Structures of p97–UBXD1 reveal extensive hexamer remodeling

With three canonical p97 binding domains in UBXD1 (VIM, UBX and PUB), we postulated that substantial interactions with p97 may define its function. UBXD1 also contains transiently structured helices, H1/H2 (residues 1–25) and H4 (residues 75–93), that weakly interact with p97 (refs. 24,25) and a helical lariat with homology to the adaptor alveolar soft part sarcoma locus (ASPL, residues 278–334) (refs. 32,33). Analysis of the UBXD1 AlphaFold model34,35 revealed structures for the canonical domains, H1/H2, H4 and the helical lariat; the remaining sequence appears unstructured (Fig. 1a,b).

Fig. 1: Cryo-EM structures of p97–UBXD1 closed and open states.

a, Domain schematics of UBXD1 and p97 (not to scale) showing reported interactions (solid lines) between conserved domains24,27,28,29 and the UBX–NTD interaction previously reported to not occur for UBXD1 (dashed line)20,27,31. b, AlphaFold model of UBXD1 showing structured regions (H1/H2, VIM, H4, PUB, helical lariat and UBX) colored as in a. c, Steady-state ATPase activity (y axis, normalized to activity at 0 nM UBXD1) of p97 at increasing concentrations of UBXD1 (x axis), resulting in a calculated IC50 of 25 nM. Data are from n = 3 independent experiments, each with three technical replicates. Data are presented as mean values from each independent experiment. d, Representative 2D class averages following the initial classification of the full p97–UBXD1 dataset, showing the p97 hexamer and no additional density for UBXD1. Scale bar, 100 Å. e,f, Final cryo-EM reconstructions of p97–UBXD1closed (e) and p97–UBXD1open (f) states with top-view 2D projections showing UBX–PUB density (*) and open p97 ring (arrow) compared to cartoon depictions of the corresponding complexes (top row); cryo-EM density maps (p97–UBXD1open is a composite map; see Methods), colored to show the p97 hexamer (light and dark blue, with protomers labeled P1–P6) and UBXD1 density for the VIM (brown), UBX (yellow) and lariat (orange) domains (bottom row). The 8 Å separation between protomers P1 and P6 is indicated for p97–UBXD1open. g, Low-pass filtered maps and fitted models of p97–UBXD1closed (left) and p97–UBXD1open (right) exhibiting low-resolution density for the PUB domain (gray).

Source data

We first determined the effect of UBXD1 on p97 ATPase activity, finding potent inhibition of p97 with a half-maximal inhibitory concentration (IC50) of 25 nM (Fig. 1c). We next analyzed UBXD1 binding to p97 by size exclusion chromatography (SEC) and SDS–PAGE analysis. Following incubation with either ADP or the slowly hydrolyzable ATP analog ATPγS, UBXD1 co-elutes with p97, indicating a stable interaction (Extended Data Fig. 1a,b). However, based on the reduced band intensity of UBXD1 compared to p97 in both conditions, UBXD1 appears to be sub-stoichiometric with respect to p97. Based on these results and the reported structural stability of the NTDs in the ADP state36, p97–UBXD1 cryo-EM samples were first prepared with ADP (Extended Data Fig. 1c). Reference-free 2D class averages show well-defined top and side views of the p97 hexamer with the NTDs in the down conformation, co-planar with the D1 ring, but with no density attributable to UBXD1 (Fig. 1d and Extended Data Fig. 1d). Next, 3D classification was performed (Extended Data Fig. 1e), resulting in the identification of two states with prominent UBXD1 density (p97–UBXD1closed and p97–UBXD1open) that were subsequently refined (Fig. 1e,f and Table 1). Individual p97 protomers are denoted P1–P6, counterclockwise, based on the asymmetry of p97–UBXD1open. High-resolution features in both states enabled the building of atomic models, with all UBXD1 domains identified except H1/H2 and H4 (Fig. 1e,f, Extended Data Fig. 2a–d, k and Supplementary Video 1).

