Ufd1 Exhibits the AAA-ATPase Fold with Two Distinct Ubiquitin Interaction Sites

Structure, Vol. 13, 995–1005, July, 2005, ©2005 Elsevier Ltd All rights reserved. DOI 10.1016/j.str.2005.04.013

Ufd1 Exhibits the AAA-ATPase Fold with Two Distinct Ubiquitin Interaction Sites Sunghyouk Park,1,5 Rivka Isaacson,2,4 Hyoung Tae Kim,3 Pamela A. Silver,4,* and Gerhard Wagner1,* 1 Department of Biological Chemistry and Molecular Pharmacology Harvard Medical School 240 Longwood Avenue Boston, Massachusetts 02115 2 Centre for Structural Biology Imperial College London South Kensington London, SW72AZ United Kingdom 3 Department of Cell Biology Harvard Medical School 240 Longwood Avenue Boston, Massachusetts 02115 4 Department of Systems Biology Harvard Medical School and The Dana-Farber Cancer Institute 44 Binney Street Boston, Massachusetts 02115 5 Department of Biochemistry College of Medicine Inha University Shinheung-dong Chung-gu Incheon, 400-712 Korea

Summary Ufd1 mediates ubiquitin fusion degradation by association with Npl4 and Cdc48/p97. The Ufd1-ubiquitin interaction is essential for transfer of substrates to the proteasome. However, the mechanism and specificity of ubiquitin recognition by Ufd1 are poorly understood due to the lack of detailed structural information. Here, we present the solution structure of yeast Ufd1 N domain and show that it has two distinct binding sites for mono- and polyubiquitin. The structure exhibits striking similarities to the Cdc48/p97 N domain. It contains the double-psi ␤ barrel motif, which is thus identified as a ubiquitin binding domain. Significantly, Ufd1 shows higher affinity toward polyubiquitin than monoubiquitin, attributable to the utilization of separate binding sites with different affinities. Further studies revealed that the Ufd1-ubiquitin interaction involves hydrophobic contacts similar to those in well-characterized ubiquitin binding proteins. Our results provide a structural basis for a previously proposed synergistic binding of polyubiquitin by Cdc48/p97 and Ufd1.

*Correspondence: [email protected] (P.A.S.); gerhard_ [email protected] (G.W.)

Introduction Regulated protein degradation through the ubiquitinproteasome pathway has been shown to be involved in a number of biologically important processes, such as development, cell cycle progression, and cancer (Adams, 2004; Cook and Tyers, 2004; Dou et al., 2003; Zhang and Laiho, 2003). Previous studies of this important pathway provided insights into the enzymatic mechanisms of substrate ubiquitination and subsequent degradation by the proteasome (Adams, 2003; Imai et al., 2003; Passmore and Barford, 2004; Pickart, 2004). However, the molecular mechanisms responsible for the recognition of ubiquitinated proteins and processes leading to proteasome-mediated degradation are currently ill-defined. An important mediator of these intermediate pathways is the essential 40 kDa protein Ufd1. It was originally discovered in screens of Saccharomyces cerevisiae mutants with defects in the ubiquitin fusion degradation pathway (Johnson et al., 1995). Although Ufd1 is conserved throughout eukaryotes, its structure is not known, and it has no recognizable sequence homology with other proteins of known structure (Johnson et al., 1995). Functionally, Ufd1 is best known for its association with the nuclear transport protein Npl4 (Meyer et al., 2000). This heterodimer is a founding member of a growing family of adaptor proteins that direct an AAAATPase (ATPases associated with diverse cellular activities), Cdc48 (p97 in mammals), to specific cellular functions (see Figure 1A for the proposed association topology). The group of adaptors includes p47 and Vcip135, which collaborate with p97 and syntaxin 5 to promote Golgi membrane fusion (Uchiyama et al., 2002). In particular, p47 and the Ufd1-Npl4 complex bind p97 in a mutually exclusive manner (Meyer et al., 2000). However, the functions and binding modes of many of these adaptor proteins have yet to be identified. A series of experiments have placed the Ufd1-Npl4Cdc48/p97 (UNC) complex on the path between ubiquitination and degradation in two well-studied processes: endoplasmic reticulum-associated degradation (ERAD) and regulated ubiquitin-mediated processing (RUP) (Bays and Hampton, 2002; Jarosch et al., 2003). ERAD is a quality control pathway in which misfolded proteins in the endoplasmic reticulum (ER) are retrotranslocated to the cytosol and degraded by the proteasome. The ERAD-substrate proteins are transferred through the ER membrane and are polyubiquitinated on the cytoplasmic face of the ER membrane. UNC recognizes the polyubiquitinated proteins and then directs them to the proteasome for degradation. In contrast, RUP involves specific proteasome-mediated regulatory cleavage of the target proteins. In yeast, for example, membrane-tethered transcription factors Spt23 and Mga2 are cleaved off the membrane by a proteasomemediated process. Here, UNC is involved in the cleavage of the substrates and the later release of the

Structure 996

Figure 1. Structure, Sequence Alignment, and p97 Association of Ufd1 (A) Schematic model of the Ufd1-Npl4-p97 complex based on the 6-fold symmetry of the p97 AAA-ATPase. See text for the domain names. (B) Ensemble of backbone traces of the 15 lowest-energy conformers of Ufd1. Regular secondary structured regions are aligned and overlaid to generate the ensemble. (C) Ribbon diagram of the lowest-energy conformer of the ensemble showing the presence of the two separate domains. The Nn subdomain is in red, and the Nc subdomain is in blue. Secondary structural elements are numbered. The two psi loops (psi1 and psi2) in the double-psi β barrel are also labeled. The position of the V94 found in ufd1-1 is shown. (D) Structure-based sequence alignment of Ufd1, its orthologs, and the p97 N domain. The alignments within the Ufd1 orthologs were generated with the ClustalX program. The alignment of p97 to the rest of the sequences is based on the experimental structure of S. cerevisiae Ufd1 reported here and the mouse p97 N domain structure reported previously (Dreveny et al., 2004) and differs from a previously proposed alignment (Golbik et al., 1999). The conserved residues are in red and boxed in blue, with the strictly conserved residues in all of the sequences in inverse shading in red. The secondary structure cartoon (red) above the alignment is for the yeast Ufd1, and the one below (blue) is for p97. The residue numbering above the alignment is for the S. cerevisiae sequence used in this study. The gene bank accession numbers of the sequences used are: S. ser, gi1717964; H. sap, gi12053683; M. mus, gi2501439; C. ele, gi2501440; A. tha, gi42573225; S. pom, gi3123677; p97, gi14488635.

