Hrs, a Mammalian Master Molecule in Vesicular Transport and Protein Sorting, Suppresses the Degradation of ESCRT Proteins Signal Transducing Adaptor Molecule 1 and 2*

The degradation and sorting of cytoplasmic and cell-surface proteins are crucial steps in the control of cellular functions. We previously identified three mammalian Vps (vacuolar protein sorting) proteins, Hrs (hepatocyte growth factor-regulated tyrosine kinase substrate) and signal transducing adaptor molecule (STAM) 1 and -2, which are tyrosine-phosphorylated upon cytokine/growth factor stimulation. Hrs and the STAMs each contain a ubiquitin-interacting motif and through formation of a complex are involved in the vesicle transport of early endosomes. To explore the mechanism and cellular function of this complex in mammalian cells, we established an Hrs-defective fibroblastoid cell line (hrs-/-); embryos with this genotype died in utero. In the hrs-/- cells only trace amounts of STAM1 and STAM2 were detected. Introduction of wild-type Hrs or an Hrs mutant with an intact STAM binding domain (Hrs-dFYVE) fully restored STAM1 and STAM2 expression, whereas mutants with no STAM binding ability (Hrs-dC2, Hrs-dM) failed to express the STAMs. This regulated control of STAM expression by Hrs was independent of transcription. Interestingly, STAM1 degradation was mediated by proteasomes and was partially dependent on the ubiquitin-interacting motif of STAM1. Revertant Hrs expression in hrs-/- cells not only led to the accumulation of ubiquitinated proteins, including intracytoplasmic vesicles, but also restored STAM1 levels in early endosomes and eliminated the enlarged endosome phenotype caused by the absence of Hrs. These results suggest that Hrs is a master molecule that controls in part the degradation of STAM1 and the accumulation of ubiquitinated proteins.

bind to specific ligands or macromolecules, are internalized from the cell surface into membrane compartments called endosomes. Although some cargos are pinched off in recycling endosomes to be sent back to the cell surface, other receptors and molecules are pinched off into endosomes that form multivesicular bodies (MVB). 1 As the maturation step proceeds, early endosomes become acidic to form late endosomes, which then fuse with acidic compartments called lysosomes, which contain many acidic proteases. Although it has not been clear how receptors that have captured their ligands are recognized and sequestered into the endocytic vesicles, a growing body of evidence suggests the involvement of monoubiquitin modification and implicates the intracytoplasmic region of the receptors (1).
The identification of vacuolar protein sorting (Vps) mutants has permitted genetic analyses in yeast to identify the proteins that make up the core machinery for MVB formation. The functional loss of Vps proteins results in deformed MVBs with unusually enlarged vesicles, called "class E" compartments, and defective vesicular trafficking. At present, 17 Vps genes are known; they are genetically categorized into four kinds of endosomal sorting complexes required for transport (ESCRT), ESCRT-0, I, II, and III. One of the ESCRT proteins involved in MVB formation is yeast Vps27. Vps27, along with Hse1, acts in an ESCRT-0 complex upstream of ESCRT-I to assist in protein sorting (2).
In studies to find molecules involved in cytokine/growth factor signaling, we identified Hrs, STAM1, and STAM2 (3)(4)(5). Hrs is a mammalian orthologue of yeast Vps27, and STAM1 and STAM2 are mammalian counterparts of yeast Hse1. We as well as others previously reported that cytokines and growth factors such as interleukin 2, granulocyte-macrophage colonystimulating factor, hepatocyte growth factor, platelet-derived growth factor, and epidermal growth factor induce the tyrosine phosphorylation of Hrs and the STAMs (3,4,6). The rapid kinetics of their phosphorylations prompted us to look for the functions of these molecules, which are immediately downstream of the receptors. Hrs possesses several functional domains, including a FYVE finger domain, which binds specifi-cally to phosphatidylinositol 3-phosphate and is important for membrane anchoring, and a clathrin binding domain, required for membrane-coated vesicle binding (7)(8)(9)(10)(11). Hrs also contains an N-terminal VHS (Vps27-Hrs-STAM) domain and a ubiquitin-interacting motif (UIM). The overexpression of Hrs, but not of a UIM deletion mutant, inhibits the recycling of a ubiquitin-fused transferrin receptor, suggesting a role for Hrs-UIM in vesicular transport (12). Furthermore, an accumulation of ligand-activated epidermal growth factor receptors within early endosomes is observed in cells overexpressing Hrs, suggesting the involvement of Hrs in ubiquitinated protein sorting (11,(13)(14)(15).
After the identification of STAM1 and STAM2 (EAST/Hbp), two closely related molecules, we showed that their roles in intracellular cytokine/growth factor signaling included induction of the proto-oncogene c-myc (3, 5, 16 -18). The STAMs have a unique structure, including a VHS, a UIM, an Src homology 3 (SH3) domain, and an ITAM (immunoreceptor tyrosine-based activation motif) (3,5). We and others reported that the ITAMs of STAM1 and STAM2 associate with Hrs through its second coiled-coil (CC2) domain (4). Like its binding partner, STAM2 binds directly to ubiquitins (19). Both Hrs and the STAMs bind Eps15, and therefore, all these molecules seem to participate in the sorting of ubiquitinated proteins into the MVB pathway (19).
To further characterize the function of the mammalian STAM-Hrs complexes in MVB formation and/or protein sorting, we established Hrs-deficient fibroblastoid cell lines. The results of introducing wild-type or a mutant Hrs into these cells suggested a novel function for Hrs in regulating the fate of ubiquitinated proteins. We demonstrate here that Hrs controls the STAM protein stability. Our demonstration of the STAM relationship with ubiquitinated proteins provides insight into how their UIM domain functions in protein sorting.

