Structural mechanism for the recognition and ubiquitination of a single nucleosome residue by Rad6–Bre1

Bre1–NCP Interaction Topology Determined by XL-MS.

To map the NCP interaction sites of Rad6 and Bre1, we used a recombinant Rad6∼Bre1 fusion, in which the C terminus of Rad6 was connected to the N terminus of the Bre1 eRING domain (extended RING; amino acids 568–700) via a flexible linker (Fig. 1A, lane1). This minimal construct specifically modifies H2B Lys123 and is biochemically stable (19). The fusion protein is more active than full-length Bre1 and Rad6 alone in an in vitro NCP ubiquitination assay, likely because the transient RING–E2 interaction is stabilized. We used NCPs reconstituted from polycistronically expressed yeast histones in Escherichia coli and the Widom 601 DNA sequence as a substrate. Titrating increasing amounts of the homobifunctional cross-linker DSS (disuccinimidyl suberate) to a 1:1 molar mixture of Rad6∼Bre1 and NCP lead to the formation of high molecular weight complexes (Fig. 1A and Fig. S1A). DSS induces nucleophilic attacks on primary amines and thereby couples lysine residues. Notably, of all histone proteins, histone H2B was preferentially cross-linked with increasing amounts of DSS (compare histone stoichiometry, e.g., lane 2 vs. 5 in Fig. 1A), which parallels the reduction and upward shift of Rad6∼Bre1. This finding indicates a preferential cross-linking and hence spatial proximity between Rad6∼Bre1 and the target protein. Subsequent MS analysis revealed 94 high-confidence interlinks and 100 intralinks among histones (Fig. S1B and Dataset S1). Ninety-eight intralinks were identified within Rad6∼Bre1 (Fig. 1B and Dataset S1), consistent with the previously reported direct interaction between Rad6 and the Bre1 eRING domain (19) as well as Bre1 homodimerization (24). Of significance, only two unique interlinks between Rad6∼Bre1 and the NCP were identified (Fig. 1B). These high-confidence linkages correspond to a single peptide within the Bre1 RING domain that cross-links exclusively to two different peptides within histone H2B and H2A. No cross-links between Rad6 and histones were found. The H2B peptide maps to the C-terminal α-helix of the protein, which harbors the Lys123 ubiquitination site, and is located surface-exposed at the DNA edge opposite of the NCP dyad. The cross-linked H2A peptide maps to a C-terminal extension at the opposite DNA edge next to the NCP dyad. Notably, despite the large number of available lysine residues in the NCP (114 Lys in total, including 54 Lys in the tails), only two lysines were specifically cross-linked to Bre1. This finding suggests that the E3 binds in a defined orientation on the disk face of the NCP. The cross-linking of one Bre1 peptide to two different histone peptides is consistent with the symmetric arrangement and the overall size of the dimeric RING domain (24), as one RING monomer could contact H2B and the other H2A. A line drawn between the identified H2A and H2B peptides (Fig. 1B) runs directly over a prominent feature of the NCP surface, the acidic patch, a groove formed by conserved aspartate and glutamate residues from H2A and H2B (Fig. 1C). Remarkably, the Bre1 RING domain features four highly conserved basic residues, which are surface-exposed (24) and could readily direct Bre1 to the NCP acidic patch. Taken together, the XL-MS–derived spatial restraints suggested a specific interaction topology that is encoded in the electrostatic complementarity of basic RING and acidic NCP residues, a hypothesis that we tested further, as described below.

The NCP Acidic Patch Directly Recruits Bre1 to Monoubiquitinate H2B.