Table 1 Cryo-EM data collection, refinement and validation statistics

In both structures, density for ADP is present in all D1 and D2 nucleotide pockets and coordinated by conserved AAA+ residues (Extended Data Fig. 1f–h). In p97–UBXD1closed, D1 and D2 are in a planar conformation and one molecule of UBXD1 is identified, binding across protomers P1 and P6. In p97–UBXD1open, a single UBXD1 is similarly bound but the interface between P1 and P6 is separated by ~8 Å relative to p97–UBXD1closed, creating an open-ring arrangement of the p97 hexamer. For both structures, strong density corresponding to the VIM and UBX domains is observed, as well as for the helical lariat that encircles the NTD of the P6 protomer (Fig. 1e,f). At a lower density threshold, the PUB domain becomes apparent and is positioned below the UBX, adjacent to the p97 D2 ring, but appears to make no substantial contact with the hexamer, probably resulting in substantial flexibility and explaining the lower resolution (Fig. 1g). Weak density resembling the VIM is also present in the NTDs of P2–P5 in both states, indicating partial binding to these protomers, which is probably a consequence of protein concentrations used for cryo-EM (Extended Data Fig. 1i). Overall, these structures identify that a singly bound UBXD1 drives ring opening at a protomer interface that is tethered by multiple UBXD1 interactions involving the VIM, UBX and lariat motif. We identify that the UBXD1 interaction buries a surprising ~3,200 Å2 of surface area, the largest of any p97–adaptor interaction that is currently structurally characterized.

During 3D classification, additional structures with different UBXD1 configurations were identified (Extended Data Figs. 1e, 2 and 3 and Table 1). A prevalent class, termed p97–UBXD1VIM, closely resembles the previously determined structure of the p97 hexamer bound to ADP (p97ADP, Cα root mean square deviation (r.m.s.d.) of 1.0 Å)36 but features additional helical density in the NTD cleft of all protomers that corresponds to the UBXD1 VIM (Extended Data Fig. 3a). Two additional states were identified, termed p97–UBXD1meta and p97–UBXD1para, in which densities for the VIM, PUB, lariat and UBX are also observed at other positions (across P2–P3 or P3–P4, respectively) in the p97 hexamer in conformations similar to p97–UBXD1closed, indicating that other configurations can occur (Extended Data Fig. 3b,c). However, the p97 open ring is observed only in the singly bound UBXD1 configuration, indicating that this conformation is driven by binding of one UBXD1. Considering the singly bound closed and open states exhibit substantial p97 conformational changes and extensive interactions by UBXD1, these states are largely the subsequent focus of this study.

UBXD1 promotes nucleotide-dependent p97 conformational changes

To identify the effects of UBXD1 binding, p97–UBXD1closed and p97–UBXD1open were aligned to p97ADP36. The r.m.s.d. values reveal extensive changes across D1 and D2 for the seam protomers (P1 and P6) in both UBXD1-bound states, although the magnitudes are greater in the open state (Fig. 2a,b, Extended Data Fig. 4a,b and Supplementary Video 2). The open state features a modest right-handed spiral with an overall elevation change of 7 Å (Extended Data Fig. 4c), largely owing to a 9° downward rotation of P1 (Fig. 2b). In addition to movements between protomers, there is a notable rotation between the small and large subdomains of D2 in protomer P1 in both states (Extended Data Fig. 4d). This rotation is particularly evident in the open state, in which the small subdomain is rotated upward by 10° relative to p97ADP (Extended Data Fig. 4d). This rotation, and the separation of P1 and P6, causes a helix (α5′) (refs. 10,37), which is normally positioned on top of α12′ of the D2 domain of the counterclockwise protomer, to disappear from the density map, probably due to increased flexibility (Extended Data Fig. 4e and Fig. 2a,b).

Fig. 2: UBXD1-mediated p97 hexamer remodeling.

a,b, The p97 hexamer and rotated side view of the seam protomers P1 and P6 for p97–UBXD1closed (a) and p97–UBXD1open (b) structures, colored according to Cα r.m.s.d. values relative to the p97ADP symmetric state (PDB 5FTK, aligned to P3 and P4). The largest changes (>15 Å, magenta with wider tubes) are identified for α12′ of P1 (labeled) in the open state, intermediate changes (∼10 Å, red) for P1 and P6 with rotations of the NTDs relative to p97ADP shown and small to no changes for the remaining regions (<5 Å, white); α5′ of P6, which disappears in the open state, is labeled in a. c, Distribution of states for the ADP and ATPγS superstoichiometric datasets (2:1 UBXD1:p97 monomer) and the ADP substoichiometric dataset (1:6 UBXD1:p97 monomer). For each dataset, all particles after 2D classification were subjected to 3D classification using the same references. Classes corresponding to junk particles were excluded and the proportions of the remaining classes were plotted. d, Schematic of p97 remodeling in various nucleotide-bound and UBXD1-bound states.

