The N-terminal Region of the Ubiquitin Regulatory X (UBX) Domain-containing Protein 1 (UBXD1) Modulates Interdomain Communication within the Valosin-containing Protein p97*

Background: p97 cooperates with cofactors to control various aspects of cellular homeostasis. Mutations at the interdomain interface cause a multisystem degenerative disorder. Results: We identified three binding epitopes on p97 for the N-terminal domain of cofactor UBXD1 (UBXD1-N), including disease-associated residues. Binding reduced p97 ATPase activity. Conclusion: UBXD1-N modulates interdomain communication and activity of p97. Significance: The polyvalent binding mode defines a new subset of p97 cofactors. Valosin-containing protein/p97 is an ATP-driven protein segregase that cooperates with distinct protein cofactors to control various aspects of cellular homeostasis. Mutations at the interface between the regulatory N-domain and the first of two ATPase domains (D1 and D2) deregulate the ATPase activity and cause a multisystem degenerative disorder, inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia/amyotrophic lateral sclerosis. Intriguingly, the mutations affect only a subset of p97-mediated pathways correlating with unbalanced cofactor interactions and most prominently compromised binding of the ubiquitin regulatory X domain-containing protein 1 (UBXD1) cofactor during endolysosomal sorting of caveolin-1. However, how the mutations impinge on the p97-cofactor interplay is unclear so far. In cell-based endosomal localization studies, we identified a critical role of the N-terminal region of UBXD1 (UBXD1-N). Biophysical studies using NMR and CD spectroscopy revealed that UBXD1-N can be classified as intrinsically disordered. NMR titration experiments confirmed a valosin-containing protein/p97 interaction motif and identified a second binding site at helices 1 and 2 of UBXD1-N as binding interfaces for p97. In reverse titration experiments, we identified two distant epitopes on the p97 N-domain that include disease-associated residues and an additional interaction between UBXD1-N and the D1D2 barrel of p97 that was confirmed by fluorescence anisotropy. Functionally, binding of UBXD1-N to p97 led to a reduction of ATPase activity and partial protection from proteolysis. These findings indicate that UBXD1-N intercalates into the p97-ND1 interface, thereby modulating interdomain communication of p97 domains and its activity with relevance for disease pathogenesis. We propose that the polyvalent binding mode characterized for UBXD1-N is a more general principle that defines a subset of p97 cofactors.


Valosin-containing protein/p97 is an ATP-driven protein segregase that cooperates with distinct protein cofactors to control various aspects of cellular homeostasis. Mutations at the interface between the regulatory N-domain and the first of two
ATPase domains (D1 and D2) deregulate the ATPase activity and cause a multisystem degenerative disorder, inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia/amyotrophic lateral sclerosis. Intriguingly, the mutations affect only a subset of p97-mediated pathways correlating with unbalanced cofactor interactions and most prominently compromised binding of the ubiquitin regulatory X domain-containing protein 1 (UBXD1) cofactor during endolysosomal sorting of caveolin-1. However, how the mutations impinge on the p97-cofactor interplay is unclear so far. In cellbased endosomal localization studies, we identified a critical role of the N-terminal region of UBXD1 (UBXD1-N). Biophysical studies using NMR and CD spectroscopy revealed that UBXD1-N can be classified as intrinsically disordered. NMR titration experiments confirmed a valosin-containing protein/ p97 interaction motif and identified a second binding site at helices 1 and 2 of UBXD1-N as binding interfaces for p97. In reverse titration experiments, we identified two distant epitopes on the p97 N-domain that include disease-associated residues and an additional interaction between UBXD1-N and the D1D2 barrel of p97 that was confirmed by fluorescence anisotropy. Functionally, binding of UBXD1-N to p97 led to a reduction of ATPase activity and partial protection from proteolysis. These findings indicate that UBXD1-N intercalates into the p97-ND1 interface, thereby modulating interdomain communication of p97 domains and its activity with relevance for disease pathogenesis. We propose that the polyvalent binding mode characterized for UBXD1-N is a more general principle that defines a subset of p97 cofactors.
Valosin-containing protein (VCP) 3 /p97 (also called Cdc48) is an abundant "ATPases associated with diverse cellular activities" (AAA)-type ATPase, which is involved in many pathways that ensure cellular homeostasis (1,2). These functions include proteasome-mediated degradation of misfolded or damaged proteins, key signaling pathways, and endolysosomal sorting and autophagy (3)(4)(5). To mediate the diverse functions, p97 cooperates with distinct cofactor proteins in different processes (1,6,7). Around 30 cofactors have been identified that bind p97 via specific interaction domains (UBX, UBX-like, PUB, and PUL) or linear motifs (VIM, SHP box, and VCP-binding motif) and serve as substrate adapters, targeting factors, or modulators of p97 activity (4,6,7). p97 is considered a "protein segregase" that structurally reorganizes proteins and separates them from binding partners or cellular structures (3,8). It acts as a macromolecular complex with its protomers building up a hexameric barrel structure. The two ATPase domains (D1 and D2) form two concentric rings, and the peripheral N-domain is coplanar with or slightly above the D1 domain, depending on the nucleotide state of the ATPase domains (9,10). Substrate adapters and other cofactors bind either to the N-domain or the C-terminal tail of p97, depending on their binding domains (7). How force is transmitted onto substrate proteins is unclear so far, but it involves movements of the N-domains relative to the D1 ring. N-domain motions are coupled to ATP hydrolysis in D2 through a chain of conformational changes that are transmitted along the D1D2 linker, the D1 domain, and the ND1 linker (9,11).
