terminated by laemmli buffer and then analyzed in parallel with uncrosslinked controls on an SDS-PAGE gel, followed by Coomassie staining. Crosslinked or uncrosslinked species were labelled alongside the gel image. b Unlabelled CoREST1NT2-HDAC2 was titrated at indicated concentrations into 40 nM fluorescently-labelled KBTBD4 variants, with or without 25 μM UM171, for MST measurements. The apparent KD values of CoREST1NT2-HDAC2 binding to KBTBD4R313PRR and KBTBD4WT-UM171, respectively, were determined to be 0.75 µM and 15 µM, while negligible binding was observed to KBTBD4WT alone. c Unlabelled UM171 was titrated at indicated concentrations into 40 nM fluorescently-labelled KBTBD4WT with 5 μM CoREST1NT2-HDAC2 for MST measurements (apparent KD = 14 µM). All MST experiments were performed in triplicates. Data points and error bars represent the means and standard deviations. Source data are provided in the Source Data file. d Various E2 enzymes were mixed with CoREST1NT2-HDAC2, KBTBD4R313PRR, neddylated CUL3, E1 and ubiquitin, and incubated in the presence of ATP and MgCl2 to measure ubiquitylation. The reactions were analyzed on an SDS-PAGE gel and visualised by Coomassie stain. Ubiquitylated HDAC2 and KBTBD4R313PRR showed higher molecular weight ladders as labelled in the figure. e Immunoblots of in vitro ubiquitylation reaction mixtures using UBE2D1 E2 enzyme. KBTBD4 variants and UM171 were added to the reaction mixtures as indicated. The ubiquitylation experiments were performed at least twice with similar results. Source data are provided in the Source Data file.
Full size image
Binding affinity measurements of the UM171-induced complex were then performed by MST assay. Titration of the CoREST1NT2-HDAC2 complex into KBTBD4WT in presence of 25 µM UM171 produced a dose-dependent interaction estimated to be one order of magnitude weaker (KD ~ 15 μM) than for the equivalent KBTBD4R313PRR complex, whereas KBTBD4WT without UM171 exhibited smaller fluorescence deviations equivalent to background noise (Fig. 3b). Reassuringly, the same binding was observed when performing the reverse titration of UM171 into a mixture of KBTBD4WT and CoREST1NT2-HDAC2 (Fig. 3c). Altogether, these experiments demonstrate that the CoREST1NT2-HDAC2 complex is a common binding entity for KBTBD4R313PRR, as well as for KBTBD4WT in the presence of UM171.
Neo-substrates recruited to CRL3KBTBD4 complexes are known to be ubiquitinated14. We therefore reconstituted an in vitro ubiquitination assay using our purified CoREST1NT2-HDAC2 substrate, as well as other purified E1, E2 and E3 components. Initial experiments using KBTBD4R313PRR revealed poly-ubiquitin chain formation in the presence of the processive UBE2D family E2s (Fig. 3d). Further investigation confirmed neomorphic ubiquitination of HDAC2 by KBTBD4R313PRR, as well as by KBTBD4WT dependent on the presence of UM171 (Fig. 3e). By contrast, auto-ubiquitination of KBTBD4 was observed in all samples independently of UM171 (Fig. 3e). Based on these collective data, we selected the CoREST1NT2-HDAC2 complex as a suitable neo-substrate for Cryo-EM studies.
Cryo-EM structures reveal HDAC2 as the direct interactor of UM171 and neomorphic mutants of KBTBD4
Structural analysis of KBTBD4-neo-substrate complexes was pursued to define how cancer mutations and UM171 could similarly co-opt KBTBD4 for neomorphic binding to CoREST1NT2-HDAC2. Cryo-EM data obtained for the complexes of wild-type and mutant KBTBD4 yielded maps with nominal resolutions of 2.7 to 3.3 Å and revealed KBTBD4-neo-substrate complexes with both 2:1 and 2:2 stoichiometry (Fig. 4, Table 1 and Supplementary Figs. 1–3). The best map at 2.7 Å was obtained for the higher affinity sample mixing KBTBD4R313PRR and CoREST1NT2-HDAC2. The modelled structure revealed a KBTBD4R313PRR dimer engaging two HDAC2 molecules in C2 symmetry (2:2 stoichiometry, Fig. 4a, b, although a rare 2D class with one HDAC2 (2:1) was also observed, Supplementary Fig. 4). The mutant KBTBD4R313PRR was resolved to near full length, consisting of BTB, BACK and Kelch domains, whereas only the histone deacetylase domain could be modelled from the full-length HDAC2 construct (Fig. 4b, c); density was also lacking for its associated CoREST1NT2 subunit. To determine the structure of the lower affinity KBTBD4WT complex mediated by UM171, we used the BS3 crosslinker to stabilise the binding of the CoREST1NT2-HDAC2 neo-substrate. This sample yielded 2:2 and 2:1 KBTBD4WT-UM171-HDAC2 complexes at 2.9 Å and 3.1 Å, respectively. A subset of the 2:1 complexes showed additional density of lower quality for a CoREST1NT2 fragment wrapping around HDAC2 at 3.3 Å (Fig. 4d–f). No such density was found in the 2:2 complex, likely due to steric constraints that would prevent CoREST1 binding in this higher order assembly (Supplementary Fig. 5). Importantly, all 3D models identified HDAC2 as the direct interactor of KBTBD4, consistent with crosslinking and MST data showing that HDAC2 was sufficient for KBTBD4 recruitment (Supplementary Fig. 6).
Fig. 4: Cryo-EM structures reveal HDAC2 as the direct interactor of UM171 and neomorphic mutants of KBTBD4.
figure 4
a Cryo-EM map of the 2:2 KBTBD4R313PRR-HDAC2 complex coloured by individual subunit (contour level = 0.15). BTB-BACK and Kelch regions in KBTBD4 are labelled. b Ribbon representation of the 2:2 KBTBD4R313PRR-HDAC2 structure with each subunit coloured as in a. The dimer interface α1 and β1 in the KBTBD4 BTB domain are labelled. c Domain organizations of KBTBD4WT, KBTBD4R313PRR and HDAC2. d–f Cryo-EM maps of UM171-induced KBTBD4WT-substrate complexes at 2:1 (KBTBD4:HDAC2, contour level = 0.08), 2:1:1 (KBTBD4:HDAC2:CoREST1, contour level = 0.08) and 2:2 (KBTBD4:HDAC2, contour level = 0.08). UM171 is coloured in yellow. Asterisk indicates the unoccupied central pocket of the KBTBD4 Kelch domain. g Ribbon representation of the KBTBD4WT homodimer binding to UM171 and HDAC2. Subunits are coloured as in d–f. Each HDAC2 subunit interacts with residues at sites 1 and 2 in the KBTBD4 dimer as boxed and labelled by text. The distance between the two Kelch domains is measured by the side chains of E381. The interaction interface between the N-terminus of KBTBD4 chain a and the Kelch domain of KBTBD4 chain b is indicated with a red box. h Close-up view of the interaction interface between the N-terminus of KBTBD4 chain a and the Kelch domain of KBTBD4 chain b. KBTBD4 chains are coloured as in g and shown in ribbon representation. Key interacting residues are shown in sticks.
Full size image
Table 1 Cryo-EM data collection and refinement statistics
Full size table
In all structures, each HDAC2 subunit was bound by both Kelch domains in the KBTBD4 dimer (Fig. 4g), further highlighting the importance of dimeric assembly in the BTB protein class30,31. Of note, the high-quality tracing of the full-length KBTBD4 chains provided additional insights into the dimerization mechanism. The N-terminal BTB domain dimerized as expected through the α1 helix and a domain-swapped β1 strand (Fig. 4b, g). However, the BACK domain of KBTBD4 was notably two helices shorter than previously observed for other family members (Supplementary Fig. 7)30,32,33,34, allowing closer proximity of the two Kelch domains. Indeed, the Kelch domains in the KBTBD4 dimer were separated by just 5.8 Å (Fig. 4g), in contrast to the 26 Å and 54 Å observed in the cryo-EM structures of KLHL22 and KBTBD2, respectively32,33,34. The more central position of the Kelch domains in KBTBD4 was further stabilised by hydrophobic interaction with an N-terminal extension of the domain-swapped β1 strand (Fig. 4g, h), showing an example of conformational crosstalk between the BTB and Kelch domains of different protomers. These structural features illustrate how each domain in KBTBD4 contributes to neo-substrate recruitment and explains the requirement for dimerization in the cellular binding data shown in Fig. 1.
