Structure of the N-terminal Domain of PEX1 AAA-ATPase

Peroxisomes are responsible for several pathways in primary metabolism, including β-oxidation and lipid biosynthesis. PEX1 and PEX6 are hexameric AAA-type ATPases, both of which are indispensable in targeting over 50 peroxisomal resident proteins from the cytosol to the peroxisomes. Although the tandem AAA-ATPase domains in the central region of PEX1 and PEX6 are highly similar, the N-terminal sequences are unique. To better understand the distinct molecular function of these two proteins, we analyzed the unique N-terminal domain (NTD) of PEX1. Extensive computational analysis revealed weak similarity (<10% identity) of PEX1 NTD to the N-terminal domains of other membrane-related type II AAA-ATPases, such as VCP (p97) and NSF. We have determined the crystal structure of mouse PEX1 NTD at 2.05-Å resolution, which clearly demonstrated that the domain belongs to the double-ψ-barrel fold family found in the other AAA-ATPases. The N-domains of both VCP and NSF are structural neighbors of PEX1 NTD with a 2.7- and 2.1-Å root mean square deviation of backbone atoms, respectively. Our findings suggest that the supradomain architecture, which is composed of a single N-terminal domain followed by tandem AAA domains, is a common feature of organellar membrane-associating AAA-ATPases. We propose that PEX1 functions as a protein unfoldase in peroxisomal biogenesis, using its N-terminal putative adaptor-binding domain.

peroxisome involves a protein transport system comprising more than 30 proteins, known as PEX gene products or peroxins (3)(4)(5). The peroxins are required for recognition, targeting, and import of peroxisomal proteins containing peroxisome targeting signals (6 -8). The PEX genes are widely conserved among most eukaryotes, from yeast to animals. In addition, genetic mutations in any of the PEX genes are a common cause of peroxisome biogenesis disorders, such as Zellweger syndrome, neonatal adrenoleukodystrophy, and infantile Refsum disease (1,9,10).
Peroxisomal protein transport is an ATP-dependent phenomenon. From over 30 peroxins, only two AAA-ATPases, 1 PEX1 and PEX6, were identified. Dysfunction of the ATPase activity of either PEX1 or PEX6 causes peroxisome biogenesis disorders (11)(12)(13)(14)(15). PEX1 or PEX6 is mutated in ϳ80% of all patients with Zellweger syndrome (15). The mutations in the PEX1 and PEX6 genes are found in different complementary groups of peroxisome biogenesis disorder patients, CG1 and CG4, respectively. PEX1 and PEX6 have been shown to associate with one another both in vitro and in vivo (16 -19). Therefore, both ATPase activities are thought to be required for normal peroxisomal matrix protein import (20).
AAA-ATPases, associated with various cellular activities, are hexameric ATP-hydrolyzing enzymes found in all three kingdoms of life (21,22). These enzymes play an important role in protein unfoldase activity, including the dissociation of protein complexes. AAA-ATPases are characterized by the presence of one or two copies of a well conserved catalytic ATPase domain comprising 220 -250 amino acids. Most AAA family proteins are classified into two representative classes, type I and type II, which contain one and two copies of the AAAcassette, respectively. In addition, an alternative classification of AAA-ATPases is possible in terms of the presence or absence of membrane integrated domains in their primary structures. Most of AAA-ATPases are cytosolic, whereas certain subfamilies contain N-terminal transmembrane helices (i.e. SF 6 based on Beyer's classification) (23). The members belonging to the cytosolic type II AAA-ATPases, NSF and VCP (p97), have been extensively studied in terms of both function and structure. These enzymes form a hexameric ring and can act as a molecular chaperone. They reside on organelle membranes, where they play an important role in maintaining organelle function (24,25). NSF and its yeast ortholog, Sec18, is responsible for heterotypic membrane fusion mainly in exocytic pathways (26 -28). VCP and yeast CDC48 are involved in endoplasmic reticulum-associated protein degradation (29) as well as remodeling of the Golgi and nuclear membrane (24,30). Although the specific target of VCP unfoldase activity is not yet known, there are three specific adaptor molecules, p47 (30), Ufd1/Npl4 (31), VCIP135 (32), and recently found VIMP (33), which differentiate the various cellular functions of VCP. In contrast, whereas the SNARE complex is known to be the specific target molecules for NSF, ␣-, ␤-, and ␥-SNAPs were identified as the most important adaptor molecule for this function (34,35).
The two AAA-domains (D1 and D2) of PEX1 and PEX6 display a sequence identity of 19 -28 and 29 -52%, respectively (Fig. 1a). In contrast, the N-terminal sequences of PEX1 and PEX6 are quite different. We hypothesized that the N-terminal region may act to discriminate the characteristic functions of the two AAA-ATPases. We based our hypothesis on the following two observations. First, distinct functions of PEX1 and PEX6 are suggested by the different genetic complementation groups of cells from patients with peroxisome biogenesis disorders (13,15,36). Second, PEX15 and PEX26, both peroxisomal membrane-integrated proteins, have been identified as specific adaptor molecules for PEX6 (37,38), but no specific adapter or putative substrate protein has been identified for PEX1 (39 -41).
In order to investigate the function of PEX1, we focused on the unique amino acid sequence at the N terminus. The sequence has been subjected to extensive bioinformatics analyses coupled with biochemical experiments. The N terminus of PEX1 (residues 1-180) was found to be possess some similarity to the N terminus of VCP and NSF (Ͻ10% amino acid identity). Subsequently, the PEX1 N-terminal domain (NTD) was shown to independently fold as a globular domain. Finally, the crystal structure of PEX1 NTD, at 2.05-Å resolution, revealed a striking similarity to that of the "N-domains" of VCP and NSF. This kind of domain architecture is known as a "supradomain," as found in many examples where two or three domains (in our case, NTD, D1, and D2) are evolutionally conserved in terms of their sequential order and biological context (42). Based on our "reversed" structural proteomics approach, a physiological function of PEX1 in peroxisome biogenesis is proposed.

EXPERIMENTAL PROCEDURES
PSI-BLAST Analysis, FORTE Analysis, FUGUE Analysis, and Multiple Sequence Alignments of Proteins-The domain analysis of PEX1 and PEX6 orthologs was initially performed using SMART (43) and PFAM (44). These servers failed to detect any domain at the N-terminal region of these proteins. To elucidate the relationship between the N-terminal region of PEX1 and homologous proteins at the primary amino acid sequence level, we performed extensive PSI-BLAST searches (45) using our in-house Linux-based PC server against the nonredundant protein sequence data base. Specifically, we performed a PSI-BLAST search on the N-terminal segment of human PEX1 (residues 1-250), using a BLOSUM62 matrix and an expect value threshold of 0.003. Sequences containing a CDC48-2 domain (i.e. mouse VCP and hamster NSF, by SMART definition) were retrieved. We also performed a reverse PSI-blast search using these sequences. Based on the fact that the CDC48 -2 domain associated with the CDC48-N domain, we subjected the N-terminal 192 residues of hamster VCP to a PSI-Blast, which successfully recovered the PEX1 sequence. A multiple alignment of sequences containing the entire N-terminal domain of PEX1 was constructed by collecting the "highest scoring pairs of sequence segments" using PSI-BLAST. These were manually realigned using Clust-alW and ClustalX (46). Based on this alignment, the C-terminal boundary of PEX1 NTD was judged to be around residue 180, in terms of the ratio of conserved amino acids within the PEX1 orthologs. This putative domain boundary was also predicted by the DomCut server (47). In order to confirm the distant relationships of PEX1 NTD to other known domain structures, we performed FORTE searches (48). FORTE is a method that implements profile-profile comparisons to predict the structure of a query sequence. The existence of a domain at the N terminus of PEX1 with similarity to the N-domain of VCP was also confirmed by FUGUE search (49,50). All protein sequences were extracted from ENTREZ.

Construction of Expression Vectors and Domain Boundary
Determination-GST fusion expression constructs containing the mouse PEX1 gene (encoding residues 3-180 or 3-131) and human PEX1 gene (encoding residues 3-180) were engineered according to the PRESAT vector methodology using an in-house pGEX-4T3-PRESAT vector (51). The GST fusion expression vectors containing the mouse PEX1 gene (encoding residues 3-263 and 3-310) were constructed according to a standard methodology. In brief, PCR-amplified cDNAs encoding the relevant regions of the PEX1 gene were first subcloned into pGEM-T vector (Promega) according to the standard T/A-cloning protocol. The DNA was subsequently subcloned into pGEX-4T3 between the BamHI-EcoRI sites.
Solubility of the GST-fused PEX1 N-terminal fragments was examined at the 10-ml culture scale. The fusion proteins were expressed in Escherichia coli strain BL21 (DE3) cultured in LBG medium at 30°C. Protein expression was induced by 1 mM isopropyl-D-thiogalactopyranoside, and the cells were harvested 3 h after induction. The cells were then resuspended in 0.5 ml of sonication buffer containing 0.15 M NaCl, 1% Triton X-100, 1 mM EDTA, and 20 mM sodium phosphate buffer (pH 7.5). Cells were disrupted by sonication, and centrifuged at 14,000 ϫ g for 5 min. Soluble and insoluble fractions were analyzed by SDS-PAGE.
Expression and Purification of PEX1 NTD-Mouse PEX1 NTD, encoding residues 3-180, was expressed as a fusion protein with glutathione S-transferase in LBG medium containing ampicillin (50 g/ml). The cells were grown to an A 600 of 0.3, and heterologous gene expression was induced by the addition of 1 mM isopropyl-D-thiogalactopyranoside. The cells were collected after 12-16 h of isopropyl-D-thiogalactopyranoside induction, pelleted, washed, and disrupted by sonication. Recombinant protein was purified by a single chromatography step using glutathione-Sepharose. The fusion protein GST-PEX1(3-180) was further purified for the structural studies. The GST tag was removed by treatment with thrombin. The protein solution was passed through a benzamidine-Sepharose 4B column and then purified by gel filtration using HiPrep 26/60 Sephacryl S-100 column (Amersham Biosciences).
Crystallization and Data Collection-Crystals of PEX1 NTD were obtained by the hanging drop vapor diffusion method using a solution of ammonium sulfate and polyethylene glycol 4000 as described elsewhere (64). The crystals belong to space group P3 2 with unit cell parameters of a ϭ b ϭ 63.5 Å and c ϭ 33.5 Å. A summary of data collection statistics is given in Table I. Diffraction data of PEX1 NTD for native crystals were collected using an FR-D rotating anode x-ray generator with an R-AXIS IV ϩϩ Imaging Plate (Rigaku, Tokyo, Japan). The mercurysoaked data were collected at Photon Factory BL-6A with a Quantum R4 CCD detector (ADSC). The wavelength was set to 1.0080 Å with a crystal-to-detector distance of 180 mm and an exposure time of 9 s per degree of oscillation. All of the diffraction data were processed with MOSFLM (52) and scaled with SCALA (53).
Structural Determination and Refinement-Molecular replacement, using the structure of mouse VCP (Protein Data Bank code 1e32) (54) or hamster NSF (Protein Data Bank code 1qcs) (55) as a search model, failed to solve the structure of mouse PEX1 NTD. This was attributed mainly to a rather low sequence homology with both of these models as well as to structural differences. The structure was eventually solved by single isomorphous replacement using the anomalous scattering (SIRAS) method. The heavy atom data were merged with Native-2 data, followed by scaling with SCALEIT (53). The heavy atom sites were searched using single-wavelength anomalous dispersion with SOLVE (56), followed by a SIR phase calculation with MLPHARE (53). The initial phases (mean figure of merit of 0.57) were improved by a density modification procedure with RESOLVE (57), resulting in a mean figure of merit of 0.83. The initial model output from RESOLVE was transferred to high resolution (2.05 Å) data (Native-1) by molecular replacement calculated with MOLREP (53). Further model building was done with O (58), and the structure was refined with CNS (59) to an R value of 21.4% and a free R value (8.2% of the data) of 25.5%. The Ramachandran plot of the final model, containing 146 amino acid residues from 13-67 to 71-179 plus one sulfate ion and 51 water molecules, shows that all of the amino acid residues are in the most favored (91.1%) position and an additional allowed (8.9%) region defined by the program PROCHECK (60). The structural determination and refinement statistics are summarized in Table I. Accessible surface areas were calculated using the Protein-Protein Interaction server (61). The figures are displayed by GRASP (62), MolMol (63), and PyMol (64). Structure comparison of PEX1 NTD against already defined structures was done using the DALI server (65).

RESULTS
Computational Sequence Analyses-Residues 1-600 of the N-terminal sequence of human PEX1, representing the preceding sequence to the D1 AAA-ATPase domain, was truncated into several pieces (ϳ250 amino acids) with various starting residues. Using these segments as query sequences, we scanned the protein nonredundant data base by PSI-Blast (45). When residues 1-250 of human PEX1 were used as a query and the threshold was set to 0.003, we recovered the N-terminal part of mouse VCP and yeast CDC48 (ϳ140 residues) at the second iteration, with E values of 1 ϫ 10 Ϫ22 and 1 ϫ 10 Ϫ18 , respectively. Further iterations recovered the N-terminal part of VAT, an archaeal homolog of VCP. All of these proteins belong to the type II AAA-ATPase family of enzymes. Since subsequent SMART analysis showed that the corresponding N-terminal sequences of VCP, CDC48, and VAT display similarity to the CDC48-2 domain, we assumed that PEX1 possesses a similar domain at the N terminus. The CDC48 -2 domain is always found to accompany the CDC48-N domain and form a single domain of ϳ180 amino acids at the N terminus of VCP, NSF, CDC-48, and VAT. This is referred to as the "N-domain." Thus, the sequence containing the entire N-domain was subjected to a reverse PSI-BLAST search. Using N-terminal mouse VCP (residues 1-192), we recovered human and Arabidopsis PEX1 proteins (e.g. ID ϭ 4505725, E ϭ 4 ϫ 10 Ϫ4 ; ID ϭ 30682405, E ϭ 2 ϫ 10 Ϫ3 , respectively) at the third iteration. Thus, on a statistical basis, we connected part of PEX1 NTD to the distant N-domain of type II AAA-ATPases. To find the boundary of this domain, the corresponding Nterminal sequences of ϳ200 amino acids from PEX1 orthologs from other organisms were aligned using ClustalX (46). The putative boundary of PEX1 NTD was defined at the position 180 Ϯ 5.
The putative PEX1 NTD sequences (residues 1-180) were subjected to analysis using the program FORTE (48). FORTE (fold recognition technique) is a profile-profile comparison tool for protein fold recognition. The program can be used to iden-tify similarities among protein families and their distant homologs. The structural similarity with the N-terminal region of VAT (Protein Data Bank code 1cz4) was suggested by the FORTE server at the highest probability (Z-score of 6.4). In addition, we found the structural similarity of PEX1 NTD to the N-terminal regions of NSF (Protein Data Bank code 1qcs) gave a Z-score of 5.0, which was the third highest value of the search. These results supported our finding based on the previous "low threshold" PSI-Blast search and a manual multiple alignment using ClustalX. Note that, when the whole N-terminal sequence (residues 1-600) of mouse PEX1 was subjected to FORTE search, no significant structural similarity was found.
Finally, the FUGUE server, which is an algorithm for sequence-structure comparison against the HOMSTRAD data base (66) was used with the N-terminal 180 residues of human PEX1 as a query. The results again showed remarkable homology of PEX1 NTD to the N-domain of mouse VCP (Protein Data Bank code 1e32), as judged by their Z-score of 17.86 (12.49 single mode).
Taking all of these results into account, we started a biochemical analysis of PEX1 NTD. The domain architectures of PEX6, PEX1, VCP, and NSF are summarized in Fig. 1a. Since sequence identity of PEX1 NTD to other distant homologs is Ͻ10% and the score is even lower at the N-and C-terminal ends of the domain, the boundary must be identified experimentally.
Determination of Domain Boundary-Recently, we proposed a PRESAT-vector methodology, using an asymmetric T/A-cloning vector for fusion protein expression as an efficient determination methodology of soluble protein domains (51). Using this methodology, four different constructs of a GST fusion of mouse PEX1 NTD (residues 3-310, 3-263, 3-180, and 3-131) were examined. Among the GST-fusion forms expressed in E. coli cells grown at 30°C, only the construct containing mouse PEX1 NTD (residues 3-180) was found in the soluble fraction. All other constructs gave insoluble protein (Fig. 2b). Thus, the putative domain boundary for PEX1 NTD was found to be around residue 180. Consequently, another GST fusion construct of human PEX1 (residues 3-180) was generated. Protein productivity and solubility of mouse and human PEX1 NTD were compared in the E. coli expression system (data not shown). Both of the fusion proteins were well expressed in E. coli (BL21(DE3)Star) at up to 40 mg/liter of LB medium. Mouse PEX1 NTD was approxi- mately twice as soluble as human PEX1 NTD, and roughly 70% of the expressed protein was recovered from the cells after optimizing conditions of protein expression and solubilization. The mouse PEX1 NTD construct was therefore used for the subsequent structural studies.
Structure of PEX1 NTD-The polypeptide chain of PEX1 NTD is folded into two globular subdomains (N-and C-terminal lobes) composed of an N-terminal double--barrel fold and a C-terminal ␤-barrel (67), referred to as the CDC48-N and CDC48-2 domains, respectively (43) (Fig. 3b). These two subdomains are connected by a short linker of four or five amino acids (Fig. 3a) to adopt a single kidney shaped molecule. In this crystal, the N-terminal residues (3-12) and residues 68 -70 and 180 were disordered. Indeed, the first 12 amino acids of the mouse and human PEX1 sequences were only conserved among mammalian PEX1 orthologs. The structure of PEX1 NTD is similar in overall folding topology to the structures of the N-domains of hamster NSF (Fig. 3c), mouse VCP (Fig. 3d), and their homologs yeast Sec18 and archaeal VAT, although the sequence identity to these proteins was rather low: 9, 6, 8, and 6%, respectively. The significant difference in sequence similarity is mainly due to insertions and deletions at the loops after ␤ 1 , ␣ 1 , ␤ 5 , ␤ 7 , and ␣ 3 as well as the region of helix ␣ 3 . Nevertheless, the hydrophobic residues from the ␤-strands, involved in the structural core, are relatively well conserved, as indicated in Fig. 1b. DALI Z-scores indicated that PEX1 NTD is most similar to the N-domain of VCP, which gave the highest Z-score of 16.3 and a root mean square deviation of 2.7 Å for the 155 aligned residues. In addition, the N-domain of NSF revealed the second highest Z-score of 16.1 with the lowest root mean square deviation of 2.1 Å. When the N-and C-lobes of PEX1 NTD were superimposed individually, the N-lobe of NSF and the C-lobe of VCP were the corresponding structural neighbors at Z-scores of 10.6 and 8.5 and root mean square deviation of 1.5 and 2.1 Å, respectively. The relative orientation between the N-and C-lobes of PEX1 NTD was tilted by ϳ5°in comparison with those of VCP, CDC48, NSF, and VAT. Therefore, the overall shape of PEX1 NTD is relatively straight and compact. The two lobes are associated with a large interdomain interface of ϳ410 Å 2 , which corresponds to 4.7% of the total surface area of PEX1 NTD. Since the averaged B-factor within each N-and C-terminal lobe was similar (27.1 and 30.9, respectively), a domain-domain hinge motion between them is unlikely to be present. The interaction between the two subdomains comprises total 25 pairs of hydrogen bonds, including 22 waterbridged and three direct hydrogen bonds, as well as six nonpolar contacts (Fig. 3f). Most of these contacts within the interface are contributed by residues from the first -loop (loop between ␤ 1 and ␤ 2 ) and helix ␣ 2 in the N-lobe and from consecutive ␣-helices and ␤-strands in the C-lobe (␣ 4 and ␤ 8 ). The residues participating in the interlobe interactions were also evolutionally conserved between most PEX1 orthologs as well as other the N-domains of AAA-ATPases (Figs. 1b and 3f).
Comparison of Loop and Surface Structure-The structure revealed a shallow groove between the N-and C-lobe along the equator of the domain, with a radius of ϳ5 Å and a length of about 26 Å. In the structure, a sulfate ion was found at the edge of a parallel ␤-sheet composed of ␤ 2 and ␤ 5 (Fig. 3e). The sulfate ion is held by four charged residues: His 30 , Arg 64 , His 65 , and Glu 75 . These residues form a small patch of charged residues, which are located close to the end of the shallow groove. The electrostatic features of the surface were calculated as shown in  Fig. 4. The surface of PEX1 NTD is generally acidic, which is consistent with the pI value of 5.7 assumed from the primary sequence. This feature is characteristic of PEX1 NTD when compared with the N-domains of NSF and VCP, which are mostly neutral or slightly basic (Fig. 4, c-f).
The bottom surface of the C-lobe of PEX1 NTD is unique in possessing an acidic cluster on helix ␣ 3 , the corresponding region in NSF and VCP comprising a loop (Fig. 4b). In addition, the residues that constitute this negatively charged area in PEX1 NTD, Asp 114 , Asp 115 , Glu 117 , and Glu 120 , are highly conserved in all PEX1 orthologs, along with Trp 116 . We propose that this surface containing the cluster of negatively charged residues may play an important role unique to PEX1. In contrast, except for Arg 135 and Lys 174 , there are very few positively charged residues exposed on the protein surface. These arginine and lysine residues are highly conserved among other PEX1 orthologs from yeast to mammal. This positively charged area is located at the end of the shallow groove.

Supradomain Structures Conserved in Type II AAA-ATPases-
In this paper, we have demonstrated that PEX1 possesses a domain architecture similar to that of other type II AAA ATPases (VCP, NSF, and its archaeal ancestor VAT), which are composed of an N-D1-D2 structure. The sequences of the AAA-cassettes were extensively analyzed to generate phylogenic relationships (23). Based on the sequence homology within these AAA-cassette regions, the domain was classified into 12 subgroups by Beyer. A further refined version of boot-strapped phylogenic tree was presented by Frohlich (68). According to these analyses, the two AAA-cassettes of PEX1, D1 and D2, belong to the same subfamilies as that of VCP, NSF, and VAT, respectively. Thus, the existence of another genetically related domain at the N terminus of PEX1 demonstrates a clear relationship to VCP, NSF, and VAT and the potential existence of a common ancestral molecule.
Recently, a concept of a persistently retained domain architecture found in multiple-domain proteins, known as a "supradomain," has been proposed by Teichman (42,69). This hypothesis focuses on the conservation of pairing and the sequential order of multiple-domain proteins in a different context. The underlying reason for the conservation of a domain pair in evolution can be explained in two ways. Either the interface between the two domains has a specific function, or the molecular functions of the two domains are related to each other in a biological context. PEX1 NTD is composed of a structurally independent N-and C-lobe, with a shallow groove between the two lobes. Similar grooves were also found in VCP, CDC48, NSF, and VAT. In the N-domain of NSF and yeast Sec18, this groove is assumed to be a adaptor binding site accepting the C-terminal helix of ␣-SNAP (55, 70, 71), although no structural evidence has been  Fig. 1b are colored in yellow. reported. We propose that this groove in PEX1 NTD may serve as a putative adaptor binding site similar to that in NSF, based on the structural similarity. Thus, the evolutional conservation of the two lobes may be explained in terms of the need to retain function at the interfacial boundary.
In contrast, the relationship between the N-domain and the D1-D2 domains is not likely to meet the criteria discussed above. The length of the amino acid segments between the N-domain and the D1-D2 domains of PEX1 significantly differs from those of the other AAA-ATPases. The lengths of linker sequences between the N-D1 domains of VCP, CDC48, NSF, Sec18, or VAT are in the range of 30 -60 residues. In these latter cases, the N-domain can directly contact with the D1 domain, as observed in VCP (72,73). In contrast, the length of the linker between these domains in PEX1 is as long as 250 residues. We found that the spermatogenesis-associated factor also belongs to this group. The architectures of type II AAA ATPase with their known subcellular localizations and/or cellular functions are summarized in Table II. We hypothesized that the relationship between the N and D1-D2 domains may be rationalized in terms of the related function of the two independent domains in a biological context. Some AAA-ATPases have an active ATP motor as well as a regulatory domain, such as an adaptor binding domain and/or substrate recognition site. We propose a model in which the N and D1-D2 domains are linked in terms of the adaptor-binding domain and an ATP motor. To date, except for PEX6, there has not been identified either target or adaptor molecule specific for PEX1 (40,41). A strategy of searching for PEX1-specific adaptors that binds PEX1 NTD is a considerable choice.
Functional Aspect of N-terminal Domain of PEX1-Comparison of the structure of PEX1 NTD with the N-domains of VCP and NSF suggests that it may possess a putative adaptor or substrate binding site. The presence of the shallow crevice between the N-and C-lobes in PEX1 NTD clearly showed the structural similarity to the N-domain of NSF, which bind to ␣-SNAP using the corresponding crevice (55,71). On the other hand, the crystal structure of the N-domain of VCP has been solved with D1 domain or D1-D2 domains, both of which were in hexameric form (Protein Data Bank codes 1e32 and 1oz4/ 1r7r, respectively). In addition, the complex with its specific adaptor p47 has been solved (1s3s). Thus, we analyzed the surface features of PEX1 NTD corresponding to the interdomain and/or intermolecular interfaces observed in VCP.
The N/D1 interface at the N-domain of VCP was mainly composed of the loop between ␤ 7 and ␣ 3 and the loop between ␤ 8 and ␤ 9 . A significant structural difference was found at this loop region, which corresponds to the extra helix ␣ 3 in PEX1 (Fig. 3d). The helix ␣ 3 comprises well conserved acidic residues. Thus, the corresponding surface of PEX1 NTD is very acidic (Fig. 4, a and b), whereas the same surface of VCP N-domain is not (Fig. 4, e and f). The corresponding surface of NSF is also not acidic (Fig. 4, c and d). Then we compared the residues involved in the N/D1 interface of VCP with the corresponding residues of PEX1 NTD (Fig. 5, a and b). Hydrophobic and neutral residues on the C-lobe of VCP N-domain, as well as four arginine residues (Arg 86 , Arg 89 , Arg 95 , and Arg 155 ) on the Nand C-lobes, contributed to the N/D1 interface. These residues were not conserved in PEX1 NTD. We further analyzed a sequence conservation of the interfacial residues on the D1 domains of VCP and PEX1. The D1/N interface at the D1 of VCP is composed of helix ␣ 13 and a loop between ␣ 10 and ␣ 11 , all of which are on the C-terminal ␣-helical subdomain of the AAA-domain. Note that this short helix ␣ 13 is characteristic of D1 but not D2 in type II AAA-ATPases. These D1 sequences are entirely conserved in VCP orthologs (Ͼ90% identity), whereas the corresponding D1 sequences of PEX1 are far less conserved in PEX1 orthologs. Thus, an interaction between the N and D1 domains in PEX1 is not likely or at least not similar to that found in VCP. This is consistent with the presence of a long  insertion (ϳ250 residues) between NTD and D1 in PEX1.
The N-terminal domain of PEX1 may be involved in interactions with ubiquitin, ubiquitin-like protein modifiers, or ubiquitin-like domains, such as Ubx. There are several lines of evidence for this possibility (1). In case of VCP, not only the Ubx domain of p47, but also VCIP135, may interact with its ubiquitinlike domain (32, 74) (2). A weak but direct binding of the Ndomain of VCP to a polyubiquitylated substrate was postulated (3,75,76). Interactions have been reported between NSF or Sec18, with the ubiquitin-like protein modifiers, GATE-16, GABARAP, and yeast Atg12 (41,(77)(78)(79)(80). Together with the structural similarity among the NTDs of PEX1, VCP, and NSF, we hypothesize that PEX1 may bind to ubiquitin or Ubx domains.
To test the hypothesis, we compared the important residues on the interface at the N-domain of VCP to p47 Ubx domain (81) with the corresponding residues of PEX1 NTD (Fig. 5, c  and d). The N-domain (VCP)/Ubx (p47) interface is located at the opposite site of a crevice between the N-and C-lobes and overlapped with the linker between them. The interface is essentially hydrophobic and composed of three conserved residues, Val 108 , Tyr 110 , and Tyr 143 , as well as some other hydrophobic residues. This hydrophobic interface is capable of accepting a hydrophobic patch on Ubx (p47). In PEX1 NTD, the corresponding surface contains only a few hydrophobic residues without forming a patch. Based on the different shape and the hydrophilic nature of this surface, PEX1 NTD is not likely to bind either Ubx or ubiquitin in a manner similar to that of the N-domain (VCP)/Ubx (p47) interface.
Analogy of Physiological Function of PEX1-PEX6 Systems at Peroxisome Biogenesis with Endoplasmic Reticulum-associated Protein Degradation and Vesicle Fusion-The striking structural similarity among NTDs of PEX1, VCP, and NSF led us to propose the involvement of PEX1 in peroxisomal biogenesis analogous to the functions of VCP and NSF. The PEX1-PEX6 complex is thought to participate in two distinct steps of peroxisomal biogenesis, membrane fusion and protein translocation. First, the complex is required for homotypic peroxisomal membrane fusion (20), in which some remnant of the peroxisomal precursor membrane comes from the endoplasmic reticulum (82). Second, the ATPase activity of the PEX1-PEX6 complex is required for the last step of peroxisome transport signal import into the peroxisome (83). The entry of PEX5-cargo complexes is ATP-independent, whereas recycling of PEX5 is an ATP-dependent step (6). The first function is similar to the common function of VCP and NSF, whereas the second function resembles that of VCP in endoplasmic reticulum-associated protein degradation. In these events, VCP and NSF may function as an ATP-dependent protein unfoldase at the organellar surface. The evident structural similarity among the N-terminal domains of PEX1, VCP, and NSF further strengthens the analogy among the functions of these molecules.
In conclusion, the similarity among NTDs of PEX1 and the FIG. 5. Comparison of surface residues of PEX1 NTD corresponding to the interdomain interfaces found in VCP. a and b, the residues found in the surface of PEX1 NTD (a) and the corresponding residues in the N/D1 interface at VCP (b) are shown. c and d, the residues found in the surface of PEX1 NTD (c) and the corresponding residues in the Ndomain (VCP)/Ubx (p47) interface (d) are shown. The acidic, basic, hydrophobic, and neutral residues are shown in red, blue, yellow, and cyan, respectively. The rotation angles in a and b are Ϫ45°along the y axis relative to Fig. 4, a and e, respectively. The rotation angles in c and d are ϩ45°along the y axis relative to Fig.  4, a and e, respectively. other membrane-related type II AAA-ATPases in terms of molecular evolution of supradomain architecture will introduce new insights into the physiological function of PEX1 in peroxisomal biogenesis.