|Home | About | Journals | Submit | Contact Us | Français|
Valosin-containing protein (VCP), also known as p97, is an AAA+ ATPase that plays an essential role in a broad array of cellular processes including the endoplasmic reticulum-associated degradation (ERAD) pathway. Recently, ERAD-specific deubiquitinating enzymes have been reported to be physically associated with VCP, although the exact mechanism is not yet clear. Among these enzymes is ovarian tumor domain-containing protein 1 (OTU1). Here, we report the structural basis for interaction between VCP and OTU1. The crystal structure of the ubiquitin regulatory X-like (UBXL) domain of OTU1 (UBXLOTU1) complexed to the N-terminal domain of VCP (NVCP) at 1.8-Å resolution reveals that UBXLOTU1 adopts a ubiquitin-like fold and binds at the interface of two subdomains of NVCP using the 39GYPP42 loop of UBXLOTU1 with the two prolines in cis- and trans-configurations, respectively. A mutagenesis study shows that this loop is not only critical for the interaction with VCP but also for its role in the ERAD pathway. Negative staining EM shows that one molecule of OTU1 binds to one VCP hexamer, and isothermal titration calorimetry suggests that the two proteins bind with a KD of 0.71 μm. Analytical size exclusion chromatography and isothermal titration calorimetry demonstrates that OTU1 can bind VCP in both the presence and absence of a heterodimer formed by ubiquitin fusion degradation protein 1 and nuclear localization protein 4.
Protein degradation is required for the elimination of faulty or unnecessary proteins as the accumulation or deposition of these proteins leads to various human diseases such as diabetes, inflammation, and neurodegenerative disorders (1). The endoplasmic reticulum-associated degradation (ERAD)3 pathway in particular serves to maintain endoplasmic reticulum homeostasis in the lumen or the membrane of the endoplasmic reticulum and belongs to the ubiquitin-proteasome pathway requiring two catalyzing processes, ubiquitination and deubiquitination (2). This pathway comprises three stages: the recognition of aberrant proteins, the retrotranslocation of the ubiquitin-modified substrates from the endoplasmic reticulum to the cytosol, and the degradation by the proteasome (3).
One of the many players in the ERAD pathway, valosin-containing protein (VCP; also known as p97 in mammals and CDC48 in yeast) is an abundant, highly conserved, and essential AAA+ ATPase. VCP governs critical steps such as retrotranslocation and transportation of substrates to the proteasome (4), and it recognizes its substrates mostly in the presence of the ubiquitin-fusion degradation protein 1 (UFD1)/nuclear localization protein 4 (NPL4) heterodimer (5). It comprises an N-terminal domain, two ATPase domains (D1 and D2), and a C-terminal tail. It forms a homohexameric ring structure with D1 and D2 stacked in a head to tail manner, and the N-terminal domains are enclosed by the D1 ring (7). In addition to the ERAD pathway, VCP is associated with a large number of independent cellular processes including transcription activation, mitosis, apoptosis, DNA repair, autophagy, endosomal sorting, and protein degradation at the outer mitochondrial membrane (see Refs. 8 and 9 and references therein). These functions are dependent on a multitude of regulatory binding partners of VCP that interact with the N-terminal domain and/or the C-terminal tail of VCP; these binding partners can be classified as either substrate-recruiting or substrate-processing proteins (10). Defects in VCP or its binding partners have been linked to several diseases (8, 9).
Recently, a number of ERAD-specific deubiquitinating enzymes have been reported to be physically associated with VCP (11, 12). Among these enzymes is ovarian tumor domain-containing protein 1 (OTU1). OTU1 is known as a deubiquitinating enzyme as well as a substrate-processing factor of VCP that is required for the retrotranslocation of the ERAD pathway (13). It also plays a role in controlling ubiquitinated or unubiquitinated substrates in the ERAD pathway (14). However, the exact mechanisms are not yet well understood. It harbors three domains: an N-terminal ubiquitin regulatory X-like (UBXL) domain, an OTU domain, and a C-terminal zinc finger domain of unknown function. It binds to the N-terminal domain of VCP (NVCP) through the UBXL domain and removes ubiquitin chains using the OTU domain (15). In particular, the UBXL domain of OTU1 (UBXLOTU1) consists of about 80 residues similar to the ubiquitin regulatory X (UBX) domain, which is one of the best characterized adaptor domains of VCP. UBX interacts with NVCP via a conserved arginine from strand β1 and the three residues (Phe-Pro-Arg) in a four-residue-long loop formed between strands β3 and β4 often referred to as the S3/S4 loop (16, 17). These residues, “R … FPR,” form the signature motif of UBX. However, UBXL has low sequence identity of less than 15% to UBX and furthermore lacks the signature motif of UBX. So far, the structure of the OTU domain from yeast complexed with ubiquitin has been produced (18), and recently it has been demonstrated that YOD1, the mammalian homolog of OTU1, has high specificity for Lys11-linked ubiquitin chains in a report that shows the complex structure of YOD1 with Lys11-linked diubiquitin (19). To understand how OTU1 interacts with VCP on a molecular level and to gain insight into how this interaction regulates the ERAD pathway, we carried out a structural study on NVCP complexed with UBXLOTU1 and performed negative staining electron microscopy (EM) and biochemical studies of the full-length proteins.