Whereas VCP/p97 knock-out is embryonic lethal, missense mutations in the ND1 interface cause systemic late onset degenerative disorders in humans called inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia/amyotrophic lateral sclerosis that affect muscle, bone, and cortical and motor neurons (12,13). The exact pathogenesis is unclear so far, but the prominent feature of affected tissues and tissue culture cell models includes defects in the endolysosomal system with impaired endosomal trafficking of caveolin-1 (CAV1) and autophagosome maturation (4,5,14,15). Intriguingly, disease-associated mutations all cluster at the ND1 interface, consistent with deregulated ATPase activity (11, 16 -18). Moreover, the mutations cause unbalanced cofactor interaction including the loss of UBXD1 binding (14,19). As UBXD1 is involved in p97-mediated endosomal sorting of CAV1 and affected by p97 mutations (14), UBXD1 is likely directly connected to the pathogenesis. Additionally, UBXD1 is unusual among the p97 cofactors as it binds p97 in two regions, at the p97 C terminus via a PUB domain and at the p97 N-domain with a short linear interaction motif termed VIM; an additional UBX domain, which usually binds the N-domain of p97, is not functional (20).
During our cell biological studies of respective binding domains, we noticed that additional moieties in the N-terminal region of UBXD1 (UBXD1-N) are involved in targeting UBXD1 to p97 on CAV1-containing endosomes. We therefore analyzed the solution structure of UBXD1-N and its interaction with p97. We discovered a yet unappreciated second binding site for UBXD1-N on the p97 N-domain including the ND1 linker. Moreover, we found that UBXD1-N also binds the D1D2 barrel. Interestingly, we observed a modulation of the p97 ATPase activity by UBXD1-N, indicating that UBXD1-N fixes the N-domain of p97 (p97-N) in a "down" position and concomitantly favors a more compact conformation for p97. Among the residues in the N-domain of p97 targeted by UBXD1-N are critical amino acids frequently affected by disease-associated mutations, suggesting that the specific binding mode of the UBXD1 cofactor to p97 may explain the compromised function of the p97⅐UBXD1 complex in the context of the disease.

Experimental Procedures
Cell-based Experiments-The human CAV1-HA and UBXD1-GFP constructs were described elsewhere (14,21,22). pEGFP-UBXD1-5xPUB-GFP was generated by QuikChange mutagenesis, resulting in the following codon changes: N184A, K193A, Y194A, K198A, and N201A in the PUB domain as described elsewhere (20). pEGFP-UBXD1-RL-GFP was mutated to code for changes R62A/L63A in the VIM as defined by Stapf et al. (23). The constructs coding for the different variants of UBXD1-GFP were cotransfected with the CAV1-HA construct in U2OS parental cells using the JetPrime reagent (Polyplus, Illkirch, France) according to the manufacturer's protocol. After 24 h of co-expression, cells were analyzed by immunofluorescence staining or Western blotting. Cells were lysed for analysis in Western blotting as described elsewhere (14). Antibodies used were anti-HA (Sigma), anti-GFP (Roche Applied Science), and anti-␣-tubulin (Sigma). For immunofluorescence staining, cells were fixed in 4% paraformaldehyde for 20 min, immunostained with an anti-HA antibody (Covance, Munster, Germany) to detect CAV1-HA, and visualized by epifluorescence microscopy. Images were taken on a Nikon Eclipse Ti microscope with the Andor DR-328G-C01-SIL camera. Image acquisition was driven by Andor IQ software (scale bar, 10 m).
Expression of Recombinant Proteins and Cell Lysis-Expression constructs were transformed into Escherichia coli BL21(DE3)T1r (Sigma-Aldrich). 100 ml of overnight cultures of transformed bacteria were resuspended in 4 liters of LB medium and grown to an A 600 of 0.8. For isotope labeling ( 15 N and/or 13 C) studies, 1 liter of LB medium was inoculated with the overnight culture and grown to an A 600 of 0.8. Subsequently, cells were pelleted by centrifugation (3,000 ϫ g, 15 min, room temperature), resuspended in 4 liters of M9 medium supplemented with 1 g/liter [ 15 N]ammonium chloride and/or 4 g/liter [ 13 C]glucose, and again grown to an A 600 of 0.8. After induction of protein expression with 250 M isopropyl ␤-D-1-thiogalactopyranoside, cells were shaken further for 6 h at 30°C for UBXD1 constructs and overnight at 30°C for p97 constructs. For cell lysis, cells were centrifuged (5,500 ϫ g, 20 min, 4°C), resuspended in low salt buffer (20 mM Tris-HCl, 50 mM NaCl, 1 mM PMSF, pH 7.4), and frozen. After a freezing cycle, cells were incubated with 300 mg of lysozyme for 1 h at 4°C followed by sonication. The cell lysate was ultracentrifuged (90,000 ϫ g, 70 min, 4°C), and the filtrated supernatant was used for purification.
Purification of Recombinant UBXD1 and p97 Constructs-The protein mixture containing GST-tagged UBXD1-N or p97-D1D2 protein was applied to a glutathione column (Macherey and Nagel, Dueren, Germany) in buffer 1 (20 mM Tris, 150 mM NaCl, pH 7.4) and eluted with buffer 1 additionally containing 20 mM glutathione. The GST tag was removed by PreScission protease. The UBXD1 constructs were separated from the GST tag by size exclusion chromatography (Superdex 75 26/600 column, GE Healthcare) combined with a glutathione column in phosphate buffer (50 mM KP i , 150 mM KCl, pH 6.5). The p97-D1D2 protein was separated from cleaved GST tag by filtration with a Centricon filter unit (30,000 molecular weight cutoff, Amicon, Millipore, Darmstadt, Germany). The protein mixture containing His 6 -tagged p97-N was applied to an anion exchange column (HiTrap Q HP, GE Healthcare) in buffer 2 (20 mM Bis-Tris, pH 6.6) and eluted with buffer 2 additionally containing 500 mM NaCl. Subsequently, the eluted protein was loaded onto a nickel-nitrilotriacetic acid column (Macherey and Nagel) in buffer 3 (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, pH 8.0) and eluted with buffer 3 supplemented with 500 mM imidazole by a gradient. For NMR spectroscopy, proteins were dialyzed against 50 mM KP i buffer, pH 6.5 for UBXD1 constructs and pH 7.2 for p97 constructs, as a compromise between high ionic strength (necessary to mimic physiological conditions) and the technical requirements for spectroscopic analyses (circular dichroism (CD), NMR, and fluorescence anisotropy).
