Characterization of Mammalian Ecm29, a 26 S Proteasome-associated Protein That Localizes to the Nucleus and Membrane Vesicles*

In addition to its thirty or so core subunits, a number of accessory proteins associate with the 26 S proteasome presumably to assist in substrate degradation or to localize the enzyme within cells. Among these proteins is ecm29p, a 200-kDa yeast protein that contains numerous HEAT repeats as well as a putative VHS domain. Higher eukaryotes possess a well conserved homolog of yeast ecm29p, and we produced antibodies to three peptides in the human Ecm29 sequence. The antibodies show that Ecm29 is present exclusively on 26 S proteasomes in HeLa cells and that Ecm29 levels vary markedly among mouse organs. Confocal immunofluorescence microscopy localizes Ecm29 to the centrosome and a subset of secretory compartments including endosomes, the ER and the ERGIC. Ecm29 is up-regulated 2–3-fold in toxin-resistant mutant CHO cells exhibiting increased rates of ER-associated degradation. Based on these results we propose that Ecm29 serves to couple the 26 S proteasome to secretory compartments engaged in quality control and to other sites of enhanced proteolysis. The covalent attachment of ubiquitin (Ub) to eukaryotic proteins is a post-translational modification of remarkable scope, complexity, and importance (1). The ubiquitin system vast array of physiological including tran-scription dissociation of Ecm29 pro- teasome. proteasomes proteasome-free proteasome:Ecm29 resedimented on glycerol gradients, assayed pepti- dase activity suc-LLVY-MCA SDS-PAGE immunoblotting Partial protease DEAE ml of linear

The covalent attachment of ubiquitin (Ub) 1 to eukaryotic proteins is a post-translational modification of remarkable scope, complexity, and importance (1). The ubiquitin system regulates a vast array of physiological processes including transcription (2)(3)(4), cell cycle traverse (5), apoptosis (6), circadian rhythms (7), and memory (8), just to name a few. Although attachment of Ub can serve non-destructive purposes (9 -11), the selective degradation of proteins is the principal way in which Ub controls cellular processes (12). The exquisite selectivity exhibited by Ub-mediated proteolysis is provided by several hundred Ub ligases, the E3 recognition components of the system (13)(14)(15). The Ub ligases collaborate with a smaller number of Ub carrier proteins, the E2s, to add poly(Ub) chains onto ⑀-amino groups of lysines in the protein substrates. The poly(Ub) chains in turn target the marked proteins to the 26 S proteasome, a large ATP-dependent protease, for degradation. The 26 S proteasome is formed by the binding of 19 S regulatory complexes (RCs) to one or both ends of the cylindrical 20 S proteasome, the proteolytic component of the 26 S enzyme. The 19 S RC provides subunits that recognize, unfold, and transfer substrate polypeptides to the central chamber of the 20 S proteasome where they are degraded (16).
The foregoing account would suggest that there are only two forms of the 26 S proteasome, one with a single 19 S RC cap and one doubly capped. However, a number of proteins associate either with the 19 S RC or with the 20 S proteasome. In eukaryotes, PA200 and PA28 (or REG) are two proteasome activators that, like the RC, bind the ends of the 20 S proteasome (17,18). Since there are two ends to the cylindrical 20 S proteasome, the enzyme can simultaneously bind a PA200 or a PA28 and a 19 S RC, thereby forming hybrid proteasomes (19). The 19 S RC associates with a larger set of proteins, many of which are components of the Ub system. For example, Ub isopeptidases form reasonably stable associations with the 26 S proteasome (20), and several Ub ligases interact with 26 S proteasomes (21)(22)(23)(24). An even larger number of proteins have been identified as interacting partners of individual RC subunits, especially the ATPases (25). Whereas some of these proteins may be substrates, others may function as adaptors that recruit substrates to the 26 S proteasome or localize the enzyme within cells (26). It now appears that 26 S proteasomes have a dynamic composition in which a core of about 30 different subunits associates with perhaps a greater number of accessory proteins that assist in substrate degradation.
Recently, several proteomic screens identified the yeast protein ecm29p as a 26 S proteasome-associated component (27,28). The association of ecm29p and the 26 S proteasome was examined by Leggett et al. (29), who found the protein associated with either the 19 S RC or the 20 S proteasome depending upon conditions. ECM29 deletion produced yeast 26 S proteasomes that dissociated in the absence of ATP, leading Leggett et al. (29) to propose that ecm29p tethers the RC to the 20 S proteasome. Higher eukaryotic genomes contain a well conserved homolog of yeast ecm29p, and this allowed us to generate anti-peptide antibodies to human Ecm29. We have used these antibodies for the initial characterization of mammalian Ecm29.
Monoclonal antibodies to the KDEL retrieval sequence of ER-resident proteins were obtained from Calbiochem. Anti-mannose-6-phosphate receptor monoclonal antibodies were from abcam. A monoclonal antibody to the human transferrin receptor (TR) was purchased from Zymed Laboratories. Monoclonal antibodies MCP21, MCP102, and MCP168 to 20 S proteasome subunits ␣2, ␤3, and ␤2, respectively were purchased from Affiniti. Monoclonal antibodies to ERGIC-53 were a kind gift from Hans-Peter Hauri (Basel, Switzerland). Anti-␣6 (MCP20) monoclonal antibodies were a gift from Klavs Hendil (Copenhagen, Denmark). Antibodies to S10b and S14 were prepared as described (30).