Source data

The closed and open states each show distinct P1–P6 interfaces (Extended Data Fig. 4f). In the closed state, the D1 domains of P1 and P6 are separated by 3 Å relative to p97ADP, while there is negligible D2 separation. In the open state, the D1 and D2 domains are separated by 8 Å and 11 Å, respectively. These changes remodel the nucleotide-binding pockets of P1 by displacing the arginine fingers from P6 (Extended Data Fig. 4d), probably precluding ATP hydrolysis in P1. To further investigate the conformational changes associated with UBXD1 binding, 3D variability analysis (Methods) was performed jointly on particles from the closed and open states (Supplementary Video 3). This reveals the transition from a planar UBXD1-bound hexamer to a partial spiral, indicating that the closed and open states may be in equilibrium and on path to substrate-bound conformations6,7.

We next sought to determine structures of p97–UBXD1 with ATPγS in order to identify nucleotide-dependent differences in UBXD1 interactions. Cryo-EM analysis of p97–UBXD1 incubated with ATPγS revealed three predominant classes: a UBXD1-bound hexamer similar to the ADP-bound closed state, a symmetric hexamer lacking UBXD1 density with NTDs in the ‘up’, ATP conformation and a symmetric hexamer similar to p97–UBXD1VIM (Extended Data Fig. 5a,b). An open state was not present in the dataset, indicating that this conformation may require a post-hydrolysis ADP state of p97. Despite the unambiguous presence of ATPγS in all nucleotide-binding pockets (Extended Data Fig. 5c), the NTDs adopt the ‘down’ conformation in maps containing UBXD1 density (the closed and VIM-only states), indicating that UBXD1 interactions favor the NTD ‘down’ state and appear to override the ‘up’ conformation observed in ATP-bound structures36.

We postulate that the symmetric VIM-only class in the ADP and ATPγS datasets (Fig. 2c and Extended Data Figs. 1e and 5a) may be a consequence of our incubation conditions involving a high, 2:1 concentration of UBXD1 relative to the p97 monomer. We predict that this results in saturated occupancy of the NTDs by the VIM helix and displacement of other interactions that define the asymmetric closed and open states. Therefore, we sought to examine UBXD1-promoted rearrangements in conditions more similar to those in vivo, in which p97 is present in excess of UBXD1 (ref. 38). To that end, we incubated p97 with a substoichiometric amount of UBXD1 (1 UBXD1 per p97 hexamer) in the presence of ADP and determined the resulting distribution of complexes by cryo-EM (Extended Data Fig. 5d–f and Methods). This analysis revealed that ~75% of the particles have no UBXD1 density, as expected given the reduced UBXD1 concentration (Fig. 2c). Strikingly, the remaining 25% are in the open state with a singly bound UBXD1, indicating that under substoichiometric conditions, UBXD1 binds predominantly as one per hexamer in the ring-open conformation. These data further support the functional significance of the open conformation, wherein binding by one UBXD1 through multiple interactions breaks the p97 interprotomer interface and opens the hexamer ring. Taken together, our findings indicate that UBXD1 may function during the p97 catalytic cycle by promoting ring opening specifically in a post-hydrolysis state (Fig. 2d).

Canonical UBXD1 interactions across three p97 protomers

The p97–UBXD1closed and p97–UBXD1open structures reveal how the conserved VIM, UBX and PUB domains of UBXD1 interact simultaneously across the p97 hexamer. In contrast to previous studies20,27,31, both VIM and UBX are identified to interact with p97, binding adjacent NTDs (Fig. 3). The 18-residue VIM helix is positioned in the NTD cleft of P1, similar to structures of isolated domains, and comprises the only major contact with this protomer (Fig. 3a,b and Extended Data Fig. 6a,b). A conserved arginine residue (R62), required for p97 binding29, projects into the NTD, potentially forming a salt bridge with D35 of the NTD and a hydrogen bond with the backbone carbonyl of A142 (ref. 29). The VIM appears anchored at its N terminus by a salt bridge between E51 of UBXD1 and K109 of the NTD and by hydrogen bonding between the backbone carbonyl of E51 and Y143 of the NTD. Additional hydrophobic contacts could further stabilize this interaction (Fig. 3b, right).