EXPERIMENTAL PROCEDURES
Targeted Disruption of hrs in Vivo-Genomic clones of the mouse hrs gene were isolated from a FixIImouse 129/Sv genomic library (Stratagene). A targeting construct for hrs was designed to replace a 0.6kilobase (HindIII-HindIII) genomic fragment encompassing exon 6, flanked by 3.2-kilobase (XbaI-XbaI) and 3.0-kilobase (EcoRI-XbaI) genomic sequences (Fig. 1A). A pGK-neo cassette flanked by a pair of loxP sequences and a diphtheria toxin A-chain (DT) gene cassette without a polyadenylation site were inserted into the construct to allow further positive/negative selection. The construct was linearized and transferred into 129/Sv-derived J1 ES cells by electroporation, and G418-resistant colonies were picked (20 -22). The occurrence of homologous recombination events for the hrs allele was examined by Southern hybridization using a 1.8-kilobase hrs genomic fragment (XbaI-BamHI) as the probe (Fig. 1B). The targeted hrs loxP-flanked (floxP) ES clone was confirmed to have a normal number of chromosomes and was injected into C57BL/6 blastocysts. The blastocysts were then transferred into foster mothers to obtain chimeric mice. By crossing the chimeras with C57BL/6 mice we obtained F 1 heterozygous mice carrying the floxP allele. The genotypes were confirmed by both genomic PCR and Southern blot analyses (data not shown). The F 1 heterozygous mice were intercrossed, but the genotyping of more than 100 offspring showed that no viable hrs floxP/floxP mice were produced. To identify hrs floxP/floxP embryos in utero, the genotypes of living embryos were determined using yolk sac DNA (23). The following oligonucleotide primers were used: Hrs-1650F (primer A), 5Ј-GAGTGAGGAGGCGGT-GTTCCCTAAACCTTG-3Ј; Hrs-1953R (primer B), 5Ј-AACATATACT-GCTGGCAAAGCATCCATACA-3Ј; Hrs-CKO1787R (primer C), 5Ј-TAT-AGCATACATTATACGAAGTTATGTCGA-3Ј. PCR was carried out by initial denaturation at 94°C for 5 min followed by 35 cycles of 30 s at 94°C, 30 s at 60°C, and 30 s at 72°C. The wild-type and mutant alleles gave rise to PCR-amplified fragments of 330 and 160 bp, respectively (Fig. 1C).
Cell Culture-HRSd and HRSd sublines transfected with wild-type Hrs or Hrs mutants were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics under 7% CO 2 in a humidified incubator. To transfect the cells with plasmid DNAs, FuGENE TM 6 transfection reagent (Roche Applied Science) was used according to the manufacturer's protocol. Where indicated lactacystin (Kyowa Medex) and E-64-d (Peptide Institute, Inc.) were added to the cell culture to a final concentration of 10 and 100 M, respectively.
Immunoprecipitation and Immunoblots-Immunoprecipitation and immunoblotting were carried out as described previously (4). In brief, cells were lysed in Nonidet P-40 lysis buffer (1% Nonidet P-40, 40 mM Tris-HCl (pH 7.5), 150 mM NaCl, 2 mM EDTA, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, and 20 g/ml aprotinin). The cell lysates were precleared of cellular debris by centrifugation (10,000 ϫ g) for 20 min at 4°C and were then subjected to immunoprecipitation with antibodies immobilized on protein A-Sepharose beads (Amersham Biosciences) at 4°C overnight. For this assay rabbit anti-Hrs polyclonal antibodies (4) and anti-ubiquitin (Santa Cruz), anti-STAM1 (TUS-1) (29), anti-STAM2 (ST2-2) (5), and anti-V5 monoclonal antibodies (Invitrogen) were used. The immunoprecipitates were then separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes (Millipore). After being blocked with 5% nonfat milk in Tris-buffered saline containing 0.1% Tween 20, the membranes were probed with the indicated primary antibodies. After another wash, the membranes were probed with horseradish peroxidase-conjugated secondary antibodies (Cell Signaling). Signals were visualized by the ECL detection system (Amersham Biosciences), and digital images were collected by a Lumi-Imager F1 (Roche Applied Science). For degradation assays, the cells were cultured in the presence of 25 g/ml cycloheximide (Wako) for the period indicated, and the cell lysates from these cultures were analyzed by Western blotting.
Northern Blots-Northern blot analyses were performed as described previously (29). In brief, total RNA was prepared using TRIzol Reagent (Invitrogen). Twenty micrograms of total RNA was fractionated by electrophoresis in a 0.8% agarose gel containing formaldehyde. The separated RNAs were then transferred onto Hybond-N ϩ membranes (Amersham Biosciences) in 10 x SSC. After UV-cross-linking, the membranes were subjected to prehybridization and hybridization in a solution containing 0.5 M sodium phosphate (pH 7.2), 1 mM EDTA, and 7% SDS at 65°C. A probe prepared by radiolabeling with [␣-32 P]dCTP using a Random Prime-DNA labeling kit (Takara Shuzo Co.) was added, and hybridization was carried out overnight. Membranes were then extensively washed with a washing buffer (2ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate), 0.1% SDS) at 65°C, and the signals were visualized with a Bio-Image Analyzer, FLA-2000 (Fuji Film). The probes used to detect murine stam1, stam2, and glyceraldehyde-3-phosphate dehydrogenase (gapdh) mRNAs were described elsewhere (29,30). Immunofluorescence Microscopy and Immunostaining-Cells were seeded into 35-mm glass-bottomed dishes (MatTek Co.) at a density of 1 ϫ 10 5 cells/dish 1 day before the transfection. DsRed1-Eps15 was then introduced as described above. After 48 h, the cells were washed twice with phosphate-buffered saline, fixed with 4% paraformaldehyde for 15 min, permeabilized for 10 min with phosphate-buffered saline containing 0.1% Triton X-100, and then blocked for 30 min with phosphatebuffered saline containing 10% fetal calf serum and 0.1% Triton X-100. For immunostaining, the fixed samples were incubated with the indicated primary antibodies, washed 3 times, and further probed with the secondary antibodies (anti-rabbit, -mouse, and -goat IgG antibodies conjugated with Alexa 488, Alexa 594, and Alexa 350, respectively (Molecular Probes)). The microscopic images were examined using the Leica DMIRBE microscope system with a PL FLUOTAR 100ϫ/1.30 -0.60 oil immersion objective.
