Vps4 Stimulatory Element of the Cofactor Vta1 Contacts the ATPase Vps4 α7 and α9 to Stimulate ATP Hydrolysis*

Background: ESCRT-III can enhance Vta1 stimulation of Vps4 ATPase activity via the Vps4 stimulatory element (VSE) of Vta1. Results: α7 and α9 of the Vps4 small AAA domain mediate VSE stimulation, contributing to Vps4 function in vivo. Conclusion: Vta1 contacts Vps4 α7 and α9 during ESCRT-III-enhanced stimulation of Vps4. Significance: These studies identify a novel mechanism of Vps4 stimulation. The endosomal sorting complexes required for transport (ESCRTs) function in a variety of membrane remodeling processes including multivesicular body sorting, abscission during cytokinesis, budding of enveloped viruses, and repair of the plasma membrane. Vps4 ATPase activity modulates ESCRT function and is itself modulated by its cofactor Vta1 and its substrate ESCRT-III. The carboxyl-terminal Vta1/SBP-1/Lip5 (VSL) domain of Vta1 binds to the Vps4 β-domain to promote Vps4 oligomerization-dependent ATP hydrolysis. Additionally, the Vps4 stimulatory element (VSE) of Vta1 contributes to enhancing Vps4 oligomer ATP hydrolysis. The VSE is also required for Vta1-dependent stimulation of Vps4 by ESCRT-III subunits. However, the manner by which the Vta1 VSE contributes to Vps4 activation is unknown. Existing structural data were used to generate a model of the Vta1 VSE in complex with Vps4. This model implicated residues within the small ATPase associated with various activities (AAA) domain, specifically α-helices 7 and 9, as relevant contact sites. Rational generation of Vps4 mutants defective for VSE-mediated stimulation, as well as intergenic compensatory mutations, support the validity of this model. These findings have uncovered the Vps4 surface responsible for coordinating ESCRT-III-stimulated Vta1 input during ESCRT function and identified a novel mechanism of Vps4 stimulation.

The endosomal sorting complexes required for transport (ESCRTs) function in a variety of membrane remodeling processes including multivesicular body sorting, abscission during cytokinesis, budding of enveloped viruses, and repair of the plasma membrane. Vps4 ATPase activity modulates ESCRT function and is itself modulated by its cofactor Vta1 and its substrate ESCRT-III. The carboxyl-terminal Vta1/SBP-1/Lip5 (VSL) domain of Vta1 binds to the Vps4 ␤-domain to promote Vps4 oligomerization-dependent ATP hydrolysis. Additionally, the Vps4 stimulatory element (VSE) of Vta1 contributes to enhancing Vps4 oligomer ATP hydrolysis. The VSE is also required for Vta1-dependent stimulation of Vps4 by ESCRT-III subunits. However, the manner by which the Vta1 VSE contributes to Vps4 activation is unknown. Existing structural data were used to generate a model of the Vta1 VSE in complex with Vps4. This model implicated residues within the small ATPase associated with various activities (AAA) domain, specifically ␣-helices 7 and 9, as relevant contact sites. Rational generation of Vps4 mutants defective for VSE-mediated stimulation, as well as intergenic compensatory mutations, support the validity of this model. These findings have uncovered the Vps4 surface responsible for coordinating ESCRT-III-stimulated Vta1 input during ESCRT function and identified a novel mechanism of Vps4 stimulation.
The endosomal sorting complexes required for transport (ESCRTs) 2 participate in multivesicular body (MVB) sorting in the endocytic pathway, abscission of the cellular bridge during cytokinesis, the budding of enveloped viruses, and the repair of small wounds in the plasma membrane (1)(2)(3)(4)(5). Most of these processes are similar in that cytosolic machinery (i.e. ESCRTs) effects membrane scission from within the neck of a vesicle budding away from the cytoplasm (e.g. MVB sorting, viral budding) or from within a membrane tubule (e.g. abscission), and this class of processes have been termed "reverse topology" membrane fission events (6). The sorting of cargo into the MVB pathway is the best understood of these ESCRT-dependent phenomena. Cargoes destined for inclusion into the MVB pathway are covalently modified with ubiquitin (7)(8)(9). These modified cargoes are recognized by ubiquitin-binding domains in the early ESCRTs (ESCRT-0, -I, and -II) and sequestered into endosomal microdomains that bud into the endosome as intralumenal vesicles (10,11). Intralumenal vesicle formation results from the activity of ESCRT-III and associated factors (e.g. Bro1⅐Doa4 and Vps4⅐Vta1 complexes) that interact on the endosomal membrane (12,13). ESCRT-III and associated factors are responsible for coordinating intralumenal vesicle formation with recycling of ubiquitin and disassembly of ESCRT-III (14 -16). ESCRTs function on the cytoplasmic face of the endosome to drive membrane vesiculation into the endosomal lumen.
The Vta1 VSE has been implicated in mediating ESCRT-IIIenhanced Vta1 stimulation of Vps4 (53), but the surface of Vps4 that is contacted by the VSE to effect increased ATP hydrolysis is unknown. Structures of the Vta1 carboxyl terminus dimer, including both the VSL and VSE (43), and the VSL dimer in complex with a Vps4 fragment containing the ␤-domain and part of the small AAA domain (55) have been determined. This information was used to model VSE interaction with Vps4 to understand Vta1 VSE stimulation (see Fig. 1B). Three residues (located on ␣7 and ␣9) of the small AAA domain were identified as a potential interaction surface: methionine 330, leucine 407, and lysine 411. Models of the Vps4 oligomer (49,50) place these residues on the outer surface of the proposed oligomer where they would be accessible to the Vta1 VSE. Vps4 mutants with these residues altered exhibited concentration-dependent ATPase activity comparable with that of WT Vps4 and were responsive to direct stimulation by ESCRT-III, indicating that these mutants do not globally compromise Vps4 structure and function. However, these Vps4 mutants exhibited defects in VSE-mediated activation by Vta1. Vps4(L407K) exhibited the most severe defects in stimulation by the Vta1 VSE, but stimulation could be restored with compensatory mutation of the Vta1 VSE. Vps4(L407K) also exhibited deficits in the maturation of the MVB cargo CPS and recycling of ESCRT-III subunit Snf7 in vivo. These observations support a model in which the Vta1 VSE contacts the Vps4 small AAA domain via ␣7 and ␣9 to mediate ESCRT-III enhancement of Vps4⅐Vta1, thereby promoting ATP hydrolysis and coordinating ESCRT function.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-pET28 Vps4 was generated by PCRamplifying the VPS4 ORF (along with 252 bp of sequence 3Ј to the STOP codon) with NdeI (5Ј) and SalI (3Ј) restriction sites and subcloning the resulting fragment into the NdeI and SalI sites of pET28b (Novagen). Mutagenesis of Vps4 was performed using the Gene Tailor site-directed mutagenesis system (Invitrogen). The NcoI to SalI fragments of these pET28 Vps4 mutants were then subcloned into the pMB28 (VPS4 in pRS416) (52) after shuttling through pRS415. All cloned PCR products and mutant plasmids were sequenced to exclude unexpected mutations. The generation of pGST-Vta1, pET28-Vta1, Vta1 fragments, Vta1(VSE⌬) mutant, and pET28-Ist1 (WT and L168A,Y172A) expression plasmids have been previously described (26,40,53). The BY4742 vps4⌬::NEO strain was obtained from Open Biosystems.
ATPase Assay-Measurement of Vps4 ATPase activity was performed as previously described (34,40). Vps4 protein concentration was assessed using a Bio-Rad protein assay, and the stock was diluted to 1 or 2 M in ATPase reaction buffer (0.1 M KOAc, 20 mM Hepes, and 5 mM MgOAc, pH 7.5). Portions of 1 or 2 M dilutions were checked by SDS-PAGE and Coomassie staining to normalize for differences between WT and mutant Vps4 proteins during rate calculations. Titration of Vps4 (0.2-1.5 M) was performed by diluting appropriate amounts of 2 M Vps4 in ATPase buffer for a total of 18 l. Reactions were initiated by the addition of ATP/[␣-32 P]ATP to 4 mM. 1-l samples were removed at 4, 8, 12, and 16 min after ATP/[␣-32 P]ATP addition and resolved by thin layer chromatography using precoated PEI Cellulose TLC glass plates (Merck) and developing buffer (0.75 M KPO 4 , pH 3.5). The plates were dried, exposed to phosphorimaging screens for 12-16 h, processed using the Typhoon FLA 7000 system (GE Life Sciences), and ADP and ATP signal was quantitated using ImageQuant software package (GE Life Sciences). The percentage of ATP hydrolysis was used to calculate ADP generated per Vps4 molecule (using corrected Vps4 concentration) per min. Vps4 (0.25 M) activity in the presence of Vta1 proteins (8 M), a concentration previously demonstrated to be saturating (53), or Ist1 (10 M, with 0.5 M Vps4) were determined similarly with the additional step that 18-l reactions were preincubated on ice for at least 30 min prior to initiation at 30°. Reactions were performed in duplicate or triplicate within experiments, and rates presented are from experiments performed on at least 2 separate days. The error bars indicate standard error from the mean. The significances of differences in rates were assessed by T tests using Prism5 (GraphPad). An example of this analysis including images of TLC plates, and determination of rates has been presented in Norgan et al. (53).
GST Pulldown Assays-GST and GST-Vta1 were coupled to glutathione-Sepharose 4B beads (GE Healthcare) in PBS with 0.05% Tween 20 (PBST). Beads were washed with PBST and with ATPase reaction buffer supplemented with 0.05% Tween 20 and 1 mM ATP. GST-or GST-Vta1-coupled beads were dispensed into 0.8-ml microspin columns (Pierce) along WT or mutant Vps4 proteins at a final concentration of 10 ng/l. Incubations were performed with 1 mM ATP at 4°C for 1 h. Samples were then spun briefly and washed three times with ATPase reaction buffer containing 0.05% Tween 20 and 1 mM ATP. Samples were eluted with 5ϫ Laemmli sample buffer, resolved by SDS-PAGE, and analyzed by staining with Coomassie Blue or Western blotting with Vps4 antisera (33).
Biochemical Analyses-Analysis of CPS transport to the vacuole by pulse-chase immunoprecipitation was performed as previously described (17). The significances of differences in maturation rates were assessed by two-way analysis of variance tests using Prism5 (GraphPad). Subcellular fractionation was performed as previously described (17,53). Briefly, 5 OD equivalents of yeast spheroplasts were resuspended at 10 OD/ml in lysis buffer (50 mM Tris, pH 7.5, 200 mM sorbitol, 2 mM EDTA with protease inhibitors), lysed by 15 strokes in a Dounce homogenizer, and subjected to a 10 min, 13,000 ϫ g spin at 4°C to separate the S13 and P13 fractions. Samples (0.04 OD equivalents) were resolved by SDS-PAGE and Western blotted for Snf7 (polyclonal antibody, 1:5,000), phosphoglycerate kinase (mAb, 1:2000) (Invitrogen), and Pep12 (mAb, 1:2000) (Invitrogen). Western blots were developed using both film and the UVP Autochemi System (Upland, CA), and quantitation was performed using ImageQuantTL software (GE Life Sciences). The error bars indicate standard error from the mean. The significances of differences in rates were assessed by t tests using Prism5 (GraphPad). Vps4 protein levels in yeast were assessed by harvesting 5 OD equivalents of rapidly growing yeast, precipitating proteins with addition of TCA to 10%, and lysing samples in 5ϫ Laemmli sample buffer with glass beads. Samples (0.5 OD equivalents) were resolved by SDS-PAGE and Western blotted for Vps4 (polyclonal antibody, 1:1,000) and phosphoglycerate kinase (mAb, 1:10,000) (Invitrogen

Residues within the Vps4 Small ATPase Domain Mediate
Vta1 VSE Activation-ESCRT-III can stimulate Vps4 ATPase activity directly (31,51,52) or indirectly through binding Vta1 and derepressing activity of the Vta1 VSE (31,53). However, the mechanisms by which these inputs are transmitted to the Vps4 AAA domain to alter ATP hydrolysis are unclear. We sought to understand this phenomenon by addressing the Vps4 AAA domain surface contacted by the Vta1 VSE. The structure of the Vta1 VSL domain dimer in complex with the small AAA and ␤-domains of Vps4 (55) was utilized to constrain a region of Vps4 that may be contacted by VSL-proximal Vta1 VSE residues ( Fig. 1A 1B). In this model, Vta1 residues implicated for the stimulatory activity of the VSE (Leu-284, Ile-287, and Met-288) (53) are proximal to Vps4 residues Met-330 and Leu-407 within the small AAA domain. Additionally, a third residue within the small AAA domain of Vps4 (Lys-411) appeared to interact with Ser-292 of Vta1, although mutation of this Vta1 residue did not compromise Vta1 stimulation of Vps4 (53). These three Vps4 residues fall onto two helices, ␣7 (Met-330) and ␣9 (Leu-407 and Lys-411), with ␣9 more proximal to the ␤-domain. Examination of yeast, fly, and mammalian sequence conservation of Vps4 ␣7 and ␣9 revealed that Leu-407 is the most conserved of the three residues across the Vps4 family (Fig. 1C). To test the model that the Vta1 VSE stimulates Vps4 via these residues, mutant forms of Vps4 were generated and analyzed for basal ATPase activity (Fig. 2), activity upon stimulation with Vta1 and ESCRT-III-activated Vta1 (Fig. 3), and direct regulation by ESCRT-III (Fig. 5).
Mutation of ␣7 and ␣9 Does Not Eliminate Vps4 ATPase Activity-WT and mutant forms of Vps4 were expressed in bacteria, purified ( Fig. 2A), and characterized for inherent ATPase activity (i.e. ATPase activity in the absence of other factors). Vps4 exhibits concentration-dependent increases in specific activity (ADP generated/Vps4/min) in the low micromolar range (34,40). This behavior is attributed to the role of oligomerization in forming the active enzyme. All forms of Vps4 displayed similar concentration-dependent ATPase activities with half-maximal activities (apparent K m , or K m, app ) between 299 nM for WT and 394 nM for Vps4(M330K) (Fig. 2B). The apparent maximal specific activities (V max, app ) of WT, Vps4(M330K), and Vps4(L407K) were similar (107, 112, and 107 ADP/Vps4/min, respectively), whereas Vps4(K411A) exhibited a reduced V max, app (76 ADP/Vps4/min). This analysis indicated that these mutant Vps4 proteins assemble into active oligomers in a manner comparable with WT Vps4, but a mutation within this portion of the small ATPase domain (K411A) subtly alters inherent ATP hydrolysis by the Vps4 oligomer.

Vta1 VSE Stimulates Vps4 via the Small AAA Domain
This profile of stimulation is consistent with previous analyses indicating that Vta1 stimulates Vps4 more robustly than the VSL domain and that Vta1(275-330) exhibits additional VSE stimulation reflective of ESCRT-III (40,53).
The two other Vps4 mutants display more complex responses (Fig. 3) Met-330 is present on ␣7, the helix more distal to the ␤-domain. These results suggested that: 1) basal VSE stimulation is mediated by contacting ␣9, 2) ESCRT-III-activated VSE stimulation is mediated by additionally contacting ␣7, and 3) L407K disrupts stimulation via ␣7 and ␣9 to prevent both modes of VSE-mediated stimulation, suggesting that this represents a relevant contact site for both modes.
These ATPase assays suggested that Vps4 ␣7 and ␣9 contribute to Vta1 interaction with Vps4. To examine this model, in vitro association was examined by GST pulldown experiments. GST-Vta1 was able to isolate WT and Vps4 mutants more effectively than GST alone under conditions analogous to the ATPase reaction (Fig. 4). However, variation of isolation was observed among the Vps4 mutants: Vps4(K411A) exhibited isolation comparable with that of Vps4(WT), but Vps4(M330K) and Vps4(L407K) were less effectively isolated. This result indicated that the VSE contributes to Vta1 interaction with Vps4 in addition to VSL-␤-domain interaction.
The ESCRT-III subunit Ist1 was used for this analysis because recombinant Ist1 is more soluble than other ESCRT-III subunits. Although the Ist1 carboxyl terminus stimulates Vps4 ATPase activity, full-length Ist1 inhibits Vps4 ATPase activity (57). However, an Ist1 mutant defective for Did2 MIM1 binding (L168A,Y172A) (26) stimulates Vps4 in a manner dependent on both the Ist1 MIM1 and the Vps4 MIT domain (Fig. 5B and data not shown). Vps4 activities in the presence of WT Ist1 or Ist1(L168A,Y172A) were assessed to examine direct ESCRT-III regulation of Vps4. WT Vps4 alone (500 nM, rather than the 250 nM concentration used in Fig. 3) exhibited activity of 107 ADP/ Vps4/min, and the addition of 10 M WT Ist1 reduced ATPase activity to 6 ADP/Vps4/min (Fig. 5A). Similarly, the activities of

Vta1 VSE Stimulates Vps4 via the Small AAA Domain
the Vps4 small AAA domain mutants were inhibited by Ist1 (101, 93, and 69 ADP/Vps4/min for M330K, L407K, and K411A alone were reduced to 4, 6, and 0.7 ADP/Vps4/min with Ist1 addition). These observations indicated that these mutations within the small AAA domain do not disrupt Ist1 inhibition of Vps4 ATPase activity.
VSE Stimulation via the Small AAA Domain Promotes Vps4 Function in Vivo-We next sought to address the significance of this VSE contact with ␣7 and ␣9 in vivo. Deletion of Vta1 only partially disrupts MVB sorting (40), and ESCRT-III activation of Vta1 only partially contributes to this defect as suggested by two observations: 1) deletion of the Vta1 MIT domains disrupts Vta1-ESCRT-III associations, but expression of this mutant in vta1⌬ yeast exhibited only weak defects in MVB sorting (31); and 2) mutation of the Vta1 VSE (L284E, I287E, and M288E; VSE⌬) compromised MVB sorting and ESCRT-III recycling in vivo, although the extents of the defects were less than vta1⌬ effects (53). These observations predicted that mutation of the Vps4 residues mediating Vta1 VSE function would also only weakly disrupt MVB sorting and ESCRT-III recycling. GST or GST-Vta1 pulldown experiments were performed with purified WT and Vps4 mutants. Eluted material was analyzed by Coomassie staining and Western blotting for Vps4. WT and Vps4 mutants were isolated with GST-Vta1 more effectively than with GST alone. Vps4(WT) and Vps4(K411A) exhibited similar isolation with GST-Vta1, but isolation of Vps4(M330K) and Vps4(L407K) was reduced. The data presented are representative of experiments performed more than three times.  OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41

JOURNAL OF BIOLOGICAL CHEMISTRY 28713
To test this prediction, M330K, L407K, and K411A forms of Vps4 were expressed in vps4⌬ cells (Fig. 6A), and the maturation kinetics of the MVB cargo CPS were evaluated as an indicator of MVB sorting pathway function (Fig. 6, B and C). CPS is synthesized as an inactive precursor (pCPS). Sorting of pCPS into the MVB pathway and subsequent delivery to the vacuolar lumen permits proteolytic activation of CPS (mCPS). The kinetic analysis of the conversion of pCPS to mCPS serves as an indicator of MVB pathway function (31,53). Loss of Vps4 (vps4⌬) disrupted MVB sorting, resulting in the persistence of pCPS (Fig. 6, B and C, and Ref. 17). Reintroduction of WT Vps4 restored CPS maturation with only 42% of pCPS remaining at 30 min (black). Reintroduction of Vps4(K411A) yielded CPS maturation kinetics (green) indistinguishable from WT Vps4. Reintroduction of Vps4(M330K) resulted in a subtle delay in CPS maturation kinetics (blue), but this effect was not significantly different from the CPS maturation kinetics with WT Vps4. By contrast, the CPS maturation kinetics with reintroduction of Vps4(L407K) were delayed from the CPS maturation kinetics with WT Vps4 (two-way analysis of variance, p value Ͻ 0.01; red). This partial defect for Vps4(L407K) is similar to partial defects in CPS maturation observed with Vta1(MIT⌬) and Vta1(VSE⌬) (31, 53). Importantly, Vps4(M330K), Vps4(L407K), and Vps4(K411A) are all expressed similarly to WT Vps4 (Fig. 6A). These results support the conclusion that Leu-407 contributes to Vps4 function in the MVB sorting pathway by permitting Vta1 stimulation via the VSE.
Vps4 disassembles ESCRT-III, and defects in Vps4 function lead to enhanced ESCRT-III membrane association (17,34). To more directly examine Vps4(L407K) function in ESCRT-III disassembly, the membrane association of the ESCRT-III subunit Snf7 was examined in vps4⌬ yeast expressing Vps4(L407K) (Fig. 6D). Loss of Vps4 (vps4⌬) resulted in 95% of Snf7 in the 13,000 ϫ g pellet, indicating increased membrane association of ESCRT-III. Reintroduction of WT Vps4 reduced Snf7 association with the 13,000 ϫ g pellet to 48%. By contrast, reintroduction of Vps4(L407K) reduced Snf7 association with the 13,000 ϫ g pellet to 56%. These observations indicate a small but significant reduction in Snf7 recycling with reintroduction of Vps4(L407K) (p value Ͻ 0.01), consistent with both predictions of the model (only a partial defect) and the Vps4(L407K) partial defect in CPS maturation. This observation supports the conclusion that L407K contributes to Vps4 recycling of ESCRT-III to enable efficient MVB sorting.

Vta1 VSE Stimulates Vps4 via the Small AAA Domain
Compensatory Mutation of the VSE Restores Stimulation of Vps4(L407K)-Although these biochemical and in vivo studies support the model that the Vta1 VSE contacts Vps4 ␣7 and ␣9 to stimulate ATP hydrolysis, one possibility is that these effects are a secondary consequence of altered small AAA domain conformation(s) rather than directly implicating these residues in mediating Vta1 VSE-Vps4 interactions. To address this possi-bility, we examined whether disrupted activity could be restored by compensatory mutations in Vta1. Characterization of the Vta1 VSE involved mutating hydrophobic residues to acidic residues (L284E, I287E, and M288E) (53), whereas Vps4 Leu-407 was mutated to a basic residue. The ability of the Vta1 L284E, I287E, and M288E triple mutant (Vta1(VSE⌬)) to compensate for the Vps4 L407K mutation was examined (Fig. 7). Vta1(VSE⌬) stimulated WT Vps4 less effectively than WT Vta1 (122 versus 177 ADP/Vps4/min; p value Ͻ 0.01), consistent with previous analysis (53). By contrast, Vta1(VSE⌬) stimulated Vps4(L407K) more effectively than WT Vta1 or VSL (149 versus 80 and 78 ADP/Vps4/min; p values Ͻ 0.05). These observations indicated that mutation of the Vta1 VSE residues to acidic residues (L284E, I287E, and M288E) compensates for mutation of Vps4 Leu-407 to a basic residue (L407K) to restore Vta1 VSE stimulation of Vps4 via the small AAA domain. This result supports the model that the Vta1 VSE contacts the proximal (␣9: Leu-407 and Lys-411) and distal (␣7: Met-330) helices of the small AAA domain to enhance Vps4 ATP hydrolysis.

DISCUSSION
Vta1 promotes Vps4 function to recycle ESCRT-III and permit efficient MVB sorting (36 -41). Vta1 stimulation of Vps4 occurs through both the Vta1 VSL domain and the neighboring VSE (31,53). The VSL domain binds to the Vps4 ␤-domain to promote oligomerization and stimulate ATP hydrolysis (39,40,41,43,55). An additional level of Vps4 stimulation is mediated by the VSE, and the VSE is also required for the enhanced stimulation that occurs when ESCRT-III subunits bind to the amino-terminal MIT domains of Vta1 (53). Our results indicate that the Vta1 VSE contacts surface residues in ␣7 and ␣9 of the small AAA domain to stimulate Vps4, with Leu-407 playing a key role in this stimulation (Fig. 8). This analysis identifies a  In the absence of ESCRT-III binding, the VSE contributes to stimulation of Vps4 ATPase activity by contacting the Vps4 ␣9 residues Leu-407 and Lys-411. However, Vta1 linker region autoinhibition of the VSE curtails the extent of stimulation in the absence of ESCRT-III such that contact with Vps4 ␣7 (Met-330) is not required. B, ESCRT-III binding Vta1 relieves autoinhibition of the VSE to facilitate contact with Vps4 ␣7 residue Met-330 to further enhance Vps4 ATPase activity. Contact with ␣9 also contributes to this ESCRT-III-enhanced stimulation but may not be required. Leu-407 of ␣9 is key for activation via both ␣7 and ␣9 and is conserved from yeast to humans. OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 novel mechanism by which the AAA domain is regulated to promote Vps4 function.

Vta1 VSE Stimulates Vps4 via the Small AAA Domain
Biochemical studies identified the Vta1 VSL domain as necessary to stimulate Vps4 (40). The VSL domain mediates dimerization of Vta1, and the structure of the Vta1 carboxyl terminus (residues 281-330) identified that residues implicated in contacting the Vps4 ␤-domain decorated two surfaces of the VSL dimer (40,43). These observations suggested that the VSL domain might stimulate Vps4 by contacting two ␤-domains from non-neighboring Vps4 subunits within the oligomer to promote oligomer formation. This model linked the importance of VSL dimerization in Vta1 function with the role of the Vta1 VSL domain in promoting Vps4 oligomerization (via the ␤-domain). However, a structure of the Vta1 VSL domain in complex with the small AAA domain of Vps4 suggested that the orientation of the VSL dimer to the ␤-domain would not permit additional contacts to other Vps4 subunits within the same oligomer (55). Instead, the VSL dimer was proposed to bridge ␤-domains across Vps4 oligomers to promote Vps4 function. Another alternative is that VSL dimer contact with a single ␤-domain facilitates Vps4 oligomerization through subtly altering conformation of the Vps4 monomer (or dimer) rather than inter-or intraoligomeric bridging of Vps4 subunits. Regardless of which mechanism is correct, these models failed to explain how Vta1 could increase the ATPase activity of the Vps4 oligomer above its apparent maximal specific activity.
Further examination of Vta1 stimulation identified an additional element (VSE) that promoted the hydrolysis above the Vps4 inherent maximal rate (maximal specific activity observed with Vps4 alone) (53). The VSE is present in the structure of the Vta1 carboxyl terminus (43) but is not included in the complex of VSL with the Vps4 ␤and small AAA domains (55). These structures were combined with the structure of Vps4 AAA domain (56) to generate a model suggesting how the VSE stimulates Vps4 (Fig. 1B). The Vta1 VSE is positioned across from two helices (␣7, ␣9) of the small AAA domain, with ␣9 proximal to the ␤-domain. Mutating residues of this proximal helix (L407K and K411A) perturbed stimulation by full-length Vta1 with only stimulation attributed to the VSL domain apparent. These observations suggested that basal Vta1 stimulation occurs through the VSE contacting ␣9 to stimulate Vps4 ATP hydrolysis beyond the inherent maximal rate (Fig. 8A). ESCRT-III subunits bind to the Vta1 amino-terminal MIT domains to potentiate stimulation via the VSE, and a truncation of Vta1 (residues 275-330) mimics this ESCRT-III activated form (53). Mutation of Vps4 Met-330 disrupted this ESCRT-III-activated VSE stimulation, although basal VSE stimulation was unaffected. These observations suggested that ESCRT-III activation promotes VSE contact with ␣7 to enhance Vps4 ATP hydrolysis (Fig. 8B). The result that Vps4(K411A) could be stimulated by Vta1(275-330) above the activities observed with full-length Vta1 or the Vta1 VSL domain suggests that VSE stimulation via ␣7 is not dependent upon stimulation via ␣9. However, Vps4(L407K) is resistant to VSE stimulation by both full-length Vta1 and Vta1(275-330) (Fig. 3). Although Vps4(L407K) was not stimulated by WT Vta1, mutation of the Vta1 VSE residues from hydrophobic to acidic residues (i.e. Vta1(VSE⌬)) compen-sated for the mutation of Vps4 Leu-407 to a basic residue. This restoration of stimulation supports the model in which the VSE residues are located in proximity to ␣7 and ␣9 to stimulate Vps4 ATPase activity, as illustrated in Fig. 1B. The ATP binding pocket is located in the cleft between the small and large AAA domains, suggesting that perturbations of ␣7 and ␣9 are transmitted to the pocket to alter the kinetics of nucleotide binding or the kinetics of nucleophilic attack of the ATP ␥-phosphate. This mechanism of contacting ␣7 and ␣9 of the small AAA domain to activate Vps4 represents a novel means to enhance AAA-ATPase activity. Further structural studies will be required to resolve the specifics of how contacting ␣7 and ␣9 accelerate nucleotide hydrolysis and/or exchange.
The ability of Vta1(VSE⌬) to stimulate Vps4(L407K) highlights the importance of Leu-407 in mediating VSE activation of Vps4. This idea is consistent with the conservation of Leu-407 from Saccharomyces cerevisiae to Drosophila melanogaster, Mus musculus, and Homo sapiens (Vps4A and Vps4B), whereas Met-330 and Lys-411 are not similarly conserved. The importance of Leu-407 is also highlighted by characterization of yeast expressing Vps4(L407K). Expression of Vps4(L407K) in vps4⌬ yeast resulted in partial defects in MVB sorting and ESCRT-III disassembly. These phenotypes are similar to the partial defects observed upon mutation of the Vta1 VSE or deletion of the Vta1 MIT domains (31,53). This similarity is consistent with Leu-407 mediating Vta1 VSE stimulation to contribute to Vps4 function in vivo. These phenotypes are also consistent with other aspects of Vps4 function (i.e. inherent ATPase activity, direct ESCRT-III stimulation, Ist1 inhibition) occurring independent of this ␣7-␣9 surface of the small AAA domain. Although ESCRT-III stimulation via Vta1 is mediated by the VSE contacting ␣7 and ␣9, direct ESCRT-III stimulation of Vps4 via the Vps4 MIT domain and the Vps4 linker region occur through a distinct mechanism.
In total, our observations support the model in which ␣7 and ␣9 of the small AAA domain are contacted by the Vta1 VSE to stimulate Vps4 ATP activity in addition to VSL association with the ␤-domain. In the absence of ESCRT-III binding by Vta1, autoinhibition within Vta1 limits the extent of VSE activity such that the VSE primarily contacts ␣9 (the helix proximal to the ␤-domain) to stimulate ATPase activity. ESCRT-III binding to the Vta1 MIT domains relieves autoinhibition within Vta1 such that the VSE makes additional contact with ␣7 (the more distal helix) to further enhance ATPase activity to drive ESCRT-III disassembly. This relief of autoinhibition could occur through subtly reorienting the flexible helix on which the VSE resides or by uncovering or reorganizing the amino-terminal portion of the VSE to permit association with ␣7 of the Vps4 small AAA domain. The model depicted in Fig. 1B likely represents the ESCRT-III-stimulated Vta1 VSE contacts with Vps4 (including interaction with Met-330), whereas the autoinhibited Vta1 VSE fails to contact ␣7 of Vps4. Formation of the Vta1 dimer is critical to this model because the VSL dimer creates a four-helix bundle with one chain contacting the ␤-domain, while the second Vta1 chain supplies the VSE to contact the small AAA domain to enhance ATP hydrolysis (Fig. 1). This model implicates VSL dimer formation in: 1) forming one surface critical for ␤-domain interaction and 2) permitting simul-

Vta1 VSE Stimulates Vps4 via the Small AAA Domain
taneous associations with the ␤-domain and the small AAA domain by distinct chains of the Vta1 dimer to fully stimulate Vps4.