To determine whether Bre1 directly binds the NCP acidic patch, we first attempted to inhibit the interaction with a viral peptide that targets the same region. The N terminus of the herpes virus latency-associated nuclear antigen (LANA) forms a hairpin that inserts into the H2A/H2B acidic patch as seen in the LANA–NCP cocrystal (25). Of note, the overall structure of the NCP is not perturbed by LANA binding. Increasing amounts of a 23-residue LANA peptide blocked binding of full-length Bre1 to the NCP in pull-down assays, consistent with LANA obstructing Bre1 access to the acidic patch. In contrast, a mutant LANA peptide (L₈R₉S₁₀ > AAA) with a reduced acidic patch affinity (25) was less efficient in competing with Bre1 for NCP binding (Fig. 2A). As a result of the impaired NCP interaction, Bre1 exhibited a reduced H2B monoubiquitination activity in an in vitro assay, whereas the mutant LANA peptide had little inhibitory effect (Fig. 2B). These results demonstrate that an accessible acidic patch is required for efficient NCP recognition and ubiquitination of H2B Lys123. To corroborate these findings, we reconstituted NCPs carrying point mutations in key acidic residues. Notably, alanine substitutions of H2A Glu57 and H2B Glu116, which lie close to the target Lys123 (Fig. 1C), impaired Bre1 binding to the mutant NCP and this effect was enhanced when both mutations were combined (Fig. 2C). Consequently, all these acidic patch mutations inhibited H2B monoubiquitination in vitro (Fig. 2D). This finding demonstrates that an accessible and intact acidic patch is required for Bre1 recruitment to the NCP in agreement with the interaction topology obtained by XL-MS.

XL-MS Captures the Transient Bre1 RING–Rad6 Interaction.

A key question that emerges is how electrostatic interactions between Bre1 and the NCP, spatially distinct from the site of catalysis, would position Rad6 such that it targets Lys123, yet no other histone Lys residue. The Bre1 RING domain directly interacts with Rad6 and is sufficient to monoubiquitinate H2B Lys123 in vitro (19). We used the known crystal structures of the Bre1 RING domain and Rad6 to generate a homology model of the E2–E3 complex (24, 26). Superposition of Bre1 RING and Rad6 onto the cCbl–UbcH7 crystal structure (27), followed by superposition onto RNF4–UbcH5∼Ub (18) generated a dimeric Bre1 RING–Rad6∼Ub model with no severe steric collisions (Fig. 3A and Fig. S2A). Importantly, the interface between Rad6 and the Bre1 RING in our model is in agreement with XL-MS–derived contact points. Notably, a single Rad6 peptide that maps to loop 4 (L4), cross-linked to two peptides located on the opposite Bre1 RING surface. The measured distance between cross-linked lysine residues in the Bre1–Rad6 model satisfies the maximal distance restraints imposed by the DSS cross-linker (<26 Å between Cα atoms) (28) (Fig. 3B) and supports the validity of the model. Other cross-links between Rad6 and Bre1 include peptides that are not present in the partial Bre1 structure (e.g., Bre1 residues 568–623) (24). Taking these data together, we find that XL-MS enabled us to monitor the highly dynamic E3–E2 interaction, providing concise spatial restraints for the Rad6–Bre1 interface.

Specific Bre1 Residues Required for Rad6 Activation and NCP Recognition.

We next sought to validate the E2–E3 interface by mutating conserved Bre1 amino acids while leaving the structurally relevant zinc-coordinating residues intact. After screening an extensive number of recombinant constructs, we finally succeeded in generating stable Leu650 > Ala and Arg675 > Asp mutants of Bre1, both of which map to the predicted Rad6 interface (Figs. 3B and 4A). To directly measure their effect on Rad6 activation, we analyzed the discharge of the ubiquitin thioester (∼Ub) from the active site of Rad6 in single turnover assays (Fig. 4A, Right, and Fig. 4B). To this end, Rad6 was charged with ubiquitin by the E1 enzyme and ATP and then treated with EDTA to inhibit recharging. Rad6 is known to undergo spontaneous Ub discharge independently of Bre1 over time (Fig. 4A, Right, compare lane 1 and 3; and Fig. 4B) (19, 29). However, adding a wild-type Bre1 eRING (amino acids 494–700) significantly accelerated the reaction. In contrast, Ub discharge was impaired in the Bre1 Arg675 > Asp and to a lesser extent in the Leu650 > Ala mutant. The involvement of Arg675 suggests a complementary charge on the Rad6 side. Indeed, three acidic Rad6 residues (Asp60, Glu61, and Glu62) are located within the cross-linked peptide on loop L4, ∼10–13 Å away from Arg675 in our model (Fig. S2B). Reduced Ub discharge from Rad6 could arise from a decreased affinity of Rad6 toward the Bre1 RING mutants. Alternatively, the mutant RING domains may specifically affect Rad6 activation. To distinguish between these possibilities, we performed in vitro binding assays. Notably, Rad6 binding to the Bre1 RING mutants was not perturbed (Fig. 4C and Fig. S2C). Thus, we conclude that these Bre1 RING residues are required for stimulating Rad6 catalytic activity, possibly through an allosteric effect. Whereas Bre1 Arg675 is located on a short α-helix with its side chain pointing laterally toward Rad6, three other conserved basic residues (Arg679, Arg681, Lys682) are part of a surface-exposed loop with their side chains oriented toward the “base” of the Bre1 RING homodimer (24) (Fig. 3 B and C). To pinpoint the function of single residues to Rad6 activation and NCP recognition, we analyzed them individually by inverting their charge. Notably, mutation of the basic Bre1 RING residues affected NCP recognition and the combined Arg681/Lys682 > Asp mutant exhibited the strongest loss of affinity (Fig. 4D), likely caused by an electrostatic repulsion with the NCP acidic patch. Interestingly, the Leu650 > Ala and Arg675 > Asp mutants were defective in both Rad6 activation and NCP recognition. This result could reflect an allosteric coupling of substrate recognition with ubiquitin discharge or other structural rearrangements. As a consequence of perturbing the NCP–RING interface, H2B ubiquitination was impaired in an in vitro assay (Fig. 4E). In sum, we could identify specific Bre1 residues required for Rad6 activation and NCP recognition, which are located at the protein interfaces suggested by XL-MS.