Fig. 3: Interactions by conserved VIM, UBX and PUB domains of UBXD1 across the p97 hexamer.

a, Sharpened map of the P1 NTD (dark blue) and the VIM helix (brown) from p97–UBXD1closed. b, Corresponding model showing VIM helix interactions with the NTD, colored as in a, with labeled interacting residues. c, Sharpened map of the P6 NTD (light blue) and UBX domain (yellow) from p97–UBXD1closed. d,e, Corresponding model of the UBX and NTD showing a conserved orientation of the S3/S4 loop45 (arrow) (d) and non-canonical structural elements Uα2, Uα3 and Uβ0 (e), colored as in c. The N-terminal (Nn) and C-terminal (Nc) lobes of the NTD are indicated. f, Overlay of PUB domains from p97–UBXD1-PUBin (gray) and p97–UBXD1-PUBout (white), aligned to the UBX (yellow) domain, showing 46° rotation of the PUB domain position. g,h, Low-pass filtered map and model of p97–UBXD1-PUBin depicting PUB domain contact with p97 and model for C-terminal HbYX tail interaction from the adjacent P5 protomer (arrow) (g) and bottom view of the hexamer map with ‘out’ (white) and ‘in’ (gray) positions of the PUB (h). i, Cartoon of p97–UBXD1closed depicting UBXD1 interactions across three p97 protomers (P1–VIM, P6–UBX and P5–PUB) through canonical p97-interacting domains.

The UBX domain is bound to the NTD of P6 in a manner similar to that of other adaptors despite mutation of the canonical phenylalanine–proline–arginine motif8 located on the S3/S4 loop to serine–glycine–glycine (Fig. 3c,d and Extended Data Fig. 6c,d). However, another arginine residue that is important for interaction with p97, R342, is conserved in UBXD1 and probably forms a hydrogen bond with the backbone carbonyl of P106. The UBX domain features an additional β-strand (Uβ0) that is positioned proximal to the N-terminal lobe of the p97 NTD (Fig. 3e). Uβ0 directly connects the UBX, PUB and helical lariat, indicating that these three domains may function together. Additionally, a C-terminal extension consisting of two α-helices (Uα2 and Uα3) connected by unstructured linkers is positioned on the apical surface of the canonical UBX and wraps over the core β-sheet.

PUB domains bind the HbYX (hydrophobic, tyrosine, any amino acid) motif located at the end of the p97 C-terminal tail28. Density for this domain is more poorly resolved in both the closed and open structures, which prompted us to perform a focused classification of this region (Extended Data Fig. 1e). Two classes show improved definition for the PUB, enabling the AlphaFold model for this region to be fit unambiguously into the density (Extended Data Fig. 6e,f). In class 1 (p97–UBXD1-PUBout), the PUB domain is positioned similarly to that in the closed model, and in class 2 (p97–UBXD1-PUBin), the PUB domain is rotated 46° about a linker connecting the PUB and UBX domains and points towards P6 (Fig. 3f, Extended Data Figs. 2 and 6f,g and Table 1). In both classes, connecting density is observed between the PUB and the bottom surface of P6, indicative of binding to the P5 C-terminal tail (Fig. 3g,h and Extended Data Fig. 6h–k). However, stronger tail density in p97–UBXD1-PUBin may indicate that binding of the HbYX motif is associated with inward rotation of the PUB domain. In summary, we identify that a single molecule of UBXD1 surprisingly interacts across three p97 protomers (P1, P5 and P6) through interactions by the VIM, UBX and PUB domains (Fig. 3i).

