The hexameric AAA ATPase Vps4 drives membrane fission by remodeling and disassembling ESCRT-III filaments. Building upon our earlier 4.3 Å resolution cryo-EM structure (Monroe et al., 2017), we now report a 3.2 Å structure of Vps4 bound to an ESCRT-III peptide substrate. The new structure reveals that the peptide approximates a β-strand conformation whose helical symmetry matches that of the five Vps4 subunits it contacts directly. Adjacent Vps4 subunits make equivalent interactions with successive substrate dipeptides through two distinct classes of side chain binding pockets formed primarily by Vps4 pore loop 1. These pockets accommodate a wide range of residues, while main chain hydrogen bonds may help dictate substrate-binding orientation. The structure supports a ‘conveyor belt’ model of translocation in which ATP binding allows a Vps4 subunit to join the growing end of the helix and engage the substrate, while hydrolysis and release promotes helix disassembly and substrate release at the lagging end.https://doi.org/10.7554/eLife.31324.001
The Endosomal Sorting Complexes Required for Transport (ESCRT) pathway drives multiple cellular membrane fission processes (Christ et al., 2017; Scourfield and Martin-Serrano, 2017) through the formation of filaments comprising different subsets of related ESCRT-III family members. ESCRT-III filaments stabilize highly curved membrane necks that resolve by fission when the filaments are remodeled by Vps4 (Monroe and Hill, 2016). Continued Vps4 activity removes ESCRT-III subunits and drives complete filament disassembly, thereby enabling subsequent rounds of ESCRT activity (Mierzwa et al., 2017; Schöneberg et al., 2017).
Vps4 is monomeric or dimeric at cytoplasmic concentrations, but forms a hexamer when active and recruited to ESCRT-III filaments (Monroe et al., 2014). Recruitment is mediated, at least in part, by binding of the Vps4 MIT (Microtubule Interacting and Transport) domains to MIM (MIT Interacting Motif) elements in the exposed tails of ESCRT-III subunits (Obita et al., 2007; Stuchell-Brereton et al., 2007; Hurley and Yang, 2008). Hexamerization is further promoted by the cofactor protein Vta1/LIP5 (Scott et al., 2005; Lottridge et al., 2006; Azmi et al., 2008; Xiao et al., 2008), whose VSL domain binds adjacent Vps4 subunits at the ring periphery (Yang and Hurley, 2010; Davies et al., 2014; Monroe et al., 2017; Sun et al., 2017). The Vps4 N-terminal MIT domain is followed by an ~40 residue flexible linker and an ATPase cassette that comprises a large ATPase domain, small ATPase domain, and a β domain (Scott et al., 2005).
Recently reported cryo-EM structures of Vps4 at overall resolutions of 4.3 Å (Monroe et al., 2017), 6.1 Å (Su et al., 2017), and 3.9 Å (Sun et al., 2017) revealed similar hexameric ‘lock washers’, in which five of the six Vps4 subunits form a helical assembly and the sixth closes the ring. This arrangement is quite different from the packing seen in multiple Vps4 crystal structures. Although the three cryo-EM structures are similar, they prompted very different mechanistic models to explain how ESCRT-III subunits are processed. Our structure, which was visualized in complex with a substrate peptide, guided the proposal that ESCRT-III substrates bind within the hexamer pore (Monroe et al., 2017), consistent with a model that ESCRT-III subunits are unfolded by translocation through the pore (Yang et al., 2015). In contrast, the observation that the Vps4 hexamer adopts ‘open’ and ‘closed’ states in the absence of Vta1/LIP5 and substrate, prompted the proposal that substrates are engaged by a single ESCRT-III subunit, and that ATP hydrolysis pulls an entire ESCRT-III subunit from the filament and positions the next Vps4 subunit to remove the ensuing ESCRT-III subunit (Su et al., 2017).
We have now determined the Vps4-ESCRT-IIIpeptide complex structure at 3.2 Å resolution. This structure shows that the highly variable side chains of the substrate bind in an equivalent way to an array of two classes of pockets that are repeated throughout the Vps4 pore. Moreover, the ESCRT-III peptide binds in a β-strand conformation in one orientation. These insights support our earlier mechanistic proposal, which may be applicable to other AAA ATPases.
The new reconstruction of the Vps4-Vta1-ESCRT-IIIpeptide-ADP·BeFx complex agrees well with our earlier 4.3 Å resolution reconstruction but now has an overall resolution of 3.2 Å, which likely reflects the use of superior microscope instrumentation (see Materials and methods). The six Vps4 subunits form a closed ring, with six Vta1 VSL domains binding around the periphery and a single ESCRT-III peptide bound in the central pore (Figure 1, Table 1, Figure 1—figure supplements 1–4, Figure 1—video 1). Vps4 subunits A-E form a helix whose symmetry approximates a 60° rotation and 6.3 Å translation between adjacent subunits, and provides the binding surface for the ESCRT-III peptide. As noted previously (Monroe et al., 2017), subunit E deviates slightly from the more exact helical symmetry of subunits A-D. The large domain of subunit F is disengaged from adjacent Vps4 subunits (and substrate) and appears to be transitioning between the two ends of the Vps4 helix. Subunit F and features at the ring periphery, including the Vps4 β domains and the Vta1 VSL domains, have weak density.
Nucleotides primarily contact one Vps4 subunit at a subunit interface, with the β-phosphate and BeFx contacting two ‘finger’ arginines, R288 and R289, from the following subunit (Figure 2, Figure 2—videos 1–5). The bound ADP·BeFx mimics ATP binding to subunits A, B, and C, whereas subunits D and E appear to bind ADP. The density of subunit F is too weak to reliably assess the presence of nucleotide, and it may be empty. The coordination of bound nucleotides is similar for all Vps4 subunits, and resembles binding to other ATPases (Wendler et al., 2012). ADP·BeFx coordination is essentially identical for A-C and is very similar for the ADP at subunit D, whereas displacement of subunit F results in loss of interaction with the finger arginines for the ADP at subunit E. In contrast to earlier proposals (Gonciarz et al., 2008), the hinge angle between large and small domains does not change substantially with the bound nucleotide, being 120–121° for subunits A-D, 117° for subunit E, and 122° for subunit F.
The AB, BC, and CD interfaces are extensive, similar to each other, and include large domain contacts with adjacent large domains and small domains (Figure 2—videos 6,7). Comparison of these interfaces with the DE subunit pair shows an ~8° rotation of subunit E which, nevertheless, maintains very similar contacts with the preceding D subunit and relative positions of pore loops (Figure 2—video 8). It is therefore uncertain why the AB, BC, and CD interfaces bind ATP (ADP·BeFx) while the similar DE interface appears to bind ADP. Although the density seems clear, we acknowledge that assigning ADP (vs. ADP·BeFx) at this site with absolute certainty will require higher resolution data. Regardless, the finding that subunit E binds ADP at the interface with subunit F further supports our mechanistic model that ATP hydrolysis at subunit D destabilizes the subunit interface to promote formation of the more open nucleotide binding site seen for subunit E (below).
In contrast to the major disruption in contacts between the large domain of subunit F and the large domains of its neighbors, contacts involving the small ATPase domains remain more similar. This is especially true for the FA interface, where the F small domain to A large domain contacts are closely superimposable with those of AB, BC, CD, and DE (Figure 2—video 9). The E small domain to F large domain contacts are mediated by the same hydrophobic contacts, albeit with a rotation of ~25° and attendant shifts of 4.5–7.5 Å. Thus, this interaction appears to be maintained throughout the reaction cycle as subunits transition from the lagging (subunit E) to the leading (subunit A) end of the Vps4 helix, and may help maintain the hexameric assembly while the hydrolysis and release of ATP disrupts the core large domain contacts that define the ESCRT-III substrate-binding site.
Interestingly, the small domain-large domain interface is the major lattice contact in all of the reported crystal structures of Vps4, which totals 22 crystallographically unique contacts (Figure 2—video 10). The archaeal Vps4 crystal contacts (Monroe et al., 2014; Caillat et al., 2015) closely resemble the FA/AB/BC/CD/DE interface, while eukaryotic crystal contacts (Scott et al., 2005; Xiao et al., 2007; Gonciarz et al., 2008; Inoue et al., 2008) overlap with the EF interface, except two of the three mouse Vps4 crystal contacts (Inoue et al., 2008), which are intermediate between the two states. These differences correlate with the presence of a Vps4 β domain and Vta1 cofactor in eukaryotes but not archaea, supporting the interpretation that Vta1 promotes formation and maintenance of the closed helical substrate-binding assembly of Vps4.
The DEIVNKVL ESCRT-III peptide is clearly defined in the new 3.2 Å map (Figure 3—video 1), with the exception of the first two side chains, D1 and E2 (Vps2 D165 and E166), which have weak density, as is typically seen for carboxylates in cryo-EM maps (Mitsuoka et al., 1999; Bartesaghi et al., 2014; Yonekura et al., 2015; Hryc et al., 2017). The assigned peptide orientation was validated by building and refining in the reverse direction, which showed a correlation coefficient of 0.85 for the assigned orientation vs. 0.81 in the reversed orientation, and EMRinger (Barad et al., 2015) scores of 3.7 for the assigned orientation vs. 1.1 for the reversed orientation (Figure 3—figure supplement 1).
The peptide adopts an extended conformation that resembles a canonical β-strand (phi −92° to −151°; psi 102° to 189° (−171°)) and packs closely against Vps4 subunits A-E (Figure 3, Figure 3—videos 2–4). Two distinct classes of side chain binding sites propagate along the pore. Odd-numbered ESCRT-III residues (D1, I3, N5, V7) bind in ‘class I’ pockets, while the side chains of even-numbered ESCRT-III residues (E2, V4, K6, L8) bind in ‘class II’ pockets. Class I pockets are formed by pore loop 1 residues K205 and W206, with substrate side chains sandwiched between W206 from successive subunits. The pocket is flanked by K205 from the first Vps4 subunit, which may also stabilize the ladder of W206 side chains through cation-π interactions. .
The class II pockets are formed by the pore loop 1 M207 residues of successive subunits, and are flanked by pore loop 2 residues of the first Vps4 subunit and, to a greater extent, by the pore loop 2 residues of the preceding subunit. The first of these pockets, which includes M207 residues of the A and B subunits, is incompletely formed because the pore loop 2 residues of the preceding subunit (F) are not yet in position. The three pore loop 2 residues that most closely approach the peptide (E245, S246, E247) do not make distinctive contacts with the ESCRT-III peptide. S246 caps the N-terminus of a Vps4 helix while E245 and E247 make non-specific contacts with peptide side chains. The different class I and class II pockets are all essentially identical (Figure 3CD) and their exposure to the highly solvated pore explains how polar side chains can be accommodated.
The refined model indicates that the NH groups of the even-numbered ESCRT-III peptide residues form hydrogen bonds with the main chain O of Vps4 K205 (subunit A, B, C, and D distances are 3.1–3.5 Å). In contrast, these distances are 3.4–4.5 Å in the model refined with the peptide in the reverse orientation. Thus, substrate NH hydrogen bonds may help define the substrate orientation by optimally positioning ESCRT-III side chains with respect to their binding pockets (Figure 3E).
The structure supports our earlier conveyor-belt model of substrate translocation (Monroe et al., 2017) but now includes more detail, especially of substrate binding. It is also consistent with other recent models (Gates et al., 2017; Monroe et al., 2017; Puchades et al., 2017; Ripstein et al., 2017), albeit with additional molecular details. We envision that the enzyme proceeds by transitioning of subunit configurations around the hexameric ring, such that one step represents transitioning of subunits F, A, B, C, D, E to the configurations of subunits A, B, C, D, E, F (Figure 4A, Figure 4—video 1). Each step comprises concomitant changes in subunit interfaces at each end of the Vps4 A-E helix. At the leading end, binding of ATP allows subunit F to pack against subunit A and thereby bind the next two residues of the substrate. At the lagging end, hydrolysis of ATP and subsequent phosphate release destabilizes the interface and drives subunit E to the transitioning configuration, thereby opening the nucleotide binding site to allow exchange of ADP by ATP. In this manner, the equivalent dipeptide-binding sites formed at the interfaces of subunits A-E track with their bound dipeptides along the pore as subunits transition from the lagging end to the leading end of the helix. Thus, the pathway can be represented either as Vps4 ‘walking’ along the substrate or as substrate being translocated through the Vps4 pore, depending upon the frame of reference.
Although our structure shows four intact dipeptide-binding sites and just one transitioning subunit, it is expected that two Vps4 subunits will be disengaged from substrate through at least part of the reaction cycle. This is because subunit E will disengage from substrate as it moves toward the F configuration while the F subunit is moving toward the substrate-binding A configuration. The specific point in the reaction cycle captured in our structure corresponds to subunit E just starting to transition toward the F state, as indicated by the slight deviation of the DE interface from the configuration of AB, BC, and CD, and consistent with our structural interpretation that ADP is bound to subunit D (above). The stability of this state relative to other conformations along the reaction coordinate will result from a combination of multiple factors, including the particular nucleotide (ADP·BeFx) and ESCRT-III peptide in the complex.
Our model implies that each step translocates two amino acid residues and hydrolyses one ATP molecule. It also explains the importance of the pore loop residues and the integral role of the class I and II pockets, because once a substrate residue binds at the top of the Vps4 helix it does not substantially change conformation until released at the bottom of the helical conveyor belt. The model is also consistent with reports that other AAA+ ATPases bind a maximum of three or four ATP molecules (Hersch et al., 2005; Horwitz et al., 2007; Yakamavich et al., 2008; Smith et al., 2011). Moreover, despite important structural differences, our mechanistic model is analogous to mechanisms established for translocation of DNA by the E1 helicase (Enemark and Joshua-Tor, 2006) and of RNA by the Rho translocase (Thomsen and Berger, 2009; Thomsen et al., 2016).
Vps4 is a member of the classic clade of AAA+ ATPases, which includes the original members of the AAA ATPase family (Iyer et al., 2004; Erzberger and Berger, 2006). These proteins are hexameric protein translocases, whose conserved pore loops emanate from equivalent structural elements, although their N-terminal domains are variable and only Vps4 has a β domain. Family members are found in a variety of contexts. For example, Vps4 is a homohexamer that forms a single ring, whereas p97/CDC48/VAT and NSF comprise two AAA ATPase cassettes that each form a hexameric ring. Others, such as HSP104/ClpB, form a double ATPase ring structure in which only one of the rings belongs to the classic clade. Still others, such as the eukaryotic proteasome ATPases, form part of a much larger complex and comprise six different, albeit related, subunits.
Sequence alignment of classic clade AAA+ ATPase family members for which structures have been reported shows that the first two of the three pore loop 1 residues that contact substrate (Vps4 K205, W206, M207) are conserved (Figure 4—figure supplement 1). W206 is always W, F, or Y, all of which could perform the role of sandwiching substrate side chains. The preceding K205 residue is usually K but sometimes M, which like K could flank the class I binding pocket. The more variable third residue, M207, is usually I, L or V, but there are examples of M (Vps4), A, F, K, R, and Y, all of which could provide hydrophobic sides to the class II binding pocket. Interestingly, the following residue, G208, is invariant, presumably because it adopts phi angles (~81 to 85°) that are only favored for glycine, and helps define a conformation that can pack against pore loop 1 residues from neighboring subunits. Pore loop 2 residues are less conserved and are typically disordered in other published structures, which is consistent with the lack of distinctive roles in contacting substrate.
Our Vps4 structure superimposes closely with the recently reported structures of VAT (Ripstein et al., 2017), HSP104 (Gates et al., 2017), and ClpB (Deville et al., 2017), which were each determined with substrate bound in their central pores, albeit at relatively low resolutions. A very recently published 3.4 Å structure of the YME1 AAA ATPase in complex with a mixed polypeptide substrate (Puchades et al., 2017) also presents a very similar structure and mechanism to that described here, although coordinates are not yet available during final preparation of this manuscript. Close superposition is also seen with multiple other classic clade AAA+ ATPases that have been visualized in the absence of substrate (Figure 4B, Figure 4—video 2). Notable exceptions include p97 and CDC48, whose structures display rotational rather than helical symmetry (Banerjee et al., 2016; Xia et al., 2016), and ClpX and HslU, which are not members of the classic clade but share notable similarities with Vps4 and have provided leading mechanistic models for protein translocating AAA+ ATPases (Olivares et al., 2016).
In contrast to these sequential models, biochemical analyses of ClpX (Martin et al., 2005) and HslU (Baytshtok et al., 2017) have argued for stochastic mechanisms by showing that just one active ATPase subunit per hexamer can drive translocation, albeit at much reduced efficiency. A possible resolution is that a hexamer that has only one active ATPase site may allow inactive subunits to diffuse through the entire helical cycle. Thus, like a 6-cylinder engine firing on just one cylinder, a single active ATPase might drive the sequential conveyor-belt model, albeit rather poorly.
In summary, the structure shows details of substrate interactions that are provided by a repeating array of dipeptide binding sites with the ESCRT-III peptide in a unique orientation. It further supports a sequential mechanism and explains the ability to translocate polypeptides with little sequence specificity. Important future priorities include testing the generality of the structural observations, mechanistic implications, and the extent to which they may apply to other AAA+ ATPases.
Sample preparation was as described (Monroe et al., 2017). Vitrified grids were loaded onto a Titan Krios (FEI) operating at 300 kV. Images were acquired using a defocus range between −1.0 to −2.2 µm. A total of 2,349 cryo-EM movies were recorded using a K2 Summit direct detector (Gatan) in counting mode with a pixel size of 1.10 Å and at a dose rate of ~7.4 e-/pixel/sec. Each movie was recorded as a stack of 40 frames accumulated over 10 s, totaling ~62e-/Å2.
Movie frames were aligned, dose weighted, and summed using MotionCor2 (Zheng et al., 2017) (Figure 1—figure supplement 1A). CTF parameters were determined on non-dose-weighted sums using gctf (Zhang, 2016). Micrographs with poor CTF cross correlation scores were excluded from downstream analyses. A total of 1,987 dose-weighted sums were used for all subsequent image processing steps. 4,429 particles were manually selected from 30 micrographs in EMAN2 using the e2boxer.py program (Tang et al., 2007) to generate preliminary 2D classes in RELION (Scheres, 2012). The non-CTF-corrected class averages were used for template-based autopicking in gautomatch. A total of 599,085 particles were extracted and used as input for full CTF-corrected image processing (Figure 1—figure supplement 1B). After multiple rounds of 2D classification, 124,743 particles were retained based on visual inspection of classes with high-resolution Vps4 features and used for an initial round of 3D classification. After 3D classification, 109,241 particles were used for RELION auto-refinement (Scheres, 2012), which generated a 4.1 Å density map of the Hcp1-Vps4 fusion complex based on the gold-standard FSC criterion (Figure 1—figure supplement 1D). To improve the resolution of Vps4, we performed signal subtraction of Hcp1 densities using the same approach as described previously (Bai et al., 2015; Monroe et al., 2017) (Figure 1—figure supplement 2). After Hcp1 signal subtraction, we performed an additional round of 3D classification, which assigned 82,225 particles into a single class with excellent Vps4 features. These particles were used for a final round of RELION auto-refinement, producing a 3.2 Å resolution density map of Vps4 (Figure 1—figure supplement 2). B-factor sharpening of −125 Å2 was applied using an automated procedure in RELION postprocessing (Rosenthal and Henderson, 2003). Local resolutions were estimated using ResMap (Kucukelbir et al., 2014) (Figure 1—figure supplement 1G).
The consensus reconstruction of the Vps4 complex revealed poor, fragmented densities for subunit F and Vta1. To improve their densities, we performed additional rounds of focused classification by generating custom soft-edged masks around their respective densities and then using RELION to classify the particles without re-alignment (Figure 1—figure supplement 3). Particles from classes with ordered densities were used for separate RELION auto-refinement reconstructions and produced lower-resolution maps that were used for rigid-body fitting of subunit F or Vta1.
Model building and refinement followed the same approach as for the earlier lower resolution structure (Monroe et al., 2017). NCS restraints were applied to Vps4 subunits A-E with the exception of residues 204–208, 227–233, and 249–271 of subunit A and residues 140–158, 171–191, and 204–205 of subunit E. For subunits A, B, C, and D, the distance between the Mg2+ and the OG of 180S was restrained to 2.0 Å. No reference model was used during refinement. The refined model was assessed using MolProbity (RRID: SCR_014226) (Chen et al., 2010) and EMRinger (Barad et al., 2015). To test for overfitting, all atoms in the refined model were randomly displaced by 0.5 Å and re-refined against one of the RELION half maps. FSC curves were generated for the re-refined model against the half map used for re-refinement (FSCwork) and against the other half map (FSCtest). The close agreement between the two curves is consistent with lack of overfitting (Figure 1—figure supplement 1E).
To validate the orientation of the 8-residue ESCRT-III peptide, it was built and refined in opposing conformations (Figure 3—figure supplement 1). RSCC scores between the models and density map were determined using UCSF Chimera (Pettersen et al., 2004). Side chain-directed model versus map calculations for the peptide were performed using EMRinger (Barad et al., 2015).
The refined model comprising the Vps4 ATPase domains of subunits A-E and ESCRT-III peptide has been deposited into the PDB (RRID: SCR_012820; PDB ID: 6BMF). The complete model, including regions not subjected to atomic refinement such as the 12 Vta1VSL domains and subunit F, has been deposited into the PDB (PDB ID: 6AP1) together with the sharpened Hcp1-subtracted map (RRID: SCR_003207, EMDB Accession Number EMD-8887). The unsharpened map, the two maps for subunit F, and the six maps for the Vta1 VSL domain have been deposited at the EMDB (RRID: SCR_003207, EMDB Accession Number(s) EMD-8888, EMD-8889, EMD-8890, EMD-8891, EMD-8892, EMD-8893, EMD-8894, EMD-8895, EMD-8896).
Covalently linked HslU hexamers support a probabilistic mechanism that links ATP hydrolysis to protein unfolding and translocationJournal of Biological Chemistry 292:5695–5704.https://doi.org/10.1074/jbc.M116.768978
Asymmetric ring structure of Vps4 required for ESCRT-III disassemblyNature Communications 6:8781.https://doi.org/10.1038/ncomms9781
MolProbity: all-atom structure validation for macromolecular crystallographyActa Crystallographica Section D Biological Crystallography 66:12–21.https://doi.org/10.1107/S0907444909042073
Cellular functions and molecular mechanisms of the ESCRT membrane-scission machineryTrends in Biochemical Sciences 42:42–56.https://doi.org/10.1016/j.tibs.2016.08.016
Vps4 stimulatory element of the cofactor Vta1 contacts the ATPase Vps4 α7 and α9 to stimulate ATP hydrolysisJournal of Biological Chemistry 289:28707–28718.https://doi.org/10.1074/jbc.M114.580696
Evolutionary relationships and structural mechanisms of AAA+ proteinsAnnual Review of Biophysics and Biomolecular Structure 35:93–114.https://doi.org/10.1146/annurev.biophys.35.040405.101933
Biochemical and structural studies of yeast Vps4 oligomerizationJournal of Molecular Biology 384:878–895.https://doi.org/10.1016/j.jmb.2008.09.066
ATP-induced structural transitions in PAN, the proteasome-regulatory ATPase complex in ArchaeaJournal of Biological Chemistry 282:22921–22929.https://doi.org/10.1074/jbc.M702846200
Evolutionary history and higher order classification of AAA+ ATPasesJournal of Structural Biology 146:11–31.https://doi.org/10.1016/j.jsb.2003.10.010
Quantifying the local resolution of cryo-EM density mapsNature Methods 11:63–65.https://doi.org/10.1038/nmeth.2727
The oligomeric state of the active Vps4 AAA ATPaseJournal of Molecular Biology 426:510–525.https://doi.org/10.1016/j.jmb.2013.09.043
Meiotic clade AAA ATPases: protein polymer disassembly machinesJournal of Molecular Biology 428:1897–1911.https://doi.org/10.1016/j.jmb.2015.11.004
Mechanistic insights into bacterial AAA+ proteases and protein-remodelling machinesNature Reviews Microbiology 14:33–44.https://doi.org/10.1038/nrmicro.2015.4
UCSF Chimera--a visualization system for exploratory research and analysisJournal of Computational Chemistry 25:1605–1612.https://doi.org/10.1002/jcc.20084
RELION: implementation of a Bayesian approach to cryo-EM structure determinationJournal of Structural Biology 180:519–530.https://doi.org/10.1016/j.jsb.2012.09.006
Reverse-topology membrane scission by the ESCRT proteinsNature Reviews Molecular Cell Biology 18:5–17.https://doi.org/10.1038/nrm.2016.121
Structural and mechanistic studies of VPS4 proteinsThe EMBO Journal 24:3658–3669.https://doi.org/10.1038/sj.emboj.7600818
Growing functions of the ESCRT machinery in cell biology and viral replicationBiochemical Society Transactions 45:613–634.https://doi.org/10.1042/BST20160479
Mechanism of Vps4 hexamer function revealed by cryo-EMScience Advances 3:e1700325.https://doi.org/10.1126/sciadv.1700325
EMAN2: an extensible image processing suite for electron microscopyJournal of Structural Biology 157:38–46.https://doi.org/10.1016/j.jsb.2006.05.009
Structure and function of the AAA+ nucleotide binding pocketBiochimica et Biophysica Acta (BBA) - Molecular Cell Research 1823:2–14.https://doi.org/10.1016/j.bbamcr.2011.06.014
Structural characterization of the ATPase reaction cycle of endosomal AAA protein Vps4Journal of Molecular Biology 374:655–670.https://doi.org/10.1016/j.jmb.2007.09.067
Asymmetric nucleotide transactions of the HslUV proteaseJournal of Molecular Biology 380:946–957.https://doi.org/10.1016/j.jmb.2008.05.070
Vps4 disassembles an ESCRT-III filament by global unfolding and processive translocationNature Structural & Molecular Biology 22:492–498.https://doi.org/10.1038/nsmb.3015
Katanin spiral and ring structures shed light on power stroke for microtubule severingNature Structural & Molecular Biology 24:717–725.https://doi.org/10.1038/nsmb.3448
Andreas MartinReviewing Editor; University of California, Berkeley, United States
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your article "The AAA ATPase Vps4 binds substrates through a repeating array of dipeptide-binding pockets" for consideration by eLife. Your article has been favorably evaluated by Michael Marletta (Senior Editor) and three reviewers, one of whom, Andreas Martin (Reviewer #1), is a member of our Board of Reviewing Editors. The following individual involved in review of your submission has agreed to reveal their identity: Gabriel C Lander (Reviewer #3).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
In this manuscript, Sundquist, Hill, and colleagues present the 3.2 Å cryo-EM structure of yeast Vps4 bound to an ESCRT-III peptide. It builds on their paper published in eLife earlier this year that provided a 4.3 Å structure of Vps4 bound to nucleotides and a substrate peptide, and proposed a spiral staircase model for the continuous conveyance of substrate through the central pore. In this previous structure, the peptide density showed contacts with multiple subunits via their pore loops, but did not reveal side-chain density or the orientation of the bound peptide.
A Krios data set now improved the resolution to 3.2 Å, which allowed the visualization of peptide side chains as well as pore loop contacts, and the proposal of a more detailed translocation model where every Vps4 subunit interacts with two amino acids through two "binding pockets" formed by the pore-1 and pore-2 loops. The new structure is thus consistent with the mechanisms the authors proposed previously, but does not reveal major new conceptual insights.
The reviewers were a bit surprised to see that there are really not many new data in this submission albeit now having a structure at 1.1 Å higher resolution, and that the authors did not take advantage of this higher resolution to address and answer important outstanding questions.
Even though it is exciting to directly visualize substrate interactions of Vps4, the advance by this manuscript in its current form is rather limited and considered not sufficient for publication in eLife. The higher resolution data should enable the authors to pursue further functional exploration and reveal new conceptual insights that go beyond of what had already been established previously for Vps4 and other AAA motors. The manuscript thus needs major revisions before publication.
1) The proposal of a directionality for ESCRT substrate translocation by Vps4 is interesting, but should be further supported by experimental data. Several related AAA motors have been shown to not have a preference for substrate orientation in the central pore, and for the conclusions here it should be considered that the described trapped structure likely represents an energy minimum, where the Vps4 subunits might arrange around the stably bound peptide. It is possible that a peptide bound in the opposite orientation could induce equivalent contacts, and that there is in fact no preference in the direction of translocation. The authors may be able to explore this through biochemical assays or structurally, for instance by using peptides of different lengths or with the inverted amino-acid sequence relative to the current one.
2) The presented translocation model also does not go significantly beyond what had been previously discussed. The authors describe a correlation between nucleotide state and subunit conformation relative to the substrate, but the structural relationship remains unexplored in the manuscript. The impressive resolution of this structure should enable the authors to explain how ATP hydrolysis and release translates to specific rearrangements of the subunits or pore loops. Reviewers were also not fully convinced by the authors' assignment of ADP- vs. ATP- (or ADP·BeFx) bound subunits, especially for the DE interface, and the presented model for ATPase-driven substrate translocation. As the authors know, reliably distinguishing between nucleotide states has been very difficult for other AAA hexamers in recent high-resolution cryo-EM studies, in part due to weaker density for beta and gamma phosphates, but also due to issues when averaging thousands of particles with potentially distinct nucleotide occupancies in some of the ATPase subunits. For the present study, it is indeed quite unexpected that neither the Mg ion nor the P-loop residues show significant changes in their orientation, and that the Arg fingers are just slightly shifted when comparing supposedly ADP·BeFx -bound and ADP-bound subunits. When discussing this surprising similarity of subunit interfaces, the authors bring up the possibility that their nucleotide assignments are wrong. The authors also try to come up with a rationale for how the very small structural differences may lead to preferential binding of ADP vs. ATP (or ADP·BeFx), but ignore in their Discussion that during normal motor function the nucleotide dictates subunit conformation, not the other way around. Several biochemical and structural studies previously suggested major conformational and functional differences between ATP- and ADP-subunits, and that the transition from ATP- to ADP-bound through hydrolysis and phosphate release may drive the power stroke of AAA motors. Interestingly, the authors claim that substantial conformational adjustments are not required to pass through the DE states. But what then is the force-generating step for translocation, if subunits can transition from ATP to ADP-bound without any major conformational changes? ADP release or ATP binding to an empty Vps4 subunit? In a revised manuscript, the authors should assess and discuss which step of the ATPase cycle likely induces conformational changes and drives the conveyer belt for forceful substrate translocation. In that context it should also be discussed how constrains in the topologically closed ring could restrict conformational changes in response to hydrolysis, which may explain the authors' findings that ADP and ADP·BeFx-bound subunits don't show major conformational differences.
3) The reviewers have expressed concerns regarding the atomic model refinement. Only statistics for chains A-E are listed in Table 1, but the deposited models contain all 6 subunits in addition to the substrate. The validation report from the PDB for the full deposition reveals statistics that are much worse than those reported in Table 1, and the selective inclusion of only chain A-E is rather misleading.https://doi.org/10.7554/eLife.31324.058
- Han Han
- Nicole Monroe
- Wesley I Sundquist
- Peter S Shen
- Christopher P Hill
- Nicole Monroe
- Nicole Monroe
- Wesley I Sundquist
- Nicole Monroe
- Wesley I Sundquist
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Electron microscopy was performed at the National Resource for Automated Molecular Microscopy and the Simons Electron Microscopy Center located at the New York Structural Biology Center, supported by grants from NIH (GM103310, S10 OD019994) and the Simons Foundation (349247), with additional support from the Agouron Institute (F00316). We thank Zhening Zhang for assistance in data collection. The Center for High Performance Computing at the University of Utah provided computing resources. We thank James Fulcher and Michael Kay for peptide synthesis, and Frank Whitby for help with model refinement.
- Andreas Martin, Reviewing Editor, University of California, Berkeley, United States
© 2017, Han et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.