Structural insights into the coupling between VCP, an essential unfoldase, and a deubiquitinase

Processing of ubiquitylated proteins for degradation and recycling can depend on valosin-containing protein (VCP, or p97), an ATPases Associated with diverse cellular Activities (AAA) unfoldase (Bodnar and Rapoport, 2017b; Ye et al., 2017). VCP extracts polyubiquitylated proteins, as well as select nonubiquitylated proteins, from macromolecular complexes, organelles, and membranes and unfolds them (Bodnar and Rapoport, 2017a; Weith et al., 2018; Ye et al., 2003; Blythe et al., 2017). Consistent with these essential functions, VCP overexpression is linked to cancer (Huryn et al., 2020) and mutations are linked to neurodegenerative disorders, such as multisystem proteopathies and vacuolar tauopathy (Darwich et al., 2020; Meyer and Weihl, 2014). Chemical inhibitors, as well as activators, of VCP have been identified and are enabling tests of therapeutic hypotheses (Jones et al., 2024; Wrobel et al., 2022; Figuerola-Conchas et al., 2020; Phan et al., 2023, Preprint; Huryn et al., 2020). In current models, K48-linked polyubiquitylated substrates are recognized by cofactors and subsequently, in an ATP hydrolysis–dependent mechanochemical cycle, threaded through the pore (hereafter, central pore) located at the center of the ring-shaped VCP hexamer (Bodnar and Rapoport, 2017a; Cooney et al., 2019; Xu et al., 2022; Twomey et al., 2019; Ji et al., 2022). For targeting to the proteasome, an unfolded substrate must retain at least part of its polyubiquitin chain (Bodnar and Rapoport, 2017a; Olszewski et al., 2019; Li et al., 2024; Ji et al., 2022). For substrate recycling, the polyubiquitin chain must be removed after recognition, extraction, and unfolding by VCP. However, we do not understand how VCP activity is coupled to the processing of polyubiquitin chains.

Deubiquitinases (DUBs) are a class of proteases, subclassified into seven families, that recognize and selectively cleave ubiquitin (Mevissen and Komander, 2017; Schauer et al., 2020). DUBs have been shown to modulate the levels of oncogenes or tumor suppressors and are currently attractive targets for therapeutics (Schauer et al., 2020; Harrigan et al., 2018). A subset of DUBs have been shown to interact with VCP, including ataxin-3, YOD1, USP13, and VCPIP1 (Sowa et al., 2009; Ernst et al., 2009; Nakayama and Kondo, 2024). In particular, VCPIP1 has been shown to play a key role in post-mitotic Golgi and ER biogenesis (Uchiyama et al., 2002) and DNA repair (Huang et al., 2020). In current models for post-mitotic membrane organization, VCPIP1 binds VCP and a second cofactor, p47, to regulate the disassembly of SNARE complexes (Wang et al., 2004). Structural data for how VCP binds any DUB are limited and include a crystal structure of a complex formed by truncated noncatalytic domains in VCP and the Saccharomyces cerevisiae DUB OTU1 (YOD1 in humans) (Kim et al., 2014). However, we lack structural models for how the protease active sites of DUBs are positioned relative to VCP’s central pore or whether VCP-DUB interactions can regulate enzymatic activities.

Here, we use single-molecule mass photometry (MP), cryogenic electron microscopy (cryo-EM), and biochemical assays to analyze the interactions between VCP and VCPIP1. We determine two cryo-EM structures of VCP-VCPIP1 complexes, in the absence of added ATP or in the presence of an ATP analog, that together reveal how VCPIP1 interacts with both the N- and C-terminal domains of VCP across distinct nucleotide states. We note that while our manuscript was in preparation, structures of VCP-VCPIP1 without the addition of nucleotide were reported (Liao et al., 2024; Shah et al., 2024, Preprint) and we compare our findings with these data. We test these structural models using mutagenesis and truncated constructs, examine competition in VCP binding between VCPIP1 and other cofactors, and show that VCPIP1’s DUB activity is stimulated in the presence of VCP. Our findings suggest a model for how the coupling of VCP with DUBs can regulate the processing of polyubiquitylated substrates.

To characterize the VCP-VCPIP1 interaction, we expressed and purified affinity tag-free wild-type (WT) human VCP (VCP^WT), which consists of a globular N-terminal domain (N), two ATPase domains (D1/D2), and a likely disordered C-terminal tail (Xia et al., 2016), and VCPIP1 (VCPIP1^WT), which is comprised of an OTU catalytic domain and a UBX-L domain (Schauer et al., 2020; Braxton and Southworth, 2023), using published procedures with some modifications (Jones et al., 2024; Wang et al., 2004, 2019) (see Materials and methods) (Fig. 1 A and Fig. S1 A).

We next used MP, which enables the analysis of the mass of single molecules in solution (Young et al., 2018). We observe a distribution of masses for VCP^WT (mean ± SD: 564 ± 36 kDa) consistent with a hexamer (Fig. 1 B). VCPIP1^WT alone revealed masses corresponding to a monomer in solution (mean ± SD: 136 ± 19 kDa) (Fig. 1 B). When VCP^WT (150 nM monomer) was mixed with VCPIP1^WT (100 nM), we observed peaks for VCPIP1 monomers and VCP-VCPIP1 complexes (Fig. 1 B). Very few (∼10% of total counts) free VCP hexamers were detected (see arrow, Fig. 1 B), consistent with a high-affinity VCP-VCPIP1 interaction. The broad peak representing the VCP-VCPIP1 complex was fit to a Gaussian distribution with a mean (mean ± SD: 930 ± 167 kDa) that corresponds to a stoichiometry of 3 VCPIP1 monomers bound to a single VCP hexamer. The standard deviation is on the order of 1 VCPIP1, suggesting that a wide range of stoichiometries is possible under these conditions and, in part, may be linked to limits in mass resolution, which for MP has been shown to decrease with increased mass (Young et al., 2018).

To examine the structure of the VCP-VCPIP1 complex, we used single-particle cryo-EM. VCP^WT was mixed with excess VCPIP1^WT without added nucleotide, and samples were processed for data collection. Micrographs revealed particles with various orientations (Fig. S1 B), and 2D classification (∼2.9 million particles) revealed classes with stacked hexagons, as would be expected for VCP (Fig. 1 C). We also observe an additional density associated with many of the stacked hexagons (Fig. 1 C, asterisk).

