Essential N-Terminal Insertion Motif Anchors the ESCRT-III Filament During MVB Vesicle Formation

Developmental Cell Article

Essential N-Terminal Insertion Motif Anchors the ESCRT-III Filament during MVB Vesicle Formation

Nicholas J. Buchkovich,1,2,3 William Mike Henne,1,2,3 Shaogeng Tang,1,2 and Scott D. Emr1,2,* 1Weill Institute for Cell and Molecular Biology, Cornell University, Ithaca, NY 14853, USA 2Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY 14853, USA 3These authors contributed equally to this study *Correspondence: [email protected] http://dx.doi.org/10.1016/j.devcel.2013.09.009

SUMMARY neck of the invaginated vesicle and is not incorporated into the nascent vesicle. How the ESCRT machinery achieves this unique The endosomal sorting complexes required for form of membrane deformation and scission is the focus of transport (ESCRTs) have emerged as key cellular extensive research. machinery that drive topologically unique membrane The ESCRT-III polymer, consisting of four core subunits, deformation and scission. Understanding how the Vps20, Snf7, Vps24, and Vps2, is believed to play a key role ESCRT-III polymer interacts with membrane, pro- in membrane invagination and vesicle scission. The transient moting and/or stabilizing membrane deformation, is nature of this hetero-oligomeric filament has made character- izing the filament difficult. However, recent efforts to visualize an important step in elucidating this sculpting mech- the ESCRT-III polymer on lipid monolayers using high-resolution anism. Using a combination of genetic and biochem- electron microscopy techniques have identified a key role for ical approaches, both in vivo and in vitro, we identify electrostatically charged lipids in modulating ESCRT-III filament two essential modules required for ESCRT-III-mem- architecture (Henne et al., 2012). This raises key questions about brane association: an electrostatic cluster and an the ESCRT-III-lipid interaction. For example, how does ESCRT- N-terminal insertion motif. Mutating either module III bind membrane, and what is the role of this interaction in in yeast causes cargo sorting defects in the MVB invaginating membrane, sculpting the vesicle, and promoting pathway. We show that the essential N-terminal scission? Understanding how the ESCRT-III subunits interact insertion motif provides a stable anchor for the with lipid is critical to elucidating the mechanism for ESCRT- ESCRT-III polymer. By replacing this N-terminal mediated membrane deformation. Here, we identify two key motif with well-characterized membrane insertion components for its membrane binding: a charged interaction surface and a membrane-inserting N terminus. Interestingly, modules, we demonstrate that the N terminus of we find that although membrane insertion modules often drive Snf7 has been tuned to maintain the topological con- curvature, the membrane-inserting motif of Snf7 has been tuned straints associated with ESCRT-III-filament-medi- to maximize its role as a membrane anchor for the ESCRT-III ated membrane invagination and vesicle formation. polymer without disrupting the topologically inverted curvature Our results provide insights into the spatially unique, associated with ILV formation. ESCRT-III-mediated membrane remodeling.

RESULTS INTRODUCTION Positively Charged Interfaces on ESCRT-III Interact with The endosomal sorting complexes required for transport Membrane (ESCRTs) are critical for sorting certain transmembrane proteins ESCRT-III subunits are structurally similar, consisting of a core into vesicles that bud into the lumen of the endosome. In addition domain, a ﬂexible linker, an autoinhibitory alpha-helix, and a to sorting cargo into intraluminal vesicles (ILVs) of multivesicular MIT-interacting motif (MIM) (Figure 1A). The core domain con- bodies (MVBs) (Babst et al., 1997, 2002a, 2002b; Katzmann sists of four tightly bundled alpha-helices. The ﬁrst elongated et al., 2001), ESCRTs drive ILV formation and scission. Addition- helix of Vps24, termed helix-1, forms a highly basic surface con- ally, the ESCRT machinery is involved in viral budding from the sisting of arginine and lysine residues (Figure 1B; Figure S1A plasma membrane and in cytokinesis, where they function as available online) (Muzio1 et al., 2006). Alignment analysis showed part of the scission machinery (Carlton and Martin-Serrano, not only a high degree of conservation of the positive charges on 2007; Garrus et al., 2001; Spitzer et al., 2006). ESCRT-depen- helix-1 between the human and yeast orthologs of Vps24 (12 and dent membrane deformation is unique when compared to other 10, respectively) but also with the ESCRT-III subunit Vps2 (11) vesicle formation events in the cell as it is topologically inverted (Figure 1B). These conserved, positively charged residues may relative to those formed by coats, such as clathrin, COPI, or constitute the membrane interaction interface during binding of COPII. Furthermore, the ESCRT machinery localizes only to the ESCRT-III to endosomes.

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 201 Developmental Cell ANCHRing ESCRT-III to Membrane

MIMMIM A helix-1 helix-2 h-3 h-4 h-5 MIM B

heh l-3 heli 3 lix-x 4 helix-5 hhellix-2 helix-1 CHMP3 23-IRKEMRVVDRQIRDIQREEEKVKRSVKDAAKKG +12 MMIMM Vps24 24-LRKNGRNIEKSLRELTVLQNKTQQLIKKSAKKN +10 ** * ** Vps2 25-LERTQRELEREKRKLELQDKKLVSEIKKSAKNG +11 * * * * Snf7 26-LREHINLLSKKQSHLRTQITNQENEARIFLTKG +6 * * * * * Changed to E for membrane binding analysis

C cell fractionation D cell fractionation E cell fractionation F % Protein Associated with P13 fraction 100% Snf7 80% S13 P13 Vps24 S13 P13 S13 P13 60% WT WT WTW Vps2 40% +1E +2E +2E % P13 +3E 20% +2E +4E +4E 0% +3E vps4∆;snf7∆ * vps4∆;vps2∆ WT +2E +4E +4E vps4 vps24∆* ∆; H cell fractionation I in vitro liposome sedimentation WT 2KE K71K79 K112 K115 BUBU G K60 core S13 P13 Snf7 90° K64 K68 1KE % Bound: 59% ±1% 21% ±3% S. cerevisiae 66-ALKKKKTI 109-EGAKAMKTI 2KE C. albicans 60-ALKRKKGY 107-QGAKAMKQI X. laevis 66-ALKRKKRY 109-FAAKAMKAA 3KE M. musculus(4B) 68-ALKRKKRY 111-YAAKAMKAA H. sapiens(4B) 68-ALKRKKRY 111-YAAKAMKAA 4KE vps4∆;snf7∆

Figure 1. Identiﬁcation of Electrostatic Membrane Binding Interfaces on ESCRT-III Subunits (A) Schematic showing helical arrangement of ESCRT-III subunits both linearly (top) and in the four-helical core bundle (bottom). Core helices and MIM (green); autoinhibitory helix-5 (yellow). (B) Helical wheel depicting positively charged interface on helix-1 of hVps24. Positively charged resides (blue); negatively charged residues (red). Cartoon depicts potential interaction with membrane. Sequence alignments of helix-1 of ESCRT-III subunits (positive charges, blue). Asterisk indicates residues mutated in this study. (C–E) Cellular fractionation analyses of vps4D and (C) vps24D, (D) vps2D, or (E) snf7D yeast exogenously expressing the respective wild-type gene (WT), or a gene with one (+1E), two (+2E), three (+3E), or four (+4E) charge inversions on helix-1. Asterisk indicates darker exposures. (F) Graphs representing percentage of protein associated with membrane fraction (P13) from data in (C)–(E). (G) Cartoon schematic and electrostatic surface model (positive residues, blue; negative residues, red) portraying the location of positively charged residues from helices-2 and -3. Sequence alignments of positive charges (blue) on this interface among eukaryotic species. (H) Cellular fraction analysis of vps4D;snf7D yeast exogenously expressing WT Snf7 or a gene with one (1KE), two (2KE), three (3KE), or four (4KE) charge inversions on interface described in (G). (I) Liposome sedimentation assay of Snf7core (WT) and Snf7core/2KE (2KE). Protein-free (unbound, U); liposome-associated protein (bound, B). See also Figure S1.

To investigate the role of helix-1 in membrane binding, we three, or four residues on helix-1 sequentially reduced the generated mutants of Snf7, Vps2, and Vps24 that had an Vps24 association with the membrane fraction (Figures 1C and increasing number of charge inversions on helix-1. In the absence 1F). Although membrane-associated Vps24 was reduced after of Vps4, which recycles ESCRT-III subunits from membrane, three or four charge inversions, we also noticed that these muta- greater than 90% of Vps24 localized to the P13 membrane fractions affected protein stability (Figure S1B). To ensure this reduc- tion in cellular fractionation experiments (Figures 1C and 1F). tion was not due to the instability of the mutant Vps24, we There was very little effect on this membrane association with a changed two positive charges to glutamate and two to neutral single charge inversion. However, inverting the charges of two, alanine residues (Vps24EEAA). This approach led to a stable

202 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

construct that, similar to the four charge inversion mutant, had the N termini of the ESCRT-III subunits by assaying the delivery reduced localization to the membrane fraction (Figure S1C). of GFP-Cps1 to the vacuole. Similar to what was previously These results show that the positive charges on helix-1 of reported for the Vps20 myristoylation mutant (Babst et al., Vps24 are important for membrane association. Accumulating 2002a), deletion of the N-terminal region upstream of helix-1 charge inversions on helix-1 of Vps2 produced similar results to on Vps20 resulted in a partial block in GFP-Cps1 sorting. The Vps24 (Figures 1D and 1F). Importantly, mutations on helix-1 of cargo accumulated both in the vacuole lumen and on the vacu- Vps2 did not abolish its ability to bind to and modify Snf7 fila- ole membrane, but not at the aberrant endosomal E dot, where ments, suggesting that they did not affect its ability to interact cargo accumulates when Vps20 is absent (Figure 2B). Similar with other ESCRT-III subunits (data not shown). This suggests to Vps20, deletion of this N-terminal region on Vps2 also resulted that both Vps2 and Vps24 utilize helix-1 for their interaction with in a partial sorting defect (Figure 2E). This was in contrast to dele- membrane. tion of the homologous N-terminal regions of Snf7 or Vps24, in Strikingly, similar mutations on Snf7 resulted in greater than which GFP-Cps1 was present at the E dot, suggesting a severe 90% of protein remaining membrane associated (Figures 1E sorting defect (Figures 2C and 2D). These data suggest that the and 1F), indicating that helix-1 may not contribute to the mem- extreme N termini of the ESCRT-III subunits are functionally brane binding of Snf7. Sequence alignments revealed that important for efficient cargo sorting. many of the helix-1-positive charges are not conserved on To better understand the partial sorting defects observed with Snf7 (Figure 1B). Because Snf7 is the most abundant ESCRT- the N-terminal deletion mutants, we used a quantitative sorting III subunit (Teis et al., 2008), understanding its membrane inter- assay that takes advantage of the pH-sensitive nature of the action is essential. An alternative membrane-binding interface GFP-derivative pHluorin and the acidic environment of the on Snf7, consisting of a series of highly conserved lysines on vacuole. By tagging the plasma membrane cargo Mup1 with helices-2 and -3, was identified from a structure generated using pHluorin and monitoring its fluorescence after internalization, the comparative structural program Modeler (Figure 1G) (Eswar one can determine whether the cargo is efficiently delivered to et al., 2008). Single or multiple charge inversions on this interface the vacuole, partially missorted, or trapped at the E dot. By impaired the sorting of Mup1-pHluorin to the vacuole lumen (Fig- calculating the percent of fluorescence that is quenched, the ure S1F). These inversions also decreased the membrane asso- efficiency of sorting can be quantified and used to assess the ciation of Snf7 by cellular fractionation (Figure 1H). These results severity of the sorting defect (Figure 2F). Using this approach, were confirmed by light microscopy, as the charge inversion we found that Vps20 and Vps2 had sorting efficiency ratings of mutants increasingly relocalized Snf7-GFP to the cytoplasm 71% and 74%, respectively, indicative of the partial sorting (Figure S1H). To further investigate these residues, we per- defect observed with GFP-Cps1. In contrast, Snf7 and Vps24 formed liposome sedimentation assays using the Snf7core had sorting efficiencies of 13% and 25% (Figure 2G), consistent domain, which is sufficient to bind membrane in vitro (Henne with severe sorting defects. These results confirm that although et al., 2012). As expected, introducing two charge inversions the N termini of all four ESCRT-III subunits are required for effi- from helix-3 onto the Snf7core background reduced membrane cient cargo sorting, they do not all contribute equally. association (Figure 1I). This provides direct in vitro confirmation Because the N terminus of Snf7 had the greatest effect on of our in vivo results, placing residues from helices-2 and -3 at cargo sorting, we replaced the N termini of the other ESCRT-III the membrane interface. Importantly, mutations on this interface subunits with the N terminus of Snf7. This resulted in completely did not disrupt the ability to form filaments, suggesting that the normal MVB sorting (Figure S2E). In contrast, replacing the N ter- altered membrane localization was not due to disrupting Snf7’s minus of Snf7 with that of Vps24 or Vps2 only weakly rescued ability to oligomerize (data not shown). Thus, Snf7 surprisingly cargo sorting (Figure S2F). This was surprising due to the homol- utilizes a different cationic interface for membrane binding, sug- ogy between the N termini of Snf7 and Vps2 (Figure 2A). Accord- gesting that ESCRT-III subunits uniquely orient in the polymer to ingly, changing a negatively charged glutamate in Vps2 to contact membrane. glycine (E5G), as it is in Snf7, greatly enhanced this rescue (Fig- ure S2F). These data confirm the differential importance of the N The N Termini of ESCRT-III Subunits Contribute to termini of ESCRT-III subunits in MVB sorting. Because Snf7 is Membrane Binding the most abundant ESCRT-III subunit and is the predominant One ESCRT-III subunit, Vps20, is N-terminally myristoylated. subunit in the ESCRT-III polymer and has the greatest N-terminal This lipid modification often works in cooperation with other lipid deletion phenotype, we focused our analysis on the N terminus binding modules, such as an electrostatic interface, to provide a of Snf7. high-affinity protein-membrane interaction. We wondered if the We next used GFP-fusion constructs to assess the effect of other ESCRT-III subunits were similarly utilizing a second mem- the N-terminal deletion on Snf7 endosomal localization. When brane-interacting module. Many membrane-associated proteins expressed in WT yeast, Snf7-GFP was located in puncta. Dele- utilize hydrophobic residues that insert into lipid bilayers to tion of the N terminus of Snf7 redistributed Snf7-GFP from its stabilize membrane binding. We noted the presence of bulky punctal localization to the cytoplasm (Figure 2H), consistent hydrophobic residues at the N termini of Snf7, Vps2, and with a role for the N terminus in associating with membrane. Vps24 (Figure 2A) and hypothesized that, similar to the myristoy- To further demonstrate that the N-terminal sorting defect was lation of Vps20, these residues stabilize membrane interactions. due to reduced membrane binding, we performed liposome Consistent with this idea, liposome binding of human Vps2 was sedimentation assays using Snf7core and DN-Snf7core. As ex- dependent on two conserved N-terminal hydrophobic residues pected, deletion of the N terminus nearly completely abolished (Bodon et al., 2011). We tested the functional significance of the ability of Snf7core to associate with membrane (Figure 2I).