Structural basis for neomorphic substrate recruitment by the cancer mutant KBTBD4R313PRR
The structure of the mutant complex shows that the Kelch domain of KBTBD4R313PRR retains a conserved β-propeller fold of six blades (I to IV) despite harbouring the R313PRR indel mutation. Each blade consists of four antiparallel beta strands (βA to βD), connected by DA and BC loops of variable length. The recurrent cancer mutation R313PRR lies within the BC loop of blade II (BC II) (Fig. 5a), in agreement with previous predictions14,15.
Fig. 5: Cryo-EM structure of the KBTBD4R313PRR-HDAC2 complex reveals the basis of neomorphic substrate recruitment in cancer.
figure 5
a Ribbon representation of the KBTBD4R313PRR Kelch domain. The Kelch β-propeller consists of blades I to VI with each blade consisting of four antiparallel β strands (labelled A-D). Loops connecting βB and βB in blade II and IV are labelled as BC II and BC IV, respectively. The R313PRR cancer mutation in BC II is shown as red spheres. b Ribbon representation of HDAC2 binding to KBTBD4R313PRR_a at site 1 (BC II) and KBTBD4R313PRR_b at site 2 (BC IV). The second HDAC2 chain in the model is a symmetric duplication of the first and, therefore, not shown for simplicity. c Close-up view of site 1 interface. HDAC2 is shown as yellow surface and KBTBD4R313PRR_a as cyan ribbon. The key interacting arginine residues in KBTBD4R313PRR_a are shown as sticks in cyan. d–f Close-up views of site 1 key interfacial residues shown in sticks with the same colour scheme as c. Hydrogen bonding is indicated by dashed lines. g Close-up view of the site 2 interface. HDAC2 is shown as yellow surface and KBTBD4R313PRR_b as grey ribbon. The key interacting residues in KBTBD4R313PRR are shown as grey sticks. h–j Close-up views of site 2. Key interfacial residues are shown in sticks with the same colour scheme as in g.
Full size image
Each HDAC2 subunit in the 2:2 model makes direct contacts with both Kelch domains in the KBTBD4R313PRR dimer by engaging the mutant BC II loop from one chain (site 1) and the BC IV loop (blade IV, site 2) from the other (Fig. 5a, b). Unexpectedly, both sites involve interactions at the outer rim of the Kelch propeller (Fig. 5b), rather than binding to the central pocket, as reported for many other BTB-Kelch proteins35,36,37,38,39,40. In contrast to the dimeric assembly of KBTBD4, no apparent intermolecular interaction is observed between the two HDAC2 chains, suggesting that the two HDAC2 recruitment events are likely independent. Given the applied C2 symmetry in the model, only one chain of HDAC2 is discussed in further structural analysis.
Site 1 involving the mutant BC II loop forms the major interaction site for HDAC2 recruitment. Here, the arginine cluster introduced by the KBTBD4R313PRR mutation and preceding R312 establish both polar and hydrophobic interactions (Fig. 5c–f). Most importantly, the indel mutation extends the length of BC II and shifts the position of KBTBD4 R312, a non-mutated residue, by 5.9 Å to the tip of the loop. This structural alteration allows the R312 side chain to insert into the catalytic pocket of HDAC2 to within 5.7 Å of the catalytic Zn2+ ion, which natively binds to acetyl-lysines in histone H3 and H4 tails41,42,43. Similar to acetyl-lysine binding44,45, KBTBD4R313PRR R312 engages HDAC2 Y304 through a hydrogen bond interaction that induces an ‘in’ position of Y304, as typically observed in active class I HDACs (Fig. 5d and Supplementary Fig. 8a, b). In addition, the R312 side chain is sandwiched between HDAC2 F151 and F206 to form cation-π interactions (Fig. 5d). Additional cation-π interaction is observed between KBTBD4R313PRR R314 and HDAC2 Y205 close to the exit of the HDAC2 catalytic pocket (Fig. 5e). The aliphatic portions of KBTBD4R313PRR R314 and R315 also form hydrophobic interactions with HDAC2 L272 (Fig. 5e), while the preceding HDAC2 R271 forms a parallel π-stacking interaction with KBTBD4R313PRR (Fig. 5f). The benzamide HDAC1/2 inhibitor BRD6929 is known to insert into the same HDAC2 catalytic site46,47 and would be predicted to compete with this binding mode (Supplementary Fig. 8c, d). Indeed, the addition of BRD6929 abrograted all KBTBD4R313PRR-HDAC2 binding in the MST assay (Supplementary Fig. 8e), further highlighting the importance of these pocket interactions. Overall, the site 1 contacts demonstrate the neomorphic interactions facilitated by the KBTBD4R313PRR cancer mutation, as well as the native surface compatibility of surrounding wild-type residues.
The site 2 interface is comprised entirely of non-mutated residues. Here, KBTBD4R313PRR L408, F410 and F411 from the BC IV loop form extended hydrophobic interactions with an HDAC2 cleft that features Y202, L212, P228, Y359 and I363 (Fig. 5g, h). Outside of this cleft, KBTBD4R313PRR L408 and F411 also interact with the aliphatic parts of HDAC2 K362 and R366 (Fig. 5i). Additionally, the binding is stabilised by a salt bridge between KBTBD4R313PRR D409 and HDAC2 R230 (Fig. 5j). These interactions suggest a substantial contribution of the native site 2 in KBTBD4R313PRR to HDAC2 recruitment, demonstrating additional surface complementarity outside the cancer hotspot mutation site.
UM171 acts as a molecular glue to recruit HDAC2 to KBTBD4WT
The cryo-EM datasets for the KBTBD4WT-UM171-HDAC2 complexes yielded 2:1 and 2:2 models displaying an RSMD value of 1.55 Å across 1055 Cα positions. Superposition of the 2:1 and 2:2 states shows that the high structural conservation extends across the protein-compound and protein-protein interaction sites (Fig. 6a), indicating that the neo-substrate recruitment mechanism in these complexes is conserved regardless of the stoichiometry. The additional binding of CoREST1, however, appears to enforce an asymmetric 2:1:1 assembly due to steric constraints, as noted above (Fig. 4d and Supplementary Fig. 5). Overall, the models reveal similar site 1 and site 2 binding interfaces to the mutant KBTBD4 complexes, with both Kelch domains in the KBTBD4WT dimer contacting each subunit of HDAC2 (Fig. 6a). The CoREST1NT2 protein wraps around HDAC2 similarly to other ELM2-SANT domain proteins29.
Fig. 6: Cryo-EM structures show UM171 recruits HDAC2 to KBTBD4WT as a molecular glue.
figure 6
a Superposition of 2:1 and 2:2 complexes at HDAC2-interacting sites 1 and 2 in ribbon representation. UM171 is shown in stick representation. KBTBD4WT BC loops involved in both interaction sites are labelled. b Chemical structure of UM171 and a close-up view of UM171 binding at site 1 in the KBTBD4WT-HDAC2 interface. UM171 is shown in yellow sticks, KBTBD4WT and HDAC2 are shown in blue and pink as surface representations. c–e Close-up views of UM171 binding showing key interfacial residues as sticks. Hydrogen bonds are indicated by dashed lines. The subunits are coloured as in b. f Surface representations show the induced-fit of UM171 bound to KBTBD4WT_a (top panel) relative to the equivalent surface in unbound KBTBD4WT_b (lower panel). Loops involved in the induced fit are labelled. g Superposition of UM171-bound KBTBD4WT_a (blue) and unbound KBTBD4WT_b (light blue) at site 1 in ribbon and stick representations. Distances are measured to show the enlarged cleft in the UM171-bound KBTBD4WT_a. h Superposition of KBTBD4WT site 2 in HDAC2-bound KBTBD4WT_b and unbound KBTBD4WT_a. KBTBD4WT chains are shown in ribbon representation. L406 in BC IV in both KBTBD4WT chains is shown in sticks and the distances between side chains is measured to demonstrate the loop movement.
