Allosteric Activation of E2-RING Finger Mediated Ubiquitylation by a Structurally-Defined Specific E2 Binding Region of gp78

The G2BR is a High Affinity Binding Site for Ube2g2

To evaluate the interaction between the gp78 G2BR (Figure 1A) and Ube2g2, isothermal titration calorimetry (ITC) was employed (Supplemental Figure 1). This confirmed the direct, high affinity interaction of Ube2g2 with a 27 amino acid G2BR peptide. The 1:1 complex of E2 and G2BR has a dissociation constant (K_d) of 21(± 4) nM (Table 1). The reaction is exothermic and the significant free energy change (ΔG = -43.72 kJ mol^-1) implies a very stable complex. The decrease in entropy reveals that part of the complex is losing some degree of freedom and likely folding into a regular structure.

‘Backside Binding’ to Ube2g2 Induces G2BR Folding

To identify the binding surface between Ube2g2 and G2BR, the interaction was examined by monitoring the NMR spectra of each molecule independently. First, the Ube2g2 preparation was validated by confirming the resonance assignments for Ube2g2 compared to a previous study (Briggman et al., 2005). A secondary structure analysis based on chemical shifts (Wishart and Sykes, 1994) combined with distance information from a ¹⁵N-edited NOESY-HSQC spectrum confirmed that our Ube2g2 preparation matched the reported crystal structure (Arai et al., 2006). Titration of G2BR into isotopically-labeled Ube2g2 (Figure 1B) yielded slow exchange NMR spectra and indicated a K_d «1μM, consistent with the ITC data. An unexpectedly large number of the Ube2g2 H^N-N peaks shifted in the presence of G2BR, requiring reassignment. The ¹³C-edited HSQC spectra of the methyl resonances of Ile, Leu, and Val residues (data not shown) indicated a localized binding site, which correlates with the largest shifts observed in Figure 1B. The secondary structure based on chemical shifts of Ube2g2:G2BR was identical to free Ube2g2 and NOESY spectra of Ube2g2:G2BR exhibited equivalent patterns of connectivities between H^N- H^N, H^N-methyl, and methyl-methyl protons compared to free Ube2g2, implying that the overall structure of Ube2g2 did not change significantly.

The binding surface was determined by chemical shift mapping on Ube2g2 and by intermolecular NOESY experiments. Chemical shift mapping for Ube2g2 indicates the primary binding site is on beta-strands β1-β3 (Supplemental Figure 2). In addition, the C-terminal regions of helices α1 and α4 are perturbed by G2BR binding. Intermolecular NOESY crosspeaks were observed between G2BR and the methyl and amide protons of Ube2g2 (Supplemental Figure 3) that correspond to the surface indicated in blue on unbound Ube2g2 in Figure 1C. These NOEs refined the contact surface to residues V25, A26, E31, E38, L40-M42, E45, E50, F51-V53, V159, L163 and L165. This surface includes the hydrophobic patch formed by V25, A26, L40-M42 and F51-V53 that is mostly conserved in other E2s (Hamilton et al., 2001; Moraes et al., 2001; Brzovic et al., 2006). The G2BR binding site also includes negatively charged residues flanking the hydrophobic core. From the perspective of the G2BR, ¹⁵N-HSQC spectra (Figure 1D) of a biosynthetically-prepared G2BR exhibited poor chemical shift dispersion, consistent with a random-coil conformation. The backbone chemical shifts were assigned in the free G2BR peptide and, except for indications of a nascent four-residue helical turn at V583-K586, the peptide lacks any regular secondary structure. The spectrum of G2BR in the presence of Ube2g2 exhibits a dramatic increase in spectral dispersion and resonance linewidths consistent with a 1:1 complex of Ube2g2:G2BR (Figure 1D). This suggests that the G2BR must fold into a compact regular secondary structure utilizing all 27 residues to fit on the defined E2 binding site. Based on this, we redefine the G2BR as amino acids 574-600 of gp78 instead of 579-600 as originally defined (Chen et al., 2006).