NMR Spectroscopy-NMR experiments were performed on a 700-MHz Ultrashield NMR spectrometer (Bruker, Ettlingen, Germany) equipped with a cryoprobe (Bruker Biospin). Samples were dissolved in 600 l (90% H 2 O, 10% D 2 O) of phosphate buffer. Standard triple resonance spectra (C␤C␣CONH/ HNC␣C␤) were used for backbone assignment of UBXD1-N (Biological Magnetic Resonance Data Bank entry 26664) and for the complementation of the assignment of p97-N (25). According to MALDI-TOF data of UBXD1-N, up to 16 C-terminal amino acids are proteolytically degraded during purification (data not shown) and could not be observed concomitantly. A contiguous stretch between Ala 31 and Glu 34 , which resides within a flexible loop, also could not be detected in the spectra. 1 H-15 N heteronuclear single quantum coherence (HSQC) spectra for titration experiments were recorded with 24 scans and 2,048 ϫ 256 data points for p97-N and six scans and 2,048 ϫ 256 data points for UBXD1-N. Data were processed with Topspin 3.0 (Bruker). Titration experiments were performed at 25°C for p97-N and 27°C for UBXD1-N by stepwise addition of different concentrations of binding partner as indicated. Intensity changes of signals were analyzed, and 1 H and 15 N shifts were combined according to the following equation (26).
Hydrogen bonds were estimated from a modified D 2 O exchange experiment with UBXD1-N constructs (amino acids 1-80 and 32-133). Because freeze-drying is not possible for intact UBXD1-N, the sample was concentrated using a Centricon filter unit, and buffer was repetitively exchanged with 100% D 2 O. For analysis of flexibility, hetero-steady-state NOE (het-NOE) spectra for UBXD1-N were recorded with the corresponding pulse program from the Bruker pulse program library. The ( 1 H)-15 N NOEs were calculated from ratios of intensities of cross-peaks in spectra measured with and without saturation.
CD Spectroscopy-CD experiments were performed on a J-170 spectrometer (Jasco). Proteins were diluted up to 0.15 mg/ml in phosphate buffer (50 mM KP i , pH 6.5). CD spectra were recorded with 30 scans at 25°C subtracted by the blank (buffer without protein). The content of secondary structure was analyzed using the CDSSTR tool with the reference set SMP56 (27)(28)(29).
ATPase Activity Assay-The ATPase activity of p97 was measured with an NADH-coupled ATPase activity assay as described elsewhere (11). 500 nM p97 protein (monomeric form) was incubated with several concentrations of cofactors as indicated in 100 mM Tris-HCl at 37°C. The decrease in the NADH absorption at 340 nm due to ATPase activity of p97 was measured on a Cary 100 Bio UV/visible spectrometer (Agilent Technologies) in real time.
Limited Proteolysis-The limited proteolysis of p97 was performed with trypsin protease (molar ratio, 1:100) in the absence and presence of several cofactors. p47 is known to hold p97-N in an upper conformation, and ATP␥S fixes the N-domain of p97 more above the D1 ring. Additionally, we tested UBXD1-N(1-133), which contains the p97-N binding sites in helix (H) 1/H2 and the VIM, as well as the modified construct UBXD1(32-133) R62A/L63A , which lacks both binding sites as a p97-N non-binding control. The measurements were per-formed in phosphate buffer (50 mM KP i , pH 6.5) as described elsewhere (11). After a 2-h incubation at 37°C, each sample was mixed with SDS loading buffer and loaded for SDS-PAGE.

Sequence Elements of UBXD1-N Outside of the VIM Are
Required to Functionally Cooperate with p97 at Endosomes-When endocytic invaginations (called caveolae) are destabilized, their major constituent, CAV1, is targeted to lysosomes for degradation (21). To mediate the process, p97 and UBXD1 are recruited to CAV1-containing endosomes, which can be stimulated by CAV1 overexpression (14,21,22). To analyze the structural determinants that govern targeting of UBXD1 to p97-regulated endosomes, we transiently overexpressed CAV1-HA with diverse variants of UBXD1-GFP (Fig. 1, A for summary schematic and B for expression analysis). Besides UBXD1 wild type and truncations of the N-terminal region, variants harboring two point mutations in the VIM (R62A/ L63A) or five point mutations in the PUB domain (5xPUB), which were previously shown to affect respective binding activities, were investigated (20,23). CAV1-HA was detected by immunofluorescence with HA antibodies, and colocalization of CAV1-HA and UBXD1-GFP was visualized by epifluorescence microscopy (Fig. 1C). UBXD1 wild type largely colocalized with CAV1 endosomes. Mutation of either the PUB domain (5xPUB) or the VIM (R62A/L63A) reduced, but did not abolish, endosome targeting. Even when both sets of mutations were combined, UBXD1 was still detectable on CAV1-positive endosomes. UBXD1 recruitment was only abolished when the PUB domain mutations were combined with full truncation of the N-terminal region covering amino acids 1-133 harboring the VIM. These data suggest a functional significance of sequence elements in UBXD1-N flanking the VIM for localization to CAV1-positive structures and indicate that they may be involved in the functional cooperation with p97.