cleaved transcription factors. UNC has other known roles in nuclear envelope assembly, the degradation of IκB, and the disassembly of the mitotic spindle (Cao et al., 2003; Hetzer et al., 2001; Hu, 2003). However, less mechanistic information is available on UNC’s functions in these processes. A lack of atomic level information on the Ufd1-Npl4 complex and its individual components has hampered our understanding of its diverse functional roles and its interaction with p97. Recently, the structure of the complex between p97 and one of its adaptor proteins,

p47, was reported (Dreveny et al., 2004; Yuan et al., 2004). The structure provided insights into the adaptor-p97 interaction, especially regarding how p47 targets p97 rather than NSF, another closely related AAA-ATPase. Additionally, a recent report addressed a possible binding mechanism between p97 and the adaptor proteins, Ufd1-Npl4 and p47 (Bruderer et al., 2004). However, no experimental structure is available for Ufd1, although sequence analyses suggested that it contains a socalled double-psi β barrel motif (Golbik et al., 1999).

Structural and Binding Studies on Ufd1 997

Studies with deletion mutants have shown that the N-terminal portion of Ufd1 contains a ubiquitin binding site, whereas the C-terminal region is important for binding Npl4 and Cdc48 (Ye et al., 2003). Here, we report the solution structure of the N-terminal fragment of Ufd1 containing residues 1–208 and its modes of ubiquitin binding. The structure of Ufd1 (1–208) bears an unexpected and striking resemblance to the N domain of p97. Ubiquitin binding experiments revealed that Ufd1 has distinct binding sites and affinities for mono- and polyubiquitin. These results shed light on the ubiquitin recognition mechanism of UNC and the reported synergistic binding of polyubiquitin by the N domains of Ufd1 and Cdc48/p97.

Table 1. Structural Statistics of the Calculated Structures Number of distance constraints All Intra or sequential Medium Long Hydrogen bond Number of dihedral angle restraints (backbone f and ψ) Rmsd from ideal covalent geometry Bonds (Å) Angles (°) Improper torsion angles (°)

The Structure of the Ufd1 N-Terminal Fragment Reveals Homology to N Domains of AAA-ATPases The structure of the Ufd1 N-terminal fragment was determined by following established multiple resonance NMR protocols (Figure 1B). More than 97% of the backbone resonances and most of the side chain resonances were assigned. Structural constraints were obtained from an array of isotope-edited NOESY spectra on uniformly and selectively labeled protein samples. The structures were calculated by using a variety of NOE distance constraints and TALOS-derived angle constraints. The final 15 lowest-energy structures had an average backbone rmsd of 0.66 Å for the regular secondary structure elements. No distance constraint violations over 0.3 Å and no angle violations larger than 3° were present in the calculated structures. The quality

0.00205 ± 0.00013 0.3521 ± 0.0088 0.2523 ± 0.0170

Quality check by PROCHECK NMR (%) Residues Residues Residues Residues

Results and Discussion Identification of the N and C Domains of Ufd1 In order to target our structural studies, we analyzed the amino acid sequence of Ufd1 and consulted previous functional studies to search for domains and their boundaries. Prediction of Ufd1 structural features has been difficult since it bears little sequence similarity to other proteins of known structure. However, the N-terminal segment was predicted to contain a so-called double-psi β barrel (Golbik et al., 1999). Functional studies with deletion mutants of Ufd1 revealed that the N-terminal UT3 portion (residues 1–211) contains a binding site for ubiquitin, whereas the C-terminal UT6 fragment (residues 215–307) encompasses the binding sites for both Npl4 and Cdc48 (Hetzer et al., 2001; Ye et al., 2003). Based on these results, we investigated the feasibility of obtaining structural information for Ufd1 by running 1D NMR spectra of the full-length protein as well as N- and C-terminal fragments (representing residues 1–208 and 209–361, respectively). The C-terminal portion of Ufd1 appears largely unstructured with a narrow resonance dispersion, whereas the N-terminal portion showed a well-dispersed spectrum characteristic of a folded domain, suitable for NMR studies (data not shown). Unfortunately, the full-length protein was ill behaved and not suitable for further structural studies. Therefore, we focused our structural and ubiquitin binding studies on the N-terminal fragment of Ufd1.

1652 929 170 492 61 250

in in in in

most favorable regions additional allowed regions generously allowed regions disallowed regions

93.8 5.1 0.0 1.1

Backbone rmsd against the mean structure (Å) Nn subdomaina (secondary structure region, all) Nc subdomainb (secondary structure region, all) All (secondary structure region, all) a b

0.592, 0.651 0.375, 0.598 0.661, 0.782

Residues 18–118. Residues 120–198.

check of the structures with PROCHECK NMR software showed good statistics in the backbone phi-psi angle distribution. The statistical summary of the calculation is listed in Table 1. The Ufd1 N-terminal fragment is composed of two readily identifiable subdomains designated as Nn and Nc subdomains (Figures 1B and 1C). The Nn subdomain adapts a double-psi β barrel fold (Castillo et al., 1999), and the Nc subdomain has a mixed α/β roll structure. Surprisingly, the overall structure of our Ufd1 fragment is very similar to the previously reported structures of the N domains of p97 and other homologous AAA-ATPases (Coles et al., 1999; Yu et al., 1999; Zhang et al., 2000). Therefore, from here forward, we call the N-terminal fragment of Ufd1 the N domain. To visualize this similarity, we overlaid the structures of the N domains of Ufd1 and p97 (see Figure 2). The backbone rmsd for the regular secondary structural parts are 1.838 Å (entire N domain), 1.707 Å (Nn subdomain), and 1.283 Å (Nc subdomain), indicating that the two proteins have similar overall domain folds. There was a prediction that Ufd1 Nn contains a double-psi β barrel (Golbik et al., 1999). This motif was indeed found in our structure, however at a different sequence location shifted by approximately 70 residues (see below). Despite the observation of the double-psi β barrel, the similarity of the entire N domains of Ufd1 and p97 is quite unexpected, because no structural similarity between an adaptor protein and an AAA-ATPase has been reported. To our knowledge, this structural similarity, combined with various biochemical data, provides new insights into the mechanism of UNC-ubiquitin binding (see below). The observed structural homology led us to revisit