Sequence of Ecm29 -The human sequence was assembled using the sequence of the KIAA0368 protein (gi:30151586) and cDNA sequences (gi:34534077 and gi:34527903) contained in the GenBank TM data base. This sequence is identical to the latest GenBank TM release of the FIG. 1. A, human Ecm29 (hEcm29) sequence. The human sequence was assembled using the sequence of the KIAA0368 protein (gi:30151586) and cDNA sequences (gi:34534077 and gi:34527903) contained in the GenBank TM data base. Sequences used to produce anti-peptide antibodies are underlined in red. A potential nuclear localization signal sequence is highlighted in green. One of three alternative N-terminal sequences in hEcm29 encoded by human, mouse, and rat ESTs derived from brain, retina, testis, and ovary is highlighted in black. Sequences homologous to conserved regions of ␤-adaptins are highlighted in orange. A region that bears resemblance to VHS domains is highlighted in pink. B, architecture of human Ecm29. We have proposed that Ecm29 is composed almost entirely of HEAT-like repeats (diamonds) (35). A region with homology to VHS domains (pink rectangle) and two helical regions homologous to conserved sequences in adaptins (orange ellipses) were identified by RPS-BLAST analysis of the oasis_sap.v1.58 data base. The VHS-like domain of hEcm29 was aligned with a structure-based alignment of the VHS domain of D. melanogaster Hrs (HrsD) (74,75)  Ecm29 Antibodies-Three peptides, Ecm29 -1 (C-RNRKESTSE-QMPSFPE); Ecm29 -2 (C-RGRPLDDIIDKLPE); and Ecm29 -3 (C-RPELQEKAALLKKTLENLE), corresponding to distinct regions of the human Ecm29 protein (see Fig. 1) were used for antibody production. Peptides were synthesized and coupled to maleimide-activated ovalbumin according to the manufacturer's instructions through an added N-terminal cysteine residue. Peptide-coupled ovalbumin was sent to Harlan Bioproducts for Science, Inc (Indianapolis, IN) for antibody production.
Distribution of Ecm29 in HeLa Cells and Mouse Organs-HeLa cells or freshly harvested mouse organs were extracted in 10 mM Tris-HCl, pH 7.5, 0.25% Triton X-100, 1 mM dithiothreitol (DTT) containing protease inhibitors and centrifuged at 4°C for 45 min at 13,000 ϫ g. Samples (30 -50 g of total protein) were separated by SDS-PAGE, transferred to nitrocellulose, and immunoblotted with anti-Ecm29-peptide antibodies. Blots were developed by enhanced chemiluminescence using the SuperSignal® West Femto reagent.
Association of Ecm29 with the 26 S Proteasome-HeLa cells (ϳ10 8 ) were homogenized in 1 ml of 5 mM Hepes, pH 7.2, 0.25 M sucrose, 0.2 mM EDTA containing protease inhibitors by 300 strokes using a tightfitting Dounce homogenizer. A 200-l aliquot of the post-mitochondrial supernatant (1.8 mg) was diluted with 50 mM sucrose in 5 mM Hepes, pH 7.2, 0.2 mM EDTA (final volume 0.3 ml) and centrifuged on a 5-ml 10 -30% glycerol gradient for 4 h at 48,000 rpm using a Beckman SW50.1 rotor. 180-l fractions were collected from the bottom and assayed for peptidase activity in the presence or absence of ATP using suc-LLVY-MCA as substrate as previously described (31). Samples (16 l) from each fraction were separated by either SDS-PAGE or native gel electrophoresis for 600 V-h at 4°C. Native gels were subsequently overlaid with 20 mM suc-LLVY-MCA for 1 h at 37°C to visualize the 26 S and 20 S proteasomes as described (32). Proteins were transferred to nitrocellulose and immunoblotted with anti-Ecm29-3.
Dissociation of Ecm29 from the 26 S Proteasome-HeLa cells were extracted in 1 ml of 5 mM Hepes, pH 7.2, 0.2 mM EDTA containing 0.25% Triton X-100 and protease inhibitors, and centrifuged. The postmitochondrial supernatant was then sedimented on a 10 -30% glycerol and 16-l aliquots of the gradient fractions were analyzed by native gel electrophoresis followed by substrate overlay and immunoblotting as described above.
Re-association of Ecm29 with the 26 S Proteasome-300-l samples of Triton X-100 HeLa cell extract were centrifuged on 5-ml 10 -30% glycerol gradients as described above. Fractions were assayed for peptidase activity and analyzed by SDS-PAGE and immunoblotting with anti-Ecm29-3 antibodies to verify dissociation of Ecm29 from the proteasome. Fractions containing 26 S proteasomes or proteasome-free Ecm29 were pooled, bovine serum albumin was added to 0.1 mg/ml, and samples were dialyzed against reticulocyte buffer (30 mM Tris-HCl, pH 7.8, 10 mM NaCl, 5 mM MgCl 2 , 1 mM DTT). Dialyzed samples were concentrated using Centricon 30 TM (Amicon) concentrators, mixed at a 26 S proteasome:Ecm29 ratio of 1.7 (v/v) in the presence of 1 mM ATP and 10% glycerol and incubated at 37°C for 30 min. Samples were finally resedimented on 10 -30% glycerol gradients, assayed for peptidase activity using suc-LLVY-MCA as substrate and analyzed by SDS-PAGE and immunoblotting with anti-Ecm29-3 antibodies.
