|Home | About | Journals | Submit | Contact Us | Français|
Schizosaccharomyces pombe Rpn10 (SpRpn10) is a proteasomal ubiquitin (Ub) receptor located within the 19 S regulatory particle where it binds to subunits of both the base and lid subparticles. We have solved the structure of full-length SpRpn10 by determining the crystal structure of the von Willebrand factor type A domain and characterizing the full-length protein by NMR. We demonstrate that the single Ub-interacting motif (UIM) of SpRpn10 forms a 1:1 complex with Lys48-linked diUb, which it binds selectively over monoUb and Lys63-linked diUb. We further show that the SpRpn10 UIM binds to SpRpn12, a subunit of the lid subparticle, with an affinity comparable with Lys48-linked diUb. This is the first observation of a UIM binding other than a Ub fold and suggests that SpRpn12 could modulate the activity of SpRpn10 as a proteasomal Ub receptor.
The post-translational modification of proteins by covalent addition of polyubiquitin chains of diverse length and linkage type is an important mechanism by which cells regulate protein activity. Fidelity in the different pathways regulated by ubiquitylation depends in part on recognition of the character of the modification: Ub3 may be added to the target protein as a monomer or in chains. Ub chains are distinguished by the identity of the lysine residue within the so-called proximal unit that forms the isopeptide bond with the C-terminal carboxyl group of the neighboring (distal) Ub in the growing Ub chain (1). Proteins that are tagged with linear chains linked through Lys48 isopeptide bonds are targeted to the proteasome for degradation (2, 3).
The 26 S eukaryotic proteasome is composed of a 20 S catalytic core and a 19 S regulatory particle (RP). The 19 S RP recognizes the polyUb signal and subsequently brings about substrate deubiquitylation, unfolding, and translocation into the 20 S proteolytic channel (3, 4). The 19 S RP can be subdivided into base and lid subcomplexes (5). The base is composed of 10 subunits: six homologous ATPases that belong to the larger class of AAA (ATPases associated with a variety of cellular activities) ATPases, two scaffold subunits, Rpn1 and Rpn2, and the proteasomal Ub receptors Rpn10 and Rpn13. In contrast, the lid subcomplex contains nine structurally diverse subunits and shares homology with the COP9/signalosome particle in terms of subunit composition and sequence (3, 6).
Rpn10 (first identified as S5a in human cells) and Rpn13/Adrm1 have been identified as proteasomal subunits that act as Ub receptors (7, 8). Rpn13 is associated with Rpn2 and was identified through its ability to mediate recruitment of the deubiquitylating enzyme UCH37 to the proteasome (9). Subsequently, Rpn13 was shown to bind monoUb and polyUb chains via its N-terminal pleckstrin-like receptor for Ub domain (10). Rpn10 interacts with both base (Rpn1/Mts4) and lid (Rpn12/Mts3) subparticles, suggesting a location within the base at the 19 S base/lid interface (5, 11, 12). It is proposed to interact with both Rpn1 and Rpn12 via its N-terminal domain. This domain has weak homology to both the A-domain of the von Willebrand factor (VWA domain), a large multimeric plasma protein that mediates cellular adhesion, and the I-domain of integrins (13). C-terminal to the VWA domain of Rpn10 is an extended Ub binding sequence. Within this region S5a encodes two UIMs (UIM1 and UIM2) (14), whereas the yeast orthologs appear to possess only one, which more closely resembles UIM1 of S5a. As well as being an integral proteasomal subunit, a substantial fraction of Rpn10 can also be isolated from the cytosol (11, 15, 16).
An additional class of Ub receptors exemplified by members of the Rad23/Rhp23 and Dsk2/Dph1 families is not composed of integral proteasome subunits but associates reversibly with it to deliver ubiquitylated substrates for degradation (3, 17, 18). They encode an N-terminal Ub-like domain that binds to Rpn1/Mts4 (12, 19) and one or more ubiquitin-associated domains (UBAs) that confer polyUb binding activity (20, 21).