We next separated particles into 3D classes and selected those containing hexagons with additional density for further processing (Fig. S1 C). A model for VCP (PDB: 8OOI) (Shein et al., 2024) could be rigid-body fit into the stacked hexagon density (∼6–8 Å resolution) (Fig. S2 A). The assignment of VCP’s D2 ring was supported by helix α12′, a prominent feature of the domain (DeLaBarre and Brunger, 2003; Braxton et al., 2023) (Fig. 1 D and Fig. S2 A). The N domain adopts “up” or “down” conformations, coaxial or coplanar with the D1 domains, respectively, depending on VCP’s nucleotide state (Bodnar and Rapoport, 2017b; Xia et al., 2016). Our data are consistent with VCP’s N domains in a “down” position, likely corresponding to a state in which ADP is bound in VCP’s D1 domain (Banerjee et al., 2016; Shein et al., 2024) (Fig. S2 A). The central pore formed by the VCP hexamer is apparent (Fig. 1 D).

The additional density below the VCP D2 domain was too large and globular to be assigned to VCP’s C-terminal tail, which is not completely accounted for in earlier structures (Pan et al., 2021; Banerjee et al., 2016; Shein et al., 2024). We initially assigned this density to VCPIP1, as an AlphaFold model (AF-Q96JH7-F1) (Jumper et al., 2021; Varadi et al., 2024) was consistent with the overall size and shape of the map (also see below). 3D classes with one, two, or three VCPIP1 protomers comprise the majority (∼80%) of cleaned particles (Fig. 1 D). In these classes, we observe one VCPIP1 protomer at the interface formed by two adjacent VCP protomers. Together, our cryo-EM data reveal VCP-VCPIP1 structures in which up to three globular domains of the VCPIP1 protomer are positioned near the exit of VCP’s central pore, the site from which unfolded substrates have been proposed to emerge (Bodnar and Rapoport, 2017a; Cooney et al., 2019; Xu et al., 2022; Twomey et al., 2019; van den Boom et al., 2023).

To obtain higher resolution structural data, we identified an asymmetric unit containing one VCPIP1 and two VCP protomers (Fig. 2 A). We employed a strategy including C3 symmetry expansion and masked 3D classification around the asymmetric unit (Fig. S1 C) to generate an overall map of ∼3.3 Å resolution (hereafter, (VCP-VCPIP1)_Apo) (Fig. 2 A; Fig. S1, D and E; and Table S1). In high-resolution regions, we observe density corresponding to secondary structure elements and side chains (Fig. 2 A). The distal ends of the map have higher noise and were not included in subsequent model building.

We next generated a model for the (VCP-VCPIP1)_Apo asymmetric unit starting from individual domains of a reported ADP-bound VCP structure (PDB: 5FTL) (Banerjee et al., 2016) and VCP C-terminal tail residues 764–775, which were resolved in a peptide substrate–bound VCP structure (PDB: 7LN6) (Pan et al., 2021). Our map had sufficient resolution to assign and build side chains into the density corresponding to the D1 and D2 domains (Fig. 2 B). We observe some density likely corresponding to ADP in the D1 and D2 domains of both VCP protomers (Fig. S1 F). However, unambiguous assignment of the nucleotide state is difficult at the current resolution, and as no nucleotide was added, we refer to this map as “apo.” The density corresponding to the VCP N domain is weak, consistent with conformational flexibility, and therefore, we did not include the N domain in our model at this stage (see below).

To model VCPIP1, we used an AlphaFold model (AF-Q96JH7-F1) (Jumper et al., 2021; Varadi et al., 2024) (see Materials and methods). A portion of VCPIP1 (residues 556–666), which we name the “stalk,” could be identified near the exit of the VCP central pore (Fig. 2, A and B; and Table S1). Due to flexibility in the density connected to the VCPIP1 stalk, we did not include this region in our model but tentatively assigned it to be the OTU domain (see below).

We next used our map and model of the asymmetric unit (Fig. 2, A and B) to identify key interactions between the VCPIP1 stalk and the VCP D2 domain (Fig. 2, C–E). First, we observe VCPIP1 F638 interacting with K754 and Y755 in the C-terminal helix on one VCP protomer (Fig. 2 D). In the adjacent VCP protomer, a second interaction between VCPIP1 (including N621/Y623) and the VCP C-terminal helix (including M757/F758) is evident (Fig. 2 E). Both of these interactions are distinct from interactions between the VCP C-terminal tail and PUB or PUL domains (Braxton and Southworth, 2023).

To analyze the contribution of these interactions to complex formation, we purified two VCP mutants, K754N and Y755H (VCP^K754N and VCP^Y755H) (Fig. S1 A). MP indicates that these VCP mutants form hexamers in solution (Fig. S2 B). When these VCP mutants were mixed with VCPIP1^WT, we observed peaks for VCPIP1 monomer and VCP hexamer (Fig. 2 F). More unbound VCP hexamer is present in comparison with VCP^WT (VCP^K754: 20%; VCP^Y755H: 27% of total counts) (Fig. 1 B and Fig. 2 F). We also observe broad peaks that may correspond to complexes between VCPIP1^WT and VCP mutants, but the low intensity suggests that binding is weaker compared to that for the WT enzymes (Fig. 2 F).

We next generated an AlphaFold model guided by the stoichiometry of VCP and VCPIP1 in the asymmetric unit (2 VCP and 1 VCPIP1) (Abramson et al., 2024) (Fig. 2 G and Fig. S4). Alignment of our model to the one generated by AlphaFold showed considerable agreement (RMSD, all atoms = ∼1.7 Å). Interestingly, the AlphaFold model also revealed an interaction between VCPIP1 and VCP’s C-terminal residues 799–806, which have been shown to interact with other cofactors (Braxton and Southworth, 2023) (Fig. 2 H). We next purified a VCP construct lacking its C-terminal tail (VCP^ΔC, residues 1–765), which includes the region of VCP’s C-terminal helix that interacts with the VCPIP1 stalk (Fig. 2, D, E, and I; and Fig. S1 A). MP analysis indicated that VCP^ΔC forms a hexamer, but does not bind VCPIP1 under the conditions tested (Fig. 2 I and Fig. S2 B). Together, these data reveal that up to three distinct contacts between VCPIP1 and VCP’s C-terminus can contribute to the formation of the VCP-VCPIP1 complex.