The Ube2g2:G2BR Crystal Structure Reveals Distant Changes that Impact E2 Loading of Ubiquitin

The Ube2g2:G2BR complex was crystallized, forming rod-shaped crystals (0.1 mm × 0.3 mm × 0.1 mm). These diffracted to 1.8 Å resolution (Table 2). The structure shows that the Ube2g2 backbone is largely unchanged in the presence of G2BR (Figure 2A), as predicted by NMR measurements. The conserved secondary structural elements include α1 = T4-L18, β1 = G23-P28, β2 = E36-M42, β3 = V53-S59, β4 = K70-F73, 3¹⁰ helix = S91-L93, α2 = V116-A128, α3 = V138–D146, and α4 = R148-L163. Strands β1-β4 form an anti-parallel β-sheet (Figure 2A and 2B). A dynamic region in the β4α2 loop is present between residues H94 and W110. This corresponds to an acidic extension (aa 96-108) that, in mammals, is limited to Ube2g2 and two other E2s [Ube2g1 and Ube2r1 (Cdc34 in yeast)]. The backbone root-mean-square deviation (rmsd) between the free and bound Ube2g2 is 1.8 Å over all residues and decreases to 0.9 Å if residues 96-108 in the β4α2 loop are excluded.

Key structural changes were observed in Ube2g2 near the active site upon binding the G2BR. Access to the active site Cys (C89) is via a channel flanked on either side by β4α2 loop and the α2α3 loop (Figure 2A). The β4α2 loop was observed to exhibit dynamic behavior for aa 96-108 in the crystal structure of free Ube2g2 (Arai et al., 2006). In fact, there were three molecules in the asymmetric unit, and the major difference between the molecules was the conformation of the β4α2 and α2α3 loops (Figure 3A). These conformations are ordered; however, they exhibit higher B-factors than the rest of the structure (Supplemental Figure 2). The dynamic region of the β4α2 loop contains a short 3₁₀ helix and generally extends away from the protein in the three conformations in the free form; however, in the Ube2g2:G2BR complex presented here, the 3₁₀ helix is gone and a region of the loop (P100-Y103) orients back towards C89. The multiple conformations of these loops, which appear to shift to a new average state in the Ube2g2:G2BR complex, and the high B-factors in the crystal structures are consistent with dynamic averaging as suggested by chemical shift effects observed in solution (Supplemental Figure 2). The present structure may represent a low energy population of the average, in which the proximity of Y103 and C89 changes from 20.8 Å in unbound Ube2g2 to 3.8 Å in the complex. In addition, the α2α3 loop (comprising N131-G135) approaches C89 in the complex. This inward conformation is stabilized in the 1.8 Å crystal structure by a new network of water-mediated hydrogen bonding, including Y103, C89, and Y83 as well as a network including E133 and S91. The structural modifications of these two loops decrease accessibility around the active site (compare Figure 1C to 4B and Figure 3B to 3C). These changes are accompanied by an altered orientation of the C89 sidechain from pointing towards the β4α2 loop in the unbound form to pointing away from the loop (~80% probability) with bound G2BR (Figure 2A and compare Figure 3B to 3C).

