Mechanism of ubiquitylation by dimeric RING ligase RNF4

Abstract

Mammalian RNF4 is a dimeric RING ubiquitin E3 ligase that ubiquitylates poly-SUMOylated proteins. We found that RNF4 bound ubiquitin-charged UbcH5a tightly but free UbcH5a weakly. To provide insight into the mechanism of RING-mediated ubiquitylation, we docked the UbcH5~ubiquitin thioester onto the RNF4 RING structure. This revealed that with E2 bound to one monomer of RNF4, the thioester-linked ubiquitin could reach across the dimer to engage the other monomer. In this model, the 'Ile44 hydrophobic patch' of ubiquitin is predicted to engage a conserved tyrosine located at the dimer interface of the RING, and mutation of these residues blocked ubiquitylation activity. Thus, dimeric RING ligases are not simply inert scaffolds that bring substrate and E2-loaded ubiquitin into close proximity. Instead, they facilitate ubiquitin transfer by preferentially binding the E2~ubiquitin thioester across the dimer and activating the thioester bond for catalysis.

Figure 1

Structure of dimeric RING domain of RNF4. (a) Single-turnover substrate ubiquitylation assay for RNF4. The UbcH5a~Ub thioester was incubated with ¹²⁵I-labeled 4 × SUMO-2 in the presence or absence of RNF4, followed by SDS-PAGE analysis and phosphorimaging. (b) Quantification of substrate ubiquitylation reactions shown in a. (c) Cartoon representation of the dimer of the RNF4 RING domain. Zinc ions are shown as grey spheres and zinc co-ordinating residues in stick representation. (d) A close up view of the dimerization interface of RNF4. Residues at the dimer interface are shown in stick representation. Atoms are color-coded as follows: nitrogens in blue, oxygens in red and sulphurs in yellow. (e) Mutational analysis of dimerization interface residues. Substrate ubiquitylation activity of dimerization mutants of RNF4 was determined using the single-turnover ubiquitylation assay described in panel a. Data represent mean ± s.d. of duplicate reactions. The upper panel shows a Coomassie blue-stained SDS-PAGE gel with purified wild-type (WT) and mutant RNF4 proteins (1 μg). (f) A model of a complex between the RNF4 RING domain and UbcH5b. Putative E2-binding residues of RNF4 that were mutated in this study are shown in stick representation. Active site cysteine of UbcH5b (Cys85) is also shown. (g) Substrate ubiquitylation activity of E2-binding mutants of RNF4. (f) A FRET-based in vitro assay to study dimerization of RNF4. ECFP-RNF4 (wild-type) was mixed with YFP or YFP-RING domain of RNF4 (wild-type or mutant as indicated) and FRET signal was measured. Data points represent mean ± s.d. of triplicate measurements.

Figure 2

Ubiquitylation by RNF4 can proceed both in cis and in trans. (a) Ubiquitylation by RNF4 can proceed in trans. Full-length RNF4 (0.55 μM) and the RNF4 RING domain (0.55 μM) were added to substrate ubiquitylation reactions as indicated and ubiquitylation activity was determined. (b) Schematic representation of the experiment in a. A heterodimer of the RNF4 RING domain and full-length RNF4 with disrupted E2-binding site (M140A R181A) should be active in a substrate ubiquitylation reaction provided ubiquitylation by RNF4 can proceed in trans. S2, SUMO-2; Ub, ubiquitin. (c) Schematic representation of a hypothesis behind the experiment in d. If ubiquitylation by RNF4 can proceed in cis, then a heterodimer of full-length RNF4 and the RNF4 RING domain with disrupted E2-binding site should possess substrate ubiquitylation activity. However, this heterodimer should be inactive if ubiquitylation can only proceed in trans. Therefore, addition of an excess of the RING domain with disrupted E2-binding site to wild-type RNF4 should result in inhibition of substrate ubiquitylation activity provided ubiquitylation cannot proceed in cis. (d) Ubiquitylation by RNF4 can proceed in cis. The RING domain of RNF4 with disrupted E2-binding site (RING M140A R181A) was added to RNF4 WT (0.275 μM) in 2, 20, and 200-times molar excess, respectively, and substrate ubiquitylation activity was determined. (e) Under the conditions used in d, substrate ubiquitylation activity is proportional to concentration of RNF4 in the reaction. In a, d, and e, the data are mean ± s.d. of duplicate reactions. The experiments were performed twice.

Figure 3

RNF4 preferentially binds ubiquitin-loaded E2 and activates the bond between E2 and ubiquitin. (a) RNF4 induces hydrolysis of E2~Ub linked via an oxyester bond. Phosphorimager scan of SDS-PAGE gels showing time course of UbcH5a (C85S)~ ¹²⁵I-Ub oxyester stability in the presence or absence of RNF4. (b) UbcH5a (C85S)~¹²⁵I-Ub oxyester was incubated with the RNF4 RING domain or full-length RNF4 (wild-type or mutant as indicated) and rates of oxyester hydrolysis were determined. Data are shown as mean ± s.d. of duplicate reactions. The experiment was performed two times. (c) RNF4 shows higher affinity for ubiquitin-charged E2 than for free E2 in a pull-down experiment. A mixture of UbcH5a (N77A C85S), ubiquitin, and the UbcH5a (N77A C85S)~Ub oxyester was incubated with MBP or MBP-RNF4 (wild-type or mutant), followed by purification on amylose beads. WB, Western blot. (d) A model of the RNF4 RING domain in complex with E2~ubiquitin thioester predicts an interaction between ubiquitin and the RING domain. The RING domain is shown in cartoon form, with one monomer in blue and the other monomer in cyan. Putative ubiquitin-interacting residues (Tyr193 and Leu152) are shown as spheres. Ubiquitin is shown in pink with the hydrophobic contact surface (Leu8, Ile44 and Val70) in black. The E2 is shown colored in wheat with the point of thioester attachment shown as a yellow sphere. (e) A close up view of the predicted interaction interface between the hydrophobic patch on ubiquitin and residues Tyr193 and Leu152 on the RING domain of RNF4.

