Distinct Modes of Ubiquitination of Peroxisome-targeting Signal Type 1 (PTS1) Receptor Pex5p Regulate PTS1 Protein Import*

Background: Pex5p translocates the cargo proteins into peroxisomes. Results: Mammalian RING peroxins, the Pex10p·Pex12p complex, catalyze Lys-linked, multiple monoubiquitination of Pex5p, assuring efficient export of Pex5p from peroxisomes. Conclusion: Distinct Lys- and Cys-linked modes of Pex5p ubiquitination control Pex5p-mediated matrix protein import. Significance: This work provides the first evidence of a functional role of mammalian RING peroxins in Pex5p ubiquitination. Peroxisome targeting signal type-1 (PTS1) receptor, Pex5p, is a key player in peroxisomal matrix protein import. Pex5p recognizes PTS1 cargoes in the cytosol, targets peroxisomes, translocates across the membrane, unloads the cargoes, and shuttles back to the cytosol. Ubiquitination of Pex5p at a conserved cysteine is required for the exit from peroxisomes. However, any potential ubiquitin ligase (E3) remains unidentified in mammals. Here, we establish an in vitro ubiquitination assay system and demonstrate that RING finger Pex10p functions as an E3 with an E2, UbcH5C. The E3 activity of Pex10p is essential for its peroxisome-restoring activity, being enhanced by another RING peroxin, Pex12p. The Pex10p·Pex12p complex catalyzes monoubiquitination of Pex5p at one of multiple lysine residues in vitro, following the dissociation of Pex5p from Pex14p and the PTS1 cargo. Several lines of evidence with lysine-to-arginine mutants of Pex5p demonstrate that Pex10p RING E3-mediated ubiquitination of Pex5p is required for its efficient export from peroxisomes to the cytosol and peroxisomal matrix protein import. RING peroxins are required for both modes of Pex5p ubiquitination, thus playing a pivotal role in Pex5p shuttling.

Peroxisomal proteins, including matrix and membrane proteins, are synthesized on free polyribosomes in the cytosol and post-translationally translocated to peroxisomes (1). Two types of peroxisome-targeting signals (PTSs), 2 PTS1 (2, 3) and PTS2 (4,5), are identified as topogenic signals for matrix proteins. In mammals, 14 PEX gene products, termed peroxins, have been isolated, 10 of which are involved in matrix protein import into peroxisomes (reviewed in Refs. 6 and 7). Pex5p and Pex7p are the cytosolic receptors for PTS1 and PTS2 proteins, respectively (8 -12). Pex5p recognizes newly synthesized PTS1 proteins in the cytosol, and the Pex5p⅐cargo complexes are targeted to peroxisome membranes by docking to membrane peroxins, Pex14p and Pex13p (13)(14)(15). Pex5p then releases the cargo proteins into the peroxisomal matrix, which is mediated by the association of Pex5p with a putative import machinery, including a docking complex (Pex14p and Pex13p) and a translocation complex comprising three RING peroxins, Pex2p, Pex10p, and Pex12p (6,7). Finally, Pex5p shuttles back to the cytosol in a manner dependent on ATP and AAA family peroxins, Pex1p and Pex6p (16,17).
Ubiquitination is a post-translational protein modification, in which the ubiquitin-activating enzyme (E1) transfers ubiquitin to a ubiquitin-conjugating enzyme (E2), and a protein-ubiquitin ligase (E3) catalyzes transfer of the ubiquitin moiety from ubiquitin E2 to the substrate (18). Ubiquitination of Pex5p has been demonstrated in yeast and mammals and conclusively regulates Pex5p function, especially in the export from peroxisome membrane to the cytosol. Yeast genetic approaches have revealed that Pex5p is ubiquitinated in two distinct modes. Monoubiquitination at the N-terminal conserved cysteine of Pex5p is required for the recycling as well as peroxisome matrix protein import (19,20). Polyubiquitination at the conserved two lysines of N-terminal region of Pex5p leads to the degradation of Pex5p (20 -23). These two types of ubiquitin modifications at the conserved cysteine and lysine(s) with similar functions are also reported in the PTS2 co-receptors, Saccharomyces cerevisiae Pex18p (24) and Pichia pastoris Pex20p (25,26). In mammals, a conserved cysteine near the N terminus of Pex5p (Cys 11 ) is monoubiquitinated via a thioester bond, being essential for Pex5p export (27)(28)(29). An E2 enzyme, yeast Pex4p, facilitates monoubiquitination of Pex5p (Cys 6 in S. cerevisiae) (19,20,30), whereas Ubc4 functions as E2 in polyubiquitination of Pex5p (21)(22)(23). The cysteine-linked ubiquitination (Cysubiquitination) of human Pex5p is reconstituted in the presence of E2 enzymes, UbcH5a/b/c, in a cell-free Pex5p import system (28). However, ubiquitination of Pex5p via a lysine residue is largely unknown in mammalian cells.
The RING (really interesting new gene) finger motif is found in over 300 proteins in mammals with diverse cellular functions and categorized as major E3 enzymes (31). RING peroxins Pex2p, Pex10p, and Pex12p possess similar domain structure with two transmembrane segments and a RING finger, showing the same membrane topology, in which the N-terminal and RING finger-containing C-terminal parts are exposed to the cytosol (see Fig. 2A) (6,7). Deficiency of any one of the RING peroxins results in typical import defects of peroxisomal matrix proteins of both the PTS1 and PTS2 types, but not membrane proteins, as seen in pex2, pex10, and pex12 cell mutants, including fibroblasts from patients with peroxisome biogenesis disorders (PBDs) (6,32). Pex5p is accumulated in peroxisome remnants in RING peroxin-impaired cell mutants (14,33), implying that RING peroxins are required at a step(s) downstream of Pex5p docking to Pex14p, most likely during the translocation of matrix proteins across the membrane (34,35). Involvement of RING peroxins in the Pex5p ubiquitination is shown in S. cerevisiae, where each RING finger of three RING peroxins has ubiquitin ligase activity as assessed by their self-ubiquitination in vitro (30,36). An in vitro ubiquitination assay and genetic analysis show that S. cerevisiae Pex12p mediates Pex4pdependent monoubiquitination of Pex5p (30,36), whereas Pex2p (30,36) and Pex10p (37) are implicated to be involved in the Ubc4-dependent polyubiquitination. Self-ubiquitination activity is shown in all three Arabidopsis thaliana RING peroxins in vitro (38). However, E3 activity of mammalian RING peroxins and their role in Pex5p ubiquitination remain unknown.
Here, we report that RING finger of human Pex10p possesses ubiquitin ligase activity with E2 UbcH5C and that the E3 activity is dramatically augmented by formation of a Pex10p complex with Pex12p. The Pex10p E3 activity is required for peroxisome biogenesis. Moreover, we demonstrate that the Pex10p⅐Pex12p complex monoubiquitinates Pex5p at multiple lysine residues in vitro. We also address two distinct types of Pex5p ubiquitination, cysteine-linked and lysinelinked ubiquitination, and their functional consequences in vivo.