The multiple states surrounding the active site seen in the crystal structures and the shift in conformational averaging in the Ube2g2:G2BR complex in solution suggest that these changes may impact interactions with E1 or ubiquitin bound to C89. To further evaluate this, we compared the structures of Ube2g2 and Ube2g2:G2BR with two reported E2-Ub structures (Hamilton et al., 2001; Eddins et al., 2006). Superposition of free Ube2g2 with the E2 component of Ubc1-Ub, which forms K48 ubiquitin chains (Chen and Pickart, 1990) and Ubc13-MMs2-Ub, which forms K63 chains (Deng et al., 2000), (Figure 3B) reveals that the C-termini of either of the two bound ubiquitins can readily be accommodated in the channel surrounding C89. However, superposition of Ube2g2:G2BR with the same two structures (Figure 3C) indicates that the area around the active site becomes occluded due to rearrangements of residues M101-Y103 and E133-G135 (within the β4α2 loop and α2α3 loop, respectively), resulting in steric clashes with the C-termini of the two ubiquitins. In Figure 3C, Y103 is buried and is not visible. These structure-based models predict direct consequences for loading of G2BR-bound Ube2g2 with ubiquitin from E1. This was tested by comparing ubiquitin loading of Ube2g2 in the presence of either a control scrambled peptide (Scr) (Figure 1A) or the G2BR. The G2BR resulted in a marked decrease in the rate of E2 loading at 30°C (Figure 3D) and almost no detectable loading at 12°C (Supplemental Figure 4). Deletion of the extended acidic portion of the β4α2 loop (aa 96-108; Ube2g2Δ96-108), which is the major contributor to steric crowding around the active site, decreased the G2BR effect by almost 75%, while G2BR binding was retained (Figure 3D; Supplemental Figure 5).

Ube2g2-Bound G2BR Forms an α-Helix with Extensive Contacts Across the Backside of the E2

The crystal structure of Ube2g2:G2BR demonstrates that the G2BR assumes a completely α-helical fold in the 1:1 complex and binds to a surface comprising the beta sheet (β1-β3) and the C-termini of α1 and α4 (Figure 2A), which is consistent with the solution NMR data (Figure 1C). The buried surface at the interface of Ube2g2 and G2BR is ~1950 Ų. The entire peptide is well ordered, and there is an extensive network of contacts between the G2BR and Ube2g2, comprising both hydrophobic and ionic or hydrogen bonding contacts (Figure 2C, 2D and 2E and Supplemental Table 1). The contact network is consistent with the thermodynamic data for formation of this complex and suggests that the recognition is highly specific. In the G2BR α-helix, there is evidence for 24 hydrogen bonds between all i and i+4 pairs beginning with the Ser 574 carbonyl oxygen to Arg 578 NH and running through Arg 596 carbonyl oxygen to Lys 600 NH. As amino acids 574-600 extends completely across the backside of Ube2g2, this suggests that extensions beyond this region do not interact with Ube2g2. This is in agreement with previous data (Chen et al., 2006) and NMR studies of aa 574-643 of gp78 (our unpublished results).

The presence of extensive contacts between Ube2g2 and G2BR suggest that single point mutations will not disrupt binding or activity. The initial characterization of the G2BR (Chen et al., 2006) indicated that the N- and C-terminal ends of the G2BR were each critical both for binding Ube2g2 and for the cellular function of gp78. To examine this observation, binding of G2BRΔN and G2BRΔC (Figure 1A) to Ube2g2 was assessed. G2BRΔC binds Ube2g2 with relatively weak affinity [K_d = 192 (± 22) μM] (Table 1). The affinity between G2BRΔN and Ube2g2 is significantly stronger [K_d = 740 (±110) nM], but nonetheless exhibits a ~35-fold decrease in affinity compared to wild-type G2BR. This confirms that the N and C terminal regions of G2BR both contribute directly to the high affinity binding observed with the intact domain. Binding of either G2BRΔN or G2BRΔC to Ube2g2 causes a subset of the observed chemical shift changes seen with the G2BR, consistent with their predicted contacts (data not shown).

On the basis of the Ube2g2:G2BR structure, we designed two peptides with four mutations each to assess the most important interactions, G2BRM4-1 and G2BRM4-2 (Figure 1A). These peptides were ¹⁵N-labeled and examined by NMR (Supplemental Figure 6). The ¹⁵N-HSQC spectra of both peptides indicated random-coil conformations characteristic of wild-type G2BR. Titration of G2BRM4-1 and G2BRM4-2 with unlabeled Ube2g2 yielded K_ds of 55 (±18) μM and 9.5 (±4) μM respectively (Table 1). These findings, combined with the data for G2BRΔN and G2BRΔC, underscore the highly distributed nature of the contacts in Ube2g2:G2BR binding.

