![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Previous Article | Next Article 
Journal of Virology, October 2006, p. 9465-9480, Vol. 80, No. 19
0022-538X/06/$08.00+0 doi:10.1128/JVI.01049-06
Copyright © 2006, American Society for Microbiology. All Rights Reserved.
Departments of Biochemistry,1 Pathology, University of Utah, Salt Lake City, Utah 841122
Received 22 May 2006/ Accepted 8 July 2006
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Genetic screens in Saccharomyces cerevisiae defined the basic machinery of MVB vesicle formation and identified 17 soluble yeast proteins that are essential for MVB biogenesis (the "class E" proteins) (40, 59). Complementary biochemical analyses revealed that MVB vesicle formation proceeds through an ordered pathway in which a series of soluble class E complexes, termed ESCRT-I (39), ESCRT-II (4), and ESCRT-III (3), are sequentially recruited to endosomal membranes, where they function in vesicle formation (Fig. 1A). Although the processes of vesicle formation and cargo incorporation are not yet understood in mechanistic detail, ESCRT-I and -II appear to function primarily as adaptors that recognize protein cargoes and help recruit ESCRT-III, which in turn appears to function more directly in protein sorting and vesicle formation.
|
The human and yeast ESCRT-II complexes are likely to be related structurally, because mammalian ESCRT-II can be isolated as a stable, soluble complex composed of three proteins that share 19 to 36% pairwise identity with their S. cerevisiae homologs (excluding the N-terminal region of Vps36p/EAP45) (37, 61). We have therefore used the crystal structure of the S. cerevisiae ESCRT-II core as a working model for the human ESCRT-II complex (Fig. 1B) (see Materials and Methods). Mammalian ESCRT-II functions are less well studied than those of the yeast complex, but the mammalian EAP20 and EAP45 proteins do colocalize with ESCRT-III components and with ubiquitylated proteins on endosomal membranes (67, 88). Like their yeast counterparts, EAP20 can bind the ESCRT-III protein, CHMP6 (46, 83, 88), and the N-terminal GLUE domain of murine EAP45 was recently shown to bind ubiquitin (67). These observations all support the idea that the mammalian and S. cerevisiae ESCRT-II complexes are functional homologs. Nevertheless, there must also be important differences between human and yeast ESCRT-II proteins. For example, the mammalian complex lacks both NZF domains and therefore must interact differently with both ESCRT-I and ubiquitin. Moreover, ESCRT-II and CHMP6 overexpression does not inhibit HIV type 1 (HIV-1) budding (46), and it was recently reported that ESCRT-II is not required for lysosomal degradation of major histocompatibility complex I (MHC-I) or epidermal growth factor receptor (EGFR), indicating that the complex is not absolutely required for membrane protein sorting or MVB vesicle formation (11). Indeed, several groups have proposed that ESCRT-II complexes in other species may perform functions that appear to be unrelated to MVB protein sorting. These include a role for mammalian ESCRT-II in transcriptional elongation (37, 61, 65) and a role for the Schizosaccharomyces pombe ESCRT-II protein EAP30 in meiotic spindle pole body amplification (35). The relationship between MVB protein sorting and these other putative ESCRT-II activities, if any, is not yet clear.
During HIV-1 assembly, human class E proteins are recruited by two short sequence motifs (termed "late domains") within the C-terminal p6 region of the Gag polyprotein (reviewed in references 15, 19, 22, 47, 48, 52, 70, and 79). The first p6gag late domain is a PTAP tetrapeptide motif that binds and recruits TSG101 (tumor susceptibility gene 101; yeast Vps23p). TSG101 is the central protein of the human ESCRT-I complex, and there is good evidence that the entire ESCRT-I complex functions directly in HIV-1 release (15, 19, 22, 24, 48). The second late domain within p6gag is a YPXnL motif (where Xn can vary in identity and length) that binds and recruits ALIX/AIP1 (yeast Bro1p) (69, 80). ALIX functions late in MVB biogenesis (56) and binds both ESCRT-I (83) and ESCRT-III (41) proteins but is not a permanent subunit of any of the three discrete ESCRT complexes. Like TSG101, ALIX functions directly in virus release, although disruptions of the PTAP-TSG101 interaction generally inhibit HIV-1 budding to a greater degree than do disruptions of the ALIX-YPXnL interaction (16, 46). As noted above, there are also indications that HIV-1 budding requires the activities of the late-acting ESCRT-III, LIP5, and VPS4 class E proteins (22, 46, 69, 83, 85). Specifically, HIV-1 release is dominantly inhibited by overexpression of VPS4 ATPase mutants and ESCRT-III proteins with large N- or C-terminal extensions (22, 46, 69, 83). HIV-1 release is also impaired by small interfering RNA (siRNA) depletion of the VPS4 binding protein LIP5 (85). It is important to note, however, that these experiments do not definitively establish a direct role for these proteins in HIV-1 budding, because the inhibitory effects could also arise from sequestration of the entire ESCRT machinery at aberrant endosomal compartments (called "class E" compartments).
In summary, HIV-1 uses the ESCRT-I complex to bud from cells and also requires the activities of downstream ESCRT-III and VPS4-LIP5 complexes. In S. cerevisiae, ESCRT-II physically bridges the ESCRT-I and ESCRT-III complexes and is essential for MVB protein sorting and vesicle formation (4). The mammalian ESCRT-II function appears to be more complex, however, and we therefore undertook this study with the goals of characterizing the biochemical properties of human ESCRT-II and testing its role in HIV release.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Plasmids. Genes for human EAP20, EAP30, EAP45, and CHMP6 used in yeast two-hybrid and biochemical experiments were amplified from expressed sequence tag clones as described previously (83). A complete list of all constructs used in this study is provided in Table S1 in the supplemental material. Genes and fragments for yeast two-hybrid experiments were cloned into pGADT7 (activation domain) and pGBKT7 (binding domain) vectors (Clontech) using 5' NdeI and 3' BamHI or BglII cloning sites. Recombinant proteins used in biochemical experiments were expressed as glutathione S-transferase (GST) C-terminal fusions from a modified pGEX vector (GE Biosciences). This vector (pGEX2T-TEV) contained a tobacco etch virus (TEV) protease cleavage site and 5' NdeI and 3' BamHI or BglII cloning sites following the GST gene. Expression constructs for deletion mutants were created by amplifying and subcloning the desired fragments or by introducing a stop codon using Quickchange mutagenesis (Stratagene). For mammalian expression, ESCRT genes were cloned as EcoRI-BamHI fragments into pcDNA3.1(–)MycHis in frame with the C-terminal Myc tag, and with a Kozak sequence, ACC, inserted between the EcoRI site and the ATG start codon.
Preparation of recombinant EAP20.
GST-EAP20 was expressed in Escherichia coli BL21(DE3) transformed with WISP05-68 (see Table S1 in the supplemental material) by induction in mid-log phase (0.4 mM IPTG [isopropyl-ß-D-thiogalactopyranoside]) and allowing the protein to accumulate for 4 hours at 23°C. The cells were lysed by lysozyme treatment (2.5 mg/ml) in 50 mM Tris, pH 8.0, 300 mM NaCl, 10% glycerol, 1 mM dithiothreitol (DTT), followed by sonication and centrifugation (at 27,000 x g for 45 min) to remove cell debris. Soluble GST-EAP20 was bound to a glutathione-Sepharose affinity matrix (GSTPrep FF; GE Healthcare), washed, and eluted with a step gradient of 20 mM glutathione. Protein fractions were dialyzed into 25 mM Tris (pH 8.0), 150 mM NaCl, and 1 mM DTT), and the fusion protein was cleaved during dialysis with 1 mg TEV protease per 100 mg protein (24 h; 23°C). The TEV cleavage reaction mixture was adjusted to 25 mM Tris (pH 8.0), 300 mM NaCl, 10% glycerol, 1 mM DTT, and free EAP20 was separated from GST and GST-EAP20 by size exclusion chromatography (Superdex 200; GE Healthcare). This purification method typically yielded 
1.5 mg of pure protein per liter of starting culture.
Preparation of recombinant CHMP6 proteins.
GST-CHMP6 expression, affinity purification, and TEV cleavage were carried out as described for GST-EAP20. Following TEV cleavage, free GST was removed by glutathione-Sepharose affinity chromatography. The solution was then diluted six-fold to reduce the salt in the buffer to 50 mM NaCl, and the CHMP6 protein was purified to homogeneity by anion-exchange chromatography on immobilized Q-Sepharose (GE Biosciences). CHMP6 eluted at 400 mM NaCl from a linear gradient of 50 mM to 1 M NaCl in 25 mM Tris (pH 8.0), 10% glycerol, 1 mM DTT.
GST pull-down assays. GST and GST-CHMP6 fusion proteins were expressed in E. coli BL21(DE3) and induced at mid-log phase with 0.4 mM IPTG. Ten-milliliter cultures of cells expressing GST-CHMP6 were pelleted; resuspended in 2 ml of 300 mM NaCl, 0.1% NP-40, 10% glycerol, 1 mM DTT, 50 mM Tris (pH 8.0); and lysed by lysozyme treatment and sonication. Soluble GST-CHMP6 was incubated with 100 µl glutathione-Sepharose resin. Approximately 30 µM purified recombinant EAP20 was added to 2 µM (equivalent) of immobilized GST-CHMP6 and incubated for 30 min at 4°C in resuspension buffer. Unbound EAP20 was removed by washing with lysis buffer, and bound proteins were released from the resin by boiling it in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer and were analyzed on Coomassie-stained 4 to 20% SDS-PAGE gradient gels.
Yeast two-hybrid binding assays. Directed yeast two-hybrid assays were performed using the Matchmaker GAL4 Yeast Two Hybrid 3 system (Clontech). Briefly, Saccharomyces cerevisiae AH-109 was cotransformed with pGADT7 or pGBKT7 cloning vector (Clontech) containing the inserts of interest. The transformed yeast colonies were grown for 3 days at 30°C on yeast extract-peptone-dextrose plates with minus Leu, minus Trp selection. Ten to 100 colonies were pooled, resuspended in a liquid culture of Sabourand dextrose broth (minus Leu, minus Trp), selected on Sabourand dextrose broth (minus Leu, minus Trp, minus Ade, minus His) plates, and allowed to grow for 3 days.
Antibody production. Antibodies against the purified recombinant proteins EAP20 (UT461 and -462) and CHMP6 (UT463 and -464) were raised in rabbits by Covance Inc., following their 87-day protocol (http://abservices.crpinc.com/ab_sampleProtocols.aspx). UT461 and UT463 were affinity purified with the respective proteins, as described previously (83).
Biosensor binding experiments.
EAP45-ubiquitin biosensor binding experiments were performed at 20°C using a BIACORE 2000 (Biacore AB, Uppsala, Sweden) equipped with a research grade CM4 sensor chip. Approximately 5,000 response units (RU) anti-GST antibody was immobilized on all four flow cells using amine-coupling chemistry (36). GST-EAP451-229, GST-EAP451-139, or recombinant GST alone from soluble E. coli lysates was diluted in running buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM DTT, 0.2 mg/ml bovine serum albumin [BSA], and 0.005% P20) and captured on the antibody surfaces at densities of 0.5 kRU. Pure GST alone was also captured as a blank reference. Ubiquitin was purified (84), diluted in running buffer, and injected in duplicate or triplicate over the four flow cells at concentrations of 0 to 1,250 µM. Responses were collected at a rate of 2 Hz during the 30-s association and dissociation phases. Binding isotherms from the equilibrium phases were fitted to simple 1:1 binding models to determine the dissociation constant and statistical fitting errors (see Fig. 4) (53). To obtain more accurate estimates for the dissociation constant and standard errors, the interaction between GST-EAP451-139 and ubiquitin was measured in six independent experiments, yielding values of 410 ± 90 µM. Biosensor binding experiments for EAP20 and GST-CHMP6 were performed as described for the EAP45-ubiquitin interaction, except that a BIACORE 3000 instrument was used and binding was examined in a buffer containing 50 mM Tris, pH 8.0, 300 mM NaCl, 10% glycerol, 1 mM DTT, 0.2 mg/ml BSA, and 0.005% P20. Anti-GST antibody was immobilized over all four flow cells (13,000 RU/cell), and CHMP6, CHMP61-100, and CHMP6101-200 were captured to densities of 2,700 RU. Purified EAP20 was diluted in running buffer and injected in triplicate at 0.03 to 560 µM concentrations.
|
Viral infectivity assays. EAP20 and CHMP6 were silenced by transfecting 5 µl of 20 µM annealed siRNAs, together with 1 µg of pSL1180 (carrier DNA) and 9 µl Lipofectamine 2000, into 293T cells in six-well plates (22). The sense strands of the siRNAs were as follows (lowercase letters denote RNA, and uppercase letters denote DNA nucleotides [nt]): EAP20 (start site, nt 218), cgauccagauuguauuagaTT; CHMP6-1 (start site, nt 257), ccaugguucagaguauugaTT; and CHMP6-2 (start site, nt 284), agaucgaaaugaaagugauTT. TSG101 and INV (TSG101) siRNA constructs have been described previously (22). The levels of protein depletion were estimated by quantifying the intensities of bands in Western blots using NIH ImageJ software.
Viral-particle production was initiated during a second siRNA transfection (t = 24 h) by replacing the 1-µg carrier DNA either with an HIV-1 vector system or with an infectious R9 HIV-1 NL4-3 plasmid. Virus was harvested 24 h after the second transfection, and the titers were measured using single-cycle MAGIC assays (70). The HIV-1 vector system contained 0.38 µg pCMV
R8.2 (54), 0.12 µg pRSV-Rev (18), 0.12 µg pMD.G (57), and 0.38 µg pWPTS-nlsLacZ (all generous gifts from D. Trono, Swiss Federal Institute of Technology). HIV-packaged lacZ vectors were harvested 48 h later and titered on HeLaM cells as described previously (85).
Western blotting. Viral supernatants and cellular proteins from transfected 293T cells were prepared using standard procedures (82). For Western blots, virions or cytoplasmic proteins were separated by 12% SDS-PAGE, transferred to nitrocellulose (TSG101) or polyvinylidene difluoride (all others) membranes in Tris-glycine-10% MetOH buffer, blocked in 1% (TSG101) or 5% milk, and incubated with antibodies diluted in Tris-buffered saline-0.1% Tween 20 buffer (anti-TSG101), or in 1% BSA or 5% nonfat milk in Tris-buffered saline-0.1% Tween 20. Proteins were detected by enhanced chemiluminescence (Pierce) or by imaging Alexa 680-nm (Molecular Probes; 1:10,000) or IRDye 800-nm (Rockland; 1:20,000) secondary antibody on an Odyssey scanner (Li-Cor Inc.). The primary antibodies used were rabbit anti-HIV CA (UT415 made against purified NL4-3 CA protein and affinity-purified; 1:15,000); rabbit anti-HIV MA from Didier Trono, Swiss Institute of Technology (1:25,000); rabbit anti-EAP20 UT461 (1:1,000); rabbit anti-CHMP6 UT463 or UT464 (1:1,000); and mouse anti-TSG101 (GeneTex; 1:1,000).
EGFR degradation assays. 293T or HeLa cells were plated on duplicate six-well plates in Dulbecco's modified Eagle's medium-10% fetal bovine serum, and transfected with either oligofectamine or Lipofectamine 2000 and specific siRNAs (see above). During the second siRNA transfection (t = 24 h), 1 µg R9 HIV plasmid was added to one set of cells (to control for knockdown efficiency and viral production), together with the second aliquot of 5 µl siRNA. In this case, virus and cells were harvested 48 h later (t = 72 h) and analyzed by Western blotting and MAGIC infectivity assays. The other set of cells, used in EGFR assays, was transfected a second time with siRNA-oligofectamine alone. Twenty-four hours after the second transfection (t = 48 h), the cells were reseeded on polylysine (Sigma)-coated coverslips and the next day were incubated in serum-free medium for 18 h. The cells were then incubated in the presence of 1 µg/ml human EGF (Molecular Probes/Invitrogen) for 1 h (HeLa) or 2 h (293T cells); fixed in 3.7% formaldehyde-PBS, pH 7.2, at room temperature for 20 min; and incubated with mouse anti-human EGFR (1:100; Neomarkers) in 0.1% saponin-1% bovine serum albumin-PBS, pH 7.2, at room temperature for 60 min. The cells were washed extensively and incubated in Alexa 488-conjugated goat anti-mouse immunoglobulin G (1:750) (Molecular Probes/Invitrogen). The cells were imaged using an epifluorescence microscope with MagnaFire software or an Olympus FV300 confocal fluorescence microscope with Fluoview 2.0.3.9 software using 0.5-µm sections. The fluorescence intensities were quantified using NIH ImageJ software.
EGF-lysosome colocalization assays. HeLa cells were plated on duplicate six-well plates in Dulbecco's modified Eagle's medium-10% fetal bovine serum and transfected twice with either oligofectamine or Lipofectamine 2000 (t = 0 and t = 24 h) with the siRNA oligonucleotides listed above and with EAP20-2 (start site, nt 477; gcacaaggccgagaucaucTT). At t = 48 h, the cells were reseeded on polylysine (Sigma)-coated coverslips, and at t = 72 h, the cells were switched to serum-free media for 18 more hours while being incubated with 0.1 mg/ml Alexa 488-conjugated dextran (molecular weight, 10,000). The cells were then washed extensively in growth medium followed by a 4-h chase in dextran-free/serum-free medium, placed at 0°C, incubated with 1 µg/ml EGF-Alexa Fluor 555-streptavidin complex (Molecular Probes/Invitrogen) for 60 min, and washed extensively in growth medium. Finally, the cells were incubated for 30 min at 37°C to allow time for EGF internalization and trafficking. Fluorescent images were captured with an Olympus FV300 confocal fluorescence microscope with Fluoview 2.0.3.9 software using 0.5-µm sections and analyzed using NIH ImageJ software. Three to seven fields of cells were analyzed for each sample, and the EGF-lysosome localization was scored manually, with a positive score given to any cell showing significant colocalization of the EGF and dextran markers.
KK3-mediated MHC-1 downregulation. EAP20 and CHMP6 were silenced as described above with the addition of either 2 µg of pTRACER-K3 or pTRACER empty control plasmid transfected at 24 and 48 h. Thirty-six hours after the second transfection, 5 x 105 cells were incubated in suspension with anti-human HLA-ABC antibody (1 µg/µl; eBioscience) for 1 h at 4°C in a solution of 10% FCS in PBS. The cells were washed once with 1 ml of PBS and then incubated with phycoerythrin-labeled goat anti-mouse immunoglobulin (1 µg/µl; BD Biosciences) in 10% FCS-PBS, followed by an additional wash with PBS. The cells were pelleted by centrifugation and resuspended in 1 ml PBS, and surface HLA-ABC levels were analyzed using the FACScan system (BD Biosciences).
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
ESCRT-I-ESCRT-II interactions. Two ESCRT-II proteins, EAP30 and EAP45, also interacted with TSG101, the central component of the ESCRT-I complex. Further mapping studies revealed that a fragment corresponding to the C-terminal region of TSG101 (termed the steadiness box [20, 42, 73]), was necessary and sufficient for both ESCRT-II protein interactions (Fig. 2A). Conversely, full-length TSG101 interacted with EAP301-173 and EAP45114-229, but not with fragments outside of these regions (Fig. 2B). Thus, the TSG101 interaction site on EAP45 maps to the linker region between the GLUE domain and the first WH repeat. The TSG101 interaction site on EAP30 spans both the long N-terminal helical extension and the WH repeat. TSG101 also interacted, albeit less robustly, with an EAP30 fragment missing the first 40 amino acids that form the segment of the N-terminal helix that extends beyond the globular body of EAP30. Hence, the full EAP30 N-terminal helical extension is not strictly required for TSG101 binding, although this helix may contribute to the interaction, particularly as the full helix is predicted to extend to EAP30 residue 55.
|
ESCRT-II-ESCRT-III interactions. In yeast, ESCRT-II recruits the downstream ESCRT-III complex through direct interactions between the ESCRT-II protein Vps25p (EAP20) and the ESCRT-III component Vps20p (CHMP6) (human homologs are in parentheses) (3, 74). Similarly, the human EAP20 and CHMP6 proteins also interacted in our two-hybrid survey of human class E proteins (Fig. 1C and 3A). Yeast two-hybrid mapping studies revealed that this interaction was mediated by the N-terminal half of CHMP6 binding to the C-terminal half of EAP20 (Fig. 3A), consistent with earlier reports (74, 88). The two C-terminal domains of EAP20 are exposed at the ends of the Y-shaped ESCRT-II complex, where they would be ideally positioned to nucleate ESCRT-III deposition through bivalent binding interactions (Fig. 1B).
|
EAP45-ubiquitin interactions. The yeast ESCRT-II complex helps sort ubiquitylated protein cargoes and binds ubiquitin through the second NZF motif (NZF-C) within the N-terminal GLUE domain of Vps36p (1). Mammalian EAP45 proteins lack NZF motifs, however, and it has therefore been unclear whether they could still interact with ubiquitylated protein cargoes. This puzzle has recently been resolved by Slagsvold and colleagues, who showed that a minimal N-terminal fragment of murine EAP45 spanning residues 1 to 139 binds directly to ubiquitin (KD = 460 µM) and also binds phosphoinositides (67).
We independently tested for ubiquitin binding activity in the human EAP45 protein using the entire region of human EAP45 that is missing from the crystal structure of yeast Vps36p (EAP451-229). Biosensor binding analyses revealed that ubiquitin bound specifically to GST-EAP451-229, with an estimated dissociation constant of 210 ± 10 µM in the experiment shown in Fig. 5. EAP451-229 binding was strongly inhibited by the Ub I44A mutation, but not by a F4A mutation, demonstrating that the exposed hydrophobic Ub I44 surface forms at least part of the EAP45 binding site. This same binding surface is also recognized by nearly all of the other ubiquitin binding proteins characterized to date (29). Ubiquitin also bound with a similar affinity to a minimal EAP451-139 construct (KD = 410 ± 90 µM; n = 6), indicating that the ubiquitin binding site is located within the N-terminal GLUE domain. These data are all in excellent agreement with recent analyses of ubiquitin binding to the murine EAP45 GLUE domain (67).
|
To test whether human ESCRT-II associated with endosomal membranes in a VPS4-dependent fashion, we examined the localization of two different ESCRT-II components, EAP20 and EAP30, in the presence of a mutant VPS4A protein (green fluorescent protein [GFP]-VPS4AK173Q) previously shown to block ATP binding and to dominantly inhibit VPS4 activity in cells (8). As shown in Fig. 6A and B, antibodies against EAP20 detected endogenous EAP20 in COS-7 cells, albeit weakly. These antibodies also cross-reacted with a second protein of 36 kDa (data not shown). We therefore also examined the localization of exogenous, Myc-tagged EAP20 to take advantage of the enhanced immunofluorescence signal of the tagged, overexpressed protein (Fig. 6C). In both cases, EAP20 localized to the enlarged endosomes induced by GFP-VPS4AK173Q overexpression. In the absence of GFP-VPS4AK173Q, both endogenous and overexpressed EAP20 were distributed throughout the cytoplasm, as has been seen for other ESCRT complexes (Fig. 6B and D). To confirm that these results accurately represented the distribution of the entire ESCRT-II complex, the localization of a second ESCRT-II component, EAP30, was also examined. Once again, the overexpressed, Myc-tagged EAP30 protein was seen throughout the cell, but it became trapped on the class E compartments induced by dominant-negative GFP-VPS4AK173Q (compare Fig. 6E and F). Hence, ESCRT-II was recruited to class E compartments, where it became trapped in the absence of VPS4 ATPase activity. Localization of the endogenous ESCRT-III protein, CHMP6, was also examined, and this protein was also strongly recruited to aberrant class E compartments (Fig. 6G and H and data not shown). Similar results were obtained using the other dominant-negative ATPase constructs, GFP-VPS4AE228Q, DsRed-VPS4BK180Q, and DsRed-VPS4BE235Q (22, 83), and with human osteosarcoma cells (data not shown). These data are all consistent with the idea that human EAP20/ESCRT-II, EAP30/ESCRT-II, and CHMP6/ESCRT-III function in vesicle formation and cargo sorting at the MVB.
|
20-fold (Fig. 7B and D). Surprisingly, however, depletion of EAP20 did not significantly affect viral release as measured in Western blot assays (Fig. 7A and C, "Virions", lanes "EAP20"), and actually increased vector titers slightly at the 24-h time point (Fig. 7B and D, bars 4), although at a later time point (48 h), HIV-1 titers were reduced twofold (data not shown). We therefore conclude that efficient depletion of EAP20 does not significantly inhibit the release or infectivity of HIV-1.
|
The effects of CHMP6 depletion on the release and infectivity of HIV-1 vectors and virus are shown in Fig. 8. As expected, virus release and infectivity were again substantially reduced upon siRNA depletion of TSG101 (Fig. 8A to D, lanes "TSG101" and bars 2), whereas treatment with an inverted control siRNA had no significant effect on virus release (defined as 100% infectivity; lanes "INV" and bars 3). Two different siRNAs were used to knock down CHMP6 (designated CHMP6-1 and CHMP6-2), and both siRNAs decreased CHMP6 levels very significantly (Fig. 8A and C, "Cells," second row) without affecting Gag protein expression or processing (third row). As with EAP20 depletion, CHMP6 depletion again failed to reduce HIV-1 release or infectivity substantially (Fig. 8A and 8C, "Virions," and B and D). Although there was some variability in multiple repetitions of this experiment, the average viral titers nevertheless fell within 40% of control levels following treatment with both CHMP6-1 and CHMP6-2 (Fig. 8B and D). At a later time point (48 h), viral infectivities were reduced four- to fivefold, possibly owing to secondary effects on general protein trafficking. We therefore conclude that normal cellular levels of CHMP6 are not required for efficient HIV-1 release or infectivity.
|
As shown in Fig. 9, depletion of TSG101/ESCRT-I, EAP20/ESCRT-II, and CHMP6/ESCRT-III significantly reduced the efficiency of EGFR degradation in response to EGF stimulation, although the block was not complete and was somewhat greater for TSG101/ESCRT-I than for EAP20/ESCRT-II and CHMP6/ESCRT-III. Total cellular EGFR protein levels and localization were assayed by immunofluorescence, using an anti-EGFR antibody. In the absence of EGF stimulation, EGFR was located almost exclusively on the surfaces of 293T cells (Fig. 9A, column 1 [from left] and inset). Following EGF stimulation for 2 h, >90% of the EGFR was internalized and degraded (column 2). Importantly, EGFR degradation was significantly reduced by treatment with siRNAs against TSG101 (column 4), EAP20 (column 5), and CHMP6 (column 6), but not by treatment with a control siRNA (column 3). Depletion of the three ESCRT components did not affect receptor internalization but rather induced intracellular accumulation of internalized receptors, consistent with a defect in MVB trafficking. We note that treatment with siRNAs against TSG101 resulted in the accumulation of EGFR in larger class E compartments (17) than were seen in cells treated with siRNAs against EAP20 or CHMP6. The EGFR trafficking defects seen in the absence of ESCRT activities were similar in 293T and HeLa cells, as shown quantitatively in Fig. 9B and C. The efficiencies of protein depletion and virus release were monitored in parallel experiments that confirmed that the targeted proteins were depleted as expected and that HIV-1 release was significantly reduced by TSG101 depletion, but not by depletion of EAP20 or CHMP6 (data not shown).
|
We examined the trafficking of EGF to the lysosome in cells depleted of TSG101/ESCRT-I, EAP20/ESCRT-II, or CHMP6/ESCRT-III. HeLa cells depleted of TSG101/ESCRT-I, EAP20/ESCRT-II, or CHMP6/ESCRT-III were incubated with the fluid phase marker Alexa 488-dextran for 18 h, followed by a 4-h chase in dextran-free medium, which allows the fluid phase marker to accumulate in lysosomes (12, 78). As shown above, the unstimulated EGFR is predominantly localized to the plasma membrane. Incubation of cells with EGF-Alexa Fluor 555 streptavidin induced the internalization of the EGFR, and after 30 min, fluorescent EGF colocalized with lysosomes in 41% of the cells examined (Fig. 10, "INV,") (86/208 cells). In contrast, cells depleted of TSG101/ESCRT-I showed no colocalization of EGF with lysosomes ("TSG101") (0/29 cells tested). Cells depleted of either EAP20/ESCRT-II or CHMP6/ESCRT-III showed intermediate phenotypes, with 7 to 10% of the cells showing EGF staining at the lysosome. To control for possible differences between different siRNA oligonucleotides, we also tested one of the siRNA oligonucleotides used to deplete EAP20 in the study by Bowers et al. This oligonucleotide (EAP20-2) also efficiently depleted EAP20 levels (Fig. 10C) and again gave an intermediate EGF localization phenotype (Fig. 10A and B, "EAP20-2") (19/258 cells positive for EAP20-2 versus 10/140 cells for EAP20-1 and 22/213 cells for CHMP6-2).
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Role of EAP20/ESCRT-II in HIV-1 release. We considered several possible explanations for the observation that EAP20/ESCRT-II depletion does not inhibit HIV-1 release. As with all siRNA depletion experiments, it is possible that even the very low levels of EAP20 remaining after depletion were still sufficient to function in HIV-1 release (and MHC-I downregulation). While we cannot rule out this possible explanation, it seems unlikely because EAP20 was efficiently depleted in our experiments (Fig. 7 and 10), and this depletion measurably inhibited EGFR degradation and lysosomal targeting (Fig. 9 and 10). Furthermore, EAP20 depletion actually produced a slight, but reproducible, increase in HIV-1 infectivity (Fig. 7), which could be explained if the endogenous ESCRT-II complex binds and competes for other factors essential for virus budding (e.g., ESCRT-I). It is also conceivable that ESCRT-II may remain active even without EAP20 subunits. This explanation again seems unlikely, because the two copies of EAP20 comprise half of the ESCRT-II subunits and form the only known link between ESCRT-II and ESCRT-III. Furthermore, ESCRT-II activity is blocked by deletion of EAP20/Vps25p in both S. cerevisiae and Drosophila (4, 76, 77). We therefore conclude that the most likely explanation for our data is that ESCRT-II is not required for efficient HIV-1 release from human cells, either because the complex does not normally participate in virus budding or because the viral budding machinery still functions efficiently in its absence.
Our results also imply that if HIV-1 utilizes the late-acting ESCRT-III and VPS4 complexes to bud from cells, then there must be other proteins or complexes, in addition to ESCRT-II, that can bridge ESCRT-I and ESCRT-III. In principle, ALIX/AIP1 could be one such bridging protein, as it can bind directly to both TSG101 (83) and to the CHMP4 proteins (38, 41, 69, 83), although simultaneous depletion of ALIX and EAP20 did not synergistically inhibit EGFR degradation (11). Similarly, interactions between the yeast ESCRT-I components, Vps37p and Vps28p, and the ESCRT-III component, Vps20p (CHMP6), have been reported (10), although we did not observe analogous ESCRT-I-ESCRT-III interactions between the isolated human proteins in two-hybrid experiments. Finally, the deubiquitylating enzyme, AMSH, binds both the upstream HRS-STAM complex (which binds to TSG101) (71) and the ESCRT-III components CHMP3 and CHMP4 (23). In principle, these (and other) bridging interactions could help to recruit ESCRT-III to sites of MVB vesicle formation and virus budding even in the absence of ESCRT-II.
Role of ESCRT-III in HIV-1 release. Models for the roles of ESCRT-III (and VPS4) in virus budding must take into account the observation that HIV-1 release is not significantly inhibited by depletion of endogenous CHMP5 and CHMP6 (reference 85 and this work), yet virus release is very potently inhibited by overexpression of dominant-negative CHMP5 or CHMP6 protein (46, 69, 70). siRNA depletion of these individual CHMP proteins therefore imposes a less severe (or less general) block in the ESCRT pathway than the dominant-negative ESCRT-III and VPS4 constructs. A likely explanation for this difference is that dominant-negative CHMP (and VPS4) constructs function by inducing the formation of aberrant endosomal compartments (class E compartments) that not only inhibit the functions of the modified CHMP proteins themselves, but also sequester other ESCRT complexes and possibly many other MVB proteins required for HIV-1 budding. Thus, we cannot rule out the possibility that wild-type CHMP and VPS4 proteins do not normally function in HIV-1 release but rather affect virus budding only when present in dominant-negative forms. However, we favor a model in which at least a subset of the ESCRT-III and VPS4 proteins do actually participate directly in virus release but where depletion of individual CHMP5 or CHMP6 proteins does not inhibit virus budding owing to (i) functional redundancy within the MVB pathway arising from the presence of 10 different CHMP proteins, (ii) the possibility that CHMP5 (and possibly other CHMP proteins) may function at a step following MVB vesicle formation (7, 66), and/or (iii) the presence of multiple distinct ESCRT pathways in human cells. It therefore continues to be important to determine precisely which subset of different ESCRT proteins is utilized directly by HIV-1 and other viruses.
Complexity of MVB vesicle formation and protein sorting in higher eukaryotes. Genetic studies have identified the basic machinery required for MVB vesicle formation and protein sorting in S. cerevisiae. Importantly, deletion of each of the six known yeast ESCRT-I and ESCRT-II subunits ultimately leads to a strong block in protein sorting and MVB vesicle formation (the class E phenotype) (40). While it seems almost certain that the basic mechanism of MVB biogenesis is highly conserved, because human cells have at least one ortholog of every known yeast class E protein (33, 51), our studies add to the growing number of indications that MVB protein-sorting/vesicle formation pathways are more complex in higher eukaryotes than in yeast. One level of increased complexity is suggested by the fact that unique yeast class E proteins often have multiple orthologs in mammalian cells. Examples include the 4 different human homologs of the single yeast Vps37p protein (6, 19, 70), the 10 different human homologs of the six yeast ESCRT-III proteins (51), and the 2 (or more) different human homologs of yeast Vps4p (60). Thus, unique proteins or complexes that are essential for MVB formation in yeast may play redundant or overlapping roles in higher eukaryotes. For example, simultaneous depletion of both VPS37B and VPS37C inhibits PTAP-dependent retrovirus budding to a greater extent than depletion of either single VPS37 paralog alone (48).
A second level of complexity is suggested by the fact that even unique ESCRT-I or ESCRT-II subunits do not appear to be required for metazoan MVB vesicle formation in all contexts. For example, a knockout of the Drosophila Vps28/ ESCRT-I protein has only modest effects on MVB morphology and no measurable effect on the downregulation of several different cell surface receptors in vivo, even though there is only a single vps28 gene in Drosophila (63). Similarly, agonist-induced lysosomal degradation of the delta opioid receptor is inhibited by dominant-negative VPS4B mutants and by depletion of HRS, but not by depletion of TSG101/ESCRT-I (31). Hence, an intact ESCRT-I complex is apparently not required for downregulation of all cell surface receptors via the MVB pathway (27, 31). Similarly, EAP20/ESCRT-II is not required for MHC-I receptor downregulation induced by Kaposi's sarcoma-associated herpesvirus KK3 (reference 11 and data not shown).
Retroviruses also differ in their requirements for ESCRT-I. HIV-1, which binds directly to TSG101, requires all of the known ESCRT-I components for efficient release (19, 22, 48), whereas equine infectious anemia virus and Moloney murine leukemia virus, which bud primarily via ALIX and the Nedd4 E3 ligases, respectively, exhibit little or no requirement for TSG101/ESCRT-I (22, 48, 62, 64, 72). Nevertheless, all retroviruses are inhibited by dominant-negative ESCRT-III and VPS4 proteins, suggesting a common use of these downstream factors (22, 46, 69, 83). This model is consistent with the idea that ESCRT-I and ESCRT-II may function primarily as adaptor complexes that help recruit cargoes into the pathway, whereas the ESCRT-III/VPS4 machinery functions more directly in vesicle formation. Indeed, the genomes of Plasmodium falciparum and Toxoplasma gondii reportedly lack ESCRT-I and ESCRT-II entirely but do encode ESCRT-III and VPS4 proteins and require VPS4 to create multivesicular bodies (87). Finally, we note that even the ESCRT-III/VPS4 machinery appears to be dispensable in some cases, as the melanosomal protein Pmel17 is sorted into MVB vesicles via an ESCRT-independent pathway that is also insensitive to inhibition by dominant-negative forms of VPS4 (75). The apparent diversity of MVB sorting pathways raises the intriguing possibility that viral systems may also have evolved to utilize such "noncanonical" MVB pathways to escape the cell (34).
Roles for mammalian ESCRT-I and ESCRT-II in receptor downregulation. Although ESCRT-I and ESCRT-II are not absolutely required for the efficient downregulation of all receptors or for the release of all retroviruses, both complexes clearly play very important roles in downregulating certain cell surface receptors. Indeed, a number of recent studies indicate that both ESCRT-I and ESCRT-II exhibit growth/tumor suppressor activities by virtue of their roles in cell surface receptor downregulation. For example, all three known ESCRT-I components (TSG101, VPS28, and VPS37) have now been implicated in cell growth and tumor suppression. TSG101 was initially identified in a genetic screen for genes with tumor suppressor activity, and depletion of TSG101 induces NIH 3T3 cell overproliferation in culture and metastatic tumor formation in nude mice (44). Similarly, deletion of erupted, the Drosophila TSG101 ortholog, inhibits Notch receptor degradation and activates the JAK-STAT (and possibly other) signaling pathway and thereby induces tissue overproliferation (50). Likewise, depletion of one of the four mammalian VPS37 proteins (6, 19, 70), VPS37A/HCRP-1, enhances the growth of cultured hepatocellular carcinoma BEL-7404 cells and elevates their invasive ability (86). Interestingly, VPS37A/HCRP-1 is also downregulated in some hepatocellular carcinomas. Finally, we have observed that depletion of VPS28 can relieve contact-inhibited growth in cultured 293T cells (J. Garrus, personal communication). ESCRT-II also appears to play an important role in growth receptor downregulation, as deletion of the Drosophila vps25 gene leads to tissue overproliferation through aberrant stabilization of the Notch and DPP (and possibly other) receptors (76, 77). Similarly, we observed that depletion of EAP20 (human VPS25) reduces the degradation rates of EGFR (Fig. 9 and 10), ferroportin (De Domenico, et al., unpublished), and presumably other cell surface receptors following ligand stimulation. Thus, just as the proteasome functions to degrade soluble proteins, the mammalian ESCRT pathway, including ESCRT-II, can play a critical role in mediating both homeostatic and regulated lysosomal degradation of integral membrane proteins.
|
ACKNOWLEDGMENTS |
|---|
We thank Chris Rodesch, David Myszka, and Rebecca Rich for assistance with immunofluorescence and biosensor experiments.
|
FOOTNOTES |
|---|
Supplemental material for this article may be found at http://jvi.asm.org.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
