MUL1-RING recruits the substrate, p53-TAD as a complex with UBE2D2-UB conjugate.

The RING domain of MUL1 (RINGMUL1 ) alone mediates ubiquitylation of the p53-transactivation domain (TADp53 ). To elucidate the mechanism underlying the simultaneous recruitment of UBE2D2 and the substrate TADp53 by RINGMUL1 , we determined the complex structure of RINGMUL1 :UBE2D2 and studied the interaction between RINGMUL1 and TADp53 in the presence of UBE2D2-UB thioester (UBE2D2~UB) mimetics. The RINGMUL1 -binding induced the closed conformation of UBE2D2S22R/C85S -UBK48R oxyester (UBE2D2RS -UBR OE ), and strongly accelerated its hydrolysis, which was suppressed by the additional N77A-mutation of UBE2D2. Interestingly, UBE2D2S22R/N77A/C85S -UBK48R oxyester (UBE2D2RAS -UBR OE ) already formed a closed conformation in the absence of RINGMUL1 . Although TADp53 exhibited weak binding for RINGMUL1 or UBE2D2 alone, its binding affinity was enhanced and even further for RINGMUL1 :UBE2D2 and RINGMUL1 :UBE2D2RAS -UBR OE , respectively. The recognition of TADp53 by RINGMUL1 as a complex with UBE2D2~UB is related to the multivalency of the binding events and underlies the ability of RINGMUL1 to ubiquitylate the intrinsically disordered protein, TADp53 .

chemical shift

chemical shift perturbation

intrinsically disordered protein

isothermal titration calorimetry

residual dipolar coupling

RING domain of MUL1

size exclusion chromatography

p53‐transactivation domain

ubiquitin

UBE2D2–UB thioester

UBE2D2–UB isopeptide

UBE2D2–UB oxyester

Introduction

Mitochondrial E3 ubiquitin ligase 1 (MUL1), located in the mitochondrial outer membrane, regulates various biological processes, including mitochondrial dynamics, cell growth, apoptosis, and mitophagy through ubiquitylation and SUMOylation [1]. It is a potential therapeutic target for Parkinson's disease because its role is similar to that of the PINK1/Parkin pathway [2, 3]. MUL1 also has different names, such as mitochondrial‐anchored protein ligase (MAPL) [4], mitochondrial ubiquitin (UB) ligase activator of NF‐κB (MULAN) [5], growth inhibition, and death E3 ligase (GIDE) [6] and Hades [7]. Sequence‐based topology analysis indicated that the major portion of MUL1, which lies between two transmembrane α‐helices (residues 9–29 and 239–259), is located in the mitochondrial intermembrane region, while the C‐terminal RING domain (RINGMUL1) faces the cytoplasm [6]. RINGMUL1 activity is critical for the ubiquitylation of mitofusin, Akt, p53, and ULK1, and responsible for the SUMOylation of dynamin‐related protein 1 (Drp1) [1]. MUL1 plays a role in apoptosis via the direct regulation of apoptosis‐associated proteins such as NF‐κB, Akt, and p53, for which the activity of RINGMUL1 is critical [6, 7, 8, 9].

Mitochondrial E3 ubiquitin ligase 1 negatively regulates the exonuclear function of p53 in the mitochondria via ubiquitylation. Moreover, results from cell‐based in vivo and in vitro studies, including pull‐down and immunoprecipitation experiments, showed that RINGMUL1 alone results in the ubiquitylation of p53 [7]. Although the six Lys residues in the C‐terminal region of p53 are ubiquitylated by Mdm2 [10], the K24 residue of p53 is specifically ubiquitylated by the action of RINGMUL1 alone [7]. It has also been reported that RINGMUL1 ubiquitylates the transactivation domain (TADp53, residues 1–73) in the presence of three E2 enzymes (UBE2D1, D2, and D3), but not by UBE2L3 (UbcH7) [11].

Ubiquitylation is generally mediated by the action of the following three enzymes: UB‐activating E1, conjugating E2, and ligase E3. E3 plays a key role in determining the target specificity and catalysing UB‐transfer from E2 to the Lys side‐chain of the target protein, and is mainly classified into two groups according to their E2‐binding domains (HECT and RING/U‐box); Transfer of the attached donor UB (UBD) directly from E2 to the substrate protein is the key feature of ubiquitylation by RING‐E3 [12]. RING‐E3 ligase generally contains another domain or region that can facilitate the recruitment of a target protein [13]. Thus, the mechanism underlying the ubiquitylation of p53 by the action of RINGMUL1 alone seems to be unique. Although our recent NMR studies have indicated that non‐labelled RINGMUL1 clearly binds 15N‐labelled TADp53 (15NTADp53), its binding affinity (K d, 1.03 mm) is too weak to support the in vivo ubiquitylation activity of RINGMUL1 alone [14].

To elucidate the mechanism by which RINGMUL1 alone results in the recruitment and ubiquitylation of TADp53, we determined the crystal structure of the RINGMUL1:UBE2D2 complex, and then studied the detailed interactions between TADp53 and the RINGMUL1 complexes with UBE2D2 and UBE2D2~UB mimetics in solution. Additionally, we reported the distinguishing features of RINGMUL1 that markedly enhanced the hydrolysis rate of UBE2D2–UB oxyester (UBE2D2–UBOE), and different dynamic natures of UBE2D2–UBOE depending on the N77 residue of UBE2D2. The higher binding affinity of TADp53 for RINGMUL1:UBE2D2–UBOE than that for RINGMUL1 or UBE2D2 alone depended on the multivalency of their binding, which resulted from the innate characteristics of the intrinsically disordered protein (IDP), TADp53.

Results

UBE2D2 exhibits a weak binding affinity for TADp53

We first estimated the binding affinity between 15NTADp53 and UBE2D2 via chemical shift perturbation (CSP) experiments using two‐dimensional (2D) 1H–15N heteronuclear single quantum correlation (HSQC), since the ubiquitylation reaction is also mediated by the action of E2. Certainly, UBE2D2 also bound to 15NTADp53 (K d, 735 ± 39 μm; Fig. 1A and Table 1), in which the region containing residues A39 to D55 (AD39) displayed a higher CSP than that exhibited by the region containing residues, S15 to N29 (SN15). The marked decrease in the peak intensities of 15NTADp53 caused by the binding of RINGMUL1 with a higher K d value, compared to that caused by the same concentration of UBE2D2, was likely from the heterogeneity of chemical shifts (CSs) induced by the structural flexibility of RINGMUL1 (Fig. 1A,B).

Fig. 1

The interactions of TADp53, UBE2D2, and RINGMUL1 for different counterparts were studied by 1H–15N HSQC experiments. The CSPs of 15NTADp53 (A–D), 15NUBE2D2 (E, F), and 15NRINGMUL1 (G–I) in the presence of interacting counterparts are shown as bar plots. The residues for which the HSQC crosspeaks were not shown in the free form and the peaks disappeared during the CSP experiments are shown as negative and green bars, respectively. (J) The AD39‐binding surfaces of 15NUBE2D2 (panel‐F) are shown following the degree of CSPs: (a) the front and right side, residues 2–16 (α1) and 94–103 (N‐terminal α3), (b) the rear side, residues 70–75 (β4‐to‐loop) and 119–126 (α4). The reference orientation was defined for UBE2D2 (top, bottom, right and left, front and rear, respectively). The length of AD39 in an extended conformation is comparable to that of the long axis of UBE2D2.