The G2BR Enhances the Affinity of Ube2g2 for the gp78 RING Finger

The positioning of the Ube2g2:G2BR interface on the ‘backside’ of Ube2g2, suggests that this interaction should not present a direct steric impediment to the interaction of Ube2g2 with the gp78 RING finger binding site (Zheng et al., 2000; Brzovic et al., 2003; Dominguez et al., 2004) (Figure 2A). To evaluate possible allosteric effects of G2BR on RING finger binding, the interaction of isotopically-labeled Ube2g2 with gp78 RING finger was monitored by NMR (Figure 4). The gp78 RING finger binding interface on Ube2g2 consists of the N-terminal end of α1, the β3β4 loop (F62-P69) and part of the β4α2 loop (W110-S115; see Figure 2A and 2B for secondary structure elements). The solution structure of the gp78 RING finger is similar to other RING finger structures (Das & Byrd, unpublished data) and a reverse labeling experiment, using isotopically-labeled gp78 RING finger, confirmed that the gp78 RING finger side of the binding surface is similar to other RING finger:E2 interactions. The interaction exhibited fast exchange (Figure 4A) and a K_d of 144 (±10) μM (Table 1, Supplemental Figure 7). This is consistent with the low affinity of many RING finger:E2 interactions (Lorick et al., 2006; Christensen et al., 2007). The affinity of the gp78 RING finger for the Ube2g2:G2BR complex was also determined. While the binding interface was unchanged and remained in fast exchange, strikingly, the K_d decreased by ~48-fold to 3 (±1) μM (Table 1, Supplemental Figure 7). Separate binding experiments confirmed that the G2BR and RING finger do not directly interact in the gp78 RING finger:Ube2g2:G2BR complex (data not shown). The allosteric relationship of Ube2g2:gp78 RING finger affinity to Ube2g2:G2BR association was tested by measuring the affinity of gp78 RING finger for Ube2g2 in the presence of G2BRΔN, which exhibits an affinity for Ube2g2 intermediate between the full G2BR and G2BRΔC. The measured K_d was 29 (±5) μM (Table 1), which indicates that even a smaller fragment of G2BR has effects that translate through Ube2g2 and increase its affinity for the gp78 RING finger.

G2BR Enhances RING Finger-Dependent Ubiquitylation by Ube2g2

gp78 is a substrate for its own ubiquitylation in cells (Fang et al., 2001; Chen et al., 2006). To begin to assess the effect of the G2BR on ubiquitylation, a GST fusion of the cytoplasmic domain of gp78 (GST-gp78C) was employed in an autoubiquitylation assay (Lorick et al., 1999). Ube2g2 was provided in ~5-fold molar excess relative to glutathione Sepharose-immobilized GST-gp78C and ubiquitylation of bead-bound material was assessed. Ubiquitylation was totally dependent on an intact gp78 RING finger (Figure 5A), and mutations of two key residues in the G2BR (L582S/L589S; gp78CL582,589S) significantly decreased ubiquitylation. A truncated form of GST-gp78C lacking the G2BR (GST-gp78CΔ577-643) also demonstrated markedly diminished ubiquitylation (Figure 5B), consistent with G2BR-dependent increased affinity of Ube2g2 for the gp78 RING finger. However, when G2BR peptide was provided ‘in trans’ there was a marked increase in ubiquitylation of truncated gp78 (Figure 5B). This rescue of activity was not observed with the G2BRΔN, G2BRΔC, G2BRM4-1, or G2BRM4-2 peptides (Figure 5C and 5D). To determine whether the G2BR peptide also enhances ubiquitylation of heterologous proteins, we took advantage of gp78’s function as an E4 in ubiquitylating proteins already modified with a single ubiquitin (Morito et al., 2008). A fusion protein of ubiquitin and GFP with a C terminal His₆ tag (Ub-GFP-His₆) was ubiquitylated by GST-gp78CΔ577-643 as assessed using Flag-tagged ubiquitin (Figure 5E). As with gp78 autoubiquitylation, ubiquitylation of Ub-GFP-His₆ by GST-gp78CΔ577-643 was substantially increased in the presence of G2BR. Thus, the effect of the G2BR applies to ubiquitylation of a heterologous substrate as well as to autoubiquitylation.