Partial Purification of Ecm29 from Bovine Brain-Bovine brain was homogenized in 10 mM Tris-HCl, pH 7.0, 25 mM KCl, 10 mM NaCl, 1.1 mM MgCl 2 , 0.1 mM EDTA, 1 mM DTT containing 0.25% Triton X-100 plus protease inhibitors, and centrifuged at 13,000 ϫ g. Glycerol was added to 10% and the sample (8 g) was batch-adsorbed to 500 ml DEAE and proteins were eluted with 5-column volumes of a 35 to 300 mM KCl linear gradient. Fractions containing Ecm29 were identified by SDS-PAGE and immunoblotting, pooled and loaded onto a 100-ml DEAE column. Proteins were eluted with 500 ml of a linear gradient from 35 to 300 mM KCl. Pooled fractions containing Ecm29 were concentrated and applied to a HiLoad 26/60 prep grade Superdex 200 column equilibrated in 25 mM Tris-HCl, pH 7.5, 75 mM KCl, 1 mM DTT, and 10% glycerol. Last, fractions containing Ecm29 were loaded onto a Mono-Q HR 5/5 column equilibrated in homogenization buffer, pH 8.5 with 10% glycerol, and proteins were eluted with a linear gradient from 35 to 750 mM KCl. A 200-l sample of the pooled Mono-Q fractions was layered atop a 5-ml 5-20% sucrose gradient and centrifuged at 39,000 rpm for 18 h at 4°C using a Beckman SW50.1 rotor. Fractions were analyzed by SDS-PAGE and immunoblotted with anti-Ecm29-3 antibodies.
Cytochalasin B-mediated Enucleation-HeLa cells were grown on 18-mm circular coverslips and enucleated after treatment with cytochalasin B by the method of Prescott et al. (33) as described previously (34). Karyoplast and cytoplast fractions were extracted in 50 l of 0.25% Triton X-100 in 10 mM Tris-HCl, pH 7.5, 1 mM ATP, 1 mM DTT plus protease inhibitors and subjected to SDS-PAGE and immunoblotting with anti-Ecm29-3 or antibodies to 26 S proteasome subunits.
Expression of Ataxin-7 Containing an 86-Residue Polyglutamine Tract-Human HeLa and L4 prostate cells were grown on coverslips for 24 h and transfected with 2 g of a plasmid encoding ataxin-7-Q 86 bearing a His 10 N-terminal tag using Lipofectamine Plus TM according to the manufacturer's instructions (Invitrogen). Two days after transfection, cells were fixed with methanol and stained with anti-His (1:40) monoclonal antibodies (Novagen) and rabbit anti-Ecm29-3 antiserum. African Green Monkey COS7 were transfected with the ataxin-7-Q 86 plasmid using Lipofectamine 2000 TM and intranuclear polyQ aggregates were detected with rabbit anti-His (1:50) polyclonal antibodies (abcam).
Expression of Tagged Versions of Mouse Ecm29 in Cultured Cells-A cDNA encoding mouse Ecm29 (96% identity to human Ecm29) cloned into the pXY vector was purchased from Open Biosystems and used as template for the polymerase chain reaction. The resulting DNA was FIG. 2. A, antibodies to human Ecm29 recognize a 205 kDa protein in HeLa cell extracts. Peptides were synthesized and coupled to ovalbumin for antibody production as described under "Experimental Procedures." HeLa cells were extracted as described under "Experimental Procedures" and samples of the total extract, soluble fraction, and the corresponding pellets were separated by SDS-PAGE and immunoblotted with anti-Ecm29 antibodies as indicated (middle and right panels). B, levels of Ecm29 in mouse organs. Homogenates of freshly harvested mouse organs were separated by SDS-PAGE and immunoblotted with anti-Ecm29 antibodies as indicated, or with anti-S10b, an ATPase of the 19 S regulatory complex. The arrows denote a high molecular weight isoform of Ecm29 present at high levels in brain. Western blot analysis of the pellet fractions revealed little Ecm29 (not shown).
subcloned at the BamHI/SalI sites of pFLAG-CMV (Sigma) or pGFP-N1 (Clontech). Transfections of HeLa, HEK293, and COS7 cells were performed with Lipofectamine 2000 TM according to the manufacturer's instructions. For immunofluorescence, HeLa and COS7 cells grown on coverslips were transfected with 5 g of total plasmid DNA, fixed in methanol at Ϫ20°C 2 days after transfection and processed as described below. HEK293 cells grown on T-25 flasks (Corning) were transfected with 10 g of pFLAG-CMV-Ecm29 and harvested 19 h after transfection. Cells were homogenized in buffer containing 0.25 M sucrose and sedimented through a 5-ml 10 -30% glycerol gradient as described above. Fractions were assayed for peptidase activity in the presence or absence of 1 mM ATP and analyzed by denaturing or non-denaturing gel electrophoresis followed by Western blotting using anti-FLAG M2 TM monoclonal antibodies (1:400).

FIG. 3. Association of human Ecm29 with the 26 S proteasome is reversible.
A, HeLa cells were homogenized in buffer containing 0.25 M sucrose or B, were extracted in buffer containing 0.25% Triton X-100. Aliquots of the post-mitochondrial supernatants (ϳ1.8 mg of total protein) were centrifuged on 5-ml 10 -30% glycerol gradients as described under "Experimental Procedures." Fractions were collected and assayed for peptidase activity in the presence (•) or absence (E) of ATP using suc-LLVY-MCA as substrate (top panels). Samples were separated by non-denaturing electrophoresis, overlaid with 20 M suc-LLVY-MCA (second panels), transferred to nitrocellulose and immunoblotted with anti-Ecm29-3 antiserum (third panels). In A, samples were also subjected to SDS-PAGE and immunoblotted with anti-Ecm29-3 (bottom panel). B clearly shows that after extraction with Triton X-100, Ecm29 is no longer associated with the 26 S proteasome. C, bovine brain was homogenized in buffer containing 0.25% Triton X-100 plus protease inhibitors and centrifuged at 13,000 ϫ g to remove debris. The supernatant fraction was then subjected to a series of chromatographic steps as described under "Experimental Procedures." Samples were analyzed by Western blotting of non-denaturing (NATIVE) and SDS-PAGE gels using anti-Ecm29-3 antibodies. D, 200-l sample of the pooled Mono-Q fractions was centrifuged on a 5-ml 5-20% sucrose gradient at 39,000 rpm for 18 h. Fractions were analyzed by SDS-PAGE and immunoblotting with anti-Ecm29-3. Sedimentation markers included catalase (232 kDa), aldolase (158 kDa), and bovine serum albumin (67 kDa). E, Triton X-100 HeLa cell extracts were centrifuged on 5-ml 10 -30% glycerol gradients and assayed as described above. Fractions were analyzed by SDS-PAGE and immunoblotting with anti-Ecm29-3 to verify dissociation of Ecm29 from the proteasome (not shown). Fractions containing stripped 26 S proteasomes or "free" Ecm29 were pooled, bovine serum albumin was added to 0.1 mg/ml, and each sample was dialyzed to remove the detergent. Dialyzed samples were concentrated, mixed at a 26 S proteasome:Ecm29 ratio of 1.7 (v/v) in the presence of 1 mM ATP and 10% glycerol, and incubated at 37°C for 30 min. Samples were then resedimented on a 10 -30% glycerol gradient, assayed for peptidase activity in the presence (•) or absence (E) of ATP and analyzed by immunoblotting with anti-Ecm29-3 (bottom panel). After dialysis and concentration, both 26 S proteasome and Ecm29 sedimented as shown in the top and middle panels. Note that Ecm29 is absent from the stripped 26 S proteasome (middle panel). However, when the enzyme was mixed with stripped Ecm29 (26 S ϩ Ecm29), a substantial amount of Ecm29 was found re-associated with the 26 S proteasome. PBS for 10 min at room temperature. Cells were blocked with 10% nonfat milk in Tris-buffered saline, pH 8.0 with 0.05% Tween 20 (TBST) and incubated with primary or secondary antibodies in blocking buffer for 1 h at 37°C. Fluorescent secondary antibodies were used at a dilution of 1:500. After incubation with antibodies, cells were washed six times in TBST for 5 min each at room temperature. Fluorescence confocal laser scanning microscopy was performed on an Olympus FVX confocal microscope equipped with argon 488 and krypton 568/647 lasers using a 60X PlanApo 1.4 NA oil objective and Fluoview 2.13.9 software. The relative amounts of Ecm29 and cellular markers at specific subcellular locations were estimated using Volocity v2.0, a volume rendering and three-dimensional quantification software (Improvision).

RESULTS
Amino Acid Sequence of Human Ecm29 (hEcm29)-From the human genome data base and EST libraries we assembled a putative full-length sequence for human Ecm29 (Fig. 1A) that is identical to the KIAA0368 protein reported in the latest GenBank TM release. Human Ecm29 is 1839 amino acids long, and Prosite analysis reveals the presence of a possible bipartite nuclear targeting signal (highlighted green in Fig. 1A). The N-terminal sequence of human Ecm29 (highlighted black in Fig. 1A) is one of three alternative N-terminal sequences encoded by human, mouse, and rat ESTs derived from brain, retina, testis, and ovary: MYHIDCRDqler (most highly represented); MAAAAASASQDELNqler; and METGSDSDqler. All known Ecm29 sequences exhibit an apparent modular structure consisting of HEAT repeats (at least 23 in hEcm29; diamonds in Fig. 1B) (35), a region with sequence homology to VHS domains (36) (highlighted pink in Fig. 1A), and two potential helical regions with homology to conserved sequences in ␤-adaptins (highlighted orange in Fig. 1A) that we call HANT domains (homology to adaptins N terminus). PSI-BLAST reiterative analysis of the GenBank TM data base using the entire human Ecm29 sequence revealed that HEAT repeats within the C-terminal half of Ecm29 show highest similarity to HEAT motifs of the translational activator GCN1 and importin ␤2 (E values Ͻ e Ϫ60 after 5 iterations) (35,47). A similar computational analysis indicated highest similarity between the VHSlike region of human Ecm29 (residues 893-1038; pink rectangle in Fig. 1B) and the VHS domain of hepatocyte growth factorregulated tyrosine kinase substrate (Hrs) (E values e Ϫ07 -e Ϫ10 after 5 iterations). However, unlike the VHS domain of Hrs and other proteins, the VHS-like region of Ecm29 is not present at the N terminus of the molecule (Fig. 1).
Antibodies to Human Ecm29 -Antibodies were generated to synthetic peptides corresponding to the three sequences underlined in red in Fig. 1. Western blots of HeLa cell extracts using the resulting antisera identified a 205-kilodalton polypeptide, the size expected for hEcm29 ( Fig. 2A). A survey of mouse organs using the three anti-peptide antibodies revealed that the levels of Ecm29 vary considerably, being particularly high in testis and brain, low in liver, and almost undetectable in heart or kidney (Fig. 2B). By contrast, the levels of S10b, a 19 S RC ATPase, were highest in testis, lowest in pancreas, and relatively uniform in the rest of the organs (Fig. 2 B). Two of the three antisera recognized a slower migrating form of Ecm29 in brain (arrows in Fig. 2B) and some mouse and human brain ESTs contain open reading frames extending beyond the initial Met residue of Ecm29 (not shown). Presumably, alternative splicing gives rise to a longer form as well as smaller Ecm29 species present in brain. The patterns generated by the three antibodies were similar in testis, lung, and spleen and in other organs (pancreas, liver, and kidney) differed at most in minor aspects leaving little doubt that Ecm29 is the protein that was detected by the three anti-peptide antibodies.
Association of hEcm29 with the 26 S Proteasome Is Detergent-sensitive and Reversible-To determine if human Ecm29 associates with the 26 S proteasome, HeLa cells were homogenized in 0.25 M sucrose, and the distribution of proteasome activity was determined after sedimentation of the post-mitochondrial supernatant fraction on a 10 -30% glycerol gradient. Samples from the various fractions were subjected to native or SDS-PAGE electrophoresis followed by Western blotting. As shown in Fig. 3A, virtually all HeLa Ecm29 is bound to the 26 S proteasome after homogenization in sucrose. This is particularly clear by comparing the location of 26 S proteasomes evident from fluorogenic peptide overlay (second panel) with the Western blot of the same gel (third panel). Except for a minor immunoreactive band (possibly the 19 S RC), directly below the two enzymatically active species of 26 S proteasome, HeLa Ecm29 migrated with the large, ATP-dependent protease. A Western blot of SDS-PAGE transfers (bottom panel) confirmed that the immunoreactivity on the native gel transfer is present as the 205-kilodalton polypeptide expected for Ecm29 (Fig. 3A, bottom panel).
A much different distribution of Ecm29 was observed after extraction of HeLa cells in buffer containing Triton X-100. Ecm29 was no longer associated with the 26 S proteasome but sedimented with apparent molecular weights ranging from 200 to 600 kDa (Fig. 3B). Western blots of native gel transfers showed several forms of hEcm29 that migrated as a smear on native gels; faster sedimenting species present in fractions 15-20 barely entered the gel whereas slower sedimenting species found in fractions 21-25 formed a distinct band. During partial purification of Ecm29 from detergent extracts of bovine For competition, antibodies were preincubated with 30 mg/ml immunizing peptide for 2 h at 4°C (C and F). Detection and fluorescence confocal laser scanning microscopy were performed as described under "Experimental Procedures." The images are integrated z-projections of twelve 0.9 m optical sections. As shown in Fig.  5, intense juxtanuclear Ecm29 staining corresponds to the centrosome. Scale bars, 10 m. brain, the protein exhibited similar properties. Ecm29 from more purified preparations (i.e. after Mono-Q chromatography) sedimented as a monomer on a sucrose gradient and migrated fastest on a native gel (Fig. 3, C and D). Presumably, slower migrating and faster sedimenting detergent-released Ecm29 molecules either form multimers or are associated with proteasome subcomplexes or possibly with other cellular components. Fractions containing proteasome-free hEcm29 were pooled, dialyzed, concentrated, and mixed with detergent-stripped HeLa 26 S proteasomes. Following re-centrifugation substantial amounts of Ecm29 once again sedimented with the 26 S pro-teasome, indicating that detergent extraction does not prevent reassociation of Ecm29 with the 26 S proteasome (Fig. 3E).
Subcellular Localization of Mammalian Ecm29 -Staining of HeLa cells with anti-Ecm29-1 or anti-Ecm29-3 resulted in predominantly cytoplasmic fluorescence that was virtually absent in cells stained with preimmune serum and was blocked by preincubation with the cognate peptide (Fig. 4). At low PMT/ gain exposures there was intense juxtanuclear staining in most HeLa cells. Confocal images of cells stained with both anti-Ecm29 and anti-␥-tubulin identified this region as the centrosome (Fig. 5, A-C). In addition to the centrosomal location,   higher PMT/gain exposures showed that HeLa Ecm29 was present in a cytoplasmic vesicular pattern. To determine whether Ecm29 was localized to a particular set of membrane compartments we collected confocal images from HeLa cells stained with anti-Ecm29 and various compartment-specific antibodies. The images in Fig. 5, D-L show that Ecm29 partially overlapped with anti-KDEL, a monoclonal antibody against the retention signal of ER-resident proteins, and with ERGIC-53, a marker of the ER-Golgi intermediate compartment (ERGIC) (37). The extent of overlap between Ecm29 and the ER/ERGIC compartments increased substantially when both anti-KDEL and anti-ERGIC-53 were incubated simultaneously with anti-Ecm29 (Fig. 5, J-L) indicating that Ecm29 is present on both compartments. By contrast, there was little overlap between Ecm29 and golgin-97 (Fig. 5, M-P), a marker for the trans-Golgi network (TGN) (38).
Although these results suggest that Ecm29 is present on a subset of ER/ERGIC compartments in the secretory pathway, the dense packing of vesicles within the cytoplasm makes this conclusion tentative. Endosomes provide a set of vesicles that can readily be distinguished following the uptake of fluorescent markers. This feature coupled with the significant Ecm29 staining peripheral to the ER/ERGIC compartments prompted us to ask whether Ecm29 also associates with endosomes. HeLa cells were incubated in medium containing Alexa® 488dextran (39,40), fixed in 2% paraformaldehyde and stained with anti-Ecm29-1 or Ecm29-2 (the epitope for anti-Ecm29-3 is destroyed by paraformaldehyde). The confocal images shown in Fig. 6, A-F demonstrate that Ecm29 associates with endosomes. As expected, vesicle-specific Ecm29 staining was absent in cells incubated with preimmune serum, and abolished when antibodies were preincubated with cognate peptide (Fig. 6, G-L). Confocal imaging of HeLa cells stained with monoclonal antibodies to the transferrin and mannose-6-phosphate receptors confirmed that Ecm29 associates with both early and late endocytic vesicles (Fig. 7). Furthermore, in HeLa cells pulselabeled with fluorescent dextran and chased for 2 h in serumfree medium, Ecm29 was associated with late juxtanuclear endosomal vesicles that had fused (data not shown). Since glycerol gradient fractionation indicated that all Ecm29 mole-  and c). Panels a-c were merged and are shown directly below (d-f). Note the presence of many triple-labeled vesicles (shown in white in panel f). B, cells incubated with a polyclonal antibody to S7, a 19 S regulatory complex subunit (a). Antibody binding was detected with Alexa® 568 goat anti-rabbit IgG and Alexa® 647 goat anti-mouse IgG and visualized by confocal microscopy as described under "Experimental Procedures." Similar results were obtained with antibodies to subunits ␣6 and ␤2 of the 20 S particle and to subunits S6Ј and S14 of the 19 S RC (data not shown). Centrosomes are indicated by arrowheads (see Fig. 5, A-C). Scale bars, 10 m.
cules are present on 26 S proteasomes (Fig. 3), we expected that 26 S proteasomes would also be found on endosomes. This proved to be the case as shown by staining with antibodies to ␤3, a 20 S proteasome subunit, and antibodies against S7 of the 19 S RC (Fig. 8, A and B). It should be noted that classical fractionation procedures involving homogenization in sucrose reveal all Ecm29 to be associated with soluble 26 S proteasomes (Fig. 3A), thereby precluding attempts to confirm the immunofluorescence data by traditional methods of subcellular fractionation. We did, however, make two additional measure-ments on the intracellular distribution of mammalian Ecm29. HEK293 cells were transfected with pFLAG-CMV expressing FLAG-tagged mouse Ecm29. Glycerol gradient sedimentation of sucrose extracts revealed that FLAG-Ecm29 associated exclusively with the 26 S proteasome (Fig. 9A), and immunofluorescence staining with anti-FLAG monoclonal antibodies produced a vesicular distribution like that observed with the three anti-peptide antibodies to Ecm29 (Fig. 9B). Furthermore, similar results were obtained upon transfection with a plasmid expressing an Ecm29-green fluorescent protein (GFP) fusion protein (see bottom panels in Fig. 12, below). Thus, we are confident that Ecm29 is a component of 26 S proteasomes that are localized on intracellular vesicles and, as shown below, present in the nucleus.
Ecm29 Is Up-regulated in Toxin-resistant CHO Cells Exhibiting Enhanced ER-associated Degradation (ERAD)-The presence of a putative VHS domain in Ecm29 and its localization on secretory compartments suggests that the protein may recruit 26 S proteasomes to membranes for the purpose of ERAD. In previous studies, mutagenized CHO cells were selected for simultaneous resistance to Pseudomonas aeruginosa exotoxin A and ricin (41). Since the two toxins are endocytosed by different pathways and modify different cellular targets, multitoxin resistance was expected to arise from changes in a common toxin processing mechanism. This was confirmed by finding that six of the isolated cell lines also exhibited nonselected resistance to cholera toxin (CT), which is thought to exploit ERAD for entry into the cytosol. In three of these toxin-resistant clones (5, 16, and 46) the established ERAD substrate ␣ 1 -antitrypsin Z accumulated in the ER and was thus FIG. 9. Expression of FLAG-tagged Ecm29 in HeLa cells. A, FLAG-Ecm29 binds the 26 S proteasome. HEK293 cells expressing mouse FLAG-Ecm29 were homogenized in buffer containing 0.25 M sucrose, and an aliquot of the post-mitochondrial supernatant (1.5 mg of total protein) was sedimented through a 5-ml 10 -30% glycerol gradient. Fractions were analyzed as described under "Experimental Procedures" and in the legend to Fig. 3. The Western blots in the third and fourth panels clearly show that FLAG-Ecm29 co-sediments with and binds to endogenous 26 S proteasomes. B, HeLa cells grown on coverslips were transfected with pFLAG-CMV-Ecm29, fixed in methanol 48 h after transfection and stained with monoclonal anti-FLAG M2™ antibodies (1:50; Sigma). Binding was visualized by confocal microscopy using Alexa® 488-labeled secondary antibodies as described under "Experimental Procedures." Scale bar, 10 m.

FIG. 10. Up-regulation of Ecm29 in cell mutants resistant to multiple toxins.
Wild type (wt) and CHO mutant cells were extracted in 10 mM Tris-HCl, pH 7.5, 0.25% Triton X-100, 1 mM DTT plus protease inhibitors, and centrifuged. 30-g samples were then separated on 10% gels and immunoblotted with anti-Ecm29-3 or antibodies to different 26 S proteasome subunits. A, survey of mutant CHO cell lines for Ecm29 expression. Note that clones 23, 24, and 38 have enhanced ERAD. The arrowhead indicates a slow migrating form of Ecm29 expressed in CHO mutant line 38 that is roughly the size of a high molecular mass brain isoform (see Fig. 2B). B, relative levels of Ecm29 and 19 S RC subunits (S10b, S14) or 20 S proteasome subunits in CHO lines. The numbers below each band (A and B) represent the amounts of Ecm29 or proteasome subunits relative to wild-type CHO cells as determined by densitometry of Western blots after SDS-PAGE and transfer to nitrocellulose. degraded less efficiently than in the wild-type parental cells, while the other three clones (23, 24, and 38) degraded ␣ 1antitrypsin Z more rapidly than wild-type CHO cells. Clones 23 and 24 also degraded the catalytic CT polypeptide more rapidly than the parental cells (42). Since overall proteasome activity was not altered in clones 23 and 24, it was concluded that toxin resistance was caused by increased coupling efficiency between toxin translocation from the ER lumen into the cytosol and toxin degradation in the cytosol (42).
Given our suspicion that Ecm29 recruits the 26 S proteasome to secretory quality control compartments, we compared Ecm29 levels in wild-type CHO and in each of the 6 toxinresistant clones. It is clear from the blots in Fig. 10A that clones 23 and 24 express 2-3-fold more Ecm29 protein than either wild-type CHO or the ERAD-deficient lines. Moreover, the increased ERAD in clones 23 and 24 is not accompanied by a general up-regulation of proteasome subunits (see Fig. 10B) or proteasome activity (42). Although the 205-kDa Ecm29 species is not increased in CHO clone 38, the other high ERAD mutant, a larger isoform of Ecm29 is uniquely present in this mutant, possibly being the same isoform as the one seen in mouse brain (see Fig. 2B). Even though the up-regulation of Ecm29 in toxinresistant CHO mutants exhibiting enhanced ERAD lends support to the idea that Ecm29 couples the 26 S proteasome to quality control compartments in the secretory/endocytic pathways, Ecm29 was not induced when HeLa cells were treated with brefeldin A, monensin, tunicamycin, or inhibitors of ER glycosidases, compounds that induce the unfolded protein response (UPR) pathway (43) (data not shown).
Mammalian Ecm29 Is Present in Karyoplasts and on Intranuclear Polyglutamine Inclusions-In most cases, nuclei stain for Ecm29 less intensely than the cytoplasm (Figs. 5, 6, and 8). However, significant nuclear staining was observed in human prostate L4 cells (see below), and the intensity of nuclear staining varied considerably in HeLa cells depending upon fixation methods as well as culture conditions (see Figs. [5][6][7][8]. Because the nuclear envelope and densely packed chromatin can make immunodetection of intranuclear components difficult, we employed cytochalasin B-mediated enucleation of HeLa cells to measure the nuclear/cytoplasmic distribution of Ecm29. As shown in Fig. 11, HeLa Ecm29 was distributed evenly between cytoplast and karyoplast fractions. The presence of Ecm29 in karyoplasts may reflect a substantial pool of intranuclear Ecm29 or Ecm29 may simply be enriched in perinuclear structures, e.g. centrosome or perinuclear ER. There is, however, no doubt that Ecm29 can enter the nucleus as shown by expression of polyglutamine-expanded ataxin-7. Expression of proteins containing an expanded polyglutamine tract (polyQ) causes a number of neurodegenerative diseases that include Huntington's and Kennedy's disease and six spinocerebellar ataxias (SCAs). One of the hallmarks of these diseases is the formation of intranuclear polyQ inclusions that accumulate components of the ubiquitin-proteasome system (44 -46). To determine whether Ecm29 accumulates in polyQ inclusions, we transfected human HeLa and L4 prostate cells with a plasmid encoding a His-tagged version of ataxin-7 containing 86 glutamines, or expressed in COS7 cells both the ataxin-7-Q 86 protein and mouse Ecm29 fused to GFP. Fixed cells were stained with anti-His to localize polyQ aggregates and either stained with anti-Ecm29-3, or visualized for the fluorescence of the Ecm29-GFP fusion protein. As seen with other components of the Ub-proteasome system, Ecm29 is indeed present on polyQ intranuclear inclusions (Fig. 12).
The Intracellular Distribution of Ecm29: A Summary-Volocity v2.0, a volume-rendering and three-dimensional quantification software, was used to estimate the relative amount of Ecm29 associated with specific compartments in HeLa cells. We found that 10% of Ecm29 was present at the centrosome, ϳ20% overlapped with ERGIC-53, about 30% of Ecm29 colocalized with the ER, and nearly 40% of Ecm29 was nuclear (Table I). By contrast, less than 4% of Ecm29 overlapped with the TGN marker. As expected, the amount of Ecm29 that overlapped with the ER/ERGIC compartments increased when cells were incubated simultaneously with both anti-KDEL and anti-ERGIC-53 (Table I). DISCUSSION Human Ecm29 Is a Non-stoichiometric Component of the 26 S Proteasome-Like its yeast homolog (29), mammalian (m) Ecm29 is a proteasome-associated protein. This is shown in Fig. 3 by the combined use of glycerol gradient sedimentation, native gel electrophoresis, and Western blotting. However, the properties of mEcm29 differ in two respects from those reported for yeast ecm29p. First, we have not observed HeLa or bovine brain Ecm29 bound to the 20 S proteasome either during purification or after attempts at reconstitution (Fig. 3 and   FIG. 11. Distribution of Ecm29 between nucleus and cytoplasm. HeLa cells grown on 18-mm circular coverslips were enucleated after treatment with cytochalasin B as described under "Experimental Procedures." Karyoplast (K) and cytoplast (C) fractions were extracted in 50 l of 0.25% Triton X-100 in 10 mM Tris-HCl, pH 7.5, 1 mM ATP, 1 mM DTT plus protease inhibitors, and subjected to SDS-PAGE (panel A) and immunoblotting with the indicated antibodies (panel B). W, whole cell extract; p80, 80-kDa cytoplasmic polypeptide recognized by anti-PA200 polyclonal antibodies (18); Nup62, nucleoporin 62. S10b and S14 are 19 S RC subunits; ␣6 is a 20 S proteasome subunit. The amount of Ecm29 in the karyoplast fractions was 47% (Exp. 1) and 48% (Exp. 2) as determined by densitometry. data not shown). Second, markedly variable expression levels in mouse organs (Fig. 2) rule out mEcm29 being a stoichiometric component of the 26 S proteasome. If mEcm29 clamps the 19 S RC to 20 S proteasomes as proposed for the yeast protein (29), it does so only in certain mammalian organs. Rather than functioning to stabilize the 26 S proteasome, we propose that mEcm29 is an adaptor that recruits the 26 S proteasome to specific intracellular locations requiring enhanced rates of protein degradation such as the centrosome, the cytoplasmic face of the ER and possibly polyQ aggregates.
Evidence That Mammalian Ecm29 Is an Adaptor-Several lines of evidence suggest that Ecm29 functions as an adaptor. Computer analyses of the Ecm29 sequence predict the protein to be composed almost exclusively of HEAT repeats (35). These protein modules form flexible, curved solenoids that enable HEAT-repeat-containing proteins to bind multiple protein ligands, and HEAT repeats are found in a number of adaptor proteins (47,48). In addition, Ecm29 contains a region homologous to VHS domains found in Hrs, STAM, and GGAs, proteins that serve as adaptors in endosomal sorting and vesicular trafficking (36,49). Finally, two regions in Ecm29 share homology to the ␤2 subunit of AP2 adaptor complexes, and potential clathrin-interacting motifs are present throughout the Ecm29 sequence (50). Thus, the sequence of Ecm29 suggests an adaptor function.
The intracellular distribution of Ecm29 also supports the idea that Ecm29 is an adaptor. Ecm29 and the 26 S proteasome are found on endosomes (Figs. 6 -8), and the presence of a VHS-like domain in Ecm29 suggests that it recruits the 26 S enzyme to endosomal membranes. Recently, Dong et al. (51) observed direct interaction between the 20 S proteasome subunit XAPC7 and Rab7, a GTPase that controls late endocytic events and is found on late endosomes. Similarly, Ecm29 may function as a molecule that recruits proteasomes to endosomes.

TABLE I Distribution of Ecm29 in HeLa cells
The relative amount of Ecm29 at each cellular location was calculated by the volume (calibrated pixels/m 3 ) of Ecm29 overlapping with each marker using Volocity v2.0. Ecm29 was detected using anti-Ecm29 -3 antibodies after methanol fixation of HeLa cells. Although proteasome activity is known to regulate the trafficking of several plasma membrane receptors (52)(53)(54)(55)(56)(57)(58), the role of the proteasome in endocytosis remains a mystery. It has been proposed that 19 S RC isopeptidase activity is needed for proper trafficking of the EGF receptor (52). Alternatively, 26 S proteasomes may degrade components involved in the endocytic process, such as c-Cbl (52), or membrane scaffolding proteins, such as PSD95 (59). Besides being found on endosomes, mEcm29 is located at two intracellular sites exhibiting high levels of Ub-dependent proteolysis: the centrosome and the ER/ERGIC compartments ( Fig. 5 and Table I). It has long been established that the centrosome is enriched in proteasomes especially when cells synthesize large amounts of abnormal proteins or when proteasomes are inhibited (60 -63). Proteasomes have likewise been shown to degrade secretory proteins in a process called ERAD (64 -67). Our finding that the intracellular distribution of mEcm29 overlaps the ER and ERGIC compartments raises the possibility that Ecm29 recruits proteasomes to membranes within these organelles. In fact, a role for Ecm29 in the ERAD pathway is indicated by overexpression of the protein in mutant CHO cells that exhibit increased ERAD activity (Fig. 10A). Localization of Ecm29 within the nucleus and on intranuclear polyQ inclusions (Fig. 12) does not argue against its playing a major role in proteasomal degradation of secretory or endocytic substrates, because the nuclear function of Ecm29 and 26 S proteasomes may be independent of their function in the cytoplasm. Moreover, a number of proteins involved in endocytosis are known to shuttle between nucleus and cytoplasm (68 -70).
Even the organ distribution of Ecm29 supports, to some extent, the idea that the protein plays a role in secretion and endocytosis. As shown in Fig. 2B, Ecm29 is particularly enriched in pancreas and brain, two organs very active in secretion and endocytosis, and it is virtually absent in heart, a non-secretory organ. On the other hand, Ecm29 is barely expressed in liver, a major secretory organ. If Ecm29 couples the 26 S proteasome to the ERAD pathway, in liver this function is presumably served by other components.
In summary, we have shown that mEcm29 is a 26 S proteasome-associated protein with varying levels of expression among mouse organs. We have also shown that Ecm29-proteasome complexes are present at centrosomes, on endosomes and on intranuclear polyQ inclusions. Up-regulation of Ecm29 in mutant CHO cells exhibiting increased ERAD and its apparent location on ER/ERGIC compartments lead us to propose that Ecm29 couples proteasomes to the ERAD quality control pathway and to functions occurring on endosomes and other membrane vesicles (71)(72)(73).