Metabolic Labeling of Cellular Proteins-For pulse-chase experiments, HRSd and 293T cells that had been transiently transfected with the indicated plasmids using FuGENE TM 6 were radiolabeled with the Pro-mix L -[ 35 S] in Vitro Cell Labeling Mix at a concentration of 100 Ci/ml (Amersham Biosciences) in Met/Cys-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum that had been predialyzed against phosphate-buffered saline for 30 min. At the end of the pulse-labeling period, the cells were washed and chased for the indicated times in complete Dulbecco's modified Eagle's medium supplemented with methionine and cysteine. Cells were collected by centrifugation, and the lysates made from them were subjected to further immunoprecipitation assays. Radioactivity was visualized and analyzed using Science Lab 2001 Image Gauge Version 4.0 software (Fuji Film).

Establishment of hrs-deficient Cell Lines by Gene Targeting-
The hrs knock-out mice have an embryonic lethal phenotype, as described previously (26,31). To analyze the biological significance of Hrs in vivo, we sought to generate conditional hrs gene-targeted mice. The targeting construct was designed to delete exon 6 of the hrs genomic sequence (Fig. 1A). Exon 6 (E6) codes for amino acids 139 -156, and therefore, the resultant hrs gene-targeted allele was expected to code for an Hrs protein with a minimal N-terminal portion that lacked part of the VHS domain (amino acids 1-138). After a positive-negative selection screen, several recombinant ES clones were selected and injected into blastocysts. The chimeric mice obtained were crossed with each other to obtain heterozygous mice (hrs ϩ/floxP ). The genotype was confirmed both by Southern blot and genomic PCR analyses (Fig. 1, B and C). Although the heterozygous mice were viable and fertile, when they were intercrossed we did not obtain viable hrs floxP/floxP offspring. Indeed, our timed breeding experiments revealed that the hrs floxP/floxP embryos were not viable after E10.5 (data not shown). We concluded that the conditional hrs floxP allele was nonfunctional. Although we were unable to establish cell lines from our original hrs gene knock-out mice (26) because the allele still produced a C-terminal-truncated Hrs (amino acids 1-454), these new mutant hrs floxP/floxP embryos were ideal for preparing mouse embryonic fibroblastoid cells (MEFs). We successfully prepared MEFs from E9.5 embryos, and among ran-domly prepared MEF cells only the hrs floxP/floxP MEFs failed to survive after a few passages; in comparison, wild-type MEFs survived for many passages. We immortalized the hrs ϩ/ϩ and hrs floxP/floxP MEFs using the SV40 large T antigen. We obtained an immortalized cell line carrying the hrs floxP/floxP genotype and a control hrs ϩ/ϩ cell line, MEFw. Next, the hrs gene and the neomycin resistance gene (neo r ) were fully inactivated by deleting the gene segment encoding exon 6. By introducing the transient expression of Cre recombinase into the cells carrying the hrs floxP/floxP genotype, we obtained a cell line, HRSd, with the hrs ␦/␦ genotype. Deletion of the region flanked by the two loxP sequences was confirmed by genomic PCR (Fig. 1D). Although the ␦ allele could potentially have produced an Nterminal 138-amino acid fragment, the defective expression of FIG. 1. Targeted disruption of the hrs gene. A, schematic restriction maps of the wild-type, targeting construct, and mutant alleles. The mutated allele is shown in its floxP and ␦ forms. E5, E6, and E7 stand for exons 5, 6, and 7 of the hrs genome. B, Xb, and E stand for the restriction sites for BamHI, XbaI, and EcoRI, respectively. The 3Ј probe used for the analytical Southern blot is shown. Primer pairs used to detect the targeted allele are shown; primers A and C for the first loxP-containing joint region and primers D and E for the neomycinresistance gene (neo r ). B, a representative genomic Southern blot analysis for the wild-type (wild) and recombinant (floxP) alleles from ES cells. Genomic DNAs extracted from ES cell clones were digested with BamHI. After gel separation and blotting, the filter was hybridized with the 3Ј probe. The 10.5-and 7.6-kilobase pair (kbp) fragments indicate wild-type and recombinant alleles, respectively. C, PCR-genotyping analysis of hrs in embryonic cell lines derived from wild-type (ϩ/ϩ), heterozygous (floxP/ϩ), and homozygous (floxP/floxP) embryos. The lengths of the PCR fragments are 330 and 160 bp for the wild-type and mutant alleles, respectively. D, Cre-mediated deletion of neo r . An immortalized fibroblastoid cell line with the floxP/floxP genotype was treated with Cre, and PCR analyses of the neo r and hrs ␦ gene were performed. E, immunoblot analyses of the Hrs protein in an immortalized knock-out (␦/␦) and a control cell line (MEFw). F, semiquantitative RT-PCR analyses of the hrs and ␤-actin expression in HRSd and MEFw cells. Templates were prepared to make a series of 3-fold dilutions.
both Hrs and the smaller protein in HRSd cells was confirmed with anti-Hrs antisera (Fig. 1E). We then performed RT-PCR experiments to determine if any mRNA containing the intact coding region within hrs could be detected. However, we could not detect any mRNA for hrs in HRSd cells (Fig. 1F). We, therefore, concluded that HRSd cells display a null phenotype. A gross microscopic analysis showed both the HRSd and MEFw cell lines displayed a similar flat cell shape (data not shown).
