Structural Basis of E2–25K/UBB+1 Interaction Leading to Proteasome Inhibition and Neurotoxicity*

E2–25K/Hip2 is an unusual ubiquitin-conjugating enzyme that interacts with the frameshift mutant of ubiquitin B (UBB+1) and has been identified as a crucial factor regulating amyloid-β neurotoxicity. To study the structural basis of the neurotoxicity mediated by the E2–25K-UBB+1 interaction, we determined the three-dimensional structures of UBB+1, E2–25K and the E2–25K/ubiquitin, and E2–25K/UBB+1 complex. The structures revealed that ubiquitin or UBB+1 is bound to E2–25K via the enzyme MGF motif and residues in α9 of the enzyme. Polyubiquitylation assays together with analyses of various E2–25K mutants showed that disrupting UBB+1 binding markedly diminishes synthesis of neurotoxic UBB+1-anchored polyubiquitin. These results suggest that the interaction between E2–25K and UBB+1 is critical for the synthesis and accumulation of UBB+1-anchored polyubiquitin, which results in proteasomal inhibition and neuronal cell death.

The pathological hallmarks of neurodegenerative diseases are the accumulation of abnormal proteins, neuritic plaques, neuropil threads, and neurofibrillary tangles (1). Although all aspects of this aggregation are considered harmful, the aberrant protein accumulation is likely the principal cause of disease. Among these proteins are presenilin-processed amyloid-␤ peptide (A␤), 4 a mutant form of ubiquitin (UBB ϩ1 ), and polyglutamine-expanded huntingtin (2). UBB ϩ1 was first identified as a frameshift mutant of the ubiquitin (Ub) B protein in the brains of neurodegenerative disease patients (3) and is composed of a Ub moiety (75 residues) with a 19-residue C-terminal extension. Neither A␤ nor UBB ϩ1 is found in young patients not suffering from dementia, but they are observed in older Alzheimer disease patients (4). The genes from which UBB ϩ1 mRNAs are transcribed contain several GAGAG motifs, and dinucleotide deletions (⌬GA) from within the GAGAG motif result in an abnormal C-terminal sequence. Normally these aberrant proteins are removed by the Ub-proteasome system (UPS), which executes the proteolytic degradation of aberrant proteins via a Ub-tagging mechanism (3,5).
Within the UPS, Ub tagging of target molecules entails enzymatic reactions catalyzed by the E1 (Ub-activating), E2 (Ubconjugating), and E3 (Ub-ligating) enzymes. Once E3 tags a target molecule with mono-or polyUb, the tagged molecule is recognized by the 26 S proteasome and degraded (6). In the healthy brain both ␤-amyloid precursor protein and UBB ϩ1 molecules are targets for the UPS and are degraded by the 26 S proteasome (7,8). In the brains of Alzheimer patients, however, both UBB ϩ1 and Ub are present within aggregation plaques also containing ␤-amyloid precursor protein, which is indicative of UPS dysfunction (9,10). When at normal basal levels, UBB ϩ1 can be removed by the UPS. But when its expression is up-regulated, UBB ϩ1 inhibits the 26 S proteasome in a dose-dependent manner, resulting in the accumulation of aberrant proteins (11). The aberrant C terminus of UBB ϩ1 prevents its activation and, therefore, subsequent ligation to substrates due to  the absence of a residue corresponding to Gly76 in Ub; instead, UBB ϩ1 serves as a scaffold for ligation of Ub molecules to produce polyUb that is anchored by UBB ϩ1 (on the unaffected Lys-48 site). These molecules are reportedly refractory to the deubiquitinating enzyme system (12). Consequently, when UBB ϩ1 -anchored polyUb is targeted to the 26 S proteasome, it acts as a functional antagonist, inhibiting the activity of the proteasome and leading to A␤ neurotoxicity (13,14). E2-25K is an unusual member of the E2 family in that it is competent to catalyze ubiquitin chain extension independent of E3 ligases (15). In addition to an E2 domain that is 22-48% identical to other human E2s, E2-25K also contains a UBA domain that is unique to this paralog (supplemental Fig. 1A). UBA domains appear to be generally involved in interactions with Ub, but the precise function of the E2-25K UBA domain is currently unclear, although evidence suggests it is important for polyubiquitylation activity. Moreover, a chimeric protein in which the UBA domain of E2-25K was fused to the E2 domain of yeast UBC4 showed no polyUb synthetic activity, suggesting that polyubiquitylation by E2-25K is dependent on the relative conformations of the E2 and UBA domains and their specific interactions (15). Interestingly, it was recently reported that an active site mutation, C92S or S86Y, or deletion of the UBA domain of E2-25K eliminated A␤ neurotoxicity (13). Based on these findings, it was proposed that UBB ϩ1 interacts with the UBA domain of E2-25K and participates in the polyubiquitylation process that produces the UBB ϩ1 -anchored polyUb chains. Details of the mechanism are not yet known, however.