Structure 998

Figure 2. Overlay of N Domains of p97 and Ufd1 The structures of N domains from p97 (PDB code 1S3S) and Ufd1, reported here, were overlaid in a ribbon diagram by using the matching residues in regular secondary structural regions. The matching residues are as shown in the alignment in Figure 1D. The p97 structure is in blue, and the Ufd1 structure is in red.

the previously reported sequence alignment (Figure 1D). Although Ufd1 does not have any readily detectable sequence similarity with other known proteins, a previous report suggested that the central domain of Ufd1 (residues 82–165) contains a double-psi β barrel domain (Golbik et al., 1999). This motif was indeed found, but was located at residues 13–119, within the Nn subdomain. The predicted central location would encompass the interdomain linker region between the Nn and Nc subdomains, which is inconsistent with the experimental structure. The revised structure-based sequence alignment presented here (Figure 1D) is used as the basis for the comparison of Ufd1 and p97 N domains (see below). Specificity for p47 Binding of p97 over Ufd1 The N domains of Ufd1 and p97 have very similar structures (Figure 2), but only that of p97 binds p47 (UBX domain of p47) (Dreveny et al., 2004; R.I., unpublished data). The crystal structure of the p97/p47 complex (Dreveny et al., 2004) reveals that p47 binds at the interface between the Nn and Nc subdomains of p97. Analysis of the superimposed structures of p97 and Ufd1 reveals that the crucial surface residues are not conserved in Ufd1 (Figure 3). Key elements of the interface between p97 and p47 include F343 in the S3–S4 loop of p47 and D35, S37, V38, and I70 in the N domain of p97. The latter residues of p97 are replaced with F44, G46, K47, and G83 in Ufd1 (Figure 3B). In the Ufd1 structure, one of the key residues, F44 (corresponding to D35 in p97), has its side chain pointing away from the Nn/Nc subdomain interface, rendering equivalent contacts with the p47 S3–S4 loop sterically unfeasible. In addition, the other residues in Ufd1 (G46, G83, and K47 along with the intervening G45) lack the hydrophobic side chains that are crucial for the p47 and p97 interaction. Another essential interaction in the p97-p47 complex is the hydrogen bond between the side chain of N345 in the S3–S4 loop of p47 and the backbone carbonyl oxygen of E141 in p97. The E141 does not have a corresponding residue in Ufd1 because it does not align well with Ufd1 residues (Figures 1D and 3C). Nevertheless, it is easy to see that an equivalent hydrogen bond is highly unlikely in the Ufd1-p47 interaction, because the corresponding regions are all involved in canonical hydrogen bonds in helix α5 of Ufd1 (Figure 3C). The binding between Ufd1 and p47 would

require the breakage of the helical hydrogen bonds, which would lead to the disruption of the helix. It was also noted that the p47 residues involved in the p97 interaction are not conserved in other AAAATPases that do not bind p47, such as NSF and VAT. However, those residues are well conserved in p97 orthologs from human to yeast. Therefore, it is likely that the N domains evolved into both adaptor proteins (Ufd1) and AAA-ATPases with different specific recognition motifs. Ufd1 Binds Mono- and Polyubiquitin at Distinct and Nonoverlapping Sites Most, if not all, of the Ufd1-Npl4-mediated p97 functions depend on interactions with some forms of ubiquitin species. In addition, previous studies have suggested that both p97 and Ufd1 have individual binding motifs for polyubiquitin, and that they bind polyubiquitin in a synergistic manner (Dai and Li, 2001; Rape et al., 2001; Ye et al., 2003). For both Ufd1 and p97, the N domains have been shown to be responsible for the ubiquitin binding (Dai and Li, 2001; Ye et al., 2003). However, there are no obvious sequence homologies between Ufd1, p97, and other known ubiquitin binding domains that could provide details about the ubiquitin binding sites. It is now well established that monoubiquitination and polyubiquitination represent distinct cellular signaling pathways (Haglund et al., 2003; Meyer et al., 2002; Pickart, 2000). Therefore, determining the underlying differences between mono- and polyubiquitin binding by p97 or Ufd1 should help us understand the function of the UNC complex. To address the mechanism of Ufd1-ubiquitin interaction, we carried out NMR binding experiments on the Ufd1 N domain with mono- and polyubiquitin. We observed chemical shift changes of 1H-15N crosspeaks upon addition of mono- or polyubiquitin (Figures 4A and 4C). When the affected residues are identified in the structure, they are localized within two distinct and nonoverlapping surface regions (Figure 4B). Polyubiquitin binding residues are located in the regions formed by the psi-2 loop, helix 2, sheet 3, sheet 6, and the α2– β4 loop, whereas the monoubiquitin binding site involves residues in sheet 2, psi-1 loop, and helix 3. Therefore, our results confirm previous reports that Ufd1 N domain binds polyubiquitin and show that the binding sites are located in the N-terminal double-psi β

Structural and Binding Studies on Ufd1 999

Figure 3. p47 Interaction with N Domains of Ufd1 and p97 (A) Schematic diagrams of possible interactions between p47 and N domains of Ufd1 and p97. “X” represents no interaction, and “O” represents real interaction. (B) The residues in the Nn subdomain of p97 involved in the hydrophobic interactions with F343 of p47 are shown in blue, and the corresponding residues in Ufd1 are shown in red. For G46 and G83 in Ufd1 lacking the side chain heavy atoms, bonds between the Hα-Cα are shown. (C) The residue (E141) in the Nc subdomain of p97 involved in the hydrogen bond with N345 in p47 is shown in blue. The hydrogen bond is indicated with a dashed line. Residues in the corresponding region in Ufd1 are drawn in red. In (B) and (C), the p97 is in blue and the Ufd1 is in red.