All human VCP and yeast OTU1 proteins were cloned into the expression vector pET28a (Novagen) that appends an N-terminal histidine tag and thrombin site to the target protein under T7 promoter control. Substitution of Gly-Tyr-Pro-Pro to Thr-Phe-Pro-Arg at positions 39–42 of the full-length OTU1 (residues 1–301) was introduced by a PCR method in clone OTU1 39TFPR42. For expression in mammalian cells, the constructs encoding pCS2-HA-OTU1 WT and pCS2-HA-OTU1 39TFPR42 were prepared. A ribophorin I mutant (RPN-IN299T) with the asparagine at position 299 replaced by threonine was cloned into the pEGFP-C1 vector (Clontech). For analytical size exclusion chromatography, the full-length UFD1 (residues 1–307) and NPL4 (residues 1–608) were cloned into the pET28a and pET22b vectors (Novagen), respectively, which are named pET28a-UFD1 and pET22b-NPL4. The glutathione S-transferase (GST) protein was fused to the N terminus of the full-length OTU1, and pGST-OTU1 was derived from pGEX-4T-1 (GE Healthcare). All proteins were expressed in Escherichia coli Rosetta (DE3). Cells were grown in LB medium at 37 °C, and protein expression was induced by 0.5 mm isopropyl β-d-1-thiogalactopyranoside at 18 °C overnight. Harvested cells were lysed in buffer A (20 mm Tris hydrochloride, pH 8.0, 100 mm sodium chloride, 0.5 mm tris(2-carboxyethyl)phosphine with 1 mm PMSF) and purified using nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography (GE Healthcare). Afterward, the histidine tag was proteolytically cleaved by addition of thrombin (Enzyme Research Laboratories), and protein was further purified using size exclusion chromatography on a HiLoad Superdex 75 or 200 26/60 column (GE Healthcare) in buffer A or buffer B (20 mm HEPES, pH 7.5, 150 mm sodium chloride). 2 mm ATP and 5 mm magnesium chloride were added to the full-length VCP (residues 1–806). A UFD1/NPL4 heterodimeric complex was formed by mixing pET28a-UFD1 and pET22b-NPL4. For the formation of complexes using the full-length proteins, VCP and either OTU1 (VCP/OTU1) or UFD1/NPL4 (VCP·UFD1/NPL4) were incubated at a molar ratio of 1:3 for 1 h at 4 °C, and then VCP/OTU1 and VCP·UFD1/NPL4 complexes were analyzed by size exclusion chromatography.
Purified complexes of NVCP and UBXLOTU1 were digested with chymotrypsin and trypsin (Sigma) at a protein-to-protease ratio of 1,000:1 in buffer A for 15 and 5 min, respectively, at 37 °C. The reactions were stopped by addition of 0.1 mm PMSF. Reaction products were resolved by 15% SDS-PAGE and visualized by subsequent Coomassie Brilliant Blue staining. The proteins were used at a concentration of 20 mg/ml for crystallization.
Crystals of crystal form I were obtained from a reservoir solution containing 14% PEG 4000, 100 mm sodium acetate, pH 5.0, and those of crystal form II were obtained from 20% PEG 2000, 10% tacsimate, pH 5.0 using the hanging drop vapor diffusion method by mixing with an equal volume of the protein solution. Crystals in a soaking solution containing 25% ethylene glycol as a cryoprotectant were flash cooled in liquid nitrogen.
Diffraction data of the two crystal forms were collected at 100 K on beamline 5C of Pohang Light Source, Republic of Korea. The data were processed and scaled using the program suite HKL2000 (20). Crystal form I in the asymmetric unit was located by molecular replacement using the program Phaser (21) with the free VCP structure (Ref. 22; Protein Data Bank code 1E32) as a search model. The structure of crystal form II was subsequently determined using the refined structure of crystal form I as an input model. Iterative model building was performed by manually fitting into the unoccupied densities with Coot (23), and refinement was carried out by CNS (24), PHENIX (25), and REFMAC5 (26). Water molecules were added with Coot, and structures of the two crystal forms were validated using PROCHECK (27). Coordinates of the crystal structures have been deposited in the Protein Data Bank under codes 4KDI and 4KDL. All figures were generated with PyMOL (28). The data collection and processing statistics for the two structures are summarized in Table 1.
