Binding of Substrates to the Central Pore of the Vps4 ATPase Is Autoinhibited by the Microtubule Interacting and Trafficking (MIT) Domain and Activated by MIT Interacting Motifs (MIMs)*

Background: The Vps4 ATPase powers the endosomal sorting complexes required for transport (ESCRT) pathway. Results: Peptide binding to hexameric Vps4 is promoted by nucleotides that can mimic ADP, ATP, and the transition state. Conclusion: ESCRT-III substrates bind Vps4 MIT domains and then bind the central pore of an asymmetric, nucleotide-bound Vps4 hexamer. Significance: Mechanistic understanding of Vps4-substrate interactions is advanced by this work. The endosomal sorting complexes required for transport (ESCRT) pathway drives reverse topology membrane fission events within multiple cellular pathways, including cytokinesis, multivesicular body biogenesis, repair of the plasma membrane, nuclear membrane vesicle formation, and HIV budding. The AAA ATPase Vps4 is recruited to membrane necks shortly before fission, where it catalyzes disassembly of the ESCRT-III lattice. The N-terminal Vps4 microtubule-interacting and trafficking (MIT) domains initially bind the C-terminal MIT-interacting motifs (MIMs) of ESCRT-III subunits, but it is unclear how the enzyme then remodels these substrates in response to ATP hydrolysis. Here, we report quantitative binding studies that demonstrate that residues from helix 5 of the Vps2p subunit of ESCRT-III bind to the central pore of an asymmetric Vps4p hexamer in a manner that is dependent upon the presence of flexible nucleotide analogs that can mimic multiple states in the ATP hydrolysis cycle. We also find that substrate engagement is autoinhibited by the Vps4p MIT domain and that this inhibition is relieved by binding of either Type 1 or Type 2 MIM elements, which bind the Vps4p MIT domain through different interfaces. These observations support the model that Vps4 substrates are initially recruited by an MIM-MIT interaction that activates the Vps4 central pore to engage substrates and generate force, thereby triggering ESCRT-III disassembly.

The endosomal sorting complexes required for transport (ESCRT) pathway drives reverse topology membrane fission events within multiple cellular pathways, including cytokinesis, multivesicular body biogenesis, repair of the plasma membrane, nuclear membrane vesicle formation, and HIV budding. The AAA ATPase Vps4 is recruited to membrane necks shortly before fission, where it catalyzes disassembly of the ESCRT-III lattice. The N-terminal Vps4 microtubule-interacting and trafficking (MIT) domains initially bind the C-terminal MIT-interacting motifs (MIMs) of ESCRT-III subunits, but it is unclear how the enzyme then remodels these substrates in response to ATP hydrolysis. Here, we report quantitative binding studies that demonstrate that residues from helix 5 of the Vps2p subunit of ESCRT-III bind to the central pore of an asymmetric Vps4p hexamer in a manner that is dependent upon the presence of flexible nucleotide analogs that can mimic multiple states in the ATP hydrolysis cycle. We also find that substrate engagement is autoinhibited by the Vps4p MIT domain and that this inhibition is relieved by binding of either Type 1 or Type 2 MIM elements, which bind the Vps4p MIT domain through different interfaces. These observations support the model that Vps4 substrates are initially recruited by an MIM-MIT interaction that activates the Vps4 central pore to engage substrates and generate force, thereby triggering ESCRT-III disassembly.
Eukaryotic Vps4 enzymes comprise the N-terminal MIT domain, an ϳ40-residue linker, a two-domain AAA ATPase cassette, a "␤-domain" that is inserted within the small domain of the ATPase cassette, and a C-terminal helix that binds against the large ATPase domain. Vps4 functions as a higherorder oligomer. Although the subunit stoichiometry and structure have been controversial (33)(34)(35)(36), we have shown that the active Vps4 enzyme is a hexamer (37), like all other well characterized Type I AAA ATPases (38,39). Our working structural model for the Vps4 hexamer is based on superposition of the known crystal structure of the Vps4 ATPase cassette (33, 37, 40 -43) onto the structure of the p97 D1 hexamer (44) and is supported by mutational analysis of proposed hexameric interface residues (37,40,41). This model places the Vps4 ␤-domains on the periphery, where they can bind the VSL domain of the Vta1p/LIP5 cofactor that promotes Vps4 assembly and stimulates ATPase activity (32,40,(45)(46)(47)(48)(49). The Vps4 hexamer has a central pore that is lined by two conserved loops: pore loop 1 and pore loop 2. Vps4 pore loop 1 displays an aromatic hydrophobic dipeptide that is conserved across AAA ATPases that have polypeptide substrates (40,41), and pore loop 2 contains a series of charged residues that are conserved and functionally important in the related ATPase spastin (50,51). Mutations in either Vps4 pore loop inhibit HIV budding (40,41), although these mutations may also destabilize the Vps4 hexamer to some extent.
Hanson and colleagues (52) have reported that overexpressed human ESCRT-III subunits CHMP2A and CHMP1B that lack their terminal MIM elements nevertheless co-sediment with human VPS4B(E235Q) from cell lysate. This interaction requires ESCRT-III helix 5 and surrounding loops, and this same region also contributes to stimulating VPS4A ATP hydrolysis when it is present within C-terminal fragments of ESCRT-III proteins (53). These observations suggest that the ESCRT-III helix 5 region may bind preferentially to the pore of the Vps4 hexamer. To investigate this idea further, we characterized the Vps4-ESCRT-III interaction using purified recombinant Saccharomyces cerevisiae proteins Vps4p and Vps2p (CHMP2 homolog). We found that the AAA ATPase cassette of a Vps4p construct that lacked the MIT domain formed a stable hexamer with the flexible nucleotide mimics ADP⅐AlF x or ADP⅐BeF x , and that this hexameric complex bound a single Vps2p helix 5 peptide in an interaction that was mediated by the Vps4p pore loop residues. Vps2p helix 5 binding was autoinhibited by the MIT domains, and this inhibition was alleviated by association of the MIT domain with an MIM1 or an MIM2 sequence. These observations, together with previous studies (21,(52)(53)(54), support a model in which ESCRT-III complexes are disassembled by pulling ESCRT-III helix 5 into the central pore of asymmetric Vps4p hexamers in a manner that is initially primed and activated by binding of MIT domains to ESCRT-III MIM sequences.

Experimental Procedures
Proteins and Peptides-Vps4p and the VSL domain of Vta1p (Vta1p VSL , residues 280 -330) were expressed and purified as described (35,37,40,41). Bacterial expression vectors are listed in Table 1. Mutations were generated by QuikChange mutagenesis and verified by DNA sequencing. Peptides were synthesized on a Prelude peptide synthesizer (Protein Technologies, Inc.) using standard Fmoc chemistry (55). Following cleavage from resin, the peptides were precipitated with icecold ether, washed thoroughly with ether, dissolved in water/ acetonitrile, and lyophilized for long-term storage. 5(6)-Carboxyfluorescein (Acros Organics) was used to fluorescently label the peptides.
For N-terminally labeled peptides, the fluorescein group was coupled to the N-terminal ␣-amine by standard coupling conditions. All other amines within the peptide were t-butyloxycarbonyl-protected. For C-terminally labeled peptides, an orthogonal lysine deprotection method was used. At the first position of the peptide (C terminus), Dde-Lys(Fmoc)-OH (AAPPTec) was coupled to the resin. Fmoc removal was performed with 20% piperidine, followed by fluorescein coupling to the ⑀-amine on the lysine side chain. The Dde group (on the lysine ␣-amine) was then removed with 3% hydrazine in N,Ndimethylformamide, and the remaining peptide was synthesized using standard protocols. Unlabeled N termini were blocked with an acetyl group, and unlabeled C termini were amides. Peptide quality was verified by reversed phase C18 HPLC. For experiments described below, peptide concentrations were determined by infrared spectroscopy using a Direct Detect spectrometer (EMS Millipore). The identities of all peptides and recombinant proteins were verified by mass spectrometry.
Analytical Size-exclusion Chromatography-A Superdex 200 column was equilibrated in running buffer (10 mM MgCl 2 , 100 mM NaCl, 20 mM HEPES, pH 7.0, and when appropriate, 3 mM nucleotide). To prepare ADP metal fluoride stock solutions, 1 M NaF was added dropwise to a 10 mM solution of ADP to a final concentration of 62.5 mM with stirring, and then 1 M AlCl 3 or BeCl 2 was slowly added to a final concentration of 12.5 mM. Vps4p proteins (200 or 50 M in subunit concentration, corre- sponding to 33 or 8.3 M hexamer, respectively) were preincubated with a 2-fold molar (subunit) excess of Vta1p VSL in running buffer at 4°C for at least 2 h. The column was calibrated using protein standards (Bio-Rad).
Fluorescence Anisotropy Binding Assays-To generate the data shown in Figs. 1B, 3A, 4B, and 6A, a dilution series of Vps4p-Vta1p VSL complex (Vta1p VSL was in 2-fold excess over Vps4p subunits, total volume 60 l) in binding buffer (10 mM MgCl 2 , 100 mM NaCl, 20 mM HEPES, pH 7.0, and where indicated, 3 mM nucleotide) was incubated with fluorescein-labeled peptides (1 nM) at 25°C for at least 3 h before measuring parallel and perpendicular fluorescence intensity using excitation/ emission wavelengths of 485/535 nm on a Tecan Infinite 200 microplate reader. Anisotropy values were calculated and plotted against the Vps4p hexamer concentration. Dissociation constants (K D values) were estimated by fitting the data to the where FA is the normalized fluorescence anisotropy and corresponds to "fraction bound," using GraphPad Prism 5 (GraphPad Software, Inc.). Error bars show the S.D. from three independent experiments and are shown for every data point in the figures.
To generate the data shown in Fig. 2, fluorescence anisotropy was measured during titration of Vps4p ⌬MIT -Vta1p VSL hexamer into 40 M peptide C in the binding buffer (above). Fraction bound was plotted versus the [Vps4p ⌬MIT hexamer]/[peptide] ratio. At these high peptide concentrations (17-fold over K D ), the stoichiometry of peptide binding to the Vps4p ⌬MIT hexamer can be estimated by the ratio at which saturation binding was reached. Data are shown for three independent experiments performed with protein from independent preparations.
GST Pulldown Assays-An N-terminal GST fusion of Vps2p residues 165-201 was used for pulldown assays. The fusion protein was expressed from a pET151-D-Topo vector with an N-terminal tobacco etch virus protease-cleavable His 6 tag and purified using the procedure described above for Vps4p proteins. Purified Vps4p at the indicated subunit concentration, Vta1p VSL (2-fold excess over Vps4p), and GST fusion proteins (1 M) were mixed in wash buffer (10 mM MgCl 2 , 100 mM NaCl, 20 mM HEPES, pH 7.0, with or without 3 mM ADP⅐AlF x ) and incubated for 3 h at 4°C. 20 l of glutathione agarose (Amersham Biosciences) was added to 1 ml of protein samples and incubated for 1 h at 4°C. Unbound proteins were removed by washing with wash buffer (four washes with 1 ml of buffer). The glutathione resin was subsequently resuspended in SDS-PAGE loading buffer and incubated for 5 min at 95°C, and bound proteins were visualized by SDS-PAGE.
ATPase Assay-To generate the data shown in Fig. 5A, Vps4p ⌬MIT (final concentration 0.3 M subunits, corresponding to 50 nM hexamer) was mixed with Vta1p VSL (2-fold excess over Vps4p subunits) and 2 mM ATP in reaction buffer (100 mM NaCl, 10 mM MgCl 2 , 1 mM DTT, and 20 mM HEPES, pH 7.5), together with the indicated concentrations of Vps2p peptide C. After 10 min, 2 volumes of malachite green solution (14 mM ammonium molybdate, 1.3 M HCl, 1.5 mM malachite green) were added, followed by 21% citric acid (same volume as the malachite green solution). Absorbance at 650 nm was mea-sured using a POLARstar OPTIMA (BMG Labtech) plate reader, and phosphate concentrations were calculated using a sodium phosphate standard curve. To generate the data shown in Fig. 4C, Vps4p ⌬MIT and its pore loop 2 mutants (final concentration 0.5 M subunits) were mixed with Vta1p VSL (2-fold excess over Vps4p subunits), and their ATPase activity was measured as described above.

Results
Vps2p Peptides Bind the ATPase Cassette of an Asymmetric Vps4p Hexamer-Quantitative fluorescence anisotropy assays were performed to test whether peptides that spanned the C-terminal end of Vps2p helix 5 (Fig. 1A) bound a pure recombinant protein complex comprising a Vps4p construct that lacked the N-terminal MIT domain (Vps4p ⌬MIT ; residues 81-437) and the Vta1p VSL domain (Vta1p VSL , residues 280 -330). Tight peptide binding (K D ϭ 2-10 M) was observed in the presence of ADP⅐AlF x or ADP⅐BeF x , but not in the presence of ATP, ADP, AMPPNP, or ATP␥S (Fig. 1B). Binding specificity was validated by the observation that a control peptide corresponding to the downstream Vps2p MIM1 motif (residues 218 -232) bound 15-fold less tightly than Vps2p helix 5 peptides A and C, and by the demonstration that different versions of peptide A that were fluorescently labeled at the N or the C terminus bound equivalently (within 2-fold, data not shown).
Peptide binding affinities to the Vps4p ⌬MIT -Vta1p VSL complex in the presence of ADP⅐AlF x were calculated assuming that Vps4p ⌬MIT is present as a hexamer, and that one Vps4p hexamer binds one peptide (see below). Peptides A (residues 165-201; K D ϭ 2.5 Ϯ 0.5 M) and C (residues 165-184; K D ϭ 2.4 Ϯ 0.5 M) bound modestly more tightly than peptide B (residues 144 -184; K D ϭ 6.1 Ϯ 2.4 M) (Fig. 1B), indicating that the minimal peptide C construct contains the primary binding site, and suggesting that the patch of highly acidic residues within the N-terminal extension of peptide B may weaken the binding slightly.
The unique ability of ADP⅐AlF x and ADP⅐BeF x to support binding likely reflects the ability of these nucleotide analogs to mimic multiple states, including ATP, ADP, and transition state-bound nucleotide complexes (56 -59), and suggests that different subunits within each Vps4p hexamer may need to bind different nucleotide states and adopt different conformations to generate a stable, asymmetric substrate-binding site. Consistent with this idea, hexameric Vps4p ⌬MIT -Vta1p VSL complexes were more stable in the presence of ADP⅐AlF x or ADP⅐BeF x than in the presence of ATP, ADP, or ATP␥S, or in the absence of nucleotide, as shown by size-exclusion chromatography (note the relative mobilities of the different complexes at 50 M Vps4p ⌬MIT subunit concentrations, corresponding to 8.3 M hexamer concentrations, Fig. 1C). Thus, Vps4p ⌬MIT is hexameric in the concentration range used for the fluorescence anisotropy binding assay in the presence of ADP⅐AlF x or ADP⅐BeF x , and the fluorescence anisotropy signal can be used as a readout to quantify peptide binding, especially since substrate binding further stabilizes hexamer formation (see below).
The Vps4p Hexamer Binds a Single Peptide-To verify the stoichiometry of binding, we performed a fluorescence anisot-ropy binding experiment in which Vps4p ⌬MIT -Vta1p VSL hexamers in the presence of ADP⅐AlF x were titrated into a solution of fluorescently labeled peptide C at high concentration (ϳ20fold above the K D ). Peptide binding saturated when the stoichiometry was close to one peptide per Vps4p hexamer (Fig. 2), supporting the idea that each asymmetric Vps4p hexamer contains a single binding site for ESCRT-III helix 5.
Pore Loop Residues Bind Directly to Vps2p-derived Peptides-To determine whether residues of the Vps4p hexamer pore contribute to binding, we tested binding to pore loop mutants that have been previously shown to inhibit HIV budding when equivalent mutations are present in human Vps4 homologs (40,41). These correspond to W206A (pore loop 1), R241A (pore loop 2), and R251A (adjacent to pore loop 2) of S. cerevisiae Vps4p (40,41). None of these mutants bound peptides with appreciable affinity (Fig. 3A). However, gel filtration experiments, performed at 100 M Vps4p ⌬MIT subunit concentration (17 M hexamer) for a more sensitive readout of hexamer dissociation, indicated that the hexamer stability of these mutants was reduced (Fig. 3B).
To identify pore loop mutations that do not destabilize the Vps4p hexamer, we substituted 10 different pore loop residues individually with alanines and assayed oligomerization by analytical size-exclusion chromatography at a Vps4p subunit concentration of 50 M (8.3 M hexamer). At this concentration, the wild-type Vps4p ⌬MIT protein elutes as a stable hexamer in  2. One Vps4p ⌬MIT hexamer binds one Vps2p helix 5 peptide. Fluorescence anisotropy was measured during titration of 40 M (ϳ20-fold over K D ) peptide C with the Vps4p ⌬MIT -Vta1p VSL hexamer. Saturation of the signal occurred when the stoichiometry of the complex was close to 1:1. Data from three independent experiments are shown here, and the indicated stoichiometry represents the ratio of Vps4p ⌬MIT -Vta1p VSL hexamer to peptide C at saturation when all data sets were fit globally. MAY (Figs. 1C and 4A). Three of the Vps4p substitution mutants (T240A, E243A, and E247A) displayed essentially unchanged oligomerization, even at low protein concentrations (Fig. 4A). These pore loop alanine substitution mutants were therefore assayed for peptide binding. Remarkably, introducing E243A or E247A point mutations increased the affinity of peptide C binding (by 3-and 4-fold, respectively, Fig. 4B). Moreover, the double mutant, Vps4p ⌬MIT (E243A, E247A), bound peptides 4-fold (peptide C) or 10-fold (peptide B, not shown) more tightly than wild-type Vps4p (Fig. 4B, lower  panel). Because different alanine pore loop substitutions can either increase or decrease peptide binding affinity (Fig. 4B) without substantially changing Vps4p hexamerization, we con-

Vps4-Substrate Interactions
clude that substrates bind the pore loops directly. This conclusion is reinforced by the observation that the double mutation, which gives the tightest binding, actually reduces hexamerization slightly (Fig. 4A, lower panel), thereby emphasizing that the mutations uncouple reduction of binding from reduction of hexamerization. Similar to the effect of removing acidic pore loop residues, we also find that introducing a positive charge (T240K) enhances peptide binding (Fig. 4B, top panel) without affecting oligomerization (Fig. 4A, top panel), whereas the T240F mutation has no effect.
We also tested the effects of the pore loop 2 glutamate mutations on the basal ATPase activity of Vps4p ⌬MIT . As shown in Fig. 4C, both the E243A and the E247A mutations reduced ATPase activity, but the double mutation did not further reduce ATPase activity (p value between E243A and the double mutant is 0.17). Although hexamerization was not impaired when assayed by gel filtration, a possible explanation is that reduced ATPase activity reflects reduced hexamerization at the concentration used to assay ATP hydrolysis (0.3 M versus 50 M subunit concentrations), whereas the increased peptide binding reflects improved interactions between the pore loop residues and the substrate. Alternatively, the reduced activity of these mutants may reflect allosteric coupling of the enzymatic active site to the central peptide-binding site (and altered coupling in the mutants).
Peptide Binding Stimulates ATPase Activity and Stabilizes Hexamerization-A previous study showed that VPS4A ATPase activity is stimulated by C-terminal ESCRT-III constructs that span the MIM motif and helix 5 (53). Having mapped relatively high affinity binding to the helix 5 region of Vps2p, we tested whether this pore-binding element could also stimulate Vps4p ⌬MIT ATPase activity. As shown in Fig. 5A, the addition of peptide C stimulated the Vps4p ⌬MIT ATPase activity by 2.5-fold at the highest concentration tested. To test the possibility that this effect resulted from increased stabilization of the Vps4p hexamer in the presence of peptide C, we characterized the oligomerization of Vps4p ⌬MIT in the presence of Vta1p VSL and ADP⅐AlF x and in the presence or absence of peptide C. Analytical gel filtration was performed at (low) Vps4p concentrations where the hexameric complex was appreciably dissociated (10 M Vps4p ⌬MIT subunits, 1.7 M hexamer). Under these conditions, Vps4p ⌬MIT eluted as a hexamer in the presence but not the absence of peptide C (Fig. 5B). Thus, peptide binding stabilizes the Vps4p hexamer. We note that ATPase activity was assayed at a concentration where inactive monomers and dimers of Vps4p ⌬MIT likely predominated over active hexamers, and the observed peptide stimulation of ATPase activity is therefore likely due, at least in part, to substrate-mediated stabilization of the Vps4p ⌬MIT hexamer.
The MIT Domain Inhibits Substrate Binding in the Absence of an MIM Interaction-To determine whether the MIT domains influence substrate binding to the Vps4p pore, we compared binding of the A, B, and C peptides to various Vps4p constructs (Fig. 6B) and found that full-length Vps4p binds each of these peptides with ϳ10-fold lower affinity than does Vps4p ⌬MIT (Fig. 6, A and C). To determine whether this inhibitory effect of the MIT domain could be alleviated by MIM binding, we designed a Vps4p variant (MIM1-Vps4p) that included an N-terminal extension of the Vps2p MIM1 motif (residues 218 -232) covalently linked to the MIT domain through a (GGGGS) 3 linker that was long enough to allow an intramolecular MIM1-MIT interaction (Fig. 6B). The equivalent construct bearing the MIM1 L225D mutation (MIM1(L225D)-Vps4p) was used as a negative control because this mutation disrupts the interaction between MIM1 and MIT (26). The MIM1-Vps4p fusion and Vps4p ⌬MIT proteins bound peptides with similar affinities, whereas the binding affinity of MIM1(L225D)-Vps4p was similar to that of full-length Vps4p (Fig. 6, A and C). Additional validation was provided by a GST affinity co-purification pulldown assay using Vps2p helix 5 residues fused to GST. This construct bound both Vps4p ⌬MIT and MIM1-Vps4p, whereas binding to both Vps4p and MIM1(L225D)-Vps4p was greatly diminished (Fig. 7). These data indicate that MIM1 binding alleviates inhibition of substrate pore binding by the MIT domains. This effect does not appear to reflect MIT domain inhibition of Vps4p oligomerization because MIM1-Vps4p and MIM1(L225D)-Vps4p displayed equivalent migration on gel filtration chromatography in the presence of ADP⅐AlF x (Fig. 6D). Essentially identical results were found for binding of peptide C to MIM2-Vps4p constructs (Vfa1p residues 183-203) (Fig. 6C) that were designed in light of the recently reported Vfa1p MIM2-MIT complex structure (60).

Discussion
ESCRT-mediated membrane fission requires remodeling and depolymerization of ESCRT-III filaments by the Vps4 AAA ATPase, which is recruited, at least in part, by binding of its N-terminal MIT domain to C-terminal ESCRT-III MIM elements. Earlier studies indicated that an additional interaction occurs for human VPS4A between the ATPase cassette and sequences from the helix 5 region of ESCRT-III subunits, that deletion of the MIT domain and linker sequence enhances ATPase activity in human VPS4A, and that mutation of pore loop residues could either increase or reduce ESCRT-III-induced ATPase activity (52,53). We have verified and extended these findings by performing quantitative binding studies using purified reagents, delineating effects due to peptide binding versus those due to stabilization or destabilization of the active hexamer, and demonstrating effects of MIM1 and MIM2 interactions on substrate engagement. The earlier studies on VPS4A were prompted in part by the relatively low ATPase activity of human VPS4 proteins (53), which suggested that fundamental differences in Vps4 regulation might exist between species but might be explained by the different assay conditions such as the presence/absence of Vta1/LIP5. In addition to providing new insights, our work using the S. cerevisiae Vps2p and Vps4p proteins therefore indicates that key principles of Vps4 regulation are likely to be conserved across species.
Our fluorescence anisotropy studies demonstrated that a 20-residue peptide from the helix 5 region of Vps2p/ESCRT-III binds the hexameric Vps4p AAA ATPase cassette with an affinity similar to that of the primary contact between the isolated Vps4p MIT domain and the Vps2p/ESCRT-III MIM (albeit without enhancing avidity effects). This interaction seems unlikely to have absolute sequence specificity, and we anticipate that many protein sequences could similarly engage the Vps4 ATPase cassette, consistent with the reports that overlapping fragments of multiple ESCRT-III family members are able to co-sediment with VPS4B(E235Q) (52) and stimulate ATPase activity of human VPS4A (53) or yeast Vps4p (61,62). We also demonstrated that binding occurs with a stoichiometry of one peptide per hexamer. This further validates the conclusion that Vps4 is active as a hexamer (37) and supports the model that substrates bind in the central Vps4 pore. We also found that peptide binding stimulates ATPase activity, at least in part by stabilizing the hexamer.
Although residues of the pore loops have previously been implicated in Vps4 function (40,41), it was formally possible that these mutations inhibited activity by reducing the stability of the active Vps4 oligomer. We have now identified point mutants in Vps4p pore loop 2 that increased binding without decreasing hexamerization, thereby solidifying the idea that substrates bind in the central pore of the hexamer. We were

Vps4-Substrate Interactions
unable to perform equivalent experiments with pore loop 1 mutants, however, because although we identified many pore loop 1 mutations that diminished binding, they also diminished hexamerization, at least to some degree.
An important conclusion from our work is that Vps4p functions as an asymmetric hexamer that translocates substrate into the central pore. This follows from our observations that a single peptide binds per hexamer, which necessarily requires an asymmetric interface, and is dependent upon the presence of ADP⅐AlF x or ADP⅐BeF x , which allow the ATPase to adopt multiple nucleotide-bound states. This is consistent with models of other ring-like ATPases, including F 1 ATPase (63), E1 helicase (64), Rho helicase (59), the 26S proteasome (65,66), the bacterial enhancer-binding protein NtrC1 (58), and N-ethylmaleimide-sensitive factor (NSF) (67), which all display asymmetry of the subunits that results in a spiral staircase arrangement of the pore loops. An attractive model is that Vps4 translocates substrate, at least partially, through the central pore, driven by changes in pore loop conformations that propagate around the hexamer ring in concert with the ATP hydrolysis cycle.
Peptide binding was not observed with ADP or ATP, which presumably gave a mixture of ATP and ADP because the binding assay was performed at a temperature that allows for hydrolysis, and nor was binding observed in the presence of nonhydrolyzable analogs of ATP. Presumably, ADP metal fluorides enable the formation of asymmetric, peptide-bound complexes but do not drive translocation, whereas binding under conditions of active ATP turnover is transient and therefore not captured by our assay. The transcriptional activator PspF, a related AAAϩ ATPase, also requires ADP⅐AlF x to form a complex with 54 that can be detected in vitro (68).
Our observations that the pore loop mutants E243A and E247A bind more tightly (especially for peptide B) indicate that the negatively charged pore residues Glu-243 and Glu-247 diminish binding of peptides rich in acidic residues (such as peptide B). Thus, the pore is not optimized to bind all peptides equally or with maximal binding affinity. Our findings are somewhat discordant with conclusions from pioneering work from the Hanson laboratory, which provided motivation for this study but also concluded that acidic residues preferentially bind (52) and stimulate ATPase activity (53) of human VPS4A. Regardless, the overall impression is that there is a pore-selective-binding element located within helix 5 of Vps2p, but also that different amino acid sequences can bind the Vps4p pore with appreciable affinity, which is consistent with the idea that once an ESCRT-III subunit is engaged, it can be processively translocated through the hexamer pore. In principle, this action could translocate the ESCRT-III substrates completely, although the extent of translocation remains to be determined experimentally.
In contrast to human VPS4A (53), deletion of the MIT domain does not seem to increase the basal rate of ATP hydrolysis for yeast Vps4p (36,61,62). However, our observation that the MIT domain inhibits peptide binding to the hexamer pore in the absence of an MIT-MIM interaction suggests a two-step model of Vps4-mediated ESCRT-III lattice disassembly (Fig. 8). We propose that the MIM-MIT interaction serves to recruit Vps4 to ESCRT-III polymers and to unmask or activate the binding site at the hexamer pore, which possesses an inherent ability to bind and translocate a wide variety of sequences starting at an internal loop, such as has been described for the proteasome (69,70). An attractive feature of this model is that it provides a mechanism for avoiding Vps4 pore engagement of inappropriate substrates. The mechanism of MIT inhibition is unclear at this time, and might include direct occlusion or indirect stabilization of an inhibited pore conformation. It is also unclear whether all of the MIT domains in a Vps4 hexamer need to be engaged by MIMs to allow binding at the pore and stimulate ATPase activity, although it is attractive to speculate that multiple MIT-MIM interactions are required because that would provide an effective mechanism for restricting activity to ESCRT-III polymers. This model provides a foundation for future structural and mechanistic studies to understand in more detail how Vps4 functions in membrane fission and the dissociation of ESCRT-III filaments.