barrel domain (Nn subdomain) (Figure 4B). These data identify the double-psi β barrel as a ubiquitin binding domain. Ufd1 Has Higher Affinity for Polyubiquitin than Monoubiquitin The binding experiment reveals that Ufd1 has a higher affinity for polyubiquitin than monoubiquitin (Figures 4A and 4C). Figure 4C shows representative peak shifts at mono- and polyubiquitin binding sites after adding the corresponding ubiquitin species. When the same mass amount of total ubiquitin was used, polyubiquitin caused significant changes in peak positions in the polyubiquitin binding sites (shown for I88; blue peaks versus yellow peaks in the right panel), whereas monoubiquitin did not elicit any appreciable changes (black peaks versus yellow peaks in both panels) (Figure 4C). Obviously, the average concentration of the polyubiquitin is much less than that of the monoubiquitin sample due to the larger molecular weight (see the inset of Fig-

ure 4A). These more pronounced changes in the peak positions in the Ufd1-polyubiquitin mixture clearly indicate Ufd1’s higher affinity for polyubiquitin than for monoubiquitin. Nevertheless, when a higher concentration of monoubiquitin was used, sizable changes were observed as well (shown for I37; red peaks versus yellow peaks in the left panel; see also Figure 4A). Significantly again, the changes were localized only at the monoubiquitin binding site. These data suggest that monoubiquitin also binds Ufd1 when present at high concentration. Due to the heterogeneous nature and possible multiple binding sites on polyubiquitin, we could not estimate the Kd values for polyubiquitin. The Kd for the Ufd1-monoubiquitin interaction was estimated to be in the 1–2 mM range. Although this affinity seems low, it is comparable to the affinities of other well-known ubiquitin binding domains lying within the several hundred micromolar Kd range (Hicke and Dunn, 2003). In an extreme case, a Kd value of 2.4 mM was reported in a

Structure 1000

Figure 4. Different Binding Sites and Affinities of the Nn Subdomain of Ufd1 to Mono- versus Polyubiquitin (A) Backbone amide chemical shift changes in 1H-15N HSQC spectra of Ufd1 upon adding mono- or polyubiquitin. Black squares, Ufd1 and polyubiquitin binding; red crosses, Ufd1 and monoubiquitin binding. The values of each point represent the differences in the peak chemical shift values from the spectra with either mono- or polyubiquitin against those from the control spectrum (Ufd1 only). The combined chemical shift change values were calculated by using the equation in the experimental section. The concentration of monoubiquitin was 4 mM (high). 800 ␮M concentration (low) did not produce any significant changes (not shown here; see Figure 4C). For polyubiquitin, the same mass amount of total ubiquitin was used as for the experiment with low concentration of monoubiquitin (800 ␮M). Therefore, in comparison, the total mass amount of monoubiquitin is five times larger than the concentration of ubiquitin domains in polyubiquitin. The horizontal dotted line indicates the cutoff values used to identify the binding sites. The inset gel picture shows the composition of the ubiquitin samples analyzed by SDS-PAGE; molecular weight marker (lane 1), the monoubiquitin (lane 2), and the polyubiquitin sample (lane 3). The asterisk in lane 2 indicates the band for E2-25K. (B) Binding site mapping onto the calculated structure of Ufd1. The binding site for monoubiquitin (MonoUb) is in red, and that for polyubiquitin (PolyUb) is in blue. The beginning and end residues of the major binding regions for PolyUb (left) and for MonoUb (right) are numbered in the corresponding color. (C) Representative overlaid HSQC spectra showing the differential binding affinities and sites. Four spectra are overlaid in each panel: yellow, Ufd1 alone; black, with monoubiquitin (low, 800 ␮M); red, with monoubiquitin (high, 4mM); blue with polyubiquitin (low, 800 ␮M of total monoubiquitin concentration). I37 (left panel in red) represents the site unique to monoubiquitin binding, and I88 (right panel in blue) represents the site unique to polyubiquitin binding. The virtually indistinguishable peak positions of the yellow and black peaks (in all two panels) compared with the significant changes between the yellow and blue peaks (right panel) show the differential binding affinities for monoversus polyubiquitin. Note that the total mass amount of ubiquitin in the blue and the black spectra is the same.

Structural and Binding Studies on Ufd1 1001

possible intramolecular interaction between the UbL domain, a close relative of ubiquitin, and the UBA1 domain (Ryu et al., 2003). Intriguingly, p97, whose N domain is very similar to that of Ufd1, has also been shown to have different affinities for ubiquitin species depending on the chain lengths (Dai and Li, 2001). However, whether p97 binds monoubiquitin remains controversial (Dai and Li, 2001; Rape et al., 2001). Based on the observed Ufd1-monoubiquitin binding and the similar structure of Ufd1 and the p97 N domain, it is very likely that p97 also binds monoubiquitin, although with weaker affinity than polyubiquitin. It is well established that Ufd1 is involved in protein degradation via the polyubiquitin pathway. Now, we show that Ufd1 can also bind monoubiquitin by using a different recognition motif from that for polyubiquitin. Monoubiquitination signaling has been implicated in receptor endocytosis or DNA repair (Haglund et al., 2003; Pickart, 2000). Thus, Ufd1 might be involved in these processes through monoubiquitin binding. Although the modulation of the p97 activities with different adaptors seems to provide some explanation for the myriad of unrelated activities of p97, there are still mysteries about the exact mechanism of the diversity of the p97 functions. Ufd1’s binding to both mono- and polyubiquitin could suggest at least partial answers to these questions. It may be speculated that further modulation of the adaptor-p97 complex depending on mono- or polyubiquitin binding would provide another layer of diversity to the functions of the complex. The Degradation-Impaired Phenotype of the ufd1-1 Strain Our suggestion of the double-psi β barrel as a ubiquitin binding domain is also consistent with the phenotypic changes of a single point mutant yeast strain. Johnson et al. (1995) reported that a yeast strain with the V94D missense mutation in UFD1 (ufd1-1) has impaired degradation of the normally short-lived UbV76-V-B-gal. Now, this phenotype can be linked to the failure in ubiquitin-mediated protein degradation in the Ufd1 pathway. Our structure shows that the mutated V94 is at the very hydrophobic core of the double-psi β barrel, forming numerous hydrophobic contacts with other core residues (see Figure 1C for the position of V94). Therefore, it is likely that the introduction of a charged residue (Asp) at this hydrophobic core leads to destabilization and/or misfolding of the β barrel domain fold, which would impair ubiquitin binding and degradation of substrate proteins. Ufd1 Nn Subdomain Recognizes a Well-Conserved Binding Motif in Monoubiquitin After defining the ubiquitin binding sites on Ufd1, we pursued to identify the Ufd1 binding sites on ubiquitin. We recorded 1H-15N-correlated NMR spectra of 15Nlabeled ubiquitin before and after addition of unlabeled Ufd1. The changes of ubiquitin resonance positions upon addition of Ufd1 are shown in Figure 5A. Similar to the situation in Ufd1, the chemical shift changes are localized to a contiguous region of the ubiquitin surface (Figure 5B). Significantly, the residues

