BAP1/ASXL1 recruitment and activation for H2A deubiquitination

Abstract

The deubiquitinating enzyme BAP1 is an important tumor suppressor that has drawn attention in the clinic since its loss leads to a variety of cancers. BAP1 is activated by ASXL1 to deubiquitinate mono-ubiquitinated H2A at K119 in Polycomb gene repression, but the mechanism of this reaction remains poorly defined. Here we show that the BAP1 C-terminal extension is important for H2A deubiquitination by auto-recruiting BAP1 to nucleosomes in a process that does not require the nucleosome acidic patch. This initial encounter-like complex is unproductive and needs to be activated by the DEUBAD domains of ASXL1, ASXL2 or ASXL3 to increase BAP1's affinity for ubiquitin on H2A, to drive the deubiquitination reaction. The reaction is specific for Polycomb modifications of H2A as the complex cannot deubiquitinate the DNA damage-dependent ubiquitination at H2A K13/15. Our results contribute to the molecular understanding of this important tumor suppressor.

Figure 1. The DEUBAD domain in ASXL proteins activates BAP1 to specifically deubiquitinate K119 on H2A.

(a) Overview of constructs. Locations of point mutations indicated with filled triangles. The asterisk highlights constructs not used in this study. (b) The DEUBAD domain but not the HARE-HTH domain of ASXL1 can stimulate BAP1 activity in an assay against minimal substrate Ub-AMC. Error bars, s.d. (n=3 independent experiments). (c) In Ub-AMC Michaelis–Menten analysis the BAP1/ASXL1^DEU complex (closed circles) has a lower K_M than BAP1 alone (open circles). Error bars, s.d. (n=3 independent experiments). (d) Anti-H2A western blot showing that the DEUBAD domain but not the HARE-HTH domain of ASXL1^DEU can stimulate BAP1 activity in a deubiquitination assay against K119 mono-ubiquitinated H2A in oligonucleosomes (time points: 0, 1, 5, 25 min). BAP1 alone reactions are indicated with ‘−'. (e) The DEUBAD domains of ASXL2 and ASXL3 can also activate BAP1 deubiquitination of oligonucleosomal H2A as shown in an anti-H2A western blot. (f) An anti-H2A western blot shows that the BAP1/ASXL1^DEU complex has low activity against H2A mono-ubiquitinated at K13/15 in oligonucleosomes, compared with H2A K119 ubiquitinated oligonucleosomes (time points: 0, 1, 5, 25 min). (g) On minimal peptide substrates derived from H2A, where either K13 or K119 are ubiquitinated chemically, the BAP1/ASXL1^DEU complex can deubiquitinate K13-linked ubiquitin as efficiently as K119-linked ubiquitin as visualized on SDS-PAGE coomassie.

Figure 2. The BAP1 ULD binds ASXL1^DEU.

(a) BAP1 and UCH-L5 are highly similar in the catalytic domain and ULD. UCH-L5 does not contain the insert that BAP1 has as indicated by a dashed line. The ULD anchor is highly conserved in a multiple sequence alignment in both BAP1 and UCH-L5 (inset, ULD anchor residues indicated with triangle) (b) Cartoon representation of the UCH-L5/RPN13^DEU structure(PDB: 4uem ref. , UCH-L5 CD in blue, UCH-L5 ULD in light blue, RPN13^DEU in green) with labels indicating equivalent BAP1 positions. The predicted location of the BAP1 insert is highlighted (c) Isothermal titration calorimetry (ITC) assays. By titrating ASXL1^DEU from the syringe to either full-length BAP1 (5 μl injections) or BAP1_1–710 or BAP1_1–670 in the cell (both 10 μl injections), we show that ASXL1^DEU binds the C-terminal ULD of BAP1 but that the BAP1 CTE is not required for ASXL1^DEU binding.

Figure 3. ASXL1^DEU activates BAP1 by increasing its affinity for ubiquitin.