VCP·UFD1/NPL4 complex was purified and concentrated to 0.05 mg/ml in buffer B. To improve its stability, the complex was chemically cross-linked with a final concentration of 0.1% EM grade glutaraldehyde (Sigma) for 10 min at 25 °C. The reaction was stopped by adding Tris, pH 8.0 to a final concentration of 25 mm.
ITC experiments were performed on an iTC200 calorimeter (GE Healthcare) at 25 °C. For the interaction between the full-length VCP and OTU1, OTU1 wild type (WT) or the mutant (39TFPR42) was placed in the syringe at 300 or 500 μm, respectively, and injected into the sample cell containing the full-length VCP at a concentration of 13 μm. The concentration of VCP/OTU1 and VCP·UFD1/NPL4 complexes was 0.01 μm in the sample cell, and UFD1/NPL4 and OTU1 were titrated to a concentration of 20 μm. Buffer B always matched between the syringe and the sample cell. The data were analyzed with the software Origin 7.0 using a one-site model.
HEK293T cells were seeded onto 12-well poly-l-lysine-coated plates at a density of 2.5 × 105 cells/well, and then transfection with expression plasmids was performed for 2 days using Effectene reagent (Qiagen). The total amount of DNA used in each well for transfection was 600 ng. Cells were treated with 10 μm MG132 for 2 h to prevent protein degradation and washed twice with complete medium. After addition of cycloheximide at a final concentration of 100 μg/ml, cells were harvested with cold Dulbecco's phosphate-buffered saline (PBS) and lysed, and cell extracts were analyzed by Western blotting with antibodies against pCS2-HA, pCS2-HA-OTU1 WT, pCS2-HA-OTU1 39TFPR42, and pEGFP-C1-RPN-IN299T.
The complex was prepared by co-purifying the full-length untagged VCP and OTU1 tagged with a hexahistidine tag at the C terminus. Freshly purified protein complex and 5-nm-diameter Ni-NTA Nanogold (Nanoprobes Inc.) were incubated together at a molar ratio of 1:10 for 10 min at 4 °C. The mixture was purified with a further chromatographic step using a Superdex 200 10/300 GL column (GE Healthcare) in buffer A. Excess Ni-NTA Nanogold was removed by centrifugation at 3,000 × g for 15 min using a 10-kDa-cutoff filter (Sartorius).
The complex of the full-length VCP and OTU1 (VCP/OTU1) labeled with gold was diluted to 20 μg/ml in buffer A. The samples were loaded onto a glow-discharged copper grid coated with a carbon film and stained with 2% uranyl acetate solution. Images were recorded on the film using a Tecnai F20 microscope operated at 200 kV equipped with a 2,048 × 2,048-pixel Gatan charge-coupled device camera. The gold-labeled proteins were selected from individual digital micrographs using a semiautomatic selection procedure, and the stoichiometry of OTU1 incorporation into VCP was determined by counting the number of gold particles. For unlabeled VCP/OTU1 complex, VCP/OTU1 complex was formed by incubating purified VCP with a 10-fold molar excess of OTU1 for 15 min at 4 °C. The complex was then further purified using size exclusion chromatography (GE Healthcare) in buffer containing 20 mm Tris hydrochloride, pH 8.0, 200 mm sodium chloride. 3 μl of VCP/OTU1 complex fraction at 0.1 μm was applied to a glow-discharged carbon-coated copper grid, and negatively stained with 2% methylamine vanadate, pH 8.0 (Nanoprobes Inc.). The specimen was examined under a Tecnai T120 microscope at 120 kV using a nominal magnification of 67,000× (1.64 Å/pixel; FEI Eindhoven, The Netherlands). Images were recorded with an FEI Eagle 4,000 × 4,000-pixel charge-coupled device camera using a defocus of 1.2 μm and an electron dose of about 30 e−/A2.
Because VCP has a preferred orientation on the EM grid, we could only obtain the hexagonal view of VCP or VCP/OTU1 complex. Hexagonal particles were automatically selected with EMAN2 boxer, and bad particles were manually excluded (29). A total of 2,839 particles were selected from 50 micrographs. Subsequent processing was performed using IMAGIC (30). To separate VCP/OTU1 complex from the heterogeneous data set containing either VCP or VCP/OTU1 complex, we executed comparative projection matching with two reference sets. One reference set was the forward projection of the crystal structure of VCP hexamer (Ref. 31; Protein Data Bank code 1R7R), and the other reference set was the forward projection of the model structure of VCP hexamer with one OTU1 molecule (Protein Data Bank codes 4KDI and 4KDL in the present work and 1R7R). These two separate data sets were respectively subjected to 10 rounds of reference-free alignment and multivariate statistical analysis followed by multivariate statistical analysis classification and averaging.
pET28a-VCP, pGST-OTU1, pET28a-UFD1, and pET22b-NPL4 were expressed and purified separately. Analytical size exclusion chromatography was carried out at 4 °C on a Superdex 200 10/300 GL column (GE Healthcare) in buffer A at a flow rate of 0.5 ml/min. Elution fractions were analyzed for complexes by SDS-PAGE and Coomassie Brilliant Blue staining.