The Isolated UBXD1-N Exhibits Four Helical Regions-Based on UBXD1 localization studies and the lack of structural studies of the whole N terminus (residues 1-154; Ref. 20), we focused on the structure of this moiety of UBXD1 (UBXD1-N). Considering the importance of the VIM (residues 52-63) in p97 binding and the N-terminal residues 1-133 in endosome recruitment, we cloned and expressed the full N-terminal domain (UBXD-N(1-154)), UBXD1-N(1-133), and a minimal construct (UBXD1-N(1-80)) including the VIM. We used the isolated UBXD1-N as a simplified but valuable representative for studying the structure and dynamics of the N-terminal region in UBXD1. The 1 H-15 N HSQC NMR spectrum of UBXD1-N(1-133) ( Fig. 2A) showed 109 NH signals, most of which were dispersed over less than 1 ppm along the 1 H dimension (ϳ7.8 -8.6 ppm). Neither extension of the N terminus of UBXD1 to Val 154 nor reduction to Arg 80 had any apparent effect on signal dispersion, indicating that all three constructs share a common core structure. As the C-terminal residues following Arg 80 FIGURE 1. Sequence elements of UBXD1-N outside of the VIM mediate UBXD1 targeting to p97-regulated endosomes. A, schematic of UBXD1 constructs used for cell assays. To monitor UBXD1 localization in U2OS cells, cells were co-transfected with CAV1-HA and different GFP fusions of UBXD1 as indicated. The asterisks represent mutations. B, Western blot analysis of expression levels of GFP-UBXD1 variants. Indicated constructs were transiently co-expressed with HA-CAV1 in U2OS cells. Cell lysates (15 g) were analyzed with the indicated antibodies. ␣-Tubulin served as a loading control. C, overexpression of CAV1-HA induces a p97-UBXD1-mediated response at endosomes. CAV1-HA was immunostained, and localization of different UBXD1 variants to CAV1-HA-positive endosomes was investigated. Note that endosome recruitment of UBXD1 was not affected by VIM (R62A/L63A (RL)) or PUB (5xPUB) mutation even in combination but was abolished by full deletion of UBXD1-N together with the PUB mutation. Scale bars, 10 m. FL, full length. DECEMBER 4, 2015 • VOLUME 290 • NUMBER 49 obviously do not influence the chemical shifts of the preceding residues 1-80, one can exclude any strong interaction between these stretches of amino acids. Based on these data and our interest in the function of UBXD1-N in endosome recruitment, we focused on the structure of UBXD1-N(1-133). Sequencebased secondary structure algorithms predicted the N terminus of UBXD1-N(1-133) to reveal a random coil fraction of 57% (42% ␣-helix and 1% ␤-sheet) (31). These predictions were also consistent with experimental data deduced from the secondary structure analysis of far-UV CD spectra of the isolated N terminus of UBXD1 (46% ␣-helix, 18% ␤-sheet, and 33% random coil; Fig. 2B). To obtain information on the core structure of the N-terminal domain, backbone atoms of UBXD1-N(1-133) were assigned to their corresponding NMR frequencies by means of HNC␣C␤ for C␣ and C␤ atoms, HNCO/HN(CA)CO for CO atoms, and 15 N total correlation spectroscopy-HSQC spectra for hydrogen atoms of the side chains. Due to the high internal dynamics of UBXD1-N, only a few medium range NOEs characterizing ␣-helices (e.g. ␣H(N) i,i ϩ 3 ; ␣H(N) i,i ϩ 4 , etc.) were found for helices H2 and H3. However, a sufficient amount of long range NOEs for structure calculation was generally not obtained from either two-dimensional or three-dimensional NOESY spectra. We refrained from calculating an NMR-based structure of UBXD1-N, although a few long range NOEs between the helical regions could be assigned (Table 1). Chemical shifts from H␣ resonances were used to predict secondary structure elements within the protein, revealing five ␣-helical regions (H1, Met 1 -Phe 5 ; H2, Ala 10 -Ser 25 ; H3, Asn 50 -Gln 65 ; H4, Asp 77 -Ala 91 ; and H5, Thr 102 -Pro 108 ) (Fig. 2C). The formation of secondary structure elements in H1, in the first part of H2, along H3 as well as partially in H4, and at the C terminus including H5 was also confirmed by a modified D 2 O exchange experiment (Fig. 2C, bars on top, H-bonds). H2 may exist as a helix-turn-helix element with the turn at Lys 15 -Ala 17 , and H4 is only poorly defined by the experimental data. As H1, H2, and H4 are predominantly defined by characteristic H␣ shift values, all three helices are considered to be transient helices. The known VIM (Ala 52 -Arg 62 ) of UBXD1-N is a part of the well defined H3. Despite the transient character of most of the helices, the mutational exchange of amino acids D11A/I12A within H2 led to significant chemical shift changes of 1 H-15 N HSQC signals belonging to amide groups of residues of H1 as well as H3/loop (L) 3 (Fig. 2D). Corresponding chemical shift changes of signals of residues from H1/H2 were observed upon exchange of amino acids R62A/L63A in H3 (Fig. 2E), revealing an interaction between these secondary structure elements. Because NMR-based structure calculation and homology modeling were not feasible, we used the information obtained from experimental data (analysis of shifts and mutational studies) to generate an ab initio model for UBXD1-N (Fig. 2F). Although the model displays interactions of helical elements within UBXD1-N, it implies a very dynamic structure, which was confirmed when retrieving information on backbone dynamics of UBXD1-N from the hetNOE of the amide resonances (Fig. 2G). For rigid tumbling proteins of about 14 kDa, these hetNOE values are higher than 0.6. Low hetNOE values are correlated with high internal flexibility. About 70% of the NH groups of UBXD1-N revealed negative values, indicating high internal mobility and, hence, flexibility of the backbone. Positive values were found for H2, H3 containing the VIM, L3, and H4. Although the positive values correlated with most of the defined partially transient secondary structural elements and point toward restricted backbone motility, the overall hetNOE values were below 0.5, indicating that UBXD1-N is a flexible region lacking a rigid three-dimensional folding topology.