Recent biophysical studies have begun to elucidate the structural principles that determine the selectivity of different Ub binding domains for Ub chains of different linkage types (22). The human proteasomal Ub receptors HsRpn13 and S5a differ in their affinity and stoichiometry of monoUb and Lys48-linked diUb binding. The HsRpn13 pleckstrin-like receptor for Ub domain has a single Ub binding site that binds to Lys48-linked diUb with a 3-fold enhanced affinity over monoUb to generate in each case a 1:1 complex (8). S5a employs its two tandem UIMs to bind monoUb with 1:2 stoichiometry (14) or Lys48-linked diUb to form a 1:1 complex (23). Schizosaccharomyces pombe and Saccharomyces cerevisiae Rpn10 each encode only a single UIM, and therefore, their binding to Lys48-linked diUb must differ from S5a. In addition, it is not known whether a single UIM can bind selectively to Lys48-linked polyUb chains over other types of chains.
The mechanism by which S5a/Rpn10 binds to both polyUb chains and to the proteasome remains largely unknown. Here, we report a hybrid crystallographic and NMR-derived structure of full-length SpRpn10 and characterize its interaction with polyUb chains and the proteasomal lid subunit SpRpn12. We show that SpRpn10 binds selectively to Lys48-linked diUb with a 1:1 stoichiometry and with a greater affinity than to monoUb or Lys63-linked diUb. NMR experiments show that, in addition to the UIM motif, the linker region between the VWA and UIM sequences is implicated in the binding of Lys48-linked diUb and that both the proximal and distal Lys48-linked Ub moieties contact the single UIM although not simultaneously. We propose that the selectivity of SpRpn10 for Lys48-linked diUb arises from two effects: sterically preferred interchange of UIM binding between the proximal and distal Ubs of Lys48-linked chains, and additional contacts made only by Lys48-linked diUb to the linker sequence that connects the VWA domain and the UIM. We also demonstrate that both the VWA domain and the UIM of SpRpn10 participate in the interaction with SpRpn12, the first indication that UIMs may interact with proteins that do not adopt a Ub fold. The residues of the SpRpn10 UIM that mediate SpRpn12 binding include the conserved LALAL motif that forms the core of the Ub binding site, strongly suggesting that the binding of SpRpn12 and Ub to SpRpn10 are mutually exclusive. SpRpn12 binds to the UIM with an affinity comparable with that of Lys48-linked diUb and considerably higher than monoUb. As such, the interaction of SpRpn12 with the SpRpn10 UIM may “gate” or otherwise modulate access of substrates to the SpRpn10 Ub binding site within the proteasome.
An untagged SpRpn10 fragment(1–193) was expressed in Escherichia coli and purified by multiple chromatography steps. A more detailed description of the purification procedure is provided in the supplemental Experimental Procedures. SpRpn10 (full-length, residues 1–243), the VWA (1–195) domain and SpRpn12 fragment(1–250) were expressed from pGEX-6P-1 in E. coli BL21(DE3) cells and purified by sequential affinity chromatography, 3C protease cleavage and size-exclusion chromatography. Full-length SpRpn10 was grown in LB or 15NH4Cl-supplemented and 13C-d-glucose-supplemented M9 minimal medium as required for NMR titrations or assignments. SpRpn10(1–195) was grown in D2O minimal medium supplemented with 15NH4Cl and 2H13C-glucose for backbone assignments. Bovine monoUb was purchased from Sigma-Aldrich. Lys48-linked polyUb chains and Lys63-linked diUb were prepared as described (24, 25). Untagged Ub mutants K48C, K63C, and Asp77 were constructed using the QuikChange (Stratagene) mutagenesis method and ligation-independent In-Fusion (Clontech) cloning into the pGEX-6P-1 backbone (GE Healthcare). Mutants were purified and then Lys48- and Lys63-linked diUb prepared as described (26). All protein concentrations were determined using extinction coefficients derived from amino acid analysis.
All SPR experiments were conducted on a BIAcore 2000 instrument at 25 °C using HBS-EP buffer (BIAcore). SpRpn10 was diluted to 10 μg ml−1 in 10 mm sodium acetate, pH 5.0, and then immobilized at 800 response units on a CM-5 sensor chip by 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide/N-hydroxysuccinimide amine coupling. PolyUb chains were injected over the ligand surface in duplicate at a flow rate of 30 μl min−1 for 30s (supplemental Fig. S1A). For each analyte concentration, the sensorgram from a mock-immobilization flow cell was subtracted to account for nonspecific binding of the analyte to the CM-5 surface and bulk refractive index change. Steady-state SPR responses were estimated from the subtracted sensorgrams. Affinities were calculated by fitting steady-state responses to a binding isotherm with a Hill coefficient to model the binding of a polyvalent analyte to the ligand.