In this report we present the three-dimensional structures of UBB ϩ1 , E2-25K, E2-25K/Ub, and E2-25K/UBB ϩ1 determined by x-ray crystallography and NMR spectroscopy. We also analyzed in detail the interaction between E2-25K and UBB ϩ1 and propose that the accumulation of UBB ϩ1 -anchored polyUb reflects a direct interaction between the E2-25K UBA domain and UBB ϩ1 and results in cell death due to proteasomal inhibition.
Protein Purification and Sample Preparation-All plasmid DNAs were transformed into E. coli strain BL21 (DE3), which was grown in LB or M9 minimal medium for 12 h at 25°C after induction with 1 mM isopropyl-1-thio-␤-D-thiogalactopyranoside. After the affinity chromatography, we removed the TRX-His tag (pET32a) or His tag (pET28b) by incubation with tobacco etch virus or thrombin protease, respectively, for 12 h at 25°C. The purified proteins were then subjected to size exclu- Crystallization and Structure Determination-E2-25K, Ub, and UBB ϩ1 were concentrated to 20, 30, and 30 mg/ml, respectively, using a centriprep (Amicon). The buffer used for E2-25K and Ub crystallization contained 20 mM HEPES-NaOH (pH 7.5) and 150 mM NaCl. The buffer used for UBB ϩ1 crystallization contained 20 mM Tris-HCl (pH 8.0) and 150 mM NaCl. E2-25K crystals were grown at 18°C in a 2-l sitting drop containing equal volumes of protein solution and mother liquor (100 mM sodium cacodylate (pH 6.6), 18% (w/v) PEG 8000, and 100 mM sodium acetate). The purified E2-25K was mixed with Ub or UBB ϩ1 at a 1:2 molar ratio. Crystals of E2-25K/Ub and E2-25K/UBB ϩ1 complexes were grown in a 2-l hanging drop containing equal volumes of the protein solution and mother liquor (E2-25K/Ub: 100 mM HEPES-NaOH (pH 7.5), 25% (w/v) PEG 3350, and 50 mM sodium fluoride; E2-25K/UBB ϩ1 : 100 mM Tris-HCl (pH 8.5) and 25% (w/v) PEG 3350). The crystals were cryoprotected in reservoir solution supplemented with 10% or 25% (v/v) glycerol. All the diffraction data were collected at 100 K. Diffraction data for E2-25K were collected at 2.4 Å resolution on a Rigaku FR-E generator equipped with a R-AXIS IV 2ϩ detector, then processed and scaled using CRYSTAL-CLEAR (Rigaku). The structure was solved using the molecular replacement method with the program AMORE (17) using the coordinates of yeast UBC1 (PDB code 1TTE) as the search model. Diffraction data for the E2-25K/Ub and E2-25K/ UBB ϩ1 complexes were collected at 2.8 and 1.8 Å resolution, respectively, at beam line 4A at the Pohang Accelerator Laboratory, Korea, then processed and scaled using HKL2000 (HKL Research) (18). The structures of the two complexes were solved using the molecular replacement method with the program MOLREP (19) using the structures of E2-25K and Ub (PDB code 1UBQ) as the search models. Native and complex structures were subjected to many cycles of manual rebuilding using the program O (20) and were refined using the programs CNS (21) and REFMAC5 (22). The final structures were analyzed using PROCHECK (23). The statistics for the structure refinement are summarized in Table 1.