Stability of STAM1 and STAM2 Is Dependent on Hrs-Our previous experiments indicated that STAM1 and STAM2 bind directly to Hrs and are involved in the vacuolar membrane transport machinery (4,32). To analyze the function of Hrs and its subdomains, we established HRSd sublines stably expressing wild-type (HRSw) or the dC2 (HRSdC2), dFYVE (HRSd-FYVE), or dM (HRSdM) mutants of Hrs (Fig. 2, A and B). We first examined the level of the Hrs-associated proteins STAM1 and STAM2 in the mutant and MEFw lines. Although significant STAM1 protein was detected in the HRSw and MEFw cells, only minimal STAM1 was detected in HRSd cells (Fig.  2C). Interestingly, HRSdFYVE cells displayed a level of STAM1 similar to the level expressed by HRSw and MEFw cells, whereas the dC2 and dM mutants exhibited severely reduced levels of STAM1 (Fig. 2C). A similarly defective level of STAM2 was observed in HRSd, HRSdC2, and HRSdM cells (Fig. 2C). These results suggest that Hrs contributes to the levels of both STAM1 and STAM2 and that the Hrs region responsible for this contribution lies within amino acids 451-574, which includes the CC2 domain.
We next investigated the mechanism underlying the severely decreased levels of STAM1 and STAM2 in HRSd, HRSdC2, and HRSdM. To determine whether the suppression of STAM1/ STAM2 was controlled at the transcriptional or post-transcriptional level, mRNAs prepared from HRSd and its sublines were subjected to Northern blotting. The results clearly indicated that there was little if any difference in the mRNA expression of STAM1 or STAM2 among these cell lines (Fig. 2D). Taken together, these results indicate that the decrease in STAM levels in the dC2 and dM Hrs mutants was post-transcriptional and was due to protein instability.
Degradation of STAM1 Is Regulated by the STAM Interaction Domain of Hrs-Since Hrs is involved in the vacuolar membrane transport machinery, we speculated that Hrs might contribute to the stability of the STAM1 protein. Therefore, we tested the possibility that Hrs is critical in regulating STAM1 degradation. To examine this possibility, HRSd cells were transiently transfected with expression vectors encoding epitopetagged STAM1 (STAM1-V5) and Hrs. We monitored STAM1 degradation in the presence of cycloheximide (to stop de novo protein synthesis) by Western blotting. Although the initial level of STAM1 decreased significantly in the presence of Hrs, the degradation rate accelerated significantly in its absence (Fig. 3A). A similar accelerated degradation in the absence of Hrs was detected when STAM2 was tested (data not shown).
To further clarify the rate of STAM1 degradation, we performed a pulse-chase analysis of STAM1 protein expression in the presence or absence of Hrs. Hrs-deficient HRSd cells were co-transfected with a STAM1 expression plasmid along with wild-type Hrs or control vectors, and the cells were then metabolically radiolabeled. A similar level of radiolabeled STAM1 protein was detected in the presence or absence of Hrs just after the labeling period (time 0), suggesting comparable protein synthesis for STAM1 irrespective of Hrs (Fig. 3B). In the presence of Hrs, the radiolabeled STAM1 was slightly increased at the initial 2-h time point after the chase, probably reflecting the ongoing de novo protein synthesis from the incorporated amino acid pool, as previously reported (33,34). At 8 h after the chase, STAM1 protein decreased to 30% of the initial level; in contrast, in the presence of Hrs, 80% of the initial amount of STAM1 was observed. We also investigated the effect of various Hrs subdomains on the stability of STAM1. After transient transfection with the STAM1 and Hrs expression constructs, irrespective of whether wild-type or mutant Hrs was used, ϳ50% of the cells co-expressed STAM1 and Hrs, with a negligible portion expressing only STAM1 (data not shown). Based on this observation, we concluded that this assay was useful for monitoring the effect of Hrs on STAM1 instability. In the absence of Hrs, a relatively fast decrease of STAM1 to 40% of its initial level was observed 6 h after the chase, which was similar to its rate of decrease in the presence of the Hrs-dM mutant (Fig. 3C). Consistent with our previous experiment, wild-type Hrs resulted in 80% of the radiolabeled STAM1 remaining after the chase. To determine if the UIM domain of Hrs was involved in the degradation of STAM1, we introduced a point mutation into Hrs, in which the leucine residue at amino acid position 265 was replaced by glutamic acid, resulting in this failure of the construct to bind ubiquitin (Hrs-mUIM, Fig. 2A). However, the degradation rate in the Lysates from the Hrs-deficient cell line (HRSd) and its sublines were used for immunoprecipitation and immunoblotting with an anti-STAM1 monoclonal antibody (TUS-1) and an anti-STAM2 monoclonal antibody (ST2-2), respectively. D, mRNA expression of STAM1 and STAM2 in HRSd and its sublines. Total RNAs were used for the detection of STAM1 (stam1) and STAM2 (stam2) mRNA. mRNA levels of the glyceraldehyde-3-phosphate dehydrogenase gene (gapdh) were monitored as the control.
presence of Hrs-mUIM was similar to the result with wild-type Hrs (Fig. 3C). These results indicate that Hrs contributes to the protein stability of STAM1, for which the STAM1-association domain of Hrs is required, but the UIM domain of Hrs is not.