All samples were prepared in 25 mm sodium phosphate, pH 6.5, 50 mm sodium chloride. Concentrations were determined spectrophotometrically at 280 nm using the experimentally determined extinction coefficients of 3,425 m−1cm−1 (diUb) and 13,118 m−1cm−1 (SpRpn10). All experiments were performed on an iTC200 microcalorimeter (MicroCal) at 25 °C as detailed in the supplemental Experimental Procedures.
Samples of Lys48-linked diUb, Lys63-linked diUb, and SpRpn10 were prepared in 25 mm sodium phosphate, pH 6.5, 50 mm sodium chloride. Six concentration ratios of diUb to SpRpn10 ranging from 1:5 to 4:1 were prepared while keeping the total protein concentration constant at 300 μm. Sedimentation velocity experiments were performed in an Optima XL-I (Beckman) analytical ultracentrifuge at 20 °C and 40,000 rpm in an An-60 Ti rotor. A280 nm measurements were taken at 10-min intervals for a total of 500 min. c(s) analysis for each sample was performed, and the weight-average sedimentation coefficient of each sample was calculated using SEDFIT(27).
A native and a sulfur single wavelength anomalous dispersion dataset were, respectively, collected from the same crystal at the ESRF and in-house. The scaled data from both experiments were used to determine the sulfur sites and to calculate electron density maps with the SHELX suite (28). The initial model was built into the electron density using COOT (29), followed by alternating cycles of refinement in PHENIX.REFINE (30) and manual building in COOT. A more detailed description of the structure solution and refinement is provided in the supplemental Experimental Procedures. The final model comprises residues 2–147 and 151–193, 264 water molecules, and three sulfate ions. 137 (96.48%) residues lie in preferred and 5 (3.52%) in allowed Ramachandran regions. Data collection and refinement statistics are given in supplemental Table S1.
NMR titrations were carried out on home-built GE/Omega spectrometers operating at 11.7, 14.1, and 17.6 T. The concentration of 15N-labeled receptor protein was between 70 and 100 μm, and 1H/15N HSQC spectra were acquired at 0, 25, 50, 75, 100, and 132.5% of ligand, respectively. Titrations with monoUb and Lys63-linked diUb were also repeated with 556 and 140 μm concentrations of 15N-labeled SpRpn10, respectively, and increasing amounts of Ub up to 10-fold excess. Detailed spectral parameters can be found in the supplemental Experimental Procedures.
2H,13C,15N-SpRpn10 VWA(1–195) was concentrated to 0.4 mm in 20 mm sodium phosphate, pH 6.0, and 5% D2O was added. The sample was left to equilibrate for 5 days at room temperature to exchange labile protons. The temperature for all NMR experiments was set to 25 °C unless otherwise stated. HNCACB, HNCA, CBCA(CO)NH, and HN(CO)CA experiments were collected on a 18.7 T (800 MHz) Varian spectrometer. Most residues (~93%) from the VWA domain were assigned using this procedure, with the exception of residues 5, 41, 102–106, 130–133, and 145. To assign residues 192–243, full-length 15N-SpRpn10 was concentrated to 0.1 mm, and 5% D2O was added and CBCA(CO)NH and CBCANH spectra were acquired on a 500 MHz Bruker spectrometer equipped with a cryoprobe.
The structural model of SpRpn10 was prepared by merging the structure of the VWA domain (residues 2–193) with a model of the UIM domain generated with the program esypred3d (31). The modeling is based on sequence alignment with S5a and using the S5a structure (Protein Data Base entry 2KDE) as a template.
We first set out to characterize the interactions of full-length SpRpn10 with monoUb, Lys48- and Lys63-linked diUb, and Lys48-linked tri- and tetraUb chains. SPR studies (Fig. 1 and supplemental Fig. S1) showed that the binding of SpRpn10 to monoUb is weak, with a measured dissociation constant (Kd) of approximately 130 μm. This value is in agreement with the values for monoUb binding to the individual S5a UIM1 and UIM2 sequences (Kd approximately 350 and 73 μm, respectively) determined by NMR spectroscopy (14, 32). Binding to Lys48-linked diUb was ~5-fold stronger (Kd, approximately 35 μm). Comparison of the observed maximal binding responses (Rmax) of the analyte and ligand can be used to determine the stoichiometry of complex formation (33). From such an analysis, we find that the UIM of SpRpn10 binds a single Lys48-linked diUb molecule. A significant increase in affinity was observed for binding of Lys48-linked tri- and tetraUb chains. However, in both cases the Rmax values were not proportional to the number of Ub units in the chain, and the Hill coefficient was much lower than 1. These effects most probably arise from the contribution to binding of a subpopulation of immobilized SpRpn10 molecules that are sufficiently close to each other to engage simultaneously a single molecule of the Lys48-linked tri- or tetraUb analyte, with correspondingly increased apparent affinity. Taking this effect into account, the SPR results suggest a model in which Lys48-linked diUb is the unit of recognition for SpRpn10.