(a) Ub-AMC hydrolysis assay of BAP1 and BAP1/ASXL1^DEU in the presence of an H2A peptide that is conjugated to ubiquitin at K119 in a non-hydrolyzable fashion. This conjugate mimics substrates and therefore functions as an inhibitor for Ub-AMC hydrolysis. BAP1 is inhibited more readily by the non-hydrolyzable ubiquitin conjugate in the presence of ASXL1^DEU. Inset shows the schematic of the non-hydrolyzable H2A ubiquitin conjugate. Error bars, s.d. (n=2 independent experiments). (b) Native gel shift assay of NCPs mono-ubiquitinated with ^TAMRAubiquitin and BAP1 or the BAP1/ASXL1^DEU complex. In the presence of ASXL1^DEU, BAP1 has a higher affinity for mono-ubiquitinated NCPs than BAP1 alone (twofold dilutions starting from 15 μM). Band visualized using the TAMRA signal (c) Using stopped-flow fluorescence polarization binding assays we confirmed that the BAP1/ASXL1^DEU complex (top) binds better to K119 mono-ubiquitinated NCPs than BAP1 alone (bottom). (d) BAP1 and BAP1/ASXL1^DEU bind unmodified NCPs with similar affinities in band-shift assays (twofold dilutions starting from 15 μM). Bands visualized using DNA staining by GelRed. (e) Anti-H2A western blot showing that the BAP1 ULD anchor mutant (D663A/R667A, DR) complex is less active in deubiquitinating oligonucleosomal H2A than the WT complex (time points: 0, 1, 5, 25 min) (f) In Ub-AMC hydrolysis assays, BAP1_DR/ASXL1^DEU complex has a threefold weaker K_M than the WT complex. Error bars, s.d. (n=3 independent experiments). (g) ASXL1_NEF^DEU (N310A/E311K/F312A, NEF) binds BAP1 with similar affinity as WT ASXL1^DEU in ITC. (h and i) ASXL1_NEF^DEU cannot activate BAP1 to the same extent as WT ASXL1^DEU using K119 mono-ubiquitinated oligonucleosomal H2A as visualized by western blot (time points: 0, 1, 5, 25 min) or Ub-AMC as a substrate. Error bars, s.d. (n=3 independent experiments). (j) Surface representation of DEUBAD domains of RPN13 (left) and INO80G (right) indicate high conservation of a region (dotted circle) between helix 4 and 5 (middle: RPN13 green, INO80G orange). Surface conservation was calculated using ConSurf.

Figure 4. The BAP1 C-terminal extension is required for H2A deubiquitination.

(a) As shown in an anti-H2A western blot, only WT BAP1 can be activated by ASXL1^DEU to cleave ubiquitin from H2A K119 in oligonucleosomes (time points: 0, 1, 5, 25 min). (b) All BAP1 variants are active on Ub-AMC. Purple bars represent rates in presence of ASXL1^DEU. Dotted black and purple lines highlight the rate of WT BAP1 or BAP1/ASXL1^DEU respectively. Error bars, s.d. (n=3 independent experiments). (c): Michaelis–Menten enzyme kinetics showing that like WT BAP1 (blue, curves from Fig. 1b), BAP1_1–710 (black) can be fully activated by ASXL1^DEU (reactions with ASXL1^DEU in filled circles) on the minimal substrate Ub-AMC. Error bars, s.d. (n=3 independent experiments).

Figure 5. The BAP1 C-terminal extension auto-recruits BAP1 to nucleosomes.

(a) The nucleosomal H2A substrate can be decomposed into two functional parts: (1) the core consisting of the histone octamers wrapped in DNA and (2) the ubiquitinated H2A tail protruding from the core. (b) Coomassie gel of DUB assay using a minimal K119 H2A-ub peptide conjugate substrate. While the BAP1_1–710/ASXL1^DEU complex is inactive on oligonucleosomal H2A, it is as active as the WT complex on the minimal H2A substrate that represents the ubiquitinated H2A tail in isolation (time points: 5, 15, 30 min). (c) Multiple sequence alignment of the C-terminus of BAP1 shows that this tail is highly positively charged. (d) Anti-H2A blot of a oligonucleosomal H2A K119 deubiquitination assay. Titrating in a synthetic peptide of the BAP1 C-terminus inhibits activity of BAP1/ASXL1^DEU (time points: 0, 1, 5, 25 min). (e) BAP1 does not require the H2A/H2B acidic patch to bind NCPs, since it shifts WT NCPs as efficient as NCPs where the acidic patch is mutated (EA-NCPs) in band-shift assays (twofold dilutions starting from 15 μM). Shift visualized by DNA staining using GelRed. (f) BAP1 that lacks its C-terminal extension (BAP1_1–710) is impaired in binding NCPs in band-shift assays compared with WT (1, 5, 10 μM of BAP1 variant).

Figure 6. Model for H2A deubiquitination by the BAP1/ASXL1 complex.

BAP1 (blue) binds nucleosomes mono-ubiquitinated at K119 using its C-terminal extension (red). This complex has low DUB activity owing to BAP1's low affinity for ubiquitin on H2A (orange). Binding of the DEUBAD domain of ASXL1 (purple) allosterically activates BAP1 by increasing its affinity for the ubiquitin moiety on H2A, driving the DUB reaction.