UBXD1-N Modulates p97 Interdomain Communication
UBXD1-N Binds the p97-N via Two Binding Sites-Based on our cell biological experiments, we wondered whether the flexible UBXD1-N moiety is capable of interacting with the p97-N. To investigate this interaction, p97-N was added step by step to 15 N-labeled UBXD1-N up to a molar ratio of 1:5, respectively. The corresponding chemical shift and intensity changes of the amide signals of residues of UBXD1-N were monitored by 1 H-15 N HSQC spectra. Upon titration of p97-N to UBXD1-N, 29 amide signals lost intensity, whereas another 12 signals underwent chemical shift changes, indicating a specific interaction of both proteins (Fig. 3A). The affected residues accumulated within two sequential stretches of UBXD1-N comprising amino acids from Lys 2 to Lys 30    signals belonging to this region disappeared in the 1 H-15 N HSQC spectrum upon addition of p97-N, whereas residues Ala 57 , Ala 70 , and Trp 71 showed only marginal chemical shifts changes. These observations suggest the direct involvement of the VIM in binding the N-domain of p97. Moreover, the stretch of interacting residues (27 amino acids) in UBXD1-N was much more extended compared with the general definition of the 10-amino acid containing VIM based on bioinformatics analysis of VIM-containing p97 cofactors (23). Most of the signals with decreased intensities belong to residues of H3 containing the VIM, whereas signals with only chemical shift changes are concentrated in H1/H2. Signal intensity loss indicates that the dynamics of complex formation is within the intermediate to slow exchange regime on the NMR time scale, pointing to high affinity binding constants of corresponding binding partners (K D values in the lower M to nM range). Changes in chemical shifts indicate that the exchange between bound and unbound conformations is within the fast exchange regime on the NMR time scale, which usually belongs to low affinity interactions (K D values in the higher M to mM range). Therefore, we concluded that the VIM and flanking residues (disappearing signals already at molar ratios of 0.2:1) bind p97-N with a higher affinity than H1/H2 (small shifts at molar ratios up to 1:5). For confirmation, the affinity of p97-N to UBXD1-N was calculated by fluorescence anisotropy to be 9 M, which is in accordance with published data (32) (Fig. 3B). The affinity for the newly identified binding region of H1/H2 was investigated using the UBXD1-N mutant R62A/L63A, which is known to interrupt the interaction between the VIM and p97-N (23). Fluorescence anisotropy studies revealed that H1/H2 bound p97-N with a K D of about 580 M independently of the VIM (Fig. 3B). This is in accordance with the observation that only signals of residues belonging to H1/H2 decreased in their intensity and/or showed chemical shift changes in a corresponding 1 H-15 N HSQC titration experiment using the UBXD1-N R62A/L63A mutant (Fig.  3C). Hence, our results confirm the presence of a high affinity binding motif (VIM) within UBXD1-N, but moreover, they provide evidence for the existence of an additional low affinity binding site.
To identify the two UBXD1-N binding epitopes on p97-N, several concentrations of UBXD1-N(1-80) (to keep the molecular weight of the binding complex as low as possible) were added gradually to 15 N-labeled p97-N up to a molar ratio of 1:5, respectively. The corresponding chemical shift changes of the amide signals of p97-N were monitored by 1 H-15 N HSQC spectra (50 mM KP i , pH 7.2). Upon binding of UBXD1-N, the intensities or chemical shifts of 48 NH signals changed. The corresponding residues were mapped onto the surface of the p97-N crystal structure including the ND1 linker (extracted from Protein Data Bank code 1E32) to visualize the binding areas (Fig.  3D). Two binding interfaces could be identified. The first interface comprises residues from the hydrophobic VIM-binding groove between the two subdomains of p97-N and neighboring residues (Val 38 -Gln 43 , Gly 54 , Thr 56 , Val 88 , Val 87 , Tyr 138 -Tyr 143 , Ile 151 -Arg 155 , Ile 175 -Ile 182 , and Cys 105 -Gly 111 ). The second interface is composed of the linker region connecting the N-domain to the D1 domain of p97 and of amino acids located in close proximity on the surface of p97-N to this linker (Arg 93 , Leu 117 , Gly 156 , and Cys 184 -Glu 195 ). To specify which interface of p97-N is bound either by H1/H2 or by the VIM of UBXD1-N, the UBXD1-N R62A/L63A mutant was added to 15 Nlabeled p97-N up to a molar ratio of 1:5, respectively. Upon binding of UBXD1-N R62A/L63A , the intensities and chemical shifts of 26 NH signals were perturbed in the corresponding 1 H-15 N HSQC spectra. Most of the amino acids reside within the ND1 linker or within its neighborhood on the surface of p97-N (Gln 50 , Lys 60 , Ser 73 , Arg 93 , Ser 101 , Leu 117 , Gly 156 /Gly 157 , Cys 184 , Lys 190 , and Glu 195 as well as Ile 119 , Asn 130 , and Asn 199 , which are not bound by the wild type; Fig. 3E). Additionally, some amino acids were affected that are not clustered and are spread over the surface (Val 39 -Gln 43 , Asp 55 , Leu 58 , Cys 105 , Leu 140 , Tyr 143 , and Thr 168 ). Thus, we conclude that the ND1 linker of p97 is preferably the binding region for H1/H2 of UBXD1-N, whereas the VIM binds the VIM groove between the two subdomains of p97-N.
The binding studies described above revealed the involvement of residues within p97-N that are mutated in patients with inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia (Arg 93 , Arg 155 , and Arg 191 ). To investigate the influence of these mutations on binding of UBXD1-N to p97-N, increasing amounts of UBXD1-N(1-80) wild type were added to a molar ratio up to 1:1 to 15 N-labeled mutant p97-N R155H (Fig. 3F) as well as 15 N-labeled mutant p97-N R95G (Fig. 3G). However, the unspecific disappearance of signals in both spectra and an instant precipitate of p97-N R155H and 1 day later of p97-N R95G occurring within the sample suggest unspecific interaction and subsequent aggregation of the corresponding complexes, which precluded further analysis.
UBXD1-N Interacts with the D1D2 Barrel of p97-In addition to the binding sites observed for p97-N in the N terminus of UBXD1, UBXD1-N also bound to the D1D2 barrel of p97 with an affinity of 270 M as demonstrated by fluorescence anisotropy measurements (Fig. 4A). Intriguingly, the binding of UBXD1-N to the p97-D1D2 barrel seems to be nucleotide-independent as the presence of ATP does not have an effect on the K D value (270 Ϯ 63 M for ADP-bound versus 251 Ϯ 53 M for ATP-bound). An interaction between cofactors and the D1D2 domain of p97 has not been described so far. To study this interaction further, several concentrations of soluble p97-D1D2 protein were added to 15 N-labeled UBXD1-N up to a molar ratio of 1:1, respectively. The corresponding chemical shift and intensity changes of the amide signals of UBXD1-N were monitored by 1 H-15 N HSQC spectra (Fig. 4B). Upon titration of p97-D1D2 to UBXD1-N, the intensity of 38 amide signals decreased significantly, indicating a specific interaction of both proteins. The interacting residues are clearly concentrated in two stretches of UBXD1-N. Thereby, p97-D1D2 binds preferably H1/L1 (Phe 4 , Phe 5 , Gln 7 , Phe 8 , Asp 11 -Phe 14 , and Ala 17 ) and H3/H4 (Glu 51 -Gln 53 , Ala 55 -Gln 65 , Gln 67 , Arg 69 -Gly 72 , Thr 78 -Arg 80 , Val 83 -Glu 86 , and Gln 88 -Glu 90 ) including the known VIM (Ala 52 -Arg 62 ) and its flanking residues. In comparison with p97-N, p97-D1D2 binds the N-terminal region of H1/H2 (H1/H2 Met 1 -Ser26 and 9 affected residues within Phe 4 -Ala 17 ), whereas p97-N is primarily bound to H2 (H2 Ala 10 -Ser 26 and 6 disappearing signals between Phe 14 and Glu 25 ). As shown, the VIM interacts with the N-domain as well DECEMBER 4, 2015 • VOLUME 290 • NUMBER 49 as with the D1D2 barrel of p97, but p97-N preferably binds the N-terminal region of the VIM, whereas the D1D2 barrel interacts with the C-terminal part of the VIM (Fig. 4C), suggesting a different binding mechanism of UBXD1-N to the N-domain and the D1D2 barrel of p97. This hypothesis is also in line with fluorescence anisotropy measurements of a UBXD1 mutant lacking a functional VIM (Fig. 4D). Although the binding of UBXD1-N R62A/L63A to p97-N is significantly reduced due to the mutation, it remains similar to the p97-D1D2 construct.

UBXD1-N Modulates p97 Interdomain Communication
UBXD1-N Inhibits the ATPase Activity of p97 and Fixes the N Domain of p97 in an Altered Conformation When Compared with p47-Our studies indicate that the isolated N terminus of UBXD1 binds in a similar region on p97-N as the cofactor p47. Because p47 is known to inhibit the p97 ATPase activity (33), we tested the ATPase activity of p97 in the presence of UBXD1-N using an NADH-coupled ATPase assay (Fig. 5A). UBXD1-N reduced the ATPase activity of p97 in a concentration-dependent manner to 7.27 mol of ATP/mg of p97/h (41%; K i ϭ 8 nM) similar to p47 (reduction to 5.59 mol of ATP/mg of p97/h; 31%; K i ϭ 7 nM). In contrast, the addition of Tris buffer did not have any significant effect on the p97 ATPase activity.
For the p97⅐p47 complex, EM-based structure analysis indicated that the N-domain of p97 is held in an "up" conformation by p47 (34). Because our data indicated that the UBXD1-N interacts with both the N-domain and the D1D2 domain of p97, we wondered whether UBXD1-N constrains p97-N in the same conformation. To investigate various conformations caused by either p47 or UBXD1-N, p97 was degraded by limited proteolysis using trypsin (Fig. 5, B and for time courses C).  (34), the corresponding conformational changes made the linker between the domains of p97 accessible for enzymatic proteolysis by trypsin. Only bands corresponding to the rigid and protected domains of p97 are left. The band for p47 below 55 kDa disappears due to partial degradation by trypsin, whereas another band at 25 kDa appears. LC-MS/MS analysis identified this band as parts of p47 containing the Shp1-eyc-p47 domain (SEP) as well as the p97-N binding UBX domain (data not shown). A modified UBXD1-N construct (amino acids 32-133, R62A/L63A), which does not bind the VIM groove of p97-N or the ND1 linker but still interacts with p97-D1D2, fails to protect p97 against proteolysis (lane 6). In this case, a similar band pattern is observed as in the presence of p47 or ATP␥S (lanes 3 and 4). Only UBXD1-N WT , which binds p97-N as well as p97-D1D2, is able to partially protect full-length p97 from proteolysis by trypsin, resulting in a different band pattern (100, 70, and 20 kDa; lane 5). The different band pattern of p97 after limited proteolysis in the presence of UBXD1-N or p47 indicates an altered conformation of the p97 hexamer. Additionally, temperature stability experiments confirmed a higher melting temperature of p97 in the presence of UBXD1-N (Ͼ88°C) in comparison with that in its absence (68°C) or in the presence of its substrate analogue ATP␥S (71°C) (Fig. 5D). Due to the high Trp content of p47 (13 Trp residues) compared with p97 (3 Trp residues), it was not possible to perform the thermal stability assay in the presence of the cofactor p47. The altered proteolysis pattern and the high

UBXD1-N Modulates p97 Interdomain Communication
increase of the p97 melting temperature in the presence of UBXD1-N indicate that UBXD1-N fixes p97-N in a conformation that is different from the p47-related up conformation. This altered conformation seems to restrict the flexibility of the N-domain and the linker regions of p97 required for the p97 ATPase activity, explaining the observed influence of UBXD1-N on the p97 ATPase activity.

Discussion
Among the p97 cofactors, UBXD1 is particular because it has a complex interaction mode with p97 that is specifically affected by disease-associated mutations in p97 (14,23). In this study, we report detailed data on the structure and binding mechanism of the critical isolated UBXD1-N moiety. Our NMR and CD spectroscopic analyses of UBXD1-N have elucidated the absence of a stable tertiary fold and the presence of stretches of mainly transient secondary structural elements with exception of the VIM. Structural predictions using the Predictor of Natural Disordered Regions algorithm (Fig. 6A) support the notion that even in full-length UBXD1 no rigid three-dimensional topology is present for the N-terminal region. The high number of disorder-supporting residues (68%; namely Ala, Arg, Gly, Gln, Ser, Glu, Lys, and Pro (35,36)) and the presence of a considerable number of charged amino acids (26%) argues for an enhanced hydrodynamic radius of UBXD1-N, leading to the short elution time in size exclusion chromatography. Taken together, these features are consistent with the interpretation that the N terminus of UBXD1 can be classified as an overall intrinsically disordered region of UBXD1. Intrinsically disordered proteins have attracted much attention during the last two decades, and thus their general regulatory role in signaling and controlling pathways is widely accepted (37)(38)(39)(40). The unique binding properties and the regulative potential of intrinsically disordered regions in proteins were recently discovered to appear also in cofactors of p97 (41)(42)(43). Moreover, proteins that are involved in p97-related pathways such as endoplasmic reticulum-associated protein degradation (44) and the ubiquitination system for protein degradation also contain unstructured amino acid stretches with regulative roles (45).
The flexible N-terminal region of UBXD1 contains the well defined ␣-helical VIM, which is known to bind the groove between the two subdomains of p97-N (23,32). This interaction was also confirmed by our NMR spectroscopic studies at the atomic level. However, consistent with our cell-based assays monitoring UBXD1 localization to p97 on CAV1 endosomes, our structural studies revealed a more complex binding mechanism than noted previously (23). We discovered a new interaction site on H1/H2 in UBXD1-N that binds the ND1 linker of p97 independently from the VIM. The H1/H2 binding region is ␣-helical and conserved among several species but unique for the UBXD1 cofactor. UBXD1 has recently been shown to also bind a population of ERGIC-53 at the plasma membrane with the very N-terminal amino acids (46). Intriguingly, this region overlaps with the H1/H2 binding site for p97, and therefore it would be interesting to clarify whether these differential interactions of UBXD1 represent alternative functions or whether they are functionally linked. Additionally, both binding regions of UBXD1-N are also involved in the interaction with the D1D2 barrel of p97. The binding of p97-N and -D1D2 seems to occur simultaneously because there is only a partial overlap of binding epitopes on UBXD1-N. Although we were not able to identify the binding interface on p97-D1D2 exactly, we assume that binding occurs to the D1 domain as UBXD1-N binds residues of the interface between the N-domain and D1 domain as well as the linker between the N-domain and the D1 domain of p97 (Fig. 6B).
The intrinsically disordered region characteristic of UBXD1-N and the absence of a defined tertiary structure may affect the ability of the protein to properly fit into the polyepitopic interface region on the p97 protein. In fact, we suggest that the intrinsic flexibility of UBXD1-N may be a structural requirement for targeting the VIM groove of p97-N while simultaneously grabbing the flexible ND1 linker and the D1D2 barrel at a very far distance. Of note, in a number of cofactors, short p97-binding motifs lie in regions without defined domain structure for example in UBE4B, small VCP-interacting protein, VCP-interacting membrane protein, Hrd1, and autocrine motility factor receptor (43,(47)(48)(49)(50). Moreover, the VIM protein small VCP-interacting protein was also shown to bind the ND1 linker (32,41). Based on our new data on UBXD1-N, we therefore propose that polyvalent binding of intrinsically disordered regions containing short p97-binding motifs is a more general feature of a number of p97 cofactors and thereby defines a subclass of cofactors.
Two additional observations in this study are consistent with the notion that UBXD1-N intercalates in a complex manner and possibly affects the flexibility of the N-domain. First, we show that UBXD1-N binding to p97 changes the protease sensitivity of p97 in comparison with p47. These changes can either occur from protection of proteolysis sites or from domain reorientation. We assume that the changed protease pattern results from domain reorientation because the proteolysis pattern of free p97 does not change upon binding of the cofactor p47, although UBXD1-N and p47 share a similar binding site on p97-N. p47, however, is known to lock p97-N in an up conformation. Considering the additional binding site of UBXD1-N on the D1D2 barrel and the fact that only two extreme positions of the p97 N-domain are realized within the hexamer, it is reasonable to assume that UBXD1-N locks p97-N in a "down" conformation toward the D1D2 barrel. Another indication for a compaction of p97⅐UBXD1-N is the UBXD1-N-mediated melting temperature increase of p97 (about 20°C). Because a long intrinsically disordered protein chain should lead to a strong positive ϪT⅐⌬S contribution to the Gibbs free enthalpy, such an unfavorable contribution can be overcompensated by additional new binding sites formed in the presence of UBXD1-N by intradomain/domain interaction as postulated to occur upon formation of the down conformation. Second and connected to the first, UBXD1-N binding to p97 inhibits its ATPase activity. Previous reports have established that p97-N motion relative to the D1 plane is coupled to ATPase activity and that locking the N-domain in a down position inhibits ATPase activity (11). Moreover, the contact between the N-domain and the D1 domain is important for the allosteric communication between substrate adaptor binding and ATPase domains (9,11,51). We therefore speculate that binding of UBXD1-N into the p97-ND1 interface may prevent p97-N motion and thus inhibit ATPase activity (Fig. 6B). An analogous explanation has been suggested for the inhibitory effect of the p47 cofactor on the ATPase activity; only in this case, p47 locks the N-domain in an up position (33,34). This, in turn, is consistent with the differential effect on protease sensitivity compared with UBXD1-N in our experiments. Following this idea of the intrinsically disordered region being a prerequisite for tethering the p97 domains and fixing them in a compact formation, the intrinsically disordered region characteristics are consecutively responsible for the decrease of p97 ATPase activity. Previously, N-terminally tagged full-length UBXD1 was reported to have no effect on p97 ATPase activity (52). We assume that the isolated UBXD1-N moiety is decoupled from regulatory elements of the protein and that pausing p97-N motions can only be a transient step in the context of the p97-mediated substrate protein remodeling reaction. Conversely, the specific inhibition of the ATPase activity of p97 by the N terminus of UBXD1 supports the notion that the intrinsic features of UBXD1-N also apply in the context of the full-length protein. Although the substoichiometric inhibition of p97 by UBXD1-N is quite unusual, such phenomena have already been described for oligomerizing protomers (53)(54)(55)(56)(57). Therefore, we hypothesize a kind of sustainability of the UBXD1-N interaction with the p97 hexamer. In any case, the specific mode of interaction between UBXD1-N and p97 points to a distinct modulation of the p97mediated segregation mechanism particularly required in UBXD1-regulated processes as opposed to p47-or Ufd1-Npl4regulated reactions (1,58).
The binding mechanism of UBXD1-N to p97 may also be relevant for p97-associated diseases. Disease-causing mutations all lie at the critical interface between the p97 N-domain and D1 domain and are believed to affect their allosteric communication (4). Crucially, we show here that UBXD1-N binding occurs at this interface and even involves residues mutated in the disease. Moreover, we showed earlier that binding of UBXD1, but not p47 or Ufd1-Npl4, which have different binding epitopes on p97, to disease-associated p97 mutants is specifically reduced (14). Consistently, compromised sorting of caveolin-1 is a dominant feature in cells expressing disease- involving the D1 domain and the ND1 linker of p97. Insertion of UBXD1-N into the ND1 interface is mediated by interaction with the N-domain, the ND1 linker, and parts of the D1D2 barrel of p97. As a result, the flexibility of the N-domain is restricted, and the ATPase activity of p97 is decreased. This mechanism may be transient in the context of a complex p97-mediated segregation reaction. The binding sites involve residues mutated in p97-associated disease, suggesting that effects on UBXD1-N binding may form the basis for the disease pathogenesis. aa, amino acid. DECEMBER 4, 2015 • VOLUME 290 • NUMBER 49 associated mutants of p97 (14). We therefore speculate that the intimate intercalation of UBXD1-N into the p97-ND1 interface constitutes the specific structural basis for the pathogenic mechanism observed in the case of p97 disease-associated mutations.

UBXD1-N Modulates p97 Interdomain Communication
From our structural analysis of the isolated N terminus of UBXD1 and further biophysical binding studies with single domains of p97, we conclude that the UBXD1-N engages in polyvalent binding to several regions within the ND1 domain interface of p97. Our findings indicate that UBXD1-N intercalates into the p97-ND1 interface, thereby modulating the interdomain communication of p97 domains and hence inhibiting its ATPase activity in a similar way to p47. We propose that such a polyvalent binding mode is facilitated by the fact that UBXD1-N is highly flexible and/or intrinsically disordered and that this is a more general principle that defines a new subset of p97 cofactors containing short p97-binding motifs in regulatory disordered regions. Moreover, the facts that UBXD1-N interaction sites include those residues frequently mutated in patients with inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia and that UBXD1 binding is affected in the disease model suggest that this particular binding mode also may have relevance for disease pathogenesis.
Author Contributions-Experiments were predominantly done by F. T. Cellular experiments were conducted by M. V., L. K., and H. K. A. M. contributed to NMR experiments. Ideas and conception of the work came from F. T., P. S. F., H. M., and P. B. The manuscript was written and revised by all authors. All authors have given approval to the final version of the manuscript.