Biochemical analyses reveal direct binding between the VCP N domain and a VCPIP1 construct containing its C-terminal UBX-L domain (Nakayama and Kondo, 2024). To analyze this interaction, we performed additional cryo-EM data processing (Fig. S1 C). In classes where the N domains were coplanar with the D1 domain (“down”), we rigid-body fit an ADP-bound VCP model (PDB: 5FTL) (Banerjee et al., 2016) and observed an additional density that could not be accounted for by the model (Fig. S2 C). We further refined this class and resolved the structure to ∼4.4 Å (hereafter, (VCP^N-VCPIP1^UBX-L)_Apo) (Fig. 3 A; and Fig. S1, C, G, and H). A model for the VCPIP1 UBX-L domain (AF-Q96JH7-F1, residues 774–851) was rigid-body fit into the additional density (Jumper et al., 2021; Varadi et al., 2024) (Fig. 3 B). Gratifyingly, our model is consistent with an x-ray crystal structure of the VCP N domain in complex with the UBX-L of OTU1, a yeast DUB (PDB: 4KDL, RMSD, all atoms = ∼1.2 Å) (Fig. 3 C). In both models, the same UBX-L residues are positioned in a cleft formed by secondary structure elements in VCP’s N domain (Kim et al., 2014) (Fig. 3, B and C).

Our results diverge from other recently reported structures of the VCP-VCPIP1 complex. One study indicates that the VCP N domain is coaxial with the D1 domain (“up”) when bound to VCPIP1, but their data do not reveal density suggesting a direct interaction with the VCPIP1 UBX-L domain (Liao et al., 2024). Another study (Shah et al., 2024, Preprint) does not specify the N domain conformation. To further examine VCPIP1 UBX-L binding to VCP in a nucleotide state that may change the N domain conformation, we performed MP experiments in the presence of AMPPNP, an ATP analog that was shown to increase binding of VCPIP1 to VCP (Nakayama and Kondo, 2024). Consistent with our MP analyses in the absence of added nucleotides (Fig. 1 B), we observe peaks for VCPIP1 monomer (mean ± SD: 149 ± 23 kDa) and VCP-VCPIP1 complex (mean ± SD: 966 ± 202 kDa) and very few (∼13% of total counts) free VCP hexamers (Fig. 3 D).

We next used cryo-EM to solve a structure of the VCP-VCPIP1 complex in the AMPPNP-bound state. Under these conditions, micrographs yielded particles in multiple orientations (Fig. S3, A and B) and initial 3D classification revealed distinct conformations of the VCP N domains (Fig. S3 C). We selected particles in which all six N domains are in the “up” position for further processing and identified an initial class of particles in which 3 VCPIP1 protomers are bound to the VCP hexamer (Fig. 3 E and Fig. S3 C).

Using a processing pipeline similar to the apo complex, we determined a ∼3.3 Å structure of the VCP-VCPIP1 asymmetric unit in the presence of AMPPNP (hereafter, (VCP-VCPIP1)_AMPPNP) (Fig. 3 F, Fig. S3, C–E, and Table S1). The overall structural organization is similar to what we observed in the absence of added nucleotide (Fig. 2 A and Fig. 3 F). We observe density likely corresponding to AMPPNP in VCP’s D1 domains and ADP in the D2 domains (Fig. S3 G). Density for the N domain, which is in its “up” conformation, was stronger than in the apo complex (Fig. 2 A) and we also observe density for the VCPIP1 UBX-L domain (Fig. 3 F). Given the limited resolution in the VCPIP1 UBX-L and OTU domains (Fig. 3 F and Fig. S3 E), we built a model for the (VCP-VCPIP1)_AMPPNP complex that includes VCP D1/D2 and VCPIP1 stalk domains (Fig. S3 F and Table S1).

To examine the interaction between “up” VCP N and VCPIP1 UBX-L domains, we performed C6 symmetry expansion and masked 3D classification on the VCP N-D1 domains and resolved this map to ∼3.8 Å (hereafter, (VCP^N-VCPIP1^UBX-L)_AMPPNP) (Fig. 3 G; and Fig. S3, C, H, and I). We rigid-body fit models for the VCP N-D1 and VCPIP1 UBX-L domains into the density and observe an upward shift in VCP’s N domain and the associated VCPIP1 UBX-L domain when compared to our model without added nucleotide (Fig. 3 H). Given these results, we reexamined 3D classes from the (VCP-VCPIP1)_Apo dataset and observed density that likely corresponds to VCPIP1’s UBX-L domain, even when the VCP N domain is in the “up” conformation (Fig. S2 C). Together, these data suggest that VCPIP1’s UBX-L can bind VCP as the N domain samples distinct conformations, independent of nucleotide state.

We next used MP assays to examine whether the VCP N-VCPIP1 UBX-L interaction is required for complex formation. A VCPIP1 construct lacking its C-terminal UBX-L domain (hereafter, VCPIP1^ΔUBX-L, residues 1–743) was expressed and purified (Fig. 3 I and Fig. S1 A). MP analysis indicated that VCPIP1^ΔUBX-L is monomeric (mean ± SD: 83 ± 14 kDa) (Fig. S2 D). When VCP^WT was mixed with VCPIP1^ΔUBX-L, we did not observe binding under the conditions of MP experiments but identified peaks corresponding to the proteins alone (mean ± SD: VCPIP1^ΔUBX-L = 83 ± 19 kDa, VCP^WT = 534 ± 69 kDa) (Fig. 3 I).

We also used immobilized metal affinity chromatography to examine the VCP N-VCPIP1 UBX-L interaction. VCP was purified with a hexahistidine tag at its N terminus (hereafter, His-VCP) (Fig. S1 A) and incubated with Ni-NTA agarose beads and VCPIP1 constructs. While we observe both VCPIP1^WT and VCPIP1^ΔUBX-L in the supernatant, or unbound to His-VCP–immobilized beads, we observe VCPIP1^WT in the eluate (Fig. 3 J). We do not observe VCPIP1^ΔUBX-L in the eluate beyond levels of nonspecific binding to the beads in the absence of His-VCP (Fig. 3 J). These data suggest that the VCPIP1 UBX-L can bind the N domain in conformations adopted during VCP’s mechanochemical cycle and that this interaction contributes to complex formation.