Full size image
UM171 is buried in the site 1 interfaces where it forms contacts with blades II and III from the KBTBD4WT Kelch domain, as well as with the HDAC2 catalytic pocket. Importantly, the maps allowed full tracing of the UM171 scaffold (Fig. 6b and Supplementary Fig. 3). The methyl tetrazole inserts into the HDAC2 catalytic pocket to form extensive interactions, including a hydrogen bond with HDAC2 Y304 and π–π interactions with HDAC2 F151, F206 and H179 (Fig. 6c). The binding is further stabilised by π–π interactions between the buried UM171 pyrimidoindole core and HDAC2 residues F151 and F206. Unlike SAHA and other HDAC inhibitors, UM171 does not coordinate the catalytic Zn2+ ion. Outside of the HDAC2 catalytic pocket, UM171 forms hydrophobic interactions with KBTBD4WT I310 and P311 from the BC II loop and L356 located at the tip of the BC III loop (Fig. 6d). Together L356 (BC III loop) and L335 (DA III loop) establish a narrow cleft for the UM171 benzyl group to anchor (Fig. 6d). Finally, direct interactions between KBTBD4WT S309, I310, D333 and HDAC2 G28, P30, F151, F206, L272 complete the ternary interaction network and highlight the native surface compatibility of the two proteins (Fig. 6e).
The asymmetric 2:1 and 2:1:1 models allowed comparison of the equivalent KBTBD4WT site 1 in both the bound and unbound states. Interestingly, UM171 was only observed in the HDAC2-bound site 1 and not in the second Kelch domain, suggesting that UM171 binding requires both proteins to be present. Indeed, UM171 did not exhibit pronounced affinity towards either KBTBD4WT or HDAC2 alone in an MST assay (Supplementary Fig. 9), highlighting the importance of the ternary interface for forming a suitable ligand-binding pocket. These features are consistent with previously defined molecular glue systems48,49, establishing UM171 a bona fide molecular glue for KBTBD4WT and HDAC2. Notably, UM171 binding exhibited an induced-fit mechanism requiring the opening of a cleft between Kelch domain blades II and III to accommodate the benzyl group of UM171 (Fig. 6f). In the UM171-bound KBTBD4WT protomer, L356 (BC III loop) is shifted a further 3.4 Å away from P311 (BC II loop) and 4.1 Å from L335 (DA III loop) relative to the unbound site 1 in the other KBTBD4WT protomer (Fig. 6g). Such induced conformational change demonstrates a plasticity of the Kelch domain that hitherto was not widely observed.
Additionally, each HDAC2 subunit engages a second KBTBD4WT protomer at site 2 involving blade IV of the second Kelch domain, as observed in the mutant complex (Fig. 6h). Similar to site 1, site 2 in the 2:1 model shows subtle conformational changes between the HDAC2-bound and unbound states. In particular, KBTBD4WT L406 in the BC IV loop shows a shift of 6.9 Å (Fig. 6h) to pack together with KBTBD4WT F408 and F409 into a shallow hydrophobic pocket on the HDAC2 surface. The equivalent F408 and F409 side chains were not resolved in the second KBTBD4WT protomer, suggesting their flexible nature when not engaged in HDAC2 interaction. Taken together, the structural analyses show that ternary complex formation is primed by native surface complementarity and stabilised by UM171, which acts as a molecular glue to bridge across the site 1 interface.
Structural mimicry of UM171 and cancer mutant KBTBD4R313PRR
The overall assemblies of the KBTBD4R313PRR-HDAC2 and KBTBD4WT-UM171-HDAC2 complexes are highly conserved (Cα RMSD < 3 Å). Site 1 superposition reveals that UM171 mimics the mutated BC II loop of KBTBD4R313PRR by engaging the same HDAC2 residues in its catalytic pocket (Fig. 7a). In particular, the methyl tetrazole of UM171 superimposes with the side chain of KBTBD4R313PRR R312 and forms similar interactions. Outside of the HDAC2 pocket, the pyrimidoindole core of UM171 overlays with the main chain atoms of the 311-PRP-313 loop of the KBTBD4R313PRR structure (Fig. 7a). Similarly, the binding pose adopted by KBTBD4WT blade IV at site 2 is conserved with its counterpart in the mutant complex despite the mutational frameshift in the preceding blade II (Fig. 7b). Such mimicry between a small molecule and natural mutation has been rarely reported and may inspire future strategies for molecular glue development utilising the knowledge of E3 ligase natural or engineered variants and their associated proteins.
Fig. 7: Structural mimicry of UM171 and cancer mutant KBTBD4R313PRR.
figure 7
Superposition of UM171 and KBTBD4R313PRR at site 1 (a) or site 2 (b). UM171 and key interacting residues are shown as sticks. Subunits are coloured as following: in the UM171 model, KBTBD4WT chains in blue and light blue, UM171 in yellow and HDAC2 in pink; in the KBTBD4R313PRR model, KBTBD4R313PRR in purple and light purple and HDAC2 in green.
Full size image
SAR of UM171 derivatives
In parallel, we carried out a systematic structure-activity relationship (SAR) analysis for UM171 revealing knowledge complementary to our structural study. We identified four sites (R1-R4) in UM171 (1) for modifications. The R1-R4 substituents are positioned on the opposite sides of UM171 and represent 3 different vectors. A naïve assessment suggested that R1 and R3 might be in closer contact with the protein, while R4 might be directed towards the solvent. To explore these sites, we designed compounds (2-15) (see the synthesis report in the Supplementary Information). The impact of the modifications was assessed in a cellular degradation assay for which HEK293 cells were incubated with these compounds at 1 µM for 16 hours. Past proteomic analyzes revealed greater degradation of CoREST and LSD1 than HDAC1/222,23,24. Therefore, the cellular assay focused on the levels of CoREST1 and LSD1, which for each compound were normalised and calculated in relation to DMSO control, as summarized in Table 2 and Supplementary Fig. 10.
Table 2 Structure-activity relationship of UM171 analogues in the degradation of CoREST1 and LSD1 in HEK 293 cells
Full size table
At the R1 site, substitution of the methyl tetrazole (1) with a 3,4-dimethylisoxazole (3) led to the complete loss of activity. In contrast, a carboxylic acid methyl ester at R1 (4) rescued the activity and caused a robust degradation of both LSD1 and CoREST1 even exceeding the activity of the reference compound UM171, suggesting smaller moieties at R1 might be favoured. However, substituting the methyl ester in (4) with a carboxylic acid moiety (5), or adding a nitrile at R1 (6) did not show detectable activity, potentially due to desolvation penalty or lack of apolar interactions. These observations validated the importance of the hydrophobic interactions observed in the cryo-EM structures between the UM171 methyl tetrazole and the HDAC2 catalytic pocket containing F151, H179, F206, L272 (Supplementary Fig. 11). However, the activity loss by an isosteric methylpyrazole moiety replacement (2) is hard to explain by the current structural information and probably merits further chemical exploration.
Similar to R1, changes at the R2 and R3 sites significantly impacted the compound activity. Substituent exchange between R1 and R2 was explored by compounds (4) and (7). Moving the methyl ester moiety to the adjacent position caused loss of activity, potentially due to a steric clash with the narrow HDAC2 catalytic pocket. For R3, replacing the benzyl moiety with a hydrogen (8) reduced the activity, validating the importance of this benzyl moiety in engaging KBTBD4 as shown in the cryo-EM structures (Fig. 6 and Supplementary Fig. 11). Compound activity is sensitive to changes at these sites, potentially owing to their proximity to the buried KBTBD4-HDAC2 interface as shown in the structures.
In contrast, changes at R4 were more tolerable, in line with our original hypothesis as well as the cryo-EM structures that this moiety is exposed to solvent. Substitution of the primary amine with dimethylamine (9) led to no perceptible change in CoREST1 or LSD1 levels, reinforcing that the activity changes observed for compounds (4), (5), and (8) were solely owing to the variability in R1 and R3 but not R4. Removing the distal basic moiety at R4 (10) resulted in reduced activity, while incorporating a basic pyrrolidine (11) or piperidine (12) ring rescued the activity with a slightly better potency observed for compound (11). As the R4 substituent is away from the complex binding interface in the cryo-EM structures, we speculate its observed impact on substrate degradation may be due to varied physico-chemical properties and cell permeability of the compound analogues. Interestingly, a carboxylic acid methyl ester at R1 consistently showed superior activity to the methyl tetrazole moiety in the presence of three different R4 substituents (comparing 4-9, 11-13 and 12-14 compound pairs), suggesting the two sites function independently and can be optimised separately.
Finally, we examined the impact of the central ring of (1). Compound (15) with an open central ring exhibited no activity, suggesting that the relative orientation of the R1-R4 substituents ensured by the UM171 pyrimidoindole core, is essential for the molecular glue effect.
Using the MST assay, we were able to observe a 3.5-fold improvement in the ternary complex affinity of the most active analogue (13) (S234984), whereas there was a 3-fold reduction for the inactive compound (2), confirming the expected correlation between ternary complex stability and cellular activity (Supplementary Fig. 12). Taken together, our experimental structural-activity relationship analysis, as well as the cryo-EM structures, extend our understanding of UM171 and will guide future rational optimisation for the drug.