affected by Ufd1 binding are either in or very close to the hydrophobic patch on the ubiquitin surface involving L8, I44, and V70 (Figures 5A and 5B). It is wellestablished that this hydrophobic region is involved in interactions between monoubiquitin and other ubiquitin binding domains, such as UBA, CUE, and NZF domains (Alam et al., 2004; Cook et al., 1994; Kang et al., 2003; Yuan et al., 2004). The monoubiquitin binding sites on the Ufd1 side are also mostly hydrophobic, involving residues C27, Y28, I30, A31, M32, I37, K39, W99, M100, M101, and G109 (residues above threshold in Figure 4A). Therefore, the mapping data suggest that the binding between monoubiquitin and Ufd1 involves specific hydrophobic interactions, similar to those in the above well-studied cases. Ufd1-Ubiquitin Interaction: The Dual Binding Site Mechanism Our Ufd1 structure and ubiquitin binding results suggest a new mechanism for ubiquitin recognition. A ubiquitin binding protein may use a single domain that contains two different sites for mono- and polyubiquitin with different affinities. The different binding affinity between mono- and polyubiquitin for a single ubiquitin binding domain has actually been demonstrated in other systems. For example, the UBA domain binds mono- and polyubiquitin with affinities that differ by two orders of magnitude (Raasi et al., 2004). However, the reason for the enhanced affinity for polyubiquitin over monoubiquitin is still poorly understood. As pointed out previously, a simple explanation based on an increased number of binding sites did not provide a satisfactory answer (Pickart, 2000). Instead, new interaction surfaces formed in polyubiquitin species are likely to be the underlying mechanism. Particularly, polyubiquitin has one less positive charge around the K-48 than monoubiquitin due to the formation of the isopeptide bond. Although we currently don’t know the conformation of the isopeptide bond in polyubiquitin, it may be speculated that the electrostatic difference between monoubiquitin and polyubiquitin has a role in the observed differential binding of the ubiquitin species. As the solution and crystal structures revealed, polyubiquitin (here, di- or tetraubiquitin) can exist in different conformations (Cook et al., 1992, 1994; Phillips et al., 2001; Varadan et al., 2004). Therefore, the recognition motifs that the various ubiquitin species present to the binding partner can be different depending on their conjugation or polymerization status (Pickart, 2000). Furthermore, a simple accessibility mechanism could not explain the higher affinity of the UBA domain for diubiquitin over monoubiquitin, because the suggested binding sites are less accessible in the diubiquitin case (Varadan et al., 2004). The above-described findings suggest that a more involved mechanism is responsible for the enhanced affinity of the polyubiquitin species. The presence of two different binding sites with different affinities, as in Ufd1, might be one of the mechanisms. This dual binding site mechanism is more likely for the larger ubiquitin binding domains such as the UEV domain (about 145 residues) than for the smaller domains (less than 50 residues) such as UIM, NZF, and UBA due to the requirement of two separate binding sites.

Structure 1002

Figure 5. Binding Sites on the Ubiquitin Molecule (A) Chemical shift changes of the ubiquitin backbone resonance peaks upon addition of Ufd1 are plotted against the residue numbers. The horizontal dotted line indicates the cutoff values used to identify the binding sites. The asterisks on residue 9 and 75 indicate that the corresponding peak residues disappeared completely upon addition of Ufd1. The bar heights of those residues were taken from the highest values within the particular binding motif. Hydrophobic binding motifs found in other interactions involving ubiquitin are shown at the upper part of the graph (see text). (B) The residues involved in Ufd1 binding are colored in red on the ribbon structure of the ubiquitin reported previously (PDB code: 1D3Z). The side chains of the residues in the “hydrophobic patch” (see text) are indicated in blue, and the K48 is shown in green.

In fact, the dual binding site mechanism seems related to an interesting observation reported previously for the UEV domain. VanDemark et al. (2001) solved the crystal structure of the UEV/E2 complex and mapped the binding site of the diubiquitin molecule as coupled through a Lys63 linkage. The diubiquitin binding sites, however, were found to be very different from the monoubiquitin binding site in a related UEV-monoubiquitin complex reported recently (Sundquist et al., 2004). As pointed out in the latter paper, the UEV domains in both reports share the same fold. Therefore, the UEV domain may represent another example for the usage of distinct sites in binding different ubiquitin forms. A Model for the Synergistic Binding of Polyubiquitin to Ufd1 and p97 As mentioned above, the N domains of p97 and Ufd1 are structurally similar, and both have N-terminal double-psi β barrels. The N domains are similar not only structurally but also functionally. They both bind monoubiquitin and polyubiquitin and have higher affinity for the latter. Therefore, we suggest that similar regions in the double-psi β barrel account for the mono- or polyubiquitin binding properties in both p97 and Ufd1 (Figure 6). With these assumptions, we can propose a simple model for the synergistic binding to polyubiquitin by Ufd1 and p97 (Figures 1A and 6). Ufd-1 binds, through its C domain, the N domain of p97 (the Ufd1Npl4 interaction also occur through the C domain of Ufd1 [Hetzer et al., 2001]). This binding puts the doublepsi β barrels of Ufd1 and the p97 in close proximity, providing avidity for synergistic binding to polyubiquitin. Additional synergy might be gained due to the hexameric nature of p97 (see Figure 1A). The six N domains

of p97 in one complex can, in theory, bind six Ufd1Npl4 molecules, providing a large capacity for polyubiquitin association. Then, the polyubiquitin chain might bend around the hexamer for binding with higher

Figure 6. Schematic Diagrams of the Mono- and Polyubiquitin Interactions by p97 and Ufd1 N Domains Bindings are indicated with arrows. Solid lines, interactions with experimental evidence; dotted lines, suggested interactions; red lines, interactions addressed in this report. For blue lines in p97ubiquitin interactions, binding regions are inferred from the Ufd1ubiquitin interactions.