To obtain atomic details on the interaction between VCP and OTU1, a number of complexes of VCP and OTU1 using various fragments containing the full -length were subjected to crystallization trials (see Fig. 1A). However, only the complex of NVCP and UBXLOTU1 yielded diffraction quality crystals in two conditions: one with residues 21–196 and the other with residues 1–187 of NVCP, referred to hereafter as crystal form I and crystal form II, respectively. Both crystal forms I and II crystallized in space group P212121 but with Z = 1 and Z = 2, respectively, and were refined at 1.81- and 1.86-Å resolution, respectively. First, the molecular replacement method using various structures such as UBX of Fas-associated factor 1 (FAF1) (Protein Data Bank code 3QX1; denoted as UBXFAF1 hereafter) and the NMR structure of UBXLOTU1 (Protein Data Bank code 2KZR) failed to give a solution. However, repeating the search using uncomplexed NVCP (residues 21–187) (Ref. 22; Protein Data Bank code 1E32) as a search model gave a solution, and the subsequent Fo − Fc difference electron density map revealed unambiguous density for bound OTU1. The structure of crystal form II was determined by the molecular replacement method using the structure of crystal form I. Residues that were not visible in the electron density map were left out in the final model; they include residues 1–20 of NVCP in all three molecules, 121–128 of NVCP in one molecule from crystal form II, 48–52 of UBXLOTU1 in crystal form I, and 49–55 in one molecule of UBXLOTU1 from crystal form II. The statistics on data collection and the refinements are given in Table 1.
The overall structure of the NVCP and UBXLOTU1 complex is shown in Fig. 1B. As seen in the figure, NVCP is composed of two subdomains (N1 and N2) connected by a seven-residue-long linker, and UBXLOTU1 binds at the interface between the N1 and N2 domains of VCP. The N1 domain (residues 23–104) consists of a double-Ψ β-barrel, and the N2 domain (residues 112–196) consists of a six-stranded β-clam with two helices on the two ends as described earlier (7). The structure of UBXLOTU1 (residues 1–73) reveals a β-grasp fold with a β-β-α-β-β-α-α-β secondary structural organization that resembles ubiquitin and ubiquitin-like proteins as well as UBX and the ubiquitin D domain (UBD) (18, 32). However, it differs significantly, especially in the conformation of the loops and the electrostatic surface, and these differences will be discussed later.
The three crystallographically independent molecules, one from crystal form I and two from crystal form II, are almost identical. When the molecule of crystal form I is taken as a reference, the two molecules in crystal form II show root mean square deviation (r.m.s.d.) values of 0.63 and 0.64 Å for NVCP and 0.45 and 0.50 Å for UBXLOTU1. When compared with NVCP used in the search model, the r.m.s.d. varies from 0.49 to 0.54 Å, suggesting that there are no structural changes upon binding of UBXLOTU1. In the case of NVCP, residues 33–34, 119–129, 155–159, and 182–187 show r.m.s.d. values greater than 3 Å, but they are typically located on the surface away from the binding interface with UBXLOTU1. These high r.m.s.d. values are due to differences in the crystal packing environment or flexibility of the molecule (data not shown). Residues 46–56 are ordered in only one of the three UBXLOTU1 molecules because there is a neighboring molecule stabilizing the structure by forming hydrogen bonds (data not shown). Conversely, the superposition of UBXLOTU1 with the earlier reported NMR structure (Protein Data Bank code 2KZR) gives an r.m.s.d. of 2.18 Å for 73 Cα atoms; this might be the reason for the failure to obtain a solution using the structure as a search model.
A DALI search using UBXLOTU1 as a search model gave UBXFAF1 (Protein Data Bank code 3QX1) as the closest match with a Z-score of 8.8, r.m.s.d. of 1.9 Å (68 Cα atoms), and sequence identity of 13%. FAF1 is a ubiquitin receptor containing multiple ubiquitin-related domains that is involved in various physiological functions such as apoptosis and NF-κB signaling. The second best match was ubiquitin (Ref. 33; Protein Data Bank code 1YIW) with a Z-score of 8.2, r.m.s.d. of 2.2 Å (63 Cα atoms), and sequence identity of 12%. Other ubiquitin-like proteins such as NEDD8 (Protein Data Bank code 3GZN), SUMO-1 (Protein Data Bank code 1A5R), SMT3 (Protein Data Bank code 3V62), and FUB1 (Protein Data Bank code 2L7R) showed Z-scores slightly less than 8, and they all had similar sequence identities as well. UBD of NPL4-like protein (Protein Data Bank code 1WF9) showed a Z-score of around 5. Although the function of NPL4-like protein is not yet clear, UBD of NPL4 (Ref. 34; Protein Data Bank code 2PJH; UBDNPL4) has been reported to assist VCP along with UFD1 in the ERAD pathway (5). Thus, it is included in the following comparison instead.