Discussion
Here, we elucidate the neomorphic mechanism of KBTBD4. Our cryo-EM and supporting data show how cancer-driving mutations remodel the KBTBD4 surface and identify HDAC1/2 as the direct interactor for recruitment of the LSD1-CoREST-HDAC1/2 complex. We found that UM171, a clinical phase 2 drug for haematopoietic stem cell transplantation, structurally mimics the cancer-mutated loop and acts as a molecular glue for wild-type KBTBD4 and HDAC1/2, resolving the long-standing puzzle of UM171’s mechanism of action.
Our structures highlight several important features of the neo-substrate recruitment. Overall, the mutant and drug-induced neo-complexes adopt highly conserved structural assemblies. Consistent with current knowledge, the binding partners exhibit native surface complementarity that is enhanced by UM171 and the cancer mutant. Both Kelch domains in the KBTBD4 homodimer provide distinct epitopes for the HDAC2 subunit to bind, demonstrating the importance of the dimer for bivalent substrate recognition, as observed in our biochemical data. Contrary to expectation, these binding sites occur at the outer rim of the Kelch β-propeller domain rather than the central pocket typically utilised by native Kelch E3 substrates32,33,35,36,37,38. UM171 acts at the ternary interface to remodel the Kelch domain surface similar to the cancer mutant. Our structure-activity relationship (SAR) analysis provides a basis for further rational design of UM171. Similar to IMiDs50, parts of the UM171 scaffold display a steep SAR, likely reflecting the relatively high contribution (~40%) of the ligand to the complex interface. Nonetheless, we identify sites within UM171 amenable to chemical modification for improved potency guided by the 3D structural models.
In cells, HDAC1/2 can engage different ELM2-SANT domain proteins to assemble into distinct repressor complexes. In addition to CoREST, UM171 induces some degradation of MIER2, but seemingly not of the NuRD complex containing MTA124. Notably, CoREST and MIER2 are degraded to a greater extent than the direct interactor HDAC1/2, suggesting that in the context of these large multi-subunit complexes HDAC1/2 is less accessible for ubiquitination. Our structures provide opportunity to model the different repressor complexes to gain potential insights into the neomorphic substrate specificity. Modelling of MTA151 predicts a steric clash between its BAH domain and KBTBD4 that may preclude its binding (Supplementary Fig. 5). Steric constraints also predict that the KBTBD4 dimer can accommodate only one HDAC1/2-CoREST pair, as observed in our 2:1:1 complex.
In support of our HDAC2-CoREST model, parallel but complementary approaches using HDAC1-CoREST have yielded mutant and UM171-dependent KBTBD4 complexes with a conserved 2:1:1 structural arrangement52,53. An interesting feature of these structures was the inclusion of inositol hexakisphosphate (InsP6), a known HDAC1/2 cofactor. InsP6 binding stabilised the HDAC1-CoREST interface as expected, but also formed a second molecular glue interaction with KBTBD4 to enhance the ternary complex affinity, as we have also observed in our MST assay (Supplementary Fig. 13).
To date, only a small number of E3s have been shown to engage a molecular glue. KBTBD4 expands this E3 repertoire to the CRL3 family and may offer a distinct interaction surface for target substrates less compatible with previously reported E3s. The exquisite mimicry between UM171 and KBTBD4R313PRR raises the intriguing prospect that mutations could help to identify other exploitable E3-target interfaces and inspire the discovery of further molecular glues. While only a few mechanistic studies exist14,54, a recent proteome-wide screen of cancer mutants has revealed that neomorphic E3 interactions may be relatively prevalent3. Understanding these mechanisms will inform cancer biology, as well as help to define other E3-neo-substrate interfaces that may be hijacked by small molecules. Learning from UM171 and KBTBD4, future molecular glue discovery can potentially be accelerated by prioritising E3 ligases selected by nature, or by wider genetic screens.
Methods
Plasmid construction
FLAG-KBTBD4WT (17-534 a.a., Uniprot Q9NVX7-2) and FLAG-KBTBD4R313PRR (17-536 a.a., cancer indel mutation introduced by site-directed mutagenesis) were subcloned into a pcNDA3.1 (+) vector with the 3xFLAG epitope tagged to its N-terminus using standard restriction enzyme cloning14. The Kelch domain of KBTBD4R313PRR (237-536 a.a.) was amplified by PCR from the full-length mutant construct and subcloned into a pcDNA3.1(+) vector with the 3xFLAG epitope tagged to its N-terminus using standard restriction enzyme cloning. Monomeric KBTBD4R313PRR was generated by introducing mutations H42A, V46D, I50T, L53E, F60S, L77K and A81E and cloned into the pCMV-3Tag-1A vector using the custom gene synthesis service provided by GenScript.
For interaction site mapping in HEK293 cells, full-length human CoREST1 (pCMV5 HA RCOR1; NM_015156.4) construct was purchased from the MRC Protein Phosphorylation and Ubiquitylation Unit (cat# DU19711, https://mrcppureagents.dundee.ac.uk). N or C-terminal fragments of CoREST1 were generated using standard PCR methods, followed by restriction enzyme cloning into a pCMV5 vector with an N-terminal HA fusion.
For biophysical assays and cryo-EM studies, the ELM2-SANT1 domain region of human CoREST1 (71-259 a.a., Uniprot Q9UKL0-1) was amplified from the full-length construct above and cloned into the mammalian expression vector pHTBV with a C-terminal FLAG/10x-histidine fusion using ligation-independent cloning. The untagged full-length HDAC2 (1-488 a.a., Uniprot Q92769-1) was cloned into the pcDNA3.1(−) vector using standard restriction enzyme cloning. For microscale thermophoresis, near full-length LSD1 (171-852 a.a., Uniprot O60341-1) was cloned using ligation-independent cloning into bacterial expression vector pNIC28-Bsa4 (GenBank accession number EF198106) which provides an N-terminal 6x-histidine tag and a TEV cleavage site.
For ubiquitination assays, full-length CUL3 (1-768 a.a., Uniprot Q13618-1) was cloned into a baculovirus transfer vector pFAST-BacTM (Invitrogen) with a C-terminal FLAG/6x-histidine tag and a TEV protease cleavage site using ligation-independent cloning. Full-length RBX1 was cloned into pFAST-BacTM with no tag. The neddylation enzyme plasmids were gifts from Brenda Schulman’s lab and were described in their previous publication55. The ubiquitin E1 and E2 plasmids were gifts from Cheryl Arrowsmith (Addgene plasmids #25213, #25598, #25145, #25470). All DNA sequences were verified by Source Bioscience Ltd.