To evaluate the effect of G2BR on substrate ubiquitylation using the full length gp78 cytoplasmic tail, gp78C and gp78CL582,589S were compared. gp78C resulted in easily detectable high molecular weight ubiquitylated Ub-GFP-His₆, which was not evident with gp78CL582,589S (Figure 5F). As gp78CL582,589S showed some persistent autoubiquitylation in Figure 5A, these reactions were carried out by combining G2BR mutants with double mutations of Ube2g2 in key contact points (Figure 5G; see Figure 2 and Supplemental Table 1 for Ube2g2:G2BR contacts). While none of the double mutants of Ube2g2 substantially decreased ubiquitylation with either gp78C or a single mutation (gp78CL582S), combining any of these E2 mutants with gp78CL582,589S resulted in a complete loss of ubiquitylation. Collectively these findings demonstrate the functional importance of structurally-defined key contact residues in the Ube2g2-G2BR interface.

G2BR Enhances Ubiquitylation by Ube2g2 with other RING Fingers

Ube2g2 also functions with other RING finger ERAD E3s (our unpublished observations). Two of these, HsHRD1 and Trc8, are implicated in human disease (Kostova et al., 2007). These E3s also function, at least in vitro, with members of the UbcH5 (Ube2d1-3) E2 family (Lorick et al., 1999; Nadav et al., 2003; Kikkert et al., 2004). HsHRD1 and Trc8 lack identifiable regions that are analogous to the G2BR. We therefore asked whether the enhanced ubiquitylation observed with the G2BR was unique to the gp78 RING finger. Addition of the G2BR to ubiquitylation reactions utilizing GST-fusions of the C-terminal RING finger-containing regions of either HsHRD1 or Trc8 resulted in a marked increase in RING finger-dependent autoubiquitylation (Figure 5H upper panel lanes 5 and 8 and Supplemental Figure 8). Ubiquitylation mediated by these E3s as well as by GST-gp78CΔ577-643 was not increased by the G2BR when UbcH5B was used as the E2 (Figure 5H, lower panel). Similarly, critical mutations in the G2BR do not decrease ubiquitylation by UbcH5B (Supplemental Figure 9). Thus, changes in the ERAD E2, Ube2g2, induced by binding of the G2BR at a site distant from where this E2 interacts with RING fingers, enhances ubiquitylation with at least three different mammalian E3s. This indicates both the specificity of the G2BR for Ube2g2 and its apparent effect on increasing productive interactions between Ube2g2 and RING fingers.

Increased Affinity of Ube2g2 for the gp78 RING Finger Accounts for the Enhanced Ubiquitylation