NMR Experiments and Backbone Dynamics-All NMR experiments were performed in a mixture of 90% H 2 O and 10% 2 H 2 O or 99% 2 H 2 O NMR buffer (50 mM Na 2 PO 4 , 100 mM NaCl, 2 mM DTT, and pH 7.0) at 298 K on a Bruker DRX 900 MHz equipped with a CryoProbe TM system. The chemical shift in 1 H was referenced directly to internal sodium 4,4-dimethyl-4-silapentane-1-sulfonate, and 13 C and 15 N were referenced indirectly to 1 H (24). All spectra were processed using XWINNMR (Bruker Biospin Corp.) and NMRpipe/NMRDraw software (25). Triple resonance experiments were executed for backbone assignment (26). The side-chain assignments were complicated by HCCH-TOCSY (two-dimensional total correlation spectroscopy) and 13 C-edited NOESY (27). 15 N-Edited NOESY-HSQC (m ϭ 120 ms) and 13 C-edited NOESY-HSQC (m ϭ 120 ms) experiments were used for gathering structure con-straints. To investigate the dynamic properties of UBB ϩ1 , the heteronuclear 1 H, 15 N NOEs were measured (28). The XNOE measurements were performed on 1024 ϫ 128 complex points for t 2 ϫ t 1 dimensions with a 5-s recycle delay.
NMR Structure Calculation of UBB ϩ1 -The solution structure of UBB ϩ1 was obtained by applying CYANA (29). Distance restraints and angle constraints were gathered from NOESY experiments and the TALOS program (30). NOE restraints consisting of 1514 unambiguous NOEs (843 short range NOEs ( I Ϫ j ϭ 1), 216 medium range NOEs (1 Յ I Ϫ j Ͻ5), and 455 long range NOEs (5 Յ I Ϫ j )) and 106 torsion angle restraints were used for structure calculations ( Table 2). The final structures were analyzed using PROCHECK (23). With the exception of residue Asp-82 in the tail region (0.01%), all residues were located on the favored and allowed regions (99.99%) of the Ramachandran plot. PyMOL (Delano Scientific LLC, San Carlos, CA) was used with APBS and MOLMOL (31) to analyze and visualize the final structures.

Structural Basis of E2-25K and UBB ؉1 Interaction
E. coli BL21 (DE3) cells, and purified on glutathione-Sepharose. Wild-type or mutant E2-25K was incubated with GST alone, GST-Ub, or GST-UBB ϩ1 in 1ϫ PBS at a molar ratio of 1:1 for 3 h at 4°C in the presence of 50 l of glutathione-Sepharose beads (Peptron). The bound proteins were eluted by boiling in SDS-PAGE sample buffer and analyzed by Western blotting using an anti-E2-25K antibody. Cell Culture, DNA Transfection, and Cell Death Assays-B103 cells (rat neuroblastoma cells) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal bovine serum. Cells were transfected using Lipofectamine TM reagent (Invitrogen) according to the manufacturer's protocol, after which their viability was assessed based on the morphology of GFP ϩ cells viewed under a fluorescence microscope (Leica DMRBE). We measured the accumulation green fluorescent protein fused to a short degron (GFPu) and Discosoma sp. red fluorescent protein (DsRed) as an index of intracellular proteasome activity in 293F cells overexpressing wild-type E2-25K or E2-25K mutants. Twenty-four hours after transfection of pDsRed plasmid and wild-type or mutant pGFPu-E2-25K plasmid into 293F cells, the cells were examined under a fluorescence microscope. 293F cells expressing only GFP (pcDNA3) were detected based on their red color; because the GFP was successfully removed by proteasome, only DsRed was detected. E2-25K transfectants were detected based on their yellow color, reflecting the combination of the green fluorescence from GFPu and red fluorescence from DsRed.
