Functional Domains and Dynamic Assembly of the Peroxin Pex14p, the Entry Site of Matrix Proteins*

The 41-kDa membrane-anchored peroxin Pex14p functions as the peroxisome targeting signal (PTS) receptor-mediated, initial import site for matrix proteins. We here identify the functional domains of Pex14p involved in the assembly of import site subcomplexes. The minimal region of Pex14p required for restoring impaired protein import in pex14 Chinese hamster ovary cell mutant lies at residues 21-260 in the primary sequence. A highly conserved N-terminal region, encompassing residues 21-70, interacts with the PTS1 receptor Pex5p, Pex13p, and Pex19p that is essential for membrane biogenesis. N-terminal residues 21-140, including a hydrophobic segment at 110-138, function as a topo-genic sequence. Site-directed mutagenesis, size fractionation, and chemical cross-linking analyses demonstrate that the coiled-coil domain at residues 156-197 regulates homodimerization of Pex14p. Moreover, AXXXA and GXXXG motifs in the transmembrane segment mediate homomeric oligomerization of Pex14p, giving rise to assembly of high molecular mass complexes and thereby assuring Pex13p-dependent localization of Pex14p to peroxisomes. Pex5p, Pex13p, and Pex19p bind to Pex14p homo-oligomers with different molecular masses, whereas cargo-unloaded Pex5p apparently disassembles Pex14p homo-oligomers. Thus, Pex14p most likely forms several distinct peroxin complexes involved in peroxisomal matrix protein import.

described (19,20). DNA transfection to CHO cells was done by lipofection (11). Peroxisomes in cells were visualized by indirect immunofluorescence light microscopy with monospecific rabbit antibodies and fluorescein isothiocyanate-or Texas Red-labeled goat anti-rabbit immunoglobulin G (IgG) antibody (Cappel) (6). As second antibodies, fluorescein isothiocyanate-labeled sheep anti-mouse IgG (Amersham Biosciences) and goat anti-guinea pig IgG (Molecular Probes) antibodies were also used. Enhanced green fluorescent protein (EGFP)-fused proteins were monitored by EGFP fluorescence.

Construction of Pex14p Fusion Proteins and Their
Variants-Construction of expression plasmids for glutathione S-transferase (GST)fused Pex14p mutants and His 6 -tagged PEX14, and introduction of various amino acid substitutions in Pex14p were performed by conventional recombinant DNA methods and described under supplementary information.
Purification of Recombinant Proteins and in Vitro Binding Assay-E. coli XL-1 Blue cells were transformed with PEX14 cDNAs in pGEX6P-2 coding for GST and GST fusion proteins of wild-type and  (upper left). Numbers designate the positions in the amino acid sequence; black bar, hydrophobic TM domain; gray bar, predicted coiled-coil domain. These constructs were verified for the activity in restoring the impaired matrix protein import and their subcellular localization in CHO pex14 mutant ZP161 cells. cDNA transfection was done with a mock vector (panels a and b) and cDNAs encoding full-length His 6 -Pex14p (panels c and d) and truncation mutants (panels e-v), as indicated. PTS1 proteins and Pex14 variants were detected with antibodies to PTS1 and RGS-His 6 , respectively. Bar, 40 m. B, His 6 -Pex14p mutants defective in peroxisome-restoring activity were verified for peroxisomal membrane targeting in CHO-K1 cells. Endogenous PMP70 and MDH were also stained using specific antibodies. Bar, 40 m. C, N-terminal part comprising residues at 21-140 is sufficient for localization to peroxisomes. Upper panel, constructs of EGFP-fused Pex14p truncated mutants; lower, EGFP-Pex14p variants were expressed in CHO-K1 cells, as indicated. EGFP-Pex14p mutants and peroxisomes were visualized by EGFP fluorescence and with anti-PTS1 antibody. Bar, 40 m. mutant GST-Pex14p in pGEX6P-2 and conventionally cultured. GST-Pex14p variants were purified as described (11). Recombinant Pex5pL, Pex13p, and Pex19p were likewise isolated from respective GST fusion proteins (6). His 6 -tagged proteins expressed in E. coli were purified as done for GST fusion proteins, except for suspending buffer: 50 mM NaH 2 PO 4 , pH 8.0, 300 mM NaCl, 20 mM imidazole, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride (PMSF). Binding of GST-Pex14p variants to Pex5pL, Pex13p, and Pex19p was assayed as described (6). Binding assays using His 6 -Pex14p and its mutants were similarly done, except that the binding buffer 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 10 mM imidazole, 1 mM PMSF was used.
Sedimentation Analysis of Recombinant and Endogenous Pex14p-5 g each of His 6 -Pex14p or its mutants were incubated with or without other peroxins and EGFP-PTS1 for 2 h at 4°C in the in vitro binding buffer. The reaction mixture was applied onto the top of continuous 1-30% glycerol gradient in the same buffer and was centrifuged at 30,000 rpm for 18 h at 4°C in a Beckman SW41Ti rotor. Organelle fraction from CHO-K1 cells were lysed with Triton X-100 in the in vitro binding buffer and analyzed. 16 or 17 aliquots of 600 l each were collected from the bottom of the tube. Typically 5 l each from each fraction was analyzed by SDS-PAGE and immunoblot. Cross-linking Experiment-Protein cross-linking was performed using an amino group-reacting reagent, bis(sulfosuccinimidyl)-suberate (BS 3 , Pierce) (23). Purified recombinant proteins in PBS, 1% Triton X-100, 1 mM PMSF were incubated at various concentrations of BS 3 for 2 h at 4°C with gentle mixing. Excess BS 3 was quenched with 250 mM of Tris-HCl, pH 6.8.
Other Methods-Immunoblot analysis was performed as described (6), using enhanced chemiluminescence Western blotting detection reagent ECL (Amersham Biosciences).

RESULTS
Functional Domain Mapping-To define functional domains of Pex14p, we constructed various deletion mutants of hexaHis (His 6 )tagged Pex14p (Fig. 1A, upper left), expressed them in CHO pex14 ZP161 cells (Fig. 1A, panel b), and verified for complementation of the impaired matrix protein import by immunofluorescence microscopy. His 6 -Pex14p expression clearly restored PTS1 import in ZP161 cells, where numerous Pex14p-positive particles were superimposable with PTS1 particles, indicative of peroxisomes (Fig. 1A, panels a-d). An N-terminal deletion mutant His 6 -Pex14p21-376 showed the PTS1 import-restoring activity as the wild type, whereas further truncated mutants such as His 6 -Pex14p41-376 showed cytosolic PTS1 proteins (Fig. 1A, panels e-h and Table 1), suggesting that the N-terminal residues at 21-40 is required for the activity. Expression of mutants, His 6 -Pex14p1-260 and His 6 -Pex14p1-200 complemented the PTS1 import defect in ZP161, but less efficiently as compared with the wild type ( Fig.  1A, panels i and j and Table 1), whereas longer deletions such as His 6 -Pex14p1-140 and His 6 -Pex14p1-106 eliminated the activity (Fig. 1A, panels l and n and Table 1). His 6 -Pex14p21-200, was as potent as His 6 -Pex14p1-200 (Fig. 1A, panel p), whereas His 6 -Pex14p121-260 apparently formed inactive aggregates (Table 1). It is noteworthy that catalase remained in the cytosol in ZP161 cells expressing His 6 -Pex14p1-200 or His 6 -Pex14p21-200, where PTS1 and PTS2 import was rescued ( Table  1). All of the Pex14p constructs competent in restoring the impaired matrix protein import were detected in PTS1-positive, re-established peroxisomes, indicating peroxisomal localization ( Fig. 1A and Table 1). Collectively, the minimal sequence of Pex14p competent in complementing the impaired protein import in pex14 cells and peroxisomal membrane targeting comprised the residues at 21-260, including the transmembrane (TM) and coiled-coil domains.
To assess this issue in regard to the TM and coiled-coil regions, we expressed in ZP161 cells four constructs: His 6 -Pex14p⌬141-240, deleted in the coiled-coil region; His 6 -Pex14p⌬110 -138, lacking the sequence containing TM segment; His 6 -Pex14p⌬110 -123, deleted in the N-terminal part of TM; and His 6 -Pex14p⌬131-138. Only His 6 -Pex14p⌬131-138 complemented the mutant phenotype of ZP161, whereas the other three showed no restoring activity (Fig. 1A, panels q-v and Table 1). Thus, it is apparent that a hydrophobic segment and its downstream coiled-coil domain are essential for the biological activity of Pex14p.
Pex14p Requires the N-terminal Portion for Peroxisomal Localization-To search for the peroxisomal targeting sequence of Pex14p, we expressed various deletion mutants of His 6 -Pex14p (Table 1) in wild- type CHO-K1 cells and verified for their subcellular localization, as in Fig. 1A. His 6 -Pex14p1-140 was localized to numerous particles, which were coincided with PMP70-stained peroxisomes (Fig. 1B, panels a and e), thereby indicating localization to peroxisomes, despite being incompetent in complementing pex14 ZP161 cells (Fig. 1A, panels k and l). Moreover, His 6 -Pex14p1-140 was integrated into membranes as His 6 -Pex14p (supplementary data, Fig. S1). In contrast, His 6 -Pex14p1-106 showed a diffused staining pattern (Fig. 1A, panel m). Taken together, the N-terminal region at 1-140 was apparently sufficient for targeting to peroxisomes. Distinct from His 6 -Pex14p21-376, His 6 -Pex14p41-376 and His 6 -Pex14p65-376 were observed in larger structures (Fig. 1B, panel d and Table 1), which were coincided with those with MDH, a mitochondrial matrix protein (Fig. 1B, panel h), suggesting that both mutants were mislocalized to mitochondria. Thus, the N-terminal part encompassing residues 21-140 of Pex14p was essential for peroxisomal localization. A middle region mutant His 6 -Pex14p121-260 apparently aggregated, and His 6 -Pex14p241-376 showed cytosolic staining (Table 1).
His 6 -Pex14p⌬110 -123 and His 6 -Pex14p⌬110 -138 were not colocalized with PMP70, indicating a defect of peroxisomal targeting, whereas His 6 -Pex14p⌬131-138 maintained the targeting activity (Fig.  1B, panels b and f and Table 1). Pex14p⌬141-240 was detected as numerous small punctate-stained particles, not colocalized with PMP70 ( Fig. 1B, panels c and g). These results strongly suggested that both the cluster of hydrophobic residues at 110 -123 and the coiled-coil region are essential for Pex14p anchoring to peroxisomal membranes.
Next, we generated several EGFP-fused N-terminal parts of Pex14p and expressed in CHO-K1 cells. EGFP-Pex14p1-140 was clearly localized to peroxisomes, as assessed by colocalization with PTS1 proteins (Fig. 1C, panels b and f), whereas EGFP was located in nucleus and cytoplasm (Fig. 1C, panel a). EGFP-Pex14p21-140 was mostly targeted to peroxisomes, but EGFP-Pex14p41-140 was detected in the cytosol and perinuclear region (Fig. 1C, panels c, d, g, and h). These results indicated that the N-terminal residues 21-140 of Pex14p play an indispensable role in translocation to peroxisomes.
Pex5p, Pex13p, and Pex19p Bind to the N-terminal Region of Pex14p-To search for Pex14p-interacting partners, we performed a pull-down assay using His 6 -Pex14p and pex14 ZP161 cell lysates. Pex5p, Pex13p, and Pex19p were specifically detected, where mitochondrial MDH was not bound ( Fig. 2A). Other peroxins such as Pex10p, Pex12p, Pex1p, and PMP70 were under the detectable level (not shown). The findings were consistent with other reports describing that Pex14p interacts with three peroxins, Pex5p, Pex13p, and Pex19p (5, 13, 14, 24). We therefore determined the precise regions involved in such interaction with three peroxins, Pex5pL, Pex13p, and Pex19p, using various truncated recombinant Pex14p variants fused to GST (Fig. 2B, left panel). Upon incubation, Pex5pL was specifically bound to the full-length and various truncated mutants of GST-Pex14p such as GST-Pex14p31-70 ( Fig. 2B, center panel). Similarly, Pex13p was bound to GST-Pex14p and several Pex14p mutants including Pex14p21-70. Together, the sequences at 31-70 and 21-70 in the N-terminal region of Pex14p are required for binding to Pex5p and Pex13p, respectively. Moreover, normal and mutant His 6 -Pex14p constructs containing the N-terminal residues at 21-70 specifically bound to Pex19p (Fig. 2B, center).
Pex19p in the Pex14p Matrix Import Subcomplexes-Pex14p forms subcomplexes with cargo-loaded Pex5p and Pex13p (6). Pex19p also interacts with Pex14p (Fig. 2, A and B) (14,15,17). However, whether Pex19p plays a role in peroxisomal matrix protein import remains elusive. We first examined by in vitro binding assays whether Pex19p alters Pex14p binding to Pex5p and Pex13p. His 6 -Pex14p bound to Pex19p (Fig. 2C, left panel, lane 5). His 6 -Pex14p equally bound to Pex5pL in the presence or absence of Pex19p (lanes 6 and 7), where Pex19p binding to His 6 -Pex14p was not detectable. Pex14p-Pex5pL interaction likely occurred with an affinity much higher than that between Pex14p and Pex19p, as noted between Pex13p and Pex14p (6). In contrast, a slightly reduced level of Pex13p bound to His 6 -Pex14p was detected in the presence of Pex19p (Fig. 2C, left, lanes 8 and 9) and apparently remained at the same level even with 100-fold excess of Pex19p to Pex13p (Fig. 2C, right). We next verified whether Pex19p binding to Pex14p modulates the complexes comprising Pex14p, Pex13p, and cargo-loaded Pex5p. GST-Pex14p was incubated with a fixed amount of Pex5p, Pex13p, PTS1-cargo EGFP-His 6 -SKL, and increasing amounts of Pex19p (Fig.   2D). Pex5p was moderately elevated with increasing amounts of Pex19p, whereas EGFP-His 6 -SKL and Pex13p were slightly decreased. Thus, it is plausible that Pex19p may regulate the assembly of import subcomplexes comprising at least Pex14p, Pex13p, and PTS1-Pex5p.
Highly Conserved N-terminal Region of Pex14p-We found that the N-terminal region is indispensable for peroxisomal localization of Pex14p ( Fig. 1) and that its highly conserved domain comprising residues at 21-70 is involved in interaction with Pex5p, Pex13p, and Pex19p (Fig. 2B). Sequence comparison of Pex14p from various species (11,12,25) showed that the N-terminal region was highly conserved (Fig. 3A). We constructed two His 6 -Pex14p mutants with Ala substitution of the most conserved residues, Phe 35 -Leu 36 and Phe 52 -Leu 53 , termed FL35/ 36AA and FL52/53AA, respectively (Fig. 3B). Both were completely eliminated in binding to Pex13p, but not to Pex19p (Fig. 3B, lower left, lanes [3][4][5]. Pex5p interaction was little or fully maintained in FL35/ 36AA and FL52/53AA, respectively. Both mutants were inactive in complementing the defective protein import when expressed in ZP161 and failed in peroxisomal targeting, rather mislocalized to mitochondria (Fig. 3B, lower right). These results strongly suggested the relevance between interaction with Pex13p and peroxisomal targeting plus function of Pex14p.
Pex14p Is Mislocalized to Mitochondria in pex13 Mutant Cells-As revealed above, Pex14p mutants defective in Pex13p binding were not localized to peroxisome membranes, rather readily mislocalized to mitochondria. To assess this notion, we examined endogenous Pex14p in PEX13-deficient fibroblasts from a patient with peroxisome biogenesis disorders possessing a homozygous nonsense mutation W234ter in Pex13p resulting in deletion of the 2nd TM and SH3 domains (26). In normal control fibroblasts, endogenous Pex14p was detected in a FIGURE 3. The N-terminal sequence of Pex14p involved in interaction with multiple peroxins is highly conserved. A, alignment of amino acid residues at 21-70 of Pex14p of mammals and other species, rat (Rn), humans (Hs), Pichia pastoris (Pp), S. cerevisiae (Sc), Neurospora crassa (Nc), Schizosaccharomyces pombe (Sp), Caenorhaditis elegans (Ce), Drosophila melanogaster (Dm), and Arabidopsis thaliana (At), using a ClustalW program. Gaps were introduced to maximize the similarity. Conserved residues, including identical or highly similar amino acid residues, were shaded. Asterisk, identical amino acid residue between all species. Colon, residue with high similarity among the species. Dot, those conserved in more than five of nine species. B, upper panel, dipeptide FL at 35-36 and 52-53 of rat His 6 -Pex14p were substituted with Ala, termed FL35/36AA and FL52/53AA, respectively. Lower left, binding assays were performed using His 6 -Pex14p, FL35/36AA, and FL52/ 53AA (1 g each) with Pex5pL, Pex13p, and Pex19p (0.2 g each), as in Fig. 2. Lanes: 1, 20% input; 2, ligand-free affinity beads; 3, His 6 -Pex14p; 4, FL35/36AA; 5, FL52/53AA. His 6 -Pex14p and FL mutants were stained with Coomassie Blue (CBB). Lower right, FL-mutants were verified for peroxisome-restoring activity and localization in ZP161 cells. Cells were immunostained using antibodies to Pex14p and PTS1. Mitochondria were stained with anti-MDH antibody. Bar, 40 m. Ps, peroxisomes; Mt, mitochondria; ϩϩ, PTS proteins were normally imported; Ϫ, not imported. Note that FL35/36AA and FL52/53AA were inactive and mostly localized to mitochondria, not to peroxisomes.
merged view with PMP70 in numerous peroxisomes (Fig. 4A, panels  a-c). In contrast, in patient fibroblasts, Pex14p was hardly merged with PMP70-positive membrane remnants (Fig. 4A, panels d-f), rather detected in a large number of tubular structures, mitochondria as visualized with MitoTracker (Fig. 4B). Thus, we concluded that Pex14p in pex13 fibroblasts is mislocalized to mitochondria, inferring that Pex14p-Pex13p interaction is required for Pex14p targeting to peroxisomes.
Oligomerization of Pex14p-Pex14p was reported to interact with itself in vitro (11), possibly via its middle region containing the coiledcoil motif (18). In addition, the hydrophobic domain of Pex14p includes 113 AXXXA 117 and 118 GXXXG 122 motifs, which are generally involved in stabilizing the intermolecular helix-helix interactions as well as the intramolecular interactions (27). To investigate a role of the coiled-coil motif in Pex14p, we constructed a His 6 -Pex14p mutant in which 10 amino acid residues in the coiled-coil motif, Val 159 , Val 163 , Val 166 , Leu 170 , Val 173 , Leu 177 , Val 184 , Leu 187 , Leu 191 , and Ala 194 at positions a and d of typical heptad repeats, were substituted with Gln (named Coil, Fig. 5A). Also, to determine whether the AXXXA and GXXXG motifs play a role in the mutual interaction of Pex14p, we constructed three mutants with A 3 L and G 3 L substitutions, termed His 6 -Pex14pAL, GL, and AL/GL (Fig. 5A). Chemical cross-linking of the purified His 6 -Pex14p mutants was performed with BS 3 . At 0.1-0.2 mM BS 3 , a band of about 130 kDa, possibly a dimer of wild-type His 6 -Pex14p with a mobility of apparent mass of ϳ56 kDa (11), was detected (Fig. 5B, upper left,  lanes 2 and 3). The amount of Pex14p dimer was maximal at 0.4 mM BS 3 (lane 4), and at higher concentration of BS 3 , the amount of Pex14p dimer was gradually reduced (lanes 5-9), with concomitant increase in higher molecular mass bands with 190 kDa and 240 kDa. Homomeric oligomers of Pex14p were also identified by cross-linking the membrane fractions each from CHO-K1, pex2 Z65, and pex12 ZP109, but not pex14 ZP110 (supplementary data, Fig. S2), strongly supporting the findings with recombinant Pex14p. The dimer and higher mass complexes of His 6 -Pex14p mutants, AL, GL, and AL/GL, were similarly but less efficiently formed, whereas only a dimeric form was detected for His 6 -Pex14p⌬110 -138 lacking the hydrophobic TM domain (respective panels, lanes 1-9). These results suggested that mutation of GXXXG and AXXXA motifs, more apparently TM deletion, affected the formation of high molecular mass complexes. More strikingly, mutant Pex14pCoil no longer gave rise to the oligomer formation even at 2 mM BS 3 (Fig. 5B, lower center), strongly suggesting that the coiled-coil domain of Pex14p is indispensable for homo-oligomerization. Crosslinking assay using a Pex14p mutant at 21-70, FL35/36AA, showed several forms of oligomers, similar to the wild type (Fig. 5B, lower right), thereby indicating that the region at 21-70 is not responsible for the oligomerization.
Next, the oligomerization of His 6 -Pex14p variants was also assessed by a sedimentation assay. Soluble wild-type 42-kDa His 6 -Pex14p and its mutants were each incubated, loaded onto a glycerol gradient, and ultracentrifuged (Fig. 5C). Wild-type His 6 -Pex14p was sedimented in fractions 4 -17, with a main peak at fractions 6 -7 corresponding to 300 -450 kDa, strongly suggesting that Pex14p formed homo-oligomeric complexes. His 6 -Pex14p variants, AL and GL, were detected in several fractions ranging 200 -600 kDa and 200 -400 kDa, respectively, indicating homo-oligomers with masses lower than those of wild-type Pex14p. Mutant AL/GL was detected mainly in fractions equivalent to 200 -300 kDa, consistent with the findings in the cross-linking assays (Fig. 5B). Thus, the AXXXA or GXXXG motifs of Pex14p are likely involved in forming high molecular mass complexes. Mutant Coil was detected mainly in two fractions, 1 and 2, corresponding to Pex14p monomer, suggesting that the coiled-coil motif plays an important role in oligomerization.
C-terminally truncated His 6 -Pex14p1-260 containing the TM and coiled-coil domains restored matrix protein import, being localized to peroxisomes in pex14 mutant cells (Fig. 1A and Table 1). In contrast, His 6 -Pex14p1-140 with only the TM domain failed to restore the protein import in pex14 mutants, despite being translocated to peroxisomes, hence implying the importance of coiled-coil domain (Table 1). Upon incubation with BS 3 , oligomeric forms of His 6 -Pex14p1-260 were indeed detected as with the wild-type, whereas Pex14p1-140 was monomeric (Fig. 5D). Furthermore, incubation with BS 3 of His 6 -oligopeptides comprising the coiled-coil sequence at 153-200 gave rise to two bands, monomer and dimer of the oligopeptides, respectively (Fig.  5E). Taken together, we conclude that the coiled-coil motif of Pex14p specifies homo-dimerization and the GXXXG and AXXXA motifs in the TM domain function in high molecular mass complex formation.
Next, we expressed Pex14p variants mutated in the coiled-coil and the GXXXG and AXXXA motifs in pex14 ZP161. PTS1 import was not restored in ZP161-expressing His 6 -Pex14pCoil, in contrast to numerous peroxisomes re-established in His 6 -Pex14p-expressing cells (Fig.  6A, panels a, b, f, and g). Expression of mutants, His 6 -Pex14pAL and His 6 -Pex14pGL, complemented the PTS1 import defect, but less efficiently than the wild type (panels c, d, h, and i), whereas His 6 -Pex14pAL/GL showed no restoring activity (panels e and j). Both His 6 -Pex14pAL and His 6 -Pex14pGL were detected in peroxisomes and partly in another type of membrane structures, apparently endoplasmic reticulum (ER) (Fig. 6A, panels c and d). His 6 -Pex14pAL/GL appeared more likely to be mislocalized to ER (Fig. 6A, panel e), as indeed confirmed by colocalization with calreticulin, an ER marker (Fig. 6B, panels e and j). pex14 ZP161 cells contain Pex13p-positive membrane remnants (11) (Fig. 6B, panels a and f). In ZP161 cells transfected with wild-type His 6 -PEX14, His 6 -Pex14p was colocalized with Pex13p to re-established peroxisomes (panels b and g). Upon expression of His 6 -Pex14pCoil, numerous Pex14p-positive particles were detected, being smaller in size but larger in number than normal peroxisomes (panels c and d). However, nearly all of such tiny particles were merged neither to Pex13p-positive, PTS1-negative (not shown) particles nor to those visualized by staining of MDH, a mitochondria marker (Fig. 6B, panels c, d, h, and i), indicative of the defect of peroxisomal membrane targeting. Essentially the same results were obtained with CHO-K1 cells (not shown). Collectively, the AXXXA and GXXXG motifs and the coiled-coil domain are essential for the peroxisomal localization and function of Pex14p, for which its oligomerization is critical.
Furthermore, these four Pex14p mutants were verified in vitro for binding to Pex5pL, Pex13p, and Pex19p. His 6 -Pex14pCoil was significantly reduced in binding to Pex13p (Fig. 6C, lane 4), whereas any mutations in the TM domain did not alter the binding to Pex5p, Pex13p, and Pex19p ( lanes 5-7). Accordingly, it is likely that oligomerization, possibly dimerization, of Pex14p is required for the interaction with Pex13p. This interpretation apparently contradicts the notion that the minimal region of Pex14p for the interaction with Pex13p resides at N-terminal residues 21-70 (see Fig. 2B). GST forms a dimer per se (28). Therefore, it is more likely that GST-fused residues 21-70 binds to Pex13p as a dimer. Alternatively, the mutation in the coiled-coil domain may affect the Pex13p binding by inducing the structural change at the N-terminal region including the residues 21-70, Pex14p Forms Distinct Subcomplexes with Its Interacting Partners-Pex14p interacted with three peroxins Pex5pL, Pex13p, and Pex19p by using almost the same regions at its N-terminal part (see Fig. 2, B and D). A similar finding was very recently reported (15). However, molecular mechanisms underlying such multi-interactions at one region of Pex14p remained unknown. To address this problem, we performed a sedimentation assay using several types of Pex14p-containing complexes. Pex14p complexes each consisting of 42-kDa His 6 -Pex14p loaded with 68-kDa Pex5pL, 44-kDa Pex13p, or 33-kDa Pex19p were isolated and analyzed by ultracentrifugation on a glycerol gradient. Mock-loaded His 6 -Pex14p was found in nearly all fractions with a major peak in fractions 5-7 corresponding to 150 -300 kDa (Fig. 7A, panel a), consistent with the pattern noted above (Fig. 5C), indicating that His 6 -Pex14p existed as homo-oligomer with various molecular masses, mainly in tetramer to octamer. In contrast, Pex5p-loaded His 6 -Pex14p was sedimented mainly into fractions 3-5 with ϳ100 kDa (panel b), indicative of a monomeric Pex14p-Pex5p complex, whereas Pex5p itself was in fractions 2 and 3 (panel a). This observation was consistent with the reports using cell-free translated Pex5p and rat liver peroxisomes (18,29). Next, Pex13p-bound His 6 -Pex14p was detected in virtually all fractions, more in fractions 6 -10 equivalent to 200 -700 kDa, whereas Pex13p was sedimented mainly in monomeric form and partly as a dimer if any (Fig. 7A, panels b and a). The peak of Pex13p-bound His 6 -Pex14p was in a higher glycerol density than that of Pex5p-bound His 6 -Pex14p, suggesting that Pex5p and Pex13p interact with Pex14p homooligomer of different sizes. Pex19p-loaded His 6 -Pex14p was partly sedimented into fractions 3 and 4, similar to Pex14p-Pex5p complexes, whereas Pex19p was in fraction 2, possibly as a monomer (panels b and a). Taken together, Pex5p and Pex19p bound to Pex14p in a monomeric or smaller size oligomer, whereas Pex13p formed larger complexes with higher mass Pex14p oligomer. Thus, it is likely that Pex14p changes the interacting partners in a manner dependent on the molecular size of Pex14p complexes. Conversely, Pex5p may disassemble Pex14p homooligomers by Pex5p-Pex14p interaction, while Pex13p and Pex19p are less likely potent in the disassembly of Pex14p oligomers because higher mass Pex14p oligomers are still detectable after loading of Pex13p and Pex19p. We further examined His 6 -Pex14p oligomerizing activity with a PTS1-cargo, EGFP-SKL-loaded Pex5p in the presence or absence of Pex13p. Incubation of His 6 -Pex14p with the PTS1-cargo-loaded Pex5p and Pex13p resulted in assembly of high molecular mass subcomplexes (Fig. 7A, upper panel c), apparently containing quaternary complexes comprising His 6 -Pex14p, Pex13p, Pex5p, and EGFP-SKL (fractions 5 and 6). Similar size oligomeric complexes, ternary complexes of His 6 -Pex14p, Pex5p, and EGFP-SKL, were detected upon incubation of His 6 -Pex14p with the cargo-loaded Pex5p (lower panel c). It was possible that size resolution in the gradient between Pex14p/Pex13p/Pex5p/EGFP-SKL and Pex14p/Pex5p/EGFP-SKL was not high enough or that the oligomeric states of Pex14p for two types of complexes were distinct.  Therefore, Pex14p oligomers are more likely disassembled only by the cargo-unloaded Pex5p.
Furthermore, we investigated whether three peroxins Pex5p, Pex13p, and Pex19p interact with distinct size Pex14p subcomplexes in vivo. Detergent-solubilized organelle fraction from wild-type CHO-K1 cells was likewise analyzed by glycerol-gradient ultracentrifugation (Fig. 7B). Endogenous Pex14p was observed as wide range mass subcomplexes such as larger ones with Pex13p and down to one with a lower range with Pex5p, whereas organelle-associated Pex19p appeared to be with monomeric Pex14p. The sedimentation pattern was similar to those seen with the incubation mixtures using recombinant proteins (Fig. 7A,  panels b and c), strongly suggesting that these three peroxins are included in Pex14p subcomplexes of different molecular masses in vivo.

DISCUSSION
Pex14p in Peroxisomal Matrix Protein Import Complexes-In the present work, we searched for functional regions of Pex14p that are: 1) required for restoring the impaired matrix protein import in pex14 cell mutant and 2) responsible for targeting to peroxisome membranes. We identified a variant, Pex14p21-260 containing the TM and coiled-coil domains as a minimal sequence for such activity. It is noteworthy that Pex14p21-200 was as competent as Pex14p21-260 in restoring the import of PTS1 and PTS2 proteins, but not of catalase, although both Pex14p mutants contained the sequence 21-70 responsible for the interaction with Pex5p, Pex13p, and Pex19p (see Table 1). Recognition of Pex5p-catalase complexes by Pex14p and a cargo-unloading step may be differentiated from the typical import of SKL-type PTS1 and PTS2 proteins. In any event, all the examined Pex14p variants evidently indicated that being active in targeting to peroxisomal membranes is prerequisite for being competent in complementing the impaired protein import in pex14 cells.
The region at 21-70 of Pex14p contains the highly conserved sequence, FLX 15 FLX 2 KGLTX 2 EI. It is noteworthy that this conserved region is required for interaction with Pex5p in protein import to peroxisomes of Saccharomyces cerevisiae (30) and glycosomes of protozoan Leishmania donovani (31) and Trypanosoma brucei (32). Through Ala substitution of conserved residues such as FL, we showed that the conserved residues are important for not only the interaction with Pex13p but also the Pex14p targeting to peroxisomes in vivo. Therefore, the recognition of and interactions with several peroxins using the identical region must be highly controlled in vivo. We indeed found that Pex14p interacted with its partners in a manner of the homo-oligomer sizes (Fig. 7), where the Pex14p monomer or dimer interacted with Pex5p and Pex19p, and the binding to Pex13p gave rise to high molecular mass hetero-oligomeric complexes. Conversely, Pex5p, but not the cargo-loaded Pex5p, mediated disassembly of Pex14p oligomer, resulting in it no longer binding to Pex13p. This was in good agreement with our earlier findings that Pex5p affects Pex14p-Pex13p complexes in a cargo-dependent manner (6).
Pex19p is mainly in the cytosol and only partly on the peroxisomal membranes (14,20). Pex19p showed a moderate effect on the complexes of Pex14p, Pex13p, and cargo-loaded Pex5p in vitro (Fig. 2). Pex19p may be functionally involved in peroxisomal matrix protein import machinery, likely playing a potential role for Pex19p in assembly of PTS receptor docking complexes as recently suggested (15). It is also possible that the interaction of Pex14p with Pex19p plays a role different from that with other peroxins, including Pex5p and Pex13p. Pex19p has recently been addressed as a cytosolic chaperone and PMP transporter (16,(33)(34)(35). 3 Targeting Pathway of Pex14p to Peroxisomes-In several peroxisomal membrane proteins, the cluster of basic residues and the adjacent or downstream transmembrane domain(s) act as the peroxisomal membrane targeting signals (mPTS) (36 -40). These mPTS-containing sequences are generally recognized by Pex19p. In this study, we found that the Pex14p N-terminal part with hydrophobic TM region binds to Pex19p and is transported to peroxisomes. However, the mutations of the highly conserved FL dipeptide sequences in this region caused mistargeting of Pex14p to mitochondria, despite their binding to Pex19p. Pex14p is likewise mislocalized to mitochondria in PEX13-deficient patient fibroblasts where peroxisomal remnants harboring PMPs such as PMP70 were normally detectable. In contrast, endogenous Pex13p is localized to peroxisomal ghosts and colocalized with PMP70 in CHO pex14 mutants, ZP110 and ZP161 (data not shown), indicating that Pex14p is not required for targeting of Pex13p and other PMPs to peroxisomal membranes. These observations raise several possibilities. First, Pex14p-Pex13p interaction is equally or more important than the Pex14p-Pex19p interaction for targeting to peroxisomes. Secondly, Pex14p may be targeted to peroxisomes by a mechanism distinct from that underlying the import of other PMPs (16,34,35,40).
Homo-oligomerization of Pex14p-Many proteins function by stably or transiently forming specific oligomers (41). Although homo-oligomers of Pex14p were identified (11,17,18), their functional consequence remained undefined. Based on our findings, the coiled-coil domain of Pex14p is most likely involved in the homo-dimerization, and the AXXXA and GXXXG motifs in the TM region are responsible for high molecular mass homo-oligomerization (Fig. 5). This is consistent with the concept that the GXXXG and AXXXA motifs, one type of the common ␣-helical interaction motifs, in the TM domain of membrane proteins mediate homo-oligomerization (27). A coiled-coil mutant of Pex14p with intact AXXXA and GXXXG motifs did not assemble into 3 T. Matsuzaki and Y. Fujiki, unpublished results. any oligomer. Furthermore, His 6 -Pex14p1-140 containing the AXXXA and GXXXG motifs could not form any homo-oligomer. Thereby, we conclude that the coiled coil-linked homo-dimerization of Pex14p and the AXXXA and GXXXG motif-mediated homo-oligomerization are independent events. It is most likely that both AXXXA and GXXXG motifs form and stabilize high molecular mass complexes after dimerization of Pex14p.
The coiled-coil mutation of Pex14p abrogated the interaction with Pex13p, resulting in mislocalization of Pex14p to numerous small membrane structures absent from PMPs such as PMP70 and Pex13p. Such membrane structures remain to be defined. In CHO pex19 mutant ZP119 and human PEX19-defective fibroblasts, peroxisomal membrane ghosts are absent despite the synthesis of PMPs, apparently because of rapid degradation of PMPs (14,42). However, only Pex14p in PMPs examined is not degraded, rather mislocalized mostly to mitochondria in pex19 and pex3 cells (14,43). Thus, Pex14p appears to be different from other PMPs in membrane targeting. We await further investigation to delineate this issue.
A Working Model of Pex14p-mediated Matrix Protein Import-Given these novel findings described above, we propose a working model for the functions of Pex14p (Fig. 8). The peroxisomal matrix protein import pathway can be divided into six steps: Step 1, cargo-loaded Pex5p first interacts with Pex14p and forms a hetero-oligomer consisting of Pex13p-Pex14p-Pex5p-PTS1 complexes.
Step 2, PTS1 proteins are unloaded into the peroxisomal matrix.
Step 4, Pex5p shuttles back to the cytosol, possibly through potential translocation complexes consisting of RING peroxins, Pex2p, Pex10p, and Pex12p (6, 44 -47). At steps 5 and 6, Pex14p forms distinct subcomplexes (7,48). Pex14p can be a component of a subcomplex at the initial step of matrix protein import, while Pex13p plays a major role in a subcomplex for translocation of Pex5p destined for shuttling back to the cytosol (6). These two types of subcomplexes may associate and dissociate in a manner dependent on Pex14p homo-oligomers, which subsequently form the initial hetero-oligomeric import machinery with Pex13p. It is plausible that several unidentified factors (X) may be involved at the PTS cargo-releasing step (step 2).