Eap45 in mammalian ESCRT-II binds ubiquitin via a phosphoinositide-interacting GLUE domain.

Ubiquitination serves as a key sorting signal in the lysosomal degradation of endocytosed receptors through the ability of ubiquitinated membrane proteins to be recognized and sorted by ubiquitin-binding proteins along the endocytic route. The ESCRT-II complex in yeast contains one such protein, Vps36, which harbors a ubiquitin-binding NZF domain and is required for vacuolar sorting of ubiquitinated membrane proteins. Surprisingly, the presumptive mammalian ortholog Eap45 lacks the ubiquitin-binding module of Vps36, and it is thus not clear whether mammalian ESCRT-II functions to bind ubiquitinated cargo. In this paper, we provide evidence that Eap45 contains a novel ubiquitin-binding domain, GLUE (GRAM-like ubiquitin-binding in Eap45), which binds ubiquitin with similar affinity and specificity as other ubiquitin-binding domains. The GLUE domain shares similarities in its primary and predicted secondary structures to phosphoinositide-binding GRAM and PH domains. Accordingly, we find that Eap45 binds to a subset of 3-phosphoinositides, suggesting that ubiquitin recognition could be coordinated with phosphoinositide binding. Furthermore, we show that Eap45 colocalizes with ubiquitinated proteins on late endosomes. These results are consistent with a role for Eap45 in endosomal sorting of ubiquitinated cargo.

Ubiquitination of proteins was until recently mainly considered to be important for targeting proteins for degradation in the 26S proteasome (1,2). The recent discovery that ubiquitination may work as a reversible modification has, however, shown that the attachment of ubiquitin is also essential for changes in protein location, activity, and interaction with binding partners. Ubiquitination is thus emerging as a very important mechanism for regulating several cellular processes, such as signal transduction, membrane trafficking, transcriptional regulation, virus budding, and DNA repair (3). Mono-and multi-ubiquitination have been reported to regulate the internalization of several membrane receptors, such as receptor tyrosine kinases and G-protein-coupled receptors, from the plasma membrane into clathrin-coated pits (4,5). In addition, ubiquitination is required for sorting endocytosed membrane proteins destined for degradation into intraluminal vesicles of multivesicular bodies (MVBs). 1 The MVBs eventually fuse with lysosomes to degrade the internal vesicles and their protein cargo (6,7).
Many of the subcomponents of the molecular machinery responsible for lysosomal sorting have now been identified. Based on successful yeast screens for proteins necessary for receptors to be transported to the vacuole (the yeast equivalent of mammalian lysosomes), several vacuolar protein sorting (Vps) proteins have been identified and shown to participate in the sorting event (18 -20). Ten of the Vps proteins have been found to exist in three distinct endosomal sorting complexes required for transport, ESCRT-I, ESCRT-II, and ESCRT-III, which appear to be recruited sequentially to the endosomes during MVB formation. Recent studies have revealed that the functions of the mammalian Vps counterparts in lysosomal degradation are conserved (6). Interestingly, these complexes are also required for budding of enveloped RNA viruses (21).
Upon delivery of ubiquitinated membrane proteins to endosomes, the Vps protein Hrs/Vps27 uses its UIM domain to recruit these proteins into microdomains of sorting endosomes (8,9,22). Several studies in yeast and mammalian cells have shown that Hrs/Vps27 is required for the recruitment of Tsg101/Vps23 in the ESCRT-I complex (23,24). Tsg101 contains a UEV domain that has been suggested to mediate delivery of cargo from the Hrs complex to subsequent ESCRT-II and -III complexes. At present, it is not clear how ESCRT-II (Vps25, Vps36, and Vps22) is recruited to MVBs. Studies in yeast have suggested that ESCRT-II is essential for MVB formation by recruiting the ESCRT-III complex (20). In addition, the yeast Vps36 subunit contains an NZF domain that binds ubiquitin and may therefore participate in cargo transfer into MVBs (16,17). ESCRT-III consists of two subcomplexes that are suggested to be involved in the final formation of the MVBs and dissociation of the ESCRT complexes from the endosomal membranes (20,25).