Proteasome Inhibition Assay-Inhibition of proteasome was examined as previously described by Song et al. (13). Briefly, pGFPu (an artificial proteasome substrate; 50 ng) together with each plasmid expressing E2-25K (500 ng) and pDsRed (50 ng) was transiently transfected into 293F cells. After 24 h the transfectants were examined under a fluorescence microscope to determine the relative numbers of GFP ϩ and DsRed ϩ cells.
Analysis of UBB ϩ1 -anchored PolyUb Synthesis-To assay UBB ϩ1 anchoring, the ubiquitylation activities of E2-25K and its mutants were measured based on an in vitro reaction run with UBB ϩ1 (50 M), , and ATP (5 mM). The 20-l reaction mixtures containing wildtype or mutant E2-25K proteins were incubated for 4 h at 37°C and analyzed by Western blotting with rat polyclonal anti-UBB ϩ1 antibody (2000:1), which was previously described by Song et al. (13).
Crystal Structure of E2-25K/Ub Complex-To understand the structural basis of Ub recognition by E2-25K, we solved the structure of the E2-25K/Ub complex (Fig. 2A). The structures of the Ub-complexed and apo forms of E2-25K were nearly identical with an r.m.s.d. of 0.77 Å for 196 C ␣ atom pairs. E2-25K binds to Ub mainly via residues in the ␣7-␣8 loop and helix ␣9 and involves about 400 Å 2 of surface area (4.1% of total surface area) (Fig. 2B). That the surface of the intermolecular contact is relatively small is consistent with the low binding affinity (K d ϭ 400 M) between E2-25K and Ub (35), as compared with other UBA domains whose affinities for  (Fig. 2B). The backbone NMR resonance assignments of FIGURE 2. Structure of the E2-25K/Ub complex. A, the crystal structure of the E2-25K/Ub complex is presented with the E2 domain, UBA domain, and Ub colored forest, pale green, and brown, respectively. Lys-14, Lys-48, and Cys-92 are presented as space-filling models. B, the interaction interface between E2-25K and Ub is presented and enlarged. The side chains of the interacting residues in E2-25K and Ub are in white and orange, respectively. The red dash lines indicate hydrogen bonds. C, 15 N-labeled E2-25K was titrated with Ub. Various molar ratios of E2-25K and Ub (1:0, 1:1, 1:2.5, 1:5, 1:7.5, and 1:10) are displayed as black (a), blue (b), cyan (c), green (d), gold (e), and red (f) peaks, respectively. D, GST pulldown assay was carried out using GST-Ub and E2-25K. Proteins were resolved by 15% SDS-PAGE and visualized by Western blotting with anti-E2-25K antibody.

Structural Basis of E2-25K and UBB ؉1 Interaction
E2-25K were completed using data from TROSY-based triple resonance experiments (supplemental Fig. 2). The TROSY-HSQC spectra for E2-25K at various molar ratios in solution with Ub were consistent with the crystal structure and showed that residues involved in intermolecular interaction experience marked chemical shift changes upon Ub binding ( Fig. 2C and  supplemental Fig. 3A). The average K d was 1.514 Ϯ 0.342 mM. Sequence comparison of the E2-25K UBA domain with other UBA domains showed that the residues forming the hydrophobic patch that interact with Ub are well conserved (supplemental Fig. 1B).
Structure of the E2-25K/UBB ϩ1 Complex-The solution structure of UBB ϩ1 shows a well structured and defined region with a Ub sequence (M1-G75) and a C-terminal structure with an aberrant tail (Tyr-76 -Gln-95) ( Fig. 3A and supplemental  Fig. 3B). When we separately superimposed the two regions using backbone atoms, the Ub region was well fitted with an r.m.s.d. of 0.54 Ϯ 0.16 Å, whereas the tail region (G75-H88) had an r.m.s.d. of 1.86 Ϯ 0.23 Å. This implies that the aberrant tail of UBB ϩ1 has a structure, although it is not well defined due to a lack of NOE information. Heteronuclear NOE data were consistent with the structural tendency of the tail region, i.e. the tail region had a heteronuclear NOE value of about 0.48 (Fig. 3B), which indicates the region has a structure that is independent of the Ub region. In addition, the two regions of UBB ϩ1 show distinctly different dynamic behaviors (supplemental Fig. 3C).