DNA oligonucleotides used for plasmid construction are listed below:
Full length FLAG-KBTBD4WT fwd CCGGCCGGATCCATGGAATCACCAGAGGAGCCTGG
rev GCATACGTCGACTTAGGCCAACACAAACTGCAAATTG
Full length FLAG-KBTBD4R313PRR mutagenesis fwd TTGTATGTGGTGGGAGGGTCCATCCCACGGCCACGGCGCATGTGGAAGT
rev ACTTCCACATGCGCCGTGGCCGTGGGATGGACCCTCCCACCACATACAA
Kelch domain of KBTBD4R313PRR (237-536 a.a.) fwd CCGGCCGGATCCATGGAGGCTTTTGCAGAGTCACTCAGGA
rev GCATACGTCGACTTAGGCCAACACAAACTGCAAATTG
CoREST1-NT1 fwd GCATACGGATCCATGCCGGCCATG
rev CCGGCCGTCGACTCACTCCTCCCGCTCCCG
CoREST1-NT2 fwd GCATACGGATCCATGAAAAGTTTGGCGGCGGC
rev CCGGCCGTCGACTCACTCCTCCCGCTCCCG
CoREST1-CT1 fwd GCATACGGATCCATGAGCGAGGATGAACTGGAAGAGG
rev CCGGCCGTCGACTCAGGAGGCAGATGCATATCTGAC
CoREST1-CT2 fwd GCATACGGATCCATGGTCAAAAAAGAAAAACATAGCACACAAGCTAAAAATAG
rev CCGGCCGTCGACTCAGGAGGCAGATGCATATCTGAC
CoREST1 (71-259 a.a) for cryo-EM fwd TTAAGAAGGAGATATACTATGTCCTGGGAGGAAGGCAGC
rev GATTGGAAGTAGAGGTTCTCTGCGCGATCCATCACACTAGTT
Full length HDAC2 (1-488 a.a.) for cryo-EM fwd TTAAGAAGGAGATATACTATGTCCTGGGAGGAAGGCAGC
rev GATTGGAAGTAGAGGTTCTCTGCGCGATCCATCACACTAGTT
Near full-length LSD1 (171-852 a.a.) fwd TTAAGAAGGAGATATACTATGCCATCGGGTGTGGAGGGC
rev GATTGGAAGTAGAGGTTCTCTGCTCACATGCTTGGGGACTGCTG
CUL3 (1-768 a.a.) fwd GATTGGAAGTAGAGGTTCTCTGCTCACATGCTTGGGGACTGCTG
rev GATTGGAAGTAGAGGTTCTCTGCTGCTACATATGTGTATACTTTGCGATC
RBX1 (1-108 a.a.) fwd TACTTCCAATCCATGGCGGCAGCGATGGATGTGG
rev TATCCACCTTTACTGTCAGTGCCCATACTTTTGGAATTCCC
Immunoprecipitation and immunoblotting
HEK293 cells (ATCC CRL-1573) were cultured in high glucose Dulbecco’s Modified Eagle’s Medium (GIBCO #41965039) with 5% penicillin-streptomycin (ThermoFisher Scientific #15070063) and 10% fetal bovine serum (Sigma-Aldrich #F9665) inside a 5% CO2 incubator at 37 °C. Cells were transfected as indicated at 60% confluency (5 µg DNA per plate using PEI-MAX 40 kDA at mass ratio 3:1). After 40 h, cells were treated with 0.1 µM MLN4924 (Calbiochem, #951950–33–7) for 4 h. Cells were washed with phosphate-buffered saline, harvested into lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 10% glycerol, 0.1% NP-40, 1 mM EDTA and 5 mM MgCl2) containing protease inhibitors (Sigma-Aldrich #11697498001), beta-glycerolphosphate, DTT, PMSF and okadaic acid, incubated on ice for 30 min and centrifuged for 30 min at 20 000 g at 4 °C. The supernatant was incubated with Anti-FLAG M2 affinity gel (Sigma-Aldrich #A2220) or HA affinity gel (Sigma-Aldrich, #E6779) at 4 °C for 3 h before being washed 3 times with lysis buffer. Immunoprecipitates were eluted using 1× LDS buffer (ThermoFisher Scientific #NP0007) supplemented with DTT and boiled for 3 min at 95 °C. Cell lysates or immunoprecipitates were resolved in 4-12% gradient Bis-Tris gels, transferred to PVDF membranes (Merck, #IPVH00010) and immunoblotted with corresponding antibodies at 1:1000 dilution – anti-FLAG (Sigma-Aldrich #F1804), anti-HA (Roche, #12013819001), anti-RCOR1 (Cell Signalling Technology, #14567), anti-LSD1 (Cell Signalling Technology, #2139) and anti-HDAC2 (Cell Signalling Technology, #5113). Data were acquired and analyzed in Image Lab 5.2.1. Source data are provided in the Source Data file.
Protein expression
5 mg DNA of KBTBD4 alone or CoREST1NT2 and HDAC2 constructs at 1:1 ratio was transfected into 1 L Expi293FTM (Thermo Fisher Scientific #A14528) cell culture in mid-log phase (2 × 106 mL−1) in Freestyle 293TM Expression Medium (Thermo Fisher Scientific #12338018) with 15 mg PEI-MAX 40 kDA. Cells were grown in an orbital shaker at 37 °C with 8% CO2 for 24 h and then 30 °C for 48 h before being harvested by centrifugation at 800 g for 12 min. The pelleted cells were washed with phosphate-buffered saline, pelleted again, then flash-frozen in liquid nitrogen for storage in a −80 °C freezer.
All bacterially expressed plasmids were transformed into E. coli strain BL21(DE3)R3-pRARE2. Cells were cultured in LB broth at 37 °C until OD600 reached 0.6. Recombinant protein expression was then induced by addition of 0.4 mM isopropyl β-D-1-thiogalactopyranoside, followed by 18 h continuous shaking at 18 °C. Cells were harvested by centrifugation and flash-frozen in liquid nitrogen for storage in a −80 °C freezer.
For baculoviral expression, recombinant baculovirus was generated by site-specific transposition in the DH10Bac™ E. coli strain (Thermo fisher, #10361012), followed by bacmid isolation and transfection into Sf9 insect cells using JetPrime® (Polyplus, 114-15). Amplified P2 virus was used to infect Sf9 insect cells at 2 × 106 mL−1. Infected cells were incubated at 27 °C with 100 rotations per minute (RPM) shaking for 48 to 72 hours before being harvested by centrifugation at 800 g in 4 °C for 12 min. The pelleted cells were washed with phosphate-buffered saline, pelleted again, then flash-frozen in liquid nitrogen for storage in a −80 °C freezer.
Protein purification
Whole mammalian cell pellets expressing KBTBD4 or CoREST1NT2-HDAC2 complex were resuspended in buffer A (50 mM 4-(2-Hydroxyethyl)piperazine-1-ethane-sulphonic acid (HEPES) pH 7.5, 150 mM NaCl, 10% glycerol, 5 mM MgCl2, 0.5 mM TCEP) supplemented with protease inhibitors and lysed by sonication. Recombinant proteins were captured by anti-FLAG M2 affinity gel, washed with buffer A supplemented with 10 mM ATP and 20 mM MgCl2, and eluted with 3xFLAG peptide (ThermoFisher #A36806). The sample was then concentrated and subjected to size-exclusion chromatography using a Superdex S200 10/300 GL increase column pre-equilibrated with buffer A. Peak fractions were pooled and concentrated to 20 μM.
For all bacterial and insect cell expressed proteins, cell pellets were resuspended in buffer B (50 mM HEPES pH 7.5, 500 mM NaCl, 5% glycerol, 5 mM imidazole, 0.5 mM TCEP). LSD1 protein was captured on nickel sepharose resin, washed with buffer B and eluted by a stepwise gradient of 30–250 mM imidazole. Further clean-up was performed by size-exclusion chromatography using a HiLoad Superdex S200 16/60 or S75 column buffered in 50 mM HEPES pH 7.5, 300 mM NaCl, 0.5 mM TCEP. Peak fractions were pooled and concentrated to 20 μM.
Microscale thermophoresis
Experiments were performed in a Monolith NT.115 instrument (NanoTemper Technologies GmbH). Purified recombinant KBTBD4 variants, as well as HDAC2 were labelled using a Monolith RED-NHS 2nd generation Protein Labelling Kit following the manufacturer’s instruction. CoREST1NT2-HDAC2, HDAC2, UM171, compounds (2), (13) or BRD6929 (MedChemExpress #HY-100719) as indicated were mixed in each experiment with 40 nM labelled KBTBD4 or 80 nM HDAC2 buffered in 50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 5 mM MgCl2, 0.05% Tween-20, 0.1 mM TCEP, with or without 40 µM inositol hexaphosphate. The reactions were incubated at room temperature for 15 min and then transferred into the standard MST capillaries for measurements at 25 °C with 20% excitation power and 20% or 40% MST power in MO. Control 1.6.1. Data were analyzed using the Monolith Evaluation software MO. Affinity Analysis 2.3 using the implemented KD model and plotted in GraphPad Prism 10.4.0. Source data are provided in the Source Data file.
Crosslinking assay
Purified proteins at equal molar ratio were mixed as indicated in 50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 5 mM MgCl2, 0.1 mM TCEP and incubated at room temperature for 15 min. Bis-sulfosuccinimidyl Suberate (BS3, ThermoFisher Scientific #21580) was then added to reach a final concentration of 4 mM. Reactions were incubated at room temperature for 45 min and then terminated by adding either 1x LDS buffer or 50 mM Tris pH 7.5.
In vitro ubiquitylation
4 μM purified full-length CUL3-RBX1 was mixed with 0.5 μM APPBP1-Uba3, 1 μM UBE2M and 60 μM NEDD8 in 50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 0.5 mM TCEP, 10 mM MgCl2 and 10 mM ATP for 30 min at room temperature for in vitro neddylation55. 100 μM ubiquitin, 100 nM UBA1, 1 μM neddylated CUL3-RBX1, 1 μM UBE2 of choice and 5 μM CoREST1NT2-HDAC2 were mixed in 50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 0.5 mM TCEP, 10 mM MgCl2 and 10 mM ATP. 1 μM KBTBD4 (WT or R313PRR) and 10 μM UM171 were added into the reactions as indicated. Reactions were incubated at room temperature for 30 min, subjected to SDS-PAGE, and then analyzed by Coomassie staining or immunoblots.