To construct His-FLAG-ClPEX5L encoding N-terminally His-FLAG-tagged Pex5p, the BglII-AxyI fragment of the PCRamplified with primers His-FL-ClP5.Fw and ClP5.AxyRv, and the AxyI-NotI fragment of ClPEX5L from pGEX/ClPEX5L (14) were cloned together into the BamHI-NotI site of pcDNAZeo/His-UbcH5C in which UbcH5C had been removed. The epitope tagging is likewise performed for the shorter form of ClPEX5, ClPEX5S. Plasmids for various KR mutants of Pex5pL were generated by replacing several or all of the 18 lysines in Pex5pL at positions, 19,28,52,83,170,205,210,218,273,293,315,360,392,449,465,499,520, and 581 with arginine at the DNA level in pcDNAZeo/His-ClPEX5L-HA (29) using a PCR-based strategy. These KR mutants of Pex5pL were also expressed as GST fusion proteins by replacing wild-type PEX5L cDNA in pGEX/ClPEX5L with those for the KR mutants. All mutations and constructions were confirmed by sequencing.
Morphological Analysis-Immunostaining for CHO cells and human fibroblasts was performed with primary antibodies and secondary antibodies labeled with Alexa 488 and 568 (Invitrogen) as described (29). Cells were observed under microscopes (LSM510; Carl Zeiss). Images were acquired and analyzed with the LSM image browser (Carl Zeiss).
Immunoprecipitation-Cells were lysed in L-buffer (20 mM Hepes-KOH, pH 7.4, 150 mM NaCl, 1 mM DTT, 1 mM EDTA, 5 M MG132 (Peptide Institute, Osaka, Japan), and protease inhibitor mixture) containing 1% Triton X-100. For immunoprecipitation of FLAG-RING peroxins, L-buffer containing 1% digitonin (Wako, Osaka, Japan) was used. After centrifugation, the soluble fraction was used for immunoprecipitation with anti-FLAG M2-agarose (Sigma) or anti-HA antibody plus protein G-Sepharose (GE Healthcare) as described (29). Proteins bound to anti-FLAG-agarose were eluted with 100 g/ml FLAG peptides (Sigma) when required. To more highly purify His-FLAG-Pex5pL, the eluate was denatured in the presence of 1% SDS, diluted with 9 volumes of buffer containing 1% Triton X-100, and then purified with TALON cobalt-chelating resin under a high stringency condition by washing with buffer containing 1 M NaCl. Proteins were eluted with buffer containing 150 mM imidazole and concentrated by precipitation with TCA and subjected to SDS-PAGE and silver staining using a Sil-Best stain kit (Nacalai Tesque, Kyoto, Japan).
Purification of Recombinant Proteins-MBP and GST fusion proteins were purified with amylose resin (New England Biolabs) and glutathione (GSH)-Sepharose beads (GE Healthcare), respectively, as described (35). Purified MBP fusion proteins were eluted by MAL buffer containing 10 mM maltose. Pex5p and its variants were cleaved and isolated from the GST fusion proteins with PreScission protease as described (14). Histagged proteins were purified by affinity chromatography using TALON cobalt-chelating resin (Invitrogen) as described (35) and eluted with buffer containing 0.2 M imidazole. In the case of His-UbcH5C, eluted His-UbcH5C was dialyzed using a Slide-A-Lyzer cassette (Pierce) in dialysis buffer (40 mM Tris-HCl, pH 7.4, 50 mM NaCl, and 0.1% Triton X-100).
In Vitro Ubiquitination Assay-Recombinant His-tagged mouse E1 was kindly provided by H. Yasuda and K. Tanaka. To verify whether RING peroxins exhibit ubiquitin ligase activity, various E2s were first prepared as soluble lysates of Escherichia coli BL21(DE3) each expressing His-tagged E2s, including UbcH2A, UbcH2B, UbcH3, UbcH4, UbcH5B, UbcH5C, UbcH6, UbcH7, and UbcH8, in PBS containing 0.1% Triton X-100 and 1 mM PMSF. The ubiquitination assay was performed at 30°C for 1 h in 25-l reaction mixtures containing 150 ng of ubiquitin (Sigma), 10 ng of E1, 2.5 l of lysates of E. coli expressing each E2, and 1.5 g of MBP-RING peroxins in 40 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 2 mM ATP, 2 mM DTT. The reaction was terminated by the addition of SDS-PAGE sample buffer and heating at 95°C for 5 min. To prepare Cys-ubiquitinated Pex5p, a ubiquitination reaction was performed as described above, except that DTT was omitted.
The ubiquitination assay was likewise performed, except that 1 g each of MBP-RING fusion proteins, Pex5pL, and Pex5pLcomplexes and 250 ng of purified His-UbcH5C as an E2 were used. FLAG-HsPex10p immunopurified from K1/FLP10 cells was also used as an E3. Wild-type ubiquitin and its mutants, K48R, K63R, and K0 (Boston Biochem), were used where indicated.
In Vitro Import and Export Assays of Pex5p-Assays for in vitro import and export of Pex5p were performed as described (17). Briefly, 35 S-labeled Pex5pL and its variants were synthesized in vitro by a coupled transcription and translation of respective cDNAs. Import reaction was performed at 26°C for 60 min with 35 S-Pex5pL and postnuclear supernatant (PNS) from a PEX5-deficient CHO mutant ZP105 (9) in the absence of ATP for 60 min at 26°C. Imported 35 S-Pex5pL was verified by its resistance to 5 g/ml Proteinase K digestion. 35 S-Pex5pL export from the organellar fraction after the import reaction was assessed in the presence of the cytosolic fraction of ZP105, ATP, and the ATP-regenerating system (17). 35 S-Pex5pL was detected by a FujiX FLA5000 Autoimaging analyzer (Fujifilm) and quantified with Image Gauge software (Fujifilm).
Proteolytic Digestion of Pex5p-One l of the reticulocyte lysates containing in vitro synthesized 35 S-Pex5pL variants was incubated on ice for 15 min in 50 l of buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM DTT) containing 0.2 g/ml Proteinase K. The reaction was terminated by adding 0.5 mg/ml PMSF and analyzed by SDS-PAGE and autoradiography as described above.
Other Methods-Subcellular fractionation of various cultured cells was performed in homogenization buffer (5 mM Hepes-KOH, pH 7.4, 0.25 M sucrose) containing protease inhibitor mixture (Roche Applied Science), 5 M MG132, and 5 mM N-ethylmaleimide, as described (29). Western blot analysis was performed using various primary antibodies described above and horseradish peroxidase-conjugated secondary antibodies to mouse or rabbit IgG (GE Healthcare), or goat IgG (Cappel) with an ECL reagent (GE Healthcare).