To better understand the structural basis of the interaction between UBB ϩ1 and E2-25K, we determined the crystal structure of the E2-25K/UBB ϩ1 complex. The overall structure of the complex was similar to that of the E2-25K/Ub complex; the r.m.s.d. between the two structures was 1.08 Å for 269 C ␣ atom pairs (Fig. 4, A and B). In addition, the r.m.s.d. between E2-25K in complex with Ub and UBB ϩ1 was 0.69 Å for 197 C ␣ atom pairs, whereas the r.m.s.d. between Ub and UBB ϩ1 in complex with E2-25K was 0.59 Å for 72 C ␣ atom pairs. Notably, the electron density for the main chain C-terminal residues (amino acids 73-75) was clearly observable in the E2-25K/UBB ϩ1 structure, whereas this region was highly disordered and invisible within the Ub complex. Among these residues, Leu-73 was highly ordered, and its side chain was clearly observable in the electron density map. Leu-73 is involved in the intermolecular interaction between E2-25K and UBB ϩ1 that enables the UBB ϩ1 molecule to tilt 15°with respect to Ub (Fig. 4C and  supplemental Fig. 4).
We next carried out a set of NMR titration experiments to examine the mode of E2-25K binding to UBB ϩ1 in solution. Fig. 5A shows the chemical shift changes in E2-25K in the presence of UBB ϩ1 . Consistent with the crystal structures, most of the changes clearly seen upon Ub binding are in the UBA domain. However, at high concentration of UBB ϩ1 , small changes are also observed in the E2 domain, presumably due to nonspecific interactions involving the flexible tail. Met-172, Gly-173, Phe-174,  in the UBA domain all showed chemical shift changes greater than 0.18 ppm, whereas Ala-171 and Thr-194 showed changes of 0.08 -0.18 ppm (Fig. 5B).
In a set of similar experiments, we also examined the effect of E2-25K binding on UBB ϩ1 . Most changes elicited by E2-25K binding occurred in ␤3, ␤4, and the ␣2-␤3 loop (Fig. 5C), which  NOVEMBER 12, 2010 • VOLUME 285 • NUMBER 46 form the hydrophobic patch (Fig. 5D). This finding confirms that intermolecular contacts between E2-25K and UBB ϩ1 in solution occur mainly between hydrophobic residues, as was observed in the crystal structure. In addition, Lys-48 and Leu-71 of UBB ϩ1 showed larger chemical shift changes than were seen with Ub titration (Fig. 5C and supplemental Fig. 5C). The rigid C-terminal region (resides 73-75) observed in the crystal structure may contribute to the dramatic change in chemical shift for Leu-71, although the chemical shift changes for residues 73-75 themselves were little affected, probably because their perturbation was not sufficient to induce significant transition of the backbone amide proton environment.

Structural Basis of E2-25K and UBB ؉1 Interaction
We next used a nonlinear curve-fitting method to determine the K d for the binding of E2-25K and UBB ϩ1 derived from the NMR titrations. Based on the binding isotherms, we calculated the plots of ⌬␦ 1H versus molar ratio for residues Met-172, Gly-173, Phe-174,   Fig. 3D and Table 1). The average K d for E2-25K binding to UBB ϩ1 was 0.939 Ϯ 0.218 mM, whereas the K d for E2-25K binding to Ub was 1.514 Ϯ 0.342 mM FIGURE 4. Structure of E2-25K/UBB ؉1 complex. A, the crystal structure of the E2-25K/UBB ϩ1 complex is displayed as a ribbon diagram. The E2 and UBA domains of E2-25K and the Ub region of UBB ϩ1 are shown in green, pale green, and navy, respectively. The red dashed line indicates the C-terminal tail region of UBB ϩ1 , which was not observed in the electrodensity map. Lys-14, Lys-48, and Cys-92 are depicted as space-filling models. B, shown is magnification of the E2-25K and UBB ϩ1 interface with specific binding residues displayed. E2-25K and UBB ϩ1 of the UBA domain are in pale green and navy, respectively. Hydrogen bonds are indicated by red dash lines. C, the overlapped structures of the E2-25K/Ub and E2-25K/UBB ϩ1 complexes are displayed. The two were superimposed on the UBA domain using backbone atoms, and the interacting residues that affect the orientation of the protein are represented. D, a GST pulldown assay for GST-tagged UBB ϩ1 with E2-25K and its mutants is presented. Proteins were resolved by 15% SDS-PAGE and visualized by Western blotting with an anti-E2-25K antibody.
To evaluate the importance of the Ub/UBB ϩ1 binding site in the assembly of UBB ϩ1 -anchored polyUb, we carried out polyubiquitylation assays with E2-25K mutants in the presence of UBB ϩ1 (Fig. 6A). An E2-25K deletion mutant lacking the UBA domain showed almost no UBB ϩ1 -anchored products, whereas wild-type E2-25K readily formed UBB ϩ1 -anchored polyUb (Ub 1 -UBB ϩ1 and Ub 1 -Ub 2 -UBB ϩ1 ). Most of the E2-25K substitution mutants studied also showed a diminished ability to catalyze the synthesis of UBB ϩ1 -anchored polyUb, indicating that the interaction between the UBA domain and UBB ϩ1 is important for synthesis of UBB ϩ1 -anchored polyUb.
E2-25K UBA-dependent Proteasome Inhibition Mediates Neuronal Cell Death-The accumulation of UBB ϩ1 -anchored polyUb was previously shown to inhibit proteasome activity (37). To determine whether there is a direct relation between proteasome inhibition and the E2-25K/UBB ϩ1 interaction, we carried out proteasome inhibition assays with wild-type E2-25K and several E2-25K mutants. We monitored intracellular proteasome activity in mammalian cells using GFPu (38). Cells expressing wild-type E2-25K contained 4-fold higher levels of GFP than cells transfected with empty vector (pcDNA3). E2-25K mutants, including a UBA deletion mutant, which impaired UBB ϩ1 binding and synthesis of UBB ϩ1 -anchored Structural Basis of E2-25K and UBB ؉1 Interaction NOVEMBER 12, 2010 • VOLUME 285 • NUMBER 46 polyUb, showed significantly lower levels of GFP accumulation (Fig. 6B). Collectively, these data strongly suggest that the E2-25K/UBB ϩ1 interaction plays a key role in mediating UPS impairment through the formation of UBB ϩ1 -anchored polyUb.
We previously reported that proteasome inhibition by UBB ϩ1 -anchored polyUb is associated with neurotoxic activity in transfected neuronal cells (13). Given the structure of E2-25K/UBB ϩ1 , we surmised that the UBA domain of E2-25K is responsible for controlling the ubiquitylation of UBB ϩ1 , which is a major factor in proteasome inhibition and UBB ϩ1induced neurotoxicity. To confirm the role of the E2-25K/ UBB ϩ1 interaction in neurotoxicity, we carried out cell death assays using various E2-25K mutants with B103 neuronal cells. Mutations within the UBA domain led to reductions in cell death, as compared with wild-type E2-25K (Fig. 6, C and D). Correlations between the reduction in cell death and the positions of the mutated residues are presented in Fig. 6, C and D. They show that the critical residues whose mutation reduces the incidence of cell death are located within the hydrophobic patch formed by helices ␣7 and ␣9. Taken together these findings suggest that the interaction between UBB ϩ1 and the UBA domain of E2-25K and the resultant formation of UBB ϩ1 -anchored polyUb are directly responsible for the neuronal cell death caused by proteasome inhibition.