Cryo-EM sample preparation, data acquisition and data processing
For the KBTBD4WT-UM171-CoREST1NT2-HDAC2 complex, 4 μM KBTBD4WT, 4 μM CoREST1NT2-HDAC2 and 40 μM UM171 were mixed in 50 mM HEPES pH 7.5, 150 mM NaCl, 5% glycerol, 5 mM MgCl2, 0.5 mM TCEP, and then crosslinked with BS3. Samples were plunge-frozen on Quantifoil Cu R1.2/1.3 300-mesh grids that had been freshly glow-discharged for 1 min. The Vitrobot for plunge-freezing was set to 100% humidity, 4 °C and 4 s blotting time.
For the KBTBD4R313PRR-CoREST1NT2-HDAC2 complex, 4 μM KBTBD4R313PRR and 4 μM CoREST1NT2-HDAC2 were mixed in 50 mM HEPES pH 7.5, 150 mM NaCl, 10% glycerol, 5 mM MgCl2, 0.5 mM TCEP, and incubated for 15 min on ice. Samples were plunge frozen on UltrAufoil R1.2/1.3 300-mesh grids which had been freshly glow-discharged for 1 min. The Vitrobot Mark IV (ThermoFisher Scientific) for plunge-freezing was set to 100% humidity, 4 °C and 2.5 s blotting time. All grids were clipped into autogrid cartridges (Thermo Fisher #1036173).
Both datasets were collected on a FEI Titan Krios (Thermo Fisher Scientific) operating at 300 kV at the Central Oxford Structural Molecular Imaging Centre (COSMIC). Super-resolution micrographs with 2x binning (0.832 Å pixel−1) were collected at ×105,000 nominal magnification at 38 or 38.6 e−/Å2 dose using a Gatan K3 direct electron detector.
For the KBTBD4WT dataset, 12,722 micrographs with defocus range −3.2 to −1.0 µm were collected and then processed in Cryosparc v4.4.1. 13,588,054 particles were picked with the blob picker function, and extracted particles were subjected to multiple cycles of 2D classification. 193,056 particles of good classes were selected to generate an ab-initio model which revealed partial occupancy for a segment suggesting heterogeneity. Particles were then manually separated based on 2D classes to generate ab-initio models, and then subjected to a heterogenous refinement which revealed distinct 2:1 and 2:2 species. The preliminary 3D volumes were used to create 2D templates for rare view particle picking with the template picker function. The new particles were subjected to 2D classifications, resulting in 264,324 and 356,933 particles of good classes used in non-uniform refinements for 2:1 (C1 symmetry) and 2:2 (C2 symmetry), respectively. 2:1 and 2:2 maps were further refined to 3.1 and 2.9 Å. Additionally, a subclass of 2:1 species containing a CoREST1 subunit was identified by a 3D variability job and further refined to 3.3 Å with 105,484 particles by local refinement.
For the KBTBD4R313PRR dataset, 13,131 micrographs with defocus range -2.8 to -1.0 µm were collected and processed similarly to the WT dataset. 295,217 particles were selected for 3D reconstitution after blob picking, template picking and multiple rounds of 2D classification. Non-uniform refinement using these particles with C2 symmetry applied reached a final resolution of 2.7 Å.
Model building and refinement
The full-length KBTBD4 model generated by AlphaFold 256, deacetylase domain of HDAC2 (7ZZT) and ELM2-SANT1 domain of a CoREST1 analogue MTA1 (4BKX) were fitted into the reconstituted EM maps using UCSF ChimeraX 1.7.157. The models were manually refined in Coot 0.8.9.258 and then auto-refined in Phenix 1.20.1-448759 using the strategies implementing real space, rigid body and atomic displacement parameters. The geometry of all the models were validated using Phenix_MolProbity.
Compound synthesis and quality controls
A detailed synthesis report can be found in the Supplementary information. All reagents obtained from commercial sources were used without further purification. Anhydrous solvents were obtained from commercial sources and used without further drying. The reactions were monitored using LC-MS and GC-MS instruments.
Analytical LC-MS: Agilent HP1200 LC with Agilent 6140 quadrupole MS, operating in positive or negative ion electrospray ionisation mode. Molecular weight scan range was 100 to 1350 m/z. Parallel UV detection was done at 210 nm and 254 nm. Samples were supplied as a 1 mM solution in MeCN or in THF/water (1:1) with 5 µL loop injection. LC-MS analyses were performed on two instruments, one of which was operated with basic, and the other with acidic eluents.
Basic LC-MS: Gemini-NX, 3 µm, C18, 50 mm × 3.00 mm i.d. column at 23 °C, at a flow rate of 1 mL min-1 using 5 mM aq. NH4HCO3 solution and MeCN as eluents.
Acidic LC-MS: ZORBAX Eclipse XDB-C18, 1.8 µm, 50 mm × 4.6 mm i.d. column at 40 °C, at a flow rate of 1 mL min-1 using water and MeCN as eluents, both containing 0.02 V/V% formic acid.
Combination gas chromatography and low-resolution mass spectrometry were performed on Agilent 6850 gas chromatograph and Agilent 5975 C mass spectrometer using 15 m × 0.25 mm column with 0.25 µm HP-5MS coating and helium as carrier gas. Ion source: EI + , 70 eV, 230 °C, quadrupole: 150 °C, interface: 300 °C.
Flash chromatography was performed on ISCO CombiFlash Rf 200i with pre-packed silica-gel cartridges (RediSep®Rf Gold High Performance).
Preparative HPLC purifications were performed on an Armen Spot Liquid Chromatography system with a Gemini-NX® 10 µm C18, 250 mm × 50 mm i.d. column running at a flow rate of 118 mL min-1 with UV diode array detection (210 – 400 nm).
1H NMR and proton-decoupled 13 C NMR measurements were performed on Bruker Avance III 500 MHz spectrometer and Bruker Avance III 400 MHz spectrometer, using DMSO-d6 or CDCl3 as solvent. 1H and 13 C NMR data are in the form of delta values, given in part per million (ppm), using the residual peak of the solvent as internal standard (DMSO-d6: 2.50 ppm (1H) / 39.5 ppm (13 C); CDCl3: 7.26 ppm (1H) / 77.0 ppm (13 C)). Splitting patterns are designated as: s (singlet), d (doublet), t (triplet), q (quartet), sp (septet), m (multiplet), br s (broad singlet), dd (doublet of doublets), td (triplet of doublets), qd (quartet of doublets). In some cases, two sets of signals appear in the spectra due to hindered rotation.
HRMS were determined on a Shimadzu IT-TOF, ion source temperature 200 °C, ESI +/−, ionisation voltage: ( ± )4.5 kV. Mass resolution min. 10000.
All obtained products had an LC purity above 96% that was corroborated by their 1H NMR spectrum unless specifically mentioned otherwise.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Cryo-EM maps and atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) under accession codes EMD-51337 and PDB 9GGM (2:2 KBTBD4R313PRR-HDAC2), EMD-51338 and PDB 9GGN (2:2 KBTBD4WT-HDAC2 with UM171), EMD-51335 and PDB 9GGL (2:1 KBTBD4WT-HDAC2 with UM171) as well as EMD-51336 and PDB 9I2C<(2:1:1 KBTBD4WT-HDAC2-CoREST1 with UM171). All other data are available in the manuscript or the supplementary materials. Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author. Source data are provided with this paper.
References
Jeffery, C. J. Proteins with neomorphic moonlighting functions in disease. IUBMB Life 63, 489–494 (2011).
CASPubMedMATHGoogle Scholar
Takiar, V., Ip, C. K., Gao, M., Mills, G. B. & Cheung, L. W. Neomorphic mutations create therapeutic challenges in cancer. Oncogene 36, 1607–1618 (2017).
CASPubMedGoogle Scholar
Mo, X. et al. Systematic discovery of mutation-directed neo-protein-protein interactions in cancer. Cell 185, 1974–1985.e1912 (2022).
CASPubMedPubMed CentralMATHGoogle Scholar
Kliche J. et al. Proteome-scale characterisation of motif-based interactome rewiring by disease mutations. Mol Syst Biol 20, 1025–1048 (2024).
Schreiber, S. L. Molecular glues and bifunctional compounds: Therapeutic modalities based on induced proximity. Cell Chem. Biol. 31, 1050–1063 (2024).
