γ2-Adaptin, a Ubiquitin-interacting Adaptor, Is a Substrate to Coupled Ubiquitination by the Ubiquitin Ligase Nedd4 and Functions in the Endosomal Pathway*

γ2-Adaptin is a putative member of the clathrin adaptor protein family with unknown physiological function. We previously reported that γ2-adaptin acts as a ubiquitin receptor by virtue of its ubiquitin-interacting motif. Here we demonstrate that this motif mediates a specific physical interaction with the ubiquitin ligase Nedd4 and promotes ubiquitination of γ2-adaptin. By mapping regions of Nedd4 involved in binding to γ2-adaptin, we identified its C2 domain to be essential, whereas the WW and HECT domains are dispensable. Consistent with this, we uncovered that the C2 domain of Nedd4 is ubiquitinated itself and as such is recruited by the ubiquitin-interacting motif of γ2-adaptin for subsequent ubiquitin conjugation. Unlike known coupled ubiquitination reactions, this novel type of interaction leads to mono- and multi/polyubiquitinated γ2-adaptin. In addition, we show that γ2-adaptin functions in the endosomal/multivesicular body (MVB) pathway. Depletion of γ2-adaptin impairs the degradation of internalized epidermal growth factor and results in defective MVB morphology characterized by significantly enlarged vesicles. These defects cannot be rescued by γ1-adaptin, a closely related homolog of γ2-adaptin, which is unable to bind ubiquitin. Together, these results indicate that γ2-adaptin may operate within the MVB sorting system in a manner different from that of classic adaptins.

␥2-Adaptin is a putative member of the clathrin adaptor protein family with unknown physiological function. We previously reported that ␥2-adaptin acts as a ubiquitin receptor by virtue of its ubiquitin-interacting motif. Here we demonstrate that this motif mediates a specific physical interaction with the ubiquitin ligase Nedd4 and promotes ubiquitination of ␥2-adaptin. By mapping regions of Nedd4 involved in binding to ␥2-adaptin, we identified its C2 domain to be essential, whereas the WW and HECT domains are dispensable. Consistent with this, we uncovered that the C2 domain of Nedd4 is ubiquitinated itself and as such is recruited by the ubiquitin-interacting motif of ␥2-adaptin for subsequent ubiquitin conjugation. Unlike known coupled ubiquitination reactions, this novel type of interaction leads to mono-and multi/polyubiquitinated ␥2-adaptin. In addition, we show that ␥2-adaptin functions in the endosomal/ multivesicular body (MVB) pathway. Depletion of ␥2-adaptin impairs the degradation of internalized epidermal growth factor and results in defective MVB morphology characterized by significantly enlarged vesicles. These defects cannot be rescued by ␥1-adaptin, a closely related homolog of ␥2-adaptin, which is unable to bind ubiquitin. Together, these results indicate that ␥2-adaptin may operate within the MVB sorting system in a manner different from that of classic adaptins.
The covalent attachment of ubiquitin marks proteins for various cellular fates and functions, including proteasomal degradation, endocytosis, endosomal sorting, DNA repair, and virus budding. One way cells interpret and transmit the information conferred by ubiquitin is through proteins that bind ubiquitin noncovalently. These ubiquitin receptors contain one or more ubiquitin binding domains (UBD), 2 of which at least sixteen have been identified to date (1,2). The first discovered UBD was the ubiquitin-interacting motif (UIM) that is found in several ubiquitin receptors controlling endocytic membrane traffic (3). This class of proteins, including eps15, epsin, Hrs, and Stam, regulates the internalization of plasma membrane proteins into the endocytic pathway as well as the sorting of proteins into the multivesicular body (MVB) in a ubiquitin-dependent manner (4 -8). A recent addition to this group of UIM-containing receptors is ␥2-adaptin, which we originally identified as a hepatitis B virus (HBV) interacting protein in a yeast two-hybrid screen (9,10).
␥2-Adaptin is classified as a member of the heterotetrameric clathrin adaptor complex (AP) family (11). APs mediate the sorting of protein cargo and their incorporation into clathrincoated transport vesicles. Four AP complexes have been identified designated AP-1 through AP-4, and they all exhibit a similar organization consisting of two large subunits, a medium-sized subunit, and a small subunit (11,12). ␥2-Adaptin is highly similar to ␥1-adaptin (one large subunit of the trans-Golgi network/early endosome adaptor AP-1) in both primary sequence and modular domain architecture (13,14). Despite their relatedness, ␥2-adaptin appears to serve a separate yet unknown function distinct from that of ␥1-adaptin. This is evidenced by embryonic lethality in mice upon disruption of the ␥1-adaptin gene (15) and by the failure of ␥1-adaptin to functionally substitute for the essential role of ␥2-adaptin played during HBV egress from human liver cells (10). The unique function of ␥2-adaptin may rely on its ubiquitin-interacting ability (10), a feature that has not been described so far for other members of the classic adaptor protein complex family.
First evidence for a ubiquitin-dependent action of ␥2-adaptin in the endosomal pathway was provided by studies investigating the host cell requirements of HBV assembly. Budding of this enveloped virus depends on the UIM domain of ␥2-adaptin which couples the viral structural components at compartments positive for the late endosomal/MVB marker CD63 (10). Interestingly, during HBV egress ␥2-adaptin acts in concert with the cellular ubiquitin ligase Nedd4. However, a direct linkage between ubiquitin recognition by ␥2-adaptin and ubiquitin conjugation by Nedd4 has not yet been established.