Figure 4

The “Ile44-centered hydrophobic patch” on ubiquitin is required for RNF4-mediated ubiquitylation. (a) Mutations in ubiquitin, used in the experiment shown in b, do not affect formation of the UbcH5a~Ub thioester. UbcH5a was incubated with ubiquitin in the presence of Ube1 and ATP, followed by nonreducing SDS-PAGE analysis. (b) Effect of mutations in ubiquitin on substrate ubiquitylation activity of RNF4. 4 × SUMO-2, radiolabeled with iodine-125, was used as substrate. Data are shown as mean ± s.d. of duplicate reactions. The experiment was performed two times. (c) Mutation I44A in ubiquitin causes a modest defect in E3-independent transfer of ubiquitin to poly-L-lysine. UbcH5a~¹²⁵I-Ub thioester was incubated with poly-L-lysine and time points were taken as indicated. Subsequently, poly-L-lysine was purified on SP-sepharose resin and radioactivity captured on the beads was quantified by γ-counting. Initial rates were determined from the first three time points. (d) The “Ile44 patch” on ubiquitin is essential for RNF4-mediated cleavage of E2~ubiquitin oxyester. The UbcH5a (C85S)~Ub oxyester was incubated in the presence or absence of RNF4 and reaction progress was analyzed by SDS-PAGE, followed by staining with SYPRO Orange. Reaction rates represent mean ± s.d. of duplicate reactions. The experiment was performed three times. (e) The “Ile44 patch” on ubiquitin is required for the interaction between ubiquitin-loaded E2 and RNF4. A mixture of UbcH5a (C85S), ubiquitin, and the UbcH5a (C85S)~Ub oxyester was briefly incubated with either MBP or MBP-RNF4 immobilized on amylose beads, followed by a quick washing step. Bound material was resolved by SDS-PAGE.

Figure 5

Tyrosine 193, a residue located at the dimer interface of the RNF4 RING domain, is required for activation of the thioester bond in the E2~ubiquitin thioester. (a) The ability of RNF4 Y193H to dimerize was assessed using a FRET-based dimerization assay. Unlabeled RNF4 was titrated into a mixture of ECFP-RNF4 and YFP-RING domain and FRET signal was measured. Data points represent mean ± s.d. of triplicate measurements. (b) Mutational analysis of the predicted ubiquitin-binding site on the RNF4 RING domain. Several mutations of Tyr193 were generated. Whereas mutations to alanine and leucine disrupted dimerization, mutations to tryptophan and histidine retained the ability of RNF4 to dimerize. Substrate ubiquitylation activity of the mutant proteins was determined using the single-turnover assay described in Fig. 1a. (c) Tyr193 in RNF4 is required for efficient hydrolysis of the E2~ubiquitin oxyester. The UbcH5a (C85S)~Ub oxyester was incubated with RNF4 (wild-type or mutant as indicated) and the rate of oxyester hydrolysis was determined. (d) Mutation Y193H in RNF4 disrupts binding to the E2~ubiquitin oxyester. A pull-down experiment was performed as described in Fig. 4e. (e) Ubiquitylation activity of RNF4 Y193H can be rescued by addition of an inactive RNF4 mutant with disrupted E2-binding site (RNF4 M140A R181A). Substrate ubiquitylation activity was determined as described in Fig. 1a. Final concentration of RNF4 mutants in the reaction was either 0.55 μM (+) or 1.1 μM (++). In panels b, c, and e, data represent mean ± s.d. of duplicate reactions. The experiments were performed two times.

Figure 6

A linear fusion of full-length RNF4 and the RNF4 RING domain shows that E2-binding site in one RING domain and Tyr193 in the other RING are both required for ubiquitylation activity. (a) A model of a linear fusion of full-length RNF4 and the RING domain of RNF4, based on the structure of the RNF4 RING domain reported herein. A short linker between the C-terminus of full-length RNF4 and the N-terminus of the RING domain is shown as a red dashed line. Tyr193 at the dimer interface and residues important for E2 binding (Met140 an Arg181) are shown in stick representation. The linear fusion protein contains one substrate-binding site, two E2-binding sites, and two Tyr193 residues, one on each side of the dimerization interface. A schematic representation of the RNF4-RING fusion protein is shown on the right-hand side. (b) Substrate ubiquitylation activity of RNF4-RING linear fusion proteins, with mutations in full-length RNF4 and/or the RNF4 RING domain as indicated. Data are shown as mean ± s.d. of duplicate reactions. The experiment was performed twice. The upper panel shows a SYPRO Orange-stained SDS-PAGE gel with purified RNF4-RING fusion proteins (0.6 μg). (c) Mutation of Tyr193 in the RING domain that does not bind E2 causes an increase in K_M and a decrease in k_cat. Michaelis-Menten kinetics were determined by varying concentration of the UbcH5a~Ub thioester. (d) Binding of the E2~ubiquitin oxyester to RNF4-RING fusion proteins correlates well with their ubiquitylation activities. A pull-down experiment was performed as in Fig. 4e.