Table 1

Binding constants (K d, μm) measured by NMR CSP and ITC (n, ΔH, kcal·mol−1) experiments.

	K d (μm)	n	ΔH (cal·mol−1)	ΔS (cal·mol−1·deg−1)
15NTADp53 + UBE2D2 a	735 ± 39	–	–	–
15NTADp53 + RINGMUL1:UBE2D2 a	171 ± 38	–	–	–
UBE2D2 + RINGMUL1	28 ± 15 b	0.71 ± 0.08 b	1479 ± 906 b	26.0 ± 4.3 b
	13 ± 3	0.65 ± 0.13	2518 ± 571	30.9
	43 ± 18	0.80 ± 0.18	858 ± 247	22.8
	28 ± 10	0.69 ± 0.20	1061 ± 381	24.4
RINGMUL1:UBE2D2 + AD39	184 ± 13 b	0.88 ± 0.01 b	−1134 ± 16 b	13.3 ± 0.1 b
	175 ± 34	0.89 ± 0.25	−1145 ± 372	13.4
	193 ± 29	0.87 ± 0.08	−1123 ± 130	13.2
RINGMUL1:UBE2D2 + TADp53	111 ± 31	1.01 ± 0.21	−673 ± 177	15.8
RINGMUL1:UBE2D2RAS–UBR OE + AD39	129 ± 22 b	1.29 ± 0.32 b	−629 ± 125 b	15.7 ± 0.7 b
	105 ± 20	1.65 ± 0.16	−598 ± 76	16.2
	151 ± 21	1.11 ± 0.16	−719 ± 122	15.1
	116 ± 16	0.95 ± 0.12	−466 ± 71	16.4
	143 ± 23	1.47 ± 0.16	−735 ± 102	15.1
RINGMUL1:UBE2D2RAS–UBR OE + TADp53	34 ± 16 b	0.67 ± 0.19 b	485 ± 71 b	20.6 ± 4.3 b
	22 ± 4	0.54 ± 0.03	380 ± 31	22.6
	26 ± 6	0.58 ± 0.05	538 ± 63	22.8
	57 ± 24	0.95 ± 0.33	516 ± 211	21.2
	33 ± 7	0.62 ± 0.05	505 ± 59	22.2

The HSQC peaks of the AD39 region of TADp53 were traced to obtain the K d values.

The average value and standard deviation of the multiple measurements.

The bold values are the average values of the multiple measurements.

Reciprocal CSP experiments of 15NUBE2D2 in the presence of TADp53 and AD39 showed that their binding surfaces of UBE2D2 were almost identical, and mainly localized at two regions (Fig. 1E,F) as follows: (a) the main location was in the vicinity of the RING‐binding site (α1 and N‐terminal α3), while (b) the other location was on the rear side of UBE2D2 (β4‐to‐loop and α4; Fig. 1J). The fact that the CSP amount of 15NUBE2D2 induced by the TADp53‐binding was considerably higher than that induced by the same concentration of AD39 indicated that the SN15 region also contributed to the binding of UBE2D2. We also ascertained the CSP amounts of 15NRINGMUL1 induced by TADp53 [14], and compared them to those corresponding to similar concentrations of AD39 and SN15. The CSP patterns of 15NRINGMUL1 demonstrated by TADp53, AD39, and SN15 were similar, but the CSP amount was higher for TADp53 (Fig. 1G–I). Since each RINGMUL1 and UBE2D2 displayed a weak affinity to TADp53, it was possible that the larger complexes (RINGMUL1:UBE2D2 or RINGMUL1:UBE2D2~UB) display a higher affinity for TADp53.

Crystal structure of RINGMUL1 in the RINGMUL1:UBE2D2 complex

The crystal structure of the RINGMUL1:UBE2D2 complex was solved at 2.7 Å resolution (Fig. 2A) by molecular replacement (MR) using the previous UBE2D2 coordinate. The structure of RINGMUL1 alone was solved at 1.8 Å resolution by the MR using the RINGMUL1 coordinate in the determined complex structure (Table S1). The crystal of UBE2D2 alone was also obtained during the crystal screens of the RINGMUL1:UBE2D2 complex (Table S1), and the structure (1.8 Å resolution) was almost identical with the previously reported structures (PDB, 2CLW, and 2ESK). Although we also attempted to co‐crystallize the RINGMUL1 complexes along with UBE2D2S22R/C85K–UBK48R isopeptide (UBE2D2RK–UBR IP), and UBE2D2S22R/N77A/C85S–UBK48R oxyester (UBE2D2RAS–UBR OE), and RINGMUL1:UBE2D2RAS–UBR OE complex with the 2,4‐dinitrophenyl dye‐linked AD39 peptide, in which the colour of the dye was used to determine the specific complex crystal containing the AD39 peptide, the results were unsuccessful.

Fig. 2

Characterizations of the RINGMUL1:UBE2D2 interaction by X‐ray and NMR. (A) Four conformers in the crystal structure of RINGMUL1:UBE2D2 are superimposed on the basis of UBE2D2. The pairwise RMSD values were calculated for the backbone atoms (CA, C, O, and N) of the well‐structured regions (UBE2D2, M1–Y145; RINGMUL1, A301–P348). The RMSD value of the RINGMUL1 part greatly increases from 0.36 to 1.81 Å, when their coordinates of four conformers were fixed to the superimposed UBE2D2. (B) The previous NMR ensemble structures (PDB, 6K2K; green) were overlaid with the six superimposed conformers of the RINGMUL1 crystal structure. (C) The 1H–15N RDC values of 15NRINGMUL1 measured in solution (open circle) were compared to the crystal structures of RINGMUL1 alone (blue) and RINGMUL1 in the complex with UBE2D2 (red). (D) The six conformers of the RINGMUL1 structure (blue) alone are superimposed with the four RINGMUL1 molecules of the RINGMUL1:UBE2D2 complex (red). The CSPs of 15NRINGMUL1 (E) and 15NUBE2D2 (F) in the presence of non‐labelled UBE2D2 and RINGMUL1, respectively are shown as bar plots with the same colours used before. (G) The CSPs of 15NUBE2D2 induced by RINGMUL1 (panel‐F) are mapped to the UBE2D2 structure. RINGMUL1 binding considerably perturbed the specific regions of UBE2D2 (α1‐to‐β2, and α3 segments) that are used for the binding of UBBS and the UBD. The residues of UBE2D2 that are close to L306 of RINGMUL1 (< 3.5 Å) are shown as a sphere model (boxed).

The failure of molecular replacement (MR) with the NMR ensemble structures of RINGMUL1 (PDB, 6k2k) could be attributed to the structural discrepancy in the region containing residues 329–334 (Fig. 2B). Since the crystal structures of RINGMUL1 are consistent with the 15N‐1H residual dipolar coupling (RDC) data of 15NRINGMUL1 measured in solution (Fig. 2C), the discrepancy might arise from missing NOE‐assignments during the previous automatic CYANA calculation. The greater deviation of the calculated RDC values among the six conformers of the higher resolution crystal structures of the RINGMUL1 alone reflected the existence of innate structural heterogeneity in solution, which substantiated the heterogeneous peak intensity of its HSQC spectrum [14]. The RDC values calculated for the four conformers of RINGMUL1 in the complex were less variable, indicating that its structure was ordered by the UBE2D2‐binding (Fig. 2C).