For some UIM-containing proteins, like eps15, epsin, and Hrs, such a link has recently been discovered by the demonstration that these proteins are themselves ubiquitinated. This modification strictly depends on the presence of their UIMs and is, therefore, referred to as coupled ubiquitination (7,8,16). In this process, UIM-directed ubiquitin binding appears to * This work was supported by Deutsche Forschungsgemeinschaft Grants SFB 490-D1 and PR 305/1-3 (to R. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. recruit ubiquitin ligases to promote the ubiquitination of the receptors themselves. Among the two major classes of ubiquitin ligases, the homologous to the E6-AP carboxy terminus (HECT) domain E3 ligases have preferentially been implicated in coupled ubiquitination (7,17). Nedd4 (neuronal precursor cell expressed developmentally down-regulated), a prototype member of this family, contains a catalytic HECT domain with an active site cysteine that forms a thiolester bond with ubiquitin, a Ca 2ϩ -dependent phospholipid binding C2 domain, and two to four WW domains interacting with proline-rich sequences, like the canonical PPXY motif (18 -20). In the case of Hrs, Nedd4 and the Nedd4like ligase AIP4 mediate its ubiquitination (7,21), whereas eps15 has been shown to recruit monoubiquitinated Nedd4 and the RING-type ubiquitin ligase Parkin for its subsequent modification (17,22). The functional significance of this UIM-regulated ubiquitin network remains to be resolved, but it has been implicated to control the activity of the endocytic ubiquitin receptors and, hence, of the endocytic machinery (7,16,23). Nedd4 family members also ubiquitinate cellular cargos to target them for endocytosis and subsequent incorporation into intraluminal vesicles of the MVBs for either degradation, lysosomal functions, or exosomal release (18,19). The sorting of ubiquitinated cargo into this class of vesicles requires the coordinated action of at least three hetero-oligomeric complexes, referred to as ESCRT (endosomal sorting complex required for transport) complexes I, II, and III, and associated proteins. Individual subunits of ESCRT-I and ESCRT-II have different UBDs and appear to function as receptors for ubiquitinated cargo and/or ubiquitinated trans-acting components (24 -26). This sorting machinery is also exploited by many enveloped viruses, including HBV, for use in viral budding (27)(28)(29)(30)(31). Accordingly, the functional inhibition of individual subunits of the ESCRT machinery results not only in malformed, dysfunctional MVB but also in a loss of function in viral budding (24,25,27,32,33). A similar phenotype in terms of perturbation of endosomal morphology and viral budding has been reported to occur upon overproduction of ␥2-adaptin in mammalian cells (29), suggesting a potential role for this UIM-bearing receptor in the endocytic MVB pathway.
To explore the functional role of ␥2-adaptin, we here characterized biological features of this protein with focusing on its ubiquitin binding ability. We found that this property allows a specific and productive interaction with the ubiquitin ligase Nedd4. Unlike well established Nedd4/substrate interactions involving the WW domains of the ligase, we identified a novel type of interaction depending on the UIM of ␥2-adaptin and on the ubiquitinated C2 domain of Nedd4. We also show that ␥2-adaptin functions in the endosomal/MVB system in a manner different from that of ␥1-adaptin.

EXPERIMENTAL PROCEDURES
DNA Constructs-The expression vectors for human ␥1-adaptin and ␥2-adaptin containing N-terminal hemagglutinin (HA) tags (13) were kindly provided by K. Nakayama (Kyoto University). Construction of the ␥2⌬UIM mutant devoid of the UIM and the ␥2⌬528 -785 mutant lacking the hinge and ear domains has been described (10). The ␥2⌬193-785 mutant encodes the first 192 amino acids of ␥2-adaptin. The human Nedd4.1 gene cloned in plasmid pGADNOT (34) was a gift from F. Bouamr (Howard Hughes Medical Institute, New York). For ectopic expression in the N-terminal FLAG-tagged form, the Nedd4.1 gene was subcloned into the expression vector p3xFLAG (Sigma-Aldrich). The Nedd4.C894S mutant carries a cysteine-to-serine substitution in the HECT domain, whereas the Nedd4⌬HECT mutant contains a deletion of this domain (amino acids 534 -900). Nedd4.1-specific domain constructs encoding the N-terminal C2 domain (Nedd4.C2; amino acids 1-161) or the central WW domain (Nedd4.WW; amino acids 210 -544) in FLAG-tagged versions were kindly provided by E. Gottwein (University of Heidelberg). For mutation of ubiquitin acceptor sites, the nine lysine residues present in the C2-encoding region were replaced by arginines (Nedd4.C2⌬K). Primer sequences and mutagenesis details will be available on request. The vector for HA-tagged ubiquitin (35) was a gift from M. Treier (EMBL, Heidelberg, Germany) provided by D. Sitterlin (University of Versailles).
Cells and Transfections-The human hepatocellular carcinoma HuH-7 and the human cervical cancer HeLa cell lines were used. Transient transfections with plasmid DNAs were performed with Lipofectamine TM Plus (Invitrogen) according to the manufacturer's instructions. Unless otherwise indicated, plasmid DNAs were used at a 1:1 ratio in cotransfection experiments and adjusted to induce moderate levels of protein expression. For transfection of cells with small interfering RNA (siRNA) plus plasmid DNA, the Lipofectamine TM RNAiMAX transfection reagent (Invitrogen) was used. Briefly, 5 ϫ 10 5 cells per well of a 6-well plate were transfected with 40 pmol of siRNA according to the protocol of the supplier. After 24 h cells were retransfected with 1 g of RNase-free plasmid DNA using Lipofectamine TM Plus, and cells were harvested after an additional 48 h. The ␥2-adaptin-specific siRNA was obtained from Qiagen and targets the nucleotide positions 2065-2085 (AAACCCTGCTTTGCTGTTAAT). The Nedd4.1-specific siRNA (Sigma-Aldrich) corresponds to nucleotide positions 2206 to 2224 (CACATCAACATGAGCTGA). As a control, a nonsense siRNA with no known homology to mammalian genes (Sigma-Aldrich) was used.