STAM1 Stability Is Mediated via Hrs Binding, and the UIM Domain of STAM1 Is Required for Its Stabilization-We next examined which STAM1 subdomains were responsible for the Hrs-mediated stabilization of STAM1. For this study, we used two V5-tagged STAM1 mutants, STAM1-DIT and STAM1-mUIM. STAM1-DIT lacked ITAM, an Hrs binding domain (16), and STAM1-mUIM carried three amino acid substitutions from Leu to Ala at amino acids 176, 182, and 184 in the UIM domain; these substitutions abrogated its ubiquitin binding ability (Fig.  4A). 293T cells were transiently co-transfected with wild-type Hrs along with wild-type STAM1 or its mutants. Whereas the level of metabolically radio-labeled STAM1-DIT gradually decreased to about 25% of its initial level during a 6-h chase, wild-type STAM1 and STAM1-mUIM showed less degradation, decreasing only to 80% by 4 h after the chase and then gradually to 70% by 6 h (Fig. 4B). Since the DIT mutant could not bind wild-type Hrs, these results suggest that a direct interac-tion between Hrs and STAM1 is required for the stability of STAM1.
To further examine the effect of Hrs on the degradation/ stability of STAM1, we used the HRSd cells. Compared with wild-type STAM1, the amount of STAM1-mUIM was significantly greater, even in the absence of Hrs (Fig. 4C). In addition, the expression levels of both STAM1 and STAM1-mUIM were much higher in the presence of Hrs than in its absence (Fig.  4C). These results suggest that the UIM domain of STAM1 is essential for the stability of STAM1, and the co-expression of Hrs significantly augments the STAM1 protein stability.
STAM1 Is a Polyubiquitinated Protein and Is Degraded in a UIM-dependent Manner by Proteasomes-Our results to this point indicated that STAM1 could be degraded by either of the two main protein degradation pathways, i.e. the lysosomal or ubiquitin-proteasomal pathways (35). To define the degradation pathway of STAM1, we first examined the effects of lactacystin, a potent and selective proteasome inhibitor, and of a lysosome inhibitor, E-64-d, on the stability of STAM1 in the absence of Hrs. HRSd cells were transiently co-transfected with wild-type STAM1 and HA-tagged ubiquitin in the presence or absence of lactacystin. Incubation with lactacystin in-

FIG. 3. Degradation of STAM1 is controlled by Hrs.
A, HRSd cells were transfected with the V5-tagged STAM1 expression vector together with Hrs or a control vector. After 48 h in culture, the cells were serum-starved for 1 h and then incubated with 25 g/ml cycloheximide. The cell lysates were immunoblotted. STAM1-V5 (arrowhead) and ␣-tubulin (tubulin) are indicated. The signal density was analyzed with a FluoroImager, and the relative densities to the initial STAM1 expression level were determined. B, HRSd cells were transfected with a V5-tagged STAM1-expression vector and the Hrs or control vectors. After 48 h in culture, the cells were radiolabeled with 100 Ci/ml [ 35 S]methionine/cysteine for 30 min at 37°C and further cultured for the indicated times. After extensive washes, the STAM1 protein was recovered by immunoprecipitation with the anti-V5 antibody and monitored for radioactivity. Signals from the V5-tagged STAM1 are indicated by arrowheads. The amounts of radioactivity were determined. C, effect of wild-type Hrs and Hrs mutants on the degradation of STAM1 in HRSd cells. HRSd cells were transfected with STAM1-V5 or the control vector and radiolabeled as described in B. Signals from V5tagged STAM1 are indicated by arrowheads. Radioactivity from STAM1 at the 0-and 6-h time points was measured, and the relative protein amount detected by the Imager is shown. creased the amount of wild-type STAM1 protein (Fig. 5A). Interestingly, STAM1-mUIM was significantly more stable than wild-type STAM1 even in the absence of lactacystin, and its stability increased profoundly in the presence of lactacystin. On the other hand, E-64-d showed little if any effect on STAM1 stability.
Since protein degradation by proteasomes is regulated by protein ubiquitination, we asked whether STAM1 was ubiquitinated. 293T cells were transiently co-transfected with wildtype STAM1 and HA-tagged ubiquitin. The co-transfected cells were subjected to immunoprecipitation with the anti-V5 antibody, and the ubiquitins in the precipitates were analyzed (Fig.  5B). Although the V5 tag contains one lysine residue, our preliminary experiments suggested that this lysine receives little if any modification by ubiquitin (data not shown). As expected, ubiquitin clearly co-precipitated with STAM1. Furthermore, the prior treatment of cells with lactacystin significantly increased the degree of ubiquitination. The protein expression levels of wild-type STAM1 and STAM1-mUIM did not differ significantly in our hands, probably because of the expression of endogenous Hrs in the 293T cells. Collectively, these results suggest that STAM1 is a polyubiquitinated protein, and mutations of its UIM domain severely abolish the ubiquitination.
STAM1⅐Hrs Complex Is Responsible for Ubiquitin Accumulation on/within the Endosomes-To clarify whether the STAMs and Hrs contribute to the accumulation of cellular ubiquitinated proteins, we co-transfected HRSd cells with the STAM1 and Hrs expression vectors and FLAG-tagged ubiquitin. Whole-cell lysates were then tested for the intracellular accumulation of ubiquitinated proteins. In the absence of Hrs, the introduction of wild-type STAM1 and DIT resulted in little if any increase in the amount of ubiquitin accumulation (Fig.  6A). Transfection of STAM1-mUIM slightly increased the ubiquitination. However, the introduction of Hrs caused a marked accumulation of ubiquitinated proteins irrespective of the presence of STAM1. Unexpectedly, a slightly smaller amount of ubiquitins accumulated in the presence of Hrs-mUIM than in the presence of wild-type Hrs even though Hrs-mUIM seemed more stable than wild-type Hrs (Fig. 6A, middle panel). These results indicate that Hrs is responsible for the intracellular accumulation of ubiquitinated proteins, independent of STAM1.