There are six and four molecules present in the asymmetric units of the RINGMUL1 and RINGMUL1:UBE2D2 crystals, respectively. Structural comparison between RINGMUL1 alone and complexes showed no appreciable change had occurred in RINGMUL1, except in the N‐ and C‐terminal regions of RINGMUL1 (Fig. 2D). R340 of RINGMUL1 belongs to the previously characterized F‐x‐(K/R) motif (F, a hydrophobic amino acid; x, the Cys of the Zn+2‐finger; I338‐C339‐R340 in RINGMUL1) of the RINGE3 domain that acts as a linchpin to enable the activation of E2~UB [15], and the side‐chain of R340 that is in close contact with Q92 of UBE2D2 in the RINGMUL1:UBE2D2 complex (Fig. 2A). Interestingly, the overlay of the four conformers of the RINGMUL1:UBE2D2 complex showed that the positional plasticity of RINGMUL1 molecules was relative to UBE2D2 (Fig. 2A). The backbone root‐mean‐square deviation (RMSD) values of RINGMUL1 and UBE2D2 among the four conformers in the complex structures are 0.59 ± 0.05 and 0.36 ± 0.04 Å, respectively. The RMSD value of four RINGMUL1 conformers after matching the positions of UBE2D2 is increased to 1.81 ± 0.46 Å.

RINGMUL1 induced CSP in the wider regions of 15NUBE2D2

Based on the complex structures, we examined the interaction between RINGMUL1 and UBE2D2 using CSP experiments. The HSQC spectrum of 15NRINGMUL1 exhibited high heterogeneity in the presence of only half‐molar UBE2D2 and many peaks disappeared (Fig. 2E and Fig. S1A), and many HSQC crosspeaks of 15NUBE2D2 also disappeared in the presence of RINGMUL1 (Fig. 2F and Fig. S1B), which were likely correlated with the positional plasticity of the crystal structures of RINGMUL1 and UBE2D2 (Fig. 2A). It has been reported that the allosteric effect of the non‐covalent binding of UB (UBBS) to the backside of UBE2D2 (α1‐to‐β2; Fig. 1J, bottom region) increases the binding affinity of RINGRNF38 to UBE2D2S–UBOE, and not to UBE2D2 (K d, 89, 73, 4.6, and 0.36 μm for UBE2D2, UBE2D2:UBBS, UBE2D2S–UBOE, and UBE2D2S–UBOE:UBBS, respectively. Moreover, it also increases the intrinsic lysine reactivity of UBE2D2~UB [16]. Interestingly, the RINGMUL1‐binding induced strong CSPs in the regions of 15NUBE2D2 (α1‐to‐β2; α2‐to‐the crossover α3), which were evidently distant from the direct RINGMUL1‐interacting regions (Fig. 2G) and were used for binding with UBBS and UBD.

L306 of RINGMUL1 fits in the cleft formed by α1 (R5, K8, E9, D12) and the N‐terminal α3 (T98) of UBE2D2 (Fig. 2G, inset), which may define the molecular basis of the occurrence of higher amounts of CSPs of UBE2D2 induced by the RINGMUL1‐binding along with the positional plasticity of RINGMUL1 in the complex. The corresponding residues of other RINGE3 domains are mostly conserved as Leu (4auq, RINGBIRC7; 5d1k, RINGRNF25; 5ulh, RINGRNF165; 5fel, RINGTRIM25; 5d0m, RINGARK2C; and 4pql, RINGRNF146), Ile (6w9d, RINGRNF12), and Met (6hpr, RINGCLAP1; 3eb6, RINGCLAP2; 4v3k, RINGRNF4; 4ap4, RINGRNF38), although these residues are changed to Cys (2yho, RINGIDOL), Ala (1fbv, RINGCBL), Gln (5mnj, RINGMDM2), and Glu (5vzw, RINGTRIM23) at certain instances. The molecular mechanism underlying the activity of UBBS remains unclear [17], and the finding that RINGMUL1 induces the exceptionally high levels of CSPs in the regions for inter‐ and intra‐molecular binding of UBBS and UBD supports the hypothesis that the binding of UBBS, UBD, and RINGMUL1 to UBE2D2 is linked allosterically.

The disappeared HSQC crosspeaks in the CSP experiments of UBE2D2 and RINGMUL1 are likely dependent on intermediate binding exchange of the NMR time scale (K d, ~ 1 μm), resulting in peak‐broadenings. Both the HSQC spectra of 15NUBE2D3 and 15NUBE2D3RS–UBOE could be traceable in the presence of E4BU [15], and the extensive peak disappearance observed in the case of RINGMUL1 was not noted. Through isothermal titration calorimetry (ITC) measurements, we determined the binding affinity to explain the reason for the larger CSP of UBE2D2 in the presence of RINGMUL1. The affinity of RINGMUL1 for UBE2D2 was relatively weaker (K d, 28 ± 15 μm) than the expected one (Table 1 and Fig. S2), since a binding event of this K d value likely causes a CS change during the CSP experiments. The binding of RINGMUL1 and UBE2D2 was an endothermic reaction (ΔH, 1497 ± 906 cal·mol−1; Table 1), and thus the formation of RINGMUL1:UBE2D2 should be entropy‐driven to accomplish the negative Gibbs free energy (ΔG = ΔH − TΔS). The K d value of RINGMUL1:UBE2D2 is still lower than those of the other RING/U‐box:E2 interactions characterized by NMR experiments (E4BU:UBE2D3RS, 97 μm; RINGBRIC7:UBE2D2, too week; RINGBRIC7:UBE2D2S–UBOE, 136 μm) [15, 18]. The higher binding affinity for UBE2D2 in addition to the positional plasticity of RINGMUL1 causes the apparent CSPs in large areas of UBE2D2. UBE2D2 might have an intrinsic dynamic motion in this area, and the RINGMUL1‐binding selects a specific conformation resulting in the CSP on a large area of UBE2D2. However, no clear relaxation dispersion curve supporting the presence of an exchanging motion of μs‐ms time scale was identified in the extreme CPMG experiment [19] of 15NUBE2D2 alone (not shown). Therefore, the large area CSP of UBE2D2 induced by the RINGMUL1‐binding does not seem to be dependent on the conformational selection mechanism of UBE2D2 itself.

Characterization of the RINGMUL1:UBE2D2~UB mimetics

Prior to investigation on the binding of RINGMUL1:UBE2D2 and RINGMUL1:UBE2D2~UB mimetics to TADp53, we first characterized the stable mimetics of UBE2D2~UB, such as UBE2D2–UBOE or UBE2D2–UB isopeptide (UBE2D2–UBIP) by HSQC experiments. Since UBE2G1C90S–UBOE synthesizes K48‐linked di‐UB even though its activity is lower than that of the thioester [20], UBE2D2C85S (UBE2D2S)–UBOE seems to be an active mimetic of UBE2D2~UB. Therefore, UBK48R, and not wild‐type UB, was used to synthesize stable mimetics via E1‐mediated conjugation. UBE2D2S22R (UBE2D2R) was used to disrupt the inter‐molecular interaction between the attached UBD and the backside of different UBE2D2 molecules [21]. To further increase the sensitivity of HSQC experiments, 2H/13C/15N‐labelled UBK48R (DCNUBR) instead of 15N‐labelled UBK48R (15NUBR) was used for the enzymatic conjugation.