Fig. 2 shows a comparison of UBXLOTU1 with UBXFAF1 (Ref. 35; Protein Data Bank code 3QQ8) and UBDNPL4 along with the structure-based sequence alignment using the alignment program STRAP (6). The r.m.s.d. obtained in the superposition of UBXLOTU1 and UBXFAF1 (UBDNPL4) is 2.21 Å (3.97 Å) for 73 Cα atoms. Although these proteins, namely UBXLOTU1, UBXFAF1, and UBDNPL4, share structural homology at the Cα level, the electrostatic surfaces are quite different (Fig. 2A). The electrostatic surface of UBXLOTU1 is particularly distinct: one side shows a positively charged patch, whereas the opposite side shows a broad negatively charged patch. The positively charged patch is formed by Lys2, Lys4, Arg37, Arg44, and Lys69 from the various strands of the β-sheet, whereas the negatively charged patch is formed by residues such as Asp24, Glu27, Asp60, and Glu61 from α1 and η1 of UBXLOTU1 (data not shown).
Despite the fact that the overall structures are similar, a detailed comparison reveals differences in four particular regions, and these are highlighted in Fig. 2, A and C. The four regions are well defined in the electron density maps and are almost the same in all three structures of crystal forms I and II. First, the β1-β2 loop, highlighted the region 1 of Fig. 2A, is tilted more toward the body of the molecule by about 20° in UBXLOTU1. Second, the α1 helix of UBXLOTU1 is shorter than that of UBXFAF1 and UBDNPL4 by nearly one turn. This is due to deletion of four residues in UBXLOTU1. In addition, UBXLOTU1 has an additional α-helix formed by residues 53–56 with an insertion of five residues after β4 (region 4), and this region extends outward from the body of the molecule. In fact, it is about 11 Å away from the corresponding loop position of UBXFAF1.
The most intriguing difference is the β3-β4 loop conformation of UBXLOTU1 (region 3 in Fig. 2A). This loop corresponds to the S3/S4 loop, which is part of the signature motif of UBX. This loop is composed of Gly39-Tyr40-Pro41-Pro42, and all the residues are well defined in the electron density map as shown in Fig. 2B. The two prolines in this loop are found in the cis- and trans-configurations, respectively. The loop has a type VIb β-turn and is stabilized by intramolecular hydrogen bonds formed between Phe38 and Gln43 (N–O, 3.07 Å; O–N, 2.83 Å) as well as π-π stacking interactions between proline and the preceding tyrosine. This is quite similar to that of UBXFAF1; i.e. the corresponding Thr-Phe-Pro-Arg forms a VIb β-turn, and the proline adopts a cis-configuration. However, the proline of UBXFAF1 does not form similar π-π stacking interactions with the preceding phenylalanine, and the distances between the first and the fourth Cα atoms in the loop are 4.96 and 3.80 Å in UBXLOTU1 and UBXFAF1, respectively, resulting in different conformations (Fig. 2B). This may be due to the additional proline in UBXLOTU1. Conversely, the proline in the S3/S4 loop of UBXp47 was reported to be in trans-configuration (36). The corresponding region in UBDNPL4 is longer by six residues and forms a 310-helix instead of the loop as shown in Fig. 2B. Taken together, the structure of UBXLOTU1 exhibits distinctive features from UBXFAF1 or UBDNPL4, and these regions, especially regions 1 and 3, appear to serve as a platform for different interactions.
To understand how UBXLOTU1 differs from UBXFAF1 and UBDNPL4 in NVCP binding, the complex structure of NVCP and UBXLOTU1 (hereafter referred to as VCP·OTU1) was compared with those of NVCP and UBXFAF1 (Ref. 35; Protein Data Bank code 3QQ8; VCP·FAF1) and NVCP and UBDNPL4 (Ref. 34; Protein Data Bank code 2PJH; VCP·NPL4). The change in accessible surface area (ΔASA) upon complex formation is 590 Å2 for VCP·OTU1; this is much smaller than the corresponding values for VCP·FAF1 and VCP·NPL4, namely 810 and 1,050 Å2, respectively. The residues involved in the interaction with NVCP are positioned in three regions, namely β1-β2, β3-β4, and β5, as indicated in Fig. 2C. The three complex structures are superposed using only the structures of NVCP. Surprisingly, the three VCP-binding partners are at significantly different positions with respect to each other; i.e. they do not overlap at all (Fig. 3A). The most pronounced difference is in the position of the α1 helix. The relative positioning of the α1 helix from VCP-binding partners is somewhat different as mentioned above (Fig. 2A) but is significantly different when the complexes are superimposed; e.g. in the complex structures, the α1 helix of UBDNPL4 is more than 5 Å shifted from that of UBXFAF1, whereas that of UBXLOTU1 is somewhat twisted from the other two (Fig. 3A).