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 203 Developmental Cell ANCHRing ESCRT-III to Membrane

A ESCRT-III core subunits: Vps20 1-MGQKSS Snf7 1-MWSSLFGWTSS Vps24 1-MDYIKKAIW Vps2 1-MSLFEWVFG

B Vps20 WT Vector ∆N-Vps20 C Snf7 WT Vector ∆N-Snf7

v v v v v v GFP-Cps1 GFP-Cps1 Luminal GFP E dot Luminal GFPE dot E dot DIC DIC

vps20∆ snf7∆

D Vps24 WT Vector ∆N-Vps24 E Vps2 WT Vector ∆N-Vps2

v v v vvv v v GFP-Cps1 GFP-Cps1 Luminal GFP E dot E dot Luminal GFP E dot DIC DIC

vps24∆ vps2∆

FHQuantitative Mup1-pHluorin G Snf7-GFP ∆N-Snf7-GFP Sorting Assay 100% WT 100% 80% 80%

Weak Sorting Defect 60% GFP 60% Intermediate Sorting Defect 40%

40% % Sorting % Sorting Strong Sorting Defect 20% 20% No Sorting - Class E 0%

0% DIC N- N- N- N- I in vitro liposome sedimentation Vps20 Snf7 Vps24 Vps2 WT ∆N SEY6210 BU BU Snf7core % Bound: 66% ±6% 4% ±3%

Figure 2. ESCRT-III Subunits Require the N Terminus for Cargo Sorting (A) Sequence alignment of ESCRT-III N termini. Hydrophobic residues (green); N-terminal myristoylated glycine (gray). (B–E) Representative images of mid-log (B) vps20D, (C) snf7D, (D) vps24D, and (E) vps2D yeast exogenously expressing GFP-Cps1 and the WT, DN, or vector control (Vec) for the corresponding ESCRT-III subunit. GFP images (upper row); composite images of GFP and DIC (bottom row). White arrows denote the GFP signal localized within the vacuole lumen and at the aberrant endosome (class E compartment). Vacuole (v). (F) Schematic of quantitative sorting scale for Mup1-pHluroin sorting assay. (G) Quantitative sorting data for DN-ESCRT-III subunits. Error bars represent standard deviation. (H) Representative images of mid-log yeast exogenously expressing Snf7-GFP or DN-Snf7-GFP. GFP images (upper row); composite images of GFP and DIC (bottom row). (I) Liposome sedimentations of Snf7core or DN-Snf7core. Protein-free (unbound, U); liposome-associated protein (bound, B). See also Figure S2.

204 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

A B Figure 3. Snf7 Requires N-Terminal Hydrophobic Residues (A) Sequence alignment of N termini of Snf7 in eukaryotic species. Hydrophobic residues (green). Snf7Snf7Snf7 F (B) Helical wheel depictions of Snf7 N terminus. S. cerevisiae 1-MWSSLFGWTSS snf7D 1-111-111-11 W (C) Representative images of mid-log yeast C. albicans 1-MWGYFFGGNSQ M exogenously expressing GFP-Cps1 and Snf7W2E, X. laevis 1-MSLIGKLFGTGGK W L Snf7F6E, or Snf7F6W. GFP images (upper row); composite M. musculus(4B) 1-MSVFGKLFGAGGG H. sapiens(4B) 1-MSVFGKLFGAGGG images of GFP and DIC (bottom row). White arrows denote luminal GFP. Vacuole (v). (D) Quantitative sorting data for Snf7, Snf7W2E, Snf7F6E, C D and Snf7F6W. Error bars represent standard deviation. W2E F6E F6W 100% (E) Cellular fractionation analyses of snf7D;vps4D yeast exogenously expressing WT Snf7, Snf74KE (4KE), and v Snf7F6E/4KE. 4KE sample from same set in Figure 1H. v 80% v (F) Representative images of mid-log yeast exogenously expressing Snf7-GFP, Snf7F6E-GFP, and Snf7F6E/4KE- GFP-Cps1 60% Luminal GFP GFP. GFP images (upper row); composite images of GFP

% Sorting and DIC (bottom row). 40% (G) Liposome sedimentation assays of Snf7core, Snf7core/W2E, Snf7core/F6E, and Snf7core/F6W. Protein-free DIC 20% (unbound, U); liposome-associated protein (bound, B). See also Figure S3. snf7∆ 0% WT W2E F6E F6W

E cell fractionation F Snf7-GFP Snf7F6E-GFP Snf7F6E/4KE-GFP phobic residues to membrane binding. Map- ping the N-terminal sequence of Snf7 on a

S13 P13 GFP helical wheel revealed an amphipathic helix, WTW with a hydrophobic surface made up of these 4KE large N-terminal residues aligned opposite of a polar interface made up of several serine res- F6E/4KE

DIC idues (Figure 3B). Substitution of the N-termi- snf7∆; vps4∆ nal tryptophans (Snf7W2E and Snf7W8E) or the F6E SEY6210 highly conserved phenylalanine (Snf7 ) with negatively charged glutamate resulted in the G in vitro liposome sedimentation in vitro liposome sedimentation aberrant sorting of both GFP-Cps1 (Figures WT W2E WT F6E F6W BUBU BUBU BU 3C and S3B) and Mup1-pHluorin (Figures 3D Snf7core Snf7core and S3C). Importantly, replacing the phenylal- % Bound: anine with other hydrophobic residues did not 60% ±3% 10% ±3% % Bound: 68% ±2% 5% ±1% 58% ±9% affect the sorting of either cargo. Point mutations of key residues on the other ESCRT-III Similar to what was observed for the human Vps2 (Bodon et al., subunits similarly resulted in reduced sorting of Mup1-pHluorin 2011), deletion of the N terminus of Vps2 also resulted in reduced (Figure S3D), thus demonstrating the importance of the hydro- membrane association, although not nearly as drastic as Snf7 phobicity of the N terminus. (Figure S2H). This provides further correlation between the We next sought to correlate the functional defect associated sorting results and membrane association. with the hydrophobicity of the N terminus with membrane asso- To ensure the altered localization was not due to oligomeriza- ciation. Importantly, addition of the F6E mutation to the cationic tion defects, we introduced a mutation on DN-Snf7 that pro- interface mutant (4KE) further reduced membrane binding, motes Snf7 filament formation (Henne et al., 2012). DN-Snf7R52E reducing the amount of protein associated with the P13 fraction was able to form filaments on EM grids, confirming that the N ter- from 52% to 24% (Figure 3E). By light microscopy, addition of minus did not affect oligomerization (Figure S2I). Furthermore, F6E to Snf7-GFP redistributed the protein from puncta to the filaments formed by Snf7R52E can be clearly visualized associ- cytoplasm, similar to what was observed for DN-Snf7-GFP ating with liposomes by electron microscopy. However, deletion (Figure 3F). Although some puncta were still present, the F6E of the N terminus abolished this association (Figure S2J). These mutation in combination with the cationic interface mutant data clearly demonstrate that the N terminus of Snf7 is important (Snf7F6E/4KE-GFP) resulted in nearly complete cytoplasmic for membrane association. localization. These results confirm a role for both the hydrophobic N terminus and the cationic interface of Snf7 in membrane The N Terminus of Snf7 Requires Hydrophobic Residues binding. By sequence alignment, we noticed the conserved position of Liposome sedimentation assays were used as an additional key hydrophobic residues at the N terminus (Figure 3A). We confirmation for the role of the N-terminal hydrophobics in mem- next wanted to investigate the contribution of these large hydro- brane binding. Both purified Snf7core/W2E and Snf7core/F6E had

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 205 Developmental Cell ANCHRing ESCRT-III to Membrane

A h-3 B 160000 F6C C 6 h-2 F6C + Lipos H118C H188C Q75C H188C + Lipos 5 120000 Buffer Liposomes Alone 4 F6C K35C N-term h-1 80000 3

Fluorescence 40000 2

0 1

500 550 600 Absorbance at 530 nm Relative 0 Wavelength F6C K35C Q75C H118C N-term h-1 h-2 h-3

D 100 mol% POPC

Filter: No Lipos 800 nm 100 nm 30 nm BUBUBUBU WT Snf7core: 8% ±1% 33% ±4% 32% ±3% 43% ±3% F6E 5% ±3% 4% ±4% 3% ±6% 3% ±3% in vitro liposome sedimentation

Figure 4. The N Terminus of Snf7 Inserts into Membrane (A) Cartoon schematic of NBD-labeled residues. (B) Emission spectra for buffer, liposomes, and Snf7core/F6C or Snf7core/H118C with and without liposomes. (C) Graph of relative absorbance at 530 nm ((protein + liposome sample) / protein alone sample) for Snf7core/F6C (N-term), Snf7core/K35C (h-1), Snf7core/Q75C (h-2), and Snf7core/H118C (h-3). (D) Liposome sedimentation assay of Snf7core and Snf7core/F6E with liposomes of varying sizes. Protein-free (unbound, U); liposome-associated protein (bound, B). Liposomes were visualized by negative-stain transmission electron microscopy. Scale bars, 100 nm. See also Figure S4. impaired binding to liposomes in vitro. In contrast, Snf7core/F6W cysteine residues introduced into the N terminus (Snf7core/F6C, associated with liposomes at similar levels as the wild-type helix-1 (Snf7core/K35C), helix-2 (Snf7core/Q75C), or helix-3 (WT) Snf7core (Figure 3G). Thus, two regions in Snf7 mediate (Snf7core/H118C)(Figures 4A and S4A). A large increase in emis- membrane interactions: a positively charged cationic interface sion intensity and a shift in maximum emission was observed and an N-terminal hydrophobic motif. upon addition of excess liposomes to the Snf7core domain labeled at the N terminus (Figures 4B, 4C, and S4B). In contrast, The N Terminus of Snf7 Inserts into the Membrane addition of liposomes to Snf7core labeled with NBD on helices-1, N-terminal amphipathic helices on proteins, such as Arf1 or the -2,or-3 resulted in only a very slight increase in intensity (Figures N-BAR family, insert into membrane bilayers to induce mem- 4B, 4C, and S4C). Thus, the N terminus shifts to a hydrophobic brane curvature (Farsad and De Camilli, 2003). Because the N environment in the presence of liposomes. terminus of Snf7 is involved in membrane binding and is depen- Membrane-inserting motifs prefer to bind liposomes with dent on hydrophobic residues, we reasoned that these residues smaller radii because of their high-membrane curvature (Bigay might insert into the membrane bilayer to generate or stabilize et al., 2005). To investigate if the Snf7 N-terminal motif could membrane deformations. To test this, we first utilized the fluores- sense membrane curvature, we tested the ability of Snf7core to cent dye 7-nitrobenz-2-oxa-1,3-diazole (NBD), which undergoes bind liposomes of different sizes. To mitigate the contribution of an increase in emission intensity and a change in its maximum the cationic interface, we used uncharged liposomes composed emission upon shifting to a hydrophobic environment (Johnson, of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) alone, 2005; Saksena et al., 2009). We labeled Snf7core with NBD on generated by extrusion through progressively smaller filter pores

206 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

in vitro liposome sedimentation ABALPS h0-NBAR ARF WT ΔN h0-NBAR BU BU B U Osh4 Rvs161 Arf1 Snf7core

% Bound: 62% ±5% 6% ±1% 39% ±8%

C 100%

100% 80%

80% 60%

60%

% Sorting 40%

% Sorting 40% 20%

20% 0% Snf7 Arf1- Arf1- Myr- Myr- 0% WT Snf7 Snf7 Snf7 Snf7 Snf7 ALPS- h0- Arf1- ∆N G2A G2A WT Snf7 NBAR Snf7 Snf7- Snf7

D E WT Snf7 h0-NBAR-Snf7 Myr-Snf7 Average ILV 50 Diameter 40 30 20 10

Diameter (nm) 0 Snf7 h0- Myr- WT NBAR Snf7 Snf7

Figure 5. Snf7 Requires a Membrane-Inserting N-Terminal Motif (A) Cartoon schematic with N-terminal helical wheels of chimeras used in quantitative sorting assay. Graph represents quantitative sorting data for snf7D yeast exogenously expressing WT Snf7, ALPS-Snf7, h0-NBAR-Snf7, Arf1-Snf7, and DN-Snf7. (B) Liposome sedimentation assays of Snf7core, DN-Snf7core, and h0-NBAR-Snf7core. Protein-free (unbound, U); liposome-associated protein (bound, B). (C) Quantitative sorting data for Snf7, Arf1-Snf7, Arf1-Snf7G2A, Myr-Snf7, and Myr-Snf7G2A. (D) Representative EM images of ILV-containing MVBs from snf7D;vam7D yeast exogenously expressing Vam7ts and Snf7 (left panel), h0-NBAR-Snf7 (middle panel), or Myr-Snf7 (right panel). Scale bar, 100 nm. (E) Graph of mean ILV size of samples in (D). N = 100. Error bars in (A), (C), and (E) represent standard deviation. See also Figure S5.

(800, 100 and 30 nm). In liposome sedimentation assays, Snf7core testing if unrelated membrane-inserting motifs could rescue bound best to liposomes prepared with the smallest ﬁlter. Lipo- the DN-Snf7-sorting defect. We replaced the N terminus of some binding was completely dependent on the N terminus, as Snf7 with the well-characterized membrane-inserting modules the Snf7core/F6E did not bind liposomes, even those with high- of the curvature-sensing ALPS motif of Osh4 (ALPS-Snf7), cur- membrane curvature (Figure 4D). This observation, in combina- vature-inducing helix-0 of Rvs161 (h0-NBAR-Snf7), or the scis- tion with the previous results, provides strong evidence that sion-promoting N terminus of Arf1 (Arf1-Snf7) (Beck et al., Snf7 contains a membrane-inserting N-terminal motif. 2011; Bigay et al., 2005; Drin et al., 2007). Both the well-characterized helix-0 of Rvs161 and the ALPS motif were able to rescue The Snf7 N Terminus Can Be Replaced by Known the sorting of Mup1-pHluorin (Figure 5A). This rescue was Membrane-Inserting Motifs dependent on key N-terminal hydrophobic residues, similar to We next investigated if the N terminus of Snf7 is a general inser- what was observed for Snf7, as substituting these residues for tion module or a sequence-speciﬁc recognition module by glutamate failed to rescue cargo sorting (Figure S5C). As

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 207 Developmental Cell ANCHRing ESCRT-III to Membrane

expected, the ability to rescue cargo sorting correlated with a invagination is negative (Figure 6A). Whereas the helical nature rescue in membrane association, as h0-NBAR-Snf7core was of the ESCRT-III oligomer (Henne et al., 2012) could stabilize able to bind liposomes in vitro (Figure 5B). the negative curvature of the invagination, the insertion of its Unexpectedly, addition of the Arf1 N-terminal amphipathic he- N-terminal amphipathic anchor would have a competing effect lix to DN-Snf7 did not rescue Mup1-pHluroin sorting (Figure 5A), and promote the opposing positive curvature. despite maintaining the ability to localize to membrane puncta The ability of an amphipathic helix to insert into and deform (Figure S5D). Because this helix is posttranslationally modified membrane is dependent upon both the residency time in mem- with a myristoyl group, we reasoned that this modification would brane and the shape (length and bulkiness) of the helix. Helices, adversely affect Snf7 function by providing too high of an affinity such as the ALPS motif, that are believed to sense, but not reaction with membrane. To test this, we replaced the N terminus actively promote, deformation contain polar residues on the of Snf7 with the myristoylation sequence of Vps20 (Myr-Snf7) hydrophilic interface (Drin et al., 2007). In contrast, amphipathic and found, surprisingly, that it was sufficient to rescue Mup1- helices that contain positively charged residues on the hydro- pHluroin sorting. This rescue was dependent on the myristoyl phillic interface have increased residency time in the membrane group, as mutating the glycine residue at position two to alanine because of the interaction of these positive charges with the (Myr-Snf7G2A) did not rescue Mup1-pHlourin sorting (Figure 5C). negatively charged lipid head groups (Drin and Antonny, 2010). Perhaps even more surprising, we found that mutation of the Insertion of an amphipathic helix with high deforming properties myristoylated glycine residue of Arf1-Snf7 to an alanine (Arf1- that would generate positive curvature would compete with the Snf7G2A) also restored sorting to wild-type levels (Figure 5C). negative curvature stabilized by the oligomer and would be Therefore, addition of the unmyristoylated, but not myristoy- detrimental to intraluminal vesicle formation. In support of this lated, Arf1 N-terminal helix to Snf7 is sufficient to rescue cargo hypothesis, the properties of the ANCHR motif of Snf7 are sorting. These data show that a myristoyl group or a known consistent with anchoring and not deformation-promoting heli- membrane-inserting amphipathic helix, but not both, are able ces, namely, it is small in size and contains only polar residues to function at the N terminus of Snf7. on the hydrophilic interface (Figure 3B). Experimentally, this We were intrigued by the rescue of cargo sorting by both the would explain why the myristoyl group and ALPS motif were bet- membrane curvature-sensing and curvature-inducing helices ter able to rescue cargo sorting than were the chimeras with and the myristoyl group. We reasoned that one role of the N ter- known membrane-deforming helices, such as the helix-0 of minus of Snf7 could be to insert into the bilayer and drive N-BAR proteins, ENTH domain, and the N terminus of Arf1 (Fig- membrane curvature during ILV morphogenesis. Alternatively, ures 5A and S5F). membrane deformation could be driven by the previously visual- To further test the effect of altering the insertion properties of ized ESCRT-III helices (Hanson et al., 2008; Henne et al., 2012). the ANCHR motif of Snf7 on intraluminal vesicle formation, we In this model, the predominant role of the N terminus is to act as a generated several N-terminal amino acid extensions. We first membrane-inserting anchor to allow the polymer assembly to added a string of positive charges (6K-Snf7) and then converted exert the mechanical forces necessary for driving membrane it to an amphipathic helix by substituting three lysines for three deformation. Because the well-characterized modules that tryptophans (6WK-Snf7; MWKKWWK). We then varied the rescue cargo sorting are predicted to deform membrane to length of the N terminus by adding lysine and tryptophan differing degrees, the first model would predict that they would residues to generate progressively longer helices with more produce intraluminal vesicles of varying morphology; the amphi- membrane deforming potential (10WK-Snf7 and 14WK-Snf7). pathic helices that promote deformation would result in smaller As expected, the 6K-Snf7 construct did not rescue sorting (Fig- vesicles. In contrast, the latter model predicts that the N terminus ure 6B), demonstrating that positive charges were insufficient to allows the mechanical force of ESCRT-III polymer assembly to anchor Snf7 and reaffirming the requirement for a membrane-in- drive deformation and that the intraluminal vesicle morphology serting element. In contrast, the 6WK-Snf7 construct sorted would be independent of the N-terminal insertion module. Mup1-pHluorin at wild-type levels. Intriguingly, increasing the Expression of either h0-NBAR-Snf7 or Myr-Snf7 resulted in length of the N-terminal motif through addition of tryptophan ILVs with wild-type size and morphology (Figures 5D and 5E), and lysine residues progressively decreased the ability of the providing support for the latter model. We thus propose that chimeras to efficiently sort Mup1-pHluorin (Figure 6B). Further the ESCRT-III oligomer is the predominant driver and stabilizer analysis of the 10WK-Snf7 mutant showed that its decreased of membrane deformation during intraluminal vesicle morpho- ability to sort cargo was not due to its mislocalization in cells (Fig- genesis. The N terminus, by inserting into membrane, provides ure S6B) or the inability to bind membrane. In fact, 10WK- a high-affinity anchor that allows for the oligomer-membrane Snf7core bound liposomes in vitro better than did both Snf7core interaction to drive this deformation. We will hereafter refer to and 6WK-Snf7core (Figure 6C), two N termini that efficiently the N terminus of Snf7 as the amphipathic N terminus containing rescue the DN-Snf7-sorting defect in vivo. hydrophobic residues motif (ANCHR). To further demonstrate the negative correlation between efficient cargo sorting and promoting membrane deformation, The Snf7 ANCHR Motif Balances Opposing Curvatures we assayed the ability of the chimeras to deform membrane. during ILV Formation The insertion of hydrophobic amino acids into membrane can Generation of an intraluminal vesicle requires maintaining an generate bilayer asymmetry and drive membrane curvature intricate balance of competing curvatures. Although the rim of (Gallop et al., 2006). In order to assay the membrane curva- the invagination contains a high degree of positive curvature, ture-generating ability of the N-terminal chimeras separate the curvature of the membrane at the inside and bottom of the from the curvature generated by the filaments, we took

208 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

Figure 6. ILV Formation Requires Maintaining Delicate Balance of Curvatures (A) Cartoon schematic demonstrating distribution of positive (blue) and negative (red) curvatures in a membrane invagination. (B) 3D-helical wheel depictions of Snf7 chimeric N termini. Quantitative sorting data for WT Snf7, 6K-Snf7, 6WK-Snf7, 10WK-Snf7, and 14WK-Snf7. (C) Liposome sedimentation assays of Snf7core, 6WK-Snf7core, and 10WK-Snf7core. Protein-free (unbound, U); liposome-associated protein (bound, B). (D) Representative EM images of samples containing liposomes and 10 mM Snf7core (left panel), h0-NBAR-Snf7core (middle panel) or 10WK-Snf7core (right panel). Scale bars, 100 nm. Red arrows highlight areas of deformation. Cartoon schematic illustrates different degrees of membrane deformation. See also Figure S6. advantage of the Snf7core domain, which is unable to form Snf7 S6C). Although some tubules were present after addition of filaments (Henne et al., 2012). Addition of purified Snf7core to 10WK-Snf7core to liposomes, most of the liposomes were con- liposomes did not result in tubulation or deformation of lipo- verted to small vesicles (Figures 6D and S6C). Tubulation has somes. In contrast, addition of purified h0-NBAR-Snf7core to been described as a highly unstable vesicle budding intermedi- liposomes resulted in membrane tubule formation, which ap- ate formed by scaffolding proteins, such as a BAR domain or peared to be stabilized by a protein coat (Figures 6D and coat protein, which can subsequently promote vesiculation

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 209 Developmental Cell ANCHRing ESCRT-III to Membrane

(Boucrot et al., 2012; Mim et al., 2012). Addition of the highly DISCUSSION deforming 10WK-Snf7core promoted this transition to vesiculation. Although whether the vesicles were independent vesicles Although ESCRT-III drives MVB morphogenesis, the mecha- or connected like beads on a string was unclear, 10WK-Snf7core nistic basis for membrane remodeling is unclear. The inability clearly promoted the highest degree of membrane remodeling. to visualize the ESCRT-III polymer at atomic resolution has hin- Notably, a similar trend in the ability to deform membrane was dered efforts to elucidate ESCRT-III’s role in membrane defor- observed when full-length constructs of the chimeras were mation and scission. Our approach to address these issues used (Figure S6D). These results are consistent with the has been to focus on the distinct functions of ESCRT-III: (1) decreasing ability of the respective chimeras to sort cargo, membrane binding, (2) oligomer assembly, (3) cargo sequestra- providing evidence that the ability to strongly induce positive tion, (4) membrane invagination, (5) ILV neck narrowing, and (6) curvature during ILV invagination decreases the ability to effi- scission. We have recently provided visual evidence of the ciently sort cargo. ESCRT-II/ESCRT-III ring architecture capable of cargo seques- tration (Henne et al., 2012). Although constructs that promote The Snf7 ANCHR Motif Alters the Kinetics of Cargo filament formation have provided insights into the nature of the Sorting ESCRT-III filament (Ghazi-Tabatabai et al., 2008; Hanson et al., The factors involved in generating and stabilizing the membrane 2008; Lata et al., 2008), other assays are required to define curvature associated with ILV morphogenesis, both positive and ESCRT-III’s role in membrane remodeling. negative, have been tuned to maintain the delicate balance of Recent studies have highlighted that ESCRT-III accumulates opposing curvatures. Disrupting this balance decreases the effi- at the highly curved neck of an invagination. This, and the fact ciency of vesicle formation and slows the kinetics of cargo sort- that ESCRT-III is not consumed during ILV morphogenesis, ing. Although steady-state sorting assays were sufficient to underscores the importance of studying ESCRT-III membrane detect mutants that adversely affect this intricate balance, we interactions. Through a series of complimentary in vivo and sought to further test these mutants using a kinetic assay to in vitro approaches, we provide a detailed analysis of ESCRT- monitor cargo sorting. We monitored the fluorescence quench- III membrane binding and remodeling. We identified two features ing of Mup1-pHluorin after methionine stimulation using a mutant shared among the four core ESCRT-III subunits important for with impaired membrane localization (Snf7F6E), with a nonde- membrane binding: a cationic interface consisting of a cluster forming N-terminal motif (Myr-Snf7) and a deformation-promot- of positively charged residues on one face of the ESCRT-III pro- ing N-terminal helix (h0-NBAR-Snf7). As expected, Myr-Snf7 teins and a membrane-inserting N-terminal ANCHR motif con- sorted with wild-type kinetics, whereas Snf7F6E exhibited a sisting of an amphipathic helix on Snf7, Vps24, and Vps2, and very slow quenching rate (Figures 7A and 7B). Interestingly, a myristoyl group in Vps20. The first feature, the cationic inter- cargo sorting by h0-NBAR-Snf7 was significantly delayed face, was unexpectedly different on Snf7, suggesting that the compared to the wild-type sample, even though under steady- subunits may uniquely orient in the ESCRT-III polymer. This het- state conditions h0-NBAR-Snf7 had the ability to significantly erogeneity underscores the functional difference among a group rescue cargo sorting (Figures 7A and 7B). Thus, although h0- of structurally similar subunits. The presence of the N-terminal NBAR-Snf7 functions, it does so less efficiently than wild-type ANCHR motif is not surprising as the generation of other cellular Snf7, consistent with the hypothesis that the presence of the vesicles is associated with membrane-inserting motifs, including NBAR helix-0 disrupts the delicate membrane curvature equilib- the topologically equivalent budding of influenza virus, which rium in ILV formation. By light microscopy, Mup1-pHluorin was requires the amphipathic helix of the M2 viral protein (Antonny found in highly dynamic green puncta in the wild-type and et al., 1997; Ford et al., 2002; Gallop et al., 2006; Lee et al., Myr-Snf7 samples, but in larger, less mobile structures in the 2005; Rossman et al., 2010). h0-NBAR-Snf7 and Snf7F6E samples, consistent with the accu- Any model to describe ESCRT-III-mediated scission needs to mulation of endosomal intermediates (Figure S7). Thus, the account for both the structural features of the ESCRT-III polymer N-terminal interaction with membrane is important for fast cargo and the properties of the various scission events it mediates. In sorting. other words, the ESCRT-III filament must have the ability to Our mutational analysis has provided key insights into the role mediate diverse scission events, from the generation of small in- of the endogenous Snf7 ANCHR motif by identifying three func- traluminal vesicles at the endosome and HIV viral particles at the tional categories of N-terminal mutants (Figure 7C). Mutants that plasma membrane, to the much larger-scale abscission during are unable to insert into membrane, for example, Snf7F6E or DN- cytokinesis. One key feature that allows ESCRT-III to mediate Snf7, cannot provide the anchoring function necessary to allow such diverse events is the regulated assembly of the ESCRT-III the ESCRT-III polymer to exert the required forces to stabilize polymer by key upstream factors, such as ESCRT-I and –II, the invagination and are therefore functionally defective (Fig- in ILV formation or the HIV Gag protein in viral budding. This ure 7D). Mutants that are able to provide a membrane anchor regulated assembly likely accounts for one remarkable feature without promoting a high degree of positive curvature, such as of ILV’s, their consistently regular size; however, future experi- Myr-Snf7 or ALPS-Snf7, are able to efficiently rescue cargo sort- ments must investigate this further. ing to near wild-type levels. In contrast, the ability to rescue Another key feature shared among these scission events re- cargo sorting decreases as the ability to deform membrane quires that ESCRT-III binds tightly to and constricts membrane, increases. This is likely due to disrupting the intricate balance acting as a membrane scaffold. This requires maintaining mem- of positive and negative curvatures associated with forming a brane binding under conditions of increased force due to mem- nascent vesicle. brane tension. The tightening spiral must therefore be anchored

210 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

A Kinetic Sorting Delay of N-terminal Snf7 B Slope Comparison of Mutants 100% Snf7 Mutants WT 100.0% F6E m = -0.65% 80% h0-NBAR 80.0% MYR

60% 60.0% m = -1.91% m = -1.03%

40% 40.0% y = -1.79%

20.0% % Initial Fluorescence 20%

0% 0.0% 840 0 20 40 60 80 100 Minutes Minutes Post Methionine Additon C Deforming Potential

G2A Myr-Snf7 Myr-Snf7 h0-NBAR-Snf7 10WK-Snf7 Snf7F6E ∆N-Snf7 Snf7 WT ALPS-Snf7 14WK-Snf7 6K-Snf7 Snf7W2E Arf1-Snf7G2A 6WK-Snf7 Arf1-Snf7 Snf7F6W Impaired Binding MVB Formation Unbalanced Curvature

Snf7 N-terminal insertion anchors ESCRT-III ionic interactions to membrane

Deformation stabilized by ESCRT-III polymer

E ESCRT-III: Polymer assembly Elongation Disassembly Membrane: Invagination Vesicle formation Scission

WT Snf7 E-II E-II

Vps4 ILV

∆N-Snf7 E-II E-II No ILV X Formation

Figure 7. The N Terminus of Snf7 Is Required for Fast Cargo Sorting (A) Kinetic sorting of mid-log cultures of snf7D yeast exogenously expressing WT Snf7, Snf7F6E, h0-NBAR-Snf7, and Myr-Snf7. (B) Plot depicting slope calculated from linear trendlines for samples in (A), 8–40 min poststimulation. Slope of trend line is shown. (C) Schematic depicting relationship between quantitative sorting efﬁciency of Snf7 N-terminal mutants and the predicted ability to deform membrane. (D) Cartoon schematic showing stabilization of invagination by ESCRT-III polymer anchored to membrane by ANCHR motif. (E) Model showing ILV formation is dependent upon N-terminal anchoring of ESCRT-III polymer to membrane. See also Figure S7.

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 211 Developmental Cell ANCHRing ESCRT-III to Membrane

to membrane and thus requires a high-affinity membrane an- ESCRT-mediated scission event? As the polymer elongates, chor. As the polymer elongates and the number of protein-mem- the counterforce due to the increasing resistance from the de- brane contacts increase, the affinity of the polymer to membrane forming membrane and the tension from the insertion of the is amplified, allowing the polymer to provide the mechanical ANCHR motif into the bud neck may restrict the elongating helix force necessary to act as a scaffold and stabilize the invagina- into a tightening 3D spiral that promotes scission at its narrow tion. Under conditions in which Snf7’s ANCHR motif is compro- end. Alternatively, the constriction of the ESCRT-III polymer mised (Snf7F6E or DN-Snf7), the ESCRT-III polymer is unable to may be coupled to the binding and disassembly reaction driven drive membrane invagination and efficiently progress to vesicle by the AAA-ATPase Vps4. The ANCHR motif of Snf7 forms a tight formation (Figure 7E). association between the ESCRT-III polymer and the membrane Our study of Snf7’s ANCHR motif addresses the balance of prior to vesicle scission and removal of the membrane-anchored curvatures associated with ILV formation. The rim of the invagi- ESCRT-III scaffold may destabilize or buckle the membrane, nation is positively curved (Figure 6A) and may initially be gener- triggering fission. Altering the properties of the ANCHR motif ated by upstream ESCRT machinery and clustered cargo (Boura may also affect the ability of Vps4 to disassemble the ESCRT- et al., 2012; Wollert and Hurley, 2010). Snf7’s ANCHR motif may III scaffold, delaying the kinetics of ILV formation. act as both a sensor of this positive curvature and a curvature The ESCRTs have emerged as cargo recognition, cargo stabilizer. Studies have shown that Snf7 localizes to both invag- sequestering, and membrane remodeling machinery, with a inations and along highly curved membranes (Fyfe et al., 2011; particular emphasis on a role for the ESCRT-III polymer in Henne et al., 2012; Wollert and Hurley, 2010). An invagination membrane deformation. Of the six distinct stages of ESCRT-III also contains negative curvature around its circumference. The activity, this study provides important insights into ESCRT- ESCRT-III polymer acts as a circular scaffold that binds the dependent membrane bending and vesicle formation. Additional membrane and stabilizes this invagination (inward deformation studies and new assays will be required to further analyze the and cargo-laden nascent vesicle). We observed that other neck narrowing and scission steps, including a precise role for well-characterized membrane-inserting wedge domains known the Vps4 ATPase in these reactions. to promote or sense positive curvature (h0-NBAR-Snf7 and ALPS-Snf7) rescue the DN-Snf7-sorting defect to varying de- EXPERIMENTAL PROCEDURES grees. Interestingly, Arf1-Snf7 did not rescue cargo sorting, likely due to its strong membrane deformation and scission-promoting Yeast Strains, Plasmids, and Protein Purification properties (Beck et al., 2011), which cause it to disrupt the intri- Strains and plasmids from this study are listed in the Supplemental Experi- mental Procedures. Snf7core and full-length constructs were cloned using the cate balance of curvatures associated with ILV formation. This is pET23d bacterial expression vector (Novagen) with an added N-terminal consistent with the reduced functionality of the synthetic 10WK- His6-tag. Vps2 core domain constructs were cloned into pGEX6P1. Constructs Snf7 chimera, which also has the ability to promote a high degree were expressed in BL21 or C41(DE3) Escherichia coli cells and purified by either of membrane deformation (Figure 6D). Furthermore, the Arf1- TALON beads (Clontech) or GSH-Sepharose beads (GE Healthcare). All expres- Snf7 chimera was able to rescue cargo sorting when its mem- sion preps were grown at 37 C until a cell OD600 of 0.6–0.8 and were then brane affinity was reduced by preventing its myristoylation induced with 1 mM IPTG and grown for 4 hr at 37 C. During purification, proteins were reconstituted in 500 mM NaCl and 20 mM HEPES (pH 7.4). TALON beads (Figure 5C). These results demonstrate that the properties of were eluted in 150 mM NaCl, 20 mM HEPES (pH 7.4), and 400 mM imidazole; different amphipathic helixes have been tuned to their specific GSH-Sepharose beads were eluted in 20 mM glutathione (Sigma-Aldrich). roles in curvature sensing and anchoring (ALPS motif, Snf7’s Samples were dialyzed overnight to remove excess imidazole or glutathione. ANCHR motif), membrane bending (h0 on N-BAR proteins), or vesicle scission (Arf1). We have exploited these helices to pro- Fluorescence Microscopy vide insights into the delicate balance of competing curvatures All fluorescent microcopy was performed on mid-log yeast. Images were that exist during ILV morphogenesis between the ESCRT-III acquired using a DeltaVision RT system (Applied Precision) with Photometrics CoolSNAP HQ Camera. Postacquisition deconvolution and image analysis polymer and the insertion events. were performed in Softworx and Photoshop (Adobe). How does the growing ESCRT-III polymer deform membrane as it assembles? We propose that the ESCRT-III polymer initially Mup1-pHluorin Quantitative Sorting Assay forms a ring on the surface of the endosome, (restricting the For steady-state analysis, mid-log cultures grown in 20 mg/ml methionine for ANCHR motifs to the circular rim of the membrane invagination), 2 hr were transferred to PBS. Mean fluorescence of 100,000 events was re- surrounding the cargo clustered by the upstream ESCRT corded on a BD Accuri C6 flow cytometer. After background subtraction, machinery. The elongating ESCRT-III filament assembles into a the WT sample was set to 100%, the vector control to 0%, and the mutant sorting efficiency was calculated based on the mean fluorescence. Each sample is three-dimensional (3D) helical spiral (corkscrew), which is represented by the mean of at least three independent readings of 100,000 anchored to membrane by the ANCHR motif, further invaginating cells. Error bars represent standard deviation. the membrane domain contained within the ESCRT-III ring. In For kinetic analysis, events were recorded for 30 s intervals every 2 min this model, the tight membrane binding of the growing filament immediately after methionine addition. The mean fluorescence of each time may keep the transmembrane protein cargo concentrated in point was plotted after background subtraction. Each time point is the average this membrane domain, ahead of the growing, narrowing fila- of three independent kinetic experiments plotted with SD. Slopes were calcu- ment, ultimately forming the vesicle at the front, leading edge, lated from the linear trendline from time points 8 to 40. of the ESCRT-III spiral. Upon scission at the leading edge, and Western Blot and Antibodies narrowest point of the filament, the concentrated cargo would Total cell lysates were prepared from 5 OD600 mid-log cultures by incubating be packaged into the nascent ILV, excluding the ESCRT-III fila- on ice 1 hr in 10% TCA. Following two acetone wash-sonication cycles, sam- ment from the nascent vesicle (Figure 7D). What promotes the ples were bead-beated 5 min in boiling buffer (50 mM Tris [pH 7.5], 1 mM

212 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. Developmental Cell ANCHRing ESCRT-III to Membrane

EDTA, and 1% SDS) and incubated 5 min at 95C. After addition of 23 urea Research Fellowship. The authors would like to particularly thank Sunil Adige, buffer (150 mM Tris [pH 6.8], 6 M urea, 6% SDS, and bromophenol blue), sam- Jeanne Quirit, William Valley, and Ning Liu for their scientiﬁc support. The au- ples were bead-beated for 5 min and heated 10 min at 65C. Samples were run thors also thank Chris Stefan, Yuxin Mao, Chris Fromme, and Leonid Timashev on 10% polyacrylamide gels and transferred to nitrocellulose membranes. for helpful discussions and their critical editing of the manuscript. Antibodies used for blotting were G6PDH and FLAG (Sigma-Aldrich), Snf7 and Vps24 (Babst et al., 1998), Vps20 (generated in-house as previously Received: June 20, 2013 described [Babst et al., 1998]), and Vps2 (a gift from David Teis). Revised: August 9, 2013 Accepted: September 12, 2013 Cellular Fractionation Analysis Published: October 17, 2013

About 15–20 OD600 of mid-log cultures were spheroplasted using 10 mg zymolyase/OD cells. Yeast were lysed using Dounce homogenizer on ice in 1 ml lysis REFERENCES buffer (50 mM Tris [pH 7.5], 1 mM EDTA, 200 mM sorbitol with protease inhib- itors). Lysates were cleared by spinning at 500 3 g for 5 min at 4 C before Antonny, B., Beraud-Dufour, S., Chardin, P., and Chabre, M. (1997). centrifugation at 13,000 3 g for 10 min at 4 C. The pellet was resuspended N-terminal hydrophobic residues of the G-protein ADP-ribosylation factor-1 in lysis buffer, and proteins were TCA precipitated from both the supernatant insert into membrane phospholipids upon GDP to GTP exchange. and pellet fractions as described under western blot section. Biochemistry 36, 4675–4684. Babst, M., Sato, T.K., Banta, L.M., and Emr, S.D. (1997). Endosomal transport Liposome Sedimentation Assays function in yeast requires a novel AAA-type ATPase, Vps4p. EMBO J. 16, To conduct the experiment, 14 mM protein and 600 mM liposomes were incu- 1820–1831. bated for 10 min and directly centrifuged in a TLA-100 (Beckman Coulter) for 10 min at 55,000 rpm at 20C. Liposomes were made in 20 mM HEPES Babst, M., Wendland, B., Estepa, E.J., and Emr, S.D. (1998). The Vps4p AAA (pH 7.4) and 150 mM NaCl by classical dehydration and sonication for ATPase regulates membrane association of a Vps protein complex required 17 1 min. The liposome composition was 60% DOPC:30% phosphatidylser- for normal endosome function. EMBO J. , 2982–2993. ine:10% phosphatidylinositol 3-phosphate. Variations of the composition Babst, M., Katzmann, D.J., Estepa-Sabal, E.J., Meerloo, T., and Emr, S.D. were 100% DOPC or 25% DOPC:75% PS. All lipids were purchased from (2002a). Escrt-III: an endosome-associated heterooligomeric protein complex Avanti or CellSignals. Where indicated, liposomes were extruded through required for mvb sorting. Dev. Cell 3, 271–282. 800, 100, or 30 nm filters. The supernatant (unbound) and pellet (bound) Babst, M., Katzmann, D.J., Snyder, W.B., Wendland, B., and Emr, S.D. were directly prepared for SDS-PAGE and visualized with Coomassie stain. (2002b). Endosome-associated complex, ESCRT-II, recruits transport Value is average of three independent sedimentation assays. machinery for protein sorting at the multivesicular body. Dev. Cell 3, 283–289. Beck, R., Prinz, S., Diestelko¨ tter-Bachert, P., Ro¨ hling, S., Adolf, F., Hoehner, Fluorescence Spectroscopy K., Welsch, S., Ronchi, P., Bru¨ gger, B., Briggs, J.A., and Wieland, F. (2011). Cysteine residues on purified Snf7core constructs were labeled with excess Coatomer and dimeric ADP ribosylation factor 1 promote distinct steps in IANBD (N,N’-dimethyl-N’-(iodoacetyl)-N’-(7-nitrobenz-2-oxa-1,3-diazole-4- membrane scission. J. Cell Biol. 194, 765–777. yl)ethylenediamine) (Molecular Probes) for 2 hr at 4C protected from light. Samples were double dialyzed into buffer containing 20 mM HEPES (pH Bigay, J., Casella, J.F., Drin, G., Mesmin, B., and Antonny, B. (2005). ArfGAP1 7.4), 150 mM NaCl, and 0.1% b-mercaptoethanol to quench and remove responds to membrane curvature through the folding of a lipid packing sensor 24 excess IANBD. All reactions were performed in 150 ml final volume in quartz motif. EMBO J. , 2244–2253. cuvettes. Liposomes were incubated in excess (250 mM) with protein for Bodon, G., Chassefeyre, R., Pernet-Gallay, K., Martinelli, N., Effantin, G., 10 min. The excitation wavelength was 468 nm and the emission was scanned Hulsik, D.L., Belly, A., Goldberg, Y., Chatellard-Causse, C., Blot, B., et al. at 500–590 nm. (2011). Charged multivesicular body protein 2B (CHMP2B) of the endosomal sorting complex required for transport-III (ESCRT-III) polymerizes into helical Electron Microscopy structures deforming the plasma membrane. J. Biol. Chem. 286, 40276– Visualization of Snf7 polymers on EM grids with or without liposomes was per- 40286. formed as previously described (Henne et al., 2012). For Snf7 liposome Boucrot, E., Pick, A., Çamdere, G., Liska, N., Evergren, E., McMahon, H.T., R52E assays, 15 mM Snf7 was mixed with 600 mM liposomes for 10 min and and Kozlov, M.M. (2012). Membrane fission is promoted by insertion of amphi- was added to EM grid. For tubulation experiments, 10 mM protein was added pathic helices and is restricted by crescent BAR domains. Cell 149, 124–136. to 600 mM liposomes and incubated for 15 min. Samples were negative stained Boura, E., Ro´ zycki,_ B., Chung, H.S., Herrick, D.Z., Canagarajah, B., Cafiso, with ammonium molybdate. D.S., Eaton, W.A., Hummer, G., and Hurley, J.H. (2012). Solution structure of Yeast cells were fixed with 2.5% (v/v) glutaraldehyde for 1 hr and sphero- the ESCRT-I and -II supercomplex: implications for membrane budding and plasted with 250 mg zymolyase and 50 ml gluculase before embedding in 2% scission. Structure 20, 874–886. ultra-low temperature agarose. Cells were incubated in 1% osmium tetroxide/1% potassium ferrocyanide (30 min), 1% thiocarbohydrazide (5 min), Carlton, J.G., and Martin-Serrano, J. (2007). Parallels between cytokinesis and 316 and 1% osmium tetroxide/1% postassium ferrocyanide (5 min). Cells were retroviral budding: a role for the ESCRT machinery. Science , 1908–1912. extensively washed in water after each treatment. After dehydration through Drin, G., and Antonny, B. (2010). Amphipathic helices and membrane curva- an ethanol series, samples were transitioned into 100% propylene oxide and ture. FEBS Lett. 584, 1840–1847. embedded in Spurr’s resin. All electron microscopy was performed on a Morg- Drin, G., Casella, J.F., Gautier, R., Boehmer, T., Schwartz, T.U., and Antonny, nani 268 transmission electron microscope (FEI) with an AMT digital camera. B. (2007). A general amphipathic a-helical motif for sensing membrane curvature. Nat. Struct. Mol. Biol. 14, 138–146. SUPPLEMENTAL INFORMATION Eswar, N., Eramian, D., Webb, B., Shen, M.Y., and Sali, A. (2008). Protein structure modeling with MODELLER. Methods Mol. Biol. 426, 145–159. Supplemental Information includes Supplemental Experimental Procedures Farsad, K., and De Camilli, P. (2003). Mechanisms of membrane deformation. and seven figures and can be found with this article online at http://dx.doi. Curr. Opin. Cell Biol. 15, 372–381. org/10.1016/j.devcel.2013.09.009. Ford, M.G., Mills, I.G., Peter, B.J., Vallis, Y., Praefcke, G.J., Evans, P.R., and ACKNOWLEDGMENTS McMahon, H.T. (2002). Curvature of clathrin-coated pits driven by epsin. Nature 419, 361–366. N.J.B. is supported by an American Cancer Society Postdoctoral Fellowship Fyfe, I., Schuh, A.L., Edwardson, J.M., and Audhya, A. (2011). Association (PF-12-062-01-DMC). W.M.H. is supported by a Sam and Nancy Fleming of the endosomal sorting complex ESCRT-II with the Vps20 subunit of

Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc. 213 Developmental Cell ANCHRing ESCRT-III to Membrane

ESCRT-III generates a curvature-sensitive complex capable of nucleating Lata, S., Schoehn, G., Jain, A., Pires, R., Piehler, J., Gottlinger, H.G., and ESCRT-III filaments. J. Biol. Chem. 286, 34262–34270. Weissenhorn, W. (2008). Helical structures of ESCRT-III are disassembled by VPS4. Science 321, 1354–1357. Gallop, J.L., Jao, C.C., Kent, H.M., Butler, P.J., Evans, P.R., Langen, R., and McMahon, H.T. (2006). Mechanism of endophilin N-BAR domain-mediated Lee, M.C., Orci, L., Hamamoto, S., Futai, E., Ravazzola, M., and Schekman, R. membrane curvature. EMBO J. 25, 2898–2910. (2005). Sar1p N-terminal helix initiates membrane curvature and completes the fission of a COPII vesicle. Cell 122, 605–617. Garrus, J.E., von Schwedler, U.K., Pornillos, O.W., Morham, S.G., Zavitz, K.H., Wang, H.E., Wettstein, D.A., Stray, K.M., Coˆ te´ , M., Rich, R.L., et al. (2001). Mim, C., Cui, H., Gawronski-Salerno, J.A., Frost, A., Lyman, E., Voth, G.A., and Tsg101 and the vacuolar protein sorting pathway are essential for HIV-1 Unger, V.M. (2012). Structural basis of membrane bending by the N-BAR 149 budding. Cell 107, 55–65. protein endophilin. Cell , 137–145. Muzio1, T., Pineda-Molina, E., Ravelli, R.B., Zamborlini, A., Usami, Y., Ghazi-Tabatabai, S., Saksena, S., Short, J.M., Pobbati, A.V., Veprintsev, D.B., Go¨ ttlinger, H., and Weissenhorn, W. (2006). Structural basis for budding by Crowther, R.A., Emr, S.D., Egelman, E.H., and Williams, R.L. (2008). Structure the ESCRT-III factor CHMP3. Dev. Cell 10, 821–830. and disassembly of filaments formed by the ESCRT-III subunit Vps24. Structure 16, 1345–1356. Rossman, J.S., Jing, X., Leser, G.P., and Lamb, R.A. (2010). Influenza virus M2 protein mediates ESCRT-independent membrane scission. Cell 142, 902–913. Hanson, P.I., Roth, R., Lin, Y., and Heuser, J.E. (2008). Plasma membrane Saksena, S., Wahlman, J., Teis, D., Johnson, A.E., and Emr, S.D. (2009). deformation by circular arrays of ESCRT-III protein filaments. J. Cell Biol. Functional reconstitution of ESCRT-III assembly and disassembly. Cell 136, 180, 389–402. 97–109. Henne, W.M., Buchkovich, N.J., Zhao, Y., and Emr, S.D. (2012). The endoso- Spitzer, C., Schellmann, S., Sabovljevic, A., Shahriari, M., Keshavaiah, C., mal sorting complex ESCRT-II mediates the assembly and architecture of Bechtold, N., Herzog, M., Mu¨ ller, S., Hanisch, F.G., and Hu¨ lskamp, M. 151 ESCRT-III helices. Cell , 356–371. (2006). The Arabidopsis elch mutant reveals functions of an ESCRT compo- Johnson, A.E. (2005). Fluorescence approaches for determining protein con- nent in cytokinesis. Development 133, 4679–4689. formations, interactions and mechanisms at membranes. Traffic 6, 1078– Teis, D., Saksena, S., and Emr, S.D. (2008). Ordered assembly of the ESCRT-III 1092. complex on endosomes is required to sequester cargo during MVB formation. Katzmann, D.J., Babst, M., and Emr, S.D. (2001). Ubiquitin-dependent sorting Dev. Cell 15, 578–589. into the multivesicular body pathway requires the function of a conserved Wollert, T., and Hurley, J.H. (2010). Molecular mechanism of multivesicular endosomal protein sorting complex, ESCRT-I. Cell 106, 145–155. body biogenesis by ESCRT complexes. Nature 464, 864–869.

214 Developmental Cell 27, 201–214, October 28, 2013 ª2013 Elsevier Inc.