CASPubMedMATHGoogle Scholar
Tsai J. M., Nowak R. P., Ebert B. L. & Fischer E. S. Targeted protein degradation: from mechanisms to clinic. Nat. Rev. Mol. Cell Biol. 25, 740–757 (2024).
Dang, L. et al. Cancer-associated IDH1 mutations produce 2-hydroxyglutarate. Nature 462, 739–744 (2009).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Fischer, E. S. et al. Structure of the DDB1-CRBN E3 ubiquitin ligase in complex with thalidomide. Nature 512, 49–53 (2014).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Matyskiela, M. E. et al. Crystal structure of the SALL4-pomalidomide-cereblon-DDB1 complex. Nat. Struct. Mol. Biol. 27, 319–322 (2020).
CASPubMedGoogle Scholar
Sievers Q. L. et al. Defining the human C2H2 zinc finger degrome targeted by thalidomide analogs through CRBN. Science 362, eaat0572 (2018).
Han T. et al. Anticancer sulfonamides target splicing by inducing RBM39 degradation via recruitment to DCAF15. Science 356, eaal3755 (2017).
Slabicki, M. et al. The CDK inhibitor CR8 acts as a molecular glue degrader that depletes cyclin K. Nature 585, 293–297 (2020).
CASPubMedPubMed CentralGoogle Scholar
Lv L. et al. Discovery of a molecular glue promoting CDK12-DDB1 interaction to trigger cyclin K degradation. Elife 9, e59994 (2020).
Chen, Z. et al. Disease-associated KBTBD4 mutations in medulloblastoma elicit neomorphic ubiquitylation activity to promote CoREST degradation. Cell Death Differ. 29, 1955–1969 (2022).
CASPubMedPubMed CentralMATHGoogle Scholar
Northcott, P. A. et al. The whole-genome landscape of medulloblastoma subtypes. Nature 547, 311–317 (2017).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Lee, J. C. et al. Recurrent KBTBD4 small in-frame insertions and absence of DROSHA deletion or DICER1 mutation differentiate pineal parenchymal tumor of intermediate differentiation (PPTID) from pineoblastoma. Acta Neuropathol. 137, 851–854 (2019).
CASPubMedPubMed CentralMATHGoogle Scholar
Yamashita, S. et al. Clinicopathologic analysis of pineal parenchymal tumors of intermediate differentiation: a multi-institutional cohort study by the Kyushu Neuro-Oncology Study Group. J. Neurooncol 162, 425–433 (2023).
CASPubMedMATHGoogle Scholar
Bastani, S., Staal, F. J. T. & Cante-Barrett, K. The quest for the holy grail: overcoming challenges in expanding human hematopoietic stem cells for clinical use. Stem Cell Investig. 10, 15 (2023).
CASPubMedPubMed CentralMATHGoogle Scholar
Fares, I. et al. Cord blood expansion. Pyrimidoindole derivatives are agonists of human hematopoietic stem cell self-renewal. Science 345, 1509–1512 (2014).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Cohen, S. et al. Improved outcomes of UM171-expanded cord blood transplantation compared with other graft sources: real-world evidence. Blood Adv. 7, 5717–5726 (2023).
CASPubMedPubMed CentralMATHGoogle Scholar
Cohen, S. et al. Hematopoietic stem cell transplantation using single UM171-expanded cord blood: a single-arm, phase 1-2 safety and feasibility study. Lancet Haematol. 7, e134–e145 (2020).
PubMedGoogle Scholar
Subramaniam, A. et al. Lysine-specific demethylase 1 A restricts ex vivo propagation of human HSCs and is a target of UM171. Blood 136, 2151–2161 (2020).
CASPubMedPubMed CentralMATHGoogle Scholar
Chagraoui, J. et al. UM171 preserves epigenetic marks that are reduced in ex vivo culture of human HSCs via Potentiation of the CLR3-KBTBD4 Complex. Cell Stem Cell 28, 48–62.e46 (2021).
CASPubMedMATHGoogle Scholar
Zemaitis, K., Ghosh, S., Hansson, J. & Subramaniam, A. The stem cell-supporting small molecule UM171 triggers Cul3-KBTBD4-mediated degradation of ELM2 domain-harboring proteins. J. Biol. Chem. 299, 104662 (2023).
CASPubMedPubMed CentralGoogle Scholar
Song, Y. et al. Mechanism of Crosstalk between the LSD1 Demethylase and HDAC1 Deacetylase in the CoREST Complex. Cell Rep. 30, 2699–2711.e2698 (2020).
CASPubMedPubMed CentralMATHGoogle Scholar
Yang, M. et al. Structural basis for CoREST-dependent demethylation of nucleosomes by the human LSD1 histone demethylase. Mol. Cell 23, 377–387 (2006).
CASPubMedMATHGoogle Scholar
Zhuang, M. et al. Structures of SPOP-substrate complexes: insights into molecular architectures of BTB-Cul3 ubiquitin ligases. Mol. Cell 36, 39–50 (2009).
CASPubMedPubMed CentralMATHGoogle Scholar
Yang, Y. et al. LSD1 coordinates with the SIN3A/HDAC complex and maintains sensitivity to chemotherapy in breast cancer. J. Mol. Cell Biol. 10, 285–301 (2018).
CASPubMedGoogle Scholar
Millard, C. J. et al. Class I HDACs share a common mechanism of regulation by inositol phosphates. Mol. Cell 51, 57–67 (2013).
CASPubMedPubMed CentralMATHGoogle Scholar
Canning, P. et al. Structural basis for Cul3 protein assembly with the BTB-Kelch family of E3 ubiquitin ligases. J. Biol. Chem. 288, 7803–7814 (2013).
CASPubMedPubMed CentralMATHGoogle Scholar
Ahmad, K. F., Engel, C. K. & Prive, G. G. Crystal structure of the BTB domain from PLZF. Proc. Natl Acad. Sci. USA 95, 12123–12128 (1998).
ADSCASPubMedPubMed CentralGoogle Scholar
Hu, Y. et al. Dynamic molecular architecture and substrate recruitment of cullin3-RING E3 ligase CRL3(KBTBD2). Nat. Struct. Mol. Biol. 31, 336–350 (2024).
CASPubMedPubMed CentralMATHGoogle Scholar
Teng, F. et al. Cryo-EM structure of the KLHL22 E3 ligase bound to an oligomeric metabolic enzyme. Structure 31, 1431–1440.e1435 (2023).
CASPubMedMATHGoogle Scholar
Wang, W. et al. A conserved N-terminal motif of CUL3 contributes to assembly and E3 ligase activity of CRL3(KLHL22). Nat. Commun. 15, 3789 (2024).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Canning, P., Sorrell, F. J. & Bullock, A. N. Structural basis of Keap1 interactions with Nrf2. Free Radic. Biol. Med 88, 101–107 (2015).
CASPubMedPubMed CentralMATHGoogle Scholar
Chen, Z., Picaud, S., Filippakopoulos, P., D’Angiolella, V. & Bullock, A. N. Structural Basis for Recruitment of DAPK1 to the KLHL20 E3 Ligase. Structure 27, 1395–1404.e1394 (2019).
CASPubMedPubMed CentralGoogle Scholar
Chen, Z. et al. Identification of a PGXPP degron motif in dishevelled and structural basis for its binding to the E3 ligase KLHL12. Open Biol. 10, 200041 (2020).
CASPubMedPubMed CentralGoogle Scholar
Schumacher, F. R., Sorrell, F. J., Alessi, D. R., Bullock, A. N. & Kurz, T. Structural and biochemical characterization of the KLHL3-WNK kinase interaction important in blood pressure regulation. Biochem J. 460, 237–246 (2014).
CASPubMedGoogle Scholar
Padmanabhan, B. et al. Structural basis for defects of Keap1 activity provoked by its point mutations in lung cancer. Mol. Cell 21, 689–700 (2006).
CASPubMedMATHGoogle Scholar
Lo, S. C., Li, X., Henzl, M. T., Beamer, L. J. & Hannink, M. Structure of the Keap1:Nrf2 interface provides mechanistic insight into Nrf2 signaling. EMBO J. 25, 3605–3617 (2006).
CASPubMedPubMed CentralGoogle Scholar
Lombardi, P. M., Cole, K. E., Dowling, D. P. & Christianson, D. W. Structure, mechanism, and inhibition of histone deacetylases and related metalloenzymes. Curr. Opin. Struct. Biol. 21, 735–743 (2011).