We next characterized the interactions between SpRpn10 and Ub by ITC and sedimentation velocity analytical ultracentrifugation (Fig. 1). ITC confirmed the initial SPR observations: SpRpn10 binds to both monoUb and Lys48-linked diUb with 1:1 stoichiometry and demonstrates a 10-fold preference for binding Lys48-linked diUb (Kd = 13.9 μm) over monoUb (Kd = 139 μm) (Fig. 1). The binding of Lys63-linked diUb to SpRpn10 showed 1:2 stoichiometry and could be modeled by Lys63-linked diUb having two equivalent binding sites for SpRpn10, each with an affinity similar to that of monoUb (134 μm). Analysis of the interaction of SpRpn10 with Lys48- and Lys63-linked diUb by sedimentation velocity analytical ultracentrifugation unambiguously confirmed that Lys48-linked diUb forms a 1:1 complex with SpRpn10 (Fig. 1, G and I). Analytical ultracentrifugation analysis of the Lys63-linked diUb·SpRpn10 complex yielded a broader profile (Fig. 1, H and I) that was consistent with the predominant species in solution being a 1:1 complex of relatively weakly binding proteins. It is likely that the 2:1 SpRpn10·Lys63-linked diUb species was not observed by sedimentation velocity analytical ultracentrifugation because the lifetime of this complex is too short for detection by this nonequilibrium method.
To characterize these complexes further, we used NMR spectroscopy and assigned the spectra of both the isolated SpRpn10 VWA domain (supplemental Fig. S2A) and the UIM in the context of the full-length protein (supplemental Fig. S2C). The more intense resonances of the UIM are not well dispersed but could be visualized separately by choosing a suitable threshold for contouring (supplemental Fig. S2C). All residues in the C-terminal UIM-containing sequence (residues 191–241) except Met218, Glu220, and Arg222 could be assigned using triple resonance methods.
MonoUb, Lys63-linked diUb, and Lys48-linked diUb were each titrated into SpRpn10 at increasing concentrations up to saturation (Fig. 2). All ligands caused significant chemical shift perturbations in the SpRpn10 spectrum, and in all titrations there was complete broadening for the signals associated with residues 207–208, 210–213, 215–217, 224–225, and 231 (Fig. 2A). Notably, residues within the LALAL motif (210–214; Fig. 2B) as well as immediately surrounding areas showed line-broadening upon the first addition of ligand. For any ligand, the largest chemical shift changes were found either adjacent to the areas of broadening or in residues 228–230. Titrations of monoUb and Lys63-linked diUb indicated fast chemical exchange on the NMR time scale, whereas titrations of Lys48-linked diUb demonstrated intermediate/slow exchange, consistent with tighter binding (Fig. 2C). The largest overall chemical shift changes accompanied addition of Lys48-linked diUb (Fig. 2A). Chemical shift changes were smallest in all titrations for residues 235–243 (<0.015 combined chemical shift change), indicating that these residues do not experience a significant change in environment upon polyUb binding.
Addition of monoUb and Lys63-linked diUb resulted in very similar shift patterns within the SpRpn10 C-terminal sequence (Fig. 2, C and D). For both ligands, residues 192–206 in the linker sequence experienced little perturbation (<0.05 parts/million), whereas residues 207–232 experienced either larger shifts (>0.05 parts/million) or peak broadening. For all resonances for which a peak could be observed in the bound state, both the direction and magnitude of the chemical shifts were virtually identical (Fig. 2C). For Lys48-linked diUb binding, residues 207–232 also experienced shifts in magnitude and direction very similar to those resulting from monoUb or Lys63-linked diUb binding, albeit with intermediate/slow exchange kinetics (Fig. 2C). In addition, Lys48-linked diUb caused significant chemical shift changes (>0.05 parts/million) and broadening in the SpRpn10 linker region (residues 192–198 and 201–205) indicative either of a unique interaction in this region or of a conformational change accompanying Lys48-linked diUb binding (Fig. 2, A and E).