The HSQC spectrum of UBE2D2S22R/C85S (UBE2D2RS)–DCNUBR OE was similar to that of free DCNUBR, except for the C‐terminal residues of UB (Figs 3A and 4A). UBE2D2RS–DCNUBR OE alone was stable in buffer solution (pH 6.5), and no apparent hydrolysis was detected after performing the HNCA experiment at 5 °C. However, the addition of RINGMUL1 extensively destabilized UBE2D2RS–DCNUBR OE, which prevented the acquisition of one clean HSQC spectrum at 25 °C. The half‐life of UBE2D2RS–DCNUBR OE in the presence of RINGMUL1 was estimated to be ~ 1.7 h, and it increased to ~ 10 h at 5 °C (Fig. 5). This half‐life was considerably less than those previously reported for the complexes of E4B minimal U‐Box and UbcH5cC85S–UB oxyester (E4BU:UBE2D3S–UBOE; 10 h at pH 5.75 and 20 °C) [15], RINGBIRC7:UBE2D2S–UBOE (hydrolysis after 1–3 days at pH 7.0 and 4 °C) [18], and UBE2G1C90S–DCNUBOE (8.8 h at pH 7.0 and 25 °C), in which the acidic loop of UBE2G1 mimics the effect of the RINGE3 domain [20].

Fig. 3

Monitoring the attached UBD of UBE2D2RS–DCNUBR OE and UBE2D2RAS–DCNUBR OE via 1H–15N HSQC experiments. All NMR experiments were performed at 5 °C, and the HSQC spectra were overlaid to that of free DCNUBR (black). The 1H–15N TROSY‐HSQC spectra of 0.1 mm UBE2D2RS–DCNUBR OE (A), RINGMUL1:UBE2D2RS–DCNUBR OE (1:1) (B), and RINGMUL1:UBE2D2RS–DCNUBR OE:AD39 (1:1:2) (C) were recorded with the reduced 15N‐dimension to decrease the acquisition time. (D) The HSQC spectrum of 0.1 mm UBE2D2RAS–DCNUBR OE shows that the N77A mutation of UBE2D2 considerably affects the conformation of UBD in the absence of RINGMUL1. The folded peaks are indicated with blue letters. When the peaks of free DCNUBR are only visible, these are indicated with the subscript F.

Fig. 4

Monitoring the UBD conformation in UBE2D2RS–DCNUBR OE and UBE2D2RAS–DCNUBR OE complexes by the HSQC experiments. The CSPs of UBE2D2RS–DCNUBR OE (A–C) and UBE2D2RAS–DCNUBR OE (D–G) in the absence and presence of interacting counterparts. The residues corresponding to the 1H–15N HSQC crosspeaks that were not accurately traced due to the absence of peaks and that disappeared during the CSP experiments are indicated with negative and green bars, respectively. The residues with the intensity ratios (I/I UB) higher than 1.5 are indicated with plus symbols. The residues for which CSP analysis was difficult due to the peak overlap are indicated with blue cross symbols. (H) Ribbon models of the open and closed forms of UBE2D2–UBOE, in which the closed model was obtained by combining two structures of our RINGMUL1:UBE2D2 and RINGBRIC7:UBE2D1RAS–UBOE (PDB, 4auq). The specific regions of UBD are indicated with different colours, and the T7 residue of UBD is indicated by spheres.

Fig. 5

The estimation of stabilities of (A) UBE2D2RS–DCNUBR OE and (B) UBE2D2RK–DCNUBR IP in the presence of RINGMUL1. The hydrolyses of UBE2D2RS–DCNUBR OE and UBE2D2RK–DCNUBR IP were assessed by SDS/PAGE analysis. A small amount of E1 fragments (E1frag and E1frag*) remained after the purification of UBE2D2~UB mimetics. The hydrolysis of UBE2D2RS–DCNUBR OE was greatly enhanced by the RINGMUL1‐binding, but that of UBE2D2RK–DCNUBR IP was not. The half‐lives of UBE2D2RS–DCNUBR OE were 1.67 ± 0.02 and 10.3 ± 1.1 h at 25 and 5 °C, respectively.

Since the isopeptide bond between K85 of UBE2D2 and G76 of UB is believed to be more stable than that of the oxyester, UBE2D2RK–DCNUBR IP was prepared following the previously reported method [22]. Although UBE2D2RK–DCNUBR IP was initially thought to be weakly hydrolysed in the presence of RINGMUL1, it was confirmed later to be stable using the SDS/PAGE analysis (Fig. 5). We also prepared UBE2D2RAS–DCNUBOE; the conserved N77 of UBE2D2 that is located near the active C85 residue, stabilizes the oxyanion intermediate during ubiquitylation [23]. The N77A mutation apparently increased the stability of UBE2D2RAS–DCNUBR OE, and any HSQC crosspeak of free DCNUBR was not detected during NMR experiments with RINGMUL1, RINGMUL1:AD39, and RINGMUL1:TADp53. Since UBE2D2RK–DCNUBR IP displayed a lower response to the binding of RINGMUL1 and AD39 compared to UBE2D2RS–DCNUBR OE likely due to the different geometry between the isopeptide and oxyester (Fig. 6A vs Fig. 3B,C), we focused on UBE2D2RS–DCNUBR OE and UBE2D2RAS–DCNUBOE as active and stable mimetics, respectively, for further studies to evaluate their detailed binding modes for RINGMUL1 and TADp53.

Fig. 6

The 1H–15N HSQC spectra of UBE2D2RK–DCNUBR IP (A) and UBE2D2RAS–DCNUBR OE (B). The folded peaks are indicated with blue letters. The HSQC spectra of 0.1 mm UBE2D2RK–DCNUBR IP and 0.1 mm UBE2D2RAS–DCNUBR OE were recorded at 5 °C, in the absence and presence of the interacting counterparts. The used concentrations of RINGMUL1, AD39, and TADp53 were 0.1, 0.2, and 0.2 mm, respectively. The HSQC spectrum of UBE2D2RK–DCNUBR IP was similar to that of UBE2D2RS–DCNUBR OE (Fig. 3A); however, changes in its spectra, induced by the binding of RINGMUL1 and RINGMUL1:AD39, were less than those of UBE2D2RS–DCNUBR OE (Fig. 3B,C), respectively.

UBE2D2RS–UBOE was different from UBE2D2RAS–UBOE in terms of closed conformation and interactions with RINGMUL1 and TADp53

We measured each clean HSQC spectrum of UBE2D2RS–DCNUBR OE in the presence of RINGMUL1 and RINGMUL1:AD39 at 5 °C. The HSQC spectra of UBE2D2RS–DCNUBR OE and free DCNUBR showed that the UBD exhibited less intramolecular interactions with UBE2D2RS and favoured an open conformation (Figs 3A and 4A), as previously shown in UBE2D3RS–15NUBOE [24]. RINGE3 activates E2~UB in a closed conformation and renders the nucleophilic attack by a Lys residue of a target protein [25]. The UBD of UBE2D2RS–DCNUBR OE likely assumed a closed conformation in the presence of RINGMUL1 (Figs 3B and 4B). The binding exchange kinetics of UBD seemed to shift from a fast (weak binding, CS‐moving) to an intermediate (stronger binding, disappearance of peak due to line‐broadening) NMR time scale. It has been well known that the residues 6–14 and 41–50 of UBD play an important role in switching to a closed conformation via an intramolecular interaction with E2 proteins [15, 24, 26]. Additionally, the RINGMUL1‐binding caused extreme peak‐broadening of the C‐terminal part of DCNUBD (residues 70–76; Figs 3B and 4B).