CASPubMedPubMed CentralMATHGoogle Scholar
Kelly, R. D. W. et al. Histone deacetylase (HDAC) 1 and 2 complexes regulate both histone acetylation and crotonylation in vivo. Sci. Rep. 8, 14690 (2018).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Hosseinzadeh, P. et al. Anchor extension: a structure-guided approach to design cyclic peptides targeting enzyme active sites. Nat. Commun. 12, 3384 (2021).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Porter, N. J., Christianson, N. H., Decroos, C. & Christianson, D. W. Structural and functional influence of the glycine-rich loop g(302)gggy on the catalytic tyrosine of histone deacetylase 8. Biochemistry 55, 6718–6729 (2016).
CASPubMedMATHGoogle Scholar
Porter, N. J. & Christianson, D. W. Structure, mechanism, and inhibition of the zinc-dependent histone deacetylases. Curr. Opin. Struct. Biol. 59, 9–18 (2019).
CASPubMedPubMed CentralMATHGoogle Scholar
Moradei, O. M. et al. Novel aminophenyl benzamide-type histone deacetylase inhibitors with enhanced potency and selectivity. J. Med Chem. 50, 5543–5546 (2007).
CASPubMedMATHGoogle Scholar
Lauffer, B. E. et al. Histone deacetylase (HDAC) inhibitor kinetic rate constants correlate with cellular histone acetylation but not transcription and cell viability. J. Biol. Chem. 288, 26926–26943 (2013).
CASPubMedPubMed CentralGoogle Scholar
Cao, S. et al. Defining molecular glues with a dual-nanobody cannabidiol sensor. Nat. Commun. 13, 815 (2022).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Kozicka, Z. et al. Design principles for cyclin K molecular glue degraders. Nat. Chem. Biol. 20, 93–102 (2024).
CASPubMedGoogle Scholar
Kozicka, Z. & Thoma, N. H. Haven’t got a glue: Protein surface variation for the design of molecular glue degraders. Cell Chem. Biol. 28, 1032–1047 (2021).
CASPubMedMATHGoogle Scholar
Millard, C. J., Fairall, L., Ragan, T. J., Savva, C. G. & Schwabe, J. W. R. The topology of chromatin-binding domains in the NuRD deacetylase complex. Nucleic Acids Res. 48, 12972–12982 (2020).
CASPubMedPubMed CentralGoogle Scholar
Yeo M. J. R. et al. UM171 glues asymmetric CRL3-HDAC1/2 assembly to degrade CoREST corepressors. Nature 639, 232–240 (2025).
Xie X. et al. Converging mechanism of UM171 and KBTBD4 neomorphic cancer mutations. Nature 639, 241–249 (2025).
Deng, Y. et al. SPOP point mutations regulate substrate preference and affect its function. Cell Death Dis. 15, 172 (2024).
CASPubMedPubMed CentralMATHGoogle Scholar
Duda, D. M. et al. Structural insights into NEDD8 activation of cullin-RING ligases: conformational control of conjugation. Cell 134, 995–1006 (2008).
CASPubMedPubMed CentralMATHGoogle Scholar
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Pettersen, E. F. et al. UCSF Chimera–a visualization system for exploratory research and analysis. J. Comput Chem. 25, 1605–1612 (2004).
CASPubMedMATHGoogle Scholar
Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr D. Biol. Crystallogr 60, 2126–2132 (2004).
ADSPubMedMATHGoogle Scholar
Liebschner, D. et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr D. Struct. Biol. 75, 861–877 (2019).
ADSCASPubMedPubMed CentralMATHGoogle Scholar
Download references
Acknowledgements
Cryo-EM studies were performed at the UK national Electron Bio-Imaging Centre (eBIC, proposal BI34631), the Central Oxford Structural Molecular Imaging Centre (COSMIC, supported by the Wellcome Trust grant no. 201536, the EPA Cephalosporin Trust and a Royal Society/Wolfson Foundation Laboratory Refurbishment grant no. WL160052), and the Oxford Particle Imaging Centre (OPIC, an Instruct-ERIC centre funded by Wellcome Trust JIF award 060208/Z/00/Z and equipment grant 093305/Z/10/Z). ZC, SCC, VD’A and ANB acknowledge funding from a Cancer Research UK grant DRCNPG-May21\100002. This study was also possible thanks to funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement number 875510 (EUbOPEN). JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA and Ontario Institute for Cancer Research, Royal Institution for the Advancement of Learning McGill University, Kungliga Tekniska Hoegskolan, and Diamond Light Source. The scientific work/research and/or results publicised in this article were reached with the sponsorship of the Gedeon Richter Talentum Foundation in the framework of the Gedeon Richter Excellence PhD Scholarship of Gedeon Richter. We thank Brenda Schulman, Cheryl Arrowsmith and Melissa Sweeney for help with constructs.
Author information
Authors and Affiliations
Centre for Medicines Discovery, University of Oxford, Oxford, OX3 7FZ, UK
Zhuoyao Chen, Gamma Chi, Xiangrong Chen, Beatriz Ralsi Montes & Alex N. Bullock
Servier Research Institute of Medicinal Chemistry, Zahony u. 7, H‐1031, Budapest, Hungary
Timea Balo, Timea Szabo, Arpad Kiss, Tibor Novak, András Herner & András Kotschy
Hevesy György Ph.D. School of Chemistry, Eötvös Loránd University, Pázmány Péter sétány 1/A, H-1117, Budapest, Hungary
Timea Balo
Wolfson Childhood Cancer Research Centre, Newcastle University Centre for Cancer, Newcastle upon Tyne, NE1 7RU, UK
Steven C. Clifford
The Institute of Genetics and Cancer, University of Edinburgh, Crewe Road South, Edinburgh, EH4 2XU, UK
Vincenzo D’Angiolella
Authors
Zhuoyao Chen
View author publications
You can also search for this author inPubMedGoogle Scholar
2. Gamma Chi
View author publications
You can also search for this author inPubMedGoogle Scholar
3. Timea Balo
View author publications
You can also search for this author inPubMedGoogle Scholar
4. Xiangrong Chen
View author publications
You can also search for this author inPubMedGoogle Scholar
5. Beatriz Ralsi Montes
View author publications
You can also search for this author inPubMedGoogle Scholar
6. Steven C. Clifford
View author publications
You can also search for this author inPubMedGoogle Scholar
7. Vincenzo D’Angiolella
View author publications
You can also search for this author inPubMedGoogle Scholar
8. Timea Szabo
View author publications
You can also search for this author inPubMedGoogle Scholar
9. Arpad Kiss
View author publications
You can also search for this author inPubMedGoogle Scholar
10. Tibor Novak
View author publications
You can also search for this author inPubMedGoogle Scholar
11. András Herner
View author publications
You can also search for this author inPubMedGoogle Scholar
12. András Kotschy
View author publications
You can also search for this author inPubMedGoogle Scholar
13. Alex N. Bullock
View author publications
You can also search for this author inPubMedGoogle Scholar
Contributions
S.C.C., V.D.’A., An.K. and A.N.B. conceived the project. Z.C., X.C. and B.R.M. undertook protein production. Cryo-EM studies were performed by Z.C. and G.C. MST experiments were performed by Z.C. and X.C. UM171 derivatives were designed and/or synthesized by T.B., T.S., Ar.K., T.N., A.H. and An.K. Cellular studies for UM171 SAR were performed by T.B. and Z.C. and analyzed by T.B., Z.C., T.S., A.H., Ar.K. and An.K. Other biochemical and cellular studies were performed by Z.C. Z.C. and G.C. prepared the figures. Z.C. wrote the initial manuscript with G.C., T.B., An.K. and A.N.B. All authors discussed and edited the manuscript.
Corresponding author
Correspondence to Alex N. Bullock.
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Supplementary Information
Reporting Summary
Transparent Peer Review file
Source data
Source Data
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and permissions
About this article
Check for updates. Verify currency and authenticity via CrossMark
Cite this article
Chen, Z., Chi, G., Balo, T. et al. Structural mimicry of UM171 and neomorphic cancer mutants co-opts E3 ligase KBTBD4 for HDAC1/2 recruitment. Nat Commun 16, 3144 (2025). https://doi.org/10.1038/s41467-025-58350-z
Download citation
Received:12 September 2024
Accepted:20 March 2025
Published:02 April 2025
DOI:https://doi.org/10.1038/s41467-025-58350-z
Share this article
Anyone you share the following link with will be able to read this content:
Get shareable link
Sorry, a shareable link is not currently available for this article.
Copy to clipboard
Provided by the Springer Nature SharedIt content-sharing initiative