To confirm that both Ub moieties in either Lys63- or Lys48-linked diUb bind to the SpRpn10 UIM, titrations were repeated with diUb in which either the proximal or distal Ub was 15N-labeled, and with 15N-labeled monoUb. Upon SpRpn10 addition, all Ub species undergo significant chemical shift perturbations on the fast/intermediate time scale. Furthermore, line broadening is observed in those residues located at the UIM binding site (centered around Leu8, Ile44, and Val70, Fig. 3). This suggests that both the proximal and distal Ub moieties of both diUbs bind to the UIM. The overall pattern of shifts observed on either monoUb or on the proximal and distal moieties of Lys63-linked diUb are very similar in terms of magnitude, distribution, and signal attenuation. This result suggests that SpRpn10 UIM engages monoUb, or the proximal or distal Ub of Lys63-linked diUb in similar fashion.
Analysis of SpRpn10-induced shifts on Lys48-linked diUb indicated that the proximal unit experiences slightly weaker chemical shift changes than does the distal unit. Furthermore, the distal Lys48-linked diUb experiences larger chemical shift changes in residues 10–12 and 35–39 compared with the other Ub moieties. Notably the C-terminal residues of the distal moiety (involved in the isopeptide bond) undergo severe broadening, which was not observed in the distal moiety of Lys63-linked diUb. However, because most of the spectral changes that accompany SpRpn10 binding are common to the other SpRpn10-Ub titrations, one can conclude that both distal and proximal Ubs of Lys48-linked diUb directly contact the SpRpn10 UIM.
Taken together, our results suggest a model in which SpRpn10 binds equivalently to monoUb or to each of the proximal and distal Ub moieties of Lys63-linked diUb to form either a 1:1 (monoUb·SpRpn10) or 1:2 (Lys63-linked diUb·SpRpn10) complex. Either the distal or proximal Ub of Lys48-linked diUb can also bind to the SpRpn10 UIM. However, in this case binding of one SpRpn10 to either Ub moiety precludes the binding of a second SpRpn10 molecule to the other Ub moiety.
SpRpn10 contains an N-terminal VWA domain connected via a short linker to a C-terminal sequence that encodes a single UIM (supplemental Fig. S4). Because attempts to crystallize the full-length protein were not successful, we independently determined the structure of the VWA domain by x-ray crystallography and characterized the UIM by NMR. We identified an untagged construct that encodes SpRpn10 residues 1–193 that crystallized readily (SpRpn10VWA) and solved its structure by sulfur single-wavelength anomalous dispersion at 1.3 Å (supplemental Table S1). The structure consists of a central six-stranded β-sheet of five parallel strands and one antiparallel strand (β3) sandwiched between two triplets of α-helices (Fig. 4A and supplemental Fig. S4C). The fold and topology are very similar to those of other VWA and I-domains despite their low sequence identities (supplemental Fig. S5).
Between β1 and α1 the Rpn10 VWA domain has a 5-amino acid insertion that is almost invariant among Rpn10 orthologs and is absent in most other VWA domains (Fig. 4 and supplemental Fig. S4). Notably, the insertion disrupts the metal ion-dependent adhesion site (MIDAS) that binds Mg2+ ions in a number of VWA domains (34). This part of the VWA fold is essential for maintaining the structural integrity of the 19 S proteasome and for conferring resistance to amino acid analogs (35, 36). Mutation of Asp11 at the end of β1 to an arginine results in the dissociation of Rpn3 and other lid subunits, and expression of mutant Rpn10 in which Asp11 is converted to any amino acid except glutamate fails to rescue the analog sensitivity of a Δrpn10 strain. Substitution by an arginine would not only prevent the formation of a hydrogen bond between the side chain of Asp11 and the backbone nitrogen of Gly115 (located at the end of β4), but might also disrupt the hydrogen bond between Ser13 and Ser116. Together, these changes could result in loss of the loop orientations necessary for interactions with the lid (Fig. 4B).