The UBXD1 helical lariat and H4 make distinct p97 interactions

The UBXD1 helical lariat is among the most striking UBXD1 binding elements owing to its intimate interaction with all three domains of the P6 protomer (Fig. 4a–e and Extended Data Fig. 7a). This domain is composed of four helices (Lα1–Lα4) inserted between Uβ0 and Uβ1 of the UBX domain that completely encircle the P6 NTD (Fig. 4a). Lα1 and Lα2 are positioned along the top of the P6 NTD; Lα1 is poorly resolved, and Lα2 projects two phenylalanine residues into the NTD (Fig. 4c). Lα3 is situated at the D1 interface of P6 and P1 and makes electrostatic contacts with residues in both the N-terminal and D1 domains of P6; in the closed state hydrophobic contacts are also made with the D1 domain of P1 (Fig. 4b,d). Lα4 contacts the P6 D2 domain and connects back to the UBX domain, completing the lariat. A short loop connects Lα3 and Lα4, further anchoring the lariat into this D2 domain (Fig. 4b,e). Additionally, the Lα3–Lα4 arrangement appears stabilized by a tripartite electrostatic network (Fig. 4e). Considering Lα3 displaces typical D1–D1 contacts between P1 and P6, we postulate that this helix probably contributes substantially to the D1 conformational changes and ring opening identified in the closed and open states of p97. Interestingly, the binding site of the Lα3–Lα4 linker overlaps with that of the p97 allosteric inhibitor UPCDC30245 (ref. 36), suggesting that occupancy of this site is a productive means to alter p97 function (Extended Data Fig. 7b).

Fig. 4: p97 remodeling interactions by UBXD1 helical lariat and VIM–H4.

a, Closed state map (from p97–UBXD1meta) showing density for the UBXD1 helical lariat (orange) and UBX (yellow) encircling the P6 NTD with Lα2, Lα3 and Lα4 interacting along the P6–P1 interprotomer interface. b, Expanded view showing Lα3 and Lα4 (orange) contacts with P6 across the NTD, D1 and D2, including putative electrostatic interactions (dashed lines). c, View of Lα2 interactions involving hydrophobic packing into the NTD using F292 and F293. d, View of the P6–P1 interface showing key contacts by Lα3 with the D1 α12 helix of protomer P1. e, View of Lα3 and Lα4 intra-lariat contacts (between R313, R318 and E326) and contacts with D2 (by L317 and T319), stabilizing the helical lariat. f, Unsharpened map of p97–UBXD1H4, showing density for H4 (green) adjacent to the VIM (brown) and along the P6–P1 interface. Shown below is an expanded view of the VIM–H4 sequence, featuring only a short, seven-amino acid linker connecting the two helices. g, Modeled view (see Extended Data Fig. 7c) of helix H4 interacting across the D2 domains at the P1–P6 interface from p97–UBXD1H4 (P1, dark blue; P6, light blue), overlaid (by alignment of the P1 D2 large subdomain) with p97–UBXD1closed (gray, showing conformational changes at the P6–P1 interface including displacement of P6 helix α5′ (red) and large rotation of P1 α12′).

Further classification of p97–UBXD1closed was performed to potentially resolve additional regions that were reported to interact with p97 (refs. 24,25) (Extended Data Fig. 1e). This analysis revealed an additional state similar to p97–UBXD1closed but with low-resolution density on top of the D2 domain of the P1 protomer (Fig. 4f, Extended Data Figs. 2 and 7c and Table 1). This region probably corresponds to H4 because of its proximity to the VIM, which is predicted to be connected to H4 by a seven-residue linker (Fig. 1b). In this structure (p97–UBXD1H4), H4 binding is associated with an upward rotation of the D2 small subdomain by ~17° relative to the closed state (Fig. 4g and Supplementary Video 4). This rotation results in the displacement of a short helix in P6 (α5′), weakening D2–D2 contacts between the seam protomers. As in p97–UBXD1open, α5′ is not present in the density map, probably due to increased flexibility (Extended Data Fig. 7d). Notably, a similar rotation of the D2 small subdomain of P1 is identified in p97–UBXD1open, supporting potential H4 occupancy (Fig. 2e). Indeed, a weak density similarly positioned atop the D2 domain was identified in the open state map at a low threshold (Extended Data Fig. 7e), indicating that H4 may be associated with the hexamer during ring opening and placing p97–UBXD1H4 as an intermediate between the closed and open states. Based on this analysis, we predict that H4 interactions play a key role in weakening D2 interprotomer contacts, thereby driving localized opening of the D2 ring.