Unexpectedly, the HSQC spectrum of UBE2D2RAS–DCNUBR OE showed that its UBD already adapted to a more closed conformation even in the absence of RINGMUL1 (Fig. 3D). The scale of the CSPs (Fig. 4D) was considerably higher than those of UBE2D2RS–DCNUBOE (Fig. 4A) and RINGMUL1:UBE2D2RS–DCNUBR OE (Fig. 4B). The N77 residue of UBE2D2 can form a hydrogen bond with the G76 carbonyl group of the attached UBD, and it was clear that the presence of N77 inhibited the formation of a closed conformation of UBD. Since UBE2D2RAS–DCNUBR OE already adopted a closed conformation, the RINGMUL1‐binding only caused a marginal CSP of DCNUBR D (Figs 4E and 6B). Moreover, the RINGMUL1‐binding also resulted in the different CSP pattern of UBE2D2RAS–DCNUBR OE from that of UBE2D2RS–DCNUBR OE. The CSPs of UBE2D2RS–DCNUBR OE mainly occurred in residues 41–50 of UBD (Fig. 4B,H), whereas those of UBE2D2RAS–DCNUBR OE mainly occurred in residues 6–14, including T9 located near the bound RINGMUL1 (Fig. 4E,H). The model structures of RINGMUL1:UBE2D2RS–UBR OE was generated via energy minimization of the hybrid model from two PDB coordinates of our RINGMUL1:UBE2D2 and RINGBRIC7:UBE2D2RAS–UBOE (PDB, 4auq). The 2D HSQC spectra for UBE2D2RAS–DCNUBR OE and free DCNUBR differ in terms of both CSP and peak intensity (Figs 3D and 4D), but those for UBE2D2RS–DCNUBR OE and RINGMUL1:UBE2D2RS–DCNUBR OE differ mostly in peak intensity (Figs 3B and 4B).

The presence of AD39 resulted in the disappearance of the many HSQC crosspeaks of RINGMUL1:UBE2D2RS–DCNUBR OE (Figs 3C and 4C). The CSPs of RINGMUL1:UBE2D2RAS–DCNUBR OE induced by the AD39‐binding (Fig. 4F) were much less than those of RINGMUL1:UBE2D2RS–DCNUBR OE (Fig. 4C). However, the binding of TADp53 resulted in higher CSP levels of RINGMUL1:UBE2D2RAS–DCNUBR OE (Fig. 4G), and the HSQC spectrum (Fig. 6B) showed increased similarity with that of RINGMUL1:UBE2D2RS–DCNUBR OE:AD39 (Fig. 3C).

RINGMUL1:UBE2D2RAS–DCNUBOE exhibits an enhanced binding affinity for TADp53 compared to UBE2D2:RINGMUL1

The K d values between RINGMUL1:UBE2D2 and TADp53 (171 ± 38 μm) was estimated by the CSP experiments with 15NTADp53 as increasing the concentration of RINGMUL1:UBE2D2, in which the HSQC peaks of the AD49 region were traced (Table 1). Similar binding affinities of TADp53 and AD39 for RINGMUL1:UBE2D2 (111 ± 31 and 184 ± 13 μm, respectively) were also confirmed via ITC experiments (Table 1 and Fig. S2). Interestingly, the ΔH value of the binding between RINGMUL1:UBE2D2 and TADp53 (−673 kcal·mol−1) was higher than that between RINGMUL1:UBE2D2 and AD39 (−1134 kcal·mol−1). Thus, the higher binding affinity of TADp53 to RINGMUL1:UBE2D2 compared to that observed with AD39 was attributed to more entropic contribution (−TΔS) arising from the SN15 region.

Although UBE2D2RAS–UBR OE did not seem to be identical to UBE2D2RS–UBR OE, it was used to study the binding thermodynamics of RINGMUL1 and TADp53 due to its high stability in solution. The CSPs of 15NTADp53 in the presence of RINGMUL1:UBE2D2RAS–UBR OE showed that the interaction region of TADp53, which was centred on a specific part of AD39, including W53 and F54, was propagated over whole regions from AD39 to SN15 (Fig. 1C,D). Therefore, the AD39 region seemed to be primarily recognized by the RINGMUL1:UBE2D2RAS–UBR OE. The 2D HSQC crosspeaks of the AD39 region specifically disappeared as increasing the concentration of RINGMUL1:UBE2D2RAS–UBR OE, and thus the K d values of RINGMUL1:UBE2D2RAS–UBR OE for AD39 and for TADp53 were determined to be 129 ± 22 μm and 34 ± 16 μm, respectively, via ITC experiments (Table 1 and Fig. S2). The molecular basis of ΔS contribution to the binding ΔG was difficult to be inferred, but the 4‐times higher binding affinity of TADp53 for RINGMUL1:UBE2D2RAS–UBOE, compared to that of AD39, was attributed to the SN15 region (ΔH, −629 and 485 kcal·mol−1; ΔS, 15.7 and 20.6 cal·mol−1·deg−1 for AD39 and TADp53, respectively).

The AD39 region of TADp53 reportedly forms an induced α‐helical structure when binding to Bcl‐XL [27] and MDM2 [28]. We examined whether the secondary structure of AD39 was changed in the presence of RINGMUL1:UBE2D2RAS–UBR OE via circular dichroism (CD) spectroscopy. The K d value of 129 μm (Table 1) assumed that ~ 0.45 fraction of the AD39 bound the RINGMUL1:UBE2D2RAS–UBR OE. However, no increased formation of an α‐helical structure was identified (Fig. 7). It is likely that TADp53 binds RINGMUL1:UBE2D2~UB without the formation of any defined secondary structure.

Fig. 7

Estimation of the secondary structure of the bound AD39 peptide to UBE2D2RAS–UBR OE by CD experiments. The protein samples were prepared in the same NMR buffer (pH 6.5, 50 mm MES, 50 mm NaCl, and 5 µm ZnSO4) and the CD experiments were performed at room temperature. The CD spectra of each separate 0.1 mm AD39 and RINGMUL1:UBE2D2RAS–UBR OE were recorded and then the spectra were arithmetically added (red). The CD spectrum of the mixed 0.1 mm RINGMUL1:UBE2D2RAS–UBR OE:AD39 was recorded for the comparison (blue). Any increased α‐helical structure of AD39 is not shown in the complex form with RINGMUL1:UBE2D2RAS–UBR OE, since no decrease of CD ellipticity at 222 nm was observed for the mixed sample.

The increased binding affinity of RINGMUL1:UBE2D2–UBOE for TADp53 depends on multivalent interactions

Tracing of TADp53‐binding surfaces via CSP experiments using 15NUBE2D2 and 15NRINGMUL1 even in the context of stable RINGMUL1:UBE2D2RAS–UBR OE was impossible, since most HSQC peaks of 15NUBE2D2RAS and 15NRINGMUL1 disappeared in the complexes due to a severe exchange peak‐broadening (not shown). Even deuterated RINGMUL1 (DCNRINGMUL1) did not result in an analysable HSQC spectrum in the presence of UBE2D2RAS–UBR OE (not shown). Therefore, we attempted to map the binding sites for AD39 sequentially using (a) RINGMUL1:15NUBE2D2 (Fig. 8A,D, red CSP colour), (b) 15NRINGMUL1:UBE2D2 (Fig. 8B,D, blue CSP colour), and (c) RINGMUL1:UBE2D2RS–DCNUBR OE (Figs 4C and 8E ), respectively. The hybrid model structure of a closed conformation (Fig. 4H) was used to show the surfaces of the attached UBD in RINGMUL1:UBE2D2–DCNUBR OE that bound to AD39.