DISCUSSION
E2-25K is known to mediate A␤ neurotoxicity by promoting the inhibition of proteasome activity, altering the ER stress response, and activating caspase-12-dependent cell death (13,38,39). In addition, accumulation of polyUb-anchored UBB ϩ1 , the synthesis of which is in part catalyzed by E2-25K, directly inhibits proteasome activity. Notably, we observed that active site mutations (C92S or S86Y) or deletion of the E2-25K UBA domain eliminated A␤ neurotoxicity (13), and we propose that by interacting with UBB ϩ1 via the UBA domain, E2-25K contributes to the formation of UBB ϩ1 -anchored polyubiquitin chains. To investigate the structural and functional role of the UBA domain in Ub/UBB ϩ1 binding, UBB ϩ1 -anchored polyUb synthesis, and proteasome inhibition-mediated A␤ neurotoxicity, we used x-ray crystallography and NMR spectroscopy to determine the three-dimensional structures of UBB ϩ1 , E2-25K, and the E2-25K/Ub and E2-25K/UBB ϩ1 complexes.
Rigid Interaction between the E2 and UBA Domains-Among E2 proteins, E2-25K and its yeast homolog, UBC1, are unique in that they contain a C-terminal UBA domain in addition to an E2 domain. The E2 and UBA domains of UBC1 are connected by a flexible, 22-residue linker and are presumably independent of one another (40,41). By contrast, the domains in E2-25K are linked by a short, six-residue tether and extensively interact via hydrophobic residues, resulting in a single globular topology. These structural differences likely affect the respective polyUb chain-building properties of E2-25K and UBC1. For instance, whereas the UBC1 UBA domain is involved in autoubiquitylation at Lys-93, E2-25K, in the absence of E3 enzymes, is responsible for the synthesis of polyUb chains that are not anchored by substrates (42,43). Haldeman et al. (43) attempted to define the function of the E2-25K UBA domain by making a chimeric protein in which the E2-25K UBA domain and its short tether were fused to the E2 domain of yeast UBC4, which has no ability to synthesize polyUb. Notably, they found that the chimeric E2 protein was also unable to catalyze polyUb synthesis. This suggests that the polyubiquitylation activity of E2-25K is dependent on the relative orientations of the E2 and UBA domains. Mastrandrea et al. (44) also addressed this issue, finding that although an S86Y mutant was unable to generate polyUb, it was more resistant to proteolytic digestion. Within the E2-25K structure, Ser-86 is located in the ␤4-␣3 loop, which also contains the active site (Cys-92). Ser-86 forms a hydrogen bond with Asp-127 in the hydrophobic environment between the two domains, and this plays a major role in stabilizing the conformation of the ␤4-␣3 loop (Fig. 1C).
Ub/UBB ϩ1 Interaction with the E2-25K UBA Domain and Polyubiquitylation-The UBA domain of E2-25K is composed of three ␣ helices. The third helix and the loop between the first two (MGF motif) mediate the interaction with Ub in four UBA-Ub complexes (UQ1, PDB code 2JY6; Ede1m PDB code 2G3Q; Dsk2p, PDB code 1WR1; EDD, PDB code 1QHO) (supplemental Fig. 6 and Refs. [45][46][47][48]. Moreover, substitution of the Met or Phe residues in the MGF motif or the Leu residue in ␣3 disrupts that interaction. The UBA domain of E2-25K binds Ub mainly via the ␣7-␣8 loop (corresponds to the ␣1-␣2 loop in other UBA structures) and helix ␣9 (corresponds to ␣3). In addition to the well studied Met-172 and Phe-174 residues in the MGF motif and the Leu-198 residue in ␣9, mutation of Val-190 and/or Thr-194 in ␣9 dramatically disrupted Ub binding, indicating helix ␣9 is as important as the MGF motif for the E2-25K UBA-Ub interaction (supplemental Fig. 5B).