The helical lariat and H4 are conserved p97-remodeling motifs

Given the rearrangements of the p97 hexamer driven by UBXD1, efforts were undertaken to identify other p97 adaptors with the UBX–helical lariat or VIM–H4 motifs. To this end, Dali searches39 against all structures in the Protein Data Bank and the AlphaFold database were performed, revealing one protein, ASPL (also called TUG or UBXD9), with a highly similar UBX–helical lariat arrangement (Fig. 5a and Extended Data Fig. 8a). Comparison of the p97–UBXD1 structures determined here to structures of an ASPL truncation (ASPL-C) bound to p97 (ref. 33) reveals a conserved interaction (Fig. 5b). Intriguingly, ASPL also inhibits p97 ATPase activity and disassembles p97 hexamers into smaller oligomers and monomers33,40. Although we find no evidence of a similar hexamer disruption in UBXD1 based on our SEC or cryo-EM analysis (Extended Data Fig. 1a–d), the split ring of the p97–UBXD1open structure is compelling as a related function of the UBX–helical lariat in the context of UBXD1 with its additional p97 binding domains.

Fig. 5: Analysis of the helical lariat and VIM–H4 as conserved p97-remodeling motifs.

a, Domain schematics of UBXD1, ASPL and SVIP (not to scale). b,c, Overlay of the UBX–helical lariat of ASPL (residues 318–495 from PDB 5IFS, colored as in a) and UBXD1 (residues 270–441 from the p97–UBXD1closed model, in white) (b) and the VIM–‘H4-like’ region of SVIP (AlphaFold model, colored as in a) and UBXD1 (residues 50–93 from the AlphaFold model, in white) (c). d, Steady-state ATPase activity of p97 as a function of ASPL-C or SVIP concentration (normalized to activity at 0 nM adaptor). Data are from n = 3 independent experiments, each with three technical replicates. Data are presented as mean values from each independent experiment. Calculated IC50 values are also shown.

Source data

Dali searches using the VIM–H4 motif did not produce any meaningful hits, probably owing to the structural simplicity of this region. However, examination of the predicted structures of other adaptors for unannotated structural elements proximal to a VIM suggested that the p97 adaptor small VCP/p97-interacting protein (SVIP) harbors an additional helix with modest similarity to UBXD1 H4 (Fig. 5c and Extended Data Fig. 8b). Given the predicted structural similarity to UBXD1 and inhibition of specific cellular functions, we reasoned that SVIP may similarly remodel p97 D2 contacts and inhibit ATPase activity.

We next purified ASPL-C33 and SVIP and analyzed their effect on p97 ATPase activity. ASPL-C potently inhibits p97 ATPase activity, with an IC50 of 96 nM (Fig. 5d). The complete loss of activity at high ASPL-C concentrations and highly cooperative inhibition (Hill slope, ~3) are probably a consequence of hexamer disassembly33. SVIP also strongly inhibits p97 ATPase activity (IC50 = 67 nM), indicating that the predicted helix C-terminal to the VIM contributes to ATPase inhibition through additional interactions with p97. Based on our structures and comparison to ASPL and SVIP, we postulate that the helical lariat and the H4 helix function as noncanonical control elements that, when paired with well-conserved binding motifs such as UBX and VIM, serve critical functions in ATPase control and p97 remodeling.

Multi-domain interactions by UBXD1 confer potent ATPase inhibition

We next sought to investigate the contribution of individual UBXD1 domains on the inhibition of p97 ATPase activity. We focused on the helical lariat and H4, given their proximity to the p97 hexamer seam, and mutated them individually and in combination (Extended Data Fig. 9a and Methods). These constructs bound p97 to a similar extent as did wild-type UBXD1 (Extended Data Fig. 9b,c) and were only modestly impaired in ATPase inhibition, indicating that disruption of these elements alone does not abolish ATPase inhibition (Fig. 6a,b,d). Curiously, the major effect of the H4 mutation (in both single-mutant and double-mutant contexts) was a ~60% increase in p97 ATPase activity at maximal inhibition, indicating a substantial loss in maximal ATPase inhibition (Fig. 6b, dashed lines). These results indicate that the lariat and H4 together contribute to p97 ATPase control, although other domains are probably also operative.

Fig. 6: Mutational analysis of UBXD1.