Structural and Binding Studies on Ufd1 1003

avidity. Interestingly, it has been reported that p47 can bind p97 in 6 to 6 stoichiometry at saturation conditions in vitro, although the crystal structure of the same complex contained fewer adaptor molecules per hexamer (Dreveny et al., 2004). Overall, the avidity-driven binding due to the proximal positioning of the double-psi β barrels and the multimeric nature of the UNC complex may be part of the mechanism of the synergistic binding of polyubiquitin to p97 and Ufd1. Experimental Procedures NMR Sample Preparation DNA encoding S. cerevisiae Ufd1 residues 1–208 was amplified from yeast genomic DNA (by using sense primer 5#-GGAATTCCA TATGTTTTCTGGCTTTAGTTC-3# and antisense primer 5#-GGCCTG ATCATTATTTGTAATCCGGTTCC-3#) and cloned into the Nde1 and BamH1 sites of Invitrogen pET15b vector (pET15b-Ufd1N). An E. coli strain (codonPlus, Stratagene, La Jolla) was transformed with pET15b-Ufd1N. The transformed E. coli were grown in M9based minimal media at 37°C and induced at OD600 = 0.6 with IPTG (50 ␮M) for 15 hr. 15N-, 15N/13C-, and 2H/15N/13C-labeled protein samples were prepared by using correspondingly labeled ammonium chloride (>99% 15N), glucose (>99% U-13C), and deuterium oxide (>99.9% 2H). The protein was purified with Ni-NTA affinity chromatography followed by anion-exchange chromatography, when necessary. All protein concentrations were determined from absorbance at 280 nm, by using extinction coefficients calculated from the primary sequence. In some preparations, the his-tag was cleaved with PreScission protease (Amersham, Piscataway), but the spectrum remained essentially the same with noncleaved protein, and the spectra from both preparations were used indiscriminately. Typical NMR samples contained approximately 200 ␮M protein in 20 mM phosphate buffer (pH 6.0) with 95% H2O and 5% D2O, containing 50 mM NaCl and 0.01% NaN3. All of the spectra were obtained at 34°C. NMR Spectroscopy NMR data were acquired by using Bruker DRX 500, Bruker DRX 600, or Varian Unity Plus 750 machines equipped with gradient triple resonance probes. All the triple resonance experiments for backbone resonance experiments were obtained by using the Bruker DRX 500 spectrometer with a cryogenic probe. Sequential assignments of the backbone resonances were obtained by using HNCA, HN(CO)CA, HNCO, and HN(CA)CO with deuterated and/or nondeuterated protein samples. Side chain assignments were obtained by using H(CCO)NH, C(CO)NH, and HCCHTOCSY spectra. For distance constraints, 15N-separated, 13C-separated NOESY-HSQC, 15N-separated HSQC-NOESY with correspondingly labeled samples, and 4D-13C-HMQC-NOESY-HSQC, 4D-13C,15N-HMQC-NOESY-HSQC were obtained with 15N-labeled, deuterated, ILV-methyl-13C-protonated samples (Gross et al., 2003). In the 4D NOSEY spectra, 13C sweep widths were set to those of the methyl carbons. Mixing times of 90 ms and 180 ms were used for protonated and deuterated samples, respectively. Because of the severe crowding of the aromatic region caused by the large number of phenylalanines, we recorded 15N-HMQCNOESY spectra on deuterated samples that contained isoleucine, leucine, and valine residues that were 13C-labeled and protonated at the methyl positions but deuterated elsewhere. In addition, the samples contained protonated phenylalanine or tyrosine residues. The spectra were recorded with the indirect spectral width set to that of the methyl region for high-resolution aromatic-methyl proton NOE analysis. All raw data were processed with nmrPipe (Delaglio et al., 1995) and were analyzed with nmrview (Johnson and Blevins, 1994) software. Structure Calculation The 3D protein structure was calculated based on various experimentally derived NOE constraints and TALOS angle restraints (Cornilescu et al., 1999). Peak intensities were categorized into strong, medium, weak, and very weak, with upper distance bounds set to

3.4, 4.0, 5.5, and 6.5 Å, respectively. Standard pseudoatom corrections were applied as needed. The lower distance bounds were set to 1.8 Å in all distance categories. Hydrogen bond constraints derived from the amide proton deuterium exchange experiment were used to constrain the NH-O distance to 2.8 Å. For helical regions, canonical hydrogen bond patterns were also used to constrain the NH-O to 2.8 Å. Initial structures were obtained by using the CYANA program (Guntert et al., 1997), and the final structures were calculated by using the CNS program (Brunger et al., 1998), implementing the conformational databases and the secondary chemical shift database described earlier (Kuszewski et al., 1995, 1996, 1997). The 15 lowest-energy conformers were picked to represent the ensemble structure, and the lowest-energy conformer was used for all of the figures. These 15 conformers do not have any NOE violations greater than 0.3 Å or angle violations larger than 3°. Ubiquitin Binding Experiment The rabbit E1 and E2-25K enzymes were purchased from Boston Biochem (Boston, MA), and the ubiquitin was purchased from Sigma (bovine ubiquitin, St. Louis, MO) and VLI research (human ubiquitin, Malvern, PA). The E2-25K synthesizes only the K-48linked chains. Polyubiquitin chains were synthesized by using the method described previously with slight changes (Chen et al., 1991). Briefly, ubiquitin was incubated with E1 (1.35 ␮g) and E225K (3 ␮g) in 50 ␮l conjugation buffer (20 mM Tris-HCl [pH 7.6], 20 mM KCl, 5 mM MgCl2, 1 mM DTT, and 10 mM ATP) at 37°C for 20 hr. The monoubiquitin sample was prepared with the same ingredients, except ATP. All the ubiquitin samples were buffer exchanged to the Ufd1 NMR sample buffer with a PD-10 gel filtration cartridge. NMR binding experiments were carried out by mixing 100 ␮l 13C15 N double-labeled Ufd1 with 140 ␮l mono- or polyubiquitin (Ufd1 final concentration = 120 ␮M). All of the samples were prepared from a single-batch preparation of Ufd1. In control experiments without ubiquitin, the presence of buffer and enzyme species did not cause any changes in the spectra. The backbone amide peak positions from the spectrum with the E1, E2, and other components for polyubiquitination reaction were identical to those from the control spectrum (Ufd1 only), showing that those additional components do not bind Ufd1 (data not shown). The combined chemical shift changes (in ppm) were calculated by (⌬δHN2 + [0.17⌬δ15N]2)−, where ⌬δHN and ⌬δ15N are the chemical shift changes for the amide protons and the amide nitrogens, respectively (Park et al., 2002).