A Model for the E3–E2∼Ub–NCP Complex.

To integrate our experimental data, we sought to develop a structural model of the enzyme-substrate complex that fulfills the constraints derived from XL-MS, mutational analyses, and the general E2 catalytic mechanism. We used the Bre1 RING homodimer modeled in contact with one Rad6∼Ub under the assumption that only a single Rad6 will target H2B Lys123 (18). The complex was manually docked onto the NCP disk surface by apposing the critical basic Bre1 Arg679 and Arg681 residues to H2B Glu116, an acidic patch residue required for the direct interaction with Bre1 (Fig. 2C). Next, the Bre1 RING–Rad6∼Ub was rotated around the DNA superhelical axis of the NCP to approach the catalytic center of Rad6 (Cys88) toward the H2B Lys123 target residue (Fig. 5 A and B). This model was refined using all-atom energy minimization and subsequent molecular dynamics (MD) equilibration in explicit solvent. Notably, the resulting structural model brings the Rad6 catalytic Cys88 within a 10.6-Å distance to the H2B Lys123 (Fig. S3A). This is the closest distance of Cys88 to any histone Lys residue in its vicinity. Taking the intrinsic flexibility of the complex into account, this distance is reasonably close to a 3–3.5 Å upper limit for a ubiquitin transfer reaction to happen. Importantly, our model is in excellent agreement with all XL-MS distance restraints and lends strong support to the suggested mechanism for H2B Lys123 ubiquitination and recognition of the acidic patch.

The NCP Acidic Patch Recruits Opposing Enzymatic Activities.

