Domain Mapping of Human PEX5 Reveals Functional and Structural Similarities to Saccharomyces cerevisiae Pex18p and Pex21p*

PEX5 functions as an import receptor for proteins with the type-1 peroxisomal targeting signal (PTS1). Although PEX5 is not involved in the import of PTS2-targeted proteins in yeast, it is essential for PTS2 protein import in mammalian cells. Human cells generate two isoforms of PEX5 through alternative splicing, PEX5S and PEX5L, and PEX5L contains an additional insert 37 amino acids long. Only one isoform, PEX5L, is involved in PTS2 protein import, and PEX5L physically interacts with PEX7, the import receptor for PTS2-con-taining proteins. In this report we map the regions of human PEX5L involved in PTS2 protein import, PEX7 interaction, and targeting to peroxisomes. These studies revealed that amino acids 1–230 of PEX5L are required for PTS2 protein import, amino acids 191–222 are sufficient for PEX7 interaction, and amino acids 1–214 are sufficient for targeting to peroxisomes. We also identified a 21-amino acid-long peptide motif of PEX5L, amino acids 209–229, that overlaps the regions sufficient for full PTS2 rescue activity and

Peroxisomes are ubiquitous organelles of eukaryotic cells that participate in a wide variety of metabolic functions (1,2). Peroxisomes lack nucleic acids, and peroxisomal proteins are encoded by nuclear genes. Enzymes that are destined for the peroxisome lumen, or matrix, are synthesized in the cytoplasm and imported post-translationally (3). Two targeting signals direct proteins into the peroxisome lumen (4). The type-1 peroxisomal targeting signal, or PTS1 1 , is found on the vast majority of matrix enzymes and consists of just three amino acids at the extreme C terminus of the enzyme (5). Although the canonical PTS1 is serine-lysine-leucine-COOH, many sequence variants of the PTS1 have been described in mammalian cells (6 -12) and even more in yeast (13,14), protozoa (15)(16)(17), and plants (18). The type-2 peroxisomal targeting signal, or PTS2, is found on only a small number of peroxisomal enzymes and is located at or near the N terminus of proteins (19,20). The canonical PTS2 is arginine-leucine-X 5 -histidine-leucine, though sequence variants of the PTS2 have also been described previously (21)(22)(23).
The peroxisome biogenesis disorders (PBDs) are a group of lethal neurological diseases caused by defects in peroxisomal matrix enzyme import and peroxisome assembly (55). The PBDs can be caused by mutations in any of at least 12 different human PEX genes, including PEX5 and PEX7 (54,56). Muta-tions in PEX5 can cause Zellweger syndrome, neonatal adrenoleukodystrophy, or infantile Refsum disease (8,27). These three diseases represent a phenotypic continuum and are characterized by a loss or reduction of virtually all peroxisomal metabolic functions (56,57). In contrast to the Zellweger spectrum of disease caused by PEX5 mutations, mutations in PEX7 cause rhizomelic chondrodysplasia punctata (36,37,39,58,59). Patients with rhizomelic chondrodysplasia punctata are characterized by defects in PTS2 protein import, normal PTS1 protein import, and a more restricted set of metabolic and developmental abnormalities that involve PTS2-targeted enzymes (55,60).
Studies of human PEX genes and cells derived from PBD patients have contributed significantly to our understanding of peroxisome biogenesis (54,56,61). This is particularly true in regard to PEX5. Human cell studies were the first to show that PEX5 is a predominantly cytoplasmic, partly peroxisomal protein (27) that cycles between the cytoplasm and peroxisome (40,42), and the crystal structure of the human PEX5⅐PTS1 complex has provided a molecular model of PTS1 recognition (32). In addition, human cell studies were the first to show that mammalian PEX5 is required for PTS2 protein import (27,29,62). The requirement for PEX5 in PTS2 protein import represents a significant departure from the PTS2 protein import pathway in yeast, in which PTS2 protein import occurs normally in the absence of PEX5 (24,25). Studies in human cells, Chinese hamster ovary cells, and transgenic mice have also established that mammalian PEX5 is expressed in at least two forms, PEX5L and PEX5S, which are generated by alternative splicing of a single PEX5 gene (27,(63)(64)(65). PEX5L contains an insert 37 amino acids long that is positioned between amino acids 214 and 215 of PEX5S. PEX5L also differs from PEX5S in that it can rescue PTS2 protein import in PEX5-deficient cells at a high frequency (64,65) and appears to interact with PEX7, the PTS2 receptor (66).
These and other results have led to the hypothesis that PEX5L plays an essential role in PTS2 protein import and that this role involves binding to PEX7 and facilitating PEX7 transport to peroxisomes (66). However, many issues remain to be addressed regarding the role of PEX5L in PTS2 protein import. In this report, we have identified sequences of PEX5L that are necessary for PTS2 protein import, interaction with PEX7, and transport to peroxisomes. We also identified a short peptide motif shared by PEX5L and the Saccharomyces cerevisiae peroxins Pex18p and Pex21p, which participate only in PTS2 protein import. These and other results raise the possibility that PEX5L may represent a functional and structural homolog of the yeast peroxins Pex18p and Pex21p.

Cell Lines and Culturing Conditions
The human skin fibroblast cell lines were kindly provided by A. B. Moser and H. W. Moser (Kennedy Krieger Institute, Baltimore, MD). All cell lines were transformed with the large T-antigen of SV40 virus as described previously (27). Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum, 2 mM L-glutamine, 100,000 units/liter penicillin, and 100 mg/liter streptomycin at 9% CO 2 . Wild-type cells have normal PTS1 and PTS2 protein import. PBD005 cells are homozygous for a mutation in PEX5, R390ter, and do not import detectable levels of PTS1 or PTS2 proteins (55). PBD054 cells are homozygous for a mutation in PEX12, S320F, have detectable PTS1 and PTS2 protein import, and import PEX5 into peroxisomes (55). PBD061 cells lack peroxisome membranes, and are homozygous for a mutation in PEX16, R176ter (55).