Acknowledgments We acknowledge Mr. Gregory Heffron for assistance with the NMR experiments, Dr. Vladimir Gelev for the precursors to make ILVmethyl 13C-prototonated samples, and Dr. Takuhiro Ito for excellent scientific communication. This research was supported by NIH (grants AI37581 and GM47467 to G.W., and grant GM36373 to P.A.S.). Maintenance of the instruments used for this research was in part supported by grant RR 00995. R.L.I. is a recipient of a Wellcome Prize Traveling Fellowship and gratefully acknowledges their support. Received: March 2, 2005 Revised: April 14, 2005 Accepted: April 14, 2005 Published: July 12, 2005 References Adams, J. (2003). The proteasome: structure, function, and role in the cell. Cancer Treat. Rev. 29 (Suppl 1), 3–9. Adams, J. (2004). The development of proteasome inhibitors as anticancer drugs. Cancer Cell 5, 417–421. Alam, S.L., Sun, J., Payne, M., Welch, B.D., Blake, B.K., Davis, D.R., Meyer, H.H., Emr, S.D., and Sundquist, W.I. (2004). Ubiquitin interactions of NZF zinc fingers. EMBO J. 23, 1411–1421. Bays, N.W., and Hampton, R.Y. (2002). Cdc48-Ufd1-Npl4: stuck in the middle with Ub. Curr. Biol. 12, R366–R371.

Structure 1004

Bruderer, R.M., Brasseur, C., and Meyer, H.H. (2004). The AAA ATPase p97/VCP interacts with its alternative co-factors, Ufd1Npl4 and p47, through a common bipartite binding mechanism. J. Biol. Chem. 279, 49609–49616. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921. Cao, K., Nakajima, R., Meyer, H.H., and Zheng, Y. (2003). The AAAATPase Cdc48/p97 regulates spindle disassembly at the end of mitosis. Cell 115, 355–367. Castillo, R.M., Mizuguchi, K., Dhanaraj, V., Albert, A., Blundell, T.L., and Murzin, A.G. (1999). A six-stranded double-psi beta barrel is shared by several protein superfamilies. Struct. Fold. Des. 7, 227– 236. Chen, Z.J., Niles, E.G., and Pickart, C.M. (1991). Isolation of a cDNA encoding a mammalian multiubiquitinating enzyme (E225K) and overexpression of the functional enzyme in Escherichia coli. J. Biol. Chem. 266, 15698–15704. Coles, M., Diercks, T., Liermann, J., Groger, A., Rockel, B., Baumeister, W., Koretke, K.K., Lupas, A., Peters, J., and Kessler, H. (1999). The solution structure of VAT-N reveals a ‘missing link’ in the evolution of complex enzymes from a simple betaalphabetabeta element. Curr. Biol. 9, 1158–1168. Cook, M.A., and Tyers, M. (2004). Cellular differentiation: the violin strikes up another tune. Curr. Biol. 14, R11–R13. Cook, W.J., Jeffrey, L.C., Carson, M., Chen, Z., and Pickart, C.M. (1992). Structure of a diubiquitin conjugate and a model for interaction with ubiquitin conjugating enzyme (E2). J. Biol. Chem. 267, 16467–16471.

Hu, X. (2003). Proteolytic signaling by TNFalpha: caspase activation and IkappaB degradation. Cytokine 21, 286–294. Imai, J., Yashiroda, H., Maruya, M., Yahara, I., and Tanaka, K. (2003). Proteasomes and molecular chaperones: cellular machinery responsible for folding and destruction of unfolded proteins. Cell Cycle 2, 585–590. Jarosch, E., Lenk, U., and Sommer, T. (2003). Endoplasmic reticulum-associated protein degradation. Int. Rev. Cytol. 223, 39–81. Johnson, B.A., and Blevins, R.A. (1994). NMRView: a computer program for the visualization and analysis of NMR data. J. Biomol. NMR 4, 603–614. Johnson, E.S., Ma, P.C., Ota, I.M., and Varshavsky, A. (1995). A proteolytic pathway that recognizes ubiquitin as a degradation signal. J. Biol. Chem. 270, 17442–17456. Kang, R.S., Daniels, C.M., Francis, S.A., Shih, S.C., Salerno, W.J., Hicke, L., and Radhakrishnan, I. (2003). Solution structure of a CUE-ubiquitin complex reveals a conserved mode of ubiquitin binding. Cell 113, 621–630. Kuszewski, J., Qin, J., Gronenborn, A.M., and Clore, G.M. (1995). The impact of direct refinement against 13C alpha and 13C beta chemical shifts on protein structure determination by NMR. J. Magn. Reson. B. 106, 92–96. Kuszewski, J., Gronenborn, A.M., and Clore, G.M. (1996). Improving the quality of NMR and crystallographic protein structures by means of a conformational database potential derived from structure databases. Protein Sci. 5, 1067–1080. Kuszewski, J., Gronenborn, A.M., and Clore, G.M. (1997). Improvements and extensions in the conformational database potential for the refinement of NMR and X-ray structures of proteins and nucleic acids. J. Magn. Reson. 125, 171–177.

Cook, W.J., Jeffrey, L.C., Kasperek, E., and Pickart, C.M. (1994). Structure of tetraubiquitin shows how multiubiquitin chains can be formed. J. Mol. Biol. 236, 601–609.

Meyer, H.H., Shorter, J.G., Seemann, J., Pappin, D., and Warren, G. (2000). A complex of mammalian ufd1 and npl4 links the AAAATPase, p97, to ubiquitin and nuclear transport pathways. EMBO J. 19, 2181–2192.

Cornilescu, G., Delaglio, F., and Bax, A. (1999). Protein backbone angle restraints from searching a database for chemical shift and sequence homology. J. Biomol. NMR 13, 289–302.

Meyer, H.H., Wang, Y., and Warren, G. (2002). Direct binding of ubiquitin conjugates by the mammalian p97 adaptor complexes, p47 and Ufd1-Npl4. EMBO J. 21, 5645–5652.