It has been widely assumed that the transfer of ubiquinated cargos into MVBs is dependent on the sequential recognition of complexes with ubiquitin binding properties, such as Hrs/ Vps27, ESCRT-I, and ESCRT-II. However, the fact that the ubiquitin-binding domain of yeast Vps36 in ESCRT-II has not been conserved in its mammalian counterpart remains as a problem for this model. The aim of this study was therefore to characterize the possible ubiquitin-binding properties of the putative mammalian Vps36 homolog, Eap45.  Fig. 3 was initially built with Clustal_X but manually edited to accommodate regions in the fungal sequences that have large insertions. This manual alignment was guided by superimposing the secondary structure elements of the myotubularin-related protein-2 (MTMR2) structure (30) with the output of the Jpred prediction.
In Vitro Transcription/Translation-pcDNA3-myc-Eap45 fulllength and deletion constructs were translated in the presence of [ 35 S]methionine (Amersham Biosciences) using the TNT® T7 coupled reticulocyte lysate system (Promega, Southampton, UK) according to the protocol from the manufacturer.
Ubiquitin Agarose and GST Pull-down Assays-Aliquots of protein-A-agarose (Sigma-Aldrich) or ubiquitin-agarose (Sigma-Aldrich) were washed three times with assay buffer (25 mM Hepes, pH 7.2, 125 mM potassium acetate, 2.5 mM magnesium acetate, 5 mM EGTA, and 1 mM dithiothreitol) before incubation with 35 S-labeled full-length Eap45 or Eap45 deletions containing 10% fetal calf serum, and 0.1% Igepal for 1 h at 4°C. Finally the beads were washed four times in assay buffer and resuspended in SDS-PAGE sample buffer. Eap45 bound to beads were detected on film or PhoshorImager screen (Amersham Biosciences). Lysates from transformed E. coli BL21 (DE3) cells containing GST, GST-ubiquitin, or GST-ubiquitin L8A/I44A were bound in Tris buffer (20 mM Tris, pH 7.5, 60 mM NaCl, and 0.01% Igepal) to aliquots of glutathione-Sepharose (Amersham Biosciences) at room temperature for 60 min. The beads were then washed with assay buffer (20 mM Hepes, pH 7, 140 mM NaCl, 2 mM CaCl 2, and 1 mM dithiothreitol), and in vitro translated Eap45 was added with assay buffer containing 0.1% Igepal. After rotation at 4°C for 60 min, the beads were washed three times with assay buffer and then analyzed by SDS-PAGE and fluorography.
Cell Culture and Transfection-HeLa cell cultures were maintained as recommended by American Type Culture Collection (Manassas, VA). For expression in mammalian cells, we used the Effectene system according to the manufacturer's instructions (Qiagen, Valencia, CA). Cells were analyzed 16 h after transfection.
Surface Plasmon Resonance-Surface plasmon biosensor binding experiments were performed at 25°C with the following buffer conditions: 10 mM Hepes, pH 7.4, 150 mM NaCl, 3 mM EDTA, and 0.005% Surfactant P20 (BIAcore, Uppsala, Sweden) using a BIAcore 2000 with CM5 research-grade sensor chip which captures anti-GST antibody (BIAcore). GST-Eap45-(1-139), GST-Eap45-(1-102), and GST alone (negative control) were captured at final densities of 1300 resonance units. Ubiquitin (Sigma-Aldrich) in a running buffer was flowed over the GST-Eap45 fragments and GST surfaces with a flow rate of 20 ml/min. The concentrations of the ubiquitin solutions were 58.7, 118, 235, 470, 705, 940, 1410, and 1880 M. A saturated response value on each concentration of ubiquitin was used as the equilibrium response value of each concentration. The equilibrium data were fit with a simple 1:1 interaction model, and the K D values were calculated with IGOR Pro 5 (WaveMetrics, Portland, OR).
Confocal Fluorescence Microscopy-Transfected HeLa cells grown on coverslips were permeabilized with 0.05% saponin, fixed with 3% paraformaldehyde, and stained for fluorescence microscopy as described previously (33). Coverslips were examined using an LSM 510 META microscope (Carl Zeiss, Jena, Germany) equipped with a Neo-Fluar 100ϫ/1.45 oil immersion objective. Image processing was done with Adobe Photoshop, Version 7.0 (Adobe Systems, Seattle, WA).
Protein-Lipid Overlay Assay-Nitrocellulose membranes containing spotted phosphoinositides and other lipids (phosphoinositide strips and phosphoinositide arrays) were purchased from Echelon Research Laboratories (Salt Lake City, UT). Membranes were blocked with 0.1% chicken egg white albumin (Sigma-Aldrich) in TBS-T (10 mM Tris-HCl, pH 8, 150 mM NaCl, 0.1% Tween 20) for 1 h at room temperature. Blocked membranes were incubated overnight at 4°C with 1 g ml Ϫ1 GST fusion proteins in TBS-T. The membranes were then washed four times for 10 min in TBS-T. After washing, membranes were incubated with monoclonal anti-GST antibody (1/5000 dilution; Sigma-Aldrich) for 2 h at room temperature followed by additional washing and incubation with horseradish peroxidase-conjugated antibody (1/15000; Jackson ImmunoResearch Laboratories). The antibodies were incubated in the presence of 0.1% of albumin. After final washing, lipid-bound GST fusion proteins were detected by chemiluminescence.

Mammalian Eap45
Binds Ubiquitin-Structural analyses of yeast Vps36 have shown that the N-terminal region contains an NZF-domain required for ubiquitin binding (17). However, the murine homolog Eap45 contains no NZF domain (Figs. 1A and 3), and it is not known whether this protein binds ubiquitin. To investigate whether the mammalian Vps36 ortholog has the ability to bind ubiquitin, we incubated in vitro translated 35 S-labeled Eap45 with ubiquitin covalently immobilized on agarose beads. SDS-PAGE analysis followed by fluorography showed that beads containing ubiquitin pulled down Eap45, whereas control beads linked with protein-A did not (Fig. 1B). This suggests that mammalian Eap45 shares the ability of yeast Vps36 to bind ubiquitin, even though it lacks the NZF domain.
To further delineate the region that was required for binding, we made various deletions of Eap45 (Fig. 1, A and B). By performing the same assay as described above for full-length Eap45, we found that the N-terminal region of Eap45-(1-154) was able to bind ubiquitin similarly to full-length Eap45. In contrast, Eap45-(140 -386) and other C-terminal deletions tested yielded ubiquitin interactions that background-level ubiquitin binding. Since a single NZF domain of yeast Vps36 is sufficient for binding to ubiquitin (17), we sought to further narrow down the ubiquitin-interacting domain within the Eap45 N-terminal region. We found that Eap45-(1-135) was able to bind ubiquitin, while Eap45-(1-102) and Eap45-(49 -154) were not (Fig. 1C). These results indicate that the 1-135 region of Eap45 mediates its binding to ubiquitin.
All previously identified ubiquitin-binding domains have been shown to interact with a hydrophobic patch of ubiquitin, the Leu-8, Ile-44, Val-70 interface (8,15,17,24). We therefore asked if Eap45 binding to ubiquitin is also mediated by the same hydrophobic interaction surface. To address this, we prepared GST fusion proteins of both wild-type ubiquitin and ubiquitin mutated in Leu-8 and Ile-44. These proteins, as well as GST alone, were immobilized on gluthatione-Sepharose beads and incubated in the presence of in vitro translated Eap45. In line with the results from the ubiquitin-agarose assay, we found that Eap45 binds wild-type ubiquitin. Importantly, we observed a strongly reduced interaction between Eap45 and the L8A/I44A ubiquitin mutant ( Fig. 2A). This suggests that the hydrophobic surface patch of ubiquitin interacts with Eap45 and underscores the requirement of this region in mediating the interaction with ubiquitin binding domains.
To measure the affinity of the binding and verify that the binding we observed of Eap45 to ubiquitin reflected a direct interaction, we performed surface plasmon resonance biosensor FIG. 1. Eap45 contains a novel ubiquitin-binding domain. A, comparison of the overall structure of Eap45 and yeast Vps36. The intensity of gray indicates the relative sequence similarity between Eap45 and yeast Vps36. The extreme Nterminal part is relatively conserved, but most of the N terminus has very low sequence identity. The Vps36 NZF domains are indicated, in which only one is functional (asterisk) (17). The parts of the Nterminal domain of yeast Vps36 and mouse Eap45 that map to the GLUE domain (see Fig. 3) are indicated above the structure with a black line. The various Eap45 deletions used in this study are shown. B, the N-terminal domain of Eap45 binds ubiquitin. Agarose-coated with ubiquitin (U) or protein-A (-) was incubated with in vitro translated, 35 Slabeled full-length Eap45 or deletion mutants for 1 h at 4°C. The beads were then washed and analyzed by SDS-PAGE and fluorography. The additional bands in some input lanes probably represent alternative translational initiations or degradation products. C, residues 1-135 of Eap45 bind ubiquitin. The experiment was performed as described for B, but various N-terminal deletion constructs were used instead. binding (BIAcore) experiments with purified proteins. GST fusion proteins of Eap45-(1-139), Eap45-(1-102), and GST alone were immobilized on BIAcore sensor chips, and various concentrations of ubiquitin were used in the flow over the GST protein surfaces to measure the affinity of the binding. Fig. 2B shows the titration curves of the Eap45 N-terminal fragments. The dissociation constant of Eap45-(1-139) for interacting with ubiquitin was 0.46 Ϯ 0.03 mM, which is slightly higher than that of the NZF domain of yeast Vps36 (0.18 mM) (17), but comparable with that of UIM (0.3 mM (34) and UEV (0.35-0.64 mM) domains (15). In contrast, the binding of ubiquitin to Eap45-(1-102) was barely detectable. These results show that Eap45 binds directly to ubiquitin with comparable affinity and specificity as other previously identified ubiquitin-binding domains (17).
The GLUE Domain, a Novel Ubiquitin-binding Domain-The structures of most ubiquitin binding domains identified so far have been solved. Surprisingly, these domains recognize ubiquitin by using very different folds. The UIM, CUE, and UBA domains use ␣-helical structural elements to interact with ubiquitin (10 -12, 35-37). In contrast, the NZF domain consists of four anti-parallel ␤-strands that are organized around a Zn(Cys) 4 cluster, in which exposed hydrophobic residues mediate binding to ubiquitin (16,17,38). Eap45 and the mammalian Vps36 homologs lack the NZF domains (17) as well as the other known ubiquitin-binding domains. To study whether the ubiquitin-binding region of Eap45 is conserved in proteins from other organisms, we used the sequence of the N-terminal ubiquitin-binding region of Eap45 (residues 1-135) in Blast and PSI-Blast searches (see "Experimental Proce-dures"). After two iterations of PSI-Blast, metazoan and fungal sequences appeared, including yeast Vps36 (E-value ϭ 0.003) with an alignment corresponding to residues 1-96 of Eap45. After three iterations of PSI-Blast, a new group of proteins appeared that does not contain the C-terminal region common to the Vps36 family. Inspection of some of these sequences (e.g. Anopheles EAA00415; GI: 55234884; E-value ϭ 0.006) revealed that they contain a GRAM domain (Ref. 39; Pfam: PF02893). In reciprocal searches using this GRAM domain as a query, members of the Vps36 family could similarly be detected. A Pfam search with Eap45 also detected a weak signal with the GRAM domain (E-value ϭ 0.035 for residues 38 -73). The sequence similarity between the Eap45 and Anopheles EAA00415 is, however, very low (only 17% identity and 37% similarity for residues 10 -122 of Eap45).
It has recently been shown that the GRAM domain of MTMR2 is part of a larger structure, which has the same fold as the PH domain (30). Using a combination of structureguided sequence alignments and secondary structure predictions (see "Experimental Procedures") we obtained an alignment of the N-terminal region of the Vps36 family (corresponding to region 1-135 of Eap45), which matches all the secondary structure elements of the MTMR2 GRAM/PH domain (Fig. 3). Strikingly, this region spans residues 1-285 of yeast Vps36 that includes the two NZF domains. It thus appears that all members of the Vps36 family contain a divergent GRAM/PH-like domain and that yeast and some other fungi have one or two NZF domains inserted in the GRAM/PH-like domain. To distinguish the Vps36 family GRAM/PH-like domain, we hereafter refer to it as the GLUE (GRAM-like ubiquitin-binding in Eap45) domain.
The GLUE Domain Binds 3-Phosphoinositides-GRAM domains are found in several membrane-associated proteins (39). One family of proteins that contain GRAM domains is the myotubularin family of 3-phosphatases. The function of these proteins in vesicular trafficking is dependent on the binding of their GRAM domains to specific phosphoinositides on endosomal membranes (30,40,41). Although there is evidence that ESCRT-II in yeast is recruited to endosomal membranes in the absence of other ESCRT complexes (19,20), previous studies have not revealed how ESCRT-II is associated with these structures. Our finding that both fungal and metazoan Vps36 proteins contain the GRAM-domain-like GLUE domain suggests that Vps36 might participate in membrane association. To investigate whether the GLUE domain binds membrane lipids, similar to GRAM domains and the related PH domains, we purified GST-Eap45-(1-154) protein and measured its binding to various lipids immobilized on nitrocellulose. As shown in Fig. 4A, we found that GST-Eap45-(1-154) bound specifically to PtdIns(3,4,5)P 3, and much weaker to PtdIns(3,4)P 2 and PtdIns(3,5)P 3, (the latter evident on higher exposure; data not shown). This finding was confirmed in phosphoinositide array experiments (Fig. 4B), in which the various lipids are titrated at different concentrations. These experiments indicate that the GLUE domain binds 3-phosphoinositides, particularly PtdIns(3,4,5)P 3 , and suggest that phosphoinositide binding could contribute to control the localization and/or activity of Eap45.
Eap45 Colocalizes with Ubiquitinated Proteins on Endosome Membranes-If Eap45 is involved in sorting of ubiquitinated cargo on endosomes, we would expect it to colocalize with such cargo. Because the localization of the ESCRT-II subunits in mammalian cells is not known, we therefore decided to investigate whether Eap45 is present on endosomes. For this purpose, we cotransfected cells with GFP-tagged Eap45 and another ESCRT-II subunit, Eap20 (ortholog of yeast Vps25), since previous studies have suggested that Vps36 expressed alone is unstable (19). In addition, we stimulated the cells with epidermal growth factor to activate endocytosis and lysosomal sorting of its receptor. Because overexpression of Vps proteins may cause their mislocalization, we only studied cells expressing very low amounts of Eap45. To visualize endosomes, we labeled the cells with an antibody to LBPA, which is a well characterized lipid marker for multivesicular late endosomes. As shown in Fig. 5, A-C, Eap45 localized on vesicular structures in the cell and partially colocalized with LBPA. Assuming that the function of Eap45 is to bind ubiquitinated cargo/protein during sorting, we asked whether ubiquitinated proteins colocalize with Eap45 on late endosomal/lysosomal structures. To address this, we transfected cells as described in Fig. 5, A-C, but stained the cells with antibodies against conjugated ubiquitin and LAMP-1 (Fig. 5, D-I). In agreement with the LBPA staining, Eap45 showed a partial overlap with LAMP-1-positive structures. We also found colocalization between Eap45 and structures containing ubiquitin. Since overexpressed ESCRT subunits are prone to aggregation (23), we cannot exclude the possibility that some of these structures represent aggregated Eap45. Importantly, however, the triple-labeling microscopy showed that several of the ubiquitin-labeled structures contained both Eap45 and LAMP-1. These results demonstrate that ESCRT-II exists in ubiquitin-enriched regions of MVBs/ late endosomes and are consistent with the idea that ESCRT-II is involved in sorting of ubiquitinated cargo. DISCUSSION In this report we have identified a novel ubiquitin-binding domain, GLUE, in the N-terminal part of the metazoan Vps36 proteins. Like other known ubiquitin-binding domains, the GLUE domain interacts with the hydrophobic surface patch of ubiquitin, and the K D value is in the high micromolar range. The GLUE domain is the first ubiquitin-binding domain shown to bind phosphoinositides, and the ability of the same domain to bind both ubiquitin and a phosphoinositide opens interesting possibilities for coordination of membrane interactions and cargo recognition. In line with a possible function of Eap45 in membrane interactions and cargo sorting, we found that it colocalizes with ubiquitinated proteins on endosome membranes.
Two recent reports revealed the 3.6-Å crystal structure of the yeast form of ESCRT-II (42)(43)(44) and indicated that the ESCRT-II core structure makes it well suited for engaging into multimeric protein networks on endosomal membranes. At the membrane, ESCRT-II can work as an interaction platform for the recruitment of other components of the sorting machinery, such as ubiquitinated cargo or the ESCRT-III complex. The low affinity for ubiquitin probably reflects the function of the ubiquitin-binding proteins during sorting of cargo into MVBs. Instead of permanently binding ubiquitin, these proteins may work as biological switches that transiently associate with ubiquitin to mediate the rapid transfer of cargo between complexes.
Although Vps36 proteins in yeast and other fungi also contain a GLUE-like domain, these have apparently acquired a novel ubiquitin-binding domain (NZF) inside their GLUE domain, leaving the latter dispensible for ubiquitin binding (see Fig. 6). There is precedence for domain insertions in PH domains in, for instance, phospholipase ␥1 (sw: PIG1_HUMAN). While other Vps proteins such as Hrs and Tsg101 contain ubiquitin-binding domains that are conserved from fungi to mammals (8,9,14,18,24,45), the Vps36 family appears to have undergone a functional and structural reorganization of its N-terminal GRAM/PH-like domain. Presumably, metazoans have kept the domain intact and employ it for ubiquitin binding, while yeast and other fungi have acquired a novel ubiquitin-binding domain (NZF) inside their GRAM/PH-like domain, leaving the latter "host domain" dispensible for ubiquitin binding.
Even though there is only a limited sequence similarity between the GLUE domain and GRAM/PH domains (Fig. 3), two additional findings argue that these may be structurally related. First, the secondary structure predictions of the GLUE domain match well with the crystal structure of the MTMR2 GRAM/PH domain. Second, like GRAM and PH domains, the GLUE domain is a phosphoinositide-binding module. We were a bit surprised to find the strongest binding in vitro to PtdIns(3,4,5)P 3 , a phosphoinositide typically produced at the plasma membrane upon agonist stimulation of cells (46). On the other hand, this phosphoinositide has also been found on endomembranes (47), and a role for PtdIns(3,4,5)P 3 in endosomal trafficking cannot be excluded. Alternatively, it is possible that, under in vivo conditions, the binding of the GLUE domain to PtdIns(3,5)P 2 , a phosphoinositide thought to be formed on endosomes (46), becomes more relevant. X-ray crystallography or NMR will be required to identify the phosphoinositide-binding interface of the GLUE domain. We note, however, that several basic residues are conserved among GLUE domains (see Fig. 3). By analogy with other phosphoinositide-binding domains, some of these are candidates for coordinating the phosphate groups of phosphoinositides.
We do not at present know the functional implications of the phosphoinositide binding, but one possibility is that the recruitment of ESCRT-II to membranes is enhanced upon cell stimulation as a consequence of increased sorting of receptors to lysosomes. Enhanced association of ESCRT-II to lipids may also be facilitated by other components of ESCRT-II, such as the N-terminal coiled-coil domain of Eap30/Vps22 (43). Interestingly, this lipid-binding region protrudes from the same end of ESCRT-II as the GLUE domain (see model in Fig. 6). Although the in vivo binding specificity to lipids often correlates with the in vitro assays, additional proteins and receptors on the membrane may regulate these interactions (48). Due to FIG. 5. Eap45 colocalizes with ubiquitin on MVBs or late endosomes. HeLa cells were transfected with GFP-Eap45 and myc-Eap20 (not shown) and then stimulated with epidermal growth factor for 45 min. The cells were then permeabilized and fixed before they were stained for LBPA (A-C) or ubiquitin and LAMP-1 (D-I). A-C, the cells were stained for confocal immuofluorescence microscopy with anti-LBPA, and the merged image is shown in C. D-I, same as A-C, but the cells were stained for ubiquitin and LAMP-1. Merged images are shown in F, H, and I. experimental limitations, we were not able to determine the effect of ubiquitin binding of the GLUE domain on its lipid association. However, the close proximity of both membrane and ubiquitin binding by the N-terminal domain of Vps36 suggests that these activities could be tightly linked and possibly cross-regulated. We thus speculate that engagement of one of the ligand binding sites may (allo)sterically affect the other. A possible role for such cross-regulation could be to prevent "empty" ESCRT-II complexes from associating with the endosome membranes in the absence of mono-ubiquitinated protein cargo. Further studies will be required to clarify this.
In summary, our studies have shown that mammalian Eap45 binds ubiquitin with similar specificity and affinity as other ubiquitin-binding domains. We find that ubiquitin-binding maps to a GRAM/PH-related domain that we call GLUE. Our findings open the possibility that binding of ESCRT-II to ubiquitinated cargo could be regulated by the presence of specific 3-phosphoinositides on the endosomal membranes or vice versa. Together with the observation that Eap45 is found in ubiquitin-enriched regions of late endosomes, our data support the idea that mammalian ESCRT-II plays a role in the sorting of ubiquitinated cargos on endosomes.  (43) is placed close to the phosphoinositide-binding site to suggest that both parts may contribute to membrane association of ESCRT-II. This model is also compatible with a possible regulatory interaction between the two GLUE domain ligands (see "Discussion" for details).