We next asked whether the presence of STAM1 and Hrs could affect the endosomal localization and accumulation of ubiquitinated proteins. Consistent with our earlier results, when HRSd cells were transfected with STAM1 and ubiquitin, only faint STAM1 labeling was observed (Fig. 6B, a). In contrast, the co-introduction of wild-type Hrs and STAM1 resulted in the colocalization of these proteins (Fig. 6B, d, e, and g). Interestingly, heavy staining of ubiquitinated proteins overlapped with the staining of the STAM1⅐Hrs complex (Fig. 6B, f  and g). Surprisingly, the introduction of Hrs-mUIM and STAM1 into HRSd cells induced enlarged endosomes in which the Hrs-mUIM and STAM1 immunoreactivity overlapped that of the ubiquitins (Fig. 6B, h-k).
Next, we examined whether the UIM of STAM1 played a role in the endosomal localization of these proteins and/or the endosomal phenotypes. When HRSd cells were transfected with STAM1-mUIM and ubiquitin, a slightly but significantly stronger signal for the STAM1-mUIM than for wild-type STAM1 was detected in the cytoplasm with a marginal increase in cytoplasmic ubiquitin expression (Fig. 6B, l-lЈ). In the presence of wild-type Hrs, the expression of STAM1-mUIM resulted in almost normal-sized endosomes that were co-localized with ubiquitins (Fig. 6B, o-r). However, when both STAM1 and Hrs were mutated at their UIM domains the HRSd cells clearly showed enlarged vesicles that co-localized with ubiquitins (Fig.  6B, s-v). These results suggest that Hrs and STAM1 co-localized with ubiquitinated proteins, and the intact UIM domain of Hrs but not of STAM1 is required for normal endosomal development. Collectively, these results suggest that Hrs is indispensable for the intracellular accumulation and endosomal localization of ubiquitins.
The UIM Domain of Hrs Is Essential for Enlarged Endosome Formation-Enlarged endosomes in the absence of Hrs have been reported (31,32). We also examined the phenotypes of early endosomes in HRSd sublines expressing wild-type Hrs or its mutants using the co-expression of an endosomal marker protein, Eps15. By fluorescence microscopy, a scattered dot-like expression pattern of Eps15 was observed within the cytoplasm of HRSw cells that corresponded to the normal early endosome size and structure (Fig. 7A, a). In contrast, HRSd cells con-tained enlarged endosomes (Fig. 7A, e), and the two Hrs mutant cell lines Hrs-dC2 and -dFYVE exhibited a significantly enlarged endosome phenotype; the endosomes in these mutants were, however, smaller than those observed in HRSd cells (Fig. 7A, b and c). Cells expressing the Hrs mutant HRSdM, which fails to interact with either STAM1 or STAM2, also manifested modestly enlarged endosomes that lacked central transparency (Fig. 7A, d). Unexpectedly, the transient introduction of the Hrs-mUIM mutant also conferred an enlarged endosome phenotype (Fig. 7B, m). Thus, enlarged endosomes were observed in the absence of Hrs and in the presence of the Hrs-dC2, -dFYVE, -dM, and -mUIM mutants.
We next asked whether the subcellular localization of STAMs was altered in the presence or absence of wild-type or mutant Hrs. To examine the localization of Hrs and STAM1, we introduced Hrs and its mutants along with the STAM1 expression vector into HRSd cells (Fig. 7B). Consistent with the data in Fig. 7A, spot-like cytoplasmic expression of the wild-type Hrs and STAM1 was observed with nearly complete co-localization (Fig. 7B, a-c). Similar results were obtained when the STAM2 expression vector was introduced into HRSd cells (data not shown). Strikingly, when the Hrs-dC2 and -dM mutants were introduced, the STAM1 and Hrs mutants localized independently; Hrs localized to the enlarged endosomes, whereas STAM1 was diffusely expressed throughout the cytoplasm (Fig.  7B, d-f and j-l). The STAM1 immunoreactivity was significantly reduced when STAM1 was co-introduced with the Hrs-dC2 and Hrs-dM mutants but not with wild-type Hrs (Fig. 7B,  e and k). Only long exposure images clearly showed the localization of STAM1 in the cytoplasm in the presence of these mutants (Fig. 7B, eЈ and kЈ). Interestingly, the Hrs-dFYVE mutant completely colocalized with STAM1, although the endosomes were enlarged and differed significantly from those seen with wild-type Hrs (Fig. 7B, g-i). These results indicate that the endosomal localization of STAM1 was dependent on its ability to bind Hrs because when it was cotransfected with Hrs mutants lacking STAM1 binding capability (HRSdC2 and HRSdM), STAM1 was expressed diffusely in the cytoplasm.

DISCUSSION
Stability of STAMs Is Dependent on Hrs-In this study we established an Hrs-deficient cell line and demonstrated the critical involvement of Hrs in the stability of both the STAM1 and STAM2 proteins. Whereas the Hrs-deficient HRSd cells showed a profound decrease in the level of STAM1 and STAM2, the stable introduction of Hrs fully restored both proteins to levels similar to those in normal MEFw cells. The drastic reduction in STAM1 and STAM2 could have been due to several possible factors, which are decreased transcription or translation, the stability of the mRNAs, or decreased protein stability. Our Northern blot analyses clearly demonstrated that the amount of STAM1 and STAM2 mRNA remained unchanged irrespective of the presence or absence of Hrs, indicating that instability of the STAM proteins accounted for the reduced levels. Similarly, transient transfection of the HRSd cells showed that the stability of both STAM1 and STAM2 was impaired. We, therefore, conclude that Hrs deficiency is likely to cause the STAM1 and STAM2 proteins to be unstable, and both proteins are therefore rapidly degraded so as to be undetectable on our regular Western blots.