We have shown that Ub and UBB ϩ1 interact with the E2-25K UBA domain in similar fashion. Within the E2-25K/ UBB ϩ1 complex, however, three additional residues were visible in the Ub region (residues 73-75, which were invisible in E2-25K/Ub structure), resulting in an increase in the surface area of the E2-25K/UBB ϩ1 interface (ϳ440 Å 2 ), as compared with the E2-25K/Ub interface (ϳ400 Å 2 ) (supplemental Fig. 4). Consistent with the greater area of interaction, UBB ϩ1 showed a 1.5-fold greater affinity for E2-25K than Ub. In particular, the ␤4-strand of UBB ϩ1 is stabilized by the tail region, resulting in an increase in binding affinity.
It is known that UBB ϩ1 expressed under basal conditions can be removed by the UPS but that up-regulated expression of UBB ϩ1 leads to inhibition of the 26 S proteasome and, subsequently, cell death (11). Inhibition of the 26 S proteasome is also dependent on formation of UBB ϩ1 -anchored polyUb chains (37). We showed that E2-25K mutation that disrupted UBB ϩ1 binding diminished the E2-25K ability to synthesize UBB ϩ1anchored polyUb (Fig. 6A). Thus, the E2-25K UBA domain, especially the MGF motif and ␣9, is critical for the synthesis of polyUb-anchored by UBB ϩ1 .
E2-25K UBA-dependent Proteasome Inhibition and Neurotoxicity-UBB ϩ1 accumulates in the brains of Alzheimer disease patients, and UBB ϩ1 -anchored polyUb potently inhibits proteasome activity, which leads to neurotoxicity (3). Notably, when transfected into neuronal cells, a UBB ϩ1 K48R mutant exhibited no ability to inhibit proteasome and no neurotoxicity (12,13,37), suggesting that inhibition of proteasome activity by UBB ϩ1 -anchored polyUb synthesized by E2-25K is associated with the observed neurotoxicity. In this report we verified that the UBB ϩ1 binding sites on E2-25K, especially the MGF motif and ␣9 in the UBA domain, are critical for UBB ϩ1 -anchored polyUb synthesis, proteasome inhibition, and neurotoxicity (Fig. 6).
Regulation of the UPS own components is important in numerous basic cellular processes. Ub, a principal component of the system, must be maintained at adequate levels to support its homeostasis under basal and stressed conditions. Ub is reportedly degraded by the proteasome via a "piggyback" mechanism. Monomeric Ub is rapidly and efficiently degraded when fused with a long C-terminal extension (Ͼ20 amino acids), which may allow entry of the protein into the 20 S catalytic chamber. By contrast, although they bind to the 26 S proteasome, shorter tailed Ubs such as UBB ϩ1 are not adequately degraded, resulting in inhibition of UPS substrate degradation. This inhibition is dependent on the UPS substrate ability to be ubiquitinated, indicating that mere binding is not sufficient for the proteasome inhibition (49). Consistent with that finding, we verified that E2-25K mutation that disrupted UBB ϩ1 binding diminished its ability to synthesize UBB ϩ1 -anchored polyUb, which affected the proteasome inhibition and neurotoxicity.
E2-25K UBA domain-mediated proteasome inhibition and the resultant neurotoxicity has also been reported in Huntington disease (2). E2-25K directly interacts with huntingtin via its UBA domain and modulates the aggregation and toxicity of expanded huntingtin, resulting in polyglutamine-induced cell death. The E2-25K UBA domain is critical for aggregation of expanded polyglutamine proteins and polyglutamine-induced cell death. We cannot rule out the possibility that E2-25Kmediated impairment of the UPS is caused by both UBB ϩ1anchored polyUb and accumulation of extended polyglutamine proteins. Further study will be required to fully understand the mechanism by which E2-25K mediates UPS impairment and accumulation of toxic molecules, such as UBB ϩ1 -anchored polyUb and extended polyglutamine proteins, within neuronal cells. We anticipate that our discoveries will lead to the development of new approaches to the treatment of Alzheimer disease, Huntington disease, and other neurodegenerative disorders. In conclusion, polyubiquitylation, proteosomal inhibition, and cellular toxicity assays based on structural information about the E2-25K/Ub and E2-25K/ UBB ϩ1 complexes revealed that interaction between the E2-25K and UBB ϩ1 plays a key role in proteosomal inhibition and neuronal cell death.