Cellular H2B Lys123∼Ub levels likely reflect the balance of ubiquitin ligase and deubiquitinase activities. This finding raises the question whether the NCP acidic patch can regulate recruitment of competing enzymes. We previously implicated solvent-exposed basic residues on the SAGA H2B DUB module (Ubp8–Sgf11–Sgf73–Sus1) in NCP recognition (14). These basic residues map to a small ZnF domain of the DUB subunit Sgf11 and were recently confirmed to target the NCP surface, specifically the NCP acidic patch (30). In light of our findings on Rad6–Bre1, we sought to test whether basic residues on Bre1 and Sgf11 indeed compete for binding to the NCP acidic patch in solution. Thus, we carried out NCP ubiquitination reactions in the presence of Sgf11, which forms a stable heterodimer with Sus1 (14) (Fig. 5C). Notably, this DUB subcomplex could effectively inhibit H2B ubiquitination by Bre1–Rad6. In contrast, Sgf11 failed to inhibit the reaction when Arg84 and Arg91 were mutated, consistent with their requirement for NCP acidic patch recognition (Fig. 5C). These observations led us to test whether the NCP acidic patch mediates the functional opposition between Rad6–Bre1 and the SAGA DUB module in vivo. To this end, we assessed global histone H2B∼Ub levels in selected acidic patch mutants in S. cerevisiae. The H2B Glu116 > Ala mutant recapitulates the ubiquitination defect seen in vitro. This result is also observed upon mutating H2A Glu93, a residue located toward the center of the NCP disk (Figs. 1C and 5D, lanes 2 and 6), and is consistent with a previous report (23). In marked contrast, H2A Glu57 and Glu65 mutations showed increased H2B Lys123∼Ub levels. This result is different from our in vitro finding in which the H2A Glu57 mutant exhibited reduced Bre1 binding and H2B ubiquitination (compare Figs. 2D, lane 2, and 5D, lane 3). This interesting observation can be explained by an interference between H2B DUB and E2/E3 activities, which are only uncovered in an in vitro system that lacks DUB activity. Remarkably, both enzymes target the NCP acidic patch, but some residues, like H2A Glu57 and Glu65, appear to be more important for DUB than E2/E3 recruitment (Fig. S3 C and D). Of note, our in vivo result on H2A Glu65 is consistent with a key structural role of this residue in contacting the basic surface of the Sgf11 ZnF, which is also important for H2B deubiquitination in vitro (30) (Fig. S3D). In sum, we establish a molecular mechanism for Rad6–Bre1 recruitment to the NCP acidic patch that occurs in competition with specific elements of the SAGA deubiquitinase.

Protein Expression and Purification.

His6–tobacco etch virus (TEV)–Rad6, Strep–Bre1 constructs, GST–TEV–Sgf11 constructs, ubiquitin, and histones were cloned, expressed, and purified as described previously (14, 19). Mutations and deletions were generated by PCR-based methods and confirmed by sequencing. Strep–Rad6∼Bre1 RING fusion protein was affinity purified on a 5-mL Strep–Tactin Superflow column (GE Healthcare) in buffer containing 50 mM Tris, pH 8.0, 100 mM NaCl, 5 mM MgCl₂, and eluted by 2.5 mM desthiobiotin. Protein was further purified by size-exclusion chromatography on a Superdex 200 HiLoad 16/60 column (GE Healthcare) equilibrated with 50 mM Hepes, pH 8.0, 100 mM NaCl, 5 mM MgCl₂.

Histone Octamer Purification and NCP Reconstitution.

A single-step purification strategy for the histone octamer was used as described previously (19). Mutations on histones were generated by PCR-based methods and confirmed by sequencing. NCPs were reconstituted from histone octamers and the 167-bp 601 WIDOM positioning sequence by salt dialysis and further purified by size-exclusion chromatography on a Superdex 200 HiLoad 16/60 column (GE Healthcare) equilibrated in 20 mM Tris⋅HCl, pH 7.5. NCPs used for cross-linking experiments were purified in 20 mM Hepes, pH 8.0.

Chemical XL-MS.

Cross-linking analysis of Rad6∼Bre1 RING:NPC was performed by mixing purified components in 1:1 molar ratio with an equimolar mixture of light and heavy-labeled (deuterated) disuccinimidyl suberate DSS-H12/D12 (Creative Molecules). A final concentration of 0.24 mM DSS was used for the cross-linking reaction at 30 °C for 30 min. The reaction was quenched by adding final 100 mM of ammonium bicarbonate incubating for 10 min at 30 °C. Digestion with lysyl endopeptidase (Wako) was performed at 35 °C, 6 M Urea for 2 h (at enzyme-substrate ratio of 1:50 wt/wt) and was followed by a second digestion with trypsin (Promega) at 35 °C overnight (at 1:50 ratio wt/wt). Digestion was stopped by the addition of 1% (vol/vol) trifluoroacetic acid (TFA). Acidified peptides were purified using C18 columns (Sep-Pak, Waters). The eluate was dried by vacuum centrifugation and reconstituted in water/acetonitrile/TFA, 75:25:0.1. Cross-linked peptides were enriched on a Superdex Peptide PC 3.2/30 column (300 × 3.2 mm) at a flow rate of 50 μL/min and water/acetonitrile/TFA, 75:25:0.1 as a mobile phase. Fractions of 100 μL were collected, dried, and reconstituted in 2% acetonitrile and 0.2% FA, and further analyzed by liquid chromatography coupled to tandem MS using an LTQ Orbitrap Elite instrument (Thermo Scientific). Cross-linked peptides were identified using xQuest (36). False-discovery rates (FDRs) were estimated by using xProphet (37) and results were filtered according to the following parameters: FDR < 0.05, minimum δ score = 0.85, MS1 tolerance window of –4 –4 ppm, ld-score > 22. The cross-links were visualized using the webserver xVis (38).