RESULTS
Three RING Peroxins Interact with Each Other-We established three stable CHO cell lines, each expressing N-terminally FLAG-tagged RING peroxins: wild-type CHO-K1 expressing FLAG-Pex10p, termed K1/FLP10; pex12 mutant ZP109 expressing FLAG-Pex12p, 109/FLP12; and pex2 Z65 expressing FLAG-Pex2p, Z65/FLP2. In these cell lines, three FLAG-RING peroxins localized to peroxisomes, and morphologically and biochemically normal peroxisomes were discernible (data not shown). Upon immunoprecipitation of FLAG-Pex10p from K1/FLP10 cells lysed with 1% digitonin, endogenous Pex2p and Pex12p were coimmunoprecipitated (Fig. 1A, lanes 3 and 7), consistent with our earlier report (35). Likewise, in the immunoprecipitates of FLAG-Pex12p and FLAG-Pex2p, the two endogenous RING peroxins were recovered (Fig. 1A, lanes 2, 4, 6, and 8). Together, we interpreted these findings to mean that three RING peroxins interact with each other. Pex14p and Pex13p, components of the matrix protein import machinery, were detected at a lower level in the immunoprecipitates of the FLAG-RING peroxins (Fig. 1A, lanes 6 -8). Moreover, a smaller amount of Pex5p was also detectable. Specificity of the interactions between RING peroxins and other peroxins was confirmed by immunoblotting of the 70-kDa peroxisomal membrane protein and cytosolic lactate dehydrogenase (Fig. 1A) Control pex12 when lysed with 1% Triton X-100 (data not shown), consistent with our earlier report (35), suggesting that three RING peroxins formed a trimeric core complex involving Pex14p, Pex13p, and Pex5p as the subcomplex. Furthermore, exogenously expressed HA 2 -Pex10p was apparently decreased in pex12 mutant ZP109 and pex2 Z65 relative that in CHO-K1 and pex14 ZP161 ( . In contrast, lower levels of three RING peroxins were detected in fibroblasts from pex1 and pex6 PBD patients (Fig. 1C, lanes 1, 5, and 6). These results suggested that the deficiency of one RING peroxin affected the stability of two other RING peroxins, as observed in yeast (30,55). RING Finger of Pex10p Is Required for the Self-ubiquitination and the Complementing Activity-RING fingers of three RING peroxins are indispensable for the complementing activity of the impaired peroxisome biogenesis in respective pex mutant cells, but the molecular mechanisms underlying such activity remain unknown. Recent findings that RING fingers in many proteins possess E3 activity (31) prompted us to verify E3 activity in the RING fingers of mammalian RING peroxins ( Fig. 2A).
In an in vitro ubiquitination assay using respective recombinant MBP-RING fusion proteins with ATP, ubiquitin, E1, and various E2s, MBP-Pex10pC gave rise to high molecular mass ladder bands in the presence of UbcH4 and UbcH5C (Fig. 2B,    Immunoblotting with anti-ubiquitin antibody confirmed that MBP-Pex10pC was modified with ubiquitin ( Fig. 2C, bottom, lanes 1 and 5). However, no ubiquitin-conjugated bands were discernible with several RING finger mutants of Pex10pC, including C273A with the first Cys mutated to Ala, H290Q identified in a pex10 PBD patient (56), and C310G, representing the eighth Cys substituted to Gly (Fig. 2C), thereby confirming the RING-dependent self-ubiquitination activity of MBP-Pex10pC in vitro, as shown in the RING domain of S. cerevisiae Pex10p (19,30). MBP-Pex2pC showed a small amount of a single band with ubiquitin modification in the presence of UbcH4 and UbcH5C (Fig. 2B, two bottom panels, lanes 4 and 6) and a lesser amount with UbcH5B (lane 5). This result strongly suggested that MBP-Pex2pC has self-ubiquitination activity in vitro, although the activity is much lower than MBP-Pex10pC. MBP-Pex12pC showed no such self-ubiquitination activity with each of the E2s examined (Fig. 2B, two middle panels), in contrast to the E3 activity of the RING finger of Pex12p in S. cerevisiae (30) and A. thaliana (38). No self-ubiquitination of each mammalian RING finger was detectable in the in vitro ubiquitination reaction with S. cerevisiae Pex4p as an E2 (data not shown).
Next, we examined the peroxisome-restoring activity of FLAG-Pex10p RING variants by transfection of their cDNAs to fibroblasts from a pex10 PBD patient of complementation group B (PBDB-01) expressing Pex10p truncated in the RING finger (43). The impaired matrix protein import was complemented by wild-type FLAG-PEX10 expression as assessed by immunostaining with anti-PTS1 antibody (Fig. 2D, a), as in our earlier report (43). In contrast, none of the RING mutants, C273A, H290Q, and C310G, restored PTS1 protein import ( Fig. 2D, b-d). Like the wild-type (Fig. 2E, a and b), FLAG-Pex10p RING mutants C273A (c and d), H290Q, and C310G (data not shown) were localized to punctate structures, peroxisomes, as verified by co-localization with PTS1 proteins in CHO-K1 cells, but apparently at a lower expression level than the wild-type (Fig. 2F, lanes 1-5), indicative of the peroxisomal localization of Pex10p RING mutants. In conclusion, the RING finger of Pex10p is not essential for localization of Pex10p to peroxisomes but is required for peroxisome-restoring activity and E3 activity of Pex10p.
Pex12p RING Finger Enhances Pex10p E3 Activity-To investigate whether Pex10p RING interacting partners, Pex2p and Pex12p, show any effects on the E3 activity of the Pex10p RING, we performed the in vitro ubiquitination assay using MBP-Pex10pC in the presence of MBP-Pex2pC and MBP-Pex12pC. The self-ubiquitination activity of MBP-Pex10pC was significantly elevated by incubation with MBP-Pex12pC in an MBP-Pex12pC-dependent manner, as compared with the reaction with a control MBP-HA 2 (Fig. 3A, top, lanes 1-4). Such augmentation of MBP-Pex10pC ubiquitination was not detectable with MBP-Pex2pC-HA 2 (lanes 5-7). MBP-Pex12pC was also highly ubiquitinated in these assays (Fig. 3A, middle, lanes 1-7). A RING mutant, MBP-Pex12pC-C304S, does not interact with Pex10p RING (35). MBP-Pex12pC-C304S showed neither enhancement of the self-ubiquitination activity of MBP-Pex10pC nor its ubiquitination (Fig. 3B), thereby suggesting that the binding of MBP-Pex10pC to Pex12p-RING was a prerequisite for augmenting its E3 activity. On the other hand, in the presence of both MBP-Pex12pC and MBP-Pex2pC-HA 2 at a different ratio, MBP-Pex12pC-dependent augmentation in the level of ubiquitinated MBP-Pex10pC and MBP-Pex12pC was similar to that of MBP-Pex10pC only with MBP-Pex12pC (Fig. 3A, lanes 8 -10), hence suggesting that MBP-Pex2pC-HA 2 did not affect the ubiquitination of MBP-Pex10pC and MBP-Pex12pC. MBP-Pex10pC and MBP-Pex12pC showed a slightly additive effect in ubiquitination of MBP-Pex2pC-HA 2 (Fig. 3A, bottom, lanes 7-9, shaded arrowhead), possessing weak selfubiquitination activity with the same E2s as for MBP-Pex2pC (data not shown), implying that the MBP-Pex10pC and MBP-Pex12pC complex enhanced the self-ubiquitination activity of MBP-Pex2pC-HA 2 and ubiquitinated MBP-Pex2pC-HA 2 as a substrate. Neither E3 activity nor an additive effect was observed in the reaction with a set of MBP-Pex12pC and MBP-Pex2pC-HA 2 (data not shown). Taken together, these results indicated that the Pex12p RING finger functioned as an activator of the E3 ligase Pex10p. Pex10p⅐Pex12p Complex Ubiquitinates PTS1 Receptor Pex5p-Monoubiquitination of the conserved cysteine residue at the N-terminal region of the PTS1 receptor Pex5p is shown to be a key regulation of its shuttling between the cytosol and peroxisomes (20,27,29,57). Although RING peroxin-dependent ubiquitination of Pex5p has been shown in yeast (30,37), a role of RING peroxin in Pex5p ubiquitination remains unknown in mammals. Therefore, we investigated whether Pex5p is a substrate of Pex10p RING E3. In an in vitro ubiquitination reaction, Pex5pL was detected in several slower migrating bands, presumably representing ubiquitinated Pex5pL, when incubated with wild-type MBP-Pex10pC plus MBP-Pex12pC (Fig. 4A,  lane 4), not with MBP-Pex10pC or MBP-Pex12pC alone (lanes 1 and 2). E3 activity of MBP-Pex10pC was not altered in the absence or presence of Pex5pL as assessed by ubiquitination of MBP-Pex10pC and MBP-Pex12pC (Fig. 4A, lanes 3 and 4). Furthermore, after the in vitro ubiquitination assay, ubiquitin-conjugated Pex5pL-FLAG was detected only in the immunoprecipitate of wild-type Pex5pL-FLAG, not in that of the K-less Pex5pL-FLAG in which all Lys residues had been replaced with Arg (data not shown), hence confirming the specific ubiquitination of Pex5pL. In contrast, such higher mass bands of Pex5pL were not discernible when the assay was performed with MBP-Pex10pC-WT plus a Pex12pC RING mutant, MBP-Pex12pC-C304S, defective in augmenting Pex10pC-WT ubiquitination (Fig. 4A, lanes 5 and 6).
Furthermore, the in vitro ubiquitination assay was likewise performed using an E3 FLAG-Pex10p that had been immunoprecipitated from K1/FLP10 cells. FLAG-Pex10p immunocomplex, apparently with a low level of endogenous Pex12p, ubiquitinated MBP-Pex12pC, but not MBP-Pex12pC-C304S, where the self-ubiquitinated forms of FLAG-Pex10p were barely detectable (Fig. 4B, lanes 1-4). Accordingly, the full-length Pex10p also exerted the ubiquitin ligase activity in vitro as Pex10pC did. Moreover, Pex5p was detected as several bands with higher masses, presumably ubiquitinated Pex5pL, as well as the authentic size, upon incubation with the FLAG-Pex10p immunocomplex (Fig. 4B, lane 5). The extent of Pex5pL ubiquitination was not altered by the addition of wild-type and C304S-type MBP-Pex12pC (Fig. 4B, lanes 6 and 7). Apparent ubiquitination of Pex5pL in the absence of MBP-Pex12pC can be explained by the presence of endogenous Pex12p in the FLAG-Pex10p immunoprecipitates as a complex (see Fig. 4B, endo. Pex12p panel). Essentially the same results were obtained with Pex5pS in place of Pex5pL (data not shown).
To determine the mode of Pex5p ubiquitination, we performed an in vitro ubiquitination assay with ubiquitin mutants, K48R and K63R, giving rise to abrogation of Lys 48 -, Lys 63linked ubiquitin chain elongation, and K0, resulting in no elongation. No difference was discernible between the wild type and mutants in higher mass bands, the ubiquitinated Pex5pL (Fig. 4D,  top, lanes 5-8), strongly suggesting that Pex5pL was monoubiquitinated at multiple lysine residues by the Pex10p⅐Pex12p complex. Such a multimonoubiquitination was also evident in the selfubiquitination of Pex10p (Fig. 4D, middle, lanes 1-8).
In vitro ubiquitination assays described above demonstrated the covalent ubiquitination of Pex5p via lysine residues. Monoubiquitination was recently identified in the conserved cysteine residue of mammalian Pex5p (27)(28)(29)58). As a step to identifying an E3 for Cys-ubiquitination of Pex5p, we examined whether RING peroxins catalyze the DTT-sensitive ubiquitination of Pex5p in vitro. As shown in Fig. 4A, MBP-Pex10pC plus MBP-Pex12pC, not MBP-Pex10pC or MBP-Pex12pC alone, exhibited ubiquitination of wild-type Pex5pL, but no difference was discernible in the slowly migrating bands in the absence or presence of DTT (Fig. 4E, lanes 1-4 and 6 -9). MBP-Pex2pC showed no effect on ubiquitination of wild-type Pex5pL by the Pex10pC⅐Pex12pC E3 complex under both conditions (Fig. 4E,  lanes 5, 10, 15, and 20), where no E3 ligase activity for Pex5pL was detectable with MBP- Pex2pC (lanes 11-14 and 16 -19). Pex5pL-C11A, a Pex5pL variant with a mutation of the ubiquitinated Cys 11 residue to Ala (29), was ubiquitinated in the presence of MBP-Pex10pC and MBP-Pex12pC, like wild-type Pex5pL, only in a DTT-resistant manner (Fig. 4E, lanes 4 and 9). Taken together, we conclude that Pex5p is a substrate of the ubiquitin ligase Pex10p⅐Pex12p complex via lysine residues in vitro. However, we could not exclude the possibility that the Pex10p⅐Pex12p RING complex requires other potential E2 enzyme to catalyze Cys-ubiquitination of Pex5p.
Pex14p, but Not Pex13p, Interferes with Pex5p Ubiquitination-To investigate whether Pex5p ubiquitination shows any effects on PTS1 protein import, Pex5p proteins were fully loaded with the interacting proteins and assayed as substrates for the in vitro ubiquitination reaction. As compared with Pex5pL, the ubiquitination of Pex5pL in a complex with Pex14p was signif-icantly inhibited and completely abrogated when mixed with Pex14p and PTS1 protein (Fig. 5A, lanes 2, 6, and 8). Pex5pL loaded with PTS1 protein was less efficiently ubiquitinated  tially regulated by its complex formation with proteins, including the cargoes and the constituent peroxins in the peroxisome import machinery. No ubiquitin modification was observed in PTS1 protein, Pex14p, and Pex13pN in the complexes with Pex5p (Fig. 5A, lanes 4, 6, 8, and 10), strongly suggesting that ubiquitin ligase activity of the Pex10p⅐Pex12p complex is highly specific to Pex5p. Next, we assessed whether the ubiquitination of Pex5p alters the affinity to the binding partners. Pex5p was initially ubiquitinated with the Pex10p⅐Pex12p complex in vitro, and the reaction mixtures were then subjected to pulldown assays using Pex5p-interacting proteins fused with GST (Fig. 5B). Both the ubiquitinated and unmodified forms of Pex5pL were detected nearly at the same level in the bound fractions to GST-EGFP-PTS1 and GST-Pex14p, but not to GST (Fig. 5B, lanes 1-8), thereby indicating that the ubiquitination did not alter the interaction of Pex5pL with PTS1 protein and Pex14p. In contrast, only non-ubiquitinated Pex5pL was detectable in the fraction bound to GST-Pex13pN (lanes 9 and 10). Together with the results described above (Fig. 5A), we interpreted these findings to mean that the Pex10p⅐Pex12p complex efficiently ubiquitinated Pex5pL in the complex with Pex13p, not Pex14p, and the ubiquitinated Pex5pL was dissociated from Pex13p.
In Vivo Ubiquitination of Pex5p via Cysteine and Lysine Residues-In mammalian cells, two types of modified Pex5p were identified: a DTT-sensitive form with monoubiquitination at a conserved cysteine residue, which is associated with peroxisomes in rat liver (28), rat hepatoma Fao cells (29,58), and HeLa cells (58) and a DTT-insensitive form with unknown modification in the cytosol in Fao and HeLa cells (29,58). We performed subcellular fractionation of HEK293 cells in homogenization buffer containing N-ethylmaleimide to inhibit deubiquitination of the ubiquitin-modified proteins (59). In immunoblot analysis of Pex5p, Pex5p in the PNS fraction was mostly recovered in the cytosolic fraction, and the remainder (about 15%) was in the organellar fraction in the absence and presence of DTT (Fig. 6A, open arrowhead; note that organellar fractions were loaded as 6-fold aliquots of PNS and cytosolic fractions), as in other mammalian cells (14,29). A Pex5p-related band with slower migration was similarly discernible in the cytosolic and organellar fractions in a DTT-resistant manner (Fig. 6A, shaded  arrowhead), suggesting a covalent modification of Pex5p. Furthermore, an additional band with a slightly lower migration adjacent to the DTT-resistant Pex5p-related band was detected only in the organellar fraction in the absence of DTT (Fig. 6A, lane 3, solid arrowhead), which then disappeared upon treatment with DTT (lane 6). During cell fractionation, N-ethylmaleimide was essentially required to detect the DTT-sensitive modification of Pex5p (Fig. 6A, lanes 7 and 8). These DTTresistant and -sensitive modifications of Pex5p were likewise detected in mouse embryonic fibroblasts and mouse leukemic monocyte macrophage cell line RAW264.7 (data not shown). The DTT-sensitive modification most likely corresponds to a monoubiquitination at the Cys 11 residue of Pex5p in Fao cells (29). Treatment with MG132, a proteasome inhibitor, gave rise to an additional DTT-resistant band of Pex5p with slower migration than Lys-ubiquitinated Pex5p (Fig. 6A, lane 11, solid  dot), where neither intracellular localization nor protein stability of Pex5p was altered (lanes 9 -12). Together, these results strongly suggested two distinct types of modification of Pex5p in mammalian cells.
To assess the DTT-resistant modification of Pex5p in vivo, endogenous Pex5p in the lysates of HEK293 cells was pulled down with GST-Pex14p and analyzed under the reducing condition (Fig. 6B). Immunoblotting with anti-Pex5p antibody showed that GST-Pex14p, not GST, recovered both endoge- In vitro Ub assay for Pex5pL pull-down with GST fusion proteins nous Pex5p as doublet bands of ϳ80 kDa, most likely containing the short and long isoforms of Pex5p (9) (Fig. 6B, top, lanes  1, 3, and 6, open arrowhead), and a slower migrating 90-kDa band of covalently modified Pex5p (solid arrowhead). When HA 2 -Ub was expressed in HEK293 cells, an HA-positive band with corresponding migration to the 90-kDa Pex5p-related band was specifically recovered with GST-Pex14p (Fig. 6B,  middle panel, lane 4, solid arrowhead), suggesting that the 90-kDa band was Pex5p covalently conjugated with HA 2 -Ub. In HEK cells expressing HA 2 -Ub, an additional mobility shift of the more slowly migrating forms of endogenous Pex5p was hardly detectable in the cell lysates (Fig. 6B, lanes 1 and 2) and both cytosolic and organellar fractions (data not shown). We interpreted these data to mean that a portion of endogenous Pex5p was covalently ubiquitinated, but the exogenously expressed HA 2 -Ub was much less efficiently utilized in the Pex5p ubiquitination. Taken together, these data strongly suggested that endogenous Pex5p was ubiquitinated at the lysine residue mainly in the cytosol, in addition to the Cys 11 ubiquitination in peroxisomes (27,29). We next verified whether Pex5p is ubiquitinated in peroxisome-deficient CHO cell mutants. In wild-type CHO-K1 cells, α-LDH most of the unmodified Pex5p was recovered in the cytosolic fraction, and about 10% was in the organellar fraction (Fig. 6C,  top panels, lanes 1-3), as in HEK cells (Fig. 6A). A small amount of DTT-sensitive Pex5p was also detectable in the peroxisomal fraction (Fig. 6C, top panels, lanes 1-3, solid arrowhead). Covalent ubiquitination of Pex5p in CHO-K1 cells was apparently much less than in HEK cells (Fig. 6A) and HeLa cells (58). It was noteworthy that Pex5p with DTT-sensitive modification, most likely Cys-ubiquitinated Pex5p, appeared to be accumulated in the organellar fractions in Pex5p export-deficient CHO mutants pex1 ZP107, pex6 ZP164, and pex26 ZP167 (17) (Fig.  6C, top, lanes 6, 9, and 12, solid arrowhead), where the amount of unmodified Pex5p in the PNS and cytosolic fractions was significantly reduced in ZP107 and ZP164 (lanes 4 -9, open  arrowhead), as reported in PEX1-and PEX6-deficient fibroblasts (33). In PEX1-deficient fibroblasts, similar DTT-sensitive modification of Pex5p accumulated in the organellar fraction and lower stability of unmodified Pex5p in the cytosolic fraction were observed (Fig. 6D, lanes 3 and 4), in contrast to normal fibroblasts (lanes 1 and 2), suggesting that an export-competent, Cys-ubiquitinated form of Pex5p accumulated in peroxisomes in Pex5p export-defective cells. On the other hand, no DTT-sensitive modification of Pex5p was discernible in the respective organellar fractions in RING peroxin-deficient CHO mutants, pex12 ZP109 and pex2 Z65 (Fig. 6C, top panels,  lanes 15 and 18), and in PEX10-deficient PBD fibroblasts (Fig.  6D, lane 6), despite relatively higher accumulation of unmodified Pex5p in these organellar fractions, as shown in fibroblasts from PBD patients defective in any of three RING peroxins (33). Pex5p in the cytosolic fraction was decreased in pex12 ZP109 and pex2 Z65 (Fig. 6C, top panels, lanes 14 and 17) and in PEX10-deficient PBD fibroblasts (Fig. 6D, lane 5), which suggests that the Pex1p⅐Pex6p AAA complex plays a protective role against proteasomal degradation of cytosolic Pex5p. A lesser but significant level of three RING peroxins was discernible in the organellar fraction of pex1 ZP107 as compared with those of wild-type CHO-K1 (Fig. 6C, bottom panels, lanes 1 and 2). However, in pex2 Z65 and pex12 ZP109, RING peroxins, Pex10p plus Pex12p and Pex2p plus Pex10p, respectively were at a significantly lower or barely detectable level (Fig. 1B) (60) as well as in PBD fibroblasts (Fig. 1C). Taken together, these results strongly suggested that RING peroxins were indispensable for Cys-ubiquitination of Pex5p in mammalian cells, as in S. cerevisiae in which the RING finger of Pex12p and Pex2p mediated mono-and polyubiquitination of Pex5p, respectively (30).
Lys-linked Monoubiquitination of Pex5p Mainly Takes Place at Lys 520 -To further investigate ubiquitin modification of Pex5p in vivo, we performed an immunoprecipitation assay using HEK cells transiently co-expressing N-terminally Hisand FLAG-tagged Chinese hamster Pex5pL (HF-Pex5pL) (29) and HA 2 -Ub. In whole cell lysates under the reducing conditions, immunoblotting with anti-His antibody visualized HF-Pex5pL in two bands: a major 90-kDa band (Fig. 7A, middle  top panels, lane 2, open arrowhead) and a 95-kDa one (solid arrowhead), probably corresponding to those of endogenous, unmodified Pex5p and a Lys-linked ubiquitinated one, respectively, as detected with anti-Pex5p antibody (top panel, lane 1 and Fig. 6, A and B). Although both bands of HF-Pex5pL were immunoprecipitated with anti-FLAG antibody (Fig. 7A, first  and second rows of panels, lane 6), only the 95-kDa protein, and not the 90-kDa one, was cross-reactive with anti-ubiquitin antibody (third row of panels, lane 6, solid arrowhead). Co-expression of HA 2 -Ub gave rise to an additional ubiquitin-positive band in the immunoprecipitate of HF-Pex5pL, indicative of HA 2 -Ub-conjugated HF-Pex5pL (Fig. 7A, first, second, and third rows of panels, lane 8, shaded arrowhead). Similar Lyslinked ubiquitination of Pex5p was observed in CHO-K1 cells transiently expressing the shorter form of Pex5p as well as the longer form (Fig. 7B, lanes 2, 3, 8, and 9). Taken together, these results strongly suggested that exogenously expressed Pex5p was, like endogenous Pex5p, covalently ubiquitinated.
As a step to delineating a role of Pex5p ubiquitination via a lysine residue, we attempted to determine the ubiquitination site(s) in Pex5pL. Ubiquitin-targeted lysine residues were not directly identified by mass spectrometric analysis (61) in tryptic digests of the modified form of HF-Pex5pL (Fig. 7A, lane 6), in which many peptide fragments derived from ubiquitin were indeed identified besides those from Pex5p (data not shown). Next, we constructed C-terminally HA-tagged Pex5pL (Pex5pL-HA) and its mutants with various numbers of Lys 3 Arg substitutions of, in total, 18 lysine residues and verified them for ubiquitination in CHO-K1 cells. In an immunoblot, the Lys-ubiquitinated, 95 kDa band of Pex5pL-HA was hardly detectable in CHO-K1 cells expressing several mutants containing K520R substitution and a K-less mutant with 18 Lys 3 Arg mutations (Fig. 7C, lanes 12-16, solid arrowhead), whereas other variants showed the 95-kDa band as wild-type Pex5pL-HA (lanes 2-11). Only the K520R mutation gave rise to no Lys-ubiquitinated form of Pex5pL-HA, as shown with the K520R mutant (Fig. 7B, lanes 5 and 11) like the K-less mutant (lanes 4 and 10), suggesting that exogenously expressed Pex5pL was mainly ubiquitinated at Lys 520 . Moreover, a KR mutant, 520K, where all lysine residues other than Lys 520 were mutated to arginine showed a distinct 95-kDa band, like those noted with wild-type Pex5pS-HA and Pex5pL-HA at a lower level (Fig. 7B, lanes 2, 3, 6, 8, 9, and 12, solid arrowheads), thereby supporting the specific ubiquitination of Pex5p at Lys 520 .
Next, we verified Lys-ubiquitination of Pex5p in several peroxisome-deficient CHO cell mutants. A distinct ubiquitin-modified, 95-kDa protein band was detected in pex12 ZP109, pex2 Z65, pex14 ZP161, pex1 ZP107, and membrane assemblydefective pex19 ZP119, each expressing wild-type Pex5pL-HA, at a higher level than in normal CHO-K1, whereas the 95-kDa band was not discernible in those expressing K520R and K-less mutants (Fig. 7D, solid arrowhead). These results suggested that Pex5p ubiquitination at Lys 520 probably took place in the cytosol in a RING peroxin-independent manner. The Pex5pL-520K mutant was likewise ubiquitinated in pex2 Z65, pex12 ZP109, and pex1 ZP107 (data not shown), supporting the cytosolic ubiquitination of Pex5p. In pex10 fibroblasts derived from a complementation group B patient with Zellweger syndrome (43), Pex5p was expressed at a low level but slightly higher than that in fibroblasts from a control, where no apparently distinct Pex5p bands were discernible between these two types of cells (data not shown).
To investigate whether Pex10p directly ubiquitinates Lys 520 of Pex5pL, various KR mutants of Pex5pL were verified by an in vitro ubiquitination assay using full-length Pex10p and Pex12p as in Fig. 4C. The Pex5pL-K520R mutant was ubiquitinated to the same extent as the wild type (Fig. 7E, lanes 1-4 and 13), hence suggesting that the Pex10p⅐Pex12p complex ubiquitinated Pex5pL not specifically at Lys 520 but at other lysine residues, consistent with RING peroxin-independent ubiquitination of Pex5pL at Lys 520 in vivo (Fig. 7D). Based on the results by thorough analyses using KR mutants, including 3KR, 6KR, 9KR, 13KR in which 3, 6, 9, and 13 lysine residues were respectively mutated to arginine as shown in Fig. 7C, bottom panel, and K-less mutants, it was most likely that the Pex10p⅐Pex12p E3 complex ubiquitinated Pex5pL preferentially at Lys 449 and Lys 499 (Fig. 7E). Essentially the same results were obtained from in vitro Pex5p ubiquitination assays using MBP-Pex10pC and MBP-Pex12pC (data not shown). We interpreted these results to mean that two different types of covalent ubiquitin modification have taken place in Pex5p: one at Lys 520 by cytosolic unknown factor(s) in a peroxisome-independent manner and the other at multiple lysine residues except for Lys 520 catalyzed by the Pex10p⅐Pex12p complex.
Ubiquitination of Pex5p via Lysine Residues Attenuates Its Export from Peroxisomes to the Cytosol-As a step to addressing a physiological role, if any, of the Lys ubiquitination of Pex5p, we investigated whether the ubiquitination modulates Pex5p interaction with its partners, including PTS1 cargoes and the docking complexes on peroxisomes, Pex14p and Pex13p. Upon immunoprecipitation of KR variants of Pex5pL-HA, Pex14p was co-immunoprecipitated at a relatively similar level as the wild-type Pex5pL-HA, His-Pex5pS-HA, and the K-less mutant (Fig. 7B, lanes 8 -12). Pex13p was likewise co-immunoprecipitated with wild-type and KR mutants of Pex5pL-HA but not efficiently with His-Pex5pS-HA. Furthermore, in in vitro binding assays, the three Pex5pL KR mutants directly bound to Pex14p, Pex13p, and a PTS1 protein, indistinguishably from the wild-type Pex5pL and Pex5pS (Fig. 7F). These results suggested that the ubiquitination of Pex5p at Lys 520 and possibly at other lysine residues did not alter Pex5p binding to the cargo proteins, Pex14p, and Pex13p.
Next, we investigated whether the Lys-ubiquitination in Pex5p is involved in the Pex5p shuttling between the cytosol and peroxisomes. In the in vitro Pex5p import assay (17) using 35 S-labeled Pex5pL and PNS fraction from pex5 ZP139 cells (9), wild-type Pex5pL and its KR mutants, including K-less, K520R, and 520K, were detected at a similar level as a Proteinase K-resistant form (Fig. 8A, top and bottom, lanes 5 and 10), indicative of peroxisomal import (3,17), hence suggesting that the Lysubiquitination of Pex5p was not involved in the import steps into peroxisomes, as well as the Cys-ubiquitination (29,58). In in vitro Pex5p export assays using organelle-associated 35 S-Pex5pL (Fig. 8A, lanes 3 and 8), the amount of wild type and K520R of 35 S-Pex5pL decreased in the organellar fraction in a time-dependent manner, apparently with a concomitant increase in the level of both types of 35 S-Pex5pL in the cytosolic fraction (Fig. 8B, left panels), thereby demonstrating the export of wild-type and K520R-type 35 S-Pex5pL from peroxisomes. Such Pex5pL export was confirmed by verifying the sensitivity of the exported 35 S-Pex5pL variants to the treatment with Proteinase K (Fig. 8B, right panels). On the other hand, K-less and 520K mutants of 35 S-Pex5pL in the organellar fraction remained nearly at the same level throughout the export reaction in a Proteinase K-resistant form (Fig. 8B, left and right panels, lanes 4 -6 and 10 -12). The export efficiency of 35 S-Pex5pL variants was determined by taking the ratio of the respective 35 S-Pex5pL mutants in the cytosolic fraction (S) to those in the cytosol plus organellar fractions (S ϩ P) (Fig. 8C). The export rates of K-less and 520K mutants were about two- thirds those of wild-type and K520R mutant (Fig. 8C). The absence of a readily visible difference between the 35 S-Pex5pL KR mutants in the cytosolic fraction was probably due to the 35 S-Pex5pL variants, including those associated with but not imported to peroxisomes as well as a nonspecifically bound one (3,17). To address the possibility that KR mutations affect the conformation of Pex5p, KR mutants of Pex5pL were assessed by partial digestion with Proteinase K. When cell-free synthesized . KR mutation affects Pex5p export from but not import to peroxisomes. A, import of KR mutants of Pex5p. An in vitro Pex5p import assay was performed with cell-free synthesized 35 S-Pex5pL variants and the PNS fraction of CHO pex5 ZP105 in the absence of ATP as described under "Materials and Methods." After the import reaction, the reaction mixtures were mock-treated (Ϫ) or treated with 5 g/ml Proteinase K (ϩ) and separated into cytosolic (S) and organellar (P) fractions. Equal aliquots of total reaction mixtures (Input) and S and P fractions were analyzed by autoradiography and immunoblotting (WB) with antibodies to Pex14p and lactate dehydrogenase (LDH). Note that lactate dehydrogenase remained in the S fraction after the treatment with Proteinase K due to its resistance to Proteinase K (17). B, export of Pex5p KR mutants. Pex5p variants imported to peroxisomes (lanes 3 and 8 in A) were verified for the export for 30 min at 26°C. At each time point, the reaction mixtures were mock-treated (Ϫ, left panels) or treated with 5 g/ml Proteinase K (ϩ, right panels), fractionated, and analyzed as in A. C, quantification of the export of Pex5p variants. 35 S-Pex5pL variant bands, prior to the Proteinase K treatment after the export assays in B, were quantitated in three independent experiments. The ratios of 35  wild-type 35 S-Pex5pL and its variants were incubated at a low concentration of Proteinase K, similar proteolytic fragment patterns were detectable between the wild-type Pex5pL and all of the KR mutants, K-less, K520R, and 520K (Fig. 8D, lanes  7-10). These band patterns were very similar to those of human normal Pex5pL after Proteinase K treatment, where three major protease-resistant bands were identified as the fragments containing the C-terminal TPR regions (62). On the other hand, PTS1 cargo did not bind to two Pex5pL mutants, one with the mutation N519K and the other with G522E, equivalent to human N526K of a PBD patient (8,63) and a pex5 CHO ZP139 (9), respectively. Upon digestion with Proteinase K of these two mutants, more distinct 42-kDa doublet bands (Fig. 8D, lanes 11  and 12, open arrowhead), but not the 35-kDa fragment (solid arrowhead), were detected. These proteolytic profiles were similar to that of human Pex5pL harboring N526K that caused the conformational change in the region preceding the C-terminal TPR domains (64). Given the finding that the K-less mutant of Pex5pL interacts with the PTS1 protein as well as Pex14p and Pex13p like wild-type Pex5p in vitro (Fig. 7F) and in vivo (Fig. 7B), the K-less and 520K mutants probably share the similar conformation with wild-type Pex5pL, whereas they are less efficiently exported from peroxisomes. Collectively, we interpreted these results to mean that the ubiquitination of Pex5p at multiple lysine residues, rather than at Lys 520 , ensures the export of Pex5p from peroxisomes to cytosol, in addition to the prerequisite role of the ubiquitination at Cys 11 in the export process (29,58). Next, we expressed KR mutants of Pex5pL in a pex5 CHO mutant ZP139 and assessed their complementing activity. K520R and K-less mutants restored the impaired PTS1 protein import in ZP139, apparently with ϳ80 and 60% efficiency, respectively, as compared with wild-type Pex5pL (Fig. 9B, left). Moreover, these two KR mutants were more significantly lower, ϳ50 and ϳ20% of the wild type, in restoring the defective import of catalase (Fig. 9, A and B, right). The 520K mutant was similar to the K-less mutant in the complementing activity (Fig.  9, A and B), hence suggesting that ubiquitin modification of Pex5pL at Lys 520 was dispensable for peroxisomal matrix protein import. Partial defects of K-less and 520K mutants in restoring the activity in peroxisomal matrix protein import (Fig. 9, A and B) were well correlated with their lower export efficiency (Fig. 8, B and C). It is more likely that the ubiquitination of Pex5p at multiple lysine residues besides Lys 520 is required for the efficient protein import to peroxisomes. To obtain more insight into the role of Lys 520 in Pex5p function, the 520K mutant yielding ubiquitination only at Lys 520 was subjected to a pull-down assay. The ubiquitinated form of the 520K mutant expressed in the cytosol was defective in the interaction with GST-EGFP-PTS1, not GST-Pex14p (Fig. 9C, lanes 1, 3,  and 7, solid arrowhead), whereas the unmodified 520K was recovered in the bound fraction of both GST-EGFP-PTS1 and GST-Pex14p with the same efficiency as a K-less mutant ( lanes  1-4, 7, and 8, open arrowhead). These results indicated that ubiquitination of Pex5p at Lys 520 specifically abrogates the binding of Pex5p to the PTS1 cargo protein, most likely in the cytosol (Fig. 7D). Therefore, Pex5p ubiquitination at Lys 520 in vivo might play a role in self-inhibition of Pex5p in the cargo recognition in the cytosol rather than a regulatory one in the export process. However, we do not exclude the possibility that Lys 520 of Pex5p plays a role in the peroxisomal protein import at unknown step(s) because K520R mutant is significantly low in the complementing activity (Figs. 9, A and B).

DISCUSSION
Identification as an E3 of Mammalian RING Peroxins-RING peroxins were postulated to function as a potential E3 about a decade ago (65,66) by the analogy to the domain structure. Our in vitro ubiquitination assay using recombinant mammalian RING peroxins demonstrated that the RING finger of Pex10p and Pex2p, but not of Pex12p, self-ubiquitinated in the presence of the E2 enzymes UbcH5C and UbcH4 (Fig. 2). In contrast, all of the three RING fingers of RING peroxins in S. cerevisiae and A. thaliana exhibited E3 activity in vitro with both Pex4p and Ubc4 (30,36) and human UbcH5B (38), respectively. Interestingly, Pex12p RING finger with no enzymatic activity significantly augments the E3 activity of Pex10p (Figs. 3 and 4), indicating that Pex12p functions as an accelerator of the Pex10p E3 enzyme. In S. cerevisiae, E3 activity was similarly enhanced by each of two pairs of RING peroxins, Pex10p⅐ Pex12p and Pex10p⅐Pex2p, in a Pex4p-and Ubc4p-dependent manner, respectively (36). However, the situation is slightly different from the mammalian case because each RING finger of the three RING peroxins of S. cerevisiae and A. thaliana has E3 ligase activity (36,38). Several RING finger proteins are known to function as heteromeric complexes, such as Mdm2⅐MdmX (67), BRCA1⅐BARD1 (68), and Ring1b⅐Bmi1 (69). In these heterocomplexes, each RING domain of MdmX, BARD1, and Bmi1 appears to lack the E3 activity (67)(68)(69). In the case of BRCA1⅐BARD1, BARD1 also structurally stabilizes the heterocomplex (70). These notions are consistent with the putative role of Pex12p in the mammalian Pex10p⅐Pex12p E3 complex. Yeast and mammalian RING peroxins share several properties, including that (i) the RING finger of Pex10p directly interacts with RING fingers of Pex2p and Pex12p (35,36), and (ii) RING peroxins form a complex in vivo, and one RING peroxin is a prerequisite for the stable expression of the other two RING peroxins ( Fig. 1) (30, 55, 71). These findings strongly support the notion that RING peroxins function as an E3 enzyme complex, whereas the specificity of E2 appears to vary in species. A typical RING finger peroxin, Pex2p, showed weak self-ubiquitination activity but no effects on the E3 activity of Pex10p in the absence and presence of Pex12p (Fig. 3). However, we do not exclude the possibility that an optimal E2, neither UbcH5C nor UbcH4, is required for the E3 activity, if any, of Pex2p and Pex12p.
Role of RING Peroxins in Pex5p Ubiquitination-We identified here that mammalian Pex5p is indeed monoubiquitinated at multiple lysine residues by the Pex10p⅐Pex12p complex in the presence of UbcH5C in vitro. Mutation in the Pex10p RING motif eliminated both the E3 activity and the peroxisome-restoring activity (Fig. 2, B and D), indicating an essential role of Pex10p-E3 activity in peroxisomal matrix protein import. Distinct roles of RING peroxins have been defined in yeast. In vitro ubiquitination assay showed that S. cerevisiae Pex12p catalyzes Pex4p-dependent monoubiquitination of Pex5p (30), which is consistent with complete loss of the monoubiquitinated form of Pex5p in the Pex12p RING domain deletion mutant (30). Both S. cerevisiae Pex2p (30) and Pex10p (37) polyubiquitinate Pex5p in vitro, suggesting that they are E3s for polyubiquitination of Pex5p. Similarly, in S. cerevisiae, the PTS2 co-receptor, Pex18p, is mono-and polyubiquitinated by distinct pairs of E2/E3 enzymes: Pex4p plus Pex12p and Pex10p and Ubc4 plus Pex2p and Pex10p, respectively (72). We found that DTT-sensitive, Cys-monoubiquitination of Pex5p identified in wild-type CHO-K1 cells is impaired in the mutants pex2 Z65 and pex12 ZP109, where all three RING peroxins are unstable (Fig. 6C).
Although the Pex10p⅐Pex12p complex is most likely a candidate for the yet undefined E3 in ubiquitination of mammalian Pex5p, the precise role(s) in the ubiquitination of Pex5p remains to be defined. In P. pastoris, all of the three RING peroxins are suggested to be required as E3 enzymes for mono-and polyubiquitination of Pex20p as well as Pex5p (26). Given the finding that RING peroxins form a complex in vivo (this study) (55,71), three RING peroxins more likely act as a functional E3 unit, facilitating a variety of enzymatic activities, such as substrate specificity for Pex5p or Pex18p⅐Pex20p, the modes of ubiquitination (i.e. mono-or polyubiquitination), and the selection of E2 enzymes, Pex4p or Ubc4.
We found that Pex10p⅐Pex12p complex-mediated Lysmonoubiquitination of Pex5p is inhibited when bound to the initial Pex5p-docking site Pex14p and completely interfered with Pex14p and Pex5p cargo, PTS1 protein (Fig. 5A). In contrast, Pex5p bound to Pex13p is readily ubiquitinated as compared with free Pex5p, where the ubiquitinated Pex5p no longer binds to Pex13p (Fig. 5, A and B). Together with our earlier findings that Pex14p and Pex13p bind to cargo-loaded Pex5p and cargo-unloaded Pex5p, respectively (49), the Pex10p⅐ Pex12p complex E3 most likely ubiquitinates Pex5p in the complex with Pex13p after being dissociated from the cargo and translocated from Pex14p, consistent with the reports that RING peroxins are located downstream of the Pex5p-cargo docking complex (14, 17, 55, 73). . Pex5p ubiquitination is involved in matrix protein import into peroxisomes. A, complementing activity of Pex5pL KR mutants. CHO pex5 ZP139 cells were transfected with plasmids, each encoding wild type, K520R, K-less, and 520K mutants of Pex5pL-HA. After a 1-day culture, cells were immunostained with anti-catalase antibody. Bar, 10 m. B, quantification of complementing activity of KR mutants. Cells restored for the impaired import of PTS1 (not shown) and catalase (A) in 100 -150 cells were counted in three independent experiments. Complementing activities of K520R, K-less, and 520K mutants were represented as percentages relative to that of wild-type with S.E. (error bars). C, ubiquitination at Lys 520 abrogates the interaction of Pex5p with the PTS1 cargo protein. Cytosolic fractions of HEK cells expressing Pex5pL-HA mutants, 520K and K-less, were incubated with GSH-Sepharose beads conjugated with GST, GST-EGFP-PTS1, and GST-Pex14pN(1-260), respectively. Proteins in the bound fractions were analyzed as in Fig. 6F by immunoblotting, as indicated on the left, and staining with Coomassie Brilliant Blue (CBB). Input, 15% cytosolic fractions (Cyto) used for the pull-down assay (lanes 1 and 2). Solid and open arrowheads, 520K mutant ubiquitinated at Lys 520 and unmodified forms of the 520K and K-less mutants, respectively.
peroxisomal protein import. Interestingly, the lysine residue corresponding to Lys 520 of Chinese hamster Pex5pL is completely conserved between Pex5p orthologues in vertebrates and A. thaliana, whereas the arginine residue instead of the lysine is conserved in fungi, Caenorhabditis elegans, and Drosophila melanogaster (data not shown). This partly explains the difference in lysine-linked ubiquitination between yeast and mammals in the ubiquitin-mediated regulation of Pex5p shuttling. Taken together, our findings shed light on the molecular mechanisms underlying Pex5p-mediated matrix protein import into peroxisomes and its regulation involving cellular homeostasis of peroxisomes.