The molecular surface of NVCP is mapped by interaction of different binding partners (Fig. 3B). The binding region common to all three VCP-binding partners is colored in brown, the unique surface to UBXLOTU1 binding is colored in blue, and the region irrelevant to UBXLOTU1, i.e. relevant to only UBX and UBD, is colored in green. As seen in Fig. 3B, interaction in the three complexes is in the same general region, but there are clear differences. In particular, Ser37 and Arg144 of NVCP are only specific to UBXLOTU1 binding. They make non-bonded contacts with Tyr40 of UBXLOTU1, which makes a hydrogen bond to the side chain of Asp35 of NVCP (Fig. 3C). Interestingly, Tyr40 is part of the S3/S4 loop. On the other hand, the green region involves residues from the β6-β7 and β11-β12 loops of NVCP that comprise mainly hydrophobic residues. They interact with the residues from the β1-β2 loop of UBXFAF1 and UBDNPL4, i.e. region 1 (Fig. 2, A and C). However, the β1-β2 loop of UBXLOTU1 clearly adopts a different conformation from the other two (Fig. 2A) and does not make any interactions with NVCP, whereas there are numerous interactions involving both hydrophobic and polar residues between NVCP and UBXFAF1 (or UBDNPL4) (Fig. 3B).
The primary interaction is via the S3/S4 loop of UBXLOTU1 as in UBX (Fig. 3C). However, binding of UBXLOTU1 is somewhat different as the aromatic ring of Tyr40 of UBXLOTU1 is packed against the following cis-Pro41 on one side and by the side chains of Val38 and Leu72 of NVCP on the other side. Also, Tyr40 is positioned further between the N1 and N2 domains of NVCP, the hydroxyl oxygen of which makes hydrogen bonds with the carboxyl oxygens of Asp35 (2.76 and 3.29 Å) of NVCP (Fig. 3C). In addition, arginine in the signature motif of UBX is replaced by threonine in UBXLOTU1 and does not seem to be conserved. In the case of VCP·FAF1, the corresponding residue Phe619 of UBXFAF1 is packed by Val38 and Ile70 of NVCP on one side, and its carbonyl oxygen makes a hydrogen bond with Arg53 of NVCP (3.23 Å). The guanidinium group of Arg621 of UBXFAF1 makes hydrogen bonds to the carbonyl oxygens of Pro137 (3.31 Å) and Leu141 (2.52 Å) of NVCP. Alternatively, in UBDNPL4, a 310-helix is formed instead of the S3/S4 loop found in UBXFAF1 and UBXLOTU1, and this facilitates NVCP binding (Figs. 2C and and33C). Another major interaction region involves the C-terminal region of the binding partners (Fig. 2C). They interact with Phe52, Arg53, Gly54, and Asp55 of NVCP, which is colored in brown (Fig. 3B). As seen in the figure, this region is equally important. However, the degree of interactions varies per the three complex structures. For example, Arg53 of NVCP makes six hydrogen bonds in VCP·OTU1, whereas the same residue of NVCP forms only three and two hydrogen bonds in VCP·FAF1 and VCP·NPL4, respectively. Taken together, our data show that UBXLOTU1 binds in a similar position of NVCP as UBXFAF1 and UBDNPL4, but the interaction mode is specific unlike the others. This implies that OTU1 is a particular binding partner of VCP.
To confirm the importance of the S3/S4 loop of OTU1 in VCP binding based on the complex structure, the mutant was made by using the full-length OTU1. The S3/S4 loop of OTU1 was replaced with that of FAF1, which showed the closest resemblance as shown above; i.e. the mutant (OTU1 39TFPR42) has the S3/S4 loop sequence corresponding to Thr618-Phe-Pro-Arg621 of FAF1. The interactions between the full-length VCP and OTU1 were measured using ITC, and the results are shown in Fig. 4A. The binding affinity (KD) between VCP and OTU1 WT is 0.71 μm, whereas the corresponding value for OTU1 39TFPR42 could not be measured possibly due to weak binding. This could be understood based on the modeling study, which showed physical clashes with NVCP when the Gly-Tyr-Pro-Pro loop of OTU1 was replaced by Thr-Phe-Pro-Arg (data not shown). In addition, the two prolines in the S3/S4 loop of OTU1 may impose rigidity in the backbone conformation and prevent conformational changes from propagating to the rest of the protein. The substitution of proline with other residues such as glycine or arginine relieves the rigidity of the backbone and increases the flexibility of the loop. Therefore, forming a negative loop of UBXLOTU1 leads to low binding affinity between VCP and OTU1.
A substrate degradation assay was carried out to check whether the inability of OTU1 binding to VCP has any influence on the ERAD pathway using ribophorin I (RPN-IN299T) as an ERAD substrate. Degradation of the substrate was monitored in the presence of cycloheximide (37). OTU1 39TFPR42 showed relatively reduced degradation and accumulation of RPN-IN299T in comparison with OTU1 WT (Fig. 4B). This result shows that disruption of the association between VCP and OTU1 leads to the negative regulation of the ERAD pathway; thus, the S3/S4 loop of OTU1 is highly specific and is involved in a unique association with VCP.
It has been reported previously that abrogation of the deubiquitinating activity of OTU1 resulted in impairment of the degradation of misfolded RI332 as an ERAD substrate, suggesting that OTU1 positively regulates the ERAD pathway (13). Here, although UBXLOTU1 has no relation to the enzymatic activity of OTU1, disruption of the interaction between VCP and OTU1 led to defects in the degradation of ERAD substrates. Therefore, it seems that UBXLOTU1 is required for the ERAD pathway. Likewise, recent reports have suggested the importance of the association of VCP and UBX-containing proteins. For instance, the S3/S4 loop mutant of UBXFAF1 lost the ability to interact with VCP as well as with ubiquitinated substrate via its ubiquitin-associated domain, resulting in the accumulation of substrate (38). Another example is UBXD7, which separately binds both VCP and substrate HIF1α and recruits the substrate to VCP for degradation. However, its UBX-lacking mutant could not interact with both VCP and substrate, thus leading to a defect in the degradation of HIF1α (39). Therefore, the association of UBX-containing proteins with VCP is involved in their function; i.e. binding of OTU1 to VCP is needed to activate OTU1 in the ERAD pathway, demonstrating the importance of UBXLOTU1. Furthermore, the S3/S4 loop of OTU1 is conserved as residues Gly-Tyr/Phe-Pro-Pro are found in other OTU1 homologs and VCP complex-interacting protein of 135 kDa (VCIP135) which is another known UBXL-containing protein (Refs. 40 and 41 and data not shown). Therefore, not only VCIP135 but also OTU1 homologs are expected to bind VCP in the same manner shown in this study.
Because VCP exists as a homohexamer, it can possibly bind six different binding partners. Therefore, it is important to characterize the stoichiometry and interaction between the full-length VCP and its binding partners. To determine the stoichiometry of VCP and OTU1, an EM study was carried out. First, gold labeling was carried out using a hexahistidine tag. The tag was added to the C terminus of OTU1 because its N-terminal domain binds to VCP. Negative staining EM images of VCP alone and in complex with the gold-labeled OTU1 were taken. About 5,800 single particles were selected from individual micrographs, and the stoichiometry between VCP and OTU1 was analyzed statistically by counting the number of gold particles. Approximately 80% of the selected particles had one gold particle, and about 15% had two (Fig. 5A). The remaining 5% were particles with more than three gold particles (data not shown). We then took images of the complex of the full-length proteins (VCP/OTU1 complex) without gold labeling to avoid the restrictive space constraints due to the gold particle, which has a diameter of 5 nm. Three types of negative staining solutions were tried, but methylamine vanadate, pH 8.0 gave the best result as shown in Fig. 5B. A total of 2,452 particles of VCP/OTU1 complex were separated successfully from 2,839 particles of a mixture containing VCP alone and VCP/OTU1 complex with the class average. When uranyl formate was used, the ratio between the complex and VCP alone was almost 1,134 to 4,231 possibly due to the acidic nature of the solution, namely pH 5.5 (data not shown). Representative two-dimensional class averages for VCP alone and VCP/OTU1 complex are shown in Fig. 5B, which clearly shows that one OTU1 molecule binds to one vertex of homohexameric VCP. ITC measurement using the full-length VCP and OTU1 showed a binding stoichiometry of n = 1.07 ± 0.02 and a KD of 0.71 μm, suggesting that the stoichiometry of OTU1 to VCP is near 1:6. This is in good agreement with the EM data above. In addition, when UBXLOTU1 was removed (data not shown) or mutated (Fig. 4A), it did not show measurable binding to VCP. This confirmed again that UBXLOTU1 is necessary for binding to VCP.