Analysis of the full-length SpRpn10 1H/15N HSQC spectrum shows that the VWA resonances are not perturbed by the presence of the UIM (supplemental Figs. S2 and S3), indicating that the two domains are not in contact with each other. The narrow distribution of UIM peaks in the hydrogen dimension indicates that this region lacks tertiary structure. This interpretation is supported by the fact that no NOE cross-peaks are observed in a three-dimensional 1H/15N NOESY-HSQC spectrum that would define contacts between the UIM and other parts of the protein. The sharpness of the UIM peaks result from faster, independent movements of the UIM compared with the VWA domain, so that its resonances are more characteristic of a peptide rather than a 27-kDa protein (supplemental Fig. 2C).
Secondary structure prediction algorithms indicate helical propensity in SpRpn10 UIM residues 208–230, consistent with CαCβ chemical shifts of the UIM obtained from our assignments that indicate a stretch of helical residues between 207 and 228. The structure of the S5a C-terminal tail (14) encodes a helix from residues 214–244 that includes the LALAL motif and corresponds to residues 209–228 in the S. pombe ortholog (Fig. 2). Our NMR analysis suggests that there are no other secondary structural elements within the SpRpn10 C-terminal sequence. Taken together, the crystallographic and NMR data suggest a structure for Rpn10 in which a compact VWA fold (residues 1–191) is connected via a long (approximately 16 amino acids), flexible linker to an extended α-helix that in solution does not interact with the VWA domain (Fig. 4C).
Rpn10 interacts with subunits of the 19 S regulatory particle, located in both the lid (Rpn9, Rpn12) and in the base (Rpt3, Rpn1) subcomplexes (36,–38). In fission yeast, deletion of Rpn10 is synthetically lethal with mutations in Rpn12, Rpn11, and Rpn1 (11). We investigated the interaction between SpRpn10 and SpRpn12 in greater detail by performing 1H/15N HSQC titrations exploiting the assigned HSQC spectra of the SpRpn10 VWA domain in isolation and the UIM sequence in the context of the full-length protein (Fig. 5). Upon addition of SpRpn12, the resonances of the SpRpn10 VWA domain show peak disappearance without shifting position, indicative of binding in the slow exchange regime (Fig. 5A). Unfortunately, neither addition of saturating amounts of SpRpn12, reacquisition of the spectra at elevated temperatures (up to 40 °C), nor the use of transverse relaxation-optimized spectroscopy could allow us to visualize the VWA signals in the bound state.
However, titration of SpRpn12 into 15N-labeled full-length SpRpn10 revealed chemical shift changes within the UIM domain, indicative of a specific interaction of SpRpn12 with this part of the protein (Fig. 5B). Residues on both sides of the LALAL motif showed the largest chemical shift changes in the UIM sequence, whereas the LALAL residues themselves were almost completely broadened upon the first addition of the ligand. A comparative analysis shows that the magnitudes and directions of the UIM chemical shift changes resulting from either SpRpn12 or monoUb titration are different, indicative of the UIM being in distinct environments in each of the bound states (Fig. 5, C and D). From this analysis, we conclude that the interaction between SpRpn12 and SpRpn10 is mediated unexpectedly by direct interactions between SpRpn12 and the SpRpn10 UIM.
To quantitate the interaction between SpRpn10 and SpRpn12, we measured the change in fluorescence anisotropy accompanying binding of unlabeled SpRpn10196–243 to either fluorescently labeled SpRpn12 or, for comparison, fluorescently labeled Lys48-linked diUb (method described in supplemental Experimental Procedures and supplemental Fig. S6). The determined affinities (18 ± 9 μm for SpRpn12 and 9 ± 2 μm for Lys48-linked diUb) demonstrate that the SpRpn10 UIM contributes significantly to the interaction between SpRpn12 and SpRpn10. The close agreement of the affinity of the interactions between Lys48-linked diUb and SpRpn10 when measured by ITC (13.9 μm) and fluorescence anisotropy (9 μm) provides a useful cross-validation of these different techniques.
Because titrations of SpRpn10 with SpRpn12 unexpectedly revealed that the binding sites for SpRpn12 and Ub within the SpRpn10 C-terminal sequence overlap, we next carried out competition experiments to determine whether Ub can displace SpRpn12 from SpRpn10. 1H/15N HSQC spectra of SpRpn10 were acquired in the SpRpn12-bound state (at 3:1 excess of SpRpn12), in the Lys48-linked diUb-bound state (at 1:1 equimolar concentrations), and at a ratio of 3:1:1 (SpRpn12·Lys48-linked diUb·SpRpn10) (Fig. 5E). Analysis of these data indicated that the spectrum of Lys48-linked diUb-bound SpRpn10 is identical to that acquired in the presence of a 3-fold molar excess of SpRpn12. The same result was obtained when the experiment was repeated with a 10-fold molar excess of monoUb. From these data, we conclude that SpRpn12 is released from the UIM in the presence of Lys48-linked diUb or excess monoUb.