Dai, R.M., and Li, C.C. (2001). Valosin-containing protein is a multiubiquitin chain-targeting factor required in ubiquitin-proteasome degradation. Nat. Cell Biol. 3, 740–744.

Park, S., Johnson, M.E., and Fung, L.W. (2002). Nuclear magnetic resonance studies of mutations at the tetramerization region of human alpha spectrin. Blood 100, 283–288.

Delaglio, F., Grzesiek, S., Vuister, G.W., Zhu, G., Pfeifer, J., and Bax, A. (1995). NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277–293.

Passmore, L.A., and Barford, D. (2004). Getting into position: the catalytic mechanisms of protein ubiquitylation. Biochem. J. 379, 513–525.

Dou, Q.P., Smith, D.M., Daniel, K.G., and Kazi, A. (2003). Interruption of tumor cell cycle progression through proteasome inhibition: implications for cancer therapy. Prog. Cell Cycle Res. 5, 441–446.

Phillips, C.L., Thrower, J., Pickart, C.M., and Hill, C.P. (2001). Structure of a new crystal form of tetraubiquitin. Acta Crystallogr. D Biol. Crystallogr. 57, 341–344.

Dreveny, I., Kondo, H., Uchiyama, K., Shaw, A., Zhang, X., and Freemont, P.S. (2004). Structural basis of the interaction between the AAA ATPase p97/VCP and its adaptor protein p47. EMBO J. 23, 1030–1039.

Pickart, C.M. (2000). Ubiquitin in chains. Trends Biochem. Sci. 25, 544–548.

Golbik, R., Lupas, A.N., Koretke, K.K., Baumeister, W., and Peters, J. (1999). The Janus face of the archaeal Cdc48/p97 homologue VAT: protein folding versus unfolding. Biol. Chem. 380, 1049–1062.

Rape, M., Hoppe, T., Gorr, I., Kalocay, M., Richly, H., and Jentsch, S. (2001). Mobilization of processed, membrane-tethered SPT23 transcription factor by CDC48(UFD1/NPL4), a ubiquitin-selective chaperone. Cell 107, 667–677.

Gross, J.D., Gelev, V.M., and Wagner, G. (2003). A sensitive and robust method for obtaining intermolecular NOEs between side chains in large protein complexes. J. Biomol. NMR 25, 235–242. Guntert, P., Mumenthaler, C., and Wuthrich, K. (1997). Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283–298. Haglund, K., Di Fiore, P.P., and Dikic, I. (2003). Distinct monoubiquitin signals in receptor endocytosis. Trends Biochem. Sci. 28, 598– 603. Hetzer, M., Meyer, H.H., Walther, T.C., Bilbao-Cortes, D., Warren, G., and Mattaj, I.W. (2001). Distinct AAA-ATPase p97 complexes function in discrete steps of nuclear assembly. Nat. Cell Biol. 3, 1086–1091. Hicke, L., and Dunn, R. (2003). Regulation of membrane protein transport by ubiquitin and ubiquitin-binding proteins. Annu. Rev. Cell Dev. Biol. 19, 141–172.

Pickart, C.M. (2004). Back to the future with ubiquitin. Cell 116, 181–190.

Raasi, S., Orlov, I., Fleming, K.G., and Pckart, C.M. (2004). Binding of polyubiquitin chains to ubiquitin-associated (UBA) domains of HHR23A. J. Mol. Biol. 341, 1367–1379. Ryu, K.S., Lee, K.J., Bae, S.H., Kim, B.K., Kim, K.A., and Choi, B.S. (2003). Binding surface mapping of intra- and interdomain interactions among hHR23B, ubiquitin, and polyubiquitin binding site 2 of S5a. J. Biol. Chem. 278, 36621–36627. Sundquist, W.I., Schubert, H.L., Kelly, B.N., Hill, G.C., Holton, J.M., and Hill, C.P. (2004). Ubiquitin recognition by the human TSG101 protein. Mol. Cell 13, 783–789. Uchiyama, K., Jokitalo, E., Kano, F., Murata, M., Zhang, X., Canas, B., Newman, R., Rabouille, C., Pappin, D., Freemont, P., and Kondo, H. (2002). VCIP135, a novel essential factor for p97/p47-mediated membrane fusion, is required for Golgi and ER assembly in vivo. J. Cell Biol. 159, 855–866.

Structural and Binding Studies on Ufd1 1005

VanDemark, A.P., Hofmann, R.M., Tsui, C., Pickart, C.M., and Wolberger, C. (2001). Molecular insights into polyubiquitin chain assembly: crystal structure of the Mms2/Ubc13 heterodimer. Cell 105, 711–720. Varadan, R., Assfalg, M., Haririnia, A., Raasi, S., Pickart, C., and Fushman, D. (2004). Solution conformation of Lys63-linked di-ubiquitin chain provides clues to functional diversity of polyubiquitin signaling. J. Biol. Chem. 279, 7055–7063. Ye, Y., Meyer, H.H., and Rapoport, T.A. (2003). Function of the p97Ufd1-Npl4 complex in retrotranslocation from the ER to the cytosol: dual recognition of nonubiquitinated polypeptide segments and polyubiquitin chains. J. Cell Biol. 162, 71–84. Yu, R.C., Jahn, R., and Brunger, A.T. (1999). NSF N-terminal domain crystal structure: models of NSF function. Mol. Cell 4, 97–107. Yuan, X., Simpson, P., McKeown, C., Kondo, H., Uchiyama, K., Wallis, R., Dreveny, I., Keetch, C., Zhang, X., Robinson, C., et al. (2004). Structure, dynamics and interactions of p47, a major adaptor of the AAA ATPase, p97. EMBO J. 23, 1463–1473. Zhang, F., and Laiho, M. (2003). On and off: proteasome and TGFbeta signaling. Exp. Cell Res. 291, 275–281. Zhang, X., Shaw, A., Bates, P.A., Newman, R.H., Gowen, B., Orlova, E., Gorman, M.A., Kondo, H., Dokurno, P., Lally, J., et al. (2000). Structure of the AAA ATPase p97. Mol. Cell 6, 1473–1484. Accession Numbers The coordinates for Ufd1 have been deposited at the Protein Data Bank under accession code 1ZC1.