Previous work by us indicated that both STAM1 and STAM2 firmly associate with Hrs through the CC2 domain of Hrs and the ITAM motif of STAM1 and STAM2 (4). Hrs probably binds the STAMs at a ratio of one to one, given that we observed only STAM1-Hrs and STAM2-Hrs heterodimeric complexes, but not a STAM1, STAM2, and Hrs trimeric complex (32). Here, neither STAM1 nor STAM2 was detected in HRSdC2 or HRSdM cells, suggesting that the region containing the CC2 domain (amino acid 348 -573) is necessary for the stability of the STAMs. Since the STAM proteins were detected in the Hrs-dFYVE mutant cell line, the FYVE domain of Hrs, which is required for Hrs to bind the inner surface of the cytoplasmic membrane through phosphatidylinositol 3,4,5-trisphosphate, is not necessary for the stability of the STAMs. The expression levels of Hrs in the Hrs-dC2 and Hrs-dM mutant cell lines, however, were markedly reduced and accompanied by the loss of STAM1 and STAM2. These results suggest that Hrs stabilizes the STAMs by directly binding to them.
The ITAM domain of STAM1 and STAM2 is required for binding to Hrs, although an additional region, called the SSM domain (in human STAM1, amino acids 297-320 and human STAM2 amino acids 286 -309) may possess auxiliary Hrs binding functions (36). Given that the Hrs-dC2 mutation resulted in the degradation of STAM1, we expected the STAM1-DIT mutant, which is incapable of binding to Hrs, to be unstable. Consistent with this, the degradation of the STAM1-DIT mutant was significantly faster than that of wild-type STAM1. Our previous report using STAM1/STAM2 double-deficient cells clearly demonstrated that Hrs protein degradation is not controlled by the STAMs (32). Furthermore, a recent report showed that an Hrs knockdown induces Tsg101 instability, suggesting that Hrs also helps stabilize Tsg101. Since Tsg101 was recently shown to interact with Hrs and is now categorized as an ESCRT-I molecule, Hrs may function to stabilize other proteins as well (37). Taken together, our results show that STAM1 is stabilized by an association with Hrs, and Hrs mutants that disrupt STAM1 binding result in STAM1 degradation.
How does the deficiency of Hrs mediate STAM1 and STAM2 protein stability? In general, there are two major protein degradation systems, the ubiquitin-proteasome and the lysosome/ vacuole pathways. These two pathways seem to regulate protein fate differentially, so that proteins with short half-lives tend to be degraded by the ubiquitin-proteasome pathway, whereas those with long half-lives are more likely to be digested by the lysosome/vacuole pathway (35). Our present results clearly indicated that the STAM1 degradation was proteasome-dependent. This was unexpected, because both STAMs and Hrs were previously shown to be resident endosomal proteins and believed to function in the endosome maturation machinery. A possible scenario, taken from recent data, is that the ESCRT complex could mediate protein degradation through a vesicular transport function (1). In this hypothesis protein ubiquitination has an indispensable function for transport. According to our present data, as long as Hrs is present ubiquitinated proteins can accumulate within the endosomes. Since the STAMs can hardly be detected in the absence of Hrs, we cannot fully exclude the possibility that STAM1 plays a role in the accumulation of ubiquitinated proteins within endosomes. Nevertheless, our data provide further evidence supporting the idea that ESCRT proteins mediate the accumulation of ubiquitinated proteins into endosomes, where the proteins are destined to be degraded in a lysosome-dependent manner.
Both the STAMs and Hrs possess a UIM domain, which was first identified as a polyubiquitin-binding site in the S5a subunit of the 26 S proteasome (38). The results of our mutational analysis suggest that the UIM domain of Hrs is not involved in the degradation of STAM1. However, the UIM domain of STAM1 is essential for its stability. A STAM1 mutant with point mutations within the UIM (STAM1-mUIM) was significantly more stable than wild-type STAM1 (STAM1-UIM), indicating that STAM1 controls its own degradation by the pro-teasome pathway. At present we have at least three hypotheses to explain the STAM1-UIM function. First, the UIM of STAM1 may contain one or more major polyubiquitinated sites. The UIM domain of STAM1 contains three lysine residues (Lys-171, Lys-178, and Lys-185) that are potentially polyubiquitinated and destined to be recognized by the 26 S proteasome for degradation. Indeed, of the three lysine residues, the first two are evolutionarily conserved from Drosophila to human. Although lysine 171 is a potential protein modification site for both ubiquitination and sumoylation, it is unlikely to play an important role in this degradation, since our UIM mutant conserves all three lysine residues (39). Second, the UIM domain of STAM1 may interact with other polyubiquitinated proteins, which will be recognized and degraded by the proteasome. However, it is unknown whether the degradation of any given protein can be mediated via interactions with another ubiquitinated protein(s). Recently, Hrs was shown to interact with AIP4, a HECT domain-containing E3 ubiquitin ligase (40). We, therefore, propose a third mechanism whereby the UIM domain of STAM1 interacts with an E3 enzyme(s) either alone or as a complex and, thus, allows STAMs to be ubiquitinated and thereby destined for degradation. Hrs is ubiquitinated by AIP4 and mediates the sorting of chemokine receptor CXCR4 into endosomes (40). In agreement with this, we also detected a significant increase in the level of the Hrs-mUIM protein compared with wild-type Hrs, suggesting its stabilization through the UIM via interactions with one or more ubiquitinase enzymes. We provide direct evidence that the UIM domain of Hrs (or some other UIM-interaction protein(s)) is not required for STAM1 degradation. To further analyze the mechanism underlying the degradation of STAMs, it will be necessary to identify an E3 enzyme that couples with the STAMs.