To further evaluate the effect of the G2BR, the rate of discharge of ubiquitin from Ube2g2 was assessed, based on a previously used approach (Petroski and Deshaies, 2005), in the presence of saturating amounts of G2BR or an equal concentration of Scr. In these reactions acceptors for ubiquitin are other proteins in the reaction mix, including bacterial proteins and excess added ubiquitin. In the absence of the RING finger, discharge of ubiquitin from Ube2g2~Ub was slow and unaffected by G2BR peptide (Figure 5I). In the presence of 8 μM gp78 RING finger, loss of Ub-bound Ube2g2 was markedly accelerated. This was significantly increased by saturating amounts of G2BR (Figure 5I). We wished to determine if this apparent increased rate of discharge can be accounted for by an increased population of Ube2g2:RING finger due to the presence of G2BR (< 5% Ube2g2 bound with Scr; > 50% bound with G2BR), or as yet unknown allosteric effects on the Ube2g2:gp78 RING finger complex mediated by the G2BR. Therefore, experiments were carried out with concentrations of gp78 RING finger at which approximately equal amounts of RING finger are bound to Ube2g2. Under these conditions no significant increase in discharge rate was observed in the presence of G2BR (Figure 5J). Thus, the G2BR-mediated increase in affinity of Ube2g2 for the RING finger provides a molecular basis for the accelerated discharge of ubiquitin from Ube2g2~Ub and, by extension, for the observed enhancement of ubiquitylation.

Structural Basis for the High Affinity G2BR Binding to Ube2g2

The NMR and X-ray crystallographic determination of the G2BR binding site on Ube2g2 provides an explanation for its high affinity interaction with gp78. The G2BR binds Ube2g2 through an extended region of the core UBC domain of Ube2g2 that includes the C-terminal ends of both the α1 and α4 helices and extensive contacts with the intervening β sheets (β1-3). This backside binding substantially overlaps the non-covalent binding site for ubiquitin on UbcH5C (Brzovic et al., 2006). While ubiquitin binds UbcH5C with a K_d of ~300 μM, which is similar to the analogous binding of ubiquitin to Ube2g2 (Das and Byrd, unpublished observations), the binding of the G2BR to Ube2g2 is of higher affinity by over 10,000-fold. The basis for this difference is explained by the interacting interfaces. Ubiquitin, characterized by a stable, ordered structure, interacts with E2s through its hydrophobic face centered on I44. The G2BR, on the other hand, folds from an unstructured state into a well-ordered α-helix upon interacting with Ube2g2. At their interface, hydrophobic sidechains along the spine of the G2BR α-helix interlock with Ube2g2 in a fashion similar to the UbcH5C:Ub complex. Additionally, charged and polar residues distributed along the G2BR α-helix, on either side of the hydrophobic spine (Figure 2C and 2D) form salt-bridges and hydrogen bonds with Ube2g2, resulting in a high-affinity complex with a K_d of ~21 nM.

Implications of Ube2g2:G2BR for Ubiquitylation and ERAD

Until now, the existence of the G2BR as a high affinity binding site for Ube2g2 had been assumed to primarily provide a means to increase the level of Ube2g2~Ub in proximity to gp78 and to enhance ubiquitylation (Chen et al., 2006). We have now uncovered additional effects of the G2BR. This highly-specific 27 aa binding site for Ube2g2 has the striking effect of increasing the affinity of Ube2g2 for the gp78 RING finger as a consequence of conformational/allosteric changes in Ube2g2. The basis for the dramatic change in K_d, from 144 μM to 3 μM, is not yet fully understood and awaits further structural studies. A striking consequence of this increased affinity is the increased association of Ube2g2 with the gp78 RING finger, thereby facilitating the discharge of ubiquitin from Ube2g2~Ub through as yet undefined RING finger-mediated effects (Ozkan et al., 2005; Petroski and Deshaies, 2005). This is reflected in enhanced ubiquitylation even when the G2BR peptide is provided ‘in trans’ to Ube2g2 together with the cytoplasmic domain of gp78 lacking the G2BR sequence. Significantly, at least in vitro, the G2BR also increases ubiquitylation by heterologous ERAD RING finger E3s. In the cell, this potential for cross-talk between different ERAD ubiquitin ligases has the potential to contribute to associations between ERAD RING finger E3s and more importantly to promote synergistic interactions among these proteins (Ye et al., 2005; Morito et al., 2008). Similarly, if the reported gp78-oligomerization-dependent formation of K48 ubiquitin chains on C89 of Ube2g2 occurs in cells (Li et al., 2009), this will be facilitated by the enhanced affinity of Ube2g2 for the gp78 RING finger with G2BRs. This domain can serve both as platforms for Ube2g2 bearing nascent ubiquitin chains and as a means to enhance transfer of ubiquitin from heterologous Ube2g2s.

In addition to the increase in RING finger binding, the Ube2g2:G2BR structure, compared to apo-Ube2g2, reveals a conformation in which there is narrowing of the channel projecting from the active site Cys89. This is largely due to repositioning of the extended mobile acidic portion of the β4α2 loop including a dramatic alteration in orientation of sidechains of M101 and Y103. All of the available structural data indicates that this region is dynamically disordered (Arai et al., 2006; Li et al., 2009; this study). One consequence of this repositioning is a substantial delay in formation of Ube2g2~Ub in the presence of G2BR. Depending on the rate limiting step(s) in vivo, slowing of E2 loading with ubiquitin in the presence of the G2BR could allow for ‘proofreading’ by deubiquitylating enzymes in either substrate selection or chain linkage. The region analogous to the β4α2 loop is important in interactions between other E2-E3 pairs (Lorick et al., 2006) as well as for Ube2g2:gp78 RING finger (this study). The acidic extension in this loop is also found in Cdc34. Mutations in acidic residues in this region in both Ube2g2 and Cdc34 translate into decreased RING finger-dependent formation of K48-linked polyubiquitin chains (Li et al., 2007; Petroski and Deshaies, 2005). Moreover, Cdc34-Ub recently has been shown to manifest ~2 fold lower K_d for the core SCF E3 than Cdc34 (Saha and Deshaies, 2008). As the net effect of the G2BR, either in cells or in vitro, is to enhance ubiquitylation, potential effects of this dynamic loop on modulating transfer of ubiquitin from Ube2g2~Ub, formation of K48 ubiquitin chains, or in differential RING finger binding of Ube2g2~Ub versus Ube2g2 may be of equal or greater significance than the pronounced inhibition of loading with ubiquitin that we observe. Addressing these possibilities and the unresolved question of whether ubiquitin can be transferred from E1 to Ube2g2 bound either to the full cytoplasmic tail of gp78 or the isolated G2BR are important questions in ERAD and in E2 function that will require additional in vitro and cellular approaches.

Existence and Implications of E2 Binding Sites Distinct from Ligase Domains

E2s interact with E1 as well as with HECT, RING and RING finger-like domains on E3s. The question now arises as to the prevalence of other functionally significant E2 interactions. A well-characterized example is the E2-like molecule Mms2 (UEV1a), which binds Ubc13 through a region distinct from the G2BR-Ube2g2 interface and facilitates formation of K63-linked ubiquitin chains (Eddins et al., 2006). Another example, mentioned above, is ubiquitin, which plays a role in the processivity of BRCA1-mediated ubiquitin chain formation, presumably through alignment of multiple UbcH5C~Ub complexes (Brzovic et al., 2006; Christensen et al., 2007). The high affinity interaction of Ube2g2:G2BR precludes backside ubiquitin binding in the context of Ube2g2 bound to gp78. However, it does not exclude that alignment of multiple E2~Ub complexes (not necessarily Ube2g2) in vivo could occur in the context of a G2BR-bound Ube2g2~Ub and play a role in chain formation.

There are several examples where regions of E3s, which are not included in their canonical ligase domains, bind E2s, e.g. Nedd4 and Ubr1p. For these, neither the sites of interaction on their respective E2s nor the functional effects of this binding are known (Madura et al., 1993; Hatakeyama et al., 1997). There is evidence that SCF E3s recruit Cdc34 through part of the C-terminal extension of this E2 (Wu et al., 2002). Yeast Ubc7p, which is the ortholog of Ube2g2, is recruited by Cue1p to the ER membrane to function with the ERAD E3s Hrd1p and Doa10p (Kostova et al., 2007). Recent findings suggest that the 180 amino acid cytoplasmic domain of Cue1p activates Ubc7p in vitro in a RING finger-independent manner through unknown mechanisms (Bazirgan and Hampton, 2008; Kostova et al., 2009). Moreover, a ~50 aa domain in Cue1p, with some sequence homology to G2BR, directly binds Ubc7p. This domain is sufficient to activate ERAD by Hrd1p in vivo, and stimulates ubiquitylation in vitro, analogous to the G2BR (Kostova et al., 2009) and thus may represent a yeast equivalent of the G2BR.

A possible parallel to gp78 and Ube2g2 comes from the SUMO E3 Nup358/RanBP2. This E3 binds the SUMO E2 (Ubc9) through regions on Ubc9 analogous to interactions between ubiquitin E3s and E2s. However, this E3 also contacts Ubc9 through a second region with similarity to the Ube2g2:G2BR interface (Reverter and Lima, 2005). As pointed out by Reverter and Lima, mutations in this interface impact Ubc9 binding and Nup358/RanBP2 function (Pichler et al., 2004; Tatham et al., 2005). It would be of interest as to whether this second site of E2:E3 interaction has analogous effects to those observed in this study.

It seems reasonable to postulate that E2 binding sites within E3 complexes, but distinct from canonical ligase domains, may be a property of a substantial fraction of E3s. To what extent these result in functionally significant allosteric effects on E2s, as demonstrated herein, now becomes an exciting area for future research.

NMR Spectroscopy

NMR samples were prepared in 50 mM Tris, 2 mM TCEP, pH 7.5 buffer and experiments were performed at 25°C. NMR spectra were acquired on 600 and 800MHz Varian INOVA spectrometers equipped with triple resonance gradient cryoprobes. Details of resonance assignments, binding titrations and intermolecular NOESY experiments are provided in the Supplemental Experimental Methods.

X-ray Diffraction

Crystallization screens were carried out with a Phoenix robot (Art Robbins Instruments). The crystals were flash-frozen in liquid N₂ after a short soak in a cryo-protection solution. A native dataset was collected at beamline 22-ID of the Advanced Photon Source, Chicago. The structure was solved by molecular replacement. The initial Fo–Fc map revealed the electron density for the G2BR. Details of crystallization, data acquisition, structure solution, model building and structure refinement are provided in Supplemental Experimental Methods.

Isothermal Titration Calorimetry

ITC was carried out using a VP-ITC microcalorimeter (MicroCal LLC, Northampton, MA) at 25°C. The typical experiment included injection of 25-27 aliquots (10 μL each) of 0.1-0.5 mM peptide solution into a 0.01-0.10 mM protein solution in the ITC cell (volume ~1.4 ml) stirring at 300 rpm. Additional details are provided in Supplemental Experimental Methods.

In Vitro Ubiquitylation, E2 loading and Discharge

Autoubiquitylation was carried out as described (Kostova et al., 2009; Lorick et al., 1999). Ubiquitylation of bacterially expressed purified Ub-GFP-His₆ (~500 nM) was carried out using Glutathione Sepharose 4B bound GST-gp78CΔ577-643. Reactions were terminated by 2% SDS (final). After 10 min samples were diluted with 8 M urea in 50 mM Tris pH 7.4 to 0.1% SDS. Ub-GFP-His₆ was isolated on Ni²⁺ beads and eluted with SDS-PAGE sample buffer after washing. ³⁵S-labeled Ube2g2 was translated in the S30 T7 bacterial TnT system (Promega) and used at <100 nM. In loading experiments Ub K0 was used. For discharge of ubiquitin from E2, E2 was loaded with wild-type ubiquitin for 10 min. Reactions were quenched with 5 mM NEM and the buffer exchanged to 50 mM Tris (pH 7.4) with Ub (80 μM), peptide (4 μM) with or without purified gp78 RING finger (aa 313-380); discharge was monitored at 25°C. Additional details are provided in Supplemental Experimental Methods.