We have demonstrated that the single UIM present in SpRpn10 binds with low micromolar affinity to Lys48-linked diUb and exhibits approximately 10-fold selectivity for Lys48-linked diUb over Lys63-linked diUb and monoUb. The ITC results demonstrate that SpRpn10 binds Lys48-linked diUb with a 1:1 stoichiometry, and our NMR results show that either the proximal or the distal Ub moiety of Lys48-linked diUb can bind to the SpRpn10 UIM (Fig. 3A). In addition, titration of Lys48-linked diUb produces significant chemical shift perturbations and line broadening for residues in the SpRpn10 linker sequence (Val192 to Phe198 and Gly201 to Asn205). A previous NMR study that compared the binding of monoUb and Lys48-linked di- and tetraUb to the ScRpn10 UIM (residues 204–268) concluded that all three Ub molecules contact essentially the same Rpn10 surface (39). Thus, S. cerevisiae and S. pombe appear to display small differences in the way their Rpn10 linker sequences engage different Ub species.
We generated a model for the Lys48-linked diUb·SpRpn10 complex using our SpRpn10 structure and that of a modeled open conformation of Lys48-linked diUb derived from a closed crystal structure (50) (Fig. 6, A and B). Binding of the proximal Ub moiety to the LALAL motif of SpRpn10 positions the distal moiety C-terminal to, and on the opposite side of the UIM helix (Fig. 6A). Binding of the distal moiety would bring the isopeptide bond and the proximal Ub into close proximity with the SpRpn10 linker and so could account for the Lys48 linkage-specific observed chemical shifts in this region. Even though this orientation might be favored, the NMR data support a model in which both alternative binding modes are represented in the bound population. Notably, unlike Lys63-linked diUb, the two UIM binding sites of Lys48-linked diUb face each other so that rebinding is favored. Taken together, the observed shifts in the SpRpn10 linker sequence and the relative positioning of the two Ub UIM binding sites could account for the increased affinity of SpRpn10 for Lys48-linked diUb (Fig. 6).
In contrast, our results demonstrate that both the proximal and distal Ub moieties of Lys63-linked diUb can independently bind to the SpRpn10 UIM employing a binding mode that is essentially indistinguishable from that of monoUb binding. Both monoUb and Lys63-linked diUb bind to 15N-labeled SpRpn10 in fast exchange on the NMR time scale yielding chemical shift perturbations that are similar in their magnitude and direction (Fig. 2). Similarly, observing 15N-proximal or 15N-distal Lys63-linked diUb, the effects of SpRpn10 binding are virtually identical (Fig. 3B) to those observed for 15N-monoUb (Fig. 3C). The results of the ITC experiments can be rationalized by a model in which the Lys63-linked Ub moieties bind independently to the SpRpn10 UIM with affinities equivalent to that of monoUb to yield a 2:1 complex. We have used our model of full-length SpRpn10 and the structure of Lys63-linked diUb (PDB code 2JF5) to generate a model for the (SpRpn10)2·Lys63-linked diUb complex (Fig. 6C). The disposition of the UIM binding sites on opposite sides of the Lys63-linked diUb molecule imposed by the linkage results in both Ub moieties being able to bind independently.
S5a has been shown to bind to Lys48-linked diUb by a mechanism in which both its UIMs simultaneously engage with a different Ub moiety (23). In the most populated structural form of the S5a·Lys48-linked diUb complex, the distal Ub binds to UIM1, and additional interactions are formed between the S5a linker and residues of the distal Ub moiety that are not observed in the S5a·monoUb structure (14). In our proposed model, SpRpn10 may also employ additional interactions with the linker sequence to bind with greater affinity to Lys48-linked diUb. Binding of Lys48-linked diUb to both SpRpn10 and S5a therefore extends the monoUb binding site and provides an explanation for the high degree of sequence conservation between Rpn10 homologs N-terminal to the LALAL motif.