Coimmunoprecipitation Analysis-To probe for ␥2-adaptin/ Nedd4 complex formation, cells were cotransfected with HAtagged ␥2-adaptin and FLAG-tagged Nedd4 constructs. After transient transfection, cells were washed with Tris-buffered saline (TBS, 50 mM Tris-HCl, pH 7.5, 150 mM NaCl) supplemented with 10 mM N-ethylmaleimide (NEM) and lysed with TBS containing 10 mM NEM, 0.5% Triton X-100, 0.01% sodium deoxycholate, and 1ϫ protease inhibitor mixture (Sigma-Aldrich) for 20 min on ice. Lysates were centrifuged for 5 min at 15,000 ϫ g and 4°C and subjected to immunoprecipitation using superparamagnetic polystyrene beads (Dynabeads M-280 Sheep anti-mouse IgG; DYNAL) that had been precoated with anti-FLAG or anti-HA antibodies in phosphatebuffered saline (PBS), 0.2% bovine serum albumin overnight at 4°C with agitation. For cross-linking the bound antibodies, the beads were washed with 0.2 M triethanolamine, pH 8.2, and incubated with diethyl pimelidate dihydrochloride (Sigma-Aldrich) in 0.2 M triethanolamine, pH 8.2, for 30 min at 20°C. After washing in 50 mM Tris-HCl, pH 7.5, the beads were suspended with PBS, 0.1% bovine serum albumin and reacted with the lysates for 3 h at 4°C. The immune complexes were washed 3 times with PBS, 0.2% Triton X-100 and once with PBS before SDS-PAGE and Western blotting analyses.
To probe for ubiquitination, cells were lysed in PBS containing 1% SDS, 10 mM N-ethylmaleimide, and protease inhibitor. After boiling for 10 min at 95°C, the lysates were adjusted to 1ϫ radioimmune precipitation assay buffer (by 10-fold dilution with PBS containing 1.1% Nonidet P-40 and 0.55% sodium deoxycholate) and cleared by centrifugation at 15,000 ϫ g for 10 min at 4°C. Samples were subjected to immunoprecipitation using a 10% protein A-Sepharose slurry or tosyl-activated, superparamagnetic polystyrene beads (DYNAL) that had been precoated with specific antibodies as outlined above. After immune complex formation for 4 h at 4°C with agitation, the slurry/beads were washed twice with PBS, 0.1% Nonidet P-40 and once with 125 mM Tris-HCl, pH 6.8, and prepared for SDS-PAGE.
Preparation of Membrane and Cytosolic Fractions-Cells were incubated with 0.1ϫ TBS for 10 min on ice, and swollen cells were disrupted by Dounce homogenization (20 strokes) and adjusted to 1ϫ TBS. After centrifugation for 20 min at 2500 ϫ g and 4°C, the postnuclear supernatant was layered on 250 mM sucrose, TBS and ultracentrifuged for 45 min at 160,000 ϫ g and 4°C in a SW60 rotor (Beckman) to separate membrane and cytosolic fractions. Proteins in the cytosolic fraction were precipitated with 10% trichloroacetic acid. The precipitates and pelleted membrane fraction were adjusted to the same volume sample buffer and analyzed by SDS-PAGE.
Immunofluorescence Microscopy-For protein immunostaining, cells grown on coverslips were fixed and permeabilized with ice-cold methanol containing 2 mM EGTA for 15 min at Ϫ20°C. After washing and blocking for 30 min in PBS containing 2% animal serum, cells were incubated with the indi-cated primary antibodies for 1 h at 37°C, rinsed with PBS, and then incubated with AlexaFluor-tagged secondary antibodies for 1 h at 37°C. DNA was stained with Hoechst 33342 (Sigma-Aldrich). Coverslips were washed with PBS and mounted onto slides using Fluoprep mounting medium (Biomerieux). For LPBA staining, cells were fixed with 3% paraformaldehyde, PBS for 20 min at room temperature, washed with PBS, and then treated with 50 mM ammonium chloride for 10 min. After blocking with PBS, 0.1% bovine serum albumin, cells were reacted with the anti-LBPA antibody in the presence of 0.05% saponin for 30 min at room temperature. After washing with PBS, cells were incubated with the Alexa-conjugated specific secondary antibody for 30 min and further processed as above. Z-stack images were acquired separately for each channel using a Zeiss Axiovert 200M microscope equipped with a Plan-Apochromat 100ϫ (1.4 NA) and a Zeiss Axiocam digital camera. Axiovision software 4.6 was used for merging pictures. Z-stack images were optically deconvoluted using the software supplied by Zeiss. Tiffs were assembled into figures using Photoshop CS2 (Adobe).
Epidermal Growth Factor (EGF) Uptake and Degradation Assay-Cells transfected with control or specific siRNA were grown on coverslips and starved in serum-free Opti-MEM medium (Invitrogen) for 1 h. Cells were then stimulated with Alexa 488-conjugated EGF (Molecular Probes) (0.02 g/ml in Opti-MEM) for 30 min at 37°C. After washing, the labeled EGF was chased by incubating the cells in serum-free medium for 3 h at 37°C. After the pulse or pulse/chase, cells were fixed in methanol and processed for fluorescence microscopy.

RESULTS
␥2-Adaptin, but Not ␥1-Adaptin, Binds Ubiquitin via Its UIM-The ubiquitously expressed human ␥2-adaptin is closely related to ␥1-adaptin and shares the domain structure, consisting of an N-terminal head and a C-terminal ear domain that are connected by a hinge region (11,13,14) (Fig. 1A). Previously, we identified a UIM motif in the head domain of ␥2-adaptin (LSLAVLNSSNV) that resembles the conserved UIM core sequence ⌽XXAXXXSXXe (⌽ denotes a large hydrophobic residue, X is any amino acid, and e denotes an acidic residue) (3). Mutational analyses revealed that the substitution of critical residues of the UIM diminished the ubiquitin binding ability of ␥2-adaptin, whereas a precise deletion of the UIM completely blocked ubiquitin binding (10). Because the head domains of ␥2and ␥1-adaptin show 69% amino acid identity (13), we inspected the sequence of ␥1-adaptin and found a similar sequence pattern (LSFALVNGNNI). This prompted us to analyze whether ␥1-adaptin could also bind ubiquitin. HuH-7 cells were transfected with a HA-tagged version of ␥1-adaptin, and lysates were incubated with ubiquitin-agarose. As controls, HA-tagged wild-type (wt) ␥2-adaptin and the UIM-deleted ␥2⌬UIM mutant were included in the ubiquitin pulldown assay. Although all three constructs were expressed at same levels, ␥1-adaptin did not interact with ubiquitin (Fig. 1B). As the ubiquitin binding ability is not conserved in ␥1-adaptin, this feature may contribute to distinct functions of the two ␥-adaptins.

JOURNAL OF BIOLOGICAL CHEMISTRY 32121
UIM-directed Ubiquitination of ␥2-Adaptin-Recent works have demonstrated that many ubiquitin receptors, especially those that contain UIM motifs, are frequently ubiquitinated themselves (7,8,16). To probe whether ␥2-adaptin shares this property, cells were transfected with HA-tagged ␥2-adaptin and lysed under denaturing conditions to prevent a possible post-lysis removal of ubiquitin. When extracts were analyzed by HA-specific immunoblotting, ␥2-adaptin appeared in its expected position of 90 kDa. Importantly, a slower migrating form of ␥2-adaptin with an apparent molecular mass of ϳ96 kDa could be reproducibly detected under these assay conditions (Fig. 1C). To test whether this 96-kDa species might represent ubiquitinated ␥2-adaptin, extracts were subjected to HA-specific immunoprecipitation, and the immune complexes were examined by Western blotting using anti-ubiquitin antibodies. These antibodies clearly recognized the 96 kDa form, thus demonstrating that a fraction of ␥2-adaptin is modified by monoubiquitination (Fig. 1C). An overexposure of the blot detected not only the 96-kDa form but also slower-migrating bands with a ladder-like appearance (data not shown), suggesting that ␥2-adaptin is modified by a mixture of mono-, multi-, and polyubiquitinated moieties. Prompted by this finding, we next analyzed the UIM-deficient ␥2⌬UIM mutant in the same experimental setting. Thereby, no evidence for a ubiquitin modification of this mutant was obtained (Fig.  1C), indicating that ubiquitination of ␥2-adaptin requires the presence of its UIM.
Previous analyses of UIM-bearing proteins have shown that their ubiquitination often occurs at sites that are N-terminal to the UIMs but not in a UIM itself (1,16). ␥2-Adaptin contains a high number of target lysine residues that hamper precise mapping of its ubiquitination site(s). For an initial mapping, we took use of a deletion mutant (␥2⌬528 -785) that lacks the C-terminal hinge and ear domain. This mutant was predominantly synthesized in its calculated mass of about 58 kDa and, importantly, appeared in a second form with a molecular weight enlarged by about 6 kDa, likely resembling ubiquitinated ␥2⌬528 -785 (Fig. 1D). Hence, ␥2-adaptin undergoes ubiquitination in its N-terminal head domain. A larger deletion encompassing not only the hinge/ear regions but also parts of the head domain including the UIM (␥2⌬193-785) consistently gave no signs of ubiquitinated species (Fig.  1D). This adds further proof that ubiquitination of ␥2-adaptin is coupled to its ubiquitin interacting capacity.
UIM-dependent Interaction of ␥2-Adaptin with Nedd4.1-For some ubiquitin receptors, their UIM-directed ubiquitination has been implicated to involve E3 ubiquitin ligases of the Nedd4-like family (7,17). This observation together with our recent demonstration that ␥2-adaptin cooperates with the Nedd4.1 ligase to promote virus egress in HBV-replicating liver cells (10) prompted us to probe for a possible link between these two proteins. To test for an in vivo interaction between ␥2-adaptin and Nedd4.1 (termed "Nedd4"), coimmunoprecipitation experiments were performed. HuH-7 cells were cotransfected with HA-tagged ␥2-adaptin and FLAG-tagged Nedd4, and cell extracts were probed with epitope-specific antibodies to show efficient synthesis of ␥2-adaptin and Nedd4. When immune complexes were isolated with antibodies against the HA-tagged ␥2-adaptin and examined by FLAG-specific Western blotting, we could easily identify coprecipitated Nedd4 (Fig.  2). This binding was specific, as only a faint background band was visible in samples prepared from cells transfected with Nedd4 alone. To examine whether the UIM of ␥2-adaptin had an impact on the association with Nedd4, lysates of ␥2⌬UIM/ Nedd4-cotransfected cells were assayed. Thereby, we observed that the ␥2⌬UIM mutant was totally blocked in Nedd4 binding. Consistent with this was the behavior of the two ␥2⌬528 -785 and ␥2⌬193-786 deletion mutants. Although the UIM-positive ␥2⌬528 -785 construct associated with Nedd4, the UIM-negative ␥2⌬193-785 mutant did not (Fig. 2). Of note, to avoid comigration of the ␥2-adaptin mutants with precipitating antibody chains, the coimmunoprecipitation assay was reversed in that Nedd4 was captured by antibodies and coprecipitated ␥2-adaptin constructs were analyzed by immunoblotting. Together, these data show a stable, UIM-directed interaction between ␥2-adaptin and Nedd4.
The C2 Domain of Nedd4, but Not the WW and HECT Domains, Mediates Interaction with ␥2-Adaptin-Because the association of ␥2-adaptin with Nedd4 involves its UIM and, presumably, binding of the UIM to ubiquitin, we reasoned that the UIM might bind to the ubiquitin moiety linked through a thiolester to the catalytic cysteine residue of Nedd4. A schematic representation of the domain structure of Nedd4 including the location of the active site cysteine within the HECT domain is shown in Fig. 3A (18 -20). To test this hypothesis, we constructed two catalytically inactive Nedd4 mutants either by substituting the active cysteine to serine (Nedd4.C894S) or by deleting the HECT domain (Nedd4⌬HECT). The mutants were efficiently synthesized in transfected HuH-7 cells, and surprisingly, they both retained the capacity to interact with ␥2-adaptin (Fig. 3B). The removal of the HECT domain even resulted in a stronger ␥2-adaptin binding as compared with the active site Nedd4 mutant. These results led us to conclude that the UIM-dependent binding of ␥2-adaptin to Nedd4 does not require (i) the ubiquitin attached to the catalytic site of the ligase, (ii) the ligase HECT structural domain, and (iii) its enzymatic activity. To understand how Nedd4 forms a stable complex with ␥2-adaptin, we made use of two other mutants carrying only its N-terminal C2 (Nedd4.C2) or central WW domains (Nedd4.WW) and tested them for an association with ␥2-adaptin. The Nedd4. C2 construct still interacted with ␥2-adaptin, whereas Nedd4.WW did not (Fig. 3B), suggesting that the C2 domain of Nedd4 plays a key role in binding ␥2-adaptin. This finding may also account for the higher affinity of Nedd4⌬HECT to ␥2-adaptin as compared with Nedd4.C894S, as the C2 domain has been shown to form an intramolecular interaction between the HECT domain in a subset of HECT-type ubiquitin ligases (37). Preventing such an interaction by deleting HECT of Nedd4 may render its C2 domain more accessible to make contacts with ␥2-adaptin.
Nedd4 Is Ubiquitinated within Its C2 Domain-The finding that the ␥2-adaptin-Nedd4 complex formation involves ubiquitin and the ligase C2 domain was unexpected and implied that Nedd4 per se and its C2 domain in particular may undergo ubiquitination. To detect such a modification, FLAG-tagged full-length Nedd4 and Nedd4.C2 were individually cotransfected with a plasmid encoding HA-tagged ubiquitin. The immunoprecipitation of lysates prepared in boiling SDS with anti-FLAG antibodies followed by immunoblotting with anti-HA antibodies demonstrated specific ubiquitinated Nedd4 forms that were absent in control-transfected cell lysates (Fig. 4A). However, because of the simultaneous analysis of Nedd4.C2 on the same gel, the resolution of these forms is imperfect. Importantly, an even more pronounced ladder of ubiquitinated forms could be detected for the Nedd4.C2 construct (Fig. 4A), implicating that this domain is a target for ubiquitin modification. To corroborate this finding, a Nedd4.C2 mutant was constructed in which all lysine residues, the known ubiquitin attachment sites, were substituted by arginines (Nedd4.C2⌬K). The wt and mutant C2 domains were cotransfected with HA-tagged ubiquitin, and lysates were prepared under denaturing conditions. The comparative analysis of the pattern of bands showed that at least three of the slower migrating species of Nedd4.C2 represented ubiquitinated forms, as they were absent in the Nedd4.C2⌬K mutant (Fig. 4B). Based on this finding, we next analyzed whether preventing ubiquitination of Nedd4.C2 might interfere with binding of ␥2-adaptin. HA-tagged ␥2-adaptin was cotransfected with FLAG-tagged Nedd4. C2 or Nedd4.C2⌬K, and cellular extracts were immunoprecipitated with anti-FLAG antibodies followed by HA-specific immunoblotting. As shown in Fig. 4C, the ubiquitination-defective Nedd4.C2⌬K mutant was almost completely blocked in ␥2-adaptin binding activity. Collectively, these data suggest a novel type of interaction that depends on the UIM of ␥2-adaptin and the ubiquitinated C2 domain of Nedd4.
␥2-Adaptin Colocalizes with Nedd4 and Recruits Its C2 Domain-For corroboration of these results, we performed deconvolution immunofluorescence microscopy to visualize the distribution of ␥2-adaptin and Nedd4 in cotransfected HuH-7 cells. ␥2-Adaptin yielded a cytoplasmic vesicular staining with some enrichment in punctuate structures (Fig. 5). Nedd4 was found mostly throughout the cytoplasm if expressed alone, and this distribution was not grossly altered upon coexpression with ␥2-adaptin. In line with our coprecipitation analyses, an overlay of the fluorescence patterns revealed an extensive degree of colocalization of Nedd4 and ␥2-adaptin. Colocalization was also found for Nedd4.C2, but differently. This protein was largely redistributed and recruited by coexpressed ␥2-adaptin and accumulated in punctuate structures (Fig. 5). Concomitantly, the distribution of ␥2-adaptin also changed, as it became substantially enriched in these structures. One likely interpretation of this finding might be that the complex formed between ␥2-adaptin and Nedd4.C2 is unable to dissociate due to the missing WW and HECT domains of the ligase. In the case of Nedd4.WW that cannot bind ␥2-adaptin, we could neither detect a significant colocalization nor a ␥2-adaptin-induced sequestration (Fig. 5). We conclude from these data that ␥2-adaptin binds and recruits Nedd4 via the C2 domain. .WW mutants are illustrated. B, Nedd4 interacts with ␥2-adaptin via C2 but not via WW and HECT. HuH-7 cells were transfected with HA-tagged ␥2-adaptin plus the indicated FLAGtagged Nedd4 mutants. Cell extracts were tested for the expression of the constructs by specific Western blotting (Input), and the migration of the proteins is indicated on the left of the panels. The input amounts correspond to 12% of the samples used for immune capture. For the extracts shown in lanes 1 and 2, the subsequent coimmunoprecipitation assay was done with anti-HA antibodies followed by FLAG-specific Western blotting (WB), whereas extracts depicted in lanes 3 and 4 were subjected to FLAG-specific immunoprecipitation (IP) followed by HA-specific WB.
␥2-Adaptin Is Ubiquitinated by Nedd4-To examine whether Nedd4 binding leads to ubiquitination of ␥2-adaptin, we used siRNA to reduce endogenous levels of Nedd4. HuH-7 cells were cotransfected with untagged ␥2-adaptin plus HAtagged ubiquitin and either a control siRNA or a Nedd4-specific siRNA, and cells were lysed under denaturing conditions. As shown in Fig. 6A, the Nedd4 siRNA effectively reduced Nedd4 expression compared with control siRNA-treated cells.
To probe for the ubiquitination status of ␥2-adaptin, lysates were immunoprecipitated with anti-␥2-adaptin antibodies followed by HA-specific Western blotting. Depletion of Nedd4 significantly diminished the amount of the ubiquitinated forms of ␥2-adaptin as compared with control siRNA-treated cells (Fig. 6A). Besides Nedd4.1, mammalian cells express the closely related Nedd4.2 isoform (19). Because this isoform should not be targeted by our silencing approach, these data suggest that ␥2-adaptin is specifically ubiquitinated by Nedd4.1. To substantiate this finding, we took use of the catalytically inactive, dominantnegative Nedd4.C894S mutant. Upon overexpression of this mutant in cells synthesizing untagged ␥2-adaptin and HA-tagged ubiquitin, the degree of ␥2-adaptin ubiquitination was dramatically reduced (Fig. 6B).
Membrane Association of ␥2-Adaptin Is Different from That of ␥1-Adaptin and Does Not Involve Nedd4 Function-Classic adaptor proteins, like ␥1-adaptin, with functions in intracellular vesicle trafficking are known to transiently interact with the cytosolic side of their donor and acceptor organelle membranes but otherwise are predominantly cytosolic (11,38). To determine whether or not ␥2-adaptin exhibits this characteristic, we compared the subcellular fractionation behavior of endogenous ␥1and ␥2-adaptin. HuH-7 homogenates were separated into soluble and membrane-associated fractions and probed by specific immunoblotting. The sufficient separation of the fractions was confirmed by assessing the distribution of the cytosolic heat shock protein Hsc70 that was exclusively found in the cytosolic pool (Fig. 7A). ␥1-Adaptin was mainly present in the soluble fraction with minor quantities in the membrane fraction, as expected (38). In contrast, ␥2-adaptin was equally distributed in both fractions, indicating that its membrane association is substantially stronger than that of ␥1-adaptin (Fig. 7A). To our surprise, its ubiquitinated forms were completely partitioned to the membrane fraction. Similar results were obtained upon ectopic expression of ␥2-adaptin (Fig. 7B, lanes 1 and 2). The UIM-deficient ␥2⌬UIM mutant yielded almost the same distribution profile as the wt protein, implicating that the UIM-directed ubiquitination of ␥2-adaptin is not responsible for its membrane localization (Fig. 7B, lanes 3  and 4). To further define elements important for recruiting FIGURE 4. Nedd4 is ubiquitinated within its C2 domain. A, HA-tagged ubiquitin was cotransfected with FLAG-tagged Nedd4, Nedd4.C2, or empty vector DNA (Control) into HuH-7 cells at 1:3 DNA ratios. Lysates prepared by boiling in SDS were analyzed by FLAG-specific Western blot (left) or by FLAG-specific immunoprecipitation followed by anti-HA immunoblotting (right), and samples were run on the same gel. Ub n denotes ubiquitinated Nedd4 forms; IgHC and IgLC indicate unspecifically stained immunoglobulin G heavy and light chains, respectively. B, HA-tagged ubiquitin (UB) was cotransfected with FLAG-tagged Nedd4.C2 or Nedd4.C2⌬K as above, and lysates were examined by FLAG-specific Western blot. The stars to the left denote ubiquitinated Nedd4.C2 molecules. C, Nedd4 ubiquitination is required for binding to ␥2-adaptin. HuH-7 cells were transfected with HA-tagged ␥2-adaptin plus the indicated FLAG-tagged Nedd4.C2 and Nedd4.C2⌬K constructs. Cell extracts were tested for the expression of the constructs by specific Western blot (Input). The input amounts correspond to 12% of the material used for immune capture performed with anti-FLAG antibodies followed by HA-specific Western blot.
␥2-adaptin to membranes, we analyzed its two C-terminaltruncated ␥2⌬528 -785 and ␥2⌬193-785 mutants and found that both mutants displayed the wt-like distribution pattern (Fig. 7B, lanes 5-8). Collectively, these results point to another distinctive feature between ␥2-adaptin and ␥1-adaptin, as they differ considerably in their partition to membrane fractions. Furthermore, to our surprise the first 192 amino acids of ␥2-adaptin are sufficient to mediate its peripheral membrane association.
Although these data pointed toward a dispensability of the UIM for ␥2-adaptin membrane association, we nonetheless tested whether Nedd4 had an impact on this property. We, therefore, investigated the distribution profile of ␥2-adaptin in Nedd4-depleted cells. Consistent with the results shown above, the knockdown of Nedd4 substantially inhibited ␥2-adaptin ubiquitination (Fig. 7C). The distribution of ␥2-adaptin to soluble and membrane-associated fractions, however, was virtually unaffected irrespective of whether Nedd4 was depleted or not. Thus, binding of ␥2-adaptin to Nedd4 is not necessary for its membrane association.
␥2-Adaptin Is Required for Normal Endosomal Morphology-Our previous works have implicated that ␥2-adaptin may function in the endosomal/MVB system, as it partially colocalizes with CD63-positive compartments in HBV-replicating liver cells and as it is able to perturb the endosomal morphology if heavily overexpressed (10,29). To further assess the physiological role of ␥2-adaptin, we reduced its endogenous expression in HuH-7 cells using siRNA (Fig. 8A) and examined the distribution of LBPA, a phospholipid enriched in MVBs/ late endosomal membranes (36). Depletion of ␥2-adaptin altered the morphology of MVBs marked by LBPA in that they were significantly enlarged and clustered (Fig. 8A). A similar endosomal enlargement phenotype was evident in ␥2-adaptin-depleted HeLa cells (data not shown), implying that it is not merely restricted to liver cells.
␥2-Adaptin Is Required for Efficient Degradation of EGF-The altered morphology of MVBs/late endosomes suggests that depletion of ␥2-adaptin may affect some steps of the endosomal membrane traffic. To provide functional data supporting this, we monitored endocytosis of EGF that is bound to its receptor at the cell surface then internalized and delivered via the MVB pathway to the lysosomal lumen for degradation (39). Control-and ␥2-adaptin-depleted cells were incubated with Alexa-conjugated EGF for 30 min and  A second transfection with untagged ␥2-adaptin and HA-tagged ubiquitin DNA was performed 1 day later, and cellular lysates were harvested by boiling in SDS after an additional 2 days. Lysates were subjected to Nedd4and ␥2-adaptin-specific Western blotting (WB) to demonstrate depletion of Nedd4 (lanes 1 and 2) and equal expression levels of ␥2-adaptin (lanes 3 and 4), respectively. The input amounts correspond to 12% of the samples used for immune capture. Diluted lysates were precipitated (IP) with anti-␥2-adaptin antibodies and detected by HA-specific Western blot (lanes 5 and 6). B, overexpression of Nedd4.C894S inhibits ␥2-adaptin ubiquitination. Untagged ␥2-adaptin plus HA-tagged ubiquitin were coexpressed with wt Nedd4 (Nedd4.wt) or the dominant-negative Nedd4.C894S mutant (Nedd4.dn). Lysates were examined by specific immunoblotting to probe for efficient protein expression (Input). The coimmunoprecipitation was done with anti-␥2adaptin antibodies followed by HA-specific Western blot.
analyzed directly or after a chase of 180 min. During the pulse, no obvious difference in EGF loading and uptake could be observed between control-and ␥2-adaptin-treated cells (Fig. 8B). However, in pulse-chased cells, disappearance of internalized EGF was almost complete in control cells, whereas in ␥2-depleted cells a substantial amount of EGF could still be detected at this time (Fig.  8B). These results indicate that reduced levels of ␥2-adaptin did not perturb the internalization of EGF from the cell surface but, rather, impaired its degradation and resulted in the accumulation of EGF to enlarged endosomal structures.

DISCUSSION
Among the typical members of the adaptor protein family, ␥2-adaptin is unique in its ubiquitin binding ability, specified by a UIM motif. Here we show that this motif serves as a signal for ubiquitination of ␥2-adaptin by recruiting the ubiquitination machinery. To understand how ubiquitination of ␥2-adaptin is coupled to ubiquitin binding, we analyzed the underlying mechanistic requirements and found a so far unrecognized feature of this process involving a stable interaction between the UIM and the ubiquitinated C2 domain of a ubiquitin ligase.
We demonstrate that ubiquitination of ␥2-adaptin depends on the presence of its UIM and requires the action of the Nedd4.1 ubiquitin ligase. Although Nedd4 family members have been suggested to play functionally redundant roles (19), the ubiquitination of ␥2-adaptin is specifically mediated by Nedd4.1, as evidenced by the lack of modification in Nedd4.1-depleted cells. Thus, neither the Nedd4.2 isoform nor other HECT-type ubiquitin ligases can functionally substitute for Nedd4.1 in this process. In support for the specific involvement of Nedd4.1, we identified a physical interaction between this ligase and ␥2-adaptin in life cells. This interaction involves the UIM of ␥2-adaptin pointing toward a dual function of this motif in ubiquitin binding and promoting ubiquitination. In this regard, ␥2-adaptin represents a new example of a subset of ubiquitin receptors, including eps15, epsin, and Hrs, which undergo UIM-directed self-ubiquitination (7,8,16). However, unlike these receptors, ␥2-adaptin becomes not only monoubiquitinated but also multi/ polyubiquitinated. One possible explanation for the apparently distinct outcomes of coupled ubiquitination is that the nature of interactions between the ubiquitin receptors and their ligases engaged may be different. In the case of eps15 that is exclusively monoubiquitinated by Nedd4, no stable interaction could be detected between the substrate and its ligase (17). This has been interpreted in such that, after catalysis of monoubiquitination, the contact between enzyme and substrate is rapidly lost, thereby restricting further modification events. In contrast, ␥2-adaptin tightly associates with Nedd4. Because Nedd4 has been shown to be capable in catalyzing both monoubiquitination and polyubiquitination (40), its high affinity to ␥2-adaptin may allow its catalytic activity to direct the addition of mono-plus multi/polyubiquitin chains.
By mapping regions of Nedd4 involved in binding to ␥2-adaptin, we found that its catalytic HECT domain in general and its active site cysteine in particular are dispensable. This indicates that the UIM of ␥2-adaptin does not recognize the thiolester-conjugated ubiquitin of Nedd4 but, rather, ubiquitin moieties attached to lysine residue(s) of the ligase. In support of this view, we identified the C2 domain of Nedd4 as the critical determinant of complex formation with ␥2-adaptin and uncovered its modification with ubiquitin. It, therefore, appears that the UIM of ␥2-adaptin targets ubiquitinated C2 of Nedd4 for subsequent ubiquitination of the adaptor. These findings confirm and extend recent observations drawn from an in vitro study in which the mechanistic basis for the coupled monoubiquitination of eps15 has been investigated (17). Similar to ␥2-adaptin, eps15 can be ubiquitinated by Nedd4 in a process that requires the ubiquitination of the ligase. If ubiquitin was  5 and 6), or ␥2⌬193-785 (lanes 7 and 8), partitioned into particulate and soluble fractions, and probed by HA-specific immunoblotting. C, membrane association of ␥2-adaptin does not require Nedd4. Cells were transfected with anti-Nedd4 siRNA (siNedd4) or control siRNA (siControl) and subsequently transfected with HA-tagged ␥2-adaptin. After separation into particulate and soluble fractions, samples were analyzed by HA-specific immunoblotting. Stars denote ubiquitinated ␥2-adaptin molecules.
covalently attached to Nedd4 in vitro, the ability of the ligase to execute monoubiquitination of eps15 was profoundly enhanced. As similar mechanisms apparently apply for the coupled ubiquitination of eps15 and ␥2-adaptin, it is tempting to speculate that this may be a general strategy employed by other UIM/UBD-containing proteins. In this respect it will be interesting to decipher the nature of the ubiquitin ligase responsible for ubiquitination of Nedd4. Moreover, our discovery that the ubiquitinated C2 domain of Nedd4 is essential for the recognition and, hence, ubiquitination of ␥2-adaptin raises an important question of whether this domain is relevant for other productive Nedd4/substrate interactions. In addition, it will be of interest to determine whether C2 domains of other HECT fam-ily members are modified with ubiquitin to regulate substrate selection. One result arguing in favor is provided by the ubiquitin ligase Rsp5, the sole member of the Nedd4 family in Saccharomyces cerevisiae, whose C2 domain has been shown to be required for ubiquitination of endosomal cargo (41). Interestingly, two lysine-rich clusters of the Rsp5 C2 domain critically contribute to cargo recognition and ubiquitination, likely by mediating precise localization of the ligase to endosomal membranes for subsequent function. However, whether this lysinerich patch is functionally modified with ubiquitin, as it is the case for the Nedd4 C2 domain, has not been experimentally addressed in that study.
The Nedd4 C2 domains are Ca 2ϩ /lipid binding modules and are thought to function as membrane recruitment domains involved in protein localization and trafficking (18 -20). At first glance it, therefore, appears possible that Nedd4 contributes to the recruitment of ␥2-adaptin to membranes. However, our results argue against this possibility because (i) a ␥2-adaptin mutant defective in Nedd4 binding still cofractionates with membranes, (ii) the membrane localization of ␥2-adaptin is not distorted in Nedd4-depleted cells, and (iii) the Nedd4 C2 domain is redistributed by ␥2-adaptin rather than vice versa. Accordingly, one possible function of ␥2-adaptin may be to act as an adaptor for Nedd4, recruiting it to membrane compartments for subsequent ubiquitination. In support, we observed that ubiquitinated ␥2-adaptin chains are exclusively present in membrane fractions. Such a role of ␥2-adaptin to localize or enrich Nedd4 in a specific cellular environment would be consistent with functions of other proteins interacting with Nedd4 family members. For example, annexin XIIIb physically interacts with the C2 domain of Nedd4, thereby recruiting the ligase to the apical plasma membrane of polarized epithelial cells (42). In the case of the lysosomal-associated multispanning membrane protein 5 (LAPTM5), Nedd4 is recruited to assist in LAPTM5 trafficking from the Golgi to the lysosome (43). Noteworthy, however, neither annexin XIIIb nor LAPTM5 become ubiquitinated by the recruited ligase which is in contrast to ␥2-adaptin. It, therefore, remains to be solved whether ␥2-adaptin binding to Nedd4, its ubiquitination by Nedd4, or both features are critical for its function.
At present, we do not know the precise function of ␥2-adaptin, but the results of this and previous works provide compelling evidence that this protein acts in the endosomal/MVB sorting and trafficking system. We show that the knockdown of ␥2-adaptin leads to a prominent decrease in the degradation of endocytosed EGF concomitant with its accumulation in enlarged vesicles. Furthermore, the loss of ␥2-adaptin results in defective MVB morphology characterized by considerably enlarged vesicle structures that stained positive for LBPA, a highly specific marker for MVBs (36). These results coincide with our previous studies demonstrating that heavily overexpression of ␥2-adaptin perturbed the normal endosomal morphology and induced aberrant, dysfunctional MVBs (29). To explain the perturbing effect of both deficit and excess ␥2-adaptin on MVB structure and function, it seems conceivable that an imbalance in the concentration of the adaptor might interfere with productive interactions required for proper MVB maintenance. Intriguingly, the phenotypes induced by up-and down-FIGURE 8. Depletion of ␥2-adaptin affects the morphology of endosomal compartments and inhibits EGF degradation. A, for analysis of MVB morphology, the distribution pattern of LPBA was monitored in control-treated HuH-7 cells (siControl) and in cells treated with specific siRNA against ␥2-adaptin (si␥2) by staining with anti-LBPA antibodies (bottom). The knockdown efficiency of endogenous ␥2-adaptin was quantitated at protein level by anti-␥2-adaptin Western blotting (top). The experiment was repeated three times, and representative images are shown. B, for analysis of EGF uptake and degradation, the fate of internalized Alexa 488-EGF was investigated in control-treated HuH-7 cells (siControl) and in cells treated with specific siRNA against ␥2-adaptin (si␥2). Cells were stimulated with labeled EGF for 30 min and examined for Alexa 488-EGF and DNA staining either directly (top) or after a chase for 3 h (bottom). Bars, 10 m. regulation of ␥2-adaptin are reminiscent to those evoked by functional loss of subunits of the MVB sorting machinery, in particular of ESCRT-III and Vps4 (24 -26, 32, 33). Hence, ␥2-adaptin may act within the ESCRT machinery by virtue of its ubiquitination and/or ubiquitin-and Nedd4 binding abilities. Several known subunits of this machinery contain ubiquitin recognition modules that are required to sort ubiquitinated cargo into the forming MVB vesicles (24 -26) and that may bind ubiquitinated trans-acting components, such as ␥2-adaptin, to coordinate cargo transport and vesicle formation. Within this network, the UIM of ␥2-adaptin may serve as a docking site for Nedd4, thereby bringing the ligase to endosomal compartments. Nedd4 family members have been implicated to function in endocytic trafficking and both viral and vesicle budding into the MVBs (27), but so far a direct association between the ligases and the ESCRT core machinery, including ESCRT-I, -II, -III, and Vps4, has not been detected (44). This has led to the suggestion that the ubiquitin ligases may interact with the ESCRT cascade via an as yet unidentified bridging factor that might be regulated by ubiquitination (28,44). Based on the results presented here, it is tempting to speculate that one such factor may be ␥2-adaptin.
Although the native function of ␥2-adaptin remains to be established, our results add new evidence for ␥2-adaptin being a unique entity distinct from ␥1-adaptin. Our silencing experiments indicate that the loss of ␥2-adaptin cannot be rescued by ␥1-adaptin with regard to both the endosomal enlargement phenotype and the delay in EGF degradation. Because ␥1-adaptin does not share the ubiquitin binding ability of ␥2-adaptin, this character likely prevents its access to the ubiquitin network and presumably to the endosomal/MVB pathway. Unlike ␥1-adaptin, ␥2-adaptin in turn may not operate as a typical vesicle-forming adaptor but, rather, as a monomeric ubiquitin receptor. In favor of this view, a ␥2-adaptin-containing AP complex has not been described, and a yeast two-hybrid screen using ␥2-adaptin as bait did not render interactions with other adaptin subunits. 3 In this respect, ␥2-adaptin would resemble the monomeric GGAs (Golgi-localizing, ␥-adaptin ear domain homology, ADP-ribosylation factor-binding proteins) that sort cargo without being major constituents of the transport vesicle (45). GGAs also possess ubiquitin-interacting activity (46,47), and GGA3 has been shown to become monoubiquitinated by the ring-type ubiquitin ligase hVPS18 in a UBD-dependent manner (48,49).
In summary, we have unraveled some mechanistic insights of ubiquitin receptor ubiquitination. Whether similar mechanisms apply for coupled ubiquitination of candidate proteins other than ␥2-adaptin remains to be determined. The finding that ␥2-adaptin appears to act in the MVB sorting pathway is another aspect that deserves further study.