Transfections, Indirect Immunofluorescence Microscopy, and Antibodies
Transfections of fibroblasts were performed by electroporation using the protocol outlined by Chang et al. (67) or using LipofectAMINE (Life Technologies) according to the manufacture's suggestions. Two days after transfection the cells were processed for indirect immunofluorescence microscopy as described in Slawecki et al. (68). Standard permeabilization was for 5 min with 1% Triton X-100, which permeabilizes both plasma and peroxisome membranes. Differential permeabilization was for 5 min with 25 g/ml digitonin, which permeabilizes the plasma membrane but does not permeabilize the peroxisome membrane. The micrographs were made using a Zeiss Axiophot microscope and Kodak Ektachrome Elite 400 or TMAX 400 film.
Polyclonal sheep anti-human catalase antibodies were obtained from The Binding Site. Polyclonal rabbit anti-GFP antibodies were purchased from CLONTECH. The monoclonal anti-myc (9E10) antibodies (69) were prepared by ammonium sulfate precipitation (25-50%, w/v) and diafiltration from serum-free cell culture supernatants. Cy3-and Cy2-conjugated secondary antibodies were obtained from Dianova. Other fluorescent antibodies were purchased from Jackson Immunochemicals. Antibodies against chloramphenicol acetyltransferase (CAT) were obtained from 5 Prime 3 3 Prime, Inc. (Boulder, CO). Anti-PEX5 antibodies were generated in rabbits against a recombinant His 6tagged version of PEX5L. The plasmid encoding full-length PEX5L, an N-terminal hexahistidinyl (His 6 ) tag and a tobacco etch virus protease cleavage side in a derivative of pET9d (Novagen) has been described before (70). This plasmid was transformed into Escherichia coli BL21(DE3), and recombinant PEX5L protein was purified under native conditions according to Schliebs et al. (70).
N-mycPEX7 (pEB13.10), containing the c-myc epitope encoding sequence upstream of the PEX7 coding region, and the multiple cloning site between the EcoRI and XbaI sites of pcDNA3 downstream of the PEX7 coding region (see PEX7pcDNA3 (37)), was cloned into the HindIII and XbaI sites of pcDNA3.1Zeo. The start codon of PEX7 was omitted. The sequence upstream of the second PEX7 codon reads as follows: (5Ј-A AGC TTC ACC ATG GAG CAG AAG CTG ATC AGC GAG GAG GAC CTG GGA TCC AGG TAC CTT CTG-3Ј). The nucleotides encoding the myc tag are underlined. PEX7 in pEGFP was created by cloning an Acc65I and ApaI fragment of pEB6.4 (PEX7 in pM, see below) into the corresponding sites of pEGFP-C1 (CLONTECH).
Yeast Two-Hybrid System-The full-length PEX5L and PEX5S coding regions were excised from pGD106 and pGD100 with NcoI (blunted) and BglII and cloned into the SmaI and BglII sites of the activating domain vector pPC86 (71). PEX5S/1-228 in pPC86 (pGD142) was generated from pJM165.1 that contained amino acids 1-298 of PEX5 in pcDNA3 by cleavage with NcoI (blunted) and XbaI and subsequent cloning into the SmaI and SpeI site of pPC86. The plasmid pJM165.1 was generated by PCR using primers DV1037 and P299ter with pGD100 as template ( Table I). The PCR fragment was cleaved with Sse8387I and XbaI and cloned into the corresponding sites of pGD100. PEX5L/1-251 in pPC86 (pEB4.2) was generated by PCR using the primer pair T7 and GD30 with pGD106 as template. The fragment was cleaved with EcoNI and NotI and cloned into the respective sites of pGD142. PEX5L/48 -298 in pPC86 (pEB2.2) was made by PCR with primers GD31 and P299ter using pGD106 as template. The fragment was cleaved with SalI and XbaI and cloned into the SalI and SpeI site of pPC86. PEX5L/139 -298 and PEX5L/ 214 -298 in pPC86 were cloned in the same way, but the sense primers GD32 and GD33 were used in the PCR reactions. The ScPEX7 DNA in pPC97 was described by Rehling et al. (34).
Mammalian Two-Hybrid System-HsPEX7 in pM (pEB6.4) was generated from pEB6.6 by excision of the full-length DNA with EcoRI and cloning it into the EcoRI site of pM. Plasmid pEB6.6 was generated by amplifying a fragment with primers GD47 and Sp6 using PEX7pcDNA3 (37) as template. This fragment was then subcloned into pGEM-T (Promega) following the manufacturer's instructions. HsPEX7 in pVP16 was cloned in a similar way generating pEB6.5. To generate PEX5L and PEX5S in pVP16 (pEB7.5 and pEB8.5), corresponding fragments were cleaved with NcoI (blunted) and BglII from pGD106 or pGD100, and cloned into the EcoRI site (blunted) and BamHI site of pVP16. HsPEX5L/1-335 in pVP16 was generated by cleaving pJM164.7 with NcoI (blunted) and XbaI and ligating the fragment into the EcoRI (blunted) and XbaI sites of pVP16. The short version HsPEX5S/1-298 in pVP16 (pEB 10.5) was derived in a similar way from pJM165.1 that encoded amino acids 1-298 in pcDNA3. HsPEX5L/1-251 in pVP16 (pEB4.5) was created by cleaving the NcoI (blunted)/XbaI fragment out of pEB4.1 and ligating it into the EcoRI (blunted) and XbaI sites of pVP16. Plasmid pEB4.1 encodes PEX5L/ 1-251 in pcDNA3 and was derived by PCR with the primer pair GD30 and T7 using pGD106 as template. An EcoNI/NotI fragment was then ligated into the corresponding sites of pGD100 to generate pEB4.1. HsPEX5L/48 -335 in pVP16 (pCK7) was generated by excising an NcoI (blunted)/XbaI fragment from pEB2.1 and cloning this fragment into the EcoRI (blunted) and XbaI sites of pVP16. To generate pEB2.1 (HsPEX5L/48 -335pcDNA3) a fragment was amplified by PCR with primers GD31 and P299ter using pGD106 as template. This fragment was cut with SalI and XbaI and cloned into the XhoI and XbaI sites of pcDNA3. The coding region of PEX5L/138 -335 was amplified from pGD106 using primers GD32 and P299ter. The fragment was cleaved with SalI and XbaI and ligated into the XhoI and XbaI sites of pcDNA3. This plasmid was cut with NcoI (blunted) and XbaI and cloned into the EcoRI (blunted) and XbaI sites of pVP16 to generate HsPEX5L/138 -335 in pVP16 (pEB3.5). The PEX5L/214 -335 encoding fragment in pVP16 (pEB1.5) was cloned in a similar way using primers GD33 and P299ter. To create HsPEX5L/191-251 in pM (pEB16.4) the plasmid pEB4.1 encoding PEX5L/1-251 in pcDNA3 was cut Bsu36I (blunted) and XbaI and cloned into the SmaI and XbaI sites of pM. The latter vector was cleaved with EcoRI and XbaI, and the similar fragment was ligated into the corresponding sites of pVP16 (pEB16.5). The coding region for HsPEX5L/191-222 was cleaved with BspMI (blunted) and SacI out of pEB16.5 and ligated into pVP16 (HindIII (blunted) and SacI) to give HsPEX5L/191-222 in pVP16 (pEB17.5). The sequence coding for HsPEX5/1-131 was amplified with the T7 primer and primer Ku582 using pGD100 as template. This fragment was cut with NcoI (blunted) and SalI and cloned into the EcoRI (blunted) and SalI sites of pVP16 to give HsPEX5/1-131pVP16. ScPEX21 in pPC86 was provided by W. Schliebs (Bochum). The coding region was cleaved with SalI (blunted) and XbaI and ligated into the SalI (blunted) and SpeI sites of pM and pVP16, respectively, to get ScPex21pM and ScPex21pVP16. HsPEX13 was cloned by reverse transcription-PCR (72) using GD144 and GD145 as PCR primers, and mRNA isolated from wild-type fibroblasts as template (Oligotex Direct mRNA kit, Qiagen). A BamHI/XbaI fragment was cloned into the corresponding sites of pVP16 and pM to give HsPEX13pVP16 and HsPEX13pM. The PEX13SH3 plasmids encode amino acids 257-447 of PEX13 in pVP16 and pM. A BamHI/XbaI fragment was generated by PCR using the primer pair Ku341/Ku342. A plasmid containing this fragment in pSK was kindly provided by G. Will (Bochum). The PEX14 plasmid in pVP16 has been described previously (52).

Immunoprecipitations
Transcription/Translation-PEX5S (pGD100), PEX5L (pGD106), and N-mycPEX7 (pEB13.10) were transcribed and translated in vitro for 1 h, using the TNT Coupled Reticulocyte Lysate system (Promega). PEX5 and PEX7 were labeled with [ 35 S]cysteine (1075 Ci/mmol) (PerkinElmer Life Sciences). Equal amounts of the PEX5L, PEX5S, and N-mycPEX7 translation reactions (10 l) were mixed and incubated together for an additional hour at 30°C. The reaction mixture was diluted to 175 l with binding buffer (20 mM Hepes, pH 7.3, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium acetate, 1 mM EDTA, 1 g/ml leupeptin, 1 g/ml pepstatin, and 1 mM phenylmethylsulfonyl fluoride, 21 g/ml NaF) containing 0.3% Triton X-100 and incubated for 2 additional hours at 4°C with 25 l of anti-mouse IgG Dynabeads (Dynal), saturated with anti-myc antibodies. Immunoprecipitates were collected using a magnet. The precipitates were washed five times with 1 ml of binding buffer with 0.3% Triton X-100, 0.005% SDS, resuspended in 20 l of SDS sample buffer, denatured for 5 min at 95°C, and separated by SDS-PAGE. Supernatant and pellet fractions were loaded with a ratio of 12.5 to 1. The gel was soaked in 0.5 M sodium salicylate for 30 min, dried, and subjected to fluorography.
Cell Lysates-PBD005 cells in 60-mm dishes were transfected with PEX5S-, PEX5L-, and N-mycPEX7-expressing plasmids using Lipo-fectAMINE as described. Two days later cells were washed with Hanks' solution and lysed in 0.8 ml of binding buffer (see above) supplemented with 0.5% Triton X-100 on ice for 30 min. The lysates were cleared at 15,000 ϫ g for 15 min. Supernatants (600 l) were incubated with 50 l of anti-myc-coated mouse IgG Dynabeads (Dynal) for 2 h. Immunoprecipitates were washed with binding buffer with 0.5% Triton and 0.02% SDS as described above, resuspended in 36 l of non-reducing SDS sample buffer, denatured for 5 min at 95°C, and separated from the beads. For analysis, samples were supplemented with 1 l of ␤-mercaptoethanol, denatured again for 5 min at 95°C, and separated by SDS-PAGE. The ratios for the supernatant and pellet fractions loaded onto the gel were 20 to 1. The proteins were transferred to a PVDF membrane and detected with anti-PEX5 antibodies using the ECL system (Amersham Pharmacia Biotech).

Mammalian Two-hybrid Methodology and ELISA
To investigate a possible interaction between human peroxins we used the mammalian MATCHMAKER two-hybrid assay kit (CLONTECH). The corresponding cDNAs were cloned behind the GAL4 binding domain sequence of the pM vector or the VP16-activating domain sequence of the pVP16 vector, as described under "Plasmids." Human wild-type skin fibroblasts (GM5756) or PEX16-deficient fibroblasts (PBD061) were seeded onto 12-well plates to reach 50 -80% confluency the next day. The cells were co-transfected with 0.24 g of DNA of pM and pVP16 derivatives together with 0.05 g of reporter plasmid pG5CAT, using 1.5 l of LipofectAMINE. Two days after transfection, the cells were lysed in 0.2 ml of lysis buffer (CAT ELISA kit (Roche Molecular Biochemicals)) for 30 min at 4°C, and the supernatants were cleared at 15,000 ϫ g for 15 min. The amount of proteins in cell lysates was estimated using the BCA protein assay reagent (Pierce) and serum albumin as standard. Lysates corresponding to 50 g of protein were used to determine the amount of chloramphenicol acetyltransferase with the CAT ELISA kit. Similar transfected samples were controlled for expression of the fusion proteins and for CAT expression by immunofluorescence microscopy, using polyclonal antibodies against the GAL4 binding domain (Santa Cruz Biotechnology) and against chloramphenicol acetyltransferase (5 Prime 3 3 Prime, Inc.).

Yeast Two-hybrid System
The cDNAs were fused to the activation domain of GAL4 in pPC86 or the GAL4 binding domain in pPC97 (71) as described above. Co-transformations of two-hybrid vectors into strain HF7c (CLONTECH) or PCY2 (71) and PCY2⌬pex14 (73) were performed with lithium acetate following a protocol from CLONTECH (Yeast Protocols Handbook). The ␤-galactosidase filter assay and the test for HIS prototrophy were performed after selection on SD plates lacking leucine and tryptophan according to Rehling et al. (34).

RESULTS
Identification of the Minimal PTS2 Import Domain of PEX5L-PBD005 cells were derived from a severely affected Zellweger patient and are homozygous for a nonsense mutation in PEX5, R390ter (27). This mutation results in the destabilization of PEX5 mRNAs, the absence of detectable PEX5 pro-

5Ј-TCCTGCAGGACCAGAATGC-3Ј
P299ter Domain Mapping of Human PEX5 tein, and eliminates import of both PTS1-and PTS2-containing proteins (27,64). Additional studies have established that mammalian cells express two forms of PEX5 that differ by 37 amino acids and that these forms are generated by alternative splicing of a single exon of the PEX5 gene (27,64,65). These forms appear to differ in their biological activities, with PEXL efficiently rescuing both PTS1 and PTS2 protein import and PEX5S involved only in PTS1 protein import. This difference can be detected using the following in vivo assay. Plasmids designed to express (a) PTS2-CAT, a plasmid designed to express bacterial chloramphenicol acetyltransferase (CAT) with a PTS2 at its N terminus (19) and (b) either PEX5S or PEX5L (64), were simultaneously introduced into PBD005 cells by electroporation. Two days after transfection, the cells were processed for indirect immunofluorescence microscopy (Fig. 1). The relative ability of PEX5S and PEX5L to mediate PTS2 protein import was calculated by comparing the percentage of cells in which PTS1 protein import was detected to the percentage of cells in which PTS2 protein (PTS2-CAT) import was detected. For PEX5L, the values were equivalent whereas PEX5S was 100 times less efficient at rescuing PTS2 protein import than it was at rescuing PTS1 protein import.
To better understand the molecular basis of PEX5-mediated PTS2 protein import, we mapped the minimal PTS2-import domain of PEX5L using the immunofluorescence-based in vivo assay described above. The C-terminal half of PEX5 is involved in binding the PTS1, and we hypothesized that it might be dispensable for PTS2 protein import. We removed this domain by cloning the upstream codons of PEX5S (1-298) and PEX5L (1-335) into a mammalian expression vector that is designed to add a c-myc epitope tag to the C terminus of each gene product. The resulting clones (PEX5S/1-298myc and PEX5L/1-335myc) contained 12 codons encoding a c-myc epitope tag (GSQN-LISQQDL-stop) in place of the last 304 codons of PEX5S and PEX5L. PEX5S/1-298myc and PEX5L/1-335myc were then tested for their ability to rescue PTS1 and PTS2 protein import in PBD005 cells (Fig. 2). These proteins were unable to rescue PTS1 protein import, and PEX5S/1-298myc was unable to rescue PTS2 protein import. However, expression of PEX5L/1-335myc rescued PTS2 protein import with nearly the same efficiency as full-length PEX5L. We generated additional Cterminal truncation mutants of PEX5S and PEX5L that removed their C-terminal 339 amino acids (PEX5S/1-263myc and PEX5L/1-300myc) or their C-terminal 375 amino acids (PEX5S/1-227myc and PEX5L/1-264myc). Neither of these proteins was capable of rescuing PTS1 protein import, but the "L" versions were capable of rescuing PTS2 protein import. Removal of the N-terminal 48 amino acids of PEX5S and PEX5L eliminated the ability of the proteins to rescue PTS1 and PTS2 protein import.
PEX5L/1-264myc contained all sequences upstream of the 37-amino acid insert, the entire insert, plus the 13 amino acids C-terminal to the 37-amino acid insert. We created four more mutants of PEX5L that removed an additional 23 (PEX5L/1-241myc), 34 (PEX5L/1-230myc), 42 (PEX5L/1-222myc), and 50 (PEX5/1-214myc) amino acids from the C terminus of PEX5L/ 1-264myc (Fig. 3). PEX5L/1-241myc and PEX5L/1-230myc lacked the C-terminal 10 and 21 amino acids of the 37-amino acid insert, respectively, but still rescued PTS2 protein import with high efficiency. The next smallest mutant that removed the C-terminal 29 amino acids of the 37-amino acid insert (PEX5L/1-222myc) displayed only a borderline ability to rescue PTS2 protein import. Loss of the entire insert (PEX5/1-214myc) eliminated the ability of PEX5 to rescue PTS2 protein import in PBD005 cells. Thus, we find that the C-terminal 21 amino acids of the insert can be removed from PEX5L without eliminating its ability to rescue PTS2 protein import.
Identification of a PEX7 Binding Domain within PEX5L-In a previous study, Fujiki and colleagues reported that PEX5L can bind PEX7 whereas PEX5S cannot (66). To map the site of interaction between PEX5L and PEX7 we first examined our ability to detect this interaction in vivo. PBD005 cells were transfected with plasmids designed to express N-mycPEX7 (a fully functional form of the protein (37)), PEX5S, PEX5L, or co-transfected with combinations of the PEX7 and PEX5 expression vectors. Two days after transfection, lysates were prepared from each transfected cell population and subjected to immunoprecipitation using anti-myc antibody beads. The resulting immunoprecipitates (IPs) were then separated by SDS-PAGE, transferred to a PVDF membrane, and probed with antibodies specific for PEX5 (Fig. 4a). PBD005 cells do not express any PEX5 proteins, and PEX5 proteins were not detected in the IP from cells expressing only N-mycPEX7. IPs from cells expressing either PEX5S or PEX5L contained a low level of PEX5, which represents our background of PEX5 binding to anti-mouse IgG Dynabeads. Relative to this background, the amount of PEX5S in the IP from cells expressing PEX5S and N-mycPEX7 was negligible. In contrast, the amount of PEX5L detected in the IP from cells expressing N-mycPEX7 and PEX5L was significantly higher than background.
The signal for PEX5L⅐PEX7 interaction detected above was not particularly robust and led us to doubt whether we could use this assay to map the region of PEX5L that mediates its interaction with PEX7. We next tested whether we could detect this interaction in a slightly more defined system using radiolabeled proteins. Specifically, we synthesized N-mycPEX7, PEX5S, and PEX5L in the presence of [ 35 S]cysteine in rabbit reticulocyte lysates for 1 h, mixed equal amounts of the respective pairs, incubated them for an additional hour at 30°C, and then subjected each sample to immunoprecipitation with antimyc antibodies. We were unable to detect interaction between PEX5L and PEX7 in this system (Fig. 4b). We also tried to assess binding between PEX5L and PEX7 using purified recombinant proteins expressed in bacteria in a variety of blot overlay, bead binding, and other protein⅐protein association assays. These experiments also failed to show an interaction between PEX5L and PEX7. These protein⅐protein interaction assays rely on a number of assumptions regarding the relative affinity of the binding reaction, the conformational constraints of the interaction, the conformational status of the bacterially expressed proteins, and the ability of the interaction to occur in the absence of other factors. Thus, these results should not be interpreted as evidence against direct PEX5L⅐PEX7 binding. They did force us, however, to search for another assay system in which we could examine the PEX5L⅐PEX7 interaction.  2. The N-terminal but not the C-terminal region of PEX5L is required for PTS2 import. PBD005 cells were co-transfected with the constructs expressing the various PEX5 proteins shown in e and with the plasmid expressing PTS2-CAT. After 2 days the cells were processed for double-indirect immunofluorescence microscopy with anti-CAT antibodies (a, c) and anti-PMP70 antibodies (b, d). PEX5S/1-227myc (a, b) did not rescue PTS2 import (a) whereas the long isoform of that protein, PEX5L/1-264myc (c, d), does restore PTS2 import (c). White boxes represent the six TPR domains involved in PTS1 binding. The amino acids encoded by the alternative exon (AE) are indicated by the gray box.
The mammalian two-hybrid protein interaction assay makes fewer assumptions regarding the variables described above, and we therefore tested whether it could be used to map the region of PEX5L that interacts with PEX7. The mammalian two-hybrid system (74) operates on the same principle as the yeast two-hybrid system but offers certain advantages. These include the ability to more easily control for expression and localization of test proteins, adaptability to a wide range of cell lines, and the ability to assay reporter gene expression on a cell-by-cell basis as well as on an averaged basis for the entire cell population. The particular assay system we employed uses the cDNA for chloramphenicol acetyltransferase (CAT) as the reporter, which is under the control of a GAL4-responsive promoter. Test proteins are expressed as fusions to either the GAL4 DNA binding domain (BD) or the transcriptional transactivation domain of VP16 (AD). Cells are then co-transfected with all three plasmids, and the interaction is scored by the extent of CAT expression in lysates prepared from transfected cells, which can be determined by ELISA as well as by indirect immunofluorescence microscopy (IIF).
Isolated expression of BD-PEX7, AD-PEX5S, and AD-PEX5L fusion proteins had no effect on GAL4-mediated expression of CAT in this assay (Table II). Co-expression of BD-PEX7 and AD-PEX5S also failed to activate the CAT reporter gene, but simultaneous expression of BD-PEX7 and AD-PEX5L led to significant expression of CAT protein, as determined by ELISA of cell lysates (Table II) and by IIF microscopy (Fig. 5). The equal expression of PEX5S and PEX5L was controlled by IIF microscopy using antibodies against PEX5 or the activating domain. This experiment was also performed in PEX16-deficient PBD061 cells, which lack peroxisome membranes, to determine whether the interaction was dependent upon the presence of peroxisome membranes or a functional peroxisomal import apparatus (75,76). The PEX7⅐PEX5L interaction was even stronger in these peroxisome-deficient cells, as determined by ELISA (Table II) and IIF microscopy (data not shown).
We next assayed the interaction between PEX7 and various deletion mutants of PEX5S and PEX5L (Table II) 4. a, PEX5L can be co-immunoprecipitated with N-myc-PEX7 from transfected fibroblasts. PEX5-deficient cells were transfected with plasmids encoding PEX5S or PEX5L and myc-PEX7. After 2 days the cells were lysed in 0.5% Triton X-100. Anti-myc antibodies coated to Dynabeads were used to precipitate myc-PEX7. The precipitate was separated by SDS-PAGE, blotted to a PVDF membrane, and probed with anti-PEX5 antibodies. The supernatants were given as a control for expression of the different PEX5 proteins. The ratio between pellet and supernatant loaded onto the gel was 20 to 1. b, immunoprecipitation of in vitro transcribed and translated proteins. PEX5L, PEX5S, and mycPEX7 were separately transcribed and translated in vitro with [ 35 S]cysteine. Equal amounts of the same protein pairs as given in a were incubated together for one additional hour and then precipitated with anti-myccoated Dynabeads. Pellet and supernatant fractions were separated on 10% SDS gels. The ratio between pellet fraction and supernatant was 12.5 to 1. The gels were dried and subjected to fluorography. The positions of the labeled proteins are indicated on the right. No significant amounts of PEX5S or PEX5L could be detected in the precipitates. results suggest that amino acids 191-222 of PEX5L are sufficient for interaction with PEX7 and that the interaction isincreased with a slightly longer fragment of PEX5L (PEX5L/191-251).
Negative results in this assay are more difficult to interpret. Loss of interaction may reflect an interesting phenomenon, such as elimination or disruption of a protein⅐protein interaction motif, but negative results can also be caused by other factors, including reduced expression, aberrant folding, and altered subcellular distribution. In these experiments, we were able to control for expression and subcellular distribution of the various AD-PEX5L fusions, which were equivalent, but controls for folding were obviously beyond the limits of our experimental approach. Nevertheless, the lack of interaction between PEX7 and fragments of PEX5 that do not contain the 37-amino acid insert of PEX5L is consistent with the hypothesis that all or part of this insert plays an important role in PEX7 interaction. In addition, the lack of interaction between PEX7 and PEX5L/214 -335 raises the possibility that some or all of the 14 amino acids upstream of the 37-amino acid insert in PEX5L are also involved in the interaction with PEX7.
The hypothesis that the PEX7 interaction domain of PEX5L is important for PTS2 protein import predicts that expression of this domain alone should inhibit PTS2 protein import. To test this hypothesis we co-transfected normal human fibroblasts with the PTS2-CAT expression vector and with either pM or pM-PEX5L/191-251. Two days after transfection the cells were processed for IIF microscopy. The percentage of cells in each population that showed import of PTS2-CAT into peroxisomes was determined by assessing PTS-CAT distribution in hundreds of transfected cells. In cells co-transfected with pM, PTS2-CAT was peroxisomal in 85% of expressing cells whereas it was peroxisomal in only 22% of cells co-transfected with pM-PEX5L/191-251. Thus, overexpression of PEX5L/  191-251 in normal human fibroblasts inhibited PTS2 protein import by ϳ75%.
Could PEX5L Represent a Functional Homolog of Yeast Pex18p/Pex21p?-Like PEX5L in mammalian cells, S. cerevisiae Pex18p and Pex21p interact with S. cerevisiae Pex7 and are required for PTS2 protein import (77). We therefore tested whether mammalian PEX5 shared any other properties with yeast Pex18p and Pex21p. These yeast peroxins have an important role in targeting S. cerevisiae Pex7p to peroxisomes (77), and we therefore investigated whether PEX5 was required for PEX7 targeting in human cells. Although yeast Pex7p has been described as a predominantly cytoplasmic, partly peroxisomal protein (34,35,38), certain tagged versions of Pex7p will actually accumulate within peroxisomes (33). A fusion protein between enhanced green fluorescence protein (EGFP) and human PEX7 (EGFP-PEX7) displays this property, as shown by its peroxisomal distribution in normal human fibroblasts (Fig. 6, a and b) that was not detectable when the cells were differentially permeabilized with digitonin indicating an intraperoxisomal localization (Fig. 6, c and d). We expressed the same EGFP-PEX7 fusion protein in PEX5-deficient PBD005 cells and found that it remained completely in the cytoplasm (Fig. 6, e and f), even though these cells contain numerous peroxisomes (27,68,78).
The hypothesis that PEX5L mediates the transport of PEX7 to peroxisomes predicts that the minimal PTS2 rescue domain of PEX5L contains at least two functional elements: one that is sufficient for interaction with PEX7 and another that is sufficient for targeting to peroxisomes. It is not a trivial issue to assay the targeting of PEX5 proteins to peroxisomes, because PEX5 is a predominantly cytoplasmic, partly peroxisomal protein (27). For example ϳ95% of the total PEX5 protein is located in the cytoplasm of normal human fibroblasts. However, certain PBD cells with mutations in other PEX genes trap PEX5 on or in peroxisomes (40,79), and we took advantage of this phenotype to determine whether the minimal PTS2 import domain of PEX5L was sufficient for peroxisomal targeting. PBD054 cells are homozygous for a missense mutation (S320F) in PEX12 and accumulate PEX5 on or in the peroxisome (79). We transfected PBD054 cells with plasmids designed to express various myc-tagged PEX5 proteins. Two days after transfection the subcellular distribution of these proteins was assessed by immunofluorescence microscopy using anti-myc antibodies (Fig. 7). PEX5/1-214 was the smallest fragment that retained peroxisomal targeting, and this region is contained within the smallest fragment of PEX5L that is sufficient for PTS2 protein import (PEX5L/1-230myc). Smaller fragments of PEX5 (amino acids 1-157 and 1-90) failed to target to peroxisomes and were instead located only in the cytoplasm.
PEX5L shares three properties with the S. cerevisiae peroxins Pex18p and Pex21p. All three proteins interact with PEX7, play an important role in PTS2 protein import, and play an important role in directing PEX7 to peroxisomes. These common properties led us to examine PEX5L, ScPex18p, and ScPex21p for any shared sequence motifs, and we identified a 21-amino acid-long peptide motif in all three proteins (Fig. 8). This motif (amino acids 209 -229) is located within the minimal PTS2 rescue domain of PEX5L (amino acids 1-230), and a PEX5L fragment that lacked the C-terminal 7 amino acids of this element (PEX5L/1-222) displayed only a borderline ability to rescue PTS2 protein import. Most of this motif also matches FIG. 6. EGFP-PEX7 has a peroxisomal localization in wild-type fibroblasts but is not targeted to peroxisomes in PEX5-deficient PBD005 cells. EGFP-PEX7 was expressed in wildtype fibroblasts (a-d) or in PEX5-deficient PBD005 cells (e, f). Permeabilization was either done with Triton X -100 (a, b, e,  f) or with digitonin (c, d), which did not permeabilize the peroxisome membrane. a, EGFP-PEX7 shows a cytoplasmic and peroxisomal distribution in wild-type cells visualized with antibodies against EGFP. b, the same cells stained for peroxisomal catalase. c and d, digitonin permeabilization reveals a peroxisomal and cytoplasmic localization of EGFP-PEX7 when the intrinsic fluorescence of EGFP, which is independent of permeabilization, is monitored (c). No peroxisomal staining is noticeable when antibodies against EGFP were used (d). This indicates that most of the peroxisomal EGFP staining accounts for intraperoxisomal EGFP-PEX7 molecules. e and f, EGFP-PEX7 has an exclusively cytoplasmic localization in PEX5-deficient PBD005 cells, when stained with antibodies against EGFP (e). The identical cell stained with antibodies against catalase, indicating the PTS1 import defect of these cells that leads to a cytoplasmic localization of catalase (f).
the smallest fragment of PEX5L that retains interaction with PEX7 in the mammalian two-hybrid assay (amino acids 191-222), indicating that it may play an important role in PEX5L⅐PEX7 interaction. The fact that this conserved motif is 8 amino acids longer than the smallest PEX7-interacting fragment and the fact that a somewhat longer PEX5 fragment PEX5L/191-251 shows an even stronger interactions may indicate that indeed the region including amino acids 209 -229 of PEX5L is most important for the interaction with PEX7. This hypothesis is supported by a recent study from the Fujiki laboratory, which showed that substitution of phenylalanine for a conserved serine of this motif (Ser-213 of human PEX5) eliminates PTS2 protein import and disrupts the PEX5L⅐PEX7 interaction (80). The hypothesis that this 21-amino acid motif may be important for interaction with PEX7 is also supported by the fact that this motif is located within regions of Pex18p FIG. 7. The N-terminal 214 amino acids are sufficient to target PEX5 to peroxisomes. PBD054 cells have a defect in PEX12 and accumulate PEX5 at or in the peroxisomes. These cells were transfected with plasmids designed to express the various myc-tagged PEX5 proteins given in c. The PEX5 distribution was monitored by immunofluorescence microscopy with anti-myc antibodies and anti-PMP70 antibodies. The immunofluorescence microscopy revealed a peroxisomal accumulation of PEX5/1-214myc (a), indicated by its colocalization with the peroxisomal membrane protein PMP70 (b). White boxes represent the six TPR domains involved in PTS1 binding. The gray box indicates the amino acids encoded by the alternative exon (AE). and Pex21p that are sufficient for interaction with yeast Pex7p (77).
The possibility that PEX5L may represent a functional homolog of S. cerevisiae Pex18p and/or Pex21p predicts that PEX5L may interact with S. cerevisiae Pex7p and that S. cerevisiae Pex18p and/or Pex21p may interact with human PEX7. Using the yeast two-hybrid assay, we observed that human PEX5L did indeed interact with yeast Pex7p whereas PEX5S did not interact with yeast Pex7p (Fig. 9). In addition, the interaction of S. cerevisiae Pex7p was mapped to the same region of PEX5L, as with human PEX7. Previous studies have reported that yeast Pex7p interacts with yeast Pex5p in the yeast twohybrid assay but that this interaction is detected only in the presence of yeast Pex14p. Presumably, Pex14p serves as a bridge between Pex5p and Pex7p, which do not contact each other directly (50). In contrast, the interaction between human PEX5L and yeast Pex7p is not reduced in the absence of Pex14p and in fact may even be stronger in the pex14-deficient two-hybrid reporter strain. We also tested whether ScPex21p interacted with human PEX7 using the mammalian two-hybrid system. We again found evidence of interaction (Table III) indicating that the interacting regions of both partners may be conserved.
The fact that the role of PEX5 in PTS2 protein import differs greatly between mammalian cells and yeast led us to consider the possibility that additional differences may exist in PTS2 protein import. We already supposed that PEX5L could transport PEX7 and its cargo to the peroxisomal membrane. The two candidates for docking this complex to the peroxisomal membrane are PEX13 and PEX14. Data available so far suggest differences between the human and yeast system. Previous studies have established that S. cerevisiae Pex7p and S. cerevisiae Pex5p both independently interact with S. cerevisiae Pex13p and Pex14p, respectively. This has been shown both in vivo and in two-hybrid assays (50). The interaction between human PEX5 and PEX14 has now been shown by several groups and is easily detectable (48,52,66). Performing coimmunoprecipitations from lysates Fujiki and co-workers found evidence for PEX13-PEX5 (66) and PEX14-PEX7 (53,66) interactions; however, there is no evidence that these interactions are direct and the interactions could not be detected in other laboratories (48,52). Using the human two hybrid-system we could easily detect the PEX5⅐PEX14 interaction (52, and this study) and the PEX5L⅐PEX7 interaction, but we find no indication for a PEX13-PEX5 interaction, even when we used the SH3 domain of PEX13 (Table IV). In addition, we did not detect an interaction between PEX13⅐PEX7 that was suffi-  FIG. 9. Interactions of ScPex7p with human PEX5 can be mapped to the same region as with HsPEX7 in a yeast two-hybrid assay and are independent of ScPex14p. The yeast strains PCY2 and PCY2⌬pex14 were doubletransformed with plasmids encoding the indicated Gal4-BD (in pPC86) and Gal4AD fusions (in pPC97) and were investigated for ␤-galactosidase activation with a filter assay after selection on SD plates lacking leucine and tryptophan (see "Experimental Procedures"). Only the PEX5L-derived proteins were able to interact with ScPex7p. This interaction was independent of yeast Pex14p.

Domain Mapping of Human PEX5
ciently high above background (Table IV). The lack of a PEX14-PEX7 interaction has been reported previously for the human system (52). Together our data support the idea that PEX14 is probably the main docking partner not only for the PTS1 receptor complex (48,66) but also mediates PTS2 import via PEX5L and PEX7 in the human system. The lack of PEX13-PEX7 and PEX14-PEX7 interactions may explain why PEX7 was not localized to the peroxisomal membrane in PEX5 deficient cells (Fig. 7).

DISCUSSION
In their initial description of PEX5, McCollum et al. (25) reported that this peroxin was not required for PTS2 protein import in the yeast Pichia pastoris. The subsequent observation that PEX5 is essential for PTS2 protein import in human cells was therefore unexpected (27,29). However, studies showing that PEX5 is also essential for PTS2 protein import in hamster cells (65) and mice (63) have firmly established the central role of PEX5 in PTS2 protein import in mammalian cells. The different roles of PEX5 in PTS2 protein import in fungi and mammals are likely to reveal core properties of the PTS2 protein import pathway, and this topic has attracted considerable attention within the field. Mammalian PEX5 RNAs are alternatively spliced to generate two isoforms of PEX5, which differ by 37 amino acids. The longer isoform, PEX5L, is the form that participates in PTS2 protein import (64,65). Recent studies from Fujiki and colleagues (48,66) have furthered our understanding by establishing that PEX7 displays preferential binding to PEX5L as compared with PEX5S. In this report we used a combination of functional complementation assays, mammalian two-hybrid assays, and protein targeting assays to identify regions of PEX5L that are sufficient for PTS2 rescue, interaction with PEX7, and targeting to peroxisomes.
PEX5L is a protein 639-amino acids long. Previous studies have established that amino acids 299 -639 are involved in PTS1 recognition (27,31,79) and interaction with PEX12 (79) and that amino acids 1-298 are involved in interaction with PEX14 (70,81). Positive results in a variety of assays have allowed us to enrich the functional domain map of PEX5L (Fig.  10) by showing that amino acids 1-230 are sufficient for rescuing PTS2 protein import in PEX5-deficient human cells, amino acids 191-222 are sufficient for interaction with PEX7, and amino acids 1-214 are sufficient for peroxisomal targeting. These results support the view of PEX5 as a modular, multidomain protein that interacts with numerous proteins during the matrix protein import process. Other protein interaction sites that remain to be mapped in PEX5 include the binding sites for PEX10 (82), Pex8p (83), and perhaps other PEX5-binding proteins that remain to be described.
Prior studies showing that PEX5L, but not PEX5S, mediated PTS2 protein import and PEX7 interaction (64 -66) suggested that the 37-amino acid insert of PEX5L played an important role in these processes. However, several lines of evidence suggest that PEX7 interaction requires additional sequences N-terminal to the insert and may require only the N-terminal portion of the insert. We observed that the C-terminal 21 amino acids of the 37-amino acid insert could be removed without significantly reducing the ability of PEX5L to rescue PTS2 protein import. We also identified a peptide motif at amino acids 209 -229 that spans the N-terminal boundary of the insert, is shared by the PEX7 binding domains of yeast Pex18p and Pex21p proteins, and lies within the smallest PTS2 rescue and PEX7 interaction domains we identified in PEX5L. Matsumura et al. (80) identified a single amino acid substitution mutation in PEX5 that eliminated only PTS2 protein import and reduced the interaction of PEX5L with PEX7. This serine to phenylalanine missense mutation alters a serine of the conserved peptide motif that is located two amino acids N-terminal of the 37-amino acid insert, lending additional support to the hypothesis that the PEX7 interaction domain spans the Nterminal boundary of the insert in PEX5L. Finally, we observed that a fragment of PEX5L lacking sequences upstream of the 37-amino acid insert failed to interact with PEX7 in the mammalian two-hybrid assay. Although there is no control for the folding of this PEX5L fragment, the fragment was properly expressed and localized and the result is at least consistent with the hypothesis that PEX7 interaction requires amino acids upstream of the insert. Additionally, we emphasize that our data do not demonstrate direct interaction between PEX5L and PEX7. In fact, we would argue that the existing data in the literature also fails to demonstrate direct PEX5L⅐PEX7 binding (66). We also note that our negative results are largely uncontrolled, can be accounted for by experimental artifact, and should not be interpreted as evidence that the PEX5L⅐PEX7 interaction is indirect.
Like yeast Pex18p and Pex21p, PEX5L is required for PTS2 protein import, interacts with PEX7, and is necessary for PEX7 transport to peroxisomes. In addition to these functional similarities, the presence of a shared peptide motif in all three proteins shows that they also share some structural similarities. These common properties suggest that PEX5L may have acquired one or more of the roles that yeast Pex18p and Pex21p perform in PTS2 protein import. One possibility is that PEX5L may have replaced either Pex18p or Pex21p, but such a simple model ignores the fact that PEX5L is essential for PTS2 protein import, whereas PTS2 protein import in yeast is blocked only by the loss of both Pex18p and Pex21p. Therefore, a more likely model would have PEX5L replacing both Pex18p and Pex21p. Another attractive possibility is that the role of these PEX7associated proteins has changed through evolution in such a way that the Pex18p-and Pex21p-like activities would both be essential for PTS2 protein import and PEX7 interaction. The identification of human orthologs of Pex18p and Pex21p would

Domain Mapping of Human PEX5
help distinguish between these various possibilities, and, although we have been unable to identify Pex18p or Pex21p genes in humans, this failure hardly represents strong evidence that Pex18p and Pex21p are absent.
A further comparison of all other known PEX5 sequences revealed that this conserved motif and, therefore, the long form of PEX5 are only found in mammals; plants (18,84) (Arabidopsis thaliana: AF074843); and protozoa Leishmania donovani (85) and Trypanosoma brucei (86) but not in Caenorhabditis elegans (C34C6.6) (87); Drosophila melanogaster (AAF45676.1); or different yeast species. For all species that share the common PEX7 binding motif there is evidence for PEX7 as in A. thaliana (AAF88113), or at least evidence for proteins with a PTS2 targeting signal as in plants and T. brucei (88). Our results suggest that PTS2 import may be similar in mammals, plants, and protozoa but different in yeasts. As for C. elegans, which appears to lack PEX7 (89), PEX5 (87) lacks the PEX7 interaction motif. In addition, orthologs of mammalian PTS2 proteins in C. elegans lack a PTS2 signal and instead are imported via a PTS1, indicating that the PTS2 pathway may be absent in C. elegans (89).