Hrs Causes Accumulation of Ubiquitinated Proteins-Previous studies have shown that the UIM domains of mammalian Hrs and yeast Vps27 are both essential for sorting ubiquitinated membrane proteins into the degradation pathway (2,12,41). Consistent with this, Hrs binds to monoubiquitinated proteins through its UIM domain (2,12,41). However, little is known about the role Hrs plays in intracellular polyubiquitination. The introduction of Hrs into HRSd cells caused the marked elevation of ubiquitinated proteins irrespective of the presence of STAM1 expression vectors. At present, the mechanism underlying the ubiquitinated protein accumulation is unclear. It is clear that the aberrantly enlarged endosomes do not account for the ubiquitin accumulation within whole-cell extracts because the class E compartment does not show any accumulation of ubiquitinated proteins in the absence of Hrs but, rather, manifests a significant decrease in ubiquitinated proteins. One possible explanation is that Hrs may control the activity of an E3 ubiquitin ligase. Cargo proteins captured by Hrs could be stabilized and/or further receive ubiquitination, probably through E3 group ubiquitinases within the Hrs complex. However, since the Hrs-mUIM mutant showed only a marginal decrease in intracellular ubiquitination, it may be that other E3 ligases are involved in ubiquitin accumulation. Since STAM1, STAM2, and Tsg101, which all belong to the ESCRT complex, possess additional ubiquitin-binding motifs, ESCRT molecules including Hrs may function as a complex to increase protein ubiquitination and/or protein stabilization. Conversely, our present data suggest that at least some ubiquitinated proteins accumulate within the endosomes. Therefore, ubiquitinated proteins destined to be degraded by the late endosome/lysosome system may be modified by Hrs. Interestingly, the overexpression of STAM2 in the presence of Hrs was recently reported to significantly increase the levels of intracellular ubiquitinated proteins (36,42). Our present data using Hrs-defective cells clearly indicate that the overexpression of STAM1 does not lead to an increase in ubiquitinated proteins. This difference might be due to the effects of excessive amounts of STAM2 and/or the presence of native Hrs in their experiment. Although further experiments are needed to clarify the mechanism, we propose that dysregulation of the STAM⅐Hrs complex may affect intracellular ubiquitination accumulation and the degradation machinery.
Enlarged Endosome Phenotype Is Dependent on Hrs-Recent studies by us (32) and others (19,42) have provided evidence that the STAMs and Hrs co-localize in early endosomes. An earlier report demonstrated that Hrs can be co-immunoprecipitated with Eps15, an endosomal marker protein (43). Hrs has also been shown to bind Eps15 through the N-terminal half of Hrs (44). The direct binding of STAM1 with Eps15 was also reported (43). Since Eps15, an early endosomal membrane protein, is involved in the endocytosis of membrane-associated receptors via its binding to AP-2 (45), both the STAMs and Hrs are also localized to the same vesicles.
The lack of Hrs triggers aberrant intracellular vesicle formation called class E compartments. This was first demonstrated in genetic analyses with yeast mutants manifesting similar large vesicular compartments (2). In this report, we also clearly demonstrated abnormally enlarged endosomes. This "mammalian class E compartment" was present in several Hrs mutants we tested that were introduced into HRSd cells. Although the degree of endosomal enlargement differed with the mutation, our present data suggest that the class E phenotype is dependent on the C-terminal, FYVE, or CC2 domain. We therefore speculate that intact Hrs is required for endosomes to have a normal appearance. Interestingly, a point mutation (L265E) within the UIM domain of Hrs abrogates the normal endosomal architecture. It is possible that an unidentified UIM interaction molecule(s) functions in the normal endosomal phenotype. Unlike the UIM of Hrs, point mutations within the UIM of STAM1 did not result in enlarged endosomes. The class E compartment phenotype can be determined, therefore, by the UIM domain of Hrs but not by that of the STAMs. It is of note that Hrs not only supported the protein stability of STAMs but also determined their intracellular localization; the absence of Hrs totally changed the localization of STAMs from the endosomes to the cytoplasm. On the other hand, Hrs, whether wild type or mutated, always localized to the endosomes. To understand how this co-localization occurs, further work will be required to identify the region(s) within Hrs that is responsible for its localization.
In Vivo Function of Hrs-We provide further evidence that Hrs is required for normal mouse development. Although our attempt to establish a conditional targeting mouse has been unsuccessful, probably because of the impaired function of the "floxed" exon 6, experiments using established cell lines carrying the null allele provided the unexpected finding that the Hrs knock-out leads to the functional extinction of the STAM1 and STAM2 proteins. Incidentally, the STAM1/ STAM2 double-knock-out mice we have generated die in utero around E11.5, because the ventral fold fails to form. 2 It is not surprising that Hrs knock-out mice with no detectable STAM1 or STAM2 protein manifest an embryonic lethal phenotype as well (26,31). Loss of Hrs and/or STAM function results in mammalian class E phenotypes, which are similar to the class E phenotypes of yeast Vps27 and Hse1. Paradoxically, an excess of Hrs protein seems to be toxic to cells, because the overexpression of Hrs leads to a similar enlargement of endosomes (13). These results suggest that the appropriate control of Hrs and STAM1/STAM2 may be critical for normal cell morphology and function. Given that Hrs is involved in the accumulation of intracellular ubiquitinated proteins, our next step should be to clarify the functional relationship between the ESCRT complex and the ubiquitination machinery.