The stoichiometries between VCP and various binding partners have been reported. In the case of UFD1/NPL4, EM studies clearly show that one UFD1/NPL4 binds one VCP hexamer (5, 42). On the contrary, for both p47 and UBXD1, three molecules of each bind to one VCP hexamer (43, 44). Interestingly, in the case of both FAF1 and UBXD7, one molecule binds one VCP hexamer only when UFD1/NPL4 is prebound to VCP (35, 38). The KD values of VCP and its binding partners are 1.7 μm for UFD1/NPL4 (35), 2.3 μm for FAF1 (35), 3.1 μm for UBXD7 (35), and 0.5 μm for p47 (43). These values are comparable with what we have seen with OTU1. It is worthwhile mentioning that in some cases bipartite binding has been reported. For example, VCP binding of UFD1/NPL4 utilizes the BS1 segment, which is a short hydrophobic stretch in the C-terminal domain of UFD1, in addition to the UBD of NPL4 as described above (45). Also, p47 uses a shp1-eyc-p47 domain, which is in the middle of p47, in addition to the C-terminal UBX domain. The shp1-eyc-p47 domain is reported to form a trimer and binds at the interface between the N and D1 domains of VCP. Besides the hierarchical binding of VCP-binding partners, some binding partners such as UFD1/NPL4, p47, and UBXD1 are reported to show mutually exclusive binding (46, 47), whereas others such as UFD1/NPL4 and UBX4 show simultaneous binding (48). This is understandable considering the existence of numerous binding partners and the diverse functions of VCP. Although the crystal structure of truncated proteins showed 1:1 binding between UBXLOTU1 and NVCP, the EM study using the full-length proteins clearly showed that one OTU1 binds to the hexameric VCP, suggesting only one OTU1 molecule is sufficient to perform its function; furthermore, it may function with other binding partners.
UFD1/NPL4 plays a crucial role in the ERAD pathway together with VCP, and some VCP-binding partners show hierarchical binding as well as mutually exclusive binding to VCP. We tested the binding pattern for VCP between OTU1 and UFD1/NPL4 using size exclusion chromatography and ITC, and the results are shown in Fig. 6. A GST-tagged OTU1 (GST-OTU1) was used to avoid possible confusion caused by molecular weights on size exclusion chromatography and SDS-PAGE. Two sets of experiments were carried out. First, GST-OTU1 was added to VCP alone, and then UFD1/NPL4 was added. Second, UFD1/NPL4 was added to VCP alone, and then OTU1 was added. As expected, both OTU1 and UFD1/NPL4 form a complex with VCP alone, but in addition, both form a quaternary complex, i.e. VCP·GST-OTU1·UFD1/NPL4 (Fig. 6A). We further determined the binding affinities and the stoichiometries between these using ITC, and the results are shown in Fig. 6B. First, UFD1/NPL4 and premade VCP/OTU1 binary complex showed an apparent KD of 3.50 μm and n = 1.05 ± 0.17, whereas OTU1 and premade VCP·UFD1/NPL4 ternary complex gave a KD of 0.41 μm and n = 1.09 ± 0.02. Both suggest that there is 1:1 binding. Interestingly, the KD values are similar to those obtained for VCP alone; e.g. the KD for UFD1/NPL4 binding to VCP was reported as 1.7 μm (35), and the KD for OTU1 binding to VCP was 0.71 μm (see Fig. 4A). This clearly shows that VCP can bind OTU1 in both the presence and absence of UFD1/NPL4 and vice versa.
In summary, we have dissected the interaction between VCP and OTU1. One OTU1 molecule binds to one VCP hexamer with similar binding affinity both in the presence and absence of UFD1/NPL4. The UBXLOTU1 binds at the interface between the two subdomains (N1 and N2) of VCP. In particular, the conformation of the Gly39-Tyr40-cis-Pro41-Pro42 loop of UBXLOTU1 is critical for NVCP binding. Although some of these properties of UBXL mimic UBX, the interaction is quite specific and has direct bearing on the proteasomal degradation of ERAD substrates. Based on these results, it is plausible that OTU1 can participate in the regulation of ubiquitinated substrates in both UFD1/NPL4-dependent and -independent manners. Our data provide insights into the interaction between the full-length VCP and OTU1 as well as the interaction between NVCP and UBXLOTU1 that could be applicable to other UBXL-containing proteins.
We thank the staff at 5C beamline, Pohang Accelerator Laboratory, Republic of Korea; Professor Y. K. Kang for interpreting the conformation of the loop; Dr. J. K. Park for VCP, UFD1, and NPL4 constructs; and S.-G. Lee for help with the EM study.
*This work was supported by Global Research Laboratory Program Grant 2013056409 of the Ministry of Science, Information and Communication Technology and Future Planning of Korea and an institutional grant from the Korea Institute of Science and Technology.
3The abbreviations used are: