Structural Details of Ufd1 Binding to p97 and Their Functional Implications in ER-Associated Degradation

The hexameric ATPase p97 has been implicated in diverse cellular processes through interactions with many different adaptor proteins at its N-terminal domain. Among these, the Ufd1-Npl4 heterodimer is a major adaptor, and the p97-Ufd1-Npl4 complex plays an essential role in endoplasmic reticulum-associated degradation (ERAD), acting as a segregase that translocates the ubiquitinated client protein from the ER membrane into the cytosol for proteasomal degradation. We determined the crystal structure of the complex of the N-terminal domain of p97 and the SHP box of Ufd1 at a resolution of 1.55 Å. The 11-residue-long SHP box of Ufd1 binds at the far-most side of the Nc lobe of the p97 N domain primarily through hydrophobic interactions, such that F225, F228, N233 and L235 of the SHP box contact hydrophobic residues on the surface of the p97 Nc lobe. Mutating these key interface residues abolished the interactions in two different binding experiments, isothermal titration calorimetry and co-immunoprecipitation. Furthermore, cycloheximide chase assays showed that these same mutations caused accumulation of tyrosinase-C89R, a well-known ERAD substrate, thus implying decreased rate of protein degradation due to their defects in ERAD function. Together, these results provide structural and biochemical insights into the interaction between p97 N domain and Ufd1 SHP box.

Introduction p97, also known as VCP (valosin-containing protein), is a hexameric ATPase of type II AAA+ family [1]. Each p97 protomer comprises an N-terminal domain (hereafter N domain), two ATPase domains in tandem (D1 and D2) which form doubly packed hexameric rings, and a short C-terminal tail region [2][3][4][5]. p97 has been implicated in a variety of cellular processes, such as ER-associated degradation, post-mitotic Golgi reassembly, cell cycle regulation, apoptosis, DNA damage response, mitochondria quality control, and autophagy [6]. p97 is among The fusion proteins were over-expressed in E. coli strain Rosetta2 (DE3) and then purified after disruption by sonication. Purification scheme is as follows: his-tag affinity column, a desalting column, cut-off of the tag by TEV protease [27], his-tag affinity column again to remove the tag, and a size-exclusion column. All columns were purchased from GE Healthcare Bioscience, USA: a HisPrep column for his-tag affinity chromatography, a HiTrap Desalting column for desalting, and a Superdex-200 column for size-exclusion chromatography. The purified fusion protein was concentrated to 20.7 mg/ml for crystallization. Peptide samples for Ufd1 SHP box were synthesized by Peptron, Korea.

Crystallization and structure determination
Crystallization conditions were initially searched using commercial screening kits, and the finally refined composition of well solution was 21%(w/v) PEG 8000, 2.9mM n-nonyl-β-thiomaltoside, 0.1 M HEPES, pH 7.5. The crystal was cryo-protected by quick soaking in a well solution containing additional 10%(v/v) glycerol. X-ray diffraction data were collected at the beamline 5C of Pohang Light Source, Korea. The raw diffraction images were processed, merged and scaled with MOSFLM and SCALA of CCP4 program suite [28]. Molecular replacement calculation was carried out by Phaser [29], and the subsequent model rebuilding and refinement were performed iteratively using Coot and Refmac [30,31]. The coordinates and structure factors have been deposited in the Protein Data Bank with the accession number 5B6C (http://www.pdb.org).

Isothermal titration calorimetry
Isothermal titration calorimetry (ITC) experiments were carried out using a ITC200 instrument (MicroCal Inc., USA). Titrations were carried out by injecting consecutive aliquots of p97 N domain (2.0 mM) into the ITC cell containing Ufd1 SHP peptide (0.050 mM) at 25°C. Binding stoichiometry, enthalpy, entropy, and binding constants were determined by fitting the data to a one-site binding model. The ITC data were fit using Origin 7.0 (MicroCal Inc., USA).

Reagents and antibodies
DMEM, cycloheximide, anti-FLAG M2 affinity gels, and mouse monoclonal antibodies (mAbs) to α-tubulin and FLAG were purchased from Sigma-Aldrich. FBS and penicillin/streptomycin were obtained from Hyclone (Logan, UT). A mouse mAb to HA (16B12) and a goat polyclonal antibody to β-actin were obtained from Covance (Richmond, CA) and Santa Cruz Biotechnology, respectively. Mouse (9B11) and rabbit (71D10) mAbs to Myc were purchased from Cell Signaling Technology. Lipofectamine 2000 and Opti-MEM I were purchased from Thermo Fisher Scientific.

Cell culture and transfection
HeLa cells were grown in DMEM supplemented with 10% FBS and penicillin/streptomycin at 37°C in a humidified atmosphere of 5% CO 2 and 95% air. Cells were grown up to 70~80% (258-275), previously identified by Bruderer et al. (2004) [23]. (b) Overall structure of p97 N domain (green) with a bound Ufd1 SHP box (pink). Secondary structural elements of the p97 N domain are noted in accordance with the previous report by Zhang et al. (2000) [3], except for β ud (β-strand undefined previously) which was newly assigned in the present study.

Western blotting and immunoprecipitation
Cells were harvested in cold lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM Na 3 VO 4 , 5 mM NaF, and 1% Triton X-100) containing protease inhibitor cocktail tablets (Roche) and lysed with a TissueLyzer II (Qiagen) for 5 min. After clearance by centrifugation (15,000 × g, 20 min, 4°C), the protein concentration of the cell lysates was determined using bicinchoninic acid protein assay reagents (Pierce, Rockford, IL). For immunoprecipitation of FLAG-Ufd1 fusion proteins, cell lysates (0.5 mg) were incubated with 20 μl anti-FLAG M2 affinity gel for 4 h at 4°C. In case of immunoprecipitation of HA-p97 fusion proteins, cell lysates (1.0 mg) were incubated with 5.0 μg of anti-HA antibody for 4 h at 4°C, and then the immune complexes were captured with 50 μl protein G Sepharose 4 Fast Flow beads (GE Healthcare Life Sciences) for an additional 2 h. All immunoprecipitated samples were washed with the cell lysis buffer five times. The starting cell lysates and immunoprecipitates in Laemmli sample buffer were separated by SDS-PAGE on 8-10% resolving gels and transferred to nitrocellulose membranes (Schleicher & Schuell Bioscience, Germany). Following blocking with 5% nonfat milk in TBS containing 0.1% Tween-20 (TBST), membrane blots were incubated with primary antibodies against FLAG, HA and α-tubulin for 2 h at room temperature, washed three times with TBST, and further incubated with HRP-conjugated secondary antibodies (Zymed Laboratories, San Francisco, CA). The resulting immune complexes were detected using SuperSignal West Pico chemiluminescent substrate (Pierce, Rockford, IL).

ER-associated degradation
For the degradation of the tyrosinase C89R mutant (Tyr C89R ), cells plated on 35 mm-dishes were transfected with Myc-Tyr C89R together with FLAG-Ufd1 or HA-p97 expression constructs (total 3 μg at a 1:1 ratio) or their relevant empty vectors for 16 h and then cell lysates were prepared as described above. For cycloheximide chase assays, the cotransfected cells were left untreated or treated with 50 μg/ml cycloheximide for up to 2 h. The protein levels of Myc-Tyr C89R and transfected proteins were analyzed by Western blotting with mouse mAbs to Myc, FLAG and HA. Band intensities of the immunoblots were measured using NIH ImageJ software (National Institutes of Health, Bethesda, MD).

Structure determination of the complex between the p97 N domain and the Ufd1 SHP box
Attempts to reconstitute and crystallize the complex between the p97 N domain and the Ufd1 SHP box were not successful for any of the constructs generated. Thus, we fused the p97 N domain ; hereafter referred to as p97N) to the Ufd1 SHP box (225-235; hereafter referred to as Ufd1SHP) in a single polypeptide chain by using a long flexible linker. We designed two different fusion constructs: p97N-(GGS) 4 -Ufd1SHP and Ufd1SHP-(GGS) 4 -p97N. Both constructs were expressed well in E. coli, and then purified. The first construct produced a well-diffracting crystal, thus allowing a data set to be collected to 1.55 Å resolution limit ( Fig 1A and Table 1). The structure was determined by molecular replacement method with the previously reported p97 N domain structure (Protein Data Bank entry 3QQ7) [32] as a search model. After positioning one molecule of the p97 N domain in the asymmetric unit, a well-defined electron density for the bound Ufd1SHP was observed in the difference map. The final refined model for p97N-Ufd1SHP complex contains residues Asn21 to Glu192 of the p97 N domain and the full 11-residue-long SHP box of Ufd1 (225-FRAFSGSGNRL-235). Four residues, Asp193 to Glu196, in the C-terminus of the p97 N domain, as well as twelve residues in the linker, (GGS) 4 , were not visible in the electron density map, even at the final stage of structure refinement, thus implying that these regions are disordered in the crystal lattice. The statistics for data collection and structure refinement are summarized in Table 1.
During the preparation of this manuscript, another research group reported a crystal structure for the complex between hexameric full-length p97 and the Ufd1 SHP box peptide at a moderate resolution of 3.08 Å, with less well-defined electron density for the SHP box, particularly for the central GSG region and the side chain orientations of the following C-terminal residues, as noted by authors [33]. In contrast, we used monomeric p97N rather than the fulllength p97 to reconstitute the complex with Ufd1SHP. In addition, we constructed a fusion polypeptide containing both p97N and Ufd1SHP in a single chain to overcome their weak interaction and achieve higher occupancy of the bound Ufd1SHP in the crystal lattice. These two points may be the reasons why we were able to achieve such a high resolution of 1.55 Å, and obtain a clearly defined electron density map for this interaction (Fig 2C). The 1.55-Å resolution structure revealed the complete atomic details of this interaction, many of which have not previously been described in the recent report [33].
High-resolution details of the interaction between the p97 N domain and the Ufd1 SHP box As reported previously [2,3,32], the p97 N domain comprises two subdomains, Nn and Nc lobes; thus, it adopts an overall bilobed structure. The current crystal structure further reveals that the Ufd1SHP (225-FRAFSGSGNRL-235) binds only the Nc lobe, consistent with the recent report [33]. The SHP box wraps around the far-most face of the Nc lobe, with all eleven residues largely forming a random coil, except for a short stretch of β-strand from Asn233 to Arg234. This short β-strand, β SHP , binds to β11 at the outer edge of the central four-stranded β-sheet (β8-β9-β7-β11) of the Nc subdomain, thereby extending it to a five-stranded sheet (β8-β9-β7-β11-β SHP : four strands from p97N and one strand from Ufd1SHP) (Fig 1B). The entire Ufd1 SHP box adopts a bent conformation to establish elaborate interactions with the curved surface of the Nc lobe (Fig 2A and 2B). Two glycine residues (Gly230 and Gly232) generate a sharp kink in the middle of the SHP box, thereby enabling the bending of Ufd1SHP upon binding to p97N. These two glycine residues, which occupy the sixth and eighth positions in the eleven-residue SHP sequence, are strictly conserved through SHP-containing p97 adaptor proteins, reflecting their critical roles (Fig 2E). The Ufd1 SHP box binds to the p97 Nc subdomain primarily through hydrophobic interactions (Fig 2B), and the most prominent interactions are established by four residues of the Ufd1 SHP box (Phe225, Phe228, Asn233 and Leu235), whose side chains are positioned toward the p97 Nc subdomain (Figs 2B and 3A). Specifically, the phenyl ring of Phe225-Ufd1 sits on the hydrophobic patch formed by His115, Leu117 and Val166, forming a T-shaped π-π stack with the imidazole ring of His115. The phenyl ring of Phe228-Ufd1 stacks with the guanidinium group of Arg113, thereby establishing a cation-π interaction, and also forms a T-shaped π-π stack with the imidazole ring of His183. The side-chain amide plane of Asn233-Ufd1 stacks in parallel with the phenyl ring of Phe131, and the Leu235-Ufd1 at the C-terminal end inserts its side chain into the hydrophobic pocket constructed by Phe139, Leu140, V176, Ile182 and Phe131. Notably, the above three hydrophobic residues of Ufd1SHP (Phe225, Phe228 and Leu235) are highly conserved throughout SHP-containing p97 adaptor proteins (Ufd1, p47 and DVC1), which have been experimentally shown to bind to the p97 N domain (Fig 2E) [23,34,35].
In addition, several hydrogen bonds are also observed between main-chain atoms. First, the two main-chain oxygen atoms of Phe228-Ufd1 and Gly230-Ufd1 are weakly hydrogen-bonded to Nε2 of His183 at distances of 3.17 and 3.42 Å, respectively. Second, regarding to the short βstrand of Ufd1SHP (β SHP , Asn233 to Arg234) which binds to β11 of the Nc subdomain, the main-chain nitrogen and oxygen atoms of Asn233-Ufd1 (β SHP ) are hydrogen-bonded to the  Fig 2B. (b, c) Isothermal titration calorimetry measurements of the interaction between the p97 N domain and the Ufd1 SHP box peptide: (b) p97N WT with Ufd1SHP mutants, and (c) p97N mutants with Ufd1SHP WT . (d, e) Reciprocal coimmunoprecipitation. HeLa cells were cotransfected with expression plasmids encoding wild-type and mutant HA-p97 and FLAG-Ufd1 as indicated. At 24 h after transfection, the cell lysates were immunoprecipitated using anti-HA antibody and protein G beads or using anti-FLAG antibody-conjugated beads. The starting lysates (input) and immunoprecipitates (IP) were analyzed by Western blotting with anti-FLAG and anti-HA antibodies.α-tubulin was included as a loading control. main-chain oxygen and nitrogen atoms of Ile182 (β11 of p97N) in a typical way of bring βstrands together into a β-sheet. Notably, the Ufd1 SHP box contains two arginine residues, Arg226-Ufd1 and Arg234-Ufd1, whose side chains point in the opposite direction from the binding interface and therefore do not establish direct interactions with any p97 residues. This should be a reason why the side chains of these two arginine residues are not so well defined in the electron density as all the other residues of the Ufd1 SHP box (Figs 2C and 3A).

Biochemical analysis of the interface
To biochemically verify the above structural observations, we performed mutational analyses for several key interface residues identified from the crystal structure. To begin with, we measured the binding affinity between the wild-types of the Ufd1 SHP box and the p97 N domain via isothermal titration calorimetry (ITC) by using a synthetic peptide of the Ufd1 SHP box and the recombinant p97 N domain (Fig 2D). The dissociation constant K d was determined to be 221 ± 36 μM, implying that the interaction is quite weak. We then assessed whether mutations in the interface residues affect the binding affinity. Firstly, three mutant Ufd1SHP peptides, carrying F225A, F228A or L235A substitutions, were titrated against the wild-type p97 N domain. Next, three mutants of the p97 N domain, F131A, R113A/H115A and I182A/H183A, were tested with the wild-type Ufd1SHP peptide. Mutating the Ufd1SHP residues reduced the binding affinity to an undetectable level in the ITC experiment (Fig 3B), and mutating the p97N residues also abolished most of the binding affinity (Fig 3C).
p97-Ufd1 binding was also explored in cultured cells by reciprocal co-immunoprecipitation. To this end, we generated mammalian expression constructs of HA-tagged p97 and FLAGtagged Ufd1 encoding full-length proteins containing the same series of mutations. After cotransfection into HeLa cells, p97 and Ufd1 in cell lysates and immunoprecipitation (IP) products were detected by Western blotting with anti-HA and anti-FLAG antibodies. First, wild-type or a mutant FLAG-Ufd1 was expressed together with wild-type HA-p97 (Fig 3D). The anti-HA IP products contained FLAG-Ufd1 WT , as expected, but the three FLAG-Ufd1 mutants (F225A, F228A and L235A) co-precipitated with HA-p97 to a much lesser degree ( Fig  3D, left). On the immunoprecipitation of the same cell lysates conversely by anti-FLAG antibody, the HA-p97 co-precipitated only weakly with FLAG-Ufd1 L235A and not at all with FLA-G-Ufd1 F225A and FLAG-Ufd1 F228A , in contrast to the strong co-precipitation with FLAG-Ufd1 WT (Fig 3D, right). The results from anti-HA and anti-FLAG IP experiments were in good agreement. Second, wild-type or mutant HA-p97 is co-expressed with wild-type FLA-G-Ufd1 (Fig 3E). Both immunoprecipitation trials using anti-FLAG and anti-HA antibody commonly showed that all three p97 mutants (F131A, R113A/H115A and I182A/H183A) completely lost the binding affinity for FLAG-Ufd1 WT .

Functional implications in ER-associated degradation
To investigate the functional significance of the structural details of the p97-Ufd1 interaction, we examined the effects of the above mutations on ER-associated degradation, a well-established cellular process involving both p97 and Ufd1. To this end, we assessed the protein level of Myc-tagged tyrosinase harboring the C89R mutation (tyrosinase-C89R or Tyr C89R ), an ERAD substrate [36], with anti-Myc immunoblotting approximately 16 hours after cotransfection with wild-type or mutant HA-tagged p97 or FLAG-tagged Ufd1 (Fig 4A and 4B). The level of Myc-Tyr C89R in the FLAG-Ufd1 WT -cotransfected sample was lower than that observed with the expression of only Myc-Tyr C89R , indicating that FLAG-Ufd1 WT enhanced the degradation of Myc-Tyr C89R (Fig 4A). In contrast, the protein level of Myc-Tyr C89R was much higher in cells coexpressing the FLAG-Ufd1 mutant than in cells coexpressing FLAG-Ufd1 WT : 4.05 ±0.37 fold for Ufd1 F225A , 4.58±0.42 fold for Ufd1 F228A and 3.46±0.25 fold for Ufd1 L235A . Similarly, the Myc-Tyr C89R protein level was decreased after coexpression of HA-p97 WT but markedly increased after the HA-p97 mutants: 4.74±0.52 fold for p97 F131A , 5.00±0.42 fold for p97 R113A/H115A and 4.70±0.46 fold for p97 I182A/H183A compared with HA-p97 WT (Fig 4B). Thus, all the tested Ufd1 and p97 interface mutants, compared with the wild types, showed substantially reduced tysoninase-C89R degradation. These results are consistent with the above binding affinity evaluation (Fig 3B-3E), in which all the same mutations greatly diminished the binding between Ufd1 and p97.
Given the well-established role for p97-Ufd1-Npl4 complex as a "segregase" that sequesters (or "segregates") substrate proteins out of the ER membrane, thereby facilitating substrate translocation into the cytoplasm for subsequent proteasomal degradation [22], p97 and Ufd1 interface mutants can be reasonably assumed to delay protein degradation of tyrosinase-C89R, resulting in its accumulation. To verify this hypothesis, we performed standard cycloheximide (CHX) chase assays in the same experimental conditions as above and measured differences in the rate of degradation of Myc-Tyr C89R between cells expressing wild-type and mutants of p97 or Ufd1 (Fig 4C-4F). As expected, Myc-Tyr C89R underwent degradation CHX time-dependently upon Ufd1 WT expression (Fig 4C and 4D). In contrast, Ufd1 F225A -, Ufd1 F228A -and Ufd1 L235A -expressing cells were much less active in degrading Myc-Tyr C89R , showing relatively prolonged retention time of the ERAD substrate (Fig 4C and 4D). Similarly, time-dependent decreases in Tyr C89R protein levels were observed in the p97 WT -expressing cells treated with CHX (Fig 4E and 4F). In a sharp contrast, however, Tyr C89R remained at relatively high levels in the p97 F131A -, p97 R113A/H115A -and p97 I182A/H183A -expressing cells (Fig 4E and 4F), which is very similar to the results obtained from Ufd1 mutants. These results imply significant defects in degradation of Tyr C89R , an ERAD substrate, without the binding of Ufd1 to p97.

Discussion
Distinctive feature of SHP box as a p97N-binding module The capacity of p97 to engage in diverse cellular processes is mediated by interacting with a variety of adaptor proteins at its N-terminal domain and C-terminal tail. Adaptors that bind the C-terminal tail utilize PUB or PUL domains as p97C-binding modules [37][38][39]. Far more adaptors bind the N-terminal domain, and five different p97N-binding modules included in these adaptors have been identified to date: UBX, UBD (also called ULD or UBX-L), VIM, VBM and SHP box [9,10]. Four p97N-binding modules (UBX, UBD, VIM, and VBM) have all been shown to bind to the cleft between the two subdomains of p97N; thus, the binding regions on the p97N surface cannot avoid overlapping to some extent [24,32,[40][41][42]. However, in the present study, the SHP box was observed to bind at the far-most side of the Nc lobe, which is distant from the inter-subdomain cleft where the other four p97N-binding modules bind, and this observation was well consistent with a recent study [33].
The SHP box has been identified as a common motif (initially called binding site 1, BS1) in two major p97 adaptors, p47 and Ufd1-Npl4 [23]. Since then, more proteins have been demonstrated to bind p97 via SHP boxes. DVC1 recruits p97 via an SHP box to sites of DNA damage to extract translesion synthesis polymerase η from the DNA repair complex [34,35]. Derlin-1, an inactive intramembrane protease of the rhomoboid family and a central component of p97-interacting ER membrane complex involved in ERAD [43,44], also binds p97 via a C-terminal SHP box [44]. ASPL/TUG also contains a SHP box to bind p97 [45] and has previously been implicated in the insulin-stimulated redistribution of the glucose transporter GLUT4 and in the assembly of the Golgi compartment [46,47]. In addition, BLAST searches using a SHP box sequence as a query return a variety of proteins (http://blast.ncbi.nlm.nih.gov; result not shown), some of which have previously been reported to interact with p97 without recognition of the SHP box, such as RanBP2 (E3 SUMO ligase) at the nuclear pore complex (a SHP box at residues 1968 to 1978: 1968-FKGFSGAGEKL-1978) [48]. Thus, the SHP box seems to be as widely used as the UBX domain, which defines a large protein family, as a general p97N-binding module [49,50]. From this perspective, our complete, atomic-detail description of the SHP-p97 interaction should lay the groundwork for future functional and biochemical studies of SHP-containing protein family members, many of which await characterization.
Assembly of p97-Ufd1-Npl4 complex and its segregase function Because the SHP box of Ufd1 and the UBD of Npl4 bind to different sites on the surface of the p97 N domain, it may be structurally possible for the two modules of an Ufd1-Npl4 heterodimer to occupy the same N domain among the six in a hexameric assembly of p97. However, it may be also possible for the two modules of Ufd1-Npl4 to bind alternatively to two different N domains in a hexameric p97, because only one Ufd1-Npl4 complex binds to the hexameric p97 [20]. To our knowledge, there is no experimental evidence exclusively supporting either scenario, and this issue might be further addressed through biochemical and/or structural analyses of the whole p97-Ufd1-Npl4 complex.
The assembly of p97-Ufd1-Npl4 complex is established through three different types of binary interactions between the three proteins (Figs 1A and 5): first, the Npl4 UBD binds to the inter-subdomain cleft of p97N; second, the Ufd1 SHP box binds to the far-most face of the Nc subdomain of p97N; and third, the Ufd1 NBM (named here as Npl4-binding motif) binds to the yet-uncharacterized central region of Npl4 [23]. The first and second interactions, Npl4UBD-p97N and Ufd1SHP-p97N, have been structurally characterized in detail through NMR and X-ray analyses, including the present study [24,33]. However, the Ufd1-Npl4 interaction has not been structurally or biochemically characterized. Previous work has revealed that the Npl4-binding motif resides between residues 258 and 275 in Ufd1, but the minimal sequence of the motif has yet to be identified [23].
The p97-Ufd1-Npl4 complex, organized by the above three binary interactions, reportedly binds to the ubiquitin chain of a client protein through the C-terminal zinc-finger motif of Npl4, the N domain of Ufd1, and the N domain of p97 [51][52][53][54]. The p97 exerts mechanical forces, which is generated from large conformational changes during ATP hydrolysis, on the ubiquitinated client protein via the Ufd1-Npl4 adaptor and thereby separates the client from its holding environment, such as a subcellular membrane or a multi-protein complex [6,22]. In the present study, we addressed one of the important structural issues for p97-Ufd1-Npl4 complex, particularly with respect to its assembly. The structural features of the interaction between the Ufd1 SHP box and the p97 N domain were clearly revealed in atomic details by determining the crystal structure at a high resolution of 1.55 Å. Furthermore, the observed structural details were corroborated by binding assays and functionally explored with respect to ER-associated degradation.  [41], the UBD of Npl4 (PDB, 2PJH) [24], the VIM of gp78 (PDB, 3TIW) [40], and the SHP of Ufd1 from the present study. (b) The p97-Ufd1-Npl4 complex is assembled by three different binary interactions. Ufd1 is colored in pink, Npl4 in blue, and p97 in gray. For simplicity, the binding of SHP (Ufd1) and UBD (Npl4) are pictured together on a single p97 N domain. However, whether two modules of the Ufd1-Npl4 complex bind to a single N domain or to two different N domains of a p97 hexamer remains unknown, as discussed in the text.