NCP Ubiquitination Assay.

For NCP ubiquitination reactions, 1.2 µM NCP, 100 nM E1, 3 µM Rad6, 36 µM ubiquitin, 18 µM Bre1, 3 mM ATP, and 0.1 mM DTT were mixed in a reaction volume of 20 µL (50 mM Tris⋅HCl, pH 8.0, 50 mM NaCl, 50 mM KCl, 10 mM MgCl₂) and incubated for 90 min at 30 °C (unless otherwise indicated in figure legends). Reactions were stopped by addition of 4× SDS loading buffer and analyzed by SDS/PAGE (NuPAGE 4–12% gels; Invitrogen, MOPS buffer) followed by immunoblotting with anti-FLAG M2 antibody (Sigma, F1804).

In Vitro Binding Assays.

E2 Discharging Assay.

Rad6 precharging was performed as described previously (19). Then, 100 µM of Bre1 eRING constructs were added to the charged E2 in buffer containing 50 mM CHES, pH 8.0, 100 mM NaCl. Samples were incubated at 30 °C for 0, 10, and 20 min. Reactions were terminated by adding stop buffer [60 mM Tris⋅HCl, pH 6.8, 2% (wt/vol) SDS, 10% (vol/vol) glycerol, 0.005% Bromphenol blue] and kept on ice until analysis by nonreducing SDS/PAGE (12% NuPAGE gels, MES buffer) and by immunoblotting with anti-His IgG2a antibody (GE Healthcare, 27471001). Blots were visualized with a ChemiDoc Touch Imaging System (Bio-Rad), and quantification of band intensities was performed with ImageLab 5.2.1 software (Bio-Rad).

Measurement of Global H2B∼Ubiquitin Levels.

FLAG-H2B yeast strains were from a mutant collection (22). Immunoprecipitation of FLAG-tagged histone H2B was performed as described previously (14).

Energy Minimization and MD Equilibration.

Two manually assembled configurations of the Bre1–Rad6/NCP complex were refined using all-atom energy minimization and subsequent MD equilibration. The yeast NCP structure was taken from the Protein Databank (PDB ID code 1ID3) together with the yeast Bre1–Rad6 homology model (based on PDB ID codes 4R7E and 1AYZ) and were protonated and repaired for the missing atoms using PDB2PQR utility (39). All further calculations were performed using GROMACS 4.5.1 package (40) and Amber99SB-ILDN force-field parameters (41). The system was placed into a 13 × 13 × 13-nm³ rectangular box, initially energy minimized and solvated with TIP3P water molecules (42). Additionally, Na⁺ and Cl⁻ ions were added to achieve electroneutrality and at the final salt concentration of 80 mM. After this step, the complete system was energy-minimized using position restraints placed on protein Ca-atoms, Zn²⁺ ions in the Bre1 cysteine pockets and DNA P-atoms, and subjected to 30,000 steps of MD with a 0.5-fs time step and subsequent 250,000 steps of MD with a 1-fs time step. A twin-range (10/14 Å) spherical cut-off function was used to truncate the van der Waals interactions. Electrostatic interactions were treated using the particle-mesh Ewald summation (real space cut-off 10 and 1.2-Å grid with fourth-order spline interpolation). MD simulations were carried out using 3D periodic boundary conditions, in the isothermal−isobaric (NPT) ensemble with an isotropic pressure of 1.013 bar and a constant temperature of 310 K. The pressure and temperature were controlled using a Nose-Hoover thermostat (43) and a Parrinello–Rahman barostat (44) with 0.5- and 10-ps relaxation parameters, respectively, and a compressibility of 4.5 × 10⁻⁵ bar⁻¹ for the barostat. Bond lengths were constrained using LINCS (45).