(A) Structural model for yeast ESCRT-I docked to an endosomal membrane. The GLUE domain (PDB code 2CAY)(Teo et al., 2006) and NZF1 domain of ESCRT-II (PDB code 2J9U) (Gill et al., 2007) are cyan, ubiquitinated Cps1 is red, and ESCRT-I subunits are colored as in Fig. 2-4. (B) Schematic diagram of the docked model, incorporating simplified models of the interacting Vps27/Hse1 and ESCRT-II complexes.

(A, B, D) Close-ups of the interface of Mvb12 (purple ribbon and stick model) with Vps23, Vps28 and Vps37 (orange, blue, and green surfaces, respectively). (C) Shows locations of the regions of the ESCRT-I heterotetramer highlighted in panels A, B and D. Mvb12 residues are labeled in white, while significant Vps23 residues are labeled in black and Vps37 residues in red. (E) Surface depiction of the ESCRT-I heterotetramer showing in dark blue residues highly conserved in orthologs of Vps23, Vps28 and Vps37. Mvb12 is depicted as a ribbon. The conservation of the Mvb12 interaction surface on the rest of the stalk suggests that there is a structural counterpart of Mvb12 in non-fungal species.

(A) ESCRT-I has a weak intrinsic ability to bind to synthetic PI(3)P-containing liposomes that depends on the N-terminal helix of Vps37. ESCRT-II binds strongly to these liposomes and recruits ESCRT-I in vitro independent of the Vps37 N-terminal helix. (B). ESCRT-I sediments with brain liposomes, indicating the interaction with the N-terminal helix does not require PI(3)P. ESCRT-I Vps37 Δ1-21 does not bind to brain liposomes, showing that the stalk and headpiece do not bind lipids. (C) The triple ESCRT-I complex with the Vps37 Δ1-21 deletion (EI₃) is compared to intact ESCRT-I for binding to synthetic PI(3)P-containing liposomes. Both complexes are recruited to the same extent by ESCRT-II. (D) Schematic of mini-Vps27 and Ub-Cps1(8-17)C constructs. (E) Sedimentation with ubiquitin-Cps1 linker-conjugated synthetic PI(3)P-containing liposomes, schematized in (F). ESCRT-II and mini-Vps27 bind to these liposomes. ESCRT-I and triple ESCRT-I do not bind in the absence of other complexes, but bind strongly in the presence of ESCRT-II and more weakly in the presence of mini-Vps27.

(A) Electron density from the density-modified experimental map (blue) contoured at 1.0 σ and from a Se anomalous difference Fourier contoured at 4.0 σ (yellow) overlaid on the refined structure in the region of the helix bundle at the distal end of the stalk. (B) Structure of ESCRT-I, Vps23, orange, Vps28, blue, Vps37, green, Mvb12, purple. (C) Structure of the headpiece in the heterotetramer, shown with (D) the previously determined trimeric core structure in the same orientation. The two structures are overlaid in (E) with the heterotetramer headpiece colored as in (C) and the trimer colored cyan. (F) The headpiece contains a small β-sheet near its junction with the stalk.
(G) The triple coiled coil is continued and stabilized by an unusual hybrid between a two-stranded coiled-coil and an extended region of Mvb12. (H) The helix bundle at the base of the stalk. Structural figures were generated with Pymol (www.pymol.org).

(A) Domain structure of full-length ESCRT-I and other constructs used in this study, and their Stokes radii R_H as determined by analytical ultracentrifugation. (B) Gel filtration of full-length ESCRT-I, monitored at 280 nm. The inset shows a Coomassie blue-stained SDS-PAGE gel of the peak fraction indicated. (C) Sedimentation equilibrium profiles of full-length ESCRT-I plotted as a distribution of A₂₈₀ vs. r at equilibrium. Data were collected at 6 (blue), 8 (green), 10 (purple) and 12 (red) krpm at a loading A₂₈₀ of 0.38 (alternate data points are shown). The solid lines show the best-fit global analysis in terms of a single ideal solute, with the corresponding residuals shown in the panels above the plot. (D) Solution structural model of intact ESCRT-I. The curved arrows indicate that the Vps23 UEV and Vps28 C-terminal domain (CTD) are conformationally dynamic.

(A) Ste3 and Sna3 sorting defects in stalk mutants. Plasmids expressing wild type or mutant forms of VPS23, MVB12 or VPS37 were introduced into vps23Δ, mvb12Δ or vps37Δ strains expressing plasmid-encoded Ste3-GFP or Sna3-GFP, as indicated. (B) Mvb12 localization depends on the integrity of the stalk. Wild type or mutant forms of VPS23 or VPS37 were expressed from plasmids in vps23Δ vps4Δ or vps37 vps4Δ strains containing chromosomally integrated Mvb12-GFP, whereas plasmid-encoded wild type and mutant forms of Mvb12-YFP were expressed in mvb12Δ vps4Δ strains. (C) Stalk mutants do not prevent Vps23 localization to the MVB. Wild type or mutant forms of VPS37 or MVB12 were expressed from plasmids in vps37Δ, mvb12Δ, or mvb12Δ vps4Δ strains containing chromosomally integrated Vps23-GFP. (D-F) Stalk mutations have differential effects on ESCRT-I assembly. (D) Detergent extracts from vps23Δ mutants containing chromosomally tagged Mvb12-GFP (upper panels) or Vps37-13myc (lower panels) and plasmids for the expression of HA-tagged wild type or mutant forms of Vps23 were subjected to immunoprecipitation with anti-HA antiserum (E) mvb12Δ mutants containing chromosomally tagged Vps23-6HA (upper panels) or Vps37-13myc (lower panels) and plasmid-borne wild type or mutant forms of Mvb12-YFP were immunoprecipitated with either anti-HA (upper panels) or anti-myc (lower panels) antiserum. (F) vps37Δ mutants containing chromosomally tagged Mvb12-GFP (upper panels) or Vps23-6HA (lower panels) and plasmids for the expression of wild type or mutant forms of Vps37-myc were immunoprecipitated with anti-myc antiserum. Co-precipitating proteins were resolved by SDS-PAGE and analyzed by western blotting with mAbs to HA, myc or GFP as indicated. Immunoprecipitates from cells expressing wild type VPS23 (D) or VPS37 (F) were loaded at 5X (D) or 2X (F) reduced levels relative to immunoprecipitates from cells expressing empty vector or mutant proteins to compare relative levels of co-purifying proteins despite differences in the stability of the mutant complexes. Bar = 2 μM.

(A) mvb12 mutants exhibit a weak CPY sorting defect. CPY missorting from wild type, vps4Δ or mvb12Δ strains with and without a MVB12-expressing plasmid (+pMVB12) was detected by colony overlay assay (Conibear and Stevens, 2002). (B) Mvb12 is recruited to the endosome in vps4Δ mutants. vps27Δ vps4Δ and vps23Δ vps4Δ strains containing integrated Mvb12-GFP and plasmid-expressed VPS27 and VPS23, or empty vector, were incubated with FM4-64 and viewed by double-label fluorescence microscopy. (C) Fluorescence microscopy of wild type and mvb12Δ strains containing Ste3-GFP, Sna3-GFP, or ALP-GFP plasmids together with a complementing MVB12 plasmid or empty vector. (D) Mvb12 recruitment does not require ESCRT-II or -III. The chromosomal copy of MVB12 was tagged with GFP in wild type, vps27Δ, vps23Δ, vps37Δ, vps28Δ, vps22Δ and vps20Δ strains and visualized in live cells. (E) Mvb12 is associated with ESCRT-I. Detergent extracts prepared from 20 OD₆₀₀ units (lanes 1–7) or 100 OD₆₀₀ (lanes 8–10) of wild type or mutant strains expressing GFP-tagged Mvb12 and/or HA-tagged Vps23 were immunoprecipitated with anti-HA antiserum and analyzed by western blotting with anti-GFP and anti-HA mAbs. Loading of lanes 9–10 was 7.5X greater than lane 8 to compare relative levels of co-purifying Mvb12-GFP despite differences in Vps23 stability. Bar = 2 μM.