Fig. 8

The binding surfaces of RINGMUL1:UBE2D2 and RINGMUL1:UBE2D2RS–UBR OE for AD39. The CSPs of 0.2 mm RINGMUL1:15NUBE2D2 (A), 15NRINGMUL1:UBE2D2 (B), and 15NRINGMUL1 (C) in the presence of 0.5 mm AD39 are shown as bar plots, respectively. The amount of CSPs induced by the AD39‐binding is higher in 15NRINGMUL1:UBE2D2 than that in 15NRINGMUL1 alone. (D) The CSPs of RINGMUL1:15NUBE2D2 (panel‐A) and 15NRINGMUL1:UBE2D2 (panel‐B) induced by the AD39‐binding are indicated on the surface model via red and blue tones, respectively. (E) The CSPs of 0.1 mm RINGMUL1:UBE2D2RS–DCNUBR OE by 0.2 mm AD39, as shown in Fig. 4C, are indicated on the surface model of RINGMUL1:UBE2D2–UBOE. The AD39‐binding also caused the CSPs in the other surfaces of UBD that are distant from the UBE2D2‐binding interface; (a) the cyan circled area, residues 13–14 and 28–34, (b) the yellow circled area, residues 58–61.

UBE2D2 exhibited the presence of two regions that interacted with AD39, wherein the main region was located in the vicinity of the RING‐binding site of UBE2D2 (α1 and N‐terminal α3; Fig. 1F,J), which was also identified in the CSPs of RINGMUL1:15NUBE2D2 (Fig. 8A,D). The AD39‐binding to the main region of UBE2D2 could additionally stabilize the closed conformation of the attached UBD by shifting RINGMUL1 close to the UBD region of UBE2D2~UB. The CSPs of 15NRINGMUL1:UBE2D2 by AD39 (Fig. 8B) showed that β1, the unstructured N‐ and C‐terminal regions of 15NRINGMUL1 were also perturbed (Fig. 8D, blue CSP colour). The higher amount of CSPs for the same concentration of AD39, compared to 15NRINGMUL1 alone (Fig. 8B,C), indicated that the presence of UBE2D2 increased the binding affinity for AD39. Although the CSPs of RINGMUL1:UBE2D2RS–DCNUBR OE induced upon the AD39‐binding were strongly coupled to the exchange motions of UBD, which also caused the significant CSPs located in the UBE2D2‐contacting region of DCNUBR D, the mapping of the CSPs of RINGMUL1:UBE2D2RS–DCNUBR OE on the surface model clearly indicated that the interface between RINGMUL1 and UBD of RINGMUL1:UBE2D2RS–DCNUBR OE complex also participated in its interaction with AD39 (Fig. 8E, marked with a cyan circle). The AD39‐binding also caused the CSPs in another surface of UBD that is distant from the UBE2D2‐binding interface (Fig. 8E, marked with yellow circle). Overall, the interaction of AD39 with RINGMUL1:UBE2D2RS–DCNUBR OE displayed a multivalency of cumulative weak bindings, which is a typical characteristic of IDP interaction [29].

TADp53 resulted in stronger CSPs of RINGMUL1:UBE2D2RAS–DCNUBR OE compared to AD39 (Fig. 4F,G). The occurrence of synergistic interactions by TADp53, from AD39 to SN15, enhanced its binding affinity for RINGMUL1:UBE2D2RAS–DCNUBR OE. The lengths of the extended AD39 region and the long horizontal axis of UBE2D2 were ~ 65 and ~ 50 Ǻ (Fig. 1J), respectively, and the binding stoichiometry (N) of RINGMUL1:UBE2D2RAS–DCNUBR OE for TADp53 determined by the ITC measurements was close to 1 (0.67 ± 0.19; Table 1). Therefore, TADp53 seemed to encompass whole interaction regions involving RINGMUL1, UBE2D2, and UBD.

Discussion

Differential characteristics of various UBE2D2~UB mimetics upon RINGMUL1 binding

Closed conformation of E2~UB reportedly plays an important role in the RING‐E3‐mediated ubiquitylation [15, 17, 30]. The UBD molecules of RING:UBE2D1RAS–UBOE (PDB, 4auq) and various other RING:E2–UBIP complexes (PDB, 4ap4, 4v3k, 5fer, 5mnj, 5vgw, and 6hpr) have the common converged position via the linchpin interactions (Fig. 9A). However, the closed conformation of UBD in the recent crystal structure of RINGRNF12:UBE2D2RK–UBIP [31] is different from those of the linchpin structures (Fig. 9B). Middleton et al. reported that direct contact between the UBD and RINGRNF12 was absent, indicating that although UBD in RINGRNF12:E2~UB might possess a range of conformational spaces, locking of UBD in the prime conformation via additional intramolecular RING‐to‐UBD linchpin contact could be critical for ubiquitylation [31]. Interestingly, the CSP pattern of UBE2D2RAS–DCNUBR OE induced by RINGMUL1 (Fig. 4E) is well matched with the linchpin structure (Fig. 9C), in which the T9 of UBD has a close contact with the RINGE3 domains. However, that of UBE2D2RS–DCNUBR OE by RINGMUL1 (Fig. 4B) is correlated with the structure of RINGRNF12:UBE2D2RK–UBIP (Fig. 9C). Although the molecular basis for the activation of UBE2D2~UB mediated by the RINGMUL1‐binding remains unknown, a special UBE2D2‐binding mechanism seems to be mediated by RINGMUL1. It may be postulated that subtle structural changes result in the occurrence of allosterically linked inter‐ and intra‐molecular interactions of RINGMUL1:UBE2D2–UBD, in which one instance among the positional plasticity of the bound RINGMUL1 favoured the transient structure of UBE2D2, thereby stabilizing a closed conformation conducive for the hydrolysis of the attached UBD.

Fig. 9

Crystal structures of various RINGE3:UBE2D2~UB mimetics and the inspection of the N77A mutation effect on a closed conformation. The presence of hydrogen‐bonds is indicated with green lines. (A) Representative crystal structures of RINGBRIC7:UBE2D1RAS–UBOE and RINGTRIM23:UBE2D2RK–UBIP (PDB, 4auq and 5vzw) are superimposed, and the positions of UBD are almost identical. (B) The recent crystal structure of RINGRNF12:UBE2D2RK–UBIP (PDB, 6w9d) shows that its closed conformation does not have the linchpin structure. (C) The CSPs of UBE2D2RS–DCNUBR OE (Fig. 4B) and UBE2D2RAS–DCNUBR OE (Fig. 4E) induced by the RINGMUL1‐binding are decorated on the hybrid model of RINGMUL1:UBE2D2–UBOE using red colour. The CSPs of UBD in the side facing the RINGMUL1 are more apparent by the N77A mutation. (D) Gibbs free energy diagram was intuitively illustrated for UBE2D2~UB and UBE2D2N77A~UB. Since UBE2D2RAS–DCNUBR OE already forms a more stable closed conformation, the ΔG system (UBD close vs UBD open) of UBE2D2N77A~UB is markedly lower than that of UBE2D2~UB. When a closed conformation of UBD is induced by the RINGMUL1‐binding (ΔG UB contact), the presence of N77 can increase the free energy at the junction of UBD and UBE2D2 (ΔG UB junction). Therefore, the overall ΔG system of RINGMUL1:UBE2D2~UB is higher than that of RINGMUL1:UBE2D2N77A~UB.