To examine the flexible region of VCPIP1 distal to the stalk-D2 interface, we processed a map containing three VCPIP1 protomers bound to the VCP hexamer in apo and AMPPNP conditions. Again, we performed C3 symmetry expansion and local refinement to obtain ∼3.4 Å resolution maps (hereafter, (VCP-3 VCPIP1)_Apo and (VCP-3 VCPIP1)_AMPPNP) (Fig. 4 A; Fig. S1, I and J; and Fig. S3, J and K). We also refined maps containing two VCPIP1 protomers without symmetry expansion (hereafter, (VCP-2 VCPIP1)_Apo and (VCP-2 VCPIP1)_AMPPNP) (Fig. 4 C). Although the interface between VCP and VCPIP1 is similar in apo and AMPPNP conditions, we observe differences between classes in which two versus three VCPIP1s are bound. In classes with three VCP-bound VCPIP1 protomers, we observe density connecting the protomers (Fig. 4, A and B). In contrast, the classes with only two VCPIP1 protomers lack this density (Fig. 4, C and D). An overlay of (VCP-3 VCPIP1)_Apo and (VCP-2 VCPIP1)_Apo maps suggests a conformational difference upon binding of three VCPIP1 protomers to VCP (Fig. 4 E). Two bound VCPIP1 protomers are arranged in a splayed conformation and, in the presence of a third protomer, are positioned closer to the central axis of the structure (Fig. 4 E). We note that one report does not observe mixed stoichiometries of the VCP-VCPIP1 complex and suggests that VCP-bound VCPIP1s function as distinct monomers (Liao et al., 2024). The varying stoichiometries of VCP-VCPIP1 complexes and the interprotomer contacts may regulate, or even limit, the formation of the VCP “spiral staircase,” a conformation adopted by the substrate-engaged unfoldase (Cooney et al., 2019; Xu et al., 2022; van den Boom et al., 2023; Twomey et al., 2019; Pan et al., 2021).

We next rigid-body fit an AlphaFold model for the VCPIP1 OTU domain (AF-Q96JH7-F1, residues 112–553) (Jumper et al., 2021; Varadi et al., 2024) into (VCP-3 VCPIP1)_Apo and (VCP-2 VCPIP1)_Apo maps (Fig. 4, F and G). These models reveal ∼3–10 Å movements in the OTU domains in complexes containing a third VCPIP1 protomer (Fig. 4 F). We also used AlphaFold to examine the interaction between VCP’s C-terminus (residues 471–806) and two or three VCPIP1 protomers (residues 112–666) (Abramson et al., 2024). The predicted models are consistent with the conformational differences we observe in our structural data (Fig. S2 E and Fig. S4).

Using our models, we identified a second characteristic of the VCPIP1 OTU domain. We observe the DUB active site, containing the catalytic triad of C219, N353, and H354, exposed to a cleft where we anticipate ubiquitin may bind (Fig. 4, G and H). Analysis of the (VCP-3 VCPIP1)_Apo map suggests that a ring formed by adjacent VCPIP1 OTU domains can accommodate folded ubiquitin, while the stalk region, where no interdomain contacts are observed, is substantially narrower (Fig. 4 I). Together, these data indicate that VCP-bound VCPIP1s undergo a conformational change that enables OTU domains to form a cavity wide enough for a folded ubiquitin and positions the protease active sites below VCP’s central pore.

To examine whether VCP-VCPIP1 interactions can alter enzyme activity, we used biochemical assays with recombinant proteins. We first tested VCP’s basal ATPase activity in the presence of VCPIP1, but without substrates, and did not observe a change in VCP’s ATPase activity across a range of VCPIP1 concentrations (0–1 μM) at which we observe VCP-VCPIP1 binding (Fig. 1 B, Fig. 3 J, and Fig. 5 A). VCPIP1 alone does not exhibit any residual ATPase activity (Fig. S2 F). The modest (<25%) increase in VCP’s ATPase activity in the presence of VCPIP1 reported recently (Liao et al., 2024) could be technical (e.g., due to an impurity copurifying with recombinant VCPIP1).

We next tested VCPIP1’s DUB activity, in the presence or absence of VCP, toward a polyubiquitylated substrate in solution. We employed VCPIP1^WT, WT or mutant VCP, and K48-linked polyubiquitylated and photocleaved mEos3.2 (hereafter, polyUb₂-mEos), an extensively used model substrate for VCP that also mimics VCPIP1’s ubiquitin linkage specificity (Blythe et al., 2019; Ritorto et al., 2014) (Fig. S1 A). VCP alone does not modify substrate chain length, while the addition of VCPIP1 alone resulted in the reduction of higher molecular weight (“long”) species and the appearance of lower molecular weight (“short”) species over time, consistent with the trimming of ubiquitin chains (Fig. 5 B). When VCP was mixed with VCPIP1 and polyUb₂-mEos, we observed even further reduction in long mEos-linked polyubiquitin chains and more short chains (Fig. 5 B). As this experiment was performed in the absence of ATP or cofactors, stimulation of VCPIP1 DUB activity by VCP is likely to be ATPase- and unfoldase-independent. To test this further, we expressed and purified VCP with an E578Q mutation (VCP^E578Q), known to suppress ATP hydrolysis (Song et al., 2003; Ye et al., 2003) (Fig. S1 A). VCP^E578Q stimulated VCPIP1 DUB activity similar to levels observed in the presence of VCP^WT (Fig. 5 B). When we used VCP^ΔC, a construct we show does not bind VCPIP1 under the conditions tested (Fig. 2 I), stimulation of VCPIP1 DUB activity was not observed (Fig. 5 B). Neither VCP^E578Q nor VCP^ΔC exhibits DUB activity alone (Fig. S2 G). These data suggest that binding to VCP is sufficient to stimulate VCPIP1’s DUB activity. Consistent with our findings, one report shows that VCP stimulates VCPIP1’s DUB activity toward rhodamine–ubiquitin, a reporter substrate (Shah et al., 2024, Preprint).

Since VCPIP1 contains a UBX-L domain that binds VCP’s N domain, we considered the possibility that VCPIP1 could compete with cofactors that also bind the N domain, particularly those that have been shown to recruit substrates to VCP, such as Ufd1/Npl4 (UN) (Ye et al., 2003; Bodnar and Rapoport, 2017a). To test this, we established a pulldown assay using streptavidin-binding peptide (SBP)–tagged Ufd1 and tagless Npl4 (hereafter, UN-SBP) and VCP^WT. VCP binds UN-SBP as indicated by its presence in the eluate from streptavidin beads (Fig. 5 C). Minimal amounts of unbound VCP are detected in the supernatant (Fig. 5 C). In the presence of increasing concentrations of VCPIP1^WT, we observe loss of VCP from the eluate and appearance of VCP in the supernatant (Fig. 5 C). Importantly, VCPIP1^ΔUBX-L does not compete with UN-SBP for binding to VCP (Fig. S2 H). Although most VCPIP1^WT does not bind UN-SBP, a fraction of VCPIP1 is bound to UN-SBP and eluted from the beads (Fig. 5 C). VCPIP1^WT exhibits minimal nonspecific binding to streptavidin beads alone (Fig. 5 C). Together, these data reveal that VCP binding stimulates VCPIP1 DUB activity and suggest a model in which VCPIP1 may compete with other cofactors for VCP binding during substrate processing.

In summary, we report two cryo-EM structures of the VCP-VCPIP1 complex where the VCPIP1 stalk binds proximal to the VCP C-terminal tail, an interaction that positions VCPIP1’s OTU domain below the exit of VCP’s central pore through which unfolded substrates are likely to emerge. We note that the arrangement of the VCP-VCPIP1 complex is similar across this and two recent studies (Liao et al., 2024; Shah et al., 2024, Preprint). Our findings also provide evidence for direct binding between VCPIP1’s UBX-L domain and VCP’s N domain in the “up” or “down” position. Additionally, our data indicate that the stoichiometry of the VCP-VCPIP1 complex can vary. Upon binding of a third VCPIP1 protomer to VCP, the VCPIP1 OTU domain may undergo a conformational change and interactions can be established between adjacent VCPIP1 protomers. Overall, the VCP-VCPIP1 complex mimics other ATP-dependent proteolysis systems such as the ClpXP and LonP complexes or the proteasome, which also have protease sites positioned where substrates emerge from ATPase mechanoenzymes (Baker and Sauer, 2012; Bard et al., 2018; Wlodawer et al., 2022).

We propose a simple model for the coupling of VCP’s unfoldase activity to VCPIP1’s DUB activity (Fig. 5 D). The unfolding or extraction of substrates is initiated through the recognition of three or more ubiquitins by a VCP–cofactor (e.g., UN) complex, followed by insertion of ubiquitin into VCP’s central pore (Bodnar and Rapoport, 2017a; Twomey et al., 2019). Up to three VCPIP1 protease sites can be positioned at the exit of VCP’s central pore, and the VCPIP1 UBX-L domains can compete with other cofactors for binding to the VCP N domains in either “up” or “down” conformations. Our pulldown experiments suggest that a fraction of VCP may bind VCPIP1 and UN simultaneously. While more work is needed to determine when VCPIP1 binds relative to substrate recruitment, it is possible that VCPIP1–cofactor competition and the variable stoichiometry of VCP-VCPIP1 complexes can allow VCP to adopt its spiral staircase conformation for substrate loading and translocation. Recruitment of all three VCPIP1s to VCP could prevent further substrate engagement, possibly by suppressing interactions with additional cofactors that recruit substrates, until ubiquitins have been cleaved. Ubiquitins must be folded for recognition and cleavage by VCPIP1 (Ji et al., 2022). We propose that ubiquitin likely remains unfolded while translocating through the narrow pore formed by three VCPIP1 stalks. Once the unfolded ubiquitin reaches the expanded cavity formed by three OTU domains, it can refold and then be cleaved by VCPIP1, whose DUB activity is stimulated upon binding to VCP. Some, or all, of the ubiquitins linked to a substrate may be cleaved before the unfolded substrate dissociates from VCP. VCPIP1 and other cofactors may again compete for VCP binding to repeat this cycle. Such coordination between VCP and VCPIP1 would enable recycling of the proteins extracted from multiprotein complexes or membranes and is likely to be important for faithful ubiquitin-dependent substrate processing and maintenance of protein homeostasis.

The plasmid for expressing WT human VCP (with an N-terminal His-tag and TEV cleavage site) in bacteria was obtained from T.-F. Chou (Caltech). To express human VCPIP1 in bacteria, the WT human VCPIP1 gene was purchased from Addgene (plasmid #22592 from Wade Harper [Harvard Medical School, Boston, MA, USA] [Sowa et al., 2009]), PCR-amplified, and cloned into a pET15b vector using the Gibson assembly. Plasmids for VCP and VCPIP1 point mutants were generated by site-directed mutagenesis. VCP^ΔC (residues 1–765) and VCPIP1^ΔUBX-L (residues 1–743) were generated by PCR amplification from the full-length genes and Gibson assembly into a pET15b vector. Ufd1 and Npl4 genes were purchased from Addgene (plasmid #117107 [Ufd1] and 117108 [Npl4] from Hemmo Meyer [University of Duisburg-Essen, Essen, Germany] [Weith et al., 2018]). UN-SBP was generated by amplification of the SBP-tag (plasmid #36943; Addgene, a gift from Cheryl Arrowsmith [University of Toronto, Toronto, Canada]) and insertion into a pET41b⁺ vector containing the Ufd1 gene (plasmid #117107; Addgene) using Gibson assembly. Plasmids for the synthesis and purification of polyUb₂-mEos were purchased from Addgene (ubiquitin: plasmid #12647 from Rachel Klevit [University of Washington, Seattle, WA, USA] [Brzovic et al., 2006]; mouse Ube1: plasmid #32534 from Jorge Eduardo Azevedo [Universidade do Porto, Porto, Portugal] [Carvalho et al., 2012]) or a gift from Andreas Martin (University of California, Berkeley, Berkeley, CA, USA) (plasmids encoding genes for His₆-Ub^G76V-Ub^G76V-mEos3.2 and a gp78-Ubc7 fusion protein) (Blythe et al., 2019).

Full-length human tagless (VCP^WT) and His-tagged VCP (His-VCP) were expressed and purified as described previously (Jones et al., 2024), but the TEV protease cleavage step was omitted from the purification of His-VCP. VCP^E578Q and VCP^ΔC were expressed and purified using the same protocol as the WT protein.

Full-length human VCPIP1^WT and VCPIP1^ΔUBX-L were expressed and purified as described previously (Wang et al., 2004), but with modifications. VCPIP1 was expressed in Escherichia coli Rosetta (DE3) pLysS cells (cat. No. 70954; Merck) and grown following the same procedure used for VCP. Cells were pelleted and resuspended in lysis buffer (50 mM HEPES, pH 7.5, 300 mM KCl, 20 mM imidazole, 5% glycerol, 5 mM β-mercaptoethanol [βME], 1 mM p-phenylmethylsulfonyl fluoride [PMSF], cOmplete EDTA-free protease inhibitor cocktail [Roche]). All subsequent purification steps were performed at 4°C. Cells were lysed using an EmulsiFlex-C5 homogenizer (Avestin, ∼6 cycles at 10,000–15,000 psi homogenization pressure). The lysate was clarified by centrifugation at 45,000 rpm for 30 min using a Type 70 Ti rotor in a Beckman Coulter Optima LE-80K ultracentrifuge. The clarified lysate was filtered and loaded onto Ni-NTA resin and incubated for 1 h. The resin was washed with ∼200 ml of wash buffer (50 mM HEPES, pH 7.5, 300 mM KCl, 20 mM imidazole, 5% glycerol, 5 mM βME) and eluted with elution buffer (50 mM HEPES, pH 7.5, 300 mM KCl, 300 mM imidazole, 5% glycerol, 5 mM βME). The eluate was treated with TEV protease and dialyzed in dialysis buffer (50 mM HEPES, pH 8.0, 100 mM KCl, 20 mM imidazole, 5% glycerol, 5 mM βME) overnight. The protein was passed three times over Ni-NTA resin pre-equilibrated with reverse Ni buffer (25 mM HEPES, pH 8.0, 100 mM KCl, 20 mM imidazole, 5% glycerol, 5 mM βME), or incubated with resin for 1 h. The flow-through was loaded onto a Q column (GE Healthcare) pre-equilibrated with low salt buffer (25 mM HEPES, pH 8.0, 100 mM KCl, 5% glycerol, 2 mM DTT) and eluted using a 0–100% gradient of high salt buffer (50 mM HEPES, pH 8.0, 500 mM KCl, 5% glycerol, 2 mM DTT) over 20 ml. Fractions containing VCPIP1 were concentrated using an Amicon Ultra 100K concentrator to 0.5 ml, centrifuged, and loaded onto a Superdex 200 Increase 10/300 column (GE Healthcare) equilibrated with gel filtration buffer (50 mM HEPES, pH 7.5, 150 mM KCl, 1 mM DTT). The fractions containing purified VCPIP1 were pooled, concentrated using an Amicon Ultra 100K concentrator, and frozen in liquid nitrogen and stored at −80°C.

Ufd1-SBP and Npl4 were expressed separately in E. coli Rosetta (DE3) pLysS cells (cat. No. 70954; Merck) grown in Miller’s LB medium at 37°C until OD₆₀₀ = ∼0.6. Protein expression was induced with 0.5 mM IPTG, and cultures were grown at 16°C (Npl4) or 30°C (Ufd1-SBP) for ∼18 h, pelleted, and resuspended in lysis buffer (50 mM Tris-HCl, pH 8.0, 200 mM KCl, 1 mM MgCl₂, 20 mM imidazole, 5 mM βME, 1 mM PMSF, cOmplete EDTA-free protease inhibitor cocktail [Roche]). Ufd1-SBP and Npl4 resuspensions were mixed in a 2:1 ratio, and all subsequent purification steps were performed at 4°C. Cells were lysed using an EmulsiFlex-C5 homogenizer (Avestin, ∼6 cycles at 10,000–15,000 psi homogenization pressure). The lysate was clarified by centrifugation at 45,000 rpm for 30 min using a Type 70 Ti rotor in a Beckman Coulter Optima LE-80K ultracentrifuge. The clarified lysate was filtered and loaded onto Ni-NTA resin and incubated for 1 h. The resin was washed with ∼200 ml of wash buffer (50 mM Tris-HCl, pH 8.0, 500 mM KCl, 1 mM MgCl₂, 20 mM imidazole, 5 mM βME) and eluted with elution buffer (50 mM Tris-HCl, pH 8.0, 200 mM KCl, 1 mM MgCl₂, 400 mM imidazole, 5 mM βME). The eluate (2 ml) was loaded onto a Superdex 200 HiLoad 16/600 column (GE Healthcare) equilibrated with gel filtration buffer (50 mM HEPES, pH 7.5, 200 mM KCl, 1 mM MgCl₂, 1 mM DTT, 5% glycerol). The fractions containing purified Ufd1-SBP/Npl4 were pooled, concentrated using an Amicon Ultra 30K concentrator, and frozen in liquid nitrogen and stored at −80°C.

Purifications of mono-Ub, Ube1, gp78-Ubc7, and His₆-Ub^G76V-Ub^G76V-mEos3.2 were performed as described previously (Blythe et al., 2017, 2019). To synthesize and purify polyUb₂-mEos, His₆-Ub^G76V-Ub^G76V-mEos3.2 (20 µM) was mixed with Ube1 (1 µM), gp78-Ubc7 (10 µM), and ubiquitin (400 µM) in 2 ml buffer (20 mM HEPES, pH 7.4, 150 mM KCl, 5 mM MgCl₂, 10 mM MgATP). Ubiquitin was added sequentially every hour for the first 6 h of incubation at 37°C. After the final addition of ubiquitin, the reaction was kept at 37°C overnight (total = 23 h). The following day, the sample was placed on ice and incubated under UV light (365 nm) for 90 min for photoconversion of mEos. The ubiquitylated, photoconverted substrate was incubated with Ni-NTA resin for 2 h at 4°C. The resin was washed with 50 ml of wash buffer (20 mM HEPES, pH 7.4, 250 mM KCl, 1 mM MgCl₂, 20 mM imidazole, 5% glycerol, 2 mM βME) and eluted with elution buffer (20 mM HEPES, pH 7.4, 250 mM KCl, 1 mM MgCl₂, 300 mM imidazole, 5% glycerol, 2 mM βME). The elution was concentrated using an Amicon Ultra 10K concentrator to 0.5 ml, centrifuged, and loaded onto a Superose 6 Increase 10/300 column (GE Healthcare) equilibrated with gel filtration buffer (20 mM HEPES, pH 7.4, 250 mM KCl, 1 mM MgCl₂, 5% glycerol, 1 mM TCEP). The fractions containing long, medium, and short polyUb₂-mEos chains were pooled, concentrated using an Amicon Ultra 10K concentrator, and frozen in liquid nitrogen and stored at −80°C. Medium-length polyUb₂-mEos chains were used for DUB assays (see below).

Full-length VCP^WT was buffer-exchanged into 50 mM HEPES, pH 7.5, 25 mM KCl, 2.5 mM MgCl₂, 2.5 mM GSH using an Amicon Ultra 100K concentrator, and concentrated to ∼1.3–2.5 mg/ml. VCP was mixed with VCPIP1^WT in the same buffer as above to obtain a final concentration of 10 µM VCP monomer and 20 µM VCPIP1. For the AMPPNP dataset, VCP was incubated with 2 mM AMPPNP on ice for five min before the addition of VCPIP1. 3 µl of the mixture was applied to a glow-discharged Quantifoil R1.2/1.3 Au 400-mesh grid, blotted for 2–3 s, and plunge-frozen into liquid ethane using a FEI Vitrobot IV operated at 4°C and 100% humidity.

Cryo-EM datasets were collected on a Titan Krios microscope (FEI), operating at 300 kV and equipped with a Gatan K3 detector and a BioQuantum energy filter in super-resolution mode using SerialEM (Mastronarde, 2005). 7,884 (Apo) or 8,153 (AMPPNP) movies were recorded at a nominal magnification of 105,000×, corresponding to a calibrated pixel size of 0.847 Å (super-resolution pixel size of 0.4235 Å/pixel). Each exposure was fractionated across 40 frames with a total electron dose of 44.6 e⁻/Ų and a total exposure time of 1.6 s (dose rate of 20 e⁻/pixel/sec). The defocus values ranged from −0.5 to −2.0 μm.

(VCP-VCPIP1)_Apo: MotionCor2 was used for dose weighting and to correct interframe movement for each pixel (Zheng et al., 2017). CTF parameters were estimated using CTF estimation in cryoSPARC (v4.4.1 or v4.5.3) (Punjani et al., 2017). Particles were automatically picked (∼2.8 million) and subjected to 2D classification. A subset of classes were selected as references for ab initio reconstruction. To clean particles, we used the resulting reconstruction as a reference for heterogeneous refinement. One class resembling the VCP hexamer (∼780,000 particles) was subjected to nonuniform refinement and 3D classification with a mask for the density below the VCP hexamer (see below). Separately, we performed a local refinement with C3 symmetry followed by symmetry expansion. Symmetry expanded particles (∼2.2 million) were used to perform local refinement and 3D classification in RELION-4.0 (Kimanius et al., 2021). 3D classification was performed using a mask surrounding the asymmetric unit consisting of two adjacent VCP protomers and one VCPIP1 protomer and enabled removal of particles lacking density for VCPIP1. We then performed supervised 3D classification using references with or without density for VCPIP1. The remaining particles that contained density for VCP and VCPIP1 (∼690,000) were subjected to CTF refinement and Bayesian polishing. Finally, we removed duplicates and subsequently performed 3D flexible refinement in cryoSPARC (Punjani and Fleet, 2023) to generate an overall map of the VCP-VCPIP1 complex at ∼3.3 Å resolution.

To improve resolution in the VCP N domain, we subjected C3 symmetry expanded particles to a second focused 3D classification with a mask surrounding the VCP N and D1 domains. Using the class with the strongest density for VCPIP1, we performed a local refinement and duplicates were removed. A final local refinement was performed to generate a map of the VCP N-D1 bound to the VCPIP1 UBX-L at ∼4.4 Å resolution.

To improve resolution in the map containing three VCPIP1 protomers bound to the VCP hexamer, we subjected cleaned particles (∼780,000 particles) to 3D classification with a mask surrounding the three VCPIP1 OTU domains. The class with three VCPIP1 protomers (∼148,000 particles) was selected for nonuniform refinement followed by C3 symmetry expansion. Symmetry expanded particles were used for local refinement to generate a map at ∼3.4 Å resolution.

(VCP-VCPIP1)_AMPPNP: We followed a processing pipeline similar to the one described above but with some modifications. After particles were cleaned and refined (∼1.2 million), 3D classification was performed using a mask focused on the VCP N domain. One class with all N domains in the “up” position (∼110,000 particles) was subjected to C3 symmetry expansion, focused 3D classification, local refinement, and 3D flexible refinement to generate a high-resolution map for the asymmetric unit (∼3.3 Å). Separately, the same particles were subjected to C6 symmetry expansion and focused 3D classification on the VCP N-D1 domains to resolve the VCP N-VCPIP1 UBX-L interaction (∼3.8 Å). Cleaned particles were also subjected to 3D classification with a mask surrounding the density for the VCPIP1 OTU domains to generate a map containing the VCP hexamer bound to three VCPIP1 protomers (∼3.4 Å).

(VCP-VCPIP1)_Apo: VCP (PDB: 5FTL and 7LN6) (Banerjee et al., 2016; Pan et al., 2021) and VCPIP1 (AF-Q96JH7-F1) (Jumper et al., 2021; Varadi et al., 2024) models were segmented into individual domains (VCP: D1 and D2, C-terminal tail residues 764–775; VCPIP1: stalk [residues 556–666]). Each segment was rigid-bodyfit into the map in UCSF Chimera (version 1.18.0) (Pettersen et al., 2004) and flexibly fit using ISOLDE in ChimeraX (version 1.8.0) (Croll, 2018; Meng et al., 2023). The docked domains were joined to generate overall VCP (monomer) and VCPIP1 models. Real-space refinement in Coot (version 0.9.8.5) (Emsley et al., 2010) with planar peptide and transpeptide restraints was used to adjust the backbone and side chains for VCPIP1 and a single VCP chain. The refined coordinates for VCP were replicated and refitted into the second protomer. The model was further refined using Phenix (version 1.21.1–5286) (Afonine et al., 2018) real-space refinement, and problem areas were fixed manually in Coot.

(VCP-VCPIP1)_AMPPNP: Coordinates for the (VCP-VCPIP1)_Apo asymmetric unit were rigid-bodyfit into the (VCP-VCPIP1)_AMPPNP asymmetric unit map in ChimeraX (version 1.8.0) (Croll, 2018; Meng et al., 2023). Real-space refinement in Coot (version 0.9.8.5) (Emsley et al., 2010) with planar peptide and transpeptide restraints was used to adjust the backbone and side chains. The model was further refined using Phenix (version 1.21.1-5286) (Afonine et al., 2018) real-space refinement, and problem areas were fixed manually in Coot.

Various combinations of VCP and VCPIP1 protein sequences were input into the AlphaFold Server (Abramson et al., 2024) as follows: VCP-VCPIP1 asymmetric unit: two copies of VCP residues 1–806, one copy of VCP residues 1–1,222; VCP D2-2 VCPIP1: six copies of VCP residues 471–806, two copies of VCPIP1 residues 112–666; VCP D2-3 VCPIP1: six copies of VCP residues 471–806, three copies of VCPIP1 residues 112–666. No ligands (i.e., nucleotides) were included.

Data were collected using a OneMP mass photometer (Refeyn) calibrated with bovine serum albumin (cat. No. 23210; Thermo Fisher Scientific), β-amylase (cat. No. A8781-1VL; Sigma-Aldrich), and thyroglobulin (cat. No. T9145-1VL; Sigma-Aldrich). Focus was adjusted using filtered (0.22 µm) buffer (50 mM HEPES, pH 7.5, 25 mM KCl, 2.5 mM MgCl₂, 2.5 mM GSH), and then, a 2X solution of protein in filtered buffer was added to the same well, diluting it twofold. Final concentrations of VCP and VCPIP1 (WT or mutants) were 150 nM monomer (25 nM hexamer) and 100 nM monomer, respectively. For conditions with added nucleotide, VCP was incubated with AMPPNP (final concentration 1 mM) on ice for 5 min before the addition of VCPIP1. Movies were acquired for 6,000 frames (60 s) using AcquireMP software (version 2022 R1) and default settings. Raw data were converted to frequency distributions with a bin size of 20 kDa, and Gaussian fits were applied using Prism 10 (GraphPad).

To examine the binding of VCPIP1 to VCP, His-VCP (500 nM monomer) was mixed with VCPIP1^WT or VCPIP1^ΔUBX-L (400 nM) in buffer (50 mM HEPES, pH 7.5, 125 mM KCl, 2.5 mM MgCl₂, 20 mM imidazole, 10 mM βME, 0.01% Tween-20, and 0.1 mg/ml BSA) and incubated at room temperature for 30 min. This mixture was added to His-tag Isolation and Pulldown Dynabeads (cat. No. 10103D; Thermo Fisher Scientific) equilibrated in buffer and incubated at room temperature for an additional 30 min. The supernatant was recovered, and beads were washed three times with buffer to remove unbound proteins. Proteins were eluted from beads using buffer supplemented with 500 mM imidazole. The input, supernatant, and eluate were analyzed by SDS-PAGE and Coomassie staining.

To examine whether VCPIP1 and UN compete for VCP binding, Streptavidin M-270 Dynabeads (cat. No. 65305; Thermo Fisher Scientific) were equilibrated with water followed by buffer (50 mM HEPES, pH 7.5, 25 mM KCl, 2.5 mM MgCl₂, 1 mM DTT, 0.01% Tween-20) before incubation with UN-SBP (200 nM) for 30 min at room temperature. Supernatants were discarded, and beads were washed twice with buffer for 5 min to remove unbound UN-SBP. Beads were incubated with VCP^WT (50 nM) for 30 min at room temperature. The beads were washed twice with buffer to remove unbound VCP prior to a 30-min incubation with or without VCPIP1 variants (VCPIP1^WT or VCPIP1^ΔUBX-L, 50 or 200 nM). 10% input and supernatant samples were collected at this step. Bound complexes were eluted from the beads in 1x SDS-PAGE loading buffer by boiling for 5 min. Samples were analyzed by SDS-PAGE and Coomassie staining. VCP levels in the eluates were quantified using densitometry and calculated relative to the condition where no VCPIP1 was added.

The ATPase activity of VCP was examined using a colorimetric assay (malachite green). VCP (100 nM monomer) was mixed with varying concentrations of VCPIP1 (0–1 μM) in 50 mM HEPES, pH 7.5, 25 mM KCl, 2.5 mM MgCl₂, 1 mM DTT, 0.01% Tween-20, and 0.1 mg/ml BSA. Upon the addition of ATP (200 μM), the reactions were incubated at room temperature for 60 min. An equal volume of malachite green reagent was added to read out phosphate production as absorbance at 620 nm using a Synergy NEO microplate reader. The absorbance from control reactions without VCP was subtracted from the corresponding reactions with VCP at the same concentration of ATP. The activity of VCP in the presence of VCPIP1 was normalized to that of VCP only.

To monitor VCPIP1 DUB activity, VCPIP1^WT (200 nM) was incubated with medium-length polyUb₂-mEos chains (20 nM) and VCP (WT or mutants) (300 nM monomer) in 20 µl of buffer (50 mM HEPES, pH 7.5, 25 mM KCl, 2.5 mM MgCl₂, 1 mM DTT, 0.01% Tween-20) for 60 min at 37°C. The proteolysis of ubiquitin chains was analyzed by western blot using an anti-Ub primary antibody (cat. No. sc-8017; Santa Cruz Biotechnology). IRDye-conjugated secondary antibody raised in goat (cat. No. 925–68070; LI-COR Biosciences) was used according to the manufacturer’s instructions.

Fig. S1 shows recombinant proteins used in this study and a summary of the cryo-EM workflow and data for the (VCP-VCPIP1)_Apo complex. Fig. S2 shows additional cryo-EM, MP, and biochemical analyses of the VCP-VCPIP1 complex. Fig. S3 shows a summary of the cryo-EM workflow and data for the (VCP-VCPIP1)_AMPPNP complex. Fig. S4 shows AlphaFold confidence metrics (predicted local distance difference test scores and predicted aligned error plots) for models used in this study. Table S1 shows single-particle cryo-EM data collection, refinement, and validation statistics.

The cryo-EM density maps and corresponding atomic models generated in this study have been deposited in the Protein Data Bank (PDB) and Electron Microscopy Data Bank (EMDB) under accession codes: (VCP-VCPIP1)_Apo asymmetric unit (PDB: 9DIL and EMDB: 46912); and (VCP-VCPIP1)_AMPPNP asymmetric unit (PDB: 9MQ6 and EMDB: 48514). All other data generated or analyzed during this study are presented in the manuscript and its supplementary information files. Source data are provided with this paper.

We thank T.-F. Chou (Caltech, Pasadena, CA, USA) for providing the H. sapiens VCP plasmid and A. Martin (University of California, Berkeley, Berkeley, CA, USA) for providing His₆-Ub^G76V-Ub^G76V-mEos3.2 and gp78-Ubc7 plasmids.

T.M. Kapoor is grateful to the National Institutes of Health (NIH) for supporting this research (GM1300234). L.E. Vostal was supported in part by the NIH T32 Chemistry-Biology Interface Training Grant to the Tri-Institutional PhD Program in Chemical Biology (GM115327 and GM136640). N.E. Dahan was supported in part by the NIH T32 Genetics and Cell Biology Training Grant (TGM144299). We are grateful to M. Ebrahim, J. Sotiris, H. Ng, and the Evelyn Gruss Lipper Cryo-Electron Microscopy Resource Center for cryo-EM support.

Author contributions: L.E. Vostal: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, validation, visualization, and writing—original draft, review, and editing. N.E. Dahan: conceptualization, formal analysis, investigation, methodology, resources, and writing—review and editing. M.J. Reynolds: investigation. L.I. Kronenberg: investigation and writing—review and editing. T.M. Kapoor: conceptualization, funding acquisition, project administration, resources, supervision, validation, and writing—original draft, review, and editing.