Conformational Changes in the Endosomal Sorting Complex Required for the Transport III Subunit Ist1 Lead to Distinct Modes of ATPase Vps4 Regulation*

Intralumenal vesicle formation of the multivesicular body is a critical step in the delivery of endocytic cargoes to the lysosome for degradation. Endosomal sorting complex required for transport III (ESCRT-III) subunits polymerize on endosomal membranes to facilitate membrane budding away from the cytoplasm to generate these intralumenal vesicles. The ATPase Vps4 remodels and disassembles ESCRT-III, but the manner in which Vps4 activity is coordinated with ESCRT-III function remains unclear. Ist1 is structurally homologous to ESCRT-III subunits and has been reported to inhibit Vps4 function despite the presence of a microtubule-interacting and trafficking domain-interacting motif (MIM) capable of stimulating Vps4 in the context of other ESCRT-III subunits. Here we report that Ist1 inhibition of Vps4 ATPase activity involves two elements in Ist1: the MIM itself and a surface containing a conserved ELYC sequence. In contrast, the MIM interaction, in concert with a more open conformation of the Ist1 core, resulted in stimulation of Vps4. Addition of the ESCRT-III subunit binding partner of Ist1, Did2, also converted Ist1 from an inhibitor to a stimulator of Vps4 ATPase activity. Finally, distinct regulation of Vps4 by Ist1 corresponded with altered ESCRT-III disassembly in vitro. Together, these data support a model in which Ist1-Did2 interactions during ESCRT-III polymerization coordinate Vps4 activity with the timing of ESCRT-III disassembly.

Intralumenal vesicle formation of the multivesicular body is a critical step in the delivery of endocytic cargoes to the lysosome for degradation. Endosomal sorting complex required for transport III (ESCRT-III) subunits polymerize on endosomal membranes to facilitate membrane budding away from the cytoplasm to generate these intralumenal vesicles. The ATPase Vps4 remodels and disassembles ESCRT-III, but the manner in which Vps4 activity is coordinated with ESCRT-III function remains unclear. Ist1 is structurally homologous to ESCRT-III subunits and has been reported to inhibit Vps4 function despite the presence of a microtubule-interacting and trafficking domain-interacting motif (MIM) capable of stimulating Vps4 in the context of other ESCRT-III subunits. Here we report that Ist1 inhibition of Vps4 ATPase activity involves two elements in Ist1: the MIM itself and a surface containing a conserved ELYC sequence. In contrast, the MIM interaction, in concert with a more open conformation of the Ist1 core, resulted in stimulation of Vps4. Addition of the ESCRT-III subunit binding partner of Ist1, Did2, also converted Ist1 from an inhibitor to a stimulator of Vps4 ATPase activity. Finally, distinct regulation of Vps4 by Ist1 corresponded with altered ESCRT-III disassembly in vitro. Together, these data support a model in which Ist1-Did2 interactions during ESCRT-III polymerization coordinate Vps4 activity with the timing of ESCRT-III disassembly.
The ESCRT-III subunit Ist1 may play a special role in coordinating Vps4 and ESCRT-III functions by exerting both posi-tive and negative regulation of Vps4 activity. Similar to Did2 and Vps2, Ist1 has been implicated as a positive regulator of Vps4. Ist1 is essential for cytokinesis in mammalian cells (8,9), synthetic genetic defects in MVB sorting are observed in ist1⌬ vta1⌬ and ist1⌬ vps60⌬ yeast strains (23,26), and a C-terminal fragment of Ist1 containing its MIM element stimulates Vps4 ATPase activity in vitro (23). By contrast, overexpression of Ist1 disrupts MVB sorting in yeast, and full-length Ist1 inhibits Vps4 ATPase activity in vitro (23), indicating that Ist1 can also negatively regulate ESCRT function. These dual activities, inhibition and stimulation of Vps4, make Ist1 unique among the ESCRT-III subunits. However, the mechanism that mediates switching between Ist1 positive and negative regulation of Vps4 are unclear.
To examine the relationships between Ist1 conformation and Vps4 regulation, a structure-function study of Ist1 was conducted. Here we report that alterations in the conformation of the Ist1 core domain altered regulation of Vps4 function. Both negative and positive regulation of Vps4 by Ist1 required MIM-MIT interactions, whereas a highly conserved ELYC region located in the Ist1 core region was required for negative regulation. These structure-function studies suggested that Ist1 MIM-Vps4 MIT domain interactions represent the primary mode of interaction between Vps4 and Ist1, whereas secondary interactions dependent upon changes in ESCRT-III core conformation modulate Vps4 function. Conversion of Ist1 from an inhibitor to a stimulator of Vps4 ATPase activity in vitro has also been observed upon addition of Did2, the ESCRT-III subunit to which Ist1 binds specifically (20,23,26,45,60). We propose that Ist1 binding to Did2 during ESCRT-III polymerization induces conformational changes in Ist1 that alter regulation of Vps4 to coordinate Vps4 and ESCRT-III functions.

Experimental Procedures
Plasmids and Strains-Yeast IST1 was amplified from Saccharomyces cerevisiae genomic DNA with the 5Ј oligomeric primer designed to remove the intron of Ist1 and cloned into the BamHI and XhoI sites of pET28b (Novagen), generating pET28-Ist1. Mutagenesis of Ist1 was performed using the Gene Tailor site-directed mutagenesis system (Invitrogen) with a pBS-Ist1 template. The pET28a Ist1(L168A,Y172A) construct was supplied by Dr. Zhaohui Xu (University of Michigan) (60). All cloned PCR products and mutant plasmids were sequenced to exclude unexpected mutations. The Ist1 promoter was amplified from yeast genomic DNA and subcloned into the NotI and BamHI sites of pRS415 (61), yielding the pRS415 Promoter(Ist1). The Ist1 coding sequences for the WT and mutants were subcloned from pET28b bacterial expression vectors into pRS415 Promoter(Ist1) via the BamHI and SalI sites. Alternatively, the BamH1 and Xho1 sites were used for pET28a Ist1(L168A,Y172A). The yeast strains used in this study included SEY6210 (62) Ist1 Antibody Generation-Purified full-length Ist1 (pET28-Ist1) lacking the His 6 tag was used for antiserum production (Covance). A New Zealand rabbit was immunized with Ist1, and test bleeds were obtained. Bleeds were tested for detection of Ist1 in WT (SEY6210) and ist1⌬ yeast strains and recombinantly expressed and purified Ist1. This polyclonal antibody detected purified WT Ist1 and Ist1 MIM mutants equivalently (data not shown).
Protein Expression and Purification-Protein expression for GST, GST-Vps4, Did2-His 6 , and His 6 -Ist1 was performed in the BL21-DE3 bacterial strain at 16°C for 16 h with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside. His 6 fusion proteins were purified by Ni 2ϩ -affinity chromatography (5 ml HiTrap chelating FF column; GE Healthcare Life Sciences), treated with thrombin to remove the His 6 tag from Ist1, incubated with ATP to dissociate chaperones, and subjected to size exclusion chromatography (SEC) (Superdex200 16/60 HiLoad for Ist1 or Superdex75 16/60 HiLoad for Did2, GE Healthcare Life Sciences) in 25 mM HEPES, 200 mM KCl (pH 7.5). Purification of Ist1 and Did2 included treatment with ATP to minimize contaminating ATPase activities, and the final purified Ist1 and Did2 proteins were analyzed to confirm the absence of significant contaminating ATPase activity ( Fig. 2C and data not shown). Purified Ist1 was run on SDS-PAGE with Benchmark Ladder (GE Healthcare Life Sciences) to confirm sample purity (Figs. 2C, 4D, 5A, 7A, and 8B). GST-Vps4 fusion protein was purified as described previously (30).
ATPase Assay-Measurement of Vps4 ATPase activity was performed as described previously (20,30,31,53), but reactions were initiated by addition of 6 mM ATP (compared with 1 or 2 mM ATP) to remain within a linear range of ATP hydrolysis for Vps4 hyperstimulated by Ist1 mutants. ATPase activities from three independent experiments are shown as mean Ϯ S.D. The significance of difference in rates was assessed by t tests using Prism5 (GraphPad). An example of this analysis, including images of TLC plates and determination of rates, is presented in Ref. 53. 500 nM Vps4 was used in all ATPase assays because this concentration of Vps4 exhibits submaximal specific activity (41-50 ADP molecules/Vps4/min), making it amenable for observing stimulation or inhibition of Vps4 ATPase activity. Maximal inhibition or stimulation of 500 nM Vps4 by WT Ist1 or Ist1 mutants was achieved at [Ist1] Յ 8 M, as ascertained from titrations from 0.5-12 M Ist1 in the presence of Vps4 (Fig. 2, A and B, and data not shown). Vps4 activity with 8 M Ist1 mutants is therefore presented for comparison (Figs. 2C, 4D, 5B, 7C, and 8A).
Limited Proteolysis-Ist1 was diluted from Ͼ100 M stocks in Ist1 purification buffer (25 mM HEPES and 200 mM KCl (pH 7.5)) to ϳ13 M in ATPase buffer (20 mM HEPES, 100 mM KOAc, and 5 mM MgOAc (pH 7.5)), incubated at 30°C for 30 min prior to addition of trypsin (Sigma-Aldrich, catalog no. T8658) at a final ratio of 1:1000 (w/w) trypsin:Ist1. 10-l samples were taken at the indicated time points (e.g. 5 min, 1 h, 2 h, and 4 h) and quenched by adding 12 l of 5ϫ Laemmli sample buffer (64) and heating at 100°C for 10 min. The samples and protein ladder (Benchmark protein ladder or Benchmark prestained protein ladder, Invitrogen) were resolved by SDS-PAGE and stained with Coomassie Blue (Bio-Rad) (Figs. 3, A and B; 4E; 5E, 6D, and 7D). Alternatively, 10-l samples were quenched by adding 12 l of 2% trifluoroacetic acid, separated by HPLC (Agilent 1200 system, Agilent Zorbax SB C18 column), and analyzed by positive-mode electron spray ionization mass spectrometry (Agilent 6224 TOF system). These data were correlated to Ist1 amino acid sequences with an accuracy of 10 ppm using Agilent Mass Hunter Qualitative Analysis/ BioConfirm software (version B.05.00).
SEC-Analytical SEC was performed with Superdex200 GL 10/300 (GE Healthcare Life Sciences) in 25 mM HEPES, 200 mM KCl (pH 7.5). This buffer was used because SEC performed with ATPase buffer resulted in reduced recovery of Ist1 and an extended lagging shoulder (data not shown). Ist1 stocks were normalized to 200 M in 25 mM HEPES, 200 mM KCl (pH 7.5), and 172 g of sample was resolved (0.75 ml/min flow rate at 4°C). The UV traces shown are representative of at least two runs. Fractions from SEC runs were collected, subjected to SDS-PAGE analysis, and visualized by Coomassie staining (data not shown). Indicated apparent molecular weights were deduced from gel filtration standards (Bio-Rad) (dashed black lines in Figs. 2, D and E; 4, A-C; 5, C and D; 7A; and 8B).
Protein-Protein Interaction-GST or GST-Vps4 was prebound to glutathione-Sepharose 4B (GE Healthcare Life Sciences) in PBS ϩ 0.05% Tween 20, incubated at 4°C for 1 h, washed with PBS ϩ 0.05% Tween 20, and equilibrated with ATPase buffer ϩ 0.05% Tween 20. Purified Ist1 proteins (2 g) were added to GST or GST-Vps4 in the presence of 0.1 mg/ml BSA. Following extensive washing with ATPase buffer ϩ 0.05% Tween 20, bound material was eluted with 20 mM Tris, 100 mM NaCl, 20 mM glutathione, and 1 mM DTT (pH 7.5). Samples were resolved by SDS-PAGE. Western blotting was performed using anti-Ist1 (this study) and Cy3-conjugated Goat anti-Rabbit IgG secondary (Life Technologies) and detected using the Typhoon FLA 7000 system. Subcellular Fractionation-Subcellular fractionation of cells grown in minimal medium was performed as described previously (23), except that cell extracts were subjected to centrifugation at 13,000 ϫ g for 10 min at 4°C to separate the soluble and pellet fractions. Samples (0.2 A 600 equivalents) were resolved by SDS-PAGE and Western-blotted for Snf7 (31) and Ist1 (this study). Phosphoglycerate kinase (Life Technologies) and Pep12 (Life Technologies) were used a markers for the soluble and pellet fractions, respectively. Western blots were developed using HRP-conjugated goat anti-rabbit IgG or goat anti-mouse (Life Technologies), along with SuperSignal West Femto and SuperSignal West Pico substrates (Thermo), and the Autochemi system (UVP, Upland, CA). Quantitation was performed with the ImageQuant software package. Subcellular fractionation data represent three independent experiments (with representative Western blots shown) and are graphed as mean Ϯ S.D.
ESCRT-III Disassembly Assay-ESCRT-III disassembly assays were performed as described previously (31). Yeast (pep4⌬ vps4⌬ or pep4⌬ ist1⌬ vta1⌬) was lysed with a Dounce in 20 mM Pipes (pH 6.8), 100 mM KCl, 50 mM KOAc, 5 mM MgOAc, and 100 mM sorbitol with protease inhibitors. The lysate was cleared by a 5-min, 300 ϫ g spin, and the cleared lysate was subjected to a 10-min spin at 10,000 ϫ g. The pelleted membranes were washed twice in lysis buffer and resuspended in disassembly assay buffer (20 mM HEPES (pH 7.4), 100 mM KOAc, 5 mM MgOAc, and 200 mM sorbitol with protease inhibitors) at 50 A 600 equivalents/ml. Reactions with various amounts of purified Vps4 and Ist1 were performed with 0.5 A 600 equivalent membranes and an ATP regeneration system (1 mM ATP) at 30°C for 10 min. Membranes were repelleted via a 10-min spin at 13,000 ϫ g, and Western blotting was performed to assess levels of Snf7 in the pellet and soluble fractions. The UVP Autochemi system and ImageQuant software package were used to quantify Snf7 levels. Disassembly assay data represent two or three independent experiments with reactions performed in duplicate or triplicate within experiments and are graphed as the mean Ϯ S.E. (representative Western blots are shown).
To assess Ist1 conformation, SEC and limited proteolysis were performed. SEC analysis of WT Ist1 (35 kDa) revealed an apparent molecular mass of 94 kDa (Fig. 2, D and E), consistent with previous analyses indicating that Ist1 behaves as an elongated monomer (23). In addition, Ist1(L168A,Y172A) and Ist1(K135A) eluted with apparent masses of 104 and 94 kDa, respectively (Fig. 2, D and E). This similarity to WT Ist1 suggests that these mutants also behave as elongated monomers, with Ist1(L168A,Y172A) adopting a subtly more open conformation.
Transient conformational changes in Ist1 were assessed by limited proteolysis with trypsin (Fig. 3). Trypsin treatment of Ist1 resulted in the initial generation of a 30-kDa fragment (indicated as A in Fig. 3). However, the stability of this fragment varied between the WT and mutant Ist1 forms. Although the A fragment from WT Ist1 persisted for more than 8 h, the A fragments from Ist1(L168A,Y172A) and Ist1(K135A) were cleaved more rapidly into smaller species (Fig. 3, A and B, fragments B-D), suggesting more open conformations. Next, mass spectrometry analyses were conducted to identify the location of trypsin cleavage in wild-type versus mutant Ist1 forms (Fig. 3,  C and D). These analyses revealed that the A fragment of WT Ist1, Ist1(L168A,Y172A), and Ist1(K135A) resulted from cleavage at Arg 241 located within the linker region between ␣5a and ␣6 (Fig. 3C), suggesting that this region was equally susceptible to trypsin. Increased proteolysis of the Ist1(L168A,Y172A) and Ist1(K135A) core occurred through cleavage sites both common with WT Ist1 (Arg 39 and Lys 158 , blue residues in Fig. 3, C

Ist1 Inhibition and Stimulation of Vps4
and D) and sites unique to these mutants (Lys 17 , Arg 22 , Lys 52 , Arg 83 , Lys 98 , Lys 130 , and Lys 178 ; red residues in Fig. 3, C and D). These data suggest that Ist1(L168A,Y172A) and Ist1(K135A) adopt more open conformations than WT Ist1 via unfolding of their core domains.

Ist1 Inhibition and Stimulation of Vps4
Next we sought to test whether a form of Ist1 with its ␣6 bound more strongly to the ␣1,2,5 groove would reduce Vps4 inhibition or stimulation (Fig. 1B, closed conformation). The Did2-MIM1 element binds with high affinity to the Ist1 ␣1,2,5 groove (60). Therefore, we replaced the Ist1 MIM element with the Did2 MIM1 element to yield the Ist1(Did2-MIM1) chimera (Fig. 5A) and tested its biochemical activities (Fig. 7). Compared with WT Ist1, Ist1(Did2-MIM1) eluted later in SEC analyses (Fig. 7A, apparent molecular mass of 69 kDa), whereas the A fragment was equally susceptible to proteolysis (Fig. 7B). These observations are consistent with Ist1(Did2-MIM1) adopting a more closed conformation and further support that cleavage at Arg 241 to generate the A fragment can occur in the closed conformation. Ist1(Did2-MIM1) failed to bind or inhibit Vps4 activity (Fig. 7C), consistent with inaccessibility of the Did2-MIM1 for interactions with the Vps4 MIT domain.
The Ist1 ELYC Region Is Essential for Inhibition of Vps4 Activity-Ist1 contains a highly conserved ELYC sequence located in ␣2 of the Ist1 core that has been suggested previously to play a role in Vps4 inhibition (Fig. 1A) (23). Several additional highly conserved, surface-exposed residues are located in this region, including alanine 82 ( 74 ELYCELLLA 82 ). To test the role of this ELYC region in Vps4 regulation, we generated Ist1(E74A) and Ist1(A82D). Ist1(E74A) and Ist1(A82D) were unable to inhibit Vps4 activity but, instead, stimulated Vps4 to greater levels than Ist1 MIM mutants (Fig. 8A). Combining the E74A and MIMa mutations led to Vps4 stimulation similar to

Ist1 Inhibition and Stimulation of Vps4
E74A alone, suggesting that additional Vps4 stimulation was MIM-MIT-independent. Ist1(E74A) eluted later in SEC analyses, suggesting a more closed conformation (Fig. 8B). However, the MIM element was still accessible, as indicated by robust binding of Ist1(E74A) to Vps4 (Fig. 6). In limited proteolysis experiments, sensitivity of the A fragment of Ist1(E74A) was not increased relative to WT Ist1, indicating that folding of the core domain was not altered (Fig. 8C). Taken together, these data suggested that the Ist1 ELYC region is involved in a secondary, weaker affinity interaction in conjunction with   Did2 Converts Ist1 from an Inhibitor to a Stimulator of Vps4 ATPase Activity-Binding of Did2 to Ist1 has been suggested to alter the conformation of the Ist1 core as Ist1 incorporates into ESCRT-III polymers (45,60). To test whether Did2 binding altered Ist1 regulation of Vps4, the effect of Ist1 on Vps4 ATPase activity in the presence of Did2 was assessed (Fig. 9). Addition of 2 or 4 M Did2 alone stimulated Vps4, consistent with previous observations (20,39). Although WT Ist1 alone inhibited Vps4 ATPase activity, mixing WT Ist1 and Did2 led to Vps4 stimulation that was greater than that observed by addition of Did2 alone. This suggests that Did2 binding to Ist1 induces a more open Ist1 conformation, thereby switching Ist1 from an inhibitor to a stimulator of Vps4 activity.
Ist1 Conformation Affects Vps4 Function-Vps4 disassembly of ESCRT-III is critical for ESCRT function (17, 24, 28 -31). Therefore, the effects of altered Ist1 conformation on Vps4 function were examined in two contexts: subcellular fractionation of ESCRT-III subunits in vivo and ESCRT-III disassembly in vitro.
Positive regulation of ESCRT-III disassembly by Ist1 in the presence of 100 nM Vps4 was examined first (Fig. 11C). Relative to WT Ist1, ESCRT-III disassembly was enhanced further by the hyperstimulatory Ist1(L168A, Y172A) mutant (p Ͻ 0.05), revealing that the increased ATPase activity observed correlates with increased Vps4 function. Two mutants (Ist1(K135A) and Ist1(E74A)) that were both defective for inhibition of Vps4 ATPase activity and retained functional MIM elements exhib-ited a similar enhancement of Snf7 disassembly compared with WT Ist1. In contrast, mutants that were defective for binding to the Vps4 MIT domain (Ist1(MIMa), Ist1(MIMb), and Ist1(Did2MIM)) and/or possessed a more folded Ist1 core domain (Ist1(K52D)) exhibited partial inhibition of disassembly activity. These results further highlight the importance of the MIM element in Ist1 as well as the conformational state of the Ist1 core region to permit positive regulation of ESCRT function.
Negative regulation of ESCRT-III disassembly was subsequently examined with 500 nM Vps4 and excess Ist1 (2 M) (Fig.  11D). Addition of WT Ist1 and Ist1(K52D) inhibited Snf7 disassembly, consistent with their abilities to inhibit Vps4 ATPase activity (Fig. 4D). By contrast, all of the other mutants examined (which either adopted a more open core conformation or contained mutations in the Ist1 ELYC region or MIM element) exhibited partial or complete deficits in inhibiting Vps4-mediated ESCRT-III disassembly. Therefore, Ist1 mutants that were defective for inhibition of Vps4 ATPase activity were similarly unable to inhibit Vps4-mediated ESCRT-III disassembly.

Discussion
In recent years, considerable progress has been made toward elucidating distinct modes of Vps4 regulation by ESCRT-III subunits (20,23,24,38,39,65). The studies presented here provide additional insights into the bimodal regulation of Vps4 by Ist1 and its effects on ESCRT function via conformational changes in Ist1.
Negative regulation of Vps4 by Ist1 results from the combination of Ist1 MIM and ELYC interactions with Vps4. How does the Ist1 MIM element contribute to inhibition of Vps4 activity whereas the MIM1 elements of Did2 and Vps2 stimulate Vps4 (20)? The Ist1 MIM element itself is functionally dis-tinct from the Did2 MIM1 because the Did2 MIM1 could not replace the Ist1 MIM element in Vps4 binding and/or regulation (Figs. 6 and 7). Although the effect on Ist1 conformation may have contributed to these effects, the Ist1 MIM has also been demonstrated to bind to MIT domains in unique ways. hIst1 ␣6 binds to the Vps4A MIT domain in a manner similar to Vps2 and Did2 MIM1 elements (9, 32), hIst1 residues upstream of ␣6 bind to a distinct surface of the Vps4A MIT domain in a MIM2 mode (9,33), and hIst1 ␣6 can bind to the MIT domain of Spartin via a different orientation, referred to as the MIM3 mode (32). These MIM2 and MIM3 modes of association may enhance the strength of Ist1-Vps4 interactions or alter presentation of the Ist1 core in a manner that permits ELYC interaction to inhibit Vps4. The ELYC motif is not found in other ESCRT-III subunits, potentially explaining why Vps4 inhibi- tion is unique to Ist1. Interactions between Vps4 and the ELYC region are weaker compared with the MIM-MIT interaction (Fig. 6), suggesting that the MIM acts as the primary binding interaction that subsequently permits additional interactions to mediate distinct Vps4 regulation (Fig. 12A).
Vps4 stimulation by Ist1 is also MIM-MIT-dependent but additionally requires a more open Ist1 core conformation (Fig.  12A). Unfolding of the Ist1 core domain in the contexts of Ist1(L168A,Y172A) and Ist1(K135A) resulted in Vps4 hyperstimulation (Figs. 2 and 3), and this effect was blocked by introducing the K52D mutation, which stabilized the Ist1 core in a more closed conformation (Fig. 4). Ist1(K52D) was examined on the basis of its similarity to hIST1(R51D), which has been demonstrated previously to be defective for homopolymerization in vitro (45). Altered Ist1 core conformation may affect Vps4 regulation in different ways. First, conformational changes may alter the presentation of secondary Vps4 regulatory elements in Ist1. For example, the Vps4 inhibitory ELYC region, which is located in ␣2 of the Ist1 core domain, may be unfolded in an open state, leading to loss of Vps4 inhibition. Alternatively, the ELYC region itself may be involved in interactions with Did2 (for instance during ESCRT-III polymerization). This notion is supported by the hIST1 ELYC region being implicated in crystal packing interactions correlated with hIST1 homopolymerization in vitro (45), which may mimic aspects of ESCRT-III heteropolymerization in vivo. Finally, the open conformation may reveal a secondary surface that acts in concert with the Ist1 MIM element to synergistically stimulate Vps4 activity. An obvious candidate for a secondary stimulatory surface is a cluster of acidic residues located in the linker region that connects ␣5-6 or ␣5, which interact with the Vps4 linker and/or pore regions to stimulate ATPase activity in other ESCRT-III subunits (38 -40). We speculate that the Ist1 conformational changes associated with hyperstimulation of Vps4 resemble changes that occur with Ist1 binding to Did2 (Fig.  12A). This model is supported by the result that addition of Did2 converted Ist1 from an inhibitor to a stimulator of Vps4 ATPase activity.
These studies also highlight a physiological link between Ist1 regulation of Vps4 ATPase activity and ESCRT-III disassembly through interactions with Did2. Addition of low concentrations of Ist1 stimulated ESCRT-III disassembly from membranes in vitro, whereas high concentrations of Ist1 resulted in potent inhibition of ESCRT-III disassembly (Fig. 11). This pattern is consistent with the model in which Ist1 association with Did2 within ESCRT-III polymers reveals positive regulation of ESCRT function, whereas saturation of this association and subsequent accumulation of soluble Ist1 leads to negative regulation of Vps4 function (Fig. 12B). This behavior may explain how reduced Ist1 protein levels in vivo, which occurs during starvation (66), promotes MVB sorting by lowering the ratio of Ist1:Did2 to minimize the soluble, inhibitory pool of Ist1.
Bimodal regulation of ESCRT function by Ist1 may be important for other cellular processes, including cytokinesis in mammalian cells (8,9). Intriguingly, hIst1 and CHMP1/hDid2 bind specifically to another MIT domain-containing AAA-ATPase, Spastin. Spastin alters microtubule-severing activities during cytokinesis and has also been implicated in mitotic spindle disassembly, nuclear envelope sealing, and/or neuron function (Refs. 67-71 and reviewed in Ref. 72). It is tempting to speculate that hIst1 may regulate Spastin ATPase activity in a manner analogous to Ist1 regulation of Vps4 in these studies, thereby positioning hIst1 as a coordinator of membrane trafficking and cytoskeletal dynamics. In both contexts, we predict that ATPase regulation by Ist1 is altered in response to binding to Did2 or ESCRT-III polymerization, which, in turn, would be dependent on physiological cues such as receptor signaling,