The activation of UBE2D2~UB via a closed conformation of UBD induced by the RINGMUL1‐binding depends critically on the presence of N77. Although detailed characterization of a closed conformation of RINGMUL1:UBE2D2RS–UBR OE remains further studies, our NMR data showed that the intra‐ and inter‐molecular interactions of UBE2D2RS–UBR OE were different from those of UBE2D2RAS–UBR OE in terms of the RINGMUL1‐ and TADp53‐bindings. To explain the different closed conformations of UBE2D2RS–DCNUBR OE and UBE2D2RAS–DCNUBR OE conceptually, the Gibbs free energy of the RINGMUL1‐induced closed conformation of UBD (ΔG system) can be divided into the overall contact energy between UBE2D2 and UBD (ΔG UB contacts) and the local energy at the junction of thioester (ΔG UB junction). The high energy state of ΔG UB junction that activates UBE2D2~UB and is critically dependent on the presence of N77 could be supported by decreasing ΔG UB contacts relying on the RINGMUL1 binding (Fig. 9D).

In vivo interaction of p53 and RINGMUL1:E2~UB

The multivalency of TADp53 originating from its innate IDP characteristics contributes to its enhanced binding affinity for RINGMUL1:UBE2D2~UB, which is supported by the estimated binding thermodynamic parameters (Table 1). The formation of the triple complex (RINGMUL1:UBE2D2~UB) provides additional surfaces for adopting TADp53. TADp53 caused the appreciable CSPs in the various regions of RINGMUL1:UBE2D2 and RINGMUL1:UBE2D2RAS–UBOE (Fig. 8D,E), which is a typical characteristic of the IDP interaction, multivalency [29], and thus the TADp53‐binding for RINGMUL1:UBE2D2RAS–UBR OE is hardly converged to one binding mode. It is likely that the IDP nature of TADp53 and its multivalency occurring during binding to RINGMUL1:UBE2D2~UB may synergistically increase the binding affinity and affect the dynamic nature of UBD. The binding stoichiometry (N) between RINGMUL1:UBE2D2RAS–UBOE and TADp53 is close to 1, and the AD39 region of TADp53 is primarily recognized as increasing the size of complexes from RINGMUL1 or UBE2D2 alone to RINGMUL1:UBE2D2RAS–UBOE. Therefore, the AD39 region is preferentially located in two interfaces (a) between the α1 of UBE2D2 and RINGMUL1, and (b) between UBD and RINGMUL1, and then the SN15 region including K24 faces the accessible junction between UBD and UBE2D2, which can render the thioester bond susceptible to the nucleophilic attack by K24.

The measured K d values between RINGMUL1:UBE2D2RAS–UBR OE and TADp53 (34 μm) may not be sufficient for achieving in vivo ubiquitylation of p53 by MUL1. However, native RINGMUL1:UBE2D2~UB could exhibit a higher affinity for TADp53, since RINGMUL1:UBE2D2RS–UBR OE displayed a stronger dynamic nature of UBD in response to the AD39‐binding than that observed with RINGMUL1:UBE2D2RAS–UBR OE. It is also well documented that the transcription‐independent pro‐apoptotic function of p53 is associated with stress‐induced translocation of p53 to the mitochondria [32, 33, 34]. In vitro physical interaction between the C‐terminal domain of p53 and negatively charged phospholipids has also been reported [35]. Therefore, the reduction in the dimensionality of the interaction from 3D to 2D in the mitochondrial outer membrane may enhance the interaction between MUL1 and p53, thereby facilitating rapid ubiquitylation of the translocated p53 for subsequent processes.

The complex formed by RINGMUL1 with UBE2D2~UB evidently aid the recruitment of TADp53 and this ubiquitylation mechanism has not been reported thus far. The in vivo regulation and proteostasis of IDPs that are dominantly observed in humans are important for understanding the mechanisms underlying the development of human diseases [29, 36]. The ubiquitylation of IDPs could be one of the most commonly observed in vivo regulation mechanisms, in which direct recognition of IDPs by other RINGE3:E2~UB complexes may also be useful.

Materials and methods

Protein expression and purification

The human RINGMUL1 domain (residues 298–352), human TADp53 (residues 1–73), and UBE2D2 conjugating enzymes were prepared following the previously reported methods [11, 14]. Non‐tagged UBK48R proteins were prepared as per the previously reported method [37]. For the preparation of NMR samples, size exclusion chromatography (SEC) was performed using the NMR‐buffer (pH 6.5, 50 mm MES, 50 mm NaCl, 5 µm ZnSO4, and 1 mm DTT). In the other cases, SEC was performed with a buffer (pH 7.5, 25 mm Tris‐HCl, 100 mm NaCl, and 1 mm DTT). The concentrations of all proteins were estimated using their extinction coefficient at 280 nm [38].

For the NMR experiments, the proteins were expressed in M9 minimal media after incubation for 6 h at 30 °C following isopropyl β‐d‐1‐thiogalactopyranoside (IPTG) induction at 0.7–0.8 OD at 600 nm. Isotope‐labelled proteins were expressed by growing Escherichia coli cells in M9 minimal media (1 L) supplemented with isotope‐labelled ammonium chloride (1 g) and glucose (2 g). CELTONE base powder (1 g), vitamins, and trace metals were added to enhance cell growth. All isotope‐labelled materials were purchased from Cambridge Isotope Laboratories Inc. The detailed composition of media and the culture method used for protein deuteration are described in previous reports [39].

The mouse E1 gene was cloned into the pRSET‐A vector, and then the plasmid was transformed into E. coli Rosetta (DE3). The E1 protein was expressed in LB medium for 6 h at 25 °C after IPTG induction at 0.7–0.8 OD at 600 nm. The cultured cells were resuspended in buffer (pH 8.0, 25 mm Tris‐HCl, 0.5 m NaCl, 10 mm 2‐mercaptoethanol, and 1 mm PMSF). Nonspecific protease activity was reduced by adding EDTA‐free protease inhibitor cocktail tablets (Sigma‐Aldrich, Seoul, Korea). After cell disruption by sonication, the supernatant was applied to a Histrap HP column (GE Healthcare, Seoul, Korea) as soon as possible. The E1 protein was eluted using 150 ml imidazole gradient with buffer (pH 8.0, 25 mm Tris‐HCl, 0.5 m NaCl, 0.5 m imidazole, and 10 mm 2‐mercaptoethanol). The protein fractions were dialysed into buffer (pH 8.0, 20 mm Tris‐HCl and 1 mm DTT), and then was applied to a Hitrap‐Q HP column (GE Healthcare). Elution was performed using a 150 ml NaCl gradient with buffer (pH 8.0, 20 mm Tris‐HCl, 1 m NaCl, and 1 mm DTT). The purified E1 protein was concentrated to ~ 2.0 mg·mL−1, roughly estimated via SDS/PAGE, and then stored in a −80 °C refrigerator.

Preparation of the UBE2D2~UB mimetics and estimation of their stabilities in solution

All mimetics, namely, UBE2D2–UBOE [20] and UBE2D2–UBIP [22], were enzymatically synthesized as previously reported methods. The mimetics were synthesized via E1‐mediated enzymatic reactions of E2 proteins (0.1 mm) in the presence of twice the amount of non‐tagged DCNUBR protein. Briefly, the reactions of UBE2D2RS and UBE2D2RAS were performed in buffer (pH 9.0, 50 mm CHES, 150 mm NaCl, 1 mm DTT, 4 mm ATP, and 5 mm MgCl2) with ~ 4 μm E1 for 6 h at 35 °C. UBE2D2–UBIP was synthesized with ~ 6 μm E1 in buffer (pH 10.0, 50 mm CAPS, 150 mm NaCl, 1 mm DTT, 4 mm ATP, and 5 mm MgCl2) for 24 h at 35 °C. The synthesized UBE2D2~UB mimetics were purified by SEC using the Superdex‐75 column in buffer (pH 6.5, 50 mm MES, and 50 mm NaCl).

Each purified 0.1 mm UBE2D2RS–UBR OE and UBE2D2RK–UBR IP was incubated with 0.15 mm RINGMUL1 in the NMR‐buffer. The reaction mixtures were aliquoted and then were incubated at 25 and 5 °C, respectively. The reaction was stopped at an appropriate time by adding the SDS sample buffer, and then was stored at −20 °C before the SDS/PAGE analysis. The protein bands of UBE2D2~UB mimetics were quantified with the ImageJ program (https://imagej.nih.gov/ij/). The half‐lives of UBE2D2RS–DCNUBR OE in the presence of RINGMUL1 were estimated by fitting to the equation of single exponential decay.

Preparation of peptide samples

SN15 (residues 15–30) and AD39 (residues 39–57) peptides, as well as the colour‐dye derivative form of AD39, were purchased from PEPTRON Inc. (Daejeon, Korea). In the dye attached AD39 peptide, the Lys residue coupled to 2,4‐dinitrophenyl dye (K‐DNP) was attached to the C‐terminal of the native form (AMDDLMLSPDDIEQWFTED/K‐DNP). All peptides were dissolved in an appropriate experimental buffer, and then an equimolar amount of NaOH as that of the Asp/Glu residues was added to maintain the solution pH value.

NMR experiments

To assign the backbone CSs of UBE2D2, the HNCACB, HN(CO)CACB, HNCO, and HN(CA)CO spectra were recorded using the Bruker 800 and 900 MHz spectrometers equipped with TCI‐cryogenic probe. The HSQC crosspeaks of the mutated proteins (UBE2D2RS, UBE2D2RAS, UBE2D2RK, and UBR) and their UB‐conjugated forms were assigned by additional HNCA experiments. In the absence of additional comments, NMR experiments were conducted in the NMR buffer (pH 6.5, 50 mm MES, 50 mm NaCl 5 μm ZnSO4, and 5% D2O) at 25 °C. The CSP data of HSQC peaks were processed using the equation, [(6×ΔH)2 + ΔN2]1/2, where ∆H and ∆N represent the CS differences of 1H and 15N, respectively. Analysis of the protein amide CSs deposited in the Biological Magnetic Resonance Bank (No. 36251) showed that the average distribution of 15N‐CSs was 6‐fold greater than that of 1H‐CSs [40]. All NMR data were processed using the NMRPipe program [41] while spectral analyses were conducted using the NMRFAM‐SPARKY program [42].

Determination of crystal structures of the RINGMUL1 and UBE2D2:RINGMUL1 complex

Detailed crystallization conditions of the RINGMUL1 protein (15 mg·mL−1) and the UBE2D2:RINGMUL1 complex (10 mg·mL−1) have already been reported [11]. The crystal of UBE2D2 alone was obtained under the conditions of the reservoir buffer containing 0.1 m sodium citrate (pH 5.6), 0.5 m ammonium sulfate, and 1.0 m lithium sulfate during the screening of the UBE2D2:RINGMUL1 crystal. The diffraction data were collected at beamline 7A at the Pohang Accelerator Laboratory, and data were indexed, scaled, and merged using the HKL‐2000 software [43].

To solve the structure of RINGMUL1, we performed molecular replacement (MR) using the Phaser program [44]. Neither the recent NMR ensemble structures of RINGMUL1 (PDB code, 6K2K) nor the homology model calculated using the Phyre2 (Protein Homology/Analogous Recognition Engine) web portal [45] yielded the correct MR results. Therefore, the crystal structure of the UBE2D2:RINGMUL1 complex was solved by MR using the coordinate of UBE2D2 (PDB code, 2ESK), and then the crystal structure of RINGMUL1 alone was solved by MR using the coordinate of RINGMUL1 in the complex. Interactive model building and structure refinement were performed using the Coot program [46] and the phenix software suite [47]. The calculated crystal structures were validated using the MolProbity web portal [48].

Biophysical analyses

All protein samples, except the peptide samples, were prepared using a buffer (pH 6.5, 50 mm MES, 50 mm NaCl, 5 µm ZnSO4, and 1 mm TCEP) through dialysis or SEC. ITC experiments were performed using the Auto‐iTC200 micro‐calorimeter (Malvern Panalytical, Malvern, UK) at 10 °C. The low concentrated protein was loaded in the sample cell, and 15–30 higher concentrated titrant protein was placed in the syringe. ITC data were analysed using the microcal origin™ software.

Circular dichroism spectra were recorded using the J‐715 CD instrument (JASCO, Tokyo, Japan) with 0.05 cm path‐length circular CD cell to measure high‐concentration protein samples (0.1 mm, > 3.0 mg·mL−1). All protein and peptide samples were prepared using a buffer (pH 6.5, 50 mm MES, 50 mm NaCl, and 5 µm ZnSO4).

Structure presentations and analyses

Visualizations of all structures were performed using the Chimera program [49]. The CSP data were presented via a colour gradient using the B‐factor column of the PDB files. The RMSD values of the structures were calculated using the AmberTools21 program package (https://ambermd.org/AmberTools.php). The evaluation of the RINGMUL1 crystal structures based on the 1H–15N RDC values was done with the calcETensor analysis module of the Xplor‐NIH program [50].

Accession numbers

The PDB coordinates of UBE2D2 (PDB ID: 7BOL), RINGMUL1 (PDB ID: 6M2D), and RINGMUL1:UBE2D2 (PDB ID: 6M2C) were deposited to protein data bank (PDB).

Conflict of interest

The authors declare no conflict of interest.

Author contributions

SWC, CKL, and KSR planned this research project. MSL and SOL performed the NMR and X‐ray experiments, respectively. JC performed most of the additional experiments. MR, MKL, JHK, and EH were involved in the design and data interpretation of additional experiments including ITC. Overall data analyses were performed by JC and KSR. The manuscript was written by JC, SWC and KSR.

Peer review

The peer review history for this article is available at https://publons.com/publon/10.1111/febs.16360.

Fig. S1. The interactions between RINGMUL1 and UBE2D2 were monitored by 1H–15N HSQC experiments.

Fig. S2. The representative ITC data of the RINGMUL1‐mediated interactions.

Table S1. Statistics of data collections and structure refinements.

Click here for additional data file.