Previous studies of the protein Rap80 have shown that two tandem UIMs can exhibit selectivity for Lys63-linked chains through a mechanism that is exquisitely dependent on the length and properties of the intervening linker region (40,–42). Although it has been demonstrated that the single UIM in ScRpn10 binds more tightly to Lys48-linked diUb than to monoUb (39), we believe this to be the first example where Lys48- versus Lys63-linkage selectivity has been described and explained for a single UIM. Other Ub binding domains have been found to be linkage-selective. For example, a subset of UBA domains employs a sandwich mechanism to bind selectively to Lys48-linked diUb (24, 43). However, other UBAs have a single Ub binding site and bind monoUb to form 1:1 complexes (44).
Two recent studies have suggested that polyUb chains linked through Lys6, Lys11, and Lys29 (45) or Lys63 (46) can target substrates to the proteasome. Our results suggest that receptor subunits such as Rpn10 can possess inherent linkage selectivity, suggesting that additional proteasomal receptors may be needed to recruit non-Lys48-linked polyUb chains. We cannot, however, exclude the possibility that Rpn10 acts as a receptor for other polyUb chain types. Although the selectivity for Lys48-linked diUb shown by SpRpn10 is substantial, it is modest compared with the approximately 100-fold selectivity reported for the Lys48-selective Mud1 UBA domain (24), and its affinity for Lys48-linked diUb is low compared with that of Rpn13, another putative proteasome polyUb receptor (8).
Electron microscopy studies of subclasses of 26 S proteasomes suggest that Rpn10 is located close to the mouth of the 19 S RP ATPase module (47). The VWA domain is a protein-protein interaction domain that is found in a variety of signaling pathways, and protein-binding sites have been mapped to various parts of the VWA domain surface (34). The extensive surface conservation of Rpn10 (supplemental Fig. S4) suggests multiple sites of protein interaction compatible with both its proposed location in the 19 S RP and its role as an extraproteasomal Ub receptor. In integrins the MIDAS motif VWA domain undergoes conformational changes upon ligand binding that are transduced to the integrin C-terminal domain so conveying allostery from one end of the domain to the other. Although the MIDAS is not conserved in SpRpn10, protein binding to the equivalent surface on the Rpn10 VWA domain either within the proteasome or in the cytoplasm might transduce an allosteric signal.
A recent report has identified four lysine residues in ScRpn10 (lysines 71, 84, 99, and 268) that undergo monoubiquitylation in response to stress (48). Lys84 was identified as the major site and its modification inhibits the ability of the ScRpn10 UIM to bind to substrates. ScRpn10 Lys84 is the sequence equivalent of SpRpn10 Asn84 that is located in the α2–α3 loop close to the pseudo-MIDAS sequence (Fig. 4) but is not conserved across species (supplemental Fig. S4). Indeed, no surface-exposed lysine residue that is conserved across species lies close in the three-dimensional structure to the site of regulatory modification in ScRpn10. Thus, it remains to be determined to what extent Rpn10 monoubiquitylation is a conserved mechanism regulating its activity across species.
Unexpectedly, we have shown that SpRpn12 interacts with both the UIM and the VWA domains of SpRpn10. The intermolecular masking of the Rpn10 UIM within the proteasome by interaction with SpRpn12 is reminiscent of the intramolecular masking of the hHR23a UBA domains by its Ub-like domain (49), binding of the ScRpn10 UIM to the Dsk2 Ub-like domain (39), and the recently described masking of the ScRpn10 UIM by VWA monoubiquitylation (48). Our results suggest that modulation of the accessibility of the Ub binding sites of Ub receptors to Ub might be a more widely employed mechanism that would offer an opportunity to regulate the processing of ubiquitylated substrates by the ubiquitin proteasome system.
We thank the staff at the ESRF (beamlines ID14-2) for providing excellent facilities and for assistance during data collection. We thank J. Boyd and N. Soffe for NMR facilities; A. Willis for protein analysis; I. Taylor for technical assistance; and C. Khoudian, J. Dean, I. Vakonakis, R. Gilbert, N. Solcan, and J. McDonnell for assistance and useful discussions. We thank D. Barford and D. Komander for constructs.
*This work was supported by the United Kingdom Medical Research Council, the Canadian Institutes of Health Research, and the Wellcome Trust.
The on-line version of this article (available at http://www.jbc.org) contains supplemental Experimental Procedures and additional references, Figs. S1–S6, and Table S1.
The atomic coordinates and structure factors (code 2X5N) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).
3The abbreviations used are: