Structural Basis of Molecular Recognition between ESCRT-III-like Protein Vps60 and AAA-ATPase Regulator Vta1 in the Multivesicular Body Pathway*♦

Background: Vps4 ATPase is stimulated by the interaction between Vta1 and Vps60, but the structural basis for this interaction remains unclear. Results: The structure of the Vta1 N-terminal domain (Vta1NTD) in complex with Vps60(128–186) was determined. Conclusion: Vps60(128–186) interacts with Vta1NTD through helices α4′ and α5′, extending over Vta1NTD MIT2 domain helices 1–3. Significance: This is a novel MIT recognition mode. The AAA-ATPase Vps4 is critical for function of the multivesicular body sorting pathway, which impacts cellular phenomena ranging from receptor down-regulation to viral budding to cytokinesis. Vps4 activity is stimulated by the interaction between Vta1 and Vps60, but the structural basis for this interaction is unclear. The fragment Vps60(128–186) was reported to display the full activity of Vps60. Vta1 interacts with Vps60 using its N-terminal domain (Vta1NTD). In this work, the structure of Vps60(128–186) in complex with Vta1NTD was determined using NMR techniques, demonstrating a novel recognition mode of the microtubule-interacting and transport (MIT) domain in which Vps60(128–186) interacts with Vta1NTD through helices α4′ and α5′, extending over Vta1NTD MIT2 domain helices 1–3. The Vps60 binding does not result in Vta1 conformational changes, further revealing the fact that Vps4 ATPase is enhanced by the interaction between Vta1 and Vps60 in an unanticipated manner.

Membrane budding away from the cytosol controls a number of biological processes important to cellular homeostasis and defenses against aging (1)(2)(3). The machinery responsible for executing this function consists of several distinct multimeric complexes known as the endosomal sorting complexes required for transport (ESCRTs) 3 (4 -6), which were originally identified in yeast and have been implicated in multivesicular body (MVB) biogenesis in plants, fungi, and animals (6,7). MVBs are formed when the late endosomal membrane invaginates and forms vesicles in the lumen, carrying selected transmembrane protein cargoes in the budding process (2,3). MVB biogenesis and fusion of an MVB with the lysosome in a later step represent a mechanism in which eukaryotic cells downregulate cell surface signaling via the endolysosomal degradation pathway (8). Components of the ESCRT machinery have been identified as potential tumor suppressors (9), mainly attributed to the involvement of the ESCRT machinery in mediating signal attenuation for activated receptors of growth factors, peptide hormones, and cytokines. The ESCRT machinery protects against age-related neurodegenerative diseases through either the canonical MVB pathway or autophagy (9,10). In addition, the ESCRT machinery also plays a pathological role in viral infection (2,11,12).
At least five distinctive multimeric complexes are involved in MVB biogenesis: ESCRT-0, ESCRT-I, ESCRT-II, ESCRT-III, and Vps4 (13,14). Their structure and function are highly conserved in all eukaryotes (9,10). ESCRT-0 is responsible for clustering of ubiquitylated cargoes to the site of MVB formation. ESCRT-I and ESCRT-II together generate membrane curva-ture and budding, whereas assembly of ESCRT-III at the bud neck catalyzes scission of the membrane. Completion of the process requires the AAA-ATPase Vps4, which disassembles ESCRT-III polymers upon ATP binding and hydrolysis (15,16). This ATP-consuming reaction is the only step in MVB biogenesis that inputs energy into the system, therefore providing the thermodynamic driving force for processing. Importantly, the role of Vps4 is conserved in all biological processes that depend on the action of the ESCRTs. Similar to other AAA-ATPases, Vps4 functions as an oligomer whose structure likely contains two conformationally distinctive hexameric rings (17). The rings contain a central pore where ESCRT-III subunits may physically interact and pass through during the disassembly process. Initial binding of Vps4 to ESCRT-III subunits requires its N-terminal microtubule-interacting and transport (MIT) domain (18). The MIT domain appears to specifically recognize short peptide sequence MIT-interacting motifs (MIMs) at or near the C-terminal end of ESCRT-III subunits (19 -24).
The in vivo activity of Vps4 is tightly regulated (25). To date, at least four proteins have been identified to bind to Vps4 and have roles in regulating its oligomerization and activity (26 -29). Did2, Ist1, and Vps60 are ESCRT-III-related proteins whose mechanisms of action on Vps4 remain to be clarified. Vta1 is a positive regulator of Vps4 by promoting Vps4 oligomerization (26,30). Structural study of Vta1 has shown that it is a molecular dimer, with each subunit folded into two terminal domains linked by a flexible linker (29). Its C-terminal domain mediates dimerization and binds to a unique ␤-domain in the Vps4 AAA domain (31,32). Its N-terminal domain (residues 1-167; Vta1NTD) ( Fig. 1) contains two tandem MIT domains, which specifically recognize Vps60 and Did2 but not other ESCRT-III subunits (27,29). The fragment Vps60(128 -186) was reported to display the full activity of Vps60, which stimulates Vps4 ATPase in a Vta1-dependent manner (27).
In this work, to investigate how Vps60 interacts with Vta1NTD, we first measured the binding affinity of Vta1NTD for Vps60(128 -186) (K d ϳ 0.7 M) using isothermal titration calorimetry assay and then determined the solution structure of Vta1NTD in complex with Vps60(128 -186). To confirm the residues involved in the interaction between Vta1NTD and Vps60(128 -186), site-directed mutations and GST pulldown experiments were performed. The structure reveals that Vps60(128 -186) interacts with Vta1NTD through a novel MIT domain recognition mode distinct from any reported mechanism.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification-DNA fragments encoding yeast Vta1 and Vps60 were amplified from Saccharomyces cerevisiae genomic DNA. Vta1NTD or Vps60(128 -186) was expressed in Escherichia coli BL21(DE3) using a modified pET28b vector with a small ubiquitin-like modifier (SUMO) protein inserted between a His 6 tag and the Vta1NTD or Vps60(128 -186) coding region, respectively. His 6 -tagged SUMO-Vta1NTD or His 6 -tagged SUMO-Vps60(128 -186) was purified by nickel-nitrilotriacetic acid affinity chromatography following standard procedures. Ulp1 protease was then added to remove the His 6 -SUMO tag, and the protein mixture was passed over a second nickel-nitrilotriacetic acid column and further purified by anion exchange chromatography on a Resource Q column (GE Healthcare). For isotope labeling, M9 minimal medium was supplemented with 15 NH 4 Cl (Cambridge Isotope Laboratories) or 15 NH 4 Cl and 2 g/liter [ 13 C]glucose (Cambridge Isotope Laboratories). Derivative proteins were purified in the same way as native proteins.
NMR Structure Calculation-Calculations were carried out using a standard simulated annealing protocol implemented in the program XPLOR-NIH 2.19. Interproton distance restraints derived from NOE intensities were grouped into three distance ranges, 1.8 -2.9, 1.8 -3.5, and 1.8 -6.0 Å, corresponding to strong, medium, and weak NOEs, respectively. The dihedral angles and were derived from the backbone chemical shifts (HN, HA, CO, and CA) using the program TALOS (36,38). Slowly exchanging amide protons, identified in the two-dimensional 15 N HSQC spectra recorded after a H 2 O buffer was exchanged with a D 2 O buffer, were used in the structure calculated with the NOE distance restraints to generate hydrogen bonds for the final structure calculation, as done in the literature (39). A total of 10 iterations (50 structures in the initial eight iterations) were performed. 100 structures were computed in the last two iterations; 20 conformers with the lowest energy were used to represent the three-dimensional structures. In the ensemble of the simulated annealing 20 structures, there was no distance constraint violation of Ͼ0.3 Å and no torsion angle violation of Ͼ3°. The final 20 structures with the lowest energy were evaluated with the programs PROCHECK-NMR and PROCHECK (40) and are summarized in Table 1. All figures were generated using the programs PyMOL and MOLMOL (41).
Isothermal Titration Calorimetry-To obtain a direct binding affinity between Vta1NTD (wild-type and mutants) and the Vps60(128 -186) peptide, wild-type Vta1NTD and mutants were titrated with the Vps60(128 -186) peptide using an iTC-200 microcalorimeter (GE Healthcare) at 25°C. All proteins and peptides were exchanged with buffer containing 20 mM sodium phosphate( pH 7.0) and 0.1 M NaCl by gel filtration chromatography, centrifuged to remove any particulates, and degassed. To obtain a direct binding affinity between Vta1NTD variants and the Vps60(128 -186) peptide, a solution of ϳ0.1 mM wild-type and mutant Vta1NTD was titrated with 1.0 mM Vps60(128 -186) peptide, respectively. The accurate concentrations of Vta1NTD and its mutants were determined using their A 280 constants.
Circular Dichroism Spectroscopy of Free Vps60(128 -186)-To probe the folding of free Vps60(128 -186), the CD experiment was performed at 25°C on a JASCO-715 spectropolarimeter (Jasco International Co., Tokyo, Japan). Data were collected at 0.1-nm intervals at a scan speed of 20 nm/min, 1-nm bandwidth, and 0.25-s response time from 250 to 190 nm. Circular quartz cells of 1-and 0.1-cm path length were used for the far-UV regions. The CD intensities are expressed as molar residue ellipticities given in units of degrees cm 2 mol Ϫ1 . The concentration of Vps60(128 -186) was ϳ10 M. The buffer conditions used for running CD spectra were 20 mM sodium phosphate and 50 mM NaCl (pH 7.5).
GST Pulldown Experiments-The experiments were performed following standard procedures in buffer containing 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 5 mM 2-mercaptoethanol. Purified wild-type Vta1NTD and mutants were incubated with GST alone or with GST-tagged Vps60(128 -186) immobilized on glutathione-agarose beads for 3 h at 4°C. The beads were then washed extensively three times with the above buffer, and bound proteins were separated by SDS-PAGE and visualized by Coomassie Blue staining.

RESULTS
NMR Structural Determination-Initially, two basic sets of NMR mixed samples were made: 1) 13 C and 15 N isotope-labeled Vta1NTD with unlabeled Vps60(128 -186) at a stoichiometric ratio of 1:1.2 and 2) 13 C and 15 N isotope-labeled Vps60(128 -186) with unlabeled Vta1NTD at a stoichiometric ratio of 1:1.2, each of them for assignment of NMR signals belonging to the corresponding 13 C-and 15 N-labeled component and its structural determination. The intermolecular NOEs could be correctly assigned by confirming signals observed in three-dimensional 13 C-F1-edited, 13 C/ 15 N-F3-filtered NOESY spectra acquired on both complex samples. In total, assignments of Ͼ96% of the main chain and 95% of the side chain atoms of the residues in the complex were completed. The NMR chemical shift changes of Vta1NTD backbone atoms 1 H and 15 N in the absence and presence of Vps60(128 -186) reveal that Vps60(128 -186) addition mainly induced Vta1NTD MIT2 domain amide 15 N and 1 H chemical shift variations in residues of the Vta1NTD MIT2 domain ( Fig.  2A), suggesting that Vps60-binding sites localize in these regions. This observation accords with the analysis of the electrostatic surface of Vta1NTD in its free state, which shows that the Vta1NTD MIT2 domain is more positively charged than the MIT1 domain (Fig. 2, B and C), suitable for negatively charged Vps60 binding.
The solution structure of the Vta1NTD-Vps60(128 -186) complex was determined by a conventional heteronuclear NMR method using 15 N-or 13 C/ 15 N-labeled protein. In total,   (Fig. 2D) suggests that Vps60(128 -186) in its free state is disordered because the cross-peaks are not dispersed, localizing mainly in the region between 8.0 and 8.5 ppm. To confirm this observation, we preformed CD spectroscopy on free Vps60 (128 -186), where negative absorption at ϳ200 nm shows a random coil conformation (Fig. 2E). Upon binding to Vta1NTD, the cross-peaks in two-dimensional NMR 1 H-15 N HSQC of Vps60(128 -186) became dispersed (Fig. 2D), suggesting that Vps60(128 -186) folds into an ordered structure, coinciding with the structure determined below. This demonstrates that Vta1NTD binding stabilizes Vps60 helix conformation.
Overall Complex Structure-The Vta1NTD-Vps60(128 -186) structure shows that the bound Vta1NTD still has two MIT domains, each of them composed of three ␣-helices (MIT1: helices ␣1, ␣2 and ␣3; and MIT2: helices ␣5, ␣6, and ␣7) (Fig. 1C), almost similar to those observed in its free state (29). Helix ␣4 was much longer in Vta1NTD bound to Vps60(128 -186) than in free Vta1NTD, which might have resulted from the conformational stabilization upon its binding to Vps60 (as demonstrated below), consistent with the secondary structure prediction based on the assignments of backbone atoms 13 C␣, 13 C␤, 13 CO, 1 H, and 15 N (supplemental Fig. S1). The backbone atoms of bound Vta1NTD had a root mean square deviation of 1.07 Å from those of free Vta1NTD (Fig. 2F), indicating that Vps60(128 -186) binding does not induce overall major conformational changes in Vta1NTD. The secondary structure prediction and perceived structural homology to ESCRT-III protein Vps24/CHMP3 suggest that Vps60(128 -186) corresponds to the fourth and the fifth helices within the Vps60 structure (42). In the current complex structure, Vps60(128 -186) indeed folds into two ␣-helices (denoted as ␣4Ј and ␣5Ј), and both helices are involved in the interaction with the MIT2 domain of Vta1NTD (Fig. 1). The two ␣-helices of bound Vps60(128 -186) adopt an overall V-shaped helix-turn-helix structure and straddle on the third helix (␣7) of the Vta1NTD MIT2 domain. The longer helix ␣4Ј consists of residues 140Ј-157Ј and interacts with ␣5 and ␣7, corresponding to the first and third helices of the Vta1NTD MIT2 domain. It runs diagonally from the N-terminal end of ␣5 to the C-terminal end of ␣7, maintaining a general direction parallel to both helices. The polypeptide chain crosses over to the other side of ␣7 near the C-terminal end of ␣7 and continues as ␣5Ј (residues 168Ј-182Ј), running nearly vertical to helices ␣6 and ␣7. The Vta1-Vps60 complex buries a total of ϳ3600-Å 2 surface area at the interface. In contrast to the MIT domains in Vps4, spastin, or AMSH (19 -24), Vps60(128 -186) interacts with Vta1NTD through helices ␣4Ј and ␣5Ј, extending over Vta1NTD MIT2 domain helices 1-3. Thus, the Vta1NTD MIT2 domain displays a fifth and novel ligand recognition mode to bind to Vps60(128 -186) ( Fig. 1D and shown under "Discussion").
To confirm that the changes in helix ␣4 of Vta1NTD are generated due to Vps60(128 -186) binding rather than crystallization, we assigned the chemical shifts of backbone atoms 13 C␣, 13 C␤, 13 CO, 1 H, and 15 N of free Vta1NTD. The secondary structure prediction based on these assignments using the programs CSI (43) and TALOS (36,38) suggests that there are  DECEMBER Fig. S1).
Mutational Analyses of the Vta1-Vps60 Interaction-Mutations were introduced into these observed binding sites to test the importance of the residues to the overall stability of the complex. As shown in Fig. 4, Table 2, and supplemental Based on the sequence alignment (Fig. 4, A and C), the Vta1NTD MIT2 domain derived from S. cerevisiae has low sequence similarity to the other organisms, whereas

Novel Mode of MIT-MIM Interaction-
The MIT domain is a versatile protein-protein interaction domain identified in proteins that have a role in vesicle trafficking, including Vps4, Vta1, AMSH, and UBPY, where they mediate interaction within the ESCRT-III complex (45). The MIT domain recognizes sequence motifs called the MIMs primarily within the ESCRT-III subunits. It has been suggested that the interaction between the MIT domain and MIM acts in regulating the disassembly of ESCRT-III, as well as in targeting specific proteins to the site of ESCRT function. At least four types of MIM (MIM1, MIM2, MIM3, and MIM4) were reported to bind to different sites on the MIT domain ( Fig. 1) (19 -24).
Vta1NTD contains two tandem MIT domains as identified in its crystal structure (29), which mediate the interaction between Vta1 and the ESCRT-III-related proteins Vps60 and Did2. Our NMR structure of Vta1NTD-Vps60(128 -186) shows that Vps60 MIM binds exclusively to the second MIT domain of Vta1. Unlike other MIMs, the Vps60 MIM sequence (residues 140 -186, defined as MIM5) is much longer and forms two helices (␣4Ј and ␣5Ј). The significant difference from the other MIMs is that Vps60 MIM5 can bind both surfaces made up of helices 5 and 7 (Vps4 MIT domain helices 1 and 3) and helices 6 and 7 (Vps4 MIT domain helices 2 and 3) of the Vta1 MIT2 domain. The Vta1 MIT2-Vps60 MIM5 contacts are a mixture of polar and hydrophobic interactions, as is the case for spastin MIT-MIM3. Thus, the structure of the Vta1-Vps60 complex provides a novel recognition mode of the MIT domain with its ligand and extends the diversity of MIT domain interaction surfaces for peptide ligands.
Vps60 Enhances Vta1 Stimulation of Vps4 in a Complex Manner-The dynamic assembly and disassembly of the ESCRT-III polymer play a critical role in ESCRT-mediated membrane deformation events and alterations of Vps4 ATPase activity. To address how Vps60 and Did2 binding enhances Vta1 stimulation of Vps4 ATPase activity, two models were presented (27). One is that their binding to the MIT2 domain results in conformation changes in Vta1; the other is that the interaction between Vta1 and Did2 or Vps60 increases the local concentration of Vta1-Vps4 in vitro. It was reported that removal of the two Vta1 tandem MIT domains (Vta1(165-330)) does not enhance the basal activation of Vps4 by Vta1, implying that Vta1 MIT domains do not autoinhibit Vps4 activation (27). The current NMR structure of the Vta1NTD-Vps60(128 -186) complex provides further evidence that Vps60 binding does not induce overall conformational changes in the N terminus of Vta1 (Fig. 2F) and thus might not lead to further structural arrangement in the C-terminal domain of Vta1. These observations suggest that Vps60 may not allosterically regulate Vta1 and thus could not potentiate its ability to activate Vps4.
Taken together, the complex structure of Vta1NTD-Vps60(128 -186) cannot account for all aspects of Vps4 activation, but it demonstrates a novel MIT recognition mode that has not been reported. Thus, to address how Vps4 ATPase is activated, further studies of the dynamics of ESCRT-III assembly and disassembly will be performed to better understand the precise function of Vta1 in the process of MVB sorting. The current structure further confirms that the interaction between Vps60 and Vta1 stimulates Vps4 ATPase in an unexpected manner.